i8042 resume timing

i8042 resume timing

[Jiri Kosina]

Hi Dmitry,

There are systems [1] that fail in i8042_resume(), as 
i8042_command(&i8042_ctr, I8042_CMD_CTL_WCTR) call returns error (not sure 
about the exact error yet, the reporter didn't yet collect i8042.debug 
output during the failure, as it is not completely reproducible), but 
retrying after small amount of time succeeds.

Looks like a buggy hardware to me, which we should workaround. Would you 
prefer DMI-based match for this retry logic with respect to CTR register 
on the resume path, or would you accept this as a general retry logic on 
resume path? (which is very probably not needed on vast majority of 
systems though).

Thanks.

[1] https://bugzilla.novell.com/show_bug.cgi?id=351119

-- 
Jiri Kosina
SUSE Labs

Re: i8042 resume timing

[Dmitry Torokhov]

Hi Jiri,

On Monday 26 May 2008 09:19, Jiri Kosina wrote:
> Hi Dmitry,
> 
> There are systems [1] that fail in i8042_resume(), as 
> i8042_command(&i8042_ctr, I8042_CMD_CTL_WCTR) call returns error (not sure 
> about the exact error yet, the reporter didn't yet collect i8042.debug 
> output during the failure, as it is not completely reproducible), but 
> retrying after small amount of time succeeds.
> 
> Looks like a buggy hardware to me, which we should workaround. Would you 
> prefer DMI-based match for this retry logic with respect to CTR register 
> on the resume path, or would you accept this as a general retry logic on 
> resume path? (which is very probably not needed on vast majority of 
> systems though).


I think just retrying a couple of times in case of failure is the most
sensible thing to do.

Thanks.

-- 
Dmitry