[RFC]md: using GFP_NOIO to allocate bio for flush request

[RFC]md: using GFP_NOIO to allocate bio for flush request

[Shaohua Li]

flush request is issued in transaction commit code path usually, so 
looks using
GFP_KERNEL to allocate memory for flush request bio falls into the classic
deadlock issue. dm already gets it right.

Signed-off-by: Shaohua Li <shli@fusionio.com>
---
  drivers/md/md.c |    2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

Index: linux/drivers/md/md.c
===================================================================
--- linux.orig/drivers/md/md.c	2012-04-13 10:03:58.395497650 +0800
+++ linux/drivers/md/md.c	2012-04-13 10:04:37.445497882 +0800
@@ -450,7 +450,7 @@ static void submit_flushes(struct work_s
  			atomic_inc(&rdev->nr_pending);
  			atomic_inc(&rdev->nr_pending);
  			rcu_read_unlock();
-			bi = bio_alloc_mddev(GFP_KERNEL, 0, mddev);
+			bi = bio_alloc_mddev(GFP_NOIO, 0, mddev);
  			bi->bi_end_io = md_end_flush;
  			bi->bi_private = rdev;
  			bi->bi_bdev = rdev->bdev;

Re: [RFC]md: using GFP_NOIO to allocate bio for flush request

[Shaohua Li]

On Mon, Apr 16, 2012 at 01:32:06PM +0800, Shaohua Li wrote:
> 
> flush request is issued in transaction commit code path usually, so
> looks using
> GFP_KERNEL to allocate memory for flush request bio falls into the classic
> deadlock issue. dm already gets it right.
ping!


> Signed-off-by: Shaohua Li <shli@fusionio.com>
> ---
>  drivers/md/md.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> Index: linux/drivers/md/md.c
> ===================================================================
> --- linux.orig/drivers/md/md.c	2012-04-13 10:03:58.395497650 +0800
> +++ linux/drivers/md/md.c	2012-04-13 10:04:37.445497882 +0800
> @@ -450,7 +450,7 @@ static void submit_flushes(struct work_s
>  			atomic_inc(&rdev->nr_pending);
>  			atomic_inc(&rdev->nr_pending);
>  			rcu_read_unlock();
> -			bi = bio_alloc_mddev(GFP_KERNEL, 0, mddev);
> +			bi = bio_alloc_mddev(GFP_NOIO, 0, mddev);
>  			bi->bi_end_io = md_end_flush;
>  			bi->bi_private = rdev;
>  			bi->bi_bdev = rdev->bdev;

Re: [RFC]md: using GFP_NOIO to allocate bio for flush request

[Dan Williams]

On Tue, May 8, 2012 at 3:18 AM, Shaohua Li <shli@kernel.org> wrote:
> On Mon, Apr 16, 2012 at 01:32:06PM +0800, Shaohua Li wrote:
>>
>> flush request is issued in transaction commit code path usually, so
>> looks using
>> GFP_KERNEL to allocate memory for flush request bio falls into the classic
>> deadlock issue. dm already gets it right.
> ping!
>

This one should probably also be queued for -stable right?

Re: [RFC]md: using GFP_NOIO to allocate bio for flush request

[NeilBrown]

[-- Attachment #1: Type: text/plain, Size: 720 bytes --]

On Tue, 8 May 2012 12:31:18 -0700 Dan Williams <dan.j.williams@intel.com>
wrote:

> On Tue, May 8, 2012 at 3:18 AM, Shaohua Li <shli@kernel.org> wrote:
> > On Mon, Apr 16, 2012 at 01:32:06PM +0800, Shaohua Li wrote:
> >>
> >> flush request is issued in transaction commit code path usually, so
> >> looks using
> >> GFP_KERNEL to allocate memory for flush request bio falls into the classic
> >> deadlock issue. dm already gets it right.
> > ping!
> >
> 
> This one should probably also be queued for -stable right?

I suspect so.

I've edited the commit-log message slightly and added a 'cc' for stable,
It is now in my for-next branch.  I'll push to Linus next week.

Thanks Shaohua,

NeilBrown

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]