journal: defer journal closes on rotate

When we rotate journals, we must set offline and close the current one,
but don't generally need to wait for this to complete.

Instead, we'll initiate an asynchronous offline via
journal_file_set_offline(oldfile, false), and add the file to a
per-server set of deferred closes to be closed later when they
won't block.

There's one complication however; journal_file_open() via
journal_file_verify_header() assumes that any writable journal in the
online state is the product of an unclean shutdown or other form of
corruption.

Thus there's a need for journal_file_open() to be aware of deferred
closes and synchronize with their completion when opening preexisting
journals for writing.  To facilitate this the deferred closes set is
supplied to the journal_file_open() function where the deferred closes
may be closed synchronously before verifying the header in such
circumstances.

10 files changed, 82 insertions, 28 deletions

diff --git a/src/journal/journal-file.c b/src/journal/journal-file.c
index 96be339d5b..f5e2952c99 100644
--- a/src/journal/journal-file.c
+++ b/src/journal/journal-file.c
@@ -39,6 +39,7 @@
 #include "parse-util.h"
 #include "random-util.h"
 #include "sd-event.h"
+#include "set.h"
 #include "string-util.h"
 #include "xattr-util.h"
 
@@ -310,6 +311,18 @@ static int journal_file_set_online(JournalFile *f) {
         }
 }
 
+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);
 
@@ -374,6 +387,15 @@ JournalFile* journal_file_close(JournalFile *f) {
         return NULL;
 }
 
+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;
@@ -2881,6 +2903,7 @@ int journal_file_open(
                 bool seal,
                 JournalMetrics *metrics,
                 MMapCache *mmap_cache,
+                Set *deferred_closes,
                 JournalFile *template,
                 JournalFile **ret) {
 
@@ -3000,6 +3023,9 @@ int journal_file_open(
         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;
@@ -3074,7 +3100,7 @@ fail:
         return r;
 }
 
-int journal_file_rotate(JournalFile **f, bool compress, bool seal) {
+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;
@@ -3114,8 +3140,13 @@ int journal_file_rotate(JournalFile **f, bool compress, bool seal) {
          * we archive them */
         old_file->defrag_on_close = true;
 
-        r = journal_file_open(old_file->path, old_file->flags, old_file->mode, compress, seal, NULL, old_file->mmap, old_file, &new_file);
-        journal_file_close(old_file);
+        r = journal_file_open(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;
@@ -3129,6 +3160,7 @@ int journal_file_open_reliably(
                 bool seal,
                 JournalMetrics *metrics,
                 MMapCache *mmap_cache,
+                Set *deferred_closes,
                 JournalFile *template,
                 JournalFile **ret) {
 
@@ -3136,7 +3168,7 @@ int journal_file_open_reliably(
         size_t l;
         _cleanup_free_ char *p = NULL;
 
-        r = journal_file_open(fname, flags, mode, compress, seal, metrics, mmap_cache, template, ret);
+        r = journal_file_open(fname, flags, mode, compress, seal, metrics, mmap_cache, deferred_closes, template, ret);
         if (!IN_SET(r,
                     -EBADMSG,           /* corrupted */
                     -ENODATA,           /* truncated */
@@ -3177,7 +3209,7 @@ int journal_file_open_reliably(
 
         log_warning_errno(r, "File %s corrupted or uncleanly shut down, renaming and replacing.", fname);
 
-        return journal_file_open(fname, flags, mode, compress, seal, metrics, mmap_cache, template, ret);
+        return journal_file_open(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) {
diff --git a/src/journal/journal-file.h b/src/journal/journal-file.h
index fad4f78bdc..9ad6013359 100644
--- a/src/journal/journal-file.h
+++ b/src/journal/journal-file.h
@@ -149,11 +149,14 @@ int journal_file_open(
                 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,
@@ -163,6 +166,7 @@ int journal_file_open_reliably(
                 bool seal,
                 JournalMetrics *metrics,
                 MMapCache *mmap_cache,
+                Set *deferred_closes,
                 JournalFile *template,
                 JournalFile **ret);
 
@@ -236,7 +240,7 @@ int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint6
 void journal_file_dump(JournalFile *f);
 void journal_file_print_header(JournalFile *f);
 
-int journal_file_rotate(JournalFile **f, bool compress, bool seal);
+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);
diff --git a/src/journal/journald-server.c b/src/journal/journald-server.c
index ac992a8b54..d5937bd013 100644
--- a/src/journal/journald-server.c
+++ b/src/journal/journald-server.c
@@ -251,9 +251,9 @@ static int open_journal(
         assert(ret);
 
         if (reliably)
-                r = journal_file_open_reliably(fname, flags, 0640, s->compress, seal, metrics, s->mmap, NULL, &f);
+                r = journal_file_open_reliably(fname, flags, 0640, s->compress, seal, metrics, s->mmap, s->deferred_closes, NULL, &f);
         else
-                r = journal_file_open(fname, flags, 0640, s->compress, seal, metrics, s->mmap, NULL, &f);
+                r = journal_file_open(fname, flags, 0640, s->compress, seal, metrics, s->mmap, s->deferred_closes, NULL, &f);
         if (r < 0)
                 return r;
 
@@ -333,7 +333,7 @@ static int do_rotate(
         if (!*f)
                 return -EINVAL;
 
-        r = journal_file_rotate(f, s->compress, seal);
+        r = journal_file_rotate(f, s->compress, seal, s->deferred_closes);
         if (r < 0)
                 if (*f)
                         log_error_errno(r, "Failed to rotate %s: %m", (*f)->path);
@@ -364,6 +364,13 @@ void server_rotate(Server *s) {
                         /* Old file has been closed and deallocated */
                         ordered_hashmap_remove(s->user_journals, k);
         }
+
+        /* Perform any deferred closes which aren't still offlining. */
+        SET_FOREACH(f, s->deferred_closes, i)
+                if (!journal_file_is_offlining(f)) {
+                        (void) set_remove(s->deferred_closes, f);
+                        (void) journal_file_close(f);
+                }
 }
 
 void server_sync(Server *s) {
@@ -1765,6 +1772,10 @@ int server_init(Server *s) {
         if (!s->mmap)
                 return log_oom();
 
+        s->deferred_closes = set_new(NULL);
+        if (!s->deferred_closes)
+                return log_oom();
+
         r = sd_event_default(&s->event);
         if (r < 0)
                 return log_error_errno(r, "Failed to create event loop: %m");
@@ -1918,6 +1929,11 @@ void server_done(Server *s) {
         JournalFile *f;
         assert(s);
 
+        if (s->deferred_closes) {
+                journal_file_close_set(s->deferred_closes);
+                set_free(s->deferred_closes);
+        }
+
         while (s->stdout_streams)
                 stdout_stream_free(s->stdout_streams);
 
diff --git a/src/journal/journald-server.h b/src/journal/journald-server.h
index b9551dda1b..e025a4cf90 100644
--- a/src/journal/journald-server.h
+++ b/src/journal/journald-server.h
@@ -130,6 +130,8 @@ struct Server {
 
         MMapCache *mmap;
 
+        Set *deferred_closes;
+
         struct udev *udev;
 
         uint64_t *kernel_seqnum;
diff --git a/src/journal/sd-journal.c b/src/journal/sd-journal.c
index 9bc4215f2e..6ff1c67f5f 100644
--- a/src/journal/sd-journal.c
+++ b/src/journal/sd-journal.c
@@ -1248,7 +1248,7 @@ static int add_any_file(sd_journal *j, const char *path) {
                 goto fail;
         }
 
-        r = journal_file_open(path, O_RDONLY, 0, false, false, NULL, j->mmap, NULL, &f);
+        r = journal_file_open(path, O_RDONLY, 0, false, false, NULL, j->mmap, NULL, NULL, &f);
         if (r < 0) {
                 log_debug_errno(r, "Failed to open journal file %s: %m", path);
                 goto fail;
diff --git a/src/journal/test-journal-flush.c b/src/journal/test-journal-flush.c
index 0c35d532c6..93dc0e0d81 100644
--- a/src/journal/test-journal-flush.c
+++ b/src/journal/test-journal-flush.c
@@ -38,7 +38,7 @@ int main(int argc, char *argv[]) {
         assert_se(mkdtemp(dn));
         fn = strappend(dn, "/test.journal");
 
-        r = journal_file_open(fn, O_CREAT|O_RDWR, 0644, false, false, NULL, NULL, NULL, &new_journal);
+        r = journal_file_open(fn, O_CREAT|O_RDWR, 0644, false, false, NULL, NULL, NULL, NULL, &new_journal);
         assert_se(r >= 0);
 
         r = sd_journal_open(&j, 0);
diff --git a/src/journal/test-journal-interleaving.c b/src/journal/test-journal-interleaving.c
index 6c6238cc4b..f887f43f0d 100644
--- a/src/journal/test-journal-interleaving.c
+++ b/src/journal/test-journal-interleaving.c
@@ -52,7 +52,7 @@ noreturn static void log_assert_errno(const char *text, int eno, const char *fil
 
 static JournalFile *test_open(const char *name) {
         JournalFile *f;
-        assert_ret(journal_file_open(name, O_RDWR|O_CREAT, 0644, true, false, NULL, NULL, NULL, &f));
+        assert_ret(journal_file_open(name, O_RDWR|O_CREAT, 0644, true, false, NULL, NULL, NULL, NULL, &f));
         return f;
 }
 
@@ -217,7 +217,7 @@ static void test_sequence_numbers(void) {
         assert_se(chdir(t) >= 0);
 
         assert_se(journal_file_open("one.journal", O_RDWR|O_CREAT, 0644,
-                                    true, false, NULL, NULL, NULL, &one) == 0);
+                                    true, false, NULL, NULL, NULL, NULL, &one) == 0);
 
         append_number(one, 1, &seqnum);
         printf("seqnum=%"PRIu64"\n", seqnum);
@@ -234,7 +234,7 @@ static void test_sequence_numbers(void) {
         memcpy(&seqnum_id, &one->header->seqnum_id, sizeof(sd_id128_t));
 
         assert_se(journal_file_open("two.journal", O_RDWR|O_CREAT, 0644,
-                                    true, false, NULL, NULL, one, &two) == 0);
+                                    true, false, NULL, NULL, NULL, one, &two) == 0);
 
         assert_se(two->header->state == STATE_ONLINE);
         assert_se(!sd_id128_equal(two->header->file_id, one->header->file_id));
@@ -265,7 +265,7 @@ static void test_sequence_numbers(void) {
         seqnum = 0;
 
         assert_se(journal_file_open("two.journal", O_RDWR, 0,
-                                    true, false, NULL, NULL, NULL, &two) == 0);
+                                    true, false, NULL, NULL, NULL, NULL, &two) == 0);
 
         assert_se(sd_id128_equal(two->header->seqnum_id, seqnum_id));
 
diff --git a/src/journal/test-journal-stream.c b/src/journal/test-journal-stream.c
index 82543bb5f9..839ea5a9a5 100644
--- a/src/journal/test-journal-stream.c
+++ b/src/journal/test-journal-stream.c
@@ -92,9 +92,9 @@ int main(int argc, char *argv[]) {
         assert_se(mkdtemp(t));
         assert_se(chdir(t) >= 0);
 
-        assert_se(journal_file_open("one.journal", O_RDWR|O_CREAT, 0666, true, false, NULL, NULL, NULL, &one) == 0);
-        assert_se(journal_file_open("two.journal", O_RDWR|O_CREAT, 0666, true, false, NULL, NULL, NULL, &two) == 0);
-        assert_se(journal_file_open("three.journal", O_RDWR|O_CREAT, 0666, true, false, NULL, NULL, NULL, &three) == 0);
+        assert_se(journal_file_open("one.journal", O_RDWR|O_CREAT, 0666, true, false, NULL, NULL, NULL, NULL, &one) == 0);
+        assert_se(journal_file_open("two.journal", O_RDWR|O_CREAT, 0666, true, false, NULL, NULL, NULL, NULL, &two) == 0);
+        assert_se(journal_file_open("three.journal", O_RDWR|O_CREAT, 0666, true, false, NULL, NULL, NULL, NULL, &three) == 0);
 
         for (i = 0; i < N_ENTRIES; i++) {
                 char *p, *q;
diff --git a/src/journal/test-journal-verify.c b/src/journal/test-journal-verify.c
index 3fec18c480..6b4643cd25 100644
--- a/src/journal/test-journal-verify.c
+++ b/src/journal/test-journal-verify.c
@@ -55,7 +55,7 @@ static int raw_verify(const char *fn, const char *verification_key) {
         JournalFile *f;
         int r;
 
-        r = journal_file_open(fn, O_RDONLY, 0666, true, !!verification_key, NULL, NULL, NULL, &f);
+        r = journal_file_open(fn, O_RDONLY, 0666, true, !!verification_key, NULL, NULL, NULL, NULL, &f);
         if (r < 0)
                 return r;
 
@@ -88,7 +88,7 @@ int main(int argc, char *argv[]) {
 
         log_info("Generating...");
 
-        assert_se(journal_file_open("test.journal", O_RDWR|O_CREAT, 0666, true, !!verification_key, NULL, NULL, NULL, &f) == 0);
+        assert_se(journal_file_open("test.journal", O_RDWR|O_CREAT, 0666, true, !!verification_key, NULL, NULL, NULL, NULL, &f) == 0);
 
         for (n = 0; n < N_ENTRIES; n++) {
                 struct iovec iovec;
@@ -111,7 +111,7 @@ int main(int argc, char *argv[]) {
 
         log_info("Verifying...");
 
-        assert_se(journal_file_open("test.journal", O_RDONLY, 0666, true, !!verification_key, NULL, NULL, NULL, &f) == 0);
+        assert_se(journal_file_open("test.journal", O_RDONLY, 0666, true, !!verification_key, NULL, NULL, NULL, NULL, &f) == 0);
         /* journal_file_print_header(f); */
         journal_file_dump(f);
 
diff --git a/src/journal/test-journal.c b/src/journal/test-journal.c
index 94d26a2573..ea685af782 100644
--- a/src/journal/test-journal.c
+++ b/src/journal/test-journal.c
@@ -42,7 +42,7 @@ static void test_non_empty(void) {
         assert_se(mkdtemp(t));
         assert_se(chdir(t) >= 0);
 
-        assert_se(journal_file_open("test.journal", O_RDWR|O_CREAT, 0666, true, true, NULL, NULL, NULL, &f) == 0);
+        assert_se(journal_file_open("test.journal", O_RDWR|O_CREAT, 0666, true, true, NULL, NULL, NULL, NULL, &f) == 0);
 
         dual_timestamp_get(&ts);
 
@@ -104,8 +104,8 @@ static void test_non_empty(void) {
 
         assert_se(journal_file_move_to_entry_by_seqnum(f, 10, DIRECTION_DOWN, &o, NULL) == 0);
 
-        journal_file_rotate(&f, true, true);
-        journal_file_rotate(&f, true, true);
+        journal_file_rotate(&f, true, true, NULL);
+        journal_file_rotate(&f, true, true, NULL);
 
         (void) journal_file_close(f);
 
@@ -131,13 +131,13 @@ static void test_empty(void) {
         assert_se(mkdtemp(t));
         assert_se(chdir(t) >= 0);
 
-        assert_se(journal_file_open("test.journal", O_RDWR|O_CREAT, 0666, false, false, NULL, NULL, NULL, &f1) == 0);
+        assert_se(journal_file_open("test.journal", O_RDWR|O_CREAT, 0666, false, false, NULL, NULL, NULL, NULL, &f1) == 0);
 
-        assert_se(journal_file_open("test-compress.journal", O_RDWR|O_CREAT, 0666, true, false, NULL, NULL, NULL, &f2) == 0);
+        assert_se(journal_file_open("test-compress.journal", O_RDWR|O_CREAT, 0666, true, false, NULL, NULL, NULL, NULL, &f2) == 0);
 
-        assert_se(journal_file_open("test-seal.journal", O_RDWR|O_CREAT, 0666, false, true, NULL, NULL, NULL, &f3) == 0);
+        assert_se(journal_file_open("test-seal.journal", O_RDWR|O_CREAT, 0666, false, true, NULL, NULL, NULL, NULL, &f3) == 0);
 
-        assert_se(journal_file_open("test-seal-compress.journal", O_RDWR|O_CREAT, 0666, true, true, NULL, NULL, NULL, &f4) == 0);
+        assert_se(journal_file_open("test-seal-compress.journal", O_RDWR|O_CREAT, 0666, true, true, NULL, NULL, NULL, NULL, &f4) == 0);
 
         journal_file_print_header(f1);
         puts("");