Hello all, The first part of this patch is wrong: it changes from an unsigned long param to __fls() to an unsigned int param in fls(). One option would be to add another implementation of flsl() or given the minimalistic usage of int_sqrt() in the kernel to keep the current code with __fls() and only change masking the lowest bit.
The only other nitpick I have come up with: Adding __attribute_const__ could be possible. best regards, Florian La Roche
