[PATCH] drm/amdgpu: Enable light SBR in XGMI+passthrough configuration

Fri Mar 12 18:00:29 UTC 2021

[AMD Public Use]

Hi , Lijo
For now we  only enable this light sbr feature for SMU  on XGMI + passthrough for Arcturus since this is the use case our customer is required and  we only verified in this configuration .  I feel  it's more reasonable to keep the logic of enable/disable in amdgpu side instead in SMU internally.   We might need to support  the SBR in bare-metal mode  in the future .  Also  we plan to use this feature  in future ASIC if the default HW reset method doesn't works stable.   Your suggestion will need to move the enable/disable logic into SMU internally which valid our original design.

Regards
Shaoyun.liu


From: Lazar, Lijo <Lijo.Lazar at amd.com>
Sent: Friday, March 12, 2021 11:54 AM
To: Liu, Shaoyun <Shaoyun.Liu at amd.com>; amd-gfx at lists.freedesktop.org; Quan, Evan <Evan.Quan at amd.com>; Zhang, Hawking <Hawking.Zhang at amd.com>
Subject: RE: [PATCH] drm/amdgpu: Enable light SBR in XGMI+passthrough configuration


[AMD Public Use]

Looks like this can be handled during post_init. It will be called as smu_post_init() happening during late_init part of smu block. You can check vangogh or navi examples on how to add your implementation.

Thanks,
Lijo

From: Liu, Shaoyun <Shaoyun.Liu at amd.com<mailto:Shaoyun.Liu at amd.com>>
Sent: Friday, March 12, 2021 8:57 PM
To: Lazar, Lijo <Lijo.Lazar at amd.com<mailto:Lijo.Lazar at amd.com>>; amd-gfx at lists.freedesktop.org<mailto:amd-gfx at lists.freedesktop.org>; Quan, Evan <Evan.Quan at amd.com<mailto:Evan.Quan at amd.com>>; Zhang, Hawking <Hawking.Zhang at amd.com<mailto:Hawking.Zhang at amd.com>>
Subject: RE: [PATCH] drm/amdgpu: Enable light SBR in XGMI+passthrough configuration


[AMD Public Use]

I don't like to add this set_light_sbr into ppt_funcs  either , but please check current swsmu  code structure ,  there is no asic specific swsmu late init function  and  there is no direct routine form  amdgpu_smu.c to smu_v11_0.c either . It requires  smu common code ->ppt_func -> smu_v11_0 for Arcturus  specific function .  So unless SMU and  PPT have a major re-structure , set_light_sbr need to go through ppt_func for now,  I think I  better  leave this  re-structure task to SMU and  PPT owner in the future .

Add  SMU and  PPT code owner  Hawking  and Quan for comments .

Regards
Shaoyun.liu


From: Lazar, Lijo <Lijo.Lazar at amd.com<mailto:Lijo.Lazar at amd.com>>
Sent: Friday, March 12, 2021 8:55 AM
To: Liu, Shaoyun <Shaoyun.Liu at amd.com<mailto:Shaoyun.Liu at amd.com>>; amd-gfx at lists.freedesktop.org<mailto:amd-gfx at lists.freedesktop.org>
Subject: RE: [PATCH] drm/amdgpu: Enable light SBR in XGMI+passthrough configuration


[AMD Public Use]

We want to keep ppt_funcs minimal. Adding everything to ppt_funcs and keeping as NULL is not the right way. Please keep the code to arcturus.

Thanks,
Lijo

From: Liu, Shaoyun <Shaoyun.Liu at amd.com<mailto:Shaoyun.Liu at amd.com>>
Sent: Friday, March 12, 2021 7:21 PM
To: Lazar, Lijo <Lijo.Lazar at amd.com<mailto:Lijo.Lazar at amd.com>>; amd-gfx at lists.freedesktop.org<mailto:amd-gfx at lists.freedesktop.org>
Subject: Re: [PATCH] drm/amdgpu: Enable light SBR in XGMI+passthrough configuration

Thanks for the comments. This light sbr solution could be applied to other asic as well. In swsmu code, It will check whether the function pointer set_light_sbr is valid before real call the function. So for other asics if the smu apply the same change, just add the ppt function pointer and we will have this support without further code change.

Thanks
Shaoyun.liu

________________________________
From: Lazar, Lijo <Lijo.Lazar at amd.com<mailto:Lijo.Lazar at amd.com>>
Sent: March 11, 2021 10:42 PM
To: Liu, Shaoyun <Shaoyun.Liu at amd.com<mailto:Shaoyun.Liu at amd.com>>; amd-gfx at lists.freedesktop.org<mailto:amd-gfx at lists.freedesktop.org> <amd-gfx at lists.freedesktop.org<mailto:amd-gfx at lists.freedesktop.org>>
Cc: Liu, Shaoyun <Shaoyun.Liu at amd.com<mailto:Shaoyun.Liu at amd.com>>
Subject: RE: [PATCH] drm/amdgpu: Enable light SBR in XGMI+passthrough configuration

[AMD Public Use]

We don't need this as a generic ppt_func. Reset functionalities are changing over programs and this could be valid only for Arcturus. Please move it to Arcturus swsmu late init.

Thanks,
Lijo

-----Original Message-----
From: amd-gfx <amd-gfx-bounces at lists.freedesktop.org<mailto:amd-gfx-bounces at lists.freedesktop.org>> On Behalf Of shaoyunl
Sent: Thursday, March 11, 2021 10:46 PM
To: amd-gfx at lists.freedesktop.org<mailto:amd-gfx at lists.freedesktop.org>
Cc: Liu, Shaoyun <Shaoyun.Liu at amd.com<mailto:Shaoyun.Liu at amd.com>>
Subject: [PATCH] drm/amdgpu: Enable light SBR in XGMI+passthrough configuration

This is to fix the commit dda9bbb26c7 where it only enable the light SMU on normal device init. This feature actually need to be enabled after ASIC been reset as well.

Signed-off-by: shaoyunl <shaoyun.liu at amd.com<mailto:shaoyun.liu at amd.com>>
Change-Id: Ie7ee02cd3ccdab3522aad9a02f681963e211ed44
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index cada3e77c7d5..fb775a9c0db1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -2513,6 +2513,9 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
         if (r)
                 DRM_ERROR("enable mgpu fan boost failed (%d).\n", r);

+       /* For XGMI + passthrough configuration , enable light SBR */
+       if (amdgpu_passthrough(adev) && adev->gmc.xgmi.num_physical_nodes > 1)
+               smu_set_light_sbr(&adev->smu, true);

         if (adev->gmc.xgmi.num_physical_nodes > 1) {
                 mutex_lock(&mgpu_info.mutex);
@@ -3615,10 +3618,6 @@ int amdgpu_device_init(struct amdgpu_device *adev,
         if (amdgpu_device_cache_pci_state(adev->pdev))
                 pci_restore_state(pdev);

-       /* Enable lightSBR on SMU in passthrough + xgmi configuration */
-       if (amdgpu_passthrough(adev) && adev->gmc.xgmi.num_physical_nodes > 1)
-               smu_set_light_sbr(&adev->smu, true);
-
         if (adev->gmc.xgmi.pending_reset)
                 queue_delayed_work(system_wq, &mgpu_info.delayed_reset_work,
                                    msecs_to_jiffies(AMDGPU_RESUME_MS));
--
2.17.1

_______________________________________________
amd-gfx mailing list
amd-gfx at lists.freedesktop.org<mailto:amd-gfx at lists.freedesktop.org>
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=04%7C01%7Clijo.lazar%40amd.com%7Cc5aedb6c2d9d49d2fee408d8e4b15b5c%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637510797685776785%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=7ndhtOVmyZcRMe3UQiGvF%2BprCdBVgo7f6IATXSbQNg4%3D&reserved=0
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/amd-gfx/attachments/20210312/87273d74/attachment-0001.htm>
