Attention is currently required from: jolly, laforge.
1 comment:
Patchset:
this was based on a discussion aeversberg and I had before he started to wrok on the patch. […]
I foresee lots of weird stuff happening only in those "super-rare events" we may not think about or which may end up with undefined behavior, eg. with one socket ending up with sqes in 2 different rings.
IMHO we should explore the possibility of, in case io_uring_get_sqe() fails, append the osmo_iofd to some sort of "wait_read_sqe_queue" which is re-attempted every time a completed() event happen on the queue.
This sounds like a lot less trouble now and in the future.
To view, visit change 40725. To unsubscribe, or for help writing mail filters, visit settings.