From 01f648202c5390f4c366793b34c27cddad4ca8d7 Mon Sep 17 00:00:00 2001
From: Jonathan Kim <jonathan.kim@amd.com>
Date: Wed, 30 Mar 2022 15:31:00 -0400
Subject: drm/amdgpu: add gfx9.4.1 hw debug mode enable and disable calls
On GFX9.4.1, the implicit wait count instruction on s_barrier is
disabled by default in the driver during normal operation for
performance requirements.
There is a hardware bug in GFX9.4.1 where if the implicit wait count
instruction after an s_barrier instruction is disabled, any wave that
hits an exception may step over the s_barrier when returning from the
trap handler with the barrier logic having no ability to be
aware of this, thereby causing other waves to wait at the barrier
indefinitely resulting in a shader hang. This bug has been corrected
for GFX9.4.2 and onward.
Since the debugger subscribes to hardware exceptions, in order to avoid
this bug, the debugger must enable implicit wait count on s_barrier
for a debug session and disable it on detach.
In order to change this setting in the in the device global SQ_CONFIG
register, the GFX pipeline must be idle. GFX9.4.1 as a compute device
will either dispatch work through the compute ring buffers used for
image post processing or through the hardware scheduler by the KFD.
Have the KGD suspend and drain the compute ring buffer, then suspend the
hardware scheduler and block any future KFD process job requests before
changing the implicit wait count setting. Once set, resume all work.
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>
---
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c | 116 +++++++++++++++++++++
1 file changed, 116 insertions(+)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index 4191af5a3f13..d2918e5c0dea 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -26,6 +26,7 @@
#include "amdgpu.h"
#include "amdgpu_amdkfd.h"
#include "amdgpu_amdkfd_arcturus.h"
+#include "amdgpu_reset.h"
#include "sdma0/sdma0_4_2_2_offset.h"
#include "sdma0/sdma0_4_2_2_sh_mask.h"
#include "sdma1/sdma1_4_2_2_offset.h"
@@ -48,6 +49,8 @@
#include "amdgpu_amdkfd_gfx_v9.h"
#include "gfxhub_v1_0.h"
#include "mmhub_v9_4.h"
+#include "gc/gc_9_0_offset.h"
+#include "gc/gc_9_0_sh_mask.h"
#define HQD_N_REGS 56
#define DUMP_REG(addr) do { \
@@ -276,6 +279,117 @@ int kgd_arcturus_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
return 0;
}
+/*
+ * Helper used to suspend/resume gfx pipe for image post process work to set
+ * barrier behaviour.
+ */
+static int suspend_resume_compute_scheduler(struct amdgpu_device *adev, bool suspend)
+{
+ int i, r = 0;
+
+ for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+ struct amdgpu_ring *ring = &adev->gfx.compute_ring[i];
+
+ if (!(ring && ring->sched.thread))
+ continue;
+
+ /* stop secheduler and drain ring. */
+ if (suspend) {
+ drm_sched_stop(&ring->sched, NULL);
+ r = amdgpu_fence_wait_empty(ring);
+ if (r)
+ goto out;
+ } else {
+ drm_sched_start(&ring->sched, false);
+ }
+ }
+
+out:
+ /* return on resume or failure to drain rings. */
+ if (!suspend || r)
+ return r;
+
+ return amdgpu_device_ip_wait_for_idle(adev, GC_HWIP);
+}
+
+static void set_barrier_auto_waitcnt(struct amdgpu_device *adev, bool enable_waitcnt)
+{
+ uint32_t data;
+
+ WRITE_ONCE(adev->barrier_has_auto_waitcnt, enable_waitcnt);
+
+ if (!down_read_trylock(&adev->reset_domain->sem))
+ return;
+
+ amdgpu_amdkfd_suspend(adev, false);
+
+ if (suspend_resume_compute_scheduler(adev, true))
+ goto out;
+
+ data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_CONFIG));
+ data = REG_SET_FIELD(data, SQ_CONFIG, DISABLE_BARRIER_WAITCNT,
+ !enable_waitcnt);
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_CONFIG), data);
+
+out:
+ suspend_resume_compute_scheduler(adev, false);
+
+ amdgpu_amdkfd_resume(adev, false);
+
+ up_read(&adev->reset_domain->sem);
+}
+
+/*
+ * restore_dbg_registers is ignored here but is a general interface requirement
+ * for devices that support GFXOFF and where the RLC save/restore list
+ * does not support hw registers for debugging i.e. the driver has to manually
+ * initialize the debug mode registers after it has disabled GFX off during the
+ * debug session.
+ */
+static uint32_t kgd_arcturus_enable_debug_trap(struct amdgpu_device *adev,
+ bool restore_dbg_registers,
+ uint32_t vmid)
+{
+ mutex_lock(&adev->grbm_idx_mutex);
+
+ kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+ set_barrier_auto_waitcnt(adev, true);
+
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+ kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+ mutex_unlock(&adev->grbm_idx_mutex);
+
+ return 0;
+}
+
+/*
+ * keep_trap_enabled is ignored here but is a general interface requirement
+ * for devices that support multi-process debugging where the performance
+ * overhead from trap temporary setup needs to be bypassed when the debug
+ * session has ended.
+ */
+static uint32_t kgd_arcturus_disable_debug_trap(struct amdgpu_device *adev,
+ bool keep_trap_enabled,
+ uint32_t vmid)
+{
+
+ mutex_lock(&adev->grbm_idx_mutex);
+
+ kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+ set_barrier_auto_waitcnt(adev, false);
+
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+ kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+ mutex_unlock(&adev->grbm_idx_mutex);
+
+ return 0;
+}
const struct kfd2kgd_calls arcturus_kfd2kgd = {
.program_sh_mem_settings = kgd_gfx_v9_program_sh_mem_settings,
.set_pasid_vmid_mapping = kgd_gfx_v9_set_pasid_vmid_mapping,
@@ -294,6 +408,8 @@ const struct kfd2kgd_calls arcturus_kfd2kgd = {
kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
.set_vm_context_page_table_base =
kgd_gfx_v9_set_vm_context_page_table_base,
+ .enable_debug_trap = kgd_arcturus_enable_debug_trap,
+ .disable_debug_trap = kgd_arcturus_disable_debug_trap,
.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings
};
--
cgit v1.2.3
From 7cee6a6824a0429a6255abe91b5af01b9a01cd03 Mon Sep 17 00:00:00 2001
From: Jonathan Kim <jonathan.kim@amd.com>
Date: Thu, 23 Mar 2023 17:17:20 -0400
Subject: drm/amdgpu: add configurable grace period for unmap queues
The HWS schedule allows a grace period for wave completion prior to
preemption for better performance by avoiding CWSR on waves that can
potentially complete quickly. The debugger, on the other hand, will
want to inspect wave status immediately after it actively triggers
preemption (a suspend function to be provided).
To minimize latency between preemption and debugger wave inspection, allow
immediate preemption by setting the grace period to 0.
Note that setting the preepmtion grace period to 0 will result in an
infinite grace period being set due to a CP FW bug so set it to 1 for now.
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>
---
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c | 2 +
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c | 2 +
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c | 43 ++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h | 6 ++
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c | 2 +
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 43 ++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h | 8 ++-
.../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 63 +++++++++++++++------
.../gpu/drm/amd/amdkfd/kfd_device_queue_manager.h | 3 +
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c | 32 +++++++++++
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c | 39 +++++++++++++
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_vi.c | 2 +
drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h | 65 ++++++++++++++++++++++
drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 5 ++
14 files changed, 295 insertions(+), 20 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
index a6f98141c29c..b811a0985050 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
@@ -82,5 +82,7 @@ const struct kfd2kgd_calls aldebaran_kfd2kgd = {
.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
.enable_debug_trap = kgd_aldebaran_enable_debug_trap,
.disable_debug_trap = kgd_aldebaran_disable_debug_trap,
+ .get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
+ .build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index d2918e5c0dea..a62bd0068515 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -410,6 +410,8 @@ const struct kfd2kgd_calls arcturus_kfd2kgd = {
kgd_gfx_v9_set_vm_context_page_table_base,
.enable_debug_trap = kgd_arcturus_enable_debug_trap,
.disable_debug_trap = kgd_arcturus_disable_debug_trap,
+ .get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
+ .build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
index 240f5006e278..98006c7021dd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
@@ -803,6 +803,47 @@ uint32_t kgd_gfx_v10_disable_debug_trap(struct amdgpu_device *adev,
return 0;
}
+/* kgd_gfx_v10_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
+ * The values read are:
+ * ib_offload_wait_time -- Wait Count for Indirect Buffer Offloads.
+ * atomic_offload_wait_time -- Wait Count for L2 and GDS Atomics Offloads.
+ * wrm_offload_wait_time -- Wait Count for WAIT_REG_MEM Offloads.
+ * gws_wait_time -- Wait Count for Global Wave Syncs.
+ * que_sleep_wait_time -- Wait Count for Dequeue Retry.
+ * sch_wave_wait_time -- Wait Count for Scheduling Wave Message.
+ * sem_rearm_wait_time -- Wait Count for Semaphore re-arm.
+ * deq_retry_wait_time -- Wait Count for Global Wave Syncs.
+ */
+void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev,
+ uint32_t *wait_times)
+
+{
+ *wait_times = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2));
+}
+
+void kgd_gfx_v10_build_grace_period_packet_info(struct amdgpu_device *adev,
+ uint32_t wait_times,
+ uint32_t grace_period,
+ uint32_t *reg_offset,
+ uint32_t *reg_data)
+{
+ *reg_data = wait_times;
+
+ /*
+ * The CP cannont handle a 0 grace period input and will result in
+ * an infinite grace period being set so set to 1 to prevent this.
+ */
+ if (grace_period == 0)
+ grace_period = 1;
+
+ *reg_data = REG_SET_FIELD(*reg_data,
+ CP_IQ_WAIT_TIME2,
+ SCH_WAVE,
+ grace_period);
+
+ *reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2);
+}
+
static void program_trap_handler_settings(struct amdgpu_device *adev,
uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr,
uint32_t inst)
@@ -848,5 +889,7 @@ const struct kfd2kgd_calls gfx_v10_kfd2kgd = {
.set_vm_context_page_table_base = set_vm_context_page_table_base,
.enable_debug_trap = kgd_gfx_v10_enable_debug_trap,
.disable_debug_trap = kgd_gfx_v10_disable_debug_trap,
+ .get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times,
+ .build_grace_period_packet_info = kgd_gfx_v10_build_grace_period_packet_info,
.program_trap_handler_settings = program_trap_handler_settings,
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
index 251d61fbde07..1e993a213646 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
@@ -26,3 +26,9 @@ uint32_t kgd_gfx_v10_enable_debug_trap(struct amdgpu_device *adev,
uint32_t kgd_gfx_v10_disable_debug_trap(struct amdgpu_device *adev,
bool keep_trap_enabled,
uint32_t vmid);
+void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev, uint32_t *wait_times);
+void kgd_gfx_v10_build_grace_period_packet_info(struct amdgpu_device *adev,
+ uint32_t wait_times,
+ uint32_t grace_period,
+ uint32_t *reg_offset,
+ uint32_t *reg_data);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
index 8b293f3dcbd2..387bdf4823c9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
@@ -672,6 +672,8 @@ const struct kfd2kgd_calls gfx_v10_3_kfd2kgd = {
.get_atc_vmid_pasid_mapping_info = get_atc_vmid_pasid_mapping_info_v10_3,
.set_vm_context_page_table_base = set_vm_context_page_table_base_v10_3,
.program_trap_handler_settings = program_trap_handler_settings_v10_3,
+ .get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times,
+ .build_grace_period_packet_info = kgd_gfx_v10_build_grace_period_packet_info,
.enable_debug_trap = kgd_gfx_v10_enable_debug_trap,
.disable_debug_trap = kgd_gfx_v10_disable_debug_trap
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
index 8d7d04704b00..829ee720cc44 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -739,6 +739,24 @@ uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
return 0;
}
+/* kgd_gfx_v9_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
+ * The values read are:
+ * ib_offload_wait_time -- Wait Count for Indirect Buffer Offloads.
+ * atomic_offload_wait_time -- Wait Count for L2 and GDS Atomics Offloads.
+ * wrm_offload_wait_time -- Wait Count for WAIT_REG_MEM Offloads.
+ * gws_wait_time -- Wait Count for Global Wave Syncs.
+ * que_sleep_wait_time -- Wait Count for Dequeue Retry.
+ * sch_wave_wait_time -- Wait Count for Scheduling Wave Message.
+ * sem_rearm_wait_time -- Wait Count for Semaphore re-arm.
+ * deq_retry_wait_time -- Wait Count for Global Wave Syncs.
+ */
+void kgd_gfx_v9_get_iq_wait_times(struct amdgpu_device *adev,
+ uint32_t *wait_times)
+
+{
+ *wait_times = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2));
+}
+
void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev,
uint32_t vmid, uint64_t page_table_base)
{
@@ -926,6 +944,29 @@ void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid,
adev->gfx.cu_info.max_waves_per_simd;
}
+void kgd_gfx_v9_build_grace_period_packet_info(struct amdgpu_device *adev,
+ uint32_t wait_times,
+ uint32_t grace_period,
+ uint32_t *reg_offset,
+ uint32_t *reg_data)
+{
+ *reg_data = wait_times;
+
+ /*
+ * The CP cannont handle a 0 grace period input and will result in
+ * an infinite grace period being set so set to 1 to prevent this.
+ */
+ if (grace_period == 0)
+ grace_period = 1;
+
+ *reg_data = REG_SET_FIELD(*reg_data,
+ CP_IQ_WAIT_TIME2,
+ SCH_WAVE,
+ grace_period);
+
+ *reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2);
+}
+
void kgd_gfx_v9_program_trap_handler_settings(struct amdgpu_device *adev,
uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr, uint32_t inst)
{
@@ -969,6 +1010,8 @@ const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
.set_vm_context_page_table_base = kgd_gfx_v9_set_vm_context_page_table_base,
.enable_debug_trap = kgd_gfx_v9_enable_debug_trap,
.disable_debug_trap = kgd_gfx_v9_disable_debug_trap,
+ .get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
+ .build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
index 9588ff055393..fed5b7f18b1a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
@@ -20,8 +20,6 @@
* OTHER DEALINGS IN THE SOFTWARE.
*/
-
-
void kgd_gfx_v9_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid,
uint32_t sh_mem_config,
uint32_t sh_mem_ape1_base, uint32_t sh_mem_ape1_limit,
@@ -73,3 +71,9 @@ uint32_t kgd_gfx_v9_enable_debug_trap(struct amdgpu_device *adev,
uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
bool keep_trap_enabled,
uint32_t vmid);
+void kgd_gfx_v9_get_iq_wait_times(struct amdgpu_device *adev, uint32_t *wait_times);
+void kgd_gfx_v9_build_grace_period_packet_info(struct amdgpu_device *adev,
+ uint32_t wait_times,
+ uint32_t grace_period,
+ uint32_t *reg_offset,
+ uint32_t *reg_data);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index 2baa0781eafc..0b88a64e61fe 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -46,10 +46,13 @@ static int set_pasid_vmid_mapping(struct device_queue_manager *dqm,
static int execute_queues_cpsch(struct device_queue_manager *dqm,
enum kfd_unmap_queues_filter filter,
- uint32_t filter_param);
+ uint32_t filter_param,
+ uint32_t grace_period);
static int unmap_queues_cpsch(struct device_queue_manager *dqm,
enum kfd_unmap_queues_filter filter,
- uint32_t filter_param, bool reset);
+ uint32_t filter_param,
+ uint32_t grace_period,
+ bool reset);
static int map_queues_cpsch(struct device_queue_manager *dqm);
@@ -866,7 +869,7 @@ static int update_queue(struct device_queue_manager *dqm, struct queue *q,
if (dqm->sched_policy != KFD_SCHED_POLICY_NO_HWS) {
if (!dqm->dev->kfd->shared_resources.enable_mes)
retval = unmap_queues_cpsch(dqm,
- KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0, false);
+ KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0, USE_DEFAULT_GRACE_PERIOD, false);
else if (prev_active)
retval = remove_queue_mes(dqm, q, &pdd->qpd);
@@ -1042,7 +1045,8 @@ static int evict_process_queues_cpsch(struct device_queue_manager *dqm,
retval = execute_queues_cpsch(dqm,
qpd->is_debug ?
KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES :
- KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
+ KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0,
+ USE_DEFAULT_GRACE_PERIOD);
out:
dqm_unlock(dqm);
@@ -1182,8 +1186,7 @@ static int restore_process_queues_cpsch(struct device_queue_manager *dqm,
}
if (!dqm->dev->kfd->shared_resources.enable_mes)
retval = execute_queues_cpsch(dqm,
- KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
-
+ KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0, USE_DEFAULT_GRACE_PERIOD);
eviction_duration = get_jiffies_64() - pdd->last_evict_timestamp;
atomic64_add(eviction_duration, &pdd->evict_duration_counter);
vm_not_acquired:
@@ -1525,6 +1528,9 @@ static int initialize_cpsch(struct device_queue_manager *dqm)
init_sdma_bitmaps(dqm);
+ if (dqm->dev->kfd2kgd->get_iq_wait_times)
+ dqm->dev->kfd2kgd->get_iq_wait_times(dqm->dev->adev,
+ &dqm->wait_times);
return 0;
}
@@ -1563,8 +1569,9 @@ static int start_cpsch(struct device_queue_manager *dqm)
dqm->is_hws_hang = false;
dqm->is_resetting = false;
dqm->sched_running = true;
+
if (!dqm->dev->kfd->shared_resources.enable_mes)
- execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
+ execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0, USE_DEFAULT_GRACE_PERIOD);
dqm_unlock(dqm);
return 0;
@@ -1589,7 +1596,7 @@ static int stop_cpsch(struct device_queue_manager *dqm)
if (!dqm->is_hws_hang) {
if (!dqm->dev->kfd->shared_resources.enable_mes)
- unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0, false);
+ unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0, USE_DEFAULT_GRACE_PERIOD, false);
else
remove_all_queues_mes(dqm);
}
@@ -1631,7 +1638,8 @@ static int create_kernel_queue_cpsch(struct device_queue_manager *dqm,
list_add(&kq->list, &qpd->priv_queue_list);
increment_queue_count(dqm, qpd, kq->queue);
qpd->is_debug = true;
- execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
+ execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0,
+ USE_DEFAULT_GRACE_PERIOD);
dqm_unlock(dqm);
return 0;
@@ -1645,7 +1653,8 @@ static void destroy_kernel_queue_cpsch(struct device_queue_manager *dqm,
list_del(&kq->list);
decrement_queue_count(dqm, qpd, kq->queue);
qpd->is_debug = false;
- execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
+ execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0,
+ USE_DEFAULT_GRACE_PERIOD);
/*
* Unconditionally decrement this counter, regardless of the queue's
* type.
@@ -1722,7 +1731,7 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
if (!dqm->dev->kfd->shared_resources.enable_mes)
retval = execute_queues_cpsch(dqm,
- KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
+ KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0, USE_DEFAULT_GRACE_PERIOD);
else
retval = add_queue_mes(dqm, q, qpd);
if (retval)
@@ -1811,7 +1820,9 @@ static int map_queues_cpsch(struct device_queue_manager *dqm)
/* dqm->lock mutex has to be locked before calling this function */
static int unmap_queues_cpsch(struct device_queue_manager *dqm,
enum kfd_unmap_queues_filter filter,
- uint32_t filter_param, bool reset)
+ uint32_t filter_param,
+ uint32_t grace_period,
+ bool reset)
{
int retval = 0;
struct mqd_manager *mqd_mgr;
@@ -1823,6 +1834,12 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm,
if (!dqm->active_runlist)
return retval;
+ if (grace_period != USE_DEFAULT_GRACE_PERIOD) {
+ retval = pm_update_grace_period(&dqm->packet_mgr, grace_period);
+ if (retval)
+ return retval;
+ }
+
retval = pm_send_unmap_queue(&dqm->packet_mgr, filter, filter_param, reset);
if (retval)
return retval;
@@ -1855,6 +1872,13 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm,
return -ETIME;
}
+ /* We need to reset the grace period value for this device */
+ if (grace_period != USE_DEFAULT_GRACE_PERIOD) {
+ if (pm_update_grace_period(&dqm->packet_mgr,
+ USE_DEFAULT_GRACE_PERIOD))
+ pr_err("Failed to reset grace period\n");
+ }
+
pm_release_ib(&dqm->packet_mgr);
dqm->active_runlist = false;
@@ -1870,7 +1894,7 @@ static int reset_queues_cpsch(struct device_queue_manager *dqm,
dqm_lock(dqm);
retval = unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_BY_PASID,
- pasid, true);
+ pasid, USE_DEFAULT_GRACE_PERIOD, true);
dqm_unlock(dqm);
return retval;
@@ -1879,13 +1903,14 @@ static int reset_queues_cpsch(struct device_queue_manager *dqm,
/* dqm->lock mutex has to be locked before calling this function */
static int execute_queues_cpsch(struct device_queue_manager *dqm,
enum kfd_unmap_queues_filter filter,
- uint32_t filter_param)
+ uint32_t filter_param,
+ uint32_t grace_period)
{
int retval;
if (dqm->is_hws_hang)
return -EIO;
- retval = unmap_queues_cpsch(dqm, filter, filter_param, false);
+ retval = unmap_queues_cpsch(dqm, filter, filter_param, grace_period, false);
if (retval)
return retval;
@@ -1943,7 +1968,8 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm,
if (!dqm->dev->kfd->shared_resources.enable_mes) {
decrement_queue_count(dqm, qpd, q);
retval = execute_queues_cpsch(dqm,
- KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
+ KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0,
+ USE_DEFAULT_GRACE_PERIOD);
if (retval == -ETIME)
qpd->reset_wavefronts = true;
} else {
@@ -2228,7 +2254,7 @@ static int process_termination_cpsch(struct device_queue_manager *dqm,
}
if (!dqm->dev->kfd->shared_resources.enable_mes)
- retval = execute_queues_cpsch(dqm, filter, 0);
+ retval = execute_queues_cpsch(dqm, filter, 0, USE_DEFAULT_GRACE_PERIOD);
if ((!dqm->is_hws_hang) && (retval || qpd->reset_wavefronts)) {
pr_warn("Resetting wave fronts (cpsch) on dev %p\n", dqm->dev);
@@ -2589,7 +2615,8 @@ int dqm_debugfs_hang_hws(struct device_queue_manager *dqm)
return r;
}
dqm->active_runlist = true;
- r = execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
+ r = execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES,
+ 0, USE_DEFAULT_GRACE_PERIOD);
dqm_unlock(dqm);
return r;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
index cd4383bb207f..d4dd3b4acbf0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
@@ -37,6 +37,7 @@
#define KFD_MES_PROCESS_QUANTUM 100000
#define KFD_MES_GANG_QUANTUM 10000
+#define USE_DEFAULT_GRACE_PERIOD 0xffffffff
struct device_process_node {
struct qcm_process_device *qpd;
@@ -259,6 +260,8 @@ struct device_queue_manager {
/* used for GFX 9.4.3 only */
uint32_t current_logical_xcc_start;
+
+ uint32_t wait_times;
};
void device_queue_manager_init_cik(
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
index 2f54172e9175..401096c103b2 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
@@ -370,6 +370,38 @@ out:
return retval;
}
+int pm_update_grace_period(struct packet_manager *pm, uint32_t grace_period)
+{
+ int retval = 0;
+ uint32_t *buffer, size;
+
+ size = pm->pmf->set_grace_period_size;
+
+ mutex_lock(&pm->lock);
+
+ if (size) {
+ kq_acquire_packet_buffer(pm->priv_queue,
+ size / sizeof(uint32_t),
+ (unsigned int **)&buffer);
+
+ if (!buffer) {
+ pr_err("Failed to allocate buffer on kernel queue\n");
+ retval = -ENOMEM;
+ goto out;
+ }
+
+ retval = pm->pmf->set_grace_period(pm, buffer, grace_period);
+ if (!retval)
+ kq_submit_packet(pm->priv_queue);
+ else
+ kq_rollback_packet(pm->priv_queue);
+ }
+
+out:
+ mutex_unlock(&pm->lock);
+ return retval;
+}
+
int pm_send_unmap_queue(struct packet_manager *pm,
enum kfd_unmap_queues_filter filter,
uint32_t filter_param, bool reset)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c
index 44cf3a5f6fdb..1fda6dcf84b1 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c
@@ -262,6 +262,41 @@ static int pm_map_queues_v9(struct packet_manager *pm, uint32_t *buffer,
return 0;
}
+static int pm_set_grace_period_v9(struct packet_manager *pm,
+ uint32_t *buffer,
+ uint32_t grace_period)
+{
+ struct pm4_mec_write_data_mmio *packet;
+ uint32_t reg_offset = 0;
+ uint32_t reg_data = 0;
+
+ pm->dqm->dev->kfd2kgd->build_grace_period_packet_info(
+ pm->dqm->dev->adev,
+ pm->dqm->wait_times,
+ grace_period,
+ ®_offset,
+ ®_data);
+
+ if (grace_period == USE_DEFAULT_GRACE_PERIOD)
+ reg_data = pm->dqm->wait_times;
+
+ packet = (struct pm4_mec_write_data_mmio *)buffer;
+ memset(buffer, 0, sizeof(struct pm4_mec_write_data_mmio));
+
+ packet->header.u32All = pm_build_pm4_header(IT_WRITE_DATA,
+ sizeof(struct pm4_mec_write_data_mmio));
+
+ packet->bitfields2.dst_sel = dst_sel___write_data__mem_mapped_register;
+ packet->bitfields2.addr_incr =
+ addr_incr___write_data__do_not_increment_address;
+
+ packet->bitfields3.dst_mmreg_addr = reg_offset;
+
+ packet->data = reg_data;
+
+ return 0;
+}
+
static int pm_unmap_queues_v9(struct packet_manager *pm, uint32_t *buffer,
enum kfd_unmap_queues_filter filter,
uint32_t filter_param, bool reset)
@@ -345,6 +380,7 @@ const struct packet_manager_funcs kfd_v9_pm_funcs = {
.set_resources = pm_set_resources_v9,
.map_queues = pm_map_queues_v9,
.unmap_queues = pm_unmap_queues_v9,
+ .set_grace_period = pm_set_grace_period_v9,
.query_status = pm_query_status_v9,
.release_mem = NULL,
.map_process_size = sizeof(struct pm4_mes_map_process),
@@ -352,6 +388,7 @@ const struct packet_manager_funcs kfd_v9_pm_funcs = {
.set_resources_size = sizeof(struct pm4_mes_set_resources),
.map_queues_size = sizeof(struct pm4_mes_map_queues),
.unmap_queues_size = sizeof(struct pm4_mes_unmap_queues),
+ .set_grace_period_size = sizeof(struct pm4_mec_write_data_mmio),
.query_status_size = sizeof(struct pm4_mes_query_status),
.release_mem_size = 0,
};
@@ -362,6 +399,7 @@ const struct packet_manager_funcs kfd_aldebaran_pm_funcs = {
.set_resources = pm_set_resources_v9,
.map_queues = pm_map_queues_v9,
.unmap_queues = pm_unmap_queues_v9,
+ .set_grace_period = pm_set_grace_period_v9,
.query_status = pm_query_status_v9,
.release_mem = NULL,
.map_process_size = sizeof(struct pm4_mes_map_process_aldebaran),
@@ -369,6 +407,7 @@ const struct packet_manager_funcs kfd_aldebaran_pm_funcs = {
.set_resources_size = sizeof(struct pm4_mes_set_resources),
.map_queues_size = sizeof(struct pm4_mes_map_queues),
.unmap_queues_size = sizeof(struct pm4_mes_unmap_queues),
+ .set_grace_period_size = sizeof(struct pm4_mec_write_data_mmio),
.query_status_size = sizeof(struct pm4_mes_query_status),
.release_mem_size = 0,
};
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_vi.c
index faf4772ed317..c1199d06d131 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_vi.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_vi.c
@@ -303,6 +303,7 @@ const struct packet_manager_funcs kfd_vi_pm_funcs = {
.set_resources = pm_set_resources_vi,
.map_queues = pm_map_queues_vi,
.unmap_queues = pm_unmap_queues_vi,
+ .set_grace_period = NULL,
.query_status = pm_query_status_vi,
.release_mem = pm_release_mem_vi,
.map_process_size = sizeof(struct pm4_mes_map_process),
@@ -310,6 +311,7 @@ const struct packet_manager_funcs kfd_vi_pm_funcs = {
.set_resources_size = sizeof(struct pm4_mes_set_resources),
.map_queues_size = sizeof(struct pm4_mes_map_queues),
.unmap_queues_size = sizeof(struct pm4_mes_unmap_queues),
+ .set_grace_period_size = 0,
.query_status_size = sizeof(struct pm4_mes_query_status),
.release_mem_size = sizeof(struct pm4_mec_release_mem)
};
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h
index 2ad708c64012..206f1960857f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h
@@ -584,6 +584,71 @@ struct pm4_mec_release_mem {
#endif
+#ifndef PM4_MEC_WRITE_DATA_DEFINED
+#define PM4_MEC_WRITE_DATA_DEFINED
+
+enum WRITE_DATA_dst_sel_enum {
+ dst_sel___write_data__mem_mapped_register = 0,
+ dst_sel___write_data__tc_l2 = 2,
+ dst_sel___write_data__gds = 3,
+ dst_sel___write_data__memory = 5,
+ dst_sel___write_data__memory_mapped_adc_persistent_state = 6,
+};
+
+enum WRITE_DATA_addr_incr_enum {
+ addr_incr___write_data__increment_address = 0,
+ addr_incr___write_data__do_not_increment_address = 1
+};
+
+enum WRITE_DATA_wr_confirm_enum {
+ wr_confirm___write_data__do_not_wait_for_write_confirmation = 0,
+ wr_confirm___write_data__wait_for_write_confirmation = 1
+};
+
+enum WRITE_DATA_cache_policy_enum {
+ cache_policy___write_data__lru = 0,
+ cache_policy___write_data__stream = 1
+};
+
+
+struct pm4_mec_write_data_mmio {
+ union {
+ union PM4_MES_TYPE_3_HEADER header; /*header */
+ unsigned int ordinal1;
+ };
+
+ union {
+ struct {
+ unsigned int reserved1:8;
+ unsigned int dst_sel:4;
+ unsigned int reserved2:4;
+ unsigned int addr_incr:1;
+ unsigned int reserved3:2;
+ unsigned int resume_vf:1;
+ unsigned int wr_confirm:1;
+ unsigned int reserved4:4;
+ unsigned int cache_policy:2;
+ unsigned int reserved5:5;
+ } bitfields2;
+ unsigned int ordinal2;
+ };
+
+ union {
+ struct {
+ unsigned int dst_mmreg_addr:18;
+ unsigned int reserved6:14;
+ } bitfields3;
+ unsigned int ordinal3;
+ };
+
+ uint32_t reserved7;
+
+ uint32_t data;
+
+};
+
+#endif
+
enum {
CACHE_FLUSH_AND_INV_TS_EVENT = 0x00000014
};
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 1b272f879b4c..4c912b7735b5 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -1350,6 +1350,8 @@ struct packet_manager_funcs {
int (*unmap_queues)(struct packet_manager *pm, uint32_t *buffer,
enum kfd_unmap_queues_filter mode,
uint32_t filter_param, bool reset);
+ int (*set_grace_period)(struct packet_manager *pm, uint32_t *buffer,
+ uint32_t grace_period);
int (*query_status)(struct packet_manager *pm, uint32_t *buffer,
uint64_t fence_address, uint64_t fence_value);
int (*release_mem)(uint64_t gpu_addr, uint32_t *buffer);
@@ -1360,6 +1362,7 @@ struct packet_manager_funcs {
int set_resources_size;
int map_queues_size;
int unmap_queues_size;
+ int set_grace_period_size;
int query_status_size;
int release_mem_size;
};
@@ -1382,6 +1385,8 @@ int pm_send_unmap_queue(struct packet_manager *pm,
void pm_release_ib(struct packet_manager *pm);
+int pm_update_grace_period(struct packet_manager *pm, uint32_t grace_period);
+
/* Following PM funcs can be shared among VI and AI */
unsigned int pm_build_pm4_header(unsigned int opcode, size_t packet_size);
--
cgit v1.2.3
From 101827e13026a981e887527620fe9710adc0e481 Mon Sep 17 00:00:00 2001
From: Jonathan Kim <jonathan.kim@amd.com>
Date: Wed, 27 Apr 2022 13:18:10 -0400
Subject: drm/amdkfd: add debug wave launch override operation
This operation allows the debugger to override the enabled HW
exceptions on the device.
On debug devices that only support the debugging of a single process,
the HW exceptions are global and set through the SPI_GDBG_TRAP_MASK
register.
Because they are global, only address watch exceptions are allowed to
be enabled. In other words, the debugger must preserve all non-address
watch exception states in normal mode operation by barring a full
replacement override or a non-address watch override request.
For multi-process debugging, all HW exception overrides are per-VMID so
all exceptions can be overridden or fully replaced.
In order for the debugger to know what is permissible, returned the
supported override mask back to the debugger along with the previously
enable overrides.
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>
---
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c | 47 ++++++++++++
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c | 2 +
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c | 55 ++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h | 10 +++
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c | 5 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c | 87 +++++++++++++++++++++-
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 55 ++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h | 10 +++
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 7 ++
drivers/gpu/drm/amd/amdkfd/kfd_debug.c | 69 +++++++++++++++++
drivers/gpu/drm/amd/amdkfd/kfd_debug.h | 6 ++
11 files changed, 351 insertions(+), 2 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
index b811a0985050..d7881bbd828d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
@@ -25,6 +25,7 @@
#include "amdgpu_amdkfd_gfx_v9.h"
#include "gc/gc_9_4_2_offset.h"
#include "gc/gc_9_4_2_sh_mask.h"
+#include <uapi/linux/kfd_ioctl.h>
/*
* Returns TRAP_EN, EXCP_EN and EXCP_REPLACE.
@@ -62,6 +63,50 @@ static uint32_t kgd_aldebaran_disable_debug_trap(struct amdgpu_device *adev,
return data;
}
+static int kgd_aldebaran_validate_trap_override_request(struct amdgpu_device *adev,
+ uint32_t trap_override,
+ uint32_t *trap_mask_supported)
+{
+ *trap_mask_supported &= KFD_DBG_TRAP_MASK_FP_INVALID |
+ KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+ KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+ KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+ KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+ KFD_DBG_TRAP_MASK_FP_INEXACT |
+ KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+ KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+ KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION;
+
+ if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR &&
+ trap_override != KFD_DBG_TRAP_OVERRIDE_REPLACE)
+ return -EPERM;
+
+ return 0;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_RPLACE. */
+static uint32_t kgd_aldebaran_set_wave_launch_trap_override(struct amdgpu_device *adev,
+ uint32_t vmid,
+ uint32_t trap_override,
+ uint32_t trap_mask_bits,
+ uint32_t trap_mask_request,
+ uint32_t *trap_mask_prev,
+ uint32_t kfd_dbg_trap_cntl_prev)
+
+{
+ uint32_t data = 0;
+
+ *trap_mask_prev = REG_GET_FIELD(kfd_dbg_trap_cntl_prev, SPI_GDBG_PER_VMID_CNTL, EXCP_EN);
+ trap_mask_bits = (trap_mask_bits & trap_mask_request) |
+ (*trap_mask_prev & ~trap_mask_request);
+
+ data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+ data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, trap_mask_bits);
+ data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, trap_override);
+
+ return data;
+}
+
const struct kfd2kgd_calls aldebaran_kfd2kgd = {
.program_sh_mem_settings = kgd_gfx_v9_program_sh_mem_settings,
.set_pasid_vmid_mapping = kgd_gfx_v9_set_pasid_vmid_mapping,
@@ -82,6 +127,8 @@ const struct kfd2kgd_calls aldebaran_kfd2kgd = {
.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
.enable_debug_trap = kgd_aldebaran_enable_debug_trap,
.disable_debug_trap = kgd_aldebaran_disable_debug_trap,
+ .validate_trap_override_request = kgd_aldebaran_validate_trap_override_request,
+ .set_wave_launch_trap_override = kgd_aldebaran_set_wave_launch_trap_override,
.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index a62bd0068515..ec2587664001 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -410,6 +410,8 @@ const struct kfd2kgd_calls arcturus_kfd2kgd = {
kgd_gfx_v9_set_vm_context_page_table_base,
.enable_debug_trap = kgd_arcturus_enable_debug_trap,
.disable_debug_trap = kgd_arcturus_disable_debug_trap,
+ .validate_trap_override_request = kgd_gfx_v9_validate_trap_override_request,
+ .set_wave_launch_trap_override = kgd_gfx_v9_set_wave_launch_trap_override,
.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
index 98006c7021dd..7ea0362dcab3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
@@ -32,6 +32,7 @@
#include "v10_structs.h"
#include "nv.h"
#include "nvd.h"
+#include <uapi/linux/kfd_ioctl.h>
enum hqd_dequeue_request_type {
NO_ACTION = 0,
@@ -803,6 +804,58 @@ uint32_t kgd_gfx_v10_disable_debug_trap(struct amdgpu_device *adev,
return 0;
}
+int kgd_gfx_v10_validate_trap_override_request(struct amdgpu_device *adev,
+ uint32_t trap_override,
+ uint32_t *trap_mask_supported)
+{
+ *trap_mask_supported &= KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH;
+
+ /* The SPI_GDBG_TRAP_MASK register is global and affects all
+ * processes. Only allow OR-ing the address-watch bit, since
+ * this only affects processes under the debugger. Other bits
+ * should stay 0 to avoid the debugger interfering with other
+ * processes.
+ */
+ if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR)
+ return -EINVAL;
+
+ return 0;
+}
+
+uint32_t kgd_gfx_v10_set_wave_launch_trap_override(struct amdgpu_device *adev,
+ uint32_t vmid,
+ uint32_t trap_override,
+ uint32_t trap_mask_bits,
+ uint32_t trap_mask_request,
+ uint32_t *trap_mask_prev,
+ uint32_t kfd_dbg_trap_cntl_prev)
+{
+ uint32_t data, wave_cntl_prev;
+
+ mutex_lock(&adev->grbm_idx_mutex);
+
+ wave_cntl_prev = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+
+ kgd_gfx_v10_set_wave_launch_stall(adev, vmid, true);
+
+ data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK));
+ *trap_mask_prev = REG_GET_FIELD(data, SPI_GDBG_TRAP_MASK, EXCP_EN);
+
+ trap_mask_bits = (trap_mask_bits & trap_mask_request) |
+ (*trap_mask_prev & ~trap_mask_request);
+
+ data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK, EXCP_EN, trap_mask_bits);
+ data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK, REPLACE, trap_override);
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), data);
+
+ /* We need to preserve wave launch mode stall settings. */
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), wave_cntl_prev);
+
+ mutex_unlock(&adev->grbm_idx_mutex);
+
+ return 0;
+}
+
/* kgd_gfx_v10_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
* The values read are:
* ib_offload_wait_time -- Wait Count for Indirect Buffer Offloads.
@@ -889,6 +942,8 @@ const struct kfd2kgd_calls gfx_v10_kfd2kgd = {
.set_vm_context_page_table_base = set_vm_context_page_table_base,
.enable_debug_trap = kgd_gfx_v10_enable_debug_trap,
.disable_debug_trap = kgd_gfx_v10_disable_debug_trap,
+ .validate_trap_override_request = kgd_gfx_v10_validate_trap_override_request,
+ .set_wave_launch_trap_override = kgd_gfx_v10_set_wave_launch_trap_override,
.get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v10_build_grace_period_packet_info,
.program_trap_handler_settings = program_trap_handler_settings,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
index 1e993a213646..57339fa12807 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
@@ -26,6 +26,16 @@ uint32_t kgd_gfx_v10_enable_debug_trap(struct amdgpu_device *adev,
uint32_t kgd_gfx_v10_disable_debug_trap(struct amdgpu_device *adev,
bool keep_trap_enabled,
uint32_t vmid);
+int kgd_gfx_v10_validate_trap_override_request(struct amdgpu_device *adev,
+ uint32_t trap_override,
+ uint32_t *trap_mask_supported);
+uint32_t kgd_gfx_v10_set_wave_launch_trap_override(struct amdgpu_device *adev,
+ uint32_t vmid,
+ uint32_t trap_override,
+ uint32_t trap_mask_bits,
+ uint32_t trap_mask_request,
+ uint32_t *trap_mask_prev,
+ uint32_t kfd_dbg_trap_cntl_prev);
void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev, uint32_t *wait_times);
void kgd_gfx_v10_build_grace_period_packet_info(struct amdgpu_device *adev,
uint32_t wait_times,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
index 387bdf4823c9..7120927fed15 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
@@ -675,5 +675,8 @@ const struct kfd2kgd_calls gfx_v10_3_kfd2kgd = {
.get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v10_build_grace_period_packet_info,
.enable_debug_trap = kgd_gfx_v10_enable_debug_trap,
- .disable_debug_trap = kgd_gfx_v10_disable_debug_trap
+ .disable_debug_trap = kgd_gfx_v10_disable_debug_trap,
+ .validate_trap_override_request = kgd_gfx_v10_validate_trap_override_request,
+ .set_wave_launch_trap_override = kgd_gfx_v10_set_wave_launch_trap_override
+
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
index cc954cf248ca..ae0c4707919f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
@@ -30,6 +30,7 @@
#include "soc15d.h"
#include "v11_structs.h"
#include "soc21.h"
+#include <uapi/linux/kfd_ioctl.h>
enum hqd_dequeue_request_type {
NO_ACTION = 0,
@@ -643,6 +644,88 @@ static uint32_t kgd_gfx_v11_disable_debug_trap(struct amdgpu_device *adev,
return data;
}
+static int kgd_gfx_v11_validate_trap_override_request(struct amdgpu_device *adev,
+ uint32_t trap_override,
+ uint32_t *trap_mask_supported)
+{
+ *trap_mask_supported &= KFD_DBG_TRAP_MASK_FP_INVALID |
+ KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+ KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+ KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+ KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+ KFD_DBG_TRAP_MASK_FP_INEXACT |
+ KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+ KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+ KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION;
+
+ if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(11, 0, 4))
+ *trap_mask_supported |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START |
+ KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+ if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR &&
+ trap_override != KFD_DBG_TRAP_OVERRIDE_REPLACE)
+ return -EPERM;
+
+ return 0;
+}
+
+static uint32_t trap_mask_map_sw_to_hw(uint32_t mask)
+{
+ uint32_t trap_on_start = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START) ? 1 : 0;
+ uint32_t trap_on_end = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END) ? 1 : 0;
+ uint32_t excp_en = mask & (KFD_DBG_TRAP_MASK_FP_INVALID |
+ KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+ KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+ KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+ KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+ KFD_DBG_TRAP_MASK_FP_INEXACT |
+ KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+ KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+ KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION);
+ uint32_t ret;
+
+ ret = REG_SET_FIELD(0, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, excp_en);
+ ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START, trap_on_start);
+ ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END, trap_on_end);
+
+ return ret;
+}
+
+static uint32_t trap_mask_map_hw_to_sw(uint32_t mask)
+{
+ uint32_t ret = REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, EXCP_EN);
+
+ if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START))
+ ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START;
+
+ if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END))
+ ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+ return ret;
+}
+
+/* Returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v11_set_wave_launch_trap_override(struct amdgpu_device *adev,
+ uint32_t vmid,
+ uint32_t trap_override,
+ uint32_t trap_mask_bits,
+ uint32_t trap_mask_request,
+ uint32_t *trap_mask_prev,
+ uint32_t kfd_dbg_trap_cntl_prev)
+{
+ uint32_t data = 0;
+
+ *trap_mask_prev = trap_mask_map_hw_to_sw(kfd_dbg_trap_cntl_prev);
+
+ data = (trap_mask_bits & trap_mask_request) | (*trap_mask_prev & ~trap_mask_request);
+ data = trap_mask_map_sw_to_hw(data);
+
+ data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+ data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, trap_override);
+
+ return data;
+}
+
const struct kfd2kgd_calls gfx_v11_kfd2kgd = {
.program_sh_mem_settings = program_sh_mem_settings_v11,
.set_pasid_vmid_mapping = set_pasid_vmid_mapping_v11,
@@ -660,5 +743,7 @@ const struct kfd2kgd_calls gfx_v11_kfd2kgd = {
.get_atc_vmid_pasid_mapping_info = NULL,
.set_vm_context_page_table_base = set_vm_context_page_table_base_v11,
.enable_debug_trap = kgd_gfx_v11_enable_debug_trap,
- .disable_debug_trap = kgd_gfx_v11_disable_debug_trap
+ .disable_debug_trap = kgd_gfx_v11_disable_debug_trap,
+ .validate_trap_override_request = kgd_gfx_v11_validate_trap_override_request,
+ .set_wave_launch_trap_override = kgd_gfx_v11_set_wave_launch_trap_override
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
index 829ee720cc44..ce5c256f0517 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -38,6 +38,7 @@
#include "soc15d.h"
#include "gfx_v9_0.h"
#include "amdgpu_amdkfd_gfx_v9.h"
+#include <uapi/linux/kfd_ioctl.h>
enum hqd_dequeue_request_type {
NO_ACTION = 0,
@@ -739,6 +740,58 @@ uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
return 0;
}
+int kgd_gfx_v9_validate_trap_override_request(struct amdgpu_device *adev,
+ uint32_t trap_override,
+ uint32_t *trap_mask_supported)
+{
+ *trap_mask_supported &= KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH;
+
+ /* The SPI_GDBG_TRAP_MASK register is global and affects all
+ * processes. Only allow OR-ing the address-watch bit, since
+ * this only affects processes under the debugger. Other bits
+ * should stay 0 to avoid the debugger interfering with other
+ * processes.
+ */
+ if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR)
+ return -EINVAL;
+
+ return 0;
+}
+
+uint32_t kgd_gfx_v9_set_wave_launch_trap_override(struct amdgpu_device *adev,
+ uint32_t vmid,
+ uint32_t trap_override,
+ uint32_t trap_mask_bits,
+ uint32_t trap_mask_request,
+ uint32_t *trap_mask_prev,
+ uint32_t kfd_dbg_cntl_prev)
+{
+ uint32_t data, wave_cntl_prev;
+
+ mutex_lock(&adev->grbm_idx_mutex);
+
+ wave_cntl_prev = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+
+ kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+ data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK));
+ *trap_mask_prev = REG_GET_FIELD(data, SPI_GDBG_TRAP_MASK, EXCP_EN);
+
+ trap_mask_bits = (trap_mask_bits & trap_mask_request) |
+ (*trap_mask_prev & ~trap_mask_request);
+
+ data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK, EXCP_EN, trap_mask_bits);
+ data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK, REPLACE, trap_override);
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), data);
+
+ /* We need to preserve wave launch mode stall settings. */
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), wave_cntl_prev);
+
+ mutex_unlock(&adev->grbm_idx_mutex);
+
+ return 0;
+}
+
/* kgd_gfx_v9_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
* The values read are:
* ib_offload_wait_time -- Wait Count for Indirect Buffer Offloads.
@@ -1010,6 +1063,8 @@ const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
.set_vm_context_page_table_base = kgd_gfx_v9_set_vm_context_page_table_base,
.enable_debug_trap = kgd_gfx_v9_enable_debug_trap,
.disable_debug_trap = kgd_gfx_v9_disable_debug_trap,
+ .validate_trap_override_request = kgd_gfx_v9_validate_trap_override_request,
+ .set_wave_launch_trap_override = kgd_gfx_v9_set_wave_launch_trap_override,
.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
index fed5b7f18b1a..76812ddd35b1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
@@ -71,6 +71,16 @@ uint32_t kgd_gfx_v9_enable_debug_trap(struct amdgpu_device *adev,
uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
bool keep_trap_enabled,
uint32_t vmid);
+int kgd_gfx_v9_validate_trap_override_request(struct amdgpu_device *adev,
+ uint32_t trap_override,
+ uint32_t *trap_mask_supported);
+uint32_t kgd_gfx_v9_set_wave_launch_trap_override(struct amdgpu_device *adev,
+ uint32_t vmid,
+ uint32_t trap_override,
+ uint32_t trap_mask_bits,
+ uint32_t trap_mask_request,
+ uint32_t *trap_mask_prev,
+ uint32_t kfd_dbg_trap_cntl_prev);
void kgd_gfx_v9_get_iq_wait_times(struct amdgpu_device *adev, uint32_t *wait_times);
void kgd_gfx_v9_build_grace_period_packet_info(struct amdgpu_device *adev,
uint32_t wait_times,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index dcf4b5f3886e..7fa249807671 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -2984,6 +2984,13 @@ static int kfd_ioctl_set_debug_trap(struct file *filep, struct kfd_process *p, v
args->set_exceptions_enabled.exception_mask);
break;
case KFD_IOC_DBG_TRAP_SET_WAVE_LAUNCH_OVERRIDE:
+ r = kfd_dbg_trap_set_wave_launch_override(target,
+ args->launch_override.override_mode,
+ args->launch_override.enable_mask,
+ args->launch_override.support_request_mask,
+ &args->launch_override.enable_mask,
+ &args->launch_override.support_request_mask);
+ break;
case KFD_IOC_DBG_TRAP_SET_WAVE_LAUNCH_MODE:
case KFD_IOC_DBG_TRAP_SUSPEND_QUEUES:
case KFD_IOC_DBG_TRAP_RESUME_QUEUES:
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
index 48a4e3cc2234..733390fb2459 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
@@ -522,6 +522,75 @@ int kfd_dbg_trap_enable(struct kfd_process *target, uint32_t fd,
return r;
}
+static int kfd_dbg_validate_trap_override_request(struct kfd_process *p,
+ uint32_t trap_override,
+ uint32_t trap_mask_request,
+ uint32_t *trap_mask_supported)
+{
+ int i = 0;
+
+ *trap_mask_supported = 0xffffffff;
+
+ for (i = 0; i < p->n_pdds; i++) {
+ struct kfd_process_device *pdd = p->pdds[i];
+ int err = pdd->dev->kfd2kgd->validate_trap_override_request(
+ pdd->dev->adev,
+ trap_override,
+ trap_mask_supported);
+
+ if (err)
+ return err;
+ }
+
+ if (trap_mask_request & ~*trap_mask_supported)
+ return -EACCES;
+
+ return 0;
+}
+
+int kfd_dbg_trap_set_wave_launch_override(struct kfd_process *target,
+ uint32_t trap_override,
+ uint32_t trap_mask_bits,
+ uint32_t trap_mask_request,
+ uint32_t *trap_mask_prev,
+ uint32_t *trap_mask_supported)
+{
+ int r = 0, i;
+
+ r = kfd_dbg_validate_trap_override_request(target,
+ trap_override,
+ trap_mask_request,
+ trap_mask_supported);
+
+ if (r)
+ return r;
+
+ for (i = 0; i < target->n_pdds; i++) {
+ struct kfd_process_device *pdd = target->pdds[i];
+
+ amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
+ pdd->spi_dbg_override = pdd->dev->kfd2kgd->set_wave_launch_trap_override(
+ pdd->dev->adev,
+ pdd->dev->vm_info.last_vmid_kfd,
+ trap_override,
+ trap_mask_bits,
+ trap_mask_request,
+ trap_mask_prev,
+ pdd->spi_dbg_override);
+ amdgpu_gfx_off_ctrl(pdd->dev->adev, true);
+
+ if (!pdd->dev->kfd->shared_resources.enable_mes)
+ r = debug_refresh_runlist(pdd->dev->dqm);
+ else
+ r = kfd_dbg_set_mes_debug_mode(pdd);
+
+ if (r)
+ break;
+ }
+
+ return r;
+}
+
void kfd_dbg_set_enabled_debug_exception_mask(struct kfd_process *target,
uint64_t exception_set_mask)
{
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
index 6c1054a08872..c9245221aa76 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
@@ -42,6 +42,12 @@ int kfd_dbg_trap_disable(struct kfd_process *target);
int kfd_dbg_trap_enable(struct kfd_process *target, uint32_t fd,
void __user *runtime_info,
uint32_t *runtime_info_size);
+int kfd_dbg_trap_set_wave_launch_override(struct kfd_process *target,
+ uint32_t trap_override,
+ uint32_t trap_mask_bits,
+ uint32_t trap_mask_request,
+ uint32_t *trap_mask_prev,
+ uint32_t *trap_mask_supported);
int kfd_dbg_send_exception_to_runtime(struct kfd_process *p,
unsigned int dev_id,
--
cgit v1.2.3
From aea1b4738bebd8092bd437ce0b03aa9587fc20a7 Mon Sep 17 00:00:00 2001
From: Jonathan Kim <jonathan.kim@amd.com>
Date: Mon, 2 May 2022 11:45:05 -0400
Subject: drm/amdkfd: add debug wave launch mode operation
Allow the debugger to set wave behaviour on to either normally operate,
halt at launch, trap on every instruction, terminate immediately or
stall on allocation.
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>
---
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c | 12 ++++++++
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c | 1 +
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c | 25 +++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h | 3 ++
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c | 3 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c | 14 ++++++++-
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 25 +++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h | 3 ++
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 3 ++
drivers/gpu/drm/amd/amdkfd/kfd_debug.c | 36 +++++++++++++++++++++-
drivers/gpu/drm/amd/amdkfd/kfd_debug.h | 2 ++
11 files changed, 124 insertions(+), 3 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
index d7881bbd828d..774ecfc3451a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
@@ -107,6 +107,17 @@ static uint32_t kgd_aldebaran_set_wave_launch_trap_override(struct amdgpu_device
return data;
}
+static uint32_t kgd_aldebaran_set_wave_launch_mode(struct amdgpu_device *adev,
+ uint8_t wave_launch_mode,
+ uint32_t vmid)
+{
+ uint32_t data = 0;
+
+ data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, LAUNCH_MODE, wave_launch_mode);
+
+ return data;
+}
+
const struct kfd2kgd_calls aldebaran_kfd2kgd = {
.program_sh_mem_settings = kgd_gfx_v9_program_sh_mem_settings,
.set_pasid_vmid_mapping = kgd_gfx_v9_set_pasid_vmid_mapping,
@@ -129,6 +140,7 @@ const struct kfd2kgd_calls aldebaran_kfd2kgd = {
.disable_debug_trap = kgd_aldebaran_disable_debug_trap,
.validate_trap_override_request = kgd_aldebaran_validate_trap_override_request,
.set_wave_launch_trap_override = kgd_aldebaran_set_wave_launch_trap_override,
+ .set_wave_launch_mode = kgd_aldebaran_set_wave_launch_mode,
.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index ec2587664001..fbdc1b7b1e42 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -412,6 +412,7 @@ const struct kfd2kgd_calls arcturus_kfd2kgd = {
.disable_debug_trap = kgd_arcturus_disable_debug_trap,
.validate_trap_override_request = kgd_gfx_v9_validate_trap_override_request,
.set_wave_launch_trap_override = kgd_gfx_v9_set_wave_launch_trap_override,
+ .set_wave_launch_mode = kgd_gfx_v9_set_wave_launch_mode,
.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
index 7ea0362dcab3..a7a6edda557f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
@@ -856,6 +856,30 @@ uint32_t kgd_gfx_v10_set_wave_launch_trap_override(struct amdgpu_device *adev,
return 0;
}
+uint32_t kgd_gfx_v10_set_wave_launch_mode(struct amdgpu_device *adev,
+ uint8_t wave_launch_mode,
+ uint32_t vmid)
+{
+ uint32_t data = 0;
+ bool is_mode_set = !!wave_launch_mode;
+
+ mutex_lock(&adev->grbm_idx_mutex);
+
+ kgd_gfx_v10_set_wave_launch_stall(adev, vmid, true);
+
+ data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
+ VMID_MASK, is_mode_set ? 1 << vmid : 0);
+ data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
+ MODE, is_mode_set ? wave_launch_mode : 0);
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL2), data);
+
+ kgd_gfx_v10_set_wave_launch_stall(adev, vmid, false);
+
+ mutex_unlock(&adev->grbm_idx_mutex);
+
+ return 0;
+}
+
/* kgd_gfx_v10_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
* The values read are:
* ib_offload_wait_time -- Wait Count for Indirect Buffer Offloads.
@@ -944,6 +968,7 @@ const struct kfd2kgd_calls gfx_v10_kfd2kgd = {
.disable_debug_trap = kgd_gfx_v10_disable_debug_trap,
.validate_trap_override_request = kgd_gfx_v10_validate_trap_override_request,
.set_wave_launch_trap_override = kgd_gfx_v10_set_wave_launch_trap_override,
+ .set_wave_launch_mode = kgd_gfx_v10_set_wave_launch_mode,
.get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v10_build_grace_period_packet_info,
.program_trap_handler_settings = program_trap_handler_settings,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
index 57339fa12807..3a6aca2b0eaa 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
@@ -36,6 +36,9 @@ uint32_t kgd_gfx_v10_set_wave_launch_trap_override(struct amdgpu_device *adev,
uint32_t trap_mask_request,
uint32_t *trap_mask_prev,
uint32_t kfd_dbg_trap_cntl_prev);
+uint32_t kgd_gfx_v10_set_wave_launch_mode(struct amdgpu_device *adev,
+ uint8_t wave_launch_mode,
+ uint32_t vmid);
void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev, uint32_t *wait_times);
void kgd_gfx_v10_build_grace_period_packet_info(struct amdgpu_device *adev,
uint32_t wait_times,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
index 7120927fed15..ed36b433a48b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
@@ -677,6 +677,7 @@ const struct kfd2kgd_calls gfx_v10_3_kfd2kgd = {
.enable_debug_trap = kgd_gfx_v10_enable_debug_trap,
.disable_debug_trap = kgd_gfx_v10_disable_debug_trap,
.validate_trap_override_request = kgd_gfx_v10_validate_trap_override_request,
- .set_wave_launch_trap_override = kgd_gfx_v10_set_wave_launch_trap_override
+ .set_wave_launch_trap_override = kgd_gfx_v10_set_wave_launch_trap_override,
+ .set_wave_launch_mode = kgd_gfx_v10_set_wave_launch_mode
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
index ae0c4707919f..9711d5128d09 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
@@ -726,6 +726,17 @@ static uint32_t kgd_gfx_v11_set_wave_launch_trap_override(struct amdgpu_device *
return data;
}
+static uint32_t kgd_gfx_v11_set_wave_launch_mode(struct amdgpu_device *adev,
+ uint8_t wave_launch_mode,
+ uint32_t vmid)
+{
+ uint32_t data = 0;
+
+ data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, LAUNCH_MODE, wave_launch_mode);
+
+ return data;
+}
+
const struct kfd2kgd_calls gfx_v11_kfd2kgd = {
.program_sh_mem_settings = program_sh_mem_settings_v11,
.set_pasid_vmid_mapping = set_pasid_vmid_mapping_v11,
@@ -745,5 +756,6 @@ const struct kfd2kgd_calls gfx_v11_kfd2kgd = {
.enable_debug_trap = kgd_gfx_v11_enable_debug_trap,
.disable_debug_trap = kgd_gfx_v11_disable_debug_trap,
.validate_trap_override_request = kgd_gfx_v11_validate_trap_override_request,
- .set_wave_launch_trap_override = kgd_gfx_v11_set_wave_launch_trap_override
+ .set_wave_launch_trap_override = kgd_gfx_v11_set_wave_launch_trap_override,
+ .set_wave_launch_mode = kgd_gfx_v11_set_wave_launch_mode
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
index ce5c256f0517..0acc0c18dfe6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -792,6 +792,30 @@ uint32_t kgd_gfx_v9_set_wave_launch_trap_override(struct amdgpu_device *adev,
return 0;
}
+uint32_t kgd_gfx_v9_set_wave_launch_mode(struct amdgpu_device *adev,
+ uint8_t wave_launch_mode,
+ uint32_t vmid)
+{
+ uint32_t data = 0;
+ bool is_mode_set = !!wave_launch_mode;
+
+ mutex_lock(&adev->grbm_idx_mutex);
+
+ kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+ data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
+ VMID_MASK, is_mode_set ? 1 << vmid : 0);
+ data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
+ MODE, is_mode_set ? wave_launch_mode : 0);
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL2), data);
+
+ kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+ mutex_unlock(&adev->grbm_idx_mutex);
+
+ return 0;
+}
+
/* kgd_gfx_v9_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
* The values read are:
* ib_offload_wait_time -- Wait Count for Indirect Buffer Offloads.
@@ -1065,6 +1089,7 @@ const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
.disable_debug_trap = kgd_gfx_v9_disable_debug_trap,
.validate_trap_override_request = kgd_gfx_v9_validate_trap_override_request,
.set_wave_launch_trap_override = kgd_gfx_v9_set_wave_launch_trap_override,
+ .set_wave_launch_mode = kgd_gfx_v9_set_wave_launch_mode,
.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
index 76812ddd35b1..18f4970ac8e4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
@@ -74,6 +74,9 @@ uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
int kgd_gfx_v9_validate_trap_override_request(struct amdgpu_device *adev,
uint32_t trap_override,
uint32_t *trap_mask_supported);
+uint32_t kgd_gfx_v9_set_wave_launch_mode(struct amdgpu_device *adev,
+ uint8_t wave_launch_mode,
+ uint32_t vmid);
uint32_t kgd_gfx_v9_set_wave_launch_trap_override(struct amdgpu_device *adev,
uint32_t vmid,
uint32_t trap_override,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 7fa249807671..a6570b124b2b 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -2992,6 +2992,9 @@ static int kfd_ioctl_set_debug_trap(struct file *filep, struct kfd_process *p, v
&args->launch_override.support_request_mask);
break;
case KFD_IOC_DBG_TRAP_SET_WAVE_LAUNCH_MODE:
+ r = kfd_dbg_trap_set_wave_launch_mode(target,
+ args->launch_mode.launch_mode);
+ break;
case KFD_IOC_DBG_TRAP_SUSPEND_QUEUES:
case KFD_IOC_DBG_TRAP_RESUME_QUEUES:
case KFD_IOC_DBG_TRAP_SET_NODE_ADDRESS_WATCH:
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
index 733390fb2459..53c3418562d4 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
@@ -301,8 +301,10 @@ void kfd_dbg_trap_deactivate(struct kfd_process *target, bool unwind, int unwind
{
int i;
- if (!unwind)
+ if (!unwind) {
cancel_work_sync(&target->debug_event_workarea);
+ kfd_dbg_trap_set_wave_launch_mode(target, 0);
+ }
for (i = 0; i < target->n_pdds; i++) {
struct kfd_process_device *pdd = target->pdds[i];
@@ -591,6 +593,38 @@ int kfd_dbg_trap_set_wave_launch_override(struct kfd_process *target,
return r;
}
+int kfd_dbg_trap_set_wave_launch_mode(struct kfd_process *target,
+ uint8_t wave_launch_mode)
+{
+ int r = 0, i;
+
+ if (wave_launch_mode != KFD_DBG_TRAP_WAVE_LAUNCH_MODE_NORMAL &&
+ wave_launch_mode != KFD_DBG_TRAP_WAVE_LAUNCH_MODE_HALT &&
+ wave_launch_mode != KFD_DBG_TRAP_WAVE_LAUNCH_MODE_DEBUG)
+ return -EINVAL;
+
+ for (i = 0; i < target->n_pdds; i++) {
+ struct kfd_process_device *pdd = target->pdds[i];
+
+ amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
+ pdd->spi_dbg_launch_mode = pdd->dev->kfd2kgd->set_wave_launch_mode(
+ pdd->dev->adev,
+ wave_launch_mode,
+ pdd->dev->vm_info.last_vmid_kfd);
+ amdgpu_gfx_off_ctrl(pdd->dev->adev, true);
+
+ if (!pdd->dev->kfd->shared_resources.enable_mes)
+ r = debug_refresh_runlist(pdd->dev->dqm);
+ else
+ r = kfd_dbg_set_mes_debug_mode(pdd);
+
+ if (r)
+ break;
+ }
+
+ return r;
+}
+
void kfd_dbg_set_enabled_debug_exception_mask(struct kfd_process *target,
uint64_t exception_set_mask)
{
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
index c9245221aa76..cb17869437c5 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
@@ -48,6 +48,8 @@ int kfd_dbg_trap_set_wave_launch_override(struct kfd_process *target,
uint32_t trap_mask_request,
uint32_t *trap_mask_prev,
uint32_t *trap_mask_supported);
+int kfd_dbg_trap_set_wave_launch_mode(struct kfd_process *target,
+ uint8_t wave_launch_mode);
int kfd_dbg_send_exception_to_runtime(struct kfd_process *p,
unsigned int dev_id,
--
cgit v1.2.3
From e0f85f4690d089cc1a60337decafb1acf7eec45e Mon Sep 17 00:00:00 2001
From: Jonathan Kim <jonathan.kim@amd.com>
Date: Fri, 6 May 2022 14:58:55 -0400
Subject: drm/amdkfd: add debug set and clear address watch points operation
Shader read, write and atomic memory operations can be alerted to the
debugger as an address watch exception.
Allow the debugger to pass in a watch point to a particular memory
address per device.
Note that there exists only 4 watch points per devices to date, so have
the KFD keep track of what watch points are allocated or not.
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>
---
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c | 51 ++++++++
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c | 2 +
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c | 78 ++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h | 8 ++
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c | 5 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c | 52 +++++++-
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 77 ++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h | 8 ++
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 24 ++++
drivers/gpu/drm/amd/amdkfd/kfd_debug.c | 136 +++++++++++++++++++++
drivers/gpu/drm/amd/amdkfd/kfd_debug.h | 8 +-
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 2 +
drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 6 +-
13 files changed, 452 insertions(+), 5 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
index 774ecfc3451a..efd6a72aab4e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
@@ -118,6 +118,55 @@ static uint32_t kgd_aldebaran_set_wave_launch_mode(struct amdgpu_device *adev,
return data;
}
+#define TCP_WATCH_STRIDE (regTCP_WATCH1_ADDR_H - regTCP_WATCH0_ADDR_H)
+static uint32_t kgd_gfx_aldebaran_set_address_watch(
+ struct amdgpu_device *adev,
+ uint64_t watch_address,
+ uint32_t watch_address_mask,
+ uint32_t watch_id,
+ uint32_t watch_mode,
+ uint32_t debug_vmid)
+{
+ uint32_t watch_address_high;
+ uint32_t watch_address_low;
+ uint32_t watch_address_cntl;
+
+ watch_address_cntl = 0;
+ watch_address_low = lower_32_bits(watch_address);
+ watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ MODE,
+ watch_mode);
+
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ MASK,
+ watch_address_mask >> 6);
+
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ VALID,
+ 1);
+
+ WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_H) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_high);
+
+ WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_L) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_low);
+
+ return watch_address_cntl;
+}
+
+uint32_t kgd_gfx_aldebaran_clear_address_watch(struct amdgpu_device *adev,
+ uint32_t watch_id)
+{
+ return 0;
+}
+
const struct kfd2kgd_calls aldebaran_kfd2kgd = {
.program_sh_mem_settings = kgd_gfx_v9_program_sh_mem_settings,
.set_pasid_vmid_mapping = kgd_gfx_v9_set_pasid_vmid_mapping,
@@ -141,6 +190,8 @@ const struct kfd2kgd_calls aldebaran_kfd2kgd = {
.validate_trap_override_request = kgd_aldebaran_validate_trap_override_request,
.set_wave_launch_trap_override = kgd_aldebaran_set_wave_launch_trap_override,
.set_wave_launch_mode = kgd_aldebaran_set_wave_launch_mode,
+ .set_address_watch = kgd_gfx_aldebaran_set_address_watch,
+ .clear_address_watch = kgd_gfx_aldebaran_clear_address_watch,
.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index fbdc1b7b1e42..6df215aba4c4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -413,6 +413,8 @@ const struct kfd2kgd_calls arcturus_kfd2kgd = {
.validate_trap_override_request = kgd_gfx_v9_validate_trap_override_request,
.set_wave_launch_trap_override = kgd_gfx_v9_set_wave_launch_trap_override,
.set_wave_launch_mode = kgd_gfx_v9_set_wave_launch_mode,
+ .set_address_watch = kgd_gfx_v9_set_address_watch,
+ .clear_address_watch = kgd_gfx_v9_clear_address_watch,
.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
index a7a6edda557f..8ad7a7779e14 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
@@ -880,6 +880,82 @@ uint32_t kgd_gfx_v10_set_wave_launch_mode(struct amdgpu_device *adev,
return 0;
}
+#define TCP_WATCH_STRIDE (mmTCP_WATCH1_ADDR_H - mmTCP_WATCH0_ADDR_H)
+uint32_t kgd_gfx_v10_set_address_watch(struct amdgpu_device *adev,
+ uint64_t watch_address,
+ uint32_t watch_address_mask,
+ uint32_t watch_id,
+ uint32_t watch_mode,
+ uint32_t debug_vmid)
+{
+ uint32_t watch_address_high;
+ uint32_t watch_address_low;
+ uint32_t watch_address_cntl;
+
+ watch_address_cntl = 0;
+
+ watch_address_low = lower_32_bits(watch_address);
+ watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ VMID,
+ debug_vmid);
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ MODE,
+ watch_mode);
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ MASK,
+ watch_address_mask >> 7);
+
+ /* Turning off this watch point until we set all the registers */
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ VALID,
+ 0);
+
+ WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_cntl);
+
+ WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_H) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_high);
+
+ WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_L) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_low);
+
+ /* Enable the watch point */
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ VALID,
+ 1);
+
+ WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_cntl);
+
+ return 0;
+}
+
+uint32_t kgd_gfx_v10_clear_address_watch(struct amdgpu_device *adev,
+ uint32_t watch_id)
+{
+ uint32_t watch_address_cntl;
+
+ watch_address_cntl = 0;
+
+ WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_cntl);
+
+ return 0;
+}
+
+
/* kgd_gfx_v10_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
* The values read are:
* ib_offload_wait_time -- Wait Count for Indirect Buffer Offloads.
@@ -969,6 +1045,8 @@ const struct kfd2kgd_calls gfx_v10_kfd2kgd = {
.validate_trap_override_request = kgd_gfx_v10_validate_trap_override_request,
.set_wave_launch_trap_override = kgd_gfx_v10_set_wave_launch_trap_override,
.set_wave_launch_mode = kgd_gfx_v10_set_wave_launch_mode,
+ .set_address_watch = kgd_gfx_v10_set_address_watch,
+ .clear_address_watch = kgd_gfx_v10_clear_address_watch,
.get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v10_build_grace_period_packet_info,
.program_trap_handler_settings = program_trap_handler_settings,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
index 3a6aca2b0eaa..e6b70196071a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
@@ -39,6 +39,14 @@ uint32_t kgd_gfx_v10_set_wave_launch_trap_override(struct amdgpu_device *adev,
uint32_t kgd_gfx_v10_set_wave_launch_mode(struct amdgpu_device *adev,
uint8_t wave_launch_mode,
uint32_t vmid);
+uint32_t kgd_gfx_v10_set_address_watch(struct amdgpu_device *adev,
+ uint64_t watch_address,
+ uint32_t watch_address_mask,
+ uint32_t watch_id,
+ uint32_t watch_mode,
+ uint32_t debug_vmid);
+uint32_t kgd_gfx_v10_clear_address_watch(struct amdgpu_device *adev,
+ uint32_t watch_id);
void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev, uint32_t *wait_times);
void kgd_gfx_v10_build_grace_period_packet_info(struct amdgpu_device *adev,
uint32_t wait_times,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
index ed36b433a48b..8c8437a4383f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
@@ -678,6 +678,7 @@ const struct kfd2kgd_calls gfx_v10_3_kfd2kgd = {
.disable_debug_trap = kgd_gfx_v10_disable_debug_trap,
.validate_trap_override_request = kgd_gfx_v10_validate_trap_override_request,
.set_wave_launch_trap_override = kgd_gfx_v10_set_wave_launch_trap_override,
- .set_wave_launch_mode = kgd_gfx_v10_set_wave_launch_mode
-
+ .set_wave_launch_mode = kgd_gfx_v10_set_wave_launch_mode,
+ .set_address_watch = kgd_gfx_v10_set_address_watch,
+ .clear_address_watch = kgd_gfx_v10_clear_address_watch
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
index 9711d5128d09..52efa690a3c2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
@@ -737,6 +737,54 @@ static uint32_t kgd_gfx_v11_set_wave_launch_mode(struct amdgpu_device *adev,
return data;
}
+#define TCP_WATCH_STRIDE (regTCP_WATCH1_ADDR_H - regTCP_WATCH0_ADDR_H)
+static uint32_t kgd_gfx_v11_set_address_watch(struct amdgpu_device *adev,
+ uint64_t watch_address,
+ uint32_t watch_address_mask,
+ uint32_t watch_id,
+ uint32_t watch_mode,
+ uint32_t debug_vmid)
+{
+ uint32_t watch_address_high;
+ uint32_t watch_address_low;
+ uint32_t watch_address_cntl;
+
+ watch_address_cntl = 0;
+ watch_address_low = lower_32_bits(watch_address);
+ watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ MODE,
+ watch_mode);
+
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ MASK,
+ watch_address_mask >> 7);
+
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ VALID,
+ 1);
+
+ WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_H) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_high);
+
+ WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_L) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_low);
+
+ return watch_address_cntl;
+}
+
+uint32_t kgd_gfx_v11_clear_address_watch(struct amdgpu_device *adev,
+ uint32_t watch_id)
+{
+ return 0;
+}
+
const struct kfd2kgd_calls gfx_v11_kfd2kgd = {
.program_sh_mem_settings = program_sh_mem_settings_v11,
.set_pasid_vmid_mapping = set_pasid_vmid_mapping_v11,
@@ -757,5 +805,7 @@ const struct kfd2kgd_calls gfx_v11_kfd2kgd = {
.disable_debug_trap = kgd_gfx_v11_disable_debug_trap,
.validate_trap_override_request = kgd_gfx_v11_validate_trap_override_request,
.set_wave_launch_trap_override = kgd_gfx_v11_set_wave_launch_trap_override,
- .set_wave_launch_mode = kgd_gfx_v11_set_wave_launch_mode
+ .set_wave_launch_mode = kgd_gfx_v11_set_wave_launch_mode,
+ .set_address_watch = kgd_gfx_v11_set_address_watch,
+ .clear_address_watch = kgd_gfx_v11_clear_address_watch
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
index 0acc0c18dfe6..51d93fb13ea3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -816,6 +816,81 @@ uint32_t kgd_gfx_v9_set_wave_launch_mode(struct amdgpu_device *adev,
return 0;
}
+#define TCP_WATCH_STRIDE (mmTCP_WATCH1_ADDR_H - mmTCP_WATCH0_ADDR_H)
+uint32_t kgd_gfx_v9_set_address_watch(struct amdgpu_device *adev,
+ uint64_t watch_address,
+ uint32_t watch_address_mask,
+ uint32_t watch_id,
+ uint32_t watch_mode,
+ uint32_t debug_vmid)
+{
+ uint32_t watch_address_high;
+ uint32_t watch_address_low;
+ uint32_t watch_address_cntl;
+
+ watch_address_cntl = 0;
+
+ watch_address_low = lower_32_bits(watch_address);
+ watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ VMID,
+ debug_vmid);
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ MODE,
+ watch_mode);
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ MASK,
+ watch_address_mask >> 6);
+
+ /* Turning off this watch point until we set all the registers */
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ VALID,
+ 0);
+
+ WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_cntl);
+
+ WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_H) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_high);
+
+ WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_L) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_low);
+
+ /* Enable the watch point */
+ watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+ TCP_WATCH0_CNTL,
+ VALID,
+ 1);
+
+ WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_cntl);
+
+ return 0;
+}
+
+uint32_t kgd_gfx_v9_clear_address_watch(struct amdgpu_device *adev,
+ uint32_t watch_id)
+{
+ uint32_t watch_address_cntl;
+
+ watch_address_cntl = 0;
+
+ WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+ (watch_id * TCP_WATCH_STRIDE)),
+ watch_address_cntl);
+
+ return 0;
+}
+
/* kgd_gfx_v9_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
* The values read are:
* ib_offload_wait_time -- Wait Count for Indirect Buffer Offloads.
@@ -1090,6 +1165,8 @@ const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
.validate_trap_override_request = kgd_gfx_v9_validate_trap_override_request,
.set_wave_launch_trap_override = kgd_gfx_v9_set_wave_launch_trap_override,
.set_wave_launch_mode = kgd_gfx_v9_set_wave_launch_mode,
+ .set_address_watch = kgd_gfx_v9_set_address_watch,
+ .clear_address_watch = kgd_gfx_v9_clear_address_watch,
.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
.build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
index 18f4970ac8e4..5f54bff0db49 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
@@ -84,6 +84,14 @@ uint32_t kgd_gfx_v9_set_wave_launch_trap_override(struct amdgpu_device *adev,
uint32_t trap_mask_request,
uint32_t *trap_mask_prev,
uint32_t kfd_dbg_trap_cntl_prev);
+uint32_t kgd_gfx_v9_set_address_watch(struct amdgpu_device *adev,
+ uint64_t watch_address,
+ uint32_t watch_address_mask,
+ uint32_t watch_id,
+ uint32_t watch_mode,
+ uint32_t debug_vmid);
+uint32_t kgd_gfx_v9_clear_address_watch(struct amdgpu_device *adev,
+ uint32_t watch_id);
void kgd_gfx_v9_get_iq_wait_times(struct amdgpu_device *adev, uint32_t *wait_times);
void kgd_gfx_v9_build_grace_period_packet_info(struct amdgpu_device *adev,
uint32_t wait_times,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 1fae97df7a1e..016724c82928 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -2880,6 +2880,7 @@ static int kfd_ioctl_set_debug_trap(struct file *filep, struct kfd_process *p, v
struct mm_struct *mm = NULL;
struct pid *pid = NULL;
struct kfd_process *target = NULL;
+ struct kfd_process_device *pdd = NULL;
int r = 0;
if (sched_policy == KFD_SCHED_POLICY_NO_HWS) {
@@ -2957,6 +2958,20 @@ static int kfd_ioctl_set_debug_trap(struct file *filep, struct kfd_process *p, v
goto unlock_out;
}
+ if (args->op == KFD_IOC_DBG_TRAP_SET_NODE_ADDRESS_WATCH ||
+ args->op == KFD_IOC_DBG_TRAP_CLEAR_NODE_ADDRESS_WATCH) {
+ int user_gpu_id = kfd_process_get_user_gpu_id(target,
+ args->op == KFD_IOC_DBG_TRAP_SET_NODE_ADDRESS_WATCH ?
+ args->set_node_address_watch.gpu_id :
+ args->clear_node_address_watch.gpu_id);
+
+ pdd = kfd_process_device_data_by_id(target, user_gpu_id);
+ if (user_gpu_id == -EINVAL || !pdd) {
+ r = -ENODEV;
+ goto unlock_out;
+ }
+ }
+
switch (args->op) {
case KFD_IOC_DBG_TRAP_ENABLE:
if (target != p)
@@ -3009,7 +3024,16 @@ static int kfd_ioctl_set_debug_trap(struct file *filep, struct kfd_process *p, v
(uint32_t *)args->resume_queues.queue_array_ptr);
break;
case KFD_IOC_DBG_TRAP_SET_NODE_ADDRESS_WATCH:
+ r = kfd_dbg_trap_set_dev_address_watch(pdd,
+ args->set_node_address_watch.address,
+ args->set_node_address_watch.mask,
+ &args->set_node_address_watch.id,
+ args->set_node_address_watch.mode);
+ break;
case KFD_IOC_DBG_TRAP_CLEAR_NODE_ADDRESS_WATCH:
+ r = kfd_dbg_trap_clear_dev_address_watch(pdd,
+ args->clear_node_address_watch.id);
+ break;
case KFD_IOC_DBG_TRAP_SET_FLAGS:
case KFD_IOC_DBG_TRAP_QUERY_DEBUG_EVENT:
case KFD_IOC_DBG_TRAP_QUERY_EXCEPTION_INFO:
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
index f4d3dfb35cb3..4b36cc8b5fb7 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
@@ -24,6 +24,8 @@
#include "kfd_device_queue_manager.h"
#include <linux/file.h>
+#define MAX_WATCH_ADDRESSES 4
+
void debug_event_write_work_handler(struct work_struct *work)
{
struct kfd_process *process;
@@ -289,6 +291,139 @@ int kfd_dbg_set_mes_debug_mode(struct kfd_process_device *pdd)
pdd->watch_points, flags);
}
+#define KFD_DEBUGGER_INVALID_WATCH_POINT_ID -1
+static int kfd_dbg_get_dev_watch_id(struct kfd_process_device *pdd, int *watch_id)
+{
+ int i;
+
+ *watch_id = KFD_DEBUGGER_INVALID_WATCH_POINT_ID;
+
+ spin_lock(&pdd->dev->kfd->watch_points_lock);
+
+ for (i = 0; i < MAX_WATCH_ADDRESSES; i++) {
+ /* device watchpoint in use so skip */
+ if ((pdd->dev->kfd->alloc_watch_ids >> i) & 0x1)
+ continue;
+
+ pdd->alloc_watch_ids |= 0x1 << i;
+ pdd->dev->kfd->alloc_watch_ids |= 0x1 << i;
+ *watch_id = i;
+ spin_unlock(&pdd->dev->kfd->watch_points_lock);
+ return 0;
+ }
+
+ spin_unlock(&pdd->dev->kfd->watch_points_lock);
+
+ return -ENOMEM;
+}
+
+static void kfd_dbg_clear_dev_watch_id(struct kfd_process_device *pdd, int watch_id)
+{
+ spin_lock(&pdd->dev->kfd->watch_points_lock);
+
+ /* process owns device watch point so safe to clear */
+ if ((pdd->alloc_watch_ids >> watch_id) & 0x1) {
+ pdd->alloc_watch_ids &= ~(0x1 << watch_id);
+ pdd->dev->kfd->alloc_watch_ids &= ~(0x1 << watch_id);
+ }
+
+ spin_unlock(&pdd->dev->kfd->watch_points_lock);
+}
+
+static bool kfd_dbg_owns_dev_watch_id(struct kfd_process_device *pdd, int watch_id)
+{
+ bool owns_watch_id = false;
+
+ spin_lock(&pdd->dev->kfd->watch_points_lock);
+ owns_watch_id = watch_id < MAX_WATCH_ADDRESSES &&
+ ((pdd->alloc_watch_ids >> watch_id) & 0x1);
+
+ spin_unlock(&pdd->dev->kfd->watch_points_lock);
+
+ return owns_watch_id;
+}
+
+int kfd_dbg_trap_clear_dev_address_watch(struct kfd_process_device *pdd,
+ uint32_t watch_id)
+{
+ int r;
+
+ if (!kfd_dbg_owns_dev_watch_id(pdd, watch_id))
+ return -EINVAL;
+
+ if (!pdd->dev->kfd->shared_resources.enable_mes) {
+ r = debug_lock_and_unmap(pdd->dev->dqm);
+ if (r)
+ return r;
+ }
+
+ amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
+ pdd->watch_points[watch_id] = pdd->dev->kfd2kgd->clear_address_watch(
+ pdd->dev->adev,
+ watch_id);
+ amdgpu_gfx_off_ctrl(pdd->dev->adev, true);
+
+ if (!pdd->dev->kfd->shared_resources.enable_mes)
+ r = debug_map_and_unlock(pdd->dev->dqm);
+ else
+ r = kfd_dbg_set_mes_debug_mode(pdd);
+
+ kfd_dbg_clear_dev_watch_id(pdd, watch_id);
+
+ return r;
+}
+
+int kfd_dbg_trap_set_dev_address_watch(struct kfd_process_device *pdd,
+ uint64_t watch_address,
+ uint32_t watch_address_mask,
+ uint32_t *watch_id,
+ uint32_t watch_mode)
+{
+ int r = kfd_dbg_get_dev_watch_id(pdd, watch_id);
+
+ if (r)
+ return r;
+
+ if (!pdd->dev->kfd->shared_resources.enable_mes) {
+ r = debug_lock_and_unmap(pdd->dev->dqm);
+ if (r) {
+ kfd_dbg_clear_dev_watch_id(pdd, *watch_id);
+ return r;
+ }
+ }
+
+ amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
+ pdd->watch_points[*watch_id] = pdd->dev->kfd2kgd->set_address_watch(
+ pdd->dev->adev,
+ watch_address,
+ watch_address_mask,
+ *watch_id,
+ watch_mode,
+ pdd->dev->vm_info.last_vmid_kfd);
+ amdgpu_gfx_off_ctrl(pdd->dev->adev, true);
+
+ if (!pdd->dev->kfd->shared_resources.enable_mes)
+ r = debug_map_and_unlock(pdd->dev->dqm);
+ else
+ r = kfd_dbg_set_mes_debug_mode(pdd);
+
+ /* HWS is broken so no point in HW rollback but release the watchpoint anyways */
+ if (r)
+ kfd_dbg_clear_dev_watch_id(pdd, *watch_id);
+
+ return 0;
+}
+
+static void kfd_dbg_clear_process_address_watch(struct kfd_process *target)
+{
+ int i, j;
+
+ for (i = 0; i < target->n_pdds; i++)
+ for (j = 0; j < MAX_WATCH_ADDRESSES; j++)
+ kfd_dbg_trap_clear_dev_address_watch(target->pdds[i], j);
+}
+
+
/* kfd_dbg_trap_deactivate:
* target: target process
* unwind: If this is unwinding a failed kfd_dbg_trap_enable()
@@ -303,6 +438,7 @@ void kfd_dbg_trap_deactivate(struct kfd_process *target, bool unwind, int unwind
if (!unwind) {
cancel_work_sync(&target->debug_event_workarea);
+ kfd_dbg_clear_process_address_watch(target);
kfd_dbg_trap_set_wave_launch_mode(target, 0);
}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
index cb17869437c5..7f0757c2af2c 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
@@ -50,7 +50,13 @@ int kfd_dbg_trap_set_wave_launch_override(struct kfd_process *target,
uint32_t *trap_mask_supported);
int kfd_dbg_trap_set_wave_launch_mode(struct kfd_process *target,
uint8_t wave_launch_mode);
-
+int kfd_dbg_trap_clear_dev_address_watch(struct kfd_process_device *pdd,
+ uint32_t watch_id);
+int kfd_dbg_trap_set_dev_address_watch(struct kfd_process_device *pdd,
+ uint64_t watch_address,
+ uint32_t watch_address_mask,
+ uint32_t *watch_id,
+ uint32_t watch_mode);
int kfd_dbg_send_exception_to_runtime(struct kfd_process *p,
unsigned int dev_id,
unsigned int queue_id,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 2c36bb578633..9fc9d32cb579 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -811,6 +811,8 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
if (kfd_resume_iommu(kfd))
goto kfd_resume_iommu_error;
+ spin_lock_init(&kfd->watch_points_lock);
+
kfd->init_complete = true;
dev_info(kfd_device, "added device %x:%x\n", kfd->adev->pdev->vendor,
kfd->adev->pdev->device);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 05da43bf233a..8ec87bc8ba82 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -381,6 +381,10 @@ struct kfd_dev {
struct kfd_node *nodes[MAX_KFD_NODES];
unsigned int num_nodes;
+
+ /* Track per device allocated watch points */
+ uint32_t alloc_watch_ids;
+ spinlock_t watch_points_lock;
};
enum kfd_mempool {
@@ -833,6 +837,7 @@ struct kfd_process_device {
uint32_t spi_dbg_override;
uint32_t spi_dbg_launch_mode;
uint32_t watch_points[4];
+ uint32_t alloc_watch_ids;
/*
* If this process has been checkpointed before, then the user
@@ -989,7 +994,6 @@ struct kfd_process {
struct semaphore runtime_enable_sema;
bool is_runtime_retry;
struct kfd_runtime_info runtime_info;
-
};
#define KFD_PROCESS_TABLE_SIZE 5 /* bits: 32 entries */
--
cgit v1.2.3
From 9bd443cb74bd47d820c3cc31ee0ed3008d004d73 Mon Sep 17 00:00:00 2001
From: Jonathan Kim <jonathan.kim@amd.com>
Date: Fri, 2 Jun 2023 13:52:04 -0400
Subject: drm/amdgpu: fix debug wait on idle for gfx9.4.1
Wait calls for amd_ip_block_type not amd_hw_ip_block_type.
Reported-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
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>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index 6df215aba4c4..625db444df1c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -309,7 +309,7 @@ out:
if (!suspend || r)
return r;
- return amdgpu_device_ip_wait_for_idle(adev, GC_HWIP);
+ return amdgpu_device_ip_wait_for_idle(adev, AMD_IP_BLOCK_TYPE_GFX);
}
static void set_barrier_auto_waitcnt(struct amdgpu_device *adev, bool enable_waitcnt)
--
cgit v1.2.3