protocols: init with standard SYSINIT(9) or VNET_SYSINIT
AbandonedPublic
Actions

Authored by glebius on Dec 17 2021, 6:56 PM.

Details

Reviewers

melifaro
glebius

Group Reviewers

network
transport
manpages

Commits

rG89128ff3e421: protocols: init with standard SYSINIT(9) or VNET_SYSINIT

Summary

The historical BSD network stack loop that rolls over domains and
over protocols has no advantages over more modern SYSINIT(9).
While doing the sweep, split global and per-VNET initializers.

Getting rid of pr_init allows to achieve several things:
o Get rid of ifdef's that protect against double foo_init() when

both INET and INET6 are compiled in.

o Isolate initializers statically to the module they init.
o Makes code easier to understand and maintain.

Diff Detail

Repository

rS FreeBSD src repository - subversion

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 43443
Build 40331: arc lint + arc unit

Event Timeline

glebius created this revision.Dec 17 2021, 6:56 PM

Herald added a reviewer: transport. · View Herald TranscriptDec 17 2021, 6:56 PM

Herald added subscribers: melifaro, donner, bz and 2 others. · View Herald Transcript

Harbormaster completed remote builds in B43443: Diff 100224.Dec 17 2021, 6:56 PM

glebius requested review of this revision.Dec 17 2021, 6:56 PM

glebius added a child revision: D33538: protocols: don't execute protosw_init() for every VNET.

Can I still set transport relevant sysctl variables in /etc/sysctl.conf?

In D33537#757884, @tuexen wrote:

Can I still set transport relevant sysctl variables in /etc/sysctl.conf?

I don't see how could this change affect these sysctls!

In D33537#757885, @glebius wrote:

In D33537#757884, @tuexen wrote:

Can I still set transport relevant sysctl variables in /etc/sysctl.conf?

I don't see how could this change affect these sysctls!

I was wondering if the initialisation happens at the same time...

Really good simplification!

Worth mentioning that de-facto global protocols state will be inited in SI_ORDER_THIRD instead of SI_ORDER_SECOND and per-VNET ip[6]_init moved to SI_ORDER_FOURTH.

Nit: worth updating updating domain(9) to reflect pr_init removal,

sys/netinet/tcp_subr.c
1485	This looks like a non-virtualized variable, why do we want to assign it on per-VNET basis? Also: maybe it's worth factoring out all hashsize calculation logic to a separate function?

Address Alexander's comments.

Herald added a reviewer: manpages. · View Herald TranscriptDec 27 2021, 8:07 PM

Harbormaster completed remote builds in B43612: Diff 100621.Dec 27 2021, 8:07 PM

Worth mentioning that de-facto global protocols state will be inited in SI_ORDER_THIRD instead of SI_ORDER_SECOND and per-VNET ip[6]_init moved to SI_ORDER_FOURTH.

Can't find a good place to document that.

Manual page change LGTM and is consistent with commit summary.