linux-kernel - Unregistering extcon providers while they are in use leads to kernel crashes

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-next>] [day] [month] [year] [list]

Message-ID: <0c877df4-c5d7-1eae-bff9-469df9651d09@redhat.com>
Date:   Wed, 21 Dec 2016 16:54:27 +0100
From:   Hans de Goede <hdegoede@...hat.com>
To:     Chanwoo Choi <cw00.choi@...sung.com>
Cc:     myungjoo.ham@...sung.com,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Unregistering extcon providers while they are in use leads to kernel
 crashes

Hi,

With the recent extcon work I've been doing I noticed that
if I want to rmmod and then insmod say extcon_axp288 I can
do so without problems even if axp288_charger is holding
a reference to the extcon device returned by extcon_get_extcon_dev.

The problem is that extcon_get_extcon_dev simply looks up
the extcon-device in the list of current registered extcon-s
and then returns a pointer to it, without any reference
counting.

The rmmod scenario can be fixed by doing a module_get from
extcon_get_extcon_dev, but that still leaves the same problem
when root manually unbinds the driver through sysfs.

A possible way fix this would be:

1) Make all extcon providers use devm_extcon_dev_allocate and document
using this to allocate an extcon_dev mandatory

2) Add a refcount to struct extcon_dev and introduce extcon_dev_get
and extcon_dev_put helpers which modify the refcount and only free
the memory on the final put (and make the evm_extcon_dev_allocate
cleanup function call extcon_dev_put)

3) On extcon_dev_unregister set a flag in the extcon_dev that it
has been free-ed, make all extcon consumer functions which take
an extcon_dev (extcon_get_state, extcon_register_notifier, etc.)
check this flag and return -ENODEV when the extcon has been unregistered

4) Make extcon_get_extcon_dev call extcon_dev_get on the returned edev
before returning it

 From here on we've fixed the crash, but we now leak the extcon_dev
when the consumer gets unbound.

5) Add a devm_extcon_get_extcon_dev which calls extcon_dev_put as the devm
cleanup function

6) Convert all extcon consumers to use devm_extcon_get_extcon_dev

Regards,

Hans