path: root/common/recipes-kernel/linux/linux-yocto-4.19.8/1285-drm-amdgpu-replace-get_user_pages-with-HMM-mirror-he.patch

From 3bf8b153e39c41e2e9d9398075fcdbc608b85d69 Mon Sep 17 00:00:00 2001
From: Philip Yang <Philip.Yang@amd.com>
Date: Thu, 13 Dec 2018 15:35:28 -0500
Subject: [PATCH 1285/2940] drm/amdgpu: replace get_user_pages with HMM mirror
 helpers
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Use HMM helper function hmm_vma_fault() to get physical pages backing
userptr and start CPU page table update track of those pages. Then use
hmm_vma_range_done() to check if those pages are updated before
amdgpu_cs_submit for gfx or before user queues are resumed for kfd.

If userptr pages are updated, for gfx, amdgpu_cs_ioctl will restart
from scratch, for kfd, restore worker is rescheduled to retry.

HMM simplify the CPU page table concurrent update check, so remove
guptasklock, mmu_invalidations, last_set_pages fields from
amdgpu_ttm_tt struct.

HMM does not pin the page (increase page ref count), so remove related
operations like release_pages(), put_page(), mark_page_dirty().

Change-Id: Ibec614d6aed9f8c414e9bab11cee29b849511c4d
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Chaudhary Amit Kumar <Chaudharyamit.Kumar@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h |   2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c      | 140 ++++++---------
 drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c     |  14 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c      |  25 ++-
 drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h      |   4 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c     | 178 ++++++++------------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h     |   3 +-
 7 files changed, 157 insertions(+), 209 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h
index 7c5f5d1601e6..a130e766cbdb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h
@@ -36,7 +36,7 @@ struct amdgpu_bo_list_entry {
 	struct amdgpu_bo_va		*bo_va;
 	uint32_t			priority;
 	struct page			**user_pages;
-	int				user_invalidated;
+	bool				user_invalidated;
 };
 
 struct amdgpu_bo_list {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index 7a4508f1ab81..adfebc79124b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -53,7 +53,6 @@ static int amdgpu_cs_user_fence_chunk(struct amdgpu_cs_parser *p,
 	p->uf_entry.tv.bo = &bo->tbo;
 	/* One for TTM and one for the CS job */
 	p->uf_entry.tv.num_shared = 2;
-	p->uf_entry.user_pages = NULL;
 
 	drm_gem_object_put_unlocked(gobj);
 
@@ -541,14 +540,14 @@ static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p,
 		if (usermm && usermm != current->mm)
 			return -EPERM;
 
-		/* Check if we have user pages and nobody bound the BO already */
-		if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm) &&
-		    lobj->user_pages) {
+		if (amdgpu_ttm_tt_is_userptr(bo->tbo.ttm) &&
+		    lobj->user_invalidated && lobj->user_pages) {
 			amdgpu_bo_placement_from_domain(bo,
 							AMDGPU_GEM_DOMAIN_CPU);
 			r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 			if (r)
 				return r;
+
 			amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm,
 						     lobj->user_pages);
 			binding_userptr = true;
@@ -579,7 +578,6 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 	struct amdgpu_bo *gds;
 	struct amdgpu_bo *gws;
 	struct amdgpu_bo *oa;
-	unsigned tries = 10;
 	int r;
 
 	INIT_LIST_HEAD(&p->validated);
@@ -615,79 +613,45 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 	if (p->uf_entry.tv.bo && !ttm_to_amdgpu_bo(p->uf_entry.tv.bo)->parent)
 		list_add(&p->uf_entry.tv.head, &p->validated);
 
-	while (1) {
-		struct list_head need_pages;
-
-		r = ttm_eu_reserve_buffers(&p->ticket, &p->validated, true,
-					   &duplicates);
-		if (unlikely(r != 0)) {
-			if (r != -ERESTARTSYS)
-				DRM_ERROR("ttm_eu_reserve_buffers failed.\n");
-			goto error_free_pages;
-		}
-
-		INIT_LIST_HEAD(&need_pages);
-		amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
-			struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
-
-			if (amdgpu_ttm_tt_userptr_invalidated(bo->tbo.ttm,
-				 &e->user_invalidated) && e->user_pages) {
-
-				/* We acquired a page array, but somebody
-				 * invalidated it. Free it and try again
-				 */
-				release_pages(e->user_pages,
-					      bo->tbo.ttm->num_pages);
-				kvfree(e->user_pages);
-				e->user_pages = NULL;
-			}
-
-			if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm) &&
-			    !e->user_pages) {
-				list_del(&e->tv.head);
-				list_add(&e->tv.head, &need_pages);
-
-				amdgpu_bo_unreserve(bo);
-			}
+	/* Get userptr backing pages. If pages are updated after registered
+	 * in amdgpu_gem_userptr_ioctl(), amdgpu_cs_list_validate() will do
+	 * amdgpu_ttm_backend_bind() to flush and invalidate new pages
+	 */
+	amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
+		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
+		bool userpage_invalidated = false;
+		int i;
+
+		e->user_pages = kvmalloc_array(bo->tbo.ttm->num_pages,
+					sizeof(struct page *),
+					GFP_KERNEL | __GFP_ZERO);
+		if (!e->user_pages) {
+			DRM_ERROR("calloc failure\n");
+			return -ENOMEM;
 		}
 
-		if (list_empty(&need_pages))
-			break;
-
-		/* Unreserve everything again. */
-		ttm_eu_backoff_reservation(&p->ticket, &p->validated);
-
-		/* We tried too many times, just abort */
-		if (!--tries) {
-			r = -EDEADLK;
-			DRM_ERROR("deadlock in %s\n", __func__);
-			goto error_free_pages;
+		r = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm, e->user_pages);
+		if (r) {
+			kvfree(e->user_pages);
+			e->user_pages = NULL;
+			return r;
 		}
 
-		/* Fill the page arrays for all userptrs. */
-		list_for_each_entry(e, &need_pages, tv.head) {
-			struct ttm_tt *ttm = e->tv.bo->ttm;
-
-			e->user_pages = kvmalloc_array(ttm->num_pages,
-							 sizeof(struct page*),
-							 GFP_KERNEL | __GFP_ZERO);
-			if (!e->user_pages) {
-				r = -ENOMEM;
-				DRM_ERROR("calloc failure in %s\n", __func__);
-				goto error_free_pages;
-			}
-
-			r = amdgpu_ttm_tt_get_user_pages(ttm, e->user_pages);
-			if (r) {
-				DRM_ERROR("amdgpu_ttm_tt_get_user_pages failed.\n");
-				kvfree(e->user_pages);
-				e->user_pages = NULL;
-				goto error_free_pages;
+		for (i = 0; i < bo->tbo.ttm->num_pages; i++) {
+			if (bo->tbo.ttm->pages[i] != e->user_pages[i]) {
+				userpage_invalidated = true;
+				break;
 			}
 		}
+		e->user_invalidated = userpage_invalidated;
+	}
 
-		/* And try again. */
-		list_splice(&need_pages, &p->validated);
+	r = ttm_eu_reserve_buffers(&p->ticket, &p->validated, true,
+				   &duplicates);
+	if (unlikely(r != 0)) {
+		if (r != -ERESTARTSYS)
+			DRM_ERROR("ttm_eu_reserve_buffers failed.\n");
+		goto out;
 	}
 
 	amdgpu_cs_get_threshold_for_moves(p->adev, &p->bytes_moved_threshold,
@@ -756,19 +720,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 error_validate:
 	if (r)
 		ttm_eu_backoff_reservation(&p->ticket, &p->validated);
-
-error_free_pages:
-
-	amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
-		if (!e->user_pages)
-			continue;
-
-		release_pages(e->user_pages,
-				bo->tbo.ttm->num_pages);
-		kvfree(e->user_pages);
-	}
-
+out:
 	return r;
 }
 
@@ -1227,7 +1179,6 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	struct amdgpu_bo_list_entry *e;
 	struct amdgpu_job *job;
 	uint64_t seq;
-
 	int r;
 
 	job = p->job;
@@ -1237,15 +1188,23 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	if (r)
 		goto error_unlock;
 
-	/* No memory allocation is allowed while holding the mn lock */
+	/* No memory allocation is allowed while holding the mn lock.
+	 * p->mn is hold until amdgpu_cs_submit is finished and fence is added
+	 * to BOs.
+	 */
 	amdgpu_mn_lock(p->mn);
+
+	/* If userptr are invalidated after amdgpu_cs_parser_bos(), return
+	 * -EAGAIN, drmIoctl in libdrm will restart the amdgpu_cs_ioctl.
+	 */
 	amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
 		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
 
-		if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm)) {
-			r = -ERESTARTSYS;
-			goto error_abort;
-		}
+		r |= !amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm);
+	}
+	if (r) {
+		r = -EAGAIN;
+		goto error_abort;
 	}
 
 	job->owner = p->filp;
@@ -1342,6 +1301,7 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 
 out:
 	amdgpu_cs_parser_fini(&parser, r, reserved_buffers);
+
 	return r;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
index 254a7c6c5724..48b0ca33ac06 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
@@ -360,26 +360,24 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 
 		r = amdgpu_bo_reserve(bo, true);
 		if (r)
-			goto free_pages;
+			goto user_pages_done;
 
 		amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT);
 		r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 		amdgpu_bo_unreserve(bo);
 		if (r)
-			goto free_pages;
+			goto user_pages_done;
 	}
 
 	r = drm_gem_handle_create(filp, gobj, &handle);
-	/* drop reference from allocate - handle holds it now */
-	drm_gem_object_put_unlocked(gobj);
 	if (r)
-		return r;
+		goto user_pages_done;
 
 	args->handle = handle;
-	return 0;
 
-free_pages:
-	release_pages(bo->tbo.ttm->pages, bo->tbo.ttm->num_pages);
+user_pages_done:
+	if (args->flags & AMDGPU_GEM_USERPTR_VALIDATE)
+		amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm);
 
 release_object:
 	drm_gem_object_put_unlocked(gobj);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c
index 090b5793daef..4ac567dc4646 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c
@@ -220,8 +220,6 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node,
 			true, false, MAX_SCHEDULE_TIMEOUT);
 		if (r <= 0)
 			DRM_ERROR("(%ld) failed to wait for user bo\n", r);
-
-		amdgpu_ttm_tt_mark_user_pages(bo->tbo.ttm);
 	}
 }
 
@@ -533,3 +531,26 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo)
 	mutex_unlock(&adev->mn_lock);
 }
 
+/* flags used by HMM internal, not related to CPU/GPU PTE flags */
+static const uint64_t hmm_range_flags[HMM_PFN_FLAG_MAX] = {
+		(1 << 0), /* HMM_PFN_VALID */
+		(1 << 1), /* HMM_PFN_WRITE */
+		0 /* HMM_PFN_DEVICE_PRIVATE */
+};
+
+static const uint64_t hmm_range_values[HMM_PFN_VALUE_MAX] = {
+		0xfffffffffffffffeUL, /* HMM_PFN_ERROR */
+		0, /* HMM_PFN_NONE */
+		0xfffffffffffffffcUL /* HMM_PFN_SPECIAL */
+};
+
+void amdgpu_hmm_init_range(struct hmm_range *range)
+{
+	if (range) {
+		range->flags = hmm_range_flags;
+		range->values = hmm_range_values;
+		range->pfn_shift = PAGE_SHIFT;
+		range->pfns = NULL;
+		INIT_LIST_HEAD(&range->list);
+	}
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h
index 0a51fd00021c..4803e216e174 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h
@@ -25,9 +25,10 @@
 #define __AMDGPU_MN_H__
 
 /*
- * MMU Notifier
+ * HMM mirror
  */
 struct amdgpu_mn;
+struct hmm_range;
 
 enum amdgpu_mn_type {
 	AMDGPU_MN_TYPE_GFX,
@@ -41,6 +42,7 @@ struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev,
 				enum amdgpu_mn_type type);
 int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr);
 void amdgpu_mn_unregister(struct amdgpu_bo *bo);
+void amdgpu_hmm_init_range(struct hmm_range *range);
 #else
 static inline void amdgpu_mn_lock(struct amdgpu_mn *mn) {}
 static inline void amdgpu_mn_unlock(struct amdgpu_mn *mn) {}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index fe853624e293..d587b374665e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -43,6 +43,7 @@
 #include <linux/pagemap.h>
 #include <linux/debugfs.h>
 #include <linux/iommu.h>
+#include <linux/hmm.h>
 #include "amdgpu.h"
 #include "amdgpu_object.h"
 #include "amdgpu_trace.h"
@@ -838,98 +839,102 @@ static unsigned long amdgpu_ttm_io_mem_pfn(struct ttm_buffer_object *bo,
 /*
  * TTM backend functions.
  */
-struct amdgpu_ttm_gup_task_list {
-	struct list_head	list;
-	struct task_struct	*task;
-};
-
 struct amdgpu_ttm_tt {
 	struct ttm_dma_tt	ttm;
 	u64			offset;
 	uint64_t		userptr;
 	struct task_struct	*usertask;
 	uint32_t		userflags;
-	spinlock_t              guptasklock;
-	struct list_head        guptasks;
-	atomic_t		mmu_invalidations;
-	uint32_t		last_set_pages;
+	struct hmm_range	range;
 };
 
 /**
- * amdgpu_ttm_tt_get_user_pages - Pin pages of memory pointed to by a USERPTR
- * pointer to memory
+ * amdgpu_ttm_tt_get_user_pages - get device accessible pages that back user
+ * memory and start HMM tracking CPU page table update
  *
- * Called by amdgpu_gem_userptr_ioctl() and amdgpu_cs_parser_bos().
- * This provides a wrapper around the get_user_pages() call to provide
- * device accessible pages that back user memory.
+ * Calling function must call amdgpu_ttm_tt_userptr_range_done() once and only
+ * once afterwards to stop HMM tracking
  */
 int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
 {
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
 	struct mm_struct *mm = gtt->usertask->mm;
-	unsigned int flags = 0;
-	unsigned pinned = 0;
-	int r;
+	unsigned long end = gtt->userptr + ttm->num_pages * PAGE_SIZE;
+	struct hmm_range *range = &gtt->range;
+	int r = 0, i;
 
 	if (!mm) /* Happens during process shutdown */
 		return -ESRCH;
 
-	if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY))
-		flags |= FOLL_WRITE;
+	amdgpu_hmm_init_range(range);
 
 	down_read(&mm->mmap_sem);
 
-	if (gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) {
-		/*
-		 * check that we only use anonymous memory to prevent problems
-		 * with writeback
-		 */
-		unsigned long end = gtt->userptr + ttm->num_pages * PAGE_SIZE;
-		struct vm_area_struct *vma;
+	range->vma = find_vma(mm, gtt->userptr);
+	if (!range_in_vma(range->vma, gtt->userptr, end))
+		r = -EFAULT;
+	else if ((gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) &&
+		range->vma->vm_file)
+		r = -EPERM;
+	if (r)
+		goto out;
 
-		vma = find_vma(mm, gtt->userptr);
-		if (!vma || vma->vm_file || vma->vm_end < end) {
-			up_read(&mm->mmap_sem);
-			return -EPERM;
-		}
+	range->pfns = kvmalloc_array(ttm->num_pages, sizeof(uint64_t),
+				     GFP_KERNEL);
+	if (range->pfns == NULL) {
+		r = -ENOMEM;
+		goto out;
 	}
+	range->start = gtt->userptr;
+	range->end = end;
 
-	/* loop enough times using contiguous pages of memory */
-	do {
-		unsigned num_pages = ttm->num_pages - pinned;
-		uint64_t userptr = gtt->userptr + pinned * PAGE_SIZE;
-		struct page **p = pages + pinned;
-		struct amdgpu_ttm_gup_task_list guptask;
+	range->pfns[0] = range->flags[HMM_PFN_VALID];
+	range->pfns[0] |= amdgpu_ttm_tt_is_readonly(ttm) ?
+				0 : range->flags[HMM_PFN_WRITE];
+	for (i = 1; i < ttm->num_pages; i++)
+		range->pfns[i] = range->pfns[0];
 
-		guptask.task = current;
-		spin_lock(&gtt->guptasklock);
-		list_add(&guptask.list, &gtt->guptasks);
-		spin_unlock(&gtt->guptasklock);
+	/* This may trigger page table update */
+	r = hmm_vma_fault(range, true);
+	if (r)
+		goto out_free_pfns;
 
-		if (mm == current->mm)
-			r = get_user_pages(userptr, num_pages, flags, p, NULL);
-		else
-			r = get_user_pages_remote(gtt->usertask,
-					mm, userptr, num_pages,
-					flags, p, NULL, NULL);
+	up_read(&mm->mmap_sem);
 
-		spin_lock(&gtt->guptasklock);
-		list_del(&guptask.list);
-		spin_unlock(&gtt->guptasklock);
+	for (i = 0; i < ttm->num_pages; i++)
+		pages[i] = hmm_pfn_to_page(range, range->pfns[i]);
 
-		if (r < 0)
-			goto release_pages;
+	return 0;
 
-		pinned += r;
+out_free_pfns:
+	kvfree(range->pfns);
+	range->pfns = NULL;
+out:
+	up_read(&mm->mmap_sem);
+	return r;
+}
 
-	} while (pinned < ttm->num_pages);
+/**
+ * amdgpu_ttm_tt_userptr_range_done - stop HMM track the CPU page table change
+ * Check if the pages backing this ttm range have been invalidated
+ *
+ * Returns: true if pages are still valid
+ */
+bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm)
+{
+	struct amdgpu_ttm_tt *gtt = (void *)ttm;
+	bool r = false;
 
-	up_read(&mm->mmap_sem);
-	return 0;
+	if (!gtt || !gtt->userptr)
+		return false;
+
+	WARN_ONCE(!gtt->range.pfns, "No user pages to check\n");
+	if (gtt->range.pfns) {
+		r = hmm_vma_range_done(&gtt->range);
+		kvfree(gtt->range.pfns);
+		gtt->range.pfns = NULL;
+	}
 
-release_pages:
-	release_pages(pages, pinned);
-	up_read(&mm->mmap_sem);
 	return r;
 }
 
@@ -942,16 +947,10 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
  */
 void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages)
 {
-	struct amdgpu_ttm_tt *gtt = (void *)ttm;
 	unsigned i;
 
-	gtt->last_set_pages = atomic_read(&gtt->mmu_invalidations);
-	for (i = 0; i < ttm->num_pages; ++i) {
-		if (ttm->pages[i])
-			put_page(ttm->pages[i]);
-
+	for (i = 0; i < ttm->num_pages; ++i)
 		ttm->pages[i] = pages ? pages[i] : NULL;
-	}
 }
 
 /**
@@ -1036,10 +1035,11 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm)
 	/* unmap the pages mapped to the device */
 	dma_unmap_sg(adev->dev, ttm->sg->sgl, ttm->sg->nents, direction);
 
-	/* mark the pages as dirty */
-	amdgpu_ttm_tt_mark_user_pages(ttm);
-
 	sg_free_table(ttm->sg);
+
+	if (gtt->range.pfns &&
+	    ttm->pages[0] == hmm_pfn_to_page(&gtt->range, gtt->range.pfns[0]))
+		WARN_ONCE(1, "Missing get_user_page_done\n");
 }
 
 int amdgpu_ttm_gart_bind(struct amdgpu_device *adev,
@@ -1390,11 +1390,6 @@ int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
 	gtt->usertask = current->group_leader;
 	get_task_struct(gtt->usertask);
 
-	spin_lock_init(&gtt->guptasklock);
-	INIT_LIST_HEAD(&gtt->guptasks);
-	atomic_set(&gtt->mmu_invalidations, 0);
-	gtt->last_set_pages = 0;
-
 	return 0;
 }
 
@@ -1423,7 +1418,6 @@ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
 				  unsigned long end)
 {
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
-	struct amdgpu_ttm_gup_task_list *entry;
 	unsigned long size;
 
 	if (gtt == NULL || !gtt->userptr)
@@ -1436,48 +1430,20 @@ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
 	if (gtt->userptr > end || gtt->userptr + size <= start)
 		return false;
 
-	/* Search the lists of tasks that hold this mapping and see
-	 * if current is one of them.  If it is return false.
-	 */
-	spin_lock(&gtt->guptasklock);
-	list_for_each_entry(entry, &gtt->guptasks, list) {
-		if (entry->task == current) {
-			spin_unlock(&gtt->guptasklock);
-			return false;
-		}
-	}
-	spin_unlock(&gtt->guptasklock);
-
-	atomic_inc(&gtt->mmu_invalidations);
-
 	return true;
 }
 
 /**
- * amdgpu_ttm_tt_userptr_invalidated - Has the ttm_tt object been invalidated?
+ * amdgpu_ttm_tt_is_userptr - Have the pages backing by userptr?
  */
-bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
-				       int *last_invalidated)
-{
-	struct amdgpu_ttm_tt *gtt = (void *)ttm;
-	int prev_invalidated = *last_invalidated;
-
-	*last_invalidated = atomic_read(&gtt->mmu_invalidations);
-	return prev_invalidated != *last_invalidated;
-}
-
-/**
- * amdgpu_ttm_tt_userptr_needs_pages - Have the pages backing this ttm_tt object
- * been invalidated since the last time they've been set?
- */
-bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm)
+bool amdgpu_ttm_tt_is_userptr(struct ttm_tt *ttm)
 {
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
 
 	if (gtt == NULL || !gtt->userptr)
 		return false;
 
-	return atomic_read(&gtt->mmu_invalidations) != gtt->last_set_pages;
+	return true;
 }
 
 /**
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
index ef0f86f688a9..9afdf06717f7 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
@@ -108,6 +108,7 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo);
 int amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo);
 
 int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages);
+bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm);
 void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages);
 void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm);
 int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
@@ -118,7 +119,7 @@ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
 				  unsigned long end);
 bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
 				       int *last_invalidated);
-bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm);
+bool amdgpu_ttm_tt_is_userptr(struct ttm_tt *ttm);
 bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm);
 uint64_t amdgpu_ttm_tt_pde_flags(struct ttm_tt *ttm, struct ttm_mem_reg *mem);
 uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm,
-- 
2.17.1