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: <20170403075133.12343-1-zboszor@pr.hu>
Date:   Mon,  3 Apr 2017 09:51:30 +0200
From:   Zoltan Boszormenyi <zboszor@...hu>
To:     linux-usb@...r.kernel.org, linux-watchdog@...r.kernel.org,
        linux-i2c@...r.kernel.org
Cc:     linux-kernel@...r.kernel.org,
        Paul Menzel <paulepanter@...rs.sourceforge.net>,
        Christian Fetzer <fetzer.ch@...il.com>,
        Jean Delvare <jdelvare@...e.com>,
        Nehal Shah <nehal-bakulchandra.shah@....com>,
        Tim Small <tim@...ss.co.uk>,
        Guenter Roeck <linux@...ck-us.net>,
        Zoltan Boszormenyi <zboszor@...hu>
Subject: [PATCH 0/3 v2] Fix regression in the sp5100_tco driver


Three drivers are accessing the same I/O ports (0xcd6 / 0xcd7) on
AMD SB800 based machines without synchronization or with excluding
each other out:
* the USB quirk for isochronous transfers on SB800 (no locking)
* sp5100_tco (request_region)
* i2c-piix4 (request_region)

Historically, the sp5100_tco watchdog driver used request_region()
for these I/O ports but an i2c-piix4 improvement for SB800 in
Linux 4.4-rc4 also added a request_region() call. Because of this
and the load order, this cause a regression and the watchdog function
became non-functional. The commit that caused the regression is:

commit 2fee61d22e606fc99ade9079fda15fdee83ec33e
Author: Christian Fetzer <fetzer.ch@...il.com>
Date:   Thu Nov 19 20:13:48 2015 +0100

    i2c: piix4: Add support for multiplexed main adapter in SB800

I was informed by Guenter Roeck <linux@...ck-us.net> that the
alternative, i.e. using request_muxed_region() can fail, either
because of a resource allocation failure, which is quite possible
with long uptimes, or because there are no guarantees that an as yet
unknown driver would also use request_muxed_region() consistently.

Because of this, a solution using a common mutex was chosen to
synchronize I/O port accesses and request_region() calls are removed
from both i2c-piix4 and sp5100_tco to make the code uniform.

This patch series implements this and restores the watchdog function.

v2: Don't introduce a new header, reference sb800_mutex explicitly

Signed-off-by: Zoltan Boszormenyi <zboszor@...hu>

 drivers/i2c/busses/i2c-piix4.c | 43 +++++++++++++------------------------------
 drivers/usb/host/pci-quirks.c  |  5 +++++
 drivers/watchdog/sp5100_tco.c  | 31 ++++++++++++++++++-------------
 3 files changed, 36 insertions(+), 43 deletions(-)

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ