Home | History | Annotate | Line # | Download | only in procfs
      1 /*	$NetBSD: procfs_vnops.c,v 1.233 2024/07/01 01:35:53 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 1.233 2024/07/01 01:35:53 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("sys"),         PFSsys,            procfs_validfile_linux },
    207 	{ DT_DIR, N("sysvipc"),     PFSsysvipc,        procfs_validfile_linux },
    208 #undef N
    209 };
    210 static const int nproc_root_targets =
    211     sizeof(proc_root_targets) / sizeof(proc_root_targets[0]);
    212 
    213 /*
    214  * List of files in the sys directory
    215  */
    216 static const struct proc_target proc_sys_targets[] = {
    217 #define N(s) sizeof(s)-1, s
    218         /*        name              type            validp */
    219 	{ DT_DIR, N("."),	PFSsys,		procfs_validfile_linux },
    220 	{ DT_DIR, N(".."),	PFSroot,	NULL },
    221 	{ DT_REG, N("fs"),	PFSsysfs,	procfs_validfile_linux },
    222 #undef N
    223 };
    224 static const int nproc_sys_targets =
    225     sizeof(proc_sys_targets) / sizeof(proc_sys_targets[0]);
    226 
    227 /*
    228  * List of files in the sys/fs directory
    229  */
    230 static const struct proc_target proc_sysfs_targets[] = {
    231 #define N(s) sizeof(s)-1, s
    232         /*        name              type            validp */
    233 	{ DT_DIR, N("."),	PFSsysfs,	procfs_validfile_linux },
    234 	{ DT_DIR, N(".."),	PFSsys,		procfs_validfile_linux },
    235 	{ DT_REG, N("mqueue"),	PFSmqueue,	procfs_validfile_linux },
    236 #undef N
    237 };
    238 static const int nproc_sysfs_targets =
    239     sizeof(proc_sysfs_targets) / sizeof(proc_sysfs_targets[0]);
    240 
    241 /*
    242  * List of files in the sys/fs/mqueue directory
    243  */
    244 static const struct proc_target proc_mqueue_targets[] = {
    245 #define N(s) sizeof(s)-1, s
    246         /*        name              	type            validp */
    247 	{ DT_DIR, N("."),		PFSmqueue,	procfs_validfile_linux },
    248 	{ DT_DIR, N(".."),		PFSsysfs,	procfs_validfile_linux },
    249 	{ DT_REG, N("msg_default"),	PFSmq_msg_def,	procfs_validfile_linux },
    250 	{ DT_REG, N("msg_max"),		PFSmq_msg_max,	procfs_validfile_linux },
    251 	{ DT_REG, N("msgsize_default"),	PFSmq_siz_def,	procfs_validfile_linux },
    252 	{ DT_REG, N("msgsize_max"),	PFSmq_siz_max,	procfs_validfile_linux },
    253 	{ DT_REG, N("queues_max"),	PFSmq_qmax,	procfs_validfile_linux },
    254 #undef N
    255 };
    256 static const int nproc_mqueue_targets =
    257     sizeof(proc_mqueue_targets) / sizeof(proc_mqueue_targets[0]);
    258 
    259 /*
    260  * List of files in the sysvipc directory
    261  */
    262 static const struct proc_target proc_sysvipc_targets[] = {
    263 #define N(s) sizeof(s)-1, s
    264         /*        name              type            validp */
    265 	{ DT_DIR, N("."),	PFSsysvipc,	NULL },
    266 	{ DT_DIR, N(".."),	PFSroot,	NULL },
    267 	{ DT_REG, N("msg"),	PFSsysvipc_msg, procfs_validfile_linux },
    268 	{ DT_REG, N("sem"),	PFSsysvipc_sem, procfs_validfile_linux },
    269 	{ DT_REG, N("shm"),	PFSsysvipc_shm, procfs_validfile_linux },
    270 #undef N
    271 };
    272 static const int nproc_sysvipc_targets =
    273     sizeof(proc_sysvipc_targets) / sizeof(proc_sysvipc_targets[0]);
    274 
    275 int	procfs_lookup(void *);
    276 int	procfs_open(void *);
    277 int	procfs_close(void *);
    278 int	procfs_access(void *);
    279 int	procfs_getattr(void *);
    280 int	procfs_setattr(void *);
    281 int	procfs_readdir(void *);
    282 int	procfs_readlink(void *);
    283 int	procfs_inactive(void *);
    284 int	procfs_reclaim(void *);
    285 int	procfs_print(void *);
    286 int	procfs_pathconf(void *);
    287 int	procfs_getpages(void *);
    288 
    289 static uint8_t fttodt(file_t *);
    290 static int atoi(const char *, size_t);
    291 
    292 /*
    293  * procfs vnode operations.
    294  */
    295 int (**procfs_vnodeop_p)(void *);
    296 const struct vnodeopv_entry_desc procfs_vnodeop_entries[] = {
    297 	{ &vop_default_desc, vn_default_error },
    298 	{ &vop_parsepath_desc, genfs_parsepath },	/* parsepath */
    299 	{ &vop_lookup_desc, procfs_lookup },		/* lookup */
    300 	{ &vop_create_desc, genfs_eopnotsupp },		/* create */
    301 	{ &vop_mknod_desc, genfs_eopnotsupp },		/* mknod */
    302 	{ &vop_open_desc, procfs_open },		/* open */
    303 	{ &vop_close_desc, procfs_close },		/* close */
    304 	{ &vop_access_desc, procfs_access },		/* access */
    305 	{ &vop_accessx_desc, genfs_accessx },		/* accessx */
    306 	{ &vop_getattr_desc, procfs_getattr },		/* getattr */
    307 	{ &vop_setattr_desc, procfs_setattr },		/* setattr */
    308 	{ &vop_read_desc, procfs_rw },			/* read */
    309 	{ &vop_write_desc, procfs_rw },			/* write */
    310 	{ &vop_fallocate_desc, genfs_eopnotsupp },	/* fallocate */
    311 	{ &vop_fdiscard_desc, genfs_eopnotsupp },	/* fdiscard */
    312 	{ &vop_fcntl_desc, genfs_fcntl },		/* fcntl */
    313 	{ &vop_ioctl_desc, genfs_enoioctl },		/* ioctl */
    314 	{ &vop_poll_desc, genfs_poll },			/* poll */
    315 	{ &vop_kqfilter_desc, genfs_kqfilter },		/* kqfilter */
    316 	{ &vop_revoke_desc, genfs_revoke },		/* revoke */
    317 	{ &vop_fsync_desc, genfs_nullop },		/* fsync */
    318 	{ &vop_seek_desc, genfs_nullop },		/* seek */
    319 	{ &vop_remove_desc, genfs_eopnotsupp },		/* remove */
    320 	{ &vop_link_desc, genfs_erofs_link },		/* link */
    321 	{ &vop_rename_desc, genfs_eopnotsupp },		/* rename */
    322 	{ &vop_mkdir_desc, genfs_eopnotsupp },		/* mkdir */
    323 	{ &vop_rmdir_desc, genfs_eopnotsupp },		/* rmdir */
    324 	{ &vop_symlink_desc, genfs_erofs_symlink },	/* symlink */
    325 	{ &vop_readdir_desc, procfs_readdir },		/* readdir */
    326 	{ &vop_readlink_desc, procfs_readlink },	/* readlink */
    327 	{ &vop_abortop_desc, genfs_abortop },		/* abortop */
    328 	{ &vop_inactive_desc, procfs_inactive },	/* inactive */
    329 	{ &vop_reclaim_desc, procfs_reclaim },		/* reclaim */
    330 	{ &vop_lock_desc, genfs_lock },			/* lock */
    331 	{ &vop_unlock_desc, genfs_unlock },		/* unlock */
    332 	{ &vop_bmap_desc, genfs_eopnotsupp },		/* bmap */
    333 	{ &vop_strategy_desc, genfs_badop },		/* strategy */
    334 	{ &vop_print_desc, procfs_print },		/* print */
    335 	{ &vop_islocked_desc, genfs_islocked },		/* islocked */
    336 	{ &vop_pathconf_desc, procfs_pathconf },	/* pathconf */
    337 	{ &vop_advlock_desc, genfs_einval },		/* advlock */
    338 	{ &vop_getpages_desc, procfs_getpages },	/* getpages */
    339 	{ &vop_putpages_desc, genfs_null_putpages },	/* putpages */
    340 	{ NULL, NULL }
    341 };
    342 const struct vnodeopv_desc procfs_vnodeop_opv_desc =
    343 	{ &procfs_vnodeop_p, procfs_vnodeop_entries };
    344 /*
    345  * set things up for doing i/o on
    346  * the pfsnode (vp).  (vp) is locked
    347  * on entry, and should be left locked
    348  * on exit.
    349  *
    350  * for procfs we don't need to do anything
    351  * in particular for i/o.  all that is done
    352  * is to support exclusive open on process
    353  * memory images.
    354  */
    355 int
    356 procfs_open(void *v)
    357 {
    358 	struct vop_open_args /* {
    359 		struct vnode *a_vp;
    360 		int  a_mode;
    361 		kauth_cred_t a_cred;
    362 	} */ *ap = v;
    363 	struct vnode *vp = ap->a_vp;
    364 	struct pfsnode *pfs = VTOPFS(vp);
    365 	struct lwp *l1;
    366 	struct proc *p2;
    367 	int error;
    368 
    369 	if ((error =
    370 	     procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &p2, ENOENT)) != 0)
    371 		return error;
    372 
    373 	l1 = curlwp;				/* tracer */
    374 
    375 #define	M2K(m)	(((m) & FREAD) && ((m) & FWRITE) ? \
    376 		 KAUTH_REQ_PROCESS_PROCFS_RW : \
    377 		 (m) & FWRITE ? KAUTH_REQ_PROCESS_PROCFS_WRITE : \
    378 		 KAUTH_REQ_PROCESS_PROCFS_READ)
    379 
    380 	mutex_enter(p2->p_lock);
    381 	error = kauth_authorize_process(l1->l_cred, KAUTH_PROCESS_PROCFS,
    382 	    p2, pfs, KAUTH_ARG(M2K(ap->a_mode)), NULL);
    383 	mutex_exit(p2->p_lock);
    384 	if (error) {
    385 		procfs_proc_unlock(p2);
    386 		return (error);
    387 	}
    388 
    389 #undef M2K
    390 
    391 	switch (pfs->pfs_type) {
    392 	case PFSmem:
    393 		if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) ||
    394 		    ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE))) {
    395 			error = EBUSY;
    396 			break;
    397 		}
    398 
    399 		if (!proc_isunder(p2, l1)) {
    400 			error = EPERM;
    401 			break;
    402 		}
    403 
    404 		if (ap->a_mode & FWRITE)
    405 			pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL);
    406 
    407 		break;
    408 
    409 	case PFSregs:
    410 	case PFSfpregs:
    411 		if (!proc_isunder(p2, l1)) {
    412 			error = EPERM;
    413 			break;
    414 		}
    415 		break;
    416 
    417 	default:
    418 		break;
    419 	}
    420 
    421 	procfs_proc_unlock(p2);
    422 	return (error);
    423 }
    424 
    425 /*
    426  * close the pfsnode (vp) after doing i/o.
    427  * (vp) is not locked on entry or exit.
    428  *
    429  * nothing to do for procfs other than undo
    430  * any exclusive open flag (see _open above).
    431  */
    432 int
    433 procfs_close(void *v)
    434 {
    435 	struct vop_close_args /* {
    436 		struct vnode *a_vp;
    437 		int  a_fflag;
    438 		kauth_cred_t a_cred;
    439 	} */ *ap = v;
    440 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
    441 
    442 	switch (pfs->pfs_type) {
    443 	case PFSmem:
    444 		if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL))
    445 			pfs->pfs_flags &= ~(FWRITE|O_EXCL);
    446 		break;
    447 
    448 	default:
    449 		break;
    450 	}
    451 
    452 	return (0);
    453 }
    454 
    455 /*
    456  * _inactive is called when the pfsnode
    457  * is vrele'd and the reference count goes
    458  * to zero.  (vp) will be on the vnode free
    459  * list, so to get it back vget() must be
    460  * used.
    461  *
    462  * (vp) is locked on entry, but must be unlocked on exit.
    463  */
    464 int
    465 procfs_inactive(void *v)
    466 {
    467 	struct vop_inactive_v2_args /* {
    468 		struct vnode *a_vp;
    469 		bool *a_recycle;
    470 	} */ *ap = v;
    471 	struct vnode *vp = ap->a_vp;
    472 	struct pfsnode *pfs = VTOPFS(vp);
    473 
    474 	mutex_enter(&proc_lock);
    475 	*ap->a_recycle = (procfs_proc_find(vp->v_mount, pfs->pfs_pid) == NULL);
    476 	mutex_exit(&proc_lock);
    477 
    478 	return (0);
    479 }
    480 
    481 /*
    482  * _reclaim is called when getnewvnode()
    483  * wants to make use of an entry on the vnode
    484  * free list.  at this time the filesystem needs
    485  * to free any private data and remove the node
    486  * from any private lists.
    487  */
    488 int
    489 procfs_reclaim(void *v)
    490 {
    491 	struct vop_reclaim_v2_args /* {
    492 		struct vnode *a_vp;
    493 	} */ *ap = v;
    494 	struct vnode *vp = ap->a_vp;
    495 	struct pfsnode *pfs = VTOPFS(vp);
    496 
    497 	VOP_UNLOCK(vp);
    498 
    499 	/*
    500 	 * To interlock with procfs_revoke_vnodes().
    501 	 */
    502 	mutex_enter(vp->v_interlock);
    503 	vp->v_data = NULL;
    504 	mutex_exit(vp->v_interlock);
    505 	procfs_hashrem(pfs);
    506 	kmem_free(pfs, sizeof(*pfs));
    507 	return 0;
    508 }
    509 
    510 /*
    511  * Return POSIX pathconf information applicable to special devices.
    512  */
    513 int
    514 procfs_pathconf(void *v)
    515 {
    516 	struct vop_pathconf_args /* {
    517 		struct vnode *a_vp;
    518 		int a_name;
    519 		register_t *a_retval;
    520 	} */ *ap = v;
    521 
    522 	switch (ap->a_name) {
    523 	case _PC_LINK_MAX:
    524 		*ap->a_retval = LINK_MAX;
    525 		return (0);
    526 	case _PC_MAX_CANON:
    527 		*ap->a_retval = MAX_CANON;
    528 		return (0);
    529 	case _PC_MAX_INPUT:
    530 		*ap->a_retval = MAX_INPUT;
    531 		return (0);
    532 	case _PC_PIPE_BUF:
    533 		*ap->a_retval = PIPE_BUF;
    534 		return (0);
    535 	case _PC_CHOWN_RESTRICTED:
    536 		*ap->a_retval = 1;
    537 		return (0);
    538 	case _PC_VDISABLE:
    539 		*ap->a_retval = _POSIX_VDISABLE;
    540 		return (0);
    541 	case _PC_SYNC_IO:
    542 		*ap->a_retval = 1;
    543 		return (0);
    544 	default:
    545 		return genfs_pathconf(ap);
    546 	}
    547 	/* NOTREACHED */
    548 }
    549 
    550 /*
    551  * _print is used for debugging.
    552  * just print a readable description
    553  * of (vp).
    554  */
    555 int
    556 procfs_print(void *v)
    557 {
    558 	struct vop_print_args /* {
    559 		struct vnode *a_vp;
    560 	} */ *ap = v;
    561 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
    562 
    563 	printf("tag VT_PROCFS, type %d, pid %d, mode %x, flags %lx\n",
    564 	    pfs->pfs_type, pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags);
    565 	return 0;
    566 }
    567 
    568 /*
    569  * Works out the path to the target process's current
    570  * working directory or chroot.  If the caller is in a chroot and
    571  * can't "reach" the target's cwd or root (or some other error
    572  * occurs), a "/" is returned for the path.
    573  */
    574 static void
    575 procfs_dir(pfstype t, struct lwp *caller, struct proc *target, char **bpp,
    576     char *path, size_t len)
    577 {
    578 	struct cwdinfo *cwdi;
    579 	struct vnode *vp, *rvp;
    580 	char *bp;
    581 
    582 	/*
    583 	 * Lock target cwdi and take a reference to the vnode
    584 	 * we are interested in to prevent it from disappearing
    585 	 * before getcwd_common() below.
    586 	 */
    587 	rw_enter(&target->p_cwdi->cwdi_lock, RW_READER);
    588 	switch (t) {
    589 	case PFScwd:
    590 		vp = target->p_cwdi->cwdi_cdir;
    591 		break;
    592 	case PFSchroot:
    593 		vp = target->p_cwdi->cwdi_rdir;
    594 		break;
    595 	default:
    596 		rw_exit(&target->p_cwdi->cwdi_lock);
    597 		return;
    598 	}
    599 	if (vp != NULL)
    600 		vref(vp);
    601 	rw_exit(&target->p_cwdi->cwdi_lock);
    602 
    603 	cwdi = caller->l_proc->p_cwdi;
    604 	rw_enter(&cwdi->cwdi_lock, RW_READER);
    605 
    606 	rvp = cwdi->cwdi_rdir;
    607 	bp = bpp ? *bpp : NULL;
    608 
    609 	/*
    610 	 * XXX: this horrible kludge avoids locking panics when
    611 	 * attempting to lookup links that point to within procfs
    612 	 */
    613 	if (vp != NULL && vp->v_tag == VT_PROCFS) {
    614 		if (bpp) {
    615 			*--bp = '/';
    616 			*bpp = bp;
    617 		}
    618 		vrele(vp);
    619 		rw_exit(&cwdi->cwdi_lock);
    620 		return;
    621 	}
    622 
    623 	if (rvp == NULL)
    624 		rvp = rootvnode;
    625 	if (vp == NULL || getcwd_common(vp, rvp, bp ? &bp : NULL, path,
    626 	    len / 2, 0, caller) != 0) {
    627 		if (bpp) {
    628 			bp = *bpp;
    629 			*--bp = '/';
    630 		}
    631 	}
    632 
    633 	if (bpp)
    634 		*bpp = bp;
    635 
    636 	if (vp != NULL)
    637 		vrele(vp);
    638 	rw_exit(&cwdi->cwdi_lock);
    639 }
    640 
    641 /*
    642  * Invent attributes for pfsnode (vp) and store
    643  * them in (vap).
    644  * Directories lengths are returned as zero since
    645  * any real length would require the genuine size
    646  * to be computed, and nothing cares anyway.
    647  *
    648  * this is relatively minimal for procfs.
    649  */
    650 int
    651 procfs_getattr(void *v)
    652 {
    653 	struct vop_getattr_args /* {
    654 		struct vnode *a_vp;
    655 		struct vattr *a_vap;
    656 		kauth_cred_t a_cred;
    657 	} */ *ap = v;
    658 	struct vnode *vp = ap->a_vp;
    659 	struct pfsnode *pfs = VTOPFS(vp);
    660 	struct vattr *vap = ap->a_vap;
    661 	struct proc *procp;
    662 	char *path, *bp, bf[16];
    663 	int error;
    664 
    665 	/* first check the process still exists */
    666 	switch (pfs->pfs_type) {
    667 	case PFSroot:
    668 	case PFScurproc:
    669 	case PFSself:
    670 		procp = NULL;
    671 		break;
    672 
    673 	default:
    674 		error =
    675 		    procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &procp, ENOENT);
    676 		if (error != 0)
    677 			return (error);
    678 		break;
    679 	}
    680 
    681 	switch (pfs->pfs_type) {
    682 	case PFStask:
    683 		if (pfs->pfs_fd == -1) {
    684 			path = NULL;
    685 			break;
    686 		}
    687 		/*FALLTHROUGH*/
    688 	case PFScwd:
    689 	case PFSchroot:
    690 		path = malloc(MAXPATHLEN + 4, M_TEMP, M_WAITOK);
    691 		if (path == NULL && procp != NULL) {
    692 			procfs_proc_unlock(procp);
    693 			return (ENOMEM);
    694 		}
    695 		break;
    696 
    697 	default:
    698 		path = NULL;
    699 		break;
    700 	}
    701 
    702 	if (procp != NULL) {
    703 		mutex_enter(procp->p_lock);
    704 		error = kauth_authorize_process(kauth_cred_get(),
    705 		    KAUTH_PROCESS_CANSEE, procp,
    706 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
    707 		mutex_exit(procp->p_lock);
    708 		if (error != 0) {
    709 		    	procfs_proc_unlock(procp);
    710 		    	if (path != NULL)
    711 		    		free(path, M_TEMP);
    712 			return (ENOENT);
    713 		}
    714 	}
    715 
    716 	error = 0;
    717 
    718 	/* start by zeroing out the attributes */
    719 	vattr_null(vap);
    720 
    721 	/* next do all the common fields */
    722 	vap->va_type = ap->a_vp->v_type;
    723 	vap->va_mode = pfs->pfs_mode;
    724 	vap->va_fileid = pfs->pfs_fileno;
    725 	vap->va_flags = 0;
    726 	vap->va_blocksize = PAGE_SIZE;
    727 
    728 	/*
    729 	 * Make all times be current TOD.
    730 	 *
    731 	 * It would be possible to get the process start
    732 	 * time from the p_stats structure, but there's
    733 	 * no "file creation" time stamp anyway, and the
    734 	 * p_stats structure is not addressable if u. gets
    735 	 * swapped out for that process.
    736 	 */
    737 	getnanotime(&vap->va_ctime);
    738 	vap->va_atime = vap->va_mtime = vap->va_ctime;
    739 	if (procp)
    740 		TIMEVAL_TO_TIMESPEC(&procp->p_stats->p_start,
    741 		    &vap->va_birthtime);
    742 	else
    743 		getnanotime(&vap->va_birthtime);
    744 
    745 	switch (pfs->pfs_type) {
    746 	case PFSmem:
    747 	case PFSregs:
    748 	case PFSfpregs:
    749 #if defined(__HAVE_PROCFS_MACHDEP) && defined(PROCFS_MACHDEP_PROTECT_CASES)
    750 	PROCFS_MACHDEP_PROTECT_CASES
    751 #endif
    752 		/*
    753 		 * If the process has exercised some setuid or setgid
    754 		 * privilege, then rip away read/write permission so
    755 		 * that only root can gain access.
    756 		 */
    757 		if (procp->p_flag & PK_SUGID)
    758 			vap->va_mode &= ~(S_IRUSR|S_IWUSR);
    759 		/* FALLTHROUGH */
    760 	case PFSstatus:
    761 	case PFSstat:
    762 	case PFSnote:
    763 	case PFSnotepg:
    764 	case PFScmdline:
    765 	case PFSenviron:
    766 	case PFSemul:
    767 	case PFSstatm:
    768 
    769 	case PFSmap:
    770 	case PFSmaps:
    771 	case PFSlimit:
    772 	case PFSlimits:
    773 	case PFSauxv:
    774 		vap->va_nlink = 1;
    775 		vap->va_uid = kauth_cred_geteuid(procp->p_cred);
    776 		vap->va_gid = kauth_cred_getegid(procp->p_cred);
    777 		break;
    778 	case PFScwd:
    779 	case PFSchroot:
    780 	case PFSmeminfo:
    781 	case PFSdevices:
    782 	case PFScpuinfo:
    783 	case PFSuptime:
    784 	case PFSmounts:
    785 	case PFScpustat:
    786 	case PFSloadavg:
    787 	case PFSversion:
    788 	case PFSexe:
    789 	case PFSself:
    790 	case PFScurproc:
    791 	case PFSroot:
    792 	case PFSmq_msg_def:
    793 	case PFSmq_msg_max:
    794 	case PFSmq_siz_def:
    795 	case PFSmq_siz_max:
    796 	case PFSmq_qmax:
    797 	case PFSsysvipc_msg:
    798 	case PFSsysvipc_sem:
    799 	case PFSsysvipc_shm:
    800 		vap->va_nlink = 1;
    801 		vap->va_uid = vap->va_gid = 0;
    802 		break;
    803 
    804 	case PFSsysvipc:
    805 		vap->va_nlink = 5;
    806 		vap->va_uid = vap->va_gid = 0;
    807 		break;
    808 
    809 	case PFSsys:	/* proc/sys only contains "fs" */
    810 	case PFSsysfs:	/* proc/sys/fs only contains "mqueue" */
    811 		vap->va_nlink = 3;
    812 		vap->va_uid = vap->va_gid = 0;
    813 		break;
    814 
    815 	case PFSmqueue:
    816 		vap->va_nlink = 7;
    817 		vap->va_uid = vap->va_gid = 0;
    818 		break;
    819 
    820 	case PFSproc:
    821 	case PFStask:
    822 	case PFSfile:
    823 	case PFSfd:
    824 		break;
    825 
    826 	default:
    827 		panic("%s: %d/1", __func__, pfs->pfs_type);
    828 	}
    829 
    830 	/*
    831 	 * now do the object specific fields
    832 	 *
    833 	 * The size could be set from struct reg, but it's hardly
    834 	 * worth the trouble, and it puts some (potentially) machine
    835 	 * dependent data into this machine-independent code.  If it
    836 	 * becomes important then this function should break out into
    837 	 * a per-file stat function in the corresponding .c file.
    838 	 */
    839 
    840 	switch (pfs->pfs_type) {
    841 	case PFSroot:
    842 		vap->va_bytes = vap->va_size = DEV_BSIZE;
    843 		break;
    844 
    845 	case PFSself:
    846 	case PFScurproc:
    847 		vap->va_bytes = vap->va_size =
    848 		    snprintf(bf, sizeof(bf), "%ld", (long)curproc->p_pid);
    849 		break;
    850 	case PFStask:
    851 		if (pfs->pfs_fd != -1) {
    852 			vap->va_nlink = 1;
    853 			vap->va_uid = 0;
    854 			vap->va_gid = 0;
    855 			vap->va_bytes = vap->va_size =
    856 			    snprintf(bf, sizeof(bf), "..");
    857 			break;
    858 		}
    859 		/*FALLTHROUGH*/
    860 	case PFSfd:
    861 		if (pfs->pfs_fd != -1) {
    862 			file_t *fp;
    863 
    864 			fp = fd_getfile2(procp, pfs->pfs_fd);
    865 			if (fp == NULL) {
    866 				error = EBADF;
    867 				break;
    868 			}
    869 			vap->va_nlink = 1;
    870 			vap->va_uid = kauth_cred_geteuid(fp->f_cred);
    871 			vap->va_gid = kauth_cred_getegid(fp->f_cred);
    872 			switch (fp->f_type) {
    873 			case DTYPE_VNODE:
    874 				vap->va_bytes = vap->va_size =
    875 				    fp->f_vnode->v_size;
    876 				break;
    877 			default:
    878 				vap->va_bytes = vap->va_size = 0;
    879 				break;
    880 			}
    881 			closef(fp);
    882 			break;
    883 		}
    884 		/*FALLTHROUGH*/
    885 	case PFSproc:
    886 		vap->va_nlink = 2;
    887 		vap->va_uid = kauth_cred_geteuid(procp->p_cred);
    888 		vap->va_gid = kauth_cred_getegid(procp->p_cred);
    889 		vap->va_bytes = vap->va_size = DEV_BSIZE;
    890 		break;
    891 
    892 	case PFSfile:
    893 		error = EOPNOTSUPP;
    894 		break;
    895 
    896 	case PFSmem:
    897 		vap->va_bytes = vap->va_size =
    898 			ctob(procp->p_vmspace->vm_tsize +
    899 				    procp->p_vmspace->vm_dsize +
    900 				    procp->p_vmspace->vm_ssize);
    901 		break;
    902 
    903 	case PFSauxv:
    904 		vap->va_bytes = vap->va_size = procp->p_execsw->es_arglen;
    905 		break;
    906 
    907 #if defined(PT_GETREGS) || defined(PT_SETREGS)
    908 	case PFSregs:
    909 		vap->va_bytes = vap->va_size = sizeof(struct reg);
    910 		break;
    911 #endif
    912 
    913 #if defined(PT_GETFPREGS) || defined(PT_SETFPREGS)
    914 	case PFSfpregs:
    915 		vap->va_bytes = vap->va_size = sizeof(struct fpreg);
    916 		break;
    917 #endif
    918 
    919 	case PFSstatus:
    920 	case PFSstat:
    921 	case PFSnote:
    922 	case PFSnotepg:
    923 	case PFScmdline:
    924 	case PFSenviron:
    925 	case PFSmeminfo:
    926 	case PFSdevices:
    927 	case PFScpuinfo:
    928 	case PFSuptime:
    929 	case PFSmounts:
    930 	case PFScpustat:
    931 	case PFSloadavg:
    932 	case PFSstatm:
    933 	case PFSversion:
    934 	case PFSsys:
    935 	case PFSsysfs:
    936 	case PFSmqueue:
    937 	case PFSmq_msg_def:
    938 	case PFSmq_msg_max:
    939 	case PFSmq_siz_def:
    940 	case PFSmq_siz_max:
    941 	case PFSmq_qmax:
    942 	case PFSsysvipc:
    943 	case PFSsysvipc_msg:
    944 	case PFSsysvipc_sem:
    945 	case PFSsysvipc_shm:
    946 		vap->va_bytes = vap->va_size = 0;
    947 		break;
    948 	case PFSlimit:
    949 	case PFSlimits:
    950 	case PFSmap:
    951 	case PFSmaps:
    952 		/*
    953 		 * Advise a larger blocksize for the map files, so that
    954 		 * they may be read in one pass.
    955 		 */
    956 		vap->va_blocksize = 4 * PAGE_SIZE;
    957 		vap->va_bytes = vap->va_size = 0;
    958 		break;
    959 
    960 	case PFScwd:
    961 	case PFSchroot:
    962 		bp = path + MAXPATHLEN;
    963 		*--bp = '\0';
    964 		procfs_dir(pfs->pfs_type, curlwp, procp, &bp, path,
    965 		     MAXPATHLEN);
    966 		vap->va_bytes = vap->va_size = strlen(bp);
    967 		break;
    968 
    969 	case PFSexe:
    970 		vap->va_bytes = vap->va_size = strlen(procp->p_path);
    971 		break;
    972 
    973 	case PFSemul:
    974 		vap->va_bytes = vap->va_size = strlen(procp->p_emul->e_name);
    975 		break;
    976 
    977 #ifdef __HAVE_PROCFS_MACHDEP
    978 	PROCFS_MACHDEP_NODETYPE_CASES
    979 		error = procfs_machdep_getattr(ap->a_vp, vap, procp);
    980 		break;
    981 #endif
    982 
    983 	default:
    984 		panic("%s: %d/2", __func__, pfs->pfs_type);
    985 	}
    986 
    987 	if (procp != NULL)
    988 		procfs_proc_unlock(procp);
    989 	if (path != NULL)
    990 		free(path, M_TEMP);
    991 
    992 	return (error);
    993 }
    994 
    995 /*ARGSUSED*/
    996 int
    997 procfs_setattr(void *v)
    998 {
    999 	/*
   1000 	 * just fake out attribute setting
   1001 	 * it's not good to generate an error
   1002 	 * return, otherwise things like creat()
   1003 	 * will fail when they try to set the
   1004 	 * file length to 0.  worse, this means
   1005 	 * that echo $note > /proc/$pid/note will fail.
   1006 	 */
   1007 
   1008 	return (0);
   1009 }
   1010 
   1011 /*
   1012  * implement access checking.
   1013  *
   1014  * actually, the check for super-user is slightly
   1015  * broken since it will allow read access to write-only
   1016  * objects.  this doesn't cause any particular trouble
   1017  * but does mean that the i/o entry points need to check
   1018  * that the operation really does make sense.
   1019  */
   1020 int
   1021 procfs_access(void *v)
   1022 {
   1023 	struct vop_access_args /* {
   1024 		struct vnode *a_vp;
   1025 		accmode_t a_accmode;
   1026 		kauth_cred_t a_cred;
   1027 	} */ *ap = v;
   1028 	struct vattr va;
   1029 	int error;
   1030 
   1031 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred)) != 0)
   1032 		return (error);
   1033 
   1034 	return kauth_authorize_vnode(ap->a_cred,
   1035 	    KAUTH_ACCESS_ACTION(ap->a_accmode, ap->a_vp->v_type, va.va_mode),
   1036 	    ap->a_vp, NULL, genfs_can_access(ap->a_vp, ap->a_cred,
   1037 	    va.va_uid, va.va_gid, va.va_mode, NULL, ap->a_accmode));
   1038 }
   1039 
   1040 /*
   1041  * lookup.  this is incredibly complicated in the
   1042  * general case, however for most pseudo-filesystems
   1043  * very little needs to be done.
   1044  *
   1045  * Locking isn't hard here, just poorly documented.
   1046  *
   1047  * If we're looking up ".", just vref the parent & return it.
   1048  *
   1049  * If we're looking up "..", unlock the parent, and lock "..". If everything
   1050  * went ok, and we're on the last component and the caller requested the
   1051  * parent locked, try to re-lock the parent. We do this to prevent lock
   1052  * races.
   1053  *
   1054  * For anything else, get the needed node. Then unlock the parent if not
   1055  * the last component or not LOCKPARENT (i.e. if we wouldn't re-lock the
   1056  * parent in the .. case).
   1057  *
   1058  * We try to exit with the parent locked in error cases.
   1059  */
   1060 int
   1061 procfs_lookup(void *v)
   1062 {
   1063 	struct vop_lookup_v2_args /* {
   1064 		struct vnode * a_dvp;
   1065 		struct vnode ** a_vpp;
   1066 		struct componentname * a_cnp;
   1067 	} */ *ap = v;
   1068 	struct componentname *cnp = ap->a_cnp;
   1069 	struct vnode **vpp = ap->a_vpp;
   1070 	struct vnode *dvp = ap->a_dvp;
   1071 	const char *pname = cnp->cn_nameptr;
   1072 	const struct proc_target *pt = NULL;
   1073 	struct vnode *fvp;
   1074 	pid_t pid, vnpid;
   1075 	struct pfsnode *pfs;
   1076 	struct proc *p = NULL;
   1077 	struct lwp *plwp;
   1078 	int i, error;
   1079 	pfstype type;
   1080 
   1081 	*vpp = NULL;
   1082 
   1083 	if ((error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred)) != 0)
   1084 		return (error);
   1085 
   1086 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
   1087 		return (EROFS);
   1088 
   1089 	if (cnp->cn_namelen == 1 && *pname == '.') {
   1090 		*vpp = dvp;
   1091 		vref(dvp);
   1092 		return (0);
   1093 	}
   1094 
   1095 	pfs = VTOPFS(dvp);
   1096 	switch (pfs->pfs_type) {
   1097 	case PFSroot:
   1098 		/*
   1099 		 * Shouldn't get here with .. in the root node.
   1100 		 */
   1101 		if (cnp->cn_flags & ISDOTDOT)
   1102 			return (EIO);
   1103 
   1104 		for (i = 0; i < nproc_root_targets; i++) {
   1105 			pt = &proc_root_targets[i];
   1106 			/*
   1107 			 * check for node match.  proc is always NULL here,
   1108 			 * so call pt_valid with constant NULL lwp.
   1109 			 */
   1110 			if (cnp->cn_namelen == pt->pt_namlen &&
   1111 			    memcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
   1112 			    (pt->pt_valid == NULL ||
   1113 			     (*pt->pt_valid)(NULL, dvp->v_mount)))
   1114 				break;
   1115 		}
   1116 
   1117 		if (i != nproc_root_targets) {
   1118 			error = procfs_allocvp(dvp->v_mount, vpp, 0,
   1119 			    pt->pt_pfstype, -1);
   1120 			return (error);
   1121 		}
   1122 
   1123 		if (CNEQ(cnp, "curproc", 7)) {
   1124 			pid = curproc->p_pid;
   1125 			vnpid = 0;
   1126 			type = PFScurproc;
   1127 		} else if (CNEQ(cnp, "self", 4)) {
   1128 			pid = curproc->p_pid;
   1129 			vnpid = 0;
   1130 			type = PFSself;
   1131 		} else {
   1132 			pid = (pid_t)atoi(pname, cnp->cn_namelen);
   1133 			vnpid = pid;
   1134 			type = PFSproc;
   1135 		}
   1136 
   1137 		if (procfs_proc_lock(dvp->v_mount, pid, &p, ESRCH) != 0)
   1138 			break;
   1139 		error = procfs_allocvp(dvp->v_mount, vpp, vnpid, type, -1);
   1140 		procfs_proc_unlock(p);
   1141 		return (error);
   1142 
   1143 	case PFSproc:
   1144 		if (cnp->cn_flags & ISDOTDOT) {
   1145 			error = procfs_allocvp(dvp->v_mount, vpp, 0, PFSroot,
   1146 			    -1);
   1147 			return (error);
   1148 		}
   1149 
   1150 		if (procfs_proc_lock(dvp->v_mount, pfs->pfs_pid, &p,
   1151 				     ESRCH) != 0)
   1152 			break;
   1153 
   1154 		mutex_enter(p->p_lock);
   1155 		LIST_FOREACH(plwp, &p->p_lwps, l_sibling) {
   1156 			if (plwp->l_stat != LSZOMB)
   1157 				break;
   1158 		}
   1159 		/* Process is exiting if no-LWPS or all LWPs are LSZOMB */
   1160 		if (plwp == NULL) {
   1161 			mutex_exit(p->p_lock);
   1162 			procfs_proc_unlock(p);
   1163 			return ESRCH;
   1164 		}
   1165 
   1166 		lwp_addref(plwp);
   1167 		mutex_exit(p->p_lock);
   1168 
   1169 		for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) {
   1170 			int found;
   1171 
   1172 			found = cnp->cn_namelen == pt->pt_namlen &&
   1173 			    memcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
   1174 			    (pt->pt_valid == NULL
   1175 			      || (*pt->pt_valid)(plwp, dvp->v_mount));
   1176 			if (found)
   1177 				break;
   1178 		}
   1179 		lwp_delref(plwp);
   1180 
   1181 		if (i == nproc_targets) {
   1182 			procfs_proc_unlock(p);
   1183 			break;
   1184 		}
   1185 		if (pt->pt_pfstype == PFSfile) {
   1186 			fvp = p->p_textvp;
   1187 			/* We already checked that it exists. */
   1188 			vref(fvp);
   1189 			procfs_proc_unlock(p);
   1190 			*vpp = fvp;
   1191 			return (0);
   1192 		}
   1193 
   1194 		error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
   1195 		    pt->pt_pfstype, -1);
   1196 		procfs_proc_unlock(p);
   1197 		return (error);
   1198 
   1199 	case PFSfd: {
   1200 		int fd;
   1201 		file_t *fp;
   1202 
   1203 		if ((error = procfs_proc_lock(dvp->v_mount, pfs->pfs_pid, &p,
   1204 					      ENOENT)) != 0)
   1205 			return error;
   1206 
   1207 		if (cnp->cn_flags & ISDOTDOT) {
   1208 			error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
   1209 			    PFSproc, -1);
   1210 			procfs_proc_unlock(p);
   1211 			return (error);
   1212 		}
   1213 		fd = atoi(pname, cnp->cn_namelen);
   1214 
   1215 		fp = fd_getfile2(p, fd);
   1216 		if (fp == NULL) {
   1217 			procfs_proc_unlock(p);
   1218 			return ENOENT;
   1219 		}
   1220 		fvp = fp->f_vnode;
   1221 
   1222 		/* Don't show directories */
   1223 		if (fp->f_type == DTYPE_VNODE && fvp->v_type != VDIR &&
   1224 		    !procfs_proc_is_linux_compat()) {
   1225 			vref(fvp);
   1226 			closef(fp);
   1227 			procfs_proc_unlock(p);
   1228 			*vpp = fvp;
   1229 			return 0;
   1230 		}
   1231 
   1232 		closef(fp);
   1233 		error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
   1234 		    PFSfd, fd);
   1235 		procfs_proc_unlock(p);
   1236 		return error;
   1237 	}
   1238 	case PFStask: {
   1239 		int xpid;
   1240 
   1241 		if ((error = procfs_proc_lock(dvp->v_mount, pfs->pfs_pid, &p,
   1242 					      ENOENT)) != 0)
   1243 			return error;
   1244 
   1245 		if (cnp->cn_flags & ISDOTDOT) {
   1246 			error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
   1247 			    PFSproc, -1);
   1248 			procfs_proc_unlock(p);
   1249 			return (error);
   1250 		}
   1251 		xpid = atoi(pname, cnp->cn_namelen);
   1252 
   1253 		if (xpid != pfs->pfs_pid) {
   1254 			procfs_proc_unlock(p);
   1255 			return ENOENT;
   1256 		}
   1257 		error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
   1258 		    PFStask, 0);
   1259 		procfs_proc_unlock(p);
   1260 		return error;
   1261 	}
   1262 	case PFSsys:
   1263 	case PFSsysfs:
   1264 	case PFSmqueue:
   1265 	case PFSsysvipc: {
   1266 		const struct proc_target *targets;
   1267 		int ntargets;
   1268 		pfstype parent;
   1269 
   1270 		switch (pfs->pfs_type) {
   1271 		case PFSsys:
   1272 			targets = proc_sys_targets;
   1273 			ntargets = nproc_sys_targets;
   1274 			parent = PFSroot;
   1275 			break;
   1276 		case PFSsysfs:
   1277 			targets = proc_sysfs_targets;
   1278 			ntargets = nproc_sysfs_targets;
   1279 			parent = PFSsys;
   1280 			break;
   1281 		case PFSmqueue:
   1282 			targets = proc_mqueue_targets;
   1283 			ntargets = nproc_mqueue_targets;
   1284 			parent = PFSsysfs;
   1285 			break;
   1286 		case PFSsysvipc:
   1287 			targets = proc_sysvipc_targets;
   1288 			ntargets = nproc_sysvipc_targets;
   1289 			parent = PFSroot;
   1290 			break;
   1291 		default:
   1292 			return (EINVAL);
   1293 		}
   1294 
   1295 		if (cnp->cn_flags & ISDOTDOT) {
   1296 			error = procfs_allocvp(dvp->v_mount, vpp, 0, parent,
   1297 			    -1);
   1298 			return (error);
   1299 		}
   1300 
   1301 		for (i = 0; i < ntargets; i++) {
   1302 			pt = &targets[i];
   1303 			/*
   1304 			 * check for node match.  proc is always NULL here,
   1305 			 * so call pt_valid with constant NULL lwp.
   1306 			 */
   1307 			if (cnp->cn_namelen == pt->pt_namlen &&
   1308 			    memcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
   1309 			    (pt->pt_valid == NULL ||
   1310 			     (*pt->pt_valid)(NULL, dvp->v_mount)))
   1311 				break;
   1312 		}
   1313 
   1314 		if (i != ntargets) {
   1315 			error = procfs_allocvp(dvp->v_mount, vpp, 0,
   1316 			    pt->pt_pfstype, -1);
   1317 			return (error);
   1318 		}
   1319 
   1320 		return (ENOENT);
   1321 	}
   1322 
   1323 	default:
   1324 		return (ENOTDIR);
   1325 	}
   1326 
   1327 	return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS);
   1328 }
   1329 
   1330 int
   1331 procfs_validfile(struct lwp *l, struct mount *mp)
   1332 {
   1333 	return l != NULL && l->l_proc != NULL && l->l_proc->p_textvp != NULL;
   1334 }
   1335 
   1336 static int
   1337 procfs_validfile_linux(struct lwp *l, struct mount *mp)
   1338 {
   1339 	return procfs_use_linux_compat(mp) &&
   1340 	    (l == NULL || l->l_proc == NULL || procfs_validfile(l, mp));
   1341 }
   1342 
   1343 struct procfs_root_readdir_ctx {
   1344 	struct uio *uiop;
   1345 	off_t *cookies;
   1346 	int ncookies;
   1347 	off_t off;
   1348 	off_t startoff;
   1349 	int error;
   1350 };
   1351 
   1352 static int
   1353 procfs_root_readdir_callback(struct proc *p, void *arg)
   1354 {
   1355 	struct procfs_root_readdir_ctx *ctxp = arg;
   1356 	struct dirent d;
   1357 	struct uio *uiop;
   1358 	int error;
   1359 
   1360 	uiop = ctxp->uiop;
   1361 	if (uiop->uio_resid < UIO_MX)
   1362 		return -1; /* no space */
   1363 
   1364 	if (kauth_authorize_process(kauth_cred_get(),
   1365 	    KAUTH_PROCESS_CANSEE, p,
   1366 	    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL) != 0)
   1367 		return 0;
   1368 
   1369 	if (ctxp->off < ctxp->startoff) {
   1370 		ctxp->off++;
   1371 		return 0;
   1372 	}
   1373 
   1374 	memset(&d, 0, UIO_MX);
   1375 	d.d_reclen = UIO_MX;
   1376 	d.d_fileno = PROCFS_FILENO(p->p_pid, PFSproc, -1);
   1377 	d.d_namlen = snprintf(d.d_name,
   1378 	    UIO_MX - offsetof(struct dirent, d_name), "%ld", (long)p->p_pid);
   1379 	d.d_type = DT_DIR;
   1380 
   1381 	mutex_exit(&proc_lock);
   1382 	error = uiomove(&d, UIO_MX, uiop);
   1383 	mutex_enter(&proc_lock);
   1384 	if (error) {
   1385 		ctxp->error = error;
   1386 		return -1;
   1387 	}
   1388 
   1389 	ctxp->ncookies++;
   1390 	if (ctxp->cookies)
   1391 		*(ctxp->cookies)++ = ctxp->off + 1;
   1392 	ctxp->off++;
   1393 
   1394 	return 0;
   1395 }
   1396 
   1397 /*
   1398  * readdir returns directory entries from pfsnode (vp).
   1399  *
   1400  * the strategy here with procfs is to generate a single
   1401  * directory entry at a time (struct dirent) and then
   1402  * copy that out to userland using uiomove.  a more efficient
   1403  * though more complex implementation, would try to minimize
   1404  * the number of calls to uiomove().  for procfs, this is
   1405  * hardly worth the added code complexity.
   1406  *
   1407  * this should just be done through read()
   1408  */
   1409 int
   1410 procfs_readdir(void *v)
   1411 {
   1412 	struct vop_readdir_args /* {
   1413 		struct vnode *a_vp;
   1414 		struct uio *a_uio;
   1415 		kauth_cred_t a_cred;
   1416 		int *a_eofflag;
   1417 		off_t **a_cookies;
   1418 		int *a_ncookies;
   1419 	} */ *ap = v;
   1420 	struct uio *uio = ap->a_uio;
   1421 	struct dirent d;
   1422 	struct pfsnode *pfs;
   1423 	off_t i;
   1424 	int error;
   1425 	off_t *cookies = NULL;
   1426 	int ncookies;
   1427 	struct vnode *vp;
   1428 	const struct proc_target *pt;
   1429 	struct procfs_root_readdir_ctx ctx;
   1430 	struct proc *p = NULL;
   1431 	struct lwp *l;
   1432 	int nfd;
   1433 	int nc = 0;
   1434 
   1435 	vp = ap->a_vp;
   1436 	pfs = VTOPFS(vp);
   1437 
   1438 	if (uio->uio_resid < UIO_MX)
   1439 		return (EINVAL);
   1440 	if (uio->uio_offset < 0)
   1441 		return (EINVAL);
   1442 
   1443 	error = 0;
   1444 	i = uio->uio_offset;
   1445 	memset(&d, 0, UIO_MX);
   1446 	d.d_reclen = UIO_MX;
   1447 	ncookies = uio->uio_resid / UIO_MX;
   1448 
   1449 	switch (pfs->pfs_type) {
   1450 	/*
   1451 	 * this is for the process-specific sub-directories.
   1452 	 * all that is needed to is copy out all the entries
   1453 	 * from the procent[] table (top of this file).
   1454 	 */
   1455 	case PFSproc: {
   1456 
   1457 		if (i >= nproc_targets)
   1458 			return 0;
   1459 
   1460 		if (procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &p, ESRCH) != 0)
   1461 			break;
   1462 
   1463 		if (ap->a_ncookies) {
   1464 			ncookies = uimin(ncookies, (nproc_targets - i));
   1465 			cookies = malloc(ncookies * sizeof (off_t),
   1466 			    M_TEMP, M_WAITOK);
   1467 			*ap->a_cookies = cookies;
   1468 		}
   1469 
   1470 		for (pt = &proc_targets[i];
   1471 		     uio->uio_resid >= UIO_MX && i < nproc_targets; pt++, i++) {
   1472 			if (pt->pt_valid) {
   1473 				/* XXXSMP LWP can disappear */
   1474 				mutex_enter(p->p_lock);
   1475 				l = LIST_FIRST(&p->p_lwps);
   1476 				KASSERT(l != NULL);
   1477 				mutex_exit(p->p_lock);
   1478 				if ((*pt->pt_valid)(l, vp->v_mount) == 0)
   1479 					continue;
   1480 			}
   1481 
   1482 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid,
   1483 			    pt->pt_pfstype, -1);
   1484 			d.d_namlen = pt->pt_namlen;
   1485 			memcpy(d.d_name, pt->pt_name, pt->pt_namlen + 1);
   1486 			d.d_type = pt->pt_type;
   1487 
   1488 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
   1489 				break;
   1490 			if (cookies)
   1491 				*cookies++ = i + 1;
   1492 		}
   1493 
   1494 		procfs_proc_unlock(p);
   1495 	    	break;
   1496 	}
   1497 	case PFSfd: {
   1498 		file_t *fp;
   1499 		int lim;
   1500 
   1501 		if ((error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &p,
   1502 					      ESRCH)) != 0)
   1503 			return error;
   1504 
   1505 		/* XXX Should this be by file as well? */
   1506 		if (kauth_authorize_process(kauth_cred_get(),
   1507 		    KAUTH_PROCESS_CANSEE, p,
   1508 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_OPENFILES), NULL,
   1509 		    NULL) != 0) {
   1510 		    	procfs_proc_unlock(p);
   1511 			return ESRCH;
   1512 		}
   1513 
   1514 		nfd = atomic_load_consume(&p->p_fd->fd_dt)->dt_nfiles;
   1515 
   1516 		lim = uimin((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
   1517 		if (i >= lim) {
   1518 		    	procfs_proc_unlock(p);
   1519 			return 0;
   1520 		}
   1521 
   1522 		if (ap->a_ncookies) {
   1523 			ncookies = uimin(ncookies, (nfd + 2 - i));
   1524 			cookies = malloc(ncookies * sizeof (off_t),
   1525 			    M_TEMP, M_WAITOK);
   1526 			*ap->a_cookies = cookies;
   1527 		}
   1528 
   1529 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
   1530 			pt = &proc_targets[i];
   1531 			d.d_namlen = pt->pt_namlen;
   1532 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid,
   1533 			    pt->pt_pfstype, -1);
   1534 			(void)memcpy(d.d_name, pt->pt_name, pt->pt_namlen + 1);
   1535 			d.d_type = pt->pt_type;
   1536 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
   1537 				break;
   1538 			if (cookies)
   1539 				*cookies++ = i + 1;
   1540 			nc++;
   1541 		}
   1542 		if (error)
   1543 			goto out;
   1544 		for (; uio->uio_resid >= UIO_MX && i < nfd; i++) {
   1545 			/* check the descriptor exists */
   1546 			if ((fp = fd_getfile2(p, i - 2)) == NULL)
   1547 				continue;
   1548 			closef(fp);
   1549 
   1550 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid, PFSfd, i - 2);
   1551 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
   1552 			    "%lld", (long long)(i - 2));
   1553 			d.d_type = fttodt(fp);
   1554 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
   1555 				break;
   1556 			if (cookies)
   1557 				*cookies++ = i + 1;
   1558 			nc++;
   1559 		}
   1560 		goto out;
   1561 	}
   1562 	case PFStask: {
   1563 
   1564 		if ((error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &p,
   1565 					      ESRCH)) != 0)
   1566 			return error;
   1567 
   1568 		nfd = 3;	/* ., .., pid */
   1569 
   1570 		if (ap->a_ncookies) {
   1571 			ncookies = uimin(ncookies, (nfd + 2 - i));
   1572 			cookies = malloc(ncookies * sizeof (off_t),
   1573 			    M_TEMP, M_WAITOK);
   1574 			*ap->a_cookies = cookies;
   1575 		}
   1576 
   1577 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
   1578 			pt = &proc_targets[i];
   1579 			d.d_namlen = pt->pt_namlen;
   1580 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid,
   1581 			    pt->pt_pfstype, -1);
   1582 			(void)memcpy(d.d_name, pt->pt_name, pt->pt_namlen + 1);
   1583 			d.d_type = pt->pt_type;
   1584 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
   1585 				break;
   1586 			if (cookies)
   1587 				*cookies++ = i + 1;
   1588 			nc++;
   1589 		}
   1590 		if (error)
   1591 			goto out;
   1592 		for (; uio->uio_resid >= UIO_MX && i < nfd; i++) {
   1593 			/* check the descriptor exists */
   1594 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid, PFStask,
   1595 			    i - 2);
   1596 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
   1597 			    "%ld", (long)pfs->pfs_pid);
   1598 			d.d_type = DT_LNK;
   1599 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
   1600 				break;
   1601 			if (cookies)
   1602 				*cookies++ = i + 1;
   1603 			nc++;
   1604 		}
   1605 		goto out;
   1606 	}
   1607 
   1608 	/*
   1609 	 * misc subdirectories
   1610 	 */
   1611 	case PFSsys:
   1612 	case PFSsysfs:
   1613 	case PFSmqueue:
   1614 	case PFSsysvipc: {
   1615 		const struct proc_target *targets;
   1616 		int ntargets;
   1617 
   1618 		switch (pfs->pfs_type) {
   1619 		case PFSsys:
   1620 			targets = proc_sys_targets;
   1621 			ntargets = nproc_sys_targets;
   1622 			break;
   1623 		case PFSsysfs:
   1624 			targets = proc_sysfs_targets;
   1625 			ntargets = nproc_sysfs_targets;
   1626 			break;
   1627 		case PFSmqueue:
   1628 			targets = proc_mqueue_targets;
   1629 			ntargets = nproc_mqueue_targets;
   1630 			break;
   1631 		case PFSsysvipc:
   1632 			targets = proc_sysvipc_targets;
   1633 			ntargets = nproc_sysvipc_targets;
   1634 			break;
   1635 		default:
   1636 			return (EINVAL);
   1637 		}
   1638 
   1639 		if ((error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &p,
   1640 					      ESRCH)) != 0)
   1641 			return error;
   1642 		if (ap->a_ncookies) {
   1643 			ncookies = uimin(ncookies, (ntargets - i));
   1644 			cookies = malloc(ncookies * sizeof (off_t),
   1645 			    M_TEMP, M_WAITOK);
   1646 			*ap->a_cookies = cookies;
   1647 		}
   1648 
   1649 		for (pt = &targets[i];
   1650 		     uio->uio_resid >= UIO_MX && i < ntargets; pt++, i++) {
   1651 			if (pt->pt_valid &&
   1652 			    (*pt->pt_valid)(NULL, vp->v_mount) == 0)
   1653 				continue;
   1654 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid,
   1655 			    pt->pt_pfstype, -1);
   1656 			d.d_namlen = pt->pt_namlen;
   1657 			memcpy(d.d_name, pt->pt_name, pt->pt_namlen + 1);
   1658 			d.d_type = pt->pt_type;
   1659 
   1660 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
   1661 				break;
   1662 			if (cookies)
   1663 				*cookies++ = i + 1;
   1664 		}
   1665 
   1666 		goto out;
   1667 	}
   1668 
   1669 	/*
   1670 	 * this is for the root of the procfs filesystem
   1671 	 * what is needed are special entries for "curproc"
   1672 	 * and "self" followed by an entry for each process
   1673 	 * on allproc.
   1674 	 */
   1675 
   1676 	case PFSroot: {
   1677 
   1678 		if (ap->a_ncookies) {
   1679 			/*
   1680 			 * XXX Potentially allocating too much space here,
   1681 			 * but I'm lazy. This loop needs some work.
   1682 			 */
   1683 			cookies = malloc(ncookies * sizeof (off_t),
   1684 			    M_TEMP, M_WAITOK);
   1685 			*ap->a_cookies = cookies;
   1686 		}
   1687 
   1688 		/* 0 ... 3 are static entries. */
   1689 		for (; i <= 3 && uio->uio_resid >= UIO_MX; i++) {
   1690 			switch (i) {
   1691 			case 0:		/* `.' */
   1692 			case 1:		/* `..' */
   1693 				d.d_fileno = PROCFS_FILENO(0, PFSroot, -1);
   1694 				d.d_namlen = i + 1;
   1695 				memcpy(d.d_name, "..", d.d_namlen);
   1696 				d.d_name[i + 1] = '\0';
   1697 				d.d_type = DT_DIR;
   1698 				break;
   1699 
   1700 			case 2:
   1701 				d.d_fileno = PROCFS_FILENO(0, PFScurproc, -1);
   1702 				d.d_namlen = sizeof("curproc") - 1;
   1703 				memcpy(d.d_name, "curproc", sizeof("curproc"));
   1704 				d.d_type = DT_LNK;
   1705 				break;
   1706 
   1707 			case 3:
   1708 				d.d_fileno = PROCFS_FILENO(0, PFSself, -1);
   1709 				d.d_namlen = sizeof("self") - 1;
   1710 				memcpy(d.d_name, "self", sizeof("self"));
   1711 				d.d_type = DT_LNK;
   1712 				break;
   1713 			}
   1714 
   1715 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
   1716 				break;
   1717 			nc++;
   1718 			if (cookies)
   1719 				*cookies++ = i + 1;
   1720 		}
   1721 		if (error)
   1722 			break;
   1723 		/* 4 ... are process entries. */
   1724 		ctx.uiop = uio;
   1725 		ctx.error = 0;
   1726 		ctx.off = 4;
   1727 		ctx.startoff = i;
   1728 		ctx.cookies = cookies;
   1729 		ctx.ncookies = nc;
   1730 		proclist_foreach_call(&allproc,
   1731 		    procfs_root_readdir_callback, &ctx);
   1732 		cookies = ctx.cookies;
   1733 		nc = ctx.ncookies;
   1734 		error = ctx.error;
   1735 		if (error)
   1736 			break;
   1737 
   1738 		/* misc entries. */
   1739 		if (i < ctx.off)
   1740 			i = ctx.off;
   1741 		if (i >= ctx.off + nproc_root_targets)
   1742 			break;
   1743 		error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &p, ESRCH);
   1744 		if (error)
   1745 			break;
   1746 		for (pt = &proc_root_targets[i - ctx.off];
   1747 		    uio->uio_resid >= UIO_MX &&
   1748 		    pt < &proc_root_targets[nproc_root_targets];
   1749 		    pt++, i++) {
   1750 			if (pt->pt_valid &&
   1751 			    (*pt->pt_valid)(NULL, vp->v_mount) == 0)
   1752 				continue;
   1753 			if (kauth_authorize_process(kauth_cred_get(),
   1754 			    KAUTH_PROCESS_CANSEE, p,
   1755 			    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY),
   1756 			    NULL, NULL) != 0)
   1757 				continue;
   1758 			d.d_fileno = PROCFS_FILENO(0, pt->pt_pfstype, -1);
   1759 			d.d_namlen = pt->pt_namlen;
   1760 			memcpy(d.d_name, pt->pt_name, pt->pt_namlen + 1);
   1761 			d.d_type = pt->pt_type;
   1762 
   1763 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
   1764 				break;
   1765 			nc++;
   1766 			if (cookies)
   1767 				*cookies++ = i + 1;
   1768 		}
   1769 out:
   1770 		KASSERT(p != NULL);
   1771 		ncookies = nc;
   1772 		procfs_proc_unlock(p);
   1773 		break;
   1774 	}
   1775 
   1776 	default:
   1777 		error = ENOTDIR;
   1778 		break;
   1779 	}
   1780 
   1781 	if (ap->a_ncookies) {
   1782 		if (error) {
   1783 			if (cookies)
   1784 				free(*ap->a_cookies, M_TEMP);
   1785 			*ap->a_ncookies = 0;
   1786 			*ap->a_cookies = NULL;
   1787 		} else
   1788 			*ap->a_ncookies = ncookies;
   1789 	}
   1790 	uio->uio_offset = i;
   1791 	return (error);
   1792 }
   1793 
   1794 /*
   1795  * readlink reads the link of `curproc' and others
   1796  */
   1797 int
   1798 procfs_readlink(void *v)
   1799 {
   1800 	struct vop_readlink_args *ap = v;
   1801 	char bf[16];		/* should be enough */
   1802 	char *bp = bf;
   1803 	char *path = NULL;
   1804 	int len = 0;
   1805 	int error = 0;
   1806 	struct vnode *vp = ap->a_vp;
   1807 	struct pfsnode *pfs = VTOPFS(vp);
   1808 	struct proc *pown = NULL;
   1809 
   1810 	if (pfs->pfs_fileno == PROCFS_FILENO(0, PFScurproc, -1))
   1811 		len = snprintf(bf, sizeof(bf), "%ld", (long)curproc->p_pid);
   1812 	else if (pfs->pfs_fileno == PROCFS_FILENO(0, PFSself, -1))
   1813 		len = snprintf(bf, sizeof(bf), "%s", "curproc");
   1814 	else if (pfs->pfs_fileno == PROCFS_FILENO(pfs->pfs_pid, PFStask, 0))
   1815 		len = snprintf(bf, sizeof(bf), "..");
   1816 	else if (pfs->pfs_fileno == PROCFS_FILENO(pfs->pfs_pid, PFSexe, -1)) {
   1817 		if ((error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &pown,
   1818 					      ESRCH)) != 0)
   1819 			return error;
   1820 		bp = pown->p_path;
   1821 		len = strlen(bp);
   1822 	} else if (pfs->pfs_fileno == PROCFS_FILENO(pfs->pfs_pid, PFScwd, -1) ||
   1823 	    pfs->pfs_fileno == PROCFS_FILENO(pfs->pfs_pid, PFSchroot, -1)) {
   1824 		if ((error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &pown,
   1825 					      ESRCH)) != 0)
   1826 			return error;
   1827 		path = malloc(MAXPATHLEN + 4, M_TEMP, M_WAITOK);
   1828 		if (path == NULL) {
   1829 			procfs_proc_unlock(pown);
   1830 			return (ENOMEM);
   1831 		}
   1832 		bp = path + MAXPATHLEN;
   1833 		*--bp = '\0';
   1834 		procfs_dir(PROCFS_TYPE(pfs->pfs_fileno), curlwp, pown,
   1835 		    &bp, path, MAXPATHLEN);
   1836 		len = strlen(bp);
   1837 	} else {
   1838 		file_t *fp;
   1839 		struct vnode *vxp;
   1840 
   1841 		if ((error = procfs_proc_lock(vp->v_mount, pfs->pfs_pid, &pown,
   1842 					      ESRCH)) != 0)
   1843 			return error;
   1844 
   1845 		fp = fd_getfile2(pown, pfs->pfs_fd);
   1846 		if (fp == NULL) {
   1847 			procfs_proc_unlock(pown);
   1848 			return EBADF;
   1849 		}
   1850 
   1851 		switch (fp->f_type) {
   1852 		case DTYPE_VNODE:
   1853 			vxp = fp->f_vnode;
   1854 			if (vxp->v_type != VDIR &&
   1855 			    !procfs_proc_is_linux_compat()) {
   1856 				error = EINVAL;
   1857 				break;
   1858 			}
   1859 			if ((path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK))
   1860 			    == NULL) {
   1861 				error = ENOMEM;
   1862 				break;
   1863 			}
   1864 			bp = path + MAXPATHLEN;
   1865 			*--bp = '\0';
   1866 
   1867 			/*
   1868 			 * XXX: kludge to avoid locking against ourselves
   1869 			 * in getcwd()
   1870 			 */
   1871 			if (vxp->v_tag == VT_PROCFS) {
   1872 				*--bp = '/';
   1873 			} else {
   1874 				rw_enter(&curproc->p_cwdi->cwdi_lock,
   1875 				    RW_READER);
   1876 				vp = curproc->p_cwdi->cwdi_rdir;
   1877 				if (vp == NULL)
   1878 					vp = rootvnode;
   1879 				error = getcwd_common(vxp, vp, &bp, path,
   1880 				    MAXPATHLEN / 2, 0, curlwp);
   1881 				rw_exit(&curproc->p_cwdi->cwdi_lock);
   1882 			}
   1883 			if (error)
   1884 				break;
   1885 			len = strlen(bp);
   1886 			break;
   1887 
   1888 		case DTYPE_MISC:
   1889 			len = snprintf(bf, sizeof(bf), "%s", "[misc]");
   1890 			break;
   1891 
   1892 		case DTYPE_KQUEUE:
   1893 			len = snprintf(bf, sizeof(bf), "%s", "[kqueue]");
   1894 			break;
   1895 
   1896 		case DTYPE_SEM:
   1897 			len = snprintf(bf, sizeof(bf), "%s", "[ksem]");
   1898 			break;
   1899 
   1900 		default:
   1901 			error = EINVAL;
   1902 			break;
   1903 		}
   1904 		closef(fp);
   1905 	}
   1906 
   1907 	if (error == 0)
   1908 		error = uiomove(bp, len, ap->a_uio);
   1909 	if (pown)
   1910 		procfs_proc_unlock(pown);
   1911 	if (path)
   1912 		free(path, M_TEMP);
   1913 	return error;
   1914 }
   1915 
   1916 int
   1917 procfs_getpages(void *v)
   1918 {
   1919 	struct vop_getpages_args /* {
   1920 		struct vnode *a_vp;
   1921 		voff_t a_offset;
   1922 		struct vm_page **a_m;
   1923 		int *a_count;
   1924 		int a_centeridx;
   1925 		vm_prot_t a_access_type;
   1926 		int a_advice;
   1927 		int a_flags;
   1928 	} */ *ap = v;
   1929 
   1930 	if ((ap->a_flags & PGO_LOCKED) == 0)
   1931 		rw_exit(ap->a_vp->v_uobj.vmobjlock);
   1932 
   1933 	return (EFAULT);
   1934 }
   1935 
   1936 /*
   1937  * convert decimal ascii to int
   1938  */
   1939 static int
   1940 atoi(const char *b, size_t len)
   1941 {
   1942 	int p = 0;
   1943 
   1944 	while (len--) {
   1945 		char c = *b++;
   1946 		if (c < '0' || c > '9')
   1947 			return -1;
   1948 		p = 10 * p + (c - '0');
   1949 	}
   1950 
   1951 	return p;
   1952 }
   1953 
   1954 /**
   1955  * convert DTYPE_XXX to corresponding DT_XXX
   1956  * matching what procfs_loadvnode() does.
   1957  */
   1958 static uint8_t
   1959 fttodt(file_t *fp)
   1960 {
   1961 	switch (fp->f_type) {
   1962 	case DTYPE_VNODE:
   1963 		switch (fp->f_vnode->v_type) {
   1964 		case VREG:	return DT_REG;
   1965 		case VDIR:	return DT_LNK;	/* symlink */
   1966 		case VBLK:	return DT_BLK;
   1967 		case VCHR:	return DT_CHR;
   1968 		case VLNK:	return DT_LNK;
   1969 		case VSOCK:	return DT_SOCK;
   1970 		case VFIFO:	return DT_FIFO;
   1971 		default:	return DT_UNKNOWN;
   1972 		}
   1973 	case DTYPE_PIPE:	return DT_FIFO;
   1974 	case DTYPE_SOCKET:	return DT_SOCK;
   1975 	case DTYPE_KQUEUE:	/*FALLTHROUGH*/
   1976 	case DTYPE_MISC:	/*FALLTHROUGH*/
   1977 	case DTYPE_SEM:		return DT_LNK;	/* symlinks */
   1978 	default:		return DT_UNKNOWN;
   1979 	}
   1980 }
   1981