Added options to support socket timestamping #504
Conversation
|
One note: The timestamps are reported as control messages/ancillary data as a field in a struct cmsghdr. |
There was a problem hiding this comment.
Sorry for taking a little while @tglane to review this, but I'm afraid due to the API surface this will take some time to review.
| target_os = "vita", | ||
| target_os = "hurd", | ||
| )))) | ||
| )] |
There was a problem hiding this comment.
Since these lists are so long, can we make it "opt-in" instead of excluding basically everything.
There was a problem hiding this comment.
Also, since it's not available on all OS, we need to use the all feature.
There was a problem hiding this comment.
Yeah you are right. I will change that to reduce the overall noise in the code.
There was a problem hiding this comment.
In 08e843b I changed the cfgattributes to be opt-in. Thanks for this remark, its so much cleaner now.
| )] | ||
| pub fn timestamp_ns(&self) -> io::Result<bool> { | ||
| unsafe { | ||
| getsockopt::<c_int>(self.as_raw(), sys::SOL_SOCKET, sys::SO_TIMESTAMPNS) |
There was a problem hiding this comment.
Is there any value in having both SO_TIMESTAMPNS and SO_TIMESTAMP? Ideally we can make a wrapper that maps it to Duration regardless of the underlying option.
There was a problem hiding this comment.
I'm not sure if I understand what you are proposing here. This API should not allow the user to set timestamps or some duration. The timestamping API allows the user to receive query timestamps for when packets were send or received (depending on the flags).
With SO_TIMESTAMP you can query packet timestamps as struct timeval with microsecond accuracy while SO_TIMESTAMPNS allows for query packet timestamps as struct timespec with nanosecond accuracy. So yes, we could probably merge them into one function with an enum to specify the accuracy but we definitely need some way to set either of these.
|
@Thomasdezeeuw I just revisited this PR and thought about your comment regarding the API design. I still don't think that a API similar to What we could do a enum enum Timestamp {
ReportMs(bool), // represents SO_TIMESTAMP
ReportNs(bool), // represents SO_TIMESTAMPNS
}But I would argue that the I`d like to hear your opinion on that one. |
This PR adds the options for a socket to generate and report timestamps when a packet was received by the hardware/kernel.
The platforms supported by
socket2vary in the support for the timestamping options but I tried to implemented this as platform independent as possible. Unfortunately all thecfgattributes produce a lot of noise in the code.For documentation check:
Closes #415