From be697aa3a78ef83a6b8d49e1f0671a002e502cd0 Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Wed, 27 Jul 2022 14:35:49 +0800
Subject: drm/amdgpu: move queue_bitmap to an independent structure (v3)
To allocate independent queue_bitmap for each XCD,
then the old bitmap policy can be continued to use
with a clear logic.
Use mec_bitmap[0] as default for all non-GC 9.4.3 IPs.
v2: squash commits to avoid breaking the build
v3: unify naming style
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 0385f7f69278..fed8bb9a721f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -162,7 +162,7 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
* clear
*/
bitmap_complement(gpu_resources.cp_queue_bitmap,
- adev->gfx.mec.queue_bitmap,
+ adev->gfx.mec_bitmap[0].queue_bitmap,
KGD_MAX_QUEUES);
/* According to linux/bitmap.h we shouldn't use bitmap_clear if
--
cgit v1.2.3
From 27fb73a0e3aa7478bcb5d2d59d65eec3c68fc165 Mon Sep 17 00:00:00 2001
From: Mukul Joshi <mukul.joshi@amd.com>
Date: Tue, 25 Apr 2023 14:19:05 -0400
Subject: drm/amdkfd: Update KFD TTM mem limit
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Use the helper function in TTM to get TTM memory
limit and set KFD's internal mem limit. This ensures
that KFD's TTM mem limit and actual TTM mem limit are
exactly same.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 3 ++-
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 6 +++++-
2 files changed, 7 insertions(+), 2 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index fed8bb9a721f..a46285841d17 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -53,7 +53,6 @@ int amdgpu_amdkfd_init(void)
amdgpu_amdkfd_total_mem_size *= si.mem_unit;
ret = kgd2kfd_init();
- amdgpu_amdkfd_gpuvm_init_mem_limits();
kfd_initialized = !ret;
return ret;
@@ -143,6 +142,8 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
int i;
int last_valid_bit;
+ amdgpu_amdkfd_gpuvm_init_mem_limits();
+
if (adev->kfd.dev) {
struct kgd2kfd_shared_resources gpu_resources = {
.compute_vmid_bitmap =
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 4432e169fae8..c3990a5eb7c6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -36,6 +36,7 @@
#include <uapi/linux/kfd_ioctl.h>
#include "amdgpu_xgmi.h"
#include "kfd_smi_events.h"
+#include <drm/ttm/ttm_tt.h>
/* Userptr restore delay, just long enough to allow consecutive VM
* changes to accumulate
@@ -110,13 +111,16 @@ void amdgpu_amdkfd_gpuvm_init_mem_limits(void)
struct sysinfo si;
uint64_t mem;
+ if (kfd_mem_limit.max_system_mem_limit)
+ return;
+
si_meminfo(&si);
mem = si.freeram - si.freehigh;
mem *= si.mem_unit;
spin_lock_init(&kfd_mem_limit.mem_limit_lock);
kfd_mem_limit.max_system_mem_limit = mem - (mem >> 4);
- kfd_mem_limit.max_ttm_mem_limit = (mem >> 1) - (mem >> 3);
+ kfd_mem_limit.max_ttm_mem_limit = ttm_tt_pages_limit() << PAGE_SHIFT;
pr_debug("Kernel memory limit %lluM, TTM limit %lluM\n",
(kfd_mem_limit.max_system_mem_limit >> 20),
(kfd_mem_limit.max_ttm_mem_limit >> 20));
--
cgit v1.2.3
From f4caf5842652f08e024741ef6d423cb0c101d863 Mon Sep 17 00:00:00 2001
From: Hawking Zhang <Hawking.Zhang@amd.com>
Date: Wed, 14 Sep 2022 16:35:50 +0800
Subject: drm/amdgpu: introduce vmhub definition for multi-partition cases (v3)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
v1: Each partition has its own gfxhub or mmhub. adjust
the num of MAX_VMHUBS and the GFXHUB/MMHUB layout (Le)
v2: re-design the AMDGPU_GFXHUB/AMDGPU_MMHUB layout (Le)
v3: apply the gfxhub/mmhub layout to new IPs (Hawking)
v4: fix up gmc11 (Alex)
v5: rebase (Alex)
Signed-off-by: Le Ma <le.ma@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 4 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 4 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 13 +++--
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 6 +--
drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 8 +--
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 8 +--
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c | 2 +-
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 4 +-
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c | 12 ++---
drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 33 ++++++-------
drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c | 26 +++++-----
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 78 +++++++++++++++---------------
drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c | 4 +-
drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/mes_v10_1.c | 2 +-
drivers/gpu/drm/amd/amdgpu/mes_v11_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c | 12 ++---
drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 4 +-
drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c | 2 +-
drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c | 4 +-
drivers/gpu/drm/amd/amdgpu/vce_v4_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 4 +-
drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c | 4 +-
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c | 8 +--
drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c | 4 +-
drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c | 2 +-
47 files changed, 204 insertions(+), 204 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index a46285841d17..f0a136d35279 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -736,7 +736,7 @@ int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
for (i = 0; i < adev->num_vmhubs; i++)
amdgpu_gmc_flush_gpu_tlb(adev, vmid, i, 0);
} else {
- amdgpu_gmc_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB_0, 0);
+ amdgpu_gmc_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB(0), 0);
}
return 0;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index 87e1a1a9f298..488b3bb6dcb1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -315,7 +315,7 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
ring->use_doorbell = true;
ring->doorbell_index = adev->doorbell_index.kiq;
ring->xcc_id = xcc_id;
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
if (xcc_id >= 1)
ring->doorbell_index = adev->doorbell_index.xcc1_kiq_start +
xcc_id - 1;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
index 4e2531758866..0a4e5fcfec6b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
@@ -670,7 +670,7 @@ void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type,
for (i = 0; i < 16; i++) {
reg = hub->vm_context0_cntl + hub->ctx_distance * i;
- tmp = (hub_type == AMDGPU_GFXHUB_0) ?
+ tmp = (hub_type == AMDGPU_GFXHUB(0)) ?
RREG32_SOC15_IP(GC, reg) :
RREG32_SOC15_IP(MMHUB, reg);
@@ -679,7 +679,7 @@ void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type,
else
tmp &= ~hub->vm_cntx_cntl_vm_fault;
- (hub_type == AMDGPU_GFXHUB_0) ?
+ (hub_type == AMDGPU_GFXHUB(0)) ?
WREG32_SOC15_IP(GC, reg, tmp) :
WREG32_SOC15_IP(MMHUB, reg, tmp);
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index b6bd667df676..c3964c14f215 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -2374,12 +2374,12 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
case AMDGPU_VM_OP_RESERVE_VMID:
/* We only have requirement to reserve vmid from gfxhub */
r = amdgpu_vmid_alloc_reserved(adev, &fpriv->vm,
- AMDGPU_GFXHUB_0);
+ AMDGPU_GFXHUB(0));
if (r)
return r;
break;
case AMDGPU_VM_OP_UNRESERVE_VMID:
- amdgpu_vmid_free_reserved(adev, &fpriv->vm, AMDGPU_GFXHUB_0);
+ amdgpu_vmid_free_reserved(adev, &fpriv->vm, AMDGPU_GFXHUB(0));
break;
default:
return -EINVAL;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
index 6f085f0b4ef3..9f5d32b0fda1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
@@ -111,11 +111,14 @@ struct amdgpu_mem_stats;
/* Reserve 4MB VRAM for page tables */
#define AMDGPU_VM_RESERVED_VRAM (8ULL << 20)
-/* max number of VMHUB */
-#define AMDGPU_MAX_VMHUBS 3
-#define AMDGPU_GFXHUB_0 0
-#define AMDGPU_MMHUB_0 1
-#define AMDGPU_MMHUB_1 2
+/*
+ * max number of VMHUB
+ * layout: max 8 GFXHUB + 4 MMHUB0 + 1 MMHUB1
+ */
+#define AMDGPU_MAX_VMHUBS 13
+#define AMDGPU_GFXHUB(x) (x)
+#define AMDGPU_MMHUB0(x) (8 + x)
+#define AMDGPU_MMHUB1(x) (8 + 4 + x)
/* Reserve 2MB at top/bottom of address space for kernel use */
#define AMDGPU_VA_RESERVED_SIZE (2ULL << 20)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index 8e86b2c23c0a..7b585141e10e 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
@@ -4461,7 +4461,7 @@ static int gfx_v10_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
else
ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1;
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe;
@@ -4490,7 +4490,7 @@ static int gfx_v10_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+ (ring_id * GFX10_MEC_HPD_SIZE);
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@@ -4978,7 +4978,7 @@ static void gfx_v10_0_constants_init(struct amdgpu_device *adev)
/* XXX SH_MEM regs */
/* where to put LDS, scratch, GPUVM in FSA64 space */
mutex_lock(&adev->srbm_mutex);
- for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
+ for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
nv_grbm_select(adev, 0, 0, 0, i);
/* CP and shaders */
WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
index f77779c31043..790df2cc3480 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
@@ -906,7 +906,7 @@ static int gfx_v11_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
else
ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1;
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe;
@@ -937,7 +937,7 @@ static int gfx_v11_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+ (ring_id * GFX11_MEC_HPD_SIZE);
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@@ -1707,7 +1707,7 @@ static void gfx_v11_0_constants_init(struct amdgpu_device *adev)
/* XXX SH_MEM regs */
/* where to put LDS, scratch, GPUVM in FSA64 space */
mutex_lock(&adev->srbm_mutex);
- for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
+ for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
soc21_grbm_select(adev, 0, 0, 0, i);
/* CP and shaders */
WREG32_SOC15(GC, 0, regSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG);
@@ -4190,7 +4190,7 @@ static int gfx_v11_0_gfxhub_enable(struct amdgpu_device *adev)
false : true;
adev->gfxhub.funcs->set_fault_enable_default(adev, value);
- amdgpu_gmc_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB_0, 0);
+ amdgpu_gmc_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB(0), 0);
return 0;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index 46577b59cb04..91814dc083c9 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -2005,7 +2005,7 @@ static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+ (ring_id * GFX9_MEC_HPD_SIZE);
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@@ -2105,7 +2105,7 @@ static int gfx_v9_0_sw_init(void *handle)
/* disable scheduler on the real ring */
ring->no_scheduler = true;
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq,
AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP,
AMDGPU_RING_PRIO_DEFAULT, NULL);
@@ -2123,7 +2123,7 @@ static int gfx_v9_0_sw_init(void *handle)
ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
ring->is_sw_ring = true;
hw_prio = amdgpu_sw_ring_priority(i);
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq,
AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP, hw_prio,
NULL);
@@ -2393,7 +2393,7 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev)
/* XXX SH_MEM regs */
/* where to put LDS, scratch, GPUVM in FSA64 space */
mutex_lock(&adev->srbm_mutex);
- for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
+ for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
soc15_grbm_select(adev, 0, 0, 0, i, 0);
/* CP and shaders */
if (i == 0) {
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c
index d648a29c33e0..ec7c049c5952 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c
@@ -1935,7 +1935,7 @@ static bool gfx_v9_4_2_query_uctl2_poison_status(struct amdgpu_device *adev)
u32 status = 0;
struct amdgpu_vmhub *hub;
- hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
status = RREG32(hub->vm_l2_pro_fault_status);
/* reset page fault status */
WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index 9d17dcfae130..f5104b982633 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -757,7 +757,7 @@ static int gfx_v9_4_3_compute_ring_init(struct amdgpu_device *adev, int ring_id,
(adev->doorbell_index.mec_ring0 + ring_id) << 1;
ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+ (ring_id * GFX9_MEC_HPD_SIZE);
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "comp_%d.%d.%d.%d",
ring->xcc_id, ring->me, ring->pipe, ring->queue);
@@ -996,7 +996,7 @@ static void gfx_v9_4_3_constants_init(struct amdgpu_device *adev)
/* XXX SH_MEM regs */
/* where to put LDS, scratch, GPUVM in FSA64 space */
mutex_lock(&adev->srbm_mutex);
- for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
+ for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
for (j = 0; j < adev->gfx.num_xcd; j++) {
soc15_grbm_select(adev, 0, 0, 0, i, j);
/* CP and shaders */
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
index ab2325f6c7ac..d94cc1ec7242 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
@@ -40,7 +40,7 @@ static void gfxhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev,
uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -247,7 +247,7 @@ static void gfxhub_v1_0_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned num_level, block_size;
uint32_t tmp;
int i;
@@ -307,7 +307,7 @@ static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v1_0_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned i;
for (i = 0 ; i < 18; ++i) {
@@ -338,7 +338,7 @@ static int gfxhub_v1_0_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v1_0_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
u32 tmp;
u32 i;
@@ -411,7 +411,7 @@ static void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev,
static void gfxhub_v1_0_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index 79af32bb078c..9c385ce3a8c4 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -42,7 +42,7 @@ static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
int i;
for (i = 0; i < adev->gfx.num_xcd; i++) {
@@ -291,7 +291,7 @@ static void gfxhub_v1_2_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned num_level, block_size;
uint32_t tmp;
int i, j;
@@ -357,7 +357,7 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v1_2_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned i, j;
for (j = 0; j < adev->gfx.num_xcd; j++) {
@@ -406,7 +406,7 @@ static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
u32 tmp;
u32 i, j;
@@ -483,7 +483,7 @@ static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
static void gfxhub_v1_2_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
index 9b3a02527318..f173a61c6c15 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
@@ -120,7 +120,7 @@ static u64 gfxhub_v2_0_get_mc_fb_offset(struct amdgpu_device *adev)
static void gfxhub_v2_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -282,7 +282,7 @@ static void gfxhub_v2_0_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
int i;
uint32_t tmp;
@@ -331,7 +331,7 @@ static void gfxhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v2_0_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned i;
for (i = 0 ; i < 18; ++i) {
@@ -360,7 +360,7 @@ static int gfxhub_v2_0_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v2_0_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
u32 tmp;
u32 i;
@@ -433,7 +433,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v2_0_vmhub_funcs = {
static void gfxhub_v2_0_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
index 4aacbbec31e2..d8fc3e8088cd 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
@@ -123,7 +123,7 @@ static u64 gfxhub_v2_1_get_mc_fb_offset(struct amdgpu_device *adev)
static void gfxhub_v2_1_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -291,7 +291,7 @@ static void gfxhub_v2_1_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v2_1_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
int i;
uint32_t tmp;
@@ -340,7 +340,7 @@ static void gfxhub_v2_1_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v2_1_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned i;
for (i = 0 ; i < 18; ++i) {
@@ -381,7 +381,7 @@ static int gfxhub_v2_1_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v2_1_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
u32 tmp;
u32 i;
@@ -462,7 +462,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v2_1_vmhub_funcs = {
static void gfxhub_v2_1_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(GC, 0,
@@ -651,7 +651,7 @@ static void gfxhub_v2_1_restore_regs(struct amdgpu_device *adev)
static void gfxhub_v2_1_halt(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
int i;
uint32_t tmp;
int time = 1000;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c
index 13712640fa46..c53147f9c9fc 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c
@@ -119,7 +119,7 @@ static u64 gfxhub_v3_0_get_mc_fb_offset(struct amdgpu_device *adev)
static void gfxhub_v3_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -290,7 +290,7 @@ static void gfxhub_v3_0_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
int i;
uint32_t tmp;
@@ -339,7 +339,7 @@ static void gfxhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v3_0_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned i;
for (i = 0 ; i < 18; ++i) {
@@ -380,7 +380,7 @@ static int gfxhub_v3_0_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v3_0_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
u32 tmp;
u32 i;
@@ -463,7 +463,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v3_0_vmhub_funcs = {
static void gfxhub_v3_0_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c
index 6e0bd628c889..ae777487d72e 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c
@@ -122,7 +122,7 @@ static u64 gfxhub_v3_0_3_get_mc_fb_offset(struct amdgpu_device *adev)
static void gfxhub_v3_0_3_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -295,7 +295,7 @@ static void gfxhub_v3_0_3_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v3_0_3_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
int i;
uint32_t tmp;
@@ -344,7 +344,7 @@ static void gfxhub_v3_0_3_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v3_0_3_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned i;
for (i = 0 ; i < 18; ++i) {
@@ -373,7 +373,7 @@ static int gfxhub_v3_0_3_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v3_0_3_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
u32 tmp;
u32 i;
@@ -451,7 +451,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v3_0_3_vmhub_funcs = {
static void gfxhub_v3_0_3_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
index 5697b66bf0de..ea2a448147e3 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
@@ -76,7 +76,7 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
switch (state) {
case AMDGPU_IRQ_STATE_DISABLE:
/* MM HUB */
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, false);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), false);
/* GFX HUB */
/* This works because this interrupt is only
* enabled at init/resume and disabled in
@@ -84,11 +84,11 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
* change over the course of suspend/resume.
*/
if (!adev->in_s0ix)
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, false);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), false);
break;
case AMDGPU_IRQ_STATE_ENABLE:
/* MM HUB */
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, true);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), true);
/* GFX HUB */
/* This works because this interrupt is only
* enabled at init/resume and disabled in
@@ -96,7 +96,7 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
* change over the course of suspend/resume.
*/
if (!adev->in_s0ix)
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, true);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), true);
break;
default:
break;
@@ -149,7 +149,7 @@ static int gmc_v10_0_process_interrupt(struct amdgpu_device *adev,
* be updated to avoid reading an incorrect value due to
* the new fast GRBM interface.
*/
- if ((entry->vmid_src == AMDGPU_GFXHUB_0) &&
+ if ((entry->vmid_src == AMDGPU_GFXHUB(0)) &&
(adev->ip_versions[GC_HWIP][0] < IP_VERSION(10, 3, 0)))
RREG32(hub->vm_l2_pro_fault_status);
@@ -212,8 +212,7 @@ static void gmc_v10_0_set_irq_funcs(struct amdgpu_device *adev)
static bool gmc_v10_0_use_invalidate_semaphore(struct amdgpu_device *adev,
uint32_t vmhub)
{
- return ((vmhub == AMDGPU_MMHUB_0 ||
- vmhub == AMDGPU_MMHUB_1) &&
+ return ((vmhub == AMDGPU_MMHUB0(0)) &&
(!amdgpu_sriov_vf(adev)));
}
@@ -249,7 +248,7 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
unsigned int i;
unsigned char hub_ip = 0;
- hub_ip = (vmhub == AMDGPU_GFXHUB_0) ?
+ hub_ip = (vmhub == AMDGPU_GFXHUB(0)) ?
GC_HWIP : MMHUB_HWIP;
spin_lock(&adev->gmc.invalidate_lock);
@@ -284,7 +283,7 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
* Issue a dummy read to wait for the ACK register to be cleared
* to avoid a false ACK due to the new fast GRBM interface.
*/
- if ((vmhub == AMDGPU_GFXHUB_0) &&
+ if ((vmhub == AMDGPU_GFXHUB(0)) &&
(adev->ip_versions[GC_HWIP][0] < IP_VERSION(10, 3, 0)))
RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_req +
hub->eng_distance * eng, hub_ip);
@@ -361,19 +360,19 @@ static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
mutex_lock(&adev->mman.gtt_window_lock);
- if (vmhub == AMDGPU_MMHUB_0) {
- gmc_v10_0_flush_vm_hub(adev, vmid, AMDGPU_MMHUB_0, 0);
+ if (vmhub == AMDGPU_MMHUB0(0)) {
+ gmc_v10_0_flush_vm_hub(adev, vmid, AMDGPU_MMHUB0(0), 0);
mutex_unlock(&adev->mman.gtt_window_lock);
return;
}
- BUG_ON(vmhub != AMDGPU_GFXHUB_0);
+ BUG_ON(vmhub != AMDGPU_GFXHUB(0));
if (!adev->mman.buffer_funcs_enabled ||
!adev->ib_pool_ready ||
amdgpu_in_reset(adev) ||
ring->sched.ready == false) {
- gmc_v10_0_flush_vm_hub(adev, vmid, AMDGPU_GFXHUB_0, 0);
+ gmc_v10_0_flush_vm_hub(adev, vmid, AMDGPU_GFXHUB(0), 0);
mutex_unlock(&adev->mman.gtt_window_lock);
return;
}
@@ -466,7 +465,7 @@ static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
i, flush_type);
} else {
gmc_v10_0_flush_gpu_tlb(adev, vmid,
- AMDGPU_GFXHUB_0, flush_type);
+ AMDGPU_GFXHUB(0), flush_type);
}
if (!adev->enable_mes)
break;
@@ -534,7 +533,7 @@ static void gmc_v10_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid
if (ring->is_mes_queue)
return;
- if (ring->vm_hub == AMDGPU_GFXHUB_0)
+ if (ring->vm_hub == AMDGPU_GFXHUB(0))
reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid;
else
reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid;
@@ -1075,9 +1074,9 @@ static int gmc_v10_0_gart_enable(struct amdgpu_device *adev)
if (!adev->in_s0ix)
adev->gfxhub.funcs->set_fault_enable_default(adev, value);
adev->mmhub.funcs->set_fault_enable_default(adev, value);
- gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB_0, 0);
+ gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB0(0), 0);
if (!adev->in_s0ix)
- gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB_0, 0);
+ gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB(0), 0);
DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
(unsigned)(adev->gmc.gart_size >> 20),
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
index 2f570fb5febe..fb2ac31cbba7 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
@@ -64,7 +64,7 @@ gmc_v11_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
switch (state) {
case AMDGPU_IRQ_STATE_DISABLE:
/* MM HUB */
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, false);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), false);
/* GFX HUB */
/* This works because this interrupt is only
* enabled at init/resume and disabled in
@@ -72,11 +72,11 @@ gmc_v11_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
* change over the course of suspend/resume.
*/
if (!adev->in_s0ix)
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, false);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), false);
break;
case AMDGPU_IRQ_STATE_ENABLE:
/* MM HUB */
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, true);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), true);
/* GFX HUB */
/* This works because this interrupt is only
* enabled at init/resume and disabled in
@@ -84,7 +84,7 @@ gmc_v11_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
* change over the course of suspend/resume.
*/
if (!adev->in_s0ix)
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, true);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), true);
break;
default:
break;
@@ -110,7 +110,7 @@ static int gmc_v11_0_process_interrupt(struct amdgpu_device *adev,
* be updated to avoid reading an incorrect value due to
* the new fast GRBM interface.
*/
- if (entry->vmid_src == AMDGPU_GFXHUB_0)
+ if (entry->vmid_src == AMDGPU_GFXHUB(0))
RREG32(hub->vm_l2_pro_fault_status);
status = RREG32(hub->vm_l2_pro_fault_status);
@@ -170,7 +170,7 @@ static void gmc_v11_0_set_irq_funcs(struct amdgpu_device *adev)
static bool gmc_v11_0_use_invalidate_semaphore(struct amdgpu_device *adev,
uint32_t vmhub)
{
- return ((vmhub == AMDGPU_MMHUB_0) &&
+ return ((vmhub == AMDGPU_MMHUB0(0)) &&
(!amdgpu_sriov_vf(adev)));
}
@@ -202,7 +202,7 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
unsigned int i;
unsigned char hub_ip = 0;
- hub_ip = (vmhub == AMDGPU_GFXHUB_0) ?
+ hub_ip = (vmhub == AMDGPU_GFXHUB(0)) ?
GC_HWIP : MMHUB_HWIP;
spin_lock(&adev->gmc.invalidate_lock);
@@ -251,7 +251,7 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
hub->eng_distance * eng, 0, hub_ip);
/* Issue additional private vm invalidation to MMHUB */
- if ((vmhub != AMDGPU_GFXHUB_0) &&
+ if ((vmhub != AMDGPU_GFXHUB(0)) &&
(hub->vm_l2_bank_select_reserved_cid2) &&
!amdgpu_sriov_vf(adev)) {
inv_req = RREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2);
@@ -284,7 +284,7 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
static void gmc_v11_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
uint32_t vmhub, uint32_t flush_type)
{
- if ((vmhub == AMDGPU_GFXHUB_0) && !adev->gfx.is_poweron)
+ if ((vmhub == AMDGPU_GFXHUB(0)) && !adev->gfx.is_poweron)
return;
/* flush hdp cache */
@@ -369,7 +369,7 @@ static int gmc_v11_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
i, flush_type);
} else {
gmc_v11_0_flush_gpu_tlb(adev, vmid,
- AMDGPU_GFXHUB_0, flush_type);
+ AMDGPU_GFXHUB(0), flush_type);
}
}
}
@@ -435,7 +435,7 @@ static void gmc_v11_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid
if (ring->is_mes_queue)
return;
- if (ring->vm_hub == AMDGPU_GFXHUB_0)
+ if (ring->vm_hub == AMDGPU_GFXHUB(0))
reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid;
else
reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT_MM) + vmid;
@@ -886,7 +886,7 @@ static int gmc_v11_0_sw_fini(void *handle)
static void gmc_v11_0_init_golden_registers(struct amdgpu_device *adev)
{
if (amdgpu_sriov_vf(adev)) {
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32(hub->vm_contexts_disable, 0);
return;
@@ -921,7 +921,7 @@ static int gmc_v11_0_gart_enable(struct amdgpu_device *adev)
false : true;
adev->mmhub.funcs->set_fault_enable_default(adev, value);
- gmc_v11_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB_0, 0);
+ gmc_v11_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB0(0), 0);
DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
(unsigned)(adev->gmc.gart_size >> 20),
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 6ae5cee9b64b..193ba4d912a6 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -491,20 +491,20 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
* fini/suspend, so the overall state doesn't
* change over the course of suspend/resume.
*/
- if (adev->in_s0ix && (j == AMDGPU_GFXHUB_0))
+ if (adev->in_s0ix && (j == AMDGPU_GFXHUB(0)))
continue;
- if (j == AMDGPU_GFXHUB_0)
- tmp = RREG32_SOC15_IP(GC, reg);
- else
+ if (j >= AMDGPU_MMHUB0(0))
tmp = RREG32_SOC15_IP(MMHUB, reg);
+ else
+ tmp = RREG32_SOC15_IP(GC, reg);
tmp &= ~bits;
- if (j == AMDGPU_GFXHUB_0)
- WREG32_SOC15_IP(GC, reg, tmp);
- else
+ if (j >= AMDGPU_MMHUB0(0))
WREG32_SOC15_IP(MMHUB, reg, tmp);
+ else
+ WREG32_SOC15_IP(GC, reg, tmp);
}
}
break;
@@ -519,20 +519,20 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
* fini/suspend, so the overall state doesn't
* change over the course of suspend/resume.
*/
- if (adev->in_s0ix && (j == AMDGPU_GFXHUB_0))
+ if (adev->in_s0ix && (j == AMDGPU_GFXHUB(0)))
continue;
- if (j == AMDGPU_GFXHUB_0)
- tmp = RREG32_SOC15_IP(GC, reg);
- else
+ if (j >= AMDGPU_MMHUB0(0))
tmp = RREG32_SOC15_IP(MMHUB, reg);
+ else
+ tmp = RREG32_SOC15_IP(GC, reg);
tmp |= bits;
- if (j == AMDGPU_GFXHUB_0)
- WREG32_SOC15_IP(GC, reg, tmp);
- else
+ if (j >= AMDGPU_MMHUB0(0))
WREG32_SOC15_IP(MMHUB, reg, tmp);
+ else
+ WREG32_SOC15_IP(GC, reg, tmp);
}
}
break;
@@ -605,13 +605,13 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
if (entry->client_id == SOC15_IH_CLIENTID_VMC) {
hub_name = "mmhub0";
- hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
} else if (entry->client_id == SOC15_IH_CLIENTID_VMC1) {
hub_name = "mmhub1";
- hub = &adev->vmhub[AMDGPU_MMHUB_1];
+ hub = &adev->vmhub[AMDGPU_MMHUB1(0)];
} else {
hub_name = "gfxhub0";
- hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
}
memset(&task_info, 0, sizeof(struct amdgpu_task_info));
@@ -636,7 +636,7 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
* be updated to avoid reading an incorrect value due to
* the new fast GRBM interface.
*/
- if ((entry->vmid_src == AMDGPU_GFXHUB_0) &&
+ if ((entry->vmid_src == AMDGPU_GFXHUB(0)) &&
(adev->ip_versions[GC_HWIP][0] < IP_VERSION(9, 4, 2)))
RREG32(hub->vm_l2_pro_fault_status);
@@ -649,7 +649,7 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
dev_err(adev->dev,
"VM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
status);
- if (hub == &adev->vmhub[AMDGPU_GFXHUB_0]) {
+ if (hub == &adev->vmhub[AMDGPU_GFXHUB(0)]) {
dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n",
cid >= ARRAY_SIZE(gfxhub_client_ids) ? "unknown" :
gfxhub_client_ids[cid],
@@ -759,8 +759,8 @@ static bool gmc_v9_0_use_invalidate_semaphore(struct amdgpu_device *adev,
adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3))
return false;
- return ((vmhub == AMDGPU_MMHUB_0 ||
- vmhub == AMDGPU_MMHUB_1) &&
+ return ((vmhub == AMDGPU_MMHUB0(0) ||
+ vmhub == AMDGPU_MMHUB1(0)) &&
(!amdgpu_sriov_vf(adev)) &&
(!(!(adev->apu_flags & AMD_APU_IS_RAVEN2) &&
(adev->apu_flags & AMD_APU_IS_PICASSO))));
@@ -849,11 +849,10 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
if (use_semaphore) {
for (j = 0; j < adev->usec_timeout; j++) {
/* a read return value of 1 means semaphore acquire */
- if (vmhub == AMDGPU_GFXHUB_0)
- tmp = RREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_sem + hub->eng_distance * eng);
- else
+ if (vmhub >= AMDGPU_MMHUB0(0))
tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_sem + hub->eng_distance * eng);
-
+ else
+ tmp = RREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_sem + hub->eng_distance * eng);
if (tmp & 0x1)
break;
udelay(1);
@@ -864,27 +863,26 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
}
do {
- if (vmhub == AMDGPU_GFXHUB_0)
- WREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
- else
+ if (vmhub >= AMDGPU_MMHUB0(0))
WREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
+ else
+ WREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
/*
* Issue a dummy read to wait for the ACK register to
* be cleared to avoid a false ACK due to the new fast
* GRBM interface.
*/
- if ((vmhub == AMDGPU_GFXHUB_0) &&
+ if ((vmhub == AMDGPU_GFXHUB(0)) &&
(adev->ip_versions[GC_HWIP][0] < IP_VERSION(9, 4, 2)))
RREG32_NO_KIQ(hub->vm_inv_eng0_req +
hub->eng_distance * eng);
for (j = 0; j < adev->usec_timeout; j++) {
- if (vmhub == AMDGPU_GFXHUB_0)
- tmp = RREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_ack + hub->eng_distance * eng);
- else
+ if (vmhub >= AMDGPU_MMHUB0(0))
tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_ack + hub->eng_distance * eng);
-
+ else
+ tmp = RREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_ack + hub->eng_distance * eng);
if (tmp & (1 << vmid))
break;
udelay(1);
@@ -900,10 +898,10 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
* add semaphore release after invalidation,
* write with 0 means semaphore release
*/
- if (vmhub == AMDGPU_GFXHUB_0)
- WREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_sem + hub->eng_distance * eng, 0);
+ if (vmhub >= AMDGPU_MMHUB0(0))
+ WREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
else
- WREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_sem + hub->eng_distance * eng, 0);
+ WREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
}
spin_unlock(&adev->gmc.invalidate_lock);
@@ -994,7 +992,7 @@ static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
i, flush_type);
} else {
gmc_v9_0_flush_gpu_tlb(adev, vmid,
- AMDGPU_GFXHUB_0, flush_type);
+ AMDGPU_GFXHUB(0), flush_type);
}
break;
}
@@ -1060,10 +1058,10 @@ static void gmc_v9_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
uint32_t reg;
/* Do nothing because there's no lut register for mmhub1. */
- if (ring->vm_hub == AMDGPU_MMHUB_1)
+ if (ring->vm_hub == AMDGPU_MMHUB1(0))
return;
- if (ring->vm_hub == AMDGPU_GFXHUB_0)
+ if (ring->vm_hub == AMDGPU_GFXHUB(0))
reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid;
else
reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid;
@@ -1947,7 +1945,7 @@ static int gmc_v9_0_hw_init(void *handle)
adev->mmhub.funcs->set_fault_enable_default(adev, value);
}
for (i = 0; i < adev->num_vmhubs; ++i) {
- if (adev->in_s0ix && (i == AMDGPU_GFXHUB_0))
+ if (adev->in_s0ix && (i == AMDGPU_GFXHUB(0)))
continue;
gmc_v9_0_flush_gpu_tlb(adev, 0, i, 0);
}
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
index a3076eb8af6a..71fe7f6f9889 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
@@ -485,7 +485,7 @@ int jpeg_v1_0_sw_init(void *handle)
return r;
ring = &adev->jpeg.inst->ring_dec;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "jpeg_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq,
0, AMDGPU_RING_PRIO_DEFAULT, NULL);
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c
index 0eddf7c824a7..3a43e42f4834 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c
@@ -86,7 +86,7 @@ static int jpeg_v2_0_sw_init(void *handle)
ring = &adev->jpeg.inst->ring_dec;
ring->use_doorbell = true;
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "jpeg_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq,
0, AMDGPU_RING_PRIO_DEFAULT, NULL);
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c
index b040f51d9aa9..259b7ba6a842 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c
@@ -128,9 +128,9 @@ static int jpeg_v2_5_sw_init(void *handle)
ring = &adev->jpeg.inst[i].ring_dec;
ring->use_doorbell = true;
if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
- ring->vm_hub = AMDGPU_MMHUB_1;
+ ring->vm_hub = AMDGPU_MMHUB1(0);
else
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1 + 8 * i;
sprintf(ring->name, "jpeg_dec_%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst[i].irq,
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c
index 1c2292cc5f2c..c55386c22311 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c
@@ -101,7 +101,7 @@ static int jpeg_v3_0_sw_init(void *handle)
ring = &adev->jpeg.inst->ring_dec;
ring->use_doorbell = true;
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "jpeg_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
AMDGPU_RING_PRIO_DEFAULT, NULL);
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c
index 77e1e64aa1d1..d7d5ffc29393 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c
@@ -108,7 +108,7 @@ static int jpeg_v4_0_sw_init(void *handle)
ring = &adev->jpeg.inst->ring_dec;
ring->use_doorbell = true;
ring->doorbell_index = amdgpu_sriov_vf(adev) ? (((adev->doorbell_index.vcn.vcn_ring0_1) << 1) + 4) : ((adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1);
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "jpeg_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c b/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c
index 4560476c7c31..f1a6abdad21b 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c
@@ -149,7 +149,7 @@ static int mes_v10_1_add_hw_queue(struct amdgpu_mes *mes,
{
struct amdgpu_device *adev = mes->adev;
union MESAPI__ADD_QUEUE mes_add_queue_pkt;
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
uint32_t vm_cntx_cntl = hub->vm_cntx_cntl;
memset(&mes_add_queue_pkt, 0, sizeof(mes_add_queue_pkt));
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
index 3adb450eec07..9791f3581786 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
@@ -164,7 +164,7 @@ static int mes_v11_0_add_hw_queue(struct amdgpu_mes *mes,
{
struct amdgpu_device *adev = mes->adev;
union MESAPI__ADD_QUEUE mes_add_queue_pkt;
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
uint32_t vm_cntx_cntl = hub->vm_cntx_cntl;
memset(&mes_add_queue_pkt, 0, sizeof(mes_add_queue_pkt));
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
index 15e7cbeae75b..fb91b31056ca 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
@@ -54,7 +54,7 @@ static u64 mmhub_v1_0_get_fb_location(struct amdgpu_device *adev)
static void mmhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -229,7 +229,7 @@ static void mmhub_v1_0_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned num_level, block_size;
uint32_t tmp;
int i;
@@ -285,7 +285,7 @@ static void mmhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v1_0_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -338,7 +338,7 @@ static int mmhub_v1_0_gart_enable(struct amdgpu_device *adev)
static void mmhub_v1_0_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -415,7 +415,7 @@ static void mmhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, bool
static void mmhub_v1_0_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c
index 73afbf2facc9..9086f2fdfaf4 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c
@@ -54,7 +54,7 @@ static u64 mmhub_v1_7_get_fb_location(struct amdgpu_device *adev)
static void mmhub_v1_7_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid, lower_32_bits(page_table_base));
@@ -261,7 +261,7 @@ static void mmhub_v1_7_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v1_7_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned num_level, block_size;
uint32_t tmp;
int i;
@@ -319,7 +319,7 @@ static void mmhub_v1_7_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v1_7_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -348,7 +348,7 @@ static int mmhub_v1_7_gart_enable(struct amdgpu_device *adev)
static void mmhub_v1_7_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -425,7 +425,7 @@ static void mmhub_v1_7_set_fault_enable_default(struct amdgpu_device *adev, bool
static void mmhub_v1_7_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
index 342d1702104c..9ec06f9db761 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
@@ -53,7 +53,7 @@ static u64 mmhub_v1_8_get_fb_location(struct amdgpu_device *adev)
static void mmhub_v1_8_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid, lower_32_bits(page_table_base));
@@ -253,7 +253,7 @@ static void mmhub_v1_8_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v1_8_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned num_level, block_size;
uint32_t tmp;
int i;
@@ -311,7 +311,7 @@ static void mmhub_v1_8_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v1_8_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -352,7 +352,7 @@ static int mmhub_v1_8_gart_enable(struct amdgpu_device *adev)
static void mmhub_v1_8_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -426,7 +426,7 @@ static void mmhub_v1_8_set_fault_enable_default(struct amdgpu_device *adev, bool
static void mmhub_v1_8_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c
index 278e32db878d..8f76c6ecf50a 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c
@@ -187,7 +187,7 @@ mmhub_v2_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
static void mmhub_v2_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET_RLC(MMHUB, 0, mmMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -362,7 +362,7 @@ static void mmhub_v2_0_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
int i;
uint32_t tmp;
@@ -412,7 +412,7 @@ static void mmhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v2_0_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -441,7 +441,7 @@ static int mmhub_v2_0_gart_enable(struct amdgpu_device *adev)
static void mmhub_v2_0_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -520,7 +520,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v2_0_vmhub_funcs = {
static void mmhub_v2_0_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c
index fcf2813e70db..8bd0fc8d9d25 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c
@@ -121,7 +121,7 @@ static void mmhub_v2_3_setup_vm_pt_regs(struct amdgpu_device *adev,
uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, mmMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid, lower_32_bits(page_table_base));
@@ -280,7 +280,7 @@ static void mmhub_v2_3_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v2_3_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
int i;
uint32_t tmp;
@@ -330,7 +330,7 @@ static void mmhub_v2_3_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v2_3_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -373,7 +373,7 @@ static int mmhub_v2_3_gart_enable(struct amdgpu_device *adev)
static void mmhub_v2_3_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -446,7 +446,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v2_3_vmhub_funcs = {
static void mmhub_v2_3_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c
index 17a792616979..441379e91cfa 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c
@@ -136,7 +136,7 @@ mmhub_v3_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
static void mmhub_v3_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -319,7 +319,7 @@ static void mmhub_v3_0_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
int i;
uint32_t tmp;
@@ -369,7 +369,7 @@ static void mmhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v3_0_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -398,7 +398,7 @@ static int mmhub_v3_0_gart_enable(struct amdgpu_device *adev)
static void mmhub_v3_0_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -477,7 +477,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v3_0_vmhub_funcs = {
static void mmhub_v3_0_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c
index 26509b6b8c24..12c7f4b46ea9 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c
@@ -138,7 +138,7 @@ static void mmhub_v3_0_1_setup_vm_pt_regs(struct amdgpu_device *adev,
uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -306,7 +306,7 @@ static void mmhub_v3_0_1_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v3_0_1_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
int i;
uint32_t tmp;
@@ -356,7 +356,7 @@ static void mmhub_v3_0_1_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v3_0_1_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -385,7 +385,7 @@ static int mmhub_v3_0_1_gart_enable(struct amdgpu_device *adev)
static void mmhub_v3_0_1_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -459,7 +459,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v3_0_1_vmhub_funcs = {
static void mmhub_v3_0_1_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c
index 26abbc6a47ab..5dadc85abf7e 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c
@@ -129,7 +129,7 @@ mmhub_v3_0_2_print_l2_protection_fault_status(struct amdgpu_device *adev,
static void mmhub_v3_0_2_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -311,7 +311,7 @@ static void mmhub_v3_0_2_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v3_0_2_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
int i;
uint32_t tmp;
@@ -361,7 +361,7 @@ static void mmhub_v3_0_2_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v3_0_2_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -390,7 +390,7 @@ static int mmhub_v3_0_2_gart_enable(struct amdgpu_device *adev)
static void mmhub_v3_0_2_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -469,7 +469,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v3_0_2_vmhub_funcs = {
static void mmhub_v3_0_2_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c
index 72083e96222f..e790f890aec6 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c
@@ -57,7 +57,7 @@ static u64 mmhub_v9_4_get_fb_location(struct amdgpu_device *adev)
static void mmhub_v9_4_setup_hubid_vm_pt_regs(struct amdgpu_device *adev, int hubid,
uint32_t vmid, uint64_t value)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0,
mmVML2VC0_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
@@ -294,7 +294,7 @@ static void mmhub_v9_4_disable_identity_aperture(struct amdgpu_device *adev,
static void mmhub_v9_4_setup_vmid_config(struct amdgpu_device *adev, int hubid)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned int num_level, block_size;
uint32_t tmp;
int i;
@@ -363,7 +363,7 @@ static void mmhub_v9_4_setup_vmid_config(struct amdgpu_device *adev, int hubid)
static void mmhub_v9_4_program_invalidation(struct amdgpu_device *adev,
int hubid)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -404,7 +404,7 @@ static int mmhub_v9_4_gart_enable(struct amdgpu_device *adev)
static void mmhub_v9_4_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i, j;
@@ -507,8 +507,8 @@ static void mmhub_v9_4_set_fault_enable_default(struct amdgpu_device *adev, bool
static void mmhub_v9_4_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub[MMHUB_NUM_INSTANCES] =
- {&adev->vmhub[AMDGPU_MMHUB_0], &adev->vmhub[AMDGPU_MMHUB_1]};
+ struct amdgpu_vmhub *hub[MMHUB_NUM_INSTANCES] = {
+ &adev->vmhub[AMDGPU_MMHUB0(0)], &adev->vmhub[AMDGPU_MMHUB1(0)]};
int i;
for (i = 0; i < MMHUB_NUM_INSTANCES; i++) {
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
index 9295ac7edd56..50b6eb9bcfda 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
@@ -1825,12 +1825,12 @@ static int sdma_v4_0_sw_init(void *handle)
/*
* On Arcturus, SDMA instance 5~7 has a different vmhub
- * type(AMDGPU_MMHUB_1).
+ * type(AMDGPU_MMHUB1).
*/
if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) && i >= 5)
- ring->vm_hub = AMDGPU_MMHUB_1;
+ ring->vm_hub = AMDGPU_MMHUB1(0);
else
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
@@ -1851,9 +1851,9 @@ static int sdma_v4_0_sw_init(void *handle)
ring->doorbell_index += 0x400;
if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) && i >= 5)
- ring->vm_hub = AMDGPU_MMHUB_1;
+ ring->vm_hub = AMDGPU_MMHUB1(0);
else
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "page%d", i);
r = amdgpu_ring_init(adev, ring, 1024,
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 64dcaa2670dd..7efe7c43fffb 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -1309,7 +1309,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
/* doorbell size is 2 dwords, get DWORD offset */
ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
@@ -1328,7 +1328,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
*/
ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
ring->doorbell_index += 0x400;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "page%d", i);
r = amdgpu_ring_init(adev, ring, 1024,
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
index 92e1299be021..a0077cf41295 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
@@ -1389,7 +1389,7 @@ static int sdma_v5_0_sw_init(void *handle)
(adev->doorbell_index.sdma_engine[0] << 1) //get DWORD offset
: (adev->doorbell_index.sdma_engine[1] << 1); // get DWORD offset
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
(i == 0) ? AMDGPU_SDMA_IRQ_INSTANCE0 :
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
index ca7e8757d78e..efa2c84ee78e 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
@@ -1253,7 +1253,7 @@ static int sdma_v5_2_sw_init(void *handle)
ring->doorbell_index =
(adev->doorbell_index.sdma_engine[i] << 1); //get DWORD offset
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
AMDGPU_SDMA_IRQ_INSTANCE0 + i,
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
index 3d9a80511a45..79d09792d2ce 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
@@ -1298,7 +1298,7 @@ static int sdma_v6_0_sw_init(void *handle)
ring->doorbell_index =
(adev->doorbell_index.sdma_engine[i] << 1); // get DWORD offset
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024,
&adev->sdma.trap_irq,
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
index e32b656b3dab..abaa4463e906 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
@@ -444,7 +444,7 @@ static int uvd_v7_0_sw_init(void *handle)
continue;
if (!amdgpu_sriov_vf(adev)) {
ring = &adev->uvd.inst[j].ring;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "uvd_%d", ring->me);
r = amdgpu_ring_init(adev, ring, 512,
&adev->uvd.inst[j].irq, 0,
@@ -455,7 +455,7 @@ static int uvd_v7_0_sw_init(void *handle)
for (i = 0; i < adev->uvd.num_enc_rings; ++i) {
ring = &adev->uvd.inst[j].ring_enc[i];
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "uvd_enc_%d.%d", ring->me, i);
if (amdgpu_sriov_vf(adev)) {
ring->use_doorbell = true;
diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c
index 57b85bb6a1e4..e0b70cd3b697 100644
--- a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c
@@ -466,7 +466,7 @@ static int vce_v4_0_sw_init(void *handle)
enum amdgpu_ring_priority_level hw_prio = amdgpu_vce_get_ring_prio(i);
ring = &adev->vce.ring[i];
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vce%d", i);
if (amdgpu_sriov_vf(adev)) {
/* DOORBELL only works under SRIOV */
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
index 761c28fa6ec1..f877c39c7cdd 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
@@ -120,7 +120,7 @@ static int vcn_v1_0_sw_init(void *handle)
return r;
ring = &adev->vcn.inst->ring_dec;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
AMDGPU_RING_PRIO_DEFAULT, NULL);
@@ -142,7 +142,7 @@ static int vcn_v1_0_sw_init(void *handle)
enum amdgpu_ring_priority_level hw_prio = amdgpu_vcn_get_enc_ring_prio(i);
ring = &adev->vcn.inst->ring_enc[i];
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_enc%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
hw_prio, NULL);
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
index 7c2b3aa48083..c975aed2f6c7 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
@@ -129,7 +129,7 @@ static int vcn_v2_0_sw_init(void *handle)
ring->use_doorbell = true;
ring->doorbell_index = adev->doorbell_index.vcn.vcn_ring0_1 << 1;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
@@ -160,7 +160,7 @@ static int vcn_v2_0_sw_init(void *handle)
ring = &adev->vcn.inst->ring_enc[i];
ring->use_doorbell = true;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
if (!amdgpu_sriov_vf(adev))
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + i;
else
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
index ab0b45d0ead1..7044bd7c9f62 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
@@ -188,9 +188,9 @@ static int vcn_v2_5_sw_init(void *handle)
(amdgpu_sriov_vf(adev) ? 2*j : 8*j);
if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
- ring->vm_hub = AMDGPU_MMHUB_1;
+ ring->vm_hub = AMDGPU_MMHUB1(0);
else
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_dec_%d", j);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[j].irq,
@@ -208,9 +208,9 @@ static int vcn_v2_5_sw_init(void *handle)
(amdgpu_sriov_vf(adev) ? (1 + i + 2*j) : (2 + i + 8*j));
if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
- ring->vm_hub = AMDGPU_MMHUB_1;
+ ring->vm_hub = AMDGPU_MMHUB1(0);
else
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_enc_%d.%d", j, i);
r = amdgpu_ring_init(adev, ring, 512,
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
index 3eab186261aa..70fefbf26c48 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
@@ -189,7 +189,7 @@ static int vcn_v3_0_sw_init(void *handle)
} else {
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 8 * i;
}
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_dec_%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
AMDGPU_RING_PRIO_DEFAULT,
@@ -213,7 +213,7 @@ static int vcn_v3_0_sw_init(void *handle)
} else {
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + j + 8 * i;
}
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_enc_%d.%d", i, j);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
hw_prio, &adev->vcn.inst[i].sched_score);
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
index bf0674039598..81446e6996df 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
@@ -149,7 +149,7 @@ static int vcn_v4_0_sw_init(void *handle)
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + i * (adev->vcn.num_enc_rings + 1) + 1;
else
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + 8 * i;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_unified_%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
--
cgit v1.2.3
From d9426c3d9b4e91dda4f1f1684f9296762fafe0de Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Mon, 20 Dec 2021 16:06:25 +0800
Subject: drm/amdgpu: add bitmask to iterate vmhubs
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
As the layout of VMHUB definition has been changed to cover multiple
XCD/AID case, the original num_vmhubs is not appropriate to do vmhub
iteration any more.
Drop num_vmhubs and introduce vmhubs_mask instead.
v2: switch to the new VMHUB layout
v3: use DECLARE_BITMAP to define vmhubs_mask
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c | 4 ++--
drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 5 +++--
drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c | 5 +++--
drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 21 ++++++++++++---------
9 files changed, 25 insertions(+), 20 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 957b18bda4a7..0f163d266812 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -829,7 +829,7 @@ struct amdgpu_device {
dma_addr_t dummy_page_addr;
struct amdgpu_vm_manager vm_manager;
struct amdgpu_vmhub vmhub[AMDGPU_MAX_VMHUBS];
- unsigned num_vmhubs;
+ DECLARE_BITMAP(vmhubs_mask, AMDGPU_MAX_VMHUBS);
/* memory management */
struct amdgpu_mman mman;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index f0a136d35279..5afbcc390d89 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -733,7 +733,7 @@ int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
if (adev->family == AMDGPU_FAMILY_AI) {
int i;
- for (i = 0; i < adev->num_vmhubs; i++)
+ for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS)
amdgpu_gmc_flush_gpu_tlb(adev, vmid, i, 0);
} else {
amdgpu_gmc_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB(0), 0);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
index 01cb89ffbd56..6b12f4a75fc3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
@@ -182,7 +182,7 @@ void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
}
mb();
amdgpu_device_flush_hdp(adev, NULL);
- for (i = 0; i < adev->num_vmhubs; i++)
+ for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS)
amdgpu_gmc_flush_gpu_tlb(adev, 0, i, 0);
drm_dev_exit(idx);
@@ -264,7 +264,7 @@ void amdgpu_gart_invalidate_tlb(struct amdgpu_device *adev)
mb();
amdgpu_device_flush_hdp(adev, NULL);
- for (i = 0; i < adev->num_vmhubs; i++)
+ for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS)
amdgpu_gmc_flush_gpu_tlb(adev, 0, i, 0);
}
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
index ea2a448147e3..ff96f11c2adf 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
@@ -460,7 +460,7 @@ static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
&queried_pasid);
if (ret && queried_pasid == pasid) {
if (all_hub) {
- for (i = 0; i < adev->num_vmhubs; i++)
+ for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS)
gmc_v10_0_flush_gpu_tlb(adev, vmid,
i, flush_type);
} else {
@@ -928,7 +928,8 @@ static int gmc_v10_0_sw_init(void *handle)
case IP_VERSION(10, 3, 6):
case IP_VERSION(10, 3, 3):
case IP_VERSION(10, 3, 7):
- adev->num_vmhubs = 2;
+ set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+ set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
/*
* To fulfill 4-level page support,
* vm size is 256TB (48bit), maximum size of Navi10/Navi14/Navi12,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
index fb2ac31cbba7..3453f1c0e066 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
@@ -364,7 +364,7 @@ static int gmc_v11_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
&queried_pasid);
if (ret && queried_pasid == pasid) {
if (all_hub) {
- for (i = 0; i < adev->num_vmhubs; i++)
+ for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS)
gmc_v11_0_flush_gpu_tlb(adev, vmid,
i, flush_type);
} else {
@@ -779,7 +779,8 @@ static int gmc_v11_0_sw_init(void *handle)
case IP_VERSION(11, 0, 2):
case IP_VERSION(11, 0, 3):
case IP_VERSION(11, 0, 4):
- adev->num_vmhubs = 2;
+ set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+ set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
/*
* To fulfill 4-level page support,
* vm size is 256TB (48bit), maximum size,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
index b7dad4e67813..aa754c95a0b3 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
@@ -808,7 +808,7 @@ static int gmc_v6_0_sw_init(void *handle)
int r;
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
- adev->num_vmhubs = 1;
+ set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
if (adev->flags & AMD_IS_APU) {
adev->gmc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
index 402960b0174e..81609a2b226f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
@@ -977,7 +977,7 @@ static int gmc_v7_0_sw_init(void *handle)
int r;
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
- adev->num_vmhubs = 1;
+ set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
if (adev->flags & AMD_IS_APU) {
adev->gmc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
index 504c1b34dab7..d48e33738a88 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
@@ -1093,7 +1093,7 @@ static int gmc_v8_0_sw_init(void *handle)
int r;
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
- adev->num_vmhubs = 1;
+ set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
if (adev->flags & AMD_IS_APU) {
adev->gmc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 193ba4d912a6..d4bfb5f8308a 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -481,7 +481,7 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
switch (state) {
case AMDGPU_IRQ_STATE_DISABLE:
- for (j = 0; j < adev->num_vmhubs; j++) {
+ for_each_set_bit(j, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS) {
hub = &adev->vmhub[j];
for (i = 0; i < 16; i++) {
reg = hub->vm_context0_cntl + i;
@@ -509,7 +509,7 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
}
break;
case AMDGPU_IRQ_STATE_ENABLE:
- for (j = 0; j < adev->num_vmhubs; j++) {
+ for_each_set_bit(j, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS) {
hub = &adev->vmhub[j];
for (i = 0; i < 16; i++) {
reg = hub->vm_context0_cntl + i;
@@ -803,7 +803,7 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
u32 j, inv_req, inv_req2, tmp;
struct amdgpu_vmhub *hub;
- BUG_ON(vmhub >= adev->num_vmhubs);
+ BUG_ON(vmhub >= AMDGPU_MAX_VMHUBS);
hub = &adev->vmhub[vmhub];
if (adev->gmc.xgmi.num_physical_nodes &&
@@ -987,7 +987,7 @@ static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
&queried_pasid);
if (ret && queried_pasid == pasid) {
if (all_hub) {
- for (i = 0; i < adev->num_vmhubs; i++)
+ for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS)
gmc_v9_0_flush_gpu_tlb(adev, vmid,
i, flush_type);
} else {
@@ -1684,7 +1684,8 @@ static int gmc_v9_0_sw_init(void *handle)
switch (adev->ip_versions[GC_HWIP][0]) {
case IP_VERSION(9, 1, 0):
case IP_VERSION(9, 2, 2):
- adev->num_vmhubs = 2;
+ set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+ set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
if (adev->rev_id == 0x0 || adev->rev_id == 0x1) {
amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
@@ -1701,8 +1702,8 @@ static int gmc_v9_0_sw_init(void *handle)
case IP_VERSION(9, 3, 0):
case IP_VERSION(9, 4, 2):
case IP_VERSION(9, 4, 3):
- adev->num_vmhubs = 2;
-
+ set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+ set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
/*
* To fulfill 4-level page support,
@@ -1718,7 +1719,9 @@ static int gmc_v9_0_sw_init(void *handle)
adev->gmc.translate_further = adev->vm_manager.num_level > 1;
break;
case IP_VERSION(9, 4, 1):
- adev->num_vmhubs = 3;
+ set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+ set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
+ set_bit(AMDGPU_MMHUB1(0), adev->vmhubs_mask);
/* Keep the vm size same with Vega20 */
amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
@@ -1944,7 +1947,7 @@ static int gmc_v9_0_hw_init(void *handle)
adev->gfxhub.funcs->set_fault_enable_default(adev, value);
adev->mmhub.funcs->set_fault_enable_default(adev, value);
}
- for (i = 0; i < adev->num_vmhubs; ++i) {
+ for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS) {
if (adev->in_s0ix && (i == AMDGPU_GFXHUB(0)))
continue;
gmc_v9_0_flush_gpu_tlb(adev, 0, i, 0);
--
cgit v1.2.3
From f87f686482c6d2d4465245356854710b01f312c1 Mon Sep 17 00:00:00 2001
From: Mukul Joshi <mukul.joshi@amd.com>
Date: Mon, 9 May 2022 22:22:20 -0400
Subject: drm/amdgpu: Add XCC inst to PASID TLB flushing
Add XCC instance to select the correct KIQ ring when
flushing TLBs on a multi-XCC setup.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Tested-by: Amber Lin <Amber.Lin@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 6 ++++--
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 3 ++-
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h | 7 ++++---
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_v7_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 12 ++++++------
drivers/gpu/drm/amd/amdkfd/kfd_process.c | 7 +++++--
9 files changed, 25 insertions(+), 18 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 5afbcc390d89..9d19c7ceda3f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -743,7 +743,9 @@ int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
}
int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
- uint16_t pasid, enum TLB_FLUSH_TYPE flush_type)
+ uint16_t pasid,
+ enum TLB_FLUSH_TYPE flush_type,
+ uint32_t inst)
{
bool all_hub = false;
@@ -751,7 +753,7 @@ int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
adev->family == AMDGPU_FAMILY_RV)
all_hub = true;
- return amdgpu_gmc_flush_gpu_tlb_pasid(adev, pasid, flush_type, all_hub);
+ return amdgpu_gmc_flush_gpu_tlb_pasid(adev, pasid, flush_type, all_hub, inst);
}
bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 01ba3589b60a..df07e212c21e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -160,7 +160,8 @@ bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev);
int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
uint16_t vmid);
int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
- uint16_t pasid, enum TLB_FLUSH_TYPE flush_type);
+ uint16_t pasid, enum TLB_FLUSH_TYPE flush_type,
+ uint32_t inst);
bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
index 6d105d7fb98b..572ef5be539f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
@@ -119,7 +119,8 @@ struct amdgpu_gmc_funcs {
uint32_t vmhub, uint32_t flush_type);
/* flush the vm tlb via pasid */
int (*flush_gpu_tlb_pasid)(struct amdgpu_device *adev, uint16_t pasid,
- uint32_t flush_type, bool all_hub);
+ uint32_t flush_type, bool all_hub,
+ uint32_t inst);
/* flush the vm tlb via ring */
uint64_t (*emit_flush_gpu_tlb)(struct amdgpu_ring *ring, unsigned vmid,
uint64_t pd_addr);
@@ -296,9 +297,9 @@ struct amdgpu_gmc {
};
#define amdgpu_gmc_flush_gpu_tlb(adev, vmid, vmhub, type) ((adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid), (vmhub), (type)))
-#define amdgpu_gmc_flush_gpu_tlb_pasid(adev, pasid, type, allhub) \
+#define amdgpu_gmc_flush_gpu_tlb_pasid(adev, pasid, type, allhub, inst) \
((adev)->gmc.gmc_funcs->flush_gpu_tlb_pasid \
- ((adev), (pasid), (type), (allhub)))
+ ((adev), (pasid), (type), (allhub), (inst)))
#define amdgpu_gmc_emit_flush_gpu_tlb(r, vmid, addr) (r)->adev->gmc.gmc_funcs->emit_flush_gpu_tlb((r), (vmid), (addr))
#define amdgpu_gmc_emit_pasid_mapping(r, vmid, pasid) (r)->adev->gmc.gmc_funcs->emit_pasid_mapping((r), (vmid), (pasid))
#define amdgpu_gmc_map_mtype(adev, flags) (adev)->gmc.gmc_funcs->map_mtype((adev),(flags))
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
index ff96f11c2adf..d76f5c8d4977 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
@@ -419,7 +419,7 @@ error_alloc:
*/
static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
uint16_t pasid, uint32_t flush_type,
- bool all_hub)
+ bool all_hub, uint32_t inst)
{
int vmid, i;
signed long r;
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
index 3453f1c0e066..4bf807d825c0 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
@@ -324,7 +324,7 @@ static void gmc_v11_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
*/
static int gmc_v11_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
uint16_t pasid, uint32_t flush_type,
- bool all_hub)
+ bool all_hub, uint32_t inst)
{
int vmid, i;
signed long r;
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
index 81609a2b226f..6f53049619cd 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
@@ -424,7 +424,7 @@ static int gmc_v7_0_mc_init(struct amdgpu_device *adev)
*/
static int gmc_v7_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
uint16_t pasid, uint32_t flush_type,
- bool all_hub)
+ bool all_hub, uint32_t inst)
{
int vmid;
unsigned int tmp;
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
index d48e33738a88..48475077ca92 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
@@ -622,7 +622,7 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev)
*/
static int gmc_v8_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
uint16_t pasid, uint32_t flush_type,
- bool all_hub)
+ bool all_hub, uint32_t inst)
{
int vmid;
unsigned int tmp;
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 6da85365e5aa..0163a761ccf0 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -924,7 +924,7 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
*/
static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
uint16_t pasid, uint32_t flush_type,
- bool all_hub)
+ bool all_hub, uint32_t inst)
{
int vmid, i;
signed long r;
@@ -932,8 +932,8 @@ static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
uint16_t queried_pasid;
bool ret;
u32 usec_timeout = amdgpu_sriov_vf(adev) ? SRIOV_USEC_TIMEOUT : adev->usec_timeout;
- struct amdgpu_ring *ring = &adev->gfx.kiq[0].ring;
- struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+ struct amdgpu_ring *ring = &adev->gfx.kiq[inst].ring;
+ struct amdgpu_kiq *kiq = &adev->gfx.kiq[inst];
if (amdgpu_in_reset(adev))
return -EIO;
@@ -953,7 +953,7 @@ static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
if (vega20_xgmi_wa)
ndw += kiq->pmf->invalidate_tlbs_size;
- spin_lock(&adev->gfx.kiq[0].ring_lock);
+ spin_lock(&adev->gfx.kiq[inst].ring_lock);
/* 2 dwords flush + 8 dwords fence */
amdgpu_ring_alloc(ring, ndw);
if (vega20_xgmi_wa)
@@ -964,13 +964,13 @@ static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
if (r) {
amdgpu_ring_undo(ring);
- spin_unlock(&adev->gfx.kiq[0].ring_lock);
+ spin_unlock(&adev->gfx.kiq[inst].ring_lock);
up_read(&adev->reset_domain->sem);
return -ETIME;
}
amdgpu_ring_commit(ring);
- spin_unlock(&adev->gfx.kiq[0].ring_lock);
+ spin_unlock(&adev->gfx.kiq[inst].ring_lock);
r = amdgpu_fence_wait_polling(ring, seq, usec_timeout);
if (r < 1) {
dev_err(adev->dev, "wait for kiq fence error: %ld.\n", r);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 888590dfa646..9b1e84d33cdc 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -2052,6 +2052,7 @@ void kfd_flush_tlb(struct kfd_process_device *pdd, enum TLB_FLUSH_TYPE type)
struct amdgpu_vm *vm = drm_priv_to_vm(pdd->drm_priv);
uint64_t tlb_seq = amdgpu_vm_tlb_seq(vm);
struct kfd_node *dev = pdd->dev;
+ int xcc = 0;
/*
* It can be that we race and lose here, but that is extremely unlikely
@@ -2069,8 +2070,10 @@ void kfd_flush_tlb(struct kfd_process_device *pdd, enum TLB_FLUSH_TYPE type)
amdgpu_amdkfd_flush_gpu_tlb_vmid(dev->adev,
pdd->qpd.vmid);
} else {
- amdgpu_amdkfd_flush_gpu_tlb_pasid(dev->adev,
- pdd->process->pasid, type);
+ for (xcc = 0; xcc < dev->num_xcc_per_node; xcc++)
+ amdgpu_amdkfd_flush_gpu_tlb_pasid(dev->adev,
+ pdd->process->pasid, type,
+ dev->start_xcc_id + xcc);
}
}
--
cgit v1.2.3
From 0c7315e7d5ef9b36ca4db32ffeb34a187cbaf231 Mon Sep 17 00:00:00 2001
From: Mukul Joshi <mukul.joshi@amd.com>
Date: Fri, 10 Jun 2022 09:41:29 -0400
Subject: drm/amdkfd: Add device repartition support
GFX9.4.3 will support dynamic repartitioning of the GPU through sysfs.
Add device repartitioning support in KFD to repartition GPU from one
mode to other.
v2: squash in fix ("drm/amdkfd: Fix warning kgd2kfd_unlock_kfd defined but not used")
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 10 ++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 13 +++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 22 +++++++++++++++++++++-
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 5 +----
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 21 +++++++++++++++++++++
5 files changed, 66 insertions(+), 5 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 9d19c7ceda3f..bbbfe9ec4adf 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -773,3 +773,13 @@ bool amdgpu_amdkfd_ras_query_utcl2_poison_status(struct amdgpu_device *adev)
else
return false;
}
+
+int amdgpu_amdkfd_check_and_lock_kfd(struct amdgpu_device *adev)
+{
+ return kgd2kfd_check_and_lock_kfd();
+}
+
+void amdgpu_amdkfd_unlock_kfd(struct amdgpu_device *adev)
+{
+ kgd2kfd_unlock_kfd();
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index df07e212c21e..d1d643a050a1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -151,6 +151,8 @@ void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev,
void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev);
void amdgpu_amdkfd_device_init(struct amdgpu_device *adev);
void amdgpu_amdkfd_device_fini_sw(struct amdgpu_device *adev);
+int amdgpu_amdkfd_check_and_lock_kfd(struct amdgpu_device *adev);
+void amdgpu_amdkfd_unlock_kfd(struct amdgpu_device *adev);
int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
enum kgd_engine_type engine,
uint32_t vmid, uint64_t gpu_addr,
@@ -373,6 +375,8 @@ int kgd2kfd_post_reset(struct kfd_dev *kfd);
void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry);
void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd);
void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask);
+int kgd2kfd_check_and_lock_kfd(void);
+void kgd2kfd_unlock_kfd(void);
#else
static inline int kgd2kfd_init(void)
{
@@ -438,5 +442,14 @@ static inline
void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask)
{
}
+
+static inline int kgd2kfd_check_and_lock_kfd(void)
+{
+ return 0;
+}
+
+static inline void kgd2kfd_unlock_kfd(void)
+{
+}
#endif
#endif /* AMDGPU_AMDKFD_H_INCLUDED */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index 069b259f384c..69bac5b801ce 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -1233,10 +1233,30 @@ static ssize_t amdgpu_gfx_set_compute_partition(struct device *dev,
return -EINVAL;
}
+ if (!adev->kfd.init_complete)
+ return -EPERM;
+
mutex_lock(&adev->gfx.partition_mutex);
- ret = adev->gfx.funcs->switch_partition_mode(adev, mode);
+ if (mode == adev->gfx.funcs->query_partition_mode(adev))
+ goto out;
+
+ ret = amdgpu_amdkfd_check_and_lock_kfd(adev);
+ if (ret)
+ goto out;
+
+ amdgpu_amdkfd_device_fini_sw(adev);
+
+ adev->gfx.funcs->switch_partition_mode(adev, mode);
+
+ amdgpu_amdkfd_device_probe(adev);
+ amdgpu_amdkfd_device_init(adev);
+ /* If KFD init failed, return failure */
+ if (!adev->kfd.init_complete)
+ ret = -EIO;
+ amdgpu_amdkfd_unlock_kfd(adev);
+out:
mutex_unlock(&adev->gfx.partition_mutex);
if (ret)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index c776fc5884de..47d8ac64e877 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -675,7 +675,7 @@ static void gfx_v9_4_3_select_me_pipe_q(struct amdgpu_device *adev,
static enum amdgpu_gfx_partition
gfx_v9_4_3_query_compute_partition(struct amdgpu_device *adev)
{
- enum amdgpu_gfx_partition mode = AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
+ enum amdgpu_gfx_partition mode = adev->gfx.partition_mode;
if (adev->nbio.funcs->get_compute_partition_mode)
mode = adev->nbio.funcs->get_compute_partition_mode(adev);
@@ -689,9 +689,6 @@ static int gfx_v9_4_3_switch_compute_partition(struct amdgpu_device *adev,
u32 tmp = 0;
int num_xcc_per_partition, i, num_xcc;
- if (mode == adev->gfx.partition_mode)
- return mode;
-
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
switch (mode) {
case AMDGPU_SPX_PARTITION_MODE:
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index eb2b44fddf74..293787290e36 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -1356,6 +1356,27 @@ unsigned int kfd_get_num_xgmi_sdma_engines(struct kfd_node *node)
kfd_get_num_sdma_engines(node);
}
+int kgd2kfd_check_and_lock_kfd(void)
+{
+ mutex_lock(&kfd_processes_mutex);
+ if (!hash_empty(kfd_processes_table) || kfd_is_locked()) {
+ mutex_unlock(&kfd_processes_mutex);
+ return -EBUSY;
+ }
+
+ ++kfd_locked;
+ mutex_unlock(&kfd_processes_mutex);
+
+ return 0;
+}
+
+void kgd2kfd_unlock_kfd(void)
+{
+ mutex_lock(&kfd_processes_mutex);
+ --kfd_locked;
+ mutex_unlock(&kfd_processes_mutex);
+}
+
#if defined(CONFIG_DEBUG_FS)
/* This function will send a package to HIQ to hang the HWS
--
cgit v1.2.3
From 2fa9ff25de08e598af051c76b216d2f073b2ee89 Mon Sep 17 00:00:00 2001
From: Philip Yang <Philip.Yang@amd.com>
Date: Thu, 9 Mar 2023 19:30:02 -0500
Subject: drm/amdgpu: KFD graphics interop support compute partition
kfd_ioctl_get_dmabuf use the amdgpu bo xcp_id to get the gpu_id of the
KFD node from the exported dmabuf_adev, and then create kfd bo on the
correct adev and KFD node when importing the amdgpu bo to KFD.
Remove function kfd_device_by_adev, it is not needed as it is the same
result as dmabuf_adev->kfd.dev->nodes[0]->id.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 4 +++-
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 2 +-
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 14 ++++++--------
drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 1 -
drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 18 ------------------
5 files changed, 10 insertions(+), 29 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index bbbfe9ec4adf..00edb13d2124 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -498,7 +498,7 @@ int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
struct amdgpu_device **dmabuf_adev,
uint64_t *bo_size, void *metadata_buffer,
size_t buffer_size, uint32_t *metadata_size,
- uint32_t *flags)
+ uint32_t *flags, int8_t *xcp_id)
{
struct dma_buf *dma_buf;
struct drm_gem_object *obj;
@@ -542,6 +542,8 @@ int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
if (bo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)
*flags |= KFD_IOC_ALLOC_MEM_FLAGS_PUBLIC;
}
+ if (xcp_id)
+ *xcp_id = bo->xcp_id;
out_put:
dma_buf_put(dma_buf);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 05c54776951b..4e6221bccffe 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -241,7 +241,7 @@ int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
struct amdgpu_device **dmabuf_adev,
uint64_t *bo_size, void *metadata_buffer,
size_t buffer_size, uint32_t *metadata_size,
- uint32_t *flags);
+ uint32_t *flags, int8_t *xcp_id);
uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst,
struct amdgpu_device *src);
int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index f85ac4dbc673..fcad90d53c9b 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -1499,6 +1499,7 @@ static int kfd_ioctl_get_dmabuf_info(struct file *filep,
struct amdgpu_device *dmabuf_adev;
void *metadata_buffer = NULL;
uint32_t flags;
+ int8_t xcp_id;
unsigned int i;
int r;
@@ -1519,17 +1520,14 @@ static int kfd_ioctl_get_dmabuf_info(struct file *filep,
r = amdgpu_amdkfd_get_dmabuf_info(dev->adev, args->dmabuf_fd,
&dmabuf_adev, &args->size,
metadata_buffer, args->metadata_size,
- &args->metadata_size, &flags);
+ &args->metadata_size, &flags, &xcp_id);
if (r)
goto exit;
- /* Reverse-lookup gpu_id from kgd pointer */
- dev = kfd_device_by_adev(dmabuf_adev);
- if (!dev) {
- r = -EINVAL;
- goto exit;
- }
- args->gpu_id = dev->id;
+ if (xcp_id >= 0)
+ args->gpu_id = dmabuf_adev->kfd.dev->nodes[xcp_id]->id;
+ else
+ args->gpu_id = dmabuf_adev->kfd.dev->nodes[0]->id;
args->flags = flags;
/* Copy metadata buffer to user mode */
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 214d950f948e..44f4d5509db6 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -1068,7 +1068,6 @@ struct kfd_topology_device *kfd_topology_device_by_proximity_domain_no_lock(
struct kfd_topology_device *kfd_topology_device_by_id(uint32_t gpu_id);
struct kfd_node *kfd_device_by_id(uint32_t gpu_id);
struct kfd_node *kfd_device_by_pci_dev(const struct pci_dev *pdev);
-struct kfd_node *kfd_device_by_adev(const struct amdgpu_device *adev);
static inline bool kfd_irq_is_from_node(struct kfd_node *node, uint32_t node_id,
uint32_t vmid)
{
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index a8e25aecf839..dbb6159344b3 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -125,24 +125,6 @@ struct kfd_node *kfd_device_by_pci_dev(const struct pci_dev *pdev)
return device;
}
-struct kfd_node *kfd_device_by_adev(const struct amdgpu_device *adev)
-{
- struct kfd_topology_device *top_dev;
- struct kfd_node *device = NULL;
-
- down_read(&topology_lock);
-
- list_for_each_entry(top_dev, &topology_device_list, list)
- if (top_dev->gpu && top_dev->gpu->adev == adev) {
- device = top_dev->gpu;
- break;
- }
-
- up_read(&topology_lock);
-
- return device;
-}
-
/* Called with write topology_lock acquired */
static void kfd_release_topology_device(struct kfd_topology_device *dev)
{
--
cgit v1.2.3
From 315e29eca57f85107cc6f687c2d510aa532fb3f0 Mon Sep 17 00:00:00 2001
From: Mukul Joshi <mukul.joshi@amd.com>
Date: Mon, 20 Mar 2023 11:21:38 -0400
Subject: drm/amdkfd: Move local_mem_info to kfd_node
We need to track memory usage on a per partition basis. To do
that, store the local memory information in KFD node instead
of kfd device.
v2: squash in fix ("amdkfd: Use mem_id to access mem_partition info")
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 17 +++++++++++++----
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 12 +++++++-----
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 7 ++++---
drivers/gpu/drm/amd/amdkfd/kfd_crat.c | 2 +-
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 7 +++++--
drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 3 ++-
drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 7 ++++---
7 files changed, 36 insertions(+), 19 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 00edb13d2124..85df73f2c85e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -428,14 +428,23 @@ uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
}
void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
- struct kfd_local_mem_info *mem_info)
+ struct kfd_local_mem_info *mem_info,
+ uint8_t xcp_id)
{
memset(mem_info, 0, sizeof(*mem_info));
- mem_info->local_mem_size_public = adev->gmc.visible_vram_size;
- mem_info->local_mem_size_private = adev->gmc.real_vram_size -
+ if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3)) {
+ if (adev->gmc.real_vram_size == adev->gmc.visible_vram_size)
+ mem_info->local_mem_size_public =
+ KFD_XCP_MEMORY_SIZE(adev, xcp_id);
+ else
+ mem_info->local_mem_size_private =
+ KFD_XCP_MEMORY_SIZE(adev, xcp_id);
+ } else {
+ mem_info->local_mem_size_public = adev->gmc.visible_vram_size;
+ mem_info->local_mem_size_private = adev->gmc.real_vram_size -
adev->gmc.visible_vram_size;
-
+ }
mem_info->vram_width = adev->gmc.vram_width;
pr_debug("Address base: %pap public 0x%llx private 0x%llx\n",
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 4e6221bccffe..4bf6f5659568 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -231,7 +231,8 @@ int amdgpu_amdkfd_remove_gws_from_process(void *info, void *mem);
uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
enum kgd_engine_type type);
void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
- struct kfd_local_mem_info *mem_info);
+ struct kfd_local_mem_info *mem_info,
+ uint8_t xcp_id);
uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev);
uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev);
@@ -334,10 +335,11 @@ void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
((adev)->xcp_mgr && (xcp_id) >= 0 ?\
(adev)->xcp_mgr->xcp[(xcp_id)].mem_id : -1)
-#define KFD_XCP_MEMORY_SIZE(n) ((n)->adev->gmc.num_mem_partitions ?\
- (n)->adev->gmc.mem_partitions[(n)->xcp->mem_id].size /\
- (n)->adev->xcp_mgr->num_xcp_per_mem_partition :\
- (n)->adev->gmc.real_vram_size)
+#define KFD_XCP_MEMORY_SIZE(adev, xcp_id)\
+ ((adev)->gmc.num_mem_partitions && (xcp_id) >= 0 ?\
+ (adev)->gmc.mem_partitions[KFD_XCP_MEM_ID((adev), (xcp_id))].size /\
+ (adev)->xcp_mgr->num_xcp_per_mem_partition :\
+ (adev)->gmc.real_vram_size)
#if IS_ENABLED(CONFIG_HSA_AMD)
void amdgpu_amdkfd_gpuvm_init_mem_limits(void);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index fcad90d53c9b..1ae867482bc7 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -1023,11 +1023,12 @@ bool kfd_dev_is_large_bar(struct kfd_node *dev)
if (dev->kfd->use_iommu_v2)
return false;
- if (dev->kfd->local_mem_info.local_mem_size_private == 0 &&
- dev->kfd->local_mem_info.local_mem_size_public > 0)
+ if (dev->local_mem_info.local_mem_size_private == 0 &&
+ dev->local_mem_info.local_mem_size_public > 0)
return true;
- if (dev->kfd->local_mem_info.local_mem_size_public == 0 && dev->kfd->adev->gmc.is_app_apu) {
+ if (dev->local_mem_info.local_mem_size_public == 0 &&
+ dev->kfd->adev->gmc.is_app_apu) {
pr_debug("APP APU, Consider like a large bar system\n");
return true;
}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
index 1aaf933f9f48..950af6820153 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
@@ -2191,7 +2191,7 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
* report the total FB size (public+private) as a single
* private heap.
*/
- local_mem_info = kdev->kfd->local_mem_info;
+ local_mem_info = kdev->local_mem_info;
sub_type_hdr = (typeof(sub_type_hdr))((char *)sub_type_hdr +
sub_type_hdr->length);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index db5b53fcdf11..d41da964d2f5 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -726,7 +726,6 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
svm_migrate_init(kfd->adev);
- amdgpu_amdkfd_get_local_mem_info(kfd->adev, &kfd->local_mem_info);
dev_info(kfd_device, "Total number of KFD nodes to be created: %d\n",
kfd->num_nodes);
@@ -756,7 +755,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
if (node->xcp) {
dev_info(kfd_device, "KFD node %d partition %d size %lldM\n",
node->node_id, node->xcp->mem_id,
- KFD_XCP_MEMORY_SIZE(node) >> 20);
+ KFD_XCP_MEMORY_SIZE(node->adev, node->node_id) >> 20);
}
if (KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 3) &&
@@ -783,6 +782,10 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
}
node->max_proc_per_quantum = max_proc_per_quantum;
atomic_set(&node->sram_ecc_flag, 0);
+
+ amdgpu_amdkfd_get_local_mem_info(kfd->adev,
+ &node->local_mem_info, node->xcp->id);
+
/* Initialize the KFD node */
if (kfd_init_node(node)) {
dev_err(kfd_device, "Error initializing KFD node\n");
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 44f4d5509db6..3bd222e8f6c3 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -313,6 +313,8 @@ struct kfd_node {
unsigned int compute_vmid_bitmap;
+ struct kfd_local_mem_info local_mem_info;
+
struct kfd_dev *kfd;
};
@@ -335,7 +337,6 @@ struct kfd_dev {
*/
struct kgd2kfd_shared_resources shared_resources;
- struct kfd_local_mem_info local_mem_info;
const struct kfd2kgd_calls *kfd2kgd;
struct mutex doorbell_mutex;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index dbb6159344b3..e0bacf017a40 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -1152,8 +1152,8 @@ static uint32_t kfd_generate_gpu_id(struct kfd_node *gpu)
if (!gpu)
return 0;
- local_mem_size = gpu->kfd->local_mem_info.local_mem_size_private +
- gpu->kfd->local_mem_info.local_mem_size_public;
+ local_mem_size = gpu->local_mem_info.local_mem_size_private +
+ gpu->local_mem_info.local_mem_size_public;
buf[0] = gpu->adev->pdev->devfn;
buf[1] = gpu->adev->pdev->subsystem_vendor |
(gpu->adev->pdev->subsystem_device << 16);
@@ -1234,7 +1234,8 @@ static void kfd_fill_mem_clk_max_info(struct kfd_topology_device *dev)
* for APUs - If CRAT from ACPI reports more than one bank, then
* all the banks will report the same mem_clk_max information
*/
- amdgpu_amdkfd_get_local_mem_info(dev->gpu->adev, &local_mem_info);
+ amdgpu_amdkfd_get_local_mem_info(dev->gpu->adev, &local_mem_info,
+ dev->gpu->xcp->id);
list_for_each_entry(mem, &dev->mem_props, list)
mem->mem_clk_max = local_mem_info.mem_clk_max;
--
cgit v1.2.3
From 9a3ce1a7a9e5372d8c275bf3fbef4456c8407145 Mon Sep 17 00:00:00 2001
From: Hawking Zhang <Hawking.Zhang@amd.com>
Date: Fri, 12 May 2023 13:22:57 +0800
Subject: drm/amdgpu: Do not access members of xcp w/o check (v2)
Not all the asic needs xcp. ensure check xcp availabity
before accessing its member.
v2: add missing change in kfd_topology.c
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Le Ma <le.ma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 8 ++++----
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 2 +-
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 2 +-
drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 2 +-
4 files changed, 7 insertions(+), 7 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 85df73f2c85e..739eb7c0d133 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -429,17 +429,17 @@ uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
struct kfd_local_mem_info *mem_info,
- uint8_t xcp_id)
+ struct amdgpu_xcp *xcp)
{
memset(mem_info, 0, sizeof(*mem_info));
- if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3)) {
+ if (xcp) {
if (adev->gmc.real_vram_size == adev->gmc.visible_vram_size)
mem_info->local_mem_size_public =
- KFD_XCP_MEMORY_SIZE(adev, xcp_id);
+ KFD_XCP_MEMORY_SIZE(adev, xcp->id);
else
mem_info->local_mem_size_private =
- KFD_XCP_MEMORY_SIZE(adev, xcp_id);
+ KFD_XCP_MEMORY_SIZE(adev, xcp->id);
} else {
mem_info->local_mem_size_public = adev->gmc.visible_vram_size;
mem_info->local_mem_size_private = adev->gmc.real_vram_size -
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 48d12dbff968..be43d71ba7ef 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -233,7 +233,7 @@ uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
enum kgd_engine_type type);
void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
struct kfd_local_mem_info *mem_info,
- uint8_t xcp_id);
+ struct amdgpu_xcp *xcp);
uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev);
uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 2cfef3f9456f..986543a000bf 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -784,7 +784,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
atomic_set(&node->sram_ecc_flag, 0);
amdgpu_amdkfd_get_local_mem_info(kfd->adev,
- &node->local_mem_info, node->xcp->id);
+ &node->local_mem_info, node->xcp);
/* Initialize the KFD node */
if (kfd_init_node(node)) {
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index e0bacf017a40..8302d8967158 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -1235,7 +1235,7 @@ static void kfd_fill_mem_clk_max_info(struct kfd_topology_device *dev)
* all the banks will report the same mem_clk_max information
*/
amdgpu_amdkfd_get_local_mem_info(dev->gpu->adev, &local_mem_info,
- dev->gpu->xcp->id);
+ dev->gpu->xcp);
list_for_each_entry(mem, &dev->mem_props, list)
mem->mem_clk_max = local_mem_info.mem_clk_max;
--
cgit v1.2.3
From 45b3a914d40e63d2c9e3a3e02fb2014be975b9b0 Mon Sep 17 00:00:00 2001
From: Alex Deucher <alexander.deucher@amd.com>
Date: Tue, 16 May 2023 17:16:30 -0400
Subject: drm/amdgpu/gmc9: fix 64 bit division in partition code
Rework logic or use do_div() to avoid problems on 32 bit.
v2: add a missing case for XCP macro
v3: fix out of bounds array access
v4: fix xcp handling harder
Acked-by: Guchun Chen <guchun.chen@amd.com> (v1)
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com> (v3)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 15 +++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 9 ++++-----
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 5 ++++-
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 11 ++++++-----
drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 8 ++++++--
5 files changed, 35 insertions(+), 13 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 739eb7c0d133..5de92c9ab18f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -794,3 +794,18 @@ void amdgpu_amdkfd_unlock_kfd(struct amdgpu_device *adev)
{
kgd2kfd_unlock_kfd();
}
+
+
+u64 amdgpu_amdkfd_xcp_memory_size(struct amdgpu_device *adev, int xcp_id)
+{
+ u64 tmp;
+ s8 mem_id = KFD_XCP_MEM_ID(adev, xcp_id);
+
+ if (adev->gmc.num_mem_partitions && xcp_id >= 0 && mem_id >= 0) {
+ tmp = adev->gmc.mem_partitions[mem_id].size;
+ do_div(tmp, adev->xcp_mgr->num_xcp_per_mem_partition);
+ return tmp;
+ } else {
+ return adev->gmc.real_vram_size;
+ }
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index be43d71ba7ef..94cc456761e5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -333,15 +333,14 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
uint64_t size, u32 alloc_flag, int8_t xcp_id);
+u64 amdgpu_amdkfd_xcp_memory_size(struct amdgpu_device *adev, int xcp_id);
+
#define KFD_XCP_MEM_ID(adev, xcp_id) \
((adev)->xcp_mgr && (xcp_id) >= 0 ?\
(adev)->xcp_mgr->xcp[(xcp_id)].mem_id : -1)
-#define KFD_XCP_MEMORY_SIZE(adev, xcp_id)\
- ((adev)->gmc.num_mem_partitions && (xcp_id) >= 0 ?\
- (adev)->gmc.mem_partitions[KFD_XCP_MEM_ID((adev), (xcp_id))].size /\
- (adev)->xcp_mgr->num_xcp_per_mem_partition :\
- (adev)->gmc.real_vram_size)
+#define KFD_XCP_MEMORY_SIZE(adev, xcp_id) amdgpu_amdkfd_xcp_memory_size((adev), (xcp_id))
+
#if IS_ENABLED(CONFIG_HSA_AMD)
void amdgpu_amdkfd_gpuvm_init_mem_limits(void);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 23101c82519a..902773ce41b5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -814,11 +814,14 @@ static void amdgpu_ttm_gart_bind_gfx9_mqd(struct amdgpu_device *adev,
struct amdgpu_ttm_tt *gtt = (void *)ttm;
uint64_t total_pages = ttm->num_pages;
int num_xcc = max(1U, adev->gfx.num_xcc_per_xcp);
- uint64_t page_idx, pages_per_xcc = total_pages / num_xcc;
+ uint64_t page_idx, pages_per_xcc;
int i;
uint64_t ctrl_flags = (flags & ~AMDGPU_PTE_MTYPE_VG10_MASK) |
AMDGPU_PTE_MTYPE_VG10(AMDGPU_MTYPE_NC);
+ pages_per_xcc = total_pages;
+ do_div(pages_per_xcc, num_xcc);
+
for (i = 0, page_idx = 0; i < num_xcc; i++, page_idx += pages_per_xcc) {
/* MQD page: use default flags */
amdgpu_gart_bind(adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 7ea80bdf8e1e..f70e666cecf2 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -1914,9 +1914,10 @@ gmc_v9_0_init_acpi_mem_ranges(struct amdgpu_device *adev,
adev->gmc.num_mem_partitions = num_ranges;
/* If there is only partition, don't use entire size */
- if (adev->gmc.num_mem_partitions == 1)
- mem_ranges[0].size =
- (mem_ranges[0].size * (mem_groups - 1) / mem_groups);
+ if (adev->gmc.num_mem_partitions == 1) {
+ mem_ranges[0].size = mem_ranges[0].size * (mem_groups - 1);
+ do_div(mem_ranges[0].size, mem_groups);
+ }
}
static void
@@ -1948,8 +1949,8 @@ gmc_v9_0_init_sw_mem_ranges(struct amdgpu_device *adev,
break;
}
- size = (adev->gmc.real_vram_size >> AMDGPU_GPU_PAGE_SHIFT) /
- adev->gmc.num_mem_partitions;
+ size = adev->gmc.real_vram_size >> AMDGPU_GPU_PAGE_SHIFT;
+ size /= adev->gmc.num_mem_partitions;
for (i = 0; i < adev->gmc.num_mem_partitions; ++i) {
mem_ranges[i].range.fpfn = start_addr;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index 206851c9e642..b0f0d31bf3e6 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -1939,10 +1939,14 @@ void svm_range_set_max_pages(struct amdgpu_device *adev)
uint64_t max_pages;
uint64_t pages, _pages;
uint64_t min_pages = 0;
- int i;
+ int i, id;
for (i = 0; i < adev->kfd.dev->num_nodes; i++) {
- pages = KFD_XCP_MEMORY_SIZE(adev, adev->kfd.dev->nodes[i]->xcp->id) >> 17;
+ if (adev->kfd.dev->nodes[i]->xcp)
+ id = adev->kfd.dev->nodes[i]->xcp->id;
+ else
+ id = -1;
+ pages = KFD_XCP_MEMORY_SIZE(adev, id) >> 17;
pages = clamp(pages, 1ULL << 9, 1ULL << 18);
pages = rounddown_pow_of_two(pages);
min_pages = min_not_zero(min_pages, pages);
--
cgit v1.2.3
From acf429dcac1440169a3b28da784cbda72f3b678b Mon Sep 17 00:00:00 2001
From: Philip Yang <Philip.Yang@amd.com>
Date: Thu, 18 May 2023 15:08:16 -0400
Subject: drm/amdkfd: Align partition memory size to page size
The compute partition memory size calculated from KFD_XCP_MEMORY_SIZE
may not align to page size if xcp_mgr->num_xcp_per_mem_partition is 6.
Change the KFD_XCP_MEMORY_SIZE macro to return page align size, so KFD
node memory size reported in sysfs is page align size, to avoid
application VRAM allocation failure because application may use the size
directly and Thunk requires the memory allocation size is page size
align.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 5de92c9ab18f..66f80b9ab0c5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -804,7 +804,7 @@ u64 amdgpu_amdkfd_xcp_memory_size(struct amdgpu_device *adev, int xcp_id)
if (adev->gmc.num_mem_partitions && xcp_id >= 0 && mem_id >= 0) {
tmp = adev->gmc.mem_partitions[mem_id].size;
do_div(tmp, adev->xcp_mgr->num_xcp_per_mem_partition);
- return tmp;
+ return ALIGN_DOWN(tmp, PAGE_SIZE);
} else {
return adev->gmc.real_vram_size;
}
--
cgit v1.2.3
From 12fb1ad70d65edc3405884792d044fa79df7244f Mon Sep 17 00:00:00 2001
From: Jonathan Kim <jonathan.kim@amd.com>
Date: Fri, 22 Apr 2022 12:26:18 -0400
Subject: drm/amdkfd: update process interrupt handling for debug events
The debugger must be notified by any debugger subscribed exception
that comes from hardware interrupts.
If a debugger session exits, any exceptions it subscribed to may still
have interrupts in the interrupt ring buffer or KGD/KFD pipeline.
To prevent a new session from inheriting stale interrupts, when a new
queue is created, open an interrupt drain and allow the IH ring to drain
from a timestamped checkpoint. Then inject a custom IV so that once
the custom IV is picked up by the KFD, it's safe to close the drain
and proceed with queue creation.
The drain must also be on debug disable as SW interrupts may still
be processed. Drain at this time and clear all the exception status.
The debugger may also not be attached nor subscibed to certain
exceptions so forward them directly to the runtime.
GFX10 also requires its own IV processing, hence the creation of
kfd_int_process_v10.c. This is because the IV from SQ interrupts are
packed into a new continguous format unlike GFX9. To make this clear,
a separate interrupting handling code file was created.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 16 +
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 2 +
drivers/gpu/drm/amd/amdkfd/Makefile | 1 +
drivers/gpu/drm/amd/amdkfd/kfd_debug.c | 84 +++++
drivers/gpu/drm/amd/amdkfd/kfd_debug.h | 6 +
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 4 +-
drivers/gpu/drm/amd/amdkfd/kfd_int_process_v10.c | 405 +++++++++++++++++++++
drivers/gpu/drm/amd/amdkfd/kfd_int_process_v11.c | 26 +-
drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c | 98 ++++-
drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 12 +
drivers/gpu/drm/amd/amdkfd/kfd_process.c | 47 +++
.../gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 4 +
12 files changed, 686 insertions(+), 19 deletions(-)
create mode 100644 drivers/gpu/drm/amd/amdkfd/kfd_int_process_v10.c
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 66f80b9ab0c5..98cd52bb005f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -777,6 +777,22 @@ void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev, bo
amdgpu_umc_poison_handler(adev, reset);
}
+int amdgpu_amdkfd_send_close_event_drain_irq(struct amdgpu_device *adev,
+ uint32_t *payload)
+{
+ int ret;
+
+ /* Device or IH ring is not ready so bail. */
+ ret = amdgpu_ih_wait_on_checkpoint_process_ts(adev, &adev->irq.ih);
+ if (ret)
+ return ret;
+
+ /* Send payload to fence KFD interrupts */
+ amdgpu_amdkfd_interrupt(adev, payload);
+
+ return 0;
+}
+
bool amdgpu_amdkfd_ras_query_utcl2_poison_status(struct amdgpu_device *adev)
{
if (adev->gfx.ras && adev->gfx.ras->query_utcl2_poison_status)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 94cc456761e5..dd740e64e6e1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -250,6 +250,8 @@ int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
struct amdgpu_device *src,
bool is_min);
int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min);
+int amdgpu_amdkfd_send_close_event_drain_irq(struct amdgpu_device *adev,
+ uint32_t *payload);
/* Read user wptr from a specified user address space with page fault
* disabled. The memory must be pinned and mapped to the hardware when
diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile
index 747754428073..2ec8f27c5366 100644
--- a/drivers/gpu/drm/amd/amdkfd/Makefile
+++ b/drivers/gpu/drm/amd/amdkfd/Makefile
@@ -53,6 +53,7 @@ AMDKFD_FILES := $(AMDKFD_PATH)/kfd_module.o \
$(AMDKFD_PATH)/kfd_events.o \
$(AMDKFD_PATH)/cik_event_interrupt.o \
$(AMDKFD_PATH)/kfd_int_process_v9.o \
+ $(AMDKFD_PATH)/kfd_int_process_v10.o \
$(AMDKFD_PATH)/kfd_int_process_v11.o \
$(AMDKFD_PATH)/kfd_smi_events.o \
$(AMDKFD_PATH)/kfd_crat.o \
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
index 17e8e9edccbf..68b657398d41 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
@@ -125,6 +125,64 @@ bool kfd_dbg_ev_raise(uint64_t event_mask,
return is_subscribed;
}
+/* set pending event queue entry from ring entry */
+bool kfd_set_dbg_ev_from_interrupt(struct kfd_node *dev,
+ unsigned int pasid,
+ uint32_t doorbell_id,
+ uint64_t trap_mask,
+ void *exception_data,
+ size_t exception_data_size)
+{
+ struct kfd_process *p;
+ bool signaled_to_debugger_or_runtime = false;
+
+ p = kfd_lookup_process_by_pasid(pasid);
+
+ if (!p)
+ return false;
+
+ if (!kfd_dbg_ev_raise(trap_mask, p, dev, doorbell_id, true,
+ exception_data, exception_data_size)) {
+ struct process_queue_manager *pqm;
+ struct process_queue_node *pqn;
+
+ if (!!(trap_mask & KFD_EC_MASK_QUEUE) &&
+ p->runtime_info.runtime_state == DEBUG_RUNTIME_STATE_ENABLED) {
+ mutex_lock(&p->mutex);
+
+ pqm = &p->pqm;
+ list_for_each_entry(pqn, &pqm->queues,
+ process_queue_list) {
+
+ if (!(pqn->q && pqn->q->device == dev &&
+ pqn->q->doorbell_id == doorbell_id))
+ continue;
+
+ kfd_send_exception_to_runtime(p, pqn->q->properties.queue_id,
+ trap_mask);
+
+ signaled_to_debugger_or_runtime = true;
+
+ break;
+ }
+
+ mutex_unlock(&p->mutex);
+ } else if (trap_mask & KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION)) {
+ kfd_dqm_evict_pasid(dev->dqm, p->pasid);
+ kfd_signal_vm_fault_event(dev, p->pasid, NULL,
+ exception_data);
+
+ signaled_to_debugger_or_runtime = true;
+ }
+ } else {
+ signaled_to_debugger_or_runtime = true;
+ }
+
+ kfd_unref_process(p);
+
+ return signaled_to_debugger_or_runtime;
+}
+
int kfd_dbg_send_exception_to_runtime(struct kfd_process *p,
unsigned int dev_id,
unsigned int queue_id,
@@ -281,6 +339,31 @@ void kfd_dbg_trap_deactivate(struct kfd_process *target, bool unwind, int unwind
kfd_dbg_set_workaround(target, false);
}
+static void kfd_dbg_clean_exception_status(struct kfd_process *target)
+{
+ struct process_queue_manager *pqm;
+ struct process_queue_node *pqn;
+ int i;
+
+ for (i = 0; i < target->n_pdds; i++) {
+ struct kfd_process_device *pdd = target->pdds[i];
+
+ kfd_process_drain_interrupts(pdd);
+
+ pdd->exception_status = 0;
+ }
+
+ pqm = &target->pqm;
+ list_for_each_entry(pqn, &pqm->queues, process_queue_list) {
+ if (!pqn->q)
+ continue;
+
+ pqn->q->properties.exception_status = 0;
+ }
+
+ target->exception_status = 0;
+}
+
int kfd_dbg_trap_disable(struct kfd_process *target)
{
if (!target->debug_trap_enabled)
@@ -304,6 +387,7 @@ int kfd_dbg_trap_disable(struct kfd_process *target)
}
target->debug_trap_enabled = false;
+ kfd_dbg_clean_exception_status(target);
kfd_unref_process(target);
return 0;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
index fca928564948..5153ccbd7fd1 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
@@ -27,6 +27,12 @@
void kfd_dbg_trap_deactivate(struct kfd_process *target, bool unwind, int unwind_count);
int kfd_dbg_trap_activate(struct kfd_process *target);
+bool kfd_set_dbg_ev_from_interrupt(struct kfd_node *dev,
+ unsigned int pasid,
+ uint32_t doorbell_id,
+ uint64_t trap_mask,
+ void *exception_data,
+ size_t exception_data_size);
bool kfd_dbg_ev_raise(uint64_t event_mask,
struct kfd_process *process, struct kfd_node *dev,
unsigned int source_id, bool use_worker,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index f0ed6e6416c3..2c36bb578633 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -140,6 +140,8 @@ static void kfd_device_info_set_event_interrupt_class(struct kfd_dev *kfd)
case IP_VERSION(9, 4, 1): /* ARCTURUS */
case IP_VERSION(9, 4, 2): /* ALDEBARAN */
case IP_VERSION(9, 4, 3): /* GC 9.4.3 */
+ kfd->device_info.event_interrupt_class = &event_interrupt_class_v9;
+ break;
case IP_VERSION(10, 3, 1): /* VANGOGH */
case IP_VERSION(10, 3, 3): /* YELLOW_CARP */
case IP_VERSION(10, 3, 6): /* GC 10.3.6 */
@@ -153,7 +155,7 @@ static void kfd_device_info_set_event_interrupt_class(struct kfd_dev *kfd)
case IP_VERSION(10, 3, 2): /* NAVY_FLOUNDER */
case IP_VERSION(10, 3, 4): /* DIMGREY_CAVEFISH */
case IP_VERSION(10, 3, 5): /* BEIGE_GOBY */
- kfd->device_info.event_interrupt_class = &event_interrupt_class_v9;
+ kfd->device_info.event_interrupt_class = &event_interrupt_class_v10;
break;
case IP_VERSION(11, 0, 0):
case IP_VERSION(11, 0, 1):
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v10.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v10.c
new file mode 100644
index 000000000000..c7991e07b6be
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v10.c
@@ -0,0 +1,405 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "kfd_events.h"
+#include "kfd_debug.h"
+#include "soc15_int.h"
+#include "kfd_device_queue_manager.h"
+
+/*
+ * GFX10 SQ Interrupts
+ *
+ * There are 3 encoding types of interrupts sourced from SQ sent as a 44-bit
+ * packet to the Interrupt Handler:
+ * Auto - Generated by the SQG (various cmd overflows, timestamps etc)
+ * Wave - Generated by S_SENDMSG through a shader program
+ * Error - HW generated errors (Illegal instructions, Memviols, EDC etc)
+ *
+ * The 44-bit packet is mapped as {context_id1[7:0],context_id0[31:0]} plus
+ * 4-bits for VMID (SOC15_VMID_FROM_IH_ENTRY) as such:
+ *
+ * - context_id1[7:6]
+ * Encoding type (0 = Auto, 1 = Wave, 2 = Error)
+ *
+ * - context_id0[24]
+ * PRIV bit indicates that Wave S_SEND or error occurred within trap
+ *
+ * - context_id0[22:0]
+ * 23-bit data with the following layout per encoding type:
+ * Auto - only context_id0[8:0] is used, which reports various interrupts
+ * generated by SQG. The rest is 0.
+ * Wave - user data sent from m0 via S_SENDMSG
+ * Error - Error type (context_id0[22:19]), Error Details (rest of bits)
+ *
+ * The other context_id bits show coordinates (SE/SH/CU/SIMD/WGP) for wave
+ * S_SENDMSG and Errors. These are 0 for Auto.
+ */
+
+enum SQ_INTERRUPT_WORD_ENCODING {
+ SQ_INTERRUPT_WORD_ENCODING_AUTO = 0x0,
+ SQ_INTERRUPT_WORD_ENCODING_INST,
+ SQ_INTERRUPT_WORD_ENCODING_ERROR,
+};
+
+enum SQ_INTERRUPT_ERROR_TYPE {
+ SQ_INTERRUPT_ERROR_TYPE_EDC_FUE = 0x0,
+ SQ_INTERRUPT_ERROR_TYPE_ILLEGAL_INST,
+ SQ_INTERRUPT_ERROR_TYPE_MEMVIOL,
+ SQ_INTERRUPT_ERROR_TYPE_EDC_FED,
+};
+
+/* SQ_INTERRUPT_WORD_AUTO_CTXID */
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE__SHIFT 0
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__WLT__SHIFT 1
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_BUF0_FULL__SHIFT 2
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_BUF1_FULL__SHIFT 3
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_UTC_ERROR__SHIFT 7
+#define SQ_INTERRUPT_WORD_AUTO_CTXID1__SE_ID__SHIFT 4
+#define SQ_INTERRUPT_WORD_AUTO_CTXID1__ENCODING__SHIFT 6
+
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_MASK 0x00000001
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__WLT_MASK 0x00000002
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_BUF0_FULL_MASK 0x00000004
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_BUF1_FULL_MASK 0x00000008
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_UTC_ERROR_MASK 0x00000080
+#define SQ_INTERRUPT_WORD_AUTO_CTXID1__SE_ID_MASK 0x030
+#define SQ_INTERRUPT_WORD_AUTO_CTXID1__ENCODING_MASK 0x0c0
+
+/* SQ_INTERRUPT_WORD_WAVE_CTXID */
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__DATA__SHIFT 0
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__SA_ID__SHIFT 23
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__PRIV__SHIFT 24
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__WAVE_ID__SHIFT 25
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__SIMD_ID__SHIFT 30
+#define SQ_INTERRUPT_WORD_WAVE_CTXID1__WGP_ID__SHIFT 0
+#define SQ_INTERRUPT_WORD_WAVE_CTXID1__SE_ID__SHIFT 4
+#define SQ_INTERRUPT_WORD_WAVE_CTXID1__ENCODING__SHIFT 6
+
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__DATA_MASK 0x000007fffff
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__SA_ID_MASK 0x0000800000
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__PRIV_MASK 0x00001000000
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__WAVE_ID_MASK 0x0003e000000
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__SIMD_ID_MASK 0x000c0000000
+#define SQ_INTERRUPT_WORD_WAVE_CTXID1__WGP_ID_MASK 0x00f
+#define SQ_INTERRUPT_WORD_WAVE_CTXID1__SE_ID_MASK 0x030
+#define SQ_INTERRUPT_WORD_WAVE_CTXID1__ENCODING_MASK 0x0c0
+
+#define KFD_CTXID0__ERR_TYPE_MASK 0x780000
+#define KFD_CTXID0__ERR_TYPE__SHIFT 19
+
+/* GFX10 SQ interrupt ENC type bit (context_id1[7:6]) for wave s_sendmsg */
+#define KFD_CONTEXT_ID1_ENC_TYPE_WAVE_MASK 0x40
+/* GFX10 SQ interrupt PRIV bit (context_id0[24]) for s_sendmsg inside trap */
+#define KFD_CONTEXT_ID0_PRIV_MASK 0x1000000
+/*
+ * The debugger will send user data(m0) with PRIV=1 to indicate it requires
+ * notification from the KFD with the following queue id (DOORBELL_ID) and
+ * trap code (TRAP_CODE).
+ */
+#define KFD_CONTEXT_ID0_DEBUG_DOORBELL_MASK 0x0003ff
+#define KFD_CONTEXT_ID0_DEBUG_TRAP_CODE_SHIFT 10
+#define KFD_CONTEXT_ID0_DEBUG_TRAP_CODE_MASK 0x07fc00
+#define KFD_DEBUG_DOORBELL_ID(ctxid0) ((ctxid0) & \
+ KFD_CONTEXT_ID0_DEBUG_DOORBELL_MASK)
+#define KFD_DEBUG_TRAP_CODE(ctxid0) (((ctxid0) & \
+ KFD_CONTEXT_ID0_DEBUG_TRAP_CODE_MASK) \
+ >> KFD_CONTEXT_ID0_DEBUG_TRAP_CODE_SHIFT)
+#define KFD_DEBUG_CP_BAD_OP_ECODE_MASK 0x3fffc00
+#define KFD_DEBUG_CP_BAD_OP_ECODE_SHIFT 10
+#define KFD_DEBUG_CP_BAD_OP_ECODE(ctxid0) (((ctxid0) & \
+ KFD_DEBUG_CP_BAD_OP_ECODE_MASK) \
+ >> KFD_DEBUG_CP_BAD_OP_ECODE_SHIFT)
+
+static void event_interrupt_poison_consumption(struct kfd_node *dev,
+ uint16_t pasid, uint16_t client_id)
+{
+ int old_poison, ret = -EINVAL;
+ struct kfd_process *p = kfd_lookup_process_by_pasid(pasid);
+
+ if (!p)
+ return;
+
+ /* all queues of a process will be unmapped in one time */
+ old_poison = atomic_cmpxchg(&p->poison, 0, 1);
+ kfd_unref_process(p);
+ if (old_poison)
+ return;
+
+ switch (client_id) {
+ case SOC15_IH_CLIENTID_SE0SH:
+ case SOC15_IH_CLIENTID_SE1SH:
+ case SOC15_IH_CLIENTID_SE2SH:
+ case SOC15_IH_CLIENTID_SE3SH:
+ case SOC15_IH_CLIENTID_UTCL2:
+ ret = kfd_dqm_evict_pasid(dev->dqm, pasid);
+ break;
+ case SOC15_IH_CLIENTID_SDMA0:
+ case SOC15_IH_CLIENTID_SDMA1:
+ case SOC15_IH_CLIENTID_SDMA2:
+ case SOC15_IH_CLIENTID_SDMA3:
+ case SOC15_IH_CLIENTID_SDMA4:
+ break;
+ default:
+ break;
+ }
+
+ kfd_signal_poison_consumed_event(dev, pasid);
+
+ /* resetting queue passes, do page retirement without gpu reset
+ * resetting queue fails, fallback to gpu reset solution
+ */
+ if (!ret) {
+ dev_warn(dev->adev->dev,
+ "RAS poison consumption, unmap queue flow succeeded: client id %d\n",
+ client_id);
+ amdgpu_amdkfd_ras_poison_consumption_handler(dev->adev, false);
+ } else {
+ dev_warn(dev->adev->dev,
+ "RAS poison consumption, fall back to gpu reset flow: client id %d\n",
+ client_id);
+ amdgpu_amdkfd_ras_poison_consumption_handler(dev->adev, true);
+ }
+}
+
+static bool event_interrupt_isr_v10(struct kfd_node *dev,
+ const uint32_t *ih_ring_entry,
+ uint32_t *patched_ihre,
+ bool *patched_flag)
+{
+ uint16_t source_id, client_id, pasid, vmid;
+ const uint32_t *data = ih_ring_entry;
+
+ source_id = SOC15_SOURCE_ID_FROM_IH_ENTRY(ih_ring_entry);
+ client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry);
+
+ /* Only handle interrupts from KFD VMIDs */
+ vmid = SOC15_VMID_FROM_IH_ENTRY(ih_ring_entry);
+ if (!KFD_IRQ_IS_FENCE(client_id, source_id) &&
+ (vmid < dev->vm_info.first_vmid_kfd ||
+ vmid > dev->vm_info.last_vmid_kfd))
+ return false;
+
+ pasid = SOC15_PASID_FROM_IH_ENTRY(ih_ring_entry);
+
+ /* Only handle clients we care about */
+ if (client_id != SOC15_IH_CLIENTID_GRBM_CP &&
+ client_id != SOC15_IH_CLIENTID_SDMA0 &&
+ client_id != SOC15_IH_CLIENTID_SDMA1 &&
+ client_id != SOC15_IH_CLIENTID_SDMA2 &&
+ client_id != SOC15_IH_CLIENTID_SDMA3 &&
+ client_id != SOC15_IH_CLIENTID_SDMA4 &&
+ client_id != SOC15_IH_CLIENTID_SDMA5 &&
+ client_id != SOC15_IH_CLIENTID_SDMA6 &&
+ client_id != SOC15_IH_CLIENTID_SDMA7 &&
+ client_id != SOC15_IH_CLIENTID_VMC &&
+ client_id != SOC15_IH_CLIENTID_VMC1 &&
+ client_id != SOC15_IH_CLIENTID_UTCL2 &&
+ client_id != SOC15_IH_CLIENTID_SE0SH &&
+ client_id != SOC15_IH_CLIENTID_SE1SH &&
+ client_id != SOC15_IH_CLIENTID_SE2SH &&
+ client_id != SOC15_IH_CLIENTID_SE3SH)
+ return false;
+
+ pr_debug("client id 0x%x, source id %d, vmid %d, pasid 0x%x. raw data:\n",
+ client_id, source_id, vmid, pasid);
+ pr_debug("%8X, %8X, %8X, %8X, %8X, %8X, %8X, %8X.\n",
+ data[0], data[1], data[2], data[3],
+ data[4], data[5], data[6], data[7]);
+
+ /* If there is no valid PASID, it's likely a bug */
+ if (WARN_ONCE(pasid == 0, "Bug: No PASID in KFD interrupt"))
+ return 0;
+
+ /* Interrupt types we care about: various signals and faults.
+ * They will be forwarded to a work queue (see below).
+ */
+ return source_id == SOC15_INTSRC_CP_END_OF_PIPE ||
+ source_id == SOC15_INTSRC_SDMA_TRAP ||
+ source_id == SOC15_INTSRC_SQ_INTERRUPT_MSG ||
+ source_id == SOC15_INTSRC_CP_BAD_OPCODE ||
+ client_id == SOC15_IH_CLIENTID_VMC ||
+ client_id == SOC15_IH_CLIENTID_VMC1 ||
+ client_id == SOC15_IH_CLIENTID_UTCL2 ||
+ KFD_IRQ_IS_FENCE(client_id, source_id);
+}
+
+static void event_interrupt_wq_v10(struct kfd_node *dev,
+ const uint32_t *ih_ring_entry)
+{
+ uint16_t source_id, client_id, pasid, vmid;
+ uint32_t context_id0, context_id1;
+ uint32_t encoding, sq_intr_err_type;
+
+ source_id = SOC15_SOURCE_ID_FROM_IH_ENTRY(ih_ring_entry);
+ client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry);
+ pasid = SOC15_PASID_FROM_IH_ENTRY(ih_ring_entry);
+ vmid = SOC15_VMID_FROM_IH_ENTRY(ih_ring_entry);
+ context_id0 = SOC15_CONTEXT_ID0_FROM_IH_ENTRY(ih_ring_entry);
+ context_id1 = SOC15_CONTEXT_ID1_FROM_IH_ENTRY(ih_ring_entry);
+
+ if (client_id == SOC15_IH_CLIENTID_GRBM_CP ||
+ client_id == SOC15_IH_CLIENTID_SE0SH ||
+ client_id == SOC15_IH_CLIENTID_SE1SH ||
+ client_id == SOC15_IH_CLIENTID_SE2SH ||
+ client_id == SOC15_IH_CLIENTID_SE3SH) {
+ if (source_id == SOC15_INTSRC_CP_END_OF_PIPE)
+ kfd_signal_event_interrupt(pasid, context_id0, 32);
+ else if (source_id == SOC15_INTSRC_SQ_INTERRUPT_MSG) {
+ encoding = REG_GET_FIELD(context_id1,
+ SQ_INTERRUPT_WORD_WAVE_CTXID1, ENCODING);
+ switch (encoding) {
+ case SQ_INTERRUPT_WORD_ENCODING_AUTO:
+ pr_debug(
+ "sq_intr: auto, se %d, ttrace %d, wlt %d, ttrac_buf0_full %d, ttrac_buf1_full %d, ttrace_utc_err %d\n",
+ REG_GET_FIELD(context_id1, SQ_INTERRUPT_WORD_AUTO_CTXID1,
+ SE_ID),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+ THREAD_TRACE),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+ WLT),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+ THREAD_TRACE_BUF0_FULL),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+ THREAD_TRACE_BUF1_FULL),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+ THREAD_TRACE_UTC_ERROR));
+ break;
+ case SQ_INTERRUPT_WORD_ENCODING_INST:
+ pr_debug("sq_intr: inst, se %d, data 0x%x, sa %d, priv %d, wave_id %d, simd_id %d, wgp_id %d\n",
+ REG_GET_FIELD(context_id1, SQ_INTERRUPT_WORD_WAVE_CTXID1,
+ SE_ID),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0,
+ DATA),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0,
+ SA_ID),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0,
+ PRIV),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0,
+ WAVE_ID),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0,
+ SIMD_ID),
+ REG_GET_FIELD(context_id1, SQ_INTERRUPT_WORD_WAVE_CTXID1,
+ WGP_ID));
+ if (context_id0 & SQ_INTERRUPT_WORD_WAVE_CTXID0__PRIV_MASK) {
+ if (kfd_set_dbg_ev_from_interrupt(dev, pasid,
+ KFD_DEBUG_DOORBELL_ID(context_id0),
+ KFD_DEBUG_TRAP_CODE(context_id0),
+ NULL, 0))
+ return;
+ }
+ break;
+ case SQ_INTERRUPT_WORD_ENCODING_ERROR:
+ sq_intr_err_type = REG_GET_FIELD(context_id0, KFD_CTXID0,
+ ERR_TYPE);
+ pr_warn("sq_intr: error, se %d, data 0x%x, sa %d, priv %d, wave_id %d, simd_id %d, wgp_id %d, err_type %d\n",
+ REG_GET_FIELD(context_id1, SQ_INTERRUPT_WORD_WAVE_CTXID1,
+ SE_ID),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0,
+ DATA),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0,
+ SA_ID),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0,
+ PRIV),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0,
+ WAVE_ID),
+ REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0,
+ SIMD_ID),
+ REG_GET_FIELD(context_id1, SQ_INTERRUPT_WORD_WAVE_CTXID1,
+ WGP_ID),
+ sq_intr_err_type);
+ if (sq_intr_err_type != SQ_INTERRUPT_ERROR_TYPE_ILLEGAL_INST &&
+ sq_intr_err_type != SQ_INTERRUPT_ERROR_TYPE_MEMVIOL) {
+ event_interrupt_poison_consumption(dev, pasid, source_id);
+ return;
+ }
+ break;
+ default:
+ break;
+ }
+ kfd_signal_event_interrupt(pasid, context_id0 & 0x7fffff, 23);
+ } else if (source_id == SOC15_INTSRC_CP_BAD_OPCODE) {
+ kfd_set_dbg_ev_from_interrupt(dev, pasid,
+ KFD_DEBUG_DOORBELL_ID(context_id0),
+ KFD_EC_MASK(KFD_DEBUG_CP_BAD_OP_ECODE(context_id0)),
+ NULL,
+ 0);
+ }
+ } else if (client_id == SOC15_IH_CLIENTID_SDMA0 ||
+ client_id == SOC15_IH_CLIENTID_SDMA1 ||
+ client_id == SOC15_IH_CLIENTID_SDMA2 ||
+ client_id == SOC15_IH_CLIENTID_SDMA3 ||
+ (client_id == SOC15_IH_CLIENTID_SDMA3_Sienna_Cichlid &&
+ KFD_GC_VERSION(dev) == IP_VERSION(10, 3, 0)) ||
+ client_id == SOC15_IH_CLIENTID_SDMA4 ||
+ client_id == SOC15_IH_CLIENTID_SDMA5 ||
+ client_id == SOC15_IH_CLIENTID_SDMA6 ||
+ client_id == SOC15_IH_CLIENTID_SDMA7) {
+ if (source_id == SOC15_INTSRC_SDMA_TRAP) {
+ kfd_signal_event_interrupt(pasid, context_id0 & 0xfffffff, 28);
+ } else if (source_id == SOC15_INTSRC_SDMA_ECC) {
+ event_interrupt_poison_consumption(dev, pasid, source_id);
+ return;
+ }
+ } else if (client_id == SOC15_IH_CLIENTID_VMC ||
+ client_id == SOC15_IH_CLIENTID_VMC1 ||
+ client_id == SOC15_IH_CLIENTID_UTCL2) {
+ struct kfd_vm_fault_info info = {0};
+ uint16_t ring_id = SOC15_RING_ID_FROM_IH_ENTRY(ih_ring_entry);
+ struct kfd_hsa_memory_exception_data exception_data;
+
+ if (client_id == SOC15_IH_CLIENTID_UTCL2 &&
+ amdgpu_amdkfd_ras_query_utcl2_poison_status(dev->adev)) {
+ event_interrupt_poison_consumption(dev, pasid, client_id);
+ return;
+ }
+
+ info.vmid = vmid;
+ info.mc_id = client_id;
+ info.page_addr = ih_ring_entry[4] |
+ (uint64_t)(ih_ring_entry[5] & 0xf) << 32;
+ info.prot_valid = ring_id & 0x08;
+ info.prot_read = ring_id & 0x10;
+ info.prot_write = ring_id & 0x20;
+
+ memset(&exception_data, 0, sizeof(exception_data));
+ exception_data.gpu_id = dev->id;
+ exception_data.va = (info.page_addr) << PAGE_SHIFT;
+ exception_data.failure.NotPresent = info.prot_valid ? 1 : 0;
+ exception_data.failure.NoExecute = info.prot_exec ? 1 : 0;
+ exception_data.failure.ReadOnly = info.prot_write ? 1 : 0;
+ exception_data.failure.imprecise = 0;
+
+ kfd_set_dbg_ev_from_interrupt(dev,
+ pasid,
+ -1,
+ KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION),
+ &exception_data,
+ sizeof(exception_data));
+ } else if (KFD_IRQ_IS_FENCE(client_id, source_id)) {
+ kfd_process_close_interrupt_drain(pasid);
+ }
+}
+
+const struct kfd_event_interrupt_class event_interrupt_class_v10 = {
+ .interrupt_isr = event_interrupt_isr_v10,
+ .interrupt_wq = event_interrupt_wq_v10,
+};
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v11.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v11.c
index c2166bf964ef..f933bd231fb9 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v11.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v11.c
@@ -26,6 +26,7 @@
#include "kfd_device_queue_manager.h"
#include "ivsrcid/vmc/irqsrcs_vmc_1_0.h"
#include "kfd_smi_events.h"
+#include "kfd_debug.h"
/*
* GFX11 SQ Interrupts
@@ -238,7 +239,7 @@ static bool event_interrupt_isr_v11(struct kfd_node *dev,
client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry);
/* Only handle interrupts from KFD VMIDs */
vmid = SOC15_VMID_FROM_IH_ENTRY(ih_ring_entry);
- if (/*!KFD_IRQ_IS_FENCE(client_id, source_id) &&*/
+ if (!KFD_IRQ_IS_FENCE(client_id, source_id) &&
(vmid < dev->vm_info.first_vmid_kfd ||
vmid > dev->vm_info.last_vmid_kfd))
return false;
@@ -267,7 +268,7 @@ static bool event_interrupt_isr_v11(struct kfd_node *dev,
source_id == SOC15_INTSRC_SQ_INTERRUPT_MSG ||
source_id == SOC15_INTSRC_CP_BAD_OPCODE ||
source_id == SOC21_INTSRC_SDMA_TRAP ||
- /* KFD_IRQ_IS_FENCE(client_id, source_id) || */
+ KFD_IRQ_IS_FENCE(client_id, source_id) ||
(((client_id == SOC21_IH_CLIENTID_VMC) ||
((client_id == SOC21_IH_CLIENTID_GFX) &&
(source_id == UTCL2_1_0__SRCID__FAULT))) &&
@@ -279,7 +280,7 @@ static void event_interrupt_wq_v11(struct kfd_node *dev,
{
uint16_t source_id, client_id, ring_id, pasid, vmid;
uint32_t context_id0, context_id1;
- uint8_t sq_int_enc, sq_int_errtype;
+ uint8_t sq_int_enc, sq_int_priv, sq_int_errtype;
struct kfd_vm_fault_info info = {0};
struct kfd_hsa_memory_exception_data exception_data;
@@ -312,9 +313,9 @@ static void event_interrupt_wq_v11(struct kfd_node *dev,
exception_data.failure.ReadOnly = info.prot_write ? 1 : 0;
exception_data.failure.imprecise = 0;
- /*kfd_set_dbg_ev_from_interrupt(dev, pasid, -1,
+ kfd_set_dbg_ev_from_interrupt(dev, pasid, -1,
KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION),
- &exception_data, sizeof(exception_data));*/
+ &exception_data, sizeof(exception_data));
kfd_smi_event_update_vmfault(dev, pasid);
/* GRBM, SDMA, SE, PMM */
@@ -324,11 +325,11 @@ static void event_interrupt_wq_v11(struct kfd_node *dev,
/* CP */
if (source_id == SOC15_INTSRC_CP_END_OF_PIPE)
kfd_signal_event_interrupt(pasid, context_id0, 32);
- /*else if (source_id == SOC15_INTSRC_CP_BAD_OPCODE)
+ else if (source_id == SOC15_INTSRC_CP_BAD_OPCODE)
kfd_set_dbg_ev_from_interrupt(dev, pasid,
KFD_CTXID0_DOORBELL_ID(context_id0),
KFD_EC_MASK(KFD_CTXID0_CP_BAD_OP_ECODE(context_id0)),
- NULL, 0);*/
+ NULL, 0);
/* SDMA */
else if (source_id == SOC21_INTSRC_SDMA_TRAP)
@@ -348,6 +349,13 @@ static void event_interrupt_wq_v11(struct kfd_node *dev,
break;
case SQ_INTERRUPT_WORD_ENCODING_INST:
print_sq_intr_info_inst(context_id0, context_id1);
+ sq_int_priv = REG_GET_FIELD(context_id0,
+ SQ_INTERRUPT_WORD_WAVE_CTXID0, PRIV);
+ if (sq_int_priv && (kfd_set_dbg_ev_from_interrupt(dev, pasid,
+ KFD_CTXID0_DOORBELL_ID(context_id0),
+ KFD_CTXID0_TRAP_CODE(context_id0),
+ NULL, 0)))
+ return;
break;
case SQ_INTERRUPT_WORD_ENCODING_ERROR:
print_sq_intr_info_error(context_id0, context_id1);
@@ -366,8 +374,8 @@ static void event_interrupt_wq_v11(struct kfd_node *dev,
kfd_signal_event_interrupt(pasid, context_id0 & 0xffffff, 24);
}
- /*} else if (KFD_IRQ_IS_FENCE(client_id, source_id)) {
- kfd_process_close_interrupt_drain(pasid);*/
+ } else if (KFD_IRQ_IS_FENCE(client_id, source_id)) {
+ kfd_process_close_interrupt_drain(pasid);
}
}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c
index 8cf58be80f4e..d5c9f30552e3 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c
@@ -23,10 +23,40 @@
#include "kfd_priv.h"
#include "kfd_events.h"
+#include "kfd_debug.h"
#include "soc15_int.h"
#include "kfd_device_queue_manager.h"
#include "kfd_smi_events.h"
+/*
+ * GFX9 SQ Interrupts
+ *
+ * There are 3 encoding types of interrupts sourced from SQ sent as a 44-bit
+ * packet to the Interrupt Handler:
+ * Auto - Generated by the SQG (various cmd overflows, timestamps etc)
+ * Wave - Generated by S_SENDMSG through a shader program
+ * Error - HW generated errors (Illegal instructions, Memviols, EDC etc)
+ *
+ * The 44-bit packet is mapped as {context_id1[7:0],context_id0[31:0]} plus
+ * 4-bits for VMID (SOC15_VMID_FROM_IH_ENTRY) as such:
+ *
+ * - context_id0[27:26]
+ * Encoding type (0 = Auto, 1 = Wave, 2 = Error)
+ *
+ * - context_id0[13]
+ * PRIV bit indicates that Wave S_SEND or error occurred within trap
+ *
+ * - {context_id1[7:0],context_id0[31:28],context_id0[11:0]}
+ * 24-bit data with the following layout per encoding type:
+ * Auto - only context_id0[8:0] is used, which reports various interrupts
+ * generated by SQG. The rest is 0.
+ * Wave - user data sent from m0 via S_SENDMSG
+ * Error - Error type (context_id1[7:4]), Error Details (rest of bits)
+ *
+ * The other context_id bits show coordinates (SE/SH/CU/SIMD/WAVE) for wave
+ * S_SENDMSG and Errors. These are 0 for Auto.
+ */
+
enum SQ_INTERRUPT_WORD_ENCODING {
SQ_INTERRUPT_WORD_ENCODING_AUTO = 0x0,
SQ_INTERRUPT_WORD_ENCODING_INST,
@@ -84,12 +114,32 @@ enum SQ_INTERRUPT_ERROR_TYPE {
#define SQ_INTERRUPT_WORD_WAVE_CTXID__SE_ID_MASK 0x03000000
#define SQ_INTERRUPT_WORD_WAVE_CTXID__ENCODING_MASK 0x0c000000
+/* GFX9 SQ interrupt 24-bit data from context_id<0,1> */
#define KFD_CONTEXT_ID_GET_SQ_INT_DATA(ctx0, ctx1) \
((ctx0 & 0xfff) | ((ctx0 >> 16) & 0xf000) | ((ctx1 << 16) & 0xff0000))
#define KFD_SQ_INT_DATA__ERR_TYPE_MASK 0xF00000
#define KFD_SQ_INT_DATA__ERR_TYPE__SHIFT 20
+/*
+ * The debugger will send user data(m0) with PRIV=1 to indicate it requires
+ * notification from the KFD with the following queue id (DOORBELL_ID) and
+ * trap code (TRAP_CODE).
+ */
+#define KFD_INT_DATA_DEBUG_DOORBELL_MASK 0x0003ff
+#define KFD_INT_DATA_DEBUG_TRAP_CODE_SHIFT 10
+#define KFD_INT_DATA_DEBUG_TRAP_CODE_MASK 0x07fc00
+#define KFD_DEBUG_DOORBELL_ID(sq_int_data) ((sq_int_data) & \
+ KFD_INT_DATA_DEBUG_DOORBELL_MASK)
+#define KFD_DEBUG_TRAP_CODE(sq_int_data) (((sq_int_data) & \
+ KFD_INT_DATA_DEBUG_TRAP_CODE_MASK) \
+ >> KFD_INT_DATA_DEBUG_TRAP_CODE_SHIFT)
+#define KFD_DEBUG_CP_BAD_OP_ECODE_MASK 0x3fffc00
+#define KFD_DEBUG_CP_BAD_OP_ECODE_SHIFT 10
+#define KFD_DEBUG_CP_BAD_OP_ECODE(ctxid0) (((ctxid0) & \
+ KFD_DEBUG_CP_BAD_OP_ECODE_MASK) \
+ >> KFD_DEBUG_CP_BAD_OP_ECODE_SHIFT)
+
static void event_interrupt_poison_consumption_v9(struct kfd_node *dev,
uint16_t pasid, uint16_t client_id)
{
@@ -168,14 +218,16 @@ static bool event_interrupt_isr_v9(struct kfd_node *dev,
uint16_t source_id, client_id, pasid, vmid;
const uint32_t *data = ih_ring_entry;
+ source_id = SOC15_SOURCE_ID_FROM_IH_ENTRY(ih_ring_entry);
+ client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry);
+
/* Only handle interrupts from KFD VMIDs */
vmid = SOC15_VMID_FROM_IH_ENTRY(ih_ring_entry);
- if (vmid < dev->vm_info.first_vmid_kfd ||
- vmid > dev->vm_info.last_vmid_kfd)
+ if (!KFD_IRQ_IS_FENCE(client_id, source_id) &&
+ (vmid < dev->vm_info.first_vmid_kfd ||
+ vmid > dev->vm_info.last_vmid_kfd))
return false;
- source_id = SOC15_SOURCE_ID_FROM_IH_ENTRY(ih_ring_entry);
- client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry);
pasid = SOC15_PASID_FROM_IH_ENTRY(ih_ring_entry);
/* Only handle clients we care about */
@@ -194,7 +246,8 @@ static bool event_interrupt_isr_v9(struct kfd_node *dev,
client_id != SOC15_IH_CLIENTID_SE0SH &&
client_id != SOC15_IH_CLIENTID_SE1SH &&
client_id != SOC15_IH_CLIENTID_SE2SH &&
- client_id != SOC15_IH_CLIENTID_SE3SH)
+ client_id != SOC15_IH_CLIENTID_SE3SH &&
+ !KFD_IRQ_IS_FENCE(client_id, source_id))
return false;
/* This is a known issue for gfx9. Under non HWS, pasid is not set
@@ -247,6 +300,7 @@ static bool event_interrupt_isr_v9(struct kfd_node *dev,
source_id == SOC15_INTSRC_SDMA_ECC ||
source_id == SOC15_INTSRC_SQ_INTERRUPT_MSG ||
source_id == SOC15_INTSRC_CP_BAD_OPCODE ||
+ KFD_IRQ_IS_FENCE(client_id, source_id) ||
((client_id == SOC15_IH_CLIENTID_VMC ||
client_id == SOC15_IH_CLIENTID_VMC1 ||
client_id == SOC15_IH_CLIENTID_UTCL2) &&
@@ -302,6 +356,13 @@ static void event_interrupt_wq_v9(struct kfd_node *dev,
REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, SIMD_ID),
REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, CU_ID),
sq_int_data);
+ if (context_id0 & SQ_INTERRUPT_WORD_WAVE_CTXID__PRIV_MASK) {
+ if (kfd_set_dbg_ev_from_interrupt(dev, pasid,
+ KFD_DEBUG_DOORBELL_ID(sq_int_data),
+ KFD_DEBUG_TRAP_CODE(sq_int_data),
+ NULL, 0))
+ return;
+ }
break;
case SQ_INTERRUPT_WORD_ENCODING_ERROR:
sq_intr_err = REG_GET_FIELD(sq_int_data, KFD_SQ_INT_DATA, ERR_TYPE);
@@ -324,8 +385,12 @@ static void event_interrupt_wq_v9(struct kfd_node *dev,
break;
}
kfd_signal_event_interrupt(pasid, context_id0 & 0xffffff, 24);
- } else if (source_id == SOC15_INTSRC_CP_BAD_OPCODE)
- kfd_signal_hw_exception_event(pasid);
+ } else if (source_id == SOC15_INTSRC_CP_BAD_OPCODE) {
+ kfd_set_dbg_ev_from_interrupt(dev, pasid,
+ KFD_DEBUG_DOORBELL_ID(context_id0),
+ KFD_EC_MASK(KFD_DEBUG_CP_BAD_OP_ECODE(context_id0)),
+ NULL, 0);
+ }
} else if (client_id == SOC15_IH_CLIENTID_SDMA0 ||
client_id == SOC15_IH_CLIENTID_SDMA1 ||
client_id == SOC15_IH_CLIENTID_SDMA2 ||
@@ -345,6 +410,7 @@ static void event_interrupt_wq_v9(struct kfd_node *dev,
client_id == SOC15_IH_CLIENTID_UTCL2) {
struct kfd_vm_fault_info info = {0};
uint16_t ring_id = SOC15_RING_ID_FROM_IH_ENTRY(ih_ring_entry);
+ struct kfd_hsa_memory_exception_data exception_data;
if (client_id == SOC15_IH_CLIENTID_UTCL2 &&
amdgpu_amdkfd_ras_query_utcl2_poison_status(dev->adev)) {
@@ -360,9 +426,23 @@ static void event_interrupt_wq_v9(struct kfd_node *dev,
info.prot_read = ring_id & 0x10;
info.prot_write = ring_id & 0x20;
+ memset(&exception_data, 0, sizeof(exception_data));
+ exception_data.gpu_id = dev->id;
+ exception_data.va = (info.page_addr) << PAGE_SHIFT;
+ exception_data.failure.NotPresent = info.prot_valid ? 1 : 0;
+ exception_data.failure.NoExecute = info.prot_exec ? 1 : 0;
+ exception_data.failure.ReadOnly = info.prot_write ? 1 : 0;
+ exception_data.failure.imprecise = 0;
+
+ kfd_set_dbg_ev_from_interrupt(dev,
+ pasid,
+ -1,
+ KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION),
+ &exception_data,
+ sizeof(exception_data));
kfd_smi_event_update_vmfault(dev, pasid);
- kfd_dqm_evict_pasid(dev->dqm, pasid);
- kfd_signal_vm_fault_event(dev, pasid, &info, NULL);
+ } else if (KFD_IRQ_IS_FENCE(client_id, source_id)) {
+ kfd_process_close_interrupt_drain(pasid);
}
}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index a02fb939614a..cd2d56e5cdf0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -963,6 +963,10 @@ struct kfd_process {
uint64_t exception_enable_mask;
uint64_t exception_status;
+ /* Used to drain stale interrupts */
+ wait_queue_head_t wait_irq_drain;
+ bool irq_drain_is_open;
+
/* shared virtual memory registered by this process */
struct svm_range_list svms;
@@ -1144,12 +1148,19 @@ int kfd_numa_node_to_apic_id(int numa_node_id);
void kfd_double_confirm_iommu_support(struct kfd_dev *gpu);
/* Interrupts */
+#define KFD_IRQ_FENCE_CLIENTID 0xff
+#define KFD_IRQ_FENCE_SOURCEID 0xff
+#define KFD_IRQ_IS_FENCE(client, source) \
+ ((client) == KFD_IRQ_FENCE_CLIENTID && \
+ (source) == KFD_IRQ_FENCE_SOURCEID)
int kfd_interrupt_init(struct kfd_node *dev);
void kfd_interrupt_exit(struct kfd_node *dev);
bool enqueue_ih_ring_entry(struct kfd_node *kfd, const void *ih_ring_entry);
bool interrupt_is_wanted(struct kfd_node *dev,
const uint32_t *ih_ring_entry,
uint32_t *patched_ihre, bool *flag);
+int kfd_process_drain_interrupts(struct kfd_process_device *pdd);
+void kfd_process_close_interrupt_drain(unsigned int pasid);
/* amdkfd Apertures */
int kfd_init_apertures(struct kfd_process *process);
@@ -1421,6 +1432,7 @@ uint64_t kfd_get_number_elems(struct kfd_dev *kfd);
/* Events */
extern const struct kfd_event_interrupt_class event_interrupt_class_cik;
extern const struct kfd_event_interrupt_class event_interrupt_class_v9;
+extern const struct kfd_event_interrupt_class event_interrupt_class_v10;
extern const struct kfd_event_interrupt_class event_interrupt_class_v11;
extern const struct kfd_device_global_init_class device_global_init_class_cik;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 3b7f219c9d06..3d3611705d41 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -862,6 +862,8 @@ struct kfd_process *kfd_create_process(struct task_struct *thread)
kfd_procfs_add_sysfs_stats(process);
kfd_procfs_add_sysfs_files(process);
kfd_procfs_add_sysfs_counters(process);
+
+ init_waitqueue_head(&process->wait_irq_drain);
}
out:
if (!IS_ERR(process))
@@ -2136,6 +2138,51 @@ void kfd_flush_tlb(struct kfd_process_device *pdd, enum TLB_FLUSH_TYPE type)
}
}
+/* assumes caller holds process lock. */
+int kfd_process_drain_interrupts(struct kfd_process_device *pdd)
+{
+ uint32_t irq_drain_fence[8];
+ int r = 0;
+
+ if (!KFD_IS_SOC15(pdd->dev))
+ return 0;
+
+ pdd->process->irq_drain_is_open = true;
+
+ memset(irq_drain_fence, 0, sizeof(irq_drain_fence));
+ irq_drain_fence[0] = (KFD_IRQ_FENCE_SOURCEID << 8) |
+ KFD_IRQ_FENCE_CLIENTID;
+ irq_drain_fence[3] = pdd->process->pasid;
+
+ /* ensure stale irqs scheduled KFD interrupts and send drain fence. */
+ if (amdgpu_amdkfd_send_close_event_drain_irq(pdd->dev->adev,
+ irq_drain_fence)) {
+ pdd->process->irq_drain_is_open = false;
+ return 0;
+ }
+
+ r = wait_event_interruptible(pdd->process->wait_irq_drain,
+ !READ_ONCE(pdd->process->irq_drain_is_open));
+ if (r)
+ pdd->process->irq_drain_is_open = false;
+
+ return r;
+}
+
+void kfd_process_close_interrupt_drain(unsigned int pasid)
+{
+ struct kfd_process *p;
+
+ p = kfd_lookup_process_by_pasid(pasid);
+
+ if (!p)
+ return;
+
+ WRITE_ONCE(p->irq_drain_is_open, false);
+ wake_up_all(&p->wait_irq_drain);
+ kfd_unref_process(p);
+}
+
struct send_exception_work_handler_workarea {
struct work_struct work;
struct kfd_process *p;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
index 43d432b5c5bc..70852a200d8f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
@@ -337,6 +337,10 @@ int pqm_create_queue(struct process_queue_manager *pqm,
kq->queue->properties.queue_id = *qid;
pqn->kq = kq;
pqn->q = NULL;
+ retval = kfd_process_drain_interrupts(pdd);
+ if (retval)
+ break;
+
retval = dev->dqm->ops.create_kernel_queue(dev->dqm,
kq, &pdd->qpd);
break;
--
cgit v1.2.3
From a70a93fa568b4f05aba548dadb673703eccf5480 Mon Sep 17 00:00:00 2001
From: Jonathan Kim <jonathan.kim@amd.com>
Date: Thu, 5 May 2022 16:15:37 -0400
Subject: drm/amdkfd: add debug suspend and resume process queues operation
In order to inspect waves from the saved context at any point during a
debug session, the debugger must be able to preempt queues to trigger
context save by suspending them.
On queue suspend, the KFD will copy the context save header information
so that the debugger can correctly crawl the appropriate size of the saved
context. The debugger must then also be allowed to resume suspended queues.
A queue that is newly created cannot be suspended because queue ids are
recycled after destruction so the debugger needs to know that this has
occurred. Query functions will be later added that will clear a given
queue of its new queue status.
A queue cannot be destroyed while it is suspended to preserve its saved
context during debugger inspection. Have queue destruction block while
a queue is suspended and unblocked when it is resumed. Likewise, if a
queue is about to be destroyed, it cannot be suspended.
Return the number of queues successfully suspended or resumed along with
a per queue status array where the upper bits per queue status show that
the request was invalid (new/destroyed queue suspend request, missing
queue) or an error occurred (HWS in a fatal state so it can't suspend or
resume queues).
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 5 +
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 1 +
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 11 +
drivers/gpu/drm/amd/amdkfd/kfd_debug.c | 7 +
.../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 447 ++++++++++++++++++++-
.../gpu/drm/amd/amdkfd/kfd_device_queue_manager.h | 10 +
drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c | 10 +
drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c | 15 +-
drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c | 14 +-
drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 5 +-
.../gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 1 +
11 files changed, 512 insertions(+), 14 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 98cd52bb005f..b4fcad0e62f7 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -772,6 +772,11 @@ bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev)
return adev->have_atomics_support;
}
+void amdgpu_amdkfd_debug_mem_fence(struct amdgpu_device *adev)
+{
+ amdgpu_device_flush_hdp(adev, NULL);
+}
+
void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev, bool reset)
{
amdgpu_umc_poison_handler(adev, reset);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index dd740e64e6e1..2d0406bff84e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -322,6 +322,7 @@ int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev,
uint64_t *mmap_offset);
int amdgpu_amdkfd_gpuvm_export_dmabuf(struct kgd_mem *mem,
struct dma_buf **dmabuf);
+void amdgpu_amdkfd_debug_mem_fence(struct amdgpu_device *adev);
int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev,
struct tile_config *config);
void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index a6570b124b2b..1fae97df7a1e 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -410,6 +410,7 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p,
pr_debug("Write ptr address == 0x%016llX\n",
args->write_pointer_address);
+ kfd_dbg_ev_raise(KFD_EC_MASK(EC_QUEUE_NEW), p, dev, queue_id, false, NULL, 0);
return 0;
err_create_queue:
@@ -2996,7 +2997,17 @@ static int kfd_ioctl_set_debug_trap(struct file *filep, struct kfd_process *p, v
args->launch_mode.launch_mode);
break;
case KFD_IOC_DBG_TRAP_SUSPEND_QUEUES:
+ r = suspend_queues(target,
+ args->suspend_queues.num_queues,
+ args->suspend_queues.grace_period,
+ args->suspend_queues.exception_mask,
+ (uint32_t *)args->suspend_queues.queue_array_ptr);
+
+ break;
case KFD_IOC_DBG_TRAP_RESUME_QUEUES:
+ r = resume_queues(target, args->resume_queues.num_queues,
+ (uint32_t *)args->resume_queues.queue_array_ptr);
+ break;
case KFD_IOC_DBG_TRAP_SET_NODE_ADDRESS_WATCH:
case KFD_IOC_DBG_TRAP_CLEAR_NODE_ADDRESS_WATCH:
case KFD_IOC_DBG_TRAP_SET_FLAGS:
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
index 53c3418562d4..f4d3dfb35cb3 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
@@ -339,6 +339,13 @@ void kfd_dbg_trap_deactivate(struct kfd_process *target, bool unwind, int unwind
}
kfd_dbg_set_workaround(target, false);
+
+ if (!unwind) {
+ int resume_count = resume_queues(target, 0, NULL);
+
+ if (resume_count)
+ pr_debug("Resumed %d queues\n", resume_count);
+ }
}
static void kfd_dbg_clean_exception_status(struct kfd_process *target)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index 44d87943e40a..bc9e81293165 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -952,6 +952,92 @@ out_unlock:
return retval;
}
+/* suspend_single_queue does not lock the dqm like the
+ * evict_process_queues_cpsch or evict_process_queues_nocpsch. You should
+ * lock the dqm before calling, and unlock after calling.
+ *
+ * The reason we don't lock the dqm is because this function may be
+ * called on multiple queues in a loop, so rather than locking/unlocking
+ * multiple times, we will just keep the dqm locked for all of the calls.
+ */
+static int suspend_single_queue(struct device_queue_manager *dqm,
+ struct kfd_process_device *pdd,
+ struct queue *q)
+{
+ bool is_new;
+
+ if (q->properties.is_suspended)
+ return 0;
+
+ pr_debug("Suspending PASID %u queue [%i]\n",
+ pdd->process->pasid,
+ q->properties.queue_id);
+
+ is_new = q->properties.exception_status & KFD_EC_MASK(EC_QUEUE_NEW);
+
+ if (is_new || q->properties.is_being_destroyed) {
+ pr_debug("Suspend: skip %s queue id %i\n",
+ is_new ? "new" : "destroyed",
+ q->properties.queue_id);
+ return -EBUSY;
+ }
+
+ q->properties.is_suspended = true;
+ if (q->properties.is_active) {
+ if (dqm->dev->kfd->shared_resources.enable_mes) {
+ int r = remove_queue_mes(dqm, q, &pdd->qpd);
+
+ if (r)
+ return r;
+ }
+
+ decrement_queue_count(dqm, &pdd->qpd, q);
+ q->properties.is_active = false;
+ }
+
+ return 0;
+}
+
+/* resume_single_queue does not lock the dqm like the functions
+ * restore_process_queues_cpsch or restore_process_queues_nocpsch. You should
+ * lock the dqm before calling, and unlock after calling.
+ *
+ * The reason we don't lock the dqm is because this function may be
+ * called on multiple queues in a loop, so rather than locking/unlocking
+ * multiple times, we will just keep the dqm locked for all of the calls.
+ */
+static int resume_single_queue(struct device_queue_manager *dqm,
+ struct qcm_process_device *qpd,
+ struct queue *q)
+{
+ struct kfd_process_device *pdd;
+
+ if (!q->properties.is_suspended)
+ return 0;
+
+ pdd = qpd_to_pdd(qpd);
+
+ pr_debug("Restoring from suspend PASID %u queue [%i]\n",
+ pdd->process->pasid,
+ q->properties.queue_id);
+
+ q->properties.is_suspended = false;
+
+ if (QUEUE_IS_ACTIVE(q->properties)) {
+ if (dqm->dev->kfd->shared_resources.enable_mes) {
+ int r = add_queue_mes(dqm, q, &pdd->qpd);
+
+ if (r)
+ return r;
+ }
+
+ q->properties.is_active = true;
+ increment_queue_count(dqm, qpd, q);
+ }
+
+ return 0;
+}
+
static int evict_process_queues_nocpsch(struct device_queue_manager *dqm,
struct qcm_process_device *qpd)
{
@@ -1926,6 +2012,31 @@ static int execute_queues_cpsch(struct device_queue_manager *dqm,
return map_queues_cpsch(dqm);
}
+static int wait_on_destroy_queue(struct device_queue_manager *dqm,
+ struct queue *q)
+{
+ struct kfd_process_device *pdd = kfd_get_process_device_data(q->device,
+ q->process);
+ int ret = 0;
+
+ if (pdd->qpd.is_debug)
+ return ret;
+
+ q->properties.is_being_destroyed = true;
+
+ if (pdd->process->debug_trap_enabled && q->properties.is_suspended) {
+ dqm_unlock(dqm);
+ mutex_unlock(&q->process->mutex);
+ ret = wait_event_interruptible(dqm->destroy_wait,
+ !q->properties.is_suspended);
+
+ mutex_lock(&q->process->mutex);
+ dqm_lock(dqm);
+ }
+
+ return ret;
+}
+
static int destroy_queue_cpsch(struct device_queue_manager *dqm,
struct qcm_process_device *qpd,
struct queue *q)
@@ -1945,11 +2056,16 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm,
q->properties.queue_id);
}
- retval = 0;
-
/* remove queue from list to prevent rescheduling after preemption */
dqm_lock(dqm);
+ retval = wait_on_destroy_queue(dqm, q);
+
+ if (retval) {
+ dqm_unlock(dqm);
+ return retval;
+ }
+
if (qpd->is_debug) {
/*
* error, currently we do not allow to destroy a queue
@@ -1996,7 +2112,14 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm,
dqm_unlock(dqm);
- /* Do free_mqd after dqm_unlock(dqm) to avoid circular locking */
+ /*
+ * Do free_mqd and raise delete event after dqm_unlock(dqm) to avoid
+ * circular locking
+ */
+ kfd_dbg_ev_raise(KFD_EC_MASK(EC_DEVICE_QUEUE_DELETE),
+ qpd->pqm->process, q->device,
+ -1, false, NULL, 0);
+
mqd_mgr->free_mqd(mqd_mgr, q->mqd, q->mqd_mem_obj);
return retval;
@@ -2461,8 +2584,10 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_node *dev)
goto out_free;
}
- if (!dqm->ops.initialize(dqm))
+ if (!dqm->ops.initialize(dqm)) {
+ init_waitqueue_head(&dqm->destroy_wait);
return dqm;
+ }
out_free:
kfree(dqm);
@@ -2602,6 +2727,320 @@ out_unlock:
return r;
}
+#define QUEUE_NOT_FOUND -1
+/* invalidate queue operation in array */
+static void q_array_invalidate(uint32_t num_queues, uint32_t *queue_ids)
+{
+ int i;
+
+ for (i = 0; i < num_queues; i++)
+ queue_ids[i] |= KFD_DBG_QUEUE_INVALID_MASK;
+}
+
+/* find queue index in array */
+static int q_array_get_index(unsigned int queue_id,
+ uint32_t num_queues,
+ uint32_t *queue_ids)
+{
+ int i;
+
+ for (i = 0; i < num_queues; i++)
+ if (queue_id == (queue_ids[i] & ~KFD_DBG_QUEUE_INVALID_MASK))
+ return i;
+
+ return QUEUE_NOT_FOUND;
+}
+
+struct copy_context_work_handler_workarea {
+ struct work_struct copy_context_work;
+ struct kfd_process *p;
+};
+
+static void copy_context_work_handler (struct work_struct *work)
+{
+ struct copy_context_work_handler_workarea *workarea;
+ struct mqd_manager *mqd_mgr;
+ struct queue *q;
+ struct mm_struct *mm;
+ struct kfd_process *p;
+ uint32_t tmp_ctl_stack_used_size, tmp_save_area_used_size;
+ int i;
+
+ workarea = container_of(work,
+ struct copy_context_work_handler_workarea,
+ copy_context_work);
+
+ p = workarea->p;
+ mm = get_task_mm(p->lead_thread);
+
+ if (!mm)
+ return;
+
+ kthread_use_mm(mm);
+ for (i = 0; i < p->n_pdds; i++) {
+ struct kfd_process_device *pdd = p->pdds[i];
+ struct device_queue_manager *dqm = pdd->dev->dqm;
+ struct qcm_process_device *qpd = &pdd->qpd;
+
+ list_for_each_entry(q, &qpd->queues_list, list) {
+ mqd_mgr = dqm->mqd_mgrs[KFD_MQD_TYPE_CP];
+
+ /* We ignore the return value from get_wave_state
+ * because
+ * i) right now, it always returns 0, and
+ * ii) if we hit an error, we would continue to the
+ * next queue anyway.
+ */
+ mqd_mgr->get_wave_state(mqd_mgr,
+ q->mqd,
+ &q->properties,
+ (void __user *) q->properties.ctx_save_restore_area_address,
+ &tmp_ctl_stack_used_size,
+ &tmp_save_area_used_size);
+ }
+ }
+ kthread_unuse_mm(mm);
+ mmput(mm);
+}
+
+static uint32_t *get_queue_ids(uint32_t num_queues, uint32_t *usr_queue_id_array)
+{
+ size_t array_size = num_queues * sizeof(uint32_t);
+ uint32_t *queue_ids = NULL;
+
+ if (!usr_queue_id_array)
+ return NULL;
+
+ queue_ids = kzalloc(array_size, GFP_KERNEL);
+ if (!queue_ids)
+ return ERR_PTR(-ENOMEM);
+
+ if (copy_from_user(queue_ids, usr_queue_id_array, array_size))
+ return ERR_PTR(-EFAULT);
+
+ return queue_ids;
+}
+
+int resume_queues(struct kfd_process *p,
+ uint32_t num_queues,
+ uint32_t *usr_queue_id_array)
+{
+ uint32_t *queue_ids = NULL;
+ int total_resumed = 0;
+ int i;
+
+ if (usr_queue_id_array) {
+ queue_ids = get_queue_ids(num_queues, usr_queue_id_array);
+
+ if (IS_ERR(queue_ids))
+ return PTR_ERR(queue_ids);
+
+ /* mask all queues as invalid. unmask per successful request */
+ q_array_invalidate(num_queues, queue_ids);
+ }
+
+ for (i = 0; i < p->n_pdds; i++) {
+ struct kfd_process_device *pdd = p->pdds[i];
+ struct device_queue_manager *dqm = pdd->dev->dqm;
+ struct qcm_process_device *qpd = &pdd->qpd;
+ struct queue *q;
+ int r, per_device_resumed = 0;
+
+ dqm_lock(dqm);
+
+ /* unmask queues that resume or already resumed as valid */
+ list_for_each_entry(q, &qpd->queues_list, list) {
+ int q_idx = QUEUE_NOT_FOUND;
+
+ if (queue_ids)
+ q_idx = q_array_get_index(
+ q->properties.queue_id,
+ num_queues,
+ queue_ids);
+
+ if (!queue_ids || q_idx != QUEUE_NOT_FOUND) {
+ int err = resume_single_queue(dqm, &pdd->qpd, q);
+
+ if (queue_ids) {
+ if (!err) {
+ queue_ids[q_idx] &=
+ ~KFD_DBG_QUEUE_INVALID_MASK;
+ } else {
+ queue_ids[q_idx] |=
+ KFD_DBG_QUEUE_ERROR_MASK;
+ break;
+ }
+ }
+
+ if (dqm->dev->kfd->shared_resources.enable_mes) {
+ wake_up_all(&dqm->destroy_wait);
+ if (!err)
+ total_resumed++;
+ } else {
+ per_device_resumed++;
+ }
+ }
+ }
+
+ if (!per_device_resumed) {
+ dqm_unlock(dqm);
+ continue;
+ }
+
+ r = execute_queues_cpsch(dqm,
+ KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES,
+ 0,
+ USE_DEFAULT_GRACE_PERIOD);
+ if (r) {
+ pr_err("Failed to resume process queues\n");
+ if (queue_ids) {
+ list_for_each_entry(q, &qpd->queues_list, list) {
+ int q_idx = q_array_get_index(
+ q->properties.queue_id,
+ num_queues,
+ queue_ids);
+
+ /* mask queue as error on resume fail */
+ if (q_idx != QUEUE_NOT_FOUND)
+ queue_ids[q_idx] |=
+ KFD_DBG_QUEUE_ERROR_MASK;
+ }
+ }
+ } else {
+ wake_up_all(&dqm->destroy_wait);
+ total_resumed += per_device_resumed;
+ }
+
+ dqm_unlock(dqm);
+ }
+
+ if (queue_ids) {
+ if (copy_to_user((void __user *)usr_queue_id_array, queue_ids,
+ num_queues * sizeof(uint32_t)))
+ pr_err("copy_to_user failed on queue resume\n");
+
+ kfree(queue_ids);
+ }
+
+ return total_resumed;
+}
+
+int suspend_queues(struct kfd_process *p,
+ uint32_t num_queues,
+ uint32_t grace_period,
+ uint64_t exception_clear_mask,
+ uint32_t *usr_queue_id_array)
+{
+ uint32_t *queue_ids = get_queue_ids(num_queues, usr_queue_id_array);
+ int total_suspended = 0;
+ int i;
+
+ if (IS_ERR(queue_ids))
+ return PTR_ERR(queue_ids);
+
+ /* mask all queues as invalid. umask on successful request */
+ q_array_invalidate(num_queues, queue_ids);
+
+ for (i = 0; i < p->n_pdds; i++) {
+ struct kfd_process_device *pdd = p->pdds[i];
+ struct device_queue_manager *dqm = pdd->dev->dqm;
+ struct qcm_process_device *qpd = &pdd->qpd;
+ struct queue *q;
+ int r, per_device_suspended = 0;
+
+ mutex_lock(&p->event_mutex);
+ dqm_lock(dqm);
+
+ /* unmask queues that suspend or already suspended */
+ list_for_each_entry(q, &qpd->queues_list, list) {
+ int q_idx = q_array_get_index(q->properties.queue_id,
+ num_queues,
+ queue_ids);
+
+ if (q_idx != QUEUE_NOT_FOUND) {
+ int err = suspend_single_queue(dqm, pdd, q);
+ bool is_mes = dqm->dev->kfd->shared_resources.enable_mes;
+
+ if (!err) {
+ queue_ids[q_idx] &= ~KFD_DBG_QUEUE_INVALID_MASK;
+ if (exception_clear_mask && is_mes)
+ q->properties.exception_status &=
+ ~exception_clear_mask;
+
+ if (is_mes)
+ total_suspended++;
+ else
+ per_device_suspended++;
+ } else if (err != -EBUSY) {
+ r = err;
+ queue_ids[q_idx] |= KFD_DBG_QUEUE_ERROR_MASK;
+ break;
+ }
+ }
+ }
+
+ if (!per_device_suspended) {
+ dqm_unlock(dqm);
+ mutex_unlock(&p->event_mutex);
+ if (total_suspended)
+ amdgpu_amdkfd_debug_mem_fence(dqm->dev->adev);
+ continue;
+ }
+
+ r = execute_queues_cpsch(dqm,
+ KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0,
+ grace_period);
+
+ if (r)
+ pr_err("Failed to suspend process queues.\n");
+ else
+ total_suspended += per_device_suspended;
+
+ list_for_each_entry(q, &qpd->queues_list, list) {
+ int q_idx = q_array_get_index(q->properties.queue_id,
+ num_queues, queue_ids);
+
+ if (q_idx == QUEUE_NOT_FOUND)
+ continue;
+
+ /* mask queue as error on suspend fail */
+ if (r)
+ queue_ids[q_idx] |= KFD_DBG_QUEUE_ERROR_MASK;
+ else if (exception_clear_mask)
+ q->properties.exception_status &=
+ ~exception_clear_mask;
+ }
+
+ dqm_unlock(dqm);
+ mutex_unlock(&p->event_mutex);
+ amdgpu_device_flush_hdp(dqm->dev->adev, NULL);
+ }
+
+ if (total_suspended) {
+ struct copy_context_work_handler_workarea copy_context_worker;
+
+ INIT_WORK_ONSTACK(
+ ©_context_worker.copy_context_work,
+ copy_context_work_handler);
+
+ copy_context_worker.p = p;
+
+ schedule_work(©_context_worker.copy_context_work);
+
+
+ flush_work(©_context_worker.copy_context_work);
+ destroy_work_on_stack(©_context_worker.copy_context_work);
+ }
+
+ if (copy_to_user((void __user *)usr_queue_id_array, queue_ids,
+ num_queues * sizeof(uint32_t)))
+ pr_err("copy_to_user failed on queue suspend\n");
+
+ kfree(queue_ids);
+
+ return total_suspended;
+}
+
int debug_lock_and_unmap(struct device_queue_manager *dqm)
{
int r;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
index bb75d93712eb..d4e6dbffe8c2 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
@@ -263,6 +263,8 @@ struct device_queue_manager {
uint32_t current_logical_xcc_start;
uint32_t wait_times;
+
+ wait_queue_head_t destroy_wait;
};
void device_queue_manager_init_cik(
@@ -290,6 +292,14 @@ int reserve_debug_trap_vmid(struct device_queue_manager *dqm,
struct qcm_process_device *qpd);
int release_debug_trap_vmid(struct device_queue_manager *dqm,
struct qcm_process_device *qpd);
+int suspend_queues(struct kfd_process *p,
+ uint32_t num_queues,
+ uint32_t grace_period,
+ uint64_t exception_clear_mask,
+ uint32_t *usr_queue_id_array);
+int resume_queues(struct kfd_process *p,
+ uint32_t num_queues,
+ uint32_t *usr_queue_id_array);
int debug_lock_and_unmap(struct device_queue_manager *dqm);
int debug_map_and_unlock(struct device_queue_manager *dqm);
int debug_refresh_runlist(struct device_queue_manager *dqm);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
index a0ac4f2fe6b5..94c0fc2e57b7 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
@@ -237,6 +237,7 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd,
u32 *save_area_used_size)
{
struct v10_compute_mqd *m;
+ struct kfd_context_save_area_header header;
m = get_mqd(mqd);
@@ -255,6 +256,15 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd,
* accessible to user mode
*/
+ header.wave_state.control_stack_size = *ctl_stack_used_size;
+ header.wave_state.wave_state_size = *save_area_used_size;
+
+ header.wave_state.wave_state_offset = m->cp_hqd_wg_state_offset;
+ header.wave_state.control_stack_offset = m->cp_hqd_cntl_stack_offset;
+
+ if (copy_to_user(ctl_stack, &header, sizeof(header.wave_state)))
+ return -EFAULT;
+
return 0;
}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c
index 9a9b4e853516..31fec5e70d13 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c
@@ -291,7 +291,7 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd,
u32 *save_area_used_size)
{
struct v11_compute_mqd *m;
- /*struct mqd_user_context_save_area_header header;*/
+ struct kfd_context_save_area_header header;
m = get_mqd(mqd);
@@ -309,16 +309,15 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd,
* it's part of the context save area that is already
* accessible to user mode
*/
-/*
- header.control_stack_size = *ctl_stack_used_size;
- header.wave_state_size = *save_area_used_size;
+ header.wave_state.control_stack_size = *ctl_stack_used_size;
+ header.wave_state.wave_state_size = *save_area_used_size;
- header.wave_state_offset = m->cp_hqd_wg_state_offset;
- header.control_stack_offset = m->cp_hqd_cntl_stack_offset;
+ header.wave_state.wave_state_offset = m->cp_hqd_wg_state_offset;
+ header.wave_state.control_stack_offset = m->cp_hqd_cntl_stack_offset;
- if (copy_to_user(ctl_stack, &header, sizeof(header)))
+ if (copy_to_user(ctl_stack, &header, sizeof(header.wave_state)))
return -EFAULT;
-*/
+
return 0;
}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
index 5b87c244e909..601bb9f68048 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
@@ -311,6 +311,7 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd,
u32 *save_area_used_size)
{
struct v9_mqd *m;
+ struct kfd_context_save_area_header header;
/* Control stack is located one page after MQD. */
void *mqd_ctl_stack = (void *)((uintptr_t)mqd + PAGE_SIZE);
@@ -322,7 +323,18 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd,
*save_area_used_size = m->cp_hqd_wg_state_offset -
m->cp_hqd_cntl_stack_size;
- if (copy_to_user(ctl_stack, mqd_ctl_stack, m->cp_hqd_cntl_stack_size))
+ header.wave_state.control_stack_size = *ctl_stack_used_size;
+ header.wave_state.wave_state_size = *save_area_used_size;
+
+ header.wave_state.wave_state_offset = m->cp_hqd_wg_state_offset;
+ header.wave_state.control_stack_offset = m->cp_hqd_cntl_stack_offset;
+
+ if (copy_to_user(ctl_stack, &header, sizeof(header.wave_state)))
+ return -EFAULT;
+
+ if (copy_to_user(ctl_stack + m->cp_hqd_cntl_stack_offset,
+ mqd_ctl_stack + m->cp_hqd_cntl_stack_offset,
+ *ctl_stack_used_size))
return -EFAULT;
return 0;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index cd2d56e5cdf0..05da43bf233a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -510,6 +510,8 @@ struct queue_properties {
uint32_t doorbell_off;
bool is_interop;
bool is_evicted;
+ bool is_suspended;
+ bool is_being_destroyed;
bool is_active;
bool is_gws;
uint32_t pm4_target_xcc;
@@ -535,7 +537,8 @@ struct queue_properties {
#define QUEUE_IS_ACTIVE(q) ((q).queue_size > 0 && \
(q).queue_address != 0 && \
(q).queue_percent > 0 && \
- !(q).is_evicted)
+ !(q).is_evicted && \
+ !(q).is_suspended)
enum mqd_update_flag {
UPDATE_FLAG_DBG_WA_ENABLE = 1,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
index 70852a200d8f..01ccab607a69 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
@@ -187,6 +187,7 @@ static int init_user_queue(struct process_queue_manager *pqm,
/* Doorbell initialized in user space*/
q_properties->doorbell_ptr = NULL;
+ q_properties->exception_status = KFD_EC_MASK(EC_QUEUE_NEW);
/* let DQM handle it*/
q_properties->vmid = 0;
--
cgit v1.2.3