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));
50 page = read_mapping_page(inode->i_mapping, pos >> PAGE_SHIFT,
55 addr = kmap_atomic(page);
56 memcpy(buf, addr + offset_in_page(pos), n);
69 * Write some verity metadata to the inode for FS_IOC_ENABLE_VERITY.
70 * kernel_write() can't be used because the file descriptor is readonly.
72 static int pagecache_write(struct inode *inode, const void *buf, size_t count,
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 = pagecache_write_begin(NULL, inode->i_mapping, pos, n, 0,
90 memcpy_to_page(page, offset_in_page(pos), buf, n);
92 res = pagecache_write_end(NULL, inode->i_mapping, pos, n, n,
106 static int ext4_begin_enable_verity(struct file *filp)
108 struct inode *inode = file_inode(filp);
109 const int credits = 2; /* superblock and inode for ext4_orphan_add() */
113 if (ext4_verity_in_progress(inode))
117 * Since the file was opened readonly, we have to initialize the jbd
118 * inode and quotas here and not rely on ->open() doing it. This must
119 * be done before evicting the inline data.
122 err = ext4_inode_attach_jinode(inode);
126 err = dquot_initialize(inode);
130 err = ext4_convert_inline_data(inode);
134 if (!ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
135 ext4_warning_inode(inode,
136 "verity is only allowed on extent-based files");
141 * ext4 uses the last allocated block to find the verity descriptor, so
142 * we must remove any other blocks past EOF which might confuse things.
144 err = ext4_truncate(inode);
148 handle = ext4_journal_start(inode, EXT4_HT_INODE, credits);
150 return PTR_ERR(handle);
152 err = ext4_orphan_add(handle, inode);
154 ext4_set_inode_state(inode, EXT4_STATE_VERITY_IN_PROGRESS);
156 ext4_journal_stop(handle);
161 * ext4 stores the verity descriptor beginning on the next filesystem block
162 * boundary after the Merkle tree. Then, the descriptor size is stored in the
163 * last 4 bytes of the last allocated filesystem block --- which is either the
164 * block in which the descriptor ends, or the next block after that if there
165 * weren't at least 4 bytes remaining.
167 * We can't simply store the descriptor in an xattr because it *must* be
168 * encrypted when ext4 encryption is used, but ext4 encryption doesn't encrypt
169 * xattrs. Also, if the descriptor includes a large signature blob it may be
170 * too large to store in an xattr without the EA_INODE feature.
172 static int ext4_write_verity_descriptor(struct inode *inode, const void *desc,
173 size_t desc_size, u64 merkle_tree_size)
175 const u64 desc_pos = round_up(ext4_verity_metadata_pos(inode) +
176 merkle_tree_size, i_blocksize(inode));
177 const u64 desc_end = desc_pos + desc_size;
178 const __le32 desc_size_disk = cpu_to_le32(desc_size);
179 const u64 desc_size_pos = round_up(desc_end + sizeof(desc_size_disk),
180 i_blocksize(inode)) -
181 sizeof(desc_size_disk);
184 err = pagecache_write(inode, desc, desc_size, desc_pos);
188 return pagecache_write(inode, &desc_size_disk, sizeof(desc_size_disk),
192 static int ext4_end_enable_verity(struct file *filp, const void *desc,
193 size_t desc_size, u64 merkle_tree_size)
195 struct inode *inode = file_inode(filp);
196 const int credits = 2; /* superblock and inode for ext4_orphan_del() */
198 struct ext4_iloc iloc;
202 * If an error already occurred (which fs/verity/ signals by passing
203 * desc == NULL), then only clean-up is needed.
208 /* Append the verity descriptor. */
209 err = ext4_write_verity_descriptor(inode, desc, desc_size,
215 * Write all pages (both data and verity metadata). Note that this must
216 * happen before clearing EXT4_STATE_VERITY_IN_PROGRESS; otherwise pages
217 * beyond i_size won't be written properly. For crash consistency, this
218 * also must happen before the verity inode flag gets persisted.
220 err = filemap_write_and_wait(inode->i_mapping);
225 * Finally, set the verity inode flag and remove the inode from the
226 * orphan list (in a single transaction).
229 handle = ext4_journal_start(inode, EXT4_HT_INODE, credits);
230 if (IS_ERR(handle)) {
231 err = PTR_ERR(handle);
235 err = ext4_orphan_del(handle, inode);
237 goto stop_and_cleanup;
239 err = ext4_reserve_inode_write(handle, inode, &iloc);
241 goto stop_and_cleanup;
243 ext4_set_inode_flag(inode, EXT4_INODE_VERITY);
244 ext4_set_inode_flags(inode);
245 err = ext4_mark_iloc_dirty(handle, inode, &iloc);
247 goto stop_and_cleanup;
249 ext4_journal_stop(handle);
251 ext4_clear_inode_state(inode, EXT4_STATE_VERITY_IN_PROGRESS);
255 ext4_journal_stop(handle);
258 * Verity failed to be enabled, so clean up by truncating any verity
259 * metadata that was written beyond i_size (both from cache and from
260 * disk), removing the inode from the orphan list (if it wasn't done
261 * already), and clearing EXT4_STATE_VERITY_IN_PROGRESS.
263 truncate_inode_pages(inode->i_mapping, inode->i_size);
264 ext4_truncate(inode);
265 ext4_orphan_del(NULL, inode);
266 ext4_clear_inode_state(inode, EXT4_STATE_VERITY_IN_PROGRESS);
270 static int ext4_get_verity_descriptor_location(struct inode *inode,
271 size_t *desc_size_ret,
274 struct ext4_ext_path *path;
275 struct ext4_extent *last_extent;
278 __le32 desc_size_disk;
284 * Descriptor size is in last 4 bytes of last allocated block.
285 * See ext4_write_verity_descriptor().
288 if (!ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
289 EXT4_ERROR_INODE(inode, "verity file doesn't use extents");
290 return -EFSCORRUPTED;
293 path = ext4_find_extent(inode, EXT_MAX_BLOCKS - 1, NULL, 0);
295 return PTR_ERR(path);
297 last_extent = path[path->p_depth].p_ext;
299 EXT4_ERROR_INODE(inode, "verity file has no extents");
300 ext4_ext_drop_refs(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_ext_drop_refs(path);
311 if (desc_size_pos < sizeof(desc_size_disk))
313 desc_size_pos -= sizeof(desc_size_disk);
315 err = pagecache_read(inode, &desc_size_disk, sizeof(desc_size_disk),
319 desc_size = le32_to_cpu(desc_size_disk);
322 * The descriptor is stored just before the desc_size_disk, but starting
323 * on a filesystem block boundary.
326 if (desc_size > INT_MAX || desc_size > desc_size_pos)
329 desc_pos = round_down(desc_size_pos - desc_size, i_blocksize(inode));
330 if (desc_pos < ext4_verity_metadata_pos(inode))
333 *desc_size_ret = desc_size;
334 *desc_pos_ret = desc_pos;
338 EXT4_ERROR_INODE(inode, "verity file corrupted; can't find descriptor");
339 return -EFSCORRUPTED;
342 static int ext4_get_verity_descriptor(struct inode *inode, void *buf,
345 size_t desc_size = 0;
349 err = ext4_get_verity_descriptor_location(inode, &desc_size, &desc_pos);
354 if (desc_size > buf_size)
356 err = pagecache_read(inode, buf, desc_size, desc_pos);
363 static struct page *ext4_read_merkle_tree_page(struct inode *inode,
366 index += ext4_verity_metadata_pos(inode) >> PAGE_SHIFT;
368 return read_mapping_page(inode->i_mapping, index, NULL);
371 static int ext4_write_merkle_tree_block(struct inode *inode, const void *buf,
372 u64 index, int log_blocksize)
374 loff_t pos = ext4_verity_metadata_pos(inode) + (index << log_blocksize);
376 return pagecache_write(inode, buf, 1 << log_blocksize, pos);
379 const struct fsverity_operations ext4_verityops = {
380 .begin_enable_verity = ext4_begin_enable_verity,
381 .end_enable_verity = ext4_end_enable_verity,
382 .get_verity_descriptor = ext4_get_verity_descriptor,
383 .read_merkle_tree_page = ext4_read_merkle_tree_page,
384 .write_merkle_tree_block = ext4_write_merkle_tree_block,