From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mx1.redhat.com ([209.132.183.28]:50501 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755016AbbCLXVL (ORCPT ); Thu, 12 Mar 2015 19:21:11 -0400 Date: Fri, 13 Mar 2015 07:21:04 +0800 From: Fam Zheng To: Paolo Bonzini Cc: linux-kernel@vger.kernel.org, Bjorn Helgaas , linux-pci@vger.kernel.org, Linux Virtualization , mst@redhat.com Subject: Re: [RFC PATCH] PCI: Disable MSI/MSI-X only if device is shutdown Message-ID: <20150312232104.GA32054@ad.nay.redhat.com> References: <1426137682-12287-1-git-send-email-famz@redhat.com> <5501BCF7.9080004@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <5501BCF7.9080004@redhat.com> Sender: linux-pci-owner@vger.kernel.org List-ID: On Thu, 03/12 17:21, Paolo Bonzini wrote: > On 12/03/2015 06:21, Fam Zheng wrote: > > If the device doesn't support shutdown, disabling interrupts may cause > > trouble. For example, virtio-scsi-pci doesn't implement shutdown, and > > after we disable MSI-X, futher notifications from device will be > > delivered to IRQ, which is unexpected. This IRQ will not be cleared, and > > may prevent us from making progress, by keep triggering interrupts. > > > > Signed-off-by: Fam Zheng > > --- > > drivers/pci/pci-driver.c | 7 ++++--- > > 1 file changed, 4 insertions(+), 3 deletions(-) > > > > diff --git a/drivers/pci/pci-driver.c b/drivers/pci/pci-driver.c > > index 3cb2210..fb29c96 100644 > > --- a/drivers/pci/pci-driver.c > > +++ b/drivers/pci/pci-driver.c > > @@ -448,10 +448,11 @@ static void pci_device_shutdown(struct device *dev) > > > > pm_runtime_resume(dev); > > > > - if (drv && drv->shutdown) > > + if (drv && drv->shutdown) { > > drv->shutdown(pci_dev); > > - pci_msi_shutdown(pci_dev); > > - pci_msix_shutdown(pci_dev); > > + pci_msi_shutdown(pci_dev); > > + pci_msix_shutdown(pci_dev); > > + } > > > > #ifdef CONFIG_KEXEC > > /* > > > > The patch may be okay, but I think the bug here is also that virtio-pci > is not defining a .shutdown callback. It should define one and call > free_irq (for INTX) and pci_disable_msix. It's not enough. The device has to know we disabled msix, otherwise it will send us IRQ, which is wrong. > > How is this related to the virtio-scsi patch that you posted? Do you > need both to fix the problem you reported? > This one should be enough. I was wrong in saying virtio-scsi hanging the shutdown is because requests not being completed, it's more because of the unexpected IRQ as explained in [1]. [1]: https://bugzilla.redhat.com/attachment.cgi?id=998391 Fam