Date: Wed, 20 Jul 2016 09:35:01 -0700 From: Gleb Smirnoff <glebius@FreeBSD.org> To: Randall Stewart <rrs@netflix.com> Cc: Randall Ray Stewart <rrs@FreeBSD.org>, src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r303037 - head/sys/kern Message-ID: <20160720163501.GY1076@FreeBSD.org> In-Reply-To: <1CA02E0F-6871-4291-9F5A-20EFF5F6315D@netflix.com> References: <201607191831.u6JIVJWq007542@repo.freebsd.org> <20160720065335.GU1076@FreeBSD.org> <1CA02E0F-6871-4291-9F5A-20EFF5F6315D@netflix.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Randall, I have just tested and r303037 brings the TCP panic back. I got two crashes during 2.5 hours. In your email [1] you are right, there is regression that (-1) return value is lost. This problem was worked on in the PR 210884, and we were very close to commiting the fix. The whole 11.0-RELEASE cycle strongly depends on this change. We don't want to release with TCP panic, and of course we want the regression described in 210884 to be fixed. Your backout mixed with extra code really made things messy. Since I don't want to go with commit war, on behalf of RE we are asking for explicit agreement to back out r303037. Then we will proceed with latest patch from 210884. Is that okay? [1] https://lists.freebsd.org/pipermail/svn-src-head/2016-July/089313.html On Wed, Jul 20, 2016 at 03:33:37PM +0200, Randall Stewart wrote: R> Gleb R> R> I wish you would have responded earlier.. I am more than glad to hand R> off all kern_timeout.c to you… please take it commit what you want to R> it and have it. I hate the code and I dislike having to touch it. R> R> Its yours.. I can assure you I will not touch it again. R> R> R R> > On Jul 20, 2016, at 8:53 AM, Gleb Smirnoff <glebius@freebsd.org> wrote: R> > R> > On Tue, Jul 19, 2016 at 06:31:19PM +0000, Randall Stewart wrote: R> > R> Author: rrs R> > R> Date: Tue Jul 19 18:31:19 2016 R> > R> New Revision: 303037 R> > R> URL: https://svnweb.freebsd.org/changeset/base/303037 R> > R> R> > R> Log: R> > R> This reverts out Gleb's changes and adds three small R> > R> fixes that I think closes up the races Gleb was R> > R> looking for. This is running quite nicely in Netflix and R> > R> now no longer causes TCP-tcb leaks. R> > R> R> > R> Differential Revision: 7135 R> > R> > Just to notice that I am completely pissed of by this commit R> > war, that you started. R> > R> > I've been testing my changes properly, I gave people time to R> > review my changes. You didn't. R> > R> > From your explanation in other emails I see that you've been R> > testing your changes with a version of FreeBSD that is a heavily R> > modified FreeBSD 10, not 11. R> > R> > The new code you mixed with revert of mine, doesn't fix the R> > problem observed. It fixes another problem that you imagined, R> > which might exist, but isn't observed. We already discussed that R> > and you didn't prove it wrong. R> > R> > Your change doesn't even revert my change completely. R> > R> > -- R> > Totus tuus, Glebius. R> R> -------- R> Randall Stewart R> rrs@netflix.com R> 803-317-4952 R> R> R> R> R> R> -- Totus tuus, Glebius.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20160720163501.GY1076>