Home | History | Annotate | Line # | Download | only in procfs
procfs_vnops.c revision 1.21
      1   1.1       pk /*
      2   1.9      cgd  * Copyright (c) 1993 Jan-Simon Pendry
      3  1.21  mycroft  * Copyright (c) 1993
      4  1.21  mycroft  *	The Regents of the University of California.  All rights reserved.
      5   1.2       pk  *
      6   1.9      cgd  * This code is derived from software contributed to Berkeley by
      7   1.9      cgd  * Jan-Simon Pendry.
      8   1.9      cgd  *
      9   1.2       pk  * Redistribution and use in source and binary forms, with or without
     10   1.2       pk  * modification, are permitted provided that the following conditions
     11   1.2       pk  * are met:
     12   1.2       pk  * 1. Redistributions of source code must retain the above copyright
     13   1.2       pk  *    notice, this list of conditions and the following disclaimer.
     14   1.2       pk  * 2. Redistributions in binary form must reproduce the above copyright
     15   1.2       pk  *    notice, this list of conditions and the following disclaimer in the
     16   1.2       pk  *    documentation and/or other materials provided with the distribution.
     17   1.2       pk  * 3. All advertising materials mentioning features or use of this software
     18   1.2       pk  *    must display the following acknowledgement:
     19   1.9      cgd  *	This product includes software developed by the University of
     20   1.9      cgd  *	California, Berkeley and its contributors.
     21   1.9      cgd  * 4. Neither the name of the University nor the names of its contributors
     22   1.9      cgd  *    may be used to endorse or promote products derived from this software
     23   1.9      cgd  *    without specific prior written permission.
     24   1.9      cgd  *
     25   1.9      cgd  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     26   1.9      cgd  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     27   1.9      cgd  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     28   1.9      cgd  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     29   1.9      cgd  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     30   1.9      cgd  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     31   1.9      cgd  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     32   1.9      cgd  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     33   1.9      cgd  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     34   1.9      cgd  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     35   1.9      cgd  * SUCH DAMAGE.
     36   1.2       pk  *
     37  1.21  mycroft  *	from: Id: procfs_vnops.c,v 3.2 1993/12/15 09:40:17 jsp Exp
     38  1.21  mycroft  *	from: @(#)procfs_vnops.c	8.6 (Berkeley) 2/7/94
     39  1.21  mycroft  *	$Id: procfs_vnops.c,v 1.21 1994/06/08 11:33:45 mycroft Exp $
     40   1.2       pk  */
     41   1.2       pk 
     42   1.2       pk /*
     43   1.9      cgd  * procfs vnode interface
     44   1.1       pk  */
     45   1.1       pk 
     46   1.8  mycroft #include <sys/param.h>
     47   1.8  mycroft #include <sys/systm.h>
     48   1.8  mycroft #include <sys/time.h>
     49   1.8  mycroft #include <sys/kernel.h>
     50   1.8  mycroft #include <sys/file.h>
     51   1.8  mycroft #include <sys/proc.h>
     52   1.8  mycroft #include <sys/vnode.h>
     53   1.8  mycroft #include <sys/namei.h>
     54   1.9      cgd #include <sys/malloc.h>
     55  1.21  mycroft #include <sys/dirent.h>
     56   1.8  mycroft #include <sys/resourcevar.h>
     57  1.14      cgd #include <sys/ptrace.h>
     58  1.21  mycroft #include <vm/vm.h>	/* for PAGE_SIZE */
     59  1.21  mycroft #include <machine/reg.h>
     60   1.9      cgd #include <miscfs/procfs/procfs.h>
     61  1.11      cgd 
     62   1.9      cgd /*
     63   1.9      cgd  * Vnode Operations.
     64   1.9      cgd  *
     65   1.9      cgd  */
     66   1.1       pk 
     67   1.1       pk /*
     68   1.9      cgd  * This is a list of the valid names in the
     69   1.9      cgd  * process-specific sub-directories.  It is
     70   1.9      cgd  * used in procfs_lookup and procfs_readdir
     71   1.9      cgd  */
     72   1.9      cgd static struct pfsnames {
     73  1.21  mycroft 	u_char	d_type;
     74  1.21  mycroft 	u_char	d_namlen;
     75   1.9      cgd 	char	d_name[PROCFS_NAMELEN];
     76   1.9      cgd 	pfstype	d_pfstype;
     77  1.21  mycroft 	int	(*d_valid) __P((struct proc *p));
     78   1.9      cgd } procent[] = {
     79   1.9      cgd #define N(s) sizeof(s)-1, s
     80   1.9      cgd 	/* namlen, nam, type */
     81  1.21  mycroft 	{ DT_DIR, N("."),	Pproc,		NULL },
     82  1.21  mycroft 	{ DT_DIR, N(".."),	Proot,		NULL },
     83  1.21  mycroft 	{ DT_REG, N("file"),	Pfile,		procfs_validfile },
     84  1.21  mycroft 	{ DT_REG, N("mem"),	Pmem,		NULL },
     85  1.21  mycroft 	{ DT_REG, N("regs"),	Pregs,		procfs_validregs },
     86  1.21  mycroft 	{ DT_REG, N("fpregs"),	Pfpregs,	procfs_validfpregs },
     87  1.21  mycroft 	{ DT_REG, N("ctl"),	Pctl,		NULL },
     88  1.21  mycroft 	{ DT_REG, N("status"),	Pstatus,	NULL },
     89  1.21  mycroft 	{ DT_REG, N("note"),	Pnote,		NULL },
     90  1.21  mycroft 	{ DT_REG, N("notepg"),	Pnotepg,	NULL },
     91   1.9      cgd #undef N
     92   1.1       pk };
     93   1.9      cgd #define Nprocent (sizeof(procent)/sizeof(procent[0]))
     94   1.9      cgd 
     95   1.9      cgd static pid_t atopid __P((const char *, u_int));
     96   1.1       pk 
     97   1.1       pk /*
     98   1.9      cgd  * set things up for doing i/o on
     99   1.9      cgd  * the pfsnode (vp).  (vp) is locked
    100   1.9      cgd  * on entry, and should be left locked
    101   1.9      cgd  * on exit.
    102   1.1       pk  *
    103   1.9      cgd  * for procfs we don't need to do anything
    104   1.9      cgd  * in particular for i/o.  all that is done
    105   1.9      cgd  * is to support exclusive open on process
    106   1.9      cgd  * memory images.
    107   1.1       pk  */
    108  1.21  mycroft procfs_open(ap)
    109  1.21  mycroft 	struct vop_open_args *ap;
    110   1.1       pk {
    111  1.21  mycroft 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
    112   1.9      cgd 
    113   1.9      cgd 	switch (pfs->pfs_type) {
    114   1.9      cgd 	case Pmem:
    115   1.9      cgd 		if (PFIND(pfs->pfs_pid) == 0)
    116   1.9      cgd 			return (ENOENT);	/* was ESRCH, jsp */
    117   1.9      cgd 
    118  1.21  mycroft 		if ((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL) ||
    119  1.21  mycroft 		    (pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE))
    120   1.9      cgd 			return (EBUSY);
    121   1.1       pk 
    122  1.21  mycroft 		if (ap->a_mode & FWRITE)
    123  1.21  mycroft 			pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL);
    124   1.1       pk 
    125   1.9      cgd 		return (0);
    126   1.1       pk 
    127   1.9      cgd 	default:
    128   1.9      cgd 		break;
    129   1.9      cgd 	}
    130   1.1       pk 
    131   1.9      cgd 	return (0);
    132   1.1       pk }
    133   1.1       pk 
    134   1.1       pk /*
    135   1.9      cgd  * close the pfsnode (vp) after doing i/o.
    136   1.9      cgd  * (vp) is not locked on entry or exit.
    137   1.9      cgd  *
    138   1.9      cgd  * nothing to do for procfs other than undo
    139   1.9      cgd  * any exclusive open flag (see _open above).
    140   1.1       pk  */
    141  1.21  mycroft procfs_close(ap)
    142  1.21  mycroft 	struct vop_close_args *ap;
    143   1.1       pk {
    144  1.21  mycroft 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
    145   1.1       pk 
    146   1.9      cgd 	switch (pfs->pfs_type) {
    147   1.9      cgd 	case Pmem:
    148  1.21  mycroft 		if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL))
    149   1.9      cgd 			pfs->pfs_flags &= ~(FWRITE|O_EXCL);
    150   1.9      cgd 		break;
    151   1.9      cgd 	}
    152   1.1       pk 
    153   1.1       pk 	return (0);
    154   1.1       pk }
    155   1.1       pk 
    156   1.1       pk /*
    157   1.9      cgd  * do an ioctl operation on pfsnode (vp).
    158   1.9      cgd  * (vp) is not locked on entry or exit.
    159   1.1       pk  */
    160  1.21  mycroft procfs_ioctl(ap)
    161  1.21  mycroft 	struct vop_ioctl_args *ap;
    162   1.1       pk {
    163   1.1       pk 
    164   1.9      cgd 	return (ENOTTY);
    165   1.1       pk }
    166   1.1       pk 
    167   1.1       pk /*
    168   1.9      cgd  * do block mapping for pfsnode (vp).
    169   1.9      cgd  * since we don't use the buffer cache
    170   1.9      cgd  * for procfs this function should never
    171   1.9      cgd  * be called.  in any case, it's not clear
    172   1.9      cgd  * what part of the kernel ever makes use
    173   1.9      cgd  * of this function.  for sanity, this is the
    174   1.9      cgd  * usual no-op bmap, although returning
    175   1.9      cgd  * (EIO) would be a reasonable alternative.
    176   1.1       pk  */
    177  1.21  mycroft procfs_bmap(ap)
    178  1.21  mycroft 	struct vop_bmap_args *ap;
    179   1.1       pk {
    180   1.1       pk 
    181  1.21  mycroft 	if (ap->a_vpp != NULL)
    182  1.21  mycroft 		*ap->a_vpp = ap->a_vp;
    183  1.21  mycroft 	if (ap->a_bnp != NULL)
    184  1.21  mycroft 		*ap->a_bnp = ap->a_bn;
    185   1.1       pk 	return (0);
    186   1.1       pk }
    187   1.1       pk 
    188   1.1       pk /*
    189   1.9      cgd  * _inactive is called when the pfsnode
    190   1.9      cgd  * is vrele'd and the reference count goes
    191   1.9      cgd  * to zero.  (vp) will be on the vnode free
    192   1.9      cgd  * list, so to get it back vget() must be
    193   1.9      cgd  * used.
    194   1.9      cgd  *
    195   1.9      cgd  * for procfs, check if the process is still
    196   1.9      cgd  * alive and if it isn't then just throw away
    197   1.9      cgd  * the vnode by calling vgone().  this may
    198   1.9      cgd  * be overkill and a waste of time since the
    199   1.9      cgd  * chances are that the process will still be
    200   1.9      cgd  * there and PFIND is not free.
    201   1.9      cgd  *
    202   1.9      cgd  * (vp) is not locked on entry or exit.
    203   1.1       pk  */
    204  1.21  mycroft procfs_inactive(ap)
    205  1.21  mycroft 	struct vop_inactive_args *ap;
    206   1.1       pk {
    207  1.21  mycroft 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
    208   1.1       pk 
    209   1.9      cgd 	if (PFIND(pfs->pfs_pid) == 0)
    210  1.21  mycroft 		vgone(ap->a_vp);
    211   1.5       pk 
    212   1.9      cgd 	return (0);
    213   1.1       pk }
    214   1.1       pk 
    215   1.1       pk /*
    216   1.9      cgd  * _reclaim is called when getnewvnode()
    217   1.9      cgd  * wants to make use of an entry on the vnode
    218   1.9      cgd  * free list.  at this time the filesystem needs
    219   1.9      cgd  * to free any private data and remove the node
    220   1.9      cgd  * from any private lists.
    221   1.1       pk  */
    222  1.21  mycroft procfs_reclaim(ap)
    223  1.21  mycroft 	struct vop_reclaim_args *ap;
    224  1.21  mycroft {
    225  1.21  mycroft 
    226  1.21  mycroft 	return (procfs_freevp(ap->a_vp));
    227  1.21  mycroft }
    228  1.21  mycroft 
    229  1.21  mycroft /*
    230  1.21  mycroft  * Return POSIX pathconf information applicable to special devices.
    231  1.21  mycroft  */
    232  1.21  mycroft procfs_pathconf(ap)
    233  1.21  mycroft 	struct vop_pathconf_args /* {
    234  1.21  mycroft 		struct vnode *a_vp;
    235  1.21  mycroft 		int a_name;
    236  1.21  mycroft 		int *a_retval;
    237  1.21  mycroft 	} */ *ap;
    238   1.1       pk {
    239   1.1       pk 
    240  1.21  mycroft 	switch (ap->a_name) {
    241  1.21  mycroft 	case _PC_LINK_MAX:
    242  1.21  mycroft 		*ap->a_retval = LINK_MAX;
    243  1.21  mycroft 		return (0);
    244  1.21  mycroft 	case _PC_MAX_CANON:
    245  1.21  mycroft 		*ap->a_retval = MAX_CANON;
    246  1.21  mycroft 		return (0);
    247  1.21  mycroft 	case _PC_MAX_INPUT:
    248  1.21  mycroft 		*ap->a_retval = MAX_INPUT;
    249  1.21  mycroft 		return (0);
    250  1.21  mycroft 	case _PC_PIPE_BUF:
    251  1.21  mycroft 		*ap->a_retval = PIPE_BUF;
    252  1.21  mycroft 		return (0);
    253  1.21  mycroft 	case _PC_CHOWN_RESTRICTED:
    254  1.21  mycroft 		*ap->a_retval = 1;
    255  1.21  mycroft 		return (0);
    256  1.21  mycroft 	case _PC_VDISABLE:
    257  1.21  mycroft 		*ap->a_retval = _POSIX_VDISABLE;
    258  1.21  mycroft 		return (0);
    259  1.21  mycroft 	default:
    260  1.21  mycroft 		return (EINVAL);
    261  1.21  mycroft 	}
    262  1.21  mycroft 	/* NOTREACHED */
    263   1.1       pk }
    264   1.1       pk 
    265   1.1       pk /*
    266   1.9      cgd  * _print is used for debugging.
    267   1.9      cgd  * just print a readable description
    268   1.9      cgd  * of (vp).
    269   1.1       pk  */
    270  1.21  mycroft procfs_print(ap)
    271  1.21  mycroft 	struct vop_print_args *ap;
    272   1.1       pk {
    273  1.21  mycroft 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
    274   1.5       pk 
    275  1.21  mycroft 	printf("tag VT_PROCFS, type %s, pid %d, mode %x, flags %x\n",
    276  1.21  mycroft 	    pfs->pfs_type, pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags);
    277   1.1       pk }
    278   1.1       pk 
    279   1.1       pk /*
    280   1.9      cgd  * _abortop is called when operations such as
    281   1.9      cgd  * rename and create fail.  this entry is responsible
    282   1.9      cgd  * for undoing any side-effects caused by the lookup.
    283   1.9      cgd  * this will always include freeing the pathname buffer.
    284   1.1       pk  */
    285  1.21  mycroft procfs_abortop(ap)
    286  1.21  mycroft 	struct vop_abortop_args *ap;
    287   1.1       pk {
    288   1.9      cgd 
    289  1.21  mycroft 	if ((ap->a_cnp->cn_flags & (HASBUF | SAVESTART)) == HASBUF)
    290  1.21  mycroft 		FREE(ap->a_cnp->cn_pnbuf, M_NAMEI);
    291   1.9      cgd 	return (0);
    292   1.1       pk }
    293   1.1       pk 
    294   1.1       pk /*
    295   1.9      cgd  * generic entry point for unsupported operations
    296   1.1       pk  */
    297   1.9      cgd procfs_badop()
    298   1.1       pk {
    299   1.1       pk 
    300   1.9      cgd 	return (EIO);
    301   1.1       pk }
    302   1.1       pk 
    303   1.1       pk /*
    304   1.9      cgd  * Invent attributes for pfsnode (vp) and store
    305   1.9      cgd  * them in (vap).
    306   1.9      cgd  * Directories lengths are returned as zero since
    307   1.9      cgd  * any real length would require the genuine size
    308   1.9      cgd  * to be computed, and nothing cares anyway.
    309   1.9      cgd  *
    310   1.9      cgd  * this is relatively minimal for procfs.
    311   1.1       pk  */
    312  1.21  mycroft procfs_getattr(ap)
    313  1.21  mycroft 	struct vop_getattr_args *ap;
    314   1.1       pk {
    315  1.21  mycroft 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
    316  1.21  mycroft 	struct vattr *vap = ap->a_vap;
    317   1.1       pk 	struct proc *procp;
    318   1.9      cgd 	int error;
    319   1.1       pk 
    320  1.21  mycroft 	/* first check the process still exists */
    321  1.21  mycroft 	switch (pfs->pfs_type) {
    322  1.21  mycroft 	case Proot:
    323  1.21  mycroft 	case Pcurproc:
    324  1.21  mycroft 		procp = 0;
    325  1.21  mycroft 		break;
    326  1.21  mycroft 
    327  1.21  mycroft 	default:
    328  1.21  mycroft 		procp = PFIND(pfs->pfs_pid);
    329  1.21  mycroft 		if (procp == 0)
    330  1.21  mycroft 			return (ENOENT);
    331  1.21  mycroft 	}
    332  1.21  mycroft 
    333  1.21  mycroft 	error = 0;
    334  1.21  mycroft 
    335   1.9      cgd 	/* start by zeroing out the attributes */
    336   1.1       pk 	VATTR_NULL(vap);
    337   1.9      cgd 
    338   1.9      cgd 	/* next do all the common fields */
    339  1.21  mycroft 	vap->va_type = ap->a_vp->v_type;
    340   1.9      cgd 	vap->va_mode = pfs->pfs_mode;
    341   1.9      cgd 	vap->va_fileid = pfs->pfs_fileno;
    342   1.9      cgd 	vap->va_flags = 0;
    343  1.19      cgd 	vap->va_blocksize = PAGE_SIZE;
    344   1.9      cgd 	vap->va_bytes = vap->va_size = 0;
    345   1.9      cgd 
    346   1.9      cgd 	/*
    347   1.9      cgd 	 * Make all times be current TOD.
    348   1.9      cgd 	 * It would be possible to get the process start
    349   1.9      cgd 	 * time from the p_stat structure, but there's
    350   1.9      cgd 	 * no "file creation" time stamp anyway, and the
    351   1.9      cgd 	 * p_stat structure is not addressible if u. gets
    352   1.9      cgd 	 * swapped out for that process.
    353  1.21  mycroft 	 *
    354  1.21  mycroft 	 * XXX
    355  1.21  mycroft 	 * Note that microtime() returns a timeval, not a timespec.
    356   1.9      cgd 	 */
    357   1.9      cgd 	microtime(&vap->va_ctime);
    358   1.9      cgd 	vap->va_atime = vap->va_mtime = vap->va_ctime;
    359   1.9      cgd 
    360  1.21  mycroft 	/*
    361  1.21  mycroft 	 * If the process has exercised some setuid or setgid
    362  1.21  mycroft 	 * privilege, then rip away read/write permission so
    363  1.21  mycroft 	 * that only root can gain access.
    364  1.21  mycroft 	 */
    365  1.21  mycroft 	switch (pfs->pfs_type) {
    366  1.21  mycroft 	case Pmem:
    367  1.21  mycroft 	case Pregs:
    368  1.21  mycroft 	case Pfpregs:
    369  1.21  mycroft 		if (procp->p_flag & P_SUGID)
    370  1.21  mycroft 			vap->va_mode &= ~((VREAD|VWRITE)|
    371  1.21  mycroft 					  ((VREAD|VWRITE)>>3)|
    372  1.21  mycroft 					  ((VREAD|VWRITE)>>6));
    373  1.21  mycroft 	case Pctl:
    374  1.21  mycroft 	case Pstatus:
    375  1.21  mycroft 	case Pnote:
    376  1.21  mycroft 	case Pnotepg:
    377  1.12       ws 		vap->va_nlink = 1;
    378  1.12       ws 		vap->va_uid = procp->p_ucred->cr_uid;
    379  1.12       ws 		vap->va_gid = procp->p_ucred->cr_gid;
    380  1.21  mycroft 		break;
    381  1.12       ws 	}
    382  1.12       ws 
    383   1.9      cgd 	/*
    384   1.9      cgd 	 * now do the object specific fields
    385   1.9      cgd 	 *
    386   1.9      cgd 	 * The size could be set from struct reg, but it's hardly
    387   1.9      cgd 	 * worth the trouble, and it puts some (potentially) machine
    388   1.9      cgd 	 * dependent data into this machine-independent code.  If it
    389   1.9      cgd 	 * becomes important then this function should break out into
    390   1.9      cgd 	 * a per-file stat function in the corresponding .c file.
    391   1.9      cgd 	 */
    392   1.1       pk 
    393   1.9      cgd 	switch (pfs->pfs_type) {
    394   1.9      cgd 	case Proot:
    395  1.21  mycroft 		/*
    396  1.21  mycroft 		 * Set nlink to 1 to tell fts(3) we don't actually know.
    397  1.21  mycroft 		 */
    398  1.21  mycroft 		vap->va_nlink = 1;
    399  1.21  mycroft 		vap->va_uid = 0;
    400  1.21  mycroft 		vap->va_gid = 0;
    401  1.21  mycroft 		vap->va_size = vap->va_bytes = DEV_BSIZE;
    402  1.21  mycroft 		break;
    403  1.21  mycroft 
    404  1.21  mycroft 	case Pcurproc: {
    405  1.21  mycroft 		char buf[16];		/* should be enough */
    406  1.21  mycroft 		vap->va_nlink = 1;
    407  1.21  mycroft 		vap->va_uid = 0;
    408  1.21  mycroft 		vap->va_gid = 0;
    409  1.21  mycroft 		vap->va_size = vap->va_bytes =
    410  1.21  mycroft 		    sprintf(buf, "%ld", (long)curproc->p_pid);
    411   1.9      cgd 		break;
    412  1.21  mycroft 	}
    413   1.9      cgd 
    414   1.9      cgd 	case Pproc:
    415   1.9      cgd 		vap->va_nlink = 2;
    416  1.21  mycroft 		vap->va_uid = procp->p_ucred->cr_uid;
    417  1.21  mycroft 		vap->va_gid = procp->p_ucred->cr_gid;
    418  1.21  mycroft 		vap->va_size = vap->va_bytes = DEV_BSIZE;
    419  1.21  mycroft 		break;
    420  1.21  mycroft 
    421  1.21  mycroft 	case Pfile:
    422  1.21  mycroft 		error = EOPNOTSUPP;
    423   1.9      cgd 		break;
    424   1.9      cgd 
    425   1.9      cgd 	case Pmem:
    426   1.9      cgd 		vap->va_bytes = vap->va_size =
    427   1.9      cgd 			ctob(procp->p_vmspace->vm_tsize +
    428   1.9      cgd 				    procp->p_vmspace->vm_dsize +
    429   1.9      cgd 				    procp->p_vmspace->vm_ssize);
    430   1.9      cgd 		break;
    431   1.9      cgd 
    432  1.18      cgd #if defined(PT_GETREGS) || defined(PT_SETREGS)
    433   1.9      cgd 	case Pregs:
    434  1.11      cgd 		vap->va_bytes = vap->va_size = sizeof(struct reg);
    435  1.18      cgd 		break;
    436  1.14      cgd #endif
    437  1.14      cgd 
    438  1.18      cgd #if defined(PT_GETFPREGS) || defined(PT_SETFPREGS)
    439  1.14      cgd 	case Pfpregs:
    440  1.14      cgd 		vap->va_bytes = vap->va_size = sizeof(struct fpreg);
    441  1.18      cgd 		break;
    442  1.14      cgd #endif
    443  1.12       ws 
    444  1.21  mycroft 	case Pctl:
    445  1.12       ws 	case Pstatus:
    446   1.9      cgd 	case Pnote:
    447   1.9      cgd 	case Pnotepg:
    448   1.9      cgd 		break;
    449   1.1       pk 
    450   1.9      cgd 	default:
    451  1.21  mycroft 		panic("procfs_getattr");
    452   1.1       pk 	}
    453   1.1       pk 
    454   1.9      cgd 	return (error);
    455   1.1       pk }
    456   1.1       pk 
    457  1.21  mycroft procfs_setattr(ap)
    458  1.21  mycroft 	struct vop_setattr_args *ap;
    459   1.5       pk {
    460   1.5       pk 	/*
    461   1.9      cgd 	 * just fake out attribute setting
    462   1.9      cgd 	 * it's not good to generate an error
    463   1.9      cgd 	 * return, otherwise things like creat()
    464   1.9      cgd 	 * will fail when they try to set the
    465   1.9      cgd 	 * file length to 0.  worse, this means
    466   1.9      cgd 	 * that echo $note > /proc/$pid/note will fail.
    467   1.5       pk 	 */
    468   1.5       pk 
    469   1.9      cgd 	return (0);
    470   1.5       pk }
    471   1.5       pk 
    472   1.9      cgd /*
    473   1.9      cgd  * implement access checking.
    474   1.9      cgd  *
    475   1.9      cgd  * something very similar to this code is duplicated
    476   1.9      cgd  * throughout the 4bsd kernel and should be moved
    477   1.9      cgd  * into kern/vfs_subr.c sometime.
    478   1.9      cgd  *
    479   1.9      cgd  * actually, the check for super-user is slightly
    480   1.9      cgd  * broken since it will allow read access to write-only
    481   1.9      cgd  * objects.  this doesn't cause any particular trouble
    482   1.9      cgd  * but does mean that the i/o entry points need to check
    483   1.9      cgd  * that the operation really does make sense.
    484   1.9      cgd  */
    485  1.21  mycroft procfs_access(ap)
    486  1.21  mycroft 	struct vop_access_args *ap;
    487   1.1       pk {
    488   1.9      cgd 	struct vattr *vap;
    489   1.1       pk 	struct vattr vattr;
    490   1.1       pk 	int error;
    491   1.1       pk 
    492   1.1       pk 	/*
    493   1.1       pk 	 * If you're the super-user,
    494   1.1       pk 	 * you always get access.
    495   1.1       pk 	 */
    496  1.21  mycroft 	if (ap->a_cred->cr_uid == 0)
    497   1.1       pk 		return (0);
    498  1.21  mycroft 
    499   1.1       pk 	vap = &vattr;
    500  1.21  mycroft 	if (error = VOP_GETATTR(ap->a_vp, vap, ap->a_cred, ap->a_p))
    501   1.1       pk 		return (error);
    502   1.9      cgd 
    503   1.1       pk 	/*
    504   1.1       pk 	 * Access check is based on only one of owner, group, public.
    505   1.1       pk 	 * If not owner, then check group. If not a member of the
    506   1.1       pk 	 * group, then check public access.
    507   1.1       pk 	 */
    508  1.21  mycroft 	if (ap->a_cred->cr_uid != vap->va_uid) {
    509   1.9      cgd 		gid_t *gp;
    510   1.9      cgd 		int i;
    511   1.9      cgd 
    512  1.21  mycroft 		ap->a_mode >>= 3;
    513  1.21  mycroft 		gp = ap->a_cred->cr_groups;
    514  1.21  mycroft 		for (i = 0; i < ap->a_cred->cr_ngroups; i++, gp++)
    515   1.1       pk 			if (vap->va_gid == *gp)
    516   1.1       pk 				goto found;
    517  1.21  mycroft 		ap->a_mode >>= 3;
    518   1.1       pk found:
    519   1.1       pk 		;
    520   1.1       pk 	}
    521   1.9      cgd 
    522  1.21  mycroft 	if ((vap->va_mode & ap->a_mode) == ap->a_mode)
    523   1.1       pk 		return (0);
    524   1.9      cgd 
    525   1.1       pk 	return (EACCES);
    526   1.1       pk }
    527   1.1       pk 
    528   1.1       pk /*
    529   1.9      cgd  * lookup.  this is incredibly complicated in the
    530   1.9      cgd  * general case, however for most pseudo-filesystems
    531   1.9      cgd  * very little needs to be done.
    532   1.9      cgd  *
    533   1.9      cgd  * unless you want to get a migraine, just make sure your
    534   1.9      cgd  * filesystem doesn't do any locking of its own.  otherwise
    535   1.9      cgd  * read and inwardly digest ufs_lookup().
    536   1.9      cgd  */
    537  1.21  mycroft procfs_lookup(ap)
    538  1.21  mycroft 	struct vop_lookup_args *ap;
    539   1.1       pk {
    540  1.21  mycroft 	struct componentname *cnp = ap->a_cnp;
    541  1.21  mycroft 	struct vnode **vpp = ap->a_vpp;
    542  1.21  mycroft 	struct vnode *dvp = ap->a_dvp;
    543  1.21  mycroft 	char *pname = cnp->cn_nameptr;
    544   1.9      cgd 	int error = 0;
    545   1.1       pk 	pid_t pid;
    546   1.1       pk 	struct vnode *nvp;
    547   1.9      cgd 	struct pfsnode *pfs;
    548   1.1       pk 	struct proc *procp;
    549   1.9      cgd 	pfstype pfs_type;
    550   1.9      cgd 	int i;
    551   1.1       pk 
    552  1.21  mycroft 	if (cnp->cn_namelen == 1 && *pname == '.') {
    553  1.21  mycroft 		*vpp = dvp;
    554   1.9      cgd 		VREF(dvp);
    555  1.21  mycroft 		/*VOP_LOCK(dvp);*/
    556   1.9      cgd 		return (0);
    557   1.9      cgd 	}
    558   1.1       pk 
    559  1.21  mycroft 	*vpp = NULL;
    560   1.1       pk 
    561   1.9      cgd 	pfs = VTOPFS(dvp);
    562   1.9      cgd 	switch (pfs->pfs_type) {
    563   1.9      cgd 	case Proot:
    564  1.21  mycroft 		if (cnp->cn_flags & ISDOTDOT)
    565   1.9      cgd 			return (EIO);
    566   1.9      cgd 
    567  1.21  mycroft 		if (CNEQ(cnp, "curproc", 7))
    568  1.21  mycroft 			return (procfs_allocvp(dvp->v_mount, vpp, 0, Pcurproc));
    569  1.12       ws 
    570  1.21  mycroft 		pid = atopid(pname, cnp->cn_namelen);
    571   1.9      cgd 		if (pid == NO_PID)
    572   1.9      cgd 			return (ENOENT);
    573   1.9      cgd 
    574   1.9      cgd 		procp = PFIND(pid);
    575   1.9      cgd 		if (procp == 0)
    576   1.9      cgd 			return (ENOENT);
    577   1.9      cgd 
    578  1.21  mycroft 		return (procfs_allocvp(dvp->v_mount, vpp, pid, Pproc));
    579   1.9      cgd 
    580   1.9      cgd 	case Pproc:
    581  1.21  mycroft 		if (cnp->cn_flags & ISDOTDOT) {
    582  1.21  mycroft 			error = procfs_root(dvp->v_mount, vpp);
    583   1.9      cgd 			return (error);
    584   1.9      cgd 		}
    585   1.1       pk 
    586   1.9      cgd 		procp = PFIND(pfs->pfs_pid);
    587   1.9      cgd 		if (procp == 0)
    588   1.9      cgd 			return (ENOENT);
    589   1.9      cgd 
    590   1.9      cgd 		for (i = 0; i < Nprocent; i++) {
    591   1.9      cgd 			struct pfsnames *dp = &procent[i];
    592   1.9      cgd 
    593  1.21  mycroft 			if (cnp->cn_namelen == dp->d_namlen &&
    594  1.18      cgd 			    bcmp(pname, dp->d_name, dp->d_namlen) == 0 &&
    595  1.18      cgd 			    (dp->d_valid == NULL || (*dp->d_valid)(procp))) {
    596   1.9      cgd 			    	pfs_type = dp->d_pfstype;
    597   1.9      cgd 				goto found;
    598   1.9      cgd 			}
    599   1.9      cgd 		}
    600   1.9      cgd 		return (ENOENT);
    601   1.1       pk 
    602   1.9      cgd 	found:
    603   1.9      cgd 		if (pfs_type == Pfile) {
    604   1.9      cgd 			nvp = procfs_findtextvp(procp);
    605  1.21  mycroft 			if (nvp == NULLVP)
    606  1.21  mycroft 				return (ENXIO);
    607  1.21  mycroft 			VREF(nvp);
    608  1.21  mycroft 			VOP_LOCK(nvp);
    609  1.21  mycroft 			*vpp = nvp;
    610  1.21  mycroft 			return (0);
    611  1.21  mycroft 		}
    612   1.1       pk 
    613  1.21  mycroft 		return (procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
    614  1.21  mycroft 		    pfs_type));
    615   1.1       pk 
    616   1.9      cgd 	default:
    617   1.9      cgd 		return (ENOTDIR);
    618   1.1       pk 	}
    619   1.1       pk }
    620   1.1       pk 
    621  1.18      cgd int
    622  1.21  mycroft procfs_validfile(p)
    623  1.21  mycroft 	struct proc *p;
    624  1.18      cgd {
    625  1.21  mycroft 
    626  1.21  mycroft 	return (procfs_findtextvp(p) != NULLVP);
    627  1.18      cgd }
    628  1.18      cgd 
    629   1.9      cgd /*
    630   1.9      cgd  * readdir returns directory entries from pfsnode (vp).
    631   1.9      cgd  *
    632   1.9      cgd  * the strategy here with procfs is to generate a single
    633   1.9      cgd  * directory entry at a time (struct pfsdent) and then
    634   1.9      cgd  * copy that out to userland using uiomove.  a more efficent
    635   1.9      cgd  * though more complex implementation, would try to minimize
    636   1.9      cgd  * the number of calls to uiomove().  for procfs, this is
    637   1.9      cgd  * hardly worth the added code complexity.
    638   1.9      cgd  *
    639   1.9      cgd  * this should just be done through read()
    640   1.9      cgd  */
    641  1.21  mycroft procfs_readdir(ap)
    642  1.21  mycroft 	struct vop_readdir_args *ap;
    643   1.1       pk {
    644  1.21  mycroft 	struct uio *uio = ap->a_uio;
    645   1.9      cgd 	struct pfsdent d;
    646   1.9      cgd 	struct pfsdent *dp = &d;
    647   1.9      cgd 	struct pfsnode *pfs;
    648   1.9      cgd 	int error;
    649   1.9      cgd 	int count;
    650   1.9      cgd 	int i;
    651   1.9      cgd 
    652  1.21  mycroft 	pfs = VTOPFS(ap->a_vp);
    653   1.1       pk 
    654   1.9      cgd 	if (uio->uio_resid < UIO_MX)
    655   1.9      cgd 		return (EINVAL);
    656   1.9      cgd 	if (uio->uio_offset & (UIO_MX-1))
    657   1.9      cgd 		return (EINVAL);
    658   1.9      cgd 	if (uio->uio_offset < 0)
    659   1.1       pk 		return (EINVAL);
    660   1.1       pk 
    661   1.9      cgd 	error = 0;
    662   1.1       pk 	count = 0;
    663   1.9      cgd 	i = uio->uio_offset / UIO_MX;
    664   1.9      cgd 
    665   1.9      cgd 	switch (pfs->pfs_type) {
    666   1.9      cgd 	/*
    667   1.9      cgd 	 * this is for the process-specific sub-directories.
    668   1.9      cgd 	 * all that is needed to is copy out all the entries
    669   1.9      cgd 	 * from the procent[] table (top of this file).
    670   1.9      cgd 	 */
    671   1.9      cgd 	case Pproc: {
    672  1.21  mycroft 		pid_t pid = pfs->pfs_pid;
    673  1.21  mycroft 		struct pfsnames *dt;
    674  1.21  mycroft 
    675  1.21  mycroft 		for (dt = &procent[i]; i < Nprocent && uio->uio_resid >= UIO_MX;
    676  1.21  mycroft 		     dt++, i++) {
    677  1.21  mycroft 			struct proc *p = PFIND(pid);
    678   1.9      cgd 
    679  1.21  mycroft 			if (p == NULL)
    680   1.9      cgd 				break;
    681   1.9      cgd 
    682  1.21  mycroft 			if (dt->d_valid && (*dt->d_valid)(p) == 0)
    683  1.21  mycroft 				continue;
    684  1.21  mycroft 
    685   1.9      cgd 			dp->d_reclen = UIO_MX;
    686  1.12       ws 			dp->d_fileno = PROCFS_FILENO(pid, dt->d_pfstype);
    687  1.21  mycroft 			dp->d_namlen = dt->d_namlen;
    688  1.21  mycroft 			bcopy(dt->d_name, dp->d_name, dt->d_namlen + 1);
    689  1.21  mycroft 			dp->d_type = dt->d_type;
    690  1.15       ws 
    691  1.21  mycroft 			if (error = uiomove((caddr_t)dp, UIO_MX, uio))
    692   1.9      cgd 				break;
    693   1.6       ws 		}
    694   1.9      cgd 
    695   1.9      cgd 	    	break;
    696   1.9      cgd 
    697   1.9      cgd 	    }
    698   1.9      cgd 
    699   1.9      cgd 	/*
    700   1.9      cgd 	 * this is for the root of the procfs filesystem
    701   1.9      cgd 	 * what is needed is a special entry for "curproc"
    702   1.9      cgd 	 * followed by an entry for each process on allproc
    703   1.9      cgd #ifdef PROCFS_ZOMBIE
    704   1.9      cgd 	 * and zombproc.
    705   1.9      cgd #endif
    706   1.9      cgd 	 */
    707   1.9      cgd 
    708   1.9      cgd 	case Proot: {
    709   1.9      cgd #ifdef PROCFS_ZOMBIE
    710   1.9      cgd 		int doingzomb = 0;
    711   1.9      cgd #endif
    712  1.21  mycroft 		int pcnt = 0;
    713  1.21  mycroft 		volatile struct proc *p = allproc;
    714   1.9      cgd 
    715  1.21  mycroft 	again:
    716  1.21  mycroft 		for (; p && uio->uio_resid >= UIO_MX; i++, pcnt++) {
    717   1.9      cgd 			bzero((char *) dp, UIO_MX);
    718   1.9      cgd 			dp->d_reclen = UIO_MX;
    719   1.9      cgd 
    720   1.9      cgd 			switch (i) {
    721  1.12       ws 			case 0:		/* `.' */
    722  1.12       ws 			case 1:		/* `..' */
    723  1.12       ws 				dp->d_fileno = PROCFS_FILENO(0, Proot);
    724  1.12       ws 				dp->d_namlen = i + 1;
    725  1.21  mycroft 				bcopy("..", dp->d_name, dp->d_namlen);
    726  1.21  mycroft 				dp->d_name[i + 1] = '\0';
    727  1.21  mycroft 				dp->d_type = DT_DIR;
    728  1.12       ws 				break;
    729  1.21  mycroft 
    730  1.12       ws 			case 2:
    731  1.21  mycroft 				dp->d_fileno = PROCFS_FILENO(0, Pcurproc);
    732   1.9      cgd 				dp->d_namlen = 7;
    733  1.21  mycroft 				bcopy("curproc", dp->d_name, 8);
    734  1.21  mycroft 				dp->d_type = DT_LNK;
    735   1.9      cgd 				break;
    736   1.9      cgd 
    737   1.9      cgd 			default:
    738  1.21  mycroft 				while (pcnt < i) {
    739  1.21  mycroft 					pcnt++;
    740  1.21  mycroft 					p = p->p_next;
    741  1.21  mycroft 					if (!p)
    742  1.21  mycroft 						goto done;
    743   1.9      cgd 				}
    744  1.21  mycroft 				dp->d_fileno = PROCFS_FILENO(p->p_pid, Pproc);
    745  1.21  mycroft 				dp->d_namlen = sprintf(dp->d_name, "%ld",
    746  1.21  mycroft 				    (long)p->p_pid);
    747  1.21  mycroft 				dp->d_type = DT_REG;
    748  1.20      cgd 				p = p->p_next;
    749   1.9      cgd 				break;
    750   1.1       pk 			}
    751  1.21  mycroft 
    752  1.21  mycroft 			if (error = uiomove((caddr_t)dp, UIO_MX, uio))
    753   1.9      cgd 				break;
    754   1.1       pk 		}
    755  1.21  mycroft 	done:
    756  1.21  mycroft 
    757  1.21  mycroft #ifdef PROCFS_ZOMBIE
    758  1.21  mycroft 		if (p == 0 && doingzomb == 0) {
    759  1.21  mycroft 			doingzomb = 1;
    760  1.21  mycroft 			p = zombproc;
    761  1.21  mycroft 			goto again;
    762  1.21  mycroft 		}
    763  1.21  mycroft #endif
    764   1.1       pk 
    765   1.9      cgd 		break;
    766   1.9      cgd 
    767   1.9      cgd 	    }
    768   1.9      cgd 
    769   1.9      cgd 	default:
    770   1.9      cgd 		error = ENOTDIR;
    771   1.9      cgd 		break;
    772   1.1       pk 	}
    773   1.9      cgd 
    774   1.9      cgd 	uio->uio_offset = i * UIO_MX;
    775   1.1       pk 
    776   1.9      cgd 	return (error);
    777  1.12       ws }
    778  1.12       ws 
    779  1.12       ws /*
    780  1.21  mycroft  * readlink reads the link of `curproc'
    781  1.12       ws  */
    782  1.21  mycroft procfs_readlink(ap)
    783  1.21  mycroft 	struct vop_readlink_args *ap;
    784  1.12       ws {
    785  1.21  mycroft 	struct uio *uio = ap->a_uio;
    786  1.12       ws 	char buf[16];		/* should be enough */
    787  1.12       ws 	int len;
    788  1.12       ws 
    789  1.21  mycroft 	if (VTOPFS(ap->a_vp)->pfs_fileno != PROCFS_FILENO(0, Pcurproc))
    790  1.21  mycroft 		return (EINVAL);
    791  1.12       ws 
    792  1.21  mycroft 	len = sprintf(buf, "%ld", (long)curproc->p_pid);
    793  1.12       ws 
    794  1.21  mycroft 	return (uiomove((caddr_t)buf, len, ap->a_uio));
    795   1.1       pk }
    796   1.1       pk 
    797   1.1       pk /*
    798   1.9      cgd  * convert decimal ascii to pid_t
    799   1.1       pk  */
    800   1.9      cgd static pid_t
    801   1.9      cgd atopid(b, len)
    802   1.9      cgd 	const char *b;
    803   1.9      cgd 	u_int len;
    804   1.1       pk {
    805   1.9      cgd 	pid_t p = 0;
    806   1.1       pk 
    807   1.1       pk 	while (len--) {
    808   1.9      cgd 		char c = *b++;
    809   1.1       pk 		if (c < '0' || c > '9')
    810   1.9      cgd 			return (NO_PID);
    811   1.9      cgd 		p = 10 * p + (c - '0');
    812   1.9      cgd 		if (p > PID_MAX)
    813   1.9      cgd 			return (NO_PID);
    814   1.1       pk 	}
    815   1.9      cgd 
    816   1.9      cgd 	return (p);
    817   1.1       pk }
    818   1.9      cgd 
    819   1.9      cgd /*
    820   1.9      cgd  * procfs vnode operations.
    821   1.9      cgd  */
    822  1.21  mycroft int (**procfs_vnodeop_p)();
    823  1.21  mycroft struct vnodeopv_entry_desc procfs_vnodeop_entries[] = {
    824  1.21  mycroft 	{ &vop_default_desc, vn_default_error },
    825  1.21  mycroft 	{ &vop_lookup_desc, procfs_lookup },		/* lookup */
    826  1.21  mycroft 	{ &vop_create_desc, procfs_create },		/* create */
    827  1.21  mycroft 	{ &vop_mknod_desc, procfs_mknod },		/* mknod */
    828  1.21  mycroft 	{ &vop_open_desc, procfs_open },		/* open */
    829  1.21  mycroft 	{ &vop_close_desc, procfs_close },		/* close */
    830  1.21  mycroft 	{ &vop_access_desc, procfs_access },		/* access */
    831  1.21  mycroft 	{ &vop_getattr_desc, procfs_getattr },		/* getattr */
    832  1.21  mycroft 	{ &vop_setattr_desc, procfs_setattr },		/* setattr */
    833  1.21  mycroft 	{ &vop_read_desc, procfs_read },		/* read */
    834  1.21  mycroft 	{ &vop_write_desc, procfs_write },		/* write */
    835  1.21  mycroft 	{ &vop_ioctl_desc, procfs_ioctl },		/* ioctl */
    836  1.21  mycroft 	{ &vop_select_desc, procfs_select },		/* select */
    837  1.21  mycroft 	{ &vop_mmap_desc, procfs_mmap },		/* mmap */
    838  1.21  mycroft 	{ &vop_fsync_desc, procfs_fsync },		/* fsync */
    839  1.21  mycroft 	{ &vop_seek_desc, procfs_seek },		/* seek */
    840  1.21  mycroft 	{ &vop_remove_desc, procfs_remove },		/* remove */
    841  1.21  mycroft 	{ &vop_link_desc, procfs_link },		/* link */
    842  1.21  mycroft 	{ &vop_rename_desc, procfs_rename },		/* rename */
    843  1.21  mycroft 	{ &vop_mkdir_desc, procfs_mkdir },		/* mkdir */
    844  1.21  mycroft 	{ &vop_rmdir_desc, procfs_rmdir },		/* rmdir */
    845  1.21  mycroft 	{ &vop_symlink_desc, procfs_symlink },		/* symlink */
    846  1.21  mycroft 	{ &vop_readdir_desc, procfs_readdir },		/* readdir */
    847  1.21  mycroft 	{ &vop_readlink_desc, procfs_readlink },	/* readlink */
    848  1.21  mycroft 	{ &vop_abortop_desc, procfs_abortop },		/* abortop */
    849  1.21  mycroft 	{ &vop_inactive_desc, procfs_inactive },	/* inactive */
    850  1.21  mycroft 	{ &vop_reclaim_desc, procfs_reclaim },		/* reclaim */
    851  1.21  mycroft 	{ &vop_lock_desc, procfs_lock },		/* lock */
    852  1.21  mycroft 	{ &vop_unlock_desc, procfs_unlock },		/* unlock */
    853  1.21  mycroft 	{ &vop_bmap_desc, procfs_bmap },		/* bmap */
    854  1.21  mycroft 	{ &vop_strategy_desc, procfs_strategy },	/* strategy */
    855  1.21  mycroft 	{ &vop_print_desc, procfs_print },		/* print */
    856  1.21  mycroft 	{ &vop_islocked_desc, procfs_islocked },	/* islocked */
    857  1.21  mycroft 	{ &vop_pathconf_desc, procfs_pathconf },	/* pathconf */
    858  1.21  mycroft 	{ &vop_advlock_desc, procfs_advlock },		/* advlock */
    859  1.21  mycroft 	{ &vop_blkatoff_desc, procfs_blkatoff },	/* blkatoff */
    860  1.21  mycroft 	{ &vop_valloc_desc, procfs_valloc },		/* valloc */
    861  1.21  mycroft 	{ &vop_vfree_desc, procfs_vfree },		/* vfree */
    862  1.21  mycroft 	{ &vop_truncate_desc, procfs_truncate },	/* truncate */
    863  1.21  mycroft 	{ &vop_update_desc, procfs_update },		/* update */
    864  1.21  mycroft 	{ (struct vnodeop_desc*)NULL, (int(*)())NULL }
    865   1.9      cgd };
    866  1.21  mycroft struct vnodeopv_desc procfs_vnodeop_opv_desc =
    867  1.21  mycroft 	{ &procfs_vnodeop_p, procfs_vnodeop_entries };
    868