Socket `recvfrom` and `sendto` missing?

[ioquatix]

I see there was a patch discussed here but it does not seem to be merged?

https://www.spinics.net/lists/io-uring/msg03217.html

What was the outcome of this patch/discussion?

[ammarfaizi2]

Not sure what happened, I just dug up the old thread here https://lore.kernel.org/io-uring/[email protected]/

I think it requires Ack from David S. Miller as it touches net files.

[axboe]

We could probably get the ack, I think it mostly stalled on people saying it's just a subset of what is already supported. What is nice about recvfrom/sendto though is that less state needs to be retained to keep the submission side stable. If someone else would be interested in picking up this work, then that would be great.

[HowHsu]

We could probably get the ack, I think it mostly stalled on people saying it's just a subset of what is already supported. What is nice about recvfrom/sendto though is that less state needs to be retained to keep the submission side stable. If someone else would be interested in picking up this work, then that would be great.

Glad to look into this, a rough glance told me that it needs to be rebased and re-think about data structure.

[ammarfaizi2]

I submitted an RFC for this. I would like to get some comments.

Link: [RFC PATCH v2 0/3] io_uring: Add sendto(2) and recvfrom(2) support

[illiliti]

Any progress?

[GalaxySnail]

AFAIK sendto is available since https://github.com/torvalds/linux/commit/516e82f0e043a1a0e8d00800ed0ffe2137cf0e7e (landed in Linux 6.1), its userspace helper is io_uring_prep_send_set_addr(3).

[illiliti]

Great. Can't wait to see recvfrom support.

[serzhiio]

Any updates on recvfrom() implementation?