[v3] NFS: Fixup allocation flags for nfsiod's __GFP_NORETRY

[PATCH v3] NFS: Fixup allocation flags for nfsiod's __GFP_NORETRY

Posted by Benjamin Coddington 5 months, 1 week ago

If the NFS client is doing writeback from a workqueue context, avoid using
__GFP_NORETRY for allocations if the task has set PF_MEMALLOC_NOIO or
PF_MEMALLOC_NOFS.  The combination of these flags makes memory allocation
failures much more likely.

We've seen those allocation failures show up when the loopback driver is
doing writeback from a workqueue to a file on NFS, where memory allocation
failure results in errors or corruption within the loopback device's
filesystem.

Suggested-by: Trond Myklebust <trondmy@kernel.org>
Fixes: 0bae835b63c5 ("NFS: Avoid writeback threads getting stuck in mempool_alloc()")
Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
Reviewed-by: Laurence Oberman <loberman@redhat.com>
Tested-by: Laurence Oberman <loberman@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
---

	On V3: fix ugly return (Thanks Paulo), add Jeff's R-b
	On V2: add missing 'Fixes' and Laurence's R-b T-b

 fs/nfs/internal.h | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index 69c2c10ee658..d8f768254f16 100644
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -671,9 +671,12 @@ nfs_write_match_verf(const struct nfs_writeverf *verf,
 
 static inline gfp_t nfs_io_gfp_mask(void)
 {
-	if (current->flags & PF_WQ_WORKER)
-		return GFP_KERNEL | __GFP_NORETRY | __GFP_NOWARN;
-	return GFP_KERNEL;
+	gfp_t ret = current_gfp_context(GFP_KERNEL);
+
+	/* For workers __GFP_NORETRY only with __GFP_IO or __GFP_FS */
+	if ((current->flags & PF_WQ_WORKER) && ret == GFP_KERNEL)
+		ret |= __GFP_NORETRY | __GFP_NOWARN;
+	return ret;
 }
 
 /*
-- 
2.47.0

Re: [PATCH v3] NFS: Fixup allocation flags for nfsiod's __GFP_NORETRY

Posted by Benjamin Coddington 4 months, 3 weeks ago

Gentle ping on this one - let me know if anyone has further review or
critique.

Ben

On 9 Jul 2025, at 21:47, Benjamin Coddington wrote:

> If the NFS client is doing writeback from a workqueue context, avoid using
> __GFP_NORETRY for allocations if the task has set PF_MEMALLOC_NOIO or
> PF_MEMALLOC_NOFS.  The combination of these flags makes memory allocation
> failures much more likely.
>
> We've seen those allocation failures show up when the loopback driver is
> doing writeback from a workqueue to a file on NFS, where memory allocation
> failure results in errors or corruption within the loopback device's
> filesystem.
>
> Suggested-by: Trond Myklebust <trondmy@kernel.org>
> Fixes: 0bae835b63c5 ("NFS: Avoid writeback threads getting stuck in mempool_alloc()")
> Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
> Reviewed-by: Laurence Oberman <loberman@redhat.com>
> Tested-by: Laurence Oberman <loberman@redhat.com>
> Reviewed-by: Jeff Layton <jlayton@kernel.org>
> ---
>
> 	On V3: fix ugly return (Thanks Paulo), add Jeff's R-b
> 	On V2: add missing 'Fixes' and Laurence's R-b T-b
>
>  fs/nfs/internal.h | 9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
> index 69c2c10ee658..d8f768254f16 100644
> --- a/fs/nfs/internal.h
> +++ b/fs/nfs/internal.h
> @@ -671,9 +671,12 @@ nfs_write_match_verf(const struct nfs_writeverf *verf,
>
>  static inline gfp_t nfs_io_gfp_mask(void)
>  {
> -	if (current->flags & PF_WQ_WORKER)
> -		return GFP_KERNEL | __GFP_NORETRY | __GFP_NOWARN;
> -	return GFP_KERNEL;
> +	gfp_t ret = current_gfp_context(GFP_KERNEL);
> +
> +	/* For workers __GFP_NORETRY only with __GFP_IO or __GFP_FS */
> +	if ((current->flags & PF_WQ_WORKER) && ret == GFP_KERNEL)
> +		ret |= __GFP_NORETRY | __GFP_NOWARN;
> +	return ret;
>  }
>
>  /*
> -- 
> 2.47.0

Re: [PATCH v3] NFS: Fixup allocation flags for nfsiod's __GFP_NORETRY

Posted by Christoph Hellwig 5 months, 1 week ago

On Wed, Jul 09, 2025 at 09:47:43PM -0400, Benjamin Coddington wrote:
> If the NFS client is doing writeback from a workqueue context, avoid using
> __GFP_NORETRY for allocations if the task has set PF_MEMALLOC_NOIO or
> PF_MEMALLOC_NOFS.  The combination of these flags makes memory allocation
> failures much more likely.

Can we take a step back and figre out why this blanket usage of
__GFP_NORETRY exists at all?

Re: [PATCH v3] NFS: Fixup allocation flags for nfsiod's __GFP_NORETRY

Posted by Benjamin Coddington 5 months, 1 week ago

On 10 Jul 2025, at 3:21, Christoph Hellwig wrote:

> On Wed, Jul 09, 2025 at 09:47:43PM -0400, Benjamin Coddington wrote:
>> If the NFS client is doing writeback from a workqueue context, avoid using
>> __GFP_NORETRY for allocations if the task has set PF_MEMALLOC_NOIO or
>> PF_MEMALLOC_NOFS.  The combination of these flags makes memory allocation
>> failures much more likely.
>
> Can we take a step back and figre out why this blanket usage of
> __GFP_NORETRY exists at all?

Added in 515dcdcd48736 there's a decent explanation which boils down to: its
usually OK for nfsiod to have an allocation failure, we want it to fail
quickly and not get hung up waiting for an allocation.

Ben