io_uring: ensure SQPOLL startup is triggered before error shutdown

syzbot reports the following hang:

INFO: task syz-executor.0:12538 can't die for more than 143 seconds.
task:syz-executor.0  state:D stack:28352 pid:12538 ppid:  8423 flags:0x00004004
Call Trace:
 context_switch kernel/sched/core.c:4324 [inline]
 __schedule+0x90c/0x21a0 kernel/sched/core.c:5075
 schedule+0xcf/0x270 kernel/sched/core.c:5154
 schedule_timeout+0x1db/0x250 kernel/time/timer.c:1868
 do_wait_for_common kernel/sched/completion.c:85 [inline]
 __wait_for_common kernel/sched/completion.c:106 [inline]
 wait_for_common kernel/sched/completion.c:117 [inline]
 wait_for_completion+0x168/0x270 kernel/sched/completion.c:138
 io_sq_thread_finish+0x96/0x580 fs/io_uring.c:7152
 io_sq_offload_create fs/io_uring.c:7929 [inline]
 io_uring_create fs/io_uring.c:9465 [inline]
 io_uring_setup+0x1fb2/0x2c20 fs/io_uring.c:9550
 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
 entry_SYSCALL_64_after_hwframe+0x44/0xae

which is due to exiting after the SQPOLL thread has been created, but
hasn't been started yet. Ensure that we always complete the startup
side when waiting for it to exit.

Reported-by: syzbot+c927c937cba8ef66dd4a@syzkaller.appspotmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

fs/io_uring.c

@@ -7141,6 +7141,7 @@ static void io_sq_thread_finish(struct io_ring_ctx *ctx)
 	struct io_sq_data *sqd = ctx->sq_data;
 
 	if (sqd) {
+		complete(&sqd->startup);
 		if (sqd->thread) {
 			wait_for_completion(&ctx->sq_thread_comp);
 			io_sq_thread_park(sqd);
@@ -7927,7 +7928,7 @@ static void io_sq_offload_start(struct io_ring_ctx *ctx)
 {
 	struct io_sq_data *sqd = ctx->sq_data;
 
-	if ((ctx->flags & IORING_SETUP_SQPOLL) && sqd->thread)
+	if (ctx->flags & IORING_SETUP_SQPOLL)
 		complete(&sqd->startup);
 }