On 10/28/2023 1:41 AM, Victor Lu wrote:
amdgpu_virt_kiq_reg_write_reg_wait is hardcoded to use MEC engine 0.
Add xcc_inst as a parameter to allow it to use different MEC engines.
v3: use first xcc for MMHUB in gmc_v9_0_flush_gpu_tlb
v2: rebase
Signed-off-by: Victor Lu <victorchengchi.lu@xxxxxxx>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c | 5 +++--
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 3 ++-
drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 26 ++++++++++++++----------
5 files changed, 22 insertions(+), 16 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
index 7a084fbfd33c..82762c61d3ec 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
@@ -73,9 +73,10 @@ void amdgpu_virt_init_setting(struct amdgpu_device *adev)
void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev,
uint32_t reg0, uint32_t reg1,
- uint32_t ref, uint32_t mask)
+ uint32_t ref, uint32_t mask,
+ uint32_t xcc_inst)
{
- struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+ struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_inst];
struct amdgpu_ring *ring = &kiq->ring;
signed long r, cnt = 0;
unsigned long flags;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
index 03c0e38b8aea..5c64258eb728 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
@@ -334,7 +334,8 @@ bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev);
void amdgpu_virt_init_setting(struct amdgpu_device *adev);
void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev,
uint32_t reg0, uint32_t rreg1,
- uint32_t ref, uint32_t mask);
+ uint32_t ref, uint32_t mask,
+ uint32_t xcc_id);
int amdgpu_virt_request_full_gpu(struct amdgpu_device *adev, bool init);
int amdgpu_virt_release_full_gpu(struct amdgpu_device *adev, bool init);
int amdgpu_virt_reset_gpu(struct amdgpu_device *adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
index d8a4fddab9c1..173237e99882 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
@@ -268,7 +268,7 @@ static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
if (adev->gfx.kiq[0].ring.sched.ready && !adev->enable_mes &&
(amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, inv_req,
- 1 << vmid);
+ 1 << vmid, 0);
return;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
index 19eaada35ede..2e4abb356e38 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
@@ -228,7 +228,7 @@ static void gmc_v11_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
if ((adev->gfx.kiq[0].ring.sched.ready || adev->mes.ring.sched.ready) &&
(amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, inv_req,
- 1 << vmid);
+ 1 << vmid, 0);
return;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 0ab9c554da78..32cc3645f02b 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -817,7 +817,7 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
uint32_t vmhub, uint32_t flush_type)
{
bool use_semaphore = gmc_v9_0_use_invalidate_semaphore(adev, vmhub);
- u32 j, inv_req, tmp, sem, req, ack;
+ u32 j, inv_req, tmp, sem, req, ack, inst;
const unsigned int eng = 17;
struct amdgpu_vmhub *hub;
@@ -832,13 +832,17 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
/* This is necessary for a HW workaround under SRIOV as well
* as GFXOFF under bare metal
*/
- if (adev->gfx.kiq[0].ring.sched.ready &&
+ if (vmhub >= AMDGPU_MMHUB0(0))
+ inst = 0;
Same GET_INST(GC, 0) comment here. Looks like this hunk can be combined
with patch 1.
Thanks,
Lijo
+ else
+ inst = vmhub;
+ if (adev->gfx.kiq[inst].ring.sched.ready &&
(amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
uint32_t req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
uint32_t ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, inv_req,
- 1 << vmid);
+ 1 << vmid, inst);
return;
}
@@ -856,9 +860,9 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
for (j = 0; j < adev->usec_timeout; j++) {
/* a read return value of 1 means semaphore acquire */
if (vmhub >= AMDGPU_MMHUB0(0))
- tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, sem, 0);
+ tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, sem, inst);
else
- tmp = RREG32_SOC15_IP_NO_KIQ(GC, sem, vmhub);
+ tmp = RREG32_SOC15_IP_NO_KIQ(GC, sem, inst);
if (tmp & 0x1)
break;
udelay(1);
@@ -869,9 +873,9 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
}
if (vmhub >= AMDGPU_MMHUB0(0))
- WREG32_SOC15_IP_NO_KIQ(MMHUB, req, inv_req, 0);
+ WREG32_SOC15_IP_NO_KIQ(MMHUB, req, inv_req, inst);
else
- WREG32_SOC15_IP_NO_KIQ(GC, req, inv_req, vmhub);
+ WREG32_SOC15_IP_NO_KIQ(GC, req, inv_req, inst);
/*
* Issue a dummy read to wait for the ACK register to
@@ -884,9 +888,9 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
for (j = 0; j < adev->usec_timeout; j++) {
if (vmhub >= AMDGPU_MMHUB0(0))
- tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, ack, 0);
+ tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, ack, inst);
else
- tmp = RREG32_SOC15_IP_NO_KIQ(GC, ack, vmhub);
+ tmp = RREG32_SOC15_IP_NO_KIQ(GC, ack, inst);
if (tmp & (1 << vmid))
break;
udelay(1);
@@ -899,9 +903,9 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
* write with 0 means semaphore release
*/
if (vmhub >= AMDGPU_MMHUB0(0))
- WREG32_SOC15_IP_NO_KIQ(MMHUB, sem, 0, 0);
+ WREG32_SOC15_IP_NO_KIQ(MMHUB, sem, 0, inst);
else
- WREG32_SOC15_IP_NO_KIQ(GC, sem, 0, vmhub);
+ WREG32_SOC15_IP_NO_KIQ(GC, sem, 0, inst);
}
spin_unlock(&adev->gmc.invalidate_lock);