Commit | Line | Data |
---|---|---|
ccd979bd MF |
1 | /* -*- mode: c; c-basic-offset: 8; -*- |
2 | * vim: noexpandtab sw=8 ts=8 sts=0: | |
3 | * | |
4 | * slot_map.c | |
5 | * | |
6 | * | |
7 | * | |
8 | * Copyright (C) 2002, 2004 Oracle. All rights reserved. | |
9 | * | |
10 | * This program is free software; you can redistribute it and/or | |
11 | * modify it under the terms of the GNU General Public | |
12 | * License as published by the Free Software Foundation; either | |
13 | * version 2 of the License, or (at your option) any later version. | |
14 | * | |
15 | * This program is distributed in the hope that it will be useful, | |
16 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
18 | * General Public License for more details. | |
19 | * | |
20 | * You should have received a copy of the GNU General Public | |
21 | * License along with this program; if not, write to the | |
22 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
23 | * Boston, MA 021110-1307, USA. | |
24 | */ | |
25 | ||
26 | #include <linux/types.h> | |
27 | #include <linux/slab.h> | |
28 | #include <linux/highmem.h> | |
ccd979bd MF |
29 | |
30 | #define MLOG_MASK_PREFIX ML_SUPER | |
31 | #include <cluster/masklog.h> | |
32 | ||
33 | #include "ocfs2.h" | |
34 | ||
35 | #include "dlmglue.h" | |
36 | #include "extent_map.h" | |
37 | #include "heartbeat.h" | |
38 | #include "inode.h" | |
39 | #include "slot_map.h" | |
40 | #include "super.h" | |
41 | #include "sysfile.h" | |
42 | ||
43 | #include "buffer_head_io.h" | |
44 | ||
fc881fa0 JB |
45 | |
46 | struct ocfs2_slot { | |
47 | int sl_valid; | |
48 | unsigned int sl_node_num; | |
49 | }; | |
50 | ||
d85b20e4 JB |
51 | struct ocfs2_slot_info { |
52 | struct inode *si_inode; | |
1c8d9a6a JB |
53 | unsigned int si_blocks; |
54 | struct buffer_head **si_bh; | |
d85b20e4 | 55 | unsigned int si_num_slots; |
fc881fa0 | 56 | struct ocfs2_slot *si_slots; |
d85b20e4 JB |
57 | }; |
58 | ||
59 | ||
fc881fa0 JB |
60 | static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si, |
61 | unsigned int node_num); | |
62 | ||
63 | static void ocfs2_invalidate_slot(struct ocfs2_slot_info *si, | |
64 | int slot_num) | |
65 | { | |
66 | BUG_ON((slot_num < 0) || (slot_num >= si->si_num_slots)); | |
67 | si->si_slots[slot_num].sl_valid = 0; | |
68 | } | |
69 | ||
70 | static void ocfs2_set_slot(struct ocfs2_slot_info *si, | |
71 | int slot_num, unsigned int node_num) | |
72 | { | |
73 | BUG_ON((slot_num < 0) || (slot_num >= si->si_num_slots)); | |
74 | BUG_ON((node_num == O2NM_INVALID_NODE_NUM) || | |
75 | (node_num >= O2NM_MAX_NODES)); | |
76 | ||
77 | si->si_slots[slot_num].sl_valid = 1; | |
78 | si->si_slots[slot_num].sl_node_num = node_num; | |
79 | } | |
ccd979bd | 80 | |
d85b20e4 JB |
81 | /* |
82 | * Post the slot information on disk into our slot_info struct. | |
83 | * Must be protected by osb_lock. | |
84 | */ | |
8e8a4603 | 85 | static void ocfs2_update_slot_info(struct ocfs2_slot_info *si) |
ccd979bd MF |
86 | { |
87 | int i; | |
fb86b1f0 | 88 | struct ocfs2_slot_map *sm; |
ccd979bd MF |
89 | |
90 | /* we don't read the slot block here as ocfs2_super_lock | |
91 | * should've made sure we have the most recent copy. */ | |
fb86b1f0 | 92 | sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data; |
ccd979bd | 93 | |
fc881fa0 | 94 | for (i = 0; i < si->si_num_slots; i++) { |
fb86b1f0 | 95 | if (le16_to_cpu(sm->sm_slots[i]) == (u16)OCFS2_INVALID_SLOT) |
fc881fa0 JB |
96 | ocfs2_invalidate_slot(si, i); |
97 | else | |
fb86b1f0 | 98 | ocfs2_set_slot(si, i, le16_to_cpu(sm->sm_slots[i])); |
fc881fa0 | 99 | } |
ccd979bd MF |
100 | } |
101 | ||
8e8a4603 MF |
102 | int ocfs2_refresh_slot_info(struct ocfs2_super *osb) |
103 | { | |
104 | int ret; | |
105 | struct ocfs2_slot_info *si = osb->slot_info; | |
8e8a4603 MF |
106 | |
107 | if (si == NULL) | |
108 | return 0; | |
109 | ||
1c8d9a6a JB |
110 | BUG_ON(si->si_blocks == 0); |
111 | BUG_ON(si->si_bh == NULL); | |
112 | ||
113 | mlog(0, "Refreshing slot map, reading %u block(s)\n", | |
114 | si->si_blocks); | |
115 | ||
116 | /* | |
117 | * We pass -1 as blocknr because we expect all of si->si_bh to | |
118 | * be !NULL. Thus, ocfs2_read_blocks() will ignore blocknr. If | |
119 | * this is not true, the read of -1 (UINT64_MAX) will fail. | |
120 | */ | |
121 | ret = ocfs2_read_blocks(osb, -1, si->si_blocks, si->si_bh, 0, | |
122 | si->si_inode); | |
d85b20e4 JB |
123 | if (ret == 0) { |
124 | spin_lock(&osb->osb_lock); | |
8e8a4603 | 125 | ocfs2_update_slot_info(si); |
d85b20e4 JB |
126 | spin_unlock(&osb->osb_lock); |
127 | } | |
8e8a4603 MF |
128 | |
129 | return ret; | |
130 | } | |
131 | ||
ccd979bd MF |
132 | /* post the our slot info stuff into it's destination bh and write it |
133 | * out. */ | |
8e8a4603 MF |
134 | static int ocfs2_update_disk_slots(struct ocfs2_super *osb, |
135 | struct ocfs2_slot_info *si) | |
ccd979bd MF |
136 | { |
137 | int status, i; | |
fb86b1f0 | 138 | struct ocfs2_slot_map *sm; |
ccd979bd | 139 | |
d85b20e4 | 140 | spin_lock(&osb->osb_lock); |
fb86b1f0 | 141 | sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data; |
fc881fa0 JB |
142 | for (i = 0; i < si->si_num_slots; i++) { |
143 | if (si->si_slots[i].sl_valid) | |
fb86b1f0 | 144 | sm->sm_slots[i] = |
fc881fa0 JB |
145 | cpu_to_le16(si->si_slots[i].sl_node_num); |
146 | else | |
fb86b1f0 | 147 | sm->sm_slots[i] = cpu_to_le16(OCFS2_INVALID_SLOT); |
fc881fa0 | 148 | } |
d85b20e4 | 149 | spin_unlock(&osb->osb_lock); |
ccd979bd | 150 | |
1c8d9a6a | 151 | status = ocfs2_write_block(osb, si->si_bh[0], si->si_inode); |
ccd979bd MF |
152 | if (status < 0) |
153 | mlog_errno(status); | |
154 | ||
155 | return status; | |
156 | } | |
157 | ||
1c8d9a6a JB |
158 | /* |
159 | * Calculate how many bytes are needed by the slot map. Returns | |
160 | * an error if the slot map file is too small. | |
161 | */ | |
162 | static int ocfs2_slot_map_physical_size(struct ocfs2_super *osb, | |
163 | struct inode *inode, | |
164 | unsigned long long *bytes) | |
165 | { | |
166 | unsigned long long bytes_needed; | |
167 | ||
168 | bytes_needed = osb->max_slots * sizeof(__le16); | |
169 | if (bytes_needed > i_size_read(inode)) { | |
170 | mlog(ML_ERROR, | |
171 | "Slot map file is too small! (size %llu, needed %llu)\n", | |
172 | i_size_read(inode), bytes_needed); | |
173 | return -ENOSPC; | |
174 | } | |
175 | ||
176 | *bytes = bytes_needed; | |
177 | return 0; | |
178 | } | |
179 | ||
fc881fa0 JB |
180 | /* try to find global node in the slot info. Returns -ENOENT |
181 | * if nothing is found. */ | |
182 | static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si, | |
183 | unsigned int node_num) | |
ccd979bd | 184 | { |
fc881fa0 | 185 | int i, ret = -ENOENT; |
ccd979bd MF |
186 | |
187 | for(i = 0; i < si->si_num_slots; i++) { | |
fc881fa0 JB |
188 | if (si->si_slots[i].sl_valid && |
189 | (node_num == si->si_slots[i].sl_node_num)) { | |
190 | ret = i; | |
ccd979bd MF |
191 | break; |
192 | } | |
193 | } | |
fc881fa0 | 194 | |
ccd979bd MF |
195 | return ret; |
196 | } | |
197 | ||
fc881fa0 JB |
198 | static int __ocfs2_find_empty_slot(struct ocfs2_slot_info *si, |
199 | int preferred) | |
ccd979bd | 200 | { |
fc881fa0 | 201 | int i, ret = -ENOSPC; |
ccd979bd | 202 | |
fc881fa0 JB |
203 | if ((preferred >= 0) && (preferred < si->si_num_slots)) { |
204 | if (!si->si_slots[preferred].sl_valid) { | |
baf4661a SM |
205 | ret = preferred; |
206 | goto out; | |
207 | } | |
208 | } | |
209 | ||
ccd979bd | 210 | for(i = 0; i < si->si_num_slots; i++) { |
fc881fa0 JB |
211 | if (!si->si_slots[i].sl_valid) { |
212 | ret = i; | |
ccd979bd MF |
213 | break; |
214 | } | |
215 | } | |
baf4661a | 216 | out: |
ccd979bd MF |
217 | return ret; |
218 | } | |
219 | ||
d85b20e4 | 220 | int ocfs2_node_num_to_slot(struct ocfs2_super *osb, unsigned int node_num) |
ccd979bd | 221 | { |
fc881fa0 | 222 | int slot; |
d85b20e4 | 223 | struct ocfs2_slot_info *si = osb->slot_info; |
ccd979bd | 224 | |
d85b20e4 JB |
225 | spin_lock(&osb->osb_lock); |
226 | slot = __ocfs2_node_num_to_slot(si, node_num); | |
227 | spin_unlock(&osb->osb_lock); | |
228 | ||
d85b20e4 JB |
229 | return slot; |
230 | } | |
231 | ||
232 | int ocfs2_slot_to_node_num_locked(struct ocfs2_super *osb, int slot_num, | |
233 | unsigned int *node_num) | |
234 | { | |
235 | struct ocfs2_slot_info *si = osb->slot_info; | |
236 | ||
237 | assert_spin_locked(&osb->osb_lock); | |
238 | ||
239 | BUG_ON(slot_num < 0); | |
240 | BUG_ON(slot_num > osb->max_slots); | |
241 | ||
fc881fa0 | 242 | if (!si->si_slots[slot_num].sl_valid) |
d85b20e4 JB |
243 | return -ENOENT; |
244 | ||
fc881fa0 | 245 | *node_num = si->si_slots[slot_num].sl_node_num; |
d85b20e4 | 246 | return 0; |
ccd979bd MF |
247 | } |
248 | ||
8e8a4603 MF |
249 | static void __ocfs2_free_slot_info(struct ocfs2_slot_info *si) |
250 | { | |
1c8d9a6a JB |
251 | unsigned int i; |
252 | ||
8e8a4603 MF |
253 | if (si == NULL) |
254 | return; | |
255 | ||
256 | if (si->si_inode) | |
257 | iput(si->si_inode); | |
1c8d9a6a JB |
258 | if (si->si_bh) { |
259 | for (i = 0; i < si->si_blocks; i++) { | |
260 | if (si->si_bh[i]) { | |
261 | brelse(si->si_bh[i]); | |
262 | si->si_bh[i] = NULL; | |
263 | } | |
264 | } | |
265 | kfree(si->si_bh); | |
266 | } | |
8e8a4603 MF |
267 | |
268 | kfree(si); | |
269 | } | |
270 | ||
fc881fa0 | 271 | int ocfs2_clear_slot(struct ocfs2_super *osb, int slot_num) |
ccd979bd | 272 | { |
8e8a4603 MF |
273 | struct ocfs2_slot_info *si = osb->slot_info; |
274 | ||
275 | if (si == NULL) | |
276 | return 0; | |
277 | ||
d85b20e4 | 278 | spin_lock(&osb->osb_lock); |
fc881fa0 | 279 | ocfs2_invalidate_slot(si, slot_num); |
d85b20e4 | 280 | spin_unlock(&osb->osb_lock); |
8e8a4603 MF |
281 | |
282 | return ocfs2_update_disk_slots(osb, osb->slot_info); | |
ccd979bd MF |
283 | } |
284 | ||
1c8d9a6a JB |
285 | static int ocfs2_map_slot_buffers(struct ocfs2_super *osb, |
286 | struct ocfs2_slot_info *si) | |
287 | { | |
288 | int status = 0; | |
289 | u64 blkno; | |
290 | unsigned long long blocks, bytes; | |
291 | unsigned int i; | |
292 | struct buffer_head *bh; | |
293 | ||
294 | status = ocfs2_slot_map_physical_size(osb, si->si_inode, &bytes); | |
295 | if (status) | |
296 | goto bail; | |
297 | ||
298 | blocks = ocfs2_blocks_for_bytes(si->si_inode->i_sb, bytes); | |
299 | BUG_ON(blocks > UINT_MAX); | |
300 | si->si_blocks = blocks; | |
301 | if (!si->si_blocks) | |
302 | goto bail; | |
303 | ||
304 | mlog(0, "Slot map needs %u buffers for %llu bytes\n", | |
305 | si->si_blocks, bytes); | |
306 | ||
307 | si->si_bh = kzalloc(sizeof(struct buffer_head *) * si->si_blocks, | |
308 | GFP_KERNEL); | |
309 | if (!si->si_bh) { | |
310 | status = -ENOMEM; | |
311 | mlog_errno(status); | |
312 | goto bail; | |
313 | } | |
314 | ||
315 | for (i = 0; i < si->si_blocks; i++) { | |
316 | status = ocfs2_extent_map_get_blocks(si->si_inode, i, | |
317 | &blkno, NULL, NULL); | |
318 | if (status < 0) { | |
319 | mlog_errno(status); | |
320 | goto bail; | |
321 | } | |
322 | ||
323 | mlog(0, "Reading slot map block %u at %llu\n", i, | |
324 | (unsigned long long)blkno); | |
325 | ||
326 | bh = NULL; /* Acquire a fresh bh */ | |
327 | status = ocfs2_read_block(osb, blkno, &bh, 0, si->si_inode); | |
328 | if (status < 0) { | |
329 | mlog_errno(status); | |
330 | goto bail; | |
331 | } | |
332 | ||
333 | si->si_bh[i] = bh; | |
334 | } | |
335 | ||
336 | bail: | |
337 | return status; | |
338 | } | |
339 | ||
ccd979bd MF |
340 | int ocfs2_init_slot_info(struct ocfs2_super *osb) |
341 | { | |
fc881fa0 | 342 | int status; |
ccd979bd | 343 | struct inode *inode = NULL; |
ccd979bd MF |
344 | struct ocfs2_slot_info *si; |
345 | ||
fc881fa0 JB |
346 | si = kzalloc(sizeof(struct ocfs2_slot_info) + |
347 | (sizeof(struct ocfs2_slot) * osb->max_slots), | |
348 | GFP_KERNEL); | |
ccd979bd MF |
349 | if (!si) { |
350 | status = -ENOMEM; | |
351 | mlog_errno(status); | |
352 | goto bail; | |
353 | } | |
354 | ||
ccd979bd | 355 | si->si_num_slots = osb->max_slots; |
fc881fa0 JB |
356 | si->si_slots = (struct ocfs2_slot *)((char *)si + |
357 | sizeof(struct ocfs2_slot_info)); | |
ccd979bd MF |
358 | |
359 | inode = ocfs2_get_system_file_inode(osb, SLOT_MAP_SYSTEM_INODE, | |
360 | OCFS2_INVALID_SLOT); | |
361 | if (!inode) { | |
362 | status = -EINVAL; | |
363 | mlog_errno(status); | |
364 | goto bail; | |
365 | } | |
366 | ||
1c8d9a6a JB |
367 | si->si_inode = inode; |
368 | status = ocfs2_map_slot_buffers(osb, si); | |
ccd979bd MF |
369 | if (status < 0) { |
370 | mlog_errno(status); | |
371 | goto bail; | |
372 | } | |
373 | ||
d85b20e4 | 374 | osb->slot_info = (struct ocfs2_slot_info *)si; |
ccd979bd MF |
375 | bail: |
376 | if (status < 0 && si) | |
8e8a4603 | 377 | __ocfs2_free_slot_info(si); |
ccd979bd MF |
378 | |
379 | return status; | |
380 | } | |
381 | ||
8e8a4603 | 382 | void ocfs2_free_slot_info(struct ocfs2_super *osb) |
ccd979bd | 383 | { |
8e8a4603 MF |
384 | struct ocfs2_slot_info *si = osb->slot_info; |
385 | ||
386 | osb->slot_info = NULL; | |
387 | __ocfs2_free_slot_info(si); | |
ccd979bd MF |
388 | } |
389 | ||
390 | int ocfs2_find_slot(struct ocfs2_super *osb) | |
391 | { | |
392 | int status; | |
fc881fa0 | 393 | int slot; |
ccd979bd MF |
394 | struct ocfs2_slot_info *si; |
395 | ||
396 | mlog_entry_void(); | |
397 | ||
398 | si = osb->slot_info; | |
399 | ||
d85b20e4 | 400 | spin_lock(&osb->osb_lock); |
ccd979bd MF |
401 | ocfs2_update_slot_info(si); |
402 | ||
ccd979bd MF |
403 | /* search for ourselves first and take the slot if it already |
404 | * exists. Perhaps we need to mark this in a variable for our | |
405 | * own journal recovery? Possibly not, though we certainly | |
406 | * need to warn to the user */ | |
407 | slot = __ocfs2_node_num_to_slot(si, osb->node_num); | |
fc881fa0 | 408 | if (slot < 0) { |
ccd979bd MF |
409 | /* if no slot yet, then just take 1st available |
410 | * one. */ | |
baf4661a | 411 | slot = __ocfs2_find_empty_slot(si, osb->preferred_slot); |
fc881fa0 | 412 | if (slot < 0) { |
d85b20e4 | 413 | spin_unlock(&osb->osb_lock); |
ccd979bd MF |
414 | mlog(ML_ERROR, "no free slots available!\n"); |
415 | status = -EINVAL; | |
416 | goto bail; | |
417 | } | |
418 | } else | |
419 | mlog(ML_NOTICE, "slot %d is already allocated to this node!\n", | |
420 | slot); | |
421 | ||
fc881fa0 | 422 | ocfs2_set_slot(si, slot, osb->node_num); |
ccd979bd | 423 | osb->slot_num = slot; |
d85b20e4 | 424 | spin_unlock(&osb->osb_lock); |
ccd979bd | 425 | |
e7607ab3 | 426 | mlog(0, "taking node slot %d\n", osb->slot_num); |
ccd979bd MF |
427 | |
428 | status = ocfs2_update_disk_slots(osb, si); | |
429 | if (status < 0) | |
430 | mlog_errno(status); | |
431 | ||
432 | bail: | |
433 | mlog_exit(status); | |
434 | return status; | |
435 | } | |
436 | ||
437 | void ocfs2_put_slot(struct ocfs2_super *osb) | |
438 | { | |
439 | int status; | |
440 | struct ocfs2_slot_info *si = osb->slot_info; | |
441 | ||
442 | if (!si) | |
443 | return; | |
444 | ||
d85b20e4 | 445 | spin_lock(&osb->osb_lock); |
ccd979bd MF |
446 | ocfs2_update_slot_info(si); |
447 | ||
fc881fa0 | 448 | ocfs2_invalidate_slot(si, osb->slot_num); |
ccd979bd | 449 | osb->slot_num = OCFS2_INVALID_SLOT; |
d85b20e4 | 450 | spin_unlock(&osb->osb_lock); |
ccd979bd MF |
451 | |
452 | status = ocfs2_update_disk_slots(osb, si); | |
453 | if (status < 0) { | |
454 | mlog_errno(status); | |
455 | goto bail; | |
456 | } | |
457 | ||
458 | bail: | |
8e8a4603 | 459 | ocfs2_free_slot_info(osb); |
ccd979bd MF |
460 | } |
461 |