Re: BUG: using smp_processor_id() in preemptible

[Avi Kivity <avi@redhat.com>]

On 06/29/2009 11:32 AM, Johannes Berg wrote:
> On Sun, 2009-06-28 at 17:11 +0300, Avi Kivity wrote:
>
>    
>>> 1) I get the $subject warning a lot, when starting kvm:
>>> [85763.262707] BUG: using smp_processor_id() in preemptible [00000000] code: kvm/13877
>>> [85763.262719] caller is kvm_write_guest_time+0x40/0x220 [kvm]
>>> [85763.262722] Pid: 13877, comm: kvm Not tainted 2.6.30-wl-26837-g0ee651a-dirty #54
>>> [85763.262725] Call Trace:
>>> [85763.262729]  [<ffffffff8041d482>] debug_smp_processor_id+0xf2/0x100
>>> [85763.262741]  [<ffffffffa0331390>] kvm_write_guest_time+0x40/0x220 [kvm]
>>> [85763.262753]  [<ffffffffa0331890>] vcpu_enter_guest+0x320/0x580 [kvm]
>>> [85763.262780]  [<ffffffffa03347f4>] __vcpu_run+0x74/0x2f0 [kvm]
>>> [85763.262792]  [<ffffffffa033571f>] kvm_arch_vcpu_ioctl_run+0x8f/0x200 [kvm]
>>> [85763.262804]  [<ffffffffa0329b48>] kvm_vcpu_ioctl+0x4b8/0x900 [kvm]
>>> [85763.262816]  [<ffffffff802f5216>] vfs_ioctl+0x36/0xb0
>>> [85763.262819]  [<ffffffff802f55f9>] do_vfs_ioctl+0x89/0x320
>>> [85763.262826]  [<ffffffff802f58df>] sys_ioctl+0x4f/0x80
>>> [85763.262830]  [<ffffffff8020b6fb>] system_call_fastpath+0x16/0x1b
>>>
>>>
>>>        
>> ISTR this was fixed...
>>
>>      
>>> That kernel version is wireless-testing, which is currently based on
>>> v2.6.30, and the -dirty is for some wireless patches I did.
>>>
>>>        
>> Please post the output of 'git merge-base wireless-testing
>> origin/master' so I can know what tree to look at.
>>      
>
> ITYM
>
> $ git merge-base wireless-testing/master linux-2.6/master
> 07a2039b8eb0af4ff464efd3dfd95de5c02648c6
> $ git describe 07a2039b8eb0af4ff464efd3dfd95de5c02648c6
> v2.6.30
>
> since my personal 'origin' branch is something completely different.
>
>    

Yes.  It was fixed in mainline by 2dea4c84bc.  I'll prepare something 
for -stable.


>> You can see where it hangs using the monitor 'info registers' and 'x/30i
>> $eip' commands.
>>      
>
> not much luck since it doesn't hang at a specific instruction:
>    

You can try mapping these with gdb (in fact, you can have gdb connect to 
qemu and do source level debugging).

-- 
error compiling committee.c: too many arguments to function