X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fjournal%2Fjournal-file.c;h=9efa82cbe094eb0a136b7a089934ef458b18cba1;hp=8f301b87f9c62d1ddd4c8e676bf07d3429db8e44;hb=361f9cbc7e6c7d3ac41892f5a5aebd1ef4049b95;hpb=8144056fa63291e5b25d91a48cd77b91e845eabf diff --git a/src/journal/journal-file.c b/src/journal/journal-file.c index 8f301b87f..9efa82cbe 100644 --- a/src/journal/journal-file.c +++ b/src/journal/journal-file.c @@ -32,8 +32,8 @@ #include "lookup3.h" #include "compress.h" -#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*16ULL) -#define DEFAULT_FIELD_HASH_TABLE_SIZE (2047ULL*16ULL) +#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) @@ -58,17 +58,24 @@ * size */ #define DEFAULT_KEEP_FREE (1024ULL*1024ULL) /* 1 MB */ -static const char signature[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' }; +/* n_data was the first entry we added after the initial file format design */ +#define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data)) #define ALIGN64(x) (((x) + 7ULL) & ~7ULL) +#define JOURNAL_HEADER_CONTAINS(h, field) \ + (le64toh((h)->header_size) >= offsetof(Header, field) + sizeof((h)->field)) + +static const char signature[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' }; + void journal_file_close(JournalFile *f) { int t; assert(f); if (f->header) { - if (f->writable) + /* Mark the file offline. Don't override the archived state if it already is set */ + if (f->writable && f->header->state == STATE_ONLINE) f->header->state = STATE_OFFLINE; munmap(f->header, PAGE_ALIGN(sizeof(Header))); @@ -107,7 +114,7 @@ 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_seqnum = template->header->tail_seqnum; } else h.seqnum_id = h.file_id; @@ -161,7 +168,8 @@ static int journal_file_verify_header(JournalFile *f) { return -EPROTONOSUPPORT; #endif - if (f->header->header_size != htole64(ALIGN64(sizeof(*(f->header))))) + /* 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->header_size) + le64toh(f->header->arena_size))) @@ -181,12 +189,15 @@ static int journal_file_verify_header(JournalFile *f) { state = f->header->state; - if (state == STATE_ONLINE) - log_debug("Journal file %s is already online. Assuming unclean closing. Ignoring.", f->path); - 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; + } } return 0; @@ -426,7 +437,7 @@ static uint64_t journal_file_seqnum(JournalFile *f, uint64_t *seqnum) { assert(f); - r = le64toh(f->header->seqnum) + 1; + r = le64toh(f->header->tail_seqnum) + 1; if (seqnum) { /* If an external seqnum counter was passed, we update @@ -439,10 +450,10 @@ static uint64_t journal_file_seqnum(JournalFile *f, uint64_t *seqnum) { *seqnum = r; } - f->header->seqnum = htole64(r); + f->header->tail_seqnum = htole64(r); - if (f->header->first_seqnum == 0) - f->header->first_seqnum = htole64(r); + if (f->header->head_seqnum == 0) + f->header->head_seqnum = htole64(r); return r; } @@ -499,7 +510,17 @@ static int journal_file_setup_data_hash_table(JournalFile *f) { assert(f); - s = DEFAULT_DATA_HASH_TABLE_SIZE; + /* We estimate that we need 1 hash table entry per 768 of + journal file and we want to make sure we never get beyond + 75% fill level. Calculate the hash table size for the + maximum file size based on these metrics. */ + + s = (f->metrics.max_size * 4 / 768 / 3) * sizeof(HashItem); + if (s < DEFAULT_DATA_HASH_TABLE_SIZE) + s = DEFAULT_DATA_HASH_TABLE_SIZE; + + log_info("Reserving %llu entries in hash table.", (unsigned long long) (s / sizeof(HashItem))); + r = journal_file_append_object(f, OBJECT_DATA_HASH_TABLE, offsetof(Object, hash_table.items) + s, @@ -596,7 +617,7 @@ static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, ui o->data.n_entries = 0; h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)); - p = le64toh(f->data_hash_table[h].head_hash_offset); + p = le64toh(f->data_hash_table[h].tail_hash_offset); if (p == 0) { /* Only entry in the hash table is easy */ f->data_hash_table[h].head_hash_offset = htole64(offset); @@ -613,6 +634,9 @@ static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, ui f->data_hash_table[h].tail_hash_offset = htole64(offset); + if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) + f->header->n_data = htole64(le64toh(f->header->n_data) + 1); + return 0; } @@ -1211,8 +1235,15 @@ static int generic_array_bisect(JournalFile *f, } } - if (k > n) + if (k > n) { + if (direction == DIRECTION_UP) { + i = n; + subtract_one = true; + goto found; + } + return 0; + } last_p = lp; @@ -1245,7 +1276,7 @@ found: *offset = p; if (idx) - *idx = t + i - (subtract_one ? 1 : 0); + *idx = t + i + (subtract_one ? -1 : 0); return 1; } @@ -1262,6 +1293,8 @@ static int generic_array_bisect_plus_one(JournalFile *f, uint64_t *idx) { int r; + bool step_back = false; + Object *o; assert(f); assert(test_object); @@ -1274,34 +1307,81 @@ static int generic_array_bisect_plus_one(JournalFile *f, r = test_object(f, extra, needle); if (r < 0) return r; - else if (r == TEST_FOUND) { - Object *o; - r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o); - if (r < 0) - return r; - - if (ret) - *ret = o; - - if (offset) - *offset = extra; + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; - if (idx) - *idx = 0; + /* if we are looking with DIRECTION_UP then we need to first + see if in the actual array there is a matching entry, and + return the last one of that. But if there isn't any we need + to return this one. Hence remember this, and return it + below. */ + if (r == TEST_LEFT) + step_back = direction == DIRECTION_UP; - return 1; - } else if (r == TEST_RIGHT) - return 0; + if (r == TEST_RIGHT) { + if (direction == DIRECTION_DOWN) + goto found; + else + return 0; + } r = generic_array_bisect(f, first, n-1, needle, test_object, direction, ret, offset, idx); - if (r > 0) + if (r == 0 && step_back) + goto found; + + if (r > 0 && idx) (*idx) ++; return r; + +found: + r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o); + if (r < 0) + return r; + + if (ret) + *ret = o; + + if (offset) + *offset = extra; + + if (idx) + *idx = 0; + + return 1; +} + +static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) { + assert(f); + assert(p > 0); + + if (p == needle) + return TEST_FOUND; + else if (p < needle) + return TEST_LEFT; + else + return TEST_RIGHT; +} + +int journal_file_move_to_entry_by_offset( + JournalFile *f, + uint64_t p, + direction_t direction, + Object **ret, + uint64_t *offset) { + + return generic_array_bisect(f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + p, + test_object_offset, + direction, + ret, offset, NULL); } + static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) { Object *o; int r; @@ -1399,16 +1479,17 @@ int journal_file_move_to_entry_by_monotonic( Object **ret, uint64_t *offset) { - char t[8+32+1] = "_BOOT_ID="; + char t[9+32+1] = "_BOOT_ID="; Object *o; int r; - sd_id128_to_string(boot_id, t + 8); + assert(f); + sd_id128_to_string(boot_id, t + 9); r = journal_file_find_data_object(f, t, strlen(t), &o, NULL); if (r < 0) return r; - else if (r == 0) + if (r == 0) return -ENOENT; return generic_array_bisect_plus_one(f, @@ -1421,18 +1502,6 @@ int journal_file_move_to_entry_by_monotonic( ret, offset, NULL); } -static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) { - assert(f); - assert(p > 0); - - if (p == needle) - return TEST_FOUND; - else if (p < needle) - return TEST_LEFT; - else - return TEST_RIGHT; -} - int journal_file_next_entry( JournalFile *f, Object *o, uint64_t p, @@ -1597,6 +1666,119 @@ int journal_file_next_entry_for_data( ret, offset); } +int journal_file_move_to_entry_by_offset_for_data( + JournalFile *f, + uint64_t data_offset, + uint64_t p, + direction_t direction, + Object **ret, uint64_t *offset) { + + int r; + Object *d; + + assert(f); + + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); + if (r < 0) + return r; + + return generic_array_bisect_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + p, + test_object_offset, + direction, + ret, offset, NULL); +} + +int journal_file_move_to_entry_by_monotonic_for_data( + JournalFile *f, + uint64_t data_offset, + sd_id128_t boot_id, + uint64_t monotonic, + direction_t direction, + Object **ret, uint64_t *offset) { + + char t[9+32+1] = "_BOOT_ID="; + Object *o, *d; + int r; + uint64_t b, z; + + assert(f); + + /* First, seek by time */ + sd_id128_to_string(boot_id, t + 9); + r = journal_file_find_data_object(f, t, strlen(t), &o, &b); + if (r < 0) + return r; + if (r == 0) + return -ENOENT; + + r = generic_array_bisect_plus_one(f, + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + le64toh(o->data.n_entries), + monotonic, + test_object_monotonic, + direction, + NULL, &z, NULL); + if (r <= 0) + return r; + + /* And now, continue seeking until we find an entry that + * exists in both bisection arrays */ + + for (;;) { + Object *qo; + uint64_t p, q; + + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); + if (r < 0) + return r; + + r = generic_array_bisect_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + z, + test_object_offset, + direction, + NULL, &p, NULL); + if (r <= 0) + return r; + + r = journal_file_move_to_object(f, OBJECT_DATA, b, &o); + if (r < 0) + return r; + + r = generic_array_bisect_plus_one(f, + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + le64toh(o->data.n_entries), + p, + test_object_offset, + direction, + &qo, &q, NULL); + + if (r <= 0) + return r; + + if (p == q) { + if (ret) + *ret = qo; + if (offset) + *offset = q; + + return 1; + } + + z = q; + } + + return 0; +} + int journal_file_move_to_entry_by_seqnum_for_data( JournalFile *f, uint64_t data_offset, @@ -1607,8 +1789,10 @@ int journal_file_move_to_entry_by_seqnum_for_data( Object *d; int r; + assert(f); + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); - if (r <= 0) + if (r < 0) return r; return generic_array_bisect_plus_one(f, @@ -1631,8 +1815,10 @@ int journal_file_move_to_entry_by_realtime_for_data( Object *d; int r; + assert(f); + r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); - if (r <= 0) + if (r < 0) return r; return generic_array_bisect_plus_one(f, @@ -1646,27 +1832,13 @@ int journal_file_move_to_entry_by_realtime_for_data( } void journal_file_dump(JournalFile *f) { - char a[33], b[33], c[33]; Object *o; int r; uint64_t p; assert(f); - printf("File Path: %s\n" - "File ID: %s\n" - "Machine ID: %s\n" - "Boot ID: %s\n" - "Arena size: %llu\n" - "Objects: %lu\n" - "Entries: %lu\n", - f->path, - sd_id128_to_string(f->header->file_id, a), - sd_id128_to_string(f->header->machine_id, b), - sd_id128_to_string(f->header->boot_id, c), - (unsigned long long) le64toh(f->header->arena_size), - (unsigned long) le64toh(f->header->n_objects), - (unsigned long) le64toh(f->header->n_entries)); + journal_file_print_header(f); p = le64toh(f->header->header_size); while (p != 0) { @@ -1722,10 +1894,73 @@ 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_SIGNED) ? " SIGNED" : "", + (f->header->compatible_flags & ~HEADER_COMPATIBLE_SIGNED) ? " ???" : "", + (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_seqnum), + (unsigned long long) le64toh(f->header->tail_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, + JournalMetrics *metrics, JournalFile *template, JournalFile **ret) { @@ -1752,10 +1987,8 @@ int journal_file_open( f->writable = (flags & O_ACCMODE) != O_RDONLY; f->prot = prot_from_flags(flags); - if (template) { - f->metrics = template->metrics; + if (template) f->compress = template->compress; - } f->path = strdup(fname); if (!f->path) { @@ -1787,7 +2020,7 @@ 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; } @@ -1806,6 +2039,12 @@ int journal_file_open( } 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; @@ -1869,7 +2108,7 @@ 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_seqnum), (unsigned long long) le64toh((*f)->header->tail_entry_realtime)); r = rename(old_file->path, p); @@ -1880,7 +2119,7 @@ int journal_file_rotate(JournalFile **f) { old_file->header->state = STATE_ARCHIVED; - r = journal_file_open(old_file->path, old_file->flags, old_file->mode, old_file, &new_file); + r = journal_file_open(old_file->path, old_file->flags, old_file->mode, NULL, old_file, &new_file); journal_file_close(old_file); *f = new_file; @@ -1891,6 +2130,7 @@ int journal_file_open_reliably( const char *fname, int flags, mode_t mode, + JournalMetrics *metrics, JournalFile *template, JournalFile **ret) { @@ -1898,7 +2138,7 @@ int journal_file_open_reliably( size_t l; char *p; - r = journal_file_open(fname, flags, mode, template, ret); + r = journal_file_open(fname, flags, mode, metrics, template, ret); if (r != -EBADMSG && /* corrupted */ r != -ENODATA && /* truncated */ r != -EHOSTDOWN && /* other machine */ @@ -1927,7 +2167,7 @@ int journal_file_open_reliably( log_warning("File %s corrupted, renaming and replacing.", fname); - return journal_file_open(fname, flags, mode, template, ret); + return journal_file_open(fname, flags, mode, metrics, template, ret); } struct vacuum_info { @@ -2149,9 +2389,6 @@ int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint6 ts.monotonic < le64toh(to->header->tail_entry_monotonic)) return -EINVAL; - if (ts.realtime < le64toh(to->header->tail_entry_realtime)) - return -EINVAL; - n = journal_file_entry_n_items(o); items = alloca(sizeof(EntryItem) * n); @@ -2284,3 +2521,110 @@ void journal_default_metrics(JournalMetrics *m, int fd) { 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; +}