inpcb: retire the inpcbinfo list lock
AcceptedPublic
Actions

Authored by glebius on Thu, Mar 19, 7:19 PM.

Details

Reviewers

markj
pouria
rrs

Group Reviewers

transport
network

Summary

With the SMR locking of inpcbs the use of this lock reduced down to the
global list and generation number. It was used only on an inpcb creation
and destruction. Use the inpcbinfo hash lock for this purpose.

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Skipped

Unit

Tests Skipped

Build Status

Buildable 71551
Build 68434: arc lint + arc unit

Event Timeline

glebius created this revision.Thu, Mar 19, 7:19 PM

Herald added 1 blocking reviewer(s): transport. · View Herald TranscriptThu, Mar 19, 7:19 PM

Herald added subscribers: melifaro, imp. · View Herald Transcript

glebius requested review of this revision.Thu, Mar 19, 7:19 PM

Harbormaster completed remote builds in B71551: Diff 173951.Thu, Mar 19, 7:19 PM

glebius added a parent revision: D55661: inpcb: apply smr_advance(9)/smr_wait(9) trick only to reusable sockets.Thu, Mar 19, 7:19 PM

glebius added a child revision: D55967: inpcb: retire the inpcb global list.

Probably the lock should be renamed to something more generic, not including "hash"?

In D55966#1281009, @markj wrote:

Probably the lock should be renamed to something more generic, not including "hash"?

On the next changeset in the stack lock usage is reduced mostly to the hash, as the list of all inpcbs goes away.

The only user of in_pcbremhash() function is now only in_pcbdrop()
should we keep it?
IMHO, since the in_pcbdrop() is small enough, maybe we should expand the in_pcbremhash() and remove it entirely.

In D55966#1282878, @pouria wrote:

The only user of in_pcbremhash() function is now only in_pcbdrop()
should we keep it?
IMHO, since the in_pcbdrop() is small enough, maybe we should expand the in_pcbremhash() and remove it entirely.

I haven't yet shared the entire patch queue. The in_pcbdrop() also goes away together with INP_DROPPED and this state is internalized to TCP. This depends on D56025. Once it is in main, I will rebase my branch and share missing patches.

rrs accepted this revision.Fri, Apr 3, 1:28 PM

This revision is now accepted and ready to land.Fri, Apr 3, 1:28 PM