From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755739Ab3LDVYU (ORCPT ); Wed, 4 Dec 2013 16:24:20 -0500 Received: from mail-wg0-f44.google.com ([74.125.82.44]:61018 "EHLO mail-wg0-f44.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754084Ab3LDVYR (ORCPT ); Wed, 4 Dec 2013 16:24:17 -0500 Message-ID: <529F9D89.9000504@linaro.org> Date: Wed, 04 Dec 2013 22:24:25 +0100 From: Daniel Lezcano User-Agent: Mozilla/5.0 (X11; Linux i686; rv:24.0) Gecko/20100101 Thunderbird/24.1.0 MIME-Version: 1.0 To: Konrad Rzeszutek Wilk CC: "Rafael J. Wysocki" , linux-pm@vger.kernel.org, linux-kernel@vger.kernel.org, xen-devel@lists.xenproject.org Subject: Re: [PATCH] cpuidle: Check for dev before deregistering it. References: <1386086398-3686-1-git-send-email-konrad.wilk@oracle.com> <3652709.SMvyaH5n9G@vostro.rjw.lan> <529EF184.3090000@linaro.org> <20131204160911.GB391@pegasus.dumpdata.com> In-Reply-To: <20131204160911.GB391@pegasus.dumpdata.com> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 12/04/2013 05:09 PM, Konrad Rzeszutek Wilk wrote: > On Wed, Dec 04, 2013 at 10:10:28AM +0100, Daniel Lezcano wrote: >> On 12/03/2013 10:33 PM, Rafael J. Wysocki wrote: >>> On Tuesday, December 03, 2013 10:59:58 AM Konrad Rzeszutek Wilk wrote: >>>> If not, we could end up in the unfortunate situation where >>>> we dereference a NULL pointer b/c we have cpuidle disabled. >>>> >>>> This is the case when booting under Xen (which uses the >>>> ACPI P/C states but disables the CPU idle driver) - and can >>>> be easily reproduced when booting with cpuidle.off=1. >>>> >>>> BUG: unable to handle kernel NULL pointer dereference at (null) >>>> IP: [] cpuidle_unregister_device+0x2a/0x90 >>>> .. snip.. >>>> Call Trace: >>>> [] acpi_processor_power_exit+0x3c/0x5c >>>> [] acpi_processor_stop+0x61/0xb6 >>>> [] __device_release_driver+0fffff81421653>] device_release_driver+0x23/0x30 >>>> [] bus_remove_device+0x108/0x180 >>>> [] device_del+0x129/0x1c0 >>>> [] ? unregister_xenbus_watch+0x1f0/0x1f0 >>>> [] device_unregister+0x1e/0x60 >>>> [] unregister_cpu+0x39/0x60 >>>> [] arch_unregister_cpu+0x23/0x30 >>>> [] handle_vcpu_hotplug_event+0xc1/0xe0 >>>> [] xenwatch_thread+0x45/0x120 >>>> [] ? abort_exclusive_wait+0xb0/0xb0 >>>> [] kthread+0xd2/0xf0 >>>> [] ? kthread_create_on_node+0x180/0x180 >>>> [] ret_from_fork+0x7c/0xb0 >>>> [] ? kthread_create_on_node+0x180/0x180 >>>> >>>> This problem also appears in 3.12 and could be a candidate for backport. >>>> >>>> CC: "Rafael J. Wysocki" >>>> CC: Daniel Lezcano >>>> CC: linux-pm@vger.kernel.org >>>> Signed-off-by: Konrad Rzeszutek Wilk >>> >>> Applied, thanks! >>> >>>> --- >>>> drivers/cpuidle/cpuidle.c | 2 +- >>>> 1 file changed, 1 insertion(+), 1 deletion(-) >>>> >>>> diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c >>>> index 2a991e4..a55e68f 100644 >>>> --- a/drivers/cpuidle/cpuidle.c >>>> +++ b/drivers/cpuidle/cpuidle.c >>>> @@ -400,7 +400,7 @@ EXPORT_SYMBOL_GPL(cpuidle_register_device); >>>> */ >>>> void cpuidle_unregister_device(struct cpuidle_device *dev) >>>> { >>>> - if (dev->registered == 0) >>>> + if (!dev || dev->registered == 0) >>>> return; >>>> >>>> cpuidle_pause_and_lock(); >> >> Oops, wait. Are we sure the problem is coming from cpuidle ? > > It is acpi_processor_power_exit assuming that the cpuidle is > initialized. It could be fixed there too, but there are multiple > entries in cpuidle where it does the : "if (!dev) return .." > so I figured this should be done as well here. I understand. From my POV the bug is coming from the acpi processor idle driver. The function acpi_processor_power_init registers the cpuidle driver and the cpuidle device when acpi_processor_registered is zero. Then it increments acpi_processor_registered preventing the next call to this function to register the driver but it will register the device. As cpuidle is disabled, the cpuidle_register_driver fails, thus the device is not registered and acpi_processor_registered is not incremented. So all calls to acpi_processor_power_init prevents the driver and the device to be registered. No problem with that. But the function acpi_processor_power_exit does not take care of the value of acpi_processor_registered and just unregister the device. Then it decrements acpi_processor_registered which is zero to -1. Trying to be immune from a NULL pointer in cpuidle_unregister_device hides bogus code from the caller. So IMO, this check shouldn't be there and the acpi_processor_power_exit function should be fixed instead. >> The cpuidle_unregister_device is called with a NULL pointer, that >> shouldn't happen. > > It does :-) >> >> Konrad, you say that could be easily reproduced. How do you produce >> it ? Unplugging a cpu ? > > Yes. > -- Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog