linux-kernel - RE: [PATCH 5.11 079/104] drm/amdgpu: enable only one high prio compute queue

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Fri, 5 Mar 2021 15:27:00 +0000
From:   "Deucher, Alexander" <Alexander.Deucher@....com>
To:     "Koenig, Christian" <Christian.Koenig@....com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
CC:     "stable@...r.kernel.org" <stable@...r.kernel.org>,
        "Das, Nirmoy" <Nirmoy.Das@....com>, Sasha Levin <sashal@...nel.org>
Subject: RE: [PATCH 5.11 079/104] drm/amdgpu: enable only one high prio
 compute queue

[AMD Public Use]

> -----Original Message-----
> From: Koenig, Christian <Christian.Koenig@....com>
> Sent: Friday, March 5, 2021 10:18 AM
> To: Deucher, Alexander <Alexander.Deucher@....com>; Greg Kroah-
> Hartman <gregkh@...uxfoundation.org>; linux-kernel@...r.kernel.org
> Cc: stable@...r.kernel.org; Das, Nirmoy <Nirmoy.Das@....com>; Sasha
> Levin <sashal@...nel.org>
> Subject: Re: [PATCH 5.11 079/104] drm/amdgpu: enable only one high prio
> compute queue
> 
> Am 05.03.21 um 15:48 schrieb Deucher, Alexander:
> > [AMD Public Use]
> >
> >> -----Original Message-----
> >> From: Koenig, Christian <Christian.Koenig@....com>
> >> Sent: Friday, March 5, 2021 8:03 AM
> >> To: Greg Kroah-Hartman <gregkh@...uxfoundation.org>; linux-
> >> kernel@...r.kernel.org
> >> Cc: stable@...r.kernel.org; Das, Nirmoy <Nirmoy.Das@....com>;
> >> Deucher, Alexander <Alexander.Deucher@....com>; Sasha Levin
> >> <sashal@...nel.org>
> >> Subject: Re: [PATCH 5.11 079/104] drm/amdgpu: enable only one high
> >> prio compute queue
> >>
> >> Mhm, I'm not sure this one needs to be backported.
> >>
> >> Why did you pick it up Greg?
> > It was picked up by Sasha's fixes checker.
> 
> Well the change who needs this isn't in any earlier kernel, isn't it?

Not sure if Sasha picked that up or not.  Would need to check that.  If it's not, this patch should be dropped.

Alex

> 
> Christian.
> 
> >
> > Alex
> >
> >
> >> Thanks,
> >> Christian.
> >>
> >> Am 05.03.21 um 13:21 schrieb Greg Kroah-Hartman:
> >>> From: Nirmoy Das <nirmoy.das@....com>
> >>>
> >>> [ Upstream commit 8c0225d79273968a65e73a4204fba023ae02714d ]
> >>>
> >>> For high priority compute to work properly we need to enable wave
> >>> limiting on gfx pipe. Wave limiting is done through writing into
> >>> mmSPI_WCL_PIPE_PERCENT_GFX register. Enable only one high priority
> >>> compute queue to avoid race condition between multiple high priority
> >>> compute queues writing that register simultaneously.
> >>>
> >>> Signed-off-by: Nirmoy Das <nirmoy.das@....com>
> >>> Acked-by: Christian König <christian.koenig@....com>
> >>> Reviewed-by: Alex Deucher <alexander.deucher@....com>
> >>> Signed-off-by: Alex Deucher <alexander.deucher@....com>
> >>> Signed-off-by: Sasha Levin <sashal@...nel.org>
> >>> ---
> >>>    drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 15 ++++++++-------
> >>>    drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h |  2 +-
> >>>    drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c  |  6 ++----
> >>>    drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c   |  6 ++----
> >>>    drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c   |  7 ++-----
> >>>    5 files changed, 15 insertions(+), 21 deletions(-)
> >>>
> >>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> >>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> >>> index cd2c676a2797..8e0a6c62322e 100644
> >>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> >>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> >>> @@ -193,15 +193,16 @@ static bool
> >> amdgpu_gfx_is_multipipe_capable(struct amdgpu_device *adev)
> >>>    }
> >>>
> >>>    bool amdgpu_gfx_is_high_priority_compute_queue(struct
> >> amdgpu_device *adev,
> >>> -					       int pipe, int queue)
> >>> +					       struct amdgpu_ring *ring)
> >>>    {
> >>> -	bool multipipe_policy = amdgpu_gfx_is_multipipe_capable(adev);
> >>> -	int cond;
> >>> -	/* Policy: alternate between normal and high priority */
> >>> -	cond = multipipe_policy ? pipe : queue;
> >>> -
> >>> -	return ((cond % 2) != 0);
> >>> +	/* Policy: use 1st queue as high priority compute queue if we
> >>> +	 * have more than one compute queue.
> >>> +	 */
> >>> +	if (adev->gfx.num_compute_rings > 1 &&
> >>> +	    ring == &adev->gfx.compute_ring[0])
> >>> +		return true;
> >>>
> >>> +	return false;
> >>>    }
> >>>
> >>>    void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device
> >> *adev)
> >>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
> >>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
> >>> index 6b5a8f4642cc..72dbcd2bc6a6 100644
> >>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
> >>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
> >>> @@ -380,7 +380,7 @@ void
> >> amdgpu_queue_mask_bit_to_mec_queue(struct amdgpu_device *adev,
> int
> >> bit,
> >>>    bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device
> *adev,
> >> int mec,
> >>>    				     int pipe, int queue);
> >>>    bool amdgpu_gfx_is_high_priority_compute_queue(struct
> >> amdgpu_device *adev,
> >>> -					       int pipe, int queue);
> >>> +					       struct amdgpu_ring *ring);
> >>>    int amdgpu_gfx_me_queue_to_bit(struct amdgpu_device *adev, int
> me,
> >>>    			       int pipe, int queue);
> >>>    void amdgpu_gfx_bit_to_me_queue(struct amdgpu_device *adev, int
> >> bit,
> >>> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
> >>> b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
> >>> index e7d6da05011f..3a291befcddc 100644
> >>> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
> >>> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
> >>> @@ -4495,8 +4495,7 @@ static int gfx_v10_0_compute_ring_init(struct
> >> amdgpu_device *adev, int ring_id,
> >>>    	irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
> >>>    		+ ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec)
> >>>    		+ ring->pipe;
> >>> -	hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev,
> >> ring->pipe,
> >>> -							    ring->queue) ?
> >>> +	hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev,
> >> ring) ?
> >>>    			AMDGPU_GFX_PIPE_PRIO_HIGH :
> >> AMDGPU_GFX_PIPE_PRIO_NORMAL;
> >>>    	/* type-2 packets are deprecated on MEC, use type-3 instead */
> >>>    	r = amdgpu_ring_init(adev, ring, 1024, @@ -6545,8 +6544,7 @@
> >>> static void gfx_v10_0_compute_mqd_set_priority(struct amdgpu_ring
> >>> *ring,
> >> struct
> >>>    	struct amdgpu_device *adev = ring->adev;
> >>>
> >>>    	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) {
> >>> -		if (amdgpu_gfx_is_high_priority_compute_queue(adev,
> >> ring->pipe,
> >>> -							      ring->queue)) {
> >>> +		if (amdgpu_gfx_is_high_priority_compute_queue(adev,
> >> ring)) {
> >>>    			mqd->cp_hqd_pipe_priority =
> >> AMDGPU_GFX_PIPE_PRIO_HIGH;
> >>>    			mqd->cp_hqd_queue_priority =
> >>>
> >> 	AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM;
> >>> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
> >>> b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
> >>> index 37639214cbbb..b0284c4659ba 100644
> >>> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
> >>> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
> >>> @@ -1923,8 +1923,7 @@ static int gfx_v8_0_compute_ring_init(struct
> >> amdgpu_device *adev, int ring_id,
> >>>    		+ ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec)
> >>>    		+ ring->pipe;
> >>>
> >>> -	hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev,
> >> ring->pipe,
> >>> -							    ring->queue) ?
> >>> +	hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev,
> >> ring) ?
> >>>    			AMDGPU_GFX_PIPE_PRIO_HIGH :
> >> AMDGPU_RING_PRIO_DEFAULT;
> >>>    	/* type-2 packets are deprecated on MEC, use type-3 instead */
> >>>    	r = amdgpu_ring_init(adev, ring, 1024, @@ -4442,8 +4441,7 @@
> >>> static void gfx_v8_0_mqd_set_priority(struct amdgpu_ring *ring,
> >>> struct vi_mqd
> >> *m
> >>>    	struct amdgpu_device *adev = ring->adev;
> >>>
> >>>    	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) {
> >>> -		if (amdgpu_gfx_is_high_priority_compute_queue(adev,
> >> ring->pipe,
> >>> -							      ring->queue)) {
> >>> +		if (amdgpu_gfx_is_high_priority_compute_queue(adev,
> >> ring)) {
> >>>    			mqd->cp_hqd_pipe_priority =
> >> AMDGPU_GFX_PIPE_PRIO_HIGH;
> >>>    			mqd->cp_hqd_queue_priority =
> >>>
> >> 	AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM;
> >>> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> >>> b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> >>> index 5f4805e4d04a..3e800193a604 100644
> >>> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> >>> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> >>> @@ -2228,8 +2228,7 @@ static int gfx_v9_0_compute_ring_init(struct
> >> amdgpu_device *adev, int ring_id,
> >>>    	irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
> >>>    		+ ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec)
> >>>    		+ ring->pipe;
> >>> -	hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev,
> >> ring->pipe,
> >>> -							    ring->queue) ?
> >>> +	hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev,
> >> ring) ?
> >>>    			AMDGPU_GFX_PIPE_PRIO_HIGH :
> >> AMDGPU_GFX_PIPE_PRIO_NORMAL;
> >>>    	/* type-2 packets are deprecated on MEC, use type-3 instead */
> >>>    	return amdgpu_ring_init(adev, ring, 1024, @@ -3391,9 +3390,7 @@
> >>> static void gfx_v9_0_mqd_set_priority(struct amdgpu_ring *ring,
> >>> struct
> >> v9_mqd *m
> >>>    	struct amdgpu_device *adev = ring->adev;
> >>>
> >>>    	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) {
> >>> -		if (amdgpu_gfx_is_high_priority_compute_queue(adev,
> >>> -							      ring->pipe,
> >>> -							      ring->queue)) {
> >>> +		if (amdgpu_gfx_is_high_priority_compute_queue(adev,
> >> ring)) {
> >>>    			mqd->cp_hqd_pipe_priority =
> >> AMDGPU_GFX_PIPE_PRIO_HIGH;
> >>>    			mqd->cp_hqd_queue_priority =
> >>>
> >> 	AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM;