Home | History | Annotate | Line # | Download | only in procfs
procfs_subr.c revision 1.36.2.2
      1 /*	$NetBSD: procfs_subr.c,v 1.36.2.2 2001/04/09 01:58:09 nathanw 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, NULL);
    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 Pmaps:	/* /proc/N/maps = -r--r--r-- */
    160 	case Pstatus:	/* /proc/N/status = -r--r--r-- */
    161 	case Pcmdline:	/* /proc/N/cmdline = -r--r--r-- */
    162 	case Pmeminfo:	/* /proc/meminfo = -r--r--r-- */
    163 	case Pcpuinfo:	/* /proc/cpuinfo = -r--r--r-- */
    164 		pfs->pfs_mode = S_IRUSR|S_IRGRP|S_IROTH;
    165 		vp->v_type = VREG;
    166 		break;
    167 
    168 	default:
    169 		panic("procfs_allocvp");
    170 	}
    171 
    172 	procfs_hashins(pfs);
    173 	uvm_vnp_setsize(vp, 0);
    174 	lockmgr(&pfs_hashlock, LK_RELEASE, NULL);
    175 
    176 	return (error);
    177 }
    178 
    179 int
    180 procfs_freevp(vp)
    181 	struct vnode *vp;
    182 {
    183 	struct pfsnode *pfs = VTOPFS(vp);
    184 
    185 	procfs_hashrem(pfs);
    186 
    187 	FREE(vp->v_data, M_TEMP);
    188 	vp->v_data = 0;
    189 	return (0);
    190 }
    191 
    192 int
    193 procfs_rw(v)
    194 	void *v;
    195 {
    196 	struct vop_read_args *ap = v;
    197 	struct vnode *vp = ap->a_vp;
    198 	struct uio *uio = ap->a_uio;
    199 	struct proc *curp = uio->uio_procp;
    200 	struct pfsnode *pfs = VTOPFS(vp);
    201 	struct lwp *l;
    202 	struct proc *p;
    203 
    204 	p = PFIND(pfs->pfs_pid);
    205 	if (p == 0)
    206 		return (EINVAL);
    207 
    208 	/* XXX NJWLWP
    209 	 * The entire procfs interface needs work to be useful to
    210 	 * a process with multiple LWPs. For the moment, we'll
    211 	 * just kluge this and fail on others.
    212 	 */
    213 
    214 	if (p->p_nlwps > 1)
    215 		switch (pfs->pfs_type) {
    216 		case Pnote:
    217 		case Pnotepg:
    218 		case Pstatus:
    219 		case Pmap:
    220 		case Pmem:
    221 		case Pcmdline:
    222 			break;
    223 		case Pregs:
    224 		case Pfpregs:
    225 		case Pctl:
    226 		default:
    227 			return (EOPNOTSUPP);
    228 		}
    229 
    230 	l = LIST_FIRST(&p->p_lwps);
    231 
    232 	switch (pfs->pfs_type) {
    233 	case Pregs:
    234 	case Pfpregs:
    235 	case Pmem:
    236 		/*
    237 		 * Do not allow init to be modified while in secure mode; it
    238 		 * could be duped into changing the security level.
    239 		 */
    240 		if (uio->uio_rw == UIO_WRITE &&
    241 		    p == initproc && securelevel > -1)
    242 			return (EPERM);
    243 		break;
    244 
    245 	default:
    246 		break;
    247 	}
    248 
    249 	switch (pfs->pfs_type) {
    250 	case Pnote:
    251 	case Pnotepg:
    252 		return (procfs_donote(curp, p, pfs, uio));
    253 
    254 	case Pregs:
    255 		return (procfs_doregs(curp, l, pfs, uio));
    256 
    257 	case Pfpregs:
    258 		return (procfs_dofpregs(curp, l, pfs, uio));
    259 
    260 	case Pctl:
    261 		return (procfs_doctl(curp, l, pfs, uio));
    262 
    263 	case Pstatus:
    264 		return (procfs_dostatus(curp, l, pfs, uio));
    265 
    266 	case Pmap:
    267 		return (procfs_domap(curp, p, pfs, uio, 0));
    268 
    269 	case Pmaps:
    270 		return (procfs_domap(curp, p, pfs, uio, 1));
    271 
    272 	case Pmem:
    273 		return (procfs_domem(curp, l, pfs, uio));
    274 
    275 	case Pcmdline:
    276 		return (procfs_docmdline(curp, p, pfs, uio));
    277 
    278 	case Pmeminfo:
    279 		return (procfs_domeminfo(curp, p, pfs, uio));
    280 	case Pcpuinfo:
    281 		return (procfs_docpuinfo(curp, p, pfs, uio));
    282 
    283 	default:
    284 		return (EOPNOTSUPP);
    285 	}
    286 }
    287 
    288 /*
    289  * Get a string from userland into (buf).  Strip a trailing
    290  * nl character (to allow easy access from the shell).
    291  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
    292  * will automatically add a nul char at the end.
    293  *
    294  * Returns 0 on success or the following errors
    295  *
    296  * EINVAL:    file offset is non-zero.
    297  * EMSGSIZE:  message is longer than kernel buffer
    298  * EFAULT:    user i/o buffer is not addressable
    299  */
    300 int
    301 vfs_getuserstr(uio, buf, buflenp)
    302 	struct uio *uio;
    303 	char *buf;
    304 	int *buflenp;
    305 {
    306 	int xlen;
    307 	int error;
    308 
    309 	if (uio->uio_offset != 0)
    310 		return (EINVAL);
    311 
    312 	xlen = *buflenp;
    313 
    314 	/* must be able to read the whole string in one go */
    315 	if (xlen < uio->uio_resid)
    316 		return (EMSGSIZE);
    317 	xlen = uio->uio_resid;
    318 
    319 	if ((error = uiomove(buf, xlen, uio)) != 0)
    320 		return (error);
    321 
    322 	/* allow multiple writes without seeks */
    323 	uio->uio_offset = 0;
    324 
    325 	/* cleanup string and remove trailing newline */
    326 	buf[xlen] = '\0';
    327 	xlen = strlen(buf);
    328 	if (xlen > 0 && buf[xlen-1] == '\n')
    329 		buf[--xlen] = '\0';
    330 	*buflenp = xlen;
    331 
    332 	return (0);
    333 }
    334 
    335 const vfs_namemap_t *
    336 vfs_findname(nm, buf, buflen)
    337 	const vfs_namemap_t *nm;
    338 	const char *buf;
    339 	int buflen;
    340 {
    341 
    342 	for (; nm->nm_name; nm++)
    343 		if (memcmp(buf, nm->nm_name, buflen+1) == 0)
    344 			return (nm);
    345 
    346 	return (0);
    347 }
    348 
    349 /*
    350  * Initialize pfsnode hash table.
    351  */
    352 void
    353 procfs_hashinit()
    354 {
    355 	lockinit(&pfs_hashlock, PINOD, "pfs_hashlock", 0, 0);
    356 	pfs_hashtbl = hashinit(desiredvnodes / 4, HASH_LIST, M_UFSMNT,
    357 	    M_WAITOK, &pfs_ihash);
    358 	simple_lock_init(&pfs_hash_slock);
    359 }
    360 
    361 /*
    362  * Free pfsnode hash table.
    363  */
    364 void
    365 procfs_hashdone()
    366 {
    367 	hashdone(pfs_hashtbl, M_UFSMNT);
    368 }
    369 
    370 struct vnode *
    371 procfs_hashget(pid, type, mp)
    372 	pid_t pid;
    373 	pfstype type;
    374 	struct mount *mp;
    375 {
    376 	struct pfsnode *pp;
    377 	struct vnode *vp;
    378 
    379 loop:
    380 	simple_lock(&pfs_hash_slock);
    381 	for (pp = PFSPIDHASH(pid)->lh_first; pp; pp = pp->pfs_hash.le_next) {
    382 		vp = PFSTOV(pp);
    383 		if (pid == pp->pfs_pid && pp->pfs_type == type &&
    384 		    vp->v_mount == mp) {
    385 			simple_lock(&vp->v_interlock);
    386 			simple_unlock(&pfs_hash_slock);
    387 			if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK))
    388 				goto loop;
    389 			return (vp);
    390 		}
    391 	}
    392 	simple_unlock(&pfs_hash_slock);
    393 	return (NULL);
    394 }
    395 
    396 /*
    397  * Insert the pfsnode into the hash table and lock it.
    398  */
    399 void
    400 procfs_hashins(pp)
    401 	struct pfsnode *pp;
    402 {
    403 	struct pfs_hashhead *ppp;
    404 
    405 	/* lock the pfsnode, then put it on the appropriate hash list */
    406 	lockmgr(&pp->pfs_vnode->v_lock, LK_EXCLUSIVE, (struct simplelock *)0);
    407 
    408 	simple_lock(&pfs_hash_slock);
    409 	ppp = PFSPIDHASH(pp->pfs_pid);
    410 	LIST_INSERT_HEAD(ppp, pp, pfs_hash);
    411 	simple_unlock(&pfs_hash_slock);
    412 }
    413 
    414 /*
    415  * Remove the pfsnode from the hash table.
    416  */
    417 void
    418 procfs_hashrem(pp)
    419 	struct pfsnode *pp;
    420 {
    421 	simple_lock(&pfs_hash_slock);
    422 	LIST_REMOVE(pp, pfs_hash);
    423 	simple_unlock(&pfs_hash_slock);
    424 }
    425 
    426 void
    427 procfs_revoke_vnodes(p, arg)
    428 	struct proc *p;
    429 	void *arg;
    430 {
    431 	struct pfsnode *pfs, *pnext;
    432 	struct vnode *vp;
    433 	struct mount *mp = (struct mount *)arg;
    434 
    435 	if (!(p->p_flag & P_SUGID))
    436 		return;
    437 
    438 	for (pfs = PFSPIDHASH(p->p_pid)->lh_first; pfs; pfs = pnext) {
    439 		vp = PFSTOV(pfs);
    440 		pnext = pfs->pfs_hash.le_next;
    441 		if (vp->v_usecount > 0 && pfs->pfs_pid == p->p_pid &&
    442 		    vp->v_mount == mp)
    443 			VOP_REVOKE(vp, REVOKEALL);
    444 	}
    445 }
    446