Merge tag 'io_uring-5.15-2021-09-25' of git://git.kernel.dk/linux-block

Pull io_uring fixes from Jens Axboe:
 "This one looks a bit bigger than it is, but that's mainly because 2/3
  of it is enabling IORING_OP_CLOSE to close direct file descriptors.

  We've had a few folks using them and finding it confusing that the way
  to close them is through using -1 for file update, this just brings
  API symmetry for direct descriptors. Hence I think we should just do
  this now and have a better API for 5.15 release. There's some room for
  de-duplicating the close code, but we're leaving that for the next
  merge window.

  Outside of that, just small fixes:

   - Poll race fixes (Hao)

   - io-wq core dump exit fix (me)

   - Reschedule around potentially intensive tctx and buffer iterators
     on teardown (me)

   - Fix for always ending up punting files update to io-wq (me)

   - Put the provided buffer meta data under memcg accounting (me)

   - Tweak for io_write(), removing dead code that was added with the
     iterator changes in this release (Pavel)"

* tag 'io_uring-5.15-2021-09-25' of git://git.kernel.dk/linux-block:
  io_uring: make OP_CLOSE consistent with direct open
  io_uring: kill extra checks in io_write()
  io_uring: don't punt files update to io-wq unconditionally
  io_uring: put provided buffer meta data under memcg accounting
  io_uring: allow conditional reschedule for intensive iterators
  io_uring: fix potential req refcount underflow
  io_uring: fix missing set of EPOLLONESHOT for CQ ring overflow
  io_uring: fix race between poll completion and cancel_hash insertion
  io-wq: ensure we exit if thread group is exiting

fs/io-wq.c

@@ -584,7 +584,8 @@ static int io_wqe_worker(void *data)
 
 			if (!get_signal(&ksig))
 				continue;
-			if (fatal_signal_pending(current))
+			if (fatal_signal_pending(current) ||
+			    signal_group_exit(current->signal))
 				break;
 			continue;
 		}

fs/io_uring.c

@@ -502,6 +502,7 @@ struct io_poll_update {
 struct io_close {
 	struct file			*file;
 	int				fd;
+	u32				file_slot;
 };
 
 struct io_timeout_data {
@@ -1098,6 +1099,8 @@ static int io_req_prep_async(struct io_kiocb *req);
 
 static int io_install_fixed_file(struct io_kiocb *req, struct file *file,
 				 unsigned int issue_flags, u32 slot_index);
+static int io_close_fixed(struct io_kiocb *req, unsigned int issue_flags);
+
 static enum hrtimer_restart io_link_timeout_fn(struct hrtimer *timer);
 
 static struct kmem_cache *req_cachep;
@@ -3605,7 +3608,6 @@ static int io_write(struct io_kiocb *req, unsigned int issue_flags)
 		iov_iter_save_state(iter, state);
 	}
 	req->result = iov_iter_count(iter);
-	ret2 = 0;
 
 	/* Ensure we clear previously set non-block flag */
 	if (!force_nonblock)
@@ -3670,8 +3672,6 @@ static int io_write(struct io_kiocb *req, unsigned int issue_flags)
 	} else {
 copy_iov:
 		iov_iter_restore(iter, state);
-		if (ret2 > 0)
-			iov_iter_advance(iter, ret2);
 		ret = io_setup_async_rw(req, iovec, inline_vecs, iter, false);
 		return ret ?: -EAGAIN;
 	}
@@ -4387,7 +4387,7 @@ static int io_add_buffers(struct io_provide_buf *pbuf, struct io_buffer **head)
 	int i, bid = pbuf->bid;
 
 	for (i = 0; i < pbuf->nbufs; i++) {
-		buf = kmalloc(sizeof(*buf), GFP_KERNEL);
+		buf = kmalloc(sizeof(*buf), GFP_KERNEL_ACCOUNT);
 		if (!buf)
 			break;
 
@@ -4594,12 +4594,16 @@ static int io_close_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
 		return -EINVAL;
 	if (sqe->ioprio || sqe->off || sqe->addr || sqe->len ||
-	    sqe->rw_flags || sqe->buf_index || sqe->splice_fd_in)
+	    sqe->rw_flags || sqe->buf_index)
 		return -EINVAL;
 	if (req->flags & REQ_F_FIXED_FILE)
 		return -EBADF;
 
 	req->close.fd = READ_ONCE(sqe->fd);
+	req->close.file_slot = READ_ONCE(sqe->file_index);
+	if (req->close.file_slot && req->close.fd)
+		return -EINVAL;
+
 	return 0;
 }
 
@@ -4611,6 +4615,11 @@ static int io_close(struct io_kiocb *req, unsigned int issue_flags)
 	struct file *file = NULL;
 	int ret = -EBADF;
 
+	if (req->close.file_slot) {
+		ret = io_close_fixed(req, issue_flags);
+		goto err;
+	}
+
 	spin_lock(&files->file_lock);
 	fdt = files_fdtable(files);
 	if (close->fd >= fdt->max_fds) {
@@ -5338,7 +5347,7 @@ static bool __io_poll_complete(struct io_kiocb *req, __poll_t mask)
 	if (req->poll.events & EPOLLONESHOT)
 		flags = 0;
 	if (!io_cqring_fill_event(ctx, req->user_data, error, flags)) {
-		req->poll.done = true;
+		req->poll.events |= EPOLLONESHOT;
 		flags = 0;
 	}
 	if (flags & IORING_CQE_F_MORE)
@@ -5367,10 +5376,15 @@ static void io_poll_task_func(struct io_kiocb *req, bool *locked)
 	} else {
 		bool done;
 
+		if (req->poll.done) {
+			spin_unlock(&ctx->completion_lock);
+			return;
+		}
 		done = __io_poll_complete(req, req->result);
 		if (done) {
 			io_poll_remove_double(req);
 			hash_del(&req->hash_node);
+			req->poll.done = true;
 		} else {
 			req->result = 0;
 			add_wait_queue(req->poll.head, &req->poll.wait);
@@ -5508,6 +5522,7 @@ static void io_async_task_func(struct io_kiocb *req, bool *locked)
 
 	hash_del(&req->hash_node);
 	io_poll_remove_double(req);
+	apoll->poll.done = true;
 	spin_unlock(&ctx->completion_lock);
 
 	if (!READ_ONCE(apoll->poll.canceled))
@@ -5828,6 +5843,7 @@ static int io_poll_add(struct io_kiocb *req, unsigned int issue_flags)
 	struct io_ring_ctx *ctx = req->ctx;
 	struct io_poll_table ipt;
 	__poll_t mask;
+	bool done;
 
 	ipt.pt._qproc = io_poll_queue_proc;
 
@@ -5836,13 +5852,13 @@ static int io_poll_add(struct io_kiocb *req, unsigned int issue_flags)
 
 	if (mask) { /* no async, we'd stolen it */
 		ipt.error = 0;
-		io_poll_complete(req, mask);
+		done = io_poll_complete(req, mask);
 	}
 	spin_unlock(&ctx->completion_lock);
 
 	if (mask) {
 		io_cqring_ev_posted(ctx);
-		if (poll->events & EPOLLONESHOT)
+		if (done)
 			io_put_req(req);
 	}
 	return ipt.error;
@@ -6333,19 +6349,16 @@ static int io_files_update(struct io_kiocb *req, unsigned int issue_flags)
 	struct io_uring_rsrc_update2 up;
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
-
 	up.offset = req->rsrc_update.offset;
 	up.data = req->rsrc_update.arg;
 	up.nr = 0;
 	up.tags = 0;
 	up.resv = 0;
 
-	mutex_lock(&ctx->uring_lock);
+	io_ring_submit_lock(ctx, !(issue_flags & IO_URING_F_NONBLOCK));
 	ret = __io_register_rsrc_update(ctx, IORING_RSRC_FILE,
 					&up, req->rsrc_update.nr_args);
-	mutex_unlock(&ctx->uring_lock);
+	io_ring_submit_unlock(ctx, !(issue_flags & IO_URING_F_NONBLOCK));
 
 	if (ret < 0)
 		req_set_fail(req);
@@ -8400,6 +8413,44 @@ static int io_install_fixed_file(struct io_kiocb *req, struct file *file,
 	return ret;
 }
 
+static int io_close_fixed(struct io_kiocb *req, unsigned int issue_flags)
+{
+	unsigned int offset = req->close.file_slot - 1;
+	struct io_ring_ctx *ctx = req->ctx;
+	struct io_fixed_file *file_slot;
+	struct file *file;
+	int ret, i;
+
+	io_ring_submit_lock(ctx, !(issue_flags & IO_URING_F_NONBLOCK));
+	ret = -ENXIO;
+	if (unlikely(!ctx->file_data))
+		goto out;
+	ret = -EINVAL;
+	if (offset >= ctx->nr_user_files)
+		goto out;
+	ret = io_rsrc_node_switch_start(ctx);
+	if (ret)
+		goto out;
+
+	i = array_index_nospec(offset, ctx->nr_user_files);
+	file_slot = io_fixed_file_slot(&ctx->file_table, i);
+	ret = -EBADF;
+	if (!file_slot->file_ptr)
+		goto out;
+
+	file = (struct file *)(file_slot->file_ptr & FFS_MASK);
+	ret = io_queue_rsrc_removal(ctx->file_data, offset, ctx->rsrc_node, file);
+	if (ret)
+		goto out;
+
+	file_slot->file_ptr = 0;
+	io_rsrc_node_switch(ctx, ctx->file_data);
+	ret = 0;
+out:
+	io_ring_submit_unlock(ctx, !(issue_flags & IO_URING_F_NONBLOCK));
+	return ret;
+}
+
 static int __io_sqe_files_update(struct io_ring_ctx *ctx,
 				 struct io_uring_rsrc_update2 *up,
 				 unsigned nr_args)
@@ -9166,8 +9217,10 @@ static void io_destroy_buffers(struct io_ring_ctx *ctx)
 	struct io_buffer *buf;
 	unsigned long index;
 
-	xa_for_each(&ctx->io_buffers, index, buf)
+	xa_for_each(&ctx->io_buffers, index, buf) {
 		__io_remove_buffers(ctx, buf, index, -1U);
+		cond_resched();
+	}
 }
 
 static void io_req_cache_free(struct list_head *list)
@@ -9665,8 +9718,10 @@ static void io_uring_clean_tctx(struct io_uring_task *tctx)
 	struct io_tctx_node *node;
 	unsigned long index;
 
-	xa_for_each(&tctx->xa, index, node)
+	xa_for_each(&tctx->xa, index, node) {
 		io_uring_del_tctx_node(index);
+		cond_resched();
+	}
 	if (wq) {
 		/*
 		 * Must be after io_uring_del_task_file() (removes nodes under