On 03/29/2017 10:27 PM, Christian König wrote:
Am 29.03.2017 um 16:17 schrieb Alex Deucher:
On Wed, Mar 29, 2017 at 4:47 AM, Christian König
<[email protected]> wrote:
Am 29.03.2017 um 10:08 schrieb Junwei Zhang:
Signed-off-by: Junwei Zhang <[email protected]>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 1ad1113..4af4ea0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -1080,6 +1080,11 @@ static void amdgpu_get_block_size(struct
amdgpu_device *adev)
static void amdgpu_check_vm_size(struct amdgpu_device *adev)
{
+ if (adev->asic_type >= CHIP_VEGA10) {
+ amdgpu_vm_size = 1U << 18; /* 256TB */
+ return;
+ }
+
NAK, that would break in a mixed environment with APU+Vega10.
Instead just go into gmc_v9_0_sw_init() and set max_pfn = 1ULL << 36 and add
a comment there why we hardcode this value for Vega10 and later.
We should fix this for real and move these variables into the driver
vm structure and then set those based on the global parameter
variables in the various gmc modules.
Yeah, that's exactly what I've suggested as well.
max_pfn is already part of the vm manager structure, so setting this different
on Vega10 than on previous generations shouldn't be a problem.
We just need to add the block size there as well and should be done with it.
Thanks for all the comments above.
Prepared another patch to fix it.
Jerry.
Christian.
Alex
Christian.
if (!amdgpu_check_pot_argument(amdgpu_vm_size)) {
dev_warn(adev->dev, "VM size (%d) must be a power of 2\n",
amdgpu_vm_size);
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
