Re: [lkp] [kallsyms] bf2d2b07db: kasan: GPF could be caused by NULL-ptr deref or user memory accessgeneral protection fault: 0000 [#1] general protection fault: 0000 [#1] PREEMPT PREEMPT KASANKASAN

[Ard Biesheuvel]

On 2 February 2016 at 04:48, Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
> On 02/01/2016 04:20 PM, Andrew Morton wrote:
>>
>> On Thu, 28 Jan 2016 09:12:15 +0800 kernel test robot
>> <ying.huang@xxxxxxxxxxxxxxx> wrote:
>>
>>> FYI, we noticed the below changes on
>>>
>>> https://git.linaro.org/people/ard.biesheuvel/linux-arm arm64-kaslr-v4a
>>> commit bf2d2b07db19001ae0bd55826025b0ba47fae0c2 ("kallsyms: add support
>>> for relative offsets in kallsyms address table")
>>>
>>>
>>>
>>> +-----------------------------------------------------------+------------+------------+
>>> |                                                           | 2c4d21df0f
>>> | bf2d2b07db |
>>>
>>> +-----------------------------------------------------------+------------+------------+
>>> | boot_successes                                            | 10
>>> | 0          |
>>> | boot_failures                                             | 6
>>> | 36         |
>>> | Kernel_panic-not_syncing:Attempted_to_kill_init!exitcode= | 2
>>> |            |
>>> | IP-Config:Auto-configuration_of_network_failed            | 4
>>> |            |
>>> | general_protection_fault:#[##]PREEMPT_PREEMPT_KASANKASAN  | 0
>>> | 36         |
>>> | general_protection_fault:#[##]                            | 0
>>> | 36         |
>>> | BUG:kernel_boot_hang                                      | 0
>>> | 36         |
>>>
>>> +-----------------------------------------------------------+------------+------------+
>>>
>>>
>>>
>>> [ 0.281636] kasan: CONFIG_KASAN_INLINE enabled
>>>
>>> [ 0.282416] kasan: GPF could be caused by NULL-ptr deref or user memory
>>> access
>>> [ 0.282416] kasan: GPF could be caused by NULL-ptr deref or user memory
>>> accessgeneral protection fault: 0000 [#1] general protection fault: 0000
>>> [#1] PREEMPT PREEMPT KASANKASAN
>>>
>>> [ 0.284561] Modules linked in:
>>> [ 0.284561] Modules linked in:
>>>
>>> [ 0.285136] CPU: 0 PID: 1 Comm: swapper Not tainted
>>> 4.5.0-rc1-00036-gbf2d2b0 #1
>>> [ 0.285136] CPU: 0 PID: 1 Comm: swapper Not tainted
>>> 4.5.0-rc1-00036-gbf2d2b0 #1
>>> [ 0.286438] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
>>> Debian-1.8.2-1 04/01/2014
>>> [ 0.286438] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
>>> Debian-1.8.2-1 04/01/2014
>>> [ 0.288000] task: ffff88000fcb0000 ti: ffff88000fcb8000 task.ti:
>>> ffff88000fcb8000
>>> [ 0.288000] task: ffff88000fcb0000 ti: ffff88000fcb8000 task.ti:
>>> ffff88000fcb8000
>>> [ 0.289287] RIP: 0010:[<ffffffff8120688c>]
>>
>>
>> I'm not sufficiently familiar with KASAN to be able to interpret this.
>> Perhaps Andrey can shed some light?
>>
>> The lack of symbol decoding is a problem.
>>
>> CONFIG_KALLSYMS=y
>> CONFIG_KALLSYMS_ALL=y
>> CONFIG_KALLSYMS_BASE_RELATIVE=y
>>
>> It should be there.  Perhaps it is not because this error is itself
>> (possibly) related to kallsyms changes?
>>
>
> Probably. I had seen something similar (crash and no symbols) with an
> earlier version of the patch,
> if the symbols were not created. There might be a message in the log,
> indicating that kallsyms
> failed to generate the symbols.
>

Please disregard. That is a working branch for my kaslr arm64 series,
in which I replaced my versions of the kallsyms patches with the
versions that have your T-b and S-o-b's
However, the branch does not have the x86 smp patch (it does have the
absolute/percpu patch since it was required for context):
http://ozlabs.org/~akpm/mmotm/broken-out/x86-kallsyms-disable-absolute-percpu-symbols-on-smp.patch

I have tried building this .config locally, and applying the above
patch gets things working again.

Thanks,
Ard.