1 // SPDX-License-Identifier: GPL-2.0
3 * fs/ext4/verity.c: fs-verity support for ext4
5 * Copyright 2019 Google LLC
9 * Implementation of fsverity_operations for ext4.
11 * ext4 stores the verity metadata (Merkle tree and fsverity_descriptor) past
12 * the end of the file, starting at the first 64K boundary beyond i_size. This
13 * approach works because (a) verity files are readonly, and (b) pages fully
14 * beyond i_size aren't visible to userspace but can be read/written internally
15 * by ext4 with only some relatively small changes to ext4. This approach
16 * avoids having to depend on the EA_INODE feature and on rearchitecturing
17 * ext4's xattr support to support paging multi-gigabyte xattrs into memory, and
18 * to support encrypting xattrs. Note that the verity metadata *must* be
19 * encrypted when the file is, since it contains hashes of the plaintext data.
21 * Using a 64K boundary rather than a 4K one keeps things ready for
22 * architectures with 64K pages, and it doesn't necessarily waste space on-disk
23 * since there can be a hole between i_size and the start of the Merkle tree.
26 #include <linux/quotaops.h>
29 #include "ext4_extents.h"
30 #include "ext4_jbd2.h"
32 static inline loff_t ext4_verity_metadata_pos(const struct inode *inode)
34 return round_up(inode->i_size, 65536);
38 * Read some verity metadata from the inode. __vfs_read() can't be used because
39 * we need to read beyond i_size.
41 static int pagecache_read(struct inode *inode, void *buf, size_t count,
45 size_t n = min_t(size_t, count,
46 PAGE_SIZE - offset_in_page(pos));
49 page = read_mapping_page(inode->i_mapping, pos >> PAGE_SHIFT,
54 memcpy_from_page(buf, page, offset_in_page(pos), n);
66 * Write some verity metadata to the inode for FS_IOC_ENABLE_VERITY.
67 * kernel_write() can't be used because the file descriptor is readonly.
69 static int pagecache_write(struct inode *inode, const void *buf, size_t count,
72 struct address_space *mapping = inode->i_mapping;
73 const struct address_space_operations *aops = mapping->a_ops;
75 if (pos + count > inode->i_sb->s_maxbytes)
79 size_t n = min_t(size_t, count,
80 PAGE_SIZE - offset_in_page(pos));
85 res = aops->write_begin(NULL, mapping, pos, n, &page, &fsdata);
89 memcpy_to_page(page, offset_in_page(pos), buf, n);
91 res = aops->write_end(NULL, mapping, pos, n, n, page, fsdata);
104 static int ext4_begin_enable_verity(struct file *filp)
106 struct inode *inode = file_inode(filp);
107 const int credits = 2; /* superblock and inode for ext4_orphan_add() */
111 if (IS_DAX(inode) || ext4_test_inode_flag(inode, EXT4_INODE_DAX))
114 if (ext4_verity_in_progress(inode))
118 * Since the file was opened readonly, we have to initialize the jbd
119 * inode and quotas here and not rely on ->open() doing it. This must
120 * be done before evicting the inline data.
123 err = ext4_inode_attach_jinode(inode);
127 err = dquot_initialize(inode);
131 err = ext4_convert_inline_data(inode);
135 if (!ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
136 ext4_warning_inode(inode,
137 "verity is only allowed on extent-based files");
142 * ext4 uses the last allocated block to find the verity descriptor, so
143 * we must remove any other blocks past EOF which might confuse things.
145 err = ext4_truncate(inode);
149 handle = ext4_journal_start(inode, EXT4_HT_INODE, credits);
151 return PTR_ERR(handle);
153 err = ext4_orphan_add(handle, inode);
155 ext4_set_inode_state(inode, EXT4_STATE_VERITY_IN_PROGRESS);
157 ext4_journal_stop(handle);
162 * ext4 stores the verity descriptor beginning on the next filesystem block
163 * boundary after the Merkle tree. Then, the descriptor size is stored in the
164 * last 4 bytes of the last allocated filesystem block --- which is either the
165 * block in which the descriptor ends, or the next block after that if there
166 * weren't at least 4 bytes remaining.
168 * We can't simply store the descriptor in an xattr because it *must* be
169 * encrypted when ext4 encryption is used, but ext4 encryption doesn't encrypt
170 * xattrs. Also, if the descriptor includes a large signature blob it may be
171 * too large to store in an xattr without the EA_INODE feature.
173 static int ext4_write_verity_descriptor(struct inode *inode, const void *desc,
174 size_t desc_size, u64 merkle_tree_size)
176 const u64 desc_pos = round_up(ext4_verity_metadata_pos(inode) +
177 merkle_tree_size, i_blocksize(inode));
178 const u64 desc_end = desc_pos + desc_size;
179 const __le32 desc_size_disk = cpu_to_le32(desc_size);
180 const u64 desc_size_pos = round_up(desc_end + sizeof(desc_size_disk),
181 i_blocksize(inode)) -
182 sizeof(desc_size_disk);
185 err = pagecache_write(inode, desc, desc_size, desc_pos);
189 return pagecache_write(inode, &desc_size_disk, sizeof(desc_size_disk),
193 static int ext4_end_enable_verity(struct file *filp, const void *desc,
194 size_t desc_size, u64 merkle_tree_size)
196 struct inode *inode = file_inode(filp);
197 const int credits = 2; /* superblock and inode for ext4_orphan_del() */
199 struct ext4_iloc iloc;
203 * If an error already occurred (which fs/verity/ signals by passing
204 * desc == NULL), then only clean-up is needed.
209 /* Append the verity descriptor. */
210 err = ext4_write_verity_descriptor(inode, desc, desc_size,
216 * Write all pages (both data and verity metadata). Note that this must
217 * happen before clearing EXT4_STATE_VERITY_IN_PROGRESS; otherwise pages
218 * beyond i_size won't be written properly. For crash consistency, this
219 * also must happen before the verity inode flag gets persisted.
221 err = filemap_write_and_wait(inode->i_mapping);
226 * Finally, set the verity inode flag and remove the inode from the
227 * orphan list (in a single transaction).
230 handle = ext4_journal_start(inode, EXT4_HT_INODE, credits);
231 if (IS_ERR(handle)) {
232 err = PTR_ERR(handle);
236 err = ext4_orphan_del(handle, inode);
238 goto stop_and_cleanup;
240 err = ext4_reserve_inode_write(handle, inode, &iloc);
242 goto stop_and_cleanup;
244 ext4_set_inode_flag(inode, EXT4_INODE_VERITY);
245 ext4_set_inode_flags(inode, false);
246 err = ext4_mark_iloc_dirty(handle, inode, &iloc);
248 goto stop_and_cleanup;
250 ext4_journal_stop(handle);
252 ext4_clear_inode_state(inode, EXT4_STATE_VERITY_IN_PROGRESS);
256 ext4_journal_stop(handle);
259 * Verity failed to be enabled, so clean up by truncating any verity
260 * metadata that was written beyond i_size (both from cache and from
261 * disk), removing the inode from the orphan list (if it wasn't done
262 * already), and clearing EXT4_STATE_VERITY_IN_PROGRESS.
264 truncate_inode_pages(inode->i_mapping, inode->i_size);
265 ext4_truncate(inode);
266 ext4_orphan_del(NULL, inode);
267 ext4_clear_inode_state(inode, EXT4_STATE_VERITY_IN_PROGRESS);
271 static int ext4_get_verity_descriptor_location(struct inode *inode,
272 size_t *desc_size_ret,
275 struct ext4_ext_path *path;
276 struct ext4_extent *last_extent;
279 __le32 desc_size_disk;
285 * Descriptor size is in last 4 bytes of last allocated block.
286 * See ext4_write_verity_descriptor().
289 if (!ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
290 EXT4_ERROR_INODE(inode, "verity file doesn't use extents");
291 return -EFSCORRUPTED;
294 path = ext4_find_extent(inode, EXT_MAX_BLOCKS - 1, NULL, 0);
296 return PTR_ERR(path);
298 last_extent = path[path->p_depth].p_ext;
300 EXT4_ERROR_INODE(inode, "verity file has no extents");
301 ext4_free_ext_path(path);
302 return -EFSCORRUPTED;
305 end_lblk = le32_to_cpu(last_extent->ee_block) +
306 ext4_ext_get_actual_len(last_extent);
307 desc_size_pos = (u64)end_lblk << inode->i_blkbits;
308 ext4_free_ext_path(path);
310 if (desc_size_pos < sizeof(desc_size_disk))
312 desc_size_pos -= sizeof(desc_size_disk);
314 err = pagecache_read(inode, &desc_size_disk, sizeof(desc_size_disk),
318 desc_size = le32_to_cpu(desc_size_disk);
321 * The descriptor is stored just before the desc_size_disk, but starting
322 * on a filesystem block boundary.
325 if (desc_size > INT_MAX || desc_size > desc_size_pos)
328 desc_pos = round_down(desc_size_pos - desc_size, i_blocksize(inode));
329 if (desc_pos < ext4_verity_metadata_pos(inode))
332 *desc_size_ret = desc_size;
333 *desc_pos_ret = desc_pos;
337 EXT4_ERROR_INODE(inode, "verity file corrupted; can't find descriptor");
338 return -EFSCORRUPTED;
341 static int ext4_get_verity_descriptor(struct inode *inode, void *buf,
344 size_t desc_size = 0;
348 err = ext4_get_verity_descriptor_location(inode, &desc_size, &desc_pos);
353 if (desc_size > buf_size)
355 err = pagecache_read(inode, buf, desc_size, desc_pos);
362 static struct page *ext4_read_merkle_tree_page(struct inode *inode,
364 unsigned long num_ra_pages)
368 index += ext4_verity_metadata_pos(inode) >> PAGE_SHIFT;
370 page = find_get_page_flags(inode->i_mapping, index, FGP_ACCESSED);
371 if (!page || !PageUptodate(page)) {
372 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, index);
376 else if (num_ra_pages > 1)
377 page_cache_ra_unbounded(&ractl, num_ra_pages, 0);
378 page = read_mapping_page(inode->i_mapping, index, NULL);
383 static int ext4_write_merkle_tree_block(struct inode *inode, const void *buf,
384 u64 index, int log_blocksize)
386 loff_t pos = ext4_verity_metadata_pos(inode) + (index << log_blocksize);
388 return pagecache_write(inode, buf, 1 << log_blocksize, pos);
391 const struct fsverity_operations ext4_verityops = {
392 .begin_enable_verity = ext4_begin_enable_verity,
393 .end_enable_verity = ext4_end_enable_verity,
394 .get_verity_descriptor = ext4_get_verity_descriptor,
395 .read_merkle_tree_page = ext4_read_merkle_tree_page,
396 .write_merkle_tree_block = ext4_write_merkle_tree_block,