#include "ec.h"
#include "error.h"
#include "extents.h"
-#include "io.h"
+#include "io_write.h"
#include "keylist.h"
#include "move.h"
#include "nocow_locking.h"
+#include "rebalance.h"
#include "subvolume.h"
-
-#include <trace/events/bcachefs.h>
-
-static int insert_snapshot_whiteouts(struct btree_trans *trans,
- enum btree_id id,
- struct bpos old_pos,
- struct bpos new_pos)
-{
- struct bch_fs *c = trans->c;
- struct btree_iter iter, iter2;
- struct bkey_s_c k, k2;
- snapshot_id_list s;
- struct bkey_i *update;
- int ret;
-
- if (!btree_type_has_snapshots(id))
- return 0;
-
- darray_init(&s);
-
- if (!bch2_snapshot_has_children(c, old_pos.snapshot))
- return 0;
-
- bch2_trans_iter_init(trans, &iter, id, old_pos,
- BTREE_ITER_NOT_EXTENTS|
- BTREE_ITER_ALL_SNAPSHOTS);
- while (1) {
- k = bch2_btree_iter_prev(&iter);
- ret = bkey_err(k);
- if (ret)
- break;
-
- if (!k.k)
- break;
-
- if (!bkey_eq(old_pos, k.k->p))
- break;
-
- if (bch2_snapshot_is_ancestor(c, k.k->p.snapshot, old_pos.snapshot) &&
- !snapshot_list_has_ancestor(c, &s, k.k->p.snapshot)) {
- struct bpos whiteout_pos = new_pos;
-
- whiteout_pos.snapshot = k.k->p.snapshot;
-
- bch2_trans_iter_init(trans, &iter2, id, whiteout_pos,
- BTREE_ITER_NOT_EXTENTS|
- BTREE_ITER_INTENT);
- k2 = bch2_btree_iter_peek_slot(&iter2);
- ret = bkey_err(k2);
-
- if (!ret && k2.k->type == KEY_TYPE_deleted) {
- update = bch2_trans_kmalloc(trans, sizeof(struct bkey_i));
- ret = PTR_ERR_OR_ZERO(update);
- if (ret)
- break;
-
- bkey_init(&update->k);
- update->k.p = whiteout_pos;
- update->k.type = KEY_TYPE_whiteout;
-
- ret = bch2_trans_update(trans, &iter2, update,
- BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
- }
- bch2_trans_iter_exit(trans, &iter2);
-
- if (ret)
- break;
-
- ret = snapshot_list_add(c, &s, k.k->p.snapshot);
- if (ret)
- break;
- }
- }
- bch2_trans_iter_exit(trans, &iter);
- darray_exit(&s);
-
- return ret;
-}
+#include "trace.h"
static void trace_move_extent_finish2(struct bch_fs *c, struct bkey_s_c k)
{
if (insert) {
i = 0;
bkey_for_each_ptr_decode(old.k, bch2_bkey_ptrs_c(old), p, entry) {
- struct bkey_s new_s;
- new_s.k = (void *) new.k;
- new_s.v = (void *) new.v;
-
if (((1U << i) & m->data_opts.rewrite_ptrs) &&
(ptr = bch2_extent_has_ptr(old, p, bkey_i_to_s(insert))) &&
!ptr->cached)
if (((1U << i) & m->data_opts.rewrite_ptrs) &&
(ptr = bch2_extent_has_ptr(old, p, bkey_i_to_s(insert))) &&
!ptr->cached) {
- bch2_bkey_drop_ptr_noerror(bkey_i_to_s(insert), ptr);
- /*
- * See comment below:
bch2_extent_ptr_set_cached(bkey_i_to_s(insert), ptr);
- */
rewrites_found |= 1U << i;
}
i++;
if (!p.ptr.cached &&
durability - ptr_durability >= m->op.opts.data_replicas) {
durability -= ptr_durability;
- bch2_bkey_drop_ptr_noerror(bkey_i_to_s(insert), &entry->ptr);
- /*
- * Currently, we're dropping unneeded replicas
- * instead of marking them as cached, since
- * cached data in stripe buckets prevents them
- * from being reused:
+
bch2_extent_ptr_set_cached(bkey_i_to_s(insert), &entry->ptr);
- */
goto restart_drop_extra_replicas;
}
}
next_pos = insert->k.p;
- if (!bkey_eq(bkey_start_pos(&insert->k), bkey_start_pos(k.k))) {
- ret = insert_snapshot_whiteouts(trans, m->btree_id, k.k->p,
- bkey_start_pos(&insert->k));
- if (ret)
- goto err;
- }
-
- if (!bkey_eq(insert->k.p, k.k->p)) {
- ret = insert_snapshot_whiteouts(trans, m->btree_id,
- k.k->p, insert->k.p);
- if (ret)
- goto err;
+ /*
+ * Check for nonce offset inconsistency:
+ * This is debug code - we've been seeing this bug rarely, and
+ * it's been hard to reproduce, so this should give us some more
+ * information when it does occur:
+ */
+ struct printbuf err = PRINTBUF;
+ int invalid = bch2_bkey_invalid(c, bkey_i_to_s_c(insert), __btree_node_type(0, m->btree_id), 0, &err);
+ printbuf_exit(&err);
+
+ if (invalid) {
+ struct printbuf buf = PRINTBUF;
+
+ prt_str(&buf, "about to insert invalid key in data update path");
+ prt_str(&buf, "\nold: ");
+ bch2_bkey_val_to_text(&buf, c, old);
+ prt_str(&buf, "\nk: ");
+ bch2_bkey_val_to_text(&buf, c, k);
+ prt_str(&buf, "\nnew: ");
+ bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(insert));
+
+ bch2_print_string_as_lines(KERN_ERR, buf.buf);
+ printbuf_exit(&buf);
+
+ bch2_fatal_error(c);
+ goto out;
}
- ret = bch2_trans_update(trans, &iter, insert,
+ ret = bch2_insert_snapshot_whiteouts(trans, m->btree_id,
+ k.k->p, bkey_start_pos(&insert->k)) ?:
+ bch2_insert_snapshot_whiteouts(trans, m->btree_id,
+ k.k->p, insert->k.p) ?:
+ bch2_bkey_set_needs_rebalance(c, insert,
+ op->opts.background_target,
+ op->opts.background_compression) ?:
+ bch2_trans_update(trans, &iter, insert,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
bch2_trans_commit(trans, &op->res,
NULL,
- BTREE_INSERT_NOCHECK_RW|
- BTREE_INSERT_NOFAIL|
+ BCH_TRANS_COMMIT_no_check_rw|
+ BCH_TRANS_COMMIT_no_enospc|
m->data_opts.btree_insert_flags);
if (!ret) {
bch2_btree_iter_set_pos(&iter, next_pos);
}
continue;
nowork:
- if (m->ctxt && m->ctxt->stats) {
+ if (m->stats && m->stats) {
BUG_ON(k.k->p.offset <= iter.pos.offset);
- atomic64_inc(&m->ctxt->stats->keys_raced);
+ atomic64_inc(&m->stats->keys_raced);
atomic64_add(k.k->p.offset - iter.pos.offset,
- &m->ctxt->stats->sectors_raced);
+ &m->stats->sectors_raced);
}
- this_cpu_add(c->counters[BCH_COUNTER_move_extent_fail], new->k.size);
+ this_cpu_inc(c->counters[BCH_COUNTER_move_extent_fail]);
bch2_btree_iter_advance(&iter);
goto next;
int bch2_data_update_index_update(struct bch_write_op *op)
{
- return bch2_trans_run(op->c, __bch2_data_update_index_update(&trans, op));
+ return bch2_trans_run(op->c, __bch2_data_update_index_update(trans, op));
}
void bch2_data_update_read_done(struct data_update *m,
&update->op.devs_have,
update->op.nr_replicas,
update->op.nr_replicas,
- update->op.alloc_reserve,
+ update->op.watermark,
0, &cl, &wp);
if (bch2_err_matches(ret, BCH_ERR_operation_blocked)) {
bch2_trans_unlock(trans);
break;
}
- if ((atomic_read(&cl.remaining) & CLOSURE_REMAINING_MASK) != 1) {
+ if (closure_nr_remaining(&cl) != 1) {
bch2_trans_unlock(trans);
closure_sync(&cl);
}
bch2_bkey_buf_reassemble(&m->k, c, k);
m->btree_id = btree_id;
m->data_opts = data_opts;
+ m->ctxt = ctxt;
+ m->stats = ctxt ? ctxt->stats : NULL;
bch2_write_op_init(&m->op, c, io_opts);
m->op.pos = bkey_start_pos(k.k);
BCH_WRITE_DATA_ENCODED|
BCH_WRITE_MOVE|
m->data_opts.write_flags;
- m->op.compression_type =
- bch2_compression_opt_to_type[io_opts.background_compression ?:
- io_opts.compression];
- if (m->data_opts.btree_insert_flags & BTREE_INSERT_USE_RESERVE)
- m->op.alloc_reserve = RESERVE_movinggc;
+ m->op.compression_opt = io_opts.background_compression ?: io_opts.compression;
+ m->op.watermark = m->data_opts.btree_insert_flags & BCH_WATERMARK_MASK;
bkey_for_each_ptr(ptrs, ptr)
percpu_ref_get(&bch_dev_bkey_exists(c, ptr->dev)->ref);
if (crc_is_compressed(p.crc))
reserve_sectors += k.k->size;
- m->op.nr_replicas += bch2_extent_ptr_durability(c, &p);
+ m->op.nr_replicas += bch2_extent_ptr_desired_durability(c, &p);
} else if (!p.ptr.cached) {
bch2_dev_list_add_dev(&m->op.devs_have, p.ptr.dev);
}
if (c->opts.nocow_enabled) {
if (ctxt) {
- move_ctxt_wait_event(ctxt, trans,
+ move_ctxt_wait_event(ctxt,
(locked = bch2_bucket_nocow_trylock(&c->nocow_locks,
PTR_BUCKET_POS(c, &p.ptr), 0)) ||
!atomic_read(&ctxt->read_sectors));
projects / bcachefs-tools-debian / blobdiff