Hi Christoph,

Am 01.08.2021 um 07:31 schrieb Michael Schmitz:
> Hi Christoph,
>> I'll back out the top commit of your series (the one removing set_fs())
>> and retest. Again, this may take a few days to show any results.
>
> No further format errors, kernel panics or lockups seen after almost a
> week of tests.
>
> That suggests something is wrong with your 'm68k: remove set_fs()' commit.
>
> I'll try adding a get_fc() and warn whenever FC does not match what your
> patch expects, maybe that can help to get a clearer picture.

See attached patches, to be applied on top of your RFC series - ran this 
for five days now, with no further errors. Will run this a while longer 
yet, but in the ordinary course of testing, I'd have seen errors by now.

Haven't seen the WARN_ON trigger once yet though, which is more than a 
little odd. Heisenbug?

I am aware that this patch defeats the purpose of the 'lets lose set_fs' 
patch series, and Linus was quite convincing in saying preemption 
couldn't be an issue so saving DFC ought not to be necessary. If there 
is anything else I can try to get to the bottom of these format errors, 
please say so.

Cheers,

	Michael