summaryrefslogtreecommitdiff
path: root/fs/bcachefs/ec.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/bcachefs/ec.c')
-rw-r--r--fs/bcachefs/ec.c1284
1 files changed, 765 insertions, 519 deletions
diff --git a/fs/bcachefs/ec.c b/fs/bcachefs/ec.c
index d503af270024..d2a5e76e6479 100644
--- a/fs/bcachefs/ec.c
+++ b/fs/bcachefs/ec.c
@@ -13,10 +13,12 @@
#include "btree_write_buffer.h"
#include "buckets.h"
#include "checksum.h"
+#include "disk_accounting.h"
#include "disk_groups.h"
#include "ec.h"
#include "error.h"
#include "io_read.h"
+#include "io_write.h"
#include "keylist.h"
#include "recovery.h"
#include "replicas.h"
@@ -24,6 +26,7 @@
#include "util.h"
#include <linux/sort.h>
+#include <linux/string_choices.h>
#ifdef __KERNEL__
@@ -106,24 +109,28 @@ struct ec_bio {
/* Stripes btree keys: */
-int bch2_stripe_invalid(struct bch_fs *c, struct bkey_s_c k,
- enum bkey_invalid_flags flags,
- struct printbuf *err)
+int bch2_stripe_validate(struct bch_fs *c, struct bkey_s_c k,
+ struct bkey_validate_context from)
{
const struct bch_stripe *s = bkey_s_c_to_stripe(k).v;
int ret = 0;
bkey_fsck_err_on(bkey_eq(k.k->p, POS_MIN) ||
- bpos_gt(k.k->p, POS(0, U32_MAX)), c, err,
- stripe_pos_bad,
+ bpos_gt(k.k->p, POS(0, U32_MAX)),
+ c, stripe_pos_bad,
"stripe at bad pos");
- bkey_fsck_err_on(bkey_val_u64s(k.k) < stripe_val_u64s(s), c, err,
- stripe_val_size_bad,
+ bkey_fsck_err_on(bkey_val_u64s(k.k) < stripe_val_u64s(s),
+ c, stripe_val_size_bad,
"incorrect value size (%zu < %u)",
bkey_val_u64s(k.k), stripe_val_u64s(s));
- ret = bch2_bkey_ptrs_invalid(c, k, flags, err);
+ bkey_fsck_err_on(s->csum_granularity_bits >= 64,
+ c, stripe_csum_granularity_bad,
+ "invalid csum granularity (%u >= 64)",
+ s->csum_granularity_bits);
+
+ ret = bch2_bkey_ptrs_validate(c, k, from);
fsck_err:
return ret;
}
@@ -131,174 +138,251 @@ fsck_err:
void bch2_stripe_to_text(struct printbuf *out, struct bch_fs *c,
struct bkey_s_c k)
{
- const struct bch_stripe *s = bkey_s_c_to_stripe(k).v;
- unsigned i, nr_data = s->nr_blocks - s->nr_redundant;
+ const struct bch_stripe *sp = bkey_s_c_to_stripe(k).v;
+ struct bch_stripe s = {};
- prt_printf(out, "algo %u sectors %u blocks %u:%u csum %u gran %u",
- s->algorithm,
- le16_to_cpu(s->sectors),
- nr_data,
- s->nr_redundant,
- s->csum_type,
- 1U << s->csum_granularity_bits);
+ memcpy(&s, sp, min(sizeof(s), bkey_val_bytes(k.k)));
- for (i = 0; i < s->nr_blocks; i++) {
- const struct bch_extent_ptr *ptr = s->ptrs + i;
- struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
- u32 offset;
- u64 b = sector_to_bucket_and_offset(ca, ptr->offset, &offset);
+ unsigned nr_data = s.nr_blocks - s.nr_redundant;
- prt_printf(out, " %u:%llu:%u", ptr->dev, b, offset);
- if (i < nr_data)
- prt_printf(out, "#%u", stripe_blockcount_get(s, i));
- prt_printf(out, " gen %u", ptr->gen);
- if (ptr_stale(ca, ptr))
- prt_printf(out, " stale");
+ prt_printf(out, "algo %u sectors %u blocks %u:%u csum ",
+ s.algorithm,
+ le16_to_cpu(s.sectors),
+ nr_data,
+ s.nr_redundant);
+ bch2_prt_csum_type(out, s.csum_type);
+ prt_str(out, " gran ");
+ if (s.csum_granularity_bits < 64)
+ prt_printf(out, "%llu", 1ULL << s.csum_granularity_bits);
+ else
+ prt_printf(out, "(invalid shift %u)", s.csum_granularity_bits);
+
+ if (s.disk_label) {
+ prt_str(out, " label");
+ bch2_disk_path_to_text(out, c, s.disk_label - 1);
+ }
+
+ for (unsigned i = 0; i < s.nr_blocks; i++) {
+ const struct bch_extent_ptr *ptr = sp->ptrs + i;
+
+ if ((void *) ptr >= bkey_val_end(k))
+ break;
+
+ prt_char(out, ' ');
+ bch2_extent_ptr_to_text(out, c, ptr);
+
+ if (s.csum_type < BCH_CSUM_NR &&
+ i < nr_data &&
+ stripe_blockcount_offset(&s, i) < bkey_val_bytes(k.k))
+ prt_printf(out, "#%u", stripe_blockcount_get(sp, i));
}
}
/* Triggers: */
-static int bch2_trans_mark_stripe_bucket(struct btree_trans *trans,
- struct bkey_s_c_stripe s,
- unsigned idx, bool deleting)
+static int __mark_stripe_bucket(struct btree_trans *trans,
+ struct bch_dev *ca,
+ struct bkey_s_c_stripe s,
+ unsigned ptr_idx, bool deleting,
+ struct bpos bucket,
+ struct bch_alloc_v4 *a,
+ enum btree_iter_update_trigger_flags flags)
{
- struct bch_fs *c = trans->c;
- const struct bch_extent_ptr *ptr = &s.v->ptrs[idx];
- struct btree_iter iter;
- struct bkey_i_alloc_v4 *a;
- enum bch_data_type data_type = idx >= s.v->nr_blocks - s.v->nr_redundant
- ? BCH_DATA_parity : 0;
- s64 sectors = data_type ? le16_to_cpu(s.v->sectors) : 0;
+ const struct bch_extent_ptr *ptr = s.v->ptrs + ptr_idx;
+ unsigned nr_data = s.v->nr_blocks - s.v->nr_redundant;
+ bool parity = ptr_idx >= nr_data;
+ enum bch_data_type data_type = parity ? BCH_DATA_parity : BCH_DATA_stripe;
+ s64 sectors = parity ? le16_to_cpu(s.v->sectors) : 0;
+ struct printbuf buf = PRINTBUF;
int ret = 0;
+ struct bch_fs *c = trans->c;
if (deleting)
sectors = -sectors;
- a = bch2_trans_start_alloc_update(trans, &iter, PTR_BUCKET_POS(c, ptr));
- if (IS_ERR(a))
- return PTR_ERR(a);
-
- ret = bch2_check_bucket_ref(trans, s.s_c, ptr, sectors, data_type,
- a->v.gen, a->v.data_type,
- a->v.dirty_sectors);
- if (ret)
- goto err;
-
if (!deleting) {
- if (bch2_trans_inconsistent_on(a->v.stripe ||
- a->v.stripe_redundancy, trans,
- "bucket %llu:%llu gen %u data type %s dirty_sectors %u: multiple stripes using same bucket (%u, %llu)",
- iter.pos.inode, iter.pos.offset, a->v.gen,
- bch2_data_type_str(a->v.data_type),
- a->v.dirty_sectors,
- a->v.stripe, s.k->p.offset)) {
- ret = -EIO;
+ if (bch2_trans_inconsistent_on(a->stripe ||
+ a->stripe_redundancy, trans,
+ "bucket %llu:%llu gen %u data type %s dirty_sectors %u: multiple stripes using same bucket (%u, %llu)\n%s",
+ bucket.inode, bucket.offset, a->gen,
+ bch2_data_type_str(a->data_type),
+ a->dirty_sectors,
+ a->stripe, s.k->p.offset,
+ (bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
+ ret = -BCH_ERR_mark_stripe;
goto err;
}
- if (bch2_trans_inconsistent_on(data_type && a->v.dirty_sectors, trans,
- "bucket %llu:%llu gen %u data type %s dirty_sectors %u: data already in stripe bucket %llu",
- iter.pos.inode, iter.pos.offset, a->v.gen,
- bch2_data_type_str(a->v.data_type),
- a->v.dirty_sectors,
- s.k->p.offset)) {
- ret = -EIO;
+ if (bch2_trans_inconsistent_on(parity && bch2_bucket_sectors_total(*a), trans,
+ "bucket %llu:%llu gen %u data type %s dirty_sectors %u cached_sectors %u: data already in parity bucket\n%s",
+ bucket.inode, bucket.offset, a->gen,
+ bch2_data_type_str(a->data_type),
+ a->dirty_sectors,
+ a->cached_sectors,
+ (bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
+ ret = -BCH_ERR_mark_stripe;
goto err;
}
-
- a->v.stripe = s.k->p.offset;
- a->v.stripe_redundancy = s.v->nr_redundant;
- a->v.data_type = BCH_DATA_stripe;
} else {
- if (bch2_trans_inconsistent_on(a->v.stripe != s.k->p.offset ||
- a->v.stripe_redundancy != s.v->nr_redundant, trans,
- "bucket %llu:%llu gen %u: not marked as stripe when deleting stripe %llu (got %u)",
- iter.pos.inode, iter.pos.offset, a->v.gen,
- s.k->p.offset, a->v.stripe)) {
- ret = -EIO;
+ if (bch2_trans_inconsistent_on(a->stripe != s.k->p.offset ||
+ a->stripe_redundancy != s.v->nr_redundant, trans,
+ "bucket %llu:%llu gen %u: not marked as stripe when deleting stripe (got %u)\n%s",
+ bucket.inode, bucket.offset, a->gen,
+ a->stripe,
+ (bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
+ ret = -BCH_ERR_mark_stripe;
+ goto err;
+ }
+
+ if (bch2_trans_inconsistent_on(a->data_type != data_type, trans,
+ "bucket %llu:%llu gen %u data type %s: wrong data type when stripe, should be %s\n%s",
+ bucket.inode, bucket.offset, a->gen,
+ bch2_data_type_str(a->data_type),
+ bch2_data_type_str(data_type),
+ (bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
+ ret = -BCH_ERR_mark_stripe;
goto err;
}
- a->v.stripe = 0;
- a->v.stripe_redundancy = 0;
- a->v.data_type = alloc_data_type(a->v, BCH_DATA_user);
+ if (bch2_trans_inconsistent_on(parity &&
+ (a->dirty_sectors != -sectors ||
+ a->cached_sectors), trans,
+ "bucket %llu:%llu gen %u dirty_sectors %u cached_sectors %u: wrong sectors when deleting parity block of stripe\n%s",
+ bucket.inode, bucket.offset, a->gen,
+ a->dirty_sectors,
+ a->cached_sectors,
+ (bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
+ ret = -BCH_ERR_mark_stripe;
+ goto err;
+ }
}
- a->v.dirty_sectors += sectors;
- if (data_type)
- a->v.data_type = !deleting ? data_type : 0;
+ if (sectors) {
+ ret = bch2_bucket_ref_update(trans, ca, s.s_c, ptr, sectors, data_type,
+ a->gen, a->data_type, &a->dirty_sectors);
+ if (ret)
+ goto err;
+ }
- ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
- if (ret)
- goto err;
+ if (!deleting) {
+ a->stripe = s.k->p.offset;
+ a->stripe_redundancy = s.v->nr_redundant;
+ alloc_data_type_set(a, data_type);
+ } else {
+ a->stripe = 0;
+ a->stripe_redundancy = 0;
+ alloc_data_type_set(a, BCH_DATA_user);
+ }
err:
- bch2_trans_iter_exit(trans, &iter);
+ printbuf_exit(&buf);
return ret;
}
static int mark_stripe_bucket(struct btree_trans *trans,
- struct bkey_s_c k,
- unsigned ptr_idx,
- unsigned flags)
+ struct bkey_s_c_stripe s,
+ unsigned ptr_idx, bool deleting,
+ enum btree_iter_update_trigger_flags flags)
{
struct bch_fs *c = trans->c;
- const struct bch_stripe *s = bkey_s_c_to_stripe(k).v;
- unsigned nr_data = s->nr_blocks - s->nr_redundant;
- bool parity = ptr_idx >= nr_data;
- enum bch_data_type data_type = parity ? BCH_DATA_parity : BCH_DATA_stripe;
- s64 sectors = parity ? le16_to_cpu(s->sectors) : 0;
- const struct bch_extent_ptr *ptr = s->ptrs + ptr_idx;
- struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
- struct bucket old, new, *g;
+ const struct bch_extent_ptr *ptr = s.v->ptrs + ptr_idx;
struct printbuf buf = PRINTBUF;
int ret = 0;
- BUG_ON(!(flags & BTREE_TRIGGER_GC));
-
- /* * XXX doesn't handle deletion */
-
- percpu_down_read(&c->mark_lock);
- g = PTR_GC_BUCKET(ca, ptr);
-
- if (g->dirty_sectors ||
- (g->stripe && g->stripe != k.k->p.offset)) {
- bch2_fs_inconsistent(c,
- "bucket %u:%zu gen %u: multiple stripes using same bucket\n%s",
- ptr->dev, PTR_BUCKET_NR(ca, ptr), g->gen,
- (bch2_bkey_val_to_text(&buf, c, k), buf.buf));
- ret = -EINVAL;
+ struct bch_dev *ca = bch2_dev_tryget(c, ptr->dev);
+ if (unlikely(!ca)) {
+ if (ptr->dev != BCH_SB_MEMBER_INVALID && !(flags & BTREE_TRIGGER_overwrite))
+ ret = -BCH_ERR_mark_stripe;
goto err;
}
- bucket_lock(g);
- old = *g;
+ struct bpos bucket = PTR_BUCKET_POS(ca, ptr);
- ret = bch2_check_bucket_ref(trans, k, ptr, sectors, data_type,
- g->gen, g->data_type,
- g->dirty_sectors);
- if (ret)
- goto err;
+ if (flags & BTREE_TRIGGER_transactional) {
+ struct bkey_i_alloc_v4 *a =
+ bch2_trans_start_alloc_update(trans, bucket, 0);
+ ret = PTR_ERR_OR_ZERO(a) ?:
+ __mark_stripe_bucket(trans, ca, s, ptr_idx, deleting, bucket, &a->v, flags);
+ }
+
+ if (flags & BTREE_TRIGGER_gc) {
+ struct bucket *g = gc_bucket(ca, bucket.offset);
+ if (bch2_fs_inconsistent_on(!g, c, "reference to invalid bucket on device %u\n %s",
+ ptr->dev,
+ (bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
+ ret = -BCH_ERR_mark_stripe;
+ goto err;
+ }
- g->data_type = data_type;
- g->dirty_sectors += sectors;
+ bucket_lock(g);
+ struct bch_alloc_v4 old = bucket_m_to_alloc(*g), new = old;
+ ret = __mark_stripe_bucket(trans, ca, s, ptr_idx, deleting, bucket, &new, flags);
+ alloc_to_bucket(g, new);
+ bucket_unlock(g);
- g->stripe = k.k->p.offset;
- g->stripe_redundancy = s->nr_redundant;
- new = *g;
+ if (!ret)
+ ret = bch2_alloc_key_to_dev_counters(trans, ca, &old, &new, flags);
+ }
err:
- bucket_unlock(g);
- if (!ret)
- bch2_dev_usage_update_m(c, ca, &old, &new);
- percpu_up_read(&c->mark_lock);
+ bch2_dev_put(ca);
printbuf_exit(&buf);
return ret;
}
+static int mark_stripe_buckets(struct btree_trans *trans,
+ struct bkey_s_c old, struct bkey_s_c new,
+ enum btree_iter_update_trigger_flags flags)
+{
+ const struct bch_stripe *old_s = old.k->type == KEY_TYPE_stripe
+ ? bkey_s_c_to_stripe(old).v : NULL;
+ const struct bch_stripe *new_s = new.k->type == KEY_TYPE_stripe
+ ? bkey_s_c_to_stripe(new).v : NULL;
+
+ BUG_ON(old_s && new_s && old_s->nr_blocks != new_s->nr_blocks);
+
+ unsigned nr_blocks = new_s ? new_s->nr_blocks : old_s->nr_blocks;
+
+ for (unsigned i = 0; i < nr_blocks; i++) {
+ if (new_s && old_s &&
+ !memcmp(&new_s->ptrs[i],
+ &old_s->ptrs[i],
+ sizeof(new_s->ptrs[i])))
+ continue;
+
+ if (new_s) {
+ int ret = mark_stripe_bucket(trans,
+ bkey_s_c_to_stripe(new), i, false, flags);
+ if (ret)
+ return ret;
+ }
+
+ if (old_s) {
+ int ret = mark_stripe_bucket(trans,
+ bkey_s_c_to_stripe(old), i, true, flags);
+ if (ret)
+ return ret;
+ }
+ }
+
+ return 0;
+}
+
+static inline void stripe_to_mem(struct stripe *m, const struct bch_stripe *s)
+{
+ m->sectors = le16_to_cpu(s->sectors);
+ m->algorithm = s->algorithm;
+ m->nr_blocks = s->nr_blocks;
+ m->nr_redundant = s->nr_redundant;
+ m->disk_label = s->disk_label;
+ m->blocks_nonempty = 0;
+
+ for (unsigned i = 0; i < s->nr_blocks; i++)
+ m->blocks_nonempty += !!stripe_blockcount_get(s, i);
+}
+
int bch2_trigger_stripe(struct btree_trans *trans,
- enum btree_id btree_id, unsigned level,
+ enum btree_id btree, unsigned level,
struct bkey_s_c old, struct bkey_s _new,
- unsigned flags)
+ enum btree_iter_update_trigger_flags flags)
{
struct bkey_s_c new = _new.s_c;
struct bch_fs *c = trans->c;
@@ -308,7 +392,15 @@ int bch2_trigger_stripe(struct btree_trans *trans,
const struct bch_stripe *new_s = new.k->type == KEY_TYPE_stripe
? bkey_s_c_to_stripe(new).v : NULL;
- if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+ if (unlikely(flags & BTREE_TRIGGER_check_repair))
+ return bch2_check_fix_ptrs(trans, btree, level, _new.s_c, flags);
+
+ BUG_ON(new_s && old_s &&
+ (new_s->nr_blocks != old_s->nr_blocks ||
+ new_s->nr_redundant != old_s->nr_redundant));
+
+
+ if (flags & (BTREE_TRIGGER_transactional|BTREE_TRIGGER_gc)) {
/*
* If the pointers aren't changing, we don't need to do anything:
*/
@@ -319,55 +411,68 @@ int bch2_trigger_stripe(struct btree_trans *trans,
new_s->nr_blocks * sizeof(struct bch_extent_ptr)))
return 0;
- BUG_ON(new_s && old_s &&
- (new_s->nr_blocks != old_s->nr_blocks ||
- new_s->nr_redundant != old_s->nr_redundant));
+ struct gc_stripe *gc = NULL;
+ if (flags & BTREE_TRIGGER_gc) {
+ gc = genradix_ptr_alloc(&c->gc_stripes, idx, GFP_KERNEL);
+ if (!gc) {
+ bch_err(c, "error allocating memory for gc_stripes, idx %llu", idx);
+ return -BCH_ERR_ENOMEM_mark_stripe;
+ }
+
+ /*
+ * This will be wrong when we bring back runtime gc: we should
+ * be unmarking the old key and then marking the new key
+ *
+ * Also: when we bring back runtime gc, locking
+ */
+ gc->alive = true;
+ gc->sectors = le16_to_cpu(new_s->sectors);
+ gc->nr_blocks = new_s->nr_blocks;
+ gc->nr_redundant = new_s->nr_redundant;
+
+ for (unsigned i = 0; i < new_s->nr_blocks; i++)
+ gc->ptrs[i] = new_s->ptrs[i];
+
+ /*
+ * gc recalculates this field from stripe ptr
+ * references:
+ */
+ memset(gc->block_sectors, 0, sizeof(gc->block_sectors));
+ }
if (new_s) {
- s64 sectors = le16_to_cpu(new_s->sectors);
+ s64 sectors = (u64) le16_to_cpu(new_s->sectors) * new_s->nr_redundant;
- struct bch_replicas_padded r;
- bch2_bkey_to_replicas(&r.e, new);
- int ret = bch2_update_replicas_list(trans, &r.e, sectors * new_s->nr_redundant);
+ struct disk_accounting_pos acc = {
+ .type = BCH_DISK_ACCOUNTING_replicas,
+ };
+ bch2_bkey_to_replicas(&acc.replicas, new);
+ int ret = bch2_disk_accounting_mod(trans, &acc, &sectors, 1, gc);
if (ret)
return ret;
+
+ if (gc)
+ memcpy(&gc->r.e, &acc.replicas, replicas_entry_bytes(&acc.replicas));
}
if (old_s) {
- s64 sectors = -((s64) le16_to_cpu(old_s->sectors));
+ s64 sectors = -((s64) le16_to_cpu(old_s->sectors)) * old_s->nr_redundant;
- struct bch_replicas_padded r;
- bch2_bkey_to_replicas(&r.e, old);
- int ret = bch2_update_replicas_list(trans, &r.e, sectors * old_s->nr_redundant);
+ struct disk_accounting_pos acc = {
+ .type = BCH_DISK_ACCOUNTING_replicas,
+ };
+ bch2_bkey_to_replicas(&acc.replicas, old);
+ int ret = bch2_disk_accounting_mod(trans, &acc, &sectors, 1, gc);
if (ret)
return ret;
}
- unsigned nr_blocks = new_s ? new_s->nr_blocks : old_s->nr_blocks;
- for (unsigned i = 0; i < nr_blocks; i++) {
- if (new_s && old_s &&
- !memcmp(&new_s->ptrs[i],
- &old_s->ptrs[i],
- sizeof(new_s->ptrs[i])))
- continue;
-
- if (new_s) {
- int ret = bch2_trans_mark_stripe_bucket(trans,
- bkey_s_c_to_stripe(new), i, false);
- if (ret)
- return ret;
- }
-
- if (old_s) {
- int ret = bch2_trans_mark_stripe_bucket(trans,
- bkey_s_c_to_stripe(old), i, true);
- if (ret)
- return ret;
- }
- }
+ int ret = mark_stripe_buckets(trans, old, new, flags);
+ if (ret)
+ return ret;
}
- if (flags & BTREE_TRIGGER_ATOMIC) {
+ if (flags & BTREE_TRIGGER_atomic) {
struct stripe *m = genradix_ptr(&c->stripes, idx);
if (!m) {
@@ -390,14 +495,7 @@ int bch2_trigger_stripe(struct btree_trans *trans,
memset(m, 0, sizeof(*m));
} else {
- m->sectors = le16_to_cpu(new_s->sectors);
- m->algorithm = new_s->algorithm;
- m->nr_blocks = new_s->nr_blocks;
- m->nr_redundant = new_s->nr_redundant;
- m->blocks_nonempty = 0;
-
- for (unsigned i = 0; i < new_s->nr_blocks; i++)
- m->blocks_nonempty += !!stripe_blockcount_get(new_s, i);
+ stripe_to_mem(m, new_s);
if (!old_s)
bch2_stripes_heap_insert(c, m, idx);
@@ -406,54 +504,6 @@ int bch2_trigger_stripe(struct btree_trans *trans,
}
}
- if (flags & BTREE_TRIGGER_GC) {
- struct gc_stripe *m =
- genradix_ptr_alloc(&c->gc_stripes, idx, GFP_KERNEL);
-
- if (!m) {
- bch_err(c, "error allocating memory for gc_stripes, idx %llu",
- idx);
- return -BCH_ERR_ENOMEM_mark_stripe;
- }
- /*
- * This will be wrong when we bring back runtime gc: we should
- * be unmarking the old key and then marking the new key
- */
- m->alive = true;
- m->sectors = le16_to_cpu(new_s->sectors);
- m->nr_blocks = new_s->nr_blocks;
- m->nr_redundant = new_s->nr_redundant;
-
- for (unsigned i = 0; i < new_s->nr_blocks; i++)
- m->ptrs[i] = new_s->ptrs[i];
-
- bch2_bkey_to_replicas(&m->r.e, new);
-
- /*
- * gc recalculates this field from stripe ptr
- * references:
- */
- memset(m->block_sectors, 0, sizeof(m->block_sectors));
-
- for (unsigned i = 0; i < new_s->nr_blocks; i++) {
- int ret = mark_stripe_bucket(trans, new, i, flags);
- if (ret)
- return ret;
- }
-
- int ret = bch2_update_replicas(c, new, &m->r.e,
- ((s64) m->sectors * m->nr_redundant),
- 0, true);
- if (ret) {
- struct printbuf buf = PRINTBUF;
-
- bch2_bkey_val_to_text(&buf, c, new);
- bch2_fs_fatal_error(c, "no replicas entry for %s", buf.buf);
- printbuf_exit(&buf);
- return ret;
- }
- }
-
return 0;
}
@@ -504,7 +554,7 @@ static void ec_stripe_buf_exit(struct ec_stripe_buf *buf)
unsigned i;
for (i = 0; i < s->v.nr_blocks; i++) {
- kvpfree(buf->data[i], buf->size << 9);
+ kvfree(buf->data[i]);
buf->data[i] = NULL;
}
}
@@ -531,7 +581,7 @@ static int ec_stripe_buf_init(struct ec_stripe_buf *buf,
memset(buf->valid, 0xFF, sizeof(buf->valid));
for (i = 0; i < v->nr_blocks; i++) {
- buf->data[i] = kvpmalloc(buf->size << 9, GFP_KERNEL);
+ buf->data[i] = kvmalloc(buf->size << 9, GFP_KERNEL);
if (!buf->data[i])
goto err;
}
@@ -604,21 +654,21 @@ static void ec_validate_checksums(struct bch_fs *c, struct ec_stripe_buf *buf)
struct bch_csum got = ec_block_checksum(buf, i, offset);
if (bch2_crc_cmp(want, got)) {
- struct printbuf err = PRINTBUF;
- struct bch_dev *ca = bch_dev_bkey_exists(c, v->ptrs[i].dev);
-
- prt_printf(&err, "stripe checksum error: expected %0llx:%0llx got %0llx:%0llx (type %s)\n",
- want.hi, want.lo,
- got.hi, got.lo,
- bch2_csum_types[v->csum_type]);
- prt_printf(&err, " for %ps at %u of\n ", (void *) _RET_IP_, i);
- bch2_bkey_val_to_text(&err, c, bkey_i_to_s_c(&buf->key));
- bch_err_ratelimited(ca, "%s", err.buf);
- printbuf_exit(&err);
+ struct bch_dev *ca = bch2_dev_tryget(c, v->ptrs[i].dev);
+ if (ca) {
+ struct printbuf err = PRINTBUF;
- clear_bit(i, buf->valid);
+ prt_str(&err, "stripe ");
+ bch2_csum_err_msg(&err, v->csum_type, want, got);
+ prt_printf(&err, " for %ps at %u of\n ", (void *) _RET_IP_, i);
+ bch2_bkey_val_to_text(&err, c, bkey_i_to_s_c(&buf->key));
+ bch_err_ratelimited(ca, "%s", err.buf);
+ printbuf_exit(&err);
+
+ bch2_io_error(ca, BCH_MEMBER_ERROR_checksum);
+ }
- bch2_io_error(ca, BCH_MEMBER_ERROR_checksum);
+ clear_bit(i, buf->valid);
break;
}
@@ -681,14 +731,16 @@ static void ec_block_endio(struct bio *bio)
? BCH_MEMBER_ERROR_write
: BCH_MEMBER_ERROR_read,
"erasure coding %s error: %s",
- bio_data_dir(bio) ? "write" : "read",
+ str_write_read(bio_data_dir(bio)),
bch2_blk_status_to_str(bio->bi_status)))
clear_bit(ec_bio->idx, ec_bio->buf->valid);
- if (ptr_stale(ca, ptr)) {
+ int stale = dev_ptr_stale(ca, ptr);
+ if (stale) {
bch_err_ratelimited(ca->fs,
- "error %s stripe: stale pointer after io",
- bio_data_dir(bio) == READ ? "reading from" : "writing to");
+ "error %s stripe: stale/invalid pointer (%i) after io",
+ bio_data_dir(bio) == READ ? "reading from" : "writing to",
+ stale);
clear_bit(ec_bio->idx, ec_bio->buf->valid);
}
@@ -703,25 +755,28 @@ static void ec_block_io(struct bch_fs *c, struct ec_stripe_buf *buf,
struct bch_stripe *v = &bkey_i_to_stripe(&buf->key)->v;
unsigned offset = 0, bytes = buf->size << 9;
struct bch_extent_ptr *ptr = &v->ptrs[idx];
- struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
enum bch_data_type data_type = idx < v->nr_blocks - v->nr_redundant
? BCH_DATA_user
: BCH_DATA_parity;
int rw = op_is_write(opf);
- if (ptr_stale(ca, ptr)) {
- bch_err_ratelimited(c,
- "error %s stripe: stale pointer",
- rw == READ ? "reading from" : "writing to");
+ struct bch_dev *ca = bch2_dev_get_ioref(c, ptr->dev, rw);
+ if (!ca) {
clear_bit(idx, buf->valid);
return;
}
- if (!bch2_dev_get_ioref(ca, rw)) {
+ int stale = dev_ptr_stale(ca, ptr);
+ if (stale) {
+ bch_err_ratelimited(c,
+ "error %s stripe: stale pointer (%i)",
+ rw == READ ? "reading from" : "writing to",
+ stale);
clear_bit(idx, buf->valid);
return;
}
+
this_cpu_add(ca->io_done->sectors[rw][data_type], buf->size);
while (offset < bytes) {
@@ -767,7 +822,7 @@ static int get_stripe_key_trans(struct btree_trans *trans, u64 idx,
int ret;
k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_stripes,
- POS(0, idx), BTREE_ITER_SLOTS);
+ POS(0, idx), BTREE_ITER_slots);
ret = bkey_err(k);
if (ret)
goto err;
@@ -782,13 +837,16 @@ err:
}
/* recovery read path: */
-int bch2_ec_read_extent(struct btree_trans *trans, struct bch_read_bio *rbio)
+int bch2_ec_read_extent(struct btree_trans *trans, struct bch_read_bio *rbio,
+ struct bkey_s_c orig_k)
{
struct bch_fs *c = trans->c;
- struct ec_stripe_buf *buf;
+ struct ec_stripe_buf *buf = NULL;
struct closure cl;
struct bch_stripe *v;
unsigned i, offset;
+ const char *msg = NULL;
+ struct printbuf msgbuf = PRINTBUF;
int ret = 0;
closure_init_stack(&cl);
@@ -801,32 +859,28 @@ int bch2_ec_read_extent(struct btree_trans *trans, struct bch_read_bio *rbio)
ret = lockrestart_do(trans, get_stripe_key_trans(trans, rbio->pick.ec.idx, buf));
if (ret) {
- bch_err_ratelimited(c,
- "error doing reconstruct read: error %i looking up stripe", ret);
- kfree(buf);
- return -EIO;
+ msg = "stripe not found";
+ goto err;
}
v = &bkey_i_to_stripe(&buf->key)->v;
if (!bch2_ptr_matches_stripe(v, rbio->pick)) {
- bch_err_ratelimited(c,
- "error doing reconstruct read: pointer doesn't match stripe");
- ret = -EIO;
+ msg = "pointer doesn't match stripe";
goto err;
}
offset = rbio->bio.bi_iter.bi_sector - v->ptrs[rbio->pick.ec.block].offset;
if (offset + bio_sectors(&rbio->bio) > le16_to_cpu(v->sectors)) {
- bch_err_ratelimited(c,
- "error doing reconstruct read: read is bigger than stripe");
- ret = -EIO;
+ msg = "read is bigger than stripe";
goto err;
}
ret = ec_stripe_buf_init(buf, offset, bio_sectors(&rbio->bio));
- if (ret)
+ if (ret) {
+ msg = "-ENOMEM";
goto err;
+ }
for (i = 0; i < v->nr_blocks; i++)
ec_block_io(c, buf, REQ_OP_READ, i, &cl);
@@ -834,9 +888,7 @@ int bch2_ec_read_extent(struct btree_trans *trans, struct bch_read_bio *rbio)
closure_sync(&cl);
if (ec_nr_failed(buf) > v->nr_redundant) {
- bch_err_ratelimited(c,
- "error doing reconstruct read: unable to read enough blocks");
- ret = -EIO;
+ msg = "unable to read enough blocks";
goto err;
}
@@ -848,10 +900,17 @@ int bch2_ec_read_extent(struct btree_trans *trans, struct bch_read_bio *rbio)
memcpy_to_bio(&rbio->bio, rbio->bio.bi_iter,
buf->data[rbio->pick.ec.block] + ((offset - buf->offset) << 9));
-err:
+out:
ec_stripe_buf_exit(buf);
kfree(buf);
return ret;
+err:
+ bch2_bkey_val_to_text(&msgbuf, c, orig_k);
+ bch_err_ratelimited(c,
+ "error doing reconstruct read: %s\n %s", msg, msgbuf.buf);
+ printbuf_exit(&msgbuf);
+ ret = -BCH_ERR_stripe_reconstruct;
+ goto out;
}
/* stripe bucket accounting: */
@@ -866,8 +925,8 @@ static int __ec_stripe_mem_alloc(struct bch_fs *c, size_t idx, gfp_t gfp)
mutex_lock(&c->ec_stripes_heap_lock);
if (n.size > h->size) {
- memcpy(n.data, h->data, h->used * sizeof(h->data[0]));
- n.used = h->used;
+ memcpy(n.data, h->data, h->nr * sizeof(h->data[0]));
+ n.nr = h->nr;
swap(*h, n);
}
mutex_unlock(&c->ec_stripes_heap_lock);
@@ -878,7 +937,7 @@ static int __ec_stripe_mem_alloc(struct bch_fs *c, size_t idx, gfp_t gfp)
if (!genradix_ptr_alloc(&c->stripes, idx, gfp))
return -BCH_ERR_ENOMEM_ec_stripe_mem_alloc;
- if (c->gc_pos.phase != GC_PHASE_NOT_RUNNING &&
+ if (c->gc_pos.phase != GC_PHASE_not_running &&
!genradix_ptr_alloc(&c->gc_stripes, idx, gfp))
return -BCH_ERR_ENOMEM_ec_stripe_mem_alloc;
@@ -958,7 +1017,7 @@ static u64 stripe_idx_to_delete(struct bch_fs *c)
lockdep_assert_held(&c->ec_stripes_heap_lock);
- if (h->used &&
+ if (h->nr &&
h->data[0].blocks_nonempty == 0 &&
!bch2_stripe_is_open(c, h->data[0].idx))
return h->data[0].idx;
@@ -966,14 +1025,6 @@ static u64 stripe_idx_to_delete(struct bch_fs *c)
return 0;
}
-static inline int ec_stripes_heap_cmp(ec_stripes_heap *h,
- struct ec_stripe_heap_entry l,
- struct ec_stripe_heap_entry r)
-{
- return ((l.blocks_nonempty > r.blocks_nonempty) -
- (l.blocks_nonempty < r.blocks_nonempty));
-}
-
static inline void ec_stripes_heap_set_backpointer(ec_stripes_heap *h,
size_t i)
{
@@ -982,12 +1033,40 @@ static inline void ec_stripes_heap_set_backpointer(ec_stripes_heap *h,
genradix_ptr(&c->stripes, h->data[i].idx)->heap_idx = i;
}
+static inline bool ec_stripes_heap_cmp(const void *l, const void *r, void __always_unused *args)
+{
+ struct ec_stripe_heap_entry *_l = (struct ec_stripe_heap_entry *)l;
+ struct ec_stripe_heap_entry *_r = (struct ec_stripe_heap_entry *)r;
+
+ return ((_l->blocks_nonempty > _r->blocks_nonempty) <
+ (_l->blocks_nonempty < _r->blocks_nonempty));
+}
+
+static inline void ec_stripes_heap_swap(void *l, void *r, void *h)
+{
+ struct ec_stripe_heap_entry *_l = (struct ec_stripe_heap_entry *)l;
+ struct ec_stripe_heap_entry *_r = (struct ec_stripe_heap_entry *)r;
+ ec_stripes_heap *_h = (ec_stripes_heap *)h;
+ size_t i = _l - _h->data;
+ size_t j = _r - _h->data;
+
+ swap(*_l, *_r);
+
+ ec_stripes_heap_set_backpointer(_h, i);
+ ec_stripes_heap_set_backpointer(_h, j);
+}
+
+static const struct min_heap_callbacks callbacks = {
+ .less = ec_stripes_heap_cmp,
+ .swp = ec_stripes_heap_swap,
+};
+
static void heap_verify_backpointer(struct bch_fs *c, size_t idx)
{
ec_stripes_heap *h = &c->ec_stripes_heap;
struct stripe *m = genradix_ptr(&c->stripes, idx);
- BUG_ON(m->heap_idx >= h->used);
+ BUG_ON(m->heap_idx >= h->nr);
BUG_ON(h->data[m->heap_idx].idx != idx);
}
@@ -997,9 +1076,7 @@ void bch2_stripes_heap_del(struct bch_fs *c,
mutex_lock(&c->ec_stripes_heap_lock);
heap_verify_backpointer(c, idx);
- heap_del(&c->ec_stripes_heap, m->heap_idx,
- ec_stripes_heap_cmp,
- ec_stripes_heap_set_backpointer);
+ min_heap_del(&c->ec_stripes_heap, m->heap_idx, &callbacks, &c->ec_stripes_heap);
mutex_unlock(&c->ec_stripes_heap_lock);
}
@@ -1007,14 +1084,15 @@ void bch2_stripes_heap_insert(struct bch_fs *c,
struct stripe *m, size_t idx)
{
mutex_lock(&c->ec_stripes_heap_lock);
- BUG_ON(heap_full(&c->ec_stripes_heap));
+ BUG_ON(min_heap_full(&c->ec_stripes_heap));
- heap_add(&c->ec_stripes_heap, ((struct ec_stripe_heap_entry) {
+ genradix_ptr(&c->stripes, idx)->heap_idx = c->ec_stripes_heap.nr;
+ min_heap_push(&c->ec_stripes_heap, &((struct ec_stripe_heap_entry) {
.idx = idx,
.blocks_nonempty = m->blocks_nonempty,
}),
- ec_stripes_heap_cmp,
- ec_stripes_heap_set_backpointer);
+ &callbacks,
+ &c->ec_stripes_heap);
heap_verify_backpointer(c, idx);
mutex_unlock(&c->ec_stripes_heap_lock);
@@ -1033,10 +1111,8 @@ void bch2_stripes_heap_update(struct bch_fs *c,
h->data[m->heap_idx].blocks_nonempty = m->blocks_nonempty;
i = m->heap_idx;
- heap_sift_up(h, i, ec_stripes_heap_cmp,
- ec_stripes_heap_set_backpointer);
- heap_sift_down(h, i, ec_stripes_heap_cmp,
- ec_stripes_heap_set_backpointer);
+ min_heap_sift_up(h, i, &callbacks, &c->ec_stripes_heap);
+ min_heap_sift_down(h, i, &callbacks, &c->ec_stripes_heap);
heap_verify_backpointer(c, idx);
@@ -1058,7 +1134,7 @@ static int ec_stripe_delete(struct btree_trans *trans, u64 idx)
int ret;
k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_stripes, POS(0, idx),
- BTREE_ITER_INTENT);
+ BTREE_ITER_intent);
ret = bkey_err(k);
if (ret)
goto err;
@@ -1100,7 +1176,7 @@ static void ec_stripe_delete_work(struct work_struct *work)
if (!idx)
break;
- int ret = bch2_trans_do(c, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
+ int ret = bch2_trans_commit_do(c, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
ec_stripe_delete(trans, idx));
bch_err_fn(c, ret);
if (ret)
@@ -1120,47 +1196,62 @@ void bch2_do_stripe_deletes(struct bch_fs *c)
/* stripe creation: */
static int ec_stripe_key_update(struct btree_trans *trans,
- struct bkey_i_stripe *new,
- bool create)
+ struct bkey_i_stripe *old,
+ struct bkey_i_stripe *new)
{
struct bch_fs *c = trans->c;
- struct btree_iter iter;
- struct bkey_s_c k;
- int ret;
+ bool create = !old;
- k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_stripes,
- new->k.p, BTREE_ITER_INTENT);
- ret = bkey_err(k);
+ struct btree_iter iter;
+ struct bkey_s_c k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_stripes,
+ new->k.p, BTREE_ITER_intent);
+ int ret = bkey_err(k);
if (ret)
goto err;
- if (k.k->type != (create ? KEY_TYPE_deleted : KEY_TYPE_stripe)) {
- bch2_fs_inconsistent(c, "error %s stripe: got existing key type %s",
- create ? "creating" : "updating",
- bch2_bkey_types[k.k->type]);
+ if (bch2_fs_inconsistent_on(k.k->type != (create ? KEY_TYPE_deleted : KEY_TYPE_stripe),
+ c, "error %s stripe: got existing key type %s",
+ create ? "creating" : "updating",
+ bch2_bkey_types[k.k->type])) {
ret = -EINVAL;
goto err;
}
if (k.k->type == KEY_TYPE_stripe) {
- const struct bch_stripe *old = bkey_s_c_to_stripe(k).v;
- unsigned i;
+ const struct bch_stripe *v = bkey_s_c_to_stripe(k).v;
- if (old->nr_blocks != new->v.nr_blocks) {
- bch_err(c, "error updating stripe: nr_blocks does not match");
- ret = -EINVAL;
- goto err;
- }
+ BUG_ON(old->v.nr_blocks != new->v.nr_blocks);
+ BUG_ON(old->v.nr_blocks != v->nr_blocks);
+
+ for (unsigned i = 0; i < new->v.nr_blocks; i++) {
+ unsigned sectors = stripe_blockcount_get(v, i);
- for (i = 0; i < new->v.nr_blocks; i++) {
- unsigned v = stripe_blockcount_get(old, i);
+ if (!bch2_extent_ptr_eq(old->v.ptrs[i], new->v.ptrs[i]) && sectors) {
+ struct printbuf buf = PRINTBUF;
- BUG_ON(v &&
- (old->ptrs[i].dev != new->v.ptrs[i].dev ||
- old->ptrs[i].gen != new->v.ptrs[i].gen ||
- old->ptrs[i].offset != new->v.ptrs[i].offset));
+ prt_printf(&buf, "stripe changed nonempty block %u", i);
+ prt_str(&buf, "\nold: ");
+ bch2_bkey_val_to_text(&buf, c, k);
+ prt_str(&buf, "\nnew: ");
+ bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&new->k_i));
+ bch2_fs_inconsistent(c, "%s", buf.buf);
+ printbuf_exit(&buf);
+ ret = -EINVAL;
+ goto err;
+ }
+
+ /*
+ * If the stripe ptr changed underneath us, it must have
+ * been dev_remove_stripes() -> * invalidate_stripe_to_dev()
+ */
+ if (!bch2_extent_ptr_eq(old->v.ptrs[i], v->ptrs[i])) {
+ BUG_ON(v->ptrs[i].dev != BCH_SB_MEMBER_INVALID);
+
+ if (bch2_extent_ptr_eq(old->v.ptrs[i], new->v.ptrs[i]))
+ new->v.ptrs[i].dev = BCH_SB_MEMBER_INVALID;
+ }
- stripe_blockcount_set(&new->v, i, v);
+ stripe_blockcount_set(&new->v, i, sectors);
}
}
@@ -1171,48 +1262,42 @@ err:
}
static int ec_stripe_update_extent(struct btree_trans *trans,
+ struct bch_dev *ca,
struct bpos bucket, u8 gen,
struct ec_stripe_buf *s,
- struct bpos *bp_pos)
+ struct bkey_s_c_backpointer bp,
+ struct bkey_buf *last_flushed)
{
struct bch_stripe *v = &bkey_i_to_stripe(&s->key)->v;
struct bch_fs *c = trans->c;
- struct bch_backpointer bp;
struct btree_iter iter;
struct bkey_s_c k;
const struct bch_extent_ptr *ptr_c;
- struct bch_extent_ptr *ptr, *ec_ptr = NULL;
+ struct bch_extent_ptr *ec_ptr = NULL;
struct bch_extent_stripe_ptr stripe_ptr;
struct bkey_i *n;
int ret, dev, block;
- ret = bch2_get_next_backpointer(trans, bucket, gen,
- bp_pos, &bp, BTREE_ITER_CACHED);
- if (ret)
- return ret;
- if (bpos_eq(*bp_pos, SPOS_MAX))
- return 0;
-
- if (bp.level) {
+ if (bp.v->level) {
struct printbuf buf = PRINTBUF;
struct btree_iter node_iter;
struct btree *b;
- b = bch2_backpointer_get_node(trans, &node_iter, *bp_pos, bp);
+ b = bch2_backpointer_get_node(trans, bp, &node_iter, last_flushed);
bch2_trans_iter_exit(trans, &node_iter);
if (!b)
return 0;
prt_printf(&buf, "found btree node in erasure coded bucket: b=%px\n", b);
- bch2_backpointer_to_text(&buf, &bp);
+ bch2_bkey_val_to_text(&buf, c, bp.s_c);
bch2_fs_inconsistent(c, "%s", buf.buf);
printbuf_exit(&buf);
return -EIO;
}
- k = bch2_backpointer_get_key(trans, &iter, *bp_pos, bp, BTREE_ITER_INTENT);
+ k = bch2_backpointer_get_key(trans, bp, &iter, BTREE_ITER_intent, last_flushed);
ret = bkey_err(k);
if (ret)
return ret;
@@ -1244,7 +1329,7 @@ static int ec_stripe_update_extent(struct btree_trans *trans,
bkey_reassemble(n, k);
- bch2_bkey_drop_ptrs(bkey_i_to_s(n), ptr, ptr->dev != dev);
+ bch2_bkey_drop_ptrs_noerror(bkey_i_to_s(n), ptr, ptr->dev != dev);
ec_ptr = bch2_bkey_has_device(bkey_i_to_s(n), dev);
BUG_ON(!ec_ptr);
@@ -1270,25 +1355,37 @@ static int ec_stripe_update_bucket(struct btree_trans *trans, struct ec_stripe_b
{
struct bch_fs *c = trans->c;
struct bch_stripe *v = &bkey_i_to_stripe(&s->key)->v;
- struct bch_extent_ptr bucket = v->ptrs[block];
- struct bpos bucket_pos = PTR_BUCKET_POS(c, &bucket);
- struct bpos bp_pos = POS_MIN;
+ struct bch_extent_ptr ptr = v->ptrs[block];
int ret = 0;
- while (1) {
- ret = commit_do(trans, NULL, NULL,
- BCH_TRANS_COMMIT_no_check_rw|
- BCH_TRANS_COMMIT_no_enospc,
- ec_stripe_update_extent(trans, bucket_pos, bucket.gen,
- s, &bp_pos));
- if (ret)
- break;
- if (bkey_eq(bp_pos, POS_MAX))
+ struct bch_dev *ca = bch2_dev_tryget(c, ptr.dev);
+ if (!ca)
+ return -EIO;
+
+ struct bpos bucket_pos = PTR_BUCKET_POS(ca, &ptr);
+
+ struct bkey_buf last_flushed;
+ bch2_bkey_buf_init(&last_flushed);
+ bkey_init(&last_flushed.k->k);
+
+ ret = for_each_btree_key_max_commit(trans, bp_iter, BTREE_ID_backpointers,
+ bucket_pos_to_bp_start(ca, bucket_pos),
+ bucket_pos_to_bp_end(ca, bucket_pos), 0, bp_k,
+ NULL, NULL,
+ BCH_TRANS_COMMIT_no_check_rw|
+ BCH_TRANS_COMMIT_no_enospc, ({
+ if (bkey_ge(bp_k.k->p, bucket_pos_to_bp(ca, bpos_nosnap_successor(bucket_pos), 0)))
break;
- bp_pos = bpos_nosnap_successor(bp_pos);
- }
+ if (bp_k.k->type != KEY_TYPE_backpointer)
+ continue;
+
+ ec_stripe_update_extent(trans, ca, bucket_pos, ptr.gen, s,
+ bkey_s_c_to_backpointer(bp_k), &last_flushed);
+ }));
+ bch2_bkey_buf_exit(&last_flushed, c);
+ bch2_dev_put(ca);
return ret;
}
@@ -1319,20 +1416,18 @@ static void zero_out_rest_of_ec_bucket(struct bch_fs *c,
unsigned block,
struct open_bucket *ob)
{
- struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
- unsigned offset = ca->mi.bucket_size - ob->sectors_free;
- int ret;
-
- if (!bch2_dev_get_ioref(ca, WRITE)) {
+ struct bch_dev *ca = bch2_dev_get_ioref(c, ob->dev, WRITE);
+ if (!ca) {
s->err = -BCH_ERR_erofs_no_writes;
return;
}
+ unsigned offset = ca->mi.bucket_size - ob->sectors_free;
memset(s->new_stripe.data[block] + (offset << 9),
0,
ob->sectors_free << 9);
- ret = blkdev_issue_zeroout(ca->disk_sb.bdev,
+ int ret = blkdev_issue_zeroout(ca->disk_sb.bdev,
ob->bucket * ca->mi.bucket_size + offset,
ob->sectors_free,
GFP_KERNEL, 0);
@@ -1414,12 +1509,14 @@ static void ec_stripe_create(struct ec_stripe_new *s)
goto err;
}
- ret = bch2_trans_do(c, &s->res, NULL,
- BCH_TRANS_COMMIT_no_check_rw|
- BCH_TRANS_COMMIT_no_enospc,
- ec_stripe_key_update(trans,
- bkey_i_to_stripe(&s->new_stripe.key),
- !s->have_existing_stripe));
+ ret = bch2_trans_commit_do(c, &s->res, NULL,
+ BCH_TRANS_COMMIT_no_check_rw|
+ BCH_TRANS_COMMIT_no_enospc,
+ ec_stripe_key_update(trans,
+ s->have_existing_stripe
+ ? bkey_i_to_stripe(&s->existing_stripe.key)
+ : NULL,
+ bkey_i_to_stripe(&s->new_stripe.key)));
bch_err_msg(c, ret, "creating stripe key");
if (ret) {
goto err;
@@ -1491,10 +1588,12 @@ void bch2_ec_do_stripe_creates(struct bch_fs *c)
bch2_write_ref_put(c, BCH_WRITE_REF_stripe_create);
}
-static void ec_stripe_set_pending(struct bch_fs *c, struct ec_stripe_head *h)
+static void ec_stripe_new_set_pending(struct bch_fs *c, struct ec_stripe_head *h)
{
struct ec_stripe_new *s = h->s;
+ lockdep_assert_held(&h->lock);
+
BUG_ON(!s->allocated && !s->err);
h->s = NULL;
@@ -1507,6 +1606,12 @@ static void ec_stripe_set_pending(struct bch_fs *c, struct ec_stripe_head *h)
ec_stripe_new_put(c, s, STRIPE_REF_io);
}
+static void ec_stripe_new_cancel(struct bch_fs *c, struct ec_stripe_head *h, int err)
+{
+ h->s->err = err;
+ ec_stripe_new_set_pending(c, h);
+}
+
void bch2_ec_bucket_cancel(struct bch_fs *c, struct open_bucket *ob)
{
struct ec_stripe_new *s = ob->ec;
@@ -1517,16 +1622,13 @@ void bch2_ec_bucket_cancel(struct bch_fs *c, struct open_bucket *ob)
void *bch2_writepoint_ec_buf(struct bch_fs *c, struct write_point *wp)
{
struct open_bucket *ob = ec_open_bucket(c, &wp->ptrs);
- struct bch_dev *ca;
- unsigned offset;
-
if (!ob)
return NULL;
BUG_ON(!ob->ec->new_stripe.data[ob->ec_idx]);
- ca = bch_dev_bkey_exists(c, ob->dev);
- offset = ca->mi.bucket_size - ob->sectors_free;
+ struct bch_dev *ca = ob_dev(c, ob);
+ unsigned offset = ca->mi.bucket_size - ob->sectors_free;
return ob->ec->new_stripe.data[ob->ec_idx] + (offset << 9);
}
@@ -1580,7 +1682,8 @@ static void ec_stripe_key_init(struct bch_fs *c,
struct bkey_i *k,
unsigned nr_data,
unsigned nr_parity,
- unsigned stripe_size)
+ unsigned stripe_size,
+ unsigned disk_label)
{
struct bkey_i_stripe *s = bkey_stripe_init(k);
unsigned u64s;
@@ -1591,7 +1694,7 @@ static void ec_stripe_key_init(struct bch_fs *c,
s->v.nr_redundant = nr_parity;
s->v.csum_granularity_bits = ilog2(c->opts.encoded_extent_max >> 9);
s->v.csum_type = BCH_CSUM_crc32c;
- s->v.pad = 0;
+ s->v.disk_label = disk_label;
while ((u64s = stripe_val_u64s(&s->v)) > BKEY_VAL_U64s_MAX) {
BUG_ON(1 << s->v.csum_granularity_bits >=
@@ -1603,7 +1706,7 @@ static void ec_stripe_key_init(struct bch_fs *c,
set_bkey_val_u64s(&s->k, u64s);
}
-static int ec_new_stripe_alloc(struct bch_fs *c, struct ec_stripe_head *h)
+static struct ec_stripe_new *ec_new_stripe_alloc(struct bch_fs *c, struct ec_stripe_head *h)
{
struct ec_stripe_new *s;
@@ -1611,7 +1714,7 @@ static int ec_new_stripe_alloc(struct bch_fs *c, struct ec_stripe_head *h)
s = kzalloc(sizeof(*s), GFP_KERNEL);
if (!s)
- return -BCH_ERR_ENOMEM_ec_new_stripe_alloc;
+ return NULL;
mutex_init(&s->lock);
closure_init(&s->iodone, NULL);
@@ -1624,40 +1727,29 @@ static int ec_new_stripe_alloc(struct bch_fs *c, struct ec_stripe_head *h)
s->nr_parity = h->redundancy;
ec_stripe_key_init(c, &s->new_stripe.key,
- s->nr_data, s->nr_parity, h->blocksize);
-
- h->s = s;
- return 0;
+ s->nr_data, s->nr_parity,
+ h->blocksize, h->disk_label);
+ return s;
}
-static struct ec_stripe_head *
-ec_new_stripe_head_alloc(struct bch_fs *c, unsigned target,
- unsigned algo, unsigned redundancy,
- enum bch_watermark watermark)
+static void ec_stripe_head_devs_update(struct bch_fs *c, struct ec_stripe_head *h)
{
- struct ec_stripe_head *h;
-
- h = kzalloc(sizeof(*h), GFP_KERNEL);
- if (!h)
- return NULL;
-
- mutex_init(&h->lock);
- BUG_ON(!mutex_trylock(&h->lock));
-
- h->target = target;
- h->algo = algo;
- h->redundancy = redundancy;
- h->watermark = watermark;
+ struct bch_devs_mask devs = h->devs;
rcu_read_lock();
- h->devs = target_rw_devs(c, BCH_DATA_user, target);
+ h->devs = target_rw_devs(c, BCH_DATA_user, h->disk_label
+ ? group_to_target(h->disk_label - 1)
+ : 0);
+ unsigned nr_devs = dev_mask_nr(&h->devs);
for_each_member_device_rcu(c, ca, &h->devs)
if (!ca->mi.durability)
__clear_bit(ca->dev_idx, h->devs.d);
+ unsigned nr_devs_with_durability = dev_mask_nr(&h->devs);
h->blocksize = pick_blocksize(c, &h->devs);
+ h->nr_active_devs = 0;
for_each_member_device_rcu(c, ca, &h->devs)
if (ca->mi.bucket_size == h->blocksize)
h->nr_active_devs++;
@@ -1668,9 +1760,50 @@ ec_new_stripe_head_alloc(struct bch_fs *c, unsigned target,
* If we only have redundancy + 1 devices, we're better off with just
* replication:
*/
- if (h->nr_active_devs < h->redundancy + 2)
- bch_err(c, "insufficient devices available to create stripe (have %u, need %u) - mismatched bucket sizes?",
- h->nr_active_devs, h->redundancy + 2);
+ h->insufficient_devs = h->nr_active_devs < h->redundancy + 2;
+
+ if (h->insufficient_devs) {
+ const char *err;
+
+ if (nr_devs < h->redundancy + 2)
+ err = NULL;
+ else if (nr_devs_with_durability < h->redundancy + 2)
+ err = "cannot use durability=0 devices";
+ else
+ err = "mismatched bucket sizes";
+
+ if (err)
+ bch_err(c, "insufficient devices available to create stripe (have %u, need %u): %s",
+ h->nr_active_devs, h->redundancy + 2, err);
+ }
+
+ struct bch_devs_mask devs_leaving;
+ bitmap_andnot(devs_leaving.d, devs.d, h->devs.d, BCH_SB_MEMBERS_MAX);
+
+ if (h->s && !h->s->allocated && dev_mask_nr(&devs_leaving))
+ ec_stripe_new_cancel(c, h, -EINTR);
+
+ h->rw_devs_change_count = c->rw_devs_change_count;
+}
+
+static struct ec_stripe_head *
+ec_new_stripe_head_alloc(struct bch_fs *c, unsigned disk_label,
+ unsigned algo, unsigned redundancy,
+ enum bch_watermark watermark)
+{
+ struct ec_stripe_head *h;
+
+ h = kzalloc(sizeof(*h), GFP_KERNEL);
+ if (!h)
+ return NULL;
+
+ mutex_init(&h->lock);
+ BUG_ON(!mutex_trylock(&h->lock));
+
+ h->disk_label = disk_label;
+ h->algo = algo;
+ h->redundancy = redundancy;
+ h->watermark = watermark;
list_add(&h->list, &c->ec_stripe_head_list);
return h;
@@ -1682,14 +1815,14 @@ void bch2_ec_stripe_head_put(struct bch_fs *c, struct ec_stripe_head *h)
h->s->allocated &&
bitmap_weight(h->s->blocks_allocated,
h->s->nr_data) == h->s->nr_data)
- ec_stripe_set_pending(c, h);
+ ec_stripe_new_set_pending(c, h);
mutex_unlock(&h->lock);
}
static struct ec_stripe_head *
__bch2_ec_stripe_head_get(struct btree_trans *trans,
- unsigned target,
+ unsigned disk_label,
unsigned algo,
unsigned redundancy,
enum bch_watermark watermark)
@@ -1707,63 +1840,84 @@ __bch2_ec_stripe_head_get(struct btree_trans *trans,
if (test_bit(BCH_FS_going_ro, &c->flags)) {
h = ERR_PTR(-BCH_ERR_erofs_no_writes);
- goto found;
+ goto err;
}
list_for_each_entry(h, &c->ec_stripe_head_list, list)
- if (h->target == target &&
+ if (h->disk_label == disk_label &&
h->algo == algo &&
h->redundancy == redundancy &&
h->watermark == watermark) {
ret = bch2_trans_mutex_lock(trans, &h->lock);
- if (ret)
+ if (ret) {
h = ERR_PTR(ret);
+ goto err;
+ }
goto found;
}
- h = ec_new_stripe_head_alloc(c, target, algo, redundancy, watermark);
+ h = ec_new_stripe_head_alloc(c, disk_label, algo, redundancy, watermark);
+ if (!h) {
+ h = ERR_PTR(-BCH_ERR_ENOMEM_stripe_head_alloc);
+ goto err;
+ }
found:
- if (!IS_ERR_OR_NULL(h) &&
- h->nr_active_devs < h->redundancy + 2) {
+ if (h->rw_devs_change_count != c->rw_devs_change_count)
+ ec_stripe_head_devs_update(c, h);
+
+ if (h->insufficient_devs) {
mutex_unlock(&h->lock);
h = NULL;
}
+err:
mutex_unlock(&c->ec_stripe_head_lock);
return h;
}
-static int new_stripe_alloc_buckets(struct btree_trans *trans, struct ec_stripe_head *h,
+static int new_stripe_alloc_buckets(struct btree_trans *trans,
+ struct ec_stripe_head *h, struct ec_stripe_new *s,
enum bch_watermark watermark, struct closure *cl)
{
struct bch_fs *c = trans->c;
struct bch_devs_mask devs = h->devs;
struct open_bucket *ob;
struct open_buckets buckets;
- struct bch_stripe *v = &bkey_i_to_stripe(&h->s->new_stripe.key)->v;
+ struct bch_stripe *v = &bkey_i_to_stripe(&s->new_stripe.key)->v;
unsigned i, j, nr_have_parity = 0, nr_have_data = 0;
bool have_cache = true;
int ret = 0;
- BUG_ON(v->nr_blocks != h->s->nr_data + h->s->nr_parity);
- BUG_ON(v->nr_redundant != h->s->nr_parity);
+ BUG_ON(v->nr_blocks != s->nr_data + s->nr_parity);
+ BUG_ON(v->nr_redundant != s->nr_parity);
+
+ /* * We bypass the sector allocator which normally does this: */
+ bitmap_and(devs.d, devs.d, c->rw_devs[BCH_DATA_user].d, BCH_SB_MEMBERS_MAX);
+
+ for_each_set_bit(i, s->blocks_gotten, v->nr_blocks) {
+ /*
+ * Note: we don't yet repair invalid blocks (failed/removed
+ * devices) when reusing stripes - we still need a codepath to
+ * walk backpointers and update all extents that point to that
+ * block when updating the stripe
+ */
+ if (v->ptrs[i].dev != BCH_SB_MEMBER_INVALID)
+ __clear_bit(v->ptrs[i].dev, devs.d);
- for_each_set_bit(i, h->s->blocks_gotten, v->nr_blocks) {
- __clear_bit(v->ptrs[i].dev, devs.d);
- if (i < h->s->nr_data)
+ if (i < s->nr_data)
nr_have_data++;
else
nr_have_parity++;
}
- BUG_ON(nr_have_data > h->s->nr_data);
- BUG_ON(nr_have_parity > h->s->nr_parity);
+ BUG_ON(nr_have_data > s->nr_data);
+ BUG_ON(nr_have_parity > s->nr_parity);
buckets.nr = 0;
- if (nr_have_parity < h->s->nr_parity) {
+ if (nr_have_parity < s->nr_parity) {
ret = bch2_bucket_alloc_set_trans(trans, &buckets,
&h->parity_stripe,
&devs,
- h->s->nr_parity,
+ s->nr_parity,
&nr_have_parity,
&have_cache, 0,
BCH_DATA_parity,
@@ -1771,14 +1925,14 @@ static int new_stripe_alloc_buckets(struct btree_trans *trans, struct ec_stripe_
cl);
open_bucket_for_each(c, &buckets, ob, i) {
- j = find_next_zero_bit(h->s->blocks_gotten,
- h->s->nr_data + h->s->nr_parity,
- h->s->nr_data);
- BUG_ON(j >= h->s->nr_data + h->s->nr_parity);
+ j = find_next_zero_bit(s->blocks_gotten,
+ s->nr_data + s->nr_parity,
+ s->nr_data);
+ BUG_ON(j >= s->nr_data + s->nr_parity);
- h->s->blocks[j] = buckets.v[i];
+ s->blocks[j] = buckets.v[i];
v->ptrs[j] = bch2_ob_ptr(c, ob);
- __set_bit(j, h->s->blocks_gotten);
+ __set_bit(j, s->blocks_gotten);
}
if (ret)
@@ -1786,11 +1940,11 @@ static int new_stripe_alloc_buckets(struct btree_trans *trans, struct ec_stripe_
}
buckets.nr = 0;
- if (nr_have_data < h->s->nr_data) {
+ if (nr_have_data < s->nr_data) {
ret = bch2_bucket_alloc_set_trans(trans, &buckets,
&h->block_stripe,
&devs,
- h->s->nr_data,
+ s->nr_data,
&nr_have_data,
&have_cache, 0,
BCH_DATA_user,
@@ -1798,13 +1952,13 @@ static int new_stripe_alloc_buckets(struct btree_trans *trans, struct ec_stripe_
cl);
open_bucket_for_each(c, &buckets, ob, i) {
- j = find_next_zero_bit(h->s->blocks_gotten,
- h->s->nr_data, 0);
- BUG_ON(j >= h->s->nr_data);
+ j = find_next_zero_bit(s->blocks_gotten,
+ s->nr_data, 0);
+ BUG_ON(j >= s->nr_data);
- h->s->blocks[j] = buckets.v[i];
+ s->blocks[j] = buckets.v[i];
v->ptrs[j] = bch2_ob_ptr(c, ob);
- __set_bit(j, h->s->blocks_gotten);
+ __set_bit(j, s->blocks_gotten);
}
if (ret)
@@ -1814,7 +1968,6 @@ static int new_stripe_alloc_buckets(struct btree_trans *trans, struct ec_stripe_
return 0;
}
-/* XXX: doesn't obey target: */
static s64 get_existing_stripe(struct bch_fs *c,
struct ec_stripe_head *head)
{
@@ -1828,7 +1981,7 @@ static s64 get_existing_stripe(struct bch_fs *c,
return -1;
mutex_lock(&c->ec_stripes_heap_lock);
- for (heap_idx = 0; heap_idx < h->used; heap_idx++) {
+ for (heap_idx = 0; heap_idx < h->nr; heap_idx++) {
/* No blocks worth reusing, stripe will just be deleted: */
if (!h->data[heap_idx].blocks_nonempty)
continue;
@@ -1837,7 +1990,8 @@ static s64 get_existing_stripe(struct bch_fs *c,
m = genradix_ptr(&c->stripes, stripe_idx);
- if (m->algorithm == head->algo &&
+ if (m->disk_label == head->disk_label &&
+ m->algorithm == head->algo &&
m->nr_redundant == head->redundancy &&
m->sectors == head->blocksize &&
m->blocks_nonempty < m->nr_blocks - m->nr_redundant &&
@@ -1850,73 +2004,78 @@ static s64 get_existing_stripe(struct bch_fs *c,
return ret;
}
-static int __bch2_ec_stripe_head_reuse(struct btree_trans *trans, struct ec_stripe_head *h)
+static int init_new_stripe_from_existing(struct bch_fs *c, struct ec_stripe_new *s)
{
- struct bch_fs *c = trans->c;
- struct bch_stripe *new_v = &bkey_i_to_stripe(&h->s->new_stripe.key)->v;
- struct bch_stripe *existing_v;
+ struct bch_stripe *new_v = &bkey_i_to_stripe(&s->new_stripe.key)->v;
+ struct bch_stripe *existing_v = &bkey_i_to_stripe(&s->existing_stripe.key)->v;
unsigned i;
- s64 idx;
- int ret;
- /*
- * If we can't allocate a new stripe, and there's no stripes with empty
- * blocks for us to reuse, that means we have to wait on copygc:
- */
- idx = get_existing_stripe(c, h);
- if (idx < 0)
- return -BCH_ERR_stripe_alloc_blocked;
-
- ret = get_stripe_key_trans(trans, idx, &h->s->existing_stripe);
- if (ret) {
- bch2_stripe_close(c, h->s);
- if (!bch2_err_matches(ret, BCH_ERR_transaction_restart))
- bch2_fs_fatal_error(c, "error reading stripe key: %s", bch2_err_str(ret));
- return ret;
- }
-
- existing_v = &bkey_i_to_stripe(&h->s->existing_stripe.key)->v;
-
- BUG_ON(existing_v->nr_redundant != h->s->nr_parity);
- h->s->nr_data = existing_v->nr_blocks -
+ BUG_ON(existing_v->nr_redundant != s->nr_parity);
+ s->nr_data = existing_v->nr_blocks -
existing_v->nr_redundant;
- ret = ec_stripe_buf_init(&h->s->existing_stripe, 0, h->blocksize);
+ int ret = ec_stripe_buf_init(&s->existing_stripe, 0, le16_to_cpu(existing_v->sectors));
if (ret) {
- bch2_stripe_close(c, h->s);
+ bch2_stripe_close(c, s);
return ret;
}
- BUG_ON(h->s->existing_stripe.size != h->blocksize);
- BUG_ON(h->s->existing_stripe.size != le16_to_cpu(existing_v->sectors));
+ BUG_ON(s->existing_stripe.size != le16_to_cpu(existing_v->sectors));
/*
* Free buckets we initially allocated - they might conflict with
* blocks from the stripe we're reusing:
*/
- for_each_set_bit(i, h->s->blocks_gotten, new_v->nr_blocks) {
- bch2_open_bucket_put(c, c->open_buckets + h->s->blocks[i]);
- h->s->blocks[i] = 0;
+ for_each_set_bit(i, s->blocks_gotten, new_v->nr_blocks) {
+ bch2_open_bucket_put(c, c->open_buckets + s->blocks[i]);
+ s->blocks[i] = 0;
}
- memset(h->s->blocks_gotten, 0, sizeof(h->s->blocks_gotten));
- memset(h->s->blocks_allocated, 0, sizeof(h->s->blocks_allocated));
+ memset(s->blocks_gotten, 0, sizeof(s->blocks_gotten));
+ memset(s->blocks_allocated, 0, sizeof(s->blocks_allocated));
- for (i = 0; i < existing_v->nr_blocks; i++) {
+ for (unsigned i = 0; i < existing_v->nr_blocks; i++) {
if (stripe_blockcount_get(existing_v, i)) {
- __set_bit(i, h->s->blocks_gotten);
- __set_bit(i, h->s->blocks_allocated);
+ __set_bit(i, s->blocks_gotten);
+ __set_bit(i, s->blocks_allocated);
}
- ec_block_io(c, &h->s->existing_stripe, READ, i, &h->s->iodone);
+ ec_block_io(c, &s->existing_stripe, READ, i, &s->iodone);
}
- bkey_copy(&h->s->new_stripe.key, &h->s->existing_stripe.key);
- h->s->have_existing_stripe = true;
+ bkey_copy(&s->new_stripe.key, &s->existing_stripe.key);
+ s->have_existing_stripe = true;
return 0;
}
-static int __bch2_ec_stripe_head_reserve(struct btree_trans *trans, struct ec_stripe_head *h)
+static int __bch2_ec_stripe_head_reuse(struct btree_trans *trans, struct ec_stripe_head *h,
+ struct ec_stripe_new *s)
+{
+ struct bch_fs *c = trans->c;
+ s64 idx;
+ int ret;
+
+ /*
+ * If we can't allocate a new stripe, and there's no stripes with empty
+ * blocks for us to reuse, that means we have to wait on copygc:
+ */
+ idx = get_existing_stripe(c, h);
+ if (idx < 0)
+ return -BCH_ERR_stripe_alloc_blocked;
+
+ ret = get_stripe_key_trans(trans, idx, &s->existing_stripe);
+ bch2_fs_fatal_err_on(ret && !bch2_err_matches(ret, BCH_ERR_transaction_restart), c,
+ "reading stripe key: %s", bch2_err_str(ret));
+ if (ret) {
+ bch2_stripe_close(c, s);
+ return ret;
+ }
+
+ return init_new_stripe_from_existing(c, s);
+}
+
+static int __bch2_ec_stripe_head_reserve(struct btree_trans *trans, struct ec_stripe_head *h,
+ struct ec_stripe_new *s)
{
struct bch_fs *c = trans->c;
struct btree_iter iter;
@@ -1925,17 +2084,21 @@ static int __bch2_ec_stripe_head_reserve(struct btree_trans *trans, struct ec_st
struct bpos start_pos = bpos_max(min_pos, POS(0, c->ec_stripe_hint));
int ret;
- if (!h->s->res.sectors) {
- ret = bch2_disk_reservation_get(c, &h->s->res,
+ if (!s->res.sectors) {
+ ret = bch2_disk_reservation_get(c, &s->res,
h->blocksize,
- h->s->nr_parity,
+ s->nr_parity,
BCH_DISK_RESERVATION_NOFAIL);
if (ret)
return ret;
}
+ /*
+ * Allocate stripe slot
+ * XXX: we're going to need a bitrange btree of free stripes
+ */
for_each_btree_key_norestart(trans, iter, BTREE_ID_stripes, start_pos,
- BTREE_ITER_SLOTS|BTREE_ITER_INTENT, k, ret) {
+ BTREE_ITER_slots|BTREE_ITER_intent, k, ret) {
if (bkey_gt(k.k->p, POS(0, U32_MAX))) {
if (start_pos.offset) {
start_pos = min_pos;
@@ -1948,7 +2111,7 @@ static int __bch2_ec_stripe_head_reserve(struct btree_trans *trans, struct ec_st
}
if (bkey_deleted(k.k) &&
- bch2_try_open_stripe(c, h->s, k.k->p.offset))
+ bch2_try_open_stripe(c, s, k.k->p.offset))
break;
}
@@ -1959,16 +2122,16 @@ static int __bch2_ec_stripe_head_reserve(struct btree_trans *trans, struct ec_st
ret = ec_stripe_mem_alloc(trans, &iter);
if (ret) {
- bch2_stripe_close(c, h->s);
+ bch2_stripe_close(c, s);
goto err;
}
- h->s->new_stripe.key.k.p = iter.pos;
+ s->new_stripe.key.k.p = iter.pos;
out:
bch2_trans_iter_exit(trans, &iter);
return ret;
err:
- bch2_disk_reservation_put(c, &h->s->res);
+ bch2_disk_reservation_put(c, &s->res);
goto out;
}
@@ -1982,29 +2145,44 @@ struct ec_stripe_head *bch2_ec_stripe_head_get(struct btree_trans *trans,
struct bch_fs *c = trans->c;
struct ec_stripe_head *h;
bool waiting = false;
+ unsigned disk_label = 0;
+ struct target t = target_decode(target);
int ret;
- h = __bch2_ec_stripe_head_get(trans, target, algo, redundancy, watermark);
+ if (t.type == TARGET_GROUP) {
+ if (t.group > U8_MAX) {
+ bch_err(c, "cannot create a stripe when disk_label > U8_MAX");
+ return NULL;
+ }
+ disk_label = t.group + 1; /* 0 == no label */
+ }
+
+ h = __bch2_ec_stripe_head_get(trans, disk_label, algo, redundancy, watermark);
if (IS_ERR_OR_NULL(h))
return h;
if (!h->s) {
- ret = ec_new_stripe_alloc(c, h);
- if (ret) {
+ h->s = ec_new_stripe_alloc(c, h);
+ if (!h->s) {
+ ret = -BCH_ERR_ENOMEM_ec_new_stripe_alloc;
bch_err(c, "failed to allocate new stripe");
goto err;
}
+
+ h->nr_created++;
}
- if (h->s->allocated)
+ struct ec_stripe_new *s = h->s;
+
+ if (s->allocated)
goto allocated;
- if (h->s->have_existing_stripe)
+ if (s->have_existing_stripe)
goto alloc_existing;
/* First, try to allocate a full stripe: */
- ret = new_stripe_alloc_buckets(trans, h, BCH_WATERMARK_stripe, NULL) ?:
- __bch2_ec_stripe_head_reserve(trans, h);
+ ret = new_stripe_alloc_buckets(trans, h, s, BCH_WATERMARK_stripe, NULL) ?:
+ __bch2_ec_stripe_head_reserve(trans, h, s);
if (!ret)
goto allocate_buf;
if (bch2_err_matches(ret, BCH_ERR_transaction_restart) ||
@@ -2016,15 +2194,15 @@ struct ec_stripe_head *bch2_ec_stripe_head_get(struct btree_trans *trans,
* existing stripe:
*/
while (1) {
- ret = __bch2_ec_stripe_head_reuse(trans, h);
+ ret = __bch2_ec_stripe_head_reuse(trans, h, s);
if (!ret)
break;
if (waiting || !cl || ret != -BCH_ERR_stripe_alloc_blocked)
goto err;
if (watermark == BCH_WATERMARK_copygc) {
- ret = new_stripe_alloc_buckets(trans, h, watermark, NULL) ?:
- __bch2_ec_stripe_head_reserve(trans, h);
+ ret = new_stripe_alloc_buckets(trans, h, s, watermark, NULL) ?:
+ __bch2_ec_stripe_head_reserve(trans, h, s);
if (ret)
goto err;
goto allocate_buf;
@@ -2042,19 +2220,19 @@ alloc_existing:
* Retry allocating buckets, with the watermark for this
* particular write:
*/
- ret = new_stripe_alloc_buckets(trans, h, watermark, cl);
+ ret = new_stripe_alloc_buckets(trans, h, s, watermark, cl);
if (ret)
goto err;
allocate_buf:
- ret = ec_stripe_buf_init(&h->s->new_stripe, 0, h->blocksize);
+ ret = ec_stripe_buf_init(&s->new_stripe, 0, h->blocksize);
if (ret)
goto err;
- h->s->allocated = true;
+ s->allocated = true;
allocated:
- BUG_ON(!h->s->idx);
- BUG_ON(!h->s->new_stripe.data[0]);
+ BUG_ON(!s->idx);
+ BUG_ON(!s->new_stripe.data[0]);
BUG_ON(trans->restarted);
return h;
err:
@@ -2062,6 +2240,73 @@ err:
return ERR_PTR(ret);
}
+/* device removal */
+
+static int bch2_invalidate_stripe_to_dev(struct btree_trans *trans, struct bkey_s_c k_a)
+{
+ struct bch_alloc_v4 a_convert;
+ const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k_a, &a_convert);
+
+ if (!a->stripe)
+ return 0;
+
+ if (a->stripe_sectors) {
+ bch_err(trans->c, "trying to invalidate device in stripe when bucket has stripe data");
+ return -BCH_ERR_invalidate_stripe_to_dev;
+ }
+
+ struct btree_iter iter;
+ struct bkey_i_stripe *s =
+ bch2_bkey_get_mut_typed(trans, &iter, BTREE_ID_stripes, POS(0, a->stripe),
+ BTREE_ITER_slots, stripe);
+ int ret = PTR_ERR_OR_ZERO(s);
+ if (ret)
+ return ret;
+
+ struct disk_accounting_pos acc = {
+ .type = BCH_DISK_ACCOUNTING_replicas,
+ };
+
+ s64 sectors = 0;
+ for (unsigned i = 0; i < s->v.nr_blocks; i++)
+ sectors -= stripe_blockcount_get(&s->v, i);
+
+ bch2_bkey_to_replicas(&acc.replicas, bkey_i_to_s_c(&s->k_i));
+ acc.replicas.data_type = BCH_DATA_user;
+ ret = bch2_disk_accounting_mod(trans, &acc, &sectors, 1, false);
+ if (ret)
+ goto err;
+
+ struct bkey_ptrs ptrs = bch2_bkey_ptrs(bkey_i_to_s(&s->k_i));
+ bkey_for_each_ptr(ptrs, ptr)
+ if (ptr->dev == k_a.k->p.inode)
+ ptr->dev = BCH_SB_MEMBER_INVALID;
+
+ sectors = -sectors;
+
+ bch2_bkey_to_replicas(&acc.replicas, bkey_i_to_s_c(&s->k_i));
+ acc.replicas.data_type = BCH_DATA_user;
+ ret = bch2_disk_accounting_mod(trans, &acc, &sectors, 1, false);
+ if (ret)
+ goto err;
+err:
+ bch2_trans_iter_exit(trans, &iter);
+ return ret;
+}
+
+int bch2_dev_remove_stripes(struct bch_fs *c, unsigned dev_idx)
+{
+ return bch2_trans_run(c,
+ for_each_btree_key_max_commit(trans, iter,
+ BTREE_ID_alloc, POS(dev_idx, 0), POS(dev_idx, U64_MAX),
+ BTREE_ITER_intent, k,
+ NULL, NULL, 0, ({
+ bch2_invalidate_stripe_to_dev(trans, k);
+ })));
+}
+
+/* startup/shutdown */
+
static void __bch2_ec_stop(struct bch_fs *c, struct bch_dev *ca)
{
struct ec_stripe_head *h;
@@ -2087,8 +2332,7 @@ static void __bch2_ec_stop(struct bch_fs *c, struct bch_dev *ca)
}
goto unlock;
found:
- h->s->err = -BCH_ERR_erofs_no_writes;
- ec_stripe_set_pending(c, h);
+ ec_stripe_new_cancel(c, h, -BCH_ERR_erofs_no_writes);
unlock:
mutex_unlock(&h->lock);
}
@@ -2125,7 +2369,7 @@ int bch2_stripes_read(struct bch_fs *c)
{
int ret = bch2_trans_run(c,
for_each_btree_key(trans, iter, BTREE_ID_stripes, POS_MIN,
- BTREE_ITER_PREFETCH, k, ({
+ BTREE_ITER_prefetch, k, ({
if (k.k->type != KEY_TYPE_stripe)
continue;
@@ -2133,17 +2377,9 @@ int bch2_stripes_read(struct bch_fs *c)
if (ret)
break;
- const struct bch_stripe *s = bkey_s_c_to_stripe(k).v;
-
struct stripe *m = genradix_ptr(&c->stripes, k.k->p.offset);
- m->sectors = le16_to_cpu(s->sectors);
- m->algorithm = s->algorithm;
- m->nr_blocks = s->nr_blocks;
- m->nr_redundant = s->nr_redundant;
- m->blocks_nonempty = 0;
- for (unsigned i = 0; i < s->nr_blocks; i++)
- m->blocks_nonempty += !!stripe_blockcount_get(s, i);
+ stripe_to_mem(m, bkey_s_c_to_stripe(k).v);
bch2_stripes_heap_insert(c, m, k.k->p.offset);
0;
@@ -2159,7 +2395,7 @@ void bch2_stripes_heap_to_text(struct printbuf *out, struct bch_fs *c)
size_t i;
mutex_lock(&c->ec_stripes_heap_lock);
- for (i = 0; i < min_t(size_t, h->used, 50); i++) {
+ for (i = 0; i < min_t(size_t, h->nr, 50); i++) {
m = genradix_ptr(&c->stripes, h->data[i].idx);
prt_printf(out, "%zu %u/%u+%u", h->data[i].idx,
@@ -2173,6 +2409,25 @@ void bch2_stripes_heap_to_text(struct printbuf *out, struct bch_fs *c)
mutex_unlock(&c->ec_stripes_heap_lock);
}
+static void bch2_new_stripe_to_text(struct printbuf *out, struct bch_fs *c,
+ struct ec_stripe_new *s)
+{
+ prt_printf(out, "\tidx %llu blocks %u+%u allocated %u ref %u %u %s obs",
+ s->idx, s->nr_data, s->nr_parity,
+ bitmap_weight(s->blocks_allocated, s->nr_data),
+ atomic_read(&s->ref[STRIPE_REF_io]),
+ atomic_read(&s->ref[STRIPE_REF_stripe]),
+ bch2_watermarks[s->h->watermark]);
+
+ struct bch_stripe *v = &bkey_i_to_stripe(&s->new_stripe.key)->v;
+ unsigned i;
+ for_each_set_bit(i, s->blocks_gotten, v->nr_blocks)
+ prt_printf(out, " %u", s->blocks[i]);
+ prt_newline(out);
+ bch2_bkey_val_to_text(out, c, bkey_i_to_s_c(&s->new_stripe.key));
+ prt_newline(out);
+}
+
void bch2_new_stripes_to_text(struct printbuf *out, struct bch_fs *c)
{
struct ec_stripe_head *h;
@@ -2180,28 +2435,21 @@ void bch2_new_stripes_to_text(struct printbuf *out, struct bch_fs *c)
mutex_lock(&c->ec_stripe_head_lock);
list_for_each_entry(h, &c->ec_stripe_head_list, list) {
- prt_printf(out, "target %u algo %u redundancy %u %s:\n",
- h->target, h->algo, h->redundancy,
- bch2_watermarks[h->watermark]);
+ prt_printf(out, "disk label %u algo %u redundancy %u %s nr created %llu:\n",
+ h->disk_label, h->algo, h->redundancy,
+ bch2_watermarks[h->watermark],
+ h->nr_created);
if (h->s)
- prt_printf(out, "\tidx %llu blocks %u+%u allocated %u\n",
- h->s->idx, h->s->nr_data, h->s->nr_parity,
- bitmap_weight(h->s->blocks_allocated,
- h->s->nr_data));
+ bch2_new_stripe_to_text(out, c, h->s);
}
mutex_unlock(&c->ec_stripe_head_lock);
prt_printf(out, "in flight:\n");
mutex_lock(&c->ec_stripe_new_lock);
- list_for_each_entry(s, &c->ec_stripe_new_list, list) {
- prt_printf(out, "\tidx %llu blocks %u+%u ref %u %u %s\n",
- s->idx, s->nr_data, s->nr_parity,
- atomic_read(&s->ref[STRIPE_REF_io]),
- atomic_read(&s->ref[STRIPE_REF_stripe]),
- bch2_watermarks[s->h->watermark]);
- }
+ list_for_each_entry(s, &c->ec_stripe_new_list, list)
+ bch2_new_stripe_to_text(out, c, s);
mutex_unlock(&c->ec_stripe_new_lock);
}
@@ -2212,11 +2460,9 @@ void bch2_fs_ec_exit(struct bch_fs *c)
while (1) {
mutex_lock(&c->ec_stripe_head_lock);
- h = list_first_entry_or_null(&c->ec_stripe_head_list,
- struct ec_stripe_head, list);
- if (h)
- list_del(&h->list);
+ h = list_pop_entry(&c->ec_stripe_head_list, struct ec_stripe_head, list);
mutex_unlock(&c->ec_stripe_head_lock);
+
if (!h)
break;
generated by cgit (git 2.34.1) at 2025-07-25 18:34:59 +0000