2014-07-09 13:51:58

by Kinglong Mee

Subject: [PATCH] NFSD: Fix bad checking of space for padding in splice read

Fix commit dc97618ddd (nfsd4: separate splice and readv cases).
xdr->end and xdr->p are type of __be32 *, so padding needs one here.

Signed-off-by: Kinglong Mee <[email protected]>
---
fs/nfsd/nfs4xdr.c | 7 ++-----
1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
index 523c666..8a2fbb2 100644
--- a/fs/nfsd/nfs4xdr.c
+++ b/fs/nfsd/nfs4xdr.c
@@ -3067,11 +3067,8 @@ static __be32 nfsd4_encode_splice_read(
__be32 nfserr;
__be32 *p = xdr->p - 2;

- /*
- * Don't inline pages unless we know there's room for eof,
- * count, and possible padding:
- */
- if (xdr->end - xdr->p < 3)
+ /* Make sure there will be room for padding if needed */
+ if (xdr->end - xdr->p < 1)
return nfserr_resource;

nfserr = nfsd_splice_read(read->rd_rqstp, file,
--
1.9.3

2014-07-11 19:19:03

by J. Bruce Fields

Subject: Re: [PATCH] NFSD: Fix bad checking of space for padding in splice read

On Fri, Jul 11, 2014 at 03:17:47PM -0400, J. Bruce Fields wrote:
> On Wed, Jul 09, 2014 at 09:51:27PM +0800, Kinglong Mee wrote:
> > Fix commit dc97618ddd (nfsd4: separate splice and readv cases).
> > xdr->end and xdr->p are type of __be32 *, so padding needs one here.
>
> I agree with the patch, but not the changelog.
>
> The problem here is that xdr->p isn't that those are __be32 *'s, but
> that xdr->p has already advanced past the eof and the count, right?
>
> Applying with an updated changelog.

(And applying for 3.17 (not 3.16) as this looks hard to hit.)

--b.


>
> --b.
>
> >
> > Signed-off-by: Kinglong Mee <[email protected]>
> > ---
> > fs/nfsd/nfs4xdr.c | 7 ++-----
> > 1 file changed, 2 insertions(+), 5 deletions(-)
> >
> > diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
> > index 523c666..8a2fbb2 100644
> > --- a/fs/nfsd/nfs4xdr.c
> > +++ b/fs/nfsd/nfs4xdr.c
> > @@ -3067,11 +3067,8 @@ static __be32 nfsd4_encode_splice_read(
> > __be32 nfserr;
> > __be32 *p = xdr->p - 2;
> >
> > - /*
> > - * Don't inline pages unless we know there's room for eof,
> > - * count, and possible padding:
> > - */
> > - if (xdr->end - xdr->p < 3)
> > + /* Make sure there will be room for padding if needed */
> > + if (xdr->end - xdr->p < 1)
> > return nfserr_resource;
> >
> > nfserr = nfsd_splice_read(read->rd_rqstp, file,
> > --
> > 1.9.3
> >

2014-07-11 19:17:50

by J. Bruce Fields

Subject: Re: [PATCH] NFSD: Fix bad checking of space for padding in splice read

On Wed, Jul 09, 2014 at 09:51:27PM +0800, Kinglong Mee wrote:
> Fix commit dc97618ddd (nfsd4: separate splice and readv cases).
> xdr->end and xdr->p are type of __be32 *, so padding needs one here.

I agree with the patch, but not the changelog.

The problem here is that xdr->p isn't that those are __be32 *'s, but
that xdr->p has already advanced past the eof and the count, right?

Applying with an updated changelog.

--b.

>
> Signed-off-by: Kinglong Mee <[email protected]>
> ---
> fs/nfsd/nfs4xdr.c | 7 ++-----
> 1 file changed, 2 insertions(+), 5 deletions(-)
>
> diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
> index 523c666..8a2fbb2 100644
> --- a/fs/nfsd/nfs4xdr.c
> +++ b/fs/nfsd/nfs4xdr.c
> @@ -3067,11 +3067,8 @@ static __be32 nfsd4_encode_splice_read(
> __be32 nfserr;
> __be32 *p = xdr->p - 2;
>
> - /*
> - * Don't inline pages unless we know there's room for eof,
> - * count, and possible padding:
> - */
> - if (xdr->end - xdr->p < 3)
> + /* Make sure there will be room for padding if needed */
> + if (xdr->end - xdr->p < 1)
> return nfserr_resource;
>
> nfserr = nfsd_splice_read(read->rd_rqstp, file,
> --
> 1.9.3
>