CIFS shares keep hanging with VirtualBox

CIFS shares keep hanging with VirtualBox

[Frédéric L. W. Meunier]

[-- Attachment #1: Type: text/plain, Size: 1193 bytes --]

I use VirtualBox (4.1.12) on Windows 7 Ultimate SP1. The guest is Linux 
(kernel 3.3.2, cifs-utils 5.3), which I boot directly from a partition.

If I use CIFS to access the Windows partitions, the connections 
constantly hang and the applications to access it deadlock until killed.

For example:

$ cd /mnt
$ ls

It hangs (the share in question is mounted under /mnt/ntfs2).

'ps' returns "3174 pts/0 D+ 0:00 ls --color=auto" and the load average 
increases until I 'kill -9' the process.

The only message I get in the logs is "CIFS VFS: Server 192.168.1.64 has 
not responded in 300 seconds. Reconnecting..."

After that, the share only works if remounted.

I don't know what's causing it, but it works perfectly from VMware 
Player using the same setup, so it looks like a bug in VirtualBox, not 
in CIFS ?

This isn't a new issue, as I started using VMware Player in July 2011 
after giving up getting CIFS to work reliably with VirtualBox.

I'm attaching a log created with 'echo 7 > /proc/fs/cifs/cifsFYI', just 
in case it helps, and am willing to do other things to, it not solve it, 
at least know what to blame. At 14:26, I killed both 'ls' processes and 
unmounted the partition.

[-- Attachment #2: cifs.txt.gz --]
[-- Type: application/octet-stream, Size: 5054 bytes --]

Re: CIFS shares keep hanging with VirtualBox

[Frédéric L. W. Meunier]

[-- Attachment #1: Type: text/plain, Size: 1185 bytes --]

On Thu, Apr 19, 2012 at 01:51:51PM -0400, Jeff Layton wrote:
> 
> On Thu, 19 Apr 2012 14:36:27 -0300 Frédéric L. W. Meunier wrote:
> 
> > I use VirtualBox (4.1.12) on Windows 7 Ultimate SP1. The guest is Linux 
> > (kernel 3.3.2, cifs-utils 5.3), which I boot directly from a partition.
> > 
> > If I use CIFS to access the Windows partitions, the connections 
> > constantly hang and the applications to access it deadlock until killed.
> 
> Looks like the server just isn't responding to echo requests. Sniffing
> the traffic between client and server might be good to confirm that:
> 
>     https://wiki.samba.org/index.php/LinuxCIFS_troubleshooting#Wire_Captures
> 
> Are you connecting to the CIFS server in windows 7 with this? Or some
> other VMware SMB server thingy?

Yes. It's the same computer. I use VirtualBox in Windows 7 to boot 
Linux, which is installed in the same hard drive. Then, in Linux, I use 
CIFS to have access to the Windows 7 partitions.

I captured a log with 'tcpdump', but I don't think it's of any use. It 
stopped logging at 16:01 but the hang happened after that. It didn't log 
anything from the problematic 'ls', 'kill -9' and 'umount /mnt/ntfs2'.

[-- Attachment #2: cifs-traffic.pcap.gz --]
[-- Type: application/octet-stream, Size: 9636 bytes --]

Re: CIFS shares keep hanging with VirtualBox

[Steve French]

On Thu, Apr 19, 2012 at 2:43 PM, Frédéric L. W. <fredlwm-fLz7/M+a0ZxNg+MwTxZMZA@public.gmane.orgt> wrote:
> On Thu, Apr 19, 2012 at 01:51:51PM -0400, Jeff Layton wrote:
>>
>> On Thu, 19 Apr 2012 14:36:27 -0300 Frédéric L. W. Meunier wrote:
>>
>> > I use VirtualBox (4.1.12) on Windows 7 Ultimate SP1. The guest is Linux
>> > (kernel 3.3.2, cifs-utils 5.3), which I boot directly from a partition.
>> >
>> > If I use CIFS to access the Windows partitions, the connections
>> > constantly hang and the applications to access it deadlock until killed.
>>
>> Looks like the server just isn't responding to echo requests. Sniffing
>> the traffic between client and server might be good to confirm that:
>>
>>     https://wiki.samba.org/index.php/LinuxCIFS_troubleshooting#Wire_Captures
>>
>> Are you connecting to the CIFS server in windows 7 with this? Or some
>> other VMware SMB server thingy?
>
> Yes. It's the same computer. I use VirtualBox in Windows 7 to boot
> Linux, which is installed in the same hard drive. Then, in Linux, I use
> CIFS to have access to the Windows 7 partitions.
>
> I captured a log with 'tcpdump', but I don't think it's of any use. It
> stopped logging at 16:01 but the hang happened after that. It didn't log
> anything from the problematic 'ls', 'kill -9' and 'umount /mnt/ntfs2'.

Maybe capture a log (with netmon on wireshark) on the windows side
to see what the trace shows.

Presumably some odd networking issue with virtual box, but
may  be useful to see which command fails/timesout by
looking at dmesg output on the client (running the
test after enabling /proc/fs/cifs/cifsFYI setting it
to 7 e.g.).

At least with the server returning maxmpx of 50, it is less
likely to be the bug with too many outstanding requests.



-- 
Thanks,

Steve

Re: CIFS shares keep hanging with VirtualBox

[Frédéric L. W. Meunier]

On Thu, Apr 19, 2012 at 03:19:13PM -0500, Steve French wrote:
> On Thu, Apr 19, 2012 at 2:43 PM, Frédéric L. W. wrote:
> >
> > I captured a log with 'tcpdump', but I don't think it's of any use. It
> > stopped logging at 16:01 but the hang happened after that. It didn't log
> > anything from the problematic 'ls', 'kill -9' and 'umount /mnt/ntfs2'.
> 
> Maybe capture a log (with netmon on wireshark) on the windows side
> to see what the trace shows.
> 
> Presumably some odd networking issue with virtual box, but
> may  be useful to see which command fails/timesout by
> looking at dmesg output on the client (running the
> test after enabling /proc/fs/cifs/cifsFYI setting it
> to 7 e.g.).
> 
> At least with the server returning maxmpx of 50, it is less
> likely to be the bug with too many outstanding requests.

I noticed that if I change the connection in VirtualBox from NAT to 
Bridged Adapter, the hangs don't occur (at least after about 2h).

Are you still interested in a log from Windows in NAT mode, where the 
hangs happen, or this is definitely an issue / "feature" in VirtualBox ?

What I don't know is if CIFS could do anything to exit gracefully 
instead of deadlocking the application trying to access the share.

Re: CIFS shares keep hanging with VirtualBox

[Steve French]

On Thu, Apr 19, 2012 at 6:39 PM, Frédéric L. W. <fredlwm-fLz7/M+a0ZxNg+MwTxZMZA@public.gmane.orgt> wrote:
> On Thu, Apr 19, 2012 at 03:19:13PM -0500, Steve French wrote:
>> On Thu, Apr 19, 2012 at 2:43 PM, Frédéric L. W. wrote:
>> >
>> > I captured a log with 'tcpdump', but I don't think it's of any use. It
>> > stopped logging at 16:01 but the hang happened after that. It didn't log
>> > anything from the problematic 'ls', 'kill -9' and 'umount /mnt/ntfs2'.
>>
>> Maybe capture a log (with netmon on wireshark) on the windows side
>> to see what the trace shows.
>>
>> Presumably some odd networking issue with virtual box, but
>> may  be useful to see which command fails/timesout by
>> looking at dmesg output on the client (running the
>> test after enabling /proc/fs/cifs/cifsFYI setting it
>> to 7 e.g.).
>>
>> At least with the server returning maxmpx of 50, it is less
>> likely to be the bug with too many outstanding requests.
>
> I noticed that if I change the connection in VirtualBox from NAT to
> Bridged Adapter, the hangs don't occur (at least after about 2h).
>
> Are you still interested in a log from Windows in NAT mode, where the
> hangs happen, or this is definitely an issue / "feature" in VirtualBox ?
>
> What I don't know is if CIFS could do anything to exit gracefully
> instead of deadlocking the application trying to access the share.

I would love to investigate in more detail (obviously if easy cifs
changes can lessen or resolve strange networking problems, we are
all for it) but I have other problems with virtualbox setup at the moment
that make it hard for me to see whether I run into the same issue.

-- 
Thanks,

Steve