 @@ -1,14 +1,14 @@
-/*	$NetBSD: pmap.c,v 1.77.2.4 2009/07/23 23:31:37 jym Exp $	*/
+/*	$NetBSD: pmap.c,v 1.77.2.5 2009/07/24 11:30:28 jym Exp $	*/
 /*
  * Copyright (c) 2007 Manuel Bouyer.
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
  * 1. Redistributions of source code must retain the above copyright
  *    notice, this list of conditions and the following disclaimer.
  * 2. Redistributions in binary form must reproduce the above copyright
  *    notice, this list of conditions and the following disclaimer in the
  *    documentation and/or other materials provided with the distribution.
  * 3. All advertising materials mentioning features or use of this software
 @@ -144,27 +144,27 @@
  *     Hibler/Jolitz pmap, as modified for FreeBSD by John S. Dyson
  *     and David Greenman.
+ *
  * [3] the Mach pmap.   this pmap, from CMU, seems to have migrated
  *     between several processors.   the VAX version was done by
  *     Avadis Tevanian, Jr., and Michael Wayne Young.    the i386
  *     version was done by Lance Berc, Mike Kupfer, Bob Baron,
  *     David Golub, and Richard Draves.    the alpha version was
  *     done by Alessandro Forin (CMU/Mach) and Chris Demetriou
  *     (NetBSD/alpha).
  */
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: pmap.c,v 1.77.2.4 2009/07/23 23:31:37 jym Exp $");
+__KERNEL_RCSID(0, "$NetBSD: pmap.c,v 1.77.2.5 2009/07/24 11:30:28 jym Exp $");
 #include "opt_user_ldt.h"
 #include "opt_lockdebug.h"
 #include "opt_multiprocessor.h"
 #include "opt_xen.h"
 #if !defined(__x86_64__)
 #include "opt_kstack_dr0.h"
 #endif /* !defined(__x86_64__) */
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/proc.h>
 #include <sys/pool.h>
 @@ -547,26 +547,32 @@ static char *csrcp, *cdstp, *zerop, *ptp
 static struct pool_cache pmap_pdp_cache;
 int	pmap_pdp_ctor(void *, void *, int);
 void	pmap_pdp_dtor(void *, void *);
 #ifdef PAE
 /* need to allocate items of 4 pages */
 void *pmap_pdp_alloc(struct pool *, int);
 void pmap_pdp_free(struct pool *, void *);
 static struct pool_allocator pmap_pdp_allocator = {
 	.pa_alloc = pmap_pdp_alloc,
 	.pa_free = pmap_pdp_free,
 	.pa_pagesz = PAGE_SIZE * PDP_SIZE,
 };
 #ifdef XEN
 /* pool for allocation of the structures tracking L2 pinned shadow pages */
 //static struct pool_cache l2_pdirpte_cache;
 #endif /* XEN */
 #endif /* PAE */
 void *vmmap; /* XXX: used by mem.c... it should really uvm_map_reserve it */
 extern vaddr_t idt_vaddr;			/* we allocate IDT early */
 extern paddr_t idt_paddr;
 #ifdef _LP64
 extern vaddr_t lo32_vaddr;
 extern vaddr_t lo32_paddr;
 #endif
 extern int end;
 @@ -742,61 +748,138 @@ pmap_is_active(struct pmap *pmap, struct
+}
 /*
  * Flush the content of APDP_PDE
  */
 static inline
 void pmap_unmap_apdp_pde(void) {
 	int i;
 	for (i = 0; i < PDP_SIZE; i++) {
 		pmap_pte_set(&APDP_PDE[i], 0);
 #ifdef PAE
-		/* clear shadow entry too */
+		/* clear current pmap shadow entries too */
     		pmap_pte_set(&APDP_PDE_SHADOW[i], 0);
 #endif
+	}
+}
 #ifdef XEN3
 /*
  * Flush all APDP entries found in pmaps
  * Required during Xen save/restore operations, as it does not
  * handle alternative recursive mappings properly
  */
 void
 pmap_unmap_all_apdp_pdes(void) {
-	// XXX JYM PAE
+	int i;
 	int s;
 	struct pmap *pm;
 	s = splvm();
 #ifdef PAE
 	/*
 	 * For PAE, there are two places where alternative recursive mappings
 	 * could be found: in the L2 shadow pages, and the "real" L2 kernel
 	 * page (pmap_kl2pd), which is unique and static.
 	 * We first clear the APDP for the current pmap. As L2 kernel page is
 	 * unique, we only need to do it once for all pmaps.
 	 */
 	pmap_unmap_apdp_pde();
 #endif
 	mutex_enter(&pmaps_lock);
 	/*
 	 * Set APDP entries to 0 in all pmaps.
 	 * Note that for PAE kernels, this only clears the APDP entries
 	 * found in the L2 shadow pages, as pmap_pdirpa() is used to obtain
 	 * the PA of the pmap->pm_pdir[] pages (forming the 4 contiguous
 	 * pages of PAE PD: 3 for user space, 1 for the L2 kernel shadow page)
 	 */
 	LIST_FOREACH(pm, &pmaps, pm_list) {
 		for (i = 0; i < PDP_SIZE; i++) {
 			xpq_queue_pte_update(
 			    xpmap_ptom(pmap_pdirpa(pm, PDIR_SLOT_APTE + i)), 0);
+		}
+	}
 	mutex_exit(&pmaps_lock);
 	xpq_flush_queue();
 	splx(s);
+}
 #ifdef PAE
 /*
  * NetBSD uses L2 shadow pages to support PAE with Xen. However, Xen does not
  * handle them correctly during save/restore, leading to incorrect page
  * tracking and pinning during restore.
  * For save/restore to succeed, two functions are introduced:
  * - pmap_map_shadow_entries(), used by resume code to set the last entry
  *   of PDIR_SLOT_PTE so that it points to the correct L2 shadow page
  *   entries to their correct values, so that pmap works again.
  * - pmap_unmap_shadow_entries(), used by suspend code to clear all
  *   PDIR_SLOT_PTE entries pointing to L2 shadow entries
  */
 void
 pmap_map_shadow_entries(void) {
 	struct pmap *pm;
 	mutex_enter(&pmaps_lock);
 	LIST_FOREACH(pm, &pmaps, pm_list) {
 		xpq_queue_pte_update(
-		    xpmap_ptom(pmap_pdirpa(pm, PDIR_SLOT_APTE)),
+		    xpmap_ptom(pmap_pdirpa(pm, PDIR_SLOT_PTE + 3)),
-);
+		    xpmap_ptom(pmap_pdirpa(pm, PDIR_SLOT_KERN)) | PG_V);
+	}
 	mutex_exit(&pmaps_lock);
 	xpq_queue_pte_update(
 	    xpmap_ptom(pmap_pdirpa(pmap_kernel(), PDIR_SLOT_PTE + 3)),
 	    xpmap_ptom(pmap_pdirpa(pmap_kernel(), PDIR_SLOT_KERN)) | PG_V);
 	xpq_flush_queue();
+}
 void
 pmap_unmap_shadow_entries(void) {
 	struct pmap *pm;
 	mutex_enter(&pmaps_lock);
 	LIST_FOREACH(pm, &pmaps, pm_list) {
 		xpq_queue_pte_update(
 		    xpmap_ptom(pmap_pdirpa(pm, PDIR_SLOT_PTE + 3)), 0);
+	}
 	mutex_exit(&pmaps_lock);
-	splx(s);
+	/* do it for pmap_kernel() too! */
 	xpq_queue_pte_update(
 	    xpmap_ptom(pmap_pdirpa(pmap_kernel(), PDIR_SLOT_PTE + 3)), 0);
 	xpq_flush_queue();
+}
 #endif /* PAE */
 #endif /* XEN3 */
 static void
 pmap_apte_flush(struct pmap *pmap)
+{
 	KASSERT(kpreempt_disabled());
 	/*
 	 * Flush the APTE mapping from all other CPUs that
 	 * are using the pmap we are using (who's APTE space
 	 * is the one we've just modified).
+	 *
 @@ -1559,27 +1642,27 @@ pmap_bootstrap(vaddr_t kva_start)
 	 * when deactivate a pmap, Xen will still consider it active.
 	 * So we set user PGD to this one to lift all protection on
 	 * the now inactive page tables set.
 	 */
 	xen_dummy_user_pgd = avail_start;
 	avail_start += PAGE_SIZE;
 	/* Zero fill it, the less checks in Xen it requires the better */
 	memset((void *) (xen_dummy_user_pgd + KERNBASE), 0, PAGE_SIZE);
 	/* Mark read-only */
 	HYPERVISOR_update_va_mapping(xen_dummy_user_pgd + KERNBASE,
 	    pmap_pa2pte(xen_dummy_user_pgd) | PG_u | PG_V, UVMF_INVLPG);
 	/* Pin as L4 */
-	xpq_queue_pin_table(xpmap_ptom_masked(xen_dummy_user_pgd));
+	xpq_queue_pin_l4_table(xpmap_ptom_masked(xen_dummy_user_pgd));
 #endif /* __x86_64__ */
 	idt_vaddr = virtual_avail;                      /* don't need pte */
 	idt_paddr = avail_start;                        /* steal a page */
 	/*
 	 * Xen require one more page as we can't store
 	 * GDT and LDT on the same page
 	 */
 	virtual_avail += 3 * PAGE_SIZE;
 	avail_start += 3 * PAGE_SIZE;
 #else /* XEN */
 	idt_vaddr = virtual_avail;			/* don't need pte */
 	idt_paddr = avail_start;			/* steal a page */
 #if defined(__x86_64__)
 @@ -1620,26 +1703,30 @@ pmap_bootstrap(vaddr_t kva_start)
 	 */
 	mutex_init(&pmaps_lock, MUTEX_DEFAULT, IPL_NONE);
 	LIST_INIT(&pmaps);
 	pmap_cpu_init_early(curcpu());
 	/*
 	 * initialize caches.
 	 */
 	pool_cache_bootstrap(&pmap_cache, sizeof(struct pmap), 0, 0, 0,
 	    "pmappl", NULL, IPL_NONE, NULL, NULL, NULL);
 #ifdef PAE
 #ifdef XEN
 	//pool_cache_bootstrap(&l2_pdirpte_cache, sizeof(struct l2_pdirpte), 0, 0,
 	//    0, "l2_pdirptepl", NULL, IPL_NONE, NULL, NULL, NULL);
 #endif /* XEN */
 	pool_cache_bootstrap(&pmap_pdp_cache, PAGE_SIZE * PDP_SIZE, 0, 0, 0,
 	    "pdppl", &pmap_pdp_allocator, IPL_NONE,
 	    pmap_pdp_ctor, pmap_pdp_dtor, NULL);
 #else /* PAE */
 	pool_cache_bootstrap(&pmap_pdp_cache, PAGE_SIZE, 0, 0, 0,
 	    "pdppl", NULL, IPL_NONE, pmap_pdp_ctor, pmap_pdp_dtor, NULL);
 #endif /* PAE */
 	pool_cache_bootstrap(&pmap_pv_cache, sizeof(struct pv_entry), 0, 0,
 	    PR_LARGECACHE, "pvpl", &pool_allocator_meta, IPL_NONE, NULL,
 	    NULL, NULL);
 	/*
 	 * ensure the TLB is sync'd with reality by flushing it...
 @@ -2212,32 +2299,32 @@ pmap_pdp_ctor(void *arg, void *v, int fl
 	for (i = 0; i < PDP_SIZE; i++, object += PAGE_SIZE) {
 		(void) pmap_extract(pmap_kernel(), object, &pdirpa);
 		/* remap this page RO */
 		pmap_kenter_pa(object, pdirpa, VM_PROT_READ);
 		pmap_update(pmap_kernel());
 		/*
 		 * pin as L2/L4 page, we have to do the page with the
 		 * PDIR_SLOT_PTE entries last
 		 */
 #ifdef PAE
 		if (i == l2tol3(PDIR_SLOT_PTE))
 			continue;
 #endif
-		xpq_queue_pin_table(xpmap_ptom_masked(pdirpa));
+		xpq_queue_pin_l2_table(xpmap_ptom_masked(pdirpa));
+	}
 #ifdef PAE
 	object = ((vaddr_t)pdir) + PAGE_SIZE  * l2tol3(PDIR_SLOT_PTE);
 	(void)pmap_extract(pmap_kernel(), object, &pdirpa);
-	xpq_queue_pin_table(xpmap_ptom_masked(pdirpa));
+	xpq_queue_pin_l2_table(xpmap_ptom_masked(pdirpa));
 #endif
 	xpq_flush_queue();
 	splx(s);
 	xen_release_ptom_lock();
 #endif /* XEN */
 	return (0);
+}
 /*
  * pmap_pdp_dtor: destructor for the PDP cache.
  */
 @@ -2302,28 +2389,32 @@ pmap_pdp_free(struct pool *pp, void *v)
 /*
  * pmap_create: create a pmap
+ *
  * => note: old pmap interface took a "size" args which allowed for
  *	the creation of "software only" pmaps (not in bsd).
  */
 struct pmap *
 pmap_create(void)
+{
 	struct pmap *pmap;
 	int i;
 #ifdef PAE
 	//struct l2_pdirpte *l2s;
 #endif /* PAE */
 	pmap = pool_cache_get(&pmap_cache, PR_WAITOK);
 	//l2p  = pool_cache_get(&l2_pdirpte_cache, PR_WAITOK);
 	/* init uvm_object */
 	for (i = 0; i < PTP_LEVELS - 1; i++) {
 		UVM_OBJ_INIT(&pmap->pm_obj[i], NULL, 1);
 		pmap->pm_ptphint[i] = NULL;
+	}
 	pmap->pm_stats.wired_count = 0;
 	pmap->pm_stats.resident_count = 1;	/* count the PDP allocd below */
 #if !defined(__x86_64__)
 	pmap->pm_hiexec = 0;
 #endif /* !defined(__x86_64__) */
 	pmap->pm_flags = 0;
 	pmap->pm_cpus = 0;
 @@ -2344,26 +2435,34 @@ pmap_create(void)
 		mutex_exit(&pmaps_lock);
 		pool_cache_destruct_object(&pmap_pdp_cache, pmap->pm_pdir);
 		goto try_again;
+	}
 #ifdef XEN3
 	xen_acquire_reader_ptom_lock();
 #endif
 #ifdef PAE
 	for (i = 0; i < PDP_SIZE; i++)
 		pmap->pm_pdirpa[i] =
 		    pmap_pte2pa(pmap->pm_pdir[PDIR_SLOT_PTE + i]);
 	/*
 	 * We now have an L2 page with a valid PDIR_SLOT_PTE pointing
 	 * to a shadow L2, so track it
 	l2p->slot_pte = pdirpa;
 	l2p->slot_pte_content = pmap_pte2pa(pmap->pm_pdir[PDIR_SLOT_PTE + i]);
 	 */
 #else
 	pmap->pm_pdirpa = pmap_pte2pa(pmap->pm_pdir[PDIR_SLOT_PTE]);
 #endif
 #ifdef XEN3
 	xen_release_ptom_lock();
 #endif
 	LIST_INSERT_HEAD(&pmaps, pmap, pm_list);
 	mutex_exit(&pmaps_lock);
 	return (pmap);

 @@ -1,14 +1,14 @@
-/*	$NetBSD: xenpmap.h,v 1.21.8.2 2009/05/13 17:18:50 jym Exp $	*/
+/*	$NetBSD: xenpmap.h,v 1.21.8.3 2009/07/24 11:30:28 jym Exp $	*/
 /*
+ *
  * Copyright (c) 2004 Christian Limpach.
  * All rights reserved.
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
  * 1. Redistributions of source code must retain the above copyright
  *    notice, this list of conditions and the following disclaimer.
  * 2. Redistributions in binary form must reproduce the above copyright
  *    notice, this list of conditions and the following disclaimer in the
 @@ -35,30 +35,54 @@
 #ifndef _XEN_XENPMAP_H_
 #define _XEN_XENPMAP_H_
 #include "opt_xen.h"
 #define	INVALID_P2M_ENTRY	(~0UL)
 void xpq_queue_machphys_update(paddr_t, paddr_t);
 void xpq_queue_invlpg(vaddr_t);
 void xpq_queue_pte_update(paddr_t, pt_entry_t);
 void xpq_queue_pt_switch(paddr_t);
 void xpq_flush_queue(void);
 void xpq_queue_set_ldt(vaddr_t, uint32_t);
 void xpq_queue_tlb_flush(void);
-void xpq_queue_pin_table(paddr_t);
+void xpq_queue_pin_table(paddr_t, unsigned int);
 void xpq_queue_unpin_table(paddr_t);
 int  xpq_update_foreign(paddr_t, pt_entry_t, int);
 static void inline
 xpq_queue_pin_l1_table(paddr_t pa) {
 	xpq_queue_pin_table(pa, MMUEXT_PIN_L1_TABLE);
+}
 static void inline
 xpq_queue_pin_l2_table(paddr_t pa) {
 	xpq_queue_pin_table(pa, MMUEXT_PIN_L2_TABLE);
+}
 #if defined(PAE) || defined(__x86_64__)
 static void inline
 xpq_queue_pin_l3_table(paddr_t pa) {
 	xpq_queue_pin_table(pa, MMUEXT_PIN_L3_TABLE);
+}
 #endif
 #if defined(__x86_64__)
 static void inline
 xpq_queue_pin_l4_table(paddr_t pa) {
 	xpq_queue_pin_table(pa, MMUEXT_PIN_L4_TABLE);
+}
 #endif
 extern unsigned long *xpmap_phys_to_machine_mapping;
 /*
  * On Xen-2, the start of the day virtual memory starts at KERNTEXTOFF
  * (0xc0100000). On Xen-3 for domain0 it starts at KERNBASE (0xc0000000).
  * So the offset between physical and virtual address is different on
  * Xen-2 and Xen-3 for domain0.
  * starting with xen-3.0.2, we can add notes so that virtual memory starts
  * at KERNBASE for domU as well.
  */
 #if defined(XEN3) && (defined(DOM0OPS) || !defined(XEN_COMPAT_030001))
 #define XPMAP_OFFSET	0
 #else

 @@ -1,14 +1,14 @@
-/*	$NetBSD: x86_xpmap.c,v 1.12.4.4 2009/07/23 23:31:37 jym Exp $	*/
+/*	$NetBSD: x86_xpmap.c,v 1.12.4.5 2009/07/24 11:30:28 jym Exp $	*/
 /*
  * Copyright (c) 2006 Mathieu Ropert <mro@adviseo.fr>
+ *
  * Permission to use, copy, modify, and distribute this software for any
  * purpose with or without fee is hereby granted, provided that the above
  * copyright notice and this permission notice appear in all copies.
+ *
  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
 @@ -69,27 +69,27 @@
  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: x86_xpmap.c,v 1.12.4.4 2009/07/23 23:31:37 jym Exp $");
+__KERNEL_RCSID(0, "$NetBSD: x86_xpmap.c,v 1.12.4.5 2009/07/24 11:30:28 jym Exp $");
 #include "opt_xen.h"
 #include "opt_ddb.h"
 #include "ksyms.h"
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/rwlock.h>
 #include <uvm/uvm.h>
 #include <machine/pmap.h>
 #include <machine/gdt.h>
 @@ -288,60 +288,41 @@ xpq_queue_pt_switch(paddr_t pa)
+{
 	struct mmuext_op op;
 	xpq_flush_queue();
 	XENPRINTK2(("xpq_queue_pt_switch: 0x%" PRIx64 " 0x%" PRIx64 "\n",
 	    (int64_t)pa, (int64_t)pa));
 	op.cmd = MMUEXT_NEW_BASEPTR;
 	op.arg1.mfn = pa >> PAGE_SHIFT;
 	if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0)
 		panic("xpq_queue_pt_switch");
+}
 void
-xpq_queue_pin_table(paddr_t pa)
+xpq_queue_pin_table(paddr_t pa, unsigned int level) {
 	struct mmuext_op op;
 	xpq_flush_queue();
 	XENPRINTK2(("xpq_queue_pin_table: 0x%" PRIx64 " 0x%" PRIx64 "\n",
 	    (int64_t)pa, (int64_t)pa));
 	op.arg1.mfn = pa >> PAGE_SHIFT;
 #if defined(__x86_64__)
 	op.cmd = MMUEXT_PIN_L4_TABLE;
 #else
 	op.cmd = MMUEXT_PIN_L2_TABLE;
 #endif
 	if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0)
 		panic("xpq_queue_pin_table");
 #ifdef PAE
 static void
 xpq_queue_pin_l3_table(paddr_t pa)
 	struct mmuext_op op;
 	xpq_flush_queue();
-	XENPRINTK2(("xpq_queue_pin_l2_table: 0x%" PRIx64 " 0x%" PRIx64 "\n",
+	XENPRINTK2(("xpq_queue_pin_table: level %u %#"PRIx64"\n",
-	    (int64_t)pa, (int64_t)pa));
+	    level, (int64_t)pa));
 	op.arg1.mfn = pa >> PAGE_SHIFT;
 	op.cmd = level;
 	op.cmd = MMUEXT_PIN_L3_TABLE;
 	if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0)
-		panic("xpq_queue_pin_table");
+		panic("xpq_queue_pin_table: level %u %#"PRIx64"\n",
 		    level, (int64_t)pa);
+}
 #endif
 void
 xpq_queue_unpin_table(paddr_t pa)
+{
 	struct mmuext_op op;
 	xpq_flush_queue();
 	XENPRINTK2(("xpq_queue_unpin_table: 0x%" PRIx64 " 0x%" PRIx64 "\n",
 	    (int64_t)pa, (int64_t)pa));
 	op.arg1.mfn = pa >> PAGE_SHIFT;
 	op.cmd = MMUEXT_UNPIN_TABLE;
 	if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0)
 		panic("xpq_queue_unpin_table");
 @@ -924,61 +905,63 @@ xen_bootstrap_tables (vaddr_t old_pgd, v
 	pde[PDIR_SLOT_PTE + 3] = xpmap_ptom_masked(addr) | PG_k | PG_V;
 	__PRINTK(("pde[%d] va 0x%lx pa 0x%lx entry 0x%" PRIx64 "\n",
 	    (int)(PDIR_SLOT_PTE + 3), pde + PAGE_SIZE * 4, (long)addr,
 	    (int64_t)pde[PDIR_SLOT_PTE + 3]));
 #endif
 	/* Mark tables RO, and pin the kernel's shadow as L2 */
 	addr = (u_long)pde - KERNBASE;
 	for (i = 0; i < 5; i++, addr += PAGE_SIZE) {
 		xen_bt_set_readonly(((vaddr_t)pde) + PAGE_SIZE * i);
 		if (i == 2 || i == 3)
 			continue;
 #if 0
 		__PRINTK(("pin L2 %d addr 0x%" PRIx64 "\n", i, (int64_t)addr));
-		xpq_queue_pin_table(xpmap_ptom_masked(addr));
+		xpq_queue_pin_l2_table(xpmap_ptom_masked(addr));
 #endif
+	}
 	if (final) {
 		addr = (u_long)pde - KERNBASE + 3 * PAGE_SIZE;
 		__PRINTK(("pin L2 %d addr 0x%" PRIx64 "\n", 2, (int64_t)addr));
-		xpq_queue_pin_table(xpmap_ptom_masked(addr));
+		xpq_queue_pin_l2_table(xpmap_ptom_masked(addr));
+	}
 #if 0
 	addr = (u_long)pde - KERNBASE + 2 * PAGE_SIZE;
 	__PRINTK(("pin L2 %d addr 0x%" PRIx64 "\n", 2, (int64_t)addr));
-	xpq_queue_pin_table(xpmap_ptom_masked(addr));
+	xpq_queue_pin_l2_table(xpmap_ptom_masked(addr));
 #endif
 #else /* PAE */
 	/* recursive entry in higher-level PD */
 	bt_pgd[PDIR_SLOT_PTE] =
 	    xpmap_ptom_masked(new_pgd - KERNBASE) | PG_k | PG_V;
 	__PRINTK(("bt_pgd[PDIR_SLOT_PTE] va 0x%lx pa 0x%" PRIx64
 	    " entry 0x%" PRIx64 "\n", new_pgd, (int64_t)new_pgd - KERNBASE,
 	    (int64_t)bt_pgd[PDIR_SLOT_PTE]));
 	/* Mark tables RO */
 	xen_bt_set_readonly((vaddr_t) pde);
 #endif
 #if PTP_LEVELS > 2 || defined(PAE)
 	xen_bt_set_readonly((vaddr_t) pdtpe);
 #endif
 #if PTP_LEVELS > 3
 	xen_bt_set_readonly(new_pgd);
 #endif
 	/* Pin the PGD */
 	__PRINTK(("pin PGD\n"));
 #ifdef PAE
 	xpq_queue_pin_l3_table(xpmap_ptom_masked(new_pgd - KERNBASE));
 #elif __x86_64__
 	xpq_queue_pin_l4_table(xpmap_ptom_masked(new_pgd - KERNBASE));
 #else
-	xpq_queue_pin_table(xpmap_ptom_masked(new_pgd - KERNBASE));
+	xpq_queue_pin_l2_table(xpmap_ptom_masked(new_pgd - KERNBASE));
 #endif
 #ifdef __i386__
 	/* Save phys. addr of PDP, for libkvm. */
 	PDPpaddr = (long)pde;
 #ifdef PAE
 	/* also save the address of the L3 page */
 	pmap_l3pd = pdtpe;
 	pmap_l3paddr = (new_pgd - KERNBASE);
 #endif /* PAE */
 #endif /* i386 */
 	/* Switch to new tables */
 	__PRINTK(("switch to PGD\n"));
 	xpq_queue_pt_switch(xpmap_ptom_masked(new_pgd - KERNBASE));

 @@ -1,14 +1,14 @@
-/*	$NetBSD: xen_machdep.c,v 1.4.12.3 2009/05/31 20:15:37 jym Exp $	*/
+/*	$NetBSD: xen_machdep.c,v 1.4.12.4 2009/07/24 11:30:28 jym Exp $	*/
 /*
  * Copyright (c) 2006 Manuel Bouyer.
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
  * 1. Redistributions of source code must retain the above copyright
  *    notice, this list of conditions and the following disclaimer.
  * 2. Redistributions in binary form must reproduce the above copyright
  *    notice, this list of conditions and the following disclaimer in the
  *    documentation and/or other materials provided with the distribution.
  * 3. All advertising materials mentioning features or use of this software
 @@ -53,27 +53,27 @@
  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: xen_machdep.c,v 1.4.12.3 2009/05/31 20:15:37 jym Exp $");
+__KERNEL_RCSID(0, "$NetBSD: xen_machdep.c,v 1.4.12.4 2009/07/24 11:30:28 jym Exp $");
 #include "opt_xen.h"
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/boot_flag.h>
 #include <sys/mount.h>
 #include <sys/reboot.h>
 #include <sys/timetc.h>
 #include <sys/sysctl.h>
 #include <sys/pmf.h>
 #include <xen/hypervisor.h>
 @@ -283,26 +283,30 @@ xen_prepare_suspend(void) {
 	kpreempt_disable();
 	xen_suspendclocks();
 	xen_acquire_writer_ptom_lock();
 	/*
 	 * Xen lazy evaluation of recursive mappings requires
 	 * to flush the APDP entries
 	 */
 	pmap_unmap_all_apdp_pdes();
 #ifdef PAE
 	pmap_unmap_shadow_entries();
 #endif
 	/*
 	 * save/restore code does not translate these MFNs to their
 	 * associated PFNs, so we must do it
 	 */
 	xen_start_info.store_mfn = mfn_to_pfn(xen_start_info.store_mfn);
 	xen_start_info.console_mfn = mfn_to_pfn(xen_start_info.console_mfn);
 	DPRINTK(("suspending domain\n"));
 	aprint_verbose("suspending domain\n");
 	/* invalidate the shared_info page */
 	if (HYPERVISOR_update_va_mapping((vaddr_t)HYPERVISOR_shared_info,
 , UVMF_INVLPG)) {
 @@ -316,26 +320,30 @@ xen_prepare_suspend(void) {
  * First operations before restoring domain context
  */
 static void
 xen_prepare_resume(void) {
 	/* map the new shared_info page */
 	if (HYPERVISOR_update_va_mapping((vaddr_t)HYPERVISOR_shared_info,
 	    xen_start_info.shared_info | PG_RW | PG_V,
 	    UVMF_INVLPG)) {
 		DPRINTK(("could not map new shared info page"));
 		HYPERVISOR_crash();
+	}
 #ifdef PAE
 	pmap_map_shadow_entries();
 #endif
 	if (xen_start_info.nr_pages != physmem) {
 		/*
 		 * XXX JYM for now, we crash - fix it with balloon when
 		 * supported
 		 */
 		DPRINTK(("xen_start_info.nr_pages != physmem"));
 		HYPERVISOR_crash();
+	}
 	xen_release_ptom_lock();
 	DPRINTK(("preparing domain resume\n"));
 	aprint_verbose("preparing domain resume\n");

 @@ -1,14 +1,14 @@
-/*	$NetBSD: xencons.c,v 1.31.2.3 2009/07/23 23:31:38 jym Exp $	*/
+/*	$NetBSD: xencons.c,v 1.31.2.4 2009/07/24 11:30:28 jym Exp $	*/
 /*
  * Copyright (c) 2006 Manuel Bouyer.
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
  * 1. Redistributions of source code must retain the above copyright
  *    notice, this list of conditions and the following disclaimer.
  * 2. Redistributions in binary form must reproduce the above copyright
  *    notice, this list of conditions and the following disclaimer in the
  *    documentation and/or other materials provided with the distribution.
  * 3. All advertising materials mentioning features or use of this software
 @@ -53,38 +53,39 @@
  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: xencons.c,v 1.31.2.3 2009/07/23 23:31:38 jym Exp $");
+__KERNEL_RCSID(0, "$NetBSD: xencons.c,v 1.31.2.4 2009/07/24 11:30:28 jym Exp $");
 #include "opt_xen.h"
 #include <sys/param.h>
 #include <sys/ioctl.h>
 #include <sys/proc.h>
 #include <sys/tty.h>
 #include <sys/systm.h>
 #include <sys/device.h>
 #include <sys/conf.h>
 #include <sys/kauth.h>
 #include <sys/kernel.h>
 #include <machine/stdarg.h>
 #include <xen/xen.h>
 #include <xen/hypervisor.h>
 #include <xen/evtchn.h>
 #ifdef XEN3
 #include <uvm/uvm.h>
 #include <machine/pmap.h>
 #include <xen/xen3-public/io/console.h>
 #else
 #include <xen/ctrl_if.h>
 #endif
 @@ -218,65 +219,63 @@ xencons_attach(device_t parent, device_t
 		xencons_resume(self, PMF_F_NONE);
+	}
 	sc->polling = 0;
 	if (!pmf_device_register(self, xencons_suspend, xencons_resume))
 		aprint_error_dev(self, "couldn't establish power handler\n");
+}
 static bool
 xencons_suspend(device_t dev PMF_FN_ARGS) {
 	int evtch;
-	if (xendomain_is_dom0()) {
+	/* dom0 console should not be suspended */
-		evtch = unbind_virq_from_evtch(VIRQ_CONSOLE);
+	if (!xendomain_is_dom0()) {
 		hypervisor_mask_event(evtch);
 		if (event_remove_handler(evtch, xencons_intr,
 		    xencons_console_device) != 0)
 			aprint_error_dev(dev,
 			    "can't remove handler: xencons_intr\n");
 	} else {
 #ifdef XEN3
 		evtch = xen_start_info.console_evtchn;
 		hypervisor_mask_event(evtch);
 		if (event_remove_handler(evtch, xencons_handler,
-		    xencons_console_device) != 0)
+		    xencons_console_device) != 0) {
 			aprint_error_dev(dev,
 			    "can't remove handler: xencons_handler\n");
+		}
 #endif
 		aprint_verbose_dev(dev, "removed event channel %d\n", evtch);
+	}
 	aprint_verbose_dev(dev, "removed event channel %d\n", evtch);
 	return true;
+}
 static bool
 xencons_resume(device_t dev PMF_FN_ARGS) {
 	int evtch = -1;
 	if (xendomain_is_dom0()) {
-		evtch = bind_virq_to_evtch(VIRQ_CONSOLE);
+	/* dom0 console resume is required only during first start-up */
-		if (event_set_handler(evtch, xencons_intr,
+		if (cold) {
-		    xencons_console_device, IPL_TTY, "xencons") != 0)
+			evtch = bind_virq_to_evtch(VIRQ_CONSOLE);
-			aprint_error_dev(dev, "can't register xencons_intr\n");
+			event_set_handler(evtch, xencons_intr,
 			    xencons_console_device, IPL_TTY, "xencons");
+		}
 	} else {
 #ifdef XEN3
 		evtch = xen_start_info.console_evtchn;
-		if (event_set_handler(evtch, xencons_handler,
+		event_set_handler(evtch, xencons_handler,
-		    xencons_console_device, IPL_TTY, "xencons") != 0)
+		    xencons_console_device, IPL_TTY, "xencons");
 			aprint_error_dev(dev, "can't register xencons_handler\n");
 #else
 		(void)ctrl_if_register_receiver(CMSG_CONSOLE,
 		    xencons_rx, 0);
 #endif
+	}
 	if (evtch != -1) {
 		aprint_verbose_dev(dev, "using event channel %d\n", evtch);
 		hypervisor_enable_event(evtch);
+	}
 	return true;
+}

 @@ -1,14 +1,14 @@
-/*	$NetBSD: pmap.h,v 1.21.2.3 2009/07/23 23:31:36 jym Exp $	*/
+/*	$NetBSD: pmap.h,v 1.21.2.4 2009/07/24 11:30:28 jym Exp $	*/
 /*
+ *
  * Copyright (c) 1997 Charles D. Cranor and Washington University.
  * All rights reserved.
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
  * 1. Redistributions of source code must retain the above copyright
  *    notice, this list of conditions and the following disclaimer.
  * 2. Redistributions in binary form must reproduce the above copyright
  *    notice, this list of conditions and the following disclaimer in the
 @@ -344,26 +344,54 @@ kvtopte(vaddr_t va)
 	return (PTE_BASE + pl1_i(va));
+}
 paddr_t vtophys(vaddr_t);
 vaddr_t	pmap_map(vaddr_t, paddr_t, paddr_t, vm_prot_t);
 void	pmap_cpu_init_early(struct cpu_info *);
 void	pmap_cpu_init_late(struct cpu_info *);
 bool	sse2_idlezero_page(void *);
 #ifdef XEN
 #ifdef PAE
 /*
  * Under PAE, Xen handles our recursive mappings to the L2 shadow pages
  * erroneously during a restore (the last entry of the PDIR_SLOT_PTE).
  * This pages are found in two places:
  * - the used ones are found in the pmaps list
  * - the unused ones (but still valid from a Xen's point of view) are cached
  *   inside the pmap_pdp_cache pool.
  * This list is not protected by locks, as it is expected to be accessed only
  * during pmap_create()/pmap_destroy(), and save/restore code, which
  * cannot run concurrently.
 struct l2_pdirpte {
 	SLIST_ENTRY(l2_pinned) l2_pdirpte_list;
 	paddr_t slot_pte;
 	paddr_t slot_pte_content;
 };
  */
 /*
  * Head of the list of all pages pinned as L2 and containing valid entry
  * in PDIR_SLOT_PTE
 SLIST_HEAD(l2_pdirpte_head, l2_pdirpte);
  */
 void	pmap_map_shadow_entries(void);
 void	pmap_unmap_shadow_entries(void);
 #endif /* PAE */
 #define XPTE_MASK	L1_FRAME
 /* XPTE_SHIFT = L1_SHIFT - log2(sizeof(pt_entry_t)) */
 #if defined(__x86_64__) || defined(PAE)
 #define XPTE_SHIFT	9
 #else
 #define XPTE_SHIFT	10
 #endif
 /* PTE access inline fuctions */
 /*
  * Get the machine address of the pointed pte
  * We use hardware MMU to get value so works only for levels 1-3