Received-SPF: pass (google.com: domain of linux-nfs+bounces-3215-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
From: Benjamin Coddington <bcodding@redhat.com>
To: Olga Kornievskaia <olga.kornievskaia@gmail.com>
Cc: trond.myklebust@hammerspace.com, anna.schumaker@netapp.com,
 linux-nfs@vger.kernel.org
Subject: Re: [PATCH 1/1] pNFS/filelayout: fixup pNfs allocation modes
Date: Wed, 08 May 2024 14:33:46 -0400
Message-ID: <B3A3E63F-E8E4-4B59-B833-7D1BE8938F52@redhat.com>
In-Reply-To: <CAN-5tyGNaWsfxo0dCGc+hW36Q9jSZXaQZgvHTOw5gGQ_HDxFqw@mail.gmail.com>
References: <20240507151545.26888-1-olga.kornievskaia@gmail.com>
 <35158E21-2724-4C1A-950F-5A6A616C862A@redhat.com>
 <CAN-5tyGNaWsfxo0dCGc+hW36Q9jSZXaQZgvHTOw5gGQ_HDxFqw@mail.gmail.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

On 8 May 2024, at 13:13, Olga Kornievskaia wrote:

> On Wed, May 8, 2024 at 11:25=E2=80=AFAM Benjamin Coddington <bcodding@r=
edhat.com> wrote:
>>
>> On 7 May 2024, at 11:15, Olga Kornievskaia wrote:
>>
>>> From: Olga Kornievskaia <kolga@netapp.com>
>>>
>>> Change left over allocation flags.
>>>
>>> Fixes: a245832aaa99 ("pNFS/files: Ensure pNFS allocation modes are co=
nsistent with nfsiod")
>>> Signed-off-by: Olga Kornievskaia <kolga@netapp.com>
>>> ---
>>>  fs/nfs/filelayout/filelayout.c | 4 ++--
>>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/fs/nfs/filelayout/filelayout.c b/fs/nfs/filelayout/filel=
ayout.c
>>> index cc2ed4b5a4fd..85d2dc9bc212 100644
>>> --- a/fs/nfs/filelayout/filelayout.c
>>> +++ b/fs/nfs/filelayout/filelayout.c
>>> @@ -875,7 +875,7 @@ filelayout_pg_init_read(struct nfs_pageio_descrip=
tor *pgio,
>>>                                                     req->wb_bytes,
>>>                                                     IOMODE_READ,
>>>                                                     false,
>>> -                                                   GFP_KERNEL);
>>> +                                                   nfs_io_gfp_mask()=
);
>>>               if (IS_ERR(pgio->pg_lseg)) {
>>>                       pgio->pg_error =3D PTR_ERR(pgio->pg_lseg);
>>>                       pgio->pg_lseg =3D NULL;
>>> @@ -899,7 +899,7 @@ filelayout_pg_init_write(struct nfs_pageio_descri=
ptor *pgio,
>>>                                                     req->wb_bytes,
>>>                                                     IOMODE_RW,
>>>                                                     false,
>>> -                                                   GFP_NOFS);
>>> +                                                   nfs_io_gfp_mask()=
);
>>>               if (IS_ERR(pgio->pg_lseg)) {
>>>                       pgio->pg_error =3D PTR_ERR(pgio->pg_lseg);
>>>                       pgio->pg_lseg =3D NULL;
>>> --
>>> 2.39.1
>>
>> Looks fine, but I didn't think you could get here from rpciod/nfsiod
>> context.  I might be missing something, how did you get here from ther=
e?
>
> I have to admit I don't fully understand (if at all) the implications
> of having the wrong flags. I also don't follow what you mean by this
> code not being executed by the rpciod/nfsiod context. This code is
> done while doing (buffered) IO and performed by the rpciod context?

I was thrown off by the Fixes: tag. The nfs_io_gfp_mask() doesn't have to=
 do
with that context per se, but rather if we're in writeback due to memory
pressure.  That's what the PF_WQ_WORKER check is for which Trond explains=

this commit 515dcdcd4873.

> In truth I was making it consistent with what flexfiles is doing for
> their pnfs_update_layout() usage.

Gotcha, makes sense to me now, thanks for helping me.

FWIW:
Reviewed-by: Benjamin Coddington <bcodding@redhat.com>

Ben