1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * The base64 encode/decode code was copied from fscrypt:
4 * Copyright (C) 2015, Google, Inc.
5 * Copyright (C) 2015, Motorola Mobility
6 * Written by Uday Savagaonkar, 2014.
7 * Modified by Jaegeuk Kim, 2015.
8 */
9 #include <linux/ceph/ceph_debug.h>
10 #include <linux/xattr.h>
11 #include <linux/fscrypt.h>
12 #include <linux/ceph/striper.h>
13
14 #include "super.h"
15 #include "mds_client.h"
16 #include "crypto.h"
17
ceph_crypt_get_context(struct inode * inode,void * ctx,size_t len)18 static int ceph_crypt_get_context(struct inode *inode, void *ctx, size_t len)
19 {
20 struct ceph_inode_info *ci = ceph_inode(inode);
21 struct ceph_fscrypt_auth *cfa = (struct ceph_fscrypt_auth *)ci->fscrypt_auth;
22 u32 ctxlen;
23
24 /* Non existent or too short? */
25 if (!cfa || (ci->fscrypt_auth_len < (offsetof(struct ceph_fscrypt_auth, cfa_blob) + 1)))
26 return -ENOBUFS;
27
28 /* Some format we don't recognize? */
29 if (le32_to_cpu(cfa->cfa_version) != CEPH_FSCRYPT_AUTH_VERSION)
30 return -ENOBUFS;
31
32 ctxlen = le32_to_cpu(cfa->cfa_blob_len);
33 if (len < ctxlen)
34 return -ERANGE;
35
36 memcpy(ctx, cfa->cfa_blob, ctxlen);
37 return ctxlen;
38 }
39
ceph_crypt_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)40 static int ceph_crypt_set_context(struct inode *inode, const void *ctx,
41 size_t len, void *fs_data)
42 {
43 int ret;
44 struct iattr attr = { };
45 struct ceph_iattr cia = { };
46 struct ceph_fscrypt_auth *cfa;
47
48 WARN_ON_ONCE(fs_data);
49
50 if (len > FSCRYPT_SET_CONTEXT_MAX_SIZE)
51 return -EINVAL;
52
53 cfa = kzalloc(sizeof(*cfa), GFP_KERNEL);
54 if (!cfa)
55 return -ENOMEM;
56
57 cfa->cfa_version = cpu_to_le32(CEPH_FSCRYPT_AUTH_VERSION);
58 cfa->cfa_blob_len = cpu_to_le32(len);
59 memcpy(cfa->cfa_blob, ctx, len);
60
61 cia.fscrypt_auth = cfa;
62
63 ret = __ceph_setattr(&nop_mnt_idmap, inode, &attr, &cia);
64 if (ret == 0)
65 inode_set_flags(inode, S_ENCRYPTED, S_ENCRYPTED);
66 kfree(cia.fscrypt_auth);
67 return ret;
68 }
69
ceph_crypt_empty_dir(struct inode * inode)70 static bool ceph_crypt_empty_dir(struct inode *inode)
71 {
72 struct ceph_inode_info *ci = ceph_inode(inode);
73
74 return ci->i_rsubdirs + ci->i_rfiles == 1;
75 }
76
ceph_get_dummy_policy(struct super_block * sb)77 static const union fscrypt_policy *ceph_get_dummy_policy(struct super_block *sb)
78 {
79 return ceph_sb_to_fs_client(sb)->fsc_dummy_enc_policy.policy;
80 }
81
82 static struct fscrypt_operations ceph_fscrypt_ops = {
83 .inode_info_offs = (int)offsetof(struct ceph_inode_info, i_crypt_info) -
84 (int)offsetof(struct ceph_inode_info, netfs.inode),
85 .needs_bounce_pages = 1,
86 .get_context = ceph_crypt_get_context,
87 .set_context = ceph_crypt_set_context,
88 .get_dummy_policy = ceph_get_dummy_policy,
89 .empty_dir = ceph_crypt_empty_dir,
90 };
91
ceph_fscrypt_set_ops(struct super_block * sb)92 void ceph_fscrypt_set_ops(struct super_block *sb)
93 {
94 fscrypt_set_ops(sb, &ceph_fscrypt_ops);
95 }
96
ceph_fscrypt_free_dummy_policy(struct ceph_fs_client * fsc)97 void ceph_fscrypt_free_dummy_policy(struct ceph_fs_client *fsc)
98 {
99 fscrypt_free_dummy_policy(&fsc->fsc_dummy_enc_policy);
100 }
101
ceph_fscrypt_prepare_context(struct inode * dir,struct inode * inode,struct ceph_acl_sec_ctx * as)102 int ceph_fscrypt_prepare_context(struct inode *dir, struct inode *inode,
103 struct ceph_acl_sec_ctx *as)
104 {
105 int ret, ctxsize;
106 bool encrypted = false;
107 struct ceph_inode_info *ci = ceph_inode(inode);
108
109 ret = fscrypt_prepare_new_inode(dir, inode, &encrypted);
110 if (ret)
111 return ret;
112 if (!encrypted)
113 return 0;
114
115 as->fscrypt_auth = kzalloc(sizeof(*as->fscrypt_auth), GFP_KERNEL);
116 if (!as->fscrypt_auth)
117 return -ENOMEM;
118
119 ctxsize = fscrypt_context_for_new_inode(as->fscrypt_auth->cfa_blob,
120 inode);
121 if (ctxsize < 0)
122 return ctxsize;
123
124 as->fscrypt_auth->cfa_version = cpu_to_le32(CEPH_FSCRYPT_AUTH_VERSION);
125 as->fscrypt_auth->cfa_blob_len = cpu_to_le32(ctxsize);
126
127 WARN_ON_ONCE(ci->fscrypt_auth);
128 kfree(ci->fscrypt_auth);
129 ci->fscrypt_auth_len = ceph_fscrypt_auth_len(as->fscrypt_auth);
130 ci->fscrypt_auth = kmemdup(as->fscrypt_auth, ci->fscrypt_auth_len,
131 GFP_KERNEL);
132 if (!ci->fscrypt_auth)
133 return -ENOMEM;
134
135 inode->i_flags |= S_ENCRYPTED;
136
137 return 0;
138 }
139
ceph_fscrypt_as_ctx_to_req(struct ceph_mds_request * req,struct ceph_acl_sec_ctx * as)140 void ceph_fscrypt_as_ctx_to_req(struct ceph_mds_request *req,
141 struct ceph_acl_sec_ctx *as)
142 {
143 swap(req->r_fscrypt_auth, as->fscrypt_auth);
144 }
145
146 /*
147 * User-created snapshots can't start with '_'. Snapshots that start with this
148 * character are special (hint: there aren't real snapshots) and use the
149 * following format:
150 *
151 * _<SNAPSHOT-NAME>_<INODE-NUMBER>
152 *
153 * where:
154 * - <SNAPSHOT-NAME> - the real snapshot name that may need to be decrypted,
155 * - <INODE-NUMBER> - the inode number (in decimal) for the actual snapshot
156 *
157 * This function parses these snapshot names and returns the inode
158 * <INODE-NUMBER>. 'name_len' will also bet set with the <SNAPSHOT-NAME>
159 * length.
160 */
parse_longname(const struct inode * parent,const char * name,int * name_len)161 static struct inode *parse_longname(const struct inode *parent,
162 const char *name, int *name_len)
163 {
164 struct ceph_client *cl = ceph_inode_to_client(parent);
165 struct inode *dir = NULL;
166 struct ceph_vino vino = { .snap = CEPH_NOSNAP };
167 char *name_end, *inode_number;
168 int ret = -EIO;
169 /* NUL-terminate */
170 char *str __free(kfree) = kmemdup_nul(name, *name_len, GFP_KERNEL);
171 if (!str)
172 return ERR_PTR(-ENOMEM);
173 /* Skip initial '_' */
174 str++;
175 name_end = strrchr(str, '_');
176 if (!name_end) {
177 doutc(cl, "failed to parse long snapshot name: %s\n", str);
178 return ERR_PTR(-EIO);
179 }
180 *name_len = (name_end - str);
181 if (*name_len <= 0) {
182 pr_err_client(cl, "failed to parse long snapshot name\n");
183 return ERR_PTR(-EIO);
184 }
185
186 /* Get the inode number */
187 inode_number = name_end + 1;
188 ret = kstrtou64(inode_number, 10, &vino.ino);
189 if (ret) {
190 doutc(cl, "failed to parse inode number: %s\n", str);
191 return ERR_PTR(ret);
192 }
193
194 /* And finally the inode */
195 dir = ceph_find_inode(parent->i_sb, vino);
196 if (!dir) {
197 /* This can happen if we're not mounting cephfs on the root */
198 dir = ceph_get_inode(parent->i_sb, vino, NULL);
199 if (IS_ERR(dir))
200 doutc(cl, "can't find inode %s (%s)\n", inode_number, name);
201 }
202 return dir;
203 }
204
ceph_encode_encrypted_dname(struct inode * parent,char * buf,int elen)205 int ceph_encode_encrypted_dname(struct inode *parent, char *buf, int elen)
206 {
207 struct ceph_client *cl = ceph_inode_to_client(parent);
208 struct inode *dir = parent;
209 char *p = buf;
210 u32 len;
211 int name_len = elen;
212 int ret;
213 u8 *cryptbuf = NULL;
214
215 /* Handle the special case of snapshot names that start with '_' */
216 if (ceph_snap(dir) == CEPH_SNAPDIR && *p == '_') {
217 dir = parse_longname(parent, p, &name_len);
218 if (IS_ERR(dir))
219 return PTR_ERR(dir);
220 p++; /* skip initial '_' */
221 }
222
223 if (!fscrypt_has_encryption_key(dir))
224 goto out;
225
226 /*
227 * Convert cleartext d_name to ciphertext. If result is longer than
228 * CEPH_NOHASH_NAME_MAX, sha256 the remaining bytes
229 *
230 * See: fscrypt_setup_filename
231 */
232 if (!fscrypt_fname_encrypted_size(dir, name_len, NAME_MAX, &len)) {
233 elen = -ENAMETOOLONG;
234 goto out;
235 }
236
237 /* Allocate a buffer appropriate to hold the result */
238 cryptbuf = kmalloc(len > CEPH_NOHASH_NAME_MAX ? NAME_MAX : len,
239 GFP_KERNEL);
240 if (!cryptbuf) {
241 elen = -ENOMEM;
242 goto out;
243 }
244
245 ret = fscrypt_fname_encrypt(dir,
246 &(struct qstr)QSTR_INIT(p, name_len),
247 cryptbuf, len);
248 if (ret) {
249 elen = ret;
250 goto out;
251 }
252
253 /* hash the end if the name is long enough */
254 if (len > CEPH_NOHASH_NAME_MAX) {
255 u8 hash[SHA256_DIGEST_SIZE];
256 u8 *extra = cryptbuf + CEPH_NOHASH_NAME_MAX;
257
258 /*
259 * hash the extra bytes and overwrite crypttext beyond that
260 * point with it
261 */
262 sha256(extra, len - CEPH_NOHASH_NAME_MAX, hash);
263 memcpy(extra, hash, SHA256_DIGEST_SIZE);
264 len = CEPH_NOHASH_NAME_MAX + SHA256_DIGEST_SIZE;
265 }
266
267 /* base64 encode the encrypted name */
268 elen = base64_encode(cryptbuf, len, p, false, BASE64_IMAP);
269 doutc(cl, "base64-encoded ciphertext name = %.*s\n", elen, p);
270
271 /* To understand the 240 limit, see CEPH_NOHASH_NAME_MAX comments */
272 WARN_ON(elen > 240);
273 if (dir != parent) // leading _ is already there; append _<inum>
274 elen += 1 + sprintf(p + elen, "_%ld", dir->i_ino);
275
276 out:
277 kfree(cryptbuf);
278 if (dir != parent) {
279 if ((inode_state_read_once(dir) & I_NEW))
280 discard_new_inode(dir);
281 else
282 iput(dir);
283 }
284 return elen;
285 }
286
287 /**
288 * ceph_fname_to_usr - convert a filename for userland presentation
289 * @fname: ceph_fname to be converted
290 * @tname: temporary name buffer to use for conversion (may be NULL)
291 * @oname: where converted name should be placed
292 * @is_nokey: set to true if key wasn't available during conversion (may be NULL)
293 *
294 * Given a filename (usually from the MDS), format it for presentation to
295 * userland. If @parent is not encrypted, just pass it back as-is.
296 *
297 * Otherwise, base64 decode the string, and then ask fscrypt to format it
298 * for userland presentation.
299 *
300 * Returns 0 on success or negative error code on error.
301 */
ceph_fname_to_usr(const struct ceph_fname * fname,struct fscrypt_str * tname,struct fscrypt_str * oname,bool * is_nokey)302 int ceph_fname_to_usr(const struct ceph_fname *fname, struct fscrypt_str *tname,
303 struct fscrypt_str *oname, bool *is_nokey)
304 {
305 struct inode *dir = fname->dir;
306 struct fscrypt_str _tname = FSTR_INIT(NULL, 0);
307 struct fscrypt_str iname;
308 char *name = fname->name;
309 int name_len = fname->name_len;
310 int ret;
311
312 /* Sanity check that the resulting name will fit in the buffer */
313 if (fname->name_len > NAME_MAX || fname->ctext_len > NAME_MAX)
314 return -EIO;
315
316 /* Handle the special case of snapshot names that start with '_' */
317 if ((ceph_snap(dir) == CEPH_SNAPDIR) && (name_len > 0) &&
318 (name[0] == '_')) {
319 dir = parse_longname(dir, name, &name_len);
320 if (IS_ERR(dir))
321 return PTR_ERR(dir);
322 name++; /* skip initial '_' */
323 }
324
325 if (!IS_ENCRYPTED(dir)) {
326 oname->name = fname->name;
327 oname->len = fname->name_len;
328 ret = 0;
329 goto out_inode;
330 }
331
332 ret = ceph_fscrypt_prepare_readdir(dir);
333 if (ret)
334 goto out_inode;
335
336 /*
337 * Use the raw dentry name as sent by the MDS instead of
338 * generating a nokey name via fscrypt.
339 */
340 if (!fscrypt_has_encryption_key(dir)) {
341 if (fname->no_copy)
342 oname->name = fname->name;
343 else
344 memcpy(oname->name, fname->name, fname->name_len);
345 oname->len = fname->name_len;
346 if (is_nokey)
347 *is_nokey = true;
348 ret = 0;
349 goto out_inode;
350 }
351
352 if (fname->ctext_len == 0) {
353 int declen;
354
355 if (!tname) {
356 ret = fscrypt_fname_alloc_buffer(NAME_MAX, &_tname);
357 if (ret)
358 goto out_inode;
359 tname = &_tname;
360 }
361
362 declen = base64_decode(name, name_len,
363 tname->name, false, BASE64_IMAP);
364 if (declen <= 0) {
365 ret = -EIO;
366 goto out;
367 }
368 iname.name = tname->name;
369 iname.len = declen;
370 } else {
371 iname.name = fname->ctext;
372 iname.len = fname->ctext_len;
373 }
374
375 ret = fscrypt_fname_disk_to_usr(dir, 0, 0, &iname, oname);
376 if (!ret && (dir != fname->dir)) {
377 char tmp_buf[BASE64_CHARS(NAME_MAX)];
378
379 name_len = snprintf(tmp_buf, sizeof(tmp_buf), "_%.*s_%ld",
380 oname->len, oname->name, dir->i_ino);
381 memcpy(oname->name, tmp_buf, name_len);
382 oname->len = name_len;
383 }
384
385 out:
386 fscrypt_fname_free_buffer(&_tname);
387 out_inode:
388 if (dir != fname->dir) {
389 if ((inode_state_read_once(dir) & I_NEW))
390 discard_new_inode(dir);
391 else
392 iput(dir);
393 }
394 return ret;
395 }
396
397 /**
398 * ceph_fscrypt_prepare_readdir - simple __fscrypt_prepare_readdir() wrapper
399 * @dir: directory inode for readdir prep
400 *
401 * Simple wrapper around __fscrypt_prepare_readdir() that will mark directory as
402 * non-complete if this call results in having the directory unlocked.
403 *
404 * Returns:
405 * 1 - if directory was locked and key is now loaded (i.e. dir is unlocked)
406 * 0 - if directory is still locked
407 * < 0 - if __fscrypt_prepare_readdir() fails
408 */
ceph_fscrypt_prepare_readdir(struct inode * dir)409 int ceph_fscrypt_prepare_readdir(struct inode *dir)
410 {
411 bool had_key = fscrypt_has_encryption_key(dir);
412 int err;
413
414 if (!IS_ENCRYPTED(dir))
415 return 0;
416
417 err = __fscrypt_prepare_readdir(dir);
418 if (err)
419 return err;
420 if (!had_key && fscrypt_has_encryption_key(dir)) {
421 /* directory just got unlocked, mark it as not complete */
422 ceph_dir_clear_complete(dir);
423 return 1;
424 }
425 return 0;
426 }
427
ceph_fscrypt_decrypt_block_inplace(const struct inode * inode,struct page * page,unsigned int len,unsigned int offs,u64 lblk_num)428 int ceph_fscrypt_decrypt_block_inplace(const struct inode *inode,
429 struct page *page, unsigned int len,
430 unsigned int offs, u64 lblk_num)
431 {
432 struct ceph_client *cl = ceph_inode_to_client(inode);
433
434 doutc(cl, "%p %llx.%llx len %u offs %u blk %llu\n", inode,
435 ceph_vinop(inode), len, offs, lblk_num);
436 return fscrypt_decrypt_block_inplace(inode, page, len, offs, lblk_num);
437 }
438
ceph_fscrypt_encrypt_block_inplace(const struct inode * inode,struct page * page,unsigned int len,unsigned int offs,u64 lblk_num)439 int ceph_fscrypt_encrypt_block_inplace(const struct inode *inode,
440 struct page *page, unsigned int len,
441 unsigned int offs, u64 lblk_num)
442 {
443 struct ceph_client *cl = ceph_inode_to_client(inode);
444
445 doutc(cl, "%p %llx.%llx len %u offs %u blk %llu\n", inode,
446 ceph_vinop(inode), len, offs, lblk_num);
447 return fscrypt_encrypt_block_inplace(inode, page, len, offs, lblk_num);
448 }
449
450 /**
451 * ceph_fscrypt_decrypt_pages - decrypt an array of pages
452 * @inode: pointer to inode associated with these pages
453 * @page: pointer to page array
454 * @off: offset into the file that the read data starts
455 * @len: max length to decrypt
456 *
457 * Decrypt an array of fscrypt'ed pages and return the amount of
458 * data decrypted. Any data in the page prior to the start of the
459 * first complete block in the read is ignored. Any incomplete
460 * crypto blocks at the end of the array are ignored (and should
461 * probably be zeroed by the caller).
462 *
463 * Returns the length of the decrypted data or a negative errno.
464 */
ceph_fscrypt_decrypt_pages(struct inode * inode,struct page ** page,u64 off,int len)465 int ceph_fscrypt_decrypt_pages(struct inode *inode, struct page **page,
466 u64 off, int len)
467 {
468 int i, num_blocks;
469 u64 baseblk = off >> CEPH_FSCRYPT_BLOCK_SHIFT;
470 int ret = 0;
471
472 /*
473 * We can't deal with partial blocks on an encrypted file, so mask off
474 * the last bit.
475 */
476 num_blocks = ceph_fscrypt_blocks(off, len & CEPH_FSCRYPT_BLOCK_MASK);
477
478 /* Decrypt each block */
479 for (i = 0; i < num_blocks; ++i) {
480 int blkoff = i << CEPH_FSCRYPT_BLOCK_SHIFT;
481 int pgidx = blkoff >> PAGE_SHIFT;
482 unsigned int pgoffs = offset_in_page(blkoff);
483 int fret;
484
485 fret = ceph_fscrypt_decrypt_block_inplace(inode, page[pgidx],
486 CEPH_FSCRYPT_BLOCK_SIZE, pgoffs,
487 baseblk + i);
488 if (fret < 0) {
489 if (ret == 0)
490 ret = fret;
491 break;
492 }
493 ret += CEPH_FSCRYPT_BLOCK_SIZE;
494 }
495 return ret;
496 }
497
498 /**
499 * ceph_fscrypt_decrypt_extents: decrypt received extents in given buffer
500 * @inode: inode associated with pages being decrypted
501 * @page: pointer to page array
502 * @off: offset into the file that the data in page[0] starts
503 * @map: pointer to extent array
504 * @ext_cnt: length of extent array
505 *
506 * Given an extent map and a page array, decrypt the received data in-place,
507 * skipping holes. Returns the offset into buffer of end of last decrypted
508 * block.
509 */
ceph_fscrypt_decrypt_extents(struct inode * inode,struct page ** page,u64 off,struct ceph_sparse_extent * map,u32 ext_cnt)510 int ceph_fscrypt_decrypt_extents(struct inode *inode, struct page **page,
511 u64 off, struct ceph_sparse_extent *map,
512 u32 ext_cnt)
513 {
514 struct ceph_client *cl = ceph_inode_to_client(inode);
515 int i, ret = 0;
516 struct ceph_inode_info *ci = ceph_inode(inode);
517 u64 objno, objoff;
518 u32 xlen;
519
520 /* Nothing to do for empty array */
521 if (ext_cnt == 0) {
522 doutc(cl, "%p %llx.%llx empty array, ret 0\n", inode,
523 ceph_vinop(inode));
524 return 0;
525 }
526
527 ceph_calc_file_object_mapping(&ci->i_layout, off, map[0].len,
528 &objno, &objoff, &xlen);
529
530 for (i = 0; i < ext_cnt; ++i) {
531 struct ceph_sparse_extent *ext = &map[i];
532 int pgsoff = ext->off - objoff;
533 int pgidx = pgsoff >> PAGE_SHIFT;
534 int fret;
535
536 if ((ext->off | ext->len) & ~CEPH_FSCRYPT_BLOCK_MASK) {
537 pr_warn_client(cl,
538 "%p %llx.%llx bad encrypted sparse extent "
539 "idx %d off %llx len %llx\n",
540 inode, ceph_vinop(inode), i, ext->off,
541 ext->len);
542 return -EIO;
543 }
544 fret = ceph_fscrypt_decrypt_pages(inode, &page[pgidx],
545 off + pgsoff, ext->len);
546 doutc(cl, "%p %llx.%llx [%d] 0x%llx~0x%llx fret %d\n", inode,
547 ceph_vinop(inode), i, ext->off, ext->len, fret);
548 if (fret < 0) {
549 if (ret == 0)
550 ret = fret;
551 break;
552 }
553 ret = pgsoff + fret;
554 }
555 doutc(cl, "ret %d\n", ret);
556 return ret;
557 }
558
559 /**
560 * ceph_fscrypt_encrypt_pages - encrypt an array of pages
561 * @inode: pointer to inode associated with these pages
562 * @page: pointer to page array
563 * @off: offset into the file that the data starts
564 * @len: max length to encrypt
565 *
566 * Encrypt an array of cleartext pages and return the amount of
567 * data encrypted. Any data in the page prior to the start of the
568 * first complete block in the read is ignored. Any incomplete
569 * crypto blocks at the end of the array are ignored.
570 *
571 * Returns the length of the encrypted data or a negative errno.
572 */
ceph_fscrypt_encrypt_pages(struct inode * inode,struct page ** page,u64 off,int len)573 int ceph_fscrypt_encrypt_pages(struct inode *inode, struct page **page, u64 off,
574 int len)
575 {
576 int i, num_blocks;
577 u64 baseblk = off >> CEPH_FSCRYPT_BLOCK_SHIFT;
578 int ret = 0;
579
580 /*
581 * We can't deal with partial blocks on an encrypted file, so mask off
582 * the last bit.
583 */
584 num_blocks = ceph_fscrypt_blocks(off, len & CEPH_FSCRYPT_BLOCK_MASK);
585
586 /* Encrypt each block */
587 for (i = 0; i < num_blocks; ++i) {
588 int blkoff = i << CEPH_FSCRYPT_BLOCK_SHIFT;
589 int pgidx = blkoff >> PAGE_SHIFT;
590 unsigned int pgoffs = offset_in_page(blkoff);
591 int fret;
592
593 fret = ceph_fscrypt_encrypt_block_inplace(inode, page[pgidx],
594 CEPH_FSCRYPT_BLOCK_SIZE, pgoffs,
595 baseblk + i);
596 if (fret < 0) {
597 if (ret == 0)
598 ret = fret;
599 break;
600 }
601 ret += CEPH_FSCRYPT_BLOCK_SIZE;
602 }
603 return ret;
604 }
605