Implement per-CPU caching for the page table page (vp) pool and the PTE

descriptor (pted) pool in the arm64 pmap implementation.  This
significantly reduces the side-effects of lock contention on the kernel
map lock that is (incorrectly) translated into excessive page daemon
wakeups.  This is not a perfect solution but it does lead to significant
speedups on machines with many CPU cores.

This requires adding a new pmap_init_percpu() function that gets called
at the point where kernel is ready to set up the per-CPU pool caches.
Dummy implementations of this function are added for all non-arm64
architectures.  Some other architectures can probably benefit from
providing an actual implementation that sets up per-CPU caches for
pmap pools as well.

ok phessler@, claudio@, miod@, patrick@

1 files changed, 2 insertions, 1 deletions

diff --git a/sys/arch/alpha/include/pmap.h b/sys/arch/alpha/include/pmap.h
index 5eddbfad573..af2437cfd22 100644
--- a/sys/arch/alpha/include/pmap.h
+++ b/sys/arch/alpha/include/pmap.h
@@ -1,4 +1,4 @@
-/* $OpenBSD: pmap.h,v 1.45 2023/04/13 15:23:21 miod Exp $ */
+/* $OpenBSD: pmap.h,v 1.46 2023/12/11 22:12:52 kettenis Exp $ */
 /* $NetBSD: pmap.h,v 1.37 2000/11/19 03:16:35 thorpej Exp $ */
 
 /*-
@@ -158,6 +158,7 @@ void	pmap_do_tlb_shootdown(struct cpu_info *, struct trapframe *);
 #define pmap_update(pmap)		/* nothing (yet) */
 
 #define pmap_proc_iflush(p, va, len)	/* nothing */
+#define pmap_init_percpu()		do { /* nothing */ } while (0)
 #define pmap_unuse_final(p)		/* nothing */
 #define	pmap_remove_holes(vm)		do { /* nothing */ } while (0)