* [PATCH v2] md/r5cache: generate R5LOG_PAYLOAD_FLUSH
@ 2017-03-10 0:24 Song Liu
2017-03-10 5:03 ` Shaohua Li
0 siblings, 1 reply; 2+ messages in thread
From: Song Liu @ 2017-03-10 0:24 UTC (permalink / raw)
To: linux-raid; +Cc: shli, neilb, kernel-team, dan.j.williams, hch, Song Liu
In r5c_finish_stripe_write_out(), R5LOG_PAYLOAD_FLUSH is append to
log->current_io.
Appending R5LOG_PAYLOAD_FLUSH in quiesce needs extra writes to
journal. To simplify the logic, we just skip R5LOG_PAYLOAD_FLUSH in
quiesce.
Even R5LOG_PAYLOAD_FLUSH supports multiple stripes per payload.
However, current implementation is one stripe per R5LOG_PAYLOAD_FLUSH,
which is simpler.
Signed-off-by: Song Liu <songliubraving@fb.com>
---
drivers/md/raid5-cache.c | 48 ++++++++++++++++++++++++++++++++++++++++++++----
1 file changed, 44 insertions(+), 4 deletions(-)
diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c
index e69f922..f148c68 100644
--- a/drivers/md/raid5-cache.c
+++ b/drivers/md/raid5-cache.c
@@ -588,10 +588,11 @@ static void r5l_log_endio(struct bio *bio)
bio_put(bio);
mempool_free(io->meta_page, log->meta_pool);
+ __r5l_set_io_unit_state(io, IO_UNIT_IO_END);
spin_lock_irqsave(&log->io_list_lock, flags);
- __r5l_set_io_unit_state(io, IO_UNIT_IO_END);
- if (log->need_cache_flush)
+
+ if (log->need_cache_flush && !list_empty(&io->stripe_list))
r5l_move_to_end_ios(log);
else
r5l_log_run_stripes(log);
@@ -619,9 +620,11 @@ static void r5l_log_endio(struct bio *bio)
bio_endio(bi);
atomic_dec(&io->pending_stripe);
}
- if (atomic_read(&io->pending_stripe) == 0)
- __r5l_stripe_write_finished(io);
}
+
+ /* finish flush only io_unit and PAYLOAD_FLUSH only io_unit */
+ if (list_empty(&io->stripe_list))
+ __r5l_stripe_write_finished(io);
}
static void r5l_do_submit_io(struct r5l_log *log, struct r5l_io_unit *io)
@@ -843,6 +846,41 @@ static void r5l_append_payload_page(struct r5l_log *log, struct page *page)
r5_reserve_log_entry(log, io);
}
+static void r5l_append_flush_payload(struct r5l_log *log, sector_t sect)
+{
+ struct mddev *mddev = log->rdev->mddev;
+ struct r5conf *conf = mddev->private;
+ struct r5l_io_unit *io;
+ struct r5l_payload_flush *payload;
+ int meta_size;
+
+ /*
+ * payload_flush requires extra writes to the journal.
+ * To avoid handling the extra IO in quiesce, just skip
+ * flush_payload
+ */
+ if (conf->quiesce)
+ return;
+
+ mutex_lock(&log->io_mutex);
+ meta_size = sizeof(struct r5l_payload_flush) + sizeof(__le64);
+
+ if (r5l_get_meta(log, meta_size)) {
+ mutex_unlock(&log->io_mutex);
+ return;
+ }
+
+ /* current implementation is one stripe per flush payload */
+ io = log->current_io;
+ payload = page_address(io->meta_page) + io->meta_offset;
+ payload->header.type = cpu_to_le16(R5LOG_PAYLOAD_FLUSH);
+ payload->header.flags = cpu_to_le16(0);
+ payload->size = cpu_to_le32(sizeof(__le64));
+ payload->flush_stripes[0] = cpu_to_le64(sect);
+ io->meta_offset += meta_size;
+ mutex_unlock(&log->io_mutex);
+}
+
static int r5l_log_stripe(struct r5l_log *log, struct stripe_head *sh,
int data_pages, int parity_pages)
{
@@ -2784,6 +2822,8 @@ void r5c_finish_stripe_write_out(struct r5conf *conf,
atomic_dec(&conf->r5c_flushing_full_stripes);
atomic_dec(&conf->r5c_cached_full_stripes);
}
+
+ r5l_append_flush_payload(log, sh->sector);
}
int
--
2.9.3
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [PATCH v2] md/r5cache: generate R5LOG_PAYLOAD_FLUSH
2017-03-10 0:24 [PATCH v2] md/r5cache: generate R5LOG_PAYLOAD_FLUSH Song Liu
@ 2017-03-10 5:03 ` Shaohua Li
0 siblings, 0 replies; 2+ messages in thread
From: Shaohua Li @ 2017-03-10 5:03 UTC (permalink / raw)
To: Song Liu; +Cc: linux-raid, shli, neilb, kernel-team, dan.j.williams, hch
On Thu, Mar 09, 2017 at 04:24:03PM -0800, Song Liu wrote:
> In r5c_finish_stripe_write_out(), R5LOG_PAYLOAD_FLUSH is append to
> log->current_io.
>
> Appending R5LOG_PAYLOAD_FLUSH in quiesce needs extra writes to
> journal. To simplify the logic, we just skip R5LOG_PAYLOAD_FLUSH in
> quiesce.
>
> Even R5LOG_PAYLOAD_FLUSH supports multiple stripes per payload.
> However, current implementation is one stripe per R5LOG_PAYLOAD_FLUSH,
> which is simpler.
>
> Signed-off-by: Song Liu <songliubraving@fb.com>
> ---
> drivers/md/raid5-cache.c | 48 ++++++++++++++++++++++++++++++++++++++++++++----
> 1 file changed, 44 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c
> index e69f922..f148c68 100644
> --- a/drivers/md/raid5-cache.c
> +++ b/drivers/md/raid5-cache.c
> @@ -588,10 +588,11 @@ static void r5l_log_endio(struct bio *bio)
>
> bio_put(bio);
> mempool_free(io->meta_page, log->meta_pool);
> + __r5l_set_io_unit_state(io, IO_UNIT_IO_END);
>
> spin_lock_irqsave(&log->io_list_lock, flags);
> - __r5l_set_io_unit_state(io, IO_UNIT_IO_END);
why move this out the lock?
> - if (log->need_cache_flush)
> +
> + if (log->need_cache_flush && !list_empty(&io->stripe_list))
> r5l_move_to_end_ios(log);
> else
> r5l_log_run_stripes(log);
> @@ -619,9 +620,11 @@ static void r5l_log_endio(struct bio *bio)
> bio_endio(bi);
> atomic_dec(&io->pending_stripe);
> }
> - if (atomic_read(&io->pending_stripe) == 0)
> - __r5l_stripe_write_finished(io);
> }
> +
> + /* finish flush only io_unit and PAYLOAD_FLUSH only io_unit */
> + if (list_empty(&io->stripe_list))
we don't hold lock here, so can't check the list. checking the pending_stripe
should work too and is safe.
Thanks,
Shaohua
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2017-03-10 5:03 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2017-03-10 0:24 [PATCH v2] md/r5cache: generate R5LOG_PAYLOAD_FLUSH Song Liu
2017-03-10 5:03 ` Shaohua Li
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).