[PATCH v4 11/19] ublk: optimize ublk_user_copy() on daemon task

[Caleb Sander Mateos <csander@purestorage.com>]

ublk user copy syscalls may be issued from any task, so they take a
reference count on the struct ublk_io to check whether it is owned by
the ublk server and prevent a concurrent UBLK_IO_COMMIT_AND_FETCH_REQ
from completing the request. However, if the user copy syscall is issued
on the io's daemon task, a concurrent UBLK_IO_COMMIT_AND_FETCH_REQ isn't
possible, so the atomic reference count dance is unnecessary. Check for
UBLK_IO_FLAG_OWNED_BY_SRV to ensure the request is dispatched to the
sever and obtain the request from ublk_io's req field instead of looking
it up on the tagset. Skip the reference count increment and decrement.
Commit 8a8fe42d765b ("ublk: optimize UBLK_IO_REGISTER_IO_BUF on daemon
task") made an analogous optimization for ublk zero copy buffer
registration.

Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
---
 drivers/block/ublk_drv.c | 23 ++++++++++++++++++-----
 1 file changed, 18 insertions(+), 5 deletions(-)

diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
index 08674c29cfdc..707a74ab083a 100644
--- a/drivers/block/ublk_drv.c
+++ b/drivers/block/ublk_drv.c
@@ -181,11 +181,11 @@ struct ublk_io {
 	/*
 	 * The number of uses of this I/O by the ublk server
 	 * if user copy or zero copy are enabled:
 	 * - UBLK_REFCOUNT_INIT from dispatch to the server
 	 *   until UBLK_IO_COMMIT_AND_FETCH_REQ
-	 * - 1 for each inflight ublk_ch_{read,write}_iter() call
+	 * - 1 for each inflight ublk_ch_{read,write}_iter() call not on task
 	 * - 1 for each io_uring registered buffer not registered on task
 	 * The I/O can only be completed once all references are dropped.
 	 * User copy and buffer registration operations are only permitted
 	 * if the reference count is nonzero.
 	 */
@@ -2689,10 +2689,11 @@ ublk_user_copy(struct kiocb *iocb, struct iov_iter *iter, int dir)
 	struct ublk_queue *ubq;
 	struct request *req;
 	struct ublk_io *io;
 	unsigned data_len;
 	bool is_integrity;
+	bool on_daemon;
 	size_t buf_off;
 	u16 tag, q_id;
 	ssize_t ret;
 
 	if (!user_backed_iter(iter))
@@ -2718,13 +2719,24 @@ ublk_user_copy(struct kiocb *iocb, struct iov_iter *iter, int dir)
 
 	if (tag >= ub->dev_info.queue_depth)
 		return -EINVAL;
 
 	io = &ubq->ios[tag];
-	req = __ublk_check_and_get_req(ub, q_id, tag, io);
-	if (!req)
-		return -EINVAL;
+	on_daemon = current == READ_ONCE(io->task);
+	if (on_daemon) {
+		/* On daemon, io can't be completed concurrently, so skip ref */
+		if (!(io->flags & UBLK_IO_FLAG_OWNED_BY_SRV))
+			return -EINVAL;
+
+		req = io->req;
+		if (!ublk_rq_has_data(req))
+			return -EINVAL;
+	} else {
+		req = __ublk_check_and_get_req(ub, q_id, tag, io);
+		if (!req)
+			return -EINVAL;
+	}
 
 	if (is_integrity) {
 		struct blk_integrity *bi = &req->q->limits.integrity;
 
 		data_len = bio_integrity_bytes(bi, blk_rq_sectors(req));
@@ -2745,11 +2757,12 @@ ublk_user_copy(struct kiocb *iocb, struct iov_iter *iter, int dir)
 		ret = ublk_copy_user_integrity(req, buf_off, iter, dir);
 	else
 		ret = ublk_copy_user_pages(req, buf_off, iter, dir);
 
 out:
-	ublk_put_req_ref(io, req);
+	if (!on_daemon)
+		ublk_put_req_ref(io, req);
 	return ret;
 }
 
 static ssize_t ublk_ch_read_iter(struct kiocb *iocb, struct iov_iter *to)
 {
-- 
2.45.2