From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755410Ab1ATLaf (ORCPT ); Thu, 20 Jan 2011 06:30:35 -0500 Received: from mail-yi0-f46.google.com ([209.85.218.46]:55916 "EHLO mail-yi0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755056Ab1ATLae convert rfc822-to-8bit (ORCPT ); Thu, 20 Jan 2011 06:30:34 -0500 DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:sender:in-reply-to:references:date :x-google-sender-auth:message-id:subject:from:to:cc:content-type :content-transfer-encoding; b=USIzYQHIVJlpY1WIJ5PvTVWOAik5M4pUJDfjDjHxnAsEvqF8NbXBKDUQxzMNFKhJtC r3NQX41TMkZYFPCaLRmfRficfec8RFRk9LUAr+ok1gzsWgayFGW/uwJgOwkHMm4MXxoi mar1z8TN+RECTkUPbyWNGgQBI/mjQrUYJ3obY= MIME-Version: 1.0 In-Reply-To: <20110120112606.GF6036@htj.dyndns.org> References: <20110119120200.GA1057@elte.hu> <20110120110635.GB6036@htj.dyndns.org> <20110120111149.GD6036@htj.dyndns.org> <1295522631.28776.155.camel@laptop> <20110120112606.GF6036@htj.dyndns.org> Date: Thu, 20 Jan 2011 13:30:33 +0200 X-Google-Sender-Auth: 2g6iGF3SoY0enFIZCuJfg45IXr0 Message-ID: Subject: Re: [PATCH 1/2] lockdep: move early boot local IRQ enable/disable status to init/main.c From: Pekka Enberg To: Tejun Heo Cc: Peter Zijlstra , Ingo Molnar , Linus Torvalds , Linux Kernel Mailing List , Thomas Gleixner , "H. Peter Anvin" , Andrew Morton , Pekka Enberg Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8BIT Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org > On Thu, Jan 20, 2011 at 12:23:51PM +0100, Peter Zijlstra wrote: >> On Thu, 2011-01-20 at 12:11 +0100, Tejun Heo wrote: >> > Hello, guys. >> > >> > These two patches remove both warnings.  The first warning was >> > completely spurious.  The second lockdep one was triggered by >> > on_each_cpu() enabling local IRQ too early. >> > >> > I don't think percpu allocator itself malfunctioned.  There was no >> > allocation failure.  It whines when percpu allocation fails but didn't >> > in the log.  Other than calling vfree() early triggering the above >> > condition, it worked okay. >> > >> > So, it's either the early enabling of local IRQ by on_each_cpu() >> > breaking something in the IRQ subsystem or something different.  Ingo, >> > can you please see whether these two patches cure the panic too? >> >> Why not use something like: system_state != SYSTEM_RUNNING ? On Thu, Jan 20, 2011 at 1:26 PM, Tejun Heo wrote: > Cuz, it's finer grained than that.  We can add a SYSTEM_* flag but I > wasn't sure whether that would be just creating more noise.  If we try > to consolidate these things, we probably should consider > oops_in_progress too.  It probably is a good idea but let's first fix > the problem at hand. Yup, SYSTEM_RUNNING happens much later in the boot sequence so we don't want to use it here. We had a long discussion about system_state when we introduced gfp_allowed_mask and IIRC people wanted to avoid more system_state flags and preferred more explicit ones like early_boot_irqs_disabled. We can do more cleanups later but for the problem at hand: Acked-by: Pekka Enberg for both patches. Pekka