nfsd4_decode_createhow4() calls nfsd4_decode_fattr4(), which allocates
refcounted struct posix_acl objects via posix_acl_alloc() and stores
them in open->op_pacl and open->op_dpacl. These pointers must be
released once the OPEN compound finishes.

When nfsd4_decode_open_claim4() returns a non-seqid-mutating error,
the dispatcher short-circuits before op_func runs:

    nfsd4_proc_compound()
      opdesc->op_func == nfsd4_open_omfg
        if (!seqid_mutating_err(ntohl(op->status)))
            return op->status;   /* nfsd4_open() never runs */
      opdesc->op_release(&op->u)  /* must still release op_pacl/op_dpacl */

Before this change OP_OPEN had no .op_release in nfsd4_ops[], and the
release pair lived inside nfsd4_open() at its out_err: label. On the
short-circuit path nfsd4_open() is never invoked, so both posix_acl
refs leak on every malformed OPEN compound that carries valid POSIX
ACL createhow4 attributes.

Add nfsd4_open_release() and wire it as .op_release for OP_OPEN.
posix_acl_release() is NULL-safe, so the single release site covers
both the normal path and the nfsd4_open_omfg short-circuit. Remove
the matching posix_acl_release() pair from nfsd4_open()'s out_err:
label to avoid double-releasing.

The compound loop has two encoding branches: nfsd4_encode_operation()
for normal ops, and nfsd4_encode_replay() for v4.0 replayed ops.
op_release was only called from nfsd4_encode_operation(), so resources
attached to op->u leak on the replay path. Add an op_release call to
the replay branch as well to ensure cleanup on every path.

Fixes: 5fc51dfc2eb1 ("NFSD: Add support for XDR decoding POSIX draft ACLs")
Signed-off-by: Chris Mason <clm@meta.com>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
---
Changes in v2:
- Ensure that op_release is called in the v4.0 replay case as well
- Link to v1: https://lore.kernel.org/r/20260531-nfsd-testing-v1-0-7bfa481b0540@kernel.org
---
 fs/nfsd/nfs4proc.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index 017474cd63b5..51998d7885ae 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -681,8 +681,6 @@ nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 	nfsd4_cleanup_open_state(cstate, open);
 	nfsd4_bump_seqid(cstate, status);
 out_err:
-	posix_acl_release(open->op_dpacl);
-	posix_acl_release(open->op_pacl);
 	return status;
 }
 
@@ -704,6 +702,13 @@ static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_stat
 	return nfsd4_open(rqstp, cstate, &op->u);
 }
 
+static void
+nfsd4_open_release(union nfsd4_op_u *u)
+{
+	posix_acl_release(u->open.op_dpacl);
+	posix_acl_release(u->open.op_pacl);
+}
+
 /*
  * filehandle-manipulating ops.
  */
@@ -3214,6 +3219,8 @@ nfsd4_proc_compound(struct svc_rqst *rqstp)
 			op->replay = &cstate->replay_owner->so_replay;
 			nfsd4_encode_replay(resp->xdr, op);
 			status = op->status = op->replay->rp_status;
+			if (op->opdesc->op_release)
+				op->opdesc->op_release(&op->u);
 		} else {
 			nfsd4_encode_operation(resp, op);
 			status = op->status;
@@ -3718,6 +3725,7 @@ static const struct nfsd4_operation nfsd4_ops[] = {
 	},
 	[OP_OPEN] = {
 		.op_func = nfsd4_open,
+		.op_release = nfsd4_open_release,
 		.op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
 		.op_name = "OP_OPEN",
 		.op_rsize_bop = nfsd4_open_rsize,

---
base-commit: 6c0004650ba248a12937ada16f9ba961b35ce2b5
change-id: 20260531-nfsd-testing-9122bf51ce95

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

On Mon, 01 Jun 2026, Jeff Layton wrote:
> nfsd4_decode_createhow4() calls nfsd4_decode_fattr4(), which allocates
> refcounted struct posix_acl objects via posix_acl_alloc() and stores
> them in open->op_pacl and open->op_dpacl. These pointers must be
> released once the OPEN compound finishes.
> 
> When nfsd4_decode_open_claim4() returns a non-seqid-mutating error,
> the dispatcher short-circuits before op_func runs:
> 
>     nfsd4_proc_compound()
>       opdesc->op_func == nfsd4_open_omfg
>         if (!seqid_mutating_err(ntohl(op->status)))
>             return op->status;   /* nfsd4_open() never runs */
>       opdesc->op_release(&op->u)  /* must still release op_pacl/op_dpacl */
> 
> Before this change OP_OPEN had no .op_release in nfsd4_ops[], and the
> release pair lived inside nfsd4_open() at its out_err: label. On the
> short-circuit path nfsd4_open() is never invoked, so both posix_acl
> refs leak on every malformed OPEN compound that carries valid POSIX
> ACL createhow4 attributes.
> 
> Add nfsd4_open_release() and wire it as .op_release for OP_OPEN.
> posix_acl_release() is NULL-safe, so the single release site covers
> both the normal path and the nfsd4_open_omfg short-circuit. Remove
> the matching posix_acl_release() pair from nfsd4_open()'s out_err:
> label to avoid double-releasing.
> 
> The compound loop has two encoding branches: nfsd4_encode_operation()
> for normal ops, and nfsd4_encode_replay() for v4.0 replayed ops.
> op_release was only called from nfsd4_encode_operation(), so resources
> attached to op->u leak on the replay path. Add an op_release call to
> the replay branch as well to ensure cleanup on every path.
> 
> Fixes: 5fc51dfc2eb1 ("NFSD: Add support for XDR decoding POSIX draft ACLs")
> Signed-off-by: Chris Mason <clm@meta.com>
> Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
> Signed-off-by: Jeff Layton <jlayton@kernel.org>
> ---
> Changes in v2:
> - Ensure that op_release is called in the v4.0 replay case as well
> - Link to v1: https://lore.kernel.org/r/20260531-nfsd-testing-v1-0-7bfa481b0540@kernel.org
> ---
>  fs/nfsd/nfs4proc.c | 12 ++++++++++--
>  1 file changed, 10 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
> index 017474cd63b5..51998d7885ae 100644
> --- a/fs/nfsd/nfs4proc.c
> +++ b/fs/nfsd/nfs4proc.c
> @@ -681,8 +681,6 @@ nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
>  	nfsd4_cleanup_open_state(cstate, open);
>  	nfsd4_bump_seqid(cstate, status);
>  out_err:
> -	posix_acl_release(open->op_dpacl);
> -	posix_acl_release(open->op_pacl);
>  	return status;
>  }
>  
> @@ -704,6 +702,13 @@ static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_stat
>  	return nfsd4_open(rqstp, cstate, &op->u);
>  }
>  
> +static void
> +nfsd4_open_release(union nfsd4_op_u *u)
> +{
> +	posix_acl_release(u->open.op_dpacl);
> +	posix_acl_release(u->open.op_pacl);
> +}
> +
>  /*
>   * filehandle-manipulating ops.
>   */
> @@ -3214,6 +3219,8 @@ nfsd4_proc_compound(struct svc_rqst *rqstp)
>  			op->replay = &cstate->replay_owner->so_replay;
>  			nfsd4_encode_replay(resp->xdr, op);
>  			status = op->status = op->replay->rp_status;
> +			if (op->opdesc->op_release)
> +				op->opdesc->op_release(&op->u);
>  		} else {
>  			nfsd4_encode_operation(resp, op);
>  			status = op->status;

I think this patch is good, but I think it would be even better if the
->op_release() call were moved out of nfsd4_encode_operation() and
places after this if-else.  Then there would be only one call-site in a
fairly obviously-correct place.
But:
  Reviewed-by: NeilBrown <neil@brown.name>
for if you just want to stick with this version.

Thanks,
NeilBrown


> @@ -3718,6 +3725,7 @@ static const struct nfsd4_operation nfsd4_ops[] = {
>  	},
>  	[OP_OPEN] = {
>  		.op_func = nfsd4_open,
> +		.op_release = nfsd4_open_release,
>  		.op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
>  		.op_name = "OP_OPEN",
>  		.op_rsize_bop = nfsd4_open_rsize,
> 
> ---
> base-commit: 6c0004650ba248a12937ada16f9ba961b35ce2b5
> change-id: 20260531-nfsd-testing-9122bf51ce95
> 
> Best regards,
> -- 
> Jeff Layton <jlayton@kernel.org>
> 
> 

On Mon, 2026-06-01 at 11:06 +1000, NeilBrown wrote:
> On Mon, 01 Jun 2026, Jeff Layton wrote:
> > nfsd4_decode_createhow4() calls nfsd4_decode_fattr4(), which allocates
> > refcounted struct posix_acl objects via posix_acl_alloc() and stores
> > them in open->op_pacl and open->op_dpacl. These pointers must be
> > released once the OPEN compound finishes.
> > 
> > When nfsd4_decode_open_claim4() returns a non-seqid-mutating error,
> > the dispatcher short-circuits before op_func runs:
> > 
> >     nfsd4_proc_compound()
> >       opdesc->op_func == nfsd4_open_omfg
> >         if (!seqid_mutating_err(ntohl(op->status)))
> >             return op->status;   /* nfsd4_open() never runs */
> >       opdesc->op_release(&op->u)  /* must still release op_pacl/op_dpacl */
> > 
> > Before this change OP_OPEN had no .op_release in nfsd4_ops[], and the
> > release pair lived inside nfsd4_open() at its out_err: label. On the
> > short-circuit path nfsd4_open() is never invoked, so both posix_acl
> > refs leak on every malformed OPEN compound that carries valid POSIX
> > ACL createhow4 attributes.
> > 
> > Add nfsd4_open_release() and wire it as .op_release for OP_OPEN.
> > posix_acl_release() is NULL-safe, so the single release site covers
> > both the normal path and the nfsd4_open_omfg short-circuit. Remove
> > the matching posix_acl_release() pair from nfsd4_open()'s out_err:
> > label to avoid double-releasing.
> > 
> > The compound loop has two encoding branches: nfsd4_encode_operation()
> > for normal ops, and nfsd4_encode_replay() for v4.0 replayed ops.
> > op_release was only called from nfsd4_encode_operation(), so resources
> > attached to op->u leak on the replay path. Add an op_release call to
> > the replay branch as well to ensure cleanup on every path.
> > 
> > Fixes: 5fc51dfc2eb1 ("NFSD: Add support for XDR decoding POSIX draft ACLs")
> > Signed-off-by: Chris Mason <clm@meta.com>
> > Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
> > Signed-off-by: Jeff Layton <jlayton@kernel.org>
> > ---
> > Changes in v2:
> > - Ensure that op_release is called in the v4.0 replay case as well
> > - Link to v1: https://lore.kernel.org/r/20260531-nfsd-testing-v1-0-7bfa481b0540@kernel.org
> > ---
> >  fs/nfsd/nfs4proc.c | 12 ++++++++++--
> >  1 file changed, 10 insertions(+), 2 deletions(-)
> > 
> > diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
> > index 017474cd63b5..51998d7885ae 100644
> > --- a/fs/nfsd/nfs4proc.c
> > +++ b/fs/nfsd/nfs4proc.c
> > @@ -681,8 +681,6 @@ nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
> >  	nfsd4_cleanup_open_state(cstate, open);
> >  	nfsd4_bump_seqid(cstate, status);
> >  out_err:
> > -	posix_acl_release(open->op_dpacl);
> > -	posix_acl_release(open->op_pacl);
> >  	return status;
> >  }
> >  
> > @@ -704,6 +702,13 @@ static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_stat
> >  	return nfsd4_open(rqstp, cstate, &op->u);
> >  }
> >  
> > +static void
> > +nfsd4_open_release(union nfsd4_op_u *u)
> > +{
> > +	posix_acl_release(u->open.op_dpacl);
> > +	posix_acl_release(u->open.op_pacl);
> > +}
> > +
> >  /*
> >   * filehandle-manipulating ops.
> >   */
> > @@ -3214,6 +3219,8 @@ nfsd4_proc_compound(struct svc_rqst *rqstp)
> >  			op->replay = &cstate->replay_owner->so_replay;
> >  			nfsd4_encode_replay(resp->xdr, op);
> >  			status = op->status = op->replay->rp_status;
> > +			if (op->opdesc->op_release)
> > +				op->opdesc->op_release(&op->u);
> >  		} else {
> >  			nfsd4_encode_operation(resp, op);
> >  			status = op->status;
> 
> I think this patch is good, but I think it would be even better if the
> ->op_release() call were moved out of nfsd4_encode_operation() and
> places after this if-else.  Then there would be only one call-site in a
> fairly obviously-correct place.
> But:
>   Reviewed-by: NeilBrown <neil@brown.name>
> for if you just want to stick with this version.
> 
> Thanks,
> NeilBrown
> 

I like that idea.

I'll be testing a pile of other patches today anyway, so I'll make this
change and test it alongside the rest.

Chuck, you can either take this one and I'll do a cleanup patch along
the lines of what Neil suggests, or I can send a v3.

> 
> > @@ -3718,6 +3725,7 @@ static const struct nfsd4_operation nfsd4_ops[] = {
> >  	},
> >  	[OP_OPEN] = {
> >  		.op_func = nfsd4_open,
> > +		.op_release = nfsd4_open_release,
> >  		.op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
> >  		.op_name = "OP_OPEN",
> >  		.op_rsize_bop = nfsd4_open_rsize,
> > 
> > ---
> > base-commit: 6c0004650ba248a12937ada16f9ba961b35ce2b5
> > change-id: 20260531-nfsd-testing-9122bf51ce95
> > 
> > Best regards,
> > -- 
> > Jeff Layton <jlayton@kernel.org>
> > 
> > 

-- 
Jeff Layton <jlayton@kernel.org>

On 6/1/26 6:41 AM, Jeff Layton wrote:
> On Mon, 2026-06-01 at 11:06 +1000, NeilBrown wrote:

>> I think this patch is good, but I think it would be even better if the
>> ->op_release() call were moved out of nfsd4_encode_operation() and
>> places after this if-else.  Then there would be only one call-site in a
>> fairly obviously-correct place.
>> But:
>>   Reviewed-by: NeilBrown <neil@brown.name>
>> for if you just want to stick with this version.
>>
>> Thanks,
>> NeilBrown
>>
> 
> I like that idea.
> 
> I'll be testing a pile of other patches today anyway, so I'll make this
> change and test it alongside the rest.
> 
> Chuck, you can either take this one and I'll do a cleanup patch along
> the lines of what Neil suggests, or I can send a v3.

v3 sounds fine.


-- 
Chuck Lever