[Cluster-devel] [PATCH] gfs2: Fix extended attribute readahead optimization

[Cluster-devel] [PATCH] gfs2: Fix extended attribute readahead optimization

[Andreas Gruenbacher]

Commit 39b0555f didn't check for a failing bio_add_page in
gfs2_submit_bhs. This could cause I/O requests to get lost, and the
affected buffer heads to stay locked forever.  Fix that by submitting
the current bio and allocating another one when bio_add_page fails.  (It
is guaranteed that we can at least add one page to a bio.)

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
---
 fs/gfs2/meta_io.c | 35 +++++++++++++++++++----------------
 1 file changed, 19 insertions(+), 16 deletions(-)

diff --git a/fs/gfs2/meta_io.c b/fs/gfs2/meta_io.c
index 950b8be..373639a5 100644
--- a/fs/gfs2/meta_io.c
+++ b/fs/gfs2/meta_io.c
@@ -216,23 +216,26 @@ static void gfs2_meta_read_endio(struct bio *bio)
 static void gfs2_submit_bhs(int op, int op_flags, struct buffer_head *bhs[],
 			    int num)
 {
-	struct buffer_head *bh = bhs[0];
-	struct bio *bio;
-	int i;
-
-	if (!num)
-		return;
-
-	bio = bio_alloc(GFP_NOIO, num);
-	bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9);
-	bio->bi_bdev = bh->b_bdev;
-	for (i = 0; i < num; i++) {
-		bh = bhs[i];
-		bio_add_page(bio, bh->b_page, bh->b_size, bh_offset(bh));
+	while (num > 0) {
+		struct buffer_head *bh = *bhs;
+		struct bio *bio;
+
+		bio = bio_alloc(GFP_NOIO, num);
+		bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9);
+		bio->bi_bdev = bh->b_bdev;
+		while (num > 0) {
+			bh = *bhs;
+			if (!bio_add_page(bio, bh->b_page, bh->b_size, bh_offset(bh))) {
+				BUG_ON(bio->bi_iter.bi_size == 0);
+				break;
+			}
+			bhs++;
+			num--;
+		}
+		bio->bi_end_io = gfs2_meta_read_endio;
+		bio_set_op_attrs(bio, op, op_flags);
+		submit_bio(bio);
 	}
-	bio->bi_end_io = gfs2_meta_read_endio;
-	bio_set_op_attrs(bio, op, op_flags);
-	submit_bio(bio);
 }
 
 /**
-- 
2.7.4

[Cluster-devel] [PATCH] gfs2: Fix extended attribute readahead optimization

[Bob Peterson]

----- Original Message -----
| Commit 39b0555f didn't check for a failing bio_add_page in
| gfs2_submit_bhs. This could cause I/O requests to get lost, and the
| affected buffer heads to stay locked forever.  Fix that by submitting
| the current bio and allocating another one when bio_add_page fails.  (It
| is guaranteed that we can at least add one page to a bio.)
| 
| Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
| ---
|  fs/gfs2/meta_io.c | 35 +++++++++++++++++++----------------
|  1 file changed, 19 insertions(+), 16 deletions(-)

Hi,

Thanks. This is now applied to the for-next branch of the linux-gfs2 tree:
https://git.kernel.org/cgit/linux/kernel/git/gfs2/linux-gfs2.git/commit/fs/gfs2?h=for-next&id=23e5671a79be00b2f1c895aa93ff40fb75a4647e

Regards,

Bob Peterson
Red Hat File Systems

[Cluster-devel] [PATCH] gfs2: Fix extended attribute readahead optimization

[Andreas Gruenbacher]

On Thu, Aug 18, 2016 at 7:06 PM, Bob Peterson <rpeterso@redhat.com> wrote:
> ----- Original Message -----
> | Commit 39b0555f didn't check for a failing bio_add_page in
> | gfs2_submit_bhs. This could cause I/O requests to get lost, and the
> | affected buffer heads to stay locked forever.  Fix that by submitting
> | the current bio and allocating another one when bio_add_page fails.  (It
> | is guaranteed that we can at least add one page to a bio.)
> |
> | Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
> | ---
> |  fs/gfs2/meta_io.c | 35 +++++++++++++++++++----------------
> |  1 file changed, 19 insertions(+), 16 deletions(-)
>
> Hi,
>
> Thanks. This is now applied to the for-next branch of the linux-gfs2 tree:
> https://git.kernel.org/cgit/linux/kernel/git/gfs2/linux-gfs2.git/commit/fs/gfs2?h=for-next&id=23e5671a79be00b2f1c895aa93ff40fb75a4647e

Should this fix be pushed upstream for 4.8 as well?

Thanks,
Andreas