Ensure 'struct thread' is aligned to a cache line

[olce]

Summary

Using the new UMA_ALIGN_CACHE_AND_MASK() facility, which allows to
simultaneously guarantee a minimum of 32 bytes of alignment (the 5 lower
bits are always 0).

For the record, to this day, here's a (possibly non-exhaustive) list of
synchronization primitives using lower bits to store flags in pointers
to thread structures:

• lockmgr, rwlock and sx all use the 5 bits directly.
• rmlock indirectly relies on sx, so can use the 5 bits.
• mtx (non-spin) relies on the 3 lower bits.

MFC after: 2 week
Sponsored by: The FreeBSD Foundation

[olce]

This is preliminary work for struct thread fields re-ordering (standalone, this commit doesn't have that much value; but older commit messages mention the intention to do it).

[markj]

sys/kern/kern_thread.c

579

This would be a good time to add a comment explaining why we require some minimum alignment (synchronization primitives expect to be able to use low bits of a thread address).

[olce]

Add a comment explaining the reason for the minimal alignment constraint.

[kib]

sys/kern/kern_thread.c

573

575

I do not think that listing the primitives and exact bits used is needed there. In particular, nobody would even guess that there is such place to fix if doing a work on the locks.

[markj]

sys/kern/kern_thread.c

575

I agree, we do not need to be so specific, the comment will just become stale over time.

[olce]

More generic comment. Move the current list of synchronization primitives to the commit message.

[olce]

sys/kern/kern_thread.c

575

I certainly would look at the alignment for allocated thread structures if I had to use one more bit in one of the synchronization primitives. And probably would I look here as well if I was removing the use of one of them. Plus I suspect these don't change very often. But OK.