Merge branch 'drm-next-4.15' of git://people.freedesktop.org/~agd5f/linux into drm-next

First feature pull for 4.15.  Highlights:
- Per VM BO support
- Lots of powerplay cleanups
- Powerplay support for CI
- pasid mgr for kfd
- interrupt infrastructure for recoverable page faults
- SR-IOV fixes
- initial GPU reset for vega10
- prime mmap support
- ttm page table debugging improvements
- lots of bug fixes

* 'drm-next-4.15' of git://people.freedesktop.org/~agd5f/linux: (232 commits)
  drm/amdgpu: clarify license in amdgpu_trace_points.c
  drm/amdgpu: Add gem_prime_mmap support
  drm/amd/powerplay: delete dead code in smumgr
  drm/amd/powerplay: delete SMUM_FIELD_MASK
  drm/amd/powerplay: delete SMUM_WAIT_INDIRECT_FIELD
  drm/amd/powerplay: delete SMUM_READ_FIELD
  drm/amd/powerplay: delete SMUM_SET_FIELD
  drm/amd/powerplay: delete SMUM_READ_VFPF_INDIRECT_FIELD
  drm/amd/powerplay: delete SMUM_WRITE_VFPF_INDIRECT_FIELD
  drm/amd/powerplay: delete SMUM_WRITE_FIELD
  drm/amd/powerplay: delete SMU_WRITE_INDIRECT_FIELD
  drm/amd/powerplay: move macros to hwmgr.h
  drm/amd/powerplay: move PHM_WAIT_VFPF_INDIRECT_FIELD to hwmgr.h
  drm/amd/powerplay: move SMUM_WAIT_VFPF_INDIRECT_FIELD_UNEQUAL to hwmgr.h
  drm/amd/powerplay: move SMUM_WAIT_INDIRECT_FIELD_UNEQUAL to hwmgr.h
  drm/amd/powerplay: add new helper functions in hwmgr.h
  drm/amd/powerplay: use SMU_IND_INDEX/DATA_11 pair
  drm/amd/powerplay: refine powerplay code.
  drm/amd/powerplay: delete dead code in hwmgr.h
  drm/amd/powerplay: refine interface in struct pp_smumgr_func
  ...

6 files changed, 83 insertions, 86 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 61fff25b4ce7..5df12b287201 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -168,13 +168,6 @@ static bool device_iommu_pasid_init(struct kfd_dev *kfd)
 	pasid_limit = min_t(unsigned int,
 			(unsigned int)(1 << kfd->device_info->max_pasid_bits),
 			iommu_info.max_pasids);
-	/*
-	 * last pasid is used for kernel queues doorbells
-	 * in the future the last pasid might be used for a kernel thread.
-	 */
-	pasid_limit = min_t(unsigned int,
-				pasid_limit,
-				kfd->doorbell_process_limit - 1);
 
 	err = amd_iommu_init_device(kfd->pdev, pasid_limit);
 	if (err < 0) {
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
index acf4d2a977ad..feb76c235b1a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
@@ -24,16 +24,15 @@
 #include <linux/mman.h>
 #include <linux/slab.h>
 #include <linux/io.h>
+#include <linux/idr.h>
 
 /*
- * This extension supports a kernel level doorbells management for
- * the kernel queues.
- * Basically the last doorbells page is devoted to kernel queues
- * and that's assures that any user process won't get access to the
- * kernel doorbells page
+ * This extension supports a kernel level doorbells management for the
+ * kernel queues using the first doorbell page reserved for the kernel.
  */
 
-#define KERNEL_DOORBELL_PASID 1
+static DEFINE_IDA(doorbell_ida);
+static unsigned int max_doorbell_slices;
 #define KFD_SIZE_OF_DOORBELL_IN_BYTES 4
 
 /*
@@ -84,13 +83,16 @@ int kfd_doorbell_init(struct kfd_dev *kfd)
 			(doorbell_aperture_size - doorbell_start_offset) /
 						doorbell_process_allocation();
 	else
-		doorbell_process_limit = 0;
+		return -ENOSPC;
+
+	if (!max_doorbell_slices ||
+	    doorbell_process_limit < max_doorbell_slices)
+		max_doorbell_slices = doorbell_process_limit;
 
 	kfd->doorbell_base = kfd->shared_resources.doorbell_physical_address +
 				doorbell_start_offset;
 
 	kfd->doorbell_id_offset = doorbell_start_offset / sizeof(u32);
-	kfd->doorbell_process_limit = doorbell_process_limit - 1;
 
 	kfd->doorbell_kernel_ptr = ioremap(kfd->doorbell_base,
 						doorbell_process_allocation());
@@ -185,11 +187,10 @@ u32 __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd,
 		return NULL;
 
 	/*
-	 * Calculating the kernel doorbell offset using "faked" kernel
-	 * pasid that allocated for kernel queues only
+	 * Calculating the kernel doorbell offset using the first
+	 * doorbell page.
 	 */
-	*doorbell_off = KERNEL_DOORBELL_PASID * (doorbell_process_allocation() /
-							sizeof(u32)) + inx;
+	*doorbell_off = kfd->doorbell_id_offset + inx;
 
 	pr_debug("Get kernel queue doorbell\n"
 			 "     doorbell offset   == 0x%08X\n"
@@ -228,11 +229,12 @@ unsigned int kfd_queue_id_to_doorbell(struct kfd_dev *kfd,
 {
 	/*
 	 * doorbell_id_offset accounts for doorbells taken by KGD.
-	 * pasid * doorbell_process_allocation/sizeof(u32) adjusts
-	 * to the process's doorbells
+	 * index * doorbell_process_allocation/sizeof(u32) adjusts to
+	 * the process's doorbells.
 	 */
 	return kfd->doorbell_id_offset +
-		process->pasid * (doorbell_process_allocation()/sizeof(u32)) +
+		process->doorbell_index
+		* doorbell_process_allocation() / sizeof(u32) +
 		queue_id;
 }
 
@@ -250,5 +252,21 @@ phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev,
 					struct kfd_process *process)
 {
 	return dev->doorbell_base +
-		process->pasid * doorbell_process_allocation();
+		process->doorbell_index * doorbell_process_allocation();
+}
+
+int kfd_alloc_process_doorbells(struct kfd_process *process)
+{
+	int r = ida_simple_get(&doorbell_ida, 1, max_doorbell_slices,
+				GFP_KERNEL);
+	if (r > 0)
+		process->doorbell_index = r;
+
+	return r;
+}
+
+void kfd_free_process_doorbells(struct kfd_process *process)
+{
+	if (process->doorbell_index)
+		ida_simple_remove(&doorbell_ida, process->doorbell_index);
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c
index 0d73bea22c45..6c5a9cab55de 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c
@@ -103,10 +103,6 @@ static int __init kfd_module_init(void)
 		return -1;
 	}
 
-	err = kfd_pasid_init();
-	if (err < 0)
-		return err;
-
 	err = kfd_chardev_init();
 	if (err < 0)
 		goto err_ioctl;
@@ -126,7 +122,6 @@ static int __init kfd_module_init(void)
 err_topology:
 	kfd_chardev_exit();
 err_ioctl:
-	kfd_pasid_exit();
 	return err;
 }
 
@@ -137,7 +132,6 @@ static void __exit kfd_module_exit(void)
 	kfd_process_destroy_wq();
 	kfd_topology_shutdown();
 	kfd_chardev_exit();
-	kfd_pasid_exit();
 	dev_info(kfd_device, "Removed module\n");
 }
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c b/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c
index 1e06de0bc673..d6a796144269 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c
@@ -20,78 +20,64 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  */
 
-#include <linux/slab.h>
 #include <linux/types.h>
 #include "kfd_priv.h"
 
-static unsigned long *pasid_bitmap;
-static unsigned int pasid_limit;
-static DEFINE_MUTEX(pasid_mutex);
-
-int kfd_pasid_init(void)
-{
-	pasid_limit = KFD_MAX_NUM_OF_PROCESSES;
-
-	pasid_bitmap = kcalloc(BITS_TO_LONGS(pasid_limit), sizeof(long),
-				GFP_KERNEL);
-	if (!pasid_bitmap)
-		return -ENOMEM;
-
-	set_bit(0, pasid_bitmap); /* PASID 0 is reserved. */
-
-	return 0;
-}
-
-void kfd_pasid_exit(void)
-{
-	kfree(pasid_bitmap);
-}
+static unsigned int pasid_bits = 16;
+static const struct kfd2kgd_calls *kfd2kgd;
 
 bool kfd_set_pasid_limit(unsigned int new_limit)
 {
-	if (new_limit < pasid_limit) {
-		bool ok;
-
-		mutex_lock(&pasid_mutex);
-
-		/* ensure that no pasids >= new_limit are in-use */
-		ok = (find_next_bit(pasid_bitmap, pasid_limit, new_limit) ==
-								pasid_limit);
-		if (ok)
-			pasid_limit = new_limit;
-
-		mutex_unlock(&pasid_mutex);
-
-		return ok;
+	if (new_limit < 2)
+		return false;
+
+	if (new_limit < (1U << pasid_bits)) {
+		if (kfd2kgd)
+			/* We've already allocated user PASIDs, too late to
+			 * change the limit
+			 */
+			return false;
+
+		while (new_limit < (1U << pasid_bits))
+			pasid_bits--;
 	}
 
 	return true;
 }
 
-inline unsigned int kfd_get_pasid_limit(void)
+unsigned int kfd_get_pasid_limit(void)
 {
-	return pasid_limit;
+	return 1U << pasid_bits;
 }
 
 unsigned int kfd_pasid_alloc(void)
 {
-	unsigned int found;
-
-	mutex_lock(&pasid_mutex);
-
-	found = find_first_zero_bit(pasid_bitmap, pasid_limit);
-	if (found == pasid_limit)
-		found = 0;
-	else
-		set_bit(found, pasid_bitmap);
+	int r;
+
+	/* Find the first best KFD device for calling KGD */
+	if (!kfd2kgd) {
+		struct kfd_dev *dev = NULL;
+		unsigned int i = 0;
+
+		while ((dev = kfd_topology_enum_kfd_devices(i)) != NULL) {
+			if (dev && dev->kfd2kgd) {
+				kfd2kgd = dev->kfd2kgd;
+				break;
+			}
+			i++;
+		}
+
+		if (!kfd2kgd)
+			return false;
+	}
 
-	mutex_unlock(&pasid_mutex);
+	r = kfd2kgd->alloc_pasid(pasid_bits);
 
-	return found;
+	return r > 0 ? r : 0;
 }
 
 void kfd_pasid_free(unsigned int pasid)
 {
-	if (!WARN_ON(pasid == 0 || pasid >= pasid_limit))
-		clear_bit(pasid, pasid_bitmap);
+	if (kfd2kgd)
+		kfd2kgd->free_pasid(pasid);
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index b397ec726400..4cb90f517906 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -157,9 +157,6 @@ struct kfd_dev {
 					 * to HW doorbell, GFX reserved some
 					 * at the start)
 					 */
-	size_t doorbell_process_limit;	/* Number of processes we have doorbell
-					 * space for.
-					 */
 	u32 __iomem *doorbell_kernel_ptr; /* This is a pointer for a doorbells
 					   * page used by kernel queue
 					   */
@@ -495,6 +492,7 @@ struct kfd_process {
 	struct rcu_head	rcu;
 
 	unsigned int pasid;
+	unsigned int doorbell_index;
 
 	/*
 	 * List of kfd_process_device structures,
@@ -583,6 +581,10 @@ void write_kernel_doorbell(u32 __iomem *db, u32 value);
 unsigned int kfd_queue_id_to_doorbell(struct kfd_dev *kfd,
 					struct kfd_process *process,
 					unsigned int queue_id);
+phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev,
+					struct kfd_process *process);
+int kfd_alloc_process_doorbells(struct kfd_process *process);
+void kfd_free_process_doorbells(struct kfd_process *process);
 
 /* GTT Sub-Allocator */
 
@@ -694,8 +696,6 @@ int pm_send_unmap_queue(struct packet_manager *pm, enum kfd_queue_type type,
 void pm_release_ib(struct packet_manager *pm);
 
 uint64_t kfd_get_number_elems(struct kfd_dev *kfd);
-phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev,
-					struct kfd_process *process);
 
 /* Events */
 extern const struct kfd_event_interrupt_class event_interrupt_class_cik;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index c74cf22a1ed9..9e65ce3c1967 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -183,6 +183,7 @@ static void kfd_process_wq_release(struct work_struct *work)
 	kfd_event_free_process(p);
 
 	kfd_pasid_free(p->pasid);
+	kfd_free_process_doorbells(p);
 
 	mutex_unlock(&p->mutex);
 
@@ -288,6 +289,9 @@ static struct kfd_process *create_process(const struct task_struct *thread)
 	if (process->pasid == 0)
 		goto err_alloc_pasid;
 
+	if (kfd_alloc_process_doorbells(process) < 0)
+		goto err_alloc_doorbells;
+
 	mutex_init(&process->mutex);
 
 	process->mm = thread->mm;
@@ -329,6 +333,8 @@ err_process_pqm_init:
 	mmu_notifier_unregister_no_release(&process->mmu_notifier, process->mm);
 err_mmu_notifier:
 	mutex_destroy(&process->mutex);
+	kfd_free_process_doorbells(process);
+err_alloc_doorbells:
 	kfd_pasid_free(process->pasid);
 err_alloc_pasid:
 	kfree(process->queues);