Passing 0 to roundup_pow_of_two would lead to wrapping around and trying to
find the last set bit on (unsigned long)(-1), which is obviously wrong.
Instead, deal with this case by rounding it up to the closest power of two
(2 ** 0).
Signed-off-by: Sasha Levin <[email protected]>
---
include/linux/log2.h | 3 +++
1 file changed, 3 insertions(+)
diff --git a/include/linux/log2.h b/include/linux/log2.h
index fd7ff3d..b6bdf0c 100644
--- a/include/linux/log2.h
+++ b/include/linux/log2.h
@@ -60,6 +60,9 @@ bool is_power_of_2(unsigned long n)
static inline __attribute__((const))
unsigned long __roundup_pow_of_two(unsigned long n)
{
+ if (n == 0)
+ return 1UL << 0;
+
return 1UL << fls_long(n - 1);
}
--
1.7.10.4
2015-12-03 18:30 GMT+03:00 Sasha Levin <[email protected]>:
> Passing 0 to roundup_pow_of_two would lead to wrapping around and trying to
> find the last set bit on (unsigned long)(-1), which is obviously wrong.
>
> Instead, deal with this case by rounding it up to the closest power of two
> (2 ** 0).
>
> Signed-off-by: Sasha Levin <[email protected]>
> ---
> include/linux/log2.h | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/include/linux/log2.h b/include/linux/log2.h
> index fd7ff3d..b6bdf0c 100644
> --- a/include/linux/log2.h
> +++ b/include/linux/log2.h
> @@ -60,6 +60,9 @@ bool is_power_of_2(unsigned long n)
> static inline __attribute__((const))
> unsigned long __roundup_pow_of_two(unsigned long n)
> {
> + if (n == 0)
> + return 1UL << 0;
> +
Perhaps we should fix callers instead?
Comment near roundup_pow_of_two() says that result is undefined when n == 0:
/**
* roundup_pow_of_two - round the given value up to nearest power of two
* @n - parameter
*
* round the given value up to the nearest power of two
* - the result is undefined when n == 0
On 12/03/2015 10:39 AM, Andrey Ryabinin wrote:
> 2015-12-03 18:30 GMT+03:00 Sasha Levin <[email protected]>:
>> > Passing 0 to roundup_pow_of_two would lead to wrapping around and trying to
>> > find the last set bit on (unsigned long)(-1), which is obviously wrong.
>> >
>> > Instead, deal with this case by rounding it up to the closest power of two
>> > (2 ** 0).
>> >
>> > Signed-off-by: Sasha Levin <[email protected]>
>> > ---
>> > include/linux/log2.h | 3 +++
>> > 1 file changed, 3 insertions(+)
>> >
>> > diff --git a/include/linux/log2.h b/include/linux/log2.h
>> > index fd7ff3d..b6bdf0c 100644
>> > --- a/include/linux/log2.h
>> > +++ b/include/linux/log2.h
>> > @@ -60,6 +60,9 @@ bool is_power_of_2(unsigned long n)
>> > static inline __attribute__((const))
>> > unsigned long __roundup_pow_of_two(unsigned long n)
>> > {
>> > + if (n == 0)
>> > + return 1UL << 0;
>> > +
> Perhaps we should fix callers instead?
> Comment near roundup_pow_of_two() says that result is undefined when n == 0:
That's how I've started doing it, but when it showed up with 3 different callers
I figured it's better to fix it at the source.
This fix would return a valid value and is working fine with the callers.
Thanks,
Sasha
2015-12-03 19:04 GMT+03:00 Sasha Levin <[email protected]>:
> On 12/03/2015 10:39 AM, Andrey Ryabinin wrote:
>> 2015-12-03 18:30 GMT+03:00 Sasha Levin <[email protected]>:
>>> > Passing 0 to roundup_pow_of_two would lead to wrapping around and trying to
>>> > find the last set bit on (unsigned long)(-1), which is obviously wrong.
>>> >
>>> > Instead, deal with this case by rounding it up to the closest power of two
>>> > (2 ** 0).
>>> >
>>> > Signed-off-by: Sasha Levin <[email protected]>
>>> > ---
>>> > include/linux/log2.h | 3 +++
>>> > 1 file changed, 3 insertions(+)
>>> >
>>> > diff --git a/include/linux/log2.h b/include/linux/log2.h
>>> > index fd7ff3d..b6bdf0c 100644
>>> > --- a/include/linux/log2.h
>>> > +++ b/include/linux/log2.h
>>> > @@ -60,6 +60,9 @@ bool is_power_of_2(unsigned long n)
>>> > static inline __attribute__((const))
>>> > unsigned long __roundup_pow_of_two(unsigned long n)
>>> > {
>>> > + if (n == 0)
>>> > + return 1UL << 0;
>>> > +
>> Perhaps we should fix callers instead?
>> Comment near roundup_pow_of_two() says that result is undefined when n == 0:
>
> That's how I've started doing it, but when it showed up with 3 different callers
> I figured it's better to fix it at the source.
>
> This fix would return a valid value and is working fine with the callers.
>
In that case patch should update the comment.
> Thanks,
> Sasha