2017-03-17 06:18:50 +00:00
|
|
|
// SPDX-License-Identifier: GPL-2.0
|
|
|
|
|
|
|
|
#include "bcachefs.h"
|
|
|
|
#include "bkey_methods.h"
|
2018-11-27 23:30:56 +00:00
|
|
|
#include "bkey_sort.h"
|
2017-03-17 06:18:50 +00:00
|
|
|
#include "btree_cache.h"
|
|
|
|
#include "btree_io.h"
|
|
|
|
#include "btree_iter.h"
|
|
|
|
#include "btree_locking.h"
|
|
|
|
#include "btree_update.h"
|
|
|
|
#include "btree_update_interior.h"
|
|
|
|
#include "buckets.h"
|
|
|
|
#include "checksum.h"
|
|
|
|
#include "debug.h"
|
|
|
|
#include "error.h"
|
|
|
|
#include "extents.h"
|
2023-09-10 22:05:17 +00:00
|
|
|
#include "io_write.h"
|
2017-03-17 06:18:50 +00:00
|
|
|
#include "journal_reclaim.h"
|
|
|
|
#include "journal_seq_blacklist.h"
|
2023-08-04 00:37:32 +00:00
|
|
|
#include "recovery.h"
|
2017-03-17 06:18:50 +00:00
|
|
|
#include "super-io.h"
|
|
|
|
#include "trace.h"
|
|
|
|
|
2020-03-27 21:38:51 +00:00
|
|
|
#include <linux/sched/mm.h>
|
|
|
|
|
2021-07-11 03:03:15 +00:00
|
|
|
void bch2_btree_node_io_unlock(struct btree *b)
|
|
|
|
{
|
|
|
|
EBUG_ON(!btree_node_write_in_flight(b));
|
|
|
|
|
2021-07-10 17:44:42 +00:00
|
|
|
clear_btree_node_write_in_flight_inner(b);
|
2021-07-11 03:03:15 +00:00
|
|
|
clear_btree_node_write_in_flight(b);
|
|
|
|
wake_up_bit(&b->flags, BTREE_NODE_write_in_flight);
|
|
|
|
}
|
|
|
|
|
|
|
|
void bch2_btree_node_io_lock(struct btree *b)
|
|
|
|
{
|
2023-03-03 05:03:01 +00:00
|
|
|
bch2_assert_btree_nodes_not_locked();
|
|
|
|
|
2021-07-11 03:03:15 +00:00
|
|
|
wait_on_bit_lock_io(&b->flags, BTREE_NODE_write_in_flight,
|
|
|
|
TASK_UNINTERRUPTIBLE);
|
|
|
|
}
|
|
|
|
|
|
|
|
void __bch2_btree_node_wait_on_read(struct btree *b)
|
|
|
|
{
|
|
|
|
wait_on_bit_io(&b->flags, BTREE_NODE_read_in_flight,
|
|
|
|
TASK_UNINTERRUPTIBLE);
|
|
|
|
}
|
|
|
|
|
|
|
|
void __bch2_btree_node_wait_on_write(struct btree *b)
|
|
|
|
{
|
|
|
|
wait_on_bit_io(&b->flags, BTREE_NODE_write_in_flight,
|
|
|
|
TASK_UNINTERRUPTIBLE);
|
|
|
|
}
|
|
|
|
|
|
|
|
void bch2_btree_node_wait_on_read(struct btree *b)
|
|
|
|
{
|
2023-03-03 05:03:01 +00:00
|
|
|
bch2_assert_btree_nodes_not_locked();
|
|
|
|
|
2021-07-11 03:03:15 +00:00
|
|
|
wait_on_bit_io(&b->flags, BTREE_NODE_read_in_flight,
|
|
|
|
TASK_UNINTERRUPTIBLE);
|
|
|
|
}
|
|
|
|
|
|
|
|
void bch2_btree_node_wait_on_write(struct btree *b)
|
|
|
|
{
|
2023-03-03 05:03:01 +00:00
|
|
|
bch2_assert_btree_nodes_not_locked();
|
|
|
|
|
2021-07-11 03:03:15 +00:00
|
|
|
wait_on_bit_io(&b->flags, BTREE_NODE_write_in_flight,
|
|
|
|
TASK_UNINTERRUPTIBLE);
|
|
|
|
}
|
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
static void verify_no_dups(struct btree *b,
|
|
|
|
struct bkey_packed *start,
|
2021-02-20 05:00:23 +00:00
|
|
|
struct bkey_packed *end)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
|
|
|
#ifdef CONFIG_BCACHEFS_DEBUG
|
2019-11-10 04:50:52 +00:00
|
|
|
struct bkey_packed *k, *p;
|
|
|
|
|
|
|
|
if (start == end)
|
|
|
|
return;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2023-03-05 04:05:55 +00:00
|
|
|
for (p = start, k = bkey_p_next(start);
|
2019-11-10 04:50:52 +00:00
|
|
|
k != end;
|
2023-03-05 04:05:55 +00:00
|
|
|
p = k, k = bkey_p_next(k)) {
|
2019-11-10 04:50:52 +00:00
|
|
|
struct bkey l = bkey_unpack_key(b, p);
|
|
|
|
struct bkey r = bkey_unpack_key(b, k);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2022-11-24 08:12:22 +00:00
|
|
|
BUG_ON(bpos_ge(l.p, bkey_start_pos(&r)));
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2019-11-10 04:50:52 +00:00
|
|
|
static void set_needs_whiteout(struct bset *i, int v)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
|
|
|
struct bkey_packed *k;
|
|
|
|
|
2023-03-05 04:05:55 +00:00
|
|
|
for (k = i->start; k != vstruct_last(i); k = bkey_p_next(k))
|
2019-11-10 04:50:52 +00:00
|
|
|
k->needs_whiteout = v;
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
2020-07-25 19:07:37 +00:00
|
|
|
static void btree_bounce_free(struct bch_fs *c, size_t size,
|
2017-03-17 06:18:50 +00:00
|
|
|
bool used_mempool, void *p)
|
|
|
|
{
|
|
|
|
if (used_mempool)
|
|
|
|
mempool_free(p, &c->btree_bounce_pool);
|
|
|
|
else
|
2024-02-01 11:35:46 +00:00
|
|
|
kvfree(p);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
2020-07-25 19:07:37 +00:00
|
|
|
static void *btree_bounce_alloc(struct bch_fs *c, size_t size,
|
2017-03-17 06:18:50 +00:00
|
|
|
bool *used_mempool)
|
|
|
|
{
|
2020-03-27 21:38:51 +00:00
|
|
|
unsigned flags = memalloc_nofs_save();
|
2017-03-17 06:18:50 +00:00
|
|
|
void *p;
|
|
|
|
|
2024-01-16 18:29:59 +00:00
|
|
|
BUG_ON(size > c->opts.btree_node_size);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
*used_mempool = false;
|
2024-02-01 11:35:46 +00:00
|
|
|
p = kvmalloc(size, __GFP_NOWARN|GFP_NOWAIT);
|
2020-03-27 21:38:51 +00:00
|
|
|
if (!p) {
|
|
|
|
*used_mempool = true;
|
2023-05-28 22:02:38 +00:00
|
|
|
p = mempool_alloc(&c->btree_bounce_pool, GFP_NOFS);
|
2020-03-27 21:38:51 +00:00
|
|
|
}
|
|
|
|
memalloc_nofs_restore(flags);
|
|
|
|
return p;
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
2019-11-29 19:08:51 +00:00
|
|
|
static void sort_bkey_ptrs(const struct btree *bt,
|
|
|
|
struct bkey_packed **ptrs, unsigned nr)
|
|
|
|
{
|
|
|
|
unsigned n = nr, a = nr / 2, b, c, d;
|
|
|
|
|
|
|
|
if (!a)
|
|
|
|
return;
|
|
|
|
|
|
|
|
/* Heap sort: see lib/sort.c: */
|
|
|
|
while (1) {
|
|
|
|
if (a)
|
|
|
|
a--;
|
|
|
|
else if (--n)
|
|
|
|
swap(ptrs[0], ptrs[n]);
|
|
|
|
else
|
|
|
|
break;
|
|
|
|
|
|
|
|
for (b = a; c = 2 * b + 1, (d = c + 1) < n;)
|
2020-11-07 17:31:20 +00:00
|
|
|
b = bch2_bkey_cmp_packed(bt,
|
2019-11-29 19:08:51 +00:00
|
|
|
ptrs[c],
|
|
|
|
ptrs[d]) >= 0 ? c : d;
|
|
|
|
if (d == n)
|
|
|
|
b = c;
|
|
|
|
|
|
|
|
while (b != a &&
|
2020-11-07 17:31:20 +00:00
|
|
|
bch2_bkey_cmp_packed(bt,
|
2019-11-29 19:08:51 +00:00
|
|
|
ptrs[a],
|
|
|
|
ptrs[b]) >= 0)
|
|
|
|
b = (b - 1) / 2;
|
|
|
|
c = b;
|
|
|
|
while (b != a) {
|
|
|
|
b = (b - 1) / 2;
|
|
|
|
swap(ptrs[b], ptrs[c]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void bch2_sort_whiteouts(struct bch_fs *c, struct btree *b)
|
|
|
|
{
|
2019-12-14 20:55:29 +00:00
|
|
|
struct bkey_packed *new_whiteouts, **ptrs, **ptrs_end, *k;
|
|
|
|
bool used_mempool = false;
|
2020-07-25 19:07:37 +00:00
|
|
|
size_t bytes = b->whiteout_u64s * sizeof(u64);
|
2019-11-29 19:08:51 +00:00
|
|
|
|
|
|
|
if (!b->whiteout_u64s)
|
|
|
|
return;
|
|
|
|
|
2020-07-25 19:07:37 +00:00
|
|
|
new_whiteouts = btree_bounce_alloc(c, bytes, &used_mempool);
|
2019-11-29 19:08:51 +00:00
|
|
|
|
2020-07-25 19:07:37 +00:00
|
|
|
ptrs = ptrs_end = ((void *) new_whiteouts + bytes);
|
2019-11-29 19:08:51 +00:00
|
|
|
|
2024-01-16 18:29:59 +00:00
|
|
|
for (k = unwritten_whiteouts_start(b);
|
|
|
|
k != unwritten_whiteouts_end(b);
|
2023-03-05 04:05:55 +00:00
|
|
|
k = bkey_p_next(k))
|
2019-12-14 20:55:29 +00:00
|
|
|
*--ptrs = k;
|
2019-11-29 19:08:51 +00:00
|
|
|
|
2019-12-14 20:55:29 +00:00
|
|
|
sort_bkey_ptrs(b, ptrs, ptrs_end - ptrs);
|
2019-11-29 19:08:51 +00:00
|
|
|
|
|
|
|
k = new_whiteouts;
|
|
|
|
|
2019-12-14 20:55:29 +00:00
|
|
|
while (ptrs != ptrs_end) {
|
2023-11-02 23:33:48 +00:00
|
|
|
bkey_p_copy(k, *ptrs);
|
2023-03-05 04:05:55 +00:00
|
|
|
k = bkey_p_next(k);
|
2019-12-14 20:55:29 +00:00
|
|
|
ptrs++;
|
2019-11-29 19:08:51 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
verify_no_dups(b, new_whiteouts,
|
2021-02-20 05:00:23 +00:00
|
|
|
(void *) ((u64 *) new_whiteouts + b->whiteout_u64s));
|
2019-11-29 19:08:51 +00:00
|
|
|
|
2024-01-16 18:29:59 +00:00
|
|
|
memcpy_u64s(unwritten_whiteouts_start(b),
|
2019-11-29 19:08:51 +00:00
|
|
|
new_whiteouts, b->whiteout_u64s);
|
|
|
|
|
2020-07-25 19:07:37 +00:00
|
|
|
btree_bounce_free(c, bytes, used_mempool, new_whiteouts);
|
2019-11-29 19:08:51 +00:00
|
|
|
}
|
|
|
|
|
2019-12-13 18:08:37 +00:00
|
|
|
static bool should_compact_bset(struct btree *b, struct bset_tree *t,
|
|
|
|
bool compacting, enum compact_mode mode)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
2019-12-13 18:08:37 +00:00
|
|
|
if (!bset_dead_u64s(b, t))
|
|
|
|
return false;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2019-12-13 18:08:37 +00:00
|
|
|
switch (mode) {
|
|
|
|
case COMPACT_LAZY:
|
|
|
|
return should_compact_bset_lazy(b, t) ||
|
|
|
|
(compacting && !bset_written(b, bset(b, t)));
|
|
|
|
case COMPACT_ALL:
|
|
|
|
return true;
|
|
|
|
default:
|
|
|
|
BUG();
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-12-13 18:08:37 +00:00
|
|
|
static bool bch2_drop_whiteouts(struct btree *b, enum compact_mode mode)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
|
|
|
struct bset_tree *t;
|
|
|
|
bool ret = false;
|
|
|
|
|
|
|
|
for_each_bset(b, t) {
|
|
|
|
struct bset *i = bset(b, t);
|
|
|
|
struct bkey_packed *k, *n, *out, *start, *end;
|
2019-12-13 18:08:37 +00:00
|
|
|
struct btree_node_entry *src = NULL, *dst = NULL;
|
|
|
|
|
|
|
|
if (t != b->set && !bset_written(b, i)) {
|
|
|
|
src = container_of(i, struct btree_node_entry, keys);
|
|
|
|
dst = max(write_block(b),
|
|
|
|
(void *) btree_bkey_last(b, t - 1));
|
|
|
|
}
|
|
|
|
|
|
|
|
if (src != dst)
|
|
|
|
ret = true;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2019-12-13 18:08:37 +00:00
|
|
|
if (!should_compact_bset(b, t, ret, mode)) {
|
|
|
|
if (src != dst) {
|
|
|
|
memmove(dst, src, sizeof(*src) +
|
|
|
|
le16_to_cpu(src->keys.u64s) *
|
|
|
|
sizeof(u64));
|
|
|
|
i = &dst->keys;
|
|
|
|
set_btree_bset(b, t, i);
|
|
|
|
}
|
2017-03-17 06:18:50 +00:00
|
|
|
continue;
|
2019-12-13 18:08:37 +00:00
|
|
|
}
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
start = btree_bkey_first(b, t);
|
|
|
|
end = btree_bkey_last(b, t);
|
|
|
|
|
2019-12-13 18:08:37 +00:00
|
|
|
if (src != dst) {
|
|
|
|
memmove(dst, src, sizeof(*src));
|
|
|
|
i = &dst->keys;
|
2017-03-17 06:18:50 +00:00
|
|
|
set_btree_bset(b, t, i);
|
|
|
|
}
|
|
|
|
|
|
|
|
out = i->start;
|
|
|
|
|
|
|
|
for (k = start; k != end; k = n) {
|
2023-03-05 04:05:55 +00:00
|
|
|
n = bkey_p_next(k);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2021-02-20 04:41:40 +00:00
|
|
|
if (!bkey_deleted(k)) {
|
2023-11-02 23:33:48 +00:00
|
|
|
bkey_p_copy(out, k);
|
2023-03-05 04:05:55 +00:00
|
|
|
out = bkey_p_next(out);
|
2019-12-13 18:08:37 +00:00
|
|
|
} else {
|
|
|
|
BUG_ON(k->needs_whiteout);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
i->u64s = cpu_to_le16((u64 *) out - i->_data);
|
2019-12-13 18:08:37 +00:00
|
|
|
set_btree_bset_end(b, t);
|
2017-03-17 06:18:50 +00:00
|
|
|
bch2_bset_set_no_aux_tree(b, t);
|
|
|
|
ret = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
bch2_verify_btree_nr_keys(b);
|
|
|
|
|
2019-12-13 18:08:37 +00:00
|
|
|
bch2_btree_build_aux_trees(b);
|
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2019-12-13 18:08:37 +00:00
|
|
|
bool bch2_compact_whiteouts(struct bch_fs *c, struct btree *b,
|
|
|
|
enum compact_mode mode)
|
|
|
|
{
|
2021-02-20 05:00:23 +00:00
|
|
|
return bch2_drop_whiteouts(b, mode);
|
2019-12-13 18:08:37 +00:00
|
|
|
}
|
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
static void btree_node_sort(struct bch_fs *c, struct btree *b,
|
|
|
|
unsigned start_idx,
|
|
|
|
unsigned end_idx,
|
|
|
|
bool filter_whiteouts)
|
|
|
|
{
|
|
|
|
struct btree_node *out;
|
2023-09-10 00:10:11 +00:00
|
|
|
struct sort_iter_stack sort_iter;
|
2017-03-17 06:18:50 +00:00
|
|
|
struct bset_tree *t;
|
|
|
|
struct bset *start_bset = bset(b, &b->set[start_idx]);
|
|
|
|
bool used_mempool = false;
|
|
|
|
u64 start_time, seq = 0;
|
2020-07-25 19:07:37 +00:00
|
|
|
unsigned i, u64s = 0, bytes, shift = end_idx - start_idx - 1;
|
2017-03-17 06:18:50 +00:00
|
|
|
bool sorting_entire_node = start_idx == 0 &&
|
|
|
|
end_idx == b->nsets;
|
|
|
|
|
2023-09-10 00:10:11 +00:00
|
|
|
sort_iter_stack_init(&sort_iter, b);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
for (t = b->set + start_idx;
|
|
|
|
t < b->set + end_idx;
|
|
|
|
t++) {
|
|
|
|
u64s += le16_to_cpu(bset(b, t)->u64s);
|
2023-09-10 00:10:11 +00:00
|
|
|
sort_iter_add(&sort_iter.iter,
|
2017-03-17 06:18:50 +00:00
|
|
|
btree_bkey_first(b, t),
|
|
|
|
btree_bkey_last(b, t));
|
|
|
|
}
|
|
|
|
|
2020-07-25 19:07:37 +00:00
|
|
|
bytes = sorting_entire_node
|
2024-01-16 18:29:59 +00:00
|
|
|
? btree_buf_bytes(b)
|
2020-07-25 19:07:37 +00:00
|
|
|
: __vstruct_bytes(struct btree_node, u64s);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2020-07-25 19:07:37 +00:00
|
|
|
out = btree_bounce_alloc(c, bytes, &used_mempool);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
start_time = local_clock();
|
|
|
|
|
2023-09-10 00:10:11 +00:00
|
|
|
u64s = bch2_sort_keys(out->keys.start, &sort_iter.iter, filter_whiteouts);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
out->keys.u64s = cpu_to_le16(u64s);
|
|
|
|
|
2020-07-25 19:07:37 +00:00
|
|
|
BUG_ON(vstruct_end(&out->keys) > (void *) out + bytes);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
if (sorting_entire_node)
|
2019-03-21 20:28:57 +00:00
|
|
|
bch2_time_stats_update(&c->times[BCH_TIME_btree_node_sort],
|
2017-03-17 06:18:50 +00:00
|
|
|
start_time);
|
|
|
|
|
|
|
|
/* Make sure we preserve bset journal_seq: */
|
|
|
|
for (t = b->set + start_idx; t < b->set + end_idx; t++)
|
|
|
|
seq = max(seq, le64_to_cpu(bset(b, t)->journal_seq));
|
|
|
|
start_bset->journal_seq = cpu_to_le64(seq);
|
|
|
|
|
|
|
|
if (sorting_entire_node) {
|
2023-09-12 22:41:22 +00:00
|
|
|
u64s = le16_to_cpu(out->keys.u64s);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2024-01-16 18:29:59 +00:00
|
|
|
BUG_ON(bytes != btree_buf_bytes(b));
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Our temporary buffer is the same size as the btree node's
|
|
|
|
* buffer, we can just swap buffers instead of doing a big
|
|
|
|
* memcpy()
|
|
|
|
*/
|
|
|
|
*out = *b->data;
|
|
|
|
out->keys.u64s = cpu_to_le16(u64s);
|
|
|
|
swap(out, b->data);
|
|
|
|
set_btree_bset(b, b->set, &b->data->keys);
|
|
|
|
} else {
|
|
|
|
start_bset->u64s = out->keys.u64s;
|
|
|
|
memcpy_u64s(start_bset->start,
|
|
|
|
out->keys.start,
|
|
|
|
le16_to_cpu(out->keys.u64s));
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = start_idx + 1; i < end_idx; i++)
|
|
|
|
b->nr.bset_u64s[start_idx] +=
|
|
|
|
b->nr.bset_u64s[i];
|
|
|
|
|
|
|
|
b->nsets -= shift;
|
|
|
|
|
|
|
|
for (i = start_idx + 1; i < b->nsets; i++) {
|
|
|
|
b->nr.bset_u64s[i] = b->nr.bset_u64s[i + shift];
|
|
|
|
b->set[i] = b->set[i + shift];
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = b->nsets; i < MAX_BSETS; i++)
|
|
|
|
b->nr.bset_u64s[i] = 0;
|
|
|
|
|
|
|
|
set_btree_bset_end(b, &b->set[start_idx]);
|
|
|
|
bch2_bset_set_no_aux_tree(b, &b->set[start_idx]);
|
|
|
|
|
2020-07-25 19:07:37 +00:00
|
|
|
btree_bounce_free(c, bytes, used_mempool, out);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
bch2_verify_btree_nr_keys(b);
|
|
|
|
}
|
|
|
|
|
|
|
|
void bch2_btree_sort_into(struct bch_fs *c,
|
|
|
|
struct btree *dst,
|
|
|
|
struct btree *src)
|
|
|
|
{
|
|
|
|
struct btree_nr_keys nr;
|
|
|
|
struct btree_node_iter src_iter;
|
|
|
|
u64 start_time = local_clock();
|
|
|
|
|
|
|
|
BUG_ON(dst->nsets != 1);
|
|
|
|
|
|
|
|
bch2_bset_set_no_aux_tree(dst, dst->set);
|
|
|
|
|
2016-07-22 03:05:06 +00:00
|
|
|
bch2_btree_node_iter_init_from_start(&src_iter, src);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2021-12-19 23:59:22 +00:00
|
|
|
nr = bch2_sort_repack(btree_bset_first(dst),
|
|
|
|
src, &src_iter,
|
|
|
|
&dst->format,
|
|
|
|
true);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2019-03-21 20:28:57 +00:00
|
|
|
bch2_time_stats_update(&c->times[BCH_TIME_btree_node_sort],
|
|
|
|
start_time);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
set_btree_bset_end(dst, dst->set);
|
|
|
|
|
|
|
|
dst->nr.live_u64s += nr.live_u64s;
|
|
|
|
dst->nr.bset_u64s[0] += nr.bset_u64s[0];
|
|
|
|
dst->nr.packed_keys += nr.packed_keys;
|
|
|
|
dst->nr.unpacked_keys += nr.unpacked_keys;
|
|
|
|
|
|
|
|
bch2_verify_btree_nr_keys(dst);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We're about to add another bset to the btree node, so if there's currently
|
|
|
|
* too many bsets - sort some of them together:
|
|
|
|
*/
|
2021-04-06 19:33:19 +00:00
|
|
|
static bool btree_node_compact(struct bch_fs *c, struct btree *b)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
|
|
|
unsigned unwritten_idx;
|
|
|
|
bool ret = false;
|
|
|
|
|
|
|
|
for (unwritten_idx = 0;
|
|
|
|
unwritten_idx < b->nsets;
|
|
|
|
unwritten_idx++)
|
2018-08-06 02:23:44 +00:00
|
|
|
if (!bset_written(b, bset(b, &b->set[unwritten_idx])))
|
2017-03-17 06:18:50 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
if (b->nsets - unwritten_idx > 1) {
|
2021-04-06 19:33:19 +00:00
|
|
|
btree_node_sort(c, b, unwritten_idx,
|
2017-03-17 06:18:50 +00:00
|
|
|
b->nsets, false);
|
|
|
|
ret = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (unwritten_idx > 1) {
|
2021-04-06 19:33:19 +00:00
|
|
|
btree_node_sort(c, b, 0, unwritten_idx, false);
|
2017-03-17 06:18:50 +00:00
|
|
|
ret = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
void bch2_btree_build_aux_trees(struct btree *b)
|
|
|
|
{
|
|
|
|
struct bset_tree *t;
|
|
|
|
|
|
|
|
for_each_bset(b, t)
|
|
|
|
bch2_bset_build_aux_tree(b, t,
|
2018-08-06 02:23:44 +00:00
|
|
|
!bset_written(b, bset(b, t)) &&
|
2017-03-17 06:18:50 +00:00
|
|
|
t == bset_tree_last(b));
|
|
|
|
}
|
|
|
|
|
2022-10-28 22:56:31 +00:00
|
|
|
/*
|
|
|
|
* If we have MAX_BSETS (3) bsets, should we sort them all down to just one?
|
|
|
|
*
|
|
|
|
* The first bset is going to be of similar order to the size of the node, the
|
|
|
|
* last bset is bounded by btree_write_set_buffer(), which is set to keep the
|
|
|
|
* memmove on insert from being too expensive: the middle bset should, ideally,
|
|
|
|
* be the geometric mean of the first and the last.
|
|
|
|
*
|
|
|
|
* Returns true if the middle bset is greater than that geometric mean:
|
|
|
|
*/
|
|
|
|
static inline bool should_compact_all(struct bch_fs *c, struct btree *b)
|
|
|
|
{
|
|
|
|
unsigned mid_u64s_bits =
|
|
|
|
(ilog2(btree_max_u64s(c)) + BTREE_WRITE_SET_U64s_BITS) / 2;
|
|
|
|
|
|
|
|
return bset_u64s(&b->set[1]) > 1U << mid_u64s_bits;
|
|
|
|
}
|
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
/*
|
|
|
|
* @bch_btree_init_next - initialize a new (unwritten) bset that can then be
|
|
|
|
* inserted into
|
|
|
|
*
|
|
|
|
* Safe to call if there already is an unwritten bset - will only add a new bset
|
|
|
|
* if @b doesn't already have one.
|
|
|
|
*
|
|
|
|
* Returns true if we sorted (i.e. invalidated iterators
|
|
|
|
*/
|
2021-08-30 18:36:03 +00:00
|
|
|
void bch2_btree_init_next(struct btree_trans *trans, struct btree *b)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
2021-07-11 03:22:06 +00:00
|
|
|
struct bch_fs *c = trans->c;
|
2017-03-17 06:18:50 +00:00
|
|
|
struct btree_node_entry *bne;
|
2021-04-06 19:33:19 +00:00
|
|
|
bool reinit_iter = false;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2023-05-21 03:57:48 +00:00
|
|
|
EBUG_ON(!six_lock_counts(&b->c.lock).n[SIX_LOCK_write]);
|
2021-04-06 19:33:19 +00:00
|
|
|
BUG_ON(bset_written(b, bset(b, &b->set[1])));
|
2022-10-28 22:56:31 +00:00
|
|
|
BUG_ON(btree_node_just_written(b));
|
2021-04-06 19:33:19 +00:00
|
|
|
|
2021-07-11 03:03:15 +00:00
|
|
|
if (b->nsets == MAX_BSETS &&
|
2022-10-28 22:56:31 +00:00
|
|
|
!btree_node_write_in_flight(b) &&
|
|
|
|
should_compact_all(c, b)) {
|
|
|
|
bch2_btree_node_write(c, b, SIX_LOCK_write,
|
|
|
|
BTREE_WRITE_init_next_bset);
|
|
|
|
reinit_iter = true;
|
2021-04-06 19:33:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (b->nsets == MAX_BSETS &&
|
|
|
|
btree_node_compact(c, b))
|
|
|
|
reinit_iter = true;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2021-04-06 19:33:19 +00:00
|
|
|
BUG_ON(b->nsets >= MAX_BSETS);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
bne = want_new_bset(c, b);
|
|
|
|
if (bne)
|
2024-01-16 18:29:59 +00:00
|
|
|
bch2_bset_init_next(b, bne);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
bch2_btree_build_aux_trees(b);
|
|
|
|
|
2021-08-30 18:36:03 +00:00
|
|
|
if (reinit_iter)
|
|
|
|
bch2_trans_node_reinit_iter(trans, b);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
2018-11-09 06:24:07 +00:00
|
|
|
static void btree_err_msg(struct printbuf *out, struct bch_fs *c,
|
2021-02-02 22:08:54 +00:00
|
|
|
struct bch_dev *ca,
|
2018-11-09 06:24:07 +00:00
|
|
|
struct btree *b, struct bset *i,
|
|
|
|
unsigned offset, int write)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
2023-01-03 22:14:07 +00:00
|
|
|
prt_printf(out, bch2_log_msg(c, "%s"),
|
|
|
|
write == READ
|
|
|
|
? "error validating btree node "
|
|
|
|
: "corrupt btree node before write ");
|
2021-02-02 22:08:54 +00:00
|
|
|
if (ca)
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_printf(out, "on %s ", ca->name);
|
|
|
|
prt_printf(out, "at btree ");
|
2023-10-20 02:49:08 +00:00
|
|
|
bch2_btree_pos_to_text(out, c, b);
|
2020-06-03 22:27:07 +00:00
|
|
|
|
2024-01-05 19:17:57 +00:00
|
|
|
prt_printf(out, "\n node offset %u/%u",
|
|
|
|
b->written, btree_ptr_sectors_written(&b->key));
|
2017-03-17 06:18:50 +00:00
|
|
|
if (i)
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_printf(out, " bset u64s %u", le16_to_cpu(i->u64s));
|
2022-09-25 22:22:54 +00:00
|
|
|
prt_str(out, ": ");
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
2023-10-25 00:44:36 +00:00
|
|
|
__printf(9, 10)
|
2023-08-03 23:36:28 +00:00
|
|
|
static int __btree_err(int ret,
|
2023-01-03 22:14:07 +00:00
|
|
|
struct bch_fs *c,
|
|
|
|
struct bch_dev *ca,
|
|
|
|
struct btree *b,
|
|
|
|
struct bset *i,
|
|
|
|
int write,
|
|
|
|
bool have_retry,
|
2023-10-25 00:44:36 +00:00
|
|
|
enum bch_sb_error_id err_type,
|
2023-01-03 22:14:07 +00:00
|
|
|
const char *fmt, ...)
|
|
|
|
{
|
|
|
|
struct printbuf out = PRINTBUF;
|
|
|
|
va_list args;
|
|
|
|
|
|
|
|
btree_err_msg(&out, c, ca, b, i, b->written, write);
|
|
|
|
|
|
|
|
va_start(args, fmt);
|
|
|
|
prt_vprintf(&out, fmt, args);
|
|
|
|
va_end(args);
|
|
|
|
|
|
|
|
if (write == WRITE) {
|
2023-02-01 20:45:45 +00:00
|
|
|
bch2_print_string_as_lines(KERN_ERR, out.buf);
|
2023-01-03 22:14:07 +00:00
|
|
|
ret = c->opts.errors == BCH_ON_ERROR_continue
|
|
|
|
? 0
|
|
|
|
: -BCH_ERR_fsck_errors_not_fixed;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2023-08-03 23:36:28 +00:00
|
|
|
if (!have_retry && ret == -BCH_ERR_btree_node_read_err_want_retry)
|
|
|
|
ret = -BCH_ERR_btree_node_read_err_fixable;
|
|
|
|
if (!have_retry && ret == -BCH_ERR_btree_node_read_err_must_retry)
|
|
|
|
ret = -BCH_ERR_btree_node_read_err_bad_node;
|
2023-02-01 20:45:45 +00:00
|
|
|
|
2023-10-25 00:44:36 +00:00
|
|
|
if (ret != -BCH_ERR_btree_node_read_err_fixable)
|
|
|
|
bch2_sb_error_count(c, err_type);
|
|
|
|
|
2023-08-03 23:36:28 +00:00
|
|
|
switch (ret) {
|
|
|
|
case -BCH_ERR_btree_node_read_err_fixable:
|
2023-10-25 00:44:36 +00:00
|
|
|
ret = bch2_fsck_err(c, FSCK_CAN_FIX, err_type, "%s", out.buf);
|
|
|
|
if (ret != -BCH_ERR_fsck_fix &&
|
|
|
|
ret != -BCH_ERR_fsck_ignore)
|
|
|
|
goto fsck_err;
|
2023-02-01 20:45:45 +00:00
|
|
|
ret = -BCH_ERR_fsck_fix;
|
2023-01-03 22:14:07 +00:00
|
|
|
break;
|
2023-08-03 23:36:28 +00:00
|
|
|
case -BCH_ERR_btree_node_read_err_want_retry:
|
|
|
|
case -BCH_ERR_btree_node_read_err_must_retry:
|
2023-02-01 20:45:45 +00:00
|
|
|
bch2_print_string_as_lines(KERN_ERR, out.buf);
|
2023-01-03 22:14:07 +00:00
|
|
|
break;
|
2023-08-03 23:36:28 +00:00
|
|
|
case -BCH_ERR_btree_node_read_err_bad_node:
|
2023-02-01 20:45:45 +00:00
|
|
|
bch2_print_string_as_lines(KERN_ERR, out.buf);
|
2024-02-06 22:24:18 +00:00
|
|
|
ret = bch2_topology_error(c);
|
2023-02-01 20:45:45 +00:00
|
|
|
break;
|
2023-08-03 23:36:28 +00:00
|
|
|
case -BCH_ERR_btree_node_read_err_incompatible:
|
2023-02-01 20:45:45 +00:00
|
|
|
bch2_print_string_as_lines(KERN_ERR, out.buf);
|
2023-01-03 22:14:07 +00:00
|
|
|
ret = -BCH_ERR_fsck_errors_not_fixed;
|
|
|
|
break;
|
2023-02-01 20:45:45 +00:00
|
|
|
default:
|
|
|
|
BUG();
|
2023-01-03 22:14:07 +00:00
|
|
|
}
|
|
|
|
out:
|
|
|
|
fsck_err:
|
|
|
|
printbuf_exit(&out);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2023-10-25 00:44:36 +00:00
|
|
|
#define btree_err(type, c, ca, b, i, _err_type, msg, ...) \
|
2017-03-17 06:18:50 +00:00
|
|
|
({ \
|
2023-10-25 00:44:36 +00:00
|
|
|
int _ret = __btree_err(type, c, ca, b, i, write, have_retry, \
|
|
|
|
BCH_FSCK_ERR_##_err_type, \
|
|
|
|
msg, ##__VA_ARGS__); \
|
2022-09-25 22:22:54 +00:00
|
|
|
\
|
2023-08-04 00:32:46 +00:00
|
|
|
if (_ret != -BCH_ERR_fsck_fix) { \
|
|
|
|
ret = _ret; \
|
2023-01-03 22:14:07 +00:00
|
|
|
goto fsck_err; \
|
2023-08-04 00:32:46 +00:00
|
|
|
} \
|
|
|
|
\
|
2023-01-03 22:32:16 +00:00
|
|
|
*saw_error = true; \
|
2017-03-17 06:18:50 +00:00
|
|
|
})
|
|
|
|
|
|
|
|
#define btree_err_on(cond, ...) ((cond) ? btree_err(__VA_ARGS__) : false)
|
|
|
|
|
2021-04-24 20:32:35 +00:00
|
|
|
/*
|
|
|
|
* When btree topology repair changes the start or end of a node, that might
|
|
|
|
* mean we have to drop keys that are no longer inside the node:
|
|
|
|
*/
|
|
|
|
__cold
|
|
|
|
void bch2_btree_node_drop_keys_outside_node(struct btree *b)
|
|
|
|
{
|
|
|
|
struct bset_tree *t;
|
|
|
|
|
|
|
|
for_each_bset(b, t) {
|
|
|
|
struct bset *i = bset(b, t);
|
|
|
|
struct bkey_packed *k;
|
|
|
|
|
2023-03-05 04:05:55 +00:00
|
|
|
for (k = i->start; k != vstruct_last(i); k = bkey_p_next(k))
|
2021-04-24 20:32:35 +00:00
|
|
|
if (bkey_cmp_left_packed(b, k, &b->data->min_key) >= 0)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (k != i->start) {
|
|
|
|
unsigned shift = (u64 *) k - (u64 *) i->start;
|
|
|
|
|
|
|
|
memmove_u64s_down(i->start, k,
|
|
|
|
(u64 *) vstruct_end(i) - (u64 *) k);
|
|
|
|
i->u64s = cpu_to_le16(le16_to_cpu(i->u64s) - shift);
|
|
|
|
set_btree_bset_end(b, t);
|
|
|
|
}
|
|
|
|
|
2023-03-05 04:05:55 +00:00
|
|
|
for (k = i->start; k != vstruct_last(i); k = bkey_p_next(k))
|
2021-04-24 20:32:35 +00:00
|
|
|
if (bkey_cmp_left_packed(b, k, &b->data->max_key) > 0)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (k != vstruct_last(i)) {
|
|
|
|
i->u64s = cpu_to_le16((u64 *) k - (u64 *) i->start);
|
|
|
|
set_btree_bset_end(b, t);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2022-08-17 18:20:48 +00:00
|
|
|
/*
|
|
|
|
* Always rebuild search trees: eytzinger search tree nodes directly
|
|
|
|
* depend on the values of min/max key:
|
|
|
|
*/
|
|
|
|
bch2_bset_set_no_aux_tree(b, b->set);
|
2021-04-24 20:32:35 +00:00
|
|
|
bch2_btree_build_aux_trees(b);
|
|
|
|
|
2023-09-12 22:41:22 +00:00
|
|
|
struct bkey_s_c k;
|
|
|
|
struct bkey unpacked;
|
|
|
|
struct btree_node_iter iter;
|
2021-04-24 20:32:35 +00:00
|
|
|
for_each_btree_node_key_unpack(b, k, &iter, &unpacked) {
|
2022-11-24 08:12:22 +00:00
|
|
|
BUG_ON(bpos_lt(k.k->p, b->data->min_key));
|
|
|
|
BUG_ON(bpos_gt(k.k->p, b->data->max_key));
|
2021-04-24 20:32:35 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-02-02 22:08:54 +00:00
|
|
|
static int validate_bset(struct bch_fs *c, struct bch_dev *ca,
|
|
|
|
struct btree *b, struct bset *i,
|
2021-07-16 16:57:27 +00:00
|
|
|
unsigned offset, unsigned sectors,
|
2023-01-03 22:32:16 +00:00
|
|
|
int write, bool have_retry, bool *saw_error)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
2020-01-07 18:29:32 +00:00
|
|
|
unsigned version = le16_to_cpu(i->version);
|
2022-02-25 18:18:19 +00:00
|
|
|
struct printbuf buf1 = PRINTBUF;
|
|
|
|
struct printbuf buf2 = PRINTBUF;
|
2017-03-17 06:18:50 +00:00
|
|
|
int ret = 0;
|
|
|
|
|
2023-06-29 00:27:07 +00:00
|
|
|
btree_err_on(!bch2_version_compatible(version),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_incompatible,
|
|
|
|
c, ca, b, i,
|
|
|
|
btree_node_unsupported_version,
|
2023-06-28 02:09:35 +00:00
|
|
|
"unsupported bset version %u.%u",
|
|
|
|
BCH_VERSION_MAJOR(version),
|
|
|
|
BCH_VERSION_MINOR(version));
|
2020-01-07 18:29:32 +00:00
|
|
|
|
2021-03-21 20:03:23 +00:00
|
|
|
if (btree_err_on(version < c->sb.version_min,
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, NULL, b, i,
|
|
|
|
btree_node_bset_older_than_sb_min,
|
2021-03-21 20:03:23 +00:00
|
|
|
"bset version %u older than superblock version_min %u",
|
|
|
|
version, c->sb.version_min)) {
|
|
|
|
mutex_lock(&c->sb_lock);
|
|
|
|
c->disk_sb.sb->version_min = cpu_to_le16(version);
|
|
|
|
bch2_write_super(c);
|
|
|
|
mutex_unlock(&c->sb_lock);
|
|
|
|
}
|
|
|
|
|
2023-06-28 02:09:35 +00:00
|
|
|
if (btree_err_on(BCH_VERSION_MAJOR(version) >
|
|
|
|
BCH_VERSION_MAJOR(c->sb.version),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, NULL, b, i,
|
|
|
|
btree_node_bset_newer_than_sb,
|
2021-03-21 20:03:23 +00:00
|
|
|
"bset version %u newer than superblock version %u",
|
|
|
|
version, c->sb.version)) {
|
|
|
|
mutex_lock(&c->sb_lock);
|
|
|
|
c->disk_sb.sb->version = cpu_to_le16(version);
|
|
|
|
bch2_write_super(c);
|
|
|
|
mutex_unlock(&c->sb_lock);
|
|
|
|
}
|
|
|
|
|
2021-03-29 04:19:05 +00:00
|
|
|
btree_err_on(BSET_SEPARATE_WHITEOUTS(i),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_incompatible,
|
|
|
|
c, ca, b, i,
|
|
|
|
btree_node_unsupported_version,
|
2021-03-29 04:19:05 +00:00
|
|
|
"BSET_SEPARATE_WHITEOUTS no longer supported");
|
|
|
|
|
2021-12-14 19:24:41 +00:00
|
|
|
if (btree_err_on(offset + sectors > btree_sectors(c),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, ca, b, i,
|
|
|
|
bset_past_end_of_btree_node,
|
2020-01-07 18:29:32 +00:00
|
|
|
"bset past end of btree node")) {
|
|
|
|
i->u64s = 0;
|
2022-02-25 18:18:19 +00:00
|
|
|
ret = 0;
|
|
|
|
goto out;
|
2020-01-07 18:29:32 +00:00
|
|
|
}
|
|
|
|
|
2021-07-16 16:57:27 +00:00
|
|
|
btree_err_on(offset && !i->u64s,
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, ca, b, i,
|
|
|
|
bset_empty,
|
2020-01-07 18:29:32 +00:00
|
|
|
"empty bset");
|
|
|
|
|
2023-10-25 00:44:36 +00:00
|
|
|
btree_err_on(BSET_OFFSET(i) && BSET_OFFSET(i) != offset,
|
|
|
|
-BCH_ERR_btree_node_read_err_want_retry,
|
|
|
|
c, ca, b, i,
|
|
|
|
bset_wrong_sector_offset,
|
2021-07-16 16:57:27 +00:00
|
|
|
"bset at wrong sector offset");
|
|
|
|
|
|
|
|
if (!offset) {
|
2020-01-07 18:29:32 +00:00
|
|
|
struct btree_node *bn =
|
|
|
|
container_of(i, struct btree_node, keys);
|
2017-03-17 06:18:50 +00:00
|
|
|
/* These indicate that we read the wrong btree node: */
|
2020-05-12 22:34:16 +00:00
|
|
|
|
|
|
|
if (b->key.k.type == KEY_TYPE_btree_ptr_v2) {
|
|
|
|
struct bch_btree_ptr_v2 *bp =
|
|
|
|
&bkey_i_to_btree_ptr_v2(&b->key)->v;
|
|
|
|
|
|
|
|
/* XXX endianness */
|
|
|
|
btree_err_on(bp->seq != bn->keys.seq,
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_must_retry,
|
|
|
|
c, ca, b, NULL,
|
|
|
|
bset_bad_seq,
|
2020-05-12 22:34:16 +00:00
|
|
|
"incorrect sequence number (wrong btree node)");
|
|
|
|
}
|
|
|
|
|
2020-01-07 18:29:32 +00:00
|
|
|
btree_err_on(BTREE_NODE_ID(bn) != b->c.btree_id,
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_must_retry,
|
|
|
|
c, ca, b, i,
|
|
|
|
btree_node_bad_btree,
|
2017-03-17 06:18:50 +00:00
|
|
|
"incorrect btree id");
|
|
|
|
|
2020-01-07 18:29:32 +00:00
|
|
|
btree_err_on(BTREE_NODE_LEVEL(bn) != b->c.level,
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_must_retry,
|
|
|
|
c, ca, b, i,
|
|
|
|
btree_node_bad_level,
|
2017-03-17 06:18:50 +00:00
|
|
|
"incorrect level");
|
|
|
|
|
2020-01-07 18:29:32 +00:00
|
|
|
if (!write)
|
|
|
|
compat_btree_node(b->c.level, b->c.btree_id, version,
|
|
|
|
BSET_BIG_ENDIAN(i), write, bn);
|
|
|
|
|
2020-02-07 18:38:02 +00:00
|
|
|
if (b->key.k.type == KEY_TYPE_btree_ptr_v2) {
|
|
|
|
struct bch_btree_ptr_v2 *bp =
|
|
|
|
&bkey_i_to_btree_ptr_v2(&b->key)->v;
|
|
|
|
|
2021-01-27 01:13:54 +00:00
|
|
|
if (BTREE_PTR_RANGE_UPDATED(bp)) {
|
|
|
|
b->data->min_key = bp->min_key;
|
|
|
|
b->data->max_key = b->key.k.p;
|
|
|
|
}
|
|
|
|
|
2022-11-24 08:12:22 +00:00
|
|
|
btree_err_on(!bpos_eq(b->data->min_key, bp->min_key),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_must_retry,
|
|
|
|
c, ca, b, NULL,
|
|
|
|
btree_node_bad_min_key,
|
2021-03-04 20:20:22 +00:00
|
|
|
"incorrect min_key: got %s should be %s",
|
2022-02-25 18:18:19 +00:00
|
|
|
(printbuf_reset(&buf1),
|
|
|
|
bch2_bpos_to_text(&buf1, bn->min_key), buf1.buf),
|
|
|
|
(printbuf_reset(&buf2),
|
|
|
|
bch2_bpos_to_text(&buf2, bp->min_key), buf2.buf));
|
2020-02-07 18:38:02 +00:00
|
|
|
}
|
|
|
|
|
2022-11-24 08:12:22 +00:00
|
|
|
btree_err_on(!bpos_eq(bn->max_key, b->key.k.p),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_must_retry,
|
|
|
|
c, ca, b, i,
|
|
|
|
btree_node_bad_max_key,
|
2021-03-04 20:20:22 +00:00
|
|
|
"incorrect max key %s",
|
2022-02-25 18:18:19 +00:00
|
|
|
(printbuf_reset(&buf1),
|
|
|
|
bch2_bpos_to_text(&buf1, bn->max_key), buf1.buf));
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2020-01-07 18:29:32 +00:00
|
|
|
if (write)
|
|
|
|
compat_btree_node(b->c.level, b->c.btree_id, version,
|
|
|
|
BSET_BIG_ENDIAN(i), write, bn);
|
|
|
|
|
2023-08-06 14:04:37 +00:00
|
|
|
btree_err_on(bch2_bkey_format_invalid(c, &bn->format, write, &buf1),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_bad_node,
|
|
|
|
c, ca, b, i,
|
|
|
|
btree_node_bad_format,
|
2023-08-03 18:42:37 +00:00
|
|
|
"invalid bkey format: %s\n %s", buf1.buf,
|
|
|
|
(printbuf_reset(&buf2),
|
|
|
|
bch2_bkey_format_to_text(&buf2, &bn->format), buf2.buf));
|
|
|
|
printbuf_reset(&buf1);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2020-01-07 18:29:32 +00:00
|
|
|
compat_bformat(b->c.level, b->c.btree_id, version,
|
|
|
|
BSET_BIG_ENDIAN(i), write,
|
|
|
|
&bn->format);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
2022-02-25 18:18:19 +00:00
|
|
|
out:
|
2020-01-07 18:29:32 +00:00
|
|
|
fsck_err:
|
2022-02-25 18:18:19 +00:00
|
|
|
printbuf_exit(&buf2);
|
|
|
|
printbuf_exit(&buf1);
|
2020-01-07 18:29:32 +00:00
|
|
|
return ret;
|
|
|
|
}
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2022-04-03 21:50:01 +00:00
|
|
|
static int bset_key_invalid(struct bch_fs *c, struct btree *b,
|
|
|
|
struct bkey_s_c k,
|
2022-04-04 01:50:25 +00:00
|
|
|
bool updated_range, int rw,
|
2022-04-03 21:50:01 +00:00
|
|
|
struct printbuf *err)
|
|
|
|
{
|
2022-04-10 04:48:36 +00:00
|
|
|
return __bch2_bkey_invalid(c, k, btree_node_type(b), READ, err) ?:
|
2023-10-25 00:44:36 +00:00
|
|
|
(!updated_range ? bch2_bkey_in_btree_node(c, b, k, err) : 0) ?:
|
2022-04-10 04:48:36 +00:00
|
|
|
(rw == WRITE ? bch2_bkey_val_invalid(c, k, READ, err) : 0);
|
2022-04-03 21:50:01 +00:00
|
|
|
}
|
|
|
|
|
2024-01-05 19:17:57 +00:00
|
|
|
static bool __bkey_valid(struct bch_fs *c, struct btree *b,
|
|
|
|
struct bset *i, struct bkey_packed *k)
|
|
|
|
{
|
|
|
|
if (bkey_p_next(k) > vstruct_last(i))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if (k->format > KEY_FORMAT_CURRENT)
|
|
|
|
return false;
|
|
|
|
|
2024-03-08 19:53:03 +00:00
|
|
|
if (k->u64s < bkeyp_key_u64s(&b->format, k))
|
|
|
|
return false;
|
|
|
|
|
2024-01-05 19:17:57 +00:00
|
|
|
struct printbuf buf = PRINTBUF;
|
|
|
|
struct bkey tmp;
|
|
|
|
struct bkey_s u = __bkey_disassemble(b, k, &tmp);
|
|
|
|
bool ret = __bch2_bkey_invalid(c, u.s_c, btree_node_type(b), READ, &buf);
|
|
|
|
printbuf_exit(&buf);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2020-01-07 18:29:32 +00:00
|
|
|
static int validate_bset_keys(struct bch_fs *c, struct btree *b,
|
2023-01-03 22:32:16 +00:00
|
|
|
struct bset *i, int write,
|
|
|
|
bool have_retry, bool *saw_error)
|
2020-01-07 18:29:32 +00:00
|
|
|
{
|
|
|
|
unsigned version = le16_to_cpu(i->version);
|
|
|
|
struct bkey_packed *k, *prev = NULL;
|
2022-04-03 21:50:01 +00:00
|
|
|
struct printbuf buf = PRINTBUF;
|
2021-04-24 20:32:35 +00:00
|
|
|
bool updated_range = b->key.k.type == KEY_TYPE_btree_ptr_v2 &&
|
|
|
|
BTREE_PTR_RANGE_UPDATED(&bkey_i_to_btree_ptr_v2(&b->key)->v);
|
2020-01-07 18:29:32 +00:00
|
|
|
int ret = 0;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
for (k = i->start;
|
|
|
|
k != vstruct_last(i);) {
|
2020-02-07 01:15:15 +00:00
|
|
|
struct bkey_s u;
|
2017-03-17 06:18:50 +00:00
|
|
|
struct bkey tmp;
|
2024-01-05 19:17:57 +00:00
|
|
|
unsigned next_good_key;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2023-03-05 04:05:55 +00:00
|
|
|
if (btree_err_on(bkey_p_next(k) > vstruct_last(i),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, NULL, b, i,
|
|
|
|
btree_node_bkey_past_bset_end,
|
2017-03-17 06:18:50 +00:00
|
|
|
"key extends past end of bset")) {
|
|
|
|
i->u64s = cpu_to_le16((u64 *) k - i->_data);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (btree_err_on(k->format > KEY_FORMAT_CURRENT,
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, NULL, b, i,
|
|
|
|
btree_node_bkey_bad_format,
|
2024-01-05 19:17:57 +00:00
|
|
|
"invalid bkey format %u", k->format))
|
|
|
|
goto drop_this_key;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2024-03-08 19:53:03 +00:00
|
|
|
if (btree_err_on(k->u64s < bkeyp_key_u64s(&b->format, k),
|
|
|
|
-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, NULL, b, i,
|
|
|
|
btree_node_bkey_bad_u64s,
|
|
|
|
"k->u64s too small (%u < %u)", k->u64s, bkeyp_key_u64s(&b->format, k)))
|
|
|
|
goto drop_this_key;
|
|
|
|
|
2020-01-07 18:29:32 +00:00
|
|
|
if (!write)
|
|
|
|
bch2_bkey_compat(b->c.level, b->c.btree_id, version,
|
|
|
|
BSET_BIG_ENDIAN(i), write,
|
|
|
|
&b->format, k);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2020-02-07 01:15:15 +00:00
|
|
|
u = __bkey_disassemble(b, k, &tmp);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2022-04-03 21:50:01 +00:00
|
|
|
printbuf_reset(&buf);
|
|
|
|
if (bset_key_invalid(c, b, u.s_c, updated_range, write, &buf)) {
|
|
|
|
printbuf_reset(&buf);
|
|
|
|
bset_key_invalid(c, b, u.s_c, updated_range, write, &buf);
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_printf(&buf, "\n ");
|
2022-04-19 03:43:08 +00:00
|
|
|
bch2_bkey_val_to_text(&buf, c, u.s_c);
|
2022-04-03 21:50:01 +00:00
|
|
|
|
2023-10-25 00:44:36 +00:00
|
|
|
btree_err(-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, NULL, b, i,
|
|
|
|
btree_node_bad_bkey,
|
|
|
|
"invalid bkey: %s", buf.buf);
|
2024-01-05 19:17:57 +00:00
|
|
|
goto drop_this_key;
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
2020-01-07 18:29:32 +00:00
|
|
|
if (write)
|
|
|
|
bch2_bkey_compat(b->c.level, b->c.btree_id, version,
|
|
|
|
BSET_BIG_ENDIAN(i), write,
|
|
|
|
&b->format, k);
|
2018-11-01 19:10:01 +00:00
|
|
|
|
2021-03-29 04:19:05 +00:00
|
|
|
if (prev && bkey_iter_cmp(b, prev, k) > 0) {
|
2019-12-30 19:37:25 +00:00
|
|
|
struct bkey up = bkey_unpack_key(b, prev);
|
2020-02-27 01:39:06 +00:00
|
|
|
|
2022-04-03 21:50:01 +00:00
|
|
|
printbuf_reset(&buf);
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_printf(&buf, "keys out of order: ");
|
2022-04-03 21:50:01 +00:00
|
|
|
bch2_bkey_to_text(&buf, &up);
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_printf(&buf, " > ");
|
2022-04-03 21:50:01 +00:00
|
|
|
bch2_bkey_to_text(&buf, u.k);
|
2020-02-27 01:39:06 +00:00
|
|
|
|
2023-10-25 00:44:36 +00:00
|
|
|
if (btree_err(-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, NULL, b, i,
|
|
|
|
btree_node_bkey_out_of_order,
|
2024-01-05 19:17:57 +00:00
|
|
|
"%s", buf.buf))
|
|
|
|
goto drop_this_key;
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
2019-12-30 19:37:25 +00:00
|
|
|
prev = k;
|
2023-03-05 04:05:55 +00:00
|
|
|
k = bkey_p_next(k);
|
2024-01-05 19:17:57 +00:00
|
|
|
continue;
|
|
|
|
drop_this_key:
|
|
|
|
next_good_key = k->u64s;
|
|
|
|
|
|
|
|
if (!next_good_key ||
|
|
|
|
(BSET_BIG_ENDIAN(i) == CPU_BIG_ENDIAN &&
|
|
|
|
version >= bcachefs_metadata_version_snapshot)) {
|
|
|
|
/*
|
|
|
|
* only do scanning if bch2_bkey_compat() has nothing to
|
|
|
|
* do
|
|
|
|
*/
|
|
|
|
|
|
|
|
if (!__bkey_valid(c, b, i, (void *) ((u64 *) k + next_good_key))) {
|
|
|
|
for (next_good_key = 1;
|
|
|
|
next_good_key < (u64 *) vstruct_last(i) - (u64 *) k;
|
|
|
|
next_good_key++)
|
|
|
|
if (__bkey_valid(c, b, i, (void *) ((u64 *) k + next_good_key)))
|
|
|
|
goto got_good_key;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* didn't find a good key, have to truncate the rest of
|
|
|
|
* the bset
|
|
|
|
*/
|
|
|
|
next_good_key = (u64 *) vstruct_last(i) - (u64 *) k;
|
|
|
|
}
|
|
|
|
got_good_key:
|
|
|
|
le16_add_cpu(&i->u64s, -next_good_key);
|
|
|
|
memmove_u64s_down(k, bkey_p_next(k), (u64 *) vstruct_end(i) - (u64 *) k);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
fsck_err:
|
2022-04-03 21:50:01 +00:00
|
|
|
printbuf_exit(&buf);
|
2017-03-17 06:18:50 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2021-02-02 22:08:54 +00:00
|
|
|
int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca,
|
2023-01-03 22:32:16 +00:00
|
|
|
struct btree *b, bool have_retry, bool *saw_error)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
|
|
|
struct btree_node_entry *bne;
|
2019-12-14 21:20:33 +00:00
|
|
|
struct sort_iter *iter;
|
2017-03-17 06:18:50 +00:00
|
|
|
struct btree_node *sorted;
|
|
|
|
struct bkey_packed *k;
|
|
|
|
struct bset *i;
|
2019-04-05 01:53:12 +00:00
|
|
|
bool used_mempool, blacklisted;
|
2021-04-24 20:32:35 +00:00
|
|
|
bool updated_range = b->key.k.type == KEY_TYPE_btree_ptr_v2 &&
|
|
|
|
BTREE_PTR_RANGE_UPDATED(&bkey_i_to_btree_ptr_v2(&b->key)->v);
|
2017-03-17 06:18:50 +00:00
|
|
|
unsigned u64s;
|
2021-07-10 17:44:42 +00:00
|
|
|
unsigned ptr_written = btree_ptr_sectors_written(&b->key);
|
2022-04-03 21:50:01 +00:00
|
|
|
struct printbuf buf = PRINTBUF;
|
2023-01-03 22:14:07 +00:00
|
|
|
int ret = 0, retry_read = 0, write = READ;
|
2024-01-03 21:42:33 +00:00
|
|
|
u64 start_time = local_clock();
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2021-03-14 23:01:14 +00:00
|
|
|
b->version_ondisk = U16_MAX;
|
2022-06-14 05:37:16 +00:00
|
|
|
/* We might get called multiple times on read retry: */
|
|
|
|
b->written = 0;
|
2021-03-14 23:01:14 +00:00
|
|
|
|
2023-05-28 22:02:38 +00:00
|
|
|
iter = mempool_alloc(&c->fill_iter, GFP_NOFS);
|
2023-09-10 00:10:11 +00:00
|
|
|
sort_iter_init(iter, b, (btree_blocks(c) + 1) * 2);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
if (bch2_meta_read_fault("btree"))
|
2023-10-25 00:44:36 +00:00
|
|
|
btree_err(-BCH_ERR_btree_node_read_err_must_retry,
|
|
|
|
c, ca, b, NULL,
|
|
|
|
btree_node_fault_injected,
|
2017-03-17 06:18:50 +00:00
|
|
|
"dynamic fault");
|
|
|
|
|
|
|
|
btree_err_on(le64_to_cpu(b->data->magic) != bset_magic(c),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_must_retry,
|
|
|
|
c, ca, b, NULL,
|
|
|
|
btree_node_bad_magic,
|
2022-02-23 15:32:43 +00:00
|
|
|
"bad magic: want %llx, got %llx",
|
|
|
|
bset_magic(c), le64_to_cpu(b->data->magic));
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2020-02-07 18:38:02 +00:00
|
|
|
if (b->key.k.type == KEY_TYPE_btree_ptr_v2) {
|
|
|
|
struct bch_btree_ptr_v2 *bp =
|
|
|
|
&bkey_i_to_btree_ptr_v2(&b->key)->v;
|
|
|
|
|
2023-12-05 20:22:25 +00:00
|
|
|
bch2_bpos_to_text(&buf, b->data->min_key);
|
|
|
|
prt_str(&buf, "-");
|
|
|
|
bch2_bpos_to_text(&buf, b->data->max_key);
|
|
|
|
|
2020-02-07 18:38:02 +00:00
|
|
|
btree_err_on(b->data->keys.seq != bp->seq,
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_must_retry,
|
|
|
|
c, ca, b, NULL,
|
|
|
|
btree_node_bad_seq,
|
2023-12-05 20:22:25 +00:00
|
|
|
"got wrong btree node (want %llx got %llx)\n"
|
|
|
|
"got btree %s level %llu pos %s",
|
|
|
|
bp->seq, b->data->keys.seq,
|
|
|
|
bch2_btree_id_str(BTREE_NODE_ID(b->data)),
|
|
|
|
BTREE_NODE_LEVEL(b->data),
|
|
|
|
buf.buf);
|
2023-10-25 00:44:36 +00:00
|
|
|
} else {
|
|
|
|
btree_err_on(!b->data->keys.seq,
|
|
|
|
-BCH_ERR_btree_node_read_err_must_retry,
|
|
|
|
c, ca, b, NULL,
|
|
|
|
btree_node_bad_seq,
|
|
|
|
"bad btree header: seq 0");
|
2020-02-07 18:38:02 +00:00
|
|
|
}
|
|
|
|
|
2021-12-14 19:24:41 +00:00
|
|
|
while (b->written < (ptr_written ?: btree_sectors(c))) {
|
2022-08-15 18:20:22 +00:00
|
|
|
unsigned sectors;
|
2017-03-17 06:18:50 +00:00
|
|
|
struct nonce nonce;
|
|
|
|
bool first = !b->written;
|
2023-10-25 20:29:37 +00:00
|
|
|
bool csum_bad;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
if (!b->written) {
|
|
|
|
i = &b->data->keys;
|
|
|
|
|
|
|
|
btree_err_on(!bch2_checksum_type_valid(c, BSET_CSUM_TYPE(i)),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_want_retry,
|
|
|
|
c, ca, b, i,
|
|
|
|
bset_unknown_csum,
|
|
|
|
"unknown checksum type %llu", BSET_CSUM_TYPE(i));
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
nonce = btree_nonce(i, b->written << 9);
|
|
|
|
|
2024-01-05 16:59:03 +00:00
|
|
|
struct bch_csum csum = csum_vstruct(c, BSET_CSUM_TYPE(i), nonce, b->data);
|
|
|
|
csum_bad = bch2_crc_cmp(b->data->csum, csum);
|
2023-10-25 20:29:37 +00:00
|
|
|
if (csum_bad)
|
|
|
|
bch2_io_error(ca, BCH_MEMBER_ERROR_checksum);
|
|
|
|
|
|
|
|
btree_err_on(csum_bad,
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_want_retry,
|
|
|
|
c, ca, b, i,
|
|
|
|
bset_bad_csum,
|
2024-01-05 16:59:03 +00:00
|
|
|
"%s",
|
|
|
|
(printbuf_reset(&buf),
|
|
|
|
bch2_csum_err_msg(&buf, BSET_CSUM_TYPE(i), b->data->csum, csum),
|
|
|
|
buf.buf));
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2022-02-19 05:42:12 +00:00
|
|
|
ret = bset_encrypt(c, i, b->written << 9);
|
|
|
|
if (bch2_fs_fatal_err_on(ret, c,
|
2024-03-18 01:51:19 +00:00
|
|
|
"decrypting btree node: %s", bch2_err_str(ret)))
|
2022-02-19 05:42:12 +00:00
|
|
|
goto fsck_err;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2021-12-11 22:13:09 +00:00
|
|
|
btree_err_on(btree_node_type_is_extents(btree_node_type(b)) &&
|
2021-02-20 05:00:23 +00:00
|
|
|
!BTREE_NODE_NEW_EXTENT_OVERWRITE(b->data),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_incompatible,
|
|
|
|
c, NULL, b, NULL,
|
|
|
|
btree_node_unsupported_version,
|
2021-02-20 05:00:23 +00:00
|
|
|
"btree node does not have NEW_EXTENT_OVERWRITE set");
|
2019-11-26 22:26:04 +00:00
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
sectors = vstruct_sectors(b->data, c->block_bits);
|
|
|
|
} else {
|
|
|
|
bne = write_block(b);
|
|
|
|
i = &bne->keys;
|
|
|
|
|
|
|
|
if (i->seq != b->data->keys.seq)
|
|
|
|
break;
|
|
|
|
|
|
|
|
btree_err_on(!bch2_checksum_type_valid(c, BSET_CSUM_TYPE(i)),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_want_retry,
|
|
|
|
c, ca, b, i,
|
|
|
|
bset_unknown_csum,
|
|
|
|
"unknown checksum type %llu", BSET_CSUM_TYPE(i));
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
nonce = btree_nonce(i, b->written << 9);
|
2024-01-05 16:59:03 +00:00
|
|
|
struct bch_csum csum = csum_vstruct(c, BSET_CSUM_TYPE(i), nonce, bne);
|
|
|
|
csum_bad = bch2_crc_cmp(bne->csum, csum);
|
2023-10-25 20:29:37 +00:00
|
|
|
if (csum_bad)
|
|
|
|
bch2_io_error(ca, BCH_MEMBER_ERROR_checksum);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2023-10-25 20:29:37 +00:00
|
|
|
btree_err_on(csum_bad,
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_want_retry,
|
|
|
|
c, ca, b, i,
|
|
|
|
bset_bad_csum,
|
2024-01-05 16:59:03 +00:00
|
|
|
"%s",
|
|
|
|
(printbuf_reset(&buf),
|
|
|
|
bch2_csum_err_msg(&buf, BSET_CSUM_TYPE(i), bne->csum, csum),
|
|
|
|
buf.buf));
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2022-02-19 05:42:12 +00:00
|
|
|
ret = bset_encrypt(c, i, b->written << 9);
|
|
|
|
if (bch2_fs_fatal_err_on(ret, c,
|
2024-03-18 01:51:19 +00:00
|
|
|
"decrypting btree node: %s", bch2_err_str(ret)))
|
2022-02-19 05:42:12 +00:00
|
|
|
goto fsck_err;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
sectors = vstruct_sectors(bne, c->block_bits);
|
|
|
|
}
|
|
|
|
|
2021-03-14 23:01:14 +00:00
|
|
|
b->version_ondisk = min(b->version_ondisk,
|
|
|
|
le16_to_cpu(i->version));
|
|
|
|
|
2021-07-16 16:57:27 +00:00
|
|
|
ret = validate_bset(c, ca, b, i, b->written, sectors,
|
2023-01-03 22:32:16 +00:00
|
|
|
READ, have_retry, saw_error);
|
2017-03-17 06:18:50 +00:00
|
|
|
if (ret)
|
|
|
|
goto fsck_err;
|
|
|
|
|
2020-01-07 18:29:32 +00:00
|
|
|
if (!b->written)
|
|
|
|
btree_node_set_format(b, b->data->format);
|
|
|
|
|
2023-01-03 22:32:16 +00:00
|
|
|
ret = validate_bset_keys(c, b, i, READ, have_retry, saw_error);
|
2020-01-07 18:29:32 +00:00
|
|
|
if (ret)
|
|
|
|
goto fsck_err;
|
|
|
|
|
|
|
|
SET_BSET_BIG_ENDIAN(i, CPU_BIG_ENDIAN);
|
|
|
|
|
2019-04-05 01:53:12 +00:00
|
|
|
blacklisted = bch2_journal_seq_is_blacklisted(c,
|
|
|
|
le64_to_cpu(i->journal_seq),
|
|
|
|
true);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2019-04-05 01:53:12 +00:00
|
|
|
btree_err_on(blacklisted && first,
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, ca, b, i,
|
|
|
|
bset_blacklisted_journal_seq,
|
2022-01-05 00:05:08 +00:00
|
|
|
"first btree node bset has blacklisted journal seq (%llu)",
|
|
|
|
le64_to_cpu(i->journal_seq));
|
2021-07-10 17:44:42 +00:00
|
|
|
|
|
|
|
btree_err_on(blacklisted && ptr_written,
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, ca, b, i,
|
|
|
|
first_bset_blacklisted_journal_seq,
|
2022-01-05 00:05:08 +00:00
|
|
|
"found blacklisted bset (journal seq %llu) in btree node at offset %u-%u/%u",
|
|
|
|
le64_to_cpu(i->journal_seq),
|
|
|
|
b->written, b->written + sectors, ptr_written);
|
|
|
|
|
|
|
|
b->written += sectors;
|
|
|
|
|
2019-04-05 01:53:12 +00:00
|
|
|
if (blacklisted && !first)
|
|
|
|
continue;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2019-12-14 21:20:33 +00:00
|
|
|
sort_iter_add(iter,
|
2022-08-15 18:20:22 +00:00
|
|
|
vstruct_idx(i, 0),
|
2019-12-14 21:20:33 +00:00
|
|
|
vstruct_last(i));
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
2021-07-10 17:44:42 +00:00
|
|
|
if (ptr_written) {
|
|
|
|
btree_err_on(b->written < ptr_written,
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_want_retry,
|
|
|
|
c, ca, b, NULL,
|
|
|
|
btree_node_data_missing,
|
2021-07-10 17:44:42 +00:00
|
|
|
"btree node data missing: expected %u sectors, found %u",
|
|
|
|
ptr_written, b->written);
|
|
|
|
} else {
|
|
|
|
for (bne = write_block(b);
|
2024-01-16 18:29:59 +00:00
|
|
|
bset_byte_offset(b, bne) < btree_buf_bytes(b);
|
2021-07-10 17:44:42 +00:00
|
|
|
bne = (void *) bne + block_bytes(c))
|
|
|
|
btree_err_on(bne->keys.seq == b->data->keys.seq &&
|
|
|
|
!bch2_journal_seq_is_blacklisted(c,
|
|
|
|
le64_to_cpu(bne->keys.journal_seq),
|
|
|
|
true),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_want_retry,
|
|
|
|
c, ca, b, NULL,
|
|
|
|
btree_node_bset_after_end,
|
2021-07-10 17:44:42 +00:00
|
|
|
"found bset signature after last bset");
|
|
|
|
}
|
2021-05-23 01:43:20 +00:00
|
|
|
|
2024-01-16 18:29:59 +00:00
|
|
|
sorted = btree_bounce_alloc(c, btree_buf_bytes(b), &used_mempool);
|
2017-03-17 06:18:50 +00:00
|
|
|
sorted->keys.u64s = 0;
|
|
|
|
|
|
|
|
set_btree_bset(b, b->set, &b->data->keys);
|
|
|
|
|
2021-02-20 05:00:23 +00:00
|
|
|
b->nr = bch2_key_sort_fix_overlapping(c, &sorted->keys, iter);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
u64s = le16_to_cpu(sorted->keys.u64s);
|
|
|
|
*sorted = *b->data;
|
|
|
|
sorted->keys.u64s = cpu_to_le16(u64s);
|
|
|
|
swap(sorted, b->data);
|
|
|
|
set_btree_bset(b, b->set, &b->data->keys);
|
|
|
|
b->nsets = 1;
|
|
|
|
|
|
|
|
BUG_ON(b->nr.live_u64s != u64s);
|
|
|
|
|
2024-01-16 18:29:59 +00:00
|
|
|
btree_bounce_free(c, btree_buf_bytes(b), used_mempool, sorted);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2021-04-24 20:32:35 +00:00
|
|
|
if (updated_range)
|
|
|
|
bch2_btree_node_drop_keys_outside_node(b);
|
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
i = &b->data->keys;
|
|
|
|
for (k = i->start; k != vstruct_last(i);) {
|
|
|
|
struct bkey tmp;
|
2020-02-07 18:38:02 +00:00
|
|
|
struct bkey_s u = __bkey_disassemble(b, k, &tmp);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2022-04-03 21:50:01 +00:00
|
|
|
printbuf_reset(&buf);
|
|
|
|
|
2022-04-04 01:50:25 +00:00
|
|
|
if (bch2_bkey_val_invalid(c, u.s_c, READ, &buf) ||
|
2020-11-02 23:20:44 +00:00
|
|
|
(bch2_inject_invalid_keys &&
|
2017-03-17 06:18:50 +00:00
|
|
|
!bversion_cmp(u.k->version, MAX_VERSION))) {
|
2022-04-03 21:50:01 +00:00
|
|
|
printbuf_reset(&buf);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_printf(&buf, "invalid bkey: ");
|
2022-04-04 01:50:25 +00:00
|
|
|
bch2_bkey_val_invalid(c, u.s_c, READ, &buf);
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_printf(&buf, "\n ");
|
2022-04-19 03:43:08 +00:00
|
|
|
bch2_bkey_val_to_text(&buf, c, u.s_c);
|
2022-04-03 21:50:01 +00:00
|
|
|
|
2023-10-25 00:44:36 +00:00
|
|
|
btree_err(-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, NULL, b, i,
|
|
|
|
btree_node_bad_bkey,
|
|
|
|
"%s", buf.buf);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
btree_keys_account_key_drop(&b->nr, 0, k);
|
|
|
|
|
|
|
|
i->u64s = cpu_to_le16(le16_to_cpu(i->u64s) - k->u64s);
|
2023-03-05 04:05:55 +00:00
|
|
|
memmove_u64s_down(k, bkey_p_next(k),
|
2017-03-17 06:18:50 +00:00
|
|
|
(u64 *) vstruct_end(i) - (u64 *) k);
|
|
|
|
set_btree_bset_end(b, b->set);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2020-02-07 18:38:02 +00:00
|
|
|
if (u.k->type == KEY_TYPE_btree_ptr_v2) {
|
|
|
|
struct bkey_s_btree_ptr_v2 bp = bkey_s_to_btree_ptr_v2(u);
|
|
|
|
|
|
|
|
bp.v->mem_ptr = 0;
|
|
|
|
}
|
|
|
|
|
2023-03-05 04:05:55 +00:00
|
|
|
k = bkey_p_next(k);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
bch2_bset_build_aux_tree(b, b->set, false);
|
|
|
|
|
2019-11-10 04:50:52 +00:00
|
|
|
set_needs_whiteout(btree_bset_first(b), true);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
btree_node_reset_sib_u64s(b);
|
2020-07-03 20:32:00 +00:00
|
|
|
|
|
|
|
bkey_for_each_ptr(bch2_bkey_ptrs(bkey_i_to_s(&b->key)), ptr) {
|
2023-09-12 22:41:22 +00:00
|
|
|
struct bch_dev *ca2 = bch_dev_bkey_exists(c, ptr->dev);
|
2020-07-03 20:32:00 +00:00
|
|
|
|
2023-09-12 22:41:22 +00:00
|
|
|
if (ca2->mi.state != BCH_MEMBER_STATE_rw)
|
2020-07-03 20:32:00 +00:00
|
|
|
set_btree_node_need_rewrite(b);
|
|
|
|
}
|
2021-07-10 17:44:42 +00:00
|
|
|
|
|
|
|
if (!ptr_written)
|
|
|
|
set_btree_node_need_rewrite(b);
|
2017-03-17 06:18:50 +00:00
|
|
|
out:
|
|
|
|
mempool_free(iter, &c->fill_iter);
|
2022-04-03 21:50:01 +00:00
|
|
|
printbuf_exit(&buf);
|
2024-01-03 21:42:33 +00:00
|
|
|
bch2_time_stats_update(&c->times[BCH_TIME_btree_node_read_done], start_time);
|
2017-03-17 06:18:50 +00:00
|
|
|
return retry_read;
|
|
|
|
fsck_err:
|
2023-08-03 23:36:28 +00:00
|
|
|
if (ret == -BCH_ERR_btree_node_read_err_want_retry ||
|
|
|
|
ret == -BCH_ERR_btree_node_read_err_must_retry)
|
2017-03-17 06:18:50 +00:00
|
|
|
retry_read = 1;
|
2023-02-01 20:45:45 +00:00
|
|
|
else
|
2017-03-17 06:18:50 +00:00
|
|
|
set_btree_node_read_error(b);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void btree_node_read_work(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct btree_read_bio *rb =
|
|
|
|
container_of(work, struct btree_read_bio, work);
|
|
|
|
struct bch_fs *c = rb->c;
|
2021-05-22 03:57:37 +00:00
|
|
|
struct btree *b = rb->b;
|
2017-03-17 06:18:50 +00:00
|
|
|
struct bch_dev *ca = bch_dev_bkey_exists(c, rb->pick.ptr.dev);
|
|
|
|
struct bio *bio = &rb->bio;
|
2018-11-01 19:28:45 +00:00
|
|
|
struct bch_io_failures failed = { .nr = 0 };
|
2022-02-25 18:18:19 +00:00
|
|
|
struct printbuf buf = PRINTBUF;
|
2021-04-24 06:47:41 +00:00
|
|
|
bool saw_error = false;
|
2022-05-26 19:19:20 +00:00
|
|
|
bool retry = false;
|
2017-03-17 06:18:50 +00:00
|
|
|
bool can_retry;
|
|
|
|
|
|
|
|
goto start;
|
|
|
|
while (1) {
|
2022-05-26 19:19:20 +00:00
|
|
|
retry = true;
|
2017-03-17 06:18:50 +00:00
|
|
|
bch_info(c, "retrying read");
|
|
|
|
ca = bch_dev_bkey_exists(c, rb->pick.ptr.dev);
|
|
|
|
rb->have_ioref = bch2_dev_get_ioref(ca, READ);
|
|
|
|
bio_reset(bio, NULL, REQ_OP_READ|REQ_SYNC|REQ_META);
|
|
|
|
bio->bi_iter.bi_sector = rb->pick.ptr.offset;
|
2024-01-16 18:29:59 +00:00
|
|
|
bio->bi_iter.bi_size = btree_buf_bytes(b);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
if (rb->have_ioref) {
|
|
|
|
bio_set_dev(bio, ca->disk_sb.bdev);
|
|
|
|
submit_bio_wait(bio);
|
|
|
|
} else {
|
|
|
|
bio->bi_status = BLK_STS_REMOVED;
|
|
|
|
}
|
|
|
|
start:
|
2022-02-25 18:18:19 +00:00
|
|
|
printbuf_reset(&buf);
|
2023-10-20 02:49:08 +00:00
|
|
|
bch2_btree_pos_to_text(&buf, c, b);
|
2023-10-25 20:29:37 +00:00
|
|
|
bch2_dev_io_err_on(bio->bi_status, ca, BCH_MEMBER_ERROR_read,
|
|
|
|
"btree read error %s for %s",
|
2022-02-25 18:18:19 +00:00
|
|
|
bch2_blk_status_to_str(bio->bi_status), buf.buf);
|
2017-03-17 06:18:50 +00:00
|
|
|
if (rb->have_ioref)
|
|
|
|
percpu_ref_put(&ca->io_ref);
|
|
|
|
rb->have_ioref = false;
|
|
|
|
|
2018-11-01 19:28:45 +00:00
|
|
|
bch2_mark_io_failure(&failed, &rb->pick);
|
|
|
|
|
2018-11-01 19:10:01 +00:00
|
|
|
can_retry = bch2_bkey_pick_read_device(c,
|
|
|
|
bkey_i_to_s_c(&b->key),
|
|
|
|
&failed, &rb->pick) > 0;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
if (!bio->bi_status &&
|
2023-01-03 22:32:16 +00:00
|
|
|
!bch2_btree_node_read_done(c, ca, b, can_retry, &saw_error)) {
|
2022-05-26 19:19:20 +00:00
|
|
|
if (retry)
|
|
|
|
bch_info(c, "retry success");
|
2017-03-17 06:18:50 +00:00
|
|
|
break;
|
2022-05-26 19:19:20 +00:00
|
|
|
}
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2021-04-24 06:47:41 +00:00
|
|
|
saw_error = true;
|
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
if (!can_retry) {
|
|
|
|
set_btree_node_read_error(b);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-03-21 20:28:57 +00:00
|
|
|
bch2_time_stats_update(&c->times[BCH_TIME_btree_node_read],
|
|
|
|
rb->start_time);
|
2017-03-17 06:18:50 +00:00
|
|
|
bio_put(&rb->bio);
|
2021-04-24 06:47:41 +00:00
|
|
|
|
2023-02-10 20:47:46 +00:00
|
|
|
if (saw_error && !btree_node_read_error(b)) {
|
2023-09-12 22:41:22 +00:00
|
|
|
printbuf_reset(&buf);
|
2023-02-10 20:47:46 +00:00
|
|
|
bch2_bpos_to_text(&buf, b->key.k.p);
|
2024-03-17 01:22:47 +00:00
|
|
|
bch_err_ratelimited(c, "%s: rewriting btree node at btree=%s level=%u %s due to error",
|
2023-10-20 02:49:08 +00:00
|
|
|
__func__, bch2_btree_id_str(b->c.btree_id), b->c.level, buf.buf);
|
2023-02-10 20:47:46 +00:00
|
|
|
|
2021-04-24 06:47:41 +00:00
|
|
|
bch2_btree_node_rewrite_async(c, b);
|
2023-02-10 20:47:46 +00:00
|
|
|
}
|
2021-04-24 06:47:41 +00:00
|
|
|
|
2023-09-12 22:41:22 +00:00
|
|
|
printbuf_exit(&buf);
|
2017-03-17 06:18:50 +00:00
|
|
|
clear_btree_node_read_in_flight(b);
|
|
|
|
wake_up_bit(&b->flags, BTREE_NODE_read_in_flight);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void btree_node_read_endio(struct bio *bio)
|
|
|
|
{
|
|
|
|
struct btree_read_bio *rb =
|
|
|
|
container_of(bio, struct btree_read_bio, bio);
|
|
|
|
struct bch_fs *c = rb->c;
|
|
|
|
|
|
|
|
if (rb->have_ioref) {
|
|
|
|
struct bch_dev *ca = bch_dev_bkey_exists(c, rb->pick.ptr.dev);
|
2022-10-22 19:59:53 +00:00
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
bch2_latency_acct(ca, rb->start_time, READ);
|
|
|
|
}
|
|
|
|
|
2021-05-22 21:37:25 +00:00
|
|
|
queue_work(c->io_complete_wq, &rb->work);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
2021-05-22 03:57:37 +00:00
|
|
|
struct btree_node_read_all {
|
|
|
|
struct closure cl;
|
|
|
|
struct bch_fs *c;
|
|
|
|
struct btree *b;
|
|
|
|
unsigned nr;
|
|
|
|
void *buf[BCH_REPLICAS_MAX];
|
|
|
|
struct bio *bio[BCH_REPLICAS_MAX];
|
2023-07-07 02:47:42 +00:00
|
|
|
blk_status_t err[BCH_REPLICAS_MAX];
|
2021-05-22 03:57:37 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
static unsigned btree_node_sectors_written(struct bch_fs *c, void *data)
|
|
|
|
{
|
|
|
|
struct btree_node *bn = data;
|
|
|
|
struct btree_node_entry *bne;
|
|
|
|
unsigned offset = 0;
|
|
|
|
|
|
|
|
if (le64_to_cpu(bn->magic) != bset_magic(c))
|
|
|
|
return 0;
|
|
|
|
|
2021-12-14 19:24:41 +00:00
|
|
|
while (offset < btree_sectors(c)) {
|
2021-05-22 03:57:37 +00:00
|
|
|
if (!offset) {
|
|
|
|
offset += vstruct_sectors(bn, c->block_bits);
|
|
|
|
} else {
|
|
|
|
bne = data + (offset << 9);
|
|
|
|
if (bne->keys.seq != bn->keys.seq)
|
|
|
|
break;
|
|
|
|
offset += vstruct_sectors(bne, c->block_bits);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return offset;
|
|
|
|
}
|
|
|
|
|
|
|
|
static bool btree_node_has_extra_bsets(struct bch_fs *c, unsigned offset, void *data)
|
|
|
|
{
|
|
|
|
struct btree_node *bn = data;
|
|
|
|
struct btree_node_entry *bne;
|
|
|
|
|
|
|
|
if (!offset)
|
|
|
|
return false;
|
|
|
|
|
2021-12-14 19:24:41 +00:00
|
|
|
while (offset < btree_sectors(c)) {
|
2021-05-22 03:57:37 +00:00
|
|
|
bne = data + (offset << 9);
|
|
|
|
if (bne->keys.seq == bn->keys.seq)
|
|
|
|
return true;
|
|
|
|
offset++;
|
|
|
|
}
|
|
|
|
|
|
|
|
return false;
|
|
|
|
return offset;
|
|
|
|
}
|
|
|
|
|
2023-11-18 00:13:27 +00:00
|
|
|
static CLOSURE_CALLBACK(btree_node_read_all_replicas_done)
|
2021-05-22 03:57:37 +00:00
|
|
|
{
|
2023-11-18 00:13:27 +00:00
|
|
|
closure_type(ra, struct btree_node_read_all, cl);
|
2021-05-22 03:57:37 +00:00
|
|
|
struct bch_fs *c = ra->c;
|
|
|
|
struct btree *b = ra->b;
|
2022-02-25 18:18:19 +00:00
|
|
|
struct printbuf buf = PRINTBUF;
|
2021-05-22 03:57:37 +00:00
|
|
|
bool dump_bset_maps = false;
|
|
|
|
bool have_retry = false;
|
2021-06-23 01:51:17 +00:00
|
|
|
int ret = 0, best = -1, write = READ;
|
2021-09-09 23:05:34 +00:00
|
|
|
unsigned i, written = 0, written2 = 0;
|
2021-05-22 03:57:37 +00:00
|
|
|
__le64 seq = b->key.k.type == KEY_TYPE_btree_ptr_v2
|
|
|
|
? bkey_i_to_btree_ptr_v2(&b->key)->v.seq : 0;
|
2023-01-03 22:32:16 +00:00
|
|
|
bool _saw_error = false, *saw_error = &_saw_error;
|
2021-05-22 03:57:37 +00:00
|
|
|
|
|
|
|
for (i = 0; i < ra->nr; i++) {
|
2021-06-23 01:51:17 +00:00
|
|
|
struct btree_node *bn = ra->buf[i];
|
|
|
|
|
2021-05-22 03:57:37 +00:00
|
|
|
if (ra->err[i])
|
|
|
|
continue;
|
|
|
|
|
2021-06-23 01:51:17 +00:00
|
|
|
if (le64_to_cpu(bn->magic) != bset_magic(c) ||
|
|
|
|
(seq && seq != bn->keys.seq))
|
|
|
|
continue;
|
2021-05-22 03:57:37 +00:00
|
|
|
|
2021-06-23 01:51:17 +00:00
|
|
|
if (best < 0) {
|
|
|
|
best = i;
|
|
|
|
written = btree_node_sectors_written(c, bn);
|
|
|
|
continue;
|
2021-05-22 03:57:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
written2 = btree_node_sectors_written(c, ra->buf[i]);
|
2023-10-25 00:44:36 +00:00
|
|
|
if (btree_err_on(written2 != written, -BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, NULL, b, NULL,
|
|
|
|
btree_node_replicas_sectors_written_mismatch,
|
2021-05-22 03:57:37 +00:00
|
|
|
"btree node sectors written mismatch: %u != %u",
|
|
|
|
written, written2) ||
|
|
|
|
btree_err_on(btree_node_has_extra_bsets(c, written2, ra->buf[i]),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, NULL, b, NULL,
|
|
|
|
btree_node_bset_after_end,
|
2021-05-22 03:57:37 +00:00
|
|
|
"found bset signature after last bset") ||
|
2021-06-23 01:51:17 +00:00
|
|
|
btree_err_on(memcmp(ra->buf[best], ra->buf[i], written << 9),
|
2023-10-25 00:44:36 +00:00
|
|
|
-BCH_ERR_btree_node_read_err_fixable,
|
|
|
|
c, NULL, b, NULL,
|
|
|
|
btree_node_replicas_data_mismatch,
|
2021-05-22 03:57:37 +00:00
|
|
|
"btree node replicas content mismatch"))
|
|
|
|
dump_bset_maps = true;
|
|
|
|
|
|
|
|
if (written2 > written) {
|
|
|
|
written = written2;
|
2021-06-23 01:51:17 +00:00
|
|
|
best = i;
|
2021-05-22 03:57:37 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
fsck_err:
|
|
|
|
if (dump_bset_maps) {
|
|
|
|
for (i = 0; i < ra->nr; i++) {
|
|
|
|
struct btree_node *bn = ra->buf[i];
|
|
|
|
struct btree_node_entry *bne = NULL;
|
|
|
|
unsigned offset = 0, sectors;
|
|
|
|
bool gap = false;
|
|
|
|
|
|
|
|
if (ra->err[i])
|
|
|
|
continue;
|
|
|
|
|
2022-02-25 18:18:19 +00:00
|
|
|
printbuf_reset(&buf);
|
|
|
|
|
2021-12-14 19:24:41 +00:00
|
|
|
while (offset < btree_sectors(c)) {
|
2021-05-22 03:57:37 +00:00
|
|
|
if (!offset) {
|
|
|
|
sectors = vstruct_sectors(bn, c->block_bits);
|
|
|
|
} else {
|
|
|
|
bne = ra->buf[i] + (offset << 9);
|
|
|
|
if (bne->keys.seq != bn->keys.seq)
|
|
|
|
break;
|
|
|
|
sectors = vstruct_sectors(bne, c->block_bits);
|
|
|
|
}
|
|
|
|
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_printf(&buf, " %u-%u", offset, offset + sectors);
|
2021-05-22 03:57:37 +00:00
|
|
|
if (bne && bch2_journal_seq_is_blacklisted(c,
|
|
|
|
le64_to_cpu(bne->keys.journal_seq), false))
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_printf(&buf, "*");
|
2021-05-22 03:57:37 +00:00
|
|
|
offset += sectors;
|
|
|
|
}
|
|
|
|
|
2021-12-14 19:24:41 +00:00
|
|
|
while (offset < btree_sectors(c)) {
|
2021-05-22 03:57:37 +00:00
|
|
|
bne = ra->buf[i] + (offset << 9);
|
|
|
|
if (bne->keys.seq == bn->keys.seq) {
|
|
|
|
if (!gap)
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_printf(&buf, " GAP");
|
2021-05-22 03:57:37 +00:00
|
|
|
gap = true;
|
|
|
|
|
|
|
|
sectors = vstruct_sectors(bne, c->block_bits);
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_printf(&buf, " %u-%u", offset, offset + sectors);
|
2021-05-22 03:57:37 +00:00
|
|
|
if (bch2_journal_seq_is_blacklisted(c,
|
|
|
|
le64_to_cpu(bne->keys.journal_seq), false))
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_printf(&buf, "*");
|
2021-05-22 03:57:37 +00:00
|
|
|
}
|
|
|
|
offset++;
|
|
|
|
}
|
|
|
|
|
2022-02-25 18:18:19 +00:00
|
|
|
bch_err(c, "replica %u:%s", i, buf.buf);
|
2021-05-22 03:57:37 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-06-23 01:51:17 +00:00
|
|
|
if (best >= 0) {
|
2024-01-16 18:29:59 +00:00
|
|
|
memcpy(b->data, ra->buf[best], btree_buf_bytes(b));
|
2023-01-03 22:32:16 +00:00
|
|
|
ret = bch2_btree_node_read_done(c, NULL, b, false, saw_error);
|
2021-06-23 01:51:17 +00:00
|
|
|
} else {
|
|
|
|
ret = -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (ret)
|
2021-05-22 03:57:37 +00:00
|
|
|
set_btree_node_read_error(b);
|
2023-01-03 22:32:16 +00:00
|
|
|
else if (*saw_error)
|
|
|
|
bch2_btree_node_rewrite_async(c, b);
|
2021-05-22 03:57:37 +00:00
|
|
|
|
|
|
|
for (i = 0; i < ra->nr; i++) {
|
|
|
|
mempool_free(ra->buf[i], &c->btree_bounce_pool);
|
|
|
|
bio_put(ra->bio[i]);
|
|
|
|
}
|
|
|
|
|
|
|
|
closure_debug_destroy(&ra->cl);
|
|
|
|
kfree(ra);
|
2022-02-25 18:18:19 +00:00
|
|
|
printbuf_exit(&buf);
|
2021-05-22 03:57:37 +00:00
|
|
|
|
|
|
|
clear_btree_node_read_in_flight(b);
|
|
|
|
wake_up_bit(&b->flags, BTREE_NODE_read_in_flight);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void btree_node_read_all_replicas_endio(struct bio *bio)
|
|
|
|
{
|
|
|
|
struct btree_read_bio *rb =
|
|
|
|
container_of(bio, struct btree_read_bio, bio);
|
|
|
|
struct bch_fs *c = rb->c;
|
|
|
|
struct btree_node_read_all *ra = rb->ra;
|
|
|
|
|
|
|
|
if (rb->have_ioref) {
|
|
|
|
struct bch_dev *ca = bch_dev_bkey_exists(c, rb->pick.ptr.dev);
|
2022-10-22 19:59:53 +00:00
|
|
|
|
2021-05-22 03:57:37 +00:00
|
|
|
bch2_latency_acct(ca, rb->start_time, READ);
|
|
|
|
}
|
|
|
|
|
|
|
|
ra->err[rb->idx] = bio->bi_status;
|
|
|
|
closure_put(&ra->cl);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* XXX This allocates multiple times from the same mempools, and can deadlock
|
|
|
|
* under sufficient memory pressure (but is only a debug path)
|
|
|
|
*/
|
|
|
|
static int btree_node_read_all_replicas(struct bch_fs *c, struct btree *b, bool sync)
|
|
|
|
{
|
|
|
|
struct bkey_s_c k = bkey_i_to_s_c(&b->key);
|
|
|
|
struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
|
|
|
|
const union bch_extent_entry *entry;
|
|
|
|
struct extent_ptr_decoded pick;
|
|
|
|
struct btree_node_read_all *ra;
|
|
|
|
unsigned i;
|
|
|
|
|
|
|
|
ra = kzalloc(sizeof(*ra), GFP_NOFS);
|
|
|
|
if (!ra)
|
2023-03-14 19:35:57 +00:00
|
|
|
return -BCH_ERR_ENOMEM_btree_node_read_all_replicas;
|
2021-05-22 03:57:37 +00:00
|
|
|
|
|
|
|
closure_init(&ra->cl, NULL);
|
|
|
|
ra->c = c;
|
|
|
|
ra->b = b;
|
|
|
|
ra->nr = bch2_bkey_nr_ptrs(k);
|
|
|
|
|
|
|
|
for (i = 0; i < ra->nr; i++) {
|
|
|
|
ra->buf[i] = mempool_alloc(&c->btree_bounce_pool, GFP_NOFS);
|
|
|
|
ra->bio[i] = bio_alloc_bioset(NULL,
|
2024-01-16 18:29:59 +00:00
|
|
|
buf_pages(ra->buf[i], btree_buf_bytes(b)),
|
2021-05-22 03:57:37 +00:00
|
|
|
REQ_OP_READ|REQ_SYNC|REQ_META,
|
|
|
|
GFP_NOFS,
|
|
|
|
&c->btree_bio);
|
|
|
|
}
|
|
|
|
|
|
|
|
i = 0;
|
|
|
|
bkey_for_each_ptr_decode(k.k, ptrs, pick, entry) {
|
|
|
|
struct bch_dev *ca = bch_dev_bkey_exists(c, pick.ptr.dev);
|
|
|
|
struct btree_read_bio *rb =
|
|
|
|
container_of(ra->bio[i], struct btree_read_bio, bio);
|
|
|
|
rb->c = c;
|
|
|
|
rb->b = b;
|
|
|
|
rb->ra = ra;
|
|
|
|
rb->start_time = local_clock();
|
|
|
|
rb->have_ioref = bch2_dev_get_ioref(ca, READ);
|
|
|
|
rb->idx = i;
|
|
|
|
rb->pick = pick;
|
|
|
|
rb->bio.bi_iter.bi_sector = pick.ptr.offset;
|
|
|
|
rb->bio.bi_end_io = btree_node_read_all_replicas_endio;
|
2024-01-16 18:29:59 +00:00
|
|
|
bch2_bio_map(&rb->bio, ra->buf[i], btree_buf_bytes(b));
|
2021-05-22 03:57:37 +00:00
|
|
|
|
|
|
|
if (rb->have_ioref) {
|
|
|
|
this_cpu_add(ca->io_done->sectors[READ][BCH_DATA_btree],
|
|
|
|
bio_sectors(&rb->bio));
|
|
|
|
bio_set_dev(&rb->bio, ca->disk_sb.bdev);
|
|
|
|
|
|
|
|
closure_get(&ra->cl);
|
|
|
|
submit_bio(&rb->bio);
|
|
|
|
} else {
|
|
|
|
ra->err[i] = BLK_STS_REMOVED;
|
|
|
|
}
|
|
|
|
|
|
|
|
i++;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (sync) {
|
|
|
|
closure_sync(&ra->cl);
|
2023-11-18 00:13:27 +00:00
|
|
|
btree_node_read_all_replicas_done(&ra->cl.work);
|
2021-05-22 03:57:37 +00:00
|
|
|
} else {
|
2021-05-22 21:37:25 +00:00
|
|
|
continue_at(&ra->cl, btree_node_read_all_replicas_done,
|
|
|
|
c->io_complete_wq);
|
2021-05-22 03:57:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2023-12-02 08:36:27 +00:00
|
|
|
void bch2_btree_node_read(struct btree_trans *trans, struct btree *b,
|
2017-03-17 06:18:50 +00:00
|
|
|
bool sync)
|
|
|
|
{
|
2023-12-02 08:36:27 +00:00
|
|
|
struct bch_fs *c = trans->c;
|
2018-10-02 15:03:39 +00:00
|
|
|
struct extent_ptr_decoded pick;
|
2017-03-17 06:18:50 +00:00
|
|
|
struct btree_read_bio *rb;
|
|
|
|
struct bch_dev *ca;
|
|
|
|
struct bio *bio;
|
|
|
|
int ret;
|
|
|
|
|
2023-12-02 08:36:27 +00:00
|
|
|
trace_and_count(c, btree_node_read, trans, b);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2021-05-22 03:57:37 +00:00
|
|
|
if (bch2_verify_all_btree_replicas &&
|
|
|
|
!btree_node_read_all_replicas(c, b, sync))
|
2022-06-07 02:09:11 +00:00
|
|
|
return;
|
2021-05-22 03:57:37 +00:00
|
|
|
|
2018-11-01 19:10:01 +00:00
|
|
|
ret = bch2_bkey_pick_read_device(c, bkey_i_to_s_c(&b->key),
|
|
|
|
NULL, &pick);
|
2022-06-07 02:09:11 +00:00
|
|
|
|
|
|
|
if (ret <= 0) {
|
|
|
|
struct printbuf buf = PRINTBUF;
|
|
|
|
|
2023-02-04 02:01:40 +00:00
|
|
|
prt_str(&buf, "btree node read error: no device to read from\n at ");
|
2023-10-20 02:49:08 +00:00
|
|
|
bch2_btree_pos_to_text(&buf, c, b);
|
2022-06-07 02:09:11 +00:00
|
|
|
bch_err(c, "%s", buf.buf);
|
|
|
|
|
2023-07-17 03:19:49 +00:00
|
|
|
if (c->recovery_passes_explicit & BIT_ULL(BCH_RECOVERY_PASS_check_topology) &&
|
|
|
|
c->curr_recovery_pass > BCH_RECOVERY_PASS_check_topology)
|
2022-06-07 02:09:11 +00:00
|
|
|
bch2_fatal_error(c);
|
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
set_btree_node_read_error(b);
|
2022-06-07 02:09:11 +00:00
|
|
|
clear_btree_node_read_in_flight(b);
|
|
|
|
wake_up_bit(&b->flags, BTREE_NODE_read_in_flight);
|
|
|
|
printbuf_exit(&buf);
|
|
|
|
return;
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
ca = bch_dev_bkey_exists(c, pick.ptr.dev);
|
|
|
|
|
|
|
|
bio = bio_alloc_bioset(NULL,
|
2024-01-16 18:29:59 +00:00
|
|
|
buf_pages(b->data, btree_buf_bytes(b)),
|
2017-03-17 06:18:50 +00:00
|
|
|
REQ_OP_READ|REQ_SYNC|REQ_META,
|
2023-05-28 22:02:38 +00:00
|
|
|
GFP_NOFS,
|
2017-03-17 06:18:50 +00:00
|
|
|
&c->btree_bio);
|
|
|
|
rb = container_of(bio, struct btree_read_bio, bio);
|
|
|
|
rb->c = c;
|
2021-05-22 03:57:37 +00:00
|
|
|
rb->b = b;
|
|
|
|
rb->ra = NULL;
|
2017-03-17 06:18:50 +00:00
|
|
|
rb->start_time = local_clock();
|
|
|
|
rb->have_ioref = bch2_dev_get_ioref(ca, READ);
|
|
|
|
rb->pick = pick;
|
|
|
|
INIT_WORK(&rb->work, btree_node_read_work);
|
|
|
|
bio->bi_iter.bi_sector = pick.ptr.offset;
|
|
|
|
bio->bi_end_io = btree_node_read_endio;
|
2024-01-16 18:29:59 +00:00
|
|
|
bch2_bio_map(bio, b->data, btree_buf_bytes(b));
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
if (rb->have_ioref) {
|
2020-07-09 22:28:11 +00:00
|
|
|
this_cpu_add(ca->io_done->sectors[READ][BCH_DATA_btree],
|
2017-03-17 06:18:50 +00:00
|
|
|
bio_sectors(bio));
|
|
|
|
bio_set_dev(bio, ca->disk_sb.bdev);
|
|
|
|
|
|
|
|
if (sync) {
|
|
|
|
submit_bio_wait(bio);
|
2024-01-03 22:32:07 +00:00
|
|
|
bch2_latency_acct(ca, rb->start_time, READ);
|
2017-03-17 06:18:50 +00:00
|
|
|
btree_node_read_work(&rb->work);
|
|
|
|
} else {
|
|
|
|
submit_bio(bio);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
bio->bi_status = BLK_STS_REMOVED;
|
|
|
|
|
|
|
|
if (sync)
|
|
|
|
btree_node_read_work(&rb->work);
|
|
|
|
else
|
2021-05-22 21:37:25 +00:00
|
|
|
queue_work(c->io_complete_wq, &rb->work);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2023-03-02 07:12:18 +00:00
|
|
|
static int __bch2_btree_root_read(struct btree_trans *trans, enum btree_id id,
|
|
|
|
const struct bkey_i *k, unsigned level)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
2023-03-02 07:12:18 +00:00
|
|
|
struct bch_fs *c = trans->c;
|
2017-03-17 06:18:50 +00:00
|
|
|
struct closure cl;
|
|
|
|
struct btree *b;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
closure_init_stack(&cl);
|
|
|
|
|
|
|
|
do {
|
2023-12-02 08:36:27 +00:00
|
|
|
ret = bch2_btree_cache_cannibalize_lock(trans, &cl);
|
2017-03-17 06:18:50 +00:00
|
|
|
closure_sync(&cl);
|
|
|
|
} while (ret);
|
|
|
|
|
2023-03-02 07:12:18 +00:00
|
|
|
b = bch2_btree_node_mem_alloc(trans, level != 0);
|
2023-12-02 08:36:27 +00:00
|
|
|
bch2_btree_cache_cannibalize_unlock(trans);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
BUG_ON(IS_ERR(b));
|
|
|
|
|
|
|
|
bkey_copy(&b->key, k);
|
|
|
|
BUG_ON(bch2_btree_node_hash_insert(&c->btree_cache, b, level, id));
|
|
|
|
|
2021-07-11 03:03:15 +00:00
|
|
|
set_btree_node_read_in_flight(b);
|
|
|
|
|
2023-12-02 08:36:27 +00:00
|
|
|
bch2_btree_node_read(trans, b, true);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
if (btree_node_read_error(b)) {
|
|
|
|
bch2_btree_node_hash_remove(&c->btree_cache, b);
|
|
|
|
|
|
|
|
mutex_lock(&c->btree_cache.lock);
|
|
|
|
list_move(&b->list, &c->btree_cache.freeable);
|
|
|
|
mutex_unlock(&c->btree_cache.lock);
|
|
|
|
|
2024-02-06 22:24:18 +00:00
|
|
|
ret = -BCH_ERR_btree_node_read_error;
|
2017-03-17 06:18:50 +00:00
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
|
|
|
bch2_btree_set_root_for_read(c, b);
|
|
|
|
err:
|
2020-06-06 16:28:01 +00:00
|
|
|
six_unlock_write(&b->c.lock);
|
|
|
|
six_unlock_intent(&b->c.lock);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2023-03-02 07:12:18 +00:00
|
|
|
int bch2_btree_root_read(struct bch_fs *c, enum btree_id id,
|
|
|
|
const struct bkey_i *k, unsigned level)
|
|
|
|
{
|
2023-09-12 21:16:02 +00:00
|
|
|
return bch2_trans_run(c, __bch2_btree_root_read(trans, id, k, level));
|
2023-03-02 07:12:18 +00:00
|
|
|
}
|
|
|
|
|
2023-12-01 04:32:20 +00:00
|
|
|
static void bch2_btree_complete_write(struct bch_fs *c, struct btree *b,
|
|
|
|
struct btree_write *w)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
|
|
|
unsigned long old, new, v = READ_ONCE(b->will_make_reachable);
|
|
|
|
|
|
|
|
do {
|
|
|
|
old = new = v;
|
|
|
|
if (!(old & 1))
|
|
|
|
break;
|
|
|
|
|
|
|
|
new &= ~1UL;
|
|
|
|
} while ((v = cmpxchg(&b->will_make_reachable, old, new)) != old);
|
|
|
|
|
|
|
|
if (old & 1)
|
|
|
|
closure_put(&((struct btree_update *) new)->cl);
|
|
|
|
|
|
|
|
bch2_journal_pin_drop(&c->journal, &w->journal);
|
|
|
|
}
|
|
|
|
|
2022-02-27 02:35:16 +00:00
|
|
|
static void __btree_node_write_done(struct bch_fs *c, struct btree *b)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
|
|
|
struct btree_write *w = btree_prev_write(b);
|
2021-07-11 20:41:14 +00:00
|
|
|
unsigned long old, new, v;
|
2022-11-17 21:03:15 +00:00
|
|
|
unsigned type = 0;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
bch2_btree_complete_write(c, b, w);
|
2021-07-11 20:41:14 +00:00
|
|
|
|
|
|
|
v = READ_ONCE(b->flags);
|
|
|
|
do {
|
|
|
|
old = new = v;
|
|
|
|
|
|
|
|
if ((old & (1U << BTREE_NODE_dirty)) &&
|
|
|
|
(old & (1U << BTREE_NODE_need_write)) &&
|
|
|
|
!(old & (1U << BTREE_NODE_never_write)) &&
|
2022-02-27 14:56:33 +00:00
|
|
|
!(old & (1U << BTREE_NODE_write_blocked)) &&
|
|
|
|
!(old & (1U << BTREE_NODE_will_make_reachable))) {
|
2021-07-11 20:41:14 +00:00
|
|
|
new &= ~(1U << BTREE_NODE_dirty);
|
|
|
|
new &= ~(1U << BTREE_NODE_need_write);
|
|
|
|
new |= (1U << BTREE_NODE_write_in_flight);
|
2021-07-10 17:44:42 +00:00
|
|
|
new |= (1U << BTREE_NODE_write_in_flight_inner);
|
2021-07-11 20:41:14 +00:00
|
|
|
new |= (1U << BTREE_NODE_just_written);
|
|
|
|
new ^= (1U << BTREE_NODE_write_idx);
|
2022-11-17 21:03:15 +00:00
|
|
|
|
|
|
|
type = new & BTREE_WRITE_TYPE_MASK;
|
|
|
|
new &= ~BTREE_WRITE_TYPE_MASK;
|
2021-07-11 20:41:14 +00:00
|
|
|
} else {
|
|
|
|
new &= ~(1U << BTREE_NODE_write_in_flight);
|
2021-07-10 17:44:42 +00:00
|
|
|
new &= ~(1U << BTREE_NODE_write_in_flight_inner);
|
2021-07-11 20:41:14 +00:00
|
|
|
}
|
|
|
|
} while ((v = cmpxchg(&b->flags, old, new)) != old);
|
|
|
|
|
|
|
|
if (new & (1U << BTREE_NODE_write_in_flight))
|
2022-11-17 21:03:15 +00:00
|
|
|
__bch2_btree_node_write(c, b, BTREE_WRITE_ALREADY_STARTED|type);
|
2022-03-10 22:35:06 +00:00
|
|
|
else
|
|
|
|
wake_up_bit(&b->flags, BTREE_NODE_write_in_flight);
|
2022-02-27 02:35:16 +00:00
|
|
|
}
|
2021-07-11 20:41:14 +00:00
|
|
|
|
2022-02-27 02:35:16 +00:00
|
|
|
static void btree_node_write_done(struct bch_fs *c, struct btree *b)
|
|
|
|
{
|
2023-09-12 21:16:02 +00:00
|
|
|
struct btree_trans *trans = bch2_trans_get(c);
|
2022-08-21 18:29:43 +00:00
|
|
|
|
2023-09-12 21:16:02 +00:00
|
|
|
btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_read);
|
2022-02-27 02:35:16 +00:00
|
|
|
__btree_node_write_done(c, b);
|
2021-07-11 20:41:14 +00:00
|
|
|
six_unlock_read(&b->c.lock);
|
2022-08-21 18:29:43 +00:00
|
|
|
|
2023-09-12 21:16:02 +00:00
|
|
|
bch2_trans_put(trans);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
2021-07-10 17:44:42 +00:00
|
|
|
static void btree_node_write_work(struct work_struct *work)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
2021-07-10 17:44:42 +00:00
|
|
|
struct btree_write_bio *wbio =
|
|
|
|
container_of(work, struct btree_write_bio, work);
|
|
|
|
struct bch_fs *c = wbio->wbio.c;
|
2017-03-17 06:18:50 +00:00
|
|
|
struct btree *b = wbio->wbio.bio.bi_private;
|
|
|
|
struct bch_extent_ptr *ptr;
|
2023-04-21 07:33:45 +00:00
|
|
|
int ret = 0;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2021-07-10 17:44:42 +00:00
|
|
|
btree_bounce_free(c,
|
|
|
|
wbio->data_bytes,
|
|
|
|
wbio->wbio.used_mempool,
|
|
|
|
wbio->data);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2021-07-10 17:44:42 +00:00
|
|
|
bch2_bkey_drop_ptrs(bkey_i_to_s(&wbio->key), ptr,
|
2018-09-30 22:28:23 +00:00
|
|
|
bch2_dev_list_has_dev(wbio->wbio.failed, ptr->dev));
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2023-12-24 02:09:34 +00:00
|
|
|
if (!bch2_bkey_nr_ptrs(bkey_i_to_s_c(&wbio->key))) {
|
2024-02-06 22:24:18 +00:00
|
|
|
ret = -BCH_ERR_btree_node_write_all_failed;
|
2017-03-17 06:18:50 +00:00
|
|
|
goto err;
|
2023-12-24 02:09:34 +00:00
|
|
|
}
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2021-07-10 17:44:42 +00:00
|
|
|
if (wbio->wbio.first_btree_write) {
|
|
|
|
if (wbio->wbio.failed.nr) {
|
|
|
|
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
ret = bch2_trans_do(c, NULL, NULL, 0,
|
2023-09-12 21:16:02 +00:00
|
|
|
bch2_btree_node_update_key_get_iter(trans, b, &wbio->key,
|
2023-06-27 21:32:48 +00:00
|
|
|
BCH_WATERMARK_reclaim|
|
2023-11-11 21:31:50 +00:00
|
|
|
BCH_TRANS_COMMIT_journal_reclaim|
|
|
|
|
BCH_TRANS_COMMIT_no_enospc|
|
|
|
|
BCH_TRANS_COMMIT_no_check_rw,
|
2023-06-27 21:32:48 +00:00
|
|
|
!wbio->wbio.failed.nr));
|
2021-07-10 17:44:42 +00:00
|
|
|
if (ret)
|
|
|
|
goto err;
|
|
|
|
}
|
2017-03-17 06:18:50 +00:00
|
|
|
out:
|
|
|
|
bio_put(&wbio->wbio.bio);
|
|
|
|
btree_node_write_done(c, b);
|
|
|
|
return;
|
|
|
|
err:
|
|
|
|
set_btree_node_noevict(b);
|
2024-03-18 01:51:19 +00:00
|
|
|
bch2_fs_fatal_err_on(!bch2_err_matches(ret, EROFS), c,
|
|
|
|
"writing btree node: %s", bch2_err_str(ret));
|
2017-03-17 06:18:50 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void btree_node_write_endio(struct bio *bio)
|
|
|
|
{
|
|
|
|
struct bch_write_bio *wbio = to_wbio(bio);
|
|
|
|
struct bch_write_bio *parent = wbio->split ? wbio->parent : NULL;
|
|
|
|
struct bch_write_bio *orig = parent ?: wbio;
|
2021-07-10 17:44:42 +00:00
|
|
|
struct btree_write_bio *wb = container_of(orig, struct btree_write_bio, wbio);
|
2017-03-17 06:18:50 +00:00
|
|
|
struct bch_fs *c = wbio->c;
|
2021-07-10 17:44:42 +00:00
|
|
|
struct btree *b = wbio->bio.bi_private;
|
2017-03-17 06:18:50 +00:00
|
|
|
struct bch_dev *ca = bch_dev_bkey_exists(c, wbio->dev);
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
if (wbio->have_ioref)
|
|
|
|
bch2_latency_acct(ca, wbio->submit_time, WRITE);
|
|
|
|
|
2023-10-25 20:29:37 +00:00
|
|
|
if (bch2_dev_io_err_on(bio->bi_status, ca, BCH_MEMBER_ERROR_write,
|
|
|
|
"btree write error: %s",
|
2020-07-21 17:34:22 +00:00
|
|
|
bch2_blk_status_to_str(bio->bi_status)) ||
|
2017-03-17 06:18:50 +00:00
|
|
|
bch2_meta_write_fault("btree")) {
|
|
|
|
spin_lock_irqsave(&c->btree_write_error_lock, flags);
|
|
|
|
bch2_dev_list_add_dev(&orig->failed, wbio->dev);
|
|
|
|
spin_unlock_irqrestore(&c->btree_write_error_lock, flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (wbio->have_ioref)
|
|
|
|
percpu_ref_put(&ca->io_ref);
|
|
|
|
|
|
|
|
if (parent) {
|
|
|
|
bio_put(bio);
|
|
|
|
bio_endio(&parent->bio);
|
2021-07-10 17:44:42 +00:00
|
|
|
return;
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
2021-07-10 17:44:42 +00:00
|
|
|
|
|
|
|
clear_btree_node_write_in_flight_inner(b);
|
|
|
|
wake_up_bit(&b->flags, BTREE_NODE_write_in_flight_inner);
|
|
|
|
INIT_WORK(&wb->work, btree_node_write_work);
|
|
|
|
queue_work(c->btree_io_complete_wq, &wb->work);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static int validate_bset_for_write(struct bch_fs *c, struct btree *b,
|
|
|
|
struct bset *i, unsigned sectors)
|
|
|
|
{
|
2022-04-03 21:50:01 +00:00
|
|
|
struct printbuf buf = PRINTBUF;
|
2023-01-03 22:32:16 +00:00
|
|
|
bool saw_error;
|
2017-03-17 06:18:50 +00:00
|
|
|
int ret;
|
|
|
|
|
2022-04-04 01:50:25 +00:00
|
|
|
ret = bch2_bkey_invalid(c, bkey_i_to_s_c(&b->key),
|
|
|
|
BKEY_TYPE_btree, WRITE, &buf);
|
2022-04-03 21:50:01 +00:00
|
|
|
|
|
|
|
if (ret)
|
|
|
|
bch2_fs_inconsistent(c, "invalid btree node key before write: %s", buf.buf);
|
|
|
|
printbuf_exit(&buf);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2023-01-03 22:32:16 +00:00
|
|
|
ret = validate_bset_keys(c, b, i, WRITE, false, &saw_error) ?:
|
|
|
|
validate_bset(c, NULL, b, i, b->written, sectors, WRITE, false, &saw_error);
|
2020-11-16 19:16:42 +00:00
|
|
|
if (ret) {
|
2017-03-17 06:18:50 +00:00
|
|
|
bch2_inconsistent_error(c);
|
2020-11-16 19:16:42 +00:00
|
|
|
dump_stack();
|
|
|
|
}
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2021-04-06 19:28:34 +00:00
|
|
|
static void btree_write_submit(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct btree_write_bio *wbio = container_of(work, struct btree_write_bio, work);
|
2023-03-05 03:36:02 +00:00
|
|
|
BKEY_PADDED_ONSTACK(k, BKEY_BTREE_PTR_VAL_U64s_MAX) tmp;
|
2021-07-10 17:44:42 +00:00
|
|
|
|
|
|
|
bkey_copy(&tmp.k, &wbio->key);
|
|
|
|
|
|
|
|
bkey_for_each_ptr(bch2_bkey_ptrs(bkey_i_to_s(&tmp.k)), ptr)
|
|
|
|
ptr->offset += wbio->sector_offset;
|
2021-04-06 19:28:34 +00:00
|
|
|
|
bcachefs: Nocow support
This adds support for nocow mode, where we do writes in-place when
possible. Patch components:
- New boolean filesystem and inode option, nocow: note that when nocow
is enabled, data checksumming and compression are implicitly disabled
- To prevent in-place writes from racing with data moves
(data_update.c) or bucket reuse (i.e. a bucket being reused and
re-allocated while a nocow write is in flight, we have a new locking
mechanism.
Buckets can be locked for either data update or data move, using a
fixed size hash table of two_state_shared locks. We don't have any
chaining, meaning updates and moves to different buckets that hash to
the same lock will wait unnecessarily - we'll want to watch for this
becoming an issue.
- The allocator path also needs to check for in-place writes in flight
to a given bucket before giving it out: thus we add another counter
to bucket_alloc_state so we can track this.
- Fsync now may need to issue cache flushes to block devices instead of
flushing the journal. We add a device bitmask to bch_inode_info,
ei_devs_need_flush, which tracks devices that need to have flushes
issued - note that this will lead to unnecessary flushes when other
codepaths have already issued flushes, we may want to replace this with
a sequence number.
- New nocow write path: look up extents, and if they're writable write
to them - otherwise fall back to the normal COW write path.
XXX: switch to sequence numbers instead of bitmask for devs needing
journal flush
XXX: ei_quota_lock being a mutex means bch2_nocow_write_done() needs to
run in process context - see if we can improve this
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2022-11-02 21:12:00 +00:00
|
|
|
bch2_submit_wbio_replicas(&wbio->wbio, wbio->wbio.c, BCH_DATA_btree,
|
|
|
|
&tmp.k, false);
|
2021-04-06 19:28:34 +00:00
|
|
|
}
|
|
|
|
|
2022-02-27 02:46:41 +00:00
|
|
|
void __bch2_btree_node_write(struct bch_fs *c, struct btree *b, unsigned flags)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
|
|
|
struct btree_write_bio *wbio;
|
|
|
|
struct bset_tree *t;
|
|
|
|
struct bset *i;
|
|
|
|
struct btree_node *bn = NULL;
|
|
|
|
struct btree_node_entry *bne = NULL;
|
2023-09-10 00:10:11 +00:00
|
|
|
struct sort_iter_stack sort_iter;
|
2017-03-17 06:18:50 +00:00
|
|
|
struct nonce nonce;
|
2020-07-25 19:07:37 +00:00
|
|
|
unsigned bytes_to_write, sectors_to_write, bytes, u64s;
|
2017-03-17 06:18:50 +00:00
|
|
|
u64 seq = 0;
|
|
|
|
bool used_mempool;
|
|
|
|
unsigned long old, new;
|
2018-11-01 19:10:01 +00:00
|
|
|
bool validate_before_checksum = false;
|
2022-10-28 21:08:41 +00:00
|
|
|
enum btree_write_type type = flags & BTREE_WRITE_TYPE_MASK;
|
2017-03-17 06:18:50 +00:00
|
|
|
void *data;
|
2022-02-19 05:42:12 +00:00
|
|
|
int ret;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2022-02-27 02:46:41 +00:00
|
|
|
if (flags & BTREE_WRITE_ALREADY_STARTED)
|
2021-07-11 20:41:14 +00:00
|
|
|
goto do_write;
|
2021-07-11 03:03:15 +00:00
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
/*
|
|
|
|
* We may only have a read lock on the btree node - the dirty bit is our
|
|
|
|
* "lock" against racing with other threads that may be trying to start
|
|
|
|
* a write, we do a write iff we clear the dirty bit. Since setting the
|
|
|
|
* dirty bit requires a write lock, we can't race with other threads
|
|
|
|
* redirtying it:
|
|
|
|
*/
|
|
|
|
do {
|
|
|
|
old = new = READ_ONCE(b->flags);
|
|
|
|
|
|
|
|
if (!(old & (1 << BTREE_NODE_dirty)))
|
|
|
|
return;
|
|
|
|
|
2022-02-27 02:46:41 +00:00
|
|
|
if ((flags & BTREE_WRITE_ONLY_IF_NEED) &&
|
|
|
|
!(old & (1 << BTREE_NODE_need_write)))
|
|
|
|
return;
|
|
|
|
|
2022-02-27 14:56:33 +00:00
|
|
|
if (old &
|
|
|
|
((1 << BTREE_NODE_never_write)|
|
|
|
|
(1 << BTREE_NODE_write_blocked)))
|
2017-03-17 06:18:50 +00:00
|
|
|
return;
|
|
|
|
|
2022-02-27 14:56:33 +00:00
|
|
|
if (b->written &&
|
|
|
|
(old & (1 << BTREE_NODE_will_make_reachable)))
|
2020-12-03 21:20:18 +00:00
|
|
|
return;
|
|
|
|
|
2022-02-27 02:46:41 +00:00
|
|
|
if (old & (1 << BTREE_NODE_write_in_flight))
|
|
|
|
return;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2022-11-17 21:03:15 +00:00
|
|
|
if (flags & BTREE_WRITE_ONLY_IF_NEED)
|
|
|
|
type = new & BTREE_WRITE_TYPE_MASK;
|
|
|
|
new &= ~BTREE_WRITE_TYPE_MASK;
|
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
new &= ~(1 << BTREE_NODE_dirty);
|
|
|
|
new &= ~(1 << BTREE_NODE_need_write);
|
|
|
|
new |= (1 << BTREE_NODE_write_in_flight);
|
2021-07-10 17:44:42 +00:00
|
|
|
new |= (1 << BTREE_NODE_write_in_flight_inner);
|
2017-03-17 06:18:50 +00:00
|
|
|
new |= (1 << BTREE_NODE_just_written);
|
|
|
|
new ^= (1 << BTREE_NODE_write_idx);
|
|
|
|
} while (cmpxchg_acquire(&b->flags, old, new) != old);
|
|
|
|
|
2021-07-11 20:41:14 +00:00
|
|
|
if (new & (1U << BTREE_NODE_need_write))
|
|
|
|
return;
|
|
|
|
do_write:
|
2022-10-28 21:08:41 +00:00
|
|
|
BUG_ON((type == BTREE_WRITE_initial) != (b->written == 0));
|
|
|
|
|
2020-11-09 18:01:52 +00:00
|
|
|
atomic_dec(&c->btree_cache.dirty);
|
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
BUG_ON(btree_node_fake(b));
|
|
|
|
BUG_ON((b->will_make_reachable != 0) != !b->written);
|
|
|
|
|
2021-12-14 19:24:41 +00:00
|
|
|
BUG_ON(b->written >= btree_sectors(c));
|
|
|
|
BUG_ON(b->written & (block_sectors(c) - 1));
|
2017-03-17 06:18:50 +00:00
|
|
|
BUG_ON(bset_written(b, btree_bset_last(b)));
|
|
|
|
BUG_ON(le64_to_cpu(b->data->magic) != bset_magic(c));
|
|
|
|
BUG_ON(memcmp(&b->data->format, &b->format, sizeof(b->format)));
|
|
|
|
|
2019-11-29 19:08:51 +00:00
|
|
|
bch2_sort_whiteouts(c, b);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2023-09-10 00:10:11 +00:00
|
|
|
sort_iter_stack_init(&sort_iter, b);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
bytes = !b->written
|
|
|
|
? sizeof(struct btree_node)
|
|
|
|
: sizeof(struct btree_node_entry);
|
|
|
|
|
|
|
|
bytes += b->whiteout_u64s * sizeof(u64);
|
|
|
|
|
|
|
|
for_each_bset(b, t) {
|
|
|
|
i = bset(b, t);
|
|
|
|
|
|
|
|
if (bset_written(b, i))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
bytes += le16_to_cpu(i->u64s) * sizeof(u64);
|
2023-09-10 00:10:11 +00:00
|
|
|
sort_iter_add(&sort_iter.iter,
|
2017-03-17 06:18:50 +00:00
|
|
|
btree_bkey_first(b, t),
|
|
|
|
btree_bkey_last(b, t));
|
|
|
|
seq = max(seq, le64_to_cpu(i->journal_seq));
|
|
|
|
}
|
|
|
|
|
2020-12-03 21:20:18 +00:00
|
|
|
BUG_ON(b->written && !seq);
|
|
|
|
|
2020-11-11 17:42:54 +00:00
|
|
|
/* bch2_varint_decode may read up to 7 bytes past the end of the buffer: */
|
|
|
|
bytes += 8;
|
|
|
|
|
2021-05-08 02:29:02 +00:00
|
|
|
/* buffer must be a multiple of the block size */
|
|
|
|
bytes = round_up(bytes, block_bytes(c));
|
|
|
|
|
2020-07-25 19:07:37 +00:00
|
|
|
data = btree_bounce_alloc(c, bytes, &used_mempool);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
if (!b->written) {
|
|
|
|
bn = data;
|
|
|
|
*bn = *b->data;
|
|
|
|
i = &bn->keys;
|
|
|
|
} else {
|
|
|
|
bne = data;
|
|
|
|
bne->keys = b->data->keys;
|
|
|
|
i = &bne->keys;
|
|
|
|
}
|
|
|
|
|
|
|
|
i->journal_seq = cpu_to_le64(seq);
|
|
|
|
i->u64s = 0;
|
|
|
|
|
2023-09-10 00:10:11 +00:00
|
|
|
sort_iter_add(&sort_iter.iter,
|
2024-01-16 18:29:59 +00:00
|
|
|
unwritten_whiteouts_start(b),
|
|
|
|
unwritten_whiteouts_end(b));
|
2021-02-20 05:00:23 +00:00
|
|
|
SET_BSET_SEPARATE_WHITEOUTS(i, false);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
b->whiteout_u64s = 0;
|
|
|
|
|
2023-09-10 00:10:11 +00:00
|
|
|
u64s = bch2_sort_keys(i->start, &sort_iter.iter, false);
|
2017-03-17 06:18:50 +00:00
|
|
|
le16_add_cpu(&i->u64s, u64s);
|
|
|
|
|
2022-10-01 04:34:02 +00:00
|
|
|
BUG_ON(!b->written && i->u64s != b->data->keys.u64s);
|
|
|
|
|
2019-11-10 04:50:52 +00:00
|
|
|
set_needs_whiteout(i, false);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
/* do we have data to write? */
|
|
|
|
if (b->written && !i->u64s)
|
|
|
|
goto nowrite;
|
|
|
|
|
|
|
|
bytes_to_write = vstruct_end(i) - data;
|
|
|
|
sectors_to_write = round_up(bytes_to_write, block_bytes(c)) >> 9;
|
|
|
|
|
2022-10-01 04:34:02 +00:00
|
|
|
if (!b->written &&
|
|
|
|
b->key.k.type == KEY_TYPE_btree_ptr_v2)
|
|
|
|
BUG_ON(btree_ptr_sectors_written(&b->key) != sectors_to_write);
|
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
memset(data + bytes_to_write, 0,
|
|
|
|
(sectors_to_write << 9) - bytes_to_write);
|
|
|
|
|
2021-12-14 19:24:41 +00:00
|
|
|
BUG_ON(b->written + sectors_to_write > btree_sectors(c));
|
2017-03-17 06:18:50 +00:00
|
|
|
BUG_ON(BSET_BIG_ENDIAN(i) != CPU_BIG_ENDIAN);
|
|
|
|
BUG_ON(i->seq != b->data->keys.seq);
|
|
|
|
|
2023-06-29 00:27:07 +00:00
|
|
|
i->version = cpu_to_le16(c->sb.version);
|
2021-07-16 16:57:27 +00:00
|
|
|
SET_BSET_OFFSET(i, b->written);
|
2017-03-17 06:18:50 +00:00
|
|
|
SET_BSET_CSUM_TYPE(i, bch2_meta_checksum_type(c));
|
|
|
|
|
2018-11-01 19:10:01 +00:00
|
|
|
if (bch2_csum_type_is_encryption(BSET_CSUM_TYPE(i)))
|
|
|
|
validate_before_checksum = true;
|
|
|
|
|
|
|
|
/* validate_bset will be modifying: */
|
bcachefs: Start using bpos.snapshot field
This patch starts treating the bpos.snapshot field like part of the key
in the btree code:
* bpos_successor() and bpos_predecessor() now include the snapshot field
* Keys in btrees that will be using snapshots (extents, inodes, dirents
and xattrs) now always have their snapshot field set to U32_MAX
The btree iterator code gets a new flag, BTREE_ITER_ALL_SNAPSHOTS, that
determines whether we're iterating over keys in all snapshots or not -
internally, this controlls whether bkey_(successor|predecessor)
increment/decrement the snapshot field, or only the higher bits of the
key.
We add a new member to struct btree_iter, iter->snapshot: when
BTREE_ITER_ALL_SNAPSHOTS is not set, iter->pos.snapshot should always
equal iter->snapshot, which will be 0 for btrees that don't use
snapshots, and alsways U32_MAX for btrees that will use snapshots
(until we enable snapshot creation).
This patch also introduces a new metadata version number, and compat
code for reading from/writing to older versions - this isn't a forced
upgrade (yet).
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2021-03-24 22:02:16 +00:00
|
|
|
if (le16_to_cpu(i->version) < bcachefs_metadata_version_current)
|
2018-11-01 19:10:01 +00:00
|
|
|
validate_before_checksum = true;
|
|
|
|
|
2017-03-17 06:18:50 +00:00
|
|
|
/* if we're going to be encrypting, check metadata validity first: */
|
2018-11-01 19:10:01 +00:00
|
|
|
if (validate_before_checksum &&
|
2017-03-17 06:18:50 +00:00
|
|
|
validate_bset_for_write(c, b, i, sectors_to_write))
|
|
|
|
goto err;
|
|
|
|
|
2022-02-19 05:42:12 +00:00
|
|
|
ret = bset_encrypt(c, i, b->written << 9);
|
|
|
|
if (bch2_fs_fatal_err_on(ret, c,
|
2024-03-18 01:51:19 +00:00
|
|
|
"encrypting btree node: %s", bch2_err_str(ret)))
|
2022-02-19 05:42:12 +00:00
|
|
|
goto err;
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
nonce = btree_nonce(i, b->written << 9);
|
|
|
|
|
|
|
|
if (bn)
|
|
|
|
bn->csum = csum_vstruct(c, BSET_CSUM_TYPE(i), nonce, bn);
|
|
|
|
else
|
|
|
|
bne->csum = csum_vstruct(c, BSET_CSUM_TYPE(i), nonce, bne);
|
|
|
|
|
|
|
|
/* if we're not encrypting, check metadata after checksumming: */
|
2018-11-01 19:10:01 +00:00
|
|
|
if (!validate_before_checksum &&
|
2017-03-17 06:18:50 +00:00
|
|
|
validate_bset_for_write(c, b, i, sectors_to_write))
|
|
|
|
goto err;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We handle btree write errors by immediately halting the journal -
|
|
|
|
* after we've done that, we can't issue any subsequent btree writes
|
|
|
|
* because they might have pointers to new nodes that failed to write.
|
|
|
|
*
|
|
|
|
* Furthermore, there's no point in doing any more btree writes because
|
|
|
|
* with the journal stopped, we're never going to update the journal to
|
|
|
|
* reflect that those writes were done and the data flushed from the
|
|
|
|
* journal:
|
|
|
|
*
|
2020-05-02 20:21:35 +00:00
|
|
|
* Also on journal error, the pending write may have updates that were
|
|
|
|
* never journalled (interior nodes, see btree_update_nodes_written()) -
|
|
|
|
* it's critical that we don't do the write in that case otherwise we
|
|
|
|
* will have updates visible that weren't in the journal:
|
|
|
|
*
|
2017-03-17 06:18:50 +00:00
|
|
|
* Make sure to update b->written so bch2_btree_init_next() doesn't
|
|
|
|
* break:
|
|
|
|
*/
|
|
|
|
if (bch2_journal_error(&c->journal) ||
|
|
|
|
c->opts.nochanges)
|
|
|
|
goto err;
|
|
|
|
|
2022-08-27 16:48:36 +00:00
|
|
|
trace_and_count(c, btree_node_write, b, bytes_to_write, sectors_to_write);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2018-11-04 00:04:54 +00:00
|
|
|
wbio = container_of(bio_alloc_bioset(NULL,
|
|
|
|
buf_pages(data, sectors_to_write << 9),
|
2019-10-21 23:38:08 +00:00
|
|
|
REQ_OP_WRITE|REQ_META,
|
2023-05-28 22:02:38 +00:00
|
|
|
GFP_NOFS,
|
2017-03-17 06:18:50 +00:00
|
|
|
&c->btree_bio),
|
|
|
|
struct btree_write_bio, wbio.bio);
|
|
|
|
wbio_init(&wbio->wbio.bio);
|
|
|
|
wbio->data = data;
|
2021-07-10 17:44:42 +00:00
|
|
|
wbio->data_bytes = bytes;
|
|
|
|
wbio->sector_offset = b->written;
|
2021-04-06 19:28:34 +00:00
|
|
|
wbio->wbio.c = c;
|
2017-03-17 06:18:50 +00:00
|
|
|
wbio->wbio.used_mempool = used_mempool;
|
2021-07-10 17:44:42 +00:00
|
|
|
wbio->wbio.first_btree_write = !b->written;
|
2017-03-17 06:18:50 +00:00
|
|
|
wbio->wbio.bio.bi_end_io = btree_node_write_endio;
|
|
|
|
wbio->wbio.bio.bi_private = b;
|
|
|
|
|
2019-07-03 23:27:42 +00:00
|
|
|
bch2_bio_map(&wbio->wbio.bio, data, sectors_to_write << 9);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
2021-04-06 19:28:34 +00:00
|
|
|
bkey_copy(&wbio->key, &b->key);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
b->written += sectors_to_write;
|
|
|
|
|
2021-07-10 17:44:42 +00:00
|
|
|
if (wbio->key.k.type == KEY_TYPE_btree_ptr_v2)
|
|
|
|
bkey_i_to_btree_ptr_v2(&wbio->key)->v.sectors_written =
|
|
|
|
cpu_to_le16(b->written);
|
|
|
|
|
2022-10-28 21:08:41 +00:00
|
|
|
atomic64_inc(&c->btree_write_stats[type].nr);
|
|
|
|
atomic64_add(bytes_to_write, &c->btree_write_stats[type].bytes);
|
2021-04-01 01:07:37 +00:00
|
|
|
|
2021-04-06 19:28:34 +00:00
|
|
|
INIT_WORK(&wbio->work, btree_write_submit);
|
2021-05-22 21:37:25 +00:00
|
|
|
queue_work(c->io_complete_wq, &wbio->work);
|
2017-03-17 06:18:50 +00:00
|
|
|
return;
|
|
|
|
err:
|
|
|
|
set_btree_node_noevict(b);
|
|
|
|
b->written += sectors_to_write;
|
|
|
|
nowrite:
|
2020-07-25 19:07:37 +00:00
|
|
|
btree_bounce_free(c, bytes, used_mempool, data);
|
2022-02-27 02:35:16 +00:00
|
|
|
__btree_node_write_done(c, b);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Work that must be done with write lock held:
|
|
|
|
*/
|
|
|
|
bool bch2_btree_post_write_cleanup(struct bch_fs *c, struct btree *b)
|
|
|
|
{
|
|
|
|
bool invalidated_iter = false;
|
|
|
|
struct btree_node_entry *bne;
|
|
|
|
struct bset_tree *t;
|
|
|
|
|
|
|
|
if (!btree_node_just_written(b))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
BUG_ON(b->whiteout_u64s);
|
|
|
|
|
|
|
|
clear_btree_node_just_written(b);
|
|
|
|
|
|
|
|
/*
|
2018-08-06 02:23:44 +00:00
|
|
|
* Note: immediately after write, bset_written() doesn't work - the
|
|
|
|
* amount of data we had to write after compaction might have been
|
|
|
|
* smaller than the offset of the last bset.
|
2017-03-17 06:18:50 +00:00
|
|
|
*
|
|
|
|
* However, we know that all bsets have been written here, as long as
|
|
|
|
* we're still holding the write lock:
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* XXX: decide if we really want to unconditionally sort down to a
|
|
|
|
* single bset:
|
|
|
|
*/
|
|
|
|
if (b->nsets > 1) {
|
2021-04-06 19:33:19 +00:00
|
|
|
btree_node_sort(c, b, 0, b->nsets, true);
|
2017-03-17 06:18:50 +00:00
|
|
|
invalidated_iter = true;
|
|
|
|
} else {
|
2019-12-13 18:08:37 +00:00
|
|
|
invalidated_iter = bch2_drop_whiteouts(b, COMPACT_ALL);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
for_each_bset(b, t)
|
2019-11-10 04:50:52 +00:00
|
|
|
set_needs_whiteout(bset(b, t), true);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
bch2_btree_verify(c, b);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If later we don't unconditionally sort down to a single bset, we have
|
|
|
|
* to ensure this is still true:
|
|
|
|
*/
|
|
|
|
BUG_ON((void *) btree_bkey_last(b, bset_tree_last(b)) > write_block(b));
|
|
|
|
|
|
|
|
bne = want_new_bset(c, b);
|
|
|
|
if (bne)
|
2024-01-16 18:29:59 +00:00
|
|
|
bch2_bset_init_next(b, bne);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
bch2_btree_build_aux_trees(b);
|
|
|
|
|
|
|
|
return invalidated_iter;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Use this one if the node is intent locked:
|
|
|
|
*/
|
|
|
|
void bch2_btree_node_write(struct bch_fs *c, struct btree *b,
|
2022-02-27 02:46:41 +00:00
|
|
|
enum six_lock_type lock_type_held,
|
|
|
|
unsigned flags)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
|
|
|
if (lock_type_held == SIX_LOCK_intent ||
|
2021-04-06 19:33:19 +00:00
|
|
|
(lock_type_held == SIX_LOCK_read &&
|
|
|
|
six_lock_tryupgrade(&b->c.lock))) {
|
2022-02-27 02:46:41 +00:00
|
|
|
__bch2_btree_node_write(c, b, flags);
|
2017-03-17 06:18:50 +00:00
|
|
|
|
|
|
|
/* don't cycle lock unnecessarily: */
|
|
|
|
if (btree_node_just_written(b) &&
|
2020-06-06 16:28:01 +00:00
|
|
|
six_trylock_write(&b->c.lock)) {
|
2017-03-17 06:18:50 +00:00
|
|
|
bch2_btree_post_write_cleanup(c, b);
|
2020-06-06 16:28:01 +00:00
|
|
|
six_unlock_write(&b->c.lock);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (lock_type_held == SIX_LOCK_read)
|
2020-06-06 16:28:01 +00:00
|
|
|
six_lock_downgrade(&b->c.lock);
|
2017-03-17 06:18:50 +00:00
|
|
|
} else {
|
2022-02-27 02:46:41 +00:00
|
|
|
__bch2_btree_node_write(c, b, flags);
|
2021-04-06 19:33:19 +00:00
|
|
|
if (lock_type_held == SIX_LOCK_write &&
|
|
|
|
btree_node_just_written(b))
|
|
|
|
bch2_btree_post_write_cleanup(c, b);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2022-04-17 21:30:49 +00:00
|
|
|
static bool __bch2_btree_flush_all(struct bch_fs *c, unsigned flag)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
|
|
|
struct bucket_table *tbl;
|
|
|
|
struct rhash_head *pos;
|
|
|
|
struct btree *b;
|
|
|
|
unsigned i;
|
2022-04-17 21:30:49 +00:00
|
|
|
bool ret = false;
|
2017-03-17 06:18:50 +00:00
|
|
|
restart:
|
|
|
|
rcu_read_lock();
|
|
|
|
for_each_cached_btree(b, c, tbl, i, pos)
|
|
|
|
if (test_bit(flag, &b->flags)) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
wait_on_bit_io(&b->flags, flag, TASK_UNINTERRUPTIBLE);
|
2022-04-17 21:30:49 +00:00
|
|
|
ret = true;
|
2017-03-17 06:18:50 +00:00
|
|
|
goto restart;
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
2022-04-17 21:30:49 +00:00
|
|
|
|
|
|
|
return ret;
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
2022-04-17 21:30:49 +00:00
|
|
|
bool bch2_btree_flush_all_reads(struct bch_fs *c)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
2022-04-17 21:30:49 +00:00
|
|
|
return __bch2_btree_flush_all(c, BTREE_NODE_read_in_flight);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
|
|
|
|
2022-04-17 21:30:49 +00:00
|
|
|
bool bch2_btree_flush_all_writes(struct bch_fs *c)
|
2017-03-17 06:18:50 +00:00
|
|
|
{
|
2022-04-17 21:30:49 +00:00
|
|
|
return __bch2_btree_flush_all(c, BTREE_NODE_write_in_flight);
|
2017-03-17 06:18:50 +00:00
|
|
|
}
|
2022-10-28 21:08:41 +00:00
|
|
|
|
2023-07-07 02:47:42 +00:00
|
|
|
static const char * const bch2_btree_write_types[] = {
|
2022-10-28 21:08:41 +00:00
|
|
|
#define x(t, n) [n] = #t,
|
|
|
|
BCH_BTREE_WRITE_TYPES()
|
|
|
|
NULL
|
|
|
|
};
|
|
|
|
|
|
|
|
void bch2_btree_write_stats_to_text(struct printbuf *out, struct bch_fs *c)
|
|
|
|
{
|
|
|
|
printbuf_tabstop_push(out, 20);
|
|
|
|
printbuf_tabstop_push(out, 10);
|
|
|
|
|
|
|
|
prt_tab(out);
|
|
|
|
prt_str(out, "nr");
|
|
|
|
prt_tab(out);
|
|
|
|
prt_str(out, "size");
|
|
|
|
prt_newline(out);
|
|
|
|
|
|
|
|
for (unsigned i = 0; i < BTREE_WRITE_TYPE_NR; i++) {
|
|
|
|
u64 nr = atomic64_read(&c->btree_write_stats[i].nr);
|
|
|
|
u64 bytes = atomic64_read(&c->btree_write_stats[i].bytes);
|
|
|
|
|
|
|
|
prt_printf(out, "%s:", bch2_btree_write_types[i]);
|
|
|
|
prt_tab(out);
|
|
|
|
prt_u64(out, nr);
|
|
|
|
prt_tab(out);
|
|
|
|
prt_human_readable_u64(out, nr ? div64_u64(bytes, nr) : 0);
|
|
|
|
prt_newline(out);
|
|
|
|
}
|
|
|
|
}
|