From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-wi0-f178.google.com ([209.85.212.178]:36067 "EHLO mail-wi0-f178.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751428AbbFXUqW (ORCPT ); Wed, 24 Jun 2015 16:46:22 -0400 Received: by wicnd19 with SMTP id nd19so146067718wic.1 for ; Wed, 24 Jun 2015 13:46:21 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: References: <20150619223325.14626.66402.stgit@bhelgaas-glaptop2.roam.corp.google.com> <20150619224243.14626.59367.stgit@bhelgaas-glaptop2.roam.corp.google.com> From: Bjorn Helgaas Date: Wed, 24 Jun 2015 15:46:02 -0500 Message-ID: Subject: Re: [PATCH 4/6] PCI: Fix Intel generic reset quirk class code check To: "linux-pci@vger.kernel.org" Cc: "x86@kernel.org" , Felipe Balbi , Yu Zhao , Huang Rui , =?UTF-8?Q?Krzysztof_Ha=C5=82asa?= , Jason Chang , Mathias Nyman Content-Type: text/plain; charset=UTF-8 Sender: linux-pci-owner@vger.kernel.org List-ID: On Fri, Jun 19, 2015 at 6:06 PM, Bjorn Helgaas wrote: > [-cc Dexuan, Matthew (bounced)] > [+cc Mathias, since MAINTAINERS mentions USB, Intel, and you in one entry :)] > > On Fri, Jun 19, 2015 at 5:42 PM, Bjorn Helgaas wrote: >> We checked for "dev->class == PCI_CLASS_SERIAL_USB", but dev->class >> contains the entire three-byte base class/sub-class/interface, while >> PCI_CLASS_SERIAL_USB is only the two-byte base class/sub-class. >> >> This error meant that we used the Intel device-specific reset on devices >> with class code 0x000c03 instead of those with class code 0x0c03xx. >> 0x000c03 is a reserved value in the 0x00 backwards compatibility base class >> and shouldn't match any devices, so I think reset_intel_generic_dev() >> always failed. >> >> Shift dev->class to discard the interface byte before comparing with >> PCI_CLASS_SERIAL_USB. >> >> Fixes: aeb30016fec3 ("PCI: add Intel USB specific reset method") >> Signed-off-by: Bjorn Helgaas >> CC: Dexuan Cui >> CC: Yu Zhao >> --- >> drivers/pci/quirks.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c >> index 8bc60c2..356a401 100644 >> --- a/drivers/pci/quirks.c >> +++ b/drivers/pci/quirks.c >> @@ -3358,7 +3358,7 @@ static int reset_intel_generic_dev(struct pci_dev *dev, int probe) >> int pos; >> >> /* only implement PCI_CLASS_SERIAL_USB at present */ >> - if (dev->class == PCI_CLASS_SERIAL_USB) { >> + if ((dev->class >> 8) == PCI_CLASS_SERIAL_USB) { > > I'm a little bit concerned about this change because it *looks* like > this has always been broken, but I assume it was tested and somebody > would have noticed. Unless we get confirmation that somebody cares about reset_intel_generic_dev() and it has been tested, I'm inclined to revert aeb30016fec3 ("PCI: add Intel USB specific reset method") instead of applying this fix. I think reset_intel_generic_dev() always returns -ENOTTY except for devices with "class == 0x000c03", which is an invalid class code. If I "fix" it by putting in the shift, it will start poking at config space on Intel USB devices. I have no way of testing that, and I haven't seen a problem report related to this, so a change is just as likely to break something as to fix something. So the safest thing looks like just dropping the quirk altogether, which shouldn't change behavior at all. >> pos = pci_find_capability(dev, PCI_CAP_ID_VNDR); >> if (!pos) >> return -ENOTTY; >>