svn commit: r254666 - head/sys/amd64/amd64

[Konstantin Belousov]

Author: kib
Date: Thu Aug 22 18:05:31 2013
New Revision: 254666
URL: http://svnweb.freebsd.org/changeset/base/254666

Log:
  Use the generation count of the pv list to work around LOR between
  pmap lock and pv list lock, and use the shared locking on
  pvh_global_lock in pmap_remove_write(), same as it was done for
  pmap_ts_referenced().
  
  Noted and reviewed by:	alc (previous version)
  Tested by:	pho
  Sponsored by:	The FreeBSD Foundation

Modified:
  head/sys/amd64/amd64/pmap.c

Modified: head/sys/amd64/amd64/pmap.c
==============================================================================
--- head/sys/amd64/amd64/pmap.c	Thu Aug 22 17:33:50 2013	(r254665)
+++ head/sys/amd64/amd64/pmap.c	Thu Aug 22 18:05:31 2013	(r254666)
@@ -4765,10 +4765,12 @@ pmap_remove_write(vm_page_t m)
 {
 	struct md_page *pvh;
 	pmap_t pmap;
+	struct rwlock *lock;
 	pv_entry_t next_pv, pv;
 	pd_entry_t *pde;
 	pt_entry_t oldpte, *pte;
 	vm_offset_t va;
+	int pvh_gen, md_gen;
 
 	KASSERT((m->oflags & VPO_UNMANAGED) == 0,
 	    ("pmap_remove_write: page %p is not managed", m));
@@ -4781,23 +4783,51 @@ pmap_remove_write(vm_page_t m)
 	VM_OBJECT_ASSERT_WLOCKED(m->object);
 	if (!vm_page_xbusied(m) && (m->aflags & PGA_WRITEABLE) == 0)
 		return;
-	rw_wlock(&pvh_global_lock);
+	rw_rlock(&pvh_global_lock);
+	lock = VM_PAGE_TO_PV_LIST_LOCK(m);
+	pvh = pa_to_pvh(VM_PAGE_TO_PHYS(m));
+retry_pv_loop:
+	rw_wlock(lock);
 	if ((m->flags & PG_FICTITIOUS) != 0)
 		goto small_mappings;
-	pvh = pa_to_pvh(VM_PAGE_TO_PHYS(m));
 	TAILQ_FOREACH_SAFE(pv, &pvh->pv_list, pv_next, next_pv) {
 		pmap = PV_PMAP(pv);
-		PMAP_LOCK(pmap);
+		if (!PMAP_TRYLOCK(pmap)) {
+			pvh_gen = pvh->pv_gen;
+			rw_wunlock(lock);
+			PMAP_LOCK(pmap);
+			rw_wlock(lock);
+			if (pvh_gen != pvh->pv_gen) {
+				PMAP_UNLOCK(pmap);
+				rw_wunlock(lock);
+				goto retry_pv_loop;
+			}
+		}
 		va = pv->pv_va;
 		pde = pmap_pde(pmap, va);
 		if ((*pde & PG_RW) != 0)
-			(void)pmap_demote_pde(pmap, pde, va);
+			(void)pmap_demote_pde_locked(pmap, pde, va, &lock);
+		KASSERT(lock == VM_PAGE_TO_PV_LIST_LOCK(m),
+		    ("inconsistent pv lock %p %p for page %p",
+		    lock, VM_PAGE_TO_PV_LIST_LOCK(m), m));
 		PMAP_UNLOCK(pmap);
 	}
 small_mappings:
 	TAILQ_FOREACH(pv, &m->md.pv_list, pv_next) {
 		pmap = PV_PMAP(pv);
-		PMAP_LOCK(pmap);
+		if (!PMAP_TRYLOCK(pmap)) {
+			pvh_gen = pvh->pv_gen;
+			md_gen = m->md.pv_gen;
+			rw_wunlock(lock);
+			PMAP_LOCK(pmap);
+			rw_wlock(lock);
+			if (pvh_gen != pvh->pv_gen ||
+			    md_gen != m->md.pv_gen) {
+				PMAP_UNLOCK(pmap);
+				rw_wunlock(lock);
+				goto retry_pv_loop;
+			}
+		}
 		pde = pmap_pde(pmap, pv->pv_va);
 		KASSERT((*pde & PG_PS) == 0,
 		    ("pmap_remove_write: found a 2mpage in page %p's pv list",
@@ -4815,8 +4845,9 @@ retry:
 		}
 		PMAP_UNLOCK(pmap);
 	}
+	rw_wunlock(lock);
 	vm_page_aflag_clear(m, PGA_WRITEABLE);
-	rw_wunlock(&pvh_global_lock);
+	rw_runlock(&pvh_global_lock);
 }
 
 /*