xref: /netbsd-src/sys/miscfs/procfs/procfs_linux.c (revision 736a4d9b7870c657841320ebee34cd39d1c751e5)
1 /*      $NetBSD: procfs_linux.c,v 1.54 2008/05/31 21:34:42 ad Exp $      */
2 
3 /*
4  * Copyright (c) 2001 Wasabi Systems, Inc.
5  * All rights reserved.
6  *
7  * Written by Frank van der Linden for Wasabi Systems, Inc.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. All advertising materials mentioning features or use of this software
18  *    must display the following acknowledgement:
19  *      This product includes software developed for the NetBSD Project by
20  *      Wasabi Systems, Inc.
21  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
22  *    or promote products derived from this software without specific prior
23  *    written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
27  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
28  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
29  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35  * POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <sys/cdefs.h>
39 __KERNEL_RCSID(0, "$NetBSD: procfs_linux.c,v 1.54 2008/05/31 21:34:42 ad Exp $");
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/time.h>
44 #include <sys/kernel.h>
45 #include <sys/proc.h>
46 #include <sys/vnode.h>
47 #include <sys/exec.h>
48 #include <sys/resource.h>
49 #include <sys/resourcevar.h>
50 #include <sys/signal.h>
51 #include <sys/signalvar.h>
52 #include <sys/tty.h>
53 #include <sys/malloc.h>
54 #include <sys/mount.h>
55 #include <sys/conf.h>
56 
57 #include <miscfs/procfs/procfs.h>
58 #include <miscfs/specfs/specdev.h>
59 
60 #include <compat/linux/common/linux_exec.h>
61 
62 #include <uvm/uvm_extern.h>
63 #include <uvm/uvm.h>
64 
65 extern struct devsw_conv *devsw_conv;
66 extern int max_devsw_convs;
67 
68 #define PGTOB(p)	((unsigned long)(p) << PAGE_SHIFT)
69 #define PGTOKB(p)	((unsigned long)(p) << (PAGE_SHIFT - 10))
70 
71 #define LBFSZ (8 * 1024)
72 
73 static void
74 get_proc_size_info(struct lwp *l, unsigned long *stext, unsigned long *etext, unsigned long *sstack)
75 {
76 	struct proc *p = l->l_proc;
77 	struct vmspace *vm;
78 	struct vm_map *map;
79 	struct vm_map_entry *entry;
80 
81 	*stext = 0;
82 	*etext = 0;
83 	*sstack = 0;
84 
85 	proc_vmspace_getref(p, &vm);
86 	map = &vm->vm_map;
87 	vm_map_lock_read(map);
88 
89 	for (entry = map->header.next; entry != &map->header;
90 	    entry = entry->next) {
91 		if (UVM_ET_ISSUBMAP(entry))
92 			continue;
93 		/* assume text is the first entry */
94 		if (*stext == *etext) {
95 			*stext = entry->start;
96 			*etext = entry->end;
97 			break;
98 		}
99 	}
100 #ifdef LINUX_USRSTACK32
101 	if (strcmp(p->p_emul->e_name, "linux32") == 0 &&
102 	    LINUX_USRSTACK32 < USRSTACK32)
103 		*sstack = (unsigned long)LINUX_USRSTACK32;
104 	else
105 #endif
106 #ifdef LINUX_USRSTACK
107 	if (strcmp(p->p_emul->e_name, "linux") == 0 &&
108 	    LINUX_USRSTACK < USRSTACK)
109 		*sstack = (unsigned long)LINUX_USRSTACK;
110 	else
111 #endif
112 #ifdef	USRSTACK32
113 	if (strstr(p->p_emul->e_name, "32") != NULL)
114 		*sstack = (unsigned long)USRSTACK32;
115 	else
116 #endif
117 		*sstack = (unsigned long)USRSTACK;
118 
119 	/*
120 	 * jdk 1.6 compares low <= addr && addr < high
121 	 * if we put addr == high, then the test fails
122 	 * so eat one page.
123 	 */
124 	*sstack -= PAGE_SIZE;
125 
126 	vm_map_unlock_read(map);
127 	uvmspace_free(vm);
128 }
129 
130 /*
131  * Linux compatible /proc/meminfo. Only active when the -o linux
132  * mountflag is used.
133  */
134 int
135 procfs_domeminfo(struct lwp *curl, struct proc *p,
136     struct pfsnode *pfs, struct uio *uio)
137 {
138 	char *bf;
139 	int len;
140 	int error = 0;
141 
142 	bf = malloc(LBFSZ, M_TEMP, M_WAITOK);
143 
144 	len = snprintf(bf, LBFSZ,
145 		"        total:    used:    free:  shared: buffers: cached:\n"
146 		"Mem:  %8lu %8lu %8lu %8lu %8lu %8lu\n"
147 		"Swap: %8lu %8lu %8lu\n"
148 		"MemTotal:  %8lu kB\n"
149 		"MemFree:   %8lu kB\n"
150 		"MemShared: %8lu kB\n"
151 		"Buffers:   %8lu kB\n"
152 		"Cached:    %8lu kB\n"
153 		"SwapTotal: %8lu kB\n"
154 		"SwapFree:  %8lu kB\n",
155 		PGTOB(uvmexp.npages),
156 		PGTOB(uvmexp.npages - uvmexp.free),
157 		PGTOB(uvmexp.free),
158 		0L,
159 		PGTOB(uvmexp.filepages),
160 		PGTOB(uvmexp.anonpages + uvmexp.filepages + uvmexp.execpages),
161 		PGTOB(uvmexp.swpages),
162 		PGTOB(uvmexp.swpginuse),
163 		PGTOB(uvmexp.swpages - uvmexp.swpginuse),
164 		PGTOKB(uvmexp.npages),
165 		PGTOKB(uvmexp.free),
166 		0L,
167 		PGTOKB(uvmexp.filepages),
168 		PGTOKB(uvmexp.anonpages + uvmexp.filepages + uvmexp.execpages),
169 		PGTOKB(uvmexp.swpages),
170 		PGTOKB(uvmexp.swpages - uvmexp.swpginuse));
171 
172 	if (len == 0)
173 		goto out;
174 
175 	error = uiomove_frombuf(bf, len, uio);
176 out:
177 	free(bf, M_TEMP);
178 	return error;
179 }
180 
181 /*
182  * Linux compatible /proc/devices. Only active when the -o linux
183  * mountflag is used.
184  */
185 int
186 procfs_dodevices(struct lwp *curl, struct proc *p,
187     struct pfsnode *pfs, struct uio *uio)
188 {
189 	char *bf;
190 	int offset = 0;
191 	int i, error = ENAMETOOLONG;
192 
193 	/* XXX elad - may need filtering. */
194 
195 	bf = malloc(LBFSZ, M_TEMP, M_WAITOK);
196 
197 	offset += snprintf(&bf[offset], LBFSZ - offset, "Character devices:\n");
198 	if (offset >= LBFSZ)
199 		goto out;
200 
201 	mutex_enter(&specfs_lock);
202 	for (i = 0; i < max_devsw_convs; i++) {
203 		if ((devsw_conv[i].d_name == NULL) ||
204 		    (devsw_conv[i].d_cmajor == -1))
205 			continue;
206 
207 		offset += snprintf(&bf[offset], LBFSZ - offset,
208 		    "%3d %s\n", devsw_conv[i].d_cmajor, devsw_conv[i].d_name);
209 		if (offset >= LBFSZ) {
210 			mutex_exit(&specfs_lock);
211 			goto out;
212 		}
213 	}
214 
215 	offset += snprintf(&bf[offset], LBFSZ - offset, "\nBlock devices:\n");
216 	if (offset >= LBFSZ) {
217 		mutex_exit(&specfs_lock);
218 		goto out;
219 	}
220 
221 	for (i = 0; i < max_devsw_convs; i++) {
222 		if ((devsw_conv[i].d_name == NULL) ||
223 		    (devsw_conv[i].d_bmajor == -1))
224 			continue;
225 
226 		offset += snprintf(&bf[offset], LBFSZ - offset,
227 		    "%3d %s\n", devsw_conv[i].d_bmajor, devsw_conv[i].d_name);
228 		if (offset >= LBFSZ) {
229 			mutex_exit(&specfs_lock);
230 			goto out;
231 		}
232 	}
233 	mutex_exit(&specfs_lock);
234 
235 	error = uiomove_frombuf(bf, offset, uio);
236 out:
237 	free(bf, M_TEMP);
238 	return error;
239 }
240 
241 /*
242  * Linux compatible /proc/stat. Only active when the -o linux
243  * mountflag is used.
244  */
245 int
246 procfs_docpustat(struct lwp *curl, struct proc *p,
247     struct pfsnode *pfs, struct uio *uio)
248 {
249 	char		*bf;
250 	int	 	 error;
251 	int	 	 len;
252 #if defined(MULTIPROCESSOR)
253         struct cpu_info *ci;
254         CPU_INFO_ITERATOR cii;
255 #endif
256 	int	 	 i;
257 
258 	error = ENAMETOOLONG;
259 	bf = malloc(LBFSZ, M_TEMP, M_WAITOK);
260 
261 	len = snprintf(bf, LBFSZ,
262 		"cpu %" PRIu64 " %" PRIu64 " %" PRIu64 " %" PRIu64 "\n",
263 		curcpu()->ci_schedstate.spc_cp_time[CP_USER],
264 		curcpu()->ci_schedstate.spc_cp_time[CP_NICE],
265 		curcpu()->ci_schedstate.spc_cp_time[CP_SYS] /*+ [CP_INTR]*/,
266 		curcpu()->ci_schedstate.spc_cp_time[CP_IDLE]);
267 	if (len == 0)
268 		goto out;
269 
270 #if defined(MULTIPROCESSOR)
271 #define ALLCPUS	CPU_INFO_FOREACH(cii, ci)
272 #define CPUNAME	ci
273 #else
274 #define ALLCPUS	; i < 1 ;
275 #define CPUNAME	curcpu()
276 #endif
277 
278 	i = 0;
279 	for (ALLCPUS) {
280 		len += snprintf(&bf[len], LBFSZ - len,
281 			"cpu%d %" PRIu64 " %" PRIu64 " %" PRIu64 " %" PRIu64
282 			"\n", i,
283 			CPUNAME->ci_schedstate.spc_cp_time[CP_USER],
284 			CPUNAME->ci_schedstate.spc_cp_time[CP_NICE],
285 			CPUNAME->ci_schedstate.spc_cp_time[CP_SYS],
286 			CPUNAME->ci_schedstate.spc_cp_time[CP_IDLE]);
287 		if (len >= LBFSZ)
288 			goto out;
289 		i += 1;
290 	}
291 
292 	len += snprintf(&bf[len], LBFSZ - len,
293 			"disk 0 0 0 0\n"
294 			"page %u %u\n"
295 			"swap %u %u\n"
296 			"intr %u\n"
297 			"ctxt %u\n"
298 			"btime %lld\n",
299 			uvmexp.pageins, uvmexp.pdpageouts,
300 			uvmexp.pgswapin, uvmexp.pgswapout,
301 			uvmexp.intrs,
302 			uvmexp.swtch,
303 			(long long)boottime.tv_sec);
304 	if (len >= LBFSZ)
305 		goto out;
306 
307 	error = uiomove_frombuf(bf, len, uio);
308 out:
309 	free(bf, M_TEMP);
310 	return error;
311 }
312 
313 /*
314  * Linux compatible /proc/loadavg. Only active when the -o linux
315  * mountflag is used.
316  */
317 int
318 procfs_doloadavg(struct lwp *curl, struct proc *p,
319     struct pfsnode *pfs, struct uio *uio)
320 {
321 	char	*bf;
322 	int 	 error;
323 	int 	 len;
324 
325 	error = ENAMETOOLONG;
326 	bf = malloc(LBFSZ, M_TEMP, M_WAITOK);
327 
328 	averunnable.fscale = FSCALE;
329 	len = snprintf(bf, LBFSZ,
330 	        "%d.%02d %d.%02d %d.%02d %d/%d %d\n",
331 		(int)(averunnable.ldavg[0] / averunnable.fscale),
332 		(int)(averunnable.ldavg[0] * 100 / averunnable.fscale % 100),
333 		(int)(averunnable.ldavg[1] / averunnable.fscale),
334 		(int)(averunnable.ldavg[1] * 100 / averunnable.fscale % 100),
335 		(int)(averunnable.ldavg[2] / averunnable.fscale),
336 		(int)(averunnable.ldavg[2] * 100 / averunnable.fscale % 100),
337 		1,		/* number of ONPROC processes */
338 		nprocs,
339 		30000);		/* last pid */
340 	if (len == 0)
341 		goto out;
342 
343 	error = uiomove_frombuf(bf, len, uio);
344 out:
345 	free(bf, M_TEMP);
346 	return error;
347 }
348 
349 /*
350  * Linux compatible /proc/<pid>/statm. Only active when the -o linux
351  * mountflag is used.
352  */
353 int
354 procfs_do_pid_statm(struct lwp *curl, struct lwp *l,
355     struct pfsnode *pfs, struct uio *uio)
356 {
357 	struct vmspace	*vm;
358 	struct proc	*p = l->l_proc;
359 	struct rusage	*ru = &p->p_stats->p_ru;
360 	char		*bf;
361 	int	 	 error;
362 	int	 	 len;
363 
364 	error = ENAMETOOLONG;
365 	bf = malloc(LBFSZ, M_TEMP, M_WAITOK);
366 
367 	/* XXX - we use values from vmspace, since dsl says that ru figures
368 	   are always 0 except for zombies. See kvm_proc.c::kvm_getproc2() */
369 	if ((error = proc_vmspace_getref(p, &vm)) != 0) {
370 		goto out;
371 	}
372 
373 	len = snprintf(bf, LBFSZ,
374 	        "%lu %lu %lu %lu %lu %lu %lu\n",
375 		(unsigned long)(vm->vm_tsize + vm->vm_dsize + vm->vm_ssize), /* size */
376 		(unsigned long)(vm->vm_rssize),	/* resident */
377 		(unsigned long)(ru->ru_ixrss),	/* shared */
378 		(unsigned long)(vm->vm_tsize),	/* text size in pages */
379 		(unsigned long)(vm->vm_dsize),	/* data size in pages */
380 		(unsigned long)(vm->vm_ssize),	/* stack size in pages */
381 		(unsigned long) 0);
382 
383 	if (len == 0)
384 		goto out;
385 
386 	error = uiomove_frombuf(bf, len, uio);
387 out:
388 	free(bf, M_TEMP);
389 	return error;
390 }
391 
392 #define USEC_2_TICKS(x)		((x) / 10000)
393 
394 /*
395  * Linux compatible /proc/<pid>/stat. Only active when the -o linux
396  * mountflag is used.
397  */
398 int
399 procfs_do_pid_stat(struct lwp *curl, struct lwp *l,
400     struct pfsnode *pfs, struct uio *uio)
401 {
402 	char *bf;
403 	struct proc *p = l->l_proc;
404 	int len;
405 	struct tty *tty = p->p_session->s_ttyp;
406 	struct rusage *ru = &p->p_stats->p_ru;
407 	struct rusage *cru = &p->p_stats->p_cru;
408 	unsigned long stext = 0, etext = 0, sstack = 0;
409 	struct timeval rt;
410 	struct vmspace	*vm;
411 	int error = 0;
412 
413 	bf = malloc(LBFSZ, M_TEMP, M_WAITOK);
414 
415 	if ((error = proc_vmspace_getref(p, &vm)) != 0) {
416 		goto out;
417 	}
418 
419 	get_proc_size_info(l, &stext, &etext, &sstack);
420 
421 	mutex_enter(proc_lock);
422 	mutex_enter(p->p_lock);
423 
424 	calcru(p, NULL, NULL, NULL, &rt);
425 
426 	len = snprintf(bf, LBFSZ,
427 	    "%d (%s) %c %d %d %d %d %d "
428 	    "%u "
429 	    "%lu %lu %lu %lu %lu %lu %lu %lu "
430 	    "%d %d %d "
431 	    "%lu %lu %lu %lu %" PRIu64 " "
432 	    "%lu %lu %lu "
433 	    "%u %u "
434 	    "%u %u %u %u "
435 	    "%lu %lu %lu %d %d\n",
436 
437 	    p->p_pid,
438 	    p->p_comm,
439 	    "0IR3SZD"[(p->p_stat > 6) ? 0 : (int)p->p_stat],
440 	    (p->p_pptr != NULL) ? p->p_pptr->p_pid : 0,
441 
442 	    p->p_pgid,
443 	    p->p_session->s_sid,
444 	    tty ? tty->t_dev : 0,
445 	    (tty && tty->t_pgrp) ? tty->t_pgrp->pg_id : 0,
446 
447 	    p->p_flag,
448 
449 	    ru->ru_minflt,
450 	    cru->ru_minflt,
451 	    ru->ru_majflt,
452 	    cru->ru_majflt,
453 	    USEC_2_TICKS(ru->ru_utime.tv_usec),
454 	    USEC_2_TICKS(ru->ru_stime.tv_usec),
455 	    USEC_2_TICKS(cru->ru_utime.tv_usec),
456 	    USEC_2_TICKS(cru->ru_stime.tv_usec),
457 
458 	    l->l_priority,				/* XXX: priority */
459 	    p->p_nice - 20,
460 	    0,
461 
462 	    rt.tv_sec,
463 	    p->p_stats->p_start.tv_sec,
464 	    (unsigned long)(vm->vm_tsize + vm->vm_dsize + vm->vm_ssize), /* size */
465 	    (unsigned long)(vm->vm_rssize),	/* resident */
466 	    p->p_rlimit[RLIMIT_RSS].rlim_cur,
467 
468 	    stext,					/* start code */
469 	    etext,					/* end code */
470 	    sstack,					/* mm start stack */
471 	    0,						/* XXX: pc */
472 	    0,						/* XXX: sp */
473 	    p->p_sigpend.sp_set.__bits[0],		/* XXX: pending */
474 	    0,						/* XXX: held */
475 	    p->p_sigctx.ps_sigignore.__bits[0],		/* ignored */
476 	    p->p_sigctx.ps_sigcatch.__bits[0],		/* caught */
477 
478 	    (unsigned long)(intptr_t)l->l_wchan,
479 	    ru->ru_nvcsw,
480 	    ru->ru_nivcsw,
481 	    p->p_exitsig,
482 	    0);						/* XXX: processor */
483 
484 	mutex_exit(p->p_lock);
485 	mutex_exit(proc_lock);
486 
487 	if (len == 0)
488 		goto out;
489 
490 	error = uiomove_frombuf(bf, len, uio);
491 out:
492 	free(bf, M_TEMP);
493 	return error;
494 }
495 
496 int
497 procfs_docpuinfo(struct lwp *curl, struct proc *p,
498     struct pfsnode *pfs, struct uio *uio)
499 {
500 	int len = LBFSZ;
501 	char *bf = malloc(len, M_TEMP, M_WAITOK);
502 	int error;
503 
504 	if (procfs_getcpuinfstr(bf, &len) < 0) {
505 		error = ENOSPC;
506 		goto done;
507 	}
508 
509 	if (len == 0) {
510 		error = 0;
511 		goto done;
512 	}
513 
514 	error = uiomove_frombuf(bf, len, uio);
515 done:
516 	free(bf, M_TEMP);
517 	return error;
518 }
519 
520 int
521 procfs_douptime(struct lwp *curl, struct proc *p,
522     struct pfsnode *pfs, struct uio *uio)
523 {
524 	char *bf;
525 	int len;
526 	struct timeval runtime;
527 	u_int64_t idle;
528 	int error = 0;
529 
530 	bf = malloc(LBFSZ, M_TEMP, M_WAITOK);
531 
532 	microuptime(&runtime);
533 	idle = curcpu()->ci_schedstate.spc_cp_time[CP_IDLE];
534 	len = snprintf(bf, LBFSZ,
535 	    "%lu.%02lu %" PRIu64 ".%02" PRIu64 "\n",
536 	    runtime.tv_sec, runtime.tv_usec / 10000,
537 	    idle / hz, (((idle % hz) * 100) / hz) % 100);
538 
539 	if (len == 0)
540 		goto out;
541 
542 	error = uiomove_frombuf(bf, len, uio);
543 out:
544 	free(bf, M_TEMP);
545 	return error;
546 }
547 
548 int
549 procfs_domounts(struct lwp *curl, struct proc *p,
550     struct pfsnode *pfs, struct uio *uio)
551 {
552 	char *bf, *mtab = NULL;
553 	const char *fsname;
554 	size_t len, mtabsz = 0;
555 	struct mount *mp, *nmp;
556 	struct statvfs *sfs;
557 	int error = 0;
558 
559 	bf = malloc(LBFSZ, M_TEMP, M_WAITOK);
560 	mutex_enter(&mountlist_lock);
561 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
562 	     mp = nmp) {
563 		if (vfs_busy(mp, &nmp)) {
564 			continue;
565 		}
566 
567 		sfs = &mp->mnt_stat;
568 
569 		/* Linux uses different names for some filesystems */
570 		fsname = sfs->f_fstypename;
571 		if (strcmp(fsname, "procfs") == 0)
572 			fsname = "proc";
573 		else if (strcmp(fsname, "ext2fs") == 0)
574 			fsname = "ext2";
575 
576 		len = snprintf(bf, LBFSZ, "%s %s %s %s%s%s%s%s%s 0 0\n",
577 			sfs->f_mntfromname,
578 			sfs->f_mntonname,
579 			fsname,
580 			(mp->mnt_flag & MNT_RDONLY) ? "ro" : "rw",
581 			(mp->mnt_flag & MNT_NOSUID) ? ",nosuid" : "",
582 			(mp->mnt_flag & MNT_NOEXEC) ? ",noexec" : "",
583 			(mp->mnt_flag & MNT_NODEV) ? ",nodev" : "",
584 			(mp->mnt_flag & MNT_SYNCHRONOUS) ? ",sync" : "",
585 			(mp->mnt_flag & MNT_NOATIME) ? ",noatime" : ""
586 			);
587 
588 		mtab = realloc(mtab, mtabsz + len, M_TEMP, M_WAITOK);
589 		memcpy(mtab + mtabsz, bf, len);
590 		mtabsz += len;
591 
592 		vfs_unbusy(mp, false, &nmp);
593 	}
594 	mutex_exit(&mountlist_lock);
595 	free(bf, M_TEMP);
596 
597 	if (mtabsz > 0) {
598 		error = uiomove_frombuf(mtab, mtabsz, uio);
599 		free(mtab, M_TEMP);
600 	}
601 
602 	return error;
603 }
604