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));
77 si->si_slots[slot_num].sl_valid = 1;
78 si->si_slots[slot_num].sl_node_num = node_num;
81 /* This version is for the extended slot map */
82 static void ocfs2_update_slot_info_extended(struct ocfs2_slot_info *si)
85 struct ocfs2_slot_map_extended *se;
88 for (b = 0; b < si->si_blocks; b++) {
89 se = (struct ocfs2_slot_map_extended *)si->si_bh[b]->b_data;
91 (i < si->si_slots_per_block) &&
92 (slotno < si->si_num_slots);
94 if (se->se_slots[i].es_valid)
95 ocfs2_set_slot(si, slotno,
96 le32_to_cpu(se->se_slots[i].es_node_num));
98 ocfs2_invalidate_slot(si, slotno);
104 * Post the slot information on disk into our slot_info struct.
105 * Must be protected by osb_lock.
107 static void ocfs2_update_slot_info_old(struct ocfs2_slot_info *si)
110 struct ocfs2_slot_map *sm;
112 sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data;
114 for (i = 0; i < si->si_num_slots; i++) {
115 if (le16_to_cpu(sm->sm_slots[i]) == (u16)OCFS2_INVALID_SLOT)
116 ocfs2_invalidate_slot(si, i);
118 ocfs2_set_slot(si, i, le16_to_cpu(sm->sm_slots[i]));
122 static void ocfs2_update_slot_info(struct ocfs2_slot_info *si)
125 * The slot data will have been refreshed when ocfs2_super_lock
129 ocfs2_update_slot_info_extended(si);
131 ocfs2_update_slot_info_old(si);
134 int ocfs2_refresh_slot_info(struct ocfs2_super *osb)
137 struct ocfs2_slot_info *si = osb->slot_info;
142 BUG_ON(si->si_blocks == 0);
143 BUG_ON(si->si_bh == NULL);
145 mlog(0, "Refreshing slot map, reading %u block(s)\n",
149 * We pass -1 as blocknr because we expect all of si->si_bh to
150 * be !NULL. Thus, ocfs2_read_blocks() will ignore blocknr. If
151 * this is not true, the read of -1 (UINT64_MAX) will fail.
153 ret = ocfs2_read_blocks(osb, -1, si->si_blocks, si->si_bh, 0,
156 spin_lock(&osb->osb_lock);
157 ocfs2_update_slot_info(si);
158 spin_unlock(&osb->osb_lock);
164 /* post the our slot info stuff into it's destination bh and write it
166 static void ocfs2_update_disk_slot_extended(struct ocfs2_slot_info *si,
168 struct buffer_head **bh)
170 int blkind = slot_num / si->si_slots_per_block;
171 int slotno = slot_num % si->si_slots_per_block;
172 struct ocfs2_slot_map_extended *se;
174 BUG_ON(blkind >= si->si_blocks);
176 se = (struct ocfs2_slot_map_extended *)si->si_bh[blkind]->b_data;
177 se->se_slots[slotno].es_valid = si->si_slots[slot_num].sl_valid;
178 if (si->si_slots[slot_num].sl_valid)
179 se->se_slots[slotno].es_node_num =
180 cpu_to_le32(si->si_slots[slot_num].sl_node_num);
181 *bh = si->si_bh[blkind];
184 static void ocfs2_update_disk_slot_old(struct ocfs2_slot_info *si,
186 struct buffer_head **bh)
189 struct ocfs2_slot_map *sm;
191 sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data;
192 for (i = 0; i < si->si_num_slots; i++) {
193 if (si->si_slots[i].sl_valid)
195 cpu_to_le16(si->si_slots[i].sl_node_num);
197 sm->sm_slots[i] = cpu_to_le16(OCFS2_INVALID_SLOT);
202 static int ocfs2_update_disk_slot(struct ocfs2_super *osb,
203 struct ocfs2_slot_info *si,
207 struct buffer_head *bh;
209 spin_lock(&osb->osb_lock);
211 ocfs2_update_disk_slot_extended(si, slot_num, &bh);
213 ocfs2_update_disk_slot_old(si, slot_num, &bh);
214 spin_unlock(&osb->osb_lock);
216 status = ocfs2_write_block(osb, bh, si->si_inode);
224 * Calculate how many bytes are needed by the slot map. Returns
225 * an error if the slot map file is too small.
227 static int ocfs2_slot_map_physical_size(struct ocfs2_super *osb,
229 unsigned long long *bytes)
231 unsigned long long bytes_needed;
233 if (ocfs2_uses_extended_slot_map(osb)) {
234 bytes_needed = osb->max_slots *
235 sizeof(struct ocfs2_extended_slot);
237 bytes_needed = osb->max_slots * sizeof(__le16);
239 if (bytes_needed > i_size_read(inode)) {
241 "Slot map file is too small! (size %llu, needed %llu)\n",
242 i_size_read(inode), bytes_needed);
246 *bytes = bytes_needed;
250 /* try to find global node in the slot info. Returns -ENOENT
251 * if nothing is found. */
252 static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
253 unsigned int node_num)
255 int i, ret = -ENOENT;
257 for(i = 0; i < si->si_num_slots; i++) {
258 if (si->si_slots[i].sl_valid &&
259 (node_num == si->si_slots[i].sl_node_num)) {
268 static int __ocfs2_find_empty_slot(struct ocfs2_slot_info *si,
271 int i, ret = -ENOSPC;
273 if ((preferred >= 0) && (preferred < si->si_num_slots)) {
274 if (!si->si_slots[preferred].sl_valid) {
280 for(i = 0; i < si->si_num_slots; i++) {
281 if (!si->si_slots[i].sl_valid) {
290 int ocfs2_node_num_to_slot(struct ocfs2_super *osb, unsigned int node_num)
293 struct ocfs2_slot_info *si = osb->slot_info;
295 spin_lock(&osb->osb_lock);
296 slot = __ocfs2_node_num_to_slot(si, node_num);
297 spin_unlock(&osb->osb_lock);
302 int ocfs2_slot_to_node_num_locked(struct ocfs2_super *osb, int slot_num,
303 unsigned int *node_num)
305 struct ocfs2_slot_info *si = osb->slot_info;
307 assert_spin_locked(&osb->osb_lock);
309 BUG_ON(slot_num < 0);
310 BUG_ON(slot_num > osb->max_slots);
312 if (!si->si_slots[slot_num].sl_valid)
315 *node_num = si->si_slots[slot_num].sl_node_num;
319 static void __ocfs2_free_slot_info(struct ocfs2_slot_info *si)
329 for (i = 0; i < si->si_blocks; i++) {
331 brelse(si->si_bh[i]);
341 int ocfs2_clear_slot(struct ocfs2_super *osb, int slot_num)
343 struct ocfs2_slot_info *si = osb->slot_info;
348 spin_lock(&osb->osb_lock);
349 ocfs2_invalidate_slot(si, slot_num);
350 spin_unlock(&osb->osb_lock);
352 return ocfs2_update_disk_slot(osb, osb->slot_info, slot_num);
355 static int ocfs2_map_slot_buffers(struct ocfs2_super *osb,
356 struct ocfs2_slot_info *si)
360 unsigned long long blocks, bytes;
362 struct buffer_head *bh;
364 status = ocfs2_slot_map_physical_size(osb, si->si_inode, &bytes);
368 blocks = ocfs2_blocks_for_bytes(si->si_inode->i_sb, bytes);
369 BUG_ON(blocks > UINT_MAX);
370 si->si_blocks = blocks;
375 si->si_slots_per_block =
376 (osb->sb->s_blocksize /
377 sizeof(struct ocfs2_extended_slot));
379 si->si_slots_per_block = osb->sb->s_blocksize / sizeof(__le16);
381 /* The size checks above should ensure this */
382 BUG_ON((osb->max_slots / si->si_slots_per_block) > blocks);
384 mlog(0, "Slot map needs %u buffers for %llu bytes\n",
385 si->si_blocks, bytes);
387 si->si_bh = kzalloc(sizeof(struct buffer_head *) * si->si_blocks,
395 for (i = 0; i < si->si_blocks; i++) {
396 status = ocfs2_extent_map_get_blocks(si->si_inode, i,
403 mlog(0, "Reading slot map block %u at %llu\n", i,
404 (unsigned long long)blkno);
406 bh = NULL; /* Acquire a fresh bh */
407 status = ocfs2_read_block(osb, blkno, &bh, 0, si->si_inode);
420 int ocfs2_init_slot_info(struct ocfs2_super *osb)
423 struct inode *inode = NULL;
424 struct ocfs2_slot_info *si;
426 si = kzalloc(sizeof(struct ocfs2_slot_info) +
427 (sizeof(struct ocfs2_slot) * osb->max_slots),
435 si->si_extended = ocfs2_uses_extended_slot_map(osb);
436 si->si_num_slots = osb->max_slots;
437 si->si_slots = (struct ocfs2_slot *)((char *)si +
438 sizeof(struct ocfs2_slot_info));
440 inode = ocfs2_get_system_file_inode(osb, SLOT_MAP_SYSTEM_INODE,
448 si->si_inode = inode;
449 status = ocfs2_map_slot_buffers(osb, si);
455 osb->slot_info = (struct ocfs2_slot_info *)si;
457 if (status < 0 && si)
458 __ocfs2_free_slot_info(si);
463 void ocfs2_free_slot_info(struct ocfs2_super *osb)
465 struct ocfs2_slot_info *si = osb->slot_info;
467 osb->slot_info = NULL;
468 __ocfs2_free_slot_info(si);
471 int ocfs2_find_slot(struct ocfs2_super *osb)
475 struct ocfs2_slot_info *si;
481 spin_lock(&osb->osb_lock);
482 ocfs2_update_slot_info(si);
484 /* search for ourselves first and take the slot if it already
485 * exists. Perhaps we need to mark this in a variable for our
486 * own journal recovery? Possibly not, though we certainly
487 * need to warn to the user */
488 slot = __ocfs2_node_num_to_slot(si, osb->node_num);
490 /* if no slot yet, then just take 1st available
492 slot = __ocfs2_find_empty_slot(si, osb->preferred_slot);
494 spin_unlock(&osb->osb_lock);
495 mlog(ML_ERROR, "no free slots available!\n");
500 mlog(ML_NOTICE, "slot %d is already allocated to this node!\n",
503 ocfs2_set_slot(si, slot, osb->node_num);
504 osb->slot_num = slot;
505 spin_unlock(&osb->osb_lock);
507 mlog(0, "taking node slot %d\n", osb->slot_num);
509 status = ocfs2_update_disk_slot(osb, si, osb->slot_num);
518 void ocfs2_put_slot(struct ocfs2_super *osb)
520 int status, slot_num;
521 struct ocfs2_slot_info *si = osb->slot_info;
526 spin_lock(&osb->osb_lock);
527 ocfs2_update_slot_info(si);
529 slot_num = osb->slot_num;
530 ocfs2_invalidate_slot(si, osb->slot_num);
531 osb->slot_num = OCFS2_INVALID_SLOT;
532 spin_unlock(&osb->osb_lock);
534 status = ocfs2_update_disk_slot(osb, si, slot_num);
541 ocfs2_free_slot_info(osb);