Does io_uring_enter have anti-deadlock magic?
I'm using io-uring through system calls (I am *not* using liburing)
I tried to deadlock my code intentionally, but it's not allowing me, which makes me think I'm blocking incorrectly. I submitted a few commands (open, stat, reads on files, a read on a pipe), and on purpose I didn't write to the pipe.
I called 'io_uring_enter(fd, 0, 1, IORING_ENTER_SQ_WAIT, 0);' in a loop, and my logging says my submission head and tail are at 13, and my consume head and tail are at 12. Then I call enter, why is it returning without processing anything? I understand there's nothing to do because I didn't write to the pipe, but shouldn't this block?
I called io_uring_setup with 'IORING_SETUP_SQPOLL | IORING_SETUP_SINGLE_ISSUER | IORING_SETUP_NO_SQARRAY'
I'm not sure what else people may want to know, but how do I block until there's work to be consumed? Removing the IORING_SETUP_SINGLE_ISSUER doesn't help.