aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2009-06-18 16:11:50 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2009-06-18 16:11:50 -0400
commit9729a6eb5878a3daa18395f2b5fb38bf9359a761 (patch)
tree7ea32e9793c659d4059c49e7da2f38da795e7679
parent5ae8606d5746bc84e19018fc3753cc1faf18843f (diff)
parent48606a9f2fc034f0b308d088c1f7ab6d407c462c (diff)
Merge branch 'for-linus' of git://neil.brown.name/md
* 'for-linus' of git://neil.brown.name/md: (39 commits) md/raid5: correctly update sync_completed when we reach max_resync md/raid5: add missing call to schedule() after prepare_to_wait() md/linear: use call_rcu to free obsolete 'conf' structures. md linear: Protecting mddev with rcu locks to avoid races md: Move check for bitmap presence to personality code. md: remove chunksize rounding from common code. md: raid0/linear: ensure device sizes are rounded to chunk size. md: move assignment of ->utime so that it never gets skipped. md: Push down reconstruction log message to personality code. md: merge reconfig and check_reshape methods. md: remove unnecessary arguments from ->reconfig method. md: raid5: check stripe cache is large enough in start_reshape md: raid0: chunk_sectors cleanups. md: fix some comments. md/raid5: Use is_power_of_2() in raid5_reconfig()/raid6_reconfig(). md: convert conf->chunk_size and conf->prev_chunk to sectors. md: Convert mddev->new_chunk to sectors. md: Make mddev->chunk_size sector-based. md: raid0 :Enables chunk size other than powers of 2. md: prepare for non-power-of-two chunk sizes ...
Diffstat
-rw-r--r--drivers/md/faulty.c21
-rw-r--r--drivers/md/linear.c218
-rw-r--r--drivers/md/linear.h12
-rw-r--r--drivers/md/md.c196
-rw-r--r--drivers/md/md.h14
-rw-r--r--drivers/md/multipath.c23
-rw-r--r--drivers/md/multipath.h6
-rw-r--r--drivers/md/raid0.c403
-rw-r--r--drivers/md/raid0.h10
-rw-r--r--drivers/md/raid1.c46
-rw-r--r--drivers/md/raid1.h6
-rw-r--r--drivers/md/raid10.c62
-rw-r--r--drivers/md/raid10.h6
-rw-r--r--drivers/md/raid5.c218
-rw-r--r--drivers/md/raid5.h8
-rw-r--r--include/linux/raid/md_p.h2
16 files changed, 588 insertions, 663 deletions
diff --git a/drivers/md/faulty.c b/drivers/md/faulty.c
index 8695809b24b0..87d88dbb667f 100644
--- a/drivers/md/faulty.c
+++ b/drivers/md/faulty.c
@@ -255,14 +255,14 @@ static void status(struct seq_file *seq, mddev_t *mddev)
255} 255}
256 256
257 257
258static int reconfig(mddev_t *mddev, int layout, int chunk_size) 258static int reshape(mddev_t *mddev)
259{ 259{
260 int mode = layout & ModeMask; 260 int mode = mddev->new_layout & ModeMask;
261 int count = layout >> ModeShift; 261 int count = mddev->new_layout >> ModeShift;
262 conf_t *conf = mddev->private; 262 conf_t *conf = mddev->private;
263 263
264 if (chunk_size != -1) 264 if (mddev->new_layout < 0)
265 return -EINVAL; 265 return 0;
266 266
267 /* new layout */ 267 /* new layout */
268 if (mode == ClearFaults) 268 if (mode == ClearFaults)
@@ -279,6 +279,7 @@ static int reconfig(mddev_t *mddev, int layout, int chunk_size)
279 atomic_set(&conf->counters[mode], count); 279 atomic_set(&conf->counters[mode], count);
280 } else 280 } else
281 return -EINVAL; 281 return -EINVAL;
282 mddev->new_layout = -1;
282 mddev->layout = -1; /* makes sure further changes come through */ 283 mddev->layout = -1; /* makes sure further changes come through */
283 return 0; 284 return 0;
284} 285}
@@ -298,8 +299,12 @@ static int run(mddev_t *mddev)
298{ 299{
299 mdk_rdev_t *rdev; 300 mdk_rdev_t *rdev;
300 int i; 301 int i;
302 conf_t *conf;
303
304 if (md_check_no_bitmap(mddev))
305 return -EINVAL;
301 306
302 conf_t *conf = kmalloc(sizeof(*conf), GFP_KERNEL); 307 conf = kmalloc(sizeof(*conf), GFP_KERNEL);
303 if (!conf) 308 if (!conf)
304 return -ENOMEM; 309 return -ENOMEM;
305 310
@@ -315,7 +320,7 @@ static int run(mddev_t *mddev)
315 md_set_array_sectors(mddev, faulty_size(mddev, 0, 0)); 320 md_set_array_sectors(mddev, faulty_size(mddev, 0, 0));
316 mddev->private = conf; 321 mddev->private = conf;
317 322
318 reconfig(mddev, mddev->layout, -1); 323 reshape(mddev);
319 324
320 return 0; 325 return 0;
321} 326}
@@ -338,7 +343,7 @@ static struct mdk_personality faulty_personality =
338 .run = run, 343 .run = run,
339 .stop = stop, 344 .stop = stop,
340 .status = status, 345 .status = status,
341 .reconfig = reconfig, 346 .check_reshape = reshape,
342 .size = faulty_size, 347 .size = faulty_size,
343}; 348};
344 349
diff --git a/drivers/md/linear.c b/drivers/md/linear.c
index 64f1f3e046e0..15c8b7b25a9b 100644
--- a/drivers/md/linear.c
+++ b/drivers/md/linear.c
@@ -27,19 +27,27 @@
27 */ 27 */
28static inline dev_info_t *which_dev(mddev_t *mddev, sector_t sector) 28static inline dev_info_t *which_dev(mddev_t *mddev, sector_t sector)
29{ 29{
30 dev_info_t *hash; 30 int lo, mid, hi;
31 linear_conf_t *conf = mddev_to_conf(mddev); 31 linear_conf_t *conf;
32 sector_t idx = sector >> conf->sector_shift; 32
33 lo = 0;
34 hi = mddev->raid_disks - 1;
35 conf = rcu_dereference(mddev->private);
33 36
34 /* 37 /*
35 * sector_div(a,b) returns the remainer and sets a to a/b 38 * Binary Search
36 */ 39 */
37 (void)sector_div(idx, conf->spacing);
38 hash = conf->hash_table[idx];
39 40
40 while (sector >= hash->num_sectors + hash->start_sector) 41 while (hi > lo) {
41 hash++; 42
42 return hash; 43 mid = (hi + lo) / 2;
44 if (sector < conf->disks[mid].end_sector)
45 hi = mid;
46 else
47 lo = mid + 1;
48 }
49
50 return conf->disks + lo;
43} 51}
44 52
45/** 53/**
@@ -59,8 +67,10 @@ static int linear_mergeable_bvec(struct request_queue *q,
59 unsigned long maxsectors, bio_sectors = bvm->bi_size >> 9; 67 unsigned long maxsectors, bio_sectors = bvm->bi_size >> 9;
60 sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev); 68 sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
61 69
70 rcu_read_lock();
62 dev0 = which_dev(mddev, sector); 71 dev0 = which_dev(mddev, sector);
63 maxsectors = dev0->num_sectors - (sector - dev0->start_sector); 72 maxsectors = dev0->end_sector - sector;
73 rcu_read_unlock();
64 74
65 if (maxsectors < bio_sectors) 75 if (maxsectors < bio_sectors)
66 maxsectors = 0; 76 maxsectors = 0;
@@ -79,46 +89,57 @@ static int linear_mergeable_bvec(struct request_queue *q,
79static void linear_unplug(struct request_queue *q) 89static void linear_unplug(struct request_queue *q)
80{ 90{
81 mddev_t *mddev = q->queuedata; 91 mddev_t *mddev = q->queuedata;
82 linear_conf_t *conf = mddev_to_conf(mddev); 92 linear_conf_t *conf;
83 int i; 93 int i;
84 94
95 rcu_read_lock();
96 conf = rcu_dereference(mddev->private);
97
85 for (i=0; i < mddev->raid_disks; i++) { 98 for (i=0; i < mddev->raid_disks; i++) {
86 struct request_queue *r_queue = bdev_get_queue(conf->disks[i].rdev->bdev); 99 struct request_queue *r_queue = bdev_get_queue(conf->disks[i].rdev->bdev);
87 blk_unplug(r_queue); 100 blk_unplug(r_queue);
88 } 101 }
102 rcu_read_unlock();
89} 103}
90 104
91static int linear_congested(void *data, int bits) 105static int linear_congested(void *data, int bits)
92{ 106{
93 mddev_t *mddev = data; 107 mddev_t *mddev = data;
94 linear_conf_t *conf = mddev_to_conf(mddev); 108 linear_conf_t *conf;
95 int i, ret = 0; 109 int i, ret = 0;
96 110
111 rcu_read_lock();
112 conf = rcu_dereference(mddev->private);
113
97 for (i = 0; i < mddev->raid_disks && !ret ; i++) { 114 for (i = 0; i < mddev->raid_disks && !ret ; i++) {
98 struct request_queue *q = bdev_get_queue(conf->disks[i].rdev->bdev); 115 struct request_queue *q = bdev_get_queue(conf->disks[i].rdev->bdev);
99 ret |= bdi_congested(&q->backing_dev_info, bits); 116 ret |= bdi_congested(&q->backing_dev_info, bits);
100 } 117 }
118
119 rcu_read_unlock();
101 return ret; 120 return ret;
102} 121}
103 122
104static sector_t linear_size(mddev_t *mddev, sector_t sectors, int raid_disks) 123static sector_t linear_size(mddev_t *mddev, sector_t sectors, int raid_disks)
105{ 124{
106 linear_conf_t *conf = mddev_to_conf(mddev); 125 linear_conf_t *conf;
126 sector_t array_sectors;
107 127
128 rcu_read_lock();
129 conf = rcu_dereference(mddev->private);
108 WARN_ONCE(sectors || raid_disks, 130 WARN_ONCE(sectors || raid_disks,
109 "%s does not support generic reshape\n", __func__); 131 "%s does not support generic reshape\n", __func__);
132 array_sectors = conf->array_sectors;
133 rcu_read_unlock();
110 134
111 return conf->array_sectors; 135 return array_sectors;
112} 136}
113 137
114static linear_conf_t *linear_conf(mddev_t *mddev, int raid_disks) 138static linear_conf_t *linear_conf(mddev_t *mddev, int raid_disks)
115{ 139{
116 linear_conf_t *conf; 140 linear_conf_t *conf;
117 dev_info_t **table;
118 mdk_rdev_t *rdev; 141 mdk_rdev_t *rdev;
119 int i, nb_zone, cnt; 142 int i, cnt;
120 sector_t min_sectors;
121 sector_t curr_sector;
122 143
123 conf = kzalloc (sizeof (*conf) + raid_disks*sizeof(dev_info_t), 144 conf = kzalloc (sizeof (*conf) + raid_disks*sizeof(dev_info_t),
124 GFP_KERNEL); 145 GFP_KERNEL);
@@ -131,6 +152,7 @@ static linear_conf_t *linear_conf(mddev_t *mddev, int raid_disks)
131 list_for_each_entry(rdev, &mddev->disks, same_set) { 152 list_for_each_entry(rdev, &mddev->disks, same_set) {
132 int j = rdev->raid_disk; 153 int j = rdev->raid_disk;
133 dev_info_t *disk = conf->disks + j; 154 dev_info_t *disk = conf->disks + j;
155 sector_t sectors;
134 156
135 if (j < 0 || j >= raid_disks || disk->rdev) { 157 if (j < 0 || j >= raid_disks || disk->rdev) {
136 printk("linear: disk numbering problem. Aborting!\n"); 158 printk("linear: disk numbering problem. Aborting!\n");
@@ -138,6 +160,11 @@ static linear_conf_t *linear_conf(mddev_t *mddev, int raid_disks)
138 } 160 }
139 161
140 disk->rdev = rdev; 162 disk->rdev = rdev;
163 if (mddev->chunk_sectors) {
164 sectors = rdev->sectors;
165 sector_div(sectors, mddev->chunk_sectors);
166 rdev->sectors = sectors * mddev->chunk_sectors;
167 }
141 168
142 blk_queue_stack_limits(mddev->queue, 169 blk_queue_stack_limits(mddev->queue,
143 rdev->bdev->bd_disk->queue); 170 rdev->bdev->bd_disk->queue);
@@ -149,102 +176,24 @@ static linear_conf_t *linear_conf(mddev_t *mddev, int raid_disks)
149 queue_max_sectors(mddev->queue) > (PAGE_SIZE>>9)) 176 queue_max_sectors(mddev->queue) > (PAGE_SIZE>>9))
150 blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9); 177 blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);
151 178
152 disk->num_sectors = rdev->sectors;
153 conf->array_sectors += rdev->sectors; 179 conf->array_sectors += rdev->sectors;
154
155 cnt++; 180 cnt++;
181
156 } 182 }
157 if (cnt != raid_disks) { 183 if (cnt != raid_disks) {
158 printk("linear: not enough drives present. Aborting!\n"); 184 printk("linear: not enough drives present. Aborting!\n");
159 goto out; 185 goto out;
160 } 186 }
161 187
162 min_sectors = conf->array_sectors;
163 sector_div(min_sectors, PAGE_SIZE/sizeof(struct dev_info *));
164 if (min_sectors == 0)
165 min_sectors = 1;
166
167 /* min_sectors is the minimum spacing that will fit the hash
168 * table in one PAGE. This may be much smaller than needed.
169 * We find the smallest non-terminal set of consecutive devices
170 * that is larger than min_sectors and use the size of that as
171 * the actual spacing
172 */
173 conf->spacing = conf->array_sectors;
174 for (i=0; i < cnt-1 ; i++) {
175 sector_t tmp = 0;
176 int j;
177 for (j = i; j < cnt - 1 && tmp < min_sectors; j++)
178 tmp += conf->disks[j].num_sectors;
179 if (tmp >= min_sectors && tmp < conf->spacing)
180 conf->spacing = tmp;
181 }
182
183 /* spacing may be too large for sector_div to work with,
184 * so we might need to pre-shift
185 */
186 conf->sector_shift = 0;
187 if (sizeof(sector_t) > sizeof(u32)) {
188 sector_t space = conf->spacing;
189 while (space > (sector_t)(~(u32)0)) {
190 space >>= 1;
191 conf->sector_shift++;
192 }
193 }
194 /* 188 /*
195 * This code was restructured to work around a gcc-2.95.3 internal 189 * Here we calculate the device offsets.
196 * compiler error. Alter it with care.
197 */ 190 */
198 { 191 conf->disks[0].end_sector = conf->disks[0].rdev->sectors;
199 sector_t sz;
200 unsigned round;
201 unsigned long base;
202
203 sz = conf->array_sectors >> conf->sector_shift;
204 sz += 1; /* force round-up */
205 base = conf->spacing >> conf->sector_shift;
206 round = sector_div(sz, base);
207 nb_zone = sz + (round ? 1 : 0);
208 }
209 BUG_ON(nb_zone > PAGE_SIZE / sizeof(struct dev_info *));
210
211 conf->hash_table = kmalloc (sizeof (struct dev_info *) * nb_zone,
212 GFP_KERNEL);
213 if (!conf->hash_table)
214 goto out;
215 192
216 /*
217 * Here we generate the linear hash table
218 * First calculate the device offsets.
219 */
220 conf->disks[0].start_sector = 0;
221 for (i = 1; i < raid_disks; i++) 193 for (i = 1; i < raid_disks; i++)
222 conf->disks[i].start_sector = 194 conf->disks[i].end_sector =
223 conf->disks[i-1].start_sector + 195 conf->disks[i-1].end_sector +
224 conf->disks[i-1].num_sectors; 196 conf->disks[i].rdev->sectors;
225
226 table = conf->hash_table;
227 i = 0;
228 for (curr_sector = 0;
229 curr_sector < conf->array_sectors;
230 curr_sector += conf->spacing) {
231
232 while (i < raid_disks-1 &&
233 curr_sector >= conf->disks[i+1].start_sector)
234 i++;
235
236 *table ++ = conf->disks + i;
237 }
238
239 if (conf->sector_shift) {
240 conf->spacing >>= conf->sector_shift;
241 /* round spacing up so that when we divide by it,
242 * we err on the side of "too-low", which is safest.
243 */
244 conf->spacing++;
245 }
246
247 BUG_ON(table - conf->hash_table > nb_zone);
248 197
249 return conf; 198 return conf;
250 199
@@ -257,6 +206,8 @@ static int linear_run (mddev_t *mddev)
257{ 206{
258 linear_conf_t *conf; 207 linear_conf_t *conf;
259 208
209 if (md_check_no_bitmap(mddev))
210 return -EINVAL;
260 mddev->queue->queue_lock = &mddev->queue->__queue_lock; 211 mddev->queue->queue_lock = &mddev->queue->__queue_lock;
261 conf = linear_conf(mddev, mddev->raid_disks); 212 conf = linear_conf(mddev, mddev->raid_disks);
262 213
@@ -272,6 +223,12 @@ static int linear_run (mddev_t *mddev)
272 return 0; 223 return 0;
273} 224}
274 225
226static void free_conf(struct rcu_head *head)
227{
228 linear_conf_t *conf = container_of(head, linear_conf_t, rcu);
229 kfree(conf);
230}
231
275static int linear_add(mddev_t *mddev, mdk_rdev_t *rdev) 232static int linear_add(mddev_t *mddev, mdk_rdev_t *rdev)
276{ 233{
277 /* Adding a drive to a linear array allows the array to grow. 234 /* Adding a drive to a linear array allows the array to grow.
@@ -282,7 +239,7 @@ static int linear_add(mddev_t *mddev, mdk_rdev_t *rdev)
282 * The current one is never freed until the array is stopped. 239 * The current one is never freed until the array is stopped.
283 * This avoids races. 240 * This avoids races.
284 */ 241 */
285 linear_conf_t *newconf; 242 linear_conf_t *newconf, *oldconf;
286 243
287 if (rdev->saved_raid_disk != mddev->raid_disks) 244 if (rdev->saved_raid_disk != mddev->raid_disks)
288 return -EINVAL; 245 return -EINVAL;
@@ -294,25 +251,29 @@ static int linear_add(mddev_t *mddev, mdk_rdev_t *rdev)
294 if (!newconf) 251 if (!newconf)
295 return -ENOMEM; 252 return -ENOMEM;
296 253
297 newconf->prev = mddev_to_conf(mddev); 254 oldconf = rcu_dereference(mddev->private);
298 mddev->private = newconf;
299 mddev->raid_disks++; 255 mddev->raid_disks++;
256 rcu_assign_pointer(mddev->private, newconf);
300 md_set_array_sectors(mddev, linear_size(mddev, 0, 0)); 257 md_set_array_sectors(mddev, linear_size(mddev, 0, 0));
301 set_capacity(mddev->gendisk, mddev->array_sectors); 258 set_capacity(mddev->gendisk, mddev->array_sectors);
259 call_rcu(&oldconf->rcu, free_conf);
302 return 0; 260 return 0;
303} 261}
304 262
305static int linear_stop (mddev_t *mddev) 263static int linear_stop (mddev_t *mddev)
306{ 264{
307 linear_conf_t *conf = mddev_to_conf(mddev); 265 linear_conf_t *conf = mddev->private;
308 266
267 /*
268 * We do not require rcu protection here since
269 * we hold reconfig_mutex for both linear_add and
270 * linear_stop, so they cannot race.
271 * We should make sure any old 'conf's are properly
272 * freed though.
273 */
274 rcu_barrier();
309 blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/ 275 blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
310 do { 276 kfree(conf);
311 linear_conf_t *t = conf->prev;
312 kfree(conf->hash_table);
313 kfree(conf);
314 conf = t;
315 } while (conf);
316 277
317 return 0; 278 return 0;
318} 279}
@@ -322,6 +283,7 @@ static int linear_make_request (struct request_queue *q, struct bio *bio)
322 const int rw = bio_data_dir(bio); 283 const int rw = bio_data_dir(bio);
323 mddev_t *mddev = q->queuedata; 284 mddev_t *mddev = q->queuedata;
324 dev_info_t *tmp_dev; 285 dev_info_t *tmp_dev;
286 sector_t start_sector;
325 int cpu; 287 int cpu;
326 288
327 if (unlikely(bio_barrier(bio))) { 289 if (unlikely(bio_barrier(bio))) {
@@ -335,33 +297,36 @@ static int linear_make_request (struct request_queue *q, struct bio *bio)
335 bio_sectors(bio)); 297 bio_sectors(bio));
336 part_stat_unlock(); 298 part_stat_unlock();
337 299
300 rcu_read_lock();
338 tmp_dev = which_dev(mddev, bio->bi_sector); 301 tmp_dev = which_dev(mddev, bio->bi_sector);
339 302 start_sector = tmp_dev->end_sector - tmp_dev->rdev->sectors;
340 if (unlikely(bio->bi_sector >= (tmp_dev->num_sectors + 303
341 tmp_dev->start_sector) 304
342 || (bio->bi_sector < 305 if (unlikely(bio->bi_sector >= (tmp_dev->end_sector)
343 tmp_dev->start_sector))) { 306 || (bio->bi_sector < start_sector))) {
344 char b[BDEVNAME_SIZE]; 307 char b[BDEVNAME_SIZE];
345 308
346 printk("linear_make_request: Sector %llu out of bounds on " 309 printk("linear_make_request: Sector %llu out of bounds on "
347 "dev %s: %llu sectors, offset %llu\n", 310 "dev %s: %llu sectors, offset %llu\n",
348 (unsigned long long)bio->bi_sector, 311 (unsigned long long)bio->bi_sector,
349 bdevname(tmp_dev->rdev->bdev, b), 312 bdevname(tmp_dev->rdev->bdev, b),
350 (unsigned long long)tmp_dev->num_sectors, 313 (unsigned long long)tmp_dev->rdev->sectors,
351 (unsigned long long)tmp_dev->start_sector); 314 (unsigned long long)start_sector);
315 rcu_read_unlock();
352 bio_io_error(bio); 316 bio_io_error(bio);
353 return 0; 317 return 0;
354 } 318 }
355 if (unlikely(bio->bi_sector + (bio->bi_size >> 9) > 319 if (unlikely(bio->bi_sector + (bio->bi_size >> 9) >
356 tmp_dev->start_sector + tmp_dev->num_sectors)) { 320 tmp_dev->end_sector)) {
357 /* This bio crosses a device boundary, so we have to 321 /* This bio crosses a device boundary, so we have to
358 * split it. 322 * split it.
359 */ 323 */
360 struct bio_pair *bp; 324 struct bio_pair *bp;
325 sector_t end_sector = tmp_dev->end_sector;
326
327 rcu_read_unlock();
361 328
362 bp = bio_split(bio, 329 bp = bio_split(bio, end_sector - bio->bi_sector);
363 tmp_dev->start_sector + tmp_dev->num_sectors
364 - bio->bi_sector);
365 330
366 if (linear_make_request(q, &bp->bio1)) 331 if (linear_make_request(q, &bp->bio1))
367 generic_make_request(&bp->bio1); 332 generic_make_request(&bp->bio1);
@@ -372,8 +337,9 @@ static int linear_make_request (struct request_queue *q, struct bio *bio)
372 } 337 }
373 338
374 bio->bi_bdev = tmp_dev->rdev->bdev; 339 bio->bi_bdev = tmp_dev->rdev->bdev;
375 bio->bi_sector = bio->bi_sector - tmp_dev->start_sector 340 bio->bi_sector = bio->bi_sector - start_sector
376 + tmp_dev->rdev->data_offset; 341 + tmp_dev->rdev->data_offset;
342 rcu_read_unlock();
377 343
378 return 1; 344 return 1;
379} 345}
@@ -381,7 +347,7 @@ static int linear_make_request (struct request_queue *q, struct bio *bio)
381static void linear_status (struct seq_file *seq, mddev_t *mddev) 347static void linear_status (struct seq_file *seq, mddev_t *mddev)
382{ 348{
383 349
384 seq_printf(seq, " %dk rounding", mddev->chunk_size/1024); 350 seq_printf(seq, " %dk rounding", mddev->chunk_sectors / 2);
385} 351}
386 352
387 353
diff --git a/drivers/md/linear.h b/drivers/md/linear.h
index bf8179587f95..0ce29b61605a 100644
--- a/drivers/md/linear.h
+++ b/drivers/md/linear.h
@@ -3,27 +3,19 @@
3 3
4struct dev_info { 4struct dev_info {
5 mdk_rdev_t *rdev; 5 mdk_rdev_t *rdev;
6 sector_t num_sectors; 6 sector_t end_sector;
7 sector_t start_sector;
8}; 7};
9 8
10typedef struct dev_info dev_info_t; 9typedef struct dev_info dev_info_t;
11 10
12struct linear_private_data 11struct linear_private_data
13{ 12{
14 struct linear_private_data *prev; /* earlier version */
15 dev_info_t **hash_table;
16 sector_t spacing;
17 sector_t array_sectors; 13 sector_t array_sectors;
18 int sector_shift; /* shift before dividing
19 * by spacing
20 */
21 dev_info_t disks[0]; 14 dev_info_t disks[0];
15 struct rcu_head rcu;
22}; 16};
23 17
24 18
25typedef struct linear_private_data linear_conf_t; 19typedef struct linear_private_data linear_conf_t;
26 20
27#define mddev_to_conf(mddev) ((linear_conf_t *) mddev->private)
28
29#endif 21#endif
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 20f6ac338349..09be637d52cb 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -440,15 +440,6 @@ static inline sector_t calc_dev_sboffset(struct block_device *bdev)
440 return MD_NEW_SIZE_SECTORS(num_sectors); 440 return MD_NEW_SIZE_SECTORS(num_sectors);
441} 441}
442 442
443static sector_t calc_num_sectors(mdk_rdev_t *rdev, unsigned chunk_size)
444{
445 sector_t num_sectors = rdev->sb_start;
446
447 if (chunk_size)
448 num_sectors &= ~((sector_t)chunk_size/512 - 1);
449 return num_sectors;
450}
451
452static int alloc_disk_sb(mdk_rdev_t * rdev) 443static int alloc_disk_sb(mdk_rdev_t * rdev)
453{ 444{
454 if (rdev->sb_page) 445 if (rdev->sb_page)
@@ -745,6 +736,24 @@ struct super_type {
745}; 736};
746 737
747/* 738/*
739 * Check that the given mddev has no bitmap.
740 *
741 * This function is called from the run method of all personalities that do not
742 * support bitmaps. It prints an error message and returns non-zero if mddev
743 * has a bitmap. Otherwise, it returns 0.
744 *
745 */
746int md_check_no_bitmap(mddev_t *mddev)
747{
748 if (!mddev->bitmap_file && !mddev->bitmap_offset)
749 return 0;
750 printk(KERN_ERR "%s: bitmaps are not supported for %s\n",
751 mdname(mddev), mddev->pers->name);
752 return 1;
753}
754EXPORT_SYMBOL(md_check_no_bitmap);
755
756/*
748 * load_super for 0.90.0 757 * load_super for 0.90.0
749 */ 758 */
750static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version) 759static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version)
@@ -797,17 +806,6 @@ static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version
797 rdev->data_offset = 0; 806 rdev->data_offset = 0;
798 rdev->sb_size = MD_SB_BYTES; 807 rdev->sb_size = MD_SB_BYTES;
799 808
800 if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
801 if (sb->level != 1 && sb->level != 4
802 && sb->level != 5 && sb->level != 6
803 && sb->level != 10) {
804 /* FIXME use a better test */
805 printk(KERN_WARNING
806 "md: bitmaps not supported for this level.\n");
807 goto abort;
808 }
809 }
810
811 if (sb->level == LEVEL_MULTIPATH) 809 if (sb->level == LEVEL_MULTIPATH)
812 rdev->desc_nr = -1; 810 rdev->desc_nr = -1;
813 else 811 else
@@ -836,7 +834,7 @@ static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version
836 else 834 else
837 ret = 0; 835 ret = 0;
838 } 836 }
839 rdev->sectors = calc_num_sectors(rdev, sb->chunk_size); 837 rdev->sectors = rdev->sb_start;
840 838
841 if (rdev->sectors < sb->size * 2 && sb->level > 1) 839 if (rdev->sectors < sb->size * 2 && sb->level > 1)
842 /* "this cannot possibly happen" ... */ 840 /* "this cannot possibly happen" ... */
@@ -866,7 +864,7 @@ static int super_90_validate(mddev_t *mddev, mdk_rdev_t *rdev)
866 mddev->minor_version = sb->minor_version; 864 mddev->minor_version = sb->minor_version;
867 mddev->patch_version = sb->patch_version; 865 mddev->patch_version = sb->patch_version;
868 mddev->external = 0; 866 mddev->external = 0;
869 mddev->chunk_size = sb->chunk_size; 867 mddev->chunk_sectors = sb->chunk_size >> 9;
870 mddev->ctime = sb->ctime; 868 mddev->ctime = sb->ctime;
871 mddev->utime = sb->utime; 869 mddev->utime = sb->utime;
872 mddev->level = sb->level; 870 mddev->level = sb->level;
@@ -883,13 +881,13 @@ static int super_90_validate(mddev_t *mddev, mdk_rdev_t *rdev)
883 mddev->delta_disks = sb->delta_disks; 881 mddev->delta_disks = sb->delta_disks;
884 mddev->new_level = sb->new_level; 882 mddev->new_level = sb->new_level;
885 mddev->new_layout = sb->new_layout; 883 mddev->new_layout = sb->new_layout;
886 mddev->new_chunk = sb->new_chunk; 884 mddev->new_chunk_sectors = sb->new_chunk >> 9;
887 } else { 885 } else {
888 mddev->reshape_position = MaxSector; 886 mddev->reshape_position = MaxSector;
889 mddev->delta_disks = 0; 887 mddev->delta_disks = 0;
890 mddev->new_level = mddev->level; 888 mddev->new_level = mddev->level;
891 mddev->new_layout = mddev->layout; 889 mddev->new_layout = mddev->layout;
892 mddev->new_chunk = mddev->chunk_size; 890 mddev->new_chunk_sectors = mddev->chunk_sectors;
893 } 891 }
894 892
895 if (sb->state & (1<<MD_SB_CLEAN)) 893 if (sb->state & (1<<MD_SB_CLEAN))
@@ -1004,7 +1002,7 @@ static void super_90_sync(mddev_t *mddev, mdk_rdev_t *rdev)
1004 sb->new_level = mddev->new_level; 1002 sb->new_level = mddev->new_level;
1005 sb->delta_disks = mddev->delta_disks; 1003 sb->delta_disks = mddev->delta_disks;
1006 sb->new_layout = mddev->new_layout; 1004 sb->new_layout = mddev->new_layout;
1007 sb->new_chunk = mddev->new_chunk; 1005 sb->new_chunk = mddev->new_chunk_sectors << 9;
1008 } 1006 }
1009 mddev->minor_version = sb->minor_version; 1007 mddev->minor_version = sb->minor_version;
1010 if (mddev->in_sync) 1008 if (mddev->in_sync)
@@ -1018,7 +1016,7 @@ static void super_90_sync(mddev_t *mddev, mdk_rdev_t *rdev)
1018 sb->recovery_cp = 0; 1016 sb->recovery_cp = 0;
1019 1017
1020 sb->layout = mddev->layout; 1018 sb->layout = mddev->layout;
1021 sb->chunk_size = mddev->chunk_size; 1019 sb->chunk_size = mddev->chunk_sectors << 9;
1022 1020
1023 if (mddev->bitmap && mddev->bitmap_file == NULL) 1021 if (mddev->bitmap && mddev->bitmap_file == NULL)
1024 sb->state |= (1<<MD_SB_BITMAP_PRESENT); 1022 sb->state |= (1<<MD_SB_BITMAP_PRESENT);
@@ -1185,17 +1183,6 @@ static int super_1_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version)
1185 bdevname(rdev->bdev,b)); 1183 bdevname(rdev->bdev,b));
1186 return -EINVAL; 1184 return -EINVAL;
1187 } 1185 }
1188 if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET)) {
1189 if (sb->level != cpu_to_le32(1) &&
1190 sb->level != cpu_to_le32(4) &&
1191 sb->level != cpu_to_le32(5) &&
1192 sb->level != cpu_to_le32(6) &&
1193 sb->level != cpu_to_le32(10)) {
1194 printk(KERN_WARNING
1195 "md: bitmaps not supported for this level.\n");
1196 return -EINVAL;
1197 }
1198 }
1199 1186
1200 rdev->preferred_minor = 0xffff; 1187 rdev->preferred_minor = 0xffff;
1201 rdev->data_offset = le64_to_cpu(sb->data_offset); 1188 rdev->data_offset = le64_to_cpu(sb->data_offset);
@@ -1248,9 +1235,6 @@ static int super_1_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version)
1248 if (rdev->sectors < le64_to_cpu(sb->data_size)) 1235 if (rdev->sectors < le64_to_cpu(sb->data_size))
1249 return -EINVAL; 1236 return -EINVAL;
1250 rdev->sectors = le64_to_cpu(sb->data_size); 1237 rdev->sectors = le64_to_cpu(sb->data_size);
1251 if (le32_to_cpu(sb->chunksize))
1252 rdev->sectors &= ~((sector_t)le32_to_cpu(sb->chunksize) - 1);
1253
1254 if (le64_to_cpu(sb->size) > rdev->sectors) 1238 if (le64_to_cpu(sb->size) > rdev->sectors)
1255 return -EINVAL; 1239 return -EINVAL;
1256 return ret; 1240 return ret;
@@ -1271,7 +1255,7 @@ static int super_1_validate(mddev_t *mddev, mdk_rdev_t *rdev)
1271 mddev->major_version = 1; 1255 mddev->major_version = 1;
1272 mddev->patch_version = 0; 1256 mddev->patch_version = 0;
1273 mddev->external = 0; 1257 mddev->external = 0;
1274 mddev->chunk_size = le32_to_cpu(sb->chunksize) << 9; 1258 mddev->chunk_sectors = le32_to_cpu(sb->chunksize);
1275 mddev->ctime = le64_to_cpu(sb->ctime) & ((1ULL << 32)-1); 1259 mddev->ctime = le64_to_cpu(sb->ctime) & ((1ULL << 32)-1);
1276 mddev->utime = le64_to_cpu(sb->utime) & ((1ULL << 32)-1); 1260 mddev->utime = le64_to_cpu(sb->utime) & ((1ULL << 32)-1);
1277 mddev->level = le32_to_cpu(sb->level); 1261 mddev->level = le32_to_cpu(sb->level);
@@ -1297,13 +1281,13 @@ static int super_1_validate(mddev_t *mddev, mdk_rdev_t *rdev)
1297 mddev->delta_disks = le32_to_cpu(sb->delta_disks); 1281 mddev->delta_disks = le32_to_cpu(sb->delta_disks);
1298 mddev->new_level = le32_to_cpu(sb->new_level); 1282 mddev->new_level = le32_to_cpu(sb->new_level);
1299 mddev->new_layout = le32_to_cpu(sb->new_layout); 1283 mddev->new_layout = le32_to_cpu(sb->new_layout);
1300 mddev->new_chunk = le32_to_cpu(sb->new_chunk)<<9; 1284 mddev->new_chunk_sectors = le32_to_cpu(sb->new_chunk);
1301 } else { 1285 } else {
1302 mddev->reshape_position = MaxSector; 1286 mddev->reshape_position = MaxSector;
1303 mddev->delta_disks = 0; 1287 mddev->delta_disks = 0;
1304 mddev->new_level = mddev->level; 1288 mddev->new_level = mddev->level;
1305 mddev->new_layout = mddev->layout; 1289 mddev->new_layout = mddev->layout;
1306 mddev->new_chunk = mddev->chunk_size; 1290 mddev->new_chunk_sectors = mddev->chunk_sectors;
1307 } 1291 }
1308 1292
1309 } else if (mddev->pers == NULL) { 1293 } else if (mddev->pers == NULL) {
@@ -1375,7 +1359,7 @@ static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev)
1375 1359
1376 sb->raid_disks = cpu_to_le32(mddev->raid_disks); 1360 sb->raid_disks = cpu_to_le32(mddev->raid_disks);
1377 sb->size = cpu_to_le64(mddev->dev_sectors); 1361 sb->size = cpu_to_le64(mddev->dev_sectors);
1378 sb->chunksize = cpu_to_le32(mddev->chunk_size >> 9); 1362 sb->chunksize = cpu_to_le32(mddev->chunk_sectors);
1379 sb->level = cpu_to_le32(mddev->level); 1363 sb->level = cpu_to_le32(mddev->level);
1380 sb->layout = cpu_to_le32(mddev->layout); 1364 sb->layout = cpu_to_le32(mddev->layout);
1381 1365
@@ -1402,7 +1386,7 @@ static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev)
1402 sb->new_layout = cpu_to_le32(mddev->new_layout); 1386 sb->new_layout = cpu_to_le32(mddev->new_layout);
1403 sb->delta_disks = cpu_to_le32(mddev->delta_disks); 1387 sb->delta_disks = cpu_to_le32(mddev->delta_disks);
1404 sb->new_level = cpu_to_le32(mddev->new_level); 1388 sb->new_level = cpu_to_le32(mddev->new_level);
1405 sb->new_chunk = cpu_to_le32(mddev->new_chunk>>9); 1389 sb->new_chunk = cpu_to_le32(mddev->new_chunk_sectors);
1406 } 1390 }
1407 1391
1408 max_dev = 0; 1392 max_dev = 0;
@@ -1897,6 +1881,7 @@ static void md_update_sb(mddev_t * mddev, int force_change)
1897 int sync_req; 1881 int sync_req;
1898 int nospares = 0; 1882 int nospares = 0;
1899 1883
1884 mddev->utime = get_seconds();
1900 if (mddev->external) 1885 if (mddev->external)
1901 return; 1886 return;
1902repeat: 1887repeat:
@@ -1926,7 +1911,6 @@ repeat:
1926 nospares = 0; 1911 nospares = 0;
1927 1912
1928 sync_req = mddev->in_sync; 1913 sync_req = mddev->in_sync;
1929 mddev->utime = get_seconds();
1930 1914
1931 /* If this is just a dirty<->clean transition, and the array is clean 1915 /* If this is just a dirty<->clean transition, and the array is clean
1932 * and 'events' is odd, we can roll back to the previous clean state */ 1916 * and 'events' is odd, we can roll back to the previous clean state */
@@ -2597,15 +2581,6 @@ static void analyze_sbs(mddev_t * mddev)
2597 clear_bit(In_sync, &rdev->flags); 2581 clear_bit(In_sync, &rdev->flags);
2598 } 2582 }
2599 } 2583 }
2600
2601
2602
2603 if (mddev->recovery_cp != MaxSector &&
2604 mddev->level >= 1)
2605 printk(KERN_ERR "md: %s: raid array is not clean"
2606 " -- starting background reconstruction\n",
2607 mdname(mddev));
2608
2609} 2584}
2610 2585
2611static void md_safemode_timeout(unsigned long data); 2586static void md_safemode_timeout(unsigned long data);
@@ -2746,7 +2721,7 @@ level_store(mddev_t *mddev, const char *buf, size_t len)
2746 if (IS_ERR(priv)) { 2721 if (IS_ERR(priv)) {
2747 mddev->new_level = mddev->level; 2722 mddev->new_level = mddev->level;
2748 mddev->new_layout = mddev->layout; 2723 mddev->new_layout = mddev->layout;
2749 mddev->new_chunk = mddev->chunk_size; 2724 mddev->new_chunk_sectors = mddev->chunk_sectors;
2750 mddev->raid_disks -= mddev->delta_disks; 2725 mddev->raid_disks -= mddev->delta_disks;
2751 mddev->delta_disks = 0; 2726 mddev->delta_disks = 0;
2752 module_put(pers->owner); 2727 module_put(pers->owner);
@@ -2764,7 +2739,7 @@ level_store(mddev_t *mddev, const char *buf, size_t len)
2764 strlcpy(mddev->clevel, pers->name, sizeof(mddev->clevel)); 2739 strlcpy(mddev->clevel, pers->name, sizeof(mddev->clevel));
2765 mddev->level = mddev->new_level; 2740 mddev->level = mddev->new_level;
2766 mddev->layout = mddev->new_layout; 2741 mddev->layout = mddev->new_layout;
2767 mddev->chunk_size = mddev->new_chunk; 2742 mddev->chunk_sectors = mddev->new_chunk_sectors;
2768 mddev->delta_disks = 0; 2743 mddev->delta_disks = 0;
2769 pers->run(mddev); 2744 pers->run(mddev);
2770 mddev_resume(mddev); 2745 mddev_resume(mddev);
@@ -2800,11 +2775,14 @@ layout_store(mddev_t *mddev, const char *buf, size_t len)
2800 2775
2801 if (mddev->pers) { 2776 if (mddev->pers) {
2802 int err; 2777 int err;
2803 if (mddev->pers->reconfig == NULL) 2778 if (mddev->pers->check_reshape == NULL)
2804 return -EBUSY; 2779 return -EBUSY;
2805 err = mddev->pers->reconfig(mddev, n, -1); 2780 mddev->new_layout = n;
2806 if (err) 2781 err = mddev->pers->check_reshape(mddev);
2782 if (err) {
2783 mddev->new_layout = mddev->layout;
2807 return err; 2784 return err;
2785 }
2808 } else { 2786 } else {
2809 mddev->new_layout = n; 2787 mddev->new_layout = n;
2810 if (mddev->reshape_position == MaxSector) 2788 if (mddev->reshape_position == MaxSector)
@@ -2857,10 +2835,11 @@ static ssize_t
2857chunk_size_show(mddev_t *mddev, char *page) 2835chunk_size_show(mddev_t *mddev, char *page)
2858{ 2836{
2859 if (mddev->reshape_position != MaxSector && 2837 if (mddev->reshape_position != MaxSector &&
2860 mddev->chunk_size != mddev->new_chunk) 2838 mddev->chunk_sectors != mddev->new_chunk_sectors)
2861 return sprintf(page, "%d (%d)\n", mddev->new_chunk, 2839 return sprintf(page, "%d (%d)\n",
2862 mddev->chunk_size); 2840 mddev->new_chunk_sectors << 9,
2863 return sprintf(page, "%d\n", mddev->chunk_size); 2841 mddev->chunk_sectors << 9);
2842 return sprintf(page, "%d\n", mddev->chunk_sectors << 9);
2864} 2843}
2865 2844
2866static ssize_t 2845static ssize_t
@@ -2874,15 +2853,18 @@ chunk_size_store(mddev_t *mddev, const char *buf, size_t len)
2874 2853
2875 if (mddev->pers) { 2854 if (mddev->pers) {
2876 int err; 2855 int err;
2877 if (mddev->pers->reconfig == NULL) 2856 if (mddev->pers->check_reshape == NULL)
2878 return -EBUSY; 2857 return -EBUSY;
2879 err = mddev->pers->reconfig(mddev, -1, n); 2858 mddev->new_chunk_sectors = n >> 9;
2880 if (err) 2859 err = mddev->pers->check_reshape(mddev);
2860 if (err) {
2861 mddev->new_chunk_sectors = mddev->chunk_sectors;
2881 return err; 2862 return err;
2863 }
2882 } else { 2864 } else {
2883 mddev->new_chunk = n; 2865 mddev->new_chunk_sectors = n >> 9;
2884 if (mddev->reshape_position == MaxSector) 2866 if (mddev->reshape_position == MaxSector)
2885 mddev->chunk_size = n; 2867 mddev->chunk_sectors = n >> 9;
2886 } 2868 }
2887 return len; 2869 return len;
2888} 2870}
@@ -3527,8 +3509,9 @@ min_sync_store(mddev_t *mddev, const char *buf, size_t len)
3527 return -EBUSY; 3509 return -EBUSY;
3528 3510
3529 /* Must be a multiple of chunk_size */ 3511 /* Must be a multiple of chunk_size */
3530 if (mddev->chunk_size) { 3512 if (mddev->chunk_sectors) {
3531 if (min & (sector_t)((mddev->chunk_size>>9)-1)) 3513 sector_t temp = min;
3514 if (sector_div(temp, mddev->chunk_sectors))
3532 return -EINVAL; 3515 return -EINVAL;
3533 } 3516 }
3534 mddev->resync_min = min; 3517 mddev->resync_min = min;
@@ -3564,8 +3547,9 @@ max_sync_store(mddev_t *mddev, const char *buf, size_t len)
3564 return -EBUSY; 3547 return -EBUSY;
3565 3548
3566 /* Must be a multiple of chunk_size */ 3549 /* Must be a multiple of chunk_size */
3567 if (mddev->chunk_size) { 3550 if (mddev->chunk_sectors) {
3568 if (max & (sector_t)((mddev->chunk_size>>9)-1)) 3551 sector_t temp = max;
3552 if (sector_div(temp, mddev->chunk_sectors))
3569 return -EINVAL; 3553 return -EINVAL;
3570 } 3554 }
3571 mddev->resync_max = max; 3555 mddev->resync_max = max;
@@ -3656,7 +3640,7 @@ reshape_position_store(mddev_t *mddev, const char *buf, size_t len)
3656 mddev->delta_disks = 0; 3640 mddev->delta_disks = 0;
3657 mddev->new_level = mddev->level; 3641 mddev->new_level = mddev->level;
3658 mddev->new_layout = mddev->layout; 3642 mddev->new_layout = mddev->layout;
3659 mddev->new_chunk = mddev->chunk_size; 3643 mddev->new_chunk_sectors = mddev->chunk_sectors;
3660 return len; 3644 return len;
3661} 3645}
3662 3646
@@ -3976,11 +3960,9 @@ static int start_dirty_degraded;
3976static int do_md_run(mddev_t * mddev) 3960static int do_md_run(mddev_t * mddev)
3977{ 3961{
3978 int err; 3962 int err;
3979 int chunk_size;
3980 mdk_rdev_t *rdev; 3963 mdk_rdev_t *rdev;
3981 struct gendisk *disk; 3964 struct gendisk *disk;
3982 struct mdk_personality *pers; 3965 struct mdk_personality *pers;
3983 char b[BDEVNAME_SIZE];
3984 3966
3985 if (list_empty(&mddev->disks)) 3967 if (list_empty(&mddev->disks))
3986 /* cannot run an array with no devices.. */ 3968 /* cannot run an array with no devices.. */
@@ -3998,38 +3980,6 @@ static int do_md_run(mddev_t * mddev)
3998 analyze_sbs(mddev); 3980 analyze_sbs(mddev);
3999 } 3981 }
4000 3982
4001 chunk_size = mddev->chunk_size;
4002
4003 if (chunk_size) {
4004 if (chunk_size > MAX_CHUNK_SIZE) {
4005 printk(KERN_ERR "too big chunk_size: %d > %d\n",
4006 chunk_size, MAX_CHUNK_SIZE);
4007 return -EINVAL;
4008 }
4009 /*
4010 * chunk-size has to be a power of 2
4011 */
4012 if ( (1 << ffz(~chunk_size)) != chunk_size) {
4013 printk(KERN_ERR "chunk_size of %d not valid\n", chunk_size);
4014 return -EINVAL;
4015 }
4016
4017 /* devices must have minimum size of one chunk */
4018 list_for_each_entry(rdev, &mddev->disks, same_set) {
4019 if (test_bit(Faulty, &rdev->flags))
4020 continue;
4021 if (rdev->sectors < chunk_size / 512) {
4022 printk(KERN_WARNING
4023 "md: Dev %s smaller than chunk_size:"
4024 " %llu < %d\n",
4025 bdevname(rdev->bdev,b),
4026 (unsigned long long)rdev->sectors,
4027 chunk_size / 512);
4028 return -EINVAL;
4029 }
4030 }
4031 }
4032
4033 if (mddev->level != LEVEL_NONE) 3983 if (mddev->level != LEVEL_NONE)
4034 request_module("md-level-%d", mddev->level); 3984 request_module("md-level-%d", mddev->level);
4035 else if (mddev->clevel[0]) 3985 else if (mddev->clevel[0])
@@ -4405,7 +4355,7 @@ static int do_md_stop(mddev_t * mddev, int mode, int is_open)
4405 mddev->flags = 0; 4355 mddev->flags = 0;
4406 mddev->ro = 0; 4356 mddev->ro = 0;
4407 mddev->metadata_type[0] = 0; 4357 mddev->metadata_type[0] = 0;
4408 mddev->chunk_size = 0; 4358 mddev->chunk_sectors = 0;
4409 mddev->ctime = mddev->utime = 0; 4359 mddev->ctime = mddev->utime = 0;
4410 mddev->layout = 0; 4360 mddev->layout = 0;
4411 mddev->max_disks = 0; 4361 mddev->max_disks = 0;
@@ -4413,7 +4363,7 @@ static int do_md_stop(mddev_t * mddev, int mode, int is_open)
4413 mddev->delta_disks = 0; 4363 mddev->delta_disks = 0;
4414 mddev->new_level = LEVEL_NONE; 4364 mddev->new_level = LEVEL_NONE;
4415 mddev->new_layout = 0; 4365 mddev->new_layout = 0;
4416 mddev->new_chunk = 0; 4366 mddev->new_chunk_sectors = 0;
4417 mddev->curr_resync = 0; 4367 mddev->curr_resync = 0;
4418 mddev->resync_mismatches = 0; 4368 mddev->resync_mismatches = 0;
4419 mddev->suspend_lo = mddev->suspend_hi = 0; 4369 mddev->suspend_lo = mddev->suspend_hi = 0;
@@ -4618,7 +4568,7 @@ static int get_array_info(mddev_t * mddev, void __user * arg)
4618 info.spare_disks = spare; 4568 info.spare_disks = spare;
4619 4569
4620 info.layout = mddev->layout; 4570 info.layout = mddev->layout;
4621 info.chunk_size = mddev->chunk_size; 4571 info.chunk_size = mddev->chunk_sectors << 9;
4622 4572
4623 if (copy_to_user(arg, &info, sizeof(info))) 4573 if (copy_to_user(arg, &info, sizeof(info)))
4624 return -EFAULT; 4574 return -EFAULT;
@@ -4843,7 +4793,7 @@ static int add_new_disk(mddev_t * mddev, mdu_disk_info_t *info)
4843 rdev->sb_start = rdev->bdev->bd_inode->i_size / 512; 4793 rdev->sb_start = rdev->bdev->bd_inode->i_size / 512;
4844 } else 4794 } else
4845 rdev->sb_start = calc_dev_sboffset(rdev->bdev); 4795 rdev->sb_start = calc_dev_sboffset(rdev->bdev);
4846 rdev->sectors = calc_num_sectors(rdev, mddev->chunk_size); 4796 rdev->sectors = rdev->sb_start;
4847 4797
4848 err = bind_rdev_to_array(rdev, mddev); 4798 err = bind_rdev_to_array(rdev, mddev);
4849 if (err) { 4799 if (err) {
@@ -4913,7 +4863,7 @@ static int hot_add_disk(mddev_t * mddev, dev_t dev)
4913 else 4863 else
4914 rdev->sb_start = rdev->bdev->bd_inode->i_size / 512; 4864 rdev->sb_start = rdev->bdev->bd_inode->i_size / 512;
4915 4865
4916 rdev->sectors = calc_num_sectors(rdev, mddev->chunk_size); 4866 rdev->sectors = rdev->sb_start;
4917 4867
4918 if (test_bit(Faulty, &rdev->flags)) { 4868 if (test_bit(Faulty, &rdev->flags)) {
4919 printk(KERN_WARNING 4869 printk(KERN_WARNING
@@ -5062,7 +5012,7 @@ static int set_array_info(mddev_t * mddev, mdu_array_info_t *info)
5062 mddev->external = 0; 5012 mddev->external = 0;
5063 5013
5064 mddev->layout = info->layout; 5014 mddev->layout = info->layout;
5065 mddev->chunk_size = info->chunk_size; 5015 mddev->chunk_sectors = info->chunk_size >> 9;
5066 5016
5067 mddev->max_disks = MD_SB_DISKS; 5017 mddev->max_disks = MD_SB_DISKS;
5068 5018
@@ -5081,7 +5031,7 @@ static int set_array_info(mddev_t * mddev, mdu_array_info_t *info)
5081 get_random_bytes(mddev->uuid, 16); 5031 get_random_bytes(mddev->uuid, 16);
5082 5032
5083 mddev->new_level = mddev->level; 5033 mddev->new_level = mddev->level;
5084 mddev->new_chunk = mddev->chunk_size; 5034 mddev->new_chunk_sectors = mddev->chunk_sectors;
5085 mddev->new_layout = mddev->layout; 5035 mddev->new_layout = mddev->layout;
5086 mddev->delta_disks = 0; 5036 mddev->delta_disks = 0;
5087 5037
@@ -5191,7 +5141,7 @@ static int update_array_info(mddev_t *mddev, mdu_array_info_t *info)
5191 mddev->level != info->level || 5141 mddev->level != info->level ||
5192/* mddev->layout != info->layout || */ 5142/* mddev->layout != info->layout || */
5193 !mddev->persistent != info->not_persistent|| 5143 !mddev->persistent != info->not_persistent||
5194 mddev->chunk_size != info->chunk_size || 5144 mddev->chunk_sectors != info->chunk_size >> 9 ||
5195 /* ignore bottom 8 bits of state, and allow SB_BITMAP_PRESENT to change */ 5145 /* ignore bottom 8 bits of state, and allow SB_BITMAP_PRESENT to change */
5196 ((state^info->state) & 0xfffffe00) 5146 ((state^info->state) & 0xfffffe00)
5197 ) 5147 )
@@ -5215,10 +5165,15 @@ static int update_array_info(mddev_t *mddev, mdu_array_info_t *info)
5215 * we don't need to do anything at the md level, the 5165 * we don't need to do anything at the md level, the
5216 * personality will take care of it all. 5166 * personality will take care of it all.
5217 */ 5167 */
5218 if (mddev->pers->reconfig == NULL) 5168 if (mddev->pers->check_reshape == NULL)
5219 return -EINVAL; 5169 return -EINVAL;
5220 else 5170 else {
5221 return mddev->pers->reconfig(mddev, info->layout, -1); 5171 mddev->new_layout = info->layout;
5172 rv = mddev->pers->check_reshape(mddev);
5173 if (rv)
5174 mddev->new_layout = mddev->layout;
5175 return rv;
5176 }
5222 } 5177 }
5223 if (info->size >= 0 && mddev->dev_sectors / 2 != info->size) 5178 if (info->size >= 0 && mddev->dev_sectors / 2 != info->size)
5224 rv = update_size(mddev, (sector_t)info->size * 2); 5179 rv = update_size(mddev, (sector_t)info->size * 2);
@@ -6717,7 +6672,8 @@ void md_check_recovery(mddev_t *mddev)
6717 */ 6672 */
6718 6673
6719 if (mddev->reshape_position != MaxSector) { 6674 if (mddev->reshape_position != MaxSector) {
6720 if (mddev->pers->check_reshape(mddev) != 0) 6675 if (mddev->pers->check_reshape == NULL ||
6676 mddev->pers->check_reshape(mddev) != 0)
6721 /* Cannot proceed */ 6677 /* Cannot proceed */
6722 goto unlock; 6678 goto unlock;
6723 set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery); 6679 set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
diff --git a/drivers/md/md.h b/drivers/md/md.h
index 8227ab909d44..9430a110db93 100644
--- a/drivers/md/md.h
+++ b/drivers/md/md.h
@@ -30,13 +30,6 @@ typedef struct mddev_s mddev_t;
30typedef struct mdk_rdev_s mdk_rdev_t; 30typedef struct mdk_rdev_s mdk_rdev_t;
31 31
32/* 32/*
33 * options passed in raidrun:
34 */
35
36/* Currently this must fit in an 'int' */
37#define MAX_CHUNK_SIZE (1<<30)
38
39/*
40 * MD's 'extended' device 33 * MD's 'extended' device
41 */ 34 */
42struct mdk_rdev_s 35struct mdk_rdev_s
@@ -145,7 +138,7 @@ struct mddev_s
145 int external; /* metadata is 138 int external; /* metadata is
146 * managed externally */ 139 * managed externally */
147 char metadata_type[17]; /* externally set*/ 140 char metadata_type[17]; /* externally set*/
148 int chunk_size; 141 int chunk_sectors;
149 time_t ctime, utime; 142 time_t ctime, utime;
150 int level, layout; 143 int level, layout;
151 char clevel[16]; 144 char clevel[16];
@@ -166,7 +159,8 @@ struct mddev_s
166 * If reshape_position is MaxSector, then no reshape is happening (yet). 159 * If reshape_position is MaxSector, then no reshape is happening (yet).
167 */ 160 */
168 sector_t reshape_position; 161 sector_t reshape_position;
169 int delta_disks, new_level, new_layout, new_chunk; 162 int delta_disks, new_level, new_layout;
163 int new_chunk_sectors;
170 164
171 struct mdk_thread_s *thread; /* management thread */ 165 struct mdk_thread_s *thread; /* management thread */
172 struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */ 166 struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */
@@ -325,7 +319,6 @@ struct mdk_personality
325 int (*check_reshape) (mddev_t *mddev); 319 int (*check_reshape) (mddev_t *mddev);
326 int (*start_reshape) (mddev_t *mddev); 320 int (*start_reshape) (mddev_t *mddev);
327 void (*finish_reshape) (mddev_t *mddev); 321 void (*finish_reshape) (mddev_t *mddev);
328 int (*reconfig) (mddev_t *mddev, int layout, int chunk_size);
329 /* quiesce moves between quiescence states 322 /* quiesce moves between quiescence states
330 * 0 - fully active 323 * 0 - fully active
331 * 1 - no new requests allowed 324 * 1 - no new requests allowed
@@ -437,5 +430,6 @@ extern void md_new_event(mddev_t *mddev);
437extern int md_allow_write(mddev_t *mddev); 430extern int md_allow_write(mddev_t *mddev);
438extern void md_wait_for_blocked_rdev(mdk_rdev_t *rdev, mddev_t *mddev); 431extern void md_wait_for_blocked_rdev(mdk_rdev_t *rdev, mddev_t *mddev);
439extern void md_set_array_sectors(mddev_t *mddev, sector_t array_sectors); 432extern void md_set_array_sectors(mddev_t *mddev, sector_t array_sectors);
433extern int md_check_no_bitmap(mddev_t *mddev);
440 434
441#endif /* _MD_MD_H */ 435#endif /* _MD_MD_H */
diff --git a/drivers/md/multipath.c b/drivers/md/multipath.c
index 4ee31aa13c40..cbe368fa6598 100644
--- a/drivers/md/multipath.c
+++ b/drivers/md/multipath.c
@@ -58,7 +58,7 @@ static void multipath_reschedule_retry (struct multipath_bh *mp_bh)
58{ 58{
59 unsigned long flags; 59 unsigned long flags;
60 mddev_t *mddev = mp_bh->mddev; 60 mddev_t *mddev = mp_bh->mddev;
61 multipath_conf_t *conf = mddev_to_conf(mddev); 61 multipath_conf_t *conf = mddev->private;
62 62
63 spin_lock_irqsave(&conf->device_lock, flags); 63 spin_lock_irqsave(&conf->device_lock, flags);
64 list_add(&mp_bh->retry_list, &conf->retry_list); 64 list_add(&mp_bh->retry_list, &conf->retry_list);
@@ -75,7 +75,7 @@ static void multipath_reschedule_retry (struct multipath_bh *mp_bh)
75static void multipath_end_bh_io (struct multipath_bh *mp_bh, int err) 75static void multipath_end_bh_io (struct multipath_bh *mp_bh, int err)
76{ 76{
77 struct bio *bio = mp_bh->master_bio; 77 struct bio *bio = mp_bh->master_bio;
78 multipath_conf_t *conf = mddev_to_conf(mp_bh->mddev); 78 multipath_conf_t *conf = mp_bh->mddev->private;
79 79
80 bio_endio(bio, err); 80 bio_endio(bio, err);
81 mempool_free(mp_bh, conf->pool); 81 mempool_free(mp_bh, conf->pool);
@@ -85,7 +85,7 @@ static void multipath_end_request(struct bio *bio, int error)
85{ 85{
86 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); 86 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
87 struct multipath_bh * mp_bh = (struct multipath_bh *)(bio->bi_private); 87 struct multipath_bh * mp_bh = (struct multipath_bh *)(bio->bi_private);
88 multipath_conf_t *conf = mddev_to_conf(mp_bh->mddev); 88 multipath_conf_t *conf = mp_bh->mddev->private;
89 mdk_rdev_t *rdev = conf->multipaths[mp_bh->path].rdev; 89 mdk_rdev_t *rdev = conf->multipaths[mp_bh->path].rdev;
90 90
91 if (uptodate) 91 if (uptodate)
@@ -107,7 +107,7 @@ static void multipath_end_request(struct bio *bio, int error)
107 107
108static void unplug_slaves(mddev_t *mddev) 108static void unplug_slaves(mddev_t *mddev)
109{ 109{
110 multipath_conf_t *conf = mddev_to_conf(mddev); 110 multipath_conf_t *conf = mddev->private;
111 int i; 111 int i;
112 112
113 rcu_read_lock(); 113 rcu_read_lock();
@@ -138,7 +138,7 @@ static void multipath_unplug(struct request_queue *q)
138static int multipath_make_request (struct request_queue *q, struct bio * bio) 138static int multipath_make_request (struct request_queue *q, struct bio * bio)
139{ 139{
140 mddev_t *mddev = q->queuedata; 140 mddev_t *mddev = q->queuedata;
141 multipath_conf_t *conf = mddev_to_conf(mddev); 141 multipath_conf_t *conf = mddev->private;
142 struct multipath_bh * mp_bh; 142 struct multipath_bh * mp_bh;
143 struct multipath_info *multipath; 143 struct multipath_info *multipath;
144 const int rw = bio_data_dir(bio); 144 const int rw = bio_data_dir(bio);
@@ -180,7 +180,7 @@ static int multipath_make_request (struct request_queue *q, struct bio * bio)
180 180
181static void multipath_status (struct seq_file *seq, mddev_t *mddev) 181static void multipath_status (struct seq_file *seq, mddev_t *mddev)
182{ 182{
183 multipath_conf_t *conf = mddev_to_conf(mddev); 183 multipath_conf_t *conf = mddev->private;
184 int i; 184 int i;
185 185
186 seq_printf (seq, " [%d/%d] [", conf->raid_disks, 186 seq_printf (seq, " [%d/%d] [", conf->raid_disks,
@@ -195,7 +195,7 @@ static void multipath_status (struct seq_file *seq, mddev_t *mddev)
195static int multipath_congested(void *data, int bits) 195static int multipath_congested(void *data, int bits)
196{ 196{
197 mddev_t *mddev = data; 197 mddev_t *mddev = data;
198 multipath_conf_t *conf = mddev_to_conf(mddev); 198 multipath_conf_t *conf = mddev->private;
199 int i, ret = 0; 199 int i, ret = 0;
200 200
201 rcu_read_lock(); 201 rcu_read_lock();
@@ -220,7 +220,7 @@ static int multipath_congested(void *data, int bits)
220 */ 220 */
221static void multipath_error (mddev_t *mddev, mdk_rdev_t *rdev) 221static void multipath_error (mddev_t *mddev, mdk_rdev_t *rdev)
222{ 222{
223 multipath_conf_t *conf = mddev_to_conf(mddev); 223 multipath_conf_t *conf = mddev->private;
224 224
225 if (conf->working_disks <= 1) { 225 if (conf->working_disks <= 1) {
226 /* 226 /*
@@ -367,7 +367,7 @@ static void multipathd (mddev_t *mddev)
367 struct multipath_bh *mp_bh; 367 struct multipath_bh *mp_bh;
368 struct bio *bio; 368 struct bio *bio;
369 unsigned long flags; 369 unsigned long flags;
370 multipath_conf_t *conf = mddev_to_conf(mddev); 370 multipath_conf_t *conf = mddev->private;
371 struct list_head *head = &conf->retry_list; 371 struct list_head *head = &conf->retry_list;
372 372
373 md_check_recovery(mddev); 373 md_check_recovery(mddev);
@@ -421,6 +421,9 @@ static int multipath_run (mddev_t *mddev)
421 struct multipath_info *disk; 421 struct multipath_info *disk;
422 mdk_rdev_t *rdev; 422 mdk_rdev_t *rdev;
423 423
424 if (md_check_no_bitmap(mddev))
425 return -EINVAL;
426
424 if (mddev->level != LEVEL_MULTIPATH) { 427 if (mddev->level != LEVEL_MULTIPATH) {
425 printk("multipath: %s: raid level not set to multipath IO (%d)\n", 428 printk("multipath: %s: raid level not set to multipath IO (%d)\n",
426 mdname(mddev), mddev->level); 429 mdname(mddev), mddev->level);
@@ -531,7 +534,7 @@ out:
531 534
532static int multipath_stop (mddev_t *mddev) 535static int multipath_stop (mddev_t *mddev)
533{ 536{
534 multipath_conf_t *conf = mddev_to_conf(mddev); 537 multipath_conf_t *conf = mddev->private;
535 538
536 md_unregister_thread(mddev->thread); 539 md_unregister_thread(mddev->thread);
537 mddev->thread = NULL; 540 mddev->thread = NULL;
diff --git a/drivers/md/multipath.h b/drivers/md/multipath.h
index 6fa70b400cda..d1c2a8d78395 100644
--- a/drivers/md/multipath.h
+++ b/drivers/md/multipath.h
@@ -19,12 +19,6 @@ struct multipath_private_data {
19typedef struct multipath_private_data multipath_conf_t; 19typedef struct multipath_private_data multipath_conf_t;
20 20
21/* 21/*
22 * this is the only point in the RAID code where we violate
23 * C type safety. mddev->private is an 'opaque' pointer.
24 */
25#define mddev_to_conf(mddev) ((multipath_conf_t *) mddev->private)
26
27/*
28 * this is our 'private' 'collective' MULTIPATH buffer head. 22 * this is our 'private' 'collective' MULTIPATH buffer head.
29 * it contains information about what kind of IO operations were started 23 * it contains information about what kind of IO operations were started
30 * for this MULTIPATH operation, and about their status: 24 * for this MULTIPATH operation, and about their status:
diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c
index 925507e7d673..ab4a489d8695 100644
--- a/drivers/md/raid0.c
+++ b/drivers/md/raid0.c
@@ -26,8 +26,8 @@
26static void raid0_unplug(struct request_queue *q) 26static void raid0_unplug(struct request_queue *q)
27{ 27{
28 mddev_t *mddev = q->queuedata; 28 mddev_t *mddev = q->queuedata;
29 raid0_conf_t *conf = mddev_to_conf(mddev); 29 raid0_conf_t *conf = mddev->private;
30 mdk_rdev_t **devlist = conf->strip_zone[0].dev; 30 mdk_rdev_t **devlist = conf->devlist;
31 int i; 31 int i;
32 32
33 for (i=0; i<mddev->raid_disks; i++) { 33 for (i=0; i<mddev->raid_disks; i++) {
@@ -40,8 +40,8 @@ static void raid0_unplug(struct request_queue *q)
40static int raid0_congested(void *data, int bits) 40static int raid0_congested(void *data, int bits)
41{ 41{
42 mddev_t *mddev = data; 42 mddev_t *mddev = data;
43 raid0_conf_t *conf = mddev_to_conf(mddev); 43 raid0_conf_t *conf = mddev->private;
44 mdk_rdev_t **devlist = conf->strip_zone[0].dev; 44 mdk_rdev_t **devlist = conf->devlist;
45 int i, ret = 0; 45 int i, ret = 0;
46 46
47 for (i = 0; i < mddev->raid_disks && !ret ; i++) { 47 for (i = 0; i < mddev->raid_disks && !ret ; i++) {
@@ -52,27 +52,60 @@ static int raid0_congested(void *data, int bits)
52 return ret; 52 return ret;
53} 53}
54 54
55/*
56 * inform the user of the raid configuration
57*/
58static void dump_zones(mddev_t *mddev)
59{
60 int j, k, h;
61 sector_t zone_size = 0;
62 sector_t zone_start = 0;
63 char b[BDEVNAME_SIZE];
64 raid0_conf_t *conf = mddev->private;
65 printk(KERN_INFO "******* %s configuration *********\n",
66 mdname(mddev));
67 h = 0;
68 for (j = 0; j < conf->nr_strip_zones; j++) {
69 printk(KERN_INFO "zone%d=[", j);
70 for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
71 printk("%s/",
72 bdevname(conf->devlist[j*mddev->raid_disks
73 + k]->bdev, b));
74 printk("]\n");
75
76 zone_size = conf->strip_zone[j].zone_end - zone_start;
77 printk(KERN_INFO " zone offset=%llukb "
78 "device offset=%llukb size=%llukb\n",
79 (unsigned long long)zone_start>>1,
80 (unsigned long long)conf->strip_zone[j].dev_start>>1,
81 (unsigned long long)zone_size>>1);
82 zone_start = conf->strip_zone[j].zone_end;
83 }
84 printk(KERN_INFO "**********************************\n\n");
85}
55 86
56static int create_strip_zones (mddev_t *mddev) 87static int create_strip_zones(mddev_t *mddev)
57{ 88{
58 int i, c, j; 89 int i, c, j, err;
59 sector_t current_start, curr_zone_start; 90 sector_t curr_zone_end, sectors;
60 sector_t min_spacing; 91 mdk_rdev_t *smallest, *rdev1, *rdev2, *rdev, **dev;
61 raid0_conf_t *conf = mddev_to_conf(mddev);
62 mdk_rdev_t *smallest, *rdev1, *rdev2, *rdev;
63 struct strip_zone *zone; 92 struct strip_zone *zone;
64 int cnt; 93 int cnt;
65 char b[BDEVNAME_SIZE]; 94 char b[BDEVNAME_SIZE];
66 95 raid0_conf_t *conf = kzalloc(sizeof(*conf), GFP_KERNEL);
67 /* 96
68 * The number of 'same size groups' 97 if (!conf)
69 */ 98 return -ENOMEM;
70 conf->nr_strip_zones = 0;
71
72 list_for_each_entry(rdev1, &mddev->disks, same_set) { 99 list_for_each_entry(rdev1, &mddev->disks, same_set) {
73 printk(KERN_INFO "raid0: looking at %s\n", 100 printk(KERN_INFO "raid0: looking at %s\n",
74 bdevname(rdev1->bdev,b)); 101 bdevname(rdev1->bdev,b));
75 c = 0; 102 c = 0;
103
104 /* round size to chunk_size */
105 sectors = rdev1->sectors;
106 sector_div(sectors, mddev->chunk_sectors);
107 rdev1->sectors = sectors * mddev->chunk_sectors;
108
76 list_for_each_entry(rdev2, &mddev->disks, same_set) { 109 list_for_each_entry(rdev2, &mddev->disks, same_set) {
77 printk(KERN_INFO "raid0: comparing %s(%llu)", 110 printk(KERN_INFO "raid0: comparing %s(%llu)",
78 bdevname(rdev1->bdev,b), 111 bdevname(rdev1->bdev,b),
@@ -103,16 +136,16 @@ static int create_strip_zones (mddev_t *mddev)
103 } 136 }
104 } 137 }
105 printk(KERN_INFO "raid0: FINAL %d zones\n", conf->nr_strip_zones); 138 printk(KERN_INFO "raid0: FINAL %d zones\n", conf->nr_strip_zones);
106 139 err = -ENOMEM;
107 conf->strip_zone = kzalloc(sizeof(struct strip_zone)* 140 conf->strip_zone = kzalloc(sizeof(struct strip_zone)*
108 conf->nr_strip_zones, GFP_KERNEL); 141 conf->nr_strip_zones, GFP_KERNEL);
109 if (!conf->strip_zone) 142 if (!conf->strip_zone)
110 return 1; 143 goto abort;
111 conf->devlist = kzalloc(sizeof(mdk_rdev_t*)* 144 conf->devlist = kzalloc(sizeof(mdk_rdev_t*)*
112 conf->nr_strip_zones*mddev->raid_disks, 145 conf->nr_strip_zones*mddev->raid_disks,
113 GFP_KERNEL); 146 GFP_KERNEL);
114 if (!conf->devlist) 147 if (!conf->devlist)
115 return 1; 148 goto abort;
116 149
117 /* The first zone must contain all devices, so here we check that 150 /* The first zone must contain all devices, so here we check that
118 * there is a proper alignment of slots to devices and find them all 151 * there is a proper alignment of slots to devices and find them all
@@ -120,7 +153,8 @@ static int create_strip_zones (mddev_t *mddev)
120 zone = &conf->strip_zone[0]; 153 zone = &conf->strip_zone[0];
121 cnt = 0; 154 cnt = 0;
122 smallest = NULL; 155 smallest = NULL;
123 zone->dev = conf->devlist; 156 dev = conf->devlist;
157 err = -EINVAL;
124 list_for_each_entry(rdev1, &mddev->disks, same_set) { 158 list_for_each_entry(rdev1, &mddev->disks, same_set) {
125 int j = rdev1->raid_disk; 159 int j = rdev1->raid_disk;
126 160
@@ -129,12 +163,12 @@ static int create_strip_zones (mddev_t *mddev)
129 "aborting!\n", j); 163 "aborting!\n", j);
130 goto abort; 164 goto abort;
131 } 165 }
132 if (zone->dev[j]) { 166 if (dev[j]) {
133 printk(KERN_ERR "raid0: multiple devices for %d - " 167 printk(KERN_ERR "raid0: multiple devices for %d - "
134 "aborting!\n", j); 168 "aborting!\n", j);
135 goto abort; 169 goto abort;
136 } 170 }
137 zone->dev[j] = rdev1; 171 dev[j] = rdev1;
138 172
139 blk_queue_stack_limits(mddev->queue, 173 blk_queue_stack_limits(mddev->queue,
140 rdev1->bdev->bd_disk->queue); 174 rdev1->bdev->bd_disk->queue);
@@ -157,34 +191,32 @@ static int create_strip_zones (mddev_t *mddev)
157 goto abort; 191 goto abort;
158 } 192 }
159 zone->nb_dev = cnt; 193 zone->nb_dev = cnt;
160 zone->sectors = smallest->sectors * cnt; 194 zone->zone_end = smallest->sectors * cnt;
161 zone->zone_start = 0;
162 195
163 current_start = smallest->sectors; 196 curr_zone_end = zone->zone_end;
164 curr_zone_start = zone->sectors;
165 197
166 /* now do the other zones */ 198 /* now do the other zones */
167 for (i = 1; i < conf->nr_strip_zones; i++) 199 for (i = 1; i < conf->nr_strip_zones; i++)
168 { 200 {
169 zone = conf->strip_zone + i; 201 zone = conf->strip_zone + i;
170 zone->dev = conf->strip_zone[i-1].dev + mddev->raid_disks; 202 dev = conf->devlist + i * mddev->raid_disks;
171 203
172 printk(KERN_INFO "raid0: zone %d\n", i); 204 printk(KERN_INFO "raid0: zone %d\n", i);
173 zone->dev_start = current_start; 205 zone->dev_start = smallest->sectors;
174 smallest = NULL; 206 smallest = NULL;
175 c = 0; 207 c = 0;
176 208
177 for (j=0; j<cnt; j++) { 209 for (j=0; j<cnt; j++) {
178 char b[BDEVNAME_SIZE]; 210 char b[BDEVNAME_SIZE];
179 rdev = conf->strip_zone[0].dev[j]; 211 rdev = conf->devlist[j];
180 printk(KERN_INFO "raid0: checking %s ...", 212 printk(KERN_INFO "raid0: checking %s ...",
181 bdevname(rdev->bdev, b)); 213 bdevname(rdev->bdev, b));
182 if (rdev->sectors <= current_start) { 214 if (rdev->sectors <= zone->dev_start) {
183 printk(KERN_INFO " nope.\n"); 215 printk(KERN_INFO " nope.\n");
184 continue; 216 continue;
185 } 217 }
186 printk(KERN_INFO " contained as device %d\n", c); 218 printk(KERN_INFO " contained as device %d\n", c);
187 zone->dev[c] = rdev; 219 dev[c] = rdev;
188 c++; 220 c++;
189 if (!smallest || rdev->sectors < smallest->sectors) { 221 if (!smallest || rdev->sectors < smallest->sectors) {
190 smallest = rdev; 222 smallest = rdev;
@@ -194,47 +226,39 @@ static int create_strip_zones (mddev_t *mddev)
194 } 226 }
195 227
196 zone->nb_dev = c; 228 zone->nb_dev = c;
197 zone->sectors = (smallest->sectors - current_start) * c; 229 sectors = (smallest->sectors - zone->dev_start) * c;
198 printk(KERN_INFO "raid0: zone->nb_dev: %d, sectors: %llu\n", 230 printk(KERN_INFO "raid0: zone->nb_dev: %d, sectors: %llu\n",
199 zone->nb_dev, (unsigned long long)zone->sectors); 231 zone->nb_dev, (unsigned long long)sectors);
200 232
201 zone->zone_start = curr_zone_start; 233 curr_zone_end += sectors;
202 curr_zone_start += zone->sectors; 234 zone->zone_end = curr_zone_end;
203 235
204 current_start = smallest->sectors;
205 printk(KERN_INFO "raid0: current zone start: %llu\n", 236 printk(KERN_INFO "raid0: current zone start: %llu\n",
206 (unsigned long long)current_start); 237 (unsigned long long)smallest->sectors);
207 }
208
209 /* Now find appropriate hash spacing.
210 * We want a number which causes most hash entries to cover
211 * at most two strips, but the hash table must be at most
212 * 1 PAGE. We choose the smallest strip, or contiguous collection
213 * of strips, that has big enough size. We never consider the last
214 * strip though as it's size has no bearing on the efficacy of the hash
215 * table.
216 */
217 conf->spacing = curr_zone_start;
218 min_spacing = curr_zone_start;
219 sector_div(min_spacing, PAGE_SIZE/sizeof(struct strip_zone*));
220 for (i=0; i < conf->nr_strip_zones-1; i++) {
221 sector_t s = 0;
222 for (j = i; j < conf->nr_strip_zones - 1 &&
223 s < min_spacing; j++)
224 s += conf->strip_zone[j].sectors;
225 if (s >= min_spacing && s < conf->spacing)
226 conf->spacing = s;
227 } 238 }
228
229 mddev->queue->unplug_fn = raid0_unplug; 239 mddev->queue->unplug_fn = raid0_unplug;
230
231 mddev->queue->backing_dev_info.congested_fn = raid0_congested; 240 mddev->queue->backing_dev_info.congested_fn = raid0_congested;
232 mddev->queue->backing_dev_info.congested_data = mddev; 241 mddev->queue->backing_dev_info.congested_data = mddev;
233 242
243 /*
244 * now since we have the hard sector sizes, we can make sure
245 * chunk size is a multiple of that sector size
246 */
247 if ((mddev->chunk_sectors << 9) % queue_logical_block_size(mddev->queue)) {
248 printk(KERN_ERR "%s chunk_size of %d not valid\n",
249 mdname(mddev),
250 mddev->chunk_sectors << 9);
251 goto abort;
252 }
234 printk(KERN_INFO "raid0: done.\n"); 253 printk(KERN_INFO "raid0: done.\n");
254 mddev->private = conf;
235 return 0; 255 return 0;
236 abort: 256abort:
237 return 1; 257 kfree(conf->strip_zone);
258 kfree(conf->devlist);
259 kfree(conf);
260 mddev->private = NULL;
261 return err;
238} 262}
239 263
240/** 264/**
@@ -252,10 +276,15 @@ static int raid0_mergeable_bvec(struct request_queue *q,
252 mddev_t *mddev = q->queuedata; 276 mddev_t *mddev = q->queuedata;
253 sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev); 277 sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
254 int max; 278 int max;
255 unsigned int chunk_sectors = mddev->chunk_size >> 9; 279 unsigned int chunk_sectors = mddev->chunk_sectors;
256 unsigned int bio_sectors = bvm->bi_size >> 9; 280 unsigned int bio_sectors = bvm->bi_size >> 9;
257 281
258 max = (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9; 282 if (is_power_of_2(chunk_sectors))
283 max = (chunk_sectors - ((sector & (chunk_sectors-1))
284 + bio_sectors)) << 9;
285 else
286 max = (chunk_sectors - (sector_div(sector, chunk_sectors)
287 + bio_sectors)) << 9;
259 if (max < 0) max = 0; /* bio_add cannot handle a negative return */ 288 if (max < 0) max = 0; /* bio_add cannot handle a negative return */
260 if (max <= biovec->bv_len && bio_sectors == 0) 289 if (max <= biovec->bv_len && bio_sectors == 0)
261 return biovec->bv_len; 290 return biovec->bv_len;
@@ -277,84 +306,28 @@ static sector_t raid0_size(mddev_t *mddev, sector_t sectors, int raid_disks)
277 return array_sectors; 306 return array_sectors;
278} 307}
279 308
280static int raid0_run (mddev_t *mddev) 309static int raid0_run(mddev_t *mddev)
281{ 310{
282 unsigned cur=0, i=0, nb_zone; 311 int ret;
283 s64 sectors;
284 raid0_conf_t *conf;
285 312
286 if (mddev->chunk_size == 0) { 313 if (mddev->chunk_sectors == 0) {
287 printk(KERN_ERR "md/raid0: non-zero chunk size required.\n"); 314 printk(KERN_ERR "md/raid0: chunk size must be set.\n");
288 return -EINVAL; 315 return -EINVAL;
289 } 316 }
290 printk(KERN_INFO "%s: setting max_sectors to %d, segment boundary to %d\n", 317 if (md_check_no_bitmap(mddev))
291 mdname(mddev), 318 return -EINVAL;
292 mddev->chunk_size >> 9, 319 blk_queue_max_sectors(mddev->queue, mddev->chunk_sectors);
293 (mddev->chunk_size>>1)-1);
294 blk_queue_max_sectors(mddev->queue, mddev->chunk_size >> 9);
295 blk_queue_segment_boundary(mddev->queue, (mddev->chunk_size>>1) - 1);
296 mddev->queue->queue_lock = &mddev->queue->__queue_lock; 320 mddev->queue->queue_lock = &mddev->queue->__queue_lock;
297 321
298 conf = kmalloc(sizeof (raid0_conf_t), GFP_KERNEL); 322 ret = create_strip_zones(mddev);
299 if (!conf) 323 if (ret < 0)
300 goto out; 324 return ret;
301 mddev->private = (void *)conf;
302
303 conf->strip_zone = NULL;
304 conf->devlist = NULL;
305 if (create_strip_zones (mddev))
306 goto out_free_conf;
307 325
308 /* calculate array device size */ 326 /* calculate array device size */
309 md_set_array_sectors(mddev, raid0_size(mddev, 0, 0)); 327 md_set_array_sectors(mddev, raid0_size(mddev, 0, 0));
310 328
311 printk(KERN_INFO "raid0 : md_size is %llu sectors.\n", 329 printk(KERN_INFO "raid0 : md_size is %llu sectors.\n",
312 (unsigned long long)mddev->array_sectors); 330 (unsigned long long)mddev->array_sectors);
313 printk(KERN_INFO "raid0 : conf->spacing is %llu sectors.\n",
314 (unsigned long long)conf->spacing);
315 {
316 sector_t s = raid0_size(mddev, 0, 0);
317 sector_t space = conf->spacing;
318 int round;
319 conf->sector_shift = 0;
320 if (sizeof(sector_t) > sizeof(u32)) {
321 /*shift down space and s so that sector_div will work */
322 while (space > (sector_t) (~(u32)0)) {
323 s >>= 1;
324 space >>= 1;
325 s += 1; /* force round-up */
326 conf->sector_shift++;
327 }
328 }
329 round = sector_div(s, (u32)space) ? 1 : 0;
330 nb_zone = s + round;
331 }
332 printk(KERN_INFO "raid0 : nb_zone is %d.\n", nb_zone);
333
334 printk(KERN_INFO "raid0 : Allocating %zu bytes for hash.\n",
335 nb_zone*sizeof(struct strip_zone*));
336 conf->hash_table = kmalloc (sizeof (struct strip_zone *)*nb_zone, GFP_KERNEL);
337 if (!conf->hash_table)
338 goto out_free_conf;
339 sectors = conf->strip_zone[cur].sectors;
340
341 conf->hash_table[0] = conf->strip_zone + cur;
342 for (i=1; i< nb_zone; i++) {
343 while (sectors <= conf->spacing) {
344 cur++;
345 sectors += conf->strip_zone[cur].sectors;
346 }
347 sectors -= conf->spacing;
348 conf->hash_table[i] = conf->strip_zone + cur;
349 }
350 if (conf->sector_shift) {
351 conf->spacing >>= conf->sector_shift;
352 /* round spacing up so when we divide by it, we
353 * err on the side of too-low, which is safest
354 */
355 conf->spacing++;
356 }
357
358 /* calculate the max read-ahead size. 331 /* calculate the max read-ahead size.
359 * For read-ahead of large files to be effective, we need to 332 * For read-ahead of large files to be effective, we need to
360 * readahead at least twice a whole stripe. i.e. number of devices 333 * readahead at least twice a whole stripe. i.e. number of devices
@@ -365,48 +338,107 @@ static int raid0_run (mddev_t *mddev)
365 * chunksize should be used in that case. 338 * chunksize should be used in that case.
366 */ 339 */
367 { 340 {
368 int stripe = mddev->raid_disks * mddev->chunk_size / PAGE_SIZE; 341 int stripe = mddev->raid_disks *
342 (mddev->chunk_sectors << 9) / PAGE_SIZE;
369 if (mddev->queue->backing_dev_info.ra_pages < 2* stripe) 343 if (mddev->queue->backing_dev_info.ra_pages < 2* stripe)
370 mddev->queue->backing_dev_info.ra_pages = 2* stripe; 344 mddev->queue->backing_dev_info.ra_pages = 2* stripe;
371 } 345 }
372 346
373
374 blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec); 347 blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec);
348 dump_zones(mddev);
375 return 0; 349 return 0;
350}
376 351
377out_free_conf: 352static int raid0_stop(mddev_t *mddev)
353{
354 raid0_conf_t *conf = mddev->private;
355
356 blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
378 kfree(conf->strip_zone); 357 kfree(conf->strip_zone);
379 kfree(conf->devlist); 358 kfree(conf->devlist);
380 kfree(conf); 359 kfree(conf);
381 mddev->private = NULL; 360 mddev->private = NULL;
382out: 361 return 0;
383 return -ENOMEM;
384} 362}
385 363
386static int raid0_stop (mddev_t *mddev) 364/* Find the zone which holds a particular offset
365 * Update *sectorp to be an offset in that zone
366 */
367static struct strip_zone *find_zone(struct raid0_private_data *conf,
368 sector_t *sectorp)
387{ 369{
388 raid0_conf_t *conf = mddev_to_conf(mddev); 370 int i;
371 struct strip_zone *z = conf->strip_zone;
372 sector_t sector = *sectorp;
373
374 for (i = 0; i < conf->nr_strip_zones; i++)
375 if (sector < z[i].zone_end) {
376 if (i)
377 *sectorp = sector - z[i-1].zone_end;
378 return z + i;
379 }
380 BUG();
381}
389 382
390 blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/ 383/*
391 kfree(conf->hash_table); 384 * remaps the bio to the target device. we separate two flows.
392 conf->hash_table = NULL; 385 * power 2 flow and a general flow for the sake of perfromance
393 kfree(conf->strip_zone); 386*/
394 conf->strip_zone = NULL; 387static mdk_rdev_t *map_sector(mddev_t *mddev, struct strip_zone *zone,
395 kfree(conf); 388 sector_t sector, sector_t *sector_offset)
396 mddev->private = NULL; 389{
390 unsigned int sect_in_chunk;
391 sector_t chunk;
392 raid0_conf_t *conf = mddev->private;
393 unsigned int chunk_sects = mddev->chunk_sectors;
394
395 if (is_power_of_2(chunk_sects)) {
396 int chunksect_bits = ffz(~chunk_sects);
397 /* find the sector offset inside the chunk */
398 sect_in_chunk = sector & (chunk_sects - 1);
399 sector >>= chunksect_bits;
400 /* chunk in zone */
401 chunk = *sector_offset;
402 /* quotient is the chunk in real device*/
403 sector_div(chunk, zone->nb_dev << chunksect_bits);
404 } else{
405 sect_in_chunk = sector_div(sector, chunk_sects);
406 chunk = *sector_offset;
407 sector_div(chunk, chunk_sects * zone->nb_dev);
408 }
409 /*
410 * position the bio over the real device
411 * real sector = chunk in device + starting of zone
412 * + the position in the chunk
413 */
414 *sector_offset = (chunk * chunk_sects) + sect_in_chunk;
415 return conf->devlist[(zone - conf->strip_zone)*mddev->raid_disks
416 + sector_div(sector, zone->nb_dev)];
417}
397 418
398 return 0; 419/*
420 * Is io distribute over 1 or more chunks ?
421*/
422static inline int is_io_in_chunk_boundary(mddev_t *mddev,
423 unsigned int chunk_sects, struct bio *bio)
424{
425 if (likely(is_power_of_2(chunk_sects))) {
426 return chunk_sects >= ((bio->bi_sector & (chunk_sects-1))
427 + (bio->bi_size >> 9));
428 } else{
429 sector_t sector = bio->bi_sector;
430 return chunk_sects >= (sector_div(sector, chunk_sects)
431 + (bio->bi_size >> 9));
432 }
399} 433}
400 434
401static int raid0_make_request (struct request_queue *q, struct bio *bio) 435static int raid0_make_request(struct request_queue *q, struct bio *bio)
402{ 436{
403 mddev_t *mddev = q->queuedata; 437 mddev_t *mddev = q->queuedata;
404 unsigned int sect_in_chunk, chunksect_bits, chunk_sects; 438 unsigned int chunk_sects;
405 raid0_conf_t *conf = mddev_to_conf(mddev); 439 sector_t sector_offset;
406 struct strip_zone *zone; 440 struct strip_zone *zone;
407 mdk_rdev_t *tmp_dev; 441 mdk_rdev_t *tmp_dev;
408 sector_t chunk;
409 sector_t sector, rsect;
410 const int rw = bio_data_dir(bio); 442 const int rw = bio_data_dir(bio);
411 int cpu; 443 int cpu;
412 444
@@ -421,11 +453,9 @@ static int raid0_make_request (struct request_queue *q, struct bio *bio)
421 bio_sectors(bio)); 453 bio_sectors(bio));
422 part_stat_unlock(); 454 part_stat_unlock();
423 455
424 chunk_sects = mddev->chunk_size >> 9; 456 chunk_sects = mddev->chunk_sectors;
425 chunksect_bits = ffz(~chunk_sects); 457 if (unlikely(!is_io_in_chunk_boundary(mddev, chunk_sects, bio))) {
426 sector = bio->bi_sector; 458 sector_t sector = bio->bi_sector;
427
428 if (unlikely(chunk_sects < (bio->bi_sector & (chunk_sects - 1)) + (bio->bi_size >> 9))) {
429 struct bio_pair *bp; 459 struct bio_pair *bp;
430 /* Sanity check -- queue functions should prevent this happening */ 460 /* Sanity check -- queue functions should prevent this happening */
431 if (bio->bi_vcnt != 1 || 461 if (bio->bi_vcnt != 1 ||
@@ -434,7 +464,12 @@ static int raid0_make_request (struct request_queue *q, struct bio *bio)
434 /* This is a one page bio that upper layers 464 /* This is a one page bio that upper layers
435 * refuse to split for us, so we need to split it. 465 * refuse to split for us, so we need to split it.
436 */ 466 */
437 bp = bio_split(bio, chunk_sects - (bio->bi_sector & (chunk_sects - 1))); 467 if (likely(is_power_of_2(chunk_sects)))
468 bp = bio_split(bio, chunk_sects - (sector &
469 (chunk_sects-1)));
470 else
471 bp = bio_split(bio, chunk_sects -
472 sector_div(sector, chunk_sects));
438 if (raid0_make_request(q, &bp->bio1)) 473 if (raid0_make_request(q, &bp->bio1))
439 generic_make_request(&bp->bio1); 474 generic_make_request(&bp->bio1);
440 if (raid0_make_request(q, &bp->bio2)) 475 if (raid0_make_request(q, &bp->bio2))
@@ -443,34 +478,14 @@ static int raid0_make_request (struct request_queue *q, struct bio *bio)
443 bio_pair_release(bp); 478 bio_pair_release(bp);
444 return 0; 479 return 0;
445 } 480 }
446
447
448 {
449 sector_t x = sector >> conf->sector_shift;
450 sector_div(x, (u32)conf->spacing);
451 zone = conf->hash_table[x];
452 }
453 481
454 while (sector >= zone->zone_start + zone->sectors) 482 sector_offset = bio->bi_sector;
455 zone++; 483 zone = find_zone(mddev->private, &sector_offset);
456 484 tmp_dev = map_sector(mddev, zone, bio->bi_sector,
457 sect_in_chunk = bio->bi_sector & (chunk_sects - 1); 485 &sector_offset);
458
459
460 {
461 sector_t x = (sector - zone->zone_start) >> chunksect_bits;
462
463 sector_div(x, zone->nb_dev);
464 chunk = x;
465
466 x = sector >> chunksect_bits;
467 tmp_dev = zone->dev[sector_div(x, zone->nb_dev)];
468 }
469 rsect = (chunk << chunksect_bits) + zone->dev_start + sect_in_chunk;
470
471 bio->bi_bdev = tmp_dev->bdev; 486 bio->bi_bdev = tmp_dev->bdev;
472 bio->bi_sector = rsect + tmp_dev->data_offset; 487 bio->bi_sector = sector_offset + zone->dev_start +
473 488 tmp_dev->data_offset;
474 /* 489 /*
475 * Let the main block layer submit the IO and resolve recursion: 490 * Let the main block layer submit the IO and resolve recursion:
476 */ 491 */
@@ -485,31 +500,35 @@ bad_map:
485 return 0; 500 return 0;
486} 501}
487 502
488static void raid0_status (struct seq_file *seq, mddev_t *mddev) 503static void raid0_status(struct seq_file *seq, mddev_t *mddev)
489{ 504{
490#undef MD_DEBUG 505#undef MD_DEBUG
491#ifdef MD_DEBUG 506#ifdef MD_DEBUG
492 int j, k, h; 507 int j, k, h;
493 char b[BDEVNAME_SIZE]; 508 char b[BDEVNAME_SIZE];
494 raid0_conf_t *conf = mddev_to_conf(mddev); 509 raid0_conf_t *conf = mddev->private;
495 510
511 sector_t zone_size;
512 sector_t zone_start = 0;
496 h = 0; 513 h = 0;
514
497 for (j = 0; j < conf->nr_strip_zones; j++) { 515 for (j = 0; j < conf->nr_strip_zones; j++) {
498 seq_printf(seq, " z%d", j); 516 seq_printf(seq, " z%d", j);
499 if (conf->hash_table[h] == conf->strip_zone+j)
500 seq_printf(seq, "(h%d)", h++);
501 seq_printf(seq, "=["); 517 seq_printf(seq, "=[");
502 for (k = 0; k < conf->strip_zone[j].nb_dev; k++) 518 for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
503 seq_printf(seq, "%s/", bdevname( 519 seq_printf(seq, "%s/", bdevname(
504 conf->strip_zone[j].dev[k]->bdev,b)); 520 conf->devlist[j*mddev->raid_disks + k]
505 521 ->bdev, b));
506 seq_printf(seq, "] zs=%d ds=%d s=%d\n", 522
507 conf->strip_zone[j].zone_start, 523 zone_size = conf->strip_zone[j].zone_end - zone_start;
508 conf->strip_zone[j].dev_start, 524 seq_printf(seq, "] ze=%lld ds=%lld s=%lld\n",
509 conf->strip_zone[j].sectors); 525 (unsigned long long)zone_start>>1,
526 (unsigned long long)conf->strip_zone[j].dev_start>>1,
527 (unsigned long long)zone_size>>1);
528 zone_start = conf->strip_zone[j].zone_end;
510 } 529 }
511#endif 530#endif
512 seq_printf(seq, " %dk chunks", mddev->chunk_size/1024); 531 seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2);
513 return; 532 return;
514} 533}
515 534
diff --git a/drivers/md/raid0.h b/drivers/md/raid0.h
index 824b12eb1d4f..91f8e876ee64 100644
--- a/drivers/md/raid0.h
+++ b/drivers/md/raid0.h
@@ -3,26 +3,18 @@
3 3
4struct strip_zone 4struct strip_zone
5{ 5{
6 sector_t zone_start; /* Zone offset in md_dev (in sectors) */ 6 sector_t zone_end; /* Start of the next zone (in sectors) */
7 sector_t dev_start; /* Zone offset in real dev (in sectors) */ 7 sector_t dev_start; /* Zone offset in real dev (in sectors) */
8 sector_t sectors; /* Zone size in sectors */
9 int nb_dev; /* # of devices attached to the zone */ 8 int nb_dev; /* # of devices attached to the zone */
10 mdk_rdev_t **dev; /* Devices attached to the zone */
11}; 9};
12 10
13struct raid0_private_data 11struct raid0_private_data
14{ 12{
15 struct strip_zone **hash_table; /* Table of indexes into strip_zone */
16 struct strip_zone *strip_zone; 13 struct strip_zone *strip_zone;
17 mdk_rdev_t **devlist; /* lists of rdevs, pointed to by strip_zone->dev */ 14 mdk_rdev_t **devlist; /* lists of rdevs, pointed to by strip_zone->dev */
18 int nr_strip_zones; 15 int nr_strip_zones;
19
20 sector_t spacing;
21 int sector_shift; /* shift this before divide by spacing */
22}; 16};
23 17
24typedef struct raid0_private_data raid0_conf_t; 18typedef struct raid0_private_data raid0_conf_t;
25 19
26#define mddev_to_conf(mddev) ((raid0_conf_t *) mddev->private)
27
28#endif 20#endif
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index e23758b4a34e..89939a7aef57 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -182,7 +182,7 @@ static void put_all_bios(conf_t *conf, r1bio_t *r1_bio)
182 182
183static void free_r1bio(r1bio_t *r1_bio) 183static void free_r1bio(r1bio_t *r1_bio)
184{ 184{
185 conf_t *conf = mddev_to_conf(r1_bio->mddev); 185 conf_t *conf = r1_bio->mddev->private;
186 186
187 /* 187 /*
188 * Wake up any possible resync thread that waits for the device 188 * Wake up any possible resync thread that waits for the device
@@ -196,7 +196,7 @@ static void free_r1bio(r1bio_t *r1_bio)
196 196
197static void put_buf(r1bio_t *r1_bio) 197static void put_buf(r1bio_t *r1_bio)
198{ 198{
199 conf_t *conf = mddev_to_conf(r1_bio->mddev); 199 conf_t *conf = r1_bio->mddev->private;
200 int i; 200 int i;
201 201
202 for (i=0; i<conf->raid_disks; i++) { 202 for (i=0; i<conf->raid_disks; i++) {
@@ -214,7 +214,7 @@ static void reschedule_retry(r1bio_t *r1_bio)
214{ 214{
215 unsigned long flags; 215 unsigned long flags;
216 mddev_t *mddev = r1_bio->mddev; 216 mddev_t *mddev = r1_bio->mddev;
217 conf_t *conf = mddev_to_conf(mddev); 217 conf_t *conf = mddev->private;
218 218
219 spin_lock_irqsave(&conf->device_lock, flags); 219 spin_lock_irqsave(&conf->device_lock, flags);
220 list_add(&r1_bio->retry_list, &conf->retry_list); 220 list_add(&r1_bio->retry_list, &conf->retry_list);
@@ -253,7 +253,7 @@ static void raid_end_bio_io(r1bio_t *r1_bio)
253 */ 253 */
254static inline void update_head_pos(int disk, r1bio_t *r1_bio) 254static inline void update_head_pos(int disk, r1bio_t *r1_bio)
255{ 255{
256 conf_t *conf = mddev_to_conf(r1_bio->mddev); 256 conf_t *conf = r1_bio->mddev->private;
257 257
258 conf->mirrors[disk].head_position = 258 conf->mirrors[disk].head_position =
259 r1_bio->sector + (r1_bio->sectors); 259 r1_bio->sector + (r1_bio->sectors);
@@ -264,7 +264,7 @@ static void raid1_end_read_request(struct bio *bio, int error)
264 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); 264 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
265 r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private); 265 r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
266 int mirror; 266 int mirror;
267 conf_t *conf = mddev_to_conf(r1_bio->mddev); 267 conf_t *conf = r1_bio->mddev->private;
268 268
269 mirror = r1_bio->read_disk; 269 mirror = r1_bio->read_disk;
270 /* 270 /*
@@ -309,7 +309,7 @@ static void raid1_end_write_request(struct bio *bio, int error)
309 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); 309 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
310 r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private); 310 r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
311 int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state); 311 int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state);
312 conf_t *conf = mddev_to_conf(r1_bio->mddev); 312 conf_t *conf = r1_bio->mddev->private;
313 struct bio *to_put = NULL; 313 struct bio *to_put = NULL;
314 314
315 315
@@ -541,7 +541,7 @@ static int read_balance(conf_t *conf, r1bio_t *r1_bio)
541 541
542static void unplug_slaves(mddev_t *mddev) 542static void unplug_slaves(mddev_t *mddev)
543{ 543{
544 conf_t *conf = mddev_to_conf(mddev); 544 conf_t *conf = mddev->private;
545 int i; 545 int i;
546 546
547 rcu_read_lock(); 547 rcu_read_lock();
@@ -573,7 +573,7 @@ static void raid1_unplug(struct request_queue *q)
573static int raid1_congested(void *data, int bits) 573static int raid1_congested(void *data, int bits)
574{ 574{
575 mddev_t *mddev = data; 575 mddev_t *mddev = data;
576 conf_t *conf = mddev_to_conf(mddev); 576 conf_t *conf = mddev->private;
577 int i, ret = 0; 577 int i, ret = 0;
578 578
579 rcu_read_lock(); 579 rcu_read_lock();
@@ -772,7 +772,7 @@ do_sync_io:
772static int make_request(struct request_queue *q, struct bio * bio) 772static int make_request(struct request_queue *q, struct bio * bio)
773{ 773{
774 mddev_t *mddev = q->queuedata; 774 mddev_t *mddev = q->queuedata;
775 conf_t *conf = mddev_to_conf(mddev); 775 conf_t *conf = mddev->private;
776 mirror_info_t *mirror; 776 mirror_info_t *mirror;
777 r1bio_t *r1_bio; 777 r1bio_t *r1_bio;
778 struct bio *read_bio; 778 struct bio *read_bio;
@@ -991,7 +991,7 @@ static int make_request(struct request_queue *q, struct bio * bio)
991 991
992static void status(struct seq_file *seq, mddev_t *mddev) 992static void status(struct seq_file *seq, mddev_t *mddev)
993{ 993{
994 conf_t *conf = mddev_to_conf(mddev); 994 conf_t *conf = mddev->private;
995 int i; 995 int i;
996 996
997 seq_printf(seq, " [%d/%d] [", conf->raid_disks, 997 seq_printf(seq, " [%d/%d] [", conf->raid_disks,
@@ -1010,7 +1010,7 @@ static void status(struct seq_file *seq, mddev_t *mddev)
1010static void error(mddev_t *mddev, mdk_rdev_t *rdev) 1010static void error(mddev_t *mddev, mdk_rdev_t *rdev)
1011{ 1011{
1012 char b[BDEVNAME_SIZE]; 1012 char b[BDEVNAME_SIZE];
1013 conf_t *conf = mddev_to_conf(mddev); 1013 conf_t *conf = mddev->private;
1014 1014
1015 /* 1015 /*
1016 * If it is not operational, then we have already marked it as dead 1016 * If it is not operational, then we have already marked it as dead
@@ -1214,7 +1214,7 @@ static void end_sync_write(struct bio *bio, int error)
1214 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); 1214 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1215 r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private); 1215 r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
1216 mddev_t *mddev = r1_bio->mddev; 1216 mddev_t *mddev = r1_bio->mddev;
1217 conf_t *conf = mddev_to_conf(mddev); 1217 conf_t *conf = mddev->private;
1218 int i; 1218 int i;
1219 int mirror=0; 1219 int mirror=0;
1220 1220
@@ -1248,7 +1248,7 @@ static void end_sync_write(struct bio *bio, int error)
1248 1248
1249static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio) 1249static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
1250{ 1250{
1251 conf_t *conf = mddev_to_conf(mddev); 1251 conf_t *conf = mddev->private;
1252 int i; 1252 int i;
1253 int disks = conf->raid_disks; 1253 int disks = conf->raid_disks;
1254 struct bio *bio, *wbio; 1254 struct bio *bio, *wbio;
@@ -1562,7 +1562,7 @@ static void raid1d(mddev_t *mddev)
1562 r1bio_t *r1_bio; 1562 r1bio_t *r1_bio;
1563 struct bio *bio; 1563 struct bio *bio;
1564 unsigned long flags; 1564 unsigned long flags;
1565 conf_t *conf = mddev_to_conf(mddev); 1565 conf_t *conf = mddev->private;
1566 struct list_head *head = &conf->retry_list; 1566 struct list_head *head = &conf->retry_list;
1567 int unplug=0; 1567 int unplug=0;
1568 mdk_rdev_t *rdev; 1568 mdk_rdev_t *rdev;
@@ -1585,7 +1585,7 @@ static void raid1d(mddev_t *mddev)
1585 spin_unlock_irqrestore(&conf->device_lock, flags); 1585 spin_unlock_irqrestore(&conf->device_lock, flags);
1586 1586
1587 mddev = r1_bio->mddev; 1587 mddev = r1_bio->mddev;
1588 conf = mddev_to_conf(mddev); 1588 conf = mddev->private;
1589 if (test_bit(R1BIO_IsSync, &r1_bio->state)) { 1589 if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
1590 sync_request_write(mddev, r1_bio); 1590 sync_request_write(mddev, r1_bio);
1591 unplug = 1; 1591 unplug = 1;
@@ -1706,7 +1706,7 @@ static int init_resync(conf_t *conf)
1706 1706
1707static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster) 1707static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
1708{ 1708{
1709 conf_t *conf = mddev_to_conf(mddev); 1709 conf_t *conf = mddev->private;
1710 r1bio_t *r1_bio; 1710 r1bio_t *r1_bio;
1711 struct bio *bio; 1711 struct bio *bio;
1712 sector_t max_sector, nr_sectors; 1712 sector_t max_sector, nr_sectors;
@@ -2052,6 +2052,10 @@ static int run(mddev_t *mddev)
2052 goto out_free_conf; 2052 goto out_free_conf;
2053 } 2053 }
2054 2054
2055 if (mddev->recovery_cp != MaxSector)
2056 printk(KERN_NOTICE "raid1: %s is not clean"
2057 " -- starting background reconstruction\n",
2058 mdname(mddev));
2055 printk(KERN_INFO 2059 printk(KERN_INFO
2056 "raid1: raid set %s active with %d out of %d mirrors\n", 2060 "raid1: raid set %s active with %d out of %d mirrors\n",
2057 mdname(mddev), mddev->raid_disks - mddev->degraded, 2061 mdname(mddev), mddev->raid_disks - mddev->degraded,
@@ -2087,7 +2091,7 @@ out:
2087 2091
2088static int stop(mddev_t *mddev) 2092static int stop(mddev_t *mddev)
2089{ 2093{
2090 conf_t *conf = mddev_to_conf(mddev); 2094 conf_t *conf = mddev->private;
2091 struct bitmap *bitmap = mddev->bitmap; 2095 struct bitmap *bitmap = mddev->bitmap;
2092 int behind_wait = 0; 2096 int behind_wait = 0;
2093 2097
@@ -2155,16 +2159,16 @@ static int raid1_reshape(mddev_t *mddev)
2155 mempool_t *newpool, *oldpool; 2159 mempool_t *newpool, *oldpool;
2156 struct pool_info *newpoolinfo; 2160 struct pool_info *newpoolinfo;
2157 mirror_info_t *newmirrors; 2161 mirror_info_t *newmirrors;
2158 conf_t *conf = mddev_to_conf(mddev); 2162 conf_t *conf = mddev->private;
2159 int cnt, raid_disks; 2163 int cnt, raid_disks;
2160 unsigned long flags; 2164 unsigned long flags;
2161 int d, d2, err; 2165 int d, d2, err;
2162 2166
2163 /* Cannot change chunk_size, layout, or level */ 2167 /* Cannot change chunk_size, layout, or level */
2164 if (mddev->chunk_size != mddev->new_chunk || 2168 if (mddev->chunk_sectors != mddev->new_chunk_sectors ||
2165 mddev->layout != mddev->new_layout || 2169 mddev->layout != mddev->new_layout ||
2166 mddev->level != mddev->new_level) { 2170 mddev->level != mddev->new_level) {
2167 mddev->new_chunk = mddev->chunk_size; 2171 mddev->new_chunk_sectors = mddev->chunk_sectors;
2168 mddev->new_layout = mddev->layout; 2172 mddev->new_layout = mddev->layout;
2169 mddev->new_level = mddev->level; 2173 mddev->new_level = mddev->level;
2170 return -EINVAL; 2174 return -EINVAL;
@@ -2252,7 +2256,7 @@ static int raid1_reshape(mddev_t *mddev)
2252 2256
2253static void raid1_quiesce(mddev_t *mddev, int state) 2257static void raid1_quiesce(mddev_t *mddev, int state)
2254{ 2258{
2255 conf_t *conf = mddev_to_conf(mddev); 2259 conf_t *conf = mddev->private;
2256 2260
2257 switch(state) { 2261 switch(state) {
2258 case 1: 2262 case 1:
diff --git a/drivers/md/raid1.h b/drivers/md/raid1.h
index 1620eea3d57c..e87b84deff68 100644
--- a/drivers/md/raid1.h
+++ b/drivers/md/raid1.h
@@ -64,12 +64,6 @@ struct r1_private_data_s {
64typedef struct r1_private_data_s conf_t; 64typedef struct r1_private_data_s conf_t;
65 65
66/* 66/*
67 * this is the only point in the RAID code where we violate
68 * C type safety. mddev->private is an 'opaque' pointer.
69 */
70#define mddev_to_conf(mddev) ((conf_t *) mddev->private)
71
72/*
73 * this is our 'private' RAID1 bio. 67 * this is our 'private' RAID1 bio.
74 * 68 *
75 * it contains information about what kind of IO operations were started 69 * it contains information about what kind of IO operations were started
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index 750550c1166f..ae12ceafe10c 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -188,7 +188,7 @@ static void put_all_bios(conf_t *conf, r10bio_t *r10_bio)
188 188
189static void free_r10bio(r10bio_t *r10_bio) 189static void free_r10bio(r10bio_t *r10_bio)
190{ 190{
191 conf_t *conf = mddev_to_conf(r10_bio->mddev); 191 conf_t *conf = r10_bio->mddev->private;
192 192
193 /* 193 /*
194 * Wake up any possible resync thread that waits for the device 194 * Wake up any possible resync thread that waits for the device
@@ -202,7 +202,7 @@ static void free_r10bio(r10bio_t *r10_bio)
202 202
203static void put_buf(r10bio_t *r10_bio) 203static void put_buf(r10bio_t *r10_bio)
204{ 204{
205 conf_t *conf = mddev_to_conf(r10_bio->mddev); 205 conf_t *conf = r10_bio->mddev->private;
206 206
207 mempool_free(r10_bio, conf->r10buf_pool); 207 mempool_free(r10_bio, conf->r10buf_pool);
208 208
@@ -213,7 +213,7 @@ static void reschedule_retry(r10bio_t *r10_bio)
213{ 213{
214 unsigned long flags; 214 unsigned long flags;
215 mddev_t *mddev = r10_bio->mddev; 215 mddev_t *mddev = r10_bio->mddev;
216 conf_t *conf = mddev_to_conf(mddev); 216 conf_t *conf = mddev->private;
217 217
218 spin_lock_irqsave(&conf->device_lock, flags); 218 spin_lock_irqsave(&conf->device_lock, flags);
219 list_add(&r10_bio->retry_list, &conf->retry_list); 219 list_add(&r10_bio->retry_list, &conf->retry_list);
@@ -245,7 +245,7 @@ static void raid_end_bio_io(r10bio_t *r10_bio)
245 */ 245 */
246static inline void update_head_pos(int slot, r10bio_t *r10_bio) 246static inline void update_head_pos(int slot, r10bio_t *r10_bio)
247{ 247{
248 conf_t *conf = mddev_to_conf(r10_bio->mddev); 248 conf_t *conf = r10_bio->mddev->private;
249 249
250 conf->mirrors[r10_bio->devs[slot].devnum].head_position = 250 conf->mirrors[r10_bio->devs[slot].devnum].head_position =
251 r10_bio->devs[slot].addr + (r10_bio->sectors); 251 r10_bio->devs[slot].addr + (r10_bio->sectors);
@@ -256,7 +256,7 @@ static void raid10_end_read_request(struct bio *bio, int error)
256 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); 256 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
257 r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private); 257 r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
258 int slot, dev; 258 int slot, dev;
259 conf_t *conf = mddev_to_conf(r10_bio->mddev); 259 conf_t *conf = r10_bio->mddev->private;
260 260
261 261
262 slot = r10_bio->read_slot; 262 slot = r10_bio->read_slot;
@@ -297,7 +297,7 @@ static void raid10_end_write_request(struct bio *bio, int error)
297 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); 297 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
298 r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private); 298 r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
299 int slot, dev; 299 int slot, dev;
300 conf_t *conf = mddev_to_conf(r10_bio->mddev); 300 conf_t *conf = r10_bio->mddev->private;
301 301
302 for (slot = 0; slot < conf->copies; slot++) 302 for (slot = 0; slot < conf->copies; slot++)
303 if (r10_bio->devs[slot].bio == bio) 303 if (r10_bio->devs[slot].bio == bio)
@@ -461,7 +461,7 @@ static int raid10_mergeable_bvec(struct request_queue *q,
461 mddev_t *mddev = q->queuedata; 461 mddev_t *mddev = q->queuedata;
462 sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev); 462 sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
463 int max; 463 int max;
464 unsigned int chunk_sectors = mddev->chunk_size >> 9; 464 unsigned int chunk_sectors = mddev->chunk_sectors;
465 unsigned int bio_sectors = bvm->bi_size >> 9; 465 unsigned int bio_sectors = bvm->bi_size >> 9;
466 466
467 max = (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9; 467 max = (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9;
@@ -596,7 +596,7 @@ rb_out:
596 596
597static void unplug_slaves(mddev_t *mddev) 597static void unplug_slaves(mddev_t *mddev)
598{ 598{
599 conf_t *conf = mddev_to_conf(mddev); 599 conf_t *conf = mddev->private;
600 int i; 600 int i;
601 601
602 rcu_read_lock(); 602 rcu_read_lock();
@@ -628,7 +628,7 @@ static void raid10_unplug(struct request_queue *q)
628static int raid10_congested(void *data, int bits) 628static int raid10_congested(void *data, int bits)
629{ 629{
630 mddev_t *mddev = data; 630 mddev_t *mddev = data;
631 conf_t *conf = mddev_to_conf(mddev); 631 conf_t *conf = mddev->private;
632 int i, ret = 0; 632 int i, ret = 0;
633 633
634 rcu_read_lock(); 634 rcu_read_lock();
@@ -788,7 +788,7 @@ static void unfreeze_array(conf_t *conf)
788static int make_request(struct request_queue *q, struct bio * bio) 788static int make_request(struct request_queue *q, struct bio * bio)
789{ 789{
790 mddev_t *mddev = q->queuedata; 790 mddev_t *mddev = q->queuedata;
791 conf_t *conf = mddev_to_conf(mddev); 791 conf_t *conf = mddev->private;
792 mirror_info_t *mirror; 792 mirror_info_t *mirror;
793 r10bio_t *r10_bio; 793 r10bio_t *r10_bio;
794 struct bio *read_bio; 794 struct bio *read_bio;
@@ -981,11 +981,11 @@ static int make_request(struct request_queue *q, struct bio * bio)
981 981
982static void status(struct seq_file *seq, mddev_t *mddev) 982static void status(struct seq_file *seq, mddev_t *mddev)
983{ 983{
984 conf_t *conf = mddev_to_conf(mddev); 984 conf_t *conf = mddev->private;
985 int i; 985 int i;
986 986
987 if (conf->near_copies < conf->raid_disks) 987 if (conf->near_copies < conf->raid_disks)
988 seq_printf(seq, " %dK chunks", mddev->chunk_size/1024); 988 seq_printf(seq, " %dK chunks", mddev->chunk_sectors / 2);
989 if (conf->near_copies > 1) 989 if (conf->near_copies > 1)
990 seq_printf(seq, " %d near-copies", conf->near_copies); 990 seq_printf(seq, " %d near-copies", conf->near_copies);
991 if (conf->far_copies > 1) { 991 if (conf->far_copies > 1) {
@@ -1006,7 +1006,7 @@ static void status(struct seq_file *seq, mddev_t *mddev)
1006static void error(mddev_t *mddev, mdk_rdev_t *rdev) 1006static void error(mddev_t *mddev, mdk_rdev_t *rdev)
1007{ 1007{
1008 char b[BDEVNAME_SIZE]; 1008 char b[BDEVNAME_SIZE];
1009 conf_t *conf = mddev_to_conf(mddev); 1009 conf_t *conf = mddev->private;
1010 1010
1011 /* 1011 /*
1012 * If it is not operational, then we have already marked it as dead 1012 * If it is not operational, then we have already marked it as dead
@@ -1215,7 +1215,7 @@ abort:
1215static void end_sync_read(struct bio *bio, int error) 1215static void end_sync_read(struct bio *bio, int error)
1216{ 1216{
1217 r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private); 1217 r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
1218 conf_t *conf = mddev_to_conf(r10_bio->mddev); 1218 conf_t *conf = r10_bio->mddev->private;
1219 int i,d; 1219 int i,d;
1220 1220
1221 for (i=0; i<conf->copies; i++) 1221 for (i=0; i<conf->copies; i++)
@@ -1253,7 +1253,7 @@ static void end_sync_write(struct bio *bio, int error)
1253 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); 1253 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1254 r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private); 1254 r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
1255 mddev_t *mddev = r10_bio->mddev; 1255 mddev_t *mddev = r10_bio->mddev;
1256 conf_t *conf = mddev_to_conf(mddev); 1256 conf_t *conf = mddev->private;
1257 int i,d; 1257 int i,d;
1258 1258
1259 for (i = 0; i < conf->copies; i++) 1259 for (i = 0; i < conf->copies; i++)
@@ -1300,7 +1300,7 @@ static void end_sync_write(struct bio *bio, int error)
1300 */ 1300 */
1301static void sync_request_write(mddev_t *mddev, r10bio_t *r10_bio) 1301static void sync_request_write(mddev_t *mddev, r10bio_t *r10_bio)
1302{ 1302{
1303 conf_t *conf = mddev_to_conf(mddev); 1303 conf_t *conf = mddev->private;
1304 int i, first; 1304 int i, first;
1305 struct bio *tbio, *fbio; 1305 struct bio *tbio, *fbio;
1306 1306
@@ -1400,7 +1400,7 @@ done:
1400 1400
1401static void recovery_request_write(mddev_t *mddev, r10bio_t *r10_bio) 1401static void recovery_request_write(mddev_t *mddev, r10bio_t *r10_bio)
1402{ 1402{
1403 conf_t *conf = mddev_to_conf(mddev); 1403 conf_t *conf = mddev->private;
1404 int i, d; 1404 int i, d;
1405 struct bio *bio, *wbio; 1405 struct bio *bio, *wbio;
1406 1406
@@ -1549,7 +1549,7 @@ static void raid10d(mddev_t *mddev)
1549 r10bio_t *r10_bio; 1549 r10bio_t *r10_bio;
1550 struct bio *bio; 1550 struct bio *bio;
1551 unsigned long flags; 1551 unsigned long flags;
1552 conf_t *conf = mddev_to_conf(mddev); 1552 conf_t *conf = mddev->private;
1553 struct list_head *head = &conf->retry_list; 1553 struct list_head *head = &conf->retry_list;
1554 int unplug=0; 1554 int unplug=0;
1555 mdk_rdev_t *rdev; 1555 mdk_rdev_t *rdev;
@@ -1572,7 +1572,7 @@ static void raid10d(mddev_t *mddev)
1572 spin_unlock_irqrestore(&conf->device_lock, flags); 1572 spin_unlock_irqrestore(&conf->device_lock, flags);
1573 1573
1574 mddev = r10_bio->mddev; 1574 mddev = r10_bio->mddev;
1575 conf = mddev_to_conf(mddev); 1575 conf = mddev->private;
1576 if (test_bit(R10BIO_IsSync, &r10_bio->state)) { 1576 if (test_bit(R10BIO_IsSync, &r10_bio->state)) {
1577 sync_request_write(mddev, r10_bio); 1577 sync_request_write(mddev, r10_bio);
1578 unplug = 1; 1578 unplug = 1;
@@ -1680,7 +1680,7 @@ static int init_resync(conf_t *conf)
1680 1680
1681static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster) 1681static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
1682{ 1682{
1683 conf_t *conf = mddev_to_conf(mddev); 1683 conf_t *conf = mddev->private;
1684 r10bio_t *r10_bio; 1684 r10bio_t *r10_bio;
1685 struct bio *biolist = NULL, *bio; 1685 struct bio *biolist = NULL, *bio;
1686 sector_t max_sector, nr_sectors; 1686 sector_t max_sector, nr_sectors;
@@ -2026,7 +2026,7 @@ static sector_t
2026raid10_size(mddev_t *mddev, sector_t sectors, int raid_disks) 2026raid10_size(mddev_t *mddev, sector_t sectors, int raid_disks)
2027{ 2027{
2028 sector_t size; 2028 sector_t size;
2029 conf_t *conf = mddev_to_conf(mddev); 2029 conf_t *conf = mddev->private;
2030 2030
2031 if (!raid_disks) 2031 if (!raid_disks)
2032 raid_disks = mddev->raid_disks; 2032 raid_disks = mddev->raid_disks;
@@ -2050,9 +2050,10 @@ static int run(mddev_t *mddev)
2050 int nc, fc, fo; 2050 int nc, fc, fo;
2051 sector_t stride, size; 2051 sector_t stride, size;
2052 2052
2053 if (mddev->chunk_size < PAGE_SIZE) { 2053 if (mddev->chunk_sectors < (PAGE_SIZE >> 9) ||
2054 !is_power_of_2(mddev->chunk_sectors)) {
2054 printk(KERN_ERR "md/raid10: chunk size must be " 2055 printk(KERN_ERR "md/raid10: chunk size must be "
2055 "at least PAGE_SIZE(%ld).\n", PAGE_SIZE); 2056 "at least PAGE_SIZE(%ld) and be a power of 2.\n", PAGE_SIZE);
2056 return -EINVAL; 2057 return -EINVAL;
2057 } 2058 }
2058 2059
@@ -2095,8 +2096,8 @@ static int run(mddev_t *mddev)
2095 conf->far_copies = fc; 2096 conf->far_copies = fc;
2096 conf->copies = nc*fc; 2097 conf->copies = nc*fc;
2097 conf->far_offset = fo; 2098 conf->far_offset = fo;
2098 conf->chunk_mask = (sector_t)(mddev->chunk_size>>9)-1; 2099 conf->chunk_mask = mddev->chunk_sectors - 1;
2099 conf->chunk_shift = ffz(~mddev->chunk_size) - 9; 2100 conf->chunk_shift = ffz(~mddev->chunk_sectors);
2100 size = mddev->dev_sectors >> conf->chunk_shift; 2101 size = mddev->dev_sectors >> conf->chunk_shift;
2101 sector_div(size, fc); 2102 sector_div(size, fc);
2102 size = size * conf->raid_disks; 2103 size = size * conf->raid_disks;
@@ -2185,6 +2186,10 @@ static int run(mddev_t *mddev)
2185 goto out_free_conf; 2186 goto out_free_conf;
2186 } 2187 }
2187 2188
2189 if (mddev->recovery_cp != MaxSector)
2190 printk(KERN_NOTICE "raid10: %s is not clean"
2191 " -- starting background reconstruction\n",
2192 mdname(mddev));
2188 printk(KERN_INFO 2193 printk(KERN_INFO
2189 "raid10: raid set %s active with %d out of %d devices\n", 2194 "raid10: raid set %s active with %d out of %d devices\n",
2190 mdname(mddev), mddev->raid_disks - mddev->degraded, 2195 mdname(mddev), mddev->raid_disks - mddev->degraded,
@@ -2204,7 +2209,8 @@ static int run(mddev_t *mddev)
2204 * maybe... 2209 * maybe...
2205 */ 2210 */
2206 { 2211 {
2207 int stripe = conf->raid_disks * (mddev->chunk_size / PAGE_SIZE); 2212 int stripe = conf->raid_disks *
2213 ((mddev->chunk_sectors << 9) / PAGE_SIZE);
2208 stripe /= conf->near_copies; 2214 stripe /= conf->near_copies;
2209 if (mddev->queue->backing_dev_info.ra_pages < 2* stripe) 2215 if (mddev->queue->backing_dev_info.ra_pages < 2* stripe)
2210 mddev->queue->backing_dev_info.ra_pages = 2* stripe; 2216 mddev->queue->backing_dev_info.ra_pages = 2* stripe;
@@ -2227,7 +2233,7 @@ out:
2227 2233
2228static int stop(mddev_t *mddev) 2234static int stop(mddev_t *mddev)
2229{ 2235{
2230 conf_t *conf = mddev_to_conf(mddev); 2236 conf_t *conf = mddev->private;
2231 2237
2232 raise_barrier(conf, 0); 2238 raise_barrier(conf, 0);
2233 lower_barrier(conf); 2239 lower_barrier(conf);
@@ -2245,7 +2251,7 @@ static int stop(mddev_t *mddev)
2245 2251
2246static void raid10_quiesce(mddev_t *mddev, int state) 2252static void raid10_quiesce(mddev_t *mddev, int state)
2247{ 2253{
2248 conf_t *conf = mddev_to_conf(mddev); 2254 conf_t *conf = mddev->private;
2249 2255
2250 switch(state) { 2256 switch(state) {
2251 case 1: 2257 case 1:
diff --git a/drivers/md/raid10.h b/drivers/md/raid10.h
index 244dbe507a54..59cd1efb8d30 100644
--- a/drivers/md/raid10.h
+++ b/drivers/md/raid10.h
@@ -62,12 +62,6 @@ struct r10_private_data_s {
62typedef struct r10_private_data_s conf_t; 62typedef struct r10_private_data_s conf_t;
63 63
64/* 64/*
65 * this is the only point in the RAID code where we violate
66 * C type safety. mddev->private is an 'opaque' pointer.
67 */
68#define mddev_to_conf(mddev) ((conf_t *) mddev->private)
69
70/*
71 * this is our 'private' RAID10 bio. 65 * this is our 'private' RAID10 bio.
72 * 66 *
73 * it contains information about what kind of IO operations were started 67 * it contains information about what kind of IO operations were started
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index bef876698232..f9f991e6e138 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -1274,8 +1274,8 @@ static sector_t raid5_compute_sector(raid5_conf_t *conf, sector_t r_sector,
1274 sector_t new_sector; 1274 sector_t new_sector;
1275 int algorithm = previous ? conf->prev_algo 1275 int algorithm = previous ? conf->prev_algo
1276 : conf->algorithm; 1276 : conf->algorithm;
1277 int sectors_per_chunk = previous ? (conf->prev_chunk >> 9) 1277 int sectors_per_chunk = previous ? conf->prev_chunk_sectors
1278 : (conf->chunk_size >> 9); 1278 : conf->chunk_sectors;
1279 int raid_disks = previous ? conf->previous_raid_disks 1279 int raid_disks = previous ? conf->previous_raid_disks
1280 : conf->raid_disks; 1280 : conf->raid_disks;
1281 int data_disks = raid_disks - conf->max_degraded; 1281 int data_disks = raid_disks - conf->max_degraded;
@@ -1480,8 +1480,8 @@ static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
1480 int raid_disks = sh->disks; 1480 int raid_disks = sh->disks;
1481 int data_disks = raid_disks - conf->max_degraded; 1481 int data_disks = raid_disks - conf->max_degraded;
1482 sector_t new_sector = sh->sector, check; 1482 sector_t new_sector = sh->sector, check;
1483 int sectors_per_chunk = previous ? (conf->prev_chunk >> 9) 1483 int sectors_per_chunk = previous ? conf->prev_chunk_sectors
1484 : (conf->chunk_size >> 9); 1484 : conf->chunk_sectors;
1485 int algorithm = previous ? conf->prev_algo 1485 int algorithm = previous ? conf->prev_algo
1486 : conf->algorithm; 1486 : conf->algorithm;
1487 sector_t stripe; 1487 sector_t stripe;
@@ -1997,8 +1997,7 @@ static void stripe_set_idx(sector_t stripe, raid5_conf_t *conf, int previous,
1997 struct stripe_head *sh) 1997 struct stripe_head *sh)
1998{ 1998{
1999 int sectors_per_chunk = 1999 int sectors_per_chunk =
2000 previous ? (conf->prev_chunk >> 9) 2000 previous ? conf->prev_chunk_sectors : conf->chunk_sectors;
2001 : (conf->chunk_size >> 9);
2002 int dd_idx; 2001 int dd_idx;
2003 int chunk_offset = sector_div(stripe, sectors_per_chunk); 2002 int chunk_offset = sector_div(stripe, sectors_per_chunk);
2004 int disks = previous ? conf->previous_raid_disks : conf->raid_disks; 2003 int disks = previous ? conf->previous_raid_disks : conf->raid_disks;
@@ -3284,7 +3283,7 @@ static void activate_bit_delay(raid5_conf_t *conf)
3284 3283
3285static void unplug_slaves(mddev_t *mddev) 3284static void unplug_slaves(mddev_t *mddev)
3286{ 3285{
3287 raid5_conf_t *conf = mddev_to_conf(mddev); 3286 raid5_conf_t *conf = mddev->private;
3288 int i; 3287 int i;
3289 3288
3290 rcu_read_lock(); 3289 rcu_read_lock();
@@ -3308,7 +3307,7 @@ static void unplug_slaves(mddev_t *mddev)
3308static void raid5_unplug_device(struct request_queue *q) 3307static void raid5_unplug_device(struct request_queue *q)
3309{ 3308{
3310 mddev_t *mddev = q->queuedata; 3309 mddev_t *mddev = q->queuedata;
3311 raid5_conf_t *conf = mddev_to_conf(mddev); 3310 raid5_conf_t *conf = mddev->private;
3312 unsigned long flags; 3311 unsigned long flags;
3313 3312
3314 spin_lock_irqsave(&conf->device_lock, flags); 3313 spin_lock_irqsave(&conf->device_lock, flags);
@@ -3327,7 +3326,7 @@ static void raid5_unplug_device(struct request_queue *q)
3327static int raid5_congested(void *data, int bits) 3326static int raid5_congested(void *data, int bits)
3328{ 3327{
3329 mddev_t *mddev = data; 3328 mddev_t *mddev = data;
3330 raid5_conf_t *conf = mddev_to_conf(mddev); 3329 raid5_conf_t *conf = mddev->private;
3331 3330
3332 /* No difference between reads and writes. Just check 3331 /* No difference between reads and writes. Just check
3333 * how busy the stripe_cache is 3332 * how busy the stripe_cache is
@@ -3352,14 +3351,14 @@ static int raid5_mergeable_bvec(struct request_queue *q,
3352 mddev_t *mddev = q->queuedata; 3351 mddev_t *mddev = q->queuedata;
3353 sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev); 3352 sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
3354 int max; 3353 int max;
3355 unsigned int chunk_sectors = mddev->chunk_size >> 9; 3354 unsigned int chunk_sectors = mddev->chunk_sectors;
3356 unsigned int bio_sectors = bvm->bi_size >> 9; 3355 unsigned int bio_sectors = bvm->bi_size >> 9;
3357 3356
3358 if ((bvm->bi_rw & 1) == WRITE) 3357 if ((bvm->bi_rw & 1) == WRITE)
3359 return biovec->bv_len; /* always allow writes to be mergeable */ 3358 return biovec->bv_len; /* always allow writes to be mergeable */
3360 3359
3361 if (mddev->new_chunk < mddev->chunk_size) 3360 if (mddev->new_chunk_sectors < mddev->chunk_sectors)
3362 chunk_sectors = mddev->new_chunk >> 9; 3361 chunk_sectors = mddev->new_chunk_sectors;
3363 max = (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9; 3362 max = (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9;
3364 if (max < 0) max = 0; 3363 if (max < 0) max = 0;
3365 if (max <= biovec->bv_len && bio_sectors == 0) 3364 if (max <= biovec->bv_len && bio_sectors == 0)
@@ -3372,11 +3371,11 @@ static int raid5_mergeable_bvec(struct request_queue *q,
3372static int in_chunk_boundary(mddev_t *mddev, struct bio *bio) 3371static int in_chunk_boundary(mddev_t *mddev, struct bio *bio)
3373{ 3372{
3374 sector_t sector = bio->bi_sector + get_start_sect(bio->bi_bdev); 3373 sector_t sector = bio->bi_sector + get_start_sect(bio->bi_bdev);
3375 unsigned int chunk_sectors = mddev->chunk_size >> 9; 3374 unsigned int chunk_sectors = mddev->chunk_sectors;
3376 unsigned int bio_sectors = bio->bi_size >> 9; 3375 unsigned int bio_sectors = bio->bi_size >> 9;
3377 3376
3378 if (mddev->new_chunk < mddev->chunk_size) 3377 if (mddev->new_chunk_sectors < mddev->chunk_sectors)
3379 chunk_sectors = mddev->new_chunk >> 9; 3378 chunk_sectors = mddev->new_chunk_sectors;
3380 return chunk_sectors >= 3379 return chunk_sectors >=
3381 ((sector & (chunk_sectors - 1)) + bio_sectors); 3380 ((sector & (chunk_sectors - 1)) + bio_sectors);
3382} 3381}
@@ -3440,7 +3439,7 @@ static void raid5_align_endio(struct bio *bi, int error)
3440 bio_put(bi); 3439 bio_put(bi);
3441 3440
3442 mddev = raid_bi->bi_bdev->bd_disk->queue->queuedata; 3441 mddev = raid_bi->bi_bdev->bd_disk->queue->queuedata;
3443 conf = mddev_to_conf(mddev); 3442 conf = mddev->private;
3444 rdev = (void*)raid_bi->bi_next; 3443 rdev = (void*)raid_bi->bi_next;
3445 raid_bi->bi_next = NULL; 3444 raid_bi->bi_next = NULL;
3446 3445
@@ -3482,7 +3481,7 @@ static int bio_fits_rdev(struct bio *bi)
3482static int chunk_aligned_read(struct request_queue *q, struct bio * raid_bio) 3481static int chunk_aligned_read(struct request_queue *q, struct bio * raid_bio)
3483{ 3482{
3484 mddev_t *mddev = q->queuedata; 3483 mddev_t *mddev = q->queuedata;
3485 raid5_conf_t *conf = mddev_to_conf(mddev); 3484 raid5_conf_t *conf = mddev->private;
3486 unsigned int dd_idx; 3485 unsigned int dd_idx;
3487 struct bio* align_bi; 3486 struct bio* align_bi;
3488 mdk_rdev_t *rdev; 3487 mdk_rdev_t *rdev;
@@ -3599,7 +3598,7 @@ static struct stripe_head *__get_priority_stripe(raid5_conf_t *conf)
3599static int make_request(struct request_queue *q, struct bio * bi) 3598static int make_request(struct request_queue *q, struct bio * bi)
3600{ 3599{
3601 mddev_t *mddev = q->queuedata; 3600 mddev_t *mddev = q->queuedata;
3602 raid5_conf_t *conf = mddev_to_conf(mddev); 3601 raid5_conf_t *conf = mddev->private;
3603 int dd_idx; 3602 int dd_idx;
3604 sector_t new_sector; 3603 sector_t new_sector;
3605 sector_t logical_sector, last_sector; 3604 sector_t logical_sector, last_sector;
@@ -3696,6 +3695,7 @@ static int make_request(struct request_queue *q, struct bio * bi)
3696 spin_unlock_irq(&conf->device_lock); 3695 spin_unlock_irq(&conf->device_lock);
3697 if (must_retry) { 3696 if (must_retry) {
3698 release_stripe(sh); 3697 release_stripe(sh);
3698 schedule();
3699 goto retry; 3699 goto retry;
3700 } 3700 }
3701 } 3701 }
@@ -3791,10 +3791,10 @@ static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped
3791 * If old and new chunk sizes differ, we need to process the 3791 * If old and new chunk sizes differ, we need to process the
3792 * largest of these 3792 * largest of these
3793 */ 3793 */
3794 if (mddev->new_chunk > mddev->chunk_size) 3794 if (mddev->new_chunk_sectors > mddev->chunk_sectors)
3795 reshape_sectors = mddev->new_chunk / 512; 3795 reshape_sectors = mddev->new_chunk_sectors;
3796 else 3796 else
3797 reshape_sectors = mddev->chunk_size / 512; 3797 reshape_sectors = mddev->chunk_sectors;
3798 3798
3799 /* we update the metadata when there is more than 3Meg 3799 /* we update the metadata when there is more than 3Meg
3800 * in the block range (that is rather arbitrary, should 3800 * in the block range (that is rather arbitrary, should
@@ -3917,7 +3917,7 @@ static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped
3917 1, &dd_idx, NULL); 3917 1, &dd_idx, NULL);
3918 last_sector = 3918 last_sector =
3919 raid5_compute_sector(conf, ((stripe_addr+reshape_sectors) 3919 raid5_compute_sector(conf, ((stripe_addr+reshape_sectors)
3920 *(new_data_disks) - 1), 3920 * new_data_disks - 1),
3921 1, &dd_idx, NULL); 3921 1, &dd_idx, NULL);
3922 if (last_sector >= mddev->dev_sectors) 3922 if (last_sector >= mddev->dev_sectors)
3923 last_sector = mddev->dev_sectors - 1; 3923 last_sector = mddev->dev_sectors - 1;
@@ -3946,7 +3946,7 @@ static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped
3946 wait_event(conf->wait_for_overlap, 3946 wait_event(conf->wait_for_overlap,
3947 atomic_read(&conf->reshape_stripes) == 0); 3947 atomic_read(&conf->reshape_stripes) == 0);
3948 mddev->reshape_position = conf->reshape_progress; 3948 mddev->reshape_position = conf->reshape_progress;
3949 mddev->curr_resync_completed = mddev->curr_resync; 3949 mddev->curr_resync_completed = mddev->curr_resync + reshape_sectors;
3950 conf->reshape_checkpoint = jiffies; 3950 conf->reshape_checkpoint = jiffies;
3951 set_bit(MD_CHANGE_DEVS, &mddev->flags); 3951 set_bit(MD_CHANGE_DEVS, &mddev->flags);
3952 md_wakeup_thread(mddev->thread); 3952 md_wakeup_thread(mddev->thread);
@@ -4129,7 +4129,7 @@ static int retry_aligned_read(raid5_conf_t *conf, struct bio *raid_bio)
4129static void raid5d(mddev_t *mddev) 4129static void raid5d(mddev_t *mddev)
4130{ 4130{
4131 struct stripe_head *sh; 4131 struct stripe_head *sh;
4132 raid5_conf_t *conf = mddev_to_conf(mddev); 4132 raid5_conf_t *conf = mddev->private;
4133 int handled; 4133 int handled;
4134 4134
4135 pr_debug("+++ raid5d active\n"); 4135 pr_debug("+++ raid5d active\n");
@@ -4185,7 +4185,7 @@ static void raid5d(mddev_t *mddev)
4185static ssize_t 4185static ssize_t
4186raid5_show_stripe_cache_size(mddev_t *mddev, char *page) 4186raid5_show_stripe_cache_size(mddev_t *mddev, char *page)
4187{ 4187{
4188 raid5_conf_t *conf = mddev_to_conf(mddev); 4188 raid5_conf_t *conf = mddev->private;
4189 if (conf) 4189 if (conf)
4190 return sprintf(page, "%d\n", conf->max_nr_stripes); 4190 return sprintf(page, "%d\n", conf->max_nr_stripes);
4191 else 4191 else
@@ -4195,7 +4195,7 @@ raid5_show_stripe_cache_size(mddev_t *mddev, char *page)
4195static ssize_t 4195static ssize_t
4196raid5_store_stripe_cache_size(mddev_t *mddev, const char *page, size_t len) 4196raid5_store_stripe_cache_size(mddev_t *mddev, const char *page, size_t len)
4197{ 4197{
4198 raid5_conf_t *conf = mddev_to_conf(mddev); 4198 raid5_conf_t *conf = mddev->private;
4199 unsigned long new; 4199 unsigned long new;
4200 int err; 4200 int err;
4201 4201
@@ -4233,7 +4233,7 @@ raid5_stripecache_size = __ATTR(stripe_cache_size, S_IRUGO | S_IWUSR,
4233static ssize_t 4233static ssize_t
4234raid5_show_preread_threshold(mddev_t *mddev, char *page) 4234raid5_show_preread_threshold(mddev_t *mddev, char *page)
4235{ 4235{
4236 raid5_conf_t *conf = mddev_to_conf(mddev); 4236 raid5_conf_t *conf = mddev->private;
4237 if (conf) 4237 if (conf)
4238 return sprintf(page, "%d\n", conf->bypass_threshold); 4238 return sprintf(page, "%d\n", conf->bypass_threshold);
4239 else 4239 else
@@ -4243,7 +4243,7 @@ raid5_show_preread_threshold(mddev_t *mddev, char *page)
4243static ssize_t 4243static ssize_t
4244raid5_store_preread_threshold(mddev_t *mddev, const char *page, size_t len) 4244raid5_store_preread_threshold(mddev_t *mddev, const char *page, size_t len)
4245{ 4245{
4246 raid5_conf_t *conf = mddev_to_conf(mddev); 4246 raid5_conf_t *conf = mddev->private;
4247 unsigned long new; 4247 unsigned long new;
4248 if (len >= PAGE_SIZE) 4248 if (len >= PAGE_SIZE)
4249 return -EINVAL; 4249 return -EINVAL;
@@ -4267,7 +4267,7 @@ raid5_preread_bypass_threshold = __ATTR(preread_bypass_threshold,
4267static ssize_t 4267static ssize_t
4268stripe_cache_active_show(mddev_t *mddev, char *page) 4268stripe_cache_active_show(mddev_t *mddev, char *page)
4269{ 4269{
4270 raid5_conf_t *conf = mddev_to_conf(mddev); 4270 raid5_conf_t *conf = mddev->private;
4271 if (conf) 4271 if (conf)
4272 return sprintf(page, "%d\n", atomic_read(&conf->active_stripes)); 4272 return sprintf(page, "%d\n", atomic_read(&conf->active_stripes));
4273 else 4273 else
@@ -4291,7 +4291,7 @@ static struct attribute_group raid5_attrs_group = {
4291static sector_t 4291static sector_t
4292raid5_size(mddev_t *mddev, sector_t sectors, int raid_disks) 4292raid5_size(mddev_t *mddev, sector_t sectors, int raid_disks)
4293{ 4293{
4294 raid5_conf_t *conf = mddev_to_conf(mddev); 4294 raid5_conf_t *conf = mddev->private;
4295 4295
4296 if (!sectors) 4296 if (!sectors)
4297 sectors = mddev->dev_sectors; 4297 sectors = mddev->dev_sectors;
@@ -4303,8 +4303,8 @@ raid5_size(mddev_t *mddev, sector_t sectors, int raid_disks)
4303 raid_disks = conf->previous_raid_disks; 4303 raid_disks = conf->previous_raid_disks;
4304 } 4304 }
4305 4305
4306 sectors &= ~((sector_t)mddev->chunk_size/512 - 1); 4306 sectors &= ~((sector_t)mddev->chunk_sectors - 1);
4307 sectors &= ~((sector_t)mddev->new_chunk/512 - 1); 4307 sectors &= ~((sector_t)mddev->new_chunk_sectors - 1);
4308 return sectors * (raid_disks - conf->max_degraded); 4308 return sectors * (raid_disks - conf->max_degraded);
4309} 4309}
4310 4310
@@ -4336,9 +4336,11 @@ static raid5_conf_t *setup_conf(mddev_t *mddev)
4336 return ERR_PTR(-EINVAL); 4336 return ERR_PTR(-EINVAL);
4337 } 4337 }
4338 4338
4339 if (!mddev->new_chunk || mddev->new_chunk % PAGE_SIZE) { 4339 if (!mddev->new_chunk_sectors ||
4340 (mddev->new_chunk_sectors << 9) % PAGE_SIZE ||
4341 !is_power_of_2(mddev->new_chunk_sectors)) {
4340 printk(KERN_ERR "raid5: invalid chunk size %d for %s\n", 4342 printk(KERN_ERR "raid5: invalid chunk size %d for %s\n",
4341 mddev->new_chunk, mdname(mddev)); 4343 mddev->new_chunk_sectors << 9, mdname(mddev));
4342 return ERR_PTR(-EINVAL); 4344 return ERR_PTR(-EINVAL);
4343 } 4345 }
4344 4346
@@ -4401,7 +4403,7 @@ static raid5_conf_t *setup_conf(mddev_t *mddev)
4401 conf->fullsync = 1; 4403 conf->fullsync = 1;
4402 } 4404 }
4403 4405
4404 conf->chunk_size = mddev->new_chunk; 4406 conf->chunk_sectors = mddev->new_chunk_sectors;
4405 conf->level = mddev->new_level; 4407 conf->level = mddev->new_level;
4406 if (conf->level == 6) 4408 if (conf->level == 6)
4407 conf->max_degraded = 2; 4409 conf->max_degraded = 2;
@@ -4411,7 +4413,7 @@ static raid5_conf_t *setup_conf(mddev_t *mddev)
4411 conf->max_nr_stripes = NR_STRIPES; 4413 conf->max_nr_stripes = NR_STRIPES;
4412 conf->reshape_progress = mddev->reshape_position; 4414 conf->reshape_progress = mddev->reshape_position;
4413 if (conf->reshape_progress != MaxSector) { 4415 if (conf->reshape_progress != MaxSector) {
4414 conf->prev_chunk = mddev->chunk_size; 4416 conf->prev_chunk_sectors = mddev->chunk_sectors;
4415 conf->prev_algo = mddev->layout; 4417 conf->prev_algo = mddev->layout;
4416 } 4418 }
4417 4419
@@ -4453,6 +4455,10 @@ static int run(mddev_t *mddev)
4453 int working_disks = 0; 4455 int working_disks = 0;
4454 mdk_rdev_t *rdev; 4456 mdk_rdev_t *rdev;
4455 4457
4458 if (mddev->recovery_cp != MaxSector)
4459 printk(KERN_NOTICE "raid5: %s is not clean"
4460 " -- starting background reconstruction\n",
4461 mdname(mddev));
4456 if (mddev->reshape_position != MaxSector) { 4462 if (mddev->reshape_position != MaxSector) {
4457 /* Check that we can continue the reshape. 4463 /* Check that we can continue the reshape.
4458 * Currently only disks can change, it must 4464 * Currently only disks can change, it must
@@ -4475,7 +4481,7 @@ static int run(mddev_t *mddev)
4475 * geometry. 4481 * geometry.
4476 */ 4482 */
4477 here_new = mddev->reshape_position; 4483 here_new = mddev->reshape_position;
4478 if (sector_div(here_new, (mddev->new_chunk>>9)* 4484 if (sector_div(here_new, mddev->new_chunk_sectors *
4479 (mddev->raid_disks - max_degraded))) { 4485 (mddev->raid_disks - max_degraded))) {
4480 printk(KERN_ERR "raid5: reshape_position not " 4486 printk(KERN_ERR "raid5: reshape_position not "
4481 "on a stripe boundary\n"); 4487 "on a stripe boundary\n");
@@ -4483,7 +4489,7 @@ static int run(mddev_t *mddev)
4483 } 4489 }
4484 /* here_new is the stripe we will write to */ 4490 /* here_new is the stripe we will write to */
4485 here_old = mddev->reshape_position; 4491 here_old = mddev->reshape_position;
4486 sector_div(here_old, (mddev->chunk_size>>9)* 4492 sector_div(here_old, mddev->chunk_sectors *
4487 (old_disks-max_degraded)); 4493 (old_disks-max_degraded));
4488 /* here_old is the first stripe that we might need to read 4494 /* here_old is the first stripe that we might need to read
4489 * from */ 4495 * from */
@@ -4498,7 +4504,7 @@ static int run(mddev_t *mddev)
4498 } else { 4504 } else {
4499 BUG_ON(mddev->level != mddev->new_level); 4505 BUG_ON(mddev->level != mddev->new_level);
4500 BUG_ON(mddev->layout != mddev->new_layout); 4506 BUG_ON(mddev->layout != mddev->new_layout);
4501 BUG_ON(mddev->chunk_size != mddev->new_chunk); 4507 BUG_ON(mddev->chunk_sectors != mddev->new_chunk_sectors);
4502 BUG_ON(mddev->delta_disks != 0); 4508 BUG_ON(mddev->delta_disks != 0);
4503 } 4509 }
4504 4510
@@ -4532,7 +4538,7 @@ static int run(mddev_t *mddev)
4532 } 4538 }
4533 4539
4534 /* device size must be a multiple of chunk size */ 4540 /* device size must be a multiple of chunk size */
4535 mddev->dev_sectors &= ~(mddev->chunk_size / 512 - 1); 4541 mddev->dev_sectors &= ~(mddev->chunk_sectors - 1);
4536 mddev->resync_max_sectors = mddev->dev_sectors; 4542 mddev->resync_max_sectors = mddev->dev_sectors;
4537 4543
4538 if (mddev->degraded > 0 && 4544 if (mddev->degraded > 0 &&
@@ -4581,7 +4587,7 @@ static int run(mddev_t *mddev)
4581 { 4587 {
4582 int data_disks = conf->previous_raid_disks - conf->max_degraded; 4588 int data_disks = conf->previous_raid_disks - conf->max_degraded;
4583 int stripe = data_disks * 4589 int stripe = data_disks *
4584 (mddev->chunk_size / PAGE_SIZE); 4590 ((mddev->chunk_sectors << 9) / PAGE_SIZE);
4585 if (mddev->queue->backing_dev_info.ra_pages < 2 * stripe) 4591 if (mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
4586 mddev->queue->backing_dev_info.ra_pages = 2 * stripe; 4592 mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
4587 } 4593 }
@@ -4678,7 +4684,8 @@ static void status(struct seq_file *seq, mddev_t *mddev)
4678 raid5_conf_t *conf = (raid5_conf_t *) mddev->private; 4684 raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
4679 int i; 4685 int i;
4680 4686
4681 seq_printf (seq, " level %d, %dk chunk, algorithm %d", mddev->level, mddev->chunk_size >> 10, mddev->layout); 4687 seq_printf(seq, " level %d, %dk chunk, algorithm %d", mddev->level,
4688 mddev->chunk_sectors / 2, mddev->layout);
4682 seq_printf (seq, " [%d/%d] [", conf->raid_disks, conf->raid_disks - mddev->degraded); 4689 seq_printf (seq, " [%d/%d] [", conf->raid_disks, conf->raid_disks - mddev->degraded);
4683 for (i = 0; i < conf->raid_disks; i++) 4690 for (i = 0; i < conf->raid_disks; i++)
4684 seq_printf (seq, "%s", 4691 seq_printf (seq, "%s",
@@ -4826,7 +4833,7 @@ static int raid5_resize(mddev_t *mddev, sector_t sectors)
4826 * any io in the removed space completes, but it hardly seems 4833 * any io in the removed space completes, but it hardly seems
4827 * worth it. 4834 * worth it.
4828 */ 4835 */
4829 sectors &= ~((sector_t)mddev->chunk_size/512 - 1); 4836 sectors &= ~((sector_t)mddev->chunk_sectors - 1);
4830 md_set_array_sectors(mddev, raid5_size(mddev, sectors, 4837 md_set_array_sectors(mddev, raid5_size(mddev, sectors,
4831 mddev->raid_disks)); 4838 mddev->raid_disks));
4832 if (mddev->array_sectors > 4839 if (mddev->array_sectors >
@@ -4843,14 +4850,37 @@ static int raid5_resize(mddev_t *mddev, sector_t sectors)
4843 return 0; 4850 return 0;
4844} 4851}
4845 4852
4846static int raid5_check_reshape(mddev_t *mddev) 4853static int check_stripe_cache(mddev_t *mddev)
4847{ 4854{
4848 raid5_conf_t *conf = mddev_to_conf(mddev); 4855 /* Can only proceed if there are plenty of stripe_heads.
4856 * We need a minimum of one full stripe,, and for sensible progress
4857 * it is best to have about 4 times that.
4858 * If we require 4 times, then the default 256 4K stripe_heads will
4859 * allow for chunk sizes up to 256K, which is probably OK.
4860 * If the chunk size is greater, user-space should request more
4861 * stripe_heads first.
4862 */
4863 raid5_conf_t *conf = mddev->private;
4864 if (((mddev->chunk_sectors << 9) / STRIPE_SIZE) * 4
4865 > conf->max_nr_stripes ||
4866 ((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4
4867 > conf->max_nr_stripes) {
4868 printk(KERN_WARNING "raid5: reshape: not enough stripes. Needed %lu\n",
4869 ((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
4870 / STRIPE_SIZE)*4);
4871 return 0;
4872 }
4873 return 1;
4874}
4875
4876static int check_reshape(mddev_t *mddev)
4877{
4878 raid5_conf_t *conf = mddev->private;
4849 4879
4850 if (mddev->delta_disks == 0 && 4880 if (mddev->delta_disks == 0 &&
4851 mddev->new_layout == mddev->layout && 4881 mddev->new_layout == mddev->layout &&
4852 mddev->new_chunk == mddev->chunk_size) 4882 mddev->new_chunk_sectors == mddev->chunk_sectors)
4853 return -EINVAL; /* nothing to do */ 4883 return 0; /* nothing to do */
4854 if (mddev->bitmap) 4884 if (mddev->bitmap)
4855 /* Cannot grow a bitmap yet */ 4885 /* Cannot grow a bitmap yet */
4856 return -EBUSY; 4886 return -EBUSY;
@@ -4869,28 +4899,15 @@ static int raid5_check_reshape(mddev_t *mddev)
4869 return -EINVAL; 4899 return -EINVAL;
4870 } 4900 }
4871 4901
4872 /* Can only proceed if there are plenty of stripe_heads. 4902 if (!check_stripe_cache(mddev))
4873 * We need a minimum of one full stripe,, and for sensible progress
4874 * it is best to have about 4 times that.
4875 * If we require 4 times, then the default 256 4K stripe_heads will
4876 * allow for chunk sizes up to 256K, which is probably OK.
4877 * If the chunk size is greater, user-space should request more
4878 * stripe_heads first.
4879 */
4880 if ((mddev->chunk_size / STRIPE_SIZE) * 4 > conf->max_nr_stripes ||
4881 (mddev->new_chunk / STRIPE_SIZE) * 4 > conf->max_nr_stripes) {
4882 printk(KERN_WARNING "raid5: reshape: not enough stripes. Needed %lu\n",
4883 (max(mddev->chunk_size, mddev->new_chunk)
4884 / STRIPE_SIZE)*4);
4885 return -ENOSPC; 4903 return -ENOSPC;
4886 }
4887 4904
4888 return resize_stripes(conf, conf->raid_disks + mddev->delta_disks); 4905 return resize_stripes(conf, conf->raid_disks + mddev->delta_disks);
4889} 4906}
4890 4907
4891static int raid5_start_reshape(mddev_t *mddev) 4908static int raid5_start_reshape(mddev_t *mddev)
4892{ 4909{
4893 raid5_conf_t *conf = mddev_to_conf(mddev); 4910 raid5_conf_t *conf = mddev->private;
4894 mdk_rdev_t *rdev; 4911 mdk_rdev_t *rdev;
4895 int spares = 0; 4912 int spares = 0;
4896 int added_devices = 0; 4913 int added_devices = 0;
@@ -4899,6 +4916,9 @@ static int raid5_start_reshape(mddev_t *mddev)
4899 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) 4916 if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
4900 return -EBUSY; 4917 return -EBUSY;
4901 4918
4919 if (!check_stripe_cache(mddev))
4920 return -ENOSPC;
4921
4902 list_for_each_entry(rdev, &mddev->disks, same_set) 4922 list_for_each_entry(rdev, &mddev->disks, same_set)
4903 if (rdev->raid_disk < 0 && 4923 if (rdev->raid_disk < 0 &&
4904 !test_bit(Faulty, &rdev->flags)) 4924 !test_bit(Faulty, &rdev->flags))
@@ -4925,8 +4945,8 @@ static int raid5_start_reshape(mddev_t *mddev)
4925 spin_lock_irq(&conf->device_lock); 4945 spin_lock_irq(&conf->device_lock);
4926 conf->previous_raid_disks = conf->raid_disks; 4946 conf->previous_raid_disks = conf->raid_disks;
4927 conf->raid_disks += mddev->delta_disks; 4947 conf->raid_disks += mddev->delta_disks;
4928 conf->prev_chunk = conf->chunk_size; 4948 conf->prev_chunk_sectors = conf->chunk_sectors;
4929 conf->chunk_size = mddev->new_chunk; 4949 conf->chunk_sectors = mddev->new_chunk_sectors;
4930 conf->prev_algo = conf->algorithm; 4950 conf->prev_algo = conf->algorithm;
4931 conf->algorithm = mddev->new_layout; 4951 conf->algorithm = mddev->new_layout;
4932 if (mddev->delta_disks < 0) 4952 if (mddev->delta_disks < 0)
@@ -5008,7 +5028,7 @@ static void end_reshape(raid5_conf_t *conf)
5008 */ 5028 */
5009 { 5029 {
5010 int data_disks = conf->raid_disks - conf->max_degraded; 5030 int data_disks = conf->raid_disks - conf->max_degraded;
5011 int stripe = data_disks * (conf->chunk_size 5031 int stripe = data_disks * ((conf->chunk_sectors << 9)
5012 / PAGE_SIZE); 5032 / PAGE_SIZE);
5013 if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe) 5033 if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
5014 conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe; 5034 conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
@@ -5022,7 +5042,7 @@ static void end_reshape(raid5_conf_t *conf)
5022static void raid5_finish_reshape(mddev_t *mddev) 5042static void raid5_finish_reshape(mddev_t *mddev)
5023{ 5043{
5024 struct block_device *bdev; 5044 struct block_device *bdev;
5025 raid5_conf_t *conf = mddev_to_conf(mddev); 5045 raid5_conf_t *conf = mddev->private;
5026 5046
5027 if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery)) { 5047 if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
5028 5048
@@ -5053,7 +5073,7 @@ static void raid5_finish_reshape(mddev_t *mddev)
5053 raid5_remove_disk(mddev, d); 5073 raid5_remove_disk(mddev, d);
5054 } 5074 }
5055 mddev->layout = conf->algorithm; 5075 mddev->layout = conf->algorithm;
5056 mddev->chunk_size = conf->chunk_size; 5076 mddev->chunk_sectors = conf->chunk_sectors;
5057 mddev->reshape_position = MaxSector; 5077 mddev->reshape_position = MaxSector;
5058 mddev->delta_disks = 0; 5078 mddev->delta_disks = 0;
5059 } 5079 }
@@ -5061,7 +5081,7 @@ static void raid5_finish_reshape(mddev_t *mddev)
5061 5081
5062static void raid5_quiesce(mddev_t *mddev, int state) 5082static void raid5_quiesce(mddev_t *mddev, int state)
5063{ 5083{
5064 raid5_conf_t *conf = mddev_to_conf(mddev); 5084 raid5_conf_t *conf = mddev->private;
5065 5085
5066 switch(state) { 5086 switch(state) {
5067 case 2: /* resume for a suspend */ 5087 case 2: /* resume for a suspend */
@@ -5111,7 +5131,7 @@ static void *raid5_takeover_raid1(mddev_t *mddev)
5111 5131
5112 mddev->new_level = 5; 5132 mddev->new_level = 5;
5113 mddev->new_layout = ALGORITHM_LEFT_SYMMETRIC; 5133 mddev->new_layout = ALGORITHM_LEFT_SYMMETRIC;
5114 mddev->new_chunk = chunksect << 9; 5134 mddev->new_chunk_sectors = chunksect;
5115 5135
5116 return setup_conf(mddev); 5136 return setup_conf(mddev);
5117} 5137}
@@ -5150,24 +5170,24 @@ static void *raid5_takeover_raid6(mddev_t *mddev)
5150} 5170}
5151 5171
5152 5172
5153static int raid5_reconfig(mddev_t *mddev, int new_layout, int new_chunk) 5173static int raid5_check_reshape(mddev_t *mddev)
5154{ 5174{
5155 /* For a 2-drive array, the layout and chunk size can be changed 5175 /* For a 2-drive array, the layout and chunk size can be changed
5156 * immediately as not restriping is needed. 5176 * immediately as not restriping is needed.
5157 * For larger arrays we record the new value - after validation 5177 * For larger arrays we record the new value - after validation
5158 * to be used by a reshape pass. 5178 * to be used by a reshape pass.
5159 */ 5179 */
5160 raid5_conf_t *conf = mddev_to_conf(mddev); 5180 raid5_conf_t *conf = mddev->private;
5181 int new_chunk = mddev->new_chunk_sectors;
5161 5182
5162 if (new_layout >= 0 && !algorithm_valid_raid5(new_layout)) 5183 if (mddev->new_layout >= 0 && !algorithm_valid_raid5(mddev->new_layout))
5163 return -EINVAL; 5184 return -EINVAL;
5164 if (new_chunk > 0) { 5185 if (new_chunk > 0) {
5165 if (new_chunk & (new_chunk-1)) 5186 if (!is_power_of_2(new_chunk))
5166 /* not a power of 2 */
5167 return -EINVAL; 5187 return -EINVAL;
5168 if (new_chunk < PAGE_SIZE) 5188 if (new_chunk < (PAGE_SIZE>>9))
5169 return -EINVAL; 5189 return -EINVAL;
5170 if (mddev->array_sectors & ((new_chunk>>9)-1)) 5190 if (mddev->array_sectors & (new_chunk-1))
5171 /* not factor of array size */ 5191 /* not factor of array size */
5172 return -EINVAL; 5192 return -EINVAL;
5173 } 5193 }
@@ -5175,49 +5195,39 @@ static int raid5_reconfig(mddev_t *mddev, int new_layout, int new_chunk)
5175 /* They look valid */ 5195 /* They look valid */
5176 5196
5177 if (mddev->raid_disks == 2) { 5197 if (mddev->raid_disks == 2) {
5178 5198 /* can make the change immediately */
5179 if (new_layout >= 0) { 5199 if (mddev->new_layout >= 0) {
5180 conf->algorithm = new_layout; 5200 conf->algorithm = mddev->new_layout;
5181 mddev->layout = mddev->new_layout = new_layout; 5201 mddev->layout = mddev->new_layout;
5182 } 5202 }
5183 if (new_chunk > 0) { 5203 if (new_chunk > 0) {
5184 conf->chunk_size = new_chunk; 5204 conf->chunk_sectors = new_chunk ;
5185 mddev->chunk_size = mddev->new_chunk = new_chunk; 5205 mddev->chunk_sectors = new_chunk;
5186 } 5206 }
5187 set_bit(MD_CHANGE_DEVS, &mddev->flags); 5207 set_bit(MD_CHANGE_DEVS, &mddev->flags);
5188 md_wakeup_thread(mddev->thread); 5208 md_wakeup_thread(mddev->thread);
5189 } else {
5190 if (new_layout >= 0)
5191 mddev->new_layout = new_layout;
5192 if (new_chunk > 0)
5193 mddev->new_chunk = new_chunk;
5194 } 5209 }
5195 return 0; 5210 return check_reshape(mddev);
5196} 5211}
5197 5212
5198static int raid6_reconfig(mddev_t *mddev, int new_layout, int new_chunk) 5213static int raid6_check_reshape(mddev_t *mddev)
5199{ 5214{
5200 if (new_layout >= 0 && !algorithm_valid_raid6(new_layout)) 5215 int new_chunk = mddev->new_chunk_sectors;
5216
5217 if (mddev->new_layout >= 0 && !algorithm_valid_raid6(mddev->new_layout))
5201 return -EINVAL; 5218 return -EINVAL;
5202 if (new_chunk > 0) { 5219 if (new_chunk > 0) {
5203 if (new_chunk & (new_chunk-1)) 5220 if (!is_power_of_2(new_chunk))
5204 /* not a power of 2 */
5205 return -EINVAL; 5221 return -EINVAL;
5206 if (new_chunk < PAGE_SIZE) 5222 if (new_chunk < (PAGE_SIZE >> 9))
5207 return -EINVAL; 5223 return -EINVAL;
5208 if (mddev->array_sectors & ((new_chunk>>9)-1)) 5224 if (mddev->array_sectors & (new_chunk-1))
5209 /* not factor of array size */ 5225 /* not factor of array size */
5210 return -EINVAL; 5226 return -EINVAL;
5211 } 5227 }
5212 5228
5213 /* They look valid */ 5229 /* They look valid */
5214 5230 return check_reshape(mddev);
5215 if (new_layout >= 0)
5216 mddev->new_layout = new_layout;
5217 if (new_chunk > 0)
5218 mddev->new_chunk = new_chunk;
5219
5220 return 0;
5221} 5231}
5222 5232
5223static void *raid5_takeover(mddev_t *mddev) 5233static void *raid5_takeover(mddev_t *mddev)
@@ -5227,8 +5237,6 @@ static void *raid5_takeover(mddev_t *mddev)
5227 * raid1 - if there are two drives. We need to know the chunk size 5237 * raid1 - if there are two drives. We need to know the chunk size
5228 * raid4 - trivial - just use a raid4 layout. 5238 * raid4 - trivial - just use a raid4 layout.
5229 * raid6 - Providing it is a *_6 layout 5239 * raid6 - Providing it is a *_6 layout
5230 *
5231 * For now, just do raid1
5232 */ 5240 */
5233 5241
5234 if (mddev->level == 1) 5242 if (mddev->level == 1)
@@ -5310,12 +5318,11 @@ static struct mdk_personality raid6_personality =
5310 .sync_request = sync_request, 5318 .sync_request = sync_request,
5311 .resize = raid5_resize, 5319 .resize = raid5_resize,
5312 .size = raid5_size, 5320 .size = raid5_size,
5313 .check_reshape = raid5_check_reshape, 5321 .check_reshape = raid6_check_reshape,
5314 .start_reshape = raid5_start_reshape, 5322 .start_reshape = raid5_start_reshape,
5315 .finish_reshape = raid5_finish_reshape, 5323 .finish_reshape = raid5_finish_reshape,
5316 .quiesce = raid5_quiesce, 5324 .quiesce = raid5_quiesce,
5317 .takeover = raid6_takeover, 5325 .takeover = raid6_takeover,
5318 .reconfig = raid6_reconfig,
5319}; 5326};
5320static struct mdk_personality raid5_personality = 5327static struct mdk_personality raid5_personality =
5321{ 5328{
@@ -5338,7 +5345,6 @@ static struct mdk_personality raid5_personality =
5338 .finish_reshape = raid5_finish_reshape, 5345 .finish_reshape = raid5_finish_reshape,
5339 .quiesce = raid5_quiesce, 5346 .quiesce = raid5_quiesce,
5340 .takeover = raid5_takeover, 5347 .takeover = raid5_takeover,
5341 .reconfig = raid5_reconfig,
5342}; 5348};
5343 5349
5344static struct mdk_personality raid4_personality = 5350static struct mdk_personality raid4_personality =
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
index 52ba99954dec..9459689c4ea0 100644
--- a/drivers/md/raid5.h
+++ b/drivers/md/raid5.h
@@ -334,7 +334,8 @@ struct raid5_private_data {
334 struct hlist_head *stripe_hashtbl; 334 struct hlist_head *stripe_hashtbl;
335 mddev_t *mddev; 335 mddev_t *mddev;
336 struct disk_info *spare; 336 struct disk_info *spare;
337 int chunk_size, level, algorithm; 337 int chunk_sectors;
338 int level, algorithm;
338 int max_degraded; 339 int max_degraded;
339 int raid_disks; 340 int raid_disks;
340 int max_nr_stripes; 341 int max_nr_stripes;
@@ -350,7 +351,8 @@ struct raid5_private_data {
350 */ 351 */
351 sector_t reshape_safe; 352 sector_t reshape_safe;
352 int previous_raid_disks; 353 int previous_raid_disks;
353 int prev_chunk, prev_algo; 354 int prev_chunk_sectors;
355 int prev_algo;
354 short generation; /* increments with every reshape */ 356 short generation; /* increments with every reshape */
355 unsigned long reshape_checkpoint; /* Time we last updated 357 unsigned long reshape_checkpoint; /* Time we last updated
356 * metadata */ 358 * metadata */
@@ -408,8 +410,6 @@ struct raid5_private_data {
408 410
409typedef struct raid5_private_data raid5_conf_t; 411typedef struct raid5_private_data raid5_conf_t;
410 412
411#define mddev_to_conf(mddev) ((raid5_conf_t *) mddev->private)
412
413/* 413/*
414 * Our supported algorithms 414 * Our supported algorithms
415 */ 415 */
diff --git a/include/linux/raid/md_p.h b/include/linux/raid/md_p.h
index 6ba830fa8538..ffa2efbbe382 100644
--- a/include/linux/raid/md_p.h
+++ b/include/linux/raid/md_p.h
@@ -232,7 +232,7 @@ struct mdp_superblock_1 {
232 __le64 reshape_position; /* next address in array-space for reshape */ 232 __le64 reshape_position; /* next address in array-space for reshape */
233 __le32 delta_disks; /* change in number of raid_disks */ 233 __le32 delta_disks; /* change in number of raid_disks */
234 __le32 new_layout; /* new layout */ 234 __le32 new_layout; /* new layout */
235 __le32 new_chunk; /* new chunk size (bytes) */ 235 __le32 new_chunk; /* new chunk size (512byte sectors) */
236 __u8 pad1[128-124]; /* set to 0 when written */ 236 __u8 pad1[128-124]; /* set to 0 when written */
237 237
238 /* constant this-device information - 64 bytes */ 238 /* constant this-device information - 64 bytes */
generated by cgit v1.2.2 (git 2.25.0) at 2025-12-30 15:24:55 -0500