All of lore.kernel.org
 help / color / mirror / Atom feed
* Useless test in alloc_multiple_bios
@ 2024-05-15 18:15 Mikulas Patocka
  2024-05-16  9:19 ` Ming Lei
  0 siblings, 1 reply; 2+ messages in thread
From: Mikulas Patocka @ 2024-05-15 18:15 UTC (permalink / raw)
  To: Mike Snitzer, Ming Lei; +Cc: dm-devel

Hi

I found this piece of code in alloc_multiple_bios:
int try = (gfp_flag & GFP_NOWAIT) ? 0 : 1;

The problem is that GFP_NOWAIT includes __GFP_KSWAPD_RECLAIM and GFP_IO 
also includes __GFP_KSWAPD_RECLAIM - so the test always returns true and 
we always start with try = 0. This code was introduced by the commit 
4a2fe2960891f1ccd7805d0973284fd44c2f12b4.

I am inclined to remove this logic at all and always start with try = 0; 
- trying to allocate bios first with GFP_NOWAIT makes no harm (and it 
improves performance because we don't need to grab the lock) and if the 
allocation fails, it is retried with GFP_NOIO.

I'd like to ask - if you think that this piece of code has some purpose, 
please describe the purpose.

Mikulas


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: Useless test in alloc_multiple_bios
  2024-05-15 18:15 Useless test in alloc_multiple_bios Mikulas Patocka
@ 2024-05-16  9:19 ` Ming Lei
  0 siblings, 0 replies; 2+ messages in thread
From: Ming Lei @ 2024-05-16  9:19 UTC (permalink / raw)
  To: Mikulas Patocka; +Cc: Mike Snitzer, dm-devel, ming.lei

On Wed, May 15, 2024 at 08:15:20PM +0200, Mikulas Patocka wrote:
> Hi
> 
> I found this piece of code in alloc_multiple_bios:
> int try = (gfp_flag & GFP_NOWAIT) ? 0 : 1;

I guess the above check is supposed to be (gfp_flag == GFP_NOWAIT)?

> 
> The problem is that GFP_NOWAIT includes __GFP_KSWAPD_RECLAIM and GFP_IO 
> also includes __GFP_KSWAPD_RECLAIM - so the test always returns true and 
> we always start with try = 0. This code was introduced by the commit 
> 4a2fe2960891f1ccd7805d0973284fd44c2f12b4.
> 
> I am inclined to remove this logic at all and always start with try = 0; 
> - trying to allocate bios first with GFP_NOWAIT makes no harm (and it 
> improves performance because we don't need to grab the lock) and if the 
> allocation fails, it is retried with GFP_NOIO.

I think it is fine to simplify the logic in this way, and now GFP_NOWAIT is
only for sending empty flush, which could be from fast path, and abnormal IO
is supposed to be in slow path.

Thanks,
Ming


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2024-05-16  9:19 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2024-05-15 18:15 Useless test in alloc_multiple_bios Mikulas Patocka
2024-05-16  9:19 ` Ming Lei

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.