Hi all,
I saw a new NFS failure in 4.13-rc2 kernel when running fstests
generic/126 on NFSv4 mounts, 4.13-rc1 kernel passed the same test.
git bisect pointed the first bad to commit bd8b2441742b.
commit bd8b2441742b49c76bec707757bd9c028ea9838e
Author: Trond Myklebust <[email protected]>
Date: Tue Jul 11 17:54:34 2017 -0400
NFS: Store the raw NFS access mask in the inode's access cache
Signed-off-by: Trond Myklebust <[email protected]>
Signed-off-by: Anna Schumaker <[email protected]>
I can reproduce it manually with these steps:
mount -t nfs -o vers=4.1 server:/export /mnt/nfs
cp /usr/bin/ls /mnt/nfs
chmod 001 /mnt/nfs/ls
chown 99:99 /mnt/nfs/ls
su testuser -c "/mnt/nfs/ls"
(testuser uid and gid are not 99.)
The last command got permission denied with 4.13-rc2.
Thanks,
Eryu
Hi Eryu,
On 07/24/2017 10:19 AM, Eryu Guan wrote:
> Hi all,
>
> I saw a new NFS failure in 4.13-rc2 kernel when running fstests
> generic/126 on NFSv4 mounts, 4.13-rc1 kernel passed the same test.
>
> git bisect pointed the first bad to commit bd8b2441742b.
>
> commit bd8b2441742b49c76bec707757bd9c028ea9838e
> Author: Trond Myklebust <[email protected]>
> Date: Tue Jul 11 17:54:34 2017 -0400
>
> NFS: Store the raw NFS access mask in the inode's access cache
>
> Signed-off-by: Trond Myklebust <[email protected]>
> Signed-off-by: Anna Schumaker <[email protected]>
>
> I can reproduce it manually with these steps:
>
> mount -t nfs -o vers=4.1 server:/export /mnt/nfs
> cp /usr/bin/ls /mnt/nfs
> chmod 001 /mnt/nfs/ls
> chown 99:99 /mnt/nfs/ls
> su testuser -c "/mnt/nfs/ls"
>
> (testuser uid and gid are not 99.)
>
> The last command got permission denied with 4.13-rc2.
Thanks for letting me know! I think I see the problem: the patch changed how the
access system call is cached but didn't update caching for NFSv4 opens (which is why
v3 still passes). Does this patch help?
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index e1a26c653e78..4893c9f9829b 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -2250,16 +2250,16 @@ static int nfs4_opendata_access(struct rpc_cred *cred,
*/
if (openflags & __FMODE_EXEC) {
/* ONLY check for exec rights */
- mask = MAY_EXEC;
+ mask = NFS4_ACCESS_EXECUTE;
} else if ((fmode & FMODE_READ) && !opendata->file_created)
- mask = MAY_READ;
+ mask = NFS4_ACCESS_READ;
cache.cred = cred;
cache.jiffies = jiffies;
nfs_access_set_mask(&cache, opendata->o_res.access_result);
nfs_access_add_cache(state->inode, &cache);
- if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
+ if ((mask & ~cache.mask & (NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE)) == 0)
return 0;
return -EACCES;
>
> Thanks,
> Eryu
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
On Tue, Jul 25, 2017 at 03:52:36PM -0400, Anna Schumaker wrote:
> Hi Eryu,
>
> On 07/24/2017 10:19 AM, Eryu Guan wrote:
> > Hi all,
> >
> > I saw a new NFS failure in 4.13-rc2 kernel when running fstests
> > generic/126 on NFSv4 mounts, 4.13-rc1 kernel passed the same test.
> >
> > git bisect pointed the first bad to commit bd8b2441742b.
> >
> > commit bd8b2441742b49c76bec707757bd9c028ea9838e
> > Author: Trond Myklebust <[email protected]>
> > Date: Tue Jul 11 17:54:34 2017 -0400
> >
> > NFS: Store the raw NFS access mask in the inode's access cache
> >
> > Signed-off-by: Trond Myklebust <[email protected]>
> > Signed-off-by: Anna Schumaker <[email protected]>
> >
> > I can reproduce it manually with these steps:
> >
> > mount -t nfs -o vers=4.1 server:/export /mnt/nfs
> > cp /usr/bin/ls /mnt/nfs
> > chmod 001 /mnt/nfs/ls
> > chown 99:99 /mnt/nfs/ls
> > su testuser -c "/mnt/nfs/ls"
> >
> > (testuser uid and gid are not 99.)
> >
> > The last command got permission denied with 4.13-rc2.
>
> Thanks for letting me know! I think I see the problem: the patch changed how the
> access system call is cached but didn't update caching for NFSv4 opens (which is why
> v3 still passes). Does this patch help?
Yes, this patch solved generic/126 failure, I tested with nfsv4.0/1/2
and nfsv3, all passed now. But I didn't do further tests against this
patch.
Thanks!
Eryu
>
>
> diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
> index e1a26c653e78..4893c9f9829b 100644
> --- a/fs/nfs/nfs4proc.c
> +++ b/fs/nfs/nfs4proc.c
> @@ -2250,16 +2250,16 @@ static int nfs4_opendata_access(struct rpc_cred *cred,
> */
> if (openflags & __FMODE_EXEC) {
> /* ONLY check for exec rights */
> - mask = MAY_EXEC;
> + mask = NFS4_ACCESS_EXECUTE;
> } else if ((fmode & FMODE_READ) && !opendata->file_created)
> - mask = MAY_READ;
> + mask = NFS4_ACCESS_READ;
>
> cache.cred = cred;
> cache.jiffies = jiffies;
> nfs_access_set_mask(&cache, opendata->o_res.access_result);
> nfs_access_add_cache(state->inode, &cache);
>
> - if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
> + if ((mask & ~cache.mask & (NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE)) == 0)
> return 0;
>
> return -EACCES;
>
>
> >
> > Thanks,
> > Eryu
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> > the body of a message to [email protected]
> > More majordomo info at http://vger.kernel.org/majordomo-info.html
> >
On 07/26/2017 01:02 AM, Eryu Guan wrote:
> On Tue, Jul 25, 2017 at 03:52:36PM -0400, Anna Schumaker wrote:
>> Hi Eryu,
>>
>> On 07/24/2017 10:19 AM, Eryu Guan wrote:
>>> Hi all,
>>>
>>> I saw a new NFS failure in 4.13-rc2 kernel when running fstests
>>> generic/126 on NFSv4 mounts, 4.13-rc1 kernel passed the same test.
>>>
>>> git bisect pointed the first bad to commit bd8b2441742b.
>>>
>>> commit bd8b2441742b49c76bec707757bd9c028ea9838e
>>> Author: Trond Myklebust <[email protected]>
>>> Date: Tue Jul 11 17:54:34 2017 -0400
>>>
>>> NFS: Store the raw NFS access mask in the inode's access cache
>>>
>>> Signed-off-by: Trond Myklebust <[email protected]>
>>> Signed-off-by: Anna Schumaker <[email protected]>
>>>
>>> I can reproduce it manually with these steps:
>>>
>>> mount -t nfs -o vers=4.1 server:/export /mnt/nfs
>>> cp /usr/bin/ls /mnt/nfs
>>> chmod 001 /mnt/nfs/ls
>>> chown 99:99 /mnt/nfs/ls
>>> su testuser -c "/mnt/nfs/ls"
>>>
>>> (testuser uid and gid are not 99.)
>>>
>>> The last command got permission denied with 4.13-rc2.
>>
>> Thanks for letting me know! I think I see the problem: the patch changed how the
>> access system call is cached but didn't update caching for NFSv4 opens (which is why
>> v3 still passes). Does this patch help?
>
> Yes, this patch solved generic/126 failure, I tested with nfsv4.0/1/2
> and nfsv3, all passed now. But I didn't do further tests against this
> patch.
Great! Thanks for testing! :)
Anna
>
> Thanks!
>
> Eryu
>
>>
>>
>> diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
>> index e1a26c653e78..4893c9f9829b 100644
>> --- a/fs/nfs/nfs4proc.c
>> +++ b/fs/nfs/nfs4proc.c
>> @@ -2250,16 +2250,16 @@ static int nfs4_opendata_access(struct rpc_cred *cred,
>> */
>> if (openflags & __FMODE_EXEC) {
>> /* ONLY check for exec rights */
>> - mask = MAY_EXEC;
>> + mask = NFS4_ACCESS_EXECUTE;
>> } else if ((fmode & FMODE_READ) && !opendata->file_created)
>> - mask = MAY_READ;
>> + mask = NFS4_ACCESS_READ;
>>
>> cache.cred = cred;
>> cache.jiffies = jiffies;
>> nfs_access_set_mask(&cache, opendata->o_res.access_result);
>> nfs_access_add_cache(state->inode, &cache);
>>
>> - if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
>> + if ((mask & ~cache.mask & (NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE)) == 0)
>> return 0;
>>
>> return -EACCES;
>>
>>
>>>
>>> Thanks,
>>> Eryu
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
>>> the body of a message to [email protected]
>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>>