1 files changed, 838 insertions, 516 deletions

diff --git a/io_uring/rsrc.c b/io_uring/rsrc.c
index 6f3b6de230bd..f2b31fb68992 100644
--- a/io_uring/rsrc.c
+++ b/io_uring/rsrc.c
@@ -9,11 +9,11 @@
 #include <linux/hugetlb.h>
 #include <linux/compat.h>
 #include <linux/io_uring.h>
+#include <linux/io_uring/cmd.h>
 
 #include <uapi/linux/io_uring.h>
 
 #include "io_uring.h"
-#include "alloc_cache.h"
 #include "openclose.h"
 #include "rsrc.h"
 #include "memmap.h"
@@ -26,20 +26,14 @@ struct io_rsrc_update {
 	u32				offset;
 };
 
-static void io_rsrc_buf_put(struct io_ring_ctx *ctx, struct io_rsrc_put *prsrc);
-static int io_sqe_buffer_register(struct io_ring_ctx *ctx, struct iovec *iov,
-				  struct io_mapped_ubuf **pimu,
-				  struct page **last_hpage);
+static struct io_rsrc_node *io_sqe_buffer_register(struct io_ring_ctx *ctx,
+			struct iovec *iov, struct page **last_hpage);
 
 /* only define max */
 #define IORING_MAX_FIXED_FILES	(1U << 20)
 #define IORING_MAX_REG_BUFFERS	(1U << 14)
 
-static const struct io_mapped_ubuf dummy_ubuf = {
-	/* set invalid range, so io_import_fixed() fails meeting it */
-	.ubuf = -1UL,
-	.len = UINT_MAX,
-};
+#define IO_CACHED_BVECS_SEGS	32
 
 int __io_account_mem(struct user_struct *user, unsigned long nr_pages)
 {
@@ -86,10 +80,21 @@ static int io_account_mem(struct io_ring_ctx *ctx, unsigned long nr_pages)
 	return 0;
 }
 
-static int io_buffer_validate(struct iovec *iov)
+int io_validate_user_buf_range(u64 uaddr, u64 ulen)
 {
-	unsigned long tmp, acct_len = iov->iov_len + (PAGE_SIZE - 1);
+	unsigned long tmp, base = (unsigned long)uaddr;
+	unsigned long acct_len = (unsigned long)PAGE_ALIGN(ulen);
 
+	/* arbitrary limit, but we need something */
+	if (ulen > SZ_1G || !ulen)
+		return -EFAULT;
+	if (check_add_overflow(base, acct_len, &tmp))
+		return -EOVERFLOW;
+	return 0;
+}
+
+static int io_buffer_validate(struct iovec *iov)
+{
 	/*
 	 * Don't impose further limits on the size and buffer
 	 * constraints here, we'll -EINVAL later when IO is
@@ -97,235 +102,120 @@ static int io_buffer_validate(struct iovec *iov)
 	 */
 	if (!iov->iov_base)
 		return iov->iov_len ? -EFAULT : 0;
-	if (!iov->iov_len)
-		return -EFAULT;
-
-	/* arbitrary limit, but we need something */
-	if (iov->iov_len > SZ_1G)
-		return -EFAULT;
-
-	if (check_add_overflow((unsigned long)iov->iov_base, acct_len, &tmp))
-		return -EOVERFLOW;
 
-	return 0;
+	return io_validate_user_buf_range((unsigned long)iov->iov_base,
+					  iov->iov_len);
 }
 
-static void io_buffer_unmap(struct io_ring_ctx *ctx, struct io_mapped_ubuf **slot)
+static void io_release_ubuf(void *priv)
 {
-	struct io_mapped_ubuf *imu = *slot;
+	struct io_mapped_ubuf *imu = priv;
 	unsigned int i;
 
-	*slot = NULL;
-	if (imu != &dummy_ubuf) {
-		if (!refcount_dec_and_test(&imu->refs))
-			return;
-		for (i = 0; i < imu->nr_bvecs; i++)
-			unpin_user_page(imu->bvec[i].bv_page);
-		if (imu->acct_pages)
-			io_unaccount_mem(ctx, imu->acct_pages);
-		kvfree(imu);
-	}
-}
+	for (i = 0; i < imu->nr_bvecs; i++) {
+		struct folio *folio = page_folio(imu->bvec[i].bv_page);
 
-static void io_rsrc_put_work(struct io_rsrc_node *node)
-{
-	struct io_rsrc_put *prsrc = &node->item;
-
-	if (prsrc->tag)
-		io_post_aux_cqe(node->ctx, prsrc->tag, 0, 0);
-
-	switch (node->type) {
-	case IORING_RSRC_FILE:
-		fput(prsrc->file);
-		break;
-	case IORING_RSRC_BUFFER:
-		io_rsrc_buf_put(node->ctx, prsrc);
-		break;
-	default:
-		WARN_ON_ONCE(1);
-		break;
+		unpin_user_folio(folio, 1);
 	}
 }
 
-void io_rsrc_node_destroy(struct io_ring_ctx *ctx, struct io_rsrc_node *node)
+static struct io_mapped_ubuf *io_alloc_imu(struct io_ring_ctx *ctx,
+					   int nr_bvecs)
 {
-	if (!io_alloc_cache_put(&ctx->rsrc_node_cache, node))
-		kfree(node);
+	if (nr_bvecs <= IO_CACHED_BVECS_SEGS)
+		return io_cache_alloc(&ctx->imu_cache, GFP_KERNEL);
+	return kvmalloc(struct_size_t(struct io_mapped_ubuf, bvec, nr_bvecs),
+			GFP_KERNEL);
 }
 
-void io_rsrc_node_ref_zero(struct io_rsrc_node *node)
-	__must_hold(&node->ctx->uring_lock)
+static void io_free_imu(struct io_ring_ctx *ctx, struct io_mapped_ubuf *imu)
 {
-	struct io_ring_ctx *ctx = node->ctx;
-
-	while (!list_empty(&ctx->rsrc_ref_list)) {
-		node = list_first_entry(&ctx->rsrc_ref_list,
-					    struct io_rsrc_node, node);
-		/* recycle ref nodes in order */
-		if (node->refs)
-			break;
-		list_del(&node->node);
-
-		if (likely(!node->empty))
-			io_rsrc_put_work(node);
-		io_rsrc_node_destroy(ctx, node);
-	}
-	if (list_empty(&ctx->rsrc_ref_list) && unlikely(ctx->rsrc_quiesce))
-		wake_up_all(&ctx->rsrc_quiesce_wq);
+	if (imu->nr_bvecs <= IO_CACHED_BVECS_SEGS)
+		io_cache_free(&ctx->imu_cache, imu);
+	else
+		kvfree(imu);
 }
 
-struct io_rsrc_node *io_rsrc_node_alloc(struct io_ring_ctx *ctx)
+static void io_buffer_unmap(struct io_ring_ctx *ctx, struct io_mapped_ubuf *imu)
 {
-	struct io_rsrc_node *ref_node;
-
-	ref_node = io_alloc_cache_get(&ctx->rsrc_node_cache);
-	if (!ref_node) {
-		ref_node = kzalloc(sizeof(*ref_node), GFP_KERNEL);
-		if (!ref_node)
-			return NULL;
-	}
+	if (!refcount_dec_and_test(&imu->refs))
+		return;
 
-	ref_node->ctx = ctx;
-	ref_node->empty = 0;
-	ref_node->refs = 1;
-	return ref_node;
+	if (imu->acct_pages)
+		io_unaccount_mem(ctx, imu->acct_pages);
+	imu->release(imu->priv);
+	io_free_imu(ctx, imu);
 }
 
-__cold static int io_rsrc_ref_quiesce(struct io_rsrc_data *data,
-				      struct io_ring_ctx *ctx)
+struct io_rsrc_node *io_rsrc_node_alloc(struct io_ring_ctx *ctx, int type)
 {
-	struct io_rsrc_node *backup;
-	DEFINE_WAIT(we);
-	int ret;
-
-	/* As We may drop ->uring_lock, other task may have started quiesce */
-	if (data->quiesce)
-		return -ENXIO;
-
-	backup = io_rsrc_node_alloc(ctx);
-	if (!backup)
-		return -ENOMEM;
-	ctx->rsrc_node->empty = true;
-	ctx->rsrc_node->type = -1;
-	list_add_tail(&ctx->rsrc_node->node, &ctx->rsrc_ref_list);
-	io_put_rsrc_node(ctx, ctx->rsrc_node);
-	ctx->rsrc_node = backup;
-
-	if (list_empty(&ctx->rsrc_ref_list))
-		return 0;
-
-	if (ctx->flags & IORING_SETUP_DEFER_TASKRUN) {
-		atomic_set(&ctx->cq_wait_nr, 1);
-		smp_mb();
+	struct io_rsrc_node *node;
+
+	node = io_cache_alloc(&ctx->node_cache, GFP_KERNEL);
+	if (node) {
+		node->type = type;
+		node->refs = 1;
+		node->tag = 0;
+		node->file_ptr = 0;
 	}
-
-	ctx->rsrc_quiesce++;
-	data->quiesce = true;
-	do {
-		prepare_to_wait(&ctx->rsrc_quiesce_wq, &we, TASK_INTERRUPTIBLE);
-		mutex_unlock(&ctx->uring_lock);
-
-		ret = io_run_task_work_sig(ctx);
-		if (ret < 0) {
-			finish_wait(&ctx->rsrc_quiesce_wq, &we);
-			mutex_lock(&ctx->uring_lock);
-			if (list_empty(&ctx->rsrc_ref_list))
-				ret = 0;
-			break;
-		}
-
-		schedule();
-		mutex_lock(&ctx->uring_lock);
-		ret = 0;
-	} while (!list_empty(&ctx->rsrc_ref_list));
-
-	finish_wait(&ctx->rsrc_quiesce_wq, &we);
-	data->quiesce = false;
-	ctx->rsrc_quiesce--;
-
-	if (ctx->flags & IORING_SETUP_DEFER_TASKRUN) {
-		atomic_set(&ctx->cq_wait_nr, 0);
-		smp_mb();
-	}
-	return ret;
+	return node;
 }
 
-static void io_free_page_table(void **table, size_t size)
+bool io_rsrc_cache_init(struct io_ring_ctx *ctx)
 {
-	unsigned i, nr_tables = DIV_ROUND_UP(size, PAGE_SIZE);
-
-	for (i = 0; i < nr_tables; i++)
-		kfree(table[i]);
-	kfree(table);
+	const int imu_cache_size = struct_size_t(struct io_mapped_ubuf, bvec,
+						 IO_CACHED_BVECS_SEGS);
+	const int node_size = sizeof(struct io_rsrc_node);
+	bool ret;
+
+	ret = io_alloc_cache_init(&ctx->node_cache, IO_ALLOC_CACHE_MAX,
+				  node_size, 0);
+	ret |= io_alloc_cache_init(&ctx->imu_cache, IO_ALLOC_CACHE_MAX,
+				   imu_cache_size, 0);
+	return ret;
 }
 
-static void io_rsrc_data_free(struct io_rsrc_data *data)
+void io_rsrc_cache_free(struct io_ring_ctx *ctx)
 {
-	size_t size = data->nr * sizeof(data->tags[0][0]);
-
-	if (data->tags)
-		io_free_page_table((void **)data->tags, size);
-	kfree(data);
+	io_alloc_cache_free(&ctx->node_cache, kfree);
+	io_alloc_cache_free(&ctx->imu_cache, kfree);
 }
 
-static __cold void **io_alloc_page_table(size_t size)
+static void io_clear_table_tags(struct io_rsrc_data *data)
 {
-	unsigned i, nr_tables = DIV_ROUND_UP(size, PAGE_SIZE);
-	size_t init_size = size;
-	void **table;
-
-	table = kcalloc(nr_tables, sizeof(*table), GFP_KERNEL_ACCOUNT);
-	if (!table)
-		return NULL;
+	int i;
 
-	for (i = 0; i < nr_tables; i++) {
-		unsigned int this_size = min_t(size_t, size, PAGE_SIZE);
+	for (i = 0; i < data->nr; i++) {
+		struct io_rsrc_node *node = data->nodes[i];
 
-		table[i] = kzalloc(this_size, GFP_KERNEL_ACCOUNT);
-		if (!table[i]) {
-			io_free_page_table(table, init_size);
-			return NULL;
-		}
-		size -= this_size;
+		if (node)
+			node->tag = 0;
 	}
-	return table;
 }
 
-__cold static int io_rsrc_data_alloc(struct io_ring_ctx *ctx, int type,
-				     u64 __user *utags,
-				     unsigned nr, struct io_rsrc_data **pdata)
+__cold void io_rsrc_data_free(struct io_ring_ctx *ctx,
+			      struct io_rsrc_data *data)
 {
-	struct io_rsrc_data *data;
-	int ret = 0;
-	unsigned i;
-
-	data = kzalloc(sizeof(*data), GFP_KERNEL);
-	if (!data)
-		return -ENOMEM;
-	data->tags = (u64 **)io_alloc_page_table(nr * sizeof(data->tags[0][0]));
-	if (!data->tags) {
-		kfree(data);
-		return -ENOMEM;
+	if (!data->nr)
+		return;
+	while (data->nr--) {
+		if (data->nodes[data->nr])
+			io_put_rsrc_node(ctx, data->nodes[data->nr]);
 	}
+	kvfree(data->nodes);
+	data->nodes = NULL;
+	data->nr = 0;
+}
 
-	data->nr = nr;
-	data->ctx = ctx;
-	data->rsrc_type = type;
-	if (utags) {
-		ret = -EFAULT;
-		for (i = 0; i < nr; i++) {
-			u64 *tag_slot = io_get_tag_slot(data, i);
-
-			if (copy_from_user(tag_slot, &utags[i],
-					   sizeof(*tag_slot)))
-				goto fail;
-		}
+__cold int io_rsrc_data_alloc(struct io_rsrc_data *data, unsigned nr)
+{
+	data->nodes = kvmalloc_array(nr, sizeof(struct io_rsrc_node *),
+					GFP_KERNEL_ACCOUNT | __GFP_ZERO);
+	if (data->nodes) {
+		data->nr = nr;
+		return 0;
 	}
-	*pdata = data;
-	return 0;
-fail:
-	io_rsrc_data_free(data);
-	return ret;
+	return -ENOMEM;
 }
 
 static int __io_sqe_files_update(struct io_ring_ctx *ctx,
@@ -334,14 +224,12 @@ static int __io_sqe_files_update(struct io_ring_ctx *ctx,
 {
 	u64 __user *tags = u64_to_user_ptr(up->tags);
 	__s32 __user *fds = u64_to_user_ptr(up->data);
-	struct io_rsrc_data *data = ctx->file_data;
-	struct io_fixed_file *file_slot;
 	int fd, i, err = 0;
 	unsigned int done;
 
-	if (!ctx->file_data)
+	if (!ctx->file_table.data.nr)
 		return -ENXIO;
-	if (up->offset + nr_args > ctx->nr_user_files)
+	if (up->offset + nr_args > ctx->file_table.data.nr)
 		return -EINVAL;
 
 	for (done = 0; done < nr_args; done++) {
@@ -359,19 +247,13 @@ static int __io_sqe_files_update(struct io_ring_ctx *ctx,
 		if (fd == IORING_REGISTER_FILES_SKIP)
 			continue;
 
-		i = array_index_nospec(up->offset + done, ctx->nr_user_files);
-		file_slot = io_fixed_file_slot(&ctx->file_table, i);
-
-		if (file_slot->file_ptr) {
-			err = io_queue_rsrc_removal(data, i,
-						    io_slot_file(file_slot));
-			if (err)
-				break;
-			file_slot->file_ptr = 0;
+		i = up->offset + done;
+		if (io_reset_rsrc_node(ctx, &ctx->file_table.data, i))
 			io_file_bitmap_clear(&ctx->file_table, i);
-		}
+
 		if (fd != -1) {
 			struct file *file = fget(fd);
+			struct io_rsrc_node *node;
 
 			if (!file) {
 				err = -EBADF;
@@ -385,8 +267,16 @@ static int __io_sqe_files_update(struct io_ring_ctx *ctx,
 				err = -EBADF;
 				break;
 			}
-			*io_get_tag_slot(data, i) = tag;
-			io_fixed_file_set(file_slot, file);
+			node = io_rsrc_node_alloc(ctx, IORING_RSRC_FILE);
+			if (!node) {
+				err = -ENOMEM;
+				fput(file);
+				break;
+			}
+			ctx->file_table.data.nodes[i] = node;
+			if (tag)
+				node->tag = tag;
+			io_fixed_file_set(node, file);
 			io_file_bitmap_set(&ctx->file_table, i);
 		}
 	}
@@ -405,13 +295,13 @@ static int __io_sqe_buffers_update(struct io_ring_ctx *ctx,
 	__u32 done;
 	int i, err;
 
-	if (!ctx->buf_data)
+	if (!ctx->buf_table.nr)
 		return -ENXIO;
-	if (up->offset + nr_args > ctx->nr_user_bufs)
+	if (up->offset + nr_args > ctx->buf_table.nr)
 		return -EINVAL;
 
 	for (done = 0; done < nr_args; done++) {
-		struct io_mapped_ubuf *imu;
+		struct io_rsrc_node *node;
 		u64 tag = 0;
 
 		uvec = u64_to_user_ptr(user_data);
@@ -427,27 +317,21 @@ static int __io_sqe_buffers_update(struct io_ring_ctx *ctx,
 		err = io_buffer_validate(iov);
 		if (err)
 			break;
-		if (!iov->iov_base && tag) {
-			err = -EINVAL;
+		node = io_sqe_buffer_register(ctx, iov, &last_hpage);
+		if (IS_ERR(node)) {
+			err = PTR_ERR(node);
 			break;
 		}
-		err = io_sqe_buffer_register(ctx, iov, &imu, &last_hpage);
-		if (err)
-			break;
-
-		i = array_index_nospec(up->offset + done, ctx->nr_user_bufs);
-		if (ctx->user_bufs[i] != &dummy_ubuf) {
-			err = io_queue_rsrc_removal(ctx->buf_data, i,
-						    ctx->user_bufs[i]);
-			if (unlikely(err)) {
-				io_buffer_unmap(ctx, &imu);
+		if (tag) {
+			if (!node) {
+				err = -EINVAL;
 				break;
 			}
-			ctx->user_bufs[i] = (struct io_mapped_ubuf *)&dummy_ubuf;
+			node->tag = tag;
 		}
-
-		ctx->user_bufs[i] = imu;
-		*io_get_tag_slot(ctx->buf_data, i) = tag;
+		i = array_index_nospec(up->offset + done, ctx->buf_table.nr);
+		io_reset_rsrc_node(ctx, &ctx->buf_table, i);
+		ctx->buf_table.nodes[i] = node;
 		if (ctx->compat)
 			user_data += sizeof(struct compat_iovec);
 		else
@@ -563,7 +447,7 @@ static int io_files_update_with_index_alloc(struct io_kiocb *req,
 	struct file *file;
 	int ret, fd;
 
-	if (!req->ctx->file_data)
+	if (!req->ctx->file_table.data.nr)
 		return -ENXIO;
 
 	for (done = 0; done < up->nr_args; done++) {
@@ -619,68 +503,37 @@ int io_files_update(struct io_kiocb *req, unsigned int issue_flags)
 	if (ret < 0)
 		req_set_fail(req);
 	io_req_set_res(req, ret, 0);
-	return IOU_OK;
+	return IOU_COMPLETE;
 }
 
-int io_queue_rsrc_removal(struct io_rsrc_data *data, unsigned idx, void *rsrc)
+void io_free_rsrc_node(struct io_ring_ctx *ctx, struct io_rsrc_node *node)
 {
-	struct io_ring_ctx *ctx = data->ctx;
-	struct io_rsrc_node *node = ctx->rsrc_node;
-	u64 *tag_slot = io_get_tag_slot(data, idx);
-
-	ctx->rsrc_node = io_rsrc_node_alloc(ctx);
-	if (unlikely(!ctx->rsrc_node)) {
-		ctx->rsrc_node = node;
-		return -ENOMEM;
-	}
-
-	node->item.rsrc = rsrc;
-	node->type = data->rsrc_type;
-	node->item.tag = *tag_slot;
-	*tag_slot = 0;
-	list_add_tail(&node->node, &ctx->rsrc_ref_list);
-	io_put_rsrc_node(ctx, node);
-	return 0;
-}
+	if (node->tag)
+		io_post_aux_cqe(ctx, node->tag, 0, 0);
 
-void __io_sqe_files_unregister(struct io_ring_ctx *ctx)
-{
-	int i;
-
-	for (i = 0; i < ctx->nr_user_files; i++) {
-		struct file *file = io_file_from_index(&ctx->file_table, i);
-
-		if (!file)
-			continue;
-		io_file_bitmap_clear(&ctx->file_table, i);
-		fput(file);
+	switch (node->type) {
+	case IORING_RSRC_FILE:
+		fput(io_slot_file(node));
+		break;
+	case IORING_RSRC_BUFFER:
+		io_buffer_unmap(ctx, node->buf);
+		break;
+	default:
+		WARN_ON_ONCE(1);
+		break;
 	}
 
-	io_free_file_tables(&ctx->file_table);
-	io_file_table_set_alloc_range(ctx, 0, 0);
-	io_rsrc_data_free(ctx->file_data);
-	ctx->file_data = NULL;
-	ctx->nr_user_files = 0;
+	io_cache_free(&ctx->node_cache, node);
 }
 
 int io_sqe_files_unregister(struct io_ring_ctx *ctx)
 {
-	unsigned nr = ctx->nr_user_files;
-	int ret;
-
-	if (!ctx->file_data)
+	if (!ctx->file_table.data.nr)
 		return -ENXIO;
 
-	/*
-	 * Quiesce may unlock ->uring_lock, and while it's not held
-	 * prevent new requests using the table.
-	 */
-	ctx->nr_user_files = 0;
-	ret = io_rsrc_ref_quiesce(ctx->file_data, ctx);
-	ctx->nr_user_files = nr;
-	if (!ret)
-		__io_sqe_files_unregister(ctx);
-	return ret;
+	io_free_file_tables(ctx, &ctx->file_table);
+	io_file_table_set_alloc_range(ctx, 0, 0);
+	return 0;
 }
 
 int io_sqe_files_register(struct io_ring_ctx *ctx, void __user *arg,
@@ -691,7 +544,7 @@ int io_sqe_files_register(struct io_ring_ctx *ctx, void __user *arg,
 	int fd, ret;
 	unsigned i;
 
-	if (ctx->file_data)
+	if (ctx->file_table.data.nr)
 		return -EBUSY;
 	if (!nr_args)
 		return -EINVAL;
@@ -699,28 +552,22 @@ int io_sqe_files_register(struct io_ring_ctx *ctx, void __user *arg,
 		return -EMFILE;
 	if (nr_args > rlimit(RLIMIT_NOFILE))
 		return -EMFILE;
-	ret = io_rsrc_data_alloc(ctx, IORING_RSRC_FILE, tags, nr_args,
-				 &ctx->file_data);
-	if (ret)
-		return ret;
-
-	if (!io_alloc_file_tables(&ctx->file_table, nr_args)) {
-		io_rsrc_data_free(ctx->file_data);
-		ctx->file_data = NULL;
+	if (!io_alloc_file_tables(ctx, &ctx->file_table, nr_args))
 		return -ENOMEM;
-	}
 
-	for (i = 0; i < nr_args; i++, ctx->nr_user_files++) {
-		struct io_fixed_file *file_slot;
+	for (i = 0; i < nr_args; i++) {
+		struct io_rsrc_node *node;
+		u64 tag = 0;
 
-		if (fds && copy_from_user(&fd, &fds[i], sizeof(fd))) {
-			ret = -EFAULT;
+		ret = -EFAULT;
+		if (tags && copy_from_user(&tag, &tags[i], sizeof(tag)))
+			goto fail;
+		if (fds && copy_from_user(&fd, &fds[i], sizeof(fd)))
 			goto fail;
-		}
 		/* allow sparse sets */
 		if (!fds || fd == -1) {
 			ret = -EINVAL;
-			if (unlikely(*io_get_tag_slot(ctx->file_data, i)))
+			if (tag)
 				goto fail;
 			continue;
 		}
@@ -737,56 +584,34 @@ int io_sqe_files_register(struct io_ring_ctx *ctx, void __user *arg,
 			fput(file);
 			goto fail;
 		}
-		file_slot = io_fixed_file_slot(&ctx->file_table, i);
-		io_fixed_file_set(file_slot, file);
+		ret = -ENOMEM;
+		node = io_rsrc_node_alloc(ctx, IORING_RSRC_FILE);
+		if (!node) {
+			fput(file);
+			goto fail;
+		}
+		if (tag)
+			node->tag = tag;
+		ctx->file_table.data.nodes[i] = node;
+		io_fixed_file_set(node, file);
 		io_file_bitmap_set(&ctx->file_table, i);
 	}
 
 	/* default it to the whole table */
-	io_file_table_set_alloc_range(ctx, 0, ctx->nr_user_files);
+	io_file_table_set_alloc_range(ctx, 0, ctx->file_table.data.nr);
 	return 0;
 fail:
-	__io_sqe_files_unregister(ctx);
+	io_clear_table_tags(&ctx->file_table.data);
+	io_sqe_files_unregister(ctx);
 	return ret;
 }
 
-static void io_rsrc_buf_put(struct io_ring_ctx *ctx, struct io_rsrc_put *prsrc)
-{
-	io_buffer_unmap(ctx, &prsrc->buf);
-	prsrc->buf = NULL;
-}
-
-void __io_sqe_buffers_unregister(struct io_ring_ctx *ctx)
-{
-	unsigned int i;
-
-	for (i = 0; i < ctx->nr_user_bufs; i++)
-		io_buffer_unmap(ctx, &ctx->user_bufs[i]);
-	kfree(ctx->user_bufs);
-	io_rsrc_data_free(ctx->buf_data);
-	ctx->user_bufs = NULL;
-	ctx->buf_data = NULL;
-	ctx->nr_user_bufs = 0;
-}
-
 int io_sqe_buffers_unregister(struct io_ring_ctx *ctx)
 {
-	unsigned nr = ctx->nr_user_bufs;
-	int ret;
-
-	if (!ctx->buf_data)
+	if (!ctx->buf_table.nr)
 		return -ENXIO;
-
-	/*
-	 * Quiesce may unlock ->uring_lock, and while it's not held
-	 * prevent new requests using the table.
-	 */
-	ctx->nr_user_bufs = 0;
-	ret = io_rsrc_ref_quiesce(ctx->buf_data, ctx);
-	ctx->nr_user_bufs = nr;
-	if (!ret)
-		__io_sqe_buffers_unregister(ctx);
-	return ret;
+	io_rsrc_data_free(ctx, &ctx->buf_table);
+	return 0;
 }
 
 /*
@@ -812,9 +637,13 @@ static bool headpage_already_acct(struct io_ring_ctx *ctx, struct page **pages,
 	}
 
 	/* check previously registered pages */
-	for (i = 0; i < ctx->nr_user_bufs; i++) {
-		struct io_mapped_ubuf *imu = ctx->user_bufs[i];
+	for (i = 0; i < ctx->buf_table.nr; i++) {
+		struct io_rsrc_node *node = ctx->buf_table.nodes[i];
+		struct io_mapped_ubuf *imu;
 
+		if (!node)
+			continue;
+		imu = node->buf;
 		for (j = 0; j < imu->nr_bvecs; j++) {
 			if (!PageCompound(imu->bvec[j].bv_page))
 				continue;
@@ -858,68 +687,60 @@ static int io_buffer_account_pin(struct io_ring_ctx *ctx, struct page **pages,
 	return ret;
 }
 
-static bool io_do_coalesce_buffer(struct page ***pages, int *nr_pages,
-				struct io_imu_folio_data *data, int nr_folios)
+static bool io_coalesce_buffer(struct page ***pages, int *nr_pages,
+				struct io_imu_folio_data *data)
 {
 	struct page **page_array = *pages, **new_array = NULL;
-	int nr_pages_left = *nr_pages, i, j;
+	unsigned nr_pages_left = *nr_pages;
+	unsigned nr_folios = data->nr_folios;
+	unsigned i, j;
 
 	/* Store head pages only*/
-	new_array = kvmalloc_array(nr_folios, sizeof(struct page *),
-					GFP_KERNEL);
+	new_array = kvmalloc_array(nr_folios, sizeof(struct page *), GFP_KERNEL);
 	if (!new_array)
 		return false;
 
-	new_array[0] = compound_head(page_array[0]);
-	/*
-	 * The pages are bound to the folio, it doesn't
-	 * actually unpin them but drops all but one reference,
-	 * which is usually put down by io_buffer_unmap().
-	 * Note, needs a better helper.
-	 */
-	if (data->nr_pages_head > 1)
-		unpin_user_pages(&page_array[1], data->nr_pages_head - 1);
-
-	j = data->nr_pages_head;
-	nr_pages_left -= data->nr_pages_head;
-	for (i = 1; i < nr_folios; i++) {
-		unsigned int nr_unpin;
-
-		new_array[i] = page_array[j];
-		nr_unpin = min_t(unsigned int, nr_pages_left - 1,
-					data->nr_pages_mid - 1);
-		if (nr_unpin)
-			unpin_user_pages(&page_array[j+1], nr_unpin);
-		j += data->nr_pages_mid;
-		nr_pages_left -= data->nr_pages_mid;
+	for (i = 0, j = 0; i < nr_folios; i++) {
+		struct page *p = compound_head(page_array[j]);
+		struct folio *folio = page_folio(p);
+		unsigned int nr;
+
+		WARN_ON_ONCE(i > 0 && p != page_array[j]);
+
+		nr = i ? data->nr_pages_mid : data->nr_pages_head;
+		nr = min(nr, nr_pages_left);
+		/* Drop all but one ref, the entire folio will remain pinned. */
+		if (nr > 1)
+			unpin_user_folio(folio, nr - 1);
+		j += nr;
+		nr_pages_left -= nr;
+		new_array[i] = p;
 	}
+
+	WARN_ON_ONCE(j != *nr_pages);
+
 	kvfree(page_array);
 	*pages = new_array;
 	*nr_pages = nr_folios;
 	return true;
 }
 
-static bool io_try_coalesce_buffer(struct page ***pages, int *nr_pages,
-					 struct io_imu_folio_data *data)
+bool io_check_coalesce_buffer(struct page **page_array, int nr_pages,
+			      struct io_imu_folio_data *data)
 {
-	struct page **page_array = *pages;
 	struct folio *folio = page_folio(page_array[0]);
 	unsigned int count = 1, nr_folios = 1;
 	int i;
 
-	if (*nr_pages <= 1)
-		return false;
-
 	data->nr_pages_mid = folio_nr_pages(folio);
-	if (data->nr_pages_mid == 1)
-		return false;
-
 	data->folio_shift = folio_shift(folio);
+	data->first_folio_page_idx = folio_page_idx(folio, page_array[0]);
+
 	/*
 	 * Check if pages are contiguous inside a folio, and all folios have
 	 * the same page count except for the head and tail.
 	 */
-	for (i = 1; i < *nr_pages; i++) {
+	for (i = 1; i < nr_pages; i++) {
 		if (page_folio(page_array[i]) == folio &&
 			page_array[i] == page_array[i-1] + 1) {
 			count++;
@@ -947,24 +768,29 @@ static bool io_try_coalesce_buffer(struct page ***pages, int *nr_pages,
 	if (nr_folios == 1)
 		data->nr_pages_head = count;
 
-	return io_do_coalesce_buffer(pages, nr_pages, data, nr_folios);
+	data->nr_folios = nr_folios;
+	return true;
 }
 
-static int io_sqe_buffer_register(struct io_ring_ctx *ctx, struct iovec *iov,
-				  struct io_mapped_ubuf **pimu,
-				  struct page **last_hpage)
+static struct io_rsrc_node *io_sqe_buffer_register(struct io_ring_ctx *ctx,
+						   struct iovec *iov,
+						   struct page **last_hpage)
 {
 	struct io_mapped_ubuf *imu = NULL;
 	struct page **pages = NULL;
+	struct io_rsrc_node *node;
 	unsigned long off;
 	size_t size;
 	int ret, nr_pages, i;
 	struct io_imu_folio_data data;
-	bool coalesced;
+	bool coalesced = false;
 
-	*pimu = (struct io_mapped_ubuf *)&dummy_ubuf;
 	if (!iov->iov_base)
-		return 0;
+		return NULL;
+
+	node = io_rsrc_node_alloc(ctx, IORING_RSRC_BUFFER);
+	if (!node)
+		return ERR_PTR(-ENOMEM);
 
 	ret = -ENOMEM;
 	pages = io_pin_pages((unsigned long) iov->iov_base, iov->iov_len,
@@ -976,29 +802,38 @@ static int io_sqe_buffer_register(struct io_ring_ctx *ctx, struct iovec *iov,
 	}
 
 	/* If it's huge page(s), try to coalesce them into fewer bvec entries */
-	coalesced = io_try_coalesce_buffer(&pages, &nr_pages, &data);
+	if (nr_pages > 1 && io_check_coalesce_buffer(pages, nr_pages, &data)) {
+		if (data.nr_pages_mid != 1)
+			coalesced = io_coalesce_buffer(&pages, &nr_pages, &data);
+	}
 
-	imu = kvmalloc(struct_size(imu, bvec, nr_pages), GFP_KERNEL);
+	imu = io_alloc_imu(ctx, nr_pages);
 	if (!imu)
 		goto done;
 
+	imu->nr_bvecs = nr_pages;
 	ret = io_buffer_account_pin(ctx, pages, nr_pages, imu, last_hpage);
-	if (ret) {
-		unpin_user_pages(pages, nr_pages);
+	if (ret)
 		goto done;
-	}
 
 	size = iov->iov_len;
 	/* store original address for later verification */
 	imu->ubuf = (unsigned long) iov->iov_base;
 	imu->len = iov->iov_len;
-	imu->nr_bvecs = nr_pages;
 	imu->folio_shift = PAGE_SHIFT;
+	imu->release = io_release_ubuf;
+	imu->priv = imu;
+	imu->is_kbuf = false;
+	imu->dir = IO_IMU_DEST | IO_IMU_SOURCE;
 	if (coalesced)
 		imu->folio_shift = data.folio_shift;
 	refcount_set(&imu->refs, 1);
-	off = (unsigned long) iov->iov_base & ((1UL << imu->folio_shift) - 1);
-	*pimu = imu;
+
+	off = (unsigned long)iov->iov_base & ~PAGE_MASK;
+	if (coalesced)
+		off += data.first_folio_page_idx << PAGE_SHIFT;
+
+	node->buf = imu;
 	ret = 0;
 
 	for (i = 0; i < nr_pages; i++) {
@@ -1010,46 +845,46 @@ static int io_sqe_buffer_register(struct io_ring_ctx *ctx, struct iovec *iov,
 		size -= vec_len;
 	}
 done:
-	if (ret)
-		kvfree(imu);
+	if (ret) {
+		if (imu)
+			io_free_imu(ctx, imu);
+		if (pages) {
+			for (i = 0; i < nr_pages; i++)
+				unpin_user_folio(page_folio(pages[i]), 1);
+		}
+		io_cache_free(&ctx->node_cache, node);
+		node = ERR_PTR(ret);
+	}
 	kvfree(pages);
-	return ret;
-}
-
-static int io_buffers_map_alloc(struct io_ring_ctx *ctx, unsigned int nr_args)
-{
-	ctx->user_bufs = kcalloc(nr_args, sizeof(*ctx->user_bufs), GFP_KERNEL);
-	return ctx->user_bufs ? 0 : -ENOMEM;
+	return node;
 }
 
 int io_sqe_buffers_register(struct io_ring_ctx *ctx, void __user *arg,
 			    unsigned int nr_args, u64 __user *tags)
 {
 	struct page *last_hpage = NULL;
-	struct io_rsrc_data *data;
+	struct io_rsrc_data data;
 	struct iovec fast_iov, *iov = &fast_iov;
 	const struct iovec __user *uvec;
 	int i, ret;
 
 	BUILD_BUG_ON(IORING_MAX_REG_BUFFERS >= (1u << 16));
 
-	if (ctx->user_bufs)
+	if (ctx->buf_table.nr)
 		return -EBUSY;
 	if (!nr_args || nr_args > IORING_MAX_REG_BUFFERS)
 		return -EINVAL;
-	ret = io_rsrc_data_alloc(ctx, IORING_RSRC_BUFFER, tags, nr_args, &data);
+	ret = io_rsrc_data_alloc(&data, nr_args);
 	if (ret)
 		return ret;
-	ret = io_buffers_map_alloc(ctx, nr_args);
-	if (ret) {
-		io_rsrc_data_free(data);
-		return ret;
-	}
 
 	if (!arg)
 		memset(iov, 0, sizeof(*iov));
 
-	for (i = 0; i < nr_args; i++, ctx->nr_user_bufs++) {
+	for (i = 0; i < nr_args; i++) {
+		struct io_rsrc_node *node;
+		u64 tag = 0;
+
 		if (arg) {
 			uvec = (struct iovec __user *) arg;
 			iov = iovec_from_user(uvec, 1, 1, &fast_iov, ctx->compat);
@@ -1066,142 +901,366 @@ int io_sqe_buffers_register(struct io_ring_ctx *ctx, void __user *arg,
 				arg += sizeof(struct iovec);
 		}
 
-		if (!iov->iov_base && *io_get_tag_slot(data, i)) {
-			ret = -EINVAL;
-			break;
+		if (tags) {
+			if (copy_from_user(&tag, &tags[i], sizeof(tag))) {
+				ret = -EFAULT;
+				break;
+			}
 		}
 
-		ret = io_sqe_buffer_register(ctx, iov, &ctx->user_bufs[i],
-					     &last_hpage);
-		if (ret)
+		node = io_sqe_buffer_register(ctx, iov, &last_hpage);
+		if (IS_ERR(node)) {
+			ret = PTR_ERR(node);
 			break;
+		}
+		if (tag) {
+			if (!node) {
+				ret = -EINVAL;
+				break;
+			}
+			node->tag = tag;
+		}
+		data.nodes[i] = node;
 	}
 
-	WARN_ON_ONCE(ctx->buf_data);
+	ctx->buf_table = data;
+	if (ret) {
+		io_clear_table_tags(&ctx->buf_table);
+		io_sqe_buffers_unregister(ctx);
+	}
+	return ret;
+}
 
-	ctx->buf_data = data;
-	if (ret)
-		__io_sqe_buffers_unregister(ctx);
+int io_buffer_register_bvec(struct io_uring_cmd *cmd, struct request *rq,
+			    void (*release)(void *), unsigned int index,
+			    unsigned int issue_flags)
+{
+	struct io_ring_ctx *ctx = cmd_to_io_kiocb(cmd)->ctx;
+	struct io_rsrc_data *data = &ctx->buf_table;
+	struct req_iterator rq_iter;
+	struct io_mapped_ubuf *imu;
+	struct io_rsrc_node *node;
+	struct bio_vec bv, *bvec;
+	u16 nr_bvecs;
+	int ret = 0;
+
+	io_ring_submit_lock(ctx, issue_flags);
+	if (index >= data->nr) {
+		ret = -EINVAL;
+		goto unlock;
+	}
+	index = array_index_nospec(index, data->nr);
+
+	if (data->nodes[index]) {
+		ret = -EBUSY;
+		goto unlock;
+	}
+
+	node = io_rsrc_node_alloc(ctx, IORING_RSRC_BUFFER);
+	if (!node) {
+		ret = -ENOMEM;
+		goto unlock;
+	}
+
+	nr_bvecs = blk_rq_nr_phys_segments(rq);
+	imu = io_alloc_imu(ctx, nr_bvecs);
+	if (!imu) {
+		kfree(node);
+		ret = -ENOMEM;
+		goto unlock;
+	}
+
+	imu->ubuf = 0;
+	imu->len = blk_rq_bytes(rq);
+	imu->acct_pages = 0;
+	imu->folio_shift = PAGE_SHIFT;
+	imu->nr_bvecs = nr_bvecs;
+	refcount_set(&imu->refs, 1);
+	imu->release = release;
+	imu->priv = rq;
+	imu->is_kbuf = true;
+	imu->dir = 1 << rq_data_dir(rq);
+
+	bvec = imu->bvec;
+	rq_for_each_bvec(bv, rq, rq_iter)
+		*bvec++ = bv;
+
+	node->buf = imu;
+	data->nodes[index] = node;
+unlock:
+	io_ring_submit_unlock(ctx, issue_flags);
 	return ret;
 }
+EXPORT_SYMBOL_GPL(io_buffer_register_bvec);
 
-int io_import_fixed(int ddir, struct iov_iter *iter,
-			   struct io_mapped_ubuf *imu,
-			   u64 buf_addr, size_t len)
+int io_buffer_unregister_bvec(struct io_uring_cmd *cmd, unsigned int index,
+			      unsigned int issue_flags)
+{
+	struct io_ring_ctx *ctx = cmd_to_io_kiocb(cmd)->ctx;
+	struct io_rsrc_data *data = &ctx->buf_table;
+	struct io_rsrc_node *node;
+	int ret = 0;
+
+	io_ring_submit_lock(ctx, issue_flags);
+	if (index >= data->nr) {
+		ret = -EINVAL;
+		goto unlock;
+	}
+	index = array_index_nospec(index, data->nr);
+
+	node = data->nodes[index];
+	if (!node) {
+		ret = -EINVAL;
+		goto unlock;
+	}
+	if (!node->buf->is_kbuf) {
+		ret = -EBUSY;
+		goto unlock;
+	}
+
+	io_put_rsrc_node(ctx, node);
+	data->nodes[index] = NULL;
+unlock:
+	io_ring_submit_unlock(ctx, issue_flags);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(io_buffer_unregister_bvec);
+
+static int validate_fixed_range(u64 buf_addr, size_t len,
+				const struct io_mapped_ubuf *imu)
 {
 	u64 buf_end;
-	size_t offset;
 
-	if (WARN_ON_ONCE(!imu))
-		return -EFAULT;
 	if (unlikely(check_add_overflow(buf_addr, (u64)len, &buf_end)))
 		return -EFAULT;
 	/* not inside the mapped region */
 	if (unlikely(buf_addr < imu->ubuf || buf_end > (imu->ubuf + imu->len)))
 		return -EFAULT;
+	if (unlikely(len > MAX_RW_COUNT))
+		return -EFAULT;
+	return 0;
+}
 
-	/*
-	 * Might not be a start of buffer, set size appropriately
-	 * and advance us to the beginning.
-	 */
-	offset = buf_addr - imu->ubuf;
-	iov_iter_bvec(iter, ddir, imu->bvec, imu->nr_bvecs, offset + len);
-
-	if (offset) {
-		/*
-		 * Don't use iov_iter_advance() here, as it's really slow for
-		 * using the latter parts of a big fixed buffer - it iterates
-		 * over each segment manually. We can cheat a bit here, because
-		 * we know that:
-		 *
-		 * 1) it's a BVEC iter, we set it up
-		 * 2) all bvecs are the same in size, except potentially the
-		 *    first and last bvec
-		 *
-		 * So just find our index, and adjust the iterator afterwards.
-		 * If the offset is within the first bvec (or the whole first
-		 * bvec, just use iov_iter_advance(). This makes it easier
-		 * since we can just skip the first segment, which may not
-		 * be folio_size aligned.
-		 */
-		const struct bio_vec *bvec = imu->bvec;
+static int io_import_kbuf(int ddir, struct iov_iter *iter,
+			  struct io_mapped_ubuf *imu, size_t len, size_t offset)
+{
+	size_t count = len + offset;
 
-		if (offset < bvec->bv_len) {
-			iter->bvec = bvec;
-			iter->count -= offset;
-			iter->iov_offset = offset;
-		} else {
-			unsigned long seg_skip;
+	iov_iter_bvec(iter, ddir, imu->bvec, imu->nr_bvecs, count);
+	iov_iter_advance(iter, offset);
 
-			/* skip first vec */
-			offset -= bvec->bv_len;
-			seg_skip = 1 + (offset >> imu->folio_shift);
+	if (count < imu->len) {
+		const struct bio_vec *bvec = iter->bvec;
 
-			iter->bvec = bvec + seg_skip;
-			iter->nr_segs -= seg_skip;
-			iter->count -= bvec->bv_len + offset;
-			iter->iov_offset = offset & ((1UL << imu->folio_shift) - 1);
+		while (len > bvec->bv_len) {
+			len -= bvec->bv_len;
+			bvec++;
 		}
+		iter->nr_segs = 1 + bvec - iter->bvec;
 	}
+	return 0;
+}
+
+static int io_import_fixed(int ddir, struct iov_iter *iter,
+			   struct io_mapped_ubuf *imu,
+			   u64 buf_addr, size_t len)
+{
+	const struct bio_vec *bvec;
+	size_t folio_mask;
+	unsigned nr_segs;
+	size_t offset;
+	int ret;
 
+	ret = validate_fixed_range(buf_addr, len, imu);
+	if (unlikely(ret))
+		return ret;
+	if (!(imu->dir & (1 << ddir)))
+		return -EFAULT;
+
+	offset = buf_addr - imu->ubuf;
+
+	if (imu->is_kbuf)
+		return io_import_kbuf(ddir, iter, imu, len, offset);
+
+	/*
+	 * Don't use iov_iter_advance() here, as it's really slow for
+	 * using the latter parts of a big fixed buffer - it iterates
+	 * over each segment manually. We can cheat a bit here for user
+	 * registered nodes, because we know that:
+	 *
+	 * 1) it's a BVEC iter, we set it up
+	 * 2) all bvecs are the same in size, except potentially the
+	 *    first and last bvec
+	 */
+	folio_mask = (1UL << imu->folio_shift) - 1;
+	bvec = imu->bvec;
+	if (offset >= bvec->bv_len) {
+		unsigned long seg_skip;
+
+		/* skip first vec */
+		offset -= bvec->bv_len;
+		seg_skip = 1 + (offset >> imu->folio_shift);
+		bvec += seg_skip;
+		offset &= folio_mask;
+	}
+	nr_segs = (offset + len + bvec->bv_offset + folio_mask) >> imu->folio_shift;
+	iov_iter_bvec(iter, ddir, bvec, nr_segs, len);
+	iter->iov_offset = offset;
 	return 0;
 }
 
-static int io_clone_buffers(struct io_ring_ctx *ctx, struct io_ring_ctx *src_ctx)
+inline struct io_rsrc_node *io_find_buf_node(struct io_kiocb *req,
+					     unsigned issue_flags)
 {
-	struct io_mapped_ubuf **user_bufs;
-	struct io_rsrc_data *data;
-	int i, ret, nbufs;
+	struct io_ring_ctx *ctx = req->ctx;
+	struct io_rsrc_node *node;
+
+	if (req->flags & REQ_F_BUF_NODE)
+		return req->buf_node;
+	req->flags |= REQ_F_BUF_NODE;
+
+	io_ring_submit_lock(ctx, issue_flags);
+	node = io_rsrc_node_lookup(&ctx->buf_table, req->buf_index);
+	if (node) {
+		node->refs++;
+		req->buf_node = node;
+		io_ring_submit_unlock(ctx, issue_flags);
+		return node;
+	}
+	req->flags &= ~REQ_F_BUF_NODE;
+	io_ring_submit_unlock(ctx, issue_flags);
+	return NULL;
+}
+
+int io_import_reg_buf(struct io_kiocb *req, struct iov_iter *iter,
+			u64 buf_addr, size_t len, int ddir,
+			unsigned issue_flags)
+{
+	struct io_rsrc_node *node;
+
+	node = io_find_buf_node(req, issue_flags);
+	if (!node)
+		return -EFAULT;
+	return io_import_fixed(ddir, iter, node->buf, buf_addr, len);
+}
+
+/* Lock two rings at once. The rings must be different! */
+static void lock_two_rings(struct io_ring_ctx *ctx1, struct io_ring_ctx *ctx2)
+{
+	if (ctx1 > ctx2)
+		swap(ctx1, ctx2);
+	mutex_lock(&ctx1->uring_lock);
+	mutex_lock_nested(&ctx2->uring_lock, SINGLE_DEPTH_NESTING);
+}
+
+/* Both rings are locked by the caller. */
+static int io_clone_buffers(struct io_ring_ctx *ctx, struct io_ring_ctx *src_ctx,
+			    struct io_uring_clone_buffers *arg)
+{
+	struct io_rsrc_data data;
+	int i, ret, off, nr;
+	unsigned int nbufs;
+
+	lockdep_assert_held(&ctx->uring_lock);
+	lockdep_assert_held(&src_ctx->uring_lock);
 
 	/*
-	 * Drop our own lock here. We'll setup the data we need and reference
-	 * the source buffers, then re-grab, check, and assign at the end.
+	 * Accounting state is shared between the two rings; that only works if
+	 * both rings are accounted towards the same counters.
 	 */
-	mutex_unlock(&ctx->uring_lock);
+	if (ctx->user != src_ctx->user || ctx->mm_account != src_ctx->mm_account)
+		return -EINVAL;
 
-	mutex_lock(&src_ctx->uring_lock);
-	ret = -ENXIO;
-	nbufs = src_ctx->nr_user_bufs;
-	if (!nbufs)
-		goto out_unlock;
-	ret = io_rsrc_data_alloc(ctx, IORING_RSRC_BUFFER, NULL, nbufs, &data);
-	if (ret)
-		goto out_unlock;
+	/* if offsets are given, must have nr specified too */
+	if (!arg->nr && (arg->dst_off || arg->src_off))
+		return -EINVAL;
+	/* not allowed unless REPLACE is set */
+	if (ctx->buf_table.nr && !(arg->flags & IORING_REGISTER_DST_REPLACE))
+		return -EBUSY;
 
-	ret = -ENOMEM;
-	user_bufs = kcalloc(nbufs, sizeof(*ctx->user_bufs), GFP_KERNEL);
-	if (!user_bufs)
-		goto out_free_data;
+	nbufs = src_ctx->buf_table.nr;
+	if (!arg->nr)
+		arg->nr = nbufs;
+	else if (arg->nr > nbufs)
+		return -EINVAL;
+	else if (arg->nr > IORING_MAX_REG_BUFFERS)
+		return -EINVAL;
+	if (check_add_overflow(arg->nr, arg->dst_off, &nbufs))
+		return -EOVERFLOW;
+	if (nbufs > IORING_MAX_REG_BUFFERS)
+		return -EINVAL;
 
-	for (i = 0; i < nbufs; i++) {
-		struct io_mapped_ubuf *src = src_ctx->user_bufs[i];
+	ret = io_rsrc_data_alloc(&data, max(nbufs, ctx->buf_table.nr));
+	if (ret)
+		return ret;
 
-		if (src != &dummy_ubuf)
-			refcount_inc(&src->refs);
-		user_bufs[i] = src;
+	/* Fill entries in data from dst that won't overlap with src */
+	for (i = 0; i < min(arg->dst_off, ctx->buf_table.nr); i++) {
+		struct io_rsrc_node *src_node = ctx->buf_table.nodes[i];
+
+		if (src_node) {
+			data.nodes[i] = src_node;
+			src_node->refs++;
+		}
 	}
 
-	/* Have a ref on the bufs now, drop src lock and re-grab our own lock */
-	mutex_unlock(&src_ctx->uring_lock);
-	mutex_lock(&ctx->uring_lock);
-	if (!ctx->user_bufs) {
-		ctx->user_bufs = user_bufs;
-		ctx->buf_data = data;
-		ctx->nr_user_bufs = nbufs;
-		return 0;
+	ret = -ENXIO;
+	nbufs = src_ctx->buf_table.nr;
+	if (!nbufs)
+		goto out_free;
+	ret = -EINVAL;
+	if (!arg->nr)
+		arg->nr = nbufs;
+	else if (arg->nr > nbufs)
+		goto out_free;
+	ret = -EOVERFLOW;
+	if (check_add_overflow(arg->nr, arg->src_off, &off))
+		goto out_free;
+	if (off > nbufs)
+		goto out_free;
+
+	off = arg->dst_off;
+	i = arg->src_off;
+	nr = arg->nr;
+	while (nr--) {
+		struct io_rsrc_node *dst_node, *src_node;
+
+		src_node = io_rsrc_node_lookup(&src_ctx->buf_table, i);
+		if (!src_node) {
+			dst_node = NULL;
+		} else {
+			dst_node = io_rsrc_node_alloc(ctx, IORING_RSRC_BUFFER);
+			if (!dst_node) {
+				ret = -ENOMEM;
+				goto out_free;
+			}
+
+			refcount_inc(&src_node->buf->refs);
+			dst_node->buf = src_node->buf;
+		}
+		data.nodes[off++] = dst_node;
+		i++;
 	}
 
-	/* someone raced setting up buffers, dump ours */
-	for (i = 0; i < nbufs; i++)
-		io_buffer_unmap(ctx, &user_bufs[i]);
-	io_rsrc_data_free(data);
-	kfree(user_bufs);
-	return -EBUSY;
-out_free_data:
-	io_rsrc_data_free(data);
-out_unlock:
-	mutex_unlock(&src_ctx->uring_lock);
-	mutex_lock(&ctx->uring_lock);
+	/*
+	 * If asked for replace, put the old table. data->nodes[] holds both
+	 * old and new nodes at this point.
+	 */
+	if (arg->flags & IORING_REGISTER_DST_REPLACE)
+		io_rsrc_data_free(ctx, &ctx->buf_table);
+
+	/*
+	 * ctx->buf_table must be empty now - either the contents are being
+	 * replaced and we just freed the table, or the contents are being
+	 * copied to a ring that does not have buffers yet (checked at function
+	 * entry).
+	 */
+	WARN_ON_ONCE(ctx->buf_table.nr);
+	ctx->buf_table = data;
+	return 0;
+
+out_free:
+	io_rsrc_data_free(ctx, &data);
 	return ret;
 }
 
@@ -1215,16 +1274,17 @@ out_unlock:
 int io_register_clone_buffers(struct io_ring_ctx *ctx, void __user *arg)
 {
 	struct io_uring_clone_buffers buf;
+	struct io_ring_ctx *src_ctx;
 	bool registered_src;
 	struct file *file;
 	int ret;
 
-	if (ctx->user_bufs || ctx->nr_user_bufs)
-		return -EBUSY;
 	if (copy_from_user(&buf, arg, sizeof(buf)))
 		return -EFAULT;
-	if (buf.flags & ~IORING_REGISTER_SRC_REGISTERED)
+	if (buf.flags & ~(IORING_REGISTER_SRC_REGISTERED|IORING_REGISTER_DST_REPLACE))
 		return -EINVAL;
+	if (!(buf.flags & IORING_REGISTER_DST_REPLACE) && ctx->buf_table.nr)
+		return -EBUSY;
 	if (memchr_inv(buf.pad, 0, sizeof(buf.pad)))
 		return -EINVAL;
 
@@ -1232,8 +1292,270 @@ int io_register_clone_buffers(struct io_ring_ctx *ctx, void __user *arg)
 	file = io_uring_register_get_file(buf.src_fd, registered_src);
 	if (IS_ERR(file))
 		return PTR_ERR(file);
-	ret = io_clone_buffers(ctx, file->private_data);
-	if (!registered_src)
-		fput(file);
+
+	src_ctx = file->private_data;
+	if (src_ctx != ctx) {
+		mutex_unlock(&ctx->uring_lock);
+		lock_two_rings(ctx, src_ctx);
+	}
+
+	ret = io_clone_buffers(ctx, src_ctx, &buf);
+
+	if (src_ctx != ctx)
+		mutex_unlock(&src_ctx->uring_lock);
+
+	fput(file);
 	return ret;
 }
+
+void io_vec_free(struct iou_vec *iv)
+{
+	if (!iv->iovec)
+		return;
+	kfree(iv->iovec);
+	iv->iovec = NULL;
+	iv->nr = 0;
+}
+
+int io_vec_realloc(struct iou_vec *iv, unsigned nr_entries)
+{
+	gfp_t gfp = GFP_KERNEL | __GFP_NOWARN;
+	struct iovec *iov;
+
+	iov = kmalloc_array(nr_entries, sizeof(iov[0]), gfp);
+	if (!iov)
+		return -ENOMEM;
+
+	io_vec_free(iv);
+	iv->iovec = iov;
+	iv->nr = nr_entries;
+	return 0;
+}
+
+static int io_vec_fill_bvec(int ddir, struct iov_iter *iter,
+				struct io_mapped_ubuf *imu,
+				struct iovec *iovec, unsigned nr_iovs,
+				struct iou_vec *vec)
+{
+	unsigned long folio_size = 1 << imu->folio_shift;
+	unsigned long folio_mask = folio_size - 1;
+	struct bio_vec *res_bvec = vec->bvec;
+	size_t total_len = 0;
+	unsigned bvec_idx = 0;
+	unsigned iov_idx;
+
+	for (iov_idx = 0; iov_idx < nr_iovs; iov_idx++) {
+		size_t iov_len = iovec[iov_idx].iov_len;
+		u64 buf_addr = (u64)(uintptr_t)iovec[iov_idx].iov_base;
+		struct bio_vec *src_bvec;
+		size_t offset;
+		int ret;
+
+		ret = validate_fixed_range(buf_addr, iov_len, imu);
+		if (unlikely(ret))
+			return ret;
+
+		if (unlikely(!iov_len))
+			return -EFAULT;
+		if (unlikely(check_add_overflow(total_len, iov_len, &total_len)))
+			return -EOVERFLOW;
+
+		offset = buf_addr - imu->ubuf;
+		/*
+		 * Only the first bvec can have non zero bv_offset, account it
+		 * here and work with full folios below.
+		 */
+		offset += imu->bvec[0].bv_offset;
+
+		src_bvec = imu->bvec + (offset >> imu->folio_shift);
+		offset &= folio_mask;
+
+		for (; iov_len; offset = 0, bvec_idx++, src_bvec++) {
+			size_t seg_size = min_t(size_t, iov_len,
+						folio_size - offset);
+
+			bvec_set_page(&res_bvec[bvec_idx],
+				      src_bvec->bv_page, seg_size, offset);
+			iov_len -= seg_size;
+		}
+	}
+	if (total_len > MAX_RW_COUNT)
+		return -EINVAL;
+
+	iov_iter_bvec(iter, ddir, res_bvec, bvec_idx, total_len);
+	return 0;
+}
+
+static int io_estimate_bvec_size(struct iovec *iov, unsigned nr_iovs,
+				 struct io_mapped_ubuf *imu)
+{
+	unsigned shift = imu->folio_shift;
+	size_t max_segs = 0;
+	unsigned i;
+
+	for (i = 0; i < nr_iovs; i++)
+		max_segs += (iov[i].iov_len >> shift) + 2;
+	return max_segs;
+}
+
+static int io_vec_fill_kern_bvec(int ddir, struct iov_iter *iter,
+				 struct io_mapped_ubuf *imu,
+				 struct iovec *iovec, unsigned nr_iovs,
+				 struct iou_vec *vec)
+{
+	const struct bio_vec *src_bvec = imu->bvec;
+	struct bio_vec *res_bvec = vec->bvec;
+	unsigned res_idx = 0;
+	size_t total_len = 0;
+	unsigned iov_idx;
+
+	for (iov_idx = 0; iov_idx < nr_iovs; iov_idx++) {
+		size_t offset = (size_t)(uintptr_t)iovec[iov_idx].iov_base;
+		size_t iov_len = iovec[iov_idx].iov_len;
+		struct bvec_iter bi = {
+			.bi_size        = offset + iov_len,
+		};
+		struct bio_vec bv;
+
+		bvec_iter_advance(src_bvec, &bi, offset);
+		for_each_mp_bvec(bv, src_bvec, bi, bi)
+			res_bvec[res_idx++] = bv;
+		total_len += iov_len;
+	}
+	iov_iter_bvec(iter, ddir, res_bvec, res_idx, total_len);
+	return 0;
+}
+
+static int iov_kern_bvec_size(const struct iovec *iov,
+			      const struct io_mapped_ubuf *imu,
+			      unsigned int *nr_seg)
+{
+	size_t offset = (size_t)(uintptr_t)iov->iov_base;
+	const struct bio_vec *bvec = imu->bvec;
+	int start = 0, i = 0;
+	size_t off = 0;
+	int ret;
+
+	ret = validate_fixed_range(offset, iov->iov_len, imu);
+	if (unlikely(ret))
+		return ret;
+
+	for (i = 0; off < offset + iov->iov_len && i < imu->nr_bvecs;
+			off += bvec[i].bv_len, i++) {
+		if (offset >= off && offset < off + bvec[i].bv_len)
+			start = i;
+	}
+	*nr_seg = i - start;
+	return 0;
+}
+
+static int io_kern_bvec_size(struct iovec *iov, unsigned nr_iovs,
+			     struct io_mapped_ubuf *imu, unsigned *nr_segs)
+{
+	unsigned max_segs = 0;
+	size_t total_len = 0;
+	unsigned i;
+	int ret;
+
+	*nr_segs = 0;
+	for (i = 0; i < nr_iovs; i++) {
+		if (unlikely(!iov[i].iov_len))
+			return -EFAULT;
+		if (unlikely(check_add_overflow(total_len, iov[i].iov_len,
+						&total_len)))
+			return -EOVERFLOW;
+		ret = iov_kern_bvec_size(&iov[i], imu, &max_segs);
+		if (unlikely(ret))
+			return ret;
+		*nr_segs += max_segs;
+	}
+	if (total_len > MAX_RW_COUNT)
+		return -EINVAL;
+	return 0;
+}
+
+int io_import_reg_vec(int ddir, struct iov_iter *iter,
+			struct io_kiocb *req, struct iou_vec *vec,
+			unsigned nr_iovs, unsigned issue_flags)
+{
+	struct io_rsrc_node *node;
+	struct io_mapped_ubuf *imu;
+	unsigned iovec_off;
+	struct iovec *iov;
+	unsigned nr_segs;
+
+	node = io_find_buf_node(req, issue_flags);
+	if (!node)
+		return -EFAULT;
+	imu = node->buf;
+	if (!(imu->dir & (1 << ddir)))
+		return -EFAULT;
+
+	iovec_off = vec->nr - nr_iovs;
+	iov = vec->iovec + iovec_off;
+
+	if (imu->is_kbuf) {
+		int ret = io_kern_bvec_size(iov, nr_iovs, imu, &nr_segs);
+
+		if (unlikely(ret))
+			return ret;
+	} else {
+		nr_segs = io_estimate_bvec_size(iov, nr_iovs, imu);
+	}
+
+	if (sizeof(struct bio_vec) > sizeof(struct iovec)) {
+		size_t bvec_bytes;
+
+		bvec_bytes = nr_segs * sizeof(struct bio_vec);
+		nr_segs = (bvec_bytes + sizeof(*iov) - 1) / sizeof(*iov);
+		nr_segs += nr_iovs;
+	}
+
+	if (nr_segs > vec->nr) {
+		struct iou_vec tmp_vec = {};
+		int ret;
+
+		ret = io_vec_realloc(&tmp_vec, nr_segs);
+		if (ret)
+			return ret;
+
+		iovec_off = tmp_vec.nr - nr_iovs;
+		memcpy(tmp_vec.iovec + iovec_off, iov, sizeof(*iov) * nr_iovs);
+		io_vec_free(vec);
+
+		*vec = tmp_vec;
+		iov = vec->iovec + iovec_off;
+		req->flags |= REQ_F_NEED_CLEANUP;
+	}
+
+	if (imu->is_kbuf)
+		return io_vec_fill_kern_bvec(ddir, iter, imu, iov, nr_iovs, vec);
+
+	return io_vec_fill_bvec(ddir, iter, imu, iov, nr_iovs, vec);
+}
+
+int io_prep_reg_iovec(struct io_kiocb *req, struct iou_vec *iv,
+		      const struct iovec __user *uvec, size_t uvec_segs)
+{
+	struct iovec *iov;
+	int iovec_off, ret;
+	void *res;
+
+	if (uvec_segs > iv->nr) {
+		ret = io_vec_realloc(iv, uvec_segs);
+		if (ret)
+			return ret;
+		req->flags |= REQ_F_NEED_CLEANUP;
+	}
+
+	/* pad iovec to the right */
+	iovec_off = iv->nr - uvec_segs;
+	iov = iv->iovec + iovec_off;
+	res = iovec_from_user(uvec, uvec_segs, uvec_segs, iov,
+			      io_is_compat(req->ctx));
+	if (IS_ERR(res))
+		return PTR_ERR(res);
+
+	req->flags |= REQ_F_IMPORT_BUFFER;
+	return 0;
+}