2024-06-11 19:41:23

by Chuck Lever

Subject: [PATCH v2] NFSD: Support write delegations in LAYOUTGET

From: Chuck Lever <[email protected]>

I noticed LAYOUTGET(LAYOUTIOMODE4_RW) returning NFS4ERR_ACCESS
unexpectedly. The NFS client had created a file with mode 0444, and
the server had returned a write delegation on the OPEN(CREATE). The
client was requesting a RW layout using the write delegation stateid
so that it could flush file modifications.

Creating a read-only file does not seem to be problematic for
NFSv4.1 without pNFS, so I began looking at NFSD's implementation of
LAYOUTGET.

The failure was because fh_verify() was doing a permission check as
part of verifying the FH presented during the LAYOUTGET. It uses the
loga_iomode value to specify the @accmode argument to fh_verify().
fh_verify(MAY_WRITE) on a file whose mode is 0444 fails with -EACCES.

To permit LAYOUT* operations in this case, add OWNER_OVERRIDE when
checking the access permission of the incoming file handle for
LAYOUTGET and LAYOUTCOMMIT.

Cc: Christoph Hellwig <[email protected]>
X-Cc: [email protected] # v6.6+
Message-Id: [email protected]
Signed-off-by: Chuck Lever <[email protected]>
---
fs/nfsd/nfs4proc.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index 46bd20fe5c0f..2e39cf2e502a 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -2269,7 +2269,7 @@ nfsd4_layoutget(struct svc_rqst *rqstp,
const struct nfsd4_layout_ops *ops;
struct nfs4_layout_stateid *ls;
__be32 nfserr;
- int accmode = NFSD_MAY_READ_IF_EXEC;
+ int accmode = NFSD_MAY_READ_IF_EXEC | NFSD_MAY_OWNER_OVERRIDE;

switch (lgp->lg_seg.iomode) {
case IOMODE_READ:
@@ -2359,7 +2359,8 @@ nfsd4_layoutcommit(struct svc_rqst *rqstp,
struct nfs4_layout_stateid *ls;
__be32 nfserr;

- nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
+ nfserr = fh_verify(rqstp, current_fh, 0,
+ NFSD_MAY_WRITE | NFSD_MAY_OWNER_OVERRIDE);
if (nfserr)
goto out;

--
2.45.1

2024-06-12 05:09:08

by Christoph Hellwig

Subject: Re: [PATCH v2] NFSD: Support write delegations in LAYOUTGET

On Tue, Jun 11, 2024 at 03:36:46PM -0400, [email protected] wrote:
> From: Chuck Lever <[email protected]>
>
> I noticed LAYOUTGET(LAYOUTIOMODE4_RW) returning NFS4ERR_ACCESS
> unexpectedly. The NFS client had created a file with mode 0444, and
> the server had returned a write delegation on the OPEN(CREATE). The
> client was requesting a RW layout using the write delegation stateid
> so that it could flush file modifications.
>
> Creating a read-only file does not seem to be problematic for
> NFSv4.1 without pNFS, so I began looking at NFSD's implementation of
> LAYOUTGET.
>
> The failure was because fh_verify() was doing a permission check as
> part of verifying the FH presented during the LAYOUTGET. It uses the
> loga_iomode value to specify the @accmode argument to fh_verify().
> fh_verify(MAY_WRITE) on a file whose mode is 0444 fails with -EACCES.
>
> To permit LAYOUT* operations in this case, add OWNER_OVERRIDE when
> checking the access permission of the incoming file handle for
> LAYOUTGET and LAYOUTCOMMIT.

This looks reasonable to me:

Reviewed-by: Christoph Hellwig <[email protected]>

2024-06-12 13:52:31

by Jeff Layton

Subject: Re: [PATCH v2] NFSD: Support write delegations in LAYOUTGET

On Tue, 2024-06-11 at 15:36 -0400, [email protected] wrote:
> From: Chuck Lever <[email protected]>
>
> I noticed LAYOUTGET(LAYOUTIOMODE4_RW) returning NFS4ERR_ACCESS
> unexpectedly. The NFS client had created a file with mode 0444, and
> the server had returned a write delegation on the OPEN(CREATE). The
> client was requesting a RW layout using the write delegation stateid
> so that it could flush file modifications.
>
> Creating a read-only file does not seem to be problematic for
> NFSv4.1 without pNFS, so I began looking at NFSD's implementation of
> LAYOUTGET.
>
> The failure was because fh_verify() was doing a permission check as
> part of verifying the FH presented during the LAYOUTGET. It uses the
> loga_iomode value to specify the @accmode argument to fh_verify().
> fh_verify(MAY_WRITE) on a file whose mode is 0444 fails with -EACCES.
>
> To permit LAYOUT* operations in this case, add OWNER_OVERRIDE when
> checking the access permission of the incoming file handle for
> LAYOUTGET and LAYOUTCOMMIT.
>
> Cc: Christoph Hellwig <[email protected]>
> X-Cc: [email protected] # v6.6+
> Message-Id: [email protected]
> Signed-off-by: Chuck Lever <[email protected]>
> ---
>  fs/nfsd/nfs4proc.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
> index 46bd20fe5c0f..2e39cf2e502a 100644
> --- a/fs/nfsd/nfs4proc.c
> +++ b/fs/nfsd/nfs4proc.c
> @@ -2269,7 +2269,7 @@ nfsd4_layoutget(struct svc_rqst *rqstp,
>   const struct nfsd4_layout_ops *ops;
>   struct nfs4_layout_stateid *ls;
>   __be32 nfserr;
> - int accmode = NFSD_MAY_READ_IF_EXEC;
> + int accmode = NFSD_MAY_READ_IF_EXEC |
> NFSD_MAY_OWNER_OVERRIDE;
>  
>   switch (lgp->lg_seg.iomode) {
>   case IOMODE_READ:
> @@ -2359,7 +2359,8 @@ nfsd4_layoutcommit(struct svc_rqst *rqstp,
>   struct nfs4_layout_stateid *ls;
>   __be32 nfserr;
>  
> - nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
> + nfserr = fh_verify(rqstp, current_fh, 0,
> +    NFSD_MAY_WRITE |
> NFSD_MAY_OWNER_OVERRIDE);
>   if (nfserr)
>   goto out;
>  

Reviewed-by: Jeff Layton <[email protected]>