drm/amdkfd: Have kfd driver use same PASID values from graphic driver

Current kfd driver has its own PASID value for a kfd process and uses it to
locate vm at interrupt handler or mapping between kfd process and vm. That
design is not working when a physical gpu device has multiple spatial
partitions, ex: adev in CPX mode. This patch has kfd driver use same pasid
values that graphic driver generated which is per vm per pasid.

These pasid values are passed to fw/hardware. We do not need change interrupt
handler though more pasid values are used. Also, pasid values at log are
replaced by user process pid; pasid values are not exposed to user. Users see
their process pids that have meaning in user space.

Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Author: xiaogang-chen-amd

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h

@@ -47,6 +47,7 @@ enum TLB_FLUSH_TYPE {
 };
 
 struct amdgpu_device;
+struct kfd_process_device;
 struct amdgpu_reset_context;
 
 enum kfd_mem_attachment_type {
@@ -299,8 +300,6 @@ bool amdgpu_amdkfd_compute_active(struct amdgpu_device *adev, uint32_t node_id);
 	(&((struct amdgpu_fpriv *)					\
 		((struct drm_file *)(drm_priv))->driver_priv)->vm)
 
-int amdgpu_amdkfd_gpuvm_set_vm_pasid(struct amdgpu_device *adev,
-				     struct amdgpu_vm *avm, u32 pasid);
 int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev,
 					struct amdgpu_vm *avm,
 					void **process_info,

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c

@@ -1529,27 +1529,6 @@ static void amdgpu_amdkfd_gpuvm_unpin_bo(struct amdgpu_bo *bo)
 	amdgpu_bo_unreserve(bo);
 }
 
-int amdgpu_amdkfd_gpuvm_set_vm_pasid(struct amdgpu_device *adev,
-				     struct amdgpu_vm *avm, u32 pasid)
-
-{
-	int ret;
-
-	/* Free the original amdgpu allocated pasid,
-	 * will be replaced with kfd allocated pasid.
-	 */
-	if (avm->pasid) {
-		amdgpu_pasid_free(avm->pasid);
-		amdgpu_vm_set_pasid(adev, avm, 0);
-	}
-
-	ret = amdgpu_vm_set_pasid(adev, avm, pasid);
-	if (ret)
-		return ret;
-
-	return 0;
-}
-
 int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev,
 					   struct amdgpu_vm *avm,
 					   void **process_info,

drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c

@@ -107,20 +107,30 @@ static void cik_event_interrupt_wq(struct kfd_node *dev,
 		kfd_signal_hw_exception_event(pasid);
 	else if (ihre->source_id == CIK_INTSRC_GFX_PAGE_INV_FAULT ||
 		ihre->source_id == CIK_INTSRC_GFX_MEM_PROT_FAULT) {
+		struct kfd_process_device *pdd = NULL;
 		struct kfd_vm_fault_info info;
+		struct kfd_process *p;
 
 		kfd_smi_event_update_vmfault(dev, pasid);
-		kfd_dqm_evict_pasid(dev->dqm, pasid);
+		p = kfd_lookup_process_by_pasid(pasid, &pdd);
+		if (!pdd)
+			return;
+
+		kfd_evict_process_device(pdd);
 
 		memset(&info, 0, sizeof(info));
 		amdgpu_amdkfd_gpuvm_get_vm_fault_info(dev->adev, &info);
-		if (!info.page_addr && !info.status)
+		if (!info.page_addr && !info.status) {
+			kfd_unref_process(p);
 			return;
+		}
 
 		if (info.vmid == vmid)
-			kfd_signal_vm_fault_event(dev, pasid, &info, NULL);
+			kfd_signal_vm_fault_event(pdd, &info, NULL);
 		else
-			kfd_signal_vm_fault_event(dev, pasid, NULL, NULL);
+			kfd_signal_vm_fault_event(pdd, &info, NULL);
+
+		kfd_unref_process(p);
 	}
 }
 

drivers/gpu/drm/amd/amdkfd/kfd_chardev.c

@@ -155,8 +155,8 @@ static int kfd_open(struct inode *inode, struct file *filep)
 	/* filep now owns the reference returned by kfd_create_process */
 	filep->private_data = process;
 
-	dev_dbg(kfd_device, "process %d opened, compat mode (32 bit) - %d\n",
-		process->pasid, process->is_32bit_user_mode);
+	dev_dbg(kfd_device, "process pid %d opened kfd node, compat mode (32 bit) - %d\n",
+		process->lead_thread->pid, process->is_32bit_user_mode);
 
 	return 0;
 }
@@ -361,8 +361,8 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p,
 		goto err_acquire_queue_buf;
 	}
 
-	pr_debug("Creating queue for PASID 0x%x on gpu 0x%x\n",
-			p->pasid,
+	pr_debug("Creating queue for process pid %d on gpu 0x%x\n",
+			p->lead_thread->pid,
 			dev->id);
 
 	err = pqm_create_queue(&p->pqm, dev, &q_properties, &queue_id,
@@ -415,9 +415,9 @@ static int kfd_ioctl_destroy_queue(struct file *filp, struct kfd_process *p,
 	int retval;
 	struct kfd_ioctl_destroy_queue_args *args = data;
 
-	pr_debug("Destroying queue id %d for pasid 0x%x\n",
+	pr_debug("Destroying queue id %d for process pid %d\n",
 				args->queue_id,
-				p->pasid);
+				p->lead_thread->pid);
 
 	mutex_lock(&p->mutex);
 
@@ -468,8 +468,8 @@ static int kfd_ioctl_update_queue(struct file *filp, struct kfd_process *p,
 	properties.pm4_target_xcc = (args->queue_percentage >> 8) & 0xFF;
 	properties.priority = args->queue_priority;
 
-	pr_debug("Updating queue id %d for pasid 0x%x\n",
-			args->queue_id, p->pasid);
+	pr_debug("Updating queue id %d for process pid %d\n",
+			args->queue_id, p->lead_thread->pid);
 
 	mutex_lock(&p->mutex);
 
@@ -695,7 +695,7 @@ static int kfd_ioctl_get_process_apertures(struct file *filp,
 	struct kfd_process_device_apertures *pAperture;
 	int i;
 
-	dev_dbg(kfd_device, "get apertures for PASID 0x%x", p->pasid);
+	dev_dbg(kfd_device, "get apertures for process pid %d", p->lead_thread->pid);
 
 	args->num_of_nodes = 0;
 
@@ -747,7 +747,8 @@ static int kfd_ioctl_get_process_apertures_new(struct file *filp,
 	int ret;
 	int i;
 
-	dev_dbg(kfd_device, "get apertures for PASID 0x%x", p->pasid);
+	dev_dbg(kfd_device, "get apertures for process pid %d",
+			p->lead_thread->pid);
 
 	if (args->num_of_nodes == 0) {
 		/* Return number of nodes, so that user space can alloacate
@@ -3365,12 +3366,12 @@ static int kfd_mmio_mmap(struct kfd_node *dev, struct kfd_process *process,
 
 	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
 
-	pr_debug("pasid 0x%x mapping mmio page\n"
+	pr_debug("process pid %d mapping mmio page\n"
 		 "     target user address == 0x%08llX\n"
 		 "     physical address    == 0x%08llX\n"
 		 "     vm_flags            == 0x%04lX\n"
 		 "     size                == 0x%04lX\n",
-		 process->pasid, (unsigned long long) vma->vm_start,
+		 process->lead_thread->pid, (unsigned long long) vma->vm_start,
 		 address, vma->vm_flags, PAGE_SIZE);
 
 	return io_remap_pfn_range(vma,

drivers/gpu/drm/amd/amdkfd/kfd_debug.c

@@ -204,11 +204,12 @@ bool kfd_set_dbg_ev_from_interrupt(struct kfd_node *dev,
 				   size_t exception_data_size)
 {
 	struct kfd_process *p;
+	struct kfd_process_device *pdd = NULL;
 	bool signaled_to_debugger_or_runtime = false;
 
-	p = kfd_lookup_process_by_pasid(pasid);
+	p = kfd_lookup_process_by_pasid(pasid, &pdd);
 
-	if (!p)
+	if (!pdd)
 		return false;
 
 	if (!kfd_dbg_ev_raise(trap_mask, p, dev, doorbell_id, true,
@@ -238,9 +239,8 @@ bool kfd_set_dbg_ev_from_interrupt(struct kfd_node *dev,
 
 			mutex_unlock(&p->mutex);
 		} else if (trap_mask & KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION)) {
-			kfd_dqm_evict_pasid(dev->dqm, p->pasid);
-			kfd_signal_vm_fault_event(dev, p->pasid, NULL,
-							exception_data);
+			kfd_evict_process_device(pdd);
+			kfd_signal_vm_fault_event(pdd, NULL, exception_data);
 
 			signaled_to_debugger_or_runtime = true;
 		}
@@ -276,8 +276,8 @@ int kfd_dbg_send_exception_to_runtime(struct kfd_process *p,
 		data = (struct kfd_hsa_memory_exception_data *)
 						pdd->vm_fault_exc_data;
 
-		kfd_dqm_evict_pasid(pdd->dev->dqm, p->pasid);
-		kfd_signal_vm_fault_event(pdd->dev, p->pasid, NULL, data);
+		kfd_evict_process_device(pdd);
+		kfd_signal_vm_fault_event(pdd, NULL, data);
 		error_reason &= ~KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION);
 	}
 

drivers/gpu/drm/amd/amdkfd/kfd_device.c

@@ -1565,7 +1565,7 @@ bool kgd2kfd_vmfault_fast_path(struct amdgpu_device *adev, struct amdgpu_iv_entr
 	u32 cam_index;
 
 	if (entry->ih == &adev->irq.ih_soft || entry->ih == &adev->irq.ih1) {
-		p = kfd_lookup_process_by_pasid(entry->pasid);
+		p = kfd_lookup_process_by_pasid(entry->pasid, NULL);
 		if (!p)
 			return true;