diff --git a/sbin/fsck_ffs/dir.c b/sbin/fsck_ffs/dir.c index cc5305c390a4..e5f0e1e7e7f4 100644 --- a/sbin/fsck_ffs/dir.c +++ b/sbin/fsck_ffs/dir.c @@ -1,1028 +1,1044 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1980, 1986, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 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. 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. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #if 0 #ifndef lint static const char sccsid[] = "@(#)dir.c 8.8 (Berkeley) 4/28/95"; #endif /* not lint */ #endif #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include "fsck.h" static struct dirtemplate emptydir = { 0, DIRBLKSIZ, DT_UNKNOWN, 0, "", 0, 0, DT_UNKNOWN, 0, "" }; static struct dirtemplate dirhead = { 0, 12, DT_DIR, 1, ".", 0, DIRBLKSIZ - 12, DT_DIR, 2, ".." }; static int chgino(struct inodesc *); static int dircheck(struct inodesc *, struct bufarea *, struct direct *); static int expanddir(struct inode *ip, char *name); static struct direct *fsck_readdir(struct inodesc *); static struct bufarea *getdirblk(ufs2_daddr_t blkno, long size); static int lftempname(char *bufp, ino_t ino); static int mkentry(struct inodesc *); /* * Propagate connected state through the tree. */ void propagate(void) { struct inoinfo **inpp, *inp; struct inoinfo **inpend; long change; inpend = &inpsort[inplast]; do { change = 0; for (inpp = inpsort; inpp < inpend; inpp++) { inp = *inpp; if (inp->i_parent == 0) continue; if (inoinfo(inp->i_parent)->ino_state == DFOUND && INO_IS_DUNFOUND(inp->i_number)) { inoinfo(inp->i_number)->ino_state = DFOUND; check_dirdepth(inp); change++; } } } while (change > 0); } /* * Check that the recorded depth of the directory is correct. */ void check_dirdepth(struct inoinfo *inp) { struct inoinfo *parentinp; struct inode ip; union dinode *dp; int saveresolved; + size_t size; static int updateasked, dirdepthupdate; if ((parentinp = getinoinfo(inp->i_parent)) == NULL) { pfatal("check_dirdepth: UNKNOWN PARENT DIR"); return; } /* * If depth is correct, nothing to do. */ if (parentinp->i_depth + 1 == inp->i_depth) return; /* * Only the root inode should have depth of 0, so if any other * directory has a depth of 0 then this is an old filesystem * that has not been tracking directory depth. Ask just once * whether it should start tracking directory depth. */ if (inp->i_depth == 0 && updateasked == 0) { updateasked = 1; if (preen) { pwarn("UPDATING FILESYSTEM TO TRACK DIRECTORY DEPTH"); dirdepthupdate = 1; } else { /* * The file system can be marked clean even if * a directory does not have the right depth. * Hence, resolved should not be cleared when * the filesystem does not update directory depths. */ saveresolved = resolved; dirdepthupdate = reply("UPDATE FILESYSTEM TO TRACK DIRECTORY DEPTH"); resolved = saveresolved; } } /* - * If we are not converting, nothing more to do. + * If we are not converting or we are running in no-write mode + * there is nothing more to do. */ - if (inp->i_depth == 0 && dirdepthupdate == 0) + if ((inp->i_depth == 0 && dirdepthupdate == 0) || + (fswritefd < 0 && bkgrdflag == 0)) return; /* * Individual directory at wrong depth. Report it and correct if * in preen mode or ask if in interactive mode. Note that if a * directory is renamed to a new location that is at a different * level in the tree, its depth will be recalculated, but none of * the directories that it contains will be updated. Thus it is * not unexpected to find directories with incorrect depths. No * operational harm will come from this though new directory * placement in the subtree may not be as optimal until the depths * of the affected directories are corrected. * * To avoid much spurious output on otherwise clean filesystems * we only generate detailed output when the debug flag is given. */ ginode(inp->i_number, &ip); dp = ip.i_dp; if (inp->i_depth != 0 && debug) { pwarn("DIRECTORY"); prtinode(&ip); printf(" DEPTH %d SHOULD BE %d", inp->i_depth, parentinp->i_depth + 1); if (preen == 0 && reply("ADJUST") == 0) { irelse(&ip); return; } if (preen) printf(" (ADJUSTED)\n"); } inp->i_depth = parentinp->i_depth + 1; - DIP_SET(dp, di_dirdepth, inp->i_depth); - inodirty(&ip); + if (bkgrdflag == 0) { + DIP_SET(dp, di_dirdepth, inp->i_depth); + inodirty(&ip); + } else { + cmd.value = inp->i_number; + cmd.size = (int64_t)inp->i_depth - DIP(dp, di_dirdepth); + if (debug) + printf("adjdepth ino %ld amt %jd\n", (long)cmd.value, + (intmax_t)cmd.size); + size = MIBSIZE; + if (sysctlnametomib("vfs.ffs.adjdepth", adjdepth, &size) < 0 || + sysctl(adjdepth, MIBSIZE, 0, 0, &cmd, sizeof cmd) == -1) + rwerror("ADJUST INODE DEPTH", cmd.value); + } irelse(&ip); } /* * Scan each entry in a directory block. */ int dirscan(struct inodesc *idesc) { struct direct *dp; struct bufarea *bp; u_int dsize, n; long blksiz; char dbuf[DIRBLKSIZ]; if (idesc->id_type != DATA) errx(EEXIT, "wrong type to dirscan %d", idesc->id_type); if (idesc->id_entryno == 0 && (idesc->id_filesize & (DIRBLKSIZ - 1)) != 0) idesc->id_filesize = roundup(idesc->id_filesize, DIRBLKSIZ); blksiz = idesc->id_numfrags * sblock.fs_fsize; if (chkrange(idesc->id_blkno, idesc->id_numfrags)) { idesc->id_filesize -= blksiz; return (SKIP); } idesc->id_loc = 0; for (dp = fsck_readdir(idesc); dp != NULL; dp = fsck_readdir(idesc)) { dsize = dp->d_reclen; if (dsize > sizeof(dbuf)) dsize = sizeof(dbuf); memmove(dbuf, dp, (size_t)dsize); idesc->id_dirp = (struct direct *)dbuf; if ((n = (*idesc->id_func)(idesc)) & ALTERED) { bp = getdirblk(idesc->id_blkno, blksiz); if (bp->b_errs != 0) return (STOP); memmove(bp->b_un.b_buf + idesc->id_loc - dsize, dbuf, (size_t)dsize); dirty(bp); sbdirty(); } if (n & STOP) return (n); } return (idesc->id_filesize > 0 ? KEEPON : STOP); } /* * Get and verify the next entry in a directory. * We also verify that if there is another entry in the block that it is * valid, so if it is not valid it can be subsumed into the current entry. */ static struct direct * fsck_readdir(struct inodesc *idesc) { struct direct *dp, *ndp; struct bufarea *bp; long size, blksiz, subsume_ndp; subsume_ndp = 0; blksiz = idesc->id_numfrags * sblock.fs_fsize; if (idesc->id_filesize <= 0 || idesc->id_loc >= blksiz) return (NULL); bp = getdirblk(idesc->id_blkno, blksiz); if (bp->b_errs != 0) return (NULL); dp = (struct direct *)(bp->b_un.b_buf + idesc->id_loc); /* * Only need to check current entry if it is the first in the * the block, as later entries will have been checked in the * previous call to this function. */ if (idesc->id_loc % DIRBLKSIZ != 0 || dircheck(idesc, bp, dp) != 0) { /* * Current entry is good, update to point at next. */ idesc->id_loc += dp->d_reclen; idesc->id_filesize -= dp->d_reclen; /* * If at end of directory block, just return this entry. */ if (idesc->id_filesize <= 0 || idesc->id_loc >= blksiz || idesc->id_loc % DIRBLKSIZ == 0) return (dp); /* * If the next entry good, return this entry. */ ndp = (struct direct *)(bp->b_un.b_buf + idesc->id_loc); if (dircheck(idesc, bp, ndp) != 0) return (dp); /* * The next entry is bad, so subsume it and the remainder * of this directory block into this entry. */ subsume_ndp = 1; } /* * Current or next entry is bad. Zap current entry or * subsume next entry into current entry as appropriate. */ size = DIRBLKSIZ - (idesc->id_loc % DIRBLKSIZ); idesc->id_loc += size; idesc->id_filesize -= size; if (idesc->id_fix == IGNORE) return (NULL); if (subsume_ndp) { memset(ndp, 0, size); dp->d_reclen += size; } else { memset(dp, 0, size); dp->d_reclen = size; } if (dofix(idesc, "DIRECTORY CORRUPTED")) dirty(bp); return (dp); } /* * Verify that a directory entry is valid. * This is a superset of the checks made in the kernel. * Also optionally clears padding and unused directory space. * * Returns 0 if the entry is bad, 1 if the entry is good. */ static int dircheck(struct inodesc *idesc, struct bufarea *bp, struct direct *dp) { size_t size; char *cp; u_int8_t namlen; int spaceleft, modified, unused; spaceleft = DIRBLKSIZ - (idesc->id_loc % DIRBLKSIZ); size = DIRSIZ(0, dp); if (dp->d_reclen == 0 || dp->d_reclen > spaceleft || dp->d_reclen < size || idesc->id_filesize < size || (dp->d_reclen & (DIR_ROUNDUP - 1)) != 0) goto bad; modified = 0; if (dp->d_ino == 0) { if (!zflag || fswritefd < 0) return (1); /* * Special case of an unused directory entry. Normally only * occurs at the beginning of a directory block when the block * contains no entries. Other than the first entry in a * directory block, the kernel coalesces unused space with * the previous entry by extending its d_reclen. However, * when cleaning up a directory, fsck may set d_ino to zero * in the middle of a directory block. If we're clearing out * directory cruft (-z flag), then make sure that all directory * space in entries with d_ino == 0 gets fully cleared. */ if (dp->d_type != 0) { dp->d_type = 0; modified = 1; } if (dp->d_namlen != 0) { dp->d_namlen = 0; modified = 1; } unused = dp->d_reclen - __offsetof(struct direct, d_name); for (cp = dp->d_name; unused > 0; unused--, cp++) { if (*cp != '\0') { *cp = '\0'; modified = 1; } } if (modified) dirty(bp); return (1); } /* * The d_type field should not be tested here. A bad type is an error * in the entry itself but is not a corruption of the directory * structure itself. So blowing away all the remaining entries in the * directory block is inappropriate. Rather the type error should be * checked in pass1 and fixed there. * * The name validation should also be done in pass1 although the * check to see if the name is longer than fits in the space * allocated for it (i.e., the *cp != '\0' fails after exiting the * loop below) then it really is a structural error that requires * the stronger action taken here. */ namlen = dp->d_namlen; if (namlen == 0 || dp->d_type > 15) goto bad; for (cp = dp->d_name, size = 0; size < namlen; size++) { if (*cp == '\0' || *cp++ == '/') goto bad; } if (*cp != '\0') goto bad; if (zflag && fswritefd >= 0) { /* * Clear unused directory entry space, including the d_name * padding. */ /* First figure the number of pad bytes. */ unused = roundup2(namlen + 1, DIR_ROUNDUP) - (namlen + 1); /* Add in the free space to the end of the record. */ unused += dp->d_reclen - DIRSIZ(0, dp); /* * Now clear out the unused space, keeping track if we actually * changed anything. */ for (cp = &dp->d_name[namlen + 1]; unused > 0; unused--, cp++) { if (*cp != '\0') { *cp = '\0'; modified = 1; } } if (modified) dirty(bp); } return (1); bad: if (debug) printf("Bad dir: ino %d reclen %d namlen %d type %d name %s\n", dp->d_ino, dp->d_reclen, dp->d_namlen, dp->d_type, dp->d_name); return (0); } void direrror(ino_t ino, const char *errmesg) { fileerror(ino, ino, errmesg); } void fileerror(ino_t cwd, ino_t ino, const char *errmesg) { struct inode ip; union dinode *dp; char pathbuf[MAXPATHLEN + 1]; pwarn("%s ", errmesg); if (ino < UFS_ROOTINO || ino > maxino) { pfatal("out-of-range inode number %ju", (uintmax_t)ino); return; } ginode(ino, &ip); dp = ip.i_dp; prtinode(&ip); printf("\n"); getpathname(pathbuf, cwd, ino); if (ftypeok(dp)) pfatal("%s=%s\n", (DIP(dp, di_mode) & IFMT) == IFDIR ? "DIR" : "FILE", pathbuf); else pfatal("NAME=%s\n", pathbuf); irelse(&ip); } void adjust(struct inodesc *idesc, int lcnt) { struct inode ip; union dinode *dp; int saveresolved; ginode(idesc->id_number, &ip); dp = ip.i_dp; if (DIP(dp, di_nlink) == lcnt) { /* * If we have not hit any unresolved problems, are running * in preen mode, and are on a file system using soft updates, * then just toss any partially allocated files. */ if (resolved && (preen || bkgrdflag) && usedsoftdep) { clri(idesc, "UNREF", 1); irelse(&ip); return; } else { /* * The file system can be marked clean even if * a file is not linked up, but is cleared. * Hence, resolved should not be cleared when * linkup is answered no, but clri is answered yes. */ saveresolved = resolved; if (linkup(idesc->id_number, (ino_t)0, NULL) == 0) { resolved = saveresolved; clri(idesc, "UNREF", 0); irelse(&ip); return; } /* * Account for the new reference created by linkup(). */ lcnt--; } } if (lcnt != 0) { pwarn("LINK COUNT %s", (lfdir == idesc->id_number) ? lfname : ((DIP(dp, di_mode) & IFMT) == IFDIR ? "DIR" : "FILE")); prtinode(&ip); printf(" COUNT %d SHOULD BE %d", DIP(dp, di_nlink), DIP(dp, di_nlink) - lcnt); if (preen || usedsoftdep) { if (lcnt < 0) { printf("\n"); pfatal("LINK COUNT INCREASING"); } if (preen) printf(" (ADJUSTED)\n"); } if (preen || reply("ADJUST") == 1) { if (bkgrdflag == 0) { DIP_SET(dp, di_nlink, DIP(dp, di_nlink) - lcnt); inodirty(&ip); } else { cmd.value = idesc->id_number; cmd.size = -lcnt; if (debug) printf("adjrefcnt ino %ld amt %lld\n", (long)cmd.value, (long long)cmd.size); if (sysctl(adjrefcnt, MIBSIZE, 0, 0, &cmd, sizeof cmd) == -1) - rwerror("ADJUST INODE", cmd.value); + rwerror("ADJUST INODE LINK COUNT", + cmd.value); } } } irelse(&ip); } static int mkentry(struct inodesc *idesc) { struct direct *dirp = idesc->id_dirp; struct direct newent; int newlen, oldlen; newent.d_namlen = strlen(idesc->id_name); newlen = DIRSIZ(0, &newent); if (dirp->d_ino != 0) oldlen = DIRSIZ(0, dirp); else oldlen = 0; if (dirp->d_reclen - oldlen < newlen) return (KEEPON); newent.d_reclen = dirp->d_reclen - oldlen; dirp->d_reclen = oldlen; dirp = (struct direct *)(((char *)dirp) + oldlen); dirp->d_ino = idesc->id_parent; /* ino to be entered is in id_parent */ dirp->d_reclen = newent.d_reclen; dirp->d_type = inoinfo(idesc->id_parent)->ino_type; dirp->d_namlen = newent.d_namlen; memmove(dirp->d_name, idesc->id_name, (size_t)newent.d_namlen + 1); return (ALTERED|STOP); } static int chgino(struct inodesc *idesc) { struct direct *dirp = idesc->id_dirp; if (memcmp(dirp->d_name, idesc->id_name, (int)dirp->d_namlen + 1)) return (KEEPON); dirp->d_ino = idesc->id_parent; dirp->d_type = inoinfo(idesc->id_parent)->ino_type; return (ALTERED|STOP); } int linkup(ino_t orphan, ino_t parentdir, char *name) { struct inode ip; union dinode *dp; int lostdir, depth; ino_t oldlfdir; struct inoinfo *inp; struct inodesc idesc; char tempname[BUFSIZ]; memset(&idesc, 0, sizeof(struct inodesc)); ginode(orphan, &ip); dp = ip.i_dp; lostdir = (DIP(dp, di_mode) & IFMT) == IFDIR; pwarn("UNREF %s ", lostdir ? "DIR" : "FILE"); prtinode(&ip); printf("\n"); if (preen && DIP(dp, di_size) == 0) { irelse(&ip); return (0); } irelse(&ip); if (cursnapshot != 0) { pfatal("FILE LINKUP IN SNAPSHOT"); return (0); } if (preen) printf(" (RECONNECTED)\n"); else if (reply("RECONNECT") == 0) return (0); if (lfdir == 0) { ginode(UFS_ROOTINO, &ip); idesc.id_name = strdup(lfname); idesc.id_type = DATA; idesc.id_func = findino; idesc.id_number = UFS_ROOTINO; if ((ckinode(ip.i_dp, &idesc) & FOUND) != 0) { lfdir = idesc.id_parent; } else { pwarn("NO lost+found DIRECTORY"); if (preen || reply("CREATE")) { lfdir = allocdir(UFS_ROOTINO, (ino_t)0, lfmode); if (lfdir != 0) { if (makeentry(UFS_ROOTINO, lfdir, lfname) != 0) { numdirs++; if (preen) printf(" (CREATED)\n"); } else { freedirino(lfdir, UFS_ROOTINO); lfdir = 0; if (preen) printf("\n"); } } } } irelse(&ip); free(idesc.id_name); if (lfdir == 0) { pfatal("SORRY. CANNOT CREATE lost+found DIRECTORY"); printf("\n\n"); return (0); } } ginode(lfdir, &ip); dp = ip.i_dp; if ((DIP(dp, di_mode) & IFMT) != IFDIR) { pfatal("lost+found IS NOT A DIRECTORY"); if (reply("REALLOCATE") == 0) { irelse(&ip); return (0); } oldlfdir = lfdir; if ((lfdir = allocdir(UFS_ROOTINO, (ino_t)0, lfmode)) == 0) { pfatal("SORRY. CANNOT CREATE lost+found DIRECTORY\n\n"); irelse(&ip); return (0); } if ((changeino(UFS_ROOTINO, lfname, lfdir, 1) & ALTERED) == 0) { pfatal("SORRY. CANNOT CREATE lost+found DIRECTORY\n\n"); irelse(&ip); return (0); } idesc.id_type = inoinfo(oldlfdir)->ino_idtype; idesc.id_func = freeblock; idesc.id_number = oldlfdir; adjust(&idesc, inoinfo(oldlfdir)->ino_linkcnt + 1); inoinfo(oldlfdir)->ino_linkcnt = 0; inodirty(&ip); irelse(&ip); ginode(lfdir, &ip); dp = ip.i_dp; } if (inoinfo(lfdir)->ino_state != DFOUND) { pfatal("SORRY. NO lost+found DIRECTORY\n\n"); irelse(&ip); return (0); } (void)lftempname(tempname, orphan); if (makeentry(lfdir, orphan, (name ? name : tempname)) == 0) { pfatal("SORRY. NO SPACE IN lost+found DIRECTORY"); printf("\n\n"); irelse(&ip); return (0); } inoinfo(orphan)->ino_linkcnt--; if (lostdir) { depth = DIP(dp, di_dirdepth) + 1; if ((changeino(orphan, "..", lfdir, depth) & ALTERED) == 0 && parentdir != (ino_t)-1) (void)makeentry(orphan, lfdir, ".."); DIP_SET(dp, di_nlink, DIP(dp, di_nlink) + 1); inodirty(&ip); inoinfo(lfdir)->ino_linkcnt++; pwarn("DIR I=%lu CONNECTED. ", (u_long)orphan); inp = getinoinfo(parentdir); if (parentdir != (ino_t)-1 && inp != NULL) { printf("PARENT WAS I=%lu\n", (u_long)parentdir); /* * If the parent directory did not have to * be replaced then because of the ordering * guarantees, has had the link count incremented * for the child, but no entry was made. This * fixes the parent link count so that fsck does * not need to be rerun. */ if ((inp->i_flags & INFO_NEW) != 0) inoinfo(parentdir)->ino_linkcnt++; } if (preen == 0) printf("\n"); } irelse(&ip); return (1); } /* * fix an entry in a directory. */ int changeino(ino_t dir, const char *name, ino_t newnum, int depth) { struct inodesc idesc; struct inode ip; int error; memset(&idesc, 0, sizeof(struct inodesc)); idesc.id_type = DATA; idesc.id_func = chgino; idesc.id_number = dir; idesc.id_fix = DONTKNOW; idesc.id_name = strdup(name); idesc.id_parent = newnum; /* new value for name */ ginode(dir, &ip); if (((error = ckinode(ip.i_dp, &idesc)) & ALTERED) && newnum != 0) { DIP_SET(ip.i_dp, di_dirdepth, depth); getinoinfo(dir)->i_depth = depth; } free(idesc.id_name); irelse(&ip); return (error); } /* * make an entry in a directory */ int makeentry(ino_t parent, ino_t ino, const char *name) { struct inode ip; union dinode *dp; struct inodesc idesc; int retval; char pathbuf[MAXPATHLEN + 1]; if (parent < UFS_ROOTINO || parent >= maxino || ino < UFS_ROOTINO || ino >= maxino) return (0); memset(&idesc, 0, sizeof(struct inodesc)); idesc.id_type = DATA; idesc.id_func = mkentry; idesc.id_number = parent; idesc.id_parent = ino; /* this is the inode to enter */ idesc.id_fix = DONTKNOW; idesc.id_name = strdup(name); ginode(parent, &ip); dp = ip.i_dp; if (DIP(dp, di_size) % DIRBLKSIZ) { DIP_SET(dp, di_size, roundup(DIP(dp, di_size), DIRBLKSIZ)); inodirty(&ip); } if ((ckinode(dp, &idesc) & ALTERED) != 0) { irelse(&ip); free(idesc.id_name); return (1); } getpathname(pathbuf, parent, parent); if (expanddir(&ip, pathbuf) == 0) { irelse(&ip); free(idesc.id_name); return (0); } retval = ckinode(dp, &idesc) & ALTERED; irelse(&ip); free(idesc.id_name); return (retval); } /* * Attempt to expand the size of a directory */ static int expanddir(struct inode *ip, char *name) { ufs2_daddr_t lastlbn, oldblk, newblk, indirblk; size_t filesize, lastlbnsize; struct bufarea *bp, *nbp; struct inodesc idesc; union dinode *dp; long cg, indiralloced; char *cp; nbp = NULL; indiralloced = newblk = indirblk = 0; memset(&idesc, 0, sizeof(struct inodesc)); idesc.id_type = ADDR; pwarn("NO SPACE LEFT IN %s", name); if (!preen && reply("EXPAND") == 0) return (0); cg = ino_to_cg(&sblock, ip->i_number); dp = ip->i_dp; filesize = DIP(dp, di_size); lastlbn = lblkno(&sblock, filesize); /* * We only expand lost+found to a single indirect block. */ if ((DIP(dp, di_mode) & IFMT) != IFDIR || filesize == 0 || lastlbn >= UFS_NDADDR + NINDIR(&sblock)) goto bad; /* * If last block is a fragment, expand it to a full size block. */ lastlbnsize = sblksize(&sblock, filesize, lastlbn); if (lastlbnsize > 0 && lastlbnsize < sblock.fs_bsize) { oldblk = DIP(dp, di_db[lastlbn]); bp = getdirblk(oldblk, lastlbnsize); if (bp->b_errs) goto bad; newblk = allocblk(cg, sblock.fs_frag, std_checkblkavail); if (newblk == 0) goto bad; nbp = getdatablk(newblk, sblock.fs_bsize, BT_DIRDATA); if (nbp->b_errs) goto bad; DIP_SET(dp, di_db[lastlbn], newblk); DIP_SET(dp, di_size, filesize + sblock.fs_bsize - lastlbnsize); DIP_SET(dp, di_blocks, DIP(dp, di_blocks) + btodb(sblock.fs_bsize - lastlbnsize)); inodirty(ip); memmove(nbp->b_un.b_buf, bp->b_un.b_buf, lastlbnsize); memset(&nbp->b_un.b_buf[lastlbnsize], 0, sblock.fs_bsize - lastlbnsize); for (cp = &nbp->b_un.b_buf[lastlbnsize]; cp < &nbp->b_un.b_buf[sblock.fs_bsize]; cp += DIRBLKSIZ) memmove(cp, &emptydir, sizeof emptydir); dirty(nbp); brelse(nbp); binval(bp); idesc.id_blkno = oldblk; idesc.id_numfrags = numfrags(&sblock, lastlbnsize); (void)freeblock(&idesc); if (preen) printf(" (EXPANDED)\n"); return (1); } if ((newblk = allocblk(cg, sblock.fs_frag, std_checkblkavail)) == 0) goto bad; bp = getdirblk(newblk, sblock.fs_bsize); if (bp->b_errs) goto bad; memset(bp->b_un.b_buf, 0, sblock.fs_bsize); for (cp = bp->b_un.b_buf; cp < &bp->b_un.b_buf[sblock.fs_bsize]; cp += DIRBLKSIZ) memmove(cp, &emptydir, sizeof emptydir); dirty(bp); if (lastlbn < UFS_NDADDR) { DIP_SET(dp, di_db[lastlbn], newblk); } else { /* * Allocate indirect block if needed. */ if ((indirblk = DIP(dp, di_ib[0])) == 0) { indirblk = allocblk(cg, sblock.fs_frag, std_checkblkavail); if (indirblk == 0) { binval(bp); goto bad; } indiralloced = 1; } nbp = getdatablk(indirblk, sblock.fs_bsize, BT_LEVEL1); if (nbp->b_errs) goto bad; if (indiralloced) { memset(nbp->b_un.b_buf, 0, sblock.fs_bsize); DIP_SET(dp, di_ib[0], indirblk); DIP_SET(dp, di_blocks, DIP(dp, di_blocks) + btodb(sblock.fs_bsize)); } IBLK_SET(nbp, lastlbn - UFS_NDADDR, newblk); dirty(nbp); brelse(nbp); } DIP_SET(dp, di_size, filesize + sblock.fs_bsize); DIP_SET(dp, di_blocks, DIP(dp, di_blocks) + btodb(sblock.fs_bsize)); inodirty(ip); if (preen) printf(" (EXPANDED)\n"); return (1); bad: pfatal(" (EXPANSION FAILED)\n"); if (nbp != NULL) { binval(bp); brelse(nbp); } if (newblk != 0) { idesc.id_blkno = newblk; idesc.id_numfrags = sblock.fs_frag; (void)freeblock(&idesc); } if (indiralloced) { idesc.id_blkno = indirblk; idesc.id_numfrags = sblock.fs_frag; (void)freeblock(&idesc); } return (0); } /* * allocate a new directory */ ino_t allocdir(ino_t parent, ino_t request, int mode) { ino_t ino; char *cp; struct inode ip; union dinode *dp; struct bufarea *bp; struct dirtemplate *dirp; struct inoinfo *inp, *parentinp; ino = allocino(request, IFDIR|mode); if (ino == 0) return (0); dirp = &dirhead; dirp->dot_ino = ino; dirp->dotdot_ino = parent; ginode(ino, &ip); dp = ip.i_dp; bp = getdirblk(DIP(dp, di_db[0]), sblock.fs_fsize); if (bp->b_errs) { freeino(ino); irelse(&ip); return (0); } memmove(bp->b_un.b_buf, dirp, sizeof(struct dirtemplate)); for (cp = &bp->b_un.b_buf[DIRBLKSIZ]; cp < &bp->b_un.b_buf[sblock.fs_fsize]; cp += DIRBLKSIZ) memmove(cp, &emptydir, sizeof emptydir); dirty(bp); DIP_SET(dp, di_nlink, 2); inodirty(&ip); if (ino == UFS_ROOTINO) { inp = cacheino(dp, ino); inp->i_parent = parent; inp->i_dotdot = parent; inp->i_flags |= INFO_NEW; inoinfo(ino)->ino_type = DT_DIR; inoinfo(ino)->ino_linkcnt = DIP(dp, di_nlink); irelse(&ip); return(ino); } if (!INO_IS_DVALID(parent)) { freeino(ino); irelse(&ip); return (0); } inp = cacheino(dp, ino); inp->i_parent = parent; inp->i_dotdot = parent; inp->i_flags |= INFO_NEW; if ((parentinp = getinoinfo(inp->i_parent)) == NULL) { pfatal("allocdir: UNKNOWN PARENT DIR"); } else { inp->i_depth = parentinp->i_depth + 1; DIP_SET(dp, di_dirdepth, inp->i_depth); } inoinfo(ino)->ino_type = DT_DIR; inoinfo(ino)->ino_state = inoinfo(parent)->ino_state; if (inoinfo(ino)->ino_state == DSTATE) { inoinfo(ino)->ino_linkcnt = DIP(dp, di_nlink); inoinfo(parent)->ino_linkcnt++; } irelse(&ip); ginode(parent, &ip); dp = ip.i_dp; DIP_SET(dp, di_nlink, DIP(dp, di_nlink) + 1); inodirty(&ip); irelse(&ip); return (ino); } /* * free a directory inode */ void freedirino(ino_t ino, ino_t parent) { struct inode ip; union dinode *dp; if (ino != parent) { ginode(parent, &ip); dp = ip.i_dp; DIP_SET(dp, di_nlink, DIP(dp, di_nlink) - 1); inodirty(&ip); irelse(&ip); } removecachedino(ino); freeino(ino); } /* * generate a temporary name for the lost+found directory. */ static int lftempname(char *bufp, ino_t ino) { ino_t in; char *cp; int namlen; cp = bufp + 2; for (in = maxino; in > 0; in /= 10) cp++; *--cp = 0; namlen = cp - bufp; in = ino; while (cp > bufp) { *--cp = (in % 10) + '0'; in /= 10; } *cp = '#'; return (namlen); } /* * Get a directory block. * Insure that it is held until another is requested. */ static struct bufarea * getdirblk(ufs2_daddr_t blkno, long size) { if (pdirbp != NULL && pdirbp->b_errs == 0) brelse(pdirbp); pdirbp = getdatablk(blkno, size, BT_DIRDATA); return (pdirbp); } diff --git a/sbin/fsck_ffs/fsck.h b/sbin/fsck_ffs/fsck.h index 092d9575dc2a..ad82c5f80da1 100644 --- a/sbin/fsck_ffs/fsck.h +++ b/sbin/fsck_ffs/fsck.h @@ -1,542 +1,543 @@ /*- * SPDX-License-Identifier: BSD-3-Clause and BSD-2-Clause * * Copyright (c) 2002 Networks Associates Technology, Inc. * All rights reserved. * * This software was developed for the FreeBSD Project by Marshall * Kirk McKusick and Network Associates Laboratories, the Security * Research Division of Network Associates, Inc. under DARPA/SPAWAR * contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS * research program. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 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. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * Copyright (c) 1980, 1986, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 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. 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. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)fsck.h 8.4 (Berkeley) 5/9/95 * $FreeBSD$ */ #ifndef _FSCK_H_ #define _FSCK_H_ #include #include #include #include #define MAXDUP 10 /* limit on dup blks (per inode) */ #define MAXBAD 10 /* limit on bad blks (per inode) */ #define MINBUFS 100 /* minimum number of buffers required */ #define INOBUFSIZE 64*1024 /* size of buffer to read inodes in pass1 */ #define ZEROBUFSIZE (dev_bsize * 128) /* size of zero buffer used by -Z */ union dinode { struct ufs1_dinode dp1; struct ufs2_dinode dp2; }; #define DIP(dp, field) \ ((sblock.fs_magic == FS_UFS1_MAGIC) ? \ (dp)->dp1.field : (dp)->dp2.field) #define DIP_SET(dp, field, val) do { \ if (sblock.fs_magic == FS_UFS1_MAGIC) \ (dp)->dp1.field = (val); \ else \ (dp)->dp2.field = (val); \ } while (0) /* * Each inode on the file system is described by the following structure. * The linkcnt is initially set to the value in the inode. Each time it * is found during the descent in passes 2, 3, and 4 the count is * decremented. Any inodes whose count is non-zero after pass 4 needs to * have its link count adjusted by the value remaining in ino_linkcnt. */ struct inostat { u_char ino_state; /* state of inode, see below */ u_char ino_type:4; /* type of inode */ u_char ino_idtype:4; /* idesc id_type, SNAP or ADDR */ u_short ino_linkcnt; /* number of links not found */ }; /* * Inode states. */ #define USTATE 0x1 /* inode not allocated */ #define FSTATE 0x2 /* inode is file */ #define FZLINK 0x3 /* inode is file with a link count of zero */ #define DSTATE 0x4 /* inode is directory */ #define DZLINK 0x5 /* inode is directory with a zero link count */ #define DFOUND 0x6 /* directory found during descent */ /* 0x7 UNUSED - see S_IS_DVALID() definition */ #define DCLEAR 0x8 /* directory is to be cleared */ #define FCLEAR 0x9 /* file is to be cleared */ /* DUNFOUND === (state == DSTATE || state == DZLINK) */ #define S_IS_DUNFOUND(state) (((state) & ~0x1) == DSTATE) /* DVALID === (state == DSTATE || state == DZLINK || state == DFOUND) */ #define S_IS_DVALID(state) (((state) & ~0x3) == DSTATE) #define INO_IS_DUNFOUND(ino) S_IS_DUNFOUND(inoinfo(ino)->ino_state) #define INO_IS_DVALID(ino) S_IS_DVALID(inoinfo(ino)->ino_state) /* * Inode state information is contained on per cylinder group lists * which are described by the following structure. */ extern struct inostatlist { long il_numalloced; /* number of inodes allocated in this cg */ struct inostat *il_stat;/* inostat info for this cylinder group */ } *inostathead; /* * Structure to reference a dinode. */ struct inode { struct bufarea *i_bp; /* buffer containing the dinode */ union dinode *i_dp; /* pointer to dinode in buffer */ ino_t i_number; /* inode number */ }; /* * Size of hash tables */ #define HASHSIZE 2048 #define HASH(x) ((x * 2654435761) & (HASHSIZE - 1)) /* * buffer cache structure. */ struct bufarea { TAILQ_ENTRY(bufarea) b_list; /* LRU buffer queue */ LIST_ENTRY(bufarea) b_hash; /* hash list */ ufs2_daddr_t b_bno; /* disk block number */ int b_size; /* size of I/O */ int b_errs; /* I/O error */ int b_flags; /* B_ flags below */ int b_type; /* BT_ type below */ int b_refcnt; /* ref count of users */ int b_index; /* for BT_LEVEL, ptr index */ /* for BT_INODES, first inum */ union { char *b_buf; /* buffer space */ ufs1_daddr_t *b_indir1; /* UFS1 indirect block */ ufs2_daddr_t *b_indir2; /* UFS2 indirect block */ struct fs *b_fs; /* super block */ struct cg *b_cg; /* cylinder group */ struct ufs1_dinode *b_dinode1; /* UFS1 inode block */ struct ufs2_dinode *b_dinode2; /* UFS2 inode block */ } b_un; }; #define IBLK(bp, i) \ ((sblock.fs_magic == FS_UFS1_MAGIC) ? \ (bp)->b_un.b_indir1[i] : (bp)->b_un.b_indir2[i]) #define IBLK_SET(bp, i, val) do { \ if (sblock.fs_magic == FS_UFS1_MAGIC) \ (bp)->b_un.b_indir1[i] = (val); \ else \ (bp)->b_un.b_indir2[i] = (val); \ } while (0) /* * Buffer flags */ #define B_DIRTY 0x00000001 /* Buffer is dirty */ /* * Type of data in buffer */ #define BT_UNKNOWN 0 /* Buffer type is unknown */ #define BT_SUPERBLK 1 /* Buffer holds a superblock */ #define BT_CYLGRP 2 /* Buffer holds a cylinder group map */ #define BT_LEVEL1 3 /* Buffer holds single level indirect */ #define BT_LEVEL2 4 /* Buffer holds double level indirect */ #define BT_LEVEL3 5 /* Buffer holds triple level indirect */ #define BT_EXTATTR 6 /* Buffer holds external attribute data */ #define BT_INODES 7 /* Buffer holds inodes */ #define BT_DIRDATA 8 /* Buffer holds directory data */ #define BT_DATA 9 /* Buffer holds user data */ #define BT_NUMBUFTYPES 10 #define BT_NAMES { \ "unknown", \ "Superblock", \ "Cylinder Group", \ "Single Level Indirect", \ "Double Level Indirect", \ "Triple Level Indirect", \ "External Attribute", \ "Inode Block", \ "Directory Contents", \ "User Data" } extern char *buftype[]; #define BT_BUFTYPE(type) \ type < BT_NUMBUFTYPES ? buftype[type] : buftype[BT_UNKNOWN] extern long readcnt[BT_NUMBUFTYPES]; extern long totalreadcnt[BT_NUMBUFTYPES]; extern struct timespec readtime[BT_NUMBUFTYPES]; extern struct timespec totalreadtime[BT_NUMBUFTYPES]; extern struct timespec startprog; extern struct bufarea *icachebp; /* inode cache buffer */ extern struct bufarea sblk; /* file system superblock */ extern struct bufarea *pdirbp; /* current directory contents */ #define dirty(bp) do { \ if (fswritefd < 0) \ pfatal("SETTING DIRTY FLAG IN READ_ONLY MODE\n"); \ else \ (bp)->b_flags |= B_DIRTY; \ } while (0) #define initbarea(bp, type) do { \ (bp)->b_bno = (ufs2_daddr_t)-4; \ (bp)->b_size = 0; \ (bp)->b_errs = 0; \ (bp)->b_flags = 0; \ (bp)->b_type = type; \ (bp)->b_refcnt = 0; \ (bp)->b_index = 0; \ } while (0) #define sbdirty() dirty(&sblk) #define sblock (*sblk.b_un.b_fs) enum fixstate {DONTKNOW, NOFIX, FIX, IGNORE}; extern ino_t cursnapshot; struct inodesc { enum fixstate id_fix; /* policy on fixing errors */ int (*id_func)(struct inodesc *); /* function to be applied to blocks of inode */ struct bufarea *id_bp; /* ckinode: buffer with indirect pointers */ union dinode *id_dp; /* ckinode: dinode being traversed */ ino_t id_number; /* inode number described */ ino_t id_parent; /* for DATA nodes, their parent */ ufs_lbn_t id_lbn; /* logical block number of current block */ ufs2_daddr_t id_blkno; /* current block number being examined */ int id_level; /* level of indirection of this block */ int id_numfrags; /* number of frags contained in block */ ufs_lbn_t id_lballoc; /* pass1: last LBN that is allocated */ off_t id_filesize; /* for DATA nodes, the size of the directory */ ufs2_daddr_t id_entryno;/* for DATA nodes, current entry number */ int id_loc; /* for DATA nodes, current location in dir */ struct direct *id_dirp; /* for DATA nodes, ptr to current entry */ char *id_name; /* for DATA nodes, name to find or enter */ char id_type; /* type of descriptor, DATA, ADDR, or SNAP */ }; /* file types */ #define DATA 1 /* a directory */ #define SNAP 2 /* a snapshot */ #define ADDR 3 /* anything but a directory or a snapshot */ /* * Linked list of duplicate blocks. * * The list is composed of two parts. The first part of the * list (from duplist through the node pointed to by muldup) * contains a single copy of each duplicate block that has been * found. The second part of the list (from muldup to the end) * contains duplicate blocks that have been found more than once. * To check if a block has been found as a duplicate it is only * necessary to search from duplist through muldup. To find the * total number of times that a block has been found as a duplicate * the entire list must be searched for occurrences of the block * in question. The following diagram shows a sample list where * w (found twice), x (found once), y (found three times), and z * (found once) are duplicate block numbers: * * w -> y -> x -> z -> y -> w -> y * ^ ^ * | | * duplist muldup */ struct dups { struct dups *next; ufs2_daddr_t dup; }; extern struct dups *duplist; /* head of dup list */ extern struct dups *muldup; /* end of unique duplicate dup block numbers */ /* * Inode cache data structures. */ struct inoinfo { SLIST_ENTRY(inoinfo) i_hash; /* hash list */ ino_t i_number; /* inode number of this entry */ ino_t i_parent; /* inode number of parent */ ino_t i_dotdot; /* inode number of `..' */ size_t i_isize; /* size of inode */ u_int i_depth; /* depth of directory from root */ u_int i_flags; /* flags, see below */ u_int i_numblks; /* size of block array in bytes */ ufs2_daddr_t i_blks[1]; /* actually longer */ }; extern SLIST_HEAD(inohash, inoinfo) *inphash; extern struct inoinfo **inpsort; /* * flags for struct inoinfo */ #define INFO_NEW 0x0000001 /* replaced broken directory */ extern long dirhash, inplast; extern unsigned long numdirs, listmax; extern long countdirs; /* number of directories we actually found */ #define MIBSIZE 3 /* size of fsck sysctl MIBs */ extern int adjblkcnt[MIBSIZE]; /* MIB cmd to adjust inode block count */ extern int adjrefcnt[MIBSIZE]; /* MIB cmd to adjust inode reference count */ extern int adjndir[MIBSIZE]; /* MIB cmd to adjust number of directories */ extern int adjnbfree[MIBSIZE]; /* MIB cmd to adjust number of free blocks */ extern int adjnifree[MIBSIZE]; /* MIB cmd to adjust number of free inodes */ extern int adjnffree[MIBSIZE]; /* MIB cmd to adjust number of free frags */ extern int adjnumclusters[MIBSIZE]; /* MIB cmd adjust number of free clusters */ +extern int adjdepth[MIBSIZE]; /* MIB cmd to adjust directory depth count */ extern int freefiles[MIBSIZE]; /* MIB cmd to free a set of files */ extern int freedirs[MIBSIZE]; /* MIB cmd to free a set of directories */ extern int freeblks[MIBSIZE]; /* MIB cmd to free a set of data blocks */ extern int setsize[MIBSIZE]; /* MIB cmd to set inode size */ extern struct fsck_cmd cmd; /* sysctl file system update commands */ extern int bkgrdcheck; /* determine if background check is possible */ extern int bkgrdsumadj; /* whether the kernel has the ability to adjust the superblock summary fields */ extern off_t bflag; /* location of alternate super block */ extern int bkgrdflag; /* use a snapshot to run on an active system */ extern char *blockmap; /* ptr to primary blk allocation map */ extern char *cdevname; /* name of device being checked */ extern int cgheader_corrupt; /* one or more CG headers are corrupt */ extern char ckclean; /* only do work if not cleanly unmounted */ extern int ckhashadd; /* check hashes to be added */ extern char *copybuf; /* buffer to copy snapshot blocks */ extern int cvtlevel; /* convert to newer file system format */ extern long dev_bsize; /* computed value of DEV_BSIZE */ extern u_int real_dev_bsize; /* actual disk sector size, not overridden */ extern int debug; /* output debugging info */ extern int Eflag; /* delete empty data blocks */ extern int fsmodified; /* 1 => write done to file system */ extern int fsreadfd; /* file descriptor for reading file system */ extern int fswritefd; /* file descriptor for writing file system */ extern char havesb; /* superblock has been read */ extern int inoopt; /* trim out unused inodes */ extern ino_t lfdir; /* lost & found directory inode number */ extern int lfmode; /* lost & found directory creation mode */ extern const char *lfname; /* lost & found directory name */ extern ufs2_daddr_t maxfsblock; /* number of blocks in the file system */ extern ino_t maxino; /* number of inodes in file system */ extern ufs2_daddr_t n_blks; /* number of blocks in use */ extern ino_t n_files; /* number of files in use */ extern char nflag; /* assume a no response */ extern char preen; /* just fix normal inconsistencies */ extern char rerun; /* rerun fsck. Only used in non-preen mode */ extern char resolved; /* cleared if unresolved changes => not clean */ extern int returntosingle; /* 1 => return to single user mode on exit */ extern long secsize; /* actual disk sector size */ extern char skipclean; /* skip clean file systems if preening */ extern int snapcnt; /* number of active snapshots */ extern struct inode snaplist[FSMAXSNAP + 1]; /* list of active snapshots */ extern char snapname[BUFSIZ]; /* when doing snapshots, the name of the file */ extern int sujrecovery; /* 1 => doing check using the journal */ extern int surrender; /* Give up if reads fail */ extern char usedsoftdep; /* just fix soft dependency inconsistencies */ extern int wantrestart; /* Restart fsck on early termination */ extern char yflag; /* assume a yes response */ extern int zflag; /* zero unused directory space */ extern int Zflag; /* zero empty data blocks */ extern volatile sig_atomic_t got_siginfo; /* received a SIGINFO */ extern volatile sig_atomic_t got_sigalarm; /* received a SIGALRM */ #define clearinode(dp) \ if (sblock.fs_magic == FS_UFS1_MAGIC) { \ (dp)->dp1 = zino.dp1; \ } else { \ (dp)->dp2 = zino.dp2; \ } extern union dinode zino; #define setbmap(blkno) setbit(blockmap, blkno) #define testbmap(blkno) isset(blockmap, blkno) #define clrbmap(blkno) clrbit(blockmap, blkno) #define STOP 0x01 #define SKIP 0x02 #define KEEPON 0x04 #define ALTERED 0x08 #define FOUND 0x10 #define EEXIT 8 /* Standard error exit. */ #define ERERUN 16 /* fsck needs to be re-run. */ #define ERESTART -1 int flushentry(void); /* * Wrapper for malloc() that flushes the cylinder group cache to try * to get space. */ static inline void* Malloc(size_t size) { void *retval; while ((retval = malloc(size)) == NULL) if (flushentry() == 0) break; return (retval); } /* * Wrapper for calloc() that flushes the cylinder group cache to try * to get space. */ static inline void* Calloc(size_t cnt, size_t size) { void *retval; while ((retval = calloc(cnt, size)) == NULL) if (flushentry() == 0) break; return (retval); } struct fstab; void adjust(struct inodesc *, int lcnt); void alarmhandler(int sig); ufs2_daddr_t allocblk(long cg, long frags, ufs2_daddr_t (*checkblkavail) (ufs2_daddr_t blkno, long frags)); ino_t allocdir(ino_t parent, ino_t request, int mode); ino_t allocino(ino_t request, int type); void binval(struct bufarea *); void blkerror(ino_t ino, const char *type, ufs2_daddr_t blk); char *blockcheck(char *name); int blread(int fd, char *buf, ufs2_daddr_t blk, long size); void bufinit(void); void blwrite(int fd, char *buf, ufs2_daddr_t blk, ssize_t size); void blerase(int fd, ufs2_daddr_t blk, long size); void blzero(int fd, ufs2_daddr_t blk, long size); void brelse(struct bufarea *); struct inoinfo *cacheino(union dinode *dp, ino_t inumber); void catch(int); void catchquit(int); void cgdirty(struct bufarea *); struct bufarea *cglookup(int cg); int changeino(ino_t dir, const char *name, ino_t newnum, int depth); void check_blkcnt(struct inode *ip); int check_cgmagic(int cg, struct bufarea *cgbp); void rebuild_cg(int cg, struct bufarea *cgbp); void check_dirdepth(struct inoinfo *inp); int chkrange(ufs2_daddr_t blk, int cnt); void ckfini(int markclean); int ckinode(union dinode *dp, struct inodesc *); void clri(struct inodesc *, const char *type, int flag); int clearentry(struct inodesc *); void copyonwrite(struct fs *, struct bufarea *, ufs2_daddr_t (*checkblkavail)(ufs2_daddr_t, long)); void direrror(ino_t ino, const char *errmesg); int dirscan(struct inodesc *); int dofix(struct inodesc *, const char *msg); int eascan(struct inodesc *, struct ufs2_dinode *dp); void fileerror(ino_t cwd, ino_t ino, const char *errmesg); void finalIOstats(void); int findino(struct inodesc *); int findname(struct inodesc *); void flush(int fd, struct bufarea *bp); int freeblock(struct inodesc *); void freedirino(ino_t ino, ino_t parent); void freeino(ino_t ino); void freeinodebuf(void); void fsckinit(void); void fsutilinit(void); int ftypeok(union dinode *dp); void getblk(struct bufarea *bp, ufs2_daddr_t blk, long size); struct bufarea *getdatablk(ufs2_daddr_t blkno, long size, int type); struct inoinfo *getinoinfo(ino_t inumber); union dinode *getnextinode(ino_t inumber, int rebuiltcg); void getpathname(char *namebuf, ino_t curdir, ino_t ino); void ginode(ino_t, struct inode *); void gjournal_check(const char *filesys); void infohandler(int sig); void irelse(struct inode *); ufs2_daddr_t ino_blkatoff(union dinode *, ino_t, ufs_lbn_t, int *, struct bufarea **); void inocleanup(void); void inodirty(struct inode *); struct inostat *inoinfo(ino_t inum); void IOstats(char *what); int linkup(ino_t orphan, ino_t parentdir, char *name); int makeentry(ino_t parent, ino_t ino, const char *name); int openfilesys(char *dev); void panic(const char *fmt, ...) __printflike(1, 2); void pass1(void); void pass1b(void); int pass1check(struct inodesc *); void pass2(void); void pass3(void); void pass4(void); void pass5(void); void pfatal(const char *fmt, ...) __printflike(1, 2); void propagate(void); void prtbuf(struct bufarea *, const char *, ...) __printflike(2, 3); void prtinode(struct inode *); void pwarn(const char *fmt, ...) __printflike(1, 2); int readsb(void); int removecachedino(ino_t); int reply(const char *question); void rwerror(const char *mesg, ufs2_daddr_t blk); void sblock_init(void); void setinodebuf(int, ino_t); int setup(char *dev); int snapblkfree(struct fs *, ufs2_daddr_t, long, ino_t, ufs2_daddr_t (*)(ufs2_daddr_t, long)); void snapremove(ino_t); void snapflush(ufs2_daddr_t (*checkblkavail)(ufs2_daddr_t, long)); ufs2_daddr_t std_checkblkavail(ufs2_daddr_t blkno, long frags); ufs2_daddr_t suj_checkblkavail(ufs2_daddr_t, long); int suj_check(const char *filesys); void update_maps(struct cg *, struct cg*, int); #endif /* !_FSCK_H_ */ diff --git a/sbin/fsck_ffs/globs.c b/sbin/fsck_ffs/globs.c index bce499754301..7aef7364eeba 100644 --- a/sbin/fsck_ffs/globs.c +++ b/sbin/fsck_ffs/globs.c @@ -1,173 +1,175 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1980, 1986, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 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. 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. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #if 0 #ifndef lint static const char copyright[] = "@(#) Copyright (c) 1980, 1986, 1993\n\ The Regents of the University of California. All rights reserved.\n"; #endif /* not lint */ #ifndef lint static char sccsid[] = "@(#)main.c 8.6 (Berkeley) 5/14/95"; #endif /* not lint */ #endif #include __FBSDID("$FreeBSD$"); #include #include #include #include #include "fsck.h" long readcnt[BT_NUMBUFTYPES]; long totalreadcnt[BT_NUMBUFTYPES]; struct timespec readtime[BT_NUMBUFTYPES]; struct timespec totalreadtime[BT_NUMBUFTYPES]; struct timespec startprog; struct bufarea sblk; /* file system superblock */ struct bufarea *pdirbp; /* current directory contents */ ino_t cursnapshot; long dirhash, inplast; unsigned long numdirs, listmax; long countdirs; /* number of directories we actually found */ int adjrefcnt[MIBSIZE]; /* MIB command to adjust inode reference cnt */ int adjblkcnt[MIBSIZE]; /* MIB command to adjust inode block count */ int setsize[MIBSIZE]; /* MIB command to set inode size */ int adjndir[MIBSIZE]; /* MIB command to adjust number of directories */ int adjnbfree[MIBSIZE]; /* MIB command to adjust number of free blocks */ int adjnifree[MIBSIZE]; /* MIB command to adjust number of free inodes */ int adjnffree[MIBSIZE]; /* MIB command to adjust number of free frags */ int adjnumclusters[MIBSIZE]; /* MIB command to adjust number of free clusters */ +int adjdepth[MIBSIZE]; /* MIB cmd to adjust directory depth count */ int freefiles[MIBSIZE]; /* MIB command to free a set of files */ int freedirs[MIBSIZE]; /* MIB command to free a set of directories */ int freeblks[MIBSIZE]; /* MIB command to free a set of data blocks */ struct fsck_cmd cmd; /* sysctl file system update commands */ char snapname[BUFSIZ]; /* when doing snapshots, the name of the file */ char *cdevname; /* name of device being checked */ long dev_bsize; /* computed value of DEV_BSIZE */ long secsize; /* actual disk sector size */ u_int real_dev_bsize; /* actual disk sector size, not overridden */ char nflag; /* assume a no response */ char yflag; /* assume a yes response */ int bkgrdflag; /* use a snapshot to run on an active system */ off_t bflag; /* location of alternate super block */ int debug; /* output debugging info */ int Eflag; /* delete empty data blocks */ int Zflag; /* zero empty data blocks */ int zflag; /* zero unused directory space */ int inoopt; /* trim out unused inodes */ char ckclean; /* only do work if not cleanly unmounted */ int cvtlevel; /* convert to newer file system format */ int ckhashadd; /* check hashes to be added */ int bkgrdcheck; /* determine if background check is possible */ int bkgrdsumadj; /* whether the kernel have ability to adjust superblock summary */ char usedsoftdep; /* just fix soft dependency inconsistencies */ char preen; /* just fix normal inconsistencies */ char rerun; /* rerun fsck. Only used in non-preen mode */ int returntosingle; /* 1 => return to single user mode on exit */ char resolved; /* cleared if unresolved changes => not clean */ char havesb; /* superblock has been read */ char skipclean; /* skip clean file systems if preening */ int fsmodified; /* 1 => write done to file system */ int fsreadfd; /* file descriptor for reading file system */ int fswritefd; /* file descriptor for writing file system */ int surrender; /* Give up if reads fail */ int wantrestart; /* Restart fsck on early termination */ ufs2_daddr_t maxfsblock; /* number of blocks in the file system */ char *blockmap; /* ptr to primary blk allocation map */ ino_t maxino; /* number of inodes in file system */ ino_t lfdir; /* lost & found directory inode number */ const char *lfname; /* lost & found directory name */ int lfmode; /* lost & found directory creation mode */ ufs2_daddr_t n_blks; /* number of blocks in use */ int cgheader_corrupt; /* one or more CG headers are corrupt */ ino_t n_files; /* number of files in use */ volatile sig_atomic_t got_siginfo; /* received a SIGINFO */ volatile sig_atomic_t got_sigalarm; /* received a SIGALRM */ union dinode zino; struct dups *duplist; struct dups *muldup; struct inostatlist *inostathead; void fsckinit(void) { bzero(readcnt, sizeof(long) * BT_NUMBUFTYPES); bzero(totalreadcnt, sizeof(long) * BT_NUMBUFTYPES); bzero(readtime, sizeof(struct timespec) * BT_NUMBUFTYPES); bzero(totalreadtime, sizeof(struct timespec) * BT_NUMBUFTYPES); bzero(&startprog, sizeof(struct timespec)); bzero(&sblk, sizeof(struct bufarea)); cursnapshot = 0; listmax = numdirs = dirhash = inplast = 0; countdirs = 0; bzero(adjrefcnt, sizeof(int) * MIBSIZE); bzero(adjblkcnt, sizeof(int) * MIBSIZE); bzero(setsize, sizeof(int) * MIBSIZE); bzero(adjndir, sizeof(int) * MIBSIZE); bzero(adjnbfree, sizeof(int) * MIBSIZE); bzero(adjnifree, sizeof(int) * MIBSIZE); bzero(adjnffree, sizeof(int) * MIBSIZE); bzero(adjnumclusters, sizeof(int) * MIBSIZE); + bzero(adjdepth, sizeof(int) * MIBSIZE); bzero(freefiles, sizeof(int) * MIBSIZE); bzero(freedirs, sizeof(int) * MIBSIZE); bzero(freeblks, sizeof(int) * MIBSIZE); bzero(&cmd, sizeof(struct fsck_cmd)); bzero(snapname, sizeof(char) * BUFSIZ); cdevname = NULL; dev_bsize = 0; secsize = 0; real_dev_bsize = 0; bkgrdsumadj = 0; usedsoftdep = 0; rerun = 0; returntosingle = 0; resolved = 0; havesb = 0; fsmodified = 0; fsreadfd = -1; fswritefd = -1; maxfsblock = 0; maxino = 0; lfdir = 0; lfname = "lost+found"; lfmode = 0700; n_blks = 0; n_files = 0; cgheader_corrupt = 0; got_siginfo = 0; got_sigalarm = 0; bzero(&zino.dp1, sizeof(struct ufs1_dinode)); bzero(&zino.dp2, sizeof(struct ufs2_dinode)); } diff --git a/sys/ufs/ffs/ffs_alloc.c b/sys/ufs/ffs/ffs_alloc.c index fbc99951bcc6..957be753ce15 100644 --- a/sys/ufs/ffs/ffs_alloc.c +++ b/sys/ufs/ffs/ffs_alloc.c @@ -1,3523 +1,3551 @@ /*- * SPDX-License-Identifier: (BSD-2-Clause AND BSD-3-Clause) * * Copyright (c) 2002 Networks Associates Technology, Inc. * All rights reserved. * * This software was developed for the FreeBSD Project by Marshall * Kirk McKusick and Network Associates Laboratories, the Security * Research Division of Network Associates, Inc. under DARPA/SPAWAR * contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS * research program * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 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. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * Copyright (c) 1982, 1986, 1989, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 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. 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. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)ffs_alloc.c 8.18 (Berkeley) 5/26/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_quota.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include typedef ufs2_daddr_t allocfcn_t(struct inode *ip, u_int cg, ufs2_daddr_t bpref, int size, int rsize); static ufs2_daddr_t ffs_alloccg(struct inode *, u_int, ufs2_daddr_t, int, int); static ufs2_daddr_t ffs_alloccgblk(struct inode *, struct buf *, ufs2_daddr_t, int); static void ffs_blkfree_cg(struct ufsmount *, struct fs *, struct vnode *, ufs2_daddr_t, long, ino_t, struct workhead *); #ifdef INVARIANTS static int ffs_checkblk(struct inode *, ufs2_daddr_t, long); #endif static ufs2_daddr_t ffs_clusteralloc(struct inode *, u_int, ufs2_daddr_t, int); static ino_t ffs_dirpref(struct inode *); static ufs2_daddr_t ffs_fragextend(struct inode *, u_int, ufs2_daddr_t, int, int); static ufs2_daddr_t ffs_hashalloc (struct inode *, u_int, ufs2_daddr_t, int, int, allocfcn_t *); static ufs2_daddr_t ffs_nodealloccg(struct inode *, u_int, ufs2_daddr_t, int, int); static ufs1_daddr_t ffs_mapsearch(struct fs *, struct cg *, ufs2_daddr_t, int); static int ffs_reallocblks_ufs1(struct vop_reallocblks_args *); static int ffs_reallocblks_ufs2(struct vop_reallocblks_args *); static void ffs_ckhash_cg(struct buf *); /* * Allocate a block in the filesystem. * * The size of the requested block is given, which must be some * multiple of fs_fsize and <= fs_bsize. * A preference may be optionally specified. If a preference is given * the following hierarchy is used to allocate a block: * 1) allocate the requested block. * 2) allocate a rotationally optimal block in the same cylinder. * 3) allocate a block in the same cylinder group. * 4) quadratically rehash into other cylinder groups, until an * available block is located. * If no block preference is given the following hierarchy is used * to allocate a block: * 1) allocate a block in the cylinder group that contains the * inode for the file. * 2) quadratically rehash into other cylinder groups, until an * available block is located. */ int ffs_alloc(struct inode *ip, ufs2_daddr_t lbn, ufs2_daddr_t bpref, int size, int flags, struct ucred *cred, ufs2_daddr_t *bnp) { struct fs *fs; struct ufsmount *ump; ufs2_daddr_t bno; u_int cg, reclaimed; int64_t delta; #ifdef QUOTA int error; #endif *bnp = 0; ump = ITOUMP(ip); fs = ump->um_fs; mtx_assert(UFS_MTX(ump), MA_OWNED); #ifdef INVARIANTS if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) { printf("dev = %s, bsize = %ld, size = %d, fs = %s\n", devtoname(ump->um_dev), (long)fs->fs_bsize, size, fs->fs_fsmnt); panic("ffs_alloc: bad size"); } if (cred == NOCRED) panic("ffs_alloc: missing credential"); #endif /* INVARIANTS */ reclaimed = 0; retry: #ifdef QUOTA UFS_UNLOCK(ump); error = chkdq(ip, btodb(size), cred, 0); if (error) return (error); UFS_LOCK(ump); #endif if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0) goto nospace; if (priv_check_cred(cred, PRIV_VFS_BLOCKRESERVE) && freespace(fs, fs->fs_minfree) - numfrags(fs, size) < 0) goto nospace; if (bpref >= fs->fs_size) bpref = 0; if (bpref == 0) cg = ino_to_cg(fs, ip->i_number); else cg = dtog(fs, bpref); bno = ffs_hashalloc(ip, cg, bpref, size, size, ffs_alloccg); if (bno > 0) { delta = btodb(size); DIP_SET(ip, i_blocks, DIP(ip, i_blocks) + delta); if (flags & IO_EXT) UFS_INODE_SET_FLAG(ip, IN_CHANGE); else UFS_INODE_SET_FLAG(ip, IN_CHANGE | IN_UPDATE); *bnp = bno; return (0); } nospace: #ifdef QUOTA UFS_UNLOCK(ump); /* * Restore user's disk quota because allocation failed. */ (void) chkdq(ip, -btodb(size), cred, FORCE); UFS_LOCK(ump); #endif if (reclaimed == 0 && (flags & IO_BUFLOCKED) == 0) { reclaimed = 1; softdep_request_cleanup(fs, ITOV(ip), cred, FLUSH_BLOCKS_WAIT); goto retry; } if (ffs_fsfail_cleanup_locked(ump, 0)) { UFS_UNLOCK(ump); return (ENXIO); } if (reclaimed > 0 && ppsratecheck(&ump->um_last_fullmsg, &ump->um_secs_fullmsg, 1)) { UFS_UNLOCK(ump); ffs_fserr(fs, ip->i_number, "filesystem full"); uprintf("\n%s: write failed, filesystem is full\n", fs->fs_fsmnt); } else { UFS_UNLOCK(ump); } return (ENOSPC); } /* * Reallocate a fragment to a bigger size * * The number and size of the old block is given, and a preference * and new size is also specified. The allocator attempts to extend * the original block. Failing that, the regular block allocator is * invoked to get an appropriate block. */ int ffs_realloccg(struct inode *ip, ufs2_daddr_t lbprev, ufs2_daddr_t bprev, ufs2_daddr_t bpref, int osize, int nsize, int flags, struct ucred *cred, struct buf **bpp) { struct vnode *vp; struct fs *fs; struct buf *bp; struct ufsmount *ump; u_int cg, request, reclaimed; int error, gbflags; ufs2_daddr_t bno; int64_t delta; vp = ITOV(ip); ump = ITOUMP(ip); fs = ump->um_fs; bp = NULL; gbflags = (flags & BA_UNMAPPED) != 0 ? GB_UNMAPPED : 0; #ifdef WITNESS gbflags |= IS_SNAPSHOT(ip) ? GB_NOWITNESS : 0; #endif mtx_assert(UFS_MTX(ump), MA_OWNED); #ifdef INVARIANTS if (vp->v_mount->mnt_kern_flag & MNTK_SUSPENDED) panic("ffs_realloccg: allocation on suspended filesystem"); if ((u_int)osize > fs->fs_bsize || fragoff(fs, osize) != 0 || (u_int)nsize > fs->fs_bsize || fragoff(fs, nsize) != 0) { printf( "dev = %s, bsize = %ld, osize = %d, nsize = %d, fs = %s\n", devtoname(ump->um_dev), (long)fs->fs_bsize, osize, nsize, fs->fs_fsmnt); panic("ffs_realloccg: bad size"); } if (cred == NOCRED) panic("ffs_realloccg: missing credential"); #endif /* INVARIANTS */ reclaimed = 0; retry: if (priv_check_cred(cred, PRIV_VFS_BLOCKRESERVE) && freespace(fs, fs->fs_minfree) - numfrags(fs, nsize - osize) < 0) { goto nospace; } if (bprev == 0) { printf("dev = %s, bsize = %ld, bprev = %jd, fs = %s\n", devtoname(ump->um_dev), (long)fs->fs_bsize, (intmax_t)bprev, fs->fs_fsmnt); panic("ffs_realloccg: bad bprev"); } UFS_UNLOCK(ump); /* * Allocate the extra space in the buffer. */ error = bread_gb(vp, lbprev, osize, NOCRED, gbflags, &bp); if (error) { return (error); } if (bp->b_blkno == bp->b_lblkno) { if (lbprev >= UFS_NDADDR) panic("ffs_realloccg: lbprev out of range"); bp->b_blkno = fsbtodb(fs, bprev); } #ifdef QUOTA error = chkdq(ip, btodb(nsize - osize), cred, 0); if (error) { brelse(bp); return (error); } #endif /* * Check for extension in the existing location. */ *bpp = NULL; cg = dtog(fs, bprev); UFS_LOCK(ump); bno = ffs_fragextend(ip, cg, bprev, osize, nsize); if (bno) { if (bp->b_blkno != fsbtodb(fs, bno)) panic("ffs_realloccg: bad blockno"); delta = btodb(nsize - osize); DIP_SET(ip, i_blocks, DIP(ip, i_blocks) + delta); if (flags & IO_EXT) UFS_INODE_SET_FLAG(ip, IN_CHANGE); else UFS_INODE_SET_FLAG(ip, IN_CHANGE | IN_UPDATE); allocbuf(bp, nsize); bp->b_flags |= B_DONE; vfs_bio_bzero_buf(bp, osize, nsize - osize); if ((bp->b_flags & (B_MALLOC | B_VMIO)) == B_VMIO) vfs_bio_set_valid(bp, osize, nsize - osize); *bpp = bp; return (0); } /* * Allocate a new disk location. */ if (bpref >= fs->fs_size) bpref = 0; switch ((int)fs->fs_optim) { case FS_OPTSPACE: /* * Allocate an exact sized fragment. Although this makes * best use of space, we will waste time relocating it if * the file continues to grow. If the fragmentation is * less than half of the minimum free reserve, we choose * to begin optimizing for time. */ request = nsize; if (fs->fs_minfree <= 5 || fs->fs_cstotal.cs_nffree > (off_t)fs->fs_dsize * fs->fs_minfree / (2 * 100)) break; log(LOG_NOTICE, "%s: optimization changed from SPACE to TIME\n", fs->fs_fsmnt); fs->fs_optim = FS_OPTTIME; break; case FS_OPTTIME: /* * At this point we have discovered a file that is trying to * grow a small fragment to a larger fragment. To save time, * we allocate a full sized block, then free the unused portion. * If the file continues to grow, the `ffs_fragextend' call * above will be able to grow it in place without further * copying. If aberrant programs cause disk fragmentation to * grow within 2% of the free reserve, we choose to begin * optimizing for space. */ request = fs->fs_bsize; if (fs->fs_cstotal.cs_nffree < (off_t)fs->fs_dsize * (fs->fs_minfree - 2) / 100) break; log(LOG_NOTICE, "%s: optimization changed from TIME to SPACE\n", fs->fs_fsmnt); fs->fs_optim = FS_OPTSPACE; break; default: printf("dev = %s, optim = %ld, fs = %s\n", devtoname(ump->um_dev), (long)fs->fs_optim, fs->fs_fsmnt); panic("ffs_realloccg: bad optim"); /* NOTREACHED */ } bno = ffs_hashalloc(ip, cg, bpref, request, nsize, ffs_alloccg); if (bno > 0) { bp->b_blkno = fsbtodb(fs, bno); if (!DOINGSOFTDEP(vp)) /* * The usual case is that a smaller fragment that * was just allocated has been replaced with a bigger * fragment or a full-size block. If it is marked as * B_DELWRI, the current contents have not been written * to disk. It is possible that the block was written * earlier, but very uncommon. If the block has never * been written, there is no need to send a BIO_DELETE * for it when it is freed. The gain from avoiding the * TRIMs for the common case of unwritten blocks far * exceeds the cost of the write amplification for the * uncommon case of failing to send a TRIM for a block * that had been written. */ ffs_blkfree(ump, fs, ump->um_devvp, bprev, (long)osize, ip->i_number, vp->v_type, NULL, (bp->b_flags & B_DELWRI) != 0 ? NOTRIM_KEY : SINGLETON_KEY); delta = btodb(nsize - osize); DIP_SET(ip, i_blocks, DIP(ip, i_blocks) + delta); if (flags & IO_EXT) UFS_INODE_SET_FLAG(ip, IN_CHANGE); else UFS_INODE_SET_FLAG(ip, IN_CHANGE | IN_UPDATE); allocbuf(bp, nsize); bp->b_flags |= B_DONE; vfs_bio_bzero_buf(bp, osize, nsize - osize); if ((bp->b_flags & (B_MALLOC | B_VMIO)) == B_VMIO) vfs_bio_set_valid(bp, osize, nsize - osize); *bpp = bp; return (0); } #ifdef QUOTA UFS_UNLOCK(ump); /* * Restore user's disk quota because allocation failed. */ (void) chkdq(ip, -btodb(nsize - osize), cred, FORCE); UFS_LOCK(ump); #endif nospace: /* * no space available */ if (reclaimed == 0 && (flags & IO_BUFLOCKED) == 0) { reclaimed = 1; UFS_UNLOCK(ump); if (bp) { brelse(bp); bp = NULL; } UFS_LOCK(ump); softdep_request_cleanup(fs, vp, cred, FLUSH_BLOCKS_WAIT); goto retry; } if (bp) brelse(bp); if (ffs_fsfail_cleanup_locked(ump, 0)) { UFS_UNLOCK(ump); return (ENXIO); } if (reclaimed > 0 && ppsratecheck(&ump->um_last_fullmsg, &ump->um_secs_fullmsg, 1)) { UFS_UNLOCK(ump); ffs_fserr(fs, ip->i_number, "filesystem full"); uprintf("\n%s: write failed, filesystem is full\n", fs->fs_fsmnt); } else { UFS_UNLOCK(ump); } return (ENOSPC); } /* * Reallocate a sequence of blocks into a contiguous sequence of blocks. * * The vnode and an array of buffer pointers for a range of sequential * logical blocks to be made contiguous is given. The allocator attempts * to find a range of sequential blocks starting as close as possible * from the end of the allocation for the logical block immediately * preceding the current range. If successful, the physical block numbers * in the buffer pointers and in the inode are changed to reflect the new * allocation. If unsuccessful, the allocation is left unchanged. The * success in doing the reallocation is returned. Note that the error * return is not reflected back to the user. Rather the previous block * allocation will be used. */ SYSCTL_NODE(_vfs, OID_AUTO, ffs, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, "FFS filesystem"); static int doasyncfree = 1; SYSCTL_INT(_vfs_ffs, OID_AUTO, doasyncfree, CTLFLAG_RW, &doasyncfree, 0, "do not force synchronous writes when blocks are reallocated"); static int doreallocblks = 1; SYSCTL_INT(_vfs_ffs, OID_AUTO, doreallocblks, CTLFLAG_RW, &doreallocblks, 0, "enable block reallocation"); static int dotrimcons = 1; SYSCTL_INT(_vfs_ffs, OID_AUTO, dotrimcons, CTLFLAG_RWTUN, &dotrimcons, 0, "enable BIO_DELETE / TRIM consolidation"); static int maxclustersearch = 10; SYSCTL_INT(_vfs_ffs, OID_AUTO, maxclustersearch, CTLFLAG_RW, &maxclustersearch, 0, "max number of cylinder group to search for contigous blocks"); #ifdef DIAGNOSTIC static int prtrealloc = 0; SYSCTL_INT(_debug, OID_AUTO, ffs_prtrealloc, CTLFLAG_RW, &prtrealloc, 0, "print out FFS filesystem block reallocation operations"); #endif int ffs_reallocblks( struct vop_reallocblks_args /* { struct vnode *a_vp; struct cluster_save *a_buflist; } */ *ap) { struct ufsmount *ump; int error; /* * We used to skip reallocating the blocks of a file into a * contiguous sequence if the underlying flash device requested * BIO_DELETE notifications, because devices that benefit from * BIO_DELETE also benefit from not moving the data. However, * the destination for the data is usually moved before the data * is written to the initially allocated location, so we rarely * suffer the penalty of extra writes. With the addition of the * consolidation of contiguous blocks into single BIO_DELETE * operations, having fewer but larger contiguous blocks reduces * the number of (slow and expensive) BIO_DELETE operations. So * when doing BIO_DELETE consolidation, we do block reallocation. * * Skip if reallocblks has been disabled globally. */ ump = ap->a_vp->v_mount->mnt_data; if ((((ump->um_flags) & UM_CANDELETE) != 0 && dotrimcons == 0) || doreallocblks == 0) return (ENOSPC); /* * We can't wait in softdep prealloc as it may fsync and recurse * here. Instead we simply fail to reallocate blocks if this * rare condition arises. */ if (DOINGSUJ(ap->a_vp)) if (softdep_prealloc(ap->a_vp, MNT_NOWAIT) != 0) return (ENOSPC); vn_seqc_write_begin(ap->a_vp); error = ump->um_fstype == UFS1 ? ffs_reallocblks_ufs1(ap) : ffs_reallocblks_ufs2(ap); vn_seqc_write_end(ap->a_vp); return (error); } static int ffs_reallocblks_ufs1( struct vop_reallocblks_args /* { struct vnode *a_vp; struct cluster_save *a_buflist; } */ *ap) { struct fs *fs; struct inode *ip; struct vnode *vp; struct buf *sbp, *ebp, *bp; ufs1_daddr_t *bap, *sbap, *ebap; struct cluster_save *buflist; struct ufsmount *ump; ufs_lbn_t start_lbn, end_lbn; ufs1_daddr_t soff, newblk, blkno; ufs2_daddr_t pref; struct indir start_ap[UFS_NIADDR + 1], end_ap[UFS_NIADDR + 1], *idp; int i, cg, len, start_lvl, end_lvl, ssize; vp = ap->a_vp; ip = VTOI(vp); ump = ITOUMP(ip); fs = ump->um_fs; /* * If we are not tracking block clusters or if we have less than 4% * free blocks left, then do not attempt to cluster. Running with * less than 5% free block reserve is not recommended and those that * choose to do so do not expect to have good file layout. */ if (fs->fs_contigsumsize <= 0 || freespace(fs, 4) < 0) return (ENOSPC); buflist = ap->a_buflist; len = buflist->bs_nchildren; start_lbn = buflist->bs_children[0]->b_lblkno; end_lbn = start_lbn + len - 1; #ifdef INVARIANTS for (i = 0; i < len; i++) if (!ffs_checkblk(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize)) panic("ffs_reallocblks: unallocated block 1"); for (i = 1; i < len; i++) if (buflist->bs_children[i]->b_lblkno != start_lbn + i) panic("ffs_reallocblks: non-logical cluster"); blkno = buflist->bs_children[0]->b_blkno; ssize = fsbtodb(fs, fs->fs_frag); for (i = 1; i < len - 1; i++) if (buflist->bs_children[i]->b_blkno != blkno + (i * ssize)) panic("ffs_reallocblks: non-physical cluster %d", i); #endif /* * If the cluster crosses the boundary for the first indirect * block, leave space for the indirect block. Indirect blocks * are initially laid out in a position after the last direct * block. Block reallocation would usually destroy locality by * moving the indirect block out of the way to make room for * data blocks if we didn't compensate here. We should also do * this for other indirect block boundaries, but it is only * important for the first one. */ if (start_lbn < UFS_NDADDR && end_lbn >= UFS_NDADDR) return (ENOSPC); /* * If the latest allocation is in a new cylinder group, assume that * the filesystem has decided to move and do not force it back to * the previous cylinder group. */ if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) != dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno))) return (ENOSPC); if (ufs_getlbns(vp, start_lbn, start_ap, &start_lvl) || ufs_getlbns(vp, end_lbn, end_ap, &end_lvl)) return (ENOSPC); /* * Get the starting offset and block map for the first block. */ if (start_lvl == 0) { sbap = &ip->i_din1->di_db[0]; soff = start_lbn; } else { idp = &start_ap[start_lvl - 1]; if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &sbp)) { brelse(sbp); return (ENOSPC); } sbap = (ufs1_daddr_t *)sbp->b_data; soff = idp->in_off; } /* * If the block range spans two block maps, get the second map. */ ebap = NULL; if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) { ssize = len; } else { #ifdef INVARIANTS if (start_lvl > 0 && start_ap[start_lvl - 1].in_lbn == idp->in_lbn) panic("ffs_reallocblk: start == end"); #endif ssize = len - (idp->in_off + 1); if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &ebp)) goto fail; ebap = (ufs1_daddr_t *)ebp->b_data; } /* * Find the preferred location for the cluster. If we have not * previously failed at this endeavor, then follow our standard * preference calculation. If we have failed at it, then pick up * where we last ended our search. */ UFS_LOCK(ump); if (ip->i_nextclustercg == -1) pref = ffs_blkpref_ufs1(ip, start_lbn, soff, sbap); else pref = cgdata(fs, ip->i_nextclustercg); /* * Search the block map looking for an allocation of the desired size. * To avoid wasting too much time, we limit the number of cylinder * groups that we will search. */ cg = dtog(fs, pref); for (i = min(maxclustersearch, fs->fs_ncg); i > 0; i--) { if ((newblk = ffs_clusteralloc(ip, cg, pref, len)) != 0) break; cg += 1; if (cg >= fs->fs_ncg) cg = 0; } /* * If we have failed in our search, record where we gave up for * next time. Otherwise, fall back to our usual search citerion. */ if (newblk == 0) { ip->i_nextclustercg = cg; UFS_UNLOCK(ump); goto fail; } ip->i_nextclustercg = -1; /* * We have found a new contiguous block. * * First we have to replace the old block pointers with the new * block pointers in the inode and indirect blocks associated * with the file. */ #ifdef DIAGNOSTIC if (prtrealloc) printf("realloc: ino %ju, lbns %jd-%jd\n\told:", (uintmax_t)ip->i_number, (intmax_t)start_lbn, (intmax_t)end_lbn); #endif blkno = newblk; for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->fs_frag) { if (i == ssize) { bap = ebap; soff = -i; } #ifdef INVARIANTS if (!ffs_checkblk(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize)) panic("ffs_reallocblks: unallocated block 2"); if (dbtofsb(fs, buflist->bs_children[i]->b_blkno) != *bap) panic("ffs_reallocblks: alloc mismatch"); #endif #ifdef DIAGNOSTIC if (prtrealloc) printf(" %d,", *bap); #endif if (DOINGSOFTDEP(vp)) { if (sbap == &ip->i_din1->di_db[0] && i < ssize) softdep_setup_allocdirect(ip, start_lbn + i, blkno, *bap, fs->fs_bsize, fs->fs_bsize, buflist->bs_children[i]); else softdep_setup_allocindir_page(ip, start_lbn + i, i < ssize ? sbp : ebp, soff + i, blkno, *bap, buflist->bs_children[i]); } *bap++ = blkno; } /* * Next we must write out the modified inode and indirect blocks. * For strict correctness, the writes should be synchronous since * the old block values may have been written to disk. In practise * they are almost never written, but if we are concerned about * strict correctness, the `doasyncfree' flag should be set to zero. * * The test on `doasyncfree' should be changed to test a flag * that shows whether the associated buffers and inodes have * been written. The flag should be set when the cluster is * started and cleared whenever the buffer or inode is flushed. * We can then check below to see if it is set, and do the * synchronous write only when it has been cleared. */ if (sbap != &ip->i_din1->di_db[0]) { if (doasyncfree) bdwrite(sbp); else bwrite(sbp); } else { UFS_INODE_SET_FLAG(ip, IN_CHANGE | IN_UPDATE); if (!doasyncfree) ffs_update(vp, 1); } if (ssize < len) { if (doasyncfree) bdwrite(ebp); else bwrite(ebp); } /* * Last, free the old blocks and assign the new blocks to the buffers. */ #ifdef DIAGNOSTIC if (prtrealloc) printf("\n\tnew:"); #endif for (blkno = newblk, i = 0; i < len; i++, blkno += fs->fs_frag) { bp = buflist->bs_children[i]; if (!DOINGSOFTDEP(vp)) /* * The usual case is that a set of N-contiguous blocks * that was just allocated has been replaced with a * set of N+1-contiguous blocks. If they are marked as * B_DELWRI, the current contents have not been written * to disk. It is possible that the blocks were written * earlier, but very uncommon. If the blocks have never * been written, there is no need to send a BIO_DELETE * for them when they are freed. The gain from avoiding * the TRIMs for the common case of unwritten blocks * far exceeds the cost of the write amplification for * the uncommon case of failing to send a TRIM for the * blocks that had been written. */ ffs_blkfree(ump, fs, ump->um_devvp, dbtofsb(fs, bp->b_blkno), fs->fs_bsize, ip->i_number, vp->v_type, NULL, (bp->b_flags & B_DELWRI) != 0 ? NOTRIM_KEY : SINGLETON_KEY); bp->b_blkno = fsbtodb(fs, blkno); #ifdef INVARIANTS if (!ffs_checkblk(ip, dbtofsb(fs, bp->b_blkno), fs->fs_bsize)) panic("ffs_reallocblks: unallocated block 3"); #endif #ifdef DIAGNOSTIC if (prtrealloc) printf(" %d,", blkno); #endif } #ifdef DIAGNOSTIC if (prtrealloc) { prtrealloc--; printf("\n"); } #endif return (0); fail: if (ssize < len) brelse(ebp); if (sbap != &ip->i_din1->di_db[0]) brelse(sbp); return (ENOSPC); } static int ffs_reallocblks_ufs2( struct vop_reallocblks_args /* { struct vnode *a_vp; struct cluster_save *a_buflist; } */ *ap) { struct fs *fs; struct inode *ip; struct vnode *vp; struct buf *sbp, *ebp, *bp; ufs2_daddr_t *bap, *sbap, *ebap; struct cluster_save *buflist; struct ufsmount *ump; ufs_lbn_t start_lbn, end_lbn; ufs2_daddr_t soff, newblk, blkno, pref; struct indir start_ap[UFS_NIADDR + 1], end_ap[UFS_NIADDR + 1], *idp; int i, cg, len, start_lvl, end_lvl, ssize; vp = ap->a_vp; ip = VTOI(vp); ump = ITOUMP(ip); fs = ump->um_fs; /* * If we are not tracking block clusters or if we have less than 4% * free blocks left, then do not attempt to cluster. Running with * less than 5% free block reserve is not recommended and those that * choose to do so do not expect to have good file layout. */ if (fs->fs_contigsumsize <= 0 || freespace(fs, 4) < 0) return (ENOSPC); buflist = ap->a_buflist; len = buflist->bs_nchildren; start_lbn = buflist->bs_children[0]->b_lblkno; end_lbn = start_lbn + len - 1; #ifdef INVARIANTS for (i = 0; i < len; i++) if (!ffs_checkblk(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize)) panic("ffs_reallocblks: unallocated block 1"); for (i = 1; i < len; i++) if (buflist->bs_children[i]->b_lblkno != start_lbn + i) panic("ffs_reallocblks: non-logical cluster"); blkno = buflist->bs_children[0]->b_blkno; ssize = fsbtodb(fs, fs->fs_frag); for (i = 1; i < len - 1; i++) if (buflist->bs_children[i]->b_blkno != blkno + (i * ssize)) panic("ffs_reallocblks: non-physical cluster %d", i); #endif /* * If the cluster crosses the boundary for the first indirect * block, do not move anything in it. Indirect blocks are * usually initially laid out in a position between the data * blocks. Block reallocation would usually destroy locality by * moving the indirect block out of the way to make room for * data blocks if we didn't compensate here. We should also do * this for other indirect block boundaries, but it is only * important for the first one. */ if (start_lbn < UFS_NDADDR && end_lbn >= UFS_NDADDR) return (ENOSPC); /* * If the latest allocation is in a new cylinder group, assume that * the filesystem has decided to move and do not force it back to * the previous cylinder group. */ if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) != dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno))) return (ENOSPC); if (ufs_getlbns(vp, start_lbn, start_ap, &start_lvl) || ufs_getlbns(vp, end_lbn, end_ap, &end_lvl)) return (ENOSPC); /* * Get the starting offset and block map for the first block. */ if (start_lvl == 0) { sbap = &ip->i_din2->di_db[0]; soff = start_lbn; } else { idp = &start_ap[start_lvl - 1]; if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &sbp)) { brelse(sbp); return (ENOSPC); } sbap = (ufs2_daddr_t *)sbp->b_data; soff = idp->in_off; } /* * If the block range spans two block maps, get the second map. */ ebap = NULL; if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) { ssize = len; } else { #ifdef INVARIANTS if (start_lvl > 0 && start_ap[start_lvl - 1].in_lbn == idp->in_lbn) panic("ffs_reallocblk: start == end"); #endif ssize = len - (idp->in_off + 1); if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &ebp)) goto fail; ebap = (ufs2_daddr_t *)ebp->b_data; } /* * Find the preferred location for the cluster. If we have not * previously failed at this endeavor, then follow our standard * preference calculation. If we have failed at it, then pick up * where we last ended our search. */ UFS_LOCK(ump); if (ip->i_nextclustercg == -1) pref = ffs_blkpref_ufs2(ip, start_lbn, soff, sbap); else pref = cgdata(fs, ip->i_nextclustercg); /* * Search the block map looking for an allocation of the desired size. * To avoid wasting too much time, we limit the number of cylinder * groups that we will search. */ cg = dtog(fs, pref); for (i = min(maxclustersearch, fs->fs_ncg); i > 0; i--) { if ((newblk = ffs_clusteralloc(ip, cg, pref, len)) != 0) break; cg += 1; if (cg >= fs->fs_ncg) cg = 0; } /* * If we have failed in our search, record where we gave up for * next time. Otherwise, fall back to our usual search citerion. */ if (newblk == 0) { ip->i_nextclustercg = cg; UFS_UNLOCK(ump); goto fail; } ip->i_nextclustercg = -1; /* * We have found a new contiguous block. * * First we have to replace the old block pointers with the new * block pointers in the inode and indirect blocks associated * with the file. */ #ifdef DIAGNOSTIC if (prtrealloc) printf("realloc: ino %ju, lbns %jd-%jd\n\told:", (uintmax_t)ip->i_number, (intmax_t)start_lbn, (intmax_t)end_lbn); #endif blkno = newblk; for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->fs_frag) { if (i == ssize) { bap = ebap; soff = -i; } #ifdef INVARIANTS if (!ffs_checkblk(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize)) panic("ffs_reallocblks: unallocated block 2"); if (dbtofsb(fs, buflist->bs_children[i]->b_blkno) != *bap) panic("ffs_reallocblks: alloc mismatch"); #endif #ifdef DIAGNOSTIC if (prtrealloc) printf(" %jd,", (intmax_t)*bap); #endif if (DOINGSOFTDEP(vp)) { if (sbap == &ip->i_din2->di_db[0] && i < ssize) softdep_setup_allocdirect(ip, start_lbn + i, blkno, *bap, fs->fs_bsize, fs->fs_bsize, buflist->bs_children[i]); else softdep_setup_allocindir_page(ip, start_lbn + i, i < ssize ? sbp : ebp, soff + i, blkno, *bap, buflist->bs_children[i]); } *bap++ = blkno; } /* * Next we must write out the modified inode and indirect blocks. * For strict correctness, the writes should be synchronous since * the old block values may have been written to disk. In practise * they are almost never written, but if we are concerned about * strict correctness, the `doasyncfree' flag should be set to zero. * * The test on `doasyncfree' should be changed to test a flag * that shows whether the associated buffers and inodes have * been written. The flag should be set when the cluster is * started and cleared whenever the buffer or inode is flushed. * We can then check below to see if it is set, and do the * synchronous write only when it has been cleared. */ if (sbap != &ip->i_din2->di_db[0]) { if (doasyncfree) bdwrite(sbp); else bwrite(sbp); } else { UFS_INODE_SET_FLAG(ip, IN_CHANGE | IN_UPDATE); if (!doasyncfree) ffs_update(vp, 1); } if (ssize < len) { if (doasyncfree) bdwrite(ebp); else bwrite(ebp); } /* * Last, free the old blocks and assign the new blocks to the buffers. */ #ifdef DIAGNOSTIC if (prtrealloc) printf("\n\tnew:"); #endif for (blkno = newblk, i = 0; i < len; i++, blkno += fs->fs_frag) { bp = buflist->bs_children[i]; if (!DOINGSOFTDEP(vp)) /* * The usual case is that a set of N-contiguous blocks * that was just allocated has been replaced with a * set of N+1-contiguous blocks. If they are marked as * B_DELWRI, the current contents have not been written * to disk. It is possible that the blocks were written * earlier, but very uncommon. If the blocks have never * been written, there is no need to send a BIO_DELETE * for them when they are freed. The gain from avoiding * the TRIMs for the common case of unwritten blocks * far exceeds the cost of the write amplification for * the uncommon case of failing to send a TRIM for the * blocks that had been written. */ ffs_blkfree(ump, fs, ump->um_devvp, dbtofsb(fs, bp->b_blkno), fs->fs_bsize, ip->i_number, vp->v_type, NULL, (bp->b_flags & B_DELWRI) != 0 ? NOTRIM_KEY : SINGLETON_KEY); bp->b_blkno = fsbtodb(fs, blkno); #ifdef INVARIANTS if (!ffs_checkblk(ip, dbtofsb(fs, bp->b_blkno), fs->fs_bsize)) panic("ffs_reallocblks: unallocated block 3"); #endif #ifdef DIAGNOSTIC if (prtrealloc) printf(" %jd,", (intmax_t)blkno); #endif } #ifdef DIAGNOSTIC if (prtrealloc) { prtrealloc--; printf("\n"); } #endif return (0); fail: if (ssize < len) brelse(ebp); if (sbap != &ip->i_din2->di_db[0]) brelse(sbp); return (ENOSPC); } /* * Allocate an inode in the filesystem. * * If allocating a directory, use ffs_dirpref to select the inode. * If allocating in a directory, the following hierarchy is followed: * 1) allocate the preferred inode. * 2) allocate an inode in the same cylinder group. * 3) quadratically rehash into other cylinder groups, until an * available inode is located. * If no inode preference is given the following hierarchy is used * to allocate an inode: * 1) allocate an inode in cylinder group 0. * 2) quadratically rehash into other cylinder groups, until an * available inode is located. */ int ffs_valloc(struct vnode *pvp, int mode, struct ucred *cred, struct vnode **vpp) { struct inode *pip; struct fs *fs; struct inode *ip; struct timespec ts; struct ufsmount *ump; ino_t ino, ipref; u_int cg; int error, reclaimed; *vpp = NULL; pip = VTOI(pvp); ump = ITOUMP(pip); fs = ump->um_fs; UFS_LOCK(ump); reclaimed = 0; retry: if (fs->fs_cstotal.cs_nifree == 0) goto noinodes; if ((mode & IFMT) == IFDIR) ipref = ffs_dirpref(pip); else ipref = pip->i_number; if (ipref >= fs->fs_ncg * fs->fs_ipg) ipref = 0; cg = ino_to_cg(fs, ipref); /* * Track number of dirs created one after another * in a same cg without intervening by files. */ if ((mode & IFMT) == IFDIR) { if (fs->fs_contigdirs[cg] < 255) fs->fs_contigdirs[cg]++; } else { if (fs->fs_contigdirs[cg] > 0) fs->fs_contigdirs[cg]--; } ino = (ino_t)ffs_hashalloc(pip, cg, ipref, mode, 0, (allocfcn_t *)ffs_nodealloccg); if (ino == 0) goto noinodes; /* * Get rid of the cached old vnode, force allocation of a new vnode * for this inode. If this fails, release the allocated ino and * return the error. */ if ((error = ffs_vgetf(pvp->v_mount, ino, LK_EXCLUSIVE, vpp, FFSV_FORCEINSMQ | FFSV_REPLACE | FFSV_NEWINODE)) != 0) { ffs_vfree(pvp, ino, mode); return (error); } /* * We got an inode, so check mode and panic if it is already allocated. */ ip = VTOI(*vpp); if (ip->i_mode) { printf("mode = 0%o, inum = %ju, fs = %s\n", ip->i_mode, (uintmax_t)ip->i_number, fs->fs_fsmnt); panic("ffs_valloc: dup alloc"); } if (DIP(ip, i_blocks) && (fs->fs_flags & FS_UNCLEAN) == 0) { /* XXX */ printf("free inode %s/%lu had %ld blocks\n", fs->fs_fsmnt, (u_long)ino, (long)DIP(ip, i_blocks)); DIP_SET(ip, i_blocks, 0); } ip->i_flags = 0; DIP_SET(ip, i_flags, 0); if ((mode & IFMT) == IFDIR) DIP_SET(ip, i_dirdepth, DIP(pip, i_dirdepth) + 1); /* * Set up a new generation number for this inode. */ while (ip->i_gen == 0 || ++ip->i_gen == 0) ip->i_gen = arc4random(); DIP_SET(ip, i_gen, ip->i_gen); if (fs->fs_magic == FS_UFS2_MAGIC) { vfs_timestamp(&ts); ip->i_din2->di_birthtime = ts.tv_sec; ip->i_din2->di_birthnsec = ts.tv_nsec; } ip->i_flag = 0; (*vpp)->v_vflag = 0; (*vpp)->v_type = VNON; if (fs->fs_magic == FS_UFS2_MAGIC) { (*vpp)->v_op = &ffs_vnodeops2; UFS_INODE_SET_FLAG(ip, IN_UFS2); } else { (*vpp)->v_op = &ffs_vnodeops1; } return (0); noinodes: if (reclaimed == 0) { reclaimed = 1; softdep_request_cleanup(fs, pvp, cred, FLUSH_INODES_WAIT); goto retry; } if (ffs_fsfail_cleanup_locked(ump, 0)) { UFS_UNLOCK(ump); return (ENXIO); } if (ppsratecheck(&ump->um_last_fullmsg, &ump->um_secs_fullmsg, 1)) { UFS_UNLOCK(ump); ffs_fserr(fs, pip->i_number, "out of inodes"); uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt); } else { UFS_UNLOCK(ump); } return (ENOSPC); } /* * Find a cylinder group to place a directory. * * The policy implemented by this algorithm is to allocate a * directory inode in the same cylinder group as its parent * directory, but also to reserve space for its files inodes * and data. Restrict the number of directories which may be * allocated one after another in the same cylinder group * without intervening allocation of files. * * If we allocate a first level directory then force allocation * in another cylinder group. */ static ino_t ffs_dirpref(struct inode *pip) { struct fs *fs; int cg, prefcg, curcg, dirsize, cgsize; int depth, range, start, end, numdirs, power, numerator, denominator; u_int avgifree, avgbfree, avgndir, curdirsize; u_int minifree, minbfree, maxndir; u_int maxcontigdirs; mtx_assert(UFS_MTX(ITOUMP(pip)), MA_OWNED); fs = ITOFS(pip); avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg; avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg; avgndir = fs->fs_cstotal.cs_ndir / fs->fs_ncg; /* * Select a preferred cylinder group to place a new directory. * If we are near the root of the filesystem we aim to spread * them out as much as possible. As we descend deeper from the * root we cluster them closer together around their parent as * we expect them to be more closely interactive. Higher-level * directories like usr/src/sys and usr/src/bin should be * separated while the directories in these areas are more * likely to be accessed together so should be closer. * * We pick a range of cylinder groups around the cylinder group * of the directory in which we are being created. The size of * the range for our search is based on our depth from the root * of our filesystem. We then probe that range based on how many * directories are already present. The first new directory is at * 1/2 (middle) of the range; the second is in the first 1/4 of the * range, then at 3/4, 1/8, 3/8, 5/8, 7/8, 1/16, 3/16, 5/16, etc. */ depth = DIP(pip, i_dirdepth); range = fs->fs_ncg / (1 << depth); curcg = ino_to_cg(fs, pip->i_number); start = curcg - (range / 2); if (start < 0) start += fs->fs_ncg; end = curcg + (range / 2); if (end >= fs->fs_ncg) end -= fs->fs_ncg; numdirs = pip->i_effnlink - 1; power = fls(numdirs); numerator = (numdirs & ~(1 << (power - 1))) * 2 + 1; denominator = 1 << power; prefcg = (curcg - (range / 2) + (range * numerator / denominator)); if (prefcg < 0) prefcg += fs->fs_ncg; if (prefcg >= fs->fs_ncg) prefcg -= fs->fs_ncg; /* * If this filesystem is not tracking directory depths, * revert to the old algorithm. */ if (depth == 0 && pip->i_number != UFS_ROOTINO) prefcg = curcg; /* * Count various limits which used for * optimal allocation of a directory inode. */ maxndir = min(avgndir + (1 << depth), fs->fs_ipg); minifree = avgifree - avgifree / 4; if (minifree < 1) minifree = 1; minbfree = avgbfree - avgbfree / 4; if (minbfree < 1) minbfree = 1; cgsize = fs->fs_fsize * fs->fs_fpg; dirsize = fs->fs_avgfilesize * fs->fs_avgfpdir; curdirsize = avgndir ? (cgsize - avgbfree * fs->fs_bsize) / avgndir : 0; if (dirsize < curdirsize) dirsize = curdirsize; if (dirsize <= 0) maxcontigdirs = 0; /* dirsize overflowed */ else maxcontigdirs = min((avgbfree * fs->fs_bsize) / dirsize, 255); if (fs->fs_avgfpdir > 0) maxcontigdirs = min(maxcontigdirs, fs->fs_ipg / fs->fs_avgfpdir); if (maxcontigdirs == 0) maxcontigdirs = 1; /* * Limit number of dirs in one cg and reserve space for * regular files, but only if we have no deficit in * inodes or space. * * We are trying to find a suitable cylinder group nearby * our preferred cylinder group to place a new directory. * We scan from our preferred cylinder group forward looking * for a cylinder group that meets our criterion. If we get * to the final cylinder group and do not find anything, * we start scanning forwards from the beginning of the * filesystem. While it might seem sensible to start scanning * backwards or even to alternate looking forward and backward, * this approach fails badly when the filesystem is nearly full. * Specifically, we first search all the areas that have no space * and finally try the one preceding that. We repeat this on * every request and in the case of the final block end up * searching the entire filesystem. By jumping to the front * of the filesystem, our future forward searches always look * in new cylinder groups so finds every possible block after * one pass over the filesystem. */ for (cg = prefcg; cg < fs->fs_ncg; cg++) if (fs->fs_cs(fs, cg).cs_ndir < maxndir && fs->fs_cs(fs, cg).cs_nifree >= minifree && fs->fs_cs(fs, cg).cs_nbfree >= minbfree) { if (fs->fs_contigdirs[cg] < maxcontigdirs) return ((ino_t)(fs->fs_ipg * cg)); } for (cg = 0; cg < prefcg; cg++) if (fs->fs_cs(fs, cg).cs_ndir < maxndir && fs->fs_cs(fs, cg).cs_nifree >= minifree && fs->fs_cs(fs, cg).cs_nbfree >= minbfree) { if (fs->fs_contigdirs[cg] < maxcontigdirs) return ((ino_t)(fs->fs_ipg * cg)); } /* * This is a backstop when we have deficit in space. */ for (cg = prefcg; cg < fs->fs_ncg; cg++) if (fs->fs_cs(fs, cg).cs_nifree >= avgifree) return ((ino_t)(fs->fs_ipg * cg)); for (cg = 0; cg < prefcg; cg++) if (fs->fs_cs(fs, cg).cs_nifree >= avgifree) break; return ((ino_t)(fs->fs_ipg * cg)); } /* * Select the desired position for the next block in a file. The file is * logically divided into sections. The first section is composed of the * direct blocks and the next fs_maxbpg blocks. Each additional section * contains fs_maxbpg blocks. * * If no blocks have been allocated in the first section, the policy is to * request a block in the same cylinder group as the inode that describes * the file. The first indirect is allocated immediately following the last * direct block and the data blocks for the first indirect immediately * follow it. * * If no blocks have been allocated in any other section, the indirect * block(s) are allocated in the same cylinder group as its inode in an * area reserved immediately following the inode blocks. The policy for * the data blocks is to place them in a cylinder group with a greater than * average number of free blocks. An appropriate cylinder group is found * by using a rotor that sweeps the cylinder groups. When a new group of * blocks is needed, the sweep begins in the cylinder group following the * cylinder group from which the previous allocation was made. The sweep * continues until a cylinder group with greater than the average number * of free blocks is found. If the allocation is for the first block in an * indirect block or the previous block is a hole, then the information on * the previous allocation is unavailable; here a best guess is made based * on the logical block number being allocated. * * If a section is already partially allocated, the policy is to * allocate blocks contiguously within the section if possible. */ ufs2_daddr_t ffs_blkpref_ufs1(struct inode *ip, ufs_lbn_t lbn, int indx, ufs1_daddr_t *bap) { struct fs *fs; u_int cg, inocg; u_int avgbfree, startcg; ufs2_daddr_t pref, prevbn; KASSERT(indx <= 0 || bap != NULL, ("need non-NULL bap")); mtx_assert(UFS_MTX(ITOUMP(ip)), MA_OWNED); fs = ITOFS(ip); /* * Allocation of indirect blocks is indicated by passing negative * values in indx: -1 for single indirect, -2 for double indirect, * -3 for triple indirect. As noted below, we attempt to allocate * the first indirect inline with the file data. For all later * indirect blocks, the data is often allocated in other cylinder * groups. However to speed random file access and to speed up * fsck, the filesystem reserves the first fs_metaspace blocks * (typically half of fs_minfree) of the data area of each cylinder * group to hold these later indirect blocks. */ inocg = ino_to_cg(fs, ip->i_number); if (indx < 0) { /* * Our preference for indirect blocks is the zone at the * beginning of the inode's cylinder group data area that * we try to reserve for indirect blocks. */ pref = cgmeta(fs, inocg); /* * If we are allocating the first indirect block, try to * place it immediately following the last direct block. */ if (indx == -1 && lbn < UFS_NDADDR + NINDIR(fs) && ip->i_din1->di_db[UFS_NDADDR - 1] != 0) pref = ip->i_din1->di_db[UFS_NDADDR - 1] + fs->fs_frag; return (pref); } /* * If we are allocating the first data block in the first indirect * block and the indirect has been allocated in the data block area, * try to place it immediately following the indirect block. */ if (lbn == UFS_NDADDR) { pref = ip->i_din1->di_ib[0]; if (pref != 0 && pref >= cgdata(fs, inocg) && pref < cgbase(fs, inocg + 1)) return (pref + fs->fs_frag); } /* * If we are at the beginning of a file, or we have already allocated * the maximum number of blocks per cylinder group, or we do not * have a block allocated immediately preceding us, then we need * to decide where to start allocating new blocks. */ if (indx == 0) { prevbn = 0; } else { prevbn = bap[indx - 1]; if (UFS_CHECK_BLKNO(ITOVFS(ip), ip->i_number, prevbn, fs->fs_bsize) != 0) prevbn = 0; } if (indx % fs->fs_maxbpg == 0 || prevbn == 0) { /* * If we are allocating a directory data block, we want * to place it in the metadata area. */ if ((ip->i_mode & IFMT) == IFDIR) return (cgmeta(fs, inocg)); /* * Until we fill all the direct and all the first indirect's * blocks, we try to allocate in the data area of the inode's * cylinder group. */ if (lbn < UFS_NDADDR + NINDIR(fs)) return (cgdata(fs, inocg)); /* * Find a cylinder with greater than average number of * unused data blocks. */ if (indx == 0 || prevbn == 0) startcg = inocg + lbn / fs->fs_maxbpg; else startcg = dtog(fs, prevbn) + 1; startcg %= fs->fs_ncg; avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg; for (cg = startcg; cg < fs->fs_ncg; cg++) if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { fs->fs_cgrotor = cg; return (cgdata(fs, cg)); } for (cg = 0; cg <= startcg; cg++) if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { fs->fs_cgrotor = cg; return (cgdata(fs, cg)); } return (0); } /* * Otherwise, we just always try to lay things out contiguously. */ return (prevbn + fs->fs_frag); } /* * Same as above, but for UFS2 */ ufs2_daddr_t ffs_blkpref_ufs2(struct inode *ip, ufs_lbn_t lbn, int indx, ufs2_daddr_t *bap) { struct fs *fs; u_int cg, inocg; u_int avgbfree, startcg; ufs2_daddr_t pref, prevbn; KASSERT(indx <= 0 || bap != NULL, ("need non-NULL bap")); mtx_assert(UFS_MTX(ITOUMP(ip)), MA_OWNED); fs = ITOFS(ip); /* * Allocation of indirect blocks is indicated by passing negative * values in indx: -1 for single indirect, -2 for double indirect, * -3 for triple indirect. As noted below, we attempt to allocate * the first indirect inline with the file data. For all later * indirect blocks, the data is often allocated in other cylinder * groups. However to speed random file access and to speed up * fsck, the filesystem reserves the first fs_metaspace blocks * (typically half of fs_minfree) of the data area of each cylinder * group to hold these later indirect blocks. */ inocg = ino_to_cg(fs, ip->i_number); if (indx < 0) { /* * Our preference for indirect blocks is the zone at the * beginning of the inode's cylinder group data area that * we try to reserve for indirect blocks. */ pref = cgmeta(fs, inocg); /* * If we are allocating the first indirect block, try to * place it immediately following the last direct block. */ if (indx == -1 && lbn < UFS_NDADDR + NINDIR(fs) && ip->i_din2->di_db[UFS_NDADDR - 1] != 0) pref = ip->i_din2->di_db[UFS_NDADDR - 1] + fs->fs_frag; return (pref); } /* * If we are allocating the first data block in the first indirect * block and the indirect has been allocated in the data block area, * try to place it immediately following the indirect block. */ if (lbn == UFS_NDADDR) { pref = ip->i_din2->di_ib[0]; if (pref != 0 && pref >= cgdata(fs, inocg) && pref < cgbase(fs, inocg + 1)) return (pref + fs->fs_frag); } /* * If we are at the beginning of a file, or we have already allocated * the maximum number of blocks per cylinder group, or we do not * have a block allocated immediately preceding us, then we need * to decide where to start allocating new blocks. */ if (indx == 0) { prevbn = 0; } else { prevbn = bap[indx - 1]; if (UFS_CHECK_BLKNO(ITOVFS(ip), ip->i_number, prevbn, fs->fs_bsize) != 0) prevbn = 0; } if (indx % fs->fs_maxbpg == 0 || prevbn == 0) { /* * If we are allocating a directory data block, we want * to place it in the metadata area. */ if ((ip->i_mode & IFMT) == IFDIR) return (cgmeta(fs, inocg)); /* * Until we fill all the direct and all the first indirect's * blocks, we try to allocate in the data area of the inode's * cylinder group. */ if (lbn < UFS_NDADDR + NINDIR(fs)) return (cgdata(fs, inocg)); /* * Find a cylinder with greater than average number of * unused data blocks. */ if (indx == 0 || prevbn == 0) startcg = inocg + lbn / fs->fs_maxbpg; else startcg = dtog(fs, prevbn) + 1; startcg %= fs->fs_ncg; avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg; for (cg = startcg; cg < fs->fs_ncg; cg++) if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { fs->fs_cgrotor = cg; return (cgdata(fs, cg)); } for (cg = 0; cg <= startcg; cg++) if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { fs->fs_cgrotor = cg; return (cgdata(fs, cg)); } return (0); } /* * Otherwise, we just always try to lay things out contiguously. */ return (prevbn + fs->fs_frag); } /* * Implement the cylinder overflow algorithm. * * The policy implemented by this algorithm is: * 1) allocate the block in its requested cylinder group. * 2) quadratically rehash on the cylinder group number. * 3) brute force search for a free block. * * Must be called with the UFS lock held. Will release the lock on success * and return with it held on failure. */ /*VARARGS5*/ static ufs2_daddr_t ffs_hashalloc(struct inode *ip, u_int cg, ufs2_daddr_t pref, int size, /* Search size for data blocks, mode for inodes */ int rsize, /* Real allocated size. */ allocfcn_t *allocator) { struct fs *fs; ufs2_daddr_t result; u_int i, icg = cg; mtx_assert(UFS_MTX(ITOUMP(ip)), MA_OWNED); #ifdef INVARIANTS if (ITOV(ip)->v_mount->mnt_kern_flag & MNTK_SUSPENDED) panic("ffs_hashalloc: allocation on suspended filesystem"); #endif fs = ITOFS(ip); /* * 1: preferred cylinder group */ result = (*allocator)(ip, cg, pref, size, rsize); if (result) return (result); /* * 2: quadratic rehash */ for (i = 1; i < fs->fs_ncg; i *= 2) { cg += i; if (cg >= fs->fs_ncg) cg -= fs->fs_ncg; result = (*allocator)(ip, cg, 0, size, rsize); if (result) return (result); } /* * 3: brute force search * Note that we start at i == 2, since 0 was checked initially, * and 1 is always checked in the quadratic rehash. */ cg = (icg + 2) % fs->fs_ncg; for (i = 2; i < fs->fs_ncg; i++) { result = (*allocator)(ip, cg, 0, size, rsize); if (result) return (result); cg++; if (cg == fs->fs_ncg) cg = 0; } return (0); } /* * Determine whether a fragment can be extended. * * Check to see if the necessary fragments are available, and * if they are, allocate them. */ static ufs2_daddr_t ffs_fragextend(struct inode *ip, u_int cg, ufs2_daddr_t bprev, int osize, int nsize) { struct fs *fs; struct cg *cgp; struct buf *bp; struct ufsmount *ump; int nffree; long bno; int frags, bbase; int i, error; u_int8_t *blksfree; ump = ITOUMP(ip); fs = ump->um_fs; if (fs->fs_cs(fs, cg).cs_nffree < numfrags(fs, nsize - osize)) return (0); frags = numfrags(fs, nsize); bbase = fragnum(fs, bprev); if (bbase > fragnum(fs, (bprev + frags - 1))) { /* cannot extend across a block boundary */ return (0); } UFS_UNLOCK(ump); if ((error = ffs_getcg(fs, ump->um_devvp, cg, 0, &bp, &cgp)) != 0) goto fail; bno = dtogd(fs, bprev); blksfree = cg_blksfree(cgp); for (i = numfrags(fs, osize); i < frags; i++) if (isclr(blksfree, bno + i)) goto fail; /* * the current fragment can be extended * deduct the count on fragment being extended into * increase the count on the remaining fragment (if any) * allocate the extended piece */ for (i = frags; i < fs->fs_frag - bbase; i++) if (isclr(blksfree, bno + i)) break; cgp->cg_frsum[i - numfrags(fs, osize)]--; if (i != frags) cgp->cg_frsum[i - frags]++; for (i = numfrags(fs, osize), nffree = 0; i < frags; i++) { clrbit(blksfree, bno + i); cgp->cg_cs.cs_nffree--; nffree++; } UFS_LOCK(ump); fs->fs_cstotal.cs_nffree -= nffree; fs->fs_cs(fs, cg).cs_nffree -= nffree; fs->fs_fmod = 1; ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); if (DOINGSOFTDEP(ITOV(ip))) softdep_setup_blkmapdep(bp, UFSTOVFS(ump), bprev, frags, numfrags(fs, osize)); bdwrite(bp); return (bprev); fail: brelse(bp); UFS_LOCK(ump); return (0); } /* * Determine whether a block can be allocated. * * Check to see if a block of the appropriate size is available, * and if it is, allocate it. */ static ufs2_daddr_t ffs_alloccg(struct inode *ip, u_int cg, ufs2_daddr_t bpref, int size, int rsize) { struct fs *fs; struct cg *cgp; struct buf *bp; struct ufsmount *ump; ufs1_daddr_t bno; ufs2_daddr_t blkno; int i, allocsiz, error, frags; u_int8_t *blksfree; ump = ITOUMP(ip); fs = ump->um_fs; if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize) return (0); UFS_UNLOCK(ump); if ((error = ffs_getcg(fs, ump->um_devvp, cg, 0, &bp, &cgp)) != 0 || (cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize)) goto fail; if (size == fs->fs_bsize) { UFS_LOCK(ump); blkno = ffs_alloccgblk(ip, bp, bpref, rsize); ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); bdwrite(bp); return (blkno); } /* * check to see if any fragments are already available * allocsiz is the size which will be allocated, hacking * it down to a smaller size if necessary */ blksfree = cg_blksfree(cgp); frags = numfrags(fs, size); for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++) if (cgp->cg_frsum[allocsiz] != 0) break; if (allocsiz == fs->fs_frag) { /* * no fragments were available, so a block will be * allocated, and hacked up */ if (cgp->cg_cs.cs_nbfree == 0) goto fail; UFS_LOCK(ump); blkno = ffs_alloccgblk(ip, bp, bpref, rsize); ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); bdwrite(bp); return (blkno); } KASSERT(size == rsize, ("ffs_alloccg: size(%d) != rsize(%d)", size, rsize)); bno = ffs_mapsearch(fs, cgp, bpref, allocsiz); if (bno < 0) goto fail; for (i = 0; i < frags; i++) clrbit(blksfree, bno + i); cgp->cg_cs.cs_nffree -= frags; cgp->cg_frsum[allocsiz]--; if (frags != allocsiz) cgp->cg_frsum[allocsiz - frags]++; UFS_LOCK(ump); fs->fs_cstotal.cs_nffree -= frags; fs->fs_cs(fs, cg).cs_nffree -= frags; fs->fs_fmod = 1; blkno = cgbase(fs, cg) + bno; ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); if (DOINGSOFTDEP(ITOV(ip))) softdep_setup_blkmapdep(bp, UFSTOVFS(ump), blkno, frags, 0); bdwrite(bp); return (blkno); fail: brelse(bp); UFS_LOCK(ump); return (0); } /* * Allocate a block in a cylinder group. * * This algorithm implements the following policy: * 1) allocate the requested block. * 2) allocate a rotationally optimal block in the same cylinder. * 3) allocate the next available block on the block rotor for the * specified cylinder group. * Note that this routine only allocates fs_bsize blocks; these * blocks may be fragmented by the routine that allocates them. */ static ufs2_daddr_t ffs_alloccgblk(struct inode *ip, struct buf *bp, ufs2_daddr_t bpref, int size) { struct fs *fs; struct cg *cgp; struct ufsmount *ump; ufs1_daddr_t bno; ufs2_daddr_t blkno; u_int8_t *blksfree; int i, cgbpref; ump = ITOUMP(ip); fs = ump->um_fs; mtx_assert(UFS_MTX(ump), MA_OWNED); cgp = (struct cg *)bp->b_data; blksfree = cg_blksfree(cgp); if (bpref == 0) { bpref = cgbase(fs, cgp->cg_cgx) + cgp->cg_rotor + fs->fs_frag; } else if ((cgbpref = dtog(fs, bpref)) != cgp->cg_cgx) { /* map bpref to correct zone in this cg */ if (bpref < cgdata(fs, cgbpref)) bpref = cgmeta(fs, cgp->cg_cgx); else bpref = cgdata(fs, cgp->cg_cgx); } /* * if the requested block is available, use it */ bno = dtogd(fs, blknum(fs, bpref)); if (ffs_isblock(fs, blksfree, fragstoblks(fs, bno))) goto gotit; /* * Take the next available block in this cylinder group. */ bno = ffs_mapsearch(fs, cgp, bpref, (int)fs->fs_frag); if (bno < 0) return (0); /* Update cg_rotor only if allocated from the data zone */ if (bno >= dtogd(fs, cgdata(fs, cgp->cg_cgx))) cgp->cg_rotor = bno; gotit: blkno = fragstoblks(fs, bno); ffs_clrblock(fs, blksfree, (long)blkno); ffs_clusteracct(fs, cgp, blkno, -1); cgp->cg_cs.cs_nbfree--; fs->fs_cstotal.cs_nbfree--; fs->fs_cs(fs, cgp->cg_cgx).cs_nbfree--; fs->fs_fmod = 1; blkno = cgbase(fs, cgp->cg_cgx) + bno; /* * If the caller didn't want the whole block free the frags here. */ size = numfrags(fs, size); if (size != fs->fs_frag) { bno = dtogd(fs, blkno); for (i = size; i < fs->fs_frag; i++) setbit(blksfree, bno + i); i = fs->fs_frag - size; cgp->cg_cs.cs_nffree += i; fs->fs_cstotal.cs_nffree += i; fs->fs_cs(fs, cgp->cg_cgx).cs_nffree += i; fs->fs_fmod = 1; cgp->cg_frsum[i]++; } /* XXX Fixme. */ UFS_UNLOCK(ump); if (DOINGSOFTDEP(ITOV(ip))) softdep_setup_blkmapdep(bp, UFSTOVFS(ump), blkno, size, 0); UFS_LOCK(ump); return (blkno); } /* * Determine whether a cluster can be allocated. * * We do not currently check for optimal rotational layout if there * are multiple choices in the same cylinder group. Instead we just * take the first one that we find following bpref. */ static ufs2_daddr_t ffs_clusteralloc(struct inode *ip, u_int cg, ufs2_daddr_t bpref, int len) { struct fs *fs; struct cg *cgp; struct buf *bp; struct ufsmount *ump; int i, run, bit, map, got, error; ufs2_daddr_t bno; u_char *mapp; int32_t *lp; u_int8_t *blksfree; ump = ITOUMP(ip); fs = ump->um_fs; if (fs->fs_maxcluster[cg] < len) return (0); UFS_UNLOCK(ump); if ((error = ffs_getcg(fs, ump->um_devvp, cg, 0, &bp, &cgp)) != 0) { UFS_LOCK(ump); return (0); } /* * Check to see if a cluster of the needed size (or bigger) is * available in this cylinder group. */ lp = &cg_clustersum(cgp)[len]; for (i = len; i <= fs->fs_contigsumsize; i++) if (*lp++ > 0) break; if (i > fs->fs_contigsumsize) { /* * This is the first time looking for a cluster in this * cylinder group. Update the cluster summary information * to reflect the true maximum sized cluster so that * future cluster allocation requests can avoid reading * the cylinder group map only to find no clusters. */ lp = &cg_clustersum(cgp)[len - 1]; for (i = len - 1; i > 0; i--) if (*lp-- > 0) break; UFS_LOCK(ump); fs->fs_maxcluster[cg] = i; brelse(bp); return (0); } /* * Search the cluster map to find a big enough cluster. * We take the first one that we find, even if it is larger * than we need as we prefer to get one close to the previous * block allocation. We do not search before the current * preference point as we do not want to allocate a block * that is allocated before the previous one (as we will * then have to wait for another pass of the elevator * algorithm before it will be read). We prefer to fail and * be recalled to try an allocation in the next cylinder group. */ if (dtog(fs, bpref) != cg) bpref = cgdata(fs, cg); else bpref = blknum(fs, bpref); bpref = fragstoblks(fs, dtogd(fs, bpref)); mapp = &cg_clustersfree(cgp)[bpref / NBBY]; map = *mapp++; bit = 1 << (bpref % NBBY); for (run = 0, got = bpref; got < cgp->cg_nclusterblks; got++) { if ((map & bit) == 0) { run = 0; } else { run++; if (run == len) break; } if ((got & (NBBY - 1)) != (NBBY - 1)) { bit <<= 1; } else { map = *mapp++; bit = 1; } } if (got >= cgp->cg_nclusterblks) { UFS_LOCK(ump); brelse(bp); return (0); } /* * Allocate the cluster that we have found. */ blksfree = cg_blksfree(cgp); for (i = 1; i <= len; i++) if (!ffs_isblock(fs, blksfree, got - run + i)) panic("ffs_clusteralloc: map mismatch"); bno = cgbase(fs, cg) + blkstofrags(fs, got - run + 1); if (dtog(fs, bno) != cg) panic("ffs_clusteralloc: allocated out of group"); len = blkstofrags(fs, len); UFS_LOCK(ump); for (i = 0; i < len; i += fs->fs_frag) if (ffs_alloccgblk(ip, bp, bno + i, fs->fs_bsize) != bno + i) panic("ffs_clusteralloc: lost block"); ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); bdwrite(bp); return (bno); } static inline struct buf * getinobuf(struct inode *ip, u_int cg, u_int32_t cginoblk, int gbflags) { struct fs *fs; fs = ITOFS(ip); return (getblk(ITODEVVP(ip), fsbtodb(fs, ino_to_fsba(fs, cg * fs->fs_ipg + cginoblk)), (int)fs->fs_bsize, 0, 0, gbflags)); } /* * Synchronous inode initialization is needed only when barrier writes do not * work as advertised, and will impose a heavy cost on file creation in a newly * created filesystem. */ static int doasyncinodeinit = 1; SYSCTL_INT(_vfs_ffs, OID_AUTO, doasyncinodeinit, CTLFLAG_RWTUN, &doasyncinodeinit, 0, "Perform inode block initialization using asynchronous writes"); /* * Determine whether an inode can be allocated. * * Check to see if an inode is available, and if it is, * allocate it using the following policy: * 1) allocate the requested inode. * 2) allocate the next available inode after the requested * inode in the specified cylinder group. */ static ufs2_daddr_t ffs_nodealloccg(struct inode *ip, u_int cg, ufs2_daddr_t ipref, int mode, int unused) { struct fs *fs; struct cg *cgp; struct buf *bp, *ibp; struct ufsmount *ump; u_int8_t *inosused, *loc; struct ufs2_dinode *dp2; int error, start, len, i; u_int32_t old_initediblk; ump = ITOUMP(ip); fs = ump->um_fs; check_nifree: if (fs->fs_cs(fs, cg).cs_nifree == 0) return (0); UFS_UNLOCK(ump); if ((error = ffs_getcg(fs, ump->um_devvp, cg, 0, &bp, &cgp)) != 0) { UFS_LOCK(ump); return (0); } restart: if (cgp->cg_cs.cs_nifree == 0) { brelse(bp); UFS_LOCK(ump); return (0); } inosused = cg_inosused(cgp); if (ipref) { ipref %= fs->fs_ipg; if (isclr(inosused, ipref)) goto gotit; } start = cgp->cg_irotor / NBBY; len = howmany(fs->fs_ipg - cgp->cg_irotor, NBBY); loc = memcchr(&inosused[start], 0xff, len); if (loc == NULL) { len = start + 1; start = 0; loc = memcchr(&inosused[start], 0xff, len); if (loc == NULL) { printf("cg = %d, irotor = %ld, fs = %s\n", cg, (long)cgp->cg_irotor, fs->fs_fsmnt); panic("ffs_nodealloccg: map corrupted"); /* NOTREACHED */ } } ipref = (loc - inosused) * NBBY + ffs(~*loc) - 1; gotit: /* * Check to see if we need to initialize more inodes. */ if (fs->fs_magic == FS_UFS2_MAGIC && ipref + INOPB(fs) > cgp->cg_initediblk && cgp->cg_initediblk < cgp->cg_niblk) { old_initediblk = cgp->cg_initediblk; /* * Free the cylinder group lock before writing the * initialized inode block. Entering the * babarrierwrite() with the cylinder group lock * causes lock order violation between the lock and * snaplk. * * Another thread can decide to initialize the same * inode block, but whichever thread first gets the * cylinder group lock after writing the newly * allocated inode block will update it and the other * will realize that it has lost and leave the * cylinder group unchanged. */ ibp = getinobuf(ip, cg, old_initediblk, GB_LOCK_NOWAIT); brelse(bp); if (ibp == NULL) { /* * The inode block buffer is already owned by * another thread, which must initialize it. * Wait on the buffer to allow another thread * to finish the updates, with dropped cg * buffer lock, then retry. */ ibp = getinobuf(ip, cg, old_initediblk, 0); brelse(ibp); UFS_LOCK(ump); goto check_nifree; } bzero(ibp->b_data, (int)fs->fs_bsize); dp2 = (struct ufs2_dinode *)(ibp->b_data); for (i = 0; i < INOPB(fs); i++) { while (dp2->di_gen == 0) dp2->di_gen = arc4random(); dp2++; } /* * Rather than adding a soft updates dependency to ensure * that the new inode block is written before it is claimed * by the cylinder group map, we just do a barrier write * here. The barrier write will ensure that the inode block * gets written before the updated cylinder group map can be * written. The barrier write should only slow down bulk * loading of newly created filesystems. */ if (doasyncinodeinit) babarrierwrite(ibp); else bwrite(ibp); /* * After the inode block is written, try to update the * cg initediblk pointer. If another thread beat us * to it, then leave it unchanged as the other thread * has already set it correctly. */ error = ffs_getcg(fs, ump->um_devvp, cg, 0, &bp, &cgp); UFS_LOCK(ump); ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); if (error != 0) return (error); if (cgp->cg_initediblk == old_initediblk) cgp->cg_initediblk += INOPB(fs); goto restart; } cgp->cg_irotor = ipref; UFS_LOCK(ump); ACTIVECLEAR(fs, cg); setbit(inosused, ipref); cgp->cg_cs.cs_nifree--; fs->fs_cstotal.cs_nifree--; fs->fs_cs(fs, cg).cs_nifree--; fs->fs_fmod = 1; if ((mode & IFMT) == IFDIR) { cgp->cg_cs.cs_ndir++; fs->fs_cstotal.cs_ndir++; fs->fs_cs(fs, cg).cs_ndir++; } UFS_UNLOCK(ump); if (DOINGSOFTDEP(ITOV(ip))) softdep_setup_inomapdep(bp, ip, cg * fs->fs_ipg + ipref, mode); bdwrite(bp); return ((ino_t)(cg * fs->fs_ipg + ipref)); } /* * Free a block or fragment. * * The specified block or fragment is placed back in the * free map. If a fragment is deallocated, a possible * block reassembly is checked. */ static void ffs_blkfree_cg(struct ufsmount *ump, struct fs *fs, struct vnode *devvp, ufs2_daddr_t bno, long size, ino_t inum, struct workhead *dephd) { struct mount *mp; struct cg *cgp; struct buf *bp; daddr_t dbn; ufs1_daddr_t fragno, cgbno; int i, blk, frags, bbase, error; u_int cg; u_int8_t *blksfree; struct cdev *dev; cg = dtog(fs, bno); if (devvp->v_type == VREG) { /* devvp is a snapshot */ MPASS(devvp->v_mount->mnt_data == ump); dev = ump->um_devvp->v_rdev; } else if (devvp->v_type == VCHR) { /* * devvp is a normal disk device * XXXKIB: devvp is not locked there, v_rdev access depends on * busy mount, which prevents mntfs devvp from reclamation. */ dev = devvp->v_rdev; } else return; #ifdef INVARIANTS if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0 || fragnum(fs, bno) + numfrags(fs, size) > fs->fs_frag) { printf("dev=%s, bno = %jd, bsize = %ld, size = %ld, fs = %s\n", devtoname(dev), (intmax_t)bno, (long)fs->fs_bsize, size, fs->fs_fsmnt); panic("ffs_blkfree_cg: bad size"); } #endif if ((u_int)bno >= fs->fs_size) { printf("bad block %jd, ino %lu\n", (intmax_t)bno, (u_long)inum); ffs_fserr(fs, inum, "bad block"); return; } if ((error = ffs_getcg(fs, devvp, cg, GB_CVTENXIO, &bp, &cgp)) != 0) { if (!ffs_fsfail_cleanup(ump, error) || !MOUNTEDSOFTDEP(UFSTOVFS(ump)) || devvp->v_type != VCHR) return; if (devvp->v_type == VREG) dbn = fragstoblks(fs, cgtod(fs, cg)); else dbn = fsbtodb(fs, cgtod(fs, cg)); error = getblkx(devvp, dbn, dbn, fs->fs_cgsize, 0, 0, 0, &bp); KASSERT(error == 0, ("getblkx failed")); softdep_setup_blkfree(UFSTOVFS(ump), bp, bno, numfrags(fs, size), dephd); bp->b_flags |= B_RELBUF | B_NOCACHE; bp->b_flags &= ~B_CACHE; bawrite(bp); return; } cgbno = dtogd(fs, bno); blksfree = cg_blksfree(cgp); UFS_LOCK(ump); if (size == fs->fs_bsize) { fragno = fragstoblks(fs, cgbno); if (!ffs_isfreeblock(fs, blksfree, fragno)) { if (devvp->v_type == VREG) { UFS_UNLOCK(ump); /* devvp is a snapshot */ brelse(bp); return; } printf("dev = %s, block = %jd, fs = %s\n", devtoname(dev), (intmax_t)bno, fs->fs_fsmnt); panic("ffs_blkfree_cg: freeing free block"); } ffs_setblock(fs, blksfree, fragno); ffs_clusteracct(fs, cgp, fragno, 1); cgp->cg_cs.cs_nbfree++; fs->fs_cstotal.cs_nbfree++; fs->fs_cs(fs, cg).cs_nbfree++; } else { bbase = cgbno - fragnum(fs, cgbno); /* * decrement the counts associated with the old frags */ blk = blkmap(fs, blksfree, bbase); ffs_fragacct(fs, blk, cgp->cg_frsum, -1); /* * deallocate the fragment */ frags = numfrags(fs, size); for (i = 0; i < frags; i++) { if (isset(blksfree, cgbno + i)) { printf("dev = %s, block = %jd, fs = %s\n", devtoname(dev), (intmax_t)(bno + i), fs->fs_fsmnt); panic("ffs_blkfree_cg: freeing free frag"); } setbit(blksfree, cgbno + i); } cgp->cg_cs.cs_nffree += i; fs->fs_cstotal.cs_nffree += i; fs->fs_cs(fs, cg).cs_nffree += i; /* * add back in counts associated with the new frags */ blk = blkmap(fs, blksfree, bbase); ffs_fragacct(fs, blk, cgp->cg_frsum, 1); /* * if a complete block has been reassembled, account for it */ fragno = fragstoblks(fs, bbase); if (ffs_isblock(fs, blksfree, fragno)) { cgp->cg_cs.cs_nffree -= fs->fs_frag; fs->fs_cstotal.cs_nffree -= fs->fs_frag; fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag; ffs_clusteracct(fs, cgp, fragno, 1); cgp->cg_cs.cs_nbfree++; fs->fs_cstotal.cs_nbfree++; fs->fs_cs(fs, cg).cs_nbfree++; } } fs->fs_fmod = 1; ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); mp = UFSTOVFS(ump); if (MOUNTEDSOFTDEP(mp) && devvp->v_type == VCHR) softdep_setup_blkfree(UFSTOVFS(ump), bp, bno, numfrags(fs, size), dephd); bdwrite(bp); } /* * Structures and routines associated with trim management. * * The following requests are passed to trim_lookup to indicate * the actions that should be taken. */ #define NEW 1 /* if found, error else allocate and hash it */ #define OLD 2 /* if not found, error, else return it */ #define REPLACE 3 /* if not found, error else unhash and reallocate it */ #define DONE 4 /* if not found, error else unhash and return it */ #define SINGLE 5 /* don't look up, just allocate it and don't hash it */ MALLOC_DEFINE(M_TRIM, "ufs_trim", "UFS trim structures"); #define TRIMLIST_HASH(ump, key) \ (&(ump)->um_trimhash[(key) & (ump)->um_trimlisthashsize]) /* * These structures describe each of the block free requests aggregated * together to make up a trim request. */ struct trim_blkreq { TAILQ_ENTRY(trim_blkreq) blkreqlist; ufs2_daddr_t bno; long size; struct workhead *pdephd; struct workhead dephd; }; /* * Description of a trim request. */ struct ffs_blkfree_trim_params { TAILQ_HEAD(, trim_blkreq) blklist; LIST_ENTRY(ffs_blkfree_trim_params) hashlist; struct task task; struct ufsmount *ump; struct vnode *devvp; ino_t inum; ufs2_daddr_t bno; long size; long key; }; static void ffs_blkfree_trim_completed(struct buf *); static void ffs_blkfree_trim_task(void *ctx, int pending __unused); static struct ffs_blkfree_trim_params *trim_lookup(struct ufsmount *, struct vnode *, ufs2_daddr_t, long, ino_t, u_long, int); static void ffs_blkfree_sendtrim(struct ffs_blkfree_trim_params *); /* * Called on trim completion to start a task to free the associated block(s). */ static void ffs_blkfree_trim_completed(struct buf *bp) { struct ffs_blkfree_trim_params *tp; tp = bp->b_fsprivate1; free(bp, M_TRIM); TASK_INIT(&tp->task, 0, ffs_blkfree_trim_task, tp); taskqueue_enqueue(tp->ump->um_trim_tq, &tp->task); } /* * Trim completion task that free associated block(s). */ static void ffs_blkfree_trim_task(void *ctx, int pending) { struct ffs_blkfree_trim_params *tp; struct trim_blkreq *blkelm; struct ufsmount *ump; tp = ctx; ump = tp->ump; while ((blkelm = TAILQ_FIRST(&tp->blklist)) != NULL) { ffs_blkfree_cg(ump, ump->um_fs, tp->devvp, blkelm->bno, blkelm->size, tp->inum, blkelm->pdephd); TAILQ_REMOVE(&tp->blklist, blkelm, blkreqlist); free(blkelm, M_TRIM); } vn_finished_secondary_write(UFSTOVFS(ump)); UFS_LOCK(ump); ump->um_trim_inflight -= 1; ump->um_trim_inflight_blks -= numfrags(ump->um_fs, tp->size); UFS_UNLOCK(ump); free(tp, M_TRIM); } /* * Lookup a trim request by inode number. * Allocate if requested (NEW, REPLACE, SINGLE). */ static struct ffs_blkfree_trim_params * trim_lookup(struct ufsmount *ump, struct vnode *devvp, ufs2_daddr_t bno, long size, ino_t inum, u_long key, int alloctype) { struct trimlist_hashhead *tphashhead; struct ffs_blkfree_trim_params *tp, *ntp; ntp = malloc(sizeof(struct ffs_blkfree_trim_params), M_TRIM, M_WAITOK); if (alloctype != SINGLE) { KASSERT(key >= FIRST_VALID_KEY, ("trim_lookup: invalid key")); UFS_LOCK(ump); tphashhead = TRIMLIST_HASH(ump, key); LIST_FOREACH(tp, tphashhead, hashlist) if (key == tp->key) break; } switch (alloctype) { case NEW: KASSERT(tp == NULL, ("trim_lookup: found trim")); break; case OLD: KASSERT(tp != NULL, ("trim_lookup: missing call to ffs_blkrelease_start()")); UFS_UNLOCK(ump); free(ntp, M_TRIM); return (tp); case REPLACE: KASSERT(tp != NULL, ("trim_lookup: missing REPLACE trim")); LIST_REMOVE(tp, hashlist); /* tp will be freed by caller */ break; case DONE: KASSERT(tp != NULL, ("trim_lookup: missing DONE trim")); LIST_REMOVE(tp, hashlist); UFS_UNLOCK(ump); free(ntp, M_TRIM); return (tp); } TAILQ_INIT(&ntp->blklist); ntp->ump = ump; ntp->devvp = devvp; ntp->bno = bno; ntp->size = size; ntp->inum = inum; ntp->key = key; if (alloctype != SINGLE) { LIST_INSERT_HEAD(tphashhead, ntp, hashlist); UFS_UNLOCK(ump); } return (ntp); } /* * Dispatch a trim request. */ static void ffs_blkfree_sendtrim(struct ffs_blkfree_trim_params *tp) { struct ufsmount *ump; struct mount *mp; struct buf *bp; /* * Postpone the set of the free bit in the cg bitmap until the * BIO_DELETE is completed. Otherwise, due to disk queue * reordering, TRIM might be issued after we reuse the block * and write some new data into it. */ ump = tp->ump; bp = malloc(sizeof(*bp), M_TRIM, M_WAITOK | M_ZERO); bp->b_iocmd = BIO_DELETE; bp->b_iooffset = dbtob(fsbtodb(ump->um_fs, tp->bno)); bp->b_iodone = ffs_blkfree_trim_completed; bp->b_bcount = tp->size; bp->b_fsprivate1 = tp; UFS_LOCK(ump); ump->um_trim_total += 1; ump->um_trim_inflight += 1; ump->um_trim_inflight_blks += numfrags(ump->um_fs, tp->size); ump->um_trim_total_blks += numfrags(ump->um_fs, tp->size); UFS_UNLOCK(ump); mp = UFSTOVFS(ump); vn_start_secondary_write(NULL, &mp, 0); g_vfs_strategy(ump->um_bo, bp); } /* * Allocate a new key to use to identify a range of blocks. */ u_long ffs_blkrelease_start(struct ufsmount *ump, struct vnode *devvp, ino_t inum) { static u_long masterkey; u_long key; if (((ump->um_flags & UM_CANDELETE) == 0) || dotrimcons == 0) return (SINGLETON_KEY); do { key = atomic_fetchadd_long(&masterkey, 1); } while (key < FIRST_VALID_KEY); (void) trim_lookup(ump, devvp, 0, 0, inum, key, NEW); return (key); } /* * Deallocate a key that has been used to identify a range of blocks. */ void ffs_blkrelease_finish(struct ufsmount *ump, u_long key) { struct ffs_blkfree_trim_params *tp; if (((ump->um_flags & UM_CANDELETE) == 0) || dotrimcons == 0) return; /* * If the vfs.ffs.dotrimcons sysctl option is enabled while * a file deletion is active, specifically after a call * to ffs_blkrelease_start() but before the call to * ffs_blkrelease_finish(), ffs_blkrelease_start() will * have handed out SINGLETON_KEY rather than starting a * collection sequence. Thus if we get a SINGLETON_KEY * passed to ffs_blkrelease_finish(), we just return rather * than trying to finish the nonexistent sequence. */ if (key == SINGLETON_KEY) { #ifdef INVARIANTS printf("%s: vfs.ffs.dotrimcons enabled on active filesystem\n", ump->um_mountp->mnt_stat.f_mntonname); #endif return; } /* * We are done with sending blocks using this key. Look up the key * using the DONE alloctype (in tp) to request that it be unhashed * as we will not be adding to it. If the key has never been used, * tp->size will be zero, so we can just free tp. Otherwise the call * to ffs_blkfree_sendtrim(tp) causes the block range described by * tp to be issued (and then tp to be freed). */ tp = trim_lookup(ump, NULL, 0, 0, 0, key, DONE); if (tp->size == 0) free(tp, M_TRIM); else ffs_blkfree_sendtrim(tp); } /* * Setup to free a block or fragment. * * Check for snapshots that might want to claim the block. * If trims are requested, prepare a trim request. Attempt to * aggregate consecutive blocks into a single trim request. */ void ffs_blkfree(struct ufsmount *ump, struct fs *fs, struct vnode *devvp, ufs2_daddr_t bno, long size, ino_t inum, enum vtype vtype, struct workhead *dephd, u_long key) { struct ffs_blkfree_trim_params *tp, *ntp; struct trim_blkreq *blkelm; /* * Check to see if a snapshot wants to claim the block. * Check that devvp is a normal disk device, not a snapshot, * it has a snapshot(s) associated with it, and one of the * snapshots wants to claim the block. */ if (devvp->v_type == VCHR && (devvp->v_vflag & VV_COPYONWRITE) && ffs_snapblkfree(fs, devvp, bno, size, inum, vtype, dephd)) { return; } /* * Nothing to delay if TRIM is not required for this block or TRIM * is disabled or the operation is performed on a snapshot. */ if (key == NOTRIM_KEY || ((ump->um_flags & UM_CANDELETE) == 0) || devvp->v_type == VREG) { ffs_blkfree_cg(ump, fs, devvp, bno, size, inum, dephd); return; } blkelm = malloc(sizeof(struct trim_blkreq), M_TRIM, M_WAITOK); blkelm->bno = bno; blkelm->size = size; if (dephd == NULL) { blkelm->pdephd = NULL; } else { LIST_INIT(&blkelm->dephd); LIST_SWAP(dephd, &blkelm->dephd, worklist, wk_list); blkelm->pdephd = &blkelm->dephd; } if (key == SINGLETON_KEY) { /* * Just a single non-contiguous piece. Use the SINGLE * alloctype to return a trim request that will not be * hashed for future lookup. */ tp = trim_lookup(ump, devvp, bno, size, inum, key, SINGLE); TAILQ_INSERT_HEAD(&tp->blklist, blkelm, blkreqlist); ffs_blkfree_sendtrim(tp); return; } /* * The callers of this function are not tracking whether or not * the blocks are contiguous. They are just saying that they * are freeing a set of blocks. It is this code that determines * the pieces of that range that are actually contiguous. * * Calling ffs_blkrelease_start() will have created an entry * that we will use. */ tp = trim_lookup(ump, devvp, bno, size, inum, key, OLD); if (tp->size == 0) { /* * First block of a potential range, set block and size * for the trim block. */ tp->bno = bno; tp->size = size; TAILQ_INSERT_HEAD(&tp->blklist, blkelm, blkreqlist); return; } /* * If this block is a continuation of the range (either * follows at the end or preceeds in the front) then we * add it to the front or back of the list and return. * * If it is not a continuation of the trim that we were * building, using the REPLACE alloctype, we request that * the old trim request (still in tp) be unhashed and a * new range started (in ntp). The ffs_blkfree_sendtrim(tp) * call causes the block range described by tp to be issued * (and then tp to be freed). */ if (bno + numfrags(fs, size) == tp->bno) { TAILQ_INSERT_HEAD(&tp->blklist, blkelm, blkreqlist); tp->bno = bno; tp->size += size; return; } else if (bno == tp->bno + numfrags(fs, tp->size)) { TAILQ_INSERT_TAIL(&tp->blklist, blkelm, blkreqlist); tp->size += size; return; } ntp = trim_lookup(ump, devvp, bno, size, inum, key, REPLACE); TAILQ_INSERT_HEAD(&ntp->blklist, blkelm, blkreqlist); ffs_blkfree_sendtrim(tp); } #ifdef INVARIANTS /* * Verify allocation of a block or fragment. Returns true if block or * fragment is allocated, false if it is free. */ static int ffs_checkblk(struct inode *ip, ufs2_daddr_t bno, long size) { struct fs *fs; struct cg *cgp; struct buf *bp; ufs1_daddr_t cgbno; int i, error, frags, free; u_int8_t *blksfree; fs = ITOFS(ip); if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) { printf("bsize = %ld, size = %ld, fs = %s\n", (long)fs->fs_bsize, size, fs->fs_fsmnt); panic("ffs_checkblk: bad size"); } if ((u_int)bno >= fs->fs_size) panic("ffs_checkblk: bad block %jd", (intmax_t)bno); error = ffs_getcg(fs, ITODEVVP(ip), dtog(fs, bno), 0, &bp, &cgp); if (error) panic("ffs_checkblk: cylinder group read failed"); blksfree = cg_blksfree(cgp); cgbno = dtogd(fs, bno); if (size == fs->fs_bsize) { free = ffs_isblock(fs, blksfree, fragstoblks(fs, cgbno)); } else { frags = numfrags(fs, size); for (free = 0, i = 0; i < frags; i++) if (isset(blksfree, cgbno + i)) free++; if (free != 0 && free != frags) panic("ffs_checkblk: partially free fragment"); } brelse(bp); return (!free); } #endif /* INVARIANTS */ /* * Free an inode. */ int ffs_vfree(struct vnode *pvp, ino_t ino, int mode) { struct ufsmount *ump; if (DOINGSOFTDEP(pvp)) { softdep_freefile(pvp, ino, mode); return (0); } ump = VFSTOUFS(pvp->v_mount); return (ffs_freefile(ump, ump->um_fs, ump->um_devvp, ino, mode, NULL)); } /* * Do the actual free operation. * The specified inode is placed back in the free map. */ int ffs_freefile(struct ufsmount *ump, struct fs *fs, struct vnode *devvp, ino_t ino, int mode, struct workhead *wkhd) { struct cg *cgp; struct buf *bp; daddr_t dbn; int error; u_int cg; u_int8_t *inosused; struct cdev *dev; ino_t cgino; cg = ino_to_cg(fs, ino); if (devvp->v_type == VREG) { /* devvp is a snapshot */ MPASS(devvp->v_mount->mnt_data == ump); dev = ump->um_devvp->v_rdev; } else if (devvp->v_type == VCHR) { /* devvp is a normal disk device */ dev = devvp->v_rdev; } else { bp = NULL; return (0); } if (ino >= fs->fs_ipg * fs->fs_ncg) panic("ffs_freefile: range: dev = %s, ino = %ju, fs = %s", devtoname(dev), (uintmax_t)ino, fs->fs_fsmnt); if ((error = ffs_getcg(fs, devvp, cg, GB_CVTENXIO, &bp, &cgp)) != 0) { if (!ffs_fsfail_cleanup(ump, error) || !MOUNTEDSOFTDEP(UFSTOVFS(ump)) || devvp->v_type != VCHR) return (error); if (devvp->v_type == VREG) dbn = fragstoblks(fs, cgtod(fs, cg)); else dbn = fsbtodb(fs, cgtod(fs, cg)); error = getblkx(devvp, dbn, dbn, fs->fs_cgsize, 0, 0, 0, &bp); KASSERT(error == 0, ("getblkx failed")); softdep_setup_inofree(UFSTOVFS(ump), bp, ino, wkhd); bp->b_flags |= B_RELBUF | B_NOCACHE; bp->b_flags &= ~B_CACHE; bawrite(bp); return (error); } inosused = cg_inosused(cgp); cgino = ino % fs->fs_ipg; if (isclr(inosused, cgino)) { printf("dev = %s, ino = %ju, fs = %s\n", devtoname(dev), (uintmax_t)ino, fs->fs_fsmnt); if (fs->fs_ronly == 0) panic("ffs_freefile: freeing free inode"); } clrbit(inosused, cgino); if (cgino < cgp->cg_irotor) cgp->cg_irotor = cgino; cgp->cg_cs.cs_nifree++; UFS_LOCK(ump); fs->fs_cstotal.cs_nifree++; fs->fs_cs(fs, cg).cs_nifree++; if ((mode & IFMT) == IFDIR) { cgp->cg_cs.cs_ndir--; fs->fs_cstotal.cs_ndir--; fs->fs_cs(fs, cg).cs_ndir--; } fs->fs_fmod = 1; ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); if (MOUNTEDSOFTDEP(UFSTOVFS(ump)) && devvp->v_type == VCHR) softdep_setup_inofree(UFSTOVFS(ump), bp, ino, wkhd); bdwrite(bp); return (0); } /* * Check to see if a file is free. * Used to check for allocated files in snapshots. */ int ffs_checkfreefile(struct fs *fs, struct vnode *devvp, ino_t ino) { struct cg *cgp; struct buf *bp; int ret, error; u_int cg; u_int8_t *inosused; cg = ino_to_cg(fs, ino); if ((devvp->v_type != VREG) && (devvp->v_type != VCHR)) return (1); if (ino >= fs->fs_ipg * fs->fs_ncg) return (1); if ((error = ffs_getcg(fs, devvp, cg, 0, &bp, &cgp)) != 0) return (1); inosused = cg_inosused(cgp); ino %= fs->fs_ipg; ret = isclr(inosused, ino); brelse(bp); return (ret); } /* * Find a block of the specified size in the specified cylinder group. * * It is a panic if a request is made to find a block if none are * available. */ static ufs1_daddr_t ffs_mapsearch(struct fs *fs, struct cg *cgp, ufs2_daddr_t bpref, int allocsiz) { ufs1_daddr_t bno; int start, len, loc, i; int blk, field, subfield, pos; u_int8_t *blksfree; /* * find the fragment by searching through the free block * map for an appropriate bit pattern */ if (bpref) start = dtogd(fs, bpref) / NBBY; else start = cgp->cg_frotor / NBBY; blksfree = cg_blksfree(cgp); len = howmany(fs->fs_fpg, NBBY) - start; loc = scanc((u_int)len, (u_char *)&blksfree[start], fragtbl[fs->fs_frag], (u_char)(1 << (allocsiz - 1 + (fs->fs_frag % NBBY)))); if (loc == 0) { len = start + 1; start = 0; loc = scanc((u_int)len, (u_char *)&blksfree[0], fragtbl[fs->fs_frag], (u_char)(1 << (allocsiz - 1 + (fs->fs_frag % NBBY)))); if (loc == 0) { printf("start = %d, len = %d, fs = %s\n", start, len, fs->fs_fsmnt); panic("ffs_alloccg: map corrupted"); /* NOTREACHED */ } } bno = (start + len - loc) * NBBY; cgp->cg_frotor = bno; /* * found the byte in the map * sift through the bits to find the selected frag */ for (i = bno + NBBY; bno < i; bno += fs->fs_frag) { blk = blkmap(fs, blksfree, bno); blk <<= 1; field = around[allocsiz]; subfield = inside[allocsiz]; for (pos = 0; pos <= fs->fs_frag - allocsiz; pos++) { if ((blk & field) == subfield) return (bno + pos); field <<= 1; subfield <<= 1; } } printf("bno = %lu, fs = %s\n", (u_long)bno, fs->fs_fsmnt); panic("ffs_alloccg: block not in map"); return (-1); } static const struct statfs * ffs_getmntstat(struct vnode *devvp) { if (devvp->v_type == VCHR) return (&devvp->v_rdev->si_mountpt->mnt_stat); return (ffs_getmntstat(VFSTOUFS(devvp->v_mount)->um_devvp)); } /* * Fetch and verify a cylinder group. */ int ffs_getcg(struct fs *fs, struct vnode *devvp, u_int cg, int flags, struct buf **bpp, struct cg **cgpp) { struct buf *bp; struct cg *cgp; const struct statfs *sfs; daddr_t blkno; int error; *bpp = NULL; *cgpp = NULL; if ((fs->fs_metackhash & CK_CYLGRP) != 0) flags |= GB_CKHASH; if (devvp->v_type == VREG) blkno = fragstoblks(fs, cgtod(fs, cg)); else blkno = fsbtodb(fs, cgtod(fs, cg)); error = breadn_flags(devvp, blkno, blkno, (int)fs->fs_cgsize, NULL, NULL, 0, NOCRED, flags, ffs_ckhash_cg, &bp); if (error != 0) return (error); cgp = (struct cg *)bp->b_data; if ((fs->fs_metackhash & CK_CYLGRP) != 0 && (bp->b_flags & B_CKHASH) != 0 && cgp->cg_ckhash != bp->b_ckhash) { sfs = ffs_getmntstat(devvp); printf("UFS %s%s (%s) cylinder checksum failed: cg %u, cgp: " "0x%x != bp: 0x%jx\n", devvp->v_type == VCHR ? "" : "snapshot of ", sfs->f_mntfromname, sfs->f_mntonname, cg, cgp->cg_ckhash, (uintmax_t)bp->b_ckhash); bp->b_flags &= ~B_CKHASH; bp->b_flags |= B_INVAL | B_NOCACHE; brelse(bp); return (EIO); } if (!cg_chkmagic(cgp) || cgp->cg_cgx != cg) { sfs = ffs_getmntstat(devvp); printf("UFS %s%s (%s)", devvp->v_type == VCHR ? "" : "snapshot of ", sfs->f_mntfromname, sfs->f_mntonname); if (!cg_chkmagic(cgp)) printf(" cg %u: bad magic number 0x%x should be 0x%x\n", cg, cgp->cg_magic, CG_MAGIC); else printf(": wrong cylinder group cg %u != cgx %u\n", cg, cgp->cg_cgx); bp->b_flags &= ~B_CKHASH; bp->b_flags |= B_INVAL | B_NOCACHE; brelse(bp); return (EIO); } bp->b_flags &= ~B_CKHASH; bp->b_xflags |= BX_BKGRDWRITE; /* * If we are using check hashes on the cylinder group then we want * to limit changing the cylinder group time to when we are actually * going to write it to disk so that its check hash remains correct * in memory. If the CK_CYLGRP flag is set the time is updated in * ffs_bufwrite() as the buffer is queued for writing. Otherwise we * update the time here as we have done historically. */ if ((fs->fs_metackhash & CK_CYLGRP) != 0) bp->b_xflags |= BX_CYLGRP; else cgp->cg_old_time = cgp->cg_time = time_second; *bpp = bp; *cgpp = cgp; return (0); } static void ffs_ckhash_cg(struct buf *bp) { uint32_t ckhash; struct cg *cgp; cgp = (struct cg *)bp->b_data; ckhash = cgp->cg_ckhash; cgp->cg_ckhash = 0; bp->b_ckhash = calculate_crc32c(~0L, bp->b_data, bp->b_bcount); cgp->cg_ckhash = ckhash; } /* * Fserr prints the name of a filesystem with an error diagnostic. * * The form of the error message is: * fs: error message */ void ffs_fserr(struct fs *fs, ino_t inum, char *cp) { struct thread *td = curthread; /* XXX */ struct proc *p = td->td_proc; log(LOG_ERR, "pid %d (%s), uid %d inumber %ju on %s: %s\n", p->p_pid, p->p_comm, td->td_ucred->cr_uid, (uintmax_t)inum, fs->fs_fsmnt, cp); } /* * This function provides the capability for the fsck program to * update an active filesystem. Fourteen operations are provided: * * adjrefcnt(inode, amt) - adjusts the reference count on the * specified inode by the specified amount. Under normal * operation the count should always go down. Decrementing * the count to zero will cause the inode to be freed. * adjblkcnt(inode, amt) - adjust the number of blocks used by the * inode by the specified amount. + * adjdepth(inode, amt) - adjust the depth of the specified directory + * inode by the specified amount. * setsize(inode, size) - set the size of the inode to the * specified size. * adjndir, adjbfree, adjifree, adjffree, adjnumclusters(amt) - * adjust the superblock summary. * freedirs(inode, count) - directory inodes [inode..inode + count - 1] * are marked as free. Inodes should never have to be marked * as in use. * freefiles(inode, count) - file inodes [inode..inode + count - 1] * are marked as free. Inodes should never have to be marked * as in use. * freeblks(blockno, size) - blocks [blockno..blockno + size - 1] * are marked as free. Blocks should never have to be marked * as in use. * setflags(flags, set/clear) - the fs_flags field has the specified * flags set (second parameter +1) or cleared (second parameter -1). * setcwd(dirinode) - set the current directory to dirinode in the * filesystem associated with the snapshot. * setdotdot(oldvalue, newvalue) - Verify that the inode number for ".." * in the current directory is oldvalue then change it to newvalue. * unlink(nameptr, oldvalue) - Verify that the inode number associated * with nameptr in the current directory is oldvalue then unlink it. */ static int sysctl_ffs_fsck(SYSCTL_HANDLER_ARGS); SYSCTL_PROC(_vfs_ffs, FFS_ADJ_REFCNT, adjrefcnt, CTLFLAG_WR | CTLTYPE_STRUCT | CTLFLAG_NEEDGIANT, 0, 0, sysctl_ffs_fsck, "S,fsck", "Adjust Inode Reference Count"); static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_BLKCNT, adjblkcnt, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Adjust Inode Used Blocks Count"); +static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_DEPTH, adjdepth, + CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, + "Adjust Directory Inode Depth"); + static SYSCTL_NODE(_vfs_ffs, FFS_SET_SIZE, setsize, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Set the inode size"); static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NDIR, adjndir, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Adjust number of directories"); static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NBFREE, adjnbfree, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Adjust number of free blocks"); static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NIFREE, adjnifree, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Adjust number of free inodes"); static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NFFREE, adjnffree, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Adjust number of free frags"); static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NUMCLUSTERS, adjnumclusters, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Adjust number of free clusters"); static SYSCTL_NODE(_vfs_ffs, FFS_DIR_FREE, freedirs, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Free Range of Directory Inodes"); static SYSCTL_NODE(_vfs_ffs, FFS_FILE_FREE, freefiles, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Free Range of File Inodes"); static SYSCTL_NODE(_vfs_ffs, FFS_BLK_FREE, freeblks, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Free Range of Blocks"); static SYSCTL_NODE(_vfs_ffs, FFS_SET_FLAGS, setflags, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Change Filesystem Flags"); static SYSCTL_NODE(_vfs_ffs, FFS_SET_CWD, setcwd, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Set Current Working Directory"); static SYSCTL_NODE(_vfs_ffs, FFS_SET_DOTDOT, setdotdot, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Change Value of .. Entry"); static SYSCTL_NODE(_vfs_ffs, FFS_UNLINK, unlink, CTLFLAG_WR | CTLFLAG_NEEDGIANT, sysctl_ffs_fsck, "Unlink a Duplicate Name"); #ifdef DIAGNOSTIC static int fsckcmds = 0; SYSCTL_INT(_debug, OID_AUTO, ffs_fsckcmds, CTLFLAG_RW, &fsckcmds, 0, "print out fsck_ffs-based filesystem update commands"); #endif /* DIAGNOSTIC */ static int sysctl_ffs_fsck(SYSCTL_HANDLER_ARGS) { struct thread *td = curthread; struct fsck_cmd cmd; struct ufsmount *ump; struct vnode *vp, *dvp, *fdvp; struct inode *ip, *dp; struct mount *mp; struct fs *fs; struct pwd *pwd; ufs2_daddr_t blkno; long blkcnt, blksize; u_long key; struct file *fp; cap_rights_t rights; int filetype, error; if (req->newptr == NULL || req->newlen > sizeof(cmd)) return (EBADRPC); if ((error = SYSCTL_IN(req, &cmd, sizeof(cmd))) != 0) return (error); if (cmd.version != FFS_CMD_VERSION) return (ERPCMISMATCH); if ((error = getvnode(td, cmd.handle, cap_rights_init_one(&rights, CAP_FSCK), &fp)) != 0) return (error); vp = fp->f_vnode; if (vp->v_type != VREG && vp->v_type != VDIR) { fdrop(fp, td); return (EINVAL); } vn_start_write(vp, &mp, V_WAIT); if (mp == NULL || strncmp(mp->mnt_stat.f_fstypename, "ufs", MFSNAMELEN)) { vn_finished_write(mp); fdrop(fp, td); return (EINVAL); } ump = VFSTOUFS(mp); if (mp->mnt_flag & MNT_RDONLY) { vn_finished_write(mp); fdrop(fp, td); return (EROFS); } fs = ump->um_fs; filetype = IFREG; switch (oidp->oid_number) { case FFS_SET_FLAGS: #ifdef DIAGNOSTIC if (fsckcmds) printf("%s: %s flags\n", mp->mnt_stat.f_mntonname, cmd.size > 0 ? "set" : "clear"); #endif /* DIAGNOSTIC */ if (cmd.size > 0) fs->fs_flags |= (long)cmd.value; else fs->fs_flags &= ~(long)cmd.value; break; case FFS_ADJ_REFCNT: #ifdef DIAGNOSTIC if (fsckcmds) { printf("%s: adjust inode %jd link count by %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value, (intmax_t)cmd.size); } #endif /* DIAGNOSTIC */ if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &vp))) break; ip = VTOI(vp); ip->i_nlink += cmd.size; DIP_SET(ip, i_nlink, ip->i_nlink); ip->i_effnlink += cmd.size; UFS_INODE_SET_FLAG(ip, IN_CHANGE | IN_MODIFIED); error = ffs_update(vp, 1); if (DOINGSOFTDEP(vp)) softdep_change_linkcnt(ip); vput(vp); break; case FFS_ADJ_BLKCNT: #ifdef DIAGNOSTIC if (fsckcmds) { printf("%s: adjust inode %jd block count by %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value, (intmax_t)cmd.size); } #endif /* DIAGNOSTIC */ if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &vp))) break; ip = VTOI(vp); DIP_SET(ip, i_blocks, DIP(ip, i_blocks) + cmd.size); UFS_INODE_SET_FLAG(ip, IN_CHANGE | IN_MODIFIED); error = ffs_update(vp, 1); vput(vp); break; + case FFS_ADJ_DEPTH: +#ifdef DIAGNOSTIC + if (fsckcmds) { + printf("%s: adjust directory inode %jd depth by %jd\n", + mp->mnt_stat.f_mntonname, (intmax_t)cmd.value, + (intmax_t)cmd.size); + } +#endif /* DIAGNOSTIC */ + if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &vp))) + break; + if (vp->v_type != VDIR) { + vput(vp); + error = ENOTDIR; + break; + } + ip = VTOI(vp); + DIP_SET(ip, i_dirdepth, DIP(ip, i_dirdepth) + cmd.size); + UFS_INODE_SET_FLAG(ip, IN_CHANGE | IN_MODIFIED); + error = ffs_update(vp, 1); + vput(vp); + break; + case FFS_SET_SIZE: #ifdef DIAGNOSTIC if (fsckcmds) { printf("%s: set inode %jd size to %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value, (intmax_t)cmd.size); } #endif /* DIAGNOSTIC */ if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &vp))) break; ip = VTOI(vp); DIP_SET(ip, i_size, cmd.size); UFS_INODE_SET_FLAG(ip, IN_SIZEMOD | IN_CHANGE | IN_MODIFIED); error = ffs_update(vp, 1); vput(vp); break; case FFS_DIR_FREE: filetype = IFDIR; /* fall through */ case FFS_FILE_FREE: #ifdef DIAGNOSTIC if (fsckcmds) { if (cmd.size == 1) printf("%s: free %s inode %ju\n", mp->mnt_stat.f_mntonname, filetype == IFDIR ? "directory" : "file", (uintmax_t)cmd.value); else printf("%s: free %s inodes %ju-%ju\n", mp->mnt_stat.f_mntonname, filetype == IFDIR ? "directory" : "file", (uintmax_t)cmd.value, (uintmax_t)(cmd.value + cmd.size - 1)); } #endif /* DIAGNOSTIC */ while (cmd.size > 0) { if ((error = ffs_freefile(ump, fs, ump->um_devvp, cmd.value, filetype, NULL))) break; cmd.size -= 1; cmd.value += 1; } break; case FFS_BLK_FREE: #ifdef DIAGNOSTIC if (fsckcmds) { if (cmd.size == 1) printf("%s: free block %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); else printf("%s: free blocks %jd-%jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value, (intmax_t)cmd.value + cmd.size - 1); } #endif /* DIAGNOSTIC */ blkno = cmd.value; blkcnt = cmd.size; blksize = fs->fs_frag - (blkno % fs->fs_frag); key = ffs_blkrelease_start(ump, ump->um_devvp, UFS_ROOTINO); while (blkcnt > 0) { if (blkcnt < blksize) blksize = blkcnt; ffs_blkfree(ump, fs, ump->um_devvp, blkno, blksize * fs->fs_fsize, UFS_ROOTINO, VDIR, NULL, key); blkno += blksize; blkcnt -= blksize; blksize = fs->fs_frag; } ffs_blkrelease_finish(ump, key); break; /* * Adjust superblock summaries. fsck(8) is expected to * submit deltas when necessary. */ case FFS_ADJ_NDIR: #ifdef DIAGNOSTIC if (fsckcmds) { printf("%s: adjust number of directories by %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DIAGNOSTIC */ fs->fs_cstotal.cs_ndir += cmd.value; break; case FFS_ADJ_NBFREE: #ifdef DIAGNOSTIC if (fsckcmds) { printf("%s: adjust number of free blocks by %+jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DIAGNOSTIC */ fs->fs_cstotal.cs_nbfree += cmd.value; break; case FFS_ADJ_NIFREE: #ifdef DIAGNOSTIC if (fsckcmds) { printf("%s: adjust number of free inodes by %+jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DIAGNOSTIC */ fs->fs_cstotal.cs_nifree += cmd.value; break; case FFS_ADJ_NFFREE: #ifdef DIAGNOSTIC if (fsckcmds) { printf("%s: adjust number of free frags by %+jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DIAGNOSTIC */ fs->fs_cstotal.cs_nffree += cmd.value; break; case FFS_ADJ_NUMCLUSTERS: #ifdef DIAGNOSTIC if (fsckcmds) { printf("%s: adjust number of free clusters by %+jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DIAGNOSTIC */ fs->fs_cstotal.cs_numclusters += cmd.value; break; case FFS_SET_CWD: #ifdef DIAGNOSTIC if (fsckcmds) { printf("%s: set current directory to inode %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DIAGNOSTIC */ if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_SHARED, &vp))) break; AUDIT_ARG_VNODE1(vp); if ((error = change_dir(vp, td)) != 0) { vput(vp); break; } VOP_UNLOCK(vp); pwd_chdir(td, vp); break; case FFS_SET_DOTDOT: #ifdef DIAGNOSTIC if (fsckcmds) { printf("%s: change .. in cwd from %jd to %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value, (intmax_t)cmd.size); } #endif /* DIAGNOSTIC */ /* * First we have to get and lock the parent directory * to which ".." points. */ error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &fdvp); if (error) break; /* * Now we get and lock the child directory containing "..". */ pwd = pwd_hold(td); dvp = pwd->pwd_cdir; if ((error = vget(dvp, LK_EXCLUSIVE)) != 0) { vput(fdvp); pwd_drop(pwd); break; } dp = VTOI(dvp); SET_I_OFFSET(dp, 12); /* XXX mastertemplate.dot_reclen */ error = ufs_dirrewrite(dp, VTOI(fdvp), (ino_t)cmd.size, DT_DIR, 0); cache_purge(fdvp); cache_purge(dvp); vput(dvp); vput(fdvp); pwd_drop(pwd); break; case FFS_UNLINK: #ifdef DIAGNOSTIC if (fsckcmds) { char buf[32]; if (copyinstr((char *)(intptr_t)cmd.value, buf,32,NULL)) strncpy(buf, "Name_too_long", 32); printf("%s: unlink %s (inode %jd)\n", mp->mnt_stat.f_mntonname, buf, (intmax_t)cmd.size); } #endif /* DIAGNOSTIC */ /* * kern_funlinkat will do its own start/finish writes and * they do not nest, so drop ours here. Setting mp == NULL * indicates that vn_finished_write is not needed down below. */ vn_finished_write(mp); mp = NULL; error = kern_funlinkat(td, AT_FDCWD, (char *)(intptr_t)cmd.value, FD_NONE, UIO_USERSPACE, 0, (ino_t)cmd.size); break; default: #ifdef DIAGNOSTIC if (fsckcmds) { printf("Invalid request %d from fsck\n", oidp->oid_number); } #endif /* DIAGNOSTIC */ error = EINVAL; break; } fdrop(fp, td); vn_finished_write(mp); return (error); } diff --git a/sys/ufs/ffs/fs.h b/sys/ufs/ffs/fs.h index 7b7a1f9f2ef6..1f2a56c26641 100644 --- a/sys/ufs/ffs/fs.h +++ b/sys/ufs/ffs/fs.h @@ -1,904 +1,905 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 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. 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. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)fs.h 8.13 (Berkeley) 3/21/95 * $FreeBSD$ */ #ifndef _UFS_FFS_FS_H_ #define _UFS_FFS_FS_H_ #include #include /* * Each disk drive contains some number of filesystems. * A filesystem consists of a number of cylinder groups. * Each cylinder group has inodes and data. * * A filesystem is described by its super-block, which in turn * describes the cylinder groups. The super-block is critical * data and is replicated in each cylinder group to protect against * catastrophic loss. This is done at `newfs' time and the critical * super-block data does not change, so the copies need not be * referenced further unless disaster strikes. * * For filesystem fs, the offsets of the various blocks of interest * are given in the super block as: * [fs->fs_sblkno] Super-block * [fs->fs_cblkno] Cylinder group block * [fs->fs_iblkno] Inode blocks * [fs->fs_dblkno] Data blocks * The beginning of cylinder group cg in fs, is given by * the ``cgbase(fs, cg)'' macro. * * Depending on the architecture and the media, the superblock may * reside in any one of four places. For tiny media where every block * counts, it is placed at the very front of the partition. Historically, * UFS1 placed it 8K from the front to leave room for the disk label and * a small bootstrap. For UFS2 it got moved to 64K from the front to leave * room for the disk label and a bigger bootstrap, and for really piggy * systems we check at 256K from the front if the first three fail. In * all cases the size of the superblock will be SBLOCKSIZE. All values are * given in byte-offset form, so they do not imply a sector size. The * SBLOCKSEARCH specifies the order in which the locations should be searched. */ #define SBLOCK_FLOPPY 0 #define SBLOCK_UFS1 8192 #define SBLOCK_UFS2 65536 #define SBLOCK_PIGGY 262144 #define SBLOCKSIZE 8192 #define SBLOCKSEARCH \ { SBLOCK_UFS2, SBLOCK_UFS1, SBLOCK_FLOPPY, SBLOCK_PIGGY, -1 } /* * Request standard superblock location in ffs_sbget(). */ #define UFS_STDSB -1 /* Search standard places for superblock */ /* * UFS_NOMSG indicates that superblock inconsistency error messages * should not be printed. It is used by programs like fsck that * want to print their own error message. * * UFS_NOCSUM causes only the superblock itself to be returned, but does * not read in any auxiliary data structures like the cylinder group * summary information. It is used by clients like glabel that just * want to check for possible filesystem types. Using UFS_NOCSUM * skips the superblock checks for csum data which allows superblocks * that have corrupted csum data to be read and used. * * UFS_NOHASHFAIL will note that the check hash is wrong but will still * return the superblock. This is used by the bootstrap code to * give the system a chance to come up so that fsck can be run to * correct the problem. * * UFS_NOWARNFAIL will warn about inconsistencies but still return the * superblock. It includes UFS_NOHASHFAIL. UFS_NOWARNFAIL is used by * programs like fsck_ffs(8) to debug broken filesystems. * * UFS_FSRONLY will only validate the superblock fields needed to * calculate where the backup filesystem superblocks are located. * If these values pass their validation tests, then the superblock * is returned. This flag is used as part of the attempt to find * alternate superblocks when using ffs_sbsearch(). */ #define UFS_NOHASHFAIL 0x0001 /* Ignore check-hash failure */ #define UFS_NOWARNFAIL 0x0003 /* Ignore non-fatal inconsistencies */ #define UFS_NOMSG 0x0004 /* Print no error message */ #define UFS_NOCSUM 0x0008 /* Read just the superblock without csum */ #define UFS_FSRONLY 0x0010 /* Validate only values needed for recovery of alternate superblocks */ #define UFS_ALTSBLK 0x1000 /* Flag used internally */ /* * Max number of fragments per block. This value is NOT tweakable. */ #define MAXFRAG 8 /* * Addresses stored in inodes are capable of addressing fragments * of `blocks'. File system blocks of at most size MAXBSIZE can * be optionally broken into 2, 4, or 8 pieces, each of which is * addressable; these pieces may be DEV_BSIZE, or some multiple of * a DEV_BSIZE unit. * * Large files consist of exclusively large data blocks. To avoid * undue wasted disk space, the last data block of a small file may be * allocated as only as many fragments of a large block as are * necessary. The filesystem format retains only a single pointer * to such a fragment, which is a piece of a single large block that * has been divided. The size of such a fragment is determinable from * information in the inode, using the ``blksize(fs, ip, lbn)'' macro. * * The filesystem records space availability at the fragment level; * to determine block availability, aligned fragments are examined. */ /* * MINBSIZE is the smallest allowable block size. * In order to insure that it is possible to create files of size * 2^32 with only two levels of indirection, MINBSIZE is set to 4096. * MINBSIZE must be big enough to hold a cylinder group block, * thus changes to (struct cg) must keep its size within MINBSIZE. * Note that super blocks are always of size SBLOCKSIZE, * and that both SBLOCKSIZE and MAXBSIZE must be >= MINBSIZE. */ #define MINBSIZE 4096 /* * The path name on which the filesystem is mounted is maintained * in fs_fsmnt. MAXMNTLEN defines the amount of space allocated in * the super block for this name. */ #define MAXMNTLEN 468 /* * The volume name for this filesystem is maintained in fs_volname. * MAXVOLLEN defines the length of the buffer allocated. */ #define MAXVOLLEN 32 /* * There is a 128-byte region in the superblock reserved for in-core * pointers to summary information. Originally this included an array * of pointers to blocks of struct csum; now there are just a few * pointers and the remaining space is padded with fs_ocsp[]. * * NOCSPTRS determines the size of this padding. Historically this * space was used to store pointers to structures that summaried * filesystem usage and layout information. However, these pointers * left various kernel pointers in the superblock which made otherwise * identical superblocks appear to have differences. So, all the * pointers in the superblock were moved to a fs_summary_info structure * reducing the superblock to having only a single pointer to this * structure. When writing the superblock to disk, this pointer is * temporarily NULL'ed out so that the kernel pointer will not appear * in the on-disk copy of the superblock. */ #define NOCSPTRS ((128 / sizeof(void *)) - 1) /* * A summary of contiguous blocks of various sizes is maintained * in each cylinder group. Normally this is set by the initial * value of fs_maxcontig. To conserve space, a maximum summary size * is set by FS_MAXCONTIG. */ #define FS_MAXCONTIG 16 /* * MINFREE gives the minimum acceptable percentage of filesystem * blocks which may be free. If the freelist drops below this level * only the superuser may continue to allocate blocks. This may * be set to 0 if no reserve of free blocks is deemed necessary, * however throughput drops by fifty percent if the filesystem * is run at between 95% and 100% full; thus the minimum default * value of fs_minfree is 5%. However, to get good clustering * performance, 10% is a better choice. hence we use 10% as our * default value. With 10% free space, fragmentation is not a * problem, so we choose to optimize for time. */ #define MINFREE 8 #define DEFAULTOPT FS_OPTTIME /* * Grigoriy Orlov has done some extensive work to fine * tune the layout preferences for directories within a filesystem. * His algorithm can be tuned by adjusting the following parameters * which tell the system the average file size and the average number * of files per directory. These defaults are well selected for typical * filesystems, but may need to be tuned for odd cases like filesystems * being used for squid caches or news spools. */ #define AVFILESIZ 16384 /* expected average file size */ #define AFPDIR 64 /* expected number of files per directory */ /* * The maximum number of snapshot nodes that can be associated * with each filesystem. This limit affects only the number of * snapshot files that can be recorded within the superblock so * that they can be found when the filesystem is mounted. However, * maintaining too many will slow the filesystem performance, so * having this limit is a good idea. */ #define FSMAXSNAP 20 /* * Used to identify special blocks in snapshots: * * BLK_NOCOPY - A block that was unallocated at the time the snapshot * was taken, hence does not need to be copied when written. * BLK_SNAP - A block held by another snapshot that is not needed by this * snapshot. When the other snapshot is freed, the BLK_SNAP entries * are converted to BLK_NOCOPY. These are needed to allow fsck to * identify blocks that are in use by other snapshots (which are * expunged from this snapshot). */ #define BLK_NOCOPY ((ufs2_daddr_t)(1)) #define BLK_SNAP ((ufs2_daddr_t)(2)) /* * Sysctl values for the fast filesystem. */ #define FFS_ADJ_REFCNT 1 /* adjust inode reference count */ #define FFS_ADJ_BLKCNT 2 /* adjust inode used block count */ #define FFS_BLK_FREE 3 /* free range of blocks in map */ #define FFS_DIR_FREE 4 /* free specified dir inodes in map */ #define FFS_FILE_FREE 5 /* free specified file inodes in map */ #define FFS_SET_FLAGS 6 /* set filesystem flags */ #define FFS_ADJ_NDIR 7 /* adjust number of directories */ #define FFS_ADJ_NBFREE 8 /* adjust number of free blocks */ #define FFS_ADJ_NIFREE 9 /* adjust number of free inodes */ #define FFS_ADJ_NFFREE 10 /* adjust number of free frags */ #define FFS_ADJ_NUMCLUSTERS 11 /* adjust number of free clusters */ #define FFS_SET_CWD 12 /* set current directory */ #define FFS_SET_DOTDOT 13 /* set inode number for ".." */ #define FFS_UNLINK 14 /* remove a name in the filesystem */ /* Was FFS_SET_INODE 15 */ /* Was FFS_SET_BUFOUTPUT 16 */ #define FFS_SET_SIZE 17 /* set inode size */ -#define FFS_MAXID 17 /* number of valid ffs ids */ +#define FFS_ADJ_DEPTH 18 /* adjust directory inode depth */ +#define FFS_MAXID 18 /* number of valid ffs ids */ /* * Command structure passed in to the filesystem to adjust filesystem values. */ #define FFS_CMD_VERSION 0x19790518 /* version ID */ struct fsck_cmd { int32_t version; /* version of command structure */ int32_t handle; /* reference to filesystem to be changed */ int64_t value; /* inode or block number to be affected */ int64_t size; /* amount or range to be adjusted */ int64_t spare; /* reserved for future use */ }; /* * A recovery structure placed at the end of the boot block area by newfs * that can be used by fsck to search for alternate superblocks. */ struct fsrecovery { int32_t fsr_magic; /* magic number */ int32_t fsr_fsbtodb; /* fsbtodb and dbtofsb shift constant */ int32_t fsr_sblkno; /* offset of super-block in filesys */ int32_t fsr_fpg; /* blocks per group * fs_frag */ u_int32_t fsr_ncg; /* number of cylinder groups */ }; /* * Per cylinder group information; summarized in blocks allocated * from first cylinder group data blocks. These blocks have to be * read in from fs_csaddr (size fs_cssize) in addition to the * super block. */ struct csum { int32_t cs_ndir; /* number of directories */ int32_t cs_nbfree; /* number of free blocks */ int32_t cs_nifree; /* number of free inodes */ int32_t cs_nffree; /* number of free frags */ }; struct csum_total { int64_t cs_ndir; /* number of directories */ int64_t cs_nbfree; /* number of free blocks */ int64_t cs_nifree; /* number of free inodes */ int64_t cs_nffree; /* number of free frags */ int64_t cs_numclusters; /* number of free clusters */ int64_t cs_spare[3]; /* future expansion */ }; /* * Pointers to super block summary information. Placed in a separate * structure so there is just one pointer in the superblock. * * The pointers in this structure are used as follows: * fs_contigdirs references an array that tracks the creation of new * directories * fs_csp references a contiguous array of struct csum for * all cylinder groups * fs_maxcluster references an array of cluster sizes that is computed * as cylinder groups are inspected * fs_active is used when creating snapshots; it points to a bitmap * of cylinder groups for which the free-block bitmap has changed * since the snapshot operation began. */ struct fs_summary_info { uint8_t *si_contigdirs; /* (u) # of contig. allocated dirs */ struct csum *si_csp; /* (u) cg summary info buffer */ int32_t *si_maxcluster; /* (u) max cluster in each cyl group */ u_int *si_active; /* (u) used by snapshots to track fs */ }; #define fs_contigdirs fs_si->si_contigdirs #define fs_csp fs_si->si_csp #define fs_maxcluster fs_si->si_maxcluster #define fs_active fs_si->si_active /* * Super block for an FFS filesystem. */ struct fs { int32_t fs_firstfield; /* historic filesystem linked list, */ int32_t fs_unused_1; /* used for incore super blocks */ int32_t fs_sblkno; /* offset of super-block in filesys */ int32_t fs_cblkno; /* offset of cyl-block in filesys */ int32_t fs_iblkno; /* offset of inode-blocks in filesys */ int32_t fs_dblkno; /* offset of first data after cg */ int32_t fs_old_cgoffset; /* cylinder group offset in cylinder */ int32_t fs_old_cgmask; /* used to calc mod fs_ntrak */ int32_t fs_old_time; /* last time written */ int32_t fs_old_size; /* number of blocks in fs */ int32_t fs_old_dsize; /* number of data blocks in fs */ u_int32_t fs_ncg; /* number of cylinder groups */ int32_t fs_bsize; /* size of basic blocks in fs */ int32_t fs_fsize; /* size of frag blocks in fs */ int32_t fs_frag; /* number of frags in a block in fs */ /* these are configuration parameters */ int32_t fs_minfree; /* minimum percentage of free blocks */ int32_t fs_old_rotdelay; /* num of ms for optimal next block */ int32_t fs_old_rps; /* disk revolutions per second */ /* these fields can be computed from the others */ int32_t fs_bmask; /* ``blkoff'' calc of blk offsets */ int32_t fs_fmask; /* ``fragoff'' calc of frag offsets */ int32_t fs_bshift; /* ``lblkno'' calc of logical blkno */ int32_t fs_fshift; /* ``numfrags'' calc number of frags */ /* these are configuration parameters */ int32_t fs_maxcontig; /* max number of contiguous blks */ int32_t fs_maxbpg; /* max number of blks per cyl group */ /* these fields can be computed from the others */ int32_t fs_fragshift; /* block to frag shift */ int32_t fs_fsbtodb; /* fsbtodb and dbtofsb shift constant */ int32_t fs_sbsize; /* actual size of super block */ int32_t fs_spare1[2]; /* old fs_csmask */ /* old fs_csshift */ int32_t fs_nindir; /* value of NINDIR */ u_int32_t fs_inopb; /* value of INOPB */ int32_t fs_old_nspf; /* value of NSPF */ /* yet another configuration parameter */ int32_t fs_optim; /* optimization preference, see below */ int32_t fs_old_npsect; /* # sectors/track including spares */ int32_t fs_old_interleave; /* hardware sector interleave */ int32_t fs_old_trackskew; /* sector 0 skew, per track */ int32_t fs_id[2]; /* unique filesystem id */ /* sizes determined by number of cylinder groups and their sizes */ int32_t fs_old_csaddr; /* blk addr of cyl grp summary area */ int32_t fs_cssize; /* size of cyl grp summary area */ int32_t fs_cgsize; /* cylinder group size */ int32_t fs_spare2; /* old fs_ntrak */ int32_t fs_old_nsect; /* sectors per track */ int32_t fs_old_spc; /* sectors per cylinder */ int32_t fs_old_ncyl; /* cylinders in filesystem */ int32_t fs_old_cpg; /* cylinders per group */ u_int32_t fs_ipg; /* inodes per group */ int32_t fs_fpg; /* blocks per group * fs_frag */ /* this data must be re-computed after crashes */ struct csum fs_old_cstotal; /* cylinder summary information */ /* these fields are cleared at mount time */ int8_t fs_fmod; /* super block modified flag */ int8_t fs_clean; /* filesystem is clean flag */ int8_t fs_ronly; /* mounted read-only flag */ int8_t fs_old_flags; /* old FS_ flags */ u_char fs_fsmnt[MAXMNTLEN]; /* name mounted on */ u_char fs_volname[MAXVOLLEN]; /* volume name */ u_int64_t fs_swuid; /* system-wide uid */ int32_t fs_pad; /* due to alignment of fs_swuid */ /* these fields retain the current block allocation info */ int32_t fs_cgrotor; /* last cg searched */ void *fs_ocsp[NOCSPTRS]; /* padding; was list of fs_cs buffers */ struct fs_summary_info *fs_si;/* In-core pointer to summary info */ int32_t fs_old_cpc; /* cyl per cycle in postbl */ int32_t fs_maxbsize; /* maximum blocking factor permitted */ int64_t fs_unrefs; /* number of unreferenced inodes */ int64_t fs_providersize; /* size of underlying GEOM provider */ int64_t fs_metaspace; /* size of area reserved for metadata */ int64_t fs_sparecon64[13]; /* old rotation block list head */ int64_t fs_sblockactualloc; /* byte offset of this superblock */ int64_t fs_sblockloc; /* byte offset of standard superblock */ struct csum_total fs_cstotal; /* (u) cylinder summary information */ ufs_time_t fs_time; /* last time written */ int64_t fs_size; /* number of blocks in fs */ int64_t fs_dsize; /* number of data blocks in fs */ ufs2_daddr_t fs_csaddr; /* blk addr of cyl grp summary area */ int64_t fs_pendingblocks; /* (u) blocks being freed */ u_int32_t fs_pendinginodes; /* (u) inodes being freed */ uint32_t fs_snapinum[FSMAXSNAP];/* list of snapshot inode numbers */ u_int32_t fs_avgfilesize; /* expected average file size */ u_int32_t fs_avgfpdir; /* expected # of files per directory */ int32_t fs_save_cgsize; /* save real cg size to use fs_bsize */ ufs_time_t fs_mtime; /* Last mount or fsck time. */ int32_t fs_sujfree; /* SUJ free list */ int32_t fs_sparecon32[21]; /* reserved for future constants */ u_int32_t fs_ckhash; /* if CK_SUPERBLOCK, its check-hash */ u_int32_t fs_metackhash; /* metadata check-hash, see CK_ below */ int32_t fs_flags; /* see FS_ flags below */ int32_t fs_contigsumsize; /* size of cluster summary array */ int32_t fs_maxsymlinklen; /* max length of an internal symlink */ int32_t fs_old_inodefmt; /* format of on-disk inodes */ u_int64_t fs_maxfilesize; /* maximum representable file size */ int64_t fs_qbmask; /* ~fs_bmask for use with 64-bit size */ int64_t fs_qfmask; /* ~fs_fmask for use with 64-bit size */ int32_t fs_state; /* validate fs_clean field */ int32_t fs_old_postblformat; /* format of positional layout tables */ int32_t fs_old_nrpos; /* number of rotational positions */ int32_t fs_spare5[2]; /* old fs_postbloff */ /* old fs_rotbloff */ int32_t fs_magic; /* magic number */ }; /* Sanity checking. */ #ifdef CTASSERT CTASSERT(sizeof(struct fs) == 1376); #endif /* * Filesystem identification */ #define FS_UFS1_MAGIC 0x011954 /* UFS1 fast filesystem magic number */ #define FS_UFS2_MAGIC 0x19540119 /* UFS2 fast filesystem magic number */ #define FS_BAD_MAGIC 0x19960408 /* UFS incomplete newfs magic number */ #define FS_42INODEFMT -1 /* 4.2BSD inode format */ #define FS_44INODEFMT 2 /* 4.4BSD inode format */ /* * Preference for optimization. */ #define FS_OPTTIME 0 /* minimize allocation time */ #define FS_OPTSPACE 1 /* minimize disk fragmentation */ /* * Filesystem flags. * * The FS_UNCLEAN flag is set by the kernel when the filesystem was * mounted with fs_clean set to zero. The FS_DOSOFTDEP flag indicates * that the filesystem should be managed by the soft updates code. * Note that the FS_NEEDSFSCK flag is set and cleared by the fsck * utility. It is set when background fsck finds an unexpected * inconsistency which requires a traditional foreground fsck to be * run. Such inconsistencies should only be found after an uncorrectable * disk error. The FS_NEEDSFSCK can also be set when a mounted filesystem * discovers an internal inconsistency such as freeing a freed inode. * A foreground fsck will clear the FS_NEEDSFSCK flag when it has * successfully cleaned up the filesystem. The kernel uses this * flag to enforce that inconsistent filesystems be mounted read-only. * * The FS_METACKHASH flag when set indicates that the kernel maintains * one or more check hashes. The actual set of supported check hashes * is stored in the fs_metackhash field. Kernels that do not support * check hashes clear the FS_METACKHASH flag to indicate that the * check hashes need to be rebuilt (by fsck) before they can be used. * * When a filesystem is mounted, any flags not included in FS_SUPPORTED * are cleared. This lets newer features know that the filesystem has * been run on an older version of the filesystem and thus that data * structures associated with those features are out-of-date and need * to be rebuilt. * * FS_ACLS indicates that POSIX.1e ACLs are administratively enabled * for the file system, so they should be loaded from extended attributes, * observed for access control purposes, and be administered by object * owners. FS_NFS4ACLS indicates that NFSv4 ACLs are administratively * enabled. This flag is mutually exclusive with FS_ACLS. FS_MULTILABEL * indicates that the TrustedBSD MAC Framework should attempt to back MAC * labels into extended attributes on the file system rather than maintain * a single mount label for all objects. */ #define FS_UNCLEAN 0x00000001 /* filesystem not clean at mount */ #define FS_DOSOFTDEP 0x00000002 /* filesystem using soft dependencies */ #define FS_NEEDSFSCK 0x00000004 /* filesystem needs sync fsck before mount */ #define FS_SUJ 0x00000008 /* Filesystem using softupdate journal */ #define FS_ACLS 0x00000010 /* file system has POSIX.1e ACLs enabled */ #define FS_MULTILABEL 0x00000020 /* file system is MAC multi-label */ #define FS_GJOURNAL 0x00000040 /* gjournaled file system */ #define FS_FLAGS_UPDATED 0x0000080 /* flags have been moved to new location */ #define FS_NFS4ACLS 0x00000100 /* file system has NFSv4 ACLs enabled */ #define FS_METACKHASH 0x00000200 /* kernel supports metadata check hashes */ #define FS_TRIM 0x00000400 /* issue BIO_DELETE for deleted blocks */ #define FS_SUPPORTED 0x00FFFFFF /* supported flags, others cleared at mount*/ /* * Things that we may someday support, but currently do not. * These flags are all cleared so we know if we ran on a kernel * that does not support them. */ #define FS_INDEXDIRS 0x01000000 /* kernel supports indexed directories */ #define FS_VARBLKSIZE 0x02000000 /* kernel supports variable block sizes */ #define FS_COOLOPT1 0x04000000 /* kernel supports cool option 1 */ #define FS_COOLOPT2 0x08000000 /* kernel supports cool option 2 */ #define FS_COOLOPT3 0x10000000 /* kernel supports cool option 3 */ #define FS_COOLOPT4 0x20000000 /* kernel supports cool option 4 */ #define FS_COOLOPT5 0x40000000 /* kernel supports cool option 5 */ #define FS_COOLOPT6 0x80000000 /* kernel supports cool option 6 */ /* * The fs_metackhash field indicates the types of metadata check-hash * that are maintained for a filesystem. Not all filesystems check-hash * all metadata. */ #define CK_SUPERBLOCK 0x0001 /* the superblock */ #define CK_CYLGRP 0x0002 /* the cylinder groups */ #define CK_INODE 0x0004 /* inodes */ #define CK_INDIR 0x0008 /* indirect blocks */ #define CK_DIR 0x0010 /* directory contents */ #define CK_SUPPORTED 0x0007 /* supported flags, others cleared at mount */ /* * The BX_FSPRIV buffer b_xflags are used to track types of data in buffers. */ #define BX_SUPERBLOCK 0x00010000 /* superblock */ #define BX_CYLGRP 0x00020000 /* cylinder groups */ #define BX_INODE 0x00040000 /* inodes */ #define BX_INDIR 0x00080000 /* indirect blocks */ #define BX_DIR 0x00100000 /* directory contents */ #define PRINT_UFS_BUF_XFLAGS "\20\25dir\24indir\23inode\22cylgrp\21superblock" /* * Macros to access bits in the fs_active array. */ #define ACTIVECGNUM(fs, cg) ((fs)->fs_active[(cg) / (NBBY * sizeof(int))]) #define ACTIVECGOFF(cg) (1 << ((cg) % (NBBY * sizeof(int)))) #define ACTIVESET(fs, cg) do { \ if ((fs)->fs_active) \ ACTIVECGNUM((fs), (cg)) |= ACTIVECGOFF((cg)); \ } while (0) #define ACTIVECLEAR(fs, cg) do { \ if ((fs)->fs_active) \ ACTIVECGNUM((fs), (cg)) &= ~ACTIVECGOFF((cg)); \ } while (0) /* * The size of a cylinder group is calculated by CGSIZE. The maximum size * is limited by the fact that cylinder groups are at most one block. * Its size is derived from the size of the maps maintained in the * cylinder group and the (struct cg) size. */ #define CGSIZE(fs) \ /* base cg */ (sizeof(struct cg) + \ /* old btotoff */ (fs)->fs_old_cpg * sizeof(int32_t) + \ /* old boff */ (fs)->fs_old_cpg * sizeof(u_int16_t) + \ /* inode map */ howmany((fs)->fs_ipg, NBBY) + \ /* block map */ howmany((fs)->fs_fpg, NBBY) + sizeof(int32_t) + \ /* if present */ ((fs)->fs_contigsumsize <= 0 ? 0 : \ /* cluster sum */ (fs)->fs_contigsumsize * sizeof(int32_t) + \ /* cluster map */ howmany(fragstoblks(fs, (fs)->fs_fpg), NBBY))) /* * The minimal number of cylinder groups that should be created. */ #define MINCYLGRPS 4 /* * Convert cylinder group to base address of its global summary info. */ #define fs_cs(fs, indx) fs_csp[indx] /* * Cylinder group block for a filesystem. */ #define CG_MAGIC 0x090255 struct cg { int32_t cg_firstfield; /* historic cyl groups linked list */ int32_t cg_magic; /* magic number */ int32_t cg_old_time; /* time last written */ u_int32_t cg_cgx; /* we are the cgx'th cylinder group */ int16_t cg_old_ncyl; /* number of cyl's this cg */ int16_t cg_old_niblk; /* number of inode blocks this cg */ u_int32_t cg_ndblk; /* number of data blocks this cg */ struct csum cg_cs; /* cylinder summary information */ u_int32_t cg_rotor; /* position of last used block */ u_int32_t cg_frotor; /* position of last used frag */ u_int32_t cg_irotor; /* position of last used inode */ u_int32_t cg_frsum[MAXFRAG]; /* counts of available frags */ int32_t cg_old_btotoff; /* (int32) block totals per cylinder */ int32_t cg_old_boff; /* (u_int16) free block positions */ u_int32_t cg_iusedoff; /* (u_int8) used inode map */ u_int32_t cg_freeoff; /* (u_int8) free block map */ u_int32_t cg_nextfreeoff; /* (u_int8) next available space */ u_int32_t cg_clustersumoff; /* (u_int32) counts of avail clusters */ u_int32_t cg_clusteroff; /* (u_int8) free cluster map */ u_int32_t cg_nclusterblks; /* number of clusters this cg */ u_int32_t cg_niblk; /* number of inode blocks this cg */ u_int32_t cg_initediblk; /* last initialized inode */ u_int32_t cg_unrefs; /* number of unreferenced inodes */ int32_t cg_sparecon32[1]; /* reserved for future use */ u_int32_t cg_ckhash; /* check-hash of this cg */ ufs_time_t cg_time; /* time last written */ int64_t cg_sparecon64[3]; /* reserved for future use */ /* actually longer - space used for cylinder group maps */ }; /* * Macros for access to cylinder group array structures */ #define cg_chkmagic(cgp) ((cgp)->cg_magic == CG_MAGIC) #define cg_inosused(cgp) \ ((u_int8_t *)((u_int8_t *)(cgp) + (cgp)->cg_iusedoff)) #define cg_blksfree(cgp) \ ((u_int8_t *)((u_int8_t *)(cgp) + (cgp)->cg_freeoff)) #define cg_clustersfree(cgp) \ ((u_int8_t *)((u_int8_t *)(cgp) + (cgp)->cg_clusteroff)) #define cg_clustersum(cgp) \ ((int32_t *)((uintptr_t)(cgp) + (cgp)->cg_clustersumoff)) /* * Turn filesystem block numbers into disk block addresses. * This maps filesystem blocks to device size blocks. */ #define fsbtodb(fs, b) ((daddr_t)(b) << (fs)->fs_fsbtodb) #define dbtofsb(fs, b) ((b) >> (fs)->fs_fsbtodb) /* * Cylinder group macros to locate things in cylinder groups. * They calc filesystem addresses of cylinder group data structures. */ #define cgbase(fs, c) (((ufs2_daddr_t)(fs)->fs_fpg) * (c)) #define cgdata(fs, c) (cgdmin(fs, c) + (fs)->fs_metaspace) /* data zone */ #define cgmeta(fs, c) (cgdmin(fs, c)) /* meta data */ #define cgdmin(fs, c) (cgstart(fs, c) + (fs)->fs_dblkno) /* 1st data */ #define cgimin(fs, c) (cgstart(fs, c) + (fs)->fs_iblkno) /* inode blk */ #define cgsblock(fs, c) (cgstart(fs, c) + (fs)->fs_sblkno) /* super blk */ #define cgtod(fs, c) (cgstart(fs, c) + (fs)->fs_cblkno) /* cg block */ #define cgstart(fs, c) \ ((fs)->fs_magic == FS_UFS2_MAGIC ? cgbase(fs, c) : \ (cgbase(fs, c) + (fs)->fs_old_cgoffset * ((c) & ~((fs)->fs_old_cgmask)))) /* * Macros for handling inode numbers: * inode number to filesystem block offset. * inode number to cylinder group number. * inode number to filesystem block address. */ #define ino_to_cg(fs, x) (((ino_t)(x)) / (fs)->fs_ipg) #define ino_to_fsba(fs, x) \ ((ufs2_daddr_t)(cgimin(fs, ino_to_cg(fs, (ino_t)(x))) + \ (blkstofrags((fs), ((((ino_t)(x)) % (fs)->fs_ipg) / INOPB(fs)))))) #define ino_to_fsbo(fs, x) (((ino_t)(x)) % INOPB(fs)) /* * Give cylinder group number for a filesystem block. * Give cylinder group block number for a filesystem block. */ #define dtog(fs, d) ((d) / (fs)->fs_fpg) #define dtogd(fs, d) ((d) % (fs)->fs_fpg) /* * Extract the bits for a block from a map. * Compute the cylinder and rotational position of a cyl block addr. */ #define blkmap(fs, map, loc) \ (((map)[(loc) / NBBY] >> ((loc) % NBBY)) & (0xff >> (NBBY - (fs)->fs_frag))) /* * The following macros optimize certain frequently calculated * quantities by using shifts and masks in place of divisions * modulos and multiplications. */ #define blkoff(fs, loc) /* calculates (loc % fs->fs_bsize) */ \ ((loc) & (fs)->fs_qbmask) #define fragoff(fs, loc) /* calculates (loc % fs->fs_fsize) */ \ ((loc) & (fs)->fs_qfmask) #define lfragtosize(fs, frag) /* calculates ((off_t)frag * fs->fs_fsize) */ \ (((off_t)(frag)) << (fs)->fs_fshift) #define lblktosize(fs, blk) /* calculates ((off_t)blk * fs->fs_bsize) */ \ (((off_t)(blk)) << (fs)->fs_bshift) /* Use this only when `blk' is known to be small, e.g., < UFS_NDADDR. */ #define smalllblktosize(fs, blk) /* calculates (blk * fs->fs_bsize) */ \ ((blk) << (fs)->fs_bshift) #define lblkno(fs, loc) /* calculates (loc / fs->fs_bsize) */ \ ((loc) >> (fs)->fs_bshift) #define numfrags(fs, loc) /* calculates (loc / fs->fs_fsize) */ \ ((loc) >> (fs)->fs_fshift) #define blkroundup(fs, size) /* calculates roundup(size, fs->fs_bsize) */ \ (((size) + (fs)->fs_qbmask) & (fs)->fs_bmask) #define fragroundup(fs, size) /* calculates roundup(size, fs->fs_fsize) */ \ (((size) + (fs)->fs_qfmask) & (fs)->fs_fmask) #define fragstoblks(fs, frags) /* calculates (frags / fs->fs_frag) */ \ ((frags) >> (fs)->fs_fragshift) #define blkstofrags(fs, blks) /* calculates (blks * fs->fs_frag) */ \ ((blks) << (fs)->fs_fragshift) #define fragnum(fs, fsb) /* calculates (fsb % fs->fs_frag) */ \ ((fsb) & ((fs)->fs_frag - 1)) #define blknum(fs, fsb) /* calculates rounddown(fsb, fs->fs_frag) */ \ ((fsb) &~ ((fs)->fs_frag - 1)) /* * Determine the number of available frags given a * percentage to hold in reserve. */ #define freespace(fs, percentreserved) \ (blkstofrags((fs), (fs)->fs_cstotal.cs_nbfree) + \ (fs)->fs_cstotal.cs_nffree - \ (((off_t)((fs)->fs_dsize)) * (percentreserved) / 100)) /* * Determining the size of a file block in the filesystem. */ #define blksize(fs, ip, lbn) \ (((lbn) >= UFS_NDADDR || (ip)->i_size >= \ (uint64_t)smalllblktosize(fs, (lbn) + 1)) \ ? (fs)->fs_bsize \ : (fragroundup(fs, blkoff(fs, (ip)->i_size)))) #define sblksize(fs, size, lbn) \ (((lbn) >= UFS_NDADDR || (size) >= ((lbn) + 1) << (fs)->fs_bshift) \ ? (fs)->fs_bsize \ : (fragroundup(fs, blkoff(fs, (size))))) /* * Number of indirects in a filesystem block. */ #define NINDIR(fs) ((fs)->fs_nindir) /* * Indirect lbns are aligned on UFS_NDADDR addresses where single indirects * are the negated address of the lowest lbn reachable, double indirects * are this lbn - 1 and triple indirects are this lbn - 2. This yields * an unusual bit order to determine level. */ static inline int lbn_level(ufs_lbn_t lbn) { if (lbn >= 0) return 0; switch (lbn & 0x3) { case 0: return (0); case 1: break; case 2: return (2); case 3: return (1); default: break; } return (-1); } static inline ufs_lbn_t lbn_offset(struct fs *fs, int level) { ufs_lbn_t res; for (res = 1; level > 0; level--) res *= NINDIR(fs); return (res); } /* * Number of inodes in a secondary storage block/fragment. */ #define INOPB(fs) ((fs)->fs_inopb) #define INOPF(fs) ((fs)->fs_inopb >> (fs)->fs_fragshift) /* * Softdep journal record format. */ #define JOP_ADDREF 1 /* Add a reference to an inode. */ #define JOP_REMREF 2 /* Remove a reference from an inode. */ #define JOP_NEWBLK 3 /* Allocate a block. */ #define JOP_FREEBLK 4 /* Free a block or a tree of blocks. */ #define JOP_MVREF 5 /* Move a reference from one off to another. */ #define JOP_TRUNC 6 /* Partial truncation record. */ #define JOP_SYNC 7 /* fsync() complete record. */ #define JREC_SIZE 32 /* Record and segment header size. */ #define SUJ_MIN (4 * 1024 * 1024) /* Minimum journal size */ #define SUJ_FILE ".sujournal" /* Journal file name */ /* * Size of the segment record header. There is at most one for each disk * block in the journal. The segment header is followed by an array of * records. fsck depends on the first element in each record being 'op' * and the second being 'ino'. Segments may span multiple disk blocks but * the header is present on each. */ struct jsegrec { uint64_t jsr_seq; /* Our sequence number */ uint64_t jsr_oldest; /* Oldest valid sequence number */ uint16_t jsr_cnt; /* Count of valid records */ uint16_t jsr_blocks; /* Count of device bsize blocks. */ uint32_t jsr_crc; /* 32bit crc of the valid space */ ufs_time_t jsr_time; /* timestamp for mount instance */ }; /* * Reference record. Records a single link count modification. */ struct jrefrec { uint32_t jr_op; uint32_t jr_ino; uint32_t jr_parent; uint16_t jr_nlink; uint16_t jr_mode; int64_t jr_diroff; uint64_t jr_unused; }; /* * Move record. Records a reference moving within a directory block. The * nlink is unchanged but we must search both locations. */ struct jmvrec { uint32_t jm_op; uint32_t jm_ino; uint32_t jm_parent; uint16_t jm_unused; int64_t jm_oldoff; int64_t jm_newoff; }; /* * Block record. A set of frags or tree of blocks starting at an indirect are * freed or a set of frags are allocated. */ struct jblkrec { uint32_t jb_op; uint32_t jb_ino; ufs2_daddr_t jb_blkno; ufs_lbn_t jb_lbn; uint16_t jb_frags; uint16_t jb_oldfrags; uint32_t jb_unused; }; /* * Truncation record. Records a partial truncation so that it may be * completed at check time. Also used for sync records. */ struct jtrncrec { uint32_t jt_op; uint32_t jt_ino; int64_t jt_size; uint32_t jt_extsize; uint32_t jt_pad[3]; }; union jrec { struct jsegrec rec_jsegrec; struct jrefrec rec_jrefrec; struct jmvrec rec_jmvrec; struct jblkrec rec_jblkrec; struct jtrncrec rec_jtrncrec; }; #ifdef CTASSERT CTASSERT(sizeof(struct jsegrec) == JREC_SIZE); CTASSERT(sizeof(struct jrefrec) == JREC_SIZE); CTASSERT(sizeof(struct jmvrec) == JREC_SIZE); CTASSERT(sizeof(struct jblkrec) == JREC_SIZE); CTASSERT(sizeof(struct jtrncrec) == JREC_SIZE); CTASSERT(sizeof(union jrec) == JREC_SIZE); #endif extern int inside[], around[]; extern u_char *fragtbl[]; /* * IOCTLs used for filesystem write suspension. */ #define UFSSUSPEND _IOW('U', 1, fsid_t) #define UFSRESUME _IO('U', 2) #endif