[PATCH] AArch64: Use bitfield in cpu_features struct (BZ #32279)

[Florian Weimer]

* Wilco Dijkstra:

>> It's probably better to shuffle things around a bit so that it's the
>> MOPS bit that, when set, produces a trap representation when interpreted
>> as bool.  Having sve and prefer_sve_ifuncs at the same time seems far
>> more likely at this point.
>
> I'm not sure I'm following. The mops and prefer_sve_ifuncs values
> would be incorrect, however we crash much earlier because all fields
> in the dynamic linker are shifted by 8 bytes. So I'm not sure whether
> we could ever get to the point of resolving ifuncs...

Hmm.  Maybe I'm wrong.  Do we even initialize these fields in the
dormant instance of ld.so?

Thanks,
Florian