Root/fs/btrfs/zlib.c

1/*
2 * Copyright (C) 2008 Oracle. All rights reserved.
3 *
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.
7 *
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.
12 *
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.
17 *
18 * Based on jffs2 zlib code:
19 * Copyright © 2001-2007 Red Hat, Inc.
20 * Created by David Woodhouse <dwmw2@infradead.org>
21 */
22
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"
34
35/* Plan: call deflate() with avail_in == *sourcelen,
36    avail_out = *dstlen - 12 and flush == Z_FINISH.
37    If it doesn't manage to finish, call it again with
38    avail_in == 0 and avail_out set to the remaining 12
39    bytes for it to clean up.
40   Q: Is 12 bytes sufficient?
41*/
42#define STREAM_END_SPACE 12
43
44struct workspace {
45    z_stream inf_strm;
46    z_stream def_strm;
47    char *buf;
48    struct list_head list;
49};
50
51static LIST_HEAD(idle_workspace);
52static DEFINE_SPINLOCK(workspace_lock);
53static unsigned long num_workspace;
54static atomic_t alloc_workspace = ATOMIC_INIT(0);
55static DECLARE_WAIT_QUEUE_HEAD(workspace_wait);
56
57/*
58 * this finds an available zlib workspace or allocates a new one
59 * NULL or an ERR_PTR is returned if things go bad.
60 */
61static struct workspace *find_zlib_workspace(void)
62{
63    struct workspace *workspace;
64    int ret;
65    int cpus = num_online_cpus();
66
67again:
68    spin_lock(&workspace_lock);
69    if (!list_empty(&idle_workspace)) {
70        workspace = list_entry(idle_workspace.next, struct workspace,
71                       list);
72        list_del(&workspace->list);
73        num_workspace--;
74        spin_unlock(&workspace_lock);
75        return workspace;
76
77    }
78    spin_unlock(&workspace_lock);
79    if (atomic_read(&alloc_workspace) > cpus) {
80        DEFINE_WAIT(wait);
81        prepare_to_wait(&workspace_wait, &wait, TASK_UNINTERRUPTIBLE);
82        if (atomic_read(&alloc_workspace) > cpus)
83            schedule();
84        finish_wait(&workspace_wait, &wait);
85        goto again;
86    }
87    atomic_inc(&alloc_workspace);
88    workspace = kzalloc(sizeof(*workspace), GFP_NOFS);
89    if (!workspace) {
90        ret = -ENOMEM;
91        goto fail;
92    }
93
94    workspace->def_strm.workspace = vmalloc(zlib_deflate_workspacesize());
95    if (!workspace->def_strm.workspace) {
96        ret = -ENOMEM;
97        goto fail;
98    }
99    workspace->inf_strm.workspace = vmalloc(zlib_inflate_workspacesize());
100    if (!workspace->inf_strm.workspace) {
101        ret = -ENOMEM;
102        goto fail_inflate;
103    }
104    workspace->buf = kmalloc(PAGE_CACHE_SIZE, GFP_NOFS);
105    if (!workspace->buf) {
106        ret = -ENOMEM;
107        goto fail_kmalloc;
108    }
109    return workspace;
110
111fail_kmalloc:
112    vfree(workspace->inf_strm.workspace);
113fail_inflate:
114    vfree(workspace->def_strm.workspace);
115fail:
116    kfree(workspace);
117    atomic_dec(&alloc_workspace);
118    wake_up(&workspace_wait);
119    return ERR_PTR(ret);
120}
121
122/*
123 * put a workspace struct back on the list or free it if we have enough
124 * idle ones sitting around
125 */
126static int free_workspace(struct workspace *workspace)
127{
128    spin_lock(&workspace_lock);
129    if (num_workspace < num_online_cpus()) {
130        list_add_tail(&workspace->list, &idle_workspace);
131        num_workspace++;
132        spin_unlock(&workspace_lock);
133        if (waitqueue_active(&workspace_wait))
134            wake_up(&workspace_wait);
135        return 0;
136    }
137    spin_unlock(&workspace_lock);
138    vfree(workspace->def_strm.workspace);
139    vfree(workspace->inf_strm.workspace);
140    kfree(workspace->buf);
141    kfree(workspace);
142
143    atomic_dec(&alloc_workspace);
144    if (waitqueue_active(&workspace_wait))
145        wake_up(&workspace_wait);
146    return 0;
147}
148
149/*
150 * cleanup function for module exit
151 */
152static void free_workspaces(void)
153{
154    struct workspace *workspace;
155    while (!list_empty(&idle_workspace)) {
156        workspace = list_entry(idle_workspace.next, struct workspace,
157                       list);
158        list_del(&workspace->list);
159        vfree(workspace->def_strm.workspace);
160        vfree(workspace->inf_strm.workspace);
161        kfree(workspace->buf);
162        kfree(workspace);
163        atomic_dec(&alloc_workspace);
164    }
165}
166
167/*
168 * given an address space and start/len, compress the bytes.
169 *
170 * pages are allocated to hold the compressed result and stored
171 * in 'pages'
172 *
173 * out_pages is used to return the number of pages allocated. There
174 * may be pages allocated even if we return an error
175 *
176 * total_in is used to return the number of bytes actually read. It
177 * may be smaller then len if we had to exit early because we
178 * ran out of room in the pages array or because we cross the
179 * max_out threshold.
180 *
181 * total_out is used to return the total number of compressed bytes
182 *
183 * max_out tells us the max number of bytes that we're allowed to
184 * stuff into pages
185 */
186int btrfs_zlib_compress_pages(struct address_space *mapping,
187                  u64 start, unsigned long len,
188                  struct page **pages,
189                  unsigned long nr_dest_pages,
190                  unsigned long *out_pages,
191                  unsigned long *total_in,
192                  unsigned long *total_out,
193                  unsigned long max_out)
194{
195    int ret;
196    struct workspace *workspace;
197    char *data_in;
198    char *cpage_out;
199    int nr_pages = 0;
200    struct page *in_page = NULL;
201    struct page *out_page = NULL;
202    int out_written = 0;
203    int in_read = 0;
204    unsigned long bytes_left;
205
206    *out_pages = 0;
207    *total_out = 0;
208    *total_in = 0;
209
210    workspace = find_zlib_workspace();
211    if (IS_ERR(workspace))
212        return -1;
213
214    if (Z_OK != zlib_deflateInit(&workspace->def_strm, 3)) {
215        printk(KERN_WARNING "deflateInit failed\n");
216        ret = -1;
217        goto out;
218    }
219
220    workspace->def_strm.total_in = 0;
221    workspace->def_strm.total_out = 0;
222
223    in_page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT);
224    data_in = kmap(in_page);
225
226    out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
227    cpage_out = kmap(out_page);
228    pages[0] = out_page;
229    nr_pages = 1;
230
231    workspace->def_strm.next_in = data_in;
232    workspace->def_strm.next_out = cpage_out;
233    workspace->def_strm.avail_out = PAGE_CACHE_SIZE;
234    workspace->def_strm.avail_in = min(len, PAGE_CACHE_SIZE);
235
236    out_written = 0;
237    in_read = 0;
238
239    while (workspace->def_strm.total_in < len) {
240        ret = zlib_deflate(&workspace->def_strm, Z_SYNC_FLUSH);
241        if (ret != Z_OK) {
242            printk(KERN_DEBUG "btrfs deflate in loop returned %d\n",
243                   ret);
244            zlib_deflateEnd(&workspace->def_strm);
245            ret = -1;
246            goto out;
247        }
248
249        /* we're making it bigger, give up */
250        if (workspace->def_strm.total_in > 8192 &&
251            workspace->def_strm.total_in <
252            workspace->def_strm.total_out) {
253            ret = -1;
254            goto out;
255        }
256        /* we need another page for writing out. Test this
257         * before the total_in so we will pull in a new page for
258         * the stream end if required
259         */
260        if (workspace->def_strm.avail_out == 0) {
261            kunmap(out_page);
262            if (nr_pages == nr_dest_pages) {
263                out_page = NULL;
264                ret = -1;
265                goto out;
266            }
267            out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
268            cpage_out = kmap(out_page);
269            pages[nr_pages] = out_page;
270            nr_pages++;
271            workspace->def_strm.avail_out = PAGE_CACHE_SIZE;
272            workspace->def_strm.next_out = cpage_out;
273        }
274        /* we're all done */
275        if (workspace->def_strm.total_in >= len)
276            break;
277
278        /* we've read in a full page, get a new one */
279        if (workspace->def_strm.avail_in == 0) {
280            if (workspace->def_strm.total_out > max_out)
281                break;
282
283            bytes_left = len - workspace->def_strm.total_in;
284            kunmap(in_page);
285            page_cache_release(in_page);
286
287            start += PAGE_CACHE_SIZE;
288            in_page = find_get_page(mapping,
289                        start >> PAGE_CACHE_SHIFT);
290            data_in = kmap(in_page);
291            workspace->def_strm.avail_in = min(bytes_left,
292                               PAGE_CACHE_SIZE);
293            workspace->def_strm.next_in = data_in;
294        }
295    }
296    workspace->def_strm.avail_in = 0;
297    ret = zlib_deflate(&workspace->def_strm, Z_FINISH);
298    zlib_deflateEnd(&workspace->def_strm);
299
300    if (ret != Z_STREAM_END) {
301        ret = -1;
302        goto out;
303    }
304
305    if (workspace->def_strm.total_out >= workspace->def_strm.total_in) {
306        ret = -1;
307        goto out;
308    }
309
310    ret = 0;
311    *total_out = workspace->def_strm.total_out;
312    *total_in = workspace->def_strm.total_in;
313out:
314    *out_pages = nr_pages;
315    if (out_page)
316        kunmap(out_page);
317
318    if (in_page) {
319        kunmap(in_page);
320        page_cache_release(in_page);
321    }
322    free_workspace(workspace);
323    return ret;
324}
325
326/*
327 * pages_in is an array of pages with compressed data.
328 *
329 * disk_start is the starting logical offset of this array in the file
330 *
331 * bvec is a bio_vec of pages from the file that we want to decompress into
332 *
333 * vcnt is the count of pages in the biovec
334 *
335 * srclen is the number of bytes in pages_in
336 *
337 * The basic idea is that we have a bio that was created by readpages.
338 * The pages in the bio are for the uncompressed data, and they may not
339 * be contiguous. They all correspond to the range of bytes covered by
340 * the compressed extent.
341 */
342int btrfs_zlib_decompress_biovec(struct page **pages_in,
343                  u64 disk_start,
344                  struct bio_vec *bvec,
345                  int vcnt,
346                  size_t srclen)
347{
348    int ret = 0;
349    int wbits = MAX_WBITS;
350    struct workspace *workspace;
351    char *data_in;
352    size_t total_out = 0;
353    unsigned long page_bytes_left;
354    unsigned long page_in_index = 0;
355    unsigned long page_out_index = 0;
356    struct page *page_out;
357    unsigned long total_pages_in = (srclen + PAGE_CACHE_SIZE - 1) /
358                    PAGE_CACHE_SIZE;
359    unsigned long buf_start;
360    unsigned long buf_offset;
361    unsigned long bytes;
362    unsigned long working_bytes;
363    unsigned long pg_offset;
364    unsigned long start_byte;
365    unsigned long current_buf_start;
366    char *kaddr;
367
368    workspace = find_zlib_workspace();
369    if (IS_ERR(workspace))
370        return -ENOMEM;
371
372    data_in = kmap(pages_in[page_in_index]);
373    workspace->inf_strm.next_in = data_in;
374    workspace->inf_strm.avail_in = min_t(size_t, srclen, PAGE_CACHE_SIZE);
375    workspace->inf_strm.total_in = 0;
376
377    workspace->inf_strm.total_out = 0;
378    workspace->inf_strm.next_out = workspace->buf;
379    workspace->inf_strm.avail_out = PAGE_CACHE_SIZE;
380    page_out = bvec[page_out_index].bv_page;
381    page_bytes_left = PAGE_CACHE_SIZE;
382    pg_offset = 0;
383
384    /* If it's deflate, and it's got no preset dictionary, then
385       we can tell zlib to skip the adler32 check. */
386    if (srclen > 2 && !(data_in[1] & PRESET_DICT) &&
387        ((data_in[0] & 0x0f) == Z_DEFLATED) &&
388        !(((data_in[0]<<8) + data_in[1]) % 31)) {
389
390        wbits = -((data_in[0] >> 4) + 8);
391        workspace->inf_strm.next_in += 2;
392        workspace->inf_strm.avail_in -= 2;
393    }
394
395    if (Z_OK != zlib_inflateInit2(&workspace->inf_strm, wbits)) {
396        printk(KERN_WARNING "inflateInit failed\n");
397        ret = -1;
398        goto out;
399    }
400    while (workspace->inf_strm.total_in < srclen) {
401        ret = zlib_inflate(&workspace->inf_strm, Z_NO_FLUSH);
402        if (ret != Z_OK && ret != Z_STREAM_END)
403            break;
404        /*
405         * buf start is the byte offset we're of the start of
406         * our workspace buffer
407         */
408        buf_start = total_out;
409
410        /* total_out is the last byte of the workspace buffer */
411        total_out = workspace->inf_strm.total_out;
412
413        working_bytes = total_out - buf_start;
414
415        /*
416         * start byte is the first byte of the page we're currently
417         * copying into relative to the start of the compressed data.
418         */
419        start_byte = page_offset(page_out) - disk_start;
420
421        if (working_bytes == 0) {
422            /* we didn't make progress in this inflate
423             * call, we're done
424             */
425            if (ret != Z_STREAM_END)
426                ret = -1;
427            break;
428        }
429
430        /* we haven't yet hit data corresponding to this page */
431        if (total_out <= start_byte)
432            goto next;
433
434        /*
435         * the start of the data we care about is offset into
436         * the middle of our working buffer
437         */
438        if (total_out > start_byte && buf_start < start_byte) {
439            buf_offset = start_byte - buf_start;
440            working_bytes -= buf_offset;
441        } else {
442            buf_offset = 0;
443        }
444        current_buf_start = buf_start;
445
446        /* copy bytes from the working buffer into the pages */
447        while (working_bytes > 0) {
448            bytes = min(PAGE_CACHE_SIZE - pg_offset,
449                    PAGE_CACHE_SIZE - buf_offset);
450            bytes = min(bytes, working_bytes);
451            kaddr = kmap_atomic(page_out, KM_USER0);
452            memcpy(kaddr + pg_offset, workspace->buf + buf_offset,
453                   bytes);
454            kunmap_atomic(kaddr, KM_USER0);
455            flush_dcache_page(page_out);
456
457            pg_offset += bytes;
458            page_bytes_left -= bytes;
459            buf_offset += bytes;
460            working_bytes -= bytes;
461            current_buf_start += bytes;
462
463            /* check if we need to pick another page */
464            if (page_bytes_left == 0) {
465                page_out_index++;
466                if (page_out_index >= vcnt) {
467                    ret = 0;
468                    goto done;
469                }
470
471                page_out = bvec[page_out_index].bv_page;
472                pg_offset = 0;
473                page_bytes_left = PAGE_CACHE_SIZE;
474                start_byte = page_offset(page_out) - disk_start;
475
476                /*
477                 * make sure our new page is covered by this
478                 * working buffer
479                 */
480                if (total_out <= start_byte)
481                    goto next;
482
483                /* the next page in the biovec might not
484                 * be adjacent to the last page, but it
485                 * might still be found inside this working
486                 * buffer. bump our offset pointer
487                 */
488                if (total_out > start_byte &&
489                    current_buf_start < start_byte) {
490                    buf_offset = start_byte - buf_start;
491                    working_bytes = total_out - start_byte;
492                    current_buf_start = buf_start +
493                        buf_offset;
494                }
495            }
496        }
497next:
498        workspace->inf_strm.next_out = workspace->buf;
499        workspace->inf_strm.avail_out = PAGE_CACHE_SIZE;
500
501        if (workspace->inf_strm.avail_in == 0) {
502            unsigned long tmp;
503            kunmap(pages_in[page_in_index]);
504            page_in_index++;
505            if (page_in_index >= total_pages_in) {
506                data_in = NULL;
507                break;
508            }
509            data_in = kmap(pages_in[page_in_index]);
510            workspace->inf_strm.next_in = data_in;
511            tmp = srclen - workspace->inf_strm.total_in;
512            workspace->inf_strm.avail_in = min(tmp,
513                               PAGE_CACHE_SIZE);
514        }
515    }
516    if (ret != Z_STREAM_END)
517        ret = -1;
518    else
519        ret = 0;
520done:
521    zlib_inflateEnd(&workspace->inf_strm);
522    if (data_in)
523        kunmap(pages_in[page_in_index]);
524out:
525    free_workspace(workspace);
526    return ret;
527}
528
529/*
530 * a less complex decompression routine. Our compressed data fits in a
531 * single page, and we want to read a single page out of it.
532 * start_byte tells us the offset into the compressed data we're interested in
533 */
534int btrfs_zlib_decompress(unsigned char *data_in,
535              struct page *dest_page,
536              unsigned long start_byte,
537              size_t srclen, size_t destlen)
538{
539    int ret = 0;
540    int wbits = MAX_WBITS;
541    struct workspace *workspace;
542    unsigned long bytes_left = destlen;
543    unsigned long total_out = 0;
544    char *kaddr;
545
546    if (destlen > PAGE_CACHE_SIZE)
547        return -ENOMEM;
548
549    workspace = find_zlib_workspace();
550    if (IS_ERR(workspace))
551        return -ENOMEM;
552
553    workspace->inf_strm.next_in = data_in;
554    workspace->inf_strm.avail_in = srclen;
555    workspace->inf_strm.total_in = 0;
556
557    workspace->inf_strm.next_out = workspace->buf;
558    workspace->inf_strm.avail_out = PAGE_CACHE_SIZE;
559    workspace->inf_strm.total_out = 0;
560    /* If it's deflate, and it's got no preset dictionary, then
561       we can tell zlib to skip the adler32 check. */
562    if (srclen > 2 && !(data_in[1] & PRESET_DICT) &&
563        ((data_in[0] & 0x0f) == Z_DEFLATED) &&
564        !(((data_in[0]<<8) + data_in[1]) % 31)) {
565
566        wbits = -((data_in[0] >> 4) + 8);
567        workspace->inf_strm.next_in += 2;
568        workspace->inf_strm.avail_in -= 2;
569    }
570
571    if (Z_OK != zlib_inflateInit2(&workspace->inf_strm, wbits)) {
572        printk(KERN_WARNING "inflateInit failed\n");
573        ret = -1;
574        goto out;
575    }
576
577    while (bytes_left > 0) {
578        unsigned long buf_start;
579        unsigned long buf_offset;
580        unsigned long bytes;
581        unsigned long pg_offset = 0;
582
583        ret = zlib_inflate(&workspace->inf_strm, Z_NO_FLUSH);
584        if (ret != Z_OK && ret != Z_STREAM_END)
585            break;
586
587        buf_start = total_out;
588        total_out = workspace->inf_strm.total_out;
589
590        if (total_out == buf_start) {
591            ret = -1;
592            break;
593        }
594
595        if (total_out <= start_byte)
596            goto next;
597
598        if (total_out > start_byte && buf_start < start_byte)
599            buf_offset = start_byte - buf_start;
600        else
601            buf_offset = 0;
602
603        bytes = min(PAGE_CACHE_SIZE - pg_offset,
604                PAGE_CACHE_SIZE - buf_offset);
605        bytes = min(bytes, bytes_left);
606
607        kaddr = kmap_atomic(dest_page, KM_USER0);
608        memcpy(kaddr + pg_offset, workspace->buf + buf_offset, bytes);
609        kunmap_atomic(kaddr, KM_USER0);
610
611        pg_offset += bytes;
612        bytes_left -= bytes;
613next:
614        workspace->inf_strm.next_out = workspace->buf;
615        workspace->inf_strm.avail_out = PAGE_CACHE_SIZE;
616    }
617
618    if (ret != Z_STREAM_END && bytes_left != 0)
619        ret = -1;
620    else
621        ret = 0;
622
623    zlib_inflateEnd(&workspace->inf_strm);
624out:
625    free_workspace(workspace);
626    return ret;
627}
628
629void btrfs_zlib_exit(void)
630{
631    free_workspaces();
632}
633

Archive Download this file



interactive