1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
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.
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.
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.
26 #include <linux/types.h>
27 #include <linux/slab.h>
28 #include <linux/highmem.h>
30 #define MLOG_MASK_PREFIX ML_SUPER
31 #include <cluster/masklog.h>
36 #include "extent_map.h"
37 #include "heartbeat.h"
43 #include "buffer_head_io.h"
48 unsigned int sl_node_num;
51 struct ocfs2_slot_info {
53 int si_slots_per_block;
54 struct inode *si_inode;
55 unsigned int si_blocks;
56 struct buffer_head **si_bh;
57 unsigned int si_num_slots;
58 struct ocfs2_slot *si_slots;
62 static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
63 unsigned int node_num);
65 static void ocfs2_invalidate_slot(struct ocfs2_slot_info *si,
68 BUG_ON((slot_num < 0) || (slot_num >= si->si_num_slots));
69 si->si_slots[slot_num].sl_valid = 0;
72 static void ocfs2_set_slot(struct ocfs2_slot_info *si,
73 int slot_num, unsigned int node_num)
75 BUG_ON((slot_num < 0) || (slot_num >= si->si_num_slots));
76 BUG_ON((node_num == O2NM_INVALID_NODE_NUM) ||
77 (node_num >= O2NM_MAX_NODES));
79 si->si_slots[slot_num].sl_valid = 1;
80 si->si_slots[slot_num].sl_node_num = node_num;
83 /* This version is for the extended slot map */
84 static void ocfs2_update_slot_info_extended(struct ocfs2_slot_info *si)
87 struct ocfs2_slot_map_extended *se;
90 for (b = 0; b < si->si_blocks; b++) {
91 se = (struct ocfs2_slot_map_extended *)si->si_bh[b]->b_data;
93 (i < si->si_slots_per_block) &&
94 (slotno < si->si_num_slots);
96 if (se->se_slots[i].es_valid)
97 ocfs2_set_slot(si, slotno,
98 le32_to_cpu(se->se_slots[i].es_node_num));
100 ocfs2_invalidate_slot(si, slotno);
106 * Post the slot information on disk into our slot_info struct.
107 * Must be protected by osb_lock.
109 static void ocfs2_update_slot_info_old(struct ocfs2_slot_info *si)
112 struct ocfs2_slot_map *sm;
114 sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data;
116 for (i = 0; i < si->si_num_slots; i++) {
117 if (le16_to_cpu(sm->sm_slots[i]) == (u16)OCFS2_INVALID_SLOT)
118 ocfs2_invalidate_slot(si, i);
120 ocfs2_set_slot(si, i, le16_to_cpu(sm->sm_slots[i]));
124 static void ocfs2_update_slot_info(struct ocfs2_slot_info *si)
127 * The slot data will have been refreshed when ocfs2_super_lock
131 ocfs2_update_slot_info_extended(si);
133 ocfs2_update_slot_info_old(si);
136 int ocfs2_refresh_slot_info(struct ocfs2_super *osb)
139 struct ocfs2_slot_info *si = osb->slot_info;
144 BUG_ON(si->si_blocks == 0);
145 BUG_ON(si->si_bh == NULL);
147 mlog(0, "Refreshing slot map, reading %u block(s)\n",
151 * We pass -1 as blocknr because we expect all of si->si_bh to
152 * be !NULL. Thus, ocfs2_read_blocks() will ignore blocknr. If
153 * this is not true, the read of -1 (UINT64_MAX) will fail.
155 ret = ocfs2_read_blocks(osb, -1, si->si_blocks, si->si_bh, 0,
158 spin_lock(&osb->osb_lock);
159 ocfs2_update_slot_info(si);
160 spin_unlock(&osb->osb_lock);
166 /* post the our slot info stuff into it's destination bh and write it
168 static void ocfs2_update_disk_slot_extended(struct ocfs2_slot_info *si,
170 struct buffer_head **bh)
172 int blkind = slot_num / si->si_slots_per_block;
173 int slotno = slot_num % si->si_slots_per_block;
174 struct ocfs2_slot_map_extended *se;
176 BUG_ON(blkind >= si->si_blocks);
178 se = (struct ocfs2_slot_map_extended *)si->si_bh[blkind]->b_data;
179 se->se_slots[slotno].es_valid = si->si_slots[slot_num].sl_valid;
180 if (si->si_slots[slot_num].sl_valid)
181 se->se_slots[slotno].es_node_num =
182 cpu_to_le32(si->si_slots[slot_num].sl_node_num);
183 *bh = si->si_bh[blkind];
186 static void ocfs2_update_disk_slot_old(struct ocfs2_slot_info *si,
188 struct buffer_head **bh)
191 struct ocfs2_slot_map *sm;
193 sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data;
194 for (i = 0; i < si->si_num_slots; i++) {
195 if (si->si_slots[i].sl_valid)
197 cpu_to_le16(si->si_slots[i].sl_node_num);
199 sm->sm_slots[i] = cpu_to_le16(OCFS2_INVALID_SLOT);
204 static int ocfs2_update_disk_slot(struct ocfs2_super *osb,
205 struct ocfs2_slot_info *si,
209 struct buffer_head *bh;
211 spin_lock(&osb->osb_lock);
213 ocfs2_update_disk_slot_extended(si, slot_num, &bh);
215 ocfs2_update_disk_slot_old(si, slot_num, &bh);
216 spin_unlock(&osb->osb_lock);
218 status = ocfs2_write_block(osb, bh, si->si_inode);
226 * Calculate how many bytes are needed by the slot map. Returns
227 * an error if the slot map file is too small.
229 static int ocfs2_slot_map_physical_size(struct ocfs2_super *osb,
231 unsigned long long *bytes)
233 unsigned long long bytes_needed;
235 if (ocfs2_uses_extended_slot_map(osb)) {
236 bytes_needed = osb->max_slots *
237 sizeof(struct ocfs2_extended_slot);
239 bytes_needed = osb->max_slots * sizeof(__le16);
241 if (bytes_needed > i_size_read(inode)) {
243 "Slot map file is too small! (size %llu, needed %llu)\n",
244 i_size_read(inode), bytes_needed);
248 *bytes = bytes_needed;
252 /* try to find global node in the slot info. Returns -ENOENT
253 * if nothing is found. */
254 static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
255 unsigned int node_num)
257 int i, ret = -ENOENT;
259 for(i = 0; i < si->si_num_slots; i++) {
260 if (si->si_slots[i].sl_valid &&
261 (node_num == si->si_slots[i].sl_node_num)) {
270 static int __ocfs2_find_empty_slot(struct ocfs2_slot_info *si,
273 int i, ret = -ENOSPC;
275 if ((preferred >= 0) && (preferred < si->si_num_slots)) {
276 if (!si->si_slots[preferred].sl_valid) {
282 for(i = 0; i < si->si_num_slots; i++) {
283 if (!si->si_slots[i].sl_valid) {
292 int ocfs2_node_num_to_slot(struct ocfs2_super *osb, unsigned int node_num)
295 struct ocfs2_slot_info *si = osb->slot_info;
297 spin_lock(&osb->osb_lock);
298 slot = __ocfs2_node_num_to_slot(si, node_num);
299 spin_unlock(&osb->osb_lock);
304 int ocfs2_slot_to_node_num_locked(struct ocfs2_super *osb, int slot_num,
305 unsigned int *node_num)
307 struct ocfs2_slot_info *si = osb->slot_info;
309 assert_spin_locked(&osb->osb_lock);
311 BUG_ON(slot_num < 0);
312 BUG_ON(slot_num > osb->max_slots);
314 if (!si->si_slots[slot_num].sl_valid)
317 *node_num = si->si_slots[slot_num].sl_node_num;
321 static void __ocfs2_free_slot_info(struct ocfs2_slot_info *si)
331 for (i = 0; i < si->si_blocks; i++) {
333 brelse(si->si_bh[i]);
343 int ocfs2_clear_slot(struct ocfs2_super *osb, int slot_num)
345 struct ocfs2_slot_info *si = osb->slot_info;
350 spin_lock(&osb->osb_lock);
351 ocfs2_invalidate_slot(si, slot_num);
352 spin_unlock(&osb->osb_lock);
354 return ocfs2_update_disk_slot(osb, osb->slot_info, slot_num);
357 static int ocfs2_map_slot_buffers(struct ocfs2_super *osb,
358 struct ocfs2_slot_info *si)
362 unsigned long long blocks, bytes;
364 struct buffer_head *bh;
366 status = ocfs2_slot_map_physical_size(osb, si->si_inode, &bytes);
370 blocks = ocfs2_blocks_for_bytes(si->si_inode->i_sb, bytes);
371 BUG_ON(blocks > UINT_MAX);
372 si->si_blocks = blocks;
377 si->si_slots_per_block =
378 (osb->sb->s_blocksize /
379 sizeof(struct ocfs2_extended_slot));
381 si->si_slots_per_block = osb->sb->s_blocksize / sizeof(__le16);
383 /* The size checks above should ensure this */
384 BUG_ON((osb->max_slots / si->si_slots_per_block) > blocks);
386 mlog(0, "Slot map needs %u buffers for %llu bytes\n",
387 si->si_blocks, bytes);
389 si->si_bh = kzalloc(sizeof(struct buffer_head *) * si->si_blocks,
397 for (i = 0; i < si->si_blocks; i++) {
398 status = ocfs2_extent_map_get_blocks(si->si_inode, i,
405 mlog(0, "Reading slot map block %u at %llu\n", i,
406 (unsigned long long)blkno);
408 bh = NULL; /* Acquire a fresh bh */
409 status = ocfs2_read_block(osb, blkno, &bh, 0, si->si_inode);
422 int ocfs2_init_slot_info(struct ocfs2_super *osb)
425 struct inode *inode = NULL;
426 struct ocfs2_slot_info *si;
428 si = kzalloc(sizeof(struct ocfs2_slot_info) +
429 (sizeof(struct ocfs2_slot) * osb->max_slots),
437 si->si_extended = ocfs2_uses_extended_slot_map(osb);
438 si->si_num_slots = osb->max_slots;
439 si->si_slots = (struct ocfs2_slot *)((char *)si +
440 sizeof(struct ocfs2_slot_info));
442 inode = ocfs2_get_system_file_inode(osb, SLOT_MAP_SYSTEM_INODE,
450 si->si_inode = inode;
451 status = ocfs2_map_slot_buffers(osb, si);
457 osb->slot_info = (struct ocfs2_slot_info *)si;
459 if (status < 0 && si)
460 __ocfs2_free_slot_info(si);
465 void ocfs2_free_slot_info(struct ocfs2_super *osb)
467 struct ocfs2_slot_info *si = osb->slot_info;
469 osb->slot_info = NULL;
470 __ocfs2_free_slot_info(si);
473 int ocfs2_find_slot(struct ocfs2_super *osb)
477 struct ocfs2_slot_info *si;
483 spin_lock(&osb->osb_lock);
484 ocfs2_update_slot_info(si);
486 /* search for ourselves first and take the slot if it already
487 * exists. Perhaps we need to mark this in a variable for our
488 * own journal recovery? Possibly not, though we certainly
489 * need to warn to the user */
490 slot = __ocfs2_node_num_to_slot(si, osb->node_num);
492 /* if no slot yet, then just take 1st available
494 slot = __ocfs2_find_empty_slot(si, osb->preferred_slot);
496 spin_unlock(&osb->osb_lock);
497 mlog(ML_ERROR, "no free slots available!\n");
502 mlog(ML_NOTICE, "slot %d is already allocated to this node!\n",
505 ocfs2_set_slot(si, slot, osb->node_num);
506 osb->slot_num = slot;
507 spin_unlock(&osb->osb_lock);
509 mlog(0, "taking node slot %d\n", osb->slot_num);
511 status = ocfs2_update_disk_slot(osb, si, osb->slot_num);
520 void ocfs2_put_slot(struct ocfs2_super *osb)
522 int status, slot_num;
523 struct ocfs2_slot_info *si = osb->slot_info;
528 spin_lock(&osb->osb_lock);
529 ocfs2_update_slot_info(si);
531 slot_num = osb->slot_num;
532 ocfs2_invalidate_slot(si, osb->slot_num);
533 osb->slot_num = OCFS2_INVALID_SLOT;
534 spin_unlock(&osb->osb_lock);
536 status = ocfs2_update_disk_slot(osb, si, slot_num);
543 ocfs2_free_slot_info(osb);