Re: [Intel-gfx] [PATCH] drm/i915/gt: Mark up intel_rps.active for racy reads

[Mika Kuoppala <mika.kuoppala@linux.intel.com>]

Chris Wilson <chris@chris-wilson.co.uk> writes:

> We read the current state of intel_rps.active outside of the lock, so
> mark up the racy access.
>
> [  525.037073] BUG: KCSAN: data-race in intel_rps_boost [i915] / intel_rps_park [i915]
> [  525.037091]
> [  525.037103] write to 0xffff8881f145efa1 of 1 bytes by task 192 on cpu 2:
> [  525.037331]  intel_rps_park+0x72/0x230 [i915]
> [  525.037552]  __gt_park+0x61/0xa0 [i915]
> [  525.037771]  ____intel_wakeref_put_last+0x42/0x90 [i915]
> [  525.037991]  __intel_wakeref_put_work+0xd3/0xf0 [i915]
> [  525.038008]  process_one_work+0x3b1/0x690
> [  525.038022]  worker_thread+0x80/0x670
> [  525.038037]  kthread+0x19a/0x1e0
> [  525.038051]  ret_from_fork+0x1f/0x30
> [  525.038062]
> [  525.038074] read to 0xffff8881f145efa1 of 1 bytes by task 733 on cpu 3:
> [  525.038304]  intel_rps_boost+0x67/0x1f0 [i915]
> [  525.038535]  i915_request_wait+0x562/0x5d0 [i915]
> [  525.038764]  i915_gem_object_wait_fence+0x81/0xa0 [i915]
> [  525.038994]  i915_gem_object_wait_reservation+0x489/0x520 [i915]
> [  525.039224]  i915_gem_wait_ioctl+0x167/0x2b0 [i915]
> [  525.039241]  drm_ioctl_kernel+0xe4/0x120
> [  525.039255]  drm_ioctl+0x297/0x4c7
> [  525.039269]  ksys_ioctl+0x89/0xb0
> [  525.039282]  __x64_sys_ioctl+0x42/0x60
> [  525.039296]  do_syscall_64+0x6e/0x2c0
> [  525.039311]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>

> ---
>  drivers/gpu/drm/i915/gt/intel_rps.c | 12 ++++++++----
>  1 file changed, 8 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_rps.c b/drivers/gpu/drm/i915/gt/intel_rps.c
> index 8794d399f5cd..506738dede16 100644
> --- a/drivers/gpu/drm/i915/gt/intel_rps.c
> +++ b/drivers/gpu/drm/i915/gt/intel_rps.c
> @@ -644,7 +644,7 @@ void intel_rps_mark_interactive(struct intel_rps *rps, bool interactive)
>  {
>  	mutex_lock(&rps->power.mutex);
>  	if (interactive) {
> -		if (!rps->power.interactive++ && rps->active)
> +		if (!rps->power.interactive++ && READ_ONCE(rps->active))
>  			rps_set_power(rps, HIGH_POWER);
>  	} else {
>  		GEM_BUG_ON(!rps->power.interactive);
> @@ -721,11 +721,15 @@ void intel_rps_unpark(struct intel_rps *rps)
>  	 * performance, jump directly to RPe as our starting frequency.
>  	 */
>  	mutex_lock(&rps->lock);
> -	rps->active = true;
> +
> +	WRITE_ONCE(rps->active, true);
> +
>  	freq = max(rps->cur_freq, rps->efficient_freq),
>  	freq = clamp(freq, rps->min_freq_softlimit, rps->max_freq_softlimit);
>  	intel_rps_set(rps, freq);
> +
>  	rps->last_adj = 0;
> +
>  	mutex_unlock(&rps->lock);
>  
>  	if (INTEL_GEN(rps_to_i915(rps)) >= 6)
> @@ -745,7 +749,7 @@ void intel_rps_park(struct intel_rps *rps)
>  	if (INTEL_GEN(i915) >= 6)
>  		rps_disable_interrupts(rps);
>  
> -	rps->active = false;
> +	WRITE_ONCE(rps->active, false);
>  	if (rps->last_freq <= rps->idle_freq)
>  		return;
>  
> @@ -772,7 +776,7 @@ void intel_rps_boost(struct i915_request *rq)
>  	struct intel_rps *rps = &rq->engine->gt->rps;
>  	unsigned long flags;
>  
> -	if (i915_request_signaled(rq) || !rps->active)
> +	if (i915_request_signaled(rq) || !READ_ONCE(rps->active))
>  		return;
>  
>  	/* Serializes with i915_request_retire() */
> -- 
> 2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx