xref: /netbsd-src/sys/miscfs/procfs/procfs_vnops.c (revision c42dbd0ed2e61fe6eda8590caa852ccf34719964)
1 /*	$NetBSD: procfs_vnops.c,v 1.232 2024/05/12 17:26:51 christos Exp $	*/
2 
3 /*-
4  * Copyright (c) 2006, 2007, 2008, 2020 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Andrew Doran.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1993, 1995
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * This code is derived from software contributed to Berkeley by
37  * Jan-Simon Pendry.
38  *
39  * Redistribution and use in source and binary forms, with or without
40  * modification, are permitted provided that the following conditions
41  * are met:
42  * 1. Redistributions of source code must retain the above copyright
43  *    notice, this list of conditions and the following disclaimer.
44  * 2. Redistributions in binary form must reproduce the above copyright
45  *    notice, this list of conditions and the following disclaimer in the
46  *    documentation and/or other materials provided with the distribution.
47  * 3. Neither the name of the University nor the names of its contributors
48  *    may be used to endorse or promote products derived from this software
49  *    without specific prior written permission.
50  *
51  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
52  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
53  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
54  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
55  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
56  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
57  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
58  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
59  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
60  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
61  * SUCH DAMAGE.
62  *
63  *	@(#)procfs_vnops.c	8.18 (Berkeley) 5/21/95
64  */
65 
66 /*
67  * Copyright (c) 1993 Jan-Simon Pendry
68  *
69  * This code is derived from software contributed to Berkeley by
70  * Jan-Simon Pendry.
71  *
72  * Redistribution and use in source and binary forms, with or without
73  * modification, are permitted provided that the following conditions
74  * are met:
75  * 1. Redistributions of source code must retain the above copyright
76  *    notice, this list of conditions and the following disclaimer.
77  * 2. Redistributions in binary form must reproduce the above copyright
78  *    notice, this list of conditions and the following disclaimer in the
79  *    documentation and/or other materials provided with the distribution.
80  * 3. All advertising materials mentioning features or use of this software
81  *    must display the following acknowledgement:
82  *	This product includes software developed by the University of
83  *	California, Berkeley and its contributors.
84  * 4. Neither the name of the University nor the names of its contributors
85  *    may be used to endorse or promote products derived from this software
86  *    without specific prior written permission.
87  *
88  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
89  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
90  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
91  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
92  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
93  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
94  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
95  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
96  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
97  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
98  * SUCH DAMAGE.
99  *
100  *	@(#)procfs_vnops.c	8.18 (Berkeley) 5/21/95
101  */
102 
103 /*
104  * procfs vnode interface
105  */
106 
107 #include <sys/cdefs.h>
108 __KERNEL_RCSID(0, "$NetBSD: procfs_vnops.c,v 1.232 2024/05/12 17:26:51 christos Exp $");
109 
110 #include <sys/param.h>
111 #include <sys/atomic.h>
112 #include <sys/systm.h>
113 #include <sys/time.h>
114 #include <sys/kernel.h>
115 #include <sys/file.h>
116 #include <sys/filedesc.h>
117 #include <sys/proc.h>
118 #include <sys/vnode.h>
119 #include <sys/namei.h>
120 #include <sys/malloc.h>
121 #include <sys/mount.h>
122 #include <sys/dirent.h>
123 #include <sys/resourcevar.h>
124 #include <sys/stat.h>
125 #include <sys/ptrace.h>
126 #include <sys/kauth.h>
127 #include <sys/exec.h>
128 
129 #include <uvm/uvm_extern.h>	/* for PAGE_SIZE */
130 
131 #include <machine/reg.h>
132 
133 #include <miscfs/genfs/genfs.h>
134 #include <miscfs/procfs/procfs.h>
135 
136 /*
137  * Vnode Operations.
138  *
139  */
140 
141 static int procfs_validfile_linux(struct lwp *, struct mount *);
142 static int procfs_root_readdir_callback(struct proc *, void *);
143 static void procfs_dir(pfstype, struct lwp *, struct proc *, char **, char *,
144     size_t);
145 
146 /*
147  * This is a list of the valid names in the
148  * process-specific sub-directories.  It is
149  * used in procfs_lookup and procfs_readdir
150  */
151 static const struct proc_target {
152 	u_char	pt_type;
153 	u_char	pt_namlen;
154 	const char	*pt_name;
155 	pfstype	pt_pfstype;
156 	int	(*pt_valid)(struct lwp *, struct mount *);
157 } proc_targets[] = {
158 #define N(s) sizeof(s)-1, s
159 	/*	  name		type		validp */
160 	{ DT_DIR, N("."),	PFSproc,	NULL },
161 	{ DT_DIR, N(".."),	PFSroot,	NULL },
162 	{ DT_DIR, N("fd"),	PFSfd,		NULL },
163 	{ DT_DIR, N("task"),	PFStask,	procfs_validfile_linux },
164 	{ DT_LNK, N("cwd"),	PFScwd,		NULL },
165 	{ DT_REG, N("emul"),	PFSemul,	NULL },
166 	{ DT_LNK, N("root"),	PFSchroot,	NULL },
167 	{ DT_REG, N("auxv"),	PFSauxv,	procfs_validauxv },
168 	{ DT_REG, N("cmdline"), PFScmdline,	NULL },
169 	{ DT_REG, N("environ"), PFSenviron,	NULL },
170 	{ DT_LNK, N("exe"),	PFSexe,		procfs_validfile },
171 	{ DT_REG, N("file"),	PFSfile,	procfs_validfile },
172 	{ DT_REG, N("fpregs"),	PFSfpregs,	procfs_validfpregs },
173 	{ DT_REG, N("limit"),	PFSlimit,	NULL },
174 	{ DT_REG, N("limits"),	PFSlimits,	procfs_validfile_linux },
175 	{ DT_REG, N("map"),	PFSmap,		procfs_validmap },
176 	{ DT_REG, N("maps"),	PFSmaps,	procfs_validmap },
177 	{ DT_REG, N("mem"),	PFSmem,		NULL },
178 	{ DT_REG, N("note"),	PFSnote,	NULL },
179 	{ DT_REG, N("notepg"),	PFSnotepg,	NULL },
180 	{ DT_REG, N("regs"),	PFSregs,	procfs_validregs },
181 	{ DT_REG, N("stat"),	PFSstat,	procfs_validfile_linux },
182 	{ DT_REG, N("statm"),	PFSstatm,	procfs_validfile_linux },
183 	{ DT_REG, N("status"),	PFSstatus,	NULL },
184 #ifdef __HAVE_PROCFS_MACHDEP
185 	PROCFS_MACHDEP_NODETYPE_DEFNS
186 #endif
187 #undef N
188 };
189 static const int nproc_targets = sizeof(proc_targets) / sizeof(proc_targets[0]);
190 
191 /*
192  * List of files in the root directory. Note: the validate function will
193  * be called with p == NULL for these ones.
194  */
195 static const struct proc_target proc_root_targets[] = {
196 #define N(s) sizeof(s)-1, s
197 	/*	  name		    type	    validp */
198 	{ DT_REG, N("meminfo"),     PFSmeminfo,        procfs_validfile_linux },
199 	{ DT_REG, N("cpuinfo"),     PFScpuinfo,        procfs_validfile_linux },
200 	{ DT_REG, N("uptime"),      PFSuptime,         procfs_validfile_linux },
201 	{ DT_REG, N("mounts"),	    PFSmounts,	       procfs_validfile_linux },
202 	{ DT_REG, N("devices"),     PFSdevices,        procfs_validfile_linux },
203 	{ DT_REG, N("stat"),	    PFScpustat,        procfs_validfile_linux },
204 	{ DT_REG, N("loadavg"),	    PFSloadavg,        procfs_validfile_linux },
205 	{ DT_REG, N("version"),     PFSversion,        procfs_validfile_linux },
206 	{ DT_DIR, N("sysvipc"),     PFSsysvipc,        procfs_validfile_linux },
207 #undef N
208 };
209 static const int nproc_root_targets =
210     sizeof(proc_root_targets) / sizeof(proc_root_targets[0]);
211 
212 /*
213  * List of files in the sysvipc directory
214  */
215 static const struct proc_target proc_sysvipc_targets[] = {
216 #define N(s) sizeof(s)-1, s
217         /*        name              type            validp */
218 	{ DT_DIR, N("."),	PFSsysvipc,	NULL },
219 	{ DT_DIR, N(".."),	PFSroot,	NULL },
220 	{ DT_REG, N("msg"),	PFSsysvipc_msg, procfs_validfile_linux },
221 	{ DT_REG, N("sem"),	PFSsysvipc_sem, procfs_validfile_linux },
222 	{ DT_REG, N("shm"),	PFSsysvipc_shm, procfs_validfile_linux },
223 #undef N
224 };
225 static const int nproc_sysvipc_targets =
226     sizeof(proc_sysvipc_targets) / sizeof(proc_sysvipc_targets[0]);
227 
228 int	procfs_lookup(void *);
229 int	procfs_open(void *);
230 int	procfs_close(void *);
231 int	procfs_access(void *);
232 int	procfs_getattr(void *);
233 int	procfs_setattr(void *);
234 int	procfs_readdir(void *);
235 int	procfs_readlink(void *);
236 int	procfs_inactive(void *);
237 int	procfs_reclaim(void *);
238 int	procfs_print(void *);
239 int	procfs_pathconf(void *);
240 int	procfs_getpages(void *);
241 
242 static uint8_t fttodt(file_t *);
243 static int atoi(const char *, size_t);
244 
245 /*
246  * procfs vnode operations.
247  */
248 int (**procfs_vnodeop_p)(void *);
249 const struct vnodeopv_entry_desc procfs_vnodeop_entries[] = {
250 	{ &vop_default_desc, vn_default_error },
251 	{ &vop_parsepath_desc, genfs_parsepath },	/* parsepath */
252 	{ &vop_lookup_desc, procfs_lookup },		/* lookup */
253 	{ &vop_create_desc, genfs_eopnotsupp },		/* create */
254 	{ &vop_mknod_desc, genfs_eopnotsupp },		/* mknod */
255 	{ &vop_open_desc, procfs_open },		/* open */
256 	{ &vop_close_desc, procfs_close },		/* close */
257 	{ &vop_access_desc, procfs_access },		/* access */
258 	{ &vop_accessx_desc, genfs_accessx },		/* accessx */
259 	{ &vop_getattr_desc, procfs_getattr },		/* getattr */
260 	{ &vop_setattr_desc, procfs_setattr },		/* setattr */
261 	{ &vop_read_desc, procfs_rw },			/* read */
262 	{ &vop_write_desc, procfs_rw },			/* write */
263 	{ &vop_fallocate_desc, genfs_eopnotsupp },	/* fallocate */
264 	{ &vop_fdiscard_desc, genfs_eopnotsupp },	/* fdiscard */
265 	{ &vop_fcntl_desc, genfs_fcntl },		/* fcntl */
266 	{ &vop_ioctl_desc, genfs_enoioctl },		/* ioctl */
267 	{ &vop_poll_desc, genfs_poll },			/* poll */
268 	{ &vop_kqfilter_desc, genfs_kqfilter },		/* kqfilter */
269 	{ &vop_revoke_desc, genfs_revoke },		/* revoke */
270 	{ &vop_fsync_desc, genfs_nullop },		/* fsync */
271 	{ &vop_seek_desc, genfs_nullop },		/* seek */
272 	{ &vop_remove_desc, genfs_eopnotsupp },		/* remove */
273 	{ &vop_link_desc, genfs_erofs_link },		/* link */
274 	{ &vop_rename_desc, genfs_eopnotsupp },		/* rename */
275 	{ &vop_mkdir_desc, genfs_eopnotsupp },		/* mkdir */
276 	{ &vop_rmdir_desc, genfs_eopnotsupp },		/* rmdir */
277 	{ &vop_symlink_desc, genfs_erofs_symlink },	/* symlink */
278 	{ &vop_readdir_desc, procfs_readdir },		/* readdir */
279 	{ &vop_readlink_desc, procfs_readlink },	/* readlink */
280 	{ &vop_abortop_desc, genfs_abortop },		/* abortop */
281 	{ &vop_inactive_desc, procfs_inactive },	/* inactive */
282 	{ &vop_reclaim_desc, procfs_reclaim },		/* reclaim */
283 	{ &vop_lock_desc, genfs_lock },			/* lock */
284 	{ &vop_unlock_desc, genfs_unlock },		/* unlock */
285 	{ &vop_bmap_desc, genfs_eopnotsupp },		/* bmap */
286 	{ &vop_strategy_desc, genfs_badop },		/* strategy */
287 	{ &vop_print_desc, procfs_print },		/* print */
288 	{ &vop_islocked_desc, genfs_islocked },		/* islocked */
289 	{ &vop_pathconf_desc, procfs_pathconf },	/* pathconf */
290 	{ &vop_advlock_desc, genfs_einval },		/* advlock */
291 	{ &vop_getpages_desc, procfs_getpages },	/* getpages */
292 	{ &vop_putpages_desc, genfs_null_putpages },	/* putpages */
293 	{ NULL, NULL }
294 };
295 const struct vnodeopv_desc procfs_vnodeop_opv_desc =
296 	{ &procfs_vnodeop_p, procfs_vnodeop_entries };
297 /*
298  * set things up for doing i/o on
299  * the pfsnode (vp).  (vp) is locked
300  * on entry, and should be left locked
301  * on exit.
302  *
303  * for procfs we don't need to do anything
304  * in particular for i/o.  all that is done
305  * is to support exclusive open on process
306  * memory images.
307  */
308 int
309 procfs_open(void *v)
310 {
311 	struct vop_open_args /* {
312 		struct vnode *a_vp;
313 		int  a_mode;
314 		kauth_cred_t a_cred;
315 	} */ *ap = v;
316 	struct vnode *vp = ap->a_vp;
317 	struct pfsnode *pfs = VTOPFS(vp);
318 	struct lwp *l1;
319 	struct proc *p2;
320 	int error;
321 
322 	if ((error =
323 	     procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &p2, ENOENT)) != 0)
324 		return error;
325 
326 	l1 = curlwp;				/* tracer */
327 
328 #define	M2K(m)	(((m) & FREAD) && ((m) & FWRITE) ? \
329 		 KAUTH_REQ_PROCESS_PROCFS_RW : \
330 		 (m) & FWRITE ? KAUTH_REQ_PROCESS_PROCFS_WRITE : \
331 		 KAUTH_REQ_PROCESS_PROCFS_READ)
332 
333 	mutex_enter(p2->p_lock);
334 	error = kauth_authorize_process(l1->l_cred, KAUTH_PROCESS_PROCFS,
335 	    p2, pfs, KAUTH_ARG(M2K(ap->a_mode)), NULL);
336 	mutex_exit(p2->p_lock);
337 	if (error) {
338 		procfs_proc_unlock(p2);
339 		return (error);
340 	}
341 
342 #undef M2K
343 
344 	switch (pfs->pfs_type) {
345 	case PFSmem:
346 		if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) ||
347 		    ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE))) {
348 			error = EBUSY;
349 			break;
350 		}
351 
352 		if (!proc_isunder(p2, l1)) {
353 			error = EPERM;
354 			break;
355 		}
356 
357 		if (ap->a_mode & FWRITE)
358 			pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL);
359 
360 		break;
361 
362 	case PFSregs:
363 	case PFSfpregs:
364 		if (!proc_isunder(p2, l1)) {
365 			error = EPERM;
366 			break;
367 		}
368 		break;
369 
370 	default:
371 		break;
372 	}
373 
374 	procfs_proc_unlock(p2);
375 	return (error);
376 }
377 
378 /*
379  * close the pfsnode (vp) after doing i/o.
380  * (vp) is not locked on entry or exit.
381  *
382  * nothing to do for procfs other than undo
383  * any exclusive open flag (see _open above).
384  */
385 int
386 procfs_close(void *v)
387 {
388 	struct vop_close_args /* {
389 		struct vnode *a_vp;
390 		int  a_fflag;
391 		kauth_cred_t a_cred;
392 	} */ *ap = v;
393 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
394 
395 	switch (pfs->pfs_type) {
396 	case PFSmem:
397 		if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL))
398 			pfs->pfs_flags &= ~(FWRITE|O_EXCL);
399 		break;
400 
401 	default:
402 		break;
403 	}
404 
405 	return (0);
406 }
407 
408 /*
409  * _inactive is called when the pfsnode
410  * is vrele'd and the reference count goes
411  * to zero.  (vp) will be on the vnode free
412  * list, so to get it back vget() must be
413  * used.
414  *
415  * (vp) is locked on entry, but must be unlocked on exit.
416  */
417 int
418 procfs_inactive(void *v)
419 {
420 	struct vop_inactive_v2_args /* {
421 		struct vnode *a_vp;
422 		bool *a_recycle;
423 	} */ *ap = v;
424 	struct vnode *vp = ap->a_vp;
425 	struct pfsnode *pfs = VTOPFS(vp);
426 
427 	mutex_enter(&proc_lock);
428 	*ap->a_recycle = (procfs_proc_find(vp->v_mount, pfs->pfs_pid) == NULL);
429 	mutex_exit(&proc_lock);
430 
431 	return (0);
432 }
433 
434 /*
435  * _reclaim is called when getnewvnode()
436  * wants to make use of an entry on the vnode
437  * free list.  at this time the filesystem needs
438  * to free any private data and remove the node
439  * from any private lists.
440  */
441 int
442 procfs_reclaim(void *v)
443 {
444 	struct vop_reclaim_v2_args /* {
445 		struct vnode *a_vp;
446 	} */ *ap = v;
447 	struct vnode *vp = ap->a_vp;
448 	struct pfsnode *pfs = VTOPFS(vp);
449 
450 	VOP_UNLOCK(vp);
451 
452 	/*
453 	 * To interlock with procfs_revoke_vnodes().
454 	 */
455 	mutex_enter(vp->v_interlock);
456 	vp->v_data = NULL;
457 	mutex_exit(vp->v_interlock);
458 	procfs_hashrem(pfs);
459 	kmem_free(pfs, sizeof(*pfs));
460 	return 0;
461 }
462 
463 /*
464  * Return POSIX pathconf information applicable to special devices.
465  */
466 int
467 procfs_pathconf(void *v)
468 {
469 	struct vop_pathconf_args /* {
470 		struct vnode *a_vp;
471 		int a_name;
472 		register_t *a_retval;
473 	} */ *ap = v;
474 
475 	switch (ap->a_name) {
476 	case _PC_LINK_MAX:
477 		*ap->a_retval = LINK_MAX;
478 		return (0);
479 	case _PC_MAX_CANON:
480 		*ap->a_retval = MAX_CANON;
481 		return (0);
482 	case _PC_MAX_INPUT:
483 		*ap->a_retval = MAX_INPUT;
484 		return (0);
485 	case _PC_PIPE_BUF:
486 		*ap->a_retval = PIPE_BUF;
487 		return (0);
488 	case _PC_CHOWN_RESTRICTED:
489 		*ap->a_retval = 1;
490 		return (0);
491 	case _PC_VDISABLE:
492 		*ap->a_retval = _POSIX_VDISABLE;
493 		return (0);
494 	case _PC_SYNC_IO:
495 		*ap->a_retval = 1;
496 		return (0);
497 	default:
498 		return genfs_pathconf(ap);
499 	}
500 	/* NOTREACHED */
501 }
502 
503 /*
504  * _print is used for debugging.
505  * just print a readable description
506  * of (vp).
507  */
508 int
509 procfs_print(void *v)
510 {
511 	struct vop_print_args /* {
512 		struct vnode *a_vp;
513 	} */ *ap = v;
514 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
515 
516 	printf("tag VT_PROCFS, type %d, pid %d, mode %x, flags %lx\n",
517 	    pfs->pfs_type, pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags);
518 	return 0;
519 }
520 
521 /*
522  * Works out the path to the target process's current
523  * working directory or chroot.  If the caller is in a chroot and
524  * can't "reach" the target's cwd or root (or some other error
525  * occurs), a "/" is returned for the path.
526  */
527 static void
528 procfs_dir(pfstype t, struct lwp *caller, struct proc *target, char **bpp,
529     char *path, size_t len)
530 {
531 	struct cwdinfo *cwdi;
532 	struct vnode *vp, *rvp;
533 	char *bp;
534 
535 	/*
536 	 * Lock target cwdi and take a reference to the vnode
537 	 * we are interested in to prevent it from disappearing
538 	 * before getcwd_common() below.
539 	 */
540 	rw_enter(&target->p_cwdi->cwdi_lock, RW_READER);
541 	switch (t) {
542 	case PFScwd:
543 		vp = target->p_cwdi->cwdi_cdir;
544 		break;
545 	case PFSchroot:
546 		vp = target->p_cwdi->cwdi_rdir;
547 		break;
548 	default:
549 		rw_exit(&target->p_cwdi->cwdi_lock);
550 		return;
551 	}
552 	if (vp != NULL)
553 		vref(vp);
554 	rw_exit(&target->p_cwdi->cwdi_lock);
555 
556 	cwdi = caller->l_proc->p_cwdi;
557 	rw_enter(&cwdi->cwdi_lock, RW_READER);
558 
559 	rvp = cwdi->cwdi_rdir;
560 	bp = bpp ? *bpp : NULL;
561 
562 	/*
563 	 * XXX: this horrible kludge avoids locking panics when
564 	 * attempting to lookup links that point to within procfs
565 	 */
566 	if (vp != NULL && vp->v_tag == VT_PROCFS) {
567 		if (bpp) {
568 			*--bp = '/';
569 			*bpp = bp;
570 		}
571 		vrele(vp);
572 		rw_exit(&cwdi->cwdi_lock);
573 		return;
574 	}
575 
576 	if (rvp == NULL)
577 		rvp = rootvnode;
578 	if (vp == NULL || getcwd_common(vp, rvp, bp ? &bp : NULL, path,
579 	    len / 2, 0, caller) != 0) {
580 		if (bpp) {
581 			bp = *bpp;
582 			*--bp = '/';
583 		}
584 	}
585 
586 	if (bpp)
587 		*bpp = bp;
588 
589 	if (vp != NULL)
590 		vrele(vp);
591 	rw_exit(&cwdi->cwdi_lock);
592 }
593 
594 /*
595  * Invent attributes for pfsnode (vp) and store
596  * them in (vap).
597  * Directories lengths are returned as zero since
598  * any real length would require the genuine size
599  * to be computed, and nothing cares anyway.
600  *
601  * this is relatively minimal for procfs.
602  */
603 int
604 procfs_getattr(void *v)
605 {
606 	struct vop_getattr_args /* {
607 		struct vnode *a_vp;
608 		struct vattr *a_vap;
609 		kauth_cred_t a_cred;
610 	} */ *ap = v;
611 	struct vnode *vp = ap->a_vp;
612 	struct pfsnode *pfs = VTOPFS(vp);
613 	struct vattr *vap = ap->a_vap;
614 	struct proc *procp;
615 	char *path, *bp, bf[16];
616 	int error;
617 
618 	/* first check the process still exists */
619 	switch (pfs->pfs_type) {
620 	case PFSroot:
621 	case PFScurproc:
622 	case PFSself:
623 		procp = NULL;
624 		break;
625 
626 	default:
627 		error =
628 		    procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &procp, ENOENT);
629 		if (error != 0)
630 			return (error);
631 		break;
632 	}
633 
634 	switch (pfs->pfs_type) {
635 	case PFStask:
636 		if (pfs->pfs_fd == -1) {
637 			path = NULL;
638 			break;
639 		}
640 		/*FALLTHROUGH*/
641 	case PFScwd:
642 	case PFSchroot:
643 		path = malloc(MAXPATHLEN + 4, M_TEMP, M_WAITOK);
644 		if (path == NULL && procp != NULL) {
645 			procfs_proc_unlock(procp);
646 			return (ENOMEM);
647 		}
648 		break;
649 
650 	default:
651 		path = NULL;
652 		break;
653 	}
654 
655 	if (procp != NULL) {
656 		mutex_enter(procp->p_lock);
657 		error = kauth_authorize_process(kauth_cred_get(),
658 		    KAUTH_PROCESS_CANSEE, procp,
659 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
660 		mutex_exit(procp->p_lock);
661 		if (error != 0) {
662 		    	procfs_proc_unlock(procp);
663 		    	if (path != NULL)
664 		    		free(path, M_TEMP);
665 			return (ENOENT);
666 		}
667 	}
668 
669 	error = 0;
670 
671 	/* start by zeroing out the attributes */
672 	vattr_null(vap);
673 
674 	/* next do all the common fields */
675 	vap->va_type = ap->a_vp->v_type;
676 	vap->va_mode = pfs->pfs_mode;
677 	vap->va_fileid = pfs->pfs_fileno;
678 	vap->va_flags = 0;
679 	vap->va_blocksize = PAGE_SIZE;
680 
681 	/*
682 	 * Make all times be current TOD.
683 	 *
684 	 * It would be possible to get the process start
685 	 * time from the p_stats structure, but there's
686 	 * no "file creation" time stamp anyway, and the
687 	 * p_stats structure is not addressable if u. gets
688 	 * swapped out for that process.
689 	 */
690 	getnanotime(&vap->va_ctime);
691 	vap->va_atime = vap->va_mtime = vap->va_ctime;
692 	if (procp)
693 		TIMEVAL_TO_TIMESPEC(&procp->p_stats->p_start,
694 		    &vap->va_birthtime);
695 	else
696 		getnanotime(&vap->va_birthtime);
697 
698 	switch (pfs->pfs_type) {
699 	case PFSmem:
700 	case PFSregs:
701 	case PFSfpregs:
702 #if defined(__HAVE_PROCFS_MACHDEP) && defined(PROCFS_MACHDEP_PROTECT_CASES)
703 	PROCFS_MACHDEP_PROTECT_CASES
704 #endif
705 		/*
706 		 * If the process has exercised some setuid or setgid
707 		 * privilege, then rip away read/write permission so
708 		 * that only root can gain access.
709 		 */
710 		if (procp->p_flag & PK_SUGID)
711 			vap->va_mode &= ~(S_IRUSR|S_IWUSR);
712 		/* FALLTHROUGH */
713 	case PFSstatus:
714 	case PFSstat:
715 	case PFSnote:
716 	case PFSnotepg:
717 	case PFScmdline:
718 	case PFSenviron:
719 	case PFSemul:
720 	case PFSstatm:
721 
722 	case PFSmap:
723 	case PFSmaps:
724 	case PFSlimit:
725 	case PFSlimits:
726 	case PFSauxv:
727 		vap->va_nlink = 1;
728 		vap->va_uid = kauth_cred_geteuid(procp->p_cred);
729 		vap->va_gid = kauth_cred_getegid(procp->p_cred);
730 		break;
731 	case PFScwd:
732 	case PFSchroot:
733 	case PFSmeminfo:
734 	case PFSdevices:
735 	case PFScpuinfo:
736 	case PFSuptime:
737 	case PFSmounts:
738 	case PFScpustat:
739 	case PFSloadavg:
740 	case PFSversion:
741 	case PFSexe:
742 	case PFSself:
743 	case PFScurproc:
744 	case PFSroot:
745 	case PFSsysvipc_msg:
746 	case PFSsysvipc_sem:
747 	case PFSsysvipc_shm:
748 		vap->va_nlink = 1;
749 		vap->va_uid = vap->va_gid = 0;
750 		break;
751 
752 	case PFSsysvipc:
753 		vap->va_nlink = 5;
754 		vap->va_uid = vap->va_gid = 0;
755 		break;
756 
757 	case PFSproc:
758 	case PFStask:
759 	case PFSfile:
760 	case PFSfd:
761 		break;
762 
763 	default:
764 		panic("%s: %d/1", __func__, pfs->pfs_type);
765 	}
766 
767 	/*
768 	 * now do the object specific fields
769 	 *
770 	 * The size could be set from struct reg, but it's hardly
771 	 * worth the trouble, and it puts some (potentially) machine
772 	 * dependent data into this machine-independent code.  If it
773 	 * becomes important then this function should break out into
774 	 * a per-file stat function in the corresponding .c file.
775 	 */
776 
777 	switch (pfs->pfs_type) {
778 	case PFSroot:
779 		vap->va_bytes = vap->va_size = DEV_BSIZE;
780 		break;
781 
782 	case PFSself:
783 	case PFScurproc:
784 		vap->va_bytes = vap->va_size =
785 		    snprintf(bf, sizeof(bf), "%ld", (long)curproc->p_pid);
786 		break;
787 	case PFStask:
788 		if (pfs->pfs_fd != -1) {
789 			vap->va_nlink = 1;
790 			vap->va_uid = 0;
791 			vap->va_gid = 0;
792 			vap->va_bytes = vap->va_size =
793 			    snprintf(bf, sizeof(bf), "..");
794 			break;
795 		}
796 		/*FALLTHROUGH*/
797 	case PFSfd:
798 		if (pfs->pfs_fd != -1) {
799 			file_t *fp;
800 
801 			fp = fd_getfile2(procp, pfs->pfs_fd);
802 			if (fp == NULL) {
803 				error = EBADF;
804 				break;
805 			}
806 			vap->va_nlink = 1;
807 			vap->va_uid = kauth_cred_geteuid(fp->f_cred);
808 			vap->va_gid = kauth_cred_getegid(fp->f_cred);
809 			switch (fp->f_type) {
810 			case DTYPE_VNODE:
811 				vap->va_bytes = vap->va_size =
812 				    fp->f_vnode->v_size;
813 				break;
814 			default:
815 				vap->va_bytes = vap->va_size = 0;
816 				break;
817 			}
818 			closef(fp);
819 			break;
820 		}
821 		/*FALLTHROUGH*/
822 	case PFSproc:
823 		vap->va_nlink = 2;
824 		vap->va_uid = kauth_cred_geteuid(procp->p_cred);
825 		vap->va_gid = kauth_cred_getegid(procp->p_cred);
826 		vap->va_bytes = vap->va_size = DEV_BSIZE;
827 		break;
828 
829 	case PFSfile:
830 		error = EOPNOTSUPP;
831 		break;
832 
833 	case PFSmem:
834 		vap->va_bytes = vap->va_size =
835 			ctob(procp->p_vmspace->vm_tsize +
836 				    procp->p_vmspace->vm_dsize +
837 				    procp->p_vmspace->vm_ssize);
838 		break;
839 
840 	case PFSauxv:
841 		vap->va_bytes = vap->va_size = procp->p_execsw->es_arglen;
842 		break;
843 
844 #if defined(PT_GETREGS) || defined(PT_SETREGS)
845 	case PFSregs:
846 		vap->va_bytes = vap->va_size = sizeof(struct reg);
847 		break;
848 #endif
849 
850 #if defined(PT_GETFPREGS) || defined(PT_SETFPREGS)
851 	case PFSfpregs:
852 		vap->va_bytes = vap->va_size = sizeof(struct fpreg);
853 		break;
854 #endif
855 
856 	case PFSstatus:
857 	case PFSstat:
858 	case PFSnote:
859 	case PFSnotepg:
860 	case PFScmdline:
861 	case PFSenviron:
862 	case PFSmeminfo:
863 	case PFSdevices:
864 	case PFScpuinfo:
865 	case PFSuptime:
866 	case PFSmounts:
867 	case PFScpustat:
868 	case PFSloadavg:
869 	case PFSstatm:
870 	case PFSversion:
871 	case PFSsysvipc:
872 	case PFSsysvipc_msg:
873 	case PFSsysvipc_sem:
874 	case PFSsysvipc_shm:
875 		vap->va_bytes = vap->va_size = 0;
876 		break;
877 	case PFSlimit:
878 	case PFSlimits:
879 	case PFSmap:
880 	case PFSmaps:
881 		/*
882 		 * Advise a larger blocksize for the map files, so that
883 		 * they may be read in one pass.
884 		 */
885 		vap->va_blocksize = 4 * PAGE_SIZE;
886 		vap->va_bytes = vap->va_size = 0;
887 		break;
888 
889 	case PFScwd:
890 	case PFSchroot:
891 		bp = path + MAXPATHLEN;
892 		*--bp = '\0';
893 		procfs_dir(pfs->pfs_type, curlwp, procp, &bp, path,
894 		     MAXPATHLEN);
895 		vap->va_bytes = vap->va_size = strlen(bp);
896 		break;
897 
898 	case PFSexe:
899 		vap->va_bytes = vap->va_size = strlen(procp->p_path);
900 		break;
901 
902 	case PFSemul:
903 		vap->va_bytes = vap->va_size = strlen(procp->p_emul->e_name);
904 		break;
905 
906 #ifdef __HAVE_PROCFS_MACHDEP
907 	PROCFS_MACHDEP_NODETYPE_CASES
908 		error = procfs_machdep_getattr(ap->a_vp, vap, procp);
909 		break;
910 #endif
911 
912 	default:
913 		panic("%s: %d/2", __func__, pfs->pfs_type);
914 	}
915 
916 	if (procp != NULL)
917 		procfs_proc_unlock(procp);
918 	if (path != NULL)
919 		free(path, M_TEMP);
920 
921 	return (error);
922 }
923 
924 /*ARGSUSED*/
925 int
926 procfs_setattr(void *v)
927 {
928 	/*
929 	 * just fake out attribute setting
930 	 * it's not good to generate an error
931 	 * return, otherwise things like creat()
932 	 * will fail when they try to set the
933 	 * file length to 0.  worse, this means
934 	 * that echo $note > /proc/$pid/note will fail.
935 	 */
936 
937 	return (0);
938 }
939 
940 /*
941  * implement access checking.
942  *
943  * actually, the check for super-user is slightly
944  * broken since it will allow read access to write-only
945  * objects.  this doesn't cause any particular trouble
946  * but does mean that the i/o entry points need to check
947  * that the operation really does make sense.
948  */
949 int
950 procfs_access(void *v)
951 {
952 	struct vop_access_args /* {
953 		struct vnode *a_vp;
954 		accmode_t a_accmode;
955 		kauth_cred_t a_cred;
956 	} */ *ap = v;
957 	struct vattr va;
958 	int error;
959 
960 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred)) != 0)
961 		return (error);
962 
963 	return kauth_authorize_vnode(ap->a_cred,
964 	    KAUTH_ACCESS_ACTION(ap->a_accmode, ap->a_vp->v_type, va.va_mode),
965 	    ap->a_vp, NULL, genfs_can_access(ap->a_vp, ap->a_cred,
966 	    va.va_uid, va.va_gid, va.va_mode, NULL, ap->a_accmode));
967 }
968 
969 /*
970  * lookup.  this is incredibly complicated in the
971  * general case, however for most pseudo-filesystems
972  * very little needs to be done.
973  *
974  * Locking isn't hard here, just poorly documented.
975  *
976  * If we're looking up ".", just vref the parent & return it.
977  *
978  * If we're looking up "..", unlock the parent, and lock "..". If everything
979  * went ok, and we're on the last component and the caller requested the
980  * parent locked, try to re-lock the parent. We do this to prevent lock
981  * races.
982  *
983  * For anything else, get the needed node. Then unlock the parent if not
984  * the last component or not LOCKPARENT (i.e. if we wouldn't re-lock the
985  * parent in the .. case).
986  *
987  * We try to exit with the parent locked in error cases.
988  */
989 int
990 procfs_lookup(void *v)
991 {
992 	struct vop_lookup_v2_args /* {
993 		struct vnode * a_dvp;
994 		struct vnode ** a_vpp;
995 		struct componentname * a_cnp;
996 	} */ *ap = v;
997 	struct componentname *cnp = ap->a_cnp;
998 	struct vnode **vpp = ap->a_vpp;
999 	struct vnode *dvp = ap->a_dvp;
1000 	const char *pname = cnp->cn_nameptr;
1001 	const struct proc_target *pt = NULL;
1002 	struct vnode *fvp;
1003 	pid_t pid, vnpid;
1004 	struct pfsnode *pfs;
1005 	struct proc *p = NULL;
1006 	struct lwp *plwp;
1007 	int i, error;
1008 	pfstype type;
1009 
1010 	*vpp = NULL;
1011 
1012 	if ((error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred)) != 0)
1013 		return (error);
1014 
1015 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
1016 		return (EROFS);
1017 
1018 	if (cnp->cn_namelen == 1 && *pname == '.') {
1019 		*vpp = dvp;
1020 		vref(dvp);
1021 		return (0);
1022 	}
1023 
1024 	pfs = VTOPFS(dvp);
1025 	switch (pfs->pfs_type) {
1026 	case PFSroot:
1027 		/*
1028 		 * Shouldn't get here with .. in the root node.
1029 		 */
1030 		if (cnp->cn_flags & ISDOTDOT)
1031 			return (EIO);
1032 
1033 		for (i = 0; i < nproc_root_targets; i++) {
1034 			pt = &proc_root_targets[i];
1035 			/*
1036 			 * check for node match.  proc is always NULL here,
1037 			 * so call pt_valid with constant NULL lwp.
1038 			 */
1039 			if (cnp->cn_namelen == pt->pt_namlen &&
1040 			    memcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
1041 			    (pt->pt_valid == NULL ||
1042 			     (*pt->pt_valid)(NULL, dvp->v_mount)))
1043 				break;
1044 		}
1045 
1046 		if (i != nproc_root_targets) {
1047 			error = procfs_allocvp(dvp->v_mount, vpp, 0,
1048 			    pt->pt_pfstype, -1);
1049 			return (error);
1050 		}
1051 
1052 		if (CNEQ(cnp, "curproc", 7)) {
1053 			pid = curproc->p_pid;
1054 			vnpid = 0;
1055 			type = PFScurproc;
1056 		} else if (CNEQ(cnp, "self", 4)) {
1057 			pid = curproc->p_pid;
1058 			vnpid = 0;
1059 			type = PFSself;
1060 		} else {
1061 			pid = (pid_t)atoi(pname, cnp->cn_namelen);
1062 			vnpid = pid;
1063 			type = PFSproc;
1064 		}
1065 
1066 		if (procfs_proc_lock(dvp->v_mount, pid, &p, ESRCH) != 0)
1067 			break;
1068 		error = procfs_allocvp(dvp->v_mount, vpp, vnpid, type, -1);
1069 		procfs_proc_unlock(p);
1070 		return (error);
1071 
1072 	case PFSproc:
1073 		if (cnp->cn_flags & ISDOTDOT) {
1074 			error = procfs_allocvp(dvp->v_mount, vpp, 0, PFSroot,
1075 			    -1);
1076 			return (error);
1077 		}
1078 
1079 		if (procfs_proc_lock(dvp->v_mount, pfs->pfs_pid, &p,
1080 				     ESRCH) != 0)
1081 			break;
1082 
1083 		mutex_enter(p->p_lock);
1084 		LIST_FOREACH(plwp, &p->p_lwps, l_sibling) {
1085 			if (plwp->l_stat != LSZOMB)
1086 				break;
1087 		}
1088 		/* Process is exiting if no-LWPS or all LWPs are LSZOMB */
1089 		if (plwp == NULL) {
1090 			mutex_exit(p->p_lock);
1091 			procfs_proc_unlock(p);
1092 			return ESRCH;
1093 		}
1094 
1095 		lwp_addref(plwp);
1096 		mutex_exit(p->p_lock);
1097 
1098 		for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) {
1099 			int found;
1100 
1101 			found = cnp->cn_namelen == pt->pt_namlen &&
1102 			    memcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
1103 			    (pt->pt_valid == NULL
1104 			      || (*pt->pt_valid)(plwp, dvp->v_mount));
1105 			if (found)
1106 				break;
1107 		}
1108 		lwp_delref(plwp);
1109 
1110 		if (i == nproc_targets) {
1111 			procfs_proc_unlock(p);
1112 			break;
1113 		}
1114 		if (pt->pt_pfstype == PFSfile) {
1115 			fvp = p->p_textvp;
1116 			/* We already checked that it exists. */
1117 			vref(fvp);
1118 			procfs_proc_unlock(p);
1119 			*vpp = fvp;
1120 			return (0);
1121 		}
1122 
1123 		error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
1124 		    pt->pt_pfstype, -1);
1125 		procfs_proc_unlock(p);
1126 		return (error);
1127 
1128 	case PFSfd: {
1129 		int fd;
1130 		file_t *fp;
1131 
1132 		if ((error = procfs_proc_lock(dvp->v_mount, pfs->pfs_pid, &p,
1133 					      ENOENT)) != 0)
1134 			return error;
1135 
1136 		if (cnp->cn_flags & ISDOTDOT) {
1137 			error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
1138 			    PFSproc, -1);
1139 			procfs_proc_unlock(p);
1140 			return (error);
1141 		}
1142 		fd = atoi(pname, cnp->cn_namelen);
1143 
1144 		fp = fd_getfile2(p, fd);
1145 		if (fp == NULL) {
1146 			procfs_proc_unlock(p);
1147 			return ENOENT;
1148 		}
1149 		fvp = fp->f_vnode;
1150 
1151 		/* Don't show directories */
1152 		if (fp->f_type == DTYPE_VNODE && fvp->v_type != VDIR &&
1153 		    !procfs_proc_is_linux_compat()) {
1154 			vref(fvp);
1155 			closef(fp);
1156 			procfs_proc_unlock(p);
1157 			*vpp = fvp;
1158 			return 0;
1159 		}
1160 
1161 		closef(fp);
1162 		error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
1163 		    PFSfd, fd);
1164 		procfs_proc_unlock(p);
1165 		return error;
1166 	}
1167 	case PFStask: {
1168 		int xpid;
1169 
1170 		if ((error = procfs_proc_lock(dvp->v_mount, pfs->pfs_pid, &p,
1171 					      ENOENT)) != 0)
1172 			return error;
1173 
1174 		if (cnp->cn_flags & ISDOTDOT) {
1175 			error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
1176 			    PFSproc, -1);
1177 			procfs_proc_unlock(p);
1178 			return (error);
1179 		}
1180 		xpid = atoi(pname, cnp->cn_namelen);
1181 
1182 		if (xpid != pfs->pfs_pid) {
1183 			procfs_proc_unlock(p);
1184 			return ENOENT;
1185 		}
1186 		error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
1187 		    PFStask, 0);
1188 		procfs_proc_unlock(p);
1189 		return error;
1190 	}
1191 	case PFSsysvipc:
1192 		if (cnp->cn_flags & ISDOTDOT) {
1193 			error = procfs_allocvp(dvp->v_mount, vpp, 0, PFSroot,
1194 			    -1);
1195 			return (error);
1196 		}
1197 
1198 		for (i = 0; i < nproc_sysvipc_targets; i++) {
1199 			pt = &proc_sysvipc_targets[i];
1200 			/*
1201 			 * check for node match.  proc is always NULL here,
1202 			 * so call pt_valid with constant NULL lwp.
1203 			 */
1204 			if (cnp->cn_namelen == pt->pt_namlen &&
1205 			    memcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
1206 			    (pt->pt_valid == NULL ||
1207 			     (*pt->pt_valid)(NULL, dvp->v_mount)))
1208 				break;
1209 		}
1210 
1211 		if (i != nproc_sysvipc_targets) {
1212 			error = procfs_allocvp(dvp->v_mount, vpp, 0,
1213 			    pt->pt_pfstype, -1);
1214 			return (error);
1215 		}
1216 
1217 		return (ENOENT);
1218 
1219 	default:
1220 		return (ENOTDIR);
1221 	}
1222 
1223 	return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS);
1224 }
1225 
1226 int
1227 procfs_validfile(struct lwp *l, struct mount *mp)
1228 {
1229 	return l != NULL && l->l_proc != NULL && l->l_proc->p_textvp != NULL;
1230 }
1231 
1232 static int
1233 procfs_validfile_linux(struct lwp *l, struct mount *mp)
1234 {
1235 	return procfs_use_linux_compat(mp) &&
1236 	    (l == NULL || l->l_proc == NULL || procfs_validfile(l, mp));
1237 }
1238 
1239 struct procfs_root_readdir_ctx {
1240 	struct uio *uiop;
1241 	off_t *cookies;
1242 	int ncookies;
1243 	off_t off;
1244 	off_t startoff;
1245 	int error;
1246 };
1247 
1248 static int
1249 procfs_root_readdir_callback(struct proc *p, void *arg)
1250 {
1251 	struct procfs_root_readdir_ctx *ctxp = arg;
1252 	struct dirent d;
1253 	struct uio *uiop;
1254 	int error;
1255 
1256 	uiop = ctxp->uiop;
1257 	if (uiop->uio_resid < UIO_MX)
1258 		return -1; /* no space */
1259 
1260 	if (kauth_authorize_process(kauth_cred_get(),
1261 	    KAUTH_PROCESS_CANSEE, p,
1262 	    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL) != 0)
1263 		return 0;
1264 
1265 	if (ctxp->off < ctxp->startoff) {
1266 		ctxp->off++;
1267 		return 0;
1268 	}
1269 
1270 	memset(&d, 0, UIO_MX);
1271 	d.d_reclen = UIO_MX;
1272 	d.d_fileno = PROCFS_FILENO(p->p_pid, PFSproc, -1);
1273 	d.d_namlen = snprintf(d.d_name,
1274 	    UIO_MX - offsetof(struct dirent, d_name), "%ld", (long)p->p_pid);
1275 	d.d_type = DT_DIR;
1276 
1277 	mutex_exit(&proc_lock);
1278 	error = uiomove(&d, UIO_MX, uiop);
1279 	mutex_enter(&proc_lock);
1280 	if (error) {
1281 		ctxp->error = error;
1282 		return -1;
1283 	}
1284 
1285 	ctxp->ncookies++;
1286 	if (ctxp->cookies)
1287 		*(ctxp->cookies)++ = ctxp->off + 1;
1288 	ctxp->off++;
1289 
1290 	return 0;
1291 }
1292 
1293 /*
1294  * readdir returns directory entries from pfsnode (vp).
1295  *
1296  * the strategy here with procfs is to generate a single
1297  * directory entry at a time (struct dirent) and then
1298  * copy that out to userland using uiomove.  a more efficient
1299  * though more complex implementation, would try to minimize
1300  * the number of calls to uiomove().  for procfs, this is
1301  * hardly worth the added code complexity.
1302  *
1303  * this should just be done through read()
1304  */
1305 int
1306 procfs_readdir(void *v)
1307 {
1308 	struct vop_readdir_args /* {
1309 		struct vnode *a_vp;
1310 		struct uio *a_uio;
1311 		kauth_cred_t a_cred;
1312 		int *a_eofflag;
1313 		off_t **a_cookies;
1314 		int *a_ncookies;
1315 	} */ *ap = v;
1316 	struct uio *uio = ap->a_uio;
1317 	struct dirent d;
1318 	struct pfsnode *pfs;
1319 	off_t i;
1320 	int error;
1321 	off_t *cookies = NULL;
1322 	int ncookies;
1323 	struct vnode *vp;
1324 	const struct proc_target *pt;
1325 	struct procfs_root_readdir_ctx ctx;
1326 	struct proc *p = NULL;
1327 	struct lwp *l;
1328 	int nfd;
1329 	int nc = 0;
1330 
1331 	vp = ap->a_vp;
1332 	pfs = VTOPFS(vp);
1333 
1334 	if (uio->uio_resid < UIO_MX)
1335 		return (EINVAL);
1336 	if (uio->uio_offset < 0)
1337 		return (EINVAL);
1338 
1339 	error = 0;
1340 	i = uio->uio_offset;
1341 	memset(&d, 0, UIO_MX);
1342 	d.d_reclen = UIO_MX;
1343 	ncookies = uio->uio_resid / UIO_MX;
1344 
1345 	switch (pfs->pfs_type) {
1346 	/*
1347 	 * this is for the process-specific sub-directories.
1348 	 * all that is needed to is copy out all the entries
1349 	 * from the procent[] table (top of this file).
1350 	 */
1351 	case PFSproc: {
1352 
1353 		if (i >= nproc_targets)
1354 			return 0;
1355 
1356 		if (procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &p, ESRCH) != 0)
1357 			break;
1358 
1359 		if (ap->a_ncookies) {
1360 			ncookies = uimin(ncookies, (nproc_targets - i));
1361 			cookies = malloc(ncookies * sizeof (off_t),
1362 			    M_TEMP, M_WAITOK);
1363 			*ap->a_cookies = cookies;
1364 		}
1365 
1366 		for (pt = &proc_targets[i];
1367 		     uio->uio_resid >= UIO_MX && i < nproc_targets; pt++, i++) {
1368 			if (pt->pt_valid) {
1369 				/* XXXSMP LWP can disappear */
1370 				mutex_enter(p->p_lock);
1371 				l = LIST_FIRST(&p->p_lwps);
1372 				KASSERT(l != NULL);
1373 				mutex_exit(p->p_lock);
1374 				if ((*pt->pt_valid)(l, vp->v_mount) == 0)
1375 					continue;
1376 			}
1377 
1378 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid,
1379 			    pt->pt_pfstype, -1);
1380 			d.d_namlen = pt->pt_namlen;
1381 			memcpy(d.d_name, pt->pt_name, pt->pt_namlen + 1);
1382 			d.d_type = pt->pt_type;
1383 
1384 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1385 				break;
1386 			if (cookies)
1387 				*cookies++ = i + 1;
1388 		}
1389 
1390 		procfs_proc_unlock(p);
1391 	    	break;
1392 	}
1393 	case PFSfd: {
1394 		file_t *fp;
1395 		int lim;
1396 
1397 		if ((error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &p,
1398 					      ESRCH)) != 0)
1399 			return error;
1400 
1401 		/* XXX Should this be by file as well? */
1402 		if (kauth_authorize_process(kauth_cred_get(),
1403 		    KAUTH_PROCESS_CANSEE, p,
1404 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_OPENFILES), NULL,
1405 		    NULL) != 0) {
1406 		    	procfs_proc_unlock(p);
1407 			return ESRCH;
1408 		}
1409 
1410 		nfd = atomic_load_consume(&p->p_fd->fd_dt)->dt_nfiles;
1411 
1412 		lim = uimin((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
1413 		if (i >= lim) {
1414 		    	procfs_proc_unlock(p);
1415 			return 0;
1416 		}
1417 
1418 		if (ap->a_ncookies) {
1419 			ncookies = uimin(ncookies, (nfd + 2 - i));
1420 			cookies = malloc(ncookies * sizeof (off_t),
1421 			    M_TEMP, M_WAITOK);
1422 			*ap->a_cookies = cookies;
1423 		}
1424 
1425 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
1426 			pt = &proc_targets[i];
1427 			d.d_namlen = pt->pt_namlen;
1428 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid,
1429 			    pt->pt_pfstype, -1);
1430 			(void)memcpy(d.d_name, pt->pt_name, pt->pt_namlen + 1);
1431 			d.d_type = pt->pt_type;
1432 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1433 				break;
1434 			if (cookies)
1435 				*cookies++ = i + 1;
1436 			nc++;
1437 		}
1438 		if (error)
1439 			goto out;
1440 		for (; uio->uio_resid >= UIO_MX && i < nfd; i++) {
1441 			/* check the descriptor exists */
1442 			if ((fp = fd_getfile2(p, i - 2)) == NULL)
1443 				continue;
1444 			closef(fp);
1445 
1446 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid, PFSfd, i - 2);
1447 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1448 			    "%lld", (long long)(i - 2));
1449 			d.d_type = fttodt(fp);
1450 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1451 				break;
1452 			if (cookies)
1453 				*cookies++ = i + 1;
1454 			nc++;
1455 		}
1456 		goto out;
1457 	}
1458 	case PFStask: {
1459 
1460 		if ((error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &p,
1461 					      ESRCH)) != 0)
1462 			return error;
1463 
1464 		nfd = 3;	/* ., .., pid */
1465 
1466 		if (ap->a_ncookies) {
1467 			ncookies = uimin(ncookies, (nfd + 2 - i));
1468 			cookies = malloc(ncookies * sizeof (off_t),
1469 			    M_TEMP, M_WAITOK);
1470 			*ap->a_cookies = cookies;
1471 		}
1472 
1473 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
1474 			pt = &proc_targets[i];
1475 			d.d_namlen = pt->pt_namlen;
1476 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid,
1477 			    pt->pt_pfstype, -1);
1478 			(void)memcpy(d.d_name, pt->pt_name, pt->pt_namlen + 1);
1479 			d.d_type = pt->pt_type;
1480 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1481 				break;
1482 			if (cookies)
1483 				*cookies++ = i + 1;
1484 			nc++;
1485 		}
1486 		if (error)
1487 			goto out;
1488 		for (; uio->uio_resid >= UIO_MX && i < nfd; i++) {
1489 			/* check the descriptor exists */
1490 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid, PFStask,
1491 			    i - 2);
1492 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1493 			    "%ld", (long)pfs->pfs_pid);
1494 			d.d_type = DT_LNK;
1495 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1496 				break;
1497 			if (cookies)
1498 				*cookies++ = i + 1;
1499 			nc++;
1500 		}
1501 		goto out;
1502 	}
1503 
1504 	/*
1505 	 * sysvipc subdirectory
1506 	 */
1507 	case PFSsysvipc: {
1508 		if ((error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &p,
1509 					      ESRCH)) != 0)
1510 			return error;
1511 		if (ap->a_ncookies) {
1512 			ncookies = uimin(ncookies, (nproc_sysvipc_targets - i));
1513 			cookies = malloc(ncookies * sizeof (off_t),
1514 			    M_TEMP, M_WAITOK);
1515 			*ap->a_cookies = cookies;
1516 		}
1517 
1518 		for (pt = &proc_sysvipc_targets[i];
1519 		     uio->uio_resid >= UIO_MX && i < nproc_sysvipc_targets; pt++, i++) {
1520 			if (pt->pt_valid &&
1521 			    (*pt->pt_valid)(NULL, vp->v_mount) == 0)
1522 				continue;
1523 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid,
1524 			    pt->pt_pfstype, -1);
1525 			d.d_namlen = pt->pt_namlen;
1526 			memcpy(d.d_name, pt->pt_name, pt->pt_namlen + 1);
1527 			d.d_type = pt->pt_type;
1528 
1529 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1530 				break;
1531 			if (cookies)
1532 				*cookies++ = i + 1;
1533 		}
1534 
1535 		goto out;
1536 	}
1537 
1538 	/*
1539 	 * this is for the root of the procfs filesystem
1540 	 * what is needed are special entries for "curproc"
1541 	 * and "self" followed by an entry for each process
1542 	 * on allproc.
1543 	 */
1544 
1545 	case PFSroot: {
1546 
1547 		if (ap->a_ncookies) {
1548 			/*
1549 			 * XXX Potentially allocating too much space here,
1550 			 * but I'm lazy. This loop needs some work.
1551 			 */
1552 			cookies = malloc(ncookies * sizeof (off_t),
1553 			    M_TEMP, M_WAITOK);
1554 			*ap->a_cookies = cookies;
1555 		}
1556 
1557 		/* 0 ... 3 are static entries. */
1558 		for (; i <= 3 && uio->uio_resid >= UIO_MX; i++) {
1559 			switch (i) {
1560 			case 0:		/* `.' */
1561 			case 1:		/* `..' */
1562 				d.d_fileno = PROCFS_FILENO(0, PFSroot, -1);
1563 				d.d_namlen = i + 1;
1564 				memcpy(d.d_name, "..", d.d_namlen);
1565 				d.d_name[i + 1] = '\0';
1566 				d.d_type = DT_DIR;
1567 				break;
1568 
1569 			case 2:
1570 				d.d_fileno = PROCFS_FILENO(0, PFScurproc, -1);
1571 				d.d_namlen = sizeof("curproc") - 1;
1572 				memcpy(d.d_name, "curproc", sizeof("curproc"));
1573 				d.d_type = DT_LNK;
1574 				break;
1575 
1576 			case 3:
1577 				d.d_fileno = PROCFS_FILENO(0, PFSself, -1);
1578 				d.d_namlen = sizeof("self") - 1;
1579 				memcpy(d.d_name, "self", sizeof("self"));
1580 				d.d_type = DT_LNK;
1581 				break;
1582 			}
1583 
1584 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1585 				break;
1586 			nc++;
1587 			if (cookies)
1588 				*cookies++ = i + 1;
1589 		}
1590 		if (error)
1591 			break;
1592 		/* 4 ... are process entries. */
1593 		ctx.uiop = uio;
1594 		ctx.error = 0;
1595 		ctx.off = 4;
1596 		ctx.startoff = i;
1597 		ctx.cookies = cookies;
1598 		ctx.ncookies = nc;
1599 		proclist_foreach_call(&allproc,
1600 		    procfs_root_readdir_callback, &ctx);
1601 		cookies = ctx.cookies;
1602 		nc = ctx.ncookies;
1603 		error = ctx.error;
1604 		if (error)
1605 			break;
1606 
1607 		/* misc entries. */
1608 		if (i < ctx.off)
1609 			i = ctx.off;
1610 		if (i >= ctx.off + nproc_root_targets)
1611 			break;
1612 		error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &p, ESRCH);
1613 		if (error)
1614 			break;
1615 		for (pt = &proc_root_targets[i - ctx.off];
1616 		    uio->uio_resid >= UIO_MX &&
1617 		    pt < &proc_root_targets[nproc_root_targets];
1618 		    pt++, i++) {
1619 			if (pt->pt_valid &&
1620 			    (*pt->pt_valid)(NULL, vp->v_mount) == 0)
1621 				continue;
1622 			if (kauth_authorize_process(kauth_cred_get(),
1623 			    KAUTH_PROCESS_CANSEE, p,
1624 			    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY),
1625 			    NULL, NULL) != 0)
1626 				continue;
1627 			d.d_fileno = PROCFS_FILENO(0, pt->pt_pfstype, -1);
1628 			d.d_namlen = pt->pt_namlen;
1629 			memcpy(d.d_name, pt->pt_name, pt->pt_namlen + 1);
1630 			d.d_type = pt->pt_type;
1631 
1632 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1633 				break;
1634 			nc++;
1635 			if (cookies)
1636 				*cookies++ = i + 1;
1637 		}
1638 out:
1639 		KASSERT(p != NULL);
1640 		ncookies = nc;
1641 		procfs_proc_unlock(p);
1642 		break;
1643 	}
1644 
1645 	default:
1646 		error = ENOTDIR;
1647 		break;
1648 	}
1649 
1650 	if (ap->a_ncookies) {
1651 		if (error) {
1652 			if (cookies)
1653 				free(*ap->a_cookies, M_TEMP);
1654 			*ap->a_ncookies = 0;
1655 			*ap->a_cookies = NULL;
1656 		} else
1657 			*ap->a_ncookies = ncookies;
1658 	}
1659 	uio->uio_offset = i;
1660 	return (error);
1661 }
1662 
1663 /*
1664  * readlink reads the link of `curproc' and others
1665  */
1666 int
1667 procfs_readlink(void *v)
1668 {
1669 	struct vop_readlink_args *ap = v;
1670 	char bf[16];		/* should be enough */
1671 	char *bp = bf;
1672 	char *path = NULL;
1673 	int len = 0;
1674 	int error = 0;
1675 	struct vnode *vp = ap->a_vp;
1676 	struct pfsnode *pfs = VTOPFS(vp);
1677 	struct proc *pown = NULL;
1678 
1679 	if (pfs->pfs_fileno == PROCFS_FILENO(0, PFScurproc, -1))
1680 		len = snprintf(bf, sizeof(bf), "%ld", (long)curproc->p_pid);
1681 	else if (pfs->pfs_fileno == PROCFS_FILENO(0, PFSself, -1))
1682 		len = snprintf(bf, sizeof(bf), "%s", "curproc");
1683 	else if (pfs->pfs_fileno == PROCFS_FILENO(pfs->pfs_pid, PFStask, 0))
1684 		len = snprintf(bf, sizeof(bf), "..");
1685 	else if (pfs->pfs_fileno == PROCFS_FILENO(pfs->pfs_pid, PFSexe, -1)) {
1686 		if ((error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &pown,
1687 					      ESRCH)) != 0)
1688 			return error;
1689 		bp = pown->p_path;
1690 		len = strlen(bp);
1691 	} else if (pfs->pfs_fileno == PROCFS_FILENO(pfs->pfs_pid, PFScwd, -1) ||
1692 	    pfs->pfs_fileno == PROCFS_FILENO(pfs->pfs_pid, PFSchroot, -1)) {
1693 		if ((error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &pown,
1694 					      ESRCH)) != 0)
1695 			return error;
1696 		path = malloc(MAXPATHLEN + 4, M_TEMP, M_WAITOK);
1697 		if (path == NULL) {
1698 			procfs_proc_unlock(pown);
1699 			return (ENOMEM);
1700 		}
1701 		bp = path + MAXPATHLEN;
1702 		*--bp = '\0';
1703 		procfs_dir(PROCFS_TYPE(pfs->pfs_fileno), curlwp, pown,
1704 		    &bp, path, MAXPATHLEN);
1705 		len = strlen(bp);
1706 	} else {
1707 		file_t *fp;
1708 		struct vnode *vxp;
1709 
1710 		if ((error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &pown,
1711 					      ESRCH)) != 0)
1712 			return error;
1713 
1714 		fp = fd_getfile2(pown, pfs->pfs_fd);
1715 		if (fp == NULL) {
1716 			procfs_proc_unlock(pown);
1717 			return EBADF;
1718 		}
1719 
1720 		switch (fp->f_type) {
1721 		case DTYPE_VNODE:
1722 			vxp = fp->f_vnode;
1723 			if (vxp->v_type != VDIR &&
1724 			    !procfs_proc_is_linux_compat()) {
1725 				error = EINVAL;
1726 				break;
1727 			}
1728 			if ((path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK))
1729 			    == NULL) {
1730 				error = ENOMEM;
1731 				break;
1732 			}
1733 			bp = path + MAXPATHLEN;
1734 			*--bp = '\0';
1735 
1736 			/*
1737 			 * XXX: kludge to avoid locking against ourselves
1738 			 * in getcwd()
1739 			 */
1740 			if (vxp->v_tag == VT_PROCFS) {
1741 				*--bp = '/';
1742 			} else {
1743 				rw_enter(&curproc->p_cwdi->cwdi_lock,
1744 				    RW_READER);
1745 				vp = curproc->p_cwdi->cwdi_rdir;
1746 				if (vp == NULL)
1747 					vp = rootvnode;
1748 				error = getcwd_common(vxp, vp, &bp, path,
1749 				    MAXPATHLEN / 2, 0, curlwp);
1750 				rw_exit(&curproc->p_cwdi->cwdi_lock);
1751 			}
1752 			if (error)
1753 				break;
1754 			len = strlen(bp);
1755 			break;
1756 
1757 		case DTYPE_MISC:
1758 			len = snprintf(bf, sizeof(bf), "%s", "[misc]");
1759 			break;
1760 
1761 		case DTYPE_KQUEUE:
1762 			len = snprintf(bf, sizeof(bf), "%s", "[kqueue]");
1763 			break;
1764 
1765 		case DTYPE_SEM:
1766 			len = snprintf(bf, sizeof(bf), "%s", "[ksem]");
1767 			break;
1768 
1769 		default:
1770 			error = EINVAL;
1771 			break;
1772 		}
1773 		closef(fp);
1774 	}
1775 
1776 	if (error == 0)
1777 		error = uiomove(bp, len, ap->a_uio);
1778 	if (pown)
1779 		procfs_proc_unlock(pown);
1780 	if (path)
1781 		free(path, M_TEMP);
1782 	return error;
1783 }
1784 
1785 int
1786 procfs_getpages(void *v)
1787 {
1788 	struct vop_getpages_args /* {
1789 		struct vnode *a_vp;
1790 		voff_t a_offset;
1791 		struct vm_page **a_m;
1792 		int *a_count;
1793 		int a_centeridx;
1794 		vm_prot_t a_access_type;
1795 		int a_advice;
1796 		int a_flags;
1797 	} */ *ap = v;
1798 
1799 	if ((ap->a_flags & PGO_LOCKED) == 0)
1800 		rw_exit(ap->a_vp->v_uobj.vmobjlock);
1801 
1802 	return (EFAULT);
1803 }
1804 
1805 /*
1806  * convert decimal ascii to int
1807  */
1808 static int
1809 atoi(const char *b, size_t len)
1810 {
1811 	int p = 0;
1812 
1813 	while (len--) {
1814 		char c = *b++;
1815 		if (c < '0' || c > '9')
1816 			return -1;
1817 		p = 10 * p + (c - '0');
1818 	}
1819 
1820 	return p;
1821 }
1822 
1823 /**
1824  * convert DTYPE_XXX to corresponding DT_XXX
1825  * matching what procfs_loadvnode() does.
1826  */
1827 static uint8_t
1828 fttodt(file_t *fp)
1829 {
1830 	switch (fp->f_type) {
1831 	case DTYPE_VNODE:
1832 		switch (fp->f_vnode->v_type) {
1833 		case VREG:	return DT_REG;
1834 		case VDIR:	return DT_LNK;	/* symlink */
1835 		case VBLK:	return DT_BLK;
1836 		case VCHR:	return DT_CHR;
1837 		case VLNK:	return DT_LNK;
1838 		case VSOCK:	return DT_SOCK;
1839 		case VFIFO:	return DT_FIFO;
1840 		default:	return DT_UNKNOWN;
1841 		}
1842 	case DTYPE_PIPE:	return DT_FIFO;
1843 	case DTYPE_SOCKET:	return DT_SOCK;
1844 	case DTYPE_KQUEUE:	/*FALLTHROUGH*/
1845 	case DTYPE_MISC:	/*FALLTHROUGH*/
1846 	case DTYPE_SEM:		return DT_LNK;	/* symlinks */
1847 	default:		return DT_UNKNOWN;
1848 	}
1849 }
1850