Home | History | Annotate | Line # | Download | only in procfs
procfs_subr.c revision 1.31.4.1
      1 /*	$NetBSD: procfs_subr.c,v 1.31.4.1 2001/03/30 21:49:23 he Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1994 Christopher G. Demetriou.  All rights reserved.
      5  * Copyright (c) 1993 Jan-Simon Pendry
      6  * Copyright (c) 1993
      7  *	The Regents of the University of California.  All rights reserved.
      8  *
      9  * This code is derived from software contributed to Berkeley by
     10  * Jan-Simon Pendry.
     11  *
     12  * Redistribution and use in source and binary forms, with or without
     13  * modification, are permitted provided that the following conditions
     14  * are met:
     15  * 1. Redistributions of source code must retain the above copyright
     16  *    notice, this list of conditions and the following disclaimer.
     17  * 2. Redistributions in binary form must reproduce the above copyright
     18  *    notice, this list of conditions and the following disclaimer in the
     19  *    documentation and/or other materials provided with the distribution.
     20  * 3. All advertising materials mentioning features or use of this software
     21  *    must display the following acknowledgement:
     22  *	This product includes software developed by the University of
     23  *	California, Berkeley and its contributors.
     24  * 4. Neither the name of the University nor the names of its contributors
     25  *    may be used to endorse or promote products derived from this software
     26  *    without specific prior written permission.
     27  *
     28  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     29  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     30  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     31  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     32  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     33  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     34  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     35  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     36  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     37  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     38  * SUCH DAMAGE.
     39  *
     40  *	@(#)procfs_subr.c	8.6 (Berkeley) 5/14/95
     41  */
     42 
     43 #include <sys/param.h>
     44 #include <sys/systm.h>
     45 #include <sys/time.h>
     46 #include <sys/kernel.h>
     47 #include <sys/proc.h>
     48 #include <sys/vnode.h>
     49 #include <sys/malloc.h>
     50 #include <sys/stat.h>
     51 
     52 #include <miscfs/procfs/procfs.h>
     53 
     54 void procfs_hashins __P((struct pfsnode *));
     55 void procfs_hashrem __P((struct pfsnode *));
     56 struct vnode *procfs_hashget __P((pid_t, pfstype, struct mount *));
     57 
     58 LIST_HEAD(pfs_hashhead, pfsnode) *pfs_hashtbl;
     59 u_long	pfs_ihash;	/* size of hash table - 1 */
     60 #define PFSPIDHASH(pid)	(&pfs_hashtbl[(pid) & pfs_ihash])
     61 
     62 struct lock pfs_hashlock;
     63 struct simplelock pfs_hash_slock;
     64 
     65 #define	ISSET(t, f)	((t) & (f))
     66 
     67 /*
     68  * allocate a pfsnode/vnode pair.  the vnode is
     69  * referenced, and locked.
     70  *
     71  * the pid, pfs_type, and mount point uniquely
     72  * identify a pfsnode.  the mount point is needed
     73  * because someone might mount this filesystem
     74  * twice.
     75  *
     76  * all pfsnodes are maintained on a singly-linked
     77  * list.  new nodes are only allocated when they cannot
     78  * be found on this list.  entries on the list are
     79  * removed when the vfs reclaim entry is called.
     80  *
     81  * a single lock is kept for the entire list.  this is
     82  * needed because the getnewvnode() function can block
     83  * waiting for a vnode to become free, in which case there
     84  * may be more than one process trying to get the same
     85  * vnode.  this lock is only taken if we are going to
     86  * call getnewvnode, since the kernel itself is single-threaded.
     87  *
     88  * if an entry is found on the list, then call vget() to
     89  * take a reference.  this is done because there may be
     90  * zero references to it and so it needs to removed from
     91  * the vnode free list.
     92  */
     93 int
     94 procfs_allocvp(mp, vpp, pid, pfs_type)
     95 	struct mount *mp;
     96 	struct vnode **vpp;
     97 	long pid;
     98 	pfstype pfs_type;
     99 {
    100 	struct pfsnode *pfs;
    101 	struct vnode *vp;
    102 	int error;
    103 
    104 	do {
    105 		if ((*vpp = procfs_hashget(pid, pfs_type, mp)) != NULL)
    106 			return (0);
    107 	} while (lockmgr(&pfs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
    108 
    109 	if ((error = getnewvnode(VT_PROCFS, mp, procfs_vnodeop_p, vpp)) != 0) {
    110 		*vpp = NULL;
    111 		lockmgr(&pfs_hashlock, LK_RELEASE, 0);
    112 		return (error);
    113 	}
    114 	vp = *vpp;
    115 
    116 	MALLOC(pfs, void *, sizeof(struct pfsnode), M_TEMP, M_WAITOK);
    117 	vp->v_data = pfs;
    118 
    119 	pfs->pfs_pid = (pid_t) pid;
    120 	pfs->pfs_type = pfs_type;
    121 	pfs->pfs_vnode = vp;
    122 	pfs->pfs_flags = 0;
    123 	pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type);
    124 
    125 	switch (pfs_type) {
    126 	case Proot:	/* /proc = dr-xr-xr-x */
    127 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
    128 		vp->v_type = VDIR;
    129 		vp->v_flag = VROOT;
    130 		break;
    131 
    132 	case Pcurproc:	/* /proc/curproc = lr-xr-xr-x */
    133 	case Pself:	/* /proc/self    = lr-xr-xr-x */
    134 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
    135 		vp->v_type = VLNK;
    136 		break;
    137 
    138 	case Pproc:	/* /proc/N = dr-xr-xr-x */
    139 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
    140 		vp->v_type = VDIR;
    141 		break;
    142 
    143 	case Pfile:	/* /proc/N/file = -rw------- */
    144 	case Pmem:	/* /proc/N/mem = -rw------- */
    145 	case Pregs:	/* /proc/N/regs = -rw------- */
    146 	case Pfpregs:	/* /proc/N/fpregs = -rw------- */
    147 		pfs->pfs_mode = S_IRUSR|S_IWUSR;
    148 		vp->v_type = VREG;
    149 		break;
    150 
    151 	case Pctl:	/* /proc/N/ctl = --w------ */
    152 	case Pnote:	/* /proc/N/note = --w------ */
    153 	case Pnotepg:	/* /proc/N/notepg = --w------ */
    154 		pfs->pfs_mode = S_IWUSR;
    155 		vp->v_type = VREG;
    156 		break;
    157 
    158 	case Pmap:	/* /proc/N/map = -r--r--r-- */
    159 	case Pstatus:	/* /proc/N/status = -r--r--r-- */
    160 	case Pcmdline:	/* /proc/N/cmdline = -r--r--r-- */
    161 	case Pmeminfo:	/* /proc/meminfo = -r--r--r-- */
    162 	case Pcpuinfo:	/* /proc/cpuinfo = -r--r--r-- */
    163 		pfs->pfs_mode = S_IRUSR|S_IRGRP|S_IROTH;
    164 		vp->v_type = VREG;
    165 		break;
    166 
    167 	default:
    168 		panic("procfs_allocvp");
    169 	}
    170 
    171 	procfs_hashins(pfs);
    172 	lockmgr(&pfs_hashlock, LK_RELEASE, 0);
    173 
    174 	return (error);
    175 }
    176 
    177 int
    178 procfs_freevp(vp)
    179 	struct vnode *vp;
    180 {
    181 	struct pfsnode *pfs = VTOPFS(vp);
    182 
    183 	procfs_hashrem(pfs);
    184 
    185 	FREE(vp->v_data, M_TEMP);
    186 	vp->v_data = 0;
    187 	return (0);
    188 }
    189 
    190 int
    191 procfs_rw(v)
    192 	void *v;
    193 {
    194 	struct vop_read_args *ap = v;
    195 	struct vnode *vp = ap->a_vp;
    196 	struct uio *uio = ap->a_uio;
    197 	struct proc *curp = uio->uio_procp;
    198 	struct pfsnode *pfs = VTOPFS(vp);
    199 	struct proc *p;
    200 
    201 	p = PFIND(pfs->pfs_pid);
    202 	if (p == 0)
    203 		return (EINVAL);
    204 
    205 	switch (pfs->pfs_type) {
    206 	case Pregs:
    207 	case Pfpregs:
    208 	case Pmem:
    209 		/*
    210 		 * Do not allow init to be modified while in secure mode; it
    211 		 * could be duped into changing the security level.
    212 		 */
    213 		if (uio->uio_rw == UIO_WRITE &&
    214 		    p == initproc && securelevel > -1)
    215 			return (EPERM);
    216 		break;
    217 
    218 	default:
    219 		break;
    220 	}
    221 
    222 	switch (pfs->pfs_type) {
    223 	case Pnote:
    224 	case Pnotepg:
    225 		return (procfs_donote(curp, p, pfs, uio));
    226 
    227 	case Pregs:
    228 		return (procfs_doregs(curp, p, pfs, uio));
    229 
    230 	case Pfpregs:
    231 		return (procfs_dofpregs(curp, p, pfs, uio));
    232 
    233 	case Pctl:
    234 		return (procfs_doctl(curp, p, pfs, uio));
    235 
    236 	case Pstatus:
    237 		return (procfs_dostatus(curp, p, pfs, uio));
    238 
    239 	case Pmap:
    240 		return (procfs_domap(curp, p, pfs, uio));
    241 
    242 	case Pmem:
    243 		return (procfs_domem(curp, p, pfs, uio));
    244 
    245 	case Pcmdline:
    246 		return (procfs_docmdline(curp, p, pfs, uio));
    247 
    248 	case Pmeminfo:
    249 		return (procfs_domeminfo(curp, p, pfs, uio));
    250 	case Pcpuinfo:
    251 		return (procfs_docpuinfo(curp, p, pfs, uio));
    252 
    253 	default:
    254 		return (EOPNOTSUPP);
    255 	}
    256 }
    257 
    258 /*
    259  * Get a string from userland into (buf).  Strip a trailing
    260  * nl character (to allow easy access from the shell).
    261  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
    262  * will automatically add a nul char at the end.
    263  *
    264  * Returns 0 on success or the following errors
    265  *
    266  * EINVAL:    file offset is non-zero.
    267  * EMSGSIZE:  message is longer than kernel buffer
    268  * EFAULT:    user i/o buffer is not addressable
    269  */
    270 int
    271 vfs_getuserstr(uio, buf, buflenp)
    272 	struct uio *uio;
    273 	char *buf;
    274 	int *buflenp;
    275 {
    276 	int xlen;
    277 	int error;
    278 
    279 	if (uio->uio_offset != 0)
    280 		return (EINVAL);
    281 
    282 	xlen = *buflenp;
    283 
    284 	/* must be able to read the whole string in one go */
    285 	if (xlen < uio->uio_resid)
    286 		return (EMSGSIZE);
    287 	xlen = uio->uio_resid;
    288 
    289 	if ((error = uiomove(buf, xlen, uio)) != 0)
    290 		return (error);
    291 
    292 	/* allow multiple writes without seeks */
    293 	uio->uio_offset = 0;
    294 
    295 	/* cleanup string and remove trailing newline */
    296 	buf[xlen] = '\0';
    297 	xlen = strlen(buf);
    298 	if (xlen > 0 && buf[xlen-1] == '\n')
    299 		buf[--xlen] = '\0';
    300 	*buflenp = xlen;
    301 
    302 	return (0);
    303 }
    304 
    305 vfs_namemap_t *
    306 vfs_findname(nm, buf, buflen)
    307 	vfs_namemap_t *nm;
    308 	char *buf;
    309 	int buflen;
    310 {
    311 
    312 	for (; nm->nm_name; nm++)
    313 		if (memcmp(buf, nm->nm_name, buflen+1) == 0)
    314 			return (nm);
    315 
    316 	return (0);
    317 }
    318 
    319 /*
    320  * Initialize pfsnode hash table.
    321  */
    322 void
    323 procfs_hashinit()
    324 {
    325 	lockinit(&pfs_hashlock, PINOD, "pfs_hashlock", 0, 0);
    326 	pfs_hashtbl = hashinit(desiredvnodes / 4, M_UFSMNT, M_WAITOK,
    327 	    &pfs_ihash);
    328 	simple_lock_init(&pfs_hash_slock);
    329 }
    330 
    331 /*
    332  * Free pfsnode hash table.
    333  */
    334 void
    335 procfs_hashdone()
    336 {
    337 	hashdone(pfs_hashtbl, M_UFSMNT);
    338 }
    339 
    340 struct vnode *
    341 procfs_hashget(pid, type, mp)
    342 	pid_t pid;
    343 	pfstype type;
    344 	struct mount *mp;
    345 {
    346 	struct pfsnode *pp;
    347 	struct vnode *vp;
    348 
    349 loop:
    350 	simple_lock(&pfs_hash_slock);
    351 	for (pp = PFSPIDHASH(pid)->lh_first; pp; pp = pp->pfs_hash.le_next) {
    352 		vp = PFSTOV(pp);
    353 		if (pid == pp->pfs_pid && pp->pfs_type == type &&
    354 		    vp->v_mount == mp) {
    355 			simple_lock(&vp->v_interlock);
    356 			simple_unlock(&pfs_hash_slock);
    357 			if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK))
    358 				goto loop;
    359 			return (vp);
    360 		}
    361 	}
    362 	simple_unlock(&pfs_hash_slock);
    363 	return (NULL);
    364 }
    365 
    366 /*
    367  * Insert the pfsnode into the hash table and lock it.
    368  */
    369 void
    370 procfs_hashins(pp)
    371 	struct pfsnode *pp;
    372 {
    373 	struct pfs_hashhead *ppp;
    374 
    375 	/* lock the pfsnode, then put it on the appropriate hash list */
    376 	lockmgr(&pp->pfs_vnode->v_lock, LK_EXCLUSIVE, (struct simplelock *)0);
    377 
    378 	simple_lock(&pfs_hash_slock);
    379 	ppp = PFSPIDHASH(pp->pfs_pid);
    380 	LIST_INSERT_HEAD(ppp, pp, pfs_hash);
    381 	simple_unlock(&pfs_hash_slock);
    382 }
    383 
    384 /*
    385  * Remove the pfsnode from the hash table.
    386  */
    387 void
    388 procfs_hashrem(pp)
    389 	struct pfsnode *pp;
    390 {
    391 	simple_lock(&pfs_hash_slock);
    392 	LIST_REMOVE(pp, pfs_hash);
    393 	simple_unlock(&pfs_hash_slock);
    394 }
    395 
    396 void
    397 procfs_revoke_vnodes(p, arg)
    398 	struct proc *p;
    399 	void *arg;
    400 {
    401 	struct pfsnode *pfs, *pnext;
    402 	struct vnode *vp;
    403 	struct mount *mp = (struct mount *)arg;
    404 
    405 	if (!(p->p_flag & P_SUGID))
    406 		return;
    407 
    408 	for (pfs = PFSPIDHASH(p->p_pid)->lh_first; pfs; pfs = pnext) {
    409 		vp = PFSTOV(pfs);
    410 		pnext = pfs->pfs_hash.le_next;
    411 		if (vp->v_usecount > 0 && pfs->pfs_pid == p->p_pid &&
    412 		    vp->v_mount == mp)
    413 			VOP_REVOKE(vp, REVOKEALL);
    414 	}
    415 }
    416