summaryrefslogtreecommitdiff
path: root/src/import/import-job.c
diff options
context:
space:
mode:
authorLennart Poettering <lennart@poettering.net>2015-01-20 01:36:11 +0100
committerLennart Poettering <lennart@poettering.net>2015-01-20 15:06:58 +0100
commit56ebfaf1ca185a93ffb372b6e1a1fa3a957d93cd (patch)
tree247317cdedacfbbc7771e450076e91f08c8ce115 /src/import/import-job.c
parenta2e0337875addaf08225fbf9b231435ba12a88b5 (diff)
import: add support for pulling raw tar balls as containers
Ubuntu provides their cloud images optionally as tarball, hence also support downloading those.
Diffstat (limited to 'src/import/import-job.c')
-rw-r--r--src/import/import-job.c629
1 files changed, 629 insertions, 0 deletions
diff --git a/src/import/import-job.c b/src/import/import-job.c
new file mode 100644
index 0000000000..5a4ea69cb2
--- /dev/null
+++ b/src/import/import-job.c
@@ -0,0 +1,629 @@
+/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
+
+/***
+ This file is part of systemd.
+
+ Copyright 2015 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 <sys/xattr.h>
+
+#include "strv.h"
+#include "import-job.h"
+
+ImportJob* import_job_unref(ImportJob *j) {
+ if (!j)
+ return NULL;
+
+ curl_glue_remove_and_free(j->glue, j->curl);
+ curl_slist_free_all(j->request_header);
+
+ safe_close(j->disk_fd);
+
+ if (j->compressed == IMPORT_JOB_XZ)
+ lzma_end(&j->xz);
+ else if (j->compressed == IMPORT_JOB_GZIP)
+ inflateEnd(&j->gzip);
+
+ free(j->url);
+ free(j->etag);
+ strv_free(j->old_etags);
+ free(j->payload);
+
+ free(j);
+
+ return NULL;
+}
+
+DEFINE_TRIVIAL_CLEANUP_FUNC(ImportJob*, import_job_unref);
+
+static void import_job_finish(ImportJob *j, int ret) {
+ assert(j);
+
+ if (j->state == IMPORT_JOB_DONE ||
+ j->state == IMPORT_JOB_FAILED)
+ return;
+
+ if (ret == 0)
+ j->state = IMPORT_JOB_DONE;
+ else {
+ j->state = IMPORT_JOB_FAILED;
+ j->error = ret;
+ }
+
+ if (j->on_finished)
+ j->on_finished(j);
+}
+
+void import_job_curl_on_finished(CurlGlue *g, CURL *curl, CURLcode result) {
+ ImportJob *j = NULL;
+ CURLcode code;
+ long status;
+ int r;
+
+ if (curl_easy_getinfo(curl, CURLINFO_PRIVATE, &j) != CURLE_OK)
+ return;
+
+ if (!j || j->state == IMPORT_JOB_DONE || j->state == IMPORT_JOB_FAILED)
+ return;
+
+ if (result != CURLE_OK) {
+ log_error("Transfer failed: %s", curl_easy_strerror(result));
+ r = -EIO;
+ goto finish;
+ }
+
+ code = curl_easy_getinfo(curl, CURLINFO_RESPONSE_CODE, &status);
+ if (code != CURLE_OK) {
+ log_error("Failed to retrieve response code: %s", curl_easy_strerror(code));
+ r = -EIO;
+ goto finish;
+ } else if (status == 304) {
+ log_info("Image already downloaded. Skipping download.");
+ r = 0;
+ goto finish;
+ } else if (status >= 300) {
+ log_error("HTTP request to %s failed with code %li.", j->url, status);
+ r = -EIO;
+ goto finish;
+ } else if (status < 200) {
+ log_error("HTTP request to %s finished with unexpected code %li.", j->url, status);
+ r = -EIO;
+ goto finish;
+ }
+
+ if (j->state != IMPORT_JOB_RUNNING) {
+ log_error("Premature connection termination.");
+ r = -EIO;
+ goto finish;
+ }
+
+ if (j->content_length != (uint64_t) -1 &&
+ j->content_length != j->written_compressed) {
+ log_error("Download truncated.");
+ r = -EIO;
+ goto finish;
+ }
+
+ if (j->disk_fd >= 0 && j->allow_sparse) {
+ /* Make sure the file size is right, in case the file was
+ * sparse and we just seeked for the last part */
+
+ if (ftruncate(j->disk_fd, j->written_uncompressed) < 0) {
+ log_error_errno(errno, "Failed to truncate file: %m");
+ r = -errno;
+ goto finish;
+ }
+
+ if (j->etag)
+ (void) fsetxattr(j->disk_fd, "user.source_etag", j->etag, strlen(j->etag), 0);
+ if (j->url)
+ (void) fsetxattr(j->disk_fd, "user.source_url", j->url, strlen(j->url), 0);
+
+ if (j->mtime != 0) {
+ struct timespec ut[2];
+
+ timespec_store(&ut[0], j->mtime);
+ ut[1] = ut[0];
+ (void) futimens(j->disk_fd, ut);
+
+ (void) fd_setcrtime(j->disk_fd, j->mtime);
+ }
+ }
+
+ r = 0;
+
+finish:
+ import_job_finish(j, r);
+}
+
+
+static int import_job_write_uncompressed(ImportJob *j, void *p, size_t sz) {
+ ssize_t n;
+
+ assert(j);
+ assert(p);
+ assert(sz > 0);
+ assert(j->disk_fd >= 0);
+
+ if (j->written_uncompressed + sz < j->written_uncompressed) {
+ log_error("File too large, overflow");
+ return -EOVERFLOW;
+ }
+
+ if (j->written_uncompressed + sz > j->uncompressed_max) {
+ log_error("File overly large, refusing");
+ return -EFBIG;
+ }
+
+ if (j->disk_fd >= 0) {
+
+ if (j->allow_sparse)
+ n = sparse_write(j->disk_fd, p, sz, 64);
+ else
+ n = write(j->disk_fd, p, sz);
+ if (n < 0) {
+ log_error_errno(errno, "Failed to write file: %m");
+ return -errno;
+ }
+ if ((size_t) n < sz) {
+ log_error("Short write");
+ return -EIO;
+ }
+ } else {
+
+ if (!GREEDY_REALLOC(j->payload, j->payload_allocated, j->payload_size + sz))
+ return log_oom();
+
+ memcpy((uint8_t*) j->payload + j->payload_size, p, sz);
+ j->payload_size += sz;
+ }
+
+ j->written_uncompressed += sz;
+
+ return 0;
+}
+
+static int import_job_write_compressed(ImportJob *j, void *p, size_t sz) {
+ int r;
+
+ assert(j);
+ assert(p);
+ assert(sz > 0);
+ assert(j->disk_fd >= 0);
+
+ if (j->written_compressed + sz < j->written_compressed) {
+ log_error("File too large, overflow");
+ return -EOVERFLOW;
+ }
+
+ if (j->written_compressed + sz > j->compressed_max) {
+ log_error("File overly large, refusing.");
+ return -EFBIG;
+ }
+
+ if (j->content_length != (uint64_t) -1 &&
+ j->written_compressed + sz > j->content_length) {
+ log_error("Content length incorrect.");
+ return -EFBIG;
+ }
+
+ switch (j->compressed) {
+
+ case IMPORT_JOB_UNCOMPRESSED:
+ r = import_job_write_uncompressed(j, p, sz);
+ if (r < 0)
+ return r;
+
+ break;
+
+ case IMPORT_JOB_XZ:
+ j->xz.next_in = p;
+ j->xz.avail_in = sz;
+
+ while (j->xz.avail_in > 0) {
+ uint8_t buffer[16 * 1024];
+ lzma_ret lzr;
+
+ j->xz.next_out = buffer;
+ j->xz.avail_out = sizeof(buffer);
+
+ lzr = lzma_code(&j->xz, LZMA_RUN);
+ if (lzr != LZMA_OK && lzr != LZMA_STREAM_END) {
+ log_error("Decompression error.");
+ return -EIO;
+ }
+
+ r = import_job_write_uncompressed(j, buffer, sizeof(buffer) - j->xz.avail_out);
+ if (r < 0)
+ return r;
+ }
+
+ break;
+
+ case IMPORT_JOB_GZIP:
+ j->gzip.next_in = p;
+ j->gzip.avail_in = sz;
+
+ while (j->gzip.avail_in > 0) {
+ uint8_t buffer[16 * 1024];
+
+ j->gzip.next_out = buffer;
+ j->gzip.avail_out = sizeof(buffer);
+
+ r = inflate(&j->gzip, Z_NO_FLUSH);
+ if (r != Z_OK && r != Z_STREAM_END) {
+ log_error("Decompression error.");
+ return -EIO;
+ }
+
+ r = import_job_write_uncompressed(j, buffer, sizeof(buffer) - j->gzip.avail_out);
+ if (r < 0)
+ return r;
+ }
+
+ break;
+
+ default:
+ assert_not_reached("Unknown compression");
+ }
+
+ j->written_compressed += sz;
+
+ return 0;
+}
+
+static int import_job_open_disk(ImportJob *j) {
+ int r;
+
+ assert(j);
+
+ if (j->on_open_disk) {
+ r = j->on_open_disk(j);
+ if (r < 0)
+ return r;
+ }
+
+ if (j->disk_fd >= 0) {
+ /* Check if we can do sparse files */
+
+ if (lseek(j->disk_fd, SEEK_SET, 0) == 0)
+ j->allow_sparse = true;
+ else {
+ if (errno != ESPIPE)
+ return log_error_errno(errno, "Failed to seek on file descriptor: %m");
+
+ j->allow_sparse = false;
+ }
+ }
+
+ return 0;
+}
+
+static int import_job_detect_compression(ImportJob *j) {
+ static const uint8_t xz_signature[] = {
+ 0xfd, '7', 'z', 'X', 'Z', 0x00
+ };
+ static const uint8_t gzip_signature[] = {
+ 0x1f, 0x8b
+ };
+
+ _cleanup_free_ uint8_t *stub = NULL;
+ size_t stub_size;
+
+ int r;
+
+ assert(j);
+
+ if (j->payload_size < MAX(sizeof(xz_signature), sizeof(gzip_signature)))
+ return 0;
+
+ if (memcmp(j->payload, xz_signature, sizeof(xz_signature)) == 0)
+ j->compressed = IMPORT_JOB_XZ;
+ else if (memcmp(j->payload, gzip_signature, sizeof(gzip_signature)) == 0)
+ j->compressed = IMPORT_JOB_GZIP;
+ else
+ j->compressed = IMPORT_JOB_UNCOMPRESSED;
+
+ log_debug("Stream is XZ compressed: %s", yes_no(j->compressed == IMPORT_JOB_XZ));
+ log_debug("Stream is GZIP compressed: %s", yes_no(j->compressed == IMPORT_JOB_GZIP));
+
+ if (j->compressed == IMPORT_JOB_XZ) {
+ lzma_ret xzr;
+
+ xzr = lzma_stream_decoder(&j->xz, UINT64_MAX, LZMA_TELL_UNSUPPORTED_CHECK);
+ if (xzr != LZMA_OK) {
+ log_error("Failed to initialize XZ decoder.");
+ return -EIO;
+ }
+ }
+ if (j->compressed == IMPORT_JOB_GZIP) {
+ r = inflateInit2(&j->gzip, 15+16);
+ if (r != Z_OK) {
+ log_error("Failed to initialize gzip decoder.");
+ return -EIO;
+ }
+ }
+
+ r = import_job_open_disk(j);
+ if (r < 0)
+ return r;
+
+ /* Now, take the payload we read so far, and decompress it */
+ stub = j->payload;
+ stub_size = j->payload_size;
+
+ j->payload = NULL;
+ j->payload_size = 0;
+
+ j->state = IMPORT_JOB_RUNNING;
+
+ r = import_job_write_compressed(j, stub, stub_size);
+ if (r < 0)
+ return r;
+
+ return 0;
+}
+
+static size_t import_job_write_callback(void *contents, size_t size, size_t nmemb, void *userdata) {
+ ImportJob *j = userdata;
+ size_t sz = size * nmemb;
+ int r;
+
+ assert(contents);
+ assert(j);
+
+ switch (j->state) {
+
+ case IMPORT_JOB_ANALYZING:
+ /* Let's first check what it actually is */
+
+ if (!GREEDY_REALLOC(j->payload, j->payload_allocated, j->payload_size + sz)) {
+ r = log_oom();
+ goto fail;
+ }
+
+ memcpy((uint8_t*) j->payload + j->payload_size, contents, sz);
+ j->payload_size += sz;
+
+ r = import_job_detect_compression(j);
+ if (r < 0)
+ goto fail;
+
+ break;
+
+ case IMPORT_JOB_RUNNING:
+
+ r = import_job_write_compressed(j, contents, sz);
+ if (r < 0)
+ goto fail;
+
+ break;
+
+ case IMPORT_JOB_DONE:
+ case IMPORT_JOB_FAILED:
+ r = -ESTALE;
+ goto fail;
+
+ default:
+ assert_not_reached("Impossible state.");
+ }
+
+ return sz;
+
+fail:
+ import_job_finish(j, r);
+ return 0;
+}
+
+static size_t import_job_header_callback(void *contents, size_t size, size_t nmemb, void *userdata) {
+ ImportJob *j = userdata;
+ size_t sz = size * nmemb;
+ _cleanup_free_ char *length = NULL, *last_modified = NULL;
+ char *etag;
+ int r;
+
+ assert(contents);
+ assert(j);
+
+ if (j->state == IMPORT_JOB_DONE || j->state == IMPORT_JOB_FAILED) {
+ r = -ESTALE;
+ goto fail;
+ }
+
+ assert(j->state == IMPORT_JOB_ANALYZING);
+
+ r = curl_header_strdup(contents, sz, "ETag:", &etag);
+ if (r < 0) {
+ log_oom();
+ goto fail;
+ }
+ if (r > 0) {
+ free(j->etag);
+ j->etag = etag;
+
+ if (strv_contains(j->old_etags, j->etag)) {
+ log_info("Image already downloaded. Skipping download.");
+ import_job_finish(j, 0);
+ return sz;
+ }
+
+ return sz;
+ }
+
+ r = curl_header_strdup(contents, sz, "Content-Length:", &length);
+ if (r < 0) {
+ log_oom();
+ goto fail;
+ }
+ if (r > 0) {
+ (void) safe_atou64(length, &j->content_length);
+
+ if (j->content_length != (uint64_t) -1) {
+ char bytes[FORMAT_BYTES_MAX];
+
+ if (j->content_length > j->compressed_max) {
+ log_error("Content too large.");
+ r = -EFBIG;
+ goto fail;
+ }
+
+ log_info("Downloading %s.", format_bytes(bytes, sizeof(bytes), j->content_length));
+ }
+
+ return sz;
+ }
+
+ r = curl_header_strdup(contents, sz, "Last-Modified:", &last_modified);
+ if (r < 0) {
+ log_oom();
+ goto fail;
+ }
+ if (r > 0) {
+ (void) curl_parse_http_time(last_modified, &j->mtime);
+ return sz;
+ }
+
+ return sz;
+
+fail:
+ import_job_finish(j, r);
+ return 0;
+}
+
+static int import_job_progress_callback(void *userdata, curl_off_t dltotal, curl_off_t dlnow, curl_off_t ultotal, curl_off_t ulnow) {
+ ImportJob *j = userdata;
+ unsigned percent;
+ usec_t n;
+
+ assert(j);
+
+ if (dltotal <= 0)
+ return 0;
+
+ percent = ((100 * dlnow) / dltotal);
+ n = now(CLOCK_MONOTONIC);
+
+ if (n > j->last_status_usec + USEC_PER_SEC &&
+ percent != j->progress_percent) {
+ char buf[FORMAT_TIMESPAN_MAX];
+
+ if (n - j->start_usec > USEC_PER_SEC && dlnow > 0) {
+ usec_t left, done;
+
+ done = n - j->start_usec;
+ left = (usec_t) (((double) done * (double) dltotal) / dlnow) - done;
+
+ log_info("Got %u%% of %s. %s left.", percent, j->url, format_timespan(buf, sizeof(buf), left, USEC_PER_SEC));
+ } else
+ log_info("Got %u%% of %s.", percent, j->url);
+
+ j->progress_percent = percent;
+ j->last_status_usec = n;
+ }
+
+ return 0;
+}
+
+int import_job_new(ImportJob **ret, const char *url, CurlGlue *glue, void *userdata) {
+ _cleanup_(import_job_unrefp) ImportJob *j = NULL;
+
+ assert(url);
+ assert(glue);
+ assert(ret);
+
+ j = new0(ImportJob, 1);
+ if (!j)
+ return -ENOMEM;
+
+ j->state = IMPORT_JOB_INIT;
+ j->disk_fd = -1;
+ j->userdata = userdata;
+ j->glue = glue;
+ j->content_length = (uint64_t) -1;
+ j->start_usec = now(CLOCK_MONOTONIC);
+ j->compressed_max = j->uncompressed_max = 8LLU * 1024LLU * 1024LLU * 1024LLU; /* 8GB */
+
+ j->url = strdup(url);
+ if (!j->url)
+ return -ENOMEM;
+
+ *ret = j;
+ j = NULL;
+
+ return 0;
+}
+
+int import_job_begin(ImportJob *j) {
+ int r;
+
+ assert(j);
+
+ if (j->state != IMPORT_JOB_INIT)
+ return -EBUSY;
+
+ r = curl_glue_make(&j->curl, j->url, j);
+ if (r < 0)
+ return r;
+
+ if (!strv_isempty(j->old_etags)) {
+ _cleanup_free_ char *cc = NULL, *hdr = NULL;
+
+ cc = strv_join(j->old_etags, ", ");
+ if (!cc)
+ return -ENOMEM;
+
+ hdr = strappend("If-None-Match: ", cc);
+ if (!hdr)
+ return -ENOMEM;
+
+ j->request_header = curl_slist_new(hdr, NULL);
+ if (!j->request_header)
+ return -ENOMEM;
+
+ if (curl_easy_setopt(j->curl, CURLOPT_HTTPHEADER, j->request_header) != CURLE_OK)
+ return -EIO;
+ }
+
+ if (curl_easy_setopt(j->curl, CURLOPT_WRITEFUNCTION, import_job_write_callback) != CURLE_OK)
+ return -EIO;
+
+ if (curl_easy_setopt(j->curl, CURLOPT_WRITEDATA, j) != CURLE_OK)
+ return -EIO;
+
+ if (curl_easy_setopt(j->curl, CURLOPT_HEADERFUNCTION, import_job_header_callback) != CURLE_OK)
+ return -EIO;
+
+ if (curl_easy_setopt(j->curl, CURLOPT_HEADERDATA, j) != CURLE_OK)
+ return -EIO;
+
+ if (curl_easy_setopt(j->curl, CURLOPT_XFERINFOFUNCTION, import_job_progress_callback) != CURLE_OK)
+ return -EIO;
+
+ if (curl_easy_setopt(j->curl, CURLOPT_XFERINFODATA, j) != CURLE_OK)
+ return -EIO;
+
+ if (curl_easy_setopt(j->curl, CURLOPT_NOPROGRESS, 0) != CURLE_OK)
+ return -EIO;
+
+ r = curl_glue_add(j->glue, j->curl);
+ if (r < 0)
+ return r;
+
+ j->state = IMPORT_JOB_ANALYZING;
+
+ return 0;
+}