* [PATCH] Revert "drm/amdgpu: Ensure the DMA engine is deactivated during set ups"
@ 2022-06-06 2:14 Guchun Chen
2022-06-06 13:27 ` Alex Deucher
0 siblings, 1 reply; 2+ messages in thread
From: Guchun Chen @ 2022-06-06 2:14 UTC (permalink / raw)
To: amd-gfx, hawking.zhang, alexander.deucher, christian.koenig
Cc: Alexander Deucher, ricetons, Guchun Chen
This reverts commit da38a66ac46e334f198afcd1b4d4554b4ddca0df.
This causes regression in GPU reset related test.
Cc: Alexander Deucher <Alexander.Deucher@amd.com>
Cc: ricetons@gmail.com
Signed-off-by: Guchun Chen <guchun.chen@amd.com>
---
drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c | 109 ++++++++++---------------
1 file changed, 45 insertions(+), 64 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
index 06b2635b142a..83c6ccaaa9e4 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
@@ -469,6 +469,7 @@ static void sdma_v5_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se
}
}
+
/**
* sdma_v5_2_gfx_stop - stop the gfx async dma engines
*
@@ -514,21 +515,17 @@ static void sdma_v5_2_rlc_stop(struct amdgpu_device *adev)
}
/**
- * sdma_v5_2_ctx_switch_enable_for_instance - start the async dma engines
- * context switch for an instance
+ * sdma_v5_2_ctx_switch_enable - stop the async dma engines context switch
*
* @adev: amdgpu_device pointer
- * @instance_idx: the index of the SDMA instance
+ * @enable: enable/disable the DMA MEs context switch.
*
- * Unhalt the async dma engines context switch.
+ * Halt or unhalt the async dma engines context switch.
*/
-static void sdma_v5_2_ctx_switch_enable_for_instance(struct amdgpu_device *adev, int instance_idx)
+static void sdma_v5_2_ctx_switch_enable(struct amdgpu_device *adev, bool enable)
{
u32 f32_cntl, phase_quantum = 0;
-
- if (WARN_ON(instance_idx >= adev->sdma.num_instances)) {
- return;
- }
+ int i;
if (amdgpu_sdma_phase_quantum) {
unsigned value = amdgpu_sdma_phase_quantum;
@@ -552,68 +549,50 @@ static void sdma_v5_2_ctx_switch_enable_for_instance(struct amdgpu_device *adev,
phase_quantum =
value << SDMA0_PHASE0_QUANTUM__VALUE__SHIFT |
unit << SDMA0_PHASE0_QUANTUM__UNIT__SHIFT;
-
- WREG32_SOC15_IP(GC,
- sdma_v5_2_get_reg_offset(adev, instance_idx, mmSDMA0_PHASE0_QUANTUM),
- phase_quantum);
- WREG32_SOC15_IP(GC,
- sdma_v5_2_get_reg_offset(adev, instance_idx, mmSDMA0_PHASE1_QUANTUM),
- phase_quantum);
- WREG32_SOC15_IP(GC,
- sdma_v5_2_get_reg_offset(adev, instance_idx, mmSDMA0_PHASE2_QUANTUM),
- phase_quantum);
}
- if (!amdgpu_sriov_vf(adev)) {
- f32_cntl = RREG32(sdma_v5_2_get_reg_offset(adev, instance_idx, mmSDMA0_CNTL));
- f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL,
- AUTO_CTXSW_ENABLE, 1);
- WREG32(sdma_v5_2_get_reg_offset(adev, instance_idx, mmSDMA0_CNTL), f32_cntl);
+ for (i = 0; i < adev->sdma.num_instances; i++) {
+ if (enable && amdgpu_sdma_phase_quantum) {
+ WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_PHASE0_QUANTUM),
+ phase_quantum);
+ WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_PHASE1_QUANTUM),
+ phase_quantum);
+ WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_PHASE2_QUANTUM),
+ phase_quantum);
+ }
+
+ if (!amdgpu_sriov_vf(adev)) {
+ f32_cntl = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL));
+ f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL,
+ AUTO_CTXSW_ENABLE, enable ? 1 : 0);
+ WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL), f32_cntl);
+ }
}
+
}
/**
- * sdma_v5_2_ctx_switch_disable_all - stop the async dma engines context switch
+ * sdma_v5_2_enable - stop the async dma engines
*
* @adev: amdgpu_device pointer
+ * @enable: enable/disable the DMA MEs.
*
- * Halt the async dma engines context switch.
+ * Halt or unhalt the async dma engines.
*/
-static void sdma_v5_2_ctx_switch_disable_all(struct amdgpu_device *adev)
+static void sdma_v5_2_enable(struct amdgpu_device *adev, bool enable)
{
u32 f32_cntl;
int i;
- if (amdgpu_sriov_vf(adev))
- return;
-
- for (i = 0; i < adev->sdma.num_instances; i++) {
- f32_cntl = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL));
- f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL,
- AUTO_CTXSW_ENABLE, 0);
- WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL), f32_cntl);
+ if (!enable) {
+ sdma_v5_2_gfx_stop(adev);
+ sdma_v5_2_rlc_stop(adev);
}
-}
-
-/**
- * sdma_v5_2_halt - stop the async dma engines
- *
- * @adev: amdgpu_device pointer
- *
- * Halt the async dma engines.
- */
-static void sdma_v5_2_halt(struct amdgpu_device *adev)
-{
- int i;
- u32 f32_cntl;
-
- sdma_v5_2_gfx_stop(adev);
- sdma_v5_2_rlc_stop(adev);
if (!amdgpu_sriov_vf(adev)) {
for (i = 0; i < adev->sdma.num_instances; i++) {
f32_cntl = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_F32_CNTL));
- f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_F32_CNTL, HALT, 1);
+ f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_F32_CNTL, HALT, enable ? 0 : 1);
WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_F32_CNTL), f32_cntl);
}
}
@@ -625,9 +604,6 @@ static void sdma_v5_2_halt(struct amdgpu_device *adev)
* @adev: amdgpu_device pointer
*
* Set up the gfx DMA ring buffers and enable them.
- * It assumes that the dma engine is stopped for each instance.
- * The function enables the engine and preemptions sequentially for each instance.
- *
* Returns 0 for success, error for failure.
*/
static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev)
@@ -769,7 +745,10 @@ static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev)
ring->sched.ready = true;
- sdma_v5_2_ctx_switch_enable_for_instance(adev, i);
+ if (amdgpu_sriov_vf(adev)) { /* bare-metal sequence doesn't need below to lines */
+ sdma_v5_2_ctx_switch_enable(adev, true);
+ sdma_v5_2_enable(adev, true);
+ }
r = amdgpu_ring_test_ring(ring);
if (r) {
@@ -813,7 +792,7 @@ static int sdma_v5_2_load_microcode(struct amdgpu_device *adev)
int i, j;
/* halt the MEs */
- sdma_v5_2_halt(adev);
+ sdma_v5_2_enable(adev, false);
for (i = 0; i < adev->sdma.num_instances; i++) {
if (!adev->sdma.instance[i].fw)
@@ -885,8 +864,8 @@ static int sdma_v5_2_start(struct amdgpu_device *adev)
int r = 0;
if (amdgpu_sriov_vf(adev)) {
- sdma_v5_2_ctx_switch_disable_all(adev);
- sdma_v5_2_halt(adev);
+ sdma_v5_2_ctx_switch_enable(adev, false);
+ sdma_v5_2_enable(adev, false);
/* set RB registers */
r = sdma_v5_2_gfx_resume(adev);
@@ -910,10 +889,12 @@ static int sdma_v5_2_start(struct amdgpu_device *adev)
amdgpu_gfx_off_ctrl(adev, false);
sdma_v5_2_soft_reset(adev);
+ /* unhalt the MEs */
+ sdma_v5_2_enable(adev, true);
+ /* enable sdma ring preemption */
+ sdma_v5_2_ctx_switch_enable(adev, true);
- /* Soft reset supposes to disable the dma engine and preemption.
- * Now start the gfx rings and rlc compute queues.
- */
+ /* start the gfx rings and rlc compute queues */
r = sdma_v5_2_gfx_resume(adev);
if (adev->in_s0ix)
amdgpu_gfx_off_ctrl(adev, true);
@@ -1447,8 +1428,8 @@ static int sdma_v5_2_hw_fini(void *handle)
if (amdgpu_sriov_vf(adev))
return 0;
- sdma_v5_2_ctx_switch_disable_all(adev);
- sdma_v5_2_halt(adev);
+ sdma_v5_2_ctx_switch_enable(adev, false);
+ sdma_v5_2_enable(adev, false);
return 0;
}
--
2.17.1
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [PATCH] Revert "drm/amdgpu: Ensure the DMA engine is deactivated during set ups"
2022-06-06 2:14 [PATCH] Revert "drm/amdgpu: Ensure the DMA engine is deactivated during set ups" Guchun Chen
@ 2022-06-06 13:27 ` Alex Deucher
0 siblings, 0 replies; 2+ messages in thread
From: Alex Deucher @ 2022-06-06 13:27 UTC (permalink / raw)
To: Guchun Chen
Cc: Deucher, Alexander, Haohui Mai, Christian Koenig, amd-gfx list,
Hawking Zhang
On Sun, Jun 5, 2022 at 10:14 PM Guchun Chen <guchun.chen@amd.com> wrote:
>
> This reverts commit da38a66ac46e334f198afcd1b4d4554b4ddca0df.
>
> This causes regression in GPU reset related test.
>
> Cc: Alexander Deucher <Alexander.Deucher@amd.com>
> Cc: ricetons@gmail.com
> Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c | 109 ++++++++++---------------
> 1 file changed, 45 insertions(+), 64 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
> index 06b2635b142a..83c6ccaaa9e4 100644
> --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
> +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
> @@ -469,6 +469,7 @@ static void sdma_v5_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se
> }
> }
>
> +
> /**
> * sdma_v5_2_gfx_stop - stop the gfx async dma engines
> *
> @@ -514,21 +515,17 @@ static void sdma_v5_2_rlc_stop(struct amdgpu_device *adev)
> }
>
> /**
> - * sdma_v5_2_ctx_switch_enable_for_instance - start the async dma engines
> - * context switch for an instance
> + * sdma_v5_2_ctx_switch_enable - stop the async dma engines context switch
> *
> * @adev: amdgpu_device pointer
> - * @instance_idx: the index of the SDMA instance
> + * @enable: enable/disable the DMA MEs context switch.
> *
> - * Unhalt the async dma engines context switch.
> + * Halt or unhalt the async dma engines context switch.
> */
> -static void sdma_v5_2_ctx_switch_enable_for_instance(struct amdgpu_device *adev, int instance_idx)
> +static void sdma_v5_2_ctx_switch_enable(struct amdgpu_device *adev, bool enable)
> {
> u32 f32_cntl, phase_quantum = 0;
> -
> - if (WARN_ON(instance_idx >= adev->sdma.num_instances)) {
> - return;
> - }
> + int i;
>
> if (amdgpu_sdma_phase_quantum) {
> unsigned value = amdgpu_sdma_phase_quantum;
> @@ -552,68 +549,50 @@ static void sdma_v5_2_ctx_switch_enable_for_instance(struct amdgpu_device *adev,
> phase_quantum =
> value << SDMA0_PHASE0_QUANTUM__VALUE__SHIFT |
> unit << SDMA0_PHASE0_QUANTUM__UNIT__SHIFT;
> -
> - WREG32_SOC15_IP(GC,
> - sdma_v5_2_get_reg_offset(adev, instance_idx, mmSDMA0_PHASE0_QUANTUM),
> - phase_quantum);
> - WREG32_SOC15_IP(GC,
> - sdma_v5_2_get_reg_offset(adev, instance_idx, mmSDMA0_PHASE1_QUANTUM),
> - phase_quantum);
> - WREG32_SOC15_IP(GC,
> - sdma_v5_2_get_reg_offset(adev, instance_idx, mmSDMA0_PHASE2_QUANTUM),
> - phase_quantum);
> }
>
> - if (!amdgpu_sriov_vf(adev)) {
> - f32_cntl = RREG32(sdma_v5_2_get_reg_offset(adev, instance_idx, mmSDMA0_CNTL));
> - f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL,
> - AUTO_CTXSW_ENABLE, 1);
> - WREG32(sdma_v5_2_get_reg_offset(adev, instance_idx, mmSDMA0_CNTL), f32_cntl);
> + for (i = 0; i < adev->sdma.num_instances; i++) {
> + if (enable && amdgpu_sdma_phase_quantum) {
> + WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_PHASE0_QUANTUM),
> + phase_quantum);
> + WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_PHASE1_QUANTUM),
> + phase_quantum);
> + WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_PHASE2_QUANTUM),
> + phase_quantum);
> + }
> +
> + if (!amdgpu_sriov_vf(adev)) {
> + f32_cntl = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL));
> + f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL,
> + AUTO_CTXSW_ENABLE, enable ? 1 : 0);
> + WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL), f32_cntl);
> + }
> }
> +
> }
>
> /**
> - * sdma_v5_2_ctx_switch_disable_all - stop the async dma engines context switch
> + * sdma_v5_2_enable - stop the async dma engines
> *
> * @adev: amdgpu_device pointer
> + * @enable: enable/disable the DMA MEs.
> *
> - * Halt the async dma engines context switch.
> + * Halt or unhalt the async dma engines.
> */
> -static void sdma_v5_2_ctx_switch_disable_all(struct amdgpu_device *adev)
> +static void sdma_v5_2_enable(struct amdgpu_device *adev, bool enable)
> {
> u32 f32_cntl;
> int i;
>
> - if (amdgpu_sriov_vf(adev))
> - return;
> -
> - for (i = 0; i < adev->sdma.num_instances; i++) {
> - f32_cntl = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL));
> - f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL,
> - AUTO_CTXSW_ENABLE, 0);
> - WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL), f32_cntl);
> + if (!enable) {
> + sdma_v5_2_gfx_stop(adev);
> + sdma_v5_2_rlc_stop(adev);
> }
> -}
> -
> -/**
> - * sdma_v5_2_halt - stop the async dma engines
> - *
> - * @adev: amdgpu_device pointer
> - *
> - * Halt the async dma engines.
> - */
> -static void sdma_v5_2_halt(struct amdgpu_device *adev)
> -{
> - int i;
> - u32 f32_cntl;
> -
> - sdma_v5_2_gfx_stop(adev);
> - sdma_v5_2_rlc_stop(adev);
>
> if (!amdgpu_sriov_vf(adev)) {
> for (i = 0; i < adev->sdma.num_instances; i++) {
> f32_cntl = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_F32_CNTL));
> - f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_F32_CNTL, HALT, 1);
> + f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_F32_CNTL, HALT, enable ? 0 : 1);
> WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_F32_CNTL), f32_cntl);
> }
> }
> @@ -625,9 +604,6 @@ static void sdma_v5_2_halt(struct amdgpu_device *adev)
> * @adev: amdgpu_device pointer
> *
> * Set up the gfx DMA ring buffers and enable them.
> - * It assumes that the dma engine is stopped for each instance.
> - * The function enables the engine and preemptions sequentially for each instance.
> - *
> * Returns 0 for success, error for failure.
> */
> static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev)
> @@ -769,7 +745,10 @@ static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev)
>
> ring->sched.ready = true;
>
> - sdma_v5_2_ctx_switch_enable_for_instance(adev, i);
> + if (amdgpu_sriov_vf(adev)) { /* bare-metal sequence doesn't need below to lines */
> + sdma_v5_2_ctx_switch_enable(adev, true);
> + sdma_v5_2_enable(adev, true);
> + }
>
> r = amdgpu_ring_test_ring(ring);
> if (r) {
> @@ -813,7 +792,7 @@ static int sdma_v5_2_load_microcode(struct amdgpu_device *adev)
> int i, j;
>
> /* halt the MEs */
> - sdma_v5_2_halt(adev);
> + sdma_v5_2_enable(adev, false);
>
> for (i = 0; i < adev->sdma.num_instances; i++) {
> if (!adev->sdma.instance[i].fw)
> @@ -885,8 +864,8 @@ static int sdma_v5_2_start(struct amdgpu_device *adev)
> int r = 0;
>
> if (amdgpu_sriov_vf(adev)) {
> - sdma_v5_2_ctx_switch_disable_all(adev);
> - sdma_v5_2_halt(adev);
> + sdma_v5_2_ctx_switch_enable(adev, false);
> + sdma_v5_2_enable(adev, false);
>
> /* set RB registers */
> r = sdma_v5_2_gfx_resume(adev);
> @@ -910,10 +889,12 @@ static int sdma_v5_2_start(struct amdgpu_device *adev)
> amdgpu_gfx_off_ctrl(adev, false);
>
> sdma_v5_2_soft_reset(adev);
> + /* unhalt the MEs */
> + sdma_v5_2_enable(adev, true);
> + /* enable sdma ring preemption */
> + sdma_v5_2_ctx_switch_enable(adev, true);
>
> - /* Soft reset supposes to disable the dma engine and preemption.
> - * Now start the gfx rings and rlc compute queues.
> - */
> + /* start the gfx rings and rlc compute queues */
> r = sdma_v5_2_gfx_resume(adev);
> if (adev->in_s0ix)
> amdgpu_gfx_off_ctrl(adev, true);
> @@ -1447,8 +1428,8 @@ static int sdma_v5_2_hw_fini(void *handle)
> if (amdgpu_sriov_vf(adev))
> return 0;
>
> - sdma_v5_2_ctx_switch_disable_all(adev);
> - sdma_v5_2_halt(adev);
> + sdma_v5_2_ctx_switch_enable(adev, false);
> + sdma_v5_2_enable(adev, false);
>
> return 0;
> }
> --
> 2.17.1
>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2022-06-06 13:27 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-06-06 2:14 [PATCH] Revert "drm/amdgpu: Ensure the DMA engine is deactivated during set ups" Guchun Chen
2022-06-06 13:27 ` Alex Deucher
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.