[PATCH] drm/amdgpu/SRIOV: Extend VF reset request wait period

[Zhao, Jiange]

[AMD Public Use]

Hi Hawking,

The process is to send request first, and then to poll and wait 6 seconds.

So the time delta between the 1-st request and the 11-th request is actually 60 seconds.

Jiange
________________________________
From: Zhang, Hawking <Hawking.Zhang at amd.com>
Sent: Tuesday, December 8, 2020 12:16 AM
To: Zhao, Jiange <Jiange.Zhao at amd.com>; amd-gfx at lists.freedesktop.org <amd-gfx at lists.freedesktop.org>
Cc: Chen, Horace <Horace.Chen at amd.com>; Zhang, Andy <Andy.Zhang at amd.com>
Subject: RE: [PATCH] drm/amdgpu/SRIOV: Extend VF reset request wait period

[AMD Public Use]

Re - Poll happens every 6 seconds and it will last for 60 seconds.
+       int ret, i = 0;
+       #define AI_MAILBOX_POLL_MSG_REP_MAX     11

The definition seems not match with your description that the polling will last for 60s.... with that fixed, the patch is

Acked-by: Hawking Zhang <Hawking.Zhang at amd.com>

Regards,
Hawking
-----Original Message-----
From: Zhao, Jiange <Jiange.Zhao at amd.com>
Sent: Monday, December 7, 2020 18:06
To: amd-gfx at lists.freedesktop.org
Cc: Zhang, Hawking <Hawking.Zhang at amd.com>; Chen, Horace <Horace.Chen at amd.com>; Zhang, Andy <Andy.Zhang at amd.com>; Zhao, Jiange <Jiange.Zhao at amd.com>
Subject: [PATCH] drm/amdgpu/SRIOV: Extend VF reset request wait period

From: Jiange Zhao <Jiange.Zhao at amd.com>

In Virtualization case, when one VF is sending too many FLR requests, hypervisor would stop responding to this VF's request for a long period of time. This is called event guard. During this period of cooling time, guest driver should wait instead of doing other things. After this period of time, guest driver would resume reset process and return to normal.

Currently, guest driver would wait 12 seconds and return fail if it doesn't get response from host.

Solution: extend this waiting time in guest driver and poll response periodically. Poll happens every 6 seconds and it will last for 60 seconds.

v2: change the max repetition times from number to macro.

Signed-off-by: Jiange Zhao <Jiange.Zhao at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c | 11 ++++++++++-  drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h |  3 ++-  drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c | 11 ++++++++++-  drivers/gpu/drm/amd/amdgpu/mxgpu_nv.h |  1 +
 4 files changed, 23 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c
index f5ce9a9f4cf5..7767ccca526b 100644
--- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c
+++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c
@@ -187,7 +187,16 @@ static int xgpu_ai_send_access_requests(struct amdgpu_device *adev,

 static int xgpu_ai_request_reset(struct amdgpu_device *adev)  {
-       return xgpu_ai_send_access_requests(adev, IDH_REQ_GPU_RESET_ACCESS);
+       int ret, i = 0;
+
+       while (i < AI_MAILBOX_POLL_MSG_REP_MAX) {
+               ret = xgpu_ai_send_access_requests(adev, IDH_REQ_GPU_RESET_ACCESS);
+               if (!ret)
+                       break;
+               i++;
+       }
+
+       return ret;
 }

 static int xgpu_ai_request_full_gpu_access(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h
index 83b453f5d717..50572635d0f8 100644
--- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h
+++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h
@@ -25,8 +25,9 @@
 #define __MXGPU_AI_H__

 #define AI_MAILBOX_POLL_ACK_TIMEDOUT    500
-#define AI_MAILBOX_POLL_MSG_TIMEDOUT   12000
+#define AI_MAILBOX_POLL_MSG_TIMEDOUT   6000
 #define AI_MAILBOX_POLL_FLR_TIMEDOUT    5000
+#define AI_MAILBOX_POLL_MSG_REP_MAX    11

 enum idh_request {
         IDH_REQ_GPU_INIT_ACCESS = 1,
diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c
index 666ed99cc14b..dd5c1e6ce009 100644
--- a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c
+++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c
@@ -200,7 +200,16 @@ static int xgpu_nv_send_access_requests(struct amdgpu_device *adev,

 static int xgpu_nv_request_reset(struct amdgpu_device *adev)  {
-       return xgpu_nv_send_access_requests(adev, IDH_REQ_GPU_RESET_ACCESS);
+       int ret, i = 0;
+
+       while (i < NV_MAILBOX_POLL_MSG_REP_MAX) {
+               ret = xgpu_nv_send_access_requests(adev, IDH_REQ_GPU_RESET_ACCESS);
+               if (!ret)
+                       break;
+               i++;
+       }
+
+       return ret;
 }

 static int xgpu_nv_request_full_gpu_access(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.h b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.h
index 52605e14a1a5..9f5808616174 100644
--- a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.h
+++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.h
@@ -27,6 +27,7 @@
 #define NV_MAILBOX_POLL_ACK_TIMEDOUT    500
 #define NV_MAILBOX_POLL_MSG_TIMEDOUT    6000
 #define NV_MAILBOX_POLL_FLR_TIMEDOUT    5000
+#define NV_MAILBOX_POLL_MSG_REP_MAX    11

 enum idh_request {
         IDH_REQ_GPU_INIT_ACCESS = 1,
--
2.25.1
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/amd-gfx/attachments/20201208/b30ce828/attachment-0001.htm>