diff options
Diffstat (limited to 'src/libsystemd/src/sd-journal')
26 files changed, 14151 insertions, 0 deletions
diff --git a/src/libsystemd/src/sd-journal/GNUmakefile b/src/libsystemd/src/sd-journal/GNUmakefile new file mode 120000 index 0000000000..13308a50cd --- /dev/null +++ b/src/libsystemd/src/sd-journal/GNUmakefile @@ -0,0 +1 @@ +../../../../GNUmakefile
\ No newline at end of file diff --git a/src/libsystemd/src/sd-journal/Makefile b/src/libsystemd/src/sd-journal/Makefile new file mode 100644 index 0000000000..d0415fb0d3 --- /dev/null +++ b/src/libsystemd/src/sd-journal/Makefile @@ -0,0 +1,42 @@ +# -*- Mode: makefile; indent-tabs-mode: t -*- +# +# This file is part of systemd. +# +# Copyright 2010-2012 Lennart Poettering +# Copyright 2010-2012 Kay Sievers +# Copyright 2013 Zbigniew Jędrzejewski-Szmek +# Copyright 2013 David Strauss +# Copyright 2016 Luke Shumaker +# +# systemd is free software; you can redistribute it and/or modify it +# under the terms of the GNU Lesser General Public License as published by +# the Free Software Foundation; either version 2.1 of the License, or +# (at your option) any later version. +# +# systemd is distributed in the hope that it will be useful, but +# WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +# Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public License +# along with systemd; If not, see <http://www.gnu.org/licenses/>. +include $(dir $(lastword $(MAKEFILE_LIST)))/../../../../config.mk +include $(topsrcdir)/build-aux/Makefile.head.mk + +audit_list_includes = -include linux/audit.h -include missing.h +ifneq ($(HAVE_AUDIT),) +audit_list_includes += -include libaudit.h +endif # HAVE_AUDIT + +$(outdir)/audit_type-list.txt: + $(AM_V_GEN)$(CPP) $(sd.ALL_CPPFLAGS) -dM $(audit_list_includes) - </dev/null | grep -vE 'AUDIT_.*(FIRST|LAST)_' | $(SED) -r -n 's/^#define\s+AUDIT_(\w+)\s+([0-9]{4})\s*$$/\1\t\2/p' | sort -k2 >$@ + +$(outdir)/audit_type-to-name.h: $(outdir)/audit_type-list.txt + $(AM_V_GEN)$(AWK) 'BEGIN{ print "const char *audit_type_to_string(int type) {\n\tswitch(type) {" } {printf " case AUDIT_%s: return \"%s\";\n", $$1, $$1 } END{ print " default: return NULL;\n\t}\n}\n" }' <$< >$@ + +# fsprg.c is a drop-in file using void pointer arithmetic +libsystemd_journal_internal_la_CFLAGS += \ + $(GCRYPT_CFLAGS) \ + -Wno-pointer-arith + +include $(topsrcdir)/build-aux/Makefile.tail.mk diff --git a/src/libsystemd/src/sd-journal/audit-type.c b/src/libsystemd/src/sd-journal/audit-type.c new file mode 100644 index 0000000000..1644e3f1b1 --- /dev/null +++ b/src/libsystemd/src/sd-journal/audit-type.c @@ -0,0 +1,31 @@ +/*** + This file is part of systemd. + + Copyright 2015 Zbigniew Jędrzejewski-Szmek + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdio.h> + +#include <linux/audit.h> +#ifdef HAVE_AUDIT +# include <libaudit.h> +#endif + +#include "systemd-basic/macro.h" +#include "systemd-basic/missing.h" + +#include "audit-type.h" +#include "audit_type-to-name.h" diff --git a/src/libsystemd/src/sd-journal/audit-type.h b/src/libsystemd/src/sd-journal/audit-type.h new file mode 100644 index 0000000000..0b4748b01f --- /dev/null +++ b/src/libsystemd/src/sd-journal/audit-type.h @@ -0,0 +1,37 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2015 Zbigniew Jędrzejewski-Szmek + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "systemd-basic/macro.h" + +const char *audit_type_to_string(int type); +int audit_type_from_string(const char *s); + +/* This is inspired by DNS TYPEnnn formatting */ +#define audit_type_name_alloca(type) \ + ({ \ + const char *_s_; \ + _s_ = audit_type_to_string(type); \ + if (!_s_) { \ + _s_ = alloca(strlen("AUDIT") + DECIMAL_STR_MAX(int)); \ + sprintf((char*) _s_, "AUDIT%04i", type); \ + } \ + _s_; \ + }) diff --git a/src/libsystemd/src/sd-journal/catalog.c b/src/libsystemd/src/sd-journal/catalog.c new file mode 100644 index 0000000000..fce4649f6f --- /dev/null +++ b/src/libsystemd/src/sd-journal/catalog.c @@ -0,0 +1,768 @@ +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <fcntl.h> +#include <locale.h> +#include <stdio.h> +#include <string.h> +#include <sys/mman.h> +#include <unistd.h> + +#include <systemd/sd-id128.h> + +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/conf-files.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/hashmap.h" +#include "systemd-basic/log.h" +#include "systemd-basic/mkdir.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/siphash24.h" +#include "systemd-basic/sparse-endian.h" +#include "systemd-basic/strbuf.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/util.h" + +#include "catalog.h" + +const char * const catalog_file_dirs[] = { + "/usr/local/lib/systemd/catalog/", + "/usr/lib/systemd/catalog/", + NULL +}; + +#define CATALOG_SIGNATURE (uint8_t[]) { 'R', 'H', 'H', 'H', 'K', 'S', 'L', 'P' } + +typedef struct CatalogHeader { + uint8_t signature[8]; /* "RHHHKSLP" */ + le32_t compatible_flags; + le32_t incompatible_flags; + le64_t header_size; + le64_t n_items; + le64_t catalog_item_size; +} CatalogHeader; + +typedef struct CatalogItem { + sd_id128_t id; + char language[32]; + le64_t offset; +} CatalogItem; + +static void catalog_hash_func(const void *p, struct siphash *state) { + const CatalogItem *i = p; + + siphash24_compress(&i->id, sizeof(i->id), state); + siphash24_compress(i->language, strlen(i->language), state); +} + +static int catalog_compare_func(const void *a, const void *b) { + const CatalogItem *i = a, *j = b; + unsigned k; + + for (k = 0; k < ELEMENTSOF(j->id.bytes); k++) { + if (i->id.bytes[k] < j->id.bytes[k]) + return -1; + if (i->id.bytes[k] > j->id.bytes[k]) + return 1; + } + + return strcmp(i->language, j->language); +} + +const struct hash_ops catalog_hash_ops = { + .hash = catalog_hash_func, + .compare = catalog_compare_func +}; + +static bool next_header(const char **s) { + const char *e; + + e = strchr(*s, '\n'); + + /* Unexpected end */ + if (!e) + return false; + + /* End of headers */ + if (e == *s) + return false; + + *s = e + 1; + return true; +} + +static const char *skip_header(const char *s) { + while (next_header(&s)) + ; + return s; +} + +static char *combine_entries(const char *one, const char *two) { + const char *b1, *b2; + size_t l1, l2, n; + char *dest, *p; + + /* Find split point of headers to body */ + b1 = skip_header(one); + b2 = skip_header(two); + + l1 = strlen(one); + l2 = strlen(two); + dest = new(char, l1 + l2 + 1); + if (!dest) { + log_oom(); + return NULL; + } + + p = dest; + + /* Headers from @one */ + n = b1 - one; + p = mempcpy(p, one, n); + + /* Headers from @two, these will only be found if not present above */ + n = b2 - two; + p = mempcpy(p, two, n); + + /* Body from @one */ + n = l1 - (b1 - one); + if (n > 0) { + memcpy(p, b1, n); + p += n; + + /* Body from @two */ + } else { + n = l2 - (b2 - two); + memcpy(p, b2, n); + p += n; + } + + assert(p - dest <= (ptrdiff_t)(l1 + l2)); + p[0] = '\0'; + return dest; +} + +static int finish_item( + Hashmap *h, + sd_id128_t id, + const char *language, + char *payload, size_t payload_size) { + + _cleanup_free_ CatalogItem *i = NULL; + _cleanup_free_ char *prev = NULL, *combined = NULL; + + assert(h); + assert(payload); + assert(payload_size > 0); + + i = new0(CatalogItem, 1); + if (!i) + return log_oom(); + + i->id = id; + if (language) { + assert(strlen(language) > 1 && strlen(language) < 32); + strcpy(i->language, language); + } + + prev = hashmap_get(h, i); + if (prev) { + /* Already have such an item, combine them */ + combined = combine_entries(payload, prev); + if (!combined) + return log_oom(); + + if (hashmap_update(h, i, combined) < 0) + return log_oom(); + combined = NULL; + } else { + /* A new item */ + combined = memdup(payload, payload_size + 1); + if (!combined) + return log_oom(); + + if (hashmap_put(h, i, combined) < 0) + return log_oom(); + i = NULL; + combined = NULL; + } + + return 0; +} + +int catalog_file_lang(const char* filename, char **lang) { + char *beg, *end, *_lang; + + end = endswith(filename, ".catalog"); + if (!end) + return 0; + + beg = end - 1; + while (beg > filename && *beg != '.' && *beg != '/' && end - beg < 32) + beg--; + + if (*beg != '.' || end <= beg + 1) + return 0; + + _lang = strndup(beg + 1, end - beg - 1); + if (!_lang) + return -ENOMEM; + + *lang = _lang; + return 1; +} + +static int catalog_entry_lang(const char* filename, int line, + const char* t, const char* deflang, char **lang) { + size_t c; + + c = strlen(t); + if (c == 0) { + log_error("[%s:%u] Language too short.", filename, line); + return -EINVAL; + } + if (c > 31) { + log_error("[%s:%u] language too long.", filename, line); + return -EINVAL; + } + + if (deflang) { + if (streq(t, deflang)) { + log_warning("[%s:%u] language specified unnecessarily", + filename, line); + return 0; + } else + log_warning("[%s:%u] language differs from default for file", + filename, line); + } + + *lang = strdup(t); + if (!*lang) + return -ENOMEM; + + return 0; +} + +int catalog_import_file(Hashmap *h, const char *path) { + _cleanup_fclose_ FILE *f = NULL; + _cleanup_free_ char *payload = NULL; + size_t payload_size = 0, payload_allocated = 0; + unsigned n = 0; + sd_id128_t id; + _cleanup_free_ char *deflang = NULL, *lang = NULL; + bool got_id = false, empty_line = true; + int r; + + assert(h); + assert(path); + + f = fopen(path, "re"); + if (!f) + return log_error_errno(errno, "Failed to open file %s: %m", path); + + r = catalog_file_lang(path, &deflang); + if (r < 0) + log_error_errno(r, "Failed to determine language for file %s: %m", path); + if (r == 1) + log_debug("File %s has language %s.", path, deflang); + + for (;;) { + char line[LINE_MAX]; + size_t line_len; + + if (!fgets(line, sizeof(line), f)) { + if (feof(f)) + break; + + return log_error_errno(errno, "Failed to read file %s: %m", path); + } + + n++; + + truncate_nl(line); + + if (line[0] == 0) { + empty_line = true; + continue; + } + + if (strchr(COMMENTS "\n", line[0])) + continue; + + if (empty_line && + strlen(line) >= 2+1+32 && + line[0] == '-' && + line[1] == '-' && + line[2] == ' ' && + (line[2+1+32] == ' ' || line[2+1+32] == '\0')) { + + bool with_language; + sd_id128_t jd; + + /* New entry */ + + with_language = line[2+1+32] != '\0'; + line[2+1+32] = '\0'; + + if (sd_id128_from_string(line + 2 + 1, &jd) >= 0) { + + if (got_id) { + if (payload_size == 0) { + log_error("[%s:%u] No payload text.", path, n); + return -EINVAL; + } + + r = finish_item(h, id, lang ?: deflang, payload, payload_size); + if (r < 0) + return r; + + lang = mfree(lang); + payload_size = 0; + } + + if (with_language) { + char *t; + + t = strstrip(line + 2 + 1 + 32 + 1); + r = catalog_entry_lang(path, n, t, deflang, &lang); + if (r < 0) + return r; + } + + got_id = true; + empty_line = false; + id = jd; + + continue; + } + } + + /* Payload */ + if (!got_id) { + log_error("[%s:%u] Got payload before ID.", path, n); + return -EINVAL; + } + + line_len = strlen(line); + if (!GREEDY_REALLOC(payload, payload_allocated, + payload_size + (empty_line ? 1 : 0) + line_len + 1 + 1)) + return log_oom(); + + if (empty_line) + payload[payload_size++] = '\n'; + memcpy(payload + payload_size, line, line_len); + payload_size += line_len; + payload[payload_size++] = '\n'; + payload[payload_size] = '\0'; + + empty_line = false; + } + + if (got_id) { + if (payload_size == 0) { + log_error("[%s:%u] No payload text.", path, n); + return -EINVAL; + } + + r = finish_item(h, id, lang ?: deflang, payload, payload_size); + if (r < 0) + return r; + } + + return 0; +} + +static int64_t write_catalog(const char *database, struct strbuf *sb, + CatalogItem *items, size_t n) { + CatalogHeader header; + _cleanup_fclose_ FILE *w = NULL; + int r; + _cleanup_free_ char *d, *p = NULL; + size_t k; + + d = dirname_malloc(database); + if (!d) + return log_oom(); + + r = mkdir_p(d, 0775); + if (r < 0) + return log_error_errno(r, "Recursive mkdir %s: %m", d); + + r = fopen_temporary(database, &w, &p); + if (r < 0) + return log_error_errno(r, "Failed to open database for writing: %s: %m", + database); + + zero(header); + memcpy(header.signature, CATALOG_SIGNATURE, sizeof(header.signature)); + header.header_size = htole64(ALIGN_TO(sizeof(CatalogHeader), 8)); + header.catalog_item_size = htole64(sizeof(CatalogItem)); + header.n_items = htole64(n); + + r = -EIO; + + k = fwrite(&header, 1, sizeof(header), w); + if (k != sizeof(header)) { + log_error("%s: failed to write header.", p); + goto error; + } + + k = fwrite(items, 1, n * sizeof(CatalogItem), w); + if (k != n * sizeof(CatalogItem)) { + log_error("%s: failed to write database.", p); + goto error; + } + + k = fwrite(sb->buf, 1, sb->len, w); + if (k != sb->len) { + log_error("%s: failed to write strings.", p); + goto error; + } + + r = fflush_and_check(w); + if (r < 0) { + log_error_errno(r, "%s: failed to write database: %m", p); + goto error; + } + + fchmod(fileno(w), 0644); + + if (rename(p, database) < 0) { + r = log_error_errno(errno, "rename (%s -> %s) failed: %m", p, database); + goto error; + } + + return ftello(w); + +error: + (void) unlink(p); + return r; +} + +int catalog_update(const char* database, const char* root, const char* const* dirs) { + _cleanup_strv_free_ char **files = NULL; + char **f; + struct strbuf *sb = NULL; + _cleanup_hashmap_free_free_free_ Hashmap *h = NULL; + _cleanup_free_ CatalogItem *items = NULL; + ssize_t offset; + char *payload; + CatalogItem *i; + Iterator j; + unsigned n; + int r; + int64_t sz; + + h = hashmap_new(&catalog_hash_ops); + sb = strbuf_new(); + + if (!h || !sb) { + r = log_oom(); + goto finish; + } + + r = conf_files_list_strv(&files, ".catalog", root, dirs); + if (r < 0) { + log_error_errno(r, "Failed to get catalog files: %m"); + goto finish; + } + + STRV_FOREACH(f, files) { + log_debug("Reading file '%s'", *f); + r = catalog_import_file(h, *f); + if (r < 0) { + log_error_errno(r, "Failed to import file '%s': %m", *f); + goto finish; + } + } + + if (hashmap_size(h) <= 0) { + log_info("No items in catalog."); + goto finish; + } else + log_debug("Found %u items in catalog.", hashmap_size(h)); + + items = new(CatalogItem, hashmap_size(h)); + if (!items) { + r = log_oom(); + goto finish; + } + + n = 0; + HASHMAP_FOREACH_KEY(payload, i, h, j) { + log_debug("Found " SD_ID128_FORMAT_STR ", language %s", + SD_ID128_FORMAT_VAL(i->id), + isempty(i->language) ? "C" : i->language); + + offset = strbuf_add_string(sb, payload, strlen(payload)); + if (offset < 0) { + r = log_oom(); + goto finish; + } + i->offset = htole64((uint64_t) offset); + items[n++] = *i; + } + + assert(n == hashmap_size(h)); + qsort_safe(items, n, sizeof(CatalogItem), catalog_compare_func); + + strbuf_complete(sb); + + sz = write_catalog(database, sb, items, n); + if (sz < 0) + r = log_error_errno(sz, "Failed to write %s: %m", database); + else { + r = 0; + log_debug("%s: wrote %u items, with %zu bytes of strings, %"PRIi64" total size.", + database, n, sb->len, sz); + } + +finish: + strbuf_cleanup(sb); + + return r; +} + +static int open_mmap(const char *database, int *_fd, struct stat *_st, void **_p) { + const CatalogHeader *h; + int fd; + void *p; + struct stat st; + + assert(_fd); + assert(_st); + assert(_p); + + fd = open(database, O_RDONLY|O_CLOEXEC); + if (fd < 0) + return -errno; + + if (fstat(fd, &st) < 0) { + safe_close(fd); + return -errno; + } + + if (st.st_size < (off_t) sizeof(CatalogHeader)) { + safe_close(fd); + return -EINVAL; + } + + p = mmap(NULL, PAGE_ALIGN(st.st_size), PROT_READ, MAP_SHARED, fd, 0); + if (p == MAP_FAILED) { + safe_close(fd); + return -errno; + } + + h = p; + if (memcmp(h->signature, CATALOG_SIGNATURE, sizeof(h->signature)) != 0 || + le64toh(h->header_size) < sizeof(CatalogHeader) || + le64toh(h->catalog_item_size) < sizeof(CatalogItem) || + h->incompatible_flags != 0 || + le64toh(h->n_items) <= 0 || + st.st_size < (off_t) (le64toh(h->header_size) + le64toh(h->catalog_item_size) * le64toh(h->n_items))) { + safe_close(fd); + munmap(p, st.st_size); + return -EBADMSG; + } + + *_fd = fd; + *_st = st; + *_p = p; + + return 0; +} + +static const char *find_id(void *p, sd_id128_t id) { + CatalogItem key, *f = NULL; + const CatalogHeader *h = p; + const char *loc; + + zero(key); + key.id = id; + + loc = setlocale(LC_MESSAGES, NULL); + if (loc && loc[0] && !streq(loc, "C") && !streq(loc, "POSIX")) { + strncpy(key.language, loc, sizeof(key.language)); + key.language[strcspn(key.language, ".@")] = 0; + + f = bsearch(&key, (const uint8_t*) p + le64toh(h->header_size), le64toh(h->n_items), le64toh(h->catalog_item_size), catalog_compare_func); + if (!f) { + char *e; + + e = strchr(key.language, '_'); + if (e) { + *e = 0; + f = bsearch(&key, (const uint8_t*) p + le64toh(h->header_size), le64toh(h->n_items), le64toh(h->catalog_item_size), catalog_compare_func); + } + } + } + + if (!f) { + zero(key.language); + f = bsearch(&key, (const uint8_t*) p + le64toh(h->header_size), le64toh(h->n_items), le64toh(h->catalog_item_size), catalog_compare_func); + } + + if (!f) + return NULL; + + return (const char*) p + + le64toh(h->header_size) + + le64toh(h->n_items) * le64toh(h->catalog_item_size) + + le64toh(f->offset); +} + +int catalog_get(const char* database, sd_id128_t id, char **_text) { + _cleanup_close_ int fd = -1; + void *p = NULL; + struct stat st = {}; + char *text = NULL; + int r; + const char *s; + + assert(_text); + + r = open_mmap(database, &fd, &st, &p); + if (r < 0) + return r; + + s = find_id(p, id); + if (!s) { + r = -ENOENT; + goto finish; + } + + text = strdup(s); + if (!text) { + r = -ENOMEM; + goto finish; + } + + *_text = text; + r = 0; + +finish: + if (p) + munmap(p, st.st_size); + + return r; +} + +static char *find_header(const char *s, const char *header) { + + for (;;) { + const char *v; + + v = startswith(s, header); + if (v) { + v += strspn(v, WHITESPACE); + return strndup(v, strcspn(v, NEWLINE)); + } + + if (!next_header(&s)) + return NULL; + } +} + +static void dump_catalog_entry(FILE *f, sd_id128_t id, const char *s, bool oneline) { + if (oneline) { + _cleanup_free_ char *subject = NULL, *defined_by = NULL; + + subject = find_header(s, "Subject:"); + defined_by = find_header(s, "Defined-By:"); + + fprintf(f, SD_ID128_FORMAT_STR " %s: %s\n", + SD_ID128_FORMAT_VAL(id), + strna(defined_by), strna(subject)); + } else + fprintf(f, "-- " SD_ID128_FORMAT_STR "\n%s\n", + SD_ID128_FORMAT_VAL(id), s); +} + + +int catalog_list(FILE *f, const char *database, bool oneline) { + _cleanup_close_ int fd = -1; + void *p = NULL; + struct stat st; + const CatalogHeader *h; + const CatalogItem *items; + int r; + unsigned n; + sd_id128_t last_id; + bool last_id_set = false; + + r = open_mmap(database, &fd, &st, &p); + if (r < 0) + return r; + + h = p; + items = (const CatalogItem*) ((const uint8_t*) p + le64toh(h->header_size)); + + for (n = 0; n < le64toh(h->n_items); n++) { + const char *s; + + if (last_id_set && sd_id128_equal(last_id, items[n].id)) + continue; + + assert_se(s = find_id(p, items[n].id)); + + dump_catalog_entry(f, items[n].id, s, oneline); + + last_id_set = true; + last_id = items[n].id; + } + + munmap(p, st.st_size); + + return 0; +} + +int catalog_list_items(FILE *f, const char *database, bool oneline, char **items) { + char **item; + int r = 0; + + STRV_FOREACH(item, items) { + sd_id128_t id; + int k; + _cleanup_free_ char *msg = NULL; + + k = sd_id128_from_string(*item, &id); + if (k < 0) { + log_error_errno(k, "Failed to parse id128 '%s': %m", *item); + if (r == 0) + r = k; + continue; + } + + k = catalog_get(database, id, &msg); + if (k < 0) { + log_full_errno(k == -ENOENT ? LOG_NOTICE : LOG_ERR, k, + "Failed to retrieve catalog entry for '%s': %m", *item); + if (r == 0) + r = k; + continue; + } + + dump_catalog_entry(f, id, msg, oneline); + } + + return r; +} diff --git a/src/libsystemd/src/sd-journal/catalog.h b/src/libsystemd/src/sd-journal/catalog.h new file mode 100644 index 0000000000..49fec555a4 --- /dev/null +++ b/src/libsystemd/src/sd-journal/catalog.h @@ -0,0 +1,36 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +#include <systemd/sd-id128.h> + +#include "systemd-basic/hashmap.h" +#include "systemd-basic/strbuf.h" + +int catalog_import_file(Hashmap *h, const char *path); +int catalog_update(const char* database, const char* root, const char* const* dirs); +int catalog_get(const char* database, sd_id128_t id, char **data); +int catalog_list(FILE *f, const char* database, bool oneline); +int catalog_list_items(FILE *f, const char* database, bool oneline, char **items); +int catalog_file_lang(const char *filename, char **lang); +extern const char * const catalog_file_dirs[]; +extern const struct hash_ops catalog_hash_ops; diff --git a/src/libsystemd/src/sd-journal/compress.c b/src/libsystemd/src/sd-journal/compress.c new file mode 100644 index 0000000000..c6c5f3c500 --- /dev/null +++ b/src/libsystemd/src/sd-journal/compress.c @@ -0,0 +1,684 @@ +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <inttypes.h> +#include <stdlib.h> +#include <string.h> +#include <sys/mman.h> +#include <unistd.h> + +#ifdef HAVE_XZ +#include <lzma.h> +#endif + +#ifdef HAVE_LZ4 +#include <lz4.h> +#include <lz4frame.h> +#endif + +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/io-util.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/sparse-endian.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/util.h" + +#include "compress.h" +#include "journal-def.h" + +#ifdef HAVE_LZ4 +DEFINE_TRIVIAL_CLEANUP_FUNC(LZ4F_compressionContext_t, LZ4F_freeCompressionContext); +DEFINE_TRIVIAL_CLEANUP_FUNC(LZ4F_decompressionContext_t, LZ4F_freeDecompressionContext); +#endif + +#define ALIGN_8(l) ALIGN_TO(l, sizeof(size_t)) + +static const char* const object_compressed_table[_OBJECT_COMPRESSED_MAX] = { + [OBJECT_COMPRESSED_XZ] = "XZ", + [OBJECT_COMPRESSED_LZ4] = "LZ4", +}; + +DEFINE_STRING_TABLE_LOOKUP(object_compressed, int); + +int compress_blob_xz(const void *src, uint64_t src_size, + void *dst, size_t dst_alloc_size, size_t *dst_size) { +#ifdef HAVE_XZ + static const lzma_options_lzma opt = { + 1u << 20u, NULL, 0, LZMA_LC_DEFAULT, LZMA_LP_DEFAULT, + LZMA_PB_DEFAULT, LZMA_MODE_FAST, 128, LZMA_MF_HC3, 4 + }; + static const lzma_filter filters[] = { + { LZMA_FILTER_LZMA2, (lzma_options_lzma*) &opt }, + { LZMA_VLI_UNKNOWN, NULL } + }; + lzma_ret ret; + size_t out_pos = 0; + + assert(src); + assert(src_size > 0); + assert(dst); + assert(dst_alloc_size > 0); + assert(dst_size); + + /* Returns < 0 if we couldn't compress the data or the + * compressed result is longer than the original */ + + if (src_size < 80) + return -ENOBUFS; + + ret = lzma_stream_buffer_encode((lzma_filter*) filters, LZMA_CHECK_NONE, NULL, + src, src_size, dst, &out_pos, dst_alloc_size); + if (ret != LZMA_OK) + return -ENOBUFS; + + *dst_size = out_pos; + return 0; +#else + return -EPROTONOSUPPORT; +#endif +} + +int compress_blob_lz4(const void *src, uint64_t src_size, + void *dst, size_t dst_alloc_size, size_t *dst_size) { +#ifdef HAVE_LZ4 + int r; + + assert(src); + assert(src_size > 0); + assert(dst); + assert(dst_alloc_size > 0); + assert(dst_size); + + /* Returns < 0 if we couldn't compress the data or the + * compressed result is longer than the original */ + + if (src_size < 9) + return -ENOBUFS; + + r = LZ4_compress_limitedOutput(src, (char*)dst + 8, src_size, (int) dst_alloc_size - 8); + if (r <= 0) + return -ENOBUFS; + + *(le64_t*) dst = htole64(src_size); + *dst_size = r + 8; + + return 0; +#else + return -EPROTONOSUPPORT; +#endif +} + + +int decompress_blob_xz(const void *src, uint64_t src_size, + void **dst, size_t *dst_alloc_size, size_t* dst_size, size_t dst_max) { + +#ifdef HAVE_XZ + _cleanup_(lzma_end) lzma_stream s = LZMA_STREAM_INIT; + lzma_ret ret; + size_t space; + + assert(src); + assert(src_size > 0); + assert(dst); + assert(dst_alloc_size); + assert(dst_size); + assert(*dst_alloc_size == 0 || *dst); + + ret = lzma_stream_decoder(&s, UINT64_MAX, 0); + if (ret != LZMA_OK) + return -ENOMEM; + + space = MIN(src_size * 2, dst_max ?: (size_t) -1); + if (!greedy_realloc(dst, dst_alloc_size, space, 1)) + return -ENOMEM; + + s.next_in = src; + s.avail_in = src_size; + + s.next_out = *dst; + s.avail_out = space; + + for (;;) { + size_t used; + + ret = lzma_code(&s, LZMA_FINISH); + + if (ret == LZMA_STREAM_END) + break; + else if (ret != LZMA_OK) + return -ENOMEM; + + if (dst_max > 0 && (space - s.avail_out) >= dst_max) + break; + else if (dst_max > 0 && space == dst_max) + return -ENOBUFS; + + used = space - s.avail_out; + space = MIN(2 * space, dst_max ?: (size_t) -1); + if (!greedy_realloc(dst, dst_alloc_size, space, 1)) + return -ENOMEM; + + s.avail_out = space - used; + s.next_out = *(uint8_t**)dst + used; + } + + *dst_size = space - s.avail_out; + return 0; +#else + return -EPROTONOSUPPORT; +#endif +} + +int decompress_blob_lz4(const void *src, uint64_t src_size, + void **dst, size_t *dst_alloc_size, size_t* dst_size, size_t dst_max) { + +#ifdef HAVE_LZ4 + char* out; + int r, size; /* LZ4 uses int for size */ + + assert(src); + assert(src_size > 0); + assert(dst); + assert(dst_alloc_size); + assert(dst_size); + assert(*dst_alloc_size == 0 || *dst); + + if (src_size <= 8) + return -EBADMSG; + + size = le64toh( *(le64_t*)src ); + if (size < 0 || (unsigned) size != le64toh(*(le64_t*)src)) + return -EFBIG; + if ((size_t) size > *dst_alloc_size) { + out = realloc(*dst, size); + if (!out) + return -ENOMEM; + *dst = out; + *dst_alloc_size = size; + } else + out = *dst; + + r = LZ4_decompress_safe((char*)src + 8, out, src_size - 8, size); + if (r < 0 || r != size) + return -EBADMSG; + + *dst_size = size; + return 0; +#else + return -EPROTONOSUPPORT; +#endif +} + +int decompress_blob(int compression, + const void *src, uint64_t src_size, + void **dst, size_t *dst_alloc_size, size_t* dst_size, size_t dst_max) { + if (compression == OBJECT_COMPRESSED_XZ) + return decompress_blob_xz(src, src_size, + dst, dst_alloc_size, dst_size, dst_max); + else if (compression == OBJECT_COMPRESSED_LZ4) + return decompress_blob_lz4(src, src_size, + dst, dst_alloc_size, dst_size, dst_max); + else + return -EBADMSG; +} + + +int decompress_startswith_xz(const void *src, uint64_t src_size, + void **buffer, size_t *buffer_size, + const void *prefix, size_t prefix_len, + uint8_t extra) { + +#ifdef HAVE_XZ + _cleanup_(lzma_end) lzma_stream s = LZMA_STREAM_INIT; + lzma_ret ret; + + /* Checks whether the decompressed blob starts with the + * mentioned prefix. The byte extra needs to follow the + * prefix */ + + assert(src); + assert(src_size > 0); + assert(buffer); + assert(buffer_size); + assert(prefix); + assert(*buffer_size == 0 || *buffer); + + ret = lzma_stream_decoder(&s, UINT64_MAX, 0); + if (ret != LZMA_OK) + return -EBADMSG; + + if (!(greedy_realloc(buffer, buffer_size, ALIGN_8(prefix_len + 1), 1))) + return -ENOMEM; + + s.next_in = src; + s.avail_in = src_size; + + s.next_out = *buffer; + s.avail_out = *buffer_size; + + for (;;) { + ret = lzma_code(&s, LZMA_FINISH); + + if (ret != LZMA_STREAM_END && ret != LZMA_OK) + return -EBADMSG; + + if (*buffer_size - s.avail_out >= prefix_len + 1) + return memcmp(*buffer, prefix, prefix_len) == 0 && + ((const uint8_t*) *buffer)[prefix_len] == extra; + + if (ret == LZMA_STREAM_END) + return 0; + + s.avail_out += *buffer_size; + + if (!(greedy_realloc(buffer, buffer_size, *buffer_size * 2, 1))) + return -ENOMEM; + + s.next_out = *(uint8_t**)buffer + *buffer_size - s.avail_out; + } + +#else + return -EPROTONOSUPPORT; +#endif +} + +int decompress_startswith_lz4(const void *src, uint64_t src_size, + void **buffer, size_t *buffer_size, + const void *prefix, size_t prefix_len, + uint8_t extra) { +#ifdef HAVE_LZ4 + /* Checks whether the decompressed blob starts with the + * mentioned prefix. The byte extra needs to follow the + * prefix */ + + int r; + size_t size; + + assert(src); + assert(src_size > 0); + assert(buffer); + assert(buffer_size); + assert(prefix); + assert(*buffer_size == 0 || *buffer); + + if (src_size <= 8) + return -EBADMSG; + + if (!(greedy_realloc(buffer, buffer_size, ALIGN_8(prefix_len + 1), 1))) + return -ENOMEM; + + r = LZ4_decompress_safe_partial((char*)src + 8, *buffer, src_size - 8, + prefix_len + 1, *buffer_size); + if (r >= 0) + size = (unsigned) r; + else { + /* lz4 always tries to decode full "sequence", so in + * pathological cases might need to decompress the + * full field. */ + r = decompress_blob_lz4(src, src_size, buffer, buffer_size, &size, 0); + if (r < 0) + return r; + } + + if (size >= prefix_len + 1) + return memcmp(*buffer, prefix, prefix_len) == 0 && + ((const uint8_t*) *buffer)[prefix_len] == extra; + else + return 0; + +#else + return -EPROTONOSUPPORT; +#endif +} + +int decompress_startswith(int compression, + const void *src, uint64_t src_size, + void **buffer, size_t *buffer_size, + const void *prefix, size_t prefix_len, + uint8_t extra) { + if (compression == OBJECT_COMPRESSED_XZ) + return decompress_startswith_xz(src, src_size, + buffer, buffer_size, + prefix, prefix_len, + extra); + else if (compression == OBJECT_COMPRESSED_LZ4) + return decompress_startswith_lz4(src, src_size, + buffer, buffer_size, + prefix, prefix_len, + extra); + else + return -EBADMSG; +} + +int compress_stream_xz(int fdf, int fdt, uint64_t max_bytes) { +#ifdef HAVE_XZ + _cleanup_(lzma_end) lzma_stream s = LZMA_STREAM_INIT; + lzma_ret ret; + uint8_t buf[BUFSIZ], out[BUFSIZ]; + lzma_action action = LZMA_RUN; + + assert(fdf >= 0); + assert(fdt >= 0); + + ret = lzma_easy_encoder(&s, LZMA_PRESET_DEFAULT, LZMA_CHECK_CRC64); + if (ret != LZMA_OK) { + log_error("Failed to initialize XZ encoder: code %u", ret); + return -EINVAL; + } + + for (;;) { + if (s.avail_in == 0 && action == LZMA_RUN) { + size_t m = sizeof(buf); + ssize_t n; + + if (max_bytes != (uint64_t) -1 && (uint64_t) m > max_bytes) + m = (size_t) max_bytes; + + n = read(fdf, buf, m); + if (n < 0) + return -errno; + if (n == 0) + action = LZMA_FINISH; + else { + s.next_in = buf; + s.avail_in = n; + + if (max_bytes != (uint64_t) -1) { + assert(max_bytes >= (uint64_t) n); + max_bytes -= n; + } + } + } + + if (s.avail_out == 0) { + s.next_out = out; + s.avail_out = sizeof(out); + } + + ret = lzma_code(&s, action); + if (ret != LZMA_OK && ret != LZMA_STREAM_END) { + log_error("Compression failed: code %u", ret); + return -EBADMSG; + } + + if (s.avail_out == 0 || ret == LZMA_STREAM_END) { + ssize_t n, k; + + n = sizeof(out) - s.avail_out; + + k = loop_write(fdt, out, n, false); + if (k < 0) + return k; + + if (ret == LZMA_STREAM_END) { + log_debug("XZ compression finished (%"PRIu64" -> %"PRIu64" bytes, %.1f%%)", + s.total_in, s.total_out, + (double) s.total_out / s.total_in * 100); + + return 0; + } + } + } +#else + return -EPROTONOSUPPORT; +#endif +} + +#define LZ4_BUFSIZE (512*1024u) + +int compress_stream_lz4(int fdf, int fdt, uint64_t max_bytes) { + +#ifdef HAVE_LZ4 + LZ4F_errorCode_t c; + _cleanup_(LZ4F_freeCompressionContextp) LZ4F_compressionContext_t ctx = NULL; + _cleanup_free_ char *buf = NULL; + char *src = NULL; + size_t size, n, total_in = 0, total_out, offset = 0, frame_size; + struct stat st; + int r; + static const LZ4F_compressOptions_t options = { + .stableSrc = 1, + }; + static const LZ4F_preferences_t preferences = { + .frameInfo.blockSizeID = 5, + }; + + c = LZ4F_createCompressionContext(&ctx, LZ4F_VERSION); + if (LZ4F_isError(c)) + return -ENOMEM; + + if (fstat(fdf, &st) < 0) + return log_debug_errno(errno, "fstat() failed: %m"); + + frame_size = LZ4F_compressBound(LZ4_BUFSIZE, &preferences); + size = frame_size + 64*1024; /* add some space for header and trailer */ + buf = malloc(size); + if (!buf) + return -ENOMEM; + + n = offset = total_out = LZ4F_compressBegin(ctx, buf, size, &preferences); + if (LZ4F_isError(n)) + return -EINVAL; + + src = mmap(NULL, st.st_size, PROT_READ, MAP_PRIVATE, fdf, 0); + if (src == MAP_FAILED) + return -errno; + + log_debug("Buffer size is %zu bytes, header size %zu bytes.", size, n); + + while (total_in < (size_t) st.st_size) { + ssize_t k; + + k = MIN(LZ4_BUFSIZE, st.st_size - total_in); + n = LZ4F_compressUpdate(ctx, buf + offset, size - offset, + src + total_in, k, &options); + if (LZ4F_isError(n)) { + r = -ENOTRECOVERABLE; + goto cleanup; + } + + total_in += k; + offset += n; + total_out += n; + + if (max_bytes != (uint64_t) -1 && total_out > (size_t) max_bytes) { + log_debug("Compressed stream longer than %"PRIu64" bytes", max_bytes); + return -EFBIG; + } + + if (size - offset < frame_size + 4) { + k = loop_write(fdt, buf, offset, false); + if (k < 0) { + r = k; + goto cleanup; + } + offset = 0; + } + } + + n = LZ4F_compressEnd(ctx, buf + offset, size - offset, &options); + if (LZ4F_isError(n)) { + r = -ENOTRECOVERABLE; + goto cleanup; + } + + offset += n; + total_out += n; + r = loop_write(fdt, buf, offset, false); + if (r < 0) + goto cleanup; + + log_debug("LZ4 compression finished (%zu -> %zu bytes, %.1f%%)", + total_in, total_out, + (double) total_out / total_in * 100); + cleanup: + munmap(src, st.st_size); + return r; +#else + return -EPROTONOSUPPORT; +#endif +} + +int decompress_stream_xz(int fdf, int fdt, uint64_t max_bytes) { + +#ifdef HAVE_XZ + _cleanup_(lzma_end) lzma_stream s = LZMA_STREAM_INIT; + lzma_ret ret; + + uint8_t buf[BUFSIZ], out[BUFSIZ]; + lzma_action action = LZMA_RUN; + + assert(fdf >= 0); + assert(fdt >= 0); + + ret = lzma_stream_decoder(&s, UINT64_MAX, 0); + if (ret != LZMA_OK) { + log_debug("Failed to initialize XZ decoder: code %u", ret); + return -ENOMEM; + } + + for (;;) { + if (s.avail_in == 0 && action == LZMA_RUN) { + ssize_t n; + + n = read(fdf, buf, sizeof(buf)); + if (n < 0) + return -errno; + if (n == 0) + action = LZMA_FINISH; + else { + s.next_in = buf; + s.avail_in = n; + } + } + + if (s.avail_out == 0) { + s.next_out = out; + s.avail_out = sizeof(out); + } + + ret = lzma_code(&s, action); + if (ret != LZMA_OK && ret != LZMA_STREAM_END) { + log_debug("Decompression failed: code %u", ret); + return -EBADMSG; + } + + if (s.avail_out == 0 || ret == LZMA_STREAM_END) { + ssize_t n, k; + + n = sizeof(out) - s.avail_out; + + if (max_bytes != (uint64_t) -1) { + if (max_bytes < (uint64_t) n) + return -EFBIG; + + max_bytes -= n; + } + + k = loop_write(fdt, out, n, false); + if (k < 0) + return k; + + if (ret == LZMA_STREAM_END) { + log_debug("XZ decompression finished (%"PRIu64" -> %"PRIu64" bytes, %.1f%%)", + s.total_in, s.total_out, + (double) s.total_out / s.total_in * 100); + + return 0; + } + } + } +#else + log_debug("Cannot decompress file. Compiled without XZ support."); + return -EPROTONOSUPPORT; +#endif +} + +int decompress_stream_lz4(int in, int out, uint64_t max_bytes) { +#ifdef HAVE_LZ4 + size_t c; + _cleanup_(LZ4F_freeDecompressionContextp) LZ4F_decompressionContext_t ctx = NULL; + _cleanup_free_ char *buf = NULL; + char *src; + struct stat st; + int r = 0; + size_t total_in = 0, total_out = 0; + + c = LZ4F_createDecompressionContext(&ctx, LZ4F_VERSION); + if (LZ4F_isError(c)) + return -ENOMEM; + + if (fstat(in, &st) < 0) + return log_debug_errno(errno, "fstat() failed: %m"); + + buf = malloc(LZ4_BUFSIZE); + if (!buf) + return -ENOMEM; + + src = mmap(NULL, st.st_size, PROT_READ, MAP_PRIVATE, in, 0); + if (src == MAP_FAILED) + return -errno; + + while (total_in < (size_t) st.st_size) { + size_t produced = LZ4_BUFSIZE; + size_t used = st.st_size - total_in; + + c = LZ4F_decompress(ctx, buf, &produced, src + total_in, &used, NULL); + if (LZ4F_isError(c)) { + r = -EBADMSG; + goto cleanup; + } + + total_in += used; + total_out += produced; + + if (max_bytes != (uint64_t) -1 && total_out > (size_t) max_bytes) { + log_debug("Decompressed stream longer than %"PRIu64" bytes", max_bytes); + r = -EFBIG; + goto cleanup; + } + + r = loop_write(out, buf, produced, false); + if (r < 0) + goto cleanup; + } + + log_debug("LZ4 decompression finished (%zu -> %zu bytes, %.1f%%)", + total_in, total_out, + (double) total_out / total_in * 100); + cleanup: + munmap(src, st.st_size); + return r; +#else + log_debug("Cannot decompress file. Compiled without LZ4 support."); + return -EPROTONOSUPPORT; +#endif +} + +int decompress_stream(const char *filename, int fdf, int fdt, uint64_t max_bytes) { + + if (endswith(filename, ".lz4")) + return decompress_stream_lz4(fdf, fdt, max_bytes); + else if (endswith(filename, ".xz")) + return decompress_stream_xz(fdf, fdt, max_bytes); + else + return -EPROTONOSUPPORT; +} diff --git a/src/libsystemd/src/sd-journal/compress.h b/src/libsystemd/src/sd-journal/compress.h new file mode 100644 index 0000000000..c138099d9a --- /dev/null +++ b/src/libsystemd/src/sd-journal/compress.h @@ -0,0 +1,85 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <unistd.h> + +#include "journal-def.h" + +const char* object_compressed_to_string(int compression); +int object_compressed_from_string(const char *compression); + +int compress_blob_xz(const void *src, uint64_t src_size, + void *dst, size_t dst_alloc_size, size_t *dst_size); +int compress_blob_lz4(const void *src, uint64_t src_size, + void *dst, size_t dst_alloc_size, size_t *dst_size); + +static inline int compress_blob(const void *src, uint64_t src_size, + void *dst, size_t dst_alloc_size, size_t *dst_size) { + int r; +#ifdef HAVE_LZ4 + r = compress_blob_lz4(src, src_size, dst, dst_alloc_size, dst_size); + if (r == 0) + return OBJECT_COMPRESSED_LZ4; +#else + r = compress_blob_xz(src, src_size, dst, dst_alloc_size, dst_size); + if (r == 0) + return OBJECT_COMPRESSED_XZ; +#endif + return r; +} + +int decompress_blob_xz(const void *src, uint64_t src_size, + void **dst, size_t *dst_alloc_size, size_t* dst_size, size_t dst_max); +int decompress_blob_lz4(const void *src, uint64_t src_size, + void **dst, size_t *dst_alloc_size, size_t* dst_size, size_t dst_max); +int decompress_blob(int compression, + const void *src, uint64_t src_size, + void **dst, size_t *dst_alloc_size, size_t* dst_size, size_t dst_max); + +int decompress_startswith_xz(const void *src, uint64_t src_size, + void **buffer, size_t *buffer_size, + const void *prefix, size_t prefix_len, + uint8_t extra); +int decompress_startswith_lz4(const void *src, uint64_t src_size, + void **buffer, size_t *buffer_size, + const void *prefix, size_t prefix_len, + uint8_t extra); +int decompress_startswith(int compression, + const void *src, uint64_t src_size, + void **buffer, size_t *buffer_size, + const void *prefix, size_t prefix_len, + uint8_t extra); + +int compress_stream_xz(int fdf, int fdt, uint64_t max_bytes); +int compress_stream_lz4(int fdf, int fdt, uint64_t max_bytes); + +int decompress_stream_xz(int fdf, int fdt, uint64_t max_size); +int decompress_stream_lz4(int fdf, int fdt, uint64_t max_size); + +#ifdef HAVE_LZ4 +# define compress_stream compress_stream_lz4 +# define COMPRESSED_EXT ".lz4" +#else +# define compress_stream compress_stream_xz +# define COMPRESSED_EXT ".xz" +#endif + +int decompress_stream(const char *filename, int fdf, int fdt, uint64_t max_bytes); diff --git a/src/libsystemd/src/sd-journal/fsprg.c b/src/libsystemd/src/sd-journal/fsprg.c new file mode 100644 index 0000000000..c0f59b28fd --- /dev/null +++ b/src/libsystemd/src/sd-journal/fsprg.c @@ -0,0 +1,375 @@ +/* + * fsprg v0.1 - (seekable) forward-secure pseudorandom generator + * Copyright (C) 2012 B. Poettering + * Contact: fsprg@point-at-infinity.org + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA + * 02110-1301 USA + */ + +/* + * See "Practical Secure Logging: Seekable Sequential Key Generators" + * by G. A. Marson, B. Poettering for details: + * + * http://eprint.iacr.org/2013/397 + */ + +#include <gcrypt.h> +#include <string.h> + +#include "systemd-gcrypt/gcrypt-util.h" + +#include "fsprg.h" + +#define ISVALID_SECPAR(secpar) (((secpar) % 16 == 0) && ((secpar) >= 16) && ((secpar) <= 16384)) +#define VALIDATE_SECPAR(secpar) assert(ISVALID_SECPAR(secpar)); + +#define RND_HASH GCRY_MD_SHA256 +#define RND_GEN_P 0x01 +#define RND_GEN_Q 0x02 +#define RND_GEN_X 0x03 + +/******************************************************************************/ + +static void mpi_export(void *buf, size_t buflen, const gcry_mpi_t x) { + unsigned len; + size_t nwritten; + + assert(gcry_mpi_cmp_ui(x, 0) >= 0); + len = (gcry_mpi_get_nbits(x) + 7) / 8; + assert(len <= buflen); + memzero(buf, buflen); + gcry_mpi_print(GCRYMPI_FMT_USG, buf + (buflen - len), len, &nwritten, x); + assert(nwritten == len); +} + +static gcry_mpi_t mpi_import(const void *buf, size_t buflen) { + gcry_mpi_t h; + unsigned len; + + assert_se(gcry_mpi_scan(&h, GCRYMPI_FMT_USG, buf, buflen, NULL) == 0); + len = (gcry_mpi_get_nbits(h) + 7) / 8; + assert(len <= buflen); + assert(gcry_mpi_cmp_ui(h, 0) >= 0); + + return h; +} + +static void uint64_export(void *buf, size_t buflen, uint64_t x) { + assert(buflen == 8); + ((uint8_t*) buf)[0] = (x >> 56) & 0xff; + ((uint8_t*) buf)[1] = (x >> 48) & 0xff; + ((uint8_t*) buf)[2] = (x >> 40) & 0xff; + ((uint8_t*) buf)[3] = (x >> 32) & 0xff; + ((uint8_t*) buf)[4] = (x >> 24) & 0xff; + ((uint8_t*) buf)[5] = (x >> 16) & 0xff; + ((uint8_t*) buf)[6] = (x >> 8) & 0xff; + ((uint8_t*) buf)[7] = (x >> 0) & 0xff; +} + +_pure_ static uint64_t uint64_import(const void *buf, size_t buflen) { + assert(buflen == 8); + return + (uint64_t)(((uint8_t*) buf)[0]) << 56 | + (uint64_t)(((uint8_t*) buf)[1]) << 48 | + (uint64_t)(((uint8_t*) buf)[2]) << 40 | + (uint64_t)(((uint8_t*) buf)[3]) << 32 | + (uint64_t)(((uint8_t*) buf)[4]) << 24 | + (uint64_t)(((uint8_t*) buf)[5]) << 16 | + (uint64_t)(((uint8_t*) buf)[6]) << 8 | + (uint64_t)(((uint8_t*) buf)[7]) << 0; +} + +/* deterministically generate from seed/idx a string of buflen pseudorandom bytes */ +static void det_randomize(void *buf, size_t buflen, const void *seed, size_t seedlen, uint32_t idx) { + gcry_md_hd_t hd, hd2; + size_t olen, cpylen; + uint32_t ctr; + + olen = gcry_md_get_algo_dlen(RND_HASH); + gcry_md_open(&hd, RND_HASH, 0); + gcry_md_write(hd, seed, seedlen); + gcry_md_putc(hd, (idx >> 24) & 0xff); + gcry_md_putc(hd, (idx >> 16) & 0xff); + gcry_md_putc(hd, (idx >> 8) & 0xff); + gcry_md_putc(hd, (idx >> 0) & 0xff); + + for (ctr = 0; buflen; ctr++) { + gcry_md_copy(&hd2, hd); + gcry_md_putc(hd2, (ctr >> 24) & 0xff); + gcry_md_putc(hd2, (ctr >> 16) & 0xff); + gcry_md_putc(hd2, (ctr >> 8) & 0xff); + gcry_md_putc(hd2, (ctr >> 0) & 0xff); + gcry_md_final(hd2); + cpylen = (buflen < olen) ? buflen : olen; + memcpy(buf, gcry_md_read(hd2, RND_HASH), cpylen); + gcry_md_close(hd2); + buf += cpylen; + buflen -= cpylen; + } + gcry_md_close(hd); +} + +/* deterministically generate from seed/idx a prime of length `bits' that is 3 (mod 4) */ +static gcry_mpi_t genprime3mod4(int bits, const void *seed, size_t seedlen, uint32_t idx) { + size_t buflen = bits / 8; + uint8_t buf[buflen]; + gcry_mpi_t p; + + assert(bits % 8 == 0); + assert(buflen > 0); + + det_randomize(buf, buflen, seed, seedlen, idx); + buf[0] |= 0xc0; /* set upper two bits, so that n=pq has maximum size */ + buf[buflen - 1] |= 0x03; /* set lower two bits, to have result 3 (mod 4) */ + + p = mpi_import(buf, buflen); + while (gcry_prime_check(p, 0)) + gcry_mpi_add_ui(p, p, 4); + + return p; +} + +/* deterministically generate from seed/idx a quadratic residue (mod n) */ +static gcry_mpi_t gensquare(const gcry_mpi_t n, const void *seed, size_t seedlen, uint32_t idx, unsigned secpar) { + size_t buflen = secpar / 8; + uint8_t buf[buflen]; + gcry_mpi_t x; + + det_randomize(buf, buflen, seed, seedlen, idx); + buf[0] &= 0x7f; /* clear upper bit, so that we have x < n */ + x = mpi_import(buf, buflen); + assert(gcry_mpi_cmp(x, n) < 0); + gcry_mpi_mulm(x, x, x, n); + return x; +} + +/* compute 2^m (mod phi(p)), for a prime p */ +static gcry_mpi_t twopowmodphi(uint64_t m, const gcry_mpi_t p) { + gcry_mpi_t phi, r; + int n; + + phi = gcry_mpi_new(0); + gcry_mpi_sub_ui(phi, p, 1); + + /* count number of used bits in m */ + for (n = 0; (1ULL << n) <= m; n++) + ; + + r = gcry_mpi_new(0); + gcry_mpi_set_ui(r, 1); + while (n) { /* square and multiply algorithm for fast exponentiation */ + n--; + gcry_mpi_mulm(r, r, r, phi); + if (m & ((uint64_t)1 << n)) { + gcry_mpi_add(r, r, r); + if (gcry_mpi_cmp(r, phi) >= 0) + gcry_mpi_sub(r, r, phi); + } + } + + gcry_mpi_release(phi); + return r; +} + +/* Decompose $x \in Z_n$ into $(xp,xq) \in Z_p \times Z_q$ using Chinese Remainder Theorem */ +static void CRT_decompose(gcry_mpi_t *xp, gcry_mpi_t *xq, const gcry_mpi_t x, const gcry_mpi_t p, const gcry_mpi_t q) { + *xp = gcry_mpi_new(0); + *xq = gcry_mpi_new(0); + gcry_mpi_mod(*xp, x, p); + gcry_mpi_mod(*xq, x, q); +} + +/* Compose $(xp,xq) \in Z_p \times Z_q$ into $x \in Z_n$ using Chinese Remainder Theorem */ +static void CRT_compose(gcry_mpi_t *x, const gcry_mpi_t xp, const gcry_mpi_t xq, const gcry_mpi_t p, const gcry_mpi_t q) { + gcry_mpi_t a, u; + + a = gcry_mpi_new(0); + u = gcry_mpi_new(0); + *x = gcry_mpi_new(0); + gcry_mpi_subm(a, xq, xp, q); + gcry_mpi_invm(u, p, q); + gcry_mpi_mulm(a, a, u, q); /* a = (xq - xp) / p (mod q) */ + gcry_mpi_mul(*x, p, a); + gcry_mpi_add(*x, *x, xp); /* x = p * ((xq - xp) / p mod q) + xp */ + gcry_mpi_release(a); + gcry_mpi_release(u); +} + +/******************************************************************************/ + +size_t FSPRG_mskinbytes(unsigned _secpar) { + VALIDATE_SECPAR(_secpar); + return 2 + 2 * (_secpar / 2) / 8; /* to store header,p,q */ +} + +size_t FSPRG_mpkinbytes(unsigned _secpar) { + VALIDATE_SECPAR(_secpar); + return 2 + _secpar / 8; /* to store header,n */ +} + +size_t FSPRG_stateinbytes(unsigned _secpar) { + VALIDATE_SECPAR(_secpar); + return 2 + 2 * _secpar / 8 + 8; /* to store header,n,x,epoch */ +} + +static void store_secpar(void *buf, uint16_t secpar) { + secpar = secpar / 16 - 1; + ((uint8_t*) buf)[0] = (secpar >> 8) & 0xff; + ((uint8_t*) buf)[1] = (secpar >> 0) & 0xff; +} + +static uint16_t read_secpar(const void *buf) { + uint16_t secpar; + secpar = + (uint16_t)(((uint8_t*) buf)[0]) << 8 | + (uint16_t)(((uint8_t*) buf)[1]) << 0; + return 16 * (secpar + 1); +} + +void FSPRG_GenMK(void *msk, void *mpk, const void *seed, size_t seedlen, unsigned _secpar) { + uint8_t iseed[FSPRG_RECOMMENDED_SEEDLEN]; + gcry_mpi_t n, p, q; + uint16_t secpar; + + VALIDATE_SECPAR(_secpar); + secpar = _secpar; + + initialize_libgcrypt(false); + + if (!seed) { + gcry_randomize(iseed, FSPRG_RECOMMENDED_SEEDLEN, GCRY_STRONG_RANDOM); + seed = iseed; + seedlen = FSPRG_RECOMMENDED_SEEDLEN; + } + + p = genprime3mod4(secpar / 2, seed, seedlen, RND_GEN_P); + q = genprime3mod4(secpar / 2, seed, seedlen, RND_GEN_Q); + + if (msk) { + store_secpar(msk + 0, secpar); + mpi_export(msk + 2 + 0 * (secpar / 2) / 8, (secpar / 2) / 8, p); + mpi_export(msk + 2 + 1 * (secpar / 2) / 8, (secpar / 2) / 8, q); + } + + if (mpk) { + n = gcry_mpi_new(0); + gcry_mpi_mul(n, p, q); + assert(gcry_mpi_get_nbits(n) == secpar); + + store_secpar(mpk + 0, secpar); + mpi_export(mpk + 2, secpar / 8, n); + + gcry_mpi_release(n); + } + + gcry_mpi_release(p); + gcry_mpi_release(q); +} + +void FSPRG_GenState0(void *state, const void *mpk, const void *seed, size_t seedlen) { + gcry_mpi_t n, x; + uint16_t secpar; + + initialize_libgcrypt(false); + + secpar = read_secpar(mpk + 0); + n = mpi_import(mpk + 2, secpar / 8); + x = gensquare(n, seed, seedlen, RND_GEN_X, secpar); + + memcpy(state, mpk, 2 + secpar / 8); + mpi_export(state + 2 + 1 * secpar / 8, secpar / 8, x); + memzero(state + 2 + 2 * secpar / 8, 8); + + gcry_mpi_release(n); + gcry_mpi_release(x); +} + +void FSPRG_Evolve(void *state) { + gcry_mpi_t n, x; + uint16_t secpar; + uint64_t epoch; + + initialize_libgcrypt(false); + + secpar = read_secpar(state + 0); + n = mpi_import(state + 2 + 0 * secpar / 8, secpar / 8); + x = mpi_import(state + 2 + 1 * secpar / 8, secpar / 8); + epoch = uint64_import(state + 2 + 2 * secpar / 8, 8); + + gcry_mpi_mulm(x, x, x, n); + epoch++; + + mpi_export(state + 2 + 1 * secpar / 8, secpar / 8, x); + uint64_export(state + 2 + 2 * secpar / 8, 8, epoch); + + gcry_mpi_release(n); + gcry_mpi_release(x); +} + +uint64_t FSPRG_GetEpoch(const void *state) { + uint16_t secpar; + secpar = read_secpar(state + 0); + return uint64_import(state + 2 + 2 * secpar / 8, 8); +} + +void FSPRG_Seek(void *state, uint64_t epoch, const void *msk, const void *seed, size_t seedlen) { + gcry_mpi_t p, q, n, x, xp, xq, kp, kq, xm; + uint16_t secpar; + + initialize_libgcrypt(false); + + secpar = read_secpar(msk + 0); + p = mpi_import(msk + 2 + 0 * (secpar / 2) / 8, (secpar / 2) / 8); + q = mpi_import(msk + 2 + 1 * (secpar / 2) / 8, (secpar / 2) / 8); + + n = gcry_mpi_new(0); + gcry_mpi_mul(n, p, q); + + x = gensquare(n, seed, seedlen, RND_GEN_X, secpar); + CRT_decompose(&xp, &xq, x, p, q); /* split (mod n) into (mod p) and (mod q) using CRT */ + + kp = twopowmodphi(epoch, p); /* compute 2^epoch (mod phi(p)) */ + kq = twopowmodphi(epoch, q); /* compute 2^epoch (mod phi(q)) */ + + gcry_mpi_powm(xp, xp, kp, p); /* compute x^(2^epoch) (mod p) */ + gcry_mpi_powm(xq, xq, kq, q); /* compute x^(2^epoch) (mod q) */ + + CRT_compose(&xm, xp, xq, p, q); /* combine (mod p) and (mod q) to (mod n) using CRT */ + + store_secpar(state + 0, secpar); + mpi_export(state + 2 + 0 * secpar / 8, secpar / 8, n); + mpi_export(state + 2 + 1 * secpar / 8, secpar / 8, xm); + uint64_export(state + 2 + 2 * secpar / 8, 8, epoch); + + gcry_mpi_release(p); + gcry_mpi_release(q); + gcry_mpi_release(n); + gcry_mpi_release(x); + gcry_mpi_release(xp); + gcry_mpi_release(xq); + gcry_mpi_release(kp); + gcry_mpi_release(kq); + gcry_mpi_release(xm); +} + +void FSPRG_GetKey(const void *state, void *key, size_t keylen, uint32_t idx) { + uint16_t secpar; + + initialize_libgcrypt(false); + + secpar = read_secpar(state + 0); + det_randomize(key, keylen, state + 2, 2 * secpar / 8 + 8, idx); +} diff --git a/src/libsystemd/src/sd-journal/fsprg.h b/src/libsystemd/src/sd-journal/fsprg.h new file mode 100644 index 0000000000..15a4e04fe8 --- /dev/null +++ b/src/libsystemd/src/sd-journal/fsprg.h @@ -0,0 +1,65 @@ +#ifndef __fsprgh__ +#define __fsprgh__ + +/* + * fsprg v0.1 - (seekable) forward-secure pseudorandom generator + * Copyright (C) 2012 B. Poettering + * Contact: fsprg@point-at-infinity.org + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA + * 02110-1301 USA + * + */ + +#include <inttypes.h> +#include <sys/types.h> + +#include "systemd-basic/macro.h" +#include "systemd-basic/util.h" + +#ifdef __cplusplus +extern "C" { +#endif + +#define FSPRG_RECOMMENDED_SECPAR 1536 +#define FSPRG_RECOMMENDED_SEEDLEN (96/8) + +size_t FSPRG_mskinbytes(unsigned secpar) _const_; +size_t FSPRG_mpkinbytes(unsigned secpar) _const_; +size_t FSPRG_stateinbytes(unsigned secpar) _const_; + +/* Setup msk and mpk. Providing seed != NULL makes this algorithm deterministic. */ +void FSPRG_GenMK(void *msk, void *mpk, const void *seed, size_t seedlen, unsigned secpar); + +/* Initialize state deterministically in dependence on seed. */ +/* Note: in case one wants to run only one GenState0 per GenMK it is safe to use + the same seed for both GenMK and GenState0. +*/ +void FSPRG_GenState0(void *state, const void *mpk, const void *seed, size_t seedlen); + +void FSPRG_Evolve(void *state); + +uint64_t FSPRG_GetEpoch(const void *state) _pure_; + +/* Seek to any arbitrary state (by providing msk together with seed from GenState0). */ +void FSPRG_Seek(void *state, uint64_t epoch, const void *msk, const void *seed, size_t seedlen); + +void FSPRG_GetKey(const void *state, void *key, size_t keylen, uint32_t idx); + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/libsystemd/src/sd-journal/journal-authenticate.c b/src/libsystemd/src/sd-journal/journal-authenticate.c new file mode 100644 index 0000000000..f3a207d88d --- /dev/null +++ b/src/libsystemd/src/sd-journal/journal-authenticate.c @@ -0,0 +1,552 @@ +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <fcntl.h> +#include <sys/mman.h> + +#include "systemd-basic/fd-util.h" +#include "systemd-basic/hexdecoct.h" +#include "systemd-gcrypt/gcrypt-util.h" + +#include "fsprg.h" +#include "journal-authenticate.h" +#include "journal-def.h" +#include "journal-file.h" + +static uint64_t journal_file_tag_seqnum(JournalFile *f) { + uint64_t r; + + assert(f); + + r = le64toh(f->header->n_tags) + 1; + f->header->n_tags = htole64(r); + + return r; +} + +int journal_file_append_tag(JournalFile *f) { + Object *o; + uint64_t p; + int r; + + assert(f); + + if (!f->seal) + return 0; + + if (!f->hmac_running) + return 0; + + assert(f->hmac); + + r = journal_file_append_object(f, OBJECT_TAG, sizeof(struct TagObject), &o, &p); + if (r < 0) + return r; + + o->tag.seqnum = htole64(journal_file_tag_seqnum(f)); + o->tag.epoch = htole64(FSPRG_GetEpoch(f->fsprg_state)); + + log_debug("Writing tag %"PRIu64" for epoch %"PRIu64"", + le64toh(o->tag.seqnum), + FSPRG_GetEpoch(f->fsprg_state)); + + /* Add the tag object itself, so that we can protect its + * header. This will exclude the actual hash value in it */ + r = journal_file_hmac_put_object(f, OBJECT_TAG, o, p); + if (r < 0) + return r; + + /* Get the HMAC tag and store it in the object */ + memcpy(o->tag.tag, gcry_md_read(f->hmac, 0), TAG_LENGTH); + f->hmac_running = false; + + return 0; +} + +int journal_file_hmac_start(JournalFile *f) { + uint8_t key[256 / 8]; /* Let's pass 256 bit from FSPRG to HMAC */ + assert(f); + + if (!f->seal) + return 0; + + if (f->hmac_running) + return 0; + + /* Prepare HMAC for next cycle */ + gcry_md_reset(f->hmac); + FSPRG_GetKey(f->fsprg_state, key, sizeof(key), 0); + gcry_md_setkey(f->hmac, key, sizeof(key)); + + f->hmac_running = true; + + return 0; +} + +static int journal_file_get_epoch(JournalFile *f, uint64_t realtime, uint64_t *epoch) { + uint64_t t; + + assert(f); + assert(epoch); + assert(f->seal); + + if (f->fss_start_usec == 0 || + f->fss_interval_usec == 0) + return -EOPNOTSUPP; + + if (realtime < f->fss_start_usec) + return -ESTALE; + + t = realtime - f->fss_start_usec; + t = t / f->fss_interval_usec; + + *epoch = t; + return 0; +} + +static int journal_file_fsprg_need_evolve(JournalFile *f, uint64_t realtime) { + uint64_t goal, epoch; + int r; + assert(f); + + if (!f->seal) + return 0; + + r = journal_file_get_epoch(f, realtime, &goal); + if (r < 0) + return r; + + epoch = FSPRG_GetEpoch(f->fsprg_state); + if (epoch > goal) + return -ESTALE; + + return epoch != goal; +} + +int journal_file_fsprg_evolve(JournalFile *f, uint64_t realtime) { + uint64_t goal, epoch; + int r; + + assert(f); + + if (!f->seal) + return 0; + + r = journal_file_get_epoch(f, realtime, &goal); + if (r < 0) + return r; + + epoch = FSPRG_GetEpoch(f->fsprg_state); + if (epoch < goal) + log_debug("Evolving FSPRG key from epoch %"PRIu64" to %"PRIu64".", epoch, goal); + + for (;;) { + if (epoch > goal) + return -ESTALE; + if (epoch == goal) + return 0; + + FSPRG_Evolve(f->fsprg_state); + epoch = FSPRG_GetEpoch(f->fsprg_state); + } +} + +int journal_file_fsprg_seek(JournalFile *f, uint64_t goal) { + void *msk; + uint64_t epoch; + + assert(f); + + if (!f->seal) + return 0; + + assert(f->fsprg_seed); + + if (f->fsprg_state) { + /* Cheaper... */ + + epoch = FSPRG_GetEpoch(f->fsprg_state); + if (goal == epoch) + return 0; + + if (goal == epoch+1) { + FSPRG_Evolve(f->fsprg_state); + return 0; + } + } else { + f->fsprg_state_size = FSPRG_stateinbytes(FSPRG_RECOMMENDED_SECPAR); + f->fsprg_state = malloc(f->fsprg_state_size); + + if (!f->fsprg_state) + return -ENOMEM; + } + + log_debug("Seeking FSPRG key to %"PRIu64".", goal); + + msk = alloca(FSPRG_mskinbytes(FSPRG_RECOMMENDED_SECPAR)); + FSPRG_GenMK(msk, NULL, f->fsprg_seed, f->fsprg_seed_size, FSPRG_RECOMMENDED_SECPAR); + FSPRG_Seek(f->fsprg_state, goal, msk, f->fsprg_seed, f->fsprg_seed_size); + return 0; +} + +int journal_file_maybe_append_tag(JournalFile *f, uint64_t realtime) { + int r; + + assert(f); + + if (!f->seal) + return 0; + + if (realtime <= 0) + realtime = now(CLOCK_REALTIME); + + r = journal_file_fsprg_need_evolve(f, realtime); + if (r <= 0) + return 0; + + r = journal_file_append_tag(f); + if (r < 0) + return r; + + r = journal_file_fsprg_evolve(f, realtime); + if (r < 0) + return r; + + return 0; +} + +int journal_file_hmac_put_object(JournalFile *f, ObjectType type, Object *o, uint64_t p) { + int r; + + assert(f); + + if (!f->seal) + return 0; + + r = journal_file_hmac_start(f); + if (r < 0) + return r; + + if (!o) { + r = journal_file_move_to_object(f, type, p, &o); + if (r < 0) + return r; + } else { + if (type > OBJECT_UNUSED && o->object.type != type) + return -EBADMSG; + } + + gcry_md_write(f->hmac, o, offsetof(ObjectHeader, payload)); + + switch (o->object.type) { + + case OBJECT_DATA: + /* All but hash and payload are mutable */ + gcry_md_write(f->hmac, &o->data.hash, sizeof(o->data.hash)); + gcry_md_write(f->hmac, o->data.payload, le64toh(o->object.size) - offsetof(DataObject, payload)); + break; + + case OBJECT_FIELD: + /* Same here */ + gcry_md_write(f->hmac, &o->field.hash, sizeof(o->field.hash)); + gcry_md_write(f->hmac, o->field.payload, le64toh(o->object.size) - offsetof(FieldObject, payload)); + break; + + case OBJECT_ENTRY: + /* All */ + gcry_md_write(f->hmac, &o->entry.seqnum, le64toh(o->object.size) - offsetof(EntryObject, seqnum)); + break; + + case OBJECT_FIELD_HASH_TABLE: + case OBJECT_DATA_HASH_TABLE: + case OBJECT_ENTRY_ARRAY: + /* Nothing: everything is mutable */ + break; + + case OBJECT_TAG: + /* All but the tag itself */ + gcry_md_write(f->hmac, &o->tag.seqnum, sizeof(o->tag.seqnum)); + gcry_md_write(f->hmac, &o->tag.epoch, sizeof(o->tag.epoch)); + break; + default: + return -EINVAL; + } + + return 0; +} + +int journal_file_hmac_put_header(JournalFile *f) { + int r; + + assert(f); + + if (!f->seal) + return 0; + + r = journal_file_hmac_start(f); + if (r < 0) + return r; + + /* All but state+reserved, boot_id, arena_size, + * tail_object_offset, n_objects, n_entries, + * tail_entry_seqnum, head_entry_seqnum, entry_array_offset, + * head_entry_realtime, tail_entry_realtime, + * tail_entry_monotonic, n_data, n_fields, n_tags, + * n_entry_arrays. */ + + gcry_md_write(f->hmac, f->header->signature, offsetof(Header, state) - offsetof(Header, signature)); + gcry_md_write(f->hmac, &f->header->file_id, offsetof(Header, boot_id) - offsetof(Header, file_id)); + gcry_md_write(f->hmac, &f->header->seqnum_id, offsetof(Header, arena_size) - offsetof(Header, seqnum_id)); + gcry_md_write(f->hmac, &f->header->data_hash_table_offset, offsetof(Header, tail_object_offset) - offsetof(Header, data_hash_table_offset)); + + return 0; +} + +int journal_file_fss_load(JournalFile *f) { + int r, fd = -1; + char *p = NULL; + struct stat st; + FSSHeader *m = NULL; + sd_id128_t machine; + + assert(f); + + if (!f->seal) + return 0; + + r = sd_id128_get_machine(&machine); + if (r < 0) + return r; + + if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/fss", + SD_ID128_FORMAT_VAL(machine)) < 0) + return -ENOMEM; + + fd = open(p, O_RDWR|O_CLOEXEC|O_NOCTTY, 0600); + if (fd < 0) { + if (errno != ENOENT) + log_error_errno(errno, "Failed to open %s: %m", p); + + r = -errno; + goto finish; + } + + if (fstat(fd, &st) < 0) { + r = -errno; + goto finish; + } + + if (st.st_size < (off_t) sizeof(FSSHeader)) { + r = -ENODATA; + goto finish; + } + + m = mmap(NULL, PAGE_ALIGN(sizeof(FSSHeader)), PROT_READ, MAP_SHARED, fd, 0); + if (m == MAP_FAILED) { + m = NULL; + r = -errno; + goto finish; + } + + if (memcmp(m->signature, FSS_HEADER_SIGNATURE, 8) != 0) { + r = -EBADMSG; + goto finish; + } + + if (m->incompatible_flags != 0) { + r = -EPROTONOSUPPORT; + goto finish; + } + + if (le64toh(m->header_size) < sizeof(FSSHeader)) { + r = -EBADMSG; + goto finish; + } + + if (le64toh(m->fsprg_state_size) != FSPRG_stateinbytes(le16toh(m->fsprg_secpar))) { + r = -EBADMSG; + goto finish; + } + + f->fss_file_size = le64toh(m->header_size) + le64toh(m->fsprg_state_size); + if ((uint64_t) st.st_size < f->fss_file_size) { + r = -ENODATA; + goto finish; + } + + if (!sd_id128_equal(machine, m->machine_id)) { + r = -EHOSTDOWN; + goto finish; + } + + if (le64toh(m->start_usec) <= 0 || + le64toh(m->interval_usec) <= 0) { + r = -EBADMSG; + goto finish; + } + + f->fss_file = mmap(NULL, PAGE_ALIGN(f->fss_file_size), PROT_READ|PROT_WRITE, MAP_SHARED, fd, 0); + if (f->fss_file == MAP_FAILED) { + f->fss_file = NULL; + r = -errno; + goto finish; + } + + f->fss_start_usec = le64toh(f->fss_file->start_usec); + f->fss_interval_usec = le64toh(f->fss_file->interval_usec); + + f->fsprg_state = (uint8_t*) f->fss_file + le64toh(f->fss_file->header_size); + f->fsprg_state_size = le64toh(f->fss_file->fsprg_state_size); + + r = 0; + +finish: + if (m) + munmap(m, PAGE_ALIGN(sizeof(FSSHeader))); + + safe_close(fd); + free(p); + + return r; +} + +int journal_file_hmac_setup(JournalFile *f) { + gcry_error_t e; + + if (!f->seal) + return 0; + + initialize_libgcrypt(true); + + e = gcry_md_open(&f->hmac, GCRY_MD_SHA256, GCRY_MD_FLAG_HMAC); + if (e != 0) + return -EOPNOTSUPP; + + return 0; +} + +int journal_file_append_first_tag(JournalFile *f) { + int r; + uint64_t p; + + if (!f->seal) + return 0; + + log_debug("Calculating first tag..."); + + r = journal_file_hmac_put_header(f); + if (r < 0) + return r; + + p = le64toh(f->header->field_hash_table_offset); + if (p < offsetof(Object, hash_table.items)) + return -EINVAL; + p -= offsetof(Object, hash_table.items); + + r = journal_file_hmac_put_object(f, OBJECT_FIELD_HASH_TABLE, NULL, p); + if (r < 0) + return r; + + p = le64toh(f->header->data_hash_table_offset); + if (p < offsetof(Object, hash_table.items)) + return -EINVAL; + p -= offsetof(Object, hash_table.items); + + r = journal_file_hmac_put_object(f, OBJECT_DATA_HASH_TABLE, NULL, p); + if (r < 0) + return r; + + r = journal_file_append_tag(f); + if (r < 0) + return r; + + return 0; +} + +int journal_file_parse_verification_key(JournalFile *f, const char *key) { + uint8_t *seed; + size_t seed_size, c; + const char *k; + int r; + unsigned long long start, interval; + + seed_size = FSPRG_RECOMMENDED_SEEDLEN; + seed = malloc(seed_size); + if (!seed) + return -ENOMEM; + + k = key; + for (c = 0; c < seed_size; c++) { + int x, y; + + while (*k == '-') + k++; + + x = unhexchar(*k); + if (x < 0) { + free(seed); + return -EINVAL; + } + k++; + y = unhexchar(*k); + if (y < 0) { + free(seed); + return -EINVAL; + } + k++; + + seed[c] = (uint8_t) (x * 16 + y); + } + + if (*k != '/') { + free(seed); + return -EINVAL; + } + k++; + + r = sscanf(k, "%llx-%llx", &start, &interval); + if (r != 2) { + free(seed); + return -EINVAL; + } + + f->fsprg_seed = seed; + f->fsprg_seed_size = seed_size; + + f->fss_start_usec = start * interval; + f->fss_interval_usec = interval; + + return 0; +} + +bool journal_file_next_evolve_usec(JournalFile *f, usec_t *u) { + uint64_t epoch; + + assert(f); + assert(u); + + if (!f->seal) + return false; + + epoch = FSPRG_GetEpoch(f->fsprg_state); + + *u = (usec_t) (f->fss_start_usec + f->fss_interval_usec * epoch + f->fss_interval_usec); + + return true; +} diff --git a/src/libsystemd/src/sd-journal/journal-authenticate.h b/src/libsystemd/src/sd-journal/journal-authenticate.h new file mode 100644 index 0000000000..6c87319ede --- /dev/null +++ b/src/libsystemd/src/sd-journal/journal-authenticate.h @@ -0,0 +1,41 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +#include "journal-file.h" + +int journal_file_append_tag(JournalFile *f); +int journal_file_maybe_append_tag(JournalFile *f, uint64_t realtime); +int journal_file_append_first_tag(JournalFile *f); + +int journal_file_hmac_setup(JournalFile *f); +int journal_file_hmac_start(JournalFile *f); +int journal_file_hmac_put_header(JournalFile *f); +int journal_file_hmac_put_object(JournalFile *f, ObjectType type, Object *o, uint64_t p); + +int journal_file_fss_load(JournalFile *f); +int journal_file_parse_verification_key(JournalFile *f, const char *key); + +int journal_file_fsprg_evolve(JournalFile *f, uint64_t realtime); +int journal_file_fsprg_seek(JournalFile *f, uint64_t epoch); + +bool journal_file_next_evolve_usec(JournalFile *f, usec_t *u); diff --git a/src/libsystemd/src/sd-journal/journal-def.h b/src/libsystemd/src/sd-journal/journal-def.h new file mode 100644 index 0000000000..1c604487fb --- /dev/null +++ b/src/libsystemd/src/sd-journal/journal-def.h @@ -0,0 +1,233 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-id128.h> + +#include "systemd-basic/macro.h" +#include "systemd-basic/sparse-endian.h" + +/* + * If you change this file you probably should also change its documentation: + * + * http://www.freedesktop.org/wiki/Software/systemd/journal-files + * + */ + +typedef struct Header Header; +typedef struct ObjectHeader ObjectHeader; +typedef union Object Object; +typedef struct DataObject DataObject; +typedef struct FieldObject FieldObject; +typedef struct EntryObject EntryObject; +typedef struct HashTableObject HashTableObject; +typedef struct EntryArrayObject EntryArrayObject; +typedef struct TagObject TagObject; +typedef struct EntryItem EntryItem; +typedef struct HashItem HashItem; +typedef struct FSSHeader FSSHeader; + +/* Object types */ +typedef enum ObjectType { + OBJECT_UNUSED, /* also serves as "any type" or "additional context" */ + OBJECT_DATA, + OBJECT_FIELD, + OBJECT_ENTRY, + OBJECT_DATA_HASH_TABLE, + OBJECT_FIELD_HASH_TABLE, + OBJECT_ENTRY_ARRAY, + OBJECT_TAG, + _OBJECT_TYPE_MAX +} ObjectType; + +/* Object flags */ +enum { + OBJECT_COMPRESSED_XZ = 1 << 0, + OBJECT_COMPRESSED_LZ4 = 1 << 1, + _OBJECT_COMPRESSED_MAX +}; + +#define OBJECT_COMPRESSION_MASK (OBJECT_COMPRESSED_XZ | OBJECT_COMPRESSED_LZ4) + +struct ObjectHeader { + uint8_t type; + uint8_t flags; + uint8_t reserved[6]; + le64_t size; + uint8_t payload[]; +} _packed_; + +struct DataObject { + ObjectHeader object; + le64_t hash; + le64_t next_hash_offset; + le64_t next_field_offset; + le64_t entry_offset; /* the first array entry we store inline */ + le64_t entry_array_offset; + le64_t n_entries; + uint8_t payload[]; +} _packed_; + +struct FieldObject { + ObjectHeader object; + le64_t hash; + le64_t next_hash_offset; + le64_t head_data_offset; + uint8_t payload[]; +} _packed_; + +struct EntryItem { + le64_t object_offset; + le64_t hash; +} _packed_; + +struct EntryObject { + ObjectHeader object; + le64_t seqnum; + le64_t realtime; + le64_t monotonic; + sd_id128_t boot_id; + le64_t xor_hash; + EntryItem items[]; +} _packed_; + +struct HashItem { + le64_t head_hash_offset; + le64_t tail_hash_offset; +} _packed_; + +struct HashTableObject { + ObjectHeader object; + HashItem items[]; +} _packed_; + +struct EntryArrayObject { + ObjectHeader object; + le64_t next_entry_array_offset; + le64_t items[]; +} _packed_; + +#define TAG_LENGTH (256/8) + +struct TagObject { + ObjectHeader object; + le64_t seqnum; + le64_t epoch; + uint8_t tag[TAG_LENGTH]; /* SHA-256 HMAC */ +} _packed_; + +union Object { + ObjectHeader object; + DataObject data; + FieldObject field; + EntryObject entry; + HashTableObject hash_table; + EntryArrayObject entry_array; + TagObject tag; +}; + +enum { + STATE_OFFLINE = 0, + STATE_ONLINE = 1, + STATE_ARCHIVED = 2, + _STATE_MAX +}; + +/* Header flags */ +enum { + HEADER_INCOMPATIBLE_COMPRESSED_XZ = 1 << 0, + HEADER_INCOMPATIBLE_COMPRESSED_LZ4 = 1 << 1, +}; + +#define HEADER_INCOMPATIBLE_ANY (HEADER_INCOMPATIBLE_COMPRESSED_XZ|HEADER_INCOMPATIBLE_COMPRESSED_LZ4) + +#if defined(HAVE_XZ) && defined(HAVE_LZ4) +# define HEADER_INCOMPATIBLE_SUPPORTED HEADER_INCOMPATIBLE_ANY +#elif defined(HAVE_XZ) +# define HEADER_INCOMPATIBLE_SUPPORTED HEADER_INCOMPATIBLE_COMPRESSED_XZ +#elif defined(HAVE_LZ4) +# define HEADER_INCOMPATIBLE_SUPPORTED HEADER_INCOMPATIBLE_COMPRESSED_LZ4 +#else +# define HEADER_INCOMPATIBLE_SUPPORTED 0 +#endif + +enum { + HEADER_COMPATIBLE_SEALED = 1 +}; + +#define HEADER_COMPATIBLE_ANY HEADER_COMPATIBLE_SEALED +#ifdef HAVE_GCRYPT +# define HEADER_COMPATIBLE_SUPPORTED HEADER_COMPATIBLE_SEALED +#else +# define HEADER_COMPATIBLE_SUPPORTED 0 +#endif + +#define HEADER_SIGNATURE ((char[]) { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' }) + +struct Header { + uint8_t signature[8]; /* "LPKSHHRH" */ + le32_t compatible_flags; + le32_t incompatible_flags; + uint8_t state; + uint8_t reserved[7]; + sd_id128_t file_id; + sd_id128_t machine_id; + sd_id128_t boot_id; /* last writer */ + sd_id128_t seqnum_id; + le64_t header_size; + le64_t arena_size; + le64_t data_hash_table_offset; + le64_t data_hash_table_size; + le64_t field_hash_table_offset; + le64_t field_hash_table_size; + le64_t tail_object_offset; + le64_t n_objects; + le64_t n_entries; + le64_t tail_entry_seqnum; + le64_t head_entry_seqnum; + le64_t entry_array_offset; + le64_t head_entry_realtime; + le64_t tail_entry_realtime; + le64_t tail_entry_monotonic; + /* Added in 187 */ + le64_t n_data; + le64_t n_fields; + /* Added in 189 */ + le64_t n_tags; + le64_t n_entry_arrays; + + /* Size: 240 */ +} _packed_; + +#define FSS_HEADER_SIGNATURE ((char[]) { 'K', 'S', 'H', 'H', 'R', 'H', 'L', 'P' }) + +struct FSSHeader { + uint8_t signature[8]; /* "KSHHRHLP" */ + le32_t compatible_flags; + le32_t incompatible_flags; + sd_id128_t machine_id; + sd_id128_t boot_id; /* last writer */ + le64_t header_size; + le64_t start_usec; + le64_t interval_usec; + le16_t fsprg_secpar; + le16_t reserved[3]; + le64_t fsprg_state_size; +} _packed_; diff --git a/src/libsystemd/src/sd-journal/journal-file.c b/src/libsystemd/src/sd-journal/journal-file.c new file mode 100644 index 0000000000..10972869c8 --- /dev/null +++ b/src/libsystemd/src/sd-journal/journal-file.c @@ -0,0 +1,3691 @@ +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <fcntl.h> +#include <pthread.h> +#include <stddef.h> +#include <sys/mman.h> +#include <sys/statvfs.h> +#include <sys/uio.h> +#include <unistd.h> + +#include <linux/fs.h> + +#include <systemd/sd-event.h> + +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/btrfs-util.h" +#include "systemd-basic/chattr-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/random-util.h" +#include "systemd-basic/set.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/xattr-util.h" + +#include "compress.h" +#include "journal-authenticate.h" +#include "journal-def.h" +#include "journal-file.h" +#include "lookup3.h" + +#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*sizeof(HashItem)) +#define DEFAULT_FIELD_HASH_TABLE_SIZE (333ULL*sizeof(HashItem)) + +#define COMPRESSION_SIZE_THRESHOLD (512ULL) + +/* This is the minimum journal file size */ +#define JOURNAL_FILE_SIZE_MIN (512ULL*1024ULL) /* 512 KiB */ + +/* These are the lower and upper bounds if we deduce the max_use value + * from the file system size */ +#define DEFAULT_MAX_USE_LOWER (1ULL*1024ULL*1024ULL) /* 1 MiB */ +#define DEFAULT_MAX_USE_UPPER (4ULL*1024ULL*1024ULL*1024ULL) /* 4 GiB */ + +/* This is the default minimal use limit, how much we'll use even if keep_free suggests otherwise. */ +#define DEFAULT_MIN_USE (1ULL*1024ULL*1024ULL) /* 1 MiB */ + +/* This is the upper bound if we deduce max_size from max_use */ +#define DEFAULT_MAX_SIZE_UPPER (128ULL*1024ULL*1024ULL) /* 128 MiB */ + +/* This is the upper bound if we deduce the keep_free value from the + * file system size */ +#define DEFAULT_KEEP_FREE_UPPER (4ULL*1024ULL*1024ULL*1024ULL) /* 4 GiB */ + +/* This is the keep_free value when we can't determine the system + * size */ +#define DEFAULT_KEEP_FREE (1024ULL*1024ULL) /* 1 MB */ + +/* This is the default maximum number of journal files to keep around. */ +#define DEFAULT_N_MAX_FILES (100) + +/* n_data was the first entry we added after the initial file format design */ +#define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data)) + +/* How many entries to keep in the entry array chain cache at max */ +#define CHAIN_CACHE_MAX 20 + +/* How much to increase the journal file size at once each time we allocate something new. */ +#define FILE_SIZE_INCREASE (8ULL*1024ULL*1024ULL) /* 8MB */ + +/* Reread fstat() of the file for detecting deletions at least this often */ +#define LAST_STAT_REFRESH_USEC (5*USEC_PER_SEC) + +/* The mmap context to use for the header we pick as one above the last defined typed */ +#define CONTEXT_HEADER _OBJECT_TYPE_MAX + +/* This may be called from a separate thread to prevent blocking the caller for the duration of fsync(). + * As a result we use atomic operations on f->offline_state for inter-thread communications with + * journal_file_set_offline() and journal_file_set_online(). */ +static void journal_file_set_offline_internal(JournalFile *f) { + assert(f); + assert(f->fd >= 0); + assert(f->header); + + for (;;) { + switch (f->offline_state) { + case OFFLINE_CANCEL: + if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_CANCEL, OFFLINE_DONE)) + continue; + return; + + case OFFLINE_AGAIN_FROM_SYNCING: + if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_SYNCING, OFFLINE_SYNCING)) + continue; + break; + + case OFFLINE_AGAIN_FROM_OFFLINING: + if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_OFFLINING, OFFLINE_SYNCING)) + continue; + break; + + case OFFLINE_SYNCING: + (void) fsync(f->fd); + + if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_OFFLINING)) + continue; + + f->header->state = f->archive ? STATE_ARCHIVED : STATE_OFFLINE; + (void) fsync(f->fd); + break; + + case OFFLINE_OFFLINING: + if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_OFFLINING, OFFLINE_DONE)) + continue; + /* fall through */ + + case OFFLINE_DONE: + return; + + case OFFLINE_JOINED: + log_debug("OFFLINE_JOINED unexpected offline state for journal_file_set_offline_internal()"); + return; + } + } +} + +static void * journal_file_set_offline_thread(void *arg) { + JournalFile *f = arg; + + journal_file_set_offline_internal(f); + + return NULL; +} + +static int journal_file_set_offline_thread_join(JournalFile *f) { + int r; + + assert(f); + + if (f->offline_state == OFFLINE_JOINED) + return 0; + + r = pthread_join(f->offline_thread, NULL); + if (r) + return -r; + + f->offline_state = OFFLINE_JOINED; + + if (mmap_cache_got_sigbus(f->mmap, f->fd)) + return -EIO; + + return 0; +} + +/* Trigger a restart if the offline thread is mid-flight in a restartable state. */ +static bool journal_file_set_offline_try_restart(JournalFile *f) { + for (;;) { + switch (f->offline_state) { + case OFFLINE_AGAIN_FROM_SYNCING: + case OFFLINE_AGAIN_FROM_OFFLINING: + return true; + + case OFFLINE_CANCEL: + if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_CANCEL, OFFLINE_AGAIN_FROM_SYNCING)) + continue; + return true; + + case OFFLINE_SYNCING: + if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_AGAIN_FROM_SYNCING)) + continue; + return true; + + case OFFLINE_OFFLINING: + if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_OFFLINING, OFFLINE_AGAIN_FROM_OFFLINING)) + continue; + return true; + + default: + return false; + } + } +} + +/* Sets a journal offline. + * + * If wait is false then an offline is dispatched in a separate thread for a + * subsequent journal_file_set_offline() or journal_file_set_online() of the + * same journal to synchronize with. + * + * If wait is true, then either an existing offline thread will be restarted + * and joined, or if none exists the offline is simply performed in this + * context without involving another thread. + */ +int journal_file_set_offline(JournalFile *f, bool wait) { + bool restarted; + int r; + + assert(f); + + if (!f->writable) + return -EPERM; + + if (!(f->fd >= 0 && f->header)) + return -EINVAL; + + /* An offlining journal is implicitly online and may modify f->header->state, + * we must also join any potentially lingering offline thread when not online. */ + if (!journal_file_is_offlining(f) && f->header->state != STATE_ONLINE) + return journal_file_set_offline_thread_join(f); + + /* Restart an in-flight offline thread and wait if needed, or join a lingering done one. */ + restarted = journal_file_set_offline_try_restart(f); + if ((restarted && wait) || !restarted) { + r = journal_file_set_offline_thread_join(f); + if (r < 0) + return r; + } + + if (restarted) + return 0; + + /* Initiate a new offline. */ + f->offline_state = OFFLINE_SYNCING; + + if (wait) /* Without using a thread if waiting. */ + journal_file_set_offline_internal(f); + else { + r = pthread_create(&f->offline_thread, NULL, journal_file_set_offline_thread, f); + if (r > 0) { + f->offline_state = OFFLINE_JOINED; + return -r; + } + } + + return 0; +} + +static int journal_file_set_online(JournalFile *f) { + bool joined = false; + + assert(f); + + if (!f->writable) + return -EPERM; + + if (!(f->fd >= 0 && f->header)) + return -EINVAL; + + while (!joined) { + switch (f->offline_state) { + case OFFLINE_JOINED: + /* No offline thread, no need to wait. */ + joined = true; + break; + + case OFFLINE_SYNCING: + if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_CANCEL)) + continue; + /* Canceled syncing prior to offlining, no need to wait. */ + break; + + case OFFLINE_AGAIN_FROM_SYNCING: + if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_SYNCING, OFFLINE_CANCEL)) + continue; + /* Canceled restart from syncing, no need to wait. */ + break; + + case OFFLINE_AGAIN_FROM_OFFLINING: + if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_OFFLINING, OFFLINE_CANCEL)) + continue; + /* Canceled restart from offlining, must wait for offlining to complete however. */ + + /* fall through to wait */ + default: { + int r; + + r = journal_file_set_offline_thread_join(f); + if (r < 0) + return r; + + joined = true; + break; + } + } + } + + if (mmap_cache_got_sigbus(f->mmap, f->fd)) + return -EIO; + + switch (f->header->state) { + case STATE_ONLINE: + return 0; + + case STATE_OFFLINE: + f->header->state = STATE_ONLINE; + (void) fsync(f->fd); + return 0; + + default: + return -EINVAL; + } +} + +bool journal_file_is_offlining(JournalFile *f) { + assert(f); + + __sync_synchronize(); + + if (f->offline_state == OFFLINE_DONE || + f->offline_state == OFFLINE_JOINED) + return false; + + return true; +} + +JournalFile* journal_file_close(JournalFile *f) { + assert(f); + +#ifdef HAVE_GCRYPT + /* Write the final tag */ + if (f->seal && f->writable) { + int r; + + r = journal_file_append_tag(f); + if (r < 0) + log_error_errno(r, "Failed to append tag when closing journal: %m"); + } +#endif + + if (f->post_change_timer) { + int enabled; + + if (sd_event_source_get_enabled(f->post_change_timer, &enabled) >= 0) + if (enabled == SD_EVENT_ONESHOT) + journal_file_post_change(f); + + (void) sd_event_source_set_enabled(f->post_change_timer, SD_EVENT_OFF); + sd_event_source_unref(f->post_change_timer); + } + + journal_file_set_offline(f, true); + + if (f->mmap && f->fd >= 0) + mmap_cache_close_fd(f->mmap, f->fd); + + if (f->fd >= 0 && f->defrag_on_close) { + + /* Be friendly to btrfs: turn COW back on again now, + * and defragment the file. We won't write to the file + * ever again, hence remove all fragmentation, and + * reenable all the good bits COW usually provides + * (such as data checksumming). */ + + (void) chattr_fd(f->fd, 0, FS_NOCOW_FL); + (void) btrfs_defrag_fd(f->fd); + } + + if (f->close_fd) + safe_close(f->fd); + free(f->path); + + mmap_cache_unref(f->mmap); + + ordered_hashmap_free_free(f->chain_cache); + +#if defined(HAVE_XZ) || defined(HAVE_LZ4) + free(f->compress_buffer); +#endif + +#ifdef HAVE_GCRYPT + if (f->fss_file) + munmap(f->fss_file, PAGE_ALIGN(f->fss_file_size)); + else + free(f->fsprg_state); + + free(f->fsprg_seed); + + if (f->hmac) + gcry_md_close(f->hmac); +#endif + + return mfree(f); +} + +void journal_file_close_set(Set *s) { + JournalFile *f; + + assert(s); + + while ((f = set_steal_first(s))) + (void) journal_file_close(f); +} + +static int journal_file_init_header(JournalFile *f, JournalFile *template) { + Header h = {}; + ssize_t k; + int r; + + assert(f); + + memcpy(h.signature, HEADER_SIGNATURE, 8); + h.header_size = htole64(ALIGN64(sizeof(h))); + + h.incompatible_flags |= htole32( + f->compress_xz * HEADER_INCOMPATIBLE_COMPRESSED_XZ | + f->compress_lz4 * HEADER_INCOMPATIBLE_COMPRESSED_LZ4); + + h.compatible_flags = htole32( + f->seal * HEADER_COMPATIBLE_SEALED); + + r = sd_id128_randomize(&h.file_id); + if (r < 0) + return r; + + if (template) { + h.seqnum_id = template->header->seqnum_id; + h.tail_entry_seqnum = template->header->tail_entry_seqnum; + } else + h.seqnum_id = h.file_id; + + k = pwrite(f->fd, &h, sizeof(h), 0); + if (k < 0) + return -errno; + + if (k != sizeof(h)) + return -EIO; + + return 0; +} + +static int fsync_directory_of_file(int fd) { + _cleanup_free_ char *path = NULL, *dn = NULL; + _cleanup_close_ int dfd = -1; + struct stat st; + int r; + + if (fstat(fd, &st) < 0) + return -errno; + + if (!S_ISREG(st.st_mode)) + return -EBADFD; + + r = fd_get_path(fd, &path); + if (r < 0) + return r; + + if (!path_is_absolute(path)) + return -EINVAL; + + dn = dirname_malloc(path); + if (!dn) + return -ENOMEM; + + dfd = open(dn, O_RDONLY|O_CLOEXEC|O_DIRECTORY); + if (dfd < 0) + return -errno; + + if (fsync(dfd) < 0) + return -errno; + + return 0; +} + +static int journal_file_refresh_header(JournalFile *f) { + sd_id128_t boot_id; + int r; + + assert(f); + assert(f->header); + + r = sd_id128_get_machine(&f->header->machine_id); + if (r < 0) + return r; + + r = sd_id128_get_boot(&boot_id); + if (r < 0) + return r; + + if (sd_id128_equal(boot_id, f->header->boot_id)) + f->tail_entry_monotonic_valid = true; + + f->header->boot_id = boot_id; + + r = journal_file_set_online(f); + + /* Sync the online state to disk */ + (void) fsync(f->fd); + + /* We likely just created a new file, also sync the directory this file is located in. */ + (void) fsync_directory_of_file(f->fd); + + return r; +} + +static int journal_file_verify_header(JournalFile *f) { + uint32_t flags; + + assert(f); + assert(f->header); + + if (memcmp(f->header->signature, HEADER_SIGNATURE, 8)) + return -EBADMSG; + + /* In both read and write mode we refuse to open files with + * incompatible flags we don't know */ + flags = le32toh(f->header->incompatible_flags); + if (flags & ~HEADER_INCOMPATIBLE_SUPPORTED) { + if (flags & ~HEADER_INCOMPATIBLE_ANY) + log_debug("Journal file %s has unknown incompatible flags %"PRIx32, + f->path, flags & ~HEADER_INCOMPATIBLE_ANY); + flags = (flags & HEADER_INCOMPATIBLE_ANY) & ~HEADER_INCOMPATIBLE_SUPPORTED; + if (flags) + log_debug("Journal file %s uses incompatible flags %"PRIx32 + " disabled at compilation time.", f->path, flags); + return -EPROTONOSUPPORT; + } + + /* When open for writing we refuse to open files with + * compatible flags, too */ + flags = le32toh(f->header->compatible_flags); + if (f->writable && (flags & ~HEADER_COMPATIBLE_SUPPORTED)) { + if (flags & ~HEADER_COMPATIBLE_ANY) + log_debug("Journal file %s has unknown compatible flags %"PRIx32, + f->path, flags & ~HEADER_COMPATIBLE_ANY); + flags = (flags & HEADER_COMPATIBLE_ANY) & ~HEADER_COMPATIBLE_SUPPORTED; + if (flags) + log_debug("Journal file %s uses compatible flags %"PRIx32 + " disabled at compilation time.", f->path, flags); + return -EPROTONOSUPPORT; + } + + if (f->header->state >= _STATE_MAX) + return -EBADMSG; + + /* The first addition was n_data, so check that we are at least this large */ + if (le64toh(f->header->header_size) < HEADER_SIZE_MIN) + return -EBADMSG; + + if (JOURNAL_HEADER_SEALED(f->header) && !JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) + return -EBADMSG; + + if ((le64toh(f->header->header_size) + le64toh(f->header->arena_size)) > (uint64_t) f->last_stat.st_size) + return -ENODATA; + + if (le64toh(f->header->tail_object_offset) > (le64toh(f->header->header_size) + le64toh(f->header->arena_size))) + return -ENODATA; + + if (!VALID64(le64toh(f->header->data_hash_table_offset)) || + !VALID64(le64toh(f->header->field_hash_table_offset)) || + !VALID64(le64toh(f->header->tail_object_offset)) || + !VALID64(le64toh(f->header->entry_array_offset))) + return -ENODATA; + + if (f->writable) { + sd_id128_t machine_id; + uint8_t state; + int r; + + r = sd_id128_get_machine(&machine_id); + if (r < 0) + return r; + + if (!sd_id128_equal(machine_id, f->header->machine_id)) + return -EHOSTDOWN; + + state = f->header->state; + + if (state == STATE_ONLINE) { + log_debug("Journal file %s is already online. Assuming unclean closing.", f->path); + return -EBUSY; + } else if (state == STATE_ARCHIVED) + return -ESHUTDOWN; + else if (state != STATE_OFFLINE) { + log_debug("Journal file %s has unknown state %i.", f->path, state); + return -EBUSY; + } + + /* Don't permit appending to files from the future. Because otherwise the realtime timestamps wouldn't + * be strictly ordered in the entries in the file anymore, and we can't have that since it breaks + * bisection. */ + if (le64toh(f->header->tail_entry_realtime) > now(CLOCK_REALTIME)) { + log_debug("Journal file %s is from the future, refusing to append new data to it that'd be older.", f->path); + return -ETXTBSY; + } + } + + f->compress_xz = JOURNAL_HEADER_COMPRESSED_XZ(f->header); + f->compress_lz4 = JOURNAL_HEADER_COMPRESSED_LZ4(f->header); + + f->seal = JOURNAL_HEADER_SEALED(f->header); + + return 0; +} + +static int journal_file_fstat(JournalFile *f) { + assert(f); + assert(f->fd >= 0); + + if (fstat(f->fd, &f->last_stat) < 0) + return -errno; + + f->last_stat_usec = now(CLOCK_MONOTONIC); + + /* Refuse appending to files that are already deleted */ + if (f->last_stat.st_nlink <= 0) + return -EIDRM; + + return 0; +} + +static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) { + uint64_t old_size, new_size; + int r; + + assert(f); + assert(f->header); + + /* We assume that this file is not sparse, and we know that + * for sure, since we always call posix_fallocate() + * ourselves */ + + if (mmap_cache_got_sigbus(f->mmap, f->fd)) + return -EIO; + + old_size = + le64toh(f->header->header_size) + + le64toh(f->header->arena_size); + + new_size = PAGE_ALIGN(offset + size); + if (new_size < le64toh(f->header->header_size)) + new_size = le64toh(f->header->header_size); + + if (new_size <= old_size) { + + /* We already pre-allocated enough space, but before + * we write to it, let's check with fstat() if the + * file got deleted, in order make sure we don't throw + * away the data immediately. Don't check fstat() for + * all writes though, but only once ever 10s. */ + + if (f->last_stat_usec + LAST_STAT_REFRESH_USEC > now(CLOCK_MONOTONIC)) + return 0; + + return journal_file_fstat(f); + } + + /* Allocate more space. */ + + if (f->metrics.max_size > 0 && new_size > f->metrics.max_size) + return -E2BIG; + + if (new_size > f->metrics.min_size && f->metrics.keep_free > 0) { + struct statvfs svfs; + + if (fstatvfs(f->fd, &svfs) >= 0) { + uint64_t available; + + available = LESS_BY((uint64_t) svfs.f_bfree * (uint64_t) svfs.f_bsize, f->metrics.keep_free); + + if (new_size - old_size > available) + return -E2BIG; + } + } + + /* Increase by larger blocks at once */ + new_size = ((new_size+FILE_SIZE_INCREASE-1) / FILE_SIZE_INCREASE) * FILE_SIZE_INCREASE; + if (f->metrics.max_size > 0 && new_size > f->metrics.max_size) + new_size = f->metrics.max_size; + + /* Note that the glibc fallocate() fallback is very + inefficient, hence we try to minimize the allocation area + as we can. */ + r = posix_fallocate(f->fd, old_size, new_size - old_size); + if (r != 0) + return -r; + + f->header->arena_size = htole64(new_size - le64toh(f->header->header_size)); + + return journal_file_fstat(f); +} + +static unsigned type_to_context(ObjectType type) { + /* One context for each type, plus one catch-all for the rest */ + assert_cc(_OBJECT_TYPE_MAX <= MMAP_CACHE_MAX_CONTEXTS); + assert_cc(CONTEXT_HEADER < MMAP_CACHE_MAX_CONTEXTS); + return type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX ? type : 0; +} + +static int journal_file_move_to(JournalFile *f, ObjectType type, bool keep_always, uint64_t offset, uint64_t size, void **ret) { + int r; + + assert(f); + assert(ret); + + if (size <= 0) + return -EINVAL; + + /* Avoid SIGBUS on invalid accesses */ + if (offset + size > (uint64_t) f->last_stat.st_size) { + /* Hmm, out of range? Let's refresh the fstat() data + * first, before we trust that check. */ + + r = journal_file_fstat(f); + if (r < 0) + return r; + + if (offset + size > (uint64_t) f->last_stat.st_size) + return -EADDRNOTAVAIL; + } + + return mmap_cache_get(f->mmap, f->fd, f->prot, type_to_context(type), keep_always, offset, size, &f->last_stat, ret); +} + +static uint64_t minimum_header_size(Object *o) { + + static const uint64_t table[] = { + [OBJECT_DATA] = sizeof(DataObject), + [OBJECT_FIELD] = sizeof(FieldObject), + [OBJECT_ENTRY] = sizeof(EntryObject), + [OBJECT_DATA_HASH_TABLE] = sizeof(HashTableObject), + [OBJECT_FIELD_HASH_TABLE] = sizeof(HashTableObject), + [OBJECT_ENTRY_ARRAY] = sizeof(EntryArrayObject), + [OBJECT_TAG] = sizeof(TagObject), + }; + + if (o->object.type >= ELEMENTSOF(table) || table[o->object.type] <= 0) + return sizeof(ObjectHeader); + + return table[o->object.type]; +} + +int journal_file_move_to_object(JournalFile *f, ObjectType type, uint64_t offset, Object **ret) { + int r; + void *t; + Object *o; + uint64_t s; + + assert(f); + assert(ret); + + /* Objects may only be located at multiple of 64 bit */ + if (!VALID64(offset)) { + log_debug("Attempt to move to object at non-64bit boundary: %" PRIu64, offset); + return -EBADMSG; + } + + /* Object may not be located in the file header */ + if (offset < le64toh(f->header->header_size)) { + log_debug("Attempt to move to object located in file header: %" PRIu64, offset); + return -EBADMSG; + } + + r = journal_file_move_to(f, type, false, offset, sizeof(ObjectHeader), &t); + if (r < 0) + return r; + + o = (Object*) t; + s = le64toh(o->object.size); + + if (s == 0) { + log_debug("Attempt to move to uninitialized object: %" PRIu64, offset); + return -EBADMSG; + } + if (s < sizeof(ObjectHeader)) { + log_debug("Attempt to move to overly short object: %" PRIu64, offset); + return -EBADMSG; + } + + if (o->object.type <= OBJECT_UNUSED) { + log_debug("Attempt to move to object with invalid type: %" PRIu64, offset); + return -EBADMSG; + } + + if (s < minimum_header_size(o)) { + log_debug("Attempt to move to truncated object: %" PRIu64, offset); + return -EBADMSG; + } + + if (type > OBJECT_UNUSED && o->object.type != type) { + log_debug("Attempt to move to object of unexpected type: %" PRIu64, offset); + return -EBADMSG; + } + + if (s > sizeof(ObjectHeader)) { + r = journal_file_move_to(f, type, false, offset, s, &t); + if (r < 0) + return r; + + o = (Object*) t; + } + + *ret = o; + return 0; +} + +static uint64_t journal_file_entry_seqnum(JournalFile *f, uint64_t *seqnum) { + uint64_t r; + + assert(f); + assert(f->header); + + r = le64toh(f->header->tail_entry_seqnum) + 1; + + if (seqnum) { + /* If an external seqnum counter was passed, we update + * both the local and the external one, and set it to + * the maximum of both */ + + if (*seqnum + 1 > r) + r = *seqnum + 1; + + *seqnum = r; + } + + f->header->tail_entry_seqnum = htole64(r); + + if (f->header->head_entry_seqnum == 0) + f->header->head_entry_seqnum = htole64(r); + + return r; +} + +int journal_file_append_object(JournalFile *f, ObjectType type, uint64_t size, Object **ret, uint64_t *offset) { + int r; + uint64_t p; + Object *tail, *o; + void *t; + + assert(f); + assert(f->header); + assert(type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX); + assert(size >= sizeof(ObjectHeader)); + assert(offset); + assert(ret); + + r = journal_file_set_online(f); + if (r < 0) + return r; + + p = le64toh(f->header->tail_object_offset); + if (p == 0) + p = le64toh(f->header->header_size); + else { + r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &tail); + if (r < 0) + return r; + + p += ALIGN64(le64toh(tail->object.size)); + } + + r = journal_file_allocate(f, p, size); + if (r < 0) + return r; + + r = journal_file_move_to(f, type, false, p, size, &t); + if (r < 0) + return r; + + o = (Object*) t; + + zero(o->object); + o->object.type = type; + o->object.size = htole64(size); + + f->header->tail_object_offset = htole64(p); + f->header->n_objects = htole64(le64toh(f->header->n_objects) + 1); + + *ret = o; + *offset = p; + + return 0; +} + +static int journal_file_setup_data_hash_table(JournalFile *f) { + uint64_t s, p; + Object *o; + int r; + + assert(f); + assert(f->header); + + /* We estimate that we need 1 hash table entry per 768 bytes + of journal file and we want to make sure we never get + beyond 75% fill level. Calculate the hash table size for + the maximum file size based on these metrics. */ + + s = (f->metrics.max_size * 4 / 768 / 3) * sizeof(HashItem); + if (s < DEFAULT_DATA_HASH_TABLE_SIZE) + s = DEFAULT_DATA_HASH_TABLE_SIZE; + + log_debug("Reserving %"PRIu64" entries in hash table.", s / sizeof(HashItem)); + + r = journal_file_append_object(f, + OBJECT_DATA_HASH_TABLE, + offsetof(Object, hash_table.items) + s, + &o, &p); + if (r < 0) + return r; + + memzero(o->hash_table.items, s); + + f->header->data_hash_table_offset = htole64(p + offsetof(Object, hash_table.items)); + f->header->data_hash_table_size = htole64(s); + + return 0; +} + +static int journal_file_setup_field_hash_table(JournalFile *f) { + uint64_t s, p; + Object *o; + int r; + + assert(f); + assert(f->header); + + /* We use a fixed size hash table for the fields as this + * number should grow very slowly only */ + + s = DEFAULT_FIELD_HASH_TABLE_SIZE; + r = journal_file_append_object(f, + OBJECT_FIELD_HASH_TABLE, + offsetof(Object, hash_table.items) + s, + &o, &p); + if (r < 0) + return r; + + memzero(o->hash_table.items, s); + + f->header->field_hash_table_offset = htole64(p + offsetof(Object, hash_table.items)); + f->header->field_hash_table_size = htole64(s); + + return 0; +} + +int journal_file_map_data_hash_table(JournalFile *f) { + uint64_t s, p; + void *t; + int r; + + assert(f); + assert(f->header); + + if (f->data_hash_table) + return 0; + + p = le64toh(f->header->data_hash_table_offset); + s = le64toh(f->header->data_hash_table_size); + + r = journal_file_move_to(f, + OBJECT_DATA_HASH_TABLE, + true, + p, s, + &t); + if (r < 0) + return r; + + f->data_hash_table = t; + return 0; +} + +int journal_file_map_field_hash_table(JournalFile *f) { + uint64_t s, p; + void *t; + int r; + + assert(f); + assert(f->header); + + if (f->field_hash_table) + return 0; + + p = le64toh(f->header->field_hash_table_offset); + s = le64toh(f->header->field_hash_table_size); + + r = journal_file_move_to(f, + OBJECT_FIELD_HASH_TABLE, + true, + p, s, + &t); + if (r < 0) + return r; + + f->field_hash_table = t; + return 0; +} + +static int journal_file_link_field( + JournalFile *f, + Object *o, + uint64_t offset, + uint64_t hash) { + + uint64_t p, h, m; + int r; + + assert(f); + assert(f->header); + assert(f->field_hash_table); + assert(o); + assert(offset > 0); + + if (o->object.type != OBJECT_FIELD) + return -EINVAL; + + m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem); + if (m <= 0) + return -EBADMSG; + + /* This might alter the window we are looking at */ + o->field.next_hash_offset = o->field.head_data_offset = 0; + + h = hash % m; + p = le64toh(f->field_hash_table[h].tail_hash_offset); + if (p == 0) + f->field_hash_table[h].head_hash_offset = htole64(offset); + else { + r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); + if (r < 0) + return r; + + o->field.next_hash_offset = htole64(offset); + } + + f->field_hash_table[h].tail_hash_offset = htole64(offset); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) + f->header->n_fields = htole64(le64toh(f->header->n_fields) + 1); + + return 0; +} + +static int journal_file_link_data( + JournalFile *f, + Object *o, + uint64_t offset, + uint64_t hash) { + + uint64_t p, h, m; + int r; + + assert(f); + assert(f->header); + assert(f->data_hash_table); + assert(o); + assert(offset > 0); + + if (o->object.type != OBJECT_DATA) + return -EINVAL; + + m = le64toh(f->header->data_hash_table_size) / sizeof(HashItem); + if (m <= 0) + return -EBADMSG; + + /* This might alter the window we are looking at */ + o->data.next_hash_offset = o->data.next_field_offset = 0; + o->data.entry_offset = o->data.entry_array_offset = 0; + o->data.n_entries = 0; + + h = hash % m; + p = le64toh(f->data_hash_table[h].tail_hash_offset); + if (p == 0) + /* Only entry in the hash table is easy */ + f->data_hash_table[h].head_hash_offset = htole64(offset); + else { + /* Move back to the previous data object, to patch in + * pointer */ + + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + o->data.next_hash_offset = htole64(offset); + } + + f->data_hash_table[h].tail_hash_offset = htole64(offset); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) + f->header->n_data = htole64(le64toh(f->header->n_data) + 1); + + return 0; +} + +int journal_file_find_field_object_with_hash( + JournalFile *f, + const void *field, uint64_t size, uint64_t hash, + Object **ret, uint64_t *offset) { + + uint64_t p, osize, h, m; + int r; + + assert(f); + assert(f->header); + assert(field && size > 0); + + /* If the field hash table is empty, we can't find anything */ + if (le64toh(f->header->field_hash_table_size) <= 0) + return 0; + + /* Map the field hash table, if it isn't mapped yet. */ + r = journal_file_map_field_hash_table(f); + if (r < 0) + return r; + + osize = offsetof(Object, field.payload) + size; + + m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem); + if (m <= 0) + return -EBADMSG; + + h = hash % m; + p = le64toh(f->field_hash_table[h].head_hash_offset); + + while (p > 0) { + Object *o; + + r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); + if (r < 0) + return r; + + if (le64toh(o->field.hash) == hash && + le64toh(o->object.size) == osize && + memcmp(o->field.payload, field, size) == 0) { + + if (ret) + *ret = o; + if (offset) + *offset = p; + + return 1; + } + + p = le64toh(o->field.next_hash_offset); + } + + return 0; +} + +int journal_file_find_field_object( + JournalFile *f, + const void *field, uint64_t size, + Object **ret, uint64_t *offset) { + + uint64_t hash; + + assert(f); + assert(field && size > 0); + + hash = hash64(field, size); + + return journal_file_find_field_object_with_hash(f, + field, size, hash, + ret, offset); +} + +int journal_file_find_data_object_with_hash( + JournalFile *f, + const void *data, uint64_t size, uint64_t hash, + Object **ret, uint64_t *offset) { + + uint64_t p, osize, h, m; + int r; + + assert(f); + assert(f->header); + assert(data || size == 0); + + /* If there's no data hash table, then there's no entry. */ + if (le64toh(f->header->data_hash_table_size) <= 0) + return 0; + + /* Map the data hash table, if it isn't mapped yet. */ + r = journal_file_map_data_hash_table(f); + if (r < 0) + return r; + + osize = offsetof(Object, data.payload) + size; + + m = le64toh(f->header->data_hash_table_size) / sizeof(HashItem); + if (m <= 0) + return -EBADMSG; + + h = hash % m; + p = le64toh(f->data_hash_table[h].head_hash_offset); + + while (p > 0) { + Object *o; + + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + if (le64toh(o->data.hash) != hash) + goto next; + + if (o->object.flags & OBJECT_COMPRESSION_MASK) { +#if defined(HAVE_XZ) || defined(HAVE_LZ4) + uint64_t l; + size_t rsize = 0; + + l = le64toh(o->object.size); + if (l <= offsetof(Object, data.payload)) + return -EBADMSG; + + l -= offsetof(Object, data.payload); + + r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK, + o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize, 0); + if (r < 0) + return r; + + if (rsize == size && + memcmp(f->compress_buffer, data, size) == 0) { + + if (ret) + *ret = o; + + if (offset) + *offset = p; + + return 1; + } +#else + return -EPROTONOSUPPORT; +#endif + } else if (le64toh(o->object.size) == osize && + memcmp(o->data.payload, data, size) == 0) { + + if (ret) + *ret = o; + + if (offset) + *offset = p; + + return 1; + } + + next: + p = le64toh(o->data.next_hash_offset); + } + + return 0; +} + +int journal_file_find_data_object( + JournalFile *f, + const void *data, uint64_t size, + Object **ret, uint64_t *offset) { + + uint64_t hash; + + assert(f); + assert(data || size == 0); + + hash = hash64(data, size); + + return journal_file_find_data_object_with_hash(f, + data, size, hash, + ret, offset); +} + +static int journal_file_append_field( + JournalFile *f, + const void *field, uint64_t size, + Object **ret, uint64_t *offset) { + + uint64_t hash, p; + uint64_t osize; + Object *o; + int r; + + assert(f); + assert(field && size > 0); + + hash = hash64(field, size); + + r = journal_file_find_field_object_with_hash(f, field, size, hash, &o, &p); + if (r < 0) + return r; + else if (r > 0) { + + if (ret) + *ret = o; + + if (offset) + *offset = p; + + return 0; + } + + osize = offsetof(Object, field.payload) + size; + r = journal_file_append_object(f, OBJECT_FIELD, osize, &o, &p); + if (r < 0) + return r; + + o->field.hash = htole64(hash); + memcpy(o->field.payload, field, size); + + r = journal_file_link_field(f, o, p, hash); + if (r < 0) + return r; + + /* The linking might have altered the window, so let's + * refresh our pointer */ + r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); + if (r < 0) + return r; + +#ifdef HAVE_GCRYPT + r = journal_file_hmac_put_object(f, OBJECT_FIELD, o, p); + if (r < 0) + return r; +#endif + + if (ret) + *ret = o; + + if (offset) + *offset = p; + + return 0; +} + +static int journal_file_append_data( + JournalFile *f, + const void *data, uint64_t size, + Object **ret, uint64_t *offset) { + + uint64_t hash, p; + uint64_t osize; + Object *o; + int r, compression = 0; + const void *eq; + + assert(f); + assert(data || size == 0); + + hash = hash64(data, size); + + r = journal_file_find_data_object_with_hash(f, data, size, hash, &o, &p); + if (r < 0) + return r; + if (r > 0) { + + if (ret) + *ret = o; + + if (offset) + *offset = p; + + return 0; + } + + osize = offsetof(Object, data.payload) + size; + r = journal_file_append_object(f, OBJECT_DATA, osize, &o, &p); + if (r < 0) + return r; + + o->data.hash = htole64(hash); + +#if defined(HAVE_XZ) || defined(HAVE_LZ4) + if (JOURNAL_FILE_COMPRESS(f) && size >= COMPRESSION_SIZE_THRESHOLD) { + size_t rsize = 0; + + compression = compress_blob(data, size, o->data.payload, size - 1, &rsize); + + if (compression >= 0) { + o->object.size = htole64(offsetof(Object, data.payload) + rsize); + o->object.flags |= compression; + + log_debug("Compressed data object %"PRIu64" -> %zu using %s", + size, rsize, object_compressed_to_string(compression)); + } else + /* Compression didn't work, we don't really care why, let's continue without compression */ + compression = 0; + } +#endif + + if (compression == 0) + memcpy_safe(o->data.payload, data, size); + + r = journal_file_link_data(f, o, p, hash); + if (r < 0) + return r; + +#ifdef HAVE_GCRYPT + r = journal_file_hmac_put_object(f, OBJECT_DATA, o, p); + if (r < 0) + return r; +#endif + + /* The linking might have altered the window, so let's + * refresh our pointer */ + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + if (!data) + eq = NULL; + else + eq = memchr(data, '=', size); + if (eq && eq > data) { + Object *fo = NULL; + uint64_t fp; + + /* Create field object ... */ + r = journal_file_append_field(f, data, (uint8_t*) eq - (uint8_t*) data, &fo, &fp); + if (r < 0) + return r; + + /* ... and link it in. */ + o->data.next_field_offset = fo->field.head_data_offset; + fo->field.head_data_offset = le64toh(p); + } + + if (ret) + *ret = o; + + if (offset) + *offset = p; + + return 0; +} + +uint64_t journal_file_entry_n_items(Object *o) { + assert(o); + + if (o->object.type != OBJECT_ENTRY) + return 0; + + return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem); +} + +uint64_t journal_file_entry_array_n_items(Object *o) { + assert(o); + + if (o->object.type != OBJECT_ENTRY_ARRAY) + return 0; + + return (le64toh(o->object.size) - offsetof(Object, entry_array.items)) / sizeof(uint64_t); +} + +uint64_t journal_file_hash_table_n_items(Object *o) { + assert(o); + + if (o->object.type != OBJECT_DATA_HASH_TABLE && + o->object.type != OBJECT_FIELD_HASH_TABLE) + return 0; + + return (le64toh(o->object.size) - offsetof(Object, hash_table.items)) / sizeof(HashItem); +} + +static int link_entry_into_array(JournalFile *f, + le64_t *first, + le64_t *idx, + uint64_t p) { + int r; + uint64_t n = 0, ap = 0, q, i, a, hidx; + Object *o; + + assert(f); + assert(f->header); + assert(first); + assert(idx); + assert(p > 0); + + a = le64toh(*first); + i = hidx = le64toh(*idx); + while (a > 0) { + + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + + n = journal_file_entry_array_n_items(o); + if (i < n) { + o->entry_array.items[i] = htole64(p); + *idx = htole64(hidx + 1); + return 0; + } + + i -= n; + ap = a; + a = le64toh(o->entry_array.next_entry_array_offset); + } + + if (hidx > n) + n = (hidx+1) * 2; + else + n = n * 2; + + if (n < 4) + n = 4; + + r = journal_file_append_object(f, OBJECT_ENTRY_ARRAY, + offsetof(Object, entry_array.items) + n * sizeof(uint64_t), + &o, &q); + if (r < 0) + return r; + +#ifdef HAVE_GCRYPT + r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, o, q); + if (r < 0) + return r; +#endif + + o->entry_array.items[i] = htole64(p); + + if (ap == 0) + *first = htole64(q); + else { + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, ap, &o); + if (r < 0) + return r; + + o->entry_array.next_entry_array_offset = htole64(q); + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) + f->header->n_entry_arrays = htole64(le64toh(f->header->n_entry_arrays) + 1); + + *idx = htole64(hidx + 1); + + return 0; +} + +static int link_entry_into_array_plus_one(JournalFile *f, + le64_t *extra, + le64_t *first, + le64_t *idx, + uint64_t p) { + + int r; + + assert(f); + assert(extra); + assert(first); + assert(idx); + assert(p > 0); + + if (*idx == 0) + *extra = htole64(p); + else { + le64_t i; + + i = htole64(le64toh(*idx) - 1); + r = link_entry_into_array(f, first, &i, p); + if (r < 0) + return r; + } + + *idx = htole64(le64toh(*idx) + 1); + return 0; +} + +static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) { + uint64_t p; + int r; + assert(f); + assert(o); + assert(offset > 0); + + p = le64toh(o->entry.items[i].object_offset); + if (p == 0) + return -EINVAL; + + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + return link_entry_into_array_plus_one(f, + &o->data.entry_offset, + &o->data.entry_array_offset, + &o->data.n_entries, + offset); +} + +static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) { + uint64_t n, i; + int r; + + assert(f); + assert(f->header); + assert(o); + assert(offset > 0); + + if (o->object.type != OBJECT_ENTRY) + return -EINVAL; + + __sync_synchronize(); + + /* Link up the entry itself */ + r = link_entry_into_array(f, + &f->header->entry_array_offset, + &f->header->n_entries, + offset); + if (r < 0) + return r; + + /* log_debug("=> %s seqnr=%"PRIu64" n_entries=%"PRIu64, f->path, o->entry.seqnum, f->header->n_entries); */ + + if (f->header->head_entry_realtime == 0) + f->header->head_entry_realtime = o->entry.realtime; + + f->header->tail_entry_realtime = o->entry.realtime; + f->header->tail_entry_monotonic = o->entry.monotonic; + + f->tail_entry_monotonic_valid = true; + + /* Link up the items */ + n = journal_file_entry_n_items(o); + for (i = 0; i < n; i++) { + r = journal_file_link_entry_item(f, o, offset, i); + if (r < 0) + return r; + } + + return 0; +} + +static int journal_file_append_entry_internal( + JournalFile *f, + const dual_timestamp *ts, + uint64_t xor_hash, + const EntryItem items[], unsigned n_items, + uint64_t *seqnum, + Object **ret, uint64_t *offset) { + uint64_t np; + uint64_t osize; + Object *o; + int r; + + assert(f); + assert(f->header); + assert(items || n_items == 0); + assert(ts); + + osize = offsetof(Object, entry.items) + (n_items * sizeof(EntryItem)); + + r = journal_file_append_object(f, OBJECT_ENTRY, osize, &o, &np); + if (r < 0) + return r; + + o->entry.seqnum = htole64(journal_file_entry_seqnum(f, seqnum)); + memcpy_safe(o->entry.items, items, n_items * sizeof(EntryItem)); + o->entry.realtime = htole64(ts->realtime); + o->entry.monotonic = htole64(ts->monotonic); + o->entry.xor_hash = htole64(xor_hash); + o->entry.boot_id = f->header->boot_id; + +#ifdef HAVE_GCRYPT + r = journal_file_hmac_put_object(f, OBJECT_ENTRY, o, np); + if (r < 0) + return r; +#endif + + r = journal_file_link_entry(f, o, np); + if (r < 0) + return r; + + if (ret) + *ret = o; + + if (offset) + *offset = np; + + return 0; +} + +void journal_file_post_change(JournalFile *f) { + assert(f); + + /* inotify() does not receive IN_MODIFY events from file + * accesses done via mmap(). After each access we hence + * trigger IN_MODIFY by truncating the journal file to its + * current size which triggers IN_MODIFY. */ + + __sync_synchronize(); + + if (ftruncate(f->fd, f->last_stat.st_size) < 0) + log_debug_errno(errno, "Failed to truncate file to its own size: %m"); +} + +static int post_change_thunk(sd_event_source *timer, uint64_t usec, void *userdata) { + assert(userdata); + + journal_file_post_change(userdata); + + return 1; +} + +static void schedule_post_change(JournalFile *f) { + sd_event_source *timer; + int enabled, r; + uint64_t now; + + assert(f); + assert(f->post_change_timer); + + timer = f->post_change_timer; + + r = sd_event_source_get_enabled(timer, &enabled); + if (r < 0) { + log_debug_errno(r, "Failed to get ftruncate timer state: %m"); + goto fail; + } + + if (enabled == SD_EVENT_ONESHOT) + return; + + r = sd_event_now(sd_event_source_get_event(timer), CLOCK_MONOTONIC, &now); + if (r < 0) { + log_debug_errno(r, "Failed to get clock's now for scheduling ftruncate: %m"); + goto fail; + } + + r = sd_event_source_set_time(timer, now+f->post_change_timer_period); + if (r < 0) { + log_debug_errno(r, "Failed to set time for scheduling ftruncate: %m"); + goto fail; + } + + r = sd_event_source_set_enabled(timer, SD_EVENT_ONESHOT); + if (r < 0) { + log_debug_errno(r, "Failed to enable scheduled ftruncate: %m"); + goto fail; + } + + return; + +fail: + /* On failure, let's simply post the change immediately. */ + journal_file_post_change(f); +} + +/* Enable coalesced change posting in a timer on the provided sd_event instance */ +int journal_file_enable_post_change_timer(JournalFile *f, sd_event *e, usec_t t) { + _cleanup_(sd_event_source_unrefp) sd_event_source *timer = NULL; + int r; + + assert(f); + assert_return(!f->post_change_timer, -EINVAL); + assert(e); + assert(t); + + r = sd_event_add_time(e, &timer, CLOCK_MONOTONIC, 0, 0, post_change_thunk, f); + if (r < 0) + return r; + + r = sd_event_source_set_enabled(timer, SD_EVENT_OFF); + if (r < 0) + return r; + + f->post_change_timer = timer; + timer = NULL; + f->post_change_timer_period = t; + + return r; +} + +static int entry_item_cmp(const void *_a, const void *_b) { + const EntryItem *a = _a, *b = _b; + + if (le64toh(a->object_offset) < le64toh(b->object_offset)) + return -1; + if (le64toh(a->object_offset) > le64toh(b->object_offset)) + return 1; + return 0; +} + +int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const struct iovec iovec[], unsigned n_iovec, uint64_t *seqnum, Object **ret, uint64_t *offset) { + unsigned i; + EntryItem *items; + int r; + uint64_t xor_hash = 0; + struct dual_timestamp _ts; + + assert(f); + assert(f->header); + assert(iovec || n_iovec == 0); + + if (!ts) { + dual_timestamp_get(&_ts); + ts = &_ts; + } + +#ifdef HAVE_GCRYPT + r = journal_file_maybe_append_tag(f, ts->realtime); + if (r < 0) + return r; +#endif + + /* alloca() can't take 0, hence let's allocate at least one */ + items = alloca(sizeof(EntryItem) * MAX(1u, n_iovec)); + + for (i = 0; i < n_iovec; i++) { + uint64_t p; + Object *o; + + r = journal_file_append_data(f, iovec[i].iov_base, iovec[i].iov_len, &o, &p); + if (r < 0) + return r; + + xor_hash ^= le64toh(o->data.hash); + items[i].object_offset = htole64(p); + items[i].hash = o->data.hash; + } + + /* Order by the position on disk, in order to improve seek + * times for rotating media. */ + qsort_safe(items, n_iovec, sizeof(EntryItem), entry_item_cmp); + + r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqnum, ret, offset); + + /* If the memory mapping triggered a SIGBUS then we return an + * IO error and ignore the error code passed down to us, since + * it is very likely just an effect of a nullified replacement + * mapping page */ + + if (mmap_cache_got_sigbus(f->mmap, f->fd)) + r = -EIO; + + if (f->post_change_timer) + schedule_post_change(f); + else + journal_file_post_change(f); + + return r; +} + +typedef struct ChainCacheItem { + uint64_t first; /* the array at the beginning of the chain */ + uint64_t array; /* the cached array */ + uint64_t begin; /* the first item in the cached array */ + uint64_t total; /* the total number of items in all arrays before this one in the chain */ + uint64_t last_index; /* the last index we looked at, to optimize locality when bisecting */ +} ChainCacheItem; + +static void chain_cache_put( + OrderedHashmap *h, + ChainCacheItem *ci, + uint64_t first, + uint64_t array, + uint64_t begin, + uint64_t total, + uint64_t last_index) { + + if (!ci) { + /* If the chain item to cache for this chain is the + * first one it's not worth caching anything */ + if (array == first) + return; + + if (ordered_hashmap_size(h) >= CHAIN_CACHE_MAX) { + ci = ordered_hashmap_steal_first(h); + assert(ci); + } else { + ci = new(ChainCacheItem, 1); + if (!ci) + return; + } + + ci->first = first; + + if (ordered_hashmap_put(h, &ci->first, ci) < 0) { + free(ci); + return; + } + } else + assert(ci->first == first); + + ci->array = array; + ci->begin = begin; + ci->total = total; + ci->last_index = last_index; +} + +static int generic_array_get( + JournalFile *f, + uint64_t first, + uint64_t i, + Object **ret, uint64_t *offset) { + + Object *o; + uint64_t p = 0, a, t = 0; + int r; + ChainCacheItem *ci; + + assert(f); + + a = first; + + /* Try the chain cache first */ + ci = ordered_hashmap_get(f->chain_cache, &first); + if (ci && i > ci->total) { + a = ci->array; + i -= ci->total; + t = ci->total; + } + + while (a > 0) { + uint64_t k; + + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + + k = journal_file_entry_array_n_items(o); + if (i < k) { + p = le64toh(o->entry_array.items[i]); + goto found; + } + + i -= k; + t += k; + a = le64toh(o->entry_array.next_entry_array_offset); + } + + return 0; + +found: + /* Let's cache this item for the next invocation */ + chain_cache_put(f->chain_cache, ci, first, a, le64toh(o->entry_array.items[0]), t, i); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + if (ret) + *ret = o; + + if (offset) + *offset = p; + + return 1; +} + +static int generic_array_get_plus_one( + JournalFile *f, + uint64_t extra, + uint64_t first, + uint64_t i, + Object **ret, uint64_t *offset) { + + Object *o; + + assert(f); + + if (i == 0) { + int r; + + r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o); + if (r < 0) + return r; + + if (ret) + *ret = o; + + if (offset) + *offset = extra; + + return 1; + } + + return generic_array_get(f, first, i-1, ret, offset); +} + +enum { + TEST_FOUND, + TEST_LEFT, + TEST_RIGHT +}; + +static int generic_array_bisect( + JournalFile *f, + uint64_t first, + uint64_t n, + uint64_t needle, + int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle), + direction_t direction, + Object **ret, + uint64_t *offset, + uint64_t *idx) { + + uint64_t a, p, t = 0, i = 0, last_p = 0, last_index = (uint64_t) -1; + bool subtract_one = false; + Object *o, *array = NULL; + int r; + ChainCacheItem *ci; + + assert(f); + assert(test_object); + + /* Start with the first array in the chain */ + a = first; + + ci = ordered_hashmap_get(f->chain_cache, &first); + if (ci && n > ci->total) { + /* Ah, we have iterated this bisection array chain + * previously! Let's see if we can skip ahead in the + * chain, as far as the last time. But we can't jump + * backwards in the chain, so let's check that + * first. */ + + r = test_object(f, ci->begin, needle); + if (r < 0) + return r; + + if (r == TEST_LEFT) { + /* OK, what we are looking for is right of the + * begin of this EntryArray, so let's jump + * straight to previously cached array in the + * chain */ + + a = ci->array; + n -= ci->total; + t = ci->total; + last_index = ci->last_index; + } + } + + while (a > 0) { + uint64_t left, right, k, lp; + + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &array); + if (r < 0) + return r; + + k = journal_file_entry_array_n_items(array); + right = MIN(k, n); + if (right <= 0) + return 0; + + i = right - 1; + lp = p = le64toh(array->entry_array.items[i]); + if (p <= 0) + r = -EBADMSG; + else + r = test_object(f, p, needle); + if (r == -EBADMSG) { + log_debug_errno(r, "Encountered invalid entry while bisecting, cutting algorithm short. (1)"); + n = i; + continue; + } + if (r < 0) + return r; + + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + if (r == TEST_RIGHT) { + left = 0; + right -= 1; + + if (last_index != (uint64_t) -1) { + assert(last_index <= right); + + /* If we cached the last index we + * looked at, let's try to not to jump + * too wildly around and see if we can + * limit the range to look at early to + * the immediate neighbors of the last + * index we looked at. */ + + if (last_index > 0) { + uint64_t x = last_index - 1; + + p = le64toh(array->entry_array.items[x]); + if (p <= 0) + return -EBADMSG; + + r = test_object(f, p, needle); + if (r < 0) + return r; + + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + if (r == TEST_RIGHT) + right = x; + else + left = x + 1; + } + + if (last_index < right) { + uint64_t y = last_index + 1; + + p = le64toh(array->entry_array.items[y]); + if (p <= 0) + return -EBADMSG; + + r = test_object(f, p, needle); + if (r < 0) + return r; + + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + if (r == TEST_RIGHT) + right = y; + else + left = y + 1; + } + } + + for (;;) { + if (left == right) { + if (direction == DIRECTION_UP) + subtract_one = true; + + i = left; + goto found; + } + + assert(left < right); + i = (left + right) / 2; + + p = le64toh(array->entry_array.items[i]); + if (p <= 0) + r = -EBADMSG; + else + r = test_object(f, p, needle); + if (r == -EBADMSG) { + log_debug_errno(r, "Encountered invalid entry while bisecting, cutting algorithm short. (2)"); + right = n = i; + continue; + } + if (r < 0) + return r; + + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + if (r == TEST_RIGHT) + right = i; + else + left = i + 1; + } + } + + if (k >= n) { + if (direction == DIRECTION_UP) { + i = n; + subtract_one = true; + goto found; + } + + return 0; + } + + last_p = lp; + + n -= k; + t += k; + last_index = (uint64_t) -1; + a = le64toh(array->entry_array.next_entry_array_offset); + } + + return 0; + +found: + if (subtract_one && t == 0 && i == 0) + return 0; + + /* Let's cache this item for the next invocation */ + chain_cache_put(f->chain_cache, ci, first, a, le64toh(array->entry_array.items[0]), t, subtract_one ? (i > 0 ? i-1 : (uint64_t) -1) : i); + + if (subtract_one && i == 0) + p = last_p; + else if (subtract_one) + p = le64toh(array->entry_array.items[i-1]); + else + p = le64toh(array->entry_array.items[i]); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + if (ret) + *ret = o; + + if (offset) + *offset = p; + + if (idx) + *idx = t + i + (subtract_one ? -1 : 0); + + return 1; +} + +static int generic_array_bisect_plus_one( + JournalFile *f, + uint64_t extra, + uint64_t first, + uint64_t n, + uint64_t needle, + int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle), + direction_t direction, + Object **ret, + uint64_t *offset, + uint64_t *idx) { + + int r; + bool step_back = false; + Object *o; + + assert(f); + assert(test_object); + + if (n <= 0) + return 0; + + /* This bisects the array in object 'first', but first checks + * an extra */ + r = test_object(f, extra, needle); + if (r < 0) + return r; + + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + /* if we are looking with DIRECTION_UP then we need to first + see if in the actual array there is a matching entry, and + return the last one of that. But if there isn't any we need + to return this one. Hence remember this, and return it + below. */ + if (r == TEST_LEFT) + step_back = direction == DIRECTION_UP; + + if (r == TEST_RIGHT) { + if (direction == DIRECTION_DOWN) + goto found; + else + return 0; + } + + r = generic_array_bisect(f, first, n-1, needle, test_object, direction, ret, offset, idx); + + if (r == 0 && step_back) + goto found; + + if (r > 0 && idx) + (*idx)++; + + return r; + +found: + r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o); + if (r < 0) + return r; + + if (ret) + *ret = o; + + if (offset) + *offset = extra; + + if (idx) + *idx = 0; + + return 1; +} + +_pure_ static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) { + assert(f); + assert(p > 0); + + if (p == needle) + return TEST_FOUND; + else if (p < needle) + return TEST_LEFT; + else + return TEST_RIGHT; +} + +static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) { + Object *o; + int r; + + assert(f); + assert(p > 0); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + if (le64toh(o->entry.seqnum) == needle) + return TEST_FOUND; + else if (le64toh(o->entry.seqnum) < needle) + return TEST_LEFT; + else + return TEST_RIGHT; +} + +int journal_file_move_to_entry_by_seqnum( + JournalFile *f, + uint64_t seqnum, + direction_t direction, + Object **ret, + uint64_t *offset) { + assert(f); + assert(f->header); + + return generic_array_bisect(f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + seqnum, + test_object_seqnum, + direction, + ret, offset, NULL); +} + +static int test_object_realtime(JournalFile *f, uint64_t p, uint64_t needle) { + Object *o; + int r; + + assert(f); + assert(p > 0); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + if (le64toh(o->entry.realtime) == needle) + return TEST_FOUND; + else if (le64toh(o->entry.realtime) < needle) + return TEST_LEFT; + else + return TEST_RIGHT; +} + +int journal_file_move_to_entry_by_realtime( + JournalFile *f, + uint64_t realtime, + direction_t direction, + Object **ret, + uint64_t *offset) { + assert(f); + assert(f->header); + + return generic_array_bisect(f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + realtime, + test_object_realtime, + direction, + ret, offset, NULL); +} + +static int test_object_monotonic(JournalFile *f, uint64_t p, uint64_t needle) { + Object *o; + int r; + + assert(f); + assert(p > 0); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + if (le64toh(o->entry.monotonic) == needle) + return TEST_FOUND; + else if (le64toh(o->entry.monotonic) < needle) + return TEST_LEFT; + else + return TEST_RIGHT; +} + +static int find_data_object_by_boot_id( + JournalFile *f, + sd_id128_t boot_id, + Object **o, + uint64_t *b) { + + char t[sizeof("_BOOT_ID=")-1 + 32 + 1] = "_BOOT_ID="; + + sd_id128_to_string(boot_id, t + 9); + return journal_file_find_data_object(f, t, sizeof(t) - 1, o, b); +} + +int journal_file_move_to_entry_by_monotonic( + JournalFile *f, + sd_id128_t boot_id, + uint64_t monotonic, + direction_t direction, + Object **ret, + uint64_t *offset) { + + Object *o; + int r; + + assert(f); + + r = find_data_object_by_boot_id(f, boot_id, &o, NULL); + if (r < 0) + return r; + if (r == 0) + return -ENOENT; + + return generic_array_bisect_plus_one(f, + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + le64toh(o->data.n_entries), + monotonic, + test_object_monotonic, + direction, + ret, offset, NULL); +} + +void journal_file_reset_location(JournalFile *f) { + f->location_type = LOCATION_HEAD; + f->current_offset = 0; + f->current_seqnum = 0; + f->current_realtime = 0; + f->current_monotonic = 0; + zero(f->current_boot_id); + f->current_xor_hash = 0; +} + +void journal_file_save_location(JournalFile *f, Object *o, uint64_t offset) { + f->location_type = LOCATION_SEEK; + f->current_offset = offset; + f->current_seqnum = le64toh(o->entry.seqnum); + f->current_realtime = le64toh(o->entry.realtime); + f->current_monotonic = le64toh(o->entry.monotonic); + f->current_boot_id = o->entry.boot_id; + f->current_xor_hash = le64toh(o->entry.xor_hash); +} + +int journal_file_compare_locations(JournalFile *af, JournalFile *bf) { + assert(af); + assert(af->header); + assert(bf); + assert(bf->header); + assert(af->location_type == LOCATION_SEEK); + assert(bf->location_type == LOCATION_SEEK); + + /* If contents and timestamps match, these entries are + * identical, even if the seqnum does not match */ + if (sd_id128_equal(af->current_boot_id, bf->current_boot_id) && + af->current_monotonic == bf->current_monotonic && + af->current_realtime == bf->current_realtime && + af->current_xor_hash == bf->current_xor_hash) + return 0; + + if (sd_id128_equal(af->header->seqnum_id, bf->header->seqnum_id)) { + + /* If this is from the same seqnum source, compare + * seqnums */ + if (af->current_seqnum < bf->current_seqnum) + return -1; + if (af->current_seqnum > bf->current_seqnum) + return 1; + + /* Wow! This is weird, different data but the same + * seqnums? Something is borked, but let's make the + * best of it and compare by time. */ + } + + if (sd_id128_equal(af->current_boot_id, bf->current_boot_id)) { + + /* If the boot id matches, compare monotonic time */ + if (af->current_monotonic < bf->current_monotonic) + return -1; + if (af->current_monotonic > bf->current_monotonic) + return 1; + } + + /* Otherwise, compare UTC time */ + if (af->current_realtime < bf->current_realtime) + return -1; + if (af->current_realtime > bf->current_realtime) + return 1; + + /* Finally, compare by contents */ + if (af->current_xor_hash < bf->current_xor_hash) + return -1; + if (af->current_xor_hash > bf->current_xor_hash) + return 1; + + return 0; +} + +static int bump_array_index(uint64_t *i, direction_t direction, uint64_t n) { + + /* Increase or decrease the specified index, in the right direction. */ + + if (direction == DIRECTION_DOWN) { + if (*i >= n - 1) + return 0; + + (*i) ++; + } else { + if (*i <= 0) + return 0; + + (*i) --; + } + + return 1; +} + +static bool check_properly_ordered(uint64_t new_offset, uint64_t old_offset, direction_t direction) { + + /* Consider it an error if any of the two offsets is uninitialized */ + if (old_offset == 0 || new_offset == 0) + return false; + + /* If we go down, the new offset must be larger than the old one. */ + return direction == DIRECTION_DOWN ? + new_offset > old_offset : + new_offset < old_offset; +} + +int journal_file_next_entry( + JournalFile *f, + uint64_t p, + direction_t direction, + Object **ret, uint64_t *offset) { + + uint64_t i, n, ofs; + int r; + + assert(f); + assert(f->header); + + n = le64toh(f->header->n_entries); + if (n <= 0) + return 0; + + if (p == 0) + i = direction == DIRECTION_DOWN ? 0 : n - 1; + else { + r = generic_array_bisect(f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + p, + test_object_offset, + DIRECTION_DOWN, + NULL, NULL, + &i); + if (r <= 0) + return r; + + r = bump_array_index(&i, direction, n); + if (r <= 0) + return r; + } + + /* And jump to it */ + for (;;) { + r = generic_array_get(f, + le64toh(f->header->entry_array_offset), + i, + ret, &ofs); + if (r > 0) + break; + if (r != -EBADMSG) + return r; + + /* OK, so this entry is borked. Most likely some entry didn't get synced to disk properly, let's see if + * the next one might work for us instead. */ + log_debug_errno(r, "Entry item %" PRIu64 " is bad, skipping over it.", i); + + r = bump_array_index(&i, direction, n); + if (r <= 0) + return r; + } + + /* Ensure our array is properly ordered. */ + if (p > 0 && !check_properly_ordered(ofs, p, direction)) { + log_debug("%s: entry array not properly ordered at entry %" PRIu64, f->path, i); + return -EBADMSG; + } + + if (offset) + *offset = ofs; + + return 1; +} + +int journal_file_next_entry_for_data( + JournalFile *f, + Object *o, uint64_t p, + uint64_t data_offset, + direction_t direction, + Object **ret, uint64_t *offset) { + + uint64_t i, n, ofs; + Object *d; + int r; + + assert(f); + assert(p > 0 || !o); + + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); + if (r < 0) + return r; + + n = le64toh(d->data.n_entries); + if (n <= 0) + return n; + + if (!o) + i = direction == DIRECTION_DOWN ? 0 : n - 1; + else { + if (o->object.type != OBJECT_ENTRY) + return -EINVAL; + + r = generic_array_bisect_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + p, + test_object_offset, + DIRECTION_DOWN, + NULL, NULL, + &i); + + if (r <= 0) + return r; + + r = bump_array_index(&i, direction, n); + if (r <= 0) + return r; + } + + for (;;) { + r = generic_array_get_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + i, + ret, &ofs); + if (r > 0) + break; + if (r != -EBADMSG) + return r; + + log_debug_errno(r, "Data entry item %" PRIu64 " is bad, skipping over it.", i); + + r = bump_array_index(&i, direction, n); + if (r <= 0) + return r; + } + + /* Ensure our array is properly ordered. */ + if (p > 0 && check_properly_ordered(ofs, p, direction)) { + log_debug("%s data entry array not properly ordered at entry %" PRIu64, f->path, i); + return -EBADMSG; + } + + if (offset) + *offset = ofs; + + return 1; +} + +int journal_file_move_to_entry_by_offset_for_data( + JournalFile *f, + uint64_t data_offset, + uint64_t p, + direction_t direction, + Object **ret, uint64_t *offset) { + + int r; + Object *d; + + assert(f); + + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); + if (r < 0) + return r; + + return generic_array_bisect_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + p, + test_object_offset, + direction, + ret, offset, NULL); +} + +int journal_file_move_to_entry_by_monotonic_for_data( + JournalFile *f, + uint64_t data_offset, + sd_id128_t boot_id, + uint64_t monotonic, + direction_t direction, + Object **ret, uint64_t *offset) { + + Object *o, *d; + int r; + uint64_t b, z; + + assert(f); + + /* First, seek by time */ + r = find_data_object_by_boot_id(f, boot_id, &o, &b); + if (r < 0) + return r; + if (r == 0) + return -ENOENT; + + r = generic_array_bisect_plus_one(f, + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + le64toh(o->data.n_entries), + monotonic, + test_object_monotonic, + direction, + NULL, &z, NULL); + if (r <= 0) + return r; + + /* And now, continue seeking until we find an entry that + * exists in both bisection arrays */ + + for (;;) { + Object *qo; + uint64_t p, q; + + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); + if (r < 0) + return r; + + r = generic_array_bisect_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + z, + test_object_offset, + direction, + NULL, &p, NULL); + if (r <= 0) + return r; + + r = journal_file_move_to_object(f, OBJECT_DATA, b, &o); + if (r < 0) + return r; + + r = generic_array_bisect_plus_one(f, + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + le64toh(o->data.n_entries), + p, + test_object_offset, + direction, + &qo, &q, NULL); + + if (r <= 0) + return r; + + if (p == q) { + if (ret) + *ret = qo; + if (offset) + *offset = q; + + return 1; + } + + z = q; + } +} + +int journal_file_move_to_entry_by_seqnum_for_data( + JournalFile *f, + uint64_t data_offset, + uint64_t seqnum, + direction_t direction, + Object **ret, uint64_t *offset) { + + Object *d; + int r; + + assert(f); + + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); + if (r < 0) + return r; + + return generic_array_bisect_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + seqnum, + test_object_seqnum, + direction, + ret, offset, NULL); +} + +int journal_file_move_to_entry_by_realtime_for_data( + JournalFile *f, + uint64_t data_offset, + uint64_t realtime, + direction_t direction, + Object **ret, uint64_t *offset) { + + Object *d; + int r; + + assert(f); + + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); + if (r < 0) + return r; + + return generic_array_bisect_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + realtime, + test_object_realtime, + direction, + ret, offset, NULL); +} + +void journal_file_dump(JournalFile *f) { + Object *o; + int r; + uint64_t p; + + assert(f); + assert(f->header); + + journal_file_print_header(f); + + p = le64toh(f->header->header_size); + while (p != 0) { + r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &o); + if (r < 0) + goto fail; + + switch (o->object.type) { + + case OBJECT_UNUSED: + printf("Type: OBJECT_UNUSED\n"); + break; + + case OBJECT_DATA: + printf("Type: OBJECT_DATA\n"); + break; + + case OBJECT_FIELD: + printf("Type: OBJECT_FIELD\n"); + break; + + case OBJECT_ENTRY: + printf("Type: OBJECT_ENTRY seqnum=%"PRIu64" monotonic=%"PRIu64" realtime=%"PRIu64"\n", + le64toh(o->entry.seqnum), + le64toh(o->entry.monotonic), + le64toh(o->entry.realtime)); + break; + + case OBJECT_FIELD_HASH_TABLE: + printf("Type: OBJECT_FIELD_HASH_TABLE\n"); + break; + + case OBJECT_DATA_HASH_TABLE: + printf("Type: OBJECT_DATA_HASH_TABLE\n"); + break; + + case OBJECT_ENTRY_ARRAY: + printf("Type: OBJECT_ENTRY_ARRAY\n"); + break; + + case OBJECT_TAG: + printf("Type: OBJECT_TAG seqnum=%"PRIu64" epoch=%"PRIu64"\n", + le64toh(o->tag.seqnum), + le64toh(o->tag.epoch)); + break; + + default: + printf("Type: unknown (%i)\n", o->object.type); + break; + } + + if (o->object.flags & OBJECT_COMPRESSION_MASK) + printf("Flags: %s\n", + object_compressed_to_string(o->object.flags & OBJECT_COMPRESSION_MASK)); + + if (p == le64toh(f->header->tail_object_offset)) + p = 0; + else + p = p + ALIGN64(le64toh(o->object.size)); + } + + return; +fail: + log_error("File corrupt"); +} + +static const char* format_timestamp_safe(char *buf, size_t l, usec_t t) { + const char *x; + + x = format_timestamp(buf, l, t); + if (x) + return x; + return " --- "; +} + +void journal_file_print_header(JournalFile *f) { + char a[33], b[33], c[33], d[33]; + char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX], z[FORMAT_TIMESTAMP_MAX]; + struct stat st; + char bytes[FORMAT_BYTES_MAX]; + + assert(f); + assert(f->header); + + printf("File Path: %s\n" + "File ID: %s\n" + "Machine ID: %s\n" + "Boot ID: %s\n" + "Sequential Number ID: %s\n" + "State: %s\n" + "Compatible Flags:%s%s\n" + "Incompatible Flags:%s%s%s\n" + "Header size: %"PRIu64"\n" + "Arena size: %"PRIu64"\n" + "Data Hash Table Size: %"PRIu64"\n" + "Field Hash Table Size: %"PRIu64"\n" + "Rotate Suggested: %s\n" + "Head Sequential Number: %"PRIu64" (%"PRIx64")\n" + "Tail Sequential Number: %"PRIu64" (%"PRIx64")\n" + "Head Realtime Timestamp: %s (%"PRIx64")\n" + "Tail Realtime Timestamp: %s (%"PRIx64")\n" + "Tail Monotonic Timestamp: %s (%"PRIx64")\n" + "Objects: %"PRIu64"\n" + "Entry Objects: %"PRIu64"\n", + f->path, + sd_id128_to_string(f->header->file_id, a), + sd_id128_to_string(f->header->machine_id, b), + sd_id128_to_string(f->header->boot_id, c), + sd_id128_to_string(f->header->seqnum_id, d), + f->header->state == STATE_OFFLINE ? "OFFLINE" : + f->header->state == STATE_ONLINE ? "ONLINE" : + f->header->state == STATE_ARCHIVED ? "ARCHIVED" : "UNKNOWN", + JOURNAL_HEADER_SEALED(f->header) ? " SEALED" : "", + (le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_ANY) ? " ???" : "", + JOURNAL_HEADER_COMPRESSED_XZ(f->header) ? " COMPRESSED-XZ" : "", + JOURNAL_HEADER_COMPRESSED_LZ4(f->header) ? " COMPRESSED-LZ4" : "", + (le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_ANY) ? " ???" : "", + le64toh(f->header->header_size), + le64toh(f->header->arena_size), + le64toh(f->header->data_hash_table_size) / sizeof(HashItem), + le64toh(f->header->field_hash_table_size) / sizeof(HashItem), + yes_no(journal_file_rotate_suggested(f, 0)), + le64toh(f->header->head_entry_seqnum), le64toh(f->header->head_entry_seqnum), + le64toh(f->header->tail_entry_seqnum), le64toh(f->header->tail_entry_seqnum), + format_timestamp_safe(x, sizeof(x), le64toh(f->header->head_entry_realtime)), le64toh(f->header->head_entry_realtime), + format_timestamp_safe(y, sizeof(y), le64toh(f->header->tail_entry_realtime)), le64toh(f->header->tail_entry_realtime), + format_timespan(z, sizeof(z), le64toh(f->header->tail_entry_monotonic), USEC_PER_MSEC), le64toh(f->header->tail_entry_monotonic), + le64toh(f->header->n_objects), + le64toh(f->header->n_entries)); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) + printf("Data Objects: %"PRIu64"\n" + "Data Hash Table Fill: %.1f%%\n", + le64toh(f->header->n_data), + 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)))); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) + printf("Field Objects: %"PRIu64"\n" + "Field Hash Table Fill: %.1f%%\n", + le64toh(f->header->n_fields), + 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)))); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_tags)) + printf("Tag Objects: %"PRIu64"\n", + le64toh(f->header->n_tags)); + if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) + printf("Entry Array Objects: %"PRIu64"\n", + le64toh(f->header->n_entry_arrays)); + + if (fstat(f->fd, &st) >= 0) + printf("Disk usage: %s\n", format_bytes(bytes, sizeof(bytes), (uint64_t) st.st_blocks * 512ULL)); +} + +static int journal_file_warn_btrfs(JournalFile *f) { + unsigned attrs; + int r; + + assert(f); + + /* Before we write anything, check if the COW logic is turned + * off on btrfs. Given our write pattern that is quite + * unfriendly to COW file systems this should greatly improve + * performance on COW file systems, such as btrfs, at the + * expense of data integrity features (which shouldn't be too + * bad, given that we do our own checksumming). */ + + r = btrfs_is_filesystem(f->fd); + if (r < 0) + return log_warning_errno(r, "Failed to determine if journal is on btrfs: %m"); + if (!r) + return 0; + + r = read_attr_fd(f->fd, &attrs); + if (r < 0) + return log_warning_errno(r, "Failed to read file attributes: %m"); + + if (attrs & FS_NOCOW_FL) { + log_debug("Detected btrfs file system with copy-on-write disabled, all is good."); + return 0; + } + + log_notice("Creating journal file %s on a btrfs file system, and copy-on-write is enabled. " + "This is likely to slow down journal access substantially, please consider turning " + "off the copy-on-write file attribute on the journal directory, using chattr +C.", f->path); + + return 1; +} + +int journal_file_open( + int fd, + const char *fname, + int flags, + mode_t mode, + bool compress, + bool seal, + JournalMetrics *metrics, + MMapCache *mmap_cache, + Set *deferred_closes, + JournalFile *template, + JournalFile **ret) { + + bool newly_created = false; + JournalFile *f; + void *h; + int r; + + assert(ret); + assert(fd >= 0 || fname); + + if ((flags & O_ACCMODE) != O_RDONLY && + (flags & O_ACCMODE) != O_RDWR) + return -EINVAL; + + if (fname) { + if (!endswith(fname, ".journal") && + !endswith(fname, ".journal~")) + return -EINVAL; + } + + f = new0(JournalFile, 1); + if (!f) + return -ENOMEM; + + f->fd = fd; + f->mode = mode; + + f->flags = flags; + f->prot = prot_from_flags(flags); + f->writable = (flags & O_ACCMODE) != O_RDONLY; +#if defined(HAVE_LZ4) + f->compress_lz4 = compress; +#elif defined(HAVE_XZ) + f->compress_xz = compress; +#endif +#ifdef HAVE_GCRYPT + f->seal = seal; +#endif + + if (mmap_cache) + f->mmap = mmap_cache_ref(mmap_cache); + else { + f->mmap = mmap_cache_new(); + if (!f->mmap) { + r = -ENOMEM; + goto fail; + } + } + + if (fname) + f->path = strdup(fname); + else /* If we don't know the path, fill in something explanatory and vaguely useful */ + asprintf(&f->path, "/proc/self/%i", fd); + if (!f->path) { + r = -ENOMEM; + goto fail; + } + + f->chain_cache = ordered_hashmap_new(&uint64_hash_ops); + if (!f->chain_cache) { + r = -ENOMEM; + goto fail; + } + + if (f->fd < 0) { + f->fd = open(f->path, f->flags|O_CLOEXEC, f->mode); + if (f->fd < 0) { + r = -errno; + goto fail; + } + + /* fds we opened here by us should also be closed by us. */ + f->close_fd = true; + } + + r = journal_file_fstat(f); + if (r < 0) + goto fail; + + if (f->last_stat.st_size == 0 && f->writable) { + + (void) journal_file_warn_btrfs(f); + + /* Let's attach the creation time to the journal file, + * so that the vacuuming code knows the age of this + * file even if the file might end up corrupted one + * day... Ideally we'd just use the creation time many + * file systems maintain for each file, but there is + * currently no usable API to query this, hence let's + * emulate this via extended attributes. If extended + * attributes are not supported we'll just skip this, + * and rely solely on mtime/atime/ctime of the file. */ + + fd_setcrtime(f->fd, 0); + +#ifdef HAVE_GCRYPT + /* Try to load the FSPRG state, and if we can't, then + * just don't do sealing */ + if (f->seal) { + r = journal_file_fss_load(f); + if (r < 0) + f->seal = false; + } +#endif + + r = journal_file_init_header(f, template); + if (r < 0) + goto fail; + + r = journal_file_fstat(f); + if (r < 0) + goto fail; + + newly_created = true; + } + + if (f->last_stat.st_size < (off_t) HEADER_SIZE_MIN) { + r = -ENODATA; + goto fail; + } + + r = mmap_cache_get(f->mmap, f->fd, f->prot, CONTEXT_HEADER, true, 0, PAGE_ALIGN(sizeof(Header)), &f->last_stat, &h); + if (r < 0) + goto fail; + + f->header = h; + + if (!newly_created) { + if (deferred_closes) + journal_file_close_set(deferred_closes); + + r = journal_file_verify_header(f); + if (r < 0) + goto fail; + } + +#ifdef HAVE_GCRYPT + if (!newly_created && f->writable) { + r = journal_file_fss_load(f); + if (r < 0) + goto fail; + } +#endif + + if (f->writable) { + if (metrics) { + journal_default_metrics(metrics, f->fd); + f->metrics = *metrics; + } else if (template) + f->metrics = template->metrics; + + r = journal_file_refresh_header(f); + if (r < 0) + goto fail; + } + +#ifdef HAVE_GCRYPT + r = journal_file_hmac_setup(f); + if (r < 0) + goto fail; +#endif + + if (newly_created) { + r = journal_file_setup_field_hash_table(f); + if (r < 0) + goto fail; + + r = journal_file_setup_data_hash_table(f); + if (r < 0) + goto fail; + +#ifdef HAVE_GCRYPT + r = journal_file_append_first_tag(f); + if (r < 0) + goto fail; +#endif + } + + if (mmap_cache_got_sigbus(f->mmap, f->fd)) { + r = -EIO; + goto fail; + } + + if (template && template->post_change_timer) { + r = journal_file_enable_post_change_timer( + f, + sd_event_source_get_event(template->post_change_timer), + template->post_change_timer_period); + + if (r < 0) + goto fail; + } + + /* The file is opened now successfully, thus we take possession of any passed in fd. */ + f->close_fd = true; + + *ret = f; + return 0; + +fail: + if (f->fd >= 0 && mmap_cache_got_sigbus(f->mmap, f->fd)) + r = -EIO; + + (void) journal_file_close(f); + + return r; +} + +int journal_file_rotate(JournalFile **f, bool compress, bool seal, Set *deferred_closes) { + _cleanup_free_ char *p = NULL; + size_t l; + JournalFile *old_file, *new_file = NULL; + int r; + + assert(f); + assert(*f); + + old_file = *f; + + if (!old_file->writable) + return -EINVAL; + + /* Is this a journal file that was passed to us as fd? If so, we synthesized a path name for it, and we refuse + * rotation, since we don't know the actual path, and couldn't rename the file hence.*/ + if (path_startswith(old_file->path, "/proc/self/fd")) + return -EINVAL; + + if (!endswith(old_file->path, ".journal")) + return -EINVAL; + + l = strlen(old_file->path); + r = asprintf(&p, "%.*s@" SD_ID128_FORMAT_STR "-%016"PRIx64"-%016"PRIx64".journal", + (int) l - 8, old_file->path, + SD_ID128_FORMAT_VAL(old_file->header->seqnum_id), + le64toh((*f)->header->head_entry_seqnum), + le64toh((*f)->header->head_entry_realtime)); + if (r < 0) + return -ENOMEM; + + /* Try to rename the file to the archived version. If the file + * already was deleted, we'll get ENOENT, let's ignore that + * case. */ + r = rename(old_file->path, p); + if (r < 0 && errno != ENOENT) + return -errno; + + /* Sync the rename to disk */ + (void) fsync_directory_of_file(old_file->fd); + + /* Set as archive so offlining commits w/state=STATE_ARCHIVED. + * Previously we would set old_file->header->state to STATE_ARCHIVED directly here, + * but journal_file_set_offline() short-circuits when state != STATE_ONLINE, which + * would result in the rotated journal never getting fsync() called before closing. + * Now we simply queue the archive state by setting an archive bit, leaving the state + * as STATE_ONLINE so proper offlining occurs. */ + old_file->archive = true; + + /* Currently, btrfs is not very good with out write patterns + * and fragments heavily. Let's defrag our journal files when + * we archive them */ + old_file->defrag_on_close = true; + + r = journal_file_open(-1, old_file->path, old_file->flags, old_file->mode, compress, seal, NULL, old_file->mmap, deferred_closes, old_file, &new_file); + + if (deferred_closes && + set_put(deferred_closes, old_file) >= 0) + (void) journal_file_set_offline(old_file, false); + else + (void) journal_file_close(old_file); + + *f = new_file; + return r; +} + +int journal_file_open_reliably( + const char *fname, + int flags, + mode_t mode, + bool compress, + bool seal, + JournalMetrics *metrics, + MMapCache *mmap_cache, + Set *deferred_closes, + JournalFile *template, + JournalFile **ret) { + + int r; + size_t l; + _cleanup_free_ char *p = NULL; + + r = journal_file_open(-1, fname, flags, mode, compress, seal, metrics, mmap_cache, deferred_closes, template, ret); + if (!IN_SET(r, + -EBADMSG, /* corrupted */ + -ENODATA, /* truncated */ + -EHOSTDOWN, /* other machine */ + -EPROTONOSUPPORT, /* incompatible feature */ + -EBUSY, /* unclean shutdown */ + -ESHUTDOWN, /* already archived */ + -EIO, /* IO error, including SIGBUS on mmap */ + -EIDRM, /* File has been deleted */ + -ETXTBSY)) /* File is from the future */ + return r; + + if ((flags & O_ACCMODE) == O_RDONLY) + return r; + + if (!(flags & O_CREAT)) + return r; + + if (!endswith(fname, ".journal")) + return r; + + /* The file is corrupted. Rotate it away and try it again (but only once) */ + + l = strlen(fname); + if (asprintf(&p, "%.*s@%016"PRIx64 "-%016"PRIx64 ".journal~", + (int) l - 8, fname, + now(CLOCK_REALTIME), + random_u64()) < 0) + return -ENOMEM; + + if (rename(fname, p) < 0) + return -errno; + + /* btrfs doesn't cope well with our write pattern and + * fragments heavily. Let's defrag all files we rotate */ + + (void) chattr_path(p, 0, FS_NOCOW_FL); + (void) btrfs_defrag(p); + + log_warning_errno(r, "File %s corrupted or uncleanly shut down, renaming and replacing.", fname); + + return journal_file_open(-1, fname, flags, mode, compress, seal, metrics, mmap_cache, deferred_closes, template, ret); +} + +int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint64_t p, uint64_t *seqnum, Object **ret, uint64_t *offset) { + uint64_t i, n; + uint64_t q, xor_hash = 0; + int r; + EntryItem *items; + dual_timestamp ts; + + assert(from); + assert(to); + assert(o); + assert(p); + + if (!to->writable) + return -EPERM; + + ts.monotonic = le64toh(o->entry.monotonic); + ts.realtime = le64toh(o->entry.realtime); + + n = journal_file_entry_n_items(o); + /* alloca() can't take 0, hence let's allocate at least one */ + items = alloca(sizeof(EntryItem) * MAX(1u, n)); + + for (i = 0; i < n; i++) { + uint64_t l, h; + le64_t le_hash; + size_t t; + void *data; + Object *u; + + q = le64toh(o->entry.items[i].object_offset); + le_hash = o->entry.items[i].hash; + + r = journal_file_move_to_object(from, OBJECT_DATA, q, &o); + if (r < 0) + return r; + + if (le_hash != o->data.hash) + return -EBADMSG; + + l = le64toh(o->object.size) - offsetof(Object, data.payload); + t = (size_t) l; + + /* We hit the limit on 32bit machines */ + if ((uint64_t) t != l) + return -E2BIG; + + if (o->object.flags & OBJECT_COMPRESSION_MASK) { +#if defined(HAVE_XZ) || defined(HAVE_LZ4) + size_t rsize = 0; + + r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK, + o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize, 0); + if (r < 0) + return r; + + data = from->compress_buffer; + l = rsize; +#else + return -EPROTONOSUPPORT; +#endif + } else + data = o->data.payload; + + r = journal_file_append_data(to, data, l, &u, &h); + if (r < 0) + return r; + + xor_hash ^= le64toh(u->data.hash); + items[i].object_offset = htole64(h); + items[i].hash = u->data.hash; + + r = journal_file_move_to_object(from, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + } + + r = journal_file_append_entry_internal(to, &ts, xor_hash, items, n, seqnum, ret, offset); + + if (mmap_cache_got_sigbus(to->mmap, to->fd)) + return -EIO; + + return r; +} + +void journal_reset_metrics(JournalMetrics *m) { + assert(m); + + /* Set everything to "pick automatic values". */ + + *m = (JournalMetrics) { + .min_use = (uint64_t) -1, + .max_use = (uint64_t) -1, + .min_size = (uint64_t) -1, + .max_size = (uint64_t) -1, + .keep_free = (uint64_t) -1, + .n_max_files = (uint64_t) -1, + }; +} + +void journal_default_metrics(JournalMetrics *m, int fd) { + char a[FORMAT_BYTES_MAX], b[FORMAT_BYTES_MAX], c[FORMAT_BYTES_MAX], d[FORMAT_BYTES_MAX], e[FORMAT_BYTES_MAX]; + struct statvfs ss; + uint64_t fs_size; + + assert(m); + assert(fd >= 0); + + if (fstatvfs(fd, &ss) >= 0) + fs_size = ss.f_frsize * ss.f_blocks; + else { + log_debug_errno(errno, "Failed to detremine disk size: %m"); + fs_size = 0; + } + + if (m->max_use == (uint64_t) -1) { + + if (fs_size > 0) { + m->max_use = PAGE_ALIGN(fs_size / 10); /* 10% of file system size */ + + if (m->max_use > DEFAULT_MAX_USE_UPPER) + m->max_use = DEFAULT_MAX_USE_UPPER; + + if (m->max_use < DEFAULT_MAX_USE_LOWER) + m->max_use = DEFAULT_MAX_USE_LOWER; + } else + m->max_use = DEFAULT_MAX_USE_LOWER; + } else { + m->max_use = PAGE_ALIGN(m->max_use); + + if (m->max_use != 0 && m->max_use < JOURNAL_FILE_SIZE_MIN*2) + m->max_use = JOURNAL_FILE_SIZE_MIN*2; + } + + if (m->min_use == (uint64_t) -1) + m->min_use = DEFAULT_MIN_USE; + + if (m->min_use > m->max_use) + m->min_use = m->max_use; + + if (m->max_size == (uint64_t) -1) { + m->max_size = PAGE_ALIGN(m->max_use / 8); /* 8 chunks */ + + if (m->max_size > DEFAULT_MAX_SIZE_UPPER) + m->max_size = DEFAULT_MAX_SIZE_UPPER; + } else + m->max_size = PAGE_ALIGN(m->max_size); + + if (m->max_size != 0) { + if (m->max_size < JOURNAL_FILE_SIZE_MIN) + m->max_size = JOURNAL_FILE_SIZE_MIN; + + if (m->max_use != 0 && m->max_size*2 > m->max_use) + m->max_use = m->max_size*2; + } + + if (m->min_size == (uint64_t) -1) + m->min_size = JOURNAL_FILE_SIZE_MIN; + else { + m->min_size = PAGE_ALIGN(m->min_size); + + if (m->min_size < JOURNAL_FILE_SIZE_MIN) + m->min_size = JOURNAL_FILE_SIZE_MIN; + + if (m->max_size != 0 && m->min_size > m->max_size) + m->max_size = m->min_size; + } + + if (m->keep_free == (uint64_t) -1) { + + if (fs_size > 0) { + m->keep_free = PAGE_ALIGN(fs_size * 3 / 20); /* 15% of file system size */ + + if (m->keep_free > DEFAULT_KEEP_FREE_UPPER) + m->keep_free = DEFAULT_KEEP_FREE_UPPER; + + } else + m->keep_free = DEFAULT_KEEP_FREE; + } + + if (m->n_max_files == (uint64_t) -1) + m->n_max_files = DEFAULT_N_MAX_FILES; + + log_debug("Fixed min_use=%s max_use=%s max_size=%s min_size=%s keep_free=%s n_max_files=%" PRIu64, + format_bytes(a, sizeof(a), m->min_use), + format_bytes(b, sizeof(b), m->max_use), + format_bytes(c, sizeof(c), m->max_size), + format_bytes(d, sizeof(d), m->min_size), + format_bytes(e, sizeof(e), m->keep_free), + m->n_max_files); +} + +int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *to) { + assert(f); + assert(f->header); + assert(from || to); + + if (from) { + if (f->header->head_entry_realtime == 0) + return -ENOENT; + + *from = le64toh(f->header->head_entry_realtime); + } + + if (to) { + if (f->header->tail_entry_realtime == 0) + return -ENOENT; + + *to = le64toh(f->header->tail_entry_realtime); + } + + return 1; +} + +int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, usec_t *from, usec_t *to) { + Object *o; + uint64_t p; + int r; + + assert(f); + assert(from || to); + + r = find_data_object_by_boot_id(f, boot_id, &o, &p); + if (r <= 0) + return r; + + if (le64toh(o->data.n_entries) <= 0) + return 0; + + if (from) { + r = journal_file_move_to_object(f, OBJECT_ENTRY, le64toh(o->data.entry_offset), &o); + if (r < 0) + return r; + + *from = le64toh(o->entry.monotonic); + } + + if (to) { + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + r = generic_array_get_plus_one(f, + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + le64toh(o->data.n_entries)-1, + &o, NULL); + if (r <= 0) + return r; + + *to = le64toh(o->entry.monotonic); + } + + return 1; +} + +bool journal_file_rotate_suggested(JournalFile *f, usec_t max_file_usec) { + assert(f); + assert(f->header); + + /* If we gained new header fields we gained new features, + * hence suggest a rotation */ + if (le64toh(f->header->header_size) < sizeof(Header)) { + log_debug("%s uses an outdated header, suggesting rotation.", f->path); + return true; + } + + /* Let's check if the hash tables grew over a certain fill + * level (75%, borrowing this value from Java's hash table + * implementation), and if so suggest a rotation. To calculate + * the fill level we need the n_data field, which only exists + * in newer versions. */ + + if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) + if (le64toh(f->header->n_data) * 4ULL > (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)) * 3ULL) { + log_debug("Data hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items, %llu file size, %"PRIu64" bytes per hash table item), suggesting rotation.", + f->path, + 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))), + le64toh(f->header->n_data), + le64toh(f->header->data_hash_table_size) / sizeof(HashItem), + (unsigned long long) f->last_stat.st_size, + f->last_stat.st_size / le64toh(f->header->n_data)); + return true; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) + if (le64toh(f->header->n_fields) * 4ULL > (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)) * 3ULL) { + log_debug("Field hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items), suggesting rotation.", + f->path, + 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))), + le64toh(f->header->n_fields), + le64toh(f->header->field_hash_table_size) / sizeof(HashItem)); + return true; + } + + /* Are the data objects properly indexed by field objects? */ + if (JOURNAL_HEADER_CONTAINS(f->header, n_data) && + JOURNAL_HEADER_CONTAINS(f->header, n_fields) && + le64toh(f->header->n_data) > 0 && + le64toh(f->header->n_fields) == 0) + return true; + + if (max_file_usec > 0) { + usec_t t, h; + + h = le64toh(f->header->head_entry_realtime); + t = now(CLOCK_REALTIME); + + if (h > 0 && t > h + max_file_usec) + return true; + } + + return false; +} diff --git a/src/libsystemd/src/sd-journal/journal-file.h b/src/libsystemd/src/sd-journal/journal-file.h new file mode 100644 index 0000000000..054fe5794b --- /dev/null +++ b/src/libsystemd/src/sd-journal/journal-file.h @@ -0,0 +1,266 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <inttypes.h> + +#ifdef HAVE_GCRYPT +#include <gcrypt.h> +#endif + +#include <systemd/sd-event.h> +#include <systemd/sd-id128.h> + +#include "systemd-basic/hashmap.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/sparse-endian.h" + +#include "journal-def.h" +#include "mmap-cache.h" + +typedef struct JournalMetrics { + /* For all these: -1 means "pick automatically", and 0 means "no limit enforced" */ + uint64_t max_size; /* how large journal files grow at max */ + uint64_t min_size; /* how large journal files grow at least */ + uint64_t max_use; /* how much disk space to use in total at max, keep_free permitting */ + uint64_t min_use; /* how much disk space to use in total at least, even if keep_free says not to */ + uint64_t keep_free; /* how much to keep free on disk */ + uint64_t n_max_files; /* how many files to keep around at max */ +} JournalMetrics; + +typedef enum direction { + DIRECTION_UP, + DIRECTION_DOWN +} direction_t; + +typedef enum LocationType { + /* The first and last entries, resp. */ + LOCATION_HEAD, + LOCATION_TAIL, + + /* We already read the entry we currently point to, and the + * next one to read should probably not be this one again. */ + LOCATION_DISCRETE, + + /* We should seek to the precise location specified, and + * return it, as we haven't read it yet. */ + LOCATION_SEEK +} LocationType; + +typedef enum OfflineState { + OFFLINE_JOINED, + OFFLINE_SYNCING, + OFFLINE_OFFLINING, + OFFLINE_CANCEL, + OFFLINE_AGAIN_FROM_SYNCING, + OFFLINE_AGAIN_FROM_OFFLINING, + OFFLINE_DONE +} OfflineState; + +typedef struct JournalFile { + int fd; + + mode_t mode; + + int flags; + int prot; + bool writable:1; + bool compress_xz:1; + bool compress_lz4:1; + bool seal:1; + bool defrag_on_close:1; + bool close_fd:1; + bool archive:1; + + bool tail_entry_monotonic_valid:1; + + direction_t last_direction; + LocationType location_type; + uint64_t last_n_entries; + + char *path; + struct stat last_stat; + usec_t last_stat_usec; + + Header *header; + HashItem *data_hash_table; + HashItem *field_hash_table; + + uint64_t current_offset; + uint64_t current_seqnum; + uint64_t current_realtime; + uint64_t current_monotonic; + sd_id128_t current_boot_id; + uint64_t current_xor_hash; + + JournalMetrics metrics; + MMapCache *mmap; + + sd_event_source *post_change_timer; + usec_t post_change_timer_period; + + OrderedHashmap *chain_cache; + + pthread_t offline_thread; + volatile OfflineState offline_state; + +#if defined(HAVE_XZ) || defined(HAVE_LZ4) + void *compress_buffer; + size_t compress_buffer_size; +#endif + +#ifdef HAVE_GCRYPT + gcry_md_hd_t hmac; + bool hmac_running; + + FSSHeader *fss_file; + size_t fss_file_size; + + uint64_t fss_start_usec; + uint64_t fss_interval_usec; + + void *fsprg_state; + size_t fsprg_state_size; + + void *fsprg_seed; + size_t fsprg_seed_size; +#endif +} JournalFile; + +int journal_file_open( + int fd, + const char *fname, + int flags, + mode_t mode, + bool compress, + bool seal, + JournalMetrics *metrics, + MMapCache *mmap_cache, + Set *deferred_closes, + JournalFile *template, + JournalFile **ret); + +int journal_file_set_offline(JournalFile *f, bool wait); +bool journal_file_is_offlining(JournalFile *f); +JournalFile* journal_file_close(JournalFile *j); +void journal_file_close_set(Set *s); + +int journal_file_open_reliably( + const char *fname, + int flags, + mode_t mode, + bool compress, + bool seal, + JournalMetrics *metrics, + MMapCache *mmap_cache, + Set *deferred_closes, + JournalFile *template, + JournalFile **ret); + +#define ALIGN64(x) (((x) + 7ULL) & ~7ULL) +#define VALID64(x) (((x) & 7ULL) == 0ULL) + +/* Use six characters to cover the offsets common in smallish journal + * files without adding too many zeros. */ +#define OFSfmt "%06"PRIx64 + +static inline bool VALID_REALTIME(uint64_t u) { + /* This considers timestamps until the year 3112 valid. That should be plenty room... */ + return u > 0 && u < (1ULL << 55); +} + +static inline bool VALID_MONOTONIC(uint64_t u) { + /* This considers timestamps until 1142 years of runtime valid. */ + return u < (1ULL << 55); +} + +static inline bool VALID_EPOCH(uint64_t u) { + /* This allows changing the key for 1142 years, every usec. */ + return u < (1ULL << 55); +} + +#define JOURNAL_HEADER_CONTAINS(h, field) \ + (le64toh((h)->header_size) >= offsetof(Header, field) + sizeof((h)->field)) + +#define JOURNAL_HEADER_SEALED(h) \ + (!!(le32toh((h)->compatible_flags) & HEADER_COMPATIBLE_SEALED)) + +#define JOURNAL_HEADER_COMPRESSED_XZ(h) \ + (!!(le32toh((h)->incompatible_flags) & HEADER_INCOMPATIBLE_COMPRESSED_XZ)) + +#define JOURNAL_HEADER_COMPRESSED_LZ4(h) \ + (!!(le32toh((h)->incompatible_flags) & HEADER_INCOMPATIBLE_COMPRESSED_LZ4)) + +int journal_file_move_to_object(JournalFile *f, ObjectType type, uint64_t offset, Object **ret); + +uint64_t journal_file_entry_n_items(Object *o) _pure_; +uint64_t journal_file_entry_array_n_items(Object *o) _pure_; +uint64_t journal_file_hash_table_n_items(Object *o) _pure_; + +int journal_file_append_object(JournalFile *f, ObjectType type, uint64_t size, Object **ret, uint64_t *offset); +int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const struct iovec iovec[], unsigned n_iovec, uint64_t *seqno, Object **ret, uint64_t *offset); + +int journal_file_find_data_object(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset); +int journal_file_find_data_object_with_hash(JournalFile *f, const void *data, uint64_t size, uint64_t hash, Object **ret, uint64_t *offset); + +int journal_file_find_field_object(JournalFile *f, const void *field, uint64_t size, Object **ret, uint64_t *offset); +int journal_file_find_field_object_with_hash(JournalFile *f, const void *field, uint64_t size, uint64_t hash, Object **ret, uint64_t *offset); + +void journal_file_reset_location(JournalFile *f); +void journal_file_save_location(JournalFile *f, Object *o, uint64_t offset); +int journal_file_compare_locations(JournalFile *af, JournalFile *bf); +int journal_file_next_entry(JournalFile *f, uint64_t p, direction_t direction, Object **ret, uint64_t *offset); + +int journal_file_next_entry_for_data(JournalFile *f, Object *o, uint64_t p, uint64_t data_offset, direction_t direction, Object **ret, uint64_t *offset); + +int journal_file_move_to_entry_by_seqnum(JournalFile *f, uint64_t seqnum, direction_t direction, Object **ret, uint64_t *offset); +int journal_file_move_to_entry_by_realtime(JournalFile *f, uint64_t realtime, direction_t direction, Object **ret, uint64_t *offset); +int journal_file_move_to_entry_by_monotonic(JournalFile *f, sd_id128_t boot_id, uint64_t monotonic, direction_t direction, Object **ret, uint64_t *offset); + +int journal_file_move_to_entry_by_offset_for_data(JournalFile *f, uint64_t data_offset, uint64_t p, direction_t direction, Object **ret, uint64_t *offset); +int journal_file_move_to_entry_by_seqnum_for_data(JournalFile *f, uint64_t data_offset, uint64_t seqnum, direction_t direction, Object **ret, uint64_t *offset); +int journal_file_move_to_entry_by_realtime_for_data(JournalFile *f, uint64_t data_offset, uint64_t realtime, direction_t direction, Object **ret, uint64_t *offset); +int journal_file_move_to_entry_by_monotonic_for_data(JournalFile *f, uint64_t data_offset, sd_id128_t boot_id, uint64_t monotonic, direction_t direction, Object **ret, uint64_t *offset); + +int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint64_t p, uint64_t *seqnum, Object **ret, uint64_t *offset); + +void journal_file_dump(JournalFile *f); +void journal_file_print_header(JournalFile *f); + +int journal_file_rotate(JournalFile **f, bool compress, bool seal, Set *deferred_closes); + +void journal_file_post_change(JournalFile *f); +int journal_file_enable_post_change_timer(JournalFile *f, sd_event *e, usec_t t); + +void journal_reset_metrics(JournalMetrics *m); +void journal_default_metrics(JournalMetrics *m, int fd); + +int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *to); +int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot, usec_t *from, usec_t *to); + +bool journal_file_rotate_suggested(JournalFile *f, usec_t max_file_usec); + +int journal_file_map_data_hash_table(JournalFile *f); +int journal_file_map_field_hash_table(JournalFile *f); + +static inline bool JOURNAL_FILE_COMPRESS(JournalFile *f) { + assert(f); + return f->compress_xz || f->compress_lz4; +} diff --git a/src/libsystemd/src/sd-journal/journal-internal.h b/src/libsystemd/src/sd-journal/journal-internal.h new file mode 100644 index 0000000000..83f38b51ed --- /dev/null +++ b/src/libsystemd/src/sd-journal/journal-internal.h @@ -0,0 +1,144 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <inttypes.h> +#include <stdbool.h> +#include <sys/types.h> + +#include <systemd/sd-id128.h> +#include <systemd/sd-journal.h> + +#include "systemd-basic/hashmap.h" +#include "systemd-basic/list.h" +#include "systemd-basic/set.h" + +#include "journal-def.h" +#include "journal-file.h" + +typedef struct Match Match; +typedef struct Location Location; +typedef struct Directory Directory; + +typedef enum MatchType { + MATCH_DISCRETE, + MATCH_OR_TERM, + MATCH_AND_TERM +} MatchType; + +struct Match { + MatchType type; + Match *parent; + LIST_FIELDS(Match, matches); + + /* For concrete matches */ + char *data; + size_t size; + le64_t le_hash; + + /* For terms */ + LIST_HEAD(Match, matches); +}; + +struct Location { + LocationType type; + + bool seqnum_set; + bool realtime_set; + bool monotonic_set; + bool xor_hash_set; + + uint64_t seqnum; + sd_id128_t seqnum_id; + + uint64_t realtime; + + uint64_t monotonic; + sd_id128_t boot_id; + + uint64_t xor_hash; +}; + +struct Directory { + char *path; + int wd; + bool is_root; +}; + +struct sd_journal { + int toplevel_fd; + + char *path; + char *prefix; + + OrderedHashmap *files; + MMapCache *mmap; + + Location current_location; + + JournalFile *current_file; + uint64_t current_field; + + Match *level0, *level1, *level2; + + pid_t original_pid; + + int inotify_fd; + unsigned current_invalidate_counter, last_invalidate_counter; + usec_t last_process_usec; + + /* Iterating through unique fields and their data values */ + char *unique_field; + JournalFile *unique_file; + uint64_t unique_offset; + + /* Iterating through known fields */ + JournalFile *fields_file; + uint64_t fields_offset; + uint64_t fields_hash_table_index; + char *fields_buffer; + size_t fields_buffer_allocated; + + int flags; + + bool on_network:1; + bool no_new_files:1; + bool no_inotify:1; + bool unique_file_lost:1; /* File we were iterating over got + removed, and there were no more + files, so sd_j_enumerate_unique + will return a value equal to 0. */ + bool fields_file_lost:1; + bool has_runtime_files:1; + bool has_persistent_files:1; + + size_t data_threshold; + + Hashmap *directories_by_path; + Hashmap *directories_by_wd; + + Hashmap *errors; +}; + +char *journal_make_match_string(sd_journal *j); +void journal_print_header(sd_journal *j); + +#define JOURNAL_FOREACH_DATA_RETVAL(j, data, l, retval) \ + for (sd_journal_restart_data(j); ((retval) = sd_journal_enumerate_data((j), &(data), &(l))) > 0; ) diff --git a/src/libsystemd/src/sd-journal/journal-send.c b/src/libsystemd/src/sd-journal/journal-send.c new file mode 100644 index 0000000000..c168902e9a --- /dev/null +++ b/src/libsystemd/src/sd-journal/journal-send.c @@ -0,0 +1,575 @@ +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <fcntl.h> +#include <printf.h> +#include <stddef.h> +#include <sys/socket.h> +#include <sys/un.h> +#include <unistd.h> + +#define SD_JOURNAL_SUPPRESS_LOCATION + +#include <systemd/sd-journal.h> + +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/io-util.h" +#include "systemd-basic/memfd-util.h" +#include "systemd-basic/socket-util.h" +#include "systemd-basic/stdio-util.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/util.h" + +#define SNDBUF_SIZE (8*1024*1024) + +#define ALLOCA_CODE_FUNC(f, func) \ + do { \ + size_t _fl; \ + const char *_func = (func); \ + char **_f = &(f); \ + _fl = strlen(_func) + 1; \ + *_f = alloca(_fl + 10); \ + memcpy(*_f, "CODE_FUNC=", 10); \ + memcpy(*_f + 10, _func, _fl); \ + } while (false) + +/* We open a single fd, and we'll share it with the current process, + * all its threads, and all its subprocesses. This means we need to + * initialize it atomically, and need to operate on it atomically + * never assuming we are the only user */ + +static int journal_fd(void) { + int fd; + static int fd_plus_one = 0; + +retry: + if (fd_plus_one > 0) + return fd_plus_one - 1; + + fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0); + if (fd < 0) + return -errno; + + fd_inc_sndbuf(fd, SNDBUF_SIZE); + + if (!__sync_bool_compare_and_swap(&fd_plus_one, 0, fd+1)) { + safe_close(fd); + goto retry; + } + + return fd; +} + +_public_ int sd_journal_print(int priority, const char *format, ...) { + int r; + va_list ap; + + va_start(ap, format); + r = sd_journal_printv(priority, format, ap); + va_end(ap); + + return r; +} + +_public_ int sd_journal_printv(int priority, const char *format, va_list ap) { + + /* FIXME: Instead of limiting things to LINE_MAX we could do a + C99 variable-length array on the stack here in a loop. */ + + char buffer[8 + LINE_MAX], p[sizeof("PRIORITY=")-1 + DECIMAL_STR_MAX(int) + 1]; + struct iovec iov[2]; + + assert_return(priority >= 0, -EINVAL); + assert_return(priority <= 7, -EINVAL); + assert_return(format, -EINVAL); + + xsprintf(p, "PRIORITY=%i", priority & LOG_PRIMASK); + + memcpy(buffer, "MESSAGE=", 8); + vsnprintf(buffer+8, sizeof(buffer) - 8, format, ap); + + /* Strip trailing whitespace, keep prefix whitespace. */ + (void) strstrip(buffer); + + /* Suppress empty lines */ + if (isempty(buffer+8)) + return 0; + + zero(iov); + IOVEC_SET_STRING(iov[0], buffer); + IOVEC_SET_STRING(iov[1], p); + + return sd_journal_sendv(iov, 2); +} + +_printf_(1, 0) static int fill_iovec_sprintf(const char *format, va_list ap, int extra, struct iovec **_iov) { + PROTECT_ERRNO; + int r, n = 0, i = 0, j; + struct iovec *iov = NULL; + + assert(_iov); + + if (extra > 0) { + n = MAX(extra * 2, extra + 4); + iov = malloc0(n * sizeof(struct iovec)); + if (!iov) { + r = -ENOMEM; + goto fail; + } + + i = extra; + } + + while (format) { + struct iovec *c; + char *buffer; + va_list aq; + + if (i >= n) { + n = MAX(i*2, 4); + c = realloc(iov, n * sizeof(struct iovec)); + if (!c) { + r = -ENOMEM; + goto fail; + } + + iov = c; + } + + va_copy(aq, ap); + if (vasprintf(&buffer, format, aq) < 0) { + va_end(aq); + r = -ENOMEM; + goto fail; + } + va_end(aq); + + VA_FORMAT_ADVANCE(format, ap); + + (void) strstrip(buffer); /* strip trailing whitespace, keep prefixing whitespace */ + + IOVEC_SET_STRING(iov[i++], buffer); + + format = va_arg(ap, char *); + } + + *_iov = iov; + + return i; + +fail: + for (j = 0; j < i; j++) + free(iov[j].iov_base); + + free(iov); + + return r; +} + +_public_ int sd_journal_send(const char *format, ...) { + int r, i, j; + va_list ap; + struct iovec *iov = NULL; + + va_start(ap, format); + i = fill_iovec_sprintf(format, ap, 0, &iov); + va_end(ap); + + if (_unlikely_(i < 0)) { + r = i; + goto finish; + } + + r = sd_journal_sendv(iov, i); + +finish: + for (j = 0; j < i; j++) + free(iov[j].iov_base); + + free(iov); + + return r; +} + +_public_ int sd_journal_sendv(const struct iovec *iov, int n) { + PROTECT_ERRNO; + int fd, r; + _cleanup_close_ int buffer_fd = -1; + struct iovec *w; + uint64_t *l; + int i, j = 0; + static const union sockaddr_union sa = { + .un.sun_family = AF_UNIX, + .un.sun_path = "/run/systemd/journal/socket", + }; + struct msghdr mh = { + .msg_name = (struct sockaddr*) &sa.sa, + .msg_namelen = SOCKADDR_UN_LEN(sa.un), + }; + ssize_t k; + bool have_syslog_identifier = false; + bool seal = true; + + assert_return(iov, -EINVAL); + assert_return(n > 0, -EINVAL); + + w = newa(struct iovec, n * 5 + 3); + l = newa(uint64_t, n); + + for (i = 0; i < n; i++) { + char *c, *nl; + + if (_unlikely_(!iov[i].iov_base || iov[i].iov_len <= 1)) + return -EINVAL; + + c = memchr(iov[i].iov_base, '=', iov[i].iov_len); + if (_unlikely_(!c || c == iov[i].iov_base)) + return -EINVAL; + + have_syslog_identifier = have_syslog_identifier || + (c == (char *) iov[i].iov_base + 17 && + startswith(iov[i].iov_base, "SYSLOG_IDENTIFIER")); + + nl = memchr(iov[i].iov_base, '\n', iov[i].iov_len); + if (nl) { + if (_unlikely_(nl < c)) + return -EINVAL; + + /* Already includes a newline? Bummer, then + * let's write the variable name, then a + * newline, then the size (64bit LE), followed + * by the data and a final newline */ + + w[j].iov_base = iov[i].iov_base; + w[j].iov_len = c - (char*) iov[i].iov_base; + j++; + + IOVEC_SET_STRING(w[j++], "\n"); + + l[i] = htole64(iov[i].iov_len - (c - (char*) iov[i].iov_base) - 1); + w[j].iov_base = &l[i]; + w[j].iov_len = sizeof(uint64_t); + j++; + + w[j].iov_base = c + 1; + w[j].iov_len = iov[i].iov_len - (c - (char*) iov[i].iov_base) - 1; + j++; + + } else + /* Nothing special? Then just add the line and + * append a newline */ + w[j++] = iov[i]; + + IOVEC_SET_STRING(w[j++], "\n"); + } + + if (!have_syslog_identifier && + string_is_safe(program_invocation_short_name)) { + + /* Implicitly add program_invocation_short_name, if it + * is not set explicitly. We only do this for + * program_invocation_short_name, and nothing else + * since everything else is much nicer to retrieve + * from the outside. */ + + IOVEC_SET_STRING(w[j++], "SYSLOG_IDENTIFIER="); + IOVEC_SET_STRING(w[j++], program_invocation_short_name); + IOVEC_SET_STRING(w[j++], "\n"); + } + + fd = journal_fd(); + if (_unlikely_(fd < 0)) + return fd; + + mh.msg_iov = w; + mh.msg_iovlen = j; + + k = sendmsg(fd, &mh, MSG_NOSIGNAL); + if (k >= 0) + return 0; + + /* Fail silently if the journal is not available */ + if (errno == ENOENT) + return 0; + + if (errno != EMSGSIZE && errno != ENOBUFS) + return -errno; + + /* Message doesn't fit... Let's dump the data in a memfd or + * temporary file and just pass a file descriptor of it to the + * other side. + * + * For the temporary files we use /dev/shm instead of /tmp + * here, since we want this to be a tmpfs, and one that is + * available from early boot on and where unprivileged users + * can create files. */ + buffer_fd = memfd_new(NULL); + if (buffer_fd < 0) { + if (buffer_fd == -ENOSYS) { + buffer_fd = open_tmpfile_unlinkable("/dev/shm", O_RDWR | O_CLOEXEC); + if (buffer_fd < 0) + return buffer_fd; + + seal = false; + } else + return buffer_fd; + } + + n = writev(buffer_fd, w, j); + if (n < 0) + return -errno; + + if (seal) { + r = memfd_set_sealed(buffer_fd); + if (r < 0) + return r; + } + + r = send_one_fd_sa(fd, buffer_fd, mh.msg_name, mh.msg_namelen, 0); + if (r == -ENOENT) + /* Fail silently if the journal is not available */ + return 0; + return r; +} + +static int fill_iovec_perror_and_send(const char *message, int skip, struct iovec iov[]) { + PROTECT_ERRNO; + size_t n, k; + + k = isempty(message) ? 0 : strlen(message) + 2; + n = 8 + k + 256 + 1; + + for (;;) { + char buffer[n]; + char* j; + + errno = 0; + j = strerror_r(_saved_errno_, buffer + 8 + k, n - 8 - k); + if (errno == 0) { + char error[sizeof("ERRNO=")-1 + DECIMAL_STR_MAX(int) + 1]; + + if (j != buffer + 8 + k) + memmove(buffer + 8 + k, j, strlen(j)+1); + + memcpy(buffer, "MESSAGE=", 8); + + if (k > 0) { + memcpy(buffer + 8, message, k - 2); + memcpy(buffer + 8 + k - 2, ": ", 2); + } + + xsprintf(error, "ERRNO=%i", _saved_errno_); + + assert_cc(3 == LOG_ERR); + IOVEC_SET_STRING(iov[skip+0], "PRIORITY=3"); + IOVEC_SET_STRING(iov[skip+1], buffer); + IOVEC_SET_STRING(iov[skip+2], error); + + return sd_journal_sendv(iov, skip + 3); + } + + if (errno != ERANGE) + return -errno; + + n *= 2; + } +} + +_public_ int sd_journal_perror(const char *message) { + struct iovec iovec[3]; + + return fill_iovec_perror_and_send(message, 0, iovec); +} + +_public_ int sd_journal_stream_fd(const char *identifier, int priority, int level_prefix) { + static const union sockaddr_union sa = { + .un.sun_family = AF_UNIX, + .un.sun_path = "/run/systemd/journal/stdout", + }; + _cleanup_close_ int fd = -1; + char *header; + size_t l; + int r; + + assert_return(priority >= 0, -EINVAL); + assert_return(priority <= 7, -EINVAL); + + fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC, 0); + if (fd < 0) + return -errno; + + r = connect(fd, &sa.sa, SOCKADDR_UN_LEN(sa.un)); + if (r < 0) + return -errno; + + if (shutdown(fd, SHUT_RD) < 0) + return -errno; + + fd_inc_sndbuf(fd, SNDBUF_SIZE); + + if (!identifier) + identifier = ""; + + l = strlen(identifier); + header = alloca(l + 1 + 1 + 2 + 2 + 2 + 2 + 2); + + memcpy(header, identifier, l); + header[l++] = '\n'; + header[l++] = '\n'; /* unit id */ + header[l++] = '0' + priority; + header[l++] = '\n'; + header[l++] = '0' + !!level_prefix; + header[l++] = '\n'; + header[l++] = '0'; + header[l++] = '\n'; + header[l++] = '0'; + header[l++] = '\n'; + header[l++] = '0'; + header[l++] = '\n'; + + r = loop_write(fd, header, l, false); + if (r < 0) + return r; + + r = fd; + fd = -1; + return r; +} + +_public_ int sd_journal_print_with_location(int priority, const char *file, const char *line, const char *func, const char *format, ...) { + int r; + va_list ap; + + va_start(ap, format); + r = sd_journal_printv_with_location(priority, file, line, func, format, ap); + va_end(ap); + + return r; +} + +_public_ int sd_journal_printv_with_location(int priority, const char *file, const char *line, const char *func, const char *format, va_list ap) { + char buffer[8 + LINE_MAX], p[sizeof("PRIORITY=")-1 + DECIMAL_STR_MAX(int) + 1]; + struct iovec iov[5]; + char *f; + + assert_return(priority >= 0, -EINVAL); + assert_return(priority <= 7, -EINVAL); + assert_return(format, -EINVAL); + + xsprintf(p, "PRIORITY=%i", priority & LOG_PRIMASK); + + memcpy(buffer, "MESSAGE=", 8); + vsnprintf(buffer+8, sizeof(buffer) - 8, format, ap); + + /* Strip trailing whitespace, keep prefixing whitespace */ + (void) strstrip(buffer); + + /* Suppress empty lines */ + if (isempty(buffer+8)) + return 0; + + /* func is initialized from __func__ which is not a macro, but + * a static const char[], hence cannot easily be prefixed with + * CODE_FUNC=, hence let's do it manually here. */ + ALLOCA_CODE_FUNC(f, func); + + zero(iov); + IOVEC_SET_STRING(iov[0], buffer); + IOVEC_SET_STRING(iov[1], p); + IOVEC_SET_STRING(iov[2], file); + IOVEC_SET_STRING(iov[3], line); + IOVEC_SET_STRING(iov[4], f); + + return sd_journal_sendv(iov, ELEMENTSOF(iov)); +} + +_public_ int sd_journal_send_with_location(const char *file, const char *line, const char *func, const char *format, ...) { + int r, i, j; + va_list ap; + struct iovec *iov = NULL; + char *f; + + va_start(ap, format); + i = fill_iovec_sprintf(format, ap, 3, &iov); + va_end(ap); + + if (_unlikely_(i < 0)) { + r = i; + goto finish; + } + + ALLOCA_CODE_FUNC(f, func); + + IOVEC_SET_STRING(iov[0], file); + IOVEC_SET_STRING(iov[1], line); + IOVEC_SET_STRING(iov[2], f); + + r = sd_journal_sendv(iov, i); + +finish: + for (j = 3; j < i; j++) + free(iov[j].iov_base); + + free(iov); + + return r; +} + +_public_ int sd_journal_sendv_with_location( + const char *file, const char *line, + const char *func, + const struct iovec *iov, int n) { + + struct iovec *niov; + char *f; + + assert_return(iov, -EINVAL); + assert_return(n > 0, -EINVAL); + + niov = alloca(sizeof(struct iovec) * (n + 3)); + memcpy(niov, iov, sizeof(struct iovec) * n); + + ALLOCA_CODE_FUNC(f, func); + + IOVEC_SET_STRING(niov[n++], file); + IOVEC_SET_STRING(niov[n++], line); + IOVEC_SET_STRING(niov[n++], f); + + return sd_journal_sendv(niov, n); +} + +_public_ int sd_journal_perror_with_location( + const char *file, const char *line, + const char *func, + const char *message) { + + struct iovec iov[6]; + char *f; + + ALLOCA_CODE_FUNC(f, func); + + IOVEC_SET_STRING(iov[0], file); + IOVEC_SET_STRING(iov[1], line); + IOVEC_SET_STRING(iov[2], f); + + return fill_iovec_perror_and_send(message, 3, iov); +} diff --git a/src/libsystemd/src/sd-journal/journal-vacuum.c b/src/libsystemd/src/sd-journal/journal-vacuum.c new file mode 100644 index 0000000000..e5d49f0b5d --- /dev/null +++ b/src/libsystemd/src/sd-journal/journal-vacuum.c @@ -0,0 +1,350 @@ +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <fcntl.h> +#include <sys/stat.h> +#include <unistd.h> + +#include <systemd/sd-id128.h> + +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/dirent-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/util.h" +#include "systemd-basic/xattr-util.h" + +#include "journal-def.h" +#include "journal-file.h" +#include "journal-vacuum.h" + +struct vacuum_info { + uint64_t usage; + char *filename; + + uint64_t realtime; + + sd_id128_t seqnum_id; + uint64_t seqnum; + bool have_seqnum; +}; + +static int vacuum_compare(const void *_a, const void *_b) { + const struct vacuum_info *a, *b; + + a = _a; + b = _b; + + if (a->have_seqnum && b->have_seqnum && + sd_id128_equal(a->seqnum_id, b->seqnum_id)) { + if (a->seqnum < b->seqnum) + return -1; + else if (a->seqnum > b->seqnum) + return 1; + else + return 0; + } + + if (a->realtime < b->realtime) + return -1; + else if (a->realtime > b->realtime) + return 1; + else if (a->have_seqnum && b->have_seqnum) + return memcmp(&a->seqnum_id, &b->seqnum_id, 16); + else + return strcmp(a->filename, b->filename); +} + +static void patch_realtime( + int fd, + const char *fn, + const struct stat *st, + unsigned long long *realtime) { + + usec_t x, crtime = 0; + + /* The timestamp was determined by the file name, but let's + * see if the file might actually be older than the file name + * suggested... */ + + assert(fd >= 0); + assert(fn); + assert(st); + assert(realtime); + + x = timespec_load(&st->st_ctim); + if (x > 0 && x != USEC_INFINITY && x < *realtime) + *realtime = x; + + x = timespec_load(&st->st_atim); + if (x > 0 && x != USEC_INFINITY && x < *realtime) + *realtime = x; + + x = timespec_load(&st->st_mtim); + if (x > 0 && x != USEC_INFINITY && x < *realtime) + *realtime = x; + + /* Let's read the original creation time, if possible. Ideally + * we'd just query the creation time the FS might provide, but + * unfortunately there's currently no sane API to query + * it. Hence let's implement this manually... */ + + if (fd_getcrtime_at(fd, fn, &crtime, 0) >= 0) { + if (crtime < *realtime) + *realtime = crtime; + } +} + +static int journal_file_empty(int dir_fd, const char *name) { + _cleanup_close_ int fd; + struct stat st; + le64_t n_entries; + ssize_t n; + + fd = openat(dir_fd, name, O_RDONLY|O_CLOEXEC|O_NOFOLLOW|O_NONBLOCK|O_NOATIME); + if (fd < 0) { + /* Maybe failed due to O_NOATIME and lack of privileges? */ + fd = openat(dir_fd, name, O_RDONLY|O_CLOEXEC|O_NOFOLLOW|O_NONBLOCK); + if (fd < 0) + return -errno; + } + + if (fstat(fd, &st) < 0) + return -errno; + + /* If an offline file doesn't even have a header we consider it empty */ + if (st.st_size < (off_t) sizeof(Header)) + return 1; + + /* If the number of entries is empty, we consider it empty, too */ + n = pread(fd, &n_entries, sizeof(n_entries), offsetof(Header, n_entries)); + if (n < 0) + return -errno; + if (n != sizeof(n_entries)) + return -EIO; + + return le64toh(n_entries) <= 0; +} + +int journal_directory_vacuum( + const char *directory, + uint64_t max_use, + uint64_t n_max_files, + usec_t max_retention_usec, + usec_t *oldest_usec, + bool verbose) { + + _cleanup_closedir_ DIR *d = NULL; + struct vacuum_info *list = NULL; + unsigned n_list = 0, i, n_active_files = 0; + size_t n_allocated = 0; + uint64_t sum = 0, freed = 0; + usec_t retention_limit = 0; + char sbytes[FORMAT_BYTES_MAX]; + struct dirent *de; + int r; + + assert(directory); + + if (max_use <= 0 && max_retention_usec <= 0 && n_max_files <= 0) + return 0; + + if (max_retention_usec > 0) { + retention_limit = now(CLOCK_REALTIME); + if (retention_limit > max_retention_usec) + retention_limit -= max_retention_usec; + else + max_retention_usec = retention_limit = 0; + } + + d = opendir(directory); + if (!d) + return -errno; + + FOREACH_DIRENT_ALL(de, d, r = -errno; goto finish) { + + unsigned long long seqnum = 0, realtime; + _cleanup_free_ char *p = NULL; + sd_id128_t seqnum_id; + bool have_seqnum; + uint64_t size; + struct stat st; + size_t q; + + if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0) { + log_debug_errno(errno, "Failed to stat file %s while vacuuming, ignoring: %m", de->d_name); + continue; + } + + if (!S_ISREG(st.st_mode)) + continue; + + q = strlen(de->d_name); + + if (endswith(de->d_name, ".journal")) { + + /* Vacuum archived files. Active files are + * left around */ + + if (q < 1 + 32 + 1 + 16 + 1 + 16 + 8) { + n_active_files++; + continue; + } + + if (de->d_name[q-8-16-1] != '-' || + de->d_name[q-8-16-1-16-1] != '-' || + de->d_name[q-8-16-1-16-1-32-1] != '@') { + n_active_files++; + continue; + } + + p = strdup(de->d_name); + if (!p) { + r = -ENOMEM; + goto finish; + } + + de->d_name[q-8-16-1-16-1] = 0; + if (sd_id128_from_string(de->d_name + q-8-16-1-16-1-32, &seqnum_id) < 0) { + n_active_files++; + continue; + } + + if (sscanf(de->d_name + q-8-16-1-16, "%16llx-%16llx.journal", &seqnum, &realtime) != 2) { + n_active_files++; + continue; + } + + have_seqnum = true; + + } else if (endswith(de->d_name, ".journal~")) { + unsigned long long tmp; + + /* Vacuum corrupted files */ + + if (q < 1 + 16 + 1 + 16 + 8 + 1) { + n_active_files++; + continue; + } + + if (de->d_name[q-1-8-16-1] != '-' || + de->d_name[q-1-8-16-1-16-1] != '@') { + n_active_files++; + continue; + } + + p = strdup(de->d_name); + if (!p) { + r = -ENOMEM; + goto finish; + } + + if (sscanf(de->d_name + q-1-8-16-1-16, "%16llx-%16llx.journal~", &realtime, &tmp) != 2) { + n_active_files++; + continue; + } + + have_seqnum = false; + } else { + /* We do not vacuum unknown files! */ + log_debug("Not vacuuming unknown file %s.", de->d_name); + continue; + } + + size = 512UL * (uint64_t) st.st_blocks; + + r = journal_file_empty(dirfd(d), p); + if (r < 0) { + log_debug_errno(r, "Failed check if %s is empty, ignoring: %m", p); + continue; + } + if (r > 0) { + /* Always vacuum empty non-online files. */ + + if (unlinkat(dirfd(d), p, 0) >= 0) { + + log_full(verbose ? LOG_INFO : LOG_DEBUG, + "Deleted empty archived journal %s/%s (%s).", directory, p, format_bytes(sbytes, sizeof(sbytes), size)); + + freed += size; + } else if (errno != ENOENT) + log_warning_errno(errno, "Failed to delete empty archived journal %s/%s: %m", directory, p); + + continue; + } + + patch_realtime(dirfd(d), p, &st, &realtime); + + if (!GREEDY_REALLOC(list, n_allocated, n_list + 1)) { + r = -ENOMEM; + goto finish; + } + + list[n_list].filename = p; + list[n_list].usage = size; + list[n_list].seqnum = seqnum; + list[n_list].realtime = realtime; + list[n_list].seqnum_id = seqnum_id; + list[n_list].have_seqnum = have_seqnum; + n_list++; + + p = NULL; + sum += size; + } + + qsort_safe(list, n_list, sizeof(struct vacuum_info), vacuum_compare); + + for (i = 0; i < n_list; i++) { + unsigned left; + + left = n_active_files + n_list - i; + + if ((max_retention_usec <= 0 || list[i].realtime >= retention_limit) && + (max_use <= 0 || sum <= max_use) && + (n_max_files <= 0 || left <= n_max_files)) + break; + + if (unlinkat(dirfd(d), list[i].filename, 0) >= 0) { + log_full(verbose ? LOG_INFO : LOG_DEBUG, "Deleted archived journal %s/%s (%s).", directory, list[i].filename, format_bytes(sbytes, sizeof(sbytes), list[i].usage)); + freed += list[i].usage; + + if (list[i].usage < sum) + sum -= list[i].usage; + else + sum = 0; + + } else if (errno != ENOENT) + log_warning_errno(errno, "Failed to delete archived journal %s/%s: %m", directory, list[i].filename); + } + + if (oldest_usec && i < n_list && (*oldest_usec == 0 || list[i].realtime < *oldest_usec)) + *oldest_usec = list[i].realtime; + + r = 0; + +finish: + for (i = 0; i < n_list; i++) + free(list[i].filename); + free(list); + + log_full(verbose ? LOG_INFO : LOG_DEBUG, "Vacuuming done, freed %s of archived journals from %s.", format_bytes(sbytes, sizeof(sbytes), freed), directory); + + return r; +} diff --git a/src/libsystemd/src/sd-journal/journal-vacuum.h b/src/libsystemd/src/sd-journal/journal-vacuum.h new file mode 100644 index 0000000000..1e76aa856d --- /dev/null +++ b/src/libsystemd/src/sd-journal/journal-vacuum.h @@ -0,0 +1,27 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <inttypes.h> +#include <stdbool.h> + +#include "systemd-basic/time-util.h" + +int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t n_max_files, usec_t max_retention_usec, usec_t *oldest_usec, bool verbose); diff --git a/src/libsystemd/src/sd-journal/journal-verify.c b/src/libsystemd/src/sd-journal/journal-verify.c new file mode 100644 index 0000000000..756dbdaa76 --- /dev/null +++ b/src/libsystemd/src/sd-journal/journal-verify.c @@ -0,0 +1,1314 @@ +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <fcntl.h> +#include <stddef.h> +#include <sys/mman.h> +#include <unistd.h> + +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/fs-util.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/terminal-util.h" +#include "systemd-basic/util.h" + +#include "compress.h" +#include "journal-authenticate.h" +#include "journal-def.h" +#include "journal-file.h" +#include "journal-verify.h" +#include "lookup3.h" + +static void draw_progress(uint64_t p, usec_t *last_usec) { + unsigned n, i, j, k; + usec_t z, x; + + if (!on_tty()) + return; + + z = now(CLOCK_MONOTONIC); + x = *last_usec; + + if (x != 0 && x + 40 * USEC_PER_MSEC > z) + return; + + *last_usec = z; + + n = (3 * columns()) / 4; + j = (n * (unsigned) p) / 65535ULL; + k = n - j; + + fputs("\r", stdout); + if (colors_enabled()) + fputs("\x1B[?25l" ANSI_HIGHLIGHT_GREEN, stdout); + + for (i = 0; i < j; i++) + fputs("\xe2\x96\x88", stdout); + + fputs(ANSI_NORMAL, stdout); + + for (i = 0; i < k; i++) + fputs("\xe2\x96\x91", stdout); + + printf(" %3"PRIu64"%%", 100U * p / 65535U); + + fputs("\r", stdout); + if (colors_enabled()) + fputs("\x1B[?25h", stdout); + + fflush(stdout); +} + +static uint64_t scale_progress(uint64_t scale, uint64_t p, uint64_t m) { + + /* Calculates scale * p / m, but handles m == 0 safely, and saturates */ + + if (p >= m || m == 0) + return scale; + + return scale * p / m; +} + +static void flush_progress(void) { + unsigned n, i; + + if (!on_tty()) + return; + + n = (3 * columns()) / 4; + + putchar('\r'); + + for (i = 0; i < n + 5; i++) + putchar(' '); + + putchar('\r'); + fflush(stdout); +} + +#define debug(_offset, _fmt, ...) do { \ + flush_progress(); \ + log_debug(OFSfmt": " _fmt, _offset, ##__VA_ARGS__); \ + } while (0) + +#define warning(_offset, _fmt, ...) do { \ + flush_progress(); \ + log_warning(OFSfmt": " _fmt, _offset, ##__VA_ARGS__); \ + } while (0) + +#define error(_offset, _fmt, ...) do { \ + flush_progress(); \ + log_error(OFSfmt": " _fmt, (uint64_t)_offset, ##__VA_ARGS__); \ + } while (0) + +#define error_errno(_offset, error, _fmt, ...) do { \ + flush_progress(); \ + log_error_errno(error, OFSfmt": " _fmt, (uint64_t)_offset, ##__VA_ARGS__); \ + } while (0) + +static int journal_file_object_verify(JournalFile *f, uint64_t offset, Object *o) { + uint64_t i; + + assert(f); + assert(offset); + assert(o); + + /* This does various superficial tests about the length an + * possible field values. It does not follow any references to + * other objects. */ + + if ((o->object.flags & OBJECT_COMPRESSED_XZ) && + o->object.type != OBJECT_DATA) { + error(offset, "Found compressed object that isn't of type DATA, which is not allowed."); + return -EBADMSG; + } + + switch (o->object.type) { + + case OBJECT_DATA: { + uint64_t h1, h2; + int compression, r; + + if (le64toh(o->data.entry_offset) == 0) + warning(offset, "Unused data (entry_offset==0)"); + + if ((le64toh(o->data.entry_offset) == 0) ^ (le64toh(o->data.n_entries) == 0)) { + error(offset, "Bad n_entries: %"PRIu64, o->data.n_entries); + return -EBADMSG; + } + + if (le64toh(o->object.size) - offsetof(DataObject, payload) <= 0) { + error(offset, "Bad object size (<= %zu): %"PRIu64, + offsetof(DataObject, payload), + le64toh(o->object.size)); + return -EBADMSG; + } + + h1 = le64toh(o->data.hash); + + compression = o->object.flags & OBJECT_COMPRESSION_MASK; + if (compression) { + _cleanup_free_ void *b = NULL; + size_t alloc = 0, b_size; + + r = decompress_blob(compression, + o->data.payload, + le64toh(o->object.size) - offsetof(Object, data.payload), + &b, &alloc, &b_size, 0); + if (r < 0) { + error_errno(offset, r, "%s decompression failed: %m", + object_compressed_to_string(compression)); + return r; + } + + h2 = hash64(b, b_size); + } else + h2 = hash64(o->data.payload, le64toh(o->object.size) - offsetof(Object, data.payload)); + + if (h1 != h2) { + error(offset, "Invalid hash (%08"PRIx64" vs. %08"PRIx64, h1, h2); + return -EBADMSG; + } + + if (!VALID64(o->data.next_hash_offset) || + !VALID64(o->data.next_field_offset) || + !VALID64(o->data.entry_offset) || + !VALID64(o->data.entry_array_offset)) { + error(offset, "Invalid offset (next_hash_offset="OFSfmt", next_field_offset="OFSfmt", entry_offset="OFSfmt", entry_array_offset="OFSfmt, + o->data.next_hash_offset, + o->data.next_field_offset, + o->data.entry_offset, + o->data.entry_array_offset); + return -EBADMSG; + } + + break; + } + + case OBJECT_FIELD: + if (le64toh(o->object.size) - offsetof(FieldObject, payload) <= 0) { + error(offset, + "Bad field size (<= %zu): %"PRIu64, + offsetof(FieldObject, payload), + le64toh(o->object.size)); + return -EBADMSG; + } + + if (!VALID64(o->field.next_hash_offset) || + !VALID64(o->field.head_data_offset)) { + error(offset, + "Invalid offset (next_hash_offset="OFSfmt", head_data_offset="OFSfmt, + o->field.next_hash_offset, + o->field.head_data_offset); + return -EBADMSG; + } + break; + + case OBJECT_ENTRY: + if ((le64toh(o->object.size) - offsetof(EntryObject, items)) % sizeof(EntryItem) != 0) { + error(offset, + "Bad entry size (<= %zu): %"PRIu64, + offsetof(EntryObject, items), + le64toh(o->object.size)); + return -EBADMSG; + } + + if ((le64toh(o->object.size) - offsetof(EntryObject, items)) / sizeof(EntryItem) <= 0) { + error(offset, + "Invalid number items in entry: %"PRIu64, + (le64toh(o->object.size) - offsetof(EntryObject, items)) / sizeof(EntryItem)); + return -EBADMSG; + } + + if (le64toh(o->entry.seqnum) <= 0) { + error(offset, + "Invalid entry seqnum: %"PRIx64, + le64toh(o->entry.seqnum)); + return -EBADMSG; + } + + if (!VALID_REALTIME(le64toh(o->entry.realtime))) { + error(offset, + "Invalid entry realtime timestamp: %"PRIu64, + le64toh(o->entry.realtime)); + return -EBADMSG; + } + + if (!VALID_MONOTONIC(le64toh(o->entry.monotonic))) { + error(offset, + "Invalid entry monotonic timestamp: %"PRIu64, + le64toh(o->entry.monotonic)); + return -EBADMSG; + } + + for (i = 0; i < journal_file_entry_n_items(o); i++) { + if (o->entry.items[i].object_offset == 0 || + !VALID64(o->entry.items[i].object_offset)) { + error(offset, + "Invalid entry item (%"PRIu64"/%"PRIu64" offset: "OFSfmt, + i, journal_file_entry_n_items(o), + o->entry.items[i].object_offset); + return -EBADMSG; + } + } + + break; + + case OBJECT_DATA_HASH_TABLE: + case OBJECT_FIELD_HASH_TABLE: + if ((le64toh(o->object.size) - offsetof(HashTableObject, items)) % sizeof(HashItem) != 0 || + (le64toh(o->object.size) - offsetof(HashTableObject, items)) / sizeof(HashItem) <= 0) { + error(offset, + "Invalid %s hash table size: %"PRIu64, + o->object.type == OBJECT_DATA_HASH_TABLE ? "data" : "field", + le64toh(o->object.size)); + return -EBADMSG; + } + + for (i = 0; i < journal_file_hash_table_n_items(o); i++) { + if (o->hash_table.items[i].head_hash_offset != 0 && + !VALID64(le64toh(o->hash_table.items[i].head_hash_offset))) { + error(offset, + "Invalid %s hash table item (%"PRIu64"/%"PRIu64") head_hash_offset: "OFSfmt, + o->object.type == OBJECT_DATA_HASH_TABLE ? "data" : "field", + i, journal_file_hash_table_n_items(o), + le64toh(o->hash_table.items[i].head_hash_offset)); + return -EBADMSG; + } + if (o->hash_table.items[i].tail_hash_offset != 0 && + !VALID64(le64toh(o->hash_table.items[i].tail_hash_offset))) { + error(offset, + "Invalid %s hash table item (%"PRIu64"/%"PRIu64") tail_hash_offset: "OFSfmt, + o->object.type == OBJECT_DATA_HASH_TABLE ? "data" : "field", + i, journal_file_hash_table_n_items(o), + le64toh(o->hash_table.items[i].tail_hash_offset)); + return -EBADMSG; + } + + if ((o->hash_table.items[i].head_hash_offset != 0) != + (o->hash_table.items[i].tail_hash_offset != 0)) { + error(offset, + "Invalid %s hash table item (%"PRIu64"/%"PRIu64"): head_hash_offset="OFSfmt" tail_hash_offset="OFSfmt, + o->object.type == OBJECT_DATA_HASH_TABLE ? "data" : "field", + i, journal_file_hash_table_n_items(o), + le64toh(o->hash_table.items[i].head_hash_offset), + le64toh(o->hash_table.items[i].tail_hash_offset)); + return -EBADMSG; + } + } + + break; + + case OBJECT_ENTRY_ARRAY: + if ((le64toh(o->object.size) - offsetof(EntryArrayObject, items)) % sizeof(le64_t) != 0 || + (le64toh(o->object.size) - offsetof(EntryArrayObject, items)) / sizeof(le64_t) <= 0) { + error(offset, + "Invalid object entry array size: %"PRIu64, + le64toh(o->object.size)); + return -EBADMSG; + } + + if (!VALID64(o->entry_array.next_entry_array_offset)) { + error(offset, + "Invalid object entry array next_entry_array_offset: "OFSfmt, + o->entry_array.next_entry_array_offset); + return -EBADMSG; + } + + for (i = 0; i < journal_file_entry_array_n_items(o); i++) + if (le64toh(o->entry_array.items[i]) != 0 && + !VALID64(le64toh(o->entry_array.items[i]))) { + error(offset, + "Invalid object entry array item (%"PRIu64"/%"PRIu64"): "OFSfmt, + i, journal_file_entry_array_n_items(o), + le64toh(o->entry_array.items[i])); + return -EBADMSG; + } + + break; + + case OBJECT_TAG: + if (le64toh(o->object.size) != sizeof(TagObject)) { + error(offset, + "Invalid object tag size: %"PRIu64, + le64toh(o->object.size)); + return -EBADMSG; + } + + if (!VALID_EPOCH(o->tag.epoch)) { + error(offset, + "Invalid object tag epoch: %"PRIu64, + o->tag.epoch); + return -EBADMSG; + } + + break; + } + + return 0; +} + +static int write_uint64(int fd, uint64_t p) { + ssize_t k; + + k = write(fd, &p, sizeof(p)); + if (k < 0) + return -errno; + if (k != sizeof(p)) + return -EIO; + + return 0; +} + +static int contains_uint64(MMapCache *m, int fd, uint64_t n, uint64_t p) { + uint64_t a, b; + int r; + + assert(m); + assert(fd >= 0); + + /* Bisection ... */ + + a = 0; b = n; + while (a < b) { + uint64_t c, *z; + + c = (a + b) / 2; + + r = mmap_cache_get(m, fd, PROT_READ|PROT_WRITE, 0, false, c * sizeof(uint64_t), sizeof(uint64_t), NULL, (void **) &z); + if (r < 0) + return r; + + if (*z == p) + return 1; + + if (a + 1 >= b) + return 0; + + if (p < *z) + b = c; + else + a = c; + } + + return 0; +} + +static int entry_points_to_data( + JournalFile *f, + int entry_fd, + uint64_t n_entries, + uint64_t entry_p, + uint64_t data_p) { + + int r; + uint64_t i, n, a; + Object *o; + bool found = false; + + assert(f); + assert(entry_fd >= 0); + + if (!contains_uint64(f->mmap, entry_fd, n_entries, entry_p)) { + error(data_p, "Data object references invalid entry at "OFSfmt, entry_p); + return -EBADMSG; + } + + r = journal_file_move_to_object(f, OBJECT_ENTRY, entry_p, &o); + if (r < 0) + return r; + + n = journal_file_entry_n_items(o); + for (i = 0; i < n; i++) + if (le64toh(o->entry.items[i].object_offset) == data_p) { + found = true; + break; + } + + if (!found) { + error(entry_p, "Data object at "OFSfmt" not referenced by linked entry", data_p); + return -EBADMSG; + } + + /* Check if this entry is also in main entry array. Since the + * main entry array has already been verified we can rely on + * its consistency. */ + + i = 0; + n = le64toh(f->header->n_entries); + a = le64toh(f->header->entry_array_offset); + + while (i < n) { + uint64_t m, u; + + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + + m = journal_file_entry_array_n_items(o); + u = MIN(n - i, m); + + if (entry_p <= le64toh(o->entry_array.items[u-1])) { + uint64_t x, y, z; + + x = 0; + y = u; + + while (x < y) { + z = (x + y) / 2; + + if (le64toh(o->entry_array.items[z]) == entry_p) + return 0; + + if (x + 1 >= y) + break; + + if (entry_p < le64toh(o->entry_array.items[z])) + y = z; + else + x = z; + } + + error(entry_p, "Entry object doesn't exist in main entry array"); + return -EBADMSG; + } + + i += u; + a = le64toh(o->entry_array.next_entry_array_offset); + } + + return 0; +} + +static int verify_data( + JournalFile *f, + Object *o, uint64_t p, + int entry_fd, uint64_t n_entries, + int entry_array_fd, uint64_t n_entry_arrays) { + + uint64_t i, n, a, last, q; + int r; + + assert(f); + assert(o); + assert(entry_fd >= 0); + assert(entry_array_fd >= 0); + + n = le64toh(o->data.n_entries); + a = le64toh(o->data.entry_array_offset); + + /* Entry array means at least two objects */ + if (a && n < 2) { + error(p, "Entry array present (entry_array_offset="OFSfmt", but n_entries=%"PRIu64")", a, n); + return -EBADMSG; + } + + if (n == 0) + return 0; + + /* We already checked that earlier */ + assert(o->data.entry_offset); + + last = q = le64toh(o->data.entry_offset); + r = entry_points_to_data(f, entry_fd, n_entries, q, p); + if (r < 0) + return r; + + i = 1; + while (i < n) { + uint64_t next, m, j; + + if (a == 0) { + error(p, "Array chain too short"); + return -EBADMSG; + } + + if (!contains_uint64(f->mmap, entry_array_fd, n_entry_arrays, a)) { + error(p, "Invalid array offset "OFSfmt, a); + return -EBADMSG; + } + + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + + next = le64toh(o->entry_array.next_entry_array_offset); + if (next != 0 && next <= a) { + error(p, "Array chain has cycle (jumps back from "OFSfmt" to "OFSfmt")", a, next); + return -EBADMSG; + } + + m = journal_file_entry_array_n_items(o); + for (j = 0; i < n && j < m; i++, j++) { + + q = le64toh(o->entry_array.items[j]); + if (q <= last) { + error(p, "Data object's entry array not sorted"); + return -EBADMSG; + } + last = q; + + r = entry_points_to_data(f, entry_fd, n_entries, q, p); + if (r < 0) + return r; + + /* Pointer might have moved, reposition */ + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + } + + a = next; + } + + return 0; +} + +static int verify_hash_table( + JournalFile *f, + int data_fd, uint64_t n_data, + int entry_fd, uint64_t n_entries, + int entry_array_fd, uint64_t n_entry_arrays, + usec_t *last_usec, + bool show_progress) { + + uint64_t i, n; + int r; + + assert(f); + assert(data_fd >= 0); + assert(entry_fd >= 0); + assert(entry_array_fd >= 0); + assert(last_usec); + + n = le64toh(f->header->data_hash_table_size) / sizeof(HashItem); + if (n <= 0) + return 0; + + r = journal_file_map_data_hash_table(f); + if (r < 0) + return log_error_errno(r, "Failed to map data hash table: %m"); + + for (i = 0; i < n; i++) { + uint64_t last = 0, p; + + if (show_progress) + draw_progress(0xC000 + scale_progress(0x3FFF, i, n), last_usec); + + p = le64toh(f->data_hash_table[i].head_hash_offset); + while (p != 0) { + Object *o; + uint64_t next; + + if (!contains_uint64(f->mmap, data_fd, n_data, p)) { + error(p, "Invalid data object at hash entry %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + next = le64toh(o->data.next_hash_offset); + if (next != 0 && next <= p) { + error(p, "Hash chain has a cycle in hash entry %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + + if (le64toh(o->data.hash) % n != i) { + error(p, "Hash value mismatch in hash entry %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + + r = verify_data(f, o, p, entry_fd, n_entries, entry_array_fd, n_entry_arrays); + if (r < 0) + return r; + + last = p; + p = next; + } + + if (last != le64toh(f->data_hash_table[i].tail_hash_offset)) { + error(p, "Tail hash pointer mismatch in hash table"); + return -EBADMSG; + } + } + + return 0; +} + +static int data_object_in_hash_table(JournalFile *f, uint64_t hash, uint64_t p) { + uint64_t n, h, q; + int r; + assert(f); + + n = le64toh(f->header->data_hash_table_size) / sizeof(HashItem); + if (n <= 0) + return 0; + + r = journal_file_map_data_hash_table(f); + if (r < 0) + return log_error_errno(r, "Failed to map data hash table: %m"); + + h = hash % n; + + q = le64toh(f->data_hash_table[h].head_hash_offset); + while (q != 0) { + Object *o; + + if (p == q) + return 1; + + r = journal_file_move_to_object(f, OBJECT_DATA, q, &o); + if (r < 0) + return r; + + q = le64toh(o->data.next_hash_offset); + } + + return 0; +} + +static int verify_entry( + JournalFile *f, + Object *o, uint64_t p, + int data_fd, uint64_t n_data) { + + uint64_t i, n; + int r; + + assert(f); + assert(o); + assert(data_fd >= 0); + + n = journal_file_entry_n_items(o); + for (i = 0; i < n; i++) { + uint64_t q, h; + Object *u; + + q = le64toh(o->entry.items[i].object_offset); + h = le64toh(o->entry.items[i].hash); + + if (!contains_uint64(f->mmap, data_fd, n_data, q)) { + error(p, "Invalid data object of entry"); + return -EBADMSG; + } + + r = journal_file_move_to_object(f, OBJECT_DATA, q, &u); + if (r < 0) + return r; + + if (le64toh(u->data.hash) != h) { + error(p, "Hash mismatch for data object of entry"); + return -EBADMSG; + } + + r = data_object_in_hash_table(f, h, q); + if (r < 0) + return r; + if (r == 0) { + error(p, "Data object missing from hash table"); + return -EBADMSG; + } + } + + return 0; +} + +static int verify_entry_array( + JournalFile *f, + int data_fd, uint64_t n_data, + int entry_fd, uint64_t n_entries, + int entry_array_fd, uint64_t n_entry_arrays, + usec_t *last_usec, + bool show_progress) { + + uint64_t i = 0, a, n, last = 0; + int r; + + assert(f); + assert(data_fd >= 0); + assert(entry_fd >= 0); + assert(entry_array_fd >= 0); + assert(last_usec); + + n = le64toh(f->header->n_entries); + a = le64toh(f->header->entry_array_offset); + while (i < n) { + uint64_t next, m, j; + Object *o; + + if (show_progress) + draw_progress(0x8000 + scale_progress(0x3FFF, i, n), last_usec); + + if (a == 0) { + error(a, "Array chain too short at %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + + if (!contains_uint64(f->mmap, entry_array_fd, n_entry_arrays, a)) { + error(a, "Invalid array %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + + next = le64toh(o->entry_array.next_entry_array_offset); + if (next != 0 && next <= a) { + error(a, "Array chain has cycle at %"PRIu64" of %"PRIu64" (jumps back from to "OFSfmt")", i, n, next); + return -EBADMSG; + } + + m = journal_file_entry_array_n_items(o); + for (j = 0; i < n && j < m; i++, j++) { + uint64_t p; + + p = le64toh(o->entry_array.items[j]); + if (p <= last) { + error(a, "Entry array not sorted at %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + last = p; + + if (!contains_uint64(f->mmap, entry_fd, n_entries, p)) { + error(a, "Invalid array entry at %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + r = verify_entry(f, o, p, data_fd, n_data); + if (r < 0) + return r; + + /* Pointer might have moved, reposition */ + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + } + + a = next; + } + + return 0; +} + +int journal_file_verify( + JournalFile *f, + const char *key, + usec_t *first_contained, usec_t *last_validated, usec_t *last_contained, + bool show_progress) { + int r; + Object *o; + uint64_t p = 0, last_epoch = 0, last_tag_realtime = 0, last_sealed_realtime = 0; + + uint64_t entry_seqnum = 0, entry_monotonic = 0, entry_realtime = 0; + sd_id128_t entry_boot_id; + bool entry_seqnum_set = false, entry_monotonic_set = false, entry_realtime_set = false, found_main_entry_array = false; + uint64_t n_weird = 0, n_objects = 0, n_entries = 0, n_data = 0, n_fields = 0, n_data_hash_tables = 0, n_field_hash_tables = 0, n_entry_arrays = 0, n_tags = 0; + usec_t last_usec = 0; + int data_fd = -1, entry_fd = -1, entry_array_fd = -1; + unsigned i; + bool found_last = false; + const char *tmp_dir = NULL; + +#ifdef HAVE_GCRYPT + uint64_t last_tag = 0; +#endif + assert(f); + + if (key) { +#ifdef HAVE_GCRYPT + r = journal_file_parse_verification_key(f, key); + if (r < 0) { + log_error("Failed to parse seed."); + return r; + } +#else + return -EOPNOTSUPP; +#endif + } else if (f->seal) + return -ENOKEY; + + r = var_tmp_dir(&tmp_dir); + if (r < 0) { + log_error_errno(r, "Failed to determine temporary directory: %m"); + goto fail; + } + + data_fd = open_tmpfile_unlinkable(tmp_dir, O_RDWR | O_CLOEXEC); + if (data_fd < 0) { + r = log_error_errno(data_fd, "Failed to create data file: %m"); + goto fail; + } + + entry_fd = open_tmpfile_unlinkable(tmp_dir, O_RDWR | O_CLOEXEC); + if (entry_fd < 0) { + r = log_error_errno(entry_fd, "Failed to create entry file: %m"); + goto fail; + } + + entry_array_fd = open_tmpfile_unlinkable(tmp_dir, O_RDWR | O_CLOEXEC); + if (entry_array_fd < 0) { + r = log_error_errno(entry_array_fd, + "Failed to create entry array file: %m"); + goto fail; + } + + if (le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_SUPPORTED) { + log_error("Cannot verify file with unknown extensions."); + r = -EOPNOTSUPP; + goto fail; + } + + for (i = 0; i < sizeof(f->header->reserved); i++) + if (f->header->reserved[i] != 0) { + error(offsetof(Header, reserved[i]), "Reserved field is non-zero"); + r = -EBADMSG; + goto fail; + } + + /* First iteration: we go through all objects, verify the + * superficial structure, headers, hashes. */ + + p = le64toh(f->header->header_size); + for (;;) { + /* Early exit if there are no objects in the file, at all */ + if (le64toh(f->header->tail_object_offset) == 0) + break; + + if (show_progress) + draw_progress(scale_progress(0x7FFF, p, le64toh(f->header->tail_object_offset)), &last_usec); + + r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &o); + if (r < 0) { + error(p, "Invalid object"); + goto fail; + } + + if (p > le64toh(f->header->tail_object_offset)) { + error(offsetof(Header, tail_object_offset), "Invalid tail object pointer"); + r = -EBADMSG; + goto fail; + } + + n_objects++; + + r = journal_file_object_verify(f, p, o); + if (r < 0) { + error_errno(p, r, "Invalid object contents: %m"); + goto fail; + } + + if ((o->object.flags & OBJECT_COMPRESSED_XZ) && + (o->object.flags & OBJECT_COMPRESSED_LZ4)) { + error(p, "Objected with double compression"); + r = -EINVAL; + goto fail; + } + + if ((o->object.flags & OBJECT_COMPRESSED_XZ) && !JOURNAL_HEADER_COMPRESSED_XZ(f->header)) { + error(p, "XZ compressed object in file without XZ compression"); + r = -EBADMSG; + goto fail; + } + + if ((o->object.flags & OBJECT_COMPRESSED_LZ4) && !JOURNAL_HEADER_COMPRESSED_LZ4(f->header)) { + error(p, "LZ4 compressed object in file without LZ4 compression"); + r = -EBADMSG; + goto fail; + } + + switch (o->object.type) { + + case OBJECT_DATA: + r = write_uint64(data_fd, p); + if (r < 0) + goto fail; + + n_data++; + break; + + case OBJECT_FIELD: + n_fields++; + break; + + case OBJECT_ENTRY: + if (JOURNAL_HEADER_SEALED(f->header) && n_tags <= 0) { + error(p, "First entry before first tag"); + r = -EBADMSG; + goto fail; + } + + r = write_uint64(entry_fd, p); + if (r < 0) + goto fail; + + if (le64toh(o->entry.realtime) < last_tag_realtime) { + error(p, "Older entry after newer tag"); + r = -EBADMSG; + goto fail; + } + + if (!entry_seqnum_set && + le64toh(o->entry.seqnum) != le64toh(f->header->head_entry_seqnum)) { + error(p, "Head entry sequence number incorrect"); + r = -EBADMSG; + goto fail; + } + + if (entry_seqnum_set && + entry_seqnum >= le64toh(o->entry.seqnum)) { + error(p, "Entry sequence number out of synchronization"); + r = -EBADMSG; + goto fail; + } + + entry_seqnum = le64toh(o->entry.seqnum); + entry_seqnum_set = true; + + if (entry_monotonic_set && + sd_id128_equal(entry_boot_id, o->entry.boot_id) && + entry_monotonic > le64toh(o->entry.monotonic)) { + error(p, "Entry timestamp out of synchronization"); + r = -EBADMSG; + goto fail; + } + + entry_monotonic = le64toh(o->entry.monotonic); + entry_boot_id = o->entry.boot_id; + entry_monotonic_set = true; + + if (!entry_realtime_set && + le64toh(o->entry.realtime) != le64toh(f->header->head_entry_realtime)) { + error(p, "Head entry realtime timestamp incorrect"); + r = -EBADMSG; + goto fail; + } + + entry_realtime = le64toh(o->entry.realtime); + entry_realtime_set = true; + + n_entries++; + break; + + case OBJECT_DATA_HASH_TABLE: + if (n_data_hash_tables > 1) { + error(p, "More than one data hash table"); + r = -EBADMSG; + goto fail; + } + + if (le64toh(f->header->data_hash_table_offset) != p + offsetof(HashTableObject, items) || + le64toh(f->header->data_hash_table_size) != le64toh(o->object.size) - offsetof(HashTableObject, items)) { + error(p, "header fields for data hash table invalid"); + r = -EBADMSG; + goto fail; + } + + n_data_hash_tables++; + break; + + case OBJECT_FIELD_HASH_TABLE: + if (n_field_hash_tables > 1) { + error(p, "More than one field hash table"); + r = -EBADMSG; + goto fail; + } + + if (le64toh(f->header->field_hash_table_offset) != p + offsetof(HashTableObject, items) || + le64toh(f->header->field_hash_table_size) != le64toh(o->object.size) - offsetof(HashTableObject, items)) { + error(p, "Header fields for field hash table invalid"); + r = -EBADMSG; + goto fail; + } + + n_field_hash_tables++; + break; + + case OBJECT_ENTRY_ARRAY: + r = write_uint64(entry_array_fd, p); + if (r < 0) + goto fail; + + if (p == le64toh(f->header->entry_array_offset)) { + if (found_main_entry_array) { + error(p, "More than one main entry array"); + r = -EBADMSG; + goto fail; + } + + found_main_entry_array = true; + } + + n_entry_arrays++; + break; + + case OBJECT_TAG: + if (!JOURNAL_HEADER_SEALED(f->header)) { + error(p, "Tag object in file without sealing"); + r = -EBADMSG; + goto fail; + } + + if (le64toh(o->tag.seqnum) != n_tags + 1) { + error(p, "Tag sequence number out of synchronization"); + r = -EBADMSG; + goto fail; + } + + if (le64toh(o->tag.epoch) < last_epoch) { + error(p, "Epoch sequence out of synchronization"); + r = -EBADMSG; + goto fail; + } + +#ifdef HAVE_GCRYPT + if (f->seal) { + uint64_t q, rt; + + debug(p, "Checking tag %"PRIu64"...", le64toh(o->tag.seqnum)); + + rt = f->fss_start_usec + o->tag.epoch * f->fss_interval_usec; + if (entry_realtime_set && entry_realtime >= rt + f->fss_interval_usec) { + error(p, "tag/entry realtime timestamp out of synchronization"); + r = -EBADMSG; + goto fail; + } + + /* OK, now we know the epoch. So let's now set + * it, and calculate the HMAC for everything + * since the last tag. */ + r = journal_file_fsprg_seek(f, le64toh(o->tag.epoch)); + if (r < 0) + goto fail; + + r = journal_file_hmac_start(f); + if (r < 0) + goto fail; + + if (last_tag == 0) { + r = journal_file_hmac_put_header(f); + if (r < 0) + goto fail; + + q = le64toh(f->header->header_size); + } else + q = last_tag; + + while (q <= p) { + r = journal_file_move_to_object(f, OBJECT_UNUSED, q, &o); + if (r < 0) + goto fail; + + r = journal_file_hmac_put_object(f, OBJECT_UNUSED, o, q); + if (r < 0) + goto fail; + + q = q + ALIGN64(le64toh(o->object.size)); + } + + /* Position might have changed, let's reposition things */ + r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &o); + if (r < 0) + goto fail; + + if (memcmp(o->tag.tag, gcry_md_read(f->hmac, 0), TAG_LENGTH) != 0) { + error(p, "Tag failed verification"); + r = -EBADMSG; + goto fail; + } + + f->hmac_running = false; + last_tag_realtime = rt; + last_sealed_realtime = entry_realtime; + } + + last_tag = p + ALIGN64(le64toh(o->object.size)); +#endif + + last_epoch = le64toh(o->tag.epoch); + + n_tags++; + break; + + default: + n_weird++; + } + + if (p == le64toh(f->header->tail_object_offset)) { + found_last = true; + break; + } + + p = p + ALIGN64(le64toh(o->object.size)); + }; + + if (!found_last && le64toh(f->header->tail_object_offset) != 0) { + error(le64toh(f->header->tail_object_offset), "Tail object pointer dead"); + r = -EBADMSG; + goto fail; + } + + if (n_objects != le64toh(f->header->n_objects)) { + error(offsetof(Header, n_objects), "Object number mismatch"); + r = -EBADMSG; + goto fail; + } + + if (n_entries != le64toh(f->header->n_entries)) { + error(offsetof(Header, n_entries), "Entry number mismatch"); + r = -EBADMSG; + goto fail; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_data) && + n_data != le64toh(f->header->n_data)) { + error(offsetof(Header, n_data), "Data number mismatch"); + r = -EBADMSG; + goto fail; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_fields) && + n_fields != le64toh(f->header->n_fields)) { + error(offsetof(Header, n_fields), "Field number mismatch"); + r = -EBADMSG; + goto fail; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_tags) && + n_tags != le64toh(f->header->n_tags)) { + error(offsetof(Header, n_tags), "Tag number mismatch"); + r = -EBADMSG; + goto fail; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays) && + n_entry_arrays != le64toh(f->header->n_entry_arrays)) { + error(offsetof(Header, n_entry_arrays), "Entry array number mismatch"); + r = -EBADMSG; + goto fail; + } + + if (!found_main_entry_array && le64toh(f->header->entry_array_offset) != 0) { + error(0, "Missing entry array"); + r = -EBADMSG; + goto fail; + } + + if (entry_seqnum_set && + entry_seqnum != le64toh(f->header->tail_entry_seqnum)) { + error(offsetof(Header, tail_entry_seqnum), "Invalid tail seqnum"); + r = -EBADMSG; + goto fail; + } + + if (entry_monotonic_set && + (!sd_id128_equal(entry_boot_id, f->header->boot_id) || + entry_monotonic != le64toh(f->header->tail_entry_monotonic))) { + error(0, "Invalid tail monotonic timestamp"); + r = -EBADMSG; + goto fail; + } + + if (entry_realtime_set && entry_realtime != le64toh(f->header->tail_entry_realtime)) { + error(0, "Invalid tail realtime timestamp"); + r = -EBADMSG; + goto fail; + } + + /* Second iteration: we follow all objects referenced from the + * two entry points: the object hash table and the entry + * array. We also check that everything referenced (directly + * or indirectly) in the data hash table also exists in the + * entry array, and vice versa. Note that we do not care for + * unreferenced objects. We only care that everything that is + * referenced is consistent. */ + + r = verify_entry_array(f, + data_fd, n_data, + entry_fd, n_entries, + entry_array_fd, n_entry_arrays, + &last_usec, + show_progress); + if (r < 0) + goto fail; + + r = verify_hash_table(f, + data_fd, n_data, + entry_fd, n_entries, + entry_array_fd, n_entry_arrays, + &last_usec, + show_progress); + if (r < 0) + goto fail; + + if (show_progress) + flush_progress(); + + mmap_cache_close_fd(f->mmap, data_fd); + mmap_cache_close_fd(f->mmap, entry_fd); + mmap_cache_close_fd(f->mmap, entry_array_fd); + + safe_close(data_fd); + safe_close(entry_fd); + safe_close(entry_array_fd); + + if (first_contained) + *first_contained = le64toh(f->header->head_entry_realtime); + if (last_validated) + *last_validated = last_sealed_realtime; + if (last_contained) + *last_contained = le64toh(f->header->tail_entry_realtime); + + return 0; + +fail: + if (show_progress) + flush_progress(); + + log_error("File corruption detected at %s:"OFSfmt" (of %llu bytes, %"PRIu64"%%).", + f->path, + p, + (unsigned long long) f->last_stat.st_size, + 100 * p / f->last_stat.st_size); + + if (data_fd >= 0) { + mmap_cache_close_fd(f->mmap, data_fd); + safe_close(data_fd); + } + + if (entry_fd >= 0) { + mmap_cache_close_fd(f->mmap, entry_fd); + safe_close(entry_fd); + } + + if (entry_array_fd >= 0) { + mmap_cache_close_fd(f->mmap, entry_array_fd); + safe_close(entry_array_fd); + } + + return r; +} diff --git a/src/libsystemd/src/sd-journal/journal-verify.h b/src/libsystemd/src/sd-journal/journal-verify.h new file mode 100644 index 0000000000..8f0eaf6daa --- /dev/null +++ b/src/libsystemd/src/sd-journal/journal-verify.h @@ -0,0 +1,24 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "journal-file.h" + +int journal_file_verify(JournalFile *f, const char *key, usec_t *first_contained, usec_t *last_validated, usec_t *last_contained, bool show_progress); diff --git a/src/libsystemd/src/sd-journal/lookup3.c b/src/libsystemd/src/sd-journal/lookup3.c new file mode 100644 index 0000000000..d8f1a4977d --- /dev/null +++ b/src/libsystemd/src/sd-journal/lookup3.c @@ -0,0 +1,1009 @@ +/* Slightly modified by Lennart Poettering, to avoid name clashes, and + * unexport a few functions. */ + +#include "lookup3.h" + +/* +------------------------------------------------------------------------------- +lookup3.c, by Bob Jenkins, May 2006, Public Domain. + +These are functions for producing 32-bit hashes for hash table lookup. +hashword(), hashlittle(), hashlittle2(), hashbig(), mix(), and final() +are externally useful functions. Routines to test the hash are included +if SELF_TEST is defined. You can use this free for any purpose. It's in +the public domain. It has no warranty. + +You probably want to use hashlittle(). hashlittle() and hashbig() +hash byte arrays. hashlittle() is faster than hashbig() on +little-endian machines. Intel and AMD are little-endian machines. +On second thought, you probably want hashlittle2(), which is identical to +hashlittle() except it returns two 32-bit hashes for the price of one. +You could implement hashbig2() if you wanted but I haven't bothered here. + +If you want to find a hash of, say, exactly 7 integers, do + a = i1; b = i2; c = i3; + mix(a,b,c); + a += i4; b += i5; c += i6; + mix(a,b,c); + a += i7; + final(a,b,c); +then use c as the hash value. If you have a variable length array of +4-byte integers to hash, use hashword(). If you have a byte array (like +a character string), use hashlittle(). If you have several byte arrays, or +a mix of things, see the comments above hashlittle(). + +Why is this so big? I read 12 bytes at a time into 3 4-byte integers, +then mix those integers. This is fast (you can do a lot more thorough +mixing with 12*3 instructions on 3 integers than you can with 3 instructions +on 1 byte), but shoehorning those bytes into integers efficiently is messy. +------------------------------------------------------------------------------- +*/ +/* #define SELF_TEST 1 */ + +#include <stdint.h> /* defines uint32_t etc */ +#include <stdio.h> /* defines printf for tests */ +#include <sys/param.h> /* attempt to define endianness */ +#include <time.h> /* defines time_t for timings in the test */ +#ifdef linux +# include <endian.h> /* attempt to define endianness */ +#endif + +/* + * My best guess at if you are big-endian or little-endian. This may + * need adjustment. + */ +#if (defined(__BYTE_ORDER) && defined(__LITTLE_ENDIAN) && \ + __BYTE_ORDER == __LITTLE_ENDIAN) || \ + (defined(i386) || defined(__i386__) || defined(__i486__) || \ + defined(__i586__) || defined(__i686__) || defined(vax) || defined(MIPSEL)) +# define HASH_LITTLE_ENDIAN 1 +# define HASH_BIG_ENDIAN 0 +#elif (defined(__BYTE_ORDER) && defined(__BIG_ENDIAN) && \ + __BYTE_ORDER == __BIG_ENDIAN) || \ + (defined(sparc) || defined(POWERPC) || defined(mc68000) || defined(sel)) +# define HASH_LITTLE_ENDIAN 0 +# define HASH_BIG_ENDIAN 1 +#else +# define HASH_LITTLE_ENDIAN 0 +# define HASH_BIG_ENDIAN 0 +#endif + +#define hashsize(n) ((uint32_t)1<<(n)) +#define hashmask(n) (hashsize(n)-1) +#define rot(x,k) (((x)<<(k)) | ((x)>>(32-(k)))) + +/* +------------------------------------------------------------------------------- +mix -- mix 3 32-bit values reversibly. + +This is reversible, so any information in (a,b,c) before mix() is +still in (a,b,c) after mix(). + +If four pairs of (a,b,c) inputs are run through mix(), or through +mix() in reverse, there are at least 32 bits of the output that +are sometimes the same for one pair and different for another pair. +This was tested for: +* pairs that differed by one bit, by two bits, in any combination + of top bits of (a,b,c), or in any combination of bottom bits of + (a,b,c). +* "differ" is defined as +, -, ^, or ~^. For + and -, I transformed + the output delta to a Gray code (a^(a>>1)) so a string of 1's (as + is commonly produced by subtraction) look like a single 1-bit + difference. +* the base values were pseudorandom, all zero but one bit set, or + all zero plus a counter that starts at zero. + +Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that +satisfy this are + 4 6 8 16 19 4 + 9 15 3 18 27 15 + 14 9 3 7 17 3 +Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing +for "differ" defined as + with a one-bit base and a two-bit delta. I +used http://burtleburtle.net/bob/hash/avalanche.html to choose +the operations, constants, and arrangements of the variables. + +This does not achieve avalanche. There are input bits of (a,b,c) +that fail to affect some output bits of (a,b,c), especially of a. The +most thoroughly mixed value is c, but it doesn't really even achieve +avalanche in c. + +This allows some parallelism. Read-after-writes are good at doubling +the number of bits affected, so the goal of mixing pulls in the opposite +direction as the goal of parallelism. I did what I could. Rotates +seem to cost as much as shifts on every machine I could lay my hands +on, and rotates are much kinder to the top and bottom bits, so I used +rotates. +------------------------------------------------------------------------------- +*/ +#define mix(a,b,c) \ +{ \ + a -= c; a ^= rot(c, 4); c += b; \ + b -= a; b ^= rot(a, 6); a += c; \ + c -= b; c ^= rot(b, 8); b += a; \ + a -= c; a ^= rot(c,16); c += b; \ + b -= a; b ^= rot(a,19); a += c; \ + c -= b; c ^= rot(b, 4); b += a; \ +} + +/* +------------------------------------------------------------------------------- +final -- final mixing of 3 32-bit values (a,b,c) into c + +Pairs of (a,b,c) values differing in only a few bits will usually +produce values of c that look totally different. This was tested for +* pairs that differed by one bit, by two bits, in any combination + of top bits of (a,b,c), or in any combination of bottom bits of + (a,b,c). +* "differ" is defined as +, -, ^, or ~^. For + and -, I transformed + the output delta to a Gray code (a^(a>>1)) so a string of 1's (as + is commonly produced by subtraction) look like a single 1-bit + difference. +* the base values were pseudorandom, all zero but one bit set, or + all zero plus a counter that starts at zero. + +These constants passed: + 14 11 25 16 4 14 24 + 12 14 25 16 4 14 24 +and these came close: + 4 8 15 26 3 22 24 + 10 8 15 26 3 22 24 + 11 8 15 26 3 22 24 +------------------------------------------------------------------------------- +*/ +#define final(a,b,c) \ +{ \ + c ^= b; c -= rot(b,14); \ + a ^= c; a -= rot(c,11); \ + b ^= a; b -= rot(a,25); \ + c ^= b; c -= rot(b,16); \ + a ^= c; a -= rot(c,4); \ + b ^= a; b -= rot(a,14); \ + c ^= b; c -= rot(b,24); \ +} + +/* +-------------------------------------------------------------------- + This works on all machines. To be useful, it requires + -- that the key be an array of uint32_t's, and + -- that the length be the number of uint32_t's in the key + + The function hashword() is identical to hashlittle() on little-endian + machines, and identical to hashbig() on big-endian machines, + except that the length has to be measured in uint32_ts rather than in + bytes. hashlittle() is more complicated than hashword() only because + hashlittle() has to dance around fitting the key bytes into registers. +-------------------------------------------------------------------- +*/ +uint32_t jenkins_hashword( +const uint32_t *k, /* the key, an array of uint32_t values */ +size_t length, /* the length of the key, in uint32_ts */ +uint32_t initval) /* the previous hash, or an arbitrary value */ +{ + uint32_t a,b,c; + + /* Set up the internal state */ + a = b = c = 0xdeadbeef + (((uint32_t)length)<<2) + initval; + + /*------------------------------------------------- handle most of the key */ + while (length > 3) + { + a += k[0]; + b += k[1]; + c += k[2]; + mix(a,b,c); + length -= 3; + k += 3; + } + + /*------------------------------------------- handle the last 3 uint32_t's */ + switch(length) /* all the case statements fall through */ + { + case 3 : c+=k[2]; + case 2 : b+=k[1]; + case 1 : a+=k[0]; + final(a,b,c); + case 0: /* case 0: nothing left to add */ + break; + } + /*------------------------------------------------------ report the result */ + return c; +} + + +/* +-------------------------------------------------------------------- +hashword2() -- same as hashword(), but take two seeds and return two +32-bit values. pc and pb must both be nonnull, and *pc and *pb must +both be initialized with seeds. If you pass in (*pb)==0, the output +(*pc) will be the same as the return value from hashword(). +-------------------------------------------------------------------- +*/ +void jenkins_hashword2 ( +const uint32_t *k, /* the key, an array of uint32_t values */ +size_t length, /* the length of the key, in uint32_ts */ +uint32_t *pc, /* IN: seed OUT: primary hash value */ +uint32_t *pb) /* IN: more seed OUT: secondary hash value */ +{ + uint32_t a,b,c; + + /* Set up the internal state */ + a = b = c = 0xdeadbeef + ((uint32_t)(length<<2)) + *pc; + c += *pb; + + /*------------------------------------------------- handle most of the key */ + while (length > 3) + { + a += k[0]; + b += k[1]; + c += k[2]; + mix(a,b,c); + length -= 3; + k += 3; + } + + /*------------------------------------------- handle the last 3 uint32_t's */ + switch(length) /* all the case statements fall through */ + { + case 3 : c+=k[2]; + case 2 : b+=k[1]; + case 1 : a+=k[0]; + final(a,b,c); + case 0: /* case 0: nothing left to add */ + break; + } + /*------------------------------------------------------ report the result */ + *pc=c; *pb=b; +} + + +/* +------------------------------------------------------------------------------- +hashlittle() -- hash a variable-length key into a 32-bit value + k : the key (the unaligned variable-length array of bytes) + length : the length of the key, counting by bytes + initval : can be any 4-byte value +Returns a 32-bit value. Every bit of the key affects every bit of +the return value. Two keys differing by one or two bits will have +totally different hash values. + +The best hash table sizes are powers of 2. There is no need to do +mod a prime (mod is sooo slow!). If you need less than 32 bits, +use a bitmask. For example, if you need only 10 bits, do + h = (h & hashmask(10)); +In which case, the hash table should have hashsize(10) elements. + +If you are hashing n strings (uint8_t **)k, do it like this: + for (i=0, h=0; i<n; ++i) h = hashlittle( k[i], len[i], h); + +By Bob Jenkins, 2006. bob_jenkins@burtleburtle.net. You may use this +code any way you wish, private, educational, or commercial. It's free. + +Use for hash table lookup, or anything where one collision in 2^^32 is +acceptable. Do NOT use for cryptographic purposes. +------------------------------------------------------------------------------- +*/ + +uint32_t jenkins_hashlittle( const void *key, size_t length, uint32_t initval) +{ + uint32_t a,b,c; /* internal state */ + union { const void *ptr; size_t i; } u; /* needed for Mac Powerbook G4 */ + + /* Set up the internal state */ + a = b = c = 0xdeadbeef + ((uint32_t)length) + initval; + + u.ptr = key; + if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) { + const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */ + + /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */ + while (length > 12) + { + a += k[0]; + b += k[1]; + c += k[2]; + mix(a,b,c); + length -= 12; + k += 3; + } + + /*----------------------------- handle the last (probably partial) block */ + /* + * "k[2]&0xffffff" actually reads beyond the end of the string, but + * then masks off the part it's not allowed to read. Because the + * string is aligned, the masked-off tail is in the same word as the + * rest of the string. Every machine with memory protection I've seen + * does it on word boundaries, so is OK with this. But VALGRIND will + * still catch it and complain. The masking trick does make the hash + * noticeably faster for short strings (like English words). + */ +#if !defined(VALGRIND) && !defined(__SANITIZE_ADDRESS__) + + switch(length) + { + case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; + case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break; + case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break; + case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break; + case 8 : b+=k[1]; a+=k[0]; break; + case 7 : b+=k[1]&0xffffff; a+=k[0]; break; + case 6 : b+=k[1]&0xffff; a+=k[0]; break; + case 5 : b+=k[1]&0xff; a+=k[0]; break; + case 4 : a+=k[0]; break; + case 3 : a+=k[0]&0xffffff; break; + case 2 : a+=k[0]&0xffff; break; + case 1 : a+=k[0]&0xff; break; + case 0 : return c; /* zero length strings require no mixing */ + } + +#else /* make valgrind happy */ + { + const uint8_t *k8 = (const uint8_t *) k; + + switch(length) + { + case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; + case 11: c+=((uint32_t)k8[10])<<16; /* fall through */ + case 10: c+=((uint32_t)k8[9])<<8; /* fall through */ + case 9 : c+=k8[8]; /* fall through */ + case 8 : b+=k[1]; a+=k[0]; break; + case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */ + case 6 : b+=((uint32_t)k8[5])<<8; /* fall through */ + case 5 : b+=k8[4]; /* fall through */ + case 4 : a+=k[0]; break; + case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */ + case 2 : a+=((uint32_t)k8[1])<<8; /* fall through */ + case 1 : a+=k8[0]; break; + case 0 : return c; + } + } + +#endif /* !valgrind */ + + } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) { + const uint16_t *k = (const uint16_t *)key; /* read 16-bit chunks */ + const uint8_t *k8; + + /*--------------- all but last block: aligned reads and different mixing */ + while (length > 12) + { + a += k[0] + (((uint32_t)k[1])<<16); + b += k[2] + (((uint32_t)k[3])<<16); + c += k[4] + (((uint32_t)k[5])<<16); + mix(a,b,c); + length -= 12; + k += 6; + } + + /*----------------------------- handle the last (probably partial) block */ + k8 = (const uint8_t *)k; + switch(length) + { + case 12: c+=k[4]+(((uint32_t)k[5])<<16); + b+=k[2]+(((uint32_t)k[3])<<16); + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 11: c+=((uint32_t)k8[10])<<16; /* fall through */ + case 10: c+=k[4]; + b+=k[2]+(((uint32_t)k[3])<<16); + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 9 : c+=k8[8]; /* fall through */ + case 8 : b+=k[2]+(((uint32_t)k[3])<<16); + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */ + case 6 : b+=k[2]; + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 5 : b+=k8[4]; /* fall through */ + case 4 : a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */ + case 2 : a+=k[0]; + break; + case 1 : a+=k8[0]; + break; + case 0 : return c; /* zero length requires no mixing */ + } + + } else { /* need to read the key one byte at a time */ + const uint8_t *k = (const uint8_t *)key; + + /*--------------- all but the last block: affect some 32 bits of (a,b,c) */ + while (length > 12) + { + a += k[0]; + a += ((uint32_t)k[1])<<8; + a += ((uint32_t)k[2])<<16; + a += ((uint32_t)k[3])<<24; + b += k[4]; + b += ((uint32_t)k[5])<<8; + b += ((uint32_t)k[6])<<16; + b += ((uint32_t)k[7])<<24; + c += k[8]; + c += ((uint32_t)k[9])<<8; + c += ((uint32_t)k[10])<<16; + c += ((uint32_t)k[11])<<24; + mix(a,b,c); + length -= 12; + k += 12; + } + + /*-------------------------------- last block: affect all 32 bits of (c) */ + switch(length) /* all the case statements fall through */ + { + case 12: c+=((uint32_t)k[11])<<24; + case 11: c+=((uint32_t)k[10])<<16; + case 10: c+=((uint32_t)k[9])<<8; + case 9 : c+=k[8]; + case 8 : b+=((uint32_t)k[7])<<24; + case 7 : b+=((uint32_t)k[6])<<16; + case 6 : b+=((uint32_t)k[5])<<8; + case 5 : b+=k[4]; + case 4 : a+=((uint32_t)k[3])<<24; + case 3 : a+=((uint32_t)k[2])<<16; + case 2 : a+=((uint32_t)k[1])<<8; + case 1 : a+=k[0]; + break; + case 0 : return c; + } + } + + final(a,b,c); + return c; +} + + +/* + * hashlittle2: return 2 32-bit hash values + * + * This is identical to hashlittle(), except it returns two 32-bit hash + * values instead of just one. This is good enough for hash table + * lookup with 2^^64 buckets, or if you want a second hash if you're not + * happy with the first, or if you want a probably-unique 64-bit ID for + * the key. *pc is better mixed than *pb, so use *pc first. If you want + * a 64-bit value do something like "*pc + (((uint64_t)*pb)<<32)". + */ +void jenkins_hashlittle2( + const void *key, /* the key to hash */ + size_t length, /* length of the key */ + uint32_t *pc, /* IN: primary initval, OUT: primary hash */ + uint32_t *pb) /* IN: secondary initval, OUT: secondary hash */ +{ + uint32_t a,b,c; /* internal state */ + union { const void *ptr; size_t i; } u; /* needed for Mac Powerbook G4 */ + + /* Set up the internal state */ + a = b = c = 0xdeadbeef + ((uint32_t)length) + *pc; + c += *pb; + + u.ptr = key; + if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) { + const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */ + + /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */ + while (length > 12) + { + a += k[0]; + b += k[1]; + c += k[2]; + mix(a,b,c); + length -= 12; + k += 3; + } + + /*----------------------------- handle the last (probably partial) block */ + /* + * "k[2]&0xffffff" actually reads beyond the end of the string, but + * then masks off the part it's not allowed to read. Because the + * string is aligned, the masked-off tail is in the same word as the + * rest of the string. Every machine with memory protection I've seen + * does it on word boundaries, so is OK with this. But VALGRIND will + * still catch it and complain. The masking trick does make the hash + * noticeably faster for short strings (like English words). + */ +#if !defined(VALGRIND) && !defined(__SANITIZE_ADDRESS__) + + switch(length) + { + case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; + case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break; + case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break; + case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break; + case 8 : b+=k[1]; a+=k[0]; break; + case 7 : b+=k[1]&0xffffff; a+=k[0]; break; + case 6 : b+=k[1]&0xffff; a+=k[0]; break; + case 5 : b+=k[1]&0xff; a+=k[0]; break; + case 4 : a+=k[0]; break; + case 3 : a+=k[0]&0xffffff; break; + case 2 : a+=k[0]&0xffff; break; + case 1 : a+=k[0]&0xff; break; + case 0 : *pc=c; *pb=b; return; /* zero length strings require no mixing */ + } + +#else /* make valgrind happy */ + + { + const uint8_t *k8 = (const uint8_t *)k; + switch(length) + { + case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; + case 11: c+=((uint32_t)k8[10])<<16; /* fall through */ + case 10: c+=((uint32_t)k8[9])<<8; /* fall through */ + case 9 : c+=k8[8]; /* fall through */ + case 8 : b+=k[1]; a+=k[0]; break; + case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */ + case 6 : b+=((uint32_t)k8[5])<<8; /* fall through */ + case 5 : b+=k8[4]; /* fall through */ + case 4 : a+=k[0]; break; + case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */ + case 2 : a+=((uint32_t)k8[1])<<8; /* fall through */ + case 1 : a+=k8[0]; break; + case 0 : *pc=c; *pb=b; return; /* zero length strings require no mixing */ + } + } + +#endif /* !valgrind */ + + } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) { + const uint16_t *k = (const uint16_t *)key; /* read 16-bit chunks */ + const uint8_t *k8; + + /*--------------- all but last block: aligned reads and different mixing */ + while (length > 12) + { + a += k[0] + (((uint32_t)k[1])<<16); + b += k[2] + (((uint32_t)k[3])<<16); + c += k[4] + (((uint32_t)k[5])<<16); + mix(a,b,c); + length -= 12; + k += 6; + } + + /*----------------------------- handle the last (probably partial) block */ + k8 = (const uint8_t *)k; + switch(length) + { + case 12: c+=k[4]+(((uint32_t)k[5])<<16); + b+=k[2]+(((uint32_t)k[3])<<16); + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 11: c+=((uint32_t)k8[10])<<16; /* fall through */ + case 10: c+=k[4]; + b+=k[2]+(((uint32_t)k[3])<<16); + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 9 : c+=k8[8]; /* fall through */ + case 8 : b+=k[2]+(((uint32_t)k[3])<<16); + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */ + case 6 : b+=k[2]; + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 5 : b+=k8[4]; /* fall through */ + case 4 : a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */ + case 2 : a+=k[0]; + break; + case 1 : a+=k8[0]; + break; + case 0 : *pc=c; *pb=b; return; /* zero length strings require no mixing */ + } + + } else { /* need to read the key one byte at a time */ + const uint8_t *k = (const uint8_t *)key; + + /*--------------- all but the last block: affect some 32 bits of (a,b,c) */ + while (length > 12) + { + a += k[0]; + a += ((uint32_t)k[1])<<8; + a += ((uint32_t)k[2])<<16; + a += ((uint32_t)k[3])<<24; + b += k[4]; + b += ((uint32_t)k[5])<<8; + b += ((uint32_t)k[6])<<16; + b += ((uint32_t)k[7])<<24; + c += k[8]; + c += ((uint32_t)k[9])<<8; + c += ((uint32_t)k[10])<<16; + c += ((uint32_t)k[11])<<24; + mix(a,b,c); + length -= 12; + k += 12; + } + + /*-------------------------------- last block: affect all 32 bits of (c) */ + switch(length) /* all the case statements fall through */ + { + case 12: c+=((uint32_t)k[11])<<24; + case 11: c+=((uint32_t)k[10])<<16; + case 10: c+=((uint32_t)k[9])<<8; + case 9 : c+=k[8]; + case 8 : b+=((uint32_t)k[7])<<24; + case 7 : b+=((uint32_t)k[6])<<16; + case 6 : b+=((uint32_t)k[5])<<8; + case 5 : b+=k[4]; + case 4 : a+=((uint32_t)k[3])<<24; + case 3 : a+=((uint32_t)k[2])<<16; + case 2 : a+=((uint32_t)k[1])<<8; + case 1 : a+=k[0]; + break; + case 0 : *pc=c; *pb=b; return; /* zero length strings require no mixing */ + } + } + + final(a,b,c); + *pc=c; *pb=b; +} + + + +/* + * hashbig(): + * This is the same as hashword() on big-endian machines. It is different + * from hashlittle() on all machines. hashbig() takes advantage of + * big-endian byte ordering. + */ +uint32_t jenkins_hashbig( const void *key, size_t length, uint32_t initval) +{ + uint32_t a,b,c; + union { const void *ptr; size_t i; } u; /* to cast key to (size_t) happily */ + + /* Set up the internal state */ + a = b = c = 0xdeadbeef + ((uint32_t)length) + initval; + + u.ptr = key; + if (HASH_BIG_ENDIAN && ((u.i & 0x3) == 0)) { + const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */ + + /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */ + while (length > 12) + { + a += k[0]; + b += k[1]; + c += k[2]; + mix(a,b,c); + length -= 12; + k += 3; + } + + /*----------------------------- handle the last (probably partial) block */ + /* + * "k[2]<<8" actually reads beyond the end of the string, but + * then shifts out the part it's not allowed to read. Because the + * string is aligned, the illegal read is in the same word as the + * rest of the string. Every machine with memory protection I've seen + * does it on word boundaries, so is OK with this. But VALGRIND will + * still catch it and complain. The masking trick does make the hash + * noticeably faster for short strings (like English words). + */ +#if !defined(VALGRIND) && !defined(__SANITIZE_ADDRESS__) + + switch(length) + { + case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; + case 11: c+=k[2]&0xffffff00; b+=k[1]; a+=k[0]; break; + case 10: c+=k[2]&0xffff0000; b+=k[1]; a+=k[0]; break; + case 9 : c+=k[2]&0xff000000; b+=k[1]; a+=k[0]; break; + case 8 : b+=k[1]; a+=k[0]; break; + case 7 : b+=k[1]&0xffffff00; a+=k[0]; break; + case 6 : b+=k[1]&0xffff0000; a+=k[0]; break; + case 5 : b+=k[1]&0xff000000; a+=k[0]; break; + case 4 : a+=k[0]; break; + case 3 : a+=k[0]&0xffffff00; break; + case 2 : a+=k[0]&0xffff0000; break; + case 1 : a+=k[0]&0xff000000; break; + case 0 : return c; /* zero length strings require no mixing */ + } + +#else /* make valgrind happy */ + + { + const uint8_t *k8 = (const uint8_t *)k; + switch(length) /* all the case statements fall through */ + { + case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; + case 11: c+=((uint32_t)k8[10])<<8; /* fall through */ + case 10: c+=((uint32_t)k8[9])<<16; /* fall through */ + case 9 : c+=((uint32_t)k8[8])<<24; /* fall through */ + case 8 : b+=k[1]; a+=k[0]; break; + case 7 : b+=((uint32_t)k8[6])<<8; /* fall through */ + case 6 : b+=((uint32_t)k8[5])<<16; /* fall through */ + case 5 : b+=((uint32_t)k8[4])<<24; /* fall through */ + case 4 : a+=k[0]; break; + case 3 : a+=((uint32_t)k8[2])<<8; /* fall through */ + case 2 : a+=((uint32_t)k8[1])<<16; /* fall through */ + case 1 : a+=((uint32_t)k8[0])<<24; break; + case 0 : return c; + } + } + +#endif /* !VALGRIND */ + + } else { /* need to read the key one byte at a time */ + const uint8_t *k = (const uint8_t *)key; + + /*--------------- all but the last block: affect some 32 bits of (a,b,c) */ + while (length > 12) + { + a += ((uint32_t)k[0])<<24; + a += ((uint32_t)k[1])<<16; + a += ((uint32_t)k[2])<<8; + a += ((uint32_t)k[3]); + b += ((uint32_t)k[4])<<24; + b += ((uint32_t)k[5])<<16; + b += ((uint32_t)k[6])<<8; + b += ((uint32_t)k[7]); + c += ((uint32_t)k[8])<<24; + c += ((uint32_t)k[9])<<16; + c += ((uint32_t)k[10])<<8; + c += ((uint32_t)k[11]); + mix(a,b,c); + length -= 12; + k += 12; + } + + /*-------------------------------- last block: affect all 32 bits of (c) */ + switch(length) /* all the case statements fall through */ + { + case 12: c+=k[11]; + case 11: c+=((uint32_t)k[10])<<8; + case 10: c+=((uint32_t)k[9])<<16; + case 9 : c+=((uint32_t)k[8])<<24; + case 8 : b+=k[7]; + case 7 : b+=((uint32_t)k[6])<<8; + case 6 : b+=((uint32_t)k[5])<<16; + case 5 : b+=((uint32_t)k[4])<<24; + case 4 : a+=k[3]; + case 3 : a+=((uint32_t)k[2])<<8; + case 2 : a+=((uint32_t)k[1])<<16; + case 1 : a+=((uint32_t)k[0])<<24; + break; + case 0 : return c; + } + } + + final(a,b,c); + return c; +} + + +#ifdef SELF_TEST + +/* used for timings */ +void driver1() +{ + uint8_t buf[256]; + uint32_t i; + uint32_t h=0; + time_t a,z; + + time(&a); + for (i=0; i<256; ++i) buf[i] = 'x'; + for (i=0; i<1; ++i) + { + h = hashlittle(&buf[0],1,h); + } + time(&z); + if (z-a > 0) printf("time %d %.8x\n", z-a, h); +} + +/* check that every input bit changes every output bit half the time */ +#define HASHSTATE 1 +#define HASHLEN 1 +#define MAXPAIR 60 +#define MAXLEN 70 +void driver2() +{ + uint8_t qa[MAXLEN+1], qb[MAXLEN+2], *a = &qa[0], *b = &qb[1]; + uint32_t c[HASHSTATE], d[HASHSTATE], i=0, j=0, k, l, m=0, z; + uint32_t e[HASHSTATE],f[HASHSTATE],g[HASHSTATE],h[HASHSTATE]; + uint32_t x[HASHSTATE],y[HASHSTATE]; + uint32_t hlen; + + printf("No more than %d trials should ever be needed \n",MAXPAIR/2); + for (hlen=0; hlen < MAXLEN; ++hlen) + { + z=0; + for (i=0; i<hlen; ++i) /*----------------------- for each input byte, */ + { + for (j=0; j<8; ++j) /*------------------------ for each input bit, */ + { + for (m=1; m<8; ++m) /*------------ for serveral possible initvals, */ + { + for (l=0; l<HASHSTATE; ++l) + e[l]=f[l]=g[l]=h[l]=x[l]=y[l]=~((uint32_t)0); + + /*---- check that every output bit is affected by that input bit */ + for (k=0; k<MAXPAIR; k+=2) + { + uint32_t finished=1; + /* keys have one bit different */ + for (l=0; l<hlen+1; ++l) {a[l] = b[l] = (uint8_t)0;} + /* have a and b be two keys differing in only one bit */ + a[i] ^= (k<<j); + a[i] ^= (k>>(8-j)); + c[0] = hashlittle(a, hlen, m); + b[i] ^= ((k+1)<<j); + b[i] ^= ((k+1)>>(8-j)); + d[0] = hashlittle(b, hlen, m); + /* check every bit is 1, 0, set, and not set at least once */ + for (l=0; l<HASHSTATE; ++l) + { + e[l] &= (c[l]^d[l]); + f[l] &= ~(c[l]^d[l]); + g[l] &= c[l]; + h[l] &= ~c[l]; + x[l] &= d[l]; + y[l] &= ~d[l]; + if (e[l]|f[l]|g[l]|h[l]|x[l]|y[l]) finished=0; + } + if (finished) break; + } + if (k>z) z=k; + if (k==MAXPAIR) + { + printf("Some bit didn't change: "); + printf("%.8x %.8x %.8x %.8x %.8x %.8x ", + e[0],f[0],g[0],h[0],x[0],y[0]); + printf("i %d j %d m %d len %d\n", i, j, m, hlen); + } + if (z==MAXPAIR) goto done; + } + } + } + done: + if (z < MAXPAIR) + { + printf("Mix success %2d bytes %2d initvals ",i,m); + printf("required %d trials\n", z/2); + } + } + printf("\n"); +} + +/* Check for reading beyond the end of the buffer and alignment problems */ +void driver3() +{ + uint8_t buf[MAXLEN+20], *b; + uint32_t len; + uint8_t q[] = "This is the time for all good men to come to the aid of their country..."; + uint32_t h; + uint8_t qq[] = "xThis is the time for all good men to come to the aid of their country..."; + uint32_t i; + uint8_t qqq[] = "xxThis is the time for all good men to come to the aid of their country..."; + uint32_t j; + uint8_t qqqq[] = "xxxThis is the time for all good men to come to the aid of their country..."; + uint32_t ref,x,y; + uint8_t *p; + + printf("Endianness. These lines should all be the same (for values filled in):\n"); + printf("%.8x %.8x %.8x\n", + hashword((const uint32_t *)q, (sizeof(q)-1)/4, 13), + hashword((const uint32_t *)q, (sizeof(q)-5)/4, 13), + hashword((const uint32_t *)q, (sizeof(q)-9)/4, 13)); + p = q; + printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n", + hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13), + hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13), + hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13), + hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13), + hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13), + hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13)); + p = &qq[1]; + printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n", + hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13), + hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13), + hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13), + hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13), + hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13), + hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13)); + p = &qqq[2]; + printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n", + hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13), + hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13), + hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13), + hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13), + hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13), + hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13)); + p = &qqqq[3]; + printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n", + hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13), + hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13), + hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13), + hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13), + hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13), + hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13)); + printf("\n"); + + /* check that hashlittle2 and hashlittle produce the same results */ + i=47; j=0; + hashlittle2(q, sizeof(q), &i, &j); + if (hashlittle(q, sizeof(q), 47) != i) + printf("hashlittle2 and hashlittle mismatch\n"); + + /* check that hashword2 and hashword produce the same results */ + len = 0xdeadbeef; + i=47, j=0; + hashword2(&len, 1, &i, &j); + if (hashword(&len, 1, 47) != i) + printf("hashword2 and hashword mismatch %x %x\n", + i, hashword(&len, 1, 47)); + + /* check hashlittle doesn't read before or after the ends of the string */ + for (h=0, b=buf+1; h<8; ++h, ++b) + { + for (i=0; i<MAXLEN; ++i) + { + len = i; + for (j=0; j<i; ++j) *(b+j)=0; + + /* these should all be equal */ + ref = hashlittle(b, len, (uint32_t)1); + *(b+i)=(uint8_t)~0; + *(b-1)=(uint8_t)~0; + x = hashlittle(b, len, (uint32_t)1); + y = hashlittle(b, len, (uint32_t)1); + if ((ref != x) || (ref != y)) + { + printf("alignment error: %.8x %.8x %.8x %d %d\n",ref,x,y, + h, i); + } + } + } +} + +/* check for problems with nulls */ + void driver4() +{ + uint8_t buf[1]; + uint32_t h,i,state[HASHSTATE]; + + + buf[0] = ~0; + for (i=0; i<HASHSTATE; ++i) state[i] = 1; + printf("These should all be different\n"); + for (i=0, h=0; i<8; ++i) + { + h = hashlittle(buf, 0, h); + printf("%2ld 0-byte strings, hash is %.8x\n", i, h); + } +} + +void driver5() +{ + uint32_t b,c; + b=0, c=0, hashlittle2("", 0, &c, &b); + printf("hash is %.8lx %.8lx\n", c, b); /* deadbeef deadbeef */ + b=0xdeadbeef, c=0, hashlittle2("", 0, &c, &b); + printf("hash is %.8lx %.8lx\n", c, b); /* bd5b7dde deadbeef */ + b=0xdeadbeef, c=0xdeadbeef, hashlittle2("", 0, &c, &b); + printf("hash is %.8lx %.8lx\n", c, b); /* 9c093ccd bd5b7dde */ + b=0, c=0, hashlittle2("Four score and seven years ago", 30, &c, &b); + printf("hash is %.8lx %.8lx\n", c, b); /* 17770551 ce7226e6 */ + b=1, c=0, hashlittle2("Four score and seven years ago", 30, &c, &b); + printf("hash is %.8lx %.8lx\n", c, b); /* e3607cae bd371de4 */ + b=0, c=1, hashlittle2("Four score and seven years ago", 30, &c, &b); + printf("hash is %.8lx %.8lx\n", c, b); /* cd628161 6cbea4b3 */ + c = hashlittle("Four score and seven years ago", 30, 0); + printf("hash is %.8lx\n", c); /* 17770551 */ + c = hashlittle("Four score and seven years ago", 30, 1); + printf("hash is %.8lx\n", c); /* cd628161 */ +} + + +int main() +{ + driver1(); /* test that the key is hashed: used for timings */ + driver2(); /* test that whole key is hashed thoroughly */ + driver3(); /* test that nothing but the key is hashed */ + driver4(); /* test hashing multiple buffers (all buffers are null) */ + driver5(); /* test the hash against known vectors */ + return 1; +} + +#endif /* SELF_TEST */ diff --git a/src/libsystemd/src/sd-journal/lookup3.h b/src/libsystemd/src/sd-journal/lookup3.h new file mode 100644 index 0000000000..0c8c0b0ae4 --- /dev/null +++ b/src/libsystemd/src/sd-journal/lookup3.h @@ -0,0 +1,22 @@ +#pragma once + +#include <inttypes.h> +#include <sys/types.h> + +#include "systemd-basic/macro.h" + +uint32_t jenkins_hashword(const uint32_t *k, size_t length, uint32_t initval) _pure_; +void jenkins_hashword2(const uint32_t *k, size_t length, uint32_t *pc, uint32_t *pb); + +uint32_t jenkins_hashlittle(const void *key, size_t length, uint32_t initval) _pure_; +void jenkins_hashlittle2(const void *key, size_t length, uint32_t *pc, uint32_t *pb); + +uint32_t jenkins_hashbig(const void *key, size_t length, uint32_t initval) _pure_; + +static inline uint64_t hash64(const void *data, size_t length) { + uint32_t a = 0, b = 0; + + jenkins_hashlittle2(data, length, &a, &b); + + return ((uint64_t) a << 32ULL) | (uint64_t) b; +} diff --git a/src/libsystemd/src/sd-journal/mmap-cache.c b/src/libsystemd/src/sd-journal/mmap-cache.c new file mode 100644 index 0000000000..3570aa1d44 --- /dev/null +++ b/src/libsystemd/src/sd-journal/mmap-cache.c @@ -0,0 +1,724 @@ +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <stdlib.h> +#include <sys/mman.h> + +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/hashmap.h" +#include "systemd-basic/list.h" +#include "systemd-basic/log.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/sigbus.h" +#include "systemd-basic/util.h" + +#include "mmap-cache.h" + +typedef struct Window Window; +typedef struct Context Context; +typedef struct FileDescriptor FileDescriptor; + +struct Window { + MMapCache *cache; + + bool invalidated:1; + bool keep_always:1; + bool in_unused:1; + + int prot; + void *ptr; + uint64_t offset; + size_t size; + + FileDescriptor *fd; + + LIST_FIELDS(Window, by_fd); + LIST_FIELDS(Window, unused); + + LIST_HEAD(Context, contexts); +}; + +struct Context { + MMapCache *cache; + unsigned id; + Window *window; + + LIST_FIELDS(Context, by_window); +}; + +struct FileDescriptor { + MMapCache *cache; + int fd; + bool sigbus; + LIST_HEAD(Window, windows); +}; + +struct MMapCache { + int n_ref; + unsigned n_windows; + + unsigned n_hit, n_missed; + + Hashmap *fds; + Context *contexts[MMAP_CACHE_MAX_CONTEXTS]; + + LIST_HEAD(Window, unused); + Window *last_unused; +}; + +#define WINDOWS_MIN 64 + +#ifdef ENABLE_DEBUG_MMAP_CACHE +/* Tiny windows increase mmap activity and the chance of exposing unsafe use. */ +# define WINDOW_SIZE (page_size()) +#else +# define WINDOW_SIZE (8ULL*1024ULL*1024ULL) +#endif + +MMapCache* mmap_cache_new(void) { + MMapCache *m; + + m = new0(MMapCache, 1); + if (!m) + return NULL; + + m->n_ref = 1; + return m; +} + +MMapCache* mmap_cache_ref(MMapCache *m) { + assert(m); + assert(m->n_ref > 0); + + m->n_ref++; + return m; +} + +static void window_unlink(Window *w) { + Context *c; + + assert(w); + + if (w->ptr) + munmap(w->ptr, w->size); + + if (w->fd) + LIST_REMOVE(by_fd, w->fd->windows, w); + + if (w->in_unused) { + if (w->cache->last_unused == w) + w->cache->last_unused = w->unused_prev; + + LIST_REMOVE(unused, w->cache->unused, w); + } + + LIST_FOREACH(by_window, c, w->contexts) { + assert(c->window == w); + c->window = NULL; + } +} + +static void window_invalidate(Window *w) { + assert(w); + + if (w->invalidated) + return; + + /* Replace the window with anonymous pages. This is useful + * when we hit a SIGBUS and want to make sure the file cannot + * trigger any further SIGBUS, possibly overrunning the sigbus + * queue. */ + + assert_se(mmap(w->ptr, w->size, w->prot, MAP_PRIVATE|MAP_ANONYMOUS|MAP_FIXED, -1, 0) == w->ptr); + w->invalidated = true; +} + +static void window_free(Window *w) { + assert(w); + + window_unlink(w); + w->cache->n_windows--; + free(w); +} + +_pure_ static bool window_matches(Window *w, int fd, int prot, uint64_t offset, size_t size) { + assert(w); + assert(fd >= 0); + assert(size > 0); + + return + w->fd && + fd == w->fd->fd && + prot == w->prot && + offset >= w->offset && + offset + size <= w->offset + w->size; +} + +static Window *window_add(MMapCache *m, FileDescriptor *fd, int prot, bool keep_always, uint64_t offset, size_t size, void *ptr) { + Window *w; + + assert(m); + assert(fd); + + if (!m->last_unused || m->n_windows <= WINDOWS_MIN) { + + /* Allocate a new window */ + w = new0(Window, 1); + if (!w) + return NULL; + m->n_windows++; + } else { + + /* Reuse an existing one */ + w = m->last_unused; + window_unlink(w); + zero(*w); + } + + w->cache = m; + w->fd = fd; + w->prot = prot; + w->keep_always = keep_always; + w->offset = offset; + w->size = size; + w->ptr = ptr; + + LIST_PREPEND(by_fd, fd->windows, w); + + return w; +} + +static void context_detach_window(Context *c) { + Window *w; + + assert(c); + + if (!c->window) + return; + + w = c->window; + c->window = NULL; + LIST_REMOVE(by_window, w->contexts, c); + + if (!w->contexts && !w->keep_always) { + /* Not used anymore? */ +#ifdef ENABLE_DEBUG_MMAP_CACHE + /* Unmap unused windows immediately to expose use-after-unmap + * by SIGSEGV. */ + window_free(w); +#else + LIST_PREPEND(unused, c->cache->unused, w); + if (!c->cache->last_unused) + c->cache->last_unused = w; + + w->in_unused = true; +#endif + } +} + +static void context_attach_window(Context *c, Window *w) { + assert(c); + assert(w); + + if (c->window == w) + return; + + context_detach_window(c); + + if (w->in_unused) { + /* Used again? */ + LIST_REMOVE(unused, c->cache->unused, w); + if (c->cache->last_unused == w) + c->cache->last_unused = w->unused_prev; + + w->in_unused = false; + } + + c->window = w; + LIST_PREPEND(by_window, w->contexts, c); +} + +static Context *context_add(MMapCache *m, unsigned id) { + Context *c; + + assert(m); + + c = m->contexts[id]; + if (c) + return c; + + c = new0(Context, 1); + if (!c) + return NULL; + + c->cache = m; + c->id = id; + + assert(!m->contexts[id]); + m->contexts[id] = c; + + return c; +} + +static void context_free(Context *c) { + assert(c); + + context_detach_window(c); + + if (c->cache) { + assert(c->cache->contexts[c->id] == c); + c->cache->contexts[c->id] = NULL; + } + + free(c); +} + +static void fd_free(FileDescriptor *f) { + assert(f); + + while (f->windows) + window_free(f->windows); + + if (f->cache) + assert_se(hashmap_remove(f->cache->fds, FD_TO_PTR(f->fd))); + + free(f); +} + +static FileDescriptor* fd_add(MMapCache *m, int fd) { + FileDescriptor *f; + int r; + + assert(m); + assert(fd >= 0); + + f = hashmap_get(m->fds, FD_TO_PTR(fd)); + if (f) + return f; + + r = hashmap_ensure_allocated(&m->fds, NULL); + if (r < 0) + return NULL; + + f = new0(FileDescriptor, 1); + if (!f) + return NULL; + + f->cache = m; + f->fd = fd; + + r = hashmap_put(m->fds, FD_TO_PTR(fd), f); + if (r < 0) + return mfree(f); + + return f; +} + +static void mmap_cache_free(MMapCache *m) { + FileDescriptor *f; + int i; + + assert(m); + + for (i = 0; i < MMAP_CACHE_MAX_CONTEXTS; i++) + if (m->contexts[i]) + context_free(m->contexts[i]); + + while ((f = hashmap_first(m->fds))) + fd_free(f); + + hashmap_free(m->fds); + + while (m->unused) + window_free(m->unused); + + free(m); +} + +MMapCache* mmap_cache_unref(MMapCache *m) { + + if (!m) + return NULL; + + assert(m->n_ref > 0); + + m->n_ref--; + if (m->n_ref == 0) + mmap_cache_free(m); + + return NULL; +} + +static int make_room(MMapCache *m) { + assert(m); + + if (!m->last_unused) + return 0; + + window_free(m->last_unused); + return 1; +} + +static int try_context( + MMapCache *m, + int fd, + int prot, + unsigned context, + bool keep_always, + uint64_t offset, + size_t size, + void **ret) { + + Context *c; + + assert(m); + assert(m->n_ref > 0); + assert(fd >= 0); + assert(size > 0); + assert(ret); + + c = m->contexts[context]; + if (!c) + return 0; + + assert(c->id == context); + + if (!c->window) + return 0; + + if (!window_matches(c->window, fd, prot, offset, size)) { + + /* Drop the reference to the window, since it's unnecessary now */ + context_detach_window(c); + return 0; + } + + if (c->window->fd->sigbus) + return -EIO; + + c->window->keep_always = c->window->keep_always || keep_always; + + *ret = (uint8_t*) c->window->ptr + (offset - c->window->offset); + return 1; +} + +static int find_mmap( + MMapCache *m, + int fd, + int prot, + unsigned context, + bool keep_always, + uint64_t offset, + size_t size, + void **ret) { + + FileDescriptor *f; + Window *w; + Context *c; + + assert(m); + assert(m->n_ref > 0); + assert(fd >= 0); + assert(size > 0); + + f = hashmap_get(m->fds, FD_TO_PTR(fd)); + if (!f) + return 0; + + assert(f->fd == fd); + + if (f->sigbus) + return -EIO; + + LIST_FOREACH(by_fd, w, f->windows) + if (window_matches(w, fd, prot, offset, size)) + break; + + if (!w) + return 0; + + c = context_add(m, context); + if (!c) + return -ENOMEM; + + context_attach_window(c, w); + w->keep_always = w->keep_always || keep_always; + + *ret = (uint8_t*) w->ptr + (offset - w->offset); + return 1; +} + +static int mmap_try_harder(MMapCache *m, void *addr, int fd, int prot, int flags, uint64_t offset, size_t size, void **res) { + void *ptr; + + assert(m); + assert(fd >= 0); + assert(res); + + for (;;) { + int r; + + ptr = mmap(addr, size, prot, flags, fd, offset); + if (ptr != MAP_FAILED) + break; + if (errno != ENOMEM) + return negative_errno(); + + r = make_room(m); + if (r < 0) + return r; + if (r == 0) + return -ENOMEM; + } + + *res = ptr; + return 0; +} + +static int add_mmap( + MMapCache *m, + int fd, + int prot, + unsigned context, + bool keep_always, + uint64_t offset, + size_t size, + struct stat *st, + void **ret) { + + uint64_t woffset, wsize; + Context *c; + FileDescriptor *f; + Window *w; + void *d; + int r; + + assert(m); + assert(m->n_ref > 0); + assert(fd >= 0); + assert(size > 0); + assert(ret); + + woffset = offset & ~((uint64_t) page_size() - 1ULL); + wsize = size + (offset - woffset); + wsize = PAGE_ALIGN(wsize); + + if (wsize < WINDOW_SIZE) { + uint64_t delta; + + delta = PAGE_ALIGN((WINDOW_SIZE - wsize) / 2); + + if (delta > offset) + woffset = 0; + else + woffset -= delta; + + wsize = WINDOW_SIZE; + } + + if (st) { + /* Memory maps that are larger then the files + underneath have undefined behavior. Hence, clamp + things to the file size if we know it */ + + if (woffset >= (uint64_t) st->st_size) + return -EADDRNOTAVAIL; + + if (woffset + wsize > (uint64_t) st->st_size) + wsize = PAGE_ALIGN(st->st_size - woffset); + } + + r = mmap_try_harder(m, NULL, fd, prot, MAP_SHARED, woffset, wsize, &d); + if (r < 0) + return r; + + c = context_add(m, context); + if (!c) + goto outofmem; + + f = fd_add(m, fd); + if (!f) + goto outofmem; + + w = window_add(m, f, prot, keep_always, woffset, wsize, d); + if (!w) + goto outofmem; + + context_detach_window(c); + c->window = w; + LIST_PREPEND(by_window, w->contexts, c); + + *ret = (uint8_t*) w->ptr + (offset - w->offset); + return 1; + +outofmem: + (void) munmap(d, wsize); + return -ENOMEM; +} + +int mmap_cache_get( + MMapCache *m, + int fd, + int prot, + unsigned context, + bool keep_always, + uint64_t offset, + size_t size, + struct stat *st, + void **ret) { + + int r; + + assert(m); + assert(m->n_ref > 0); + assert(fd >= 0); + assert(size > 0); + assert(ret); + assert(context < MMAP_CACHE_MAX_CONTEXTS); + + /* Check whether the current context is the right one already */ + r = try_context(m, fd, prot, context, keep_always, offset, size, ret); + if (r != 0) { + m->n_hit++; + return r; + } + + /* Search for a matching mmap */ + r = find_mmap(m, fd, prot, context, keep_always, offset, size, ret); + if (r != 0) { + m->n_hit++; + return r; + } + + m->n_missed++; + + /* Create a new mmap */ + return add_mmap(m, fd, prot, context, keep_always, offset, size, st, ret); +} + +unsigned mmap_cache_get_hit(MMapCache *m) { + assert(m); + + return m->n_hit; +} + +unsigned mmap_cache_get_missed(MMapCache *m) { + assert(m); + + return m->n_missed; +} + +static void mmap_cache_process_sigbus(MMapCache *m) { + bool found = false; + FileDescriptor *f; + Iterator i; + int r; + + assert(m); + + /* Iterate through all triggered pages and mark their files as + * invalidated */ + for (;;) { + bool ours; + void *addr; + + r = sigbus_pop(&addr); + if (_likely_(r == 0)) + break; + if (r < 0) { + log_error_errno(r, "SIGBUS handling failed: %m"); + abort(); + } + + ours = false; + HASHMAP_FOREACH(f, m->fds, i) { + Window *w; + + LIST_FOREACH(by_fd, w, f->windows) { + if ((uint8_t*) addr >= (uint8_t*) w->ptr && + (uint8_t*) addr < (uint8_t*) w->ptr + w->size) { + found = ours = f->sigbus = true; + break; + } + } + + if (ours) + break; + } + + /* Didn't find a matching window, give up */ + if (!ours) { + log_error("Unknown SIGBUS page, aborting."); + abort(); + } + } + + /* The list of triggered pages is now empty. Now, let's remap + * all windows of the triggered file to anonymous maps, so + * that no page of the file in question is triggered again, so + * that we can be sure not to hit the queue size limit. */ + if (_likely_(!found)) + return; + + HASHMAP_FOREACH(f, m->fds, i) { + Window *w; + + if (!f->sigbus) + continue; + + LIST_FOREACH(by_fd, w, f->windows) + window_invalidate(w); + } +} + +bool mmap_cache_got_sigbus(MMapCache *m, int fd) { + FileDescriptor *f; + + assert(m); + assert(fd >= 0); + + mmap_cache_process_sigbus(m); + + f = hashmap_get(m->fds, FD_TO_PTR(fd)); + if (!f) + return false; + + return f->sigbus; +} + +void mmap_cache_close_fd(MMapCache *m, int fd) { + FileDescriptor *f; + + assert(m); + assert(fd >= 0); + + /* Make sure that any queued SIGBUS are first dispatched, so + * that we don't end up with a SIGBUS entry we cannot relate + * to any existing memory map */ + + mmap_cache_process_sigbus(m); + + f = hashmap_get(m->fds, FD_TO_PTR(fd)); + if (!f) + return; + + fd_free(f); +} diff --git a/src/libsystemd/src/sd-journal/mmap-cache.h b/src/libsystemd/src/sd-journal/mmap-cache.h new file mode 100644 index 0000000000..199d944647 --- /dev/null +++ b/src/libsystemd/src/sd-journal/mmap-cache.h @@ -0,0 +1,49 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> +#include <sys/stat.h> + +/* One context per object type, plus one of the header, plus one "additional" one */ +#define MMAP_CACHE_MAX_CONTEXTS 9 + +typedef struct MMapCache MMapCache; + +MMapCache* mmap_cache_new(void); +MMapCache* mmap_cache_ref(MMapCache *m); +MMapCache* mmap_cache_unref(MMapCache *m); + +int mmap_cache_get( + MMapCache *m, + int fd, + int prot, + unsigned context, + bool keep_always, + uint64_t offset, + size_t size, + struct stat *st, + void **ret); +void mmap_cache_close_fd(MMapCache *m, int fd); + +unsigned mmap_cache_get_hit(MMapCache *m); +unsigned mmap_cache_get_missed(MMapCache *m); + +bool mmap_cache_got_sigbus(MMapCache *m, int fd); diff --git a/src/libsystemd/src/sd-journal/sd-journal.c b/src/libsystemd/src/sd-journal/sd-journal.c new file mode 100644 index 0000000000..68ef58b79b --- /dev/null +++ b/src/libsystemd/src/sd-journal/sd-journal.c @@ -0,0 +1,3006 @@ +/*** + This file is part of systemd. + + Copyright 2011 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <fcntl.h> +#include <inttypes.h> +#include <poll.h> +#include <stddef.h> +#include <sys/inotify.h> +#include <sys/vfs.h> +#include <unistd.h> + +#include <linux/magic.h> + +#include <systemd/sd-journal.h> + +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/dirent-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/formats-util.h" +#include "systemd-basic/fs-util.h" +#include "systemd-basic/hashmap.h" +#include "systemd-basic/hostname-util.h" +#include "systemd-basic/io-util.h" +#include "systemd-basic/list.h" +#include "systemd-basic/missing.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/replace-var.h" +#include "systemd-basic/stat-util.h" +#include "systemd-basic/stdio-util.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" + +#include "catalog.h" +#include "compress.h" +#include "journal-def.h" +#include "journal-file.h" +#include "journal-internal.h" +#include "lookup3.h" + +#define JOURNAL_FILES_MAX 7168 + +#define JOURNAL_FILES_RECHECK_USEC (2 * USEC_PER_SEC) + +#define REPLACE_VAR_MAX 256 + +#define DEFAULT_DATA_THRESHOLD (64*1024) + +static void remove_file_real(sd_journal *j, JournalFile *f); + +static bool journal_pid_changed(sd_journal *j) { + assert(j); + + /* We don't support people creating a journal object and + * keeping it around over a fork(). Let's complain. */ + + return j->original_pid != getpid(); +} + +static int journal_put_error(sd_journal *j, int r, const char *path) { + char *copy; + int k; + + /* Memorize an error we encountered, and store which + * file/directory it was generated from. Note that we store + * only *one* path per error code, as the error code is the + * key into the hashmap, and the path is the value. This means + * we keep track only of all error kinds, but not of all error + * locations. This has the benefit that the hashmap cannot + * grow beyond bounds. + * + * We return an error here only if we didn't manage to + * memorize the real error. */ + + if (r >= 0) + return r; + + k = hashmap_ensure_allocated(&j->errors, NULL); + if (k < 0) + return k; + + if (path) { + copy = strdup(path); + if (!copy) + return -ENOMEM; + } else + copy = NULL; + + k = hashmap_put(j->errors, INT_TO_PTR(r), copy); + if (k < 0) { + free(copy); + + if (k == -EEXIST) + return 0; + + return k; + } + + return 0; +} + +static void detach_location(sd_journal *j) { + Iterator i; + JournalFile *f; + + assert(j); + + j->current_file = NULL; + j->current_field = 0; + + ORDERED_HASHMAP_FOREACH(f, j->files, i) + journal_file_reset_location(f); +} + +static void reset_location(sd_journal *j) { + assert(j); + + detach_location(j); + zero(j->current_location); +} + +static void init_location(Location *l, LocationType type, JournalFile *f, Object *o) { + assert(l); + assert(type == LOCATION_DISCRETE || type == LOCATION_SEEK); + assert(f); + assert(o->object.type == OBJECT_ENTRY); + + l->type = type; + l->seqnum = le64toh(o->entry.seqnum); + l->seqnum_id = f->header->seqnum_id; + l->realtime = le64toh(o->entry.realtime); + l->monotonic = le64toh(o->entry.monotonic); + l->boot_id = o->entry.boot_id; + l->xor_hash = le64toh(o->entry.xor_hash); + + l->seqnum_set = l->realtime_set = l->monotonic_set = l->xor_hash_set = true; +} + +static void set_location(sd_journal *j, JournalFile *f, Object *o) { + assert(j); + assert(f); + assert(o); + + init_location(&j->current_location, LOCATION_DISCRETE, f, o); + + j->current_file = f; + j->current_field = 0; + + /* Let f know its candidate entry was picked. */ + assert(f->location_type == LOCATION_SEEK); + f->location_type = LOCATION_DISCRETE; +} + +static int match_is_valid(const void *data, size_t size) { + const char *b, *p; + + assert(data); + + if (size < 2) + return false; + + if (startswith(data, "__")) + return false; + + b = data; + for (p = b; p < b + size; p++) { + + if (*p == '=') + return p > b; + + if (*p == '_') + continue; + + if (*p >= 'A' && *p <= 'Z') + continue; + + if (*p >= '0' && *p <= '9') + continue; + + return false; + } + + return false; +} + +static bool same_field(const void *_a, size_t s, const void *_b, size_t t) { + const uint8_t *a = _a, *b = _b; + size_t j; + + for (j = 0; j < s && j < t; j++) { + + if (a[j] != b[j]) + return false; + + if (a[j] == '=') + return true; + } + + assert_not_reached("\"=\" not found"); +} + +static Match *match_new(Match *p, MatchType t) { + Match *m; + + m = new0(Match, 1); + if (!m) + return NULL; + + m->type = t; + + if (p) { + m->parent = p; + LIST_PREPEND(matches, p->matches, m); + } + + return m; +} + +static void match_free(Match *m) { + assert(m); + + while (m->matches) + match_free(m->matches); + + if (m->parent) + LIST_REMOVE(matches, m->parent->matches, m); + + free(m->data); + free(m); +} + +static void match_free_if_empty(Match *m) { + if (!m || m->matches) + return; + + match_free(m); +} + +_public_ int sd_journal_add_match(sd_journal *j, const void *data, size_t size) { + Match *l3, *l4, *add_here = NULL, *m; + le64_t le_hash; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(data, -EINVAL); + + if (size == 0) + size = strlen(data); + + assert_return(match_is_valid(data, size), -EINVAL); + + /* level 0: AND term + * level 1: OR terms + * level 2: AND terms + * level 3: OR terms + * level 4: concrete matches */ + + if (!j->level0) { + j->level0 = match_new(NULL, MATCH_AND_TERM); + if (!j->level0) + return -ENOMEM; + } + + if (!j->level1) { + j->level1 = match_new(j->level0, MATCH_OR_TERM); + if (!j->level1) + return -ENOMEM; + } + + if (!j->level2) { + j->level2 = match_new(j->level1, MATCH_AND_TERM); + if (!j->level2) + return -ENOMEM; + } + + assert(j->level0->type == MATCH_AND_TERM); + assert(j->level1->type == MATCH_OR_TERM); + assert(j->level2->type == MATCH_AND_TERM); + + le_hash = htole64(hash64(data, size)); + + LIST_FOREACH(matches, l3, j->level2->matches) { + assert(l3->type == MATCH_OR_TERM); + + LIST_FOREACH(matches, l4, l3->matches) { + assert(l4->type == MATCH_DISCRETE); + + /* Exactly the same match already? Then ignore + * this addition */ + if (l4->le_hash == le_hash && + l4->size == size && + memcmp(l4->data, data, size) == 0) + return 0; + + /* Same field? Then let's add this to this OR term */ + if (same_field(data, size, l4->data, l4->size)) { + add_here = l3; + break; + } + } + + if (add_here) + break; + } + + if (!add_here) { + add_here = match_new(j->level2, MATCH_OR_TERM); + if (!add_here) + goto fail; + } + + m = match_new(add_here, MATCH_DISCRETE); + if (!m) + goto fail; + + m->le_hash = le_hash; + m->size = size; + m->data = memdup(data, size); + if (!m->data) + goto fail; + + detach_location(j); + + return 0; + +fail: + match_free_if_empty(add_here); + match_free_if_empty(j->level2); + match_free_if_empty(j->level1); + match_free_if_empty(j->level0); + + return -ENOMEM; +} + +_public_ int sd_journal_add_conjunction(sd_journal *j) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + if (!j->level0) + return 0; + + if (!j->level1) + return 0; + + if (!j->level1->matches) + return 0; + + j->level1 = NULL; + j->level2 = NULL; + + return 0; +} + +_public_ int sd_journal_add_disjunction(sd_journal *j) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + if (!j->level0) + return 0; + + if (!j->level1) + return 0; + + if (!j->level2) + return 0; + + if (!j->level2->matches) + return 0; + + j->level2 = NULL; + return 0; +} + +static char *match_make_string(Match *m) { + char *p = NULL, *r; + Match *i; + bool enclose = false; + + if (!m) + return strdup("none"); + + if (m->type == MATCH_DISCRETE) + return strndup(m->data, m->size); + + LIST_FOREACH(matches, i, m->matches) { + char *t, *k; + + t = match_make_string(i); + if (!t) + return mfree(p); + + if (p) { + k = strjoin(p, m->type == MATCH_OR_TERM ? " OR " : " AND ", t, NULL); + free(p); + free(t); + + if (!k) + return NULL; + + p = k; + + enclose = true; + } else + p = t; + } + + if (enclose) { + r = strjoin("(", p, ")", NULL); + free(p); + return r; + } + + return p; +} + +char *journal_make_match_string(sd_journal *j) { + assert(j); + + return match_make_string(j->level0); +} + +_public_ void sd_journal_flush_matches(sd_journal *j) { + if (!j) + return; + + if (j->level0) + match_free(j->level0); + + j->level0 = j->level1 = j->level2 = NULL; + + detach_location(j); +} + +_pure_ static int compare_with_location(JournalFile *f, Location *l) { + assert(f); + assert(l); + assert(f->location_type == LOCATION_SEEK); + assert(l->type == LOCATION_DISCRETE || l->type == LOCATION_SEEK); + + if (l->monotonic_set && + sd_id128_equal(f->current_boot_id, l->boot_id) && + l->realtime_set && + f->current_realtime == l->realtime && + l->xor_hash_set && + f->current_xor_hash == l->xor_hash) + return 0; + + if (l->seqnum_set && + sd_id128_equal(f->header->seqnum_id, l->seqnum_id)) { + + if (f->current_seqnum < l->seqnum) + return -1; + if (f->current_seqnum > l->seqnum) + return 1; + } + + if (l->monotonic_set && + sd_id128_equal(f->current_boot_id, l->boot_id)) { + + if (f->current_monotonic < l->monotonic) + return -1; + if (f->current_monotonic > l->monotonic) + return 1; + } + + if (l->realtime_set) { + + if (f->current_realtime < l->realtime) + return -1; + if (f->current_realtime > l->realtime) + return 1; + } + + if (l->xor_hash_set) { + + if (f->current_xor_hash < l->xor_hash) + return -1; + if (f->current_xor_hash > l->xor_hash) + return 1; + } + + return 0; +} + +static int next_for_match( + sd_journal *j, + Match *m, + JournalFile *f, + uint64_t after_offset, + direction_t direction, + Object **ret, + uint64_t *offset) { + + int r; + uint64_t np = 0; + Object *n; + + assert(j); + assert(m); + assert(f); + + if (m->type == MATCH_DISCRETE) { + uint64_t dp; + + r = journal_file_find_data_object_with_hash(f, m->data, m->size, le64toh(m->le_hash), NULL, &dp); + if (r <= 0) + return r; + + return journal_file_move_to_entry_by_offset_for_data(f, dp, after_offset, direction, ret, offset); + + } else if (m->type == MATCH_OR_TERM) { + Match *i; + + /* Find the earliest match beyond after_offset */ + + LIST_FOREACH(matches, i, m->matches) { + uint64_t cp; + + r = next_for_match(j, i, f, after_offset, direction, NULL, &cp); + if (r < 0) + return r; + else if (r > 0) { + if (np == 0 || (direction == DIRECTION_DOWN ? cp < np : cp > np)) + np = cp; + } + } + + if (np == 0) + return 0; + + } else if (m->type == MATCH_AND_TERM) { + Match *i, *last_moved; + + /* Always jump to the next matching entry and repeat + * this until we find an offset that matches for all + * matches. */ + + if (!m->matches) + return 0; + + r = next_for_match(j, m->matches, f, after_offset, direction, NULL, &np); + if (r <= 0) + return r; + + assert(direction == DIRECTION_DOWN ? np >= after_offset : np <= after_offset); + last_moved = m->matches; + + LIST_LOOP_BUT_ONE(matches, i, m->matches, last_moved) { + uint64_t cp; + + r = next_for_match(j, i, f, np, direction, NULL, &cp); + if (r <= 0) + return r; + + assert(direction == DIRECTION_DOWN ? cp >= np : cp <= np); + if (direction == DIRECTION_DOWN ? cp > np : cp < np) { + np = cp; + last_moved = i; + } + } + } + + assert(np > 0); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, np, &n); + if (r < 0) + return r; + + if (ret) + *ret = n; + if (offset) + *offset = np; + + return 1; +} + +static int find_location_for_match( + sd_journal *j, + Match *m, + JournalFile *f, + direction_t direction, + Object **ret, + uint64_t *offset) { + + int r; + + assert(j); + assert(m); + assert(f); + + if (m->type == MATCH_DISCRETE) { + uint64_t dp; + + r = journal_file_find_data_object_with_hash(f, m->data, m->size, le64toh(m->le_hash), NULL, &dp); + if (r <= 0) + return r; + + /* FIXME: missing: find by monotonic */ + + if (j->current_location.type == LOCATION_HEAD) + return journal_file_next_entry_for_data(f, NULL, 0, dp, DIRECTION_DOWN, ret, offset); + if (j->current_location.type == LOCATION_TAIL) + return journal_file_next_entry_for_data(f, NULL, 0, dp, DIRECTION_UP, ret, offset); + if (j->current_location.seqnum_set && sd_id128_equal(j->current_location.seqnum_id, f->header->seqnum_id)) + return journal_file_move_to_entry_by_seqnum_for_data(f, dp, j->current_location.seqnum, direction, ret, offset); + if (j->current_location.monotonic_set) { + r = journal_file_move_to_entry_by_monotonic_for_data(f, dp, j->current_location.boot_id, j->current_location.monotonic, direction, ret, offset); + if (r != -ENOENT) + return r; + } + if (j->current_location.realtime_set) + return journal_file_move_to_entry_by_realtime_for_data(f, dp, j->current_location.realtime, direction, ret, offset); + + return journal_file_next_entry_for_data(f, NULL, 0, dp, direction, ret, offset); + + } else if (m->type == MATCH_OR_TERM) { + uint64_t np = 0; + Object *n; + Match *i; + + /* Find the earliest match */ + + LIST_FOREACH(matches, i, m->matches) { + uint64_t cp; + + r = find_location_for_match(j, i, f, direction, NULL, &cp); + if (r < 0) + return r; + else if (r > 0) { + if (np == 0 || (direction == DIRECTION_DOWN ? np > cp : np < cp)) + np = cp; + } + } + + if (np == 0) + return 0; + + r = journal_file_move_to_object(f, OBJECT_ENTRY, np, &n); + if (r < 0) + return r; + + if (ret) + *ret = n; + if (offset) + *offset = np; + + return 1; + + } else { + Match *i; + uint64_t np = 0; + + assert(m->type == MATCH_AND_TERM); + + /* First jump to the last match, and then find the + * next one where all matches match */ + + if (!m->matches) + return 0; + + LIST_FOREACH(matches, i, m->matches) { + uint64_t cp; + + r = find_location_for_match(j, i, f, direction, NULL, &cp); + if (r <= 0) + return r; + + if (np == 0 || (direction == DIRECTION_DOWN ? cp > np : cp < np)) + np = cp; + } + + return next_for_match(j, m, f, np, direction, ret, offset); + } +} + +static int find_location_with_matches( + sd_journal *j, + JournalFile *f, + direction_t direction, + Object **ret, + uint64_t *offset) { + + int r; + + assert(j); + assert(f); + assert(ret); + assert(offset); + + if (!j->level0) { + /* No matches is simple */ + + if (j->current_location.type == LOCATION_HEAD) + return journal_file_next_entry(f, 0, DIRECTION_DOWN, ret, offset); + if (j->current_location.type == LOCATION_TAIL) + return journal_file_next_entry(f, 0, DIRECTION_UP, ret, offset); + if (j->current_location.seqnum_set && sd_id128_equal(j->current_location.seqnum_id, f->header->seqnum_id)) + return journal_file_move_to_entry_by_seqnum(f, j->current_location.seqnum, direction, ret, offset); + if (j->current_location.monotonic_set) { + r = journal_file_move_to_entry_by_monotonic(f, j->current_location.boot_id, j->current_location.monotonic, direction, ret, offset); + if (r != -ENOENT) + return r; + } + if (j->current_location.realtime_set) + return journal_file_move_to_entry_by_realtime(f, j->current_location.realtime, direction, ret, offset); + + return journal_file_next_entry(f, 0, direction, ret, offset); + } else + return find_location_for_match(j, j->level0, f, direction, ret, offset); +} + +static int next_with_matches( + sd_journal *j, + JournalFile *f, + direction_t direction, + Object **ret, + uint64_t *offset) { + + assert(j); + assert(f); + assert(ret); + assert(offset); + + /* No matches is easy. We simple advance the file + * pointer by one. */ + if (!j->level0) + return journal_file_next_entry(f, f->current_offset, direction, ret, offset); + + /* If we have a match then we look for the next matching entry + * with an offset at least one step larger */ + return next_for_match(j, j->level0, f, + direction == DIRECTION_DOWN ? f->current_offset + 1 + : f->current_offset - 1, + direction, ret, offset); +} + +static int next_beyond_location(sd_journal *j, JournalFile *f, direction_t direction) { + Object *c; + uint64_t cp, n_entries; + int r; + + assert(j); + assert(f); + + n_entries = le64toh(f->header->n_entries); + + /* If we hit EOF before, we don't need to look into this file again + * unless direction changed or new entries appeared. */ + if (f->last_direction == direction && f->location_type == LOCATION_TAIL && + n_entries == f->last_n_entries) + return 0; + + f->last_n_entries = n_entries; + + if (f->last_direction == direction && f->current_offset > 0) { + /* LOCATION_SEEK here means we did the work in a previous + * iteration and the current location already points to a + * candidate entry. */ + if (f->location_type != LOCATION_SEEK) { + r = next_with_matches(j, f, direction, &c, &cp); + if (r <= 0) + return r; + + journal_file_save_location(f, c, cp); + } + } else { + f->last_direction = direction; + + r = find_location_with_matches(j, f, direction, &c, &cp); + if (r <= 0) + return r; + + journal_file_save_location(f, c, cp); + } + + /* OK, we found the spot, now let's advance until an entry + * that is actually different from what we were previously + * looking at. This is necessary to handle entries which exist + * in two (or more) journal files, and which shall all be + * suppressed but one. */ + + for (;;) { + bool found; + + if (j->current_location.type == LOCATION_DISCRETE) { + int k; + + k = compare_with_location(f, &j->current_location); + + found = direction == DIRECTION_DOWN ? k > 0 : k < 0; + } else + found = true; + + if (found) + return 1; + + r = next_with_matches(j, f, direction, &c, &cp); + if (r <= 0) + return r; + + journal_file_save_location(f, c, cp); + } +} + +static int real_journal_next(sd_journal *j, direction_t direction) { + JournalFile *f, *new_file = NULL; + Iterator i; + Object *o; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + ORDERED_HASHMAP_FOREACH(f, j->files, i) { + bool found; + + r = next_beyond_location(j, f, direction); + if (r < 0) { + log_debug_errno(r, "Can't iterate through %s, ignoring: %m", f->path); + remove_file_real(j, f); + continue; + } else if (r == 0) { + f->location_type = LOCATION_TAIL; + continue; + } + + if (!new_file) + found = true; + else { + int k; + + k = journal_file_compare_locations(f, new_file); + + found = direction == DIRECTION_DOWN ? k < 0 : k > 0; + } + + if (found) + new_file = f; + } + + if (!new_file) + return 0; + + r = journal_file_move_to_object(new_file, OBJECT_ENTRY, new_file->current_offset, &o); + if (r < 0) + return r; + + set_location(j, new_file, o); + + return 1; +} + +_public_ int sd_journal_next(sd_journal *j) { + return real_journal_next(j, DIRECTION_DOWN); +} + +_public_ int sd_journal_previous(sd_journal *j) { + return real_journal_next(j, DIRECTION_UP); +} + +static int real_journal_next_skip(sd_journal *j, direction_t direction, uint64_t skip) { + int c = 0, r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + if (skip == 0) { + /* If this is not a discrete skip, then at least + * resolve the current location */ + if (j->current_location.type != LOCATION_DISCRETE) + return real_journal_next(j, direction); + + return 0; + } + + do { + r = real_journal_next(j, direction); + if (r < 0) + return r; + + if (r == 0) + return c; + + skip--; + c++; + } while (skip > 0); + + return c; +} + +_public_ int sd_journal_next_skip(sd_journal *j, uint64_t skip) { + return real_journal_next_skip(j, DIRECTION_DOWN, skip); +} + +_public_ int sd_journal_previous_skip(sd_journal *j, uint64_t skip) { + return real_journal_next_skip(j, DIRECTION_UP, skip); +} + +_public_ int sd_journal_get_cursor(sd_journal *j, char **cursor) { + Object *o; + int r; + char bid[33], sid[33]; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(cursor, -EINVAL); + + if (!j->current_file || j->current_file->current_offset <= 0) + return -EADDRNOTAVAIL; + + r = journal_file_move_to_object(j->current_file, OBJECT_ENTRY, j->current_file->current_offset, &o); + if (r < 0) + return r; + + sd_id128_to_string(j->current_file->header->seqnum_id, sid); + sd_id128_to_string(o->entry.boot_id, bid); + + if (asprintf(cursor, + "s=%s;i=%"PRIx64";b=%s;m=%"PRIx64";t=%"PRIx64";x=%"PRIx64, + sid, le64toh(o->entry.seqnum), + bid, le64toh(o->entry.monotonic), + le64toh(o->entry.realtime), + le64toh(o->entry.xor_hash)) < 0) + return -ENOMEM; + + return 0; +} + +_public_ int sd_journal_seek_cursor(sd_journal *j, const char *cursor) { + const char *word, *state; + size_t l; + unsigned long long seqnum, monotonic, realtime, xor_hash; + bool + seqnum_id_set = false, + seqnum_set = false, + boot_id_set = false, + monotonic_set = false, + realtime_set = false, + xor_hash_set = false; + sd_id128_t seqnum_id, boot_id; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(!isempty(cursor), -EINVAL); + + FOREACH_WORD_SEPARATOR(word, l, cursor, ";", state) { + char *item; + int k = 0; + + if (l < 2 || word[1] != '=') + return -EINVAL; + + item = strndup(word, l); + if (!item) + return -ENOMEM; + + switch (word[0]) { + + case 's': + seqnum_id_set = true; + k = sd_id128_from_string(item+2, &seqnum_id); + break; + + case 'i': + seqnum_set = true; + if (sscanf(item+2, "%llx", &seqnum) != 1) + k = -EINVAL; + break; + + case 'b': + boot_id_set = true; + k = sd_id128_from_string(item+2, &boot_id); + break; + + case 'm': + monotonic_set = true; + if (sscanf(item+2, "%llx", &monotonic) != 1) + k = -EINVAL; + break; + + case 't': + realtime_set = true; + if (sscanf(item+2, "%llx", &realtime) != 1) + k = -EINVAL; + break; + + case 'x': + xor_hash_set = true; + if (sscanf(item+2, "%llx", &xor_hash) != 1) + k = -EINVAL; + break; + } + + free(item); + + if (k < 0) + return k; + } + + if ((!seqnum_set || !seqnum_id_set) && + (!monotonic_set || !boot_id_set) && + !realtime_set) + return -EINVAL; + + reset_location(j); + + j->current_location.type = LOCATION_SEEK; + + if (realtime_set) { + j->current_location.realtime = (uint64_t) realtime; + j->current_location.realtime_set = true; + } + + if (seqnum_set && seqnum_id_set) { + j->current_location.seqnum = (uint64_t) seqnum; + j->current_location.seqnum_id = seqnum_id; + j->current_location.seqnum_set = true; + } + + if (monotonic_set && boot_id_set) { + j->current_location.monotonic = (uint64_t) monotonic; + j->current_location.boot_id = boot_id; + j->current_location.monotonic_set = true; + } + + if (xor_hash_set) { + j->current_location.xor_hash = (uint64_t) xor_hash; + j->current_location.xor_hash_set = true; + } + + return 0; +} + +_public_ int sd_journal_test_cursor(sd_journal *j, const char *cursor) { + int r; + Object *o; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(!isempty(cursor), -EINVAL); + + if (!j->current_file || j->current_file->current_offset <= 0) + return -EADDRNOTAVAIL; + + r = journal_file_move_to_object(j->current_file, OBJECT_ENTRY, j->current_file->current_offset, &o); + if (r < 0) + return r; + + for (;;) { + _cleanup_free_ char *item = NULL; + unsigned long long ll; + sd_id128_t id; + int k = 0; + + r = extract_first_word(&cursor, &item, ";", EXTRACT_DONT_COALESCE_SEPARATORS); + if (r < 0) + return r; + + if (r == 0) + break; + + if (strlen(item) < 2 || item[1] != '=') + return -EINVAL; + + switch (item[0]) { + + case 's': + k = sd_id128_from_string(item+2, &id); + if (k < 0) + return k; + if (!sd_id128_equal(id, j->current_file->header->seqnum_id)) + return 0; + break; + + case 'i': + if (sscanf(item+2, "%llx", &ll) != 1) + return -EINVAL; + if (ll != le64toh(o->entry.seqnum)) + return 0; + break; + + case 'b': + k = sd_id128_from_string(item+2, &id); + if (k < 0) + return k; + if (!sd_id128_equal(id, o->entry.boot_id)) + return 0; + break; + + case 'm': + if (sscanf(item+2, "%llx", &ll) != 1) + return -EINVAL; + if (ll != le64toh(o->entry.monotonic)) + return 0; + break; + + case 't': + if (sscanf(item+2, "%llx", &ll) != 1) + return -EINVAL; + if (ll != le64toh(o->entry.realtime)) + return 0; + break; + + case 'x': + if (sscanf(item+2, "%llx", &ll) != 1) + return -EINVAL; + if (ll != le64toh(o->entry.xor_hash)) + return 0; + break; + } + } + + return 1; +} + + +_public_ int sd_journal_seek_monotonic_usec(sd_journal *j, sd_id128_t boot_id, uint64_t usec) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + reset_location(j); + j->current_location.type = LOCATION_SEEK; + j->current_location.boot_id = boot_id; + j->current_location.monotonic = usec; + j->current_location.monotonic_set = true; + + return 0; +} + +_public_ int sd_journal_seek_realtime_usec(sd_journal *j, uint64_t usec) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + reset_location(j); + j->current_location.type = LOCATION_SEEK; + j->current_location.realtime = usec; + j->current_location.realtime_set = true; + + return 0; +} + +_public_ int sd_journal_seek_head(sd_journal *j) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + reset_location(j); + j->current_location.type = LOCATION_HEAD; + + return 0; +} + +_public_ int sd_journal_seek_tail(sd_journal *j) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + reset_location(j); + j->current_location.type = LOCATION_TAIL; + + return 0; +} + +static void check_network(sd_journal *j, int fd) { + struct statfs sfs; + + assert(j); + + if (j->on_network) + return; + + if (fstatfs(fd, &sfs) < 0) + return; + + j->on_network = + F_TYPE_EQUAL(sfs.f_type, CIFS_MAGIC_NUMBER) || + F_TYPE_EQUAL(sfs.f_type, CODA_SUPER_MAGIC) || + F_TYPE_EQUAL(sfs.f_type, NCP_SUPER_MAGIC) || + F_TYPE_EQUAL(sfs.f_type, NFS_SUPER_MAGIC) || + F_TYPE_EQUAL(sfs.f_type, SMB_SUPER_MAGIC); +} + +static bool file_has_type_prefix(const char *prefix, const char *filename) { + const char *full, *tilded, *atted; + + full = strjoina(prefix, ".journal"); + tilded = strjoina(full, "~"); + atted = strjoina(prefix, "@"); + + return streq(filename, full) || + streq(filename, tilded) || + startswith(filename, atted); +} + +static bool file_type_wanted(int flags, const char *filename) { + assert(filename); + + if (!endswith(filename, ".journal") && !endswith(filename, ".journal~")) + return false; + + /* no flags set → every type is OK */ + if (!(flags & (SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER))) + return true; + + if (flags & SD_JOURNAL_SYSTEM && file_has_type_prefix("system", filename)) + return true; + + if (flags & SD_JOURNAL_CURRENT_USER) { + char prefix[5 + DECIMAL_STR_MAX(uid_t) + 1]; + + xsprintf(prefix, "user-"UID_FMT, getuid()); + + if (file_has_type_prefix(prefix, filename)) + return true; + } + + return false; +} + +static bool path_has_prefix(sd_journal *j, const char *path, const char *prefix) { + assert(j); + assert(path); + assert(prefix); + + if (j->toplevel_fd >= 0) + return false; + + return path_startswith(path, prefix); +} + +static const char *skip_slash(const char *p) { + + if (!p) + return NULL; + + while (*p == '/') + p++; + + return p; +} + +static int add_any_file(sd_journal *j, int fd, const char *path) { + JournalFile *f = NULL; + bool close_fd = false; + int r, k; + + assert(j); + assert(fd >= 0 || path); + + if (path && ordered_hashmap_get(j->files, path)) + return 0; + + if (ordered_hashmap_size(j->files) >= JOURNAL_FILES_MAX) { + log_debug("Too many open journal files, not adding %s.", path); + r = -ETOOMANYREFS; + goto fail; + } + + if (fd < 0 && j->toplevel_fd >= 0) { + + /* If there's a top-level fd defined, open the file relative to this now. (Make the path relative, + * explicitly, since otherwise openat() ignores the first argument.) */ + + fd = openat(j->toplevel_fd, skip_slash(path), O_RDONLY|O_CLOEXEC); + if (fd < 0) { + r = log_debug_errno(errno, "Failed to open journal file %s: %m", path); + goto fail; + } + + close_fd = true; + } + + r = journal_file_open(fd, path, O_RDONLY, 0, false, false, NULL, j->mmap, NULL, NULL, &f); + if (r < 0) { + if (close_fd) + safe_close(fd); + log_debug_errno(r, "Failed to open journal file %s: %m", path); + goto fail; + } + + /* journal_file_dump(f); */ + + r = ordered_hashmap_put(j->files, f->path, f); + if (r < 0) { + f->close_fd = close_fd; + (void) journal_file_close(f); + goto fail; + } + + if (!j->has_runtime_files && path_has_prefix(j, f->path, "/run")) + j->has_runtime_files = true; + else if (!j->has_persistent_files && path_has_prefix(j, f->path, "/var")) + j->has_persistent_files = true; + + log_debug("File %s added.", f->path); + + check_network(j, f->fd); + + j->current_invalidate_counter++; + + return 0; + +fail: + k = journal_put_error(j, r, path); + if (k < 0) + return k; + + return r; +} + +static int add_file(sd_journal *j, const char *prefix, const char *filename) { + const char *path; + + assert(j); + assert(prefix); + assert(filename); + + if (j->no_new_files) + return 0; + + if (!file_type_wanted(j->flags, filename)) + return 0; + + path = strjoina(prefix, "/", filename); + return add_any_file(j, -1, path); +} + +static void remove_file(sd_journal *j, const char *prefix, const char *filename) { + const char *path; + JournalFile *f; + + assert(j); + assert(prefix); + assert(filename); + + path = strjoina(prefix, "/", filename); + f = ordered_hashmap_get(j->files, path); + if (!f) + return; + + remove_file_real(j, f); +} + +static void remove_file_real(sd_journal *j, JournalFile *f) { + assert(j); + assert(f); + + ordered_hashmap_remove(j->files, f->path); + + log_debug("File %s removed.", f->path); + + if (j->current_file == f) { + j->current_file = NULL; + j->current_field = 0; + } + + if (j->unique_file == f) { + /* Jump to the next unique_file or NULL if that one was last */ + j->unique_file = ordered_hashmap_next(j->files, j->unique_file->path); + j->unique_offset = 0; + if (!j->unique_file) + j->unique_file_lost = true; + } + + if (j->fields_file == f) { + j->fields_file = ordered_hashmap_next(j->files, j->fields_file->path); + j->fields_offset = 0; + if (!j->fields_file) + j->fields_file_lost = true; + } + + (void) journal_file_close(f); + + j->current_invalidate_counter++; +} + +static int dirname_is_machine_id(const char *fn) { + sd_id128_t id, machine; + int r; + + r = sd_id128_get_machine(&machine); + if (r < 0) + return r; + + r = sd_id128_from_string(fn, &id); + if (r < 0) + return r; + + return sd_id128_equal(id, machine); +} + +static int add_directory(sd_journal *j, const char *prefix, const char *dirname) { + _cleanup_free_ char *path = NULL; + _cleanup_closedir_ DIR *d = NULL; + struct dirent *de = NULL; + Directory *m; + int r, k; + + assert(j); + assert(prefix); + + /* Adds a journal file directory to watch. If the directory is already tracked this updates the inotify watch + * and reenumerates directory contents */ + + if (dirname) + path = strjoin(prefix, "/", dirname, NULL); + else + path = strdup(prefix); + if (!path) { + r = -ENOMEM; + goto fail; + } + + log_debug("Considering directory %s.", path); + + /* We consider everything local that is in a directory for the local machine ID, or that is stored in /run */ + if ((j->flags & SD_JOURNAL_LOCAL_ONLY) && + !((dirname && dirname_is_machine_id(dirname) > 0) || path_has_prefix(j, path, "/run"))) + return 0; + + + if (j->toplevel_fd < 0) + d = opendir(path); + else + /* Open the specified directory relative to the toplevel fd. Enforce that the path specified is + * relative, by dropping the initial slash */ + d = xopendirat(j->toplevel_fd, skip_slash(path), 0); + if (!d) { + r = log_debug_errno(errno, "Failed to open directory %s: %m", path); + goto fail; + } + + m = hashmap_get(j->directories_by_path, path); + if (!m) { + m = new0(Directory, 1); + if (!m) { + r = -ENOMEM; + goto fail; + } + + m->is_root = false; + m->path = path; + + if (hashmap_put(j->directories_by_path, m->path, m) < 0) { + free(m); + r = -ENOMEM; + goto fail; + } + + path = NULL; /* avoid freeing in cleanup */ + j->current_invalidate_counter++; + + log_debug("Directory %s added.", m->path); + + } else if (m->is_root) + return 0; + + if (m->wd <= 0 && j->inotify_fd >= 0) { + /* Watch this directory, if it not being watched yet. */ + + m->wd = inotify_add_watch_fd(j->inotify_fd, dirfd(d), + IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB|IN_DELETE| + IN_DELETE_SELF|IN_MOVE_SELF|IN_UNMOUNT|IN_MOVED_FROM| + IN_ONLYDIR); + + if (m->wd > 0 && hashmap_put(j->directories_by_wd, INT_TO_PTR(m->wd), m) < 0) + inotify_rm_watch(j->inotify_fd, m->wd); + } + + FOREACH_DIRENT_ALL(de, d, r = log_debug_errno(errno, "Failed to read directory %s: %m", m->path); goto fail) { + + if (dirent_is_file_with_suffix(de, ".journal") || + dirent_is_file_with_suffix(de, ".journal~")) + (void) add_file(j, m->path, de->d_name); + } + + check_network(j, dirfd(d)); + + return 0; + +fail: + k = journal_put_error(j, r, path ?: prefix); + if (k < 0) + return k; + + return r; +} + +static int add_root_directory(sd_journal *j, const char *p, bool missing_ok) { + + _cleanup_closedir_ DIR *d = NULL; + struct dirent *de; + Directory *m; + int r, k; + + assert(j); + + /* Adds a root directory to our set of directories to use. If the root directory is already in the set, we + * update the inotify logic, and renumerate the directory entries. This call may hence be called to initially + * populate the set, as well as to update it later. */ + + if (p) { + /* If there's a path specified, use it. */ + + if ((j->flags & SD_JOURNAL_RUNTIME_ONLY) && + !path_has_prefix(j, p, "/run")) + return -EINVAL; + + if (j->prefix) + p = strjoina(j->prefix, p); + + if (j->toplevel_fd < 0) + d = opendir(p); + else + d = xopendirat(j->toplevel_fd, skip_slash(p), 0); + + if (!d) { + if (errno == ENOENT && missing_ok) + return 0; + + r = log_debug_errno(errno, "Failed to open root directory %s: %m", p); + goto fail; + } + } else { + int dfd; + + /* If there's no path specified, then we use the top-level fd itself. We duplicate the fd here, since + * opendir() will take possession of the fd, and close it, which we don't want. */ + + p = "."; /* store this as "." in the directories hashmap */ + + dfd = fcntl(j->toplevel_fd, F_DUPFD_CLOEXEC, 3); + if (dfd < 0) { + r = -errno; + goto fail; + } + + d = fdopendir(dfd); + if (!d) { + r = -errno; + safe_close(dfd); + goto fail; + } + + rewinddir(d); + } + + m = hashmap_get(j->directories_by_path, p); + if (!m) { + m = new0(Directory, 1); + if (!m) { + r = -ENOMEM; + goto fail; + } + + m->is_root = true; + + m->path = strdup(p); + if (!m->path) { + free(m); + r = -ENOMEM; + goto fail; + } + + if (hashmap_put(j->directories_by_path, m->path, m) < 0) { + free(m->path); + free(m); + r = -ENOMEM; + goto fail; + } + + j->current_invalidate_counter++; + + log_debug("Root directory %s added.", m->path); + + } else if (!m->is_root) + return 0; + + if (m->wd <= 0 && j->inotify_fd >= 0) { + + m->wd = inotify_add_watch_fd(j->inotify_fd, dirfd(d), + IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB|IN_DELETE| + IN_ONLYDIR); + + if (m->wd > 0 && hashmap_put(j->directories_by_wd, INT_TO_PTR(m->wd), m) < 0) + inotify_rm_watch(j->inotify_fd, m->wd); + } + + if (j->no_new_files) + return 0; + + FOREACH_DIRENT_ALL(de, d, r = log_debug_errno(errno, "Failed to read directory %s: %m", m->path); goto fail) { + sd_id128_t id; + + if (dirent_is_file_with_suffix(de, ".journal") || + dirent_is_file_with_suffix(de, ".journal~")) + (void) add_file(j, m->path, de->d_name); + else if (IN_SET(de->d_type, DT_DIR, DT_LNK, DT_UNKNOWN) && + sd_id128_from_string(de->d_name, &id) >= 0) + (void) add_directory(j, m->path, de->d_name); + } + + check_network(j, dirfd(d)); + + return 0; + +fail: + k = journal_put_error(j, r, p); + if (k < 0) + return k; + + return r; +} + +static void remove_directory(sd_journal *j, Directory *d) { + assert(j); + + if (d->wd > 0) { + hashmap_remove(j->directories_by_wd, INT_TO_PTR(d->wd)); + + if (j->inotify_fd >= 0) + inotify_rm_watch(j->inotify_fd, d->wd); + } + + hashmap_remove(j->directories_by_path, d->path); + + if (d->is_root) + log_debug("Root directory %s removed.", d->path); + else + log_debug("Directory %s removed.", d->path); + + free(d->path); + free(d); +} + +static int add_search_paths(sd_journal *j) { + + static const char search_paths[] = + "/run/log/journal\0" + "/var/log/journal\0"; + const char *p; + + assert(j); + + /* We ignore most errors here, since the idea is to only open + * what's actually accessible, and ignore the rest. */ + + NULSTR_FOREACH(p, search_paths) + (void) add_root_directory(j, p, true); + + return 0; +} + +static int add_current_paths(sd_journal *j) { + Iterator i; + JournalFile *f; + + assert(j); + assert(j->no_new_files); + + /* Simply adds all directories for files we have open as directories. We don't expect errors here, so we + * treat them as fatal. */ + + ORDERED_HASHMAP_FOREACH(f, j->files, i) { + _cleanup_free_ char *dir; + int r; + + dir = dirname_malloc(f->path); + if (!dir) + return -ENOMEM; + + r = add_directory(j, dir, NULL); + if (r < 0) + return r; + } + + return 0; +} + +static int allocate_inotify(sd_journal *j) { + assert(j); + + if (j->inotify_fd < 0) { + j->inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC); + if (j->inotify_fd < 0) + return -errno; + } + + return hashmap_ensure_allocated(&j->directories_by_wd, NULL); +} + +static sd_journal *journal_new(int flags, const char *path) { + sd_journal *j; + + j = new0(sd_journal, 1); + if (!j) + return NULL; + + j->original_pid = getpid(); + j->toplevel_fd = -1; + j->inotify_fd = -1; + j->flags = flags; + j->data_threshold = DEFAULT_DATA_THRESHOLD; + + if (path) { + char *t; + + t = strdup(path); + if (!t) + goto fail; + + if (flags & SD_JOURNAL_OS_ROOT) + j->prefix = t; + else + j->path = t; + } + + j->files = ordered_hashmap_new(&string_hash_ops); + j->directories_by_path = hashmap_new(&string_hash_ops); + j->mmap = mmap_cache_new(); + if (!j->files || !j->directories_by_path || !j->mmap) + goto fail; + + return j; + +fail: + sd_journal_close(j); + return NULL; +} + +#define OPEN_ALLOWED_FLAGS \ + (SD_JOURNAL_LOCAL_ONLY | \ + SD_JOURNAL_RUNTIME_ONLY | \ + SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER) + +_public_ int sd_journal_open(sd_journal **ret, int flags) { + sd_journal *j; + int r; + + assert_return(ret, -EINVAL); + assert_return((flags & ~OPEN_ALLOWED_FLAGS) == 0, -EINVAL); + + j = journal_new(flags, NULL); + if (!j) + return -ENOMEM; + + r = add_search_paths(j); + if (r < 0) + goto fail; + + *ret = j; + return 0; + +fail: + sd_journal_close(j); + + return r; +} + +#define OPEN_CONTAINER_ALLOWED_FLAGS \ + (SD_JOURNAL_LOCAL_ONLY | SD_JOURNAL_SYSTEM) + +_public_ int sd_journal_open_container(sd_journal **ret, const char *machine, int flags) { + _cleanup_free_ char *root = NULL, *class = NULL; + sd_journal *j; + char *p; + int r; + + /* This is pretty much deprecated, people should use machined's OpenMachineRootDirectory() call instead in + * combination with sd_journal_open_directory_fd(). */ + + assert_return(machine, -EINVAL); + assert_return(ret, -EINVAL); + assert_return((flags & ~OPEN_CONTAINER_ALLOWED_FLAGS) == 0, -EINVAL); + assert_return(machine_name_is_valid(machine), -EINVAL); + + p = strjoina("/run/systemd/machines/", machine); + r = parse_env_file(p, NEWLINE, "ROOT", &root, "CLASS", &class, NULL); + if (r == -ENOENT) + return -EHOSTDOWN; + if (r < 0) + return r; + if (!root) + return -ENODATA; + + if (!streq_ptr(class, "container")) + return -EIO; + + j = journal_new(flags, root); + if (!j) + return -ENOMEM; + + r = add_search_paths(j); + if (r < 0) + goto fail; + + *ret = j; + return 0; + +fail: + sd_journal_close(j); + return r; +} + +#define OPEN_DIRECTORY_ALLOWED_FLAGS \ + (SD_JOURNAL_OS_ROOT | \ + SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER ) + +_public_ int sd_journal_open_directory(sd_journal **ret, const char *path, int flags) { + sd_journal *j; + int r; + + assert_return(ret, -EINVAL); + assert_return(path, -EINVAL); + assert_return((flags & ~OPEN_DIRECTORY_ALLOWED_FLAGS) == 0, -EINVAL); + + j = journal_new(flags, path); + if (!j) + return -ENOMEM; + + if (flags & SD_JOURNAL_OS_ROOT) + r = add_search_paths(j); + else + r = add_root_directory(j, path, false); + if (r < 0) + goto fail; + + *ret = j; + return 0; + +fail: + sd_journal_close(j); + return r; +} + +_public_ int sd_journal_open_files(sd_journal **ret, const char **paths, int flags) { + sd_journal *j; + const char **path; + int r; + + assert_return(ret, -EINVAL); + assert_return(flags == 0, -EINVAL); + + j = journal_new(flags, NULL); + if (!j) + return -ENOMEM; + + STRV_FOREACH(path, paths) { + r = add_any_file(j, -1, *path); + if (r < 0) + goto fail; + } + + j->no_new_files = true; + + *ret = j; + return 0; + +fail: + sd_journal_close(j); + return r; +} + +#define OPEN_DIRECTORY_FD_ALLOWED_FLAGS \ + (SD_JOURNAL_OS_ROOT | \ + SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER ) + +_public_ int sd_journal_open_directory_fd(sd_journal **ret, int fd, int flags) { + sd_journal *j; + struct stat st; + int r; + + assert_return(ret, -EINVAL); + assert_return(fd >= 0, -EBADF); + assert_return((flags & ~OPEN_DIRECTORY_FD_ALLOWED_FLAGS) == 0, -EINVAL); + + if (fstat(fd, &st) < 0) + return -errno; + + if (!S_ISDIR(st.st_mode)) + return -EBADFD; + + j = journal_new(flags, NULL); + if (!j) + return -ENOMEM; + + j->toplevel_fd = fd; + + if (flags & SD_JOURNAL_OS_ROOT) + r = add_search_paths(j); + else + r = add_root_directory(j, NULL, false); + if (r < 0) + goto fail; + + *ret = j; + return 0; + +fail: + sd_journal_close(j); + return r; +} + +_public_ int sd_journal_open_files_fd(sd_journal **ret, int fds[], unsigned n_fds, int flags) { + Iterator iterator; + JournalFile *f; + sd_journal *j; + unsigned i; + int r; + + assert_return(ret, -EINVAL); + assert_return(n_fds > 0, -EBADF); + assert_return(flags == 0, -EINVAL); + + j = journal_new(flags, NULL); + if (!j) + return -ENOMEM; + + for (i = 0; i < n_fds; i++) { + struct stat st; + + if (fds[i] < 0) { + r = -EBADF; + goto fail; + } + + if (fstat(fds[i], &st) < 0) { + r = -errno; + goto fail; + } + + if (!S_ISREG(st.st_mode)) { + r = -EBADFD; + goto fail; + } + + r = add_any_file(j, fds[i], NULL); + if (r < 0) + goto fail; + } + + j->no_new_files = true; + j->no_inotify = true; + + *ret = j; + return 0; + +fail: + /* If we fail, make sure we don't take possession of the files we managed to make use of successfully, and they + * remain open */ + ORDERED_HASHMAP_FOREACH(f, j->files, iterator) + f->close_fd = false; + + sd_journal_close(j); + return r; +} + +_public_ void sd_journal_close(sd_journal *j) { + Directory *d; + JournalFile *f; + char *p; + + if (!j) + return; + + sd_journal_flush_matches(j); + + while ((f = ordered_hashmap_steal_first(j->files))) + (void) journal_file_close(f); + + ordered_hashmap_free(j->files); + + while ((d = hashmap_first(j->directories_by_path))) + remove_directory(j, d); + + while ((d = hashmap_first(j->directories_by_wd))) + remove_directory(j, d); + + hashmap_free(j->directories_by_path); + hashmap_free(j->directories_by_wd); + + safe_close(j->inotify_fd); + + if (j->mmap) { + log_debug("mmap cache statistics: %u hit, %u miss", mmap_cache_get_hit(j->mmap), mmap_cache_get_missed(j->mmap)); + mmap_cache_unref(j->mmap); + } + + while ((p = hashmap_steal_first(j->errors))) + free(p); + hashmap_free(j->errors); + + free(j->path); + free(j->prefix); + free(j->unique_field); + free(j->fields_buffer); + free(j); +} + +_public_ int sd_journal_get_realtime_usec(sd_journal *j, uint64_t *ret) { + Object *o; + JournalFile *f; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(ret, -EINVAL); + + f = j->current_file; + if (!f) + return -EADDRNOTAVAIL; + + if (f->current_offset <= 0) + return -EADDRNOTAVAIL; + + r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); + if (r < 0) + return r; + + *ret = le64toh(o->entry.realtime); + return 0; +} + +_public_ int sd_journal_get_monotonic_usec(sd_journal *j, uint64_t *ret, sd_id128_t *ret_boot_id) { + Object *o; + JournalFile *f; + int r; + sd_id128_t id; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + f = j->current_file; + if (!f) + return -EADDRNOTAVAIL; + + if (f->current_offset <= 0) + return -EADDRNOTAVAIL; + + r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); + if (r < 0) + return r; + + if (ret_boot_id) + *ret_boot_id = o->entry.boot_id; + else { + r = sd_id128_get_boot(&id); + if (r < 0) + return r; + + if (!sd_id128_equal(id, o->entry.boot_id)) + return -ESTALE; + } + + if (ret) + *ret = le64toh(o->entry.monotonic); + + return 0; +} + +static bool field_is_valid(const char *field) { + const char *p; + + assert(field); + + if (isempty(field)) + return false; + + if (startswith(field, "__")) + return false; + + for (p = field; *p; p++) { + + if (*p == '_') + continue; + + if (*p >= 'A' && *p <= 'Z') + continue; + + if (*p >= '0' && *p <= '9') + continue; + + return false; + } + + return true; +} + +_public_ int sd_journal_get_data(sd_journal *j, const char *field, const void **data, size_t *size) { + JournalFile *f; + uint64_t i, n; + size_t field_length; + int r; + Object *o; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(field, -EINVAL); + assert_return(data, -EINVAL); + assert_return(size, -EINVAL); + assert_return(field_is_valid(field), -EINVAL); + + f = j->current_file; + if (!f) + return -EADDRNOTAVAIL; + + if (f->current_offset <= 0) + return -EADDRNOTAVAIL; + + r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); + if (r < 0) + return r; + + field_length = strlen(field); + + n = journal_file_entry_n_items(o); + for (i = 0; i < n; i++) { + uint64_t p, l; + le64_t le_hash; + size_t t; + int compression; + + p = le64toh(o->entry.items[i].object_offset); + le_hash = o->entry.items[i].hash; + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + if (le_hash != o->data.hash) + return -EBADMSG; + + l = le64toh(o->object.size) - offsetof(Object, data.payload); + + compression = o->object.flags & OBJECT_COMPRESSION_MASK; + if (compression) { +#if defined(HAVE_XZ) || defined(HAVE_LZ4) + r = decompress_startswith(compression, + o->data.payload, l, + &f->compress_buffer, &f->compress_buffer_size, + field, field_length, '='); + if (r < 0) + log_debug_errno(r, "Cannot decompress %s object of length %"PRIu64" at offset "OFSfmt": %m", + object_compressed_to_string(compression), l, p); + else if (r > 0) { + + size_t rsize; + + r = decompress_blob(compression, + o->data.payload, l, + &f->compress_buffer, &f->compress_buffer_size, &rsize, + j->data_threshold); + if (r < 0) + return r; + + *data = f->compress_buffer; + *size = (size_t) rsize; + + return 0; + } +#else + return -EPROTONOSUPPORT; +#endif + } else if (l >= field_length+1 && + memcmp(o->data.payload, field, field_length) == 0 && + o->data.payload[field_length] == '=') { + + t = (size_t) l; + + if ((uint64_t) t != l) + return -E2BIG; + + *data = o->data.payload; + *size = t; + + return 0; + } + + r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); + if (r < 0) + return r; + } + + return -ENOENT; +} + +static int return_data(sd_journal *j, JournalFile *f, Object *o, const void **data, size_t *size) { + size_t t; + uint64_t l; + int compression; + + l = le64toh(o->object.size) - offsetof(Object, data.payload); + t = (size_t) l; + + /* We can't read objects larger than 4G on a 32bit machine */ + if ((uint64_t) t != l) + return -E2BIG; + + compression = o->object.flags & OBJECT_COMPRESSION_MASK; + if (compression) { +#if defined(HAVE_XZ) || defined(HAVE_LZ4) + size_t rsize; + int r; + + r = decompress_blob(compression, + o->data.payload, l, &f->compress_buffer, + &f->compress_buffer_size, &rsize, j->data_threshold); + if (r < 0) + return r; + + *data = f->compress_buffer; + *size = (size_t) rsize; +#else + return -EPROTONOSUPPORT; +#endif + } else { + *data = o->data.payload; + *size = t; + } + + return 0; +} + +_public_ int sd_journal_enumerate_data(sd_journal *j, const void **data, size_t *size) { + JournalFile *f; + uint64_t p, n; + le64_t le_hash; + int r; + Object *o; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(data, -EINVAL); + assert_return(size, -EINVAL); + + f = j->current_file; + if (!f) + return -EADDRNOTAVAIL; + + if (f->current_offset <= 0) + return -EADDRNOTAVAIL; + + r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); + if (r < 0) + return r; + + n = journal_file_entry_n_items(o); + if (j->current_field >= n) + return 0; + + p = le64toh(o->entry.items[j->current_field].object_offset); + le_hash = o->entry.items[j->current_field].hash; + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + if (le_hash != o->data.hash) + return -EBADMSG; + + r = return_data(j, f, o, data, size); + if (r < 0) + return r; + + j->current_field++; + + return 1; +} + +_public_ void sd_journal_restart_data(sd_journal *j) { + if (!j) + return; + + j->current_field = 0; +} + +_public_ int sd_journal_get_fd(sd_journal *j) { + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + if (j->no_inotify) + return -EMEDIUMTYPE; + + if (j->inotify_fd >= 0) + return j->inotify_fd; + + r = allocate_inotify(j); + if (r < 0) + return r; + + log_debug("Reiterating files to get inotify watches established"); + + /* Iterate through all dirs again, to add them to the + * inotify */ + if (j->no_new_files) + r = add_current_paths(j); + else if (j->flags & SD_JOURNAL_OS_ROOT) + r = add_search_paths(j); + else if (j->toplevel_fd >= 0) + r = add_root_directory(j, NULL, false); + else if (j->path) + r = add_root_directory(j, j->path, true); + else + r = add_search_paths(j); + if (r < 0) + return r; + + return j->inotify_fd; +} + +_public_ int sd_journal_get_events(sd_journal *j) { + int fd; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + fd = sd_journal_get_fd(j); + if (fd < 0) + return fd; + + return POLLIN; +} + +_public_ int sd_journal_get_timeout(sd_journal *j, uint64_t *timeout_usec) { + int fd; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(timeout_usec, -EINVAL); + + fd = sd_journal_get_fd(j); + if (fd < 0) + return fd; + + if (!j->on_network) { + *timeout_usec = (uint64_t) -1; + return 0; + } + + /* If we are on the network we need to regularly check for + * changes manually */ + + *timeout_usec = j->last_process_usec + JOURNAL_FILES_RECHECK_USEC; + return 1; +} + +static void process_inotify_event(sd_journal *j, struct inotify_event *e) { + Directory *d; + + assert(j); + assert(e); + + /* Is this a subdirectory we watch? */ + d = hashmap_get(j->directories_by_wd, INT_TO_PTR(e->wd)); + if (d) { + sd_id128_t id; + + if (!(e->mask & IN_ISDIR) && e->len > 0 && + (endswith(e->name, ".journal") || + endswith(e->name, ".journal~"))) { + + /* Event for a journal file */ + + if (e->mask & (IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB)) + (void) add_file(j, d->path, e->name); + else if (e->mask & (IN_DELETE|IN_MOVED_FROM|IN_UNMOUNT)) + remove_file(j, d->path, e->name); + + } else if (!d->is_root && e->len == 0) { + + /* Event for a subdirectory */ + + if (e->mask & (IN_DELETE_SELF|IN_MOVE_SELF|IN_UNMOUNT)) + remove_directory(j, d); + + } else if (d->is_root && (e->mask & IN_ISDIR) && e->len > 0 && sd_id128_from_string(e->name, &id) >= 0) { + + /* Event for root directory */ + + if (e->mask & (IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB)) + (void) add_directory(j, d->path, e->name); + } + + return; + } + + if (e->mask & IN_IGNORED) + return; + + log_debug("Unknown inotify event."); +} + +static int determine_change(sd_journal *j) { + bool b; + + assert(j); + + b = j->current_invalidate_counter != j->last_invalidate_counter; + j->last_invalidate_counter = j->current_invalidate_counter; + + return b ? SD_JOURNAL_INVALIDATE : SD_JOURNAL_APPEND; +} + +_public_ int sd_journal_process(sd_journal *j) { + bool got_something = false; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + j->last_process_usec = now(CLOCK_MONOTONIC); + + for (;;) { + union inotify_event_buffer buffer; + struct inotify_event *e; + ssize_t l; + + l = read(j->inotify_fd, &buffer, sizeof(buffer)); + if (l < 0) { + if (errno == EAGAIN || errno == EINTR) + return got_something ? determine_change(j) : SD_JOURNAL_NOP; + + return -errno; + } + + got_something = true; + + FOREACH_INOTIFY_EVENT(e, buffer, l) + process_inotify_event(j, e); + } +} + +_public_ int sd_journal_wait(sd_journal *j, uint64_t timeout_usec) { + int r; + uint64_t t; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + if (j->inotify_fd < 0) { + + /* This is the first invocation, hence create the + * inotify watch */ + r = sd_journal_get_fd(j); + if (r < 0) + return r; + + /* The journal might have changed since the context + * object was created and we weren't watching before, + * hence don't wait for anything, and return + * immediately. */ + return determine_change(j); + } + + r = sd_journal_get_timeout(j, &t); + if (r < 0) + return r; + + if (t != (uint64_t) -1) { + usec_t n; + + n = now(CLOCK_MONOTONIC); + t = t > n ? t - n : 0; + + if (timeout_usec == (uint64_t) -1 || timeout_usec > t) + timeout_usec = t; + } + + do { + r = fd_wait_for_event(j->inotify_fd, POLLIN, timeout_usec); + } while (r == -EINTR); + + if (r < 0) + return r; + + return sd_journal_process(j); +} + +_public_ int sd_journal_get_cutoff_realtime_usec(sd_journal *j, uint64_t *from, uint64_t *to) { + Iterator i; + JournalFile *f; + bool first = true; + uint64_t fmin = 0, tmax = 0; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(from || to, -EINVAL); + assert_return(from != to, -EINVAL); + + ORDERED_HASHMAP_FOREACH(f, j->files, i) { + usec_t fr, t; + + r = journal_file_get_cutoff_realtime_usec(f, &fr, &t); + if (r == -ENOENT) + continue; + if (r < 0) + return r; + if (r == 0) + continue; + + if (first) { + fmin = fr; + tmax = t; + first = false; + } else { + fmin = MIN(fr, fmin); + tmax = MAX(t, tmax); + } + } + + if (from) + *from = fmin; + if (to) + *to = tmax; + + return first ? 0 : 1; +} + +_public_ int sd_journal_get_cutoff_monotonic_usec(sd_journal *j, sd_id128_t boot_id, uint64_t *from, uint64_t *to) { + Iterator i; + JournalFile *f; + bool found = false; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(from || to, -EINVAL); + assert_return(from != to, -EINVAL); + + ORDERED_HASHMAP_FOREACH(f, j->files, i) { + usec_t fr, t; + + r = journal_file_get_cutoff_monotonic_usec(f, boot_id, &fr, &t); + if (r == -ENOENT) + continue; + if (r < 0) + return r; + if (r == 0) + continue; + + if (found) { + if (from) + *from = MIN(fr, *from); + if (to) + *to = MAX(t, *to); + } else { + if (from) + *from = fr; + if (to) + *to = t; + found = true; + } + } + + return found; +} + +void journal_print_header(sd_journal *j) { + Iterator i; + JournalFile *f; + bool newline = false; + + assert(j); + + ORDERED_HASHMAP_FOREACH(f, j->files, i) { + if (newline) + putchar('\n'); + else + newline = true; + + journal_file_print_header(f); + } +} + +_public_ int sd_journal_get_usage(sd_journal *j, uint64_t *bytes) { + Iterator i; + JournalFile *f; + uint64_t sum = 0; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(bytes, -EINVAL); + + ORDERED_HASHMAP_FOREACH(f, j->files, i) { + struct stat st; + + if (fstat(f->fd, &st) < 0) + return -errno; + + sum += (uint64_t) st.st_blocks * 512ULL; + } + + *bytes = sum; + return 0; +} + +_public_ int sd_journal_query_unique(sd_journal *j, const char *field) { + char *f; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(!isempty(field), -EINVAL); + assert_return(field_is_valid(field), -EINVAL); + + f = strdup(field); + if (!f) + return -ENOMEM; + + free(j->unique_field); + j->unique_field = f; + j->unique_file = NULL; + j->unique_offset = 0; + j->unique_file_lost = false; + + return 0; +} + +_public_ int sd_journal_enumerate_unique(sd_journal *j, const void **data, size_t *l) { + size_t k; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(data, -EINVAL); + assert_return(l, -EINVAL); + assert_return(j->unique_field, -EINVAL); + + k = strlen(j->unique_field); + + if (!j->unique_file) { + if (j->unique_file_lost) + return 0; + + j->unique_file = ordered_hashmap_first(j->files); + if (!j->unique_file) + return 0; + + j->unique_offset = 0; + } + + for (;;) { + JournalFile *of; + Iterator i; + Object *o; + const void *odata; + size_t ol; + bool found; + int r; + + /* Proceed to next data object in the field's linked list */ + if (j->unique_offset == 0) { + r = journal_file_find_field_object(j->unique_file, j->unique_field, k, &o, NULL); + if (r < 0) + return r; + + j->unique_offset = r > 0 ? le64toh(o->field.head_data_offset) : 0; + } else { + r = journal_file_move_to_object(j->unique_file, OBJECT_DATA, j->unique_offset, &o); + if (r < 0) + return r; + + j->unique_offset = le64toh(o->data.next_field_offset); + } + + /* We reached the end of the list? Then start again, with the next file */ + if (j->unique_offset == 0) { + j->unique_file = ordered_hashmap_next(j->files, j->unique_file->path); + if (!j->unique_file) + return 0; + + continue; + } + + /* We do not use OBJECT_DATA context here, but OBJECT_UNUSED + * instead, so that we can look at this data object at the same + * time as one on another file */ + r = journal_file_move_to_object(j->unique_file, OBJECT_UNUSED, j->unique_offset, &o); + if (r < 0) + return r; + + /* Let's do the type check by hand, since we used 0 context above. */ + if (o->object.type != OBJECT_DATA) { + log_debug("%s:offset " OFSfmt ": object has type %d, expected %d", + j->unique_file->path, j->unique_offset, + o->object.type, OBJECT_DATA); + return -EBADMSG; + } + + r = return_data(j, j->unique_file, o, &odata, &ol); + if (r < 0) + return r; + + /* Check if we have at least the field name and "=". */ + if (ol <= k) { + log_debug("%s:offset " OFSfmt ": object has size %zu, expected at least %zu", + j->unique_file->path, j->unique_offset, + ol, k + 1); + return -EBADMSG; + } + + if (memcmp(odata, j->unique_field, k) || ((const char*) odata)[k] != '=') { + log_debug("%s:offset " OFSfmt ": object does not start with \"%s=\"", + j->unique_file->path, j->unique_offset, + j->unique_field); + return -EBADMSG; + } + + /* OK, now let's see if we already returned this data + * object by checking if it exists in the earlier + * traversed files. */ + found = false; + ORDERED_HASHMAP_FOREACH(of, j->files, i) { + if (of == j->unique_file) + break; + + /* Skip this file it didn't have any fields indexed */ + if (JOURNAL_HEADER_CONTAINS(of->header, n_fields) && le64toh(of->header->n_fields) <= 0) + continue; + + r = journal_file_find_data_object_with_hash(of, odata, ol, le64toh(o->data.hash), NULL, NULL); + if (r < 0) + return r; + if (r > 0) { + found = true; + break; + } + } + + if (found) + continue; + + r = return_data(j, j->unique_file, o, data, l); + if (r < 0) + return r; + + return 1; + } +} + +_public_ void sd_journal_restart_unique(sd_journal *j) { + if (!j) + return; + + j->unique_file = NULL; + j->unique_offset = 0; + j->unique_file_lost = false; +} + +_public_ int sd_journal_enumerate_fields(sd_journal *j, const char **field) { + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(field, -EINVAL); + + if (!j->fields_file) { + if (j->fields_file_lost) + return 0; + + j->fields_file = ordered_hashmap_first(j->files); + if (!j->fields_file) + return 0; + + j->fields_hash_table_index = 0; + j->fields_offset = 0; + } + + for (;;) { + JournalFile *f, *of; + Iterator i; + uint64_t m; + Object *o; + size_t sz; + bool found; + + f = j->fields_file; + + if (j->fields_offset == 0) { + bool eof = false; + + /* We are not yet positioned at any field. Let's pick the first one */ + r = journal_file_map_field_hash_table(f); + if (r < 0) + return r; + + m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem); + for (;;) { + if (j->fields_hash_table_index >= m) { + /* Reached the end of the hash table, go to the next file. */ + eof = true; + break; + } + + j->fields_offset = le64toh(f->field_hash_table[j->fields_hash_table_index].head_hash_offset); + + if (j->fields_offset != 0) + break; + + /* Empty hash table bucket, go to next one */ + j->fields_hash_table_index++; + } + + if (eof) { + /* Proceed with next file */ + j->fields_file = ordered_hashmap_next(j->files, f->path); + if (!j->fields_file) { + *field = NULL; + return 0; + } + + j->fields_offset = 0; + j->fields_hash_table_index = 0; + continue; + } + + } else { + /* We are already positioned at a field. If so, let's figure out the next field from it */ + + r = journal_file_move_to_object(f, OBJECT_FIELD, j->fields_offset, &o); + if (r < 0) + return r; + + j->fields_offset = le64toh(o->field.next_hash_offset); + if (j->fields_offset == 0) { + /* Reached the end of the hash table chain */ + j->fields_hash_table_index++; + continue; + } + } + + /* We use OBJECT_UNUSED here, so that the iterator below doesn't remove our mmap window */ + r = journal_file_move_to_object(f, OBJECT_UNUSED, j->fields_offset, &o); + if (r < 0) + return r; + + /* Because we used OBJECT_UNUSED above, we need to do our type check manually */ + if (o->object.type != OBJECT_FIELD) { + log_debug("%s:offset " OFSfmt ": object has type %i, expected %i", f->path, j->fields_offset, o->object.type, OBJECT_FIELD); + return -EBADMSG; + } + + sz = le64toh(o->object.size) - offsetof(Object, field.payload); + + /* Let's see if we already returned this field name before. */ + found = false; + ORDERED_HASHMAP_FOREACH(of, j->files, i) { + if (of == f) + break; + + /* Skip this file it didn't have any fields indexed */ + if (JOURNAL_HEADER_CONTAINS(of->header, n_fields) && le64toh(of->header->n_fields) <= 0) + continue; + + r = journal_file_find_field_object_with_hash(of, o->field.payload, sz, le64toh(o->field.hash), NULL, NULL); + if (r < 0) + return r; + if (r > 0) { + found = true; + break; + } + } + + if (found) + continue; + + /* Check if this is really a valid string containing no NUL byte */ + if (memchr(o->field.payload, 0, sz)) + return -EBADMSG; + + if (sz > j->data_threshold) + sz = j->data_threshold; + + if (!GREEDY_REALLOC(j->fields_buffer, j->fields_buffer_allocated, sz + 1)) + return -ENOMEM; + + memcpy(j->fields_buffer, o->field.payload, sz); + j->fields_buffer[sz] = 0; + + if (!field_is_valid(j->fields_buffer)) + return -EBADMSG; + + *field = j->fields_buffer; + return 1; + } +} + +_public_ void sd_journal_restart_fields(sd_journal *j) { + if (!j) + return; + + j->fields_file = NULL; + j->fields_hash_table_index = 0; + j->fields_offset = 0; + j->fields_file_lost = false; +} + +_public_ int sd_journal_reliable_fd(sd_journal *j) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + return !j->on_network; +} + +static char *lookup_field(const char *field, void *userdata) { + sd_journal *j = userdata; + const void *data; + size_t size, d; + int r; + + assert(field); + assert(j); + + r = sd_journal_get_data(j, field, &data, &size); + if (r < 0 || + size > REPLACE_VAR_MAX) + return strdup(field); + + d = strlen(field) + 1; + + return strndup((const char*) data + d, size - d); +} + +_public_ int sd_journal_get_catalog(sd_journal *j, char **ret) { + const void *data; + size_t size; + sd_id128_t id; + _cleanup_free_ char *text = NULL, *cid = NULL; + char *t; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(ret, -EINVAL); + + r = sd_journal_get_data(j, "MESSAGE_ID", &data, &size); + if (r < 0) + return r; + + cid = strndup((const char*) data + 11, size - 11); + if (!cid) + return -ENOMEM; + + r = sd_id128_from_string(cid, &id); + if (r < 0) + return r; + + r = catalog_get(CATALOG_DATABASE, id, &text); + if (r < 0) + return r; + + t = replace_var(text, lookup_field, j); + if (!t) + return -ENOMEM; + + *ret = t; + return 0; +} + +_public_ int sd_journal_get_catalog_for_message_id(sd_id128_t id, char **ret) { + assert_return(ret, -EINVAL); + + return catalog_get(CATALOG_DATABASE, id, ret); +} + +_public_ int sd_journal_set_data_threshold(sd_journal *j, size_t sz) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + j->data_threshold = sz; + return 0; +} + +_public_ int sd_journal_get_data_threshold(sd_journal *j, size_t *sz) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(sz, -EINVAL); + + *sz = j->data_threshold; + return 0; +} + +_public_ int sd_journal_has_runtime_files(sd_journal *j) { + assert_return(j, -EINVAL); + + return j->has_runtime_files; +} + +_public_ int sd_journal_has_persistent_files(sd_journal *j) { + assert_return(j, -EINVAL); + + return j->has_persistent_files; +} |