On Mon, Mar 5, 2012 at 1:53 PM, Bryan Freed <bfreed@xxxxxxxxxxxx> wrote:
> Looks good to me, too,
Thanks for reviewing!
> but couldn't we get rid of one fail state (fail3) by
> reversing the order of schedule_delayed_work() and iio_device_register()?
My focus on this patch was to get a driver to load/unload properly and
avoid restructuring bits that can be deferred to later patches.
I had the same thought and considered this. The problem is I don't
understand or have time to investigate the locking that is hinted at
in the comment preceding the call to schedule_delayed_work(). I don't
understand what locking they are referring to since interrupts are in
fact enabled by the time the delayed work function is invoked.
I have to wonder if the comment is left over from when the function
was directly called instead of "delayed work". ie race condition
exists regardless of when we call schedule_delayed_work().
> Reviewed-by: Bryan Freed <bfreed@xxxxxxxxxxxx>
Thanks! :)
cheers,
grant
>
> bryan.
>
> On Mon, Mar 5, 2012 at 9:15 AM, Grant Grundler <grundler@xxxxxxxxxxxx>
> wrote:
>>
>> tsl2563 probe function has two minor issues with it's error handling
>> paths:
>> 1) it is silent (did not report errors to dmesg)
>> 2) did not return failure code (mixed up use of ret and err)
>>
>> and two major issues:
>> 3) goto fail2 would corrupt a free memory pool ("double free")
>> 4) device registration failure did NOT cancel/flush delayed work.
>> (and thus dereference a freed data structure later)
>>
>> The "double free" is subtle and was introduced with this change:
>> Author: Jonathan Cameron <jic23@xxxxxxxxx>
>> Date: Mon Apr 18 12:58:55 2011 +0100
>> staging:iio:tsl2563 take advantage of new iio_device_allocate private
>> data.
>>
>> Originally, chip was allocated seperately. Now it's appended to the
>> indio_dev by iio_allocate_device(sizeof(*chip)). So we only need one
>> kfree call as well (in iio_free_device()).
>>
>> Signed-off-by: Grant Grundler <grundler@xxxxxxxxxxxx>
>> ---
>> Gory details of tracking this down are here:
>> http://crosbug.com/26819
>>
>> Despite iio_device_registration failing, system can at least now boot.
>> Will follow up with a fix to "double register : in_intensity_both_raw"
>> error that is included in the bug report.
>>
>> drivers/staging/iio/light/tsl2563.c | 39
>> +++++++++++++++++++++++-----------
>> 1 files changed, 26 insertions(+), 13 deletions(-)
>>
>> diff --git a/drivers/staging/iio/light/tsl2563.c
>> b/drivers/staging/iio/light/tsl2563.c
>> index 7e984bc..bf6498b 100644
>> --- a/drivers/staging/iio/light/tsl2563.c
>> +++ b/drivers/staging/iio/light/tsl2563.c
>> @@ -705,7 +705,6 @@ static int __devinit tsl2563_probe(struct i2c_client
>> *client,
>> struct tsl2563_chip *chip;
>> struct tsl2563_platform_data *pdata = client->dev.platform_data;
>> int err = 0;
>> - int ret;
>> u8 id = 0;
>>
>> indio_dev = iio_allocate_device(sizeof(*chip));
>> @@ -719,13 +718,15 @@ static int __devinit tsl2563_probe(struct i2c_client
>> *client,
>>
>> err = tsl2563_detect(chip);
>> if (err) {
>> - dev_err(&client->dev, "device not found, error %d\n",
>> -err);
>> + dev_err(&client->dev, "detect error %d\n", -err);
>> goto fail1;
>> }
>>
>> err = tsl2563_read_id(chip, &id);
>> - if (err)
>> + if (err) {
>> + dev_err(&client->dev, "read id error %d\n", -err);
>> goto fail1;
>> + }
>>
>> mutex_init(&chip->lock);
>>
>> @@ -748,40 +749,52 @@ static int __devinit tsl2563_probe(struct i2c_client
>> *client,
>> indio_dev->num_channels = ARRAY_SIZE(tsl2563_channels);
>> indio_dev->dev.parent = &client->dev;
>> indio_dev->modes = INDIO_DIRECT_MODE;
>> +
>> if (client->irq)
>> indio_dev->info = &tsl2563_info;
>> else
>> indio_dev->info = &tsl2563_info_no_irq;
>> +
>> if (client->irq) {
>> - ret = request_threaded_irq(client->irq,
>> + err = request_threaded_irq(client->irq,
>> NULL,
>> &tsl2563_event_handler,
>> IRQF_TRIGGER_RISING |
>> IRQF_ONESHOT,
>> "tsl2563_event",
>> indio_dev);
>> - if (ret)
>> - goto fail2;
>> + if (err) {
>> + dev_err(&client->dev, "irq request error %d\n",
>> -err);
>> + goto fail1;
>> + }
>> }
>> +
>> err = tsl2563_configure(chip);
>> - if (err)
>> - goto fail3;
>> + if (err) {
>> + dev_err(&client->dev, "configure error %d\n", -err);
>> + goto fail2;
>> + }
>>
>> INIT_DELAYED_WORK(&chip->poweroff_work, tsl2563_poweroff_work);
>> +
>> /* The interrupt cannot yet be enabled so this is fine without lock
>> */
>> schedule_delayed_work(&chip->poweroff_work, 5 * HZ);
>>
>> - ret = iio_device_register(indio_dev);
>> - if (ret)
>> + err = iio_device_register(indio_dev);
>> + if (err) {
>> + dev_err(&client->dev, "iio registration error %d\n",
>> -err);
>> goto fail3;
>> + }
>>
>> return 0;
>> +
>> fail3:
>> + cancel_delayed_work(&chip->poweroff_work);
>> + flush_scheduled_work();
>> +fail2:
>> if (client->irq)
>> free_irq(client->irq, indio_dev);
>> -fail2:
>> - iio_free_device(indio_dev);
>> fail1:
>> - kfree(chip);
>> + iio_free_device(indio_dev);
>> return err;
>> }
>>
>> --
>> 1.7.3.4
>>
>
_______________________________________________
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxx
http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
