If this is set, io_uring will never use an IPI to deliver a task_work
notification. This can be used in the common case where a single task or
thread communicates with the ring, and doesn't rely on
io_uring_cqe_peek().
This provides a noticeable win in performance, both from eliminating
the IPI itself, but also from avoiding interrupting the submitting
task unnecessarily.
Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
---
fs/io_uring.c | 8 +++++---
include/uapi/linux/io_uring.h | 1 +
2 files changed, 6 insertions(+), 3 deletions(-)
diff --git a/fs/io_uring.c b/fs/io_uring.c
index 20297fe4300b..43634cd5c79d 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -11352,9 +11352,10 @@ static __cold int io_uring_create(unsigned entries, struct io_uring_params *p,
ctx->user = get_uid(current_user());
/*
- * For SQPOLL, we just need a wakeup, always.
+ * For SQPOLL, we just need a wakeup, always. For !SQPOLL, if
+ * NO_RESCHED is set, then IPIs are never needed by the app.
*/
- if (ctx->flags & IORING_SETUP_SQPOLL)
+ if (ctx->flags & (IORING_SETUP_SQPOLL|IORING_SETUP_NO_RESCHED))
ctx->notify_method = TWA_SIGNAL_NO_IPI;
else
ctx->notify_method = TWA_SIGNAL;
@@ -11456,7 +11457,8 @@ static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
if (p.flags & ~(IORING_SETUP_IOPOLL | IORING_SETUP_SQPOLL |
IORING_SETUP_SQ_AFF | IORING_SETUP_CQSIZE |
IORING_SETUP_CLAMP | IORING_SETUP_ATTACH_WQ |
- IORING_SETUP_R_DISABLED | IORING_SETUP_SUBMIT_ALL))
+ IORING_SETUP_R_DISABLED | IORING_SETUP_SUBMIT_ALL |
+ IORING_SETUP_NO_RESCHED))
return -EINVAL;
return io_uring_create(entries, &p, params);
diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index 980d82eb196e..8a32230aa6f4 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -102,6 +102,7 @@ enum {
#define IORING_SETUP_ATTACH_WQ (1U << 5) /* attach to existing wq */
#define IORING_SETUP_R_DISABLED (1U << 6) /* start with ring disabled */
#define IORING_SETUP_SUBMIT_ALL (1U << 7) /* continue submit on error */
+#define IORING_SETUP_NO_RESCHED (1U << 8) /* work doesn't need resched */
enum {
IORING_OP_NOP,
--
2.35.1
