#include "btree_update.h"
#include "btree_update_interior.h"
#include "btree_gc.h"
+#include "btree_write_buffer.h"
#include "buckets.h"
#include "buckets_waiting_for_journal.h"
#include "clock.h"
#include "error.h"
#include "lru.h"
#include "recovery.h"
+#include "trace.h"
#include "varint.h"
#include <linux/kthread.h>
#include <linux/rcupdate.h>
#include <linux/sched/task.h>
#include <linux/sort.h>
-#include <trace/events/bcachefs.h>
/* Persistent alloc info: */
struct bkey_s_c_alloc_v4 a = bkey_s_c_to_alloc_v4(k);
int rw = flags & WRITE;
- if (alloc_v4_u64s(a.v) != bkey_val_u64s(k.k)) {
- prt_printf(err, "bad val size (%lu != %u)",
- bkey_val_u64s(k.k), alloc_v4_u64s(a.v));
+ if (alloc_v4_u64s(a.v) > bkey_val_u64s(k.k)) {
+ prt_printf(err, "bad val size (%u > %lu)",
+ alloc_v4_u64s(a.v), bkey_val_u64s(k.k));
return -BCH_ERR_invalid_bkey;
}
prt_newline(out);
prt_printf(out, "io_time[WRITE] %llu", a->io_time[WRITE]);
prt_newline(out);
+ prt_printf(out, "fragmentation %llu", a->fragmentation_lru);
+ prt_newline(out);
+ prt_printf(out, "bp_start %llu", BCH_ALLOC_V4_BACKPOINTERS_START(a));
+ prt_newline(out);
- if (k.k->type == KEY_TYPE_alloc_v4) {
+ if (BCH_ALLOC_V4_NR_BACKPOINTERS(a)) {
struct bkey_s_c_alloc_v4 a_raw = bkey_s_c_to_alloc_v4(k);
const struct bch_backpointer *bps = alloc_v4_backpointers_c(a_raw.v);
- prt_printf(out, "bp_start %llu", BCH_ALLOC_V4_BACKPOINTERS_START(a_raw.v));
- prt_newline(out);
-
prt_printf(out, "backpointers: %llu", BCH_ALLOC_V4_NR_BACKPOINTERS(a_raw.v));
printbuf_indent_add(out, 2);
if (src < dst)
memset(src, 0, dst - src);
+
+ SET_BCH_ALLOC_V4_NR_BACKPOINTERS(out, 0);
} else {
struct bkey_alloc_unpacked u = bch2_alloc_unpack(k);
__bch2_alloc_to_v4_mut(struct btree_trans *trans, struct bkey_s_c k)
{
struct bkey_i_alloc_v4 *ret;
+
+ ret = bch2_trans_kmalloc(trans, max(bkey_bytes(k.k), sizeof(struct bkey_i_alloc_v4)));
+ if (IS_ERR(ret))
+ return ret;
+
if (k.k->type == KEY_TYPE_alloc_v4) {
- struct bkey_s_c_alloc_v4 a = bkey_s_c_to_alloc_v4(k);
- unsigned bytes = sizeof(struct bkey_i_alloc_v4) +
- BCH_ALLOC_V4_NR_BACKPOINTERS(a.v) *
- sizeof(struct bch_backpointer);
void *src, *dst;
- /*
- * Reserve space for one more backpointer here:
- * Not sketchy at doing it this way, nope...
- */
- ret = bch2_trans_kmalloc(trans, bytes + sizeof(struct bch_backpointer));
- if (IS_ERR(ret))
- return ret;
-
bkey_reassemble(&ret->k_i, k);
src = alloc_v4_backpointers(&ret->v);
SET_BCH_ALLOC_V4_BACKPOINTERS_START(&ret->v, BCH_ALLOC_V4_U64s);
dst = alloc_v4_backpointers(&ret->v);
- memmove(dst, src, BCH_ALLOC_V4_NR_BACKPOINTERS(&ret->v) *
- sizeof(struct bch_backpointer));
if (src < dst)
memset(src, 0, dst - src);
+
+ SET_BCH_ALLOC_V4_NR_BACKPOINTERS(&ret->v, 0);
set_alloc_v4_u64s(ret);
} else {
- ret = bch2_trans_kmalloc(trans, sizeof(struct bkey_i_alloc_v4) +
- sizeof(struct bch_backpointer));
- if (IS_ERR(ret))
- return ret;
-
bkey_alloc_v4_init(&ret->k_i);
ret->k.p = k.k->p;
bch2_alloc_to_v4(k, &ret->v);
static inline struct bkey_i_alloc_v4 *bch2_alloc_to_v4_mut_inlined(struct btree_trans *trans, struct bkey_s_c k)
{
+ struct bkey_s_c_alloc_v4 a;
+
if (likely(k.k->type == KEY_TYPE_alloc_v4) &&
- BCH_ALLOC_V4_BACKPOINTERS_START(bkey_s_c_to_alloc_v4(k).v) == BCH_ALLOC_V4_U64s) {
- /*
- * Reserve space for one more backpointer here:
- * Not sketchy at doing it this way, nope...
- */
- struct bkey_i_alloc_v4 *ret =
- bch2_trans_kmalloc_nomemzero(trans, bkey_bytes(k.k) + sizeof(struct bch_backpointer));
- if (!IS_ERR(ret))
- bkey_reassemble(&ret->k_i, k);
- return ret;
- }
+ ((a = bkey_s_c_to_alloc_v4(k), true) &&
+ BCH_ALLOC_V4_NR_BACKPOINTERS(a.v) == 0))
+ return bch2_bkey_make_mut_noupdate_typed(trans, k, alloc_v4);
return __bch2_alloc_to_v4_mut(trans, k);
}
struct bkey_i_alloc_v4 *a;
int ret;
- bch2_trans_iter_init(trans, iter, BTREE_ID_alloc, pos,
+ k = bch2_bkey_get_iter(trans, iter, BTREE_ID_alloc, pos,
BTREE_ITER_WITH_UPDATES|
BTREE_ITER_CACHED|
BTREE_ITER_INTENT);
- k = bch2_btree_iter_peek_slot(iter);
ret = bkey_err(k);
if (unlikely(ret))
- goto err;
+ return ERR_PTR(ret);
a = bch2_alloc_to_v4_mut_inlined(trans, k);
ret = PTR_ERR_OR_ZERO(a);
ret = commit_do(&trans, NULL, NULL,
BTREE_INSERT_NOFAIL|
BTREE_INSERT_LAZY_RW,
- __bch2_btree_insert(&trans, BTREE_ID_bucket_gens, &g.k_i));
+ __bch2_btree_insert(&trans, BTREE_ID_bucket_gens, &g.k_i, 0));
if (ret)
break;
have_bucket_gens_key = false;
ret = commit_do(&trans, NULL, NULL,
BTREE_INSERT_NOFAIL|
BTREE_INSERT_LAZY_RW,
- __bch2_btree_insert(&trans, BTREE_ID_bucket_gens, &g.k_i));
+ __bch2_btree_insert(&trans, BTREE_ID_bucket_gens, &g.k_i, 0));
bch2_trans_exit(&trans);
return 0;
}
- bch2_trans_iter_init(trans, &iter, btree,
+ old = bch2_bkey_get_iter(trans, &iter, btree,
bkey_start_pos(&k->k),
BTREE_ITER_INTENT);
- old = bch2_btree_iter_peek_slot(&iter);
ret = bkey_err(old);
if (ret)
- goto err;
+ return ret;
if (ca->mi.freespace_initialized &&
test_bit(BCH_FS_CHECK_ALLOC_DONE, &c->flags) &&
bch2_trans_inconsistent_on(old.k->type != old_type, trans,
- "incorrect key when %s %s btree (got %s should be %s)\n"
+ "incorrect key when %s %s:%llu:%llu:0 (got %s should be %s)\n"
" for %s",
set ? "setting" : "clearing",
bch2_btree_ids[btree],
+ iter.pos.inode,
+ iter.pos.offset,
bch2_bkey_types[old.k->type],
bch2_bkey_types[old_type],
(bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf))) {
if (ret)
return ret;
- bch2_trans_iter_init(trans, &iter, BTREE_ID_bucket_gens, pos,
- BTREE_ITER_INTENT|
- BTREE_ITER_WITH_UPDATES);
- k = bch2_btree_iter_peek_slot(&iter);
+ k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_bucket_gens, pos,
+ BTREE_ITER_INTENT|
+ BTREE_ITER_WITH_UPDATES);
ret = bkey_err(k);
if (ret)
- goto err;
+ return ret;
if (k.k->type != KEY_TYPE_bucket_gens) {
bkey_bucket_gens_init(&g->k_i);
g->v.gens[offset] = gen;
ret = bch2_trans_update(trans, &iter, &g->k_i, 0);
-err:
bch2_trans_iter_exit(trans, &iter);
return ret;
}
!new_a->io_time[READ])
new_a->io_time[READ] = max_t(u64, 1, atomic64_read(&c->io_clock[READ].now));
- old_lru = alloc_lru_idx(*old_a);
- new_lru = alloc_lru_idx(*new_a);
+ old_lru = alloc_lru_idx_read(*old_a);
+ new_lru = alloc_lru_idx_read(*new_a);
if (old_lru != new_lru) {
ret = bch2_lru_change(trans, new->k.p.inode,
return ret;
}
+ new_a->fragmentation_lru = alloc_lru_idx_fragmentation(*new_a,
+ bch_dev_bkey_exists(c, new->k.p.inode));
+
+ if (old_a->fragmentation_lru != new_a->fragmentation_lru) {
+ ret = bch2_lru_change(trans,
+ BCH_LRU_FRAGMENTATION_START,
+ bucket_to_u64(new->k.p),
+ old_a->fragmentation_lru, new_a->fragmentation_lru);
+ if (ret)
+ return ret;
+ }
+
if (old_a->gen != new_a->gen) {
ret = bch2_bucket_gen_update(trans, new->k.p, new_a->gen);
if (ret)
struct bpos next;
bch2_trans_copy_iter(&iter2, iter);
- k = bch2_btree_iter_peek_upto(&iter2,
- bkey_min(bkey_min(end,
- iter->path->l[0].b->key.k.p),
- POS(iter->pos.inode, iter->pos.offset + U32_MAX - 1)));
+
+ if (!bpos_eq(iter->path->l[0].b->key.k.p, SPOS_MAX))
+ end = bkey_min(end, bpos_nosnap_successor(iter->path->l[0].b->key.k.p));
+
+ end = bkey_min(end, POS(iter->pos.inode, iter->pos.offset + U32_MAX - 1));
+
+ /*
+ * btree node min/max is a closed interval, upto takes a half
+ * open interval:
+ */
+ k = bch2_btree_iter_peek_upto(&iter2, end);
next = iter2.pos;
bch2_trans_iter_exit(iter->trans, &iter2);
iter = bucket->inode;
ca = __bch2_next_dev(c, &iter, NULL);
if (ca)
- bucket->offset = ca->mi.first_bucket;
+ *bucket = POS(ca->dev_idx, ca->mi.first_bucket);
rcu_read_unlock();
return ca != NULL;
return ret;
}
-static int bch2_check_discard_freespace_key(struct btree_trans *trans,
- struct btree_iter *iter)
+static int __bch2_check_discard_freespace_key(struct btree_trans *trans,
+ struct btree_iter *iter)
{
struct bch_fs *c = trans->c;
struct btree_iter alloc_iter;
pos.offset &= ~(~0ULL << 56);
genbits = iter->pos.offset & (~0ULL << 56);
- bch2_trans_iter_init(trans, &alloc_iter, BTREE_ID_alloc, pos, 0);
+ alloc_k = bch2_bkey_get_iter(trans, &alloc_iter, BTREE_ID_alloc, pos, 0);
+ ret = bkey_err(alloc_k);
+ if (ret)
+ return ret;
if (fsck_err_on(!bch2_dev_bucket_exists(c, pos), c,
"entry in %s btree for nonexistant dev:bucket %llu:%llu",
bch2_btree_ids[iter->btree_id], pos.inode, pos.offset))
goto delete;
- alloc_k = bch2_btree_iter_peek_slot(&alloc_iter);
- ret = bkey_err(alloc_k);
- if (ret)
- goto err;
-
a = bch2_alloc_to_v4(alloc_k, &a_convert);
if (fsck_err_on(a->data_type != state ||
(state == BCH_DATA_free &&
genbits != alloc_freespace_genbits(*a)), c,
- "%s\n incorrectly set in %s index (free %u, genbits %llu should be %llu)",
+ "%s\n incorrectly set at %s:%llu:%llu:0 (free %u, genbits %llu should be %llu)",
(bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf),
bch2_btree_ids[iter->btree_id],
+ iter->pos.inode,
+ iter->pos.offset,
a->data_type == state,
genbits >> 56, alloc_freespace_genbits(*a) >> 56))
goto delete;
out:
-err:
fsck_err:
+ set_btree_iter_dontneed(&alloc_iter);
bch2_trans_iter_exit(trans, &alloc_iter);
printbuf_exit(&buf);
return ret;
delete:
- ret = bch2_btree_delete_extent_at(trans, iter,
- iter->btree_id == BTREE_ID_freespace ? 1 : 0, 0);
+ ret = bch2_btree_delete_extent_at(trans, iter,
+ iter->btree_id == BTREE_ID_freespace ? 1 : 0, 0) ?:
+ bch2_trans_commit(trans, NULL, NULL,
+ BTREE_INSERT_NOFAIL|BTREE_INSERT_LAZY_RW);
goto out;
}
+static int bch2_check_discard_freespace_key(struct btree_trans *trans,
+ struct btree_iter *iter,
+ struct bpos end)
+{
+ if (!btree_node_type_is_extents(iter->btree_id)) {
+ return __bch2_check_discard_freespace_key(trans, iter);
+ } else {
+ int ret;
+
+ while (!bkey_eq(iter->pos, end) &&
+ !(ret = btree_trans_too_many_iters(trans) ?:
+ __bch2_check_discard_freespace_key(trans, iter)))
+ bch2_btree_iter_set_pos(iter, bpos_nosnap_successor(iter->pos));
+
+ return ret;
+ }
+}
+
/*
* We've already checked that generation numbers in the bucket_gens btree are
* valid for buckets that exist; this just checks for keys for nonexistent
u64 start = bucket_gens_pos_to_alloc(k.k->p, 0).offset;
u64 end = bucket_gens_pos_to_alloc(bpos_nosnap_successor(k.k->p), 0).offset;
u64 b;
- bool need_update = false;
+ bool need_update = false, dev_exists;
struct printbuf buf = PRINTBUF;
int ret = 0;
BUG_ON(k.k->type != KEY_TYPE_bucket_gens);
bkey_reassemble(&g.k_i, k);
- if (fsck_err_on(!bch2_dev_exists2(c, k.k->p.inode), c,
- "bucket_gens key for invalid device:\n %s",
- (bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
- ret = bch2_btree_delete_at(trans, iter, 0);
+ /* if no bch_dev, skip out whether we repair or not */
+ dev_exists = bch2_dev_exists2(c, k.k->p.inode);
+ if (!dev_exists) {
+ if (fsck_err_on(!dev_exists, c,
+ "bucket_gens key for invalid device:\n %s",
+ (bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
+ ret = bch2_btree_delete_at(trans, iter, 0);
+ }
goto out;
}
if (ret < 0)
goto err;
- ret = for_each_btree_key_commit(&trans, iter,
+ ret = for_each_btree_key2(&trans, iter,
BTREE_ID_need_discard, POS_MIN,
BTREE_ITER_PREFETCH, k,
- NULL, NULL, BTREE_INSERT_NOFAIL|BTREE_INSERT_LAZY_RW,
- bch2_check_discard_freespace_key(&trans, &iter)) ?:
- for_each_btree_key_commit(&trans, iter,
+ bch2_check_discard_freespace_key(&trans, &iter, k.k->p)) ?:
+ for_each_btree_key2(&trans, iter,
BTREE_ID_freespace, POS_MIN,
BTREE_ITER_PREFETCH, k,
- NULL, NULL, BTREE_INSERT_NOFAIL|BTREE_INSERT_LAZY_RW,
- bch2_check_discard_freespace_key(&trans, &iter)) ?:
+ bch2_check_discard_freespace_key(&trans, &iter, k.k->p)) ?:
for_each_btree_key_commit(&trans, iter,
BTREE_ID_bucket_gens, POS_MIN,
BTREE_ITER_PREFETCH, k,
struct btree_iter lru_iter;
struct bch_alloc_v4 a_convert;
const struct bch_alloc_v4 *a;
- struct bkey_s_c alloc_k, k;
+ struct bkey_s_c alloc_k, lru_k;
struct printbuf buf = PRINTBUF;
int ret;
if (a->data_type != BCH_DATA_cached)
return 0;
- bch2_trans_iter_init(trans, &lru_iter, BTREE_ID_lru,
+ lru_k = bch2_bkey_get_iter(trans, &lru_iter, BTREE_ID_lru,
lru_pos(alloc_k.k->p.inode,
bucket_to_u64(alloc_k.k->p),
a->io_time[READ]), 0);
- k = bch2_btree_iter_peek_slot(&lru_iter);
- ret = bkey_err(k);
+ ret = bkey_err(lru_k);
if (ret)
- goto err;
+ return ret;
if (fsck_err_on(!a->io_time[READ], c,
"cached bucket with read_time 0\n"
" %s",
(printbuf_reset(&buf),
bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf)) ||
- fsck_err_on(k.k->type != KEY_TYPE_set, c,
+ fsck_err_on(lru_k.k->type != KEY_TYPE_set, c,
"missing lru entry\n"
" %s",
(printbuf_reset(&buf),
goto out;
}
- bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc,
- need_discard_iter->pos,
- BTREE_ITER_CACHED);
- k = bch2_btree_iter_peek_slot(&iter);
+ k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_alloc,
+ need_discard_iter->pos,
+ BTREE_ITER_CACHED);
ret = bkey_err(k);
if (ret)
goto out;
void bch2_do_discards(struct bch_fs *c)
{
if (bch2_write_ref_tryget(c, BCH_WRITE_REF_discard) &&
- !queue_work(system_long_wq, &c->discard_work))
+ !queue_work(c->write_ref_wq, &c->discard_work))
bch2_write_ref_put(c, BCH_WRITE_REF_discard);
}
goto err;
}
+ if (bch2_bucket_is_open_safe(c, bucket.inode, bucket.offset))
+ return 0;
+
a = bch2_trans_start_alloc_update(trans, &alloc_iter, bucket);
ret = PTR_ERR_OR_ZERO(a);
if (ret)
goto out;
- if (lru_pos_time(lru_iter->pos) != alloc_lru_idx(a->v)) {
- prt_str(&buf, "alloc key does not point back to lru entry when invalidating bucket:");
- goto err;
- }
+ /* We expect harmless races here due to the btree write buffer: */
+ if (lru_pos_time(lru_iter->pos) != alloc_lru_idx_read(a->v))
+ goto out;
- if (a->v.data_type != BCH_DATA_cached) {
- prt_str(&buf, "lru entry points to non cached bucket:");
- goto err;
- }
+ BUG_ON(a->v.data_type != BCH_DATA_cached);
if (!a->v.cached_sectors)
bch_err(c, "invalidating empty bucket, confused");
bch2_trans_init(&trans, c, 0, 0);
+ ret = bch2_btree_write_buffer_flush(&trans);
+ if (ret)
+ goto err;
+
for_each_member_device(ca, c, i) {
s64 nr_to_invalidate =
should_invalidate_buckets(ca, bch2_dev_usage_read(ca));
break;
}
}
-
+err:
bch2_trans_exit(&trans);
bch2_write_ref_put(c, BCH_WRITE_REF_invalidate);
}
void bch2_do_invalidates(struct bch_fs *c)
{
if (bch2_write_ref_tryget(c, BCH_WRITE_REF_invalidate) &&
- !queue_work(system_long_wq, &c->invalidate_work))
+ !queue_work(c->write_ref_wq, &c->invalidate_work))
bch2_write_ref_put(c, BCH_WRITE_REF_invalidate);
}
-static int bch2_dev_freespace_init(struct bch_fs *c, struct bch_dev *ca)
+static int bch2_dev_freespace_init(struct bch_fs *c, struct bch_dev *ca,
+ unsigned long *last_updated)
{
struct btree_trans trans;
struct btree_iter iter;
* freespace/need_discard/need_gc_gens btrees as needed:
*/
while (1) {
+ if (*last_updated + HZ * 10 < jiffies) {
+ bch_info(ca, "%s: currently at %llu/%llu",
+ __func__, iter.pos.offset, ca->mi.nbuckets);
+ *last_updated = jiffies;
+ }
+
bch2_trans_begin(&trans);
if (bkey_ge(iter.pos, end)) {
freespace->k.p = k.k->p;
freespace->k.size = k.k->size;
- ret = __bch2_btree_insert(&trans, BTREE_ID_freespace, freespace) ?:
+ ret = __bch2_btree_insert(&trans, BTREE_ID_freespace, freespace, 0) ?:
bch2_trans_commit(&trans, NULL, NULL,
BTREE_INSERT_LAZY_RW|
BTREE_INSERT_NOFAIL);
unsigned i;
int ret = 0;
bool doing_init = false;
+ unsigned long last_updated = jiffies;
/*
* We can crash during the device add path, so we need to check this on
doing_init = true;
}
- ret = bch2_dev_freespace_init(c, ca);
+ ret = bch2_dev_freespace_init(c, ca, &last_updated);
if (ret) {
percpu_ref_put(&ca->ref);
return ret;
*/
bch2_recalc_capacity(c);
- /* Next, close write points that point to this device... */
- for (i = 0; i < ARRAY_SIZE(c->write_points); i++)
- bch2_writepoint_stop(c, ca, &c->write_points[i]);
-
- bch2_writepoint_stop(c, ca, &c->copygc_write_point);
- bch2_writepoint_stop(c, ca, &c->rebalance_write_point);
- bch2_writepoint_stop(c, ca, &c->btree_write_point);
-
- mutex_lock(&c->btree_reserve_cache_lock);
- while (c->btree_reserve_cache_nr) {
- struct btree_alloc *a =
- &c->btree_reserve_cache[--c->btree_reserve_cache_nr];
-
- bch2_open_buckets_put(c, &a->ob);
- }
- mutex_unlock(&c->btree_reserve_cache_lock);
-
- while (1) {
- struct open_bucket *ob;
-
- spin_lock(&c->freelist_lock);
- if (!ca->open_buckets_partial_nr) {
- spin_unlock(&c->freelist_lock);
- break;
- }
- ob = c->open_buckets +
- ca->open_buckets_partial[--ca->open_buckets_partial_nr];
- ob->on_partial_list = false;
- spin_unlock(&c->freelist_lock);
-
- bch2_open_bucket_put(c, ob);
- }
-
- bch2_ec_stop_dev(c, ca);
+ bch2_open_buckets_stop(c, ca, false);
/*
* Wake up threads that were blocked on allocation, so they can notice
projects / bcachefs-tools-debian / blobdiff