Am 22.06.22 um 12:07 schrieb Pierre-Eric Pelloux-Prayer:
Without this change amdgpu_ttm_training_data_block_init tries
to allocate at the end of the real amount of RAM, which
then fails like this if amdgpu.vramlimit=XXXX is used:
[drm:amdgpu_ttm_init [amdgpu]] *ERROR* alloc c2p_bo failed(-12)!
[drm:amdgpu_device_init.cold [amdgpu]] *ERROR* sw_init of IP block <gmc_v10_0> failed -12
amdgpu: amdgpu_device_ip_init failed
amdgpu: Fatal error during GPU init
Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@xxxxxxx>
That looks like a real bug fix to me, so I think we should add "CC:
stable..." tag.
Apart from that Reviewed-by: Christian König <christian.koenig@xxxxxxx>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index be0efaae79a9..952e99e6d07e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -1621,9 +1621,9 @@ static void amdgpu_ttm_training_data_block_init(struct amdgpu_device *adev)
memset(ctx, 0, sizeof(*ctx));
ctx->c2p_train_data_offset =
- ALIGN((adev->gmc.mc_vram_size - adev->mman.discovery_tmr_size - SZ_1M), SZ_1M);
+ ALIGN((adev->gmc.real_vram_size - adev->mman.discovery_tmr_size - SZ_1M), SZ_1M);
ctx->p2c_train_data_offset =
- (adev->gmc.mc_vram_size - GDDR6_MEM_TRAINING_OFFSET);
+ (adev->gmc.real_vram_size - GDDR6_MEM_TRAINING_OFFSET);
ctx->train_data_size =
GDDR6_MEM_TRAINING_DATA_SIZE_IN_BYTES;
base-commit: a81bcfc756bcaa9e8bb46262f910504fa5290aab
