source: mainline/uspace/srv/bd/hr/raid0.c@ 372a9fc

Last change on this file since 372a9fc was 50603405, checked in by Miroslav Cimerman <mc@…>, 7 months ago

hr: metadata format agnostic superblock ops

Put metadata specific code behind a new hr_superblock_ops_t
interface, that allows to easily add support for new metadata
formats.
  • Property mode set to 100644
File size: 9.1 KB
Line 
1/*
2 * Copyright (c) 2025 Miroslav Cimerman
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 *
9 * - Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * - Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * - The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29/** @addtogroup hr
30 * @{
31 */
32/**
33 * @file
34 */
35
36#include <abi/ipc/ipc.h>
37#include <bd_srv.h>
38#include <block.h>
39#include <errno.h>
40#include <hr.h>
41#include <io/log.h>
42#include <ipc/hr.h>
43#include <ipc/services.h>
44#include <loc.h>
45#include <task.h>
46#include <stdio.h>
47#include <stdlib.h>
48#include <str_error.h>
49
50#include "io.h"
51#include "superblock.h"
52#include "util.h"
53#include "var.h"
54
55static void hr_raid0_update_vol_status(hr_volume_t *);
56static void hr_raid0_state_callback(hr_volume_t *, size_t, errno_t);
57static errno_t hr_raid0_bd_op(hr_bd_op_type_t, bd_srv_t *, aoff64_t, size_t,
58 void *, const void *, size_t);
59
60/* bdops */
61static errno_t hr_raid0_bd_open(bd_srvs_t *, bd_srv_t *);
62static errno_t hr_raid0_bd_close(bd_srv_t *);
63static errno_t hr_raid0_bd_read_blocks(bd_srv_t *, aoff64_t, size_t, void *,
64 size_t);
65static errno_t hr_raid0_bd_sync_cache(bd_srv_t *, aoff64_t, size_t);
66static errno_t hr_raid0_bd_write_blocks(bd_srv_t *, aoff64_t, size_t,
67 const void *, size_t);
68static errno_t hr_raid0_bd_get_block_size(bd_srv_t *, size_t *);
69static errno_t hr_raid0_bd_get_num_blocks(bd_srv_t *, aoff64_t *);
70
71static bd_ops_t hr_raid0_bd_ops = {
72 .open = hr_raid0_bd_open,
73 .close = hr_raid0_bd_close,
74 .sync_cache = hr_raid0_bd_sync_cache,
75 .read_blocks = hr_raid0_bd_read_blocks,
76 .write_blocks = hr_raid0_bd_write_blocks,
77 .get_block_size = hr_raid0_bd_get_block_size,
78 .get_num_blocks = hr_raid0_bd_get_num_blocks
79};
80
81extern loc_srv_t *hr_srv;
82
83errno_t hr_raid0_create(hr_volume_t *new_volume)
84{
85 HR_DEBUG("%s()", __func__);
86
87 assert(new_volume->level == HR_LVL_0);
88
89 if (new_volume->extent_no < 2) {
90 HR_ERROR("RAID 0 array needs at least 2 devices\n");
91 return EINVAL;
92 }
93
94 hr_raid0_update_vol_status(new_volume);
95 if (new_volume->status != HR_VOL_ONLINE)
96 return EINVAL;
97
98 bd_srvs_init(&new_volume->hr_bds);
99 new_volume->hr_bds.ops = &hr_raid0_bd_ops;
100 new_volume->hr_bds.sarg = new_volume;
101
102 new_volume->state_callback = hr_raid0_state_callback;
103
104 return EOK;
105}
106
107/*
108 * Called only once in volume's lifetime.
109 */
110errno_t hr_raid0_init(hr_volume_t *vol)
111{
112 HR_DEBUG("%s()", __func__);
113
114 assert(vol->level == HR_LVL_0);
115
116 uint64_t truncated_blkno = vol->extents[0].blkno;
117 for (size_t i = 1; i < vol->extent_no; i++) {
118 if (vol->extents[i].blkno < truncated_blkno)
119 truncated_blkno = vol->extents[i].blkno;
120 }
121
122 uint64_t total_blkno = truncated_blkno * vol->extent_no;
123
124 vol->truncated_blkno = truncated_blkno;
125 vol->nblocks = total_blkno;
126 vol->data_offset = vol->meta_ops->get_data_offset();
127
128 vol->data_blkno = total_blkno;
129 /* count md blocks */
130 vol->data_blkno -= vol->meta_ops->get_size() * vol->extent_no;
131
132 vol->strip_size = HR_STRIP_SIZE;
133
134 return EOK;
135}
136
137void hr_raid0_status_event(hr_volume_t *vol)
138{
139 HR_DEBUG("%s()", __func__);
140
141 hr_raid0_update_vol_status(vol);
142}
143
144static errno_t hr_raid0_bd_open(bd_srvs_t *bds, bd_srv_t *bd)
145{
146 HR_DEBUG("%s()", __func__);
147
148 hr_volume_t *vol = bd->srvs->sarg;
149
150 atomic_fetch_add_explicit(&vol->open_cnt, 1, memory_order_relaxed);
151
152 return EOK;
153}
154
155static errno_t hr_raid0_bd_close(bd_srv_t *bd)
156{
157 HR_DEBUG("%s()", __func__);
158
159 hr_volume_t *vol = bd->srvs->sarg;
160
161 atomic_fetch_sub_explicit(&vol->open_cnt, 1, memory_order_relaxed);
162
163 return EOK;
164}
165
166static errno_t hr_raid0_bd_sync_cache(bd_srv_t *bd, aoff64_t ba, size_t cnt)
167{
168 return hr_raid0_bd_op(HR_BD_SYNC, bd, ba, cnt, NULL, NULL, 0);
169}
170
171static errno_t hr_raid0_bd_read_blocks(bd_srv_t *bd, aoff64_t ba, size_t cnt,
172 void *buf, size_t size)
173{
174 return hr_raid0_bd_op(HR_BD_READ, bd, ba, cnt, buf, NULL, size);
175}
176
177static errno_t hr_raid0_bd_write_blocks(bd_srv_t *bd, aoff64_t ba, size_t cnt,
178 const void *data, size_t size)
179{
180 return hr_raid0_bd_op(HR_BD_WRITE, bd, ba, cnt, NULL, data, size);
181}
182
183static errno_t hr_raid0_bd_get_block_size(bd_srv_t *bd, size_t *rsize)
184{
185 hr_volume_t *vol = bd->srvs->sarg;
186
187 *rsize = vol->bsize;
188 return EOK;
189}
190
191static errno_t hr_raid0_bd_get_num_blocks(bd_srv_t *bd, aoff64_t *rnb)
192{
193 hr_volume_t *vol = bd->srvs->sarg;
194
195 *rnb = vol->data_blkno;
196 return EOK;
197}
198
199static void hr_raid0_update_vol_status(hr_volume_t *vol)
200{
201 fibril_mutex_lock(&vol->md_lock);
202
203 vol->meta_ops->inc_counter(vol->in_mem_md);
204 /* TODO: save right away */
205
206 fibril_mutex_unlock(&vol->md_lock);
207
208 fibril_rwlock_read_lock(&vol->states_lock);
209
210 hr_vol_status_t old_state = vol->status;
211
212 for (size_t i = 0; i < vol->extent_no; i++) {
213 if (vol->extents[i].status != HR_EXT_ONLINE) {
214 fibril_rwlock_read_unlock(&vol->states_lock);
215
216 if (old_state != HR_VOL_FAULTY) {
217 fibril_rwlock_write_lock(&vol->states_lock);
218 hr_update_vol_status(vol, HR_VOL_FAULTY);
219 fibril_rwlock_write_unlock(&vol->states_lock);
220 }
221 return;
222 }
223 }
224 fibril_rwlock_read_unlock(&vol->states_lock);
225
226 if (old_state != HR_VOL_ONLINE) {
227 fibril_rwlock_write_lock(&vol->states_lock);
228 hr_update_vol_status(vol, HR_VOL_ONLINE);
229 fibril_rwlock_write_unlock(&vol->states_lock);
230 }
231}
232
233static void hr_raid0_state_callback(hr_volume_t *vol, size_t extent, errno_t rc)
234{
235 if (rc == EOK)
236 return;
237
238 fibril_rwlock_write_lock(&vol->states_lock);
239
240 switch (rc) {
241 case ENOENT:
242 hr_update_ext_status(vol, extent, HR_EXT_MISSING);
243 break;
244 default:
245 hr_update_ext_status(vol, extent, HR_EXT_FAILED);
246 }
247
248 hr_update_vol_status(vol, HR_VOL_FAULTY);
249
250 fibril_rwlock_write_unlock(&vol->states_lock);
251}
252
253static errno_t hr_raid0_bd_op(hr_bd_op_type_t type, bd_srv_t *bd, aoff64_t ba,
254 size_t cnt, void *dst, const void *src, size_t size)
255{
256 hr_volume_t *vol = bd->srvs->sarg;
257 errno_t rc;
258 uint64_t phys_block, len;
259 size_t left;
260 const uint8_t *data_write = src;
261 uint8_t *data_read = dst;
262
263 fibril_rwlock_read_lock(&vol->states_lock);
264 if (vol->status != HR_VOL_ONLINE) {
265 fibril_rwlock_read_unlock(&vol->states_lock);
266 return EIO;
267 }
268 fibril_rwlock_read_unlock(&vol->states_lock);
269
270 /* propagate sync */
271 if (type == HR_BD_SYNC && ba == 0 && cnt == 0) {
272 hr_fgroup_t *group = hr_fgroup_create(vol->fge,
273 vol->extent_no);
274 if (group == NULL)
275 return ENOMEM;
276
277 for (size_t i = 0; i < vol->extent_no; i++) {
278 hr_io_t *io = hr_fgroup_alloc(group);
279 io->extent = i;
280 io->ba = ba;
281 io->cnt = cnt;
282 io->type = type;
283 io->vol = vol;
284
285 hr_fgroup_submit(group, hr_io_worker, io);
286 }
287
288 size_t bad;
289 rc = hr_fgroup_wait(group, NULL, &bad);
290 if (rc == ENOMEM)
291 return ENOMEM;
292
293 if (bad > 0)
294 return EIO;
295
296 return EOK;
297 }
298
299 if (type == HR_BD_READ || type == HR_BD_WRITE)
300 if (size < cnt * vol->bsize)
301 return EINVAL;
302
303 rc = hr_check_ba_range(vol, cnt, ba);
304 if (rc != EOK)
305 return rc;
306
307 uint64_t strip_size = vol->strip_size / vol->bsize; /* in blocks */
308 uint64_t strip_no = ba / strip_size;
309 uint64_t extent = strip_no % vol->extent_no;
310 uint64_t stripe = strip_no / vol->extent_no;
311 uint64_t strip_off = ba % strip_size;
312
313 left = cnt;
314
315 /* calculate how many strips does the IO span */
316 size_t end_strip_no = (ba + cnt - 1) / strip_size;
317 size_t span = end_strip_no - strip_no + 1;
318
319 hr_fgroup_t *group = hr_fgroup_create(vol->fge, span);
320 if (group == NULL)
321 return ENOMEM;
322
323 while (left != 0) {
324 phys_block = stripe * strip_size + strip_off;
325 cnt = min(left, strip_size - strip_off);
326 len = vol->bsize * cnt;
327 hr_add_ba_offset(vol, &phys_block);
328
329 hr_io_t *io = hr_fgroup_alloc(group);
330 io->extent = extent;
331 io->data_write = data_write;
332 io->data_read = data_read;
333 io->ba = phys_block;
334 io->cnt = cnt;
335 io->type = type;
336 io->vol = vol;
337
338 hr_fgroup_submit(group, hr_io_worker, io);
339
340 left -= cnt;
341 if (left == 0)
342 break;
343
344 if (type == HR_BD_READ)
345 data_read += len;
346 else if (type == HR_BD_WRITE)
347 data_write += len;
348
349 strip_off = 0;
350 extent++;
351 if (extent >= vol->extent_no) {
352 stripe++;
353 extent = 0;
354 }
355 }
356
357 size_t bad;
358 rc = hr_fgroup_wait(group, NULL, &bad);
359 if (rc == ENOMEM && type == HR_BD_READ)
360 return ENOMEM;
361
362 if (bad > 0)
363 return EIO;
364
365 return EOK;
366}
367
368/** @}
369 */
Note: See TracBrowser for help on using the repository browser.