xref: /csrg-svn/sys/kern/vfs_lookup.c (revision 41337)
123401Smckusick /*
237715Smckusick  * Copyright (c) 1982, 1986, 1989 Regents of the University of California.
337715Smckusick  * All rights reserved.
423401Smckusick  *
537715Smckusick  * Redistribution and use in source and binary forms are permitted
637715Smckusick  * provided that the above copyright notice and this paragraph are
737715Smckusick  * duplicated in all such forms and that any documentation,
837715Smckusick  * advertising materials, and other materials related to such
937715Smckusick  * distribution and use acknowledge that the software was developed
1037715Smckusick  * by the University of California, Berkeley.  The name of the
1137715Smckusick  * University may not be used to endorse or promote products derived
1237715Smckusick  * from this software without specific prior written permission.
1337715Smckusick  * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
1437715Smckusick  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
1537715Smckusick  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
1637715Smckusick  *
17*41337Smckusick  *	@(#)vfs_lookup.c	7.20 (Berkeley) 05/03/90
1823401Smckusick  */
1930Sbill 
2017100Sbloom #include "param.h"
2137715Smckusick #include "time.h"
2237715Smckusick #include "namei.h"
2337715Smckusick #include "vnode.h"
2417100Sbloom #include "mount.h"
2537715Smckusick #include "errno.h"
2631650Smckusick #include "malloc.h"
2737715Smckusick 
2837582Smarc #ifdef KTRACE
2937715Smckusick #include "user.h"
3037582Smarc #include "proc.h"
3137582Smarc #include "ktrace.h"
3237582Smarc #endif
3330Sbill 
3430Sbill /*
3527268Smckusick  * Convert a pathname into a pointer to a locked inode.
367534Sroot  * This is a very central and rather complicated routine.
3730Sbill  *
3837715Smckusick  * The flag argument is LOOKUP, CREATE, RENAME, or DELETE depending on
3937715Smckusick  * whether the name is to be looked up, created, renamed, or deleted.
4037715Smckusick  * When CREATE, RENAME, or DELETE is specified, information usable in
4137715Smckusick  * creating, renaming, or deleting a directory entry may be calculated.
4237715Smckusick  * If flag has LOCKPARENT or'ed into it and the target of the pathname
4337715Smckusick  * exists, namei returns both the target and its parent directory locked.
4437715Smckusick  * When creating or renaming and LOCKPARENT is specified, the target may not
4537715Smckusick  * be ".".  When deleting and LOCKPARENT is specified, the target may be ".".
469166Ssam  *
4716688Smckusick  * The FOLLOW flag is set when symbolic links are to be followed
489166Ssam  * when they occur at the end of the name translation process.
4927268Smckusick  * Symbolic links are always followed for all other pathname
5027268Smckusick  * components other than the last.
519166Ssam  *
5227268Smckusick  * The segflg defines whether the name is to be copied from user
5327268Smckusick  * space or kernel space.
5427268Smckusick  *
5515798Smckusick  * Overall outline of namei:
5615798Smckusick  *
577534Sroot  *	copy in name
587534Sroot  *	get starting directory
597534Sroot  * dirloop:
6016688Smckusick  *	copy next component of name to ndp->ni_dent
617534Sroot  *	handle degenerate case where name is null string
6237715Smckusick  *	if .. and on mounted filesys, find parent
6337715Smckusick  *	call lookup routine for next component name
6437715Smckusick  *	  directory vnode returned in ni_dvp, unlocked unless LOCKPARENT set
6537715Smckusick  *	  component vnode returned in ni_vp (if it exists), locked.
667534Sroot  *	if symbolic link, massage name in buffer and continue at dirloop
6737715Smckusick  *	if result inode is mounted on, find mounted on vnode
687534Sroot  *	if more components of name, do next level at dirloop
6937715Smckusick  *	return the answer in ni_vp as locked vnode;
7037715Smckusick  *	  if LOCKPARENT set, return locked parent in ni_dvp
719166Ssam  *
7237715Smckusick  * NOTE: (LOOKUP | LOCKPARENT) currently returns the parent vnode unlocked.
7330Sbill  */
7416688Smckusick namei(ndp)
7516688Smckusick 	register struct nameidata *ndp;
7630Sbill {
777534Sroot 	register char *cp;		/* pointer into pathname argument */
7837715Smckusick 	register struct vnode *dp = 0;	/* the directory we are searching */
7937715Smckusick 	register int i;		   	/* Temp counter */
8037715Smckusick 	struct vnode *tdp;		/* saved dp */
8137715Smckusick 	struct mount *mp;		/* mount table entry */
8218109Smckusick 	int docache;			/* == 0 do not cache last component */
8337715Smckusick 	int flag;			/* LOOKUP, CREATE, RENAME or DELETE */
8437715Smckusick 	int wantparent;			/* 1 => wantparent or lockparent flag */
8538001Smckusick 	int lockparent;			/* 1 => lockparent flag */
86*41337Smckusick 	int getbuf;			/* 1 => Malloc a pathname buffer */
87*41337Smckusick 	int rdonly;			/* mounted read-only flag bit(s) */
8837715Smckusick 	int error = 0;
8930Sbill 
90*41337Smckusick 	/*
91*41337Smckusick 	 * Setup: break out flag bits into variables.
92*41337Smckusick 	 */
9339758Smckusick 	ndp->ni_dvp = NULL;
9437715Smckusick 	flag = ndp->ni_nameiop & OPFLAG;
9537715Smckusick 	wantparent = ndp->ni_nameiop & (LOCKPARENT|WANTPARENT);
9638001Smckusick 	lockparent = ndp->ni_nameiop & LOCKPARENT;
9716688Smckusick 	docache = (ndp->ni_nameiop & NOCACHE) ^ NOCACHE;
98*41337Smckusick 	getbuf = (ndp->ni_nameiop & HASBUF) ^ HASBUF;
9937715Smckusick 	if (flag == DELETE || wantparent)
10015798Smckusick 		docache = 0;
101*41337Smckusick 	rdonly = M_RDONLY;
102*41337Smckusick 	if (ndp->ni_nameiop & REMOTE)
103*41337Smckusick 		rdonly |= M_EXRDONLY;
10430Sbill 	/*
1057534Sroot 	 * Get a buffer for the name to be translated, and copy the
1067534Sroot 	 * name into the buffer.
1075972Swnj 	 */
108*41337Smckusick 	if (getbuf) {
109*41337Smckusick 		MALLOC(ndp->ni_pnbuf, caddr_t, MAXPATHLEN, M_NAMEI, M_WAITOK);
110*41337Smckusick 		if (ndp->ni_segflg == UIO_SYSSPACE)
111*41337Smckusick 			error = copystr(ndp->ni_dirp, ndp->ni_pnbuf,
112*41337Smckusick 				    MAXPATHLEN, &ndp->ni_pathlen);
113*41337Smckusick 		else
114*41337Smckusick 			error = copyinstr(ndp->ni_dirp, ndp->ni_pnbuf,
115*41337Smckusick 				    MAXPATHLEN, &ndp->ni_pathlen);
116*41337Smckusick 		if (error) {
117*41337Smckusick 			free(ndp->ni_pnbuf, M_NAMEI);
118*41337Smckusick 			ndp->ni_vp = NULL;
119*41337Smckusick 			return (error);
120*41337Smckusick 		}
121*41337Smckusick 		ndp->ni_ptr = ndp->ni_pnbuf;
1225972Swnj 	}
12337715Smckusick 	ndp->ni_loopcnt = 0;
12437715Smckusick 	dp = ndp->ni_cdir;
12538347Smckusick 	VREF(dp);
12637582Smarc #ifdef KTRACE
12737582Smarc 	if (KTRPOINT(u.u_procp, KTR_NAMEI))
12837715Smckusick 		ktrnamei(u.u_procp->p_tracep, ndp->ni_pnbuf);
12937582Smarc #endif
1307534Sroot 
13137715Smckusick start:
1325972Swnj 	/*
1337534Sroot 	 * Get starting directory.
13437715Smckusick 	 * Done at start of translation and after symbolic link.
13530Sbill 	 */
13637715Smckusick 	if (*ndp->ni_ptr == '/') {
13737715Smckusick 		vrele(dp);
13837715Smckusick 		while (*ndp->ni_ptr == '/') {
13937715Smckusick 			ndp->ni_ptr++;
14037715Smckusick 			ndp->ni_pathlen--;
14137715Smckusick 		}
14237715Smckusick 		if ((dp = ndp->ni_rdir) == NULL)
14330Sbill 			dp = rootdir;
14438347Smckusick 		VREF(dp);
14537715Smckusick 	}
14637715Smckusick 	VOP_LOCK(dp);
14718027Smckusick 	ndp->ni_endoff = 0;
1487534Sroot 
1497534Sroot 	/*
1507534Sroot 	 * We come to dirloop to search a new directory.
1517534Sroot 	 */
1526571Smckusic dirloop:
15330Sbill 	/*
15416688Smckusick 	 * Copy next component of name to ndp->ni_dent.
15537715Smckusick 	 * XXX kern_exec looks at d_name
15637715Smckusick 	 * ??? The ni_hash value may be useful for vfs_cache
15737715Smckusick 	 * XXX There must be the last component of the filename left
15837715Smckusick 	 * somewhere accessible via. ndp for NFS (and any other stateless file
15937715Smckusick 	 * systems) in case they are doing a CREATE. The "Towards a..." noted
16037715Smckusick 	 * that ni_ptr would be left pointing to the last component, but since
16137715Smckusick 	 * the ni_pnbuf gets free'd, that is not a good idea.
1627534Sroot 	 */
163*41337Smckusick 	if (getbuf) {
164*41337Smckusick 		ndp->ni_hash = 0;
165*41337Smckusick 		for (cp = ndp->ni_ptr, i = 0; *cp != 0 && *cp != '/'; cp++) {
166*41337Smckusick 			if (i >= MAXNAMLEN) {
167*41337Smckusick 				error = ENAMETOOLONG;
168*41337Smckusick 				goto bad;
169*41337Smckusick 			}
170*41337Smckusick 			if (*cp & 0200)
171*41337Smckusick 				if ((*cp&0377) == ('/'|0200) ||
172*41337Smckusick 				    flag != DELETE) {
173*41337Smckusick 					error = EINVAL;
174*41337Smckusick 					goto bad;
175*41337Smckusick 				}
176*41337Smckusick 			ndp->ni_dent.d_name[i++] = *cp;
177*41337Smckusick 			ndp->ni_hash += (unsigned char)*cp * i;
17837743Smckusick 		}
179*41337Smckusick 		ndp->ni_namelen = i;
180*41337Smckusick 		ndp->ni_dent.d_namlen = i;
181*41337Smckusick 		ndp->ni_dent.d_name[i] = '\0';
182*41337Smckusick 		ndp->ni_pathlen -= i;
183*41337Smckusick 		ndp->ni_next = cp;
18437743Smckusick #ifdef NAMEI_DIAGNOSTIC
185*41337Smckusick 		printf("{%s}: ", ndp->ni_dent.d_name);
18637743Smckusick #endif
1875972Swnj 	}
188*41337Smckusick 	cp = ndp->ni_next;
18937715Smckusick 	ndp->ni_makeentry = 1;
19018109Smckusick 	if (*cp == '\0' && docache == 0)
19137715Smckusick 		ndp->ni_makeentry = 0;
19237715Smckusick 	ndp->ni_isdotdot = (ndp->ni_namelen == 2 &&
19337715Smckusick 		ndp->ni_dent.d_name[1] == '.' && ndp->ni_dent.d_name[0] == '.');
1947534Sroot 
1957534Sroot 	/*
1967534Sroot 	 * Check for degenerate name (e.g. / or "")
1977534Sroot 	 * which is a way of talking about a directory,
1987534Sroot 	 * e.g. like "/." or ".".
1997534Sroot 	 */
20037715Smckusick 	if (ndp->ni_ptr[0] == '\0') {
20137715Smckusick 		if (flag != LOOKUP || wantparent) {
20237715Smckusick 			error = EISDIR;
2037534Sroot 			goto bad;
2045972Swnj 		}
205*41337Smckusick 		if (getbuf)
206*41337Smckusick 			free(ndp->ni_pnbuf, M_NAMEI);
20737715Smckusick 		if (!(ndp->ni_nameiop & LOCKLEAF))
20837715Smckusick 			VOP_UNLOCK(dp);
20937715Smckusick 		ndp->ni_vp = dp;
21037715Smckusick 		return (0);
2115972Swnj 	}
2127534Sroot 
2136571Smckusic 	/*
21437715Smckusick 	 * Handle "..": two special cases.
21537715Smckusick 	 * 1. If at root directory (e.g. after chroot)
21637715Smckusick 	 *    then ignore it so can't get out.
21737715Smckusick 	 * 2. If this vnode is the root of a mounted
21837715Smckusick 	 *    file system, then replace it with the
21937715Smckusick 	 *    vnode which was mounted on so we take the
22037715Smckusick 	 *    .. in the other file system.
22136547Smckusick 	 */
22237715Smckusick 	if (ndp->ni_isdotdot) {
22336547Smckusick 		for (;;) {
22437715Smckusick 			if (dp == ndp->ni_rdir || dp == rootdir) {
22537715Smckusick 				ndp->ni_dvp = dp;
22638390Smckusick 				ndp->ni_vp = dp;
22738347Smckusick 				VREF(dp);
22837715Smckusick 				goto nextname;
22936547Smckusick 			}
230*41337Smckusick 			if ((dp->v_flag & VROOT) == 0 ||
231*41337Smckusick 				(ndp->ni_nameiop & NOCROSSMOUNT))
23236547Smckusick 				break;
23337715Smckusick 			tdp = dp;
23437715Smckusick 			dp = dp->v_mount->m_vnodecovered;
23537715Smckusick 			vput(tdp);
23638390Smckusick 			VREF(dp);
23737715Smckusick 			VOP_LOCK(dp);
23836547Smckusick 		}
23936547Smckusick 	}
24036547Smckusick 
24136547Smckusick 	/*
24215798Smckusick 	 * We now have a segment name to search for, and a directory to search.
24315798Smckusick 	 */
24437715Smckusick 	if (error = VOP_LOOKUP(dp, ndp)) {
24537715Smckusick 		if (ndp->ni_vp != NULL)
24637715Smckusick 			panic("leaf should be empty");
24737582Smarc #ifdef NAMEI_DIAGNOSTIC
24837715Smckusick 		printf("not found\n");
24937582Smarc #endif
25038581Smckusick 		if (flag == LOOKUP || flag == DELETE ||
25138581Smckusick 		    error != ENOENT || *cp != 0)
25238581Smckusick 			goto bad;
2535972Swnj 		/*
25437715Smckusick 		 * If creating and at end of pathname, then can consider
25537715Smckusick 		 * allowing file to be created.
2565972Swnj 		 */
257*41337Smckusick 		if (ndp->ni_dvp->v_mount->m_flag & rdonly) {
25837715Smckusick 			error = EROFS;
2597534Sroot 			goto bad;
26038581Smckusick 		}
2615972Swnj 		/*
26237715Smckusick 		 * We return with ni_vp NULL to indicate that the entry
26337715Smckusick 		 * doesn't currently exist, leaving a pointer to the
26437715Smckusick 		 * (possibly locked) directory inode in ndp->ni_dvp.
2655972Swnj 		 */
266*41337Smckusick 		if (getbuf)
267*41337Smckusick 			FREE(ndp->ni_pnbuf, M_NAMEI);
26837715Smckusick 		return (0);	/* should this be ENOENT? */
2697534Sroot 	}
27037582Smarc #ifdef NAMEI_DIAGNOSTIC
27137715Smckusick 	printf("found\n");
27237582Smarc #endif
2737534Sroot 
2747534Sroot 	/*
27537715Smckusick 	 * Check for symbolic link
2767534Sroot 	 */
27737715Smckusick 	dp = ndp->ni_vp;
27837715Smckusick 	if ((dp->v_type == VLNK) &&
27937715Smckusick 	    ((ndp->ni_nameiop & FOLLOW) || *ndp->ni_next == '/')) {
28037715Smckusick 		struct iovec aiov;
28137715Smckusick 		struct uio auio;
28237715Smckusick 		int linklen;
2837534Sroot 
284*41337Smckusick 		if (!getbuf)
285*41337Smckusick 			panic("namei: unexpected symlink");
28637715Smckusick 		if (++ndp->ni_loopcnt > MAXSYMLINKS) {
28737715Smckusick 			error = ELOOP;
28837715Smckusick 			goto bad2;
28937715Smckusick 		}
29038215Smckusick 		if (ndp->ni_pathlen > 1)
29137715Smckusick 			MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
2927534Sroot 		else
29337715Smckusick 			cp = ndp->ni_pnbuf;
29437715Smckusick 		aiov.iov_base = cp;
29537715Smckusick 		aiov.iov_len = MAXPATHLEN;
29637715Smckusick 		auio.uio_iov = &aiov;
29737715Smckusick 		auio.uio_iovcnt = 1;
29837715Smckusick 		auio.uio_offset = 0;
29937715Smckusick 		auio.uio_rw = UIO_READ;
30037715Smckusick 		auio.uio_segflg = UIO_SYSSPACE;
30137715Smckusick 		auio.uio_resid = MAXPATHLEN;
30237715Smckusick 		if (error = VOP_READLINK(dp, &auio, ndp->ni_cred)) {
30338215Smckusick 			if (ndp->ni_pathlen > 1)
30437715Smckusick 				free(cp, M_NAMEI);
30537715Smckusick 			goto bad2;
3069166Ssam 		}
30737715Smckusick 		linklen = MAXPATHLEN - auio.uio_resid;
30837715Smckusick 		if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
30938215Smckusick 			if (ndp->ni_pathlen > 1)
31037715Smckusick 				free(cp, M_NAMEI);
31137715Smckusick 			error = ENAMETOOLONG;
31212011Smckusick 			goto bad2;
31315798Smckusick 		}
31438215Smckusick 		if (ndp->ni_pathlen > 1) {
31538215Smckusick 			bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
31637715Smckusick 			FREE(ndp->ni_pnbuf, M_NAMEI);
31737715Smckusick 			ndp->ni_pnbuf = cp;
31838215Smckusick 		} else
31938001Smckusick 			ndp->ni_pnbuf[linklen] = '\0';
32037715Smckusick 		ndp->ni_ptr = cp;
32137715Smckusick 		vput(dp);
32237715Smckusick 		dp = ndp->ni_dvp;
32338393Smckusick 		if (lockparent && ndp->ni_pathlen == 1)
32438001Smckusick 			VOP_UNLOCK(dp);
32538393Smckusick 		ndp->ni_pathlen += linklen;
32637715Smckusick 		goto start;
32715798Smckusick 	}
32815798Smckusick 
3297534Sroot 	/*
33037715Smckusick 	 * Check to see if the vnode has been mounted on;
33137715Smckusick 	 * if so find the root of the mounted file system.
3327534Sroot 	 */
33337715Smckusick mntloop:
334*41337Smckusick 	while (dp->v_type == VDIR && (mp = dp->v_mountedhere) &&
335*41337Smckusick 	       (ndp->ni_nameiop & NOCROSSMOUNT) == 0) {
33637715Smckusick 		while(mp->m_flag & M_MLOCK) {
33737715Smckusick 			mp->m_flag |= M_MWAIT;
33837715Smckusick 			sleep((caddr_t)mp, PVFS);
33937715Smckusick 			goto mntloop;
34021014Smckusick 		}
34137715Smckusick 		error = VFS_ROOT(dp->v_mountedhere, &tdp);
34237715Smckusick 		if (error)
3437534Sroot 			goto bad2;
34437715Smckusick 		vput(dp);
34537715Smckusick 		ndp->ni_vp = dp = tdp;
34630Sbill 	}
3477534Sroot 
34837715Smckusick nextname:
34930Sbill 	/*
3507534Sroot 	 * Not a symbolic link.  If more pathname,
3517534Sroot 	 * continue at next component, else return.
35230Sbill 	 */
35337715Smckusick 	ndp->ni_ptr = ndp->ni_next;
35437715Smckusick 	if (*ndp->ni_ptr == '/') {
35537715Smckusick 		while (*ndp->ni_ptr == '/') {
35637715Smckusick 			ndp->ni_ptr++;
35737715Smckusick 			ndp->ni_pathlen--;
35837715Smckusick 		}
35937715Smckusick 		vrele(ndp->ni_dvp);
3607534Sroot 		goto dirloop;
36130Sbill 	}
3627534Sroot 	/*
36338400Smckusick 	 * Check for read-only file systems.
3647534Sroot 	 */
36538400Smckusick 	if (flag == DELETE || flag == RENAME) {
36638267Smckusick 		/*
36738400Smckusick 		 * Disallow directory write attempts on read-only
36838400Smckusick 		 * file systems.
36938267Smckusick 		 */
370*41337Smckusick 		if ((dp->v_mount->m_flag & rdonly) ||
371*41337Smckusick 		    (wantparent && (ndp->ni_dvp->v_mount->m_flag & rdonly))) {
37238267Smckusick 			error = EROFS;
37338267Smckusick 			goto bad2;
37438267Smckusick 		}
37538267Smckusick 	}
37637715Smckusick 	if (!wantparent)
37737715Smckusick 		vrele(ndp->ni_dvp);
37837715Smckusick 	if ((ndp->ni_nameiop & LOCKLEAF) == 0)
37937715Smckusick 		VOP_UNLOCK(dp);
380*41337Smckusick 	if (getbuf)
381*41337Smckusick 		FREE(ndp->ni_pnbuf, M_NAMEI);
38237715Smckusick 	return (0);
3837534Sroot 
38437715Smckusick bad2:
38538001Smckusick 	if (lockparent && *ndp->ni_next == '\0')
38638001Smckusick 		VOP_UNLOCK(ndp->ni_dvp);
38737715Smckusick 	vrele(ndp->ni_dvp);
38837715Smckusick bad:
38937715Smckusick 	vput(dp);
39037715Smckusick 	ndp->ni_vp = NULL;
391*41337Smckusick 	if (getbuf)
392*41337Smckusick 		FREE(ndp->ni_pnbuf, M_NAMEI);
39310849Ssam 	return (error);
3945972Swnj }
395