diff options
author | Jonathan Kim <jonathan.kim@amd.com> | 2022-09-01 18:27:15 +0300 |
---|---|---|
committer | Alex Deucher <alexander.deucher@amd.com> | 2023-06-09 19:35:52 +0300 |
commit | 69a8c3ae2dea84a6d571e4c1aad306f630f3ccfd (patch) | |
tree | 17a0ba33f8f739f1cfeb7971438b55d32849e41b /drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c | |
parent | 218895820e6fccade42a7c3ab9c0a44dec0a1ebc (diff) | |
download | linux-69a8c3ae2dea84a6d571e4c1aad306f630f3ccfd.tar.xz |
drm/amdkfd: apply trap workaround for gfx11
Due to a HW bug, waves in only half the shader arrays can enter trap.
When starting a debug session, relocate all waves to the first shader
array of each shader engine and mask off the 2nd shader array as
unavailable.
When ending a debug session, re-enable the 2nd shader array per
shader engine.
User CU masking per queue cannot be guaranteed to remain functional
if requested during debugging (e.g. user cu mask requests only 2nd shader
array as an available resource leading to zero HW resources available)
nor can runtime be alerted of any of these changes during execution.
Make user CU masking and debugging mutual exclusive with respect to
availability.
If the debugger tries to attach to a process with a user cu masked
queue, return the runtime status as enabled but busy.
If the debugger tries to attach and fails to reallocate queue waves to
the first shader array of each shader engine, return the runtime status
as enabled but with an error.
In addition, like any other mutli-process debug supported devices,
disable trap temporary setup per-process to avoid performance impact from
setup overhead.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c')
-rw-r--r-- | drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c | 42 |
1 files changed, 31 insertions, 11 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c index 3a48bbc589fe..9a9b4e853516 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c @@ -46,15 +46,33 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd, { struct v11_compute_mqd *m; uint32_t se_mask[KFD_MAX_NUM_SE] = {0}; + bool has_wa_flag = minfo && (minfo->update_flag & (UPDATE_FLAG_DBG_WA_ENABLE | + UPDATE_FLAG_DBG_WA_DISABLE)); - if (!minfo || (minfo->update_flag != UPDATE_FLAG_CU_MASK) || - !minfo->cu_mask.ptr) + if (!minfo || !(has_wa_flag || minfo->cu_mask.ptr)) return; + m = get_mqd(mqd); + + if (has_wa_flag) { + uint32_t wa_mask = minfo->update_flag == UPDATE_FLAG_DBG_WA_ENABLE ? + 0xffff : 0xffffffff; + + m->compute_static_thread_mgmt_se0 = wa_mask; + m->compute_static_thread_mgmt_se1 = wa_mask; + m->compute_static_thread_mgmt_se2 = wa_mask; + m->compute_static_thread_mgmt_se3 = wa_mask; + m->compute_static_thread_mgmt_se4 = wa_mask; + m->compute_static_thread_mgmt_se5 = wa_mask; + m->compute_static_thread_mgmt_se6 = wa_mask; + m->compute_static_thread_mgmt_se7 = wa_mask; + + return; + } + mqd_symmetrically_map_cu_mask(mm, minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask); - m = get_mqd(mqd); m->compute_static_thread_mgmt_se0 = se_mask[0]; m->compute_static_thread_mgmt_se1 = se_mask[1]; m->compute_static_thread_mgmt_se2 = se_mask[2]; @@ -109,6 +127,7 @@ static void init_mqd(struct mqd_manager *mm, void **mqd, uint64_t addr; struct v11_compute_mqd *m; int size; + uint32_t wa_mask = q->is_dbg_wa ? 0xffff : 0xffffffff; m = (struct v11_compute_mqd *) mqd_mem_obj->cpu_ptr; addr = mqd_mem_obj->gpu_addr; @@ -122,14 +141,15 @@ static void init_mqd(struct mqd_manager *mm, void **mqd, m->header = 0xC0310800; m->compute_pipelinestat_enable = 1; - m->compute_static_thread_mgmt_se0 = 0xFFFFFFFF; - m->compute_static_thread_mgmt_se1 = 0xFFFFFFFF; - m->compute_static_thread_mgmt_se2 = 0xFFFFFFFF; - m->compute_static_thread_mgmt_se3 = 0xFFFFFFFF; - m->compute_static_thread_mgmt_se4 = 0xFFFFFFFF; - m->compute_static_thread_mgmt_se5 = 0xFFFFFFFF; - m->compute_static_thread_mgmt_se6 = 0xFFFFFFFF; - m->compute_static_thread_mgmt_se7 = 0xFFFFFFFF; + + m->compute_static_thread_mgmt_se0 = wa_mask; + m->compute_static_thread_mgmt_se1 = wa_mask; + m->compute_static_thread_mgmt_se2 = wa_mask; + m->compute_static_thread_mgmt_se3 = wa_mask; + m->compute_static_thread_mgmt_se4 = wa_mask; + m->compute_static_thread_mgmt_se5 = wa_mask; + m->compute_static_thread_mgmt_se6 = wa_mask; + m->compute_static_thread_mgmt_se7 = wa_mask; m->cp_hqd_persistent_state = CP_HQD_PERSISTENT_STATE__PRELOAD_REQ_MASK | 0x55 << CP_HQD_PERSISTENT_STATE__PRELOAD_SIZE__SHIFT; |