2 * Copyright (C) 2008 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
18 * Based on jffs2 zlib code:
19 * Copyright © 2001-2007 Red Hat, Inc.
20 * Created by David Woodhouse <dwmw2@infradead.org>
23 #include <linux/kernel.h>
24 #include <linux/slab.h>
25 #include <linux/zlib.h>
26 #include <linux/zutil.h>
27 #include <linux/vmalloc.h>
28 #include <linux/init.h>
29 #include <linux/err.h>
30 #include <linux/sched.h>
31 #include <linux/pagemap.h>
32 #include <linux/bio.h>
33 #include "compression.h"
38 struct list_head list
;
41 static void zlib_free_workspace(struct list_head
*ws
)
43 struct workspace
*workspace
= list_entry(ws
, struct workspace
, list
);
45 vfree(workspace
->strm
.workspace
);
46 kfree(workspace
->buf
);
50 static struct list_head
*zlib_alloc_workspace(void)
52 struct workspace
*workspace
;
55 workspace
= kzalloc(sizeof(*workspace
), GFP_NOFS
);
57 return ERR_PTR(-ENOMEM
);
59 workspacesize
= max(zlib_deflate_workspacesize(MAX_WBITS
, MAX_MEM_LEVEL
),
60 zlib_inflate_workspacesize());
61 workspace
->strm
.workspace
= vmalloc(workspacesize
);
62 workspace
->buf
= kmalloc(PAGE_CACHE_SIZE
, GFP_NOFS
);
63 if (!workspace
->strm
.workspace
|| !workspace
->buf
)
66 INIT_LIST_HEAD(&workspace
->list
);
68 return &workspace
->list
;
70 zlib_free_workspace(&workspace
->list
);
71 return ERR_PTR(-ENOMEM
);
74 static int zlib_compress_pages(struct list_head
*ws
,
75 struct address_space
*mapping
,
76 u64 start
, unsigned long len
,
78 unsigned long nr_dest_pages
,
79 unsigned long *out_pages
,
80 unsigned long *total_in
,
81 unsigned long *total_out
,
82 unsigned long max_out
)
84 struct workspace
*workspace
= list_entry(ws
, struct workspace
, list
);
89 struct page
*in_page
= NULL
;
90 struct page
*out_page
= NULL
;
91 unsigned long bytes_left
;
97 if (Z_OK
!= zlib_deflateInit(&workspace
->strm
, 3)) {
98 printk(KERN_WARNING
"BTRFS: deflateInit failed\n");
103 workspace
->strm
.total_in
= 0;
104 workspace
->strm
.total_out
= 0;
106 in_page
= find_get_page(mapping
, start
>> PAGE_CACHE_SHIFT
);
107 data_in
= kmap(in_page
);
109 out_page
= alloc_page(GFP_NOFS
| __GFP_HIGHMEM
);
110 if (out_page
== NULL
) {
114 cpage_out
= kmap(out_page
);
118 workspace
->strm
.next_in
= data_in
;
119 workspace
->strm
.next_out
= cpage_out
;
120 workspace
->strm
.avail_out
= PAGE_CACHE_SIZE
;
121 workspace
->strm
.avail_in
= min(len
, PAGE_CACHE_SIZE
);
123 while (workspace
->strm
.total_in
< len
) {
124 ret
= zlib_deflate(&workspace
->strm
, Z_SYNC_FLUSH
);
126 printk(KERN_DEBUG
"BTRFS: deflate in loop returned %d\n",
128 zlib_deflateEnd(&workspace
->strm
);
133 /* we're making it bigger, give up */
134 if (workspace
->strm
.total_in
> 8192 &&
135 workspace
->strm
.total_in
<
136 workspace
->strm
.total_out
) {
140 /* we need another page for writing out. Test this
141 * before the total_in so we will pull in a new page for
142 * the stream end if required
144 if (workspace
->strm
.avail_out
== 0) {
146 if (nr_pages
== nr_dest_pages
) {
151 out_page
= alloc_page(GFP_NOFS
| __GFP_HIGHMEM
);
152 if (out_page
== NULL
) {
156 cpage_out
= kmap(out_page
);
157 pages
[nr_pages
] = out_page
;
159 workspace
->strm
.avail_out
= PAGE_CACHE_SIZE
;
160 workspace
->strm
.next_out
= cpage_out
;
163 if (workspace
->strm
.total_in
>= len
)
166 /* we've read in a full page, get a new one */
167 if (workspace
->strm
.avail_in
== 0) {
168 if (workspace
->strm
.total_out
> max_out
)
171 bytes_left
= len
- workspace
->strm
.total_in
;
173 page_cache_release(in_page
);
175 start
+= PAGE_CACHE_SIZE
;
176 in_page
= find_get_page(mapping
,
177 start
>> PAGE_CACHE_SHIFT
);
178 data_in
= kmap(in_page
);
179 workspace
->strm
.avail_in
= min(bytes_left
,
181 workspace
->strm
.next_in
= data_in
;
184 workspace
->strm
.avail_in
= 0;
185 ret
= zlib_deflate(&workspace
->strm
, Z_FINISH
);
186 zlib_deflateEnd(&workspace
->strm
);
188 if (ret
!= Z_STREAM_END
) {
193 if (workspace
->strm
.total_out
>= workspace
->strm
.total_in
) {
199 *total_out
= workspace
->strm
.total_out
;
200 *total_in
= workspace
->strm
.total_in
;
202 *out_pages
= nr_pages
;
208 page_cache_release(in_page
);
213 static int zlib_decompress_biovec(struct list_head
*ws
, struct page
**pages_in
,
215 struct bio_vec
*bvec
,
219 struct workspace
*workspace
= list_entry(ws
, struct workspace
, list
);
221 int wbits
= MAX_WBITS
;
223 size_t total_out
= 0;
224 unsigned long page_in_index
= 0;
225 unsigned long page_out_index
= 0;
226 unsigned long total_pages_in
= DIV_ROUND_UP(srclen
, PAGE_CACHE_SIZE
);
227 unsigned long buf_start
;
228 unsigned long pg_offset
;
230 data_in
= kmap(pages_in
[page_in_index
]);
231 workspace
->strm
.next_in
= data_in
;
232 workspace
->strm
.avail_in
= min_t(size_t, srclen
, PAGE_CACHE_SIZE
);
233 workspace
->strm
.total_in
= 0;
235 workspace
->strm
.total_out
= 0;
236 workspace
->strm
.next_out
= workspace
->buf
;
237 workspace
->strm
.avail_out
= PAGE_CACHE_SIZE
;
240 /* If it's deflate, and it's got no preset dictionary, then
241 we can tell zlib to skip the adler32 check. */
242 if (srclen
> 2 && !(data_in
[1] & PRESET_DICT
) &&
243 ((data_in
[0] & 0x0f) == Z_DEFLATED
) &&
244 !(((data_in
[0]<<8) + data_in
[1]) % 31)) {
246 wbits
= -((data_in
[0] >> 4) + 8);
247 workspace
->strm
.next_in
+= 2;
248 workspace
->strm
.avail_in
-= 2;
251 if (Z_OK
!= zlib_inflateInit2(&workspace
->strm
, wbits
)) {
252 printk(KERN_WARNING
"BTRFS: inflateInit failed\n");
255 while (workspace
->strm
.total_in
< srclen
) {
256 ret
= zlib_inflate(&workspace
->strm
, Z_NO_FLUSH
);
257 if (ret
!= Z_OK
&& ret
!= Z_STREAM_END
)
260 buf_start
= total_out
;
261 total_out
= workspace
->strm
.total_out
;
263 /* we didn't make progress in this inflate call, we're done */
264 if (buf_start
== total_out
)
267 ret2
= btrfs_decompress_buf2page(workspace
->buf
, buf_start
,
268 total_out
, disk_start
,
270 &page_out_index
, &pg_offset
);
276 workspace
->strm
.next_out
= workspace
->buf
;
277 workspace
->strm
.avail_out
= PAGE_CACHE_SIZE
;
279 if (workspace
->strm
.avail_in
== 0) {
281 kunmap(pages_in
[page_in_index
]);
283 if (page_in_index
>= total_pages_in
) {
287 data_in
= kmap(pages_in
[page_in_index
]);
288 workspace
->strm
.next_in
= data_in
;
289 tmp
= srclen
- workspace
->strm
.total_in
;
290 workspace
->strm
.avail_in
= min(tmp
,
294 if (ret
!= Z_STREAM_END
)
299 zlib_inflateEnd(&workspace
->strm
);
301 kunmap(pages_in
[page_in_index
]);
303 btrfs_clear_biovec_end(bvec
, vcnt
, page_out_index
, pg_offset
);
307 static int zlib_decompress(struct list_head
*ws
, unsigned char *data_in
,
308 struct page
*dest_page
,
309 unsigned long start_byte
,
310 size_t srclen
, size_t destlen
)
312 struct workspace
*workspace
= list_entry(ws
, struct workspace
, list
);
314 int wbits
= MAX_WBITS
;
315 unsigned long bytes_left
;
316 unsigned long total_out
= 0;
317 unsigned long pg_offset
= 0;
320 destlen
= min_t(unsigned long, destlen
, PAGE_SIZE
);
321 bytes_left
= destlen
;
323 workspace
->strm
.next_in
= data_in
;
324 workspace
->strm
.avail_in
= srclen
;
325 workspace
->strm
.total_in
= 0;
327 workspace
->strm
.next_out
= workspace
->buf
;
328 workspace
->strm
.avail_out
= PAGE_CACHE_SIZE
;
329 workspace
->strm
.total_out
= 0;
330 /* If it's deflate, and it's got no preset dictionary, then
331 we can tell zlib to skip the adler32 check. */
332 if (srclen
> 2 && !(data_in
[1] & PRESET_DICT
) &&
333 ((data_in
[0] & 0x0f) == Z_DEFLATED
) &&
334 !(((data_in
[0]<<8) + data_in
[1]) % 31)) {
336 wbits
= -((data_in
[0] >> 4) + 8);
337 workspace
->strm
.next_in
+= 2;
338 workspace
->strm
.avail_in
-= 2;
341 if (Z_OK
!= zlib_inflateInit2(&workspace
->strm
, wbits
)) {
342 printk(KERN_WARNING
"BTRFS: inflateInit failed\n");
346 while (bytes_left
> 0) {
347 unsigned long buf_start
;
348 unsigned long buf_offset
;
351 ret
= zlib_inflate(&workspace
->strm
, Z_NO_FLUSH
);
352 if (ret
!= Z_OK
&& ret
!= Z_STREAM_END
)
355 buf_start
= total_out
;
356 total_out
= workspace
->strm
.total_out
;
358 if (total_out
== buf_start
) {
363 if (total_out
<= start_byte
)
366 if (total_out
> start_byte
&& buf_start
< start_byte
)
367 buf_offset
= start_byte
- buf_start
;
371 bytes
= min(PAGE_CACHE_SIZE
- pg_offset
,
372 PAGE_CACHE_SIZE
- buf_offset
);
373 bytes
= min(bytes
, bytes_left
);
375 kaddr
= kmap_atomic(dest_page
);
376 memcpy(kaddr
+ pg_offset
, workspace
->buf
+ buf_offset
, bytes
);
377 kunmap_atomic(kaddr
);
382 workspace
->strm
.next_out
= workspace
->buf
;
383 workspace
->strm
.avail_out
= PAGE_CACHE_SIZE
;
386 if (ret
!= Z_STREAM_END
&& bytes_left
!= 0)
391 zlib_inflateEnd(&workspace
->strm
);
394 * this should only happen if zlib returned fewer bytes than we
395 * expected. btrfs_get_block is responsible for zeroing from the
396 * end of the inline extent (destlen) to the end of the page
398 if (pg_offset
< destlen
) {
399 kaddr
= kmap_atomic(dest_page
);
400 memset(kaddr
+ pg_offset
, 0, destlen
- pg_offset
);
401 kunmap_atomic(kaddr
);
406 const struct btrfs_compress_op btrfs_zlib_compress
= {
407 .alloc_workspace
= zlib_alloc_workspace
,
408 .free_workspace
= zlib_free_workspace
,
409 .compress_pages
= zlib_compress_pages
,
410 .decompress_biovec
= zlib_decompress_biovec
,
411 .decompress
= zlib_decompress
,