summaryrefslogtreecommitdiff
path: root/fs/bcachefs/journal_io.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/bcachefs/journal_io.c')
-rw-r--r--fs/bcachefs/journal_io.c223
1 files changed, 129 insertions, 94 deletions
diff --git a/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c
index fb35dd336331..11c39e0c34f4 100644
--- a/fs/bcachefs/journal_io.c
+++ b/fs/bcachefs/journal_io.c
@@ -17,6 +17,9 @@
#include "sb-clean.h"
#include "trace.h"
+#include <linux/ioprio.h>
+#include <linux/string_choices.h>
+
void bch2_journal_pos_from_member_info_set(struct bch_fs *c)
{
lockdep_assert_held(&c->sb_lock);
@@ -299,7 +302,7 @@ static void journal_entry_err_msg(struct printbuf *out,
journal_entry_err_msg(&_buf, version, jset, entry); \
prt_printf(&_buf, msg, ##__VA_ARGS__); \
\
- switch (flags & BCH_VALIDATE_write) { \
+ switch (from.flags & BCH_VALIDATE_write) { \
case READ: \
mustfix_fsck_err(c, _err, "%s", _buf.buf); \
break; \
@@ -325,11 +328,11 @@ static void journal_entry_err_msg(struct printbuf *out,
static int journal_validate_key(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
- unsigned level, enum btree_id btree_id,
struct bkey_i *k,
- unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from,
+ unsigned version, int big_endian)
{
+ enum bch_validate_flags flags = from.flags;
int write = flags & BCH_VALIDATE_write;
void *next = vstruct_next(entry);
int ret = 0;
@@ -364,11 +367,10 @@ static int journal_validate_key(struct bch_fs *c,
}
if (!write)
- bch2_bkey_compat(level, btree_id, version, big_endian,
+ bch2_bkey_compat(from.level, from.btree, version, big_endian,
write, NULL, bkey_to_packed(k));
- ret = bch2_bkey_validate(c, bkey_i_to_s_c(k),
- __btree_node_type(level, btree_id), write);
+ ret = bch2_bkey_validate(c, bkey_i_to_s_c(k), from);
if (ret == -BCH_ERR_fsck_delete_bkey) {
le16_add_cpu(&entry->u64s, -((u16) k->k.u64s));
memmove(k, bkey_next(k), next - (void *) bkey_next(k));
@@ -379,7 +381,7 @@ static int journal_validate_key(struct bch_fs *c,
goto fsck_err;
if (write)
- bch2_bkey_compat(level, btree_id, version, big_endian,
+ bch2_bkey_compat(from.level, from.btree, version, big_endian,
write, NULL, bkey_to_packed(k));
fsck_err:
return ret;
@@ -389,16 +391,15 @@ static int journal_entry_btree_keys_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
struct bkey_i *k = entry->start;
+ from.level = entry->level;
+ from.btree = entry->btree_id;
+
while (k != vstruct_last(entry)) {
- int ret = journal_validate_key(c, jset, entry,
- entry->level,
- entry->btree_id,
- k, version, big_endian,
- flags|BCH_VALIDATE_journal);
+ int ret = journal_validate_key(c, jset, entry, k, from, version, big_endian);
if (ret == FSCK_DELETED_KEY)
continue;
else if (ret)
@@ -421,7 +422,8 @@ static void journal_entry_btree_keys_to_text(struct printbuf *out, struct bch_fs
bch2_prt_jset_entry_type(out, entry->type);
prt_str(out, ": ");
}
- prt_printf(out, "btree=%s l=%u ", bch2_btree_id_str(entry->btree_id), entry->level);
+ bch2_btree_id_level_to_text(out, entry->btree_id, entry->level);
+ prt_char(out, ' ');
bch2_bkey_val_to_text(out, c, bkey_i_to_s_c(k));
first = false;
}
@@ -431,11 +433,15 @@ static int journal_entry_btree_root_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
struct bkey_i *k = entry->start;
int ret = 0;
+ from.root = true;
+ from.level = entry->level + 1;
+ from.btree = entry->btree_id;
+
if (journal_entry_err_on(!entry->u64s ||
le16_to_cpu(entry->u64s) != k->k.u64s,
c, version, jset, entry,
@@ -452,8 +458,7 @@ static int journal_entry_btree_root_validate(struct bch_fs *c,
return 0;
}
- ret = journal_validate_key(c, jset, entry, 1, entry->btree_id, k,
- version, big_endian, flags);
+ ret = journal_validate_key(c, jset, entry, k, from, version, big_endian);
if (ret == FSCK_DELETED_KEY)
ret = 0;
fsck_err:
@@ -470,7 +475,7 @@ static int journal_entry_prio_ptrs_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
/* obsolete, don't care: */
return 0;
@@ -485,7 +490,7 @@ static int journal_entry_blacklist_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
int ret = 0;
@@ -512,7 +517,7 @@ static int journal_entry_blacklist_v2_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
struct jset_entry_blacklist_v2 *bl_entry;
int ret = 0;
@@ -554,7 +559,7 @@ static int journal_entry_usage_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
struct jset_entry_usage *u =
container_of(entry, struct jset_entry_usage, entry);
@@ -588,7 +593,7 @@ static int journal_entry_data_usage_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
struct jset_entry_data_usage *u =
container_of(entry, struct jset_entry_data_usage, entry);
@@ -632,7 +637,7 @@ static int journal_entry_clock_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
struct jset_entry_clock *clock =
container_of(entry, struct jset_entry_clock, entry);
@@ -665,14 +670,14 @@ static void journal_entry_clock_to_text(struct printbuf *out, struct bch_fs *c,
struct jset_entry_clock *clock =
container_of(entry, struct jset_entry_clock, entry);
- prt_printf(out, "%s=%llu", clock->rw ? "write" : "read", le64_to_cpu(clock->time));
+ prt_printf(out, "%s=%llu", str_write_read(clock->rw), le64_to_cpu(clock->time));
}
static int journal_entry_dev_usage_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
struct jset_entry_dev_usage *u =
container_of(entry, struct jset_entry_dev_usage, entry);
@@ -729,7 +734,7 @@ static int journal_entry_log_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
return 0;
}
@@ -738,19 +743,19 @@ static void journal_entry_log_to_text(struct printbuf *out, struct bch_fs *c,
struct jset_entry *entry)
{
struct jset_entry_log *l = container_of(entry, struct jset_entry_log, entry);
- unsigned bytes = vstruct_bytes(entry) - offsetof(struct jset_entry_log, d);
- prt_printf(out, "%.*s", bytes, l->d);
+ prt_printf(out, "%.*s", jset_entry_log_msg_bytes(l), l->d);
}
static int journal_entry_overwrite_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
+ from.flags = 0;
return journal_entry_btree_keys_validate(c, jset, entry,
- version, big_endian, READ);
+ version, big_endian, from);
}
static void journal_entry_overwrite_to_text(struct printbuf *out, struct bch_fs *c,
@@ -763,10 +768,10 @@ static int journal_entry_write_buffer_keys_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
return journal_entry_btree_keys_validate(c, jset, entry,
- version, big_endian, READ);
+ version, big_endian, from);
}
static void journal_entry_write_buffer_keys_to_text(struct printbuf *out, struct bch_fs *c,
@@ -779,7 +784,7 @@ static int journal_entry_datetime_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
unsigned bytes = vstruct_bytes(entry);
unsigned expected = 16;
@@ -809,7 +814,7 @@ static void journal_entry_datetime_to_text(struct printbuf *out, struct bch_fs *
struct jset_entry_ops {
int (*validate)(struct bch_fs *, struct jset *,
struct jset_entry *, unsigned, int,
- enum bch_validate_flags);
+ struct bkey_validate_context);
void (*to_text)(struct printbuf *, struct bch_fs *, struct jset_entry *);
};
@@ -827,11 +832,11 @@ int bch2_journal_entry_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
- enum bch_validate_flags flags)
+ struct bkey_validate_context from)
{
return entry->type < BCH_JSET_ENTRY_NR
? bch2_jset_entry_ops[entry->type].validate(c, jset, entry,
- version, big_endian, flags)
+ version, big_endian, from)
: 0;
}
@@ -849,10 +854,18 @@ void bch2_journal_entry_to_text(struct printbuf *out, struct bch_fs *c,
static int jset_validate_entries(struct bch_fs *c, struct jset *jset,
enum bch_validate_flags flags)
{
+ struct bkey_validate_context from = {
+ .flags = flags,
+ .from = BKEY_VALIDATE_journal,
+ .journal_seq = le64_to_cpu(jset->seq),
+ };
+
unsigned version = le32_to_cpu(jset->version);
int ret = 0;
vstruct_for_each(jset, entry) {
+ from.journal_offset = (u64 *) entry - jset->_data;
+
if (journal_entry_err_on(vstruct_next(entry) > vstruct_last(jset),
c, version, jset, entry,
journal_entry_past_jset_end,
@@ -861,8 +874,8 @@ static int jset_validate_entries(struct bch_fs *c, struct jset *jset,
break;
}
- ret = bch2_journal_entry_validate(c, jset, entry,
- version, JSET_BIG_ENDIAN(jset), flags);
+ ret = bch2_journal_entry_validate(c, jset, entry, version,
+ JSET_BIG_ENDIAN(jset), from);
if (ret)
break;
}
@@ -875,13 +888,17 @@ static int jset_validate(struct bch_fs *c,
struct jset *jset, u64 sector,
enum bch_validate_flags flags)
{
- unsigned version;
+ struct bkey_validate_context from = {
+ .flags = flags,
+ .from = BKEY_VALIDATE_journal,
+ .journal_seq = le64_to_cpu(jset->seq),
+ };
int ret = 0;
if (le64_to_cpu(jset->magic) != jset_magic(c))
return JOURNAL_ENTRY_NONE;
- version = le32_to_cpu(jset->version);
+ unsigned version = le32_to_cpu(jset->version);
if (journal_entry_err_on(!bch2_version_compatible(version),
c, version, jset, NULL,
jset_unsupported_version,
@@ -926,15 +943,16 @@ static int jset_validate_early(struct bch_fs *c,
unsigned bucket_sectors_left,
unsigned sectors_read)
{
- size_t bytes = vstruct_bytes(jset);
- unsigned version;
- enum bch_validate_flags flags = BCH_VALIDATE_journal;
+ struct bkey_validate_context from = {
+ .from = BKEY_VALIDATE_journal,
+ .journal_seq = le64_to_cpu(jset->seq),
+ };
int ret = 0;
if (le64_to_cpu(jset->magic) != jset_magic(c))
return JOURNAL_ENTRY_NONE;
- version = le32_to_cpu(jset->version);
+ unsigned version = le32_to_cpu(jset->version);
if (journal_entry_err_on(!bch2_version_compatible(version),
c, version, jset, NULL,
jset_unsupported_version,
@@ -947,6 +965,7 @@ static int jset_validate_early(struct bch_fs *c,
return -EINVAL;
}
+ size_t bytes = vstruct_bytes(jset);
if (bytes > (sectors_read << 9) &&
sectors_read < bucket_sectors_left)
return JOURNAL_ENTRY_REREAD;
@@ -1231,8 +1250,6 @@ int bch2_journal_read(struct bch_fs *c,
* those entries will be blacklisted:
*/
genradix_for_each_reverse(&c->journal_entries, radix_iter, _i) {
- enum bch_validate_flags flags = BCH_VALIDATE_journal;
-
i = *_i;
if (journal_replay_ignore(i))
@@ -1252,6 +1269,10 @@ int bch2_journal_read(struct bch_fs *c,
continue;
}
+ struct bkey_validate_context from = {
+ .from = BKEY_VALIDATE_journal,
+ .journal_seq = le64_to_cpu(i->j.seq),
+ };
if (journal_entry_err_on(le64_to_cpu(i->j.last_seq) > le64_to_cpu(i->j.seq),
c, le32_to_cpu(i->j.version), &i->j, NULL,
jset_last_seq_newer_than_seq,
@@ -1411,27 +1432,50 @@ fsck_err:
/* journal write: */
+static void journal_advance_devs_to_next_bucket(struct journal *j,
+ struct dev_alloc_list *devs,
+ unsigned sectors, u64 seq)
+{
+ struct bch_fs *c = container_of(j, struct bch_fs, journal);
+
+ darray_for_each(*devs, i) {
+ struct bch_dev *ca = rcu_dereference(c->devs[*i]);
+ if (!ca)
+ continue;
+
+ struct journal_device *ja = &ca->journal;
+
+ if (sectors > ja->sectors_free &&
+ sectors <= ca->mi.bucket_size &&
+ bch2_journal_dev_buckets_available(j, ja,
+ journal_space_discarded)) {
+ ja->cur_idx = (ja->cur_idx + 1) % ja->nr;
+ ja->sectors_free = ca->mi.bucket_size;
+
+ /*
+ * ja->bucket_seq[ja->cur_idx] must always have
+ * something sensible:
+ */
+ ja->bucket_seq[ja->cur_idx] = le64_to_cpu(seq);
+ }
+ }
+}
+
static void __journal_write_alloc(struct journal *j,
struct journal_buf *w,
- struct dev_alloc_list *devs_sorted,
+ struct dev_alloc_list *devs,
unsigned sectors,
unsigned *replicas,
unsigned replicas_want)
{
struct bch_fs *c = container_of(j, struct bch_fs, journal);
- struct journal_device *ja;
- struct bch_dev *ca;
- unsigned i;
- if (*replicas >= replicas_want)
- return;
-
- for (i = 0; i < devs_sorted->nr; i++) {
- ca = rcu_dereference(c->devs[devs_sorted->devs[i]]);
+ darray_for_each(*devs, i) {
+ struct bch_dev *ca = rcu_dereference(c->devs[*i]);
if (!ca)
continue;
- ja = &ca->journal;
+ struct journal_device *ja = &ca->journal;
/*
* Check that we can use this device, and aren't already using
@@ -1477,65 +1521,53 @@ static int journal_write_alloc(struct journal *j, struct journal_buf *w)
{
struct bch_fs *c = container_of(j, struct bch_fs, journal);
struct bch_devs_mask devs;
- struct journal_device *ja;
- struct bch_dev *ca;
struct dev_alloc_list devs_sorted;
unsigned sectors = vstruct_sectors(w->data, c->block_bits);
unsigned target = c->opts.metadata_target ?:
c->opts.foreground_target;
- unsigned i, replicas = 0, replicas_want =
+ unsigned replicas = 0, replicas_want =
READ_ONCE(c->opts.metadata_replicas);
unsigned replicas_need = min_t(unsigned, replicas_want,
READ_ONCE(c->opts.metadata_replicas_required));
+ bool advance_done = false;
rcu_read_lock();
-retry:
- devs = target_rw_devs(c, BCH_DATA_journal, target);
- devs_sorted = bch2_dev_alloc_list(c, &j->wp.stripe, &devs);
+ /* We might run more than once if we have to stop and do discards: */
+ struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(bkey_i_to_s_c(&w->key));
+ bkey_for_each_ptr(ptrs, p) {
+ struct bch_dev *ca = bch2_dev_rcu_noerror(c, p->dev);
+ if (ca)
+ replicas += ca->mi.durability;
+ }
- __journal_write_alloc(j, w, &devs_sorted,
- sectors, &replicas, replicas_want);
+retry_target:
+ devs = target_rw_devs(c, BCH_DATA_journal, target);
+ devs_sorted = bch2_dev_alloc_list(c, &j->wp.stripe, &devs);
+retry_alloc:
+ __journal_write_alloc(j, w, &devs_sorted, sectors, &replicas, replicas_want);
- if (replicas >= replicas_want)
+ if (likely(replicas >= replicas_want))
goto done;
- for (i = 0; i < devs_sorted.nr; i++) {
- ca = rcu_dereference(c->devs[devs_sorted.devs[i]]);
- if (!ca)
- continue;
-
- ja = &ca->journal;
-
- if (sectors > ja->sectors_free &&
- sectors <= ca->mi.bucket_size &&
- bch2_journal_dev_buckets_available(j, ja,
- journal_space_discarded)) {
- ja->cur_idx = (ja->cur_idx + 1) % ja->nr;
- ja->sectors_free = ca->mi.bucket_size;
-
- /*
- * ja->bucket_seq[ja->cur_idx] must always have
- * something sensible:
- */
- ja->bucket_seq[ja->cur_idx] = le64_to_cpu(w->data->seq);
- }
+ if (!advance_done) {
+ journal_advance_devs_to_next_bucket(j, &devs_sorted, sectors, w->data->seq);
+ advance_done = true;
+ goto retry_alloc;
}
- __journal_write_alloc(j, w, &devs_sorted,
- sectors, &replicas, replicas_want);
-
if (replicas < replicas_want && target) {
/* Retry from all devices: */
target = 0;
- goto retry;
+ advance_done = false;
+ goto retry_target;
}
done:
rcu_read_unlock();
BUG_ON(bkey_val_u64s(&w->key.k) > BCH_REPLICAS_MAX);
- return replicas >= replicas_need ? 0 : -EROFS;
+ return replicas >= replicas_need ? 0 : -BCH_ERR_insufficient_journal_devices;
}
static void journal_buf_realloc(struct journal *j, struct journal_buf *buf)
@@ -1732,6 +1764,7 @@ static CLOSURE_CALLBACK(journal_write_submit)
bio->bi_iter.bi_sector = ptr->offset;
bio->bi_end_io = journal_write_endio;
bio->bi_private = ca;
+ bio->bi_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_RT, 0);
BUG_ON(bio->bi_iter.bi_sector == ca->prev_journal_sector);
ca->prev_journal_sector = bio->bi_iter.bi_sector;
@@ -2023,19 +2056,21 @@ CLOSURE_CALLBACK(bch2_journal_write)
bch2_journal_do_discards(j);
}
- if (ret) {
+ if (ret && !bch2_journal_error(j)) {
struct printbuf buf = PRINTBUF;
buf.atomic++;
- prt_printf(&buf, bch2_fmt(c, "Unable to allocate journal write at seq %llu: %s"),
+ prt_printf(&buf, bch2_fmt(c, "Unable to allocate journal write at seq %llu for %zu sectors: %s"),
le64_to_cpu(w->data->seq),
+ vstruct_sectors(w->data, c->block_bits),
bch2_err_str(ret));
__bch2_journal_debug_to_text(&buf, j);
spin_unlock(&j->lock);
bch2_print_string_as_lines(KERN_ERR, buf.buf);
printbuf_exit(&buf);
- goto err;
}
+ if (ret)
+ goto err;
/*
* write is allocated, no longer need to account for it in
generated by cgit v1.2.3 (git 2.39.0) at 2025-05-01 19:20:07 +0300