#include <fcntl.h>
#include <stddef.h>
+#ifdef HAVE_XATTR
+#include <attr/xattr.h>
+#endif
+
#include "journal-def.h"
#include "journal-file.h"
#include "journal-authenticate.h"
#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 */
/* n_data was the first entry we added after the initial file format design */
#define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data))
+/* 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) {
+ assert(f);
+
+ if (!f->writable)
+ return -EPERM;
+
+ if (!(f->fd >= 0 && f->header))
+ return -EINVAL;
+
+ switch(f->header->state) {
+ case STATE_ONLINE:
+ return 0;
+
+ case STATE_OFFLINE:
+ f->header->state = STATE_ONLINE;
+ fsync(f->fd);
+ return 0;
+
+ default:
+ return -EINVAL;
+ }
+}
+
+int journal_file_set_offline(JournalFile *f) {
+ assert(f);
+
+ if (!f->writable)
+ return -EPERM;
+
+ if (!(f->fd >= 0 && f->header))
+ return -EINVAL;
+
+ if (f->header->state != STATE_ONLINE)
+ return 0;
+
+ fsync(f->fd);
+
+ f->header->state = STATE_OFFLINE;
+
+ fsync(f->fd);
+
+ return 0;
+}
+
void journal_file_close(JournalFile *f) {
assert(f);
+#ifdef HAVE_GCRYPT
/* Write the final tag */
if (f->seal && f->writable)
journal_file_append_tag(f);
+#endif
/* Sync everything to disk, before we mark the file offline */
if (f->mmap && f->fd >= 0)
mmap_cache_close_fd(f->mmap, f->fd);
- if (f->writable && f->fd >= 0)
- fdatasync(f->fd);
-
- if (f->header) {
- /* 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;
+ journal_file_set_offline(f);
+ if (f->header)
munmap(f->header, PAGE_ALIGN(sizeof(Header)));
- }
if (f->fd >= 0)
close_nointr_nofail(f->fd);
if (f->mmap)
mmap_cache_unref(f->mmap);
+ hashmap_free_free(f->chain_cache);
+
#ifdef HAVE_XZ
free(f->compress_buffer);
#endif
f->header->boot_id = boot_id;
- f->header->state = STATE_ONLINE;
+ journal_file_set_online(f);
/* Sync the online state to disk */
msync(f->header, PAGE_ALIGN(sizeof(Header)), MS_SYNC);
if (le64toh(f->header->header_size) < HEADER_SIZE_MIN)
return -EBADMSG;
- if ((le32toh(f->header->compatible_flags) & HEADER_COMPATIBLE_SEALED) &&
- !JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays))
+ if (JOURNAL_HEADER_SEALED(f->header) && !JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays))
return -EBADMSG;
if ((le64toh(f->header->header_size) + le64toh(f->header->arena_size)) > (uint64_t) f->last_stat.st_size)
if (le64toh(f->header->tail_object_offset) > (le64toh(f->header->header_size) + le64toh(f->header->arena_size)))
return -ENODATA;
- if (!VALID64(f->header->data_hash_table_offset) ||
- !VALID64(f->header->field_hash_table_offset) ||
- !VALID64(f->header->tail_object_offset) ||
- !VALID64(f->header->entry_array_offset))
+ if (!VALID64(le64toh(f->header->data_hash_table_offset)) ||
+ !VALID64(le64toh(f->header->field_hash_table_offset)) ||
+ !VALID64(le64toh(f->header->tail_object_offset)) ||
+ !VALID64(le64toh(f->header->entry_array_offset)))
+ return -ENODATA;
+
+ if (le64toh(f->header->data_hash_table_offset) < le64toh(f->header->header_size) ||
+ le64toh(f->header->field_hash_table_offset) < le64toh(f->header->header_size) ||
+ le64toh(f->header->tail_object_offset) < le64toh(f->header->header_size) ||
+ le64toh(f->header->entry_array_offset) < le64toh(f->header->header_size))
return -ENODATA;
if (f->writable) {
}
}
- f->compress = !!(le32toh(f->header->incompatible_flags) & HEADER_INCOMPATIBLE_COMPRESSED);
+ f->compress = JOURNAL_HEADER_COMPRESSED(f->header);
- if (f->writable)
- f->seal = !!(le32toh(f->header->compatible_flags) & HEADER_COMPATIBLE_SEALED);
+ f->seal = JOURNAL_HEADER_SEALED(f->header);
return 0;
}
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;
return 0;
}
-static int journal_file_move_to(JournalFile *f, int context, uint64_t offset, uint64_t size, void **ret) {
+static int journal_file_move_to(JournalFile *f, int context, bool keep_always, uint64_t offset, uint64_t size, void **ret) {
assert(f);
assert(ret);
+ if (size <= 0)
+ return -EINVAL;
+
/* 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
return -EADDRNOTAVAIL;
}
- return mmap_cache_get(f->mmap, f->fd, f->prot, context, offset, size, ret);
+ return mmap_cache_get(f->mmap, f->fd, f->prot, context, keep_always, offset, size, &f->last_stat, ret);
}
static uint64_t minimum_header_size(Object *o) {
/* 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);
+ r = journal_file_move_to(f, context, false, offset, sizeof(ObjectHeader), &t);
if (r < 0)
return r;
if (s < minimum_header_size(o))
return -EBADMSG;
- if (type >= 0 && o->object.type != type)
+ if (type > 0 && o->object.type != type)
return -EBADMSG;
if (s > sizeof(ObjectHeader)) {
- r = journal_file_move_to(f, o->object.type, offset, s, &t);
+ r = journal_file_move_to(f, o->object.type, false, offset, s, &t);
if (r < 0)
return r;
assert(offset);
assert(ret);
+ r = journal_file_set_online(f);
+ if (r < 0)
+ return r;
+
p = le64toh(f->header->tail_object_offset);
if (p == 0)
p = le64toh(f->header->header_size);
if (r < 0)
return r;
- r = journal_file_move_to(f, type, p, size, &t);
+ r = journal_file_move_to(f, type, false, p, size, &t);
if (r < 0)
return r;
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)));
+ log_debug("Reserving %llu entries in hash table.", (unsigned long long) (s / sizeof(HashItem)));
r = journal_file_append_object(f,
OBJECT_DATA_HASH_TABLE,
assert(f);
+ /* We use a fixed size hash table for the fields as this
+ * number should grow very slowly only */
+
s = DEFAULT_FIELD_HASH_TABLE_SIZE;
r = journal_file_append_object(f,
OBJECT_FIELD_HASH_TABLE,
r = journal_file_move_to(f,
OBJECT_DATA_HASH_TABLE,
+ true,
p, s,
&t);
if (r < 0)
r = journal_file_move_to(f,
OBJECT_FIELD_HASH_TABLE,
+ true,
p, s,
&t);
if (r < 0)
return 0;
}
-static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, uint64_t hash) {
+static int journal_file_link_field(
+ 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 == OBJECT_DATA);
+
+ if (o->object.type != OBJECT_FIELD)
+ return -EINVAL;
+
+ /* This might alter the window we are looking at */
+
+ o->field.next_hash_offset = o->field.head_data_offset = 0;
+
+ h = hash % (le64toh(f->header->field_hash_table_size) / sizeof(HashItem));
+ p = le64toh(f->field_hash_table[h].tail_hash_offset);
+ if (p == 0)
+ f->field_hash_table[h].head_hash_offset = htole64(offset);
+ else {
+ r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o);
+ if (r < 0)
+ return r;
+
+ o->field.next_hash_offset = htole64(offset);
+ }
+
+ f->field_hash_table[h].tail_hash_offset = htole64(offset);
+
+ if (JOURNAL_HEADER_CONTAINS(f->header, n_fields))
+ f->header->n_fields = htole64(le64toh(f->header->n_fields) + 1);
+
+ return 0;
+}
+
+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);
+
+ if (o->object.type != OBJECT_DATA)
+ return -EINVAL;
/* This might alter the window we are looking at */
h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem));
p = le64toh(f->data_hash_table[h].tail_hash_offset);
- if (p == 0) {
+ if (p == 0)
/* Only entry in the hash table is easy */
f->data_hash_table[h].head_hash_offset = htole64(offset);
- } else {
+ else {
/* Move back to the previous data object, to patch in
* pointer */
return 0;
}
+int journal_file_find_field_object_with_hash(
+ JournalFile *f,
+ const void *field, uint64_t size, uint64_t hash,
+ Object **ret, uint64_t *offset) {
+
+ uint64_t p, osize, h;
+ int r;
+
+ assert(f);
+ assert(field && size > 0);
+
+ osize = offsetof(Object, field.payload) + size;
+
+ if (f->header->field_hash_table_size == 0)
+ return -EBADMSG;
+
+ h = hash % (le64toh(f->header->field_hash_table_size) / sizeof(HashItem));
+ p = le64toh(f->field_hash_table[h].head_hash_offset);
+
+ while (p > 0) {
+ Object *o;
+
+ r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o);
+ if (r < 0)
+ return r;
+
+ if (le64toh(o->field.hash) == hash &&
+ le64toh(o->object.size) == osize &&
+ memcmp(o->field.payload, field, size) == 0) {
+
+ if (ret)
+ *ret = o;
+ if (offset)
+ *offset = p;
+
+ return 1;
+ }
+
+ p = le64toh(o->field.next_hash_offset);
+ }
+
+ return 0;
+}
+
+int journal_file_find_field_object(
+ JournalFile *f,
+ const void *field, uint64_t size,
+ Object **ret, uint64_t *offset) {
+
+ uint64_t hash;
+
+ assert(f);
+ assert(field && size > 0);
+
+ hash = hash64(field, size);
+
+ return journal_file_find_field_object_with_hash(f,
+ field, size, hash,
+ ret, offset);
+}
+
int journal_file_find_data_object_with_hash(
JournalFile *f,
const void *data, uint64_t size, uint64_t hash,
l -= offsetof(Object, data.payload);
- if (!uncompress_blob(o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize))
+ if (!uncompress_blob(o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize, 0))
return -EBADMSG;
if (rsize == size &&
ret, offset);
}
+static int journal_file_append_field(
+ JournalFile *f,
+ const void *field, uint64_t size,
+ Object **ret, uint64_t *offset) {
+
+ uint64_t hash, p;
+ uint64_t osize;
+ Object *o;
+ int r;
+
+ assert(f);
+ assert(field && size > 0);
+
+ hash = hash64(field, size);
+
+ r = journal_file_find_field_object_with_hash(f, field, 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, field.payload) + size;
+ r = journal_file_append_object(f, OBJECT_FIELD, osize, &o, &p);
+
+ o->field.hash = htole64(hash);
+ memcpy(o->field.payload, field, size);
+
+ r = journal_file_link_field(f, o, p, hash);
+ 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_FIELD, p, &o);
+ if (r < 0)
+ return r;
+
+#ifdef HAVE_GCRYPT
+ r = journal_file_hmac_put_object(f, OBJECT_FIELD, o, p);
+ if (r < 0)
+ return r;
+#endif
+
+ if (ret)
+ *ret = o;
+
+ if (offset)
+ *offset = p;
+
+ return 0;
+}
+
static int journal_file_append_data(
JournalFile *f,
const void *data, uint64_t size,
Object *o;
int r;
bool compressed = false;
+ const void *eq;
assert(f);
assert(data || size == 0);
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;
+ eq = memchr(data, '=', size);
+ if (eq && eq > data) {
+ uint64_t fp;
+ Object *fo;
+
+ /* Create field object ... */
+ r = journal_file_append_field(f, data, (uint8_t*) eq - (uint8_t*) data, &fo, &fp);
+ if (r < 0)
+ return r;
+
+ /* ... and link it in. */
+ o->data.next_field_offset = fo->field.head_data_offset;
+ fo->field.head_data_offset = le64toh(p);
+ }
+
+#ifdef HAVE_GCRYPT
+ r = journal_file_hmac_put_object(f, OBJECT_DATA, o, p);
+ if (r < 0)
+ return r;
+#endif
+
if (ret)
*ret = o;
uint64_t journal_file_entry_n_items(Object *o) {
assert(o);
- assert(o->object.type == OBJECT_ENTRY);
+
+ if (o->object.type != OBJECT_ENTRY)
+ return 0;
return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem);
}
uint64_t journal_file_entry_array_n_items(Object *o) {
assert(o);
- assert(o->object.type == OBJECT_ENTRY_ARRAY);
+
+ if (o->object.type != OBJECT_ENTRY_ARRAY)
+ return 0;
return (le64toh(o->object.size) - offsetof(Object, entry_array.items)) / sizeof(uint64_t);
}
uint64_t journal_file_hash_table_n_items(Object *o) {
assert(o);
- assert(o->object.type == OBJECT_DATA_HASH_TABLE ||
- o->object.type == OBJECT_FIELD_HASH_TABLE);
+
+ if (o->object.type != OBJECT_DATA_HASH_TABLE &&
+ o->object.type != OBJECT_FIELD_HASH_TABLE)
+ return 0;
return (le64toh(o->object.size) - offsetof(Object, hash_table.items)) / sizeof(HashItem);
}
if (r < 0)
return r;
- r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, q);
+#ifdef HAVE_GCRYPT
+ r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, o, q);
if (r < 0)
return r;
+#endif
o->entry_array.items[i] = htole64(p);
assert(f);
assert(o);
assert(offset > 0);
- assert(o->object.type == OBJECT_ENTRY);
+
+ if (o->object.type != OBJECT_ENTRY)
+ return -EINVAL;
__sync_synchronize();
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);
+#ifdef HAVE_GCRYPT
+ r = journal_file_hmac_put_object(f, OBJECT_ENTRY, o, np);
if (r < 0)
return r;
+#endif
r = journal_file_link_entry(f, o, np);
if (r < 0)
__sync_synchronize();
if (ftruncate(f->fd, f->last_stat.st_size) < 0)
- log_error("Failed to to truncate file to its own size: %m");
+ log_error("Failed to truncate file to its own size: %m");
+}
+
+static int entry_item_cmp(const void *_a, const void *_b) {
+ const EntryItem *a = _a, *b = _b;
+
+ if (le64toh(a->object_offset) < le64toh(b->object_offset))
+ return -1;
+ if (le64toh(a->object_offset) > le64toh(b->object_offset))
+ return 1;
+ return 0;
}
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) {
assert(f);
assert(iovec || n_iovec == 0);
- if (!f->writable)
- return -EPERM;
-
if (!ts) {
dual_timestamp_get(&_ts);
ts = &_ts;
ts->monotonic < le64toh(f->header->tail_entry_monotonic))
return -EINVAL;
+#ifdef HAVE_GCRYPT
r = journal_file_maybe_append_tag(f, ts->realtime);
if (r < 0)
return r;
+#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;
items[i].hash = o->data.hash;
}
+ /* Order by the position on disk, in order to improve seek
+ * times for rotating media. */
+ qsort(items, n_iovec, sizeof(EntryItem), entry_item_cmp);
+
r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqnum, ret, offset);
journal_file_post_change(f);
return r;
}
+typedef struct ChainCacheItem {
+ uint64_t first; /* the array at the begin of the chain */
+ 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 */
+} ChainCacheItem;
+
+static void chain_cache_put(
+ Hashmap *h,
+ ChainCacheItem *ci,
+ uint64_t first,
+ uint64_t array,
+ uint64_t begin,
+ uint64_t total) {
+
+ if (!ci) {
+ /* If the chain item to cache for this chain is the
+ * first one it's not worth caching anything */
+ if (array == first)
+ return;
+
+ if (hashmap_size(h) >= CHAIN_CACHE_MAX)
+ ci = hashmap_steal_first(h);
+ else {
+ ci = new(ChainCacheItem, 1);
+ if (!ci)
+ return;
+ }
+
+ ci->first = first;
+
+ if (hashmap_put(h, &ci->first, ci) < 0) {
+ free(ci);
+ return;
+ }
+ } else
+ assert(ci->first == first);
+
+ ci->array = array;
+ ci->begin = begin;
+ ci->total = total;
+}
+
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;
+ uint64_t p = 0, a, t = 0;
int r;
+ ChainCacheItem *ci;
assert(f);
a = first;
+
+ /* Try the chain cache first */
+ ci = hashmap_get(f->chain_cache, &first);
+ if (ci && i > ci->total) {
+ a = ci->array;
+ i -= ci->total;
+ t = ci->total;
+ }
+
while (a > 0) {
- uint64_t n;
+ uint64_t k;
r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o);
if (r < 0)
return r;
- n = journal_file_entry_array_n_items(o);
- if (i < n) {
+ k = journal_file_entry_array_n_items(o);
+ if (i < k) {
p = le64toh(o->entry_array.items[i]);
- break;
+ goto found;
}
- i -= n;
+ i -= k;
+ t += k;
a = le64toh(o->entry_array.next_entry_array_offset);
}
- if (a <= 0 || p <= 0)
- return 0;
+ return 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);
r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
if (r < 0)
bool subtract_one = false;
Object *o, *array = NULL;
int r;
+ ChainCacheItem *ci;
assert(f);
assert(test_object);
+ /* Start with the first array in the chain */
a = first;
+
+ ci = hashmap_get(f->chain_cache, &first);
+ if (ci && n > ci->total) {
+ /* Ah, we have iterated this bisection array chain
+ * previously! Let's see if we can skip ahead in the
+ * chain, as far as the last time. But we can't jump
+ * backwards in the chain, so let's check that
+ * first. */
+
+ r = test_object(f, ci->begin, needle);
+ if (r < 0)
+ return r;
+
+ if (r == TEST_LEFT) {
+ /* OK, what we are looking for is right of th
+ * 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;
+ }
+ }
+
while (a > 0) {
uint64_t left, right, k, lp;
if (subtract_one && t == 0 && i == 0)
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);
+
if (subtract_one && i == 0)
p = last_p;
else if (subtract_one)
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)
printf("Type: OBJECT_DATA\n");
break;
+ case OBJECT_FIELD:
+ printf("Type: OBJECT_FIELD\n");
+ break;
+
case OBJECT_ENTRY:
- printf("Type: OBJECT_ENTRY %llu %llu %llu\n",
+ 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));
break;
case OBJECT_TAG:
- printf("Type: OBJECT_TAG %llu\n",
- (unsigned long long) le64toh(o->tag.seqnum));
+ printf("Type: OBJECT_TAG seqnum=%llu epoch=%llu\n",
+ (unsigned long long) le64toh(o->tag.seqnum),
+ (unsigned long long) le64toh(o->tag.epoch));
+ break;
+
+ default:
+ printf("Type: unknown (%u)\n", o->object.type);
break;
}
void journal_file_print_header(JournalFile *f) {
char a[33], b[33], c[33];
char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX];
+ struct stat st;
+ char bytes[FORMAT_BYTES_MAX];
assert(f);
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_SEALED) ? " SEALED" : "",
- (f->header->compatible_flags & ~HEADER_COMPATIBLE_SEALED) ? " ???" : "",
- (f->header->incompatible_flags & HEADER_INCOMPATIBLE_COMPRESSED) ? " COMPRESSED" : "",
- (f->header->incompatible_flags & ~HEADER_INCOMPATIBLE_COMPRESSED) ? " ???" : "",
+ JOURNAL_HEADER_SEALED(f->header) ? " SEALED" : "",
+ (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),
- yes_no(journal_file_rotate_suggested(f)),
+ 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)),
if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays))
printf("Entry Array Objects: %llu\n",
(unsigned long long) 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_open(
bool newly_created = false;
assert(fname);
+ assert(ret);
if ((flags & O_ACCMODE) != O_RDONLY &&
(flags & O_ACCMODE) != O_RDWR)
f->flags = flags;
f->prot = prot_from_flags(flags);
f->writable = (flags & O_ACCMODE) != O_RDONLY;
+#ifdef HAVE_XZ
f->compress = compress;
+#endif
+#ifdef HAVE_GCRYPT
f->seal = seal;
+#endif
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);
+ f->mmap = mmap_cache_new();
if (!f->mmap) {
r = -ENOMEM;
goto fail;
goto fail;
}
+ f->chain_cache = hashmap_new(uint64_hash_func, uint64_compare_func);
+ if (!f->chain_cache) {
+ r = -ENOMEM;
+ goto fail;
+ }
+
f->fd = open(f->path, f->flags|O_CLOEXEC, f->mode);
if (f->fd < 0) {
r = -errno;
}
if (f->last_stat.st_size == 0 && f->writable) {
- newly_created = true;
+#ifdef HAVE_XATTR
+ uint64_t crtime;
+
+ /* Let's attach the creation time to the journal file,
+ * so that the vacuuming code knows the age of this
+ * file even if the file might end up corrupted one
+ * day... Ideally we'd just use the creation time many
+ * file systems maintain for each file, but there is
+ * currently no usable API to query this, hence let's
+ * emulate this via extended attributes. If extended
+ * attributes are not supported we'll just skip this,
+ * and rely solely on mtime/atime/ctime of the file.*/
+
+ crtime = htole64((uint64_t) now(CLOCK_REALTIME));
+ fsetxattr(f->fd, "user.crtime_usec", &crtime, sizeof(crtime), XATTR_CREATE);
+#endif
+#ifdef HAVE_GCRYPT
/* Try to load the FSPRG state, and if we can't, then
* just don't do sealing */
- r = journal_file_fss_load(f);
- if (r < 0)
- f->seal = false;
+ if (f->seal) {
+ r = journal_file_fss_load(f);
+ if (r < 0)
+ f->seal = false;
+ }
+#endif
r = journal_file_init_header(f, template);
if (r < 0)
r = -errno;
goto fail;
}
+
+ newly_created = true;
}
if (f->last_stat.st_size < (off_t) HEADER_SIZE_MIN) {
goto fail;
}
+#ifdef HAVE_GCRYPT
if (!newly_created && f->writable) {
r = journal_file_fss_load(f);
if (r < 0)
goto fail;
}
+#endif
if (f->writable) {
if (metrics) {
goto fail;
}
+#ifdef HAVE_GCRYPT
r = journal_file_hmac_setup(f);
if (r < 0)
goto fail;
+#endif
if (newly_created) {
r = journal_file_setup_field_hash_table(f);
if (r < 0)
goto fail;
+#ifdef HAVE_GCRYPT
r = journal_file_append_first_tag(f);
if (r < 0)
goto fail;
+#endif
}
r = journal_file_map_field_hash_table(f);
if (r < 0)
goto fail;
- if (ret)
- *ret = f;
-
+ *ret = f;
return 0;
fail:
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_entry_seqnum),
- (unsigned long long) le64toh((*f)->header->tail_entry_realtime));
+ (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);
metrics, mmap_cache, template, ret);
}
-
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;
#ifdef HAVE_XZ
uint64_t rsize;
- if (!uncompress_blob(o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize))
+ if (!uncompress_blob(o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize, 0))
return -EBADMSG;
data = from->compress_buffer;
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;
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));
+ log_debug("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) {
}
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;
return 1;
}
-bool journal_file_rotate_suggested(JournalFile *f) {
+bool journal_file_rotate_suggested(JournalFile *f, usec_t max_file_usec) {
assert(f);
/* If we gained new header fields we gained new features,
return true;
}
+ /* Are the data objects properly indexed by field objects? */
+ if (JOURNAL_HEADER_CONTAINS(f->header, n_data) &&
+ JOURNAL_HEADER_CONTAINS(f->header, n_fields) &&
+ le64toh(f->header->n_data) > 0 &&
+ le64toh(f->header->n_fields) == 0)
+ return true;
+
+ if (max_file_usec > 0) {
+ usec_t t, h;
+
+ h = le64toh(f->header->head_entry_realtime);
+ t = now(CLOCK_REALTIME);
+
+ if (h > 0 && t > h + max_file_usec)
+ return true;
+ }
+
return false;
}
~ianmdlvl / elogind.git / blobdiff