Re: [PATCH 0/2] ovl: support NFS as lower layer

[sa <sa-dev@rainbow.by>]

On 29.06.2015 17:35, Miklos Szeredi wrote:
> On Sun, Jun 28, 2015 at 05:38:35PM +0300, sa wrote:
>> On 04.06.2015 16:29, Miklos Szeredi wrote:
>>
>>> Two small patches implementing this follow.  Comments and testing welcome.
>>>    git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/vfs.git overlayfs-next
>>> ---
>>>   fs/overlayfs/super.c | 81 +++++++++++++++++++++++++++++++++++++++++++++++-----
>>>   1 file changed, 74 insertions(+), 7 deletions(-)
>>
>> Thank you, it works in our environment mostly as expected, except
>> of the following:
>> files already existing in readonly NFS lowerdir can't be opened
>> for writing (but can be deleted). Newly created files are writable
>> as expected. The same lowerdir works with aufs without problem.
>
> Is the lower fs exported read-only on the NFS server?
>
> If so, can you please apply the following patch and specify
> 'default_permissions' mount option?


It works fine.

(sid)root@usbsid:/var# cat test
00000
(sid)root@usbsid:/var# echo 11111 > test
(sid)root@usbsid:/var# cat test
11111
(sid)root@usbsid:/var# echo 22222 >> test
(sid)root@usbsid:/var# cat test
11111
22222
(sid)root@usbsid:/var# mount | grep overlay
overlay on /var type overlay 
(rw,relatime,lowerdir=/auvar.ro,upperdir=/auvar.rw/.upperdir,workdir=/auvar.rw/.workdir,default_permissions)

Lower fs exported with (ro,async,no_root_squash,no_subtree_check).


-- 
sa