Re: [Intel-gfx] [PATCH 2/2] drm/i915/guc: Don't deadlock busyness stats vs reset

[John Harrison]

On 10/31/2022 05:51, Tvrtko Ursulin wrote:
On 31/10/2022 10:09, Tvrtko Ursulin wrote:
On 28/10/2022 20:46, john.c.harri...@intel.com wrote:
From: John Harrison <john.c.harri...@intel.com>
The engine busyness stats has a worker function to do things like
64bit extend the 32bit hardware counters. The GuC's reset prepare
function flushes out this worker function to ensure no corruption
happens during the reset. Unforunately, the worker function has an
infinite wait for active resets to finish before doing its work. Thus
a deadlock would occur if the worker function had actually started
just as the reset starts.

Update the worker to abort if a reset is in progress rather than
waiting for it to complete. It will still acquire the reset lock in
the case where a reset was not already in progress. So the processing
is still safe from corruption, but the deadlock can no longer occur.

Signed-off-by: John Harrison <john.c.harri...@intel.com>
---
  drivers/gpu/drm/i915/gt/intel_reset.c             | 15 
++++++++++++++-
  drivers/gpu/drm/i915/gt/intel_reset.h             |  1 +
  drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c |  6 ++++--
  3 files changed, 19 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c 
b/drivers/gpu/drm/i915/gt/intel_reset.c
index 3159df6cdd492..2f48c6e4420ea 100644
--- a/drivers/gpu/drm/i915/gt/intel_reset.c
+++ b/drivers/gpu/drm/i915/gt/intel_reset.c
@@ -1407,7 +1407,7 @@ void intel_gt_handle_error(struct intel_gt *gt,
      intel_runtime_pm_put(gt->uncore->rpm, wakeref);
  }
-int intel_gt_reset_trylock(struct intel_gt *gt, int *srcu)
+static int _intel_gt_reset_trylock(struct intel_gt *gt, int *srcu, 
bool retry)
  {
      might_lock(&gt->reset.backoff_srcu);
      might_sleep();
@@ -1416,6 +1416,9 @@ int intel_gt_reset_trylock(struct intel_gt 
*gt, int *srcu)
      while (test_bit(I915_RESET_BACKOFF, &gt->reset.flags)) {
          rcu_read_unlock();
+        if (!retry)
+            return -EBUSY;
+
          if (wait_event_interruptible(gt->reset.queue,
                           !test_bit(I915_RESET_BACKOFF,
                                 &gt->reset.flags)))

Would it be more obvious to rename the existing semantics to 
intel_gt_reset_interruptible(), while the flavour you add in this 
patch truly is trylock? I am not sure, since it's all a bit special, 
but trylock sure feels confusing if it can sleep forever...
To me, it would seem totally more obvious to have a function called 
'trylock' not wait forever until it can manage to acquire the lock. 
However, according to '2caffbf1176256 drm/i915: Revoke mmaps and prevent 
access to fence registers across reset', the current behaviour is 
exactly how the code was originally written and intended. It hasn't just 
mutated into some confused evolution a thousand patches later. So I 
figure there is some subtle but important reason why it was named how it 
is named and yet does what it does. Therefore it seemed safest to not 
change it unnecessarily.


Oh and might_sleep() shouldn't be there with the trylock version - I 
mean any flavour of the real trylock.
You mean if the code is split into two completely separate functions? Or 
do you just mean to wrap the might_sleep() call with 'if(!retry)'?

And just to be totally clear, the unconditional call to rcu_read_lock() 
is not something that can sleep? One doesn't need a might_sleep() before 
doing that lock?

John.



Regards,

Tvrtko