* [PATCH v2] mm: honor FGP_NOWAIT for page cache page allocation
@ 2022-07-02 15:43 Jens Axboe
2022-07-03 5:41 ` Shakeel Butt
2022-07-03 8:37 ` Matthew Wilcox
0 siblings, 2 replies; 4+ messages in thread
From: Jens Axboe @ 2022-07-02 15:43 UTC (permalink / raw)
To: Matthew Wilcox, linux-mm, Shakeel Butt; +Cc: Stefan Roesch
If we're creating a page cache page with FGP_CREAT but FGP_NOWAIT is
set, we should dial back the gfp flags to avoid frivolous blocking
which is trivial to hit in low memory conditions:
[ 10.117661] __schedule+0x8c/0x550
[ 10.118305] schedule+0x58/0xa0
[ 10.118897] schedule_timeout+0x30/0xdc
[ 10.119610] __wait_for_common+0x88/0x114
[ 10.120348] wait_for_completion+0x1c/0x24
[ 10.121103] __flush_work.isra.0+0x16c/0x19c
[ 10.121896] flush_work+0xc/0x14
[ 10.122496] __drain_all_pages+0x144/0x218
[ 10.123267] drain_all_pages+0x10/0x18
[ 10.123941] __alloc_pages+0x464/0x9e4
[ 10.124633] __folio_alloc+0x18/0x3c
[ 10.125294] __filemap_get_folio+0x17c/0x204
[ 10.126084] iomap_write_begin+0xf8/0x428
[ 10.126829] iomap_file_buffered_write+0x144/0x24c
[ 10.127710] xfs_file_buffered_write+0xe8/0x248
[ 10.128553] xfs_file_write_iter+0xa8/0x120
[ 10.129324] io_write+0x16c/0x38c
[ 10.129940] io_issue_sqe+0x70/0x1cc
[ 10.130617] io_queue_sqe+0x18/0xfc
[ 10.131277] io_submit_sqes+0x5d4/0x600
[ 10.131946] __arm64_sys_io_uring_enter+0x224/0x600
[ 10.132752] invoke_syscall.constprop.0+0x70/0xc0
[ 10.133616] do_el0_svc+0xd0/0x118
[ 10.134238] el0_svc+0x78/0xa0
Clear IO, FS, and reclaim flags and mark the allocation as GFP_NOWAIT and
add __GFP_NOWARN to avoid polluting dmesg with pointless allocations
failures. A caller with FGP_NOWAIT must be expected to handle the
resulting -EAGAIN return and retry from a suitable context without NOWAIT
set.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
V2:
- Improve commit message
- Add GFP_NOWAIT to allow kswapd wakeup
mm/filemap.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/mm/filemap.c b/mm/filemap.c
index ffdfbc8b0e3c..254931a6e3ed 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1988,6 +1988,10 @@ struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
gfp |= __GFP_WRITE;
if (fgp_flags & FGP_NOFS)
gfp &= ~__GFP_FS;
+ if (fgp_flags & FGP_NOWAIT) {
+ gfp &= ~GFP_KERNEL;
+ gfp |= GFP_NOWAIT | __GFP_NOWARN;
+ }
folio = filemap_alloc_folio(gfp, 0);
if (!folio)
--
2.35.1
--
Jens Axboe
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH v2] mm: honor FGP_NOWAIT for page cache page allocation
2022-07-02 15:43 [PATCH v2] mm: honor FGP_NOWAIT for page cache page allocation Jens Axboe
@ 2022-07-03 5:41 ` Shakeel Butt
2022-07-03 8:37 ` Matthew Wilcox
1 sibling, 0 replies; 4+ messages in thread
From: Shakeel Butt @ 2022-07-03 5:41 UTC (permalink / raw)
To: Jens Axboe; +Cc: Matthew Wilcox, linux-mm, Stefan Roesch
On Sat, Jul 02, 2022 at 09:43:32AM -0600, Jens Axboe wrote:
> If we're creating a page cache page with FGP_CREAT but FGP_NOWAIT is
> set, we should dial back the gfp flags to avoid frivolous blocking
> which is trivial to hit in low memory conditions:
>
> [ 10.117661] __schedule+0x8c/0x550
> [ 10.118305] schedule+0x58/0xa0
> [ 10.118897] schedule_timeout+0x30/0xdc
> [ 10.119610] __wait_for_common+0x88/0x114
> [ 10.120348] wait_for_completion+0x1c/0x24
> [ 10.121103] __flush_work.isra.0+0x16c/0x19c
> [ 10.121896] flush_work+0xc/0x14
> [ 10.122496] __drain_all_pages+0x144/0x218
> [ 10.123267] drain_all_pages+0x10/0x18
> [ 10.123941] __alloc_pages+0x464/0x9e4
> [ 10.124633] __folio_alloc+0x18/0x3c
> [ 10.125294] __filemap_get_folio+0x17c/0x204
> [ 10.126084] iomap_write_begin+0xf8/0x428
> [ 10.126829] iomap_file_buffered_write+0x144/0x24c
> [ 10.127710] xfs_file_buffered_write+0xe8/0x248
> [ 10.128553] xfs_file_write_iter+0xa8/0x120
> [ 10.129324] io_write+0x16c/0x38c
> [ 10.129940] io_issue_sqe+0x70/0x1cc
> [ 10.130617] io_queue_sqe+0x18/0xfc
> [ 10.131277] io_submit_sqes+0x5d4/0x600
> [ 10.131946] __arm64_sys_io_uring_enter+0x224/0x600
> [ 10.132752] invoke_syscall.constprop.0+0x70/0xc0
> [ 10.133616] do_el0_svc+0xd0/0x118
> [ 10.134238] el0_svc+0x78/0xa0
>
> Clear IO, FS, and reclaim flags and mark the allocation as GFP_NOWAIT and
> add __GFP_NOWARN to avoid polluting dmesg with pointless allocations
> failures. A caller with FGP_NOWAIT must be expected to handle the
> resulting -EAGAIN return and retry from a suitable context without NOWAIT
> set.
>
> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH v2] mm: honor FGP_NOWAIT for page cache page allocation
2022-07-02 15:43 [PATCH v2] mm: honor FGP_NOWAIT for page cache page allocation Jens Axboe
2022-07-03 5:41 ` Shakeel Butt
@ 2022-07-03 8:37 ` Matthew Wilcox
2022-07-03 12:51 ` Jens Axboe
1 sibling, 1 reply; 4+ messages in thread
From: Matthew Wilcox @ 2022-07-03 8:37 UTC (permalink / raw)
To: Jens Axboe; +Cc: linux-mm, Shakeel Butt, Stefan Roesch
On Sat, Jul 02, 2022 at 09:43:32AM -0600, Jens Axboe wrote:
> +++ b/mm/filemap.c
> @@ -1988,6 +1988,10 @@ struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
> gfp |= __GFP_WRITE;
> if (fgp_flags & FGP_NOFS)
> gfp &= ~__GFP_FS;
> + if (fgp_flags & FGP_NOWAIT) {
> + gfp &= ~GFP_KERNEL;
> + gfp |= GFP_NOWAIT | __GFP_NOWARN;
> + }
Wouldn't it be simpler to write:
if (fgp_flags & FGP_NOWAIT) {
gfp &= ~__GFP_DIRECT_RECLAIM;
gfp |= __GFP_NOWARN;
}
(I'd prefer it if the caller passed in the GFP flags that it actually
wants, but looking at the patches that prompted this, that seems rather
annoying to do)
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH v2] mm: honor FGP_NOWAIT for page cache page allocation
2022-07-03 8:37 ` Matthew Wilcox
@ 2022-07-03 12:51 ` Jens Axboe
0 siblings, 0 replies; 4+ messages in thread
From: Jens Axboe @ 2022-07-03 12:51 UTC (permalink / raw)
To: Matthew Wilcox; +Cc: linux-mm, Shakeel Butt, Stefan Roesch
On 7/3/22 2:37 AM, Matthew Wilcox wrote:
> On Sat, Jul 02, 2022 at 09:43:32AM -0600, Jens Axboe wrote:
>> +++ b/mm/filemap.c
>> @@ -1988,6 +1988,10 @@ struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
>> gfp |= __GFP_WRITE;
>> if (fgp_flags & FGP_NOFS)
>> gfp &= ~__GFP_FS;
>> + if (fgp_flags & FGP_NOWAIT) {
>> + gfp &= ~GFP_KERNEL;
>> + gfp |= GFP_NOWAIT | __GFP_NOWARN;
>> + }
>
> Wouldn't it be simpler to write:
>
> if (fgp_flags & FGP_NOWAIT) {
> gfp &= ~__GFP_DIRECT_RECLAIM;
> gfp |= __GFP_NOWARN;
> }
That won't clear IO or FS, though? We really just want GFP_NOWAIT |
__GFP_NOWARN here, but didn't want to make assumptions on whatever else
would be set in gfp already.
> (I'd prefer it if the caller passed in the GFP flags that it actually
> wants, but looking at the patches that prompted this, that seems rather
> annoying to do)
Yes me too, but that's a lot of annoying churn...
--
Jens Axboe
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2022-07-03 12:51 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-07-02 15:43 [PATCH v2] mm: honor FGP_NOWAIT for page cache page allocation Jens Axboe
2022-07-03 5:41 ` Shakeel Butt
2022-07-03 8:37 ` Matthew Wilcox
2022-07-03 12:51 ` Jens Axboe
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.