cvs commit: src/sys/netinet tcp_syncache.c

[Peter Wemm]

Mike Silbersack wrote:
> 
> On Mon, 10 Nov 2003, M. Warner Losh wrote:
> 
> > In message: <200311102359.hAANx2xu023566 at mta7.pltn13.pbi.net>
> >             Jeffrey Hsu <hsu at FreeBSD.org> writes:
> > : Are you kidding me?  Or do you really not understand why it's not
> > : Giant-free ready yet?
> >
> > While it may be obvious to you, it seems that a sentence or two about
> > how it isn't giant-free would be useful.
> >
> > Warner
> 
> If I'm not mistaken, syncache_timer calls syncache_respond, which then
> goes and interacts with the routing table and ip stack.  Along with the
> rest of the non-locked code, it sure looks like it's not-MPSAFE to me.
> 
> It seems "obvious" that the code isn't giant-free, given that there are no
> locking assertions in it whatsoever.

Sure, but that's what the commit message should have said.  eg:

"Revert rev 1.40.  Mark TCP syncache timer as not Giant-free ready yet.
 It still calls syncache_respond which interacts with the routing table
 and ip stack which is not locked yet."

It isn't that hard.

----------------------------
revision 1.40
date: 2003/07/17 11:19:25;  author: hsu;  state: Exp;  lines: +1 -1
Drop Giant around syncache timer processing.
----------------------------
revision 1.45
date: 2003/11/10 20:42:04;  author: hsu;  state: Exp;  lines: +1 -1
Mark TCP syncache timer as not Giant-free ready yet.
----------------------------

Considering the history, an explanation in the commit log is in order. Both
commit messages have plenty of conviction but there is no explanation of
the contradiction.

Cheers,
-Peter
--
Peter Wemm - peter at wemm.org; peter at FreeBSD.org; peter at yahoo-inc.com
"All of this is for nothing if we don't go to the stars" - JMS/B5