X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?a=blobdiff_plain;f=src%2Fjournal%2Fjournal-file.c;h=3bb1e90fb324b9f86b3a8c76c87cf641a480fa0e;hb=56e81f7ca8276e40f8c88c4c30713a5b54009613;hp=9235e5fea2df5711acbe92ce141567ef175c7665;hpb=d98cc1f29fbf31ccc500d6e20c29b636b9af7e0f;p=elogind.git diff --git a/src/journal/journal-file.c b/src/journal/journal-file.c index 9235e5fea..3bb1e90fb 100644 --- a/src/journal/journal-file.c +++ b/src/journal/journal-file.c @@ -29,6 +29,7 @@ #include "journal-def.h" #include "journal-file.h" +#include "journal-authenticate.h" #include "lookup3.h" #include "compress.h" #include "fsprg.h" @@ -36,8 +37,6 @@ #define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*sizeof(HashItem)) #define DEFAULT_FIELD_HASH_TABLE_SIZE (333ULL*sizeof(HashItem)) -#define DEFAULT_WINDOW_SIZE (8ULL*1024ULL*1024ULL) - #define COMPRESSION_SIZE_THRESHOLD (512ULL) /* This is the minimum journal file size */ @@ -62,17 +61,7 @@ /* 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) { - int t; - assert(f); /* Write the final tag */ @@ -80,9 +69,8 @@ void journal_file_close(JournalFile *f) { journal_file_append_tag(f); /* Sync everything to disk, before we mark the file offline */ - for (t = 0; t < _WINDOW_MAX; t++) - if (f->windows[t].ptr) - munmap(f->windows[t].ptr, f->windows[t].size); + if (f->mmap && f->fd >= 0) + mmap_cache_close_fd(f->mmap, f->fd); if (f->writable && f->fd >= 0) fdatasync(f->fd); @@ -100,13 +88,20 @@ void journal_file_close(JournalFile *f) { 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->fsprg_file) + munmap(f->fsprg_file, PAGE_ALIGN(f->fsprg_file_size)); + else if (f->fsprg_state) + free(f->fsprg_state); + + free(f->fsprg_seed); if (f->hmac) gcry_md_close(f->hmac); @@ -138,7 +133,7 @@ static int journal_file_init_header(JournalFile *f, JournalFile *template) { if (template) { h.seqnum_id = template->header->seqnum_id; - h.tail_seqnum = template->header->tail_seqnum; + h.tail_entry_seqnum = template->header->tail_entry_seqnum; } else h.seqnum_id = h.file_id; @@ -212,6 +207,10 @@ static int journal_file_verify_header(JournalFile *f) { if (le64toh(f->header->header_size) < HEADER_SIZE_MIN) return -EBADMSG; + if ((le32toh(f->header->compatible_flags) & HEADER_COMPATIBLE_AUTHENTICATED) && + !JOURNAL_HEADER_CONTAINS(f->header, n_tags)) + return -EBADMSG; + if ((uint64_t) f->last_stat.st_size < (le64toh(f->header->header_size) + le64toh(f->header->arena_size))) return -ENODATA; @@ -297,6 +296,8 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) if (r != 0) return -r; + mmap_cache_close_fd_range(f->mmap, f->fd, old_size); + if (fstat(f->fd, &f->last_stat) < 0) return -errno; @@ -305,59 +306,11 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) return 0; } -static int journal_file_map( - JournalFile *f, - uint64_t offset, - uint64_t size, - void **_window, - uint64_t *_woffset, - uint64_t *_wsize, - void **ret) { - - uint64_t woffset, wsize; - void *window; - +static int journal_file_move_to(JournalFile *f, int context, uint64_t offset, uint64_t size, void **ret) { assert(f); - assert(size > 0); assert(ret); - woffset = offset & ~((uint64_t) page_size() - 1ULL); - wsize = size + (offset - woffset); - wsize = PAGE_ALIGN(wsize); - /* Avoid SIGBUS on invalid accesses */ - if (woffset + wsize > (uint64_t) PAGE_ALIGN(f->last_stat.st_size)) - return -EADDRNOTAVAIL; - - window = mmap(NULL, wsize, f->prot, MAP_SHARED, f->fd, woffset); - if (window == MAP_FAILED) - return -errno; - - if (_window) - *_window = window; - - if (_woffset) - *_woffset = woffset; - - if (_wsize) - *_wsize = wsize; - - *ret = (uint8_t*) window + (offset - woffset); - - return 0; -} - -static int journal_file_move_to(JournalFile *f, int wt, uint64_t offset, uint64_t size, void **ret) { - void *p = NULL; - uint64_t delta; - int r; - Window *w; - - assert(f); - assert(ret); - assert(wt >= 0); - assert(wt < _WINDOW_MAX); - if (offset + size > (uint64_t) f->last_stat.st_size) { /* Hmm, out of range? Let's refresh the fstat() data * first, before we trust that check. */ @@ -367,74 +320,25 @@ static int journal_file_move_to(JournalFile *f, int wt, uint64_t offset, uint64_ return -EADDRNOTAVAIL; } - w = f->windows + wt; - - if (_likely_(w->ptr && - w->offset <= offset && - w->offset + w->size >= offset + size)) { - - *ret = (uint8_t*) w->ptr + (offset - w->offset); - return 0; - } - - if (w->ptr) { - if (munmap(w->ptr, w->size) < 0) - return -errno; - - w->ptr = NULL; - w->size = w->offset = 0; - } - - if (size < DEFAULT_WINDOW_SIZE) { - /* If the default window size is larger then what was - * asked for extend the mapping a bit in the hope to - * minimize needed remappings later on. We add half - * the window space before and half behind the - * requested mapping */ - - delta = (DEFAULT_WINDOW_SIZE - size) / 2; - - if (delta > offset) - delta = offset; - - offset -= delta; - size = DEFAULT_WINDOW_SIZE; - } else - delta = 0; - - if (offset + size > (uint64_t) f->last_stat.st_size) - size = (uint64_t) f->last_stat.st_size - offset; - - if (size <= 0) - return -EADDRNOTAVAIL; - - r = journal_file_map(f, - offset, size, - &w->ptr, &w->offset, &w->size, - &p); - - if (r < 0) - return r; - - *ret = (uint8_t*) p + delta; - return 0; + return mmap_cache_get(f->mmap, f->fd, f->prot, context, offset, size, ret); } -static bool verify_hash(Object *o) { - uint64_t h1, h2; +static uint64_t minimum_header_size(Object *o) { - assert(o); + static uint64_t table[] = { + [OBJECT_DATA] = sizeof(DataObject), + [OBJECT_FIELD] = sizeof(FieldObject), + [OBJECT_ENTRY] = sizeof(EntryObject), + [OBJECT_DATA_HASH_TABLE] = sizeof(HashTableObject), + [OBJECT_FIELD_HASH_TABLE] = sizeof(HashTableObject), + [OBJECT_ENTRY_ARRAY] = sizeof(EntryArrayObject), + [OBJECT_TAG] = sizeof(TagObject), + }; - if (o->object.type == 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; + if (o->object.type >= ELEMENTSOF(table) || table[o->object.type] <= 0) + return sizeof(ObjectHeader); - return h1 == h2; + return table[o->object.type]; } int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Object **ret) { @@ -442,12 +346,15 @@ int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Objec void *t; Object *o; uint64_t s; + unsigned context; assert(f); assert(ret); - assert(type < _OBJECT_TYPE_MAX); - r = journal_file_move_to(f, type >= 0 ? type : WINDOW_UNKNOWN, offset, sizeof(ObjectHeader), &t); + /* One context for each type, plus one catch-all for the rest */ + context = type > 0 && type < _OBJECT_TYPE_MAX ? type : 0; + + r = journal_file_move_to(f, context, offset, sizeof(ObjectHeader), &t); if (r < 0) return r; @@ -457,6 +364,12 @@ int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Objec if (s < sizeof(ObjectHeader)) return -EBADMSG; + if (o->object.type <= OBJECT_UNUSED) + return -EBADMSG; + + if (s < minimum_header_size(o)) + return -EBADMSG; + if (type >= 0 && o->object.type != type) return -EBADMSG; @@ -468,9 +381,6 @@ int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Objec o = (Object*) t; } - if (!verify_hash(o)) - return -EBADMSG; - *ret = o; return 0; } @@ -480,7 +390,7 @@ static uint64_t journal_file_entry_seqnum(JournalFile *f, uint64_t *seqnum) { assert(f); - r = le64toh(f->header->tail_seqnum) + 1; + r = le64toh(f->header->tail_entry_seqnum) + 1; if (seqnum) { /* If an external seqnum counter was passed, we update @@ -493,21 +403,22 @@ static uint64_t journal_file_entry_seqnum(JournalFile *f, uint64_t *seqnum) { *seqnum = r; } - f->header->tail_seqnum = htole64(r); + f->header->tail_entry_seqnum = htole64(r); - if (f->header->head_seqnum == 0) - f->header->head_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, int type, uint64_t size, Object **ret, uint64_t *offset) { +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); @@ -613,7 +524,7 @@ static int journal_file_map_data_hash_table(JournalFile *f) { s = le64toh(f->header->data_hash_table_size); r = journal_file_move_to(f, - WINDOW_DATA_HASH_TABLE, + OBJECT_DATA_HASH_TABLE, p, s, &t); if (r < 0) @@ -634,7 +545,7 @@ static int journal_file_map_field_hash_table(JournalFile *f) { s = le64toh(f->header->field_hash_table_size); r = journal_file_move_to(f, - WINDOW_FIELD_HASH_TABLE, + OBJECT_FIELD_HASH_TABLE, p, s, &t); if (r < 0) @@ -862,7 +773,7 @@ uint64_t journal_file_entry_n_items(Object *o) { return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem); } -static uint64_t journal_file_entry_array_n_items(Object *o) { +uint64_t journal_file_entry_array_n_items(Object *o) { assert(o); assert(o->object.type == OBJECT_ENTRY_ARRAY); @@ -932,6 +843,9 @@ static int link_entry_into_array(JournalFile *f, o->entry_array.next_entry_array_offset = htole64(q); } + if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) + f->header->n_entry_arrays = htole64(le64toh(f->header->n_entry_arrays) + 1); + *idx = htole64(hidx + 1); return 0; @@ -1889,412 +1803,6 @@ int journal_file_move_to_entry_by_realtime_for_data( 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_seqnum, offsetof(Header, head_entry_realtime) - offsetof(Header, head_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; -} - void journal_file_dump(JournalFile *f) { Object *o; int r; @@ -2403,8 +1911,8 @@ void journal_file_print_header(JournalFile *f) { (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_seqnum), - (unsigned long long) le64toh(f->header->tail_seqnum), + (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))); @@ -2428,6 +1936,7 @@ int journal_file_open( bool compress, bool authenticate, JournalMetrics *metrics, + MMapCache *mmap_cache, JournalFile *template, JournalFile **ret) { @@ -2457,6 +1966,19 @@ int journal_file_open( 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) { r = -ENOMEM; @@ -2594,7 +2116,7 @@ int journal_file_rotate(JournalFile **f, bool compress, bool authenticate) { 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->tail_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); @@ -2605,7 +2127,7 @@ int journal_file_rotate(JournalFile **f, bool compress, bool authenticate) { old_file->header->state = STATE_ARCHIVED; - r = journal_file_open(old_file->path, old_file->flags, old_file->mode, compress, authenticate, NULL, 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; @@ -2619,6 +2141,7 @@ int journal_file_open_reliably( bool compress, bool authenticate, JournalMetrics *metrics, + MMapCache *mmap_cache, JournalFile *template, JournalFile **ret) { @@ -2626,7 +2149,8 @@ int journal_file_open_reliably( size_t l; char *p; - r = journal_file_open(fname, flags, mode, compress, authenticate, metrics, template, ret); + r = journal_file_open(fname, flags, mode, compress, authenticate, + metrics, mmap_cache, template, ret); if (r != -EBADMSG && /* corrupted */ r != -ENODATA && /* truncated */ r != -EHOSTDOWN && /* other machine */ @@ -2660,206 +2184,10 @@ int journal_file_open_reliably( log_warning("File %s corrupted or uncleanly shut down, renaming and replacing.", fname); - return journal_file_open(fname, flags, mode, compress, authenticate, metrics, template, ret); + return journal_file_open(fname, flags, mode, compress, authenticate, + metrics, mmap_cache, template, ret); } -struct vacuum_info { - off_t usage; - char *filename; - - uint64_t realtime; - sd_id128_t seqnum_id; - uint64_t seqnum; - - bool have_seqnum; -}; - -static int vacuum_compare(const void *_a, const void *_b) { - const struct vacuum_info *a, *b; - - a = _a; - b = _b; - - 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) - return 1; - else - return 0; - } - - if (a->realtime < b->realtime) - return -1; - else if (a->realtime > b->realtime) - return 1; - 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) { - DIR *d; - int r = 0; - struct vacuum_info *list = NULL; - unsigned n_list = 0, n_allocated = 0, i; - uint64_t sum = 0; - - assert(directory); - - if (max_use <= 0) - return 0; - - d = opendir(directory); - if (!d) - return -errno; - - for (;;) { - int k; - struct dirent buf, *de; - size_t q; - struct stat st; - char *p; - unsigned long long seqnum = 0, realtime; - sd_id128_t seqnum_id; - bool have_seqnum; - - k = readdir_r(d, &buf, &de); - if (k != 0) { - r = -k; - goto finish; - } - - if (!de) - break; - - 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 (endswith(de->d_name, ".journal")) { - - /* Vacuum archived files */ - - if (q < 1 + 32 + 1 + 16 + 1 + 16 + 8) - 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; - } - - 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; - } - - have_seqnum = false; - } else - continue; - - if (n_list >= n_allocated) { - struct vacuum_info *j; - - n_allocated = MAX(n_allocated * 2U, 8U); - j = realloc(list, n_allocated * sizeof(struct vacuum_info)); - if (!j) { - free(p); - r = -ENOMEM; - goto finish; - } - - list = j; - } - - list[n_list].filename = p; - 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 ++; - } - - if (n_list > 0) - qsort(list, n_list, sizeof(struct vacuum_info), vacuum_compare); - - for(i = 0; i < n_list; i++) { - struct statvfs ss; - - if (fstatvfs(dirfd(d), &ss) < 0) { - r = -errno; - goto finish; - } - - if (sum <= max_use && - (uint64_t) ss.f_bavail * (uint64_t) ss.f_bsize >= min_free) - break; - - if (unlinkat(dirfd(d), list[i].filename, 0) >= 0) { - 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); - } - -finish: - for (i = 0; i < n_list; i++) - free(list[i].filename); - - 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;