blob: cba8128953186edd304430c2cbf8b219693789b1 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 raid0.c : Multiple Devices driver for Linux
NeilBrownf72ffdd2014-09-30 14:23:59 +10003 Copyright (C) 1994-96 Marc ZYNGIER
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 <zyngier@ufr-info-p7.ibp.fr> or
5 <maz@gloups.fdn.fr>
NeilBrownf72ffdd2014-09-30 14:23:59 +10006 Copyright (C) 1999, 2000 Ingo Molnar, Red Hat
Linus Torvalds1da177e2005-04-16 15:20:36 -07007
8 RAID-0 management functions.
9
10 This program is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 2, or (at your option)
13 any later version.
NeilBrownf72ffdd2014-09-30 14:23:59 +100014
Linus Torvalds1da177e2005-04-16 15:20:36 -070015 You should have received a copy of the GNU General Public License
16 (for example /usr/src/linux/COPYING); if not, write to the Free
NeilBrownf72ffdd2014-09-30 14:23:59 +100017 Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
Linus Torvalds1da177e2005-04-16 15:20:36 -070018*/
19
NeilBrownbff61972009-03-31 14:33:13 +110020#include <linux/blkdev.h>
NeilBrownbff61972009-03-31 14:33:13 +110021#include <linux/seq_file.h>
Paul Gortmaker056075c2011-07-03 13:58:33 -040022#include <linux/module.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090023#include <linux/slab.h>
NeilBrown109e3762016-11-18 13:22:04 +110024#include <trace/events/block.h>
NeilBrown43b2e5d2009-03-31 14:33:13 +110025#include "md.h"
Christoph Hellwigef740c32009-03-31 14:27:03 +110026#include "raid0.h"
Trela, Maciej9af204c2010-03-08 16:02:44 +110027#include "raid5.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070028
NeilBrownca39e8182019-09-09 16:30:02 +100029static int default_layout = 0;
30module_param(default_layout, int, 0644);
31
Shaohua Li394ed8e2017-01-04 16:10:19 -080032#define UNSUPPORTED_MDDEV_FLAGS \
33 ((1L << MD_HAS_JOURNAL) | \
34 (1L << MD_JOURNAL_CLEAN) | \
Artur Paszkiewiczea0213e2017-03-09 09:59:57 +010035 (1L << MD_FAILFAST_SUPPORTED) |\
Pawel Baldysiakddc08822017-08-16 17:13:45 +020036 (1L << MD_HAS_PPL) | \
37 (1L << MD_HAS_MULTIPLE_PPLS))
Shaohua Li394ed8e2017-01-04 16:10:19 -080038
NeilBrown5c675f82014-12-15 12:56:56 +110039static int raid0_congested(struct mddev *mddev, int bits)
NeilBrown26be34d2006-10-03 01:15:53 -070040{
NeilBrowne373ab12011-10-11 16:48:59 +110041 struct r0conf *conf = mddev->private;
NeilBrown3cb03002011-10-11 16:45:26 +110042 struct md_rdev **devlist = conf->devlist;
NeilBrown84707f32010-03-16 17:23:35 +110043 int raid_disks = conf->strip_zone[0].nb_dev;
NeilBrown26be34d2006-10-03 01:15:53 -070044 int i, ret = 0;
45
NeilBrown84707f32010-03-16 17:23:35 +110046 for (i = 0; i < raid_disks && !ret ; i++) {
Jens Axboe165125e2007-07-24 09:28:11 +020047 struct request_queue *q = bdev_get_queue(devlist[i]->bdev);
NeilBrown26be34d2006-10-03 01:15:53 -070048
Jan Karadc3b17c2017-02-02 15:56:50 +010049 ret |= bdi_congested(q->backing_dev_info, bits);
NeilBrown26be34d2006-10-03 01:15:53 -070050 }
51 return ret;
52}
53
raz ben yehuda46994192009-06-16 17:00:54 +100054/*
55 * inform the user of the raid configuration
56*/
NeilBrownfd01b882011-10-11 16:47:53 +110057static void dump_zones(struct mddev *mddev)
raz ben yehuda46994192009-06-16 17:00:54 +100058{
NeilBrown50de8df2011-10-07 14:23:22 +110059 int j, k;
raz ben yehuda46994192009-06-16 17:00:54 +100060 sector_t zone_size = 0;
61 sector_t zone_start = 0;
62 char b[BDEVNAME_SIZE];
NeilBrowne373ab12011-10-11 16:48:59 +110063 struct r0conf *conf = mddev->private;
NeilBrown84707f32010-03-16 17:23:35 +110064 int raid_disks = conf->strip_zone[0].nb_dev;
NeilBrown76603882016-11-02 14:16:50 +110065 pr_debug("md: RAID0 configuration for %s - %d zone%s\n",
66 mdname(mddev),
67 conf->nr_strip_zones, conf->nr_strip_zones==1?"":"s");
raz ben yehuda46994192009-06-16 17:00:54 +100068 for (j = 0; j < conf->nr_strip_zones; j++) {
NeilBrown76603882016-11-02 14:16:50 +110069 char line[200];
70 int len = 0;
71
raz ben yehuda46994192009-06-16 17:00:54 +100072 for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
Saurabh Sengar3b0a2bd2022-08-23 11:51:04 -070073 len += scnprintf(line+len, 200-len, "%s%s", k?"/":"",
74 bdevname(conf->devlist[j*raid_disks
NeilBrown76603882016-11-02 14:16:50 +110075 + k]->bdev, b));
76 pr_debug("md: zone%d=[%s]\n", j, line);
raz ben yehuda46994192009-06-16 17:00:54 +100077
78 zone_size = conf->strip_zone[j].zone_end - zone_start;
NeilBrown76603882016-11-02 14:16:50 +110079 pr_debug(" zone-offset=%10lluKB, device-offset=%10lluKB, size=%10lluKB\n",
raz ben yehuda46994192009-06-16 17:00:54 +100080 (unsigned long long)zone_start>>1,
81 (unsigned long long)conf->strip_zone[j].dev_start>>1,
82 (unsigned long long)zone_size>>1);
83 zone_start = conf->strip_zone[j].zone_end;
84 }
raz ben yehuda46994192009-06-16 17:00:54 +100085}
86
NeilBrowne373ab12011-10-11 16:48:59 +110087static int create_strip_zones(struct mddev *mddev, struct r0conf **private_conf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070088{
NeilBrowna9f326e2009-09-23 18:06:41 +100089 int i, c, err;
NeilBrown49f357a22009-06-16 16:50:35 +100090 sector_t curr_zone_end, sectors;
NeilBrown3cb03002011-10-11 16:45:26 +110091 struct md_rdev *smallest, *rdev1, *rdev2, *rdev, **dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 struct strip_zone *zone;
93 int cnt;
94 char b[BDEVNAME_SIZE];
NeilBrown50de8df2011-10-07 14:23:22 +110095 char b2[BDEVNAME_SIZE];
NeilBrowne373ab12011-10-11 16:48:59 +110096 struct r0conf *conf = kzalloc(sizeof(*conf), GFP_KERNEL);
Mikulas Patocka0c7a7d82020-01-15 08:35:25 -050097 unsigned blksize = 512;
Andre Nolled7b0032009-06-16 16:47:36 +100098
Dan Carpenter7dedd152016-04-14 12:31:49 +030099 *private_conf = ERR_PTR(-ENOMEM);
Andre Nolled7b0032009-06-16 16:47:36 +1000100 if (!conf)
101 return -ENOMEM;
NeilBrowndafb20f2012-03-19 12:46:39 +1100102 rdev_for_each(rdev1, mddev) {
NeilBrown50de8df2011-10-07 14:23:22 +1100103 pr_debug("md/raid0:%s: looking at %s\n",
104 mdname(mddev),
105 bdevname(rdev1->bdev, b));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 c = 0;
NeilBrown13f26822009-06-18 08:48:55 +1000107
108 /* round size to chunk_size */
109 sectors = rdev1->sectors;
110 sector_div(sectors, mddev->chunk_sectors);
111 rdev1->sectors = sectors * mddev->chunk_sectors;
112
NeilBrown199dc6e2015-08-03 13:11:47 +1000113 blksize = max(blksize, queue_logical_block_size(
114 rdev1->bdev->bd_disk->queue));
115
NeilBrowndafb20f2012-03-19 12:46:39 +1100116 rdev_for_each(rdev2, mddev) {
NeilBrown50de8df2011-10-07 14:23:22 +1100117 pr_debug("md/raid0:%s: comparing %s(%llu)"
118 " with %s(%llu)\n",
119 mdname(mddev),
120 bdevname(rdev1->bdev,b),
121 (unsigned long long)rdev1->sectors,
122 bdevname(rdev2->bdev,b2),
123 (unsigned long long)rdev2->sectors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124 if (rdev2 == rdev1) {
NeilBrown50de8df2011-10-07 14:23:22 +1100125 pr_debug("md/raid0:%s: END\n",
126 mdname(mddev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127 break;
128 }
Andre Nolldd8ac332009-03-31 14:33:13 +1100129 if (rdev2->sectors == rdev1->sectors) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 /*
131 * Not unique, don't count it as a new
132 * group
133 */
NeilBrown50de8df2011-10-07 14:23:22 +1100134 pr_debug("md/raid0:%s: EQUAL\n",
135 mdname(mddev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 c = 1;
137 break;
138 }
NeilBrown50de8df2011-10-07 14:23:22 +1100139 pr_debug("md/raid0:%s: NOT EQUAL\n",
140 mdname(mddev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 }
142 if (!c) {
NeilBrown50de8df2011-10-07 14:23:22 +1100143 pr_debug("md/raid0:%s: ==> UNIQUE\n",
144 mdname(mddev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 conf->nr_strip_zones++;
NeilBrown50de8df2011-10-07 14:23:22 +1100146 pr_debug("md/raid0:%s: %d zones\n",
147 mdname(mddev), conf->nr_strip_zones);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148 }
149 }
NeilBrown50de8df2011-10-07 14:23:22 +1100150 pr_debug("md/raid0:%s: FINAL %d zones\n",
151 mdname(mddev), conf->nr_strip_zones);
NeilBrownca39e8182019-09-09 16:30:02 +1000152
NeilBrown199dc6e2015-08-03 13:11:47 +1000153 /*
154 * now since we have the hard sector sizes, we can make sure
155 * chunk size is a multiple of that sector size
156 */
157 if ((mddev->chunk_sectors << 9) % blksize) {
NeilBrown76603882016-11-02 14:16:50 +1100158 pr_warn("md/raid0:%s: chunk_size of %d not multiple of block size %d\n",
159 mdname(mddev),
160 mddev->chunk_sectors << 9, blksize);
NeilBrown199dc6e2015-08-03 13:11:47 +1000161 err = -EINVAL;
162 goto abort;
163 }
164
Andre Nolled7b0032009-06-16 16:47:36 +1000165 err = -ENOMEM;
NeilBrown9ffae0c2006-01-06 00:20:32 -0800166 conf->strip_zone = kzalloc(sizeof(struct strip_zone)*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167 conf->nr_strip_zones, GFP_KERNEL);
168 if (!conf->strip_zone)
Andre Nolled7b0032009-06-16 16:47:36 +1000169 goto abort;
NeilBrown3cb03002011-10-11 16:45:26 +1100170 conf->devlist = kzalloc(sizeof(struct md_rdev*)*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 conf->nr_strip_zones*mddev->raid_disks,
172 GFP_KERNEL);
173 if (!conf->devlist)
Andre Nolled7b0032009-06-16 16:47:36 +1000174 goto abort;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 /* The first zone must contain all devices, so here we check that
177 * there is a proper alignment of slots to devices and find them all
178 */
179 zone = &conf->strip_zone[0];
180 cnt = 0;
181 smallest = NULL;
NeilBrownb4145792009-06-16 16:50:52 +1000182 dev = conf->devlist;
Andre Nolled7b0032009-06-16 16:47:36 +1000183 err = -EINVAL;
NeilBrowndafb20f2012-03-19 12:46:39 +1100184 rdev_for_each(rdev1, mddev) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 int j = rdev1->raid_disk;
186
NeilBrowne93f68a2010-06-15 09:36:03 +0100187 if (mddev->level == 10) {
Trela, Maciej9af204c2010-03-08 16:02:44 +1100188 /* taking over a raid10-n2 array */
189 j /= 2;
NeilBrowne93f68a2010-06-15 09:36:03 +0100190 rdev1->new_raid_disk = j;
191 }
Trela, Maciej9af204c2010-03-08 16:02:44 +1100192
Krzysztof Wojcikfc3a08b2011-01-31 13:47:13 +1100193 if (mddev->level == 1) {
194 /* taiking over a raid1 array-
195 * we have only one active disk
196 */
197 j = 0;
198 rdev1->new_raid_disk = j;
199 }
200
NeilBrownf96c9f32013-02-21 15:50:07 +1100201 if (j < 0) {
NeilBrown76603882016-11-02 14:16:50 +1100202 pr_warn("md/raid0:%s: remove inactive devices before converting to RAID0\n",
203 mdname(mddev));
NeilBrownf96c9f32013-02-21 15:50:07 +1100204 goto abort;
205 }
206 if (j >= mddev->raid_disks) {
NeilBrown76603882016-11-02 14:16:50 +1100207 pr_warn("md/raid0:%s: bad disk number %d - aborting!\n",
208 mdname(mddev), j);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 goto abort;
210 }
NeilBrownb4145792009-06-16 16:50:52 +1000211 if (dev[j]) {
NeilBrown76603882016-11-02 14:16:50 +1100212 pr_warn("md/raid0:%s: multiple devices for %d - aborting!\n",
213 mdname(mddev), j);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 goto abort;
215 }
NeilBrownb4145792009-06-16 16:50:52 +1000216 dev[j] = rdev1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217
Andre Nolldd8ac332009-03-31 14:33:13 +1100218 if (!smallest || (rdev1->sectors < smallest->sectors))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 smallest = rdev1;
220 cnt++;
221 }
222 if (cnt != mddev->raid_disks) {
NeilBrown76603882016-11-02 14:16:50 +1100223 pr_warn("md/raid0:%s: too few disks (%d of %d) - aborting!\n",
224 mdname(mddev), cnt, mddev->raid_disks);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 goto abort;
226 }
227 zone->nb_dev = cnt;
NeilBrown49f357a22009-06-16 16:50:35 +1000228 zone->zone_end = smallest->sectors * cnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229
NeilBrown49f357a22009-06-16 16:50:35 +1000230 curr_zone_end = zone->zone_end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231
232 /* now do the other zones */
233 for (i = 1; i < conf->nr_strip_zones; i++)
234 {
NeilBrowna9f326e2009-09-23 18:06:41 +1000235 int j;
236
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 zone = conf->strip_zone + i;
NeilBrownb4145792009-06-16 16:50:52 +1000238 dev = conf->devlist + i * mddev->raid_disks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
NeilBrown50de8df2011-10-07 14:23:22 +1100240 pr_debug("md/raid0:%s: zone %d\n", mdname(mddev), i);
NeilBrownd27a43ab2009-06-16 16:46:46 +1000241 zone->dev_start = smallest->sectors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 smallest = NULL;
243 c = 0;
244
245 for (j=0; j<cnt; j++) {
NeilBrownb4145792009-06-16 16:50:52 +1000246 rdev = conf->devlist[j];
NeilBrownd27a43ab2009-06-16 16:46:46 +1000247 if (rdev->sectors <= zone->dev_start) {
NeilBrown50de8df2011-10-07 14:23:22 +1100248 pr_debug("md/raid0:%s: checking %s ... nope\n",
249 mdname(mddev),
250 bdevname(rdev->bdev, b));
Andre Nolldd8ac332009-03-31 14:33:13 +1100251 continue;
252 }
NeilBrown50de8df2011-10-07 14:23:22 +1100253 pr_debug("md/raid0:%s: checking %s ..."
254 " contained as device %d\n",
255 mdname(mddev),
256 bdevname(rdev->bdev, b), c);
NeilBrownb4145792009-06-16 16:50:52 +1000257 dev[c] = rdev;
Andre Nolldd8ac332009-03-31 14:33:13 +1100258 c++;
259 if (!smallest || rdev->sectors < smallest->sectors) {
260 smallest = rdev;
NeilBrown50de8df2011-10-07 14:23:22 +1100261 pr_debug("md/raid0:%s: (%llu) is smallest!.\n",
262 mdname(mddev),
263 (unsigned long long)rdev->sectors);
Andre Nolldd8ac332009-03-31 14:33:13 +1100264 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 }
266
267 zone->nb_dev = c;
NeilBrown49f357a22009-06-16 16:50:35 +1000268 sectors = (smallest->sectors - zone->dev_start) * c;
NeilBrown50de8df2011-10-07 14:23:22 +1100269 pr_debug("md/raid0:%s: zone->nb_dev: %d, sectors: %llu\n",
270 mdname(mddev),
271 zone->nb_dev, (unsigned long long)sectors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272
NeilBrown49f357a22009-06-16 16:50:35 +1000273 curr_zone_end += sectors;
NeilBrownd27a43ab2009-06-16 16:46:46 +1000274 zone->zone_end = curr_zone_end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275
NeilBrown50de8df2011-10-07 14:23:22 +1100276 pr_debug("md/raid0:%s: current zone start: %llu\n",
277 mdname(mddev),
278 (unsigned long long)smallest->sectors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280
Pascal Hambourgc2e66812022-04-13 08:53:56 +0200281 if (conf->nr_strip_zones == 1 || conf->strip_zone[1].nb_dev == 1) {
282 conf->layout = RAID0_ORIG_LAYOUT;
283 } else if (mddev->layout == RAID0_ORIG_LAYOUT ||
284 mddev->layout == RAID0_ALT_MULTIZONE_LAYOUT) {
285 conf->layout = mddev->layout;
286 } else if (default_layout == RAID0_ORIG_LAYOUT ||
287 default_layout == RAID0_ALT_MULTIZONE_LAYOUT) {
288 conf->layout = default_layout;
289 } else {
290 pr_err("md/raid0:%s: cannot assemble multi-zone RAID0 with default_layout setting\n",
291 mdname(mddev));
292 pr_err("md/raid0: please set raid0.default_layout to 1 or 2\n");
293 err = -EOPNOTSUPP;
294 goto abort;
295 }
296
Jason Baron7bf29ed2023-06-23 14:05:23 -0400297 if (conf->layout == RAID0_ORIG_LAYOUT) {
298 for (i = 1; i < conf->nr_strip_zones; i++) {
299 sector_t first_sector = conf->strip_zone[i-1].zone_end;
300
301 sector_div(first_sector, mddev->chunk_sectors);
302 zone = conf->strip_zone + i;
303 /* disk_shift is first disk index used in the zone */
304 zone->disk_shift = sector_div(first_sector,
305 zone->nb_dev);
306 }
307 }
308
NeilBrown50de8df2011-10-07 14:23:22 +1100309 pr_debug("md/raid0:%s: done.\n", mdname(mddev));
Trela, Maciej9af204c2010-03-08 16:02:44 +1100310 *private_conf = conf;
311
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 return 0;
Andre Noll5568a602009-06-16 16:47:21 +1000313abort:
Andre Nolled7b0032009-06-16 16:47:36 +1000314 kfree(conf->strip_zone);
315 kfree(conf->devlist);
316 kfree(conf);
NeilBrown58ebb342013-02-21 15:36:38 +1100317 *private_conf = ERR_PTR(err);
Andre Nolled7b0032009-06-16 16:47:36 +1000318 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319}
320
NeilBrownba13da42012-03-19 12:46:39 +1100321/* Find the zone which holds a particular offset
322 * Update *sectorp to be an offset in that zone
323 */
324static struct strip_zone *find_zone(struct r0conf *conf,
325 sector_t *sectorp)
326{
327 int i;
328 struct strip_zone *z = conf->strip_zone;
329 sector_t sector = *sectorp;
330
331 for (i = 0; i < conf->nr_strip_zones; i++)
332 if (sector < z[i].zone_end) {
333 if (i)
334 *sectorp = sector - z[i-1].zone_end;
335 return z + i;
336 }
337 BUG();
338}
339
340/*
341 * remaps the bio to the target device. we separate two flows.
NeilBrown47d68972015-04-10 13:19:04 +1000342 * power 2 flow and a general flow for the sake of performance
NeilBrownba13da42012-03-19 12:46:39 +1100343*/
344static struct md_rdev *map_sector(struct mddev *mddev, struct strip_zone *zone,
345 sector_t sector, sector_t *sector_offset)
346{
347 unsigned int sect_in_chunk;
348 sector_t chunk;
349 struct r0conf *conf = mddev->private;
350 int raid_disks = conf->strip_zone[0].nb_dev;
351 unsigned int chunk_sects = mddev->chunk_sectors;
352
353 if (is_power_of_2(chunk_sects)) {
354 int chunksect_bits = ffz(~chunk_sects);
355 /* find the sector offset inside the chunk */
356 sect_in_chunk = sector & (chunk_sects - 1);
357 sector >>= chunksect_bits;
358 /* chunk in zone */
359 chunk = *sector_offset;
360 /* quotient is the chunk in real device*/
361 sector_div(chunk, zone->nb_dev << chunksect_bits);
362 } else{
363 sect_in_chunk = sector_div(sector, chunk_sects);
364 chunk = *sector_offset;
365 sector_div(chunk, chunk_sects * zone->nb_dev);
366 }
367 /*
368 * position the bio over the real device
369 * real sector = chunk in device + starting of zone
370 * + the position in the chunk
371 */
372 *sector_offset = (chunk * chunk_sects) + sect_in_chunk;
373 return conf->devlist[(zone - conf->strip_zone)*raid_disks
374 + sector_div(sector, zone->nb_dev)];
375}
376
NeilBrownfd01b882011-10-11 16:47:53 +1100377static sector_t raid0_size(struct mddev *mddev, sector_t sectors, int raid_disks)
Dan Williams80c3a6c2009-03-17 18:10:40 -0700378{
379 sector_t array_sectors = 0;
NeilBrown3cb03002011-10-11 16:45:26 +1100380 struct md_rdev *rdev;
Dan Williams80c3a6c2009-03-17 18:10:40 -0700381
382 WARN_ONCE(sectors || raid_disks,
383 "%s does not support generic reshape\n", __func__);
384
NeilBrowndafb20f2012-03-19 12:46:39 +1100385 rdev_for_each(rdev, mddev)
NeilBrowna6468532013-02-21 14:33:17 +1100386 array_sectors += (rdev->sectors &
387 ~(sector_t)(mddev->chunk_sectors-1));
Dan Williams80c3a6c2009-03-17 18:10:40 -0700388
389 return array_sectors;
390}
391
NeilBrownafa0f552014-12-15 12:56:58 +1100392static void raid0_free(struct mddev *mddev, void *priv);
majianpeng0366ef82012-04-02 09:48:37 +1000393
NeilBrownfd01b882011-10-11 16:47:53 +1100394static int raid0_run(struct mddev *mddev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395{
NeilBrowne373ab12011-10-11 16:48:59 +1100396 struct r0conf *conf;
Andre Noll5568a602009-06-16 16:47:21 +1000397 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398
Andre Noll9d8f0362009-06-18 08:45:01 +1000399 if (mddev->chunk_sectors == 0) {
NeilBrown76603882016-11-02 14:16:50 +1100400 pr_warn("md/raid0:%s: chunk size must be set.\n", mdname(mddev));
NeilBrown2604b702006-01-06 00:20:36 -0800401 return -EINVAL;
402 }
Andre Noll0894cc32009-06-18 08:49:23 +1000403 if (md_check_no_bitmap(mddev))
404 return -EINVAL;
Heinz Mauelshagen753f2852015-02-13 19:48:01 +0100405
Trela, Maciej9af204c2010-03-08 16:02:44 +1100406 /* if private is not null, we are here after takeover */
407 if (mddev->private == NULL) {
408 ret = create_strip_zones(mddev, &conf);
409 if (ret < 0)
410 return ret;
411 mddev->private = conf;
412 }
413 conf = mddev->private;
NeilBrown199dc6e2015-08-03 13:11:47 +1000414 if (mddev->queue) {
415 struct md_rdev *rdev;
416 bool discard_supported = false;
417
NeilBrown199dc6e2015-08-03 13:11:47 +1000418 blk_queue_max_hw_sectors(mddev->queue, mddev->chunk_sectors);
419 blk_queue_max_write_same_sectors(mddev->queue, mddev->chunk_sectors);
Christoph Hellwig3deff1a2017-04-05 19:21:03 +0200420 blk_queue_max_write_zeroes_sectors(mddev->queue, mddev->chunk_sectors);
Shaohua Li29efc392017-05-07 17:36:24 -0700421 blk_queue_max_discard_sectors(mddev->queue, UINT_MAX);
NeilBrown199dc6e2015-08-03 13:11:47 +1000422
423 blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9);
424 blk_queue_io_opt(mddev->queue,
425 (mddev->chunk_sectors << 9) * mddev->raid_disks);
426
NeilBrown66eefe52015-09-24 15:47:47 +1000427 rdev_for_each(rdev, mddev) {
428 disk_stack_limits(mddev->gendisk, rdev->bdev,
429 rdev->data_offset << 9);
430 if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
431 discard_supported = true;
432 }
NeilBrown199dc6e2015-08-03 13:11:47 +1000433 if (!discard_supported)
434 queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
435 else
436 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
437 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438
439 /* calculate array device size */
Dan Williams1f403622009-03-31 14:59:03 +1100440 md_set_array_sectors(mddev, raid0_size(mddev, 0, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441
NeilBrown76603882016-11-02 14:16:50 +1100442 pr_debug("md/raid0:%s: md_size is %llu sectors.\n",
443 mdname(mddev),
444 (unsigned long long)mddev->array_sectors);
Heinz Mauelshagen753f2852015-02-13 19:48:01 +0100445
446 if (mddev->queue) {
447 /* calculate the max read-ahead size.
448 * For read-ahead of large files to be effective, we need to
449 * readahead at least twice a whole stripe. i.e. number of devices
450 * multiplied by chunk size times 2.
451 * If an individual device has an ra_pages greater than the
452 * chunk size, then we will not drive that device as hard as it
453 * wants. We consider this a configuration error: a larger
454 * chunksize should be used in that case.
455 */
Andre Noll9d8f0362009-06-18 08:45:01 +1000456 int stripe = mddev->raid_disks *
457 (mddev->chunk_sectors << 9) / PAGE_SIZE;
Jan Karadc3b17c2017-02-02 15:56:50 +0100458 if (mddev->queue->backing_dev_info->ra_pages < 2* stripe)
459 mddev->queue->backing_dev_info->ra_pages = 2* stripe;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 }
461
raz ben yehuda46994192009-06-16 17:00:54 +1000462 dump_zones(mddev);
majianpeng0366ef82012-04-02 09:48:37 +1000463
464 ret = md_integrity_register(mddev);
majianpeng0366ef82012-04-02 09:48:37 +1000465
466 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467}
468
NeilBrownafa0f552014-12-15 12:56:58 +1100469static void raid0_free(struct mddev *mddev, void *priv)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470{
NeilBrownafa0f552014-12-15 12:56:58 +1100471 struct r0conf *conf = priv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472
Jesper Juhl990a8ba2005-06-21 17:17:30 -0700473 kfree(conf->strip_zone);
Andre Nollfb5ab4b2009-06-16 16:48:19 +1000474 kfree(conf->devlist);
Jesper Juhl990a8ba2005-06-21 17:17:30 -0700475 kfree(conf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476}
477
raz ben yehudafbb704e2009-06-16 17:02:05 +1000478/*
479 * Is io distribute over 1 or more chunks ?
480*/
NeilBrownfd01b882011-10-11 16:47:53 +1100481static inline int is_io_in_chunk_boundary(struct mddev *mddev,
raz ben yehudafbb704e2009-06-16 17:02:05 +1000482 unsigned int chunk_sects, struct bio *bio)
483{
NeilBrownd6e412e2009-06-18 08:47:00 +1000484 if (likely(is_power_of_2(chunk_sects))) {
Kent Overstreet4f024f32013-10-11 15:44:27 -0700485 return chunk_sects >=
486 ((bio->bi_iter.bi_sector & (chunk_sects-1))
Kent Overstreetaa8b57a2013-02-05 15:19:29 -0800487 + bio_sectors(bio));
raz ben yehudafbb704e2009-06-16 17:02:05 +1000488 } else{
Kent Overstreet4f024f32013-10-11 15:44:27 -0700489 sector_t sector = bio->bi_iter.bi_sector;
raz ben yehudafbb704e2009-06-16 17:02:05 +1000490 return chunk_sects >= (sector_div(sector, chunk_sects)
Kent Overstreetaa8b57a2013-02-05 15:19:29 -0800491 + bio_sectors(bio));
raz ben yehudafbb704e2009-06-16 17:02:05 +1000492 }
493}
494
Jason Baron7bf29ed2023-06-23 14:05:23 -0400495/*
496 * Convert disk_index to the disk order in which it is read/written.
497 * For example, if we have 4 disks, they are numbered 0,1,2,3. If we
498 * write the disks starting at disk 3, then the read/write order would
499 * be disk 3, then 0, then 1, and then disk 2 and we want map_disk_shift()
500 * to map the disks as follows 0,1,2,3 => 1,2,3,0. So disk 0 would map
501 * to 1, 1 to 2, 2 to 3, and 3 to 0. That way we can compare disks in
502 * that 'output' space to understand the read/write disk ordering.
503 */
504static int map_disk_shift(int disk_index, int num_disks, int disk_shift)
505{
506 return ((disk_index + num_disks - disk_shift) % num_disks);
507}
508
Shaohua Li29efc392017-05-07 17:36:24 -0700509static void raid0_handle_discard(struct mddev *mddev, struct bio *bio)
510{
511 struct r0conf *conf = mddev->private;
512 struct strip_zone *zone;
513 sector_t start = bio->bi_iter.bi_sector;
514 sector_t end;
515 unsigned int stripe_size;
516 sector_t first_stripe_index, last_stripe_index;
517 sector_t start_disk_offset;
518 unsigned int start_disk_index;
519 sector_t end_disk_offset;
520 unsigned int end_disk_index;
521 unsigned int disk;
Jason Baron7bf29ed2023-06-23 14:05:23 -0400522 sector_t orig_start, orig_end;
Shaohua Li29efc392017-05-07 17:36:24 -0700523
Jason Baron7bf29ed2023-06-23 14:05:23 -0400524 orig_start = start;
Shaohua Li29efc392017-05-07 17:36:24 -0700525 zone = find_zone(conf, &start);
526
527 if (bio_end_sector(bio) > zone->zone_end) {
528 struct bio *split = bio_split(bio,
529 zone->zone_end - bio->bi_iter.bi_sector, GFP_NOIO,
530 mddev->bio_set);
531 bio_chain(split, bio);
532 generic_make_request(bio);
533 bio = split;
534 end = zone->zone_end;
535 } else
536 end = bio_end_sector(bio);
537
Jason Baron7bf29ed2023-06-23 14:05:23 -0400538 orig_end = end;
Shaohua Li29efc392017-05-07 17:36:24 -0700539 if (zone != conf->strip_zone)
540 end = end - zone[-1].zone_end;
541
542 /* Now start and end is the offset in zone */
543 stripe_size = zone->nb_dev * mddev->chunk_sectors;
544
545 first_stripe_index = start;
546 sector_div(first_stripe_index, stripe_size);
547 last_stripe_index = end;
548 sector_div(last_stripe_index, stripe_size);
549
Jason Baron7bf29ed2023-06-23 14:05:23 -0400550 /* In the first zone the original and alternate layouts are the same */
551 if ((conf->layout == RAID0_ORIG_LAYOUT) && (zone != conf->strip_zone)) {
552 sector_div(orig_start, mddev->chunk_sectors);
553 start_disk_index = sector_div(orig_start, zone->nb_dev);
554 start_disk_index = map_disk_shift(start_disk_index,
555 zone->nb_dev,
556 zone->disk_shift);
557 sector_div(orig_end, mddev->chunk_sectors);
558 end_disk_index = sector_div(orig_end, zone->nb_dev);
559 end_disk_index = map_disk_shift(end_disk_index,
560 zone->nb_dev, zone->disk_shift);
561 } else {
562 start_disk_index = (int)(start - first_stripe_index * stripe_size) /
563 mddev->chunk_sectors;
564 end_disk_index = (int)(end - last_stripe_index * stripe_size) /
565 mddev->chunk_sectors;
566 }
Shaohua Li29efc392017-05-07 17:36:24 -0700567 start_disk_offset = ((int)(start - first_stripe_index * stripe_size) %
568 mddev->chunk_sectors) +
569 first_stripe_index * mddev->chunk_sectors;
Shaohua Li29efc392017-05-07 17:36:24 -0700570 end_disk_offset = ((int)(end - last_stripe_index * stripe_size) %
571 mddev->chunk_sectors) +
572 last_stripe_index * mddev->chunk_sectors;
573
574 for (disk = 0; disk < zone->nb_dev; disk++) {
575 sector_t dev_start, dev_end;
576 struct bio *discard_bio = NULL;
577 struct md_rdev *rdev;
Jason Baron7bf29ed2023-06-23 14:05:23 -0400578 int compare_disk;
Shaohua Li29efc392017-05-07 17:36:24 -0700579
Jason Baron7bf29ed2023-06-23 14:05:23 -0400580 compare_disk = map_disk_shift(disk, zone->nb_dev,
581 zone->disk_shift);
582
583 if (compare_disk < start_disk_index)
Shaohua Li29efc392017-05-07 17:36:24 -0700584 dev_start = (first_stripe_index + 1) *
585 mddev->chunk_sectors;
Jason Baron7bf29ed2023-06-23 14:05:23 -0400586 else if (compare_disk > start_disk_index)
Shaohua Li29efc392017-05-07 17:36:24 -0700587 dev_start = first_stripe_index * mddev->chunk_sectors;
588 else
589 dev_start = start_disk_offset;
590
Jason Baron7bf29ed2023-06-23 14:05:23 -0400591 if (compare_disk < end_disk_index)
Shaohua Li29efc392017-05-07 17:36:24 -0700592 dev_end = (last_stripe_index + 1) * mddev->chunk_sectors;
Jason Baron7bf29ed2023-06-23 14:05:23 -0400593 else if (compare_disk > end_disk_index)
Shaohua Li29efc392017-05-07 17:36:24 -0700594 dev_end = last_stripe_index * mddev->chunk_sectors;
595 else
596 dev_end = end_disk_offset;
597
598 if (dev_end <= dev_start)
599 continue;
600
601 rdev = conf->devlist[(zone - conf->strip_zone) *
602 conf->strip_zone[0].nb_dev + disk];
603 if (__blkdev_issue_discard(rdev->bdev,
604 dev_start + zone->dev_start + rdev->data_offset,
605 dev_end - dev_start, GFP_NOIO, 0, &discard_bio) ||
606 !discard_bio)
607 continue;
608 bio_chain(discard_bio, bio);
Shaohua Li8a8e6f82017-08-18 10:27:59 -0700609 bio_clone_blkcg_association(discard_bio, bio);
Shaohua Li29efc392017-05-07 17:36:24 -0700610 if (mddev->gendisk)
611 trace_block_bio_remap(bdev_get_queue(rdev->bdev),
612 discard_bio, disk_devt(mddev->gendisk),
613 bio->bi_iter.bi_sector);
614 generic_make_request(discard_bio);
615 }
616 bio_endio(bio);
617}
618
NeilBrowncc27b0c2017-06-05 16:49:39 +1000619static bool raid0_make_request(struct mddev *mddev, struct bio *bio)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620{
NeilBrownca39e8182019-09-09 16:30:02 +1000621 struct r0conf *conf = mddev->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 struct strip_zone *zone;
NeilBrown3cb03002011-10-11 16:45:26 +1100623 struct md_rdev *tmp_dev;
NeilBrownf00d7c82017-04-05 14:05:51 +1000624 sector_t bio_sector;
625 sector_t sector;
NeilBrownca39e8182019-09-09 16:30:02 +1000626 sector_t orig_sector;
NeilBrownf00d7c82017-04-05 14:05:51 +1000627 unsigned chunk_sects;
628 unsigned sectors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629
Jens Axboe1eff9d32016-08-05 15:35:16 -0600630 if (unlikely(bio->bi_opf & REQ_PREFLUSH)) {
Tejun Heoe9c74692010-09-03 11:56:18 +0200631 md_flush_request(mddev, bio);
NeilBrowncc27b0c2017-06-05 16:49:39 +1000632 return true;
NeilBrowne5dcdd82005-09-09 16:23:41 -0700633 }
634
Shaohua Li29efc392017-05-07 17:36:24 -0700635 if (unlikely((bio_op(bio) == REQ_OP_DISCARD))) {
636 raid0_handle_discard(mddev, bio);
NeilBrowncc27b0c2017-06-05 16:49:39 +1000637 return true;
Shaohua Li29efc392017-05-07 17:36:24 -0700638 }
639
NeilBrownf00d7c82017-04-05 14:05:51 +1000640 bio_sector = bio->bi_iter.bi_sector;
641 sector = bio_sector;
642 chunk_sects = mddev->chunk_sectors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643
NeilBrownf00d7c82017-04-05 14:05:51 +1000644 sectors = chunk_sects -
645 (likely(is_power_of_2(chunk_sects))
646 ? (sector & (chunk_sects-1))
647 : sector_div(sector, chunk_sects));
Christoph Hellwig5a7bbad2011-09-12 12:12:01 +0200648
NeilBrownf00d7c82017-04-05 14:05:51 +1000649 /* Restore due to sector_div */
650 sector = bio_sector;
Eric Worka8115772015-05-18 23:26:23 -0700651
NeilBrownf00d7c82017-04-05 14:05:51 +1000652 if (sectors < bio_sectors(bio)) {
653 struct bio *split = bio_split(bio, sectors, GFP_NOIO, mddev->bio_set);
654 bio_chain(split, bio);
655 generic_make_request(bio);
656 bio = split;
657 }
Shaohua Lic83057a2012-10-11 13:25:44 +1100658
NeilBrownca39e8182019-09-09 16:30:02 +1000659 orig_sector = sector;
NeilBrownf00d7c82017-04-05 14:05:51 +1000660 zone = find_zone(mddev->private, &sector);
NeilBrownca39e8182019-09-09 16:30:02 +1000661 switch (conf->layout) {
662 case RAID0_ORIG_LAYOUT:
663 tmp_dev = map_sector(mddev, zone, orig_sector, &sector);
664 break;
665 case RAID0_ALT_MULTIZONE_LAYOUT:
666 tmp_dev = map_sector(mddev, zone, sector, &sector);
667 break;
668 default:
Dan Carpenter8f873b82019-09-21 09:00:31 +0300669 WARN(1, "md/raid0:%s: Invalid layout\n", mdname(mddev));
NeilBrownca39e8182019-09-09 16:30:02 +1000670 bio_io_error(bio);
671 return true;
672 }
673
Christoph Hellwig74d46992017-08-23 19:10:32 +0200674 bio_set_dev(bio, tmp_dev->bdev);
NeilBrownf00d7c82017-04-05 14:05:51 +1000675 bio->bi_iter.bi_sector = sector + zone->dev_start +
676 tmp_dev->data_offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
Shaohua Li29efc392017-05-07 17:36:24 -0700678 if (mddev->gendisk)
Christoph Hellwig74d46992017-08-23 19:10:32 +0200679 trace_block_bio_remap(bio->bi_disk->queue, bio,
680 disk_devt(mddev->gendisk), bio_sector);
Shaohua Li29efc392017-05-07 17:36:24 -0700681 mddev_check_writesame(mddev, bio);
682 mddev_check_write_zeroes(mddev, bio);
683 generic_make_request(bio);
NeilBrowncc27b0c2017-06-05 16:49:39 +1000684 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685}
NeilBrown8299d7f2007-10-16 23:30:53 -0700686
NeilBrownfd01b882011-10-11 16:47:53 +1100687static void raid0_status(struct seq_file *seq, struct mddev *mddev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688{
Andre Noll9d8f0362009-06-18 08:45:01 +1000689 seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690 return;
691}
692
NeilBrownfd01b882011-10-11 16:47:53 +1100693static void *raid0_takeover_raid45(struct mddev *mddev)
Trela, Maciej9af204c2010-03-08 16:02:44 +1100694{
NeilBrown3cb03002011-10-11 16:45:26 +1100695 struct md_rdev *rdev;
NeilBrowne373ab12011-10-11 16:48:59 +1100696 struct r0conf *priv_conf;
Trela, Maciej9af204c2010-03-08 16:02:44 +1100697
698 if (mddev->degraded != 1) {
NeilBrown76603882016-11-02 14:16:50 +1100699 pr_warn("md/raid0:%s: raid5 must be degraded! Degraded disks: %d\n",
700 mdname(mddev),
701 mddev->degraded);
Trela, Maciej9af204c2010-03-08 16:02:44 +1100702 return ERR_PTR(-EINVAL);
703 }
704
NeilBrowndafb20f2012-03-19 12:46:39 +1100705 rdev_for_each(rdev, mddev) {
Trela, Maciej9af204c2010-03-08 16:02:44 +1100706 /* check slot number for a disk */
707 if (rdev->raid_disk == mddev->raid_disks-1) {
NeilBrown76603882016-11-02 14:16:50 +1100708 pr_warn("md/raid0:%s: raid5 must have missing parity disk!\n",
709 mdname(mddev));
Trela, Maciej9af204c2010-03-08 16:02:44 +1100710 return ERR_PTR(-EINVAL);
711 }
NeilBrowneea136d2013-06-26 11:55:20 +1000712 rdev->sectors = mddev->dev_sectors;
Trela, Maciej9af204c2010-03-08 16:02:44 +1100713 }
714
715 /* Set new parameters */
716 mddev->new_level = 0;
Maciej Trela001048a2010-06-16 11:55:14 +0100717 mddev->new_layout = 0;
Trela, Maciej9af204c2010-03-08 16:02:44 +1100718 mddev->new_chunk_sectors = mddev->chunk_sectors;
719 mddev->raid_disks--;
720 mddev->delta_disks = -1;
721 /* make sure it will be not marked as dirty */
722 mddev->recovery_cp = MaxSector;
Shaohua Li394ed8e2017-01-04 16:10:19 -0800723 mddev_clear_unsupported_flags(mddev, UNSUPPORTED_MDDEV_FLAGS);
Trela, Maciej9af204c2010-03-08 16:02:44 +1100724
725 create_strip_zones(mddev, &priv_conf);
Shaohua Li6995f0b2016-12-08 15:48:17 -0800726
Trela, Maciej9af204c2010-03-08 16:02:44 +1100727 return priv_conf;
728}
729
NeilBrownfd01b882011-10-11 16:47:53 +1100730static void *raid0_takeover_raid10(struct mddev *mddev)
Trela, Maciej9af204c2010-03-08 16:02:44 +1100731{
NeilBrowne373ab12011-10-11 16:48:59 +1100732 struct r0conf *priv_conf;
Trela, Maciej9af204c2010-03-08 16:02:44 +1100733
734 /* Check layout:
735 * - far_copies must be 1
736 * - near_copies must be 2
737 * - disks number must be even
738 * - all mirrors must be already degraded
739 */
740 if (mddev->layout != ((1 << 8) + 2)) {
NeilBrown76603882016-11-02 14:16:50 +1100741 pr_warn("md/raid0:%s:: Raid0 cannot takeover layout: 0x%x\n",
742 mdname(mddev),
743 mddev->layout);
Trela, Maciej9af204c2010-03-08 16:02:44 +1100744 return ERR_PTR(-EINVAL);
745 }
746 if (mddev->raid_disks & 1) {
NeilBrown76603882016-11-02 14:16:50 +1100747 pr_warn("md/raid0:%s: Raid0 cannot takeover Raid10 with odd disk number.\n",
748 mdname(mddev));
Trela, Maciej9af204c2010-03-08 16:02:44 +1100749 return ERR_PTR(-EINVAL);
750 }
751 if (mddev->degraded != (mddev->raid_disks>>1)) {
NeilBrown76603882016-11-02 14:16:50 +1100752 pr_warn("md/raid0:%s: All mirrors must be already degraded!\n",
753 mdname(mddev));
Trela, Maciej9af204c2010-03-08 16:02:44 +1100754 return ERR_PTR(-EINVAL);
755 }
756
757 /* Set new parameters */
758 mddev->new_level = 0;
Maciej Trela001048a2010-06-16 11:55:14 +0100759 mddev->new_layout = 0;
Trela, Maciej9af204c2010-03-08 16:02:44 +1100760 mddev->new_chunk_sectors = mddev->chunk_sectors;
761 mddev->delta_disks = - mddev->raid_disks / 2;
762 mddev->raid_disks += mddev->delta_disks;
763 mddev->degraded = 0;
764 /* make sure it will be not marked as dirty */
765 mddev->recovery_cp = MaxSector;
Shaohua Li394ed8e2017-01-04 16:10:19 -0800766 mddev_clear_unsupported_flags(mddev, UNSUPPORTED_MDDEV_FLAGS);
Trela, Maciej9af204c2010-03-08 16:02:44 +1100767
768 create_strip_zones(mddev, &priv_conf);
Trela, Maciej9af204c2010-03-08 16:02:44 +1100769 return priv_conf;
770}
771
NeilBrownfd01b882011-10-11 16:47:53 +1100772static void *raid0_takeover_raid1(struct mddev *mddev)
Krzysztof Wojcikfc3a08b2011-01-31 13:47:13 +1100773{
NeilBrowne373ab12011-10-11 16:48:59 +1100774 struct r0conf *priv_conf;
Jes Sorensen24b961f2012-04-01 23:48:38 +1000775 int chunksect;
Krzysztof Wojcikfc3a08b2011-01-31 13:47:13 +1100776
777 /* Check layout:
778 * - (N - 1) mirror drives must be already faulty
779 */
780 if ((mddev->raid_disks - 1) != mddev->degraded) {
NeilBrown76603882016-11-02 14:16:50 +1100781 pr_err("md/raid0:%s: (N - 1) mirrors drives must be already faulty!\n",
Krzysztof Wojcikfc3a08b2011-01-31 13:47:13 +1100782 mdname(mddev));
783 return ERR_PTR(-EINVAL);
784 }
785
Jes Sorensen24b961f2012-04-01 23:48:38 +1000786 /*
787 * a raid1 doesn't have the notion of chunk size, so
788 * figure out the largest suitable size we can use.
789 */
790 chunksect = 64 * 2; /* 64K by default */
791
792 /* The array must be an exact multiple of chunksize */
793 while (chunksect && (mddev->array_sectors & (chunksect - 1)))
794 chunksect >>= 1;
795
796 if ((chunksect << 9) < PAGE_SIZE)
797 /* array size does not allow a suitable chunk size */
798 return ERR_PTR(-EINVAL);
799
Krzysztof Wojcikfc3a08b2011-01-31 13:47:13 +1100800 /* Set new parameters */
801 mddev->new_level = 0;
802 mddev->new_layout = 0;
Jes Sorensen24b961f2012-04-01 23:48:38 +1000803 mddev->new_chunk_sectors = chunksect;
804 mddev->chunk_sectors = chunksect;
Krzysztof Wojcikfc3a08b2011-01-31 13:47:13 +1100805 mddev->delta_disks = 1 - mddev->raid_disks;
Krzysztof Wojcikf7bee802011-02-14 10:01:41 +1100806 mddev->raid_disks = 1;
Krzysztof Wojcikfc3a08b2011-01-31 13:47:13 +1100807 /* make sure it will be not marked as dirty */
808 mddev->recovery_cp = MaxSector;
Shaohua Li394ed8e2017-01-04 16:10:19 -0800809 mddev_clear_unsupported_flags(mddev, UNSUPPORTED_MDDEV_FLAGS);
Krzysztof Wojcikfc3a08b2011-01-31 13:47:13 +1100810
811 create_strip_zones(mddev, &priv_conf);
812 return priv_conf;
813}
814
NeilBrownfd01b882011-10-11 16:47:53 +1100815static void *raid0_takeover(struct mddev *mddev)
Trela, Maciej9af204c2010-03-08 16:02:44 +1100816{
817 /* raid0 can take over:
Maciej Trela049d6c12010-06-16 11:56:12 +0100818 * raid4 - if all data disks are active.
Trela, Maciej9af204c2010-03-08 16:02:44 +1100819 * raid5 - providing it is Raid4 layout and one disk is faulty
820 * raid10 - assuming we have all necessary active disks
Krzysztof Wojcikfc3a08b2011-01-31 13:47:13 +1100821 * raid1 - with (N -1) mirror drives faulty
Trela, Maciej9af204c2010-03-08 16:02:44 +1100822 */
NeilBrowna8461a62014-08-06 16:34:27 +1000823
824 if (mddev->bitmap) {
NeilBrown76603882016-11-02 14:16:50 +1100825 pr_warn("md/raid0: %s: cannot takeover array with bitmap\n",
826 mdname(mddev));
NeilBrowna8461a62014-08-06 16:34:27 +1000827 return ERR_PTR(-EBUSY);
828 }
Maciej Trela049d6c12010-06-16 11:56:12 +0100829 if (mddev->level == 4)
830 return raid0_takeover_raid45(mddev);
831
Trela, Maciej9af204c2010-03-08 16:02:44 +1100832 if (mddev->level == 5) {
833 if (mddev->layout == ALGORITHM_PARITY_N)
Maciej Trela049d6c12010-06-16 11:56:12 +0100834 return raid0_takeover_raid45(mddev);
Trela, Maciej9af204c2010-03-08 16:02:44 +1100835
NeilBrown76603882016-11-02 14:16:50 +1100836 pr_warn("md/raid0:%s: Raid can only takeover Raid5 with layout: %d\n",
837 mdname(mddev), ALGORITHM_PARITY_N);
Trela, Maciej9af204c2010-03-08 16:02:44 +1100838 }
839
840 if (mddev->level == 10)
841 return raid0_takeover_raid10(mddev);
842
Krzysztof Wojcikfc3a08b2011-01-31 13:47:13 +1100843 if (mddev->level == 1)
844 return raid0_takeover_raid1(mddev);
845
NeilBrown76603882016-11-02 14:16:50 +1100846 pr_warn("Takeover from raid%i to raid0 not supported\n",
Krzysztof Wojcikfc3a08b2011-01-31 13:47:13 +1100847 mddev->level);
848
Trela, Maciej9af204c2010-03-08 16:02:44 +1100849 return ERR_PTR(-EINVAL);
850}
851
NeilBrown2fc45ef2017-10-19 12:49:15 +1100852static void raid0_quiesce(struct mddev *mddev, int quiesce)
Trela, Maciej9af204c2010-03-08 16:02:44 +1100853{
854}
855
NeilBrown84fc4b52011-10-11 16:49:58 +1100856static struct md_personality raid0_personality=
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857{
858 .name = "raid0",
NeilBrown2604b702006-01-06 00:20:36 -0800859 .level = 0,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 .owner = THIS_MODULE,
861 .make_request = raid0_make_request,
862 .run = raid0_run,
NeilBrownafa0f552014-12-15 12:56:58 +1100863 .free = raid0_free,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 .status = raid0_status,
Dan Williams80c3a6c2009-03-17 18:10:40 -0700865 .size = raid0_size,
Trela, Maciej9af204c2010-03-08 16:02:44 +1100866 .takeover = raid0_takeover,
867 .quiesce = raid0_quiesce,
NeilBrown5c675f82014-12-15 12:56:56 +1100868 .congested = raid0_congested,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869};
870
871static int __init raid0_init (void)
872{
NeilBrown2604b702006-01-06 00:20:36 -0800873 return register_md_personality (&raid0_personality);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874}
875
876static void raid0_exit (void)
877{
NeilBrown2604b702006-01-06 00:20:36 -0800878 unregister_md_personality (&raid0_personality);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879}
880
881module_init(raid0_init);
882module_exit(raid0_exit);
883MODULE_LICENSE("GPL");
NeilBrown0efb9e62009-12-14 12:49:58 +1100884MODULE_DESCRIPTION("RAID0 (striping) personality for MD");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885MODULE_ALIAS("md-personality-2"); /* RAID0 */
NeilBrownd9d166c2006-01-06 00:20:51 -0800886MODULE_ALIAS("md-raid0");
NeilBrown2604b702006-01-06 00:20:36 -0800887MODULE_ALIAS("md-level-0");