[Intel-gfx] [PATCH] drm/i915: Limit the backpressure for i915_request allocation

[Chris Wilson]

Quoting Tvrtko Ursulin (2018-09-12 14:34:16)
> 
> On 12/09/2018 12:11, Chris Wilson wrote:
> > If we try and fail to allocate a i915_request, we apply some
> > backpressure on the clients to throttle the memory allocations coming
> > from i915.ko. Currently, we wait until completely idle, but this is far
> > too heavy and leads to some situations where the only escape is to
> > declare a client hung and reset the GPU. The intent is to only ratelimit
> > the allocation requests, so we need only wait for a jiffie before using
> > the normal direct reclaim.
> > 
> > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=106680
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> > Cc: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
> > ---
> >   drivers/gpu/drm/i915/i915_request.c | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
> > index 09ed48833b54..588bc5a4d18b 100644
> > --- a/drivers/gpu/drm/i915/i915_request.c
> > +++ b/drivers/gpu/drm/i915/i915_request.c
> > @@ -736,7 +736,7 @@ i915_request_alloc(struct intel_engine_cs *engine, struct i915_gem_context *ctx)
> >               ret = i915_gem_wait_for_idle(i915,
> >                                            I915_WAIT_LOCKED |
> >                                            I915_WAIT_INTERRUPTIBLE,
> > -                                          MAX_SCHEDULE_TIMEOUT);
> > +                                          1);
> >               if (ret)
> >                       goto err_unreserve;
> >   
> > 
> 
> What is the remaining value of even trying to wait for idle, instead of 
> maybe just i915_request_retire and sleep for a jiffie? The intention 
> would potentially read clearer since it is questionable there is any 
> relationship with idle and rate limiting clients. In fact, now that I 
> think of it, waiting for idle is a nice way to starve an unlucky client 
> forever.

Better to starve the unlucky client, than to allow the entire system to
grind to a halt.

One caveat to using RCU is that it is our responsibility to apply
backpressure as none is applied by the vm.
-Chris