io_uring: assign ctx->ring_file and ctx->ring_fd early in io_uring_create()
RFC PATCH 1/1 io_uring: assign ctx->ring_file and ctx->ring_fd early in io_uring_create()
While testing the feature which supports multiple rings to share same poll thread by specifying same cpu, fio often hangs in fio_ioring_getevents() infinitely. After some investigations, I find it's because ctx->ring_file is NULL, so io_sq_thread doest not submit this ctx's requests.
While ring initialization completes, IORING_SQ_NEED_WAKEUP will be not set initially, app may already submit some reqs and the corresponding io_sq_thread may already see these reqs, but will not submit them because ring_file is NULL. Some apps just check IORING_SQ_NEED_WAKEUP beforce submitting reqs, for example, fio just checks it in fio_ioring_commit() and will not check it in fio_ioring_getevents(), then fio will hang forever. To fix this issue, we assign ctx->ring_file and ctx->ring_fd early in io_uring_create().
--- fs/io_uring.c | 5 +++++ 1 file changed, 5 insertions(+)
fs/io_uring.c +5 -0
@@ -9034,6 +9034,11 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p, goto err; } +if (p->flags &IORING_SETUP_SQPOLL ) { + ctx->ring_fd = fd; + ctx->ring_file = file; + } + ret =io_sq_offload_create (ctx, p);if (ret) goto err;
Signature · 1 line
2.17.2