summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
diff options
context:
space:
mode:
authorDave Airlie <airlied@redhat.com>2019-10-30 05:46:00 +1000
committerDave Airlie <airlied@redhat.com>2019-10-30 05:46:09 +1000
commit60845e34f0c5c19a9e86af477b429993952f585b (patch)
tree6b2fed16145e80c3904ffae2d996efedb80db3a0 /drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
parent3275a71e76fac5bc276f0d60e027b18c2e8d7a5b (diff)
parent0e04ad7d1857670944786a8465930a049aaf995f (diff)
Merge tag 'drm-next-5.5-2019-10-25' of git://people.freedesktop.org/~agd5f/linux into drm-next
drm-next-5.5-2019-10-25: amdgpu: - BACO support for CI and VI asics - Quick memory training support for navi - MSI-X support - RAS fixes - Display AVI infoframe fixes - Display ref clock fixes for renoir - Fix number of audio endpoints in renoir - Fix for discovery tables - Powerplay fixes - Documentation fixes - Misc cleanups radeon: - revert a PPC fix which broke x86 Signed-off-by: Dave Airlie <airlied@redhat.com> From: Alex Deucher <alexdeucher@gmail.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191025221020.203546-1-alexander.deucher@amd.com
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c')
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c119
1 files changed, 110 insertions, 9 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 0976fffd3fee..a61b0d9ab986 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -1652,6 +1652,88 @@ static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev)
&adev->fw_vram_usage.va);
}
+/*
+ * Memoy training reservation functions
+ */
+
+/**
+ * amdgpu_ttm_training_reserve_vram_fini - free memory training reserved vram
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * free memory training reserved vram if it has been reserved.
+ */
+static int amdgpu_ttm_training_reserve_vram_fini(struct amdgpu_device *adev)
+{
+ struct psp_memory_training_context *ctx = &adev->psp.mem_train_ctx;
+
+ ctx->init = PSP_MEM_TRAIN_NOT_SUPPORT;
+ amdgpu_bo_free_kernel(&ctx->c2p_bo, NULL, NULL);
+ ctx->c2p_bo = NULL;
+
+ amdgpu_bo_free_kernel(&ctx->p2c_bo, NULL, NULL);
+ ctx->p2c_bo = NULL;
+
+ return 0;
+}
+
+/**
+ * amdgpu_ttm_training_reserve_vram_init - create bo vram reservation from memory training
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * create bo vram reservation from memory training.
+ */
+static int amdgpu_ttm_training_reserve_vram_init(struct amdgpu_device *adev)
+{
+ int ret;
+ struct psp_memory_training_context *ctx = &adev->psp.mem_train_ctx;
+
+ memset(ctx, 0, sizeof(*ctx));
+ if (!adev->fw_vram_usage.mem_train_support) {
+ DRM_DEBUG("memory training does not support!\n");
+ return 0;
+ }
+
+ ctx->c2p_train_data_offset = adev->fw_vram_usage.mem_train_fb_loc;
+ ctx->p2c_train_data_offset = (adev->gmc.mc_vram_size - GDDR6_MEM_TRAINING_OFFSET);
+ ctx->train_data_size = GDDR6_MEM_TRAINING_DATA_SIZE_IN_BYTES;
+
+ DRM_DEBUG("train_data_size:%llx,p2c_train_data_offset:%llx,c2p_train_data_offset:%llx.\n",
+ ctx->train_data_size,
+ ctx->p2c_train_data_offset,
+ ctx->c2p_train_data_offset);
+
+ ret = amdgpu_bo_create_kernel_at(adev,
+ ctx->p2c_train_data_offset,
+ ctx->train_data_size,
+ AMDGPU_GEM_DOMAIN_VRAM,
+ &ctx->p2c_bo,
+ NULL);
+ if (ret) {
+ DRM_ERROR("alloc p2c_bo failed(%d)!\n", ret);
+ goto Err_out;
+ }
+
+ ret = amdgpu_bo_create_kernel_at(adev,
+ ctx->c2p_train_data_offset,
+ ctx->train_data_size,
+ AMDGPU_GEM_DOMAIN_VRAM,
+ &ctx->c2p_bo,
+ NULL);
+ if (ret) {
+ DRM_ERROR("alloc c2p_bo failed(%d)!\n", ret);
+ goto Err_out;
+ }
+
+ ctx->init = PSP_MEM_TRAIN_RESERVE_SUCCESS;
+ return 0;
+
+Err_out:
+ amdgpu_ttm_training_reserve_vram_fini(adev);
+ return ret;
+}
+
/**
* amdgpu_ttm_init - Init the memory management (ttm) as well as various
* gtt/vram related fields.
@@ -1726,6 +1808,14 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
return r;
}
+ /*
+ *The reserved vram for memory training must be pinned to the specified
+ *place on the VRAM, so reserve it early.
+ */
+ r = amdgpu_ttm_training_reserve_vram_init(adev);
+ if (r)
+ return r;
+
/* allocate memory as required for VGA
* This is used for VGA emulation and pre-OS scanout buffers to
* avoid display artifacts while transitioning between pre-OS
@@ -1736,6 +1826,20 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
NULL, &stolen_vga_buf);
if (r)
return r;
+
+ /*
+ * reserve one TMR (64K) memory at the top of VRAM which holds
+ * IP Discovery data and is protected by PSP.
+ */
+ r = amdgpu_bo_create_kernel_at(adev,
+ adev->gmc.real_vram_size - DISCOVERY_TMR_SIZE,
+ DISCOVERY_TMR_SIZE,
+ AMDGPU_GEM_DOMAIN_VRAM,
+ &adev->discovery_memory,
+ NULL);
+ if (r)
+ return r;
+
DRM_INFO("amdgpu: %uM of VRAM memory ready\n",
(unsigned) (adev->gmc.real_vram_size / (1024 * 1024)));
@@ -1800,6 +1904,9 @@ void amdgpu_ttm_late_init(struct amdgpu_device *adev)
void *stolen_vga_buf;
/* return the VGA stolen memory (if any) back to VRAM */
amdgpu_bo_free_kernel(&adev->stolen_vga_memory, NULL, &stolen_vga_buf);
+
+ /* return the IP Discovery TMR memory back to VRAM */
+ amdgpu_bo_free_kernel(&adev->discovery_memory, NULL, NULL);
}
/**
@@ -1811,6 +1918,7 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev)
return;
amdgpu_ttm_debugfs_fini(adev);
+ amdgpu_ttm_training_reserve_vram_fini(adev);
amdgpu_ttm_fw_reserve_vram_fini(adev);
if (adev->mman.aper_base_kaddr)
iounmap(adev->mman.aper_base_kaddr);
@@ -1907,10 +2015,7 @@ static int amdgpu_map_buffer(struct ttm_buffer_object *bo,
*addr += (u64)window * AMDGPU_GTT_MAX_TRANSFER_SIZE *
AMDGPU_GPU_PAGE_SIZE;
- num_dw = adev->mman.buffer_funcs->copy_num_dw;
- while (num_dw & 0x7)
- num_dw++;
-
+ num_dw = ALIGN(adev->mman.buffer_funcs->copy_num_dw, 8);
num_bytes = num_pages * 8;
r = amdgpu_job_alloc_with_ib(adev, num_dw * 4 + num_bytes, &job);
@@ -1970,11 +2075,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
max_bytes = adev->mman.buffer_funcs->copy_max_bytes;
num_loops = DIV_ROUND_UP(byte_count, max_bytes);
- num_dw = num_loops * adev->mman.buffer_funcs->copy_num_dw;
-
- /* for IB padding */
- while (num_dw & 0x7)
- num_dw++;
+ num_dw = ALIGN(num_loops * adev->mman.buffer_funcs->copy_num_dw, 8);
r = amdgpu_job_alloc_with_ib(adev, num_dw * 4, &job);
if (r)