On 2018-06-27 20:22, Leon Romanovsky wrote:

On Wed, Jun 27, 2018 at 12:10:12PM -0600, Jason Gunthorpe wrote:

quoted

On Wed, Jun 27, 2018 at 11:36:03AM +0200, Rasmus Villemoes wrote:

quoted

   Well, the types you can check at compile-time, the values not, so you
   still have to define the result, i.e. contents of *d, for negative
   values (even if we decide that "overflow" should always be signalled in
   that case).

Why do a need to define a 'result' beyond whatever the not-undefined
behavior shift expression produces?

Well, perhaps you don't, it's just that the other check_*_overflow have
that behaviour (which they inherit from gcc's builtins), and it's a
nice-to-have. But I see that it's hard to come up with something
sensible in all the "doesn't make sense" cases. When writing the tests
for test_overflow.c, you can of course just omit the comparison to
the/an expected result in the overflow case.

quoted

/*
 * Compute *d = (a << s)
 *
 * Returns true if '*d' cannot hold the result or 'a << s' doesn't make sense.
 * - 'a << s' causes bits to be lost when stored in d
 * - 's' is garbage (eg negative) or so large that a << s is guarenteed to be 0
 * - 'a' is negative
 * - 'a << s' sets the sign bit, if any, in '*d'
 * *d is not defined if false is returned.
 */
#define check_shift_overflow(a, s, d)                                          \
	({                                                                     \
		typeof(a) _a = a;                                              \
		typeof(s) _s = s;                                              \
		typeof(d) _d = d;                                              \
		u64 _a_full = _a;                                              \
		unsigned int _to_shift =                                       \
			_s >= 0 && _s < 8 * sizeof(*d) ? _s : 0;               \
                                                                               \
		*_d = (_a_full << _to_shift);                                  \
                                                                               \
		(_to_shift != _s || *_d < 0 || _a < 0 ||                       \
		 (*_d >> _to_shift) != a);                                     \
	})

That last a still needs to be _a. Other than that, I don't see anything
wrong with this version.

quoted

int main(int argc, const char *argv[])
{
	int32_t s32;
	uint32_t u32;

	assert(check_shift_overflow(1, 0, &s32) == false && s32 == (1 << 0));

[...]>> 	assert(check_shift_overflow(0xFFFFFFFF, 1, &s32) == true);

Please add these in some form to test_overflow.c, but do also include
cases where a and *d have different width, e.g. check_shift_overflow(1,
32, &s64) should be ok, while check_shift_overflow(65432, 0, &s16)
should not.

Rasmus