From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753394AbdHPCPz (ORCPT ); Tue, 15 Aug 2017 22:15:55 -0400 Received: from mga02.intel.com ([134.134.136.20]:40148 "EHLO mga02.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752382AbdHPCPy (ORCPT ); Tue, 15 Aug 2017 22:15:54 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.41,380,1498546800"; d="scan'208";a="1163041537" Subject: Re: [PATCH v3 5/5] usb: xhci: Handle USB transaction error on address command To: Mathias Nyman References: <1502419296-8212-1-git-send-email-baolu.lu@linux.intel.com> <1502419296-8212-6-git-send-email-baolu.lu@linux.intel.com> <5992DB45.80803@linux.intel.com> Cc: linux-usb@vger.kernel.org, linux-kernel@vger.kernel.org, zhengjun.xing@linux.intel.com From: Lu Baolu Message-ID: <5993AAD8.4060106@linux.intel.com> Date: Wed, 16 Aug 2017 10:15:52 +0800 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.5.1 MIME-Version: 1.0 In-Reply-To: <5992DB45.80803@linux.intel.com> Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi, On 08/15/2017 07:30 PM, Mathias Nyman wrote: > On 11.08.2017 05:41, Lu Baolu wrote: >> Xhci driver handles USB transaction errors on transfer events, >> but transaction errors are possible on address device command >> completion events as well. >> >> The xHCI specification (section 4.6.5) says: A USB Transaction >> Error Completion Code for an Address Device Command may be due >> to a Stall response from a device. Software should issue a Disable >> Slot Command for the Device Slot then an Enable Slot Command to >> recover from this error. >> >> This patch handles USB transaction errors on address command >> completion events. The related discussion threads can be found >> through below links. >> >> http://marc.info/?l=linux-usb&m=149362010728921&w=2 >> http://marc.info/?l=linux-usb&m=149252752825755&w=2 >> >> Suggested-by: Mathias Nyman >> Signed-off-by: Lu Baolu >> --- >> drivers/usb/host/xhci.c | 5 +++++ >> 1 file changed, 5 insertions(+) >> >> diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c >> index d6b728d..95780f8 100644 >> --- a/drivers/usb/host/xhci.c >> +++ b/drivers/usb/host/xhci.c >> @@ -3822,6 +3822,11 @@ static int xhci_setup_device(struct usb_hcd *hcd, struct usb_device *udev, >> break; >> case COMP_USB_TRANSACTION_ERROR: >> dev_warn(&udev->dev, "Device not responding to setup %s.\n", act); >> + >> + ret = xhci_disable_slot(xhci, udev->slot_id); >> + if (!ret) >> + xhci_alloc_dev(hcd, udev); > > Might be a xhci->mutex locking issue here, > both xhci_setup_device() and xhci_alloc_dev() take xhci->mutex > I will apply xhci->mutex in this patch for code consistency, but I think we can drop xhci->mutex (in a separated patch) anyway. xhci->mutex was introduced to protect two race sources of xhci->slot_id and xhci->addr_dev by below commit: commit a00918d0521df1c7a2ec9143142a3ea998c8526d Author: Chris Bainbridge Date: Tue May 19 16:30:51 2015 +0300 usb: host: xhci: add mutex for non-thread-safe data Regression in commit 638139eb95d2 ("usb: hub: allow to process more usb hub events in parallel") The regression resulted in intermittent failure to initialise a 10-port hub (with three internal VL812 4-port hub controllers) on boot, with a failure rate of around 8%, due to multiple race conditions when accessing addr_dev and slot_id in struct xhci_hcd. This regression also exposed a problem with xhci_setup_device, which "should be protected by the usb_address0_mutex" but no longer is due to commit 6fecd4f2a58c ("USB: separate usb_address0 mutexes for each bus") With separate buses (and locks) it is no longer the case that a single lock will protect xhci_setup_device from accesses by two parallel threads processing events on the two buses. Fix this by adding a mutex to protect addr_dev and slot_id in struct xhci_hcd, and by making the assignment of slot_id atomic. [--cut---] We have already removed these two race sources after that by below two commits: c2d3d49 usb: xhci: move slot_id from xhci_hcd to xhci_command structure 87e44f2 usb: xhci: remove the use of xhci->addr_dev So we don't need xhci->mutex any more. I will try to do this in a separated patch with more tests. For now, I will add xhci->mutex for code consistency. Best regards, Lu Baolu