Merge tag 'io_uring-6.0-2022-08-26' of git://git.kernel.dk/linux-block

Pull io_uring fixes from Jens Axboe:

 - Add missing header file to the MAINTAINERS entry for io_uring (Ammar)

 - liburing and the kernel ship the same io_uring.h header, but one
   change we've had for a long time only in liburing is to ensure it's
   C++ safe. Add extern C around it, so we can more easily sync them in
   the future (Ammar)

 - Fix an off-by-one in the sync cancel added in this merge window (me)

 - Error handling fix for passthrough (Kanchan)

 - Fix for address saving for async execution for the zc tx support
   (Pavel)

 - Fix ordering for TCP zc notifications, so we always have them ordered
   correctly between "data was sent" and "data was acked". This isn't
   strictly needed with the notification slots, but we've been pondering
   disabling the slot support for 6.0 - and if we do, then we do require
   the ordering to be sane. Regardless of that, it's the sane thing to
   do in terms of API (Pavel)

 - Minor cleanup for indentation and lockdep annotation (Pavel)

* tag 'io_uring-6.0-2022-08-26' of git://git.kernel.dk/linux-block:
  io_uring/net: save address for sendzc async execution
  io_uring: conditional ->async_data allocation
  io_uring/notif: order notif vs send CQEs
  io_uring/net: fix indentation
  io_uring/net: fix zc send link failing
  io_uring/net: fix must_hold annotation
  io_uring: fix submission-failure handling for uring-cmd
  io_uring: fix off-by-one in sync cancelation file check
  io_uring: uapi: Add `extern "C"` in io_uring.h for liburing
  MAINTAINERS: Add `include/linux/io_uring_types.h`

10 files changed, 74 insertions, 17 deletions

diff --git a/MAINTAINERS b/MAINTAINERS
index af4848466a08..07403024f3f2 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -10659,6 +10659,7 @@ T:	git git://git.kernel.dk/linux-block
 T:	git git://git.kernel.dk/liburing
 F:	io_uring/
 F:	include/linux/io_uring.h
+F:	include/linux/io_uring_types.h
 F:	include/uapi/linux/io_uring.h
 F:	tools/io_uring/
 
diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index 1463cfecb56b..9e0b5c8d92ce 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -12,6 +12,10 @@
 #include <linux/types.h>
 #include <linux/time_types.h>
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+
 /*
  * IO submission data structure (Submission Queue Entry)
  */
@@ -661,4 +665,8 @@ struct io_uring_recvmsg_out {
 	__u32 flags;
 };
 
+#ifdef __cplusplus
+}
+#endif
+
 #endif
diff --git a/io_uring/cancel.c b/io_uring/cancel.c
index e4e1dc0325f0..5fc5d3e80fcb 100644
--- a/io_uring/cancel.c
+++ b/io_uring/cancel.c
@@ -218,7 +218,7 @@ static int __io_sync_cancel(struct io_uring_task *tctx,
 	    (cd->flags & IORING_ASYNC_CANCEL_FD_FIXED)) {
 		unsigned long file_ptr;
 
-		if (unlikely(fd > ctx->nr_user_files))
+		if (unlikely(fd >= ctx->nr_user_files))
 			return -EBADF;
 		fd = array_index_nospec(fd, ctx->nr_user_files);
 		file_ptr = io_fixed_file_slot(&ctx->file_table, fd)->file_ptr;
diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index ebfdb2212ec2..77616279000b 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -1450,9 +1450,10 @@ int io_req_prep_async(struct io_kiocb *req)
 		return 0;
 	if (WARN_ON_ONCE(req_has_async_data(req)))
 		return -EFAULT;
-	if (io_alloc_async_data(req))
-		return -EAGAIN;
-
+	if (!io_op_defs[req->opcode].manual_alloc) {
+		if (io_alloc_async_data(req))
+			return -EAGAIN;
+	}
 	return def->prep_async(req);
 }
 
diff --git a/io_uring/net.c b/io_uring/net.c
index f8cdf1dc3863..0af8a02df580 100644
--- a/io_uring/net.c
+++ b/io_uring/net.c
@@ -182,6 +182,37 @@ static int io_sendmsg_copy_hdr(struct io_kiocb *req,
 					&iomsg->free_iov);
 }
 
+int io_sendzc_prep_async(struct io_kiocb *req)
+{
+	struct io_sendzc *zc = io_kiocb_to_cmd(req, struct io_sendzc);
+	struct io_async_msghdr *io;
+	int ret;
+
+	if (!zc->addr || req_has_async_data(req))
+		return 0;
+	if (io_alloc_async_data(req))
+		return -ENOMEM;
+
+	io = req->async_data;
+	ret = move_addr_to_kernel(zc->addr, zc->addr_len, &io->addr);
+	return ret;
+}
+
+static int io_setup_async_addr(struct io_kiocb *req,
+			      struct sockaddr_storage *addr,
+			      unsigned int issue_flags)
+{
+	struct io_async_msghdr *io;
+
+	if (!addr || req_has_async_data(req))
+		return -EAGAIN;
+	if (io_alloc_async_data(req))
+		return -ENOMEM;
+	io = req->async_data;
+	memcpy(&io->addr, addr, sizeof(io->addr));
+	return -EAGAIN;
+}
+
 int io_sendmsg_prep_async(struct io_kiocb *req)
 {
 	int ret;
@@ -944,7 +975,7 @@ static int io_sg_from_iter(struct sock *sk, struct sk_buff *skb,
 
 int io_sendzc(struct io_kiocb *req, unsigned int issue_flags)
 {
-	struct sockaddr_storage address;
+	struct sockaddr_storage __address, *addr = NULL;
 	struct io_ring_ctx *ctx = req->ctx;
 	struct io_sendzc *zc = io_kiocb_to_cmd(req, struct io_sendzc);
 	struct io_notif_slot *notif_slot;
@@ -978,10 +1009,17 @@ int io_sendzc(struct io_kiocb *req, unsigned int issue_flags)
 	msg.msg_namelen = 0;
 
 	if (zc->addr) {
-		ret = move_addr_to_kernel(zc->addr, zc->addr_len, &address);
-		if (unlikely(ret < 0))
-			return ret;
-		msg.msg_name = (struct sockaddr *)&address;
+		if (req_has_async_data(req)) {
+			struct io_async_msghdr *io = req->async_data;
+
+			msg.msg_name = addr = &io->addr;
+		} else {
+			ret = move_addr_to_kernel(zc->addr, zc->addr_len, &__address);
+			if (unlikely(ret < 0))
+				return ret;
+			msg.msg_name = (struct sockaddr *)&__address;
+			addr = &__address;
+		}
 		msg.msg_namelen = zc->addr_len;
 	}
 
@@ -989,7 +1027,7 @@ int io_sendzc(struct io_kiocb *req, unsigned int issue_flags)
 		ret = io_import_fixed(WRITE, &msg.msg_iter, req->imu,
 					(u64)(uintptr_t)zc->buf, zc->len);
 		if (unlikely(ret))
-				return ret;
+			return ret;
 	} else {
 		ret = import_single_range(WRITE, zc->buf, zc->len, &iov,
 					  &msg.msg_iter);
@@ -1013,16 +1051,18 @@ int io_sendzc(struct io_kiocb *req, unsigned int issue_flags)
 
 	if (unlikely(ret < min_ret)) {
 		if (ret == -EAGAIN && (issue_flags & IO_URING_F_NONBLOCK))
-			return -EAGAIN;
+			return io_setup_async_addr(req, addr, issue_flags);
+
 		if (ret > 0 && io_net_retry(sock, msg.msg_flags)) {
 			zc->len -= ret;
 			zc->buf += ret;
 			zc->done_io += ret;
 			req->flags |= REQ_F_PARTIAL_IO;
-			return -EAGAIN;
+			return io_setup_async_addr(req, addr, issue_flags);
 		}
 		if (ret == -ERESTARTSYS)
 			ret = -EINTR;
+		req_set_fail(req);
 	} else if (zc->flags & IORING_RECVSEND_NOTIF_FLUSH) {
 		io_notif_slot_flush_submit(notif_slot, 0);
 	}
diff --git a/io_uring/net.h b/io_uring/net.h
index 7c438d39c089..f91f56c6eeac 100644
--- a/io_uring/net.h
+++ b/io_uring/net.h
@@ -31,6 +31,7 @@ struct io_async_connect {
 int io_shutdown_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
 int io_shutdown(struct io_kiocb *req, unsigned int issue_flags);
 
+int io_sendzc_prep_async(struct io_kiocb *req);
 int io_sendmsg_prep_async(struct io_kiocb *req);
 void io_sendmsg_recvmsg_cleanup(struct io_kiocb *req);
 int io_sendmsg_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
diff --git a/io_uring/notif.c b/io_uring/notif.c
index 977736e82c1a..96f076b175e0 100644
--- a/io_uring/notif.c
+++ b/io_uring/notif.c
@@ -73,7 +73,7 @@ struct io_kiocb *io_alloc_notif(struct io_ring_ctx *ctx,
 }
 
 void io_notif_slot_flush(struct io_notif_slot *slot)
-	__must_hold(&ctx->uring_lock)
+	__must_hold(&slot->notif->ctx->uring_lock)
 {
 	struct io_kiocb *notif = slot->notif;
 	struct io_notif_data *nd = io_notif_to_data(notif);
@@ -81,8 +81,10 @@ void io_notif_slot_flush(struct io_notif_slot *slot)
 	slot->notif = NULL;
 
 	/* drop slot's master ref */
-	if (refcount_dec_and_test(&nd->uarg.refcnt))
-		io_notif_complete(notif);
+	if (refcount_dec_and_test(&nd->uarg.refcnt)) {
+		notif->io_task_work.func = __io_notif_complete_tw;
+		io_req_task_work_add(notif);
+	}
 }
 
 __cold int io_notif_unregister(struct io_ring_ctx *ctx)
diff --git a/io_uring/opdef.c b/io_uring/opdef.c
index 72dd2b2d8a9d..41410126c1c6 100644
--- a/io_uring/opdef.c
+++ b/io_uring/opdef.c
@@ -478,13 +478,15 @@ const struct io_op_def io_op_defs[] = {
 		.pollout		= 1,
 		.audit_skip		= 1,
 		.ioprio			= 1,
+		.manual_alloc		= 1,
 #if defined(CONFIG_NET)
+		.async_size		= sizeof(struct io_async_msghdr),
 		.prep			= io_sendzc_prep,
 		.issue			= io_sendzc,
+		.prep_async		= io_sendzc_prep_async,
 #else
 		.prep			= io_eopnotsupp_prep,
 #endif
-
 	},
 };
 
diff --git a/io_uring/opdef.h b/io_uring/opdef.h
index ece8ed4f96c4..763c6e54e2ee 100644
--- a/io_uring/opdef.h
+++ b/io_uring/opdef.h
@@ -25,6 +25,8 @@ struct io_op_def {
 	unsigned		ioprio : 1;
 	/* supports iopoll */
 	unsigned		iopoll : 1;
+	/* opcode specific path will handle ->async_data allocation if needed */
+	unsigned		manual_alloc : 1;
 	/* size of async data needed, if any */
 	unsigned short		async_size;
 
diff --git a/io_uring/uring_cmd.c b/io_uring/uring_cmd.c
index 8e0cc2d9205e..b9989ae7b957 100644
--- a/io_uring/uring_cmd.c
+++ b/io_uring/uring_cmd.c
@@ -112,7 +112,7 @@ int io_uring_cmd(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 ret;
 	}
 
 	return IOU_ISSUE_SKIP_COMPLETE;