#define COMPRESSION_SIZE_THRESHOLD (512ULL)
/* This is the minimum journal file size */
-#define JOURNAL_FILE_SIZE_MIN (64ULL*1024ULL) /* 64 KiB */
+#define JOURNAL_FILE_SIZE_MIN (4ULL*1024ULL*1024ULL) /* 4 MiB */
/* These are the lower and upper bounds if we deduce the max_use value
* from the file system size */
/* How many entries to keep in the entry array chain cache at max */
#define CHAIN_CACHE_MAX 20
-int journal_file_set_online(JournalFile *f) {
+/* How much to increase the journal file size at once each time we allocate something new. */
+#define FILE_SIZE_INCREASE (8ULL*1024ULL*1024ULL) /* 8MB */
+
+static int journal_file_set_online(JournalFile *f) {
assert(f);
if (!f->writable)
journal_file_set_online(f);
/* Sync the online state to disk */
- msync(f->header, PAGE_ALIGN(sizeof(Header)), MS_SYNC);
- fdatasync(f->fd);
+ fsync(f->fd);
return 0;
}
if (new_size <= old_size)
return 0;
- if (f->metrics.max_size > 0 &&
- new_size > f->metrics.max_size)
+ if (f->metrics.max_size > 0 && new_size > f->metrics.max_size)
return -E2BIG;
- if (new_size > f->metrics.min_size &&
- f->metrics.keep_free > 0) {
+ if (new_size > f->metrics.min_size && f->metrics.keep_free > 0) {
struct statvfs svfs;
if (fstatvfs(f->fd, &svfs) >= 0) {
}
}
+ /* Increase by larger blocks at once */
+ new_size = ((new_size+FILE_SIZE_INCREASE-1) / FILE_SIZE_INCREASE) * FILE_SIZE_INCREASE;
+ if (f->metrics.max_size > 0 && new_size > f->metrics.max_size)
+ new_size = f->metrics.max_size;
+
/* Note that the glibc fallocate() fallback is very
inefficient, hence we try to minimize the allocation area
as we can. */
static uint64_t minimum_header_size(Object *o) {
- static uint64_t table[] = {
+ static const uint64_t table[] = {
[OBJECT_DATA] = sizeof(DataObject),
[OBJECT_FIELD] = sizeof(FieldObject),
[OBJECT_ENTRY] = sizeof(EntryObject),
if (s < DEFAULT_DATA_HASH_TABLE_SIZE)
s = DEFAULT_DATA_HASH_TABLE_SIZE;
- log_debug("Reserving %llu entries in hash table.", (unsigned long long) (s / sizeof(HashItem)));
+ log_debug("Reserving %"PRIu64" entries in hash table.", s / sizeof(HashItem));
r = journal_file_append_object(f,
OBJECT_DATA_HASH_TABLE,
osize = offsetof(Object, field.payload) + size;
r = journal_file_append_object(f, OBJECT_FIELD, osize, &o, &p);
+ if (r < 0)
+ return r;
o->field.hash = htole64(hash);
memcpy(o->field.payload, field, size);
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);
+ log_debug("Compressed data object %"PRIu64" -> %"PRIu64, size, rsize);
}
}
#endif
if (r < 0)
return r;
- eq = memchr(data, '=', size);
+ if (!data)
+ eq = NULL;
+ else
+ eq = memchr(data, '=', size);
if (eq && eq > data) {
uint64_t fp;
Object *fo;
if (r < 0)
return r;
- /* log_debug("=> %s seqnr=%lu n_entries=%lu", f->path, (unsigned long) o->entry.seqnum, (unsigned long) f->header->n_entries); */
+ /* log_debug("=> %s seqnr=%"PRIu64" n_entries=%"PRIu64, f->path, o->entry.seqnum, f->header->n_entries); */
if (f->header->head_entry_realtime == 0)
f->header->head_entry_realtime = o->entry.realtime;
#endif
/* alloca() can't take 0, hence let's allocate at least one */
- items = alloca(sizeof(EntryItem) * MAX(1, n_iovec));
+ items = alloca(sizeof(EntryItem) * MAX(1u, n_iovec));
for (i = 0; i < n_iovec; i++) {
uint64_t p;
/* Order by the position on disk, in order to improve seek
* times for rotating media. */
- qsort(items, n_iovec, sizeof(EntryItem), entry_item_cmp);
+ qsort_safe(items, n_iovec, sizeof(EntryItem), entry_item_cmp);
r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqnum, ret, offset);
uint64_t array; /* the cached array */
uint64_t begin; /* the first item in the cached array */
uint64_t total; /* the total number of items in all arrays before this one in the chain */
+ uint64_t last_index; /* the last index we looked at, to optimize locality when bisecting */
} ChainCacheItem;
static void chain_cache_put(
uint64_t first,
uint64_t array,
uint64_t begin,
- uint64_t total) {
+ uint64_t total,
+ uint64_t last_index) {
if (!ci) {
/* If the chain item to cache for this chain is the
ci->array = array;
ci->begin = begin;
ci->total = total;
+ ci->last_index = last_index;
}
-static int generic_array_get(JournalFile *f,
- uint64_t first,
- uint64_t i,
- 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 p = 0, a, t = 0;
found:
/* Let's cache this item for the next invocation */
- chain_cache_put(f->chain_cache, ci, first, a, o->entry_array.items[0], t);
+ chain_cache_put(f->chain_cache, ci, first, a, o->entry_array.items[0], t, i);
r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
if (r < 0)
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) {
+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;
TEST_RIGHT
};
-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;
+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, last_index = (uint64_t) -1;
bool subtract_one = false;
Object *o, *array = NULL;
int r;
return r;
if (r == TEST_LEFT) {
- /* OK, what we are looking for is right of th
+ /* OK, what we are looking for is right of the
* begin of this EntryArray, so let's jump
* straight to previously cached array in the
* chain */
a = ci->array;
n -= ci->total;
t = ci->total;
+ last_index = ci->last_index;
}
}
if (r == TEST_RIGHT) {
left = 0;
right -= 1;
+
+ if (last_index != (uint64_t) -1) {
+ assert(last_index <= right);
+
+ /* If we cached the last index we
+ * looked at, let's try to not to jump
+ * too wildly around and see if we can
+ * limit the range to look at early to
+ * the immediate neighbors of the last
+ * index we looked at. */
+
+ if (last_index > 0) {
+ uint64_t x = last_index - 1;
+
+ p = le64toh(array->entry_array.items[x]);
+ 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 = x;
+ else
+ left = x + 1;
+ }
+
+ if (last_index < right) {
+ uint64_t y = last_index + 1;
+
+ p = le64toh(array->entry_array.items[y]);
+ 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 = y;
+ else
+ left = y + 1;
+ }
+ }
+
for (;;) {
if (left == right) {
if (direction == DIRECTION_UP)
}
assert(left < right);
-
i = (left + right) / 2;
+
p = le64toh(array->entry_array.items[i]);
if (p <= 0)
return -EBADMSG;
n -= k;
t += k;
+ last_index = (uint64_t) -1;
a = le64toh(array->entry_array.next_entry_array_offset);
}
return 0;
/* Let's cache this item for the next invocation */
- chain_cache_put(f->chain_cache, ci, first, a, array->entry_array.items[0], t);
+ chain_cache_put(f->chain_cache, ci, first, a, array->entry_array.items[0], t, subtract_one ? (i > 0 ? i-1 : (uint64_t) -1) : i);
if (subtract_one && i == 0)
p = last_p;
return 1;
}
-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) {
+
+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;
return 1;
}
-static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) {
+_pure_ static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) {
assert(f);
assert(p > 0);
return TEST_RIGHT;
}
+static inline int find_data_object_by_boot_id(
+ JournalFile *f,
+ sd_id128_t boot_id,
+ Object **o,
+ uint64_t *b) {
+ char t[sizeof("_BOOT_ID=")-1 + 32 + 1] = "_BOOT_ID=";
+
+ sd_id128_to_string(boot_id, t + 9);
+ return journal_file_find_data_object(f, t, sizeof(t) - 1, o, b);
+}
+
int journal_file_move_to_entry_by_monotonic(
JournalFile *f,
sd_id128_t boot_id,
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);
+ r = find_data_object_by_boot_id(f, boot_id, &o, NULL);
if (r < 0)
return r;
if (r == 0)
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);
+ r = find_data_object_by_boot_id(f, boot_id, &o, &b);
if (r < 0)
return r;
if (r == 0)
break;
case OBJECT_ENTRY:
- printf("Type: OBJECT_ENTRY seqnum=%llu monotonic=%llu realtime=%llu\n",
- (unsigned long long) le64toh(o->entry.seqnum),
- (unsigned long long) le64toh(o->entry.monotonic),
- (unsigned long long) le64toh(o->entry.realtime));
+ printf("Type: OBJECT_ENTRY seqnum=%"PRIu64" monotonic=%"PRIu64" realtime=%"PRIu64"\n",
+ le64toh(o->entry.seqnum),
+ le64toh(o->entry.monotonic),
+ le64toh(o->entry.realtime));
break;
case OBJECT_FIELD_HASH_TABLE:
break;
case OBJECT_TAG:
- printf("Type: OBJECT_TAG seqnum=%llu epoch=%llu\n",
- (unsigned long long) le64toh(o->tag.seqnum),
- (unsigned long long) le64toh(o->tag.epoch));
+ printf("Type: OBJECT_TAG seqnum=%"PRIu64" epoch=%"PRIu64"\n",
+ le64toh(o->tag.seqnum),
+ le64toh(o->tag.epoch));
break;
default:
log_error("File corrupt");
}
+static const char* format_timestamp_safe(char *buf, size_t l, usec_t t) {
+ const char *x;
+
+ x = format_timestamp(buf, l, t);
+ if (x)
+ return x;
+ return " --- ";
+}
+
void journal_file_print_header(JournalFile *f) {
- char a[33], b[33], c[33];
- char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX];
+ char a[33], b[33], c[33], d[33];
+ char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX], z[FORMAT_TIMESTAMP_MAX];
struct stat st;
char bytes[FORMAT_BYTES_MAX];
"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"
+ "Header size: %"PRIu64"\n"
+ "Arena size: %"PRIu64"\n"
+ "Data Hash Table Size: %"PRIu64"\n"
+ "Field Hash Table Size: %"PRIu64"\n"
"Rotate Suggested: %s\n"
- "Head Sequential Number: %llu\n"
- "Tail Sequential Number: %llu\n"
+ "Head Sequential Number: %"PRIu64"\n"
+ "Tail Sequential Number: %"PRIu64"\n"
"Head Realtime Timestamp: %s\n"
"Tail Realtime Timestamp: %s\n"
- "Objects: %llu\n"
- "Entry Objects: %llu\n",
+ "Tail Monotonic Timestamp: %s\n"
+ "Objects: %"PRIu64"\n"
+ "Entry Objects: %"PRIu64"\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),
+ sd_id128_to_string(f->header->seqnum_id, d),
f->header->state == STATE_OFFLINE ? "OFFLINE" :
f->header->state == STATE_ONLINE ? "ONLINE" :
f->header->state == STATE_ARCHIVED ? "ARCHIVED" : "UNKNOWN",
(le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_SEALED) ? " ???" : "",
JOURNAL_HEADER_COMPRESSED(f->header) ? " COMPRESSED" : "",
(le32toh(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),
+ le64toh(f->header->header_size),
+ le64toh(f->header->arena_size),
+ le64toh(f->header->data_hash_table_size) / sizeof(HashItem),
+ le64toh(f->header->field_hash_table_size) / sizeof(HashItem),
yes_no(journal_file_rotate_suggested(f, 0)),
- (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)),
- (unsigned long long) le64toh(f->header->n_objects),
- (unsigned long long) le64toh(f->header->n_entries));
+ le64toh(f->header->head_entry_seqnum),
+ le64toh(f->header->tail_entry_seqnum),
+ format_timestamp_safe(x, sizeof(x), le64toh(f->header->head_entry_realtime)),
+ format_timestamp_safe(y, sizeof(y), le64toh(f->header->tail_entry_realtime)),
+ format_timespan(z, sizeof(z), le64toh(f->header->tail_entry_monotonic), USEC_PER_MSEC),
+ le64toh(f->header->n_objects),
+ le64toh(f->header->n_entries));
if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
- printf("Data Objects: %llu\n"
+ printf("Data Objects: %"PRIu64"\n"
"Data Hash Table Fill: %.1f%%\n",
- (unsigned long long) le64toh(f->header->n_data),
+ 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"
+ printf("Field Objects: %"PRIu64"\n"
"Field Hash Table Fill: %.1f%%\n",
- (unsigned long long) le64toh(f->header->n_fields),
+ le64toh(f->header->n_fields),
100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))));
if (JOURNAL_HEADER_CONTAINS(f->header, n_tags))
- printf("Tag Objects: %llu\n",
- (unsigned long long) le64toh(f->header->n_tags));
+ printf("Tag Objects: %"PRIu64"\n",
+ le64toh(f->header->n_tags));
if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays))
- printf("Entry Array Objects: %llu\n",
- (unsigned long long) le64toh(f->header->n_entry_arrays));
+ printf("Entry Array Objects: %"PRIu64"\n",
+ le64toh(f->header->n_entry_arrays));
if (fstat(f->fd, &st) >= 0)
printf("Disk usage: %s\n", format_bytes(bytes, sizeof(bytes), (off_t) st.st_blocks * 512ULL));
}
int journal_file_rotate(JournalFile **f, bool compress, bool seal) {
- char *p;
+ _cleanup_free_ char *p = NULL;
size_t l;
JournalFile *old_file, *new_file = NULL;
int r;
return -EINVAL;
l = strlen(old_file->path);
-
- p = new(char, l + 1 + 32 + 1 + 16 + 1 + 16 + 1);
- if (!p)
+ r = asprintf(&p, "%.*s@" SD_ID128_FORMAT_STR "-%016"PRIx64"-%016"PRIx64".journal",
+ (int) l - 8, old_file->path,
+ SD_ID128_FORMAT_VAL(old_file->header->seqnum_id),
+ le64toh((*f)->header->head_entry_seqnum),
+ le64toh((*f)->header->head_entry_realtime));
+ if (r < 0)
return -ENOMEM;
- memcpy(p, old_file->path, l - 8);
- p[l-8] = '@';
- 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->head_entry_seqnum),
- (unsigned long long) le64toh((*f)->header->head_entry_realtime));
-
r = rename(old_file->path, p);
- free(p);
-
if (r < 0)
return -errno;
int r;
size_t l;
- char *p;
+ _cleanup_free_ char *p = NULL;
r = journal_file_open(fname, flags, mode, compress, seal,
metrics, mmap_cache, template, ret);
l = strlen(fname);
if (asprintf(&p, "%.*s@%016llx-%016llx.journal~",
- (int) (l-8), fname,
+ (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;
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);
+ /* alloca() can't take 0, hence let's allocate at least one */
+ items = alloca(sizeof(EntryItem) * MAX(1u, n));
for (i = 0; i < n; i++) {
uint64_t l, h;
if (m->keep_free == (uint64_t) -1) {
if (fs_size > 0) {
- m->keep_free = PAGE_ALIGN(fs_size / 20); /* 5% of file system size */
+ m->keep_free = PAGE_ALIGN(fs_size * 3 / 20); /* 15% of file system size */
if (m->keep_free > DEFAULT_KEEP_FREE_UPPER)
m->keep_free = DEFAULT_KEEP_FREE_UPPER;
}
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);
+ r = find_data_object_by_boot_id(f, boot_id, &o, &p);
if (r <= 0)
return r;
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.",
+ log_debug("Data hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items, %llu file size, %"PRIu64" 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)));
+ le64toh(f->header->n_data),
+ le64toh(f->header->data_hash_table_size) / sizeof(HashItem),
+ (unsigned long long) f->last_stat.st_size,
+ 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.",
+ log_debug("Field hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" 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)));
+ le64toh(f->header->n_fields),
+ le64toh(f->header->field_hash_table_size) / sizeof(HashItem));
return true;
}
~ianmdlvl / elogind.git / blobdiff