On a dual socket POWER9:

@luporl we want a LE 64-bit number here.

Since mail/thunderbird now builds and works on powerpc64, is this OBE? Or is more needed?

Looks fine to me

Maybe after it's in, someone can try to figure out why enabling superpages causes a 2% performance penalty in buildworld.

Maybe we should stop playing whack-a-mole and just use an overkill bounce page that can hold the full name field (2048 bytes?)

Is there any more work needed on this?

I like the refactoring!

Seems fine to me. It's consistent with all other archs.

Looks good, glad you got the speed back!

This is mostly a series of experiments to see what can reduce the locking contention on the 'page pv' lock (PV_LOCK()/PV_PAGE_LOCK()) since those seem to be pretty heavily contended, as measured with lockstat(1).

Move pc_flags to a better spot.

One last thing: did you try a debug session between the initrd syms setup and pmap_bootstrap() to make sure it doesn't panic due to being unable to access the DMAP?

Try again...

Fix booke, which doesn't use ifuncs for copyin/copyout. Use a "naked function"
definition instead, since it's really treated as a secondary return path.

Address bdragon's feedback.

Fix the flag value for PC_FLAG_NOSRS.

Once more, with compile

Upload the real diff...

Committed as r361544.

Apply to ps3.

Fix bugs caused by my misunderstanding of how ifuncs work.

I think this is now OBE.

Copy the pc_flags before unleashing the CPUs, instead of recomputing on each CPU.