Re: syscall rmdir hangs with autofs

[Sebastian Hetze <s.hetze@linux-ag.com>]

On Mon, Jul 19, 2010 at 06:03:30PM +0300, Avi Kivity wrote:
> On 07/19/2010 05:47 PM, Sebastian Hetze wrote:
>> On Mon, Jul 19, 2010 at 05:00:52PM +0300, Avi Kivity wrote:
>>    
>>> On 07/19/2010 04:45 PM, Sebastian Hetze wrote:
>>>      
>>>>        
>>>>> It looks like VFS breakage.  Does this happen with older guest kernels?
>>>>>
>>>>>          
>>>> currently I can confirm this problem with 2.6.31-22-generic-pae back to
>>>> 2.6.31-16-generic-pae. What guest kernel version would you suggest?
>>>>
>>>>        
>>> Can you try it on a non-virtualized system?
>>>      
>> This automount setup has been running non-virtualized for quite a while
>> without problems until last year. Never seen or heared about similar hangs
>> with automount before.
>> We have not been able to reproduce this error in laboratory conditions.
>> The productive system has 2TB data and the problem occurs totally
>> unpredictable with real live workload.
>> And there is no easy way to switch back to non-virtualized unless we
>> declare this whole project a failure.
>>    
>
> Ahem.
>
> What's your hardware platform?  EPT/NPT capable?  Host kernel version?

Intel S5520HC Board with 2 Xeon CPU E5520, HT enabled
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe
syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good
xtopology nonstop_tsc aperfmperf pni dtes64 monitor ds_cpl vmx est tm2
ssse3 cx16 xtpr pdcm dca sse4_1 sse4_2 popcnt lahf_lm ida tpr_shadow
vnmi flexpriority ept vpid

(so yes, EPT is available)

host kernel is vanilla 2.6.34