diff mbox series

Reply: [PATCH v3] dm verity: don't prefetch hash blocks for already-verified data

Message ID 6b185db43b044bd9ae96036825dee631@transsion.com (mailing list archive)
State Not Applicable, archived
Delegated to: Mike Snitzer
Headers show
Series Reply: [PATCH v3] dm verity: don't prefetch hash blocks for already-verified data | expand

Commit Message

xianrong.zhou(周先荣) Jan. 7, 2020, 6:58 a.m. UTC 
Hey Eric:

	Thank you very much. I am quit happy you improvement of and send this patch.

Subject: [PATCH v3] dm verity: don't prefetch hash blocks for already-verified data

From: "xianrong.zhou" <xianrong.zhou@transsion.com>

Try to skip prefetching hash blocks that won't be needed due to the "check_at_most_once" option being enabled and the corresponding data blocks already having been verified.

Since prefetching operates on a range of data blocks, do this by just trimming the two ends of the range.  This doesn't skip every unneeded hash block, since data blocks in the middle of the range could also be unneeded, and hash blocks are still prefetched in large clusters as controlled by dm_verity_prefetch_cluster.  But it can still help a lot.

In a test on Android Q launching 91 apps every 15s repeated 21 times, prefetching was only done for 447177/4776629 = 9.36% of data blocks.

Tested-by: ruxian.feng <ruxian.feng@transsion.com>
Co-developed-by: yuanjiong.gao <yuanjiong.gao@transsion.com>
Signed-off-by: yuanjiong.gao <yuanjiong.gao@transsion.com>
Signed-off-by: xianrong.zhou <xianrong.zhou@transsion.com>
[EB: simplified the 'while' loops and improved the commit message]
Signed-off-by: Eric Biggers <ebiggers@google.com>
---
 drivers/md/dm-verity-target.c | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

 
 	INIT_WORK(&pw->work, verity_prefetch_io);
 	pw->v = v;
-	pw->block = io->block;
-	pw->n_blocks = io->n_blocks;
+	pw->block = block;
+	pw->n_blocks = n_blocks;
 	queue_work(v->verify_wq, &pw->work);
 }
 
--
2.24.1


--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel

Comments

Mike Snitzer Jan. 7, 2020, 3:11 p.m. UTC | #1 
Please fix your mail client so that when you reply to a message you
don't start a new thread.


--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
diff mbox series

Patch

diff --git a/drivers/md/dm-verity-target.c b/drivers/md/dm-verity-target.c index 4fb33e7562c5..0d61e9c67986 100644
--- a/drivers/md/dm-verity-target.c
+++ b/drivers/md/dm-verity-target.c
@@ -611,8 +611,22 @@  static void verity_prefetch_io(struct work_struct *work)
 
 static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io)  {
+	sector_t block = io->block;
+	unsigned int n_blocks = io->n_blocks;
 	struct dm_verity_prefetch_work *pw;
 
+	if (v->validated_blocks) {
+		while (n_blocks && test_bit(block, v->validated_blocks)) {
+			block++;
+			n_blocks--;
+		}
+		while (n_blocks && test_bit(block + n_blocks - 1,
+					    v->validated_blocks))
+			n_blocks--;
+		if (!n_blocks)
+			return;
+	}
+
 	pw = kmalloc(sizeof(struct dm_verity_prefetch_work),
 		GFP_NOIO | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN);
 
@@ -621,8 +635,8 @@  static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io)