Date: Sat, 10 Jun 2000 03:25:43 -0400 (EDT) From: Bosko Milekic <bmilekic@dsuper.net> To: Mike Silbersack <silby@silby.com> Cc: hackers@FreeBSD.ORG, jlemon@freebsd.org Subject: Re: Mbuf waiting mfc to 3 Message-ID: <Pine.BSF.4.21.0006100306330.23301-200000@jehovah.technokratis.com> In-Reply-To: <Pine.BSF.4.21.0006092012390.13026-100000@achilles.silby.com>
next in thread | previous in thread | raw e-mail | index | archive | help
[-- Attachment #1 --] [re-directed to --hackers, as more appropriate there, also jlemon@freebsd.org is in the CC, make _SURE_ to remove him from there before you post ANY replies!!!] Mike, your patch looks fine. However, I found a bug in /sys/netkey code. (and it's related to the wait stuff, although I don't believe it directly concerns your patch so, as far as I'm concerned, your stuff is ready to go in.) However, I believe this code is only for 4.x and -CURRENT people. keysock.c's key_sendup() does a silly thing with the mbuf allocation. Attached is a patch that fixes it. This applies to v 1.2 of the file. here is the Id: /* KAME @(#)$Id: keysock.c,v 1.2 1999/08/16 19:30:36 shin Exp $ */ (Jlemon, can you commit this?) Oh yeah, and please also commit pr=18471 as it's been sitting there for a while. Thanks in advance, Bosko. On Fri, 9 Jun 2000, Mike Silbersack wrote: > Well, it's been nearly a month since I posted the mbuf waiting MFC for 3.4 > to -net, although I haven't heard any complaints about it messing up > systems, there have been a few complaints on bugtraq of mbuf > exhaustion attacks which would be much less serious with it. :) > > In any case, the patch is still available at > http://www.silby.com/patches/mbuf-wait-mfc-2.patch for review. I'm > fairly confident in its reliability, but I'd prefer a few more people to > test it if they have the time. If there are no negative complaints, I'd > like to get it committed before the end of next week to ensure that we > don't miss getting it into 3.5. > > There are no changes between this patch and the last one I posted other > than a single version line I had messed up in the previous one, so if > you're currently testing that one, there's no need to download this > one. Please post your experiences with it in any case, though. > > The small memory leak I alluded to in my previous posting of the patch has > been found and committed seperately (as it affected 3,4, and 5.) So, > please CVSUP before testing this patch to ensure you're seeing its true > colors. > > Thanks, > > Mike "Silby" Silbersack -- Bosko Milekic * Voice/Mobile: 514.865.7738 * Pager: 514.921.0237 bmilekic@technokratis.com * http://www.technokratis.com/ [-- Attachment #2 --] --- keysock.old.c Sat Jun 10 03:09:05 2000 +++ keysock.c Sat Jun 10 03:13:43 2000 @@ -419,18 +419,25 @@ while (tlen > 0) { if (tlen == len) { MGETHDR(n, M_DONTWAIT, MT_DATA); + if (n == NULL) { + if (m) m_freem(m); + return ENOBUFS; + } n->m_len = MHLEN; } else { MGET(n, M_DONTWAIT, MT_DATA); + if (n == NULL) { + if (m) m_freem(m); + return ENOBUFS; + } n->m_len = MLEN; } - if (!n) - return ENOBUFS; + if (tlen > MCLBYTES) { /*XXX better threshold? */ MCLGET(n, M_DONTWAIT); if ((n->m_flags & M_EXT) == 0) { m_free(n); - m_freem(m); + if (m) m_freem(m); return ENOBUFS; } n->m_len = MCLBYTES;
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0006100306330.23301-200000>