From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Message-ID: <50C7932A.2030104@metafoo.de> Date: Tue, 11 Dec 2012 21:10:18 +0100 From: Lars-Peter Clausen MIME-Version: 1.0 To: Hartmut Knaack CC: Jonathan Cameron , Sascha Hauer , linux-iio@vger.kernel.org Subject: Re: [PATCH 0/2] fixes for adt7410 References: <50032AD8.9040103@gmx.de> <50033218.7030602@metafoo.de> <20120716083428.GD30009@pengutronix.de> <5003D404.4040005@kernel.org> <50C7902C.2060804@gmx.de> In-Reply-To: <50C7902C.2060804@gmx.de> Content-Type: text/plain; charset=ISO-8859-15 List-ID: On 12/11/2012 08:57 PM, Hartmut Knaack wrote: > Jonathan Cameron schrieb: >> On 7/16/2012 9:34 AM, Sascha Hauer wrote: >>> On Sun, Jul 15, 2012 at 11:11:52PM +0200, Lars-Peter Clausen wrote: >>>> On 07/15/2012 10:40 PM, Hartmut Knaack wrote: >>>>> This patchset makes the ADT7410 usable. Main reason was the following error when trying to register one of these devices: >>>>> [ 180.945561] BUG: unable to handle kernel NULL pointer dereference at (null) >>>>> [ 180.945592] IP: [] iio_device_register_eventset+0x380/0x3a0 [industrialio] >>>>> This happens, since in industrialio-events.c __iio_add_event_config_attrs (which does a INIT_LIST_HEAD) is not called, if the channels attribute of the device is not set. As a result, list_for_each_entry causes those NULL pointer dereference problems. So, before trying to access those list elements, we check for their existence (also in unregister functions). >>>>> Now the adt7410.c ended in some complex changes - my apologies for this: >>>>> >>>>> * check for exact values provided through sysfs, otherwise output a verbose error/usage message for: sample mode, resolution, event mode >>>>> * adt7410_show_id: I don't see a point in masking out some LSBs, if they get shifted to oblivion, anyway -> no more need for ADT7410_MANUFACTORY_ID_MASK >>>>> * adt7410_convert_temperature: in 13 bit mode, the driver assumed the data to be stored in bits 0-12, but according to the data sheet, it is stored in bits 3-15. Temperature readings were always around 200(°C), while real temperature was something around 20(°C). Simple fix: mask out bits 0-2 in 13 bit mode and use the 16 bit routines from that point on. >>>>> * adt7410_set_t_bound: handle float values provided through sysfs properly. Also, simplify treatment of 13 bit values by masking out bits 0-2. >>>>> * adt7410_probe, adt7410_remove: fix another possible NULL pointer dereference issue, in case no platform data is provided. Sync chip->config with the config register >>>> Hi, >>>> >>>> Sascha Hauer posted a couple of patches a while ago which seem to fix a >>>> similar set of issues. >>>> >>>> See: http://www.spinics.net/lists/linux-iio/msg05947.html and >>>> http://www.spinics.net/lists/linux-iio/msg05955.html >>> Indeed, most if not all of the mentioned issues should be fixed already >>> with these patches. >>> >> Just to reiterate what I said to Sascha at the time... This driver does >> not belong in IIO. It is very much a hardware monitoring part and so >> wants to move to hwmon. We aren't going to take it out of staging into >> drivers/iio and I'm rather unwilling to take any 'new' features into the >> staging version. Someone with hardware who is interesting needs to bite >> the bullet and convert / rewrite this driver as a hwmon device. >> > FYI, I created a basic hwmon driver for ADT7410, which now made it into linux 3.7. > Take care Ah, great, now we have two drivers which bind to the same device... - Lars