svn commit: r280759 - head/sys/netinet

Gleb Smirnoff glebius at FreeBSD.org
Sun Mar 29 21:21:48 UTC 2015


On Sun, Mar 29, 2015 at 02:15:43PM -0700, Adrian Chadd wrote:
A> > To illustrate the probability of 1 instruction race, I would suggest to
A> > look at PCPU_ADD() implementation for all arches except amd64/i386 and
A> > find out that it is prone to this race. Hitting the race in PCPU_ADD()
A> > will usually lead to mangling vm_meter statistics. Did we ever observe
A> > this in practice? No.
A> 
A> I dunno. Has anyone actually looked at that? :)
A> 
A> On a busy system you may find those small race windows exist enough to
A> be annoying. Case in point - the recent callout code race. The race
A> between "read" "modify" "write" was tiny, but enough to break things.

The difference is that in callout it lead to an endless loop, and in
discussed case it leads to IP ID reuse, which already is happening all
the time.

-- 
Totus tuus, Glebius.


More information about the svn-src-head mailing list