Re: Kernel Oops in cdc_acm

[Oliver Neukum <oneukum@suse.com>]

Am Dienstag, den 26.05.2020, 21:57 +0200 schrieb Jean Rene Dawin:
> Oliver Neukum wrote on Tue 26/05/20 13:13:
> > Hi,
> > 
> > may I ask whether you did the test with removing the battery twice with
> > an older kernel? Could you please go back to
> > f6cc6093a729ede1ff5658b493237c42b82ba107
> > and repeat the test of a second battery removal with that state?
> > I just cannot find anything pointing to a change that could cause
> > this issue within that time.
> 
> Hi,
> 
> testing with f6cc6093a729ede1ff5658b493237c42b82ba107 looks like this:

OK, we have two possibilities here. Either
a4e7279cd1d19f48f0af2a10ed020febaa9ac092 or
0afccd7601514c4b83d8cc58c740089cc447051d

have had a really wierd effect, or they introduced a bug
that hid a later bug. Can I ask you to run a complicated test
to decide between these possibilities?

Could you test a4e7279cd1d19f48f0af2a10ed020febaa9ac092
together with the patch I sent you applied on top?

> Interesting is, that at the second time, the usb disconnect message
> conmes first, and then a message from cdc_acm acm_port_activate.

Basically you are deliberately creating a race condition between error
handling and disconnect. That is a feature of the HC design.

	Regards
		Oliver