Introduce per-swap area accounting in the VM system, and export

this information via the vm.nswapdev sysctl (number of swap areas)
and vm.swapdevX nodes (where X is the device), which contain the MIBs
dev, blocks, used, and flags.  These changes are required to allow
top and other userland swap-monitoring utilities to run without
setgid kmem.

Submitted by:	Thomas Moestl <tmoestl@gmx.net>
Reviewed by:	freebsd-audit

4 files changed, 44 insertions, 3 deletions

diff --git a/sys/sys/conf.h b/sys/sys/conf.h
index 81ea6af..81b94bc 100644
--- a/sys/sys/conf.h
+++ b/sys/sys/conf.h
@@ -232,6 +232,7 @@ struct swdevt {
 	udev_t	sw_dev;			/* For quasibogus swapdev reporting */
 	int	sw_flags;
 	int	sw_nblks;
+	int     sw_used;
 	struct	vnode *sw_vp;
 	dev_t	sw_device;
 };
diff --git a/sys/sys/linedisc.h b/sys/sys/linedisc.h
index 81ea6af..81b94bc 100644
--- a/sys/sys/linedisc.h
+++ b/sys/sys/linedisc.h
@@ -232,6 +232,7 @@ struct swdevt {
 	udev_t	sw_dev;			/* For quasibogus swapdev reporting */
 	int	sw_flags;
 	int	sw_nblks;
+	int     sw_used;
 	struct	vnode *sw_vp;
 	dev_t	sw_device;
 };
diff --git a/sys/vm/swap_pager.c b/sys/vm/swap_pager.c
index 9a0ca56..d62ada2 100644
--- a/sys/vm/swap_pager.c
+++ b/sys/vm/swap_pager.c
@@ -125,11 +125,16 @@ static struct swblock **swhash;
 static int swhash_mask;
 static int swap_async_max = 4;	/* maximum in-progress async I/O's	*/
 
-extern struct vnode *swapdev_vp;	/* from vm_swap.c */
+/* from vm_swap.c */
+extern struct vnode *swapdev_vp;
+extern struct swdevt *swdevt;
+extern int nswdev;
 
 SYSCTL_INT(_vm, OID_AUTO, swap_async_max,
         CTLFLAG_RW, &swap_async_max, 0, "Maximum running async swap ops");
 
+#define BLK2DEVIDX(blk) (nswdev > 1 ? blk / dmmax % nswdev : 0)
+
 /*
  * "named" and "unnamed" anon region objects.  Try to reduce the overhead
  * of searching a named list by hashing it just a little.
@@ -490,6 +495,8 @@ swp_pager_getswapspace(npages)
 		}
 	} else {
 		vm_swap_size -= npages;
+		/* per-swap area stats */
+		swdevt[BLK2DEVIDX(blk)].sw_used += npages;
 		swp_sizecheck();
 	}
 	return(blk);
@@ -517,6 +524,8 @@ swp_pager_freeswapspace(blk, npages)
 {
 	blist_free(swapblist, blk, npages);
 	vm_swap_size += npages;
+	/* per-swap area stats */
+	swdevt[BLK2DEVIDX(blk)].sw_used -= npages;
 	swp_sizecheck();
 }
 
diff --git a/sys/vm/vm_swap.c b/sys/vm/vm_swap.c
index fada384..5904395 100644
--- a/sys/vm/vm_swap.c
+++ b/sys/vm/vm_swap.c
@@ -51,6 +51,7 @@
 #include <sys/lock.h>
 #include <sys/conf.h>
 #include <sys/stat.h>
+#include <sys/sysctl.h>
 #include <vm/vm.h>
 #include <vm/vm_extern.h>
 #include <vm/vm_zone.h>
@@ -64,10 +65,13 @@
 #define NSWAPDEV	4
 #endif
 static struct swdevt should_be_malloced[NSWAPDEV];
-static struct swdevt *swdevt = should_be_malloced;
+struct swdevt *swdevt = should_be_malloced;
 static int nswap;		/* first block after the interleaved devs */
-static int nswdev = NSWAPDEV;
+int nswdev = NSWAPDEV;
 int vm_swap_size;
+static int ncswdev = 0;         /* # of configured swap devs */
+SYSCTL_INT(_vm, OID_AUTO, nswapdev, CTLFLAG_RD, 
+   &ncswdev, 0, "Number of swap devices");
 
 static int swapdev_strategy __P((struct vop_strategy_args *ap));
 struct vnode *swapdev_vp;
@@ -246,6 +250,8 @@ swaponvp(p, vp, dev, nblks)
 	swblk_t dvbase;
 	int error;
 	u_long aligned_nblks;
+	struct sysctl_oid *oid;
+	char name[11];
 
 	if (!swapdev_vp) {
 		error = getnewvnode(VT_NON, NULL, swapdev_vnodeop_p,
@@ -305,6 +311,7 @@ swaponvp(p, vp, dev, nblks)
 	sp->sw_device = dev;
 	sp->sw_flags |= SW_FREED;
 	sp->sw_nblks = nblks;
+	sp->sw_used = 0;
 
 	/*
 	 * nblks, nswap, and dmmax are PAGE_SIZE'd parameters now, not
@@ -327,6 +334,29 @@ swaponvp(p, vp, dev, nblks)
 		blist_free(swapblist, vsbase, blk);
 		vm_swap_size += blk;
 	}
+	/* 
+	 * Add sysctl entries for new swap area. XXX: if some day swap devices 
+	 * can be removed, add an oid member to swdevt.
+	 */
+	if (snprintf(name, sizeof(name), "swapdev%d", ncswdev) < sizeof(name)) {
+		oid = SYSCTL_ADD_NODE(NULL, SYSCTL_STATIC_CHILDREN(_vm), 
+		    OID_AUTO, name, CTLFLAG_RD, NULL, "Swap device stats");
+		if (oid != NULL) {
+			SYSCTL_ADD_INT(NULL, SYSCTL_CHILDREN(oid), OID_AUTO, 
+			    "flags", CTLFLAG_RD, &sp->sw_flags, 0, "Flags");
+			SYSCTL_ADD_INT(NULL, SYSCTL_CHILDREN(oid), OID_AUTO, 
+			    "nblks", CTLFLAG_RD, &sp->sw_nblks, 0, 
+			    "Number of blocks");
+			SYSCTL_ADD_INT(NULL, SYSCTL_CHILDREN(oid), OID_AUTO, 
+			    "used", CTLFLAG_RD, &sp->sw_used, 0, 
+			    "Number of blocks in use");
+			SYSCTL_ADD_OPAQUE(NULL, SYSCTL_CHILDREN(oid), OID_AUTO, 
+			    "dev", CTLFLAG_RD, &sp->sw_dev, sizeof(sp->sw_dev), 
+			    "T,dev_t", "Device");
+		}
+	} else
+		printf("XXX: swaponvp() name buffer too small!\n");
+	ncswdev++;
 
 	return (0);
 }