#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"
#include "lookup3.h"
#include "compress.h"
+#include "fsprg.h"
#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 */
/* 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)
+/* How many entries to keep in the entry array chain cache at max */
+#define CHAIN_CACHE_MAX 20
-#define JOURNAL_HEADER_CONTAINS(h, field) \
- (le64toh((h)->header_size) >= offsetof(Header, field) + sizeof((h)->field))
+void journal_file_close(JournalFile *f) {
+ assert(f);
-static const char signature[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' };
+#ifdef HAVE_GCRYPT
+ /* Write the final tag */
+ if (f->seal && f->writable)
+ journal_file_append_tag(f);
+#endif
-void journal_file_close(JournalFile *f) {
- int t;
+ /* Sync everything to disk, before we mark the file offline */
+ if (f->mmap && f->fd >= 0)
+ mmap_cache_close_fd(f->mmap, f->fd);
- assert(f);
+ 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 */
munmap(f->header, PAGE_ALIGN(sizeof(Header)));
}
- for (t = 0; t < _WINDOW_MAX; t++)
- if (f->windows[t].ptr)
- munmap(f->windows[t].ptr, f->windows[t].size);
-
if (f->fd >= 0)
close_nointr_nofail(f->fd);
free(f->path);
+ if (f->mmap)
+ mmap_cache_unref(f->mmap);
+
+ hashmap_free_free(f->chain_cache);
+
#ifdef HAVE_XZ
free(f->compress_buffer);
#endif
+#ifdef HAVE_GCRYPT
+ if (f->fss_file)
+ munmap(f->fss_file, PAGE_ALIGN(f->fss_file_size));
+ else if (f->fsprg_state)
+ free(f->fsprg_state);
+
+ free(f->fsprg_seed);
+
+ if (f->hmac)
+ gcry_md_close(f->hmac);
+#endif
+
free(f);
}
assert(f);
zero(h);
- memcpy(h.signature, signature, 8);
+ memcpy(h.signature, HEADER_SIGNATURE, 8);
h.header_size = htole64(ALIGN64(sizeof(h)));
+ h.incompatible_flags =
+ htole32(f->compress ? HEADER_INCOMPATIBLE_COMPRESSED : 0);
+
+ h.compatible_flags =
+ htole32(f->seal ? HEADER_COMPATIBLE_SEALED : 0);
+
r = sd_id128_randomize(&h.file_id);
if (r < 0)
return r;
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;
f->header->state = STATE_ONLINE;
- __sync_synchronize();
+ /* Sync the online state to disk */
+ msync(f->header, PAGE_ALIGN(sizeof(Header)), MS_SYNC);
+ fdatasync(f->fd);
return 0;
}
static int journal_file_verify_header(JournalFile *f) {
assert(f);
- if (memcmp(f->header, signature, 8))
+ if (memcmp(f->header->signature, HEADER_SIGNATURE, 8))
return -EBADMSG;
+ /* In both read and write mode we refuse to open files with
+ * incompatible flags we don't know */
#ifdef HAVE_XZ
- if ((le64toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_COMPRESSED) != 0)
+ if ((le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_COMPRESSED) != 0)
return -EPROTONOSUPPORT;
#else
if (f->header->incompatible_flags != 0)
return -EPROTONOSUPPORT;
#endif
+ /* When open for writing we refuse to open files with
+ * compatible flags, too */
+ if (f->writable) {
+#ifdef HAVE_GCRYPT
+ if ((le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_SEALED) != 0)
+ return -EPROTONOSUPPORT;
+#else
+ if (f->header->compatible_flags != 0)
+ return -EPROTONOSUPPORT;
+#endif
+ }
+
+ if (f->header->state >= _STATE_MAX)
+ return -EBADMSG;
+
/* 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)))
+ 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)
+ return -ENODATA;
+
+ if (le64toh(f->header->tail_object_offset) > (le64toh(f->header->header_size) + le64toh(f->header->arena_size)))
+ return -ENODATA;
+
+ 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 = JOURNAL_HEADER_COMPRESSED(f->header);
+
+ f->seal = JOURNAL_HEADER_SEALED(f->header);
+
return 0;
}
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, bool keep_always, 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);
+ if (size <= 0)
+ return -EINVAL;
/* 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. */
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, keep_always, offset, size, &f->last_stat, 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) {
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);
+ /* Objects may only be located at multiple of 64 bit */
+ if (!VALID64(offset))
+ return -EFAULT;
+
+ /* 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, false, offset, sizeof(ObjectHeader), &t);
if (r < 0)
return r;
if (s < sizeof(ObjectHeader))
return -EBADMSG;
- if (type >= 0 && o->object.type != type)
+ if (o->object.type <= OBJECT_UNUSED)
+ return -EBADMSG;
+
+ if (s < minimum_header_size(o))
+ return -EBADMSG;
+
+ if (type > 0 && o->object.type != type)
return -EBADMSG;
if (s > sizeof(ObjectHeader)) {
- r = journal_file_move_to(f, o->object.type, offset, s, &t);
+ r = journal_file_move_to(f, o->object.type, false, offset, s, &t);
if (r < 0)
return r;
o = (Object*) t;
}
- if (!verify_hash(o))
- return -EBADMSG;
-
*ret = o;
return 0;
}
-static uint64_t journal_file_seqnum(JournalFile *f, uint64_t *seqnum) {
+static uint64_t journal_file_entry_seqnum(JournalFile *f, uint64_t *seqnum) {
uint64_t r;
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
*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);
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,
s = le64toh(f->header->data_hash_table_size);
r = journal_file_move_to(f,
- WINDOW_DATA_HASH_TABLE,
+ OBJECT_DATA_HASH_TABLE,
+ true,
p, s,
&t);
if (r < 0)
s = le64toh(f->header->field_hash_table_size);
r = journal_file_move_to(f,
- WINDOW_FIELD_HASH_TABLE,
+ 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,
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);
o->object.size = htole64(offsetof(Object, data.payload) + rsize);
o->object.flags |= OBJECT_COMPRESSED;
- f->header->incompatible_flags = htole32(le32toh(f->header->incompatible_flags) | HEADER_INCOMPATIBLE_COMPRESSED);
-
log_debug("Compressed data object %lu -> %lu", (unsigned long) size, (unsigned long) rsize);
}
}
#endif
- if (!compressed)
+ if (!compressed && size > 0)
memcpy(o->data.payload, data, size);
r = journal_file_link_data(f, o, p, hash);
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);
}
-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);
+
+ 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);
+
+ 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);
+}
+
static int link_entry_into_array(JournalFile *f,
le64_t *first,
le64_t *idx,
if (r < 0)
return r;
+#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);
if (ap == 0)
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;
assert(f);
assert(o);
assert(offset > 0);
- assert(o->object.type == OBJECT_ENTRY);
+
+ if (o->object.type != OBJECT_ENTRY)
+ return -EINVAL;
__sync_synchronize();
if (r < 0)
return r;
- o->entry.seqnum = htole64(journal_file_seqnum(f, seqnum));
+ o->entry.seqnum = htole64(journal_file_entry_seqnum(f, seqnum));
memcpy(o->entry.items, items, n_items * sizeof(EntryItem));
o->entry.realtime = htole64(ts->realtime);
o->entry.monotonic = htole64(ts->monotonic);
o->entry.xor_hash = htole64(xor_hash);
o->entry.boot_id = f->header->boot_id;
+#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)
return r;
__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) {
ts->monotonic < le64toh(f->header->tail_entry_monotonic))
return -EINVAL;
- items = alloca(sizeof(EntryItem) * n_iovec);
+#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));
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 (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)
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));
printf("Type: OBJECT_ENTRY_ARRAY\n");
break;
- case OBJECT_SIGNATURE:
- printf("Type: OBJECT_SIGNATURE\n");
+ 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));
+ 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);
"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",
+ "Tail Realtime Timestamp: %s\n"
+ "Objects: %llu\n"
+ "Entry Objects: %llu\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) ? " ???" : "",
+ f->header->state == STATE_OFFLINE ? "OFFLINE" :
+ f->header->state == STATE_ONLINE ? "ONLINE" :
+ f->header->state == STATE_ARCHIVED ? "ARCHIVED" : "UNKNOWN",
+ 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),
- (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),
+ 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)));
+ 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));
if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
printf("Data 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))));
+
+ if (JOURNAL_HEADER_CONTAINS(f->header, n_tags))
+ printf("Tag Objects: %llu\n",
+ (unsigned long long) 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));
+
+ 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(
const char *fname,
int flags,
mode_t mode,
+ bool compress,
+ bool seal,
JournalMetrics *metrics,
+ MMapCache *mmap_cache,
JournalFile *template,
JournalFile **ret) {
bool newly_created = false;
assert(fname);
+ assert(ret);
if ((flags & O_ACCMODE) != O_RDONLY &&
(flags & O_ACCMODE) != O_RDWR)
return -EINVAL;
- if (!endswith(fname, ".journal"))
+ if (!endswith(fname, ".journal") &&
+ !endswith(fname, ".journal~"))
return -EINVAL;
f = new0(JournalFile, 1);
return -ENOMEM;
f->fd = -1;
- f->flags = flags;
f->mode = mode;
- f->writable = (flags & O_ACCMODE) != O_RDONLY;
+
+ f->flags = flags;
f->prot = prot_from_flags(flags);
+ f->writable = (flags & O_ACCMODE) != O_RDONLY;
+#ifdef HAVE_XZ
+ f->compress = compress;
+#endif
+#ifdef HAVE_GCRYPT
+ f->seal = seal;
+#endif
- if (template)
- f->compress = template->compress;
+ if (mmap_cache)
+ f->mmap = mmap_cache_ref(mmap_cache);
+ else {
+ f->mmap = mmap_cache_new();
+ if (!f->mmap) {
+ r = -ENOMEM;
+ goto fail;
+ }
+ }
f->path = strdup(fname);
if (!f->path) {
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 */
+ 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) {
journal_default_metrics(metrics, f->fd);
goto fail;
}
- if (newly_created) {
+#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;
r = journal_file_setup_data_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:
return r;
}
-int journal_file_rotate(JournalFile **f) {
+int journal_file_rotate(JournalFile **f, bool compress, bool seal) {
char *p;
size_t l;
JournalFile *old_file, *new_file = NULL;
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_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);
old_file->header->state = STATE_ARCHIVED;
- r = journal_file_open(old_file->path, old_file->flags, old_file->mode, NULL, old_file, &new_file);
+ r = journal_file_open(old_file->path, old_file->flags, old_file->mode, compress, seal, NULL, old_file->mmap, old_file, &new_file);
journal_file_close(old_file);
*f = new_file;
const char *fname,
int flags,
mode_t mode,
+ bool compress,
+ bool seal,
JournalMetrics *metrics,
+ MMapCache *mmap_cache,
JournalFile *template,
JournalFile **ret) {
size_t l;
char *p;
- r = journal_file_open(fname, flags, mode, metrics, template, ret);
+ r = journal_file_open(fname, flags, mode, compress, seal,
+ metrics, mmap_cache, template, ret);
if (r != -EBADMSG && /* corrupted */
r != -ENODATA && /* truncated */
r != -EHOSTDOWN && /* other machine */
- r != -EPROTONOSUPPORT) /* incompatible feature */
+ r != -EPROTONOSUPPORT && /* incompatible feature */
+ r != -EBUSY && /* unclean shutdown */
+ r != -ESHUTDOWN /* already archived */)
return r;
if ((flags & O_ACCMODE) == O_RDONLY)
if (!(flags & O_CREAT))
return r;
+ if (!endswith(fname, ".journal"))
+ return r;
+
/* The file is corrupted. Rotate it away and try it again (but only once) */
l = strlen(fname);
if (r < 0)
return -errno;
- log_warning("File %s corrupted, renaming and replacing.", fname);
-
- return journal_file_open(fname, flags, mode, metrics, template, ret);
-}
-
-struct vacuum_info {
- off_t usage;
- char *filename;
-
- uint64_t realtime;
- sd_id128_t seqnum_id;
- uint64_t seqnum;
+ log_warning("File %s corrupted or uncleanly shut down, renaming and replacing.", fname);
- 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);
+ return journal_file_open(fname, flags, mode, compress, seal,
+ metrics, mmap_cache, template, ret);
}
-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 ++;
- }
-
- 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;
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) {
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,
* hence suggest a rotation */
- if (le64toh(f->header->header_size) < sizeof(Header))
+ 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
* 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)
+ 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)
+ 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;
+ }
+
+ /* 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