/* * Check for a valid change to a users allocation. * Issue an error message if appropriate. */ static int chkdqchg(struct inode *ip, ufs2_daddr_t change, struct ucred *cred, int type, int *warn) { struct dquot *dq = ip->i_dquot[type]; ufs2_daddr_t ncurblocks = dq->dq_curblocks + change; /* * If user would exceed their hard limit, disallow space allocation. */ if (ncurblocks >= dq->dq_bhardlimit && dq->dq_bhardlimit) { if ((dq->dq_flags & DQ_BLKS) == 0 && ip->i_uid == cred->cr_uid) { dq->dq_flags |= DQ_BLKS; DQI_UNLOCK(dq); uprintf("\n%s: write failed, %s disk limit reached\n", ITOVFS(ip)->mnt_stat.f_mntonname, quotatypes[type]); return (EDQUOT); } DQI_UNLOCK(dq); return (EDQUOT); } /* * If user is over their soft limit for too long, disallow space * allocation. Reset time limit as they cross their soft limit. */ if (ncurblocks >= dq->dq_bsoftlimit && dq->dq_bsoftlimit) { if (dq->dq_curblocks < dq->dq_bsoftlimit) { dq->dq_btime = time_second + ITOUMP(ip)->um_btime[type]; if (ip->i_uid == cred->cr_uid) *warn = 1; return (0); } if (time_second > dq->dq_btime) { if ((dq->dq_flags & DQ_BLKS) == 0 && ip->i_uid == cred->cr_uid) { dq->dq_flags |= DQ_BLKS; DQI_UNLOCK(dq); uprintf("\n%s: write failed, %s " "disk quota exceeded for too long\n", ITOVFS(ip)->mnt_stat.f_mntonname, quotatypes[type]); return (EDQUOT); } DQI_UNLOCK(dq); return (EDQUOT); } } return (0); }
/* * On filesystems with quotas enabled, it is an error for a file to change * size and not to have a dquot structure associated with it. */ static void chkdquot(struct inode *ip) { struct ufsmount *ump; struct vnode *vp; int i; ump = ITOUMP(ip); vp = ITOV(ip); /* * Disk quotas must be turned off for system files. Currently * these are snapshots and quota files. */ if ((vp->v_vflag & VV_SYSTEM) != 0) return; /* * XXX: Turn off quotas for files with a negative UID or GID. * This prevents the creation of 100GB+ quota files. */ if ((int)ip->i_uid < 0 || (int)ip->i_gid < 0) return; UFS_LOCK(ump); for (i = 0; i < MAXQUOTAS; i++) { if (ump->um_quotas[i] == NULLVP || (ump->um_qflags[i] & (QTF_OPENING|QTF_CLOSING))) continue; if (ip->i_dquot[i] == NODQUOT) { UFS_UNLOCK(ump); vn_printf(ITOV(ip), "chkdquot: missing dquot "); panic("chkdquot: missing dquot"); } } UFS_UNLOCK(ump); }
/* * Check the inode limit, applying corrective action. */ int chkiq(struct inode *ip, int change, struct ucred *cred, int flags) { struct dquot *dq; int i, error, warn, do_check; #ifdef DIAGNOSTIC if ((flags & CHOWN) == 0) chkdquot(ip); #endif if (change == 0) return (0); if (change < 0) { for (i = 0; i < MAXQUOTAS; i++) { if ((dq = ip->i_dquot[i]) == NODQUOT) continue; DQI_LOCK(dq); DQI_WAIT(dq, PINOD+1, "chkiq1"); if (dq->dq_curinodes >= -change) dq->dq_curinodes += change; else dq->dq_curinodes = 0; dq->dq_flags &= ~DQ_INODS; dq->dq_flags |= DQ_MOD; DQI_UNLOCK(dq); } return (0); } if ((flags & FORCE) == 0 && priv_check_cred(cred, PRIV_VFS_EXCEEDQUOTA, 0)) do_check = 1; else do_check = 0; for (i = 0; i < MAXQUOTAS; i++) { if ((dq = ip->i_dquot[i]) == NODQUOT) continue; warn = 0; DQI_LOCK(dq); DQI_WAIT(dq, PINOD+1, "chkiq2"); if (do_check) { error = chkiqchg(ip, change, cred, i, &warn); if (error) { /* * Roll back user quota changes when * group quota failed. */ while (i > 0) { --i; dq = ip->i_dquot[i]; if (dq == NODQUOT) continue; DQI_LOCK(dq); DQI_WAIT(dq, PINOD+1, "chkiq3"); if (dq->dq_curinodes >= change) dq->dq_curinodes -= change; else dq->dq_curinodes = 0; dq->dq_flags &= ~DQ_INODS; dq->dq_flags |= DQ_MOD; DQI_UNLOCK(dq); } return (error); } } /* Reset timer when crossing soft limit */ if (dq->dq_curinodes + change >= dq->dq_isoftlimit && dq->dq_curinodes < dq->dq_isoftlimit) dq->dq_itime = time_second + ITOUMP(ip)->um_itime[i]; dq->dq_curinodes += change; dq->dq_flags |= DQ_MOD; DQI_UNLOCK(dq); if (warn) uprintf("\n%s: warning, %s inode quota exceeded\n", ITOVFS(ip)->mnt_stat.f_mntonname, quotatypes[i]); } return (0); }
/* * Update disk usage, and take corrective action. */ int chkdq(struct inode *ip, ufs2_daddr_t change, struct ucred *cred, int flags) { struct dquot *dq; ufs2_daddr_t ncurblocks; struct vnode *vp = ITOV(ip); int i, error, warn, do_check; /* * Disk quotas must be turned off for system files. Currently * snapshot and quota files. */ if ((vp->v_vflag & VV_SYSTEM) != 0) return (0); /* * XXX: Turn off quotas for files with a negative UID or GID. * This prevents the creation of 100GB+ quota files. */ if ((int)ip->i_uid < 0 || (int)ip->i_gid < 0) return (0); #ifdef DIAGNOSTIC if ((flags & CHOWN) == 0) chkdquot(ip); #endif if (change == 0) return (0); if (change < 0) { for (i = 0; i < MAXQUOTAS; i++) { if ((dq = ip->i_dquot[i]) == NODQUOT) continue; DQI_LOCK(dq); DQI_WAIT(dq, PINOD+1, "chkdq1"); ncurblocks = dq->dq_curblocks + change; if (ncurblocks >= 0) dq->dq_curblocks = ncurblocks; else dq->dq_curblocks = 0; dq->dq_flags &= ~DQ_BLKS; dq->dq_flags |= DQ_MOD; DQI_UNLOCK(dq); } return (0); } if ((flags & FORCE) == 0 && priv_check_cred(cred, PRIV_VFS_EXCEEDQUOTA, 0)) do_check = 1; else do_check = 0; for (i = 0; i < MAXQUOTAS; i++) { if ((dq = ip->i_dquot[i]) == NODQUOT) continue; warn = 0; DQI_LOCK(dq); DQI_WAIT(dq, PINOD+1, "chkdq2"); if (do_check) { error = chkdqchg(ip, change, cred, i, &warn); if (error) { /* * Roll back user quota changes when * group quota failed. */ while (i > 0) { --i; dq = ip->i_dquot[i]; if (dq == NODQUOT) continue; DQI_LOCK(dq); DQI_WAIT(dq, PINOD+1, "chkdq3"); ncurblocks = dq->dq_curblocks - change; if (ncurblocks >= 0) dq->dq_curblocks = ncurblocks; else dq->dq_curblocks = 0; dq->dq_flags &= ~DQ_BLKS; dq->dq_flags |= DQ_MOD; DQI_UNLOCK(dq); } return (error); } } /* Reset timer when crossing soft limit */ if (dq->dq_curblocks + change >= dq->dq_bsoftlimit && dq->dq_curblocks < dq->dq_bsoftlimit) dq->dq_btime = time_second + ITOUMP(ip)->um_btime[i]; dq->dq_curblocks += change; dq->dq_flags |= DQ_MOD; DQI_UNLOCK(dq); if (warn) uprintf("\n%s: warning, %s disk quota exceeded\n", ITOVFS(ip)->mnt_stat.f_mntonname, quotatypes[i]); } return (0); }
/* * Update the access, modified, and inode change times as specified by the * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode * to disk if the IN_MODIFIED flag is set (it may be set initially, or by * the timestamp update). The IN_LAZYMOD flag is set to force a write * later if not now. The IN_LAZYACCESS is set instead of IN_MODIFIED if the fs * is currently being suspended (or is suspended) and vnode has been accessed. * If we write now, then clear IN_MODIFIED, IN_LAZYACCESS and IN_LAZYMOD to * reflect the presumably successful write, and if waitfor is set, then wait * for the write to complete. */ int ffs_update (vnode *vp, int waitfor) { int error = 0; print("HARVEY TODO: %s\n", __func__); #if 0 struct fs *fs; struct buf *bp; struct inode *ip; int flags, error; ASSERT_VOP_ELOCKED(vp, "ffs_update"); ufs_itimes(vp); ip = VTOI(vp); if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0) return (0); ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED); fs = ITOFS(ip); if (fs->fs_ronly && ITOUMP(ip)->um_fsckpid == 0) return (0); /* * If we are updating a snapshot and another process is currently * writing the buffer containing the inode for this snapshot then * a deadlock can occur when it tries to check the snapshot to see * if that block needs to be copied. Thus when updating a snapshot * we check to see if the buffer is already locked, and if it is * we drop the snapshot lock until the buffer has been written * and is available to us. We have to grab a reference to the * snapshot vnode to prevent it from being removed while we are * waiting for the buffer. */ flags = 0; if (IS_SNAPSHOT(ip)) flags = GB_LOCK_NOWAIT; loop: error = breadn_flags(ITODEVVP(ip), fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), (int) fs->fs_bsize, 0, 0, 0, NOCRED, flags, &bp); if (error != 0) { if (error != EBUSY) return (error); KASSERT((IS_SNAPSHOT(ip)), ("EBUSY from non-snapshot")); /* * Wait for our inode block to become available. * * Hold a reference to the vnode to protect against * ffs_snapgone(). Since we hold a reference, it can only * get reclaimed (VI_DOOMED flag) in a forcible downgrade * or unmount. For an unmount, the entire filesystem will be * gone, so we cannot attempt to touch anything associated * with it while the vnode is unlocked; all we can do is * pause briefly and try again. If when we relock the vnode * we discover that it has been reclaimed, updating it is no * longer necessary and we can just return an error. */ vref(vp); VOP_UNLOCK(vp, 0); pause("ffsupd", 1); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); vrele(vp); if ((vp->v_iflag & VI_DOOMED) != 0) return (ENOENT); goto loop; } if (DOINGSOFTDEP(vp)) softdep_update_inodeblock(ip, bp, waitfor); else if (ip->i_effnlink != ip->i_nlink) panic("ffs_update: bad link cnt"); if (I_IS_UFS1(ip)) { *((struct ufs1_dinode *)bp->b_data + ino_to_fsbo(fs, ip->i_number)) = *ip->i_din1; /* XXX: FIX? The entropy here is desirable, but the harvesting may be expensive */ random_harvest_queue(&(ip->i_din1), sizeof(ip->i_din1), 1, RANDOM_FS_ATIME); } else { *((struct ufs2_dinode *)bp->b_data + ino_to_fsbo(fs, ip->i_number)) = *ip->i_din2; /* XXX: FIX? The entropy here is desirable, but the harvesting may be expensive */ random_harvest_queue(&(ip->i_din2), sizeof(ip->i_din2), 1, RANDOM_FS_ATIME); } if (waitfor) error = bwrite(bp); else if (vm_page_count_severe() || buf_dirty_count_severe()) { bawrite(bp); error = 0; } else { if (bp->b_bufsize == fs->fs_bsize) bp->b_flags |= B_CLUSTEROK; bdwrite(bp); error = 0; } #endif // 0 return (error); }
/* * Balloc defines the structure of filesystem storage * by allocating the physical blocks on a device given * the inode and the logical block number in a file. * This is the allocation strategy for UFS1. Below is * the allocation strategy for UFS2. */ int ffs_balloc_ufs1(struct vnode *vp, off_t startoffset, int size, struct ucred *cred, int flags, struct buf **bpp) { struct inode *ip; struct ufs1_dinode *dp; ufs_lbn_t lbn, lastlbn; struct fs *fs; ufs1_daddr_t nb; struct buf *bp, *nbp; struct ufsmount *ump; struct indir indirs[NIADDR + 2]; int deallocated, osize, nsize, num, i, error; ufs2_daddr_t newb; ufs1_daddr_t *bap, pref; ufs1_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1]; ufs2_daddr_t *lbns_remfree, lbns[NIADDR + 1]; int unwindidx = -1; int saved_inbdflush; static struct timeval lastfail; static int curfail; int gbflags, reclaimed; ip = VTOI(vp); dp = ip->i_din1; fs = ITOFS(ip); ump = ITOUMP(ip); lbn = lblkno(fs, startoffset); size = blkoff(fs, startoffset) + size; reclaimed = 0; if (size > fs->fs_bsize) panic("ffs_balloc_ufs1: blk too big"); *bpp = NULL; if (flags & IO_EXT) return (EOPNOTSUPP); if (lbn < 0) return (EFBIG); gbflags = (flags & BA_UNMAPPED) != 0 ? GB_UNMAPPED : 0; if (DOINGSOFTDEP(vp)) softdep_prealloc(vp, MNT_WAIT); /* * If the next write will extend the file into a new block, * and the file is currently composed of a fragment * this fragment has to be extended to be a full block. */ lastlbn = lblkno(fs, ip->i_size); if (lastlbn < NDADDR && lastlbn < lbn) { nb = lastlbn; osize = blksize(fs, ip, nb); if (osize < fs->fs_bsize && osize > 0) { UFS_LOCK(ump); error = ffs_realloccg(ip, nb, dp->di_db[nb], ffs_blkpref_ufs1(ip, lastlbn, (int)nb, &dp->di_db[0]), osize, (int)fs->fs_bsize, flags, cred, &bp); if (error) return (error); if (DOINGSOFTDEP(vp)) softdep_setup_allocdirect(ip, nb, dbtofsb(fs, bp->b_blkno), dp->di_db[nb], fs->fs_bsize, osize, bp); ip->i_size = smalllblktosize(fs, nb + 1); dp->di_size = ip->i_size; dp->di_db[nb] = dbtofsb(fs, bp->b_blkno); ip->i_flag |= IN_CHANGE | IN_UPDATE; if (flags & IO_SYNC) bwrite(bp); else if (DOINGASYNC(vp)) bdwrite(bp); else bawrite(bp); } } /* * The first NDADDR blocks are direct blocks */ if (lbn < NDADDR) { if (flags & BA_METAONLY) panic("ffs_balloc_ufs1: BA_METAONLY for direct block"); nb = dp->di_db[lbn]; if (nb != 0 && ip->i_size >= smalllblktosize(fs, lbn + 1)) { error = bread(vp, lbn, fs->fs_bsize, NOCRED, &bp); if (error) { brelse(bp); return (error); } bp->b_blkno = fsbtodb(fs, nb); *bpp = bp; return (0); } if (nb != 0) { /* * Consider need to reallocate a fragment. */ osize = fragroundup(fs, blkoff(fs, ip->i_size)); nsize = fragroundup(fs, size); if (nsize <= osize) { error = bread(vp, lbn, osize, NOCRED, &bp); if (error) { brelse(bp); return (error); } bp->b_blkno = fsbtodb(fs, nb); } else { UFS_LOCK(ump); error = ffs_realloccg(ip, lbn, dp->di_db[lbn], ffs_blkpref_ufs1(ip, lbn, (int)lbn, &dp->di_db[0]), osize, nsize, flags, cred, &bp); if (error) return (error); if (DOINGSOFTDEP(vp)) softdep_setup_allocdirect(ip, lbn, dbtofsb(fs, bp->b_blkno), nb, nsize, osize, bp); } } else { if (ip->i_size < smalllblktosize(fs, lbn + 1)) nsize = fragroundup(fs, size); else nsize = fs->fs_bsize; UFS_LOCK(ump); error = ffs_alloc(ip, lbn, ffs_blkpref_ufs1(ip, lbn, (int)lbn, &dp->di_db[0]), nsize, flags, cred, &newb); if (error) return (error); bp = getblk(vp, lbn, nsize, 0, 0, gbflags); bp->b_blkno = fsbtodb(fs, newb); if (flags & BA_CLRBUF) vfs_bio_clrbuf(bp); if (DOINGSOFTDEP(vp)) softdep_setup_allocdirect(ip, lbn, newb, 0, nsize, 0, bp); } dp->di_db[lbn] = dbtofsb(fs, bp->b_blkno); ip->i_flag |= IN_CHANGE | IN_UPDATE; *bpp = bp; return (0); } /* * Determine the number of levels of indirection. */ pref = 0; if ((error = ufs_getlbns(vp, lbn, indirs, &num)) != 0) return(error); #ifdef INVARIANTS if (num < 1) panic ("ffs_balloc_ufs1: ufs_getlbns returned indirect block"); #endif saved_inbdflush = curthread_pflags_set(TDP_INBDFLUSH); /* * Fetch the first indirect block allocating if necessary. */ --num; nb = dp->di_ib[indirs[0].in_off]; allocib = NULL; allocblk = allociblk; lbns_remfree = lbns; if (nb == 0) { UFS_LOCK(ump); pref = ffs_blkpref_ufs1(ip, lbn, -indirs[0].in_off - 1, (ufs1_daddr_t *)0); if ((error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, flags, cred, &newb)) != 0) { curthread_pflags_restore(saved_inbdflush); return (error); } pref = newb + fs->fs_frag; nb = newb; MPASS(allocblk < allociblk + nitems(allociblk)); MPASS(lbns_remfree < lbns + nitems(lbns)); *allocblk++ = nb; *lbns_remfree++ = indirs[1].in_lbn; bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0, gbflags); bp->b_blkno = fsbtodb(fs, nb); vfs_bio_clrbuf(bp); if (DOINGSOFTDEP(vp)) { softdep_setup_allocdirect(ip, NDADDR + indirs[0].in_off, newb, 0, fs->fs_bsize, 0, bp); bdwrite(bp); } else if ((flags & IO_SYNC) == 0 && DOINGASYNC(vp)) { if (bp->b_bufsize == fs->fs_bsize) bp->b_flags |= B_CLUSTEROK; bdwrite(bp); } else { if ((error = bwrite(bp)) != 0) goto fail; } allocib = &dp->di_ib[indirs[0].in_off]; *allocib = nb; ip->i_flag |= IN_CHANGE | IN_UPDATE; } /* * Fetch through the indirect blocks, allocating as necessary. */ retry: for (i = 1;;) { error = bread(vp, indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp); if (error) { brelse(bp); goto fail; } bap = (ufs1_daddr_t *)bp->b_data; nb = bap[indirs[i].in_off]; if (i == num) break; i += 1; if (nb != 0) { bqrelse(bp); continue; } UFS_LOCK(ump); /* * If parent indirect has just been allocated, try to cluster * immediately following it. */ if (pref == 0) pref = ffs_blkpref_ufs1(ip, lbn, i - num - 1, (ufs1_daddr_t *)0); if ((error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, flags | IO_BUFLOCKED, cred, &newb)) != 0) { brelse(bp); if (DOINGSOFTDEP(vp) && ++reclaimed == 1) { UFS_LOCK(ump); softdep_request_cleanup(fs, vp, cred, FLUSH_BLOCKS_WAIT); UFS_UNLOCK(ump); goto retry; } if (ppsratecheck(&lastfail, &curfail, 1)) { ffs_fserr(fs, ip->i_number, "filesystem full"); uprintf("\n%s: write failed, filesystem " "is full\n", fs->fs_fsmnt); } goto fail; } pref = newb + fs->fs_frag; nb = newb; MPASS(allocblk < allociblk + nitems(allociblk)); MPASS(lbns_remfree < lbns + nitems(lbns)); *allocblk++ = nb; *lbns_remfree++ = indirs[i].in_lbn; nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0, 0); nbp->b_blkno = fsbtodb(fs, nb); vfs_bio_clrbuf(nbp); if (DOINGSOFTDEP(vp)) { softdep_setup_allocindir_meta(nbp, ip, bp, indirs[i - 1].in_off, nb); bdwrite(nbp); } else if ((flags & IO_SYNC) == 0 && DOINGASYNC(vp)) { if (nbp->b_bufsize == fs->fs_bsize) nbp->b_flags |= B_CLUSTEROK; bdwrite(nbp); } else { if ((error = bwrite(nbp)) != 0) { brelse(bp); goto fail; } } bap[indirs[i - 1].in_off] = nb; if (allocib == NULL && unwindidx < 0) unwindidx = i - 1; /* * If required, write synchronously, otherwise use * delayed write. */ if (flags & IO_SYNC) { bwrite(bp); } else { if (bp->b_bufsize == fs->fs_bsize) bp->b_flags |= B_CLUSTEROK; bdwrite(bp); } } /* * If asked only for the indirect block, then return it. */ if (flags & BA_METAONLY) { curthread_pflags_restore(saved_inbdflush); *bpp = bp; return (0); } /* * Get the data block, allocating if necessary. */ if (nb == 0) { UFS_LOCK(ump); /* * If allocating metadata at the front of the cylinder * group and parent indirect block has just been allocated, * then cluster next to it if it is the first indirect in * the file. Otherwise it has been allocated in the metadata * area, so we want to find our own place out in the data area. */ if (pref == 0 || (lbn > NDADDR && fs->fs_metaspace != 0)) pref = ffs_blkpref_ufs1(ip, lbn, indirs[i].in_off, &bap[0]); error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, flags | IO_BUFLOCKED, cred, &newb); if (error) { brelse(bp); if (DOINGSOFTDEP(vp) && ++reclaimed == 1) { UFS_LOCK(ump); softdep_request_cleanup(fs, vp, cred, FLUSH_BLOCKS_WAIT); UFS_UNLOCK(ump); goto retry; } if (ppsratecheck(&lastfail, &curfail, 1)) { ffs_fserr(fs, ip->i_number, "filesystem full"); uprintf("\n%s: write failed, filesystem " "is full\n", fs->fs_fsmnt); } goto fail; } nb = newb; MPASS(allocblk < allociblk + nitems(allociblk)); MPASS(lbns_remfree < lbns + nitems(lbns)); *allocblk++ = nb; *lbns_remfree++ = lbn; nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0, gbflags); nbp->b_blkno = fsbtodb(fs, nb); if (flags & BA_CLRBUF) vfs_bio_clrbuf(nbp); if (DOINGSOFTDEP(vp)) softdep_setup_allocindir_page(ip, lbn, bp, indirs[i].in_off, nb, 0, nbp); bap[indirs[i].in_off] = nb; /* * If required, write synchronously, otherwise use * delayed write. */ if (flags & IO_SYNC) { bwrite(bp); } else { if (bp->b_bufsize == fs->fs_bsize) bp->b_flags |= B_CLUSTEROK; bdwrite(bp); } curthread_pflags_restore(saved_inbdflush); *bpp = nbp; return (0); } brelse(bp); if (flags & BA_CLRBUF) { int seqcount = (flags & BA_SEQMASK) >> BA_SEQSHIFT; if (seqcount != 0 && (vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0 && !(vm_page_count_severe() || buf_dirty_count_severe())) { error = cluster_read(vp, ip->i_size, lbn, (int)fs->fs_bsize, NOCRED, MAXBSIZE, seqcount, gbflags, &nbp); } else { error = bread_gb(vp, lbn, (int)fs->fs_bsize, NOCRED, gbflags, &nbp); } if (error) { brelse(nbp); goto fail; } } else {