2016-11-24 13:39:09

by Dan Carpenter

Subject: [patch] iio: tsl2583: make array large enough

This array is supposed to have 10 elements. Smatch complains that with
the current code we can have n == max_ints and read beyond the end of
the array.

Fixes: ac4f6eee8fe8 ("staging: iio: TAOS tsl258x: Device driver")
Signed-off-by: Dan Carpenter <[email protected]>

diff --git a/drivers/iio/light/tsl2583.c b/drivers/iio/light/tsl2583.c
index 0b87f6a..a78b602 100644
--- a/drivers/iio/light/tsl2583.c
+++ b/drivers/iio/light/tsl2583.c
@@ -565,7 +565,7 @@ static ssize_t in_illuminance_lux_table_store(struct device *dev,
struct iio_dev *indio_dev = dev_to_iio_dev(dev);
struct tsl2583_chip *chip = iio_priv(indio_dev);
const unsigned int max_ints = TSL2583_MAX_LUX_TABLE_ENTRIES * 3;
- int value[TSL2583_MAX_LUX_TABLE_ENTRIES * 3];
+ int value[TSL2583_MAX_LUX_TABLE_ENTRIES * 3 + 1];
int ret = -EINVAL;
unsigned int n;

2016-11-24 15:48:54

by Brian Masney

Subject: Re: [patch] iio: tsl2583: make array large enough

On Thu, Nov 24, 2016 at 04:38:07PM +0300, Dan Carpenter wrote:
> This array is supposed to have 10 elements. Smatch complains that with
> the current code we can have n == max_ints and read beyond the end of
> the array.
>
> Fixes: ac4f6eee8fe8 ("staging: iio: TAOS tsl258x: Device driver")
> Signed-off-by: Dan Carpenter <[email protected]>
>
> diff --git a/drivers/iio/light/tsl2583.c b/drivers/iio/light/tsl2583.c
> index 0b87f6a..a78b602 100644
> --- a/drivers/iio/light/tsl2583.c
> +++ b/drivers/iio/light/tsl2583.c
> @@ -565,7 +565,7 @@ static ssize_t in_illuminance_lux_table_store(struct device *dev,
> struct iio_dev *indio_dev = dev_to_iio_dev(dev);
> struct tsl2583_chip *chip = iio_priv(indio_dev);
> const unsigned int max_ints = TSL2583_MAX_LUX_TABLE_ENTRIES * 3;
> - int value[TSL2583_MAX_LUX_TABLE_ENTRIES * 3];
> + int value[TSL2583_MAX_LUX_TABLE_ENTRIES * 3 + 1];
> int ret = -EINVAL;
> unsigned int n;
>

Acked-by: Brian Masney <[email protected]>

2016-11-24 16:54:35

by walter harms

Subject: Re: [patch] iio: tsl2583: make array large enough

Am 24.11.2016 16:48, schrieb Brian Masney:
> On Thu, Nov 24, 2016 at 04:38:07PM +0300, Dan Carpenter wrote:
>> This array is supposed to have 10 elements. Smatch complains that with
>> the current code we can have n == max_ints and read beyond the end of
>> the array.
>>
>> Fixes: ac4f6eee8fe8 ("staging: iio: TAOS tsl258x: Device driver")
>> Signed-off-by: Dan Carpenter <[email protected]>
>>
>> diff --git a/drivers/iio/light/tsl2583.c b/drivers/iio/light/tsl2583.c
>> index 0b87f6a..a78b602 100644
>> --- a/drivers/iio/light/tsl2583.c
>> +++ b/drivers/iio/light/tsl2583.c
>> @@ -565,7 +565,7 @@ static ssize_t in_illuminance_lux_table_store(struct device *dev,
>> struct iio_dev *indio_dev = dev_to_iio_dev(dev);
>> struct tsl2583_chip *chip = iio_priv(indio_dev);
>> const unsigned int max_ints = TSL2583_MAX_LUX_TABLE_ENTRIES * 3;
>> - int value[TSL2583_MAX_LUX_TABLE_ENTRIES * 3];
>> + int value[TSL2583_MAX_LUX_TABLE_ENTRIES * 3 + 1];
>> int ret = -EINVAL;
>> unsigned int n;
>>
>

sorry i did not notice that bevor ..
there is a
max_ints = TSL2583_MAX_LUX_TABLE_ENTRIES * 3

IMHO this should read either:
int value[max_ints+1];
or
max_ints=ARRAY_SIZE(value)-1;

(my personal favorite is dropping max_ints completely).

re,
wh

2016-11-24 17:51:30

by Brian Masney

Subject: Re: [patch] iio: tsl2583: make array large enough

On Thu, Nov 24, 2016 at 05:54:17PM +0100, walter harms wrote:
>
>
> Am 24.11.2016 16:48, schrieb Brian Masney:
> > On Thu, Nov 24, 2016 at 04:38:07PM +0300, Dan Carpenter wrote:
> >> This array is supposed to have 10 elements. Smatch complains that with
> >> the current code we can have n == max_ints and read beyond the end of
> >> the array.
> >>
> >> Fixes: ac4f6eee8fe8 ("staging: iio: TAOS tsl258x: Device driver")
> >> Signed-off-by: Dan Carpenter <[email protected]>
> >>
> >> diff --git a/drivers/iio/light/tsl2583.c b/drivers/iio/light/tsl2583.c
> >> index 0b87f6a..a78b602 100644
> >> --- a/drivers/iio/light/tsl2583.c
> >> +++ b/drivers/iio/light/tsl2583.c
> >> @@ -565,7 +565,7 @@ static ssize_t in_illuminance_lux_table_store(struct device *dev,
> >> struct iio_dev *indio_dev = dev_to_iio_dev(dev);
> >> struct tsl2583_chip *chip = iio_priv(indio_dev);
> >> const unsigned int max_ints = TSL2583_MAX_LUX_TABLE_ENTRIES * 3;
> >> - int value[TSL2583_MAX_LUX_TABLE_ENTRIES * 3];
> >> + int value[TSL2583_MAX_LUX_TABLE_ENTRIES * 3 + 1];
> >> int ret = -EINVAL;
> >> unsigned int n;
> >>
> >
>
> sorry i did not notice that bevor ..
> there is a
> max_ints = TSL2583_MAX_LUX_TABLE_ENTRIES * 3
>
> IMHO this should read either:
> int value[max_ints+1];

I originally went this route when I refactored the function, however
running make C=1 yields the following warnings:

drivers/iio/light/tsl2583.c:568:19: warning: Variable length array is
used.
drivers/iio/light/tsl2583.c:574:26: error: cannot size expression

That is why I went with the current implementation.

> or
> max_ints=ARRAY_SIZE(value)-1;
>
> (my personal favorite is dropping max_ints completely).

The max_ints value is also shown in the error message if the user passes
in too many or too few entries in the per device lux table. I wanted the
user to see the maximum allowable number without having to dig through
the kernel source code. Without it, I would have had to duplicate the
TSL2583_MAX_LUX_TABLE_ENTRIES * 3 statement a third time.

Brian

2016-11-24 20:12:59

by Jonathan Cameron

Subject: Re: [patch] iio: tsl2583: make array large enough

On 24/11/16 17:51, Brian Masney wrote:
> On Thu, Nov 24, 2016 at 05:54:17PM +0100, walter harms wrote:
>>
>>
>> Am 24.11.2016 16:48, schrieb Brian Masney:
>>> On Thu, Nov 24, 2016 at 04:38:07PM +0300, Dan Carpenter wrote:
>>>> This array is supposed to have 10 elements. Smatch complains that with
>>>> the current code we can have n == max_ints and read beyond the end of
>>>> the array.
>>>>
>>>> Fixes: ac4f6eee8fe8 ("staging: iio: TAOS tsl258x: Device driver")
>>>> Signed-off-by: Dan Carpenter <[email protected]>
>>>>
>>>> diff --git a/drivers/iio/light/tsl2583.c b/drivers/iio/light/tsl2583.c
>>>> index 0b87f6a..a78b602 100644
>>>> --- a/drivers/iio/light/tsl2583.c
>>>> +++ b/drivers/iio/light/tsl2583.c
>>>> @@ -565,7 +565,7 @@ static ssize_t in_illuminance_lux_table_store(struct device *dev,
>>>> struct iio_dev *indio_dev = dev_to_iio_dev(dev);
>>>> struct tsl2583_chip *chip = iio_priv(indio_dev);
>>>> const unsigned int max_ints = TSL2583_MAX_LUX_TABLE_ENTRIES * 3;
>>>> - int value[TSL2583_MAX_LUX_TABLE_ENTRIES * 3];
>>>> + int value[TSL2583_MAX_LUX_TABLE_ENTRIES * 3 + 1];
>>>> int ret = -EINVAL;
>>>> unsigned int n;
>>>>
>>>
>>
>> sorry i did not notice that bevor ..
>> there is a
>> max_ints = TSL2583_MAX_LUX_TABLE_ENTRIES * 3
>>
>> IMHO this should read either:
>> int value[max_ints+1];
>
> I originally went this route when I refactored the function, however
> running make C=1 yields the following warnings:
>
> drivers/iio/light/tsl2583.c:568:19: warning: Variable length array is
> used.
> drivers/iio/light/tsl2583.c:574:26: error: cannot size expression
>
> That is why I went with the current implementation.
>
>> or
>> max_ints=ARRAY_SIZE(value)-1;
>>
>> (my personal favorite is dropping max_ints completely).
>
> The max_ints value is also shown in the error message if the user passes
> in too many or too few entries in the per device lux table. I wanted the
> user to see the maximum allowable number without having to dig through
> the kernel source code. Without it, I would have had to duplicate the
> TSL2583_MAX_LUX_TABLE_ENTRIES * 3 statement a third time.
>
> Brian
>
I'm taking this as is, because it'll just squeeze into my last
pull request to Greg. Sorry for not waiting for this discussion
to finish first!

Applied to the togreg branch of iio.git and pushed out as testing for the
autobuilders to very briefly play with it.

Thanks,

Jonathan

2016-11-25 08:53:56

by walter harms

Subject: Re: [patch] iio: tsl2583: make array large enough

Am 24.11.2016 18:51, schrieb Brian Masney:
> On Thu, Nov 24, 2016 at 05:54:17PM +0100, walter harms wrote:
>>
>>
>> Am 24.11.2016 16:48, schrieb Brian Masney:
>>> On Thu, Nov 24, 2016 at 04:38:07PM +0300, Dan Carpenter wrote:
>>>> This array is supposed to have 10 elements. Smatch complains that with
>>>> the current code we can have n == max_ints and read beyond the end of
>>>> the array.
>>>>
>>>> Fixes: ac4f6eee8fe8 ("staging: iio: TAOS tsl258x: Device driver")
>>>> Signed-off-by: Dan Carpenter <[email protected]>
>>>>
>>>> diff --git a/drivers/iio/light/tsl2583.c b/drivers/iio/light/tsl2583.c
>>>> index 0b87f6a..a78b602 100644
>>>> --- a/drivers/iio/light/tsl2583.c
>>>> +++ b/drivers/iio/light/tsl2583.c
>>>> @@ -565,7 +565,7 @@ static ssize_t in_illuminance_lux_table_store(struct device *dev,
>>>> struct iio_dev *indio_dev = dev_to_iio_dev(dev);
>>>> struct tsl2583_chip *chip = iio_priv(indio_dev);
>>>> const unsigned int max_ints = TSL2583_MAX_LUX_TABLE_ENTRIES * 3;
>>>> - int value[TSL2583_MAX_LUX_TABLE_ENTRIES * 3];
>>>> + int value[TSL2583_MAX_LUX_TABLE_ENTRIES * 3 + 1];
>>>> int ret = -EINVAL;
>>>> unsigned int n;
>>>>
>>>
>>
>> sorry i did not notice that bevor ..
>> there is a
>> max_ints = TSL2583_MAX_LUX_TABLE_ENTRIES * 3
>>
>> IMHO this should read either:
>> int value[max_ints+1];
>
> I originally went this route when I refactored the function, however
> running make C=1 yields the following warnings:
>
> drivers/iio/light/tsl2583.c:568:19: warning: Variable length array is
> used.
> drivers/iio/light/tsl2583.c:574:26: error: cannot size expression
>
> That is why I went with the current implementation.
>
>> or
>> max_ints=ARRAY_SIZE(value)-1;
>>
>> (my personal favorite is dropping max_ints completely).
>
> The max_ints value is also shown in the error message if the user passes
> in too many or too few entries in the per device lux table. I wanted the
> user to see the maximum allowable number without having to dig through
> the kernel source code. Without it, I would have had to duplicate the
> TSL2583_MAX_LUX_TABLE_ENTRIES * 3 statement a third time.
>
> Brian
>


Hello Brian, thanks for the replay,

i have no problem when people so such things intentional but some times
people do this unintentional. When i review code, i see it as part of my
job to challenge constructs where i get a strange feeling.

re,
wh