public-mirrors/linux
1
0
Fork 0
mirror of git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git synced 2025-08-05 16:54:27 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

drm/amdgpu: add amdgpu soft reset

Browse source 
Check gpu status first, if MC/VMC/DISPLAY hang, directly triger full reset.
If engine hangs, then triger engine soft reset, if soft reset fails, will
fallback to full reset.

Signed-off-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit is contained in:
Chunming Zhou 2016-07-15 15:57:13 +08:00 • committed by Alex Deucher
parent 1057f20c22
commit 35d782feae
2 changed files with 88 additions and 18 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

104
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
View file

@ -1962,7 +1962,8 @@ int amdgpu_pre_soft_reset(struct amdgpu_device *adev)
for (i = 0; i < adev->num_ip_blocks; i++) { for (i = 0; i < adev->num_ip_blocks; i++) {
if (!adev->ip_block_status[i].valid) if (!adev->ip_block_status[i].valid)
continue; continue;
if (adev->ip_blocks[i].funcs->pre_soft_reset) { if (adev->ip_block_status[i].hang &&
adev->ip_blocks[i].funcs->pre_soft_reset) {
r = adev->ip_blocks[i].funcs->pre_soft_reset(adev); r = adev->ip_blocks[i].funcs->pre_soft_reset(adev);
if (r) if (r)
return r; return r;
@ -1972,6 +1973,58 @@ int amdgpu_pre_soft_reset(struct amdgpu_device *adev)
return 0; return 0;
} }
static bool amdgpu_need_full_reset(struct amdgpu_device *adev)
{
if (adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang ||
adev->ip_block_status[AMD_IP_BLOCK_TYPE_IH].hang ||
adev->ip_block_status[AMD_IP_BLOCK_TYPE_SMC].hang ||
adev->ip_block_status[AMD_IP_BLOCK_TYPE_GFX].hang ||
adev->ip_block_status[AMD_IP_BLOCK_TYPE_SDMA].hang ||
adev->ip_block_status[AMD_IP_BLOCK_TYPE_UVD].hang ||
adev->ip_block_status[AMD_IP_BLOCK_TYPE_VCE].hang ||
adev->ip_block_status[AMD_IP_BLOCK_TYPE_ACP].hang ||
adev->ip_block_status[AMD_IP_BLOCK_TYPE_DCE].hang) {
DRM_INFO("Some block need full reset!\n");
return true;
}
return false;
}
static int amdgpu_soft_reset(struct amdgpu_device *adev)
{
int i, r = 0;
for (i = 0; i < adev->num_ip_blocks; i++) {
if (!adev->ip_block_status[i].valid)
continue;
if (adev->ip_block_status[i].hang &&
adev->ip_blocks[i].funcs->soft_reset) {
r = adev->ip_blocks[i].funcs->soft_reset(adev);
if (r)
return r;
}
}
return 0;
}
static int amdgpu_post_soft_reset(struct amdgpu_device *adev)
{
int i, r = 0;
for (i = 0; i < adev->num_ip_blocks; i++) {
if (!adev->ip_block_status[i].valid)
continue;
if (adev->ip_block_status[i].hang &&
adev->ip_blocks[i].funcs->post_soft_reset)
r = adev->ip_blocks[i].funcs->post_soft_reset(adev);
if (r)
return r;
}
return 0;
}
/** /**
* amdgpu_gpu_reset - reset the asic * amdgpu_gpu_reset - reset the asic
* *
@ -1984,6 +2037,7 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev)
{ {
int i, r; int i, r;
int resched; int resched;
bool need_full_reset;
if (!amdgpu_check_soft_reset(adev)) { if (!amdgpu_check_soft_reset(adev)) {
DRM_INFO("No hardware hang detected. Did some blocks stall?\n"); DRM_INFO("No hardware hang detected. Did some blocks stall?\n");
@ -2007,28 +2061,42 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev)
/* after all hw jobs are reset, hw fence is meaningless, so force_completion */ /* after all hw jobs are reset, hw fence is meaningless, so force_completion */
amdgpu_fence_driver_force_completion(adev); amdgpu_fence_driver_force_completion(adev);
/* save scratch */ need_full_reset = amdgpu_need_full_reset(adev);
amdgpu_atombios_scratch_regs_save(adev);
r = amdgpu_suspend(adev); if (!need_full_reset) {
amdgpu_pre_soft_reset(adev);
r = amdgpu_soft_reset(adev);
amdgpu_post_soft_reset(adev);
if (r || amdgpu_check_soft_reset(adev)) {
DRM_INFO("soft reset failed, will fallback to full reset!\n");
need_full_reset = true;
}
}
if (need_full_reset) {
/* save scratch */
amdgpu_atombios_scratch_regs_save(adev);
r = amdgpu_suspend(adev);
retry: retry:
/* Disable fb access */ /* Disable fb access */
if (adev->mode_info.num_crtc) { if (adev->mode_info.num_crtc) {
struct amdgpu_mode_mc_save save; struct amdgpu_mode_mc_save save;
amdgpu_display_stop_mc_access(adev, &save); amdgpu_display_stop_mc_access(adev, &save);
amdgpu_wait_for_idle(adev, AMD_IP_BLOCK_TYPE_GMC); amdgpu_wait_for_idle(adev, AMD_IP_BLOCK_TYPE_GMC);
} }
r = amdgpu_asic_reset(adev); r = amdgpu_asic_reset(adev);
/* post card */ /* post card */
amdgpu_atom_asic_init(adev->mode_info.atom_context); amdgpu_atom_asic_init(adev->mode_info.atom_context);
if (!r) { if (!r) {
dev_info(adev->dev, "GPU reset succeeded, trying to resume\n"); dev_info(adev->dev, "GPU reset succeeded, trying to resume\n");
r = amdgpu_resume(adev); r = amdgpu_resume(adev);
}
/* restore scratch */
amdgpu_atombios_scratch_regs_restore(adev);
} }
/* restore scratch */
amdgpu_atombios_scratch_regs_restore(adev);
if (!r) { if (!r) {
r = amdgpu_ib_ring_tests(adev); r = amdgpu_ib_ring_tests(adev);
if (r) { if (r) {

2
drivers/gpu/drm/amd/include/amd_shared.h
View file

@ -165,6 +165,8 @@ struct amd_ip_funcs {
int (*pre_soft_reset)(void *handle); int (*pre_soft_reset)(void *handle);
/* soft reset the IP block */ /* soft reset the IP block */
int (*soft_reset)(void *handle); int (*soft_reset)(void *handle);
/* post soft reset the IP block */
int (*post_soft_reset)(void *handle);
/* enable/disable cg for the IP block */ /* enable/disable cg for the IP block */
int (*set_clockgating_state)(void *handle, int (*set_clockgating_state)(void *handle,
enum amd_clockgating_state state); enum amd_clockgating_state state);