Skip to contentSkip to conversation

drm/amdgpu: fix structurally dead code vcn_v2_5_hw_init

PATCH In review Gustavo A. R. Silva 2 participants 7 messages Oct 1, 2019 – Oct 2, 2019

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/amdgpu: fix structurally dead code vcn_v2_5_hw_init

Notice that there is a *continue* statement in the middle of the for loop and that prevents the code below from ever being reached:

	r = amdgpu_ring_test_ring(ring);
	if (r) {
		ring->sched.ready = false;
		goto done;
	}

Fix this by removing the continue statement and updating ring->sched.ready to true before calling amdgpu_ring_test_ring(ring).

Notice that this fix is based on
commit 1b61de45dfaf ("drm/amdgpu: add initial VCN2.0 support (v2)")

Addresses-Coverity-ID 1485608 ("Structurally dead code")

Fixes 28c17d72072b ("drm/amdgpu: add VCN2.5 basic supports")

---

Any feedback is greatly appreciated.

 drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +3 -2
index 395c2259f979..47b0dcd59e13 100644
@@ -258,6 +258,7 @@ static int vcn_v2_5_hw_init(void *handle)
 		adev->nbio_funcs->vcn_doorbell_range(adev, ring->use_doorbell,
 						     ring->doorbell_index, j);
 
+		ring->sched.ready = true;
 		r = amdgpu_ring_test_ring(ring);
 		if (r) {
 			ring->sched.ready = false;
@@ -266,8 +267,7 @@ static int vcn_v2_5_hw_init(void *handle)
 
 		for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
 			ring = &adev->vcn.inst[j].ring_enc[i];
-			ring->sched.ready = false;
-			continue;
+			ring->sched.ready = true;
 			r = amdgpu_ring_test_ring(ring);
 			if (r) {
 				ring->sched.ready = false;
@@ -276,6 +276,7 @@ static int vcn_v2_5_hw_init(void *handle)
 		}
 
 		ring = &adev->vcn.inst[j].ring_jpeg;
+		ring->sched.ready = true;
 		r = amdgpu_ring_test_ring(ring);
 		if (r) {
 			ring->sched.ready = false;
Signature · 1 line
2.23.0
#1 lore raw
Quoted 30 lines · Gustavo A. R. Silva

Notice that there is a *continue* statement in the middle of the for loop and that prevents the code below from ever being reached:

	r = amdgpu_ring_test_ring(ring);
	if (r) {
		ring->sched.ready = false;
		goto done;
	}

Fix this by removing the continue statement and updating ring->sched.ready to true before calling amdgpu_ring_test_ring(ring).

Notice that this fix is based on
commit 1b61de45dfaf ("drm/amdgpu: add initial VCN2.0 support (v2)")

Addresses-Coverity-ID 1485608 ("Structurally dead code")

Fixes 28c17d72072b ("drm/amdgpu: add VCN2.5 basic supports")

---

Any feedback is greatly appreciated.

  drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c | 5 +++--
  1 file changed, 3 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +0 -0
index 395c2259f979..47b0dcd59e13 100644

drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c static int vcn_v2_5_hw_init(void *handle) · line 261

Quoted 4 lines
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -0
@@ -258,6 +258,7 @@ static int vcn_v2_5_hw_init(void *handle)
  		adev->nbio_funcs->vcn_doorbell_range(adev, ring->use_doorbell,
  						     ring->doorbell_index, j);
 
+		ring->sched.ready = true;

This is redundant. all the sched->ready is initialized as true, please refer to drm_sched_init()

drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c static int vcn_v2_5_hw_init(void *handle) · line 270

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -2
  		r = amdgpu_ring_test_ring(ring);
  		if (r) {
  			ring->sched.ready = false;
@@ -266,8 +267,7 @@ static int vcn_v2_5_hw_init(void *handle)
 
  		for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
  			ring = &adev->vcn.inst[j].ring_enc[i];
-			ring->sched.ready = false;
-			continue;
+			ring->sched.ready = true;

Because the VCN 2.5 FW still has issue for encode, so we have it disabled here.

Regards,

Leo

drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c static int vcn_v2_5_hw_init(void *handle) · line 282

Quoted 11 lines
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -0
  			r = amdgpu_ring_test_ring(ring);
  			if (r) {
  				ring->sched.ready = false;
@@ -276,6 +276,7 @@ static int vcn_v2_5_hw_init(void *handle)
  		}
 
  		ring = &adev->vcn.inst[j].ring_jpeg;
+		ring->sched.ready = true;
  		r = amdgpu_ring_test_ring(ring);
  		if (r) {
  			ring->sched.ready = false;

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#2 lore raw

Hi,

Quoted 35 lines · Leo
Quoted 33 lines · Gustavo A. R. Silva

Notice that there is a *continue* statement in the middle of the for loop and that prevents the code below from ever being reached:

	r = amdgpu_ring_test_ring(ring);
	if (r) {
		ring->sched.ready = false;
		goto done;
	}

Fix this by removing the continue statement and updating ring->sched.ready to true before calling amdgpu_ring_test_ring(ring).

Notice that this fix is based on
commit 1b61de45dfaf ("drm/amdgpu: add initial VCN2.0 support (v2)")

Addresses-Coverity-ID 1485608 ("Structurally dead code")

Fixes 28c17d72072b ("drm/amdgpu: add VCN2.5 basic supports")

---

Any feedback is greatly appreciated.

  drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c | 5 +++--
  1 file changed, 3 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +0 -0
index 395c2259f979..47b0dcd59e13 100644
@@ -258,6 +258,7 @@ static int vcn_v2_5_hw_init(void *handle)
  		adev->nbio_funcs->vcn_doorbell_range(adev, ring->use_doorbell,
  						     ring->doorbell_index, j);
 
Quoted 5 lines

drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c static int vcn_v2_5_hw_init(void *handle) · line 261

Quoted 1 line
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -0
+		ring->sched.ready = true;

This is redundant. all the sched->ready is initialized as true, please refer to drm_sched_init()

I see... so in the following commit 1b61de45dfaf ("drm/amdgpu: add initial VCN2.0 support (v2)") that line is also redundant?

Quoted 6 lines
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +0 -0
  		r = amdgpu_ring_test_ring(ring);
  		if (r) {
  			ring->sched.ready = false;
@@ -266,8 +267,7 @@ static int vcn_v2_5_hw_init(void *handle)
 
Quoted 9 lines

drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c static int vcn_v2_5_hw_init(void *handle) · line 270

Quoted 5 lines
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -2
  		for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
  			ring = &adev->vcn.inst[j].ring_enc[i];
-			ring->sched.ready = false;
-			continue;
+			ring->sched.ready = true;

Because the VCN 2.5 FW still has issue for encode, so we have it disabled here.

OK. So, maybe we can add a comment pointing that out?

Thanks
--
Gustavo

Quoted 12 lines
Quoted 12 lines
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -0
  			r = amdgpu_ring_test_ring(ring);
  			if (r) {
  				ring->sched.ready = false;
@@ -276,6 +276,7 @@ static int vcn_v2_5_hw_init(void *handle)
  		}
 
  		ring = &adev->vcn.inst[j].ring_jpeg;
+		ring->sched.ready = true;
  		r = amdgpu_ring_test_ring(ring);
  		if (r) {
  			ring->sched.ready = false;
#3 lore raw
Quoted 37 lines · Gustavo A. R. Silva

Hi,

Quoted 34 lines · Leo
Quoted 33 lines · Gustavo A. R. Silva

Notice that there is a *continue* statement in the middle of the for loop and that prevents the code below from ever being reached:

	r = amdgpu_ring_test_ring(ring);
	if (r) {
		ring->sched.ready = false;
		goto done;
	}

Fix this by removing the continue statement and updating ring->sched.ready to true before calling amdgpu_ring_test_ring(ring).

Notice that this fix is based on
commit 1b61de45dfaf ("drm/amdgpu: add initial VCN2.0 support (v2)")

Addresses-Coverity-ID 1485608 ("Structurally dead code")

Fixes 28c17d72072b ("drm/amdgpu: add VCN2.5 basic supports")

---

Any feedback is greatly appreciated.

   drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c | 5 +++--
   1 file changed, 3 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +0 -0
index 395c2259f979..47b0dcd59e13 100644
@@ -258,6 +258,7 @@ static int vcn_v2_5_hw_init(void *handle)
   		adev->nbio_funcs->vcn_doorbell_range(adev, ring->use_doorbell,
   						     ring->doorbell_index, j);
 
Quoted 6 lines
Quoted 4 lines

drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c static int vcn_v2_5_hw_init(void *handle) · line 261

Quoted 1 line
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -0
+		ring->sched.ready = true;

This is redundant. all the sched->ready is initialized as true, please refer to drm_sched_init()

I see... so in the following commit 1b61de45dfaf ("drm/amdgpu: add initial VCN2.0 support (v2)") that line is also redundant?

Yes.

Quoted 13 lines
Quoted 5 lines
Quoted 5 lines
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +0 -0
   		r = amdgpu_ring_test_ring(ring);
   		if (r) {
   			ring->sched.ready = false;
@@ -266,8 +267,7 @@ static int vcn_v2_5_hw_init(void *handle)
 
Quoted 8 lines

drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c static int vcn_v2_5_hw_init(void *handle) · line 270

Quoted 5 lines
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -2
   		for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
   			ring = &adev->vcn.inst[j].ring_enc[i];
-			ring->sched.ready = false;
-			continue;
+			ring->sched.ready = true;

Because the VCN 2.5 FW still has issue for encode, so we have it disabled here.

Quoted 1 line

OK. So, maybe we can add a comment pointing that out?

That could be better.

Thanks,
Leo

Quoted 11 lines

Thanks
--
Gustavo

Quoted 6 lines
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +0 -0
   			r = amdgpu_ring_test_ring(ring);
   			if (r) {
   				ring->sched.ready = false;
@@ -276,6 +276,7 @@ static int vcn_v2_5_hw_init(void *handle)
   		}
 
Quoted 5 lines
Quoted 5 lines

drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c static int vcn_v2_5_hw_init(void *handle) · line 282

Quoted 5 lines
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -0
   		ring = &adev->vcn.inst[j].ring_jpeg;
+		ring->sched.ready = true;
   		r = amdgpu_ring_test_ring(ring);
   		if (r) {
   			ring->sched.ready = false;

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#4 lore raw
Quoted 9 lines · Leo
Quoted 6 lines
Quoted 4 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+		ring->sched.ready = true;

This is redundant. all the sched->ready is initialized as true, please refer to drm_sched_init()

I see... so in the following commit 1b61de45dfaf ("drm/amdgpu: add initial VCN2.0 support (v2)") that line is also redundant?

Yes.

OK.

Quoted 14 lines
Quoted 14 lines
Quoted 14 lines

drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c static int vcn_v2_5_hw_init(void *handle) · line 270

Quoted 11 lines
   		r = amdgpu_ring_test_ring(ring);
   		if (r) {
   			ring->sched.ready = false;
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -2
@@ -266,8 +267,7 @@ static int vcn_v2_5_hw_init(void *handle)
 
   		for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
   			ring = &adev->vcn.inst[j].ring_enc[i];
-			ring->sched.ready = false;
-			continue;
+			ring->sched.ready = true;

Because the VCN 2.5 FW still has issue for encode, so we have it disabled here.

Quoted 4 lines
Quoted 1 line

OK. So, maybe we can add a comment pointing that out?

That could be better.

Great. I'm glad it's not a bug. I'll write a patch for that so other people don't waste time taking a look.

I appreciate your feedback.
Thanks
--
Gustavo
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#5 lore raw
Quoted 26 lines · Gustavo A. R. Silva
Quoted 8 lines · Leo
Quoted 6 lines
Quoted 4 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+		ring->sched.ready = true;

This is redundant. all the sched->ready is initialized as true, please refer to drm_sched_init()

I see... so in the following commit 1b61de45dfaf ("drm/amdgpu: add initial VCN2.0 support (v2)") that line is also redundant?

Yes.

OK.

Quoted 13 lines
Quoted 13 lines
Quoted 13 lines

drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c static int vcn_v2_5_hw_init(void *handle) · line 270

Quoted 10 lines
    		r = amdgpu_ring_test_ring(ring);
    		if (r) {
    			ring->sched.ready = false;
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -2
@@ -266,8 +267,7 @@ static int vcn_v2_5_hw_init(void *handle)
 
    		for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
    			ring = &adev->vcn.inst[j].ring_enc[i];
-			ring->sched.ready = false;
-			continue;
+			ring->sched.ready = true;

Because the VCN 2.5 FW still has issue for encode, so we have it disabled here.

Quoted 5 lines
Quoted 3 lines
Quoted 1 line

OK. So, maybe we can add a comment pointing that out?

That could be better.

Great. I'm glad it's not a bug. I'll write a patch for that so other people don't waste time taking a look.

Thanks, just sent two patches to add comment, and long with the patch to make VCN ring ready properly.

Leo

Quoted 5 lines

I appreciate your feedback.
Thanks
--
Gustavo

#6 lore raw
Quoted 9 lines · Leo
Quoted 5 lines
Quoted 3 lines
Quoted 1 line

OK. So, maybe we can add a comment pointing that out?

That could be better.

Great. I'm glad it's not a bug. I'll write a patch for that so other people don't waste time taking a look.

Thanks, just sent two patches to add comment, and long with the patch to make VCN ring ready properly.

Awesome. Thank you. I would just add a commit log to this patch:

[PATCH 2/2] drm/amdgpu: add a comment to VCN 2.5 encode ring

I'd update the subject to: drm/amdgpu: add code comment in vcn_v2_5_hw_init and add this as a commit log: Add a comment to VCN 2.5 encode ring

Also, I think it's important to follow the process and CC all the people and lists below:

$ scripts/get_maintainer.pl --nokeywords --nogit --nogit-fallback drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
Alex Deucher <[email protected]> (supporter:RADEON and AMDGPU DRM DRIVERS)
"Christian König" <[email protected]> (supporter:RADEON and AMDGPU DRM DRIVERS)
"David (ChunMing) Zhou" <[email protected]> (supporter:RADEON and AMDGPU DRM DRIVERS)
David Airlie <[email protected]> (maintainer:DRM DRIVERS)
Daniel Vetter <[email protected]> (maintainer:DRM DRIVERS)
[email protected] (open list:RADEON and AMDGPU DRM DRIVERS)
[email protected] (open list:DRM DRIVERS)
[email protected] (open list)

Thanks
--
Gustavo