bybrooklyn/linux
1
0
Fork 0
mirror of https://github.com/torvalds/linux.git synced 2026-04-18 14:53:58 -04:00
Code Issues Packages Projects Releases Wiki Activity

drm/amdgpu: Handle GPUVM fault storms

Browse Source 
When many wavefronts cause VM faults at the same time, it can
overwhelm the interrupt handler and cause IH ring overflows before
the driver can notify or kill the faulting application.

As a workaround I'm introducing limited per-VM fault credit. After
that number of VM faults have occurred, further VM faults are
filtered out at the prescreen stage of processing.

This depends on the PASID in the interrupt packet, so it currently
only works for KFD contexts.

Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit is contained in:
Felix Kuehling
2017-09-21 16:26:41 -04:00
committed by Alex Deucher
parent 1bab0fc01b
commit c98171ccf6
7 changed files with 112 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
drivers/gpu/drm/amd/amdgpu/cz_ih.c
View File

@@ -216,8 +216,23 @@ static u32 cz_ih_get_wptr(struct amdgpu_device *adev)
*/
static bool cz_ih_prescreen_iv(struct amdgpu_device *adev)
{
/* Process all interrupts */
return true;
u32 ring_index = adev->irq.ih.rptr >> 2;
u16 pasid;
switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) {
case 146:
case 147:
pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16;
if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid))
return true;
break;
default:
/* Not a VM fault */
return true;
}
adev->irq.ih.rptr += 16;
return false;
}
/**