chiark / gitweb /
import: add support for pulling raw tar balls as containers
authorLennart Poettering <lennart@poettering.net>
Tue, 20 Jan 2015 00:36:11 +0000 (01:36 +0100)
committerLennart Poettering <lennart@poettering.net>
Tue, 20 Jan 2015 14:06:58 +0000 (15:06 +0100)
Ubuntu provides their cloud images optionally as tarball, hence also
support downloading those.

Makefile.am
src/import/import-dkr.c
src/import/import-job.c [new file with mode: 0644]
src/import/import-job.h [new file with mode: 0644]
src/import/import-raw.c
src/import/import-tar.c [new file with mode: 0644]
src/import/import-tar.h [new file with mode: 0644]
src/import/import-util.c [new file with mode: 0644]
src/import/import-util.h [new file with mode: 0644]
src/import/import.c

index 788e634..f165042 100644 (file)
@@ -5246,8 +5246,14 @@ systemd_import_SOURCES = \
        src/import/import.c \
        src/import/import-raw.c \
        src/import/import-raw.h \
+       src/import/import-tar.c \
+       src/import/import-tar.h \
        src/import/import-dkr.c \
        src/import/import-dkr.h \
+       src/import/import-job.c \
+       src/import/import-job.h \
+       src/import/import-util.c \
+       src/import/import-util.h \
        src/import/curl-util.c \
        src/import/curl-util.h \
        src/import/aufs-util.c \
index b54a1a6..8dfd270 100644 (file)
@@ -422,7 +422,7 @@ static void dkr_import_name_maybe_finish(DkrImportName *name) {
                         return;
                 }
 
-                log_info("Created new local image %s.", p);
+                log_info("Created new local image '%s'.", name->local);
         }
 
         dkr_import_finish(name->import, 0);
diff --git a/src/import/import-job.c b/src/import/import-job.c
new file mode 100644 (file)
index 0000000..5a4ea69
--- /dev/null
@@ -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;
+}
diff --git a/src/import/import-job.h b/src/import/import-job.h
new file mode 100644 (file)
index 0000000..843daa2
--- /dev/null
@@ -0,0 +1,101 @@
+/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
+
+#pragma once
+
+/***
+  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 <lzma.h>
+#include <zlib.h>
+
+#include "macro.h"
+#include "curl-util.h"
+
+typedef struct ImportJob ImportJob;
+
+typedef void (*ImportJobFinished)(ImportJob *job);
+typedef int (*ImportJobOpenDisk)(ImportJob *job);
+
+typedef enum ImportJobState {
+        IMPORT_JOB_INIT,
+        IMPORT_JOB_ANALYZING, /* Still reading into ->payload, to figure out what we have */
+        IMPORT_JOB_RUNNING,  /* Writing to destination */
+        IMPORT_JOB_DONE,
+        IMPORT_JOB_FAILED,
+        _IMPORT_JOB_STATE_MAX,
+        _IMPORT_JOB_STATE_INVALID = -1,
+} ImportJobState;
+
+typedef enum ImportJobCompression {
+        IMPORT_JOB_UNCOMPRESSED,
+        IMPORT_JOB_XZ,
+        IMPORT_JOB_GZIP,
+        _IMPORT_JOB_COMPRESSION_MAX,
+        _IMPORT_JOB_COMPRESSION_INVALID = -1,
+} ImportJobCompression;
+
+struct ImportJob {
+        ImportJobState state;
+        int error;
+
+        char *url;
+
+        void *userdata;
+        ImportJobFinished on_finished;
+        ImportJobOpenDisk on_open_disk;
+
+        CurlGlue *glue;
+        CURL *curl;
+        struct curl_slist *request_header;
+
+        char *etag;
+        char **old_etags;
+
+        uint64_t content_length;
+        uint64_t written_compressed;
+        uint64_t written_uncompressed;
+
+        uint64_t uncompressed_max;
+        uint64_t compressed_max;
+
+        uint8_t *payload;
+        size_t payload_size;
+        size_t payload_allocated;
+
+        int disk_fd;
+
+        usec_t mtime;
+
+        ImportJobCompression compressed;
+        lzma_stream xz;
+        z_stream gzip;
+
+        unsigned progress_percent;
+        usec_t start_usec;
+        usec_t last_status_usec;
+
+        bool allow_sparse;
+};
+
+int import_job_new(ImportJob **job, const char *url, CurlGlue *glue, void *userdata);
+ImportJob* import_job_unref(ImportJob *job);
+
+int import_job_begin(ImportJob *j);
+
+void import_job_curl_on_finished(CurlGlue *g, CURL *curl, CURLcode result);
index 486157a..1fe27b6 100644 (file)
@@ -30,6 +30,7 @@
 #include "qcow2-util.h"
 #include "strv.h"
 #include "copy.h"
+#include "import-util.h"
 #include "import-raw.h"
 
 typedef struct RawImportFile RawImportFile;
@@ -691,17 +692,6 @@ static int raw_import_file_progress_callback(void *userdata, curl_off_t dltotal,
         return 0;
 }
 
-static bool etag_is_valid(const char *etag) {
-
-        if (!endswith(etag, "\""))
-                return false;
-
-        if (!startswith(etag, "\"") && !startswith(etag, "W/\""))
-                return false;
-
-        return true;
-}
-
 static int raw_import_file_find_old_etags(RawImportFile *f) {
         _cleanup_free_ char *escaped_url = NULL;
         _cleanup_closedir_ DIR *d = NULL;
@@ -751,7 +741,7 @@ static int raw_import_file_find_old_etags(RawImportFile *f) {
                 if (!u)
                         return -ENOMEM;
 
-                if (!etag_is_valid(u)) {
+                if (!http_etag_is_valid(u)) {
                         free(u);
                         continue;
                 }
diff --git a/src/import/import-tar.c b/src/import/import-tar.c
new file mode 100644 (file)
index 0000000..43227f6
--- /dev/null
@@ -0,0 +1,296 @@
+/*-*- 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/prctl.h>
+#include <curl/curl.h>
+
+#include "hashmap.h"
+#include "utf8.h"
+#include "strv.h"
+#include "copy.h"
+#include "btrfs-util.h"
+#include "util.h"
+#include "macro.h"
+#include "mkdir.h"
+#include "curl-util.h"
+#include "import-job.h"
+#include "import-util.h"
+#include "import-tar.h"
+
+struct TarImport {
+        sd_event *event;
+        CurlGlue *glue;
+
+        char *image_root;
+
+        ImportJob *tar_job;
+
+        TarImportFinished on_finished;
+        void *userdata;
+
+        bool finished;
+
+        char *local;
+        bool force_local;
+
+        pid_t tar_pid;
+
+        char *temp_path;
+        char *final_path;
+};
+
+TarImport* tar_import_unref(TarImport *i) {
+        if (!i)
+                return NULL;
+
+        if (i->tar_pid > 0) {
+                kill(i->tar_pid, SIGKILL);
+                wait_for_terminate(i->tar_pid, NULL);
+        }
+
+        import_job_unref(i->tar_job);
+
+        curl_glue_unref(i->glue);
+        sd_event_unref(i->event);
+
+        if (i->temp_path) {
+                (void) btrfs_subvol_remove(i->temp_path);
+                (void) rm_rf_dangerous(i->temp_path, false, true, false);
+        }
+
+        free(i->final_path);
+        free(i->image_root);
+        free(i->local);
+
+        free(i);
+
+        return NULL;
+}
+
+int tar_import_new(TarImport **ret, sd_event *event, const char *image_root, TarImportFinished on_finished, void *userdata) {
+        _cleanup_(tar_import_unrefp) TarImport *i = NULL;
+        int r;
+
+        assert(ret);
+        assert(event);
+
+        i = new0(TarImport, 1);
+        if (!i)
+                return -ENOMEM;
+
+        i->on_finished = on_finished;
+        i->userdata = userdata;
+
+        i->image_root = strdup(image_root ?: "/var/lib/machines");
+        if (!i->image_root)
+                return -ENOMEM;
+
+        if (event)
+                i->event = sd_event_ref(event);
+        else {
+                r = sd_event_default(&i->event);
+                if (r < 0)
+                        return r;
+        }
+
+        r = curl_glue_new(&i->glue, i->event);
+        if (r < 0)
+                return r;
+
+        i->glue->on_finished = import_job_curl_on_finished;
+        i->glue->userdata = i;
+
+        *ret = i;
+        i = NULL;
+
+        return 0;
+}
+
+static void tar_import_job_on_finished(ImportJob *j) {
+        TarImport *i;
+        int r;
+
+        assert(j);
+        assert(j->userdata);
+
+        i = j->userdata;
+
+        if (j->error != 0) {
+                r = j->error;
+                goto finish;
+        }
+
+        /* This is invoked if either the download completed
+         * successfully, or the download was skipped because we
+         * already have the etag. */
+
+        j->disk_fd = safe_close(j->disk_fd);
+
+        if (i->tar_pid > 0) {
+                r = wait_for_terminate_and_warn("tar", i->tar_pid, true);
+                i->tar_pid = 0;
+                if (r < 0)
+                        goto finish;
+        }
+
+        if (i->temp_path) {
+                r = import_make_read_only(i->temp_path);
+                if (r < 0)
+                        goto finish;
+
+                if (rename(i->temp_path, i->final_path) < 0) {
+                        r = log_error_errno(errno, "Failed to rename to final image name: %m");
+                        goto finish;
+                }
+        }
+
+        if (i->local) {
+                if (!i->final_path) {
+                        r = import_make_path(j->url, j->etag, i->image_root, ".tar-", NULL, &i->final_path);
+                        if (r < 0)
+                                goto finish;
+                }
+
+                r = import_make_local_copy(i->final_path, i->image_root, i->local, i->force_local);
+                if (r < 0)
+                        goto finish;
+        }
+
+        r = 0;
+
+finish:
+        i->finished = true;
+
+        if (i->on_finished)
+                i->on_finished(i, r, i->userdata);
+        else
+                sd_event_exit(i->event, r);
+}
+
+static int tar_import_job_on_open_disk(ImportJob *j) {
+        _cleanup_close_pair_ int pipefd[2] = { -1 , -1 };
+        TarImport *i;
+        int r;
+
+        assert(j);
+        assert(j->userdata);
+
+        i = j->userdata;
+
+        r = import_make_path(j->url, j->etag, i->image_root, ".tar-", NULL, &i->final_path);
+        if (r < 0)
+                return log_oom();
+
+        r = tempfn_random(i->final_path, &i->temp_path);
+        if (r < 0)
+                return log_oom();
+
+        mkdir_parents_label(i->temp_path, 0700);
+
+        r = btrfs_subvol_make(i->temp_path);
+        if (r == -ENOTTY) {
+                if (mkdir(i->temp_path, 0755) < 0)
+                        return log_error_errno(errno, "Failed to create directory %s: %m", i->temp_path);
+        } else if (r < 0)
+                return log_error_errno(errno, "Failed to create subvolume %s: %m", i->temp_path);
+
+        if (pipe2(pipefd, O_CLOEXEC) < 0)
+                return log_error_errno(errno, "Failed to create pipe for tar: %m");
+
+        i->tar_pid = fork();
+        if (i->tar_pid < 0)
+                return log_error_errno(errno, "Failed to fork off tar: %m");
+        if (i->tar_pid == 0) {
+                int null_fd;
+
+                reset_all_signal_handlers();
+                reset_signal_mask();
+                assert_se(prctl(PR_SET_PDEATHSIG, SIGTERM) == 0);
+
+                pipefd[1] = safe_close(pipefd[1]);
+
+                if (dup2(pipefd[0], STDIN_FILENO) != STDIN_FILENO) {
+                        log_error_errno(errno, "Failed to dup2() fd: %m");
+                        _exit(EXIT_FAILURE);
+                }
+
+                if (pipefd[0] != STDIN_FILENO)
+                        safe_close(pipefd[0]);
+
+                null_fd = open("/dev/null", O_WRONLY|O_NOCTTY);
+                if (null_fd < 0) {
+                        log_error_errno(errno, "Failed to open /dev/null: %m");
+                        _exit(EXIT_FAILURE);
+                }
+
+                if (dup2(null_fd, STDOUT_FILENO) != STDOUT_FILENO) {
+                        log_error_errno(errno, "Failed to dup2() fd: %m");
+                        _exit(EXIT_FAILURE);
+                }
+
+                if (null_fd != STDOUT_FILENO)
+                        safe_close(null_fd);
+
+                execlp("tar", "tar", "--numeric-owner", "-C", i->temp_path, "-px", NULL);
+                _exit(EXIT_FAILURE);
+        }
+
+        pipefd[0] = safe_close(pipefd[0]);
+
+        j->disk_fd = pipefd[1];
+        pipefd[1] = -1;
+
+        return 0;
+}
+
+int tar_import_pull(TarImport *i, const char *url, const char *local, bool force_local) {
+        int r;
+
+        assert(i);
+
+        if (i->tar_job)
+                return -EBUSY;
+
+        if (!http_url_is_valid(url))
+                return -EINVAL;
+
+        if (local && !machine_name_is_valid(local))
+                return -EINVAL;
+
+        r = free_and_strdup(&i->local, local);
+        if (r < 0)
+                return r;
+
+        i->force_local = force_local;
+
+        r = import_job_new(&i->tar_job, url, i->glue, i);
+        if (r < 0)
+                return r;
+
+        i->tar_job->on_finished = tar_import_job_on_finished;
+        i->tar_job->on_open_disk = tar_import_job_on_open_disk;
+
+        r = import_find_old_etags(url, i->image_root, DT_DIR, ".tar-", NULL, &i->tar_job->old_etags);
+        if (r < 0)
+                return r;
+
+        return import_job_begin(i->tar_job);
+}
diff --git a/src/import/import-tar.h b/src/import/import-tar.h
new file mode 100644 (file)
index 0000000..6a7477f
--- /dev/null
@@ -0,0 +1,36 @@
+/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
+
+#pragma once
+
+/***
+  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 "sd-event.h"
+#include "macro.h"
+
+typedef struct TarImport TarImport;
+
+typedef void (*TarImportFinished)(TarImport *import, int error, void *userdata);
+
+int tar_import_new(TarImport **import, sd_event *event, const char *image_root, TarImportFinished on_finished, void *userdata);
+TarImport* tar_import_unref(TarImport *import);
+
+DEFINE_TRIVIAL_CLEANUP_FUNC(TarImport*, tar_import_unref);
+
+int tar_import_pull(TarImport *import, const char *rul, const char *local, bool force_local);
diff --git a/src/import/import-util.c b/src/import/import-util.c
new file mode 100644 (file)
index 0000000..24c8ac1
--- /dev/null
@@ -0,0 +1,201 @@
+/*-*- 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 "util.h"
+#include "strv.h"
+#include "copy.h"
+#include "btrfs-util.h"
+#include "import-util.h"
+
+#define FILENAME_ESCAPE "/.#\"\'"
+
+bool http_etag_is_valid(const char *etag) {
+        if (!endswith(etag, "\""))
+                return false;
+
+        if (!startswith(etag, "\"") && !startswith(etag, "W/\""))
+                return false;
+
+        return true;
+}
+
+int import_find_old_etags(const char *url, const char *image_root, int dt, const char *prefix, const char *suffix, char ***etags) {
+        _cleanup_free_ char *escaped_url = NULL;
+        _cleanup_closedir_ DIR *d = NULL;
+        _cleanup_strv_free_ char **l = NULL;
+        struct dirent *de;
+        int r;
+
+        assert(url);
+        assert(etags);
+
+        if (!image_root)
+                image_root = "/var/lib/machines";
+
+        escaped_url = xescape(url, FILENAME_ESCAPE);
+        if (!escaped_url)
+                return -ENOMEM;
+
+        d = opendir(image_root);
+        if (!d) {
+                if (errno == ENOENT) {
+                        *etags = NULL;
+                        return 0;
+                }
+
+                return -errno;
+        }
+
+        FOREACH_DIRENT_ALL(de, d, return -errno) {
+                const char *a, *b;
+                char *u;
+
+                if (de->d_type != DT_UNKNOWN &&
+                    de->d_type != dt)
+                        continue;
+
+                if (prefix) {
+                        a = startswith(de->d_name, prefix);
+                        if (!a)
+                                continue;
+                } else
+                        a = de->d_name;
+
+                a = startswith(a, escaped_url);
+                if (!a)
+                        continue;
+
+                a = startswith(a, ".");
+                if (!a)
+                        continue;
+
+                if (suffix) {
+                        b = endswith(de->d_name, suffix);
+                        if (!b)
+                                continue;
+                } else
+                        b = strchr(de->d_name, 0);
+
+                if (a >= b)
+                        continue;
+
+                u = cunescape_length(a, b - a);
+                if (!u)
+                        return -ENOMEM;
+
+                if (!http_etag_is_valid(u)) {
+                        free(u);
+                        continue;
+                }
+
+                r = strv_consume(&l, u);
+                if (r < 0)
+                        return r;
+        }
+
+        *etags = l;
+        l = NULL;
+
+        return 0;
+}
+
+int import_make_local_copy(const char *final, const char *image_root, const char *local, bool force_local) {
+        const char *p;
+        int r;
+
+        assert(final);
+        assert(local);
+
+        if (!image_root)
+                image_root = "/var/lib/machines";
+
+        p = strappenda(image_root, "/", local);
+
+        if (force_local) {
+                (void) btrfs_subvol_remove(p);
+                (void) rm_rf_dangerous(p, false, true, false);
+        }
+
+        r = btrfs_subvol_snapshot(final, p, false, false);
+        if (r == -ENOTTY) {
+                r = copy_tree(final, p, false);
+                if (r < 0)
+                        return log_error_errno(r, "Failed to copy image: %m");
+        } else if (r < 0)
+                return log_error_errno(r, "Failed to create local image: %m");
+
+        log_info("Created new local image '%s'.", local);
+
+        return 0;
+}
+
+int import_make_read_only(const char *path) {
+        int r;
+
+        r = btrfs_subvol_set_read_only(path, true);
+        if (r == -ENOTTY) {
+                struct stat st;
+
+                r = stat(path, &st);
+                if (r < 0)
+                        return log_error_errno(errno, "Failed to stat temporary image: %m");
+
+                if (chmod(path, st.st_mode & 0755) < 0)
+                        return log_error_errno(errno, "Failed to chmod() final image: %m");
+
+                return 0;
+        }
+        if (r < 0)
+                return log_error_errno(r, "Failed to mark final image read-only: %m");
+
+        return 0;
+}
+
+int import_make_path(const char *url, const char *etag, const char *image_root, const char *prefix, const char *suffix, char **ret) {
+        _cleanup_free_ char *escaped_url = NULL;
+        char *path;
+
+        assert(url);
+        assert(ret);
+
+        if (!image_root)
+                image_root = "/var/lib/machines";
+
+        escaped_url = xescape(url, FILENAME_ESCAPE);
+        if (!escaped_url)
+                return -ENOMEM;
+
+        if (etag) {
+                _cleanup_free_ char *escaped_etag = NULL;
+
+                escaped_etag = xescape(etag, FILENAME_ESCAPE);
+                if (!escaped_etag)
+                        return -ENOMEM;
+
+                path = strjoin(image_root, "/", strempty(prefix), escaped_url, ".", escaped_etag, strempty(suffix), NULL);
+        } else
+                path = strjoin(image_root, "/", strempty(prefix), escaped_url, strempty(suffix), NULL);
+        if (!path)
+                return -ENOMEM;
+
+        *ret = path;
+        return 0;
+}
diff --git a/src/import/import-util.h b/src/import/import-util.h
new file mode 100644 (file)
index 0000000..ad5ab50
--- /dev/null
@@ -0,0 +1,31 @@
+/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
+
+#pragma once
+
+/***
+  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 <stdbool.h>
+
+bool http_etag_is_valid(const char *etag);
+
+int import_make_local_copy(const char *final, const char *root, const char *local, bool force_local);
+int import_find_old_etags(const char *url, const char *root, int dt, const char *prefix, const char *suffix, char ***etags);
+int import_make_read_only(const char *path);
+int import_make_path(const char *url, const char *etag, const char *image_root, const char *prefix, const char *suffix, char **ret);
index 9b10de5..c0fc224 100644 (file)
@@ -25,6 +25,8 @@
 #include "event-util.h"
 #include "verbs.h"
 #include "build.h"
+#include "machine-image.h"
+#include "import-tar.h"
 #include "import-raw.h"
 #include "import-dkr.h"
 
@@ -33,6 +35,144 @@ static const char *arg_image_root = "/var/lib/machines";
 
 static const char* arg_dkr_index_url = DEFAULT_DKR_INDEX_URL;
 
+static void on_tar_finished(TarImport *import, int error, void *userdata) {
+        sd_event *event = userdata;
+        assert(import);
+
+        if (error == 0)
+                log_info("Operation completed successfully.");
+        else
+                log_error_errno(error, "Operation failed: %m");
+
+        sd_event_exit(event, error);
+}
+
+static int url_final_component(const char *url, char **ret) {
+        const char *e, *p;
+        char *s;
+
+        e = strchrnul(url, '?');
+
+        while (e > url && e[-1] == '/')
+                        e--;
+
+        p = e;
+        while (p > url && p[-1] != '/')
+                p--;
+
+        if (e <= p)
+                return -EINVAL;
+
+        s = strndup(p, e - p);
+        if (!s)
+                return -ENOMEM;
+
+        *ret = s;
+        return 0;
+}
+
+static int strip_tar_suffixes(const char *name, char **ret) {
+        const char *e;
+        char *s;
+
+        e = endswith(name, ".tar");
+        if (!e)
+                e = endswith(name, ".tar.gz");
+        if (!e)
+                e = endswith(name, ".tar.xz");
+        if (!e)
+                e = endswith(name, ".tgz");
+        if (!e)
+                e = strchr(name, 0);
+
+        if (e <= name)
+                return -EINVAL;
+
+        s = strndup(name, e - name);
+        if (!s)
+                return -ENOMEM;
+
+        *ret = s;
+        return 0;
+}
+
+static int pull_tar(int argc, char *argv[], void *userdata) {
+        _cleanup_(tar_import_unrefp) TarImport *import = NULL;
+        _cleanup_event_unref_ sd_event *event = NULL;
+        const char *url, *local;
+        _cleanup_free_ char *l = NULL, *ll = NULL;
+        int r;
+
+        url = argv[1];
+        if (!http_url_is_valid(url)) {
+                log_error("URL '%s' is not valid.", url);
+                return -EINVAL;
+        }
+
+        if (argc >= 3)
+                local = argv[2];
+        else {
+                r = url_final_component(url, &l);
+                if (r < 0)
+                        return log_error_errno(r, "Failed get final component of URL: %m");
+
+                local = l;
+        }
+
+        if (isempty(local) || streq(local, "-"))
+                local = NULL;
+
+        if (local) {
+                r = strip_tar_suffixes(local, &ll);
+                if (r < 0)
+                        return log_oom();
+
+                local = ll;
+
+                if (!machine_name_is_valid(local)) {
+                        log_error("Local image name '%s' is not valid.", local);
+                        return -EINVAL;
+                }
+
+                if (!arg_force) {
+                        r = image_find(local, NULL);
+                        if (r < 0)
+                                return log_error_errno(r, "Failed to check whether image '%s' exists: %m", local);
+                        else if (r > 0) {
+                                log_error_errno(EEXIST, "Image '%s' already exists.", local);
+                                return -EEXIST;
+                        }
+                }
+
+                log_info("Pulling '%s', saving as '%s'.", url, local);
+        } else
+                log_info("Pulling '%s'.", url);
+
+        r = sd_event_default(&event);
+        if (r < 0)
+                return log_error_errno(r, "Failed to allocate event loop: %m");
+
+        assert_se(sigprocmask_many(SIG_BLOCK, SIGTERM, SIGINT, -1) == 0);
+        sd_event_add_signal(event, NULL, SIGTERM, NULL,  NULL);
+        sd_event_add_signal(event, NULL, SIGINT, NULL, NULL);
+
+        r = tar_import_new(&import, event, arg_image_root, on_tar_finished, event);
+        if (r < 0)
+                return log_error_errno(r, "Failed to allocate importer: %m");
+
+        r = tar_import_pull(import, url, local, arg_force);
+        if (r < 0)
+                return log_error_errno(r, "Failed to pull image: %m");
+
+        r = sd_event_loop(event);
+        if (r < 0)
+                return log_error_errno(r, "Failed to run event loop: %m");
+
+        log_info("Exiting.");
+
+        return 0;
+}
+
 static void on_raw_finished(RawImport *import, int error, void *userdata) {
         sd_event *event = userdata;
         assert(import);
@@ -40,7 +180,7 @@ static void on_raw_finished(RawImport *import, int error, void *userdata) {
         if (error == 0)
                 log_info("Operation completed successfully.");
         else
-                log_info_errno(error, "Operation failed: %m");
+                log_error_errno(error, "Operation failed: %m");
 
         sd_event_exit(event, error);
 }
@@ -173,7 +313,7 @@ static void on_dkr_finished(DkrImport *import, int error, void *userdata) {
         if (error == 0)
                 log_info("Operation completed successfully.");
         else
-                log_info_errno(error, "Operation failed: %m");
+                log_error_errno(error, "Operation failed: %m");
 
         sd_event_exit(event, error);
 }
@@ -277,8 +417,9 @@ static int help(int argc, char *argv[], void *userdata) {
                "     --image-root=            Image root directory\n"
                "     --dkr-index-url=URL      Specify index URL to use for downloads\n\n"
                "Commands:\n"
-               "  pull-dkr REMOTE [NAME]      Download a DKR image\n"
-               "  pull-raw URL [NAME]         Download a RAW image\n",
+               "  pull-tar URL                Download a TAR image\n"
+               "  pull-raw URL [NAME]         Download a RAW image\n"
+               "  pull-dkr REMOTE [NAME]      Download a DKR image\n",
                program_invocation_short_name);
 
         return 0;
@@ -350,8 +491,9 @@ static int import_main(int argc, char *argv[]) {
 
         static const Verb verbs[] = {
                 { "help",     VERB_ANY, VERB_ANY, 0, help     },
-                { "pull-dkr", 2,        3,        0, pull_dkr },
+                { "pull-tar", 2,        3,        0, pull_tar },
                 { "pull-raw", 2,        3,        0, pull_raw },
+                { "pull-dkr", 2,        3,        0, pull_dkr },
                 {}
         };