On 6/10/25 7:01 AM, Chuck Lever wrote:

On 6/10/25 9:59 AM, Jeff Layton wrote:

quoted

On Tue, 2025-06-10 at 09:52 -0400, Chuck Lever wrote:

quoted

On 6/10/25 9:50 AM, Jeff Layton wrote:

quoted

On Tue, 2025-06-10 at 06:41 -0700, Dai Ngo wrote:

quoted

When the client sends an OPEN with claim type CLAIM_DELEG_CUR_FH or
CLAIM_DELEGATION_CUR, the delegation stateid and the file handle
must belongs to the same file, otherwise return NFS4ERR_BAD_STATEID.

Signed-off-by: Dai Ngo <dai.ngo@oracle.com>
---
  fs/nfsd/nfs4state.c | 5 +++++
  1 file changed, 5 insertions(+)

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 59a693f22452..be2ee641a22d 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c

@@ -6318,6 +6318,11 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf
  		status = nfs4_check_deleg(cl, open, &dp);
  		if (status)
  			goto out;
+		if (dp && nfsd4_is_deleg_cur(open) &&
+				(dp->dl_stid.sc_file != fp)) {
+			status = nfserr_bad_stateid;
+			goto out;
+		}
  		stp = nfsd4_find_and_lock_existing_open(fp, open);
  	} else {
  		open->op_file = NULL;

This seems like a good idea. I wonder if BAD_STATEID is the right error
here. It is a valid stateid, after all, it just doesn't match the
current_fh. Maybe this should be nfserr_inval ?

I agree, NFS4ERR_BAD_STATEID /might/ cause a loop, so that needs to be
tested. BAD_STATEID is mandated by the spec, so if we choose to return
a different status code here, it needs a comment explaining why.

Oh, I didn't realize that error was mandated, but you're right.
RFC8881, section 8.2.4:

- If the selected table entry does not match the current filehandle,
return NFS4ERR_BAD_STATEID.

I guess we're stuck with reporting that unless we want to amend the
spec.

It is spec-mandated behavior, but we are always free to ignore the
spec. I'm OK with NFS4ERR_INVAL if it results in better behavior
(as long as there is a comment explaining why we deviate from the
mandate).

Since the Linux client does not behave this way I can not test if this
error get us into a loop. I used pynfs to force this behavior.

However, here is the comment in nfs4_do_open:

                 /*
                  * BAD_STATEID on OPEN means that the server cancelled our
                  * state before it received the OPEN_CONFIRM.
                  * Recover by retrying the request as per the discussion
                  * on Page 181 of RFC3530.
                  */

So it guess BAD_STATEID will  get the client and server into a loop.
I'll change error to NFS4ERR_INVAL and add a comment in the code.

-Dai

quoted

quoted

quoted

In any case, whatever we decide:

Reviewed-by: Jeff Layton <jlayton@kernel.org>