linux-kernel - Re: [PATCH v2 1/6] ACPI / hotplug: Fix theoretical race in acpi_hotplug_notify

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Sun, 02 Feb 2014 18:01:31 +0100
From:	"Rafael J. Wysocki" <rjw@...ysocki.net>
To:	ACPI Devel Maling List <linux-acpi@...r.kernel.org>
Cc:	Bjorn Helgaas <bhelgaas@...gle.com>, Aaron Lu <aaron.lu@...el.com>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Linux PCI <linux-pci@...r.kernel.org>,
	Mika Westerberg <mika.westerberg@...ux.intel.com>,
	Robert Moore <robert.moore@...el.com>
Subject: Re: [PATCH v2 1/6] ACPI / hotplug: Fix theoretical race in acpi_hotplug_notify_cb()

On Sunday, February 02, 2014 01:54:02 AM Rafael J. Wysocki wrote:
> From: Rafael J. Wysocki <rafael.j.wysocki@...el.com>
> 
> There is a slight possibility for the ACPI device object pointed to
> by adev in acpi_hotplug_notify_cb() to become invalid between the
> acpi_bus_get_device() that it comes from and the subsequent get_device().
> Namely, if acpi_scan_drop_device() runs concurrently with respect to
> acpi_hotplug_notify_cb() and acpi_device_del_list is not empty,
> acpi_device_del_work_fn() may delete the device object in question
> without waiting for the ACPI events workqueue to drain, which very
> well may happen right after a successful execution of
> acpi_bus_get_device() in acpi_hotplug_notify_cb().
> 
> To prevent that from happening, run acpi_bus_get_device() and the
> subsequent get_device() in acpi_hotplug_notify_cb() under
> acpi_device_del_lock, so that the deletion of the given device
> object cannot be queued up by acpi_scan_drop_device() between the
> two.
> 
> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@...el.com>
> ---
>  drivers/acpi/scan.c |   14 ++++++++------
>  1 file changed, 8 insertions(+), 6 deletions(-)
> 
> Index: linux-pm/drivers/acpi/scan.c
> ===================================================================
> --- linux-pm.orig/drivers/acpi/scan.c
> +++ linux-pm/drivers/acpi/scan.c
> @@ -41,6 +41,8 @@ static DEFINE_MUTEX(acpi_scan_lock);
>  static LIST_HEAD(acpi_scan_handlers_list);
>  DEFINE_MUTEX(acpi_device_lock);
>  LIST_HEAD(acpi_wakeup_device_list);
> +static LIST_HEAD(acpi_device_del_list);
> +static DEFINE_MUTEX(acpi_device_del_lock);
>  
>  struct acpi_device_bus_id{
>  	char bus_id[15];
> @@ -488,9 +490,6 @@ static void acpi_hotplug_notify_cb(acpi_
>  	struct acpi_device *adev;
>  	acpi_status status;
>  
> -	if (acpi_bus_get_device(handle, &adev))
> -		goto err_out;
> -
>  	switch (type) {
>  	case ACPI_NOTIFY_BUS_CHECK:
>  		acpi_handle_debug(handle, "ACPI_NOTIFY_BUS_CHECK event\n");
> @@ -512,7 +511,13 @@ static void acpi_hotplug_notify_cb(acpi_
>  		/* non-hotplug event; possibly handled by other handler */
>  		return;
>  	}
> +	mutex_lock(&acpi_device_del_lock);
> +	if (acpi_bus_get_device(handle, &adev)) {
> +		mutex_unlock(&acpi_device_del_lock);
> +		goto err_out;
> +	}
>  	get_device(&adev->dev);
> +	mutex_unlock(&acpi_device_del_lock);
>  	status = acpi_hotplug_execute(acpi_device_hotplug, adev, type);
>  	if (ACPI_SUCCESS(status))
>  		return;

Well, that would have been good if it hand't been broken. :-(

acpi_scan_drop_device() which acquires acpi_device_del_lock is called under
the ACPICA's namespace mutex and acpi_bus_get_device() above acquires that
mutex, so this leads to a classical ABBA deadlock scenario.  Bummer.

And I haven't been able to convince myself that what we're doing in
acpi_hotplug_notify_cb() is actually safe without any locking.  Not to
mention acpi_bus_notify() for that matter.  Moreover, the *only* safe
way to do that I'm seeing at the moment is to call the get_device()
under the ACPICA's namespace mutex, before it is released in
acpi_get_data().

Of course, ACPICA will need to be modified slightly for that to be
possible (sorry, Bob), but at least that *should* work, so I have a
new version of this patchset doing just that.  I'll send it out
shortly.

Thanks!

-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/