Rename pmap_collect() to pmap_pv_reclaim() and rewrite it such that it no

longer uses the active and inactive paging queues.  Instead, the pmap now
maintains an LRU-ordered list of pv entry pages, and pmap_pv_reclaim() uses
this list to select pv entries for reclamation.

Note: The old pmap_collect() tried to avoid reclaiming mappings for pages
that have either a hold_count or a busy field that is non-zero.  However,
this isn't necessary for correctness, and the locking in pmap_collect() was
insufficient to guarantee that such mappings weren't reclaimed.  The new
pmap_pv_reclaim() doesn't even try.

Tested by:	sbruno
MFC after:	5 weeks

1 files changed, 121 insertions, 59 deletions

diff --git a/sys/i386/xen/pmap.c b/sys/i386/xen/pmap.c
index 189f311..0f25e3c 100644
--- a/sys/i386/xen/pmap.c
+++ b/sys/i386/xen/pmap.c
@@ -230,6 +230,7 @@ static int pat_works;			/* Is page attribute table sane? */
 /*
  * Data for the pv entry allocation mechanism
  */
+static TAILQ_HEAD(pch, pv_chunk) pv_chunks = TAILQ_HEAD_INITIALIZER(pv_chunks);
 static int pv_entry_count = 0, pv_entry_max = 0, pv_entry_high_water = 0;
 static int shpgperproc = PMAP_SHPGPERPROC;
 
@@ -1958,64 +1959,133 @@ SYSCTL_LONG(_vm_pmap, OID_AUTO, pv_entry_allocs, CTLFLAG_RD, &pv_entry_allocs, 0
 	"Current number of pv entry allocs");
 SYSCTL_INT(_vm_pmap, OID_AUTO, pv_entry_spare, CTLFLAG_RD, &pv_entry_spare, 0,
 	"Current number of spare pv entries");
-
-static int pmap_collect_inactive, pmap_collect_active;
-
-SYSCTL_INT(_vm_pmap, OID_AUTO, pmap_collect_inactive, CTLFLAG_RD, &pmap_collect_inactive, 0,
-	"Current number times pmap_collect called on inactive queue");
-SYSCTL_INT(_vm_pmap, OID_AUTO, pmap_collect_active, CTLFLAG_RD, &pmap_collect_active, 0,
-	"Current number times pmap_collect called on active queue");
 #endif
 
 /*
  * We are in a serious low memory condition.  Resort to
  * drastic measures to free some pages so we can allocate
- * another pv entry chunk.  This is normally called to
- * unmap inactive pages, and if necessary, active pages.
+ * another pv entry chunk.
  */
-static void
-pmap_collect(pmap_t locked_pmap, struct vpgqueues *vpq)
+static vm_page_t
+pmap_pv_reclaim(pmap_t locked_pmap)
 {
+	struct pch newtail;
+	struct pv_chunk *pc;
 	pmap_t pmap;
 	pt_entry_t *pte, tpte;
-	pv_entry_t next_pv, pv;
+	pv_entry_t pv;
 	vm_offset_t va;
-	vm_page_t m, free;
-
+	vm_page_t free, m, m_pc;
+	uint32_t inuse, freemask;
+	int bit, field, freed;
+
+	PMAP_LOCK_ASSERT(locked_pmap, MA_OWNED);
+	pmap = NULL;
+	free = m_pc = NULL;
+	TAILQ_INIT(&newtail);
 	sched_pin();
-	TAILQ_FOREACH(m, &vpq->pl, pageq) {
-		if ((m->flags & PG_MARKER) != 0 || m->hold_count || m->busy)
-			continue;
-		TAILQ_FOREACH_SAFE(pv, &m->md.pv_list, pv_list, next_pv) {
-			va = pv->pv_va;
-			pmap = PV_PMAP(pv);
+	while ((pc = TAILQ_FIRST(&pv_chunks)) != NULL && (pv_vafree == 0 ||
+	    free == NULL)) {
+		TAILQ_REMOVE(&pv_chunks, pc, pc_lru);
+		if (pmap != pc->pc_pmap) {
+			if (pmap != NULL) {
+				pmap_invalidate_all(pmap);
+				if (pmap != locked_pmap)
+					PMAP_UNLOCK(pmap);
+			}
+			pmap = pc->pc_pmap;
 			/* Avoid deadlock and lock recursion. */
 			if (pmap > locked_pmap)
 				PMAP_LOCK(pmap);
-			else if (pmap != locked_pmap && !PMAP_TRYLOCK(pmap))
+			else if (pmap != locked_pmap && !PMAP_TRYLOCK(pmap)) {
+				pmap = NULL;
+				TAILQ_INSERT_TAIL(&newtail, pc, pc_lru);
 				continue;
-			pmap->pm_stats.resident_count--;
-			pte = pmap_pte_quick(pmap, va);
-			tpte = pte_load_clear(pte);
-			KASSERT((tpte & PG_W) == 0,
-			    ("pmap_collect: wired pte %#jx", (uintmax_t)tpte));
-			if (tpte & PG_A)
-				vm_page_aflag_set(m, PGA_REFERENCED);
-			if ((tpte & (PG_M | PG_RW)) == (PG_M | PG_RW))
-				vm_page_dirty(m);
-			free = NULL;
-			pmap_unuse_pt(pmap, va, &free);
-			pmap_invalidate_page(pmap, va);
-			pmap_free_zero_pages(free);
-			TAILQ_REMOVE(&m->md.pv_list, pv, pv_list);
-			free_pv_entry(pmap, pv);
-			if (pmap != locked_pmap)
-				PMAP_UNLOCK(pmap);
+			}
+		}
+
+		/*
+		 * Destroy every non-wired, 4 KB page mapping in the chunk.
+		 */
+		freed = 0;
+		for (field = 0; field < _NPCM; field++) {
+			freemask = 0;
+			for (inuse = ~pc->pc_map[field] & pc_freemask[field];
+			    inuse != 0; inuse &= ~(1UL << bit)) {
+				bit = bsfl(inuse);
+				pv = &pc->pc_pventry[field * 32 + bit];
+				va = pv->pv_va;
+				pte = pmap_pte_quick(pmap, va);
+				if ((*pte & PG_W) != 0)
+					continue;
+				tpte = pte_load_clear(pte);
+				if ((tpte & PG_G) != 0)
+					pmap_invalidate_page(pmap, va);
+				m = PHYS_TO_VM_PAGE(tpte & PG_FRAME);
+				if ((tpte & (PG_M | PG_RW)) == (PG_M | PG_RW))
+					vm_page_dirty(m);
+				if ((tpte & PG_A) != 0)
+					vm_page_aflag_set(m, PGA_REFERENCED);
+				TAILQ_REMOVE(&m->md.pv_list, pv, pv_list);
+				if (TAILQ_EMPTY(&m->md.pv_list))
+					vm_page_aflag_clear(m, PGA_WRITEABLE);
+				pmap_unuse_pt(pmap, va, &free);
+				freemask |= 1UL << bit;
+				freed++;
+			}
+			pc->pc_map[field] |= freemask;
+		}
+		if (freed == 0) {
+			TAILQ_INSERT_TAIL(&newtail, pc, pc_lru);
+			continue;
+		}
+		pmap->pm_stats.resident_count -= freed;
+		PV_STAT(pv_entry_frees += freed);
+		PV_STAT(pv_entry_spare += freed);
+		pv_entry_count -= freed;
+		TAILQ_REMOVE(&pmap->pm_pvchunk, pc, pc_list);
+		for (field = 0; field < _NPCM; field++)
+			if (pc->pc_map[field] != pc_freemask[field]) {
+				TAILQ_INSERT_HEAD(&pmap->pm_pvchunk, pc,
+				    pc_list);
+				TAILQ_INSERT_TAIL(&newtail, pc, pc_lru);
+
+				/*
+				 * One freed pv entry in locked_pmap is
+				 * sufficient.
+				 */
+				if (pmap == locked_pmap)
+					goto out;
+				break;
+			}
+		if (field == _NPCM) {
+			PV_STAT(pv_entry_spare -= _NPCPV);
+			PV_STAT(pc_chunk_count--);
+			PV_STAT(pc_chunk_frees++);
+			/* Entire chunk is free; return it. */
+			m_pc = PHYS_TO_VM_PAGE(pmap_kextract((vm_offset_t)pc));
+			pmap_qremove((vm_offset_t)pc, 1);
+			pmap_ptelist_free(&pv_vafree, (vm_offset_t)pc);
+			break;
 		}
-		if (TAILQ_EMPTY(&m->md.pv_list))
-			vm_page_aflag_clear(m, PGA_WRITEABLE);
 	}
+out:
 	sched_unpin();
+	TAILQ_CONCAT(&pv_chunks, &newtail, pc_lru);
+	if (pmap != NULL) {
+		pmap_invalidate_all(pmap);
+		if (pmap != locked_pmap)
+			PMAP_UNLOCK(pmap);
+	}
+	if (m_pc == NULL && pv_vafree != 0 && free != NULL) {
+		m_pc = free;
+		free = m_pc->right;
+		/* Recycle a freed page table page. */
+		m_pc->wire_count = 1;
+		atomic_add_int(&cnt.v_wire_count, 1);
+	}
+	pmap_free_zero_pages(free);
+	return (m_pc);
 }
 
 
@@ -2046,6 +2116,7 @@ free_pv_entry(pmap_t pmap, pv_entry_t pv)
 			TAILQ_INSERT_HEAD(&pmap->pm_pvchunk, pc, pc_list);
 			return;
 		}
+ 	TAILQ_REMOVE(&pv_chunks, pc, pc_lru);
 	PV_STAT(pv_entry_spare -= _NPCPV);
 	PV_STAT(pc_chunk_count--);
 	PV_STAT(pc_chunk_frees++);
@@ -2066,7 +2137,6 @@ get_pv_entry(pmap_t pmap, int try)
 {
 	static const struct timeval printinterval = { 60, 0 };
 	static struct timeval lastprint;
-	struct vpgqueues *pq;
 	int bit, field;
 	pv_entry_t pv;
 	struct pv_chunk *pc;
@@ -2081,7 +2151,6 @@ get_pv_entry(pmap_t pmap, int try)
 			printf("Approaching the limit on PV entries, consider "
 			    "increasing either the vm.pmap.shpgperproc or the "
 			    "vm.pmap.pv_entry_max tunable.\n");
-	pq = NULL;
 retry:
 	pc = TAILQ_FIRST(&pmap->pm_pvchunk);
 	if (pc != NULL) {
@@ -2102,6 +2171,10 @@ retry:
 				}
 			TAILQ_REMOVE(&pmap->pm_pvchunk, pc, pc_list);
 			TAILQ_INSERT_TAIL(&pmap->pm_pvchunk, pc, pc_list);
+			if (pc != TAILQ_LAST(&pv_chunks, pch)) {
+				TAILQ_REMOVE(&pv_chunks, pc, pc_lru);
+				TAILQ_INSERT_TAIL(&pv_chunks, pc, pc_lru);
+			}
 			PV_STAT(pv_entry_spare--);
 			return (pv);
 		}
@@ -2111,29 +2184,16 @@ retry:
 	 * queues lock.  If "pv_vafree" is currently non-empty, it will
 	 * remain non-empty until pmap_ptelist_alloc() completes.
 	 */
-	if (pv_vafree == 0 || (m = vm_page_alloc(NULL, 0, (pq ==
-	    &vm_page_queues[PQ_ACTIVE] ? VM_ALLOC_SYSTEM : VM_ALLOC_NORMAL) |
+	if (pv_vafree == 0 || (m = vm_page_alloc(NULL, 0, VM_ALLOC_NORMAL |
 	    VM_ALLOC_NOOBJ | VM_ALLOC_WIRED)) == NULL) {
 		if (try) {
 			pv_entry_count--;
 			PV_STAT(pc_chunk_tryfail++);
 			return (NULL);
 		}
-		/*
-		 * Reclaim pv entries: At first, destroy mappings to
-		 * inactive pages.  After that, if a pv chunk entry
-		 * is still needed, destroy mappings to active pages.
-		 */
-		if (pq == NULL) {
-			PV_STAT(pmap_collect_inactive++);
-			pq = &vm_page_queues[PQ_INACTIVE];
-		} else if (pq == &vm_page_queues[PQ_INACTIVE]) {
-			PV_STAT(pmap_collect_active++);
-			pq = &vm_page_queues[PQ_ACTIVE];
-		} else
-			panic("get_pv_entry: increase vm.pmap.shpgperproc");
-		pmap_collect(pmap, pq);
-		goto retry;
+		m = pmap_pv_reclaim(pmap);
+		if (m == NULL)
+			goto retry;
 	}
 	PV_STAT(pc_chunk_count++);
 	PV_STAT(pc_chunk_allocs++);
@@ -2145,6 +2205,7 @@ retry:
 	pc->pc_map[0] = pc_freemask[0] & ~1ul;	/* preallocated bit 0 */
 	for (field = 1; field < _NPCM; field++)
 		pc->pc_map[field] = pc_freemask[field];
+	TAILQ_INSERT_TAIL(&pv_chunks, pc, pc_lru);
 	pv = &pc->pc_pventry[0];
 	TAILQ_INSERT_HEAD(&pmap->pm_pvchunk, pc, pc_list);
 	PV_STAT(pv_entry_spare += _NPCPV - 1);
@@ -3535,6 +3596,7 @@ pmap_remove_pages(pmap_t pmap)
 			PV_STAT(pc_chunk_count--);
 			PV_STAT(pc_chunk_frees++);
 			TAILQ_REMOVE(&pmap->pm_pvchunk, pc, pc_list);
+			TAILQ_REMOVE(&pv_chunks, pc, pc_lru);
 			m = PHYS_TO_VM_PAGE(pmap_kextract((vm_offset_t)pc));
 			pmap_qremove((vm_offset_t)pc, 1);
 			vm_page_unwire(m, 0);