From: David Yat Sin <david.yat...@amd.com>

We need the SPI_GDBG_PER_VMID_CNTL.TRAP_EN bit to be set during PC
Sampling so that the TTMP registers are valid inside the sampling data.
runtime_info.ttmp_setup will be cleared when the user application
does the AMDKFD_IOC_RUNTIME_ENABLE ioctl without
KFD_RUNTIME_ENABLE_MODE_ENABLE_MASK flag on exit.

It is also not valid to have the debugger attached to a process while PC
sampling is enabled so adding some checks to prevent this.

Signed-off-by: David Yat Sin <david.yat...@amd.com>
Signed-off-by: James Zhu <james....@amd.com>
---
 drivers/gpu/drm/amd/amdkfd/kfd_chardev.c     | 30 ++++++--------------
 drivers/gpu/drm/amd/amdkfd/kfd_debug.c       | 26 +++++++++++++++++
 drivers/gpu/drm/amd/amdkfd/kfd_debug.h       |  3 ++
 drivers/gpu/drm/amd/amdkfd/kfd_pc_sampling.c | 13 +++++++++
 drivers/gpu/drm/amd/amdkfd/kfd_priv.h        |  3 ++
 5 files changed, 54 insertions(+), 21 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c 
b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index d9cac97c54c0..bc37f3ee2c66 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -2804,26 +2804,9 @@ static int runtime_enable(struct kfd_process *p, 
uint64_t r_debug,
 
        p->runtime_info.runtime_state = DEBUG_RUNTIME_STATE_ENABLED;
        p->runtime_info.r_debug = r_debug;
-       p->runtime_info.ttmp_setup = enable_ttmp_setup;
 
-       if (p->runtime_info.ttmp_setup) {
-               for (i = 0; i < p->n_pdds; i++) {
-                       struct kfd_process_device *pdd = p->pdds[i];
-
-                       if (!kfd_dbg_is_rlc_restore_supported(pdd->dev)) {
-                               amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
-                               pdd->dev->kfd2kgd->enable_debug_trap(
-                                               pdd->dev->adev,
-                                               true,
-                                               
pdd->dev->vm_info.last_vmid_kfd);
-                       } else if (kfd_dbg_is_per_vmid_supported(pdd->dev)) {
-                               pdd->spi_dbg_override = 
pdd->dev->kfd2kgd->enable_debug_trap(
-                                               pdd->dev->adev,
-                                               false,
-                                               0);
-                       }
-               }
-       }
+       if (enable_ttmp_setup)
+               kfd_dbg_enable_ttmp_setup(p);
 
 retry:
        if (p->debug_trap_enabled) {
@@ -2972,10 +2955,10 @@ static int kfd_ioctl_set_debug_trap(struct file *filep, 
struct kfd_process *p, v
                goto out;
        }
 
-       /* Check if target is still PTRACED. */
        rcu_read_lock();
+       /* Check if target is still PTRACED. */
        if (target != p && args->op != KFD_IOC_DBG_TRAP_DISABLE
-                               && ptrace_parent(target->lead_thread) != 
current) {
+                       && ptrace_parent(target->lead_thread) != current) {
                pr_err("PID %i is not PTRACED and cannot be debugged\n", 
args->pid);
                r = -EPERM;
        }
@@ -2985,6 +2968,11 @@ static int kfd_ioctl_set_debug_trap(struct file *filep, 
struct kfd_process *p, v
                goto out;
 
        mutex_lock(&target->mutex);
+       if (!!target->pc_sampling_ref) {
+               pr_debug("Cannot enable debug trap on PID:%d because PC 
Sampling active\n", args->pid);
+               r = -EBUSY;
+               goto unlock_out;
+       }
 
        if (args->op != KFD_IOC_DBG_TRAP_ENABLE && !target->debug_trap_enabled) 
{
                pr_err("PID %i not debug enabled for op %i\n", args->pid, 
args->op);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c 
b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
index d889e3545120..8d836c65c636 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
@@ -1120,3 +1120,29 @@ void kfd_dbg_set_enabled_debug_exception_mask(struct 
kfd_process *target,
 
        mutex_unlock(&target->event_mutex);
 }
+
+void kfd_dbg_enable_ttmp_setup(struct kfd_process *p)
+{
+       int i;
+
+       if (p->runtime_info.ttmp_setup)
+               return;
+
+       p->runtime_info.ttmp_setup = true;
+       for (i = 0; i < p->n_pdds; i++) {
+               struct kfd_process_device *pdd = p->pdds[i];
+
+               if (!kfd_dbg_is_rlc_restore_supported(pdd->dev)) {
+                       amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
+                       pdd->dev->kfd2kgd->enable_debug_trap(
+                                       pdd->dev->adev,
+                                       true,
+                                       pdd->dev->vm_info.last_vmid_kfd);
+               } else if (kfd_dbg_is_per_vmid_supported(pdd->dev)) {
+                       pdd->spi_dbg_override = 
pdd->dev->kfd2kgd->enable_debug_trap(
+                                       pdd->dev->adev,
+                                       false,
+                                       0);
+               }
+       }
+}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h 
b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
index fd0ff64d4184..d7ce0b119dd0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
@@ -90,6 +90,9 @@ int kfd_dbg_trap_device_snapshot(struct kfd_process *target,
 
 void kfd_dbg_set_enabled_debug_exception_mask(struct kfd_process *target,
                                        uint64_t exception_set_mask);
+
+void kfd_dbg_enable_ttmp_setup(struct kfd_process *p);
+
 /*
  * If GFX off is enabled, chips that do not support RLC restore for the debug
  * registers will disable GFX off temporarily for the entire debug session.
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pc_sampling.c 
b/drivers/gpu/drm/amd/amdkfd/kfd_pc_sampling.c
index 783844ddd82f..4777cfefbb08 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_pc_sampling.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_pc_sampling.c
@@ -24,6 +24,7 @@
 #include "kfd_priv.h"
 #include "amdgpu_amdkfd.h"
 #include "kfd_pc_sampling.h"
+#include "kfd_debug.h"
 #include "kfd_device_queue_manager.h"
 
 struct supported_pc_sample_info {
@@ -312,6 +313,14 @@ static int kfd_pc_sample_create(struct kfd_process_device 
*pdd,
        pcs_entry->pdd = pdd;
        user_args->trace_id = (uint32_t)i;
 
+       /*
+        * Set SPI_GDBG_PER_VMID_CNTL.TRAP_EN so that TTMP registers are valid 
in the sampling data
+        * p->runtime_info.ttmp_setup will be cleared when user application 
calls runtime_disable
+        * on exit.
+        */
+       kfd_dbg_enable_ttmp_setup(pdd->process);
+       pdd->process->pc_sampling_ref++;
+
        pr_debug("alloc pcs_entry = %p, trace_id = 0x%x on gpu 0x%x", 
pcs_entry, i, pdd->dev->id);
 
        return 0;
@@ -323,6 +332,7 @@ static int kfd_pc_sample_destroy(struct kfd_process_device 
*pdd, uint32_t trace_
        pr_debug("free pcs_entry = %p, trace_id = 0x%x on gpu 0x%x",
                pcs_entry, trace_id, pdd->dev->id);
 
+       pdd->process->pc_sampling_ref--;
        mutex_lock(&pdd->dev->pcs_data.mutex);
        pdd->dev->pcs_data.hosttrap_entry.base.use_count--;
        idr_remove(&pdd->dev->pcs_data.hosttrap_entry.base.pc_sampling_idr, 
trace_id);
@@ -381,6 +391,9 @@ int kfd_pc_sample(struct kfd_process_device *pdd,
                if (!pcs_entry ||
                        pcs_entry->pdd != pdd)
                        return -EINVAL;
+       } else if (pdd->process->debug_trap_enabled) {
+               pr_debug("Cannot have PC Sampling and debug trap 
simultaneously");
+               return -EBUSY;
        }
 
        switch (args->op) {
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h 
b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 372727b5d945..cce99478118b 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -1021,6 +1021,9 @@ struct kfd_process {
        struct semaphore runtime_enable_sema;
        bool is_runtime_retry;
        struct kfd_runtime_info runtime_info;
+
+       /* Indicates process' PC Sampling ref cnt*/
+       uint32_t pc_sampling_ref;
 };
 
 #define KFD_PROCESS_TABLE_SIZE 5 /* bits: 32 entries */
-- 
2.25.1

Reply via email to