On 18.10.2023 20:00, Julien Grall wrote:
> On 18/10/2023 15:58, Jan Beulich wrote:
>> From: Manuel Bouyer <[email protected]>
>>
>> The use of rcu_lock_domain_by_id() right in switch_serial_input() makes
>> assumptions about domain IDs which don't hold when in shim mode: The
>> sole (initial) domain there has a non-zero ID. Obtain the real domain ID
>> in that case (generalized as get_initial_domain_id() returns zero when
>> not in shim mode).
>>
>> Note that console_input_domain() isn't altered, for not being used when
>> in shim mode (or more generally on x86).
>
> I think it would be worth to either add a comment in
> console_input_domain() and/or #ifdef the code. In any case...
I have such a patch already, but intend to submit only for post-4.18.
>> Fixes: c2581c58bec9 ("xen/console: skip switching serial input to non
>> existing domains")
>> Signed-off-by: Manuel Bouyer <[email protected]>
>> Signed-off-by: Jan Beulich <[email protected]>
>
> ...
>
> Reviewed-by: Julien Grall <[email protected]>
Thanks.
> Also, should we consider it for xen 4.18? (I notice there is no for-4.18
> tag).
Of course we should, as it's a regression. I probably should have added
the tag, despite my dislike for such. Intention was imo nevertheless
clear, by me having Cc-ed Henry.
Jan
