io_uring: set error code to be ENOMEM when io_alloc_async_ctx() fails
PATCH 1/1 io_uring: set error code to be ENOMEM when io_alloc_async_ctx() fails
We should return ENOMEM for memory allocation failures, fix this issue for io_alloc_async_ctx() calls.
--- fs/io_uring.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
fs/io_uring.c +3 -3
@@ -4993,7 +4993,7 @@ static int io_req_defer(struct io_kiocb *req, const struct io_uring_sqe *sqe)return 0 ;if (!req->io &&io_alloc_async_ctx (req)) -return -EAGAIN ; +return -ENOMEM ; ret =io_req_defer_prep (req, sqe);if (ret <0 )
@@ -5695,7 +5695,7 @@ static bool io_submit_sqe(struct io_kiocb *req, const struct io_uring_sqe *sqe, ctx->drain_next =1 ; }if (io_alloc_async_ctx (req)) { - ret = -EAGAIN ; + ret = -ENOMEM ; goto err_req; }
@@ -5723,7 +5723,7 @@ static bool io_submit_sqe(struct io_kiocb *req, const struct io_uring_sqe *sqe,INIT_LIST_HEAD (&req->link_list);if (io_alloc_async_ctx (req)) { - ret = -EAGAIN ; + ret = -ENOMEM ; goto err_req; } ret =io_req_defer_prep (req, sqe);
Signature · 1 line
2.17.2
Jens Axboe
maintainerQuoted 2 lines · Xiaoguang Wang
We should return ENOMEM for memory allocation failures, fix this issue for
io_alloc_async_ctx()calls.
It's not uncommon to have out-of-memory turn into -EAGAIN for the application for runtime allocations, indicating that the application could feasibly try again and hope for a better outcome (maybe after freeing memory).
The error code is also documented as such in the io_uring_enter.2 man page.
Signature · 1 line
Jens Axboe
hi,
Quoted 8 lines · Xiaoguang Wang
Quoted 2 lines
We should return ENOMEM for memory allocation failures, fix this issue for
io_alloc_async_ctx()calls.It's not uncommon to have out-of-memory turn into
-EAGAINfor the application for runtime allocations, indicating that the application could feasibly try again and hope for a better outcome (maybe after freeing memory).
OK, i see, thanks.
Quoted 3 lines
The error code is also documented as such in the io_uring_enter.2 man page.
Yeah, I checked it just now, and I should read the man page carefully again :)
Regards,
Xiaoguang Wang