4 files changed, 70 insertions, 78 deletions

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 4e9f1507ffd9..bface798ee59 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -1867,10 +1867,6 @@ static int mlx5_ib_alloc_ucontext(struct ib_ucontext *uctx,
 	if (err)
 		goto out_sys_pages;
 
-	if (ibdev->attrs.device_cap_flags & IB_DEVICE_ON_DEMAND_PAGING)
-		context->ibucontext.invalidate_range =
-			&mlx5_ib_invalidate_range;
-
 	if (req.flags & MLX5_IB_ALLOC_UCTX_DEVX) {
 		err = mlx5_ib_devx_create(dev, true);
 		if (err < 0)
@@ -1999,11 +1995,6 @@ static void mlx5_ib_dealloc_ucontext(struct ib_ucontext *ibcontext)
 	struct mlx5_ib_dev *dev = to_mdev(ibcontext->device);
 	struct mlx5_bfreg_info *bfregi;
 
-	/* All umem's must be destroyed before destroying the ucontext. */
-	mutex_lock(&ibcontext->per_mm_list_lock);
-	WARN_ON(!list_empty(&ibcontext->per_mm_list));
-	mutex_unlock(&ibcontext->per_mm_list_lock);
-
 	bfregi = &context->bfregi;
 	mlx5_ib_dealloc_transport_domain(dev, context->tdn, context->devx_uid);
 
diff --git a/drivers/infiniband/hw/mlx5/mem.c b/drivers/infiniband/hw/mlx5/mem.c
index a40e0abf2338..b5aece786b36 100644
--- a/drivers/infiniband/hw/mlx5/mem.c
+++ b/drivers/infiniband/hw/mlx5/mem.c
@@ -56,19 +56,6 @@ void mlx5_ib_cont_pages(struct ib_umem *umem, u64 addr,
 	struct scatterlist *sg;
 	int entry;
 
-	if (umem->is_odp) {
-		struct ib_umem_odp *odp = to_ib_umem_odp(umem);
-		unsigned int page_shift = odp->page_shift;
-
-		*ncont = ib_umem_odp_num_pages(odp);
-		*count = *ncont << (page_shift - PAGE_SHIFT);
-		*shift = page_shift;
-		if (order)
-			*order = ilog2(roundup_pow_of_two(*ncont));
-
-		return;
-	}
-
 	addr = addr >> PAGE_SHIFT;
 	tmp = (unsigned long)addr;
 	m = find_first_bit(&tmp, BITS_PER_LONG);
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 3401f5f6792e..1eff031ef048 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -784,19 +784,37 @@ static int mr_umem_get(struct mlx5_ib_dev *dev, struct ib_udata *udata,
 		       int *ncont, int *order)
 {
 	struct ib_umem *u;
-	int err;
 
 	*umem = NULL;
 
-	u = ib_umem_get(udata, start, length, access_flags, 0);
-	err = PTR_ERR_OR_ZERO(u);
-	if (err) {
-		mlx5_ib_dbg(dev, "umem get failed (%d)\n", err);
-		return err;
+	if (access_flags & IB_ACCESS_ON_DEMAND) {
+		struct ib_umem_odp *odp;
+
+		odp = ib_umem_odp_get(udata, start, length, access_flags);
+		if (IS_ERR(odp)) {
+			mlx5_ib_dbg(dev, "umem get failed (%ld)\n",
+				    PTR_ERR(odp));
+			return PTR_ERR(odp);
+		}
+
+		u = &odp->umem;
+
+		*page_shift = odp->page_shift;
+		*ncont = ib_umem_odp_num_pages(odp);
+		*npages = *ncont << (*page_shift - PAGE_SHIFT);
+		if (order)
+			*order = ilog2(roundup_pow_of_two(*ncont));
+	} else {
+		u = ib_umem_get(udata, start, length, access_flags, 0);
+		if (IS_ERR(u)) {
+			mlx5_ib_dbg(dev, "umem get failed (%ld)\n", PTR_ERR(u));
+			return PTR_ERR(u);
+		}
+
+		mlx5_ib_cont_pages(u, start, MLX5_MKEY_PAGE_SHIFT_MASK, npages,
+				   page_shift, ncont, order);
 	}
 
-	mlx5_ib_cont_pages(u, start, MLX5_MKEY_PAGE_SHIFT_MASK, npages,
-			   page_shift, ncont, order);
 	if (!*npages) {
 		mlx5_ib_warn(dev, "avoid zero region\n");
 		ib_umem_release(u);
@@ -1599,7 +1617,7 @@ static void dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 		/* Wait for all running page-fault handlers to finish. */
 		synchronize_srcu(&dev->mr_srcu);
 		/* Destroy all page mappings */
-		if (umem_odp->page_list)
+		if (!umem_odp->is_implicit_odp)
 			mlx5_ib_invalidate_range(umem_odp,
 						 ib_umem_start(umem_odp),
 						 ib_umem_end(umem_odp));
@@ -1610,7 +1628,7 @@ static void dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 		 * so that there will not be any invalidations in
 		 * flight, looking at the *mr struct.
 		 */
-		ib_umem_release(umem);
+		ib_umem_odp_release(umem_odp);
 		atomic_sub(npages, &dev->mdev->priv.reg_pages);
 
 		/* Avoid double-freeing the umem. */
diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
index 0a59912a4cef..dd26e7acb37e 100644
--- a/drivers/infiniband/hw/mlx5/odp.c
+++ b/drivers/infiniband/hw/mlx5/odp.c
@@ -184,7 +184,7 @@ void mlx5_odp_populate_klm(struct mlx5_klm *pklm, size_t offset,
 	for (i = 0; i < nentries; i++, pklm++) {
 		pklm->bcount = cpu_to_be32(MLX5_IMR_MTT_SIZE);
 		va = (offset + i) * MLX5_IMR_MTT_SIZE;
-		if (odp && odp->umem.address == va) {
+		if (odp && ib_umem_start(odp) == va) {
 			struct mlx5_ib_mr *mtt = odp->private;
 
 			pklm->key = cpu_to_be32(mtt->ibmr.lkey);
@@ -206,7 +206,7 @@ static void mr_leaf_free_action(struct work_struct *work)
 	mr->parent = NULL;
 	synchronize_srcu(&mr->dev->mr_srcu);
 
-	ib_umem_release(&odp->umem);
+	ib_umem_odp_release(odp);
 	if (imr->live)
 		mlx5_ib_update_xlt(imr, idx, 1, 0,
 				   MLX5_IB_UPD_XLT_INDIRECT |
@@ -386,7 +386,7 @@ static void mlx5_ib_page_fault_resume(struct mlx5_ib_dev *dev,
 }
 
 static struct mlx5_ib_mr *implicit_mr_alloc(struct ib_pd *pd,
-					    struct ib_umem *umem,
+					    struct ib_umem_odp *umem_odp,
 					    bool ksm, int access_flags)
 {
 	struct mlx5_ib_dev *dev = to_mdev(pd->device);
@@ -404,7 +404,7 @@ static struct mlx5_ib_mr *implicit_mr_alloc(struct ib_pd *pd,
 	mr->dev = dev;
 	mr->access_flags = access_flags;
 	mr->mmkey.iova = 0;
-	mr->umem = umem;
+	mr->umem = &umem_odp->umem;
 
 	if (ksm) {
 		err = mlx5_ib_update_xlt(mr, 0,
@@ -464,18 +464,17 @@ next_mr:
 		if (nentries)
 			nentries++;
 	} else {
-		odp = ib_alloc_odp_umem(odp_mr, addr,
-					MLX5_IMR_MTT_SIZE);
+		odp = ib_umem_odp_alloc_child(odp_mr, addr, MLX5_IMR_MTT_SIZE);
 		if (IS_ERR(odp)) {
 			mutex_unlock(&odp_mr->umem_mutex);
 			return ERR_CAST(odp);
 		}
 
-		mtt = implicit_mr_alloc(mr->ibmr.pd, &odp->umem, 0,
+		mtt = implicit_mr_alloc(mr->ibmr.pd, odp, 0,
 					mr->access_flags);
 		if (IS_ERR(mtt)) {
 			mutex_unlock(&odp_mr->umem_mutex);
-			ib_umem_release(&odp->umem);
+			ib_umem_odp_release(odp);
 			return ERR_CAST(mtt);
 		}
 
@@ -497,7 +496,7 @@ next_mr:
 	addr += MLX5_IMR_MTT_SIZE;
 	if (unlikely(addr < io_virt + bcnt)) {
 		odp = odp_next(odp);
-		if (odp && odp->umem.address != addr)
+		if (odp && ib_umem_start(odp) != addr)
 			odp = NULL;
 		goto next_mr;
 	}
@@ -521,19 +520,19 @@ struct mlx5_ib_mr *mlx5_ib_alloc_implicit_mr(struct mlx5_ib_pd *pd,
 					     int access_flags)
 {
 	struct mlx5_ib_mr *imr;
-	struct ib_umem *umem;
+	struct ib_umem_odp *umem_odp;
 
-	umem = ib_umem_get(udata, 0, 0, access_flags, 0);
-	if (IS_ERR(umem))
-		return ERR_CAST(umem);
+	umem_odp = ib_umem_odp_alloc_implicit(udata, access_flags);
+	if (IS_ERR(umem_odp))
+		return ERR_CAST(umem_odp);
 
-	imr = implicit_mr_alloc(&pd->ibpd, umem, 1, access_flags);
+	imr = implicit_mr_alloc(&pd->ibpd, umem_odp, 1, access_flags);
 	if (IS_ERR(imr)) {
-		ib_umem_release(umem);
+		ib_umem_odp_release(umem_odp);
 		return ERR_CAST(imr);
 	}
 
-	imr->umem = umem;
+	imr->umem = &umem_odp->umem;
 	init_waitqueue_head(&imr->q_leaf_free);
 	atomic_set(&imr->num_leaf_free, 0);
 	atomic_set(&imr->num_pending_prefetch, 0);
@@ -541,34 +540,31 @@ struct mlx5_ib_mr *mlx5_ib_alloc_implicit_mr(struct mlx5_ib_pd *pd,
 	return imr;
 }
 
-static int mr_leaf_free(struct ib_umem_odp *umem_odp, u64 start, u64 end,
-			void *cookie)
+void mlx5_ib_free_implicit_mr(struct mlx5_ib_mr *imr)
 {
-	struct mlx5_ib_mr *mr = umem_odp->private, *imr = cookie;
-
-	if (mr->parent != imr)
-		return 0;
-
-	ib_umem_odp_unmap_dma_pages(umem_odp, ib_umem_start(umem_odp),
-				    ib_umem_end(umem_odp));
+	struct ib_ucontext_per_mm *per_mm = mr_to_per_mm(imr);
+	struct rb_node *node;
 
-	if (umem_odp->dying)
-		return 0;
+	down_read(&per_mm->umem_rwsem);
+	for (node = rb_first_cached(&per_mm->umem_tree); node;
+	     node = rb_next(node)) {
+		struct ib_umem_odp *umem_odp =
+			rb_entry(node, struct ib_umem_odp, interval_tree.rb);
+		struct mlx5_ib_mr *mr = umem_odp->private;
 
-	WRITE_ONCE(umem_odp->dying, 1);
-	atomic_inc(&imr->num_leaf_free);
-	schedule_work(&umem_odp->work);
+		if (mr->parent != imr)
+			continue;
 
-	return 0;
-}
+		ib_umem_odp_unmap_dma_pages(umem_odp, ib_umem_start(umem_odp),
+					    ib_umem_end(umem_odp));
 
-void mlx5_ib_free_implicit_mr(struct mlx5_ib_mr *imr)
-{
-	struct ib_ucontext_per_mm *per_mm = mr_to_per_mm(imr);
+		if (umem_odp->dying)
+			continue;
 
-	down_read(&per_mm->umem_rwsem);
-	rbt_ib_umem_for_each_in_range(&per_mm->umem_tree, 0, ULLONG_MAX,
-				      mr_leaf_free, true, imr);
+		WRITE_ONCE(umem_odp->dying, 1);
+		atomic_inc(&imr->num_leaf_free);
+		schedule_work(&umem_odp->work);
+	}
 	up_read(&per_mm->umem_rwsem);
 
 	wait_event(imr->q_leaf_free, !atomic_read(&imr->num_leaf_free));
@@ -589,7 +585,7 @@ static int pagefault_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr,
 	struct ib_umem_odp *odp;
 	size_t size;
 
-	if (!odp_mr->page_list) {
+	if (odp_mr->is_implicit_odp) {
 		odp = implicit_mr_get_data(mr, io_virt, bcnt);
 
 		if (IS_ERR(odp))
@@ -607,7 +603,7 @@ next_mr:
 	start_idx = (io_virt - (mr->mmkey.iova & page_mask)) >> page_shift;
 	access_mask = ODP_READ_ALLOWED_BIT;
 
-	if (prefetch && !downgrade && !mr->umem->writable) {
+	if (prefetch && !downgrade && !odp->umem.writable) {
 		/* prefetch with write-access must
 		 * be supported by the MR
 		 */
@@ -615,7 +611,7 @@ next_mr:
 		goto out;
 	}
 
-	if (mr->umem->writable && !downgrade)
+	if (odp->umem.writable && !downgrade)
 		access_mask |= ODP_WRITE_ALLOWED_BIT;
 
 	current_seq = READ_ONCE(odp->notifiers_seq);
@@ -625,8 +621,8 @@ next_mr:
 	 */
 	smp_rmb();
 
-	ret = ib_umem_odp_map_dma_pages(to_ib_umem_odp(mr->umem), io_virt, size,
-					access_mask, current_seq);
+	ret = ib_umem_odp_map_dma_pages(odp, io_virt, size, access_mask,
+					current_seq);
 
 	if (ret < 0)
 		goto out;
@@ -634,8 +630,7 @@ next_mr:
 	np = ret;
 
 	mutex_lock(&odp->umem_mutex);
-	if (!ib_umem_mmu_notifier_retry(to_ib_umem_odp(mr->umem),
-					current_seq)) {
+	if (!ib_umem_mmu_notifier_retry(odp, current_seq)) {
 		/*
 		 * No need to check whether the MTTs really belong to
 		 * this MR, since ib_umem_odp_map_dma_pages already
@@ -668,7 +663,7 @@ next_mr:
 
 		io_virt += size;
 		next = odp_next(odp);
-		if (unlikely(!next || next->umem.address != io_virt)) {
+		if (unlikely(!next || ib_umem_start(next) != io_virt)) {
 			mlx5_ib_dbg(dev, "next implicit leaf removed at 0x%llx. got %p\n",
 				    io_virt, next);
 			return -EAGAIN;
@@ -1618,6 +1613,7 @@ void mlx5_odp_init_mr_cache_entry(struct mlx5_cache_ent *ent)
 
 static const struct ib_device_ops mlx5_ib_dev_odp_ops = {
 	.advise_mr = mlx5_ib_advise_mr,
+	.invalidate_range = mlx5_ib_invalidate_range,
 };
 
 int mlx5_ib_odp_init_one(struct mlx5_ib_dev *dev)