Merge branch 'drm-next-4.16' of git://people.freedesktop.org/~agd5f/linux into drm-next

Last few updates for 4.16:
- Misc fixes for amdgpu
- Enable swapout for reserved BOs during allocation for ttm
- Misc cleanups for ttm

* 'drm-next-4.16' of git://people.freedesktop.org/~agd5f/linux: (24 commits)
  drm/amdgpu: Correct the IB size of bo update mapping.
  drm/ttm: enable swapout for reserved BOs during allocation
  drm/ttm: add new function to check if bo is allowable to evict or swapout
  drm/ttm: use an operation ctx for ttm_tt_bind
  drm/ttm: use an operation ctx for ttm_tt_populate in ttm_bo_driver (v2)
  drm/ttm: use an operation ctx for ttm_mem_global_alloc_page
  drm/ttm: use an operation ctx for ttm_mem_global_alloc
  drm/ttm: call ttm_bo_swapout directly when ttm shrink
  drm/vmwgfx: remove the default io_mem_pfn set
  drm/virtio: remove the default io_mem_pfn set
  drm/radeon: remove the default io_mem_pfn set
  drm/qxl: remove the default io_mem_pfn set
  drm/nouveau: remove the default io_mem_pfn set
  drm/mgag200: remove the default io_mem_pfn set
  drm/cirrus: remove the default io_mem_pfn set
  drm/bochs: remove the default io_mem_pfn set
  drm/ast: remove the default io_mem_pfn set
  drm/ttm: add ttm_bo_io_mem_pfn to check io_mem_pfn
  drm/amdgpu: fix VM faults with per VM BOs
  drm/ttm: drop the spin in delayed delete if the trylock doesn't work
  ...

drivers/gpu/drm/amd/amdgpu/Makefile

@@ -52,7 +52,8 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \
 	amdgpu_prime.o amdgpu_vm.o amdgpu_ib.o amdgpu_pll.o \
 	amdgpu_ucode.o amdgpu_bo_list.o amdgpu_ctx.o amdgpu_sync.o \
 	amdgpu_gtt_mgr.o amdgpu_vram_mgr.o amdgpu_virt.o amdgpu_atomfirmware.o \
-	amdgpu_queue_mgr.o amdgpu_vf_error.o amdgpu_sched.o amdgpu_debugfs.o
+	amdgpu_queue_mgr.o amdgpu_vf_error.o amdgpu_sched.o amdgpu_debugfs.o \
+	amdgpu_ids.o
 
 # add asic specific block
 amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o cik_ih.o kv_smc.o kv_dpm.o \

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

@@ -351,7 +351,7 @@ struct amdgpu_gart_funcs {
 	/* get the pde for a given mc addr */
 	void (*get_vm_pde)(struct amdgpu_device *adev, int level,
 			   u64 *dst, u64 *flags);
-	uint32_t (*get_invalidate_req)(unsigned int vm_id);
+	uint32_t (*get_invalidate_req)(unsigned int vmid);
 };
 
 /* provided by the ih block */
@@ -1125,7 +1125,7 @@ struct amdgpu_job {
 	void			*owner;
 	uint64_t		fence_ctx; /* the fence_context this job uses */
 	bool                    vm_needs_flush;
-	unsigned		vm_id;
+	unsigned		vmid;
 	uint64_t		vm_pd_addr;
 	uint32_t		gds_base, gds_size;
 	uint32_t		gws_base, gws_size;
@@ -1850,7 +1850,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
 #define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r))
 #define amdgpu_ring_get_wptr(r) (r)->funcs->get_wptr((r))
 #define amdgpu_ring_set_wptr(r) (r)->funcs->set_wptr((r))
-#define amdgpu_ring_emit_ib(r, ib, vm_id, c) (r)->funcs->emit_ib((r), (ib), (vm_id), (c))
+#define amdgpu_ring_emit_ib(r, ib, vmid, c) (r)->funcs->emit_ib((r), (ib), (vmid), (c))
 #define amdgpu_ring_emit_pipeline_sync(r) (r)->funcs->emit_pipeline_sync((r))
 #define amdgpu_ring_emit_vm_flush(r, vmid, addr) (r)->funcs->emit_vm_flush((r), (vmid), (addr))
 #define amdgpu_ring_emit_fence(r, addr, seq, flags) (r)->funcs->emit_fence((r), (addr), (seq), (flags))

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c

@@ -176,8 +176,8 @@ static const struct kfd2kgd_calls kfd2kgd = {
 	.get_local_mem_info = get_local_mem_info,
 	.get_gpu_clock_counter = get_gpu_clock_counter,
 	.get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
-	.alloc_pasid = amdgpu_vm_alloc_pasid,
-	.free_pasid = amdgpu_vm_free_pasid,
+	.alloc_pasid = amdgpu_pasid_alloc,
+	.free_pasid = amdgpu_pasid_free,
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
 	.init_pipeline = kgd_init_pipeline,

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c

@@ -135,8 +135,8 @@ static const struct kfd2kgd_calls kfd2kgd = {
 	.get_local_mem_info = get_local_mem_info,
 	.get_gpu_clock_counter = get_gpu_clock_counter,
 	.get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
-	.alloc_pasid = amdgpu_vm_alloc_pasid,
-	.free_pasid = amdgpu_vm_free_pasid,
+	.alloc_pasid = amdgpu_pasid_alloc,
+	.free_pasid = amdgpu_pasid_free,
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
 	.init_pipeline = kgd_init_pipeline,

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

@@ -801,6 +801,8 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 				else
 					strcpy(fw_name, "amdgpu/vega10_smc.bin");
 				break;
+			case CHIP_CARRIZO:
+			case CHIP_STONEY:
 			case CHIP_RAVEN:
 				adev->pm.fw_version = info->version;
 				return 0;

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

@@ -149,7 +149,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 		return -EINVAL;
 	}
 
-	if (vm && !job->vm_id) {
+	if (vm && !job->vmid) {
 		dev_err(adev->dev, "VM IB without ID\n");
 		return -EINVAL;
 	}
@@ -211,7 +211,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 			!amdgpu_sriov_vf(adev)) /* for SRIOV preemption, Preamble CE ib must be inserted anyway */
 			continue;
 
-		amdgpu_ring_emit_ib(ring, ib, job ? job->vm_id : 0,
+		amdgpu_ring_emit_ib(ring, ib, job ? job->vmid : 0,
 				    need_ctx_switch);
 		need_ctx_switch = false;
 	}
@@ -229,9 +229,8 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 	r = amdgpu_fence_emit(ring, f);
 	if (r) {
 		dev_err(adev->dev, "failed to emit fence (%d)\n", r);
-		if (job && job->vm_id)
-			amdgpu_vm_reset_id(adev, ring->funcs->vmhub,
-					   job->vm_id);
+		if (job && job->vmid)
+			amdgpu_vmid_reset(adev, ring->funcs->vmhub, job->vmid);
 		amdgpu_ring_undo(ring);
 		return r;
 	}

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

+/*
+ * Copyright 2017 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu_ids.h"
+
+#include <linux/idr.h>
+#include <linux/dma-fence-array.h>
+#include <drm/drmP.h>
+
+#include "amdgpu.h"
+#include "amdgpu_trace.h"
+
+/*
+ * PASID manager
+ *
+ * PASIDs are global address space identifiers that can be shared
+ * between the GPU, an IOMMU and the driver. VMs on different devices
+ * may use the same PASID if they share the same address
+ * space. Therefore PASIDs are allocated using a global IDA. VMs are
+ * looked up from the PASID per amdgpu_device.
+ */
+static DEFINE_IDA(amdgpu_pasid_ida);
+
+/**
+ * amdgpu_pasid_alloc - Allocate a PASID
+ * @bits: Maximum width of the PASID in bits, must be at least 1
+ *
+ * Allocates a PASID of the given width while keeping smaller PASIDs
+ * available if possible.
+ *
+ * Returns a positive integer on success. Returns %-EINVAL if bits==0.
+ * Returns %-ENOSPC if no PASID was available. Returns %-ENOMEM on
+ * memory allocation failure.
+ */
+int amdgpu_pasid_alloc(unsigned int bits)
+{
+	int pasid = -EINVAL;
+
+	for (bits = min(bits, 31U); bits > 0; bits--) {
+		pasid = ida_simple_get(&amdgpu_pasid_ida,
+				       1U << (bits - 1), 1U << bits,
+				       GFP_KERNEL);
+		if (pasid != -ENOSPC)
+			break;
+	}
+
+	return pasid;
+}
+
+/**
+ * amdgpu_pasid_free - Free a PASID
+ * @pasid: PASID to free
+ */
+void amdgpu_pasid_free(unsigned int pasid)
+{
+	ida_simple_remove(&amdgpu_pasid_ida, pasid);
+}
+
+/*
+ * VMID manager
+ *
+ * VMIDs are a per VMHUB identifier for page tables handling.
+ */
+
+/**
+ * amdgpu_vmid_had_gpu_reset - check if reset occured since last use
+ *
+ * @adev: amdgpu_device pointer
+ * @id: VMID structure
+ *
+ * Check if GPU reset occured since last use of the VMID.
+ */
+bool amdgpu_vmid_had_gpu_reset(struct amdgpu_device *adev,
+			       struct amdgpu_vmid *id)
+{
+	return id->current_gpu_reset_count !=
+		atomic_read(&adev->gpu_reset_counter);
+}
+
+/* idr_mgr->lock must be held */
+static int amdgpu_vmid_grab_reserved_locked(struct amdgpu_vm *vm,
+					    struct amdgpu_ring *ring,
+					    struct amdgpu_sync *sync,
+					    struct dma_fence *fence,
+					    struct amdgpu_job *job)
+{
+	struct amdgpu_device *adev = ring->adev;
+	unsigned vmhub = ring->funcs->vmhub;
+	uint64_t fence_context = adev->fence_context + ring->idx;
+	struct amdgpu_vmid *id = vm->reserved_vmid[vmhub];
+	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	struct dma_fence *updates = sync->last_vm_update;
+	int r = 0;
+	struct dma_fence *flushed, *tmp;
+	bool needs_flush = vm->use_cpu_for_update;
+
+	flushed  = id->flushed_updates;
+	if ((amdgpu_vmid_had_gpu_reset(adev, id)) ||
+	    (atomic64_read(&id->owner) != vm->entity.fence_context) ||
+	    (job->vm_pd_addr != id->pd_gpu_addr) ||
+	    (updates && (!flushed || updates->context != flushed->context ||
+			dma_fence_is_later(updates, flushed))) ||
+	    (!id->last_flush || (id->last_flush->context != fence_context &&
+				 !dma_fence_is_signaled(id->last_flush)))) {
+		needs_flush = true;
+		/* to prevent one context starved by another context */
+		id->pd_gpu_addr = 0;
+		tmp = amdgpu_sync_peek_fence(&id->active, ring);
+		if (tmp) {
+			r = amdgpu_sync_fence(adev, sync, tmp, false);
+			return r;
+		}
+	}
+
+	/* Good we can use this VMID. Remember this submission as
+	* user of the VMID.
+	*/
+	r = amdgpu_sync_fence(ring->adev, &id->active, fence, false);
+	if (r)
+		goto out;
+
+	if (updates && (!flushed || updates->context != flushed->context ||
+			dma_fence_is_later(updates, flushed))) {
+		dma_fence_put(id->flushed_updates);
+		id->flushed_updates = dma_fence_get(updates);
+	}
+	id->pd_gpu_addr = job->vm_pd_addr;
+	atomic64_set(&id->owner, vm->entity.fence_context);
+	job->vm_needs_flush = needs_flush;
+	if (needs_flush) {
+		dma_fence_put(id->last_flush);
+		id->last_flush = NULL;
+	}
+	job->vmid = id - id_mgr->ids;
+	trace_amdgpu_vm_grab_id(vm, ring, job);
+out:
+	return r;
+}
+
+/**
+ * amdgpu_vm_grab_id - allocate the next free VMID
+ *
+ * @vm: vm to allocate id for
+ * @ring: ring we want to submit job to
+ * @sync: sync object where we add dependencies
+ * @fence: fence protecting ID from reuse
+ *
+ * Allocate an id for the vm, adding fences to the sync obj as necessary.
+ */
+int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
+		     struct amdgpu_sync *sync, struct dma_fence *fence,
+		     struct amdgpu_job *job)
+{
+	struct amdgpu_device *adev = ring->adev;
+	unsigned vmhub = ring->funcs->vmhub;
+	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	uint64_t fence_context = adev->fence_context + ring->idx;
+	struct dma_fence *updates = sync->last_vm_update;
+	struct amdgpu_vmid *id, *idle;
+	struct dma_fence **fences;
+	unsigned i;
+	int r = 0;
+
+	mutex_lock(&id_mgr->lock);
+	if (vm->reserved_vmid[vmhub]) {
+		r = amdgpu_vmid_grab_reserved_locked(vm, ring, sync, fence, job);
+		mutex_unlock(&id_mgr->lock);
+		return r;
+	}
+	fences = kmalloc_array(sizeof(void *), id_mgr->num_ids, GFP_KERNEL);
+	if (!fences) {
+		mutex_unlock(&id_mgr->lock);
+		return -ENOMEM;
+	}
+	/* Check if we have an idle VMID */
+	i = 0;
+	list_for_each_entry(idle, &id_mgr->ids_lru, list) {
+		fences[i] = amdgpu_sync_peek_fence(&idle->active, ring);
+		if (!fences[i])
+			break;
+		++i;
+	}
+
+	/* If we can't find a idle VMID to use, wait till one becomes available */
+	if (&idle->list == &id_mgr->ids_lru) {
+		u64 fence_context = adev->vm_manager.fence_context + ring->idx;
+		unsigned seqno = ++adev->vm_manager.seqno[ring->idx];
+		struct dma_fence_array *array;
+		unsigned j;
+
+		for (j = 0; j < i; ++j)
+			dma_fence_get(fences[j]);
+
+		array = dma_fence_array_create(i, fences, fence_context,
+					   seqno, true);
+		if (!array) {
+			for (j = 0; j < i; ++j)
+				dma_fence_put(fences[j]);
+			kfree(fences);
+			r = -ENOMEM;
+			goto error;
+		}
+
+
+		r = amdgpu_sync_fence(ring->adev, sync, &array->base, false);
+		dma_fence_put(&array->base);
+		if (r)
+			goto error;
+
+		mutex_unlock(&id_mgr->lock);
+		return 0;
+
+	}
+	kfree(fences);
+
+	job->vm_needs_flush = vm->use_cpu_for_update;
+	/* Check if we can use a VMID already assigned to this VM */
+	list_for_each_entry_reverse(id, &id_mgr->ids_lru, list) {
+		struct dma_fence *flushed;
+		bool needs_flush = vm->use_cpu_for_update;
+
+		/* Check all the prerequisites to using this VMID */
+		if (amdgpu_vmid_had_gpu_reset(adev, id))
+			continue;
+
+		if (atomic64_read(&id->owner) != vm->entity.fence_context)
+			continue;
+
+		if (job->vm_pd_addr != id->pd_gpu_addr)
+			continue;
+
+		if (!id->last_flush ||
+		    (id->last_flush->context != fence_context &&
+		     !dma_fence_is_signaled(id->last_flush)))
+			needs_flush = true;
+
+		flushed  = id->flushed_updates;
+		if (updates && (!flushed || dma_fence_is_later(updates, flushed)))
+			needs_flush = true;
+
+		/* Concurrent flushes are only possible starting with Vega10 */
+		if (adev->asic_type < CHIP_VEGA10 && needs_flush)
+			continue;
+
+		/* Good we can use this VMID. Remember this submission as
+		 * user of the VMID.
+		 */
+		r = amdgpu_sync_fence(ring->adev, &id->active, fence, false);
+		if (r)
+			goto error;
+
+		if (updates && (!flushed || dma_fence_is_later(updates, flushed))) {
+			dma_fence_put(id->flushed_updates);
+			id->flushed_updates = dma_fence_get(updates);
+		}
+
+		if (needs_flush)
+			goto needs_flush;
+		else
+			goto no_flush_needed;
+
+	};
+
+	/* Still no ID to use? Then use the idle one found earlier */
+	id = idle;
+
+	/* Remember this submission as user of the VMID */
+	r = amdgpu_sync_fence(ring->adev, &id->active, fence, false);
+	if (r)
+		goto error;
+
+	id->pd_gpu_addr = job->vm_pd_addr;
+	dma_fence_put(id->flushed_updates);
+	id->flushed_updates = dma_fence_get(updates);
+	atomic64_set(&id->owner, vm->entity.fence_context);
+
+needs_flush:
+	job->vm_needs_flush = true;
+	dma_fence_put(id->last_flush);
+	id->last_flush = NULL;
+
+no_flush_needed:
+	list_move_tail(&id->list, &id_mgr->ids_lru);
+
+	job->vmid = id - id_mgr->ids;
+	trace_amdgpu_vm_grab_id(vm, ring, job);
+
+error:
+	mutex_unlock(&id_mgr->lock);
+	return r;
+}
+
+int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev,
+			       struct amdgpu_vm *vm,
+			       unsigned vmhub)
+{
+	struct amdgpu_vmid_mgr *id_mgr;
+	struct amdgpu_vmid *idle;
+	int r = 0;
+
+	id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	mutex_lock(&id_mgr->lock);
+	if (vm->reserved_vmid[vmhub])
+		goto unlock;
+	if (atomic_inc_return(&id_mgr->reserved_vmid_num) >
+	    AMDGPU_VM_MAX_RESERVED_VMID) {
+		DRM_ERROR("Over limitation of reserved vmid\n");
+		atomic_dec(&id_mgr->reserved_vmid_num);
+		r = -EINVAL;
+		goto unlock;
+	}
+	/* Select the first entry VMID */
+	idle = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vmid, list);
+	list_del_init(&idle->list);
+	vm->reserved_vmid[vmhub] = idle;
+	mutex_unlock(&id_mgr->lock);
+
+	return 0;
+unlock:
+	mutex_unlock(&id_mgr->lock);
+	return r;
+}
+
+void amdgpu_vmid_free_reserved(struct amdgpu_device *adev,
+			       struct amdgpu_vm *vm,
+			       unsigned vmhub)
+{
+	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+
+	mutex_lock(&id_mgr->lock);
+	if (vm->reserved_vmid[vmhub]) {
+		list_add(&vm->reserved_vmid[vmhub]->list,
+			&id_mgr->ids_lru);
+		vm->reserved_vmid[vmhub] = NULL;
+		atomic_dec(&id_mgr->reserved_vmid_num);
+	}
+	mutex_unlock(&id_mgr->lock);
+}
+
+/**
+ * amdgpu_vmid_reset - reset VMID to zero
+ *
+ * @adev: amdgpu device structure
+ * @vmid: vmid number to use
+ *
+ * Reset saved GDW, GWS and OA to force switch on next flush.
+ */
+void amdgpu_vmid_reset(struct amdgpu_device *adev, unsigned vmhub,
+		       unsigned vmid)
+{
+	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	struct amdgpu_vmid *id = &id_mgr->ids[vmid];
+
+	atomic64_set(&id->owner, 0);
+	id->gds_base = 0;
+	id->gds_size = 0;
+	id->gws_base = 0;
+	id->gws_size = 0;
+	id->oa_base = 0;
+	id->oa_size = 0;
+}
+
+/**
+ * amdgpu_vmid_reset_all - reset VMID to zero
+ *
+ * @adev: amdgpu device structure
+ *
+ * Reset VMID to force flush on next use
+ */
+void amdgpu_vmid_reset_all(struct amdgpu_device *adev)
+{
+	unsigned i, j;
+
+	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
+		struct amdgpu_vmid_mgr *id_mgr =
+			&adev->vm_manager.id_mgr[i];
+
+		for (j = 1; j < id_mgr->num_ids; ++j)
+			amdgpu_vmid_reset(adev, i, j);
+	}
+}
+
+/**
+ * amdgpu_vmid_mgr_init - init the VMID manager
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Initialize the VM manager structures
+ */
+void amdgpu_vmid_mgr_init(struct amdgpu_device *adev)
+{
+	unsigned i, j;
+
+	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
+		struct amdgpu_vmid_mgr *id_mgr =
+			&adev->vm_manager.id_mgr[i];
+
+		mutex_init(&id_mgr->lock);
+		INIT_LIST_HEAD(&id_mgr->ids_lru);
+		atomic_set(&id_mgr->reserved_vmid_num, 0);
+
+		/* skip over VMID 0, since it is the system VM */
+		for (j = 1; j < id_mgr->num_ids; ++j) {
+			amdgpu_vmid_reset(adev, i, j);
+			amdgpu_sync_create(&id_mgr->ids[i].active);
+			list_add_tail(&id_mgr->ids[j].list, &id_mgr->ids_lru);
+		}
+	}
+
+	adev->vm_manager.fence_context =
+		dma_fence_context_alloc(AMDGPU_MAX_RINGS);
+	for (i = 0; i < AMDGPU_MAX_RINGS; ++i)
+		adev->vm_manager.seqno[i] = 0;
+}
+
+/**
+ * amdgpu_vmid_mgr_fini - cleanup VM manager
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Cleanup the VM manager and free resources.
+ */
+void amdgpu_vmid_mgr_fini(struct amdgpu_device *adev)
+{
+	unsigned i, j;
+
+	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
+		struct amdgpu_vmid_mgr *id_mgr =
+			&adev->vm_manager.id_mgr[i];
+
+		mutex_destroy(&id_mgr->lock);
+		for (j = 0; j < AMDGPU_NUM_VMID; ++j) {
+			struct amdgpu_vmid *id = &id_mgr->ids[j];
+
+			amdgpu_sync_free(&id->active);
+			dma_fence_put(id->flushed_updates);
+			dma_fence_put(id->last_flush);
+		}
+	}
+}

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

+/*
+ * Copyright 2017 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __AMDGPU_IDS_H__
+#define __AMDGPU_IDS_H__
+
+#include <linux/types.h>
+#include <linux/mutex.h>
+#include <linux/list.h>
+#include <linux/dma-fence.h>
+
+#include "amdgpu_sync.h"
+
+/* maximum number of VMIDs */
+#define AMDGPU_NUM_VMID	16
+
+struct amdgpu_device;
+struct amdgpu_vm;
+struct amdgpu_ring;
+struct amdgpu_sync;
+struct amdgpu_job;
+
+struct amdgpu_vmid {
+	struct list_head	list;
+	struct amdgpu_sync	active;
+	struct dma_fence	*last_flush;
+	atomic64_t		owner;
+
+	uint64_t		pd_gpu_addr;
+	/* last flushed PD/PT update */
+	struct dma_fence	*flushed_updates;
+
+	uint32_t                current_gpu_reset_count;
+
+	uint32_t		gds_base;
+	uint32_t		gds_size;
+	uint32_t		gws_base;
+	uint32_t		gws_size;
+	uint32_t		oa_base;
+	uint32_t		oa_size;
+};
+
+struct amdgpu_vmid_mgr {
+	struct mutex		lock;
+	unsigned		num_ids;
+	struct list_head	ids_lru;
+	struct amdgpu_vmid	ids[AMDGPU_NUM_VMID];
+	atomic_t		reserved_vmid_num;
+};
+
+int amdgpu_pasid_alloc(unsigned int bits);
+void amdgpu_pasid_free(unsigned int pasid);
+
+bool amdgpu_vmid_had_gpu_reset(struct amdgpu_device *adev,
+			       struct amdgpu_vmid *id);
+int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev,
+			       struct amdgpu_vm *vm,
+			       unsigned vmhub);
+void amdgpu_vmid_free_reserved(struct amdgpu_device *adev,
+			       struct amdgpu_vm *vm,
+			       unsigned vmhub);
+int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
+		     struct amdgpu_sync *sync, struct dma_fence *fence,
+		     struct amdgpu_job *job);
+void amdgpu_vmid_reset(struct amdgpu_device *adev, unsigned vmhub,
+		       unsigned vmid);
+void amdgpu_vmid_reset_all(struct amdgpu_device *adev);
+
+void amdgpu_vmid_mgr_init(struct amdgpu_device *adev);
+void amdgpu_vmid_mgr_fini(struct amdgpu_device *adev);
+
+#endif

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

@@ -105,8 +105,8 @@ struct amdgpu_iv_entry {
 	unsigned client_id;
 	unsigned src_id;
 	unsigned ring_id;
-	unsigned vm_id;
-	unsigned vm_id_src;
+	unsigned vmid;
+	unsigned vmid_src;
 	uint64_t timestamp;
 	unsigned timestamp_src;
 	unsigned pas_id;

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

@@ -158,12 +158,12 @@ static struct dma_fence *amdgpu_job_dependency(struct drm_sched_job *sched_job,
 		}
 	}
 
-	while (fence == NULL && vm && !job->vm_id) {
+	while (fence == NULL && vm && !job->vmid) {
 		struct amdgpu_ring *ring = job->ring;
 
-		r = amdgpu_vm_grab_id(vm, ring, &job->sync,
-				      &job->base.s_fence->finished,
-				      job);
+		r = amdgpu_vmid_grab(vm, ring, &job->sync,
+				     &job->base.s_fence->finished,
+				     job);
 		if (r)
 			DRM_ERROR("Error getting VM ID (%d)\n", r);
 

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

@@ -121,11 +121,11 @@ struct amdgpu_ring_funcs {
 	/* command emit functions */
 	void (*emit_ib)(struct amdgpu_ring *ring,
 			struct amdgpu_ib *ib,
-			unsigned vm_id, bool ctx_switch);
+			unsigned vmid, bool ctx_switch);
 	void (*emit_fence)(struct amdgpu_ring *ring, uint64_t addr,
 			   uint64_t seq, unsigned flags);
 	void (*emit_pipeline_sync)(struct amdgpu_ring *ring);
-	void (*emit_vm_flush)(struct amdgpu_ring *ring, unsigned vm_id,
+	void (*emit_vm_flush)(struct amdgpu_ring *ring, unsigned vmid,
 			      uint64_t pd_addr);
 	void (*emit_hdp_flush)(struct amdgpu_ring *ring);
 	void (*emit_hdp_invalidate)(struct amdgpu_ring *ring);

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

@@ -82,8 +82,8 @@ TRACE_EVENT(amdgpu_iv,
 			     __field(unsigned, client_id)
 			     __field(unsigned, src_id)
 			     __field(unsigned, ring_id)
-			     __field(unsigned, vm_id)
-			     __field(unsigned, vm_id_src)
+			     __field(unsigned, vmid)
+			     __field(unsigned, vmid_src)
 			     __field(uint64_t, timestamp)
 			     __field(unsigned, timestamp_src)
 			     __field(unsigned, pas_id)
@@ -93,8 +93,8 @@ TRACE_EVENT(amdgpu_iv,
 			   __entry->client_id = iv->client_id;
 			   __entry->src_id = iv->src_id;
 			   __entry->ring_id = iv->ring_id;
-			   __entry->vm_id = iv->vm_id;
-			   __entry->vm_id_src = iv->vm_id_src;
+			   __entry->vmid = iv->vmid;
+			   __entry->vmid_src = iv->vmid_src;
 			   __entry->timestamp = iv->timestamp;
 			   __entry->timestamp_src = iv->timestamp_src;
 			   __entry->pas_id = iv->pas_id;
@@ -103,9 +103,9 @@ TRACE_EVENT(amdgpu_iv,
 			   __entry->src_data[2] = iv->src_data[2];
 			   __entry->src_data[3] = iv->src_data[3];
 			   ),
-	    TP_printk("client_id:%u src_id:%u ring:%u vm_id:%u timestamp: %llu pas_id:%u src_data: %08x %08x %08x %08x\n",
+	    TP_printk("client_id:%u src_id:%u ring:%u vmid:%u timestamp: %llu pas_id:%u src_data: %08x %08x %08x %08x\n",
 		      __entry->client_id, __entry->src_id,
-		      __entry->ring_id, __entry->vm_id,
+		      __entry->ring_id, __entry->vmid,
 		      __entry->timestamp, __entry->pas_id,
 		      __entry->src_data[0], __entry->src_data[1],
 		      __entry->src_data[2], __entry->src_data[3])
@@ -219,7 +219,7 @@ TRACE_EVENT(amdgpu_vm_grab_id,
 	    TP_STRUCT__entry(
 			     __field(struct amdgpu_vm *, vm)
 			     __field(u32, ring)
-			     __field(u32, vm_id)
+			     __field(u32, vmid)
 			     __field(u32, vm_hub)
 			     __field(u64, pd_addr)
 			     __field(u32, needs_flush)
@@ -228,13 +228,13 @@ TRACE_EVENT(amdgpu_vm_grab_id,
 	    TP_fast_assign(
 			   __entry->vm = vm;
 			   __entry->ring = ring->idx;
-			   __entry->vm_id = job->vm_id;
+			   __entry->vmid = job->vmid;
 			   __entry->vm_hub = ring->funcs->vmhub,
 			   __entry->pd_addr = job->vm_pd_addr;
 			   __entry->needs_flush = job->vm_needs_flush;
 			   ),
 	    TP_printk("vm=%p, ring=%u, id=%u, hub=%u, pd_addr=%010Lx needs_flush=%u",
-		      __entry->vm, __entry->ring, __entry->vm_id,
+		      __entry->vm, __entry->ring, __entry->vmid,
 		      __entry->vm_hub, __entry->pd_addr, __entry->needs_flush)
 );
 
@@ -357,24 +357,24 @@ TRACE_EVENT(amdgpu_vm_copy_ptes,
 );
 
 TRACE_EVENT(amdgpu_vm_flush,
-	    TP_PROTO(struct amdgpu_ring *ring, unsigned vm_id,
+	    TP_PROTO(struct amdgpu_ring *ring, unsigned vmid,
 		     uint64_t pd_addr),
-	    TP_ARGS(ring, vm_id, pd_addr),
+	    TP_ARGS(ring, vmid, pd_addr),
 	    TP_STRUCT__entry(
 			     __field(u32, ring)
-			     __field(u32, vm_id)
+			     __field(u32, vmid)
 			     __field(u32, vm_hub)
 			     __field(u64, pd_addr)
 			     ),
 
 	    TP_fast_assign(
 			   __entry->ring = ring->idx;
-			   __entry->vm_id = vm_id;
+			   __entry->vmid = vmid;
 			   __entry->vm_hub = ring->funcs->vmhub;
 			   __entry->pd_addr = pd_addr;
 			   ),
 	    TP_printk("ring=%u, id=%u, hub=%u, pd_addr=%010Lx",
-		      __entry->ring, __entry->vm_id,
+		      __entry->ring, __entry->vmid,
 		      __entry->vm_hub,__entry->pd_addr)
 );
 

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

@@ -497,7 +497,7 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, bool evict,
 		goto out_cleanup;
 	}
 
-	r = ttm_tt_bind(bo->ttm, &tmp_mem);
+	r = ttm_tt_bind(bo->ttm, &tmp_mem, ctx);
 	if (unlikely(r)) {
 		goto out_cleanup;
 	}
@@ -990,7 +990,8 @@ static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_bo_device *bdev,
 	return &gtt->ttm.ttm;
 }
 
-static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm)
+static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm,
+			struct ttm_operation_ctx *ctx)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev);
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
@@ -1018,11 +1019,11 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm)
 
 #ifdef CONFIG_SWIOTLB
 	if (swiotlb_nr_tbl()) {
-		return ttm_dma_populate(&gtt->ttm, adev->dev);
+		return ttm_dma_populate(&gtt->ttm, adev->dev, ctx);
 	}
 #endif
 
-	return ttm_populate_and_map_pages(adev->dev, &gtt->ttm);
+	return ttm_populate_and_map_pages(adev->dev, &gtt->ttm, ctx);
 }
 
 static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm)

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

@@ -991,7 +991,7 @@ int amdgpu_vce_ring_parse_cs_vm(struct amdgpu_cs_parser *p, uint32_t ib_idx)
  *
  */
 void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib,
-			     unsigned vm_id, bool ctx_switch)
+			     unsigned vmid, bool ctx_switch)
 {
 	amdgpu_ring_write(ring, VCE_CMD_IB);
 	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));

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

@@ -63,7 +63,7 @@ void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp);
 int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx);
 int amdgpu_vce_ring_parse_cs_vm(struct amdgpu_cs_parser *p, uint32_t ib_idx);
 void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib,
-			     unsigned vm_id, bool ctx_switch);
+			     unsigned vmid, bool ctx_switch);
 void amdgpu_vce_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
 				unsigned flags);
 int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring);

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

@@ -33,52 +33,6 @@
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 
-/*
- * PASID manager
- *
- * PASIDs are global address space identifiers that can be shared
- * between the GPU, an IOMMU and the driver. VMs on different devices
- * may use the same PASID if they share the same address
- * space. Therefore PASIDs are allocated using a global IDA. VMs are
- * looked up from the PASID per amdgpu_device.
- */
-static DEFINE_IDA(amdgpu_vm_pasid_ida);
-
-/**
- * amdgpu_vm_alloc_pasid - Allocate a PASID
- * @bits: Maximum width of the PASID in bits, must be at least 1
- *
- * Allocates a PASID of the given width while keeping smaller PASIDs
- * available if possible.
- *
- * Returns a positive integer on success. Returns %-EINVAL if bits==0.
- * Returns %-ENOSPC if no PASID was available. Returns %-ENOMEM on
- * memory allocation failure.
- */
-int amdgpu_vm_alloc_pasid(unsigned int bits)
-{
-	int pasid = -EINVAL;
-
-	for (bits = min(bits, 31U); bits > 0; bits--) {
-		pasid = ida_simple_get(&amdgpu_vm_pasid_ida,
-				       1U << (bits - 1), 1U << bits,
-				       GFP_KERNEL);
-		if (pasid != -ENOSPC)
-			break;
-	}
-
-	return pasid;
-}
-
-/**
- * amdgpu_vm_free_pasid - Free a PASID
- * @pasid: PASID to free
- */
-void amdgpu_vm_free_pasid(unsigned int pasid)
-{
-	ida_simple_remove(&amdgpu_vm_pasid_ida, pasid);
-}
-
 /*
  * GPUVM
  * GPUVM is similar to the legacy gart on older asics, however
@@ -447,286 +401,6 @@ int amdgpu_vm_alloc_pts(struct amdgpu_device *adev,
 				      adev->vm_manager.root_level);
 }
 
-/**
- * amdgpu_vm_had_gpu_reset - check if reset occured since last use
- *
- * @adev: amdgpu_device pointer
- * @id: VMID structure
- *
- * Check if GPU reset occured since last use of the VMID.
- */
-static bool amdgpu_vm_had_gpu_reset(struct amdgpu_device *adev,
-				    struct amdgpu_vm_id *id)
-{
-	return id->current_gpu_reset_count !=
-		atomic_read(&adev->gpu_reset_counter);
-}
-
-static bool amdgpu_vm_reserved_vmid_ready(struct amdgpu_vm *vm, unsigned vmhub)
-{
-	return !!vm->reserved_vmid[vmhub];
-}
-
-/* idr_mgr->lock must be held */
-static int amdgpu_vm_grab_reserved_vmid_locked(struct amdgpu_vm *vm,
-					       struct amdgpu_ring *ring,
-					       struct amdgpu_sync *sync,
-					       struct dma_fence *fence,
-					       struct amdgpu_job *job)
-{
-	struct amdgpu_device *adev = ring->adev;
-	unsigned vmhub = ring->funcs->vmhub;
-	uint64_t fence_context = adev->fence_context + ring->idx;
-	struct amdgpu_vm_id *id = vm->reserved_vmid[vmhub];
-	struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[vmhub];
-	struct dma_fence *updates = sync->last_vm_update;
-	int r = 0;
-	struct dma_fence *flushed, *tmp;
-	bool needs_flush = vm->use_cpu_for_update;
-
-	flushed  = id->flushed_updates;
-	if ((amdgpu_vm_had_gpu_reset(adev, id)) ||
-	    (atomic64_read(&id->owner) != vm->client_id) ||
-	    (job->vm_pd_addr != id->pd_gpu_addr) ||
-	    (updates && (!flushed || updates->context != flushed->context ||
-			dma_fence_is_later(updates, flushed))) ||
-	    (!id->last_flush || (id->last_flush->context != fence_context &&
-				 !dma_fence_is_signaled(id->last_flush)))) {
-		needs_flush = true;
-		/* to prevent one context starved by another context */
-		id->pd_gpu_addr = 0;
-		tmp = amdgpu_sync_peek_fence(&id->active, ring);
-		if (tmp) {
-			r = amdgpu_sync_fence(adev, sync, tmp, false);
-			return r;
-		}
-	}
-
-	/* Good we can use this VMID. Remember this submission as
-	* user of the VMID.
-	*/
-	r = amdgpu_sync_fence(ring->adev, &id->active, fence, false);
-	if (r)
-		goto out;
-
-	if (updates && (!flushed || updates->context != flushed->context ||
-			dma_fence_is_later(updates, flushed))) {
-		dma_fence_put(id->flushed_updates);
-		id->flushed_updates = dma_fence_get(updates);
-	}
-	id->pd_gpu_addr = job->vm_pd_addr;
-	atomic64_set(&id->owner, vm->client_id);
-	job->vm_needs_flush = needs_flush;
-	if (needs_flush) {
-		dma_fence_put(id->last_flush);
-		id->last_flush = NULL;
-	}
-	job->vm_id = id - id_mgr->ids;
-	trace_amdgpu_vm_grab_id(vm, ring, job);
-out:
-	return r;
-}
-
-/**
- * amdgpu_vm_grab_id - allocate the next free VMID
- *
- * @vm: vm to allocate id for
- * @ring: ring we want to submit job to
- * @sync: sync object where we add dependencies
- * @fence: fence protecting ID from reuse
- *
- * Allocate an id for the vm, adding fences to the sync obj as necessary.
- */
-int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
-		      struct amdgpu_sync *sync, struct dma_fence *fence,
-		      struct amdgpu_job *job)
-{
-	struct amdgpu_device *adev = ring->adev;
-	unsigned vmhub = ring->funcs->vmhub;
-	struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[vmhub];
-	uint64_t fence_context = adev->fence_context + ring->idx;
-	struct dma_fence *updates = sync->last_vm_update;
-	struct amdgpu_vm_id *id, *idle;
-	struct dma_fence **fences;
-	unsigned i;
-	int r = 0;
-
-	mutex_lock(&id_mgr->lock);
-	if (amdgpu_vm_reserved_vmid_ready(vm, vmhub)) {
-		r = amdgpu_vm_grab_reserved_vmid_locked(vm, ring, sync, fence, job);
-		mutex_unlock(&id_mgr->lock);
-		return r;
-	}
-	fences = kmalloc_array(sizeof(void *), id_mgr->num_ids, GFP_KERNEL);
-	if (!fences) {
-		mutex_unlock(&id_mgr->lock);
-		return -ENOMEM;
-	}
-	/* Check if we have an idle VMID */
-	i = 0;
-	list_for_each_entry(idle, &id_mgr->ids_lru, list) {
-		fences[i] = amdgpu_sync_peek_fence(&idle->active, ring);
-		if (!fences[i])
-			break;
-		++i;
-	}
-
-	/* If we can't find a idle VMID to use, wait till one becomes available */
-	if (&idle->list == &id_mgr->ids_lru) {
-		u64 fence_context = adev->vm_manager.fence_context + ring->idx;
-		unsigned seqno = ++adev->vm_manager.seqno[ring->idx];
-		struct dma_fence_array *array;
-		unsigned j;
-
-		for (j = 0; j < i; ++j)
-			dma_fence_get(fences[j]);
-
-		array = dma_fence_array_create(i, fences, fence_context,
-					   seqno, true);
-		if (!array) {
-			for (j = 0; j < i; ++j)
-				dma_fence_put(fences[j]);
-			kfree(fences);
-			r = -ENOMEM;
-			goto error;
-		}
-
-
-		r = amdgpu_sync_fence(ring->adev, sync, &array->base, false);
-		dma_fence_put(&array->base);
-		if (r)
-			goto error;
-
-		mutex_unlock(&id_mgr->lock);
-		return 0;
-
-	}
-	kfree(fences);
-
-	job->vm_needs_flush = vm->use_cpu_for_update;
-	/* Check if we can use a VMID already assigned to this VM */
-	list_for_each_entry_reverse(id, &id_mgr->ids_lru, list) {
-		struct dma_fence *flushed;
-		bool needs_flush = vm->use_cpu_for_update;
-
-		/* Check all the prerequisites to using this VMID */
-		if (amdgpu_vm_had_gpu_reset(adev, id))
-			continue;
-
-		if (atomic64_read(&id->owner) != vm->client_id)
-			continue;
-
-		if (job->vm_pd_addr != id->pd_gpu_addr)
-			continue;
-
-		if (!id->last_flush ||
-		    (id->last_flush->context != fence_context &&
-		     !dma_fence_is_signaled(id->last_flush)))
-			needs_flush = true;
-
-		flushed  = id->flushed_updates;
-		if (updates && (!flushed || dma_fence_is_later(updates, flushed)))
-			needs_flush = true;
-
-		/* Concurrent flushes are only possible starting with Vega10 */
-		if (adev->asic_type < CHIP_VEGA10 && needs_flush)
-			continue;
-
-		/* Good we can use this VMID. Remember this submission as
-		 * user of the VMID.
-		 */
-		r = amdgpu_sync_fence(ring->adev, &id->active, fence, false);
-		if (r)
-			goto error;
-
-		if (updates && (!flushed || dma_fence_is_later(updates, flushed))) {
-			dma_fence_put(id->flushed_updates);
-			id->flushed_updates = dma_fence_get(updates);
-		}
-
-		if (needs_flush)
-			goto needs_flush;
-		else
-			goto no_flush_needed;
-
-	};
-
-	/* Still no ID to use? Then use the idle one found earlier */
-	id = idle;
-
-	/* Remember this submission as user of the VMID */
-	r = amdgpu_sync_fence(ring->adev, &id->active, fence, false);
-	if (r)
-		goto error;
-
-	id->pd_gpu_addr = job->vm_pd_addr;
-	dma_fence_put(id->flushed_updates);
-	id->flushed_updates = dma_fence_get(updates);
-	atomic64_set(&id->owner, vm->client_id);
-
-needs_flush:
-	job->vm_needs_flush = true;
-	dma_fence_put(id->last_flush);
-	id->last_flush = NULL;
-
-no_flush_needed:
-	list_move_tail(&id->list, &id_mgr->ids_lru);
-
-	job->vm_id = id - id_mgr->ids;
-	trace_amdgpu_vm_grab_id(vm, ring, job);
-
-error:
-	mutex_unlock(&id_mgr->lock);
-	return r;
-}
-
-static void amdgpu_vm_free_reserved_vmid(struct amdgpu_device *adev,
-					  struct amdgpu_vm *vm,
-					  unsigned vmhub)
-{
-	struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[vmhub];
-
-	mutex_lock(&id_mgr->lock);
-	if (vm->reserved_vmid[vmhub]) {
-		list_add(&vm->reserved_vmid[vmhub]->list,
-			&id_mgr->ids_lru);
-		vm->reserved_vmid[vmhub] = NULL;
-		atomic_dec(&id_mgr->reserved_vmid_num);
-	}
-	mutex_unlock(&id_mgr->lock);
-}
-
-static int amdgpu_vm_alloc_reserved_vmid(struct amdgpu_device *adev,
-					 struct amdgpu_vm *vm,
-					 unsigned vmhub)
-{
-	struct amdgpu_vm_id_manager *id_mgr;
-	struct amdgpu_vm_id *idle;
-	int r = 0;
-
-	id_mgr = &adev->vm_manager.id_mgr[vmhub];
-	mutex_lock(&id_mgr->lock);
-	if (vm->reserved_vmid[vmhub])
-		goto unlock;
-	if (atomic_inc_return(&id_mgr->reserved_vmid_num) >
-	    AMDGPU_VM_MAX_RESERVED_VMID) {
-		DRM_ERROR("Over limitation of reserved vmid\n");
-		atomic_dec(&id_mgr->reserved_vmid_num);
-		r = -EINVAL;
-		goto unlock;
-	}
-	/* Select the first entry VMID */
-	idle = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vm_id, list);
-	list_del_init(&idle->list);
-	vm->reserved_vmid[vmhub] = idle;
-	mutex_unlock(&id_mgr->lock);
-
-	return 0;
-unlock:
-	mutex_unlock(&id_mgr->lock);
-	return r;
-}
-
 /**
  * amdgpu_vm_check_compute_bug - check whether asic has compute vm bug
  *
@@ -767,14 +441,14 @@ bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring,
 {
 	struct amdgpu_device *adev = ring->adev;
 	unsigned vmhub = ring->funcs->vmhub;
-	struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[vmhub];
-	struct amdgpu_vm_id *id;
+	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	struct amdgpu_vmid *id;
 	bool gds_switch_needed;
 	bool vm_flush_needed = job->vm_needs_flush || ring->has_compute_vm_bug;
 
-	if (job->vm_id == 0)
+	if (job->vmid == 0)
 		return false;
-	id = &id_mgr->ids[job->vm_id];
+	id = &id_mgr->ids[job->vmid];
 	gds_switch_needed = ring->funcs->emit_gds_switch && (
 		id->gds_base != job->gds_base ||
 		id->gds_size != job->gds_size ||
@@ -783,7 +457,7 @@ bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring,
 		id->oa_base != job->oa_base ||
 		id->oa_size != job->oa_size);
 
-	if (amdgpu_vm_had_gpu_reset(adev, id))
+	if (amdgpu_vmid_had_gpu_reset(adev, id))
 		return true;
 
 	return vm_flush_needed || gds_switch_needed;
@@ -798,7 +472,7 @@ static bool amdgpu_vm_is_large_bar(struct amdgpu_device *adev)
  * amdgpu_vm_flush - hardware flush the vm
  *
  * @ring: ring to use for flush
- * @vm_id: vmid number to use
+ * @vmid: vmid number to use
  * @pd_addr: address of the page directory
  *
  * Emit a VM flush when it is necessary.
@@ -807,8 +481,8 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_
 {
 	struct amdgpu_device *adev = ring->adev;
 	unsigned vmhub = ring->funcs->vmhub;
-	struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[vmhub];
-	struct amdgpu_vm_id *id = &id_mgr->ids[job->vm_id];
+	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	struct amdgpu_vmid *id = &id_mgr->ids[job->vmid];
 	bool gds_switch_needed = ring->funcs->emit_gds_switch && (
 		id->gds_base != job->gds_base ||
 		id->gds_size != job->gds_size ||
@@ -820,7 +494,7 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_
 	unsigned patch_offset = 0;
 	int r;
 
-	if (amdgpu_vm_had_gpu_reset(adev, id)) {
+	if (amdgpu_vmid_had_gpu_reset(adev, id)) {
 		gds_switch_needed = true;
 		vm_flush_needed = true;
 	}
@@ -837,8 +511,8 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_
 	if (ring->funcs->emit_vm_flush && vm_flush_needed) {
 		struct dma_fence *fence;
 
-		trace_amdgpu_vm_flush(ring, job->vm_id, job->vm_pd_addr);
-		amdgpu_ring_emit_vm_flush(ring, job->vm_id, job->vm_pd_addr);
+		trace_amdgpu_vm_flush(ring, job->vmid, job->vm_pd_addr);
+		amdgpu_ring_emit_vm_flush(ring, job->vmid, job->vm_pd_addr);
 
 		r = amdgpu_fence_emit(ring, &fence);
 		if (r)
@@ -858,7 +532,7 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_
 		id->gws_size = job->gws_size;
 		id->oa_base = job->oa_base;
 		id->oa_size = job->oa_size;
-		amdgpu_ring_emit_gds_switch(ring, job->vm_id, job->gds_base,
+		amdgpu_ring_emit_gds_switch(ring, job->vmid, job->gds_base,
 					    job->gds_size, job->gws_base,
 					    job->gws_size, job->oa_base,
 					    job->oa_size);
@@ -875,49 +549,6 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_
 	return 0;
 }
 
-/**
- * amdgpu_vm_reset_id - reset VMID to zero
- *
- * @adev: amdgpu device structure
- * @vm_id: vmid number to use
- *
- * Reset saved GDW, GWS and OA to force switch on next flush.
- */
-void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vmhub,
-			unsigned vmid)
-{
-	struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[vmhub];
-	struct amdgpu_vm_id *id = &id_mgr->ids[vmid];
-
-	atomic64_set(&id->owner, 0);
-	id->gds_base = 0;
-	id->gds_size = 0;
-	id->gws_base = 0;
-	id->gws_size = 0;
-	id->oa_base = 0;
-	id->oa_size = 0;
-}
-
-/**
- * amdgpu_vm_reset_all_id - reset VMID to zero
- *
- * @adev: amdgpu device structure
- *
- * Reset VMID to force flush on next use
- */
-void amdgpu_vm_reset_all_ids(struct amdgpu_device *adev)
-{
-	unsigned i, j;
-
-	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
-		struct amdgpu_vm_id_manager *id_mgr =
-			&adev->vm_manager.id_mgr[i];
-
-		for (j = 1; j < id_mgr->num_ids; ++j)
-			amdgpu_vm_reset_id(adev, i, j);
-	}
-}
-
 /**
  * amdgpu_vm_bo_find - find the bo_va for a specific vm & bo
  *
@@ -1569,13 +1200,19 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
          *
          * The second command is for the shadow pagetables.
 	 */
-	ncmds = ((nptes >> min(adev->vm_manager.block_size, 11u)) + 1) * 2;
+	if (vm->root.base.bo->shadow)
+		ncmds = ((nptes >> min(adev->vm_manager.block_size, 11u)) + 1) * 2;
+	else
+		ncmds = ((nptes >> min(adev->vm_manager.block_size, 11u)) + 1);
 
 	/* padding, etc. */
 	ndw = 64;
 
 	/* one PDE write for each huge page */
-	ndw += ((nptes >> adev->vm_manager.block_size) + 1) * 6;
+	if (vm->root.base.bo->shadow)
+		ndw += ((nptes >> adev->vm_manager.block_size) + 1) * 6 * 2;
+	else
+		ndw += ((nptes >> adev->vm_manager.block_size) + 1) * 6;
 
 	if (pages_addr) {
 		/* copy commands needed */
@@ -2114,8 +1751,26 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
 	INIT_LIST_HEAD(&bo_va->valids);
 	INIT_LIST_HEAD(&bo_va->invalids);
 
-	if (bo)
-		list_add_tail(&bo_va->base.bo_list, &bo->va);
+	if (!bo)
+		return bo_va;
+
+	list_add_tail(&bo_va->base.bo_list, &bo->va);
+
+	if (bo->tbo.resv != vm->root.base.bo->tbo.resv)
+		return bo_va;
+
+	if (bo->preferred_domains &
+	    amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type))
+		return bo_va;
+
+	/*
+	 * We checked all the prerequisites, but it looks like this per VM BO
+	 * is currently evicted. add the BO to the evicted list to make sure it
+	 * is validated on next VM use to avoid fault.
+	 * */
+	spin_lock(&vm->status_lock);
+	list_move_tail(&bo_va->base.vm_status, &vm->evicted);
+	spin_unlock(&vm->status_lock);
 
 	return bo_va;
 }
@@ -2625,7 +2280,6 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	uint64_t init_pde_value = 0;
 
 	vm->va = RB_ROOT_CACHED;
-	vm->client_id = atomic64_inc_return(&adev->vm_manager.client_counter);
 	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
 		vm->reserved_vmid[i] = NULL;
 	spin_lock_init(&vm->status_lock);
@@ -2819,7 +2473,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	amdgpu_bo_unref(&root);
 	dma_fence_put(vm->last_update);
 	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
-		amdgpu_vm_free_reserved_vmid(adev, vm, i);
+		amdgpu_vmid_free_reserved(adev, vm, i);
 }
 
 /**
@@ -2861,23 +2515,9 @@ bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev,
  */
 void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 {
-	unsigned i, j;
-
-	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
-		struct amdgpu_vm_id_manager *id_mgr =
-			&adev->vm_manager.id_mgr[i];
+	unsigned i;
 
-		mutex_init(&id_mgr->lock);
-		INIT_LIST_HEAD(&id_mgr->ids_lru);
-		atomic_set(&id_mgr->reserved_vmid_num, 0);
-
-		/* skip over VMID 0, since it is the system VM */
-		for (j = 1; j < id_mgr->num_ids; ++j) {
-			amdgpu_vm_reset_id(adev, i, j);
-			amdgpu_sync_create(&id_mgr->ids[i].active);
-			list_add_tail(&id_mgr->ids[j].list, &id_mgr->ids_lru);
-		}
-	}
+	amdgpu_vmid_mgr_init(adev);
 
 	adev->vm_manager.fence_context =
 		dma_fence_context_alloc(AMDGPU_MAX_RINGS);
@@ -2885,7 +2525,6 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 		adev->vm_manager.seqno[i] = 0;
 
 	atomic_set(&adev->vm_manager.vm_pte_next_ring, 0);
-	atomic64_set(&adev->vm_manager.client_counter, 0);
 	spin_lock_init(&adev->vm_manager.prt_lock);
 	atomic_set(&adev->vm_manager.num_prt_users, 0);
 
@@ -2918,24 +2557,10 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev)
  */
 void amdgpu_vm_manager_fini(struct amdgpu_device *adev)
 {
-	unsigned i, j;
-
 	WARN_ON(!idr_is_empty(&adev->vm_manager.pasid_idr));
 	idr_destroy(&adev->vm_manager.pasid_idr);
 
-	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
-		struct amdgpu_vm_id_manager *id_mgr =
-			&adev->vm_manager.id_mgr[i];
-
-		mutex_destroy(&id_mgr->lock);
-		for (j = 0; j < AMDGPU_NUM_VM; ++j) {
-			struct amdgpu_vm_id *id = &id_mgr->ids[j];
-
-			amdgpu_sync_free(&id->active);
-			dma_fence_put(id->flushed_updates);
-			dma_fence_put(id->last_flush);
-		}
-	}
+	amdgpu_vmid_mgr_fini(adev);
 }
 
 int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
@@ -2948,13 +2573,12 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	switch (args->in.op) {
 	case AMDGPU_VM_OP_RESERVE_VMID:
 		/* current, we only have requirement to reserve vmid from gfxhub */
-		r = amdgpu_vm_alloc_reserved_vmid(adev, &fpriv->vm,
-						  AMDGPU_GFXHUB);
+		r = amdgpu_vmid_alloc_reserved(adev, &fpriv->vm, AMDGPU_GFXHUB);
 		if (r)
 			return r;
 		break;
 	case AMDGPU_VM_OP_UNRESERVE_VMID:
-		amdgpu_vm_free_reserved_vmid(adev, &fpriv->vm, AMDGPU_GFXHUB);
+		amdgpu_vmid_free_reserved(adev, &fpriv->vm, AMDGPU_GFXHUB);
 		break;
 	default:
 		return -EINVAL;

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

@@ -31,6 +31,7 @@
 
 #include "amdgpu_sync.h"
 #include "amdgpu_ring.h"
+#include "amdgpu_ids.h"
 
 struct amdgpu_bo_va;
 struct amdgpu_job;
@@ -40,9 +41,6 @@ struct amdgpu_bo_list_entry;
  * GPUVM handling
  */
 
-/* maximum number of VMIDs */
-#define AMDGPU_NUM_VM	16
-
 /* Maximum number of PTEs the hardware can write with one command */
 #define AMDGPU_VM_MAX_UPDATE_SIZE	0x3FFFF
 
@@ -193,11 +191,9 @@ struct amdgpu_vm {
 	/* Scheduler entity for page table updates */
 	struct drm_sched_entity	entity;
 
-	/* client id and PASID (TODO: replace client_id with PASID) */
-	u64                     client_id;
 	unsigned int		pasid;
 	/* dedicated to vm */
-	struct amdgpu_vm_id	*reserved_vmid[AMDGPU_MAX_VMHUBS];
+	struct amdgpu_vmid	*reserved_vmid[AMDGPU_MAX_VMHUBS];
 
 	/* Flag to indicate if VM tables are updated by CPU or GPU (SDMA) */
 	bool                    use_cpu_for_update;
@@ -212,37 +208,9 @@ struct amdgpu_vm {
 	unsigned int		fault_credit;
 };
 
-struct amdgpu_vm_id {
-	struct list_head	list;
-	struct amdgpu_sync	active;
-	struct dma_fence		*last_flush;
-	atomic64_t		owner;
-
-	uint64_t		pd_gpu_addr;
-	/* last flushed PD/PT update */
-	struct dma_fence		*flushed_updates;
-
-	uint32_t                current_gpu_reset_count;
-
-	uint32_t		gds_base;
-	uint32_t		gds_size;
-	uint32_t		gws_base;
-	uint32_t		gws_size;
-	uint32_t		oa_base;
-	uint32_t		oa_size;
-};
-
-struct amdgpu_vm_id_manager {
-	struct mutex		lock;
-	unsigned		num_ids;
-	struct list_head	ids_lru;
-	struct amdgpu_vm_id	ids[AMDGPU_NUM_VM];
-	atomic_t		reserved_vmid_num;
-};
-
 struct amdgpu_vm_manager {
 	/* Handling of VMIDs */
-	struct amdgpu_vm_id_manager		id_mgr[AMDGPU_MAX_VMHUBS];
+	struct amdgpu_vmid_mgr			id_mgr[AMDGPU_MAX_VMHUBS];
 
 	/* Handling of VM fences */
 	u64					fence_context;
@@ -260,8 +228,6 @@ struct amdgpu_vm_manager {
 	struct amdgpu_ring                      *vm_pte_rings[AMDGPU_MAX_RINGS];
 	unsigned				vm_pte_num_rings;
 	atomic_t				vm_pte_next_ring;
-	/* client id counter */
-	atomic64_t				client_counter;
 
 	/* partial resident texture handling */
 	spinlock_t				prt_lock;
@@ -280,8 +246,6 @@ struct amdgpu_vm_manager {
 	spinlock_t				pasid_lock;
 };
 
-int amdgpu_vm_alloc_pasid(unsigned int bits);
-void amdgpu_vm_free_pasid(unsigned int pasid);
 void amdgpu_vm_manager_init(struct amdgpu_device *adev);
 void amdgpu_vm_manager_fini(struct amdgpu_device *adev);
 int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
@@ -299,13 +263,7 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 int amdgpu_vm_alloc_pts(struct amdgpu_device *adev,
 			struct amdgpu_vm *vm,
 			uint64_t saddr, uint64_t size);
-int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
-		      struct amdgpu_sync *sync, struct dma_fence *fence,
-		      struct amdgpu_job *job);
 int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_pipe_sync);
-void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vmhub,
-			unsigned vmid);
-void amdgpu_vm_reset_all_ids(struct amdgpu_device *adev);
 int amdgpu_vm_update_directories(struct amdgpu_device *adev,
 				 struct amdgpu_vm *vm);
 int amdgpu_vm_clear_freed(struct amdgpu_device *adev,

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

@@ -280,7 +280,7 @@ static void cik_ih_decode_iv(struct amdgpu_device *adev,
 	entry->src_id = dw[0] & 0xff;
 	entry->src_data[0] = dw[1] & 0xfffffff;
 	entry->ring_id = dw[2] & 0xff;
-	entry->vm_id = (dw[2] >> 8) & 0xff;
+	entry->vmid = (dw[2] >> 8) & 0xff;
 	entry->pas_id = (dw[2] >> 16) & 0xffff;
 
 	/* wptr/rptr are in bytes! */

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

@@ -221,9 +221,9 @@ static void cik_sdma_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
  */
 static void cik_sdma_ring_emit_ib(struct amdgpu_ring *ring,
 				  struct amdgpu_ib *ib,
-				  unsigned vm_id, bool ctx_switch)
+				  unsigned vmid, bool ctx_switch)
 {
-	u32 extra_bits = vm_id & 0xf;
+	u32 extra_bits = vmid & 0xf;
 
 	/* IB packet must end on a 8 DW boundary */
 	cik_sdma_ring_insert_nop(ring, (12 - (lower_32_bits(ring->wptr) & 7)) % 8);
@@ -880,23 +880,23 @@ static void cik_sdma_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
  * using sDMA (CIK).
  */
 static void cik_sdma_ring_emit_vm_flush(struct amdgpu_ring *ring,
-					unsigned vm_id, uint64_t pd_addr)
+					unsigned vmid, uint64_t pd_addr)
 {
 	u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(0) |
 			  SDMA_POLL_REG_MEM_EXTRA_FUNC(0)); /* always */
 
 	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
-	if (vm_id < 8) {
-		amdgpu_ring_write(ring, (mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + vm_id));
+	if (vmid < 8) {
+		amdgpu_ring_write(ring, (mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + vmid));
 	} else {
-		amdgpu_ring_write(ring, (mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vm_id - 8));
+		amdgpu_ring_write(ring, (mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8));
 	}
 	amdgpu_ring_write(ring, pd_addr >> 12);
 
 	/* flush TLB */
 	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
 	amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST);
-	amdgpu_ring_write(ring, 1 << vm_id);
+	amdgpu_ring_write(ring, 1 << vmid);
 
 	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_POLL_REG_MEM, 0, extra_bits));
 	amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST << 2);

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

@@ -259,7 +259,7 @@ static void cz_ih_decode_iv(struct amdgpu_device *adev,
 	entry->src_id = dw[0] & 0xff;
 	entry->src_data[0] = dw[1] & 0xfffffff;
 	entry->ring_id = dw[2] & 0xff;
-	entry->vm_id = (dw[2] >> 8) & 0xff;
+	entry->vmid = (dw[2] >> 8) & 0xff;
 	entry->pas_id = (dw[2] >> 16) & 0xffff;
 
 	/* wptr/rptr are in bytes! */