Darrick J. Wong | c2fc338 | 2017-10-17 21:37:47 -0700 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (C) 2017 Oracle. All Rights Reserved. |
| 3 | * |
| 4 | * Author: Darrick J. Wong <darrick.wong@oracle.com> |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or |
| 7 | * modify it under the terms of the GNU General Public License |
| 8 | * as published by the Free Software Foundation; either version 2 |
| 9 | * of the License, or (at your option) any later version. |
| 10 | * |
| 11 | * This program is distributed in the hope that it would be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 14 | * GNU General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU General Public License |
| 17 | * along with this program; if not, write the Free Software Foundation, |
| 18 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA. |
| 19 | */ |
| 20 | #include "xfs.h" |
| 21 | #include "xfs_fs.h" |
| 22 | #include "xfs_shared.h" |
| 23 | #include "xfs_format.h" |
| 24 | #include "xfs_trans_resv.h" |
| 25 | #include "xfs_mount.h" |
| 26 | #include "xfs_defer.h" |
| 27 | #include "xfs_btree.h" |
| 28 | #include "xfs_bit.h" |
| 29 | #include "xfs_log_format.h" |
| 30 | #include "xfs_trans.h" |
| 31 | #include "xfs_sb.h" |
| 32 | #include "xfs_inode.h" |
| 33 | #include "xfs_inode_fork.h" |
| 34 | #include "xfs_alloc.h" |
| 35 | #include "xfs_bmap.h" |
| 36 | #include "xfs_quota.h" |
| 37 | #include "xfs_qm.h" |
| 38 | #include "xfs_dquot.h" |
| 39 | #include "xfs_dquot_item.h" |
| 40 | #include "scrub/xfs_scrub.h" |
| 41 | #include "scrub/scrub.h" |
| 42 | #include "scrub/common.h" |
| 43 | #include "scrub/trace.h" |
| 44 | |
| 45 | /* Convert a scrub type code to a DQ flag, or return 0 if error. */ |
| 46 | static inline uint |
| 47 | xfs_scrub_quota_to_dqtype( |
| 48 | struct xfs_scrub_context *sc) |
| 49 | { |
| 50 | switch (sc->sm->sm_type) { |
| 51 | case XFS_SCRUB_TYPE_UQUOTA: |
| 52 | return XFS_DQ_USER; |
| 53 | case XFS_SCRUB_TYPE_GQUOTA: |
| 54 | return XFS_DQ_GROUP; |
| 55 | case XFS_SCRUB_TYPE_PQUOTA: |
| 56 | return XFS_DQ_PROJ; |
| 57 | default: |
| 58 | return 0; |
| 59 | } |
| 60 | } |
| 61 | |
| 62 | /* Set us up to scrub a quota. */ |
| 63 | int |
| 64 | xfs_scrub_setup_quota( |
| 65 | struct xfs_scrub_context *sc, |
| 66 | struct xfs_inode *ip) |
| 67 | { |
| 68 | uint dqtype; |
| 69 | |
| 70 | /* |
| 71 | * If userspace gave us an AG number or inode data, they don't |
| 72 | * know what they're doing. Get out. |
| 73 | */ |
| 74 | if (sc->sm->sm_agno || sc->sm->sm_ino || sc->sm->sm_gen) |
| 75 | return -EINVAL; |
| 76 | |
| 77 | dqtype = xfs_scrub_quota_to_dqtype(sc); |
| 78 | if (dqtype == 0) |
| 79 | return -EINVAL; |
| 80 | if (!xfs_this_quota_on(sc->mp, dqtype)) |
| 81 | return -ENOENT; |
| 82 | return 0; |
| 83 | } |
| 84 | |
| 85 | /* Quotas. */ |
| 86 | |
| 87 | /* Scrub the fields in an individual quota item. */ |
| 88 | STATIC void |
| 89 | xfs_scrub_quota_item( |
| 90 | struct xfs_scrub_context *sc, |
| 91 | uint dqtype, |
| 92 | struct xfs_dquot *dq, |
| 93 | xfs_dqid_t id) |
| 94 | { |
| 95 | struct xfs_mount *mp = sc->mp; |
| 96 | struct xfs_disk_dquot *d = &dq->q_core; |
| 97 | struct xfs_quotainfo *qi = mp->m_quotainfo; |
| 98 | xfs_fileoff_t offset; |
| 99 | unsigned long long bsoft; |
| 100 | unsigned long long isoft; |
| 101 | unsigned long long rsoft; |
| 102 | unsigned long long bhard; |
| 103 | unsigned long long ihard; |
| 104 | unsigned long long rhard; |
| 105 | unsigned long long bcount; |
| 106 | unsigned long long icount; |
| 107 | unsigned long long rcount; |
| 108 | xfs_ino_t fs_icount; |
| 109 | |
| 110 | offset = id * qi->qi_dqperchunk; |
| 111 | |
| 112 | /* |
| 113 | * We fed $id and DQNEXT into the xfs_qm_dqget call, which means |
| 114 | * that the actual dquot we got must either have the same id or |
| 115 | * the next higher id. |
| 116 | */ |
| 117 | if (id > be32_to_cpu(d->d_id)) |
| 118 | xfs_scrub_fblock_set_corrupt(sc, XFS_DATA_FORK, offset); |
| 119 | |
| 120 | /* Did we get the dquot type we wanted? */ |
| 121 | if (dqtype != (d->d_flags & XFS_DQ_ALLTYPES)) |
| 122 | xfs_scrub_fblock_set_corrupt(sc, XFS_DATA_FORK, offset); |
| 123 | |
| 124 | if (d->d_pad0 != cpu_to_be32(0) || d->d_pad != cpu_to_be16(0)) |
| 125 | xfs_scrub_fblock_set_corrupt(sc, XFS_DATA_FORK, offset); |
| 126 | |
| 127 | /* Check the limits. */ |
| 128 | bhard = be64_to_cpu(d->d_blk_hardlimit); |
| 129 | ihard = be64_to_cpu(d->d_ino_hardlimit); |
| 130 | rhard = be64_to_cpu(d->d_rtb_hardlimit); |
| 131 | |
| 132 | bsoft = be64_to_cpu(d->d_blk_softlimit); |
| 133 | isoft = be64_to_cpu(d->d_ino_softlimit); |
| 134 | rsoft = be64_to_cpu(d->d_rtb_softlimit); |
| 135 | |
| 136 | /* |
| 137 | * Warn if the hard limits are larger than the fs. |
| 138 | * Administrators can do this, though in production this seems |
| 139 | * suspect, which is why we flag it for review. |
| 140 | * |
| 141 | * Complain about corruption if the soft limit is greater than |
| 142 | * the hard limit. |
| 143 | */ |
| 144 | if (bhard > mp->m_sb.sb_dblocks) |
| 145 | xfs_scrub_fblock_set_warning(sc, XFS_DATA_FORK, offset); |
| 146 | if (bsoft > bhard) |
| 147 | xfs_scrub_fblock_set_corrupt(sc, XFS_DATA_FORK, offset); |
| 148 | |
| 149 | if (ihard > mp->m_maxicount) |
| 150 | xfs_scrub_fblock_set_warning(sc, XFS_DATA_FORK, offset); |
| 151 | if (isoft > ihard) |
| 152 | xfs_scrub_fblock_set_corrupt(sc, XFS_DATA_FORK, offset); |
| 153 | |
| 154 | if (rhard > mp->m_sb.sb_rblocks) |
| 155 | xfs_scrub_fblock_set_warning(sc, XFS_DATA_FORK, offset); |
| 156 | if (rsoft > rhard) |
| 157 | xfs_scrub_fblock_set_corrupt(sc, XFS_DATA_FORK, offset); |
| 158 | |
| 159 | /* Check the resource counts. */ |
| 160 | bcount = be64_to_cpu(d->d_bcount); |
| 161 | icount = be64_to_cpu(d->d_icount); |
| 162 | rcount = be64_to_cpu(d->d_rtbcount); |
| 163 | fs_icount = percpu_counter_sum(&mp->m_icount); |
| 164 | |
| 165 | /* |
| 166 | * Check that usage doesn't exceed physical limits. However, on |
| 167 | * a reflink filesystem we're allowed to exceed physical space |
| 168 | * if there are no quota limits. |
| 169 | */ |
| 170 | if (xfs_sb_version_hasreflink(&mp->m_sb)) { |
| 171 | if (mp->m_sb.sb_dblocks < bcount) |
| 172 | xfs_scrub_fblock_set_warning(sc, XFS_DATA_FORK, |
| 173 | offset); |
| 174 | } else { |
| 175 | if (mp->m_sb.sb_dblocks < bcount) |
| 176 | xfs_scrub_fblock_set_corrupt(sc, XFS_DATA_FORK, |
| 177 | offset); |
| 178 | } |
| 179 | if (icount > fs_icount || rcount > mp->m_sb.sb_rblocks) |
| 180 | xfs_scrub_fblock_set_corrupt(sc, XFS_DATA_FORK, offset); |
| 181 | |
| 182 | /* |
| 183 | * We can violate the hard limits if the admin suddenly sets a |
| 184 | * lower limit than the actual usage. However, we flag it for |
| 185 | * admin review. |
| 186 | */ |
| 187 | if (id != 0 && bhard != 0 && bcount > bhard) |
| 188 | xfs_scrub_fblock_set_warning(sc, XFS_DATA_FORK, offset); |
| 189 | if (id != 0 && ihard != 0 && icount > ihard) |
| 190 | xfs_scrub_fblock_set_warning(sc, XFS_DATA_FORK, offset); |
| 191 | if (id != 0 && rhard != 0 && rcount > rhard) |
| 192 | xfs_scrub_fblock_set_warning(sc, XFS_DATA_FORK, offset); |
| 193 | } |
| 194 | |
| 195 | /* Scrub all of a quota type's items. */ |
| 196 | int |
| 197 | xfs_scrub_quota( |
| 198 | struct xfs_scrub_context *sc) |
| 199 | { |
| 200 | struct xfs_bmbt_irec irec = { 0 }; |
| 201 | struct xfs_mount *mp = sc->mp; |
| 202 | struct xfs_inode *ip; |
| 203 | struct xfs_quotainfo *qi = mp->m_quotainfo; |
| 204 | struct xfs_dquot *dq; |
| 205 | xfs_fileoff_t max_dqid_off; |
| 206 | xfs_fileoff_t off = 0; |
| 207 | xfs_dqid_t id = 0; |
| 208 | uint dqtype; |
| 209 | int nimaps; |
| 210 | int error; |
| 211 | |
| 212 | if (!XFS_IS_QUOTA_RUNNING(mp) || !XFS_IS_QUOTA_ON(mp)) |
| 213 | return -ENOENT; |
| 214 | |
| 215 | mutex_lock(&qi->qi_quotaofflock); |
| 216 | dqtype = xfs_scrub_quota_to_dqtype(sc); |
| 217 | if (!xfs_this_quota_on(sc->mp, dqtype)) { |
| 218 | error = -ENOENT; |
| 219 | goto out_unlock_quota; |
| 220 | } |
| 221 | |
| 222 | /* Attach to the quota inode and set sc->ip so that reporting works. */ |
| 223 | ip = xfs_quota_inode(sc->mp, dqtype); |
| 224 | sc->ip = ip; |
| 225 | |
| 226 | /* Look for problem extents. */ |
| 227 | xfs_ilock(ip, XFS_ILOCK_EXCL); |
| 228 | if (ip->i_d.di_flags & XFS_DIFLAG_REALTIME) { |
| 229 | xfs_scrub_ino_set_corrupt(sc, sc->ip->i_ino, NULL); |
| 230 | goto out_unlock_inode; |
| 231 | } |
| 232 | max_dqid_off = ((xfs_dqid_t)-1) / qi->qi_dqperchunk; |
| 233 | while (1) { |
| 234 | if (xfs_scrub_should_terminate(sc, &error)) |
| 235 | break; |
| 236 | |
| 237 | off = irec.br_startoff + irec.br_blockcount; |
| 238 | nimaps = 1; |
| 239 | error = xfs_bmapi_read(ip, off, -1, &irec, &nimaps, |
| 240 | XFS_BMAPI_ENTIRE); |
| 241 | if (!xfs_scrub_fblock_process_error(sc, XFS_DATA_FORK, off, |
| 242 | &error)) |
| 243 | goto out_unlock_inode; |
| 244 | if (!nimaps) |
| 245 | break; |
| 246 | if (irec.br_startblock == HOLESTARTBLOCK) |
| 247 | continue; |
| 248 | |
| 249 | /* Check the extent record doesn't point to crap. */ |
| 250 | if (irec.br_startblock + irec.br_blockcount <= |
| 251 | irec.br_startblock) |
| 252 | xfs_scrub_fblock_set_corrupt(sc, XFS_DATA_FORK, |
| 253 | irec.br_startoff); |
| 254 | if (!xfs_verify_fsbno(mp, irec.br_startblock) || |
| 255 | !xfs_verify_fsbno(mp, irec.br_startblock + |
| 256 | irec.br_blockcount - 1)) |
| 257 | xfs_scrub_fblock_set_corrupt(sc, XFS_DATA_FORK, |
| 258 | irec.br_startoff); |
| 259 | |
| 260 | /* |
| 261 | * Unwritten extents or blocks mapped above the highest |
| 262 | * quota id shouldn't happen. |
| 263 | */ |
| 264 | if (isnullstartblock(irec.br_startblock) || |
| 265 | irec.br_startoff > max_dqid_off || |
| 266 | irec.br_startoff + irec.br_blockcount > max_dqid_off + 1) |
| 267 | xfs_scrub_fblock_set_corrupt(sc, XFS_DATA_FORK, off); |
| 268 | } |
| 269 | xfs_iunlock(ip, XFS_ILOCK_EXCL); |
| 270 | if (sc->sm->sm_flags & XFS_SCRUB_OFLAG_CORRUPT) |
| 271 | goto out; |
| 272 | |
| 273 | /* Check all the quota items. */ |
| 274 | while (id < ((xfs_dqid_t)-1ULL)) { |
| 275 | if (xfs_scrub_should_terminate(sc, &error)) |
| 276 | break; |
| 277 | |
| 278 | error = xfs_qm_dqget(mp, NULL, id, dqtype, XFS_QMOPT_DQNEXT, |
| 279 | &dq); |
| 280 | if (error == -ENOENT) |
| 281 | break; |
| 282 | if (!xfs_scrub_fblock_process_error(sc, XFS_DATA_FORK, |
| 283 | id * qi->qi_dqperchunk, &error)) |
| 284 | break; |
| 285 | |
| 286 | xfs_scrub_quota_item(sc, dqtype, dq, id); |
| 287 | |
| 288 | id = be32_to_cpu(dq->q_core.d_id) + 1; |
| 289 | xfs_qm_dqput(dq); |
| 290 | if (!id) |
| 291 | break; |
| 292 | } |
| 293 | |
| 294 | out: |
| 295 | /* We set sc->ip earlier, so make sure we clear it now. */ |
| 296 | sc->ip = NULL; |
| 297 | out_unlock_quota: |
| 298 | mutex_unlock(&qi->qi_quotaofflock); |
| 299 | return error; |
| 300 | |
| 301 | out_unlock_inode: |
| 302 | xfs_iunlock(ip, XFS_ILOCK_EXCL); |
| 303 | goto out; |
| 304 | } |