Copyright 2011 Lennart Poettering
systemd is free software; you can redistribute it and/or modify it
- under the terms of the GNU General Public License as published by
- the Free Software Foundation; either version 2 of the License, or
+ under the terms of the GNU Lesser General Public License as published by
+ the Free Software Foundation; either version 2.1 of the License, or
(at your option) any later version.
systemd is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
+ Lesser General Public License for more details.
- You should have received a copy of the GNU General Public License
+ You should have received a copy of the GNU Lesser General Public License
along with systemd; If not, see <http://www.gnu.org/licenses/>.
***/
#include "journal-file.h"
#include "lookup3.h"
#include "compress.h"
+#include "fsprg.h"
-#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*16ULL)
-#define DEFAULT_FIELD_HASH_TABLE_SIZE (2047ULL*16ULL)
+#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*sizeof(HashItem))
+#define DEFAULT_FIELD_HASH_TABLE_SIZE (333ULL*sizeof(HashItem))
-#define DEFAULT_WINDOW_SIZE (128ULL*1024ULL*1024ULL)
+#define DEFAULT_WINDOW_SIZE (8ULL*1024ULL*1024ULL)
-#define COMPRESSION_SIZE_THRESHOLD (64ULL)
+#define COMPRESSION_SIZE_THRESHOLD (512ULL)
-static const char signature[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' };
+/* This is the minimum journal file size */
+#define JOURNAL_FILE_SIZE_MIN (64ULL*1024ULL) /* 64 KiB */
+
+/* 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 (128ULL*1024ULL*1024ULL) /* 128 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 */
+
+/* n_data was the first entry we added after the initial file format design */
+#define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data))
#define ALIGN64(x) (((x) + 7ULL) & ~7ULL)
+#define JOURNAL_HEADER_CONTAINS(h, field) \
+ (le64toh((h)->header_size) >= offsetof(Header, field) + sizeof((h)->field))
+
+static int journal_file_maybe_append_tag(JournalFile *f, uint64_t realtime);
+static int journal_file_hmac_put_object(JournalFile *f, int type, uint64_t p);
+
void journal_file_close(JournalFile *f) {
int t;
assert(f);
- if (f->header && f->writable)
- f->header->state = STATE_OFFLINE;
-
+ /* Write the final tag */
+ if (f->authenticate)
+ journal_file_append_tag(f);
+ /* Sync everything to disk, before we mark the file offline */
for (t = 0; t < _WINDOW_MAX; t++)
if (f->windows[t].ptr)
munmap(f->windows[t].ptr, f->windows[t].size);
+ 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;
+
+ munmap(f->header, PAGE_ALIGN(sizeof(Header)));
+ }
+
if (f->fd >= 0)
close_nointr_nofail(f->fd);
free(f->compress_buffer);
#endif
+#ifdef HAVE_GCRYPT
+ if (f->fsprg_header)
+ munmap(f->fsprg_header, PAGE_ALIGN(f->fsprg_size));
+
+ if (f->hmac)
+ gcry_md_close(f->hmac);
+#endif
+
free(f);
}
assert(f);
zero(h);
- memcpy(h.signature, signature, 8);
- h.arena_offset = htole64(ALIGN64(sizeof(h)));
+ 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->authenticate ? HEADER_COMPATIBLE_AUTHENTICATED : 0);
r = sd_id128_randomize(&h.file_id);
if (r < 0)
if (template) {
h.seqnum_id = template->header->seqnum_id;
- h.seqnum = template->header->seqnum;
+ h.tail_seqnum = template->header->tail_seqnum;
} else
h.seqnum_id = h.file_id;
f->header->boot_id = boot_id;
f->header->state = STATE_ONLINE;
+
+ /* 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
- if ((uint64_t) f->last_stat.st_size < (le64toh(f->header->arena_offset) + le64toh(f->header->arena_size)))
+ /* 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_AUTHENTICATED) != 0)
+ return -EPROTONOSUPPORT;
+#else
+ if (f->header->compatible_flags != 0)
+ return -EPROTONOSUPPORT;
+#endif
+ }
+
+ /* 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)))
return -ENODATA;
if (f->writable) {
- uint32_t state;
+ uint8_t state;
sd_id128_t machine_id;
int r;
state = f->header->state;
- if (state == STATE_ONLINE)
- log_debug("Journal file %s is already online. Assuming unclean closing. Ignoring.", f->path);
- else if (state == STATE_ARCHIVED)
+ if (state == STATE_ONLINE) {
+ log_debug("Journal file %s is already online. Assuming unclean closing.", f->path);
+ return -EBUSY;
+ } else if (state == STATE_ARCHIVED)
return -ESHUTDOWN;
- else if (state != STATE_OFFLINE)
- log_debug("Journal file %s has unknown state %u. Ignoring.", f->path, state);
+ else if (state != STATE_OFFLINE) {
+ log_debug("Journal file %s has unknown state %u.", f->path, state);
+ return -EBUSY;
+ }
}
+ f->compress = !!(le32toh(f->header->incompatible_flags) & HEADER_INCOMPATIBLE_COMPRESSED);
+ f->authenticate = !!(le32toh(f->header->compatible_flags) & HEADER_COMPATIBLE_AUTHENTICATED);
+
return 0;
}
static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) {
uint64_t old_size, new_size;
+ int r;
assert(f);
* ourselves */
old_size =
- le64toh(f->header->arena_offset) +
+ le64toh(f->header->header_size) +
le64toh(f->header->arena_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 < le64toh(f->header->header_size))
+ new_size = le64toh(f->header->header_size);
if (new_size <= old_size)
return 0;
/* 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;
+ r = posix_fallocate(f->fd, old_size, new_size - old_size);
+ if (r != 0)
+ return -r;
if (fstat(f->fd, &f->last_stat) < 0)
return -errno;
- f->header->arena_size = new_size - htole64(f->header->arena_offset);
+ f->header->arena_size = htole64(new_size - le64toh(f->header->header_size));
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;
}
static int journal_file_move_to(JournalFile *f, int wt, uint64_t offset, uint64_t size, void **ret) {
- void *p;
+ void *p = NULL;
uint64_t delta;
int r;
Window *w;
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. */
+
+ if (fstat(f->fd, &f->last_stat) < 0 ||
+ offset + size > (uint64_t) f->last_stat.st_size)
+ return -EADDRNOTAVAIL;
+ }
+
w = f->windows + wt;
if (_likely_(w->ptr &&
* the window space before and half behind the
* requested mapping */
- delta = PAGE_ALIGN((DEFAULT_WINDOW_SIZE - size) / 2);
+ delta = (DEFAULT_WINDOW_SIZE - size) / 2;
- if (offset < delta)
+ if (delta > offset)
delta = offset;
offset -= delta;
- size += (DEFAULT_WINDOW_SIZE - 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,
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->seqnum) + 1;
+ r = le64toh(f->header->tail_seqnum) + 1;
if (seqnum) {
/* If an external seqnum counter was passed, we update
*seqnum = r;
}
- f->header->seqnum = htole64(r);
+ f->header->tail_seqnum = htole64(r);
- if (f->header->first_seqnum == 0)
- f->header->first_seqnum = htole64(r);
+ if (f->header->head_seqnum == 0)
+ f->header->head_seqnum = htole64(r);
return r;
}
p = le64toh(f->header->tail_object_offset);
if (p == 0)
- p = le64toh(f->header->arena_offset);
+ p = le64toh(f->header->header_size);
else {
r = journal_file_move_to_object(f, -1, p, &tail);
if (r < 0)
assert(f);
- s = DEFAULT_DATA_HASH_TABLE_SIZE;
+ /* We estimate that we need 1 hash table entry per 768 of
+ journal file and we want to make sure we never get beyond
+ 75% fill level. Calculate the hash table size for the
+ maximum file size based on these metrics. */
+
+ s = (f->metrics.max_size * 4 / 768 / 3) * sizeof(HashItem);
+ if (s < DEFAULT_DATA_HASH_TABLE_SIZE)
+ s = DEFAULT_DATA_HASH_TABLE_SIZE;
+
+ log_info("Reserving %llu entries in hash table.", (unsigned long long) (s / sizeof(HashItem)));
+
r = journal_file_append_object(f,
OBJECT_DATA_HASH_TABLE,
offsetof(Object, hash_table.items) + s,
assert(offset > 0);
assert(o->object.type == OBJECT_DATA);
+ /* This might alter the window we are looking at */
+
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;
h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem));
- p = le64toh(f->data_hash_table[h].head_hash_offset);
+ p = le64toh(f->data_hash_table[h].tail_hash_offset);
if (p == 0) {
/* Only entry in the hash table is easy */
f->data_hash_table[h].head_hash_offset = htole64(offset);
} else {
- /* Temporarily move back to the previous data object,
- * to patch in pointer */
+ /* Move back to the previous data object, to patch in
+ * pointer */
r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
if (r < 0)
return r;
o->data.next_hash_offset = htole64(offset);
-
- r = journal_file_move_to_object(f, OBJECT_DATA, offset, &o);
- if (r < 0)
- return r;
}
f->data_hash_table[h].tail_hash_offset = htole64(offset);
+ if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
+ f->header->n_data = htole64(le64toh(f->header->n_data) + 1);
+
return 0;
}
JournalFile *f,
const void *data, uint64_t size, uint64_t hash,
Object **ret, uint64_t *offset) {
+
uint64_t p, osize, h;
int r;
return r;
if (le64toh(o->data.hash) != hash)
- return -EBADMSG;
+ goto next;
if (o->object.flags & OBJECT_COMPRESSED) {
#ifdef HAVE_XZ
return 1;
}
+ next:
p = le64toh(o->data.next_hash_offset);
}
ret, offset);
}
-static int journal_file_append_data(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *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;
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;
+ 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;
+
if (ret)
*ret = o;
uint64_t journal_file_entry_n_items(Object *o) {
assert(o);
- assert(o->object.type == htole64(OBJECT_ENTRY));
+ assert(o->object.type == OBJECT_ENTRY);
return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem);
}
static uint64_t journal_file_entry_array_n_items(Object *o) {
assert(o);
- assert(o->object.type == htole64(OBJECT_ENTRY_ARRAY));
+ assert(o->object.type == 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,
+ le64_t *first,
+ le64_t *idx,
uint64_t p) {
int r;
uint64_t n = 0, ap = 0, q, i, a, hidx;
if (r < 0)
return r;
+ r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, q);
+ if (r < 0)
+ return r;
+
o->entry_array.items[i] = htole64(p);
if (ap == 0)
- *first = q;
+ *first = htole64(q);
else {
r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, ap, &o);
if (r < 0)
}
static int link_entry_into_array_plus_one(JournalFile *f,
- uint64_t *extra,
- uint64_t *first,
- uint64_t *idx,
+ le64_t *extra,
+ le64_t *first,
+ le64_t *idx,
uint64_t p) {
int r;
if (*idx == 0)
*extra = htole64(p);
else {
- uint64_t i;
+ le64_t i;
- i = le64toh(*idx) - 1;
+ i = htole64(le64toh(*idx) - 1);
r = link_entry_into_array(f, first, &i, p);
if (r < 0)
return r;
assert(offset > 0);
assert(o->object.type == OBJECT_ENTRY);
+ __sync_synchronize();
+
/* Link up the entry itself */
r = link_entry_into_array(f,
&f->header->entry_array_offset,
if (r < 0)
return r;
- log_error("=> %s seqnr=%lu n_entries=%lu", f->path, (unsigned long) o->entry.seqnum, (unsigned long) f->header->n_entries);
+ /* log_debug("=> %s seqnr=%lu n_entries=%lu", f->path, (unsigned long) o->entry.seqnum, (unsigned long) f->header->n_entries); */
if (f->header->head_entry_realtime == 0)
f->header->head_entry_realtime = o->entry.realtime;
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;
+ r = journal_file_hmac_put_object(f, OBJECT_ENTRY, np);
+ if (r < 0)
+ return r;
+
r = journal_file_link_entry(f, o, np);
if (r < 0)
return r;
return 0;
}
-static void journal_file_post_change(JournalFile *f) {
+void journal_file_post_change(JournalFile *f) {
assert(f);
/* inotify() does not receive IN_MODIFY events from file
ts->monotonic < le64toh(f->header->tail_entry_monotonic))
return -EINVAL;
- if (ts->realtime < le64toh(f->header->tail_entry_realtime))
- return -EINVAL;
+ r = journal_file_maybe_append_tag(f, ts->realtime);
+ if (r < 0)
+ return r;
- items = new(EntryItem, n_iovec);
- if (!items)
- return -ENOMEM;
+ /* 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;
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);
journal_file_post_change(f);
-finish:
- free(items);
-
return r;
}
Object **ret, uint64_t *offset) {
Object *o;
- uint64_t p, a;
+ uint64_t p = 0, a;
int r;
assert(f);
}
}
- if (k > n)
+ if (k > n) {
+ if (direction == DIRECTION_UP) {
+ i = n;
+ subtract_one = true;
+ goto found;
+ }
+
return 0;
+ }
last_p = lp;
*offset = p;
if (idx)
- *idx = t + i - (subtract_one ? 1 : 0);
+ *idx = t + i + (subtract_one ? -1 : 0);
return 1;
}
uint64_t *idx) {
int r;
+ bool step_back = false;
+ Object *o;
assert(f);
assert(test_object);
r = test_object(f, extra, needle);
if (r < 0)
return r;
- else if (r == TEST_FOUND) {
- Object *o;
-
- r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o);
- if (r < 0)
- return r;
-
- if (ret)
- *ret = o;
- if (offset)
- *offset = extra;
+ if (r == TEST_FOUND)
+ r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
- if (idx)
- *idx = 0;
+ /* if we are looking with DIRECTION_UP then we need to first
+ see if in the actual array there is a matching entry, and
+ return the last one of that. But if there isn't any we need
+ to return this one. Hence remember this, and return it
+ below. */
+ if (r == TEST_LEFT)
+ step_back = direction == DIRECTION_UP;
- return 1;
- } else if (r == TEST_RIGHT)
- return 0;
+ if (r == TEST_RIGHT) {
+ if (direction == DIRECTION_DOWN)
+ goto found;
+ else
+ return 0;
+ }
r = generic_array_bisect(f, first, n-1, needle, test_object, direction, ret, offset, idx);
- if (r > 0)
+ if (r == 0 && step_back)
+ goto found;
+
+ if (r > 0 && idx)
(*idx) ++;
return r;
+
+found:
+ r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o);
+ if (r < 0)
+ return r;
+
+ if (ret)
+ *ret = o;
+
+ if (offset)
+ *offset = extra;
+
+ if (idx)
+ *idx = 0;
+
+ return 1;
+}
+
+static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) {
+ assert(f);
+ assert(p > 0);
+
+ if (p == needle)
+ return TEST_FOUND;
+ else if (p < needle)
+ return TEST_LEFT;
+ else
+ return TEST_RIGHT;
+}
+
+int journal_file_move_to_entry_by_offset(
+ JournalFile *f,
+ uint64_t p,
+ direction_t direction,
+ Object **ret,
+ uint64_t *offset) {
+
+ return generic_array_bisect(f,
+ le64toh(f->header->entry_array_offset),
+ le64toh(f->header->n_entries),
+ p,
+ test_object_offset,
+ direction,
+ ret, offset, NULL);
}
+
static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) {
Object *o;
int r;
Object **ret,
uint64_t *offset) {
- char t[8+32+1] = "_BOOT_ID=";
+ char t[9+32+1] = "_BOOT_ID=";
Object *o;
int r;
- sd_id128_to_string(boot_id, t + 8);
+ assert(f);
+ sd_id128_to_string(boot_id, t + 9);
r = journal_file_find_data_object(f, t, strlen(t), &o, NULL);
if (r < 0)
return r;
- else if (r == 0)
+ if (r == 0)
return -ENOENT;
return generic_array_bisect_plus_one(f,
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,
ret, offset);
}
+int journal_file_move_to_entry_by_offset_for_data(
+ JournalFile *f,
+ uint64_t data_offset,
+ uint64_t p,
+ direction_t direction,
+ Object **ret, uint64_t *offset) {
+
+ int r;
+ Object *d;
+
+ assert(f);
+
+ r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
+ if (r < 0)
+ return r;
+
+ return generic_array_bisect_plus_one(f,
+ le64toh(d->data.entry_offset),
+ le64toh(d->data.entry_array_offset),
+ le64toh(d->data.n_entries),
+ p,
+ test_object_offset,
+ direction,
+ ret, offset, NULL);
+}
+
+int journal_file_move_to_entry_by_monotonic_for_data(
+ JournalFile *f,
+ uint64_t data_offset,
+ sd_id128_t boot_id,
+ uint64_t monotonic,
+ direction_t direction,
+ Object **ret, uint64_t *offset) {
+
+ char t[9+32+1] = "_BOOT_ID=";
+ Object *o, *d;
+ int r;
+ uint64_t b, z;
+
+ assert(f);
+
+ /* First, seek by time */
+ sd_id128_to_string(boot_id, t + 9);
+ r = journal_file_find_data_object(f, t, strlen(t), &o, &b);
+ if (r < 0)
+ return r;
+ if (r == 0)
+ return -ENOENT;
+
+ r = generic_array_bisect_plus_one(f,
+ le64toh(o->data.entry_offset),
+ le64toh(o->data.entry_array_offset),
+ le64toh(o->data.n_entries),
+ monotonic,
+ test_object_monotonic,
+ direction,
+ NULL, &z, NULL);
+ if (r <= 0)
+ return r;
+
+ /* And now, continue seeking until we find an entry that
+ * exists in both bisection arrays */
+
+ for (;;) {
+ Object *qo;
+ uint64_t p, q;
+
+ r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
+ if (r < 0)
+ return r;
+
+ r = generic_array_bisect_plus_one(f,
+ le64toh(d->data.entry_offset),
+ le64toh(d->data.entry_array_offset),
+ le64toh(d->data.n_entries),
+ z,
+ test_object_offset,
+ direction,
+ NULL, &p, NULL);
+ if (r <= 0)
+ return r;
+
+ r = journal_file_move_to_object(f, OBJECT_DATA, b, &o);
+ if (r < 0)
+ return r;
+
+ r = generic_array_bisect_plus_one(f,
+ le64toh(o->data.entry_offset),
+ le64toh(o->data.entry_array_offset),
+ le64toh(o->data.n_entries),
+ p,
+ test_object_offset,
+ direction,
+ &qo, &q, NULL);
+
+ if (r <= 0)
+ return r;
+
+ if (p == q) {
+ if (ret)
+ *ret = qo;
+ if (offset)
+ *offset = q;
+
+ return 1;
+ }
+
+ z = q;
+ }
+
+ return 0;
+}
+
int journal_file_move_to_entry_by_seqnum_for_data(
JournalFile *f,
uint64_t data_offset,
Object *d;
int r;
+ assert(f);
+
r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
- if (r <= 0)
+ if (r < 0)
return r;
return generic_array_bisect_plus_one(f,
Object *d;
int r;
+ assert(f);
+
r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
- if (r <= 0)
+ if (r < 0)
return r;
return generic_array_bisect_plus_one(f,
ret, offset, NULL);
}
-void journal_file_dump(JournalFile *f) {
- char a[33], b[33], c[33];
- Object *o;
- int r;
- uint64_t p;
-
+static void *fsprg_state(JournalFile *f) {
+ uint64_t a, b;
assert(f);
- printf("File Path: %s\n"
- "File ID: %s\n"
- "Machine ID: %s\n"
- "Boot ID: %s\n"
- "Arena size: %llu\n"
- "Objects: %lu\n"
- "Entries: %lu\n",
- f->path,
- sd_id128_to_string(f->header->file_id, a),
- sd_id128_to_string(f->header->machine_id, b),
- sd_id128_to_string(f->header->boot_id, c),
- (unsigned long long) le64toh(f->header->arena_size),
- (unsigned long) le64toh(f->header->n_objects),
- (unsigned long) le64toh(f->header->n_entries));
+ if (!f->authenticate)
+ return NULL;
- p = le64toh(f->header->arena_offset);
- while (p != 0) {
- r = journal_file_move_to_object(f, -1, p, &o);
- if (r < 0)
- goto fail;
+ a = le64toh(f->fsprg_header->header_size);
+ b = le64toh(f->fsprg_header->state_size);
- switch (o->object.type) {
+ if (a + b > f->fsprg_size)
+ return NULL;
- case OBJECT_UNUSED:
- printf("Type: OBJECT_UNUSED\n");
- break;
+ return (uint8_t*) f->fsprg_header + a;
+}
+
+static uint64_t journal_file_tag_seqnum(JournalFile *f) {
+ uint64_t r;
+
+ assert(f);
+
+ r = le64toh(f->header->n_tags) + 1;
+ f->header->n_tags = htole64(r);
+
+ return r;
+}
+
+int journal_file_append_tag(JournalFile *f) {
+ Object *o;
+ uint64_t p;
+ int r;
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ if (!f->hmac_running)
+ return 0;
+
+ log_debug("Writing tag for epoch %llu\n", (unsigned long long) FSPRG_GetEpoch(fsprg_state(f)));
+
+ assert(f->hmac);
+
+ r = journal_file_append_object(f, OBJECT_TAG, sizeof(struct TagObject), &o, &p);
+ if (r < 0)
+ return r;
+
+ o->tag.seqnum = htole64(journal_file_tag_seqnum(f));
+
+ /* Add the tag object itself, so that we can protect its
+ * header. This will exclude the actual hash value in it */
+ r = journal_file_hmac_put_object(f, OBJECT_TAG, p);
+ if (r < 0)
+ return r;
+
+ /* Get the HMAC tag and store it in the object */
+ memcpy(o->tag.tag, gcry_md_read(f->hmac, 0), TAG_LENGTH);
+ f->hmac_running = false;
+
+ return 0;
+}
+
+static int journal_file_hmac_start(JournalFile *f) {
+ uint8_t key[256 / 8]; /* Let's pass 256 bit from FSPRG to HMAC */
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ if (f->hmac_running)
+ return 0;
+
+ /* Prepare HMAC for next cycle */
+ gcry_md_reset(f->hmac);
+ FSPRG_GetKey(fsprg_state(f), key, sizeof(key), 0);
+ gcry_md_setkey(f->hmac, key, sizeof(key));
+
+ f->hmac_running = true;
+
+ return 0;
+}
+
+static int journal_file_get_epoch(JournalFile *f, uint64_t realtime, uint64_t *epoch) {
+ uint64_t t;
+
+ assert(f);
+ assert(epoch);
+ assert(f->authenticate);
+
+ if (le64toh(f->fsprg_header->fsprg_start_usec) == 0 ||
+ le64toh(f->fsprg_header->fsprg_interval_usec) == 0)
+ return -ENOTSUP;
+
+ if (realtime < le64toh(f->fsprg_header->fsprg_start_usec))
+ return -ESTALE;
+
+ t = realtime - le64toh(f->fsprg_header->fsprg_start_usec);
+ t = t / le64toh(f->fsprg_header->fsprg_interval_usec);
+
+ *epoch = t;
+ return 0;
+}
+
+static int journal_file_need_evolve(JournalFile *f, uint64_t realtime) {
+ uint64_t goal, epoch;
+ int r;
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ r = journal_file_get_epoch(f, realtime, &goal);
+ if (r < 0)
+ return r;
+
+ epoch = FSPRG_GetEpoch(fsprg_state(f));
+ if (epoch > goal)
+ return -ESTALE;
+
+ return epoch != goal;
+}
+
+static int journal_file_evolve(JournalFile *f, uint64_t realtime) {
+ uint64_t goal, epoch;
+ int r;
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ r = journal_file_get_epoch(f, realtime, &goal);
+ if (r < 0)
+ return r;
+
+ epoch = FSPRG_GetEpoch(fsprg_state(f));
+ if (epoch < goal)
+ log_debug("Evolving FSPRG key from epoch %llu to %llu.", (unsigned long long) epoch, (unsigned long long) goal);
+
+ for (;;) {
+ if (epoch > goal)
+ return -ESTALE;
+ if (epoch == goal)
+ return 0;
+
+ FSPRG_Evolve(fsprg_state(f));
+ epoch = FSPRG_GetEpoch(fsprg_state(f));
+ }
+}
+
+static int journal_file_maybe_append_tag(JournalFile *f, uint64_t realtime) {
+ int r;
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ r = journal_file_need_evolve(f, realtime);
+ if (r <= 0)
+ return 0;
+
+ r = journal_file_append_tag(f);
+ if (r < 0)
+ return r;
+
+ r = journal_file_evolve(f, realtime);
+ if (r < 0)
+ return r;
+
+ r = journal_file_hmac_start(f);
+ if (r < 0)
+ return r;
+
+ return 0;
+}
+
+static int journal_file_hmac_put_object(JournalFile *f, int type, uint64_t p) {
+ int r;
+ Object *o;
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ r = journal_file_hmac_start(f);
+ if (r < 0)
+ return r;
+
+ r = journal_file_move_to_object(f, type, p, &o);
+ if (r < 0)
+ return r;
+
+ gcry_md_write(f->hmac, o, offsetof(ObjectHeader, payload));
+
+ switch (o->object.type) {
+
+ case OBJECT_DATA:
+ /* All but: hash and payload are mutable */
+ gcry_md_write(f->hmac, &o->data.hash, sizeof(o->data.hash));
+ gcry_md_write(f->hmac, o->data.payload, le64toh(o->object.size) - offsetof(DataObject, payload));
+ break;
+
+ case OBJECT_ENTRY:
+ /* All */
+ gcry_md_write(f->hmac, &o->entry.seqnum, le64toh(o->object.size) - offsetof(EntryObject, seqnum));
+ break;
+
+ case OBJECT_FIELD_HASH_TABLE:
+ case OBJECT_DATA_HASH_TABLE:
+ case OBJECT_ENTRY_ARRAY:
+ /* Nothing: everything is mutable */
+ break;
+
+ case OBJECT_TAG:
+ /* All but the tag itself */
+ gcry_md_write(f->hmac, &o->tag.seqnum, sizeof(o->tag.seqnum));
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+static int journal_file_hmac_put_header(JournalFile *f) {
+ int r;
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ r = journal_file_hmac_start(f);
+ if (r < 0)
+ return r;
+
+ /* All but state+reserved, boot_id, arena_size,
+ * tail_object_offset, n_objects, n_entries, tail_seqnum,
+ * head_entry_realtime, tail_entry_realtime,
+ * tail_entry_monotonic, n_data, n_fields, header_tag */
+
+ gcry_md_write(f->hmac, f->header->signature, offsetof(Header, state) - offsetof(Header, signature));
+ gcry_md_write(f->hmac, &f->header->file_id, offsetof(Header, boot_id) - offsetof(Header, file_id));
+ gcry_md_write(f->hmac, &f->header->seqnum_id, offsetof(Header, arena_size) - offsetof(Header, seqnum_id));
+ gcry_md_write(f->hmac, &f->header->data_hash_table_offset, offsetof(Header, tail_object_offset) - offsetof(Header, data_hash_table_offset));
+ gcry_md_write(f->hmac, &f->header->head_seqnum, offsetof(Header, head_entry_realtime) - offsetof(Header, head_seqnum));
+
+ return 0;
+}
+
+static int journal_file_load_fsprg(JournalFile *f) {
+ int r, fd = -1;
+ char *p = NULL;
+ struct stat st;
+ FSPRGHeader *m = NULL;
+ sd_id128_t machine;
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ r = sd_id128_get_machine(&machine);
+ if (r < 0)
+ return r;
+
+ if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/fsprg",
+ SD_ID128_FORMAT_VAL(machine)) < 0)
+ return -ENOMEM;
+
+ fd = open(p, O_RDWR|O_CLOEXEC|O_NOCTTY, 0600);
+ if (fd < 0) {
+ log_error("Failed to open %s: %m", p);
+ r = -errno;
+ goto finish;
+ }
+
+ if (fstat(fd, &st) < 0) {
+ r = -errno;
+ goto finish;
+ }
+
+ if (st.st_size < (off_t) sizeof(FSPRGHeader)) {
+ r = -ENODATA;
+ goto finish;
+ }
+
+ m = mmap(NULL, PAGE_ALIGN(sizeof(FSPRGHeader)), PROT_READ, MAP_SHARED, fd, 0);
+ if (m == MAP_FAILED) {
+ m = NULL;
+ r = -errno;
+ goto finish;
+ }
+
+ if (memcmp(m->signature, FSPRG_HEADER_SIGNATURE, 8) != 0) {
+ r = -EBADMSG;
+ goto finish;
+ }
+
+ if (m->incompatible_flags != 0) {
+ r = -EPROTONOSUPPORT;
+ goto finish;
+ }
+
+ if (le64toh(m->header_size) < sizeof(FSPRGHeader)) {
+ r = -EBADMSG;
+ goto finish;
+ }
+
+ if (le64toh(m->state_size) != FSPRG_stateinbytes(m->secpar)) {
+ r = -EBADMSG;
+ goto finish;
+ }
+
+ f->fsprg_size = le64toh(m->header_size) + le64toh(m->state_size);
+ if ((uint64_t) st.st_size < f->fsprg_size) {
+ r = -ENODATA;
+ goto finish;
+ }
+
+ if (!sd_id128_equal(machine, m->machine_id)) {
+ r = -EHOSTDOWN;
+ goto finish;
+ }
+
+ if (le64toh(m->fsprg_start_usec) <= 0 ||
+ le64toh(m->fsprg_interval_usec) <= 0) {
+ r = -EBADMSG;
+ goto finish;
+ }
+
+ f->fsprg_header = mmap(NULL, PAGE_ALIGN(f->fsprg_size), PROT_READ|PROT_WRITE, MAP_SHARED, fd, 0);
+ if (f->fsprg_header == MAP_FAILED) {
+ f->fsprg_header = NULL;
+ r = -errno;
+ goto finish;
+ }
+
+ r = 0;
+
+finish:
+ if (m)
+ munmap(m, PAGE_ALIGN(sizeof(FSPRGHeader)));
+
+ if (fd >= 0)
+ close_nointr_nofail(fd);
+
+ free(p);
+ return r;
+}
+
+static int journal_file_setup_hmac(JournalFile *f) {
+ gcry_error_t e;
+
+ if (!f->authenticate)
+ return 0;
+
+ e = gcry_md_open(&f->hmac, GCRY_MD_SHA256, GCRY_MD_FLAG_HMAC);
+ if (e != 0)
+ return -ENOTSUP;
+
+ return 0;
+}
+
+static int journal_file_append_first_tag(JournalFile *f) {
+ int r;
+ uint64_t p;
+
+ if (!f->authenticate)
+ return 0;
+
+ log_debug("Calculating first tag...");
+
+ r = journal_file_hmac_put_header(f);
+ if (r < 0)
+ return r;
+
+ p = le64toh(f->header->field_hash_table_offset);
+ if (p < offsetof(Object, hash_table.items))
+ return -EINVAL;
+ p -= offsetof(Object, hash_table.items);
+
+ r = journal_file_hmac_put_object(f, OBJECT_FIELD_HASH_TABLE, p);
+ if (r < 0)
+ return r;
+
+ p = le64toh(f->header->data_hash_table_offset);
+ if (p < offsetof(Object, hash_table.items))
+ return -EINVAL;
+ p -= offsetof(Object, hash_table.items);
+
+ r = journal_file_hmac_put_object(f, OBJECT_DATA_HASH_TABLE, p);
+ if (r < 0)
+ return r;
+
+ r = journal_file_append_tag(f);
+ if (r < 0)
+ return r;
+
+ return 0;
+}
+
+void journal_file_dump(JournalFile *f) {
+ Object *o;
+ int r;
+ uint64_t p;
+
+ assert(f);
+
+ journal_file_print_header(f);
+
+ p = le64toh(f->header->header_size);
+ while (p != 0) {
+ r = journal_file_move_to_object(f, -1, p, &o);
+ if (r < 0)
+ goto fail;
+
+ switch (o->object.type) {
+
+ case OBJECT_UNUSED:
+ printf("Type: OBJECT_UNUSED\n");
+ break;
case OBJECT_DATA:
printf("Type: OBJECT_DATA\n");
case OBJECT_ENTRY_ARRAY:
printf("Type: OBJECT_ENTRY_ARRAY\n");
break;
+
+ case OBJECT_TAG:
+ printf("Type: OBJECT_TAG %llu\n",
+ (unsigned long long) le64toh(o->tag.seqnum));
+ break;
}
if (o->object.flags & OBJECT_COMPRESSED)
log_error("File corrupt");
}
+void journal_file_print_header(JournalFile *f) {
+ char a[33], b[33], c[33];
+ char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX];
+
+ assert(f);
+
+ printf("File Path: %s\n"
+ "File ID: %s\n"
+ "Machine ID: %s\n"
+ "Boot ID: %s\n"
+ "Sequential Number ID: %s\n"
+ "State: %s\n"
+ "Compatible Flags:%s%s\n"
+ "Incompatible Flags:%s%s\n"
+ "Header size: %llu\n"
+ "Arena size: %llu\n"
+ "Data Hash Table Size: %llu\n"
+ "Field Hash Table Size: %llu\n"
+ "Objects: %llu\n"
+ "Entry Objects: %llu\n"
+ "Rotate Suggested: %s\n"
+ "Head Sequential Number: %llu\n"
+ "Tail Sequential Number: %llu\n"
+ "Head Realtime Timestamp: %s\n"
+ "Tail Realtime Timestamp: %s\n",
+ f->path,
+ sd_id128_to_string(f->header->file_id, a),
+ sd_id128_to_string(f->header->machine_id, b),
+ sd_id128_to_string(f->header->boot_id, c),
+ sd_id128_to_string(f->header->seqnum_id, c),
+ f->header->state == STATE_OFFLINE ? "offline" :
+ f->header->state == STATE_ONLINE ? "online" :
+ f->header->state == STATE_ARCHIVED ? "archived" : "unknown",
+ (f->header->compatible_flags & HEADER_COMPATIBLE_AUTHENTICATED) ? " AUTHENTICATED" : "",
+ (f->header->compatible_flags & ~HEADER_COMPATIBLE_AUTHENTICATED) ? " ???" : "",
+ (f->header->incompatible_flags & HEADER_INCOMPATIBLE_COMPRESSED) ? " COMPRESSED" : "",
+ (f->header->incompatible_flags & ~HEADER_INCOMPATIBLE_COMPRESSED) ? " ???" : "",
+ (unsigned long long) le64toh(f->header->header_size),
+ (unsigned long long) le64toh(f->header->arena_size),
+ (unsigned long long) le64toh(f->header->data_hash_table_size) / sizeof(HashItem),
+ (unsigned long long) le64toh(f->header->field_hash_table_size) / sizeof(HashItem),
+ (unsigned long long) le64toh(f->header->n_objects),
+ (unsigned long long) le64toh(f->header->n_entries),
+ yes_no(journal_file_rotate_suggested(f)),
+ (unsigned long long) le64toh(f->header->head_seqnum),
+ (unsigned long long) le64toh(f->header->tail_seqnum),
+ format_timestamp(x, sizeof(x), le64toh(f->header->head_entry_realtime)),
+ format_timestamp(y, sizeof(y), le64toh(f->header->tail_entry_realtime)));
+
+ if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
+ printf("Data Objects: %llu\n"
+ "Data Hash Table Fill: %.1f%%\n",
+ (unsigned long long) le64toh(f->header->n_data),
+ 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))));
+
+ if (JOURNAL_HEADER_CONTAINS(f->header, n_fields))
+ printf("Field Objects: %llu\n"
+ "Field Hash Table Fill: %.1f%%\n",
+ (unsigned long long) le64toh(f->header->n_fields),
+ 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))));
+}
+
int journal_file_open(
const char *fname,
int flags,
mode_t mode,
+ bool compress,
+ bool authenticate,
+ JournalMetrics *metrics,
JournalFile *template,
JournalFile **ret) {
(flags & O_ACCMODE) != O_RDWR)
return -EINVAL;
+ if (!endswith(fname, ".journal"))
+ return -EINVAL;
+
f = new0(JournalFile, 1);
if (!f)
return -ENOMEM;
f->fd = -1;
- f->flags = flags;
f->mode = mode;
- f->writable = (flags & O_ACCMODE) != O_RDONLY;
- f->prot = prot_from_flags(flags);
- f->metrics.max_size = DEFAULT_MAX_SIZE;
- f->metrics.min_size = DEFAULT_MIN_SIZE;
- f->metrics.keep_free = DEFAULT_KEEP_FREE;
+ f->flags = flags;
+ f->prot = prot_from_flags(flags);
+ f->writable = (flags & O_ACCMODE) != O_RDONLY;
+ f->compress = compress;
+ f->authenticate = authenticate;
f->path = strdup(fname);
if (!f->path) {
if (f->last_stat.st_size == 0 && f->writable) {
newly_created = true;
+ /* Try to load the FSPRG state, and if we can't, then
+ * just don't do authentication */
+ r = journal_file_load_fsprg(f);
+ if (r < 0)
+ f->authenticate = false;
+
r = journal_file_init_header(f, template);
if (r < 0)
goto fail;
}
}
- if (f->last_stat.st_size < (off_t) sizeof(Header)) {
+ if (f->last_stat.st_size < (off_t) HEADER_SIZE_MIN) {
r = -EIO;
goto fail;
}
goto fail;
}
+ if (!newly_created && f->writable) {
+ r = journal_file_load_fsprg(f);
+ if (r < 0)
+ goto fail;
+ }
+
if (f->writable) {
+ if (metrics) {
+ journal_default_metrics(metrics, f->fd);
+ f->metrics = *metrics;
+ } else if (template)
+ f->metrics = template->metrics;
+
r = journal_file_refresh_header(f);
if (r < 0)
goto fail;
+
+ r = journal_file_setup_hmac(f);
+ if (r < 0)
+ goto fail;
}
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;
+
+ r = journal_file_append_first_tag(f);
+ if (r < 0)
+ goto fail;
}
r = journal_file_map_field_hash_table(f);
return r;
}
-int journal_file_rotate(JournalFile **f) {
+int journal_file_rotate(JournalFile **f, bool compress, bool authenticate) {
char *p;
size_t l;
JournalFile *old_file, *new_file = NULL;
l = strlen(old_file->path);
- p = new(char, l + 1 + 16 + 1 + 32 + 1 + 16 + 1);
+ p = new(char, l + 1 + 32 + 1 + 16 + 1 + 16 + 1);
if (!p)
return -ENOMEM;
sd_id128_to_string(old_file->header->seqnum_id, p + l - 8 + 1);
snprintf(p + l - 8 + 1 + 32, 1 + 16 + 1 + 16 + 8 + 1,
"-%016llx-%016llx.journal",
- (unsigned long long) le64toh((*f)->header->seqnum),
+ (unsigned long long) le64toh((*f)->header->tail_seqnum),
(unsigned long long) le64toh((*f)->header->tail_entry_realtime));
r = rename(old_file->path, p);
if (r < 0)
return -errno;
- old_file->header->state = le32toh(STATE_ARCHIVED);
+ old_file->header->state = STATE_ARCHIVED;
- r = journal_file_open(old_file->path, old_file->flags, old_file->mode, old_file, &new_file);
+ r = journal_file_open(old_file->path, old_file->flags, old_file->mode, compress, authenticate, NULL, old_file, &new_file);
journal_file_close(old_file);
*f = new_file;
return r;
}
+int journal_file_open_reliably(
+ const char *fname,
+ int flags,
+ mode_t mode,
+ bool compress,
+ bool authenticate,
+ JournalMetrics *metrics,
+ JournalFile *template,
+ JournalFile **ret) {
+
+ int r;
+ size_t l;
+ char *p;
+
+ r = journal_file_open(fname, flags, mode, compress, authenticate, metrics, template, ret);
+ if (r != -EBADMSG && /* corrupted */
+ r != -ENODATA && /* truncated */
+ r != -EHOSTDOWN && /* other machine */
+ r != -EPROTONOSUPPORT && /* incompatible feature */
+ r != -EBUSY && /* unclean shutdown */
+ r != -ESHUTDOWN /* already archived */)
+ return r;
+
+ if ((flags & O_ACCMODE) == O_RDONLY)
+ return r;
+
+ 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 (asprintf(&p, "%.*s@%016llx-%016llx.journal~",
+ (int) (l-8), fname,
+ (unsigned long long) now(CLOCK_REALTIME),
+ random_ull()) < 0)
+ return -ENOMEM;
+
+ r = rename(fname, p);
+ free(p);
+ if (r < 0)
+ return -errno;
+
+ log_warning("File %s corrupted or uncleanly shut down, renaming and replacing.", fname);
+
+ return journal_file_open(fname, flags, mode, compress, authenticate, metrics, template, ret);
+}
+
struct vacuum_info {
off_t usage;
char *filename;
uint64_t realtime;
sd_id128_t seqnum_id;
uint64_t seqnum;
+
+ bool have_seqnum;
};
static int vacuum_compare(const void *_a, const void *_b) {
a = _a;
b = _b;
- if (sd_id128_equal(a->seqnum_id, b->seqnum_id)) {
+ 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 if (a->realtime > b->realtime)
return 1;
- else
+ else if (a->have_seqnum && b->have_seqnum)
return memcmp(&a->seqnum_id, &b->seqnum_id, 16);
+ else
+ return strcmp(a->filename, b->filename);
}
int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t min_free) {
assert(directory);
if (max_use <= 0)
- max_use = DEFAULT_MAX_USE;
+ return 0;
d = opendir(directory);
if (!d)
size_t q;
struct stat st;
char *p;
- unsigned long long seqnum, realtime;
+ unsigned long long seqnum = 0, realtime;
sd_id128_t seqnum_id;
+ bool have_seqnum;
k = readdir_r(d, &buf, &de);
if (k != 0) {
if (!de)
break;
- if (!dirent_is_file_with_suffix(de, ".journal"))
+ 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 (q < 1 + 32 + 1 + 16 + 1 + 16 + 8)
- continue;
+ if (endswith(de->d_name, ".journal")) {
- 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;
+ /* Vacuum archived files */
- if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0)
- continue;
+ if (q < 1 + 32 + 1 + 16 + 1 + 16 + 8)
+ continue;
- if (!S_ISREG(st.st_mode))
- 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;
- }
+ 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;
- }
+ 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 (sscanf(de->d_name + q-8-16-1-16, "%16llx-%16llx.journal", &seqnum, &realtime) != 2) {
- free(p);
+ 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;
}
list[n_list].filename = p;
- list[n_list].usage = (uint64_t) st.st_blksize * (uint64_t) st.st_blocks;
+ 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);
+ if (n_list > 0)
+ qsort(list, n_list, sizeof(struct vacuum_info), vacuum_compare);
for(i = 0; i < n_list; i++) {
struct statvfs ss;
break;
if (unlinkat(dirfd(d), list[i].filename, 0) >= 0) {
- log_debug("Deleted archived journal %s/%s.", directory, list[i].filename);
+ 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);
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;
+
+ n = journal_file_entry_n_items(o);
+ items = alloca(sizeof(EntryItem) * n);
+
+ for (i = 0; i < n; i++) {
+ uint64_t l, h;
+ le64_t le_hash;
+ 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[FORMAT_BYTES_MAX], b[FORMAT_BYTES_MAX], c[FORMAT_BYTES_MAX], d[FORMAT_BYTES_MAX];
+
+ 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_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));
+}
+
+int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *to) {
+ assert(f);
+ assert(from || to);
+
+ if (from) {
+ if (f->header->head_entry_realtime == 0)
+ return -ENOENT;
+
+ *from = le64toh(f->header->head_entry_realtime);
+ }
+
+ if (to) {
+ if (f->header->tail_entry_realtime == 0)
+ return -ENOENT;
+
+ *to = le64toh(f->header->tail_entry_realtime);
+ }
+
+ return 1;
+}
+
+int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, usec_t *from, usec_t *to) {
+ char t[9+32+1] = "_BOOT_ID=";
+ Object *o;
+ uint64_t p;
+ int r;
+
+ assert(f);
+ assert(from || to);
+
+ sd_id128_to_string(boot_id, t + 9);
+
+ r = journal_file_find_data_object(f, t, strlen(t), &o, &p);
+ if (r <= 0)
+ return r;
+
+ if (le64toh(o->data.n_entries) <= 0)
+ return 0;
+
+ if (from) {
+ r = journal_file_move_to_object(f, OBJECT_ENTRY, le64toh(o->data.entry_offset), &o);
+ if (r < 0)
+ return r;
+
+ *from = le64toh(o->entry.monotonic);
+ }
+
+ if (to) {
+ r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
+ if (r < 0)
+ return r;
+
+ r = generic_array_get_plus_one(f,
+ le64toh(o->data.entry_offset),
+ le64toh(o->data.entry_array_offset),
+ le64toh(o->data.n_entries)-1,
+ &o, NULL);
+ if (r <= 0)
+ return r;
+
+ *to = le64toh(o->entry.monotonic);
+ }
+
+ return 1;
+}
+
+bool journal_file_rotate_suggested(JournalFile *f) {
+ assert(f);
+
+ /* If we gained new header fields we gained new features,
+ * hence suggest a rotation */
+ if (le64toh(f->header->header_size) < sizeof(Header)) {
+ log_debug("%s uses an outdated header, suggesting rotation.", f->path);
+ return true;
+ }
+
+ /* Let's check if the hash tables grew over a certain fill
+ * level (75%, borrowing this value from Java's hash table
+ * implementation), and if so suggest a rotation. To calculate
+ * the fill level we need the n_data field, which only exists
+ * in newer versions. */
+
+ if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
+ if (le64toh(f->header->n_data) * 4ULL > (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)) * 3ULL) {
+ log_debug("Data hash table of %s has a fill level at %.1f (%llu of %llu items, %llu file size, %llu bytes per hash table item), suggesting rotation.",
+ f->path,
+ 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))),
+ (unsigned long long) le64toh(f->header->n_data),
+ (unsigned long long) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)),
+ (unsigned long long) (f->last_stat.st_size),
+ (unsigned long long) (f->last_stat.st_size / le64toh(f->header->n_data)));
+ return true;
+ }
+
+ if (JOURNAL_HEADER_CONTAINS(f->header, n_fields))
+ if (le64toh(f->header->n_fields) * 4ULL > (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)) * 3ULL) {
+ log_debug("Field hash table of %s has a fill level at %.1f (%llu of %llu items), suggesting rotation.",
+ f->path,
+ 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))),
+ (unsigned long long) le64toh(f->header->n_fields),
+ (unsigned long long) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)));
+ return true;
+ }
+
+ return false;
+}
~ianmdlvl / elogind.git / blobdiff