* [PATCH 1/2] drm/amdgpu: Support MES suspend_all_sdma_gangs
2026-03-23 18:44 [PATCH 0/2] Support SDMA queue reset Amber Lin
@ 2026-03-23 18:44 ` Amber Lin
2026-03-23 19:25 ` Alex Deucher
0 siblings, 1 reply; 4+ messages in thread
From: Amber Lin @ 2026-03-23 18:44 UTC (permalink / raw)
To: amd-gfx; +Cc: Michael.Chen, Shaoyun.Liu, Amber Lin
suspend_all_sdma_gangs is supported in new MES firmware for gfx 12.1
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c | 3 +++
drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h | 1 +
drivers/gpu/drm/amd/amdgpu/mes_v12_1.c | 1 +
3 files changed, 5 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
index fd6b40d9da58..dfe18f0a3501 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
@@ -311,6 +311,9 @@ int amdgpu_mes_suspend(struct amdgpu_device *adev, uint32_t xcc_id)
memset(&input, 0x0, sizeof(struct mes_suspend_gang_input));
input.suspend_all_gangs = 1;
input.xcc_id = xcc_id;
+ if ((amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(12, 1, 0)) &&
+ ((adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 0x71))
+ input.suspend_all_sdma_gangs = 1;
/*
* Avoid taking any other locks under MES lock to avoid circular
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
index 44fa4d73bce8..2d08e33eb1e9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
@@ -298,6 +298,7 @@ struct mes_unmap_legacy_queue_input {
struct mes_suspend_gang_input {
uint32_t xcc_id;
bool suspend_all_gangs;
+ bool suspend_all_sdma_gangs;
uint64_t gang_context_addr;
uint64_t suspend_fence_addr;
uint32_t suspend_fence_value;
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c b/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c
index ac9e26b8bb52..8a0c3dc0ecb7 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c
@@ -479,6 +479,7 @@ static int mes_v12_1_suspend_gang(struct amdgpu_mes *mes,
mes_suspend_gang_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
mes_suspend_gang_pkt.suspend_all_gangs = input->suspend_all_gangs;
+ mes_suspend_gang_pkt.suspend_all_sdma_gangs = input->suspend_all_sdma_gangs;
mes_suspend_gang_pkt.gang_context_addr = input->gang_context_addr;
mes_suspend_gang_pkt.suspend_fence_addr = input->suspend_fence_addr;
mes_suspend_gang_pkt.suspend_fence_value = input->suspend_fence_value;
--
2.43.0
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH 1/2] drm/amdgpu: Support MES suspend_all_sdma_gangs
2026-03-23 18:44 ` [PATCH 1/2] drm/amdgpu: Support MES suspend_all_sdma_gangs Amber Lin
@ 2026-03-23 19:25 ` Alex Deucher
0 siblings, 0 replies; 4+ messages in thread
From: Alex Deucher @ 2026-03-23 19:25 UTC (permalink / raw)
To: Amber Lin; +Cc: amd-gfx, Michael.Chen, Shaoyun.Liu
On Mon, Mar 23, 2026 at 2:54 PM Amber Lin <Amber.Lin@amd.com> wrote:
>
> suspend_all_sdma_gangs is supported in new MES firmware for gfx 12.1
>
> Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c | 3 +++
> drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h | 1 +
> drivers/gpu/drm/amd/amdgpu/mes_v12_1.c | 1 +
> 3 files changed, 5 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
> index fd6b40d9da58..dfe18f0a3501 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
> @@ -311,6 +311,9 @@ int amdgpu_mes_suspend(struct amdgpu_device *adev, uint32_t xcc_id)
> memset(&input, 0x0, sizeof(struct mes_suspend_gang_input));
> input.suspend_all_gangs = 1;
> input.xcc_id = xcc_id;
> + if ((amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(12, 1, 0)) &&
> + ((adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 0x71))
> + input.suspend_all_sdma_gangs = 1;
>
> /*
> * Avoid taking any other locks under MES lock to avoid circular
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
> index 44fa4d73bce8..2d08e33eb1e9 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
> @@ -298,6 +298,7 @@ struct mes_unmap_legacy_queue_input {
> struct mes_suspend_gang_input {
> uint32_t xcc_id;
> bool suspend_all_gangs;
> + bool suspend_all_sdma_gangs;
> uint64_t gang_context_addr;
> uint64_t suspend_fence_addr;
> uint32_t suspend_fence_value;
> diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c b/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c
> index ac9e26b8bb52..8a0c3dc0ecb7 100644
> --- a/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c
> +++ b/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c
> @@ -479,6 +479,7 @@ static int mes_v12_1_suspend_gang(struct amdgpu_mes *mes,
> mes_suspend_gang_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
>
> mes_suspend_gang_pkt.suspend_all_gangs = input->suspend_all_gangs;
> + mes_suspend_gang_pkt.suspend_all_sdma_gangs = input->suspend_all_sdma_gangs;
> mes_suspend_gang_pkt.gang_context_addr = input->gang_context_addr;
> mes_suspend_gang_pkt.suspend_fence_addr = input->suspend_fence_addr;
> mes_suspend_gang_pkt.suspend_fence_value = input->suspend_fence_value;
> --
> 2.43.0
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH 1/2] drm/amdgpu: Support MES suspend_all_sdma_gangs
@ 2026-05-13 2:59 Amber Lin
2026-05-13 2:59 ` [PATCH 2/2] drm/amdkfd: Enable SDMA queue reset on gfx v12.1 Amber Lin
0 siblings, 1 reply; 4+ messages in thread
From: Amber Lin @ 2026-05-13 2:59 UTC (permalink / raw)
To: amd-gfx; +Cc: alexander.deucher, Amber Lin, Michael Chen
suspend_all_sdma_gangs is supported in new MES firmware for gfx 12.1
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Michael Chen<michael.chen@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c | 3 +++
drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h | 1 +
drivers/gpu/drm/amd/amdgpu/mes_v12_1.c | 1 +
3 files changed, 5 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
index 610dc2ca67a2..c9467b26e42c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
@@ -312,6 +312,9 @@ int amdgpu_mes_suspend(struct amdgpu_device *adev, u32 xcc_id)
memset(&input, 0x0, sizeof(struct mes_suspend_gang_input));
input.suspend_all_gangs = 1;
input.xcc_id = xcc_id;
+ if ((amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(12, 1, 0)) &&
+ ((adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 0x71))
+ input.suspend_all_sdma_gangs = 1;
/*
* Avoid taking any other locks under MES lock to avoid circular
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
index 31cad8a5bb9f..93990d4990f2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
@@ -298,6 +298,7 @@ struct mes_unmap_legacy_queue_input {
struct mes_suspend_gang_input {
uint32_t xcc_id;
bool suspend_all_gangs;
+ bool suspend_all_sdma_gangs;
uint64_t gang_context_addr;
uint64_t suspend_fence_addr;
uint32_t suspend_fence_value;
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c b/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c
index 4217b3fea0f3..b169e577e583 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c
@@ -480,6 +480,7 @@ static int mes_v12_1_suspend_gang(struct amdgpu_mes *mes,
mes_suspend_gang_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
mes_suspend_gang_pkt.suspend_all_gangs = input->suspend_all_gangs;
+ mes_suspend_gang_pkt.suspend_all_sdma_gangs = input->suspend_all_sdma_gangs;
mes_suspend_gang_pkt.gang_context_addr = input->gang_context_addr;
mes_suspend_gang_pkt.suspend_fence_addr = input->suspend_fence_addr;
mes_suspend_gang_pkt.suspend_fence_value = input->suspend_fence_value;
--
2.43.0
^ permalink raw reply related [flat|nested] 4+ messages in thread
* [PATCH 2/2] drm/amdkfd: Enable SDMA queue reset on gfx v12.1
2026-05-13 2:59 [PATCH 1/2] drm/amdgpu: Support MES suspend_all_sdma_gangs Amber Lin
@ 2026-05-13 2:59 ` Amber Lin
0 siblings, 0 replies; 4+ messages in thread
From: Amber Lin @ 2026-05-13 2:59 UTC (permalink / raw)
To: amd-gfx; +Cc: alexander.deucher, Amber Lin, Michael Chen
After suspend/resume sdma_gang is supported on MES 12.1, SDMA queue reset
is supported too.
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Reviewed-by: Michael Chen<michael.chen@amd.com>
---
drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index 397e916e11b6..f37c2d3c8416 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -514,7 +514,8 @@ static ssize_t node_show(struct kobject *kobj, struct attribute *attr,
dev->node_props.capability |=
HSA_CAP_AQL_QUEUE_DOUBLE_MAP;
- if (KFD_GC_VERSION(dev->gpu) < IP_VERSION(10, 0, 0) &&
+ if ((KFD_GC_VERSION(dev->gpu) < IP_VERSION(10, 0, 0) ||
+ KFD_GC_VERSION(dev->gpu) == IP_VERSION(12, 1, 0)) &&
(dev->gpu->adev->sdma.supported_reset & AMDGPU_RESET_TYPE_PER_QUEUE))
dev->node_props.capability2 |= HSA_CAP2_PER_SDMA_QUEUE_RESET_SUPPORTED;
--
2.43.0
^ permalink raw reply related [flat|nested] 4+ messages in thread
end of thread, other threads:[~2026-05-13 3:00 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-05-13 2:59 [PATCH 1/2] drm/amdgpu: Support MES suspend_all_sdma_gangs Amber Lin
2026-05-13 2:59 ` [PATCH 2/2] drm/amdkfd: Enable SDMA queue reset on gfx v12.1 Amber Lin
-- strict thread matches above, loose matches on Subject: below --
2026-03-23 18:44 [PATCH 0/2] Support SDMA queue reset Amber Lin
2026-03-23 18:44 ` [PATCH 1/2] drm/amdgpu: Support MES suspend_all_sdma_gangs Amber Lin
2026-03-23 19:25 ` Alex Deucher
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox