+// SPDX-License-Identifier: GPL-2.0
/*
* Code for manipulating bucket marks for garbage collection.
*
* Copyright 2014 Datera, Inc.
- *
- * Bucket states:
- * - free bucket: mark == 0
- * The bucket contains no data and will not be read
- *
- * - allocator bucket: owned_by_allocator == 1
- * The bucket is on a free list, or it is an open bucket
- *
- * - cached bucket: owned_by_allocator == 0 &&
- * dirty_sectors == 0 &&
- * cached_sectors > 0
- * The bucket contains data but may be safely discarded as there are
- * enough replicas of the data on other cache devices, or it has been
- * written back to the backing device
- *
- * - dirty bucket: owned_by_allocator == 0 &&
- * dirty_sectors > 0
- * The bucket contains data that we must not discard (either only copy,
- * or one of the 'main copies' for data requiring multiple replicas)
- *
- * - metadata bucket: owned_by_allocator == 0 && is_metadata == 1
- * This is a btree node, journal or gen/prio bucket
- *
- * Lifecycle:
- *
- * bucket invalidated => bucket on freelist => open bucket =>
- * [dirty bucket =>] cached bucket => bucket invalidated => ...
- *
- * Note that cache promotion can skip the dirty bucket step, as data
- * is copied from a deeper tier to a shallower tier, onto a cached
- * bucket.
- * Note also that a cached bucket can spontaneously become dirty --
- * see below.
- *
- * Only a traversal of the key space can determine whether a bucket is
- * truly dirty or cached.
- *
- * Transitions:
- *
- * - free => allocator: bucket was invalidated
- * - cached => allocator: bucket was invalidated
- *
- * - allocator => dirty: open bucket was filled up
- * - allocator => cached: open bucket was filled up
- * - allocator => metadata: metadata was allocated
- *
- * - dirty => cached: dirty sectors were copied to a deeper tier
- * - dirty => free: dirty sectors were overwritten or moved (copy gc)
- * - cached => free: cached sectors were overwritten
- *
- * - metadata => free: metadata was freed
- *
- * Oddities:
- * - cached => dirty: a device was removed so formerly replicated data
- * is no longer sufficiently replicated
- * - free => cached: cannot happen
- * - free => dirty: cannot happen
- * - free => metadata: cannot happen
*/
#include "bcachefs.h"
-#include "alloc.h"
+#include "alloc_background.h"
+#include "backpointers.h"
+#include "bset.h"
#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"
#include "movinggc.h"
+#include "recovery.h"
+#include "reflink.h"
+#include "replicas.h"
+#include "subvolume.h"
+#include "trace.h"
#include <linux/preempt.h>
-#include <trace/events/bcachefs.h>
-
-#ifdef DEBUG_BUCKETS
-#define lg_local_lock lg_global_lock
-#define lg_local_unlock lg_global_unlock
+static inline void fs_usage_data_type_to_base(struct bch_fs_usage *fs_usage,
+ enum bch_data_type data_type,
+ s64 sectors)
+{
+ switch (data_type) {
+ case BCH_DATA_btree:
+ fs_usage->btree += sectors;
+ break;
+ case BCH_DATA_user:
+ case BCH_DATA_parity:
+ fs_usage->data += sectors;
+ break;
+ case BCH_DATA_cached:
+ fs_usage->cached += sectors;
+ break;
+ default:
+ break;
+ }
+}
-static void bch2_fs_stats_verify(struct bch_fs *c)
+void bch2_fs_usage_initialize(struct bch_fs *c)
{
- struct bch_fs_usage stats =
- __bch2_fs_usage_read(c);
+ struct bch_fs_usage *usage;
+ struct bch_dev *ca;
unsigned i;
- for (i = 0; i < ARRAY_SIZE(stats.s); i++) {
- if ((s64) stats.s[i].data[S_META] < 0)
- panic("replicas %u meta underflow: %lli\n",
- i + 1, stats.s[i].data[S_META]);
+ percpu_down_write(&c->mark_lock);
+ usage = c->usage_base;
- if ((s64) stats.s[i].data[S_DIRTY] < 0)
- panic("replicas %u dirty underflow: %lli\n",
- i + 1, stats.s[i].data[S_DIRTY]);
+ for (i = 0; i < ARRAY_SIZE(c->usage); i++)
+ bch2_fs_usage_acc_to_base(c, i);
- if ((s64) stats.s[i].persistent_reserved < 0)
- panic("replicas %u reserved underflow: %lli\n",
- i + 1, stats.s[i].persistent_reserved);
+ for (i = 0; i < BCH_REPLICAS_MAX; i++)
+ usage->reserved += usage->persistent_reserved[i];
+
+ for (i = 0; i < c->replicas.nr; i++) {
+ struct bch_replicas_entry *e =
+ cpu_replicas_entry(&c->replicas, i);
+
+ fs_usage_data_type_to_base(usage, e->data_type, usage->replicas[i]);
+ }
+
+ for_each_member_device(ca, c, i) {
+ struct bch_dev_usage dev = bch2_dev_usage_read(ca);
+
+ usage->hidden += (dev.d[BCH_DATA_sb].buckets +
+ dev.d[BCH_DATA_journal].buckets) *
+ ca->mi.bucket_size;
}
- if ((s64) stats.online_reserved < 0)
- panic("sectors_online_reserved underflow: %lli\n",
- stats.online_reserved);
+ percpu_up_write(&c->mark_lock);
}
-static void bch2_dev_stats_verify(struct bch_dev *ca)
+static inline struct bch_dev_usage *dev_usage_ptr(struct bch_dev *ca,
+ unsigned journal_seq,
+ bool gc)
{
- struct bch_dev_usage stats =
- __bch2_dev_usage_read(ca);
- u64 n = ca->mi.nbuckets - ca->mi.first_bucket;
- unsigned i;
+ BUG_ON(!gc && !journal_seq);
- for (i = 0; i < ARRAY_SIZE(stats.buckets); i++)
- BUG_ON(stats.buckets[i] > n);
- BUG_ON(stats.buckets_alloc > n);
- BUG_ON(stats.buckets_unavailable > n);
+ return this_cpu_ptr(gc
+ ? ca->usage_gc
+ : ca->usage[journal_seq & JOURNAL_BUF_MASK]);
}
-static void bch2_disk_reservations_verify(struct bch_fs *c, int flags)
+void bch2_dev_usage_read_fast(struct bch_dev *ca, struct bch_dev_usage *usage)
{
- if (!(flags & BCH_DISK_RESERVATION_NOFAIL)) {
- u64 used = __bch2_fs_sectors_used(c);
- u64 cached = 0;
- u64 avail = atomic64_read(&c->sectors_available);
- int cpu;
+ struct bch_fs *c = ca->fs;
+ unsigned seq, i, u64s = dev_usage_u64s();
- for_each_possible_cpu(cpu)
- cached += per_cpu_ptr(c->usage_percpu, cpu)->available_cache;
+ do {
+ seq = read_seqcount_begin(&c->usage_lock);
+ memcpy(usage, ca->usage_base, u64s * sizeof(u64));
+ for (i = 0; i < ARRAY_SIZE(ca->usage); i++)
+ acc_u64s_percpu((u64 *) usage, (u64 __percpu *) ca->usage[i], u64s);
+ } while (read_seqcount_retry(&c->usage_lock, seq));
+}
- if (used + avail + cached > c->capacity)
- panic("used %llu avail %llu cached %llu capacity %llu\n",
- used, avail, cached, c->capacity);
- }
+u64 bch2_fs_usage_read_one(struct bch_fs *c, u64 *v)
+{
+ ssize_t offset = v - (u64 *) c->usage_base;
+ unsigned i, seq;
+ u64 ret;
+
+ BUG_ON(offset < 0 || offset >= fs_usage_u64s(c));
+ percpu_rwsem_assert_held(&c->mark_lock);
+
+ do {
+ seq = read_seqcount_begin(&c->usage_lock);
+ ret = *v;
+
+ for (i = 0; i < ARRAY_SIZE(c->usage); i++)
+ ret += percpu_u64_get((u64 __percpu *) c->usage[i] + offset);
+ } while (read_seqcount_retry(&c->usage_lock, seq));
+
+ return ret;
}
-#else
+struct bch_fs_usage_online *bch2_fs_usage_read(struct bch_fs *c)
+{
+ struct bch_fs_usage_online *ret;
+ unsigned nr_replicas = READ_ONCE(c->replicas.nr);
+ unsigned seq, i;
+retry:
+ ret = kmalloc(__fs_usage_online_u64s(nr_replicas) * sizeof(u64), GFP_KERNEL);
+ if (unlikely(!ret))
+ return NULL;
+
+ percpu_down_read(&c->mark_lock);
+
+ if (nr_replicas != c->replicas.nr) {
+ nr_replicas = c->replicas.nr;
+ percpu_up_read(&c->mark_lock);
+ kfree(ret);
+ goto retry;
+ }
-static void bch2_fs_stats_verify(struct bch_fs *c) {}
-static void bch2_dev_stats_verify(struct bch_dev *ca) {}
-static void bch2_disk_reservations_verify(struct bch_fs *c, int flags) {}
+ ret->online_reserved = percpu_u64_get(c->online_reserved);
-#endif
+ do {
+ seq = read_seqcount_begin(&c->usage_lock);
+ unsafe_memcpy(&ret->u, c->usage_base,
+ __fs_usage_u64s(nr_replicas) * sizeof(u64),
+ "embedded variable length struct");
+ for (i = 0; i < ARRAY_SIZE(c->usage); i++)
+ acc_u64s_percpu((u64 *) &ret->u, (u64 __percpu *) c->usage[i],
+ __fs_usage_u64s(nr_replicas));
+ } while (read_seqcount_retry(&c->usage_lock, seq));
-/*
- * Clear journal_seq_valid for buckets for which it's not needed, to prevent
- * wraparound:
- */
-void bch2_bucket_seq_cleanup(struct bch_fs *c)
+ return ret;
+}
+
+void bch2_fs_usage_acc_to_base(struct bch_fs *c, unsigned idx)
{
- u16 last_seq_ondisk = c->journal.last_seq_ondisk;
struct bch_dev *ca;
- struct bucket_array *buckets;
- struct bucket *g;
- struct bucket_mark m;
- unsigned i;
+ unsigned i, u64s = fs_usage_u64s(c);
- for_each_member_device(ca, c, i) {
- down_read(&ca->bucket_lock);
- buckets = bucket_array(ca);
+ BUG_ON(idx >= ARRAY_SIZE(c->usage));
- for_each_bucket(g, buckets) {
- bucket_cmpxchg(g, m, ({
- if (!m.journal_seq_valid ||
- bucket_needs_journal_commit(m, last_seq_ondisk))
- break;
+ preempt_disable();
+ write_seqcount_begin(&c->usage_lock);
- m.journal_seq_valid = 0;
- }));
- }
- up_read(&ca->bucket_lock);
+ acc_u64s_percpu((u64 *) c->usage_base,
+ (u64 __percpu *) c->usage[idx], u64s);
+ percpu_memset(c->usage[idx], 0, u64s * sizeof(u64));
+
+ rcu_read_lock();
+ for_each_member_device_rcu(ca, c, i, NULL) {
+ u64s = dev_usage_u64s();
+
+ acc_u64s_percpu((u64 *) ca->usage_base,
+ (u64 __percpu *) ca->usage[idx], u64s);
+ percpu_memset(ca->usage[idx], 0, u64s * sizeof(u64));
}
+ rcu_read_unlock();
+
+ write_seqcount_end(&c->usage_lock);
+ preempt_enable();
}
-#define bch2_usage_add(_acc, _stats) \
-do { \
- typeof(_acc) _a = (_acc), _s = (_stats); \
- unsigned i; \
- \
- for (i = 0; i < sizeof(*_a) / sizeof(u64); i++) \
- ((u64 *) (_a))[i] += ((u64 *) (_s))[i]; \
-} while (0)
+void bch2_fs_usage_to_text(struct printbuf *out,
+ struct bch_fs *c,
+ struct bch_fs_usage_online *fs_usage)
+{
+ unsigned i;
-#define bch2_usage_read_raw(_stats) \
-({ \
- typeof(*this_cpu_ptr(_stats)) _acc; \
- int cpu; \
- \
- memset(&_acc, 0, sizeof(_acc)); \
- \
- for_each_possible_cpu(cpu) \
- bch2_usage_add(&_acc, per_cpu_ptr((_stats), cpu)); \
- \
- _acc; \
-})
+ prt_printf(out, "capacity:\t\t\t%llu\n", c->capacity);
+
+ prt_printf(out, "hidden:\t\t\t\t%llu\n",
+ fs_usage->u.hidden);
+ prt_printf(out, "data:\t\t\t\t%llu\n",
+ fs_usage->u.data);
+ prt_printf(out, "cached:\t\t\t\t%llu\n",
+ fs_usage->u.cached);
+ prt_printf(out, "reserved:\t\t\t%llu\n",
+ fs_usage->u.reserved);
+ prt_printf(out, "nr_inodes:\t\t\t%llu\n",
+ fs_usage->u.nr_inodes);
+ prt_printf(out, "online reserved:\t\t%llu\n",
+ fs_usage->online_reserved);
+
+ for (i = 0;
+ i < ARRAY_SIZE(fs_usage->u.persistent_reserved);
+ i++) {
+ prt_printf(out, "%u replicas:\n", i + 1);
+ prt_printf(out, "\treserved:\t\t%llu\n",
+ fs_usage->u.persistent_reserved[i]);
+ }
+
+ for (i = 0; i < c->replicas.nr; i++) {
+ struct bch_replicas_entry *e =
+ cpu_replicas_entry(&c->replicas, i);
+
+ prt_printf(out, "\t");
+ bch2_replicas_entry_to_text(out, e);
+ prt_printf(out, ":\t%llu\n", fs_usage->u.replicas[i]);
+ }
+}
-#define bch2_usage_read_cached(_c, _cached, _uncached) \
-({ \
- typeof(_cached) _ret; \
- unsigned _seq; \
- \
- do { \
- _seq = read_seqcount_begin(&(_c)->gc_pos_lock); \
- _ret = (_c)->gc_pos.phase == GC_PHASE_DONE \
- ? bch2_usage_read_raw(_uncached) \
- : (_cached); \
- } while (read_seqcount_retry(&(_c)->gc_pos_lock, _seq)); \
- \
- _ret; \
-})
+static u64 reserve_factor(u64 r)
+{
+ return r + (round_up(r, (1 << RESERVE_FACTOR)) >> RESERVE_FACTOR);
+}
-struct bch_dev_usage __bch2_dev_usage_read(struct bch_dev *ca)
+u64 bch2_fs_sectors_used(struct bch_fs *c, struct bch_fs_usage_online *fs_usage)
{
- return bch2_usage_read_raw(ca->usage_percpu);
+ return min(fs_usage->u.hidden +
+ fs_usage->u.btree +
+ fs_usage->u.data +
+ reserve_factor(fs_usage->u.reserved +
+ fs_usage->online_reserved),
+ c->capacity);
}
-struct bch_dev_usage bch2_dev_usage_read(struct bch_fs *c, struct bch_dev *ca)
+static struct bch_fs_usage_short
+__bch2_fs_usage_read_short(struct bch_fs *c)
{
- return bch2_usage_read_cached(c, ca->usage_cached, ca->usage_percpu);
+ struct bch_fs_usage_short ret;
+ u64 data, reserved;
+
+ ret.capacity = c->capacity -
+ bch2_fs_usage_read_one(c, &c->usage_base->hidden);
+
+ data = bch2_fs_usage_read_one(c, &c->usage_base->data) +
+ bch2_fs_usage_read_one(c, &c->usage_base->btree);
+ reserved = bch2_fs_usage_read_one(c, &c->usage_base->reserved) +
+ percpu_u64_get(c->online_reserved);
+
+ ret.used = min(ret.capacity, data + reserve_factor(reserved));
+ ret.free = ret.capacity - ret.used;
+
+ ret.nr_inodes = bch2_fs_usage_read_one(c, &c->usage_base->nr_inodes);
+
+ return ret;
}
-struct bch_fs_usage
-__bch2_fs_usage_read(struct bch_fs *c)
+struct bch_fs_usage_short
+bch2_fs_usage_read_short(struct bch_fs *c)
{
- return bch2_usage_read_raw(c->usage_percpu);
+ struct bch_fs_usage_short ret;
+
+ percpu_down_read(&c->mark_lock);
+ ret = __bch2_fs_usage_read_short(c);
+ percpu_up_read(&c->mark_lock);
+
+ return ret;
}
-struct bch_fs_usage
-bch2_fs_usage_read(struct bch_fs *c)
+void bch2_dev_usage_init(struct bch_dev *ca)
{
- return bch2_usage_read_cached(c,
- c->usage_cached,
- c->usage_percpu);
+ ca->usage_base->d[BCH_DATA_free].buckets = ca->mi.nbuckets - ca->mi.first_bucket;
}
-struct fs_usage_sum {
- u64 data;
- u64 reserved;
-};
+static inline int bucket_sectors_fragmented(struct bch_dev *ca,
+ struct bch_alloc_v4 a)
+{
+ return a.dirty_sectors
+ ? max(0, (int) ca->mi.bucket_size - (int) a.dirty_sectors)
+ : 0;
+}
-static inline struct fs_usage_sum __fs_usage_sum(struct bch_fs_usage stats)
+static void bch2_dev_usage_update(struct bch_fs *c, struct bch_dev *ca,
+ struct bch_alloc_v4 old,
+ struct bch_alloc_v4 new,
+ u64 journal_seq, bool gc)
{
- struct fs_usage_sum sum = { 0 };
- unsigned i;
+ struct bch_fs_usage *fs_usage;
+ struct bch_dev_usage *u;
- for (i = 0; i < ARRAY_SIZE(stats.s); i++) {
- sum.data += (stats.s[i].data[S_META] +
- stats.s[i].data[S_DIRTY]) * (i + 1);
- sum.reserved += stats.s[i].persistent_reserved * (i + 1);
- }
+ preempt_disable();
+ fs_usage = fs_usage_ptr(c, journal_seq, gc);
- sum.reserved += stats.online_reserved;
- return sum;
-}
+ 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;
-#define RESERVE_FACTOR 6
+ u = dev_usage_ptr(ca, journal_seq, gc);
-static u64 reserve_factor(u64 r)
+ u->d[old.data_type].buckets--;
+ u->d[new.data_type].buckets++;
+
+ 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 += 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();
+}
+
+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)
{
- return r + (round_up(r, (1 << RESERVE_FACTOR)) >> RESERVE_FACTOR);
+ 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 u64 avail_factor(u64 r)
+static inline int __update_replicas(struct bch_fs *c,
+ struct bch_fs_usage *fs_usage,
+ struct bch_replicas_entry *r,
+ s64 sectors)
{
- return (r << RESERVE_FACTOR) / (1 << RESERVE_FACTOR) + 1;
+ int idx = bch2_replicas_entry_idx(c, r);
+
+ if (idx < 0)
+ return -1;
+
+ fs_usage_data_type_to_base(fs_usage, r->data_type, sectors);
+ fs_usage->replicas[idx] += sectors;
+ return 0;
}
-u64 __bch2_fs_sectors_used(struct bch_fs *c, struct bch_fs_usage stats)
+static inline int update_replicas(struct bch_fs *c, struct bkey_s_c k,
+ struct bch_replicas_entry *r, s64 sectors,
+ unsigned journal_seq, bool gc)
{
- struct fs_usage_sum sum = __fs_usage_sum(stats);
+ struct bch_fs_usage *fs_usage;
+ int idx, ret = 0;
+ struct printbuf buf = PRINTBUF;
+
+ percpu_down_read(&c->mark_lock);
+
+ idx = bch2_replicas_entry_idx(c, r);
+ if (idx < 0 &&
+ fsck_err(c, ptr_to_missing_replicas_entry,
+ "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);
+ percpu_down_read(&c->mark_lock);
+
+ if (ret)
+ goto err;
+ idx = bch2_replicas_entry_idx(c, r);
+ }
+ if (idx < 0) {
+ ret = -1;
+ goto err;
+ }
- return sum.data + reserve_factor(sum.reserved);
+ preempt_disable();
+ fs_usage = fs_usage_ptr(c, journal_seq, gc);
+ fs_usage_data_type_to_base(fs_usage, r->data_type, sectors);
+ fs_usage->replicas[idx] += sectors;
+ preempt_enable();
+err:
+fsck_err:
+ percpu_up_read(&c->mark_lock);
+ printbuf_exit(&buf);
+ return ret;
}
-u64 bch2_fs_sectors_used(struct bch_fs *c, struct bch_fs_usage stats)
+static inline int update_cached_sectors(struct bch_fs *c,
+ struct bkey_s_c k,
+ unsigned dev, s64 sectors,
+ unsigned journal_seq, bool gc)
{
- return min(c->capacity, __bch2_fs_sectors_used(c, stats));
+ struct bch_replicas_padded r;
+
+ bch2_replicas_entry_cached(&r.e, dev);
+
+ return update_replicas(c, k, &r.e, sectors, journal_seq, gc);
}
-u64 bch2_fs_sectors_free(struct bch_fs *c, struct bch_fs_usage stats)
+static int __replicas_deltas_realloc(struct btree_trans *trans, unsigned more,
+ gfp_t gfp)
{
- return avail_factor(c->capacity - bch2_fs_sectors_used(c, stats));
+ struct replicas_delta_list *d = trans->fs_usage_deltas;
+ unsigned new_size = d ? (d->size + more) * 2 : 128;
+ unsigned alloc_size = sizeof(*d) + new_size;
+
+ WARN_ON_ONCE(alloc_size > REPLICAS_DELTA_LIST_MAX);
+
+ if (!d || d->used + more > d->size) {
+ d = krealloc(d, alloc_size, gfp|__GFP_ZERO);
+
+ if (unlikely(!d)) {
+ if (alloc_size > REPLICAS_DELTA_LIST_MAX)
+ return -ENOMEM;
+
+ d = mempool_alloc(&trans->c->replicas_delta_pool, gfp);
+ if (!d)
+ return -ENOMEM;
+
+ memset(d, 0, REPLICAS_DELTA_LIST_MAX);
+
+ if (trans->fs_usage_deltas)
+ memcpy(d, trans->fs_usage_deltas,
+ trans->fs_usage_deltas->size + sizeof(*d));
+
+ new_size = REPLICAS_DELTA_LIST_MAX - sizeof(*d);
+ kfree(trans->fs_usage_deltas);
+ }
+
+ d->size = new_size;
+ trans->fs_usage_deltas = d;
+ }
+
+ return 0;
}
-static inline int is_unavailable_bucket(struct bucket_mark m)
+int bch2_replicas_deltas_realloc(struct btree_trans *trans, unsigned more)
{
- return !is_available_bucket(m);
+ return allocate_dropping_locks_errcode(trans,
+ __replicas_deltas_realloc(trans, more, _gfp));
}
-static inline int is_fragmented_bucket(struct bucket_mark m,
- struct bch_dev *ca)
+static inline int update_replicas_list(struct btree_trans *trans,
+ struct bch_replicas_entry *r,
+ s64 sectors)
{
- if (!m.owned_by_allocator &&
- m.data_type == BCH_DATA_USER &&
- bucket_sectors_used(m))
- return max_t(int, 0, (int) ca->mi.bucket_size -
- bucket_sectors_used(m));
+ struct replicas_delta_list *d;
+ struct replicas_delta *n;
+ unsigned b;
+ int ret;
+
+ if (!sectors)
+ return 0;
+
+ b = replicas_entry_bytes(r) + 8;
+ ret = bch2_replicas_deltas_realloc(trans, b);
+ if (ret)
+ return ret;
+
+ d = trans->fs_usage_deltas;
+ n = (void *) d->d + d->used;
+ n->delta = sectors;
+ unsafe_memcpy((void *) n + offsetof(struct replicas_delta, r),
+ r, replicas_entry_bytes(r),
+ "flexible array member embedded in strcuct with padding");
+ bch2_replicas_entry_sort(&n->r);
+ d->used += b;
return 0;
}
-static inline enum bch_data_type bucket_type(struct bucket_mark m)
+static inline int update_cached_sectors_list(struct btree_trans *trans,
+ unsigned dev, s64 sectors)
{
- return m.cached_sectors && !m.dirty_sectors
- ? BCH_DATA_CACHED
- : m.data_type;
+ struct bch_replicas_padded r;
+
+ bch2_replicas_entry_cached(&r.e, dev);
+
+ return update_replicas_list(trans, &r.e, sectors);
}
-static bool bucket_became_unavailable(struct bch_fs *c,
- struct bucket_mark old,
- struct bucket_mark new)
+int bch2_mark_alloc(struct btree_trans *trans,
+ enum btree_id btree, unsigned level,
+ struct bkey_s_c old, struct bkey_s_c new,
+ unsigned flags)
{
- return is_available_bucket(old) &&
- !is_available_bucket(new) &&
- (!c || c->gc_pos.phase == GC_PHASE_DONE);
+ bool gc = flags & BTREE_TRIGGER_GC;
+ u64 journal_seq = trans->journal_res.seq;
+ u64 bucket_journal_seq;
+ struct bch_fs *c = trans->c;
+ struct bch_alloc_v4 old_a_convert, new_a_convert;
+ const struct bch_alloc_v4 *old_a, *new_a;
+ struct bch_dev *ca;
+ int ret = 0;
+
+ /*
+ * alloc btree is read in by bch2_alloc_read, not gc:
+ */
+ if ((flags & BTREE_TRIGGER_GC) &&
+ !(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);
+
+ old_a = bch2_alloc_to_v4(old, &old_a_convert);
+ new_a = bch2_alloc_to_v4(new, &new_a_convert);
+
+ bucket_journal_seq = new_a->journal_seq;
+
+ if ((flags & BTREE_TRIGGER_INSERT) &&
+ 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;
+
+ EBUG_ON(!journal_seq);
+
+ /*
+ * If the btree updates referring to a bucket weren't flushed
+ * 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 = bucket_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;
+ }
+
+ if (!data_type_is_empty(old_a->data_type) &&
+ data_type_is_empty(new_a->data_type) &&
+ bucket_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,
+ bucket_journal_seq);
+ if (ret) {
+ bch2_fs_fatal_error(c,
+ "error setting bucket_needs_journal_commit: %i", ret);
+ return ret;
+ }
+ }
+
+ percpu_down_read(&c->mark_lock);
+ if (!gc && new_a->gen != old_a->gen)
+ *bucket_gen(ca, new.k->p.offset) = new_a->gen;
+
+ bch2_dev_usage_update(c, ca, *old_a, *new_a, journal_seq, gc);
+
+ if (gc) {
+ struct bucket *g = gc_bucket(ca, new.k->p.offset);
+
+ bucket_lock(g);
+
+ 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;
+
+ bucket_unlock(g);
+ }
+ percpu_up_read(&c->mark_lock);
+
+ /*
+ * need to know if we're getting called from the invalidate path or
+ * not:
+ */
+
+ if ((flags & BTREE_TRIGGER_BUCKET_INVALIDATE) &&
+ old_a->cached_sectors) {
+ ret = update_cached_sectors(c, new, ca->dev_idx,
+ -((s64) old_a->cached_sectors),
+ journal_seq, gc);
+ if (ret) {
+ bch2_fs_fatal_error(c, "%s(): no replicas entry while updating cached sectors",
+ __func__);
+ return ret;
+ }
+ }
+
+ 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 &&
+ (!bucket_journal_seq || bucket_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);
+
+ return 0;
}
-void bch2_fs_usage_apply(struct bch_fs *c,
- struct bch_fs_usage *stats,
- struct disk_reservation *disk_res,
- struct gc_pos gc_pos)
+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 fs_usage_sum sum = __fs_usage_sum(*stats);
- s64 added = sum.data + sum.reserved;
+ struct bucket old, new, *g;
+ int ret = 0;
+
+ BUG_ON(!(flags & BTREE_TRIGGER_GC));
+ BUG_ON(data_type != BCH_DATA_sb &&
+ data_type != BCH_DATA_journal);
/*
- * Not allowed to reduce sectors_available except by getting a
- * reservation:
+ * Backup superblock might be past the end of our normal usable space:
*/
- BUG_ON(added > (s64) (disk_res ? disk_res->sectors : 0));
+ if (b >= ca->mi.nbuckets)
+ return 0;
- if (added > 0) {
- disk_res->sectors -= added;
- stats->online_reserved -= added;
+ percpu_down_read(&c->mark_lock);
+ g = gc_bucket(ca, b);
+
+ 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;
}
- percpu_down_read_preempt_disable(&c->usage_lock);
- /* online_reserved not subject to gc: */
- this_cpu_ptr(c->usage_percpu)->online_reserved +=
- stats->online_reserved;
- stats->online_reserved = 0;
- if (!gc_will_visit(c, gc_pos))
- bch2_usage_add(this_cpu_ptr(c->usage_percpu), stats);
+ 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;
+}
- bch2_fs_stats_verify(c);
- percpu_up_read_preempt_enable(&c->usage_lock);
+static int check_bucket_ref(struct btree_trans *trans,
+ struct bkey_s_c k,
+ const struct bch_extent_ptr *ptr,
+ s64 sectors, enum bch_data_type ptr_data_type,
+ u8 b_gen, u8 bucket_data_type,
+ u32 dirty_sectors, u32 cached_sectors)
+{
+ struct bch_fs *c = trans->c;
+ struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
+ size_t bucket_nr = PTR_BUCKET_NR(ca, ptr);
+ u32 bucket_sectors = !ptr->cached
+ ? dirty_sectors
+ : cached_sectors;
+ struct printbuf buf = PRINTBUF;
+ int ret = 0;
+
+ if (bucket_data_type == BCH_DATA_cached)
+ bucket_data_type = BCH_DATA_user;
+
+ if ((bucket_data_type == BCH_DATA_stripe && ptr_data_type == BCH_DATA_user) ||
+ (bucket_data_type == BCH_DATA_user && ptr_data_type == BCH_DATA_stripe))
+ bucket_data_type = ptr_data_type = BCH_DATA_stripe;
+
+ if (gen_after(ptr->gen, b_gen)) {
+ bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
+ BCH_FSCK_ERR_ptr_gen_newer_than_bucket_gen,
+ "bucket %u:%zu gen %u data type %s: ptr gen %u newer than bucket gen\n"
+ "while marking %s",
+ ptr->dev, bucket_nr, b_gen,
+ bch2_data_types[bucket_data_type ?: ptr_data_type],
+ ptr->gen,
+ (bch2_bkey_val_to_text(&buf, c, k), buf.buf));
+ ret = -EIO;
+ goto err;
+ }
+
+ if (gen_cmp(b_gen, ptr->gen) > BUCKET_GC_GEN_MAX) {
+ bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
+ BCH_FSCK_ERR_ptr_too_stale,
+ "bucket %u:%zu gen %u data type %s: ptr gen %u too stale\n"
+ "while marking %s",
+ ptr->dev, bucket_nr, b_gen,
+ bch2_data_types[bucket_data_type ?: ptr_data_type],
+ ptr->gen,
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf));
+ ret = -EIO;
+ goto err;
+ }
+
+ if (b_gen != ptr->gen && !ptr->cached) {
+ bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
+ BCH_FSCK_ERR_stale_dirty_ptr,
+ "bucket %u:%zu gen %u (mem gen %u) data type %s: stale dirty ptr (gen %u)\n"
+ "while marking %s",
+ ptr->dev, bucket_nr, b_gen,
+ *bucket_gen(ca, bucket_nr),
+ bch2_data_types[bucket_data_type ?: ptr_data_type],
+ ptr->gen,
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf));
+ ret = -EIO;
+ goto err;
+ }
+
+ if (b_gen != ptr->gen) {
+ ret = 1;
+ goto out;
+ }
+
+ 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,
+ BCH_FSCK_ERR_ptr_bucket_data_type_mismatch,
+ "bucket %u:%zu gen %u different types of data in same bucket: %s, %s\n"
+ "while marking %s",
+ ptr->dev, bucket_nr, b_gen,
+ bch2_data_types[bucket_data_type],
+ bch2_data_types[ptr_data_type],
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf));
+ ret = -EIO;
+ goto err;
+ }
- memset(stats, 0, sizeof(*stats));
+ if ((u64) bucket_sectors + sectors > U32_MAX) {
+ bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
+ BCH_FSCK_ERR_bucket_sector_count_overflow,
+ "bucket %u:%zu gen %u data type %s sector count overflow: %u + %lli > U32_MAX\n"
+ "while marking %s",
+ ptr->dev, bucket_nr, b_gen,
+ bch2_data_types[bucket_data_type ?: ptr_data_type],
+ bucket_sectors, sectors,
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf));
+ ret = -EIO;
+ goto err;
+ }
+out:
+ printbuf_exit(&buf);
+ return ret;
+err:
+ bch2_dump_trans_updates(trans);
+ goto out;
}
-static void bch2_dev_usage_update(struct bch_fs *c, struct bch_dev *ca,
- struct bucket_mark old, struct bucket_mark new)
+static int mark_stripe_bucket(struct btree_trans *trans,
+ struct bkey_s_c k,
+ unsigned ptr_idx,
+ unsigned flags)
{
- struct bch_dev_usage *dev_usage;
+ 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;
+ 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;
+ struct printbuf buf = PRINTBUF;
+ int ret = 0;
+
+ BUG_ON(!(flags & BTREE_TRIGGER_GC));
- if (c)
- percpu_rwsem_assert_held(&c->usage_lock);
+ /* * XXX doesn't handle deletion */
- if (old.data_type && new.data_type &&
- old.data_type != new.data_type) {
- BUG_ON(!c);
+ 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,
- "different types of data in same bucket: %s, %s",
- bch2_data_types[old.data_type],
- bch2_data_types[new.data_type]);
+ "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;
+ goto err;
}
- dev_usage = this_cpu_ptr(ca->usage_percpu);
+ bucket_lock(g);
+ old = *g;
+
+ ret = check_bucket_ref(trans, k, ptr, sectors, data_type,
+ g->gen, g->data_type,
+ g->dirty_sectors, g->cached_sectors);
+ if (ret)
+ goto err;
- dev_usage->buckets[bucket_type(old)]--;
- dev_usage->buckets[bucket_type(new)]++;
+ g->data_type = data_type;
+ g->dirty_sectors += sectors;
- dev_usage->buckets_alloc +=
- (int) new.owned_by_allocator - (int) old.owned_by_allocator;
- dev_usage->buckets_unavailable +=
- is_unavailable_bucket(new) - is_unavailable_bucket(old);
+ g->stripe = k.k->p.offset;
+ g->stripe_redundancy = s->nr_redundant;
+ 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);
+ printbuf_exit(&buf);
+ return ret;
+}
- dev_usage->sectors[old.data_type] -= old.dirty_sectors;
- dev_usage->sectors[new.data_type] += new.dirty_sectors;
- dev_usage->sectors[BCH_DATA_CACHED] +=
- (int) new.cached_sectors - (int) old.cached_sectors;
- dev_usage->sectors_fragmented +=
- is_fragmented_bucket(new, ca) - is_fragmented_bucket(old, ca);
+static int __mark_pointer(struct btree_trans *trans,
+ 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,
+ u32 *dirty_sectors, u32 *cached_sectors)
+{
+ u32 *dst_sectors = !ptr->cached
+ ? dirty_sectors
+ : cached_sectors;
+ int ret = check_bucket_ref(trans, k, ptr, sectors, ptr_data_type,
+ bucket_gen, *bucket_data_type,
+ *dirty_sectors, *cached_sectors);
+
+ if (ret)
+ return ret;
+
+ *dst_sectors += sectors;
+ *bucket_data_type = *dirty_sectors || *cached_sectors
+ ? ptr_data_type : 0;
+ return 0;
+}
- if (!is_available_bucket(old) && is_available_bucket(new))
- bch2_wake_allocator(ca);
+static int bch2_mark_pointer(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c k,
+ struct extent_ptr_decoded p,
+ s64 sectors,
+ unsigned flags)
+{
+ u64 journal_seq = trans->journal_res.seq;
+ struct bch_fs *c = trans->c;
+ struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev);
+ struct bucket old, new, *g;
+ enum bch_data_type data_type = bkey_ptr_data_type(btree_id, level, k, p);
+ u8 bucket_data_type;
+ int ret = 0;
+
+ BUG_ON(!(flags & BTREE_TRIGGER_GC));
+
+ percpu_down_read(&c->mark_lock);
+ g = PTR_GC_BUCKET(ca, &p.ptr);
+ 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);
- bch2_dev_stats_verify(ca);
+ return ret;
}
-#define bucket_data_cmpxchg(c, ca, g, new, expr) \
-({ \
- struct bucket_mark _old = bucket_cmpxchg(g, new, expr); \
- \
- bch2_dev_usage_update(c, ca, _old, new); \
- _old; \
-})
+static int bch2_mark_stripe_ptr(struct btree_trans *trans,
+ struct bkey_s_c k,
+ struct bch_extent_stripe_ptr p,
+ enum bch_data_type data_type,
+ s64 sectors,
+ unsigned flags)
+{
+ struct bch_fs *c = trans->c;
+ struct bch_replicas_padded r;
+ struct gc_stripe *m;
+
+ BUG_ON(!(flags & BTREE_TRIGGER_GC));
+
+ m = genradix_ptr_alloc(&c->gc_stripes, p.idx, GFP_KERNEL);
+ if (!m) {
+ bch_err(c, "error allocating memory for gc_stripes, idx %llu",
+ (u64) p.idx);
+ return -BCH_ERR_ENOMEM_mark_stripe_ptr;
+ }
+
+ mutex_lock(&c->ec_stripes_heap_lock);
+
+ if (!m || !m->alive) {
+ mutex_unlock(&c->ec_stripes_heap_lock);
+ bch_err_ratelimited(c, "pointer to nonexistent stripe %llu",
+ (u64) p.idx);
+ bch2_inconsistent_error(c);
+ return -EIO;
+ }
-bool bch2_invalidate_bucket(struct bch_fs *c, struct bch_dev *ca,
- size_t b, struct bucket_mark *old)
+ m->block_sectors[p.block] += sectors;
+
+ r = m->r;
+ mutex_unlock(&c->ec_stripes_heap_lock);
+
+ r.e.data_type = data_type;
+ update_replicas(c, k, &r.e, sectors, trans->journal_res.seq, true);
+
+ return 0;
+}
+
+static int __mark_extent(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c k, unsigned flags)
{
- struct bucket *g;
- struct bucket_mark new;
+ u64 journal_seq = trans->journal_res.seq;
+ struct bch_fs *c = trans->c;
+ struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
+ const union bch_extent_entry *entry;
+ struct extent_ptr_decoded p;
+ struct bch_replicas_padded r;
+ enum bch_data_type data_type = bkey_is_btree_ptr(k.k)
+ ? BCH_DATA_btree
+ : BCH_DATA_user;
+ s64 sectors = bkey_is_btree_ptr(k.k)
+ ? btree_sectors(c)
+ : k.k->size;
+ s64 dirty_sectors = 0;
+ bool stale;
+ int ret;
+
+ BUG_ON(!(flags & BTREE_TRIGGER_GC));
+
+ r.e.data_type = data_type;
+ r.e.nr_devs = 0;
+ r.e.nr_required = 1;
+
+ bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
+ s64 disk_sectors = ptr_disk_sectors(sectors, p);
+
+ if (flags & BTREE_TRIGGER_OVERWRITE)
+ disk_sectors = -disk_sectors;
+
+ ret = bch2_mark_pointer(trans, btree_id, level, k, p, disk_sectors, flags);
+ if (ret < 0)
+ return ret;
+
+ stale = ret > 0;
+
+ if (p.ptr.cached) {
+ if (!stale) {
+ ret = update_cached_sectors(c, k, p.ptr.dev,
+ disk_sectors, journal_seq, true);
+ if (ret) {
+ bch2_fs_fatal_error(c, "%s(): no replicas entry while updating cached sectors",
+ __func__);
+ return ret;
+ }
+ }
+ } else if (!p.has_ec) {
+ dirty_sectors += disk_sectors;
+ r.e.devs[r.e.nr_devs++] = p.ptr.dev;
+ } else {
+ ret = bch2_mark_stripe_ptr(trans, k, p.ec, data_type,
+ disk_sectors, flags);
+ if (ret)
+ return ret;
+
+ /*
+ * There may be other dirty pointers in this extent, but
+ * if so they're not required for mounting if we have an
+ * erasure coded pointer in this extent:
+ */
+ r.e.nr_required = 0;
+ }
+ }
+
+ if (r.e.nr_devs) {
+ ret = update_replicas(c, k, &r.e, dirty_sectors, journal_seq, true);
+ if (ret) {
+ struct printbuf buf = PRINTBUF;
+
+ bch2_bkey_val_to_text(&buf, c, k);
+ bch2_fs_fatal_error(c, "%s(): no replicas entry for %s", __func__, buf.buf);
+ printbuf_exit(&buf);
+ return ret;
+ }
+ }
+
+ return 0;
+}
+
+int bch2_mark_extent(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c old, struct bkey_s_c new,
+ unsigned flags)
+{
+ return mem_trigger_run_overwrite_then_insert(__mark_extent, trans, btree_id, level, old, new, flags);
+}
+
+int bch2_mark_stripe(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ 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;
+ u64 idx = new.k->p.offset;
+ 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;
+ unsigned i;
+ int ret;
+
+ BUG_ON(gc && old_s);
+
+ if (!gc) {
+ struct stripe *m = genradix_ptr(&c->stripes, idx);
+
+ if (!m) {
+ struct printbuf buf1 = PRINTBUF;
+ struct printbuf buf2 = PRINTBUF;
- percpu_rwsem_assert_held(&c->usage_lock);
+ 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.buf, buf2.buf);
+ printbuf_exit(&buf2);
+ printbuf_exit(&buf1);
+ bch2_inconsistent_error(c);
+ return -1;
+ }
+
+ if (!new_s) {
+ bch2_stripes_heap_del(c, m, idx);
+
+ 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 (i = 0; i < new_s->nr_blocks; i++)
+ m->blocks_nonempty += !!stripe_blockcount_get(new_s, i);
+
+ if (!old_s)
+ bch2_stripes_heap_insert(c, m, idx);
+ else
+ bch2_stripes_heap_update(c, m, idx);
+ }
+ } else {
+ 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 (i = 0; i < new_s->nr_blocks; i++)
+ m->ptrs[i] = new_s->ptrs[i];
+
+ bch2_bkey_to_replicas(&m->r.e, new);
- g = bucket(ca, b);
+ /*
+ * gc recalculates this field from stripe ptr
+ * references:
+ */
+ memset(m->block_sectors, 0, sizeof(m->block_sectors));
- *old = bucket_data_cmpxchg(c, ca, g, new, ({
- if (!is_available_bucket(new)) {
- percpu_up_read_preempt_enable(&c->usage_lock);
- return false;
+ for (i = 0; i < new_s->nr_blocks; i++) {
+ ret = mark_stripe_bucket(trans, new, i, flags);
+ if (ret)
+ return ret;
}
- new.owned_by_allocator = 1;
- new.data_type = 0;
- new.cached_sectors = 0;
- new.dirty_sectors = 0;
- new.gen++;
- }));
+ ret = update_replicas(c, new, &m->r.e,
+ ((s64) m->sectors * m->nr_redundant),
+ journal_seq, gc);
+ 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;
+ }
+ }
- if (!old->owned_by_allocator && old->cached_sectors)
- trace_invalidate(ca, bucket_to_sector(ca, b),
- old->cached_sectors);
- return true;
+ return 0;
}
-void bch2_mark_alloc_bucket(struct bch_fs *c, struct bch_dev *ca,
- size_t b, bool owned_by_allocator,
- struct gc_pos pos, unsigned flags)
+static int __mark_reservation(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c k, unsigned flags)
{
- struct bucket *g;
- struct bucket_mark old, new;
+ struct bch_fs *c = trans->c;
+ struct bch_fs_usage *fs_usage;
+ unsigned replicas = bkey_s_c_to_reservation(k).v->nr_replicas;
+ s64 sectors = (s64) k.k->size;
+
+ BUG_ON(!(flags & BTREE_TRIGGER_GC));
+
+ if (flags & BTREE_TRIGGER_OVERWRITE)
+ sectors = -sectors;
+ sectors *= replicas;
- percpu_rwsem_assert_held(&c->usage_lock);
- g = bucket(ca, b);
+ percpu_down_read(&c->mark_lock);
+ preempt_disable();
- if (!(flags & BCH_BUCKET_MARK_GC_LOCK_HELD) &&
- gc_will_visit(c, pos))
- return;
+ fs_usage = fs_usage_ptr(c, trans->journal_res.seq, flags & BTREE_TRIGGER_GC);
+ replicas = clamp_t(unsigned, replicas, 1,
+ ARRAY_SIZE(fs_usage->persistent_reserved));
- old = bucket_data_cmpxchg(c, ca, g, new, ({
- new.owned_by_allocator = owned_by_allocator;
- }));
+ fs_usage->reserved += sectors;
+ fs_usage->persistent_reserved[replicas - 1] += sectors;
- BUG_ON(!owned_by_allocator && !old.owned_by_allocator &&
- c->gc_pos.phase == GC_PHASE_DONE);
+ preempt_enable();
+ percpu_up_read(&c->mark_lock);
+
+ return 0;
}
-#define saturated_add(ca, dst, src, max) \
-do { \
- BUG_ON((int) (dst) + (src) < 0); \
- if ((dst) == (max)) \
- ; \
- else if ((dst) + (src) <= (max)) \
- dst += (src); \
- else { \
- dst = (max); \
- trace_sectors_saturated(ca); \
- } \
-} while (0)
+int bch2_mark_reservation(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c old, struct bkey_s_c new,
+ unsigned flags)
+{
+ return mem_trigger_run_overwrite_then_insert(__mark_reservation, trans, btree_id, level, old, new, flags);
+}
-void bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca,
- size_t b, enum bch_data_type type,
- unsigned sectors, struct gc_pos pos,
- unsigned flags)
+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 bucket *g;
- struct bucket_mark old, new;
+ 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;
- BUG_ON(!type);
+ if (r_idx >= c->reflink_gc_nr)
+ goto not_found;
- if (likely(c)) {
- percpu_rwsem_assert_held(&c->usage_lock);
+ r = genradix_ptr(&c->reflink_gc_table, r_idx);
+ next_idx = min(next_idx, r->offset - r->size);
+ if (*idx < next_idx)
+ goto not_found;
- if (!(flags & BCH_BUCKET_MARK_GC_LOCK_HELD) &&
- gc_will_visit(c, pos))
- return;
+ BUG_ON((s64) r->refcount + add < 0);
+
+ r->refcount += add;
+ *idx = r->offset;
+ return 0;
+not_found:
+ if (fsck_err(c, reflink_p_to_missing_reflink_v,
+ "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;
+
+ new = bch2_trans_kmalloc(trans, sizeof(*new));
+ ret = PTR_ERR_OR_ZERO(new);
+ if (ret)
+ goto err;
+
+ bkey_init(&new->k);
+ new->k.type = KEY_TYPE_error;
+ 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(trans, BTREE_ID_extents, &new->k_i,
+ BTREE_TRIGGER_NORUN);
}
- rcu_read_lock();
+ *idx = next_idx;
+err:
+fsck_err:
+ printbuf_exit(&buf);
+ return ret;
+}
- g = bucket(ca, b);
- old = bucket_data_cmpxchg(c, ca, g, new, ({
- saturated_add(ca, new.dirty_sectors, sectors,
- GC_MAX_SECTORS_USED);
- new.data_type = type;
- }));
+static int __mark_reflink_p(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c k, unsigned flags)
+{
+ struct bch_fs *c = trans->c;
+ 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), start = idx;
+ u64 end = le64_to_cpu(p.v->idx) + p.k->size;
+ int ret = 0;
+
+ BUG_ON(!(flags & BTREE_TRIGGER_GC));
+
+ if (c->sb.version_upgrade_complete >= bcachefs_metadata_version_reflink_p_fix) {
+ idx -= le32_to_cpu(p.v->front_pad);
+ end += le32_to_cpu(p.v->back_pad);
+ }
- rcu_read_unlock();
+ l = 0;
+ r = c->reflink_gc_nr;
+ while (l < r) {
+ m = l + (r - l) / 2;
+
+ ref = genradix_ptr(&c->reflink_gc_table, m);
+ if (ref->offset <= idx)
+ l = m + 1;
+ else
+ r = m;
+ }
- BUG_ON(!(flags & BCH_BUCKET_MARK_MAY_MAKE_UNAVAILABLE) &&
- bucket_became_unavailable(c, old, new));
+ while (idx < end && !ret)
+ ret = __bch2_mark_reflink_p(trans, p, start, end,
+ &idx, flags, l++);
+
+ return ret;
}
-/* Reverting this until the copygc + compression issue is fixed: */
+int bch2_mark_reflink_p(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c old, struct bkey_s_c new,
+ unsigned flags)
+{
+ return mem_trigger_run_overwrite_then_insert(__mark_reflink_p, trans, btree_id, level, old, new, flags);
+}
-static int __disk_sectors(struct bch_extent_crc_unpacked crc, unsigned sectors)
+void bch2_trans_fs_usage_revert(struct btree_trans *trans,
+ struct replicas_delta_list *deltas)
{
- if (!sectors)
- return 0;
+ struct bch_fs *c = trans->c;
+ struct bch_fs_usage *dst;
+ struct replicas_delta *d, *top = (void *) deltas->d + deltas->used;
+ s64 added = 0;
+ unsigned i;
+
+ percpu_down_read(&c->mark_lock);
+ preempt_disable();
+ dst = fs_usage_ptr(c, trans->journal_res.seq, false);
+
+ /* revert changes: */
+ for (d = deltas->d; d != top; d = replicas_delta_next(d)) {
+ switch (d->r.data_type) {
+ case BCH_DATA_btree:
+ case BCH_DATA_user:
+ case BCH_DATA_parity:
+ added += d->delta;
+ }
+ BUG_ON(__update_replicas(c, dst, &d->r, -d->delta));
+ }
- return max(1U, DIV_ROUND_UP(sectors * crc.compressed_size,
- crc.uncompressed_size));
+ dst->nr_inodes -= deltas->nr_inodes;
+
+ for (i = 0; i < BCH_REPLICAS_MAX; i++) {
+ added -= deltas->persistent_reserved[i];
+ dst->reserved -= deltas->persistent_reserved[i];
+ dst->persistent_reserved[i] -= deltas->persistent_reserved[i];
+ }
+
+ if (added > 0) {
+ trans->disk_res->sectors += added;
+ this_cpu_add(*c->online_reserved, added);
+ }
+
+ preempt_enable();
+ percpu_up_read(&c->mark_lock);
}
-/*
- * Checking against gc's position has to be done here, inside the cmpxchg()
- * loop, to avoid racing with the start of gc clearing all the marks - GC does
- * that with the gc pos seqlock held.
- */
-static void bch2_mark_pointer(struct bch_fs *c,
- struct bkey_s_c_extent e,
- const struct bch_extent_ptr *ptr,
- struct bch_extent_crc_unpacked crc,
- s64 sectors, enum s_alloc type,
- struct bch_fs_usage *stats,
- u64 journal_seq, unsigned flags)
-{
- struct bucket_mark old, new;
- unsigned saturated;
- struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
- struct bucket *g = PTR_BUCKET(ca, ptr);
- enum bch_data_type data_type = type == S_META
- ? BCH_DATA_BTREE : BCH_DATA_USER;
- u64 v;
+int bch2_trans_fs_usage_apply(struct btree_trans *trans,
+ struct replicas_delta_list *deltas)
+{
+ struct bch_fs *c = trans->c;
+ static int warned_disk_usage = 0;
+ bool warn = false;
+ u64 disk_res_sectors = trans->disk_res ? trans->disk_res->sectors : 0;
+ struct replicas_delta *d, *d2;
+ struct replicas_delta *top = (void *) deltas->d + deltas->used;
+ struct bch_fs_usage *dst;
+ s64 added = 0, should_not_have_added;
+ unsigned i;
- if (crc.compression_type) {
- unsigned old_sectors, new_sectors;
+ percpu_down_read(&c->mark_lock);
+ preempt_disable();
+ dst = fs_usage_ptr(c, trans->journal_res.seq, false);
- if (sectors > 0) {
- old_sectors = 0;
- new_sectors = sectors;
- } else {
- old_sectors = e.k->size;
- new_sectors = e.k->size + sectors;
+ for (d = deltas->d; d != top; d = replicas_delta_next(d)) {
+ switch (d->r.data_type) {
+ case BCH_DATA_btree:
+ case BCH_DATA_user:
+ case BCH_DATA_parity:
+ added += d->delta;
}
- sectors = -__disk_sectors(crc, old_sectors)
- +__disk_sectors(crc, new_sectors);
+ if (__update_replicas(c, dst, &d->r, d->delta))
+ goto need_mark;
}
- if (flags & BCH_BUCKET_MARK_GC_WILL_VISIT) {
- if (journal_seq)
- bucket_cmpxchg(g, new, ({
- new.journal_seq_valid = 1;
- new.journal_seq = journal_seq;
- }));
+ dst->nr_inodes += deltas->nr_inodes;
- return;
+ for (i = 0; i < BCH_REPLICAS_MAX; i++) {
+ added += deltas->persistent_reserved[i];
+ dst->reserved += deltas->persistent_reserved[i];
+ dst->persistent_reserved[i] += deltas->persistent_reserved[i];
}
- v = atomic64_read(&g->_mark.v);
- do {
- new.v.counter = old.v.counter = v;
- saturated = 0;
+ /*
+ * Not allowed to reduce sectors_available except by getting a
+ * reservation:
+ */
+ should_not_have_added = added - (s64) disk_res_sectors;
+ if (unlikely(should_not_have_added > 0)) {
+ u64 old, new, v = atomic64_read(&c->sectors_available);
+
+ do {
+ old = v;
+ new = max_t(s64, 0, old - should_not_have_added);
+ } while ((v = atomic64_cmpxchg(&c->sectors_available,
+ old, new)) != old);
+
+ added -= should_not_have_added;
+ warn = true;
+ }
- /*
- * Check this after reading bucket mark to guard against
- * the allocator invalidating a bucket after we've already
- * checked the gen
- */
- if (gen_after(new.gen, ptr->gen)) {
- BUG_ON(!test_bit(BCH_FS_ALLOC_READ_DONE, &c->flags));
- EBUG_ON(!ptr->cached &&
- test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags));
- return;
- }
+ if (added > 0) {
+ trans->disk_res->sectors -= added;
+ this_cpu_sub(*c->online_reserved, added);
+ }
- if (!ptr->cached &&
- new.dirty_sectors == GC_MAX_SECTORS_USED &&
- sectors < 0)
- saturated = -sectors;
+ preempt_enable();
+ percpu_up_read(&c->mark_lock);
- if (ptr->cached)
- saturated_add(ca, new.cached_sectors, sectors,
- GC_MAX_SECTORS_USED);
- else
- saturated_add(ca, new.dirty_sectors, sectors,
- GC_MAX_SECTORS_USED);
+ if (unlikely(warn) && !xchg(&warned_disk_usage, 1))
+ bch2_trans_inconsistent(trans,
+ "disk usage increased %lli more than %llu sectors reserved)",
+ should_not_have_added, disk_res_sectors);
+ return 0;
+need_mark:
+ /* revert changes: */
+ for (d2 = deltas->d; d2 != d; d2 = replicas_delta_next(d2))
+ BUG_ON(__update_replicas(c, dst, &d2->r, -d2->delta));
+
+ preempt_enable();
+ percpu_up_read(&c->mark_lock);
+ return -1;
+}
+
+/* trans_mark: */
+
+static inline int bch2_trans_mark_pointer(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c k, struct extent_ptr_decoded p,
+ unsigned flags)
+{
+ bool insert = !(flags & BTREE_TRIGGER_OVERWRITE);
+ struct btree_iter iter;
+ struct bkey_i_alloc_v4 *a;
+ struct bpos bucket;
+ struct bch_backpointer bp;
+ s64 sectors;
+ int ret;
+
+ bch2_extent_ptr_to_bp(trans->c, btree_id, level, k, p, &bucket, &bp);
+ sectors = bp.bucket_len;
+ if (!insert)
+ sectors = -sectors;
+
+ a = bch2_trans_start_alloc_update(trans, &iter, bucket);
+ if (IS_ERR(a))
+ return PTR_ERR(a);
+
+ ret = __mark_pointer(trans, k, &p.ptr, sectors, bp.data_type,
+ 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);
- if (!new.dirty_sectors &&
- !new.cached_sectors) {
- new.data_type = 0;
+ if (ret)
+ return ret;
- if (journal_seq) {
- new.journal_seq_valid = 1;
- new.journal_seq = journal_seq;
+ if (!p.ptr.cached) {
+ ret = bch2_bucket_backpointer_mod(trans, bucket, bp, k, insert);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
+static int bch2_trans_mark_stripe_ptr(struct btree_trans *trans,
+ struct extent_ptr_decoded p,
+ s64 sectors, enum bch_data_type data_type)
+{
+ struct btree_iter iter;
+ struct bkey_i_stripe *s;
+ struct bch_replicas_padded r;
+ int ret = 0;
+
+ s = bch2_bkey_get_mut_typed(trans, &iter,
+ BTREE_ID_stripes, POS(0, p.ec.idx),
+ BTREE_ITER_WITH_UPDATES, stripe);
+ ret = PTR_ERR_OR_ZERO(s);
+ if (unlikely(ret)) {
+ bch2_trans_inconsistent_on(bch2_err_matches(ret, ENOENT), trans,
+ "pointer to nonexistent stripe %llu",
+ (u64) p.ec.idx);
+ goto err;
+ }
+
+ if (!bch2_ptr_matches_stripe(&s->v, p)) {
+ bch2_trans_inconsistent(trans,
+ "stripe pointer doesn't match stripe %llu",
+ (u64) p.ec.idx);
+ ret = -EIO;
+ goto err;
+ }
+
+ stripe_blockcount_set(&s->v, p.ec.block,
+ stripe_blockcount_get(&s->v, p.ec.block) +
+ sectors);
+
+ bch2_bkey_to_replicas(&r.e, bkey_i_to_s_c(&s->k_i));
+ r.e.data_type = data_type;
+ ret = update_replicas_list(trans, &r.e, sectors);
+err:
+ bch2_trans_iter_exit(trans, &iter);
+ return ret;
+}
+
+static int __trans_mark_extent(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c k, unsigned flags)
+{
+ struct bch_fs *c = trans->c;
+ struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
+ const union bch_extent_entry *entry;
+ struct extent_ptr_decoded p;
+ struct bch_replicas_padded r;
+ enum bch_data_type data_type = bkey_is_btree_ptr(k.k)
+ ? BCH_DATA_btree
+ : BCH_DATA_user;
+ s64 sectors = bkey_is_btree_ptr(k.k)
+ ? btree_sectors(c)
+ : k.k->size;
+ s64 dirty_sectors = 0;
+ bool stale;
+ int ret = 0;
+
+ r.e.data_type = data_type;
+ r.e.nr_devs = 0;
+ r.e.nr_required = 1;
+
+ bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
+ s64 disk_sectors = ptr_disk_sectors(sectors, p);
+
+ if (flags & BTREE_TRIGGER_OVERWRITE)
+ disk_sectors = -disk_sectors;
+
+ ret = bch2_trans_mark_pointer(trans, btree_id, level, k, p, flags);
+ if (ret < 0)
+ return ret;
+
+ stale = ret > 0;
+
+ if (p.ptr.cached) {
+ if (!stale) {
+ ret = update_cached_sectors_list(trans, p.ptr.dev,
+ disk_sectors);
+ if (ret)
+ return ret;
}
+ } else if (!p.has_ec) {
+ dirty_sectors += disk_sectors;
+ r.e.devs[r.e.nr_devs++] = p.ptr.dev;
} else {
- new.data_type = data_type;
+ ret = bch2_trans_mark_stripe_ptr(trans, p,
+ disk_sectors, data_type);
+ if (ret)
+ return ret;
+
+ r.e.nr_required = 0;
}
+ }
+
+ if (r.e.nr_devs)
+ ret = update_replicas_list(trans, &r.e, dirty_sectors);
+
+ return ret;
+}
+
+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;
+ int mod = (int) bch2_bkey_needs_rebalance(c, bkey_i_to_s_c(new)) -
+ (int) bch2_bkey_needs_rebalance(c, old);
+
+ if (mod) {
+ int ret = bch2_btree_bit_mod(trans, BTREE_ID_rebalance_work, new->k.p, mod > 0);
+ if (ret)
+ return ret;
+ }
+
+ return trigger_run_overwrite_then_insert(__trans_mark_extent, trans, btree_id, level, old, new, flags);
+}
+
+static int bch2_trans_mark_stripe_bucket(struct btree_trans *trans,
+ struct bkey_s_c_stripe s,
+ unsigned idx, bool deleting)
+{
+ 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;
+ int ret = 0;
+
+ 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 = check_bucket_ref(trans, s.s_c, ptr, sectors, data_type,
+ a->v.gen, a->v.data_type,
+ a->v.dirty_sectors, a->v.cached_sectors);
+ if (ret)
+ goto err;
- if (flags & BCH_BUCKET_MARK_NOATOMIC) {
- g->_mark = new;
- break;
+ 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_types[a->v.data_type],
+ a->v.dirty_sectors,
+ a->v.stripe, s.k->p.offset)) {
+ ret = -EIO;
+ goto err;
}
- } while ((v = atomic64_cmpxchg(&g->_mark.v,
- old.v.counter,
- new.v.counter)) != old.v.counter);
-
- bch2_dev_usage_update(c, ca, old, new);
-
- BUG_ON(!(flags & BCH_BUCKET_MARK_MAY_MAKE_UNAVAILABLE) &&
- bucket_became_unavailable(c, old, new));
-
- if (saturated &&
- atomic_long_add_return(saturated,
- &ca->saturated_count) >=
- bucket_to_sector(ca, ca->free_inc.size)) {
- if (c->gc_thread) {
- trace_gc_sectors_saturated(c);
- wake_up_process(c->gc_thread);
+
+ 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_types[a->v.data_type],
+ a->v.dirty_sectors,
+ s.k->p.offset)) {
+ ret = -EIO;
+ 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;
+ goto err;
}
+
+ a->v.stripe = 0;
+ a->v.stripe_redundancy = 0;
+ a->v.data_type = alloc_data_type(a->v, BCH_DATA_user);
}
+
+ a->v.dirty_sectors += sectors;
+ if (data_type)
+ a->v.data_type = !deleting ? data_type : 0;
+
+ ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
+ if (ret)
+ goto err;
+err:
+ bch2_trans_iter_exit(trans, &iter);
+ return ret;
}
-void bch2_mark_key(struct bch_fs *c, struct bkey_s_c k,
- s64 sectors, bool metadata,
- struct gc_pos pos,
- struct bch_fs_usage *stats,
- u64 journal_seq, 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)
{
+ 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).v;
+ if (new->k.type == KEY_TYPE_stripe)
+ new_s = &bkey_i_to_stripe(new)->v;
+
/*
- * synchronization w.r.t. GC:
- *
- * Normally, bucket sector counts/marks are updated on the fly, as
- * references are added/removed from the btree, the lists of buckets the
- * allocator owns, other metadata buckets, etc.
- *
- * When GC is in progress and going to mark this reference, we do _not_
- * mark this reference here, to avoid double counting - GC will count it
- * when it gets to it.
- *
- * To know whether we should mark a given reference (GC either isn't
- * running, or has already marked references at this position) we
- * construct a total order for everything GC walks. Then, we can simply
- * compare the position of the reference we're marking - @pos - with
- * GC's current position. If GC is going to mark this reference, GC's
- * current position will be less than @pos; if GC's current position is
- * greater than @pos GC has either already walked this position, or
- * isn't running.
- *
- * To avoid racing with GC's position changing, we have to deal with
- * - GC's position being set to GC_POS_MIN when GC starts:
- * usage_lock guards against this
- * - GC's position overtaking @pos: we guard against this with
- * whatever lock protects the data structure the reference lives in
- * (e.g. the btree node lock, or the relevant allocator lock).
+ * If the pointers aren't changing, we don't need to do anything:
*/
+ 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 && old_s &&
+ (new_s->nr_blocks != old_s->nr_blocks ||
+ new_s->nr_redundant != old_s->nr_redundant));
+
+ nr_blocks = new_s ? new_s->nr_blocks : old_s->nr_blocks;
- percpu_down_read_preempt_disable(&c->usage_lock);
- if (!(flags & BCH_BUCKET_MARK_GC_LOCK_HELD) &&
- gc_will_visit(c, pos))
- flags |= BCH_BUCKET_MARK_GC_WILL_VISIT;
-
- if (!stats)
- stats = this_cpu_ptr(c->usage_percpu);
-
- switch (k.k->type) {
- case BCH_EXTENT:
- case BCH_EXTENT_CACHED: {
- struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
- const struct bch_extent_ptr *ptr;
- struct bch_extent_crc_unpacked crc;
- enum s_alloc type = metadata ? S_META : S_DIRTY;
- unsigned replicas = 0;
-
- BUG_ON(metadata && bkey_extent_is_cached(e.k));
- BUG_ON(!sectors);
-
- extent_for_each_ptr_crc(e, ptr, crc) {
- bch2_mark_pointer(c, e, ptr, crc, sectors, type,
- stats, journal_seq, flags);
- replicas += !ptr->cached;
+ if (new_s) {
+ s64 sectors = le16_to_cpu(new_s->sectors);
+
+ bch2_bkey_to_replicas(&r.e, bkey_i_to_s_c(new));
+ ret = update_replicas_list(trans, &r.e, sectors * new_s->nr_redundant);
+ if (ret)
+ return ret;
+ }
+
+ if (old_s) {
+ s64 sectors = -((s64) le16_to_cpu(old_s->sectors));
+
+ bch2_bkey_to_replicas(&r.e, old);
+ ret = update_replicas_list(trans, &r.e, sectors * old_s->nr_redundant);
+ if (ret)
+ return ret;
+ }
+
+ for (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) {
+ ret = bch2_trans_mark_stripe_bucket(trans,
+ bkey_i_to_s_c_stripe(new), i, false);
+ if (ret)
+ break;
}
- if (replicas) {
- BUG_ON(replicas - 1 > ARRAY_SIZE(stats->s));
- stats->s[replicas - 1].data[type] += sectors;
+ if (old_s) {
+ ret = bch2_trans_mark_stripe_bucket(trans,
+ bkey_s_c_to_stripe(old), i, true);
+ if (ret)
+ break;
}
- break;
}
- case BCH_RESERVATION: {
- struct bkey_s_c_reservation r = bkey_s_c_to_reservation(k);
- if (r.v->nr_replicas) {
- BUG_ON(r.v->nr_replicas - 1 > ARRAY_SIZE(stats->s));
- stats->s[r.v->nr_replicas - 1].persistent_reserved += sectors;
- }
- break;
+ return ret;
+}
+
+static int __trans_mark_reservation(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c k, unsigned flags)
+{
+ unsigned replicas = bkey_s_c_to_reservation(k).v->nr_replicas;
+ s64 sectors = (s64) k.k->size;
+ struct replicas_delta_list *d;
+ int ret;
+
+ if (flags & BTREE_TRIGGER_OVERWRITE)
+ sectors = -sectors;
+ sectors *= replicas;
+
+ ret = bch2_replicas_deltas_realloc(trans, 0);
+ if (ret)
+ return ret;
+
+ d = trans->fs_usage_deltas;
+ replicas = clamp_t(unsigned, replicas, 1,
+ ARRAY_SIZE(d->persistent_reserved));
+
+ d->persistent_reserved[replicas - 1] += sectors;
+ return 0;
+}
+
+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)
+{
+ return trigger_run_overwrite_then_insert(__trans_mark_reservation, trans, btree_id, level, old, new, flags);
+}
+
+static int trans_mark_reflink_p_segment(struct btree_trans *trans,
+ struct bkey_s_c_reflink_p p,
+ u64 *idx, unsigned flags)
+{
+ struct bch_fs *c = trans->c;
+ struct btree_iter iter;
+ struct bkey_i *k;
+ __le64 *refcount;
+ int add = !(flags & BTREE_TRIGGER_OVERWRITE) ? 1 : -1;
+ struct printbuf buf = PRINTBUF;
+ int ret;
+
+ k = bch2_bkey_get_mut_noupdate(trans, &iter,
+ BTREE_ID_reflink, POS(0, *idx),
+ BTREE_ITER_WITH_UPDATES);
+ ret = PTR_ERR_OR_ZERO(k);
+ if (ret)
+ goto err;
+
+ refcount = bkey_refcount(k);
+ if (!refcount) {
+ 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.buf);
+ ret = -EIO;
+ goto err;
}
+
+ if (!*refcount && (flags & BTREE_TRIGGER_OVERWRITE)) {
+ 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.buf);
+ ret = -EIO;
+ goto err;
}
- percpu_up_read_preempt_enable(&c->usage_lock);
+
+ if (flags & BTREE_TRIGGER_INSERT) {
+ struct bch_reflink_p *v = (struct bch_reflink_p *) p.v;
+ u64 pad;
+
+ pad = max_t(s64, le32_to_cpu(v->front_pad),
+ le64_to_cpu(v->idx) - bkey_start_offset(&k->k));
+ BUG_ON(pad > U32_MAX);
+ v->front_pad = cpu_to_le32(pad);
+
+ pad = max_t(s64, le32_to_cpu(v->back_pad),
+ k->k.p.offset - p.k->size - le64_to_cpu(v->idx));
+ BUG_ON(pad > U32_MAX);
+ v->back_pad = cpu_to_le32(pad);
+ }
+
+ le64_add_cpu(refcount, add);
+
+ bch2_btree_iter_set_pos_to_extent_start(&iter);
+ ret = bch2_trans_update(trans, &iter, k, 0);
+ if (ret)
+ goto err;
+
+ *idx = k->k.p.offset;
+err:
+ bch2_trans_iter_exit(trans, &iter);
+ printbuf_exit(&buf);
+ return ret;
}
-/* Disk reservations: */
+static int __trans_mark_reflink_p(struct btree_trans *trans,
+ enum btree_id btree_id, unsigned level,
+ struct bkey_s_c k, unsigned flags)
+{
+ struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k);
+ u64 idx, end_idx;
+ int ret = 0;
+
+ idx = le64_to_cpu(p.v->idx) - le32_to_cpu(p.v->front_pad);
+ end_idx = le64_to_cpu(p.v->idx) + p.k->size +
+ le32_to_cpu(p.v->back_pad);
+
+ while (idx < end_idx && !ret)
+ ret = trans_mark_reflink_p_segment(trans, p, &idx, flags);
+ return ret;
+}
+
+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)
+{
+ if (flags & BTREE_TRIGGER_INSERT) {
+ struct bch_reflink_p *v = &bkey_i_to_reflink_p(new)->v;
+
+ v->front_pad = v->back_pad = 0;
+ }
+
+ return trigger_run_overwrite_then_insert(__trans_mark_reflink_p, trans, btree_id, level, old, new, flags);
+}
+
+static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
+ struct bch_dev *ca, size_t b,
+ enum bch_data_type type,
+ unsigned sectors)
+{
+ struct bch_fs *c = trans->c;
+ struct btree_iter iter;
+ struct bkey_i_alloc_v4 *a;
+ int ret = 0;
+
+ /*
+ * Backup superblock might be past the end of our normal usable space:
+ */
+ if (b >= ca->mi.nbuckets)
+ return 0;
+
+ a = bch2_trans_start_alloc_update(trans, &iter, POS(ca->dev_idx, b));
+ if (IS_ERR(a))
+ return PTR_ERR(a);
+
+ if (a->v.data_type && type && a->v.data_type != type) {
+ bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
+ BCH_FSCK_ERR_bucket_metadata_type_mismatch,
+ "bucket %llu:%llu gen %u different types of data in same bucket: %s, %s\n"
+ "while marking %s",
+ 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 err;
+ }
-static u64 __recalc_sectors_available(struct bch_fs *c)
+ if (a->v.data_type != type ||
+ a->v.dirty_sectors != sectors) {
+ a->v.data_type = type;
+ a->v.dirty_sectors = sectors;
+ ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
+ }
+err:
+ bch2_trans_iter_exit(trans, &iter);
+ return ret;
+}
+
+int bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
+ struct bch_dev *ca, size_t b,
+ enum bch_data_type type,
+ unsigned sectors)
+{
+ return commit_do(trans, NULL, NULL, 0,
+ __bch2_trans_mark_metadata_bucket(trans, ca, b, type, sectors));
+}
+
+static int bch2_trans_mark_metadata_sectors(struct btree_trans *trans,
+ struct bch_dev *ca,
+ u64 start, u64 end,
+ enum bch_data_type type,
+ u64 *bucket, unsigned *bucket_sectors)
{
- int cpu;
+ do {
+ u64 b = sector_to_bucket(ca, start);
+ unsigned sectors =
+ min_t(u64, bucket_to_sector(ca, b + 1), end) - start;
+
+ if (b != *bucket && *bucket_sectors) {
+ int ret = bch2_trans_mark_metadata_bucket(trans, ca, *bucket,
+ type, *bucket_sectors);
+ if (ret)
+ return ret;
+
+ *bucket_sectors = 0;
+ }
- for_each_possible_cpu(cpu)
- per_cpu_ptr(c->usage_percpu, cpu)->available_cache = 0;
+ *bucket = b;
+ *bucket_sectors += sectors;
+ start += sectors;
+ } while (start < end);
- return bch2_fs_sectors_free(c, bch2_fs_usage_read(c));
+ return 0;
}
-/* Used by gc when it's starting: */
-void bch2_recalc_sectors_available(struct bch_fs *c)
+static int __bch2_trans_mark_dev_sb(struct btree_trans *trans,
+ struct bch_dev *ca)
{
- percpu_down_write(&c->usage_lock);
- atomic64_set(&c->sectors_available, __recalc_sectors_available(c));
- percpu_up_write(&c->usage_lock);
+ struct bch_sb_layout *layout = &ca->disk_sb.sb->layout;
+ u64 bucket = 0;
+ unsigned i, bucket_sectors = 0;
+ int ret;
+
+ for (i = 0; i < layout->nr_superblocks; i++) {
+ u64 offset = le64_to_cpu(layout->sb_offset[i]);
+
+ if (offset == BCH_SB_SECTOR) {
+ ret = bch2_trans_mark_metadata_sectors(trans, ca,
+ 0, BCH_SB_SECTOR,
+ BCH_DATA_sb, &bucket, &bucket_sectors);
+ if (ret)
+ return ret;
+ }
+
+ ret = bch2_trans_mark_metadata_sectors(trans, ca, offset,
+ offset + (1 << layout->sb_max_size_bits),
+ BCH_DATA_sb, &bucket, &bucket_sectors);
+ if (ret)
+ return ret;
+ }
+
+ if (bucket_sectors) {
+ ret = bch2_trans_mark_metadata_bucket(trans, ca,
+ bucket, BCH_DATA_sb, bucket_sectors);
+ if (ret)
+ return ret;
+ }
+
+ for (i = 0; i < ca->journal.nr; i++) {
+ ret = bch2_trans_mark_metadata_bucket(trans, ca,
+ ca->journal.buckets[i],
+ BCH_DATA_journal, ca->mi.bucket_size);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
}
-void __bch2_disk_reservation_put(struct bch_fs *c, struct disk_reservation *res)
+int bch2_trans_mark_dev_sb(struct bch_fs *c, struct bch_dev *ca)
{
- percpu_down_read_preempt_disable(&c->usage_lock);
- this_cpu_sub(c->usage_percpu->online_reserved,
- res->sectors);
+ int ret = bch2_trans_run(c, __bch2_trans_mark_dev_sb(trans, ca));
+
+ if (ret)
+ bch_err_fn(c, ret);
+ return ret;
+}
- bch2_fs_stats_verify(c);
- percpu_up_read_preempt_enable(&c->usage_lock);
+int bch2_trans_mark_dev_sbs(struct bch_fs *c)
+{
+ struct bch_dev *ca;
+ unsigned i;
- res->sectors = 0;
+ for_each_online_member(ca, c, i) {
+ int ret = bch2_trans_mark_dev_sb(c, ca);
+ if (ret) {
+ percpu_ref_put(&ca->ref);
+ return ret;
+ }
+ }
+
+ return 0;
}
+/* Disk reservations: */
+
#define SECTORS_CACHE 1024
-int bch2_disk_reservation_add(struct bch_fs *c, struct disk_reservation *res,
- unsigned sectors, int flags)
+int __bch2_disk_reservation_add(struct bch_fs *c, struct disk_reservation *res,
+ u64 sectors, int flags)
{
- struct bch_fs_usage *stats;
+ struct bch_fs_pcpu *pcpu;
u64 old, v, get;
s64 sectors_available;
int ret;
- percpu_down_read_preempt_disable(&c->usage_lock);
- stats = this_cpu_ptr(c->usage_percpu);
+ percpu_down_read(&c->mark_lock);
+ preempt_disable();
+ pcpu = this_cpu_ptr(c->pcpu);
- if (sectors <= stats->available_cache)
+ if (sectors <= pcpu->sectors_available)
goto out;
v = atomic64_read(&c->sectors_available);
get = min((u64) sectors + SECTORS_CACHE, old);
if (get < sectors) {
- percpu_up_read_preempt_enable(&c->usage_lock);
+ preempt_enable();
goto recalculate;
}
} while ((v = atomic64_cmpxchg(&c->sectors_available,
old, old - get)) != old);
- stats->available_cache += get;
+ pcpu->sectors_available += get;
out:
- stats->available_cache -= sectors;
- stats->online_reserved += sectors;
- res->sectors += sectors;
+ pcpu->sectors_available -= sectors;
+ this_cpu_add(*c->online_reserved, sectors);
+ res->sectors += sectors;
- bch2_disk_reservations_verify(c, flags);
- bch2_fs_stats_verify(c);
- percpu_up_read_preempt_enable(&c->usage_lock);
+ preempt_enable();
+ percpu_up_read(&c->mark_lock);
return 0;
recalculate:
- /*
- * GC recalculates sectors_available when it starts, so that hopefully
- * we don't normally end up blocking here:
- */
-
- /*
- * Piss fuck, we can be called from extent_insert_fixup() with btree
- * locks held:
- */
-
- if (!(flags & BCH_DISK_RESERVATION_GC_LOCK_HELD)) {
- if (!(flags & BCH_DISK_RESERVATION_BTREE_LOCKS_HELD))
- down_read(&c->gc_lock);
- else if (!down_read_trylock(&c->gc_lock))
- return -EINTR;
- }
+ mutex_lock(&c->sectors_available_lock);
- percpu_down_write(&c->usage_lock);
- sectors_available = __recalc_sectors_available(c);
+ percpu_u64_set(&c->pcpu->sectors_available, 0);
+ sectors_available = avail_factor(__bch2_fs_usage_read_short(c).free);
if (sectors <= sectors_available ||
(flags & BCH_DISK_RESERVATION_NOFAIL)) {
atomic64_set(&c->sectors_available,
max_t(s64, 0, sectors_available - sectors));
- stats->online_reserved += sectors;
- res->sectors += sectors;
+ this_cpu_add(*c->online_reserved, sectors);
+ res->sectors += sectors;
ret = 0;
-
- bch2_disk_reservations_verify(c, flags);
} else {
atomic64_set(&c->sectors_available, sectors_available);
- ret = -ENOSPC;
+ ret = -BCH_ERR_ENOSPC_disk_reservation;
}
- bch2_fs_stats_verify(c);
- percpu_up_write(&c->usage_lock);
-
- if (!(flags & BCH_DISK_RESERVATION_GC_LOCK_HELD))
- up_read(&c->gc_lock);
+ mutex_unlock(&c->sectors_available_lock);
+ percpu_up_read(&c->mark_lock);
return ret;
}
/* Startup/shutdown: */
-static void buckets_free_rcu(struct rcu_head *rcu)
+static void bucket_gens_free_rcu(struct rcu_head *rcu)
{
- struct bucket_array *buckets =
- container_of(rcu, struct bucket_array, rcu);
+ struct bucket_gens *buckets =
+ container_of(rcu, struct bucket_gens, rcu);
- kvpfree(buckets,
- sizeof(struct bucket_array) +
- buckets->nbuckets * sizeof(struct bucket));
+ 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;
- unsigned long *buckets_dirty = NULL;
- u8 *oldest_gens = NULL;
- alloc_fifo free[RESERVE_NR];
- alloc_fifo free_inc;
- alloc_heap alloc_heap;
- copygc_heap copygc_heap;
-
- size_t btree_reserve = DIV_ROUND_UP(BTREE_NODE_RESERVE,
- ca->mi.bucket_size / c->opts.btree_node_size);
- /* XXX: these should be tunable */
- size_t reserve_none = max_t(size_t, 4, ca->mi.nbuckets >> 9);
- size_t copygc_reserve = max_t(size_t, 16, ca->mi.nbuckets >> 7);
- size_t free_inc_reserve = copygc_reserve / 2;
- bool resize = ca->buckets != NULL,
- start_copygc = ca->copygc_thread != NULL;
- int ret = -ENOMEM;
- unsigned i;
+ struct bucket_gens *bucket_gens = NULL, *old_bucket_gens = NULL;
+ unsigned long *buckets_nouse = NULL;
+ bool resize = ca->bucket_gens != NULL;
+ int ret;
- memset(&free, 0, sizeof(free));
- memset(&free_inc, 0, sizeof(free_inc));
- memset(&alloc_heap, 0, sizeof(alloc_heap));
- memset(©gc_heap, 0, sizeof(copygc_heap));
-
- if (!(buckets = kvpmalloc(sizeof(struct bucket_array) +
- nbuckets * sizeof(struct bucket),
- GFP_KERNEL|__GFP_ZERO)) ||
- !(oldest_gens = kvpmalloc(nbuckets * sizeof(u8),
- GFP_KERNEL|__GFP_ZERO)) ||
- !(buckets_dirty = kvpmalloc(BITS_TO_LONGS(nbuckets) *
+ if (!(bucket_gens = kvpmalloc(sizeof(struct bucket_gens) + nbuckets,
+ GFP_KERNEL|__GFP_ZERO))) {
+ ret = -BCH_ERR_ENOMEM_bucket_gens;
+ goto err;
+ }
+
+ if ((c->opts.buckets_nouse &&
+ !(buckets_nouse = kvpmalloc(BITS_TO_LONGS(nbuckets) *
sizeof(unsigned long),
- GFP_KERNEL|__GFP_ZERO)) ||
- !init_fifo(&free[RESERVE_BTREE], btree_reserve, GFP_KERNEL) ||
- !init_fifo(&free[RESERVE_MOVINGGC],
- copygc_reserve, GFP_KERNEL) ||
- !init_fifo(&free[RESERVE_NONE], reserve_none, GFP_KERNEL) ||
- !init_fifo(&free_inc, free_inc_reserve, GFP_KERNEL) ||
- !init_heap(&alloc_heap, free_inc_reserve, GFP_KERNEL) ||
- !init_heap(©gc_heap, copygc_reserve, GFP_KERNEL))
+ GFP_KERNEL|__GFP_ZERO)))) {
+ ret = -BCH_ERR_ENOMEM_buckets_nouse;
goto err;
+ }
- buckets->first_bucket = ca->mi.first_bucket;
- buckets->nbuckets = nbuckets;
+ bucket_gens->first_bucket = ca->mi.first_bucket;
+ bucket_gens->nbuckets = nbuckets;
- bch2_copygc_stop(ca);
+ bch2_copygc_stop(c);
if (resize) {
down_write(&c->gc_lock);
down_write(&ca->bucket_lock);
- percpu_down_write(&c->usage_lock);
+ 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);
-
- memcpy(buckets->b,
- old_buckets->b,
- n * sizeof(struct bucket));
- memcpy(oldest_gens,
- ca->oldest_gens,
- n * sizeof(u8));
- memcpy(buckets_dirty,
- ca->buckets_dirty,
- BITS_TO_LONGS(n) * sizeof(unsigned long));
+ size_t n = min(bucket_gens->nbuckets, old_bucket_gens->nbuckets);
+
+ memcpy(bucket_gens->b,
+ old_bucket_gens->b,
+ n);
+ if (buckets_nouse)
+ memcpy(buckets_nouse,
+ ca->buckets_nouse,
+ BITS_TO_LONGS(n) * sizeof(unsigned long));
}
- rcu_assign_pointer(ca->buckets, buckets);
- buckets = old_buckets;
-
- swap(ca->oldest_gens, oldest_gens);
- swap(ca->buckets_dirty, buckets_dirty);
+ rcu_assign_pointer(ca->bucket_gens, bucket_gens);
+ bucket_gens = old_bucket_gens;
- if (resize)
- percpu_up_write(&c->usage_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);
-
- /* and we shut down copygc: */
- swap(ca->copygc_heap, copygc_heap);
+ 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);
}
- if (start_copygc &&
- bch2_copygc_start(c, ca))
- bch_err(ca, "error restarting copygc thread");
-
ret = 0;
err:
- free_heap(©gc_heap);
- free_heap(&alloc_heap);
- free_fifo(&free_inc);
- for (i = 0; i < RESERVE_NR; i++)
- free_fifo(&free[i]);
- kvpfree(buckets_dirty,
+ kvpfree(buckets_nouse,
BITS_TO_LONGS(nbuckets) * sizeof(unsigned long));
- kvpfree(oldest_gens,
- nbuckets * sizeof(u8));
- if (buckets)
- call_rcu(&old_buckets->rcu, buckets_free_rcu);
+ if (bucket_gens)
+ call_rcu(&bucket_gens->rcu, bucket_gens_free_rcu);
return ret;
}
{
unsigned i;
- free_heap(&ca->copygc_heap);
- 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_dirty,
+ kvpfree(ca->buckets_nouse,
BITS_TO_LONGS(ca->mi.nbuckets) * sizeof(unsigned long));
- kvpfree(ca->oldest_gens, ca->mi.nbuckets * sizeof(u8));
- kvpfree(rcu_dereference_protected(ca->buckets, 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);
- free_percpu(ca->usage_percpu);
+ for (i = 0; i < ARRAY_SIZE(ca->usage); i++)
+ free_percpu(ca->usage[i]);
+ kfree(ca->usage_base);
}
int bch2_dev_buckets_alloc(struct bch_fs *c, struct bch_dev *ca)
{
- if (!(ca->usage_percpu = alloc_percpu(struct bch_dev_usage)))
- return -ENOMEM;
+ unsigned i;
+
+ ca->usage_base = kzalloc(sizeof(struct bch_dev_usage), GFP_KERNEL);
+ if (!ca->usage_base)
+ return -BCH_ERR_ENOMEM_usage_init;
+
+ for (i = 0; i < ARRAY_SIZE(ca->usage); i++) {
+ ca->usage[i] = alloc_percpu(struct bch_dev_usage);
+ if (!ca->usage[i])
+ return -BCH_ERR_ENOMEM_usage_init;
+ }
- return bch2_dev_buckets_resize(c, ca, ca->mi.nbuckets);;
+ return bch2_dev_buckets_resize(c, ca, ca->mi.nbuckets);
}