* [PATCH v7 01/10] md/r5cache: Check array size in r5l_init_log
From: Song Liu @ 2016-11-17 23:24 UTC (permalink / raw)
To: linux-raid
Cc: neilb, shli, kernel-team, dan.j.williams, hch, liuzhengyuang521,
liuzhengyuan, Song Liu
In-Reply-To: <20161117232445.1798305-1-songliubraving@fb.com>
Currently, r5l_write_stripe checks meta size for each stripe write,
which is not necessary.
With this patch, r5l_init_log checks maximal meta size of the array,
which is (r5l_meta_block + raid_disks x r5l_payload_data_parity).
If this is too big to fit in one page, r5l_init_log aborts.
With current meta data, r5l_log support raid_disks up to 203.
Signed-off-by: Song Liu <songliubraving@fb.com>
---
drivers/md/raid5-cache.c | 26 ++++++++++++++++----------
1 file changed, 16 insertions(+), 10 deletions(-)
diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c
index f73672b..33fc850 100644
--- a/drivers/md/raid5-cache.c
+++ b/drivers/md/raid5-cache.c
@@ -441,7 +441,6 @@ int r5l_write_stripe(struct r5l_log *log, struct stripe_head *sh)
{
int write_disks = 0;
int data_pages, parity_pages;
- int meta_size;
int reserve;
int i;
int ret = 0;
@@ -473,15 +472,6 @@ int r5l_write_stripe(struct r5l_log *log, struct stripe_head *sh)
parity_pages = 1 + !!(sh->qd_idx >= 0);
data_pages = write_disks - parity_pages;
- meta_size =
- ((sizeof(struct r5l_payload_data_parity) + sizeof(__le32))
- * data_pages) +
- sizeof(struct r5l_payload_data_parity) +
- sizeof(__le32) * parity_pages;
- /* Doesn't work with very big raid array */
- if (meta_size + sizeof(struct r5l_meta_block) > PAGE_SIZE)
- return -EINVAL;
-
set_bit(STRIPE_LOG_TRAPPED, &sh->state);
/*
* The stripe must enter state machine again to finish the write, so
@@ -1197,6 +1187,22 @@ int r5l_init_log(struct r5conf *conf, struct md_rdev *rdev)
if (PAGE_SIZE != 4096)
return -EINVAL;
+
+ /*
+ * The PAGE_SIZE must be big enough to hold 1 r5l_meta_block and
+ * raid_disks r5l_payload_data_parity.
+ *
+ * Write journal and cache does not work for very big array
+ * (raid_disks > 203)
+ */
+ if (sizeof(struct r5l_meta_block) +
+ ((sizeof(struct r5l_payload_data_parity) + sizeof(__le32)) *
+ conf->raid_disks) > PAGE_SIZE) {
+ pr_err("md/raid:%s: write journal/cache doesn't work for array with %d disks\n",
+ mdname(conf->mddev), conf->raid_disks);
+ return -EINVAL;
+ }
+
log = kzalloc(sizeof(*log), GFP_KERNEL);
if (!log)
return -ENOMEM;
--
2.9.3
^ permalink raw reply related
* [PATCH v7 02/10] md/r5cache: move some code to raid5.h
From: Song Liu @ 2016-11-17 23:24 UTC (permalink / raw)
To: linux-raid
Cc: neilb, shli, kernel-team, dan.j.williams, hch, liuzhengyuang521,
liuzhengyuan, Song Liu
In-Reply-To: <20161117232445.1798305-1-songliubraving@fb.com>
Move some define and inline functions to raid5.h, so they can be
used in raid5-cache.c
Signed-off-by: Song Liu <songliubraving@fb.com>
---
drivers/md/raid5.c | 71 -------------------------------------------------
drivers/md/raid5.h | 77 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 77 insertions(+), 71 deletions(-)
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index df88656..34895f3 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -70,19 +70,6 @@ module_param(devices_handle_discard_safely, bool, 0644);
MODULE_PARM_DESC(devices_handle_discard_safely,
"Set to Y if all devices in each array reliably return zeroes on reads from discarded regions");
static struct workqueue_struct *raid5_wq;
-/*
- * Stripe cache
- */
-
-#define NR_STRIPES 256
-#define STRIPE_SIZE PAGE_SIZE
-#define STRIPE_SHIFT (PAGE_SHIFT - 9)
-#define STRIPE_SECTORS (STRIPE_SIZE>>9)
-#define IO_THRESHOLD 1
-#define BYPASS_THRESHOLD 1
-#define NR_HASH (PAGE_SIZE / sizeof(struct hlist_head))
-#define HASH_MASK (NR_HASH - 1)
-#define MAX_STRIPE_BATCH 8
static inline struct hlist_head *stripe_hash(struct r5conf *conf, sector_t sect)
{
@@ -126,64 +113,6 @@ static inline void unlock_all_device_hash_locks_irq(struct r5conf *conf)
local_irq_enable();
}
-/* bio's attached to a stripe+device for I/O are linked together in bi_sector
- * order without overlap. There may be several bio's per stripe+device, and
- * a bio could span several devices.
- * When walking this list for a particular stripe+device, we must never proceed
- * beyond a bio that extends past this device, as the next bio might no longer
- * be valid.
- * This function is used to determine the 'next' bio in the list, given the sector
- * of the current stripe+device
- */
-static inline struct bio *r5_next_bio(struct bio *bio, sector_t sector)
-{
- int sectors = bio_sectors(bio);
- if (bio->bi_iter.bi_sector + sectors < sector + STRIPE_SECTORS)
- return bio->bi_next;
- else
- return NULL;
-}
-
-/*
- * We maintain a biased count of active stripes in the bottom 16 bits of
- * bi_phys_segments, and a count of processed stripes in the upper 16 bits
- */
-static inline int raid5_bi_processed_stripes(struct bio *bio)
-{
- atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
- return (atomic_read(segments) >> 16) & 0xffff;
-}
-
-static inline int raid5_dec_bi_active_stripes(struct bio *bio)
-{
- atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
- return atomic_sub_return(1, segments) & 0xffff;
-}
-
-static inline void raid5_inc_bi_active_stripes(struct bio *bio)
-{
- atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
- atomic_inc(segments);
-}
-
-static inline void raid5_set_bi_processed_stripes(struct bio *bio,
- unsigned int cnt)
-{
- atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
- int old, new;
-
- do {
- old = atomic_read(segments);
- new = (old & 0xffff) | (cnt << 16);
- } while (atomic_cmpxchg(segments, old, new) != old);
-}
-
-static inline void raid5_set_bi_stripes(struct bio *bio, unsigned int cnt)
-{
- atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
- atomic_set(segments, cnt);
-}
-
/* Find first data disk in a raid6 stripe */
static inline int raid6_d0(struct stripe_head *sh)
{
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
index 57ec49f..ffc13c4 100644
--- a/drivers/md/raid5.h
+++ b/drivers/md/raid5.h
@@ -410,6 +410,83 @@ struct disk_info {
struct md_rdev *rdev, *replacement;
};
+/*
+ * Stripe cache
+ */
+
+#define NR_STRIPES 256
+#define STRIPE_SIZE PAGE_SIZE
+#define STRIPE_SHIFT (PAGE_SHIFT - 9)
+#define STRIPE_SECTORS (STRIPE_SIZE>>9)
+#define IO_THRESHOLD 1
+#define BYPASS_THRESHOLD 1
+#define NR_HASH (PAGE_SIZE / sizeof(struct hlist_head))
+#define HASH_MASK (NR_HASH - 1)
+#define MAX_STRIPE_BATCH 8
+
+/* bio's attached to a stripe+device for I/O are linked together in bi_sector
+ * order without overlap. There may be several bio's per stripe+device, and
+ * a bio could span several devices.
+ * When walking this list for a particular stripe+device, we must never proceed
+ * beyond a bio that extends past this device, as the next bio might no longer
+ * be valid.
+ * This function is used to determine the 'next' bio in the list, given the
+ * sector of the current stripe+device
+ */
+static inline struct bio *r5_next_bio(struct bio *bio, sector_t sector)
+{
+ int sectors = bio_sectors(bio);
+
+ if (bio->bi_iter.bi_sector + sectors < sector + STRIPE_SECTORS)
+ return bio->bi_next;
+ else
+ return NULL;
+}
+
+/*
+ * We maintain a biased count of active stripes in the bottom 16 bits of
+ * bi_phys_segments, and a count of processed stripes in the upper 16 bits
+ */
+static inline int raid5_bi_processed_stripes(struct bio *bio)
+{
+ atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+
+ return (atomic_read(segments) >> 16) & 0xffff;
+}
+
+static inline int raid5_dec_bi_active_stripes(struct bio *bio)
+{
+ atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+
+ return atomic_sub_return(1, segments) & 0xffff;
+}
+
+static inline void raid5_inc_bi_active_stripes(struct bio *bio)
+{
+ atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+
+ atomic_inc(segments);
+}
+
+static inline void raid5_set_bi_processed_stripes(struct bio *bio,
+ unsigned int cnt)
+{
+ atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+ int old, new;
+
+ do {
+ old = atomic_read(segments);
+ new = (old & 0xffff) | (cnt << 16);
+ } while (atomic_cmpxchg(segments, old, new) != old);
+}
+
+static inline void raid5_set_bi_stripes(struct bio *bio, unsigned int cnt)
+{
+ atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+
+ atomic_set(segments, cnt);
+}
+
/* NOTE NR_STRIPE_HASH_LOCKS must remain below 64.
* This is because we sometimes take all the spinlocks
* and creating that much locking depth can cause
--
2.9.3
^ permalink raw reply related
* [PATCH v7 03/10] md/r5cache: State machine for raid5-cache write back mode
From: Song Liu @ 2016-11-17 23:24 UTC (permalink / raw)
To: linux-raid
Cc: neilb, shli, kernel-team, dan.j.williams, hch, liuzhengyuang521,
liuzhengyuan, Song Liu
In-Reply-To: <20161117232445.1798305-1-songliubraving@fb.com>
This patch adds state machine for raid5-cache. With log device, the
raid456 array could operate in two different modes (r5c_journal_mode):
- write-back (R5C_MODE_WRITE_BACK)
- write-through (R5C_MODE_WRITE_THROUGH)
Existing code of raid5-cache only has write-through mode. For write-back
cache, it is necessary to extend the state machine.
With write-back cache, every stripe could operate in two different
phases:
- caching
- writing-out
In caching phase, the stripe handles writes as:
- write to journal
- return IO
In writing-out phase, the stripe behaviors as a stripe in write through
mode R5C_MODE_WRITE_THROUGH.
STRIPE_R5C_CACHING is added to sh->state to differentiate caching and
writing-out phase.
Please note: this is a "no-op" patch for raid5-cache write-through
mode.
The following detailed explanation is copied from the raid5-cache.c:
/*
* raid5 cache state machine
*
* With rhe RAID cache, each stripe works in two phases:
* - caching phase
* - writing-out phase
*
* These two phases are controlled by bit STRIPE_R5C_CACHING:
* if STRIPE_R5C_CACHING == 0, the stripe is in writing-out phase
* if STRIPE_R5C_CACHING == 1, the stripe is in caching phase
*
* When there is no journal, or the journal is in write-through mode,
* the stripe is always in writing-out phase.
*
* For write-back journal, the stripe is sent to caching phase on write
* (r5c_handle_stripe_dirtying). r5c_make_stripe_write_out() kicks off
* the write-out phase by clearing STRIPE_R5C_CACHING.
*
* Stripes in caching phase do not write the raid disks. Instead, all
* writes are committed from the log device. Therefore, a stripe in
* caching phase handles writes as:
* - write to log device
* - return IO
*
* Stripes in writing-out phase handle writes as:
* - calculate parity
* - write pending data and parity to journal
* - write data and parity to raid disks
* - return IO for pending writes
*/
Signed-off-by: Song Liu <songliubraving@fb.com>
---
drivers/md/raid5-cache.c | 143 ++++++++++++++++++++++++++++++++++++++++++++++-
drivers/md/raid5.c | 45 +++++++++++++--
drivers/md/raid5.h | 31 +++++++++-
3 files changed, 211 insertions(+), 8 deletions(-)
diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c
index 33fc850..02a5544 100644
--- a/drivers/md/raid5-cache.c
+++ b/drivers/md/raid5-cache.c
@@ -40,6 +40,47 @@
*/
#define R5L_POOL_SIZE 4
+/*
+ * r5c journal modes of the array: write-back or write-through.
+ * write-through mode has identical behavior as existing log only
+ * implementation.
+ */
+enum r5c_journal_mode {
+ R5C_JOURNAL_MODE_WRITE_THROUGH = 0,
+ R5C_JOURNAL_MODE_WRITE_BACK = 1,
+};
+
+/*
+ * raid5 cache state machine
+ *
+ * With rhe RAID cache, each stripe works in two phases:
+ * - caching phase
+ * - writing-out phase
+ *
+ * These two phases are controlled by bit STRIPE_R5C_CACHING:
+ * if STRIPE_R5C_CACHING == 0, the stripe is in writing-out phase
+ * if STRIPE_R5C_CACHING == 1, the stripe is in caching phase
+ *
+ * When there is no journal, or the journal is in write-through mode,
+ * the stripe is always in writing-out phase.
+ *
+ * For write-back journal, the stripe is sent to caching phase on write
+ * (r5c_try_caching_write). r5c_make_stripe_write_out() kicks off
+ * the write-out phase by clearing STRIPE_R5C_CACHING.
+ *
+ * Stripes in caching phase do not write the raid disks. Instead, all
+ * writes are committed from the log device. Therefore, a stripe in
+ * caching phase handles writes as:
+ * - write to log device
+ * - return IO
+ *
+ * Stripes in writing-out phase handle writes as:
+ * - calculate parity
+ * - write pending data and parity to journal
+ * - write data and parity to raid disks
+ * - return IO for pending writes
+ */
+
struct r5l_log {
struct md_rdev *rdev;
@@ -96,6 +137,9 @@ struct r5l_log {
spinlock_t no_space_stripes_lock;
bool need_cache_flush;
+
+ /* for r5c_cache */
+ enum r5c_journal_mode r5c_journal_mode;
};
/*
@@ -133,6 +177,12 @@ enum r5l_io_unit_state {
IO_UNIT_STRIPE_END = 3, /* stripes data finished writing to raid */
};
+bool r5c_is_writeback(struct r5l_log *log)
+{
+ return (log != NULL &&
+ log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_BACK);
+}
+
static sector_t r5l_ring_add(struct r5l_log *log, sector_t start, sector_t inc)
{
start += inc;
@@ -168,12 +218,51 @@ static void __r5l_set_io_unit_state(struct r5l_io_unit *io,
io->state = state;
}
+/*
+ * Put the stripe into writing-out phase by clearing STRIPE_R5C_CACHING.
+ * This function should only be called in write-back mode.
+ */
+static void r5c_make_stripe_write_out(struct stripe_head *sh)
+{
+ struct r5conf *conf = sh->raid_conf;
+ struct r5l_log *log = conf->log;
+
+ BUG_ON(!r5c_is_writeback(log));
+
+ WARN_ON(!test_bit(STRIPE_R5C_CACHING, &sh->state));
+ clear_bit(STRIPE_R5C_CACHING, &sh->state);
+}
+
+/*
+ * Setting proper flags after writing (or flushing) data and/or parity to the
+ * log device. This is called from r5l_log_endio() or r5l_log_flush_endio().
+ */
+static void r5c_finish_cache_stripe(struct stripe_head *sh)
+{
+ struct r5l_log *log = sh->raid_conf->log;
+
+ if (log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH) {
+ BUG_ON(test_bit(STRIPE_R5C_CACHING, &sh->state));
+ /*
+ * Set R5_InJournal for parity dev[pd_idx]. This means
+ * all data AND parity in the journal. For RAID 6, it is
+ * NOT necessary to set the flag for dev[qd_idx], as the
+ * two parities are written out together.
+ */
+ set_bit(R5_InJournal, &sh->dev[sh->pd_idx].flags);
+ } else
+ BUG(); /* write-back logic in next patch */
+}
+
static void r5l_io_run_stripes(struct r5l_io_unit *io)
{
struct stripe_head *sh, *next;
list_for_each_entry_safe(sh, next, &io->stripe_list, log_list) {
list_del_init(&sh->log_list);
+
+ r5c_finish_cache_stripe(sh);
+
set_bit(STRIPE_HANDLE, &sh->state);
raid5_release_stripe(sh);
}
@@ -412,18 +501,19 @@ static int r5l_log_stripe(struct r5l_log *log, struct stripe_head *sh,
r5l_append_payload_page(log, sh->dev[i].page);
}
- if (sh->qd_idx >= 0) {
+ if (parity_pages == 2) {
r5l_append_payload_meta(log, R5LOG_PAYLOAD_PARITY,
sh->sector, sh->dev[sh->pd_idx].log_checksum,
sh->dev[sh->qd_idx].log_checksum, true);
r5l_append_payload_page(log, sh->dev[sh->pd_idx].page);
r5l_append_payload_page(log, sh->dev[sh->qd_idx].page);
- } else {
+ } else if (parity_pages == 1) {
r5l_append_payload_meta(log, R5LOG_PAYLOAD_PARITY,
sh->sector, sh->dev[sh->pd_idx].log_checksum,
0, false);
r5l_append_payload_page(log, sh->dev[sh->pd_idx].page);
- }
+ } else /* Just writing data, not parity, in caching phase */
+ BUG_ON(parity_pages != 0);
list_add_tail(&sh->log_list, &io->stripe_list);
atomic_inc(&io->pending_stripe);
@@ -455,6 +545,8 @@ int r5l_write_stripe(struct r5l_log *log, struct stripe_head *sh)
return -EAGAIN;
}
+ WARN_ON(test_bit(STRIPE_R5C_CACHING, &sh->state));
+
for (i = 0; i < sh->disks; i++) {
void *addr;
@@ -1112,6 +1204,49 @@ static void r5l_write_super(struct r5l_log *log, sector_t cp)
set_bit(MD_CHANGE_DEVS, &mddev->flags);
}
+/*
+ * Try handle write operation in caching phase. This function should only
+ * be called in write-back mode.
+ *
+ * If all outstanding writes can be handled in caching phase, returns 0
+ * If writes requires write-out phase, call r5c_make_stripe_write_out()
+ * and returns -EAGAIN
+ */
+int r5c_try_caching_write(struct r5conf *conf,
+ struct stripe_head *sh,
+ struct stripe_head_state *s,
+ int disks)
+{
+ struct r5l_log *log = conf->log;
+
+ BUG_ON(!r5c_is_writeback(log));
+
+ /* more write-back logic in next patches */
+ r5c_make_stripe_write_out(sh);
+ return -EAGAIN;
+}
+
+/*
+ * clean up the stripe (clear R5_InJournal for dev[pd_idx] etc.) after the
+ * stripe is committed to RAID disks.
+ */
+void r5c_finish_stripe_write_out(struct r5conf *conf,
+ struct stripe_head *sh,
+ struct stripe_head_state *s)
+{
+ if (!conf->log ||
+ !test_bit(R5_InJournal, &sh->dev[sh->pd_idx].flags))
+ return;
+
+ WARN_ON(test_bit(STRIPE_R5C_CACHING, &sh->state));
+ clear_bit(R5_InJournal, &sh->dev[sh->pd_idx].flags);
+
+ if (conf->log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH)
+ return;
+ BUG(); /* write-back logic in following patches */
+}
+
+
static int r5l_load_log(struct r5l_log *log)
{
struct md_rdev *rdev = log->rdev;
@@ -1249,6 +1384,8 @@ int r5l_init_log(struct r5conf *conf, struct md_rdev *rdev)
INIT_LIST_HEAD(&log->no_space_stripes);
spin_lock_init(&log->no_space_stripes_lock);
+ log->r5c_journal_mode = R5C_JOURNAL_MODE_WRITE_THROUGH;
+
if (r5l_load_log(log))
goto error;
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 34895f3..7c98eb0 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -4107,6 +4107,9 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
if (rdev && !test_bit(Faulty, &rdev->flags))
do_recovery = 1;
}
+
+ if (test_bit(R5_InJournal, &dev->flags))
+ s->injournal++;
}
if (test_bit(STRIPE_SYNCING, &sh->state)) {
/* If there is a failed device being replaced,
@@ -4386,14 +4389,47 @@ static void handle_stripe(struct stripe_head *sh)
|| s.expanding)
handle_stripe_fill(sh, &s, disks);
- /* Now to consider new write requests and what else, if anything
- * should be read. We do not handle new writes when:
+ /*
+ * When the stripe finishes full journal write cycle (write to journal
+ * and raid disk), this is the clean up procedure so it is ready for
+ * next operation.
+ */
+ r5c_finish_stripe_write_out(conf, sh, &s);
+
+ /*
+ * Now to consider new write requests, cache write back and what else,
+ * if anything should be read. We do not handle new writes when:
* 1/ A 'write' operation (copy+xor) is already in flight.
* 2/ A 'check' operation is in flight, as it may clobber the parity
* block.
+ * 3/ A r5c cache log write is in flight.
*/
- if (s.to_write && !sh->reconstruct_state && !sh->check_state)
- handle_stripe_dirtying(conf, sh, &s, disks);
+
+ if (!sh->reconstruct_state && !sh->check_state && !sh->log_io) {
+ if (!r5c_is_writeback(conf->log)) {
+ if (s.to_write)
+ handle_stripe_dirtying(conf, sh, &s, disks);
+ } else { /* write back cache */
+ int ret = 0;
+
+ /* First, try handle writes in caching phase */
+ if (s.to_write)
+ ret = r5c_try_caching_write(conf, sh, &s,
+ disks);
+ /*
+ * If caching phase failed: ret == -EAGAIN
+ * OR
+ * stripe under reclaim: !caching && injournal
+ *
+ * fall back to handle_stripe_dirtying()
+ */
+ if (ret == -EAGAIN ||
+ /* stripe under reclaim: !caching && injournal */
+ (!test_bit(STRIPE_R5C_CACHING, &sh->state) &&
+ s.injournal > 0))
+ handle_stripe_dirtying(conf, sh, &s, disks);
+ }
+ }
/* maybe we need to check and possibly fix the parity for this stripe
* Any reads will already have been scheduled, so we just see if enough
@@ -5110,6 +5146,7 @@ static void raid5_make_request(struct mddev *mddev, struct bio * bi)
* data on failed drives.
*/
if (rw == READ && mddev->degraded == 0 &&
+ !r5c_is_writeback(conf->log) &&
mddev->reshape_position == MaxSector) {
bi = chunk_aligned_read(mddev, bi);
if (!bi)
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
index ffc13c4..c9590a8 100644
--- a/drivers/md/raid5.h
+++ b/drivers/md/raid5.h
@@ -264,6 +264,7 @@ struct stripe_head_state {
int syncing, expanding, expanded, replacing;
int locked, uptodate, to_read, to_write, failed, written;
int to_fill, compute, req_compute, non_overwrite;
+ int injournal;
int failed_num[2];
int p_failed, q_failed;
int dec_preread_active;
@@ -313,6 +314,11 @@ enum r5dev_flags {
*/
R5_Discard, /* Discard the stripe */
R5_SkipCopy, /* Don't copy data from bio to stripe cache */
+ R5_InJournal, /* data being written is in the journal device.
+ * if R5_InJournal is set for parity pd_idx, all the
+ * data and parity being written are in the journal
+ * device
+ */
};
/*
@@ -345,7 +351,23 @@ enum {
STRIPE_BITMAP_PENDING, /* Being added to bitmap, don't add
* to batch yet.
*/
- STRIPE_LOG_TRAPPED, /* trapped into log */
+ STRIPE_LOG_TRAPPED, /* trapped into log (see raid5-cache.c)
+ * this bit is used in two scenarios:
+ *
+ * 1. write-out phase
+ * set in first entry of r5l_write_stripe
+ * clear in second entry of r5l_write_stripe
+ * used to bypass logic in handle_stripe
+ *
+ * 2. caching phase
+ * set in r5c_try_caching_write()
+ * clear when journal write is done
+ * used to initiate r5c_cache_data()
+ * also used to bypass logic in handle_stripe
+ */
+ STRIPE_R5C_CACHING, /* the stripe is in caching phase
+ * see more detail in the raid5-cache.c
+ */
};
#define STRIPE_EXPAND_SYNC_FLAGS \
@@ -710,4 +732,11 @@ extern void r5l_stripe_write_finished(struct stripe_head *sh);
extern int r5l_handle_flush_request(struct r5l_log *log, struct bio *bio);
extern void r5l_quiesce(struct r5l_log *log, int state);
extern bool r5l_log_disk_error(struct r5conf *conf);
+extern bool r5c_is_writeback(struct r5l_log *log);
+extern int
+r5c_try_caching_write(struct r5conf *conf, struct stripe_head *sh,
+ struct stripe_head_state *s, int disks);
+extern void
+r5c_finish_stripe_write_out(struct r5conf *conf, struct stripe_head *sh,
+ struct stripe_head_state *s);
#endif
--
2.9.3
^ permalink raw reply related
* [PATCH v7 04/10] md/r5cache: caching phase of r5cache
From: Song Liu @ 2016-11-17 23:24 UTC (permalink / raw)
To: linux-raid
Cc: neilb, shli, kernel-team, dan.j.williams, hch, liuzhengyuang521,
liuzhengyuan, Song Liu
In-Reply-To: <20161117232445.1798305-1-songliubraving@fb.com>
As described in previous patch, write back cache operates in two
phases: caching and writing-out. The caching phase works as:
1. write data to journal
(r5c_handle_stripe_dirtying, r5c_cache_data)
2. call bio_endio
(r5c_handle_data_cached, r5c_return_dev_pending_writes).
Then the writing-out phase is as:
1. Mark the stripe as write-out (r5c_make_stripe_write_out)
2. Calcualte parity (reconstruct or RMW)
3. Write parity (and maybe some other data) to journal device
4. Write data and parity to RAID disks
This patch implements caching phase. The cache is integrated with
stripe cache of raid456. It leverages code of r5l_log to write
data to journal device.
Writing-out phase of the cache is implemented in the next patch.
With r5cache, write operation does not wait for parity calculation
and write out, so the write latency is lower (1 write to journal
device vs. read and then write to raid disks). Also, r5cache will
reduce RAID overhead (multipile IO due to read-modify-write of
parity) and provide more opportunities of full stripe writes.
This patch adds 2 flags to stripe_head.state:
- STRIPE_R5C_PARTIAL_STRIPE,
- STRIPE_R5C_FULL_STRIPE,
Instead of inactive_list, stripes with cached data are tracked in
r5conf->r5c_full_stripe_list and r5conf->r5c_partial_stripe_list.
STRIPE_R5C_FULL_STRIPE and STRIPE_R5C_PARTIAL_STRIPE are flags for
stripes in these lists. Note: stripes in r5c_full/partial_stripe_list
are not considered as "active".
For RMW, the code allocates an extra page for each data block
being updated. This is stored in r5dev->orig_page and the old data
is read into it. Then the prexor calculation subtracts ->orig_page
from the parity block, and the reconstruct calculation adds the
->page data back into the parity block.
r5cache naturally excludes SkipCopy. When the array has write back
cache, async_copy_data() will not skip copy.
There are some known limitations of the cache implementation:
1. Write cache only covers full page writes (R5_OVERWRITE). Writes
of smaller granularity are write through.
2. Only one log io (sh->log_io) for each stripe at anytime. Later
writes for the same stripe have to wait. This can be improved by
moving log_io to r5dev.
3. With writeback cache, read path must enter state machine, which
is a significant bottleneck for some workloads.
4. There is no per stripe checkpoint (with r5l_payload_flush) in
the log, so recovery code has to replay more than necessary data
(sometimes all the log from last_checkpoint). This reduces
availability of the array.
This patch includes a fix proposed by ZhengYuan Liu
<liuzhengyuan@kylinos.cn>
Signed-off-by: Song Liu <songliubraving@fb.com>
---
drivers/md/raid5-cache.c | 242 +++++++++++++++++++++++++++++++++++++++++++++--
drivers/md/raid5.c | 152 ++++++++++++++++++++++++-----
drivers/md/raid5.h | 19 +++-
3 files changed, 381 insertions(+), 32 deletions(-)
diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c
index 02a5544..19c5af9 100644
--- a/drivers/md/raid5-cache.c
+++ b/drivers/md/raid5-cache.c
@@ -20,6 +20,7 @@
#include <linux/random.h>
#include "md.h"
#include "raid5.h"
+#include "bitmap.h"
/*
* metadata/data stored in disk with 4k size unit (a block) regardless
@@ -218,6 +219,43 @@ static void __r5l_set_io_unit_state(struct r5l_io_unit *io,
io->state = state;
}
+static void
+r5c_return_dev_pending_writes(struct r5conf *conf, struct r5dev *dev,
+ struct bio_list *return_bi)
+{
+ struct bio *wbi, *wbi2;
+
+ wbi = dev->written;
+ dev->written = NULL;
+ while (wbi && wbi->bi_iter.bi_sector <
+ dev->sector + STRIPE_SECTORS) {
+ wbi2 = r5_next_bio(wbi, dev->sector);
+ if (!raid5_dec_bi_active_stripes(wbi)) {
+ md_write_end(conf->mddev);
+ bio_list_add(return_bi, wbi);
+ }
+ wbi = wbi2;
+ }
+}
+
+void r5c_handle_cached_data_endio(struct r5conf *conf,
+ struct stripe_head *sh, int disks, struct bio_list *return_bi)
+{
+ int i;
+
+ for (i = sh->disks; i--; ) {
+ if (sh->dev[i].written) {
+ set_bit(R5_UPTODATE, &sh->dev[i].flags);
+ r5c_return_dev_pending_writes(conf, &sh->dev[i],
+ return_bi);
+ bitmap_endwrite(conf->mddev->bitmap, sh->sector,
+ STRIPE_SECTORS,
+ !test_bit(STRIPE_DEGRADED, &sh->state),
+ 0);
+ }
+ }
+}
+
/*
* Put the stripe into writing-out phase by clearing STRIPE_R5C_CACHING.
* This function should only be called in write-back mode.
@@ -231,6 +269,44 @@ static void r5c_make_stripe_write_out(struct stripe_head *sh)
WARN_ON(!test_bit(STRIPE_R5C_CACHING, &sh->state));
clear_bit(STRIPE_R5C_CACHING, &sh->state);
+
+ if (!test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+ atomic_inc(&conf->preread_active_stripes);
+
+ if (test_and_clear_bit(STRIPE_R5C_PARTIAL_STRIPE, &sh->state)) {
+ BUG_ON(atomic_read(&conf->r5c_cached_partial_stripes) == 0);
+ atomic_dec(&conf->r5c_cached_partial_stripes);
+ }
+
+ if (test_and_clear_bit(STRIPE_R5C_FULL_STRIPE, &sh->state)) {
+ BUG_ON(atomic_read(&conf->r5c_cached_full_stripes) == 0);
+ atomic_dec(&conf->r5c_cached_full_stripes);
+ }
+}
+
+static void r5c_handle_data_cached(struct stripe_head *sh)
+{
+ int i;
+
+ for (i = sh->disks; i--; )
+ if (test_and_clear_bit(R5_Wantwrite, &sh->dev[i].flags)) {
+ set_bit(R5_InJournal, &sh->dev[i].flags);
+ clear_bit(R5_LOCKED, &sh->dev[i].flags);
+ }
+ clear_bit(STRIPE_LOG_TRAPPED, &sh->state);
+}
+
+/*
+ * this journal write must contain full parity,
+ * it may also contain some data pages
+ */
+static void r5c_handle_parity_cached(struct stripe_head *sh)
+{
+ int i;
+
+ for (i = sh->disks; i--; )
+ if (test_bit(R5_InJournal, &sh->dev[i].flags))
+ set_bit(R5_Wantwrite, &sh->dev[i].flags);
}
/*
@@ -250,8 +326,12 @@ static void r5c_finish_cache_stripe(struct stripe_head *sh)
* two parities are written out together.
*/
set_bit(R5_InJournal, &sh->dev[sh->pd_idx].flags);
- } else
- BUG(); /* write-back logic in next patch */
+ } else if (test_bit(STRIPE_R5C_CACHING, &sh->state)) {
+ r5c_handle_data_cached(sh);
+ } else {
+ r5c_handle_parity_cached(sh);
+ set_bit(R5_InJournal, &sh->dev[sh->pd_idx].flags);
+ }
}
static void r5l_io_run_stripes(struct r5l_io_unit *io)
@@ -491,7 +571,8 @@ static int r5l_log_stripe(struct r5l_log *log, struct stripe_head *sh,
io = log->current_io;
for (i = 0; i < sh->disks; i++) {
- if (!test_bit(R5_Wantwrite, &sh->dev[i].flags))
+ if (!test_bit(R5_Wantwrite, &sh->dev[i].flags) ||
+ test_bit(R5_InJournal, &sh->dev[i].flags))
continue;
if (i == sh->pd_idx || i == sh->qd_idx)
continue;
@@ -550,8 +631,10 @@ int r5l_write_stripe(struct r5l_log *log, struct stripe_head *sh)
for (i = 0; i < sh->disks; i++) {
void *addr;
- if (!test_bit(R5_Wantwrite, &sh->dev[i].flags))
+ if (!test_bit(R5_Wantwrite, &sh->dev[i].flags) ||
+ test_bit(R5_InJournal, &sh->dev[i].flags))
continue;
+
write_disks++;
/* checksum is already calculated in last run */
if (test_bit(STRIPE_LOG_TRAPPED, &sh->state))
@@ -817,7 +900,6 @@ static void r5l_write_super_and_discard_space(struct r5l_log *log,
}
}
-
static void r5l_do_reclaim(struct r5l_log *log)
{
sector_t reclaim_target = xchg(&log->reclaim_target, 0);
@@ -1218,12 +1300,80 @@ int r5c_try_caching_write(struct r5conf *conf,
int disks)
{
struct r5l_log *log = conf->log;
+ int i;
+ struct r5dev *dev;
+ int to_cache = 0;
BUG_ON(!r5c_is_writeback(log));
- /* more write-back logic in next patches */
- r5c_make_stripe_write_out(sh);
- return -EAGAIN;
+ if (!test_bit(STRIPE_R5C_CACHING, &sh->state)) {
+ /*
+ * There are two different scenarios here:
+ * 1. The stripe has some data cached, and it is sent to
+ * write-out phase for reclaim
+ * 2. The stripe is clean, and this is the first write
+ *
+ * For 1, return -EAGAIN, so we continue with
+ * handle_stripe_dirtying().
+ *
+ * For 2, set STRIPE_R5C_CACHING and continue with caching
+ * write.
+ */
+
+ /* case 1: anything injournal or anything in written */
+ if (s->injournal > 0 || s->written > 0)
+ return -EAGAIN;
+ /* case 2 */
+ set_bit(STRIPE_R5C_CACHING, &sh->state);
+ }
+
+ for (i = disks; i--; ) {
+ dev = &sh->dev[i];
+ /* if non-overwrite, use writing-out phase */
+ if (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags) &&
+ !test_bit(R5_InJournal, &dev->flags)) {
+ r5c_make_stripe_write_out(sh);
+ return -EAGAIN;
+ }
+ }
+
+ for (i = disks; i--; ) {
+ dev = &sh->dev[i];
+ if (dev->towrite) {
+ set_bit(R5_Wantwrite, &dev->flags);
+ set_bit(R5_Wantdrain, &dev->flags);
+ set_bit(R5_LOCKED, &dev->flags);
+ to_cache++;
+ }
+ }
+
+ if (to_cache) {
+ set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
+ /*
+ * set STRIPE_LOG_TRAPPED, which triggers r5c_cache_data()
+ * in ops_run_io(). STRIPE_LOG_TRAPPED will be cleared in
+ * r5c_handle_data_cached()
+ */
+ set_bit(STRIPE_LOG_TRAPPED, &sh->state);
+ }
+
+ return 0;
+}
+
+/*
+ * free extra pages (orig_page) we allocated for prexor
+ */
+void r5c_release_extra_page(struct stripe_head *sh)
+{
+ int i;
+
+ for (i = sh->disks; i--; )
+ if (sh->dev[i].page != sh->dev[i].orig_page) {
+ struct page *p = sh->dev[i].orig_page;
+
+ sh->dev[i].orig_page = sh->dev[i].page;
+ put_page(p);
+ }
}
/*
@@ -1234,6 +1384,9 @@ void r5c_finish_stripe_write_out(struct r5conf *conf,
struct stripe_head *sh,
struct stripe_head_state *s)
{
+ int i;
+ int do_wakeup = 0;
+
if (!conf->log ||
!test_bit(R5_InJournal, &sh->dev[sh->pd_idx].flags))
return;
@@ -1243,7 +1396,78 @@ void r5c_finish_stripe_write_out(struct r5conf *conf,
if (conf->log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH)
return;
- BUG(); /* write-back logic in following patches */
+
+ for (i = sh->disks; i--; ) {
+ clear_bit(R5_InJournal, &sh->dev[i].flags);
+ if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags))
+ do_wakeup = 1;
+ }
+
+ /*
+ * analyse_stripe() runs before r5c_finish_stripe_write_out(),
+ * We updated R5_InJournal, so we also update s->injournal.
+ */
+ s->injournal = 0;
+
+ if (test_and_clear_bit(STRIPE_FULL_WRITE, &sh->state))
+ if (atomic_dec_and_test(&conf->pending_full_writes))
+ md_wakeup_thread(conf->mddev->thread);
+
+ if (do_wakeup)
+ wake_up(&conf->wait_for_overlap);
+}
+
+int
+r5c_cache_data(struct r5l_log *log, struct stripe_head *sh,
+ struct stripe_head_state *s)
+{
+ int pages = 0;
+ int reserve;
+ int i;
+ int ret = 0;
+
+ BUG_ON(!log);
+
+ for (i = 0; i < sh->disks; i++) {
+ void *addr;
+
+ if (!test_bit(R5_Wantwrite, &sh->dev[i].flags))
+ continue;
+ addr = kmap_atomic(sh->dev[i].page);
+ sh->dev[i].log_checksum = crc32c_le(log->uuid_checksum,
+ addr, PAGE_SIZE);
+ kunmap_atomic(addr);
+ pages++;
+ }
+ WARN_ON(pages == 0);
+
+ /*
+ * The stripe must enter state machine again to call endio, so
+ * don't delay.
+ */
+ clear_bit(STRIPE_DELAYED, &sh->state);
+ atomic_inc(&sh->count);
+
+ mutex_lock(&log->io_mutex);
+ /* meta + data */
+ reserve = (1 + pages) << (PAGE_SHIFT - 9);
+ if (!r5l_has_free_space(log, reserve)) {
+ spin_lock(&log->no_space_stripes_lock);
+ list_add_tail(&sh->log_list, &log->no_space_stripes);
+ spin_unlock(&log->no_space_stripes_lock);
+
+ r5l_wake_reclaim(log, reserve);
+ } else {
+ ret = r5l_log_stripe(log, sh, pages, 0);
+ if (ret) {
+ spin_lock_irq(&log->io_list_lock);
+ list_add_tail(&sh->log_list, &log->no_mem_stripes);
+ spin_unlock_irq(&log->io_list_lock);
+ }
+ }
+
+ mutex_unlock(&log->io_mutex);
+ return 0;
}
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 7c98eb0..f535ce2 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -218,8 +218,17 @@ static void raid5_wakeup_stripe_thread(struct stripe_head *sh)
static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh,
struct list_head *temp_inactive_list)
{
+ int i;
+ int injournal = 0; /* number of date pages with R5_InJournal */
+
BUG_ON(!list_empty(&sh->lru));
BUG_ON(atomic_read(&conf->active_stripes)==0);
+
+ if (r5c_is_writeback(conf->log))
+ for (i = sh->disks; i--; )
+ if (test_bit(R5_InJournal, &sh->dev[i].flags))
+ injournal++;
+
if (test_bit(STRIPE_HANDLE, &sh->state)) {
if (test_bit(STRIPE_DELAYED, &sh->state) &&
!test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
@@ -245,8 +254,29 @@ static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh,
< IO_THRESHOLD)
md_wakeup_thread(conf->mddev->thread);
atomic_dec(&conf->active_stripes);
- if (!test_bit(STRIPE_EXPANDING, &sh->state))
- list_add_tail(&sh->lru, temp_inactive_list);
+ if (!test_bit(STRIPE_EXPANDING, &sh->state)) {
+ if (!r5c_is_writeback(conf->log))
+ list_add_tail(&sh->lru, temp_inactive_list);
+ else {
+ WARN_ON(test_bit(R5_InJournal, &sh->dev[sh->pd_idx].flags));
+ if (injournal == 0)
+ list_add_tail(&sh->lru, temp_inactive_list);
+ else if (injournal == conf->raid_disks - conf->max_degraded) {
+ /* full stripe */
+ if (!test_and_set_bit(STRIPE_R5C_FULL_STRIPE, &sh->state))
+ atomic_inc(&conf->r5c_cached_full_stripes);
+ if (test_and_clear_bit(STRIPE_R5C_PARTIAL_STRIPE, &sh->state))
+ atomic_dec(&conf->r5c_cached_partial_stripes);
+ list_add_tail(&sh->lru, &conf->r5c_full_stripe_list);
+ } else {
+ /* partial stripe */
+ if (!test_and_set_bit(STRIPE_R5C_PARTIAL_STRIPE,
+ &sh->state))
+ atomic_inc(&conf->r5c_cached_partial_stripes);
+ list_add_tail(&sh->lru, &conf->r5c_partial_stripe_list);
+ }
+ }
+ }
}
}
@@ -830,8 +860,17 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
might_sleep();
- if (r5l_write_stripe(conf->log, sh) == 0)
- return;
+ if (!test_bit(STRIPE_R5C_CACHING, &sh->state)) {
+ /* writing out phase */
+ if (r5l_write_stripe(conf->log, sh) == 0)
+ return;
+ } else { /* caching phase */
+ if (test_bit(STRIPE_LOG_TRAPPED, &sh->state)) {
+ r5c_cache_data(conf->log, sh, s);
+ return;
+ }
+ }
+
for (i = disks; i--; ) {
int op, op_flags = 0;
int replace_only = 0;
@@ -1044,7 +1083,7 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
static struct dma_async_tx_descriptor *
async_copy_data(int frombio, struct bio *bio, struct page **page,
sector_t sector, struct dma_async_tx_descriptor *tx,
- struct stripe_head *sh)
+ struct stripe_head *sh, int no_skipcopy)
{
struct bio_vec bvl;
struct bvec_iter iter;
@@ -1084,7 +1123,8 @@ async_copy_data(int frombio, struct bio *bio, struct page **page,
if (frombio) {
if (sh->raid_conf->skip_copy &&
b_offset == 0 && page_offset == 0 &&
- clen == STRIPE_SIZE)
+ clen == STRIPE_SIZE &&
+ !no_skipcopy)
*page = bio_page;
else
tx = async_memcpy(*page, bio_page, page_offset,
@@ -1166,7 +1206,7 @@ static void ops_run_biofill(struct stripe_head *sh)
while (rbi && rbi->bi_iter.bi_sector <
dev->sector + STRIPE_SECTORS) {
tx = async_copy_data(0, rbi, &dev->page,
- dev->sector, tx, sh);
+ dev->sector, tx, sh, 0);
rbi = r5_next_bio(rbi, dev->sector);
}
}
@@ -1293,10 +1333,15 @@ static int set_syndrome_sources(struct page **srcs,
if (i == sh->qd_idx || i == sh->pd_idx ||
(srctype == SYNDROME_SRC_ALL) ||
(srctype == SYNDROME_SRC_WANT_DRAIN &&
- test_bit(R5_Wantdrain, &dev->flags)) ||
+ (test_bit(R5_Wantdrain, &dev->flags) ||
+ test_bit(R5_InJournal, &dev->flags))) ||
(srctype == SYNDROME_SRC_WRITTEN &&
- dev->written))
- srcs[slot] = sh->dev[i].page;
+ dev->written)) {
+ if (test_bit(R5_InJournal, &dev->flags))
+ srcs[slot] = sh->dev[i].orig_page;
+ else
+ srcs[slot] = sh->dev[i].page;
+ }
i = raid6_next_disk(i, disks);
} while (i != d0_idx);
@@ -1475,6 +1520,13 @@ static void ops_complete_prexor(void *stripe_head_ref)
pr_debug("%s: stripe %llu\n", __func__,
(unsigned long long)sh->sector);
+
+ if (r5c_is_writeback(sh->raid_conf->log))
+ /*
+ * raid5-cache write back uses orig_page during prexor.
+ * After prexor, it is time to free orig_page
+ */
+ r5c_release_extra_page(sh);
}
static struct dma_async_tx_descriptor *
@@ -1496,7 +1548,9 @@ ops_run_prexor5(struct stripe_head *sh, struct raid5_percpu *percpu,
for (i = disks; i--; ) {
struct r5dev *dev = &sh->dev[i];
/* Only process blocks that are known to be uptodate */
- if (test_bit(R5_Wantdrain, &dev->flags))
+ if (test_bit(R5_InJournal, &dev->flags))
+ xor_srcs[count++] = dev->orig_page;
+ else if (test_bit(R5_Wantdrain, &dev->flags))
xor_srcs[count++] = dev->page;
}
@@ -1530,6 +1584,7 @@ ops_run_prexor6(struct stripe_head *sh, struct raid5_percpu *percpu,
static struct dma_async_tx_descriptor *
ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
{
+ struct r5conf *conf = sh->raid_conf;
int disks = sh->disks;
int i;
struct stripe_head *head_sh = sh;
@@ -1547,6 +1602,11 @@ ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
again:
dev = &sh->dev[i];
+ /*
+ * clear R5_InJournal, so when rewriting a page in
+ * journal, it is not skipped by r5l_log_stripe()
+ */
+ clear_bit(R5_InJournal, &dev->flags);
spin_lock_irq(&sh->stripe_lock);
chosen = dev->towrite;
dev->towrite = NULL;
@@ -1566,8 +1626,10 @@ ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
set_bit(R5_Discard, &dev->flags);
else {
tx = async_copy_data(1, wbi, &dev->page,
- dev->sector, tx, sh);
- if (dev->page != dev->orig_page) {
+ dev->sector, tx, sh,
+ r5c_is_writeback(conf->log));
+ if (dev->page != dev->orig_page &&
+ !r5c_is_writeback(conf->log)) {
set_bit(R5_SkipCopy, &dev->flags);
clear_bit(R5_UPTODATE, &dev->flags);
clear_bit(R5_OVERWRITE, &dev->flags);
@@ -1675,7 +1737,8 @@ ops_run_reconstruct5(struct stripe_head *sh, struct raid5_percpu *percpu,
xor_dest = xor_srcs[count++] = sh->dev[pd_idx].page;
for (i = disks; i--; ) {
struct r5dev *dev = &sh->dev[i];
- if (head_sh->dev[i].written)
+ if (head_sh->dev[i].written ||
+ test_bit(R5_InJournal, &head_sh->dev[i].flags))
xor_srcs[count++] = dev->page;
}
} else {
@@ -2796,6 +2859,13 @@ schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
int level = conf->level;
if (rcw) {
+ /*
+ * In some cases, handle_stripe_dirtying initially decided to
+ * run rmw and allocates extra page for prexor. However, rcw is
+ * cheaper later on. We need to free the extra page now,
+ * because we won't be able to do that in ops_complete_prexor().
+ */
+ r5c_release_extra_page(sh);
for (i = disks; i--; ) {
struct r5dev *dev = &sh->dev[i];
@@ -2806,6 +2876,9 @@ schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
if (!expand)
clear_bit(R5_UPTODATE, &dev->flags);
s->locked++;
+ } else if (test_bit(R5_InJournal, &dev->flags)) {
+ set_bit(R5_LOCKED, &dev->flags);
+ s->locked++;
}
}
/* if we are not expanding this is a proper write request, and
@@ -2845,6 +2918,9 @@ schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
set_bit(R5_LOCKED, &dev->flags);
clear_bit(R5_UPTODATE, &dev->flags);
s->locked++;
+ } else if (test_bit(R5_InJournal, &dev->flags)) {
+ set_bit(R5_LOCKED, &dev->flags);
+ s->locked++;
}
}
if (!s->locked)
@@ -3516,9 +3592,12 @@ static void handle_stripe_dirtying(struct r5conf *conf,
} else for (i = disks; i--; ) {
/* would I have to read this buffer for read_modify_write */
struct r5dev *dev = &sh->dev[i];
- if ((dev->towrite || i == sh->pd_idx || i == sh->qd_idx) &&
+ if ((dev->towrite || i == sh->pd_idx || i == sh->qd_idx ||
+ test_bit(R5_InJournal, &dev->flags)) &&
!test_bit(R5_LOCKED, &dev->flags) &&
- !(test_bit(R5_UPTODATE, &dev->flags) ||
+ !((test_bit(R5_UPTODATE, &dev->flags) &&
+ (!test_bit(R5_InJournal, &dev->flags) ||
+ dev->page != dev->orig_page)) ||
test_bit(R5_Wantcompute, &dev->flags))) {
if (test_bit(R5_Insync, &dev->flags))
rmw++;
@@ -3530,13 +3609,15 @@ static void handle_stripe_dirtying(struct r5conf *conf,
i != sh->pd_idx && i != sh->qd_idx &&
!test_bit(R5_LOCKED, &dev->flags) &&
!(test_bit(R5_UPTODATE, &dev->flags) ||
- test_bit(R5_Wantcompute, &dev->flags))) {
+ test_bit(R5_InJournal, &dev->flags) ||
+ test_bit(R5_Wantcompute, &dev->flags))) {
if (test_bit(R5_Insync, &dev->flags))
rcw++;
else
rcw += 2*disks;
}
}
+
pr_debug("for sector %llu, rmw=%d rcw=%d\n",
(unsigned long long)sh->sector, rmw, rcw);
set_bit(STRIPE_HANDLE, &sh->state);
@@ -3548,10 +3629,24 @@ static void handle_stripe_dirtying(struct r5conf *conf,
(unsigned long long)sh->sector, rmw);
for (i = disks; i--; ) {
struct r5dev *dev = &sh->dev[i];
- if ((dev->towrite || i == sh->pd_idx || i == sh->qd_idx) &&
+ if (test_bit(R5_InJournal, &dev->flags) &&
+ dev->page == dev->orig_page &&
+ !test_bit(R5_LOCKED, &sh->dev[sh->pd_idx].flags)) {
+ /* alloc page for prexor */
+ dev->orig_page = alloc_page(GFP_NOIO);
+
+ /* will handle failure in a later patch*/
+ BUG_ON(!dev->orig_page);
+ }
+
+ if ((dev->towrite ||
+ i == sh->pd_idx || i == sh->qd_idx ||
+ test_bit(R5_InJournal, &dev->flags)) &&
!test_bit(R5_LOCKED, &dev->flags) &&
- !(test_bit(R5_UPTODATE, &dev->flags) ||
- test_bit(R5_Wantcompute, &dev->flags)) &&
+ !((test_bit(R5_UPTODATE, &dev->flags) &&
+ (!test_bit(R5_InJournal, &dev->flags) ||
+ dev->page != dev->orig_page)) ||
+ test_bit(R5_Wantcompute, &dev->flags)) &&
test_bit(R5_Insync, &dev->flags)) {
if (test_bit(STRIPE_PREREAD_ACTIVE,
&sh->state)) {
@@ -3577,6 +3672,7 @@ static void handle_stripe_dirtying(struct r5conf *conf,
i != sh->pd_idx && i != sh->qd_idx &&
!test_bit(R5_LOCKED, &dev->flags) &&
!(test_bit(R5_UPTODATE, &dev->flags) ||
+ test_bit(R5_InJournal, &dev->flags) ||
test_bit(R5_Wantcompute, &dev->flags))) {
rcw++;
if (test_bit(R5_Insync, &dev->flags) &&
@@ -3616,7 +3712,7 @@ static void handle_stripe_dirtying(struct r5conf *conf,
*/
if ((s->req_compute || !test_bit(STRIPE_COMPUTE_RUN, &sh->state)) &&
(s->locked == 0 && (rcw == 0 || rmw == 0) &&
- !test_bit(STRIPE_BIT_DELAY, &sh->state)))
+ !test_bit(STRIPE_BIT_DELAY, &sh->state)))
schedule_reconstruction(sh, s, rcw == 0, 0);
}
@@ -4110,6 +4206,8 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
if (test_bit(R5_InJournal, &dev->flags))
s->injournal++;
+ if (test_bit(R5_InJournal, &dev->flags) && dev->written)
+ s->just_cached++;
}
if (test_bit(STRIPE_SYNCING, &sh->state)) {
/* If there is a failed device being replaced,
@@ -4338,7 +4436,8 @@ static void handle_stripe(struct stripe_head *sh)
struct r5dev *dev = &sh->dev[i];
if (test_bit(R5_LOCKED, &dev->flags) &&
(i == sh->pd_idx || i == sh->qd_idx ||
- dev->written)) {
+ dev->written || test_bit(R5_InJournal,
+ &dev->flags))) {
pr_debug("Writing block %d\n", i);
set_bit(R5_Wantwrite, &dev->flags);
if (prexor)
@@ -4378,6 +4477,10 @@ static void handle_stripe(struct stripe_head *sh)
test_bit(R5_Discard, &qdev->flags))))))
handle_stripe_clean_event(conf, sh, disks, &s.return_bi);
+ if (s.just_cached)
+ r5c_handle_cached_data_endio(conf, sh, disks, &s.return_bi);
+ r5l_stripe_write_finished(sh);
+
/* Now we might consider reading some blocks, either to check/generate
* parity, or to satisfy requests
* or to load a block that is being partially written.
@@ -6499,6 +6602,11 @@ static struct r5conf *setup_conf(struct mddev *mddev)
for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++)
INIT_LIST_HEAD(conf->temp_inactive_list + i);
+ atomic_set(&conf->r5c_cached_full_stripes, 0);
+ INIT_LIST_HEAD(&conf->r5c_full_stripe_list);
+ atomic_set(&conf->r5c_cached_partial_stripes, 0);
+ INIT_LIST_HEAD(&conf->r5c_partial_stripe_list);
+
conf->level = mddev->new_level;
conf->chunk_sectors = mddev->new_chunk_sectors;
if (raid5_alloc_percpu(conf) != 0)
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
index c9590a8..73c1833 100644
--- a/drivers/md/raid5.h
+++ b/drivers/md/raid5.h
@@ -264,7 +264,7 @@ struct stripe_head_state {
int syncing, expanding, expanded, replacing;
int locked, uptodate, to_read, to_write, failed, written;
int to_fill, compute, req_compute, non_overwrite;
- int injournal;
+ int injournal, just_cached;
int failed_num[2];
int p_failed, q_failed;
int dec_preread_active;
@@ -368,6 +368,12 @@ enum {
STRIPE_R5C_CACHING, /* the stripe is in caching phase
* see more detail in the raid5-cache.c
*/
+ STRIPE_R5C_PARTIAL_STRIPE, /* in r5c cache (to-be/being handled or
+ * in conf->r5c_partial_stripe_list)
+ */
+ STRIPE_R5C_FULL_STRIPE, /* in r5c cache (to-be/being handled or
+ * in conf->r5c_full_stripe_list)
+ */
};
#define STRIPE_EXPAND_SYNC_FLAGS \
@@ -618,6 +624,12 @@ struct r5conf {
*/
atomic_t active_stripes;
struct list_head inactive_list[NR_STRIPE_HASH_LOCKS];
+
+ atomic_t r5c_cached_full_stripes;
+ struct list_head r5c_full_stripe_list;
+ atomic_t r5c_cached_partial_stripes;
+ struct list_head r5c_partial_stripe_list;
+
atomic_t empty_inactive_list_nr;
struct llist_head released_stripes;
wait_queue_head_t wait_for_quiescent;
@@ -739,4 +751,9 @@ r5c_try_caching_write(struct r5conf *conf, struct stripe_head *sh,
extern void
r5c_finish_stripe_write_out(struct r5conf *conf, struct stripe_head *sh,
struct stripe_head_state *s);
+extern void r5c_release_extra_page(struct stripe_head *sh);
+extern void r5c_handle_cached_data_endio(struct r5conf *conf,
+ struct stripe_head *sh, int disks, struct bio_list *return_bi);
+extern int r5c_cache_data(struct r5l_log *log, struct stripe_head *sh,
+ struct stripe_head_state *s);
#endif
--
2.9.3
^ permalink raw reply related
* [PATCH v7 05/10] md/r5cache: write-out phase and reclaim support
From: Song Liu @ 2016-11-17 23:24 UTC (permalink / raw)
To: linux-raid
Cc: neilb, shli, kernel-team, dan.j.williams, hch, liuzhengyuang521,
liuzhengyuan, Song Liu
In-Reply-To: <20161117232445.1798305-1-songliubraving@fb.com>
There are two limited resources, stripe cache and journal disk space.
For better performance, we priotize reclaim of full stripe writes.
To free up more journal space, we free earliest data on the journal.
In current implementation, reclaim happens when:
1. Periodically (every R5C_RECLAIM_WAKEUP_INTERVAL, 30 seconds) reclaim
if there is no reclaim in the past 5 seconds.
2. when there are R5C_FULL_STRIPE_FLUSH_BATCH (256) cached full stripes,
or cached stripes is enough for a full stripe (chunk size / 4k)
(r5c_check_cached_full_stripe)
3. when there is pressure on stripe cache (r5c_check_stripe_cache_usage)
4. when there is pressure on journal space (r5l_write_stripe, r5c_cache_data)
r5c_do_reclaim() contains new logic of reclaim.
For stripe cache:
When stripe cache pressure is high (more than 3/4 stripes are cached,
or there is empty inactive lists), flush all full stripe. If fewer
than R5C_RECLAIM_STRIPE_GROUP (NR_STRIPE_HASH_LOCKS * 2) full stripes
are flushed, flush some paritial stripes. When stripe cache pressure
is moderate (1/2 to 3/4 of stripes are cached), flush all full stripes.
For log space:
To avoid deadlock due to log space, we need to reserve enough space
to flush cached data. The size of required log space depends on total
number of cached stripes (stripe_in_journal_count). In current
implementation, the writing-out phase automatically include pending
data writes with parity writes (similar to write through case).
Therefore, we need up to (conf->raid_disks + 1) pages for each cached
stripe (1 page for meta data, raid_disks pages for all data and
parity). r5c_log_required_to_flush_cache() calculates log space
required to flush cache. In the following, we refer to the space
calculated by r5c_log_required_to_flush_cache() as
reclaim_required_space.
Two flags are added to r5conf->cache_state: R5C_LOG_TIGHT and
R5C_LOG_CRITICAL. R5C_LOG_TIGHT is set when free space on the log
device is less than 3x of reclaim_required_space. R5C_LOG_CRITICAL
is set when free space on the log device is less than 2x of
reclaim_required_space.
r5c_cache keeps all data in cache (not fully committed to RAID) in
a list (stripe_in_journal_list). These stripes are in the order of their
first appearance on the journal. So the log tail (last_checkpoint)
should point to the journal_start of the first item in the list.
When R5C_LOG_TIGHT is set, r5l_reclaim_thread starts flushing out
stripes at the head of stripe_in_journal. When R5C_LOG_CRITICAL is
set, the state machine only writes data that are already in the
log device (in stripe_in_journal_list).
This patch includes a fix to improve performance by
Shaohua Li <shli@fb.com>.
Signed-off-by: Song Liu <songliubraving@fb.com>
---
drivers/md/raid5-cache.c | 411 +++++++++++++++++++++++++++++++++++++++++++----
drivers/md/raid5.c | 21 +++
drivers/md/raid5.h | 39 +++--
3 files changed, 430 insertions(+), 41 deletions(-)
diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c
index 19c5af9..7dec2a0 100644
--- a/drivers/md/raid5-cache.c
+++ b/drivers/md/raid5-cache.c
@@ -29,12 +29,21 @@
#define BLOCK_SECTORS (8)
/*
- * reclaim runs every 1/4 disk size or 10G reclaimable space. This can prevent
- * recovery scans a very long log
+ * log->max_free_space is min(1/4 disk size, 10G reclaimable space).
+ *
+ * In write through mode, the reclaim runs every log->max_free_space.
+ * This can prevent the recovery scans for too long
*/
#define RECLAIM_MAX_FREE_SPACE (10 * 1024 * 1024 * 2) /* sector */
#define RECLAIM_MAX_FREE_SPACE_SHIFT (2)
+/* wake up reclaim thread periodically */
+#define R5C_RECLAIM_WAKEUP_INTERVAL (30 * HZ)
+/* start flush with these full stripes */
+#define R5C_FULL_STRIPE_FLUSH_BATCH 256
+/* reclaim stripes in groups */
+#define R5C_RECLAIM_STRIPE_GROUP (NR_STRIPE_HASH_LOCKS * 2)
+
/*
* We only need 2 bios per I/O unit to make progress, but ensure we
* have a few more available to not get too tight.
@@ -141,6 +150,12 @@ struct r5l_log {
/* for r5c_cache */
enum r5c_journal_mode r5c_journal_mode;
+
+ /* all stripes in r5cache, in the order of seq at sh->log_start */
+ struct list_head stripe_in_journal_list;
+
+ spinlock_t stripe_in_journal_lock;
+ atomic_t stripe_in_journal_count;
};
/*
@@ -256,11 +271,109 @@ void r5c_handle_cached_data_endio(struct r5conf *conf,
}
}
+/* Check whether we should flush some stripes to free up stripe cache */
+void r5c_check_stripe_cache_usage(struct r5conf *conf)
+{
+ int total_cached;
+
+ if (!r5c_is_writeback(conf->log))
+ return;
+
+ total_cached = atomic_read(&conf->r5c_cached_partial_stripes) +
+ atomic_read(&conf->r5c_cached_full_stripes);
+
+ /*
+ * The following condition is true for either of the following:
+ * - stripe cache pressure high:
+ * total_cached > 3/4 min_nr_stripes ||
+ * empty_inactive_list_nr > 0
+ * - stripe cache pressure moderate:
+ * total_cached > 1/2 min_nr_stripes
+ */
+ if (total_cached > conf->min_nr_stripes * 1 / 2 ||
+ atomic_read(&conf->empty_inactive_list_nr) > 0)
+ r5l_wake_reclaim(conf->log, 0);
+}
+
+/*
+ * flush cache when there are R5C_FULL_STRIPE_FLUSH_BATCH or more full
+ * stripes in the cache
+ */
+void r5c_check_cached_full_stripe(struct r5conf *conf)
+{
+ if (!r5c_is_writeback(conf->log))
+ return;
+
+ /*
+ * wake up reclaim for R5C_FULL_STRIPE_FLUSH_BATCH cached stripes
+ * or a full stripe (chunk size / 4k stripes).
+ */
+ if (atomic_read(&conf->r5c_cached_full_stripes) >=
+ min(R5C_FULL_STRIPE_FLUSH_BATCH,
+ conf->chunk_sectors >> STRIPE_SHIFT))
+ r5l_wake_reclaim(conf->log, 0);
+}
+
+/*
+ * Total log space (in sectors) needed to flush all data in cache
+ *
+ * Currently, writing-out phase automatically includes all pending writes
+ * to the same sector. So the reclaim of each stripe takes up to
+ * (conf->raid_disks + 1) pages of log space.
+ *
+ * To totally avoid deadlock due to log space, the code reserves
+ * (conf->raid_disks + 1) pages for each stripe in cache, which is not
+ * necessary in most cases.
+ *
+ * To improve this, we will need writing-out phase to be able to NOT include
+ * pending writes, which will reduce the requirement to
+ * (conf->max_degraded + 1) pages per stripe in cache.
+ */
+static sector_t r5c_log_required_to_flush_cache(struct r5conf *conf)
+{
+ struct r5l_log *log = conf->log;
+
+ if (!r5c_is_writeback(log))
+ return 0;
+
+ return BLOCK_SECTORS * (conf->raid_disks + 1) *
+ atomic_read(&log->stripe_in_journal_count);
+}
+
+/*
+ * evaluate log space usage and update R5C_LOG_TIGHT and R5C_LOG_CRITICAL
+ *
+ * R5C_LOG_TIGHT is set when free space on the log device is less than 3x of
+ * reclaim_required_space. R5C_LOG_CRITICAL is set when free space on the log
+ * device is less than 2x of reclaim_required_space.
+ */
+static inline void r5c_update_log_state(struct r5l_log *log)
+{
+ struct r5conf *conf = log->rdev->mddev->private;
+ sector_t free_space;
+ sector_t reclaim_space;
+
+ if (!r5c_is_writeback(log))
+ return;
+
+ free_space = r5l_ring_distance(log, log->log_start,
+ log->last_checkpoint);
+ reclaim_space = r5c_log_required_to_flush_cache(conf);
+ if (free_space < 2 * reclaim_space)
+ set_bit(R5C_LOG_CRITICAL, &conf->cache_state);
+ else
+ clear_bit(R5C_LOG_CRITICAL, &conf->cache_state);
+ if (free_space < 3 * reclaim_space)
+ set_bit(R5C_LOG_TIGHT, &conf->cache_state);
+ else
+ clear_bit(R5C_LOG_TIGHT, &conf->cache_state);
+}
+
/*
* Put the stripe into writing-out phase by clearing STRIPE_R5C_CACHING.
* This function should only be called in write-back mode.
*/
-static void r5c_make_stripe_write_out(struct stripe_head *sh)
+void r5c_make_stripe_write_out(struct stripe_head *sh)
{
struct r5conf *conf = sh->raid_conf;
struct r5l_log *log = conf->log;
@@ -440,6 +553,7 @@ static void r5_reserve_log_entry(struct r5l_log *log, struct r5l_io_unit *io)
{
log->log_start = r5l_ring_add(log, log->log_start, BLOCK_SECTORS);
+ r5c_update_log_state(log);
/*
* If we filled up the log device start from the beginning again,
* which will require a new bio.
@@ -600,21 +714,43 @@ static int r5l_log_stripe(struct r5l_log *log, struct stripe_head *sh,
atomic_inc(&io->pending_stripe);
sh->log_io = io;
+ if (log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH)
+ return 0;
+
+ if (sh->log_start == MaxSector) {
+ BUG_ON(!list_empty(&sh->r5c));
+ sh->log_start = io->log_start;
+ spin_lock_irq(&log->stripe_in_journal_lock);
+ list_add_tail(&sh->r5c,
+ &log->stripe_in_journal_list);
+ spin_unlock_irq(&log->stripe_in_journal_lock);
+ atomic_inc(&log->stripe_in_journal_count);
+ }
return 0;
}
-static void r5l_wake_reclaim(struct r5l_log *log, sector_t space);
+/* add stripe to no_space_stripes, and then wake up reclaim */
+static inline void r5l_add_no_space_stripe(struct r5l_log *log,
+ struct stripe_head *sh)
+{
+ spin_lock(&log->no_space_stripes_lock);
+ list_add_tail(&sh->log_list, &log->no_space_stripes);
+ spin_unlock(&log->no_space_stripes_lock);
+}
+
/*
* running in raid5d, where reclaim could wait for raid5d too (when it flushes
* data from log to raid disks), so we shouldn't wait for reclaim here
*/
int r5l_write_stripe(struct r5l_log *log, struct stripe_head *sh)
{
+ struct r5conf *conf = sh->raid_conf;
int write_disks = 0;
int data_pages, parity_pages;
int reserve;
int i;
int ret = 0;
+ bool wake_reclaim = false;
if (!log)
return -EAGAIN;
@@ -658,22 +794,49 @@ int r5l_write_stripe(struct r5l_log *log, struct stripe_head *sh)
mutex_lock(&log->io_mutex);
/* meta + data */
reserve = (1 + write_disks) << (PAGE_SHIFT - 9);
- if (!r5l_has_free_space(log, reserve)) {
- spin_lock(&log->no_space_stripes_lock);
- list_add_tail(&sh->log_list, &log->no_space_stripes);
- spin_unlock(&log->no_space_stripes_lock);
- r5l_wake_reclaim(log, reserve);
- } else {
- ret = r5l_log_stripe(log, sh, data_pages, parity_pages);
- if (ret) {
- spin_lock_irq(&log->io_list_lock);
- list_add_tail(&sh->log_list, &log->no_mem_stripes);
- spin_unlock_irq(&log->io_list_lock);
+ if (log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH) {
+ if (!r5l_has_free_space(log, reserve)) {
+ r5l_add_no_space_stripe(log, sh);
+ wake_reclaim = true;
+ } else {
+ ret = r5l_log_stripe(log, sh, data_pages, parity_pages);
+ if (ret) {
+ spin_lock_irq(&log->io_list_lock);
+ list_add_tail(&sh->log_list,
+ &log->no_mem_stripes);
+ spin_unlock_irq(&log->io_list_lock);
+ }
+ }
+ } else { /* R5C_JOURNAL_MODE_WRITE_BACK */
+ /*
+ * log space critical, do not process stripes that are
+ * not in cache yet (sh->log_start == MaxSector).
+ */
+ if (test_bit(R5C_LOG_CRITICAL, &conf->cache_state) &&
+ sh->log_start == MaxSector) {
+ r5l_add_no_space_stripe(log, sh);
+ wake_reclaim = true;
+ reserve = 0;
+ } else if (!r5l_has_free_space(log, reserve)) {
+ if (sh->log_start == log->last_checkpoint)
+ BUG();
+ else
+ r5l_add_no_space_stripe(log, sh);
+ } else {
+ ret = r5l_log_stripe(log, sh, data_pages, parity_pages);
+ if (ret) {
+ spin_lock_irq(&log->io_list_lock);
+ list_add_tail(&sh->log_list,
+ &log->no_mem_stripes);
+ spin_unlock_irq(&log->io_list_lock);
+ }
}
}
mutex_unlock(&log->io_mutex);
+ if (wake_reclaim)
+ r5l_wake_reclaim(log, reserve);
return 0;
}
@@ -720,10 +883,40 @@ static void r5l_run_no_space_stripes(struct r5l_log *log)
spin_unlock(&log->no_space_stripes_lock);
}
+/*
+ * calculate new last_checkpoint
+ * for write through mode, returns log->next_checkpoint
+ * for write back, returns log_start of first sh in stripe_in_journal_list
+ */
+static sector_t r5c_calculate_new_cp(struct r5conf *conf)
+{
+ struct stripe_head *sh;
+ struct r5l_log *log = conf->log;
+ sector_t new_cp;
+ unsigned long flags;
+
+ if (log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH)
+ return log->next_checkpoint;
+
+ spin_lock_irqsave(&log->stripe_in_journal_lock, flags);
+ if (list_empty(&conf->log->stripe_in_journal_list)) {
+ /* all stripes flushed */
+ spin_unlock(&log->stripe_in_journal_lock);
+ return log->next_checkpoint;
+ }
+ sh = list_first_entry(&conf->log->stripe_in_journal_list,
+ struct stripe_head, r5c);
+ new_cp = sh->log_start;
+ spin_unlock_irqrestore(&log->stripe_in_journal_lock, flags);
+ return new_cp;
+}
+
static sector_t r5l_reclaimable_space(struct r5l_log *log)
{
+ struct r5conf *conf = log->rdev->mddev->private;
+
return r5l_ring_distance(log, log->last_checkpoint,
- log->next_checkpoint);
+ r5c_calculate_new_cp(conf));
}
static void r5l_run_no_mem_stripe(struct r5l_log *log)
@@ -769,6 +962,7 @@ static bool r5l_complete_finished_ios(struct r5l_log *log)
static void __r5l_stripe_write_finished(struct r5l_io_unit *io)
{
struct r5l_log *log = io->log;
+ struct r5conf *conf = log->rdev->mddev->private;
unsigned long flags;
spin_lock_irqsave(&log->io_list_lock, flags);
@@ -779,7 +973,8 @@ static void __r5l_stripe_write_finished(struct r5l_io_unit *io)
return;
}
- if (r5l_reclaimable_space(log) > log->max_free_space)
+ if (r5l_reclaimable_space(log) > log->max_free_space ||
+ test_bit(R5C_LOG_TIGHT, &conf->cache_state))
r5l_wake_reclaim(log, 0);
spin_unlock_irqrestore(&log->io_list_lock, flags);
@@ -900,14 +1095,146 @@ static void r5l_write_super_and_discard_space(struct r5l_log *log,
}
}
+/*
+ * r5c_flush_stripe moves stripe from cached list to handle_list. When called,
+ * the stripe must be on r5c_cached_full_stripes or r5c_cached_partial_stripes.
+ *
+ * must hold conf->device_lock
+ */
+static void r5c_flush_stripe(struct r5conf *conf, struct stripe_head *sh)
+{
+ BUG_ON(list_empty(&sh->lru));
+ BUG_ON(!test_bit(STRIPE_R5C_CACHING, &sh->state));
+ BUG_ON(test_bit(STRIPE_HANDLE, &sh->state));
+
+ /*
+ * The stripe is not ON_RELEASE_LIST, so it is safe to call
+ * raid5_release_stripe() while holding conf->device_lock
+ */
+ BUG_ON(test_bit(STRIPE_ON_RELEASE_LIST, &sh->state));
+ assert_spin_locked(&conf->device_lock);
+
+ list_del_init(&sh->lru);
+ atomic_inc(&sh->count);
+
+ set_bit(STRIPE_HANDLE, &sh->state);
+ atomic_inc(&conf->active_stripes);
+ r5c_make_stripe_write_out(sh);
+
+ if (!test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+ atomic_inc(&conf->preread_active_stripes);
+ raid5_release_stripe(sh);
+}
+
+/*
+ * if num == 0, flush all full stripes
+ * if num > 0, flush all full stripes. If less than num full stripes are
+ * flushed, flush some partial stripes until totally num stripes are
+ * flushed or there is no more cached stripes.
+ */
+void r5c_flush_cache(struct r5conf *conf, int num)
+{
+ int count;
+ struct stripe_head *sh, *next;
+
+ assert_spin_locked(&conf->device_lock);
+ if (!conf->log)
+ return;
+
+ count = 0;
+ list_for_each_entry_safe(sh, next, &conf->r5c_full_stripe_list, lru) {
+ r5c_flush_stripe(conf, sh);
+ count++;
+ }
+
+ if (count >= num)
+ return;
+ list_for_each_entry_safe(sh, next,
+ &conf->r5c_partial_stripe_list, lru) {
+ r5c_flush_stripe(conf, sh);
+ if (++count >= num)
+ break;
+ }
+}
+
+static void r5c_do_reclaim(struct r5conf *conf)
+{
+ struct r5l_log *log = conf->log;
+ struct stripe_head *sh;
+ int count = 0;
+ unsigned long flags;
+ int total_cached;
+ int stripes_to_flush;
+
+ if (!r5c_is_writeback(log))
+ return;
+
+ total_cached = atomic_read(&conf->r5c_cached_partial_stripes) +
+ atomic_read(&conf->r5c_cached_full_stripes);
+
+ if (total_cached > conf->min_nr_stripes * 3 / 4 ||
+ atomic_read(&conf->empty_inactive_list_nr) > 0)
+ /*
+ * if stripe cache pressure high, flush all full stripes and
+ * some partial stripes
+ */
+ stripes_to_flush = R5C_RECLAIM_STRIPE_GROUP;
+ else if (total_cached > conf->min_nr_stripes * 1 / 2 ||
+ atomic_read(&conf->r5c_cached_full_stripes) >
+ R5C_FULL_STRIPE_FLUSH_BATCH)
+ /*
+ * if stripe cache pressure moderate, or if there is many full
+ * stripes,flush all full stripes
+ */
+ stripes_to_flush = 0;
+ else
+ /* no need to flush */
+ stripes_to_flush = -1;
+
+ if (stripes_to_flush >= 0) {
+ spin_lock_irqsave(&conf->device_lock, flags);
+ r5c_flush_cache(conf, stripes_to_flush);
+ spin_unlock_irqrestore(&conf->device_lock, flags);
+ }
+
+ /* if log space is tight, flush stripes on stripe_in_journal_list */
+ if (test_bit(R5C_LOG_TIGHT, &conf->cache_state)) {
+ spin_lock_irqsave(&log->stripe_in_journal_lock, flags);
+ spin_lock(&conf->device_lock);
+ list_for_each_entry(sh, &log->stripe_in_journal_list, r5c) {
+ /*
+ * stripes on stripe_in_journal_list could be in any
+ * state of the stripe_cache state machine. In this
+ * case, we only want to flush stripe on
+ * r5c_cached_full/partial_stripes. The following
+ * condition makes sure the stripe is on one of the
+ * two lists.
+ */
+ if (!list_empty(&sh->lru) &&
+ !test_bit(STRIPE_HANDLE, &sh->state) &&
+ atomic_read(&sh->count) == 0) {
+ r5c_flush_stripe(conf, sh);
+ }
+ if (count++ >= R5C_RECLAIM_STRIPE_GROUP)
+ break;
+ }
+ spin_unlock(&conf->device_lock);
+ spin_unlock_irqrestore(&log->stripe_in_journal_lock, flags);
+ }
+ md_wakeup_thread(conf->mddev->thread);
+}
+
static void r5l_do_reclaim(struct r5l_log *log)
{
+ struct r5conf *conf = log->rdev->mddev->private;
sector_t reclaim_target = xchg(&log->reclaim_target, 0);
sector_t reclaimable;
sector_t next_checkpoint;
- u64 next_cp_seq;
+ bool write_super;
spin_lock_irq(&log->io_list_lock);
+ write_super = r5l_reclaimable_space(log) > log->max_free_space ||
+ reclaim_target != 0 || !list_empty(&log->no_space_stripes);
/*
* move proper io_unit to reclaim list. We should not change the order.
* reclaimable/unreclaimable io_unit can be mixed in the list, we
@@ -928,12 +1255,12 @@ static void r5l_do_reclaim(struct r5l_log *log)
log->io_list_lock);
}
- next_checkpoint = log->next_checkpoint;
- next_cp_seq = log->next_cp_seq;
+ next_checkpoint = r5c_calculate_new_cp(conf);
spin_unlock_irq(&log->io_list_lock);
BUG_ON(reclaimable < 0);
- if (reclaimable == 0)
+
+ if (reclaimable == 0 || !write_super)
return;
/*
@@ -945,7 +1272,7 @@ static void r5l_do_reclaim(struct r5l_log *log)
mutex_lock(&log->io_mutex);
log->last_checkpoint = next_checkpoint;
- log->last_cp_seq = next_cp_seq;
+ r5c_update_log_state(log);
mutex_unlock(&log->io_mutex);
r5l_run_no_space_stripes(log);
@@ -959,14 +1286,17 @@ static void r5l_reclaim_thread(struct md_thread *thread)
if (!log)
return;
+ r5c_do_reclaim(conf);
r5l_do_reclaim(log);
}
-static void r5l_wake_reclaim(struct r5l_log *log, sector_t space)
+void r5l_wake_reclaim(struct r5l_log *log, sector_t space)
{
unsigned long target;
unsigned long new = (unsigned long)space; /* overflow in theory */
+ if (!log)
+ return;
do {
target = log->reclaim_target;
if (new < target)
@@ -990,11 +1320,12 @@ void r5l_quiesce(struct r5l_log *log, int state)
return;
log->reclaim_thread = md_register_thread(r5l_reclaim_thread,
log->rdev->mddev, "reclaim");
+ log->reclaim_thread->timeout = R5C_RECLAIM_WAKEUP_INTERVAL;
} else if (state == 1) {
/* make sure r5l_write_super_and_discard_space exits */
mddev = log->rdev->mddev;
wake_up(&mddev->sb_wait);
- r5l_wake_reclaim(log, -1L);
+ r5l_wake_reclaim(log, MaxSector);
md_unregister_thread(&log->reclaim_thread);
r5l_do_reclaim(log);
}
@@ -1415,12 +1746,22 @@ void r5c_finish_stripe_write_out(struct r5conf *conf,
if (do_wakeup)
wake_up(&conf->wait_for_overlap);
+
+ if (conf->log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH)
+ return;
+
+ spin_lock_irq(&conf->log->stripe_in_journal_lock);
+ list_del_init(&sh->r5c);
+ spin_unlock_irq(&conf->log->stripe_in_journal_lock);
+ sh->log_start = MaxSector;
+ atomic_dec(&conf->log->stripe_in_journal_count);
}
int
r5c_cache_data(struct r5l_log *log, struct stripe_head *sh,
struct stripe_head_state *s)
{
+ struct r5conf *conf = sh->raid_conf;
int pages = 0;
int reserve;
int i;
@@ -1451,12 +1792,15 @@ r5c_cache_data(struct r5l_log *log, struct stripe_head *sh,
mutex_lock(&log->io_mutex);
/* meta + data */
reserve = (1 + pages) << (PAGE_SHIFT - 9);
- if (!r5l_has_free_space(log, reserve)) {
- spin_lock(&log->no_space_stripes_lock);
- list_add_tail(&sh->log_list, &log->no_space_stripes);
- spin_unlock(&log->no_space_stripes_lock);
- r5l_wake_reclaim(log, reserve);
+ if (test_bit(R5C_LOG_CRITICAL, &conf->cache_state) &&
+ sh->log_start == MaxSector)
+ r5l_add_no_space_stripe(log, sh);
+ else if (!r5l_has_free_space(log, reserve)) {
+ if (sh->log_start == log->last_checkpoint)
+ BUG();
+ else
+ r5l_add_no_space_stripe(log, sh);
} else {
ret = r5l_log_stripe(log, sh, pages, 0);
if (ret) {
@@ -1470,7 +1814,6 @@ r5c_cache_data(struct r5l_log *log, struct stripe_head *sh,
return 0;
}
-
static int r5l_load_log(struct r5l_log *log)
{
struct md_rdev *rdev = log->rdev;
@@ -1530,6 +1873,9 @@ static int r5l_load_log(struct r5l_log *log)
log->max_free_space = RECLAIM_MAX_FREE_SPACE;
log->last_checkpoint = cp;
log->next_checkpoint = cp;
+ mutex_lock(&log->io_mutex);
+ r5c_update_log_state(log);
+ mutex_unlock(&log->io_mutex);
__free_page(page);
@@ -1601,6 +1947,8 @@ int r5l_init_log(struct r5conf *conf, struct md_rdev *rdev)
log->rdev->mddev, "reclaim");
if (!log->reclaim_thread)
goto reclaim_thread;
+ log->reclaim_thread->timeout = R5C_RECLAIM_WAKEUP_INTERVAL;
+
init_waitqueue_head(&log->iounit_wait);
INIT_LIST_HEAD(&log->no_mem_stripes);
@@ -1609,6 +1957,9 @@ int r5l_init_log(struct r5conf *conf, struct md_rdev *rdev)
spin_lock_init(&log->no_space_stripes_lock);
log->r5c_journal_mode = R5C_JOURNAL_MODE_WRITE_THROUGH;
+ INIT_LIST_HEAD(&log->stripe_in_journal_list);
+ spin_lock_init(&log->stripe_in_journal_lock);
+ atomic_set(&log->stripe_in_journal_count, 0);
if (r5l_load_log(log))
goto error;
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index f535ce2..90638ba 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -228,6 +228,16 @@ static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh,
for (i = sh->disks; i--; )
if (test_bit(R5_InJournal, &sh->dev[i].flags))
injournal++;
+ /*
+ * When quiesce in r5c write back, set STRIPE_HANDLE for stripes with
+ * data in journal, so they are not released to cached lists
+ */
+ if (conf->quiesce && r5c_is_writeback(conf->log) &&
+ !test_bit(STRIPE_HANDLE, &sh->state) && injournal != 0) {
+ if (test_bit(STRIPE_R5C_CACHING, &sh->state))
+ r5c_make_stripe_write_out(sh);
+ set_bit(STRIPE_HANDLE, &sh->state);
+ }
if (test_bit(STRIPE_HANDLE, &sh->state)) {
if (test_bit(STRIPE_DELAYED, &sh->state) &&
@@ -268,6 +278,7 @@ static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh,
if (test_and_clear_bit(STRIPE_R5C_PARTIAL_STRIPE, &sh->state))
atomic_dec(&conf->r5c_cached_partial_stripes);
list_add_tail(&sh->lru, &conf->r5c_full_stripe_list);
+ r5c_check_cached_full_stripe(conf);
} else {
/* partial stripe */
if (!test_and_set_bit(STRIPE_R5C_PARTIAL_STRIPE,
@@ -639,9 +650,12 @@ raid5_get_active_stripe(struct r5conf *conf, sector_t sector,
}
if (noblock && sh == NULL)
break;
+
+ r5c_check_stripe_cache_usage(conf);
if (!sh) {
set_bit(R5_INACTIVE_BLOCKED,
&conf->cache_state);
+ r5l_wake_reclaim(conf->log, 0);
wait_event_lock_irq(
conf->wait_for_stripe,
!list_empty(conf->inactive_list + hash) &&
@@ -1992,7 +2006,9 @@ static struct stripe_head *alloc_stripe(struct kmem_cache *sc, gfp_t gfp,
spin_lock_init(&sh->batch_lock);
INIT_LIST_HEAD(&sh->batch_list);
INIT_LIST_HEAD(&sh->lru);
+ INIT_LIST_HEAD(&sh->r5c);
atomic_set(&sh->count, 1);
+ sh->log_start = MaxSector;
for (i = 0; i < disks; i++) {
struct r5dev *dev = &sh->dev[i];
@@ -4759,6 +4775,10 @@ static int raid5_congested(struct mddev *mddev, int bits)
if (test_bit(R5_INACTIVE_BLOCKED, &conf->cache_state))
return 1;
+
+ /* Also checks whether there is pressure on r5cache log space */
+ if (test_bit(R5C_LOG_TIGHT, &conf->cache_state))
+ return 1;
if (conf->quiesce)
return 1;
if (atomic_read(&conf->empty_inactive_list_nr))
@@ -7661,6 +7681,7 @@ static void raid5_quiesce(struct mddev *mddev, int state)
/* '2' tells resync/reshape to pause so that all
* active stripes can drain
*/
+ r5c_flush_cache(conf, INT_MAX);
conf->quiesce = 2;
wait_event_cmd(conf->wait_for_quiescent,
atomic_read(&conf->active_stripes) == 0 &&
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
index 73c1833..35b4c0f 100644
--- a/drivers/md/raid5.h
+++ b/drivers/md/raid5.h
@@ -226,6 +226,8 @@ struct stripe_head {
struct r5l_io_unit *log_io;
struct list_head log_list;
+ sector_t log_start; /* first meta block on the journal */
+ struct list_head r5c; /* for r5c_cache->stripe_in_journal */
/**
* struct stripe_operations
* @target - STRIPE_OP_COMPUTE_BLK target
@@ -537,6 +539,27 @@ struct r5worker_group {
int stripes_cnt;
};
+enum r5_cache_state {
+ R5_INACTIVE_BLOCKED, /* release of inactive stripes blocked,
+ * waiting for 25% to be free
+ */
+ R5_ALLOC_MORE, /* It might help to allocate another
+ * stripe.
+ */
+ R5_DID_ALLOC, /* A stripe was allocated, don't allocate
+ * more until at least one has been
+ * released. This avoids flooding
+ * the cache.
+ */
+ R5C_LOG_TIGHT, /* log device space tight, need to
+ * prioritize stripes at last_checkpoint
+ */
+ R5C_LOG_CRITICAL, /* log device is running out of space,
+ * only process stripes that are already
+ * occupying the log
+ */
+};
+
struct r5conf {
struct hlist_head *stripe_hashtbl;
/* only protect corresponding hash list and inactive_list */
@@ -636,17 +659,6 @@ struct r5conf {
wait_queue_head_t wait_for_stripe;
wait_queue_head_t wait_for_overlap;
unsigned long cache_state;
-#define R5_INACTIVE_BLOCKED 1 /* release of inactive stripes blocked,
- * waiting for 25% to be free
- */
-#define R5_ALLOC_MORE 2 /* It might help to allocate another
- * stripe.
- */
-#define R5_DID_ALLOC 4 /* A stripe was allocated, don't allocate
- * more until at least one has been
- * released. This avoids flooding
- * the cache.
- */
struct shrinker shrinker;
int pool_size; /* number of disks in stripeheads in pool */
spinlock_t device_lock;
@@ -752,8 +764,13 @@ extern void
r5c_finish_stripe_write_out(struct r5conf *conf, struct stripe_head *sh,
struct stripe_head_state *s);
extern void r5c_release_extra_page(struct stripe_head *sh);
+extern void r5l_wake_reclaim(struct r5l_log *log, sector_t space);
extern void r5c_handle_cached_data_endio(struct r5conf *conf,
struct stripe_head *sh, int disks, struct bio_list *return_bi);
extern int r5c_cache_data(struct r5l_log *log, struct stripe_head *sh,
struct stripe_head_state *s);
+extern void r5c_make_stripe_write_out(struct stripe_head *sh);
+extern void r5c_flush_cache(struct r5conf *conf, int num);
+extern void r5c_check_stripe_cache_usage(struct r5conf *conf);
+extern void r5c_check_cached_full_stripe(struct r5conf *conf);
#endif
--
2.9.3
^ permalink raw reply related
* [PATCH v7 06/10] md/r5cache: sysfs entry journal_mode
From: Song Liu @ 2016-11-17 23:24 UTC (permalink / raw)
To: linux-raid
Cc: neilb, shli, kernel-team, dan.j.williams, hch, liuzhengyuang521,
liuzhengyuan, Song Liu
In-Reply-To: <20161117232445.1798305-1-songliubraving@fb.com>
With write cache, journal_mode is the knob to switch between
write-back and write-through.
Below is an example:
root@virt-test:~/# cat /sys/block/md0/md/journal_mode
[write-through] write-back
root@virt-test:~/# echo write-back > /sys/block/md0/md/journal_mode
root@virt-test:~/# cat /sys/block/md0/md/journal_mode
write-through [write-back]
Signed-off-by: Song Liu <songliubraving@fb.com>
---
drivers/md/raid5-cache.c | 65 ++++++++++++++++++++++++++++++++++++++++++++++++
drivers/md/raid5.c | 1 +
drivers/md/raid5.h | 1 +
3 files changed, 67 insertions(+)
diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c
index 7dec2a0..b9ad0e8 100644
--- a/drivers/md/raid5-cache.c
+++ b/drivers/md/raid5-cache.c
@@ -60,6 +60,8 @@ enum r5c_journal_mode {
R5C_JOURNAL_MODE_WRITE_BACK = 1,
};
+static char *r5c_journal_mode_str[] = {"write-through",
+ "write-back"};
/*
* raid5 cache state machine
*
@@ -1617,6 +1619,69 @@ static void r5l_write_super(struct r5l_log *log, sector_t cp)
set_bit(MD_CHANGE_DEVS, &mddev->flags);
}
+static ssize_t r5c_journal_mode_show(struct mddev *mddev, char *page)
+{
+ struct r5conf *conf = mddev->private;
+ int ret;
+
+ if (!conf->log)
+ return 0;
+
+ switch (conf->log->r5c_journal_mode) {
+ case R5C_JOURNAL_MODE_WRITE_THROUGH:
+ ret = snprintf(
+ page, PAGE_SIZE, "[%s] %s\n",
+ r5c_journal_mode_str[R5C_JOURNAL_MODE_WRITE_THROUGH],
+ r5c_journal_mode_str[R5C_JOURNAL_MODE_WRITE_BACK]);
+ break;
+ case R5C_JOURNAL_MODE_WRITE_BACK:
+ ret = snprintf(
+ page, PAGE_SIZE, "%s [%s]\n",
+ r5c_journal_mode_str[R5C_JOURNAL_MODE_WRITE_THROUGH],
+ r5c_journal_mode_str[R5C_JOURNAL_MODE_WRITE_BACK]);
+ break;
+ default:
+ ret = 0;
+ }
+ return ret;
+}
+
+static ssize_t r5c_journal_mode_store(struct mddev *mddev,
+ const char *page, size_t length)
+{
+ struct r5conf *conf = mddev->private;
+ struct r5l_log *log = conf->log;
+ int val = -1, i;
+ int len = length;
+
+ if (!log)
+ return -ENODEV;
+
+ if (len && page[len - 1] == '\n')
+ len -= 1;
+ for (i = 0; i < ARRAY_SIZE(r5c_journal_mode_str); i++)
+ if (strlen(r5c_journal_mode_str[i]) == len &&
+ strncmp(page, r5c_journal_mode_str[i], len) == 0) {
+ val = i;
+ break;
+ }
+ if (val < R5C_JOURNAL_MODE_WRITE_THROUGH ||
+ val > R5C_JOURNAL_MODE_WRITE_BACK)
+ return -EINVAL;
+
+ mddev_suspend(mddev);
+ conf->log->r5c_journal_mode = val;
+ mddev_resume(mddev);
+
+ pr_debug("md/raid:%s: setting r5c cache mode to %d: %s\n",
+ mdname(mddev), val, r5c_journal_mode_str[val]);
+ return length;
+}
+
+struct md_sysfs_entry
+r5c_journal_mode = __ATTR(journal_mode, 0644,
+ r5c_journal_mode_show, r5c_journal_mode_store);
+
/*
* Try handle write operation in caching phase. This function should only
* be called in write-back mode.
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 90638ba..17cf98e 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -6319,6 +6319,7 @@ static struct attribute *raid5_attrs[] = {
&raid5_group_thread_cnt.attr,
&raid5_skip_copy.attr,
&raid5_rmw_level.attr,
+ &r5c_journal_mode.attr,
NULL,
};
static struct attribute_group raid5_attrs_group = {
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
index 35b4c0f..a698113 100644
--- a/drivers/md/raid5.h
+++ b/drivers/md/raid5.h
@@ -773,4 +773,5 @@ extern void r5c_make_stripe_write_out(struct stripe_head *sh);
extern void r5c_flush_cache(struct r5conf *conf, int num);
extern void r5c_check_stripe_cache_usage(struct r5conf *conf);
extern void r5c_check_cached_full_stripe(struct r5conf *conf);
+extern struct md_sysfs_entry r5c_journal_mode;
#endif
--
2.9.3
^ permalink raw reply related
* [PATCH v7 07/10] md/r5cache: refactoring journal recovery code
From: Song Liu @ 2016-11-17 23:24 UTC (permalink / raw)
To: linux-raid
Cc: neilb, shli, kernel-team, dan.j.williams, hch, liuzhengyuang521,
liuzhengyuan, Song Liu
In-Reply-To: <20161117232445.1798305-1-songliubraving@fb.com>
1. rename r5l_read_meta_block() as r5l_recovery_read_meta_block();
2. pull the code that initialize r5l_meta_block from
r5l_log_write_empty_meta_block() to a separate function
r5l_recovery_create_empty_meta_block(), so that we can reuse this
piece of code.
Signed-off-by: Song Liu <songliubraving@fb.com>
---
drivers/md/raid5-cache.c | 27 ++++++++++++++++++---------
1 file changed, 18 insertions(+), 9 deletions(-)
diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c
index b9ad0e8..390be84 100644
--- a/drivers/md/raid5-cache.c
+++ b/drivers/md/raid5-cache.c
@@ -1356,8 +1356,8 @@ struct r5l_recovery_ctx {
u64 seq; /* recovery position seq */
};
-static int r5l_read_meta_block(struct r5l_log *log,
- struct r5l_recovery_ctx *ctx)
+static int r5l_recovery_read_meta_block(struct r5l_log *log,
+ struct r5l_recovery_ctx *ctx)
{
struct page *page = ctx->meta_page;
struct r5l_meta_block *mb;
@@ -1530,7 +1530,7 @@ static void r5l_recovery_flush_log(struct r5l_log *log,
struct r5l_recovery_ctx *ctx)
{
while (1) {
- if (r5l_read_meta_block(log, ctx))
+ if (r5l_recovery_read_meta_block(log, ctx))
return;
if (r5l_recovery_flush_one_meta(log, ctx))
return;
@@ -1539,17 +1539,16 @@ static void r5l_recovery_flush_log(struct r5l_log *log,
}
}
-static int r5l_log_write_empty_meta_block(struct r5l_log *log, sector_t pos,
- u64 seq)
+static void
+r5l_recovery_create_empty_meta_block(struct r5l_log *log,
+ struct page *page,
+ sector_t pos, u64 seq)
{
- struct page *page;
struct r5l_meta_block *mb;
u32 crc;
- page = alloc_page(GFP_KERNEL | __GFP_ZERO);
- if (!page)
- return -ENOMEM;
mb = page_address(page);
+ clear_page(mb);
mb->magic = cpu_to_le32(R5LOG_MAGIC);
mb->version = R5LOG_VERSION;
mb->meta_size = cpu_to_le32(sizeof(struct r5l_meta_block));
@@ -1557,7 +1556,17 @@ static int r5l_log_write_empty_meta_block(struct r5l_log *log, sector_t pos,
mb->position = cpu_to_le64(pos);
crc = crc32c_le(log->uuid_checksum, mb, PAGE_SIZE);
mb->checksum = cpu_to_le32(crc);
+}
+static int r5l_log_write_empty_meta_block(struct r5l_log *log, sector_t pos,
+ u64 seq)
+{
+ struct page *page;
+
+ page = alloc_page(GFP_KERNEL);
+ if (!page)
+ return -ENOMEM;
+ r5l_recovery_create_empty_meta_block(log, page, pos, seq);
if (!sync_page_io(log->rdev, pos, PAGE_SIZE, page, REQ_OP_WRITE,
WRITE_FUA, false)) {
__free_page(page);
--
2.9.3
^ permalink raw reply related
* [PATCH v7 08/10] md/r5cache: r5cache recovery: part 1
From: Song Liu @ 2016-11-17 23:24 UTC (permalink / raw)
To: linux-raid
Cc: neilb, shli, kernel-team, dan.j.williams, hch, liuzhengyuang521,
liuzhengyuan, Song Liu
In-Reply-To: <20161117232445.1798305-1-songliubraving@fb.com>
Recovery of write-back cache has different logic to write-through only
cache. Specifically, for write-back cache, the recovery need to scan
through all active journal entries before flushing data out. Therefore,
large portion of the recovery logic is rewritten here.
To make the diffs cleaner, we split the rewrite as follows:
1. In this patch, we:
- add new data to r5l_recovery_ctx
- add new functions to recovery write-back cache
The new functions are not used in this patch, so this patch does not
change the behavior of recovery.
2. In next patch, we:
- modify main recovery procedure r5l_recovery_log() to call new
functions
- remove old functions
With cache feature, there are 2 different scenarios of recovery:
1. Data-Parity stripe: a stripe with complete parity in journal.
2. Data-Only stripe: a stripe with only data in journal (or partial
parity).
The code differentiate Data-Parity stripe from Data-Only stripe with
flag STRIPE_R5C_CACHING.
For Data-Parity stripes, we use the same procedure as raid5 journal,
where all the data and parity are replayed to the RAID devices.
For Data-Only strips, we need to finish complete calculate parity and
finish the full reconstruct write or RMW write. For simplicity, in
the recovery, we load the stripe to stripe cache. Once the array is
started, the stripe cache state machine will handle these stripes
through normal write path.
r5c_recovery_flush_log contains the main procedure of recovery. The
recovery code first scans through the journal and loads data to
stripe cache. The code keeps tracks of all these stripes in a list
(use sh->lru and ctx->cached_list), stripes in the list are
organized in the order of its first appearance on the journal.
During the scan, the recovery code assesses each stripe as
Data-Parity or Data-Only.
During scan, the array may run out of stripe cache. In these cases,
the recovery code will also call raid5_set_cache_size to increase
stripe cache size. If the array still runs out of stripe cache
because there isn't enough memory, the array will not assemble.
At the end of scan, the recovery code replays all Data-Parity
stripes, and sets proper states for Data-Only stripes. The recovery
code also increases seq number by 10 and rewrites all Data-Only
stripes to journal. This is to avoid confusion after repeated
crashes. More details is explained in raid5-cache.c before
r5c_recovery_rewrite_data_only_stripes().
Signed-off-by: Song Liu <songliubraving@fb.com>
---
drivers/md/raid5-cache.c | 602 +++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 602 insertions(+)
diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c
index 390be84..3aa7ecc 100644
--- a/drivers/md/raid5-cache.c
+++ b/drivers/md/raid5-cache.c
@@ -1,5 +1,6 @@
/*
* Copyright (C) 2015 Shaohua Li <shli@fb.com>
+ * Copyright (C) 2016 Song Liu <songliubraving@fb.com>
*
* This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License,
@@ -1354,6 +1355,9 @@ struct r5l_recovery_ctx {
sector_t meta_total_blocks; /* total size of current meta and data */
sector_t pos; /* recovery position */
u64 seq; /* recovery position seq */
+ int data_parity_stripes; /* number of data_parity stripes */
+ int data_only_stripes; /* number of data_only stripes */
+ struct list_head cached_list;
};
static int r5l_recovery_read_meta_block(struct r5l_log *log,
@@ -1576,6 +1580,590 @@ static int r5l_log_write_empty_meta_block(struct r5l_log *log, sector_t pos,
return 0;
}
+/*
+ * r5l_recovery_load_data and r5l_recovery_load_parity uses flag R5_Wantwrite
+ * to mark valid (potentially not flushed) data in the journal.
+ *
+ * We already verified checksum in r5l_recovery_verify_data_checksum_for_mb,
+ * so there should not be any mismatch here.
+ */
+static void r5l_recovery_load_data(struct r5l_log *log,
+ struct stripe_head *sh,
+ struct r5l_recovery_ctx *ctx,
+ struct r5l_payload_data_parity *payload,
+ sector_t log_offset)
+{
+ struct mddev *mddev = log->rdev->mddev;
+ struct r5conf *conf = mddev->private;
+ int dd_idx;
+
+ raid5_compute_sector(conf,
+ le64_to_cpu(payload->location), 0,
+ &dd_idx, sh);
+ sync_page_io(log->rdev, log_offset, PAGE_SIZE,
+ sh->dev[dd_idx].page, REQ_OP_READ, 0, false);
+ sh->dev[dd_idx].log_checksum =
+ le32_to_cpu(payload->checksum[0]);
+ ctx->meta_total_blocks += BLOCK_SECTORS;
+
+ set_bit(R5_Wantwrite, &sh->dev[dd_idx].flags);
+ set_bit(STRIPE_R5C_CACHING, &sh->state);
+}
+
+static void r5l_recovery_load_parity(struct r5l_log *log,
+ struct stripe_head *sh,
+ struct r5l_recovery_ctx *ctx,
+ struct r5l_payload_data_parity *payload,
+ sector_t log_offset)
+{
+ struct mddev *mddev = log->rdev->mddev;
+ struct r5conf *conf = mddev->private;
+
+ ctx->meta_total_blocks += BLOCK_SECTORS * conf->max_degraded;
+ sync_page_io(log->rdev, log_offset, PAGE_SIZE,
+ sh->dev[sh->pd_idx].page, REQ_OP_READ, 0, false);
+ sh->dev[sh->pd_idx].log_checksum =
+ le32_to_cpu(payload->checksum[0]);
+ set_bit(R5_Wantwrite, &sh->dev[sh->pd_idx].flags);
+
+ if (sh->qd_idx >= 0) {
+ sync_page_io(log->rdev,
+ r5l_ring_add(log, log_offset, BLOCK_SECTORS),
+ PAGE_SIZE, sh->dev[sh->qd_idx].page,
+ REQ_OP_READ, 0, false);
+ sh->dev[sh->qd_idx].log_checksum =
+ le32_to_cpu(payload->checksum[1]);
+ set_bit(R5_Wantwrite, &sh->dev[sh->qd_idx].flags);
+ }
+ clear_bit(STRIPE_R5C_CACHING, &sh->state);
+}
+
+static void r5l_recovery_reset_stripe(struct stripe_head *sh)
+{
+ int i;
+
+ sh->state = 0;
+ sh->log_start = MaxSector;
+ for (i = sh->disks; i--; )
+ sh->dev[i].flags = 0;
+}
+
+static void
+r5l_recovery_replay_one_stripe(struct r5conf *conf,
+ struct stripe_head *sh,
+ struct r5l_recovery_ctx *ctx)
+{
+ struct md_rdev *rdev, *rrdev;
+ int disk_index;
+ int data_count = 0;
+
+ for (disk_index = 0; disk_index < sh->disks; disk_index++) {
+ if (!test_bit(R5_Wantwrite, &sh->dev[disk_index].flags))
+ continue;
+ if (disk_index == sh->qd_idx || disk_index == sh->pd_idx)
+ continue;
+ data_count++;
+ }
+
+ /*
+ * stripes that only have parity must have been flushed
+ * before the crash that we are now recovering from, so
+ * there is nothing more to recovery.
+ */
+ if (data_count == 0)
+ goto out;
+
+ for (disk_index = 0; disk_index < sh->disks; disk_index++) {
+ if (!test_bit(R5_Wantwrite, &sh->dev[disk_index].flags))
+ continue;
+
+ /* in case device is broken */
+ rcu_read_lock();
+ rdev = rcu_dereference(conf->disks[disk_index].rdev);
+ if (rdev) {
+ atomic_inc(&rdev->nr_pending);
+ rcu_read_unlock();
+ sync_page_io(rdev, sh->sector, PAGE_SIZE,
+ sh->dev[disk_index].page, REQ_OP_WRITE, 0,
+ false);
+ rdev_dec_pending(rdev, rdev->mddev);
+ rcu_read_lock();
+ }
+ rrdev = rcu_dereference(conf->disks[disk_index].replacement);
+ if (rrdev) {
+ atomic_inc(&rrdev->nr_pending);
+ rcu_read_unlock();
+ sync_page_io(rrdev, sh->sector, PAGE_SIZE,
+ sh->dev[disk_index].page, REQ_OP_WRITE, 0,
+ false);
+ rdev_dec_pending(rrdev, rrdev->mddev);
+ rcu_read_lock();
+ }
+ rcu_read_unlock();
+ }
+ ctx->data_parity_stripes++;
+out:
+ r5l_recovery_reset_stripe(sh);
+}
+
+static struct stripe_head *
+r5c_recovery_alloc_stripe(struct r5conf *conf,
+ struct list_head *recovery_list,
+ sector_t stripe_sect,
+ sector_t log_start)
+{
+ struct stripe_head *sh;
+
+ sh = raid5_get_active_stripe(conf, stripe_sect, 0, 1, 0);
+ if (!sh)
+ return NULL; /* no more stripe available */
+
+ r5l_recovery_reset_stripe(sh);
+ sh->log_start = log_start;
+
+ return sh;
+}
+
+static struct stripe_head *
+r5c_recovery_lookup_stripe(struct list_head *list, sector_t sect)
+{
+ struct stripe_head *sh;
+
+ list_for_each_entry(sh, list, lru)
+ if (sh->sector == sect)
+ return sh;
+ return NULL;
+}
+
+static void
+r5c_recovery_drop_stripes(struct list_head *cached_stripe_list,
+ struct r5l_recovery_ctx *ctx)
+{
+ struct stripe_head *sh, *next;
+
+ list_for_each_entry_safe(sh, next, cached_stripe_list, lru) {
+ r5l_recovery_reset_stripe(sh);
+ list_del_init(&sh->lru);
+ raid5_release_stripe(sh);
+ }
+}
+
+static void
+r5c_recovery_replay_stripes(struct list_head *cached_stripe_list,
+ struct r5l_recovery_ctx *ctx)
+{
+ struct stripe_head *sh, *next;
+
+ list_for_each_entry_safe(sh, next, cached_stripe_list, lru)
+ if (!test_bit(STRIPE_R5C_CACHING, &sh->state)) {
+ r5l_recovery_replay_one_stripe(sh->raid_conf, sh, ctx);
+ list_del_init(&sh->lru);
+ raid5_release_stripe(sh);
+ }
+}
+
+/* if matches return 0; otherwise return -EINVAL */
+static int
+r5l_recovery_verify_data_checksum(struct r5l_log *log, struct page *page,
+ sector_t log_offset, __le32 log_checksum)
+{
+ void *addr;
+ u32 checksum;
+
+ sync_page_io(log->rdev, log_offset, PAGE_SIZE,
+ page, REQ_OP_READ, 0, false);
+ addr = kmap_atomic(page);
+ checksum = crc32c_le(log->uuid_checksum, addr, PAGE_SIZE);
+ kunmap_atomic(addr);
+ return (le32_to_cpu(log_checksum) == checksum) ? 0 : -EINVAL;
+}
+
+/*
+ * before loading data to stripe cache, we need verify checksum for all data,
+ * if there is mismatch for any data page, we drop all data in the mata block
+ */
+static int
+r5l_recovery_verify_data_checksum_for_mb(struct r5l_log *log,
+ struct r5l_recovery_ctx *ctx)
+{
+ struct mddev *mddev = log->rdev->mddev;
+ struct r5conf *conf = mddev->private;
+ struct r5l_meta_block *mb = page_address(ctx->meta_page);
+ sector_t mb_offset = sizeof(struct r5l_meta_block);
+ sector_t log_offset = r5l_ring_add(log, ctx->pos, BLOCK_SECTORS);
+ struct page *page;
+ struct r5l_payload_data_parity *payload;
+
+ page = alloc_page(GFP_KERNEL);
+ if (!page)
+ return -ENOMEM;
+
+ while (mb_offset < le32_to_cpu(mb->meta_size)) {
+ payload = (void *)mb + mb_offset;
+
+ if (payload->header.type == R5LOG_PAYLOAD_DATA) {
+ if (r5l_recovery_verify_data_checksum(
+ log, page, log_offset,
+ payload->checksum[0]) < 0)
+ goto mismatch;
+ } else if (payload->header.type == R5LOG_PAYLOAD_PARITY) {
+ if (r5l_recovery_verify_data_checksum(
+ log, page, log_offset,
+ payload->checksum[0]) < 0)
+ goto mismatch;
+ if (conf->max_degraded == 2 && /* q for RAID 6 */
+ r5l_recovery_verify_data_checksum(
+ log, page,
+ r5l_ring_add(log, log_offset,
+ BLOCK_SECTORS),
+ payload->checksum[1]) < 0)
+ goto mismatch;
+ } else /* not R5LOG_PAYLOAD_DATA or R5LOG_PAYLOAD_PARITY */
+ goto mismatch;
+
+ log_offset = r5l_ring_add(log, log_offset,
+ le32_to_cpu(payload->size));
+
+ mb_offset += sizeof(struct r5l_payload_data_parity) +
+ sizeof(__le32) *
+ (le32_to_cpu(payload->size) >> (PAGE_SHIFT - 9));
+ }
+
+ put_page(page);
+ return 0;
+
+mismatch:
+ put_page(page);
+ return -EINVAL;
+}
+
+/*
+ * Analyze all data/parity pages in one meta block
+ * Returns:
+ * 0 for success
+ * -EINVAL for unknown playload type
+ * -EAGAIN for checksum mismatch of data page
+ * -ENOMEM for run out of memory (alloc_page failed or run out of stripes)
+ */
+static int
+r5c_recovery_analyze_meta_block(struct r5l_log *log,
+ struct r5l_recovery_ctx *ctx,
+ struct list_head *cached_stripe_list)
+{
+ struct mddev *mddev = log->rdev->mddev;
+ struct r5conf *conf = mddev->private;
+ struct r5l_meta_block *mb;
+ struct r5l_payload_data_parity *payload;
+ int mb_offset;
+ sector_t log_offset;
+ sector_t stripe_sect;
+ struct stripe_head *sh;
+ int ret;
+
+ /*
+ * for mismatch in data blocks, we will drop all data in this mb, but
+ * we will still read next mb for other data with FLUSH flag, as
+ * io_unit could finish out of order.
+ */
+ ret = r5l_recovery_verify_data_checksum_for_mb(log, ctx);
+ if (ret == -EINVAL)
+ return -EAGAIN;
+ else if (ret)
+ return ret; /* -ENOMEM duo to alloc_page() failed */
+
+ mb = page_address(ctx->meta_page);
+ mb_offset = sizeof(struct r5l_meta_block);
+ log_offset = r5l_ring_add(log, ctx->pos, BLOCK_SECTORS);
+
+ while (mb_offset < le32_to_cpu(mb->meta_size)) {
+ int dd;
+
+ payload = (void *)mb + mb_offset;
+ stripe_sect = (payload->header.type == R5LOG_PAYLOAD_DATA) ?
+ raid5_compute_sector(
+ conf, le64_to_cpu(payload->location), 0, &dd,
+ NULL)
+ : le64_to_cpu(payload->location);
+
+ sh = r5c_recovery_lookup_stripe(cached_stripe_list,
+ stripe_sect);
+
+ if (!sh) {
+ sh = r5c_recovery_alloc_stripe(conf, cached_stripe_list,
+ stripe_sect, ctx->pos);
+ /*
+ * cannot get stripe from raid5_get_active_stripe
+ * try replay some stripes
+ */
+ if (!sh) {
+ r5c_recovery_replay_stripes(
+ cached_stripe_list, ctx);
+ sh = r5c_recovery_alloc_stripe(
+ conf, cached_stripe_list,
+ stripe_sect, ctx->pos);
+ }
+ if (!sh) {
+ pr_debug("md/raid:%s: Increasing stripe cache size to %d to recovery data on journal.\n",
+ mdname(mddev),
+ conf->min_nr_stripes * 2);
+ raid5_set_cache_size(mddev,
+ conf->min_nr_stripes * 2);
+ sh = r5c_recovery_alloc_stripe(
+ conf, cached_stripe_list, stripe_sect,
+ ctx->pos);
+ }
+ if (!sh) {
+ pr_err("md/raid:%s: Cannot get enough stripes due to memory pressure. Recovery failed.\n",
+ mdname(mddev));
+ return -ENOMEM;
+ }
+ list_add_tail(&sh->lru, cached_stripe_list);
+ }
+
+ if (payload->header.type == R5LOG_PAYLOAD_DATA) {
+ if (!test_bit(STRIPE_R5C_CACHING, &sh->state)) {
+ r5l_recovery_replay_one_stripe(conf, sh, ctx);
+ r5l_recovery_reset_stripe(sh);
+ sh->log_start = ctx->pos;
+ list_move_tail(&sh->lru, cached_stripe_list);
+ }
+ r5l_recovery_load_data(log, sh, ctx, payload,
+ log_offset);
+ } else if (payload->header.type == R5LOG_PAYLOAD_PARITY)
+ r5l_recovery_load_parity(log, sh, ctx, payload,
+ log_offset);
+ else
+ return -EINVAL;
+
+ log_offset = r5l_ring_add(log, log_offset,
+ le32_to_cpu(payload->size));
+
+ mb_offset += sizeof(struct r5l_payload_data_parity) +
+ sizeof(__le32) *
+ (le32_to_cpu(payload->size) >> (PAGE_SHIFT - 9));
+ }
+
+ return 0;
+}
+
+/*
+ * Load the stripe into cache. The stripe will be written out later by
+ * the stripe cache state machine.
+ */
+static void r5c_recovery_load_one_stripe(struct r5l_log *log,
+ struct stripe_head *sh)
+{
+ struct r5conf *conf = sh->raid_conf;
+ struct r5dev *dev;
+ int i;
+
+ for (i = sh->disks; i--; ) {
+ dev = sh->dev + i;
+ if (test_and_clear_bit(R5_Wantwrite, &dev->flags)) {
+ set_bit(R5_InJournal, &dev->flags);
+ set_bit(R5_UPTODATE, &dev->flags);
+ }
+ }
+ set_bit(STRIPE_R5C_PARTIAL_STRIPE, &sh->state);
+ atomic_inc(&conf->r5c_cached_partial_stripes);
+ list_add_tail(&sh->r5c, &log->stripe_in_journal_list);
+}
+
+/*
+ * Scan through the log for all to-be-flushed data
+ *
+ * For stripes with data and parity, namely Data-Parity stripe
+ * (STRIPE_R5C_CACHING == 0), we simply replay all the writes.
+ *
+ * For stripes with only data, namely Data-Only stripe
+ * (STRIPE_R5C_CACHING == 1), we load them to stripe cache state machine.
+ *
+ * For a stripe, if we see data after parity, we should discard all previous
+ * data and parity for this stripe, as these data are already flushed to
+ * the array.
+ *
+ * At the end of the scan, we return the new journal_tail, which points to
+ * first data-only stripe on the journal device, or next invalid meta block.
+ */
+static int r5c_recovery_flush_log(struct r5l_log *log,
+ struct r5l_recovery_ctx *ctx)
+{
+ struct stripe_head *sh, *next;
+ int ret = 0;
+
+ /* scan through the log */
+ while (1) {
+ if (r5l_recovery_read_meta_block(log, ctx))
+ break;
+
+ ret = r5c_recovery_analyze_meta_block(log, ctx,
+ &ctx->cached_list);
+ /*
+ * -EAGAIN means mismatch in data block, in this case, we still
+ * try scan the next metablock
+ */
+ if (ret && ret != -EAGAIN)
+ break; /* ret == -EINVAL or -ENOMEM */
+ ctx->seq++;
+ ctx->pos = r5l_ring_add(log, ctx->pos, ctx->meta_total_blocks);
+ }
+
+ if (ret == -ENOMEM) {
+ r5c_recovery_drop_stripes(&ctx->cached_list, ctx);
+ return ret;
+ }
+
+ /* replay data-parity stripes */
+ r5c_recovery_replay_stripes(&ctx->cached_list, ctx);
+
+ /* load data-only stripes to stripe cache */
+ list_for_each_entry_safe(sh, next, &ctx->cached_list, lru) {
+ WARN_ON(!test_bit(STRIPE_R5C_CACHING, &sh->state));
+ r5c_recovery_load_one_stripe(log, sh);
+ list_del_init(&sh->lru);
+ raid5_release_stripe(sh);
+ ctx->data_only_stripes++;
+ }
+
+ return 0;
+}
+
+/*
+ * we did a recovery. Now ctx.pos points to an invalid meta block. New
+ * log will start here. but we can't let superblock point to last valid
+ * meta block. The log might looks like:
+ * | meta 1| meta 2| meta 3|
+ * meta 1 is valid, meta 2 is invalid. meta 3 could be valid. If
+ * superblock points to meta 1, we write a new valid meta 2n. if crash
+ * happens again, new recovery will start from meta 1. Since meta 2n is
+ * valid now, recovery will think meta 3 is valid, which is wrong.
+ * The solution is we create a new meta in meta2 with its seq == meta
+ * 1's seq + 10 and let superblock points to meta2. The same recovery will
+ * not think meta 3 is a valid meta, because its seq doesn't match
+ */
+
+/*
+ * Before recovery, the log looks like the following
+ *
+ * ---------------------------------------------
+ * | valid log | invalid log |
+ * ---------------------------------------------
+ * ^
+ * |- log->last_checkpoint
+ * |- log->last_cp_seq
+ *
+ * Now we scan through the log until we see invalid entry
+ *
+ * ---------------------------------------------
+ * | valid log | invalid log |
+ * ---------------------------------------------
+ * ^ ^
+ * |- log->last_checkpoint |- ctx->pos
+ * |- log->last_cp_seq |- ctx->seq
+ *
+ * From this point, we need to increase seq number by 10 to avoid
+ * confusing next recovery.
+ *
+ * ---------------------------------------------
+ * | valid log | invalid log |
+ * ---------------------------------------------
+ * ^ ^
+ * |- log->last_checkpoint |- ctx->pos+1
+ * |- log->last_cp_seq |- ctx->seq+11
+ *
+ * However, it is not safe to start the state machine yet, because data only
+ * parities are not yet secured in RAID. To save these data only parities, we
+ * rewrite them from seq+11.
+ *
+ * -----------------------------------------------------------------
+ * | valid log | data only stripes | invalid log |
+ * -----------------------------------------------------------------
+ * ^ ^
+ * |- log->last_checkpoint |- ctx->pos+n
+ * |- log->last_cp_seq |- ctx->seq+10+n
+ *
+ * If failure happens again during this process, the recovery can safe start
+ * again from log->last_checkpoint.
+ *
+ * Once data only stripes are rewritten to journal, we move log_tail
+ *
+ * -----------------------------------------------------------------
+ * | old log | data only stripes | invalid log |
+ * -----------------------------------------------------------------
+ * ^ ^
+ * |- log->last_checkpoint |- ctx->pos+n
+ * |- log->last_cp_seq |- ctx->seq+10+n
+ *
+ * Then we can safely start the state machine. If failure happens from this
+ * point on, the recovery will start from new log->last_checkpoint.
+ */
+static int
+r5c_recovery_rewrite_data_only_stripes(struct r5l_log *log,
+ struct r5l_recovery_ctx *ctx)
+{
+ struct stripe_head *sh;
+ struct mddev *mddev = log->rdev->mddev;
+ struct page *page;
+
+ page = alloc_page(GFP_KERNEL);
+ if (!page) {
+ pr_err("md/raid:%s: cannot allocate memory to rewrite data only stripes\n",
+ mdname(mddev));
+ return -ENOMEM;
+ }
+
+ ctx->seq += 10;
+ list_for_each_entry(sh, &ctx->cached_list, lru) {
+ struct r5l_meta_block *mb;
+ int i;
+ int offset;
+ sector_t write_pos;
+
+ WARN_ON(!test_bit(STRIPE_R5C_CACHING, &sh->state));
+ r5l_recovery_create_empty_meta_block(log, page,
+ ctx->pos, ctx->seq);
+ mb = page_address(page);
+ offset = le32_to_cpu(mb->meta_size);
+ write_pos = ctx->pos + BLOCK_SECTORS;
+
+ for (i = sh->disks; i--; ) {
+ struct r5dev *dev = &sh->dev[i];
+ struct r5l_payload_data_parity *payload;
+ void *addr;
+
+ if (test_bit(R5_InJournal, &dev->flags)) {
+ payload = (void *)mb + offset;
+ payload->header.type = cpu_to_le16(
+ R5LOG_PAYLOAD_DATA);
+ payload->size = BLOCK_SECTORS;
+ payload->location = cpu_to_le64(
+ raid5_compute_blocknr(sh, i, 0));
+ addr = kmap_atomic(dev->page);
+ payload->checksum[0] = cpu_to_le32(
+ crc32c_le(log->uuid_checksum, addr,
+ PAGE_SIZE));
+ kunmap_atomic(addr);
+ sync_page_io(log->rdev, write_pos, PAGE_SIZE,
+ dev->page, REQ_OP_WRITE, 0, false);
+ write_pos = r5l_ring_add(log, write_pos,
+ BLOCK_SECTORS);
+ offset += sizeof(__le32) +
+ sizeof(struct r5l_payload_data_parity);
+
+ }
+ }
+ mb->meta_size = cpu_to_le32(offset);
+ mb->checksum = crc32c_le(log->uuid_checksum, mb, PAGE_SIZE);
+ sync_page_io(log->rdev, ctx->pos, PAGE_SIZE, page,
+ REQ_OP_WRITE, WRITE_FUA, false);
+ sh->log_start = ctx->pos;
+ ctx->pos = write_pos;
+ ctx->seq += 1;
+ }
+ __free_page(page);
+ return 0;
+}
+
static int r5l_recovery_log(struct r5l_log *log)
{
struct r5l_recovery_ctx ctx;
@@ -1583,6 +2171,10 @@ static int r5l_recovery_log(struct r5l_log *log)
ctx.pos = log->last_checkpoint;
ctx.seq = log->last_cp_seq;
ctx.meta_page = alloc_page(GFP_KERNEL);
+ ctx.data_only_stripes = 0;
+ ctx.data_parity_stripes = 0;
+ INIT_LIST_HEAD(&ctx.cached_list);
+
if (!ctx.meta_page)
return -ENOMEM;
@@ -1617,6 +2209,16 @@ static int r5l_recovery_log(struct r5l_log *log)
log->log_start = ctx.pos;
log->seq = ctx.seq;
}
+
+ /*
+ * This is to suppress "function defined but not used" warning.
+ * It will be removed when the two functions are used (next patch).
+ */
+ if (!log) {
+ r5c_recovery_flush_log(log, &ctx);
+ r5c_recovery_rewrite_data_only_stripes(log, &ctx);
+ }
+
return 0;
}
--
2.9.3
^ permalink raw reply related
* [PATCH v7 09/10] md/r5cache: r5cache recovery: part 2
From: Song Liu @ 2016-11-17 23:24 UTC (permalink / raw)
To: linux-raid
Cc: neilb, shli, kernel-team, dan.j.williams, hch, liuzhengyuang521,
liuzhengyuan, Song Liu
In-Reply-To: <20161117232445.1798305-1-songliubraving@fb.com>
1. In previous patch, we:
- add new data to r5l_recovery_ctx
- add new functions to recovery write-back cache
The new functions are not used in this patch, so this patch does not
change the behavior of recovery.
2. In this patchpatch, we:
- modify main recovery procedure r5l_recovery_log() to call new
functions
- remove old functions
Signed-off-by: Song Liu <songliubraving@fb.com>
---
drivers/md/raid5-cache.c | 210 ++++++-----------------------------------------
drivers/md/raid5.c | 3 +-
2 files changed, 26 insertions(+), 187 deletions(-)
diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c
index 3aa7ecc..6b99570 100644
--- a/drivers/md/raid5-cache.c
+++ b/drivers/md/raid5-cache.c
@@ -1393,156 +1393,6 @@ static int r5l_recovery_read_meta_block(struct r5l_log *log,
return 0;
}
-static int r5l_recovery_flush_one_stripe(struct r5l_log *log,
- struct r5l_recovery_ctx *ctx,
- sector_t stripe_sect,
- int *offset)
-{
- struct r5conf *conf = log->rdev->mddev->private;
- struct stripe_head *sh;
- struct r5l_payload_data_parity *payload;
- int disk_index;
-
- sh = raid5_get_active_stripe(conf, stripe_sect, 0, 0, 0);
- while (1) {
- sector_t log_offset = r5l_ring_add(log, ctx->pos,
- ctx->meta_total_blocks);
- payload = page_address(ctx->meta_page) + *offset;
-
- if (le16_to_cpu(payload->header.type) == R5LOG_PAYLOAD_DATA) {
- raid5_compute_sector(conf,
- le64_to_cpu(payload->location), 0,
- &disk_index, sh);
-
- sync_page_io(log->rdev, log_offset, PAGE_SIZE,
- sh->dev[disk_index].page, REQ_OP_READ, 0,
- false);
- sh->dev[disk_index].log_checksum =
- le32_to_cpu(payload->checksum[0]);
- set_bit(R5_Wantwrite, &sh->dev[disk_index].flags);
- } else {
- disk_index = sh->pd_idx;
- sync_page_io(log->rdev, log_offset, PAGE_SIZE,
- sh->dev[disk_index].page, REQ_OP_READ, 0,
- false);
- sh->dev[disk_index].log_checksum =
- le32_to_cpu(payload->checksum[0]);
- set_bit(R5_Wantwrite, &sh->dev[disk_index].flags);
-
- if (sh->qd_idx >= 0) {
- disk_index = sh->qd_idx;
- sync_page_io(log->rdev,
- r5l_ring_add(log, log_offset, BLOCK_SECTORS),
- PAGE_SIZE, sh->dev[disk_index].page,
- REQ_OP_READ, 0, false);
- sh->dev[disk_index].log_checksum =
- le32_to_cpu(payload->checksum[1]);
- set_bit(R5_Wantwrite,
- &sh->dev[disk_index].flags);
- }
- }
-
- ctx->meta_total_blocks += le32_to_cpu(payload->size);
- *offset += sizeof(struct r5l_payload_data_parity) +
- sizeof(__le32) *
- (le32_to_cpu(payload->size) >> (PAGE_SHIFT - 9));
- if (le16_to_cpu(payload->header.type) == R5LOG_PAYLOAD_PARITY)
- break;
- }
-
- for (disk_index = 0; disk_index < sh->disks; disk_index++) {
- void *addr;
- u32 checksum;
-
- if (!test_bit(R5_Wantwrite, &sh->dev[disk_index].flags))
- continue;
- addr = kmap_atomic(sh->dev[disk_index].page);
- checksum = crc32c_le(log->uuid_checksum, addr, PAGE_SIZE);
- kunmap_atomic(addr);
- if (checksum != sh->dev[disk_index].log_checksum)
- goto error;
- }
-
- for (disk_index = 0; disk_index < sh->disks; disk_index++) {
- struct md_rdev *rdev, *rrdev;
-
- if (!test_and_clear_bit(R5_Wantwrite,
- &sh->dev[disk_index].flags))
- continue;
-
- /* in case device is broken */
- rcu_read_lock();
- rdev = rcu_dereference(conf->disks[disk_index].rdev);
- if (rdev) {
- atomic_inc(&rdev->nr_pending);
- rcu_read_unlock();
- sync_page_io(rdev, stripe_sect, PAGE_SIZE,
- sh->dev[disk_index].page, REQ_OP_WRITE, 0,
- false);
- rdev_dec_pending(rdev, rdev->mddev);
- rcu_read_lock();
- }
- rrdev = rcu_dereference(conf->disks[disk_index].replacement);
- if (rrdev) {
- atomic_inc(&rrdev->nr_pending);
- rcu_read_unlock();
- sync_page_io(rrdev, stripe_sect, PAGE_SIZE,
- sh->dev[disk_index].page, REQ_OP_WRITE, 0,
- false);
- rdev_dec_pending(rrdev, rrdev->mddev);
- rcu_read_lock();
- }
- rcu_read_unlock();
- }
- raid5_release_stripe(sh);
- return 0;
-
-error:
- for (disk_index = 0; disk_index < sh->disks; disk_index++)
- sh->dev[disk_index].flags = 0;
- raid5_release_stripe(sh);
- return -EINVAL;
-}
-
-static int r5l_recovery_flush_one_meta(struct r5l_log *log,
- struct r5l_recovery_ctx *ctx)
-{
- struct r5conf *conf = log->rdev->mddev->private;
- struct r5l_payload_data_parity *payload;
- struct r5l_meta_block *mb;
- int offset;
- sector_t stripe_sector;
-
- mb = page_address(ctx->meta_page);
- offset = sizeof(struct r5l_meta_block);
-
- while (offset < le32_to_cpu(mb->meta_size)) {
- int dd;
-
- payload = (void *)mb + offset;
- stripe_sector = raid5_compute_sector(conf,
- le64_to_cpu(payload->location), 0, &dd, NULL);
- if (r5l_recovery_flush_one_stripe(log, ctx, stripe_sector,
- &offset))
- return -EINVAL;
- }
- return 0;
-}
-
-/* copy data/parity from log to raid disks */
-static void r5l_recovery_flush_log(struct r5l_log *log,
- struct r5l_recovery_ctx *ctx)
-{
- while (1) {
- if (r5l_recovery_read_meta_block(log, ctx))
- return;
- if (r5l_recovery_flush_one_meta(log, ctx))
- return;
- ctx->seq++;
- ctx->pos = r5l_ring_add(log, ctx->pos, ctx->meta_total_blocks);
- }
-}
-
static void
r5l_recovery_create_empty_meta_block(struct r5l_log *log,
struct page *page,
@@ -2166,7 +2016,9 @@ r5c_recovery_rewrite_data_only_stripes(struct r5l_log *log,
static int r5l_recovery_log(struct r5l_log *log)
{
+ struct mddev *mddev = log->rdev->mddev;
struct r5l_recovery_ctx ctx;
+ int ret;
ctx.pos = log->last_checkpoint;
ctx.seq = log->last_cp_seq;
@@ -2178,47 +2030,33 @@ static int r5l_recovery_log(struct r5l_log *log)
if (!ctx.meta_page)
return -ENOMEM;
- r5l_recovery_flush_log(log, &ctx);
+ ret = r5c_recovery_flush_log(log, &ctx);
__free_page(ctx.meta_page);
- /*
- * we did a recovery. Now ctx.pos points to an invalid meta block. New
- * log will start here. but we can't let superblock point to last valid
- * meta block. The log might looks like:
- * | meta 1| meta 2| meta 3|
- * meta 1 is valid, meta 2 is invalid. meta 3 could be valid. If
- * superblock points to meta 1, we write a new valid meta 2n. if crash
- * happens again, new recovery will start from meta 1. Since meta 2n is
- * valid now, recovery will think meta 3 is valid, which is wrong.
- * The solution is we create a new meta in meta2 with its seq == meta
- * 1's seq + 10 and let superblock points to meta2. The same recovery will
- * not think meta 3 is a valid meta, because its seq doesn't match
- */
- if (ctx.seq > log->last_cp_seq) {
- int ret;
-
- ret = r5l_log_write_empty_meta_block(log, ctx.pos, ctx.seq + 10);
- if (ret)
- return ret;
- log->seq = ctx.seq + 11;
- log->log_start = r5l_ring_add(log, ctx.pos, BLOCK_SECTORS);
- r5l_write_super(log, ctx.pos);
- log->last_checkpoint = ctx.pos;
- log->next_checkpoint = ctx.pos;
- } else {
- log->log_start = ctx.pos;
- log->seq = ctx.seq;
- }
+ if (ret)
+ return ret;
- /*
- * This is to suppress "function defined but not used" warning.
- * It will be removed when the two functions are used (next patch).
- */
- if (!log) {
- r5c_recovery_flush_log(log, &ctx);
- r5c_recovery_rewrite_data_only_stripes(log, &ctx);
+ if ((ctx.data_only_stripes == 0) && (ctx.data_parity_stripes == 0))
+ pr_debug("md/raid:%s: starting from clean shutdown\n",
+ mdname(mddev));
+ else {
+ pr_debug("md/raid:%s: recoverying %d data-only stripes and %d data-parity stripes\n",
+ mdname(mddev), ctx.data_only_stripes,
+ ctx.data_parity_stripes);
+
+ if (ctx.data_only_stripes > 0)
+ if (r5c_recovery_rewrite_data_only_stripes(log, &ctx)) {
+ pr_err("md/raid:%s: failed to rewrite stripes to journal\n",
+ mdname(mddev));
+ return -EIO;
+ }
}
+ log->log_start = ctx.pos;
+ log->next_checkpoint = ctx.pos;
+ log->seq = ctx.seq;
+ r5l_log_write_empty_meta_block(log, ctx.pos, ctx.seq);
+ r5l_write_super(log, ctx.pos);
return 0;
}
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 17cf98e..aa4968c 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -7100,7 +7100,8 @@ static int raid5_run(struct mddev *mddev)
pr_debug("md/raid:%s: using device %s as journal\n",
mdname(mddev), bdevname(journal_dev->bdev, b));
- r5l_init_log(conf, journal_dev);
+ if (r5l_init_log(conf, journal_dev))
+ goto abort;
}
return 0;
--
2.9.3
^ permalink raw reply related
* [PATCH v7 10/10] md/r5cache: handle SYNC and FUA
From: Song Liu @ 2016-11-17 23:24 UTC (permalink / raw)
To: linux-raid
Cc: neilb, shli, kernel-team, dan.j.williams, hch, liuzhengyuang521,
liuzhengyuan, Song Liu
In-Reply-To: <20161117232445.1798305-1-songliubraving@fb.com>
With raid5 cache, we committing data from journal device. When
there is flush request, we need to flush journal device's cache.
This was not needed in raid5 journal, because we will flush the
journal before committing data to raid disks.
This is similar to FUA, except that we also need flush journal for
FUA. Otherwise, corruptions in earlier meta data will stop recovery
from reaching FUA data.
Signed-off-by: Song Liu <songliubraving@fb.com>
---
drivers/md/raid5-cache.c | 162 +++++++++++++++++++++++++++++++++++++++++------
drivers/md/raid5.c | 8 +++
drivers/md/raid5.h | 1 +
3 files changed, 153 insertions(+), 18 deletions(-)
diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c
index 6b99570..308f06f 100644
--- a/drivers/md/raid5-cache.c
+++ b/drivers/md/raid5-cache.c
@@ -19,6 +19,7 @@
#include <linux/raid/md_p.h>
#include <linux/crc32c.h>
#include <linux/random.h>
+#include <trace/events/block.h>
#include "md.h"
#include "raid5.h"
#include "bitmap.h"
@@ -159,6 +160,9 @@ struct r5l_log {
spinlock_t stripe_in_journal_lock;
atomic_t stripe_in_journal_count;
+
+ /* to submit async io_units, to fulfill ordering of flush */
+ struct work_struct deferred_io_work;
};
/*
@@ -185,6 +189,18 @@ struct r5l_io_unit {
int state;
bool need_split_bio;
+ struct bio *split_bio;
+
+ unsigned int has_flush:1; /* include flush request */
+ unsigned int has_fua:1; /* include fua request */
+ unsigned int has_null_flush:1; /* include empty flush request */
+ /*
+ * io isn't sent yet, flush/fua request can only be submitted till it's
+ * the first IO in running_ios list
+ */
+ unsigned int io_deferred:1;
+
+ struct bio_list flush_barriers; /* size == 0 flush bios */
};
/* r5l_io_unit state */
@@ -494,9 +510,11 @@ static void r5l_move_to_end_ios(struct r5l_log *log)
}
}
+static void __r5l_stripe_write_finished(struct r5l_io_unit *io);
static void r5l_log_endio(struct bio *bio)
{
struct r5l_io_unit *io = bio->bi_private;
+ struct r5l_io_unit *io_deferred;
struct r5l_log *log = io->log;
unsigned long flags;
@@ -512,18 +530,89 @@ static void r5l_log_endio(struct bio *bio)
r5l_move_to_end_ios(log);
else
r5l_log_run_stripes(log);
+ if (!list_empty(&log->running_ios)) {
+ /*
+ * FLUSH/FUA io_unit is deferred because of ordering, now we
+ * can dispatch it
+ */
+ io_deferred = list_first_entry(&log->running_ios,
+ struct r5l_io_unit, log_sibling);
+ if (io_deferred->io_deferred)
+ schedule_work(&log->deferred_io_work);
+ }
+
spin_unlock_irqrestore(&log->io_list_lock, flags);
if (log->need_cache_flush)
md_wakeup_thread(log->rdev->mddev->thread);
+
+ if (io->has_null_flush) {
+ struct bio *bi;
+
+ WARN_ON(bio_list_empty(&io->flush_barriers));
+ while ((bi = bio_list_pop(&io->flush_barriers)) != NULL) {
+ bio_endio(bi);
+ atomic_dec(&io->pending_stripe);
+ }
+ if (atomic_read(&io->pending_stripe) == 0)
+ __r5l_stripe_write_finished(io);
+ }
+}
+
+static void r5l_do_submit_io(struct r5l_log *log, struct r5l_io_unit *io)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&log->io_list_lock, flags);
+ __r5l_set_io_unit_state(io, IO_UNIT_IO_START);
+ spin_unlock_irqrestore(&log->io_list_lock, flags);
+
+ if (io->has_flush)
+ bio_set_op_attrs(io->current_bio, REQ_OP_WRITE, WRITE_FLUSH);
+ if (io->has_fua)
+ bio_set_op_attrs(io->current_bio, REQ_OP_WRITE, WRITE_FUA);
+ submit_bio(io->current_bio);
+
+ if (!io->split_bio)
+ return;
+
+ if (io->has_flush)
+ bio_set_op_attrs(io->split_bio, REQ_OP_WRITE, WRITE_FLUSH);
+ if (io->has_fua)
+ bio_set_op_attrs(io->split_bio, REQ_OP_WRITE, WRITE_FUA);
+ submit_bio(io->split_bio);
+}
+
+/* deferred io_unit will be dispatched here */
+static void r5l_submit_io_async(struct work_struct *work)
+{
+ struct r5l_log *log = container_of(work, struct r5l_log,
+ deferred_io_work);
+ struct r5l_io_unit *io = NULL;
+ unsigned long flags;
+
+ spin_lock_irqsave(&log->io_list_lock, flags);
+ if (!list_empty(&log->running_ios)) {
+ io = list_first_entry(&log->running_ios, struct r5l_io_unit,
+ log_sibling);
+ if (!io->io_deferred)
+ io = NULL;
+ else
+ io->io_deferred = 0;
+ }
+ spin_unlock_irqrestore(&log->io_list_lock, flags);
+ if (io)
+ r5l_do_submit_io(log, io);
}
static void r5l_submit_current_io(struct r5l_log *log)
{
struct r5l_io_unit *io = log->current_io;
+ struct bio *bio;
struct r5l_meta_block *block;
unsigned long flags;
u32 crc;
+ bool do_submit = true;
if (!io)
return;
@@ -532,13 +621,20 @@ static void r5l_submit_current_io(struct r5l_log *log)
block->meta_size = cpu_to_le32(io->meta_offset);
crc = crc32c_le(log->uuid_checksum, block, PAGE_SIZE);
block->checksum = cpu_to_le32(crc);
+ bio = io->current_bio;
log->current_io = NULL;
spin_lock_irqsave(&log->io_list_lock, flags);
- __r5l_set_io_unit_state(io, IO_UNIT_IO_START);
+ if (io->has_flush || io->has_fua) {
+ if (io != list_first_entry(&log->running_ios,
+ struct r5l_io_unit, log_sibling)) {
+ io->io_deferred = 1;
+ do_submit = false;
+ }
+ }
spin_unlock_irqrestore(&log->io_list_lock, flags);
-
- submit_bio(io->current_bio);
+ if (do_submit)
+ r5l_do_submit_io(log, io);
}
static struct bio *r5l_bio_alloc(struct r5l_log *log)
@@ -583,6 +679,7 @@ static struct r5l_io_unit *r5l_new_meta(struct r5l_log *log)
io->log = log;
INIT_LIST_HEAD(&io->log_sibling);
INIT_LIST_HEAD(&io->stripe_list);
+ bio_list_init(&io->flush_barriers);
io->state = IO_UNIT_RUNNING;
io->meta_page = mempool_alloc(log->meta_pool, GFP_NOIO);
@@ -653,12 +750,11 @@ static void r5l_append_payload_page(struct r5l_log *log, struct page *page)
struct r5l_io_unit *io = log->current_io;
if (io->need_split_bio) {
- struct bio *prev = io->current_bio;
-
+ BUG_ON(io->split_bio);
+ io->split_bio = io->current_bio;
io->current_bio = r5l_bio_alloc(log);
- bio_chain(io->current_bio, prev);
-
- submit_bio(prev);
+ bio_chain(io->current_bio, io->split_bio);
+ io->need_split_bio = false;
}
if (!bio_add_page(io->current_bio, page, PAGE_SIZE, 0))
@@ -687,12 +783,23 @@ static int r5l_log_stripe(struct r5l_log *log, struct stripe_head *sh,
io = log->current_io;
+ if (test_and_clear_bit(STRIPE_R5C_PREFLUSH, &sh->state))
+ io->has_flush = 1;
+
for (i = 0; i < sh->disks; i++) {
if (!test_bit(R5_Wantwrite, &sh->dev[i].flags) ||
test_bit(R5_InJournal, &sh->dev[i].flags))
continue;
if (i == sh->pd_idx || i == sh->qd_idx)
continue;
+ if (test_bit(R5_WantFUA, &sh->dev[i].flags)) {
+ io->has_fua = 1;
+ /*
+ * we need to flush journal to make sure recovery can
+ * reach the data with fua flag
+ */
+ io->has_flush = 1;
+ }
r5l_append_payload_meta(log, R5LOG_PAYLOAD_DATA,
raid5_compute_blocknr(sh, i, 0),
sh->dev[i].log_checksum, 0, false);
@@ -856,17 +963,34 @@ int r5l_handle_flush_request(struct r5l_log *log, struct bio *bio)
{
if (!log)
return -ENODEV;
- /*
- * we flush log disk cache first, then write stripe data to raid disks.
- * So if bio is finished, the log disk cache is flushed already. The
- * recovery guarantees we can recovery the bio from log disk, so we
- * don't need to flush again
- */
- if (bio->bi_iter.bi_size == 0) {
- bio_endio(bio);
- return 0;
+
+ if (log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH) {
+ /*
+ * in write through (journal only)
+ * we flush log disk cache first, then write stripe data to
+ * raid disks. So if bio is finished, the log disk cache is
+ * flushed already. The recovery guarantees we can recovery
+ * the bio from log disk, so we don't need to flush again
+ */
+ if (bio->bi_iter.bi_size == 0) {
+ bio_endio(bio);
+ return 0;
+ }
+ bio->bi_opf &= ~REQ_PREFLUSH;
+ } else {
+ /* write back (with cache) */
+ if (bio->bi_iter.bi_size == 0) {
+ mutex_lock(&log->io_mutex);
+ r5l_get_meta(log, 0);
+ bio_list_add(&log->current_io->flush_barriers, bio);
+ log->current_io->has_flush = 1;
+ log->current_io->has_null_flush = 1;
+ atomic_inc(&log->current_io->pending_stripe);
+ r5l_submit_current_io(log);
+ mutex_unlock(&log->io_mutex);
+ return 0;
+ }
}
- bio->bi_opf &= ~REQ_PREFLUSH;
return -EAGAIN;
}
@@ -2470,6 +2594,8 @@ int r5l_init_log(struct r5conf *conf, struct md_rdev *rdev)
INIT_LIST_HEAD(&log->no_space_stripes);
spin_lock_init(&log->no_space_stripes_lock);
+ INIT_WORK(&log->deferred_io_work, r5l_submit_io_async);
+
log->r5c_journal_mode = R5C_JOURNAL_MODE_WRITE_THROUGH;
INIT_LIST_HEAD(&log->stripe_in_journal_list);
spin_lock_init(&log->stripe_in_journal_lock);
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index aa4968c..9eafbb0 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -5248,6 +5248,7 @@ static void raid5_make_request(struct mddev *mddev, struct bio * bi)
int remaining;
DEFINE_WAIT(w);
bool do_prepare;
+ bool do_flush = false;
if (unlikely(bi->bi_opf & REQ_PREFLUSH)) {
int ret = r5l_handle_flush_request(conf->log, bi);
@@ -5259,6 +5260,7 @@ static void raid5_make_request(struct mddev *mddev, struct bio * bi)
return;
}
/* ret == -EAGAIN, fallback */
+ do_flush = true;
}
md_write_start(mddev, bi);
@@ -5398,6 +5400,12 @@ static void raid5_make_request(struct mddev *mddev, struct bio * bi)
do_prepare = true;
goto retry;
}
+ if (do_flush) {
+ set_bit(STRIPE_R5C_PREFLUSH, &sh->state);
+ /* we only need flush for one stripe */
+ do_flush = false;
+ }
+
set_bit(STRIPE_HANDLE, &sh->state);
clear_bit(STRIPE_DELAYED, &sh->state);
if ((!sh->batch_head || sh == sh->batch_head) &&
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
index a698113..d13fe45 100644
--- a/drivers/md/raid5.h
+++ b/drivers/md/raid5.h
@@ -376,6 +376,7 @@ enum {
STRIPE_R5C_FULL_STRIPE, /* in r5c cache (to-be/being handled or
* in conf->r5c_full_stripe_list)
*/
+ STRIPE_R5C_PREFLUSH, /* need to flush journal device */
};
#define STRIPE_EXPAND_SYNC_FLAGS \
--
2.9.3
^ permalink raw reply related
* Re: [PATCH 1/2] raid5-cache: update superblock at shutdown/reboot
From: NeilBrown @ 2016-11-18 0:01 UTC (permalink / raw)
To: Shaohua Li; +Cc: Shaohua Li, linux-raid, songliubraving, Zhengyuan Liu
In-Reply-To: <20161117181353.z7mgiynosyb664oc@kernel.org>
[-- Attachment #1: Type: text/plain, Size: 2804 bytes --]
On Fri, Nov 18 2016, Shaohua Li wrote:
> On Thu, Nov 17, 2016 at 04:18:15PM +1100, Neil Brown wrote:
>> On Thu, Nov 17 2016, Shaohua Li wrote:
>>
>> > Currently raid5-cache update superblock in .quiesce. But since at
>> > shutdown/reboot, .quiesce is called with reconfig mutex locked,
>> > superblock isn't guaranteed to be called in reclaim thread (see
>> > 8e018c21da3). This will make assemble do unnecessary journal recovery.
>> > It doesn't corrupt data but is annoying. This adds an extra hook to
>> > guarantee journal is flushed to raid disks. And since this hook is
>> > called before superblock update, this will guarantee we have a uptodate
>> > superblock in shutdown/reboot
>>
>> Hi.
>> I don't quite follow some of the reasoning here.
>> In particular, the ->stop_writes() that you have implemented
>> does almost exactly the same thing as r5l_quiesce(1).
>> So why not simply call ->quiesce(mddev, 1) in __md_stop_writes()??
>> You probably need to also call ->quiesce(mddev, 0) to keep things
>> balanced.
>
> reboot (md_notify_reboot) doesn't call .quiesce, maybe we should do though. And
> in stop, we hold reconfig_mutex before calling .quiesce. And with commit
> 8e018c21da3, r5l_write_super_and_discard_space tries to hold the reconfig_mutex
> before write super, which it can't hold, so superblock write is skipped. After
> .quiesce we don't write superblock. To fix the shutdown case, we can add a
> superblock write after .quiesce. But I think it's more generic to add a
> ->stop_writes since it will work for the reboot case.
I hadn't quite processed that this was about md_notify_reboot().
I would be very wary of optimizing this code. It should certainly avoid
data loss, but anything more doesn't belong here.
During a clean shutdown the array should be stopped properly.
md_notify_reboot() is only meant for minimizing damaged caused by a
hasty "reboot -f -n".
A "clean" shutdown currently includes systemd/mdadm.shutdown (in the
mdadm package) running "mdadm --wait-clean --scan".
"mdadm --wait-clean" changes the "safe_mode_delay" so that the array
will become "clean" more quickly.
Possibly we should add something to that to trigger a flush of the
journal, and to wait for the flush to complete.
>
>> Also you have introduced a static mutex (which isn't my favourite sort
>> of thing) without giving any explanation why in the changelog comment.
>> So I cannot easily see if that addition is at all justified.
>
> Now it's possible both the reclaim thread and the thread calling
> __md_stop_writes run into r5l_do_reclaim. The mutex is trying to avoid races.
> I'll add comments there.
It seems to me there is already quite a bit of locking in there... I'll
wait to read the comments though.
NeilBrown
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 800 bytes --]
^ permalink raw reply
* Re: [md PATCH 1/4] md: add block tracing for bio_remapping
From: NeilBrown @ 2016-11-18 0:45 UTC (permalink / raw)
To: Shaohua Li; +Cc: linux-raid
In-Reply-To: <20161117180423.tzmeqpgklihgh7vp@kernel.org>
[-- Attachment #1: Type: text/plain, Size: 2291 bytes --]
On Fri, Nov 18 2016, Shaohua Li wrote:
> On Thu, Nov 17, 2016 at 04:33:33PM +1100, Neil Brown wrote:
>> On Thu, Nov 17 2016, Shaohua Li wrote:
>>
>> > On Mon, Nov 14, 2016 at 04:30:21PM +1100, Neil Brown wrote:
>> >> The block tracing infrastructure (accessed with blktrace/blkparse)
>> >> supports the tracing of mapping bios from one device to another.
>> >> This is currently used when a bio in a partition is mapped to the
>> >> whole device, when bios are mapped by dm, and for mapping in md/raid5.
>> >> Other md personalities do not include this tracing yet, so add it.
>> >>
>> >> When a read-error is detected we redirect the request to a different device.
>> >> This could justifiably be seen as a new mapping for the originial bio,
>> >> or a secondary mapping for the bio that errors. This patch uses
>> >> the second option.
>> >>
>> >> When md is used under dm-raid, the mappings are not traced as we do
>> >> not have access to the block device number of the parent.
>> >
>> > Looks the the original sector (the last parameter of trace_block_bio_remap)
>> > isn't correct.
>> > - in linear/raid0, bio_split already updated bio->bi_iter.bi_sector
>>
>> Oh yes, of course. in the common case 'split == bio' so when
>> split->bi_iter.bi_sector is adjusted, bio->.... is as well.
>> I'll fix that, and also add calls to trace_block_split() as appropriate.
>>
>>
>> > - in raid1/raid10, r1_bio->sector is updated before the bio is sent.
>>
>> Here I really think my code is correct. r1_bio->sector is always the
>> address in the array of the request. It is only set once for each
>> r1_bio, and that is before the call to trace_block_io_remap().
>
> Oh, you are right, sorry. I think moving the trace_remap right before we add
> the bio to plug or pending list is better. It will make the code simpler. The
> timing of the trace doesn't matter.
I agree the timing doesn't matter
The reason I didn't do that before is that bio->bi_bdev points to the
rdev when the write bio is queued to the pending list. I'm not sure if
trace_block_bio_remap() accesses ->bi_bdev, but it could.
So I'm not completely sure that it makes the code simpler, but I'll post
a version like that to see what you think.
Thanks,
NeilBrown
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 800 bytes --]
^ permalink raw reply
* Re: [PATCH 1/2] raid5-cache: update superblock at shutdown/reboot
From: Shaohua Li @ 2016-11-18 1:41 UTC (permalink / raw)
To: NeilBrown; +Cc: Shaohua Li, linux-raid, songliubraving, Zhengyuan Liu
In-Reply-To: <8760nlhd18.fsf@notabene.neil.brown.name>
On Fri, Nov 18, 2016 at 11:01:07AM +1100, Neil Brown wrote:
> On Fri, Nov 18 2016, Shaohua Li wrote:
>
> > On Thu, Nov 17, 2016 at 04:18:15PM +1100, Neil Brown wrote:
> >> On Thu, Nov 17 2016, Shaohua Li wrote:
> >>
> >> > Currently raid5-cache update superblock in .quiesce. But since at
> >> > shutdown/reboot, .quiesce is called with reconfig mutex locked,
> >> > superblock isn't guaranteed to be called in reclaim thread (see
> >> > 8e018c21da3). This will make assemble do unnecessary journal recovery.
> >> > It doesn't corrupt data but is annoying. This adds an extra hook to
> >> > guarantee journal is flushed to raid disks. And since this hook is
> >> > called before superblock update, this will guarantee we have a uptodate
> >> > superblock in shutdown/reboot
> >>
> >> Hi.
> >> I don't quite follow some of the reasoning here.
> >> In particular, the ->stop_writes() that you have implemented
> >> does almost exactly the same thing as r5l_quiesce(1).
> >> So why not simply call ->quiesce(mddev, 1) in __md_stop_writes()??
> >> You probably need to also call ->quiesce(mddev, 0) to keep things
> >> balanced.
> >
> > reboot (md_notify_reboot) doesn't call .quiesce, maybe we should do though. And
> > in stop, we hold reconfig_mutex before calling .quiesce. And with commit
> > 8e018c21da3, r5l_write_super_and_discard_space tries to hold the reconfig_mutex
> > before write super, which it can't hold, so superblock write is skipped. After
> > .quiesce we don't write superblock. To fix the shutdown case, we can add a
> > superblock write after .quiesce. But I think it's more generic to add a
> > ->stop_writes since it will work for the reboot case.
>
> I hadn't quite processed that this was about md_notify_reboot().
> I would be very wary of optimizing this code. It should certainly avoid
> data loss, but anything more doesn't belong here.
> During a clean shutdown the array should be stopped properly.
> md_notify_reboot() is only meant for minimizing damaged caused by a
> hasty "reboot -f -n".
yep, this isn't the priority. So do you still suggest we ignore the reboot case
and add the journal flush after .quiesce() is called in stop?
> A "clean" shutdown currently includes systemd/mdadm.shutdown (in the
> mdadm package) running "mdadm --wait-clean --scan".
> "mdadm --wait-clean" changes the "safe_mode_delay" so that the array
> will become "clean" more quickly.
> Possibly we should add something to that to trigger a flush of the
> journal, and to wait for the flush to complete.
I'm not sure how we can do this. But adding a different state in 'array_state'
indicating journal isn't clean makes sense.
Thanks,
Shaohua
^ permalink raw reply
* Re: Newly-created arrays don't auto-assemble - related to hostname change?
From: Glenn Enright @ 2016-11-18 2:03 UTC (permalink / raw)
To: mdraid
In-Reply-To: <014d01d24129$7a9b2c20$6fd18460$@wnsdev.com>
Could the /usr/share/mdadm/mkconf script help? It can be used to print
out the running equivalent of mdadm.conf
It might be a good thing to use for comparisons, I use it a fair bit
in my scripting with good results.
Regards, Glenn
On 18 November 2016 at 12:22, Peter Sangas <pete@wnsdev.com> wrote:
> Andy, Your question as prompted me to think about the following: I'm using Ubuntu 16 and have a running system with RAID1. If I change the hostname of the system do I need to make any changes to /etc/mdadm/mdadm.conf file and if so how do I do that?
>
> I see the host name is listed at the end of /etc/mdadm/mdadm.conf (name=hostname:0) for example.
>
> Thank you,
> Pete
>
>
> -----Original Message-----
> From: Andy Smith [mailto:andy@strugglers.net]
> Sent: Wednesday, November 16, 2016 7:53 PM
> To: linux-raid@vger.kernel.org
> Subject: Newly-created arrays don't auto-assemble - related to hostname change?
>
> Hi,
>
> I feel I am missing something very simple here, as I usually don't have this issue, but here goes…
>
> I've got a Debian jessie host on which I created four arrays during install (md{0,1,2,3}), using the Debian installer and partman. These auto-assemble fine.
>
> After install the name of the server was changed from "tbd" to "jfd". Another array was then created (md5), added to /etc/mdadm/mdadm.conf and the initramfs was rebuilt (update-initramfs -u).
>
> md5 does not auto-assemble. It can be started fine after boot with:
>
> # mdadm --assemble /dev/md5
>
> or:
>
> # mdadm --incremental /dev/sdc
> # mdadm --incremental /dev/sdd
>
> /etc/mdadm/mdadm.conf:
>
> DEVICE /dev/sd*
> CREATE owner=root group=disk mode=0660 auto=yes
> HOMEHOST <ignore>
> MAILADDR root
> ARRAY /dev/md/0 metadata=1.2 UUID=400bac1d:e2c5d6ef:fea3b8c8:bcb70f8f
> ARRAY /dev/md/1 metadata=1.2 UUID=e29c8b89:705f0116:d888f77e:2b6e32f5
> ARRAY /dev/md/2 metadata=1.2 UUID=039b3427:4be5157a:6e2d53bd:fe898803
> ARRAY /dev/md/3 metadata=1.2 UUID=30f745ce:7ed41b53:4df72181:7406ea1d
> ARRAY /dev/md/5 metadata=1.2 UUID=957030cf:c09f023d:ceaebb27:e546f095
>
> I've unpacked the initramfs and looked at the mdadm.conf in there and it is identical.
>
> Initially HOMEHOST was set to <system> (the default), but I noticed when looking at --detail that md5 has:
>
> Name : jfd:5 (local to host jfd)
>
> whereas the others have:
>
> Name : tbd:0
>
> …so I changed it to <ignore> to see if that would help. It didn't.
>
> So, I'd really appreciate any hints as to what I've missed here!
>
> Here follows --detail and --examine of md5 and its members, then the contents of /proc/mdstat after I have manually assembled md5.
>
> $ sudo mdadm --detail /dev/md5
> /dev/md5:
> Version : 1.2
> Creation Time : Thu Nov 17 02:35:15 2016
> Raid Level : raid10
> Array Size : 1875243008 (1788.37 GiB 1920.25 GB)
> Used Dev Size : 1875243008 (1788.37 GiB 1920.25 GB)
> Raid Devices : 2
> Total Devices : 2
> Persistence : Superblock is persistent
>
> Intent Bitmap : Internal
>
> Update Time : Thu Nov 17 02:35:15 2016
> State : clean
> Active Devices : 2
> Working Devices : 2
> Failed Devices : 0
> Spare Devices : 0
>
> Layout : far=2
> Chunk Size : 512K
>
> Name : jfd:5 (local to host jfd)
> UUID : 957030cf:c09f023d:ceaebb27:e546f095
> Events : 0
>
> Number Major Minor RaidDevice State
> 0 8 48 0 active sync /dev/sdd
> 1 8 32 1 active sync /dev/sdc
>
> $ sudo mdadm --examine /dev/sd{c,d}
> /dev/sdc:
> Magic : a92b4efc
> Version : 1.2
> Feature Map : 0x1
> Array UUID : 957030cf:c09f023d:ceaebb27:e546f095
> Name : jfd:5 (local to host jfd)
> Creation Time : Thu Nov 17 02:35:15 2016
> Raid Level : raid10
> Raid Devices : 2
>
> Avail Dev Size : 3750486704 (1788.37 GiB 1920.25 GB)
> Array Size : 1875243008 (1788.37 GiB 1920.25 GB)
> Used Dev Size : 3750486016 (1788.37 GiB 1920.25 GB)
> Data Offset : 262144 sectors
> Super Offset : 8 sectors
> Unused Space : before=262056 sectors, after=688 sectors
> State : clean
> Device UUID : 4ac82c29:2d109465:7fff9b22:8c411c1e
>
> Internal Bitmap : 8 sectors from superblock
> Update Time : Thu Nov 17 02:35:15 2016
> Bad Block Log : 512 entries available at offset 72 sectors
> Checksum : 96d669f1 - correct
> Events : 0
>
> Layout : far=2
> Chunk Size : 512K
>
> Device Role : Active device 1
> Array State : AA ('A' == active, '.' == missing, 'R' == replacing)
> /dev/sdd:
> Magic : a92b4efc
> Version : 1.2
> Feature Map : 0x1
> Array UUID : 957030cf:c09f023d:ceaebb27:e546f095
> Name : jfd:5 (local to host jfd)
> Creation Time : Thu Nov 17 02:35:15 2016
> Raid Level : raid10
> Raid Devices : 2
> Avail Dev Size : 3750486704 (1788.37 GiB 1920.25 GB)
> Array Size : 1875243008 (1788.37 GiB 1920.25 GB)
> Used Dev Size : 3750486016 (1788.37 GiB 1920.25 GB)
> Data Offset : 262144 sectors
> Super Offset : 8 sectors
> Unused Space : before=262056 sectors, after=688 sectors
> State : clean
> Device UUID : 3a067652:6e88afae:82722342:0036bae0
>
> Internal Bitmap : 8 sectors from superblock
> Update Time : Thu Nov 17 02:35:15 2016
> Bad Block Log : 512 entries available at offset 72 sectors
> Checksum : eb608799 - correct
> Events : 0
>
> Layout : far=2
> Chunk Size : 512K
>
> Device Role : Active device 0
> Array State : AA ('A' == active, '.' == missing, 'R' == replacing)
>
> $ cat /proc/mdstat
> Personalities : [raid1] [raid10]
> md5 : active (auto-read-only) raid10 sdd[0] sdc[1]
> 1875243008 blocks super 1.2 512K chunks 2 far-copies [2/2] [UU]
> bitmap: 0/14 pages [0KB], 65536KB chunk
>
> md3 : active raid10 sda5[0] sdb5[1]
> 12199936 blocks super 1.2 512K chunks 2 far-copies [2/2] [UU]
>
> md2 : active (auto-read-only) raid10 sda3[0] sdb3[1]
> 975872 blocks super 1.2 512K chunks 2 far-copies [2/2] [UU]
>
> md1 : active raid10 sda2[0] sdb2[1]
> 1951744 blocks super 1.2 512K chunks 2 far-copies [2/2] [UU]
>
> md0 : active raid1 sda1[0] sdb1[1]
> 498368 blocks super 1.2 [2/2] [UU]
>
> unused devices: <none>
>
> Cheers,
> Andy
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply
* [PATCH] block: call trace_block_split() from bio_split()
From: NeilBrown @ 2016-11-18 2:14 UTC (permalink / raw)
To: Jens Axboe, Christoph Hellwig; +Cc: linux-block, linux-raid, linux-kernel
[-- Attachment #1: Type: text/plain, Size: 1588 bytes --]
Somewhere around
Commit: 20d0189b1012 ("block: Introduce new bio_split()")
and
Commit: 4b1faf931650 ("block: Kill bio_pair_split()")
in 3.14 we lost the call to trace_block_split() from bio_split().
Commit: cda22646adaa ("block: add call to split trace point")
in 4.5 added it back for blk_queue_split(), but not for other users of
bio_split(), and particularly not for md/raid.
This patch moves the trace_block_split() call from blk_queue_split()
to bio_split().
As blk_queue_split() calls bio_split() (via various helper functions)
the same events that were traced before will still be traced.
Signed-off-by: NeilBrown <neilb@suse.com>
---
block/bio.c | 1 +
block/blk-merge.c | 1 -
2 files changed, 1 insertion(+), 1 deletion(-)
diff --git a/block/bio.c b/block/bio.c
index db85c5753a76..212ea95a7401 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -1804,6 +1804,7 @@ struct bio *bio_split(struct bio *bio, int sectors,
bio_integrity_trim(split, 0, sectors);
bio_advance(bio, split->bi_iter.bi_size);
+ trace_block_split(bdev_get_queue(bio->bi_bdev), split, bio->bi_iter.bi_sector);
return split;
}
diff --git a/block/blk-merge.c b/block/blk-merge.c
index 2642e5fc8b69..82cdd35a9f07 100644
--- a/block/blk-merge.c
+++ b/block/blk-merge.c
@@ -217,7 +217,6 @@ void blk_queue_split(struct request_queue *q, struct bio **bio,
split->bi_opf |= REQ_NOMERGE;
bio_chain(split, *bio);
- trace_block_split(q, split, (*bio)->bi_iter.bi_sector);
generic_make_request(*bio);
*bio = split;
}
--
2.10.2
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 800 bytes --]
^ permalink raw reply related
* [PATCH v2] md: add block tracing for bio_remapping
From: NeilBrown @ 2016-11-18 2:22 UTC (permalink / raw)
To: Shaohua Li; +Cc: linux-raid
[-- Attachment #1: Type: text/plain, Size: 10897 bytes --]
The block tracing infrastructure (accessed with blktrace/blkparse)
supports the tracing of mapping bios from one device to another.
This is currently used when a bio in a partition is mapped to the
whole device, when bios are mapped by dm, and for mapping in md/raid5.
Other md personalities do not include this tracing yet, so add it.
When a read-error is detected we redirect the request to a different device.
This could justifiably be seen as a new mapping for the originial bio,
or a secondary mapping for the bio that errors. This patch uses
the second option.
When md is used under dm-raid, the mappings are not traced as we do
not have access to the block device number of the parent.
Signed-off-by: NeilBrown <neilb@suse.com>
---
This is the revised version based on discussions.
Now uses correct sector for linear and raid0, and code for raid1/raid10
rearranged a bit.
drivers/md/linear.c | 18 ++++++++++++------
drivers/md/raid0.c | 13 ++++++++++---
drivers/md/raid1.c | 26 ++++++++++++++++++++++++--
drivers/md/raid10.c | 29 +++++++++++++++++++++++++++--
4 files changed, 73 insertions(+), 13 deletions(-)
diff --git a/drivers/md/linear.c b/drivers/md/linear.c
index 9c7d4f5483ea..5975c9915684 100644
--- a/drivers/md/linear.c
+++ b/drivers/md/linear.c
@@ -21,6 +21,7 @@
#include <linux/seq_file.h>
#include <linux/module.h>
#include <linux/slab.h>
+#include <trace/events/block.h>
#include "md.h"
#include "linear.h"
@@ -227,22 +228,22 @@ static void linear_make_request(struct mddev *mddev, struct bio *bio)
}
do {
- tmp_dev = which_dev(mddev, bio->bi_iter.bi_sector);
+ sector_t bio_sector = bio->bi_iter.bi_sector;
+ tmp_dev = which_dev(mddev, bio_sector);
start_sector = tmp_dev->end_sector - tmp_dev->rdev->sectors;
end_sector = tmp_dev->end_sector;
data_offset = tmp_dev->rdev->data_offset;
bio->bi_bdev = tmp_dev->rdev->bdev;
- if (unlikely(bio->bi_iter.bi_sector >= end_sector ||
- bio->bi_iter.bi_sector < start_sector))
+ if (unlikely(bio_sector >= end_sector ||
+ bio_sector < start_sector))
goto out_of_bounds;
if (unlikely(bio_end_sector(bio) > end_sector)) {
/* This bio crosses a device boundary, so we have to
* split it.
*/
- split = bio_split(bio, end_sector -
- bio->bi_iter.bi_sector,
+ split = bio_split(bio, end_sector - bio_sector,
GFP_NOIO, fs_bio_set);
bio_chain(split, bio);
} else {
@@ -256,8 +257,13 @@ static void linear_make_request(struct mddev *mddev, struct bio *bio)
!blk_queue_discard(bdev_get_queue(split->bi_bdev)))) {
/* Just ignore it */
bio_endio(split);
- } else
+ } else {
+ if (mddev->gendisk)
+ trace_block_bio_remap(bdev_get_queue(split->bi_bdev),
+ split, disk_devt(mddev->gendisk),
+ bio_sector);
generic_make_request(split);
+ }
} while (split != bio);
return;
diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c
index b3ba77a3c3bc..e628f187e5ad 100644
--- a/drivers/md/raid0.c
+++ b/drivers/md/raid0.c
@@ -21,6 +21,7 @@
#include <linux/seq_file.h>
#include <linux/module.h>
#include <linux/slab.h>
+#include <trace/events/block.h>
#include "md.h"
#include "raid0.h"
#include "raid5.h"
@@ -463,7 +464,8 @@ static void raid0_make_request(struct mddev *mddev, struct bio *bio)
}
do {
- sector_t sector = bio->bi_iter.bi_sector;
+ sector_t bio_sector = bio->bi_iter.bi_sector;
+ sector_t sector = bio_sector;
unsigned chunk_sects = mddev->chunk_sectors;
unsigned sectors = chunk_sects -
@@ -472,7 +474,7 @@ static void raid0_make_request(struct mddev *mddev, struct bio *bio)
: sector_div(sector, chunk_sects));
/* Restore due to sector_div */
- sector = bio->bi_iter.bi_sector;
+ sector = bio_sector;
if (sectors < bio_sectors(bio)) {
split = bio_split(bio, sectors, GFP_NOIO, fs_bio_set);
@@ -491,8 +493,13 @@ static void raid0_make_request(struct mddev *mddev, struct bio *bio)
!blk_queue_discard(bdev_get_queue(split->bi_bdev)))) {
/* Just ignore it */
bio_endio(split);
- } else
+ } else {
+ if (mddev->gendisk)
+ trace_block_bio_remap(bdev_get_queue(split->bi_bdev),
+ split, disk_devt(mddev->gendisk),
+ bio_sector);
generic_make_request(split);
+ }
} while (split != bio);
}
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index 9ac61cd85e5c..2dc1934925ec 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -37,6 +37,7 @@
#include <linux/module.h>
#include <linux/seq_file.h>
#include <linux/ratelimit.h>
+#include <trace/events/block.h>
#include "md.h"
#include "raid1.h"
#include "bitmap.h"
@@ -1162,6 +1163,11 @@ static void raid1_make_request(struct mddev *mddev, struct bio * bio)
bio_set_op_attrs(read_bio, op, do_sync);
read_bio->bi_private = r1_bio;
+ if (mddev->gendisk)
+ trace_block_bio_remap(bdev_get_queue(read_bio->bi_bdev),
+ read_bio, disk_devt(mddev->gendisk),
+ r1_bio->sector);
+
if (max_sectors < r1_bio->sectors) {
/* could not read all from this device, so we will
* need another r1_bio.
@@ -1367,13 +1373,20 @@ static void raid1_make_request(struct mddev *mddev, struct bio * bio)
mbio->bi_iter.bi_sector = (r1_bio->sector +
conf->mirrors[i].rdev->data_offset);
- mbio->bi_bdev = (void*)conf->mirrors[i].rdev;
+ mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
mbio->bi_end_io = raid1_end_write_request;
bio_set_op_attrs(mbio, op, do_flush_fua | do_sync);
mbio->bi_private = r1_bio;
atomic_inc(&r1_bio->remaining);
+ if (mddev->gendisk)
+ trace_block_bio_remap(bdev_get_queue(mbio->bi_bdev),
+ mbio, disk_devt(mddev->gendisk),
+ r1_bio->sector);
+ /* flush_pending_writes() needs access to the rdev so...*/
+ mbio->bi_bdev = (void*)conf->mirrors[i].rdev;
+
cb = blk_check_plugged(raid1_unplug, mddev, sizeof(*plug));
if (cb)
plug = container_of(cb, struct raid1_plug_cb, cb);
@@ -2290,6 +2303,8 @@ static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
struct bio *bio;
char b[BDEVNAME_SIZE];
struct md_rdev *rdev;
+ dev_t bio_dev;
+ sector_t bio_sector;
clear_bit(R1BIO_ReadError, &r1_bio->state);
/* we got a read error. Maybe the drive is bad. Maybe just
@@ -2303,6 +2318,8 @@ static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
bio = r1_bio->bios[r1_bio->read_disk];
bdevname(bio->bi_bdev, b);
+ bio_dev = bio->bi_bdev->bd_dev;
+ bio_sector = conf->mirrors[r1_bio->read_disk].rdev->data_offset + r1_bio->sector;
bio_put(bio);
r1_bio->bios[r1_bio->read_disk] = NULL;
@@ -2353,6 +2370,8 @@ static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
else
mbio->bi_phys_segments++;
spin_unlock_irq(&conf->device_lock);
+ trace_block_bio_remap(bdev_get_queue(bio->bi_bdev),
+ bio, bio_dev, bio_sector);
generic_make_request(bio);
bio = NULL;
@@ -2367,8 +2386,11 @@ static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
sectors_handled;
goto read_more;
- } else
+ } else {
+ trace_block_bio_remap(bdev_get_queue(bio->bi_bdev),
+ bio, bio_dev, bio_sector);
generic_make_request(bio);
+ }
}
}
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index 5290be3d5c26..c63041ec9415 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -25,6 +25,7 @@
#include <linux/seq_file.h>
#include <linux/ratelimit.h>
#include <linux/kthread.h>
+#include <trace/events/block.h>
#include "md.h"
#include "raid10.h"
#include "raid0.h"
@@ -1165,6 +1166,10 @@ static void __make_request(struct mddev *mddev, struct bio *bio)
bio_set_op_attrs(read_bio, op, do_sync);
read_bio->bi_private = r10_bio;
+ if (mddev->gendisk)
+ trace_block_bio_remap(bdev_get_queue(read_bio->bi_bdev),
+ read_bio, disk_devt(mddev->gendisk),
+ r10_bio->sector);
if (max_sectors < r10_bio->sectors) {
/* Could not read all from this device, so we will
* need another r10_bio.
@@ -1367,11 +1372,17 @@ static void __make_request(struct mddev *mddev, struct bio *bio)
mbio->bi_iter.bi_sector = (r10_bio->devs[i].addr+
choose_data_offset(r10_bio,
rdev));
- mbio->bi_bdev = (void*)rdev;
+ mbio->bi_bdev = rdev->bdev;
mbio->bi_end_io = raid10_end_write_request;
bio_set_op_attrs(mbio, op, do_sync | do_fua);
mbio->bi_private = r10_bio;
+ if (conf->mddev->gendisk)
+ trace_block_bio_remap(bdev_get_queue(mbio->bi_bdev),
+ mbio, disk_devt(conf->mddev->gendisk),
+ r10_bio->sector);
+ mbio->bi_bdev = (void*)rdev;
+
atomic_inc(&r10_bio->remaining);
cb = blk_check_plugged(raid10_unplug, mddev,
@@ -1409,11 +1420,17 @@ static void __make_request(struct mddev *mddev, struct bio *bio)
mbio->bi_iter.bi_sector = (r10_bio->devs[i].addr +
choose_data_offset(
r10_bio, rdev));
- mbio->bi_bdev = (void*)rdev;
+ mbio->bi_bdev = rdev->bdev;
mbio->bi_end_io = raid10_end_write_request;
bio_set_op_attrs(mbio, op, do_sync | do_fua);
mbio->bi_private = r10_bio;
+ if (conf->mddev->gendisk)
+ trace_block_bio_remap(bdev_get_queue(mbio->bi_bdev),
+ mbio, disk_devt(conf->mddev->gendisk),
+ r10_bio->sector);
+ mbio->bi_bdev = (void*)rdev;
+
atomic_inc(&r10_bio->remaining);
spin_lock_irqsave(&conf->device_lock, flags);
bio_list_add(&conf->pending_bio_list, mbio);
@@ -2496,6 +2513,8 @@ static void handle_read_error(struct mddev *mddev, struct r10bio *r10_bio)
char b[BDEVNAME_SIZE];
unsigned long do_sync;
int max_sectors;
+ dev_t bio_dev;
+ sector_t bio_last_sector;
/* we got a read error. Maybe the drive is bad. Maybe just
* the block and we can fix it.
@@ -2507,6 +2526,8 @@ static void handle_read_error(struct mddev *mddev, struct r10bio *r10_bio)
*/
bio = r10_bio->devs[slot].bio;
bdevname(bio->bi_bdev, b);
+ bio_dev = bio->bi_bdev->bd_dev;
+ bio_last_sector = r10_bio->devs[slot].addr + rdev->data_offset + r10_bio->sectors;
bio_put(bio);
r10_bio->devs[slot].bio = NULL;
@@ -2546,6 +2567,10 @@ static void handle_read_error(struct mddev *mddev, struct r10bio *r10_bio)
bio_set_op_attrs(bio, REQ_OP_READ, do_sync);
bio->bi_private = r10_bio;
bio->bi_end_io = raid10_end_read_request;
+ trace_block_bio_remap(bdev_get_queue(bio->bi_bdev),
+ bio, bio_dev,
+ bio_last_sector - r10_bio->sectors);
+
if (max_sectors < r10_bio->sectors) {
/* Drat - have to split this up more */
struct bio *mbio = r10_bio->master_bio;
--
2.10.2
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 800 bytes --]
^ permalink raw reply related
* Re: Newly-created arrays don't auto-assemble - related to hostname change?
From: Andy Smith @ 2016-11-18 2:31 UTC (permalink / raw)
To: NeilBrown; +Cc: linux-raid
In-Reply-To: <87d1hthgm7.fsf@notabene.neil.brown.name>
Hi Neil,
On Fri, Nov 18, 2016 at 09:43:44AM +1100, NeilBrown wrote:
> Something you could try, after boot and while the arrays are still not
> assembled, is
>
> echo change > /sys/block/sdc/uevent
> echo change > /sys/block/sdd/uevent
>
> That should cause udev to assemble the array.
Nothing appeared to happen. There was nothing interesting in syslog
or systemd's journal. Here's the output from "udevadm monitor":
$ sudo udevadm monitor
monitor will print the received events for:
UDEV - the event which udev sends out after rule processing
KERNEL - the kernel uevent
KERNEL[12376.719100] change /devices/pci0000:00/0000:00:01.0/0000:01:00.0/host10/port-10:0/end_device-10:0/target10:0:0/10:0:0:0/block/sdc (block)
UDEV [12376.722996] change /devices/pci0000:00/0000:00:01.0/0000:01:00.0/host10/port-10:0/end_device-10:0/target10:0:0/10:0:0:0/block/sdc (block)
KERNEL[12388.329566] change /devices/pci0000:00/0000:00:01.0/0000:01:00.0/host10/port-10:1/end_device-10:1/target10:0:1/10:0:1:0/block/sdd (block)
UDEV [12388.333441] change /devices/pci0000:00/0000:00:01.0/0000:01:00.0/host10/port-10:1/end_device-10:1/target10:0:1/10:0:1:0/block/sdd (block)
> If that doesn't work, the looking over the udev logs, and possibly
> turning on extra udev logging, might lead to an answer.
I don't know how to do that, but I'll look into it - thanks.
Cheers,
Andy
^ permalink raw reply
* Re: Newly-created arrays don't auto-assemble - related to hostname change?
From: NeilBrown @ 2016-11-18 3:02 UTC (permalink / raw)
To: Andy Smith; +Cc: linux-raid
In-Reply-To: <20161118023144.GC1804@bitfolk.com>
[-- Attachment #1: Type: text/plain, Size: 1606 bytes --]
On Fri, Nov 18 2016, Andy Smith wrote:
> Hi Neil,
>
> On Fri, Nov 18, 2016 at 09:43:44AM +1100, NeilBrown wrote:
>> Something you could try, after boot and while the arrays are still not
>> assembled, is
>>
>> echo change > /sys/block/sdc/uevent
>> echo change > /sys/block/sdd/uevent
>>
>> That should cause udev to assemble the array.
>
> Nothing appeared to happen. There was nothing interesting in syslog
> or systemd's journal. Here's the output from "udevadm monitor":
OK, we are getting closer. That is good.
>
> $ sudo udevadm monitor
> monitor will print the received events for:
> UDEV - the event which udev sends out after rule processing
> KERNEL - the kernel uevent
>
> KERNEL[12376.719100] change /devices/pci0000:00/0000:00:01.0/0000:01:00.0/host10/port-10:0/end_device-10:0/target10:0:0/10:0:0:0/block/sdc (block)
> UDEV [12376.722996] change /devices/pci0000:00/0000:00:01.0/0000:01:00.0/host10/port-10:0/end_device-10:0/target10:0:0/10:0:0:0/block/sdc (block)
> KERNEL[12388.329566] change /devices/pci0000:00/0000:00:01.0/0000:01:00.0/host10/port-10:1/end_device-10:1/target10:0:1/10:0:1:0/block/sdd (block)
> UDEV [12388.333441] change /devices/pci0000:00/0000:00:01.0/0000:01:00.0/host10/port-10:1/end_device-10:1/target10:0:1/10:0:1:0/block/sdd (block)
>
>> If that doesn't work, the looking over the udev logs, and possibly
>> turning on extra udev logging, might lead to an answer.
>
> I don't know how to do that, but I'll look into it - thanks.
udevadm control -l debug
might help.
NeilBrown
>
> Cheers,
> Andy
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 800 bytes --]
^ permalink raw reply
* Re: Newly-created arrays don't auto-assemble - related to hostname change?
From: Andy Smith @ 2016-11-18 3:47 UTC (permalink / raw)
To: NeilBrown; +Cc: linux-raid
In-Reply-To: <87poltfq2d.fsf@notabene.neil.brown.name>
Hi Neil,
I've managed to workaround the issue.
In scrolling back on the serial console to view the early boot
messages and debug logging from udev, I noticed that only sd{a,b}
drives were detected during the initramfs stage. sd{c,d,e,f} were
absent.
This server has two internal flash devices, then the rest of the
drives are on a SAS controller. Evidently the initramfs does not
contain the driver for the SAS controller (mpt3sas). Possibly the
initramfs script correctly deduced that the system could be booted
with only sd{a,b} and did not bother including mpt3sas.
Anyway, I added mpt3sas to /etc/initramfs-tools/modules and rebuilt
the initramfs, and now all drives appear at boot time and all arrays
are assembled:
Begin: Loading essential drivers ... done.
Begin: Running /scripts/init-premount ... done.
Begin: Mounting root file system ... Begin: Running /scripts/local-top ... Begin: Assembling all MD arrays ... [ 40.1
58317] random: nonblocking pool is initialized
[ 40.161795] md: bind<sdf1>
[ 40.162161] md: bind<sde1>
[ 40.163226] md: raid1 personality registered for level 1
[ 40.163654] md/raid1:md0: active with 2 out of 2 mirrors
[ 40.163745] md0: detected capacity change from 0 to 510328832
[ 40.164259] md0: unknown partition table
mdadm: /dev/md/0 has been started with 2 drives.
[ 40.176662] md: bind<sdf2>
[ 40.177235] md: bind<sde2>
[ 40.178332] md: raid10 personality registered for level 10
[ 40.178656] md/raid10:md1: active with 2 out of 2 devices
[ 40.178746] md1: detected capacity change from 0 to 1998585856
[ 40.179170] md1: unknown partition table
mdadm: /dev/md/1 has been started with 2 drives.
[ 40.189887] md: md2 stopped.
[ 40.191292] md: bind<sdf3>
[ 40.191498] md: bind<sde3>
[ 40.192705] md/raid10:md2: active with 2 out of 2 devices
[ 40.192797] md2: detected capacity change from 0 to 999292928
[ 40.193128] md2: unknown partition table
mdadm: /dev/md/2 has been started with 2 drives.
[ 40.204234] md: md3 stopped.
[ 40.205278] md: bind<sdf5>
[ 40.205695] md: bind<sde5>
[ 40.206613] md/raid10:md3: active with 2 out of 2 devices
[ 40.206704] md3: detected capacity change from 0 to 12492734464
[ 40.207094] md3: unknown partition table
mdadm: /dev/md/3 has been started with 2 drives.
[ 40.218963] md: md5 stopped.
[ 40.223807] sdb: unknown partition table
[ 40.228841] sda: unknown partition table
[ 40.229044] md: bind<sda>
[ 40.229613] md: bind<sdb>
[ 40.234024] sdb: unknown partition table
[ 40.243686] md/raid10:md5: active with 2 out of 2 devices
[ 40.243867] created bitmap (14 pages) for device md5
[ 40.244684] md5: bitmap initialized from disk: read 1 pages, set 0 of 28614 bits
[ 40.245376] md5: detected capacity change from 0 to 1920248840192
[ 40.248331] md5: unknown partition table
mdadm: /dev/md/5 has been started with 2 drives.
Success: assembled all arrays.
done.
[ 40.260413] device-mapper: uevent: version 1.0.3
[ 40.260540] device-mapper: ioctl: 4.27.0-ioctl (2013-10-30) initialised: dm-devel@redhat.com
done.
Begin: Running /scripts/local-premount ... [ 40.265067] PM: Starting manual resume from disk
done.
Begin: Will now check root file system ... fsck from util-linux 2.25.2
[/sbin/fsck.ext4 (1) -- /dev/md1] fsck.ext4 -a -C0 /dev/md1
root: clean, 44775/122160 files, 265568/487936 blocks
done.
[ 40.296866] EXT4-fs (md1): mounted filesystem with ordered data mode. Opts: (null)
done.
Begin: Running /scripts/local-bottom ... done.
Begin: Running /scripts/init-bottom ... done.
[ 40.350701] systemd[1]: systemd 215 running in system mode. (+PAM +AUDIT +SELINUX +IMA +SYSVINIT +LIBCRYPTSETUP +GCR
YPT +ACL +XZ -SECCOMP -APPARMOR)
[ 40.350843] systemd[1]: Detected virtualization 'xen'.
[ 40.350920] systemd[1]: Detected architecture 'x86-64'.
Welcome to Debian GNU/Linux 8 (jessie)!
Am I right in thinking this is not incremental assembly by udev, but
kernel auto-assembly?
Also, in the previous configuration, mpt3sas was being loaded once
the root filesystem had been mounted. All the drives on the SAS
controller were then available, so udev should have assembled md5 at
that point, right? So I think there is still a problem here, just
one which I have worked around in some other way.
If that is the case, would you like me to continue debugging udev?
Cheers,
Andy
^ permalink raw reply
* Re: [PATCH 1/2] raid5-cache: update superblock at shutdown/reboot
From: NeilBrown @ 2016-11-18 3:49 UTC (permalink / raw)
To: Shaohua Li; +Cc: Shaohua Li, linux-raid, songliubraving, Zhengyuan Liu
In-Reply-To: <20161118014151.d6f3garrxovzrz66@kernel.org>
[-- Attachment #1: Type: text/plain, Size: 4336 bytes --]
On Fri, Nov 18 2016, Shaohua Li wrote:
> On Fri, Nov 18, 2016 at 11:01:07AM +1100, Neil Brown wrote:
>> On Fri, Nov 18 2016, Shaohua Li wrote:
>>
>> > On Thu, Nov 17, 2016 at 04:18:15PM +1100, Neil Brown wrote:
>> >> On Thu, Nov 17 2016, Shaohua Li wrote:
>> >>
>> >> > Currently raid5-cache update superblock in .quiesce. But since at
>> >> > shutdown/reboot, .quiesce is called with reconfig mutex locked,
>> >> > superblock isn't guaranteed to be called in reclaim thread (see
>> >> > 8e018c21da3). This will make assemble do unnecessary journal recovery.
>> >> > It doesn't corrupt data but is annoying. This adds an extra hook to
>> >> > guarantee journal is flushed to raid disks. And since this hook is
>> >> > called before superblock update, this will guarantee we have a uptodate
>> >> > superblock in shutdown/reboot
>> >>
>> >> Hi.
>> >> I don't quite follow some of the reasoning here.
>> >> In particular, the ->stop_writes() that you have implemented
>> >> does almost exactly the same thing as r5l_quiesce(1).
>> >> So why not simply call ->quiesce(mddev, 1) in __md_stop_writes()??
>> >> You probably need to also call ->quiesce(mddev, 0) to keep things
>> >> balanced.
>> >
>> > reboot (md_notify_reboot) doesn't call .quiesce, maybe we should do though. And
>> > in stop, we hold reconfig_mutex before calling .quiesce. And with commit
>> > 8e018c21da3, r5l_write_super_and_discard_space tries to hold the reconfig_mutex
>> > before write super, which it can't hold, so superblock write is skipped. After
>> > .quiesce we don't write superblock. To fix the shutdown case, we can add a
>> > superblock write after .quiesce. But I think it's more generic to add a
>> > ->stop_writes since it will work for the reboot case.
>>
>> I hadn't quite processed that this was about md_notify_reboot().
>> I would be very wary of optimizing this code. It should certainly avoid
>> data loss, but anything more doesn't belong here.
>> During a clean shutdown the array should be stopped properly.
>> md_notify_reboot() is only meant for minimizing damaged caused by a
>> hasty "reboot -f -n".
>
> yep, this isn't the priority. So do you still suggest we ignore the reboot case
> and add the journal flush after .quiesce() is called in stop?
Not sure... is there a problem we are trying to solve?
mddev_detach() calls ->quiesce() so things get cleaned up when the array
is stopped. md_set_readonly() only calls __md_stop_writes() though.
I would probably support adding a ->quiesce() call to __md_stop_writes,
except that r5l_quiesce() registers a new reclaim_thread() so when
__md_stop_writes() is followed by mddev_detach(), the thread would be
killed, then recreated, then killed, then recreated, then finally killed
in r5l_exit_log().
It would be nice if we could either
1/ not kill the thread, just freeze it, or
2/ have it start up some other way.
e.g. get raid5d() to start the thread if it isn't running and
conf->quiesce is 0, and array isn't clean.
>
>> A "clean" shutdown currently includes systemd/mdadm.shutdown (in the
>> mdadm package) running "mdadm --wait-clean --scan".
>> "mdadm --wait-clean" changes the "safe_mode_delay" so that the array
>> will become "clean" more quickly.
>> Possibly we should add something to that to trigger a flush of the
>> journal, and to wait for the flush to complete.
>
> I'm not sure how we can do this. But adding a different state in 'array_state'
> indicating journal isn't clean makes sense.
Possible, though as mdmon uses array_state, we would need to be careful.
Of course mdmon never looks at an array with a journal, so that wouldn't
actually be a problem.
Ignoring the new caching code for the moment, whenever the array is
"clean", the journal is irrelevant. Possibly we could arrange that if
the array is found to be "clean" on startup, the journal is ignored?
For that to work with the caching code, we would need to hold off
marking the array as 'clean' until the journal is flushed. I think that
is probably a good idea.
When safe_mode_delay is set to 0 (or 1) we should probably set the cache
flush time down from 30 seconds to (nearly) zero too. That way, the
current WaitClean() code would continue to work.
NeilBrown
>
> Thanks,
> Shaohua
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 800 bytes --]
^ permalink raw reply
* Re: Newly-created arrays don't auto-assemble - related to hostname change?
From: NeilBrown @ 2016-11-18 4:08 UTC (permalink / raw)
To: Andy Smith; +Cc: linux-raid
In-Reply-To: <20161118034716.GI21587@bitfolk.com>
[-- Attachment #1: Type: text/plain, Size: 1260 bytes --]
On Fri, Nov 18 2016, Andy Smith wrote:
>
> Am I right in thinking this is not incremental assembly by udev, but
> kernel auto-assembly?
That isn't kernel auto-assembly, but it doesn't look like incremental
assembly either.
/usr/share/initramfs-tools/scripts/local-block/mdadm
runs
mdadm -q --assemble --scan --no-degraded
to start all the arrays that the initramfs can find. So it is
mdadm-based auto-assembly.
>
> Also, in the previous configuration, mpt3sas was being loaded once
> the root filesystem had been mounted. All the drives on the SAS
> controller were then available, so udev should have assembled md5 at
> that point, right? So I think there is still a problem here, just
> one which I have worked around in some other way.
Yes, I think there is still a problem. When the mpt3sas is loaded udev
should trigger the arrays to be created.
>
> If that is the case, would you like me to continue debugging udev?
Up to you, but I have an idea.
The udev rules files depends on 'blkid' having been run.
/lib/udev/rules.d/60-persistent-storage.rules
does this, but not for
KERNEL=="fd*|mtd*|nbd*|gnbd*|btibm*|dm-*|md*|zram*|mmcblk[0-9]*rpmb"
... though that wouldn't apply to you.
what does
udevadm info /dev/sdc
report?
NeilBrown
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 800 bytes --]
^ permalink raw reply
* Re: Newly-created arrays don't auto-assemble - related to hostname change?
From: Andy Smith @ 2016-11-18 4:17 UTC (permalink / raw)
To: NeilBrown; +Cc: linux-raid
In-Reply-To: <87k2c1fn0o.fsf@notabene.neil.brown.name>
Hi Neil,
On Fri, Nov 18, 2016 at 03:08:23PM +1100, NeilBrown wrote:
> Up to you, but I have an idea.
> The udev rules files depends on 'blkid' having been run.
> /lib/udev/rules.d/60-persistent-storage.rules
> does this, but not for
> KERNEL=="fd*|mtd*|nbd*|gnbd*|btibm*|dm-*|md*|zram*|mmcblk[0-9]*rpmb"
>
> ... though that wouldn't apply to you.
>
> what does
> udevadm info /dev/sdc
(Since mpt3sas got loaded early the device identifiers have all
changed; what was sd{a,b} have now shifted to the end as sd{e,f}, so
the two members of md5 are now sd{a,b})
$ sudo udevadm info /dev/sda
P: /devices/pci0000:00/0000:00:01.0/0000:01:00.0/host0/port-0:0/end_device-0:0/target0:0:0/0:0:0:0/block/sda
N: sda
S: disk/by-id/ata-SAMSUNG_MZ7KM1T9HAJM-00005_S2HNNAAH200633
S: disk/by-id/wwn-0x5002538c0007e7a8
S: disk/by-path/pci-0000:01:00.0-sas-0x4433221100000000-lun-0
E: DEVLINKS=/dev/disk/by-id/ata-SAMSUNG_MZ7KM1T9HAJM-00005_S2HNNAAH200633 /dev/disk/by-id/wwn-0x5002538c0007e7a8 /dev/disk/by-path/pci-0000:01:00.0-sas-0x4433221100000000-lun-0
E: DEVNAME=/dev/sda
E: DEVPATH=/devices/pci0000:00/0000:00:01.0/0000:01:00.0/host0/port-0:0/end_device-0:0/target0:0:0/0:0:0:0/block/sda
E: DEVTYPE=disk
E: ID_ATA=1
E: ID_ATA_DOWNLOAD_MICROCODE=1
E: ID_ATA_FEATURE_SET_HPA=1
E: ID_ATA_FEATURE_SET_HPA_ENABLED=1
E: ID_ATA_FEATURE_SET_PM=1
E: ID_ATA_FEATURE_SET_PM_ENABLED=1
E: ID_ATA_FEATURE_SET_SECURITY=1
E: ID_ATA_FEATURE_SET_SECURITY_ENABLED=0
E: ID_ATA_FEATURE_SET_SECURITY_ENHANCED_ERASE_UNIT_MIN=32
E: ID_ATA_FEATURE_SET_SECURITY_ERASE_UNIT_MIN=32
E: ID_ATA_FEATURE_SET_SMART=1
E: ID_ATA_FEATURE_SET_SMART_ENABLED=1
E: ID_ATA_ROTATION_RATE_RPM=0
E: ID_ATA_SATA=1
E: ID_ATA_SATA_SIGNAL_RATE_GEN1=1
E: ID_ATA_SATA_SIGNAL_RATE_GEN2=1
E: ID_ATA_WRITE_CACHE=1
E: ID_ATA_WRITE_CACHE_ENABLED=1
E: ID_BUS=ata
E: ID_FS_LABEL=tbd:5
E: ID_FS_LABEL_ENC=tbd:5
E: ID_FS_TYPE=linux_raid_member
E: ID_FS_USAGE=raid
E: ID_FS_UUID=957030cf-c09f-023d-ceae-bb27e546f095
E: ID_FS_UUID_ENC=957030cf-c09f-023d-ceae-bb27e546f095
E: ID_FS_UUID_SUB=4ac82c29-2d10-9465-7fff-9b228c411c1e
E: ID_FS_UUID_SUB_ENC=4ac82c29-2d10-9465-7fff-9b228c411c1e
E: ID_FS_VERSION=1.2
E: ID_MODEL=SAMSUNG_MZ7KM1T9HAJM-00005
E: ID_MODEL_ENC=SAMSUNG\x20MZ7KM1T9HAJM-00005\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20
E: ID_PATH=pci-0000:01:00.0-sas-0x4433221100000000-lun-0
E: ID_PATH_TAG=pci-0000_01_00_0-sas-0x4433221100000000-lun-0
E: ID_REVISION=GXM1003Q
E: ID_SERIAL=SAMSUNG_MZ7KM1T9HAJM-00005_S2HNNAAH200633
E: ID_SERIAL_SHORT=S2HNNAAH200633
E: ID_TYPE=disk
E: ID_WWN=0x5002538c0007e7a8
E: ID_WWN_WITH_EXTENSION=0x5002538c0007e7a8
E: MAJOR=8
E: MINOR=0
E: SUBSYSTEM=block
E: TAGS=:systemd:
E: UDEV_LOG=7
E: USEC_INITIALIZED=38597
Cheers,
Andy
^ permalink raw reply
* [PATCH/RFC] add "failfast" support for raid1/raid10.
From: NeilBrown @ 2016-11-18 5:16 UTC (permalink / raw)
To: Shaohua Li; +Cc: linux-raid, linux-block, Christoph Hellwig, linux-kernel, hare
Hi,
I've been sitting on these patches for a while because although they
solve a real problem, it is a fairly limited use-case, and I don't
really like some of the details.
So I'm posting them as RFC in the hope that a different perspective
might help me like them better, or find a better approach.
The core idea is that when you have multiple copies of data
(i.e. mirrored drives) it doesn't make sense to wait for a read from
a drive that seems to be having problems. It will probably be faster
to just cancel that read, and read from the other device.
Similarly, in some circumstances, it might be better to fail a drive
that is being slow to respond to writes, rather than cause all writes
to be very slow.
The particular context where this comes up is when mirroring across
storage arrays, where the storage arrays can temporarily take an
unusually long time to respond to requests (firmware updates have
been mentioned). As the array will have redundancy internally, there
is little risk to the data. The mirrored pair is really only for
disaster recovery, and it is deemed better to lose the last few
minutes of updates in the case of a serious disaster, rather than
occasionally having latency issues because one array needs to do some
maintenance for a few minutes. The particular storage arrays in
question are DASD devices which are part of the s390 ecosystem.
Linux block layer has "failfast" flags to direct drivers to fail more
quickly. These patches allow devices in an md array to be given a
"failfast" flag, which will cause IO requests to be marked as
"failfast" providing there is another device available. Once the
array becomes degraded, we stop using failfast, as that could result
in data loss.
I don't like the whole "failfast" concept because it is not at all
clear how fast "fast" is. In fact, these block-layer flags are
really a misnomer. They should be "noretry" flags.
REQ_FAILFAST_DEV means "don't retry requests which reported an error
which seems to come from the device.
REQ_FAILFAST_TRANSPORT means "don't retry requests which seem to
indicate a problem with the transport, rather than the device"
REQ_FAILFAST_DRIVER means .... I'm not exactly sure. I think it
means whatever a particular driver wants it to mean, basically "I
cannot seem to handle this right now, just resend and I'll probably
be more in control next time". It seems to be for internal-use only.
Multipath code uses REQ_FAILFAST_TRANSPORT only, which makes sense.
btrfs uses REQ_FAILFAST_DEV only (for read-ahead) which doesn't seem
to make sense.... why would you ever use _DEV without _TRANSPORT?
None of these actually change the timeouts in the driver or in the
device, which is what I would expect for "failfast", so to get real
"fast failure" you need to enable failfast, and adjust the timeouts.
That is what we do for our customers with DASD.
Anyway, it seems to make sense to use _TRANSPORT and _DEV for
requests from md where there is somewhere to fall-back on.
If we get an error from a "failfast" request, and the array is still
non-degraded, we just fail the device. We don't try to repair read
errors (which is pointless on storage arrays).
It is assumed that some user-space code will notice the failure,
monitor the device to see when it becomes available again, and then
--re-add it. Assuming the array has a bitmap, the --re-add should be
fast and the array will become optimal again without experiencing
excessive latencies.
My two main concerns are:
- does this functionality have any use-case outside of mirrored
storage arrays, and are there other storage arrays which
occasionally inserted excessive latency (seems like a serious
misfeature to me, but I know few of the details)?
- would it be at all possible to have "real" failfast functionality
in the block layer? I.e. something that is based on time rather
than retry count. Maybe in some cases a retry would be
appropriate if the first failure was very fast.
I.e. it would reduce timeouts and decide on retries based on
elapsed time rather than number of attempts.
With this would come the question of "how fast is fast" and I
don't have a really good answer. Maybe md would need to set a
timeout, which it would double whenever it got failures on all
drives. Otherwise the timeout would drift towards (say) 10 times
the typical response time.
So: comments most welcome. As I say, this does address a genuine
need. Just find it hard to like it :-(
Thanks,
NeilBrown
---
NeilBrown (6):
md/failfast: add failfast flag for md to be used by some personalities.
md: Use REQ_FAILFAST_* on metadata writes where appropriate
md/raid1: add failfast handling for reads.
md/raid1: add failfast handling for writes.
md/raid10: add failfast handling for reads.
md/raid10: add failfast handling for writes.
drivers/md/bitmap.c | 15 ++++++--
drivers/md/md.c | 71 +++++++++++++++++++++++++++++++-----
drivers/md/md.h | 27 +++++++++++++-
drivers/md/raid1.c | 79 ++++++++++++++++++++++++++++++++++------
drivers/md/raid1.h | 1 +
drivers/md/raid10.c | 79 +++++++++++++++++++++++++++++++++++++---
drivers/md/raid10.h | 2 +
include/uapi/linux/raid/md_p.h | 7 +++-
8 files changed, 249 insertions(+), 32 deletions(-)
--
Signature
^ permalink raw reply
* [md PATCH 1/6] md/failfast: add failfast flag for md to be used by some personalities.
From: NeilBrown @ 2016-11-18 5:16 UTC (permalink / raw)
To: Shaohua Li; +Cc: linux-raid, linux-block, Christoph Hellwig, linux-kernel, hare
In-Reply-To: <147944614789.3302.1959091446949640579.stgit@noble>
This patch just adds a 'failfast' per-device flag which can be stored
in v0.90 or v1.x metadata.
The flag is not used yet but the intent is that it can be used for
mirrored (raid1/raid10) arrays where low latency is more important
than keeping all devices on-line.
Setting the flag for a device effectively gives permission for that
device to be marked as Faulty and excluded from the array on the first
error. The underlying driver will be directed not to retry requests
that result in failures. There is a proviso that the device must not
be marked faulty if that would cause the array as a whole to fail, it
may only be marked Faulty if the array remains functional, but is
degraded.
Failures on read requests will cause the device to be marked
as Faulty immediately so that further reads will avoid that
device. No attempt will be made to correct read errors by
over-writing with the correct data.
It is expected that if transient errors, such as cable unplug, are
possible, then something in user-space will revalidate failed
devices and re-add them when they appear to be working again.
Signed-off-by: NeilBrown <neilb@suse.com>
---
drivers/md/md.c | 27 +++++++++++++++++++++++++++
drivers/md/md.h | 6 ++++++
include/uapi/linux/raid/md_p.h | 7 ++++++-
3 files changed, 39 insertions(+), 1 deletion(-)
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 1f1c7f007b68..0d1a9fd9d1c1 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -1163,6 +1163,8 @@ static int super_90_validate(struct mddev *mddev, struct md_rdev *rdev)
}
if (desc->state & (1<<MD_DISK_WRITEMOSTLY))
set_bit(WriteMostly, &rdev->flags);
+ if (desc->state & (1<<MD_DISK_FAILFAST))
+ set_bit(FailFast, &rdev->flags);
} else /* MULTIPATH are always insync */
set_bit(In_sync, &rdev->flags);
return 0;
@@ -1288,6 +1290,8 @@ static void super_90_sync(struct mddev *mddev, struct md_rdev *rdev)
}
if (test_bit(WriteMostly, &rdev2->flags))
d->state |= (1<<MD_DISK_WRITEMOSTLY);
+ if (test_bit(FailFast, &rdev2->flags))
+ d->state |= (1<<MD_DISK_FAILFAST);
}
/* now set the "removed" and "faulty" bits on any missing devices */
for (i=0 ; i < mddev->raid_disks ; i++) {
@@ -1672,6 +1676,8 @@ static int super_1_validate(struct mddev *mddev, struct md_rdev *rdev)
}
if (sb->devflags & WriteMostly1)
set_bit(WriteMostly, &rdev->flags);
+ if (sb->devflags & FailFast1)
+ set_bit(FailFast, &rdev->flags);
if (le32_to_cpu(sb->feature_map) & MD_FEATURE_REPLACEMENT)
set_bit(Replacement, &rdev->flags);
} else /* MULTIPATH are always insync */
@@ -1710,6 +1716,10 @@ static void super_1_sync(struct mddev *mddev, struct md_rdev *rdev)
sb->chunksize = cpu_to_le32(mddev->chunk_sectors);
sb->level = cpu_to_le32(mddev->level);
sb->layout = cpu_to_le32(mddev->layout);
+ if (test_bit(FailFast, &rdev->flags))
+ sb->devflags |= FailFast1;
+ else
+ sb->devflags &= ~FailFast1;
if (test_bit(WriteMostly, &rdev->flags))
sb->devflags |= WriteMostly1;
@@ -2554,6 +2564,8 @@ state_show(struct md_rdev *rdev, char *page)
len += sprintf(page+len, "replacement%s", sep);
if (test_bit(ExternalBbl, &flags))
len += sprintf(page+len, "external_bbl%s", sep);
+ if (test_bit(FailFast, &flags))
+ len += sprintf(page+len, "failfast%s", sep);
if (len)
len -= strlen(sep);
@@ -2576,6 +2588,7 @@ state_store(struct md_rdev *rdev, const char *buf, size_t len)
* so that it gets rebuilt based on bitmap
* write_error - sets WriteErrorSeen
* -write_error - clears WriteErrorSeen
+ * {,-}failfast - set/clear FailFast
*/
int err = -EINVAL;
if (cmd_match(buf, "faulty") && rdev->mddev->pers) {
@@ -2634,6 +2647,12 @@ state_store(struct md_rdev *rdev, const char *buf, size_t len)
} else if (cmd_match(buf, "insync") && rdev->raid_disk == -1) {
set_bit(In_sync, &rdev->flags);
err = 0;
+ } else if (cmd_match(buf, "failfast")) {
+ set_bit(FailFast, &rdev->flags);
+ err = 0;
+ } else if (cmd_match(buf, "-failfast")) {
+ clear_bit(FailFast, &rdev->flags);
+ err = 0;
} else if (cmd_match(buf, "-insync") && rdev->raid_disk >= 0 &&
!test_bit(Journal, &rdev->flags)) {
if (rdev->mddev->pers == NULL) {
@@ -5939,6 +5958,8 @@ static int get_disk_info(struct mddev *mddev, void __user * arg)
info.state |= (1<<MD_DISK_JOURNAL);
if (test_bit(WriteMostly, &rdev->flags))
info.state |= (1<<MD_DISK_WRITEMOSTLY);
+ if (test_bit(FailFast, &rdev->flags))
+ info.state |= (1<<MD_DISK_FAILFAST);
} else {
info.major = info.minor = 0;
info.raid_disk = -1;
@@ -6046,6 +6067,10 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
set_bit(WriteMostly, &rdev->flags);
else
clear_bit(WriteMostly, &rdev->flags);
+ if (info->state & (1<<MD_DISK_FAILFAST))
+ set_bit(FailFast, &rdev->flags);
+ else
+ clear_bit(FailFast, &rdev->flags);
if (info->state & (1<<MD_DISK_JOURNAL)) {
struct md_rdev *rdev2;
@@ -6135,6 +6160,8 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
if (info->state & (1<<MD_DISK_WRITEMOSTLY))
set_bit(WriteMostly, &rdev->flags);
+ if (info->state & (1<<MD_DISK_FAILFAST))
+ set_bit(FailFast, &rdev->flags);
if (!mddev->persistent) {
pr_debug("md: nonpersistent superblock ...\n");
diff --git a/drivers/md/md.h b/drivers/md/md.h
index af6b33c30d2d..bc6712ef8c81 100644
--- a/drivers/md/md.h
+++ b/drivers/md/md.h
@@ -171,6 +171,12 @@ enum flag_bits {
ExternalBbl, /* External metadata provides bad
* block management for a disk
*/
+ FailFast, /* Minimal retries should be attempted on
+ * this device, so use REQ_FAILFAST_DEV.
+ * Also don't try to repair failed reads.
+ * It is expects that no bad block log
+ * is present.
+ */
};
static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
diff --git a/include/uapi/linux/raid/md_p.h b/include/uapi/linux/raid/md_p.h
index c3e654c6d518..9930f3e9040f 100644
--- a/include/uapi/linux/raid/md_p.h
+++ b/include/uapi/linux/raid/md_p.h
@@ -84,6 +84,10 @@
#define MD_DISK_CANDIDATE 5 /* disk is added as spare (local) until confirmed
* For clustered enviroments only.
*/
+#define MD_DISK_FAILFAST 10 /* Send REQ_FAILFAST if there are multiple
+ * devices available - and don't try to
+ * correct read errors.
+ */
#define MD_DISK_WRITEMOSTLY 9 /* disk is "write-mostly" is RAID1 config.
* read requests will only be sent here in
@@ -265,8 +269,9 @@ struct mdp_superblock_1 {
__le32 dev_number; /* permanent identifier of this device - not role in raid */
__le32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */
__u8 device_uuid[16]; /* user-space setable, ignored by kernel */
- __u8 devflags; /* per-device flags. Only one defined...*/
+ __u8 devflags; /* per-device flags. Only two defined...*/
#define WriteMostly1 1 /* mask for writemostly flag in above */
+#define FailFast1 2 /* Should avoid retries and fixups and just fail */
/* Bad block log. If there are any bad blocks the feature flag is set.
* If offset and size are non-zero, that space is reserved and available
*/
^ permalink raw reply related
* [md PATCH 2/6] md: Use REQ_FAILFAST_* on metadata writes where appropriate
From: NeilBrown @ 2016-11-18 5:16 UTC (permalink / raw)
To: Shaohua Li; +Cc: linux-raid, linux-block, Christoph Hellwig, linux-kernel, hare
In-Reply-To: <147944614789.3302.1959091446949640579.stgit@noble>
This can only be supported on personalities which ensure
that md_error() never causes an array to enter the 'failed'
state. i.e. if marking a device Faulty would cause some
data to be inaccessible, the device is status is left as
non-Faulty. This is true for RAID1 and RAID10.
If we get a failure writing metadata but the device doesn't
fail, it must be the last device so we re-write without
FAILFAST to improve chance of success. We also flag the
device as LastDev so that future metadata updates don't
waste time on failfast writes.
Signed-off-by: NeilBrown <neilb@suse.com>
---
drivers/md/bitmap.c | 15 ++++++++++++---
drivers/md/md.c | 44 ++++++++++++++++++++++++++++++++++----------
drivers/md/md.h | 21 ++++++++++++++++++++-
drivers/md/raid1.c | 1 +
drivers/md/raid10.c | 1 +
5 files changed, 68 insertions(+), 14 deletions(-)
diff --git a/drivers/md/bitmap.c b/drivers/md/bitmap.c
index cf77cbf9ed22..c4621571b718 100644
--- a/drivers/md/bitmap.c
+++ b/drivers/md/bitmap.c
@@ -209,11 +209,13 @@ static struct md_rdev *next_active_rdev(struct md_rdev *rdev, struct mddev *mdde
static int write_sb_page(struct bitmap *bitmap, struct page *page, int wait)
{
- struct md_rdev *rdev = NULL;
+ struct md_rdev *rdev;
struct block_device *bdev;
struct mddev *mddev = bitmap->mddev;
struct bitmap_storage *store = &bitmap->storage;
+restart:
+ rdev = NULL;
while ((rdev = next_active_rdev(rdev, mddev)) != NULL) {
int size = PAGE_SIZE;
loff_t offset = mddev->bitmap_info.offset;
@@ -269,8 +271,8 @@ static int write_sb_page(struct bitmap *bitmap, struct page *page, int wait)
page);
}
- if (wait)
- md_super_wait(mddev);
+ if (wait && md_super_wait(mddev) < 0)
+ goto restart;
return 0;
bad_alignment:
@@ -428,6 +430,13 @@ static void bitmap_wait_writes(struct bitmap *bitmap)
wait_event(bitmap->write_wait,
atomic_read(&bitmap->pending_writes)==0);
else
+ /* Note that we ignore the return value. The writes
+ * might have failed, but that would just mean that
+ * some bits which should be cleared haven't been,
+ * which is safe. The relevant bitmap blocks will
+ * probably get written again, but there is no great
+ * loss if they aren't.
+ */
md_super_wait(bitmap->mddev);
}
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 0d1a9fd9d1c1..047e7db1381b 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -726,7 +726,13 @@ static void super_written(struct bio *bio)
if (bio->bi_error) {
pr_err("md: super_written gets error=%d\n", bio->bi_error);
md_error(mddev, rdev);
- }
+ if (!test_bit(Faulty, &rdev->flags)
+ && (bio->bi_opf & MD_FAILFAST)) {
+ set_bit(MD_NEED_REWRITE, &mddev->flags);
+ set_bit(LastDev, &rdev->flags);
+ }
+ } else
+ clear_bit(LastDev, &rdev->flags);
if (atomic_dec_and_test(&mddev->pending_writes))
wake_up(&mddev->sb_wait);
@@ -743,7 +749,13 @@ void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
* if zero is reached.
* If an error occurred, call md_error
*/
- struct bio *bio = bio_alloc_mddev(GFP_NOIO, 1, mddev);
+ struct bio *bio;
+ int ff = 0;
+
+ if (test_bit(Faulty, &rdev->flags))
+ return;
+
+ bio = bio_alloc_mddev(GFP_NOIO, 1, mddev);
atomic_inc(&rdev->nr_pending);
@@ -752,16 +764,24 @@ void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
bio_add_page(bio, page, size, 0);
bio->bi_private = rdev;
bio->bi_end_io = super_written;
- bio_set_op_attrs(bio, REQ_OP_WRITE, WRITE_FLUSH_FUA);
+
+ if (test_bit(MD_FAILFAST_SUPPORTED, &mddev->flags) &&
+ test_bit(FailFast, &rdev->flags) &&
+ !test_bit(LastDev, &rdev->flags))
+ ff = MD_FAILFAST;
+ bio_set_op_attrs(bio, REQ_OP_WRITE, WRITE_FLUSH_FUA | ff);
atomic_inc(&mddev->pending_writes);
submit_bio(bio);
}
-void md_super_wait(struct mddev *mddev)
+int md_super_wait(struct mddev *mddev)
{
/* wait for all superblock writes that were scheduled to complete */
wait_event(mddev->sb_wait, atomic_read(&mddev->pending_writes)==0);
+ if (test_and_clear_bit(MD_NEED_REWRITE, &mddev->flags))
+ return -EAGAIN;
+ return 0;
}
int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
@@ -1333,9 +1353,10 @@ super_90_rdev_size_change(struct md_rdev *rdev, sector_t num_sectors)
if (IS_ENABLED(CONFIG_LBDAF) && (u64)num_sectors >= (2ULL << 32) &&
rdev->mddev->level >= 1)
num_sectors = (sector_t)(2ULL << 32) - 2;
- md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
+ do
+ md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
rdev->sb_page);
- md_super_wait(rdev->mddev);
+ while (md_super_wait(rdev->mddev) < 0);
return num_sectors;
}
@@ -1876,9 +1897,10 @@ super_1_rdev_size_change(struct md_rdev *rdev, sector_t num_sectors)
sb->data_size = cpu_to_le64(num_sectors);
sb->super_offset = rdev->sb_start;
sb->sb_csum = calc_sb_1_csum(sb);
- md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
- rdev->sb_page);
- md_super_wait(rdev->mddev);
+ do
+ md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
+ rdev->sb_page);
+ while (md_super_wait(rdev->mddev) < 0);
return num_sectors;
}
@@ -2413,6 +2435,7 @@ void md_update_sb(struct mddev *mddev, int force_change)
pr_debug("md: updating %s RAID superblock on device (in sync %d)\n",
mdname(mddev), mddev->in_sync);
+rewrite:
bitmap_update_sb(mddev->bitmap);
rdev_for_each(rdev, mddev) {
char b[BDEVNAME_SIZE];
@@ -2444,7 +2467,8 @@ void md_update_sb(struct mddev *mddev, int force_change)
/* only need to write one superblock... */
break;
}
- md_super_wait(mddev);
+ if (md_super_wait(mddev) < 0)
+ goto rewrite;
/* if there was a failure, MD_CHANGE_DEVS was set, and we re-write super */
if (mddev_is_clustered(mddev) && ret == 0)
diff --git a/drivers/md/md.h b/drivers/md/md.h
index bc6712ef8c81..5c08f84101fa 100644
--- a/drivers/md/md.h
+++ b/drivers/md/md.h
@@ -30,6 +30,16 @@
#define MaxSector (~(sector_t)0)
/*
+ * These flags should really be called "NO_RETRY" rather than
+ * "FAILFAST" because they don't make any promise about time lapse,
+ * only about the number of retries, which will be zero.
+ * REQ_FAILFAST_DRIVER is not included because
+ * Commit: 4a27446f3e39 ("[SCSI] modify scsi to handle new fail fast flags.")
+ * seems to suggest that the errors it avoids retrying should usually
+ * be retried.
+ */
+#define MD_FAILFAST (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT)
+/*
* MD's 'extended' device
*/
struct md_rdev {
@@ -177,6 +187,10 @@ enum flag_bits {
* It is expects that no bad block log
* is present.
*/
+ LastDev, /* Seems to be the last working dev as
+ * it didn't fail, so don't use FailFast
+ * any more for metadata
+ */
};
static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
@@ -213,6 +227,11 @@ enum mddev_flags {
MD_CLUSTER_RESYNC_LOCKED, /* cluster raid only, which means node
* already took resync lock, need to
* release the lock */
+ MD_FAILFAST_SUPPORTED, /* Using MD_FAILFAST on metadata writes is
+ * supported as calls to md_error() will
+ * never cause the array to become failed.
+ */
+ MD_NEED_REWRITE, /* metadata write needs to be repeated */
};
#define MD_UPDATE_SB_FLAGS (BIT(MD_CHANGE_DEVS) | \
BIT(MD_CHANGE_CLEAN) | \
@@ -628,7 +647,7 @@ extern int mddev_congested(struct mddev *mddev, int bits);
extern void md_flush_request(struct mddev *mddev, struct bio *bio);
extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
sector_t sector, int size, struct page *page);
-extern void md_super_wait(struct mddev *mddev);
+extern int md_super_wait(struct mddev *mddev);
extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
struct page *page, int op, int op_flags,
bool metadata_op);
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index 96474f0af1b8..cfd73730e6af 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -2989,6 +2989,7 @@ static int raid1_run(struct mddev *mddev)
mddev->thread = conf->thread;
conf->thread = NULL;
mddev->private = conf;
+ set_bit(MD_FAILFAST_SUPPORTED, &mddev->flags);
md_set_array_sectors(mddev, raid1_size(mddev, 0, 0));
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index b53610c59104..763ca45b6b32 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -3728,6 +3728,7 @@ static int raid10_run(struct mddev *mddev)
size = raid10_size(mddev, 0, 0);
md_set_array_sectors(mddev, size);
mddev->resync_max_sectors = size;
+ set_bit(MD_FAILFAST_SUPPORTED, &mddev->flags);
if (mddev->queue) {
int stripe = conf->geo.raid_disks *
^ permalink raw reply related
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox