Nouveau - Jul 2014 - [PATCH 09/17] drm/radeon: use common fence implementation for fences

Am 23.07.2014 11:55, schrieb Maarten Lankhorst:
> op 23-07-14 11:47, Christian K?nig schreef:
>> Am 23.07.2014 11:44, schrieb Daniel Vetter:
>>> On Wed, Jul 23, 2014 at 11:39 AM, Daniel Vetter <daniel.vetter
at ffwll.ch> wrote:
>>>> The scheduler needs to keep track of a lot of fences, so I
think we'll
>>>> have to register callbacks, not a simple wait function. We must
keep
>>>> track of all the non-i915 fences for all oustanding batches.
Also, the
>>>> scheduler doesn't eliminate the hw queue, only keep it much
slower so
>>>> that we can sneak in higher priority things.
>>>>
>>>> Really, scheduler or not is orthogonal.
>>> Also see my other comment about interactions between wait_fence and
>>> the i915 reset logic. We can't actually use it from within the
>>> scheduler code since that would deadlock.
>> Yeah, I see. You would need some way to abort the waiting on other
devices fences in case of a lockup.
>>
>> What about an userspace thread to offload waiting and command
submission to?
> You would still need enable_signaling, else polling on the dma-buf
wouldn't work. ;-)
> Can't wait synchronously with multiple shared fences, need to poll for
that.
No you don't. Just make a list of fences you need to wait for and wait 
for each after another. But having an thread for each command submission 
context doesn't sounds like the best solution anyway.
> And the dma-buf would still have fences belonging to both drivers, and it
would still call from outside the driver.
Calling from outside the driver is fine as long as the driver can do 
everything necessary to complete it's work and isn't forced into any 
ugly hacks and things that are not 100% reliable.

So I don't see much other approach as integrating recovery code for not 
firing interrupts and some kind of lockup handling into the fence code 
as well.

Christian.
>
> ~Maarten
>

On Wed, Jul 23, 2014 at 12:13 PM, Christian K?nig
<christian.koenig at amd.com> wrote:
>
>> And the dma-buf would still have fences belonging to both drivers, and
it
>> would still call from outside the driver.
>
>
> Calling from outside the driver is fine as long as the driver can do
> everything necessary to complete it's work and isn't forced into
any ugly
> hacks and things that are not 100% reliable.
>
> So I don't see much other approach as integrating recovery code for not
> firing interrupts and some kind of lockup handling into the fence code as
> well.
That approach doesn't really work at that well since every driver has
it's own reset semantics. And we're trying to move away from global
reset to fine-grained reset. So stop-the-world reset is out of
fashion, at least for i915. As you said, reset is normal in gpus and
we're trying to make reset less invasive. I really don't see a point
in imposing a reset scheme upon all drivers and I think you have about
as much motivation to convert radeon to the scheme used by i915 as
I'll have for converting to the one used by radeon. If it would fit at
all.

I guess for radeon we just have to add tons of insulation by punting
all callbacks to work items so that radeon can do whatever it wants.
Plus start a delayed_work based fallback when ->enable_signalling is
called to make sure we work on platforms that lack interrupts.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

Am 23.07.2014 12:52, schrieb Daniel Vetter:
> On Wed, Jul 23, 2014 at 12:13 PM, Christian K?nig
> <christian.koenig at amd.com> wrote:
>>> And the dma-buf would still have fences belonging to both drivers,
and it
>>> would still call from outside the driver.
>>
>> Calling from outside the driver is fine as long as the driver can do
>> everything necessary to complete it's work and isn't forced
into any ugly
>> hacks and things that are not 100% reliable.
>>
>> So I don't see much other approach as integrating recovery code for
not
>> firing interrupts and some kind of lockup handling into the fence code
as
>> well.
> That approach doesn't really work at that well since every driver has
> it's own reset semantics. And we're trying to move away from global
> reset to fine-grained reset. So stop-the-world reset is out of
> fashion, at least for i915. As you said, reset is normal in gpus and
> we're trying to make reset less invasive. I really don't see a
point
> in imposing a reset scheme upon all drivers and I think you have about
> as much motivation to convert radeon to the scheme used by i915 as
> I'll have for converting to the one used by radeon. If it would fit at
> all.
Oh my! No, I didn't wanted to suggest any global reset infrastructure.

My idea was more that the fence framework provides a 
fence->process_signaling callback that is periodically called after 
enable_signaling is called to trigger manual signal processing in the 
driver.

This would both be suitable as a fallback in case of not working 
interrupts as well as a chance for any driver to do necessary lockup 
handling.

Christian.
> I guess for radeon we just have to add tons of insulation by punting
> all callbacks to work items so that radeon can do whatever it wants.
> Plus start a delayed_work based fallback when ->enable_signalling is
> called to make sure we work on platforms that lack interrupts.
> -Daniel