machine.c revision 28819
1/*
2 * top - a top users display for Unix
3 *
4 * SYNOPSIS:  For FreeBSD-2.x system
5 *
6 * DESCRIPTION:
7 * Originally written for BSD4.4 system by Christos Zoulas.
8 * Ported to FreeBSD 2.x by Steven Wallace && Wolfram Schneider
9 *
10 * This is the machine-dependent module for FreeBSD 2.2
11 * Works for:
12 *	FreeBSD 2.2, and probably FreeBSD 2.1.x
13 *
14 * LIBS: -lkvm
15 *
16 * AUTHOR:  Christos Zoulas <christos@ee.cornell.edu>
17 *          Steven Wallace  <swallace@freebsd.org>
18 *          Wolfram Schneider <wosch@FreeBSD.org>
19 *
20 * $Id: machine.c,v 1.5 1997/07/14 09:06:46 peter Exp $
21 */
22
23
24#include <sys/types.h>
25#include <sys/signal.h>
26#include <sys/param.h>
27
28#include "os.h"
29#include <stdio.h>
30#include <nlist.h>
31#include <math.h>
32#include <kvm.h>
33#include <pwd.h>
34#include <sys/errno.h>
35#include <sys/sysctl.h>
36#include <sys/dkstat.h>
37#include <sys/file.h>
38#include <sys/time.h>
39#include <sys/proc.h>
40#include <sys/user.h>
41#include <sys/vmmeter.h>
42
43/* Swap */
44#include <stdlib.h>
45#include <sys/rlist.h>
46#include <sys/conf.h>
47
48#include <osreldate.h> /* for changes in kernel structures */
49
50#include "top.h"
51#include "machine.h"
52
53static int check_nlist __P((struct nlist *));
54static int getkval __P((unsigned long, int *, int, char *));
55extern char* printable __P((char *));
56int swapmode __P((int *retavail, int *retfree));
57static int smpmode;
58static int namelength;
59static int cmdlength;
60
61
62/* get_process_info passes back a handle.  This is what it looks like: */
63
64struct handle
65{
66    struct kinfo_proc **next_proc;	/* points to next valid proc pointer */
67    int remaining;		/* number of pointers remaining */
68};
69
70/* declarations for load_avg */
71#include "loadavg.h"
72
73#define PP(pp, field) ((pp)->kp_proc . field)
74#define EP(pp, field) ((pp)->kp_eproc . field)
75#define VP(pp, field) ((pp)->kp_eproc.e_vm . field)
76
77/* define what weighted cpu is.  */
78#define weighted_cpu(pct, pp) (PP((pp), p_swtime) == 0 ? 0.0 : \
79			 ((pct) / (1.0 - exp(PP((pp), p_swtime) * logcpu))))
80
81/* what we consider to be process size: */
82#define PROCSIZE(pp) (VP((pp), vm_tsize) + VP((pp), vm_dsize) + VP((pp), vm_ssize))
83
84/* definitions for indices in the nlist array */
85
86
87static struct nlist nlst[] = {
88#define X_CCPU		0
89    { "_ccpu" },		/* 0 */
90#define X_CP_TIME	1
91    { "_cp_time" },		/* 1 */
92#define X_HZ		2
93    { "_hz" },		        /* 2 */
94#define X_STATHZ	3
95    { "_stathz" },		/* 3 */
96#define X_AVENRUN	4
97    { "_averunnable" },		/* 4 */
98
99/* Swap */
100#define VM_SWAPLIST	5
101	{ "_swaplist" },/* list of free swap areas */
102#define VM_SWDEVT	6
103	{ "_swdevt" },	/* list of swap devices and sizes */
104#define VM_NSWAP	7
105	{ "_nswap" },	/* size of largest swap device */
106#define VM_NSWDEV	8
107	{ "_nswdev" },	/* number of swap devices */
108#define VM_DMMAX	9
109	{ "_dmmax" },	/* maximum size of a swap block */
110#define X_BUFSPACE	10
111	{ "_bufspace" },	/* K in buffer cache */
112#define X_CNT           11
113    { "_cnt" },		        /* struct vmmeter cnt */
114
115/* Last pid */
116#define X_LASTPID	12
117    { "_nextpid" },
118    { 0 }
119};
120
121/*
122 *  These definitions control the format of the per-process area
123 */
124
125static char smp_header[] =
126  "  PID %-*.*s PRI NICE  SIZE    RES STATE C   TIME   WCPU    CPU COMMAND";
127
128#define smp_Proc_format \
129	"%5d %-*.*s %3d %3d%7s %6s %-6.6s%1x%7s %5.2f%% %5.2f%% %.*s"
130
131static char up_header[] =
132  "  PID %-*.*s PRI NICE  SIZE    RES STATE    TIME   WCPU    CPU COMMAND";
133
134#define up_Proc_format \
135	"%5d %-*.*s %3d %3d%7s %6s %-6.6s%.0d%7s %5.2f%% %5.2f%% %.*s"
136
137
138
139/* process state names for the "STATE" column of the display */
140/* the extra nulls in the string "run" are for adding a slash and
141   the processor number when needed */
142
143char *state_abbrev[] =
144{
145    "", "START", "RUN\0\0\0", "SLEEP", "STOP", "ZOMB",
146};
147
148
149static kvm_t *kd;
150
151/* values that we stash away in _init and use in later routines */
152
153static double logcpu;
154
155/* these are retrieved from the kernel in _init */
156
157static          long hz;
158static load_avg  ccpu;
159
160/* these are offsets obtained via nlist and used in the get_ functions */
161
162static unsigned long cp_time_offset;
163static unsigned long avenrun_offset;
164static unsigned long lastpid_offset;
165static long lastpid;
166static unsigned long cnt_offset;
167static unsigned long bufspace_offset;
168static long cnt;
169
170/* these are for calculating cpu state percentages */
171
172static long cp_time[CPUSTATES];
173static long cp_old[CPUSTATES];
174static long cp_diff[CPUSTATES];
175
176/* these are for detailing the process states */
177
178int process_states[6];
179char *procstatenames[] = {
180    "", " starting, ", " running, ", " sleeping, ", " stopped, ",
181    " zombie, ",
182    NULL
183};
184
185/* these are for detailing the cpu states */
186
187int cpu_states[CPUSTATES];
188char *cpustatenames[] = {
189    "user", "nice", "system", "interrupt", "idle", NULL
190};
191
192/* these are for detailing the memory statistics */
193
194int memory_stats[7];
195char *memorynames[] = {
196    "K Active, ", "K Inact, ", "K Wired, ", "K Cache, ", "K Buf, ", "K Free",
197    NULL
198};
199
200int swap_stats[7];
201char *swapnames[] = {
202/*   0           1            2           3            4       5 */
203    "K Total, ", "K Used, ", "K Free, ", "% Inuse, ", "K In, ", "K Out",
204    NULL
205};
206
207
208/* these are for keeping track of the proc array */
209
210static int nproc;
211static int onproc = -1;
212static int pref_len;
213static struct kinfo_proc *pbase;
214static struct kinfo_proc **pref;
215
216/* these are for getting the memory statistics */
217
218static int pageshift;		/* log base 2 of the pagesize */
219
220/* define pagetok in terms of pageshift */
221
222#define pagetok(size) ((size) << pageshift)
223
224/* useful externals */
225long percentages();
226
227int
228machine_init(statics)
229
230struct statics *statics;
231
232{
233    register int i = 0;
234    register int pagesize;
235    int modelen;
236    struct passwd *pw;
237
238    modelen = sizeof(smpmode);
239    if ((sysctlbyname("machdep.smp_active", &smpmode, &modelen, NULL, 0) < 0 &&
240         sysctlbyname("smp.smp_active", &smpmode, &modelen, NULL, 0) < 0) ||
241	modelen != sizeof(smpmode))
242	    smpmode = 0;
243
244    while ((pw = getpwent()) != NULL) {
245	if (strlen(pw->pw_name) > namelength)
246	    namelength = strlen(pw->pw_name);
247    }
248    if (namelength < 8)
249	namelength = 8;
250    if (namelength > 16)
251	namelength = 16;
252
253    if ((kd = kvm_open(NULL, NULL, NULL, O_RDONLY, "kvm_open")) == NULL)
254	return -1;
255
256
257    /* get the list of symbols we want to access in the kernel */
258    (void) kvm_nlist(kd, nlst);
259    if (nlst[0].n_type == 0)
260    {
261	fprintf(stderr, "top: nlist failed\n");
262	return(-1);
263    }
264
265    /* make sure they were all found */
266    if (i > 0 && check_nlist(nlst) > 0)
267    {
268	return(-1);
269    }
270
271    /* get the symbol values out of kmem */
272    (void) getkval(nlst[X_STATHZ].n_value, (int *)(&hz), sizeof(hz), "!");
273    if (!hz) {
274	(void) getkval(nlst[X_HZ].n_value, (int *)(&hz), sizeof(hz),
275		       nlst[X_HZ].n_name);
276    }
277
278    (void) getkval(nlst[X_CCPU].n_value,   (int *)(&ccpu),	sizeof(ccpu),
279	    nlst[X_CCPU].n_name);
280
281    /* stash away certain offsets for later use */
282    cp_time_offset = nlst[X_CP_TIME].n_value;
283    avenrun_offset = nlst[X_AVENRUN].n_value;
284    lastpid_offset =  nlst[X_LASTPID].n_value;
285    cnt_offset = nlst[X_CNT].n_value;
286    bufspace_offset = nlst[X_BUFSPACE].n_value;
287
288    /* this is used in calculating WCPU -- calculate it ahead of time */
289    logcpu = log(loaddouble(ccpu));
290
291    pbase = NULL;
292    pref = NULL;
293    nproc = 0;
294    onproc = -1;
295    /* get the page size with "getpagesize" and calculate pageshift from it */
296    pagesize = getpagesize();
297    pageshift = 0;
298    while (pagesize > 1)
299    {
300	pageshift++;
301	pagesize >>= 1;
302    }
303
304    /* we only need the amount of log(2)1024 for our conversion */
305    pageshift -= LOG1024;
306
307    /* fill in the statics information */
308    statics->procstate_names = procstatenames;
309    statics->cpustate_names = cpustatenames;
310    statics->memory_names = memorynames;
311    statics->swap_names = swapnames;
312
313    /* all done! */
314    return(0);
315}
316
317char *format_header(uname_field)
318
319register char *uname_field;
320
321{
322    register char *ptr;
323    static char Header[128];
324
325    snprintf(Header, sizeof(Header), smpmode ? smp_header : up_header,
326	     namelength, namelength, uname_field);
327
328    cmdlength = 80 - strlen(Header) + 6;
329
330    return Header;
331}
332
333static int swappgsin = -1;
334static int swappgsout = -1;
335extern struct timeval timeout;
336
337void
338get_system_info(si)
339
340struct system_info *si;
341
342{
343    long total;
344    load_avg avenrun[3];
345
346    /* get the cp_time array */
347    (void) getkval(cp_time_offset, (int *)cp_time, sizeof(cp_time),
348		   nlst[X_CP_TIME].n_name);
349    (void) getkval(avenrun_offset, (int *)avenrun, sizeof(avenrun),
350		   nlst[X_AVENRUN].n_name);
351
352    (void) getkval(lastpid_offset, (int *)(&lastpid), sizeof(lastpid),
353		   "!");
354
355    /* convert load averages to doubles */
356    {
357	register int i;
358	register double *infoloadp;
359	load_avg *avenrunp;
360
361#ifdef notyet
362	struct loadavg sysload;
363	int size;
364	getkerninfo(KINFO_LOADAVG, &sysload, &size, 0);
365#endif
366
367	infoloadp = si->load_avg;
368	avenrunp = avenrun;
369	for (i = 0; i < 3; i++)
370	{
371#ifdef notyet
372	    *infoloadp++ = ((double) sysload.ldavg[i]) / sysload.fscale;
373#endif
374	    *infoloadp++ = loaddouble(*avenrunp++);
375	}
376    }
377
378    /* convert cp_time counts to percentages */
379    total = percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff);
380
381    /* sum memory & swap statistics */
382    {
383	struct vmmeter sum;
384	static unsigned int swap_delay = 0;
385	static int swapavail = 0;
386	static int swapfree = 0;
387	static int bufspace = 0;
388
389        (void) getkval(cnt_offset, (int *)(&sum), sizeof(sum),
390		   "_cnt");
391        (void) getkval(bufspace_offset, (int *)(&bufspace), sizeof(bufspace),
392		   "_bufspace");
393
394	/* convert memory stats to Kbytes */
395	memory_stats[0] = pagetok(sum.v_active_count);
396	memory_stats[1] = pagetok(sum.v_inactive_count);
397	memory_stats[2] = pagetok(sum.v_wire_count);
398	memory_stats[3] = pagetok(sum.v_cache_count);
399	memory_stats[4] = bufspace / 1024;
400	memory_stats[5] = pagetok(sum.v_free_count);
401	memory_stats[6] = -1;
402
403	/* first interval */
404        if (swappgsin < 0) {
405	    swap_stats[4] = 0;
406	    swap_stats[5] = 0;
407	}
408
409	/* compute differences between old and new swap statistic */
410	else {
411	    swap_stats[4] = pagetok(((sum.v_swappgsin - swappgsin)));
412	    swap_stats[5] = pagetok(((sum.v_swappgsout - swappgsout)));
413	}
414
415        swappgsin = sum.v_swappgsin;
416	swappgsout = sum.v_swappgsout;
417
418	/* call CPU heavy swapmode() only for changes */
419        if (swap_stats[4] > 0 || swap_stats[5] > 0 || swap_delay == 0) {
420	    swap_stats[3] = swapmode(&swapavail, &swapfree);
421	    swap_stats[0] = swapavail;
422	    swap_stats[1] = swapavail - swapfree;
423	    swap_stats[2] = swapfree;
424	}
425        swap_delay = 1;
426	swap_stats[6] = -1;
427    }
428
429    /* set arrays and strings */
430    si->cpustates = cpu_states;
431    si->memory = memory_stats;
432    si->swap = swap_stats;
433
434
435    if(lastpid > 0) {
436	si->last_pid = lastpid;
437    } else {
438	si->last_pid = -1;
439    }
440}
441
442static struct handle handle;
443
444caddr_t get_process_info(si, sel, compare)
445
446struct system_info *si;
447struct process_select *sel;
448int (*compare)();
449
450{
451    register int i;
452    register int total_procs;
453    register int active_procs;
454    register struct kinfo_proc **prefp;
455    register struct kinfo_proc *pp;
456
457    /* these are copied out of sel for speed */
458    int show_idle;
459    int show_system;
460    int show_uid;
461    int show_command;
462
463
464    pbase = kvm_getprocs(kd, KERN_PROC_ALL, 0, &nproc);
465    if (nproc > onproc)
466	pref = (struct kinfo_proc **) realloc(pref, sizeof(struct kinfo_proc *)
467		* (onproc = nproc));
468    if (pref == NULL || pbase == NULL) {
469	(void) fprintf(stderr, "top: Out of memory.\n");
470	quit(23);
471    }
472    /* get a pointer to the states summary array */
473    si->procstates = process_states;
474
475    /* set up flags which define what we are going to select */
476    show_idle = sel->idle;
477    show_system = sel->system;
478    show_uid = sel->uid != -1;
479    show_command = sel->command != NULL;
480
481    /* count up process states and get pointers to interesting procs */
482    total_procs = 0;
483    active_procs = 0;
484    memset((char *)process_states, 0, sizeof(process_states));
485    prefp = pref;
486    for (pp = pbase, i = 0; i < nproc; pp++, i++)
487    {
488	/*
489	 *  Place pointers to each valid proc structure in pref[].
490	 *  Process slots that are actually in use have a non-zero
491	 *  status field.  Processes with P_SYSTEM set are system
492	 *  processes---these get ignored unless show_sysprocs is set.
493	 */
494	if (PP(pp, p_stat) != 0 &&
495	    (show_system || ((PP(pp, p_flag) & P_SYSTEM) == 0)))
496	{
497	    total_procs++;
498	    process_states[(unsigned char) PP(pp, p_stat)]++;
499	    if ((PP(pp, p_stat) != SZOMB) &&
500		(show_idle || (PP(pp, p_pctcpu) != 0) ||
501		 (PP(pp, p_stat) == SRUN)) &&
502		(!show_uid || EP(pp, e_pcred.p_ruid) == (uid_t)sel->uid))
503	    {
504		*prefp++ = pp;
505		active_procs++;
506	    }
507	}
508    }
509
510    /* if requested, sort the "interesting" processes */
511    if (compare != NULL)
512    {
513	qsort((char *)pref, active_procs, sizeof(struct kinfo_proc *), compare);
514    }
515
516    /* remember active and total counts */
517    si->p_total = total_procs;
518    si->p_active = pref_len = active_procs;
519
520    /* pass back a handle */
521    handle.next_proc = pref;
522    handle.remaining = active_procs;
523    return((caddr_t)&handle);
524}
525
526char fmt[128];		/* static area where result is built */
527
528char *format_next_process(handle, get_userid)
529
530caddr_t handle;
531char *(*get_userid)();
532
533{
534    register struct kinfo_proc *pp;
535    register long cputime;
536    register double pct;
537    struct handle *hp;
538    char status[16];
539
540    /* find and remember the next proc structure */
541    hp = (struct handle *)handle;
542    pp = *(hp->next_proc++);
543    hp->remaining--;
544
545
546    /* get the process's user struct and set cputime */
547    if ((PP(pp, p_flag) & P_INMEM) == 0) {
548	/*
549	 * Print swapped processes as <pname>
550	 */
551	char *comm = PP(pp, p_comm);
552#define COMSIZ sizeof(PP(pp, p_comm))
553	char buf[COMSIZ];
554	(void) strncpy(buf, comm, COMSIZ);
555	comm[0] = '<';
556	(void) strncpy(&comm[1], buf, COMSIZ - 2);
557	comm[COMSIZ - 2] = '\0';
558	(void) strncat(comm, ">", COMSIZ - 1);
559	comm[COMSIZ - 1] = '\0';
560    }
561
562#if 0
563    /* This does not produce the correct results */
564    cputime = PP(pp, p_uticks) + PP(pp, p_sticks) + PP(pp, p_iticks);
565#endif
566    cputime = PP(pp, p_rtime).tv_sec;	/* This does not count interrupts */
567
568    /* calculate the base for cpu percentages */
569    pct = pctdouble(PP(pp, p_pctcpu));
570
571    /* generate "STATE" field */
572    switch (PP(pp, p_stat)) {
573	case SRUN:
574	    if (smpmode && PP(pp, p_oncpu) >= 0)
575		sprintf(status, "CPU%d", PP(pp, p_oncpu));
576	    else
577		strcpy(status, "RUN");
578	    break;
579	case SSLEEP:
580	    if (PP(pp, p_wmesg) != NULL) {
581		sprintf(status, "%.6s", EP(pp, e_wmesg));
582		break;
583	    }
584	    /* fall through */
585	default:
586	    sprintf(status, "%.6s", state_abbrev[(unsigned char) PP(pp, p_stat)]);
587	    break;
588    }
589
590    /* format this entry */
591    sprintf(fmt,
592	    smpmode ? smp_Proc_format : up_Proc_format,
593	    PP(pp, p_pid),
594	    namelength, namelength,
595	    (*get_userid)(EP(pp, e_pcred.p_ruid)),
596	    PP(pp, p_priority) - PZERO,
597	    PP(pp, p_nice) - NZERO,
598	    format_k2(pagetok(PROCSIZE(pp))),
599	    format_k2(pagetok(VP(pp, vm_rssize))),
600	    status,
601	    smpmode ? PP(pp, p_lastcpu) : 0,
602	    format_time(cputime),
603	    10000.0 * weighted_cpu(pct, pp) / hz,
604	    10000.0 * pct / hz,
605	    cmdlength,
606	    printable(PP(pp, p_comm)));
607
608    /* return the result */
609    return(fmt);
610}
611
612
613/*
614 * check_nlist(nlst) - checks the nlist to see if any symbols were not
615 *		found.  For every symbol that was not found, a one-line
616 *		message is printed to stderr.  The routine returns the
617 *		number of symbols NOT found.
618 */
619
620static int check_nlist(nlst)
621
622register struct nlist *nlst;
623
624{
625    register int i;
626
627    /* check to see if we got ALL the symbols we requested */
628    /* this will write one line to stderr for every symbol not found */
629
630    i = 0;
631    while (nlst->n_name != NULL)
632    {
633	if (nlst->n_type == 0)
634	{
635	    /* this one wasn't found */
636	    (void) fprintf(stderr, "kernel: no symbol named `%s'\n",
637			   nlst->n_name);
638	    i = 1;
639	}
640	nlst++;
641    }
642
643    return(i);
644}
645
646
647/*
648 *  getkval(offset, ptr, size, refstr) - get a value out of the kernel.
649 *	"offset" is the byte offset into the kernel for the desired value,
650 *  	"ptr" points to a buffer into which the value is retrieved,
651 *  	"size" is the size of the buffer (and the object to retrieve),
652 *  	"refstr" is a reference string used when printing error meessages,
653 *	    if "refstr" starts with a '!', then a failure on read will not
654 *  	    be fatal (this may seem like a silly way to do things, but I
655 *  	    really didn't want the overhead of another argument).
656 *
657 */
658
659static int getkval(offset, ptr, size, refstr)
660
661unsigned long offset;
662int *ptr;
663int size;
664char *refstr;
665
666{
667    if (kvm_read(kd, offset, (char *) ptr, size) != size)
668    {
669	if (*refstr == '!')
670	{
671	    return(0);
672	}
673	else
674	{
675	    fprintf(stderr, "top: kvm_read for %s: %s\n",
676		refstr, strerror(errno));
677	    quit(23);
678	}
679    }
680    return(1);
681}
682
683/* comparison routine for qsort */
684
685/*
686 *  proc_compare - comparison function for "qsort"
687 *	Compares the resource consumption of two processes using five
688 *  	distinct keys.  The keys (in descending order of importance) are:
689 *  	percent cpu, cpu ticks, state, resident set size, total virtual
690 *  	memory usage.  The process states are ordered as follows (from least
691 *  	to most important):  WAIT, zombie, sleep, stop, start, run.  The
692 *  	array declaration below maps a process state index into a number
693 *  	that reflects this ordering.
694 */
695
696static unsigned char sorted_state[] =
697{
698    0,	/* not used		*/
699    3,	/* sleep		*/
700    1,	/* ABANDONED (WAIT)	*/
701    6,	/* run			*/
702    5,	/* start		*/
703    2,	/* zombie		*/
704    4	/* stop			*/
705};
706
707int
708proc_compare(pp1, pp2)
709
710struct proc **pp1;
711struct proc **pp2;
712
713{
714    register struct kinfo_proc *p1;
715    register struct kinfo_proc *p2;
716    register int result;
717    register pctcpu lresult;
718
719    /* remove one level of indirection */
720    p1 = *(struct kinfo_proc **) pp1;
721    p2 = *(struct kinfo_proc **) pp2;
722
723    /* compare percent cpu (pctcpu) */
724    if ((lresult = PP(p2, p_pctcpu) - PP(p1, p_pctcpu)) == 0)
725    {
726	/* use cpticks to break the tie */
727	if ((result = PP(p2, p_cpticks) - PP(p1, p_cpticks)) == 0)
728	{
729	    /* use process state to break the tie */
730	    if ((result = sorted_state[(unsigned char) PP(p2, p_stat)] -
731			  sorted_state[(unsigned char) PP(p1, p_stat)])  == 0)
732	    {
733		/* use priority to break the tie */
734		if ((result = PP(p2, p_priority) - PP(p1, p_priority)) == 0)
735		{
736		    /* use resident set size (rssize) to break the tie */
737		    if ((result = VP(p2, vm_rssize) - VP(p1, vm_rssize)) == 0)
738		    {
739			/* use total memory to break the tie */
740			result = PROCSIZE(p2) - PROCSIZE(p1);
741		    }
742		}
743	    }
744	}
745    }
746    else
747    {
748	result = lresult < 0 ? -1 : 1;
749    }
750
751    return(result);
752}
753
754
755/*
756 * proc_owner(pid) - returns the uid that owns process "pid", or -1 if
757 *		the process does not exist.
758 *		It is EXTREMLY IMPORTANT that this function work correctly.
759 *		If top runs setuid root (as in SVR4), then this function
760 *		is the only thing that stands in the way of a serious
761 *		security problem.  It validates requests for the "kill"
762 *		and "renice" commands.
763 */
764
765int proc_owner(pid)
766
767int pid;
768
769{
770    register int cnt;
771    register struct kinfo_proc **prefp;
772    register struct kinfo_proc *pp;
773
774    prefp = pref;
775    cnt = pref_len;
776    while (--cnt >= 0)
777    {
778	pp = *prefp++;
779	if (PP(pp, p_pid) == (pid_t)pid)
780	{
781	    return((int)EP(pp, e_pcred.p_ruid));
782	}
783    }
784    return(-1);
785}
786
787
788/*
789 * swapmode is based on a program called swapinfo written
790 * by Kevin Lahey <kml@rokkaku.atl.ga.us>.
791 */
792
793#define	SVAR(var) __STRING(var)	/* to force expansion */
794#define	KGET(idx, var)							\
795	KGET1(idx, &var, sizeof(var), SVAR(var))
796#define	KGET1(idx, p, s, msg)						\
797	KGET2(nlst[idx].n_value, p, s, msg)
798#define	KGET2(addr, p, s, msg)						\
799	if (kvm_read(kd, (u_long)(addr), p, s) != s) {		        \
800		warnx("cannot read %s: %s", msg, kvm_geterr(kd));       \
801		return (0);                                             \
802       }
803#define	KGETRET(addr, p, s, msg)					\
804	if (kvm_read(kd, (u_long)(addr), p, s) != s) {			\
805		warnx("cannot read %s: %s", msg, kvm_geterr(kd));	\
806		return (0);						\
807	}
808
809
810int
811swapmode(retavail, retfree)
812	int *retavail;
813	int *retfree;
814{
815	char *header;
816	int hlen, nswap, nswdev, dmmax;
817	int i, div, avail, nfree, npfree, used;
818	struct swdevt *sw;
819	long blocksize, *perdev;
820	u_long ptr;
821	struct rlist head;
822#if __FreeBSD_version >= 220000
823	struct rlisthdr swaplist;
824#else
825	struct rlist *swaplist;
826#endif
827	struct rlist *swapptr;
828
829	/*
830	 * Counter for error messages. If we reach the limit,
831	 * stop reading information from swap devices and
832	 * return zero. This prevent endless 'bad address'
833	 * messages.
834	 */
835	static warning = 10;
836
837	if (warning <= 0) {
838	    /* a single warning */
839	    if (!warning) {
840		warning--;
841		fprintf(stderr,
842			"Too much errors, stop reading swap devices ...\n");
843		(void)sleep(3);
844	    }
845	    return(0);
846	}
847	warning--; /* decrease counter, see end of function */
848
849	KGET(VM_NSWAP, nswap);
850	if (!nswap) {
851		fprintf(stderr, "No swap space available\n");
852		return(0);
853	}
854
855	KGET(VM_NSWDEV, nswdev);
856	KGET(VM_DMMAX, dmmax);
857	KGET1(VM_SWAPLIST, &swaplist, sizeof(swaplist), "swaplist");
858	if ((sw = (struct swdevt *)malloc(nswdev * sizeof(*sw))) == NULL ||
859	    (perdev = (long *)malloc(nswdev * sizeof(*perdev))) == NULL)
860		err(1, "malloc");
861	KGET1(VM_SWDEVT, &ptr, sizeof ptr, "swdevt");
862	KGET2(ptr, sw, nswdev * sizeof(*sw), "*swdevt");
863
864	/* Count up swap space. */
865	nfree = 0;
866	memset(perdev, 0, nswdev * sizeof(*perdev));
867#if  __FreeBSD_version >= 220000
868	swapptr = swaplist.rlh_list;
869	while (swapptr) {
870#else
871	while (swaplist) {
872#endif
873		int	top, bottom, next_block;
874#if  __FreeBSD_version >= 220000
875		KGET2(swapptr, &head, sizeof(struct rlist), "swapptr");
876#else
877		KGET2(swaplist, &head, sizeof(struct rlist), "swaplist");
878#endif
879
880		top = head.rl_end;
881		bottom = head.rl_start;
882
883		nfree += top - bottom + 1;
884
885		/*
886		 * Swap space is split up among the configured disks.
887		 *
888		 * For interleaved swap devices, the first dmmax blocks
889		 * of swap space some from the first disk, the next dmmax
890		 * blocks from the next, and so on up to nswap blocks.
891		 *
892		 * The list of free space joins adjacent free blocks,
893		 * ignoring device boundries.  If we want to keep track
894		 * of this information per device, we'll just have to
895		 * extract it ourselves.
896		 */
897		while (top / dmmax != bottom / dmmax) {
898			next_block = ((bottom + dmmax) / dmmax);
899			perdev[(bottom / dmmax) % nswdev] +=
900				next_block * dmmax - bottom;
901			bottom = next_block * dmmax;
902		}
903		perdev[(bottom / dmmax) % nswdev] +=
904			top - bottom + 1;
905
906#if  __FreeBSD_version >= 220000
907		swapptr = head.rl_next;
908#else
909		swaplist = head.rl_next;
910#endif
911	}
912
913	header = getbsize(&hlen, &blocksize);
914	div = blocksize / 512;
915	avail = npfree = 0;
916	for (i = 0; i < nswdev; i++) {
917		int xsize, xfree;
918
919		/*
920		 * Don't report statistics for partitions which have not
921		 * yet been activated via swapon(8).
922		 */
923
924		xsize = sw[i].sw_nblks;
925		xfree = perdev[i];
926		used = xsize - xfree;
927		npfree++;
928		avail += xsize;
929	}
930
931	/*
932	 * If only one partition has been set up via swapon(8), we don't
933	 * need to bother with totals.
934	 */
935	*retavail = avail / 2;
936	*retfree = nfree / 2;
937	used = avail - nfree;
938	free(sw); free(perdev);
939
940	/* increase counter, no errors occurs */
941	warning++;
942
943	return  (int)(((double)used / (double)avail * 100.0) + 0.5);
944}
945