On Tue, Sep 14, 2021 at 11:45 AM Nick Desaulniers
[off-list ref] wrote:

Any issues passing an loff_t (aka long long) to __ffs which expects an
unsigned long for ilp32 targets?

No.

We literally _just_ checked that the value is a power-of-two, and that
it's in the range [1024, PAGE_SIZE].

There was never anything "loff_t" about bitmask at any point.

Any issues modifying the sysfs interface? Perhaps something in
userspace relies on parsing those strings?

See my comment about how it could use DEFINE_SHOW_ATTRIBUTE() to
always show the bits as a value.

But it's not even sysfs. It's debugfs. So nobody _should_ have relied
on any of this anyway.

Of course, "should have" is just a dream world - but the point is that
if somebody complains, it's very fixable.

           Linus