AMD-GFX Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: "Christian König" <ckoenig.leichtzumerken-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
To: "Liu, Leo" <Leo.Liu-5C7GfCeVMHo@public.gmane.org>,
	"Feng, Kenneth" <Kenneth.Feng-5C7GfCeVMHo@public.gmane.org>,
	"amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org"
	<amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org>
Subject: Re: [PATCH] drm/amd/amdgpu: workaround for UVD busy status
Date: Wed, 14 Aug 2019 19:20:59 +0200	[thread overview]
Message-ID: <d7dcc4e1-57b2-dc75-5a29-ddf4e12d1d6d@gmail.com> (raw)
In-Reply-To: <d47e5780-afda-98a4-e95f-ef1ede95dd95-5C7GfCeVMHo@public.gmane.org>

Am 14.08.19 um 18:58 schrieb Liu, Leo:
> On 2019-08-14 7:11 a.m., Kenneth Feng wrote:
>> On Vega20, tools depends on UVD_STATUS.VCPU_REPORT bit0
>> to decide if UVD instances are in busy state or idle state.
>> This workaround fixes the issue that tools always fetch the
>> UVD instances state as busy state no matter if there is a UVD work.
> The VCPU_REPORT is really for FW to report status to driver, not the
> other way around. Sounds to me this is a FW bug.

Agreed, but I wanted to double check the underlying background.

Looks like this is just for internal testing and shouldn't be upstreamed,
Christian.

>
> Regards,
>
> Leo
>
>
>> Signed-off-by: Kenneth Feng <kenneth.feng@amd.com>
>> ---
>>    drivers/gpu/drm/amd/amdgpu/amdgpu.h        |  3 +++
>>    drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 23 +++++++++++++++++++++++
>>    drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c    | 30 ++++++++++++++++++++++++++++--
>>    drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c      | 19 +++++++++++++++++++
>>    drivers/gpu/drm/amd/include/amd_shared.h   |  1 +
>>    5 files changed, 74 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
>> index 4d096ff..6e5a41b 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
>> @@ -259,6 +259,9 @@ int amdgpu_device_ip_set_clockgating_state(void *dev,
>>    int amdgpu_device_ip_set_powergating_state(void *dev,
>>    					   enum amd_ip_block_type block_type,
>>    					   enum amd_powergating_state state);
>> +int amdgpu_device_ip_set_instance_state(void *dev,
>> +					   enum amd_ip_block_type block_type,
>> +					   bool busy_state, int inst);
>>    void amdgpu_device_ip_get_clockgating_state(struct amdgpu_device *adev,
>>    					    u32 *flags);
>>    int amdgpu_device_ip_wait_for_idle(struct amdgpu_device *adev,
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> index 93ed3cb..e65e251 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> @@ -1127,6 +1127,29 @@ int amdgpu_device_ip_set_powergating_state(void *dev,
>>    	return r;
>>    }
>>    
>> +int amdgpu_device_ip_set_instance_state(void *dev,
>> +					   enum amd_ip_block_type block_type,
>> +					   bool busy_state, int inst)
>> +{
>> +	struct amdgpu_device *adev = dev;
>> +	int i, r = 0;
>> +
>> +	for (i = 0; i < adev->num_ip_blocks; i++) {
>> +		if (!adev->ip_blocks[i].status.valid)
>> +			continue;
>> +		if (adev->ip_blocks[i].version->type != block_type)
>> +			continue;
>> +		if (!adev->ip_blocks[i].version->funcs->set_instance_state)
>> +			continue;
>> +		r = adev->ip_blocks[i].version->funcs->set_instance_state(
>> +			(void *)adev, busy_state, inst);
>> +		if (r)
>> +			DRM_ERROR("set_instance_state of IP block <%s> failed %d\n",
>> +				  adev->ip_blocks[i].version->funcs->name, r);
>> +	}
>> +	return r;
>> +}
>> +
>>    /**
>>     * amdgpu_device_ip_get_clockgating_state - get the CG state
>>     *
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
>> index 4e5d13e4..d3496ab 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
>> @@ -1180,13 +1180,22 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work)
>>    	struct amdgpu_device *adev =
>>    		container_of(work, struct amdgpu_device, uvd.idle_work.work);
>>    	unsigned fences = 0, i, j;
>> +	unsigned *fences_inst = (unsigned *)kzalloc(adev->uvd.num_uvd_inst*sizeof(unsigned), GFP_KERNEL);
>>    
>>    	for (i = 0; i < adev->uvd.num_uvd_inst; ++i) {
>>    		if (adev->uvd.harvest_config & (1 << i))
>>    			continue;
>> -		fences += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring);
>> +		fences_inst[i] = amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring);
>> +		fences += fences_inst[i];
>>    		for (j = 0; j < adev->uvd.num_enc_rings; ++j) {
>> -			fences += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring_enc[j]);
>> +			fences_inst[i] += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring_enc[j]);
>> +			fences += fences_inst[i];
>> +		}
>> +		/* workaround for the tools to show UVD busy status */
>> +		if (fences_inst[i] == 0) {
>> +			amdgpu_device_ip_set_instance_state(adev, AMD_IP_BLOCK_TYPE_UVD,
>> +							       false, i);
>> +			printk("Kenneth - set instance idle!\n");
>>    		}
>>    	}
>>    
>> @@ -1210,10 +1219,27 @@ void amdgpu_uvd_ring_begin_use(struct amdgpu_ring *ring)
>>    {
>>    	struct amdgpu_device *adev = ring->adev;
>>    	bool set_clocks;
>> +	unsigned i,j;
>> +	unsigned *fences_inst = (unsigned *)kzalloc(adev->uvd.num_uvd_inst*sizeof(unsigned), GFP_KERNEL);
>>    
>>    	if (amdgpu_sriov_vf(adev))
>>    		return;
>>    
>> +	for (i = 0; i < adev->uvd.num_uvd_inst; ++i) {
>> +		printk("Kenneth - run after set_clock!\n");
>> +		if (adev->uvd.harvest_config & (1 << i))
>> +			continue;
>> +		fences_inst[i] = amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring);
>> +		for (j = 0; j < adev->uvd.num_enc_rings; ++j)
>> +			fences_inst[i] += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring_enc[j]);
>> +		/* workaround for the tools to show UVD busy */
>> +		if (fences_inst[i] != 0) {
>> +			amdgpu_device_ip_set_instance_state(adev, AMD_IP_BLOCK_TYPE_UVD,
>> +						       true, i);
>> +			printk("Kenneth - run in setting instance state!\n");
>> +		}
>> +	}
>> +
>>    	set_clocks = !cancel_delayed_work_sync(&adev->uvd.idle_work);
>>    	if (set_clocks) {
>>    		if (adev->pm.dpm_enabled) {
>> diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
>> index 2f3d4e8..bf0f33a 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
>> @@ -1738,6 +1738,24 @@ static int uvd_v7_0_set_clockgating_state(void *handle,
>>    	return 0;
>>    }
>>    
>> +/* workaround for tools to fetch the UVD busy status */
>> +static int uvd_v7_0_set_instance_state(void *handle, bool busy_state, int inst)
>> +{
>> +	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
>> +
>> +	if(inst >= adev->uvd.num_uvd_inst)
>> +		return -EINVAL;
>> +
>> +	if(busy_state)
>> +		WREG32_P(SOC15_REG_OFFSET(UVD, inst, mmUVD_STATUS), (1 << UVD_STATUS__VCPU_REPORT__SHIFT),
>> +				~(1 << UVD_STATUS__VCPU_REPORT__SHIFT));
>> +	else
>> +		WREG32_P(SOC15_REG_OFFSET(UVD, inst, mmUVD_STATUS), 0,
>> +				~(1 << UVD_STATUS__VCPU_REPORT__SHIFT));
>> +
>> +	return 0;
>> +}
>> +
>>    const struct amd_ip_funcs uvd_v7_0_ip_funcs = {
>>    	.name = "uvd_v7_0",
>>    	.early_init = uvd_v7_0_early_init,
>> @@ -1756,6 +1774,7 @@ const struct amd_ip_funcs uvd_v7_0_ip_funcs = {
>>    	.post_soft_reset = NULL /* uvd_v7_0_post_soft_reset */,
>>    	.set_clockgating_state = uvd_v7_0_set_clockgating_state,
>>    	.set_powergating_state = NULL /* uvd_v7_0_set_powergating_state */,
>> +	.set_instance_state = uvd_v7_0_set_instance_state,
>>    };
>>    
>>    static const struct amdgpu_ring_funcs uvd_v7_0_ring_vm_funcs = {
>> diff --git a/drivers/gpu/drm/amd/include/amd_shared.h b/drivers/gpu/drm/amd/include/amd_shared.h
>> index 779c9e7..8cc54c9 100644
>> --- a/drivers/gpu/drm/amd/include/amd_shared.h
>> +++ b/drivers/gpu/drm/amd/include/amd_shared.h
>> @@ -198,6 +198,7 @@ struct amd_ip_funcs {
>>    	void (*get_clockgating_state)(void *handle, u32 *flags);
>>    	/** @enable_umd_pstate: enable UMD powerstate */
>>    	int (*enable_umd_pstate)(void *handle, enum amd_dpm_forced_level *level);
>> +	int (*set_instance_state)(void *handle, bool busy_state, int inst);
>>    };
>>    
>>    
> _______________________________________________
> amd-gfx mailing list
> amd-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

      parent reply	other threads:[~2019-08-14 17:20 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-08-14 11:11 [PATCH] drm/amd/amdgpu: workaround for UVD busy status Kenneth Feng
     [not found] ` <1565781071-5919-1-git-send-email-kenneth.feng-5C7GfCeVMHo@public.gmane.org>
2019-08-14 11:15   ` Christian König
     [not found]     ` <aa7f826f-17b3-26da-90e2-b52f0e6263a3-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2019-08-14 11:59       ` Feng, Kenneth
2019-08-14 16:58   ` Liu, Leo
     [not found]     ` <d47e5780-afda-98a4-e95f-ef1ede95dd95-5C7GfCeVMHo@public.gmane.org>
2019-08-14 17:20       ` Christian König [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=d7dcc4e1-57b2-dc75-5a29-ddf4e12d1d6d@gmail.com \
    --to=ckoenig.leichtzumerken-re5jqeeqqe8avxtiumwx3w@public.gmane.org \
    --cc=Kenneth.Feng-5C7GfCeVMHo@public.gmane.org \
    --cc=Leo.Liu-5C7GfCeVMHo@public.gmane.org \
    --cc=amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org \
    --cc=christian.koenig-5C7GfCeVMHo@public.gmane.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox