#include "journal.h"
#include "journal_io.h"
#include "journal_reclaim.h"
+#include "journal_sb.h"
#include "journal_seq_blacklist.h"
-#include "super-io.h"
+#include "trace.h"
-#include <trace/events/bcachefs.h>
+static const char * const bch2_journal_errors[] = {
+#define x(n) #n,
+ JOURNAL_ERRORS()
+#undef x
+ NULL
+};
static inline bool journal_seq_unwritten(struct journal *j, u64 seq)
{
static void journal_pin_list_init(struct journal_entry_pin_list *p, int count)
{
- INIT_LIST_HEAD(&p->list);
- INIT_LIST_HEAD(&p->key_cache_list);
+ unsigned i;
+ for (i = 0; i < ARRAY_SIZE(p->list); i++)
+ INIT_LIST_HEAD(&p->list[i]);
INIT_LIST_HEAD(&p->flushed);
atomic_set(&p->count, count);
p->devs.nr = 0;
}
+/*
+ * Detect stuck journal conditions and trigger shutdown. Technically the journal
+ * can end up stuck for a variety of reasons, such as a blocked I/O, journal
+ * reservation lockup, etc. Since this is a fatal error with potentially
+ * unpredictable characteristics, we want to be fairly conservative before we
+ * decide to shut things down.
+ *
+ * Consider the journal stuck when it appears full with no ability to commit
+ * btree transactions, to discard journal buckets, nor acquire priority
+ * (reserved watermark) reservation.
+ */
+static inline bool
+journal_error_check_stuck(struct journal *j, int error, unsigned flags)
+{
+ struct bch_fs *c = container_of(j, struct bch_fs, journal);
+ bool stuck = false;
+ struct printbuf buf = PRINTBUF;
+
+ if (!(error == JOURNAL_ERR_journal_full ||
+ error == JOURNAL_ERR_journal_pin_full) ||
+ nr_unwritten_journal_entries(j) ||
+ (flags & BCH_WATERMARK_MASK) != BCH_WATERMARK_reclaim)
+ return stuck;
+
+ spin_lock(&j->lock);
+
+ if (j->can_discard) {
+ spin_unlock(&j->lock);
+ return stuck;
+ }
+
+ stuck = true;
+
+ /*
+ * The journal shutdown path will set ->err_seq, but do it here first to
+ * serialize against concurrent failures and avoid duplicate error
+ * reports.
+ */
+ if (j->err_seq) {
+ spin_unlock(&j->lock);
+ return stuck;
+ }
+ j->err_seq = journal_cur_seq(j);
+ spin_unlock(&j->lock);
+
+ bch_err(c, "Journal stuck! Hava a pre-reservation but journal full (error %s)",
+ bch2_journal_errors[error]);
+ bch2_journal_debug_to_text(&buf, j);
+ bch_err(c, "%s", buf.buf);
+
+ printbuf_reset(&buf);
+ bch2_journal_pins_to_text(&buf, j);
+ bch_err(c, "Journal pins:\n%s", buf.buf);
+ printbuf_exit(&buf);
+
+ bch2_fatal_error(c);
+ dump_stack();
+
+ return stuck;
+}
+
/* journal entry close/open: */
void __bch2_journal_buf_put(struct journal *j)
__journal_entry_close(j, JOURNAL_ENTRY_ERROR_VAL);
if (!j->err_seq)
j->err_seq = journal_cur_seq(j);
+ journal_wake(j);
spin_unlock(&j->lock);
}
/*
* should _only_ called from journal_res_get() - when we actually want a
* journal reservation - journal entry is open means journal is dirty:
- *
- * returns:
- * 0: success
- * -ENOSPC: journal currently full, must invoke reclaim
- * -EAGAIN: journal blocked, must wait
- * -EROFS: insufficient rw devices or journal error
*/
static int journal_entry_open(struct journal *j)
{
BUG_ON(BCH_SB_CLEAN(c->disk_sb.sb));
if (j->blocked)
- return cur_entry_blocked;
+ return JOURNAL_ERR_blocked;
if (j->cur_entry_error)
return j->cur_entry_error;
if (bch2_journal_error(j))
- return cur_entry_insufficient_devices; /* -EROFS */
+ return JOURNAL_ERR_insufficient_devices; /* -EROFS */
if (!fifo_free(&j->pin))
- return cur_entry_journal_pin_full;
+ return JOURNAL_ERR_journal_pin_full;
- if (nr_unwritten_journal_entries(j) == ARRAY_SIZE(j->buf) - 1)
- return cur_entry_max_in_flight;
+ if (nr_unwritten_journal_entries(j) == ARRAY_SIZE(j->buf))
+ return JOURNAL_ERR_max_in_flight;
BUG_ON(!j->cur_entry_sectors);
journal_entry_overhead(j);
u64s = clamp_t(int, u64s, 0, JOURNAL_ENTRY_CLOSED_VAL - 1);
- if (u64s <= 0)
- return cur_entry_journal_full;
+ if (u64s <= (ssize_t) j->early_journal_entries.nr)
+ return JOURNAL_ERR_journal_full;
+
+ if (fifo_empty(&j->pin) && j->reclaim_thread)
+ wake_up_process(j->reclaim_thread);
/*
* The fifo_push() needs to happen at the same time as j->seq is
buf->data->seq = cpu_to_le64(journal_cur_seq(j));
buf->data->u64s = 0;
+ if (j->early_journal_entries.nr) {
+ memcpy(buf->data->_data, j->early_journal_entries.data,
+ j->early_journal_entries.nr * sizeof(u64));
+ le32_add_cpu(&buf->data->u64s, j->early_journal_entries.nr);
+ }
+
/*
* Must be set before marking the journal entry as open:
*/
BUG_ON(new.idx != (journal_cur_seq(j) & JOURNAL_BUF_MASK));
journal_state_inc(&new);
- new.cur_entry_offset = 0;
+
+ /* Handle any already added entries */
+ new.cur_entry_offset = le32_to_cpu(buf->data->u64s);
} while ((v = atomic64_cmpxchg(&j->reservations.counter,
old.v, new.v)) != old.v);
&j->write_work,
msecs_to_jiffies(c->opts.journal_flush_delay));
journal_wake(j);
+
+ if (j->early_journal_entries.nr)
+ darray_exit(&j->early_journal_entries);
return 0;
}
{
struct journal *j = container_of(work, struct journal, write_work.work);
struct bch_fs *c = container_of(j, struct bch_fs, journal);
- struct journal_buf *buf;
long delta;
spin_lock(&j->lock);
- buf = journal_cur_buf(j);
- delta = buf->expires - jiffies;
+ if (!__journal_entry_is_open(j->reservations))
+ goto unlock;
+
+ delta = journal_cur_buf(j)->expires - jiffies;
if (delta > 0)
mod_delayed_work(c->io_complete_wq, &j->write_work, delta);
else
__journal_entry_close(j, JOURNAL_ENTRY_CLOSED_VAL);
+unlock:
spin_unlock(&j->lock);
}
return 0;
if (bch2_journal_error(j))
- return -EROFS;
+ return -BCH_ERR_erofs_journal_err;
spin_lock(&j->lock);
+ /* check once more in case somebody else shut things down... */
+ if (bch2_journal_error(j)) {
+ spin_unlock(&j->lock);
+ return -BCH_ERR_erofs_journal_err;
+ }
+
/*
* Recheck after taking the lock, so we don't race with another thread
* that just did journal_entry_open() and call journal_entry_close()
return 0;
}
- if (!(flags & JOURNAL_RES_GET_RESERVED) &&
- !test_bit(JOURNAL_MAY_GET_UNRESERVED, &j->flags)) {
+ if ((flags & BCH_WATERMARK_MASK) < j->watermark) {
/*
* Don't want to close current journal entry, just need to
* invoke reclaim:
*/
- ret = cur_entry_journal_full;
+ ret = JOURNAL_ERR_journal_full;
goto unlock;
}
__journal_entry_close(j, JOURNAL_ENTRY_CLOSED_VAL);
ret = journal_entry_open(j);
- if (ret == cur_entry_max_in_flight)
- trace_journal_entry_full(c);
+ if (ret == JOURNAL_ERR_max_in_flight)
+ trace_and_count(c, journal_entry_full, c);
unlock:
- if ((ret && ret != cur_entry_insufficient_devices) &&
+ if ((ret && ret != JOURNAL_ERR_insufficient_devices) &&
!j->res_get_blocked_start) {
j->res_get_blocked_start = local_clock() ?: 1;
- trace_journal_full(c);
+ trace_and_count(c, journal_full, c);
}
can_discard = j->can_discard;
if (!ret)
goto retry;
-
- if ((ret == cur_entry_journal_full ||
- ret == cur_entry_journal_pin_full) &&
- !can_discard &&
- !nr_unwritten_journal_entries(j) &&
- (flags & JOURNAL_RES_GET_RESERVED)) {
- struct printbuf buf = PRINTBUF;
-
- bch_err(c, "Journal stuck! Hava a pre-reservation but journal full");
-
- bch2_journal_debug_to_text(&buf, j);
- bch_err(c, "%s", buf.buf);
-
- printbuf_reset(&buf);
- bch2_journal_pins_to_text(&buf, j);
- bch_err(c, "Journal pins:\n%s", buf.buf);
-
- printbuf_exit(&buf);
- bch2_fatal_error(c);
- dump_stack();
- }
+ if (journal_error_check_stuck(j, ret, flags))
+ ret = -BCH_ERR_journal_res_get_blocked;
/*
* Journal is full - can't rely on reclaim from work item due to
* freezing:
*/
- if ((ret == cur_entry_journal_full ||
- ret == cur_entry_journal_pin_full) &&
+ if ((ret == JOURNAL_ERR_journal_full ||
+ ret == JOURNAL_ERR_journal_pin_full) &&
!(flags & JOURNAL_RES_GET_NONBLOCK)) {
if (can_discard) {
bch2_journal_do_discards(j);
}
}
- return ret == cur_entry_insufficient_devices ? -EROFS : -EAGAIN;
+ return ret == JOURNAL_ERR_insufficient_devices
+ ? -EROFS
+ : -BCH_ERR_journal_res_get_blocked;
}
/*
int ret;
closure_wait_event(&j->async_wait,
- (ret = __journal_res_get(j, res, flags)) != -EAGAIN ||
+ (ret = __journal_res_get(j, res, flags)) !=
+ -BCH_ERR_journal_res_get_blocked||
(flags & JOURNAL_RES_GET_NONBLOCK));
return ret;
}
return ret ?: ret2 < 0 ? ret2 : 0;
}
-int bch2_journal_meta(struct journal *j)
-{
- struct journal_buf *buf;
- struct journal_res res;
- int ret;
-
- memset(&res, 0, sizeof(res));
-
- ret = bch2_journal_res_get(j, &res, jset_u64s(0), 0);
- if (ret)
- return ret;
-
- buf = j->buf + (res.seq & JOURNAL_BUF_MASK);
- buf->must_flush = true;
-
- if (!buf->flush_time) {
- buf->flush_time = local_clock() ?: 1;
- buf->expires = jiffies;
- }
-
- bch2_journal_res_put(j, &res);
-
- return bch2_journal_flush_seq(j, res.seq);
-}
-
/*
* bch2_journal_flush_async - if there is an open journal entry, or a journal
* still being written, write it and wait for the write to complete
return ret;
}
+int bch2_journal_meta(struct journal *j)
+{
+ struct journal_buf *buf;
+ struct journal_res res;
+ int ret;
+
+ memset(&res, 0, sizeof(res));
+
+ ret = bch2_journal_res_get(j, &res, jset_u64s(0), 0);
+ if (ret)
+ return ret;
+
+ buf = j->buf + (res.seq & JOURNAL_BUF_MASK);
+ buf->must_flush = true;
+
+ if (!buf->flush_time) {
+ buf->flush_time = local_clock() ?: 1;
+ buf->expires = jiffies;
+ }
+
+ bch2_journal_res_put(j, &res);
+
+ return bch2_journal_flush_seq(j, res.seq);
+}
+
/* block/unlock the journal: */
void bch2_journal_unblock(struct journal *j)
{
struct bch_fs *c = ca->fs;
struct journal_device *ja = &ca->journal;
- struct bch_sb_field_journal *journal_buckets;
u64 *new_bucket_seq = NULL, *new_buckets = NULL;
+ struct open_bucket **ob = NULL;
+ long *bu = NULL;
+ unsigned i, pos, nr_got = 0, nr_want = nr - ja->nr;
int ret = 0;
- /* don't handle reducing nr of buckets yet: */
- if (nr <= ja->nr)
- return 0;
+ BUG_ON(nr <= ja->nr);
- new_buckets = kzalloc(nr * sizeof(u64), GFP_KERNEL);
- new_bucket_seq = kzalloc(nr * sizeof(u64), GFP_KERNEL);
- if (!new_buckets || !new_bucket_seq) {
- ret = -ENOMEM;
- goto err;
+ bu = kcalloc(nr_want, sizeof(*bu), GFP_KERNEL);
+ ob = kcalloc(nr_want, sizeof(*ob), GFP_KERNEL);
+ new_buckets = kcalloc(nr, sizeof(u64), GFP_KERNEL);
+ new_bucket_seq = kcalloc(nr, sizeof(u64), GFP_KERNEL);
+ if (!bu || !ob || !new_buckets || !new_bucket_seq) {
+ ret = -BCH_ERR_ENOMEM_set_nr_journal_buckets;
+ goto err_free;
}
- journal_buckets = bch2_sb_resize_journal(&ca->disk_sb,
- nr + sizeof(*journal_buckets) / sizeof(u64));
- if (!journal_buckets) {
- ret = -ENOSPC;
- goto err;
+ for (nr_got = 0; nr_got < nr_want; nr_got++) {
+ if (new_fs) {
+ bu[nr_got] = bch2_bucket_alloc_new_fs(ca);
+ if (bu[nr_got] < 0) {
+ ret = -BCH_ERR_ENOSPC_bucket_alloc;
+ break;
+ }
+ } else {
+ ob[nr_got] = bch2_bucket_alloc(c, ca, BCH_WATERMARK_normal, cl);
+ ret = PTR_ERR_OR_ZERO(ob[nr_got]);
+ if (ret)
+ break;
+
+ ret = bch2_trans_run(c,
+ bch2_trans_mark_metadata_bucket(&trans, ca,
+ ob[nr_got]->bucket, BCH_DATA_journal,
+ ca->mi.bucket_size));
+ if (ret) {
+ bch2_open_bucket_put(c, ob[nr_got]);
+ bch_err(c, "error marking new journal buckets: %s", bch2_err_str(ret));
+ break;
+ }
+
+ bu[nr_got] = ob[nr_got]->bucket;
+ }
}
- /*
- * We may be called from the device add path, before the new device has
- * actually been added to the running filesystem:
- */
- if (!new_fs)
- spin_lock(&c->journal.lock);
+ if (!nr_got)
+ goto err_free;
+
+ /* Don't return an error if we successfully allocated some buckets: */
+ ret = 0;
+
+ if (c) {
+ bch2_journal_flush_all_pins(&c->journal);
+ bch2_journal_block(&c->journal);
+ mutex_lock(&c->sb_lock);
+ }
memcpy(new_buckets, ja->buckets, ja->nr * sizeof(u64));
memcpy(new_bucket_seq, ja->bucket_seq, ja->nr * sizeof(u64));
- swap(new_buckets, ja->buckets);
- swap(new_bucket_seq, ja->bucket_seq);
- if (!new_fs)
- spin_unlock(&c->journal.lock);
+ BUG_ON(ja->discard_idx > ja->nr);
- while (ja->nr < nr) {
- struct open_bucket *ob = NULL;
- unsigned pos;
- long b;
+ pos = ja->discard_idx ?: ja->nr;
- if (new_fs) {
- b = bch2_bucket_alloc_new_fs(ca);
- if (b < 0) {
- ret = -ENOSPC;
- goto err;
- }
- } else {
- rcu_read_lock();
- ob = bch2_bucket_alloc(c, ca, RESERVE_NONE,
- false, cl);
- rcu_read_unlock();
- if (IS_ERR(ob)) {
- ret = cl ? -EAGAIN : -ENOSPC;
- goto err;
- }
+ memmove(new_buckets + pos + nr_got,
+ new_buckets + pos,
+ sizeof(new_buckets[0]) * (ja->nr - pos));
+ memmove(new_bucket_seq + pos + nr_got,
+ new_bucket_seq + pos,
+ sizeof(new_bucket_seq[0]) * (ja->nr - pos));
- b = ob->bucket;
- }
+ for (i = 0; i < nr_got; i++) {
+ new_buckets[pos + i] = bu[i];
+ new_bucket_seq[pos + i] = 0;
+ }
- if (c)
- spin_lock(&c->journal.lock);
+ nr = ja->nr + nr_got;
- /*
- * XXX
- * For resize at runtime, we should be writing the new
- * superblock before inserting into the journal array
- */
+ ret = bch2_journal_buckets_to_sb(c, ca, new_buckets, nr);
+ if (ret)
+ goto err_unblock;
- pos = ja->nr ? (ja->cur_idx + 1) % ja->nr : 0;
- __array_insert_item(ja->buckets, ja->nr, pos);
- __array_insert_item(ja->bucket_seq, ja->nr, pos);
- __array_insert_item(journal_buckets->buckets, ja->nr, pos);
- ja->nr++;
-
- ja->buckets[pos] = b;
- ja->bucket_seq[pos] = 0;
- journal_buckets->buckets[pos] = cpu_to_le64(b);
-
- if (pos <= ja->discard_idx)
- ja->discard_idx = (ja->discard_idx + 1) % ja->nr;
- if (pos <= ja->dirty_idx_ondisk)
- ja->dirty_idx_ondisk = (ja->dirty_idx_ondisk + 1) % ja->nr;
- if (pos <= ja->dirty_idx)
- ja->dirty_idx = (ja->dirty_idx + 1) % ja->nr;
- if (pos <= ja->cur_idx)
- ja->cur_idx = (ja->cur_idx + 1) % ja->nr;
-
- if (c)
- spin_unlock(&c->journal.lock);
-
- if (!new_fs) {
- ret = bch2_trans_do(c, NULL, NULL, BTREE_INSERT_NOFAIL,
- bch2_trans_mark_metadata_bucket(&trans, ca,
- b, BCH_DATA_journal,
- ca->mi.bucket_size));
+ if (!new_fs)
+ bch2_write_super(c);
- bch2_open_bucket_put(c, ob);
+ /* Commit: */
+ if (c)
+ spin_lock(&c->journal.lock);
- if (ret)
- goto err;
- }
+ swap(new_buckets, ja->buckets);
+ swap(new_bucket_seq, ja->bucket_seq);
+ ja->nr = nr;
+
+ if (pos <= ja->discard_idx)
+ ja->discard_idx = (ja->discard_idx + nr_got) % ja->nr;
+ if (pos <= ja->dirty_idx_ondisk)
+ ja->dirty_idx_ondisk = (ja->dirty_idx_ondisk + nr_got) % ja->nr;
+ if (pos <= ja->dirty_idx)
+ ja->dirty_idx = (ja->dirty_idx + nr_got) % ja->nr;
+ if (pos <= ja->cur_idx)
+ ja->cur_idx = (ja->cur_idx + nr_got) % ja->nr;
+
+ if (c)
+ spin_unlock(&c->journal.lock);
+err_unblock:
+ if (c) {
+ bch2_journal_unblock(&c->journal);
+ mutex_unlock(&c->sb_lock);
}
-err:
- bch2_sb_resize_journal(&ca->disk_sb,
- ja->nr + sizeof(*journal_buckets) / sizeof(u64));
+
+ if (ret && !new_fs)
+ for (i = 0; i < nr_got; i++)
+ bch2_trans_run(c,
+ bch2_trans_mark_metadata_bucket(&trans, ca,
+ bu[i], BCH_DATA_free, 0));
+err_free:
+ if (!new_fs)
+ for (i = 0; i < nr_got; i++)
+ bch2_open_bucket_put(c, ob[i]);
+
kfree(new_bucket_seq);
kfree(new_buckets);
-
+ kfree(ob);
+ kfree(bu);
return ret;
}
{
struct journal_device *ja = &ca->journal;
struct closure cl;
- unsigned current_nr;
- int ret;
+ int ret = 0;
closure_init_stack(&cl);
- do {
- struct disk_reservation disk_res = { 0, 0 };
+ down_write(&c->state_lock);
- closure_sync(&cl);
+ /* don't handle reducing nr of buckets yet: */
+ if (nr < ja->nr)
+ goto unlock;
- mutex_lock(&c->sb_lock);
- current_nr = ja->nr;
+ while (ja->nr < nr) {
+ struct disk_reservation disk_res = { 0, 0 };
/*
* note: journal buckets aren't really counted as _sectors_ used yet, so
* we don't need the disk reservation to avoid the BUG_ON() in buckets.c
* when space used goes up without a reservation - but we do need the
* reservation to ensure we'll actually be able to allocate:
+ *
+ * XXX: that's not right, disk reservations only ensure a
+ * filesystem-wide allocation will succeed, this is a device
+ * specific allocation - we can hang here:
*/
- if (bch2_disk_reservation_get(c, &disk_res,
- bucket_to_sector(ca, nr - ja->nr), 1, 0)) {
- mutex_unlock(&c->sb_lock);
- return -ENOSPC;
- }
+ ret = bch2_disk_reservation_get(c, &disk_res,
+ bucket_to_sector(ca, nr - ja->nr), 1, 0);
+ if (ret)
+ break;
ret = __bch2_set_nr_journal_buckets(ca, nr, false, &cl);
bch2_disk_reservation_put(c, &disk_res);
- if (ja->nr != current_nr)
- bch2_write_super(c);
- mutex_unlock(&c->sb_lock);
- } while (ret == -EAGAIN);
+ closure_sync(&cl);
+
+ if (ret && ret != -BCH_ERR_bucket_alloc_blocked)
+ break;
+ }
+ if (ret)
+ bch_err_fn(c, ret);
+unlock:
+ up_write(&c->state_lock);
return ret;
}
int bch2_dev_journal_alloc(struct bch_dev *ca)
{
unsigned nr;
+ int ret;
- if (dynamic_fault("bcachefs:add:journal_alloc"))
- return -ENOMEM;
+ if (dynamic_fault("bcachefs:add:journal_alloc")) {
+ ret = -BCH_ERR_ENOMEM_set_nr_journal_buckets;
+ goto err;
+ }
/* 1/128th of the device by default: */
nr = ca->mi.nbuckets >> 7;
min(1 << 13,
(1 << 24) / ca->mi.bucket_size));
- return __bch2_set_nr_journal_buckets(ca, nr, true, NULL);
+ ret = __bch2_set_nr_journal_buckets(ca, nr, true, NULL);
+err:
+ if (ret)
+ bch_err_fn(ca, ret);
+ return ret;
}
/* startup/shutdown: */
seq++) {
struct journal_buf *buf = journal_seq_to_buf(j, seq);
- if (bch2_bkey_has_device(bkey_i_to_s_c(&buf->key), dev_idx))
+ if (bch2_bkey_has_device_c(bkey_i_to_s_c(&buf->key), dev_idx))
ret = true;
}
spin_unlock(&j->lock);
void bch2_fs_journal_stop(struct journal *j)
{
+ bch2_journal_reclaim_stop(j);
bch2_journal_flush_all_pins(j);
wait_event(j->wait, journal_entry_close(j));
j->last_empty_seq != journal_cur_seq(j));
cancel_delayed_work_sync(&j->write_work);
- bch2_journal_reclaim_stop(j);
}
-int bch2_fs_journal_start(struct journal *j, u64 cur_seq,
- struct list_head *journal_entries)
+int bch2_fs_journal_start(struct journal *j, u64 cur_seq)
{
struct bch_fs *c = container_of(j, struct bch_fs, journal);
struct journal_entry_pin_list *p;
- struct journal_replay *i;
+ struct journal_replay *i, **_i;
+ struct genradix_iter iter;
+ bool had_entries = false;
+ unsigned ptr;
u64 last_seq = cur_seq, nr, seq;
- if (!list_empty(journal_entries))
- last_seq = le64_to_cpu(list_last_entry(journal_entries,
- struct journal_replay, list)->j.last_seq);
+ genradix_for_each_reverse(&c->journal_entries, iter, _i) {
+ i = *_i;
+
+ if (!i || i->ignore)
+ continue;
+
+ last_seq = le64_to_cpu(i->j.last_seq);
+ break;
+ }
nr = cur_seq - last_seq;
init_fifo(&j->pin, roundup_pow_of_two(nr + 1), GFP_KERNEL);
if (!j->pin.data) {
bch_err(c, "error reallocating journal fifo (%llu open entries)", nr);
- return -ENOMEM;
+ return -BCH_ERR_ENOMEM_journal_pin_fifo;
}
}
j->pin.back = cur_seq;
atomic64_set(&j->seq, cur_seq - 1);
- if (list_empty(journal_entries))
- j->last_empty_seq = cur_seq - 1;
-
fifo_for_each_entry_ptr(p, &j->pin, seq)
journal_pin_list_init(p, 1);
- list_for_each_entry(i, journal_entries, list) {
- unsigned ptr;
+ genradix_for_each(&c->journal_entries, iter, _i) {
+ i = *_i;
+
+ if (!i || i->ignore)
+ continue;
seq = le64_to_cpu(i->j.seq);
BUG_ON(seq >= cur_seq);
p->devs.nr = 0;
for (ptr = 0; ptr < i->nr_ptrs; ptr++)
bch2_dev_list_add_dev(&p->devs, i->ptrs[ptr].dev);
+
+ had_entries = true;
}
- if (list_empty(journal_entries))
+ if (!had_entries)
j->last_empty_seq = cur_seq;
spin_lock(&j->lock);
struct journal_device *ja = &ca->journal;
struct bch_sb_field_journal *journal_buckets =
bch2_sb_get_journal(sb);
- unsigned i;
+ struct bch_sb_field_journal_v2 *journal_buckets_v2 =
+ bch2_sb_get_journal_v2(sb);
+ unsigned i, nr_bvecs;
+
+ ja->nr = 0;
- ja->nr = bch2_nr_journal_buckets(journal_buckets);
+ if (journal_buckets_v2) {
+ unsigned nr = bch2_sb_field_journal_v2_nr_entries(journal_buckets_v2);
+
+ for (i = 0; i < nr; i++)
+ ja->nr += le64_to_cpu(journal_buckets_v2->d[i].nr);
+ } else if (journal_buckets) {
+ ja->nr = bch2_nr_journal_buckets(journal_buckets);
+ }
ja->bucket_seq = kcalloc(ja->nr, sizeof(u64), GFP_KERNEL);
if (!ja->bucket_seq)
- return -ENOMEM;
+ return -BCH_ERR_ENOMEM_dev_journal_init;
+
+ nr_bvecs = DIV_ROUND_UP(JOURNAL_ENTRY_SIZE_MAX, PAGE_SIZE);
- ca->journal.bio = bio_kmalloc(GFP_KERNEL,
- DIV_ROUND_UP(JOURNAL_ENTRY_SIZE_MAX, PAGE_SIZE));
+ ca->journal.bio = bio_kmalloc(nr_bvecs, GFP_KERNEL);
if (!ca->journal.bio)
- return -ENOMEM;
+ return -BCH_ERR_ENOMEM_dev_journal_init;
+
+ bio_init(ca->journal.bio, NULL, ca->journal.bio->bi_inline_vecs, nr_bvecs, 0);
ja->buckets = kcalloc(ja->nr, sizeof(u64), GFP_KERNEL);
if (!ja->buckets)
- return -ENOMEM;
-
- for (i = 0; i < ja->nr; i++)
- ja->buckets[i] = le64_to_cpu(journal_buckets->buckets[i]);
+ return -BCH_ERR_ENOMEM_dev_journal_init;
+
+ if (journal_buckets_v2) {
+ unsigned nr = bch2_sb_field_journal_v2_nr_entries(journal_buckets_v2);
+ unsigned j, dst = 0;
+
+ for (i = 0; i < nr; i++)
+ for (j = 0; j < le64_to_cpu(journal_buckets_v2->d[i].nr); j++)
+ ja->buckets[dst++] =
+ le64_to_cpu(journal_buckets_v2->d[i].start) + j;
+ } else if (journal_buckets) {
+ for (i = 0; i < ja->nr; i++)
+ ja->buckets[i] = le64_to_cpu(journal_buckets->buckets[i]);
+ }
return 0;
}
{
unsigned i;
+ darray_exit(&j->early_journal_entries);
+
for (i = 0; i < ARRAY_SIZE(j->buf); i++)
kvpfree(j->buf[i].data, j->buf[i].buf_size);
free_fifo(&j->pin);
{ .cur_entry_offset = JOURNAL_ENTRY_CLOSED_VAL }).v);
if (!(init_fifo(&j->pin, JOURNAL_PIN, GFP_KERNEL))) {
- ret = -ENOMEM;
+ ret = -BCH_ERR_ENOMEM_journal_pin_fifo;
goto out;
}
j->buf[i].buf_size = JOURNAL_ENTRY_SIZE_MIN;
j->buf[i].data = kvpmalloc(j->buf[i].buf_size, GFP_KERNEL);
if (!j->buf[i].data) {
- ret = -ENOMEM;
+ ret = -BCH_ERR_ENOMEM_journal_buf;
goto out;
}
}
u64 seq;
unsigned i;
+ if (!out->nr_tabstops)
+ printbuf_tabstop_push(out, 24);
out->atomic++;
- out->tabstops[0] = 24;
rcu_read_lock();
s = READ_ONCE(j->reservations);
- pr_buf(out, "dirty journal entries:\t%llu\n", fifo_used(&j->pin));
- pr_buf(out, "seq:\t\t\t%llu\n", journal_cur_seq(j));
- pr_buf(out, "seq_ondisk:\t\t%llu\n", j->seq_ondisk);
- pr_buf(out, "last_seq:\t\t%llu\n", journal_last_seq(j));
- pr_buf(out, "last_seq_ondisk:\t%llu\n", j->last_seq_ondisk);
- pr_buf(out, "flushed_seq_ondisk:\t%llu\n", j->flushed_seq_ondisk);
- pr_buf(out, "prereserved:\t\t%u/%u\n", j->prereserved.reserved, j->prereserved.remaining);
- pr_buf(out, "each entry reserved:\t%u\n", j->entry_u64s_reserved);
- pr_buf(out, "nr flush writes:\t%llu\n", j->nr_flush_writes);
- pr_buf(out, "nr noflush writes:\t%llu\n", j->nr_noflush_writes);
- pr_buf(out, "nr direct reclaim:\t%llu\n", j->nr_direct_reclaim);
- pr_buf(out, "nr background reclaim:\t%llu\n", j->nr_background_reclaim);
- pr_buf(out, "reclaim kicked:\t\t%u\n", j->reclaim_kicked);
- pr_buf(out, "reclaim runs in:\t%u ms\n", time_after(j->next_reclaim, now)
+ prt_printf(out, "dirty journal entries:\t%llu/%llu\n", fifo_used(&j->pin), j->pin.size);
+ prt_printf(out, "seq:\t\t\t%llu\n", journal_cur_seq(j));
+ prt_printf(out, "seq_ondisk:\t\t%llu\n", j->seq_ondisk);
+ prt_printf(out, "last_seq:\t\t%llu\n", journal_last_seq(j));
+ prt_printf(out, "last_seq_ondisk:\t%llu\n", j->last_seq_ondisk);
+ prt_printf(out, "flushed_seq_ondisk:\t%llu\n", j->flushed_seq_ondisk);
+ prt_printf(out, "prereserved:\t\t%u/%u\n", j->prereserved.reserved, j->prereserved.remaining);
+ prt_printf(out, "watermark:\t\t%s\n", bch2_watermarks[j->watermark]);
+ prt_printf(out, "each entry reserved:\t%u\n", j->entry_u64s_reserved);
+ prt_printf(out, "nr flush writes:\t%llu\n", j->nr_flush_writes);
+ prt_printf(out, "nr noflush writes:\t%llu\n", j->nr_noflush_writes);
+ prt_printf(out, "nr direct reclaim:\t%llu\n", j->nr_direct_reclaim);
+ prt_printf(out, "nr background reclaim:\t%llu\n", j->nr_background_reclaim);
+ prt_printf(out, "reclaim kicked:\t\t%u\n", j->reclaim_kicked);
+ prt_printf(out, "reclaim runs in:\t%u ms\n", time_after(j->next_reclaim, now)
? jiffies_to_msecs(j->next_reclaim - jiffies) : 0);
- pr_buf(out, "current entry sectors:\t%u\n", j->cur_entry_sectors);
- pr_buf(out, "current entry error:\t%u\n", j->cur_entry_error);
- pr_buf(out, "current entry:\t\t");
+ prt_printf(out, "current entry sectors:\t%u\n", j->cur_entry_sectors);
+ prt_printf(out, "current entry error:\t%s\n", bch2_journal_errors[j->cur_entry_error]);
+ prt_printf(out, "current entry:\t\t");
switch (s.cur_entry_offset) {
case JOURNAL_ENTRY_ERROR_VAL:
- pr_buf(out, "error");
+ prt_printf(out, "error");
break;
case JOURNAL_ENTRY_CLOSED_VAL:
- pr_buf(out, "closed");
+ prt_printf(out, "closed");
break;
default:
- pr_buf(out, "%u/%u", s.cur_entry_offset, j->cur_entry_u64s);
+ prt_printf(out, "%u/%u", s.cur_entry_offset, j->cur_entry_u64s);
break;
}
- pr_newline(out);
+ prt_newline(out);
for (seq = journal_cur_seq(j);
seq >= journal_last_unwritten_seq(j);
--seq) {
i = seq & JOURNAL_BUF_MASK;
- pr_buf(out, "unwritten entry:");
- pr_tab(out);
- pr_buf(out, "%llu", seq);
- pr_newline(out);
- pr_indent_push(out, 2);
+ prt_printf(out, "unwritten entry:");
+ prt_tab(out);
+ prt_printf(out, "%llu", seq);
+ prt_newline(out);
+ printbuf_indent_add(out, 2);
- pr_buf(out, "refcount:");
- pr_tab(out);
- pr_buf(out, "%u", journal_state_count(s, i));
- pr_newline(out);
+ prt_printf(out, "refcount:");
+ prt_tab(out);
+ prt_printf(out, "%u", journal_state_count(s, i));
+ prt_newline(out);
- pr_buf(out, "sectors:");
- pr_tab(out);
- pr_buf(out, "%u", j->buf[i].sectors);
- pr_newline(out);
+ prt_printf(out, "sectors:");
+ prt_tab(out);
+ prt_printf(out, "%u", j->buf[i].sectors);
+ prt_newline(out);
- pr_buf(out, "expires");
- pr_tab(out);
- pr_buf(out, "%li jiffies", j->buf[i].expires - jiffies);
- pr_newline(out);
+ prt_printf(out, "expires");
+ prt_tab(out);
+ prt_printf(out, "%li jiffies", j->buf[i].expires - jiffies);
+ prt_newline(out);
- pr_indent_pop(out, 2);
+ printbuf_indent_sub(out, 2);
}
- pr_buf(out,
+ prt_printf(out,
"replay done:\t\t%i\n",
test_bit(JOURNAL_REPLAY_DONE, &j->flags));
- pr_buf(out, "space:\n");
- pr_buf(out, "\tdiscarded\t%u:%u\n",
+ prt_printf(out, "space:\n");
+ prt_printf(out, "\tdiscarded\t%u:%u\n",
j->space[journal_space_discarded].next_entry,
j->space[journal_space_discarded].total);
- pr_buf(out, "\tclean ondisk\t%u:%u\n",
+ prt_printf(out, "\tclean ondisk\t%u:%u\n",
j->space[journal_space_clean_ondisk].next_entry,
j->space[journal_space_clean_ondisk].total);
- pr_buf(out, "\tclean\t\t%u:%u\n",
+ prt_printf(out, "\tclean\t\t%u:%u\n",
j->space[journal_space_clean].next_entry,
j->space[journal_space_clean].total);
- pr_buf(out, "\ttotal\t\t%u:%u\n",
+ prt_printf(out, "\ttotal\t\t%u:%u\n",
j->space[journal_space_total].next_entry,
j->space[journal_space_total].total);
if (!ja->nr)
continue;
- pr_buf(out, "dev %u:\n", i);
- pr_buf(out, "\tnr\t\t%u\n", ja->nr);
- pr_buf(out, "\tbucket size\t%u\n", ca->mi.bucket_size);
- pr_buf(out, "\tavailable\t%u:%u\n", bch2_journal_dev_buckets_available(j, ja, journal_space_discarded), ja->sectors_free);
- pr_buf(out, "\tdiscard_idx\t%u\n", ja->discard_idx);
- pr_buf(out, "\tdirty_ondisk\t%u (seq %llu)\n", ja->dirty_idx_ondisk, ja->bucket_seq[ja->dirty_idx_ondisk]);
- pr_buf(out, "\tdirty_idx\t%u (seq %llu)\n", ja->dirty_idx, ja->bucket_seq[ja->dirty_idx]);
- pr_buf(out, "\tcur_idx\t\t%u (seq %llu)\n", ja->cur_idx, ja->bucket_seq[ja->cur_idx]);
+ prt_printf(out, "dev %u:\n", i);
+ prt_printf(out, "\tnr\t\t%u\n", ja->nr);
+ prt_printf(out, "\tbucket size\t%u\n", ca->mi.bucket_size);
+ prt_printf(out, "\tavailable\t%u:%u\n", bch2_journal_dev_buckets_available(j, ja, journal_space_discarded), ja->sectors_free);
+ prt_printf(out, "\tdiscard_idx\t%u\n", ja->discard_idx);
+ prt_printf(out, "\tdirty_ondisk\t%u (seq %llu)\n", ja->dirty_idx_ondisk, ja->bucket_seq[ja->dirty_idx_ondisk]);
+ prt_printf(out, "\tdirty_idx\t%u (seq %llu)\n", ja->dirty_idx, ja->bucket_seq[ja->dirty_idx]);
+ prt_printf(out, "\tcur_idx\t\t%u (seq %llu)\n", ja->cur_idx, ja->bucket_seq[ja->cur_idx]);
}
rcu_read_unlock();
{
struct journal_entry_pin_list *pin_list;
struct journal_entry_pin *pin;
+ unsigned i;
spin_lock(&j->lock);
*seq = max(*seq, j->pin.front);
pin_list = journal_seq_pin(j, *seq);
- pr_buf(out, "%llu: count %u", *seq, atomic_read(&pin_list->count));
- pr_newline(out);
- pr_indent_push(out, 2);
-
- list_for_each_entry(pin, &pin_list->list, list) {
- pr_buf(out, "\t%px %ps", pin, pin->flush);
- pr_newline(out);
- }
+ prt_printf(out, "%llu: count %u", *seq, atomic_read(&pin_list->count));
+ prt_newline(out);
+ printbuf_indent_add(out, 2);
- list_for_each_entry(pin, &pin_list->key_cache_list, list) {
- pr_buf(out, "\t%px %ps", pin, pin->flush);
- pr_newline(out);
- }
+ for (i = 0; i < ARRAY_SIZE(pin_list->list); i++)
+ list_for_each_entry(pin, &pin_list->list[i], list) {
+ prt_printf(out, "\t%px %ps", pin, pin->flush);
+ prt_newline(out);
+ }
if (!list_empty(&pin_list->flushed)) {
- pr_buf(out, "flushed:");
- pr_newline(out);
+ prt_printf(out, "flushed:");
+ prt_newline(out);
}
list_for_each_entry(pin, &pin_list->flushed, list) {
- pr_buf(out, "\t%px %ps", pin, pin->flush);
- pr_newline(out);
+ prt_printf(out, "\t%px %ps", pin, pin->flush);
+ prt_newline(out);
}
- pr_indent_pop(out, 2);
+ printbuf_indent_sub(out, 2);
--out->atomic;
spin_unlock(&j->lock);