1 // SPDX-License-Identifier: GPL-2.0 2 #include <linux/ceph/ceph_debug.h> 3 #include <linux/in.h> 4 5 #include "super.h" 6 #include "mds_client.h" 7 #include "ioctl.h" 8 #include <linux/ceph/striper.h> 9 #include <linux/fscrypt.h> 10 11 /* 12 * ioctls 13 */ 14 15 /* 16 * get and set the file layout 17 */ 18 static long ceph_ioctl_get_layout(struct file *file, void __user *arg) 19 { 20 struct ceph_inode_info *ci = ceph_inode(file_inode(file)); 21 struct ceph_ioctl_layout l; 22 int err; 23 24 err = ceph_do_getattr(file_inode(file), CEPH_STAT_CAP_LAYOUT, false); 25 if (!err) { 26 l.stripe_unit = ci->i_layout.stripe_unit; 27 l.stripe_count = ci->i_layout.stripe_count; 28 l.object_size = ci->i_layout.object_size; 29 l.data_pool = ci->i_layout.pool_id; 30 l.preferred_osd = -1; 31 if (copy_to_user(arg, &l, sizeof(l))) 32 return -EFAULT; 33 } 34 35 return err; 36 } 37 38 static long __validate_layout(struct ceph_mds_client *mdsc, 39 struct ceph_ioctl_layout *l) 40 { 41 int i, err; 42 43 /* validate striping parameters */ 44 if ((l->object_size & ~PAGE_MASK) || 45 (l->stripe_unit & ~PAGE_MASK) || 46 ((unsigned)l->stripe_unit != 0 && 47 ((unsigned)l->object_size % (unsigned)l->stripe_unit))) 48 return -EINVAL; 49 50 /* make sure it's a valid data pool */ 51 mutex_lock(&mdsc->mutex); 52 err = -EINVAL; 53 for (i = 0; i < mdsc->mdsmap->m_num_data_pg_pools; i++) 54 if (mdsc->mdsmap->m_data_pg_pools[i] == l->data_pool) { 55 err = 0; 56 break; 57 } 58 mutex_unlock(&mdsc->mutex); 59 if (err) 60 return err; 61 62 return 0; 63 } 64 65 static long ceph_ioctl_set_layout(struct file *file, void __user *arg) 66 { 67 struct inode *inode = file_inode(file); 68 struct ceph_mds_client *mdsc = ceph_sb_to_fs_client(inode->i_sb)->mdsc; 69 struct ceph_mds_request *req; 70 struct ceph_ioctl_layout l; 71 struct ceph_inode_info *ci = ceph_inode(file_inode(file)); 72 struct ceph_ioctl_layout nl; 73 int err; 74 75 if (copy_from_user(&l, arg, sizeof(l))) 76 return -EFAULT; 77 78 /* validate changed params against current layout */ 79 err = ceph_do_getattr(file_inode(file), CEPH_STAT_CAP_LAYOUT, false); 80 if (err) 81 return err; 82 83 memset(&nl, 0, sizeof(nl)); 84 if (l.stripe_count) 85 nl.stripe_count = l.stripe_count; 86 else 87 nl.stripe_count = ci->i_layout.stripe_count; 88 if (l.stripe_unit) 89 nl.stripe_unit = l.stripe_unit; 90 else 91 nl.stripe_unit = ci->i_layout.stripe_unit; 92 if (l.object_size) 93 nl.object_size = l.object_size; 94 else 95 nl.object_size = ci->i_layout.object_size; 96 if (l.data_pool) 97 nl.data_pool = l.data_pool; 98 else 99 nl.data_pool = ci->i_layout.pool_id; 100 101 /* this is obsolete, and always -1 */ 102 nl.preferred_osd = -1; 103 104 err = __validate_layout(mdsc, &nl); 105 if (err) 106 return err; 107 108 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETLAYOUT, 109 USE_AUTH_MDS); 110 if (IS_ERR(req)) 111 return PTR_ERR(req); 112 req->r_inode = inode; 113 ihold(inode); 114 req->r_num_caps = 1; 115 116 req->r_inode_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_FILE_EXCL; 117 118 req->r_args.setlayout.layout.fl_stripe_unit = 119 cpu_to_le32(l.stripe_unit); 120 req->r_args.setlayout.layout.fl_stripe_count = 121 cpu_to_le32(l.stripe_count); 122 req->r_args.setlayout.layout.fl_object_size = 123 cpu_to_le32(l.object_size); 124 req->r_args.setlayout.layout.fl_pg_pool = cpu_to_le32(l.data_pool); 125 126 err = ceph_mdsc_do_request(mdsc, NULL, req); 127 ceph_mdsc_put_request(req); 128 return err; 129 } 130 131 /* 132 * Set a layout policy on a directory inode. All items in the tree 133 * rooted at this inode will inherit this layout on creation, 134 * (It doesn't apply retroactively ) 135 * unless a subdirectory has its own layout policy. 136 */ 137 static long ceph_ioctl_set_layout_policy (struct file *file, void __user *arg) 138 { 139 struct inode *inode = file_inode(file); 140 struct ceph_mds_request *req; 141 struct ceph_ioctl_layout l; 142 int err; 143 struct ceph_mds_client *mdsc = ceph_sb_to_fs_client(inode->i_sb)->mdsc; 144 145 /* copy and validate */ 146 if (copy_from_user(&l, arg, sizeof(l))) 147 return -EFAULT; 148 149 err = __validate_layout(mdsc, &l); 150 if (err) 151 return err; 152 153 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETDIRLAYOUT, 154 USE_AUTH_MDS); 155 156 if (IS_ERR(req)) 157 return PTR_ERR(req); 158 req->r_inode = inode; 159 ihold(inode); 160 req->r_num_caps = 1; 161 162 req->r_args.setlayout.layout.fl_stripe_unit = 163 cpu_to_le32(l.stripe_unit); 164 req->r_args.setlayout.layout.fl_stripe_count = 165 cpu_to_le32(l.stripe_count); 166 req->r_args.setlayout.layout.fl_object_size = 167 cpu_to_le32(l.object_size); 168 req->r_args.setlayout.layout.fl_pg_pool = 169 cpu_to_le32(l.data_pool); 170 171 err = ceph_mdsc_do_request(mdsc, inode, req); 172 ceph_mdsc_put_request(req); 173 return err; 174 } 175 176 /* 177 * Return object name, size/offset information, and location (OSD 178 * number, network address) for a given file offset. 179 */ 180 static long ceph_ioctl_get_dataloc(struct file *file, void __user *arg) 181 { 182 struct ceph_ioctl_dataloc dl; 183 struct inode *inode = file_inode(file); 184 struct ceph_inode_info *ci = ceph_inode(inode); 185 struct ceph_osd_client *osdc = 186 &ceph_sb_to_fs_client(inode->i_sb)->client->osdc; 187 struct ceph_object_locator oloc; 188 CEPH_DEFINE_OID_ONSTACK(oid); 189 u32 xlen; 190 u64 tmp; 191 struct ceph_pg pgid; 192 int r; 193 194 /* copy and validate */ 195 if (copy_from_user(&dl, arg, sizeof(dl))) 196 return -EFAULT; 197 198 down_read(&osdc->lock); 199 ceph_calc_file_object_mapping(&ci->i_layout, dl.file_offset, 1, 200 &dl.object_no, &dl.object_offset, &xlen); 201 dl.file_offset -= dl.object_offset; 202 dl.object_size = ci->i_layout.object_size; 203 dl.block_size = ci->i_layout.stripe_unit; 204 205 /* block_offset = object_offset % block_size */ 206 tmp = dl.object_offset; 207 dl.block_offset = do_div(tmp, dl.block_size); 208 209 snprintf(dl.object_name, sizeof(dl.object_name), "%llx.%08llx", 210 ceph_ino(inode), dl.object_no); 211 212 oloc.pool = ci->i_layout.pool_id; 213 oloc.pool_ns = ceph_try_get_string(ci->i_layout.pool_ns); 214 ceph_oid_printf(&oid, "%s", dl.object_name); 215 216 r = ceph_object_locator_to_pg(osdc->osdmap, &oid, &oloc, &pgid); 217 218 ceph_oloc_destroy(&oloc); 219 if (r < 0) { 220 up_read(&osdc->lock); 221 return r; 222 } 223 224 dl.osd = ceph_pg_to_acting_primary(osdc->osdmap, &pgid); 225 if (dl.osd >= 0) { 226 struct ceph_entity_addr *a = 227 ceph_osd_addr(osdc->osdmap, dl.osd); 228 if (a) 229 memcpy(&dl.osd_addr, &a->in_addr, sizeof(dl.osd_addr)); 230 } else { 231 memset(&dl.osd_addr, 0, sizeof(dl.osd_addr)); 232 } 233 up_read(&osdc->lock); 234 235 /* send result back to user */ 236 if (copy_to_user(arg, &dl, sizeof(dl))) 237 return -EFAULT; 238 239 return 0; 240 } 241 242 static long ceph_ioctl_lazyio(struct file *file) 243 { 244 struct ceph_file_info *fi = file->private_data; 245 struct inode *inode = file_inode(file); 246 struct ceph_inode_info *ci = ceph_inode(inode); 247 struct ceph_mds_client *mdsc = ceph_inode_to_fs_client(inode)->mdsc; 248 struct ceph_client *cl = mdsc->fsc->client; 249 bool is_file_already_lazy = false; 250 251 spin_lock(&ci->i_ceph_lock); 252 if ((fi->fmode & CEPH_FILE_MODE_LAZY) == 0) { 253 fi->fmode |= CEPH_FILE_MODE_LAZY; 254 ci->i_nr_by_mode[ffs(CEPH_FILE_MODE_LAZY)]++; 255 __ceph_touch_fmode(ci, mdsc, fi->fmode); 256 } else { 257 is_file_already_lazy = true; 258 } 259 spin_unlock(&ci->i_ceph_lock); 260 261 if (is_file_already_lazy) { 262 doutc(cl, "file %p %p %llx.%llx already lazy\n", file, inode, 263 ceph_vinop(inode)); 264 } else { 265 doutc(cl, "file %p %p %llx.%llx marked lazy\n", file, inode, 266 ceph_vinop(inode)); 267 268 ceph_check_caps(ci, 0); 269 } 270 271 return 0; 272 } 273 274 static long ceph_ioctl_syncio(struct file *file) 275 { 276 struct ceph_file_info *fi = file->private_data; 277 278 fi->flags |= CEPH_F_SYNC; 279 return 0; 280 } 281 282 static int vet_mds_for_fscrypt(struct file *file) 283 { 284 int i, ret = -EOPNOTSUPP; 285 struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(file_inode(file)->i_sb); 286 287 mutex_lock(&mdsc->mutex); 288 for (i = 0; i < mdsc->max_sessions; i++) { 289 struct ceph_mds_session *s = mdsc->sessions[i]; 290 291 if (!s) 292 continue; 293 if (test_bit(CEPHFS_FEATURE_ALTERNATE_NAME, &s->s_features)) 294 ret = 0; 295 break; 296 } 297 mutex_unlock(&mdsc->mutex); 298 return ret; 299 } 300 301 static long ceph_set_encryption_policy(struct file *file, unsigned long arg) 302 { 303 int ret, got = 0; 304 struct inode *inode = file_inode(file); 305 struct ceph_inode_info *ci = ceph_inode(inode); 306 307 /* encrypted directories can't have striped layout */ 308 if (ci->i_layout.stripe_count > 1) 309 return -EINVAL; 310 311 ret = vet_mds_for_fscrypt(file); 312 if (ret) 313 return ret; 314 315 /* 316 * Ensure we hold these caps so that we _know_ that the rstats check 317 * in the empty_dir check is reliable. 318 */ 319 ret = ceph_get_caps(file, CEPH_CAP_FILE_SHARED, 0, -1, &got); 320 if (ret) 321 return ret; 322 323 ret = fscrypt_ioctl_set_policy(file, (const void __user *)arg); 324 if (got) 325 ceph_put_cap_refs(ci, got); 326 327 return ret; 328 } 329 330 static const char *ceph_ioctl_cmd_name(const unsigned int cmd) 331 { 332 switch (cmd) { 333 case CEPH_IOC_GET_LAYOUT: 334 return "get_layout"; 335 case CEPH_IOC_SET_LAYOUT: 336 return "set_layout"; 337 case CEPH_IOC_SET_LAYOUT_POLICY: 338 return "set_layout_policy"; 339 case CEPH_IOC_GET_DATALOC: 340 return "get_dataloc"; 341 case CEPH_IOC_LAZYIO: 342 return "lazyio"; 343 case CEPH_IOC_SYNCIO: 344 return "syncio"; 345 case FS_IOC_SET_ENCRYPTION_POLICY: 346 return "set_encryption_policy"; 347 case FS_IOC_GET_ENCRYPTION_POLICY: 348 return "get_encryption_policy"; 349 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 350 return "get_encryption_policy_ex"; 351 case FS_IOC_ADD_ENCRYPTION_KEY: 352 return "add_encryption_key"; 353 case FS_IOC_REMOVE_ENCRYPTION_KEY: 354 return "remove_encryption_key"; 355 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 356 return "remove_encryption_key_all_users"; 357 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 358 return "get_encryption_key_status"; 359 case FS_IOC_GET_ENCRYPTION_NONCE: 360 return "get_encryption_nonce"; 361 default: 362 return "unknown"; 363 } 364 } 365 366 long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 367 { 368 struct inode *inode = file_inode(file); 369 struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode); 370 int ret; 371 372 doutc(fsc->client, "file %p %p %llx.%llx cmd %s arg %lu\n", file, 373 inode, ceph_vinop(inode), ceph_ioctl_cmd_name(cmd), arg); 374 switch (cmd) { 375 case CEPH_IOC_GET_LAYOUT: 376 return ceph_ioctl_get_layout(file, (void __user *)arg); 377 378 case CEPH_IOC_SET_LAYOUT: 379 return ceph_ioctl_set_layout(file, (void __user *)arg); 380 381 case CEPH_IOC_SET_LAYOUT_POLICY: 382 return ceph_ioctl_set_layout_policy(file, (void __user *)arg); 383 384 case CEPH_IOC_GET_DATALOC: 385 return ceph_ioctl_get_dataloc(file, (void __user *)arg); 386 387 case CEPH_IOC_LAZYIO: 388 return ceph_ioctl_lazyio(file); 389 390 case CEPH_IOC_SYNCIO: 391 return ceph_ioctl_syncio(file); 392 393 case FS_IOC_SET_ENCRYPTION_POLICY: 394 return ceph_set_encryption_policy(file, arg); 395 396 case FS_IOC_GET_ENCRYPTION_POLICY: 397 ret = vet_mds_for_fscrypt(file); 398 if (ret) 399 return ret; 400 return fscrypt_ioctl_get_policy(file, (void __user *)arg); 401 402 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 403 ret = vet_mds_for_fscrypt(file); 404 if (ret) 405 return ret; 406 return fscrypt_ioctl_get_policy_ex(file, (void __user *)arg); 407 408 case FS_IOC_ADD_ENCRYPTION_KEY: 409 ret = vet_mds_for_fscrypt(file); 410 if (ret) 411 return ret; 412 return fscrypt_ioctl_add_key(file, (void __user *)arg); 413 414 case FS_IOC_REMOVE_ENCRYPTION_KEY: 415 return fscrypt_ioctl_remove_key(file, (void __user *)arg); 416 417 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 418 return fscrypt_ioctl_remove_key_all_users(file, 419 (void __user *)arg); 420 421 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 422 return fscrypt_ioctl_get_key_status(file, (void __user *)arg); 423 424 case FS_IOC_GET_ENCRYPTION_NONCE: 425 ret = vet_mds_for_fscrypt(file); 426 if (ret) 427 return ret; 428 return fscrypt_ioctl_get_nonce(file, (void __user *)arg); 429 } 430 431 return -ENOTTY; 432 } 433