Re: [PATCH] perf/x86/intel/uncore: fix IMC missing box initialization

[Stephane Eranian]

On Fri, Apr 24, 2015 at 12:22 PM, Bjorn Helgaas <bhelgaas@xxxxxxxxxx> wrote:
> On Fri, Apr 24, 2015 at 9:38 AM, Stephane Eranian <eranian@xxxxxxxxxx> wrote:
>> On Fri, Apr 24, 2015 at 7:03 AM, Vince Weaver <vincent.weaver@xxxxxxxxx> wrote:
>>> On Fri, 24 Apr 2015, Vince Weaver wrote:
>>>
>>>> [   79.473121] kernel BUG at mm/vmalloc.c:1335!
>>>> [   79.477705] invalid opcode: 0000 [#1] SMP
>>> ...
>>>> [   79.694391] Call Trace:
>>>> [   79.696997]  <IRQ>
>>>> [   79.699090]  [<ffffffff811b2130>] get_vm_area_caller+0x40/0x50
>>>> [   79.705505]  [<ffffffff81039f4d>] ? snb_uncore_imc_init_box+0x6d/0x90
>>>
>>> This maps to:
>>>
>>>         static void snb_uncore_imc_init_box(struct intel_uncore_box *box) {
>>>                 ...
>>>                 box->io_addr = ioremap(addr, SNB_UNCORE_PCI_IMC_MAP_SIZE);
>>>
>>> The machine I am seeing this on is a Haswell desktop, Intel 6/60/3.
>>>
>> I also got that on one Lenovo IvyBridge laptop but I had just updated the BIOS.
>> It was working before with older kernels. I am not sure what is going on here.
>> It will be easier to debug once I get my desktop Haswell back and setup with
>> serial console. Need to verify that the address of the BAR is the same compared
>> with older kernels.
>>
>> Maybe Bjorn can shed some lights. Why would ioremap() die like this in the
>> latest 4.0 tree.
>
>   1328  static struct vm_struct *__get_vm_area_node(unsigned long size,
>   1329                  unsigned long align, unsigned long flags,
> unsigned long start,
>   1330                  unsigned long end, int node, gfp_t gfp_mask,
> const void *caller)
>   1331  {
>   1332          struct vmap_area *va;
>   1333          struct vm_struct *area;
>   1334
>   1335          BUG_ON(in_interrupt());           <-----------------
>
> Is there some perfmon initialization happening in interrupt context?

May be in IPI context if I read Vince's stack trace.

This leads me to believe that this patch:

commit c05199e5a57a579fea1e8fa65e2b511ceb524ffc
Author: Kan Liang <kan.liang@xxxxxxxxx>
Date:   Tue Jan 20 04:54:25 2015 +0000

    perf/x86/intel/uncore: Move uncore_box_init() out of driver initialization

If I revert it, I bet things will work again.
--
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/