- If swap metadata does not fit into the KVM, reduce the number of

struct swblock entries by dividing the number of the entries by 2
until the swap metadata fits.

- Reject swapon(2) upon failure of swap_zone allocation.

This is just a temporary fix. Better solutions include:
(suggested by:	dillon)

o reserving swap in SWAP_META_PAGES chunks, and
o swapping the swblock structures themselves.

Reviewed by:	alfred, dillon

11 files changed, 76 insertions, 27 deletions

diff --git a/sys/amd64/amd64/machdep.c b/sys/amd64/amd64/machdep.c
index a6f92b1..7b537ff 100644
--- a/sys/amd64/amd64/machdep.c
+++ b/sys/amd64/amd64/machdep.c
@@ -260,6 +260,7 @@ cpu_startup(dummy)
 	vm_size_t size = 0;
 	int firstaddr;
 	vm_offset_t minaddr;
+	int physmem_est;
 
 	if (boothowto & RB_VERBOSE)
 		bootverbose++;
@@ -329,6 +330,16 @@ again:
 	valloc(callwheel, struct callout_tailq, callwheelsize);
 
 	/*
+	 * Discount the physical memory larger than the size of kernel_map
+	 * to avoid eating up all of KVA space.
+	 */
+	if (kernel_map->first_free == NULL) {
+		printf("Warning: no free entries in kernel_map.\n");
+		physmem_est = physmem;
+	} else
+		physmem_est = min(physmem, kernel_map->max_offset - kernel_map->min_offset);
+
+	/*
 	 * The nominal buffer size (and minimum KVA allocation) is BKVASIZE.
 	 * For the first 64MB of ram nominally allocate sufficient buffers to
 	 * cover 1/4 of our ram.  Beyond the first 64MB allocate additional
@@ -340,10 +351,10 @@ again:
 		int factor = 4 * BKVASIZE / PAGE_SIZE;
 
 		nbuf = 50;
-		if (physmem > 1024)
-			nbuf += min((physmem - 1024) / factor, 16384 / factor);
-		if (physmem > 16384)
-			nbuf += (physmem - 16384) * 2 / (factor * 5);
+		if (physmem_est > 1024)
+			nbuf += min((physmem_est - 1024) / factor, 16384 / factor);
+		if (physmem_est > 16384)
+			nbuf += (physmem_est - 16384) * 2 / (factor * 5);
 	}
 
 	/*
diff --git a/sys/i386/i386/machdep.c b/sys/i386/i386/machdep.c
index a6f92b1..7b537ff 100644
--- a/sys/i386/i386/machdep.c
+++ b/sys/i386/i386/machdep.c
@@ -260,6 +260,7 @@ cpu_startup(dummy)
 	vm_size_t size = 0;
 	int firstaddr;
 	vm_offset_t minaddr;
+	int physmem_est;
 
 	if (boothowto & RB_VERBOSE)
 		bootverbose++;
@@ -329,6 +330,16 @@ again:
 	valloc(callwheel, struct callout_tailq, callwheelsize);
 
 	/*
+	 * Discount the physical memory larger than the size of kernel_map
+	 * to avoid eating up all of KVA space.
+	 */
+	if (kernel_map->first_free == NULL) {
+		printf("Warning: no free entries in kernel_map.\n");
+		physmem_est = physmem;
+	} else
+		physmem_est = min(physmem, kernel_map->max_offset - kernel_map->min_offset);
+
+	/*
 	 * The nominal buffer size (and minimum KVA allocation) is BKVASIZE.
 	 * For the first 64MB of ram nominally allocate sufficient buffers to
 	 * cover 1/4 of our ram.  Beyond the first 64MB allocate additional
@@ -340,10 +351,10 @@ again:
 		int factor = 4 * BKVASIZE / PAGE_SIZE;
 
 		nbuf = 50;
-		if (physmem > 1024)
-			nbuf += min((physmem - 1024) / factor, 16384 / factor);
-		if (physmem > 16384)
-			nbuf += (physmem - 16384) * 2 / (factor * 5);
+		if (physmem_est > 1024)
+			nbuf += min((physmem_est - 1024) / factor, 16384 / factor);
+		if (physmem_est > 16384)
+			nbuf += (physmem_est - 16384) * 2 / (factor * 5);
 	}
 
 	/*
diff --git a/sys/vm/default_pager.c b/sys/vm/default_pager.c
index 7ea507f..f5d88a5 100644
--- a/sys/vm/default_pager.c
+++ b/sys/vm/default_pager.c
@@ -46,6 +46,7 @@
 #include <vm/vm_object.h>
 #include <vm/vm_page.h>
 #include <vm/vm_pager.h>
+#include <vm/vm_zone.h>
 #include <vm/swap_pager.h>
 
 static vm_object_t default_pager_alloc __P((void *, vm_ooffset_t, vm_prot_t,
diff --git a/sys/vm/swap_pager.c b/sys/vm/swap_pager.c
index 0b88efd..9a0ca56 100644
--- a/sys/vm/swap_pager.c
+++ b/sys/vm/swap_pager.c
@@ -90,13 +90,16 @@
 
 #include "opt_swap.h"
 #include <vm/vm.h>
+#include <vm/pmap.h>
+#include <vm/vm_map.h>
+#include <vm/vm_kern.h>
 #include <vm/vm_object.h>
 #include <vm/vm_page.h>
 #include <vm/vm_pager.h>
 #include <vm/vm_pageout.h>
+#include <vm/vm_zone.h>
 #include <vm/swap_pager.h>
 #include <vm/vm_extern.h>
-#include <vm/vm_zone.h>
 
 #define SWM_FREE	0x02	/* free, period			*/
 #define SWM_POP		0x04	/* pop out			*/
@@ -273,7 +276,7 @@ swap_pager_init()
 void
 swap_pager_swap_init()
 {
-	int n;
+	int n, n2;
 
 	/*
 	 * Number of in-transit swap bp operations.  Don't
@@ -311,15 +314,23 @@ swap_pager_swap_init()
 	 * can hold 16 pages, so this is probably overkill.
 	 */
 
-	n = cnt.v_page_count * 2;
-
-	swap_zone = zinit(
-	    "SWAPMETA", 
-	    sizeof(struct swblock), 
-	    n,
-	    ZONE_INTERRUPT, 
-	    1
-	);
+	n = min(cnt.v_page_count, (kernel_map->max_offset - kernel_map->min_offset) / PAGE_SIZE) * 2;
+	n2 = n;
+
+	while (n > 0
+	       && (swap_zone = zinit(
+		       "SWAPMETA", 
+		       sizeof(struct swblock), 
+		       n,
+		       ZONE_INTERRUPT, 
+		       1
+		       )) == NULL)
+		n >>= 1;
+	if (swap_zone == NULL)
+		printf("WARNING: failed to init swap_zone!\n");
+	if (n2 != n)
+		printf("Swap zone entries reduced to %d.\n", n);
+	n2 = n;
 
 	/*
 	 * Initialize our meta-data hash table.  The swapper does not need to
@@ -330,7 +341,7 @@ swap_pager_swap_init()
 	 *	swhash_mask:	hash table index mask
 	 */
 
-	for (n = 1; n < cnt.v_page_count / 4; n <<= 1)
+	for (n = 1; n < n2 ; n <<= 1)
 		;
 
 	swhash = malloc(sizeof(struct swblock *) * n, M_VMPGDATA, M_WAITOK | M_ZERO);
diff --git a/sys/vm/swap_pager.h b/sys/vm/swap_pager.h
index ea57132..b0fbe3a 100644
--- a/sys/vm/swap_pager.h
+++ b/sys/vm/swap_pager.h
@@ -84,6 +84,7 @@ struct swblock {
 extern struct pagerlst swap_pager_un_object_list;
 extern int swap_pager_full;
 extern struct blist *swapblist;
+extern vm_zone_t swap_zone;
 
 void swap_pager_putpages __P((vm_object_t, vm_page_t *, int, boolean_t, int *));
 boolean_t swap_pager_haspage __P((vm_object_t object, vm_pindex_t pindex, int *before, int *after));
diff --git a/sys/vm/vm_kern.c b/sys/vm/vm_kern.c
index ee9e7e4..14b9fbe 100644
--- a/sys/vm/vm_kern.c
+++ b/sys/vm/vm_kern.c
@@ -489,4 +489,3 @@ kmem_init(start, end)
 	/* ... and ending with the completion of the above `insert' */
 	vm_map_unlock(m);
 }
-
diff --git a/sys/vm/vm_map.c b/sys/vm/vm_map.c
index 8aa25e2..8cc61b1 100644
--- a/sys/vm/vm_map.c
+++ b/sys/vm/vm_map.c
@@ -86,8 +86,8 @@
 #include <vm/vm_pager.h>
 #include <vm/vm_kern.h>
 #include <vm/vm_extern.h>
-#include <vm/swap_pager.h>
 #include <vm/vm_zone.h>
+#include <vm/swap_pager.h>
 
 /*
  *	Virtual memory maps provide for the mapping, protection,
diff --git a/sys/vm/vm_object.c b/sys/vm/vm_object.c
index ddb9ab7..3eb5243 100644
--- a/sys/vm/vm_object.c
+++ b/sys/vm/vm_object.c
@@ -84,10 +84,10 @@
 #include <vm/vm_page.h>
 #include <vm/vm_pageout.h>
 #include <vm/vm_pager.h>
+#include <vm/vm_zone.h>
 #include <vm/swap_pager.h>
 #include <vm/vm_kern.h>
 #include <vm/vm_extern.h>
-#include <vm/vm_zone.h>
 
 static void	vm_object_qcollapse __P((vm_object_t object));
 
diff --git a/sys/vm/vm_pageout.c b/sys/vm/vm_pageout.c
index 6a52166..dbea3d6 100644
--- a/sys/vm/vm_pageout.c
+++ b/sys/vm/vm_pageout.c
@@ -93,6 +93,7 @@
 #include <vm/vm_map.h>
 #include <vm/vm_pageout.h>
 #include <vm/vm_pager.h>
+#include <vm/vm_zone.h>
 #include <vm/swap_pager.h>
 #include <vm/vm_extern.h>
 
diff --git a/sys/vm/vm_swap.c b/sys/vm/vm_swap.c
index 8691c22..fada384 100644
--- a/sys/vm/vm_swap.c
+++ b/sys/vm/vm_swap.c
@@ -53,6 +53,7 @@
 #include <sys/stat.h>
 #include <vm/vm.h>
 #include <vm/vm_extern.h>
+#include <vm/vm_zone.h>
 #include <vm/swap_pager.h>
 
 /*
@@ -194,6 +195,13 @@ swapon(p, uap)
 	if (error)
 		return (error);
 
+	/*
+	 * Swap metadata may not fit in the KVM if we have physical
+	 * memory of >1GB.
+	 */
+	if (swap_zone == NULL)
+		return (ENOMEM);
+
 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->name, p);
 	error = namei(&nd);
 	if (error)
diff --git a/sys/vm/vm_zone.c b/sys/vm/vm_zone.c
index c592c91..b5ade30 100644
--- a/sys/vm/vm_zone.c
+++ b/sys/vm/vm_zone.c
@@ -80,8 +80,11 @@ int
 zinitna(vm_zone_t z, vm_object_t obj, char *name, int size,
 	int nentries, int flags, int zalloc)
 {
-	int totsize;
+	int totsize, oldzflags;
+	vm_zone_t oldzlist;
 
+	oldzflags = z->zflags;
+	oldzlist = zlist;
 	if ((z->zflags & ZONE_BOOT) == 0) {
 		z->zsize = (size + ZONE_ROUNDING - 1) & ~(ZONE_ROUNDING - 1);
 		simple_lock_init(&z->zlock);
@@ -112,8 +115,12 @@ zinitna(vm_zone_t z, vm_object_t obj, char *name, int size,
 		zone_kmem_kvaspace += totsize;
 
 		z->zkva = kmem_alloc_pageable(kernel_map, totsize);
-		if (z->zkva == 0)
+		if (z->zkva == 0) {
+			/* Clean up the zlist in case we messed it. */
+			if ((oldzflags & ZONE_BOOT) == 0)
+				zlist = oldzlist;
 			return 0;
+		}
 
 		z->zpagemax = totsize / PAGE_SIZE;
 		if (obj == NULL) {
@@ -156,11 +163,10 @@ zinit(char *name, int size, int nentries, int flags, int zalloc)
 {
 	vm_zone_t z;
 
-	z = (vm_zone_t) malloc(sizeof (struct vm_zone), M_ZONE, M_NOWAIT);
+	z = (vm_zone_t) malloc(sizeof (struct vm_zone), M_ZONE, M_NOWAIT | M_ZERO);
 	if (z == NULL)
 		return NULL;
 
-	z->zflags = 0;
 	if (zinitna(z, NULL, name, size, nentries, flags, zalloc) == 0) {
 		free(z, M_ZONE);
 		return NULL;