kvm_getswapinfo.c revision 43173 
1/*
2 * Copyright (c) 1999, Matthew Dillon.  All Rights Reserved.
3 *
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided under the terms of the BSD
6 * Copyright as found in /usr/src/COPYRIGHT in the FreeBSD source tree.
7 */
8
9#ifndef lint
10static const char copyright[] =
11    "@(#) Copyright (c) 1999\n"
12    "Matthew Dillon.  All rights reserved.\n";
13#endif /* not lint */
14
15#ifndef lint
16static const char rcsid[] =
17	"$Id: kvm_getswapinfo.c,v 1.2 1999/01/22 10:57:03 dillon Exp $";
18#endif /* not lint */
19
20#include <sys/param.h>
21#include <sys/time.h>
22#include <sys/vnode.h>
23#include <sys/ucred.h>
24#include <sys/stat.h>
25#include <sys/conf.h>
26#include <sys/rlist.h>
27#include <sys/blist.h>
28
29#include <err.h>
30#include <fcntl.h>
31#include <kvm.h>
32#include <limits.h>
33#include <nlist.h>
34#include <stdio.h>
35#include <stdlib.h>
36#include <string.h>
37#include <unistd.h>
38
39static struct nlist kvm_swap_nl[] = {
40	{ "_swaplist" },	/* old style swap list		*/
41	{ "_swapblist" },	/* new radix swap list		*/
42	{ "_swdevt" },		/* list of swap devices and sizes */
43	{ "_nswdev" },		/* number of swap devices */
44	{ "_dmmax" },		/* maximum size of a swap block */
45	{ "" }
46};
47
48#define NL_SWAPLIST	0
49#define NL_SWAPBLIST	1
50#define NL_SWDEVT	2
51#define NL_NSWDEV	3
52#define NL_DMMAX	4
53
54static int kvm_swap_nl_cached = 0;
55static int nswdev;
56static int unswdev;
57static int dmmax;
58static int type;
59
60static void getswapinfo_old(kvm_t *kd, kvm_swap_t swap_ary, int swap_max, int flags);
61static void getswapinfo_radix(kvm_t *kd, kvm_swap_t swap_ary, int swap_max, int flags);
62
63#define	SVAR(var) __STRING(var)	/* to force expansion */
64#define	KGET(idx, var)							\
65	KGET1(idx, &var, sizeof(var), SVAR(var))
66#define	KGET1(idx, p, s, msg)						\
67	KGET2(kvm_swap_nl[idx].n_value, p, s, msg)
68#define	KGET2(addr, p, s, msg)						\
69	if (kvm_read(kd, (u_long)(addr), p, s) != s)			\
70		warnx("cannot read %s: %s", msg, kvm_geterr(kd))
71#define	KGETN(idx, var)							\
72	KGET1N(idx, &var, sizeof(var), SVAR(var))
73#define	KGET1N(idx, p, s, msg)						\
74	KGET2N(kvm_swap_nl[idx].n_value, p, s, msg)
75#define	KGET2N(addr, p, s, msg)						\
76	((kvm_read(kd, (u_long)(addr), p, s) == s) ? 1 : 0)
77#define	KGETRET(addr, p, s, msg)					\
78	if (kvm_read(kd, (u_long)(addr), p, s) != s) {			\
79		warnx("cannot read %s: %s", msg, kvm_geterr(kd));	\
80		return (0);						\
81	}
82
83int
84kvm_getswapinfo(
85	kvm_t *kd,
86	struct kvm_swap *swap_ary,
87	int swap_max,
88	int flags
89) {
90	int ti = 0;
91
92	/*
93	 * clear cache
94	 */
95	if (kd == NULL) {
96		kvm_swap_nl_cached = 0;
97		return(0);
98	}
99
100	/*
101	 * namelist
102	 */
103	if (kvm_swap_nl_cached == 0) {
104		struct swdevt *sw;
105
106		if (kvm_nlist(kd, kvm_swap_nl) < 0)
107			return(-1);
108
109		/*
110		 * required entries
111		 */
112
113		if (
114		    kvm_swap_nl[NL_SWDEVT].n_value == 0 ||
115		    kvm_swap_nl[NL_NSWDEV].n_value == 0 ||
116		    kvm_swap_nl[NL_DMMAX].n_value == 0
117		) {
118			return(-1);
119		}
120
121		/*
122		 * get globals, type of swap
123		 */
124
125		KGET(NL_NSWDEV, nswdev);
126		KGET(NL_DMMAX, dmmax);
127
128		if (kvm_swap_nl[NL_SWAPLIST].n_value)
129			type = 1;
130
131		if (kvm_swap_nl[NL_SWAPBLIST].n_value)
132			type = 2;
133
134		/*
135		 * figure out how many actual swap devices are enabled
136		 */
137
138		KGET(NL_SWDEVT, sw);
139		for (unswdev = nswdev - 1; unswdev >= 0; --unswdev) {
140			struct swdevt swinfo;
141
142			KGET2(&sw[unswdev], &swinfo, sizeof(swinfo), "swinfo");
143			if (swinfo.sw_nblks)
144				break;
145		}
146		++unswdev;
147
148		kvm_swap_nl_cached = 1;
149	}
150
151
152	{
153		struct swdevt *sw;
154		int i;
155
156		ti = unswdev;
157		if (ti >= swap_max)
158			ti = swap_max - 1;
159
160		if (ti >= 0)
161			bzero(swap_ary, sizeof(struct kvm_swap) * (ti + 1));
162
163		KGET(NL_SWDEVT, sw);
164		for (i = 0; i < unswdev; ++i) {
165			struct swdevt swinfo;
166			int ttl;
167
168			KGET2(&sw[i], &swinfo, sizeof(swinfo), "swinfo");
169
170			/*
171			 * old style: everything in DEV_BSIZE'd chunks,
172			 * convert to pages.
173			 *
174			 * new style: swinfo in DEV_BSIZE'd chunks but dmmax
175			 * in pages.
176			 */
177
178			if (type == 1)
179				ttl = dbtoc(swinfo.sw_nblks);
180			else
181				ttl = swinfo.sw_nblks;
182
183			if (ttl == 0)
184				continue;
185
186			if (i < ti) {
187				swap_ary[i].ksw_total = ttl;
188				swap_ary[i].ksw_used = ttl;
189				swap_ary[i].ksw_flags = swinfo.sw_flags;
190				if (swinfo.sw_dev == NODEV) {
191					snprintf(
192					    swap_ary[i].ksw_devname,
193					    sizeof(swap_ary[i].ksw_devname),
194					    "%s",
195					    "[NFS swap]"
196					);
197				} else {
198					snprintf(
199					    swap_ary[i].ksw_devname,
200					    sizeof(swap_ary[i].ksw_devname),
201					    "%s%s",
202					    ((flags & SWIF_DEV_PREFIX) ? "/dev/" : ""),
203					    devname(swinfo.sw_dev, S_IFBLK)
204					);
205				}
206			}
207			if (ti >= 0) {
208				swap_ary[ti].ksw_total += ttl;
209				swap_ary[ti].ksw_used += ttl;
210			}
211		}
212	}
213
214	switch(type) {
215	case 1:
216		getswapinfo_old(kd, swap_ary, swap_max, flags);
217		break;
218	case 2:
219		getswapinfo_radix(kd, swap_ary, swap_max, flags);
220		break;
221	default:
222		ti = -1;
223		break;
224	}
225	return(ti);
226}
227
228/*
229 * scanradix() - support routine for radix scanner
230 */
231
232#define TABME	tab, tab, ""
233
234static int
235scanradix(
236	blmeta_t *scan,
237	daddr_t blk,
238	daddr_t radix,
239	daddr_t skip,
240	daddr_t count,
241	kvm_t *kd,
242	int dmmax,
243	int nswdev,
244	kvm_swap_t swap_ary,
245	int swap_max,
246	int tab,
247	int flags
248) {
249	blmeta_t meta;
250	int ti = (unswdev >= swap_max) ? swap_max - 1 : unswdev;
251
252	KGET2(scan, &meta, sizeof(meta), "blmeta_t");
253
254	/*
255	 * Terminator
256	 */
257	if (meta.bm_bighint == (daddr_t)-1) {
258		if (flags & SWIF_DUMP_TREE) {
259			printf("%*.*s(0x%06x,%d) Terminator\n",
260			    TABME,
261			    blk,
262			    radix
263			);
264		}
265		return(-1);
266	}
267
268	if (radix == BLIST_BMAP_RADIX) {
269		/*
270		 * Leaf bitmap
271		 */
272		int i;
273
274		if (flags & SWIF_DUMP_TREE) {
275			printf("%*.*s(0x%06x,%d) Bitmap %08x big=%d\n",
276			    TABME,
277			    blk,
278			    radix,
279			    (int)meta.u.bmu_bitmap,
280			    meta.bm_bighint
281			);
282		}
283
284		/*
285		 * If not all allocated, count.
286		 */
287		if (meta.u.bmu_bitmap != 0) {
288			for (i = 0; i < BLIST_BMAP_RADIX && i < count; ++i) {
289				/*
290				 * A 0 bit means allocated
291				 */
292				if ((meta.u.bmu_bitmap & (1 << i))) {
293					int t = 0;
294
295					if (nswdev)
296						t = (blk + i) / dmmax % nswdev;
297					if (t < ti)
298						--swap_ary[t].ksw_used;
299					if (ti >= 0)
300						--swap_ary[ti].ksw_used;
301				}
302			}
303		}
304	} else if (meta.u.bmu_avail == radix) {
305		/*
306		 * Meta node if all free
307		 */
308		if (flags & SWIF_DUMP_TREE) {
309			printf("%*.*s(0x%06x,%d) Submap ALL-FREE {\n",
310			    TABME,
311			    blk,
312			    radix,
313			    (int)meta.u.bmu_avail,
314			    meta.bm_bighint
315			);
316		}
317		/*
318		 * Note: both dmmax and radix are powers of 2.  However, dmmax
319		 * may be larger then radix so use a smaller increment if
320		 * necessary.
321		 */
322		{
323			int t;
324			int tinc = dmmax;
325
326			while (tinc > radix)
327				tinc >>= 1;
328
329			for (t = blk; t < blk + radix; t += tinc) {
330				int u = (nswdev) ? (t / dmmax % nswdev) : 0;
331
332				if (u < ti)
333					swap_ary[u].ksw_used -= tinc;
334				if (ti >= 0)
335					swap_ary[ti].ksw_used -= tinc;
336			}
337		}
338	} else if (meta.u.bmu_avail == 0) {
339		/*
340		 * Meta node if all used
341		 */
342		if (flags & SWIF_DUMP_TREE) {
343			printf("%*.*s(0x%06x,%d) Submap ALL-ALLOCATED\n",
344			    TABME,
345			    blk,
346			    radix,
347			    (int)meta.u.bmu_avail,
348			    meta.bm_bighint
349			);
350		}
351	} else {
352		/*
353		 * Meta node if not all free
354		 */
355		int i;
356		int next_skip;
357
358		if (flags & SWIF_DUMP_TREE) {
359			printf("%*.*s(0x%06x,%d) Submap avail=%d big=%d {\n",
360			    TABME,
361			    blk,
362			    radix,
363			    (int)meta.u.bmu_avail,
364			    meta.bm_bighint
365			);
366		}
367
368		radix >>= BLIST_META_RADIX_SHIFT;
369		next_skip = skip >> BLIST_META_RADIX_SHIFT;
370
371		for (i = 1; i <= skip; i += next_skip) {
372			int r;
373			daddr_t vcount = (count > radix) ? radix : count;
374
375			r = scanradix(
376			    &scan[i],
377			    blk,
378			    radix,
379			    next_skip - 1,
380			    vcount,
381			    kd,
382			    dmmax,
383			    nswdev,
384			    swap_ary,
385			    swap_max,
386			    tab + 4,
387			    flags
388			);
389			if (r < 0)
390				break;
391			blk += radix;
392		}
393		if (flags & SWIF_DUMP_TREE) {
394			printf("%*.*s}\n", TABME);
395		}
396	}
397	return(0);
398}
399
400static void
401getswapinfo_radix(kvm_t *kd, kvm_swap_t swap_ary, int swap_max, int flags)
402{
403	struct blist *swapblist = NULL;
404	struct blist blcopy = { 0 };
405
406	KGET(NL_SWAPBLIST, swapblist);
407	KGET2(swapblist, &blcopy, sizeof(blcopy), "*swapblist");
408
409	if (flags & SWIF_DUMP_TREE) {
410		printf("radix tree: %d/%d/%d blocks, %dK wired\n",
411			blcopy.bl_free,
412			blcopy.bl_blocks,
413			blcopy.bl_radix,
414			(blcopy.bl_rootblks * sizeof(blmeta_t) + 1023)/
415			    1024
416		);
417	}
418	scanradix(
419	    blcopy.bl_root,
420	    0,
421	    blcopy.bl_radix,
422	    blcopy.bl_skip,
423	    blcopy.bl_rootblks,
424	    kd,
425	    dmmax,
426	    nswdev,
427	    swap_ary,
428	    swap_max,
429	    0,
430	    flags
431	);
432}
433
434static void
435getswapinfo_old(kvm_t *kd, kvm_swap_t swap_ary, int swap_max, int flags)
436{
437	struct rlist *swapptr;
438	struct rlisthdr swaplist;
439	int ti = (unswdev >= swap_max) ? swap_max - 1 : unswdev;
440
441	KGET(NL_SWAPLIST, swaplist);
442
443	swapptr = swaplist.rlh_list;
444
445	while (swapptr) {
446		int	top;
447		int	bottom;
448		int	next_block;
449		int	t;
450		int	v;
451		struct rlist head;
452
453		KGET2(swapptr, &head, sizeof(head), "swapptr");
454
455		top = head.rl_end;
456		bottom = head.rl_start;
457
458		/*
459		 * Handle interleave indexing
460		 */
461
462		while (top / dmmax != bottom / dmmax) {
463			next_block = ((bottom + dmmax) / dmmax);
464
465			t = (bottom / dmmax) % nswdev;
466			v = next_block * dmmax - bottom;
467
468			if (t < ti)
469				swap_ary[t].ksw_used -= dbtoc(v);
470			if (ti >= 0)
471				swap_ary[ti].ksw_used -= dbtoc(v);
472
473			bottom = next_block * dmmax;
474		}
475
476		t = (bottom / dmmax) % nswdev;
477		v = top - bottom + 1;
478
479		if (t < ti)
480			swap_ary[t].ksw_used -= dbtoc(v);
481		if (ti >= 0)
482			swap_ary[ti].ksw_used -= dbtoc(v);
483
484		swapptr = head.rl_next;
485	}
486}
487
488