Please note that diffs are not public domain; they are subject to the copyright notices on the relevant files. =================================================================== RCS file: /ftp/cvs/cvsroot/src/sys/ufs/ufs/ufs_lookup.c,v rcsdiff: /ftp/cvs/cvsroot/src/sys/ufs/ufs/ufs_lookup.c,v: warning: Unknown phrases like `commitid ...;' are present. retrieving revision 1.12 retrieving revision 1.98.4.2 diff -u -p -r1.12 -r1.98.4.2 --- src/sys/ufs/ufs/ufs_lookup.c 1997/05/12 19:04:16 1.12 +++ src/sys/ufs/ufs/ufs_lookup.c 2008/12/13 01:15:41 1.98.4.2 @@ -1,4 +1,4 @@ -/* $NetBSD: ufs_lookup.c,v 1.12 1997/05/12 19:04:16 kleink Exp $ */ +/* $NetBSD: ufs_lookup.c,v 1.98.4.2 2008/12/13 01:15:41 haad Exp $ */ /* * Copyright (c) 1989, 1993 @@ -17,11 +17,7 @@ * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. - * 3. All advertising materials mentioning features or use of this software - * must display the following acknowledgement: - * This product includes software developed by the University of - * California, Berkeley and its contributors. - * 4. Neither the name of the University nor the names of its contributors + * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * @@ -40,28 +36,46 @@ * @(#)ufs_lookup.c 8.9 (Berkeley) 8/11/94 */ +#include +__KERNEL_RCSID(0, "$NetBSD: ufs_lookup.c,v 1.98.4.2 2008/12/13 01:15:41 haad Exp $"); + +#ifdef _KERNEL_OPT +#include "opt_ffs.h" +#include "fs_ffs.h" +#endif + #include #include #include #include #include +#include #include #include +#include +#include +#include +#include +#include +#include -#include #include #include +#ifdef UFS_DIRHASH +#include +#endif #include #include +#include +#include -struct nchstats nchstats; #ifdef DIAGNOSTIC int dirchk = 1; #else int dirchk = 0; #endif -#define FSFMT(vp) ((vp)->v_mount->mnt_maxsymlinklen <= 0) +#define FSFMT(vp) (((vp)->v_mount->mnt_iflag & IMNT_DTYPE) == 0) /* * Convert a component of a pathname into a pointer to a locked inode. @@ -97,18 +111,17 @@ int dirchk = 0; * nor deleting, add name to cache */ int -ufs_lookup(v) - void *v; +ufs_lookup(void *v) { struct vop_lookup_args /* { struct vnode *a_dvp; struct vnode **a_vpp; struct componentname *a_cnp; } */ *ap = v; - register struct vnode *vdp; /* vnode for directory being searched */ - register struct inode *dp; /* inode for directory being searched */ + struct vnode *vdp = ap->a_dvp; /* vnode for directory being searched */ + struct inode *dp = VTOI(vdp); /* inode for directory being searched */ struct buf *bp; /* a buffer of directory entries */ - register struct direct *ep; /* the current directory entry */ + struct direct *ep; /* the current directory entry */ int entryoffsetinblock; /* offset of ep in bp's buffer */ enum {NONE, COMPACT, FOUND} slotstatus; doff_t slotoffset; /* offset of area with free space */ @@ -122,29 +135,33 @@ ufs_lookup(v) struct vnode *tdp; /* returned by VFS_VGET */ doff_t enduseful; /* pointer past last used dir slot */ u_long bmask; /* block offset mask */ - int lockparent; /* 1 => lockparent flag is set */ - int wantparent; /* 1 => wantparent or lockparent flag */ int namlen, error; struct vnode **vpp = ap->a_vpp; struct componentname *cnp = ap->a_cnp; - struct ucred *cred = cnp->cn_cred; - int flags = cnp->cn_flags; + kauth_cred_t cred = cnp->cn_cred; + int flags; int nameiop = cnp->cn_nameiop; + struct ufsmount *ump = dp->i_ump; + const int needswap = UFS_MPNEEDSWAP(ump); + int dirblksiz = ump->um_dirblksiz; + ino_t foundino; + + flags = cnp->cn_flags; bp = NULL; slotoffset = -1; *vpp = NULL; - vdp = ap->a_dvp; - dp = VTOI(vdp); - lockparent = flags & LOCKPARENT; - wantparent = flags & (LOCKPARENT|WANTPARENT); - + endsearch = 0; /* silence compiler warning */ /* * Check accessiblity of directory. */ - if ((error = VOP_ACCESS(vdp, VEXEC, cred, cnp->cn_proc)) != 0) + if ((error = VOP_ACCESS(vdp, VEXEC, cred)) != 0) return (error); + if ((flags & ISLASTCN) && (vdp->v_mount->mnt_flag & MNT_RDONLY) && + (nameiop == DELETE || nameiop == RENAME)) + return (EROFS); + /* * We now have a segment name to search for, and a directory to search. * @@ -152,51 +169,12 @@ ufs_lookup(v) * check the name cache to see if the directory/name pair * we are looking for is known already. */ - if ((error = cache_lookup(vdp, vpp, cnp)) != 0) { - int vpid; /* capability number of vnode */ - - if (error == ENOENT) - return (error); - /* - * Get the next vnode in the path. - * See comment below starting `Step through' for - * an explaination of the locking protocol. - */ - pdp = vdp; - dp = VTOI(*vpp); - vdp = *vpp; - vpid = vdp->v_id; - if (pdp == vdp) { /* lookup on "." */ - VREF(vdp); - error = 0; - } else if (flags & ISDOTDOT) { - VOP_UNLOCK(pdp); - error = vget(vdp, 1); - if (!error && lockparent && (flags & ISLASTCN)) - error = VOP_LOCK(pdp); - } else { - error = vget(vdp, 1); - if (!lockparent || error || !(flags & ISLASTCN)) - VOP_UNLOCK(pdp); - } - /* - * Check that the capability number did not change - * while we were waiting for the lock. - */ - if (!error) { - if (vpid == vdp->v_id) - return (0); - vput(vdp); - if (lockparent && pdp != vdp && (flags & ISLASTCN)) - VOP_UNLOCK(pdp); - } - if ((error = VOP_LOCK(pdp)) != 0) - return (error); - vdp = pdp; - dp = VTOI(pdp); - *vpp = NULL; + if ((error = cache_lookup(vdp, vpp, cnp)) >= 0) { + return (error); } + fstrans_start(vdp->v_mount, FSTRANS_SHARED); + /* * Suppress search for slots unless creating * file and at end of pathname, in which case @@ -208,8 +186,7 @@ ufs_lookup(v) if ((nameiop == CREATE || nameiop == RENAME) && (flags & ISLASTCN)) { slotstatus = NONE; - slotneeded = (sizeof(struct direct) - MAXNAMLEN + - cnp->cn_namelen + 3) &~ 3; + slotneeded = DIRECTSIZ(cnp->cn_namelen); } /* @@ -223,44 +200,87 @@ ufs_lookup(v) * profiling time and hence has been removed in the interest * of simplicity. */ - bmask = VFSTOUFS(vdp->v_mount)->um_mountp->mnt_stat.f_iosize - 1; + bmask = vdp->v_mount->mnt_stat.f_iosize - 1; + +#ifdef UFS_DIRHASH + /* + * Use dirhash for fast operations on large directories. The logic + * to determine whether to hash the directory is contained within + * ufsdirhash_build(); a zero return means that it decided to hash + * this directory and it successfully built up the hash table. + */ + if (ufsdirhash_build(dp) == 0) { + /* Look for a free slot if needed. */ + enduseful = dp->i_size; + if (slotstatus != FOUND) { + slotoffset = ufsdirhash_findfree(dp, slotneeded, + &slotsize); + if (slotoffset >= 0) { + slotstatus = COMPACT; + enduseful = ufsdirhash_enduseful(dp); + if (enduseful < 0) + enduseful = dp->i_size; + } + } + /* Look up the component. */ + numdirpasses = 1; + entryoffsetinblock = 0; /* silence compiler warning */ + switch (ufsdirhash_lookup(dp, cnp->cn_nameptr, cnp->cn_namelen, + &dp->i_offset, &bp, nameiop == DELETE ? &prevoff : NULL)) { + case 0: + ep = (struct direct *)((char *)bp->b_data + + (dp->i_offset & bmask)); + goto foundentry; + case ENOENT: + dp->i_offset = roundup(dp->i_size, dirblksiz); + goto notfound; + default: + /* Something failed; just do a linear search. */ + break; + } + } +#endif /* UFS_DIRHASH */ + if (nameiop != LOOKUP || dp->i_diroff == 0 || - dp->i_diroff > dp->i_size) { + dp->i_diroff >= dp->i_size) { entryoffsetinblock = 0; dp->i_offset = 0; numdirpasses = 1; } else { dp->i_offset = dp->i_diroff; if ((entryoffsetinblock = dp->i_offset & bmask) && - (error = VOP_BLKATOFF(vdp, (off_t)dp->i_offset, NULL, &bp))) - return (error); + (error = ufs_blkatoff(vdp, (off_t)dp->i_offset, + NULL, &bp, false))) + goto out; numdirpasses = 2; nchstats.ncs_2passes++; } prevoff = dp->i_offset; - endsearch = roundup(dp->i_size, DIRBLKSIZ); + endsearch = roundup(dp->i_size, dirblksiz); enduseful = 0; searchloop: while (dp->i_offset < endsearch) { + if (curcpu()->ci_schedstate.spc_flags & SPCF_SHOULDYIELD) + preempt(); /* * If necessary, get the next directory block. */ if ((dp->i_offset & bmask) == 0) { if (bp != NULL) - brelse(bp); - error = VOP_BLKATOFF(vdp, (off_t)dp->i_offset, NULL, - &bp); + brelse(bp, 0); + error = ufs_blkatoff(vdp, (off_t)dp->i_offset, NULL, + &bp, false); if (error) - return (error); + goto out; entryoffsetinblock = 0; } /* - * If still looking for a slot, and at a DIRBLKSIZE + * If still looking for a slot, and at a DIRBLKSIZ * boundary, have to start looking for free space again. */ if (slotstatus == NONE && - (entryoffsetinblock & (DIRBLKSIZ - 1)) == 0) { + (entryoffsetinblock & (dirblksiz - 1)) == 0) { slotoffset = -1; slotfreespace = 0; } @@ -271,13 +291,14 @@ searchloop: * directory. Complete checks can be run by patching * "dirchk" to be true. */ + KASSERT(bp != NULL); ep = (struct direct *)((char *)bp->b_data + entryoffsetinblock); if (ep->d_reclen == 0 || (dirchk && ufs_dirbadentry(vdp, ep, entryoffsetinblock))) { int i; ufs_dirbad(dp, dp->i_offset, "mangled entry"); - i = DIRBLKSIZ - (entryoffsetinblock & (DIRBLKSIZ - 1)); + i = dirblksiz - (entryoffsetinblock & (dirblksiz - 1)); dp->i_offset += i; entryoffsetinblock += i; continue; @@ -290,15 +311,16 @@ searchloop: * compaction is viable. */ if (slotstatus != FOUND) { - int size = ep->d_reclen; + int size = ufs_rw16(ep->d_reclen, needswap); if (ep->d_ino != 0) - size -= DIRSIZ(FSFMT(vdp), ep); + size -= DIRSIZ(FSFMT(vdp), ep, needswap); if (size > 0) { if (size >= slotneeded) { slotstatus = FOUND; slotoffset = dp->i_offset; - slotsize = ep->d_reclen; + slotsize = ufs_rw16(ep->d_reclen, + needswap); } else if (slotstatus == NONE) { slotfreespace += size; if (slotoffset == -1) @@ -306,7 +328,9 @@ searchloop: if (slotfreespace >= slotneeded) { slotstatus = COMPACT; slotsize = dp->i_offset + - ep->d_reclen - slotoffset; + ufs_rw16(ep->d_reclen, + needswap) - + slotoffset; } } } @@ -316,27 +340,33 @@ searchloop: * Check for a name match. */ if (ep->d_ino) { -# if (BYTE_ORDER == LITTLE_ENDIAN) - if (vdp->v_mount->mnt_maxsymlinklen > 0) - namlen = ep->d_namlen; - else - namlen = ep->d_type; -# else +#if (BYTE_ORDER == LITTLE_ENDIAN) + if (FSFMT(vdp) && needswap == 0) + namlen = ep->d_type; + else + namlen = ep->d_namlen; +#else + if (FSFMT(vdp) && needswap != 0) + namlen = ep->d_type; + else namlen = ep->d_namlen; -# endif +#endif if (namlen == cnp->cn_namelen && - !bcmp(cnp->cn_nameptr, ep->d_name, - (unsigned)namlen)) { + !memcmp(cnp->cn_nameptr, ep->d_name, + (unsigned)namlen)) { +#ifdef UFS_DIRHASH +foundentry: +#endif /* * Save directory entry's inode number and * reclen in ndp->ni_ufs area, and release * directory buffer. */ - if (vdp->v_mount->mnt_maxsymlinklen > 0 && - ep->d_type == DT_WHT) { + if (!FSFMT(vdp) && ep->d_type == DT_WHT) { slotstatus = FOUND; slotoffset = dp->i_offset; - slotsize = ep->d_reclen; + slotsize = ufs_rw16(ep->d_reclen, + needswap); dp->i_reclen = slotsize; /* * This is used to set dp->i_endoff, @@ -359,15 +389,14 @@ searchloop: numdirpasses--; goto notfound; } - dp->i_ino = ep->d_ino; - dp->i_reclen = ep->d_reclen; - brelse(bp); + foundino = ufs_rw32(ep->d_ino, needswap); + dp->i_reclen = ufs_rw16(ep->d_reclen, needswap); goto found; } } prevoff = dp->i_offset; - dp->i_offset += ep->d_reclen; - entryoffsetinblock += ep->d_reclen; + dp->i_offset += ufs_rw16(ep->d_reclen, needswap); + entryoffsetinblock += ufs_rw16(ep->d_reclen, needswap); if (ep->d_ino) enduseful = dp->i_offset; } @@ -383,7 +412,7 @@ notfound: goto searchloop; } if (bp != NULL) - brelse(bp); + brelse(bp, 0); /* * If creating, and at end of pathname and current * directory has not been removed, then can consider @@ -393,20 +422,14 @@ notfound: (nameiop == DELETE && (ap->a_cnp->cn_flags & DOWHITEOUT) && (ap->a_cnp->cn_flags & ISWHITEOUT))) && - (flags & ISLASTCN) && dp->i_nlink != 0) { - /* - * Creation of files on a read-only mounted file system - * is pointless, so don't proceed any further. - */ - if (vdp->v_mount->mnt_flag & MNT_RDONLY) - return (EROFS); + (flags & ISLASTCN) && dp->i_ffs_effnlink != 0) { /* * Access for write is interpreted as allowing * creation of files in the directory. */ - error = VOP_ACCESS(vdp, VWRITE, cred, cnp->cn_proc); + error = VOP_ACCESS(vdp, VWRITE, cred); if (error) - return (error); + goto out; /* * Return an indication of where the new directory * entry should be put. If we didn't find a slot, @@ -417,12 +440,12 @@ notfound: * dp->i_offset + dp->i_count. */ if (slotstatus == NONE) { - dp->i_offset = roundup(dp->i_size, DIRBLKSIZ); + dp->i_offset = roundup(dp->i_size, dirblksiz); dp->i_count = 0; enduseful = dp->i_offset; } else if (nameiop == DELETE) { dp->i_offset = slotoffset; - if ((dp->i_offset & (DIRBLKSIZ - 1)) == 0) + if ((dp->i_offset & (dirblksiz - 1)) == 0) dp->i_count = 0; else dp->i_count = dp->i_offset - prevoff; @@ -432,8 +455,10 @@ notfound: if (enduseful < slotoffset + slotsize) enduseful = slotoffset + slotsize; } - dp->i_endoff = roundup(enduseful, DIRBLKSIZ); + dp->i_endoff = roundup(enduseful, dirblksiz); +#if 0 /* commented out by dbj. none of the on disk fields changed */ dp->i_flag |= IN_CHANGE | IN_UPDATE; +#endif /* * We return with the directory locked, so that * the parameters we set up above will still be @@ -448,16 +473,16 @@ notfound: * information cannot be used. */ cnp->cn_flags |= SAVENAME; - if (!lockparent) - VOP_UNLOCK(vdp); - return (EJUSTRETURN); + error = EJUSTRETURN; + goto out; } /* * Insert name into cache (as non-existent) if appropriate. */ if ((cnp->cn_flags & MAKEENTRY) && nameiop != CREATE) cache_enter(vdp, *vpp, cnp); - return (ENOENT); + error = ENOENT; + goto out; found: if (numdirpasses == 2) @@ -466,11 +491,14 @@ found: * Check that directory length properly reflects presence * of this entry. */ - if (entryoffsetinblock + DIRSIZ(FSFMT(vdp), ep) > dp->i_size) { + if (dp->i_offset + DIRSIZ(FSFMT(vdp), ep, needswap) > dp->i_size) { ufs_dirbad(dp, dp->i_offset, "i_size too small"); - dp->i_size = entryoffsetinblock + DIRSIZ(FSFMT(vdp), ep); + dp->i_size = dp->i_offset + DIRSIZ(FSFMT(vdp), ep, needswap); + DIP_ASSIGN(dp, size, dp->i_size); dp->i_flag |= IN_CHANGE | IN_UPDATE; + UFS_WAPBL_UPDATE(vdp, NULL, NULL, UPDATE_DIROP); } + brelse(bp, 0); /* * Found component in pathname. @@ -478,40 +506,43 @@ found: * in the cache as to where the entry was found. */ if ((flags & ISLASTCN) && nameiop == LOOKUP) - dp->i_diroff = dp->i_offset &~ (DIRBLKSIZ - 1); + dp->i_diroff = dp->i_offset &~ (dirblksiz - 1); /* * If deleting, and at end of pathname, return * parameters which can be used to remove file. - * If the wantparent flag isn't set, we return only - * the directory (in ndp->ni_dvp), otherwise we go - * on and lock the inode, being careful with ".". + * Lock the inode, being careful with ".". */ if (nameiop == DELETE && (flags & ISLASTCN)) { /* * Write access to directory required to delete files. */ - error = VOP_ACCESS(vdp, VWRITE, cred, cnp->cn_proc); + error = VOP_ACCESS(vdp, VWRITE, cred); if (error) - return (error); + goto out; /* * Return pointer to current entry in dp->i_offset, * and distance past previous entry (if there * is a previous entry in this block) in dp->i_count. * Save directory inode pointer in ndp->ni_dvp for dirremove(). */ - if ((dp->i_offset & (DIRBLKSIZ - 1)) == 0) + if ((dp->i_offset & (dirblksiz - 1)) == 0) dp->i_count = 0; else dp->i_count = dp->i_offset - prevoff; - if (dp->i_number == dp->i_ino) { + if (dp->i_number == foundino) { VREF(vdp); *vpp = vdp; - return (0); + error = 0; + goto out; } - error = VFS_VGET(vdp->v_mount, dp->i_ino, &tdp); + if (flags & ISDOTDOT) + VOP_UNLOCK(vdp, 0); /* race to get the inode */ + error = VFS_VGET(vdp->v_mount, foundino, &tdp); + if (flags & ISDOTDOT) + vn_lock(vdp, LK_EXCLUSIVE | LK_RETRY); if (error) - return (error); + goto out; /* * If directory is "sticky", then user must own * the directory, or the file in it, else she @@ -519,16 +550,17 @@ found: * implements append-only directories. */ if ((dp->i_mode & ISVTX) && - cred->cr_uid != 0 && - cred->cr_uid != dp->i_uid && - VTOI(tdp)->i_uid != cred->cr_uid) { + kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, + NULL) != 0 && + kauth_cred_geteuid(cred) != dp->i_uid && + VTOI(tdp)->i_uid != kauth_cred_geteuid(cred)) { vput(tdp); - return (EPERM); + error = EPERM; + goto out; } *vpp = tdp; - if (!lockparent) - VOP_UNLOCK(vdp); - return (0); + error = 0; + goto out; } /* @@ -537,25 +569,29 @@ found: * Must get inode of directory entry to verify it's a * regular file, or empty directory. */ - if (nameiop == RENAME && wantparent && - (flags & ISLASTCN)) { - error = VOP_ACCESS(vdp, VWRITE, cred, cnp->cn_proc); + if (nameiop == RENAME && (flags & ISLASTCN)) { + error = VOP_ACCESS(vdp, VWRITE, cred); if (error) - return (error); + goto out; /* * Careful about locking second inode. * This can only occur if the target is ".". */ - if (dp->i_number == dp->i_ino) - return (EISDIR); - error = VFS_VGET(vdp->v_mount, dp->i_ino, &tdp); + if (dp->i_number == foundino) { + error = EISDIR; + goto out; + } + if (flags & ISDOTDOT) + VOP_UNLOCK(vdp, 0); /* race to get the inode */ + error = VFS_VGET(vdp->v_mount, foundino, &tdp); + if (flags & ISDOTDOT) + vn_lock(vdp, LK_EXCLUSIVE | LK_RETRY); if (error) - return (error); + goto out; *vpp = tdp; cnp->cn_flags |= SAVENAME; - if (!lockparent) - VOP_UNLOCK(vdp); - return (0); + error = 0; + goto out; } /* @@ -579,27 +615,20 @@ found: */ pdp = vdp; if (flags & ISDOTDOT) { - VOP_UNLOCK(pdp); /* race to get the inode */ - error = VFS_VGET(vdp->v_mount, dp->i_ino, &tdp); + VOP_UNLOCK(pdp, 0); /* race to get the inode */ + error = VFS_VGET(vdp->v_mount, foundino, &tdp); + vn_lock(pdp, LK_EXCLUSIVE | LK_RETRY); if (error) { - VOP_LOCK(pdp); - return (error); - } - if (lockparent && (flags & ISLASTCN) && - (error = VOP_LOCK(pdp))) { - vput(tdp); - return (error); + goto out; } *vpp = tdp; - } else if (dp->i_number == dp->i_ino) { + } else if (dp->i_number == foundino) { VREF(vdp); /* we want ourself, ie "." */ *vpp = vdp; } else { - error = VFS_VGET(vdp->v_mount, dp->i_ino, &tdp); + error = VFS_VGET(vdp->v_mount, foundino, &tdp); if (error) - return (error); - if (!lockparent || !(flags & ISLASTCN)) - VOP_UNLOCK(pdp); + goto out; *vpp = tdp; } @@ -608,21 +637,23 @@ found: */ if (cnp->cn_flags & MAKEENTRY) cache_enter(vdp, *vpp, cnp); - return (0); + error = 0; + +out: + fstrans_done(vdp->v_mount); + return error; } void -ufs_dirbad(ip, offset, how) - struct inode *ip; - doff_t offset; - char *how; +ufs_dirbad(struct inode *ip, doff_t offset, const char *how) { struct mount *mp; mp = ITOV(ip)->v_mount; - printf("%s: bad dir ino %d at offset %d: %s\n", - mp->mnt_stat.f_mntonname, ip->i_number, offset, how); - if ((mp->mnt_stat.f_flags & MNT_RDONLY) == 0) + printf("%s: bad dir ino %llu at offset %d: %s\n", + mp->mnt_stat.f_mntonname, (unsigned long long)ip->i_number, + offset, how); + if ((mp->mnt_stat.f_flag & MNT_RDONLY) == 0) panic("bad dir"); } @@ -631,31 +662,42 @@ ufs_dirbad(ip, offset, how) * record length must be multiple of 4 * entry must fit in rest of its DIRBLKSIZ block * record must be large enough to contain entry - * name is not longer than MAXNAMLEN + * name is not longer than FFS_MAXNAMLEN * name must be as long as advertised, and null terminated */ int -ufs_dirbadentry(dp, ep, entryoffsetinblock) - struct vnode *dp; - register struct direct *ep; - int entryoffsetinblock; +ufs_dirbadentry(struct vnode *dp, struct direct *ep, int entryoffsetinblock) { - register int i; + int i; int namlen; - -# if (BYTE_ORDER == LITTLE_ENDIAN) - if (dp->v_mount->mnt_maxsymlinklen > 0) - namlen = ep->d_namlen; - else - namlen = ep->d_type; -# else + struct ufsmount *ump = VFSTOUFS(dp->v_mount); + const int needswap = UFS_MPNEEDSWAP(ump); + int dirblksiz = ump->um_dirblksiz; + +#if (BYTE_ORDER == LITTLE_ENDIAN) + if (FSFMT(dp) && needswap == 0) + namlen = ep->d_type; + else + namlen = ep->d_namlen; +#else + if (FSFMT(dp) && needswap != 0) + namlen = ep->d_type; + else namlen = ep->d_namlen; -# endif - if ((ep->d_reclen & 0x3) != 0 || - ep->d_reclen > DIRBLKSIZ - (entryoffsetinblock & (DIRBLKSIZ - 1)) || - ep->d_reclen < DIRSIZ(FSFMT(dp), ep) || namlen > MAXNAMLEN) { +#endif + if ((ufs_rw16(ep->d_reclen, needswap) & 0x3) != 0 || + ufs_rw16(ep->d_reclen, needswap) > + dirblksiz - (entryoffsetinblock & (dirblksiz - 1)) || + ufs_rw16(ep->d_reclen, needswap) < + DIRSIZ(FSFMT(dp), ep, needswap) || + namlen > FFS_MAXNAMLEN) { /*return (1); */ - printf("First bad\n"); + printf("First bad, reclen=%#x, DIRSIZ=%lu, namlen=%d, " + "flags=%#x, entryoffsetinblock=%d, dirblksiz = %d\n", + ufs_rw16(ep->d_reclen, needswap), + (u_long)DIRSIZ(FSFMT(dp), ep, needswap), + namlen, dp->v_mount->mnt_flag, entryoffsetinblock, + dirblksiz); goto bad; } if (ep->d_ino == 0) @@ -674,66 +716,63 @@ bad: } /* - * Write a directory entry after a call to namei, using the parameters - * that it left in nameidata. The argument ip is the inode which the new - * directory entry will refer to. Dvp is a pointer to the directory to - * be written, which was left locked by namei. Remaining parameters - * (dp->i_offset, dp->i_count) indicate how the space for the new - * entry is to be obtained. + * Construct a new directory entry after a call to namei, using the + * parameters that it left in the componentname argument cnp. The + * argument ip is the inode to which the new directory entry will refer. */ -int -ufs_direnter(ip, dvp, cnp) - struct inode *ip; - struct vnode *dvp; - register struct componentname *cnp; +void +ufs_makedirentry(struct inode *ip, struct componentname *cnp, + struct direct *newdirp) { - register struct inode *dp; - struct direct newdir; - #ifdef DIAGNOSTIC if ((cnp->cn_flags & SAVENAME) == 0) - panic("direnter: missing name"); + panic("makedirentry: missing name"); #endif - dp = VTOI(dvp); - newdir.d_ino = ip->i_number; - newdir.d_namlen = cnp->cn_namelen; - bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1); - if (dvp->v_mount->mnt_maxsymlinklen > 0) - newdir.d_type = IFTODT(ip->i_mode); - else { - newdir.d_type = 0; -# if (BYTE_ORDER == LITTLE_ENDIAN) - { u_char tmp = newdir.d_namlen; - newdir.d_namlen = newdir.d_type; - newdir.d_type = tmp; } -# endif - } - return (ufs_direnter2(dvp, &newdir, cnp->cn_cred, cnp->cn_proc)); + newdirp->d_ino = ip->i_number; + newdirp->d_namlen = cnp->cn_namelen; + memcpy(newdirp->d_name, cnp->cn_nameptr, (size_t)cnp->cn_namelen); + newdirp->d_name[cnp->cn_namelen] = '\0'; + if (FSFMT(ITOV(ip))) + newdirp->d_type = 0; + else + newdirp->d_type = IFTODT(ip->i_mode); } /* - * Common entry point for directory entry removal used by ufs_direnter - * and ufs_whiteout + * Write a directory entry after a call to namei, using the parameters + * that it left in nameidata. The argument dirp is the new directory + * entry contents. Dvp is a pointer to the directory to be written, + * which was left locked by namei. Remaining parameters (dp->i_offset, + * dp->i_count) indicate how the space for the new entry is to be obtained. + * Non-null bp indicates that a directory is being created (for the + * soft dependency code). */ int -ufs_direnter2(dvp, dirp, cr, p) - struct vnode *dvp; - struct direct *dirp; - struct ucred *cr; - struct proc *p; +ufs_direnter(struct vnode *dvp, struct vnode *tvp, struct direct *dirp, + struct componentname *cnp, struct buf *newdirbp) { + kauth_cred_t cr; + struct lwp *l; int newentrysize; struct inode *dp; struct buf *bp; - struct iovec aiov; - struct uio auio; u_int dsize; struct direct *ep, *nep; - int error, loc, spacefree; + int error, ret, blkoff, loc, spacefree, flags; char *dirbuf; + struct timespec ts; + struct ufsmount *ump = VFSTOUFS(dvp->v_mount); + const int needswap = UFS_MPNEEDSWAP(ump); + int dirblksiz = ump->um_dirblksiz; + + UFS_WAPBL_JLOCK_ASSERT(dvp->v_mount); + + error = 0; + cr = cnp->cn_cred; + l = curlwp; dp = VTOI(dvp); - newentrysize = DIRSIZ(FSFMT(dvp), dirp); + newentrysize = DIRSIZ(0, dirp, 0); if (dp->i_count == 0) { /* @@ -742,101 +781,246 @@ ufs_direnter2(dvp, dirp, cr, p) * be on a directory block boundary and we will write the * new entry into a fresh block. */ - if (dp->i_offset & (DIRBLKSIZ - 1)) - panic("ufs_direnter2: newblk"); - auio.uio_offset = dp->i_offset; - dirp->d_reclen = DIRBLKSIZ; - auio.uio_resid = newentrysize; - aiov.iov_len = newentrysize; - aiov.iov_base = (caddr_t)dirp; - auio.uio_iov = &aiov; - auio.uio_iovcnt = 1; - auio.uio_rw = UIO_WRITE; - auio.uio_segflg = UIO_SYSSPACE; - auio.uio_procp = (struct proc *)0; - error = VOP_WRITE(dvp, &auio, IO_SYNC, cr); - if (DIRBLKSIZ > - VFSTOUFS(dvp->v_mount)->um_mountp->mnt_stat.f_bsize) - /* XXX should grow with balloc() */ - panic("ufs_direnter2: frag size"); - else if (!error) { - dp->i_size = roundup(dp->i_size, DIRBLKSIZ); - dp->i_flag |= IN_CHANGE; + if (dp->i_offset & (dirblksiz - 1)) + panic("ufs_direnter: newblk"); + flags = B_CLRBUF; + if (!DOINGSOFTDEP(dvp)) + flags |= B_SYNC; + if ((error = UFS_BALLOC(dvp, (off_t)dp->i_offset, dirblksiz, + cr, flags, &bp)) != 0) { + if (DOINGSOFTDEP(dvp) && newdirbp != NULL) + bdwrite(newdirbp); + return (error); + } + dp->i_size = dp->i_offset + dirblksiz; + DIP_ASSIGN(dp, size, dp->i_size); + dp->i_flag |= IN_CHANGE | IN_UPDATE; + uvm_vnp_setsize(dvp, dp->i_size); + dirp->d_reclen = ufs_rw16(dirblksiz, needswap); + dirp->d_ino = ufs_rw32(dirp->d_ino, needswap); + if (FSFMT(dvp)) { +#if (BYTE_ORDER == LITTLE_ENDIAN) + if (needswap == 0) { +#else + if (needswap != 0) { +#endif + u_char tmp = dirp->d_namlen; + dirp->d_namlen = dirp->d_type; + dirp->d_type = tmp; + } + } + blkoff = dp->i_offset & (ump->um_mountp->mnt_stat.f_iosize - 1); + memcpy((char *)bp->b_data + blkoff, dirp, newentrysize); +#ifdef UFS_DIRHASH + if (dp->i_dirhash != NULL) { + ufsdirhash_newblk(dp, dp->i_offset); + ufsdirhash_add(dp, dirp, dp->i_offset); + ufsdirhash_checkblock(dp, (char *)bp->b_data + blkoff, + dp->i_offset); + } +#endif + if (DOINGSOFTDEP(dvp)) { + /* + * Ensure that the entire newly allocated block is a + * valid directory so that future growth within the + * block does not have to ensure that the block is + * written before the inode. + */ + blkoff += dirblksiz; + while (blkoff < bp->b_bcount) { + ((struct direct *) + ((char *)bp->b_data + blkoff))->d_reclen = dirblksiz; + blkoff += dirblksiz; + } + if (softdep_setup_directory_add(bp, dp, dp->i_offset, + ufs_rw32(dirp->d_ino, needswap), newdirbp, 1) == 0) { + bdwrite(bp); + vfs_timestamp(&ts); + return UFS_UPDATE(dvp, &ts, &ts, UPDATE_DIROP); + } + /* We have just allocated a directory block in an + * indirect block. Rather than tracking when it gets + * claimed by the inode, we simply do a VOP_FSYNC + * now to ensure that it is there (in case the user + * does a future fsync). Note that we have to unlock + * the inode for the entry that we just entered, as + * the VOP_FSYNC may need to lock other inodes which + * can lead to deadlock if we also hold a lock on + * the newly entered node. + */ + error = VOP_BWRITE(bp); + if (error != 0) + return (error); + if (tvp != NULL) + VOP_UNLOCK(tvp, 0); + error = VOP_FSYNC(dvp, l->l_cred, FSYNC_WAIT, 0, 0); + if (tvp != 0) + vn_lock(tvp, LK_EXCLUSIVE | LK_RETRY); + return (error); + } else { + error = VOP_BWRITE(bp); } + vfs_timestamp(&ts); + ret = UFS_UPDATE(dvp, &ts, &ts, UPDATE_DIROP); + if (error == 0) + return (ret); return (error); } /* - * If dp->i_count is non-zero, then namei found space - * for the new entry in the range dp->i_offset to - * dp->i_offset + dp->i_count in the directory. - * To use this space, we may have to compact the entries located - * there, by copying them together towards the beginning of the - * block, leaving the free space in one usable chunk at the end. + * If dp->i_count is non-zero, then namei found space for the new + * entry in the range dp->i_offset to dp->i_offset + dp->i_count + * in the directory. To use this space, we may have to compact + * the entries located there, by copying them together towards the + * beginning of the block, leaving the free space in one usable + * chunk at the end. */ /* * Increase size of directory if entry eats into new space. * This should never push the size past a new multiple of - * DIRBLKSIZE. + * DIRBLKSIZ. * * N.B. - THIS IS AN ARTIFACT OF 4.2 AND SHOULD NEVER HAPPEN. */ - if (dp->i_offset + dp->i_count > dp->i_size) + if (dp->i_offset + dp->i_count > dp->i_size) { dp->i_size = dp->i_offset + dp->i_count; + DIP_ASSIGN(dp, size, dp->i_size); + dp->i_flag |= IN_CHANGE | IN_UPDATE; + UFS_WAPBL_UPDATE(dvp, NULL, NULL, UPDATE_DIROP); + } /* * Get the block containing the space for the new directory entry. */ - error = VOP_BLKATOFF(dvp, (off_t)dp->i_offset, &dirbuf, &bp); - if (error) + error = ufs_blkatoff(dvp, (off_t)dp->i_offset, &dirbuf, &bp, true); + if (error) { + if (DOINGSOFTDEP(dvp) && newdirbp != NULL) + bdwrite(newdirbp); return (error); + } /* * Find space for the new entry. In the simple case, the entry at * offset base will have the space. If it does not, then namei * arranged that compacting the region dp->i_offset to - * dp->i_offset + dp->i_count would yield the - * space. + * dp->i_offset + dp->i_count would yield the space. */ ep = (struct direct *)dirbuf; - dsize = DIRSIZ(FSFMT(dvp), ep); - spacefree = ep->d_reclen - dsize; - for (loc = ep->d_reclen; loc < dp->i_count; ) { + dsize = (ep->d_ino != 0) ? DIRSIZ(FSFMT(dvp), ep, needswap) : 0; + spacefree = ufs_rw16(ep->d_reclen, needswap) - dsize; + for (loc = ufs_rw16(ep->d_reclen, needswap); loc < dp->i_count; ) { + uint16_t reclen; + nep = (struct direct *)(dirbuf + loc); - if (ep->d_ino) { - /* trim the existing slot */ - ep->d_reclen = dsize; - ep = (struct direct *)((char *)ep + dsize); - } else { - /* overwrite; nothing there; header is ours */ - spacefree += dsize; + + /* Trim the existing slot (NB: dsize may be zero). */ + ep->d_reclen = ufs_rw16(dsize, needswap); + ep = (struct direct *)((char *)ep + dsize); + + reclen = ufs_rw16(nep->d_reclen, needswap); + loc += reclen; + if (nep->d_ino == 0) { + /* + * A mid-block unused entry. Such entries are + * never created by the kernel, but fsck_ffs + * can create them (and it doesn't fix them). + * + * Add up the free space, and initialise the + * relocated entry since we don't memcpy it. + */ + spacefree += reclen; + ep->d_ino = 0; + dsize = 0; + continue; } - dsize = DIRSIZ(FSFMT(dvp), nep); - spacefree += nep->d_reclen - dsize; - loc += nep->d_reclen; - bcopy((caddr_t)nep, (caddr_t)ep, dsize); + dsize = DIRSIZ(FSFMT(dvp), nep, needswap); + spacefree += reclen - dsize; +#ifdef UFS_DIRHASH + if (dp->i_dirhash != NULL) + ufsdirhash_move(dp, nep, + dp->i_offset + ((char *)nep - dirbuf), + dp->i_offset + ((char *)ep - dirbuf)); +#endif + if (DOINGSOFTDEP(dvp)) + softdep_change_directoryentry_offset(dp, dirbuf, + (void *)nep, (void *)ep, dsize); + else + memcpy((void *)ep, (void *)nep, dsize); } /* + * Here, `ep' points to a directory entry containing `dsize' in-use + * bytes followed by `spacefree' unused bytes. If ep->d_ino == 0, + * then the entry is completely unused (dsize == 0). The value + * of ep->d_reclen is always indeterminate. + * * Update the pointer fields in the previous entry (if any), * copy in the new entry, and write out the block. */ if (ep->d_ino == 0 || - (ep->d_ino == WINO && - bcmp(ep->d_name, dirp->d_name, dirp->d_namlen) == 0)) { + (ufs_rw32(ep->d_ino, needswap) == WINO && + memcmp(ep->d_name, dirp->d_name, dirp->d_namlen) == 0)) { if (spacefree + dsize < newentrysize) - panic("ufs_direnter2: compact1"); + panic("ufs_direnter: compact1"); dirp->d_reclen = spacefree + dsize; } else { if (spacefree < newentrysize) - panic("ufs_direnter2: compact2"); + panic("ufs_direnter: compact2"); dirp->d_reclen = spacefree; - ep->d_reclen = dsize; + ep->d_reclen = ufs_rw16(dsize, needswap); ep = (struct direct *)((char *)ep + dsize); } - bcopy((caddr_t)dirp, (caddr_t)ep, (u_int)newentrysize); - error = VOP_BWRITE(bp); + dirp->d_reclen = ufs_rw16(dirp->d_reclen, needswap); + dirp->d_ino = ufs_rw32(dirp->d_ino, needswap); + if (FSFMT(dvp)) { +#if (BYTE_ORDER == LITTLE_ENDIAN) + if (needswap == 0) { +#else + if (needswap != 0) { +#endif + u_char tmp = dirp->d_namlen; + dirp->d_namlen = dirp->d_type; + dirp->d_type = tmp; + } + } +#ifdef UFS_DIRHASH + if (dp->i_dirhash != NULL && (ep->d_ino == 0 || + dirp->d_reclen == spacefree)) + ufsdirhash_add(dp, dirp, dp->i_offset + ((char *)ep - dirbuf)); +#endif + memcpy((void *)ep, (void *)dirp, (u_int)newentrysize); +#ifdef UFS_DIRHASH + if (dp->i_dirhash != NULL) + ufsdirhash_checkblock(dp, dirbuf - + (dp->i_offset & (dirblksiz - 1)), + dp->i_offset & ~(dirblksiz - 1)); +#endif + if (DOINGSOFTDEP(dvp)) { + softdep_setup_directory_add(bp, dp, + dp->i_offset + (char *)ep - dirbuf, + ufs_rw32(dirp->d_ino, needswap), newdirbp, 0); + bdwrite(bp); + } else { + error = VOP_BWRITE(bp); + } dp->i_flag |= IN_CHANGE | IN_UPDATE; - if (!error && dp->i_endoff && dp->i_endoff < dp->i_size) - error = VOP_TRUNCATE(dvp, (off_t)dp->i_endoff, IO_SYNC, cr, p); + /* + * If all went well, and the directory can be shortened, proceed + * with the truncation. Note that we have to unlock the inode for + * the entry that we just entered, as the truncation may need to + * lock other inodes which can lead to deadlock if we also hold a + * lock on the newly entered node. + */ + if (error == 0 && dp->i_endoff && dp->i_endoff < dp->i_size) { + if (DOINGSOFTDEP(dvp) && (tvp != NULL)) + VOP_UNLOCK(tvp, 0); +#ifdef UFS_DIRHASH + if (dp->i_dirhash != NULL) + ufsdirhash_dirtrunc(dp, dp->i_endoff); +#endif + (void) UFS_TRUNCATE(dvp, (off_t)dp->i_endoff, IO_SYNC, cr); + if (DOINGSOFTDEP(dvp) && (tvp != NULL)) + vn_lock(tvp, LK_EXCLUSIVE | LK_RETRY); + } + UFS_WAPBL_UPDATE(dvp, NULL, NULL, UPDATE_DIROP); return (error); } @@ -853,55 +1037,99 @@ ufs_direnter2(dvp, dirp, cr, p) * to the size of the previous entry. */ int -ufs_dirremove(dvp, cnp) - struct vnode *dvp; - struct componentname *cnp; +ufs_dirremove(struct vnode *dvp, struct inode *ip, int flags, int isrmdir) { - register struct inode *dp; + struct inode *dp = VTOI(dvp); struct direct *ep; struct buf *bp; int error; +#ifdef FFS_EI + const int needswap = UFS_MPNEEDSWAP(dp->i_ump); +#endif - dp = VTOI(dvp); + UFS_WAPBL_JLOCK_ASSERT(dvp->v_mount); - if (cnp->cn_flags & DOWHITEOUT) { + if (flags & DOWHITEOUT) { /* * Whiteout entry: set d_ino to WINO. */ - error = VOP_BLKATOFF(dvp, (off_t)dp->i_offset, (char **)&ep, - &bp); + error = ufs_blkatoff(dvp, (off_t)dp->i_offset, (void *)&ep, + &bp, true); if (error) return (error); - ep->d_ino = WINO; + ep->d_ino = ufs_rw32(WINO, needswap); ep->d_type = DT_WHT; - error = VOP_BWRITE(bp); - dp->i_flag |= IN_CHANGE | IN_UPDATE; - return (error); + goto out; } + if ((error = ufs_blkatoff(dvp, + (off_t)(dp->i_offset - dp->i_count), (void *)&ep, &bp, true)) != 0) + return (error); + +#ifdef UFS_DIRHASH + /* + * Remove the dirhash entry. This is complicated by the fact + * that `ep' is the previous entry when dp->i_count != 0. + */ + if (dp->i_dirhash != NULL) + ufsdirhash_remove(dp, (dp->i_count == 0) ? ep : + (struct direct *)((char *)ep + + ufs_rw16(ep->d_reclen, needswap)), dp->i_offset); +#endif + if (dp->i_count == 0) { /* * First entry in block: set d_ino to zero. */ - error = VOP_BLKATOFF(dvp, (off_t)dp->i_offset, (char **)&ep, - &bp); - if (error) - return (error); ep->d_ino = 0; + } else { + /* + * Collapse new free space into previous entry. + */ + ep->d_reclen = + ufs_rw16(ufs_rw16(ep->d_reclen, needswap) + dp->i_reclen, + needswap); + } + +#ifdef UFS_DIRHASH + if (dp->i_dirhash != NULL) { + int dirblksiz = ip->i_ump->um_dirblksiz; + ufsdirhash_checkblock(dp, (char *)ep - + ((dp->i_offset - dp->i_count) & (dirblksiz - 1)), + dp->i_offset & ~(dirblksiz - 1)); + } +#endif + +out: + if (DOINGSOFTDEP(dvp)) { + if (ip) { + ip->i_ffs_effnlink--; + softdep_change_linkcnt(ip); + softdep_setup_remove(bp, dp, ip, isrmdir); + } + bdwrite(bp); + } else { + if (ip) { + ip->i_ffs_effnlink--; + ip->i_nlink--; + DIP_ASSIGN(ip, nlink, ip->i_nlink); + ip->i_flag |= IN_CHANGE; + UFS_WAPBL_UPDATE(ITOV(ip), NULL, NULL, 0); + } error = VOP_BWRITE(bp); - dp->i_flag |= IN_CHANGE | IN_UPDATE; - return (error); } + dp->i_flag |= IN_CHANGE | IN_UPDATE; +#ifdef FFS /* - * Collapse new free space into previous entry. + * If the last named reference to a snapshot goes away, + * drop its snapshot reference so that it will be reclaimed + * when last open reference goes away. */ - error = VOP_BLKATOFF(dvp, (off_t)(dp->i_offset - dp->i_count), - (char **)&ep, &bp); - if (error) - return (error); - ep->d_reclen += dp->i_reclen; - error = VOP_BWRITE(bp); - dp->i_flag |= IN_CHANGE | IN_UPDATE; + if (ip != 0 && (ip->i_flags & SF_SNAPSHOT) != 0 && + ip->i_ffs_effnlink == 0) + ffs_snapgone(ip); + UFS_WAPBL_UPDATE(dvp, NULL, NULL, 0); +#endif return (error); } @@ -911,23 +1139,43 @@ ufs_dirremove(dvp, cnp) * set up by a call to namei. */ int -ufs_dirrewrite(dp, ip, cnp) - struct inode *dp, *ip; - struct componentname *cnp; +ufs_dirrewrite(struct inode *dp, struct inode *oip, ino_t newinum, int newtype, + int isrmdir, int iflags) { struct buf *bp; struct direct *ep; struct vnode *vdp = ITOV(dp); int error; - error = VOP_BLKATOFF(vdp, (off_t)dp->i_offset, (char **)&ep, &bp); + error = ufs_blkatoff(vdp, (off_t)dp->i_offset, (void *)&ep, &bp, true); if (error) return (error); - ep->d_ino = ip->i_number; - if (vdp->v_mount->mnt_maxsymlinklen > 0) - ep->d_type = IFTODT(ip->i_mode); - error = VOP_BWRITE(bp); - dp->i_flag |= IN_CHANGE | IN_UPDATE; + ep->d_ino = ufs_rw32(newinum, UFS_MPNEEDSWAP(dp->i_ump)); + if (!FSFMT(vdp)) + ep->d_type = newtype; + oip->i_ffs_effnlink--; + if (DOINGSOFTDEP(vdp)) { + softdep_change_linkcnt(oip); + softdep_setup_directory_change(bp, dp, oip, newinum, isrmdir); + bdwrite(bp); + } else { + oip->i_nlink--; + DIP_ASSIGN(oip, nlink, oip->i_nlink); + oip->i_flag |= IN_CHANGE; + UFS_WAPBL_UPDATE(ITOV(oip), NULL, NULL, UPDATE_DIROP); + error = VOP_BWRITE(bp); + } + dp->i_flag |= iflags; +#ifdef FFS + /* + * If the last named reference to a snapshot goes away, + * drop its snapshot reference so that it will be reclaimed + * when last open reference goes away. + */ + if ((oip->i_flags & SF_SNAPSHOT) != 0 && oip->i_ffs_effnlink == 0) + ffs_snapgone(oip); + UFS_WAPBL_UPDATE(vdp, NULL, NULL, UPDATE_DIROP); +#endif return (error); } @@ -941,20 +1189,20 @@ ufs_dirrewrite(dp, ip, cnp) * NB: does not handle corrupted directories. */ int -ufs_dirempty(ip, parentino, cred) - register struct inode *ip; - ino_t parentino; - struct ucred *cred; +ufs_dirempty(struct inode *ip, ino_t parentino, kauth_cred_t cred) { - register off_t off; + doff_t off; struct dirtemplate dbuf; - register struct direct *dp = (struct direct *)&dbuf; - int error, count, namlen; + struct direct *dp = (struct direct *)&dbuf; + int error, namlen; + size_t count; + const int needswap = UFS_IPNEEDSWAP(ip); #define MINDIRSIZ (sizeof (struct dirtemplate) / 2) - for (off = 0; off < ip->i_size; off += dp->d_reclen) { - error = vn_rdwr(UIO_READ, ITOV(ip), (caddr_t)dp, MINDIRSIZ, off, - UIO_SYSSPACE, IO_NODELOCKED, cred, &count, (struct proc *)0); + for (off = 0; off < ip->i_size; + off += ufs_rw16(dp->d_reclen, needswap)) { + error = vn_rdwr(UIO_READ, ITOV(ip), (void *)dp, MINDIRSIZ, off, + UIO_SYSSPACE, IO_NODELOCKED, cred, &count, NULL); /* * Since we read MINDIRSIZ, residual must * be 0 unless we're at end of file. @@ -965,17 +1213,20 @@ ufs_dirempty(ip, parentino, cred) if (dp->d_reclen == 0) return (0); /* skip empty entries */ - if (dp->d_ino == 0 || dp->d_ino == WINO) + if (dp->d_ino == 0 || ufs_rw32(dp->d_ino, needswap) == WINO) continue; /* accept only "." and ".." */ -# if (BYTE_ORDER == LITTLE_ENDIAN) - if (ITOV(ip)->v_mount->mnt_maxsymlinklen > 0) - namlen = dp->d_namlen; - else - namlen = dp->d_type; -# else +#if (BYTE_ORDER == LITTLE_ENDIAN) + if (FSFMT(ITOV(ip)) && needswap == 0) + namlen = dp->d_type; + else namlen = dp->d_namlen; -# endif +#else + if (FSFMT(ITOV(ip)) && needswap != 0) + namlen = dp->d_type; + else + namlen = dp->d_namlen; +#endif if (namlen > 2) return (0); if (dp->d_name[0] != '.') @@ -985,9 +1236,11 @@ ufs_dirempty(ip, parentino, cred) * 1 implies ".", 2 implies ".." if second * char is also "." */ - if (namlen == 1) + if (namlen == 1 && + ufs_rw32(dp->d_ino, needswap) == ip->i_number) continue; - if (dp->d_name[1] == '.' && dp->d_ino == parentino) + if (dp->d_name[1] == '.' && + ufs_rw32(dp->d_ino, needswap) == parentino) continue; return (0); } @@ -1000,13 +1253,12 @@ ufs_dirempty(ip, parentino, cred) * The target is always vput before returning. */ int -ufs_checkpath(source, target, cred) - struct inode *source, *target; - struct ucred *cred; +ufs_checkpath(struct inode *source, struct inode *target, kauth_cred_t cred) { - struct vnode *vp; + struct vnode *vp = ITOV(target); int error, rootino, namlen; struct dirtemplate dirbuf; + const int needswap = UFS_MPNEEDSWAP(target->i_ump); vp = ITOV(target); if (target->i_number == source->i_number) { @@ -1023,33 +1275,37 @@ ufs_checkpath(source, target, cred) error = ENOTDIR; break; } - error = vn_rdwr(UIO_READ, vp, (caddr_t)&dirbuf, - sizeof (struct dirtemplate), (off_t)0, UIO_SYSSPACE, - IO_NODELOCKED, cred, (int *)0, (struct proc *)0); + error = vn_rdwr(UIO_READ, vp, (void *)&dirbuf, + sizeof (struct dirtemplate), (off_t)0, UIO_SYSSPACE, + IO_NODELOCKED, cred, NULL, NULL); if (error != 0) break; -# if (BYTE_ORDER == LITTLE_ENDIAN) - if (vp->v_mount->mnt_maxsymlinklen > 0) - namlen = dirbuf.dotdot_namlen; - else - namlen = dirbuf.dotdot_type; -# else +#if (BYTE_ORDER == LITTLE_ENDIAN) + if (FSFMT(vp) && needswap == 0) + namlen = dirbuf.dotdot_type; + else namlen = dirbuf.dotdot_namlen; -# endif +#else + if (FSFMT(vp) && needswap != 0) + namlen = dirbuf.dotdot_type; + else + namlen = dirbuf.dotdot_namlen; +#endif if (namlen != 2 || dirbuf.dotdot_name[0] != '.' || dirbuf.dotdot_name[1] != '.') { error = ENOTDIR; break; } - if (dirbuf.dotdot_ino == source->i_number) { + if (ufs_rw32(dirbuf.dotdot_ino, needswap) == source->i_number) { error = EINVAL; break; } - if (dirbuf.dotdot_ino == rootino) + if (ufs_rw32(dirbuf.dotdot_ino, needswap) == rootino) break; vput(vp); - error = VFS_VGET(vp->v_mount, dirbuf.dotdot_ino, &vp); + error = VFS_VGET(vp->v_mount, + ufs_rw32(dirbuf.dotdot_ino, needswap), &vp); if (error) { vp = NULL; break; @@ -1063,3 +1319,71 @@ out: vput(vp); return (error); } + +#define UFS_DIRRABLKS 0 +int ufs_dirrablks = UFS_DIRRABLKS; + +/* + * ufs_blkatoff: Return buffer with the contents of block "offset" from + * the beginning of directory "vp". If "res" is non-zero, fill it in with + * a pointer to the remaining space in the directory. If the caller intends + * to modify the buffer returned, "modify" must be true. + */ + +int +ufs_blkatoff(struct vnode *vp, off_t offset, char **res, struct buf **bpp, + bool modify) +{ + struct inode *ip; + struct buf *bp; + daddr_t lbn; + const int dirrablks = ufs_dirrablks; + daddr_t *blks; + int *blksizes; + int run, error; + struct mount *mp = vp->v_mount; + const int bshift = mp->mnt_fs_bshift; + const int bsize = 1 << bshift; + off_t eof; + + blks = kmem_alloc((1 + dirrablks) * sizeof(daddr_t), KM_SLEEP); + blksizes = kmem_alloc((1 + dirrablks) * sizeof(int), KM_SLEEP); + ip = VTOI(vp); + KASSERT(vp->v_size == ip->i_size); + GOP_SIZE(vp, vp->v_size, &eof, 0); + lbn = offset >> bshift; + + for (run = 0; run <= dirrablks;) { + const off_t curoff = lbn << bshift; + const int size = MIN(eof - curoff, bsize); + + if (size == 0) { + break; + } + KASSERT(curoff < eof); + blks[run] = lbn; + blksizes[run] = size; + lbn++; + run++; + if (size != bsize) { + break; + } + } + KASSERT(run >= 1); + error = breadn(vp, blks[0], blksizes[0], &blks[1], &blksizes[1], + run - 1, NOCRED, (modify ? B_MODIFY : 0), &bp); + if (error != 0) { + brelse(bp, 0); + *bpp = NULL; + goto out; + } + if (res) { + *res = (char *)bp->b_data + (offset & (bsize - 1)); + } + *bpp = bp; + + out: + kmem_free(blks, (1 + dirrablks) * sizeof(daddr_t)); + kmem_free(blksizes, (1 + dirrablks) * sizeof(int)); + return error; +}