diff options
author | Theodore Ts'o <tytso@mit.edu> | 2009-05-17 15:38:01 -0400 |
---|---|---|
committer | Theodore Ts'o <tytso@mit.edu> | 2009-05-17 15:38:01 -0400 |
commit | 6fd058f7791087648c683eb8572edf3be3c4c23c (patch) | |
tree | 0d80791532d2d022c91f20013003716eaf0afb40 /fs/ext4/block_validity.c | |
parent | 2ac3b6e00acb46406c993d57921f86a594aafe08 (diff) |
ext4: Add a comprehensive block validity check to ext4_get_blocks()
To catch filesystem bugs or corruption which could lead to the
filesystem getting severly damaged, this patch adds a facility for
tracking all of the filesystem metadata blocks by contiguous regions
in a red-black tree. This allows quick searching of the tree to
locate extents which might overlap with filesystem metadata blocks.
This facility is also used by the multi-block allocator to assure that
it is not allocating blocks out of the system zone, as well as by the
routines used when reading indirect blocks and extents information
from disk to make sure their contents are valid.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Diffstat (limited to 'fs/ext4/block_validity.c')
-rw-r--r-- | fs/ext4/block_validity.c | 244 |
1 files changed, 244 insertions, 0 deletions
diff --git a/fs/ext4/block_validity.c b/fs/ext4/block_validity.c new file mode 100644 index 000000000000..50784ef07563 --- /dev/null +++ b/fs/ext4/block_validity.c | |||
@@ -0,0 +1,244 @@ | |||
1 | /* | ||
2 | * linux/fs/ext4/block_validity.c | ||
3 | * | ||
4 | * Copyright (C) 2009 | ||
5 | * Theodore Ts'o (tytso@mit.edu) | ||
6 | * | ||
7 | * Track which blocks in the filesystem are metadata blocks that | ||
8 | * should never be used as data blocks by files or directories. | ||
9 | */ | ||
10 | |||
11 | #include <linux/time.h> | ||
12 | #include <linux/fs.h> | ||
13 | #include <linux/namei.h> | ||
14 | #include <linux/quotaops.h> | ||
15 | #include <linux/buffer_head.h> | ||
16 | #include <linux/module.h> | ||
17 | #include <linux/swap.h> | ||
18 | #include <linux/pagemap.h> | ||
19 | #include <linux/version.h> | ||
20 | #include <linux/blkdev.h> | ||
21 | #include <linux/mutex.h> | ||
22 | #include "ext4.h" | ||
23 | |||
24 | struct ext4_system_zone { | ||
25 | struct rb_node node; | ||
26 | ext4_fsblk_t start_blk; | ||
27 | unsigned int count; | ||
28 | }; | ||
29 | |||
30 | static struct kmem_cache *ext4_system_zone_cachep; | ||
31 | |||
32 | int __init init_ext4_system_zone(void) | ||
33 | { | ||
34 | ext4_system_zone_cachep = KMEM_CACHE(ext4_system_zone, | ||
35 | SLAB_RECLAIM_ACCOUNT); | ||
36 | if (ext4_system_zone_cachep == NULL) | ||
37 | return -ENOMEM; | ||
38 | return 0; | ||
39 | } | ||
40 | |||
41 | void exit_ext4_system_zone(void) | ||
42 | { | ||
43 | kmem_cache_destroy(ext4_system_zone_cachep); | ||
44 | } | ||
45 | |||
46 | static inline int can_merge(struct ext4_system_zone *entry1, | ||
47 | struct ext4_system_zone *entry2) | ||
48 | { | ||
49 | if ((entry1->start_blk + entry1->count) == entry2->start_blk) | ||
50 | return 1; | ||
51 | return 0; | ||
52 | } | ||
53 | |||
54 | /* | ||
55 | * Mark a range of blocks as belonging to the "system zone" --- that | ||
56 | * is, filesystem metadata blocks which should never be used by | ||
57 | * inodes. | ||
58 | */ | ||
59 | static int add_system_zone(struct ext4_sb_info *sbi, | ||
60 | ext4_fsblk_t start_blk, | ||
61 | unsigned int count) | ||
62 | { | ||
63 | struct ext4_system_zone *new_entry = NULL, *entry; | ||
64 | struct rb_node **n = &sbi->system_blks.rb_node, *node; | ||
65 | struct rb_node *parent = NULL, *new_node = NULL; | ||
66 | |||
67 | while (*n) { | ||
68 | parent = *n; | ||
69 | entry = rb_entry(parent, struct ext4_system_zone, node); | ||
70 | if (start_blk < entry->start_blk) | ||
71 | n = &(*n)->rb_left; | ||
72 | else if (start_blk >= (entry->start_blk + entry->count)) | ||
73 | n = &(*n)->rb_right; | ||
74 | else { | ||
75 | if (start_blk + count > (entry->start_blk + | ||
76 | entry->count)) | ||
77 | entry->count = (start_blk + count - | ||
78 | entry->start_blk); | ||
79 | new_node = *n; | ||
80 | new_entry = rb_entry(new_node, struct ext4_system_zone, | ||
81 | node); | ||
82 | break; | ||
83 | } | ||
84 | } | ||
85 | |||
86 | if (!new_entry) { | ||
87 | new_entry = kmem_cache_alloc(ext4_system_zone_cachep, | ||
88 | GFP_KERNEL); | ||
89 | if (!new_entry) | ||
90 | return -ENOMEM; | ||
91 | new_entry->start_blk = start_blk; | ||
92 | new_entry->count = count; | ||
93 | new_node = &new_entry->node; | ||
94 | |||
95 | rb_link_node(new_node, parent, n); | ||
96 | rb_insert_color(new_node, &sbi->system_blks); | ||
97 | } | ||
98 | |||
99 | /* Can we merge to the left? */ | ||
100 | node = rb_prev(new_node); | ||
101 | if (node) { | ||
102 | entry = rb_entry(node, struct ext4_system_zone, node); | ||
103 | if (can_merge(entry, new_entry)) { | ||
104 | new_entry->start_blk = entry->start_blk; | ||
105 | new_entry->count += entry->count; | ||
106 | rb_erase(node, &sbi->system_blks); | ||
107 | kmem_cache_free(ext4_system_zone_cachep, entry); | ||
108 | } | ||
109 | } | ||
110 | |||
111 | /* Can we merge to the right? */ | ||
112 | node = rb_next(new_node); | ||
113 | if (node) { | ||
114 | entry = rb_entry(node, struct ext4_system_zone, node); | ||
115 | if (can_merge(new_entry, entry)) { | ||
116 | new_entry->count += entry->count; | ||
117 | rb_erase(node, &sbi->system_blks); | ||
118 | kmem_cache_free(ext4_system_zone_cachep, entry); | ||
119 | } | ||
120 | } | ||
121 | return 0; | ||
122 | } | ||
123 | |||
124 | static void debug_print_tree(struct ext4_sb_info *sbi) | ||
125 | { | ||
126 | struct rb_node *node; | ||
127 | struct ext4_system_zone *entry; | ||
128 | int first = 1; | ||
129 | |||
130 | printk(KERN_INFO "System zones: "); | ||
131 | node = rb_first(&sbi->system_blks); | ||
132 | while (node) { | ||
133 | entry = rb_entry(node, struct ext4_system_zone, node); | ||
134 | printk("%s%llu-%llu", first ? "" : ", ", | ||
135 | entry->start_blk, entry->start_blk + entry->count - 1); | ||
136 | first = 0; | ||
137 | node = rb_next(node); | ||
138 | } | ||
139 | printk("\n"); | ||
140 | } | ||
141 | |||
142 | int ext4_setup_system_zone(struct super_block *sb) | ||
143 | { | ||
144 | ext4_group_t ngroups = ext4_get_groups_count(sb); | ||
145 | struct ext4_sb_info *sbi = EXT4_SB(sb); | ||
146 | struct ext4_group_desc *gdp; | ||
147 | ext4_group_t i; | ||
148 | int flex_size = ext4_flex_bg_size(sbi); | ||
149 | int ret; | ||
150 | |||
151 | if (!test_opt(sb, BLOCK_VALIDITY)) { | ||
152 | if (EXT4_SB(sb)->system_blks.rb_node) | ||
153 | ext4_release_system_zone(sb); | ||
154 | return 0; | ||
155 | } | ||
156 | if (EXT4_SB(sb)->system_blks.rb_node) | ||
157 | return 0; | ||
158 | |||
159 | for (i=0; i < ngroups; i++) { | ||
160 | if (ext4_bg_has_super(sb, i) && | ||
161 | ((i < 5) || ((i % flex_size) == 0))) | ||
162 | add_system_zone(sbi, ext4_group_first_block_no(sb, i), | ||
163 | sbi->s_gdb_count + 1); | ||
164 | gdp = ext4_get_group_desc(sb, i, NULL); | ||
165 | ret = add_system_zone(sbi, ext4_block_bitmap(sb, gdp), 1); | ||
166 | if (ret) | ||
167 | return ret; | ||
168 | ret = add_system_zone(sbi, ext4_inode_bitmap(sb, gdp), 1); | ||
169 | if (ret) | ||
170 | return ret; | ||
171 | ret = add_system_zone(sbi, ext4_inode_table(sb, gdp), | ||
172 | sbi->s_itb_per_group); | ||
173 | if (ret) | ||
174 | return ret; | ||
175 | } | ||
176 | |||
177 | if (test_opt(sb, DEBUG)) | ||
178 | debug_print_tree(EXT4_SB(sb)); | ||
179 | return 0; | ||
180 | } | ||
181 | |||
182 | /* Called when the filesystem is unmounted */ | ||
183 | void ext4_release_system_zone(struct super_block *sb) | ||
184 | { | ||
185 | struct rb_node *n = EXT4_SB(sb)->system_blks.rb_node; | ||
186 | struct rb_node *parent; | ||
187 | struct ext4_system_zone *entry; | ||
188 | |||
189 | while (n) { | ||
190 | /* Do the node's children first */ | ||
191 | if (n->rb_left) { | ||
192 | n = n->rb_left; | ||
193 | continue; | ||
194 | } | ||
195 | if (n->rb_right) { | ||
196 | n = n->rb_right; | ||
197 | continue; | ||
198 | } | ||
199 | /* | ||
200 | * The node has no children; free it, and then zero | ||
201 | * out parent's link to it. Finally go to the | ||
202 | * beginning of the loop and try to free the parent | ||
203 | * node. | ||
204 | */ | ||
205 | parent = rb_parent(n); | ||
206 | entry = rb_entry(n, struct ext4_system_zone, node); | ||
207 | kmem_cache_free(ext4_system_zone_cachep, entry); | ||
208 | if (!parent) | ||
209 | EXT4_SB(sb)->system_blks.rb_node = NULL; | ||
210 | else if (parent->rb_left == n) | ||
211 | parent->rb_left = NULL; | ||
212 | else if (parent->rb_right == n) | ||
213 | parent->rb_right = NULL; | ||
214 | n = parent; | ||
215 | } | ||
216 | EXT4_SB(sb)->system_blks.rb_node = NULL; | ||
217 | } | ||
218 | |||
219 | /* | ||
220 | * Returns 1 if the passed-in block region (start_blk, | ||
221 | * start_blk+count) is valid; 0 if some part of the block region | ||
222 | * overlaps with filesystem metadata blocks. | ||
223 | */ | ||
224 | int ext4_data_block_valid(struct ext4_sb_info *sbi, ext4_fsblk_t start_blk, | ||
225 | unsigned int count) | ||
226 | { | ||
227 | struct ext4_system_zone *entry; | ||
228 | struct rb_node *n = sbi->system_blks.rb_node; | ||
229 | |||
230 | if ((start_blk <= le32_to_cpu(sbi->s_es->s_first_data_block)) || | ||
231 | (start_blk + count > ext4_blocks_count(sbi->s_es))) | ||
232 | return 0; | ||
233 | while (n) { | ||
234 | entry = rb_entry(n, struct ext4_system_zone, node); | ||
235 | if (start_blk + count - 1 < entry->start_blk) | ||
236 | n = n->rb_left; | ||
237 | else if (start_blk >= (entry->start_blk + entry->count)) | ||
238 | n = n->rb_right; | ||
239 | else | ||
240 | return 0; | ||
241 | } | ||
242 | return 1; | ||
243 | } | ||
244 | |||