Re: commit 762380a "block: add notion of a chunk size for request merging" stops io on btrfs

[Jens Axboe]

On 2014-06-17 14:35, Konstantinos Skarlatos wrote:
> Hi all,
> with 3.16-rc1  rsync stops writing to my btrfs filesystem and stays at a
> D+ state.
> git bisect showed that the problematic commit is:
>
> 762380ad9322951cea4ce9d24864265f9c66a916 is the first bad commit
> commit 762380ad9322951cea4ce9d24864265f9c66a916
> Author: Jens Axboe <axboe@xxxxxx>
> Date:   Thu Jun 5 13:38:39 2014 -0600
>
>      block: add notion of a chunk size for request merging
>
>      Some drivers have different limits on what size a request should
>      optimally be, depending on the offset of the request. Similar to
>      dividing a device into chunks. Add a setting that allows the driver
>      to inform the block layer of such a chunk size. The block layer will
>      then prevent merging across the chunks.
>
>      This is needed to optimally support NVMe with a non-zero stripe size.
>
>      Signed-off-by: Jens Axboe <axboe@xxxxxx>

That's odd, should not have any effect since nobody enables stripe sizes 
in the kernel. I'll double check, perhaps it's not always being cleared.

Ah wait, does the attached help?


--
Jens Axboe

diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 31e11051f1ba..713f8b62b435 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -920,7 +920,7 @@ static inline unsigned int blk_max_size_offset(struct request_queue *q,
 					       sector_t offset)
 {
 	if (!q->limits.chunk_sectors)
-		return q->limits.max_hw_sectors;
+		return q->limits.max_sectors;
 
 	return q->limits.chunk_sectors -
 			(offset & (q->limits.chunk_sectors - 1));