1/* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or https://opensource.org/licenses/CDDL-1.0. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21/* 22 * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved. 23 */ 24 25#include <sys/zfs_context.h> 26#include <sys/vnode.h> 27#include <sys/sa.h> 28#include <sys/zfs_acl.h> 29#include <sys/zfs_sa.h> 30#include <sys/dmu_objset.h> 31#include <sys/sa_impl.h> 32#include <sys/zfeature.h> 33 34/* 35 * ZPL attribute registration table. 36 * Order of attributes doesn't matter 37 * a unique value will be assigned for each 38 * attribute that is file system specific 39 * 40 * This is just the set of ZPL attributes that this 41 * version of ZFS deals with natively. The file system 42 * could have other attributes stored in files, but they will be 43 * ignored. The SA framework will preserve them, just that 44 * this version of ZFS won't change or delete them. 45 */ 46 47const sa_attr_reg_t zfs_attr_table[ZPL_END+1] = { 48 {"ZPL_ATIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 0}, 49 {"ZPL_MTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 1}, 50 {"ZPL_CTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 2}, 51 {"ZPL_CRTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 3}, 52 {"ZPL_GEN", sizeof (uint64_t), SA_UINT64_ARRAY, 4}, 53 {"ZPL_MODE", sizeof (uint64_t), SA_UINT64_ARRAY, 5}, 54 {"ZPL_SIZE", sizeof (uint64_t), SA_UINT64_ARRAY, 6}, 55 {"ZPL_PARENT", sizeof (uint64_t), SA_UINT64_ARRAY, 7}, 56 {"ZPL_LINKS", sizeof (uint64_t), SA_UINT64_ARRAY, 8}, 57 {"ZPL_XATTR", sizeof (uint64_t), SA_UINT64_ARRAY, 9}, 58 {"ZPL_RDEV", sizeof (uint64_t), SA_UINT64_ARRAY, 10}, 59 {"ZPL_FLAGS", sizeof (uint64_t), SA_UINT64_ARRAY, 11}, 60 {"ZPL_UID", sizeof (uint64_t), SA_UINT64_ARRAY, 12}, 61 {"ZPL_GID", sizeof (uint64_t), SA_UINT64_ARRAY, 13}, 62 {"ZPL_PAD", sizeof (uint64_t) * 4, SA_UINT64_ARRAY, 14}, 63 {"ZPL_ZNODE_ACL", 88, SA_UINT8_ARRAY, 15}, 64 {"ZPL_DACL_COUNT", sizeof (uint64_t), SA_UINT64_ARRAY, 0}, 65 {"ZPL_SYMLINK", 0, SA_UINT8_ARRAY, 0}, 66 {"ZPL_SCANSTAMP", 32, SA_UINT8_ARRAY, 0}, 67 {"ZPL_DACL_ACES", 0, SA_ACL, 0}, 68 {"ZPL_DXATTR", 0, SA_UINT8_ARRAY, 0}, 69 {"ZPL_PROJID", sizeof (uint64_t), SA_UINT64_ARRAY, 0}, 70 {NULL, 0, 0, 0} 71}; 72 73 74#ifdef _KERNEL 75static int zfs_zil_saxattr = 1; 76 77int 78zfs_sa_readlink(znode_t *zp, zfs_uio_t *uio) 79{ 80 dmu_buf_t *db = sa_get_db(zp->z_sa_hdl); 81 size_t bufsz; 82 int error; 83 84 bufsz = zp->z_size; 85 if (bufsz + ZFS_OLD_ZNODE_PHYS_SIZE <= db->db_size) { 86 error = zfs_uiomove((caddr_t)db->db_data + 87 ZFS_OLD_ZNODE_PHYS_SIZE, 88 MIN((size_t)bufsz, zfs_uio_resid(uio)), UIO_READ, uio); 89 } else { 90 dmu_buf_t *dbp; 91 if ((error = dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id, 92 0, FTAG, &dbp, DMU_READ_NO_PREFETCH)) == 0) { 93 error = zfs_uiomove(dbp->db_data, 94 MIN((size_t)bufsz, zfs_uio_resid(uio)), UIO_READ, 95 uio); 96 dmu_buf_rele(dbp, FTAG); 97 } 98 } 99 return (error); 100} 101 102void 103zfs_sa_symlink(znode_t *zp, char *link, int len, dmu_tx_t *tx) 104{ 105 dmu_buf_t *db = sa_get_db(zp->z_sa_hdl); 106 107 if (ZFS_OLD_ZNODE_PHYS_SIZE + len <= dmu_bonus_max()) { 108 VERIFY0(dmu_set_bonus(db, len + ZFS_OLD_ZNODE_PHYS_SIZE, tx)); 109 if (len) { 110 memcpy((caddr_t)db->db_data + 111 ZFS_OLD_ZNODE_PHYS_SIZE, link, len); 112 } 113 } else { 114 dmu_buf_t *dbp; 115 116 zfs_grow_blocksize(zp, len, tx); 117 VERIFY0(dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id, 0, FTAG, &dbp, 118 DMU_READ_NO_PREFETCH)); 119 120 dmu_buf_will_dirty(dbp, tx); 121 122 ASSERT3U(len, <=, dbp->db_size); 123 memcpy(dbp->db_data, link, len); 124 dmu_buf_rele(dbp, FTAG); 125 } 126} 127 128void 129zfs_sa_get_scanstamp(znode_t *zp, xvattr_t *xvap) 130{ 131 zfsvfs_t *zfsvfs = ZTOZSB(zp); 132 xoptattr_t *xoap; 133 134 ASSERT(MUTEX_HELD(&zp->z_lock)); 135 VERIFY((xoap = xva_getxoptattr(xvap)) != NULL); 136 if (zp->z_is_sa) { 137 if (sa_lookup(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs), 138 &xoap->xoa_av_scanstamp, 139 sizeof (xoap->xoa_av_scanstamp)) != 0) 140 return; 141 } else { 142 dmu_object_info_t doi; 143 dmu_buf_t *db = sa_get_db(zp->z_sa_hdl); 144 int len; 145 146 if (!(zp->z_pflags & ZFS_BONUS_SCANSTAMP)) 147 return; 148 149 sa_object_info(zp->z_sa_hdl, &doi); 150 len = sizeof (xoap->xoa_av_scanstamp) + 151 ZFS_OLD_ZNODE_PHYS_SIZE; 152 153 if (len <= doi.doi_bonus_size) { 154 (void) memcpy(xoap->xoa_av_scanstamp, 155 (caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE, 156 sizeof (xoap->xoa_av_scanstamp)); 157 } 158 } 159 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); 160} 161 162void 163zfs_sa_set_scanstamp(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) 164{ 165 zfsvfs_t *zfsvfs = ZTOZSB(zp); 166 xoptattr_t *xoap; 167 168 ASSERT(MUTEX_HELD(&zp->z_lock)); 169 VERIFY((xoap = xva_getxoptattr(xvap)) != NULL); 170 if (zp->z_is_sa) 171 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs), 172 &xoap->xoa_av_scanstamp, 173 sizeof (xoap->xoa_av_scanstamp), tx)); 174 else { 175 dmu_object_info_t doi; 176 dmu_buf_t *db = sa_get_db(zp->z_sa_hdl); 177 int len; 178 179 sa_object_info(zp->z_sa_hdl, &doi); 180 len = sizeof (xoap->xoa_av_scanstamp) + 181 ZFS_OLD_ZNODE_PHYS_SIZE; 182 if (len > doi.doi_bonus_size) 183 VERIFY(dmu_set_bonus(db, len, tx) == 0); 184 (void) memcpy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE, 185 xoap->xoa_av_scanstamp, sizeof (xoap->xoa_av_scanstamp)); 186 187 zp->z_pflags |= ZFS_BONUS_SCANSTAMP; 188 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), 189 &zp->z_pflags, sizeof (uint64_t), tx)); 190 } 191} 192 193int 194zfs_sa_get_xattr(znode_t *zp) 195{ 196 zfsvfs_t *zfsvfs = ZTOZSB(zp); 197 char *obj; 198 int size; 199 int error; 200 201 ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock)); 202 ASSERT(!zp->z_xattr_cached); 203 ASSERT(zp->z_is_sa); 204 205 error = sa_size(zp->z_sa_hdl, SA_ZPL_DXATTR(zfsvfs), &size); 206 if (error) { 207 if (error == ENOENT) 208 return nvlist_alloc(&zp->z_xattr_cached, 209 NV_UNIQUE_NAME, KM_SLEEP); 210 else 211 return (error); 212 } 213 214 obj = vmem_alloc(size, KM_SLEEP); 215 216 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DXATTR(zfsvfs), obj, size); 217 if (error == 0) 218 error = nvlist_unpack(obj, size, &zp->z_xattr_cached, KM_SLEEP); 219 220 vmem_free(obj, size); 221 222 return (error); 223} 224 225int 226zfs_sa_set_xattr(znode_t *zp, const char *name, const void *value, size_t vsize) 227{ 228 zfsvfs_t *zfsvfs = ZTOZSB(zp); 229 zilog_t *zilog; 230 dmu_tx_t *tx; 231 char *obj; 232 size_t size; 233 int error, logsaxattr = 0; 234 235 ASSERT(RW_WRITE_HELD(&zp->z_xattr_lock)); 236 ASSERT(zp->z_xattr_cached); 237 ASSERT(zp->z_is_sa); 238 239 error = nvlist_size(zp->z_xattr_cached, &size, NV_ENCODE_XDR); 240 if ((error == 0) && (size > SA_ATTR_MAX_LEN)) 241 error = SET_ERROR(EFBIG); 242 if (error) 243 goto out; 244 245 obj = vmem_alloc(size, KM_SLEEP); 246 247 error = nvlist_pack(zp->z_xattr_cached, &obj, &size, 248 NV_ENCODE_XDR, KM_SLEEP); 249 if (error) 250 goto out_free; 251 252 zilog = zfsvfs->z_log; 253 254 /* 255 * Users enable ZIL logging of xattr=sa operations by enabling the 256 * SPA_FEATURE_ZILSAXATTR feature on the pool. Feature is activated 257 * during zil_process_commit_list/zil_create, if enabled. 258 */ 259 if (spa_feature_is_enabled(zfsvfs->z_os->os_spa, 260 SPA_FEATURE_ZILSAXATTR) && zfs_zil_saxattr) 261 logsaxattr = 1; 262 263 tx = dmu_tx_create(zfsvfs->z_os); 264 dmu_tx_hold_sa_create(tx, size); 265 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); 266 267 error = dmu_tx_assign(tx, TXG_WAIT); 268 if (error) { 269 dmu_tx_abort(tx); 270 } else { 271 int count = 0; 272 sa_bulk_attr_t bulk[2]; 273 uint64_t ctime[2]; 274 275 if (logsaxattr) 276 zfs_log_setsaxattr(zilog, tx, TX_SETSAXATTR, zp, name, 277 value, vsize); 278 279 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime); 280 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DXATTR(zfsvfs), 281 NULL, obj, size); 282 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), 283 NULL, &ctime, 16); 284 VERIFY0(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx)); 285 286 dmu_tx_commit(tx); 287 if (logsaxattr && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 288 zil_commit(zilog, 0); 289 } 290out_free: 291 vmem_free(obj, size); 292out: 293 return (error); 294} 295 296/* 297 * I'm not convinced we should do any of this upgrade. 298 * since the SA code can read both old/new znode formats 299 * with probably little to no performance difference. 300 * 301 * All new files will be created with the new format. 302 */ 303 304void 305zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx) 306{ 307 dmu_buf_t *db = sa_get_db(hdl); 308 znode_t *zp = sa_get_userdata(hdl); 309 zfsvfs_t *zfsvfs = ZTOZSB(zp); 310 int count = 0; 311 sa_bulk_attr_t *bulk, *sa_attrs; 312 zfs_acl_locator_cb_t locate = { 0 }; 313 uint64_t uid, gid, mode, rdev, xattr, parent, tmp_gen; 314 uint64_t crtime[2], mtime[2], ctime[2], atime[2]; 315 uint64_t links; 316 zfs_acl_phys_t znode_acl; 317 char scanstamp[AV_SCANSTAMP_SZ]; 318 boolean_t drop_lock = B_FALSE; 319 320 /* 321 * No upgrade if ACL isn't cached 322 * since we won't know which locks are held 323 * and ready the ACL would require special "locked" 324 * interfaces that would be messy 325 */ 326 if (zp->z_acl_cached == NULL || Z_ISLNK(ZTOTYPE(zp))) 327 return; 328 329 /* 330 * If the z_lock is held and we aren't the owner 331 * the just return since we don't want to deadlock 332 * trying to update the status of z_is_sa. This 333 * file can then be upgraded at a later time. 334 * 335 * Otherwise, we know we are doing the 336 * sa_update() that caused us to enter this function. 337 */ 338 if (MUTEX_NOT_HELD(&zp->z_lock)) { 339 if (mutex_tryenter(&zp->z_lock) == 0) 340 return; 341 else 342 drop_lock = B_TRUE; 343 } 344 345 /* First do a bulk query of the attributes that aren't cached */ 346 bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); 347 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); 348 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); 349 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); 350 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16); 351 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); 352 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8); 353 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL, &xattr, 8); 354 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL, &rdev, 8); 355 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8); 356 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8); 357 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &tmp_gen, 8); 358 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL, 359 &znode_acl, 88); 360 361 if (sa_bulk_lookup_locked(hdl, bulk, count) != 0) 362 goto done; 363 364 if (dmu_objset_projectquota_enabled(hdl->sa_os) && 365 !(zp->z_pflags & ZFS_PROJID)) { 366 zp->z_pflags |= ZFS_PROJID; 367 zp->z_projid = ZFS_DEFAULT_PROJID; 368 } 369 370 /* 371 * While the order here doesn't matter its best to try and organize 372 * it is such a way to pick up an already existing layout number 373 */ 374 count = 0; 375 sa_attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); 376 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); 377 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zfsvfs), NULL, 378 &zp->z_size, 8); 379 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zfsvfs), 380 NULL, &tmp_gen, 8); 381 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8); 382 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8); 383 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zfsvfs), 384 NULL, &parent, 8); 385 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL, 386 &zp->z_pflags, 8); 387 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zfsvfs), NULL, 388 &atime, 16); 389 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zfsvfs), NULL, 390 &mtime, 16); 391 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zfsvfs), NULL, 392 &ctime, 16); 393 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL, 394 &crtime, 16); 395 links = ZTONLNK(zp); 396 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zfsvfs), NULL, 397 &links, 8); 398 if (dmu_objset_projectquota_enabled(hdl->sa_os)) 399 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PROJID(zfsvfs), NULL, 400 &zp->z_projid, 8); 401 if (Z_ISBLK(ZTOTYPE(zp)) || Z_ISCHR(ZTOTYPE(zp))) 402 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zfsvfs), NULL, 403 &rdev, 8); 404 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), NULL, 405 &zp->z_acl_cached->z_acl_count, 8); 406 407 if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID) 408 zfs_acl_xform(zp, zp->z_acl_cached, CRED()); 409 410 locate.cb_aclp = zp->z_acl_cached; 411 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zfsvfs), 412 zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes); 413 414 if (xattr) 415 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zfsvfs), 416 NULL, &xattr, 8); 417 418 /* if scanstamp then add scanstamp */ 419 420 if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) { 421 memcpy(scanstamp, 422 (caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE, 423 AV_SCANSTAMP_SZ); 424 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zfsvfs), 425 NULL, scanstamp, AV_SCANSTAMP_SZ); 426 zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP; 427 } 428 429 VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0); 430 VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs, 431 count, tx) == 0); 432 if (znode_acl.z_acl_extern_obj) 433 VERIFY(0 == dmu_object_free(zfsvfs->z_os, 434 znode_acl.z_acl_extern_obj, tx)); 435 436 zp->z_is_sa = B_TRUE; 437 kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END); 438done: 439 kmem_free(bulk, sizeof (sa_bulk_attr_t) * ZPL_END); 440 if (drop_lock) 441 mutex_exit(&zp->z_lock); 442} 443 444void 445zfs_sa_upgrade_txholds(dmu_tx_t *tx, znode_t *zp) 446{ 447 if (!ZTOZSB(zp)->z_use_sa || zp->z_is_sa) 448 return; 449 450 451 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); 452 453 if (zfs_external_acl(zp)) { 454 dmu_tx_hold_free(tx, zfs_external_acl(zp), 0, 455 DMU_OBJECT_END); 456 } 457} 458 459ZFS_MODULE_PARAM(zfs, zfs_, zil_saxattr, INT, ZMOD_RW, 460 "Disable xattr=sa extended attribute logging in ZIL by settng 0."); 461 462EXPORT_SYMBOL(zfs_attr_table); 463EXPORT_SYMBOL(zfs_sa_readlink); 464EXPORT_SYMBOL(zfs_sa_symlink); 465EXPORT_SYMBOL(zfs_sa_get_scanstamp); 466EXPORT_SYMBOL(zfs_sa_set_scanstamp); 467EXPORT_SYMBOL(zfs_sa_get_xattr); 468EXPORT_SYMBOL(zfs_sa_set_xattr); 469EXPORT_SYMBOL(zfs_sa_upgrade); 470EXPORT_SYMBOL(zfs_sa_upgrade_txholds); 471 472#endif 473