#include "btree_gc.h"
#include "btree_update.h"
#include "buckets.h"
+#include "buckets_waiting_for_journal.h"
#include "ec.h"
#include "error.h"
#include "inode.h"
}
}
-/*
- * Clear journal_seq_valid for buckets for which it's not needed, to prevent
- * wraparound:
- */
-void bch2_bucket_seq_cleanup(struct bch_fs *c)
-{
- u64 journal_seq = atomic64_read(&c->journal.seq);
- u16 last_seq_ondisk = c->journal.flushed_seq_ondisk;
- struct bch_dev *ca;
- struct bucket_array *buckets;
- struct bucket *g;
- struct bucket_mark m;
- unsigned i;
-
- if (journal_seq - c->last_bucket_seq_cleanup <
- (1U << (BUCKET_JOURNAL_SEQ_BITS - 2)))
- return;
-
- c->last_bucket_seq_cleanup = journal_seq;
-
- for_each_member_device(ca, c, i) {
- down_read(&ca->bucket_lock);
- buckets = bucket_array(ca);
-
- for_each_bucket(g, buckets) {
- bucket_cmpxchg(g, m, ({
- if (!m.journal_seq_valid ||
- bucket_needs_journal_commit(m, last_seq_ondisk))
- break;
-
- m.journal_seq_valid = 0;
- }));
- }
- up_read(&ca->bucket_lock);
- }
-}
-
void bch2_fs_usage_initialize(struct bch_fs *c)
{
struct bch_fs_usage *usage;
return ret;
}
-static inline int is_unavailable_bucket(struct bucket_mark m)
+void bch2_dev_usage_init(struct bch_dev *ca)
{
- return !is_available_bucket(m);
+ ca->usage_base->d[BCH_DATA_free].buckets = ca->mi.nbuckets - ca->mi.first_bucket;
}
static inline int bucket_sectors_fragmented(struct bch_dev *ca,
- struct bucket_mark m)
+ struct bch_alloc_v4 a)
{
- return bucket_sectors_used(m)
- ? max(0, (int) ca->mi.bucket_size - (int) bucket_sectors_used(m))
+ return a.dirty_sectors
+ ? max(0, (int) ca->mi.bucket_size - (int) a.dirty_sectors)
: 0;
}
-static inline int is_stripe_data_bucket(struct bucket_mark m)
-{
- return m.stripe && m.data_type != BCH_DATA_parity;
-}
-
-static inline enum bch_data_type bucket_type(struct bucket_mark m)
-{
- return m.cached_sectors && !m.dirty_sectors
- ? BCH_DATA_cached
- : m.data_type;
-}
-
-static inline void account_bucket(struct bch_fs_usage *fs_usage,
- struct bch_dev_usage *dev_usage,
- enum bch_data_type type,
- int nr, s64 size)
-{
- if (type == BCH_DATA_sb || type == BCH_DATA_journal)
- fs_usage->hidden += size;
-
- dev_usage->d[type].buckets += nr;
-}
-
static void bch2_dev_usage_update(struct bch_fs *c, struct bch_dev *ca,
- struct bucket_mark old, struct bucket_mark new,
+ struct bch_alloc_v4 old,
+ struct bch_alloc_v4 new,
u64 journal_seq, bool gc)
{
struct bch_fs_usage *fs_usage;
preempt_disable();
fs_usage = fs_usage_ptr(c, journal_seq, gc);
- u = dev_usage_ptr(ca, journal_seq, gc);
- if (bucket_type(old))
- account_bucket(fs_usage, u, bucket_type(old),
- -1, -ca->mi.bucket_size);
+ if (data_type_is_hidden(old.data_type))
+ fs_usage->hidden -= ca->mi.bucket_size;
+ if (data_type_is_hidden(new.data_type))
+ fs_usage->hidden += ca->mi.bucket_size;
+
+ u = dev_usage_ptr(ca, journal_seq, gc);
- if (bucket_type(new))
- account_bucket(fs_usage, u, bucket_type(new),
- 1, ca->mi.bucket_size);
+ u->d[old.data_type].buckets--;
+ u->d[new.data_type].buckets++;
- u->buckets_ec += (int) new.stripe - (int) old.stripe;
- u->buckets_unavailable +=
- is_unavailable_bucket(new) - is_unavailable_bucket(old);
+ u->buckets_ec -= (int) !!old.stripe;
+ u->buckets_ec += (int) !!new.stripe;
u->d[old.data_type].sectors -= old.dirty_sectors;
u->d[new.data_type].sectors += new.dirty_sectors;
- u->d[BCH_DATA_cached].sectors +=
- (int) new.cached_sectors - (int) old.cached_sectors;
+
+ u->d[BCH_DATA_cached].sectors += new.cached_sectors;
+ u->d[BCH_DATA_cached].sectors -= old.cached_sectors;
u->d[old.data_type].fragmented -= bucket_sectors_fragmented(ca, old);
u->d[new.data_type].fragmented += bucket_sectors_fragmented(ca, new);
preempt_enable();
+}
- if (!is_available_bucket(old) && is_available_bucket(new))
- bch2_wake_allocator(ca);
+static void bch2_dev_usage_update_m(struct bch_fs *c, struct bch_dev *ca,
+ struct bucket old, struct bucket new,
+ u64 journal_seq, bool gc)
+{
+ struct bch_alloc_v4 old_a = {
+ .gen = old.gen,
+ .data_type = old.data_type,
+ .dirty_sectors = old.dirty_sectors,
+ .cached_sectors = old.cached_sectors,
+ .stripe = old.stripe,
+ };
+ struct bch_alloc_v4 new_a = {
+ .gen = new.gen,
+ .data_type = new.data_type,
+ .dirty_sectors = new.dirty_sectors,
+ .cached_sectors = new.cached_sectors,
+ .stripe = new.stripe,
+ };
+
+ bch2_dev_usage_update(c, ca, old_a, new_a, journal_seq, gc);
}
static inline int __update_replicas(struct bch_fs *c,
{
struct bch_fs_usage __percpu *fs_usage;
int idx, ret = 0;
- char buf[200];
+ struct printbuf buf = PRINTBUF;
percpu_down_read(&c->mark_lock);
+ buf.atomic++;
idx = bch2_replicas_entry_idx(c, r);
if (idx < 0 &&
- (test_bit(BCH_FS_REBUILD_REPLICAS, &c->flags) ||
- fsck_err(c, "no replicas entry\n"
- " while marking %s",
- (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf)))) {
+ fsck_err(c, "no replicas entry\n"
+ " while marking %s",
+ (bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
percpu_up_read(&c->mark_lock);
ret = bch2_mark_replicas(c, r);
- if (ret)
- return ret;
-
percpu_down_read(&c->mark_lock);
+
+ if (ret)
+ goto err;
idx = bch2_replicas_entry_idx(c, r);
}
if (idx < 0) {
err:
fsck_err:
percpu_up_read(&c->mark_lock);
+ printbuf_exit(&buf);
return ret;
}
update_replicas_list(trans, &r.e, sectors);
}
-void bch2_mark_alloc_bucket(struct bch_fs *c, struct bch_dev *ca,
- size_t b, bool owned_by_allocator)
-{
- struct bucket *g = bucket(ca, b);
- struct bucket_mark old, new;
-
- old = bucket_cmpxchg(g, new, ({
- new.owned_by_allocator = owned_by_allocator;
- }));
-
- BUG_ON(owned_by_allocator == old.owned_by_allocator);
-}
-
-static int bch2_mark_alloc(struct btree_trans *trans,
- struct bkey_s_c old, struct bkey_s_c new,
- unsigned flags)
+int bch2_mark_alloc(struct btree_trans *trans,
+ struct bkey_s_c old, struct bkey_s_c new,
+ unsigned flags)
{
bool gc = flags & BTREE_TRIGGER_GC;
u64 journal_seq = trans->journal_res.seq;
struct bch_fs *c = trans->c;
- struct bkey_alloc_unpacked old_u = bch2_alloc_unpack(old);
- struct bkey_alloc_unpacked new_u = bch2_alloc_unpack(new);
+ struct bch_alloc_v4 old_a, new_a;
struct bch_dev *ca;
- struct bucket *g;
- struct bucket_mark old_m, m;
int ret = 0;
/*
!(flags & BTREE_TRIGGER_BUCKET_INVALIDATE))
return 0;
+ if (bch2_trans_inconsistent_on(!bch2_dev_bucket_exists(c, new.k->p), trans,
+ "alloc key for invalid device or bucket"))
+ return -EIO;
+
+ ca = bch_dev_bkey_exists(c, new.k->p.inode);
+
+ bch2_alloc_to_v4(old, &old_a);
+ bch2_alloc_to_v4(new, &new_a);
+
if ((flags & BTREE_TRIGGER_INSERT) &&
- !old_u.data_type != !new_u.data_type &&
- new.k->type == KEY_TYPE_alloc_v3) {
- struct bch_alloc_v3 *v = (struct bch_alloc_v3 *) new.v;
+ data_type_is_empty(old_a.data_type) !=
+ data_type_is_empty(new_a.data_type) &&
+ new.k->type == KEY_TYPE_alloc_v4) {
+ struct bch_alloc_v4 *v = (struct bch_alloc_v4 *) new.v;
BUG_ON(!journal_seq);
* before the bucket became empty again, then the we don't have
* to wait on a journal flush before we can reuse the bucket:
*/
- v->journal_seq = !new_u.data_type &&
- bch2_journal_noflush_seq(&c->journal, journal_seq)
- ? 0 : cpu_to_le64(journal_seq);
+ new_a.journal_seq = data_type_is_empty(new_a.data_type) &&
+ (journal_seq == v->journal_seq ||
+ bch2_journal_noflush_seq(&c->journal, v->journal_seq))
+ ? 0 : journal_seq;
+ v->journal_seq = new_a.journal_seq;
}
- ca = bch_dev_bkey_exists(c, new.k->p.inode);
+ if (!data_type_is_empty(old_a.data_type) &&
+ data_type_is_empty(new_a.data_type) &&
+ new_a.journal_seq) {
+ ret = bch2_set_bucket_needs_journal_commit(&c->buckets_waiting_for_journal,
+ c->journal.flushed_seq_ondisk,
+ new.k->p.inode, new.k->p.offset,
+ new_a.journal_seq);
+ if (ret) {
+ bch2_fs_fatal_error(c,
+ "error setting bucket_needs_journal_commit: %i", ret);
+ return ret;
+ }
+ }
- if (new.k->p.offset >= ca->mi.nbuckets)
- return 0;
+ if (new_a.data_type == BCH_DATA_free &&
+ (!new_a.journal_seq || new_a.journal_seq < c->journal.flushed_seq_ondisk))
+ closure_wake_up(&c->freelist_wait);
+
+ if (new_a.data_type == BCH_DATA_need_discard &&
+ (!new_a.journal_seq || new_a.journal_seq < c->journal.flushed_seq_ondisk))
+ bch2_do_discards(c);
+
+ if (old_a.data_type != BCH_DATA_cached &&
+ new_a.data_type == BCH_DATA_cached &&
+ should_invalidate_buckets(ca, bch2_dev_usage_read(ca)))
+ bch2_do_invalidates(c);
+
+ if (new_a.data_type == BCH_DATA_need_gc_gens)
+ bch2_do_gc_gens(c);
percpu_down_read(&c->mark_lock);
- if (!gc && new_u.gen != old_u.gen)
- *bucket_gen(ca, new.k->p.offset) = new_u.gen;
+ if (!gc && new_a.gen != old_a.gen)
+ *bucket_gen(ca, new.k->p.offset) = new_a.gen;
- g = __bucket(ca, new.k->p.offset, gc);
+ bch2_dev_usage_update(c, ca, old_a, new_a, journal_seq, gc);
- old_m = bucket_cmpxchg(g, m, ({
- m.gen = new_u.gen;
- m.data_type = new_u.data_type;
- m.dirty_sectors = new_u.dirty_sectors;
- m.cached_sectors = new_u.cached_sectors;
- m.stripe = new_u.stripe != 0;
+ if (gc) {
+ struct bucket *g = gc_bucket(ca, new.k->p.offset);
- if (journal_seq) {
- m.journal_seq_valid = 1;
- m.journal_seq = journal_seq;
- }
- }));
+ bucket_lock(g);
- bch2_dev_usage_update(c, ca, old_m, m, journal_seq, gc);
+ g->gen_valid = 1;
+ g->gen = new_a.gen;
+ g->data_type = new_a.data_type;
+ g->stripe = new_a.stripe;
+ g->stripe_redundancy = new_a.stripe_redundancy;
+ g->dirty_sectors = new_a.dirty_sectors;
+ g->cached_sectors = new_a.cached_sectors;
- g->io_time[READ] = new_u.read_time;
- g->io_time[WRITE] = new_u.write_time;
- g->oldest_gen = new_u.oldest_gen;
- g->gen_valid = 1;
- g->stripe = new_u.stripe;
- g->stripe_redundancy = new_u.stripe_redundancy;
+ bucket_unlock(g);
+ }
percpu_up_read(&c->mark_lock);
/*
*/
if ((flags & BTREE_TRIGGER_BUCKET_INVALIDATE) &&
- old_m.cached_sectors) {
+ old_a.cached_sectors) {
ret = update_cached_sectors(c, new, ca->dev_idx,
- -old_m.cached_sectors,
+ -old_a.cached_sectors,
journal_seq, gc);
if (ret) {
bch2_fs_fatal_error(c, "bch2_mark_alloc(): no replicas entry while updating cached sectors");
return ret;
}
-
- trace_invalidate(ca, bucket_to_sector(ca, new.k->p.offset),
- old_m.cached_sectors);
}
return 0;
}
-#define checked_add(a, b) \
-({ \
- unsigned _res = (unsigned) (a) + (b); \
- bool overflow = _res > U16_MAX; \
- if (overflow) \
- _res = U16_MAX; \
- (a) = _res; \
- overflow; \
-})
-
-void bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca,
- size_t b, enum bch_data_type data_type,
- unsigned sectors, struct gc_pos pos,
- unsigned flags)
+int bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca,
+ size_t b, enum bch_data_type data_type,
+ unsigned sectors, struct gc_pos pos,
+ unsigned flags)
{
- struct bucket *g;
- struct bucket_mark old, new;
- bool overflow;
+ struct bucket old, new, *g;
+ int ret = 0;
BUG_ON(!(flags & BTREE_TRIGGER_GC));
BUG_ON(data_type != BCH_DATA_sb &&
* Backup superblock might be past the end of our normal usable space:
*/
if (b >= ca->mi.nbuckets)
- return;
+ return 0;
percpu_down_read(&c->mark_lock);
g = gc_bucket(ca, b);
- old = bucket_cmpxchg(g, new, ({
- new.data_type = data_type;
- overflow = checked_add(new.dirty_sectors, sectors);
- }));
-
- bch2_fs_inconsistent_on(old.data_type &&
- old.data_type != data_type, c,
- "different types of data in same bucket: %s, %s",
- bch2_data_types[old.data_type],
- bch2_data_types[data_type]);
-
- bch2_fs_inconsistent_on(overflow, c,
- "bucket %u:%zu gen %u data type %s sector count overflow: %u + %u > U16_MAX",
- ca->dev_idx, b, new.gen,
- bch2_data_types[old.data_type ?: data_type],
- old.dirty_sectors, sectors);
-
- bch2_dev_usage_update(c, ca, old, new, 0, true);
+
+ bucket_lock(g);
+ old = *g;
+
+ if (bch2_fs_inconsistent_on(g->data_type &&
+ g->data_type != data_type, c,
+ "different types of data in same bucket: %s, %s",
+ bch2_data_types[g->data_type],
+ bch2_data_types[data_type])) {
+ ret = -EIO;
+ goto err;
+ }
+
+ if (bch2_fs_inconsistent_on((u64) g->dirty_sectors + sectors > ca->mi.bucket_size, c,
+ "bucket %u:%zu gen %u data type %s sector count overflow: %u + %u > bucket size",
+ ca->dev_idx, b, g->gen,
+ bch2_data_types[g->data_type ?: data_type],
+ g->dirty_sectors, sectors)) {
+ ret = -EIO;
+ goto err;
+ }
+
+
+ g->data_type = data_type;
+ g->dirty_sectors += sectors;
+ new = *g;
+err:
+ bucket_unlock(g);
+ if (!ret)
+ bch2_dev_usage_update_m(c, ca, old, new, 0, true);
percpu_up_read(&c->mark_lock);
+ return ret;
}
static s64 ptr_disk_sectors(s64 sectors, struct extent_ptr_decoded p)
{
EBUG_ON(sectors < 0);
- return p.crc.compression_type &&
- p.crc.compression_type != BCH_COMPRESSION_TYPE_incompressible
+ return crc_is_compressed(p.crc)
? DIV_ROUND_UP_ULL(sectors * p.crc.compressed_size,
- p.crc.uncompressed_size)
+ p.crc.uncompressed_size)
: sectors;
}
struct bkey_s_c k,
const struct bch_extent_ptr *ptr,
s64 sectors, enum bch_data_type ptr_data_type,
- u8 bucket_gen, u8 bucket_data_type,
- u16 dirty_sectors, u16 cached_sectors)
+ u8 b_gen, u8 bucket_data_type,
+ u32 dirty_sectors, u32 cached_sectors)
{
- size_t bucket_nr = PTR_BUCKET_NR(bch_dev_bkey_exists(c, ptr->dev), ptr);
+ struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
+ size_t bucket_nr = PTR_BUCKET_NR(ca, ptr);
u16 bucket_sectors = !ptr->cached
? dirty_sectors
: cached_sectors;
- char buf[200];
+ struct printbuf buf = PRINTBUF;
+ int ret = 0;
- if (gen_after(ptr->gen, bucket_gen)) {
+ if (bucket_data_type == BCH_DATA_cached)
+ bucket_data_type = BCH_DATA_user;
+
+ if (gen_after(ptr->gen, b_gen)) {
bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
"bucket %u:%zu gen %u data type %s: ptr gen %u newer than bucket gen\n"
"while marking %s",
- ptr->dev, bucket_nr, bucket_gen,
+ ptr->dev, bucket_nr, b_gen,
bch2_data_types[bucket_data_type ?: ptr_data_type],
ptr->gen,
- (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf));
- return -EIO;
+ (bch2_bkey_val_to_text(&buf, c, k), buf.buf));
+ ret = -EIO;
+ goto err;
}
- if (gen_cmp(bucket_gen, ptr->gen) > BUCKET_GC_GEN_MAX) {
+ if (gen_cmp(b_gen, ptr->gen) > BUCKET_GC_GEN_MAX) {
bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
"bucket %u:%zu gen %u data type %s: ptr gen %u too stale\n"
"while marking %s",
- ptr->dev, bucket_nr, bucket_gen,
+ ptr->dev, bucket_nr, b_gen,
bch2_data_types[bucket_data_type ?: ptr_data_type],
ptr->gen,
- (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf));
- return -EIO;
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf));
+ ret = -EIO;
+ goto err;
}
- if (bucket_gen != ptr->gen && !ptr->cached) {
+ if (b_gen != ptr->gen && !ptr->cached) {
bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
- "bucket %u:%zu gen %u data type %s: stale dirty ptr (gen %u)\n"
+ "bucket %u:%zu gen %u (mem gen %u) data type %s: stale dirty ptr (gen %u)\n"
"while marking %s",
- ptr->dev, bucket_nr, bucket_gen,
+ ptr->dev, bucket_nr, b_gen,
+ *bucket_gen(ca, bucket_nr),
bch2_data_types[bucket_data_type ?: ptr_data_type],
ptr->gen,
- (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf));
- return -EIO;
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf));
+ ret = -EIO;
+ goto err;
}
- if (bucket_gen != ptr->gen)
- return 1;
+ if (b_gen != ptr->gen) {
+ ret = 1;
+ goto err;
+ }
- if (bucket_data_type && ptr_data_type &&
+ if (!data_type_is_empty(bucket_data_type) &&
+ ptr_data_type &&
bucket_data_type != ptr_data_type) {
bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
"bucket %u:%zu gen %u different types of data in same bucket: %s, %s\n"
"while marking %s",
- ptr->dev, bucket_nr, bucket_gen,
+ ptr->dev, bucket_nr, b_gen,
bch2_data_types[bucket_data_type],
bch2_data_types[ptr_data_type],
- (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf));
- return -EIO;
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf));
+ ret = -EIO;
+ goto err;
}
- if ((unsigned) (bucket_sectors + sectors) > U16_MAX) {
+ if ((unsigned) (bucket_sectors + sectors) > U32_MAX) {
bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
"bucket %u:%zu gen %u data type %s sector count overflow: %u + %lli > U16_MAX\n"
"while marking %s",
- ptr->dev, bucket_nr, bucket_gen,
+ ptr->dev, bucket_nr, b_gen,
bch2_data_types[bucket_data_type ?: ptr_data_type],
bucket_sectors, sectors,
- (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf));
- return -EIO;
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf));
+ ret = -EIO;
+ goto err;
}
-
- return 0;
+err:
+ printbuf_exit(&buf);
+ return ret;
}
static int mark_stripe_bucket(struct btree_trans *trans,
struct bkey_s_c k,
unsigned ptr_idx,
- u64 journal_seq, unsigned flags)
+ unsigned flags)
{
struct bch_fs *c = trans->c;
+ u64 journal_seq = trans->journal_res.seq;
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;
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 *g;
- struct bucket_mark new, old;
- char buf[200];
+ struct bucket old, new, *g;
+ struct printbuf buf = PRINTBUF;
int ret = 0;
BUG_ON(!(flags & BTREE_TRIGGER_GC));
/* * XXX doesn't handle deletion */
percpu_down_read(&c->mark_lock);
+ buf.atomic++;
g = PTR_GC_BUCKET(ca, ptr);
- if (g->mark.dirty_sectors ||
+ 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->mark.gen,
- (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf));
+ ptr->dev, PTR_BUCKET_NR(ca, ptr), g->gen,
+ (bch2_bkey_val_to_text(&buf, c, k), buf.buf));
ret = -EINVAL;
goto err;
}
- old = bucket_cmpxchg(g, new, ({
- ret = check_bucket_ref(c, k, ptr, sectors, data_type,
- new.gen, new.data_type,
- new.dirty_sectors, new.cached_sectors);
- if (ret)
- goto err;
-
- new.dirty_sectors += sectors;
- if (data_type)
- new.data_type = data_type;
+ bucket_lock(g);
+ old = *g;
- if (journal_seq) {
- new.journal_seq_valid = 1;
- new.journal_seq = journal_seq;
- }
+ ret = check_bucket_ref(c, k, ptr, sectors, data_type,
+ g->gen, g->data_type,
+ g->dirty_sectors, g->cached_sectors);
+ if (ret)
+ goto err;
- new.stripe = true;
- }));
+ if (data_type)
+ g->data_type = data_type;
+ g->dirty_sectors += sectors;
g->stripe = k.k->p.offset;
g->stripe_redundancy = s->nr_redundant;
-
- bch2_dev_usage_update(c, ca, old, new, journal_seq, true);
+ new = *g;
err:
+ bucket_unlock(g);
+ if (!ret)
+ bch2_dev_usage_update_m(c, ca, old, new, journal_seq, true);
percpu_up_read(&c->mark_lock);
-
- return 0;
+ printbuf_exit(&buf);
+ return ret;
}
static int __mark_pointer(struct btree_trans *trans,
const struct bch_extent_ptr *ptr,
s64 sectors, enum bch_data_type ptr_data_type,
u8 bucket_gen, u8 *bucket_data_type,
- u16 *dirty_sectors, u16 *cached_sectors)
+ u32 *dirty_sectors, u32 *cached_sectors)
{
- u16 *dst_sectors = !ptr->cached
+ u32 *dst_sectors = !ptr->cached
? dirty_sectors
: cached_sectors;
int ret = check_bucket_ref(trans->c, k, ptr, sectors, ptr_data_type,
{
u64 journal_seq = trans->journal_res.seq;
struct bch_fs *c = trans->c;
- struct bucket_mark old, new;
struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev);
- struct bucket *g;
+ struct bucket old, new, *g;
u8 bucket_data_type;
- u64 v;
int ret = 0;
BUG_ON(!(flags & BTREE_TRIGGER_GC));
percpu_down_read(&c->mark_lock);
g = PTR_GC_BUCKET(ca, &p.ptr);
-
- v = atomic64_read(&g->_mark.v);
- do {
- new.v.counter = old.v.counter = v;
- bucket_data_type = new.data_type;
-
- ret = __mark_pointer(trans, k, &p.ptr, sectors,
- data_type, new.gen,
- &bucket_data_type,
- &new.dirty_sectors,
- &new.cached_sectors);
- if (ret)
- goto err;
-
- new.data_type = bucket_data_type;
-
- if (journal_seq) {
- new.journal_seq_valid = 1;
- new.journal_seq = journal_seq;
- }
-
- if (flags & BTREE_TRIGGER_NOATOMIC) {
- g->_mark = new;
- break;
- }
- } while ((v = atomic64_cmpxchg(&g->_mark.v,
- old.v.counter,
- new.v.counter)) != old.v.counter);
-
- bch2_dev_usage_update(c, ca, old, new, journal_seq, true);
-err:
+ bucket_lock(g);
+ old = *g;
+
+ bucket_data_type = g->data_type;
+ ret = __mark_pointer(trans, k, &p.ptr, sectors,
+ data_type, g->gen,
+ &bucket_data_type,
+ &g->dirty_sectors,
+ &g->cached_sectors);
+ if (!ret)
+ g->data_type = bucket_data_type;
+
+ new = *g;
+ bucket_unlock(g);
+ if (!ret)
+ bch2_dev_usage_update_m(c, ca, old, new, journal_seq, true);
percpu_up_read(&c->mark_lock);
return ret;
return 0;
}
-static int bch2_mark_extent(struct btree_trans *trans,
- struct bkey_s_c old, struct bkey_s_c new,
- unsigned flags)
+int bch2_mark_extent(struct btree_trans *trans,
+ struct bkey_s_c old, struct bkey_s_c new,
+ unsigned flags)
{
u64 journal_seq = trans->journal_res.seq;
struct bch_fs *c = trans->c;
if (r.e.nr_devs) {
ret = update_replicas(c, k, &r.e, dirty_sectors, journal_seq, true);
if (ret) {
- char buf[200];
+ struct printbuf buf = PRINTBUF;
- bch2_bkey_val_to_text(&PBUF(buf), c, k);
- bch2_fs_fatal_error(c, "no replicas entry for %s", buf);
+ bch2_bkey_val_to_text(&buf, c, k);
+ bch2_fs_fatal_error(c, "no replicas entry for %s", buf.buf);
+ printbuf_exit(&buf);
return ret;
}
}
return 0;
}
-static int bch2_mark_stripe(struct btree_trans *trans,
- struct bkey_s_c old, struct bkey_s_c new,
- unsigned flags)
+int bch2_mark_stripe(struct btree_trans *trans,
+ struct bkey_s_c old, struct bkey_s_c new,
+ unsigned flags)
{
bool gc = flags & BTREE_TRIGGER_GC;
u64 journal_seq = trans->journal_res.seq;
struct stripe *m = genradix_ptr(&c->stripes, idx);
if (!m || (old_s && !m->alive)) {
- char buf1[200], buf2[200];
+ struct printbuf buf1 = PRINTBUF;
+ struct printbuf buf2 = PRINTBUF;
- bch2_bkey_val_to_text(&PBUF(buf1), c, old);
- bch2_bkey_val_to_text(&PBUF(buf2), c, new);
+ bch2_bkey_val_to_text(&buf1, c, old);
+ bch2_bkey_val_to_text(&buf2, c, new);
bch_err_ratelimited(c, "error marking nonexistent stripe %llu while marking\n"
"old %s\n"
- "new %s", idx, buf1, buf2);
+ "new %s", idx, buf1.buf, buf2.buf);
+ printbuf_exit(&buf2);
+ printbuf_exit(&buf1);
bch2_inconsistent_error(c);
return -1;
}
memset(m->block_sectors, 0, sizeof(m->block_sectors));
for (i = 0; i < new_s->nr_blocks; i++) {
- ret = mark_stripe_bucket(trans, new, i, journal_seq, flags);
+ ret = mark_stripe_bucket(trans, new, i, flags);
if (ret)
return ret;
}
((s64) m->sectors * m->nr_redundant),
journal_seq, gc);
if (ret) {
- char buf[200];
+ struct printbuf buf = PRINTBUF;
- bch2_bkey_val_to_text(&PBUF(buf), c, new);
- bch2_fs_fatal_error(c, "no replicas entry for %s", buf);
+ 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;
}
-static int bch2_mark_inode(struct btree_trans *trans,
- struct bkey_s_c old, struct bkey_s_c new,
- unsigned flags)
+int bch2_mark_inode(struct btree_trans *trans,
+ struct bkey_s_c old, struct bkey_s_c new,
+ unsigned flags)
{
struct bch_fs *c = trans->c;
struct bch_fs_usage __percpu *fs_usage;
return 0;
}
-static int bch2_mark_reservation(struct btree_trans *trans,
- struct bkey_s_c old, struct bkey_s_c new,
- unsigned flags)
+int bch2_mark_reservation(struct btree_trans *trans,
+ struct bkey_s_c old, struct bkey_s_c new,
+ unsigned flags)
{
struct bch_fs *c = trans->c;
struct bkey_s_c k = flags & BTREE_TRIGGER_OVERWRITE ? old: new;
return 0;
}
-static s64 __bch2_mark_reflink_p(struct bch_fs *c, struct bkey_s_c_reflink_p p,
+static s64 __bch2_mark_reflink_p(struct btree_trans *trans,
+ struct bkey_s_c_reflink_p p,
+ u64 start, u64 end,
u64 *idx, unsigned flags, size_t r_idx)
{
+ struct bch_fs *c = trans->c;
struct reflink_gc *r;
int add = !(flags & BTREE_TRIGGER_OVERWRITE) ? 1 : -1;
+ u64 next_idx = end;
s64 ret = 0;
+ struct printbuf buf = PRINTBUF;
if (r_idx >= c->reflink_gc_nr)
goto not_found;
r = genradix_ptr(&c->reflink_gc_table, r_idx);
- if (*idx < r->offset - r->size)
+ next_idx = min(next_idx, r->offset - r->size);
+ if (*idx < next_idx)
goto not_found;
BUG_ON((s64) r->refcount + add < 0);
*idx = r->offset;
return 0;
not_found:
- *idx = U64_MAX;
- ret = -EIO;
-
- /*
- * XXX: we're replacing the entire reflink pointer with an error
- * key, we should just be replacing the part that was missing:
- */
- if (fsck_err(c, "%llu:%llu len %u points to nonexistent indirect extent %llu",
- p.k->p.inode, p.k->p.offset, p.k->size, *idx)) {
+ if (fsck_err(c, "pointer to missing indirect extent\n"
+ " %s\n"
+ " missing range %llu-%llu",
+ (bch2_bkey_val_to_text(&buf, c, p.s_c), buf.buf),
+ *idx, next_idx)) {
struct bkey_i_error new;
bkey_init(&new.k);
new.k.type = KEY_TYPE_error;
- new.k.p = p.k->p;
- new.k.size = p.k->size;
- ret = bch2_journal_key_insert(c, BTREE_ID_extents, 0, &new.k_i);
+ new.k.p = bkey_start_pos(p.k);
+ new.k.p.offset += *idx - start;
+ bch2_key_resize(&new.k, next_idx - *idx);
+ ret = __bch2_btree_insert(trans, BTREE_ID_extents, &new.k_i);
}
+
+ *idx = next_idx;
fsck_err:
+ printbuf_exit(&buf);
return ret;
}
-static int bch2_mark_reflink_p(struct btree_trans *trans,
- struct bkey_s_c old, struct bkey_s_c new,
- unsigned flags)
+int bch2_mark_reflink_p(struct btree_trans *trans,
+ struct bkey_s_c old, struct bkey_s_c new,
+ unsigned flags)
{
struct bch_fs *c = trans->c;
struct bkey_s_c k = flags & BTREE_TRIGGER_OVERWRITE ? old: new;
struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k);
struct reflink_gc *ref;
size_t l, r, m;
- u64 idx = le64_to_cpu(p.v->idx);
+ u64 idx = le64_to_cpu(p.v->idx), start = idx;
u64 end = le64_to_cpu(p.v->idx) + p.k->size;
int ret = 0;
}
while (idx < end && !ret)
- ret = __bch2_mark_reflink_p(c, p, &idx, flags, l++);
-
- return ret;
-}
-
-int bch2_mark_key(struct btree_trans *trans,
- struct bkey_s_c old,
- struct bkey_s_c new,
- unsigned flags)
-{
- struct bkey_s_c k = flags & BTREE_TRIGGER_OVERWRITE ? old: new;
-
- switch (k.k->type) {
- case KEY_TYPE_alloc:
- case KEY_TYPE_alloc_v2:
- case KEY_TYPE_alloc_v3:
- return bch2_mark_alloc(trans, old, new, flags);
- case KEY_TYPE_btree_ptr:
- case KEY_TYPE_btree_ptr_v2:
- case KEY_TYPE_extent:
- case KEY_TYPE_reflink_v:
- return bch2_mark_extent(trans, old, new, flags);
- case KEY_TYPE_stripe:
- return bch2_mark_stripe(trans, old, new, flags);
- case KEY_TYPE_inode:
- case KEY_TYPE_inode_v2:
- return bch2_mark_inode(trans, old, new, flags);
- case KEY_TYPE_reservation:
- return bch2_mark_reservation(trans, old, new, flags);
- case KEY_TYPE_reflink_p:
- return bch2_mark_reflink_p(trans, old, new, flags);
- case KEY_TYPE_snapshot:
- return bch2_mark_snapshot(trans, old, new, flags);
- default:
- return 0;
- }
-}
-
-int bch2_mark_update(struct btree_trans *trans, struct btree_path *path,
- struct bkey_i *new, unsigned flags)
-{
- struct bkey _deleted = KEY(0, 0, 0);
- struct bkey_s_c deleted = (struct bkey_s_c) { &_deleted, NULL };
- struct bkey_s_c old;
- struct bkey unpacked;
- int ret;
-
- _deleted.p = path->pos;
-
- if (unlikely(flags & BTREE_TRIGGER_NORUN))
- return 0;
-
- if (!btree_node_type_needs_gc(path->btree_id))
- return 0;
-
- old = bch2_btree_path_peek_slot(path, &unpacked);
-
- if (old.k->type == new->k.type &&
- ((1U << old.k->type) & BTREE_TRIGGER_WANTS_OLD_AND_NEW)) {
- ret = bch2_mark_key(trans, old, bkey_i_to_s_c(new),
- BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE|flags);
- } else {
- ret = bch2_mark_key(trans, deleted, bkey_i_to_s_c(new),
- BTREE_TRIGGER_INSERT|flags) ?:
- bch2_mark_key(trans, old, deleted,
- BTREE_TRIGGER_OVERWRITE|flags);
- }
+ ret = __bch2_mark_reflink_p(trans, p, start, end,
+ &idx, flags, l++);
return ret;
}
{
struct bch_fs *c = trans->c;
struct btree_insert_entry *i;
- char buf[200];
+ struct printbuf buf = PRINTBUF;
bch_err(c, "disk usage increased %lli more than %u sectors reserved",
should_not_have_added, disk_res_sectors);
trans_for_each_update(trans, i) {
+ struct bkey_s_c old = { &i->old_k, i->old_v };
+
pr_err("while inserting");
- bch2_bkey_val_to_text(&PBUF(buf), c, bkey_i_to_s_c(i->k));
- pr_err("%s", buf);
+ printbuf_reset(&buf);
+ bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(i->k));
+ pr_err(" %s", buf.buf);
pr_err("overlapping with");
-
- if (!i->cached) {
- struct bkey u;
- struct bkey_s_c k = bch2_btree_path_peek_slot(i->path, &u);
-
- bch2_bkey_val_to_text(&PBUF(buf), c, k);
- pr_err("%s", buf);
- } else {
- struct bkey_cached *ck = (void *) i->path->l[0].b;
-
- if (ck->valid) {
- bch2_bkey_val_to_text(&PBUF(buf), c, bkey_i_to_s_c(ck->k));
- pr_err("%s", buf);
- }
- }
+ printbuf_reset(&buf);
+ bch2_bkey_val_to_text(&buf, c, old);
+ pr_err(" %s", buf.buf);
}
+
__WARN();
+ printbuf_exit(&buf);
}
int bch2_trans_fs_usage_apply(struct btree_trans *trans,
/* trans_mark: */
-static int bch2_trans_start_alloc_update(struct btree_trans *trans, struct btree_iter *iter,
- const struct bch_extent_ptr *ptr,
- struct bkey_alloc_unpacked *u)
-{
- struct bch_fs *c = trans->c;
- struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
- struct bpos pos = POS(ptr->dev, PTR_BUCKET_NR(ca, ptr));
- struct bkey_i *update = btree_trans_peek_updates(trans, BTREE_ID_alloc, pos);
- int ret;
-
- bch2_trans_iter_init(trans, iter, BTREE_ID_alloc, pos,
- BTREE_ITER_CACHED|
- BTREE_ITER_CACHED_NOFILL|
- BTREE_ITER_INTENT);
- ret = bch2_btree_iter_traverse(iter);
- if (ret) {
- bch2_trans_iter_exit(trans, iter);
- return ret;
- }
-
- *u = update && !bpos_cmp(update->k.p, pos)
- ? bch2_alloc_unpack(bkey_i_to_s_c(update))
- : alloc_mem_to_key(c, iter);
-
- return 0;
-}
-
static int bch2_trans_mark_pointer(struct btree_trans *trans,
struct bkey_s_c k, struct extent_ptr_decoded p,
s64 sectors, enum bch_data_type data_type)
{
struct btree_iter iter;
- struct bkey_alloc_unpacked u;
+ struct bkey_i_alloc_v4 *a;
int ret;
- ret = bch2_trans_start_alloc_update(trans, &iter, &p.ptr, &u);
- if (ret)
- return ret;
+ a = bch2_trans_start_alloc_update(trans, &iter, PTR_BUCKET_POS(trans->c, &p.ptr));
+ if (IS_ERR(a))
+ return PTR_ERR(a);
ret = __mark_pointer(trans, k, &p.ptr, sectors, data_type,
- u.gen, &u.data_type,
- &u.dirty_sectors, &u.cached_sectors);
- if (ret)
- goto out;
-
- ret = bch2_alloc_write(trans, &iter, &u, 0);
- if (ret)
- goto out;
-out:
+ a->v.gen, &a->v.data_type,
+ &a->v.dirty_sectors, &a->v.cached_sectors) ?:
+ bch2_trans_update(trans, &iter, &a->k_i, 0);
bch2_trans_iter_exit(trans, &iter);
return ret;
}
struct extent_ptr_decoded p,
s64 sectors, enum bch_data_type data_type)
{
- struct bch_fs *c = trans->c;
struct btree_iter iter;
struct bkey_s_c k;
struct bkey_i_stripe *s;
goto err;
if (k.k->type != KEY_TYPE_stripe) {
- bch2_fs_inconsistent(c,
+ bch2_trans_inconsistent(trans,
"pointer to nonexistent stripe %llu",
(u64) p.ec.idx);
- bch2_inconsistent_error(c);
ret = -EIO;
goto err;
}
if (!bch2_ptr_matches_stripe(bkey_s_c_to_stripe(k).v, p)) {
- bch2_fs_inconsistent(c,
+ bch2_trans_inconsistent(trans,
"stripe pointer doesn't match stripe %llu",
(u64) p.ec.idx);
ret = -EIO;
return ret;
}
-static int bch2_trans_mark_extent(struct btree_trans *trans,
- struct bkey_s_c k, unsigned flags)
+int bch2_trans_mark_extent(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c old, struct bkey_i *new,
+ unsigned flags)
{
struct bch_fs *c = trans->c;
+ struct bkey_s_c k = flags & BTREE_TRIGGER_OVERWRITE
+ ? old
+ : bkey_i_to_s_c(new);
struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
const union bch_extent_entry *entry;
struct extent_ptr_decoded p;
struct bch_fs *c = trans->c;
const struct bch_extent_ptr *ptr = &s.v->ptrs[idx];
struct btree_iter iter;
- struct bkey_alloc_unpacked u;
+ 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;
if (deleting)
sectors = -sectors;
- ret = bch2_trans_start_alloc_update(trans, &iter, ptr, &u);
- if (ret)
- return ret;
+ a = bch2_trans_start_alloc_update(trans, &iter, PTR_BUCKET_POS(c, ptr));
+ if (IS_ERR(a))
+ return PTR_ERR(a);
ret = check_bucket_ref(c, s.s_c, ptr, sectors, data_type,
- u.gen, u.data_type,
- u.dirty_sectors, u.cached_sectors);
+ a->v.gen, a->v.data_type,
+ a->v.dirty_sectors, a->v.cached_sectors);
if (ret)
goto err;
if (!deleting) {
- if (bch2_fs_inconsistent_on(u.stripe ||
- u.stripe_redundancy, c,
+ 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, u.gen,
- bch2_data_types[u.data_type],
- u.dirty_sectors,
- u.stripe, s.k->p.offset)) {
+ iter.pos.inode, iter.pos.offset, a->v.gen,
+ bch2_data_types[a->v.data_type],
+ a->v.dirty_sectors,
+ a->v.stripe, s.k->p.offset)) {
ret = -EIO;
goto err;
}
- if (bch2_fs_inconsistent_on(data_type && u.dirty_sectors, c,
+ 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, u.gen,
- bch2_data_types[u.data_type],
- u.dirty_sectors,
+ iter.pos.inode, iter.pos.offset, a->v.gen,
+ bch2_data_types[a->v.data_type],
+ a->v.dirty_sectors,
s.k->p.offset)) {
ret = -EIO;
goto err;
}
- u.stripe = s.k->p.offset;
- u.stripe_redundancy = s.v->nr_redundant;
+ a->v.stripe = s.k->p.offset;
+ a->v.stripe_redundancy = s.v->nr_redundant;
} else {
- if (bch2_fs_inconsistent_on(u.stripe != s.k->p.offset ||
- u.stripe_redundancy != s.v->nr_redundant, c,
+ 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, u.gen,
- s.k->p.offset, u.stripe)) {
+ iter.pos.inode, iter.pos.offset, a->v.gen,
+ s.k->p.offset, a->v.stripe)) {
ret = -EIO;
goto err;
}
- u.stripe = 0;
- u.stripe_redundancy = 0;
+ a->v.stripe = 0;
+ a->v.stripe_redundancy = 0;
}
- u.dirty_sectors += sectors;
+ a->v.dirty_sectors += sectors;
if (data_type)
- u.data_type = !deleting ? data_type : 0;
+ a->v.data_type = !deleting ? data_type : 0;
- ret = bch2_alloc_write(trans, &iter, &u, 0);
+ ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
if (ret)
goto err;
err:
return ret;
}
-static int bch2_trans_mark_stripe(struct btree_trans *trans,
- struct bkey_s_c old, struct bkey_s_c new,
- unsigned flags)
+int bch2_trans_mark_stripe(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c old, struct bkey_i *new,
+ unsigned flags)
{
- struct bkey_s_c_stripe old_s = { .k = NULL };
- struct bkey_s_c_stripe new_s = { .k = NULL };
+ const struct bch_stripe *old_s = NULL;
+ struct bch_stripe *new_s = NULL;
struct bch_replicas_padded r;
unsigned i, nr_blocks;
int ret = 0;
if (old.k->type == KEY_TYPE_stripe)
- old_s = bkey_s_c_to_stripe(old);
- if (new.k->type == KEY_TYPE_stripe)
- new_s = bkey_s_c_to_stripe(new);
+ old_s = bkey_s_c_to_stripe(old).v;
+ if (new->k.type == KEY_TYPE_stripe)
+ new_s = &bkey_i_to_stripe(new)->v;
/*
* If the pointers aren't changing, we don't need to do anything:
*/
- if (new_s.k && old_s.k &&
- new_s.v->nr_blocks == old_s.v->nr_blocks &&
- new_s.v->nr_redundant == old_s.v->nr_redundant &&
- !memcmp(old_s.v->ptrs, new_s.v->ptrs,
- new_s.v->nr_blocks * sizeof(struct bch_extent_ptr)))
+ if (new_s && old_s &&
+ new_s->nr_blocks == old_s->nr_blocks &&
+ new_s->nr_redundant == old_s->nr_redundant &&
+ !memcmp(old_s->ptrs, new_s->ptrs,
+ new_s->nr_blocks * sizeof(struct bch_extent_ptr)))
return 0;
- BUG_ON(new_s.k && old_s.k &&
- (new_s.v->nr_blocks != old_s.v->nr_blocks ||
- new_s.v->nr_redundant != old_s.v->nr_redundant));
+ BUG_ON(new_s && old_s &&
+ (new_s->nr_blocks != old_s->nr_blocks ||
+ new_s->nr_redundant != old_s->nr_redundant));
- nr_blocks = new_s.k ? new_s.v->nr_blocks : old_s.v->nr_blocks;
+ nr_blocks = new_s ? new_s->nr_blocks : old_s->nr_blocks;
- if (new_s.k) {
- s64 sectors = le16_to_cpu(new_s.v->sectors);
+ if (new_s) {
+ s64 sectors = le16_to_cpu(new_s->sectors);
- bch2_bkey_to_replicas(&r.e, new);
- update_replicas_list(trans, &r.e, sectors * new_s.v->nr_redundant);
+ bch2_bkey_to_replicas(&r.e, bkey_i_to_s_c(new));
+ update_replicas_list(trans, &r.e, sectors * new_s->nr_redundant);
}
- if (old_s.k) {
- s64 sectors = -((s64) le16_to_cpu(old_s.v->sectors));
+ if (old_s) {
+ s64 sectors = -((s64) le16_to_cpu(old_s->sectors));
bch2_bkey_to_replicas(&r.e, old);
- update_replicas_list(trans, &r.e, sectors * old_s.v->nr_redundant);
+ update_replicas_list(trans, &r.e, sectors * old_s->nr_redundant);
}
for (i = 0; i < nr_blocks; i++) {
- if (new_s.k && old_s.k &&
- !memcmp(&new_s.v->ptrs[i],
- &old_s.v->ptrs[i],
- sizeof(new_s.v->ptrs[i])))
+ if (new_s && old_s &&
+ !memcmp(&new_s->ptrs[i],
+ &old_s->ptrs[i],
+ sizeof(new_s->ptrs[i])))
continue;
- if (new_s.k) {
- ret = bch2_trans_mark_stripe_bucket(trans, new_s, i, false);
+ if (new_s) {
+ ret = bch2_trans_mark_stripe_bucket(trans,
+ bkey_i_to_s_c_stripe(new), i, false);
if (ret)
break;
}
- if (old_s.k) {
- ret = bch2_trans_mark_stripe_bucket(trans, old_s, i, true);
+ if (old_s) {
+ ret = bch2_trans_mark_stripe_bucket(trans,
+ bkey_s_c_to_stripe(old), i, true);
if (ret)
break;
}
return ret;
}
-static int bch2_trans_mark_inode(struct btree_trans *trans,
- struct bkey_s_c old,
- struct bkey_s_c new,
- unsigned flags)
+int bch2_trans_mark_inode(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c old,
+ struct bkey_i *new,
+ unsigned flags)
{
- int nr = bkey_is_inode(new.k) - bkey_is_inode(old.k);
+ int nr = bkey_is_inode(&new->k) - bkey_is_inode(old.k);
if (nr) {
struct replicas_delta_list *d =
return 0;
}
-static int bch2_trans_mark_reservation(struct btree_trans *trans,
- struct bkey_s_c k, unsigned flags)
+int bch2_trans_mark_reservation(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c old,
+ struct bkey_i *new,
+ unsigned flags)
{
+ struct bkey_s_c k = flags & BTREE_TRIGGER_OVERWRITE
+ ? old
+ : bkey_i_to_s_c(new);
unsigned replicas = bkey_s_c_to_reservation(k).v->nr_replicas;
s64 sectors = (s64) k.k->size;
struct replicas_delta_list *d;
struct bkey_i *n;
__le64 *refcount;
int add = !(flags & BTREE_TRIGGER_OVERWRITE) ? 1 : -1;
- char buf[200];
+ struct printbuf buf = PRINTBUF;
int ret;
bch2_trans_iter_init(trans, &iter, BTREE_ID_reflink, POS(0, *idx),
refcount = bkey_refcount(n);
if (!refcount) {
- bch2_bkey_val_to_text(&PBUF(buf), c, p.s_c);
- bch2_fs_inconsistent(c,
+ bch2_bkey_val_to_text(&buf, c, p.s_c);
+ bch2_trans_inconsistent(trans,
"nonexistent indirect extent at %llu while marking\n %s",
- *idx, buf);
+ *idx, buf.buf);
ret = -EIO;
goto err;
}
if (!*refcount && (flags & BTREE_TRIGGER_OVERWRITE)) {
- bch2_bkey_val_to_text(&PBUF(buf), c, p.s_c);
- bch2_fs_inconsistent(c,
+ bch2_bkey_val_to_text(&buf, c, p.s_c);
+ bch2_trans_inconsistent(trans,
"indirect extent refcount underflow at %llu while marking\n %s",
- *idx, buf);
+ *idx, buf.buf);
ret = -EIO;
goto err;
}
le64_add_cpu(refcount, add);
- if (!*refcount) {
- n->k.type = KEY_TYPE_deleted;
- set_bkey_val_u64s(&n->k, 0);
- }
-
bch2_btree_iter_set_pos_to_extent_start(&iter);
ret = bch2_trans_update(trans, &iter, n, 0);
if (ret)
*idx = k.k->p.offset;
err:
bch2_trans_iter_exit(trans, &iter);
+ printbuf_exit(&buf);
return ret;
}
-static int bch2_trans_mark_reflink_p(struct btree_trans *trans,
- struct bkey_s_c k, unsigned flags)
+int bch2_trans_mark_reflink_p(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c old,
+ struct bkey_i *new,
+ unsigned flags)
{
+ struct bkey_s_c k = flags & BTREE_TRIGGER_OVERWRITE
+ ? old
+ : bkey_i_to_s_c(new);
struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k);
u64 idx, end_idx;
int ret = 0;
return ret;
}
-int bch2_trans_mark_key(struct btree_trans *trans, struct bkey_s_c old,
- struct bkey_s_c new, unsigned flags)
-{
- struct bkey_s_c k = flags & BTREE_TRIGGER_OVERWRITE ? old: new;
-
- switch (k.k->type) {
- case KEY_TYPE_btree_ptr:
- case KEY_TYPE_btree_ptr_v2:
- case KEY_TYPE_extent:
- case KEY_TYPE_reflink_v:
- return bch2_trans_mark_extent(trans, k, flags);
- case KEY_TYPE_stripe:
- return bch2_trans_mark_stripe(trans, old, new, flags);
- case KEY_TYPE_inode:
- case KEY_TYPE_inode_v2:
- return bch2_trans_mark_inode(trans, old, new, flags);
- case KEY_TYPE_reservation:
- return bch2_trans_mark_reservation(trans, k, flags);
- case KEY_TYPE_reflink_p:
- return bch2_trans_mark_reflink_p(trans, k, flags);
- default:
- return 0;
- }
-}
-
static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
struct bch_dev *ca, size_t b,
enum bch_data_type type,
{
struct bch_fs *c = trans->c;
struct btree_iter iter;
- struct bkey_alloc_unpacked u;
- struct bch_extent_ptr ptr = {
- .dev = ca->dev_idx,
- .offset = bucket_to_sector(ca, b),
- };
+ struct bkey_i_alloc_v4 *a;
int ret = 0;
/*
if (b >= ca->mi.nbuckets)
return 0;
- ret = bch2_trans_start_alloc_update(trans, &iter, &ptr, &u);
- if (ret)
- return ret;
+ a = bch2_trans_start_alloc_update(trans, &iter, POS(ca->dev_idx, b));
+ if (IS_ERR(a))
+ return PTR_ERR(a);
- if (u.data_type && u.data_type != type) {
+ if (a->v.data_type && a->v.data_type != type) {
bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
"bucket %llu:%llu gen %u different types of data in same bucket: %s, %s\n"
"while marking %s",
- iter.pos.inode, iter.pos.offset, u.gen,
- bch2_data_types[u.data_type],
+ iter.pos.inode, iter.pos.offset, a->v.gen,
+ bch2_data_types[a->v.data_type],
bch2_data_types[type],
bch2_data_types[type]);
ret = -EIO;
goto out;
}
- u.data_type = type;
- u.dirty_sectors = sectors;
+ a->v.data_type = type;
+ a->v.dirty_sectors = sectors;
- ret = bch2_alloc_write(trans, &iter, &u, 0);
+ ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
if (ret)
goto out;
out:
/* Startup/shutdown: */
-static void buckets_free_rcu(struct rcu_head *rcu)
-{
- struct bucket_array *buckets =
- container_of(rcu, struct bucket_array, rcu);
-
- kvpfree(buckets,
- sizeof(struct bucket_array) +
- buckets->nbuckets * sizeof(struct bucket));
-}
-
static void bucket_gens_free_rcu(struct rcu_head *rcu)
{
struct bucket_gens *buckets =
container_of(rcu, struct bucket_gens, rcu);
- kvpfree(buckets, sizeof(struct bucket_array) + buckets->nbuckets);
+ kvpfree(buckets, sizeof(*buckets) + buckets->nbuckets);
}
int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
{
- struct bucket_array *buckets = NULL, *old_buckets = NULL;
struct bucket_gens *bucket_gens = NULL, *old_bucket_gens = NULL;
unsigned long *buckets_nouse = NULL;
- alloc_fifo free[RESERVE_NR];
- alloc_fifo free_inc;
- alloc_heap alloc_heap;
-
- size_t btree_reserve = DIV_ROUND_UP(BTREE_NODE_RESERVE,
- ca->mi.bucket_size / btree_sectors(c));
- /* XXX: these should be tunable */
- size_t reserve_none = max_t(size_t, 1, nbuckets >> 9);
- size_t copygc_reserve = max_t(size_t, 2, nbuckets >> 6);
- size_t free_inc_nr = max(max_t(size_t, 1, nbuckets >> 12),
- btree_reserve * 2);
- bool resize = ca->buckets[0] != NULL;
+ bool resize = ca->bucket_gens != NULL;
int ret = -ENOMEM;
- unsigned i;
-
- memset(&free, 0, sizeof(free));
- memset(&free_inc, 0, sizeof(free_inc));
- memset(&alloc_heap, 0, sizeof(alloc_heap));
- if (!(buckets = kvpmalloc(sizeof(struct bucket_array) +
- nbuckets * sizeof(struct bucket),
+ if (!(bucket_gens = kvpmalloc(sizeof(struct bucket_gens) + nbuckets,
GFP_KERNEL|__GFP_ZERO)) ||
- !(bucket_gens = kvpmalloc(sizeof(struct bucket_gens) + nbuckets,
- GFP_KERNEL|__GFP_ZERO)) ||
- !(buckets_nouse = kvpmalloc(BITS_TO_LONGS(nbuckets) *
+ (c->opts.buckets_nouse &&
+ !(buckets_nouse = kvpmalloc(BITS_TO_LONGS(nbuckets) *
sizeof(unsigned long),
- GFP_KERNEL|__GFP_ZERO)) ||
- !init_fifo(&free[RESERVE_MOVINGGC],
- copygc_reserve, GFP_KERNEL) ||
- !init_fifo(&free[RESERVE_NONE], reserve_none, GFP_KERNEL) ||
- !init_fifo(&free_inc, free_inc_nr, GFP_KERNEL) ||
- !init_heap(&alloc_heap, ALLOC_SCAN_BATCH(ca) << 1, GFP_KERNEL))
+ GFP_KERNEL|__GFP_ZERO))))
goto err;
- buckets->first_bucket = ca->mi.first_bucket;
- buckets->nbuckets = nbuckets;
bucket_gens->first_bucket = ca->mi.first_bucket;
bucket_gens->nbuckets = nbuckets;
percpu_down_write(&c->mark_lock);
}
- old_buckets = bucket_array(ca);
old_bucket_gens = rcu_dereference_protected(ca->bucket_gens, 1);
if (resize) {
- size_t n = min(buckets->nbuckets, old_buckets->nbuckets);
+ size_t n = min(bucket_gens->nbuckets, old_bucket_gens->nbuckets);
- memcpy(buckets->b,
- old_buckets->b,
- n * sizeof(struct bucket));
memcpy(bucket_gens->b,
old_bucket_gens->b,
n);
- memcpy(buckets_nouse,
- ca->buckets_nouse,
- BITS_TO_LONGS(n) * sizeof(unsigned long));
+ if (buckets_nouse)
+ memcpy(buckets_nouse,
+ ca->buckets_nouse,
+ BITS_TO_LONGS(n) * sizeof(unsigned long));
}
- rcu_assign_pointer(ca->buckets[0], buckets);
rcu_assign_pointer(ca->bucket_gens, bucket_gens);
- buckets = old_buckets;
bucket_gens = old_bucket_gens;
swap(ca->buckets_nouse, buckets_nouse);
+ nbuckets = ca->mi.nbuckets;
+
if (resize) {
percpu_up_write(&c->mark_lock);
+ up_write(&ca->bucket_lock);
up_write(&c->gc_lock);
}
- spin_lock(&c->freelist_lock);
- for (i = 0; i < RESERVE_NR; i++) {
- fifo_move(&free[i], &ca->free[i]);
- swap(ca->free[i], free[i]);
- }
- fifo_move(&free_inc, &ca->free_inc);
- swap(ca->free_inc, free_inc);
- spin_unlock(&c->freelist_lock);
-
- /* with gc lock held, alloc_heap can't be in use: */
- swap(ca->alloc_heap, alloc_heap);
-
- nbuckets = ca->mi.nbuckets;
-
- if (resize)
- up_write(&ca->bucket_lock);
-
ret = 0;
err:
- free_heap(&alloc_heap);
- free_fifo(&free_inc);
- for (i = 0; i < RESERVE_NR; i++)
- free_fifo(&free[i]);
kvpfree(buckets_nouse,
BITS_TO_LONGS(nbuckets) * sizeof(unsigned long));
if (bucket_gens)
- call_rcu(&old_buckets->rcu, bucket_gens_free_rcu);
- if (buckets)
- call_rcu(&old_buckets->rcu, buckets_free_rcu);
+ call_rcu(&bucket_gens->rcu, bucket_gens_free_rcu);
return ret;
}
{
unsigned i;
- free_heap(&ca->alloc_heap);
- free_fifo(&ca->free_inc);
- for (i = 0; i < RESERVE_NR; i++)
- free_fifo(&ca->free[i]);
kvpfree(ca->buckets_nouse,
BITS_TO_LONGS(ca->mi.nbuckets) * sizeof(unsigned long));
- kvpfree(rcu_dereference_protected(ca->buckets[0], 1),
- sizeof(struct bucket_array) +
- ca->mi.nbuckets * sizeof(struct bucket));
+ kvpfree(rcu_dereference_protected(ca->bucket_gens, 1),
+ sizeof(struct bucket_gens) + ca->mi.nbuckets);
for (i = 0; i < ARRAY_SIZE(ca->usage); i++)
free_percpu(ca->usage[i]);
projects / bcachefs-tools-debian / blobdiff