On Sat, Jan 4, 2014 at 9:21 PM, Bart Van Assche <bvanass...@acm.org> wrote:
> Avoid that bus_unregister() triggers a use-after-free with
> CONFIG_DEBUG_KOBJECT_RELEASE=y. This patch avoids that the
> following sequence triggers a kernel crash with memory poisoning
> enabled:
> * bus_register()
> * driver_register()
> * driver_unregister()
> * bus_unregister()
>
> The above sequence causes the bus private data to be freed from
> inside the bus_unregister() call although it is not guaranteed in
> that function that the reference count on the bus private data has
> dropped to zero. As an example, with CONFIG_DEBUG_KOBJECT_RELEASE=y
> the ${bus}/drivers kobject is still holding a reference on
> bus->p->subsys.kobj via its parent pointer at the time the bus
> private data is freed. Fix this by deferring freeing the bus private
> data until the last kobject_put() call on bus->p->subsys.kobj.
>
> The kernel oops triggered by the above sequence and with memory
> poisoning enabled and that is fixed by this patch is as follows:
>
> general protection fault: 0000 [#1] PREEMPT SMP
> CPU: 3 PID: 2711 Comm: kworker/3:32 Tainted: G W O 3.13.0-rc4-debug+
> #1
> Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
> Workqueue: events kobject_delayed_cleanup
> task: ffff880037f866d0 ti: ffff88003b638000 task.ti: ffff88003b638000
> Call Trace:
> [<ffffffff81263105>] ? kobject_get_path+0x25/0x100
> [<ffffffff81264354>] kobject_uevent_env+0x134/0x600
> [<ffffffff8126482b>] kobject_uevent+0xb/0x10
> [<ffffffff81262fa2>] kobject_delayed_cleanup+0xc2/0x1b0
> [<ffffffff8106c047>] process_one_work+0x217/0x700
> [<ffffffff8106bfdb>] ? process_one_work+0x1ab/0x700
> [<ffffffff8106c64b>] worker_thread+0x11b/0x3a0
> [<ffffffff8106c530>] ? process_one_work+0x700/0x700
> [<ffffffff81074b70>] kthread+0xf0/0x110
> [<ffffffff81074a80>] ? insert_kthread_work+0x80/0x80
> [<ffffffff815673bc>] ret_from_fork+0x7c/0xb0
> [<ffffffff81074a80>] ? insert_kthread_work+0x80/0x80
> Code: 89 f8 48 89 e5 f6 82 c0 27 63 81 20 74 15 0f 1f 44 00 00 48 83 c0 01 0f
> b6 10 f6 82 c0 27 63 81 20 75 f0 5d c3 66 0f 1f 44 00 00 <80> 3f 00 55 48 89
> e5 74 15 48 89 f8 0f 1f 40 00 48 83 c0 01 80
> RIP [<ffffffff81267ed0>] strlen+0x0/0x30
> RSP <ffff88003b639c70>
> ---[ end trace 210f883ef80376aa ]---
>
> Signed-off-by: Bart Van Assche <bvanass...@acm.org>
> ---
> drivers/base/bus.c | 13 +++++++++++--
> 1 file changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/base/bus.c b/drivers/base/bus.c
> index 73f6c29..59dc808 100644
> --- a/drivers/base/bus.c
> +++ b/drivers/base/bus.c
> @@ -146,8 +146,19 @@ void bus_remove_file(struct bus_type *bus, struct
> bus_attribute *attr)
> }
> EXPORT_SYMBOL_GPL(bus_remove_file);
>
> +static void bus_release(struct kobject *kobj)
> +{
> + struct subsys_private *priv =
> + container_of(kobj, typeof(*priv), subsys.kobj);
> + struct bus_type *bus = priv->bus;
> +
> + kfree(priv);
> + bus->p = NULL;
> +}
> +
> static struct kobj_type bus_ktype = {
> .sysfs_ops = &bus_sysfs_ops,
> + .release = bus_release,
> };
>
> static int bus_uevent_filter(struct kset *kset, struct kobject *kobj)
> @@ -953,8 +964,6 @@ void bus_unregister(struct bus_type *bus)
> kset_unregister(bus->p->devices_kset);
> bus_remove_file(bus, &bus_attr_uevent);
> kset_unregister(&bus->p->subsys);
> - kfree(bus->p);
> - bus->p = NULL;
> }
> EXPORT_SYMBOL_GPL(bus_unregister);
Looks the fix is fine,
Acked-by: Ming Lei <ming....@canonical.com>
Thanks,
--
Ming Lei
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
