machine.c revision 1.11 
1/*	$OpenBSD: machine.c,v 1.11 1998/08/21 13:55:23 kstailey Exp $	*/
2
3/*
4 * top - a top users display for Unix
5 *
6 * SYNOPSIS:  For an OpenBSD system
7 *
8 * DESCRIPTION:
9 * This is the machine-dependent module for OpenBSD
10 * Tested on:
11 *	i386
12 *
13 * LIBS: -lkvm
14 *
15 * TERMCAP: -ltermlib
16 *
17 * CFLAGS: -DHAVE_GETOPT -DORDER
18 *
19 * AUTHOR:  Thorsten Lockert <tholo@sigmasoft.com>
20 *          Adapted from BSD4.4 by Christos Zoulas <christos@ee.cornell.edu>
21 *          Patch for process wait display by Jarl F. Greipsland <jarle@idt.unit.no>
22 *	    Patch for -DORDER by Kenneth Stailey <kstailey@disclosure.com>
23 */
24
25#include <sys/types.h>
26#include <sys/signal.h>
27#include <sys/param.h>
28
29#define DOSWAP
30
31#include <stdio.h>
32#include <stdlib.h>
33#include <string.h>
34#include <limits.h>
35#include <err.h>
36#include <nlist.h>
37#include <math.h>
38#include <kvm.h>
39#include <unistd.h>
40#include <sys/errno.h>
41#include <sys/sysctl.h>
42#include <sys/dir.h>
43#include <sys/dkstat.h>
44#include <sys/file.h>
45#include <sys/time.h>
46#include <sys/resource.h>
47
48#ifdef DOSWAP
49#include <err.h>
50#include <sys/map.h>
51#include <sys/conf.h>
52#endif
53
54static int check_nlist __P((struct nlist *));
55static int getkval __P((unsigned long, int *, int, char *));
56static int swapmode __P((int *, int *));
57
58#include "top.h"
59#include "display.h"
60#include "machine.h"
61#include "utils.h"
62
63/* get_process_info passes back a handle.  This is what it looks like: */
64
65struct handle
66{
67    struct kinfo_proc **next_proc;	/* points to next valid proc pointer */
68    int remaining;		/* number of pointers remaining */
69};
70
71/* declarations for load_avg */
72#include "loadavg.h"
73
74#define PP(pp, field) ((pp)->kp_proc . field)
75#define EP(pp, field) ((pp)->kp_eproc . field)
76#define VP(pp, field) ((pp)->kp_eproc.e_vm . field)
77
78/* what we consider to be process size: */
79#define PROCSIZE(pp) (VP((pp), vm_tsize) + VP((pp), vm_dsize) + VP((pp), vm_ssize))
80
81/* definitions for indices in the nlist array */
82#define X_CP_TIME	0
83#define X_HZ		1
84
85#ifdef DOSWAP
86#define	VM_SWAPMAP	2
87#define	VM_NSWAPMAP	3
88#define	VM_SWDEVT	4
89#define	VM_NSWAP	5
90#define	VM_NSWDEV	6
91#define	VM_DMMAX	7
92#define	VM_NISWAP	8
93#define	VM_NISWDEV	9
94#endif
95
96static struct nlist nlst[] = {
97    { "_cp_time" },		/* 0 */
98    { "_hz" },			/* 1 */
99#ifdef DOSWAP
100    { "_swapmap" },		/* 2 */
101    { "_nswapmap" },		/* 3 */
102    { "_swdevt" },		/* 4 */
103    { "_nswap" },		/* 5 */
104    { "_nswdev" },		/* 6 */
105    { "_dmmax" },		/* 7 */
106    { "_niswap" },		/* 8 */
107    { "_niswdev" },		/* 9 */
108#endif
109    { 0 }
110};
111
112/*
113 *  These definitions control the format of the per-process area
114 */
115
116static char header[] =
117  "  PID X        PRI NICE  SIZE   RES STATE WAIT     TIME    CPU COMMAND";
118/* 0123456   -- field to fill in starts at header+6 */
119#define UNAME_START 6
120
121#define Proc_format \
122	"%5d %-8.8s %3d %4d %5s %5s %-5s %-6.6s %6s %5.2f%% %.14s"
123
124
125/* process state names for the "STATE" column of the display */
126/* the extra nulls in the string "run" are for adding a slash and
127   the processor number when needed */
128
129char *state_abbrev[] =
130{
131    "", "start", "run\0\0\0", "sleep", "stop", "zomb",
132};
133
134
135static kvm_t *kd;
136
137/* these are retrieved from the kernel in _init */
138
139static          int hz;
140
141/* these are offsets obtained via nlist and used in the get_ functions */
142
143static unsigned long cp_time_offset;
144
145/* these are for calculating cpu state percentages */
146static int cp_time[CPUSTATES];
147static int cp_old[CPUSTATES];
148static int cp_diff[CPUSTATES];
149
150/* these are for detailing the process states */
151
152int process_states[7];
153char *procstatenames[] = {
154    "", " starting, ", " running, ", " idle, ", " stopped, ", " zombie, ",
155    NULL
156};
157
158/* these are for detailing the cpu states */
159
160int cpu_states[CPUSTATES];
161char *cpustatenames[] = {
162    "user", "nice", "system", "interrupt", "idle", NULL
163};
164
165/* these are for detailing the memory statistics */
166
167int memory_stats[8];
168char *memorynames[] = {
169    "Real: ", "K/", "K act/tot  ", "Free: ", "K  ",
170#ifdef DOSWAP
171    "Swap: ", "K/", "K used/tot",
172#endif
173    NULL
174};
175
176#ifdef ORDER
177/* these are names given to allowed sorting orders -- first is default */
178
179char *ordernames[] = {"cpu", "size", "res", "time", "pri", NULL};
180#endif
181
182/* these are for keeping track of the proc array */
183
184static int nproc;
185static int onproc = -1;
186static int pref_len;
187static struct kinfo_proc *pbase;
188static struct kinfo_proc **pref;
189
190/* these are for getting the memory statistics */
191
192static int pageshift;		/* log base 2 of the pagesize */
193
194/* define pagetok in terms of pageshift */
195
196#define pagetok(size) ((size) << pageshift)
197
198int
199machine_init(statics)
200
201struct statics *statics;
202
203{
204    register int i = 0;
205    register int pagesize;
206    char errbuf[_POSIX2_LINE_MAX];
207
208    if ((kd = kvm_openfiles(NULL, NULL, NULL, O_RDONLY, errbuf)) == NULL) {
209	warnx("%s", errbuf);
210	return(-1);
211    }
212
213    setegid(getgid());
214    setgid(getgid());
215
216    /* get the list of symbols we want to access in the kernel */
217    if (kvm_nlist(kd, nlst) <= 0) {
218	warnx("nlist failed");
219	return(-1);
220    }
221
222    /* make sure they were all found */
223    if (i > 0 && check_nlist(nlst) > 0)
224	return(-1);
225
226    /* get the symbol values out of kmem */
227    (void) getkval(nlst[X_HZ].n_value,     (int *)(&hz),	sizeof(hz),
228	    nlst[X_HZ].n_name);
229
230    /* stash away certain offsets for later use */
231    cp_time_offset = nlst[X_CP_TIME].n_value;
232
233    pbase = NULL;
234    pref = NULL;
235    onproc = -1;
236    nproc = 0;
237
238    /* get the page size with "getpagesize" and calculate pageshift from it */
239    pagesize = getpagesize();
240    pageshift = 0;
241    while (pagesize > 1)
242    {
243	pageshift++;
244	pagesize >>= 1;
245    }
246
247    /* we only need the amount of log(2)1024 for our conversion */
248    pageshift -= LOG1024;
249
250    /* fill in the statics information */
251    statics->procstate_names = procstatenames;
252    statics->cpustate_names = cpustatenames;
253    statics->memory_names = memorynames;
254#ifdef ORDER
255    statics->order_names = ordernames;
256#endif
257
258    /* all done! */
259    return(0);
260}
261
262char *format_header(uname_field)
263
264register char *uname_field;
265
266{
267    register char *ptr;
268
269    ptr = header + UNAME_START;
270    while (*uname_field != '\0')
271    {
272	*ptr++ = *uname_field++;
273    }
274
275    return(header);
276}
277
278void
279get_system_info(si)
280
281struct system_info *si;
282
283{
284    int total;
285
286    /* get the cp_time array */
287    (void) getkval(cp_time_offset, (int *)cp_time, sizeof(cp_time),
288		   "_cp_time");
289
290    /* convert load averages to doubles */
291    {
292	register int i;
293	register double *infoloadp;
294	struct loadavg sysload;
295	size_t size = sizeof(sysload);
296	static int mib[] = { CTL_VM, VM_LOADAVG };
297
298	if (sysctl(mib, 2, &sysload, &size, NULL, 0) < 0) {
299	    warn("sysctl failed");
300	    bzero(&total, sizeof(total));
301	}
302
303	infoloadp = si->load_avg;
304	for (i = 0; i < 3; i++)
305	    *infoloadp++ = ((double) sysload.ldavg[i]) / sysload.fscale;
306    }
307
308    /* convert cp_time counts to percentages */
309    total = percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff);
310
311    /* sum memory statistics */
312    {
313	struct vmtotal total;
314	size_t size = sizeof(total);
315	static int mib[] = { CTL_VM, VM_METER };
316
317	/* get total -- systemwide main memory usage structure */
318	if (sysctl(mib, 2, &total, &size, NULL, 0) < 0) {
319	    warn("sysctl failed");
320	    bzero(&total, sizeof(total));
321	}
322	/* convert memory stats to Kbytes */
323	memory_stats[0] = -1;
324	memory_stats[1] = pagetok(total.t_arm);
325	memory_stats[2] = pagetok(total.t_rm);
326	memory_stats[3] = -1;
327	memory_stats[4] = pagetok(total.t_free);
328	memory_stats[5] = -1;
329#ifdef DOSWAP
330	if (!swapmode(&memory_stats[6], &memory_stats[7])) {
331	    memory_stats[6] = 0;
332	    memory_stats[7] = 0;
333	}
334#endif
335    }
336
337    /* set arrays and strings */
338    si->cpustates = cpu_states;
339    si->memory = memory_stats;
340    si->last_pid = -1;
341}
342
343static struct handle handle;
344
345caddr_t get_process_info(si, sel, compare)
346
347struct system_info *si;
348struct process_select *sel;
349int (*compare) __P((const void *, const void *));
350
351{
352    register int i;
353    register int total_procs;
354    register int active_procs;
355    register struct kinfo_proc **prefp;
356    register struct kinfo_proc *pp;
357
358    /* these are copied out of sel for speed */
359    int show_idle;
360    int show_system;
361    int show_uid;
362    int show_command;
363
364
365    if ((pbase = kvm_getprocs(kd, KERN_PROC_ALL, 0, &nproc)) == NULL) {
366	warnx("%s", kvm_geterr(kd));
367	quit(23);
368    }
369    if (nproc > onproc)
370	pref = (struct kinfo_proc **) realloc(pref, sizeof(struct kinfo_proc *)
371		* (onproc = nproc));
372    if (pref == NULL) {
373	warnx("Out of memory.");
374	quit(23);
375    }
376    /* get a pointer to the states summary array */
377    si->procstates = process_states;
378
379    /* set up flags which define what we are going to select */
380    show_idle = sel->idle;
381    show_system = sel->system;
382    show_uid = sel->uid != -1;
383    show_command = sel->command != NULL;
384
385    /* count up process states and get pointers to interesting procs */
386    total_procs = 0;
387    active_procs = 0;
388    memset((char *)process_states, 0, sizeof(process_states));
389    prefp = pref;
390    for (pp = pbase, i = 0; i < nproc; pp++, i++)
391    {
392	/*
393	 *  Place pointers to each valid proc structure in pref[].
394	 *  Process slots that are actually in use have a non-zero
395	 *  status field.  Processes with SSYS set are system
396	 *  processes---these get ignored unless show_sysprocs is set.
397	 */
398	if (PP(pp, p_stat) != 0 &&
399	    (show_system || ((PP(pp, p_flag) & P_SYSTEM) == 0)))
400	{
401	    total_procs++;
402	    process_states[(unsigned char) PP(pp, p_stat)]++;
403	    if ((PP(pp, p_stat) != SZOMB) &&
404		(show_idle || (PP(pp, p_pctcpu) != 0) ||
405		 (PP(pp, p_stat) == SRUN)) &&
406		(!show_uid || EP(pp, e_pcred.p_ruid) == (uid_t)sel->uid))
407	    {
408		*prefp++ = pp;
409		active_procs++;
410	    }
411	}
412    }
413
414    /* if requested, sort the "interesting" processes */
415    if (compare != NULL)
416    {
417	qsort((char *)pref, active_procs, sizeof(struct kinfo_proc *), compare);
418    }
419
420    /* remember active and total counts */
421    si->p_total = total_procs;
422    si->p_active = pref_len = active_procs;
423
424    /* pass back a handle */
425    handle.next_proc = pref;
426    handle.remaining = active_procs;
427    return((caddr_t)&handle);
428}
429
430char fmt[MAX_COLS];		/* static area where result is built */
431
432char *format_next_process(handle, get_userid)
433
434caddr_t handle;
435char *(*get_userid)();
436
437{
438    register struct kinfo_proc *pp;
439    register int cputime;
440    register double pct;
441    struct handle *hp;
442    char waddr[sizeof(void *) * 2 + 3];	/* Hexify void pointer */
443    char *p_wait;
444
445    /* find and remember the next proc structure */
446    hp = (struct handle *)handle;
447    pp = *(hp->next_proc++);
448    hp->remaining--;
449
450
451    /* get the process's user struct and set cputime */
452    if ((PP(pp, p_flag) & P_INMEM) == 0) {
453	/*
454	 * Print swapped processes as <pname>
455	 */
456	char *comm = PP(pp, p_comm);
457#define COMSIZ sizeof(PP(pp, p_comm))
458	char buf[COMSIZ];
459	(void) strncpy(buf, comm, COMSIZ);
460	comm[0] = '<';
461	(void) strncpy(&comm[1], buf, COMSIZ - 2);
462	comm[COMSIZ - 2] = '\0';
463	(void) strncat(comm, ">", COMSIZ - 1);
464	comm[COMSIZ - 1] = '\0';
465    }
466
467    cputime = (PP(pp, p_uticks) + PP(pp, p_sticks) + PP(pp, p_iticks)) / hz;
468
469    /* calculate the base for cpu percentages */
470    pct = pctdouble(PP(pp, p_pctcpu));
471
472    if (PP(pp, p_wchan))
473        if (PP(pp, p_wmesg))
474	    p_wait = EP(pp, e_wmesg);
475	else {
476	    snprintf(waddr, sizeof(waddr), "%lx",
477		(unsigned long)(PP(pp, p_wchan)) & ~KERNBASE);
478	    p_wait = waddr;
479        }
480    else
481	p_wait = "-";
482
483    /* format this entry */
484    snprintf(fmt, MAX_COLS,
485	    Proc_format,
486	    PP(pp, p_pid),
487	    (*get_userid)(EP(pp, e_pcred.p_ruid)),
488	    PP(pp, p_priority) - PZERO,
489	    PP(pp, p_nice) - NZERO,
490	    format_k(pagetok(PROCSIZE(pp))),
491	    format_k(pagetok(VP(pp, vm_rssize))),
492	    (PP(pp, p_stat) == SSLEEP && PP(pp, p_slptime) > MAXSLP)
493	     ? "idle" : state_abbrev[(unsigned char) PP(pp, p_stat)],
494	    p_wait,
495	    format_time(cputime),
496	    100.0 * pct,
497	    printable(PP(pp, p_comm)));
498
499    /* return the result */
500    return(fmt);
501}
502
503
504/*
505 * check_nlist(nlst) - checks the nlist to see if any symbols were not
506 *		found.  For every symbol that was not found, a one-line
507 *		message is printed to stderr.  The routine returns the
508 *		number of symbols NOT found.
509 */
510
511static int check_nlist(nlst)
512
513register struct nlist *nlst;
514
515{
516    register int i;
517
518    /* check to see if we got ALL the symbols we requested */
519    /* this will write one line to stderr for every symbol not found */
520
521    i = 0;
522    while (nlst->n_name != NULL)
523    {
524	if (nlst->n_type == 0)
525	{
526	    /* this one wasn't found */
527	    (void) fprintf(stderr, "kernel: no symbol named `%s'\n",
528			   nlst->n_name);
529	    i = 1;
530	}
531	nlst++;
532    }
533
534    return(i);
535}
536
537
538/*
539 *  getkval(offset, ptr, size, refstr) - get a value out of the kernel.
540 *	"offset" is the byte offset into the kernel for the desired value,
541 *  	"ptr" points to a buffer into which the value is retrieved,
542 *  	"size" is the size of the buffer (and the object to retrieve),
543 *  	"refstr" is a reference string used when printing error meessages,
544 *	    if "refstr" starts with a '!', then a failure on read will not
545 *  	    be fatal (this may seem like a silly way to do things, but I
546 *  	    really didn't want the overhead of another argument).
547 *
548 */
549
550static int getkval(offset, ptr, size, refstr)
551
552unsigned long offset;
553int *ptr;
554int size;
555char *refstr;
556
557{
558    if (kvm_read(kd, offset, (char *) ptr, size) != size)
559    {
560	if (*refstr == '!')
561	{
562	    return(0);
563	}
564	else
565	{
566	    warn("kvm_read for %s", refstr);
567	    quit(23);
568	}
569    }
570    return(1);
571}
572
573/* comparison routine for qsort */
574
575static unsigned char sorted_state[] =
576{
577    0,	/* not used		*/
578    4,	/* start		*/
579    5,	/* run			*/
580    2,	/* sleep		*/
581    3,	/* stop			*/
582    1	/* zombie		*/
583};
584
585#ifdef ORDER
586
587/*
588 *  proc_compares - comparison functions for "qsort"
589 */
590
591/*
592 * First, the possible comparison keys.  These are defined in such a way
593 * that they can be merely listed in the source code to define the actual
594 * desired ordering.
595 */
596
597
598#define ORDERKEY_PCTCPU \
599	if (lresult = PP(p2, p_pctcpu) - PP(p1, p_pctcpu), \
600           (result = lresult > 0 ? 1 : lresult < 0 ? -1 : 0) == 0)
601#define ORDERKEY_CPUTIME \
602	if ((result = PP(p2, p_rtime.tv_sec) - PP(p1, p_rtime.tv_sec)) == 0) \
603		if ((result = PP(p2, p_rtime.tv_usec) - \
604		     PP(p1, p_rtime.tv_usec)) == 0)
605#define ORDERKEY_STATE \
606	if ((result = sorted_state[(unsigned char) PP(p2, p_stat)] - \
607                      sorted_state[(unsigned char) PP(p1, p_stat)])  == 0)
608#define ORDERKEY_PRIO \
609	if ((result = PP(p2, p_priority) - PP(p1, p_priority)) == 0)
610#define ORDERKEY_RSSIZE \
611	if ((result = VP(p2, vm_rssize) - VP(p1, vm_rssize)) == 0)
612#define ORDERKEY_MEM \
613	if ((result = PROCSIZE(p2) - PROCSIZE(p1)) == 0)
614
615
616/* compare_cpu - the comparison function for sorting by cpu percentage */
617
618int
619compare_cpu(v1, v2)
620
621const void *v1, *v2;
622
623{
624    register struct proc **pp1 = (struct proc **)v1;
625    register struct proc **pp2 = (struct proc **)v2;
626    register struct kinfo_proc *p1;
627    register struct kinfo_proc *p2;
628    register int result;
629    register pctcpu lresult;
630
631    /* remove one level of indirection */
632    p1 = *(struct kinfo_proc **) pp1;
633    p2 = *(struct kinfo_proc **) pp2;
634
635    ORDERKEY_PCTCPU
636    ORDERKEY_CPUTIME
637    ORDERKEY_STATE
638    ORDERKEY_PRIO
639    ORDERKEY_RSSIZE
640    ORDERKEY_MEM
641    ;
642    return(result);
643}
644
645/* compare_size - the comparison function for sorting by total memory usage */
646
647int
648compare_size(v1, v2)
649
650const void *v1, *v2;
651
652{
653    register struct proc **pp1 = (struct proc **)v1;
654    register struct proc **pp2 = (struct proc **)v2;
655    register struct kinfo_proc *p1;
656    register struct kinfo_proc *p2;
657    register int result;
658    register pctcpu lresult;
659
660    /* remove one level of indirection */
661    p1 = *(struct kinfo_proc **) pp1;
662    p2 = *(struct kinfo_proc **) pp2;
663
664    ORDERKEY_MEM
665    ORDERKEY_RSSIZE
666    ORDERKEY_PCTCPU
667    ORDERKEY_CPUTIME
668    ORDERKEY_STATE
669    ORDERKEY_PRIO
670    ;
671
672    return(result);
673}
674
675/* compare_res - the comparison function for sorting by resident set size */
676
677int
678compare_res(v1, v2)
679
680const void *v1, *v2;
681
682{
683    register struct proc **pp1 = (struct proc **)v1;
684    register struct proc **pp2 = (struct proc **)v2;
685    register struct kinfo_proc *p1;
686    register struct kinfo_proc *p2;
687    register int result;
688    register pctcpu lresult;
689
690    /* remove one level of indirection */
691    p1 = *(struct kinfo_proc **) pp1;
692    p2 = *(struct kinfo_proc **) pp2;
693
694    ORDERKEY_RSSIZE
695    ORDERKEY_MEM
696    ORDERKEY_PCTCPU
697    ORDERKEY_CPUTIME
698    ORDERKEY_STATE
699    ORDERKEY_PRIO
700    ;
701
702    return(result);
703}
704
705/* compare_time - the comparison function for sorting by CPU time */
706
707int
708compare_time(v1, v2)
709
710const void *v1, *v2;
711
712{
713    register struct proc **pp1 = (struct proc **)v1;
714    register struct proc **pp2 = (struct proc **)v2;
715    register struct kinfo_proc *p1;
716    register struct kinfo_proc *p2;
717    register int result;
718    register pctcpu lresult;
719
720    /* remove one level of indirection */
721    p1 = *(struct kinfo_proc **) pp1;
722    p2 = *(struct kinfo_proc **) pp2;
723
724    ORDERKEY_CPUTIME
725    ORDERKEY_PCTCPU
726    ORDERKEY_STATE
727    ORDERKEY_PRIO
728    ORDERKEY_MEM
729    ORDERKEY_RSSIZE
730    ;
731
732    return(result);
733}
734
735/* compare_prio - the comparison function for sorting by CPU time */
736
737int
738compare_prio(v1, v2)
739
740const void *v1, *v2;
741
742{
743    register struct proc **pp1 = (struct proc **)v1;
744    register struct proc **pp2 = (struct proc **)v2;
745    register struct kinfo_proc *p1;
746    register struct kinfo_proc *p2;
747    register int result;
748    register pctcpu lresult;
749
750    /* remove one level of indirection */
751    p1 = *(struct kinfo_proc **) pp1;
752    p2 = *(struct kinfo_proc **) pp2;
753
754    ORDERKEY_PRIO
755    ORDERKEY_PCTCPU
756    ORDERKEY_CPUTIME
757    ORDERKEY_STATE
758    ORDERKEY_RSSIZE
759    ORDERKEY_MEM
760    ;
761
762    return(result);
763}
764
765int (*proc_compares[])() = {
766    compare_cpu,
767    compare_size,
768    compare_res,
769    compare_time,
770    compare_prio,
771    NULL
772};
773#else
774/*
775 *  proc_compare - comparison function for "qsort"
776 *	Compares the resource consumption of two processes using five
777 *  	distinct keys.  The keys (in descending order of importance) are:
778 *  	percent cpu, cpu ticks, state, resident set size, total virtual
779 *  	memory usage.  The process states are ordered as follows (from least
780 *  	to most important):  zombie, sleep, stop, start, run.  The array
781 *  	declaration below maps a process state index into a number that
782 *  	reflects this ordering.
783 */
784
785int
786proc_compare(v1, v2)
787
788const void *v1, *v2;
789
790{
791    register struct proc **pp1 = (struct proc **)v1;
792    register struct proc **pp2 = (struct proc **)v2;
793    register struct kinfo_proc *p1;
794    register struct kinfo_proc *p2;
795    register int result;
796    register pctcpu lresult;
797
798    /* remove one level of indirection */
799    p1 = *(struct kinfo_proc **) pp1;
800    p2 = *(struct kinfo_proc **) pp2;
801
802    /* compare percent cpu (pctcpu) */
803    if ((lresult = PP(p2, p_pctcpu) - PP(p1, p_pctcpu)) == 0)
804    {
805	/* use CPU usage to break the tie */
806	if ((result = PP(p2, p_rtime).tv_sec - PP(p1, p_rtime).tv_sec) == 0)
807	{
808	    /* use process state to break the tie */
809	    if ((result = sorted_state[(unsigned char) PP(p2, p_stat)] -
810			  sorted_state[(unsigned char) PP(p1, p_stat)])  == 0)
811	    {
812		/* use priority to break the tie */
813		if ((result = PP(p2, p_priority) - PP(p1, p_priority)) == 0)
814		{
815		    /* use resident set size (rssize) to break the tie */
816		    if ((result = VP(p2, vm_rssize) - VP(p1, vm_rssize)) == 0)
817		    {
818			/* use total memory to break the tie */
819			result = PROCSIZE(p2) - PROCSIZE(p1);
820		    }
821		}
822	    }
823	}
824    }
825    else
826    {
827	result = lresult < 0 ? -1 : 1;
828    }
829
830    return(result);
831}
832#endif
833
834/*
835 * proc_owner(pid) - returns the uid that owns process "pid", or -1 if
836 *		the process does not exist.
837 *		It is EXTREMLY IMPORTANT that this function work correctly.
838 *		If top runs setuid root (as in SVR4), then this function
839 *		is the only thing that stands in the way of a serious
840 *		security problem.  It validates requests for the "kill"
841 *		and "renice" commands.
842 */
843
844int proc_owner(pid)
845
846pid_t pid;
847
848{
849    register int cnt;
850    register struct kinfo_proc **prefp;
851    register struct kinfo_proc *pp;
852
853    prefp = pref;
854    cnt = pref_len;
855    while (--cnt >= 0)
856    {
857	pp = *prefp++;
858	if (PP(pp, p_pid) == pid)
859	{
860	    return((int)EP(pp, e_pcred.p_ruid));
861	}
862    }
863    return(-1);
864}
865
866#ifdef DOSWAP
867/*
868 * swapmode is based on a program called swapinfo written
869 * by Kevin Lahey <kml@rokkaku.atl.ga.us>.
870 */
871
872#define	SVAR(var) __STRING(var)	/* to force expansion */
873#define	KGET(idx, var)							\
874	KGET1(idx, &var, sizeof(var), SVAR(var))
875#define	KGET1(idx, p, s, msg)						\
876	KGET2(nlst[idx].n_value, p, s, msg)
877#define	KGET2(addr, p, s, msg)						\
878	if (kvm_read(kd, (u_long)(addr), p, s) != s)			\
879		warnx("cannot read %s: %s", msg, kvm_geterr(kd))
880
881static int
882swapmode(used, total)
883int *used;
884int *total;
885{
886	int nswap, nswdev, dmmax, nswapmap, niswap, niswdev;
887	int s, e, i, l, nfree;
888	struct swdevt *sw;
889	long *perdev;
890	struct map *swapmap, *kswapmap;
891	struct mapent *mp, *freemp;
892
893	KGET(VM_NSWAP, nswap);
894	KGET(VM_NSWDEV, nswdev);
895	KGET(VM_DMMAX, dmmax);
896	KGET(VM_NSWAPMAP, nswapmap);
897	KGET(VM_SWAPMAP, kswapmap);	/* kernel `swapmap' is a pointer */
898	if (nswap == 0) {
899		*used = 0;
900		*total = 0;
901		return (1);
902	}
903	if ((sw = malloc(nswdev * sizeof(*sw))) == NULL ||
904	    (perdev = malloc(nswdev * sizeof(*perdev))) == NULL ||
905	    (freemp = mp = malloc(nswapmap * sizeof(*mp))) == NULL)
906		err(1, "malloc");
907	KGET1(VM_SWDEVT, sw, nswdev * sizeof(*sw), "swdevt");
908	KGET2((long)kswapmap, mp, nswapmap * sizeof(*mp), "swapmap");
909
910	/* Supports sequential swap */
911	if (nlst[VM_NISWAP].n_value != 0) {
912		KGET(VM_NISWAP, niswap);
913		KGET(VM_NISWDEV, niswdev);
914	} else {
915		niswap = nswap;
916		niswdev = nswdev;
917	}
918
919	/* First entry in map is `struct map'; rest are mapent's. */
920	swapmap = (struct map *)mp;
921	if (nswapmap != swapmap->m_limit - (struct mapent *)kswapmap)
922		errx(1, "panic: nswapmap goof");
923
924	/* Count up swap space. */
925	nfree = 0;
926	memset(perdev, 0, nswdev * sizeof(*perdev));
927	for (mp++; mp->m_addr != 0; mp++) {
928		s = mp->m_addr;			/* start of swap region */
929		e = mp->m_addr + mp->m_size;	/* end of region */
930		nfree += mp->m_size;
931
932		/*
933		 * Swap space is split up among the configured disks.
934		 *
935		 * For interleaved swap devices, the first dmmax blocks
936		 * of swap space some from the first disk, the next dmmax
937		 * blocks from the next, and so on up to niswap blocks.
938		 *
939		 * Sequential swap devices follow the interleaved devices
940		 * (i.e. blocks starting at niswap) in the order in which
941		 * they appear in the swdev table.  The size of each device
942		 * will be a multiple of dmmax.
943		 *
944		 * The list of free space joins adjacent free blocks,
945		 * ignoring device boundries.  If we want to keep track
946		 * of this information per device, we'll just have to
947		 * extract it ourselves.  We know that dmmax-sized chunks
948		 * cannot span device boundaries (interleaved or sequential)
949		 * so we loop over such chunks assigning them to devices.
950		 */
951		i = -1;
952		while (s < e) {		/* XXX this is inefficient */
953			int bound = roundup(s+1, dmmax);
954
955			if (bound > e)
956				bound = e;
957			if (bound <= niswap) {
958				/* Interleaved swap chunk. */
959				if (i == -1)
960					i = (s / dmmax) % niswdev;
961				perdev[i] += bound - s;
962				if (++i >= niswdev)
963					i = 0;
964			} else {
965				/* Sequential swap chunk. */
966				if (i < niswdev) {
967					i = niswdev;
968					l = niswap + sw[i].sw_nblks;
969				}
970				while (s >= l) {
971					/* XXX don't die on bogus blocks */
972					if (i == nswdev-1)
973						break;
974					l += sw[++i].sw_nblks;
975				}
976				perdev[i] += bound - s;
977			}
978			s = bound;
979		}
980	}
981
982	*total = 0;
983	for (i = 0; i < nswdev; i++) {
984		int xsize, xfree;
985
986		xsize = sw[i].sw_nblks;
987		xfree = perdev[i];
988		*total += xsize;
989	}
990
991	/*
992	 * If only one partition has been set up via swapon(8), we don't
993	 * need to bother with totals.
994	 */
995#if DEV_BSHIFT < 10
996	*used = (*total - nfree) >> (10 - DEV_BSHIFT);
997	*total >>= 10 - DEV_BSHIFT;
998#elif DEV_BSHIFT > 10
999	*used = (*total - nfree) >> (DEV_BSHIFT - 10);
1000	*total >>= DEV_BSHIFT - 10;
1001#endif
1002	free (sw); free (freemp); free (perdev);
1003	return 1;
1004}
1005#endif
1006