From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Return-Path: Date: Wed, 9 Nov 2016 14:54:59 -0500 From: Keith Busch To: Prarit Bhargava Cc: Bjorn Helgaas , linux-pci@vger.kernel.org, alex.williamson@redhat.com, darcari@redhat.com, mstowe@redhat.com, bhelgaas@google.com, lukas@wunner.de, mika.westerberg@linux.intel.com Subject: Re: [PATCH] pci: Only disable MSI/X and enable INTx if shutdown function has been called Message-ID: <20161109195458.GA3409@localhost.localdomain> References: <1478627867-28795-1-git-send-email-prarit@redhat.com> <20161109170529.GJ14322@bhelgaas-glaptop.roam.corp.google.com> <58237AB7.8080301@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <58237AB7.8080301@redhat.com> List-ID: On Wed, Nov 09, 2016 at 02:36:23PM -0500, Prarit Bhargava wrote: > > > On 11/09/2016 12:05 PM, Bjorn Helgaas wrote: > > Hi Prarit, > > > > Is there a bugzilla or other archive of configuration/dmesg/other info > > [I have only added Bjorn and myself to the BZ below. Please feel free to add > yourself.] > > Bjorn, unfortunately this won't be caught in a dmesg log because the filesystem > is unmounted by the time we shutdown the PCI devices in the halt/reboot path. > > The trace is only available from serial console, and I have opened up > > https://bugzilla.kernel.org/show_bug.cgi?id=187351 > > to track this. > > I have added some additional links to other bugzillas which seem to show the > same behavior, and included a full serial console capture of the boot to the BZ. > The end of the log shows the unhandled irq stack trace for irq 16. Just researching where this came from, the behavior to shutdown msi/msix and enable intx was done in commit d52877c7b1 for some kexec issue, and I think that was wrong in the first place. We shouldn't be changing interrupt configuration out from under the drivers.