RE: [PATCH bpf] bpf: Fix DEVMAP_HASH overflow check on 32-bit arches

["Toke Høiland-Jørgensen" <toke@redhat.com>]

John Fastabend <john.fastabend@gmail.com> writes:

> Toke Høiland-Jørgensen wrote:
>> John Fastabend <john.fastabend@gmail.com> writes:
>> 
>> > Toke Høiland-Jørgensen wrote:
>> >> The devmap code allocates a number hash buckets equal to the next power of two
>> >> of the max_entries value provided when creating the map. When rounding up to the
>> >> next power of two, the 32-bit variable storing the number of buckets can
>> >> overflow, and the code checks for overflow by checking if the truncated 32-bit value
>> >> is equal to 0. However, on 32-bit arches the rounding up itself can overflow
>> >> mid-way through, because it ends up doing a left-shift of 32 bits on an unsigned
>> >> long value. If the size of an unsigned long is four bytes, this is undefined
>> >> behaviour, so there is no guarantee that we'll end up with a nice and tidy
>> >> 0-value at the end.
>
> Hi Toke, dumb question where is this left-shift noted above? It looks
> like fls_long tries to account by having a check for sizeof(l) == 4.
> I'm asking mostly because I've found a few more spots without this
> check.

That check in fls_long only switches between too different
implementations of the fls op itself (fls() vs fls64()). AFAICT this is
mostly meaningful for the generic (non-ASM) version that iterates over
the bits instead of just emitting a single instruction.

The shift is in the caller:

static inline __attribute__((const))
unsigned long __roundup_pow_of_two(unsigned long n)
{
	return 1UL << fls_long(n - 1);
}

If this is called with a value > 0x80000000, fls_long() will (correctly)
return 32, leading to the ub[0] shift when sizeof(unsigned long) == 4.

-Toke

[0] https://wiki.sei.cmu.edu/confluence/display/c/int34-c.+do+not+shift+an+expression+by+a+negative+number+of+bits+or+by+greater+than+or+equal+to+the+number+of+bits+that+exist+in+the+operand