#include "journal-def.h"
#include "journal-file.h"
#include "lookup3.h"
+#include "compress.h"
-#define DEFAULT_ARENA_MAX_SIZE (16ULL*1024ULL*1024ULL*1024ULL)
-#define DEFAULT_ARENA_MIN_SIZE (256ULL*1024ULL)
-#define DEFAULT_ARENA_KEEP_FREE (1ULL*1024ULL*1024ULL)
+#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*16ULL)
+#define DEFAULT_FIELD_HASH_TABLE_SIZE (2047ULL*16ULL)
-#define DEFAULT_MAX_USE (16ULL*1024ULL*1024ULL*16ULL)
+#define DEFAULT_WINDOW_SIZE (128ULL*1024ULL*1024ULL)
-#define DEFAULT_HASH_TABLE_SIZE (2047ULL*16ULL)
-#define DEFAULT_BISECT_TABLE_SIZE ((DEFAULT_ARENA_MAX_SIZE/(64ULL*1024ULL))*8ULL)
+#define COMPRESSION_SIZE_THRESHOLD (64ULL)
-#define DEFAULT_WINDOW_SIZE (128ULL*1024ULL*1024ULL)
+/* This is the minimum journal file size */
+#define JOURNAL_FILE_SIZE_MIN (64ULL*1024ULL)
+
+/* These are the lower and upper bounds if we deduce the max_use value
+ * from the file system size */
+#define DEFAULT_MAX_USE_LOWER (1ULL*1024ULL*1024ULL) /* 1 MiB */
+#define DEFAULT_MAX_USE_UPPER (4ULL*1024ULL*1024ULL*1024ULL) /* 4 GiB */
+
+/* This is the upper bound if we deduce max_size from max_use */
+#define DEFAULT_MAX_SIZE_UPPER (16ULL*1024ULL*1024ULL) /* 16 MiB */
+
+/* This is the upper bound if we deduce the keep_free value from the
+ * file system size */
+#define DEFAULT_KEEP_FREE_UPPER (4ULL*1024ULL*1024ULL*1024ULL) /* 4 GiB */
+
+/* This is the keep_free value when we can't determine the system
+ * size */
+#define DEFAULT_KEEP_FREE (1024ULL*1024ULL) /* 1 MB */
static const char signature[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' };
#define ALIGN64(x) (((x) + 7ULL) & ~7ULL)
void journal_file_close(JournalFile *f) {
- assert(f);
-
- if (f->header) {
- if (f->writable && f->header->state == htole32(STATE_ONLINE))
- f->header->state = htole32(STATE_OFFLINE);
+ int t;
- munmap(f->header, PAGE_ALIGN(sizeof(Header)));
- }
+ assert(f);
- if (f->hash_table_window)
- munmap(f->hash_table_window, f->hash_table_window_size);
+ if (f->header && f->writable)
+ f->header->state = STATE_OFFLINE;
- if (f->bisect_table_window)
- munmap(f->bisect_table_window, f->bisect_table_window_size);
- if (f->window)
- munmap(f->window, f->window_size);
+ 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);
+
+#ifdef HAVE_XZ
+ free(f->compress_buffer);
+#endif
+
free(f);
}
zero(h);
memcpy(h.signature, signature, 8);
h.arena_offset = htole64(ALIGN64(sizeof(h)));
- h.arena_max_size = htole64(DEFAULT_ARENA_MAX_SIZE);
- h.arena_min_size = htole64(DEFAULT_ARENA_MIN_SIZE);
- h.arena_keep_free = htole64(DEFAULT_ARENA_KEEP_FREE);
r = sd_id128_randomize(&h.file_id);
if (r < 0)
static int journal_file_refresh_header(JournalFile *f) {
int r;
+ sd_id128_t boot_id;
assert(f);
if (r < 0)
return r;
- r = sd_id128_get_boot(&f->header->boot_id);
+ r = sd_id128_get_boot(&boot_id);
if (r < 0)
return r;
- f->header->state = htole32(STATE_ONLINE);
+ if (sd_id128_equal(boot_id, f->header->boot_id))
+ f->tail_entry_monotonic_valid = true;
+
+ f->header->boot_id = boot_id;
+
+ f->header->state = STATE_ONLINE;
return 0;
}
if (memcmp(f->header, signature, 8))
return -EBADMSG;
+#ifdef HAVE_XZ
+ if ((le64toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_COMPRESSED) != 0)
+ return -EPROTONOSUPPORT;
+#else
if (f->header->incompatible_flags != 0)
return -EPROTONOSUPPORT;
+#endif
if ((uint64_t) f->last_stat.st_size < (le64toh(f->header->arena_offset) + le64toh(f->header->arena_size)))
return -ENODATA;
if (!sd_id128_equal(machine_id, f->header->machine_id))
return -EHOSTDOWN;
- state = le32toh(f->header->state);
+ state = f->header->state;
if (state == STATE_ONLINE)
log_debug("Journal file %s is already online. Assuming unclean closing. Ignoring.", f->path);
}
static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) {
- uint64_t asize;
uint64_t old_size, new_size;
assert(f);
- if (offset < le64toh(f->header->arena_offset))
- return -EINVAL;
-
- new_size = PAGE_ALIGN(offset + size);
-
/* We assume that this file is not sparse, and we know that
* for sure, since we always call posix_fallocate()
* ourselves */
le64toh(f->header->arena_offset) +
le64toh(f->header->arena_size);
- if (old_size >= new_size)
+ new_size = PAGE_ALIGN(offset + size);
+ if (new_size < le64toh(f->header->arena_offset))
+ new_size = le64toh(f->header->arena_offset);
+
+ if (new_size <= old_size)
return 0;
- asize = new_size - le64toh(f->header->arena_offset);
+ if (f->metrics.max_size > 0 &&
+ new_size > f->metrics.max_size)
+ return -E2BIG;
- if (asize > le64toh(f->header->arena_min_size)) {
+ if (new_size > f->metrics.min_size &&
+ f->metrics.keep_free > 0) {
struct statvfs svfs;
if (fstatvfs(f->fd, &svfs) >= 0) {
available = svfs.f_bfree * svfs.f_bsize;
- if (available >= f->header->arena_keep_free)
- available -= f->header->arena_keep_free;
+ if (available >= f->metrics.keep_free)
+ available -= f->metrics.keep_free;
else
available = 0;
}
}
- if (asize > le64toh(f->header->arena_max_size))
- return -E2BIG;
-
+ /* Note that the glibc fallocate() fallback is very
+ inefficient, hence we try to minimize the allocation area
+ as we can. */
if (posix_fallocate(f->fd, old_size, new_size - old_size) < 0)
return -errno;
if (fstat(f->fd, &f->last_stat) < 0)
return -errno;
- f->header->arena_size = htole64(asize);
+ f->header->arena_size = new_size - htole64(f->header->arena_offset);
return 0;
}
wsize = size + (offset - woffset);
wsize = PAGE_ALIGN(wsize);
+ /* Avoid SIGBUS on invalid accesses */
+ if (woffset + wsize > (uint64_t) PAGE_ALIGN(f->last_stat.st_size))
+ return -EADDRNOTAVAIL;
+
window = mmap(NULL, wsize, f->prot, MAP_SHARED, f->fd, woffset);
if (window == MAP_FAILED)
return -errno;
return 0;
}
-static int journal_file_move_to(JournalFile *f, uint64_t offset, uint64_t size, void **ret) {
- void *p;
+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);
+
+ w = f->windows + wt;
- if (_likely_(f->window &&
- f->window_offset <= offset &&
- f->window_offset+f->window_size >= offset + size)) {
+ if (_likely_(w->ptr &&
+ w->offset <= offset &&
+ w->offset + w->size >= offset + size)) {
- *ret = (uint8_t*) f->window + (offset - f->window_offset);
+ *ret = (uint8_t*) w->ptr + (offset - w->offset);
return 0;
}
- if (f->window) {
- if (munmap(f->window, f->window_size) < 0)
+ if (w->ptr) {
+ if (munmap(w->ptr, w->size) < 0)
return -errno;
- f->window = NULL;
- f->window_size = f->window_offset = 0;
+ w->ptr = NULL;
+ w->size = w->offset = 0;
}
if (size < DEFAULT_WINDOW_SIZE) {
} else
delta = 0;
+ if (offset > (uint64_t) f->last_stat.st_size)
+ return -EADDRNOTAVAIL;
+
+ if (offset + size > (uint64_t) f->last_stat.st_size)
+ size = PAGE_ALIGN((uint64_t) f->last_stat.st_size - offset);
+
+ if (size <= 0)
+ return -EADDRNOTAVAIL;
+
r = journal_file_map(f,
offset, size,
- &f->window, &f->window_offset, &f->window_size,
- & p);
+ &w->ptr, &w->offset, &w->size,
+ &p);
if (r < 0)
return r;
}
static bool verify_hash(Object *o) {
- uint64_t t;
+ uint64_t h1, h2;
assert(o);
- t = le64toh(o->object.type);
- if (t == OBJECT_DATA) {
- uint64_t s, h1, h2;
-
- s = le64toh(o->object.size);
-
+ if (o->object.type == OBJECT_DATA && !(o->object.flags & OBJECT_COMPRESSED)) {
h1 = le64toh(o->data.hash);
- h2 = hash64(o->data.payload, s - offsetof(Object, data.payload));
-
- return h1 == h2;
- }
+ 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;
- return true;
+ return h1 == h2;
}
-int journal_file_move_to_object(JournalFile *f, uint64_t offset, int type, Object **ret) {
+int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Object **ret) {
int r;
void *t;
Object *o;
assert(f);
assert(ret);
+ assert(type < _OBJECT_TYPE_MAX);
- r = journal_file_move_to(f, offset, sizeof(ObjectHeader), &t);
+ r = journal_file_move_to(f, type >= 0 ? type : WINDOW_UNKNOWN, offset, sizeof(ObjectHeader), &t);
if (r < 0)
return r;
if (s < sizeof(ObjectHeader))
return -EBADMSG;
- if (type >= 0 && le64toh(o->object.type) != type)
+ if (type >= 0 && o->object.type != type)
return -EBADMSG;
if (s > sizeof(ObjectHeader)) {
- r = journal_file_move_to(f, offset, s, &t);
+ r = journal_file_move_to(f, o->object.type, offset, s, &t);
if (r < 0)
return r;
r = le64toh(f->header->seqnum) + 1;
if (seqnum) {
- /* If an external seqno counter was passed, we update
+ /* If an external seqnum counter was passed, we update
* both the local and the external one, and set it to
* the maximum of both */
f->header->seqnum = htole64(r);
+ if (f->header->first_seqnum == 0)
+ f->header->first_seqnum = htole64(r);
+
return r;
}
-static int journal_file_append_object(JournalFile *f, uint64_t size, Object **ret, uint64_t *offset) {
+static 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;
assert(ret);
p = le64toh(f->header->tail_object_offset);
-
if (p == 0)
p = le64toh(f->header->arena_offset);
else {
- r = journal_file_move_to_object(f, p, -1, &tail);
+ r = journal_file_move_to_object(f, -1, p, &tail);
if (r < 0)
return r;
if (r < 0)
return r;
- r = journal_file_move_to(f, p, size, &t);
+ r = journal_file_move_to(f, type, p, size, &t);
if (r < 0)
return r;
o = (Object*) t;
zero(o->object);
- o->object.type = htole64(OBJECT_UNUSED);
- zero(o->object.reserved);
+ o->object.type = type;
o->object.size = htole64(size);
f->header->tail_object_offset = htole64(p);
- if (f->header->head_object_offset == 0)
- f->header->head_object_offset = htole64(p);
-
f->header->n_objects = htole64(le64toh(f->header->n_objects) + 1);
*ret = o;
return 0;
}
-static int journal_file_setup_hash_table(JournalFile *f) {
+static int journal_file_setup_data_hash_table(JournalFile *f) {
uint64_t s, p;
Object *o;
int r;
assert(f);
- s = DEFAULT_HASH_TABLE_SIZE;
- r = journal_file_append_object(f, offsetof(Object, hash_table.table) + s, &o, &p);
+ s = DEFAULT_DATA_HASH_TABLE_SIZE;
+ r = journal_file_append_object(f,
+ OBJECT_DATA_HASH_TABLE,
+ offsetof(Object, hash_table.items) + s,
+ &o, &p);
if (r < 0)
return r;
- o->object.type = htole64(OBJECT_HASH_TABLE);
- memset(o->hash_table.table, 0, s);
+ memset(o->hash_table.items, 0, s);
- f->header->hash_table_offset = htole64(p + offsetof(Object, hash_table.table));
- f->header->hash_table_size = htole64(s);
+ f->header->data_hash_table_offset = htole64(p + offsetof(Object, hash_table.items));
+ f->header->data_hash_table_size = htole64(s);
return 0;
}
-static int journal_file_setup_bisect_table(JournalFile *f) {
+static int journal_file_setup_field_hash_table(JournalFile *f) {
uint64_t s, p;
Object *o;
int r;
assert(f);
- s = DEFAULT_BISECT_TABLE_SIZE;
- r = journal_file_append_object(f, offsetof(Object, bisect_table.table) + s, &o, &p);
+ s = DEFAULT_FIELD_HASH_TABLE_SIZE;
+ r = journal_file_append_object(f,
+ OBJECT_FIELD_HASH_TABLE,
+ offsetof(Object, hash_table.items) + s,
+ &o, &p);
if (r < 0)
return r;
- o->object.type = htole64(OBJECT_BISECT_TABLE);
- memset(o->bisect_table.table, 0, s);
+ memset(o->hash_table.items, 0, s);
- f->header->bisect_table_offset = htole64(p + offsetof(Object, bisect_table.table));
- f->header->bisect_table_size = htole64(s);
+ f->header->field_hash_table_offset = htole64(p + offsetof(Object, hash_table.items));
+ f->header->field_hash_table_size = htole64(s);
return 0;
}
-static int journal_file_map_hash_table(JournalFile *f) {
+static int journal_file_map_data_hash_table(JournalFile *f) {
uint64_t s, p;
void *t;
int r;
assert(f);
- p = le64toh(f->header->hash_table_offset);
- s = le64toh(f->header->hash_table_size);
+ p = le64toh(f->header->data_hash_table_offset);
+ s = le64toh(f->header->data_hash_table_size);
- r = journal_file_map(f,
- p, s,
- &f->hash_table_window, NULL, &f->hash_table_window_size,
- &t);
+ r = journal_file_move_to(f,
+ WINDOW_DATA_HASH_TABLE,
+ p, s,
+ &t);
if (r < 0)
return r;
- f->hash_table = t;
+ f->data_hash_table = t;
return 0;
}
-static int journal_file_map_bisect_table(JournalFile *f) {
+static int journal_file_map_field_hash_table(JournalFile *f) {
uint64_t s, p;
void *t;
int r;
assert(f);
- p = le64toh(f->header->bisect_table_offset);
- s = le64toh(f->header->bisect_table_size);
-
- r = journal_file_map(f,
- p, s,
- &f->bisect_table_window, NULL, &f->bisect_table_window_size,
- &t);
+ p = le64toh(f->header->field_hash_table_offset);
+ s = le64toh(f->header->field_hash_table_size);
+ r = journal_file_move_to(f,
+ WINDOW_FIELD_HASH_TABLE,
+ p, s,
+ &t);
if (r < 0)
return r;
- f->bisect_table = t;
+ f->field_hash_table = t;
return 0;
}
-static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, uint64_t hash_index) {
- uint64_t p;
+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);
- assert(o->object.type == htole64(OBJECT_DATA));
+ assert(o->object.type == OBJECT_DATA);
- o->data.head_entry_offset = o->data.tail_entry_offset = 0;
- o->data.next_hash_offset = 0;
+ o->data.next_hash_offset = o->data.next_field_offset = 0;
+ o->data.entry_offset = o->data.entry_array_offset = 0;
+ o->data.n_entries = 0;
- p = le64toh(f->hash_table[hash_index].tail_hash_offset);
+ h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem));
+ p = le64toh(f->data_hash_table[h].head_hash_offset);
if (p == 0) {
/* Only entry in the hash table is easy */
-
- o->data.prev_hash_offset = 0;
- f->hash_table[hash_index].head_hash_offset = htole64(offset);
+ f->data_hash_table[h].head_hash_offset = htole64(offset);
} else {
- o->data.prev_hash_offset = htole64(p);
-
/* Temporarily move back to the previous data object,
* to patch in pointer */
- r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
+ r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
if (r < 0)
return r;
- o->data.next_hash_offset = offset;
+ o->data.next_hash_offset = htole64(offset);
- r = journal_file_move_to_object(f, offset, OBJECT_DATA, &o);
+ r = journal_file_move_to_object(f, OBJECT_DATA, offset, &o);
if (r < 0)
return r;
}
- f->hash_table[hash_index].tail_hash_offset = htole64(offset);
+ f->data_hash_table[h].tail_hash_offset = htole64(offset);
return 0;
}
-static int journal_file_append_data(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset) {
- uint64_t hash, h, p, np;
- uint64_t osize;
- Object *o;
+int journal_file_find_data_object_with_hash(
+ JournalFile *f,
+ const void *data, uint64_t size, uint64_t hash,
+ Object **ret, uint64_t *offset) {
+ uint64_t p, osize, h;
int r;
assert(f);
osize = offsetof(Object, data.payload) + size;
- hash = hash64(data, size);
- h = hash % (le64toh(f->header->hash_table_size) / sizeof(HashItem));
- p = le64toh(f->hash_table[h].head_hash_offset);
+ if (f->header->data_hash_table_size == 0)
+ return -EBADMSG;
- while (p != 0) {
- /* Look for this data object in the hash table */
+ h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem));
+ p = le64toh(f->data_hash_table[h].head_hash_offset);
- r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
+ while (p > 0) {
+ Object *o;
+
+ r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
if (r < 0)
return r;
- if (le64toh(o->object.size) == osize &&
- memcmp(o->data.payload, data, size) == 0) {
+ if (le64toh(o->data.hash) != hash)
+ goto next;
- if (le64toh(o->data.hash) != hash)
+ if (o->object.flags & OBJECT_COMPRESSED) {
+#ifdef HAVE_XZ
+ uint64_t l, rsize;
+
+ l = le64toh(o->object.size);
+ if (l <= offsetof(Object, data.payload))
return -EBADMSG;
+ l -= offsetof(Object, data.payload);
+
+ if (!uncompress_blob(o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize))
+ return -EBADMSG;
+
+ if (rsize == size &&
+ memcmp(f->compress_buffer, data, size) == 0) {
+
+ if (ret)
+ *ret = o;
+
+ if (offset)
+ *offset = p;
+
+ return 1;
+ }
+#else
+ return -EPROTONOSUPPORT;
+#endif
+
+ } else if (le64toh(o->object.size) == osize &&
+ memcmp(o->data.payload, data, size) == 0) {
+
if (ret)
*ret = o;
if (offset)
*offset = p;
- return 0;
+ return 1;
}
+ next:
p = le64toh(o->data.next_hash_offset);
}
- r = journal_file_append_object(f, osize, &o, &np);
+ return 0;
+}
+
+int journal_file_find_data_object(
+ JournalFile *f,
+ const void *data, uint64_t size,
+ Object **ret, uint64_t *offset) {
+
+ uint64_t hash;
+
+ assert(f);
+ assert(data || size == 0);
+
+ hash = hash64(data, size);
+
+ return journal_file_find_data_object_with_hash(f,
+ data, size, hash,
+ ret, offset);
+}
+
+static int journal_file_append_data(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset) {
+ uint64_t hash, p;
+ uint64_t osize;
+ Object *o;
+ int r;
+ bool compressed = false;
+
+ assert(f);
+ assert(data || size == 0);
+
+ hash = hash64(data, size);
+
+ r = journal_file_find_data_object_with_hash(f, data, 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, data.payload) + size;
+ r = journal_file_append_object(f, OBJECT_DATA, osize, &o, &p);
if (r < 0)
return r;
- o->object.type = htole64(OBJECT_DATA);
o->data.hash = htole64(hash);
- memcpy(o->data.payload, data, size);
- r = journal_file_link_data(f, o, np, h);
+#ifdef HAVE_XZ
+ if (f->compress &&
+ size >= COMPRESSION_SIZE_THRESHOLD) {
+ uint64_t rsize;
+
+ compressed = compress_blob(data, size, o->data.payload, &rsize);
+
+ if (compressed) {
+ 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)
+ memcpy(o->data.payload, data, size);
+
+ r = journal_file_link_data(f, o, p, hash);
if (r < 0)
return r;
*ret = o;
if (offset)
- *offset = np;
+ *offset = p;
return 0;
}
return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem);
}
-static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) {
- uint64_t p, q;
+static uint64_t journal_file_entry_array_n_items(Object *o) {
+ assert(o);
+ assert(o->object.type == htole64(OBJECT_ENTRY_ARRAY));
+
+ return (le64toh(o->object.size) - offsetof(Object, entry_array.items)) / sizeof(uint64_t);
+}
+
+static int link_entry_into_array(JournalFile *f,
+ uint64_t *first,
+ uint64_t *idx,
+ uint64_t p) {
int r;
+ uint64_t n = 0, ap = 0, q, i, a, hidx;
+ Object *o;
+
assert(f);
- assert(o);
- assert(offset > 0);
+ assert(first);
+ assert(idx);
+ assert(p > 0);
- p = le64toh(o->entry.items[i].object_offset);
- if (p == 0)
- return -EINVAL;
+ a = le64toh(*first);
+ i = hidx = le64toh(*idx);
+ while (a > 0) {
- o->entry.items[i].next_entry_offset = 0;
+ r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o);
+ if (r < 0)
+ return r;
- /* Move to the data object */
- r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
+ n = journal_file_entry_array_n_items(o);
+ if (i < n) {
+ o->entry_array.items[i] = htole64(p);
+ *idx = htole64(hidx + 1);
+ return 0;
+ }
+
+ i -= n;
+ ap = a;
+ a = le64toh(o->entry_array.next_entry_array_offset);
+ }
+
+ if (hidx > n)
+ n = (hidx+1) * 2;
+ else
+ n = n * 2;
+
+ if (n < 4)
+ n = 4;
+
+ r = journal_file_append_object(f, OBJECT_ENTRY_ARRAY,
+ offsetof(Object, entry_array.items) + n * sizeof(uint64_t),
+ &o, &q);
if (r < 0)
return r;
- q = le64toh(o->data.tail_entry_offset);
- o->data.tail_entry_offset = htole64(offset);
+ o->entry_array.items[i] = htole64(p);
- if (q == 0)
- o->data.head_entry_offset = htole64(offset);
+ if (ap == 0)
+ *first = q;
else {
- uint64_t n, j;
-
- /* Move to previous entry */
- r = journal_file_move_to_object(f, q, OBJECT_ENTRY, &o);
+ r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, ap, &o);
if (r < 0)
return r;
- n = journal_file_entry_n_items(o);
- for (j = 0; j < n; j++)
- if (le64toh(o->entry.items[j].object_offset) == p)
- break;
+ o->entry_array.next_entry_array_offset = htole64(q);
+ }
+
+ *idx = htole64(hidx + 1);
- if (j >= n)
- return -EBADMSG;
+ return 0;
+}
- o->entry.items[j].next_entry_offset = offset;
+static int link_entry_into_array_plus_one(JournalFile *f,
+ uint64_t *extra,
+ uint64_t *first,
+ uint64_t *idx,
+ uint64_t p) {
+
+ int r;
+
+ assert(f);
+ assert(extra);
+ assert(first);
+ assert(idx);
+ assert(p > 0);
+
+ if (*idx == 0)
+ *extra = htole64(p);
+ else {
+ uint64_t i;
+
+ i = le64toh(*idx) - 1;
+ r = link_entry_into_array(f, first, &i, p);
+ if (r < 0)
+ return r;
}
- /* Move back to original entry */
- r = journal_file_move_to_object(f, offset, OBJECT_ENTRY, &o);
+ *idx = htole64(le64toh(*idx) + 1);
+ return 0;
+}
+
+static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) {
+ uint64_t p;
+ int r;
+ assert(f);
+ assert(o);
+ assert(offset > 0);
+
+ p = le64toh(o->entry.items[i].object_offset);
+ if (p == 0)
+ return -EINVAL;
+
+ r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
if (r < 0)
return r;
- o->entry.items[i].prev_entry_offset = q;
- return 0;
+ return link_entry_into_array_plus_one(f,
+ &o->data.entry_offset,
+ &o->data.entry_array_offset,
+ &o->data.n_entries,
+ offset);
}
static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) {
- uint64_t p, i, n, k, a, b;
+ uint64_t n, i;
int r;
assert(f);
assert(o);
assert(offset > 0);
- assert(o->object.type == htole64(OBJECT_ENTRY));
+ assert(o->object.type == OBJECT_ENTRY);
/* Link up the entry itself */
- p = le64toh(f->header->tail_entry_offset);
+ r = link_entry_into_array(f,
+ &f->header->entry_array_offset,
+ &f->header->n_entries,
+ offset);
+ if (r < 0)
+ return r;
- o->entry.prev_entry_offset = f->header->tail_entry_offset;
- o->entry.next_entry_offset = 0;
+ log_error("=> %s seqnr=%lu n_entries=%lu", f->path, (unsigned long) o->entry.seqnum, (unsigned long) f->header->n_entries);
- if (p == 0) {
- f->header->head_entry_offset = htole64(offset);
+ if (f->header->head_entry_realtime == 0)
f->header->head_entry_realtime = o->entry.realtime;
- } else {
- /* Temporarily move back to the previous entry, to
- * patch in pointer */
-
- r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
- if (r < 0)
- return r;
-
- o->entry.next_entry_offset = htole64(offset);
-
- r = journal_file_move_to_object(f, offset, OBJECT_ENTRY, &o);
- if (r < 0)
- return r;
- }
- f->header->tail_entry_offset = htole64(offset);
f->header->tail_entry_realtime = o->entry.realtime;
+ f->header->tail_entry_monotonic = o->entry.monotonic;
+
+ f->tail_entry_monotonic_valid = true;
/* Link up the items */
n = journal_file_entry_n_items(o);
return r;
}
- /* Link up the entry in the bisect table */
- n = le64toh(f->header->bisect_table_size) / sizeof(uint64_t);
- k = le64toh(f->header->arena_max_size) / n;
-
- a = (le64toh(f->header->last_bisect_offset) + k - 1) / k;
- b = offset / k;
-
- for (; a <= b; a++)
- f->bisect_table[a] = htole64(offset);
-
- f->header->last_bisect_offset = htole64(offset + le64toh(o->object.size));
-
return 0;
}
const dual_timestamp *ts,
uint64_t xor_hash,
const EntryItem items[], unsigned n_items,
- uint64_t *seqno,
+ uint64_t *seqnum,
Object **ret, uint64_t *offset) {
uint64_t np;
uint64_t osize;
assert(f);
assert(items || n_items == 0);
+ assert(ts);
osize = offsetof(Object, entry.items) + (n_items * sizeof(EntryItem));
- r = journal_file_append_object(f, osize, &o, &np);
+ r = journal_file_append_object(f, OBJECT_ENTRY, osize, &o, &np);
if (r < 0)
return r;
- o->object.type = htole64(OBJECT_ENTRY);
- o->entry.seqnum = htole64(journal_file_seqnum(f, seqno));
+ o->entry.seqnum = htole64(journal_file_seqnum(f, seqnum));
memcpy(o->entry.items, items, n_items * sizeof(EntryItem));
- o->entry.realtime = htole64(ts ? ts->realtime : now(CLOCK_REALTIME));
- o->entry.monotonic = htole64(ts ? ts->monotonic : now(CLOCK_MONOTONIC));
+ 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;
return 0;
}
-int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const struct iovec iovec[], unsigned n_iovec, uint64_t *seqno, Object **ret, uint64_t *offset) {
+void journal_file_post_change(JournalFile *f) {
+ assert(f);
+
+ /* inotify() does not receive IN_MODIFY events from file
+ * accesses done via mmap(). After each access we hence
+ * trigger IN_MODIFY by truncating the journal file to its
+ * current size which triggers IN_MODIFY. */
+
+ __sync_synchronize();
+
+ if (ftruncate(f->fd, f->last_stat.st_size) < 0)
+ log_error("Failed to to truncate file to its own size: %m");
+}
+
+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) {
unsigned i;
EntryItem *items;
int r;
uint64_t xor_hash = 0;
+ struct dual_timestamp _ts;
assert(f);
assert(iovec || n_iovec == 0);
- items = new(EntryItem, n_iovec);
- if (!items)
- return -ENOMEM;
+ if (!f->writable)
+ return -EPERM;
+
+ if (!ts) {
+ dual_timestamp_get(&_ts);
+ ts = &_ts;
+ }
+
+ if (f->tail_entry_monotonic_valid &&
+ ts->monotonic < le64toh(f->header->tail_entry_monotonic))
+ return -EINVAL;
+
+ if (ts->realtime < le64toh(f->header->tail_entry_realtime))
+ return -EINVAL;
+
+ items = alloca(sizeof(EntryItem) * n_iovec);
for (i = 0; i < n_iovec; i++) {
uint64_t p;
r = journal_file_append_data(f, iovec[i].iov_base, iovec[i].iov_len, &o, &p);
if (r < 0)
- goto finish;
+ return r;
xor_hash ^= le64toh(o->data.hash);
items[i].object_offset = htole64(p);
items[i].hash = o->data.hash;
}
- r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqno, ret, offset);
+ r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqnum, ret, offset);
-finish:
- free(items);
+ journal_file_post_change(f);
return r;
}
-int journal_file_move_to_entry(JournalFile *f, uint64_t seqnum, 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 lower, upper, p, n, k;
+ uint64_t p = 0, a;
int r;
assert(f);
- n = le64toh(f->header->bisect_table_size) / sizeof(uint64_t);
- k = le64toh(f->header->arena_max_size) / n;
-
- lower = 0;
- upper = le64toh(f->header->last_bisect_offset)/k+1;
+ a = first;
+ while (a > 0) {
+ uint64_t n;
- while (lower < upper) {
- k = (upper + lower) / 2;
- p = le64toh(f->bisect_table[k]);
+ r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o);
+ if (r < 0)
+ return r;
- if (p == 0) {
- upper = k;
- continue;
+ n = journal_file_entry_array_n_items(o);
+ if (i < n) {
+ p = le64toh(o->entry_array.items[i]);
+ break;
}
- r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
+ i -= n;
+ a = le64toh(o->entry_array.next_entry_array_offset);
+ }
+
+ if (a <= 0 || p <= 0)
+ return 0;
+
+ r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
+ if (r < 0)
+ return r;
+
+ if (ret)
+ *ret = o;
+
+ if (offset)
+ *offset = p;
+
+ 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) {
+
+ Object *o;
+
+ assert(f);
+
+ if (i == 0) {
+ int r;
+
+ r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o);
if (r < 0)
return r;
- if (o->entry.seqnum == seqnum) {
- if (ret)
- *ret = o;
+ if (ret)
+ *ret = o;
- if (offset)
- *offset = p;
+ if (offset)
+ *offset = extra;
- return 1;
- } else if (seqnum < o->entry.seqnum)
- upper = k;
- else if (seqnum > o->entry.seqnum)
- lower = k+1;
+ return 1;
}
- assert(lower == upper);
+ return generic_array_get(f, first, i-1, ret, offset);
+}
- if (lower <= 0)
- return 0;
+enum {
+ TEST_FOUND,
+ TEST_LEFT,
+ TEST_RIGHT
+};
- /* The object we are looking for is between
- * bisect_table[lower-1] and bisect_table[lower] */
+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;
+ bool subtract_one = false;
+ Object *o, *array = NULL;
+ int r;
- p = le64toh(f->bisect_table[lower-1]);
+ assert(f);
+ assert(test_object);
- for (;;) {
- r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
+ a = first;
+ while (a > 0) {
+ uint64_t left, right, k, lp;
+
+ r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &array);
if (r < 0)
return r;
- if (o->entry.seqnum == seqnum) {
- if (ret)
- *ret = o;
+ k = journal_file_entry_array_n_items(array);
+ right = MIN(k, n);
+ if (right <= 0)
+ return 0;
- if (offset)
- *offset = p;
+ i = right - 1;
+ lp = p = le64toh(array->entry_array.items[i]);
+ if (p <= 0)
+ return -EBADMSG;
- return 1;
+ r = test_object(f, p, needle);
+ if (r < 0)
+ return r;
- } if (seqnum < o->entry.seqnum)
- return 0;
+ if (r == TEST_FOUND)
+ r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
- if (o->entry.next_entry_offset == 0)
+ if (r == TEST_RIGHT) {
+ left = 0;
+ right -= 1;
+ for (;;) {
+ if (left == right) {
+ if (direction == DIRECTION_UP)
+ subtract_one = true;
+
+ i = left;
+ goto found;
+ }
+
+ assert(left < right);
+
+ i = (left + right) / 2;
+ p = le64toh(array->entry_array.items[i]);
+ 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 = i;
+ else
+ left = i + 1;
+ }
+ }
+
+ if (k > n)
return 0;
- p = le64toh(o->entry.next_entry_offset);
+ last_p = lp;
+
+ n -= k;
+ t += k;
+ a = le64toh(array->entry_array.next_entry_array_offset);
}
return 0;
+
+found:
+ if (subtract_one && t == 0 && i == 0)
+ return 0;
+
+ if (subtract_one && i == 0)
+ p = last_p;
+ else if (subtract_one)
+ p = le64toh(array->entry_array.items[i-1]);
+ else
+ p = le64toh(array->entry_array.items[i]);
+
+ r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
+ if (r < 0)
+ return r;
+
+ if (ret)
+ *ret = o;
+
+ if (offset)
+ *offset = p;
+
+ if (idx)
+ *idx = t + i - (subtract_one ? 1 : 0);
+
+ return 1;
}
-int journal_file_next_entry(JournalFile *f, Object *o, direction_t direction, Object **ret, uint64_t *offset) {
- uint64_t np;
+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;
assert(f);
+ assert(test_object);
- if (!o)
- np = le64toh(direction == DIRECTION_DOWN ?
- f->header->head_entry_offset :
- f->header->tail_entry_offset);
- else {
- if (le64toh(o->object.type) != OBJECT_ENTRY)
- return -EINVAL;
+ if (n <= 0)
+ return 0;
- np = le64toh(direction == DIRECTION_DOWN ?
- o->entry.next_entry_offset :
- o->entry.prev_entry_offset);
- }
+ /* This bisects the array in object 'first', but first checks
+ * an extra */
+ 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 (np == 0)
+ if (offset)
+ *offset = extra;
+
+ if (idx)
+ *idx = 0;
+
+ return 1;
+ } else if (r == TEST_RIGHT)
return 0;
- r = journal_file_move_to_object(f, np, OBJECT_ENTRY, &o);
+ r = generic_array_bisect(f, first, n-1, needle, test_object, direction, ret, offset, idx);
+
+ if (r > 0)
+ (*idx) ++;
+
+ return r;
+}
+
+static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) {
+ Object *o;
+ int r;
+
+ assert(f);
+ assert(p > 0);
+
+ r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
if (r < 0)
return r;
- if (ret)
- *ret = o;
+ if (le64toh(o->entry.seqnum) == needle)
+ return TEST_FOUND;
+ else if (le64toh(o->entry.seqnum) < needle)
+ return TEST_LEFT;
+ else
+ return TEST_RIGHT;
+}
- if (offset)
- *offset = np;
+int journal_file_move_to_entry_by_seqnum(
+ JournalFile *f,
+ uint64_t seqnum,
+ direction_t direction,
+ Object **ret,
+ uint64_t *offset) {
+
+ return generic_array_bisect(f,
+ le64toh(f->header->entry_array_offset),
+ le64toh(f->header->n_entries),
+ seqnum,
+ test_object_seqnum,
+ direction,
+ ret, offset, NULL);
+}
- return 1;
+static int test_object_realtime(JournalFile *f, uint64_t p, uint64_t needle) {
+ Object *o;
+ int r;
+
+ assert(f);
+ assert(p > 0);
+
+ r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
+ if (r < 0)
+ return r;
+
+ if (le64toh(o->entry.realtime) == needle)
+ return TEST_FOUND;
+ else if (le64toh(o->entry.realtime) < needle)
+ return TEST_LEFT;
+ else
+ return TEST_RIGHT;
}
-int journal_file_prev_entry(JournalFile *f, Object *o, Object **ret, uint64_t *offset) {
- uint64_t np;
+int journal_file_move_to_entry_by_realtime(
+ JournalFile *f,
+ uint64_t realtime,
+ direction_t direction,
+ Object **ret,
+ uint64_t *offset) {
+
+ return generic_array_bisect(f,
+ le64toh(f->header->entry_array_offset),
+ le64toh(f->header->n_entries),
+ realtime,
+ test_object_realtime,
+ direction,
+ ret, offset, NULL);
+}
+
+static int test_object_monotonic(JournalFile *f, uint64_t p, uint64_t needle) {
+ Object *o;
int r;
assert(f);
+ assert(p > 0);
+
+ r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
+ if (r < 0)
+ return r;
+
+ if (le64toh(o->entry.monotonic) == needle)
+ return TEST_FOUND;
+ else if (le64toh(o->entry.monotonic) < needle)
+ return TEST_LEFT;
+ else
+ return TEST_RIGHT;
+}
+
+int journal_file_move_to_entry_by_monotonic(
+ JournalFile *f,
+ sd_id128_t boot_id,
+ uint64_t monotonic,
+ direction_t direction,
+ Object **ret,
+ uint64_t *offset) {
+
+ char t[8+32+1] = "_BOOT_ID=";
+ Object *o;
+ int r;
+
+ sd_id128_to_string(boot_id, t + 8);
+
+ r = journal_file_find_data_object(f, t, strlen(t), &o, NULL);
+ if (r < 0)
+ return r;
+ else if (r == 0)
+ return -ENOENT;
+
+ return 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,
+ 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,
+ direction_t direction,
+ Object **ret, uint64_t *offset) {
+
+ uint64_t i, n;
+ int r;
+
+ assert(f);
+ assert(p > 0 || !o);
+
+ n = le64toh(f->header->n_entries);
+ if (n <= 0)
+ return 0;
if (!o)
- np = le64toh(f->header->tail_entry_offset);
+ i = direction == DIRECTION_DOWN ? 0 : n - 1;
else {
- if (le64toh(o->object.type) != OBJECT_ENTRY)
+ if (o->object.type != OBJECT_ENTRY)
return -EINVAL;
- np = le64toh(o->entry.prev_entry_offset);
+ r = generic_array_bisect(f,
+ le64toh(f->header->entry_array_offset),
+ le64toh(f->header->n_entries),
+ p,
+ test_object_offset,
+ DIRECTION_DOWN,
+ NULL, NULL,
+ &i);
+ if (r <= 0)
+ return r;
+
+ if (direction == DIRECTION_DOWN) {
+ if (i >= n - 1)
+ return 0;
+
+ i++;
+ } else {
+ if (i <= 0)
+ return 0;
+
+ i--;
+ }
}
- if (np == 0)
- return 0;
+ /* And jump to it */
+ return generic_array_get(f,
+ le64toh(f->header->entry_array_offset),
+ i,
+ ret, offset);
+}
- r = journal_file_move_to_object(f, np, OBJECT_ENTRY, &o);
- if (r < 0)
+int journal_file_skip_entry(
+ JournalFile *f,
+ Object *o, uint64_t p,
+ int64_t skip,
+ Object **ret, uint64_t *offset) {
+
+ uint64_t i, n;
+ int r;
+
+ assert(f);
+ assert(o);
+ assert(p > 0);
+
+ if (o->object.type != OBJECT_ENTRY)
+ return -EINVAL;
+
+ r = generic_array_bisect(f,
+ le64toh(f->header->entry_array_offset),
+ le64toh(f->header->n_entries),
+ p,
+ test_object_offset,
+ DIRECTION_DOWN,
+ NULL, NULL,
+ &i);
+ if (r <= 0)
return r;
- if (ret)
- *ret = o;
+ /* Calculate new index */
+ if (skip < 0) {
+ if ((uint64_t) -skip >= i)
+ i = 0;
+ else
+ i = i - (uint64_t) -skip;
+ } else
+ i += (uint64_t) skip;
- if (offset)
- *offset = np;
+ n = le64toh(f->header->n_entries);
+ if (n <= 0)
+ return -EBADMSG;
- return 1;
+ if (i >= n)
+ i = n-1;
+
+ return generic_array_get(f,
+ le64toh(f->header->entry_array_offset),
+ i,
+ ret, offset);
}
-int journal_file_find_first_entry(JournalFile *f, const void *data, uint64_t size, direction_t direction, Object **ret, uint64_t *offset) {
- uint64_t p, osize, hash, h;
+int journal_file_next_entry_for_data(
+ JournalFile *f,
+ Object *o, uint64_t p,
+ uint64_t data_offset,
+ direction_t direction,
+ Object **ret, uint64_t *offset) {
+
+ uint64_t n, i;
int r;
+ Object *d;
assert(f);
- assert(data || size == 0);
+ assert(p > 0 || !o);
- osize = offsetof(Object, data.payload) + size;
+ r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
+ if (r < 0)
+ return r;
- hash = hash64(data, size);
- h = hash % (le64toh(f->header->hash_table_size) / sizeof(HashItem));
- p = le64toh(f->hash_table[h].head_hash_offset);
+ n = le64toh(d->data.n_entries);
+ if (n <= 0)
+ return n;
- while (p != 0) {
- Object *o;
+ if (!o)
+ i = direction == DIRECTION_DOWN ? 0 : n - 1;
+ else {
+ if (o->object.type != OBJECT_ENTRY)
+ return -EINVAL;
- r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
- if (r < 0)
+ r = 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_DOWN,
+ NULL, NULL,
+ &i);
+
+ if (r <= 0)
return r;
- if (le64toh(o->object.size) == osize &&
- memcmp(o->data.payload, data, size) == 0) {
+ if (direction == DIRECTION_DOWN) {
+ if (i >= n - 1)
+ return 0;
- if (le64toh(o->data.hash) != hash)
- return -EBADMSG;
+ i++;
+ } else {
+ if (i <= 0)
+ return 0;
- p = le64toh(direction == DIRECTION_DOWN ?
- o->data.head_entry_offset :
- o->data.tail_entry_offset);
+ i--;
+ }
- if (p == 0)
- return 0;
+ }
- r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
- if (r < 0)
- return r;
+ return generic_array_get_plus_one(f,
+ le64toh(d->data.entry_offset),
+ le64toh(d->data.entry_array_offset),
+ i,
+ ret, offset);
+}
- if (ret)
- *ret = o;
+int journal_file_move_to_entry_by_seqnum_for_data(
+ JournalFile *f,
+ uint64_t data_offset,
+ uint64_t seqnum,
+ direction_t direction,
+ Object **ret, uint64_t *offset) {
- if (offset)
- *offset = p;
+ Object *d;
+ int r;
- return 1;
- }
+ r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
+ if (r <= 0)
+ return r;
- p = le64toh(o->data.next_hash_offset);
- }
+ return generic_array_bisect_plus_one(f,
+ le64toh(d->data.entry_offset),
+ le64toh(d->data.entry_array_offset),
+ le64toh(d->data.n_entries),
+ seqnum,
+ test_object_seqnum,
+ direction,
+ ret, offset, NULL);
+}
- return 0;
+int journal_file_move_to_entry_by_realtime_for_data(
+ JournalFile *f,
+ uint64_t data_offset,
+ uint64_t realtime,
+ direction_t direction,
+ Object **ret, uint64_t *offset) {
+
+ Object *d;
+ int r;
+
+ 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),
+ realtime,
+ test_object_realtime,
+ direction,
+ ret, offset, NULL);
}
void journal_file_dump(JournalFile *f) {
assert(f);
- printf("File ID: %s\n"
+ printf("File Path: %s\n"
+ "File ID: %s\n"
"Machine ID: %s\n"
"Boot ID: %s\n"
- "Arena size: %llu\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 long) le64toh(f->header->arena_size),
+ (unsigned long) le64toh(f->header->n_objects),
+ (unsigned long) le64toh(f->header->n_entries));
- p = le64toh(f->header->head_object_offset);
+ p = le64toh(f->header->arena_offset);
while (p != 0) {
- r = journal_file_move_to_object(f, p, -1, &o);
+ r = journal_file_move_to_object(f, -1, p, &o);
if (r < 0)
goto fail;
(unsigned long long) le64toh(o->entry.realtime));
break;
- case OBJECT_HASH_TABLE:
- printf("Type: OBJECT_HASH_TABLE\n");
+ case OBJECT_FIELD_HASH_TABLE:
+ printf("Type: OBJECT_FIELD_HASH_TABLE\n");
break;
- case OBJECT_BISECT_TABLE:
- printf("Type: OBJECT_BISECT_TABLE\n");
+ case OBJECT_DATA_HASH_TABLE:
+ printf("Type: OBJECT_DATA_HASH_TABLE\n");
+ break;
+
+ case OBJECT_ENTRY_ARRAY:
+ printf("Type: OBJECT_ENTRY_ARRAY\n");
break;
}
+ if (o->object.flags & OBJECT_COMPRESSED)
+ printf("Flags: COMPRESSED\n");
+
if (p == le64toh(f->header->tail_object_offset))
p = 0;
else
if (newly_created) {
- r = journal_file_setup_hash_table(f);
+ r = journal_file_setup_field_hash_table(f);
if (r < 0)
goto fail;
- r = journal_file_setup_bisect_table(f);
+ r = journal_file_setup_data_hash_table(f);
if (r < 0)
goto fail;
}
- r = journal_file_map_hash_table(f);
+ r = journal_file_map_field_hash_table(f);
if (r < 0)
goto fail;
- r = journal_file_map_bisect_table(f);
+ r = journal_file_map_data_hash_table(f);
if (r < 0)
goto fail;
assert(directory);
if (max_use <= 0)
- max_use = DEFAULT_MAX_USE;
+ return 0;
d = opendir(directory);
if (!d)
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;
+ uint64_t q, xor_hash = 0;
+ int r;
+ EntryItem *items;
+ dual_timestamp ts;
+
+ assert(from);
+ assert(to);
+ assert(o);
+ assert(p);
+
+ if (!to->writable)
+ return -EPERM;
+
+ 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;
+
+ if (ts.realtime < le64toh(to->header->tail_entry_realtime))
+ return -EINVAL;
+
+ n = journal_file_entry_n_items(o);
+ items = alloca(sizeof(EntryItem) * n);
+
+ for (i = 0; i < n; i++) {
+ uint64_t le_hash, l, h;
+ size_t t;
+ void *data;
+ Object *u;
+
+ q = le64toh(o->entry.items[i].object_offset);
+ le_hash = o->entry.items[i].hash;
+
+ r = journal_file_move_to_object(from, OBJECT_DATA, q, &o);
+ if (r < 0)
+ return r;
+
+ if (le_hash != o->data.hash)
+ return -EBADMSG;
+
+ l = le64toh(o->object.size) - offsetof(Object, data.payload);
+ t = (size_t) l;
+
+ /* We hit the limit on 32bit machines */
+ if ((uint64_t) t != l)
+ return -E2BIG;
+
+ if (o->object.flags & OBJECT_COMPRESSED) {
+#ifdef HAVE_XZ
+ uint64_t rsize;
+
+ if (!uncompress_blob(o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize))
+ return -EBADMSG;
+
+ data = from->compress_buffer;
+ l = rsize;
+#else
+ return -EPROTONOSUPPORT;
+#endif
+ } else
+ data = o->data.payload;
+
+ r = journal_file_append_data(to, data, l, &u, &h);
+ if (r < 0)
+ return r;
+
+ xor_hash ^= le64toh(u->data.hash);
+ items[i].object_offset = htole64(h);
+ items[i].hash = u->data.hash;
+
+ r = journal_file_move_to_object(from, OBJECT_ENTRY, p, &o);
+ if (r < 0)
+ return r;
+ }
+
+ return journal_file_append_entry_internal(to, &ts, xor_hash, items, n, seqnum, ret, offset);
+}
+
+void journal_default_metrics(JournalMetrics *m, int fd) {
+ uint64_t fs_size = 0;
+ struct statvfs ss;
+ char a[64], b[64], c[64], d[64];
+
+ assert(m);
+ assert(fd >= 0);
+
+ if (fstatvfs(fd, &ss) >= 0)
+ fs_size = ss.f_frsize * ss.f_blocks;
+
+ if (m->max_use == (uint64_t) -1) {
+
+ if (fs_size > 0) {
+ m->max_use = PAGE_ALIGN(fs_size / 10); /* 10% of file system size */
+
+ if (m->max_use > DEFAULT_MAX_USE_UPPER)
+ m->max_use = DEFAULT_MAX_USE_UPPER;
+
+ if (m->max_use < DEFAULT_MAX_USE_LOWER)
+ m->max_use = DEFAULT_MAX_USE_LOWER;
+ } else
+ m->max_use = DEFAULT_MAX_USE_LOWER;
+ } else {
+ m->max_use = PAGE_ALIGN(m->max_use);
+
+ if (m->max_use < JOURNAL_FILE_SIZE_MIN*2)
+ m->max_use = JOURNAL_FILE_SIZE_MIN*2;
+ }
+
+ if (m->max_size == (uint64_t) -1) {
+ m->max_size = PAGE_ALIGN(m->max_use / 8); /* 8 chunks */
+
+ if (m->max_size > DEFAULT_MAX_SIZE_UPPER)
+ m->max_size = DEFAULT_MAX_SIZE_UPPER;
+ } else
+ m->max_size = PAGE_ALIGN(m->max_size);
+
+ if (m->max_size < JOURNAL_FILE_SIZE_MIN)
+ m->max_size = JOURNAL_FILE_SIZE_MIN;
+
+ if (m->max_size*2 > m->max_use)
+ m->max_use = m->max_size*2;
+
+ if (m->min_size == (uint64_t) -1)
+ m->min_size = JOURNAL_FILE_SIZE_MIN;
+ else {
+ m->min_size = PAGE_ALIGN(m->min_size);
+
+ if (m->min_size < JOURNAL_FILE_SIZE_MIN)
+ m->min_size = JOURNAL_FILE_SIZE_MIN;
+
+ if (m->min_size > m->max_size)
+ m->max_size = m->min_size;
+ }
+
+ if (m->keep_free == (uint64_t) -1) {
+
+ if (fs_size > 0) {
+ m->keep_free = PAGE_ALIGN(fs_size / 20); /* 5% of file system size */
+
+ if (m->keep_free > DEFAULT_KEEP_FREE_UPPER)
+ m->keep_free = DEFAULT_KEEP_FREE_UPPER;
+
+ } else
+ m->keep_free = DEFAULT_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));
+}