Jens Axboe

> I have changed the code to use io_uring_for_each_cqe vs io_uring_peek_batch_cqe and it doesn't segfault, though its not perfect yet. That doesn't really help me, as I'd be interested in...

> Not sure why but io_uring_peek_cqe(ring, cqe) returns 4294967285 vs 0 on success and -EAGAIN on failure 4294967285 is -EAGAIN, it sounds like a casting issue on your side where...

Looking at your earlier segfaults, are you sure it isn't some random memory issues you are seeing? Eg a bug in the other code, or something like that. They don't...

> io_uring_for_each_cqe kinda works! But there are certain time where cqe doesn't get populated as well. Things like that make me very suspicious, because while the peek batch API is...

I still think this is exceedingly likely an issue in your code or interface. Given that there are demonstrable signedness issues etc, I have to say I don't have a...

You should fold it into a single patch and force push it again, it doesn't make a lot of sense to have a broken state in the upstream tree.

Apart from that one patch adding the invalidating state, looks fine to me. Maybe we just defer that decision to later and you just drop it from now? Or if...

Thanks everyone, now merged.