Merge tag 'drm-amdkfd-next-2016-09-19' of git://people.freedesktop.org/~gabbayo/linux into drm-next

This is amdkfd's pull request for kernel 4.9. It contains a fix to a possible
infinite loop bug and a couple of other minor "cleaning" patches.

* tag 'drm-amdkfd-next-2016-09-19' of git://people.freedesktop.org/~gabbayo/linux:
  drm/amdkfd: Pass 'struct queue_propertices' by reference
  drm/amdkfd: Unify multiple calls to pr_debug() into one
  drm/amdkfd: Fix possible infinite loop
  drm/amdkfd: Reuse function to find a process through pasid
  drm/amdkfd: Add some missing memset zero'ing in queue init func
  drm/amdkfd: Tidy up kfd_generate_gpu_id() uint64_t bitshift unpack

9 files changed, 65 insertions, 67 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
index 362bedc9e507..1a0a5f7cccbc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
@@ -103,11 +103,11 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 				uint32_t pipe_id, uint32_t queue_id);
 
 static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type,
-				unsigned int timeout, uint32_t pipe_id,
+				unsigned int utimeout, uint32_t pipe_id,
 				uint32_t queue_id);
 static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd);
 static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
-				unsigned int timeout);
+				unsigned int utimeout);
 static int kgd_address_watch_disable(struct kgd_dev *kgd);
 static int kgd_address_watch_execute(struct kgd_dev *kgd,
 					unsigned int watch_point_id,
@@ -437,11 +437,12 @@ static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
 }
 
 static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type,
-				unsigned int timeout, uint32_t pipe_id,
+				unsigned int utimeout, uint32_t pipe_id,
 				uint32_t queue_id)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t temp;
+	int timeout = utimeout;
 
 	acquire_queue(kgd, pipe_id, queue_id);
 	WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, 0);
@@ -452,9 +453,8 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type,
 		temp = RREG32(mmCP_HQD_ACTIVE);
 		if (temp & CP_HQD_ACTIVE__ACTIVE_MASK)
 			break;
-		if (timeout == 0) {
-			pr_err("kfd: cp queue preemption time out (%dms)\n",
-				temp);
+		if (timeout <= 0) {
+			pr_err("kfd: cp queue preemption time out.\n");
 			release_queue(kgd);
 			return -ETIME;
 		}
@@ -467,12 +467,13 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type,
 }
 
 static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
-				unsigned int timeout)
+				unsigned int utimeout)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct cik_sdma_rlc_registers *m;
 	uint32_t sdma_base_addr;
 	uint32_t temp;
+	int timeout = utimeout;
 
 	m = get_sdma_mqd(mqd);
 	sdma_base_addr = get_sdma_base_addr(m);
@@ -485,7 +486,7 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
 		temp = RREG32(sdma_base_addr + mmSDMA0_RLC0_CONTEXT_STATUS);
 		if (temp & SDMA0_STATUS_REG__RB_CMD_IDLE__SHIFT)
 			break;
-		if (timeout == 0)
+		if (timeout <= 0)
 			return -ETIME;
 		msleep(20);
 		timeout -= 20;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
index 04b744d64b57..6697612239c2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
@@ -62,10 +62,10 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 		uint32_t pipe_id, uint32_t queue_id);
 static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd);
 static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type,
-				unsigned int timeout, uint32_t pipe_id,
+				unsigned int utimeout, uint32_t pipe_id,
 				uint32_t queue_id);
 static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
-				unsigned int timeout);
+				unsigned int utimeout);
 static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid);
 static int kgd_address_watch_disable(struct kgd_dev *kgd);
 static int kgd_address_watch_execute(struct kgd_dev *kgd,
@@ -349,11 +349,12 @@ static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
 }
 
 static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type,
-				unsigned int timeout, uint32_t pipe_id,
+				unsigned int utimeout, uint32_t pipe_id,
 				uint32_t queue_id)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t temp;
+	int timeout = utimeout;
 
 	acquire_queue(kgd, pipe_id, queue_id);
 
@@ -363,9 +364,8 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type,
 		temp = RREG32(mmCP_HQD_ACTIVE);
 		if (temp & CP_HQD_ACTIVE__ACTIVE_MASK)
 			break;
-		if (timeout == 0) {
-			pr_err("kfd: cp queue preemption time out (%dms)\n",
-				temp);
+		if (timeout <= 0) {
+			pr_err("kfd: cp queue preemption time out.\n");
 			release_queue(kgd);
 			return -ETIME;
 		}
@@ -378,12 +378,13 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type,
 }
 
 static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
-				unsigned int timeout)
+				unsigned int utimeout)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct cik_sdma_rlc_registers *m;
 	uint32_t sdma_base_addr;
 	uint32_t temp;
+	int timeout = utimeout;
 
 	m = get_sdma_mqd(mqd);
 	sdma_base_addr = get_sdma_base_addr(m);
@@ -396,7 +397,7 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
 		temp = RREG32(sdma_base_addr + mmSDMA0_RLC0_CONTEXT_STATUS);
 		if (temp & SDMA0_STATUS_REG__RB_CMD_IDLE__SHIFT)
 			break;
-		if (timeout == 0)
+		if (timeout <= 0)
 			return -ETIME;
 		msleep(20);
 		timeout -= 20;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
index e621eba63126..ad494b875311 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
@@ -142,13 +142,15 @@ int kfd_doorbell_mmap(struct kfd_process *process, struct vm_area_struct *vma)
 
 	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
 
-	pr_debug("mapping doorbell page:\n");
-	pr_debug("     target user address == 0x%08llX\n",
-			(unsigned long long) vma->vm_start);
-	pr_debug("     physical address    == 0x%08llX\n", address);
-	pr_debug("     vm_flags            == 0x%04lX\n", vma->vm_flags);
-	pr_debug("     size                == 0x%04lX\n",
-			 doorbell_process_allocation());
+	pr_debug("kfd: mapping doorbell page in %s\n"
+		 "     target user address == 0x%08llX\n"
+		 "     physical address    == 0x%08llX\n"
+		 "     vm_flags            == 0x%04lX\n"
+		 "     size                == 0x%04lX\n",
+		 __func__,
+		 (unsigned long long) vma->vm_start, address, vma->vm_flags,
+		 doorbell_process_allocation());
+
 
 	return io_remap_pfn_range(vma,
 				vma->vm_start,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c
index 9beae87aadd5..d135cd002a95 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c
@@ -47,6 +47,9 @@ static bool initialize(struct kernel_queue *kq, struct kfd_dev *dev,
 	pr_debug("amdkfd: In func %s initializing queue type %d size %d\n",
 			__func__, KFD_QUEUE_TYPE_HIQ, queue_size);
 
+	memset(&prop, 0, sizeof(prop));
+	memset(&nop, 0, sizeof(nop));
+
 	nop.opcode = IT_NOP;
 	nop.type = PM4_TYPE_3;
 	nop.u32all |= PM4_COUNT_ZERO;
@@ -121,7 +124,7 @@ static bool initialize(struct kernel_queue *kq, struct kfd_dev *dev,
 	prop.eop_ring_buffer_address = kq->eop_gpu_addr;
 	prop.eop_ring_buffer_size = PAGE_SIZE;
 
-	if (init_queue(&kq->queue, prop) != 0)
+	if (init_queue(&kq->queue, &prop) != 0)
 		goto err_init_queue;
 
 	kq->queue->device = dev;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 80113c335966..4750cabe4252 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -619,7 +619,7 @@ int kfd_init_apertures(struct kfd_process *process);
 /* Queue Context Management */
 struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd);
 
-int init_queue(struct queue **q, struct queue_properties properties);
+int init_queue(struct queue **q, const struct queue_properties *properties);
 void uninit_queue(struct queue *q);
 void print_queue_properties(struct queue_properties *q);
 void print_queue(struct queue *q);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 4f3849ac8c07..ef7c8de7060e 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -404,58 +404,47 @@ void kfd_unbind_process_from_device(struct kfd_dev *dev, unsigned int pasid)
 {
 	struct kfd_process *p;
 	struct kfd_process_device *pdd;
-	int idx, i;
 
 	BUG_ON(dev == NULL);
 
-	idx = srcu_read_lock(&kfd_processes_srcu);
-
 	/*
 	 * Look for the process that matches the pasid. If there is no such
 	 * process, we either released it in amdkfd's own notifier, or there
 	 * is a bug. Unfortunately, there is no way to tell...
 	 */
-	hash_for_each_rcu(kfd_processes_table, i, p, kfd_processes)
-		if (p->pasid == pasid) {
-
-			srcu_read_unlock(&kfd_processes_srcu, idx);
-
-			pr_debug("Unbinding process %d from IOMMU\n", pasid);
+	p = kfd_lookup_process_by_pasid(pasid);
+	if (!p)
+		return;
 
-			mutex_lock(&p->mutex);
-
-			if ((dev->dbgmgr) && (dev->dbgmgr->pasid == p->pasid))
-				kfd_dbgmgr_destroy(dev->dbgmgr);
-
-			pqm_uninit(&p->pqm);
+	pr_debug("Unbinding process %d from IOMMU\n", pasid);
 
-			pdd = kfd_get_process_device_data(dev, p);
+	if ((dev->dbgmgr) && (dev->dbgmgr->pasid == p->pasid))
+		kfd_dbgmgr_destroy(dev->dbgmgr);
 
-			if (!pdd) {
-				mutex_unlock(&p->mutex);
-				return;
-			}
+	pqm_uninit(&p->pqm);
 
-			if (pdd->reset_wavefronts) {
-				dbgdev_wave_reset_wavefronts(pdd->dev, p);
-				pdd->reset_wavefronts = false;
-			}
+	pdd = kfd_get_process_device_data(dev, p);
 
-			/*
-			 * Just mark pdd as unbound, because we still need it
-			 * to call amd_iommu_unbind_pasid() in when the
-			 * process exits.
-			 * We don't call amd_iommu_unbind_pasid() here
-			 * because the IOMMU called us.
-			 */
-			pdd->bound = false;
+	if (!pdd) {
+		mutex_unlock(&p->mutex);
+		return;
+	}
 
-			mutex_unlock(&p->mutex);
+	if (pdd->reset_wavefronts) {
+		dbgdev_wave_reset_wavefronts(pdd->dev, p);
+		pdd->reset_wavefronts = false;
+	}
 
-			return;
-		}
+	/*
+	 * Just mark pdd as unbound, because we still need it
+	 * to call amd_iommu_unbind_pasid() in when the
+	 * process exits.
+	 * We don't call amd_iommu_unbind_pasid() here
+	 * because the IOMMU called us.
+	 */
+	pdd->bound = false;
 
-	srcu_read_unlock(&kfd_processes_srcu, idx);
+	mutex_unlock(&p->mutex);
 }
 
 struct kfd_process_device *kfd_get_first_process_device_data(struct kfd_process *p)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
index 7b69070f7ecc..e1fb40b84c72 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
@@ -129,7 +129,7 @@ static int create_cp_queue(struct process_queue_manager *pqm,
 	q_properties->vmid = 0;
 	q_properties->queue_id = qid;
 
-	retval = init_queue(q, *q_properties);
+	retval = init_queue(q, q_properties);
 	if (retval != 0)
 		goto err_init_queue;
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_queue.c
index 9a0c90b0702e..0ab197077f2d 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_queue.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_queue.c
@@ -63,7 +63,7 @@ void print_queue(struct queue *q)
 	pr_debug("Queue Device Address: 0x%p\n", q->device);
 }
 
-int init_queue(struct queue **q, struct queue_properties properties)
+int init_queue(struct queue **q, const struct queue_properties *properties)
 {
 	struct queue *tmp;
 
@@ -73,7 +73,7 @@ int init_queue(struct queue **q, struct queue_properties properties)
 	if (!tmp)
 		return -ENOMEM;
 
-	memcpy(&tmp->properties, &properties, sizeof(struct queue_properties));
+	memcpy(&tmp->properties, properties, sizeof(struct queue_properties));
 
 	*q = tmp;
 	return 0;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index 884c96f50c3d..1e5064749959 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -1090,19 +1090,21 @@ static uint32_t kfd_generate_gpu_id(struct kfd_dev *gpu)
 {
 	uint32_t hashout;
 	uint32_t buf[7];
+	uint64_t local_mem_size;
 	int i;
 
 	if (!gpu)
 		return 0;
 
+	local_mem_size = gpu->kfd2kgd->get_vmem_size(gpu->kgd);
+
 	buf[0] = gpu->pdev->devfn;
 	buf[1] = gpu->pdev->subsystem_vendor;
 	buf[2] = gpu->pdev->subsystem_device;
 	buf[3] = gpu->pdev->device;
 	buf[4] = gpu->pdev->bus->number;
-	buf[5] = (uint32_t)(gpu->kfd2kgd->get_vmem_size(gpu->kgd)
-			& 0xffffffff);
-	buf[6] = (uint32_t)(gpu->kfd2kgd->get_vmem_size(gpu->kgd) >> 32);
+	buf[5] = lower_32_bits(local_mem_size);
+	buf[6] = upper_32_bits(local_mem_size);
 
 	for (i = 0, hashout = 0; i < 7; i++)
 		hashout ^= hash_32(buf[i], KFD_GPU_ID_HASH_WIDTH);