[PATCH] nfs: fix bitmap decoder to handle a 3rd word

[PATCH] nfs: fix bitmap decoder to handle a 3rd word

[Jeff Layton]

It only decodes the first two words at this point. Have it decode the
third word as well. Without this, the client doesn't send delegated
timestamps in the CB_GETATTR response.

Fixes: 43df7110f4a9 ("NFSv4: Add CB_GETATTR support for delegated attributes")
Signed-off-by: Jeff Layton <jlayton@kernel.org>
---
Found this while working on the delstid patches for nfsd.
---
 fs/nfs/callback_xdr.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/nfs/callback_xdr.c b/fs/nfs/callback_xdr.c
index 29c49a7e5fe1..246470306172 100644
--- a/fs/nfs/callback_xdr.c
+++ b/fs/nfs/callback_xdr.c
@@ -118,7 +118,9 @@ static __be32 decode_bitmap(struct xdr_stream *xdr, uint32_t *bitmap)
 	if (likely(attrlen > 0))
 		bitmap[0] = ntohl(*p++);
 	if (attrlen > 1)
-		bitmap[1] = ntohl(*p);
+		bitmap[1] = ntohl(*p++);
+	if (attrlen > 2)
+		bitmap[2] = ntohl(*p);
 	return 0;
 }
 

---
base-commit: b311c1b497e51a628aa89e7cb954481e5f9dced2
change-id: 20240821-nfs-6-11-188bb4e1f1dd

Best regards,
-- 
Jeff Layton <jlayton@kernel.org>

Re: [PATCH] nfs: fix bitmap decoder to handle a 3rd word

[Trond Myklebust]

On Wed, 2024-08-21 at 08:16 -0400, Jeff Layton wrote:
> It only decodes the first two words at this point. Have it decode the
> third word as well. Without this, the client doesn't send delegated
> timestamps in the CB_GETATTR response.
> 
> Fixes: 43df7110f4a9 ("NFSv4: Add CB_GETATTR support for delegated
> attributes")
> Signed-off-by: Jeff Layton <jlayton@kernel.org>
> ---
> Found this while working on the delstid patches for nfsd.
> ---
>  fs/nfs/callback_xdr.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/nfs/callback_xdr.c b/fs/nfs/callback_xdr.c
> index 29c49a7e5fe1..246470306172 100644
> --- a/fs/nfs/callback_xdr.c
> +++ b/fs/nfs/callback_xdr.c
> @@ -118,7 +118,9 @@ static __be32 decode_bitmap(struct xdr_stream
> *xdr, uint32_t *bitmap)
>  	if (likely(attrlen > 0))
>  		bitmap[0] = ntohl(*p++);
>  	if (attrlen > 1)
> -		bitmap[1] = ntohl(*p);
> +		bitmap[1] = ntohl(*p++);
> +	if (attrlen > 2)
> +		bitmap[2] = ntohl(*p);
>  	return 0;
>  }
>  
> 
> ---
> base-commit: b311c1b497e51a628aa89e7cb954481e5f9dced2
> change-id: 20240821-nfs-6-11-188bb4e1f1dd
> 
> Best regards,

Why do we need this? I'm not really understanding which callback
attributes we'd want to return in that range.

-- 
Trond Myklebust
Linux NFS client maintainer, Hammerspace
trond.myklebust@hammerspace.com

Re: [PATCH] nfs: fix bitmap decoder to handle a 3rd word

[Jeff Layton]

On Wed, 2024-08-21 at 14:58 +0000, Trond Myklebust wrote:
> On Wed, 2024-08-21 at 08:16 -0400, Jeff Layton wrote:
> > It only decodes the first two words at this point. Have it decode the
> > third word as well. Without this, the client doesn't send delegated
> > timestamps in the CB_GETATTR response.
> > 
> > Fixes: 43df7110f4a9 ("NFSv4: Add CB_GETATTR support for delegated
> > attributes")
> > Signed-off-by: Jeff Layton <jlayton@kernel.org>
> > ---
> > Found this while working on the delstid patches for nfsd.
> > ---
> >  fs/nfs/callback_xdr.c | 4 +++-
> >  1 file changed, 3 insertions(+), 1 deletion(-)
> > 
> > diff --git a/fs/nfs/callback_xdr.c b/fs/nfs/callback_xdr.c
> > index 29c49a7e5fe1..246470306172 100644
> > --- a/fs/nfs/callback_xdr.c
> > +++ b/fs/nfs/callback_xdr.c
> > @@ -118,7 +118,9 @@ static __be32 decode_bitmap(struct xdr_stream
> > *xdr, uint32_t *bitmap)
> >  	if (likely(attrlen > 0))
> >  		bitmap[0] = ntohl(*p++);
> >  	if (attrlen > 1)
> > -		bitmap[1] = ntohl(*p);
> > +		bitmap[1] = ntohl(*p++);
> > +	if (attrlen > 2)
> > +		bitmap[2] = ntohl(*p);
> >  	return 0;
> >  }
> >  
> > 
> > ---
> > base-commit: b311c1b497e51a628aa89e7cb954481e5f9dced2
> > change-id: 20240821-nfs-6-11-188bb4e1f1dd
> > 
> > Best regards,
> 
> Why do we need this? I'm not really understanding which callback
> attributes we'd want to return in that range.
> 

(note that there is a v2 that fixes a potential buffer overrun with
this change. We'll want that one)

FATTR4_WORD2_TIME_DELEG_ACCESS and FATTR4_WORD2_TIME_DELEG_MODIFY. 
43df7110f4a90 added support for those, but the client doesn't see the
server's request for them without this change.

-- 
Jeff Layton <jlayton@kernel.org>

Re: [PATCH] nfs: fix bitmap decoder to handle a 3rd word

[Trond Myklebust]

On Wed, 2024-08-21 at 11:01 -0400, Jeff Layton wrote:
> On Wed, 2024-08-21 at 14:58 +0000, Trond Myklebust wrote:
> > On Wed, 2024-08-21 at 08:16 -0400, Jeff Layton wrote:
> > > It only decodes the first two words at this point. Have it decode
> > > the
> > > third word as well. Without this, the client doesn't send
> > > delegated
> > > timestamps in the CB_GETATTR response.
> > > 
> > > Fixes: 43df7110f4a9 ("NFSv4: Add CB_GETATTR support for delegated
> > > attributes")
> > > Signed-off-by: Jeff Layton <jlayton@kernel.org>
> > > ---
> > > Found this while working on the delstid patches for nfsd.
> > > ---
> > >  fs/nfs/callback_xdr.c | 4 +++-
> > >  1 file changed, 3 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/fs/nfs/callback_xdr.c b/fs/nfs/callback_xdr.c
> > > index 29c49a7e5fe1..246470306172 100644
> > > --- a/fs/nfs/callback_xdr.c
> > > +++ b/fs/nfs/callback_xdr.c
> > > @@ -118,7 +118,9 @@ static __be32 decode_bitmap(struct xdr_stream
> > > *xdr, uint32_t *bitmap)
> > >  	if (likely(attrlen > 0))
> > >  		bitmap[0] = ntohl(*p++);
> > >  	if (attrlen > 1)
> > > -		bitmap[1] = ntohl(*p);
> > > +		bitmap[1] = ntohl(*p++);
> > > +	if (attrlen > 2)
> > > +		bitmap[2] = ntohl(*p);
> > >  	return 0;
> > >  }
> > >  
> > > 
> > > ---
> > > base-commit: b311c1b497e51a628aa89e7cb954481e5f9dced2
> > > change-id: 20240821-nfs-6-11-188bb4e1f1dd
> > > 
> > > Best regards,
> > 
> > Why do we need this? I'm not really understanding which callback
> > attributes we'd want to return in that range.
> > 
> 
> (note that there is a v2 that fixes a potential buffer overrun with
> this change. We'll want that one)
> 
> FATTR4_WORD2_TIME_DELEG_ACCESS and FATTR4_WORD2_TIME_DELEG_MODIFY. 
> 43df7110f4a90 added support for those, but the client doesn't see the
> server's request for them without this change.
> 

D'oh! Acked...

-- 
Trond Myklebust
Linux NFS client maintainer, Hammerspace
trond.myklebust@hammerspace.com

Re: [PATCH] nfs: fix bitmap decoder to handle a 3rd word

[Trond Myklebust]

On Wed, 2024-08-21 at 15:37 +0000, Trond Myklebust wrote:
> On Wed, 2024-08-21 at 11:01 -0400, Jeff Layton wrote:
> > On Wed, 2024-08-21 at 14:58 +0000, Trond Myklebust wrote:
> > > On Wed, 2024-08-21 at 08:16 -0400, Jeff Layton wrote:
> > > > It only decodes the first two words at this point. Have it
> > > > decode
> > > > the
> > > > third word as well. Without this, the client doesn't send
> > > > delegated
> > > > timestamps in the CB_GETATTR response.
> > > > 
> > > > Fixes: 43df7110f4a9 ("NFSv4: Add CB_GETATTR support for
> > > > delegated
> > > > attributes")
> > > > Signed-off-by: Jeff Layton <jlayton@kernel.org>
> > > > ---
> > > > Found this while working on the delstid patches for nfsd.
> > > > ---
> > > >  fs/nfs/callback_xdr.c | 4 +++-
> > > >  1 file changed, 3 insertions(+), 1 deletion(-)
> > > > 
> > > > diff --git a/fs/nfs/callback_xdr.c b/fs/nfs/callback_xdr.c
> > > > index 29c49a7e5fe1..246470306172 100644
> > > > --- a/fs/nfs/callback_xdr.c
> > > > +++ b/fs/nfs/callback_xdr.c
> > > > @@ -118,7 +118,9 @@ static __be32 decode_bitmap(struct
> > > > xdr_stream
> > > > *xdr, uint32_t *bitmap)
> > > >  	if (likely(attrlen > 0))
> > > >  		bitmap[0] = ntohl(*p++);
> > > >  	if (attrlen > 1)
> > > > -		bitmap[1] = ntohl(*p);
> > > > +		bitmap[1] = ntohl(*p++);
> > > > +	if (attrlen > 2)
> > > > +		bitmap[2] = ntohl(*p);
> > > >  	return 0;
> > > >  }
> > > >  
> > > > 
> > > > ---
> > > > base-commit: b311c1b497e51a628aa89e7cb954481e5f9dced2
> > > > change-id: 20240821-nfs-6-11-188bb4e1f1dd
> > > > 
> > > > Best regards,
> > > 
> > > Why do we need this? I'm not really understanding which callback
> > > attributes we'd want to return in that range.
> > > 
> > 
> > (note that there is a v2 that fixes a potential buffer overrun with
> > this change. We'll want that one)
> > 
> > FATTR4_WORD2_TIME_DELEG_ACCESS and FATTR4_WORD2_TIME_DELEG_MODIFY. 
> > 43df7110f4a90 added support for those, but the client doesn't see
> > the
> > server's request for them without this change.
> > 
> 
> D'oh! Acked...
> 

Sorry. I mean Ack for the v2 that also fixes the resulting potential
for an array overflow in decode_recallany_args.

-- 
Trond Myklebust
Linux NFS client maintainer, Hammerspace
trond.myklebust@hammerspace.com