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 23 /* 24 * Bulk Stat 25 * ========= 26 * 27 * Use the inode walking functions to fill out struct xfs_bulkstat for every 28 * allocated inode, then pass the stat information to some externally provided 29 * iteration function. 30 */ 31 32 struct xfs_bstat_chunk { 33 bulkstat_one_fmt_pf formatter; 34 struct xfs_ibulk *breq; 35 struct xfs_bulkstat *buf; 36 }; 37 38 /* 39 * Fill out the bulkstat info for a single inode and report it somewhere. 40 * 41 * bc->breq->lastino is effectively the inode cursor as we walk through the 42 * filesystem. Therefore, we update it any time we need to move the cursor 43 * forward, regardless of whether or not we're sending any bstat information 44 * back to userspace. If the inode is internal metadata or, has been freed 45 * out from under us, we just simply keep going. 46 * 47 * However, if any other type of error happens we want to stop right where we 48 * are so that userspace will call back with exact number of the bad inode and 49 * we can send back an error code. 50 * 51 * Note that if the formatter tells us there's no space left in the buffer we 52 * move the cursor forward and abort the walk. 53 */ 54 STATIC int 55 xfs_bulkstat_one_int( 56 struct xfs_mount *mp, 57 struct user_namespace *mnt_userns, 58 struct xfs_trans *tp, 59 xfs_ino_t ino, 60 struct xfs_bstat_chunk *bc) 61 { 62 struct user_namespace *sb_userns = mp->m_super->s_user_ns; 63 struct xfs_icdinode *dic; /* dinode core info pointer */ 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 dic = &ip->i_d; 85 86 /* xfs_iget returns the following without needing 87 * further change. 88 */ 89 buf->bs_projectid = ip->i_d.di_projid; 90 buf->bs_ino = ino; 91 buf->bs_uid = from_kuid(sb_userns, i_uid_into_mnt(mnt_userns, inode)); 92 buf->bs_gid = from_kgid(sb_userns, i_gid_into_mnt(mnt_userns, inode)); 93 buf->bs_size = dic->di_size; 94 95 buf->bs_nlink = inode->i_nlink; 96 buf->bs_atime = inode->i_atime.tv_sec; 97 buf->bs_atime_nsec = inode->i_atime.tv_nsec; 98 buf->bs_mtime = inode->i_mtime.tv_sec; 99 buf->bs_mtime_nsec = inode->i_mtime.tv_nsec; 100 buf->bs_ctime = inode->i_ctime.tv_sec; 101 buf->bs_ctime_nsec = inode->i_ctime.tv_nsec; 102 buf->bs_btime = dic->di_crtime.tv_sec; 103 buf->bs_btime_nsec = dic->di_crtime.tv_nsec; 104 buf->bs_gen = inode->i_generation; 105 buf->bs_mode = inode->i_mode; 106 107 buf->bs_xflags = xfs_ip2xflags(ip); 108 buf->bs_extsize_blks = dic->di_extsize; 109 buf->bs_extents = xfs_ifork_nextents(&ip->i_df); 110 xfs_bulkstat_health(ip, buf); 111 buf->bs_aextents = xfs_ifork_nextents(ip->i_afp); 112 buf->bs_forkoff = XFS_IFORK_BOFF(ip); 113 buf->bs_version = XFS_BULKSTAT_VERSION_V5; 114 115 if (xfs_sb_version_has_v3inode(&mp->m_sb)) { 116 if (dic->di_flags2 & XFS_DIFLAG2_COWEXTSIZE) 117 buf->bs_cowextsize_blks = dic->di_cowextsize; 118 } 119 120 switch (ip->i_df.if_format) { 121 case XFS_DINODE_FMT_DEV: 122 buf->bs_rdev = sysv_encode_dev(inode->i_rdev); 123 buf->bs_blksize = BLKDEV_IOSIZE; 124 buf->bs_blocks = 0; 125 break; 126 case XFS_DINODE_FMT_LOCAL: 127 buf->bs_rdev = 0; 128 buf->bs_blksize = mp->m_sb.sb_blocksize; 129 buf->bs_blocks = 0; 130 break; 131 case XFS_DINODE_FMT_EXTENTS: 132 case XFS_DINODE_FMT_BTREE: 133 buf->bs_rdev = 0; 134 buf->bs_blksize = mp->m_sb.sb_blocksize; 135 buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks; 136 break; 137 } 138 xfs_iunlock(ip, XFS_ILOCK_SHARED); 139 xfs_irele(ip); 140 141 error = bc->formatter(bc->breq, buf); 142 if (error == -ECANCELED) 143 goto out_advance; 144 if (error) 145 goto out; 146 147 out_advance: 148 /* 149 * Advance the cursor to the inode that comes after the one we just 150 * looked at. We want the caller to move along if the bulkstat 151 * information was copied successfully; if we tried to grab the inode 152 * but it's no longer allocated; or if it's internal metadata. 153 */ 154 bc->breq->startino = ino + 1; 155 out: 156 return error; 157 } 158 159 /* Bulkstat a single inode. */ 160 int 161 xfs_bulkstat_one( 162 struct xfs_ibulk *breq, 163 bulkstat_one_fmt_pf formatter) 164 { 165 struct xfs_bstat_chunk bc = { 166 .formatter = formatter, 167 .breq = breq, 168 }; 169 int error; 170 171 ASSERT(breq->icount == 1); 172 173 bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat), 174 KM_MAYFAIL); 175 if (!bc.buf) 176 return -ENOMEM; 177 178 error = xfs_bulkstat_one_int(breq->mp, breq->mnt_userns, NULL, 179 breq->startino, &bc); 180 181 kmem_free(bc.buf); 182 183 /* 184 * If we reported one inode to userspace then we abort because we hit 185 * the end of the buffer. Don't leak that back to userspace. 186 */ 187 if (error == -ECANCELED) 188 error = 0; 189 190 return error; 191 } 192 193 static int 194 xfs_bulkstat_iwalk( 195 struct xfs_mount *mp, 196 struct xfs_trans *tp, 197 xfs_ino_t ino, 198 void *data) 199 { 200 struct xfs_bstat_chunk *bc = data; 201 int error; 202 203 error = xfs_bulkstat_one_int(mp, bc->breq->mnt_userns, tp, ino, data); 204 /* bulkstat just skips over missing inodes */ 205 if (error == -ENOENT || error == -EINVAL) 206 return 0; 207 return error; 208 } 209 210 /* 211 * Check the incoming lastino parameter. 212 * 213 * We allow any inode value that could map to physical space inside the 214 * filesystem because if there are no inodes there, bulkstat moves on to the 215 * next chunk. In other words, the magic agino value of zero takes us to the 216 * first chunk in the AG, and an agino value past the end of the AG takes us to 217 * the first chunk in the next AG. 218 * 219 * Therefore we can end early if the requested inode is beyond the end of the 220 * filesystem or doesn't map properly. 221 */ 222 static inline bool 223 xfs_bulkstat_already_done( 224 struct xfs_mount *mp, 225 xfs_ino_t startino) 226 { 227 xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, startino); 228 xfs_agino_t agino = XFS_INO_TO_AGINO(mp, startino); 229 230 return agno >= mp->m_sb.sb_agcount || 231 startino != XFS_AGINO_TO_INO(mp, agno, agino); 232 } 233 234 /* Return stat information in bulk (by-inode) for the filesystem. */ 235 int 236 xfs_bulkstat( 237 struct xfs_ibulk *breq, 238 bulkstat_one_fmt_pf formatter) 239 { 240 struct xfs_bstat_chunk bc = { 241 .formatter = formatter, 242 .breq = breq, 243 }; 244 int error; 245 246 if (breq->mnt_userns != &init_user_ns) { 247 xfs_warn_ratelimited(breq->mp, 248 "bulkstat not supported inside of idmapped mounts."); 249 return -EINVAL; 250 } 251 if (xfs_bulkstat_already_done(breq->mp, breq->startino)) 252 return 0; 253 254 bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat), 255 KM_MAYFAIL); 256 if (!bc.buf) 257 return -ENOMEM; 258 259 error = xfs_iwalk(breq->mp, NULL, breq->startino, breq->flags, 260 xfs_bulkstat_iwalk, breq->icount, &bc); 261 262 kmem_free(bc.buf); 263 264 /* 265 * We found some inodes, so clear the error status and return them. 266 * The lastino pointer will point directly at the inode that triggered 267 * any error that occurred, so on the next call the error will be 268 * triggered again and propagated to userspace as there will be no 269 * formatted inodes in the buffer. 270 */ 271 if (breq->ocount > 0) 272 error = 0; 273 274 return error; 275 } 276 277 /* Convert bulkstat (v5) to bstat (v1). */ 278 void 279 xfs_bulkstat_to_bstat( 280 struct xfs_mount *mp, 281 struct xfs_bstat *bs1, 282 const struct xfs_bulkstat *bstat) 283 { 284 /* memset is needed here because of padding holes in the structure. */ 285 memset(bs1, 0, sizeof(struct xfs_bstat)); 286 bs1->bs_ino = bstat->bs_ino; 287 bs1->bs_mode = bstat->bs_mode; 288 bs1->bs_nlink = bstat->bs_nlink; 289 bs1->bs_uid = bstat->bs_uid; 290 bs1->bs_gid = bstat->bs_gid; 291 bs1->bs_rdev = bstat->bs_rdev; 292 bs1->bs_blksize = bstat->bs_blksize; 293 bs1->bs_size = bstat->bs_size; 294 bs1->bs_atime.tv_sec = bstat->bs_atime; 295 bs1->bs_mtime.tv_sec = bstat->bs_mtime; 296 bs1->bs_ctime.tv_sec = bstat->bs_ctime; 297 bs1->bs_atime.tv_nsec = bstat->bs_atime_nsec; 298 bs1->bs_mtime.tv_nsec = bstat->bs_mtime_nsec; 299 bs1->bs_ctime.tv_nsec = bstat->bs_ctime_nsec; 300 bs1->bs_blocks = bstat->bs_blocks; 301 bs1->bs_xflags = bstat->bs_xflags; 302 bs1->bs_extsize = XFS_FSB_TO_B(mp, bstat->bs_extsize_blks); 303 bs1->bs_extents = bstat->bs_extents; 304 bs1->bs_gen = bstat->bs_gen; 305 bs1->bs_projid_lo = bstat->bs_projectid & 0xFFFF; 306 bs1->bs_forkoff = bstat->bs_forkoff; 307 bs1->bs_projid_hi = bstat->bs_projectid >> 16; 308 bs1->bs_sick = bstat->bs_sick; 309 bs1->bs_checked = bstat->bs_checked; 310 bs1->bs_cowextsize = XFS_FSB_TO_B(mp, bstat->bs_cowextsize_blks); 311 bs1->bs_dmevmask = 0; 312 bs1->bs_dmstate = 0; 313 bs1->bs_aextents = bstat->bs_aextents; 314 } 315 316 struct xfs_inumbers_chunk { 317 inumbers_fmt_pf formatter; 318 struct xfs_ibulk *breq; 319 }; 320 321 /* 322 * INUMBERS 323 * ======== 324 * This is how we export inode btree records to userspace, so that XFS tools 325 * can figure out where inodes are allocated. 326 */ 327 328 /* 329 * Format the inode group structure and report it somewhere. 330 * 331 * Similar to xfs_bulkstat_one_int, lastino is the inode cursor as we walk 332 * through the filesystem so we move it forward unless there was a runtime 333 * error. If the formatter tells us the buffer is now full we also move the 334 * cursor forward and abort the walk. 335 */ 336 STATIC int 337 xfs_inumbers_walk( 338 struct xfs_mount *mp, 339 struct xfs_trans *tp, 340 xfs_agnumber_t agno, 341 const struct xfs_inobt_rec_incore *irec, 342 void *data) 343 { 344 struct xfs_inumbers inogrp = { 345 .xi_startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino), 346 .xi_alloccount = irec->ir_count - irec->ir_freecount, 347 .xi_allocmask = ~irec->ir_free, 348 .xi_version = XFS_INUMBERS_VERSION_V5, 349 }; 350 struct xfs_inumbers_chunk *ic = data; 351 int error; 352 353 error = ic->formatter(ic->breq, &inogrp); 354 if (error && error != -ECANCELED) 355 return error; 356 357 ic->breq->startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino) + 358 XFS_INODES_PER_CHUNK; 359 return error; 360 } 361 362 /* 363 * Return inode number table for the filesystem. 364 */ 365 int 366 xfs_inumbers( 367 struct xfs_ibulk *breq, 368 inumbers_fmt_pf formatter) 369 { 370 struct xfs_inumbers_chunk ic = { 371 .formatter = formatter, 372 .breq = breq, 373 }; 374 int error = 0; 375 376 if (xfs_bulkstat_already_done(breq->mp, breq->startino)) 377 return 0; 378 379 error = xfs_inobt_walk(breq->mp, NULL, breq->startino, breq->flags, 380 xfs_inumbers_walk, breq->icount, &ic); 381 382 /* 383 * We found some inode groups, so clear the error status and return 384 * them. The lastino pointer will point directly at the inode that 385 * triggered any error that occurred, so on the next call the error 386 * will be triggered again and propagated to userspace as there will be 387 * no formatted inode groups in the buffer. 388 */ 389 if (breq->ocount > 0) 390 error = 0; 391 392 return error; 393 } 394 395 /* Convert an inumbers (v5) struct to a inogrp (v1) struct. */ 396 void 397 xfs_inumbers_to_inogrp( 398 struct xfs_inogrp *ig1, 399 const struct xfs_inumbers *ig) 400 { 401 /* memset is needed here because of padding holes in the structure. */ 402 memset(ig1, 0, sizeof(struct xfs_inogrp)); 403 ig1->xi_startino = ig->xi_startino; 404 ig1->xi_alloccount = ig->xi_alloccount; 405 ig1->xi_allocmask = ig->xi_allocmask; 406 } 407