Re: context_loss_count error value

[Kevin Hilman <khilman@ti.com>]

Tomi Valkeinen <tomi.valkeinen@ti.com> writes:

> On Wed, 2011-05-18 at 12:50 +0200, Kevin Hilman wrote:
>> Tomi Valkeinen <tomi.valkeinen@ti.com> writes:
>> 
>> > Hi Kevin,
>> >
>> > I was fixing DSS context loss handling which is a bit broken, and while
>> > testing on OMAP3 Overo, with -rc7 and omap2plus_defconfig, I noticed
>> > that get_context_loss_count() seems to always return 0.
>> >
>> > 0 should be returned when an error happens, and as far as I see in
>> > pwrdm_get_context_loss_count(), no error is happening but the DSS
>> > context has just never been lost and the returned count is thus 0.
>> >
>> > Is this correct? And what happens when the count wraps and goes back to
>> > zero, does the function return 0 in that case?
>> 
>> Hmm, you're right.  zero is actually documented as the error return
>> value (even though it's not really checked.)
>> 
>> Since driver's should only every care about the *difference* in value
>> between two calls to context_loss_count(), this might not be a big deal,
>> but a proper fix is probably to have the state counters start at one.
>
> But if there happens an error in get_context_loss_count(), for whatever
> reason, I'd guess it's safer from the driver's perspective to assume
> that a context restore _is_ needed. If the driver handles zero value as
> a normal return value, it would mean that the driver never restores
> context if get_context_loss_count() returns 0 for all calls.

Looking closer at the code, a zero return happens only when

1) no hwmod associated to omap_device
2) no power domain associated to hwmod
3) power domain has not (yet) lost context

None of these are actually error conditions per-se, and in all cases, it
indidates that context has not been lost (or we can't tell if context
has been lost.) 

So I think the current code is correct.

Are you finding a case where HW context has actually been lost and the
powerdomain context loss is still at zero?

Kevin