alex/bcachefs-tools
1
0
Fork 0
mirror of https://github.com/koverstreet/bcachefs-tools.git synced 2025-02-23 00:00:02 +03:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

Update bcachefs sources to 26494335d114 bcachefs: improve move_gap()

Browse Source 
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This commit is contained in:
Kent Overstreet 2024-02-25 21:11:52 -05:00
parent 55a2b65bd0
commit 62ab2e5216
17 changed files with 75 additions and 128 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
.bcachefs_revision
View File

@ -1 +1 @@
a623b3063d5d6ab6652314028531060d5a0d192e
26494335d114f7813a7fc499bbacb4a74d613b6f

8
c_src/cmd_list_journal.c
View File

@ -72,14 +72,10 @@ static bool entry_matches_transaction_filter(struct jset_entry *entry,
{
if (entry->type == BCH_JSET_ENTRY_btree_root ||
entry->type == BCH_JSET_ENTRY_btree_keys ||
entry->type == BCH_JSET_ENTRY_overwrite) {
struct bkey_i *k;
entry->type == BCH_JSET_ENTRY_overwrite)
jset_entry_for_each_key(entry, k)
if (bkey_matches_filter(filter, entry, k))
return true;
}
return false;
}
@ -100,8 +96,6 @@ static bool should_print_transaction(struct jset_entry *entry, struct jset_entry
static bool should_print_entry(struct jset_entry *entry, d_btree_id filter)
{
struct bkey_i *k;
if (!filter.nr)
return true;

6
libbcachefs/bcachefs.h
View File

@ -668,6 +668,8 @@ struct journal_seq_blacklist_table {
};
struct journal_keys {
/* must match layout in darray_types.h */
size_t nr, size;
struct journal_key {
u64 journal_seq;
u32 journal_offset;
@ -676,15 +678,13 @@ struct journal_keys {
bool allocated;
bool overwritten;
struct bkey_i *k;
} *d;
} *data;
/*
* Gap buffer: instead of all the empty space in the array being at the
* end of the buffer - from @nr to @size - the empty space is at @gap.
* This means that sequential insertions are O(n) instead of O(n^2).
*/
size_t gap;
size_t nr;
size_t size;
atomic_t ref;
bool initial_ref_held;
};

4
libbcachefs/btree_iter.c
View File

@ -2156,7 +2156,9 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
* isn't monotonically increasing before FILTER_SNAPSHOTS, and
* that's what we check against in extents mode:
*/
if (k.k->p.inode > end.inode)
if (unlikely(!(iter->flags & BTREE_ITER_IS_EXTENTS)
? bkey_gt(k.k->p, end)
: k.k->p.inode > end.inode))
goto end;
if (iter->update_path &&

126
libbcachefs/btree_journal_iter.c
View File

@ -42,7 +42,7 @@ static inline size_t idx_to_pos(struct journal_keys *keys, size_t idx)
static inline struct journal_key *idx_to_key(struct journal_keys *keys, size_t idx)
{
return keys->d + idx_to_pos(keys, idx);
return keys->data + idx_to_pos(keys, idx);
}
static size_t __bch2_journal_key_search(struct journal_keys *keys,
@ -182,10 +182,10 @@ int bch2_journal_key_insert_take(struct bch_fs *c, enum btree_id id,
BUG_ON(test_bit(BCH_FS_rw, &c->flags));
if (idx < keys->size &&
journal_key_cmp(&n, &keys->d[idx]) == 0) {
if (keys->d[idx].allocated)
kfree(keys->d[idx].k);
keys->d[idx] = n;
journal_key_cmp(&n, &keys->data[idx]) == 0) {
if (keys->data[idx].allocated)
kfree(keys->data[idx].k);
keys->data[idx] = n;
return 0;
}
@ -198,17 +198,17 @@ int bch2_journal_key_insert_take(struct bch_fs *c, enum btree_id id,
.size = max_t(size_t, keys->size, 8) * 2,
};
new_keys.d = kvmalloc_array(new_keys.size, sizeof(new_keys.d[0]), GFP_KERNEL);
if (!new_keys.d) {
new_keys.data = kvmalloc_array(new_keys.size, sizeof(new_keys.data[0]), GFP_KERNEL);
if (!new_keys.data) {
bch_err(c, "%s: error allocating new key array (size %zu)",
__func__, new_keys.size);
return -BCH_ERR_ENOMEM_journal_key_insert;
}
/* Since @keys was full, there was no gap: */
memcpy(new_keys.d, keys->d, sizeof(keys->d[0]) * keys->nr);
kvfree(keys->d);
keys->d = new_keys.d;
memcpy(new_keys.data, keys->data, sizeof(keys->data[0]) * keys->nr);
kvfree(keys->data);
keys->data = new_keys.data;
keys->nr = new_keys.nr;
keys->size = new_keys.size;
@ -218,11 +218,10 @@ int bch2_journal_key_insert_take(struct bch_fs *c, enum btree_id id,
journal_iters_move_gap(c, keys->gap, idx);
move_gap(keys->d, keys->nr, keys->size, keys->gap, idx);
keys->gap = idx;
move_gap(keys, idx);
keys->nr++;
keys->d[keys->gap++] = n;
keys->data[keys->gap++] = n;
journal_iters_fix(c);
@ -269,10 +268,10 @@ void bch2_journal_key_overwritten(struct bch_fs *c, enum btree_id btree,
size_t idx = bch2_journal_key_search(keys, btree, level, pos);
if (idx < keys->size &&
keys->d[idx].btree_id == btree &&
keys->d[idx].level == level &&
bpos_eq(keys->d[idx].k->k.p, pos))
keys->d[idx].overwritten = true;
keys->data[idx].btree_id == btree &&
keys->data[idx].level == level &&
bpos_eq(keys->data[idx].k->k.p, pos))
keys->data[idx].overwritten = true;
}
static void bch2_journal_iter_advance(struct journal_iter *iter)
@ -286,16 +285,16 @@ static void bch2_journal_iter_advance(struct journal_iter *iter)
static struct bkey_s_c bch2_journal_iter_peek(struct journal_iter *iter)
{
struct journal_key *k = iter->keys->d + iter->idx;
struct journal_key *k = iter->keys->data + iter->idx;
while (k < iter->keys->d + iter->keys->size &&
while (k < iter->keys->data + iter->keys->size &&
k->btree_id == iter->btree_id &&
k->level == iter->level) {
if (!k->overwritten)
return bkey_i_to_s_c(k->k);
bch2_journal_iter_advance(iter);
k = iter->keys->d + iter->idx;
k = iter->keys->data + iter->idx;
}
return bkey_s_c_null;
@ -467,22 +466,20 @@ static int journal_sort_key_cmp(const void *_l, const void *_r)
void bch2_journal_keys_put(struct bch_fs *c)
{
struct journal_keys *keys = &c->journal_keys;
struct journal_key *i;
BUG_ON(atomic_read(&keys->ref) <= 0);
if (!atomic_dec_and_test(&keys->ref))
return;
move_gap(keys->d, keys->nr, keys->size, keys->gap, keys->nr);
keys->gap = keys->nr;
move_gap(keys, keys->nr);
for (i = keys->d; i < keys->d + keys->nr; i++)
darray_for_each(*keys, i)
if (i->allocated)
kfree(i->k);
kvfree(keys->d);
keys->d = NULL;
kvfree(keys->data);
keys->data = NULL;
keys->nr = keys->gap = keys->size = 0;
bch2_journal_entries_free(c);
@ -490,62 +487,27 @@ void bch2_journal_keys_put(struct bch_fs *c)
static void __journal_keys_sort(struct journal_keys *keys)
{
struct journal_key *src, *dst;
sort(keys->data, keys->nr, sizeof(keys->data[0]), journal_sort_key_cmp, NULL);
sort(keys->d, keys->nr, sizeof(keys->d[0]), journal_sort_key_cmp, NULL);
struct journal_key *dst = keys->data;
src = dst = keys->d;
while (src < keys->d + keys->nr) {
while (src + 1 < keys->d + keys->nr &&
darray_for_each(*keys, src) {
if (src + 1 < &darray_top(*keys) &&
!journal_key_cmp(src, src + 1))
src++;
continue;
*dst++ = *src++;
*dst++ = *src;
}
keys->nr = dst - keys->d;
keys->nr = dst - keys->data;
}
int bch2_journal_keys_sort(struct bch_fs *c)
{
struct genradix_iter iter;
struct journal_replay *i, **_i;
struct jset_entry *entry;
struct bkey_i *k;
struct journal_keys *keys = &c->journal_keys;
size_t nr_keys = 0, nr_read = 0;
genradix_for_each(&c->journal_entries, iter, _i) {
i = *_i;
if (!i || i->ignore)
continue;
for_each_jset_key(k, entry, &i->j)
nr_keys++;
}
if (!nr_keys)
return 0;
keys->size = roundup_pow_of_two(nr_keys);
keys->d = kvmalloc_array(keys->size, sizeof(keys->d[0]), GFP_KERNEL);
if (!keys->d) {
bch_err(c, "Failed to allocate buffer for sorted journal keys (%zu keys); trying slowpath",
nr_keys);
do {
keys->size >>= 1;
keys->d = kvmalloc_array(keys->size, sizeof(keys->d[0]), GFP_KERNEL);
} while (!keys->d && keys->size > nr_keys / 8);
if (!keys->d) {
bch_err(c, "Failed to allocate %zu size buffer for sorted journal keys; exiting",
keys->size);
return -BCH_ERR_ENOMEM_journal_keys_sort;
}
}
size_t nr_read = 0;
genradix_for_each(&c->journal_entries, iter, _i) {
i = *_i;
@ -556,17 +518,7 @@ int bch2_journal_keys_sort(struct bch_fs *c)
cond_resched();
for_each_jset_key(k, entry, &i->j) {
if (keys->nr == keys->size) {
__journal_keys_sort(keys);
if (keys->nr > keys->size * 7 / 8) {
bch_err(c, "Too many journal keys for slowpath; have %zu compacted, buf size %zu, processed %zu/%zu",
keys->nr, keys->size, nr_read, nr_keys);
return -BCH_ERR_ENOMEM_journal_keys_sort;
}
}
keys->d[keys->nr++] = (struct journal_key) {
struct journal_key n = (struct journal_key) {
.btree_id = entry->btree_id,
.level = entry->level,
.k = k,
@ -574,6 +526,18 @@ int bch2_journal_keys_sort(struct bch_fs *c)
.journal_offset = k->_data - i->j._data,
};
if (darray_push(keys, n)) {
__journal_keys_sort(keys);
if (keys->nr * 8 > keys->size * 7) {
bch_err(c, "Too many journal keys for slowpath; have %zu compacted, buf size %zu, processed %zu keys at seq %llu",
keys->nr, keys->size, nr_read, le64_to_cpu(i->j.seq));
return -BCH_ERR_ENOMEM_journal_keys_sort;
}
BUG_ON(darray_push(keys, n));
}
nr_read++;
}
}
@ -581,6 +545,6 @@ int bch2_journal_keys_sort(struct bch_fs *c)
__journal_keys_sort(keys);
keys->gap = keys->nr;
bch_verbose(c, "Journal keys: %zu read, %zu after sorting and compacting", nr_keys, keys->nr);
bch_verbose(c, "Journal keys: %zu read, %zu after sorting and compacting", nr_read, keys->nr);
return 0;
}

2
libbcachefs/btree_update.c
View File

@ -848,7 +848,7 @@ __bch2_fs_log_msg(struct bch_fs *c, unsigned commit_flags, const char *fmt,
if (ret)
goto err;
if (!test_bit(JOURNAL_RUNNING, &c->journal.flags)) {
if (!test_bit(JOURNAL_STARTED, &c->journal.flags)) {
ret = darray_make_room(&c->journal.early_journal_entries, jset_u64s(u64s));
if (ret)
goto err;

2
libbcachefs/btree_write_buffer.c
View File

@ -574,8 +574,6 @@ void bch2_journal_keys_to_write_buffer_end(struct bch_fs *c, struct journal_keys
static int bch2_journal_keys_to_write_buffer(struct bch_fs *c, struct journal_buf *buf)
{
struct journal_keys_to_wb dst;
struct jset_entry *entry;
struct bkey_i *k;
int ret = 0;
bch2_journal_keys_to_write_buffer_start(c, &dst, le64_to_cpu(buf->data->seq));

5
libbcachefs/journal.c
View File

@ -1183,13 +1183,12 @@ void bch2_fs_journal_stop(struct journal *j)
bch2_journal_meta(j);
journal_quiesce(j);
cancel_delayed_work_sync(&j->write_work);
BUG_ON(!bch2_journal_error(j) &&
test_bit(JOURNAL_REPLAY_DONE, &j->flags) &&
j->last_empty_seq != journal_cur_seq(j));
clear_bit(JOURNAL_RUNNING, &j->flags);
cancel_delayed_work_sync(&j->write_work);
}
int bch2_fs_journal_start(struct journal *j, u64 cur_seq)
@ -1263,7 +1262,7 @@ int bch2_fs_journal_start(struct journal *j, u64 cur_seq)
spin_lock(&j->lock);
set_bit(JOURNAL_RUNNING, &j->flags);
set_bit(JOURNAL_STARTED, &j->flags);
j->last_flush_write = jiffies;
j->reservations.idx = j->reservations.unwritten_idx = journal_cur_seq(j);

4
libbcachefs/journal.h
View File

@ -372,7 +372,7 @@ static inline int bch2_journal_res_get(struct journal *j, struct journal_res *re
int ret;
EBUG_ON(res->ref);
EBUG_ON(!test_bit(JOURNAL_RUNNING, &j->flags));
EBUG_ON(!test_bit(JOURNAL_STARTED, &j->flags));
res->u64s = u64s;
@ -418,7 +418,7 @@ struct bch_dev;
static inline void bch2_journal_set_replay_done(struct journal *j)
{
BUG_ON(!test_bit(JOURNAL_RUNNING, &j->flags));
BUG_ON(!test_bit(JOURNAL_STARTED, &j->flags));
set_bit(JOURNAL_REPLAY_DONE, &j->flags);
}

3
libbcachefs/journal_io.c
View File

@ -40,7 +40,6 @@ static void bch2_journal_replay_to_text(struct printbuf *out, struct bch_fs *c,
bch2_journal_ptrs_to_text(out, c, j);
struct jset_entry *entry;
for_each_jset_entry_type(entry, &j->j, BCH_JSET_ENTRY_datetime) {
struct jset_entry_datetime *datetime =
container_of(entry, struct jset_entry_datetime, entry);
@ -394,7 +393,6 @@ static int journal_entry_btree_keys_validate(struct bch_fs *c,
static void journal_entry_btree_keys_to_text(struct printbuf *out, struct bch_fs *c,
struct jset_entry *entry)
{
struct bkey_i *k;
bool first = true;
jset_entry_for_each_key(entry, k) {
@ -1814,7 +1812,6 @@ static int bch2_journal_write_prep(struct journal *j, struct journal_buf *w)
if (!wb.wb)
bch2_journal_keys_to_write_buffer_start(c, &wb, seq);
struct bkey_i *k;
jset_entry_for_each_key(i, k) {
ret = bch2_journal_key_to_wb(c, &wb, i->btree_id, k);
if (ret) {

4
libbcachefs/journal_io.h
View File

@ -39,12 +39,12 @@ static inline struct jset_entry *__jset_entry_type_next(struct jset *jset,
}
#define for_each_jset_entry_type(entry, jset, type) \
for (entry = (jset)->start; \
for (struct jset_entry *entry = (jset)->start; \
(entry = __jset_entry_type_next(jset, entry, type)); \
entry = vstruct_next(entry))
#define jset_entry_for_each_key(_e, _k) \
for (_k = (_e)->start; \
for (struct bkey_i *_k = (_e)->start; \
_k < vstruct_last(_e); \
_k = bkey_next(_k))

2
libbcachefs/journal_reclaim.c
View File

@ -831,7 +831,7 @@ bool bch2_journal_flush_pins(struct journal *j, u64 seq_to_flush)
/* time_stats this */
bool did_work = false;
if (!test_bit(JOURNAL_RUNNING, &j->flags))
if (!test_bit(JOURNAL_STARTED, &j->flags))
return false;
closure_wait_event(&j->async_wait,

2
libbcachefs/journal_types.h
View File

@ -131,7 +131,7 @@ enum journal_space_from {
enum journal_flags {
JOURNAL_REPLAY_DONE,
JOURNAL_RUNNING,
JOURNAL_STARTED,
JOURNAL_MAY_SKIP_FLUSH,
JOURNAL_NEED_FLUSH_WRITE,
};

15
libbcachefs/recovery.c
View File

@ -57,8 +57,8 @@ static void drop_alloc_keys(struct journal_keys *keys)
size_t src, dst;
for (src = 0, dst = 0; src < keys->nr; src++)
if (!btree_id_is_alloc(keys->d[src].btree_id))
keys->d[dst++] = keys->d[src];
if (!btree_id_is_alloc(keys->data[src].btree_id))
keys->data[dst++] = keys->data[src];
keys->nr = dst;
}
@ -70,9 +70,7 @@ static void drop_alloc_keys(struct journal_keys *keys)
*/
static void zero_out_btree_mem_ptr(struct journal_keys *keys)
{
struct journal_key *i;
for (i = keys->d; i < keys->d + keys->nr; i++)
darray_for_each(*keys, i)
if (i->k->k.type == KEY_TYPE_btree_ptr_v2)
bkey_i_to_btree_ptr_v2(i->k)->v.mem_ptr = 0;
}
@ -166,11 +164,9 @@ static int bch2_journal_replay(struct bch_fs *c)
* efficient - better locality of btree access - but some might fail if
* that would cause a journal deadlock.
*/
for (size_t i = 0; i < keys->nr; i++) {
darray_for_each(*keys, k) {
cond_resched();
struct journal_key *k = keys->d + i;
/* Skip fastpath if we're low on space in the journal */
ret = c->journal.watermark ? -1 :
commit_do(trans, NULL, NULL,
@ -524,8 +520,7 @@ static int bch2_set_may_go_rw(struct bch_fs *c)
* setting journal_key->overwritten: it will be accessed by multiple
* threads
*/
move_gap(keys->d, keys->nr, keys->size, keys->gap, keys->nr);
keys->gap = keys->nr;
move_gap(keys, keys->nr);
set_bit(BCH_FS_may_go_rw, &c->flags);

6
libbcachefs/super.c
View File

@ -286,13 +286,8 @@ static void __bch2_fs_read_only(struct bch_fs *c)
if (test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags) &&
!test_bit(BCH_FS_emergency_ro, &c->flags))
set_bit(BCH_FS_clean_shutdown, &c->flags);
bch2_fs_journal_stop(&c->journal);
bch_info(c, "%sshutdown complete, journal seq %llu",
test_bit(BCH_FS_clean_shutdown, &c->flags) ? "" : "un",
c->journal.seq_ondisk);
/*
* After stopping journal:
*/
@ -466,7 +461,6 @@ static int __bch2_fs_read_write(struct bch_fs *c, bool early)
* at least one non-flush write in the journal or recovery will fail:
*/
set_bit(JOURNAL_NEED_FLUSH_WRITE, &c->journal.flags);
set_bit(JOURNAL_RUNNING, &c->journal.flags);
for_each_rw_member(c, ca)
bch2_dev_allocator_add(c, ca);

2
libbcachefs/util.c
View File

@ -289,7 +289,7 @@ int bch2_save_backtrace(bch_stacktrace *stack, struct task_struct *task, unsigne
do {
nr_entries = stack_trace_save_tsk(task, stack->data, stack->size, skipnr + 1);
} while (nr_entries == stack->size &&
!(ret = darray_make_room(stack, stack->size * 2)));
!(ret = darray_make_room_gfp(stack, stack->size * 2, gfp)));
stack->nr = nr_entries;
up_read(&task->signal->exec_update_lock);

8
libbcachefs/util.h
View File

@ -651,8 +651,12 @@ static inline void __move_gap(void *array, size_t element_size,
}
/* Move the gap in a gap buffer: */
#define move_gap(_array, _nr, _size, _old_gap, _new_gap) \
__move_gap(_array, sizeof(_array[0]), _nr, _size, _old_gap, _new_gap)
#define move_gap(_d, _new_gap) \
do { \
__move_gap((_d)->data, sizeof((_d)->data[0]), \
(_d)->nr, (_d)->size, (_d)->gap, _new_gap); \
(_d)->gap = _new_gap; \
} while (0)
#define bubble_sort(_base, _nr, _cmp) \
do { \