summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorArunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>2023-10-10 07:35:06 -0700
committerAlex Deucher <alexander.deucher@amd.com>2023-11-17 09:29:53 -0500
commitc8031019dc95e3ab7cc0b09f1894c5f52dc0c187 (patch)
tree55eddd0b4ed3ecf7977d72bdfdec1402300ec7d1
parente8c2d3e25b844ad8f7c8b269a7cfd65285329264 (diff)
drm/amdgpu: Implement a new 64bit sequence memory driver
Developed a new driver which allocates a 64bit memory on each request in sequence order. At the moment, user queue fence memory is the main consumer of this seq64 driver. v2: Worked on review comments from Christian for the following modifications - Move driver name from "semaphore" to "seq64" - Remove unnecessary PT/PD mapping - Move enable_mes check into init/fini functions. v3: Worked on review comments from Christian - drop enable_mes check - use DECLARE_BITMAP for bit array - added kerneldoc for seq64 v4: Worked on review comments from Christian - Rename amdgpu_seq64_get name with amdgpu_seq64_alloc v5: Worked on review comments from Christian - Fix seq64 lockdep warning - move fpriv->seq64_va check into amdgpu_seq64_unmap() - make the function amdgpu_seq64_unmap() return as void. - reserve the buffers as not interruptible. v6: port to drm_exec (Alex) v7: disable for now (Arun) Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
-rw-r--r--drivers/gpu/drm/amd/amdgpu/Makefile2
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu.h5
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu_device.c7
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c2
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c247
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h49
6 files changed, 311 insertions, 1 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile
index 2afecc55090f..260e32ef7bae 100644
--- a/drivers/gpu/drm/amd/amdgpu/Makefile
+++ b/drivers/gpu/drm/amd/amdgpu/Makefile
@@ -80,7 +80,7 @@ amdgpu-y += amdgpu_device.o amdgpu_doorbell_mgr.o amdgpu_kms.o \
amdgpu_umc.o smu_v11_0_i2c.o amdgpu_fru_eeprom.o amdgpu_rap.o \
amdgpu_fw_attestation.o amdgpu_securedisplay.o \
amdgpu_eeprom.o amdgpu_mca.o amdgpu_psp_ta.o amdgpu_lsdma.o \
- amdgpu_ring_mux.o amdgpu_xcp.o
+ amdgpu_ring_mux.o amdgpu_xcp.o amdgpu_seq64.o
amdgpu-$(CONFIG_PROC_FS) += amdgpu_fdinfo.o
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 9d92ca157677..4d100f8d75bf 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -109,6 +109,7 @@
#include "amdgpu_mca.h"
#include "amdgpu_ras.h"
#include "amdgpu_xcp.h"
+#include "amdgpu_seq64.h"
#define MAX_GPU_INSTANCE 64
@@ -468,6 +469,7 @@ struct amdgpu_fpriv {
struct amdgpu_vm vm;
struct amdgpu_bo_va *prt_va;
struct amdgpu_bo_va *csa_va;
+ struct amdgpu_bo_va *seq64_va;
struct mutex bo_list_lock;
struct idr bo_list_handles;
struct amdgpu_ctx_mgr ctx_mgr;
@@ -986,6 +988,9 @@ struct amdgpu_device {
/* GDS */
struct amdgpu_gds gds;
+ /* for userq and VM fences */
+ struct amdgpu_seq64 seq64;
+
/* KFD */
struct amdgpu_kfd_dev kfd;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 7eeaf0aa7f81..151f8b950b72 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -2676,6 +2676,12 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
goto init_failed;
}
}
+
+ r = amdgpu_seq64_init(adev);
+ if (r) {
+ DRM_ERROR("allocate seq64 failed %d\n", r);
+ goto init_failed;
+ }
}
}
@@ -3138,6 +3144,7 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
amdgpu_device_wb_fini(adev);
amdgpu_device_mem_scratch_fini(adev);
amdgpu_ib_pool_fini(adev);
+ amdgpu_seq64_fini(adev);
}
r = adev->ip_blocks[i].version->funcs->sw_fini((void *)adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
index 583cf03950cd..b5ebafd4a3ad 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
@@ -1428,6 +1428,8 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev,
fpriv->csa_va = NULL;
}
+ amdgpu_seq64_unmap(adev, fpriv);
+
pasid = fpriv->vm.pasid;
pd = amdgpu_bo_ref(fpriv->vm.root.bo);
if (!WARN_ON(amdgpu_bo_reserve(pd, true))) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c
new file mode 100644
index 000000000000..f3de02193138
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c
@@ -0,0 +1,247 @@
+// SPDX-License-Identifier: MIT
+/*
+ * 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 "amdgpu.h"
+#include "amdgpu_seq64.h"
+
+#include <drm/drm_exec.h>
+
+/**
+ * DOC: amdgpu_seq64
+ *
+ * amdgpu_seq64 allocates a 64bit memory on each request in sequence order.
+ * seq64 driver is required for user queue fence memory allocation, TLB
+ * counters and VM updates. It has maximum count of 32768 64 bit slots.
+ */
+
+/**
+ * amdgpu_seq64_map - Map the seq64 memory to VM
+ *
+ * @adev: amdgpu_device pointer
+ * @vm: vm pointer
+ * @bo_va: bo_va pointer
+ * @seq64_addr: seq64 vaddr start address
+ * @size: seq64 pool size
+ *
+ * Map the seq64 memory to the given VM.
+ *
+ * Returns:
+ * 0 on success or a negative error code on failure
+ */
+int amdgpu_seq64_map(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+ struct amdgpu_bo_va **bo_va, u64 seq64_addr,
+ uint32_t size)
+{
+ struct amdgpu_bo *bo;
+ struct drm_exec exec;
+ int r;
+
+ bo = adev->seq64.sbo;
+ if (!bo)
+ return -EINVAL;
+
+ drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT);
+ drm_exec_until_all_locked(&exec) {
+ r = amdgpu_vm_lock_pd(vm, &exec, 0);
+ if (likely(!r))
+ r = drm_exec_lock_obj(&exec, &bo->tbo.base);
+ drm_exec_retry_on_contention(&exec);
+ if (unlikely(r))
+ goto error;
+ }
+
+ *bo_va = amdgpu_vm_bo_add(adev, vm, bo);
+ if (!*bo_va) {
+ r = -ENOMEM;
+ goto error;
+ }
+
+ r = amdgpu_vm_bo_map(adev, *bo_va, seq64_addr, 0, size,
+ AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE |
+ AMDGPU_PTE_EXECUTABLE);
+ if (r) {
+ DRM_ERROR("failed to do bo_map on userq sem, err=%d\n", r);
+ amdgpu_vm_bo_del(adev, *bo_va);
+ goto error;
+ }
+
+ r = amdgpu_vm_bo_update(adev, *bo_va, false);
+ if (r) {
+ DRM_ERROR("failed to do vm_bo_update on userq sem\n");
+ amdgpu_vm_bo_del(adev, *bo_va);
+ goto error;
+ }
+
+error:
+ drm_exec_fini(&exec);
+ return r;
+}
+
+/**
+ * amdgpu_seq64_unmap - Unmap the seq64 memory
+ *
+ * @adev: amdgpu_device pointer
+ * @fpriv: DRM file private
+ *
+ * Unmap the seq64 memory from the given VM.
+ */
+void amdgpu_seq64_unmap(struct amdgpu_device *adev, struct amdgpu_fpriv *fpriv)
+{
+ struct amdgpu_vm *vm;
+ struct amdgpu_bo *bo;
+ struct drm_exec exec;
+ int r;
+
+ if (!fpriv->seq64_va)
+ return;
+
+ bo = adev->seq64.sbo;
+ if (!bo)
+ return;
+
+ vm = &fpriv->vm;
+
+ drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT);
+ drm_exec_until_all_locked(&exec) {
+ r = amdgpu_vm_lock_pd(vm, &exec, 0);
+ if (likely(!r))
+ r = drm_exec_lock_obj(&exec, &bo->tbo.base);
+ drm_exec_retry_on_contention(&exec);
+ if (unlikely(r))
+ goto error;
+ }
+
+ amdgpu_vm_bo_del(adev, fpriv->seq64_va);
+
+ fpriv->seq64_va = NULL;
+
+error:
+ drm_exec_fini(&exec);
+}
+
+/**
+ * amdgpu_seq64_alloc - Allocate a 64 bit memory
+ *
+ * @adev: amdgpu_device pointer
+ * @gpu_addr: allocated gpu VA start address
+ * @cpu_addr: allocated cpu VA start address
+ *
+ * Alloc a 64 bit memory from seq64 pool.
+ *
+ * Returns:
+ * 0 on success or a negative error code on failure
+ */
+int amdgpu_seq64_alloc(struct amdgpu_device *adev, u64 *gpu_addr,
+ u64 **cpu_addr)
+{
+ unsigned long bit_pos;
+ u32 offset;
+
+ bit_pos = find_first_zero_bit(adev->seq64.used, adev->seq64.num_sem);
+
+ if (bit_pos < adev->seq64.num_sem) {
+ __set_bit(bit_pos, adev->seq64.used);
+ offset = bit_pos << 6; /* convert to qw offset */
+ } else {
+ return -EINVAL;
+ }
+
+ *gpu_addr = offset + AMDGPU_SEQ64_VADDR_START;
+ *cpu_addr = offset + adev->seq64.cpu_base_addr;
+
+ return 0;
+}
+
+/**
+ * amdgpu_seq64_free - Free the given 64 bit memory
+ *
+ * @adev: amdgpu_device pointer
+ * @gpu_addr: gpu start address to be freed
+ *
+ * Free the given 64 bit memory from seq64 pool.
+ *
+ */
+void amdgpu_seq64_free(struct amdgpu_device *adev, u64 gpu_addr)
+{
+ u32 offset;
+
+ offset = gpu_addr - AMDGPU_SEQ64_VADDR_START;
+
+ offset >>= 6;
+ if (offset < adev->seq64.num_sem)
+ __clear_bit(offset, adev->seq64.used);
+}
+
+/**
+ * amdgpu_seq64_fini - Cleanup seq64 driver
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Free the memory space allocated for seq64.
+ *
+ */
+void amdgpu_seq64_fini(struct amdgpu_device *adev)
+{
+ amdgpu_bo_free_kernel(&adev->seq64.sbo,
+ NULL,
+ (void **)&adev->seq64.cpu_base_addr);
+}
+
+/**
+ * amdgpu_seq64_init - Initialize seq64 driver
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Allocate the required memory space for seq64.
+ *
+ * Returns:
+ * 0 on success or a negative error code on failure
+ */
+int amdgpu_seq64_init(struct amdgpu_device *adev)
+{
+ int r;
+
+ if (adev->seq64.sbo)
+ return 0;
+
+ /*
+ * AMDGPU_MAX_SEQ64_SLOTS * sizeof(u64) * 8 = AMDGPU_MAX_SEQ64_SLOTS
+ * 64bit slots
+ */
+ r = amdgpu_bo_create_kernel(adev, AMDGPU_SEQ64_SIZE,
+ PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
+ &adev->seq64.sbo, NULL,
+ (void **)&adev->seq64.cpu_base_addr);
+ if (r) {
+ dev_warn(adev->dev, "(%d) create seq64 failed\n", r);
+ return r;
+ }
+
+ memset(adev->seq64.cpu_base_addr, 0, AMDGPU_SEQ64_SIZE);
+
+ adev->seq64.num_sem = AMDGPU_MAX_SEQ64_SLOTS;
+ memset(&adev->seq64.used, 0, sizeof(adev->seq64.used));
+
+ return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h
new file mode 100644
index 000000000000..2196e72be508
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h
@@ -0,0 +1,49 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * 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.
+ *
+ */
+
+#ifndef __AMDGPU_SEQ64_H__
+#define __AMDGPU_SEQ64_H__
+
+#define AMDGPU_SEQ64_SIZE (2ULL << 20)
+#define AMDGPU_MAX_SEQ64_SLOTS (AMDGPU_SEQ64_SIZE / (sizeof(u64) * 8))
+#define AMDGPU_SEQ64_VADDR_OFFSET 0x50000
+#define AMDGPU_SEQ64_VADDR_START (AMDGPU_VA_RESERVED_SIZE + AMDGPU_SEQ64_VADDR_OFFSET)
+
+struct amdgpu_seq64 {
+ struct amdgpu_bo *sbo;
+ u32 num_sem;
+ u64 *cpu_base_addr;
+ DECLARE_BITMAP(used, AMDGPU_MAX_SEQ64_SLOTS);
+};
+
+void amdgpu_seq64_fini(struct amdgpu_device *adev);
+int amdgpu_seq64_init(struct amdgpu_device *adev);
+int amdgpu_seq64_alloc(struct amdgpu_device *adev, u64 *gpu_addr, u64 **cpu_addr);
+void amdgpu_seq64_free(struct amdgpu_device *adev, u64 gpu_addr);
+int amdgpu_seq64_map(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+ struct amdgpu_bo_va **bo_va, u64 seq64_addr, uint32_t size);
+void amdgpu_seq64_unmap(struct amdgpu_device *adev, struct amdgpu_fpriv *fpriv);
+
+#endif
+