Add latest NETLINK_* constants

[rruuaanng]

Feature or enhancement

Proposal:

This issue suggests adding new netlink constants

They are usually used in the proto parameter of the socket.

e.g.

/* Kernel swap userspace event, such as device hotplugging */
socket(AF_NETLINK, SOCK_RAW, NETLINK_KOBJECT_UEVENT)
/* iSCSI protocal */
socket(AF_NETLINK, SOCK_RAW, NETLINK_ISCSI)
/* socket diagnostics, used to query status words */
socket(AF_NETLINK, SOCK_RAW, NETLINK_SOCK_DIAG)
/* used to interact with audit subsystem */
socket(AF_NETLINK, SOCK_RAW, NETLINK_AUDIT)
/* NETLINK_SOCK_DIAG's packaging*/
socket(AF_NETLINK, SOCK_RAW, NETLINK_INET_DIAG)
/* SCSI device interaction related, such as status words, events */
socket(AF_NETLINK, SOCK_RAW, NETLINK_SCSITRANSPORT)
/*  used to interact with the security-enhanced subsystem */
socket(AF_NETLINK, SOCK_RAW, NETLINK_SELINUX);

ref:
https://github.com/torvalds/linux/blob/master/include/uapi/linux/netlink.h

I'll submitte a PR.

Has this already been discussed elsewhere?

No response given

Links to previous discussion of this feature:

No response

Linked PRs

• gh-127197: Add NETLINK_* constants to socket module #127198

[erlend-aasland]

Given that it is extremely inconvenient to use raw sockets in a high-level language, compared to a low-level language, do we really need these? It sounds to me like we're adding code that will probably end up being unused.

[rruuaanng]

Maybe some users will attempt to use them, as the simplicity of high-level languages makes operations on raw socket more straightforward. e.g.

https://github.com/Cisco-Talos/mutiny-fuzzer/blob/3f347d292ebf7e4dff596f54757663346222aeb1/mutiny.py#L174
https://github.com/gsauthof/utility/blob/9cdf912fb4daf82c96394e04c95a42ae61a1f16d/ping.py#L57
https://github.com/hrvach/babyping/blob/fb7bcb70db25916fd76bf3a46cb59bd46b12a1a5/start.py#L43

Edit

When we provide these, users in need will seek them out, rather than searching and then discovering that these new constants are unavailable, only to end up complaining, "This is really disappointing."

[erlend-aasland]

That is not a likely scenario. Raw sockets programming is a niche; most users will probably use a library instead. Nobody has complained about the lack of these; check the issue tracker; there is no mention of a single one of these constants. I'm far from convinced.

[rruuaanng]

Raw sockets programming is a niche;

But users can also use non-raw sockets. And the modules that utilize these relatively new Netlink constants are mostly third-party libraries designed to work with Netlink, as you mentioned. Once CPython natively supports them, users will have the option to rely on the more portable standard library instead of using locally defined constants.

from netlink (7) man page
https://man7.org/linux/man-pages/man7/netlink.7.html

Netlink is a datagram-oriented service.  Both SOCK_RAW and
SOCK_DGRAM are valid values for socket_type.

e.g.

https://github.com/kuba-moo/meteor/blob/5234b2f89c6fbcf8e762ed308f7e478dc0b2f23c/bin/monitor_tc#L17

https://github.com/bplotnick/oomscored/blob/bbbfc0f40e44ca7ef8640682362a026a5efc1c9f/oomscoreadj.py#L175

[erlend-aasland]

I'm still not convinced. From a simple Google search, it looks to me there are several Python libraries that provide netlink APIs. There has been no requests to add these in the past. I think we'll pass on this for now. If the need arises, we can reconsider later.