NFS4 acl ops do not work with standard kernel

With kernel 2.6.34 from http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.34-rc3-lucid/linux-image-2.6.34-020634rc3-generic_2.6.34-020634rc3_i386.deb the issue is resolved:

root@bloch:/mnt/nfs4# mount -t nfs4 -o rw,acl netapp-sim02:/vol/test test
root@bloch:/mnt/nfs4# cd test/
root@bloch:/mnt/nfs4/test# l
total 16
drwxrwxrwx 4 root nogroup 4096 Apr 7 14:46 ./
drwxr-xr-x 5 root root 4096 Apr 8 14:48 ../
drwxrwxrwx 10 root nogroup 4096 Apr 15 10:00 .snapshot/
drwxr-xr-x 2 root nogroup 4096 Apr 7 14:46 test/
root@bloch:/mnt/nfs4/test# nfs4_getfacl test/
A::OWNER@:rwaDxtTnNcCy
D::OWNER@:
A:g:GROUP@:rxtncy
D:g:GROUP@:waDTC
A::EVERYONE@:rxtncy
D::EVERYONE@:waDTC

I can confirm this bug.

Is there a chance the fix in the kernel package will become a StableReleaseUpdate? In my opinion it should. Otherwise this whole package is useless.

I'd like to second that. Deploying a custom kernel is doable, but it's a pain.

Hi Robert,

If you could also please test the latest upstream kernel available that would be great. It will allow additional upstream developers to examine the issue. Refer to https://wiki.ubuntu.com/KernelMainlineBuilds . Once you've tested the upstream kernel, please remove the 'needs-upstream-testing' tag. This can be done by clicking on the yellow pencil icon next to the tag located at the bottom of the bug description and deleting the 'needs-upstream-testing' text. Please let us know your results.

Thanks in advance.

    [This is an automated message. Apologies if it has reached you inappropriately; please just reply to this message indicating so.]

I can also confirm after installing the latest upstream kernel (linux-image-2.6.35-999-generic_2.6.35-999.201006281005_amd64) the problem is solved.

I've been struggling with this issue for a couple of months as well.
I used to solve it by patching and recompiling the kernel.
See here for more details:
  http://linux-nfs.org/pipermail/nfsv4/2009-November/011643.html

The issue has been fixed in vanilla kernel from 2.6.33.2. Also see:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=patch;\
h=462d60577a997aa87c935ae4521bd303733a9f2b

All it takes is to apply the patch below. Strangely enough, Debian Lenny has fixed as.
As have a lot of other distributions, especially server oriented. After all, it was diagnosed and fixed in Nov 2009.

diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
index 83ad47c..32b11c0 100644
--- a/fs/nfs/nfs4xdr.c
+++ b/fs/nfs/nfs4xdr.c
@@ -2096,7 +2096,7 @@ nfs4_xdr_enc_getacl(struct rpc_rqst *req, __be32 *p,
  encode_compound_hdr(&xdr, req, &hdr);
  encode_sequence(&xdr, &args->seq_args, &hdr);
  encode_putfh(&xdr, args->fh, &hdr);
- replen = hdr.replen + nfs4_fattr_bitmap_maxsz + 1;
+ replen = hdr.replen + op_decode_hdr_maxsz + nfs4_fattr_bitmap_maxsz + 1;
  encode_getattr_two(&xdr, FATTR4_WORD0_ACL, 0, &hdr);

  xdr_inline_pages(&req->rq_rcv_buf, replen << 2,

I can confirm that the ppa kernels 2.6.33 and up fix the problem, but they should as the vanilla kernel has this fix.
However, running a stock Ubuntu Lucid kernel with the fix instead of a ppa kernel is much preferred.

Since lucid is an LTS version, couldn't this be fixed in the standard kernel?

Looks like 462d60577a997aa87c935ae4521bd303733a9f2 hasn't gone into the linux-2.6.32.y tree; not sure why. Potential for an SRU here..

Sorry, that should be d327cf7449e6fd5cbac784c641770e9366faa386.

Hi Greg,

Looks like this fixes an issue on 2.6.32.17 -
https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/562913 . The
patch has hit mainline (as of 2.6.33, in commit
d327cf7449e6fd5cbac784c641770e9366faa386), but has missed stable.

Bruce - let me know if there's any reason this shouldn't go in.

Please consider for inclusion in the 2.6.32.y stable series.

Regards,

Jeremy

>From 394cc62815fdac2b3effe952588630c8c3e0629f Mon Sep 17 00:00:00 2001
From: J. Bruce Fields <email address hidden>
Date: Thu, 3 Dec 2009 08:10:17 -0500
Subject: [PATCH] Re: acl trouble after upgrading ubuntu

Subject: [PATCH] nfs: fix acl decoding

Commit 28f566942c6b1d929f5e240e69e7081b77b238d3 "NFS: use dynamically
computed compound_hdr.replen for xdr_inline_pages offset" accidentally
changed the amount of space to allow for the acl reply, resulting in an
IO error on attempts to get an acl.

Reported-by: Paul Rudin <email address hidden>
Cc: Benny Halevy <email address hidden>
Signed-off-by: J. Bruce Fields <email address hidden>
Signed-off-by: Trond Myklebust <email address hidden>
---
 fs/nfs/nfs4xdr.c | 2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
index 5ec74cd..e81b2bf 100644
--- a/fs/nfs/nfs4xdr.c
+++ b/fs/nfs/nfs4xdr.c
@@ -2096,7 +2096,7 @@ nfs4_xdr_enc_getacl(struct rpc_rqst *req, __be32 *p,
  encode_compound_hdr(&xdr, req, &hdr);
  encode_sequence(&xdr, &args->seq_args, &hdr);
  encode_putfh(&xdr, args->fh, &hdr);
- replen = hdr.replen + nfs4_fattr_bitmap_maxsz + 1;
+ replen = hdr.replen + op_decode_hdr_maxsz + nfs4_fattr_bitmap_maxsz + 1;
  encode_getattr_two(&xdr, FATTR4_WORD0_ACL, 0, &hdr);

  xdr_inline_pages(&req->rq_rcv_buf, replen << 2,
--
1.7.0.4

On Tue, Aug 03, 2010 at 02:37:18PM +0800, Jeremy Kerr wrote:
> Hi Greg,
>
> Looks like this fixes an issue on 2.6.32.17 -
> https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/562913 . The
> patch has hit mainline (as of 2.6.33, in commit
> d327cf7449e6fd5cbac784c641770e9366faa386), but has missed stable.

Now queued up.

thanks,

greg k-h

On Tue, Aug 03, 2010 at 02:37:18PM +0800, Jeremy Kerr wrote:
> Hi Greg,
>
> Looks like this fixes an issue on 2.6.32.17 -
> https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/562913 . The
> patch has hit mainline (as of 2.6.33, in commit
> d327cf7449e6fd5cbac784c641770e9366faa386), but has missed stable.
>
> Bruce - let me know if there's any reason this shouldn't go in.

Apologies, yes, I think it should have.

--b.

>
> Please consider for inclusion in the 2.6.32.y stable series.
>
> Regards,
>
>
> Jeremy
>
>
> >From 394cc62815fdac2b3effe952588630c8c3e0629f Mon Sep 17 00:00:00 2001
> From: J. Bruce Fields <email address hidden>
> Date: Thu, 3 Dec 2009 08:10:17 -0500
> Subject: [PATCH] Re: acl trouble after upgrading ubuntu
>
> Subject: [PATCH] nfs: fix acl decoding
>
> Commit 28f566942c6b1d929f5e240e69e7081b77b238d3 "NFS: use dynamically
> computed compound_hdr.replen for xdr_inline_pages offset" accidentally
> changed the amount of space to allow for the acl reply, resulting in an
> IO error on attempts to get an acl.
>
> Reported-by: Paul Rudin <email address hidden>
> Cc: Benny Halevy <email address hidden>
> Signed-off-by: J. Bruce Fields <email address hidden>
> Signed-off-by: Trond Myklebust <email address hidden>
> ---
> fs/nfs/nfs4xdr.c | 2 +-
> 1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
> index 5ec74cd..e81b2bf 100644
> --- a/fs/nfs/nfs4xdr.c
> +++ b/fs/nfs/nfs4xdr.c
> @@ -2096,7 +2096,7 @@ nfs4_xdr_enc_getacl(struct rpc_rqst *req, __be32 *p,
> encode_compound_hdr(&xdr, req, &hdr);
> encode_sequence(&xdr, &args->seq_args, &hdr);
> encode_putfh(&xdr, args->fh, &hdr);
> - replen = hdr.replen + nfs4_fattr_bitmap_maxsz + 1;
> + replen = hdr.replen + op_decode_hdr_maxsz + nfs4_fattr_bitmap_maxsz + 1;
> encode_getattr_two(&xdr, FATTR4_WORD0_ACL, 0, &hdr);
>
> xdr_inline_pages(&req->rq_rcv_buf, replen << 2,
> --
> 1.7.0.4
>
>
>

Since this bug is a showstopper for us (we are in the process of setting up an Ubuntu 10.04 NFS v4 server with ~1500 Ubuntu 10.04 client machines) I would very much like to know if the mentioned patch will be making it into the kernel for lucid (1) at all and (2) if there has been a decision to include it is there a time estimate on when it will make it into the mainstream?

Thanks for your time and great work!

This is now fixed in lucid's linux-image-2.6.32-25-server package.

We have been trying to get NFS4 ACLs working and this issue greatly affects us (showstopper for workstation deployment). I can confirm the bug is fixed in the server kernel image (tried with 2.6.32-27.49), but is seems to be still unpatched in the generic image (tried version 2.6.32-27.29). When will this be patched in the generic kernel?

The patch mention in comment #10 has been in Lucid since 2.6.32-25.43 (all kernels with that number are built from the same source so server and generic are the same). Did a quick test on the current Lucid kernel and getting ACLs seems to work. Marking as fixed. If this issue does not seem to be fixed, feel free to re-open or create a new bug.