summaryrefslogtreecommitdiff
path: root/fs/bcachefs/io_read.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/bcachefs/io_read.c')
-rw-r--r--fs/bcachefs/io_read.c114
1 files changed, 66 insertions, 48 deletions
diff --git a/fs/bcachefs/io_read.c b/fs/bcachefs/io_read.c
index ebf39ef72fb2..4531c9ab3e12 100644
--- a/fs/bcachefs/io_read.c
+++ b/fs/bcachefs/io_read.c
@@ -93,21 +93,24 @@ static const struct rhashtable_params bch_promote_params = {
static inline int should_promote(struct bch_fs *c, struct bkey_s_c k,
struct bpos pos,
struct bch_io_opts opts,
- unsigned flags)
+ unsigned flags,
+ struct bch_io_failures *failed)
{
- BUG_ON(!opts.promote_target);
+ if (!failed) {
+ BUG_ON(!opts.promote_target);
- if (!(flags & BCH_READ_MAY_PROMOTE))
- return -BCH_ERR_nopromote_may_not;
+ if (!(flags & BCH_READ_MAY_PROMOTE))
+ return -BCH_ERR_nopromote_may_not;
- if (bch2_bkey_has_target(c, k, opts.promote_target))
- return -BCH_ERR_nopromote_already_promoted;
+ if (bch2_bkey_has_target(c, k, opts.promote_target))
+ return -BCH_ERR_nopromote_already_promoted;
- if (bkey_extent_is_unwritten(k))
- return -BCH_ERR_nopromote_unwritten;
+ if (bkey_extent_is_unwritten(k))
+ return -BCH_ERR_nopromote_unwritten;
- if (bch2_target_congested(c, opts.promote_target))
- return -BCH_ERR_nopromote_congested;
+ if (bch2_target_congested(c, opts.promote_target))
+ return -BCH_ERR_nopromote_congested;
+ }
if (rhashtable_lookup_fast(&c->promote_table, &pos,
bch_promote_params))
@@ -164,7 +167,8 @@ static struct promote_op *__promote_alloc(struct btree_trans *trans,
struct extent_ptr_decoded *pick,
struct bch_io_opts opts,
unsigned sectors,
- struct bch_read_bio **rbio)
+ struct bch_read_bio **rbio,
+ struct bch_io_failures *failed)
{
struct bch_fs *c = trans->c;
struct promote_op *op = NULL;
@@ -217,14 +221,28 @@ static struct promote_op *__promote_alloc(struct btree_trans *trans,
bio = &op->write.op.wbio.bio;
bio_init(bio, NULL, bio->bi_inline_vecs, pages, 0);
+ struct data_update_opts update_opts = {};
+
+ if (!failed) {
+ update_opts.target = opts.promote_target;
+ update_opts.extra_replicas = 1;
+ update_opts.write_flags = BCH_WRITE_ALLOC_NOWAIT|BCH_WRITE_CACHED;
+ } else {
+ update_opts.target = opts.foreground_target;
+
+ struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
+ unsigned i = 0;
+ bkey_for_each_ptr(ptrs, ptr) {
+ if (bch2_dev_io_failures(failed, ptr->dev))
+ update_opts.rewrite_ptrs |= BIT(i);
+ i++;
+ }
+ }
+
ret = bch2_data_update_init(trans, NULL, NULL, &op->write,
writepoint_hashed((unsigned long) current),
opts,
- (struct data_update_opts) {
- .target = opts.promote_target,
- .extra_replicas = 1,
- .write_flags = BCH_WRITE_ALLOC_NOWAIT|BCH_WRITE_CACHED,
- },
+ update_opts,
btree_id, k);
/*
* possible errors: -BCH_ERR_nocow_lock_blocked,
@@ -258,10 +276,17 @@ static struct promote_op *promote_alloc(struct btree_trans *trans,
unsigned flags,
struct bch_read_bio **rbio,
bool *bounce,
- bool *read_full)
+ bool *read_full,
+ struct bch_io_failures *failed)
{
struct bch_fs *c = trans->c;
- bool promote_full = *read_full || READ_ONCE(c->promote_whole_extents);
+ /*
+ * if failed != NULL we're not actually doing a promote, we're
+ * recovering from an io/checksum error
+ */
+ bool promote_full = (failed ||
+ *read_full ||
+ READ_ONCE(c->promote_whole_extents));
/* data might have to be decompressed in the write path: */
unsigned sectors = promote_full
? max(pick->crc.compressed_size, pick->crc.live_size)
@@ -272,7 +297,7 @@ static struct promote_op *promote_alloc(struct btree_trans *trans,
struct promote_op *promote;
int ret;
- ret = should_promote(c, k, pos, opts, flags);
+ ret = should_promote(c, k, pos, opts, flags, failed);
if (ret)
goto nopromote;
@@ -280,7 +305,7 @@ static struct promote_op *promote_alloc(struct btree_trans *trans,
k.k->type == KEY_TYPE_reflink_v
? BTREE_ID_reflink
: BTREE_ID_extents,
- k, pos, pick, opts, sectors, rbio);
+ k, pos, pick, opts, sectors, rbio, failed);
ret = PTR_ERR_OR_ZERO(promote);
if (ret)
goto nopromote;
@@ -910,9 +935,9 @@ retry_pick:
bounce = true;
}
- if (orig->opts.promote_target)
+ if (orig->opts.promote_target)// || failed)
promote = promote_alloc(trans, iter, k, &pick, orig->opts, flags,
- &rbio, &bounce, &read_full);
+ &rbio, &bounce, &read_full, failed);
if (!read_full) {
EBUG_ON(crc_is_compressed(pick.crc));
@@ -1122,34 +1147,27 @@ void __bch2_read(struct bch_fs *c, struct bch_read_bio *rbio,
struct btree_iter iter;
struct bkey_buf sk;
struct bkey_s_c k;
- u32 snapshot;
int ret;
BUG_ON(flags & BCH_READ_NODECODE);
bch2_bkey_buf_init(&sk);
-retry:
- bch2_trans_begin(trans);
- iter = (struct btree_iter) { NULL };
-
- ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
- if (ret)
- goto err;
-
bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
- SPOS(inum.inum, bvec_iter.bi_sector, snapshot),
+ POS(inum.inum, bvec_iter.bi_sector),
BTREE_ITER_slots);
+
while (1) {
unsigned bytes, sectors, offset_into_extent;
enum btree_id data_btree = BTREE_ID_extents;
- /*
- * read_extent -> io_time_reset may cause a transaction restart
- * without returning an error, we need to check for that here:
- */
- ret = bch2_trans_relock(trans);
+ bch2_trans_begin(trans);
+
+ u32 snapshot;
+ ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
if (ret)
- break;
+ goto err;
+
+ bch2_btree_iter_set_snapshot(&iter, snapshot);
bch2_btree_iter_set_pos(&iter,
POS(inum.inum, bvec_iter.bi_sector));
@@ -1157,7 +1175,7 @@ retry:
k = bch2_btree_iter_peek_slot(&iter);
ret = bkey_err(k);
if (ret)
- break;
+ goto err;
offset_into_extent = iter.pos.offset -
bkey_start_offset(k.k);
@@ -1168,7 +1186,7 @@ retry:
ret = bch2_read_indirect_extent(trans, &data_btree,
&offset_into_extent, &sk);
if (ret)
- break;
+ goto err;
k = bkey_i_to_s_c(sk.k);
@@ -1188,7 +1206,7 @@ retry:
data_btree, k,
offset_into_extent, failed, flags);
if (ret)
- break;
+ goto err;
if (flags & BCH_READ_LAST_FRAGMENT)
break;
@@ -1198,16 +1216,16 @@ retry:
ret = btree_trans_too_many_iters(trans);
if (ret)
+ goto err;
+err:
+ if (ret &&
+ !bch2_err_matches(ret, BCH_ERR_transaction_restart) &&
+ ret != READ_RETRY &&
+ ret != READ_RETRY_AVOID)
break;
}
-err:
- bch2_trans_iter_exit(trans, &iter);
-
- if (bch2_err_matches(ret, BCH_ERR_transaction_restart) ||
- ret == READ_RETRY ||
- ret == READ_RETRY_AVOID)
- goto retry;
+ bch2_trans_iter_exit(trans, &iter);
bch2_trans_put(trans);
bch2_bkey_buf_exit(&sk, c);
generated by cgit (git 2.34.1) at 2025-06-06 07:42:06 +0000