Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Content-Type: text/plain;
        charset=us-ascii
Mime-Version: 1.0 (Mac OS X Mail 13.4 \(3608.120.23.2.4\))
Subject: Re: [PATCH 15/16] nfsd: Fixes for nfsd4_encode_read_plus_data()
From:   Chuck Lever <chucklever@gmail.com>
In-Reply-To: <55246c5db6ec550827f1f230bd980760db28a689.camel@hammerspace.com>
Date:   Wed, 9 Dec 2020 11:57:56 -0500
Cc:     Linux NFS Mailing List <linux-nfs@vger.kernel.org>
Content-Transfer-Encoding: quoted-printable
Message-Id: <DF972250-F1EA-4996-8CEB-18171BD346C2@gmail.com>
References: <20201209144801.700778-1-trondmy@kernel.org>
 <20201209144801.700778-2-trondmy@kernel.org>
 <20201209144801.700778-3-trondmy@kernel.org>
 <20201209144801.700778-4-trondmy@kernel.org>
 <20201209144801.700778-5-trondmy@kernel.org>
 <20201209144801.700778-6-trondmy@kernel.org>
 <20201209144801.700778-7-trondmy@kernel.org>
 <20201209144801.700778-8-trondmy@kernel.org>
 <20201209144801.700778-9-trondmy@kernel.org>
 <20201209144801.700778-10-trondmy@kernel.org>
 <20201209144801.700778-11-trondmy@kernel.org>
 <20201209144801.700778-12-trondmy@kernel.org>
 <20201209144801.700778-13-trondmy@kernel.org>
 <20201209144801.700778-14-trondmy@kernel.org>
 <20201209144801.700778-15-trondmy@kernel.org>
 <20201209144801.700778-16-trondmy@kernel.org>
 <4BA2AAC7-C4F5-4740-B10F-D34022A58722@gmail.com>
 <55246c5db6ec550827f1f230bd980760db28a689.camel@hammerspace.com>
To:     Trond Myklebust <trondmy@hammerspace.com>
Precedence: bulk


> On Dec 9, 2020, at 11:39 AM, Trond Myklebust <trondmy@hammerspace.com> =
wrote:
>=20
> On Wed, 2020-12-09 at 11:16 -0500, Chuck Lever wrote:
>> Hey-
>>=20
>>> On Dec 9, 2020, at 9:48 AM, trondmy@kernel.org wrote:
>>>=20
>>> From: Trond Myklebust <trond.myklebust@hammerspace.com>
>>>=20
>>> Ensure that we encode the data payload + padding, and that we
>>> truncate
>>> the preallocated buffer to the actual read size.
>>=20
>> Did you intend to merge 15/16 and 16/16 through your tree?
>=20
> No. They can go through the nfsd tree. I included them here because
> they are necessary in order to pass the xfstests.

Would it be OK if they went in 5.11-rc? I've got the initial
merge tag prepared already. If they can't wait, let me know.


>> Can the patch descriptions say a little more about why these
>> changes are necessary? If they fix a misbehavior, describe
>> the problem.
>=20
> It's the same problem and solution that exists in the READ code.
>=20
> nfsd_readv() doesn't necessarily return the same number of bytes that
> we requested and preallocated buffer space for. So to deal with that,
> we have to truncate the preallocated buffer.

Huh. I thought it was doing that already? Oh, that's just for
the cases where the server returns an error status. The
READ_PLUS encoder incorrectly does not do that truncation for
short READs... got it.


> Finally, we have to write zeros into the padding bytes after the read
> buffer.

Right. Then the problem statement is that the server's READ_PLUS
XDR encoder isn't padding the read buffer properly.

Quibble: perhaps these are two separate issues that each deserve
their own patches with Fixes: tags (and if you re-post these,
please add a Fixes: tag to 16/16 too).

Thanks!


>>> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
>>> ---
>>> fs/nfsd/nfs4xdr.c | 5 +++++
>>> 1 file changed, 5 insertions(+)
>>>=20
>>> diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
>>> index 833a2c64dfe8..26f6e277101d 100644
>>> --- a/fs/nfsd/nfs4xdr.c
>>> +++ b/fs/nfsd/nfs4xdr.c
>>> @@ -4632,6 +4632,7 @@ nfsd4_encode_read_plus_data(struct
>>> nfsd4_compoundres *resp,
>>>                             resp->rqstp->rq_vec, read->rd_vlen,
>>> maxcount, eof);
>>>         if (nfserr)
>>>                 return nfserr;
>>> +       xdr_truncate_encode(xdr, starting_len + 16 +
>>> xdr_align_size(*maxcount));
>>>=20
>>>         tmp =3D htonl(NFS4_CONTENT_DATA);
>>>         write_bytes_to_xdr_buf(xdr->buf, starting_len,      &tmp, =20=

>>> 4);
>>> @@ -4639,6 +4640,10 @@ nfsd4_encode_read_plus_data(struct
>>> nfsd4_compoundres *resp,
>>>         write_bytes_to_xdr_buf(xdr->buf, starting_len + 4,  &tmp64,
>>> 8);
>>>         tmp =3D htonl(*maxcount);
>>>         write_bytes_to_xdr_buf(xdr->buf, starting_len + 12, &tmp, =20=

>>> 4);
>>> +
>>> +       tmp =3D xdr_zero;
>>> +       write_bytes_to_xdr_buf(xdr->buf, starting_len + 16 +
>>> *maxcount, &tmp,
>>> +                              xdr_pad_size(*maxcount));
>>>         return nfs_ok;
>>> }
>=20
> --=20
> Trond Myklebust
> Linux NFS client maintainer, Hammerspace
> trond.myklebust@hammerspace.com

--
Chuck Lever
chucklever@gmail.com