1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc. 4 * All Rights Reserved. 5 */ 6 #include "xfs.h" 7 #include "xfs_fs.h" 8 #include "xfs_shared.h" 9 #include "xfs_format.h" 10 #include "xfs_log_format.h" 11 #include "xfs_trans_resv.h" 12 #include "xfs_mount.h" 13 #include "xfs_inode.h" 14 #include "xfs_btree.h" 15 #include "xfs_ialloc.h" 16 #include "xfs_ialloc_btree.h" 17 #include "xfs_iwalk.h" 18 #include "xfs_itable.h" 19 #include "xfs_error.h" 20 #include "xfs_icache.h" 21 #include "xfs_health.h" 22 #include "xfs_trans.h" 23 24 /* 25 * Bulk Stat 26 * ========= 27 * 28 * Use the inode walking functions to fill out struct xfs_bulkstat for every 29 * allocated inode, then pass the stat information to some externally provided 30 * iteration function. 31 */ 32 33 struct xfs_bstat_chunk { 34 bulkstat_one_fmt_pf formatter; 35 struct xfs_ibulk *breq; 36 struct xfs_bulkstat *buf; 37 }; 38 39 /* 40 * Fill out the bulkstat info for a single inode and report it somewhere. 41 * 42 * bc->breq->lastino is effectively the inode cursor as we walk through the 43 * filesystem. Therefore, we update it any time we need to move the cursor 44 * forward, regardless of whether or not we're sending any bstat information 45 * back to userspace. If the inode is internal metadata or, has been freed 46 * out from under us, we just simply keep going. 47 * 48 * However, if any other type of error happens we want to stop right where we 49 * are so that userspace will call back with exact number of the bad inode and 50 * we can send back an error code. 51 * 52 * Note that if the formatter tells us there's no space left in the buffer we 53 * move the cursor forward and abort the walk. 54 */ 55 STATIC int 56 xfs_bulkstat_one_int( 57 struct xfs_mount *mp, 58 struct user_namespace *mnt_userns, 59 struct xfs_trans *tp, 60 xfs_ino_t ino, 61 struct xfs_bstat_chunk *bc) 62 { 63 struct user_namespace *sb_userns = mp->m_super->s_user_ns; 64 struct xfs_inode *ip; /* incore inode pointer */ 65 struct inode *inode; 66 struct xfs_bulkstat *buf = bc->buf; 67 int error = -EINVAL; 68 69 if (xfs_internal_inum(mp, ino)) 70 goto out_advance; 71 72 error = xfs_iget(mp, tp, ino, 73 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED), 74 XFS_ILOCK_SHARED, &ip); 75 if (error == -ENOENT || error == -EINVAL) 76 goto out_advance; 77 if (error) 78 goto out; 79 80 ASSERT(ip != NULL); 81 ASSERT(ip->i_imap.im_blkno != 0); 82 inode = VFS_I(ip); 83 84 /* xfs_iget returns the following without needing 85 * further change. 86 */ 87 buf->bs_projectid = ip->i_projid; 88 buf->bs_ino = ino; 89 buf->bs_uid = from_kuid(sb_userns, i_uid_into_mnt(mnt_userns, inode)); 90 buf->bs_gid = from_kgid(sb_userns, i_gid_into_mnt(mnt_userns, inode)); 91 buf->bs_size = ip->i_disk_size; 92 93 buf->bs_nlink = inode->i_nlink; 94 buf->bs_atime = inode->i_atime.tv_sec; 95 buf->bs_atime_nsec = inode->i_atime.tv_nsec; 96 buf->bs_mtime = inode->i_mtime.tv_sec; 97 buf->bs_mtime_nsec = inode->i_mtime.tv_nsec; 98 buf->bs_ctime = inode->i_ctime.tv_sec; 99 buf->bs_ctime_nsec = inode->i_ctime.tv_nsec; 100 buf->bs_gen = inode->i_generation; 101 buf->bs_mode = inode->i_mode; 102 103 buf->bs_xflags = xfs_ip2xflags(ip); 104 buf->bs_extsize_blks = ip->i_extsize; 105 buf->bs_extents = xfs_ifork_nextents(&ip->i_df); 106 xfs_bulkstat_health(ip, buf); 107 buf->bs_aextents = xfs_ifork_nextents(ip->i_afp); 108 buf->bs_forkoff = XFS_IFORK_BOFF(ip); 109 buf->bs_version = XFS_BULKSTAT_VERSION_V5; 110 111 if (xfs_has_v3inodes(mp)) { 112 buf->bs_btime = ip->i_crtime.tv_sec; 113 buf->bs_btime_nsec = ip->i_crtime.tv_nsec; 114 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE) 115 buf->bs_cowextsize_blks = ip->i_cowextsize; 116 } 117 118 switch (ip->i_df.if_format) { 119 case XFS_DINODE_FMT_DEV: 120 buf->bs_rdev = sysv_encode_dev(inode->i_rdev); 121 buf->bs_blksize = BLKDEV_IOSIZE; 122 buf->bs_blocks = 0; 123 break; 124 case XFS_DINODE_FMT_LOCAL: 125 buf->bs_rdev = 0; 126 buf->bs_blksize = mp->m_sb.sb_blocksize; 127 buf->bs_blocks = 0; 128 break; 129 case XFS_DINODE_FMT_EXTENTS: 130 case XFS_DINODE_FMT_BTREE: 131 buf->bs_rdev = 0; 132 buf->bs_blksize = mp->m_sb.sb_blocksize; 133 buf->bs_blocks = ip->i_nblocks + ip->i_delayed_blks; 134 break; 135 } 136 xfs_iunlock(ip, XFS_ILOCK_SHARED); 137 xfs_irele(ip); 138 139 error = bc->formatter(bc->breq, buf); 140 if (error == -ECANCELED) 141 goto out_advance; 142 if (error) 143 goto out; 144 145 out_advance: 146 /* 147 * Advance the cursor to the inode that comes after the one we just 148 * looked at. We want the caller to move along if the bulkstat 149 * information was copied successfully; if we tried to grab the inode 150 * but it's no longer allocated; or if it's internal metadata. 151 */ 152 bc->breq->startino = ino + 1; 153 out: 154 return error; 155 } 156 157 /* Bulkstat a single inode. */ 158 int 159 xfs_bulkstat_one( 160 struct xfs_ibulk *breq, 161 bulkstat_one_fmt_pf formatter) 162 { 163 struct xfs_bstat_chunk bc = { 164 .formatter = formatter, 165 .breq = breq, 166 }; 167 struct xfs_trans *tp; 168 int error; 169 170 if (breq->mnt_userns != &init_user_ns) { 171 xfs_warn_ratelimited(breq->mp, 172 "bulkstat not supported inside of idmapped mounts."); 173 return -EINVAL; 174 } 175 176 ASSERT(breq->icount == 1); 177 178 bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat), 179 KM_MAYFAIL); 180 if (!bc.buf) 181 return -ENOMEM; 182 183 /* 184 * Grab an empty transaction so that we can use its recursive buffer 185 * locking abilities to detect cycles in the inobt without deadlocking. 186 */ 187 error = xfs_trans_alloc_empty(breq->mp, &tp); 188 if (error) 189 goto out; 190 191 error = xfs_bulkstat_one_int(breq->mp, breq->mnt_userns, tp, 192 breq->startino, &bc); 193 xfs_trans_cancel(tp); 194 out: 195 kmem_free(bc.buf); 196 197 /* 198 * If we reported one inode to userspace then we abort because we hit 199 * the end of the buffer. Don't leak that back to userspace. 200 */ 201 if (error == -ECANCELED) 202 error = 0; 203 204 return error; 205 } 206 207 static int 208 xfs_bulkstat_iwalk( 209 struct xfs_mount *mp, 210 struct xfs_trans *tp, 211 xfs_ino_t ino, 212 void *data) 213 { 214 struct xfs_bstat_chunk *bc = data; 215 int error; 216 217 error = xfs_bulkstat_one_int(mp, bc->breq->mnt_userns, tp, ino, data); 218 /* bulkstat just skips over missing inodes */ 219 if (error == -ENOENT || error == -EINVAL) 220 return 0; 221 return error; 222 } 223 224 /* 225 * Check the incoming lastino parameter. 226 * 227 * We allow any inode value that could map to physical space inside the 228 * filesystem because if there are no inodes there, bulkstat moves on to the 229 * next chunk. In other words, the magic agino value of zero takes us to the 230 * first chunk in the AG, and an agino value past the end of the AG takes us to 231 * the first chunk in the next AG. 232 * 233 * Therefore we can end early if the requested inode is beyond the end of the 234 * filesystem or doesn't map properly. 235 */ 236 static inline bool 237 xfs_bulkstat_already_done( 238 struct xfs_mount *mp, 239 xfs_ino_t startino) 240 { 241 xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, startino); 242 xfs_agino_t agino = XFS_INO_TO_AGINO(mp, startino); 243 244 return agno >= mp->m_sb.sb_agcount || 245 startino != XFS_AGINO_TO_INO(mp, agno, agino); 246 } 247 248 /* Return stat information in bulk (by-inode) for the filesystem. */ 249 int 250 xfs_bulkstat( 251 struct xfs_ibulk *breq, 252 bulkstat_one_fmt_pf formatter) 253 { 254 struct xfs_bstat_chunk bc = { 255 .formatter = formatter, 256 .breq = breq, 257 }; 258 struct xfs_trans *tp; 259 int error; 260 261 if (breq->mnt_userns != &init_user_ns) { 262 xfs_warn_ratelimited(breq->mp, 263 "bulkstat not supported inside of idmapped mounts."); 264 return -EINVAL; 265 } 266 if (xfs_bulkstat_already_done(breq->mp, breq->startino)) 267 return 0; 268 269 bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat), 270 KM_MAYFAIL); 271 if (!bc.buf) 272 return -ENOMEM; 273 274 /* 275 * Grab an empty transaction so that we can use its recursive buffer 276 * locking abilities to detect cycles in the inobt without deadlocking. 277 */ 278 error = xfs_trans_alloc_empty(breq->mp, &tp); 279 if (error) 280 goto out; 281 282 error = xfs_iwalk(breq->mp, tp, breq->startino, breq->flags, 283 xfs_bulkstat_iwalk, breq->icount, &bc); 284 xfs_trans_cancel(tp); 285 out: 286 kmem_free(bc.buf); 287 288 /* 289 * We found some inodes, so clear the error status and return them. 290 * The lastino pointer will point directly at the inode that triggered 291 * any error that occurred, so on the next call the error will be 292 * triggered again and propagated to userspace as there will be no 293 * formatted inodes in the buffer. 294 */ 295 if (breq->ocount > 0) 296 error = 0; 297 298 return error; 299 } 300 301 /* Convert bulkstat (v5) to bstat (v1). */ 302 void 303 xfs_bulkstat_to_bstat( 304 struct xfs_mount *mp, 305 struct xfs_bstat *bs1, 306 const struct xfs_bulkstat *bstat) 307 { 308 /* memset is needed here because of padding holes in the structure. */ 309 memset(bs1, 0, sizeof(struct xfs_bstat)); 310 bs1->bs_ino = bstat->bs_ino; 311 bs1->bs_mode = bstat->bs_mode; 312 bs1->bs_nlink = bstat->bs_nlink; 313 bs1->bs_uid = bstat->bs_uid; 314 bs1->bs_gid = bstat->bs_gid; 315 bs1->bs_rdev = bstat->bs_rdev; 316 bs1->bs_blksize = bstat->bs_blksize; 317 bs1->bs_size = bstat->bs_size; 318 bs1->bs_atime.tv_sec = bstat->bs_atime; 319 bs1->bs_mtime.tv_sec = bstat->bs_mtime; 320 bs1->bs_ctime.tv_sec = bstat->bs_ctime; 321 bs1->bs_atime.tv_nsec = bstat->bs_atime_nsec; 322 bs1->bs_mtime.tv_nsec = bstat->bs_mtime_nsec; 323 bs1->bs_ctime.tv_nsec = bstat->bs_ctime_nsec; 324 bs1->bs_blocks = bstat->bs_blocks; 325 bs1->bs_xflags = bstat->bs_xflags; 326 bs1->bs_extsize = XFS_FSB_TO_B(mp, bstat->bs_extsize_blks); 327 bs1->bs_extents = bstat->bs_extents; 328 bs1->bs_gen = bstat->bs_gen; 329 bs1->bs_projid_lo = bstat->bs_projectid & 0xFFFF; 330 bs1->bs_forkoff = bstat->bs_forkoff; 331 bs1->bs_projid_hi = bstat->bs_projectid >> 16; 332 bs1->bs_sick = bstat->bs_sick; 333 bs1->bs_checked = bstat->bs_checked; 334 bs1->bs_cowextsize = XFS_FSB_TO_B(mp, bstat->bs_cowextsize_blks); 335 bs1->bs_dmevmask = 0; 336 bs1->bs_dmstate = 0; 337 bs1->bs_aextents = bstat->bs_aextents; 338 } 339 340 struct xfs_inumbers_chunk { 341 inumbers_fmt_pf formatter; 342 struct xfs_ibulk *breq; 343 }; 344 345 /* 346 * INUMBERS 347 * ======== 348 * This is how we export inode btree records to userspace, so that XFS tools 349 * can figure out where inodes are allocated. 350 */ 351 352 /* 353 * Format the inode group structure and report it somewhere. 354 * 355 * Similar to xfs_bulkstat_one_int, lastino is the inode cursor as we walk 356 * through the filesystem so we move it forward unless there was a runtime 357 * error. If the formatter tells us the buffer is now full we also move the 358 * cursor forward and abort the walk. 359 */ 360 STATIC int 361 xfs_inumbers_walk( 362 struct xfs_mount *mp, 363 struct xfs_trans *tp, 364 xfs_agnumber_t agno, 365 const struct xfs_inobt_rec_incore *irec, 366 void *data) 367 { 368 struct xfs_inumbers inogrp = { 369 .xi_startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino), 370 .xi_alloccount = irec->ir_count - irec->ir_freecount, 371 .xi_allocmask = ~irec->ir_free, 372 .xi_version = XFS_INUMBERS_VERSION_V5, 373 }; 374 struct xfs_inumbers_chunk *ic = data; 375 int error; 376 377 error = ic->formatter(ic->breq, &inogrp); 378 if (error && error != -ECANCELED) 379 return error; 380 381 ic->breq->startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino) + 382 XFS_INODES_PER_CHUNK; 383 return error; 384 } 385 386 /* 387 * Return inode number table for the filesystem. 388 */ 389 int 390 xfs_inumbers( 391 struct xfs_ibulk *breq, 392 inumbers_fmt_pf formatter) 393 { 394 struct xfs_inumbers_chunk ic = { 395 .formatter = formatter, 396 .breq = breq, 397 }; 398 struct xfs_trans *tp; 399 int error = 0; 400 401 if (xfs_bulkstat_already_done(breq->mp, breq->startino)) 402 return 0; 403 404 /* 405 * Grab an empty transaction so that we can use its recursive buffer 406 * locking abilities to detect cycles in the inobt without deadlocking. 407 */ 408 error = xfs_trans_alloc_empty(breq->mp, &tp); 409 if (error) 410 goto out; 411 412 error = xfs_inobt_walk(breq->mp, tp, breq->startino, breq->flags, 413 xfs_inumbers_walk, breq->icount, &ic); 414 xfs_trans_cancel(tp); 415 out: 416 417 /* 418 * We found some inode groups, so clear the error status and return 419 * them. The lastino pointer will point directly at the inode that 420 * triggered any error that occurred, so on the next call the error 421 * will be triggered again and propagated to userspace as there will be 422 * no formatted inode groups in the buffer. 423 */ 424 if (breq->ocount > 0) 425 error = 0; 426 427 return error; 428 } 429 430 /* Convert an inumbers (v5) struct to a inogrp (v1) struct. */ 431 void 432 xfs_inumbers_to_inogrp( 433 struct xfs_inogrp *ig1, 434 const struct xfs_inumbers *ig) 435 { 436 /* memset is needed here because of padding holes in the structure. */ 437 memset(ig1, 0, sizeof(struct xfs_inogrp)); 438 ig1->xi_startino = ig->xi_startino; 439 ig1->xi_alloccount = ig->xi_alloccount; 440 ig1->xi_allocmask = ig->xi_allocmask; 441 } 442