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=934c043aff61ee4a1308bcb82f53d77d7e3ddbe8;hb=f65425cbc4385b4f5b20ded641cbb07ad6a0938f;hpb=0ac38b707212e9aa40e25d65ffbae648cc9116f5 diff --git a/src/journal/journal-file.c b/src/journal/journal-file.c index 934c043af..7beedb4a2 100644 --- a/src/journal/journal-file.c +++ b/src/journal/journal-file.c @@ -6,16 +6,16 @@ Copyright 2011 Lennart Poettering systemd is free software; you can redistribute it and/or modify it - under the terms of the GNU General Public License as published by - the Free Software Foundation; either version 2 of the License, or + 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 - General Public License for more details. + Lesser General Public License for more details. - You should have received a copy of the GNU General Public License + You should have received a copy of the GNU Lesser General Public License along with systemd; If not, see . ***/ @@ -30,45 +30,86 @@ #include "journal-def.h" #include "journal-file.h" #include "lookup3.h" +#include "compress.h" +#include "fsprg.h" -#define DEFAULT_ARENA_MAX_SIZE (16ULL*1024ULL*1024ULL*1024ULL) -#define DEFAULT_ARENA_MIN_SIZE (256ULL*1024ULL) -#define DEFAULT_ARENA_KEEP_FREE (1ULL*1024ULL*1024ULL) +#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*sizeof(HashItem)) +#define DEFAULT_FIELD_HASH_TABLE_SIZE (333ULL*sizeof(HashItem)) -#define DEFAULT_MAX_USE (16ULL*1024ULL*1024ULL*16ULL) +#define COMPRESSION_SIZE_THRESHOLD (512ULL) -#define DEFAULT_HASH_TABLE_SIZE (2047ULL*16ULL) -#define DEFAULT_BISECT_TABLE_SIZE ((DEFAULT_ARENA_MAX_SIZE/(64ULL*1024ULL))*8ULL) +/* This is the minimum journal file size */ +#define JOURNAL_FILE_SIZE_MIN (64ULL*1024ULL) /* 64 KiB */ -#define DEFAULT_WINDOW_SIZE (128ULL*1024ULL*1024ULL) +/* These are the lower and upper bounds if we deduce the max_use value + * from the file system size */ +#define DEFAULT_MAX_USE_LOWER (1ULL*1024ULL*1024ULL) /* 1 MiB */ +#define DEFAULT_MAX_USE_UPPER (4ULL*1024ULL*1024ULL*1024ULL) /* 4 GiB */ -static const char signature[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' }; +/* This is the upper bound if we deduce max_size from max_use */ +#define DEFAULT_MAX_SIZE_UPPER (128ULL*1024ULL*1024ULL) /* 128 MiB */ + +/* This is the upper bound if we deduce the keep_free value from the + * file system size */ +#define DEFAULT_KEEP_FREE_UPPER (4ULL*1024ULL*1024ULL*1024ULL) /* 4 GiB */ + +/* This is the keep_free value when we can't determine the system + * size */ +#define DEFAULT_KEEP_FREE (1024ULL*1024ULL) /* 1 MB */ + +/* 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) +#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); - if (f->header) { - if (f->writable && f->header->state == htole32(STATE_ONLINE)) - f->header->state = htole32(STATE_OFFLINE); + /* Write the final tag */ + if (f->authenticate) + journal_file_append_tag(f); - munmap(f->header, PAGE_ALIGN(sizeof(Header))); - } + /* 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->hash_table_window) - munmap(f->hash_table_window, f->hash_table_window_size); + if (f->writable && f->fd >= 0) + fdatasync(f->fd); - if (f->bisect_table_window) - munmap(f->bisect_table_window, f->bisect_table_window_size); + if (f->header) { + /* 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; - if (f->window) - munmap(f->window, f->window_size); + munmap(f->header, PAGE_ALIGN(sizeof(Header))); + } 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); } @@ -80,11 +121,14 @@ static int journal_file_init_header(JournalFile *f, JournalFile *template) { assert(f); zero(h); - memcpy(h.signature, signature, 8); - h.arena_offset = htole64(ALIGN64(sizeof(h))); - h.arena_max_size = htole64(DEFAULT_ARENA_MAX_SIZE); - h.arena_min_size = htole64(DEFAULT_ARENA_MIN_SIZE); - h.arena_keep_free = htole64(DEFAULT_ARENA_KEEP_FREE); + 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) @@ -92,7 +136,7 @@ static int journal_file_init_header(JournalFile *f, JournalFile *template) { 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; @@ -108,6 +152,7 @@ static int journal_file_init_header(JournalFile *f, JournalFile *template) { static int journal_file_refresh_header(JournalFile *f) { int r; + sd_id128_t boot_id; assert(f); @@ -115,28 +160,65 @@ static int journal_file_refresh_header(JournalFile *f) { if (r < 0) return r; - r = sd_id128_get_boot(&f->header->boot_id); + r = sd_id128_get_boot(&boot_id); if (r < 0) return r; - f->header->state = htole32(STATE_ONLINE); + if (sd_id128_equal(boot_id, f->header->boot_id)) + f->tail_entry_monotonic_valid = true; + + f->header->boot_id = boot_id; + + f->header->state = STATE_ONLINE; + + /* Sync the online state to disk */ + msync(f->header, PAGE_ALIGN(sizeof(Header)), MS_SYNC); + fdatasync(f->fd); + return 0; } 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 ((le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_COMPRESSED) != 0) + return -EPROTONOSUPPORT; +#else if (f->header->incompatible_flags != 0) return -EPROTONOSUPPORT; +#endif + + /* 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 ((uint64_t) f->last_stat.st_size < (le64toh(f->header->arena_offset) + le64toh(f->header->arena_size))) + 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))) return -ENODATA; if (f->writable) { - uint32_t state; + uint8_t state; sd_id128_t machine_id; int r; @@ -147,44 +229,52 @@ static int journal_file_verify_header(JournalFile *f) { if (!sd_id128_equal(machine_id, f->header->machine_id)) return -EHOSTDOWN; - state = le32toh(f->header->state); + state = f->header->state; - if (state == STATE_ONLINE) - log_debug("Journal file %s is already online. Assuming unclean closing. Ignoring.", f->path); - 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; } static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) { - uint64_t asize; uint64_t old_size, new_size; + int r; assert(f); - if (offset < le64toh(f->header->arena_offset)) - return -EINVAL; - - new_size = PAGE_ALIGN(offset + size); - /* We assume that this file is not sparse, and we know that * for sure, since we always call posix_fallocate() * ourselves */ old_size = - le64toh(f->header->arena_offset) + + le64toh(f->header->header_size) + le64toh(f->header->arena_size); - if (old_size >= new_size) + new_size = PAGE_ALIGN(offset + size); + if (new_size < le64toh(f->header->header_size)) + new_size = le64toh(f->header->header_size); + + if (new_size <= old_size) return 0; - asize = new_size - le64toh(f->header->arena_offset); + if (f->metrics.max_size > 0 && + new_size > f->metrics.max_size) + return -E2BIG; - if (asize > le64toh(f->header->arena_min_size)) { + if (new_size > f->metrics.min_size && + f->metrics.keep_free > 0) { struct statvfs svfs; if (fstatvfs(f->fd, &svfs) >= 0) { @@ -192,8 +282,8 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) available = svfs.f_bfree * svfs.f_bsize; - if (available >= f->header->arena_keep_free) - available -= f->header->arena_keep_free; + if (available >= f->metrics.keep_free) + available -= f->metrics.keep_free; else available = 0; @@ -202,141 +292,89 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) } } - if (asize > le64toh(f->header->arena_max_size)) - return -E2BIG; + /* Note that the glibc fallocate() fallback is very + inefficient, hence we try to minimize the allocation area + as we can. */ + r = posix_fallocate(f->fd, old_size, new_size - old_size); + if (r != 0) + return -r; - if (posix_fallocate(f->fd, old_size, new_size - old_size) < 0) - return -errno; + mmap_cache_close_fd_range(f->mmap, f->fd, old_size); if (fstat(f->fd, &f->last_stat) < 0) return -errno; - f->header->arena_size = htole64(asize); - - 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; - - assert(f); - assert(size > 0); - assert(ret); - - woffset = offset & ~((uint64_t) page_size() - 1ULL); - wsize = size + (offset - woffset); - wsize = PAGE_ALIGN(wsize); - - 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); + f->header->arena_size = htole64(new_size - le64toh(f->header->header_size)); return 0; } -static int journal_file_move_to(JournalFile *f, uint64_t offset, uint64_t size, void **ret) { - void *p; - uint64_t delta; - int r; - +static int journal_file_move_to(JournalFile *f, int context, uint64_t offset, uint64_t size, void **ret) { assert(f); assert(ret); - if (_likely_(f->window && - f->window_offset <= offset && - f->window_offset+f->window_size >= offset + size)) { - - *ret = (uint8_t*) f->window + (offset - f->window_offset); - return 0; - } - - if (f->window) { - if (munmap(f->window, f->window_size) < 0) - return -errno; + /* Avoid SIGBUS on invalid accesses */ + if (offset + size > (uint64_t) f->last_stat.st_size) { + /* Hmm, out of range? Let's refresh the fstat() data + * first, before we trust that check. */ - f->window = NULL; - f->window_size = f->window_offset = 0; + if (fstat(f->fd, &f->last_stat) < 0 || + offset + size > (uint64_t) f->last_stat.st_size) + return -EADDRNOTAVAIL; } - 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 = PAGE_ALIGN((DEFAULT_WINDOW_SIZE - size) / 2); - - if (offset < delta) - delta = offset; - - offset -= delta; - size += (DEFAULT_WINDOW_SIZE - delta); - } else - delta = 0; - - r = journal_file_map(f, - offset, size, - &f->window, &f->window_offset, &f->window_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) { - uint64_t t; + uint64_t h1, h2; assert(o); - t = le64toh(o->object.type); - if (t == OBJECT_DATA) { - uint64_t s, h1, h2; + if (o->object.type == OBJECT_DATA && !(o->object.flags & OBJECT_COMPRESSED)) { + h1 = le64toh(o->data.hash); + h2 = hash64(o->data.payload, le64toh(o->object.size) - offsetof(Object, data.payload)); + } else if (o->object.type == OBJECT_FIELD) { + h1 = le64toh(o->field.hash); + h2 = hash64(o->field.payload, le64toh(o->object.size) - offsetof(Object, field.payload)); + } else + return true; - s = le64toh(o->object.size); + return h1 == h2; +} - h1 = le64toh(o->data.hash); - h2 = hash64(o->data.payload, s - offsetof(Object, data.payload)); +static uint64_t minimum_header_size(Object *o) { - return h1 == h2; - } + 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), + }; - return true; + 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, uint64_t offset, int type, Object **ret) { +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); - r = journal_file_move_to(f, 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; @@ -346,11 +384,17 @@ int journal_file_move_to_object(JournalFile *f, uint64_t offset, int type, Objec if (s < sizeof(ObjectHeader)) return -EBADMSG; - if (type >= 0 && le64toh(o->object.type) != type) + 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; if (s > sizeof(ObjectHeader)) { - r = journal_file_move_to(f, offset, s, &t); + r = journal_file_move_to(f, o->object.type, offset, s, &t); if (r < 0) return r; @@ -364,34 +408,49 @@ int journal_file_move_to_object(JournalFile *f, uint64_t offset, int type, Objec return 0; } -static uint64_t journal_file_seqnum(JournalFile *f) { +static uint64_t journal_file_entry_seqnum(JournalFile *f, uint64_t *seqnum) { uint64_t r; assert(f); - r = le64toh(f->header->seqnum) + 1; - f->header->seqnum = htole64(r); + r = le64toh(f->header->tail_entry_seqnum) + 1; + + if (seqnum) { + /* If an external seqnum counter was passed, we update + * both the local and the external one, and set it to + * the maximum of both */ + + if (*seqnum + 1 > r) + r = *seqnum + 1; + + *seqnum = r; + } + + f->header->tail_entry_seqnum = htole64(r); + + if (f->header->head_entry_seqnum == 0) + f->header->head_entry_seqnum = htole64(r); return r; } -static int journal_file_append_object(JournalFile *f, uint64_t size, Object **ret, uint64_t *offset) { +static int journal_file_append_object(JournalFile *f, int type, uint64_t size, Object **ret, uint64_t *offset) { int r; uint64_t p; Object *tail, *o; void *t; assert(f); + assert(type > 0 && type < _OBJECT_TYPE_MAX); assert(size >= sizeof(ObjectHeader)); assert(offset); assert(ret); p = le64toh(f->header->tail_object_offset); - if (p == 0) - p = le64toh(f->header->arena_offset); + p = le64toh(f->header->header_size); else { - r = journal_file_move_to_object(f, p, -1, &tail); + r = journal_file_move_to_object(f, -1, p, &tail); if (r < 0) return r; @@ -402,21 +461,17 @@ static int journal_file_append_object(JournalFile *f, uint64_t size, Object **re if (r < 0) return r; - r = journal_file_move_to(f, p, size, &t); + r = journal_file_move_to(f, type, p, size, &t); if (r < 0) return r; o = (Object*) t; zero(o->object); - o->object.type = htole64(OBJECT_UNUSED); - zero(o->object.reserved); + o->object.type = type; o->object.size = htole64(size); f->header->tail_object_offset = htole64(p); - if (f->header->head_object_offset == 0) - f->header->head_object_offset = htole64(p); - f->header->n_objects = htole64(le64toh(f->header->n_objects) + 1); *ret = o; @@ -425,135 +480,149 @@ static int journal_file_append_object(JournalFile *f, uint64_t size, Object **re return 0; } -static int journal_file_setup_hash_table(JournalFile *f) { +static int journal_file_setup_data_hash_table(JournalFile *f) { uint64_t s, p; Object *o; int r; assert(f); - s = DEFAULT_HASH_TABLE_SIZE; - r = journal_file_append_object(f, offsetof(Object, hash_table.table) + s, &o, &p); + /* 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, + &o, &p); if (r < 0) return r; - o->object.type = htole64(OBJECT_HASH_TABLE); - memset(o->hash_table.table, 0, s); + memset(o->hash_table.items, 0, s); - f->header->hash_table_offset = htole64(p + offsetof(Object, hash_table.table)); - f->header->hash_table_size = htole64(s); + f->header->data_hash_table_offset = htole64(p + offsetof(Object, hash_table.items)); + f->header->data_hash_table_size = htole64(s); return 0; } -static int journal_file_setup_bisect_table(JournalFile *f) { +static int journal_file_setup_field_hash_table(JournalFile *f) { uint64_t s, p; Object *o; int r; assert(f); - s = DEFAULT_BISECT_TABLE_SIZE; - r = journal_file_append_object(f, offsetof(Object, bisect_table.table) + s, &o, &p); + s = DEFAULT_FIELD_HASH_TABLE_SIZE; + r = journal_file_append_object(f, + OBJECT_FIELD_HASH_TABLE, + offsetof(Object, hash_table.items) + s, + &o, &p); if (r < 0) return r; - o->object.type = htole64(OBJECT_BISECT_TABLE); - memset(o->bisect_table.table, 0, s); + memset(o->hash_table.items, 0, s); - f->header->bisect_table_offset = htole64(p + offsetof(Object, bisect_table.table)); - f->header->bisect_table_size = htole64(s); + f->header->field_hash_table_offset = htole64(p + offsetof(Object, hash_table.items)); + f->header->field_hash_table_size = htole64(s); return 0; } -static int journal_file_map_hash_table(JournalFile *f) { +static int journal_file_map_data_hash_table(JournalFile *f) { uint64_t s, p; void *t; int r; assert(f); - p = le64toh(f->header->hash_table_offset); - s = le64toh(f->header->hash_table_size); + p = le64toh(f->header->data_hash_table_offset); + s = le64toh(f->header->data_hash_table_size); - r = journal_file_map(f, - p, s, - &f->hash_table_window, NULL, &f->hash_table_window_size, - &t); + r = journal_file_move_to(f, + OBJECT_DATA_HASH_TABLE, + p, s, + &t); if (r < 0) return r; - f->hash_table = t; + f->data_hash_table = t; return 0; } -static int journal_file_map_bisect_table(JournalFile *f) { +static int journal_file_map_field_hash_table(JournalFile *f) { uint64_t s, p; void *t; int r; assert(f); - p = le64toh(f->header->bisect_table_offset); - s = le64toh(f->header->bisect_table_size); - - r = journal_file_map(f, - p, s, - &f->bisect_table_window, NULL, &f->bisect_table_window_size, - &t); + p = le64toh(f->header->field_hash_table_offset); + s = le64toh(f->header->field_hash_table_size); + r = journal_file_move_to(f, + OBJECT_FIELD_HASH_TABLE, + p, s, + &t); if (r < 0) return r; - f->bisect_table = t; + f->field_hash_table = t; return 0; } -static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, uint64_t hash_index) { - uint64_t p; +static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, uint64_t hash) { + uint64_t p, h; int r; assert(f); assert(o); assert(offset > 0); - assert(o->object.type == htole64(OBJECT_DATA)); + assert(o->object.type == OBJECT_DATA); + + /* This might alter the window we are looking at */ - o->data.head_entry_offset = o->data.tail_entry_offset = 0; - o->data.next_hash_offset = 0; + o->data.next_hash_offset = o->data.next_field_offset = 0; + o->data.entry_offset = o->data.entry_array_offset = 0; + o->data.n_entries = 0; - p = le64toh(f->hash_table[hash_index].tail_hash_offset); + h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)); + p = le64toh(f->data_hash_table[h].tail_hash_offset); if (p == 0) { /* Only entry in the hash table is easy */ - - o->data.prev_hash_offset = 0; - f->hash_table[hash_index].head_hash_offset = htole64(offset); + f->data_hash_table[h].head_hash_offset = htole64(offset); } else { - o->data.prev_hash_offset = htole64(p); - - /* Temporarily move back to the previous data object, - * to patch in pointer */ + /* Move back to the previous data object, to patch in + * pointer */ - r = journal_file_move_to_object(f, p, OBJECT_DATA, &o); + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); if (r < 0) return r; - o->data.next_hash_offset = offset; - - r = journal_file_move_to_object(f, offset, OBJECT_DATA, &o); - if (r < 0) - return r; + o->data.next_hash_offset = htole64(offset); } - f->hash_table[hash_index].tail_hash_offset = htole64(offset); + f->data_hash_table[h].tail_hash_offset = htole64(offset); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) + f->header->n_data = htole64(le64toh(f->header->n_data) + 1); return 0; } -static int journal_file_append_data(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset) { - uint64_t hash, h, p, np; - uint64_t osize; - Object *o; +int journal_file_find_data_object_with_hash( + JournalFile *f, + const void *data, uint64_t size, uint64_t hash, + Object **ret, uint64_t *offset) { + + uint64_t p, osize, h; int r; assert(f); @@ -561,44 +630,153 @@ static int journal_file_append_data(JournalFile *f, const void *data, uint64_t s osize = offsetof(Object, data.payload) + size; - hash = hash64(data, size); - h = hash % (le64toh(f->header->hash_table_size) / sizeof(HashItem)); - p = le64toh(f->hash_table[h].head_hash_offset); + if (f->header->data_hash_table_size == 0) + return -EBADMSG; - while (p != 0) { - /* Look for this data object in the hash table */ + h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)); + p = le64toh(f->data_hash_table[h].head_hash_offset); + + while (p > 0) { + Object *o; - r = journal_file_move_to_object(f, p, OBJECT_DATA, &o); + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); if (r < 0) return r; - if (le64toh(o->object.size) == osize && - memcmp(o->data.payload, data, size) == 0) { + if (le64toh(o->data.hash) != hash) + goto next; + + if (o->object.flags & OBJECT_COMPRESSED) { +#ifdef HAVE_XZ + uint64_t l, rsize; + + l = le64toh(o->object.size); + if (l <= offsetof(Object, data.payload)) + return -EBADMSG; + + l -= offsetof(Object, data.payload); - if (le64toh(o->data.hash) != hash) + if (!uncompress_blob(o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize)) return -EBADMSG; + if (rsize == size && + memcmp(f->compress_buffer, data, size) == 0) { + + if (ret) + *ret = o; + + if (offset) + *offset = p; + + return 1; + } +#else + return -EPROTONOSUPPORT; +#endif + + } else if (le64toh(o->object.size) == osize && + memcmp(o->data.payload, data, size) == 0) { + if (ret) *ret = o; if (offset) *offset = p; - return 0; + return 1; } + next: p = le64toh(o->data.next_hash_offset); } - r = journal_file_append_object(f, osize, &o, &np); + return 0; +} + +int journal_file_find_data_object( + JournalFile *f, + const void *data, uint64_t size, + Object **ret, uint64_t *offset) { + + uint64_t hash; + + assert(f); + assert(data || size == 0); + + hash = hash64(data, size); + + return journal_file_find_data_object_with_hash(f, + data, size, hash, + ret, offset); +} + +static int journal_file_append_data( + JournalFile *f, + const void *data, uint64_t size, + Object **ret, uint64_t *offset) { + + uint64_t hash, p; + uint64_t osize; + Object *o; + int r; + bool compressed = false; + + assert(f); + assert(data || size == 0); + + hash = hash64(data, size); + + r = journal_file_find_data_object_with_hash(f, data, size, hash, &o, &p); + if (r < 0) + return r; + else if (r > 0) { + + if (ret) + *ret = o; + + if (offset) + *offset = p; + + return 0; + } + + osize = offsetof(Object, data.payload) + size; + r = journal_file_append_object(f, OBJECT_DATA, osize, &o, &p); if (r < 0) return r; - o->object.type = htole64(OBJECT_DATA); o->data.hash = htole64(hash); - memcpy(o->data.payload, data, size); - r = journal_file_link_data(f, o, np, h); +#ifdef HAVE_XZ + if (f->compress && + size >= COMPRESSION_SIZE_THRESHOLD) { + uint64_t rsize; + + compressed = compress_blob(data, size, o->data.payload, &rsize); + + if (compressed) { + o->object.size = htole64(offsetof(Object, data.payload) + rsize); + o->object.flags |= OBJECT_COMPRESSED; + + log_debug("Compressed data object %lu -> %lu", (unsigned long) size, (unsigned long) rsize); + } + } +#endif + + 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); if (r < 0) return r; @@ -606,104 +784,172 @@ static int journal_file_append_data(JournalFile *f, const void *data, uint64_t s *ret = o; if (offset) - *offset = np; + *offset = p; return 0; } uint64_t journal_file_entry_n_items(Object *o) { assert(o); - assert(o->object.type == htole64(OBJECT_ENTRY)); + assert(o->object.type == OBJECT_ENTRY); return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem); } -static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) { - uint64_t p, q; +static uint64_t journal_file_entry_array_n_items(Object *o) { + assert(o); + assert(o->object.type == OBJECT_ENTRY_ARRAY); + + return (le64toh(o->object.size) - offsetof(Object, entry_array.items)) / sizeof(uint64_t); +} + +static int link_entry_into_array(JournalFile *f, + le64_t *first, + le64_t *idx, + uint64_t p) { int r; + uint64_t n = 0, ap = 0, q, i, a, hidx; + Object *o; + assert(f); - assert(o); - assert(offset > 0); + assert(first); + assert(idx); + assert(p > 0); - p = le64toh(o->entry.items[i].object_offset); - if (p == 0) - return -EINVAL; + a = le64toh(*first); + i = hidx = le64toh(*idx); + while (a > 0) { + + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; - o->entry.items[i].next_entry_offset = 0; + n = journal_file_entry_array_n_items(o); + if (i < n) { + o->entry_array.items[i] = htole64(p); + *idx = htole64(hidx + 1); + return 0; + } + + i -= n; + ap = a; + a = le64toh(o->entry_array.next_entry_array_offset); + } + + if (hidx > n) + n = (hidx+1) * 2; + else + n = n * 2; + + if (n < 4) + n = 4; + + r = journal_file_append_object(f, OBJECT_ENTRY_ARRAY, + offsetof(Object, entry_array.items) + n * sizeof(uint64_t), + &o, &q); + if (r < 0) + return r; - /* Move to the data object */ - r = journal_file_move_to_object(f, p, OBJECT_DATA, &o); + r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, q); if (r < 0) return r; - q = le64toh(o->data.tail_entry_offset); - o->data.tail_entry_offset = htole64(offset); + o->entry_array.items[i] = htole64(p); - if (q == 0) - o->data.head_entry_offset = htole64(offset); + if (ap == 0) + *first = htole64(q); else { - uint64_t n, j; - - /* Move to previous entry */ - r = journal_file_move_to_object(f, q, OBJECT_ENTRY, &o); + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, ap, &o); if (r < 0) return r; - n = journal_file_entry_n_items(o); - for (j = 0; j < n; j++) - if (le64toh(o->entry.items[j].object_offset) == p) - break; + o->entry_array.next_entry_array_offset = htole64(q); + } + + *idx = htole64(hidx + 1); - if (j >= n) - return -EBADMSG; + return 0; +} - o->entry.items[j].next_entry_offset = offset; +static int link_entry_into_array_plus_one(JournalFile *f, + le64_t *extra, + le64_t *first, + le64_t *idx, + uint64_t p) { + + int r; + + assert(f); + assert(extra); + assert(first); + assert(idx); + assert(p > 0); + + if (*idx == 0) + *extra = htole64(p); + else { + le64_t i; + + i = htole64(le64toh(*idx) - 1); + r = link_entry_into_array(f, first, &i, p); + if (r < 0) + return r; } - /* Move back to original entry */ - r = journal_file_move_to_object(f, offset, OBJECT_ENTRY, &o); + *idx = htole64(le64toh(*idx) + 1); + return 0; +} + +static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) { + uint64_t p; + int r; + assert(f); + assert(o); + assert(offset > 0); + + p = le64toh(o->entry.items[i].object_offset); + if (p == 0) + return -EINVAL; + + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); if (r < 0) return r; - o->entry.items[i].prev_entry_offset = q; - return 0; + return link_entry_into_array_plus_one(f, + &o->data.entry_offset, + &o->data.entry_array_offset, + &o->data.n_entries, + offset); } static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) { - uint64_t p, i, n, k, a, b; + uint64_t n, i; int r; assert(f); assert(o); assert(offset > 0); - assert(o->object.type == htole64(OBJECT_ENTRY)); + assert(o->object.type == OBJECT_ENTRY); + + __sync_synchronize(); /* Link up the entry itself */ - p = le64toh(f->header->tail_entry_offset); + r = link_entry_into_array(f, + &f->header->entry_array_offset, + &f->header->n_entries, + offset); + if (r < 0) + return r; - o->entry.prev_entry_offset = f->header->tail_entry_offset; - o->entry.next_entry_offset = 0; + /* log_debug("=> %s seqnr=%lu n_entries=%lu", f->path, (unsigned long) o->entry.seqnum, (unsigned long) f->header->n_entries); */ - if (p == 0) { - f->header->head_entry_offset = htole64(offset); + if (f->header->head_entry_realtime == 0) f->header->head_entry_realtime = o->entry.realtime; - } else { - /* Temporarily move back to the previous entry, to - * patch in pointer */ - - r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o); - if (r < 0) - return r; - o->entry.next_entry_offset = htole64(offset); - - r = journal_file_move_to_object(f, offset, OBJECT_ENTRY, &o); - if (r < 0) - return r; - } - - f->header->tail_entry_offset = htole64(offset); f->header->tail_entry_realtime = o->entry.realtime; + f->header->tail_entry_monotonic = o->entry.monotonic; + + f->tail_entry_monotonic_valid = true; /* Link up the items */ n = journal_file_entry_n_items(o); @@ -713,18 +959,6 @@ static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) { return r; } - /* Link up the entry in the bisect table */ - n = le64toh(f->header->bisect_table_size) / sizeof(uint64_t); - k = le64toh(f->header->arena_max_size) / n; - - a = (le64toh(f->header->last_bisect_offset) + k - 1) / k; - b = offset / k; - - for (; a <= b; a++) - f->bisect_table[a] = htole64(offset); - - f->header->last_bisect_offset = htole64(offset + le64toh(o->object.size)); - return 0; } @@ -733,6 +967,7 @@ static int journal_file_append_entry_internal( const dual_timestamp *ts, uint64_t xor_hash, const EntryItem items[], unsigned n_items, + uint64_t *seqnum, Object **ret, uint64_t *offset) { uint64_t np; uint64_t osize; @@ -741,21 +976,25 @@ static int journal_file_append_entry_internal( assert(f); assert(items || n_items == 0); + assert(ts); osize = offsetof(Object, entry.items) + (n_items * sizeof(EntryItem)); - r = journal_file_append_object(f, osize, &o, &np); + r = journal_file_append_object(f, OBJECT_ENTRY, osize, &o, &np); if (r < 0) return r; - o->object.type = htole64(OBJECT_ENTRY); - o->entry.seqnum = htole64(journal_file_seqnum(f)); + o->entry.seqnum = htole64(journal_file_entry_seqnum(f, seqnum)); memcpy(o->entry.items, items, n_items * sizeof(EntryItem)); - o->entry.realtime = ts ? htole64(ts->realtime) : 0; - o->entry.monotonic = ts ? htole64(ts->monotonic) : 0; + 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; @@ -769,18 +1008,48 @@ static int journal_file_append_entry_internal( return 0; } -int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const struct iovec iovec[], unsigned n_iovec, Object **ret, uint64_t *offset) { - unsigned i; - EntryItem *items; - int r; +void journal_file_post_change(JournalFile *f) { + assert(f); + + /* inotify() does not receive IN_MODIFY events from file + * accesses done via mmap(). After each access we hence + * trigger IN_MODIFY by truncating the journal file to its + * current size which triggers IN_MODIFY. */ + + __sync_synchronize(); + + if (ftruncate(f->fd, f->last_stat.st_size) < 0) + log_error("Failed to to truncate file to its own size: %m"); +} + +int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const struct iovec iovec[], unsigned n_iovec, uint64_t *seqnum, Object **ret, uint64_t *offset) { + unsigned i; + EntryItem *items; + int r; uint64_t xor_hash = 0; + struct dual_timestamp _ts; assert(f); assert(iovec || n_iovec == 0); - items = new(EntryItem, n_iovec); - if (!items) - return -ENOMEM; + if (!f->writable) + return -EPERM; + + if (!ts) { + dual_timestamp_get(&_ts); + ts = &_ts; + } + + if (f->tail_entry_monotonic_valid && + ts->monotonic < le64toh(f->header->tail_entry_monotonic)) + return -EINVAL; + + 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; @@ -788,115 +1057,279 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st r = journal_file_append_data(f, iovec[i].iov_base, iovec[i].iov_len, &o, &p); if (r < 0) - goto finish; + return r; xor_hash ^= le64toh(o->data.hash); items[i].object_offset = htole64(p); + items[i].hash = o->data.hash; } - r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, ret, offset); + r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqnum, ret, offset); -finish: - free(items); + journal_file_post_change(f); return r; } -int journal_file_move_to_entry(JournalFile *f, uint64_t seqnum, Object **ret, uint64_t *offset) { +static int generic_array_get(JournalFile *f, + uint64_t first, + uint64_t i, + Object **ret, uint64_t *offset) { + Object *o; - uint64_t lower, upper, p, n, k; + uint64_t p = 0, a; int r; assert(f); - n = le64toh(f->header->bisect_table_size) / sizeof(uint64_t); - k = le64toh(f->header->arena_max_size) / n; + a = first; + while (a > 0) { + uint64_t n; - lower = 0; - upper = le64toh(f->header->last_bisect_offset)/k+1; - - while (lower < upper) { - k = (upper + lower) / 2; - p = le64toh(f->bisect_table[k]); + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; - if (p == 0) { - upper = k; - continue; + n = journal_file_entry_array_n_items(o); + if (i < n) { + p = le64toh(o->entry_array.items[i]); + break; } - r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o); + i -= n; + a = le64toh(o->entry_array.next_entry_array_offset); + } + + if (a <= 0 || p <= 0) + return 0; + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + if (ret) + *ret = o; + + if (offset) + *offset = p; + + return 1; +} + +static int generic_array_get_plus_one(JournalFile *f, + uint64_t extra, + uint64_t first, + uint64_t i, + Object **ret, uint64_t *offset) { + + Object *o; + + assert(f); + + if (i == 0) { + int r; + + r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o); if (r < 0) return r; - if (o->entry.seqnum == seqnum) { - if (ret) - *ret = o; + if (ret) + *ret = o; - if (offset) - *offset = p; + if (offset) + *offset = extra; - return 1; - } else if (seqnum < o->entry.seqnum) - upper = k; - else if (seqnum > o->entry.seqnum) - lower = k+1; + return 1; } - assert(lower == upper); + return generic_array_get(f, first, i-1, ret, offset); +} + +enum { + TEST_FOUND, + TEST_LEFT, + TEST_RIGHT +}; - if (lower <= 0) - return 0; +static int generic_array_bisect(JournalFile *f, + uint64_t first, + uint64_t n, + uint64_t needle, + int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle), + direction_t direction, + Object **ret, + uint64_t *offset, + uint64_t *idx) { + + uint64_t a, p, t = 0, i = 0, last_p = 0; + bool subtract_one = false; + Object *o, *array = NULL; + int r; - /* The object we are looking for is between - * bisect_table[lower-1] and bisect_table[lower] */ + assert(f); + assert(test_object); - p = le64toh(f->bisect_table[lower-1]); + a = first; + while (a > 0) { + uint64_t left, right, k, lp; - for (;;) { - r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o); + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &array); if (r < 0) return r; - if (o->entry.seqnum == seqnum) { - if (ret) - *ret = o; + k = journal_file_entry_array_n_items(array); + right = MIN(k, n); + if (right <= 0) + return 0; - if (offset) - *offset = p; + i = right - 1; + lp = p = le64toh(array->entry_array.items[i]); + if (p <= 0) + return -EBADMSG; - return 1; + r = test_object(f, p, needle); + if (r < 0) + return r; - } if (seqnum < o->entry.seqnum) - return 0; + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + if (r == TEST_RIGHT) { + left = 0; + right -= 1; + for (;;) { + if (left == right) { + if (direction == DIRECTION_UP) + subtract_one = true; + + i = left; + goto found; + } + + assert(left < right); + + i = (left + right) / 2; + p = le64toh(array->entry_array.items[i]); + if (p <= 0) + return -EBADMSG; + + r = test_object(f, p, needle); + if (r < 0) + return r; + + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + if (r == TEST_RIGHT) + right = i; + else + left = i + 1; + } + } + + if (k > n) { + if (direction == DIRECTION_UP) { + i = n; + subtract_one = true; + goto found; + } - if (o->entry.next_entry_offset == 0) return 0; + } - p = le64toh(o->entry.next_entry_offset); + last_p = lp; + + n -= k; + t += k; + a = le64toh(array->entry_array.next_entry_array_offset); } return 0; + +found: + if (subtract_one && t == 0 && i == 0) + return 0; + + if (subtract_one && i == 0) + p = last_p; + else if (subtract_one) + p = le64toh(array->entry_array.items[i-1]); + else + p = le64toh(array->entry_array.items[i]); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + if (ret) + *ret = o; + + if (offset) + *offset = p; + + if (idx) + *idx = t + i + (subtract_one ? -1 : 0); + + return 1; } -int journal_file_next_entry(JournalFile *f, Object *o, Object **ret, uint64_t *offset) { - uint64_t np; +static int generic_array_bisect_plus_one(JournalFile *f, + uint64_t extra, + uint64_t first, + uint64_t n, + uint64_t needle, + int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle), + direction_t direction, + Object **ret, + uint64_t *offset, + uint64_t *idx) { + int r; + bool step_back = false; + Object *o; assert(f); + assert(test_object); - if (!o) - np = le64toh(f->header->head_entry_offset); - else { - if (le64toh(o->object.type) != OBJECT_ENTRY) - return -EINVAL; + if (n <= 0) + return 0; + + /* This bisects the array in object 'first', but first checks + * an extra */ + r = test_object(f, extra, needle); + if (r < 0) + return r; + + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + /* if we are looking with DIRECTION_UP then we need to first + see if in the actual array there is a matching entry, and + return the last one of that. But if there isn't any we need + to return this one. Hence remember this, and return it + below. */ + if (r == TEST_LEFT) + step_back = direction == DIRECTION_UP; - np = le64toh(o->entry.next_entry_offset); + if (r == TEST_RIGHT) { + if (direction == DIRECTION_DOWN) + goto found; + else + return 0; } - if (np == 0) - return 0; + r = generic_array_bisect(f, first, n-1, needle, test_object, direction, ret, offset, idx); + + if (r == 0 && step_back) + goto found; - r = journal_file_move_to_object(f, np, OBJECT_ENTRY, &o); + if (r > 0 && idx) + (*idx) ++; + + return r; + +found: + r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o); if (r < 0) return r; @@ -904,160 +1337,1437 @@ int journal_file_next_entry(JournalFile *f, Object *o, Object **ret, uint64_t *o *ret = o; if (offset) - *offset = np; + *offset = extra; + + if (idx) + *idx = 0; return 1; } -int journal_file_prev_entry(JournalFile *f, Object *o, Object **ret, uint64_t *offset) { - uint64_t np; +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; + + assert(f); + assert(p > 0); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + if (le64toh(o->entry.seqnum) == needle) + return TEST_FOUND; + else if (le64toh(o->entry.seqnum) < needle) + return TEST_LEFT; + else + return TEST_RIGHT; +} + +int journal_file_move_to_entry_by_seqnum( + JournalFile *f, + uint64_t seqnum, + direction_t direction, + Object **ret, + uint64_t *offset) { + + return generic_array_bisect(f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + seqnum, + test_object_seqnum, + direction, + ret, offset, NULL); +} + +static int test_object_realtime(JournalFile *f, uint64_t p, uint64_t needle) { + Object *o; + int r; + + assert(f); + assert(p > 0); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + if (le64toh(o->entry.realtime) == needle) + return TEST_FOUND; + else if (le64toh(o->entry.realtime) < needle) + return TEST_LEFT; + else + return TEST_RIGHT; +} + +int journal_file_move_to_entry_by_realtime( + JournalFile *f, + uint64_t realtime, + direction_t direction, + Object **ret, + uint64_t *offset) { + + return generic_array_bisect(f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + realtime, + test_object_realtime, + direction, + ret, offset, NULL); +} + +static int test_object_monotonic(JournalFile *f, uint64_t p, uint64_t needle) { + Object *o; int r; assert(f); + assert(p > 0); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + if (le64toh(o->entry.monotonic) == needle) + return TEST_FOUND; + else if (le64toh(o->entry.monotonic) < needle) + return TEST_LEFT; + else + return TEST_RIGHT; +} + +int journal_file_move_to_entry_by_monotonic( + JournalFile *f, + sd_id128_t boot_id, + uint64_t monotonic, + direction_t direction, + Object **ret, + uint64_t *offset) { + + char t[9+32+1] = "_BOOT_ID="; + Object *o; + int r; + + 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; + if (r == 0) + return -ENOENT; + + return generic_array_bisect_plus_one(f, + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + le64toh(o->data.n_entries), + monotonic, + test_object_monotonic, + direction, + ret, offset, NULL); +} + +int journal_file_next_entry( + JournalFile *f, + Object *o, uint64_t p, + direction_t direction, + Object **ret, uint64_t *offset) { + + uint64_t i, n; + int r; + + assert(f); + assert(p > 0 || !o); + + n = le64toh(f->header->n_entries); + if (n <= 0) + return 0; if (!o) - np = le64toh(f->header->tail_entry_offset); + i = direction == DIRECTION_DOWN ? 0 : n - 1; else { - if (le64toh(o->object.type) != OBJECT_ENTRY) + if (o->object.type != OBJECT_ENTRY) return -EINVAL; - np = le64toh(o->entry.prev_entry_offset); + r = generic_array_bisect(f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + p, + test_object_offset, + DIRECTION_DOWN, + NULL, NULL, + &i); + if (r <= 0) + return r; + + if (direction == DIRECTION_DOWN) { + if (i >= n - 1) + return 0; + + i++; + } else { + if (i <= 0) + return 0; + + i--; + } } - if (np == 0) - return 0; + /* And jump to it */ + return generic_array_get(f, + le64toh(f->header->entry_array_offset), + i, + ret, offset); +} - r = journal_file_move_to_object(f, np, OBJECT_ENTRY, &o); - if (r < 0) +int journal_file_skip_entry( + JournalFile *f, + Object *o, uint64_t p, + int64_t skip, + Object **ret, uint64_t *offset) { + + uint64_t i, n; + int r; + + assert(f); + assert(o); + assert(p > 0); + + if (o->object.type != OBJECT_ENTRY) + return -EINVAL; + + r = generic_array_bisect(f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + p, + test_object_offset, + DIRECTION_DOWN, + NULL, NULL, + &i); + if (r <= 0) return r; - if (ret) - *ret = o; + /* Calculate new index */ + if (skip < 0) { + if ((uint64_t) -skip >= i) + i = 0; + else + i = i - (uint64_t) -skip; + } else + i += (uint64_t) skip; - if (offset) - *offset = np; + n = le64toh(f->header->n_entries); + if (n <= 0) + return -EBADMSG; - return 1; + if (i >= n) + i = n-1; + + return generic_array_get(f, + le64toh(f->header->entry_array_offset), + i, + ret, offset); } -int journal_file_find_first_entry(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset) { - uint64_t p, osize, hash, h; +int journal_file_next_entry_for_data( + JournalFile *f, + Object *o, uint64_t p, + uint64_t data_offset, + direction_t direction, + Object **ret, uint64_t *offset) { + + uint64_t n, i; int r; + Object *d; assert(f); - assert(data || size == 0); + assert(p > 0 || !o); - osize = offsetof(Object, data.payload) + size; + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); + if (r < 0) + return r; - hash = hash64(data, size); - h = hash % (le64toh(f->header->hash_table_size) / sizeof(HashItem)); - p = le64toh(f->hash_table[h].head_hash_offset); + n = le64toh(d->data.n_entries); + if (n <= 0) + return n; - while (p != 0) { - Object *o; + if (!o) + i = direction == DIRECTION_DOWN ? 0 : n - 1; + else { + if (o->object.type != OBJECT_ENTRY) + return -EINVAL; - r = journal_file_move_to_object(f, p, OBJECT_DATA, &o); - if (r < 0) + r = generic_array_bisect_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + p, + test_object_offset, + DIRECTION_DOWN, + NULL, NULL, + &i); + + if (r <= 0) return r; - if (le64toh(o->object.size) == osize && - memcmp(o->data.payload, data, size) == 0) { - - if (le64toh(o->data.hash) != hash) - return -EBADMSG; + if (direction == DIRECTION_DOWN) { + if (i >= n - 1) + return 0; - if (o->data.head_entry_offset == 0) + i++; + } else { + if (i <= 0) return 0; - p = le64toh(o->data.head_entry_offset); - r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o); - if (r < 0) - return r; + i--; + } - if (ret) - *ret = o; + } - if (offset) - *offset = p; + return generic_array_get_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + i, + ret, offset); +} - return 1; - } +int journal_file_move_to_entry_by_offset_for_data( + JournalFile *f, + uint64_t data_offset, + uint64_t p, + direction_t direction, + Object **ret, uint64_t *offset) { - p = le64toh(o->data.next_hash_offset); - } + int r; + Object *d; - return 0; + 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_find_last_entry(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset) { - uint64_t p, osize, hash, h; +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); - assert(data || size == 0); - osize = offsetof(Object, data.payload) + size; + /* 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, + uint64_t seqnum, + direction_t direction, + Object **ret, uint64_t *offset) { + + Object *d; + int r; + + assert(f); + + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); + if (r < 0) + return r; + + return generic_array_bisect_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + seqnum, + test_object_seqnum, + direction, + ret, offset, NULL); +} + +int journal_file_move_to_entry_by_realtime_for_data( + JournalFile *f, + uint64_t data_offset, + uint64_t realtime, + direction_t direction, + Object **ret, uint64_t *offset) { + + Object *d; + int r; + + assert(f); + + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); + if (r < 0) + return r; + + return generic_array_bisect_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + realtime, + test_object_realtime, + direction, + ret, offset, NULL); +} + +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; + } + } - hash = hash64(data, size); - h = hash % (le64toh(f->header->hash_table_size) / sizeof(HashItem)); - p = le64toh(f->hash_table[h].tail_hash_offset); + } - while (p != 0) { - Object *o; + 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; + } - r = journal_file_move_to_object(f, p, OBJECT_DATA, &o); - if (r < 0) - return r; + if (p == le64toh(f->header->tail_object_offset)) + p = 0; + else + p = p + ALIGN64(le64toh(o->object.size)); + } - if (le64toh(o->object.size) == osize && - memcmp(o->data.payload, data, size) == 0) { + flush_progress(); - if (le64toh(o->data.hash) != hash) - return -EBADMSG; + 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); - if (o->data.tail_entry_offset == 0) - return 0; + close_nointr_nofail(data_fd); + close_nointr_nofail(entry_fd); + close_nointr_nofail(entry_array_fd); - p = le64toh(o->data.tail_entry_offset); - r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o); - if (r < 0) - return r; + return 0; - if (ret) - *ret = o; +fail: + flush_progress(); - if (offset) - *offset = p; + 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)); - return 1; - } + if (data_fd >= 0) { + mmap_cache_close_fd(f->mmap, data_fd); + close_nointr_nofail(data_fd); + } - p = le64toh(o->data.prev_hash_offset); + if (entry_fd >= 0) { + mmap_cache_close_fd(f->mmap, entry_fd); + close_nointr_nofail(entry_fd); } - return 0; + 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) { - char a[33], b[33], c[33]; Object *o; int r; uint64_t p; assert(f); - printf("File ID: %s\n" - "Machine ID: %s\n" - "Boot ID: %s\n" - "Arena size: %llu\n", - 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)); + journal_file_print_header(f); - p = le64toh(f->header->head_object_offset); + p = le64toh(f->header->header_size); while (p != 0) { - r = journal_file_move_to_object(f, p, -1, &o); + r = journal_file_move_to_object(f, -1, p, &o); if (r < 0) goto fail; @@ -1072,18 +2782,33 @@ void journal_file_dump(JournalFile *f) { break; case OBJECT_ENTRY: - printf("Type: OBJECT_ENTRY %llu\n", (unsigned long long) le64toh(o->entry.seqnum)); + printf("Type: OBJECT_ENTRY %llu %llu %llu\n", + (unsigned long long) le64toh(o->entry.seqnum), + (unsigned long long) le64toh(o->entry.monotonic), + (unsigned long long) le64toh(o->entry.realtime)); + break; + + case OBJECT_FIELD_HASH_TABLE: + printf("Type: OBJECT_FIELD_HASH_TABLE\n"); + break; + + case OBJECT_DATA_HASH_TABLE: + printf("Type: OBJECT_DATA_HASH_TABLE\n"); break; - case OBJECT_HASH_TABLE: - printf("Type: OBJECT_HASH_TABLE\n"); + case OBJECT_ENTRY_ARRAY: + printf("Type: OBJECT_ENTRY_ARRAY\n"); break; - case OBJECT_BISECT_TABLE: - printf("Type: OBJECT_BISECT_TABLE\n"); + case OBJECT_TAG: + printf("Type: OBJECT_TAG %llu\n", + (unsigned long long) le64toh(o->tag.seqnum)); break; } + if (o->object.flags & OBJECT_COMPRESSED) + printf("Flags: COMPRESSED\n"); + if (p == le64toh(f->header->tail_object_offset)) p = 0; else @@ -1095,10 +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) { @@ -1112,15 +2903,34 @@ int journal_file_open( (flags & O_ACCMODE) != O_RDWR) return -EINVAL; + if (!endswith(fname, ".journal")) + return -EINVAL; + f = new0(JournalFile, 1); if (!f) 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 (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); if (!f->path) { @@ -1142,6 +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; @@ -1152,7 +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; } @@ -1170,28 +2986,47 @@ 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; - r = journal_file_setup_hash_table(f); + r = journal_file_setup_data_hash_table(f); if (r < 0) goto fail; - r = journal_file_setup_bisect_table(f); + r = journal_file_append_first_tag(f); if (r < 0) goto fail; } - r = journal_file_map_hash_table(f); + r = journal_file_map_field_hash_table(f); if (r < 0) goto fail; - r = journal_file_map_bisect_table(f); + r = journal_file_map_data_hash_table(f); if (r < 0) goto fail; @@ -1206,7 +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; @@ -1225,7 +3060,7 @@ int journal_file_rotate(JournalFile **f) { l = strlen(old_file->path); - p = new(char, l + 1 + 16 + 1 + 32 + 1 + 16 + 1); + p = new(char, l + 1 + 32 + 1 + 16 + 1 + 16 + 1); if (!p) return -ENOMEM; @@ -1234,7 +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); @@ -1243,15 +3078,67 @@ int journal_file_rotate(JournalFile **f) { if (r < 0) return -errno; - old_file->header->state = le32toh(STATE_ARCHIVED); + 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; return r; } +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) { + + int r; + size_t l; + char *p; + + 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 != -EBUSY && /* unclean shutdown */ + r != -ESHUTDOWN /* already archived */) + return r; + + if ((flags & O_ACCMODE) == O_RDONLY) + return r; + + if (!(flags & O_CREAT)) + return r; + + if (!endswith(fname, ".journal")) + return r; + + /* The file is corrupted. Rotate it away and try it again (but only once) */ + + l = strlen(fname); + if (asprintf(&p, "%.*s@%016llx-%016llx.journal~", + (int) (l-8), fname, + (unsigned long long) now(CLOCK_REALTIME), + random_ull()) < 0) + return -ENOMEM; + + r = rename(fname, p); + free(p); + if (r < 0) + return -errno; + + log_warning("File %s corrupted or uncleanly shut down, renaming and replacing.", fname); + + return journal_file_open(fname, flags, mode, compress, authenticate, metrics, mmap, template, ret); +} + struct vacuum_info { off_t usage; char *filename; @@ -1259,6 +3146,8 @@ struct vacuum_info { uint64_t realtime; sd_id128_t seqnum_id; uint64_t seqnum; + + bool have_seqnum; }; static int vacuum_compare(const void *_a, const void *_b) { @@ -1267,7 +3156,8 @@ static int vacuum_compare(const void *_a, const void *_b) { a = _a; b = _b; - if (sd_id128_equal(a->seqnum_id, b->seqnum_id)) { + if (a->have_seqnum && b->have_seqnum && + sd_id128_equal(a->seqnum_id, b->seqnum_id)) { if (a->seqnum < b->seqnum) return -1; else if (a->seqnum > b->seqnum) @@ -1280,8 +3170,10 @@ static int vacuum_compare(const void *_a, const void *_b) { return -1; else if (a->realtime > b->realtime) return 1; - else + else if (a->have_seqnum && b->have_seqnum) return memcmp(&a->seqnum_id, &b->seqnum_id, 16); + else + return strcmp(a->filename, b->filename); } int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t min_free) { @@ -1294,7 +3186,7 @@ int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t m assert(directory); if (max_use <= 0) - max_use = DEFAULT_MAX_USE; + return 0; d = opendir(directory); if (!d) @@ -1306,8 +3198,9 @@ int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t m size_t q; struct stat st; char *p; - unsigned long long seqnum, realtime; + unsigned long long seqnum = 0, realtime; sd_id128_t seqnum_id; + bool have_seqnum; k = readdir_r(d, &buf, &de); if (k != 0) { @@ -1318,41 +3211,71 @@ int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t m if (!de) break; - if (!dirent_is_file_with_suffix(de, ".journal")) + if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0) + continue; + + if (!S_ISREG(st.st_mode)) continue; q = strlen(de->d_name); - if (q < 1 + 32 + 1 + 16 + 1 + 16 + 8) - continue; + if (endswith(de->d_name, ".journal")) { - if (de->d_name[q-8-16-1] != '-' || - de->d_name[q-8-16-1-16-1] != '-' || - de->d_name[q-8-16-1-16-1-32-1] != '@') - continue; + /* Vacuum archived files */ - if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0) - continue; + if (q < 1 + 32 + 1 + 16 + 1 + 16 + 8) + continue; - if (!S_ISREG(st.st_mode)) - continue; + if (de->d_name[q-8-16-1] != '-' || + de->d_name[q-8-16-1-16-1] != '-' || + de->d_name[q-8-16-1-16-1-32-1] != '@') + continue; - p = strdup(de->d_name); - if (!p) { - r = -ENOMEM; - goto finish; - } + p = strdup(de->d_name); + if (!p) { + r = -ENOMEM; + goto finish; + } - de->d_name[q-8-16-1-16-1] = 0; - if (sd_id128_from_string(de->d_name + q-8-16-1-16-1-32, &seqnum_id) < 0) { - free(p); - continue; - } + de->d_name[q-8-16-1-16-1] = 0; + if (sd_id128_from_string(de->d_name + q-8-16-1-16-1-32, &seqnum_id) < 0) { + free(p); + continue; + } + + if (sscanf(de->d_name + q-8-16-1-16, "%16llx-%16llx.journal", &seqnum, &realtime) != 2) { + free(p); + continue; + } + + have_seqnum = true; + + } else if (endswith(de->d_name, ".journal~")) { + unsigned long long tmp; + + /* Vacuum corrupted files */ + + if (q < 1 + 16 + 1 + 16 + 8 + 1) + continue; + + if (de->d_name[q-1-8-16-1] != '-' || + de->d_name[q-1-8-16-1-16-1] != '@') + continue; + + p = strdup(de->d_name); + if (!p) { + r = -ENOMEM; + goto finish; + } + + if (sscanf(de->d_name + q-1-8-16-1-16, "%16llx-%16llx.journal~", &realtime, &tmp) != 2) { + free(p); + continue; + } - if (sscanf(de->d_name + q-8-16-1-16, "%16llx-%16llx.journal", &seqnum, &realtime) != 2) { - free(p); + have_seqnum = false; + } else continue; - } if (n_list >= n_allocated) { struct vacuum_info *j; @@ -1369,17 +3292,19 @@ int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t m } list[n_list].filename = p; - list[n_list].usage = (uint64_t) st.st_blksize * (uint64_t) st.st_blocks; + list[n_list].usage = 512UL * (uint64_t) st.st_blocks; list[n_list].seqnum = seqnum; list[n_list].realtime = realtime; list[n_list].seqnum_id = seqnum_id; + list[n_list].have_seqnum = have_seqnum; sum += list[n_list].usage; 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; @@ -1394,7 +3319,7 @@ int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t m break; if (unlinkat(dirfd(d), list[i].filename, 0) >= 0) { - log_debug("Deleted archived journal %s/%s.", directory, list[i].filename); + log_info("Deleted archived journal %s/%s.", directory, list[i].filename); sum -= list[i].usage; } else if (errno != ENOENT) log_warning("Failed to delete %s/%s: %m", directory, list[i].filename); @@ -1406,5 +3331,270 @@ finish: free(list); + if (d) + closedir(d); + return r; } + +int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint64_t p, uint64_t *seqnum, Object **ret, uint64_t *offset) { + uint64_t i, n; + uint64_t q, xor_hash = 0; + int r; + EntryItem *items; + dual_timestamp ts; + + assert(from); + assert(to); + assert(o); + assert(p); + + if (!to->writable) + return -EPERM; + + ts.monotonic = le64toh(o->entry.monotonic); + ts.realtime = le64toh(o->entry.realtime); + + if (to->tail_entry_monotonic_valid && + ts.monotonic < le64toh(to->header->tail_entry_monotonic)) + return -EINVAL; + + n = journal_file_entry_n_items(o); + items = alloca(sizeof(EntryItem) * n); + + for (i = 0; i < n; i++) { + uint64_t l, h; + le64_t le_hash; + size_t t; + void *data; + Object *u; + + q = le64toh(o->entry.items[i].object_offset); + le_hash = o->entry.items[i].hash; + + r = journal_file_move_to_object(from, OBJECT_DATA, q, &o); + if (r < 0) + return r; + + if (le_hash != o->data.hash) + return -EBADMSG; + + l = le64toh(o->object.size) - offsetof(Object, data.payload); + t = (size_t) l; + + /* We hit the limit on 32bit machines */ + if ((uint64_t) t != l) + return -E2BIG; + + if (o->object.flags & OBJECT_COMPRESSED) { +#ifdef HAVE_XZ + uint64_t rsize; + + if (!uncompress_blob(o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize)) + return -EBADMSG; + + data = from->compress_buffer; + l = rsize; +#else + return -EPROTONOSUPPORT; +#endif + } else + data = o->data.payload; + + r = journal_file_append_data(to, data, l, &u, &h); + if (r < 0) + return r; + + xor_hash ^= le64toh(u->data.hash); + items[i].object_offset = htole64(h); + items[i].hash = u->data.hash; + + r = journal_file_move_to_object(from, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + } + + return journal_file_append_entry_internal(to, &ts, xor_hash, items, n, seqnum, ret, offset); +} + +void journal_default_metrics(JournalMetrics *m, int fd) { + uint64_t fs_size = 0; + struct statvfs ss; + char a[FORMAT_BYTES_MAX], b[FORMAT_BYTES_MAX], c[FORMAT_BYTES_MAX], d[FORMAT_BYTES_MAX]; + + assert(m); + assert(fd >= 0); + + if (fstatvfs(fd, &ss) >= 0) + fs_size = ss.f_frsize * ss.f_blocks; + + if (m->max_use == (uint64_t) -1) { + + if (fs_size > 0) { + m->max_use = PAGE_ALIGN(fs_size / 10); /* 10% of file system size */ + + if (m->max_use > DEFAULT_MAX_USE_UPPER) + m->max_use = DEFAULT_MAX_USE_UPPER; + + if (m->max_use < DEFAULT_MAX_USE_LOWER) + m->max_use = DEFAULT_MAX_USE_LOWER; + } else + m->max_use = DEFAULT_MAX_USE_LOWER; + } else { + m->max_use = PAGE_ALIGN(m->max_use); + + if (m->max_use < JOURNAL_FILE_SIZE_MIN*2) + m->max_use = JOURNAL_FILE_SIZE_MIN*2; + } + + if (m->max_size == (uint64_t) -1) { + m->max_size = PAGE_ALIGN(m->max_use / 8); /* 8 chunks */ + + if (m->max_size > DEFAULT_MAX_SIZE_UPPER) + m->max_size = DEFAULT_MAX_SIZE_UPPER; + } else + m->max_size = PAGE_ALIGN(m->max_size); + + if (m->max_size < JOURNAL_FILE_SIZE_MIN) + m->max_size = JOURNAL_FILE_SIZE_MIN; + + if (m->max_size*2 > m->max_use) + m->max_use = m->max_size*2; + + if (m->min_size == (uint64_t) -1) + m->min_size = JOURNAL_FILE_SIZE_MIN; + else { + m->min_size = PAGE_ALIGN(m->min_size); + + if (m->min_size < JOURNAL_FILE_SIZE_MIN) + m->min_size = JOURNAL_FILE_SIZE_MIN; + + if (m->min_size > m->max_size) + m->max_size = m->min_size; + } + + if (m->keep_free == (uint64_t) -1) { + + if (fs_size > 0) { + m->keep_free = PAGE_ALIGN(fs_size / 20); /* 5% of file system size */ + + if (m->keep_free > DEFAULT_KEEP_FREE_UPPER) + m->keep_free = DEFAULT_KEEP_FREE_UPPER; + + } else + m->keep_free = DEFAULT_KEEP_FREE; + } + + log_info("Fixed max_use=%s max_size=%s min_size=%s keep_free=%s", + format_bytes(a, sizeof(a), m->max_use), + format_bytes(b, sizeof(b), m->max_size), + format_bytes(c, sizeof(c), m->min_size), + format_bytes(d, sizeof(d), m->keep_free)); +} + +int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *to) { + assert(f); + assert(from || to); + + if (from) { + if (f->header->head_entry_realtime == 0) + return -ENOENT; + + *from = le64toh(f->header->head_entry_realtime); + } + + if (to) { + if (f->header->tail_entry_realtime == 0) + return -ENOENT; + + *to = le64toh(f->header->tail_entry_realtime); + } + + return 1; +} + +int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, usec_t *from, usec_t *to) { + char t[9+32+1] = "_BOOT_ID="; + Object *o; + uint64_t p; + int r; + + assert(f); + assert(from || to); + + sd_id128_to_string(boot_id, t + 9); + + r = journal_file_find_data_object(f, t, strlen(t), &o, &p); + if (r <= 0) + return r; + + if (le64toh(o->data.n_entries) <= 0) + return 0; + + if (from) { + r = journal_file_move_to_object(f, OBJECT_ENTRY, le64toh(o->data.entry_offset), &o); + if (r < 0) + return r; + + *from = le64toh(o->entry.monotonic); + } + + if (to) { + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + r = generic_array_get_plus_one(f, + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + le64toh(o->data.n_entries)-1, + &o, NULL); + if (r <= 0) + return r; + + *to = le64toh(o->entry.monotonic); + } + + return 1; +} + +bool journal_file_rotate_suggested(JournalFile *f) { + 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; +}