svn commit: r350427 - head/sys/arm64/arm64

[Mark Johnston]

Author: markj
Date: Mon Jul 29 21:21:53 2019
New Revision: 350427
URL: https://svnweb.freebsd.org/changeset/base/350427

Log:
  Have arm64's pmap_fault() handle WnR faults on dirty PTEs.
  
  If we take a WnR permission fault on a managed, writeable and dirty
  PTE, simply return success without calling the main fault handler.  This
  situation can occur if multiple threads simultaneously access a clean
  writeable mapping and trigger WnR faults; losers of the race to mark the
  PTE dirty would end up calling the main fault handler, which had no work
  to do.
  
  Reported by:	alc
  Reviewed by:	alc
  MFC with:	r350004
  Sponsored by:	The FreeBSD Foundation
  Differential Revision:	https://reviews.freebsd.org/D21097

Modified:
  head/sys/arm64/arm64/pmap.c

Modified: head/sys/arm64/arm64/pmap.c
==============================================================================
--- head/sys/arm64/arm64/pmap.c	Mon Jul 29 20:50:26 2019	(r350426)
+++ head/sys/arm64/arm64/pmap.c	Mon Jul 29 21:21:53 2019	(r350427)
@@ -5743,7 +5743,7 @@ pmap_sync_icache(pmap_t pmap, vm_offset_t va, vm_size_
 int
 pmap_fault(pmap_t pmap, uint64_t esr, uint64_t far)
 {
-	pt_entry_t *pte;
+	pt_entry_t pte, *ptep;
 	register_t intr;
 	uint64_t ec, par;
 	int lvl, rv;
@@ -5767,9 +5767,9 @@ pmap_fault(pmap_t pmap, uint64_t esr, uint64_t far)
 	case ISS_DATA_DFSC_AFF_L2:
 	case ISS_DATA_DFSC_AFF_L3:
 		PMAP_LOCK(pmap);
-		pte = pmap_pte(pmap, far, &lvl);
-		if (pte != NULL) {
-			pmap_set_bits(pte, ATTR_AF);
+		ptep = pmap_pte(pmap, far, &lvl);
+		if (ptep != NULL) {
+			pmap_set_bits(ptep, ATTR_AF);
 			rv = KERN_SUCCESS;
 			/*
 			 * XXXMJ as an optimization we could mark the entry
@@ -5785,12 +5785,13 @@ pmap_fault(pmap_t pmap, uint64_t esr, uint64_t far)
 		    (esr & ISS_DATA_WnR) == 0)
 			return (rv);
 		PMAP_LOCK(pmap);
-		pte = pmap_pte(pmap, far, &lvl);
-		if (pte != NULL &&
-		    (pmap_load(pte) & (ATTR_AP_RW_BIT | ATTR_SW_DBM)) ==
-		    (ATTR_AP(ATTR_AP_RO) | ATTR_SW_DBM)) {
-			pmap_clear_bits(pte, ATTR_AP_RW_BIT);
-			pmap_invalidate_page(pmap, far);
+		ptep = pmap_pte(pmap, far, &lvl);
+		if (ptep != NULL &&
+		    ((pte = pmap_load(ptep)) & ATTR_SW_DBM) != 0) {
+			if ((pte & ATTR_AP_RW_BIT) == ATTR_AP(ATTR_AP_RO)) {
+				pmap_clear_bits(ptep, ATTR_AP_RW_BIT);
+				pmap_invalidate_page(pmap, far);
+			}
 			rv = KERN_SUCCESS;
 		}
 		PMAP_UNLOCK(pmap);