1 files changed, 191 insertions, 63 deletions

diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index 3451a82adba..1df856f7856 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -28,8 +28,9 @@
  * Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
  */
 
-#include "ttm/ttm_bo_driver.h"
-#include "ttm/ttm_placement.h"
+#include <drm/ttm/ttm_bo_driver.h>
+#include <drm/ttm/ttm_placement.h>
+#include <drm/drm_vma_manager.h>
 #include <linux/io.h>
 #include <linux/highmem.h>
 #include <linux/wait.h>
@@ -39,18 +40,11 @@
 
 void ttm_bo_free_old_node(struct ttm_buffer_object *bo)
 {
-	struct ttm_mem_reg *old_mem = &bo->mem;
-
-	if (old_mem->mm_node) {
-		spin_lock(&bo->glob->lru_lock);
-		drm_mm_put_block(old_mem->mm_node);
-		spin_unlock(&bo->glob->lru_lock);
-	}
-	old_mem->mm_node = NULL;
+	ttm_bo_mem_put(bo, &bo->mem);
 }
 
 int ttm_bo_move_ttm(struct ttm_buffer_object *bo,
-		    bool evict, bool no_wait_reserve,
+		    bool evict,
 		    bool no_wait_gpu, struct ttm_mem_reg *new_mem)
 {
 	struct ttm_tt *ttm = bo->ttm;
@@ -82,37 +76,128 @@ int ttm_bo_move_ttm(struct ttm_buffer_object *bo,
 }
 EXPORT_SYMBOL(ttm_bo_move_ttm);
 
-int ttm_mem_io_reserve(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem)
+int ttm_mem_io_lock(struct ttm_mem_type_manager *man, bool interruptible)
 {
-	int ret;
+	if (likely(man->io_reserve_fastpath))
+		return 0;
+
+	if (interruptible)
+		return mutex_lock_interruptible(&man->io_reserve_mutex);
+
+	mutex_lock(&man->io_reserve_mutex);
+	return 0;
+}
+EXPORT_SYMBOL(ttm_mem_io_lock);
+
+void ttm_mem_io_unlock(struct ttm_mem_type_manager *man)
+{
+	if (likely(man->io_reserve_fastpath))
+		return;
 
-	if (!mem->bus.io_reserved) {
-		mem->bus.io_reserved = true;
+	mutex_unlock(&man->io_reserve_mutex);
+}
+EXPORT_SYMBOL(ttm_mem_io_unlock);
+
+static int ttm_mem_io_evict(struct ttm_mem_type_manager *man)
+{
+	struct ttm_buffer_object *bo;
+
+	if (!man->use_io_reserve_lru || list_empty(&man->io_reserve_lru))
+		return -EAGAIN;
+
+	bo = list_first_entry(&man->io_reserve_lru,
+			      struct ttm_buffer_object,
+			      io_reserve_lru);
+	list_del_init(&bo->io_reserve_lru);
+	ttm_bo_unmap_virtual_locked(bo);
+
+	return 0;
+}
+
+
+int ttm_mem_io_reserve(struct ttm_bo_device *bdev,
+		       struct ttm_mem_reg *mem)
+{
+	struct ttm_mem_type_manager *man = &bdev->man[mem->mem_type];
+	int ret = 0;
+
+	if (!bdev->driver->io_mem_reserve)
+		return 0;
+	if (likely(man->io_reserve_fastpath))
+		return bdev->driver->io_mem_reserve(bdev, mem);
+
+	if (bdev->driver->io_mem_reserve &&
+	    mem->bus.io_reserved_count++ == 0) {
+retry:
 		ret = bdev->driver->io_mem_reserve(bdev, mem);
+		if (ret == -EAGAIN) {
+			ret = ttm_mem_io_evict(man);
+			if (ret == 0)
+				goto retry;
+		}
+	}
+	return ret;
+}
+EXPORT_SYMBOL(ttm_mem_io_reserve);
+
+void ttm_mem_io_free(struct ttm_bo_device *bdev,
+		     struct ttm_mem_reg *mem)
+{
+	struct ttm_mem_type_manager *man = &bdev->man[mem->mem_type];
+
+	if (likely(man->io_reserve_fastpath))
+		return;
+
+	if (bdev->driver->io_mem_reserve &&
+	    --mem->bus.io_reserved_count == 0 &&
+	    bdev->driver->io_mem_free)
+		bdev->driver->io_mem_free(bdev, mem);
+
+}
+EXPORT_SYMBOL(ttm_mem_io_free);
+
+int ttm_mem_io_reserve_vm(struct ttm_buffer_object *bo)
+{
+	struct ttm_mem_reg *mem = &bo->mem;
+	int ret;
+
+	if (!mem->bus.io_reserved_vm) {
+		struct ttm_mem_type_manager *man =
+			&bo->bdev->man[mem->mem_type];
+
+		ret = ttm_mem_io_reserve(bo->bdev, mem);
 		if (unlikely(ret != 0))
 			return ret;
+		mem->bus.io_reserved_vm = true;
+		if (man->use_io_reserve_lru)
+			list_add_tail(&bo->io_reserve_lru,
+				      &man->io_reserve_lru);
 	}
 	return 0;
 }
 
-void ttm_mem_io_free(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem)
+void ttm_mem_io_free_vm(struct ttm_buffer_object *bo)
 {
-	if (bdev->driver->io_mem_reserve) {
-		if (mem->bus.io_reserved) {
-			mem->bus.io_reserved = false;
-			bdev->driver->io_mem_free(bdev, mem);
-		}
+	struct ttm_mem_reg *mem = &bo->mem;
+
+	if (mem->bus.io_reserved_vm) {
+		mem->bus.io_reserved_vm = false;
+		list_del_init(&bo->io_reserve_lru);
+		ttm_mem_io_free(bo->bdev, mem);
 	}
 }
 
-int ttm_mem_reg_ioremap(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem,
+static int ttm_mem_reg_ioremap(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem,
 			void **virtual)
 {
+	struct ttm_mem_type_manager *man = &bdev->man[mem->mem_type];
 	int ret;
 	void *addr;
 
 	*virtual = NULL;
+	(void) ttm_mem_io_lock(man, false);
 	ret = ttm_mem_io_reserve(bdev, mem);
+	ttm_mem_io_unlock(man);
 	if (ret || !mem->bus.is_iomem)
 		return ret;
 
@@ -124,7 +209,9 @@ int ttm_mem_reg_ioremap(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem,
 		else
 			addr = ioremap_nocache(mem->bus.base + mem->bus.offset, mem->bus.size);
 		if (!addr) {
+			(void) ttm_mem_io_lock(man, false);
 			ttm_mem_io_free(bdev, mem);
+			ttm_mem_io_unlock(man);
 			return -ENOMEM;
 		}
 	}
@@ -132,7 +219,7 @@ int ttm_mem_reg_ioremap(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem,
 	return 0;
 }
 
-void ttm_mem_reg_iounmap(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem,
+static void ttm_mem_reg_iounmap(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem,
 			 void *virtual)
 {
 	struct ttm_mem_type_manager *man;
@@ -141,7 +228,9 @@ void ttm_mem_reg_iounmap(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem,
 
 	if (virtual && mem->bus.addr == NULL)
 		iounmap(virtual);
+	(void) ttm_mem_io_lock(man, false);
 	ttm_mem_io_free(bdev, mem);
+	ttm_mem_io_unlock(man);
 }
 
 static int ttm_copy_io_page(void *dst, void *src, unsigned long page)
@@ -161,7 +250,7 @@ static int ttm_copy_io_ttm_page(struct ttm_tt *ttm, void *src,
 				unsigned long page,
 				pgprot_t prot)
 {
-	struct page *d = ttm_tt_get_page(ttm, page);
+	struct page *d = ttm->pages[page];
 	void *dst;
 
 	if (!d)
@@ -170,7 +259,7 @@ static int ttm_copy_io_ttm_page(struct ttm_tt *ttm, void *src,
 	src = (void *)((unsigned long)src + (page << PAGE_SHIFT));
 
 #ifdef CONFIG_X86
-	dst = kmap_atomic_prot(d, KM_USER0, prot);
+	dst = kmap_atomic_prot(d, prot);
 #else
 	if (pgprot_val(prot) != pgprot_val(PAGE_KERNEL))
 		dst = vmap(&d, 1, 0, prot);
@@ -183,7 +272,7 @@ static int ttm_copy_io_ttm_page(struct ttm_tt *ttm, void *src,
 	memcpy_fromio(dst, src, PAGE_SIZE);
 
 #ifdef CONFIG_X86
-	kunmap_atomic(dst, KM_USER0);
+	kunmap_atomic(dst);
 #else
 	if (pgprot_val(prot) != pgprot_val(PAGE_KERNEL))
 		vunmap(dst);
@@ -198,7 +287,7 @@ static int ttm_copy_ttm_io_page(struct ttm_tt *ttm, void *dst,
 				unsigned long page,
 				pgprot_t prot)
 {
-	struct page *s = ttm_tt_get_page(ttm, page);
+	struct page *s = ttm->pages[page];
 	void *src;
 
 	if (!s)
@@ -206,7 +295,7 @@ static int ttm_copy_ttm_io_page(struct ttm_tt *ttm, void *dst,
 
 	dst = (void *)((unsigned long)dst + (page << PAGE_SHIFT));
 #ifdef CONFIG_X86
-	src = kmap_atomic_prot(s, KM_USER0, prot);
+	src = kmap_atomic_prot(s, prot);
 #else
 	if (pgprot_val(prot) != pgprot_val(PAGE_KERNEL))
 		src = vmap(&s, 1, 0, prot);
@@ -219,7 +308,7 @@ static int ttm_copy_ttm_io_page(struct ttm_tt *ttm, void *dst,
 	memcpy_toio(dst, src, PAGE_SIZE);
 
 #ifdef CONFIG_X86
-	kunmap_atomic(src, KM_USER0);
+	kunmap_atomic(src);
 #else
 	if (pgprot_val(prot) != pgprot_val(PAGE_KERNEL))
 		vunmap(src);
@@ -231,7 +320,7 @@ static int ttm_copy_ttm_io_page(struct ttm_tt *ttm, void *dst,
 }
 
 int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
-		       bool evict, bool no_wait_reserve, bool no_wait_gpu,
+		       bool evict, bool no_wait_gpu,
 		       struct ttm_mem_reg *new_mem)
 {
 	struct ttm_bo_device *bdev = bo->bdev;
@@ -254,17 +343,36 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
 	if (ret)
 		goto out;
 
+	/*
+	 * Single TTM move. NOP.
+	 */
 	if (old_iomap == NULL && new_iomap == NULL)
 		goto out2;
-	if (old_iomap == NULL && ttm == NULL)
+
+	/*
+	 * Don't move nonexistent data. Clear destination instead.
+	 */
+	if (old_iomap == NULL &&
+	    (ttm == NULL || (ttm->state == tt_unpopulated &&
+			     !(ttm->page_flags & TTM_PAGE_FLAG_SWAPPED)))) {
+		memset_io(new_iomap, 0, new_mem->num_pages*PAGE_SIZE);
 		goto out2;
+	}
+
+	/*
+	 * TTM might be null for moves within the same region.
+	 */
+	if (ttm && ttm->state == tt_unpopulated) {
+		ret = ttm->bdev->driver->ttm_tt_populate(ttm);
+		if (ret)
+			goto out1;
+	}
 
 	add = 0;
 	dir = 1;
 
 	if ((old_mem->mem_type == new_mem->mem_type) &&
-	    (new_mem->mm_node->start <
-	     old_mem->mm_node->start + old_mem->mm_node->size)) {
+	    (new_mem->start < old_mem->start + old_mem->size)) {
 		dir = -1;
 		add = new_mem->num_pages - 1;
 	}
@@ -288,8 +396,7 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
 	}
 	mb();
 out2:
-	ttm_bo_free_old_node(bo);
-
+	old_copy = *old_mem;
 	*old_mem = *new_mem;
 	new_mem->mm_node = NULL;
 
@@ -300,9 +407,15 @@ out2:
 	}
 
 out1:
-	ttm_mem_reg_iounmap(bdev, new_mem, new_iomap);
+	ttm_mem_reg_iounmap(bdev, old_mem, new_iomap);
 out:
 	ttm_mem_reg_iounmap(bdev, &old_copy, old_iomap);
+
+	/*
+	 * On error, keep the mm node!
+	 */
+	if (!ret)
+		ttm_bo_mem_put(bo, &old_copy);
 	return ret;
 }
 EXPORT_SYMBOL(ttm_bo_move_memcpy);
@@ -333,8 +446,9 @@ static int ttm_buffer_object_transfer(struct ttm_buffer_object *bo,
 	struct ttm_buffer_object *fbo;
 	struct ttm_bo_device *bdev = bo->bdev;
 	struct ttm_bo_driver *driver = bdev->driver;
+	int ret;
 
-	fbo = kzalloc(sizeof(*fbo), GFP_KERNEL);
+	fbo = kmalloc(sizeof(*fbo), GFP_KERNEL);
 	if (!fbo)
 		return -ENOMEM;
 
@@ -345,18 +459,27 @@ static int ttm_buffer_object_transfer(struct ttm_buffer_object *bo,
 	 * TODO: Explicit member copy would probably be better here.
 	 */
 
-	spin_lock_init(&fbo->lock);
-	init_waitqueue_head(&fbo->event_queue);
 	INIT_LIST_HEAD(&fbo->ddestroy);
 	INIT_LIST_HEAD(&fbo->lru);
 	INIT_LIST_HEAD(&fbo->swap);
-	fbo->vm_node = NULL;
+	INIT_LIST_HEAD(&fbo->io_reserve_lru);
+	drm_vma_node_reset(&fbo->vma_node);
 	atomic_set(&fbo->cpu_writers, 0);
 
-	fbo->sync_obj = driver->sync_obj_ref(bo->sync_obj);
+	spin_lock(&bdev->fence_lock);
+	if (bo->sync_obj)
+		fbo->sync_obj = driver->sync_obj_ref(bo->sync_obj);
+	else
+		fbo->sync_obj = NULL;
+	spin_unlock(&bdev->fence_lock);
 	kref_init(&fbo->list_kref);
 	kref_init(&fbo->kref);
 	fbo->destroy = &ttm_transfered_destroy;
+	fbo->acc_size = 0;
+	fbo->resv = &fbo->ttm_resv;
+	reservation_object_init(fbo->resv);
+	ret = ww_mutex_trylock(&fbo->resv->lock);
+	WARN_ON(!ret);
 
 	*new_obj = fbo;
 	return 0;
@@ -383,7 +506,7 @@ pgprot_t ttm_io_prot(uint32_t caching_flags, pgprot_t tmp)
 	else
 		tmp = pgprot_noncached(tmp);
 #endif
-#if defined(__sparc__)
+#if defined(__sparc__) || defined(__mips__)
 	if (!(caching_flags & TTM_PL_FLAG_CACHED))
 		tmp = pgprot_noncached(tmp);
 #endif
@@ -420,10 +543,16 @@ static int ttm_bo_kmap_ttm(struct ttm_buffer_object *bo,
 {
 	struct ttm_mem_reg *mem = &bo->mem; pgprot_t prot;
 	struct ttm_tt *ttm = bo->ttm;
-	struct page *d;
-	int i;
+	int ret;
 
 	BUG_ON(!ttm);
+
+	if (ttm->state == tt_unpopulated) {
+		ret = ttm->bdev->driver->ttm_tt_populate(ttm);
+		if (ret)
+			return ret;
+	}
+
 	if (num_pages == 1 && (mem->placement & TTM_PL_FLAG_CACHED)) {
 		/*
 		 * We're mapping a single page, and the desired
@@ -431,18 +560,9 @@ static int ttm_bo_kmap_ttm(struct ttm_buffer_object *bo,
 		 */
 
 		map->bo_kmap_type = ttm_bo_map_kmap;
-		map->page = ttm_tt_get_page(ttm, start_page);
+		map->page = ttm->pages[start_page];
 		map->virtual = kmap(map->page);
 	} else {
-	    /*
-	     * Populate the part we're mapping;
-	     */
-		for (i = start_page; i < start_page + num_pages; ++i) {
-			d = ttm_tt_get_page(ttm, i);
-			if (!d)
-				return -ENOMEM;
-		}
-
 		/*
 		 * We need to use vmap to get the desired page protection
 		 * or to make the buffer object look contiguous.
@@ -461,6 +581,8 @@ int ttm_bo_kmap(struct ttm_buffer_object *bo,
 		unsigned long start_page, unsigned long num_pages,
 		struct ttm_bo_kmap_obj *map)
 {
+	struct ttm_mem_type_manager *man =
+		&bo->bdev->man[bo->mem.mem_type];
 	unsigned long offset, size;
 	int ret;
 
@@ -472,10 +594,12 @@ int ttm_bo_kmap(struct ttm_buffer_object *bo,
 	if (start_page > bo->num_pages)
 		return -EINVAL;
 #if 0
-	if (num_pages > 1 && !DRM_SUSER(DRM_CURPROC))
+	if (num_pages > 1 && !capable(CAP_SYS_ADMIN))
 		return -EPERM;
 #endif
+	(void) ttm_mem_io_lock(man, false);
 	ret = ttm_mem_io_reserve(bo->bdev, &bo->mem);
+	ttm_mem_io_unlock(man);
 	if (ret)
 		return ret;
 	if (!bo->mem.bus.is_iomem) {
@@ -490,12 +614,15 @@ EXPORT_SYMBOL(ttm_bo_kmap);
 
 void ttm_bo_kunmap(struct ttm_bo_kmap_obj *map)
 {
+	struct ttm_buffer_object *bo = map->bo;
+	struct ttm_mem_type_manager *man =
+		&bo->bdev->man[bo->mem.mem_type];
+
 	if (!map->virtual)
 		return;
 	switch (map->bo_kmap_type) {
 	case ttm_bo_map_iomap:
 		iounmap(map->virtual);
-		ttm_mem_io_free(map->bo->bdev, &map->bo->mem);
 		break;
 	case ttm_bo_map_vmap:
 		vunmap(map->virtual);
@@ -508,6 +635,9 @@ void ttm_bo_kunmap(struct ttm_bo_kmap_obj *map)
 	default:
 		BUG();
 	}
+	(void) ttm_mem_io_lock(man, false);
+	ttm_mem_io_free(map->bo->bdev, &map->bo->mem);
+	ttm_mem_io_unlock(man);
 	map->virtual = NULL;
 	map->page = NULL;
 }
@@ -515,8 +645,7 @@ EXPORT_SYMBOL(ttm_bo_kunmap);
 
 int ttm_bo_move_accel_cleanup(struct ttm_buffer_object *bo,
 			      void *sync_obj,
-			      void *sync_obj_arg,
-			      bool evict, bool no_wait_reserve,
+			      bool evict,
 			      bool no_wait_gpu,
 			      struct ttm_mem_reg *new_mem)
 {
@@ -528,28 +657,27 @@ int ttm_bo_move_accel_cleanup(struct ttm_buffer_object *bo,
 	struct ttm_buffer_object *ghost_obj;
 	void *tmp_obj = NULL;
 
-	spin_lock(&bo->lock);
+	spin_lock(&bdev->fence_lock);
 	if (bo->sync_obj) {
 		tmp_obj = bo->sync_obj;
 		bo->sync_obj = NULL;
 	}
 	bo->sync_obj = driver->sync_obj_ref(sync_obj);
-	bo->sync_obj_arg = sync_obj_arg;
 	if (evict) {
 		ret = ttm_bo_wait(bo, false, false, false);
-		spin_unlock(&bo->lock);
+		spin_unlock(&bdev->fence_lock);
 		if (tmp_obj)
 			driver->sync_obj_unref(&tmp_obj);
 		if (ret)
 			return ret;
 
-		ttm_bo_free_old_node(bo);
 		if ((man->flags & TTM_MEMTYPE_FLAG_FIXED) &&
 		    (bo->ttm != NULL)) {
 			ttm_tt_unbind(bo->ttm);
 			ttm_tt_destroy(bo->ttm);
 			bo->ttm = NULL;
 		}
+		ttm_bo_free_old_node(bo);
 	} else {
 		/**
 		 * This should help pipeline ordinary buffer moves.
@@ -560,7 +688,7 @@ int ttm_bo_move_accel_cleanup(struct ttm_buffer_object *bo,
 		 */
 
 		set_bit(TTM_BO_PRIV_FLAG_MOVING, &bo->priv_flags);
-		spin_unlock(&bo->lock);
+		spin_unlock(&bdev->fence_lock);
 		if (tmp_obj)
 			driver->sync_obj_unref(&tmp_obj);