1 // SPDX-License-Identifier: GPL-2.0
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.
9 #include <linux/ceph/ceph_debug.h>
10 #include <linux/xattr.h>
11 #include <linux/fscrypt.h>
12 #include <linux/ceph/striper.h>
15 #include "mds_client.h"
19 * The base64url encoding used by fscrypt includes the '_' character, which may
20 * cause problems in snapshot names (which can not start with '_'). Thus, we
21 * used the base64 encoding defined for IMAP mailbox names (RFC 3501) instead,
22 * which replaces '-' and '_' by '+' and ','.
24 static const char base64_table[65] =
25 "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+,";
27 int ceph_base64_encode(const u8 *src, int srclen, char *dst)
34 for (i = 0; i < srclen; i++) {
35 ac = (ac << 8) | src[i];
39 *cp++ = base64_table[(ac >> bits) & 0x3f];
43 *cp++ = base64_table[(ac << (6 - bits)) & 0x3f];
47 int ceph_base64_decode(const char *src, int srclen, u8 *dst)
54 for (i = 0; i < srclen; i++) {
55 const char *p = strchr(base64_table, src[i]);
57 if (p == NULL || src[i] == 0)
59 ac = (ac << 6) | (p - base64_table);
63 *bp++ = (u8)(ac >> bits);
66 if (ac & ((1 << bits) - 1))
71 static int ceph_crypt_get_context(struct inode *inode, void *ctx, size_t len)
73 struct ceph_inode_info *ci = ceph_inode(inode);
74 struct ceph_fscrypt_auth *cfa = (struct ceph_fscrypt_auth *)ci->fscrypt_auth;
77 /* Non existent or too short? */
78 if (!cfa || (ci->fscrypt_auth_len < (offsetof(struct ceph_fscrypt_auth, cfa_blob) + 1)))
81 /* Some format we don't recognize? */
82 if (le32_to_cpu(cfa->cfa_version) != CEPH_FSCRYPT_AUTH_VERSION)
85 ctxlen = le32_to_cpu(cfa->cfa_blob_len);
89 memcpy(ctx, cfa->cfa_blob, ctxlen);
93 static int ceph_crypt_set_context(struct inode *inode, const void *ctx,
94 size_t len, void *fs_data)
97 struct iattr attr = { };
98 struct ceph_iattr cia = { };
99 struct ceph_fscrypt_auth *cfa;
101 WARN_ON_ONCE(fs_data);
103 if (len > FSCRYPT_SET_CONTEXT_MAX_SIZE)
106 cfa = kzalloc(sizeof(*cfa), GFP_KERNEL);
110 cfa->cfa_version = cpu_to_le32(CEPH_FSCRYPT_AUTH_VERSION);
111 cfa->cfa_blob_len = cpu_to_le32(len);
112 memcpy(cfa->cfa_blob, ctx, len);
114 cia.fscrypt_auth = cfa;
116 ret = __ceph_setattr(&nop_mnt_idmap, inode, &attr, &cia);
118 inode_set_flags(inode, S_ENCRYPTED, S_ENCRYPTED);
119 kfree(cia.fscrypt_auth);
123 static bool ceph_crypt_empty_dir(struct inode *inode)
125 struct ceph_inode_info *ci = ceph_inode(inode);
127 return ci->i_rsubdirs + ci->i_rfiles == 1;
130 static const union fscrypt_policy *ceph_get_dummy_policy(struct super_block *sb)
132 return ceph_sb_to_fs_client(sb)->fsc_dummy_enc_policy.policy;
135 static struct fscrypt_operations ceph_fscrypt_ops = {
136 .needs_bounce_pages = 1,
137 .get_context = ceph_crypt_get_context,
138 .set_context = ceph_crypt_set_context,
139 .get_dummy_policy = ceph_get_dummy_policy,
140 .empty_dir = ceph_crypt_empty_dir,
143 void ceph_fscrypt_set_ops(struct super_block *sb)
145 fscrypt_set_ops(sb, &ceph_fscrypt_ops);
148 void ceph_fscrypt_free_dummy_policy(struct ceph_fs_client *fsc)
150 fscrypt_free_dummy_policy(&fsc->fsc_dummy_enc_policy);
153 int ceph_fscrypt_prepare_context(struct inode *dir, struct inode *inode,
154 struct ceph_acl_sec_ctx *as)
157 bool encrypted = false;
158 struct ceph_inode_info *ci = ceph_inode(inode);
160 ret = fscrypt_prepare_new_inode(dir, inode, &encrypted);
166 as->fscrypt_auth = kzalloc(sizeof(*as->fscrypt_auth), GFP_KERNEL);
167 if (!as->fscrypt_auth)
170 ctxsize = fscrypt_context_for_new_inode(as->fscrypt_auth->cfa_blob,
175 as->fscrypt_auth->cfa_version = cpu_to_le32(CEPH_FSCRYPT_AUTH_VERSION);
176 as->fscrypt_auth->cfa_blob_len = cpu_to_le32(ctxsize);
178 WARN_ON_ONCE(ci->fscrypt_auth);
179 kfree(ci->fscrypt_auth);
180 ci->fscrypt_auth_len = ceph_fscrypt_auth_len(as->fscrypt_auth);
181 ci->fscrypt_auth = kmemdup(as->fscrypt_auth, ci->fscrypt_auth_len,
183 if (!ci->fscrypt_auth)
186 inode->i_flags |= S_ENCRYPTED;
191 void ceph_fscrypt_as_ctx_to_req(struct ceph_mds_request *req,
192 struct ceph_acl_sec_ctx *as)
194 swap(req->r_fscrypt_auth, as->fscrypt_auth);
198 * User-created snapshots can't start with '_'. Snapshots that start with this
199 * character are special (hint: there aren't real snapshots) and use the
202 * _<SNAPSHOT-NAME>_<INODE-NUMBER>
205 * - <SNAPSHOT-NAME> - the real snapshot name that may need to be decrypted,
206 * - <INODE-NUMBER> - the inode number (in decimal) for the actual snapshot
208 * This function parses these snapshot names and returns the inode
209 * <INODE-NUMBER>. 'name_len' will also bet set with the <SNAPSHOT-NAME>
212 static struct inode *parse_longname(const struct inode *parent,
213 const char *name, int *name_len)
215 struct ceph_client *cl = ceph_inode_to_client(parent);
216 struct inode *dir = NULL;
217 struct ceph_vino vino = { .snap = CEPH_NOSNAP };
220 int orig_len = *name_len;
223 /* Skip initial '_' */
225 name_end = strrchr(name, '_');
227 doutc(cl, "failed to parse long snapshot name: %s\n", name);
228 return ERR_PTR(-EIO);
230 *name_len = (name_end - name);
231 if (*name_len <= 0) {
232 pr_err_client(cl, "failed to parse long snapshot name\n");
233 return ERR_PTR(-EIO);
236 /* Get the inode number */
237 inode_number = kmemdup_nul(name_end + 1,
238 orig_len - *name_len - 2,
241 return ERR_PTR(-ENOMEM);
242 ret = kstrtou64(inode_number, 10, &vino.ino);
244 doutc(cl, "failed to parse inode number: %s\n", name);
249 /* And finally the inode */
250 dir = ceph_find_inode(parent->i_sb, vino);
252 /* This can happen if we're not mounting cephfs on the root */
253 dir = ceph_get_inode(parent->i_sb, vino, NULL);
255 doutc(cl, "can't find inode %s (%s)\n", inode_number, name);
263 int ceph_encode_encrypted_dname(struct inode *parent, struct qstr *d_name,
266 struct ceph_client *cl = ceph_inode_to_client(parent);
267 struct inode *dir = parent;
275 iname.name = d_name->name;
276 name_len = d_name->len;
278 /* Handle the special case of snapshot names that start with '_' */
279 if ((ceph_snap(dir) == CEPH_SNAPDIR) && (name_len > 0) &&
280 (iname.name[0] == '_')) {
281 dir = parse_longname(parent, iname.name, &name_len);
284 iname.name++; /* skip initial '_' */
286 iname.len = name_len;
288 if (!fscrypt_has_encryption_key(dir)) {
289 memcpy(buf, d_name->name, d_name->len);
295 * Convert cleartext d_name to ciphertext. If result is longer than
296 * CEPH_NOHASH_NAME_MAX, sha256 the remaining bytes
298 * See: fscrypt_setup_filename
300 if (!fscrypt_fname_encrypted_size(dir, iname.len, NAME_MAX, &len)) {
301 elen = -ENAMETOOLONG;
305 /* Allocate a buffer appropriate to hold the result */
306 cryptbuf = kmalloc(len > CEPH_NOHASH_NAME_MAX ? NAME_MAX : len,
313 ret = fscrypt_fname_encrypt(dir, &iname, cryptbuf, len);
319 /* hash the end if the name is long enough */
320 if (len > CEPH_NOHASH_NAME_MAX) {
321 u8 hash[SHA256_DIGEST_SIZE];
322 u8 *extra = cryptbuf + CEPH_NOHASH_NAME_MAX;
325 * hash the extra bytes and overwrite crypttext beyond that
328 sha256(extra, len - CEPH_NOHASH_NAME_MAX, hash);
329 memcpy(extra, hash, SHA256_DIGEST_SIZE);
330 len = CEPH_NOHASH_NAME_MAX + SHA256_DIGEST_SIZE;
333 /* base64 encode the encrypted name */
334 elen = ceph_base64_encode(cryptbuf, len, buf);
335 doutc(cl, "base64-encoded ciphertext name = %.*s\n", elen, buf);
337 /* To understand the 240 limit, see CEPH_NOHASH_NAME_MAX comments */
339 if ((elen > 0) && (dir != parent)) {
340 char tmp_buf[NAME_MAX];
342 elen = snprintf(tmp_buf, sizeof(tmp_buf), "_%.*s_%ld",
343 elen, buf, dir->i_ino);
344 memcpy(buf, tmp_buf, elen);
350 if ((dir->i_state & I_NEW))
351 discard_new_inode(dir);
358 int ceph_encode_encrypted_fname(struct inode *parent, struct dentry *dentry,
361 WARN_ON_ONCE(!fscrypt_has_encryption_key(parent));
363 return ceph_encode_encrypted_dname(parent, &dentry->d_name, buf);
367 * ceph_fname_to_usr - convert a filename for userland presentation
368 * @fname: ceph_fname to be converted
369 * @tname: temporary name buffer to use for conversion (may be NULL)
370 * @oname: where converted name should be placed
371 * @is_nokey: set to true if key wasn't available during conversion (may be NULL)
373 * Given a filename (usually from the MDS), format it for presentation to
374 * userland. If @parent is not encrypted, just pass it back as-is.
376 * Otherwise, base64 decode the string, and then ask fscrypt to format it
377 * for userland presentation.
379 * Returns 0 on success or negative error code on error.
381 int ceph_fname_to_usr(const struct ceph_fname *fname, struct fscrypt_str *tname,
382 struct fscrypt_str *oname, bool *is_nokey)
384 struct inode *dir = fname->dir;
385 struct fscrypt_str _tname = FSTR_INIT(NULL, 0);
386 struct fscrypt_str iname;
387 char *name = fname->name;
388 int name_len = fname->name_len;
391 /* Sanity check that the resulting name will fit in the buffer */
392 if (fname->name_len > NAME_MAX || fname->ctext_len > NAME_MAX)
395 /* Handle the special case of snapshot names that start with '_' */
396 if ((ceph_snap(dir) == CEPH_SNAPDIR) && (name_len > 0) &&
398 dir = parse_longname(dir, name, &name_len);
401 name++; /* skip initial '_' */
404 if (!IS_ENCRYPTED(dir)) {
405 oname->name = fname->name;
406 oname->len = fname->name_len;
411 ret = ceph_fscrypt_prepare_readdir(dir);
416 * Use the raw dentry name as sent by the MDS instead of
417 * generating a nokey name via fscrypt.
419 if (!fscrypt_has_encryption_key(dir)) {
421 oname->name = fname->name;
423 memcpy(oname->name, fname->name, fname->name_len);
424 oname->len = fname->name_len;
431 if (fname->ctext_len == 0) {
435 ret = fscrypt_fname_alloc_buffer(NAME_MAX, &_tname);
441 declen = ceph_base64_decode(name, name_len, tname->name);
446 iname.name = tname->name;
449 iname.name = fname->ctext;
450 iname.len = fname->ctext_len;
453 ret = fscrypt_fname_disk_to_usr(dir, 0, 0, &iname, oname);
454 if (!ret && (dir != fname->dir)) {
455 char tmp_buf[CEPH_BASE64_CHARS(NAME_MAX)];
457 name_len = snprintf(tmp_buf, sizeof(tmp_buf), "_%.*s_%ld",
458 oname->len, oname->name, dir->i_ino);
459 memcpy(oname->name, tmp_buf, name_len);
460 oname->len = name_len;
464 fscrypt_fname_free_buffer(&_tname);
466 if (dir != fname->dir) {
467 if ((dir->i_state & I_NEW))
468 discard_new_inode(dir);
476 * ceph_fscrypt_prepare_readdir - simple __fscrypt_prepare_readdir() wrapper
477 * @dir: directory inode for readdir prep
479 * Simple wrapper around __fscrypt_prepare_readdir() that will mark directory as
480 * non-complete if this call results in having the directory unlocked.
483 * 1 - if directory was locked and key is now loaded (i.e. dir is unlocked)
484 * 0 - if directory is still locked
485 * < 0 - if __fscrypt_prepare_readdir() fails
487 int ceph_fscrypt_prepare_readdir(struct inode *dir)
489 bool had_key = fscrypt_has_encryption_key(dir);
492 if (!IS_ENCRYPTED(dir))
495 err = __fscrypt_prepare_readdir(dir);
498 if (!had_key && fscrypt_has_encryption_key(dir)) {
499 /* directory just got unlocked, mark it as not complete */
500 ceph_dir_clear_complete(dir);
506 int ceph_fscrypt_decrypt_block_inplace(const struct inode *inode,
507 struct page *page, unsigned int len,
508 unsigned int offs, u64 lblk_num)
510 struct ceph_client *cl = ceph_inode_to_client(inode);
512 doutc(cl, "%p %llx.%llx len %u offs %u blk %llu\n", inode,
513 ceph_vinop(inode), len, offs, lblk_num);
514 return fscrypt_decrypt_block_inplace(inode, page, len, offs, lblk_num);
517 int ceph_fscrypt_encrypt_block_inplace(const struct inode *inode,
518 struct page *page, unsigned int len,
519 unsigned int offs, u64 lblk_num,
522 struct ceph_client *cl = ceph_inode_to_client(inode);
524 doutc(cl, "%p %llx.%llx len %u offs %u blk %llu\n", inode,
525 ceph_vinop(inode), len, offs, lblk_num);
526 return fscrypt_encrypt_block_inplace(inode, page, len, offs, lblk_num,
531 * ceph_fscrypt_decrypt_pages - decrypt an array of pages
532 * @inode: pointer to inode associated with these pages
533 * @page: pointer to page array
534 * @off: offset into the file that the read data starts
535 * @len: max length to decrypt
537 * Decrypt an array of fscrypt'ed pages and return the amount of
538 * data decrypted. Any data in the page prior to the start of the
539 * first complete block in the read is ignored. Any incomplete
540 * crypto blocks at the end of the array are ignored (and should
541 * probably be zeroed by the caller).
543 * Returns the length of the decrypted data or a negative errno.
545 int ceph_fscrypt_decrypt_pages(struct inode *inode, struct page **page,
549 u64 baseblk = off >> CEPH_FSCRYPT_BLOCK_SHIFT;
553 * We can't deal with partial blocks on an encrypted file, so mask off
556 num_blocks = ceph_fscrypt_blocks(off, len & CEPH_FSCRYPT_BLOCK_MASK);
558 /* Decrypt each block */
559 for (i = 0; i < num_blocks; ++i) {
560 int blkoff = i << CEPH_FSCRYPT_BLOCK_SHIFT;
561 int pgidx = blkoff >> PAGE_SHIFT;
562 unsigned int pgoffs = offset_in_page(blkoff);
565 fret = ceph_fscrypt_decrypt_block_inplace(inode, page[pgidx],
566 CEPH_FSCRYPT_BLOCK_SIZE, pgoffs,
573 ret += CEPH_FSCRYPT_BLOCK_SIZE;
579 * ceph_fscrypt_decrypt_extents: decrypt received extents in given buffer
580 * @inode: inode associated with pages being decrypted
581 * @page: pointer to page array
582 * @off: offset into the file that the data in page[0] starts
583 * @map: pointer to extent array
584 * @ext_cnt: length of extent array
586 * Given an extent map and a page array, decrypt the received data in-place,
587 * skipping holes. Returns the offset into buffer of end of last decrypted
590 int ceph_fscrypt_decrypt_extents(struct inode *inode, struct page **page,
591 u64 off, struct ceph_sparse_extent *map,
594 struct ceph_client *cl = ceph_inode_to_client(inode);
596 struct ceph_inode_info *ci = ceph_inode(inode);
600 /* Nothing to do for empty array */
602 doutc(cl, "%p %llx.%llx empty array, ret 0\n", inode,
607 ceph_calc_file_object_mapping(&ci->i_layout, off, map[0].len,
608 &objno, &objoff, &xlen);
610 for (i = 0; i < ext_cnt; ++i) {
611 struct ceph_sparse_extent *ext = &map[i];
612 int pgsoff = ext->off - objoff;
613 int pgidx = pgsoff >> PAGE_SHIFT;
616 if ((ext->off | ext->len) & ~CEPH_FSCRYPT_BLOCK_MASK) {
618 "%p %llx.%llx bad encrypted sparse extent "
619 "idx %d off %llx len %llx\n",
620 inode, ceph_vinop(inode), i, ext->off,
624 fret = ceph_fscrypt_decrypt_pages(inode, &page[pgidx],
625 off + pgsoff, ext->len);
626 doutc(cl, "%p %llx.%llx [%d] 0x%llx~0x%llx fret %d\n", inode,
627 ceph_vinop(inode), i, ext->off, ext->len, fret);
635 doutc(cl, "ret %d\n", ret);
640 * ceph_fscrypt_encrypt_pages - encrypt an array of pages
641 * @inode: pointer to inode associated with these pages
642 * @page: pointer to page array
643 * @off: offset into the file that the data starts
644 * @len: max length to encrypt
645 * @gfp: gfp flags to use for allocation
647 * Decrypt an array of cleartext pages and return the amount of
648 * data encrypted. Any data in the page prior to the start of the
649 * first complete block in the read is ignored. Any incomplete
650 * crypto blocks at the end of the array are ignored.
652 * Returns the length of the encrypted data or a negative errno.
654 int ceph_fscrypt_encrypt_pages(struct inode *inode, struct page **page, u64 off,
658 u64 baseblk = off >> CEPH_FSCRYPT_BLOCK_SHIFT;
662 * We can't deal with partial blocks on an encrypted file, so mask off
665 num_blocks = ceph_fscrypt_blocks(off, len & CEPH_FSCRYPT_BLOCK_MASK);
667 /* Encrypt each block */
668 for (i = 0; i < num_blocks; ++i) {
669 int blkoff = i << CEPH_FSCRYPT_BLOCK_SHIFT;
670 int pgidx = blkoff >> PAGE_SHIFT;
671 unsigned int pgoffs = offset_in_page(blkoff);
674 fret = ceph_fscrypt_encrypt_block_inplace(inode, page[pgidx],
675 CEPH_FSCRYPT_BLOCK_SIZE, pgoffs,
682 ret += CEPH_FSCRYPT_BLOCK_SIZE;