Re: [BUG] cpu hotplug vs scheduler

[Avi Kivity <avi@qumranet.com>]

Heiko Carstens wrote:
> On Wed, May 21, 2008 at 03:42:51PM +0300, Avi Kivity wrote:
>   
>> Heiko Carstens wrote:
>>     
>>> I think you fixed this specific bug. Does cpu hotplug stresstest now work
>>> for you?
>>>       
>> Which bug, the order-2 allocs?  I certainly didn't fix it.
>>
>> Last I tried, cpu hotplug still failed, but it was some time ago as  
>> patches flow.
>>     
>
> Ah no. I thought "x86: fix crash on cpu hotplug on pat-incapable machines"
> would have fixed the cpu hotplug bug for you.
>
>   

No, that only fixes an immediate oops when running in a virtual machine 
(which doesn't have pat).

> I was actually talking of the pick_next_task_fair incarnation of one of
> the cpu hotplug bugs.
>
>   

That hasn't been addressed, AFAIK.

> s390 backtrace on 2.6.25 looks like this.
>
>     <1>Unable to handle kernel pointer dereference at virtual kernel address 00000200000db000
>     <4>Oops: 003b [#1] PREEMPT SMP 
>     <4>Modules linked in: dm_multipath sunrpc qeth_l2 dm_mod cu3088 qeth ccwgroup
>     <4>CPU: 0 Not tainted 2.6.25-27.x.20080422-s390xdefault #1
>     <4>Process migration/0 (pid: 5806, task: 00000000aac3a838, ksp: 000000017a2f3dd0)
>     <4>Krnl PSW : 0400200180000000 0000000000125f48 (pick_next_task_fair+0x34/0xd4)
>     <4>           R:0 T:1 IO:0 EX:0 Key:0 M:0 W:0 P:0 AS:0 CC:2 PM:0 EA:3
>     <4>Krnl GPRS: 0000079e029c6ecb 0000000000000000 000000008548ea60 000000008548d700
>     <4>           00000000004526f8 0000000000000597 0000000000000000 0000000000000000
>     <4>           0000000000672a00 0000000000000000 00000000aac3a838 000000017a2f3d10
>     <4>           00000200000db85a 00000000004672a0 000000017a2f3d50 000000017a2f3d10
>     <4>Krnl Code: 0000000000125f38: e3e0f0980024        stg     %r14,152(%r15)
>     <4>           0000000000125f3e: d507d000c010        clc     0(8,%r13),16(%r12)
>     <4>           0000000000125f44: a784004e            brc     8,125fe0
>     <4>          >0000000000125f48: d507d000c030        clc     0(8,%r13),48(%r12)
>     <4>           0000000000125f4e: b904002c            lgr     %r2,%r12
>     <4>           0000000000125f52: a7990000            lghi    %r9,0
>     <4>           0000000000125f56: a7840033            brc     8,125fbc
>     <4>           0000000000125f5a: c0e5ffffee23        brasl   %r14,123ba0
>     <4>Call Trace:
>     <4>([<00000000aac3a838>] 0xaac3a838)
>     <4> [<000000000045270c>] schedule+0x178/0x978
>     <4> [<000000000012fe86>] migration_thread+0x1aa/0x30c
>     <4> [<0000000000152ad8>] kthread+0x68/0xa0
>     <4> [<000000000010a0be>] kernel_thread_starter+0x6/0xc
>     <4> [<000000000010a0b8>] kernel_thread_starter+0x0/0xc
>     <4>Last Breaking-Event-Address:
>     <4> [<0000000000125fc6>] pick_next_task_fair+0xb2/0xd4
>   

Seems to be the same bug, yes.

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