blob: df73375c160a6cb039444920be2a5c5a23bd44e5 [file] [log] [blame]
NeilBrown32a76272005-06-21 17:17:14 -07001/*
2 * bitmap.c two-level bitmap (C) Peter T. Breuer (ptb@ot.uc3m.es) 2003
3 *
4 * bitmap_create - sets up the bitmap structure
5 * bitmap_destroy - destroys the bitmap structure
6 *
7 * additions, Copyright (C) 2003-2004, Paul Clements, SteelEye Technology, Inc.:
8 * - added disk storage for bitmap
9 * - changes to allow various bitmap chunk sizes
NeilBrown32a76272005-06-21 17:17:14 -070010 */
11
12/*
13 * Still to do:
14 *
15 * flush after percent set rather than just time based. (maybe both).
NeilBrown32a76272005-06-21 17:17:14 -070016 */
17
NeilBrownbff61972009-03-31 14:33:13 +110018#include <linux/blkdev.h>
NeilBrown32a76272005-06-21 17:17:14 -070019#include <linux/module.h>
NeilBrown32a76272005-06-21 17:17:14 -070020#include <linux/errno.h>
21#include <linux/slab.h>
22#include <linux/init.h>
NeilBrown32a76272005-06-21 17:17:14 -070023#include <linux/timer.h>
24#include <linux/sched.h>
25#include <linux/list.h>
26#include <linux/file.h>
27#include <linux/mount.h>
28#include <linux/buffer_head.h>
NeilBrown57148962012-03-19 12:46:40 +110029#include <linux/seq_file.h>
NeilBrown43b2e5d2009-03-31 14:33:13 +110030#include "md.h"
Christoph Hellwigef740c32009-03-31 14:27:03 +110031#include "bitmap.h"
NeilBrown32a76272005-06-21 17:17:14 -070032
NeilBrownac2f40b2010-06-01 19:37:31 +100033static inline char *bmname(struct bitmap *bitmap)
NeilBrown32a76272005-06-21 17:17:14 -070034{
35 return bitmap->mddev ? mdname(bitmap->mddev) : "mdX";
36}
37
NeilBrown32a76272005-06-21 17:17:14 -070038/*
NeilBrown32a76272005-06-21 17:17:14 -070039 * check a page and, if necessary, allocate it (or hijack it if the alloc fails)
40 *
41 * 1) check to see if this page is allocated, if it's not then try to alloc
42 * 2) if the alloc fails, set the page's hijacked flag so we'll use the
43 * page pointer directly as a counter
44 *
45 * if we find our page, we increment the page's refcount so that it stays
46 * allocated while we're using it
47 */
NeilBrown40cffcc2012-05-22 13:55:24 +100048static int bitmap_checkpage(struct bitmap_counts *bitmap,
NeilBrownac2f40b2010-06-01 19:37:31 +100049 unsigned long page, int create)
NeilBrownee305ac2009-09-23 18:06:44 +100050__releases(bitmap->lock)
51__acquires(bitmap->lock)
NeilBrown32a76272005-06-21 17:17:14 -070052{
53 unsigned char *mappage;
54
55 if (page >= bitmap->pages) {
NeilBrown1187cf02009-03-31 14:27:02 +110056 /* This can happen if bitmap_start_sync goes beyond
57 * End-of-device while looking for a whole page.
58 * It is harmless.
59 */
NeilBrown32a76272005-06-21 17:17:14 -070060 return -EINVAL;
61 }
62
NeilBrown32a76272005-06-21 17:17:14 -070063 if (bitmap->bp[page].hijacked) /* it's hijacked, don't try to alloc */
64 return 0;
65
66 if (bitmap->bp[page].map) /* page is already allocated, just return */
67 return 0;
68
69 if (!create)
70 return -ENOENT;
71
NeilBrown32a76272005-06-21 17:17:14 -070072 /* this page has not been allocated yet */
73
NeilBrownac2f40b2010-06-01 19:37:31 +100074 spin_unlock_irq(&bitmap->lock);
NeilBrown792a1d42012-03-19 12:46:41 +110075 mappage = kzalloc(PAGE_SIZE, GFP_NOIO);
NeilBrownac2f40b2010-06-01 19:37:31 +100076 spin_lock_irq(&bitmap->lock);
77
78 if (mappage == NULL) {
NeilBrown40cffcc2012-05-22 13:55:24 +100079 pr_debug("md/bitmap: map page allocation failed, hijacking\n");
NeilBrown32a76272005-06-21 17:17:14 -070080 /* failed - set the hijacked flag so that we can use the
81 * pointer as a counter */
NeilBrown32a76272005-06-21 17:17:14 -070082 if (!bitmap->bp[page].map)
83 bitmap->bp[page].hijacked = 1;
NeilBrownac2f40b2010-06-01 19:37:31 +100084 } else if (bitmap->bp[page].map ||
85 bitmap->bp[page].hijacked) {
NeilBrown32a76272005-06-21 17:17:14 -070086 /* somebody beat us to getting the page */
NeilBrown792a1d42012-03-19 12:46:41 +110087 kfree(mappage);
NeilBrown32a76272005-06-21 17:17:14 -070088 return 0;
NeilBrownac2f40b2010-06-01 19:37:31 +100089 } else {
90
91 /* no page was in place and we have one, so install it */
92
93 bitmap->bp[page].map = mappage;
94 bitmap->missing_pages--;
NeilBrown32a76272005-06-21 17:17:14 -070095 }
NeilBrown32a76272005-06-21 17:17:14 -070096 return 0;
97}
98
NeilBrown32a76272005-06-21 17:17:14 -070099/* if page is completely empty, put it back on the free list, or dealloc it */
100/* if page was hijacked, unmark the flag so it might get alloced next time */
101/* Note: lock should be held when calling this */
NeilBrown40cffcc2012-05-22 13:55:24 +1000102static void bitmap_checkfree(struct bitmap_counts *bitmap, unsigned long page)
NeilBrown32a76272005-06-21 17:17:14 -0700103{
104 char *ptr;
105
106 if (bitmap->bp[page].count) /* page is still busy */
107 return;
108
109 /* page is no longer in use, it can be released */
110
111 if (bitmap->bp[page].hijacked) { /* page was hijacked, undo this now */
112 bitmap->bp[page].hijacked = 0;
113 bitmap->bp[page].map = NULL;
NeilBrownac2f40b2010-06-01 19:37:31 +1000114 } else {
115 /* normal case, free the page */
116 ptr = bitmap->bp[page].map;
117 bitmap->bp[page].map = NULL;
118 bitmap->missing_pages++;
NeilBrown792a1d42012-03-19 12:46:41 +1100119 kfree(ptr);
NeilBrown32a76272005-06-21 17:17:14 -0700120 }
NeilBrown32a76272005-06-21 17:17:14 -0700121}
122
NeilBrown32a76272005-06-21 17:17:14 -0700123/*
124 * bitmap file handling - read and write the bitmap file and its superblock
125 */
126
NeilBrown32a76272005-06-21 17:17:14 -0700127/*
128 * basic page I/O operations
129 */
130
NeilBrowna654b9d82005-06-21 17:17:27 -0700131/* IO operations when bitmap is stored near all superblocks */
NeilBrown27581e52012-05-22 13:55:08 +1000132static int read_sb_page(struct mddev *mddev, loff_t offset,
133 struct page *page,
134 unsigned long index, int size)
NeilBrowna654b9d82005-06-21 17:17:27 -0700135{
136 /* choose a good rdev and read the page from there */
137
NeilBrown3cb03002011-10-11 16:45:26 +1100138 struct md_rdev *rdev;
NeilBrowna654b9d82005-06-21 17:17:27 -0700139 sector_t target;
NeilBrowna654b9d82005-06-21 17:17:27 -0700140
NeilBrowndafb20f2012-03-19 12:46:39 +1100141 rdev_for_each(rdev, mddev) {
NeilBrownb2d444d2005-11-08 21:39:31 -0800142 if (! test_bit(In_sync, &rdev->flags)
143 || test_bit(Faulty, &rdev->flags))
NeilBrownab904d62005-09-09 16:23:52 -0700144 continue;
145
Jonathan Brassowccebd4c2011-01-14 09:14:33 +1100146 target = offset + index * (PAGE_SIZE/512);
NeilBrowna654b9d82005-06-21 17:17:27 -0700147
NeilBrown2b193362010-10-27 15:16:40 +1100148 if (sync_page_io(rdev, target,
Martin K. Petersene1defc42009-05-22 17:17:49 -0400149 roundup(size, bdev_logical_block_size(rdev->bdev)),
Jonathan Brassowccebd4c2011-01-14 09:14:33 +1100150 page, READ, true)) {
NeilBrownab904d62005-09-09 16:23:52 -0700151 page->index = index;
NeilBrown27581e52012-05-22 13:55:08 +1000152 return 0;
NeilBrownab904d62005-09-09 16:23:52 -0700153 }
154 }
NeilBrown27581e52012-05-22 13:55:08 +1000155 return -EIO;
NeilBrowna654b9d82005-06-21 17:17:27 -0700156}
157
NeilBrownfd01b882011-10-11 16:47:53 +1100158static struct md_rdev *next_active_rdev(struct md_rdev *rdev, struct mddev *mddev)
NeilBrownb2d2c4c2008-09-01 12:48:13 +1000159{
160 /* Iterate the disks of an mddev, using rcu to protect access to the
161 * linked list, and raising the refcount of devices we return to ensure
162 * they don't disappear while in use.
163 * As devices are only added or removed when raid_disk is < 0 and
164 * nr_pending is 0 and In_sync is clear, the entries we return will
165 * still be in the same position on the list when we re-enter
Michael Wangfd177482012-10-11 13:43:21 +1100166 * list_for_each_entry_continue_rcu.
NeilBrownb2d2c4c2008-09-01 12:48:13 +1000167 */
NeilBrownb2d2c4c2008-09-01 12:48:13 +1000168 rcu_read_lock();
169 if (rdev == NULL)
170 /* start at the beginning */
Michael Wangfd177482012-10-11 13:43:21 +1100171 rdev = list_entry_rcu(&mddev->disks, struct md_rdev, same_set);
NeilBrownb2d2c4c2008-09-01 12:48:13 +1000172 else {
173 /* release the previous rdev and start from there. */
174 rdev_dec_pending(rdev, mddev);
NeilBrownb2d2c4c2008-09-01 12:48:13 +1000175 }
Michael Wangfd177482012-10-11 13:43:21 +1100176 list_for_each_entry_continue_rcu(rdev, &mddev->disks, same_set) {
NeilBrownb2d2c4c2008-09-01 12:48:13 +1000177 if (rdev->raid_disk >= 0 &&
NeilBrownb2d2c4c2008-09-01 12:48:13 +1000178 !test_bit(Faulty, &rdev->flags)) {
179 /* this is a usable devices */
180 atomic_inc(&rdev->nr_pending);
181 rcu_read_unlock();
182 return rdev;
183 }
184 }
185 rcu_read_unlock();
186 return NULL;
187}
188
NeilBrownab6085c2007-05-23 13:58:10 -0700189static int write_sb_page(struct bitmap *bitmap, struct page *page, int wait)
NeilBrowna654b9d82005-06-21 17:17:27 -0700190{
NeilBrown3cb03002011-10-11 16:45:26 +1100191 struct md_rdev *rdev = NULL;
Jonathan Brassowa6ff7e02011-01-14 09:14:34 +1100192 struct block_device *bdev;
NeilBrownfd01b882011-10-11 16:47:53 +1100193 struct mddev *mddev = bitmap->mddev;
NeilBrown1ec885c2012-05-22 13:55:10 +1000194 struct bitmap_storage *store = &bitmap->storage;
NeilBrowna654b9d82005-06-21 17:17:27 -0700195
NeilBrownb2d2c4c2008-09-01 12:48:13 +1000196 while ((rdev = next_active_rdev(rdev, mddev)) != NULL) {
NeilBrownac2f40b2010-06-01 19:37:31 +1000197 int size = PAGE_SIZE;
198 loff_t offset = mddev->bitmap_info.offset;
Jonathan Brassowa6ff7e02011-01-14 09:14:34 +1100199
200 bdev = (rdev->meta_bdev) ? rdev->meta_bdev : rdev->bdev;
201
NeilBrown9b1215c2012-05-22 13:55:11 +1000202 if (page->index == store->file_pages-1) {
203 int last_page_size = store->bytes & (PAGE_SIZE-1);
204 if (last_page_size == 0)
205 last_page_size = PAGE_SIZE;
206 size = roundup(last_page_size,
Jonathan Brassowa6ff7e02011-01-14 09:14:34 +1100207 bdev_logical_block_size(bdev));
NeilBrown9b1215c2012-05-22 13:55:11 +1000208 }
NeilBrownac2f40b2010-06-01 19:37:31 +1000209 /* Just make sure we aren't corrupting data or
210 * metadata
211 */
212 if (mddev->external) {
213 /* Bitmap could be anywhere. */
214 if (rdev->sb_start + offset + (page->index
215 * (PAGE_SIZE/512))
216 > rdev->data_offset
217 &&
218 rdev->sb_start + offset
219 < (rdev->data_offset + mddev->dev_sectors
220 + (PAGE_SIZE/512)))
221 goto bad_alignment;
222 } else if (offset < 0) {
223 /* DATA BITMAP METADATA */
224 if (offset
225 + (long)(page->index * (PAGE_SIZE/512))
226 + size/512 > 0)
227 /* bitmap runs in to metadata */
228 goto bad_alignment;
229 if (rdev->data_offset + mddev->dev_sectors
230 > rdev->sb_start + offset)
231 /* data runs in to bitmap */
232 goto bad_alignment;
233 } else if (rdev->sb_start < rdev->data_offset) {
234 /* METADATA BITMAP DATA */
235 if (rdev->sb_start
236 + offset
237 + page->index*(PAGE_SIZE/512) + size/512
238 > rdev->data_offset)
239 /* bitmap runs in to data */
240 goto bad_alignment;
241 } else {
242 /* DATA METADATA BITMAP - no problems */
243 }
244 md_super_write(mddev, rdev,
245 rdev->sb_start + offset
246 + page->index * (PAGE_SIZE/512),
247 size,
248 page);
NeilBrownb2d2c4c2008-09-01 12:48:13 +1000249 }
NeilBrowna654b9d82005-06-21 17:17:27 -0700250
251 if (wait)
NeilBrowna9701a32005-11-08 21:39:34 -0800252 md_super_wait(mddev);
NeilBrowna654b9d82005-06-21 17:17:27 -0700253 return 0;
NeilBrown4b809912008-07-21 17:05:25 +1000254
255 bad_alignment:
NeilBrown4b809912008-07-21 17:05:25 +1000256 return -EINVAL;
NeilBrowna654b9d82005-06-21 17:17:27 -0700257}
258
NeilBrown4ad13662007-07-17 04:06:13 -0700259static void bitmap_file_kick(struct bitmap *bitmap);
NeilBrown32a76272005-06-21 17:17:14 -0700260/*
NeilBrowna654b9d82005-06-21 17:17:27 -0700261 * write out a page to a file
NeilBrown32a76272005-06-21 17:17:14 -0700262 */
NeilBrown4ad13662007-07-17 04:06:13 -0700263static void write_page(struct bitmap *bitmap, struct page *page, int wait)
NeilBrown32a76272005-06-21 17:17:14 -0700264{
NeilBrownd785a062006-06-26 00:27:48 -0700265 struct buffer_head *bh;
NeilBrown32a76272005-06-21 17:17:14 -0700266
NeilBrown1ec885c2012-05-22 13:55:10 +1000267 if (bitmap->storage.file == NULL) {
NeilBrownf0d76d72007-07-17 04:06:12 -0700268 switch (write_sb_page(bitmap, page, wait)) {
269 case -EINVAL:
NeilBrownb405fe92012-05-22 13:55:15 +1000270 set_bit(BITMAP_WRITE_ERROR, &bitmap->flags);
NeilBrownf0d76d72007-07-17 04:06:12 -0700271 }
NeilBrown4ad13662007-07-17 04:06:13 -0700272 } else {
NeilBrowna654b9d82005-06-21 17:17:27 -0700273
NeilBrown4ad13662007-07-17 04:06:13 -0700274 bh = page_buffers(page);
NeilBrownc7084432006-01-06 00:20:45 -0800275
NeilBrown4ad13662007-07-17 04:06:13 -0700276 while (bh && bh->b_blocknr) {
277 atomic_inc(&bitmap->pending_writes);
278 set_buffer_locked(bh);
279 set_buffer_mapped(bh);
Jens Axboe721a9602011-03-09 11:56:30 +0100280 submit_bh(WRITE | REQ_SYNC, bh);
NeilBrown4ad13662007-07-17 04:06:13 -0700281 bh = bh->b_this_page;
282 }
NeilBrown32a76272005-06-21 17:17:14 -0700283
NeilBrownac2f40b2010-06-01 19:37:31 +1000284 if (wait)
NeilBrown4ad13662007-07-17 04:06:13 -0700285 wait_event(bitmap->write_wait,
286 atomic_read(&bitmap->pending_writes)==0);
NeilBrown32a76272005-06-21 17:17:14 -0700287 }
NeilBrownb405fe92012-05-22 13:55:15 +1000288 if (test_bit(BITMAP_WRITE_ERROR, &bitmap->flags))
NeilBrown4ad13662007-07-17 04:06:13 -0700289 bitmap_file_kick(bitmap);
NeilBrown32a76272005-06-21 17:17:14 -0700290}
291
NeilBrownd785a062006-06-26 00:27:48 -0700292static void end_bitmap_write(struct buffer_head *bh, int uptodate)
NeilBrown32a76272005-06-21 17:17:14 -0700293{
NeilBrownd785a062006-06-26 00:27:48 -0700294 struct bitmap *bitmap = bh->b_private;
NeilBrownd785a062006-06-26 00:27:48 -0700295
NeilBrownb405fe92012-05-22 13:55:15 +1000296 if (!uptodate)
297 set_bit(BITMAP_WRITE_ERROR, &bitmap->flags);
NeilBrownd785a062006-06-26 00:27:48 -0700298 if (atomic_dec_and_test(&bitmap->pending_writes))
299 wake_up(&bitmap->write_wait);
300}
301
302/* copied from buffer.c */
303static void
304__clear_page_buffers(struct page *page)
305{
306 ClearPagePrivate(page);
307 set_page_private(page, 0);
308 page_cache_release(page);
309}
310static void free_buffers(struct page *page)
311{
NeilBrown27581e52012-05-22 13:55:08 +1000312 struct buffer_head *bh;
NeilBrownd785a062006-06-26 00:27:48 -0700313
NeilBrown27581e52012-05-22 13:55:08 +1000314 if (!PagePrivate(page))
315 return;
316
317 bh = page_buffers(page);
NeilBrownd785a062006-06-26 00:27:48 -0700318 while (bh) {
319 struct buffer_head *next = bh->b_this_page;
320 free_buffer_head(bh);
321 bh = next;
322 }
323 __clear_page_buffers(page);
324 put_page(page);
325}
326
327/* read a page from a file.
328 * We both read the page, and attach buffers to the page to record the
329 * address of each block (using bmap). These addresses will be used
330 * to write the block later, completely bypassing the filesystem.
331 * This usage is similar to how swap files are handled, and allows us
332 * to write to a file with no concerns of memory allocation failing.
333 */
NeilBrown27581e52012-05-22 13:55:08 +1000334static int read_page(struct file *file, unsigned long index,
335 struct bitmap *bitmap,
336 unsigned long count,
337 struct page *page)
NeilBrownd785a062006-06-26 00:27:48 -0700338{
NeilBrown27581e52012-05-22 13:55:08 +1000339 int ret = 0;
Josef Sipekc649bb92006-12-08 02:37:19 -0800340 struct inode *inode = file->f_path.dentry->d_inode;
NeilBrownd785a062006-06-26 00:27:48 -0700341 struct buffer_head *bh;
342 sector_t block;
NeilBrown32a76272005-06-21 17:17:14 -0700343
NeilBrown36a4e1f2011-10-07 14:23:17 +1100344 pr_debug("read bitmap file (%dB @ %llu)\n", (int)PAGE_SIZE,
345 (unsigned long long)index << PAGE_SHIFT);
NeilBrown32a76272005-06-21 17:17:14 -0700346
NeilBrownd785a062006-06-26 00:27:48 -0700347 bh = alloc_page_buffers(page, 1<<inode->i_blkbits, 0);
348 if (!bh) {
NeilBrown27581e52012-05-22 13:55:08 +1000349 ret = -ENOMEM;
NeilBrownd785a062006-06-26 00:27:48 -0700350 goto out;
351 }
352 attach_page_buffers(page, bh);
353 block = index << (PAGE_SHIFT - inode->i_blkbits);
354 while (bh) {
355 if (count == 0)
356 bh->b_blocknr = 0;
357 else {
358 bh->b_blocknr = bmap(inode, block);
359 if (bh->b_blocknr == 0) {
360 /* Cannot use this file! */
NeilBrown27581e52012-05-22 13:55:08 +1000361 ret = -EINVAL;
NeilBrownd785a062006-06-26 00:27:48 -0700362 goto out;
363 }
364 bh->b_bdev = inode->i_sb->s_bdev;
365 if (count < (1<<inode->i_blkbits))
366 count = 0;
367 else
368 count -= (1<<inode->i_blkbits);
NeilBrown32a76272005-06-21 17:17:14 -0700369
NeilBrownd785a062006-06-26 00:27:48 -0700370 bh->b_end_io = end_bitmap_write;
371 bh->b_private = bitmap;
NeilBrownce25c312006-06-26 00:27:49 -0700372 atomic_inc(&bitmap->pending_writes);
373 set_buffer_locked(bh);
374 set_buffer_mapped(bh);
375 submit_bh(READ, bh);
NeilBrownd785a062006-06-26 00:27:48 -0700376 }
377 block++;
378 bh = bh->b_this_page;
379 }
NeilBrownd785a062006-06-26 00:27:48 -0700380 page->index = index;
NeilBrownce25c312006-06-26 00:27:49 -0700381
382 wait_event(bitmap->write_wait,
383 atomic_read(&bitmap->pending_writes)==0);
NeilBrownb405fe92012-05-22 13:55:15 +1000384 if (test_bit(BITMAP_WRITE_ERROR, &bitmap->flags))
NeilBrown27581e52012-05-22 13:55:08 +1000385 ret = -EIO;
NeilBrown32a76272005-06-21 17:17:14 -0700386out:
NeilBrown27581e52012-05-22 13:55:08 +1000387 if (ret)
388 printk(KERN_ALERT "md: bitmap read error: (%dB @ %llu): %d\n",
NeilBrown2d1f3b52006-01-06 00:20:31 -0800389 (int)PAGE_SIZE,
390 (unsigned long long)index << PAGE_SHIFT,
NeilBrown27581e52012-05-22 13:55:08 +1000391 ret);
392 return ret;
NeilBrown32a76272005-06-21 17:17:14 -0700393}
394
395/*
396 * bitmap file superblock operations
397 */
398
399/* update the event counter and sync the superblock to disk */
NeilBrown4ad13662007-07-17 04:06:13 -0700400void bitmap_update_sb(struct bitmap *bitmap)
NeilBrown32a76272005-06-21 17:17:14 -0700401{
402 bitmap_super_t *sb;
NeilBrown32a76272005-06-21 17:17:14 -0700403
404 if (!bitmap || !bitmap->mddev) /* no bitmap for this array */
NeilBrown4ad13662007-07-17 04:06:13 -0700405 return;
NeilBrownece5cff2009-12-14 12:49:56 +1100406 if (bitmap->mddev->bitmap_info.external)
407 return;
NeilBrown1ec885c2012-05-22 13:55:10 +1000408 if (!bitmap->storage.sb_page) /* no superblock */
NeilBrown4ad13662007-07-17 04:06:13 -0700409 return;
NeilBrown1ec885c2012-05-22 13:55:10 +1000410 sb = kmap_atomic(bitmap->storage.sb_page);
NeilBrown32a76272005-06-21 17:17:14 -0700411 sb->events = cpu_to_le64(bitmap->mddev->events);
NeilBrown8258c532011-05-11 14:26:30 +1000412 if (bitmap->mddev->events < bitmap->events_cleared)
Neil Browna0da84f2008-06-28 08:31:22 +1000413 /* rocking back to read-only */
414 bitmap->events_cleared = bitmap->mddev->events;
NeilBrown8258c532011-05-11 14:26:30 +1000415 sb->events_cleared = cpu_to_le64(bitmap->events_cleared);
416 sb->state = cpu_to_le32(bitmap->flags);
NeilBrown43a70502009-12-14 12:49:55 +1100417 /* Just in case these have been changed via sysfs: */
418 sb->daemon_sleep = cpu_to_le32(bitmap->mddev->bitmap_info.daemon_sleep/HZ);
419 sb->write_behind = cpu_to_le32(bitmap->mddev->bitmap_info.max_write_behind);
NeilBrownb81a0402012-05-22 13:55:26 +1000420 /* This might have been changed by a reshape */
421 sb->sync_size = cpu_to_le64(bitmap->mddev->resync_max_sectors);
422 sb->chunksize = cpu_to_le32(bitmap->mddev->bitmap_info.chunksize);
NeilBrown1dff2b82012-05-22 13:55:34 +1000423 sb->sectors_reserved = cpu_to_le32(bitmap->mddev->
424 bitmap_info.space);
Cong Wangb2f46e62011-11-28 13:25:44 +0800425 kunmap_atomic(sb);
NeilBrown1ec885c2012-05-22 13:55:10 +1000426 write_page(bitmap, bitmap->storage.sb_page, 1);
NeilBrown32a76272005-06-21 17:17:14 -0700427}
428
429/* print out the bitmap file superblock */
430void bitmap_print_sb(struct bitmap *bitmap)
431{
432 bitmap_super_t *sb;
433
NeilBrown1ec885c2012-05-22 13:55:10 +1000434 if (!bitmap || !bitmap->storage.sb_page)
NeilBrown32a76272005-06-21 17:17:14 -0700435 return;
NeilBrown1ec885c2012-05-22 13:55:10 +1000436 sb = kmap_atomic(bitmap->storage.sb_page);
NeilBrown32a76272005-06-21 17:17:14 -0700437 printk(KERN_DEBUG "%s: bitmap file superblock:\n", bmname(bitmap));
NeilBrowna2cff262005-06-21 17:17:20 -0700438 printk(KERN_DEBUG " magic: %08x\n", le32_to_cpu(sb->magic));
439 printk(KERN_DEBUG " version: %d\n", le32_to_cpu(sb->version));
440 printk(KERN_DEBUG " uuid: %08x.%08x.%08x.%08x\n",
NeilBrown32a76272005-06-21 17:17:14 -0700441 *(__u32 *)(sb->uuid+0),
442 *(__u32 *)(sb->uuid+4),
443 *(__u32 *)(sb->uuid+8),
444 *(__u32 *)(sb->uuid+12));
NeilBrowna2cff262005-06-21 17:17:20 -0700445 printk(KERN_DEBUG " events: %llu\n",
NeilBrown32a76272005-06-21 17:17:14 -0700446 (unsigned long long) le64_to_cpu(sb->events));
NeilBrowna2cff262005-06-21 17:17:20 -0700447 printk(KERN_DEBUG "events cleared: %llu\n",
NeilBrown32a76272005-06-21 17:17:14 -0700448 (unsigned long long) le64_to_cpu(sb->events_cleared));
NeilBrowna2cff262005-06-21 17:17:20 -0700449 printk(KERN_DEBUG " state: %08x\n", le32_to_cpu(sb->state));
450 printk(KERN_DEBUG " chunksize: %d B\n", le32_to_cpu(sb->chunksize));
451 printk(KERN_DEBUG " daemon sleep: %ds\n", le32_to_cpu(sb->daemon_sleep));
452 printk(KERN_DEBUG " sync size: %llu KB\n",
453 (unsigned long long)le64_to_cpu(sb->sync_size)/2);
NeilBrown4b6d2872005-09-09 16:23:47 -0700454 printk(KERN_DEBUG "max write behind: %d\n", le32_to_cpu(sb->write_behind));
Cong Wangb2f46e62011-11-28 13:25:44 +0800455 kunmap_atomic(sb);
NeilBrown32a76272005-06-21 17:17:14 -0700456}
457
Jonathan Brassow9c810752011-06-08 17:59:30 -0500458/*
459 * bitmap_new_disk_sb
460 * @bitmap
461 *
462 * This function is somewhat the reverse of bitmap_read_sb. bitmap_read_sb
463 * reads and verifies the on-disk bitmap superblock and populates bitmap_info.
464 * This function verifies 'bitmap_info' and populates the on-disk bitmap
465 * structure, which is to be written to disk.
466 *
467 * Returns: 0 on success, -Exxx on error
468 */
469static int bitmap_new_disk_sb(struct bitmap *bitmap)
470{
471 bitmap_super_t *sb;
472 unsigned long chunksize, daemon_sleep, write_behind;
473 int err = -EINVAL;
474
NeilBrown1ec885c2012-05-22 13:55:10 +1000475 bitmap->storage.sb_page = alloc_page(GFP_KERNEL);
476 if (IS_ERR(bitmap->storage.sb_page)) {
477 err = PTR_ERR(bitmap->storage.sb_page);
478 bitmap->storage.sb_page = NULL;
Jonathan Brassow9c810752011-06-08 17:59:30 -0500479 return err;
480 }
NeilBrown1ec885c2012-05-22 13:55:10 +1000481 bitmap->storage.sb_page->index = 0;
Jonathan Brassow9c810752011-06-08 17:59:30 -0500482
NeilBrown1ec885c2012-05-22 13:55:10 +1000483 sb = kmap_atomic(bitmap->storage.sb_page);
Jonathan Brassow9c810752011-06-08 17:59:30 -0500484
485 sb->magic = cpu_to_le32(BITMAP_MAGIC);
486 sb->version = cpu_to_le32(BITMAP_MAJOR_HI);
487
488 chunksize = bitmap->mddev->bitmap_info.chunksize;
489 BUG_ON(!chunksize);
490 if (!is_power_of_2(chunksize)) {
Cong Wangb2f46e62011-11-28 13:25:44 +0800491 kunmap_atomic(sb);
Jonathan Brassow9c810752011-06-08 17:59:30 -0500492 printk(KERN_ERR "bitmap chunksize not a power of 2\n");
493 return -EINVAL;
494 }
495 sb->chunksize = cpu_to_le32(chunksize);
496
497 daemon_sleep = bitmap->mddev->bitmap_info.daemon_sleep;
498 if (!daemon_sleep ||
499 (daemon_sleep < 1) || (daemon_sleep > MAX_SCHEDULE_TIMEOUT)) {
500 printk(KERN_INFO "Choosing daemon_sleep default (5 sec)\n");
501 daemon_sleep = 5 * HZ;
502 }
503 sb->daemon_sleep = cpu_to_le32(daemon_sleep);
504 bitmap->mddev->bitmap_info.daemon_sleep = daemon_sleep;
505
506 /*
507 * FIXME: write_behind for RAID1. If not specified, what
508 * is a good choice? We choose COUNTER_MAX / 2 arbitrarily.
509 */
510 write_behind = bitmap->mddev->bitmap_info.max_write_behind;
511 if (write_behind > COUNTER_MAX)
512 write_behind = COUNTER_MAX / 2;
513 sb->write_behind = cpu_to_le32(write_behind);
514 bitmap->mddev->bitmap_info.max_write_behind = write_behind;
515
516 /* keep the array size field of the bitmap superblock up to date */
517 sb->sync_size = cpu_to_le64(bitmap->mddev->resync_max_sectors);
518
519 memcpy(sb->uuid, bitmap->mddev->uuid, 16);
520
NeilBrownb405fe92012-05-22 13:55:15 +1000521 set_bit(BITMAP_STALE, &bitmap->flags);
NeilBrown84e92342012-05-22 13:55:14 +1000522 sb->state = cpu_to_le32(bitmap->flags);
Jonathan Brassow9c810752011-06-08 17:59:30 -0500523 bitmap->events_cleared = bitmap->mddev->events;
524 sb->events_cleared = cpu_to_le64(bitmap->mddev->events);
525
Cong Wangb2f46e62011-11-28 13:25:44 +0800526 kunmap_atomic(sb);
Jonathan Brassow9c810752011-06-08 17:59:30 -0500527
528 return 0;
529}
530
NeilBrown32a76272005-06-21 17:17:14 -0700531/* read the superblock from the bitmap file and initialize some bitmap fields */
532static int bitmap_read_sb(struct bitmap *bitmap)
533{
534 char *reason = NULL;
535 bitmap_super_t *sb;
NeilBrown4b6d2872005-09-09 16:23:47 -0700536 unsigned long chunksize, daemon_sleep, write_behind;
NeilBrown32a76272005-06-21 17:17:14 -0700537 unsigned long long events;
NeilBrown1dff2b82012-05-22 13:55:34 +1000538 unsigned long sectors_reserved = 0;
NeilBrown32a76272005-06-21 17:17:14 -0700539 int err = -EINVAL;
NeilBrown27581e52012-05-22 13:55:08 +1000540 struct page *sb_page;
NeilBrown32a76272005-06-21 17:17:14 -0700541
NeilBrown1ec885c2012-05-22 13:55:10 +1000542 if (!bitmap->storage.file && !bitmap->mddev->bitmap_info.offset) {
NeilBrownef99bf42012-05-22 13:55:08 +1000543 chunksize = 128 * 1024 * 1024;
544 daemon_sleep = 5 * HZ;
545 write_behind = 0;
NeilBrownb405fe92012-05-22 13:55:15 +1000546 set_bit(BITMAP_STALE, &bitmap->flags);
NeilBrownef99bf42012-05-22 13:55:08 +1000547 err = 0;
548 goto out_no_sb;
549 }
NeilBrown32a76272005-06-21 17:17:14 -0700550 /* page 0 is the superblock, read it... */
NeilBrown27581e52012-05-22 13:55:08 +1000551 sb_page = alloc_page(GFP_KERNEL);
552 if (!sb_page)
553 return -ENOMEM;
NeilBrown1ec885c2012-05-22 13:55:10 +1000554 bitmap->storage.sb_page = sb_page;
NeilBrown27581e52012-05-22 13:55:08 +1000555
NeilBrown1ec885c2012-05-22 13:55:10 +1000556 if (bitmap->storage.file) {
557 loff_t isize = i_size_read(bitmap->storage.file->f_mapping->host);
NeilBrownf49d5e62007-01-26 00:57:03 -0800558 int bytes = isize > PAGE_SIZE ? PAGE_SIZE : isize;
559
NeilBrown1ec885c2012-05-22 13:55:10 +1000560 err = read_page(bitmap->storage.file, 0,
NeilBrown27581e52012-05-22 13:55:08 +1000561 bitmap, bytes, sb_page);
NeilBrownf49d5e62007-01-26 00:57:03 -0800562 } else {
NeilBrown27581e52012-05-22 13:55:08 +1000563 err = read_sb_page(bitmap->mddev,
564 bitmap->mddev->bitmap_info.offset,
565 sb_page,
566 0, sizeof(bitmap_super_t));
NeilBrowna654b9d82005-06-21 17:17:27 -0700567 }
NeilBrown27581e52012-05-22 13:55:08 +1000568 if (err)
NeilBrown32a76272005-06-21 17:17:14 -0700569 return err;
NeilBrown32a76272005-06-21 17:17:14 -0700570
NeilBrown27581e52012-05-22 13:55:08 +1000571 sb = kmap_atomic(sb_page);
NeilBrown32a76272005-06-21 17:17:14 -0700572
NeilBrown32a76272005-06-21 17:17:14 -0700573 chunksize = le32_to_cpu(sb->chunksize);
NeilBrown1b04be92009-12-14 12:49:53 +1100574 daemon_sleep = le32_to_cpu(sb->daemon_sleep) * HZ;
NeilBrown4b6d2872005-09-09 16:23:47 -0700575 write_behind = le32_to_cpu(sb->write_behind);
NeilBrown1dff2b82012-05-22 13:55:34 +1000576 sectors_reserved = le32_to_cpu(sb->sectors_reserved);
NeilBrown32a76272005-06-21 17:17:14 -0700577
578 /* verify that the bitmap-specific fields are valid */
579 if (sb->magic != cpu_to_le32(BITMAP_MAGIC))
580 reason = "bad magic";
NeilBrownbd926c62005-11-08 21:39:32 -0800581 else if (le32_to_cpu(sb->version) < BITMAP_MAJOR_LO ||
582 le32_to_cpu(sb->version) > BITMAP_MAJOR_HI)
NeilBrown32a76272005-06-21 17:17:14 -0700583 reason = "unrecognized superblock version";
NeilBrown1187cf02009-03-31 14:27:02 +1100584 else if (chunksize < 512)
NeilBrown7dd5d342006-01-06 00:20:39 -0800585 reason = "bitmap chunksize too small";
Jonathan Brassowd7445402011-06-08 18:01:10 -0500586 else if (!is_power_of_2(chunksize))
NeilBrown32a76272005-06-21 17:17:14 -0700587 reason = "bitmap chunksize not a power of 2";
NeilBrown1b04be92009-12-14 12:49:53 +1100588 else if (daemon_sleep < 1 || daemon_sleep > MAX_SCHEDULE_TIMEOUT)
NeilBrown7dd5d342006-01-06 00:20:39 -0800589 reason = "daemon sleep period out of range";
NeilBrown4b6d2872005-09-09 16:23:47 -0700590 else if (write_behind > COUNTER_MAX)
591 reason = "write-behind limit out of range (0 - 16383)";
NeilBrown32a76272005-06-21 17:17:14 -0700592 if (reason) {
593 printk(KERN_INFO "%s: invalid bitmap file superblock: %s\n",
594 bmname(bitmap), reason);
595 goto out;
596 }
597
598 /* keep the array size field of the bitmap superblock up to date */
599 sb->sync_size = cpu_to_le64(bitmap->mddev->resync_max_sectors);
600
NeilBrown278c1ca2012-03-19 12:46:40 +1100601 if (bitmap->mddev->persistent) {
602 /*
603 * We have a persistent array superblock, so compare the
604 * bitmap's UUID and event counter to the mddev's
605 */
606 if (memcmp(sb->uuid, bitmap->mddev->uuid, 16)) {
607 printk(KERN_INFO
608 "%s: bitmap superblock UUID mismatch\n",
609 bmname(bitmap));
610 goto out;
611 }
612 events = le64_to_cpu(sb->events);
613 if (events < bitmap->mddev->events) {
614 printk(KERN_INFO
615 "%s: bitmap file is out of date (%llu < %llu) "
616 "-- forcing full recovery\n",
617 bmname(bitmap), events,
618 (unsigned long long) bitmap->mddev->events);
NeilBrownb405fe92012-05-22 13:55:15 +1000619 set_bit(BITMAP_STALE, &bitmap->flags);
NeilBrown278c1ca2012-03-19 12:46:40 +1100620 }
621 }
NeilBrown32a76272005-06-21 17:17:14 -0700622
NeilBrown32a76272005-06-21 17:17:14 -0700623 /* assign fields using values from superblock */
NeilBrown4f2e6392006-10-21 10:24:09 -0700624 bitmap->flags |= le32_to_cpu(sb->state);
NeilBrownbd926c62005-11-08 21:39:32 -0800625 if (le32_to_cpu(sb->version) == BITMAP_MAJOR_HOSTENDIAN)
NeilBrownb405fe92012-05-22 13:55:15 +1000626 set_bit(BITMAP_HOSTENDIAN, &bitmap->flags);
NeilBrown32a76272005-06-21 17:17:14 -0700627 bitmap->events_cleared = le64_to_cpu(sb->events_cleared);
628 err = 0;
629out:
Cong Wangb2f46e62011-11-28 13:25:44 +0800630 kunmap_atomic(sb);
NeilBrownef99bf42012-05-22 13:55:08 +1000631out_no_sb:
NeilBrownb405fe92012-05-22 13:55:15 +1000632 if (test_bit(BITMAP_STALE, &bitmap->flags))
NeilBrownef99bf42012-05-22 13:55:08 +1000633 bitmap->events_cleared = bitmap->mddev->events;
634 bitmap->mddev->bitmap_info.chunksize = chunksize;
635 bitmap->mddev->bitmap_info.daemon_sleep = daemon_sleep;
636 bitmap->mddev->bitmap_info.max_write_behind = write_behind;
NeilBrown1dff2b82012-05-22 13:55:34 +1000637 if (bitmap->mddev->bitmap_info.space == 0 ||
638 bitmap->mddev->bitmap_info.space > sectors_reserved)
639 bitmap->mddev->bitmap_info.space = sectors_reserved;
NeilBrown32a76272005-06-21 17:17:14 -0700640 if (err)
641 bitmap_print_sb(bitmap);
642 return err;
643}
644
NeilBrown32a76272005-06-21 17:17:14 -0700645/*
646 * general bitmap file operations
647 */
648
NeilBrownece5cff2009-12-14 12:49:56 +1100649/*
650 * on-disk bitmap:
651 *
652 * Use one bit per "chunk" (block set). We do the disk I/O on the bitmap
653 * file a page at a time. There's a superblock at the start of the file.
654 */
NeilBrown32a76272005-06-21 17:17:14 -0700655/* calculate the index of the page that contains this bit */
NeilBrown1ec885c2012-05-22 13:55:10 +1000656static inline unsigned long file_page_index(struct bitmap_storage *store,
657 unsigned long chunk)
NeilBrown32a76272005-06-21 17:17:14 -0700658{
NeilBrown1ec885c2012-05-22 13:55:10 +1000659 if (store->sb_page)
NeilBrownece5cff2009-12-14 12:49:56 +1100660 chunk += sizeof(bitmap_super_t) << 3;
661 return chunk >> PAGE_BIT_SHIFT;
NeilBrown32a76272005-06-21 17:17:14 -0700662}
663
664/* calculate the (bit) offset of this bit within a page */
NeilBrown1ec885c2012-05-22 13:55:10 +1000665static inline unsigned long file_page_offset(struct bitmap_storage *store,
666 unsigned long chunk)
NeilBrown32a76272005-06-21 17:17:14 -0700667{
NeilBrown1ec885c2012-05-22 13:55:10 +1000668 if (store->sb_page)
NeilBrownece5cff2009-12-14 12:49:56 +1100669 chunk += sizeof(bitmap_super_t) << 3;
670 return chunk & (PAGE_BITS - 1);
NeilBrown32a76272005-06-21 17:17:14 -0700671}
672
673/*
674 * return a pointer to the page in the filemap that contains the given bit
675 *
676 * this lookup is complicated by the fact that the bitmap sb might be exactly
677 * 1 page (e.g., x86) or less than 1 page -- so the bitmap might start on page
678 * 0 or page 1
679 */
NeilBrown1ec885c2012-05-22 13:55:10 +1000680static inline struct page *filemap_get_page(struct bitmap_storage *store,
Jonathan Brassow3520fa42011-07-27 11:00:37 +1000681 unsigned long chunk)
NeilBrown32a76272005-06-21 17:17:14 -0700682{
NeilBrown1ec885c2012-05-22 13:55:10 +1000683 if (file_page_index(store, chunk) >= store->file_pages)
NeilBrownac2f40b2010-06-01 19:37:31 +1000684 return NULL;
NeilBrown1ec885c2012-05-22 13:55:10 +1000685 return store->filemap[file_page_index(store, chunk)
686 - file_page_index(store, 0)];
NeilBrown32a76272005-06-21 17:17:14 -0700687}
688
NeilBrownd1244cb2012-05-22 13:55:12 +1000689static int bitmap_storage_alloc(struct bitmap_storage *store,
690 unsigned long chunks, int with_super)
691{
692 int pnum;
693 unsigned long num_pages;
694 unsigned long bytes;
695
696 bytes = DIV_ROUND_UP(chunks, 8);
697 if (with_super)
698 bytes += sizeof(bitmap_super_t);
699
700 num_pages = DIV_ROUND_UP(bytes, PAGE_SIZE);
701
702 store->filemap = kmalloc(sizeof(struct page *)
703 * num_pages, GFP_KERNEL);
704 if (!store->filemap)
705 return -ENOMEM;
706
707 if (with_super && !store->sb_page) {
NeilBrownd60b4792012-05-22 13:55:25 +1000708 store->sb_page = alloc_page(GFP_KERNEL|__GFP_ZERO);
NeilBrownd1244cb2012-05-22 13:55:12 +1000709 if (store->sb_page == NULL)
710 return -ENOMEM;
711 store->sb_page->index = 0;
712 }
713 pnum = 0;
714 if (store->sb_page) {
715 store->filemap[0] = store->sb_page;
716 pnum = 1;
717 }
718 for ( ; pnum < num_pages; pnum++) {
NeilBrownd60b4792012-05-22 13:55:25 +1000719 store->filemap[pnum] = alloc_page(GFP_KERNEL|__GFP_ZERO);
NeilBrownd1244cb2012-05-22 13:55:12 +1000720 if (!store->filemap[pnum]) {
721 store->file_pages = pnum;
722 return -ENOMEM;
723 }
724 store->filemap[pnum]->index = pnum;
725 }
726 store->file_pages = pnum;
727
728 /* We need 4 bits per page, rounded up to a multiple
729 * of sizeof(unsigned long) */
730 store->filemap_attr = kzalloc(
731 roundup(DIV_ROUND_UP(num_pages*4, 8), sizeof(unsigned long)),
732 GFP_KERNEL);
733 if (!store->filemap_attr)
734 return -ENOMEM;
735
736 store->bytes = bytes;
737
738 return 0;
739}
740
NeilBrownfae7d322012-05-22 13:55:21 +1000741static void bitmap_file_unmap(struct bitmap_storage *store)
NeilBrown32a76272005-06-21 17:17:14 -0700742{
743 struct page **map, *sb_page;
NeilBrown32a76272005-06-21 17:17:14 -0700744 int pages;
NeilBrownfae7d322012-05-22 13:55:21 +1000745 struct file *file;
NeilBrown32a76272005-06-21 17:17:14 -0700746
NeilBrownfae7d322012-05-22 13:55:21 +1000747 file = store->file;
NeilBrown1ec885c2012-05-22 13:55:10 +1000748 map = store->filemap;
NeilBrown1ec885c2012-05-22 13:55:10 +1000749 pages = store->file_pages;
NeilBrown1ec885c2012-05-22 13:55:10 +1000750 sb_page = store->sb_page;
NeilBrown32a76272005-06-21 17:17:14 -0700751
752 while (pages--)
NeilBrownece5cff2009-12-14 12:49:56 +1100753 if (map[pages] != sb_page) /* 0 is sb_page, release it below */
NeilBrownd785a062006-06-26 00:27:48 -0700754 free_buffers(map[pages]);
NeilBrown32a76272005-06-21 17:17:14 -0700755 kfree(map);
NeilBrownfae7d322012-05-22 13:55:21 +1000756 kfree(store->filemap_attr);
NeilBrown32a76272005-06-21 17:17:14 -0700757
NeilBrownd785a062006-06-26 00:27:48 -0700758 if (sb_page)
759 free_buffers(sb_page);
NeilBrown32a76272005-06-21 17:17:14 -0700760
NeilBrownd785a062006-06-26 00:27:48 -0700761 if (file) {
Josef Sipekc649bb92006-12-08 02:37:19 -0800762 struct inode *inode = file->f_path.dentry->d_inode;
Andrew Mortonfc0ecff2007-02-10 01:45:39 -0800763 invalidate_mapping_pages(inode->i_mapping, 0, -1);
NeilBrown32a76272005-06-21 17:17:14 -0700764 fput(file);
NeilBrownd785a062006-06-26 00:27:48 -0700765 }
NeilBrown32a76272005-06-21 17:17:14 -0700766}
767
NeilBrown32a76272005-06-21 17:17:14 -0700768/*
769 * bitmap_file_kick - if an error occurs while manipulating the bitmap file
770 * then it is no longer reliable, so we stop using it and we mark the file
771 * as failed in the superblock
772 */
773static void bitmap_file_kick(struct bitmap *bitmap)
774{
775 char *path, *ptr = NULL;
776
NeilBrownb405fe92012-05-22 13:55:15 +1000777 if (!test_and_set_bit(BITMAP_STALE, &bitmap->flags)) {
NeilBrown4ad13662007-07-17 04:06:13 -0700778 bitmap_update_sb(bitmap);
NeilBrown32a76272005-06-21 17:17:14 -0700779
NeilBrown1ec885c2012-05-22 13:55:10 +1000780 if (bitmap->storage.file) {
NeilBrown4ad13662007-07-17 04:06:13 -0700781 path = kmalloc(PAGE_SIZE, GFP_KERNEL);
782 if (path)
NeilBrown1ec885c2012-05-22 13:55:10 +1000783 ptr = d_path(&bitmap->storage.file->f_path,
784 path, PAGE_SIZE);
Christoph Hellwig6bcfd602008-05-23 13:04:34 -0700785
NeilBrown4ad13662007-07-17 04:06:13 -0700786 printk(KERN_ALERT
787 "%s: kicking failed bitmap file %s from array!\n",
Christoph Hellwig6bcfd602008-05-23 13:04:34 -0700788 bmname(bitmap), IS_ERR(ptr) ? "" : ptr);
NeilBrown32a76272005-06-21 17:17:14 -0700789
NeilBrown4ad13662007-07-17 04:06:13 -0700790 kfree(path);
791 } else
792 printk(KERN_ALERT
793 "%s: disabling internal bitmap due to errors\n",
794 bmname(bitmap));
NeilBrowna654b9d82005-06-21 17:17:27 -0700795 }
NeilBrown32a76272005-06-21 17:17:14 -0700796}
797
798enum bitmap_page_attr {
NeilBrownac2f40b2010-06-01 19:37:31 +1000799 BITMAP_PAGE_DIRTY = 0, /* there are set bits that need to be synced */
NeilBrown5a537df2011-09-21 15:37:46 +1000800 BITMAP_PAGE_PENDING = 1, /* there are bits that are being cleaned.
801 * i.e. counter is 1 or 2. */
NeilBrownac2f40b2010-06-01 19:37:31 +1000802 BITMAP_PAGE_NEEDWRITE = 2, /* there are cleared bits that need to be synced */
NeilBrown32a76272005-06-21 17:17:14 -0700803};
804
NeilBrownd1891222012-05-22 13:55:09 +1000805static inline void set_page_attr(struct bitmap *bitmap, int pnum,
806 enum bitmap_page_attr attr)
NeilBrown32a76272005-06-21 17:17:14 -0700807{
NeilBrownbdfd1142012-05-22 13:55:22 +1000808 set_bit((pnum<<2) + attr, bitmap->storage.filemap_attr);
NeilBrown32a76272005-06-21 17:17:14 -0700809}
810
NeilBrownd1891222012-05-22 13:55:09 +1000811static inline void clear_page_attr(struct bitmap *bitmap, int pnum,
812 enum bitmap_page_attr attr)
NeilBrown32a76272005-06-21 17:17:14 -0700813{
NeilBrownbdfd1142012-05-22 13:55:22 +1000814 clear_bit((pnum<<2) + attr, bitmap->storage.filemap_attr);
NeilBrown32a76272005-06-21 17:17:14 -0700815}
816
NeilBrownbdfd1142012-05-22 13:55:22 +1000817static inline int test_page_attr(struct bitmap *bitmap, int pnum,
818 enum bitmap_page_attr attr)
NeilBrown32a76272005-06-21 17:17:14 -0700819{
NeilBrown1ec885c2012-05-22 13:55:10 +1000820 return test_bit((pnum<<2) + attr, bitmap->storage.filemap_attr);
NeilBrown32a76272005-06-21 17:17:14 -0700821}
822
NeilBrownbdfd1142012-05-22 13:55:22 +1000823static inline int test_and_clear_page_attr(struct bitmap *bitmap, int pnum,
824 enum bitmap_page_attr attr)
825{
826 return test_and_clear_bit((pnum<<2) + attr,
827 bitmap->storage.filemap_attr);
828}
NeilBrown32a76272005-06-21 17:17:14 -0700829/*
830 * bitmap_file_set_bit -- called before performing a write to the md device
831 * to set (and eventually sync) a particular bit in the bitmap file
832 *
833 * we set the bit immediately, then we record the page number so that
834 * when an unplug occurs, we can flush the dirty pages out to disk
835 */
836static void bitmap_file_set_bit(struct bitmap *bitmap, sector_t block)
837{
838 unsigned long bit;
Jonathan Brassow3520fa42011-07-27 11:00:37 +1000839 struct page *page;
NeilBrown32a76272005-06-21 17:17:14 -0700840 void *kaddr;
NeilBrown40cffcc2012-05-22 13:55:24 +1000841 unsigned long chunk = block >> bitmap->counts.chunkshift;
NeilBrown32a76272005-06-21 17:17:14 -0700842
NeilBrown1ec885c2012-05-22 13:55:10 +1000843 page = filemap_get_page(&bitmap->storage, chunk);
Jonathan Brassow3520fa42011-07-27 11:00:37 +1000844 if (!page)
845 return;
NeilBrown1ec885c2012-05-22 13:55:10 +1000846 bit = file_page_offset(&bitmap->storage, chunk);
NeilBrown32a76272005-06-21 17:17:14 -0700847
Jonathan Brassow3520fa42011-07-27 11:00:37 +1000848 /* set the bit */
Cong Wangb2f46e62011-11-28 13:25:44 +0800849 kaddr = kmap_atomic(page);
NeilBrownb405fe92012-05-22 13:55:15 +1000850 if (test_bit(BITMAP_HOSTENDIAN, &bitmap->flags))
Jonathan Brassow3520fa42011-07-27 11:00:37 +1000851 set_bit(bit, kaddr);
852 else
NeilBrown63c68262012-05-22 13:55:23 +1000853 test_and_set_bit_le(bit, kaddr);
Cong Wangb2f46e62011-11-28 13:25:44 +0800854 kunmap_atomic(kaddr);
NeilBrown36a4e1f2011-10-07 14:23:17 +1100855 pr_debug("set file bit %lu page %lu\n", bit, page->index);
NeilBrown32a76272005-06-21 17:17:14 -0700856 /* record page number so it gets flushed to disk when unplug occurs */
NeilBrownd1891222012-05-22 13:55:09 +1000857 set_page_attr(bitmap, page->index, BITMAP_PAGE_DIRTY);
NeilBrown32a76272005-06-21 17:17:14 -0700858}
859
NeilBrownef99bf42012-05-22 13:55:08 +1000860static void bitmap_file_clear_bit(struct bitmap *bitmap, sector_t block)
861{
862 unsigned long bit;
863 struct page *page;
864 void *paddr;
NeilBrown40cffcc2012-05-22 13:55:24 +1000865 unsigned long chunk = block >> bitmap->counts.chunkshift;
NeilBrownef99bf42012-05-22 13:55:08 +1000866
NeilBrown1ec885c2012-05-22 13:55:10 +1000867 page = filemap_get_page(&bitmap->storage, chunk);
NeilBrownef99bf42012-05-22 13:55:08 +1000868 if (!page)
869 return;
NeilBrown1ec885c2012-05-22 13:55:10 +1000870 bit = file_page_offset(&bitmap->storage, chunk);
NeilBrownef99bf42012-05-22 13:55:08 +1000871 paddr = kmap_atomic(page);
NeilBrownb405fe92012-05-22 13:55:15 +1000872 if (test_bit(BITMAP_HOSTENDIAN, &bitmap->flags))
NeilBrownef99bf42012-05-22 13:55:08 +1000873 clear_bit(bit, paddr);
874 else
NeilBrown63c68262012-05-22 13:55:23 +1000875 test_and_clear_bit_le(bit, paddr);
NeilBrownef99bf42012-05-22 13:55:08 +1000876 kunmap_atomic(paddr);
NeilBrownd1891222012-05-22 13:55:09 +1000877 if (!test_page_attr(bitmap, page->index, BITMAP_PAGE_NEEDWRITE)) {
878 set_page_attr(bitmap, page->index, BITMAP_PAGE_PENDING);
NeilBrownef99bf42012-05-22 13:55:08 +1000879 bitmap->allclean = 0;
880 }
881}
882
NeilBrown32a76272005-06-21 17:17:14 -0700883/* this gets called when the md device is ready to unplug its underlying
884 * (slave) device queues -- before we let any writes go down, we need to
885 * sync the dirty pages of the bitmap file to disk */
NeilBrown4ad13662007-07-17 04:06:13 -0700886void bitmap_unplug(struct bitmap *bitmap)
NeilBrown32a76272005-06-21 17:17:14 -0700887{
NeilBrown74667122012-05-22 13:55:19 +1000888 unsigned long i;
NeilBrownec7a3192006-06-26 00:27:45 -0700889 int dirty, need_write;
NeilBrown32a76272005-06-21 17:17:14 -0700890 int wait = 0;
891
NeilBrown62f82fa2012-05-22 13:55:21 +1000892 if (!bitmap || !bitmap->storage.filemap ||
893 test_bit(BITMAP_STALE, &bitmap->flags))
NeilBrown4ad13662007-07-17 04:06:13 -0700894 return;
NeilBrown32a76272005-06-21 17:17:14 -0700895
896 /* look at each page to see if there are any set bits that need to be
897 * flushed out to disk */
NeilBrown1ec885c2012-05-22 13:55:10 +1000898 for (i = 0; i < bitmap->storage.file_pages; i++) {
NeilBrownbdfd1142012-05-22 13:55:22 +1000899 if (!bitmap->storage.filemap)
NeilBrown4ad13662007-07-17 04:06:13 -0700900 return;
NeilBrownbdfd1142012-05-22 13:55:22 +1000901 dirty = test_and_clear_page_attr(bitmap, i, BITMAP_PAGE_DIRTY);
902 need_write = test_and_clear_page_attr(bitmap, i,
903 BITMAP_PAGE_NEEDWRITE);
904 if (dirty || need_write) {
NeilBrownd1891222012-05-22 13:55:09 +1000905 clear_page_attr(bitmap, i, BITMAP_PAGE_PENDING);
NeilBrownbdfd1142012-05-22 13:55:22 +1000906 write_page(bitmap, bitmap->storage.filemap[i], 0);
907 }
NeilBrownec7a3192006-06-26 00:27:45 -0700908 if (dirty)
NeilBrown32a76272005-06-21 17:17:14 -0700909 wait = 1;
NeilBrown32a76272005-06-21 17:17:14 -0700910 }
911 if (wait) { /* if any writes were performed, we need to wait on them */
NeilBrown1ec885c2012-05-22 13:55:10 +1000912 if (bitmap->storage.file)
NeilBrownd785a062006-06-26 00:27:48 -0700913 wait_event(bitmap->write_wait,
914 atomic_read(&bitmap->pending_writes)==0);
NeilBrown0b79ccf2006-06-26 00:27:44 -0700915 else
NeilBrowna9701a32005-11-08 21:39:34 -0800916 md_super_wait(bitmap->mddev);
NeilBrown32a76272005-06-21 17:17:14 -0700917 }
NeilBrownb405fe92012-05-22 13:55:15 +1000918 if (test_bit(BITMAP_WRITE_ERROR, &bitmap->flags))
NeilBrownd785a062006-06-26 00:27:48 -0700919 bitmap_file_kick(bitmap);
NeilBrown32a76272005-06-21 17:17:14 -0700920}
NeilBrownac2f40b2010-06-01 19:37:31 +1000921EXPORT_SYMBOL(bitmap_unplug);
NeilBrown32a76272005-06-21 17:17:14 -0700922
NeilBrown6a079972005-09-09 16:23:44 -0700923static void bitmap_set_memory_bits(struct bitmap *bitmap, sector_t offset, int needed);
NeilBrown32a76272005-06-21 17:17:14 -0700924/* * bitmap_init_from_disk -- called at bitmap_create time to initialize
925 * the in-memory bitmap from the on-disk bitmap -- also, sets up the
926 * memory mapping of the bitmap file
927 * Special cases:
928 * if there's no bitmap file, or if the bitmap file had been
929 * previously kicked from the array, we mark all the bits as
930 * 1's in order to cause a full resync.
NeilBrown6a079972005-09-09 16:23:44 -0700931 *
932 * We ignore all bits for sectors that end earlier than 'start'.
933 * This is used when reading an out-of-date bitmap...
NeilBrown32a76272005-06-21 17:17:14 -0700934 */
NeilBrown6a079972005-09-09 16:23:44 -0700935static int bitmap_init_from_disk(struct bitmap *bitmap, sector_t start)
NeilBrown32a76272005-06-21 17:17:14 -0700936{
937 unsigned long i, chunks, index, oldindex, bit;
NeilBrown27581e52012-05-22 13:55:08 +1000938 struct page *page = NULL;
NeilBrownd1244cb2012-05-22 13:55:12 +1000939 unsigned long bit_cnt = 0;
NeilBrown32a76272005-06-21 17:17:14 -0700940 struct file *file;
NeilBrownd1244cb2012-05-22 13:55:12 +1000941 unsigned long offset;
NeilBrown32a76272005-06-21 17:17:14 -0700942 int outofdate;
943 int ret = -ENOSPC;
NeilBrownea03aff2006-01-06 00:20:34 -0800944 void *paddr;
NeilBrown1ec885c2012-05-22 13:55:10 +1000945 struct bitmap_storage *store = &bitmap->storage;
NeilBrown32a76272005-06-21 17:17:14 -0700946
NeilBrown40cffcc2012-05-22 13:55:24 +1000947 chunks = bitmap->counts.chunks;
NeilBrown1ec885c2012-05-22 13:55:10 +1000948 file = store->file;
NeilBrown32a76272005-06-21 17:17:14 -0700949
NeilBrownef99bf42012-05-22 13:55:08 +1000950 if (!file && !bitmap->mddev->bitmap_info.offset) {
951 /* No permanent bitmap - fill with '1s'. */
NeilBrown1ec885c2012-05-22 13:55:10 +1000952 store->filemap = NULL;
953 store->file_pages = 0;
NeilBrownef99bf42012-05-22 13:55:08 +1000954 for (i = 0; i < chunks ; i++) {
955 /* if the disk bit is set, set the memory bit */
NeilBrown40cffcc2012-05-22 13:55:24 +1000956 int needed = ((sector_t)(i+1) << (bitmap->counts.chunkshift)
NeilBrownef99bf42012-05-22 13:55:08 +1000957 >= start);
958 bitmap_set_memory_bits(bitmap,
NeilBrown40cffcc2012-05-22 13:55:24 +1000959 (sector_t)i << bitmap->counts.chunkshift,
NeilBrownef99bf42012-05-22 13:55:08 +1000960 needed);
961 }
962 return 0;
963 }
NeilBrown32a76272005-06-21 17:17:14 -0700964
NeilBrownb405fe92012-05-22 13:55:15 +1000965 outofdate = test_bit(BITMAP_STALE, &bitmap->flags);
NeilBrown32a76272005-06-21 17:17:14 -0700966 if (outofdate)
967 printk(KERN_INFO "%s: bitmap file is out of date, doing full "
968 "recovery\n", bmname(bitmap));
969
NeilBrownd1244cb2012-05-22 13:55:12 +1000970 if (file && i_size_read(file->f_mapping->host) < store->bytes) {
NeilBrown32a76272005-06-21 17:17:14 -0700971 printk(KERN_INFO "%s: bitmap file too short %lu < %lu\n",
NeilBrownd1244cb2012-05-22 13:55:12 +1000972 bmname(bitmap),
973 (unsigned long) i_size_read(file->f_mapping->host),
974 store->bytes);
NeilBrown4ad13662007-07-17 04:06:13 -0700975 goto err;
NeilBrown32a76272005-06-21 17:17:14 -0700976 }
NeilBrownbc7f77d2005-06-21 17:17:17 -0700977
NeilBrown32a76272005-06-21 17:17:14 -0700978 oldindex = ~0L;
NeilBrownd1244cb2012-05-22 13:55:12 +1000979 offset = 0;
980 if (!bitmap->mddev->bitmap_info.external)
981 offset = sizeof(bitmap_super_t);
NeilBrown32a76272005-06-21 17:17:14 -0700982
983 for (i = 0; i < chunks; i++) {
NeilBrownbd926c62005-11-08 21:39:32 -0800984 int b;
NeilBrown1ec885c2012-05-22 13:55:10 +1000985 index = file_page_index(&bitmap->storage, i);
986 bit = file_page_offset(&bitmap->storage, i);
NeilBrown32a76272005-06-21 17:17:14 -0700987 if (index != oldindex) { /* this is a new page, read it in */
NeilBrownd785a062006-06-26 00:27:48 -0700988 int count;
NeilBrown32a76272005-06-21 17:17:14 -0700989 /* unmap the old page, we're done with it */
NeilBrownd1244cb2012-05-22 13:55:12 +1000990 if (index == store->file_pages-1)
991 count = store->bytes - index * PAGE_SIZE;
NeilBrownd785a062006-06-26 00:27:48 -0700992 else
993 count = PAGE_SIZE;
NeilBrown1ec885c2012-05-22 13:55:10 +1000994 page = store->filemap[index];
NeilBrown27581e52012-05-22 13:55:08 +1000995 if (file)
996 ret = read_page(file, index, bitmap,
997 count, page);
998 else
999 ret = read_sb_page(
1000 bitmap->mddev,
1001 bitmap->mddev->bitmap_info.offset,
1002 page,
1003 index, count);
1004
1005 if (ret)
NeilBrown4ad13662007-07-17 04:06:13 -07001006 goto err;
NeilBrowna654b9d82005-06-21 17:17:27 -07001007
NeilBrown32a76272005-06-21 17:17:14 -07001008 oldindex = index;
NeilBrown32a76272005-06-21 17:17:14 -07001009
1010 if (outofdate) {
1011 /*
1012 * if bitmap is out of date, dirty the
NeilBrownac2f40b2010-06-01 19:37:31 +10001013 * whole page and write it out
NeilBrown32a76272005-06-21 17:17:14 -07001014 */
Cong Wangb2f46e62011-11-28 13:25:44 +08001015 paddr = kmap_atomic(page);
NeilBrownea03aff2006-01-06 00:20:34 -08001016 memset(paddr + offset, 0xff,
NeilBrown6a079972005-09-09 16:23:44 -07001017 PAGE_SIZE - offset);
Cong Wangb2f46e62011-11-28 13:25:44 +08001018 kunmap_atomic(paddr);
NeilBrown4ad13662007-07-17 04:06:13 -07001019 write_page(bitmap, page, 1);
1020
1021 ret = -EIO;
NeilBrownb405fe92012-05-22 13:55:15 +10001022 if (test_bit(BITMAP_WRITE_ERROR,
1023 &bitmap->flags))
NeilBrown4ad13662007-07-17 04:06:13 -07001024 goto err;
NeilBrown32a76272005-06-21 17:17:14 -07001025 }
NeilBrown32a76272005-06-21 17:17:14 -07001026 }
Cong Wangb2f46e62011-11-28 13:25:44 +08001027 paddr = kmap_atomic(page);
NeilBrownb405fe92012-05-22 13:55:15 +10001028 if (test_bit(BITMAP_HOSTENDIAN, &bitmap->flags))
NeilBrownea03aff2006-01-06 00:20:34 -08001029 b = test_bit(bit, paddr);
NeilBrownbd926c62005-11-08 21:39:32 -08001030 else
Akinobu Mita6b33aff2011-03-23 16:42:13 -07001031 b = test_bit_le(bit, paddr);
Cong Wangb2f46e62011-11-28 13:25:44 +08001032 kunmap_atomic(paddr);
NeilBrownbd926c62005-11-08 21:39:32 -08001033 if (b) {
NeilBrown32a76272005-06-21 17:17:14 -07001034 /* if the disk bit is set, set the memory bit */
NeilBrown40cffcc2012-05-22 13:55:24 +10001035 int needed = ((sector_t)(i+1) << bitmap->counts.chunkshift
NeilBrowndb305e52009-05-07 12:49:06 +10001036 >= start);
1037 bitmap_set_memory_bits(bitmap,
NeilBrown40cffcc2012-05-22 13:55:24 +10001038 (sector_t)i << bitmap->counts.chunkshift,
NeilBrowndb305e52009-05-07 12:49:06 +10001039 needed);
NeilBrown32a76272005-06-21 17:17:14 -07001040 bit_cnt++;
1041 }
NeilBrown27581e52012-05-22 13:55:08 +10001042 offset = 0;
NeilBrown32a76272005-06-21 17:17:14 -07001043 }
1044
NeilBrown32a76272005-06-21 17:17:14 -07001045 printk(KERN_INFO "%s: bitmap initialized from disk: "
NeilBrownd1244cb2012-05-22 13:55:12 +10001046 "read %lu pages, set %lu of %lu bits\n",
NeilBrown1ec885c2012-05-22 13:55:10 +10001047 bmname(bitmap), store->file_pages,
NeilBrownd1244cb2012-05-22 13:55:12 +10001048 bit_cnt, chunks);
NeilBrown32a76272005-06-21 17:17:14 -07001049
NeilBrown4ad13662007-07-17 04:06:13 -07001050 return 0;
1051
1052 err:
1053 printk(KERN_INFO "%s: bitmap initialisation failed: %d\n",
1054 bmname(bitmap), ret);
NeilBrown32a76272005-06-21 17:17:14 -07001055 return ret;
1056}
1057
NeilBrowna654b9d82005-06-21 17:17:27 -07001058void bitmap_write_all(struct bitmap *bitmap)
1059{
1060 /* We don't actually write all bitmap blocks here,
1061 * just flag them as needing to be written
1062 */
NeilBrownec7a3192006-06-26 00:27:45 -07001063 int i;
NeilBrowna654b9d82005-06-21 17:17:27 -07001064
NeilBrown1ec885c2012-05-22 13:55:10 +10001065 if (!bitmap || !bitmap->storage.filemap)
NeilBrownef99bf42012-05-22 13:55:08 +10001066 return;
NeilBrown1ec885c2012-05-22 13:55:10 +10001067 if (bitmap->storage.file)
NeilBrownef99bf42012-05-22 13:55:08 +10001068 /* Only one copy, so nothing needed */
1069 return;
1070
NeilBrown1ec885c2012-05-22 13:55:10 +10001071 for (i = 0; i < bitmap->storage.file_pages; i++)
NeilBrownd1891222012-05-22 13:55:09 +10001072 set_page_attr(bitmap, i,
NeilBrownec7a3192006-06-26 00:27:45 -07001073 BITMAP_PAGE_NEEDWRITE);
NeilBrown2585f3e2011-09-21 15:37:46 +10001074 bitmap->allclean = 0;
NeilBrowna654b9d82005-06-21 17:17:27 -07001075}
1076
NeilBrown40cffcc2012-05-22 13:55:24 +10001077static void bitmap_count_page(struct bitmap_counts *bitmap,
1078 sector_t offset, int inc)
NeilBrown32a76272005-06-21 17:17:14 -07001079{
NeilBrown61a0d802012-03-19 12:46:41 +11001080 sector_t chunk = offset >> bitmap->chunkshift;
NeilBrown32a76272005-06-21 17:17:14 -07001081 unsigned long page = chunk >> PAGE_COUNTER_SHIFT;
1082 bitmap->bp[page].count += inc;
NeilBrown32a76272005-06-21 17:17:14 -07001083 bitmap_checkfree(bitmap, page);
1084}
NeilBrownbf07bb72012-05-22 13:55:06 +10001085
NeilBrown40cffcc2012-05-22 13:55:24 +10001086static void bitmap_set_pending(struct bitmap_counts *bitmap, sector_t offset)
NeilBrownbf07bb72012-05-22 13:55:06 +10001087{
1088 sector_t chunk = offset >> bitmap->chunkshift;
1089 unsigned long page = chunk >> PAGE_COUNTER_SHIFT;
1090 struct bitmap_page *bp = &bitmap->bp[page];
1091
1092 if (!bp->pending)
1093 bp->pending = 1;
1094}
1095
NeilBrown40cffcc2012-05-22 13:55:24 +10001096static bitmap_counter_t *bitmap_get_counter(struct bitmap_counts *bitmap,
NeilBrown57dab0b2010-10-19 10:03:39 +11001097 sector_t offset, sector_t *blocks,
NeilBrown32a76272005-06-21 17:17:14 -07001098 int create);
1099
1100/*
1101 * bitmap daemon -- periodically wakes up to clean bits and flush pages
1102 * out to disk
1103 */
1104
NeilBrownfd01b882011-10-11 16:47:53 +11001105void bitmap_daemon_work(struct mddev *mddev)
NeilBrown32a76272005-06-21 17:17:14 -07001106{
NeilBrownaa5cbd12009-12-14 12:49:46 +11001107 struct bitmap *bitmap;
NeilBrownaa3163f2005-06-21 17:17:22 -07001108 unsigned long j;
NeilBrownbf07bb72012-05-22 13:55:06 +10001109 unsigned long nextpage;
NeilBrown57dab0b2010-10-19 10:03:39 +11001110 sector_t blocks;
NeilBrown40cffcc2012-05-22 13:55:24 +10001111 struct bitmap_counts *counts;
NeilBrown32a76272005-06-21 17:17:14 -07001112
NeilBrownaa5cbd12009-12-14 12:49:46 +11001113 /* Use a mutex to guard daemon_work against
1114 * bitmap_destroy.
1115 */
NeilBrownc3d97142009-12-14 12:49:52 +11001116 mutex_lock(&mddev->bitmap_info.mutex);
NeilBrownaa5cbd12009-12-14 12:49:46 +11001117 bitmap = mddev->bitmap;
1118 if (bitmap == NULL) {
NeilBrownc3d97142009-12-14 12:49:52 +11001119 mutex_unlock(&mddev->bitmap_info.mutex);
NeilBrown4ad13662007-07-17 04:06:13 -07001120 return;
NeilBrownaa5cbd12009-12-14 12:49:46 +11001121 }
NeilBrown42a04b52009-12-14 12:49:53 +11001122 if (time_before(jiffies, bitmap->daemon_lastrun
NeilBrown2e61ebb2011-12-23 10:17:50 +11001123 + mddev->bitmap_info.daemon_sleep))
NeilBrown7be3dfe2008-03-10 11:43:48 -07001124 goto done;
1125
NeilBrown32a76272005-06-21 17:17:14 -07001126 bitmap->daemon_lastrun = jiffies;
NeilBrown8311c292008-03-04 14:29:30 -08001127 if (bitmap->allclean) {
NeilBrown2e61ebb2011-12-23 10:17:50 +11001128 mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
NeilBrownaa5cbd12009-12-14 12:49:46 +11001129 goto done;
NeilBrown8311c292008-03-04 14:29:30 -08001130 }
1131 bitmap->allclean = 1;
NeilBrown32a76272005-06-21 17:17:14 -07001132
NeilBrownbf07bb72012-05-22 13:55:06 +10001133 /* Any file-page which is PENDING now needs to be written.
1134 * So set NEEDWRITE now, then after we make any last-minute changes
1135 * we will write it.
1136 */
NeilBrown1ec885c2012-05-22 13:55:10 +10001137 for (j = 0; j < bitmap->storage.file_pages; j++)
NeilBrownbdfd1142012-05-22 13:55:22 +10001138 if (test_and_clear_page_attr(bitmap, j,
1139 BITMAP_PAGE_PENDING))
NeilBrownd1891222012-05-22 13:55:09 +10001140 set_page_attr(bitmap, j,
NeilBrownbf07bb72012-05-22 13:55:06 +10001141 BITMAP_PAGE_NEEDWRITE);
NeilBrownbf07bb72012-05-22 13:55:06 +10001142
1143 if (bitmap->need_sync &&
1144 mddev->bitmap_info.external == 0) {
1145 /* Arrange for superblock update as well as
1146 * other changes */
1147 bitmap_super_t *sb;
1148 bitmap->need_sync = 0;
NeilBrown1ec885c2012-05-22 13:55:10 +10001149 if (bitmap->storage.filemap) {
1150 sb = kmap_atomic(bitmap->storage.sb_page);
NeilBrownef99bf42012-05-22 13:55:08 +10001151 sb->events_cleared =
1152 cpu_to_le64(bitmap->events_cleared);
1153 kunmap_atomic(sb);
NeilBrownd1891222012-05-22 13:55:09 +10001154 set_page_attr(bitmap, 0,
NeilBrownef99bf42012-05-22 13:55:08 +10001155 BITMAP_PAGE_NEEDWRITE);
1156 }
NeilBrownbf07bb72012-05-22 13:55:06 +10001157 }
1158 /* Now look at the bitmap counters and if any are '2' or '1',
1159 * decrement and handle accordingly.
1160 */
NeilBrown40cffcc2012-05-22 13:55:24 +10001161 counts = &bitmap->counts;
1162 spin_lock_irq(&counts->lock);
NeilBrownbf07bb72012-05-22 13:55:06 +10001163 nextpage = 0;
NeilBrown40cffcc2012-05-22 13:55:24 +10001164 for (j = 0; j < counts->chunks; j++) {
NeilBrown32a76272005-06-21 17:17:14 -07001165 bitmap_counter_t *bmc;
NeilBrown40cffcc2012-05-22 13:55:24 +10001166 sector_t block = (sector_t)j << counts->chunkshift;
Jonathan Brassow3520fa42011-07-27 11:00:37 +10001167
NeilBrownbf07bb72012-05-22 13:55:06 +10001168 if (j == nextpage) {
1169 nextpage += PAGE_COUNTER_RATIO;
NeilBrown40cffcc2012-05-22 13:55:24 +10001170 if (!counts->bp[j >> PAGE_COUNTER_SHIFT].pending) {
NeilBrownbf07bb72012-05-22 13:55:06 +10001171 j |= PAGE_COUNTER_MASK;
NeilBrownaa3163f2005-06-21 17:17:22 -07001172 continue;
1173 }
NeilBrown40cffcc2012-05-22 13:55:24 +10001174 counts->bp[j >> PAGE_COUNTER_SHIFT].pending = 0;
NeilBrown32a76272005-06-21 17:17:14 -07001175 }
NeilBrown40cffcc2012-05-22 13:55:24 +10001176 bmc = bitmap_get_counter(counts,
NeilBrownef99bf42012-05-22 13:55:08 +10001177 block,
NeilBrowndb305e52009-05-07 12:49:06 +10001178 &blocks, 0);
NeilBrown32a76272005-06-21 17:17:14 -07001179
NeilBrownbf07bb72012-05-22 13:55:06 +10001180 if (!bmc) {
1181 j |= PAGE_COUNTER_MASK;
1182 continue;
1183 }
1184 if (*bmc == 1 && !bitmap->need_sync) {
1185 /* We can clear the bit */
NeilBrownbf07bb72012-05-22 13:55:06 +10001186 *bmc = 0;
NeilBrown40cffcc2012-05-22 13:55:24 +10001187 bitmap_count_page(counts, block, -1);
NeilBrownef99bf42012-05-22 13:55:08 +10001188 bitmap_file_clear_bit(bitmap, block);
NeilBrownbf07bb72012-05-22 13:55:06 +10001189 } else if (*bmc && *bmc <= 2) {
1190 *bmc = 1;
NeilBrown40cffcc2012-05-22 13:55:24 +10001191 bitmap_set_pending(counts, block);
NeilBrown2585f3e2011-09-21 15:37:46 +10001192 bitmap->allclean = 0;
NeilBrown32a76272005-06-21 17:17:14 -07001193 }
NeilBrown32a76272005-06-21 17:17:14 -07001194 }
NeilBrown40cffcc2012-05-22 13:55:24 +10001195 spin_unlock_irq(&counts->lock);
NeilBrown32a76272005-06-21 17:17:14 -07001196
NeilBrownbf07bb72012-05-22 13:55:06 +10001197 /* Now start writeout on any page in NEEDWRITE that isn't DIRTY.
1198 * DIRTY pages need to be written by bitmap_unplug so it can wait
1199 * for them.
1200 * If we find any DIRTY page we stop there and let bitmap_unplug
1201 * handle all the rest. This is important in the case where
1202 * the first blocking holds the superblock and it has been updated.
1203 * We mustn't write any other blocks before the superblock.
1204 */
NeilBrown62f82fa2012-05-22 13:55:21 +10001205 for (j = 0;
1206 j < bitmap->storage.file_pages
1207 && !test_bit(BITMAP_STALE, &bitmap->flags);
1208 j++) {
NeilBrownbf07bb72012-05-22 13:55:06 +10001209
NeilBrownd1891222012-05-22 13:55:09 +10001210 if (test_page_attr(bitmap, j,
NeilBrownbf07bb72012-05-22 13:55:06 +10001211 BITMAP_PAGE_DIRTY))
1212 /* bitmap_unplug will handle the rest */
1213 break;
NeilBrownbdfd1142012-05-22 13:55:22 +10001214 if (test_and_clear_page_attr(bitmap, j,
1215 BITMAP_PAGE_NEEDWRITE)) {
NeilBrown1ec885c2012-05-22 13:55:10 +10001216 write_page(bitmap, bitmap->storage.filemap[j], 0);
NeilBrownbf07bb72012-05-22 13:55:06 +10001217 }
1218 }
NeilBrownbf07bb72012-05-22 13:55:06 +10001219
NeilBrown7be3dfe2008-03-10 11:43:48 -07001220 done:
NeilBrown8311c292008-03-04 14:29:30 -08001221 if (bitmap->allclean == 0)
NeilBrown2e61ebb2011-12-23 10:17:50 +11001222 mddev->thread->timeout =
1223 mddev->bitmap_info.daemon_sleep;
NeilBrownc3d97142009-12-14 12:49:52 +11001224 mutex_unlock(&mddev->bitmap_info.mutex);
NeilBrown32a76272005-06-21 17:17:14 -07001225}
1226
NeilBrown40cffcc2012-05-22 13:55:24 +10001227static bitmap_counter_t *bitmap_get_counter(struct bitmap_counts *bitmap,
NeilBrown57dab0b2010-10-19 10:03:39 +11001228 sector_t offset, sector_t *blocks,
NeilBrown32a76272005-06-21 17:17:14 -07001229 int create)
NeilBrownee305ac2009-09-23 18:06:44 +10001230__releases(bitmap->lock)
1231__acquires(bitmap->lock)
NeilBrown32a76272005-06-21 17:17:14 -07001232{
1233 /* If 'create', we might release the lock and reclaim it.
1234 * The lock must have been taken with interrupts enabled.
1235 * If !create, we don't release the lock.
1236 */
NeilBrown61a0d802012-03-19 12:46:41 +11001237 sector_t chunk = offset >> bitmap->chunkshift;
NeilBrown32a76272005-06-21 17:17:14 -07001238 unsigned long page = chunk >> PAGE_COUNTER_SHIFT;
1239 unsigned long pageoff = (chunk & PAGE_COUNTER_MASK) << COUNTER_BYTE_SHIFT;
1240 sector_t csize;
NeilBrownef425672010-06-01 19:37:33 +10001241 int err;
NeilBrown32a76272005-06-21 17:17:14 -07001242
NeilBrownef425672010-06-01 19:37:33 +10001243 err = bitmap_checkpage(bitmap, page, create);
1244
1245 if (bitmap->bp[page].hijacked ||
1246 bitmap->bp[page].map == NULL)
NeilBrown61a0d802012-03-19 12:46:41 +11001247 csize = ((sector_t)1) << (bitmap->chunkshift +
NeilBrownef425672010-06-01 19:37:33 +10001248 PAGE_COUNTER_SHIFT - 1);
1249 else
NeilBrown61a0d802012-03-19 12:46:41 +11001250 csize = ((sector_t)1) << bitmap->chunkshift;
NeilBrownef425672010-06-01 19:37:33 +10001251 *blocks = csize - (offset & (csize - 1));
1252
1253 if (err < 0)
NeilBrown32a76272005-06-21 17:17:14 -07001254 return NULL;
NeilBrownef425672010-06-01 19:37:33 +10001255
NeilBrown32a76272005-06-21 17:17:14 -07001256 /* now locked ... */
1257
1258 if (bitmap->bp[page].hijacked) { /* hijacked pointer */
1259 /* should we use the first or second counter field
1260 * of the hijacked pointer? */
1261 int hi = (pageoff > PAGE_COUNTER_MASK);
NeilBrown32a76272005-06-21 17:17:14 -07001262 return &((bitmap_counter_t *)
1263 &bitmap->bp[page].map)[hi];
NeilBrownef425672010-06-01 19:37:33 +10001264 } else /* page is allocated */
NeilBrown32a76272005-06-21 17:17:14 -07001265 return (bitmap_counter_t *)
1266 &(bitmap->bp[page].map[pageoff]);
NeilBrown32a76272005-06-21 17:17:14 -07001267}
1268
NeilBrown4b6d2872005-09-09 16:23:47 -07001269int bitmap_startwrite(struct bitmap *bitmap, sector_t offset, unsigned long sectors, int behind)
NeilBrown32a76272005-06-21 17:17:14 -07001270{
NeilBrownac2f40b2010-06-01 19:37:31 +10001271 if (!bitmap)
1272 return 0;
NeilBrown4b6d2872005-09-09 16:23:47 -07001273
1274 if (behind) {
Paul Clements696fcd52010-03-08 16:02:37 +11001275 int bw;
NeilBrown4b6d2872005-09-09 16:23:47 -07001276 atomic_inc(&bitmap->behind_writes);
Paul Clements696fcd52010-03-08 16:02:37 +11001277 bw = atomic_read(&bitmap->behind_writes);
1278 if (bw > bitmap->behind_writes_used)
1279 bitmap->behind_writes_used = bw;
1280
NeilBrown36a4e1f2011-10-07 14:23:17 +11001281 pr_debug("inc write-behind count %d/%lu\n",
1282 bw, bitmap->mddev->bitmap_info.max_write_behind);
NeilBrown4b6d2872005-09-09 16:23:47 -07001283 }
1284
NeilBrown32a76272005-06-21 17:17:14 -07001285 while (sectors) {
NeilBrown57dab0b2010-10-19 10:03:39 +11001286 sector_t blocks;
NeilBrown32a76272005-06-21 17:17:14 -07001287 bitmap_counter_t *bmc;
1288
NeilBrown40cffcc2012-05-22 13:55:24 +10001289 spin_lock_irq(&bitmap->counts.lock);
1290 bmc = bitmap_get_counter(&bitmap->counts, offset, &blocks, 1);
NeilBrown32a76272005-06-21 17:17:14 -07001291 if (!bmc) {
NeilBrown40cffcc2012-05-22 13:55:24 +10001292 spin_unlock_irq(&bitmap->counts.lock);
NeilBrown32a76272005-06-21 17:17:14 -07001293 return 0;
1294 }
1295
Namhyung Kim27d5ea02011-06-09 11:42:57 +10001296 if (unlikely(COUNTER(*bmc) == COUNTER_MAX)) {
Neil Brownda6e1a32007-02-08 14:20:37 -08001297 DEFINE_WAIT(__wait);
1298 /* note that it is safe to do the prepare_to_wait
1299 * after the test as long as we do it before dropping
1300 * the spinlock.
1301 */
1302 prepare_to_wait(&bitmap->overflow_wait, &__wait,
1303 TASK_UNINTERRUPTIBLE);
NeilBrown40cffcc2012-05-22 13:55:24 +10001304 spin_unlock_irq(&bitmap->counts.lock);
NeilBrownf54a9d02012-08-02 08:33:20 +10001305 schedule();
Neil Brownda6e1a32007-02-08 14:20:37 -08001306 finish_wait(&bitmap->overflow_wait, &__wait);
1307 continue;
1308 }
1309
NeilBrownac2f40b2010-06-01 19:37:31 +10001310 switch (*bmc) {
NeilBrown32a76272005-06-21 17:17:14 -07001311 case 0:
1312 bitmap_file_set_bit(bitmap, offset);
NeilBrown40cffcc2012-05-22 13:55:24 +10001313 bitmap_count_page(&bitmap->counts, offset, 1);
NeilBrown32a76272005-06-21 17:17:14 -07001314 /* fall through */
1315 case 1:
1316 *bmc = 2;
1317 }
Neil Brownda6e1a32007-02-08 14:20:37 -08001318
NeilBrown32a76272005-06-21 17:17:14 -07001319 (*bmc)++;
1320
NeilBrown40cffcc2012-05-22 13:55:24 +10001321 spin_unlock_irq(&bitmap->counts.lock);
NeilBrown32a76272005-06-21 17:17:14 -07001322
1323 offset += blocks;
1324 if (sectors > blocks)
1325 sectors -= blocks;
NeilBrownac2f40b2010-06-01 19:37:31 +10001326 else
1327 sectors = 0;
NeilBrown32a76272005-06-21 17:17:14 -07001328 }
1329 return 0;
1330}
NeilBrownac2f40b2010-06-01 19:37:31 +10001331EXPORT_SYMBOL(bitmap_startwrite);
NeilBrown32a76272005-06-21 17:17:14 -07001332
1333void bitmap_endwrite(struct bitmap *bitmap, sector_t offset, unsigned long sectors,
NeilBrown4b6d2872005-09-09 16:23:47 -07001334 int success, int behind)
NeilBrown32a76272005-06-21 17:17:14 -07001335{
NeilBrownac2f40b2010-06-01 19:37:31 +10001336 if (!bitmap)
1337 return;
NeilBrown4b6d2872005-09-09 16:23:47 -07001338 if (behind) {
NeilBrowne5551902010-03-31 11:21:44 +11001339 if (atomic_dec_and_test(&bitmap->behind_writes))
1340 wake_up(&bitmap->behind_wait);
NeilBrown36a4e1f2011-10-07 14:23:17 +11001341 pr_debug("dec write-behind count %d/%lu\n",
1342 atomic_read(&bitmap->behind_writes),
1343 bitmap->mddev->bitmap_info.max_write_behind);
NeilBrown4b6d2872005-09-09 16:23:47 -07001344 }
1345
NeilBrown32a76272005-06-21 17:17:14 -07001346 while (sectors) {
NeilBrown57dab0b2010-10-19 10:03:39 +11001347 sector_t blocks;
NeilBrown32a76272005-06-21 17:17:14 -07001348 unsigned long flags;
1349 bitmap_counter_t *bmc;
1350
NeilBrown40cffcc2012-05-22 13:55:24 +10001351 spin_lock_irqsave(&bitmap->counts.lock, flags);
1352 bmc = bitmap_get_counter(&bitmap->counts, offset, &blocks, 0);
NeilBrown32a76272005-06-21 17:17:14 -07001353 if (!bmc) {
NeilBrown40cffcc2012-05-22 13:55:24 +10001354 spin_unlock_irqrestore(&bitmap->counts.lock, flags);
NeilBrown32a76272005-06-21 17:17:14 -07001355 return;
1356 }
1357
NeilBrown961902c2011-12-23 09:57:48 +11001358 if (success && !bitmap->mddev->degraded &&
Neil Browna0da84f2008-06-28 08:31:22 +10001359 bitmap->events_cleared < bitmap->mddev->events) {
1360 bitmap->events_cleared = bitmap->mddev->events;
1361 bitmap->need_sync = 1;
NeilBrown5ff5aff2010-06-01 19:37:32 +10001362 sysfs_notify_dirent_safe(bitmap->sysfs_can_clear);
Neil Browna0da84f2008-06-28 08:31:22 +10001363 }
1364
Namhyung Kim27d5ea02011-06-09 11:42:57 +10001365 if (!success && !NEEDED(*bmc))
NeilBrown32a76272005-06-21 17:17:14 -07001366 *bmc |= NEEDED_MASK;
1367
Namhyung Kim27d5ea02011-06-09 11:42:57 +10001368 if (COUNTER(*bmc) == COUNTER_MAX)
Neil Brownda6e1a32007-02-08 14:20:37 -08001369 wake_up(&bitmap->overflow_wait);
1370
NeilBrown32a76272005-06-21 17:17:14 -07001371 (*bmc)--;
NeilBrown2585f3e2011-09-21 15:37:46 +10001372 if (*bmc <= 2) {
NeilBrown40cffcc2012-05-22 13:55:24 +10001373 bitmap_set_pending(&bitmap->counts, offset);
NeilBrown2585f3e2011-09-21 15:37:46 +10001374 bitmap->allclean = 0;
1375 }
NeilBrown40cffcc2012-05-22 13:55:24 +10001376 spin_unlock_irqrestore(&bitmap->counts.lock, flags);
NeilBrown32a76272005-06-21 17:17:14 -07001377 offset += blocks;
1378 if (sectors > blocks)
1379 sectors -= blocks;
NeilBrownac2f40b2010-06-01 19:37:31 +10001380 else
1381 sectors = 0;
NeilBrown32a76272005-06-21 17:17:14 -07001382 }
1383}
NeilBrownac2f40b2010-06-01 19:37:31 +10001384EXPORT_SYMBOL(bitmap_endwrite);
NeilBrown32a76272005-06-21 17:17:14 -07001385
NeilBrown57dab0b2010-10-19 10:03:39 +11001386static int __bitmap_start_sync(struct bitmap *bitmap, sector_t offset, sector_t *blocks,
NeilBrown1187cf02009-03-31 14:27:02 +11001387 int degraded)
NeilBrown32a76272005-06-21 17:17:14 -07001388{
1389 bitmap_counter_t *bmc;
1390 int rv;
1391 if (bitmap == NULL) {/* FIXME or bitmap set as 'failed' */
1392 *blocks = 1024;
1393 return 1; /* always resync if no bitmap */
1394 }
NeilBrown40cffcc2012-05-22 13:55:24 +10001395 spin_lock_irq(&bitmap->counts.lock);
1396 bmc = bitmap_get_counter(&bitmap->counts, offset, blocks, 0);
NeilBrown32a76272005-06-21 17:17:14 -07001397 rv = 0;
1398 if (bmc) {
1399 /* locked */
1400 if (RESYNC(*bmc))
1401 rv = 1;
1402 else if (NEEDED(*bmc)) {
1403 rv = 1;
NeilBrown6a806c52005-07-15 03:56:35 -07001404 if (!degraded) { /* don't set/clear bits if degraded */
1405 *bmc |= RESYNC_MASK;
1406 *bmc &= ~NEEDED_MASK;
1407 }
NeilBrown32a76272005-06-21 17:17:14 -07001408 }
1409 }
NeilBrown40cffcc2012-05-22 13:55:24 +10001410 spin_unlock_irq(&bitmap->counts.lock);
NeilBrown32a76272005-06-21 17:17:14 -07001411 return rv;
1412}
1413
NeilBrown57dab0b2010-10-19 10:03:39 +11001414int bitmap_start_sync(struct bitmap *bitmap, sector_t offset, sector_t *blocks,
NeilBrown1187cf02009-03-31 14:27:02 +11001415 int degraded)
1416{
1417 /* bitmap_start_sync must always report on multiples of whole
1418 * pages, otherwise resync (which is very PAGE_SIZE based) will
1419 * get confused.
1420 * So call __bitmap_start_sync repeatedly (if needed) until
1421 * At least PAGE_SIZE>>9 blocks are covered.
1422 * Return the 'or' of the result.
1423 */
1424 int rv = 0;
NeilBrown57dab0b2010-10-19 10:03:39 +11001425 sector_t blocks1;
NeilBrown1187cf02009-03-31 14:27:02 +11001426
1427 *blocks = 0;
1428 while (*blocks < (PAGE_SIZE>>9)) {
1429 rv |= __bitmap_start_sync(bitmap, offset,
1430 &blocks1, degraded);
1431 offset += blocks1;
1432 *blocks += blocks1;
1433 }
1434 return rv;
1435}
NeilBrownac2f40b2010-06-01 19:37:31 +10001436EXPORT_SYMBOL(bitmap_start_sync);
NeilBrown1187cf02009-03-31 14:27:02 +11001437
NeilBrown57dab0b2010-10-19 10:03:39 +11001438void bitmap_end_sync(struct bitmap *bitmap, sector_t offset, sector_t *blocks, int aborted)
NeilBrown32a76272005-06-21 17:17:14 -07001439{
1440 bitmap_counter_t *bmc;
1441 unsigned long flags;
NeilBrownac2f40b2010-06-01 19:37:31 +10001442
1443 if (bitmap == NULL) {
NeilBrown32a76272005-06-21 17:17:14 -07001444 *blocks = 1024;
1445 return;
1446 }
NeilBrown40cffcc2012-05-22 13:55:24 +10001447 spin_lock_irqsave(&bitmap->counts.lock, flags);
1448 bmc = bitmap_get_counter(&bitmap->counts, offset, blocks, 0);
NeilBrown32a76272005-06-21 17:17:14 -07001449 if (bmc == NULL)
1450 goto unlock;
1451 /* locked */
NeilBrown32a76272005-06-21 17:17:14 -07001452 if (RESYNC(*bmc)) {
1453 *bmc &= ~RESYNC_MASK;
1454
1455 if (!NEEDED(*bmc) && aborted)
1456 *bmc |= NEEDED_MASK;
1457 else {
NeilBrown2585f3e2011-09-21 15:37:46 +10001458 if (*bmc <= 2) {
NeilBrown40cffcc2012-05-22 13:55:24 +10001459 bitmap_set_pending(&bitmap->counts, offset);
NeilBrown2585f3e2011-09-21 15:37:46 +10001460 bitmap->allclean = 0;
1461 }
NeilBrown32a76272005-06-21 17:17:14 -07001462 }
1463 }
1464 unlock:
NeilBrown40cffcc2012-05-22 13:55:24 +10001465 spin_unlock_irqrestore(&bitmap->counts.lock, flags);
NeilBrown32a76272005-06-21 17:17:14 -07001466}
NeilBrownac2f40b2010-06-01 19:37:31 +10001467EXPORT_SYMBOL(bitmap_end_sync);
NeilBrown32a76272005-06-21 17:17:14 -07001468
1469void bitmap_close_sync(struct bitmap *bitmap)
1470{
1471 /* Sync has finished, and any bitmap chunks that weren't synced
1472 * properly have been aborted. It remains to us to clear the
1473 * RESYNC bit wherever it is still on
1474 */
1475 sector_t sector = 0;
NeilBrown57dab0b2010-10-19 10:03:39 +11001476 sector_t blocks;
NeilBrownb47490c2008-02-06 01:39:50 -08001477 if (!bitmap)
1478 return;
NeilBrown32a76272005-06-21 17:17:14 -07001479 while (sector < bitmap->mddev->resync_max_sectors) {
1480 bitmap_end_sync(bitmap, sector, &blocks, 0);
NeilBrownb47490c2008-02-06 01:39:50 -08001481 sector += blocks;
NeilBrown32a76272005-06-21 17:17:14 -07001482 }
1483}
NeilBrownac2f40b2010-06-01 19:37:31 +10001484EXPORT_SYMBOL(bitmap_close_sync);
NeilBrown32a76272005-06-21 17:17:14 -07001485
NeilBrownb47490c2008-02-06 01:39:50 -08001486void bitmap_cond_end_sync(struct bitmap *bitmap, sector_t sector)
1487{
1488 sector_t s = 0;
NeilBrown57dab0b2010-10-19 10:03:39 +11001489 sector_t blocks;
NeilBrownb47490c2008-02-06 01:39:50 -08001490
1491 if (!bitmap)
1492 return;
1493 if (sector == 0) {
1494 bitmap->last_end_sync = jiffies;
1495 return;
1496 }
1497 if (time_before(jiffies, (bitmap->last_end_sync
NeilBrown1b04be92009-12-14 12:49:53 +11001498 + bitmap->mddev->bitmap_info.daemon_sleep)))
NeilBrownb47490c2008-02-06 01:39:50 -08001499 return;
1500 wait_event(bitmap->mddev->recovery_wait,
1501 atomic_read(&bitmap->mddev->recovery_active) == 0);
1502
NeilBrown75d3da42011-01-14 09:14:34 +11001503 bitmap->mddev->curr_resync_completed = sector;
NeilBrown070dc6d2010-08-30 17:33:34 +10001504 set_bit(MD_CHANGE_CLEAN, &bitmap->mddev->flags);
NeilBrown40cffcc2012-05-22 13:55:24 +10001505 sector &= ~((1ULL << bitmap->counts.chunkshift) - 1);
NeilBrownb47490c2008-02-06 01:39:50 -08001506 s = 0;
1507 while (s < sector && s < bitmap->mddev->resync_max_sectors) {
1508 bitmap_end_sync(bitmap, s, &blocks, 0);
1509 s += blocks;
1510 }
1511 bitmap->last_end_sync = jiffies;
NeilBrownacb180b2009-04-14 16:28:34 +10001512 sysfs_notify(&bitmap->mddev->kobj, NULL, "sync_completed");
NeilBrownb47490c2008-02-06 01:39:50 -08001513}
NeilBrownac2f40b2010-06-01 19:37:31 +10001514EXPORT_SYMBOL(bitmap_cond_end_sync);
NeilBrownb47490c2008-02-06 01:39:50 -08001515
NeilBrown6a079972005-09-09 16:23:44 -07001516static void bitmap_set_memory_bits(struct bitmap *bitmap, sector_t offset, int needed)
NeilBrown32a76272005-06-21 17:17:14 -07001517{
1518 /* For each chunk covered by any of these sectors, set the
NeilBrownef99bf42012-05-22 13:55:08 +10001519 * counter to 2 and possibly set resync_needed. They should all
NeilBrown32a76272005-06-21 17:17:14 -07001520 * be 0 at this point
1521 */
NeilBrown193f1c92005-08-04 12:53:33 -07001522
NeilBrown57dab0b2010-10-19 10:03:39 +11001523 sector_t secs;
NeilBrown193f1c92005-08-04 12:53:33 -07001524 bitmap_counter_t *bmc;
NeilBrown40cffcc2012-05-22 13:55:24 +10001525 spin_lock_irq(&bitmap->counts.lock);
1526 bmc = bitmap_get_counter(&bitmap->counts, offset, &secs, 1);
NeilBrown193f1c92005-08-04 12:53:33 -07001527 if (!bmc) {
NeilBrown40cffcc2012-05-22 13:55:24 +10001528 spin_unlock_irq(&bitmap->counts.lock);
NeilBrown193f1c92005-08-04 12:53:33 -07001529 return;
NeilBrown32a76272005-06-21 17:17:14 -07001530 }
NeilBrownac2f40b2010-06-01 19:37:31 +10001531 if (!*bmc) {
NeilBrown915c4202011-12-23 10:17:51 +11001532 *bmc = 2 | (needed ? NEEDED_MASK : 0);
NeilBrown40cffcc2012-05-22 13:55:24 +10001533 bitmap_count_page(&bitmap->counts, offset, 1);
1534 bitmap_set_pending(&bitmap->counts, offset);
NeilBrown2585f3e2011-09-21 15:37:46 +10001535 bitmap->allclean = 0;
NeilBrown193f1c92005-08-04 12:53:33 -07001536 }
NeilBrown40cffcc2012-05-22 13:55:24 +10001537 spin_unlock_irq(&bitmap->counts.lock);
NeilBrown32a76272005-06-21 17:17:14 -07001538}
1539
Paul Clements9b1d1da2006-10-03 01:15:49 -07001540/* dirty the memory and file bits for bitmap chunks "s" to "e" */
1541void bitmap_dirty_bits(struct bitmap *bitmap, unsigned long s, unsigned long e)
1542{
1543 unsigned long chunk;
1544
1545 for (chunk = s; chunk <= e; chunk++) {
NeilBrown40cffcc2012-05-22 13:55:24 +10001546 sector_t sec = (sector_t)chunk << bitmap->counts.chunkshift;
Paul Clements9b1d1da2006-10-03 01:15:49 -07001547 bitmap_set_memory_bits(bitmap, sec, 1);
1548 bitmap_file_set_bit(bitmap, sec);
NeilBrownffa23322009-12-14 12:49:56 +11001549 if (sec < bitmap->mddev->recovery_cp)
1550 /* We are asserting that the array is dirty,
1551 * so move the recovery_cp address back so
1552 * that it is obvious that it is dirty
1553 */
1554 bitmap->mddev->recovery_cp = sec;
Paul Clements9b1d1da2006-10-03 01:15:49 -07001555 }
1556}
1557
NeilBrown32a76272005-06-21 17:17:14 -07001558/*
NeilBrown6b8b3e82005-08-04 12:53:35 -07001559 * flush out any pending updates
1560 */
NeilBrownfd01b882011-10-11 16:47:53 +11001561void bitmap_flush(struct mddev *mddev)
NeilBrown6b8b3e82005-08-04 12:53:35 -07001562{
1563 struct bitmap *bitmap = mddev->bitmap;
NeilBrown42a04b52009-12-14 12:49:53 +11001564 long sleep;
NeilBrown6b8b3e82005-08-04 12:53:35 -07001565
1566 if (!bitmap) /* there was no bitmap */
1567 return;
1568
1569 /* run the daemon_work three time to ensure everything is flushed
1570 * that can be
1571 */
NeilBrown1b04be92009-12-14 12:49:53 +11001572 sleep = mddev->bitmap_info.daemon_sleep * 2;
NeilBrown42a04b52009-12-14 12:49:53 +11001573 bitmap->daemon_lastrun -= sleep;
NeilBrownaa5cbd12009-12-14 12:49:46 +11001574 bitmap_daemon_work(mddev);
NeilBrown42a04b52009-12-14 12:49:53 +11001575 bitmap->daemon_lastrun -= sleep;
NeilBrownaa5cbd12009-12-14 12:49:46 +11001576 bitmap_daemon_work(mddev);
NeilBrown42a04b52009-12-14 12:49:53 +11001577 bitmap->daemon_lastrun -= sleep;
NeilBrownaa5cbd12009-12-14 12:49:46 +11001578 bitmap_daemon_work(mddev);
NeilBrown6b8b3e82005-08-04 12:53:35 -07001579 bitmap_update_sb(bitmap);
1580}
1581
1582/*
NeilBrown32a76272005-06-21 17:17:14 -07001583 * free memory that was allocated
1584 */
NeilBrown3178b0d2005-09-09 16:23:50 -07001585static void bitmap_free(struct bitmap *bitmap)
NeilBrown32a76272005-06-21 17:17:14 -07001586{
1587 unsigned long k, pages;
1588 struct bitmap_page *bp;
NeilBrown32a76272005-06-21 17:17:14 -07001589
1590 if (!bitmap) /* there was no bitmap */
1591 return;
1592
NeilBrownfae7d322012-05-22 13:55:21 +10001593 /* Shouldn't be needed - but just in case.... */
1594 wait_event(bitmap->write_wait,
1595 atomic_read(&bitmap->pending_writes) == 0);
1596
1597 /* release the bitmap file */
1598 bitmap_file_unmap(&bitmap->storage);
NeilBrown32a76272005-06-21 17:17:14 -07001599
NeilBrown40cffcc2012-05-22 13:55:24 +10001600 bp = bitmap->counts.bp;
1601 pages = bitmap->counts.pages;
NeilBrown32a76272005-06-21 17:17:14 -07001602
1603 /* free all allocated memory */
1604
NeilBrown32a76272005-06-21 17:17:14 -07001605 if (bp) /* deallocate the page memory */
1606 for (k = 0; k < pages; k++)
1607 if (bp[k].map && !bp[k].hijacked)
1608 kfree(bp[k].map);
1609 kfree(bp);
1610 kfree(bitmap);
1611}
NeilBrownaa5cbd12009-12-14 12:49:46 +11001612
NeilBrownfd01b882011-10-11 16:47:53 +11001613void bitmap_destroy(struct mddev *mddev)
NeilBrown3178b0d2005-09-09 16:23:50 -07001614{
1615 struct bitmap *bitmap = mddev->bitmap;
1616
1617 if (!bitmap) /* there was no bitmap */
1618 return;
1619
NeilBrownc3d97142009-12-14 12:49:52 +11001620 mutex_lock(&mddev->bitmap_info.mutex);
NeilBrown3178b0d2005-09-09 16:23:50 -07001621 mddev->bitmap = NULL; /* disconnect from the md device */
NeilBrownc3d97142009-12-14 12:49:52 +11001622 mutex_unlock(&mddev->bitmap_info.mutex);
NeilBrownb15c2e52006-01-06 00:20:16 -08001623 if (mddev->thread)
1624 mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
NeilBrown3178b0d2005-09-09 16:23:50 -07001625
NeilBrownece5cff2009-12-14 12:49:56 +11001626 if (bitmap->sysfs_can_clear)
1627 sysfs_put(bitmap->sysfs_can_clear);
1628
NeilBrown3178b0d2005-09-09 16:23:50 -07001629 bitmap_free(bitmap);
1630}
NeilBrown32a76272005-06-21 17:17:14 -07001631
1632/*
1633 * initialize the bitmap structure
1634 * if this returns an error, bitmap_destroy must be called to do clean up
1635 */
NeilBrownfd01b882011-10-11 16:47:53 +11001636int bitmap_create(struct mddev *mddev)
NeilBrown32a76272005-06-21 17:17:14 -07001637{
1638 struct bitmap *bitmap;
NeilBrown1f593902009-04-20 11:50:24 +10001639 sector_t blocks = mddev->resync_max_sectors;
NeilBrownc3d97142009-12-14 12:49:52 +11001640 struct file *file = mddev->bitmap_info.file;
NeilBrown32a76272005-06-21 17:17:14 -07001641 int err;
NeilBrown5ff5aff2010-06-01 19:37:32 +10001642 struct sysfs_dirent *bm = NULL;
NeilBrown32a76272005-06-21 17:17:14 -07001643
Alexey Dobriyan5f6e3c832006-10-11 01:22:26 -07001644 BUILD_BUG_ON(sizeof(bitmap_super_t) != 256);
NeilBrown32a76272005-06-21 17:17:14 -07001645
NeilBrownc3d97142009-12-14 12:49:52 +11001646 BUG_ON(file && mddev->bitmap_info.offset);
NeilBrowna654b9d82005-06-21 17:17:27 -07001647
NeilBrown9ffae0c2006-01-06 00:20:32 -08001648 bitmap = kzalloc(sizeof(*bitmap), GFP_KERNEL);
NeilBrown32a76272005-06-21 17:17:14 -07001649 if (!bitmap)
1650 return -ENOMEM;
1651
NeilBrown40cffcc2012-05-22 13:55:24 +10001652 spin_lock_init(&bitmap->counts.lock);
NeilBrownce25c312006-06-26 00:27:49 -07001653 atomic_set(&bitmap->pending_writes, 0);
1654 init_waitqueue_head(&bitmap->write_wait);
Neil Brownda6e1a32007-02-08 14:20:37 -08001655 init_waitqueue_head(&bitmap->overflow_wait);
NeilBrowne5551902010-03-31 11:21:44 +11001656 init_waitqueue_head(&bitmap->behind_wait);
NeilBrownce25c312006-06-26 00:27:49 -07001657
NeilBrown32a76272005-06-21 17:17:14 -07001658 bitmap->mddev = mddev;
NeilBrown32a76272005-06-21 17:17:14 -07001659
NeilBrown5ff5aff2010-06-01 19:37:32 +10001660 if (mddev->kobj.sd)
1661 bm = sysfs_get_dirent(mddev->kobj.sd, NULL, "bitmap");
NeilBrownece5cff2009-12-14 12:49:56 +11001662 if (bm) {
Eric W. Biederman3ff195b2010-03-30 11:31:26 -07001663 bitmap->sysfs_can_clear = sysfs_get_dirent(bm, NULL, "can_clear");
NeilBrownece5cff2009-12-14 12:49:56 +11001664 sysfs_put(bm);
1665 } else
1666 bitmap->sysfs_can_clear = NULL;
1667
NeilBrown1ec885c2012-05-22 13:55:10 +10001668 bitmap->storage.file = file;
NeilBrownce25c312006-06-26 00:27:49 -07001669 if (file) {
1670 get_file(file);
NeilBrownae8fa282009-10-16 15:56:01 +11001671 /* As future accesses to this file will use bmap,
1672 * and bypass the page cache, we must sync the file
1673 * first.
1674 */
Christoph Hellwig8018ab02010-03-22 17:32:25 +01001675 vfs_fsync(file, 1);
NeilBrownce25c312006-06-26 00:27:49 -07001676 }
NeilBrown42a04b52009-12-14 12:49:53 +11001677 /* read superblock from bitmap file (this sets mddev->bitmap_info.chunksize) */
Jonathan Brassow9c810752011-06-08 17:59:30 -05001678 if (!mddev->bitmap_info.external) {
1679 /*
1680 * If 'MD_ARRAY_FIRST_USE' is set, then device-mapper is
1681 * instructing us to create a new on-disk bitmap instance.
1682 */
1683 if (test_and_clear_bit(MD_ARRAY_FIRST_USE, &mddev->flags))
1684 err = bitmap_new_disk_sb(bitmap);
1685 else
1686 err = bitmap_read_sb(bitmap);
1687 } else {
NeilBrownece5cff2009-12-14 12:49:56 +11001688 err = 0;
1689 if (mddev->bitmap_info.chunksize == 0 ||
1690 mddev->bitmap_info.daemon_sleep == 0)
1691 /* chunksize and time_base need to be
1692 * set first. */
1693 err = -EINVAL;
1694 }
NeilBrown32a76272005-06-21 17:17:14 -07001695 if (err)
NeilBrown3178b0d2005-09-09 16:23:50 -07001696 goto error;
NeilBrown32a76272005-06-21 17:17:14 -07001697
NeilBrown624ce4f2009-12-14 12:49:56 +11001698 bitmap->daemon_lastrun = jiffies;
NeilBrownd60b4792012-05-22 13:55:25 +10001699 err = bitmap_resize(bitmap, blocks, mddev->bitmap_info.chunksize, 1);
1700 if (err)
NeilBrown3178b0d2005-09-09 16:23:50 -07001701 goto error;
NeilBrown32a76272005-06-21 17:17:14 -07001702
NeilBrown69e51b42010-06-01 19:37:35 +10001703 printk(KERN_INFO "created bitmap (%lu pages) for device %s\n",
NeilBrownd60b4792012-05-22 13:55:25 +10001704 bitmap->counts.pages, bmname(bitmap));
NeilBrown69e51b42010-06-01 19:37:35 +10001705
1706 mddev->bitmap = bitmap;
NeilBrownb405fe92012-05-22 13:55:15 +10001707 return test_bit(BITMAP_WRITE_ERROR, &bitmap->flags) ? -EIO : 0;
NeilBrown69e51b42010-06-01 19:37:35 +10001708
1709 error:
1710 bitmap_free(bitmap);
1711 return err;
1712}
1713
NeilBrownfd01b882011-10-11 16:47:53 +11001714int bitmap_load(struct mddev *mddev)
NeilBrown69e51b42010-06-01 19:37:35 +10001715{
1716 int err = 0;
Jonathan Brassow3520fa42011-07-27 11:00:37 +10001717 sector_t start = 0;
NeilBrown69e51b42010-06-01 19:37:35 +10001718 sector_t sector = 0;
1719 struct bitmap *bitmap = mddev->bitmap;
1720
1721 if (!bitmap)
1722 goto out;
1723
1724 /* Clear out old bitmap info first: Either there is none, or we
1725 * are resuming after someone else has possibly changed things,
1726 * so we should forget old cached info.
1727 * All chunks should be clean, but some might need_sync.
1728 */
1729 while (sector < mddev->resync_max_sectors) {
NeilBrown57dab0b2010-10-19 10:03:39 +11001730 sector_t blocks;
NeilBrown69e51b42010-06-01 19:37:35 +10001731 bitmap_start_sync(bitmap, sector, &blocks, 0);
1732 sector += blocks;
1733 }
1734 bitmap_close_sync(bitmap);
1735
Jonathan Brassow3520fa42011-07-27 11:00:37 +10001736 if (mddev->degraded == 0
1737 || bitmap->events_cleared == mddev->events)
1738 /* no need to keep dirty bits to optimise a
1739 * re-add of a missing device */
1740 start = mddev->recovery_cp;
NeilBrown69e51b42010-06-01 19:37:35 +10001741
NeilBrownafbaa902012-04-12 16:05:06 +10001742 mutex_lock(&mddev->bitmap_info.mutex);
Jonathan Brassow3520fa42011-07-27 11:00:37 +10001743 err = bitmap_init_from_disk(bitmap, start);
NeilBrownafbaa902012-04-12 16:05:06 +10001744 mutex_unlock(&mddev->bitmap_info.mutex);
Jonathan Brassow3520fa42011-07-27 11:00:37 +10001745
NeilBrown32a76272005-06-21 17:17:14 -07001746 if (err)
NeilBrown69e51b42010-06-01 19:37:35 +10001747 goto out;
NeilBrownb405fe92012-05-22 13:55:15 +10001748 clear_bit(BITMAP_STALE, &bitmap->flags);
NeilBrownef99bf42012-05-22 13:55:08 +10001749
1750 /* Kick recovery in case any bits were set */
1751 set_bit(MD_RECOVERY_NEEDED, &bitmap->mddev->recovery);
NeilBrown3178b0d2005-09-09 16:23:50 -07001752
NeilBrown1b04be92009-12-14 12:49:53 +11001753 mddev->thread->timeout = mddev->bitmap_info.daemon_sleep;
NeilBrown9cd30fd2009-12-14 12:49:54 +11001754 md_wakeup_thread(mddev->thread);
NeilBrownb15c2e52006-01-06 00:20:16 -08001755
NeilBrown4ad13662007-07-17 04:06:13 -07001756 bitmap_update_sb(bitmap);
1757
NeilBrownb405fe92012-05-22 13:55:15 +10001758 if (test_bit(BITMAP_WRITE_ERROR, &bitmap->flags))
NeilBrown69e51b42010-06-01 19:37:35 +10001759 err = -EIO;
1760out:
NeilBrown3178b0d2005-09-09 16:23:50 -07001761 return err;
NeilBrown32a76272005-06-21 17:17:14 -07001762}
NeilBrown69e51b42010-06-01 19:37:35 +10001763EXPORT_SYMBOL_GPL(bitmap_load);
NeilBrown32a76272005-06-21 17:17:14 -07001764
NeilBrown57148962012-03-19 12:46:40 +11001765void bitmap_status(struct seq_file *seq, struct bitmap *bitmap)
1766{
1767 unsigned long chunk_kb;
NeilBrown40cffcc2012-05-22 13:55:24 +10001768 struct bitmap_counts *counts;
NeilBrown57148962012-03-19 12:46:40 +11001769
1770 if (!bitmap)
1771 return;
1772
NeilBrown40cffcc2012-05-22 13:55:24 +10001773 counts = &bitmap->counts;
1774
NeilBrown57148962012-03-19 12:46:40 +11001775 chunk_kb = bitmap->mddev->bitmap_info.chunksize >> 10;
1776 seq_printf(seq, "bitmap: %lu/%lu pages [%luKB], "
1777 "%lu%s chunk",
NeilBrown40cffcc2012-05-22 13:55:24 +10001778 counts->pages - counts->missing_pages,
1779 counts->pages,
1780 (counts->pages - counts->missing_pages)
NeilBrown57148962012-03-19 12:46:40 +11001781 << (PAGE_SHIFT - 10),
1782 chunk_kb ? chunk_kb : bitmap->mddev->bitmap_info.chunksize,
1783 chunk_kb ? "KB" : "B");
NeilBrown1ec885c2012-05-22 13:55:10 +10001784 if (bitmap->storage.file) {
NeilBrown57148962012-03-19 12:46:40 +11001785 seq_printf(seq, ", file: ");
NeilBrown1ec885c2012-05-22 13:55:10 +10001786 seq_path(seq, &bitmap->storage.file->f_path, " \t\n");
NeilBrown57148962012-03-19 12:46:40 +11001787 }
1788
1789 seq_printf(seq, "\n");
NeilBrown57148962012-03-19 12:46:40 +11001790}
1791
NeilBrownd60b4792012-05-22 13:55:25 +10001792int bitmap_resize(struct bitmap *bitmap, sector_t blocks,
1793 int chunksize, int init)
1794{
1795 /* If chunk_size is 0, choose an appropriate chunk size.
1796 * Then possibly allocate new storage space.
1797 * Then quiesce, copy bits, replace bitmap, and re-start
1798 *
1799 * This function is called both to set up the initial bitmap
1800 * and to resize the bitmap while the array is active.
1801 * If this happens as a result of the array being resized,
1802 * chunksize will be zero, and we need to choose a suitable
1803 * chunksize, otherwise we use what we are given.
1804 */
1805 struct bitmap_storage store;
1806 struct bitmap_counts old_counts;
1807 unsigned long chunks;
1808 sector_t block;
1809 sector_t old_blocks, new_blocks;
1810 int chunkshift;
1811 int ret = 0;
1812 long pages;
1813 struct bitmap_page *new_bp;
1814
1815 if (chunksize == 0) {
1816 /* If there is enough space, leave the chunk size unchanged,
1817 * else increase by factor of two until there is enough space.
1818 */
1819 long bytes;
1820 long space = bitmap->mddev->bitmap_info.space;
1821
1822 if (space == 0) {
1823 /* We don't know how much space there is, so limit
1824 * to current size - in sectors.
1825 */
1826 bytes = DIV_ROUND_UP(bitmap->counts.chunks, 8);
1827 if (!bitmap->mddev->bitmap_info.external)
1828 bytes += sizeof(bitmap_super_t);
1829 space = DIV_ROUND_UP(bytes, 512);
1830 bitmap->mddev->bitmap_info.space = space;
1831 }
1832 chunkshift = bitmap->counts.chunkshift;
1833 chunkshift--;
1834 do {
1835 /* 'chunkshift' is shift from block size to chunk size */
1836 chunkshift++;
1837 chunks = DIV_ROUND_UP_SECTOR_T(blocks, 1 << chunkshift);
1838 bytes = DIV_ROUND_UP(chunks, 8);
1839 if (!bitmap->mddev->bitmap_info.external)
1840 bytes += sizeof(bitmap_super_t);
1841 } while (bytes > (space << 9));
1842 } else
1843 chunkshift = ffz(~chunksize) - BITMAP_BLOCK_SHIFT;
1844
1845 chunks = DIV_ROUND_UP_SECTOR_T(blocks, 1 << chunkshift);
1846 memset(&store, 0, sizeof(store));
1847 if (bitmap->mddev->bitmap_info.offset || bitmap->mddev->bitmap_info.file)
1848 ret = bitmap_storage_alloc(&store, chunks,
1849 !bitmap->mddev->bitmap_info.external);
1850 if (ret)
1851 goto err;
1852
1853 pages = DIV_ROUND_UP(chunks, PAGE_COUNTER_RATIO);
1854
1855 new_bp = kzalloc(pages * sizeof(*new_bp), GFP_KERNEL);
1856 ret = -ENOMEM;
1857 if (!new_bp) {
1858 bitmap_file_unmap(&store);
1859 goto err;
1860 }
1861
1862 if (!init)
1863 bitmap->mddev->pers->quiesce(bitmap->mddev, 1);
1864
1865 store.file = bitmap->storage.file;
1866 bitmap->storage.file = NULL;
1867
1868 if (store.sb_page && bitmap->storage.sb_page)
1869 memcpy(page_address(store.sb_page),
1870 page_address(bitmap->storage.sb_page),
1871 sizeof(bitmap_super_t));
1872 bitmap_file_unmap(&bitmap->storage);
1873 bitmap->storage = store;
1874
1875 old_counts = bitmap->counts;
1876 bitmap->counts.bp = new_bp;
1877 bitmap->counts.pages = pages;
1878 bitmap->counts.missing_pages = pages;
1879 bitmap->counts.chunkshift = chunkshift;
1880 bitmap->counts.chunks = chunks;
1881 bitmap->mddev->bitmap_info.chunksize = 1 << (chunkshift +
1882 BITMAP_BLOCK_SHIFT);
1883
1884 blocks = min(old_counts.chunks << old_counts.chunkshift,
1885 chunks << chunkshift);
1886
1887 spin_lock_irq(&bitmap->counts.lock);
1888 for (block = 0; block < blocks; ) {
1889 bitmap_counter_t *bmc_old, *bmc_new;
1890 int set;
1891
1892 bmc_old = bitmap_get_counter(&old_counts, block,
1893 &old_blocks, 0);
1894 set = bmc_old && NEEDED(*bmc_old);
1895
1896 if (set) {
1897 bmc_new = bitmap_get_counter(&bitmap->counts, block,
1898 &new_blocks, 1);
1899 if (*bmc_new == 0) {
1900 /* need to set on-disk bits too. */
1901 sector_t end = block + new_blocks;
1902 sector_t start = block >> chunkshift;
1903 start <<= chunkshift;
1904 while (start < end) {
1905 bitmap_file_set_bit(bitmap, block);
1906 start += 1 << chunkshift;
1907 }
1908 *bmc_new = 2;
1909 bitmap_count_page(&bitmap->counts,
1910 block, 1);
1911 bitmap_set_pending(&bitmap->counts,
1912 block);
1913 }
1914 *bmc_new |= NEEDED_MASK;
1915 if (new_blocks < old_blocks)
1916 old_blocks = new_blocks;
1917 }
1918 block += old_blocks;
1919 }
1920
1921 if (!init) {
1922 int i;
1923 while (block < (chunks << chunkshift)) {
1924 bitmap_counter_t *bmc;
1925 bmc = bitmap_get_counter(&bitmap->counts, block,
1926 &new_blocks, 1);
1927 if (bmc) {
1928 /* new space. It needs to be resynced, so
1929 * we set NEEDED_MASK.
1930 */
1931 if (*bmc == 0) {
1932 *bmc = NEEDED_MASK | 2;
1933 bitmap_count_page(&bitmap->counts,
1934 block, 1);
1935 bitmap_set_pending(&bitmap->counts,
1936 block);
1937 }
1938 }
1939 block += new_blocks;
1940 }
1941 for (i = 0; i < bitmap->storage.file_pages; i++)
1942 set_page_attr(bitmap, i, BITMAP_PAGE_DIRTY);
1943 }
1944 spin_unlock_irq(&bitmap->counts.lock);
1945
1946 if (!init) {
1947 bitmap_unplug(bitmap);
1948 bitmap->mddev->pers->quiesce(bitmap->mddev, 0);
1949 }
1950 ret = 0;
1951err:
1952 return ret;
1953}
1954EXPORT_SYMBOL_GPL(bitmap_resize);
1955
NeilBrown43a70502009-12-14 12:49:55 +11001956static ssize_t
NeilBrownfd01b882011-10-11 16:47:53 +11001957location_show(struct mddev *mddev, char *page)
NeilBrown43a70502009-12-14 12:49:55 +11001958{
1959 ssize_t len;
NeilBrownac2f40b2010-06-01 19:37:31 +10001960 if (mddev->bitmap_info.file)
NeilBrown43a70502009-12-14 12:49:55 +11001961 len = sprintf(page, "file");
NeilBrownac2f40b2010-06-01 19:37:31 +10001962 else if (mddev->bitmap_info.offset)
NeilBrown43a70502009-12-14 12:49:55 +11001963 len = sprintf(page, "%+lld", (long long)mddev->bitmap_info.offset);
NeilBrownac2f40b2010-06-01 19:37:31 +10001964 else
NeilBrown43a70502009-12-14 12:49:55 +11001965 len = sprintf(page, "none");
1966 len += sprintf(page+len, "\n");
1967 return len;
1968}
1969
1970static ssize_t
NeilBrownfd01b882011-10-11 16:47:53 +11001971location_store(struct mddev *mddev, const char *buf, size_t len)
NeilBrown43a70502009-12-14 12:49:55 +11001972{
1973
1974 if (mddev->pers) {
1975 if (!mddev->pers->quiesce)
1976 return -EBUSY;
1977 if (mddev->recovery || mddev->sync_thread)
1978 return -EBUSY;
1979 }
1980
1981 if (mddev->bitmap || mddev->bitmap_info.file ||
1982 mddev->bitmap_info.offset) {
1983 /* bitmap already configured. Only option is to clear it */
1984 if (strncmp(buf, "none", 4) != 0)
1985 return -EBUSY;
1986 if (mddev->pers) {
1987 mddev->pers->quiesce(mddev, 1);
1988 bitmap_destroy(mddev);
1989 mddev->pers->quiesce(mddev, 0);
1990 }
1991 mddev->bitmap_info.offset = 0;
1992 if (mddev->bitmap_info.file) {
1993 struct file *f = mddev->bitmap_info.file;
1994 mddev->bitmap_info.file = NULL;
1995 restore_bitmap_write_access(f);
1996 fput(f);
1997 }
1998 } else {
1999 /* No bitmap, OK to set a location */
2000 long long offset;
2001 if (strncmp(buf, "none", 4) == 0)
2002 /* nothing to be done */;
2003 else if (strncmp(buf, "file:", 5) == 0) {
2004 /* Not supported yet */
2005 return -EINVAL;
2006 } else {
2007 int rv;
2008 if (buf[0] == '+')
2009 rv = strict_strtoll(buf+1, 10, &offset);
2010 else
2011 rv = strict_strtoll(buf, 10, &offset);
2012 if (rv)
2013 return rv;
2014 if (offset == 0)
2015 return -EINVAL;
NeilBrownece5cff2009-12-14 12:49:56 +11002016 if (mddev->bitmap_info.external == 0 &&
2017 mddev->major_version == 0 &&
NeilBrown43a70502009-12-14 12:49:55 +11002018 offset != mddev->bitmap_info.default_offset)
2019 return -EINVAL;
2020 mddev->bitmap_info.offset = offset;
2021 if (mddev->pers) {
2022 mddev->pers->quiesce(mddev, 1);
2023 rv = bitmap_create(mddev);
NeilBrown4474ca42012-03-19 12:46:37 +11002024 if (!rv)
2025 rv = bitmap_load(mddev);
NeilBrown43a70502009-12-14 12:49:55 +11002026 if (rv) {
2027 bitmap_destroy(mddev);
2028 mddev->bitmap_info.offset = 0;
2029 }
2030 mddev->pers->quiesce(mddev, 0);
2031 if (rv)
2032 return rv;
2033 }
2034 }
2035 }
2036 if (!mddev->external) {
2037 /* Ensure new bitmap info is stored in
2038 * metadata promptly.
2039 */
2040 set_bit(MD_CHANGE_DEVS, &mddev->flags);
2041 md_wakeup_thread(mddev->thread);
2042 }
2043 return len;
2044}
2045
2046static struct md_sysfs_entry bitmap_location =
2047__ATTR(location, S_IRUGO|S_IWUSR, location_show, location_store);
2048
NeilBrown6409bb02012-05-22 13:55:07 +10002049/* 'bitmap/space' is the space available at 'location' for the
2050 * bitmap. This allows the kernel to know when it is safe to
2051 * resize the bitmap to match a resized array.
2052 */
2053static ssize_t
2054space_show(struct mddev *mddev, char *page)
2055{
2056 return sprintf(page, "%lu\n", mddev->bitmap_info.space);
2057}
2058
2059static ssize_t
2060space_store(struct mddev *mddev, const char *buf, size_t len)
2061{
2062 unsigned long sectors;
2063 int rv;
2064
2065 rv = kstrtoul(buf, 10, &sectors);
2066 if (rv)
2067 return rv;
2068
2069 if (sectors == 0)
2070 return -EINVAL;
2071
2072 if (mddev->bitmap &&
NeilBrown9b1215c2012-05-22 13:55:11 +10002073 sectors < (mddev->bitmap->storage.bytes + 511) >> 9)
NeilBrown6409bb02012-05-22 13:55:07 +10002074 return -EFBIG; /* Bitmap is too big for this small space */
2075
2076 /* could make sure it isn't too big, but that isn't really
2077 * needed - user-space should be careful.
2078 */
2079 mddev->bitmap_info.space = sectors;
2080 return len;
2081}
2082
2083static struct md_sysfs_entry bitmap_space =
2084__ATTR(space, S_IRUGO|S_IWUSR, space_show, space_store);
2085
NeilBrown43a70502009-12-14 12:49:55 +11002086static ssize_t
NeilBrownfd01b882011-10-11 16:47:53 +11002087timeout_show(struct mddev *mddev, char *page)
NeilBrown43a70502009-12-14 12:49:55 +11002088{
2089 ssize_t len;
2090 unsigned long secs = mddev->bitmap_info.daemon_sleep / HZ;
2091 unsigned long jifs = mddev->bitmap_info.daemon_sleep % HZ;
NeilBrownac2f40b2010-06-01 19:37:31 +10002092
NeilBrown43a70502009-12-14 12:49:55 +11002093 len = sprintf(page, "%lu", secs);
2094 if (jifs)
2095 len += sprintf(page+len, ".%03u", jiffies_to_msecs(jifs));
2096 len += sprintf(page+len, "\n");
2097 return len;
2098}
2099
2100static ssize_t
NeilBrownfd01b882011-10-11 16:47:53 +11002101timeout_store(struct mddev *mddev, const char *buf, size_t len)
NeilBrown43a70502009-12-14 12:49:55 +11002102{
2103 /* timeout can be set at any time */
2104 unsigned long timeout;
2105 int rv = strict_strtoul_scaled(buf, &timeout, 4);
2106 if (rv)
2107 return rv;
2108
2109 /* just to make sure we don't overflow... */
2110 if (timeout >= LONG_MAX / HZ)
2111 return -EINVAL;
2112
2113 timeout = timeout * HZ / 10000;
2114
2115 if (timeout >= MAX_SCHEDULE_TIMEOUT)
2116 timeout = MAX_SCHEDULE_TIMEOUT-1;
2117 if (timeout < 1)
2118 timeout = 1;
2119 mddev->bitmap_info.daemon_sleep = timeout;
2120 if (mddev->thread) {
2121 /* if thread->timeout is MAX_SCHEDULE_TIMEOUT, then
2122 * the bitmap is all clean and we don't need to
2123 * adjust the timeout right now
2124 */
2125 if (mddev->thread->timeout < MAX_SCHEDULE_TIMEOUT) {
2126 mddev->thread->timeout = timeout;
2127 md_wakeup_thread(mddev->thread);
2128 }
2129 }
2130 return len;
2131}
2132
2133static struct md_sysfs_entry bitmap_timeout =
2134__ATTR(time_base, S_IRUGO|S_IWUSR, timeout_show, timeout_store);
2135
2136static ssize_t
NeilBrownfd01b882011-10-11 16:47:53 +11002137backlog_show(struct mddev *mddev, char *page)
NeilBrown43a70502009-12-14 12:49:55 +11002138{
2139 return sprintf(page, "%lu\n", mddev->bitmap_info.max_write_behind);
2140}
2141
2142static ssize_t
NeilBrownfd01b882011-10-11 16:47:53 +11002143backlog_store(struct mddev *mddev, const char *buf, size_t len)
NeilBrown43a70502009-12-14 12:49:55 +11002144{
2145 unsigned long backlog;
2146 int rv = strict_strtoul(buf, 10, &backlog);
2147 if (rv)
2148 return rv;
2149 if (backlog > COUNTER_MAX)
2150 return -EINVAL;
2151 mddev->bitmap_info.max_write_behind = backlog;
2152 return len;
2153}
2154
2155static struct md_sysfs_entry bitmap_backlog =
2156__ATTR(backlog, S_IRUGO|S_IWUSR, backlog_show, backlog_store);
2157
2158static ssize_t
NeilBrownfd01b882011-10-11 16:47:53 +11002159chunksize_show(struct mddev *mddev, char *page)
NeilBrown43a70502009-12-14 12:49:55 +11002160{
2161 return sprintf(page, "%lu\n", mddev->bitmap_info.chunksize);
2162}
2163
2164static ssize_t
NeilBrownfd01b882011-10-11 16:47:53 +11002165chunksize_store(struct mddev *mddev, const char *buf, size_t len)
NeilBrown43a70502009-12-14 12:49:55 +11002166{
2167 /* Can only be changed when no bitmap is active */
2168 int rv;
2169 unsigned long csize;
2170 if (mddev->bitmap)
2171 return -EBUSY;
2172 rv = strict_strtoul(buf, 10, &csize);
2173 if (rv)
2174 return rv;
2175 if (csize < 512 ||
2176 !is_power_of_2(csize))
2177 return -EINVAL;
2178 mddev->bitmap_info.chunksize = csize;
2179 return len;
2180}
2181
2182static struct md_sysfs_entry bitmap_chunksize =
2183__ATTR(chunksize, S_IRUGO|S_IWUSR, chunksize_show, chunksize_store);
2184
NeilBrownfd01b882011-10-11 16:47:53 +11002185static ssize_t metadata_show(struct mddev *mddev, char *page)
NeilBrownece5cff2009-12-14 12:49:56 +11002186{
2187 return sprintf(page, "%s\n", (mddev->bitmap_info.external
2188 ? "external" : "internal"));
2189}
2190
NeilBrownfd01b882011-10-11 16:47:53 +11002191static ssize_t metadata_store(struct mddev *mddev, const char *buf, size_t len)
NeilBrownece5cff2009-12-14 12:49:56 +11002192{
2193 if (mddev->bitmap ||
2194 mddev->bitmap_info.file ||
2195 mddev->bitmap_info.offset)
2196 return -EBUSY;
2197 if (strncmp(buf, "external", 8) == 0)
2198 mddev->bitmap_info.external = 1;
2199 else if (strncmp(buf, "internal", 8) == 0)
2200 mddev->bitmap_info.external = 0;
2201 else
2202 return -EINVAL;
2203 return len;
2204}
2205
2206static struct md_sysfs_entry bitmap_metadata =
2207__ATTR(metadata, S_IRUGO|S_IWUSR, metadata_show, metadata_store);
2208
NeilBrownfd01b882011-10-11 16:47:53 +11002209static ssize_t can_clear_show(struct mddev *mddev, char *page)
NeilBrownece5cff2009-12-14 12:49:56 +11002210{
2211 int len;
2212 if (mddev->bitmap)
2213 len = sprintf(page, "%s\n", (mddev->bitmap->need_sync ?
2214 "false" : "true"));
2215 else
2216 len = sprintf(page, "\n");
2217 return len;
2218}
2219
NeilBrownfd01b882011-10-11 16:47:53 +11002220static ssize_t can_clear_store(struct mddev *mddev, const char *buf, size_t len)
NeilBrownece5cff2009-12-14 12:49:56 +11002221{
2222 if (mddev->bitmap == NULL)
2223 return -ENOENT;
2224 if (strncmp(buf, "false", 5) == 0)
2225 mddev->bitmap->need_sync = 1;
2226 else if (strncmp(buf, "true", 4) == 0) {
2227 if (mddev->degraded)
2228 return -EBUSY;
2229 mddev->bitmap->need_sync = 0;
2230 } else
2231 return -EINVAL;
2232 return len;
2233}
2234
2235static struct md_sysfs_entry bitmap_can_clear =
2236__ATTR(can_clear, S_IRUGO|S_IWUSR, can_clear_show, can_clear_store);
2237
Paul Clements696fcd52010-03-08 16:02:37 +11002238static ssize_t
NeilBrownfd01b882011-10-11 16:47:53 +11002239behind_writes_used_show(struct mddev *mddev, char *page)
Paul Clements696fcd52010-03-08 16:02:37 +11002240{
2241 if (mddev->bitmap == NULL)
2242 return sprintf(page, "0\n");
2243 return sprintf(page, "%lu\n",
2244 mddev->bitmap->behind_writes_used);
2245}
2246
2247static ssize_t
NeilBrownfd01b882011-10-11 16:47:53 +11002248behind_writes_used_reset(struct mddev *mddev, const char *buf, size_t len)
Paul Clements696fcd52010-03-08 16:02:37 +11002249{
2250 if (mddev->bitmap)
2251 mddev->bitmap->behind_writes_used = 0;
2252 return len;
2253}
2254
2255static struct md_sysfs_entry max_backlog_used =
2256__ATTR(max_backlog_used, S_IRUGO | S_IWUSR,
2257 behind_writes_used_show, behind_writes_used_reset);
2258
NeilBrown43a70502009-12-14 12:49:55 +11002259static struct attribute *md_bitmap_attrs[] = {
2260 &bitmap_location.attr,
NeilBrown6409bb02012-05-22 13:55:07 +10002261 &bitmap_space.attr,
NeilBrown43a70502009-12-14 12:49:55 +11002262 &bitmap_timeout.attr,
2263 &bitmap_backlog.attr,
2264 &bitmap_chunksize.attr,
NeilBrownece5cff2009-12-14 12:49:56 +11002265 &bitmap_metadata.attr,
2266 &bitmap_can_clear.attr,
Paul Clements696fcd52010-03-08 16:02:37 +11002267 &max_backlog_used.attr,
NeilBrown43a70502009-12-14 12:49:55 +11002268 NULL
2269};
2270struct attribute_group md_bitmap_group = {
2271 .name = "bitmap",
2272 .attrs = md_bitmap_attrs,
2273};
2274