Re: [Regression 3.16-rc kernel] -55% reduction in throughput for OLTP benchmark

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, 2014-07-29 at 09:49 -0700, Christoph Hellwig wrote:
> On Tue, Jul 29, 2014 at 09:35:02AM -0700, Tim Chen wrote:
> > On Tue, 2014-07-29 at 07:19 -0700, Christoph Hellwig wrote:
> > > Hi Tim,
> > > 
> > > I think the problem is that sdio.size doesn't get set set with
> > > the new dio code.  Please try the patch below to fix this by not
> > > relying on that field:
> > 
> > Thanks for your quick response.  We'll test this patch and let you
> > know the results.
> 
> There actually was a little bug in it, please try the version below
> instead:
> 

Christoph, we're glad to report that this patch does fix the problem.
Wonder if Al or someone can pick it up.  It is a large regression
that should not be left unfixed for 3.16.

Tim


> 
> diff --git a/fs/direct-io.c b/fs/direct-io.c
> index 194d0d1..17e39b0 100644
> --- a/fs/direct-io.c
> +++ b/fs/direct-io.c
> @@ -71,7 +71,6 @@ struct dio_submit {
>  					   been performed at the start of a
>  					   write */
>  	int pages_in_io;		/* approximate total IO pages */
> -	size_t	size;			/* total request size (doesn't change)*/
>  	sector_t block_in_file;		/* Current offset into the underlying
>  					   file in dio_block units. */
>  	unsigned blocks_available;	/* At block_in_file.  changes */
> @@ -1104,7 +1103,8 @@ do_blockdev_direct_IO(int rw, struct kiocb *iocb, struct inode *inode,
>  	unsigned blkbits = i_blkbits;
>  	unsigned blocksize_mask = (1 << blkbits) - 1;
>  	ssize_t retval = -EINVAL;
> -	loff_t end = offset + iov_iter_count(iter);
> +	size_t count = iov_iter_count(iter);
> +	loff_t end = offset + count;
>  	struct dio *dio;
>  	struct dio_submit sdio = { 0, };
>  	struct buffer_head map_bh = { 0, };
> @@ -1287,10 +1287,9 @@ do_blockdev_direct_IO(int rw, struct kiocb *iocb, struct inode *inode,
>  	 */
>  	BUG_ON(retval == -EIOCBQUEUED);
>  	if (dio->is_async && retval == 0 && dio->result &&
> -	    ((rw == READ) || (dio->result == sdio.size)))
> +	    (rw == READ || dio->result == count))
>  		retval = -EIOCBQUEUED;
> -
> -	if (retval != -EIOCBQUEUED)
> +	else
>  		dio_await_completion(dio);
>  
>  	if (drop_refcount(dio) == 0) {
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/


--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [Linux Ext4 Filesystem]     [Union Filesystem]     [Filesystem Testing]     [Ceph Users]     [Ecryptfs]     [AutoFS]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux Cachefs]     [Reiser Filesystem]     [Linux RAID]     [Samba]     [Device Mapper]     [CEPH Development]
  Powered by Linux