From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Date: Fri, 14 Jun 2013 07:54:06 +0200 From: Willy Tarreau To: Ben Hutchings Cc: linux-kernel@vger.kernel.org, stable@vger.kernel.org, Paul Gortmaker , Eric Dumazet Subject: Re: Linux 2.6.32.61 Message-ID: <20130614055406.GA880@1wt.eu> References: <20130610101557.GA3235@1wt.eu> <1371183115.10688.78.camel@deadeye.wl.decadent.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1371183115.10688.78.camel@deadeye.wl.decadent.org.uk> Sender: linux-kernel-owner@vger.kernel.org List-ID: Hi Ben, On Fri, Jun 14, 2013 at 05:11:55AM +0100, Ben Hutchings wrote: > On Mon, 2013-06-10 at 12:15 +0200, Willy Tarreau wrote: > [...] > > Eric Dumazet (8): > [...] > > softirq: reduce latencies > [...] > > This one apparently introduces a regression, fixed in mainline by commit > 34376a50fb1f 'Fix lockup related to stop_machine being stuck in > __do_softirq.' > > I'm a bit surprised that you picked it for 2.6.32 as it wasn't applied > to any other branch older than 3.9. Interesting. It was discussed on stable@ in March between Eric and Paul about the pertinence of merging it into 2.6.32 and 2.6.34 (both CCed). The thread was '[PATCH] net: reduce net_rx_action() latency to 2 HZ'. It is supposed to fix a but introduced in 2.6.29 by commit 24f8b2385 (net: increase receive packet quantum). I didn't notice that it was not in more recent versions. Eric, do you suggest that I revert this patch or that I also include the mainline fix above to fix the regression ? If the latter, should Paul and Greg also take the two patches ? Thanks, Willy