Re: [lkp-robot] [rcu] b332151a29: kernel_BUG_at_mm/slab.c

[Jens Axboe]

On 01/20/2017 09:09 AM, Sebastian Andrzej Siewior wrote:
> On 2017-01-20 08:32:37 [-0800], Jens Axboe wrote:
>> That's alright, sounds like it's not a -next regression, but rather something
>> that is already broken. I can reproduce a lot of breakage if I enable
>> CONFIG_DEBUG_TEST_DRIVER_REMOVE, in fact my system doesn't boot at all. This
>> is the first bug:
>>
>> [   18.247895] ------------[ cut here ]------------
>> [   18.247907] WARNING: CPU: 21 PID: 2223 at drivers/ata/libata-core.c:6522 ata_host_detach+0x11b]
>> [   18.247908] Modules linked in: igb(+) ahci(+) libahci i2c_algo_bit dca libata nvme(+) nvme_core
>> [   18.247917] CPU: 21 PID: 2223 Comm: systemd-udevd Tainted: G        W       4.10.0-rc4+ #30
>> [   18.247919] Hardware name: Dell Inc. PowerEdge T630/0NT78X, BIOS 2.3.4 11/09/2016
>> [   18.247919] Call Trace:
>> [   18.247928]  dump_stack+0x68/0x93
>> [   18.247934]  __warn+0xc6/0xe0
>> [   18.247937]  warn_slowpath_null+0x18/0x20
>> [   18.247943]  ata_host_detach+0x11b/0x120 [libata]
> â
> 
>> and it's even more downhill from there. That option is marked unstable, are we
>> expecting it to work right now?
> 
> Well, as per 248ff0216543 ("driver core: Make Kconfig text for
> DEBUG_TEST_DRIVER_REMOVE stronger"):
> 
> |   The current state of driver removal is not great.
> |   CONFIG_DEBUG_TEST_DRIVER_REMOVE finds lots of errors. The help text
> |   currently undersells exactly how many errors this option will find. Add
> |   a bit more description to indicate this option shouldn't be turned on
> |   unless you actually want to debug driver removal. The text can be
> |   changed later when more drivers are fixed up.
> 
> so it looks like the option is working but it uncovers bugs. I've put
> you in TO because the breakage in kvm test went away after I disabled
> the MQ support in SCSI. So I *assumed* that MQ was not doing something
> right in the removal path. I don't know if this libata-core backtrace is
> a false positive or not.

Sure, I get that, my question is just if it's always finding valid bugs,
or if the test itself is buggy. The fact that I can't boot anything after
enabling it makes me suspicious.

Or maybe the state of load/remove/load is just pretty sad.

-- 
Jens Axboe