X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fjournal%2Fjournal-file.c;h=7beedb4a254cc4edebd8025d5c6fd8489515ab4c;hp=83f603ac821b0af1ac521a6b8b992bd754ffe362;hb=f65425cbc4385b4f5b20ded641cbb07ad6a0938f;hpb=ecf68b1d3833c2ee3156a813ad42cf155dc4ace4 diff --git a/src/journal/journal-file.c b/src/journal/journal-file.c index 83f603ac8..7beedb4a2 100644 --- a/src/journal/journal-file.c +++ b/src/journal/journal-file.c @@ -31,11 +31,10 @@ #include "journal-file.h" #include "lookup3.h" #include "compress.h" +#include "fsprg.h" -#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*16ULL) -#define DEFAULT_FIELD_HASH_TABLE_SIZE (2047ULL*16ULL) - -#define DEFAULT_WINDOW_SIZE (8ULL*1024ULL*1024ULL) +#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*sizeof(HashItem)) +#define DEFAULT_FIELD_HASH_TABLE_SIZE (333ULL*sizeof(HashItem)) #define COMPRESSION_SIZE_THRESHOLD (512ULL) @@ -58,35 +57,59 @@ * size */ #define DEFAULT_KEEP_FREE (1024ULL*1024ULL) /* 1 MB */ -static const char signature[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' }; +/* n_data was the first entry we added after the initial file format design */ +#define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data)) #define ALIGN64(x) (((x) + 7ULL) & ~7ULL) -void journal_file_close(JournalFile *f) { - int t; +#define JOURNAL_HEADER_CONTAINS(h, field) \ + (le64toh((h)->header_size) >= offsetof(Header, field) + sizeof((h)->field)) +static int journal_file_maybe_append_tag(JournalFile *f, uint64_t realtime); +static int journal_file_hmac_put_object(JournalFile *f, int type, uint64_t p); + +void journal_file_close(JournalFile *f) { assert(f); + /* Write the final tag */ + if (f->authenticate) + journal_file_append_tag(f); + + /* Sync everything to disk, before we mark the file offline */ + if (f->mmap && f->fd >= 0) + mmap_cache_close_fd(f->mmap, f->fd); + + if (f->writable && f->fd >= 0) + fdatasync(f->fd); + if (f->header) { - if (f->writable) + /* Mark the file offline. Don't override the archived state if it already is set */ + if (f->writable && f->header->state == STATE_ONLINE) f->header->state = STATE_OFFLINE; munmap(f->header, PAGE_ALIGN(sizeof(Header))); } - for (t = 0; t < _WINDOW_MAX; t++) - if (f->windows[t].ptr) - munmap(f->windows[t].ptr, f->windows[t].size); - if (f->fd >= 0) close_nointr_nofail(f->fd); free(f->path); + if (f->mmap) + mmap_cache_unref(f->mmap); + #ifdef HAVE_XZ free(f->compress_buffer); #endif +#ifdef HAVE_GCRYPT + if (f->fsprg_header) + munmap(f->fsprg_header, PAGE_ALIGN(f->fsprg_size)); + + if (f->hmac) + gcry_md_close(f->hmac); +#endif + free(f); } @@ -98,16 +121,22 @@ static int journal_file_init_header(JournalFile *f, JournalFile *template) { assert(f); zero(h); - memcpy(h.signature, signature, 8); + memcpy(h.signature, HEADER_SIGNATURE, 8); h.header_size = htole64(ALIGN64(sizeof(h))); + h.incompatible_flags = + htole32(f->compress ? HEADER_INCOMPATIBLE_COMPRESSED : 0); + + h.compatible_flags = + htole32(f->authenticate ? HEADER_COMPATIBLE_AUTHENTICATED : 0); + r = sd_id128_randomize(&h.file_id); if (r < 0) return r; if (template) { h.seqnum_id = template->header->seqnum_id; - h.seqnum = template->header->seqnum; + h.tail_entry_seqnum = template->header->tail_entry_seqnum; } else h.seqnum_id = h.file_id; @@ -142,7 +171,9 @@ static int journal_file_refresh_header(JournalFile *f) { f->header->state = STATE_ONLINE; - __sync_synchronize(); + /* Sync the online state to disk */ + msync(f->header, PAGE_ALIGN(sizeof(Header)), MS_SYNC); + fdatasync(f->fd); return 0; } @@ -150,18 +181,37 @@ static int journal_file_refresh_header(JournalFile *f) { static int journal_file_verify_header(JournalFile *f) { assert(f); - if (memcmp(f->header, signature, 8)) + 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 */ #ifdef HAVE_XZ - if ((le64toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_COMPRESSED) != 0) + if ((le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_COMPRESSED) != 0) return -EPROTONOSUPPORT; #else if (f->header->incompatible_flags != 0) return -EPROTONOSUPPORT; #endif - if (f->header->header_size != htole64(ALIGN64(sizeof(*(f->header))))) + /* When open for writing we refuse to open files with + * compatible flags, too */ + if (f->writable) { +#ifdef HAVE_GCRYPT + if ((le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_AUTHENTICATED) != 0) + return -EPROTONOSUPPORT; +#else + if (f->header->compatible_flags != 0) + return -EPROTONOSUPPORT; +#endif + } + + /* 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 ((le32toh(f->header->compatible_flags) & HEADER_COMPATIBLE_AUTHENTICATED) && + !JOURNAL_HEADER_CONTAINS(f->header, n_tags)) return -EBADMSG; if ((uint64_t) f->last_stat.st_size < (le64toh(f->header->header_size) + le64toh(f->header->arena_size))) @@ -181,15 +231,20 @@ static int journal_file_verify_header(JournalFile *f) { state = f->header->state; - if (state == STATE_ONLINE) - log_debug("Journal file %s is already online. Assuming unclean closing. Ignoring.", f->path); - /* FIXME: immediately rotate */ - else if (state == STATE_ARCHIVED) + 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 %u. Ignoring.", f->path, state); + else if (state != STATE_OFFLINE) { + log_debug("Journal file %s has unknown state %u.", f->path, state); + return -EBUSY; + } } + f->compress = !!(le32toh(f->header->incompatible_flags) & HEADER_INCOMPATIBLE_COMPRESSED); + f->authenticate = !!(le32toh(f->header->compatible_flags) & HEADER_COMPATIBLE_AUTHENTICATED); + return 0; } @@ -244,6 +299,8 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) if (r != 0) return -r; + mmap_cache_close_fd_range(f->mmap, f->fd, old_size); + if (fstat(f->fd, &f->last_stat) < 0) return -errno; @@ -252,59 +309,11 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) return 0; } -static int journal_file_map( - JournalFile *f, - uint64_t offset, - uint64_t size, - void **_window, - uint64_t *_woffset, - uint64_t *_wsize, - void **ret) { - - uint64_t woffset, wsize; - void *window; - +static int journal_file_move_to(JournalFile *f, int context, uint64_t offset, uint64_t size, void **ret) { assert(f); - assert(size > 0); assert(ret); - woffset = offset & ~((uint64_t) page_size() - 1ULL); - wsize = size + (offset - woffset); - wsize = PAGE_ALIGN(wsize); - /* Avoid SIGBUS on invalid accesses */ - if (woffset + wsize > (uint64_t) PAGE_ALIGN(f->last_stat.st_size)) - return -EADDRNOTAVAIL; - - window = mmap(NULL, wsize, f->prot, MAP_SHARED, f->fd, woffset); - if (window == MAP_FAILED) - return -errno; - - if (_window) - *_window = window; - - if (_woffset) - *_woffset = woffset; - - if (_wsize) - *_wsize = wsize; - - *ret = (uint8_t*) window + (offset - woffset); - - return 0; -} - -static int journal_file_move_to(JournalFile *f, int wt, uint64_t offset, uint64_t size, void **ret) { - void *p = NULL; - uint64_t delta; - int r; - Window *w; - - assert(f); - assert(ret); - assert(wt >= 0); - assert(wt < _WINDOW_MAX); - 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. */ @@ -314,57 +323,7 @@ static int journal_file_move_to(JournalFile *f, int wt, uint64_t offset, uint64_ return -EADDRNOTAVAIL; } - w = f->windows + wt; - - if (_likely_(w->ptr && - w->offset <= offset && - w->offset + w->size >= offset + size)) { - - *ret = (uint8_t*) w->ptr + (offset - w->offset); - return 0; - } - - if (w->ptr) { - if (munmap(w->ptr, w->size) < 0) - return -errno; - - w->ptr = NULL; - w->size = w->offset = 0; - } - - if (size < DEFAULT_WINDOW_SIZE) { - /* If the default window size is larger then what was - * asked for extend the mapping a bit in the hope to - * minimize needed remappings later on. We add half - * the window space before and half behind the - * requested mapping */ - - delta = (DEFAULT_WINDOW_SIZE - size) / 2; - - if (delta > offset) - delta = offset; - - offset -= delta; - size = DEFAULT_WINDOW_SIZE; - } else - delta = 0; - - if (offset + size > (uint64_t) f->last_stat.st_size) - size = (uint64_t) f->last_stat.st_size - offset; - - if (size <= 0) - return -EADDRNOTAVAIL; - - r = journal_file_map(f, - offset, size, - &w->ptr, &w->offset, &w->size, - &p); - - if (r < 0) - return r; - - *ret = (uint8_t*) p + delta; - return 0; + return mmap_cache_get(f->mmap, f->fd, f->prot, context, offset, size, ret); } static bool verify_hash(Object *o) { @@ -384,17 +343,38 @@ static bool verify_hash(Object *o) { return h1 == h2; } +static uint64_t minimum_header_size(Object *o) { + + static 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, int type, uint64_t offset, Object **ret) { int r; void *t; Object *o; uint64_t s; + unsigned context; assert(f); assert(ret); - assert(type < _OBJECT_TYPE_MAX); - r = journal_file_move_to(f, type >= 0 ? type : WINDOW_UNKNOWN, offset, sizeof(ObjectHeader), &t); + /* One context for each type, plus one catch-all for the rest */ + context = type > 0 && type < _OBJECT_TYPE_MAX ? type : 0; + + r = journal_file_move_to(f, context, offset, sizeof(ObjectHeader), &t); if (r < 0) return r; @@ -404,6 +384,12 @@ int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Objec if (s < sizeof(ObjectHeader)) return -EBADMSG; + if (o->object.type <= OBJECT_UNUSED) + return -EBADMSG; + + if (s < minimum_header_size(o)) + return -EBADMSG; + if (type >= 0 && o->object.type != type) return -EBADMSG; @@ -422,12 +408,12 @@ int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Objec return 0; } -static uint64_t journal_file_seqnum(JournalFile *f, uint64_t *seqnum) { +static uint64_t journal_file_entry_seqnum(JournalFile *f, uint64_t *seqnum) { uint64_t r; assert(f); - r = le64toh(f->header->seqnum) + 1; + r = le64toh(f->header->tail_entry_seqnum) + 1; if (seqnum) { /* If an external seqnum counter was passed, we update @@ -440,10 +426,10 @@ static uint64_t journal_file_seqnum(JournalFile *f, uint64_t *seqnum) { *seqnum = r; } - f->header->seqnum = htole64(r); + f->header->tail_entry_seqnum = htole64(r); - if (f->header->first_seqnum == 0) - f->header->first_seqnum = htole64(r); + if (f->header->head_entry_seqnum == 0) + f->header->head_entry_seqnum = htole64(r); return r; } @@ -455,6 +441,7 @@ static int journal_file_append_object(JournalFile *f, int type, uint64_t size, O void *t; assert(f); + assert(type > 0 && type < _OBJECT_TYPE_MAX); assert(size >= sizeof(ObjectHeader)); assert(offset); assert(ret); @@ -500,7 +487,17 @@ static int journal_file_setup_data_hash_table(JournalFile *f) { assert(f); - s = DEFAULT_DATA_HASH_TABLE_SIZE; + /* We estimate that we need 1 hash table entry per 768 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_info("Reserving %llu entries in hash table.", (unsigned long long) (s / sizeof(HashItem))); + r = journal_file_append_object(f, OBJECT_DATA_HASH_TABLE, offsetof(Object, hash_table.items) + s, @@ -550,7 +547,7 @@ static int journal_file_map_data_hash_table(JournalFile *f) { s = le64toh(f->header->data_hash_table_size); r = journal_file_move_to(f, - WINDOW_DATA_HASH_TABLE, + OBJECT_DATA_HASH_TABLE, p, s, &t); if (r < 0) @@ -571,7 +568,7 @@ static int journal_file_map_field_hash_table(JournalFile *f) { s = le64toh(f->header->field_hash_table_size); r = journal_file_move_to(f, - WINDOW_FIELD_HASH_TABLE, + OBJECT_FIELD_HASH_TABLE, p, s, &t); if (r < 0) @@ -614,6 +611,9 @@ static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, ui 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; } @@ -758,20 +758,22 @@ static int journal_file_append_data( o->object.size = htole64(offsetof(Object, data.payload) + rsize); o->object.flags |= OBJECT_COMPRESSED; - f->header->incompatible_flags = htole32(le32toh(f->header->incompatible_flags) | HEADER_INCOMPATIBLE_COMPRESSED); - log_debug("Compressed data object %lu -> %lu", (unsigned long) size, (unsigned long) rsize); } } #endif - if (!compressed) + if (!compressed && size > 0) memcpy(o->data.payload, data, size); r = journal_file_link_data(f, o, p, hash); if (r < 0) return r; + r = journal_file_hmac_put_object(f, OBJECT_DATA, p); + 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_DATA, p, &o); @@ -848,6 +850,10 @@ static int link_entry_into_array(JournalFile *f, if (r < 0) return r; + r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, q); + if (r < 0) + return r; + o->entry_array.items[i] = htole64(p); if (ap == 0) @@ -978,13 +984,17 @@ static int journal_file_append_entry_internal( if (r < 0) return r; - o->entry.seqnum = htole64(journal_file_seqnum(f, seqnum)); + o->entry.seqnum = htole64(journal_file_entry_seqnum(f, seqnum)); memcpy(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; + r = journal_file_hmac_put_object(f, OBJECT_ENTRY, np); + if (r < 0) + return r; + r = journal_file_link_entry(f, o, np); if (r < 0) return r; @@ -1034,7 +1044,12 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st ts->monotonic < le64toh(f->header->tail_entry_monotonic)) return -EINVAL; - items = alloca(sizeof(EntryItem) * n_iovec); + r = journal_file_maybe_append_tag(f, ts->realtime); + if (r < 0) + return r; + + /* alloca() can't take 0, hence let's allocate at least one */ + items = alloca(sizeof(EntryItem) * MAX(1, n_iovec)); for (i = 0; i < n_iovec; i++) { uint64_t p; @@ -1212,8 +1227,15 @@ static int generic_array_bisect(JournalFile *f, } } - if (k > n) + if (k > n) { + if (direction == DIRECTION_UP) { + i = n; + subtract_one = true; + goto found; + } + return 0; + } last_p = lp; @@ -1246,7 +1268,7 @@ found: *offset = p; if (idx) - *idx = t + i - (subtract_one ? 1 : 0); + *idx = t + i + (subtract_one ? -1 : 0); return 1; } @@ -1263,6 +1285,8 @@ static int generic_array_bisect_plus_one(JournalFile *f, uint64_t *idx) { int r; + bool step_back = false; + Object *o; assert(f); assert(test_object); @@ -1279,33 +1303,77 @@ static int generic_array_bisect_plus_one(JournalFile *f, if (r == TEST_FOUND) r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; - if (r == TEST_RIGHT) { - Object *o; - - 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; + /* 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; - return 1; + 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; } +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; +} + +int journal_file_move_to_entry_by_offset( + JournalFile *f, + uint64_t p, + direction_t direction, + Object **ret, + uint64_t *offset) { + + return generic_array_bisect(f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + p, + test_object_offset, + direction, + ret, offset, NULL); +} + + static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) { Object *o; int r; @@ -1407,12 +1475,13 @@ int journal_file_move_to_entry_by_monotonic( Object *o; int r; - sd_id128_to_string(boot_id, t + 9); + assert(f); + sd_id128_to_string(boot_id, t + 9); r = journal_file_find_data_object(f, t, strlen(t), &o, NULL); if (r < 0) return r; - else if (r == 0) + if (r == 0) return -ENOENT; return generic_array_bisect_plus_one(f, @@ -1425,18 +1494,6 @@ int journal_file_move_to_entry_by_monotonic( ret, offset, NULL); } -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; -} - int journal_file_next_entry( JournalFile *f, Object *o, uint64_t p, @@ -1601,6 +1658,119 @@ int journal_file_next_entry_for_data( ret, 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 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) { + + char t[9+32+1] = "_BOOT_ID="; + Object *o, *d; + int r; + uint64_t b, z; + + assert(f); + + /* First, seek by time */ + sd_id128_to_string(boot_id, t + 9); + r = journal_file_find_data_object(f, t, strlen(t), &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; + } + + return 0; +} + int journal_file_move_to_entry_by_seqnum_for_data( JournalFile *f, uint64_t data_offset, @@ -1611,8 +1781,10 @@ int journal_file_move_to_entry_by_seqnum_for_data( Object *d; int r; + assert(f); + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); - if (r <= 0) + if (r < 0) return r; return generic_array_bisect_plus_one(f, @@ -1635,8 +1807,10 @@ int journal_file_move_to_entry_by_realtime_for_data( Object *d; int r; + assert(f); + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); - if (r <= 0) + if (r < 0) return r; return generic_array_bisect_plus_one(f, @@ -1649,28 +1823,947 @@ int journal_file_move_to_entry_by_realtime_for_data( ret, offset, NULL); } -void journal_file_dump(JournalFile *f) { - char a[33], b[33], c[33]; - Object *o; +static void *fsprg_state(JournalFile *f) { + uint64_t a, b; + assert(f); + + if (!f->authenticate) + return NULL; + + a = le64toh(f->fsprg_header->header_size); + b = le64toh(f->fsprg_header->state_size); + + if (a + b > f->fsprg_size) + return NULL; + + return (uint8_t*) f->fsprg_header + a; +} + +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->authenticate) + return 0; + + if (!f->hmac_running) + return 0; + + log_debug("Writing tag for epoch %llu\n", (unsigned long long) FSPRG_GetEpoch(fsprg_state(f))); + + 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)); + + /* 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, 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; +} + +static 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->authenticate) + return 0; + + if (f->hmac_running) + return 0; + + /* Prepare HMAC for next cycle */ + gcry_md_reset(f->hmac); + FSPRG_GetKey(fsprg_state(f), 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->authenticate); + + if (le64toh(f->fsprg_header->fsprg_start_usec) == 0 || + le64toh(f->fsprg_header->fsprg_interval_usec) == 0) + return -ENOTSUP; + + if (realtime < le64toh(f->fsprg_header->fsprg_start_usec)) + return -ESTALE; + + t = realtime - le64toh(f->fsprg_header->fsprg_start_usec); + t = t / le64toh(f->fsprg_header->fsprg_interval_usec); + + *epoch = t; + return 0; +} + +static int journal_file_need_evolve(JournalFile *f, uint64_t realtime) { + uint64_t goal, epoch; + int r; + assert(f); + + if (!f->authenticate) + return 0; + + r = journal_file_get_epoch(f, realtime, &goal); + if (r < 0) + return r; + + epoch = FSPRG_GetEpoch(fsprg_state(f)); + if (epoch > goal) + return -ESTALE; + + return epoch != goal; +} + +static int journal_file_evolve(JournalFile *f, uint64_t realtime) { + uint64_t goal, epoch; + int r; + + assert(f); + + if (!f->authenticate) + return 0; + + r = journal_file_get_epoch(f, realtime, &goal); + if (r < 0) + return r; + + epoch = FSPRG_GetEpoch(fsprg_state(f)); + if (epoch < goal) + log_debug("Evolving FSPRG key from epoch %llu to %llu.", (unsigned long long) epoch, (unsigned long long) goal); + + for (;;) { + if (epoch > goal) + return -ESTALE; + if (epoch == goal) + return 0; + + FSPRG_Evolve(fsprg_state(f)); + epoch = FSPRG_GetEpoch(fsprg_state(f)); + } +} + +static int journal_file_maybe_append_tag(JournalFile *f, uint64_t realtime) { + int r; + + assert(f); + + if (!f->authenticate) + return 0; + + r = journal_file_need_evolve(f, realtime); + if (r <= 0) + return 0; + + r = journal_file_append_tag(f); + if (r < 0) + return r; + + r = journal_file_evolve(f, realtime); + if (r < 0) + return r; + + r = journal_file_hmac_start(f); + if (r < 0) + return r; + + return 0; +} + +static int journal_file_hmac_put_object(JournalFile *f, int type, uint64_t p) { + int r; + Object *o; + + assert(f); + + if (!f->authenticate) + return 0; + + r = journal_file_hmac_start(f); + if (r < 0) + return r; + + r = journal_file_move_to_object(f, type, p, &o); + if (r < 0) + return r; + + 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_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)); + break; + default: + return -EINVAL; + } + + return 0; +} + +static int journal_file_hmac_put_header(JournalFile *f) { + int r; + + assert(f); + + if (!f->authenticate) + 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_seqnum, + * head_entry_realtime, tail_entry_realtime, + * tail_entry_monotonic, n_data, n_fields, header_tag */ + + 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)); + gcry_md_write(f->hmac, &f->header->head_entry_seqnum, offsetof(Header, head_entry_realtime) - offsetof(Header, head_entry_seqnum)); + + return 0; +} + +static int journal_file_load_fsprg(JournalFile *f) { + int r, fd = -1; + char *p = NULL; + struct stat st; + FSPRGHeader *m = NULL; + sd_id128_t machine; + + assert(f); + + if (!f->authenticate) + return 0; + + r = sd_id128_get_machine(&machine); + if (r < 0) + return r; + + if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/fsprg", + SD_ID128_FORMAT_VAL(machine)) < 0) + return -ENOMEM; + + fd = open(p, O_RDWR|O_CLOEXEC|O_NOCTTY, 0600); + if (fd < 0) { + log_error("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(FSPRGHeader)) { + r = -ENODATA; + goto finish; + } + + m = mmap(NULL, PAGE_ALIGN(sizeof(FSPRGHeader)), PROT_READ, MAP_SHARED, fd, 0); + if (m == MAP_FAILED) { + m = NULL; + r = -errno; + goto finish; + } + + if (memcmp(m->signature, FSPRG_HEADER_SIGNATURE, 8) != 0) { + r = -EBADMSG; + goto finish; + } + + if (m->incompatible_flags != 0) { + r = -EPROTONOSUPPORT; + goto finish; + } + + if (le64toh(m->header_size) < sizeof(FSPRGHeader)) { + r = -EBADMSG; + goto finish; + } + + if (le64toh(m->state_size) != FSPRG_stateinbytes(m->secpar)) { + r = -EBADMSG; + goto finish; + } + + f->fsprg_size = le64toh(m->header_size) + le64toh(m->state_size); + if ((uint64_t) st.st_size < f->fsprg_size) { + r = -ENODATA; + goto finish; + } + + if (!sd_id128_equal(machine, m->machine_id)) { + r = -EHOSTDOWN; + goto finish; + } + + if (le64toh(m->fsprg_start_usec) <= 0 || + le64toh(m->fsprg_interval_usec) <= 0) { + r = -EBADMSG; + goto finish; + } + + f->fsprg_header = mmap(NULL, PAGE_ALIGN(f->fsprg_size), PROT_READ|PROT_WRITE, MAP_SHARED, fd, 0); + if (f->fsprg_header == MAP_FAILED) { + f->fsprg_header = NULL; + r = -errno; + goto finish; + } + + r = 0; + +finish: + if (m) + munmap(m, PAGE_ALIGN(sizeof(FSPRGHeader))); + + if (fd >= 0) + close_nointr_nofail(fd); + + free(p); + return r; +} + +static int journal_file_setup_hmac(JournalFile *f) { + gcry_error_t e; + + if (!f->authenticate) + return 0; + + e = gcry_md_open(&f->hmac, GCRY_MD_SHA256, GCRY_MD_FLAG_HMAC); + if (e != 0) + return -ENOTSUP; + + return 0; +} + +static int journal_file_append_first_tag(JournalFile *f) { + int r; + uint64_t p; + + if (!f->authenticate) + 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, 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, p); + if (r < 0) + return r; + + r = journal_file_append_tag(f); + if (r < 0) + return r; + + return 0; +} + +static int journal_file_object_verify(JournalFile *f, Object *o) { + assert(f); + assert(o); + + /* This does various superficial tests about the length an + * possible field values. It does not follow any references to + * other objects. */ + + switch (o->object.type) { + case OBJECT_DATA: + if (le64toh(o->data.entry_offset) <= 0 || + le64toh(o->data.n_entries) <= 0) + return -EBADMSG; + + if (le64toh(o->object.size) - offsetof(DataObject, payload) <= 0) + return -EBADMSG; + break; + + case OBJECT_FIELD: + if (le64toh(o->object.size) - offsetof(FieldObject, payload) <= 0) + return -EBADMSG; + break; + + case OBJECT_ENTRY: + if ((le64toh(o->object.size) - offsetof(EntryObject, items)) % sizeof(EntryItem) != 0) + return -EBADMSG; + + if ((le64toh(o->object.size) - offsetof(EntryObject, items)) / sizeof(EntryItem) <= 0) + return -EBADMSG; + + if (le64toh(o->entry.seqnum) <= 0 || + le64toh(o->entry.realtime) <= 0) + return -EBADMSG; + + break; + + case OBJECT_DATA_HASH_TABLE: + case OBJECT_FIELD_HASH_TABLE: + if ((le64toh(o->object.size) - offsetof(HashTableObject, items)) % sizeof(HashItem) != 0) + return -EBADMSG; + + break; + + case OBJECT_ENTRY_ARRAY: + if ((le64toh(o->object.size) - offsetof(EntryArrayObject, items)) % sizeof(le64_t) != 0) + return -EBADMSG; + + break; + + case OBJECT_TAG: + if (le64toh(o->object.size) != sizeof(TagObject)) + return -EBADMSG; + break; + } + + return 0; +} + +static void draw_progress(uint64_t p, usec_t *last_usec) { + unsigned n, i, j, k; + usec_t z, x; + + if (!isatty(STDOUT_FILENO)) + 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\x1B[?25l", stdout); + + for (i = 0; i < j; i++) + fputs("\xe2\x96\x88", stdout); + + for (i = 0; i < k; i++) + fputs("\xe2\x96\x91", stdout); + + printf(" %3lu%%", 100LU * (unsigned long) p / 65535LU); + + fputs("\r\x1B[?25h", stdout); + fflush(stdout); +} + +static void flush_progress(void) { + unsigned n, i; + + if (!isatty(STDOUT_FILENO)) + return; + + n = (3 * columns()) / 4; + + putchar('\r'); + + for (i = 0; i < n + 5; i++) + putchar(' '); + + putchar('\r'); + fflush(stdout); +} + +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, 0, c * sizeof(uint64_t), sizeof(uint64_t), (void **) &z); + if (r < 0) + return r; + + if (*z == p) + return 1; + + if (p < *z) + b = c; + else + a = c; + } + + return 0; +} + +int journal_file_verify(JournalFile *f, const char *key) { + int r; + Object *o; + uint64_t p = 0; + uint64_t tag_seqnum = 0, 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; + usec_t last_usec = 0; + int data_fd = -1, entry_fd = -1, entry_array_fd = -1; + char data_path[] = "/var/tmp/journal-data-XXXXXX", + entry_path[] = "/var/tmp/journal-entry-XXXXXX", + entry_array_path[] = "/var/tmp/journal-entry-array-XXXXXX"; + + assert(f); + + data_fd = mkostemp(data_path, O_CLOEXEC); + if (data_fd < 0) { + log_error("Failed to create data file: %m"); + goto fail; + } + unlink(data_path); + + entry_fd = mkostemp(entry_path, O_CLOEXEC); + if (entry_fd < 0) { + log_error("Failed to create entry file: %m"); + goto fail; + } + unlink(entry_path); + + entry_array_fd = mkostemp(entry_array_path, O_CLOEXEC); + if (entry_array_fd < 0) { + log_error("Failed to create entry array file: %m"); + goto fail; + } + unlink(entry_array_path); + + /* First iteration: we go through all objects, verify the + * superficial structure, headers, hashes. */ + + r = journal_file_hmac_put_header(f); + if (r < 0) { + log_error("Failed to calculate HMAC of header."); + goto fail; + } + + p = le64toh(f->header->header_size); + while (p != 0) { + draw_progress((0x7FFF * p) / le64toh(f->header->tail_object_offset), &last_usec); + + r = journal_file_move_to_object(f, -1, p, &o); + if (r < 0) { + log_error("Invalid object at %llu", (unsigned long long) p); + goto fail; + } + + if (le64toh(f->header->tail_object_offset) < p) { + log_error("Invalid tail object pointer."); + r = -EBADMSG; + goto fail; + } + + n_objects ++; + + r = journal_file_object_verify(f, o); + if (r < 0) { + log_error("Invalid object contents at %llu", (unsigned long long) p); + goto fail; + } + + r = journal_file_hmac_put_object(f, -1, p); + if (r < 0) { + log_error("Failed to calculate HMAC at %llu", (unsigned long long) p); + goto fail; + } + + if (o->object.flags & OBJECT_COMPRESSED && + !(le32toh(f->header->incompatible_flags) & HEADER_INCOMPATIBLE_COMPRESSED)) { + log_error("Compressed object without compression at %llu", (unsigned long long) p); + r = -EBADMSG; + goto fail; + } + + if (o->object.flags & OBJECT_COMPRESSED && + o->object.type != OBJECT_DATA) { + log_error("Compressed non-data object at %llu", (unsigned long long) p); + r = -EBADMSG; + goto fail; + } + + if (o->object.type == OBJECT_TAG) { + + if (!(le32toh(f->header->compatible_flags) & HEADER_COMPATIBLE_AUTHENTICATED)) { + log_error("Tag object without authentication at %llu", (unsigned long long) p); + r = -EBADMSG; + goto fail; + } + + if (le64toh(o->tag.seqnum) != tag_seqnum) { + log_error("Tag sequence number out of synchronization at %llu", (unsigned long long) p); + r = -EBADMSG; + goto fail; + } + + } else if (o->object.type == OBJECT_ENTRY) { + + r = write_uint64(entry_fd, p); + if (r < 0) + goto fail; + + if (!entry_seqnum_set && + le64toh(o->entry.seqnum) != le64toh(f->header->head_entry_seqnum)) { + log_error("Head entry sequence number incorrect"); + r = -EBADMSG; + goto fail; + } + + if (entry_seqnum_set && + entry_seqnum >= le64toh(o->entry.seqnum)) { + log_error("Entry sequence number out of synchronization at %llu", (unsigned long long) p); + 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)) { + log_error("Entry timestamp out of synchronization at %llu", (unsigned long long) p); + 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)) { + log_error("Head entry realtime timestamp incorrect"); + r = -EBADMSG; + goto fail; + } + + entry_realtime = le64toh(o->entry.realtime); + entry_realtime_set = true; + + n_entries ++; + } else if (o->object.type == 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) { + log_error("More than one main entry array at %llu", (unsigned long long) p); + r = -EBADMSG; + goto fail; + } + + found_main_entry_array = true; + } + + n_entry_arrays++; + + } else if (o->object.type == OBJECT_DATA) { + + r = write_uint64(data_fd, p); + if (r < 0) + goto fail; + + n_data++; + + } else if (o->object.type == OBJECT_FIELD) + n_fields++; + else if (o->object.type == OBJECT_DATA_HASH_TABLE) { + n_data_hash_tables++; + + if (n_data_hash_tables > 1) { + log_error("More than one data hash table at %llu", (unsigned long long) p); + 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)) { + log_error("Header fields for data hash table invalid."); + r = -EBADMSG; + goto fail; + } + } else if (o->object.type == OBJECT_FIELD_HASH_TABLE) { + n_field_hash_tables++; + + if (n_field_hash_tables > 1) { + log_error("More than one field hash table at %llu", (unsigned long long) p); + 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)) { + log_error("Header fields for field hash table invalid."); + r = -EBADMSG; + goto fail; + } + } else if (o->object.type >= _OBJECT_TYPE_MAX) + n_weird ++; + + if (p == le64toh(f->header->tail_object_offset)) + p = 0; + else + p = p + ALIGN64(le64toh(o->object.size)); + } + + if (n_objects != le64toh(f->header->n_objects)) { + log_error("Object number mismatch"); + r = -EBADMSG; + goto fail; + } + + if (n_entries != le64toh(f->header->n_entries)) { + log_error("Entry number mismatch"); + r = -EBADMSG; + goto fail; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_data) && + n_data != le64toh(f->header->n_data)) { + log_error("Data number mismatch"); + r = -EBADMSG; + goto fail; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_fields) && + n_fields != le64toh(f->header->n_fields)) { + log_error("Field number mismatch"); + r = -EBADMSG; + goto fail; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_tags) && + tag_seqnum != le64toh(f->header->n_tags)) { + log_error("Tag number mismatch"); + r = -EBADMSG; + goto fail; + } + + if (n_data_hash_tables != 1) { + log_error("Missing data hash table"); + r = -EBADMSG; + goto fail; + } + + if (n_field_hash_tables != 1) { + log_error("Missing field hash table"); + r = -EBADMSG; + goto fail; + } + + if (!found_main_entry_array) { + log_error("Missing entry array"); + r = -EBADMSG; + goto fail; + } + + if (entry_seqnum_set && + entry_seqnum != le64toh(f->header->tail_entry_seqnum)) { + log_error("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))) { + log_error("Invalid tail monotonic timestamp"); + r = -EBADMSG; + goto fail; + } + + if (entry_realtime_set && entry_realtime != le64toh(f->header->tail_entry_realtime)) { + log_error("Invalid tail realtime timestamp"); + r = -EBADMSG; + goto fail; + } + + /* Second iteration: we go through all objects again, this + * time verify all pointers. */ + + p = le64toh(f->header->header_size); + while (p != 0) { + draw_progress(0x8000 + (0x7FFF * p) / le64toh(f->header->tail_object_offset), &last_usec); + + r = journal_file_move_to_object(f, -1, p, &o); + if (r < 0) { + log_error("Invalid object at %llu", (unsigned long long) p); + goto fail; + } + + if (o->object.type == OBJECT_ENTRY_ARRAY) { + uint64_t i = 0, n; + + if (le64toh(o->entry_array.next_entry_array_offset) != 0 && + !contains_uint64(f->mmap, entry_array_fd, n_entry_arrays, le64toh(o->entry_array.next_entry_array_offset))) { + log_error("Entry array chains up to invalid next array at %llu", (unsigned long long) p); + r = -EBADMSG; + goto fail; + } + + n = journal_file_entry_array_n_items(o); + for (i = 0; i < n; i++) { + if (le64toh(o->entry_array.items[i]) != 0 && + !contains_uint64(f->mmap, entry_fd, n_entries, le64toh(o->entry_array.items[i]))) { + + log_error("Entry array points to invalid next array at %llu", (unsigned long long) p); + r = -EBADMSG; + goto fail; + } + } + + } + + r = journal_file_move_to_object(f, -1, p, &o); + if (r < 0) { + log_error("Invalid object at %llu", (unsigned long long) p); + goto fail; + } + + if (p == le64toh(f->header->tail_object_offset)) + p = 0; + else + p = p + ALIGN64(le64toh(o->object.size)); + } + + 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); + + close_nointr_nofail(data_fd); + close_nointr_nofail(entry_fd); + close_nointr_nofail(entry_array_fd); + + return 0; + +fail: + flush_progress(); + + log_error("File corruption detected at %s:%llu (of %llu, %llu%%).", + f->path, + (unsigned long long) p, + (unsigned long long) f->last_stat.st_size, + (unsigned long long) (100 * p / f->last_stat.st_size)); + + if (data_fd >= 0) { + mmap_cache_close_fd(f->mmap, data_fd); + close_nointr_nofail(data_fd); + } + + if (entry_fd >= 0) { + mmap_cache_close_fd(f->mmap, entry_fd); + close_nointr_nofail(entry_fd); + } + + if (entry_array_fd >= 0) { + mmap_cache_close_fd(f->mmap, entry_array_fd); + close_nointr_nofail(entry_array_fd); + } + + return r; +} + +void journal_file_dump(JournalFile *f) { + Object *o; int r; uint64_t p; assert(f); - printf("File Path: %s\n" - "File ID: %s\n" - "Machine ID: %s\n" - "Boot ID: %s\n" - "Arena size: %llu\n" - "Objects: %lu\n" - "Entries: %lu\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), - (unsigned long long) le64toh(f->header->arena_size), - (unsigned long) le64toh(f->header->n_objects), - (unsigned long) le64toh(f->header->n_entries)); + journal_file_print_header(f); p = le64toh(f->header->header_size); while (p != 0) { @@ -1707,8 +2800,9 @@ void journal_file_dump(JournalFile *f) { printf("Type: OBJECT_ENTRY_ARRAY\n"); break; - case OBJECT_SIGNATURE: - printf("Type: OBJECT_SIGNATURE\n"); + case OBJECT_TAG: + printf("Type: OBJECT_TAG %llu\n", + (unsigned long long) le64toh(o->tag.seqnum)); break; } @@ -1726,10 +2820,76 @@ fail: log_error("File corrupt"); } +void journal_file_print_header(JournalFile *f) { + char a[33], b[33], c[33]; + char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX]; + + assert(f); + + 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\n" + "Header size: %llu\n" + "Arena size: %llu\n" + "Data Hash Table Size: %llu\n" + "Field Hash Table Size: %llu\n" + "Objects: %llu\n" + "Entry Objects: %llu\n" + "Rotate Suggested: %s\n" + "Head Sequential Number: %llu\n" + "Tail Sequential Number: %llu\n" + "Head Realtime Timestamp: %s\n" + "Tail Realtime Timestamp: %s\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, c), + f->header->state == STATE_OFFLINE ? "offline" : + f->header->state == STATE_ONLINE ? "online" : + f->header->state == STATE_ARCHIVED ? "archived" : "unknown", + (f->header->compatible_flags & HEADER_COMPATIBLE_AUTHENTICATED) ? " AUTHENTICATED" : "", + (f->header->compatible_flags & ~HEADER_COMPATIBLE_AUTHENTICATED) ? " ???" : "", + (f->header->incompatible_flags & HEADER_INCOMPATIBLE_COMPRESSED) ? " COMPRESSED" : "", + (f->header->incompatible_flags & ~HEADER_INCOMPATIBLE_COMPRESSED) ? " ???" : "", + (unsigned long long) le64toh(f->header->header_size), + (unsigned long long) le64toh(f->header->arena_size), + (unsigned long long) le64toh(f->header->data_hash_table_size) / sizeof(HashItem), + (unsigned long long) le64toh(f->header->field_hash_table_size) / sizeof(HashItem), + (unsigned long long) le64toh(f->header->n_objects), + (unsigned long long) le64toh(f->header->n_entries), + yes_no(journal_file_rotate_suggested(f)), + (unsigned long long) le64toh(f->header->head_entry_seqnum), + (unsigned long long) le64toh(f->header->tail_entry_seqnum), + format_timestamp(x, sizeof(x), le64toh(f->header->head_entry_realtime)), + format_timestamp(y, sizeof(y), le64toh(f->header->tail_entry_realtime))); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) + printf("Data Objects: %llu\n" + "Data Hash Table Fill: %.1f%%\n", + (unsigned long long) 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: %llu\n" + "Field Hash Table Fill: %.1f%%\n", + (unsigned long long) le64toh(f->header->n_fields), + 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)))); +} + int journal_file_open( const char *fname, int flags, mode_t mode, + bool compress, + bool authenticate, + JournalMetrics *metrics, + MMapCache *mmap_cache, JournalFile *template, JournalFile **ret) { @@ -1751,14 +2911,25 @@ int journal_file_open( return -ENOMEM; f->fd = -1; - f->flags = flags; f->mode = mode; - f->writable = (flags & O_ACCMODE) != O_RDONLY; + + f->flags = flags; f->prot = prot_from_flags(flags); + f->writable = (flags & O_ACCMODE) != O_RDONLY; + f->compress = compress; + f->authenticate = authenticate; - if (template) { - f->metrics = template->metrics; - f->compress = template->compress; + if (mmap_cache) + f->mmap = mmap_cache_ref(mmap_cache); + else { + /* One context for each type, plus the zeroth catchall + * context. One fd for the file plus one for each type + * (which we need during verification */ + f->mmap = mmap_cache_new(_OBJECT_TYPE_MAX, 1 + _OBJECT_TYPE_MAX); + if (!f->mmap) { + r = -ENOMEM; + goto fail; + } } f->path = strdup(fname); @@ -1781,6 +2952,12 @@ int journal_file_open( if (f->last_stat.st_size == 0 && f->writable) { newly_created = true; + /* Try to load the FSPRG state, and if we can't, then + * just don't do authentication */ + r = journal_file_load_fsprg(f); + if (r < 0) + f->authenticate = false; + r = journal_file_init_header(f, template); if (r < 0) goto fail; @@ -1791,7 +2968,7 @@ int journal_file_open( } } - if (f->last_stat.st_size < (off_t) sizeof(Header)) { + if (f->last_stat.st_size < (off_t) HEADER_SIZE_MIN) { r = -EIO; goto fail; } @@ -1809,14 +2986,29 @@ int journal_file_open( goto fail; } + if (!newly_created && f->writable) { + r = journal_file_load_fsprg(f); + if (r < 0) + goto fail; + } + 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; + + r = journal_file_setup_hmac(f); + if (r < 0) + goto fail; } if (newly_created) { - r = journal_file_setup_field_hash_table(f); if (r < 0) goto fail; @@ -1824,6 +3016,10 @@ int journal_file_open( r = journal_file_setup_data_hash_table(f); if (r < 0) goto fail; + + r = journal_file_append_first_tag(f); + if (r < 0) + goto fail; } r = journal_file_map_field_hash_table(f); @@ -1845,7 +3041,7 @@ fail: return r; } -int journal_file_rotate(JournalFile **f) { +int journal_file_rotate(JournalFile **f, bool compress, bool authenticate) { char *p; size_t l; JournalFile *old_file, *new_file = NULL; @@ -1873,7 +3069,7 @@ int journal_file_rotate(JournalFile **f) { sd_id128_to_string(old_file->header->seqnum_id, p + l - 8 + 1); snprintf(p + l - 8 + 1 + 32, 1 + 16 + 1 + 16 + 8 + 1, "-%016llx-%016llx.journal", - (unsigned long long) le64toh((*f)->header->seqnum), + (unsigned long long) le64toh((*f)->header->tail_entry_seqnum), (unsigned long long) le64toh((*f)->header->tail_entry_realtime)); r = rename(old_file->path, p); @@ -1884,7 +3080,7 @@ int journal_file_rotate(JournalFile **f) { old_file->header->state = STATE_ARCHIVED; - r = journal_file_open(old_file->path, old_file->flags, old_file->mode, old_file, &new_file); + r = journal_file_open(old_file->path, old_file->flags, old_file->mode, compress, authenticate, NULL, old_file->mmap, old_file, &new_file); journal_file_close(old_file); *f = new_file; @@ -1895,6 +3091,10 @@ int journal_file_open_reliably( const char *fname, int flags, mode_t mode, + bool compress, + bool authenticate, + JournalMetrics *metrics, + MMapCache *mmap, JournalFile *template, JournalFile **ret) { @@ -1902,11 +3102,13 @@ int journal_file_open_reliably( size_t l; char *p; - r = journal_file_open(fname, flags, mode, template, ret); + r = journal_file_open(fname, flags, mode, compress, authenticate, metrics, mmap, template, ret); if (r != -EBADMSG && /* corrupted */ r != -ENODATA && /* truncated */ r != -EHOSTDOWN && /* other machine */ - r != -EPROTONOSUPPORT) /* incompatible feature */ + r != -EPROTONOSUPPORT && /* incompatible feature */ + r != -EBUSY && /* unclean shutdown */ + r != -ESHUTDOWN /* already archived */) return r; if ((flags & O_ACCMODE) == O_RDONLY) @@ -1915,6 +3117,9 @@ int journal_file_open_reliably( 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); @@ -1929,9 +3134,9 @@ int journal_file_open_reliably( if (r < 0) return -errno; - log_warning("File %s corrupted, renaming and replacing.", fname); + log_warning("File %s corrupted or uncleanly shut down, renaming and replacing.", fname); - return journal_file_open(fname, flags, mode, template, ret); + return journal_file_open(fname, flags, mode, compress, authenticate, metrics, mmap, template, ret); } struct vacuum_info { @@ -2098,7 +3303,8 @@ int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t m n_list ++; } - qsort(list, n_list, sizeof(struct vacuum_info), vacuum_compare); + if (n_list > 0) + qsort(list, n_list, sizeof(struct vacuum_info), vacuum_compare); for(i = 0; i < n_list; i++) { struct statvfs ss; @@ -2153,9 +3359,6 @@ int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint6 ts.monotonic < le64toh(to->header->tail_entry_monotonic)) return -EINVAL; - if (ts.realtime < le64toh(to->header->tail_entry_realtime)) - return -EINVAL; - n = journal_file_entry_n_items(o); items = alloca(sizeof(EntryItem) * n); @@ -2290,26 +3493,21 @@ void journal_default_metrics(JournalMetrics *m, int fd) { } int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *to) { - Object *o; - int r; - assert(f); assert(from || to); if (from) { - r = journal_file_next_entry(f, NULL, 0, DIRECTION_DOWN, &o, NULL); - if (r <= 0) - return r; + if (f->header->head_entry_realtime == 0) + return -ENOENT; - *from = le64toh(o->entry.realtime); + *from = le64toh(f->header->head_entry_realtime); } if (to) { - r = journal_file_next_entry(f, NULL, 0, DIRECTION_UP, &o, NULL); - if (r <= 0) - return r; + if (f->header->tail_entry_realtime == 0) + return -ENOENT; - *to = le64toh(o->entry.realtime); + *to = le64toh(f->header->tail_entry_realtime); } return 1; @@ -2359,3 +3557,44 @@ int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, u return 1; } + +bool journal_file_rotate_suggested(JournalFile *f) { + assert(f); + + /* 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 (%llu of %llu items, %llu file size, %llu 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))), + (unsigned long long) le64toh(f->header->n_data), + (unsigned long long) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)), + (unsigned long long) (f->last_stat.st_size), + (unsigned long long) (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 (%llu of %llu items), suggesting rotation.", + f->path, + 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))), + (unsigned long long) le64toh(f->header->n_fields), + (unsigned long long) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))); + return true; + } + + return false; +}