Re: 5.1.0-rc4: Oops in __rpc_execute() when trying to boot from NFS

[Daniel Mack]

On 9/4/2019 6:55 PM, Trond Myklebust wrote:
> On Tue, 2019-04-09 at 18:25 +0200, Daniel Mack wrote:
>> On 8/4/2019 8:51 PM, Trond Myklebust wrote:
>>> On Mon, 2019-04-08 at 19:01 +0200, Daniel Mack wrote:
>>>> Hi,
>>>>
>>>> I'm seeing the Oops below when trying to boot 5.1.0-rc4 on an ARM
>>>> PXA3xx
>>>> platform. v5.0 did not show this effect with the same cmdline.
>>>>
>>> Please do bisect if that is at all practical. I'm having trouble
>>> interpreting this Oops.
>>
>> Here you go:
>>
>> 009a82f6437490c262584d65a14094a818bcb747 is the first bad commit
>> commit 009a82f6437490c262584d65a14094a818bcb747
>> Author: Trond Myklebust <trond.myklebust@xxxxxxxxxxxxxxx>
>> Date:   Sat Mar 9 12:07:17 2019 -0500
>>
>>     SUNRPC: Micro-optimise when the task is known not to be sleeping
>>
>>     In cases where we know the task is not sleeping, try to optimise
>>     away the indirect call to task->tk_action() by replacing it with
>>     a direct call.
>>     Only change tail calls, to allow gcc to perform tail call
>>     elimination.
> 
> Ah... It looks like we explicitly turn off tail call optimisation in
> some ARM configs, so this might be a stack overflow.
> 
> Does your config file have THUMB2_AVOID_R_ARM_THM_JUMP11 set?

Nope. I don't even have THUMB2_KERNEL.

In the meantime, I tried to trace that with some printks, but the bug
appears evasive, and the backtrace changes as soon as I modify the
timing. Hmm.

Happy to test patches if you have any idea.


Thanks,
Daniel