Skip to contentSkip to conversation

drm/amdgpu: use PCI_IRQ_MSI_TYPES where appropriate

PATCH Applied Piotr Stankiewicz 4 participants 9 messages Jun 2, 2020

Subsystems (2)

v1 of 4 Applied Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH 7/15 drm/amdgpu: use PCI_IRQ_MSI_TYPES where appropriate

Seeing as there is shorthand available to use when asking for any type of interrupt, or any type of message signalled interrupt, leverage it.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +4 -4
index 5ed4227f304b..6dbe173a9fd4 100644
@@ -251,11 +251,11 @@ int amdgpu_irq_init(struct amdgpu_device *adev)
 		int nvec = pci_msix_vec_count(adev->pdev);
 		unsigned int flags;
 
-		if (nvec <= 0) {
+		if (nvec > 0)
+			flags = PCI_IRQ_MSI_TYPES;
+		else
 			flags = PCI_IRQ_MSI;
-		} else {
-			flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
-		}
+
 		/* we only need one vector */
 		nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1, flags);
 		if (nvec > 0) {
Signature · 1 line
2.17.2
#1 lore raw
Quoted 19 lines · Piotr Stankiewicz

Seeing as there is shorthand available to use when asking for any type of interrupt, or any type of message signalled interrupt, leverage it.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +0 -1
index 5ed4227f304b..6dbe173a9fd4 100644
@@ -251,11 +251,11 @@ int amdgpu_irq_init(struct amdgpu_device *adev)
                int nvec = pci_msix_vec_count(adev->pdev);
                unsigned int flags;
 
-               if (nvec <= 0) {

drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c int amdgpu_irq_init(struct amdgpu_device *adev) · line 260

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +4 -3
+               if (nvec > 0)
+                       flags = PCI_IRQ_MSI_TYPES;
+               else
                        flags = PCI_IRQ_MSI;
-               } else {
-                       flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
-               }
+
                /* we only need one vector */
                nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1, flags);

I'm not sure if you have seen my last comment internally about this patch.

I don't understand why we need these pci_msix_vec_count() followed by conditional at all.
Perhaps we may simple drop all these and supply flag directly?

But OTOH, I don't know the initial motivation, so, the above patch is non-intrusive and keeps original logic.

Quoted 4 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +0 -1
                if (nvec > 0) {
--

2.17.2

Signature · 6 lines
With Best Regards,
Andy Shevchenko
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#2 lore raw
Quoted 47 lines

-----Original Message-----
From: Andy Shevchenko <[email protected]>
Sent: Tuesday, June 2, 2020 11:49 AM
To: Stankiewicz, Piotr <[email protected]>
Cc: Alex Deucher <[email protected]>; Christian König <[email protected]>; David Zhou <[email protected]>; David Airlie <[email protected]>; Daniel Vetter <[email protected]>; amd- [email protected]; dri-devel <[email protected]>; Linux Kernel Mailing List <[email protected]>
Subject: Re: [PATCH 07/15] drm/amdgpu: use PCI_IRQ_MSI_TYPES where appropriate

Review

Quoted 11 lines · Piotr Stankiewicz

Seeing as there is shorthand available to use when asking for any type of interrupt, or any type of message signalled interrupt, leverage it.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
(quoted diff) +0 -0

b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c

Quoted 8 lines

index 5ed4227f304b..6dbe173a9fd4 100644

drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +0 -1
@@ -251,11 +251,11 @@ int amdgpu_irq_init(struct amdgpu_device *adev)
                int nvec = pci_msix_vec_count(adev->pdev);
                unsigned int flags;
 
-               if (nvec <= 0) {

drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c int amdgpu_irq_init(struct amdgpu_device *adev) · line 260

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +4 -3
+               if (nvec > 0)
+                       flags = PCI_IRQ_MSI_TYPES;
+               else
                        flags = PCI_IRQ_MSI;
-               } else {
-                       flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
-               }
+
                /* we only need one vector */
                nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1, flags);

I'm not sure if you have seen my last comment internally about this patch.

Quoted 7 lines

I don't understand why we need these pci_msix_vec_count() followed by conditional at all.
Perhaps we may simple drop all these and supply flag directly?

But OTOH, I don't know the initial motivation, so, the above patch is non-intrusive and keeps original logic.

Sorry, I must have misunderstood or missed that comment. I am happy to do a V2 if dropping the conditional is preferable.

Quoted 9 lines
Quoted 4 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +0 -1
                if (nvec > 0) {
--

2.17.2

--
With Best Regards,
Andy Shevchenko

BR,
Piotr Stankiewicz
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#3 lore raw
Quoted 13 lines · Piotr
Quoted 13 lines

-----Original Message-----
From: Andy Shevchenko <[email protected]>
Sent: Tuesday, June 2, 2020 11:49 AM
To: Stankiewicz, Piotr <[email protected]>
Cc: Alex Deucher <[email protected]>; Christian König <[email protected]>; David Zhou <[email protected]>; David Airlie <[email protected]>; Daniel Vetter <[email protected]>; amd- [email protected]; dri-devel <[email protected]>; Linux Kernel Mailing List <[email protected]>
Subject: Re: [PATCH 07/15] drm/amdgpu: use PCI_IRQ_MSI_TYPES where appropriate
On Tue, Jun 2, 2020 at 12:24 PM Piotr Stankiewicz <[email protected]> wrote:

...

Quoted 21 lines
Quoted 21 lines
Quoted 4 lines
                int nvec = pci_msix_vec_count(adev->pdev);
                unsigned int flags;
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +0 -1
-               if (nvec <= 0) {

drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c int amdgpu_irq_init(struct amdgpu_device *adev) · line 260

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +4 -3
+               if (nvec > 0)
+                       flags = PCI_IRQ_MSI_TYPES;
+               else
                        flags = PCI_IRQ_MSI;
-               } else {
-                       flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
-               }
+
                /* we only need one vector */
                nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1, flags);

I'm not sure if you have seen my last comment internally about this patch.

I don't understand why we need these pci_msix_vec_count() followed by conditional at all.
Perhaps we may simple drop all these and supply flag directly?

Quoted 6 lines
Quoted 3 lines

But OTOH, I don't know the initial motivation, so, the above patch is non-intrusive and keeps original logic.

Sorry, I must have misunderstood or missed that comment. I am happy to do a V2 if dropping the conditional is preferable.

Let's wait for AMD people to confirm either.

Signature · 6 lines
With Best Regards,
Andy Shevchenko
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#4 lore raw
In reply to Piotr Stankiewicz
Quoted 29 lines

-----Original Message-----
From: dri-devel <[email protected]> On Behalf Of Piotr Stankiewicz
Sent: Tuesday, June 2, 2020 5:21 AM
To: Alex Deucher <[email protected]>; Christian König <[email protected]>; David Zhou <[email protected]>; David Airlie <[email protected]>; Daniel Vetter <[email protected]>
Cc: Stankiewicz, Piotr <[email protected]>; dri- [email protected]; [email protected]; linux- [email protected]
Subject: [PATCH 07/15] drm/amdgpu: use PCI_IRQ_MSI_TYPES where appropriate

Seeing as there is shorthand available to use when asking for any type of interrupt, or any type of message signalled interrupt, leverage it.

---
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c | 8 ++++----

1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +0 -0
index 5ed4227f304b..6dbe173a9fd4 100644
@@ -251,11 +251,11 @@ int amdgpu_irq_init(struct amdgpu_device *adev)
 		int nvec = pci_msix_vec_count(adev->pdev);

drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c int amdgpu_irq_init(struct amdgpu_device *adev) · line 258

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +3 -4
 		unsigned int flags;
 
-		if (nvec <= 0) {
+		if (nvec > 0)
+			flags = PCI_IRQ_MSI_TYPES;
+		else
 			flags = PCI_IRQ_MSI;
-		} else {
-			flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
-		}

Minor nit:

Is it really necessary to set do this check? Can flags just be set?

I.e.:

	flags = PCI_IRQ_MSI_TYPES;

pci_alloc_irq_vector() tries stuff in order. If MSIX is not available, it will try MSI.

M

Quoted 11 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +1 -1
+
 		/* we only need one vector */
 		nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1, flags);
 		if (nvec > 0) {
--

2.17.2

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

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

#5 lore raw
Quoted 12 lines · Michael J
Quoted 12 lines

-----Original Message-----
From: dri-devel <[email protected]> On Behalf Of Piotr Stankiewicz
Sent: Tuesday, June 2, 2020 5:21 AM
To: Alex Deucher <[email protected]>; Christian König <[email protected]>; David Zhou <[email protected]>; David Airlie <[email protected]>; Daniel Vetter <[email protected]>
Cc: Stankiewicz, Piotr <[email protected]>; dri- [email protected]; [email protected]; linux- [email protected]
Subject: [PATCH 07/15] drm/amdgpu: use PCI_IRQ_MSI_TYPES where appropriate

...

Quoted 14 lines

drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c int amdgpu_irq_init(struct amdgpu_device *adev) · line 258

Quoted 11 lines
              int nvec = pci_msix_vec_count(adev->pdev);
              unsigned int flags;
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +3 -4
-              if (nvec <= 0) {
+              if (nvec > 0)
+                      flags = PCI_IRQ_MSI_TYPES;
+              else
                      flags = PCI_IRQ_MSI;
-              } else {
-                      flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
-              }

Minor nit:

Quoted 8 lines

Is it really necessary to set do this check? Can flags just be set?

I.e.:

        flags = PCI_IRQ_MSI_TYPES;

pci_alloc_irq_vector() tries stuff in order. If MSIX is not available, it will try MSI.

That's also what I proposed earlier. But I suggested as well to wait for AMD people to confirm that neither pci_msix_vec_count() nor flags is needed and we can directly supply MSI_TYPES to the below call.

Quoted 2 lines
Quoted 2 lines
              /* we only need one vector */
              nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1, flags);
Signature · 6 lines
With Best Regards,
Andy Shevchenko
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Reviewed-by Tested-by Tom St Denis Reviewed-by Shaoyun liu
#6 lore raw
Quoted 37 lines · Andy Shevchenko
Quoted 12 lines · Michael J
Quoted 12 lines

-----Original Message-----
From: dri-devel <[email protected]> On Behalf Of Piotr Stankiewicz
Sent: Tuesday, June 2, 2020 5:21 AM
To: Alex Deucher <[email protected]>; Christian König <[email protected]>; David Zhou <[email protected]>; David Airlie <[email protected]>; Daniel Vetter <[email protected]>
Cc: Stankiewicz, Piotr <[email protected]>; dri- [email protected]; [email protected]; linux- [email protected]
Subject: [PATCH 07/15] drm/amdgpu: use PCI_IRQ_MSI_TYPES where appropriate

...

Quoted 20 lines

drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c int amdgpu_irq_init(struct amdgpu_device *adev) · line 258

Quoted 11 lines
              int nvec = pci_msix_vec_count(adev->pdev);
              unsigned int flags;
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +3 -4
-              if (nvec <= 0) {
+              if (nvec > 0)
+                      flags = PCI_IRQ_MSI_TYPES;
+              else
                      flags = PCI_IRQ_MSI;
-              } else {
-                      flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
-              }

Minor nit:

Is it really necessary to set do this check? Can flags just be set?

I.e.:

        flags = PCI_IRQ_MSI_TYPES;
Quoted 7 lines
Quoted 2 lines

pci_alloc_irq_vector() tries stuff in order. If MSIX is not available, it will try MSI.

That's also what I proposed earlier. But I suggested as well to wait for AMD people to confirm that neither pci_msix_vec_count() nor flags is needed and we can directly supply MSI_TYPES to the below call.

I think it was leftover from debugging and just to be careful. We had some issues when we originally enabled MSI-X on certain boards. The fix was to just allocate a single vector (since that is all we use anyway) and we were using the wrong irq (pdev->irq vs pci_irq_vector(pdev, 0)). For reference, the original patch to add MSI-X:

commit bd660f4f111161f60392dd02424c3a3d2240dc2f

Author: shaoyunl <[email protected]>
Date:   Tue Oct 1 15:52:31 2019 -0400

drm/amdgpu : enable msix for amdgpu driver

We might used out of the msi resources in some cloud project which have a lot gpu devices(including PF and VF), msix can provide enough resources from system level view

And the fix:

commit 8a745c7ff2ddb8511ef760b4d9cb4cf56a15fc8d

Author: Alex Deucher <[email protected]>
Date:   Thu Oct 3 10:34:30 2019 -0500

drm/amdgpu: improve MSI-X handling (v3)

Check the number of supported vectors and fall back to MSI if we return or error or 0 MSI-X vectors.

v2: only allocate one vector. We can't currently use more than one anyway.

v3: install the irq on vector 0.

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

#7 lore raw
Quoted 34 lines · Alex Deucher
Quoted 24 lines · Andy Shevchenko
Quoted 23 lines · Michael J
Quoted 6 lines

From: dri-devel <[email protected]> On Behalf Of Piotr Stankiewicz

              int nvec = pci_msix_vec_count(adev->pdev);
              unsigned int flags;

drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c int amdgpu_irq_init(struct amdgpu_device *adev) · line 258

Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +3 -4
-              if (nvec <= 0) {
+              if (nvec > 0)
+                      flags = PCI_IRQ_MSI_TYPES;
+              else
                      flags = PCI_IRQ_MSI;
-              } else {
-                      flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
-              }

Minor nit:

Is it really necessary to set do this check? Can flags just be set?

I.e.:

        flags = PCI_IRQ_MSI_TYPES;
Quoted 7 lines
Quoted 2 lines

pci_alloc_irq_vector() tries stuff in order. If MSIX is not available, it will try MSI.

That's also what I proposed earlier. But I suggested as well to wait for AMD people to confirm that neither pci_msix_vec_count() nor flags is needed and we can directly supply MSI_TYPES to the below call.

Quoted 5 lines

I think it was leftover from debugging and just to be careful. We had some issues when we originally enabled MSI-X on certain boards. The fix was to just allocate a single vector (since that is all we use anyway) and we were using the wrong irq (pdev->irq vs pci_irq_vector(pdev, 0)).

Do you agree that simple

  nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1, PCI_IRQ_MSI_TYPES);

will work and we can remove that leftover?

Signature · 6 lines
With Best Regards,
Andy Shevchenko
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#8 lore raw
Quoted 42 lines · Andy Shevchenko
Quoted 34 lines · Alex Deucher
Quoted 24 lines · Andy Shevchenko
Quoted 23 lines · Michael J
Quoted 6 lines

From: dri-devel <[email protected]> On Behalf Of Piotr Stankiewicz

              int nvec = pci_msix_vec_count(adev->pdev);
              unsigned int flags;

Review

Quoted 8 lines
(quoted diff) +3 -4
-              if (nvec <= 0) {
+              if (nvec > 0)
+                      flags = PCI_IRQ_MSI_TYPES;
+              else
                      flags = PCI_IRQ_MSI;
-              } else {
-                      flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
-              }

Minor nit:

Is it really necessary to set do this check? Can flags just be set?

I.e.:

        flags = PCI_IRQ_MSI_TYPES;
Quoted 7 lines
Quoted 2 lines

pci_alloc_irq_vector() tries stuff in order. If MSIX is not available, it will try MSI.

That's also what I proposed earlier. But I suggested as well to wait for AMD people to confirm that neither pci_msix_vec_count() nor flags is needed and we can directly supply MSI_TYPES to the below call.

Quoted 5 lines

I think it was leftover from debugging and just to be careful. We had some issues when we originally enabled MSI-X on certain boards. The fix was to just allocate a single vector (since that is all we use anyway) and we were using the wrong irq (pdev->irq vs pci_irq_vector(pdev, 0)).

Quoted 5 lines

Do you agree that simple

  nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1, PCI_IRQ_MSI_TYPES);

will work and we can remove that leftover?

Yes, I believe so. Tom, can you give this a quick spin on raven just in case if you get a chance? Something like this:

(quoted diff) +0 -0

b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
index 0cc4c67f95f7..c59111b57cc2 100644

drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +2 -8
@@ -248,16 +248,10 @@ int amdgpu_irq_init(struct amdgpu_device *adev)
        adev->irq.msi_enabled = false;
 
        if (amdgpu_msi_ok(adev)) {
-               int nvec = pci_msix_vec_count(adev->pdev);
-               unsigned int flags;
+               int nvec;
 
-               if (nvec <= 0) {
-                       flags = PCI_IRQ_MSI;
-               } else {
-                       flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
-               }
                /* we only need one vector */
-               nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1, flags);
+               nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1,
PCI_IRQ_MSI | PCI_IRQ_MSIX);
                if (nvec > 0) {
                        adev->irq.msi_enabled = true;
                        dev_dbg(adev->dev, "using MSI/MSI-X.\n");

Thanks,

Alex