linux-raid.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Jonathan Derrick <jonathan.derrick@linux.dev>
To: Song Liu <song@kernel.org>
Cc: <linux-raid@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
	jonathan.derrick@solidigm.com, jonathanx.sk.derrick@intel.com,
	Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>,
	Jonathan Derrick <jonathan.derrick@linux.dev>
Subject: [PATCH v2 1/3] md/bitmap: Add chunk-threshold unplugging
Date: Thu, 13 Oct 2022 16:41:49 -0600	[thread overview]
Message-ID: <20221013224151.300-2-jonathan.derrick@linux.dev> (raw)
In-Reply-To: <20221013224151.300-1-jonathan.derrick@linux.dev>

Add a mechanism to allow bitmap unplugging and flushing to wait until it
has surpassed a defined threshold of dirty chunks. This allows certain
high I/O write workloads to make good forward progress between bitmap
updates or provide reliable bitmap consistency. The default behavior is
previous behavior of always unplugging when called.

Signed-off-by: Jonathan Derrick <jonathan.derrick@linux.dev>
---
 drivers/md/md-bitmap.c | 35 +++++++++++++++++++++++++++++++----
 drivers/md/md-bitmap.h |  1 +
 drivers/md/md.h        |  1 +
 3 files changed, 33 insertions(+), 4 deletions(-)

diff --git a/drivers/md/md-bitmap.c b/drivers/md/md-bitmap.c
index bf6dffadbe6f..c5c77f8371a8 100644
--- a/drivers/md/md-bitmap.c
+++ b/drivers/md/md-bitmap.c
@@ -1004,7 +1004,7 @@ static int md_bitmap_file_test_bit(struct bitmap *bitmap, sector_t block)
 /* this gets called when the md device is ready to unplug its underlying
  * (slave) device queues -- before we let any writes go down, we need to
  * sync the dirty pages of the bitmap file to disk */
-void md_bitmap_unplug(struct bitmap *bitmap)
+static void __md_bitmap_unplug(struct bitmap *bitmap)
 {
 	unsigned long i;
 	int dirty, need_write;
@@ -1038,6 +1038,33 @@ void md_bitmap_unplug(struct bitmap *bitmap)
 	if (test_bit(BITMAP_WRITE_ERROR, &bitmap->flags))
 		md_bitmap_file_kick(bitmap);
 }
+
+/*
+ * Conditional unplug based on user-defined parameter
+ * Defaults to unconditional behavior
+ */
+void md_bitmap_unplug(struct bitmap *bitmap)
+{
+	unsigned int flush_threshold = bitmap->mddev->bitmap_info.flush_threshold;
+
+	if (!flush_threshold) {
+		__md_bitmap_unplug(bitmap);
+	} else {
+		struct bitmap_page *bp = bitmap->counts.bp;
+		unsigned long pages = bitmap->counts.pages;
+		unsigned long k, count = 0;
+
+		for (k = 0; k < pages; k++)
+			if (bp[k].map && !bp[k].hijacked)
+				count += bp[k].count;
+
+		if (count - bitmap->unplugged_count > flush_threshold) {
+			bitmap->unplugged_count = count;
+			md_bitmap_daemon_work(&bitmap->mddev->daemon_timer);
+			__md_bitmap_unplug(bitmap);
+		}
+	}
+}
 EXPORT_SYMBOL(md_bitmap_unplug);
 
 static void md_bitmap_set_memory_bits(struct bitmap *bitmap, sector_t offset, int needed);
@@ -2012,9 +2039,9 @@ int md_bitmap_copy_from_slot(struct mddev *mddev, int slot,
 		for (i = 0; i < bitmap->storage.file_pages; i++)
 			if (test_page_attr(bitmap, i, BITMAP_PAGE_PENDING))
 				set_page_attr(bitmap, i, BITMAP_PAGE_NEEDWRITE);
-		md_bitmap_unplug(bitmap);
+		__md_bitmap_unplug(bitmap);
 	}
-	md_bitmap_unplug(mddev->bitmap);
+	__md_bitmap_unplug(mddev->bitmap);
 	*low = lo;
 	*high = hi;
 	md_bitmap_free(bitmap);
@@ -2246,7 +2273,7 @@ int md_bitmap_resize(struct bitmap *bitmap, sector_t blocks,
 	spin_unlock_irq(&bitmap->counts.lock);
 
 	if (!init) {
-		md_bitmap_unplug(bitmap);
+		__md_bitmap_unplug(bitmap);
 		bitmap->mddev->pers->quiesce(bitmap->mddev, 0);
 	}
 	ret = 0;
diff --git a/drivers/md/md-bitmap.h b/drivers/md/md-bitmap.h
index cfd7395de8fd..49a93d8ff307 100644
--- a/drivers/md/md-bitmap.h
+++ b/drivers/md/md-bitmap.h
@@ -223,6 +223,7 @@ struct bitmap {
 	unsigned long daemon_lastrun; /* jiffies of last run */
 	unsigned long last_end_sync; /* when we lasted called end_sync to
 				      * update bitmap with resync progress */
+	unsigned long unplugged_count; /* last dirty count from md_bitmap_unplug */
 
 	atomic_t pending_writes; /* pending writes to the bitmap file */
 	wait_queue_head_t write_wait;
diff --git a/drivers/md/md.h b/drivers/md/md.h
index b4e2d8b87b61..1a558cb18bd4 100644
--- a/drivers/md/md.h
+++ b/drivers/md/md.h
@@ -501,6 +501,7 @@ struct mddev {
 		int			external;
 		int			nodes; /* Maximum number of nodes in the cluster */
 		char                    cluster_name[64]; /* Name of the cluster */
+		unsigned int		flush_threshold; /* how many dirty chunks between updates */
 	} bitmap_info;
 
 	atomic_t			max_corr_read_errors; /* max read retries */
-- 
2.31.1


  reply	other threads:[~2022-10-13 22:45 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-10-13 22:41 [PATCH v2 0/3] Bitmap percentage flushing Jonathan Derrick
2022-10-13 22:41 ` Jonathan Derrick [this message]
2022-10-14  1:11   ` [PATCH v2 1/3] md/bitmap: Add chunk-threshold unplugging Jonathan Derrick
2022-10-13 22:41 ` [PATCH v2 2/3] md/bitmap: Add sysfs interface for flush threshold Jonathan Derrick
2022-10-13 22:41 ` [PATCH v2 3/3] md/bitmap: Convert daemon_work to proper timer Jonathan Derrick
2022-10-14 21:10 ` [PATCH v2 0/3] Bitmap percentage flushing John Stoffel
2022-10-15 22:27   ` Jonathan Derrick

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20221013224151.300-2-jonathan.derrick@linux.dev \
    --to=jonathan.derrick@linux.dev \
    --cc=jonathan.derrick@solidigm.com \
    --cc=jonathanx.sk.derrick@intel.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-raid@vger.kernel.org \
    --cc=mariusz.tkaczyk@linux.intel.com \
    --cc=song@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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).