#include "bkey_methods.h"
#include "btree_gc.h"
#include "btree_update.h"
+#include "btree_write_buffer.h"
#include "buckets.h"
#include "error.h"
#include "journal.h"
#include "journal_reclaim.h"
#include "journal_sb.h"
#include "journal_seq_blacklist.h"
-
-#include <trace/events/bcachefs.h>
-
-#define x(n) #n,
-static const char * const bch2_journal_watermarks[] = {
- JOURNAL_WATERMARKS()
- NULL
-};
+#include "trace.h"
static const char * const bch2_journal_errors[] = {
+#define x(n) #n,
JOURNAL_ERRORS()
+#undef x
NULL
};
-#undef x
static inline bool journal_seq_unwritten(struct journal *j, u64 seq)
{
return __journal_entry_is_open(j->reservations);
}
+static void bch2_journal_buf_to_text(struct printbuf *out, struct journal *j, u64 seq)
+{
+ union journal_res_state s = READ_ONCE(j->reservations);
+ unsigned i = seq & JOURNAL_BUF_MASK;
+ struct journal_buf *buf = j->buf + i;
+
+ prt_printf(out, "seq:");
+ prt_tab(out);
+ prt_printf(out, "%llu", seq);
+ prt_newline(out);
+ printbuf_indent_add(out, 2);
+
+ prt_printf(out, "refcount:");
+ prt_tab(out);
+ prt_printf(out, "%u", journal_state_count(s, i));
+ prt_newline(out);
+
+ prt_printf(out, "size:");
+ prt_tab(out);
+ prt_human_readable_u64(out, vstruct_bytes(buf->data));
+ prt_newline(out);
+
+ prt_printf(out, "expires");
+ prt_tab(out);
+ prt_printf(out, "%li jiffies", buf->expires - jiffies);
+ prt_newline(out);
+
+ if (buf->write_done)
+ prt_printf(out, "write done\n");
+ else if (buf->write_allocated)
+ prt_printf(out, "write allocated\n");
+ else if (buf->write_started)
+ prt_printf(out, "write started\n");
+
+ printbuf_indent_sub(out, 2);
+}
+
+static void bch2_journal_bufs_to_text(struct printbuf *out, struct journal *j)
+{
+ if (!out->nr_tabstops)
+ printbuf_tabstop_push(out, 24);
+
+ for (u64 seq = journal_last_unwritten_seq(j);
+ seq <= journal_cur_seq(j);
+ seq++)
+ bch2_journal_buf_to_text(out, j, seq);
+ prt_printf(out, "last buf %s\n", journal_entry_is_open(j) ? "open" : "closed");
+}
+
static inline struct journal_buf *
journal_seq_to_buf(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;
}
-/* journal entry close/open: */
-
-void __bch2_journal_buf_put(struct journal *j)
+/*
+ * 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;
- closure_call(&j->io, bch2_journal_write, c->io_complete_wq, NULL);
+ 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;
+}
+
+void bch2_journal_do_writes(struct journal *j)
+{
+ for (u64 seq = journal_last_unwritten_seq(j);
+ seq <= journal_cur_seq(j);
+ seq++) {
+ unsigned idx = seq & JOURNAL_BUF_MASK;
+ struct journal_buf *w = j->buf + idx;
+
+ if (w->write_started && !w->write_allocated)
+ break;
+ if (w->write_started)
+ continue;
+
+ if (!journal_state_count(j->reservations, idx)) {
+ w->write_started = true;
+ closure_call(&w->io, bch2_journal_write, j->wq, NULL);
+ }
+
+ break;
+ }
+}
+
+/*
+ * Final processing when the last reference of a journal buffer has been
+ * dropped. Drop the pin list reference acquired at journal entry open and write
+ * the buffer, if requested.
+ */
+void bch2_journal_buf_put_final(struct journal *j, u64 seq)
+{
+ lockdep_assert_held(&j->lock);
+
+ if (__bch2_journal_pin_put(j, seq))
+ bch2_journal_reclaim_fast(j);
+ bch2_journal_do_writes(j);
}
/*
* We don't close a journal_buf until the next journal_buf is finished writing,
* and can be opened again - this also initializes the next journal_buf:
*/
-static void __journal_entry_close(struct journal *j, unsigned closed_val)
+static void __journal_entry_close(struct journal *j, unsigned closed_val, bool trace)
{
struct bch_fs *c = container_of(j, struct bch_fs, journal);
struct journal_buf *buf = journal_cur_buf(j);
/* Close out old buffer: */
buf->data->u64s = cpu_to_le32(old.cur_entry_offset);
+ if (trace_journal_entry_close_enabled() && trace) {
+ struct printbuf pbuf = PRINTBUF;
+ pbuf.atomic++;
+
+ prt_str(&pbuf, "entry size: ");
+ prt_human_readable_u64(&pbuf, vstruct_bytes(buf->data));
+ prt_newline(&pbuf);
+ bch2_prt_task_backtrace(&pbuf, current, 1, GFP_NOWAIT);
+ trace_journal_entry_close(c, pbuf.buf);
+ printbuf_exit(&pbuf);
+ }
+
sectors = vstruct_blocks_plus(buf->data, c->block_bits,
buf->u64s_reserved) << c->block_bits;
BUG_ON(sectors > buf->sectors);
buf->data->last_seq = cpu_to_le64(buf->last_seq);
BUG_ON(buf->last_seq > le64_to_cpu(buf->data->seq));
- __bch2_journal_pin_put(j, le64_to_cpu(buf->data->seq));
-
cancel_delayed_work(&j->write_work);
bch2_journal_space_available(j);
- bch2_journal_buf_put(j, old.idx);
+ __bch2_journal_buf_put(j, old.idx, le64_to_cpu(buf->data->seq));
}
void bch2_journal_halt(struct journal *j)
{
spin_lock(&j->lock);
- __journal_entry_close(j, JOURNAL_ENTRY_ERROR_VAL);
+ __journal_entry_close(j, JOURNAL_ENTRY_ERROR_VAL, true);
if (!j->err_seq)
j->err_seq = journal_cur_seq(j);
+ journal_wake(j);
spin_unlock(&j->lock);
}
/* Don't close it yet if we already have a write in flight: */
if (ret)
- __journal_entry_close(j, JOURNAL_ENTRY_CLOSED_VAL);
+ __journal_entry_close(j, JOURNAL_ENTRY_CLOSED_VAL, true);
else if (nr_unwritten_journal_entries(j)) {
struct journal_buf *buf = journal_cur_buf(j);
return ret;
}
-static bool journal_entry_close(struct journal *j)
+bool bch2_journal_entry_close(struct journal *j)
{
bool ret;
/*
* 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)
{
if (!fifo_free(&j->pin))
return JOURNAL_ERR_journal_pin_full;
- if (nr_unwritten_journal_entries(j) == ARRAY_SIZE(j->buf) - 1)
+ 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)
+ if (u64s <= (ssize_t) j->early_journal_entries.nr)
return JOURNAL_ERR_journal_full;
if (fifo_empty(&j->pin) && j->reclaim_thread)
atomic64_inc(&j->seq);
journal_pin_list_init(fifo_push_ref(&j->pin), 1);
+ BUG_ON(j->pin.back - 1 != atomic64_read(&j->seq));
+
BUG_ON(j->buf + (journal_cur_seq(j) & JOURNAL_BUF_MASK) != buf);
bkey_extent_init(&buf->key);
- buf->noflush = false;
- buf->must_flush = false;
- buf->separate_flush = false;
- buf->flush_time = 0;
+ buf->noflush = false;
+ buf->must_flush = false;
+ buf->separate_flush = false;
+ buf->flush_time = 0;
+ buf->need_flush_to_write_buffer = true;
+ buf->write_started = false;
+ buf->write_allocated = false;
+ buf->write_done = false;
memset(buf->data, 0, sizeof(*buf->data));
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);
- if (j->res_get_blocked_start)
- bch2_time_stats_update(j->blocked_time,
- j->res_get_blocked_start);
- j->res_get_blocked_start = 0;
-
- mod_delayed_work(c->io_complete_wq,
- &j->write_work,
- msecs_to_jiffies(c->opts.journal_flush_delay));
+ if (nr_unwritten_journal_entries(j) == 1)
+ mod_delayed_work(j->wq,
+ &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;
}
bool ret = atomic64_read(&j->seq) == j->seq_ondisk;
if (!ret)
- journal_entry_close(j);
+ bch2_journal_entry_close(j);
return ret;
}
static void journal_write_work(struct work_struct *work)
{
struct journal *j = container_of(work, struct journal, write_work.work);
- struct bch_fs *c = container_of(j, struct bch_fs, journal);
- long delta;
spin_lock(&j->lock);
- if (!__journal_entry_is_open(j->reservations))
- goto unlock;
+ if (__journal_entry_is_open(j->reservations)) {
+ long delta = journal_cur_buf(j)->expires - jiffies;
- 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:
+ if (delta > 0)
+ mod_delayed_work(j->wq, &j->write_work, delta);
+ else
+ __journal_entry_close(j, JOURNAL_ENTRY_CLOSED_VAL, true);
+ }
spin_unlock(&j->lock);
}
if (journal_res_get_fast(j, res, flags))
return 0;
+ if ((flags & BCH_WATERMARK_MASK) < j->watermark) {
+ ret = JOURNAL_ERR_journal_full;
+ can_discard = j->can_discard;
+ goto out;
+ }
+
+ if (j->blocked)
+ return -BCH_ERR_journal_res_get_blocked;
+
if (bch2_journal_error(j))
- return -EROFS;
+ return -BCH_ERR_erofs_journal_err;
+
+ if (nr_unwritten_journal_entries(j) == ARRAY_SIZE(j->buf) && !journal_entry_is_open(j)) {
+ ret = JOURNAL_ERR_max_in_flight;
+ goto out;
+ }
spin_lock(&j->lock);
/*
* Recheck after taking the lock, so we don't race with another thread
- * that just did journal_entry_open() and call journal_entry_close()
+ * that just did journal_entry_open() and call bch2_journal_entry_close()
* unnecessarily
*/
if (journal_res_get_fast(j, res, flags)) {
- spin_unlock(&j->lock);
- return 0;
- }
-
- if ((flags & JOURNAL_WATERMARK_MASK) < j->watermark) {
- /*
- * Don't want to close current journal entry, just need to
- * invoke reclaim:
- */
- ret = JOURNAL_ERR_journal_full;
+ ret = 0;
goto unlock;
}
buf->buf_size < JOURNAL_ENTRY_SIZE_MAX)
j->buf_size_want = max(j->buf_size_want, buf->buf_size << 1);
- __journal_entry_close(j, JOURNAL_ENTRY_CLOSED_VAL);
- ret = journal_entry_open(j);
-
- if (ret == JOURNAL_ERR_max_in_flight)
- trace_and_count(c, journal_entry_full, c);
+ __journal_entry_close(j, JOURNAL_ENTRY_CLOSED_VAL, false);
+ ret = journal_entry_open(j) ?: JOURNAL_ERR_retry;
unlock:
- if ((ret && ret != JOURNAL_ERR_insufficient_devices) &&
- !j->res_get_blocked_start) {
- j->res_get_blocked_start = local_clock() ?: 1;
- trace_and_count(c, journal_full, c);
- }
-
can_discard = j->can_discard;
spin_unlock(&j->lock);
-
- if (!ret)
+out:
+ if (ret == JOURNAL_ERR_retry)
goto retry;
+ if (!ret)
+ return 0;
- if ((ret == JOURNAL_ERR_journal_full ||
- ret == JOURNAL_ERR_journal_pin_full) &&
- !can_discard &&
- !nr_unwritten_journal_entries(j) &&
- (flags & JOURNAL_WATERMARK_MASK) == JOURNAL_WATERMARK_reserved) {
- struct printbuf buf = PRINTBUF;
-
- bch_err(c, "Journal stuck! Hava a pre-reservation but journal full (ret %s)",
- bch2_journal_errors[ret]);
-
- bch2_journal_debug_to_text(&buf, j);
- bch_err(c, "%s", buf.buf);
+ if (journal_error_check_stuck(j, ret, flags))
+ ret = -BCH_ERR_journal_res_get_blocked;
- printbuf_reset(&buf);
- bch2_journal_pins_to_text(&buf, j);
- bch_err(c, "Journal pins:\n%s", buf.buf);
+ if (ret == JOURNAL_ERR_max_in_flight &&
+ track_event_change(&c->times[BCH_TIME_blocked_journal_max_in_flight], true)) {
+ struct printbuf buf = PRINTBUF;
+ prt_printf(&buf, "seq %llu\n", journal_cur_seq(j));
+ bch2_journal_bufs_to_text(&buf, j);
+ trace_journal_entry_full(c, buf.buf);
printbuf_exit(&buf);
- bch2_fatal_error(c);
- dump_stack();
+ count_event(c, journal_entry_full);
}
/*
}
}
- return ret == JOURNAL_ERR_insufficient_devices ? -EROFS : -EAGAIN;
+ return ret == JOURNAL_ERR_insufficient_devices
+ ? -BCH_ERR_erofs_journal_err
+ : -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;
}
-/* journal_preres: */
-
-static bool journal_preres_available(struct journal *j,
- struct journal_preres *res,
- unsigned new_u64s,
- unsigned flags)
-{
- bool ret = bch2_journal_preres_get_fast(j, res, new_u64s, flags, true);
-
- if (!ret && mutex_trylock(&j->reclaim_lock)) {
- bch2_journal_reclaim(j);
- mutex_unlock(&j->reclaim_lock);
- }
-
- return ret;
-}
-
-int __bch2_journal_preres_get(struct journal *j,
- struct journal_preres *res,
- unsigned new_u64s,
- unsigned flags)
-{
- int ret;
-
- closure_wait_event(&j->preres_wait,
- (ret = bch2_journal_error(j)) ||
- journal_preres_available(j, res, new_u64s, flags));
- return ret;
-}
-
/* journal_entry_res: */
void bch2_journal_entry_res_resize(struct journal *j,
/*
* Not enough room in current journal entry, have to flush it:
*/
- __journal_entry_close(j, JOURNAL_ENTRY_CLOSED_VAL);
+ __journal_entry_close(j, JOURNAL_ENTRY_CLOSED_VAL, true);
} else {
journal_cur_buf(j)->u64s_reserved += d;
}
/**
* bch2_journal_flush_seq_async - wait for a journal entry to be written
+ * @j: journal object
+ * @seq: seq to flush
+ * @parent: closure object to wait with
+ * Returns: 1 if @seq has already been flushed, 0 if @seq is being flushed,
+ * -EIO if @seq will never be flushed
*
- * like bch2_journal_wait_on_seq, except that it triggers a write immediately if
+ * Like bch2_journal_wait_on_seq, except that it triggers a write immediately if
* necessary
*/
int bch2_journal_flush_seq_async(struct journal *j, u64 seq,
struct journal_res res = { 0 };
if (journal_entry_is_open(j))
- __journal_entry_close(j, JOURNAL_ENTRY_CLOSED_VAL);
+ __journal_entry_close(j, JOURNAL_ENTRY_CLOSED_VAL, true);
spin_unlock(&j->lock);
ret = wait_event_interruptible(j->wait, (ret2 = bch2_journal_flush_seq_async(j, seq, NULL)));
if (!ret)
- bch2_time_stats_update(j->flush_seq_time, start_time);
+ time_stats_update(j->flush_seq_time, start_time);
return ret ?: ret2 < 0 ? ret2 : 0;
}
return bch2_journal_flush_seq(j, res.seq);
}
-int bch2_journal_log_msg(struct journal *j, const char *fmt, ...)
-{
- struct jset_entry_log *entry;
- struct journal_res res = { 0 };
- unsigned msglen, u64s;
- va_list args;
- int ret;
-
- va_start(args, fmt);
- msglen = vsnprintf(NULL, 0, fmt, args) + 1;
- va_end(args);
-
- u64s = jset_u64s(DIV_ROUND_UP(msglen, sizeof(u64)));
-
- ret = bch2_journal_res_get(j, &res, u64s, 0);
- if (ret)
- return ret;
-
- entry = container_of(journal_res_entry(j, &res),
- struct jset_entry_log, entry);;
- memset(entry, 0, u64s * sizeof(u64));
- entry->entry.type = BCH_JSET_ENTRY_log;
- entry->entry.u64s = u64s - 1;
-
- va_start(args, fmt);
- vsnprintf(entry->d, INT_MAX, fmt, args);
- va_end(args);
-
- 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)
journal_quiesce(j);
}
+static struct journal_buf *__bch2_next_write_buffer_flush_journal_buf(struct journal *j, u64 max_seq)
+{
+ struct journal_buf *ret = NULL;
+
+ mutex_lock(&j->buf_lock);
+ spin_lock(&j->lock);
+ max_seq = min(max_seq, journal_cur_seq(j));
+
+ for (u64 seq = journal_last_unwritten_seq(j);
+ seq <= max_seq;
+ seq++) {
+ unsigned idx = seq & JOURNAL_BUF_MASK;
+ struct journal_buf *buf = j->buf + idx;
+
+ if (buf->need_flush_to_write_buffer) {
+ if (seq == journal_cur_seq(j))
+ __journal_entry_close(j, JOURNAL_ENTRY_CLOSED_VAL, true);
+
+ union journal_res_state s;
+ s.v = atomic64_read_acquire(&j->reservations.counter);
+
+ ret = journal_state_count(s, idx)
+ ? ERR_PTR(-EAGAIN)
+ : buf;
+ break;
+ }
+ }
+
+ spin_unlock(&j->lock);
+ if (IS_ERR_OR_NULL(ret))
+ mutex_unlock(&j->buf_lock);
+ return ret;
+}
+
+struct journal_buf *bch2_next_write_buffer_flush_journal_buf(struct journal *j, u64 max_seq)
+{
+ struct journal_buf *ret;
+
+ wait_event(j->wait, (ret = __bch2_next_write_buffer_flush_journal_buf(j, max_seq)) != ERR_PTR(-EAGAIN));
+ return ret;
+}
+
/* allocate journal on a device: */
static int __bch2_set_nr_journal_buckets(struct bch_dev *ca, unsigned nr,
u64 *new_bucket_seq = NULL, *new_buckets = NULL;
struct open_bucket **ob = NULL;
long *bu = NULL;
- unsigned i, nr_got = 0, nr_want = nr - ja->nr;
- unsigned old_nr = ja->nr;
- unsigned old_discard_idx = ja->discard_idx;
- unsigned old_dirty_idx_ondisk = ja->dirty_idx_ondisk;
- unsigned old_dirty_idx = ja->dirty_idx;
- unsigned old_cur_idx = ja->cur_idx;
+ unsigned i, pos, nr_got = 0, nr_want = nr - ja->nr;
int ret = 0;
- if (c) {
- bch2_journal_flush_all_pins(&c->journal);
- bch2_journal_block(&c->journal);
- }
+ BUG_ON(nr <= ja->nr);
- bu = kzalloc(nr_want * sizeof(*bu), GFP_KERNEL);
- ob = kzalloc(nr_want * sizeof(*ob), GFP_KERNEL);
- new_buckets = kzalloc(nr * sizeof(u64), GFP_KERNEL);
- new_bucket_seq = kzalloc(nr * sizeof(u64), GFP_KERNEL);
+ 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 = -ENOMEM;
- goto err_unblock;
+ ret = -BCH_ERR_ENOMEM_set_nr_journal_buckets;
+ goto err_free;
}
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 = -ENOSPC;
+ ret = -BCH_ERR_ENOSPC_bucket_alloc;
break;
}
} else {
- ob[nr_got] = bch2_bucket_alloc(c, ca, RESERVE_none,
- false, cl);
- if (IS_ERR(ob[nr_got])) {
- ret = cl ? -EAGAIN : -ENOSPC;
+ 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_msg(c, ret, "marking new journal buckets");
break;
}
}
if (!nr_got)
- goto err_unblock;
+ goto err_free;
- /*
- * 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);
+ /* 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);
+
+ BUG_ON(ja->discard_idx > ja->nr);
+
+ pos = ja->discard_idx ?: ja->nr;
+
+ 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));
for (i = 0; i < nr_got; i++) {
- unsigned pos = ja->discard_idx ?: ja->nr;
- long b = bu[i];
-
- __array_insert_item(ja->buckets, ja->nr, pos);
- __array_insert_item(ja->bucket_seq, ja->nr, pos);
- ja->nr++;
-
- ja->buckets[pos] = b;
- ja->bucket_seq[pos] = 0;
-
- 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;
+ new_buckets[pos + i] = bu[i];
+ new_bucket_seq[pos + i] = 0;
}
- ret = bch2_journal_buckets_to_sb(c, ca);
- if (ret) {
- /* Revert: */
- swap(new_buckets, ja->buckets);
- swap(new_bucket_seq, ja->bucket_seq);
- ja->nr = old_nr;
- ja->discard_idx = old_discard_idx;
- ja->dirty_idx_ondisk = old_dirty_idx_ondisk;
- ja->dirty_idx = old_dirty_idx;
- ja->cur_idx = old_cur_idx;
- }
+ nr = ja->nr + nr_got;
+
+ ret = bch2_journal_buckets_to_sb(c, ca, new_buckets, nr);
+ if (ret)
+ goto err_unblock;
if (!new_fs)
- spin_unlock(&c->journal.lock);
+ bch2_write_super(c);
+ /* Commit: */
if (c)
- bch2_journal_unblock(&c->journal);
+ 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 (!new_fs) {
- for (i = 0; i < nr_got; i++) {
- ret = bch2_trans_run(c,
- bch2_trans_mark_metadata_bucket(&trans, ca,
- bu[i], BCH_DATA_journal,
- ca->mi.bucket_size));
- if (ret) {
- bch2_fs_inconsistent(c, "error marking new journal buckets: %i", ret);
- goto err;
- }
- }
+ 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:
- if (ob && !new_fs)
+
+ 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_buckets);
kfree(ob);
kfree(bu);
-
return ret;
-err_unblock:
- if (c)
- bch2_journal_unblock(&c->journal);
- goto err;
}
/*
{
struct journal_device *ja = &ca->journal;
struct closure cl;
- unsigned current_nr;
int ret = 0;
- /* don't handle reducing nr of buckets yet: */
- if (nr < ja->nr)
- return 0;
-
closure_init_stack(&cl);
- while (ja->nr != nr && (ret == 0 || ret == -EAGAIN)) {
- 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, 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);
+ closure_sync(&cl);
+
+ if (ret && ret != -BCH_ERR_bucket_alloc_blocked)
+ break;
}
+ bch_err_fn(c, ret);
+unlock:
+ up_write(&c->state_lock);
return ret;
}
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));
- if (ca->fs)
- mutex_lock(&ca->fs->sb_lock);
-
ret = __bch2_set_nr_journal_buckets(ca, nr, true, NULL);
+err:
+ bch_err_fn(ca, ret);
+ return ret;
+}
- if (ca->fs)
- mutex_unlock(&ca->fs->sb_lock);
+int bch2_fs_journal_alloc(struct bch_fs *c)
+{
+ for_each_online_member(c, ca) {
+ if (ca->journal.nr)
+ continue;
- return ret;
+ int ret = bch2_dev_journal_alloc(ca);
+ if (ret) {
+ percpu_ref_put(&ca->io_ref);
+ return ret;
+ }
+ }
+
+ return 0;
}
/* 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);
bch2_journal_reclaim_stop(j);
bch2_journal_flush_all_pins(j);
- wait_event(j->wait, journal_entry_close(j));
+ wait_event(j->wait, bch2_journal_entry_close(j));
/*
* Always write a new journal entry, to make sure the clock hands are up
struct journal_replay *i, **_i;
struct genradix_iter iter;
bool had_entries = false;
- unsigned ptr;
u64 last_seq = cur_seq, nr, seq;
genradix_for_each_reverse(&c->journal_entries, iter, _i) {
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;
}
}
p = journal_seq_pin(j, seq);
p->devs.nr = 0;
- for (ptr = 0; ptr < i->nr_ptrs; ptr++)
- bch2_dev_list_add_dev(&p->devs, i->ptrs[ptr].dev);
+ darray_for_each(i->ptrs, ptr)
+ bch2_dev_list_add_dev(&p->devs, ptr->dev);
had_entries = true;
}
void bch2_dev_journal_exit(struct bch_dev *ca)
{
- kfree(ca->journal.bio);
- kfree(ca->journal.buckets);
- kfree(ca->journal.bucket_seq);
+ struct journal_device *ja = &ca->journal;
- ca->journal.bio = NULL;
- ca->journal.buckets = NULL;
- ca->journal.bucket_seq = NULL;
+ for (unsigned i = 0; i < ARRAY_SIZE(ja->bio); i++) {
+ kfree(ja->bio[i]);
+ ja->bio[i] = NULL;
+ }
+
+ kfree(ja->buckets);
+ kfree(ja->bucket_seq);
+ ja->buckets = NULL;
+ ja->bucket_seq = NULL;
}
int bch2_dev_journal_init(struct bch_dev *ca, struct bch_sb *sb)
{
struct journal_device *ja = &ca->journal;
struct bch_sb_field_journal *journal_buckets =
- bch2_sb_get_journal(sb);
+ bch2_sb_field_get(sb, journal);
struct bch_sb_field_journal_v2 *journal_buckets_v2 =
- bch2_sb_get_journal_v2(sb);
- unsigned i, nr_bvecs;
+ bch2_sb_field_get(sb, journal_v2);
ja->nr = 0;
if (journal_buckets_v2) {
unsigned nr = bch2_sb_field_journal_v2_nr_entries(journal_buckets_v2);
- for (i = 0; i < nr; i++)
+ for (unsigned 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);
+ unsigned nr_bvecs = DIV_ROUND_UP(JOURNAL_ENTRY_SIZE_MAX, PAGE_SIZE);
- ca->journal.bio = bio_kmalloc(nr_bvecs, GFP_KERNEL);
- if (!ca->journal.bio)
- return -ENOMEM;
+ for (unsigned i = 0; i < ARRAY_SIZE(ja->bio); i++) {
+ ja->bio[i] = kmalloc(struct_size(ja->bio[i], bio.bi_inline_vecs,
+ nr_bvecs), GFP_KERNEL);
+ if (!ja->bio[i])
+ return -BCH_ERR_ENOMEM_dev_journal_init;
- bio_init(ca->journal.bio, NULL, ca->journal.bio->bi_inline_vecs, nr_bvecs, 0);
+ ja->bio[i]->ca = ca;
+ ja->bio[i]->buf_idx = i;
+ bio_init(&ja->bio[i]->bio, NULL, ja->bio[i]->bio.bi_inline_vecs, nr_bvecs, 0);
+ }
ja->buckets = kcalloc(ja->nr, sizeof(u64), GFP_KERNEL);
if (!ja->buckets)
- return -ENOMEM;
+ 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;
+ unsigned dst = 0;
- for (i = 0; i < nr; i++)
- for (j = 0; j < le64_to_cpu(journal_buckets_v2->d[i].nr); j++)
+ for (unsigned i = 0; i < nr; i++)
+ for (unsigned 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++)
+ for (unsigned i = 0; i < ja->nr; i++)
ja->buckets[i] = le64_to_cpu(journal_buckets->buckets[i]);
}
void bch2_fs_journal_exit(struct journal *j)
{
- unsigned i;
+ if (j->wq)
+ destroy_workqueue(j->wq);
- for (i = 0; i < ARRAY_SIZE(j->buf); i++)
- kvpfree(j->buf[i].data, j->buf[i].buf_size);
+ darray_exit(&j->early_journal_entries);
+
+ for (unsigned i = 0; i < ARRAY_SIZE(j->buf); i++)
+ kvfree(j->buf[i].data);
free_fifo(&j->pin);
}
int bch2_fs_journal_init(struct journal *j)
{
- struct bch_fs *c = container_of(j, struct bch_fs, journal);
static struct lock_class_key res_key;
- unsigned i;
- int ret = 0;
-
- pr_verbose_init(c->opts, "");
+ mutex_init(&j->buf_lock);
spin_lock_init(&j->lock);
spin_lock_init(&j->err_lock);
init_waitqueue_head(&j->wait);
((union journal_res_state)
{ .cur_entry_offset = JOURNAL_ENTRY_CLOSED_VAL }).v);
- if (!(init_fifo(&j->pin, JOURNAL_PIN, GFP_KERNEL))) {
- ret = -ENOMEM;
- goto out;
- }
+ if (!(init_fifo(&j->pin, JOURNAL_PIN, GFP_KERNEL)))
+ return -BCH_ERR_ENOMEM_journal_pin_fifo;
- for (i = 0; i < ARRAY_SIZE(j->buf); i++) {
+ for (unsigned i = 0; i < ARRAY_SIZE(j->buf); i++) {
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;
- goto out;
- }
+ j->buf[i].data = kvmalloc(j->buf[i].buf_size, GFP_KERNEL);
+ if (!j->buf[i].data)
+ return -BCH_ERR_ENOMEM_journal_buf;
+ j->buf[i].idx = i;
}
j->pin.front = j->pin.back = 1;
-out:
- pr_verbose_init(c->opts, "ret %i", ret);
- return ret;
+
+ j->wq = alloc_workqueue("bcachefs_journal",
+ WQ_HIGHPRI|WQ_FREEZABLE|WQ_UNBOUND|WQ_MEM_RECLAIM, 512);
+ if (!j->wq)
+ return -BCH_ERR_ENOMEM_fs_other_alloc;
+ return 0;
}
/* debug: */
{
struct bch_fs *c = container_of(j, struct bch_fs, journal);
union journal_res_state s;
- struct bch_dev *ca;
unsigned long now = jiffies;
- u64 seq;
- unsigned i;
+ u64 nr_writes = j->nr_flush_writes + j->nr_noflush_writes;
if (!out->nr_tabstops)
printbuf_tabstop_push(out, 24);
rcu_read_lock();
s = READ_ONCE(j->reservations);
- prt_printf(out, "dirty journal entries:\t%llu/%llu\n",fifo_used(&j->pin), j->pin.size);
+ 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:\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_journal_watermarks[j->watermark]);
- prt_printf(out, "each entry reserved:\t%u\n", j->entry_u64s_reserved);
+ prt_printf(out, "flushed_seq_ondisk:\t%llu\n", j->flushed_seq_ondisk);
+ 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 noflush writes:\t%llu\n", j->nr_noflush_writes);
+ prt_printf(out, "average write size:\t");
+ prt_human_readable_u64(out, nr_writes ? div64_u64(j->entry_bytes_written, nr_writes) : 0);
+ prt_newline(out);
+ 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)
+ prt_printf(out, "reclaim runs in:\t%u ms\n", time_after(j->next_reclaim, now)
? jiffies_to_msecs(j->next_reclaim - jiffies) : 0);
- 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 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) {
}
prt_newline(out);
-
- for (seq = journal_cur_seq(j);
- seq >= journal_last_unwritten_seq(j);
- --seq) {
- i = seq & JOURNAL_BUF_MASK;
-
- prt_printf(out, "unwritten entry:");
- prt_tab(out);
- prt_printf(out, "%llu", seq);
- prt_newline(out);
- printbuf_indent_add(out, 2);
-
- prt_printf(out, "refcount:");
- prt_tab(out);
- prt_printf(out, "%u", journal_state_count(s, i));
- prt_newline(out);
-
- prt_printf(out, "sectors:");
- prt_tab(out);
- prt_printf(out, "%u", j->buf[i].sectors);
- prt_newline(out);
-
- prt_printf(out, "expires");
- prt_tab(out);
- prt_printf(out, "%li jiffies", j->buf[i].expires - jiffies);
- prt_newline(out);
-
- printbuf_indent_sub(out, 2);
- }
+ prt_printf(out, "unwritten entries:");
+ prt_newline(out);
+ bch2_journal_bufs_to_text(out, j);
prt_printf(out,
"replay done:\t\t%i\n",
j->space[journal_space_total].next_entry,
j->space[journal_space_total].total);
- for_each_member_device_rcu(ca, c, i,
- &c->rw_devs[BCH_DATA_journal]) {
+ for_each_member_device_rcu(c, ca, &c->rw_devs[BCH_DATA_journal]) {
struct journal_device *ja = &ca->journal;
if (!test_bit(ca->dev_idx, c->rw_devs[BCH_DATA_journal].d))
if (!ja->nr)
continue;
- prt_printf(out, "dev %u:\n", i);
+ prt_printf(out, "dev %u:\n", ca->dev_idx);
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_newline(out);
printbuf_indent_add(out, 2);
- list_for_each_entry(pin, &pin_list->list, list) {
- prt_printf(out, "\t%px %ps", pin, pin->flush);
- prt_newline(out);
- }
-
- list_for_each_entry(pin, &pin_list->key_cache_list, list) {
- prt_printf(out, "\t%px %ps", pin, pin->flush);
- prt_newline(out);
- }
+ for (unsigned 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)) {
prt_printf(out, "flushed:");