Re: [2.6.39] CIFS write failures where 2.6.38 works

[Jeff Layton]

On Thu, 23 Jun 2011 13:25:36 +0200
Helge Hafting <helge.hafting@xxxxxxx> wrote:

> On 22. juni 2011 22:36, Jeff Layton wrote:
> > On Thu, 9 Jun 2011 18:28:45 -0400
> > Jeff Layton<jlayton@xxxxxxxxxx>  wrote:
> >
> >> On Fri, 03 Jun 2011 17:11:05 +0200
> >> Helge Hafting<helge.hafting@xxxxxxx>  wrote:
> >>
> >>> On 03. juni 2011 12:15, Suresh Jayaraman wrote:
> >>>> [Cc linux-cifs@xxxxxxxxxxxxxxx]
> >>>>
> >>>> On 06/01/2011 03:41 PM, Helge Hafting wrote:
> >>>>> At work I use cifs for accessing a windows server. This has worked fine
> >>>>> for a long time, up to and including Debian's 2.6.38-2.
> >>>>>
> >>>>> I just installed Debians's 2.6.39-1, and had to give up on it.
> >>>>> Mounting CIFS works, and I can see the files. But if I
> >>>>> try to make a new file (with cp), I get a long delay.
> >>>>
> >>>> What is the security mechanism you are using? If you seeing the problem
> >>>> with ntlm, could you try using ntlmv2 and see whether the problem is
> >>>> reproducible?
> >>>
> >>> In the beginning, I did not specify the mechanism. So, whatever the
> >>> default is.
> >>>
> >>> The fstab entry was like this:
> >>> \\servername\resource /mountpoint cifs
> >>> domain=MYDOMAIN,credentials=/etc/fstabcred,rw,noauto,iocharset=utf8,uid=username,gid=group,sockopt=TCP_NODELAY,users,file_mode=0640,dir_mode=0750,relatime
> >>> 0 0
> >>>
> >>> I looked at cifs options, and tried to add "sign" and "sec=ntlmv2i". It
> >>> made no difference. Still failure with 2.6.39, and mounting with these
> >>> new options works fine with 2.6.38
> >>>
> >>
> >> I think we need to understand what's happening on the wire. Are you
> >> still able to reproduce this? If so, can you turn up debug logging and
> >> reproduce this?. Instructions for how to do that are here:
> >>
> >>      http://wiki.samba.org/index.php/LinuxCIFS_troubleshooting#Enabling_Debugging
> >>
> >> Also, it looks like someone opened a bug at kernel.org too:
> >>
> >>      https://bugzilla.kernel.org/show_bug.cgi?id=36952
> >>
> >> ...so if you can attach the resulting log there, that would be great.
> >>
> >
> > I think that this is probably due to the change that added the
> > page_mkwrite function to cifs.ko. Prior to that, cifs did single-page
> > writes on signed connections. Now we do multi-page writes and windows
> > servers apparently reject large write calls on signed connections.
> >
> > One way to test this theory would be to set the wsize to something
> > smaller when you mount. For instance:
> >
> >      wsize=16384
> >
> > ...assuming that doesn't go over the server's MaxBufferSize, then that
> > should act as a workaround. Can you try that and let me know if it
> > helps?
> 
> 
> Yes, that seemed to fix it. I added wsize=16384 and mounted using 
> debians 2.6.39-1-amd64 kernel.
> 
> I tried a recursive copy of 26MB from one directory tree to another on 
> that mount. It completed in 24s with no error messages. 1MB/s is not 
> much, but there may be 40 other users.
> 
> The server runs windows 2008r2, 64-bit.
> 

Great, thanks for testing that. I've already sent Steve French a patch
that should fix this in 3.0, but we'll need to do something different
for 2.6.39 stable. I'll roll up a patch for that soon.

BTW, 3.0 should have much improved write performance, even with signing
enabled as that adds the capability to do async writes.

-- 
Jeff Layton <jlayton@xxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/