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(inode: file_inode(f: file)); |
21 | struct ceph_ioctl_layout l; |
22 | int err; |
23 | |
24 | err = ceph_do_getattr(inode: file_inode(f: file), CEPH_STAT_CAP_LAYOUT, force: 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(to: arg, from: &l, n: 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(lock: &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(f: file); |
68 | struct ceph_mds_client *mdsc = ceph_sb_to_fs_client(sb: inode->i_sb)->mdsc; |
69 | struct ceph_mds_request *req; |
70 | struct ceph_ioctl_layout l; |
71 | struct ceph_inode_info *ci = ceph_inode(inode: file_inode(f: file)); |
72 | struct ceph_ioctl_layout nl; |
73 | int err; |
74 | |
75 | if (copy_from_user(to: &l, from: arg, n: sizeof(l))) |
76 | return -EFAULT; |
77 | |
78 | /* validate changed params against current layout */ |
79 | err = ceph_do_getattr(inode: file_inode(f: file), CEPH_STAT_CAP_LAYOUT, force: 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, l: &nl); |
105 | if (err) |
106 | return err; |
107 | |
108 | req = ceph_mdsc_create_request(mdsc, op: CEPH_MDS_OP_SETLAYOUT, |
109 | mode: USE_AUTH_MDS); |
110 | if (IS_ERR(ptr: req)) |
111 | return PTR_ERR(ptr: 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(f: 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(sb: inode->i_sb)->mdsc; |
144 | |
145 | /* copy and validate */ |
146 | if (copy_from_user(to: &l, from: arg, n: sizeof(l))) |
147 | return -EFAULT; |
148 | |
149 | err = __validate_layout(mdsc, l: &l); |
150 | if (err) |
151 | return err; |
152 | |
153 | req = ceph_mdsc_create_request(mdsc, op: CEPH_MDS_OP_SETDIRLAYOUT, |
154 | mode: USE_AUTH_MDS); |
155 | |
156 | if (IS_ERR(ptr: req)) |
157 | return PTR_ERR(ptr: 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, dir: 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(f: file); |
184 | struct ceph_inode_info *ci = ceph_inode(inode); |
185 | struct ceph_osd_client *osdc = |
186 | &ceph_sb_to_fs_client(sb: 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(to: &dl, from: arg, n: sizeof(dl))) |
196 | return -EFAULT; |
197 | |
198 | down_read(sem: &osdc->lock); |
199 | ceph_calc_file_object_mapping(l: &ci->i_layout, off: dl.file_offset, len: 1, |
200 | objno: &dl.object_no, objoff: &dl.object_offset, xlen: &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(buf: dl.object_name, size: sizeof(dl.object_name), fmt: "%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: &oid, fmt: "%s" , dl.object_name); |
215 | |
216 | r = ceph_object_locator_to_pg(osdmap: osdc->osdmap, oid: &oid, oloc: &oloc, raw_pgid: &pgid); |
217 | |
218 | ceph_oloc_destroy(oloc: &oloc); |
219 | if (r < 0) { |
220 | up_read(sem: &osdc->lock); |
221 | return r; |
222 | } |
223 | |
224 | dl.osd = ceph_pg_to_acting_primary(osdmap: osdc->osdmap, raw_pgid: &pgid); |
225 | if (dl.osd >= 0) { |
226 | struct ceph_entity_addr *a = |
227 | ceph_osd_addr(map: osdc->osdmap, osd: 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(sem: &osdc->lock); |
234 | |
235 | /* send result back to user */ |
236 | if (copy_to_user(to: arg, from: &dl, n: 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(f: 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 | |
250 | if ((fi->fmode & CEPH_FILE_MODE_LAZY) == 0) { |
251 | spin_lock(lock: &ci->i_ceph_lock); |
252 | fi->fmode |= CEPH_FILE_MODE_LAZY; |
253 | ci->i_nr_by_mode[ffs(CEPH_FILE_MODE_LAZY)]++; |
254 | __ceph_touch_fmode(ci, mdsc, fmode: fi->fmode); |
255 | spin_unlock(lock: &ci->i_ceph_lock); |
256 | doutc(cl, "file %p %p %llx.%llx marked lazy\n" , file, inode, |
257 | ceph_vinop(inode)); |
258 | |
259 | ceph_check_caps(ci, flags: 0); |
260 | } else { |
261 | doutc(cl, "file %p %p %llx.%llx already lazy\n" , file, inode, |
262 | ceph_vinop(inode)); |
263 | } |
264 | return 0; |
265 | } |
266 | |
267 | static long ceph_ioctl_syncio(struct file *file) |
268 | { |
269 | struct ceph_file_info *fi = file->private_data; |
270 | |
271 | fi->flags |= CEPH_F_SYNC; |
272 | return 0; |
273 | } |
274 | |
275 | static int vet_mds_for_fscrypt(struct file *file) |
276 | { |
277 | int i, ret = -EOPNOTSUPP; |
278 | struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(sb: file_inode(f: file)->i_sb); |
279 | |
280 | mutex_lock(&mdsc->mutex); |
281 | for (i = 0; i < mdsc->max_sessions; i++) { |
282 | struct ceph_mds_session *s = mdsc->sessions[i]; |
283 | |
284 | if (!s) |
285 | continue; |
286 | if (test_bit(CEPHFS_FEATURE_ALTERNATE_NAME, &s->s_features)) |
287 | ret = 0; |
288 | break; |
289 | } |
290 | mutex_unlock(lock: &mdsc->mutex); |
291 | return ret; |
292 | } |
293 | |
294 | static long ceph_set_encryption_policy(struct file *file, unsigned long arg) |
295 | { |
296 | int ret, got = 0; |
297 | struct inode *inode = file_inode(f: file); |
298 | struct ceph_inode_info *ci = ceph_inode(inode); |
299 | |
300 | /* encrypted directories can't have striped layout */ |
301 | if (ci->i_layout.stripe_count > 1) |
302 | return -EINVAL; |
303 | |
304 | ret = vet_mds_for_fscrypt(file); |
305 | if (ret) |
306 | return ret; |
307 | |
308 | /* |
309 | * Ensure we hold these caps so that we _know_ that the rstats check |
310 | * in the empty_dir check is reliable. |
311 | */ |
312 | ret = ceph_get_caps(filp: file, CEPH_CAP_FILE_SHARED, want: 0, endoff: -1, got: &got); |
313 | if (ret) |
314 | return ret; |
315 | |
316 | ret = fscrypt_ioctl_set_policy(filp: file, arg: (const void __user *)arg); |
317 | if (got) |
318 | ceph_put_cap_refs(ci, had: got); |
319 | |
320 | return ret; |
321 | } |
322 | |
323 | static const char *ceph_ioctl_cmd_name(const unsigned int cmd) |
324 | { |
325 | switch (cmd) { |
326 | case CEPH_IOC_GET_LAYOUT: |
327 | return "get_layout" ; |
328 | case CEPH_IOC_SET_LAYOUT: |
329 | return "set_layout" ; |
330 | case CEPH_IOC_SET_LAYOUT_POLICY: |
331 | return "set_layout_policy" ; |
332 | case CEPH_IOC_GET_DATALOC: |
333 | return "get_dataloc" ; |
334 | case CEPH_IOC_LAZYIO: |
335 | return "lazyio" ; |
336 | case CEPH_IOC_SYNCIO: |
337 | return "syncio" ; |
338 | case FS_IOC_SET_ENCRYPTION_POLICY: |
339 | return "set_encryption_policy" ; |
340 | case FS_IOC_GET_ENCRYPTION_POLICY: |
341 | return "get_encryption_policy" ; |
342 | case FS_IOC_GET_ENCRYPTION_POLICY_EX: |
343 | return "get_encryption_policy_ex" ; |
344 | case FS_IOC_ADD_ENCRYPTION_KEY: |
345 | return "add_encryption_key" ; |
346 | case FS_IOC_REMOVE_ENCRYPTION_KEY: |
347 | return "remove_encryption_key" ; |
348 | case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: |
349 | return "remove_encryption_key_all_users" ; |
350 | case FS_IOC_GET_ENCRYPTION_KEY_STATUS: |
351 | return "get_encryption_key_status" ; |
352 | case FS_IOC_GET_ENCRYPTION_NONCE: |
353 | return "get_encryption_nonce" ; |
354 | default: |
355 | return "unknown" ; |
356 | } |
357 | } |
358 | |
359 | long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg) |
360 | { |
361 | struct inode *inode = file_inode(f: file); |
362 | struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode); |
363 | int ret; |
364 | |
365 | doutc(fsc->client, "file %p %p %llx.%llx cmd %s arg %lu\n" , file, |
366 | inode, ceph_vinop(inode), ceph_ioctl_cmd_name(cmd), arg); |
367 | switch (cmd) { |
368 | case CEPH_IOC_GET_LAYOUT: |
369 | return ceph_ioctl_get_layout(file, arg: (void __user *)arg); |
370 | |
371 | case CEPH_IOC_SET_LAYOUT: |
372 | return ceph_ioctl_set_layout(file, arg: (void __user *)arg); |
373 | |
374 | case CEPH_IOC_SET_LAYOUT_POLICY: |
375 | return ceph_ioctl_set_layout_policy(file, arg: (void __user *)arg); |
376 | |
377 | case CEPH_IOC_GET_DATALOC: |
378 | return ceph_ioctl_get_dataloc(file, arg: (void __user *)arg); |
379 | |
380 | case CEPH_IOC_LAZYIO: |
381 | return ceph_ioctl_lazyio(file); |
382 | |
383 | case CEPH_IOC_SYNCIO: |
384 | return ceph_ioctl_syncio(file); |
385 | |
386 | case FS_IOC_SET_ENCRYPTION_POLICY: |
387 | return ceph_set_encryption_policy(file, arg); |
388 | |
389 | case FS_IOC_GET_ENCRYPTION_POLICY: |
390 | ret = vet_mds_for_fscrypt(file); |
391 | if (ret) |
392 | return ret; |
393 | return fscrypt_ioctl_get_policy(filp: file, arg: (void __user *)arg); |
394 | |
395 | case FS_IOC_GET_ENCRYPTION_POLICY_EX: |
396 | ret = vet_mds_for_fscrypt(file); |
397 | if (ret) |
398 | return ret; |
399 | return fscrypt_ioctl_get_policy_ex(filp: file, arg: (void __user *)arg); |
400 | |
401 | case FS_IOC_ADD_ENCRYPTION_KEY: |
402 | ret = vet_mds_for_fscrypt(file); |
403 | if (ret) |
404 | return ret; |
405 | return fscrypt_ioctl_add_key(filp: file, arg: (void __user *)arg); |
406 | |
407 | case FS_IOC_REMOVE_ENCRYPTION_KEY: |
408 | return fscrypt_ioctl_remove_key(filp: file, arg: (void __user *)arg); |
409 | |
410 | case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: |
411 | return fscrypt_ioctl_remove_key_all_users(filp: file, |
412 | arg: (void __user *)arg); |
413 | |
414 | case FS_IOC_GET_ENCRYPTION_KEY_STATUS: |
415 | return fscrypt_ioctl_get_key_status(filp: file, arg: (void __user *)arg); |
416 | |
417 | case FS_IOC_GET_ENCRYPTION_NONCE: |
418 | ret = vet_mds_for_fscrypt(file); |
419 | if (ret) |
420 | return ret; |
421 | return fscrypt_ioctl_get_nonce(filp: file, arg: (void __user *)arg); |
422 | } |
423 | |
424 | return -ENOTTY; |
425 | } |
426 | |