From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from mx1.redhat.com ([209.132.183.28]) by bombadil.infradead.org with esmtps (Exim 4.80.1 #2 (Red Hat Linux)) id 1avie6-0006K2-9d for kexec@lists.infradead.org; Thu, 28 Apr 2016 09:54:06 +0000 Date: Thu, 28 Apr 2016 17:53:41 +0800 From: Baoquan He Subject: Re: [PATCH 1/3] kexec: don't invoke OOM-killer for control page allocation Message-ID: <20160428095341.GA10691@x1.redhat.com> References: <20160414195938.GV19428@n2100.arm.linux.org.uk> <20160418053253.GA3602@x1.redhat.com> <20160418083932.GA19428@n2100.arm.linux.org.uk> <20160418101214.GA6694@x1.redhat.com> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: <20160418101214.GA6694@x1.redhat.com> List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Sender: "kexec" Errors-To: kexec-bounces+dwmw2=infradead.org@lists.infradead.org To: Russell King - ARM Linux Cc: Fenghua Yu , Tony Luck , linux-ia64@vger.kernel.org, Eric Biederman , kexec@lists.infradead.org On 04/18/16 at 06:12pm, Baoquan He wrote: > On 04/18/16 at 09:39am, Russell King - ARM Linux wrote: > > On Mon, Apr 18, 2016 at 01:32:53PM +0800, Baoquan He wrote: > > > On 04/14/16 at 09:00pm, Russell King wrote: > > > > If we are unable to find a suitable page when allocating the control > > > > page, do not invoke the OOM-killer: killing processes probably isn't > > > > going to help. > > > > > > Originally kexec was made to reboot to test kernel quickly. If 1st > > > kernel is palyed and hurted in a bad state and developer want to discard > > > it and take a quick reboot, why don't we have a best try to make a > > > successful kexec load? > > > > And if it kills off every process trying to get a suitable page, > > which then means you can't do anything other than power cycle, > > that's okay? > > Yes, I agree on that it's non-sense if every process is killed. But will > each kexec load which need OOM-killer go that far? And there's only one > page (if 32 bit) or 2 pages (if 64 bit) for control page, it may not > need kill that many processes to pick one. Sorry, I was wrong. Those control pages are imtermediate pages which is used to copy kenrel and initrd from user buffer to their final place, they should be dozens of MB. So I am fine with this change, ack it. Acked-by: Baoquan He Thanks Baoquan _______________________________________________ kexec mailing list kexec@lists.infradead.org http://lists.infradead.org/mailman/listinfo/kexec