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 #include <cluster/masklog.h>
35 #include "extent_map.h"
36 #include "heartbeat.h"
41 #include "ocfs2_trace.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 trace_ocfs2_refresh_slot_info(si->si_blocks);
148 * We pass -1 as blocknr because we expect all of si->si_bh to
149 * be !NULL. Thus, ocfs2_read_blocks() will ignore blocknr. If
150 * this is not true, the read of -1 (UINT64_MAX) will fail.
152 ret = ocfs2_read_blocks(INODE_CACHE(si->si_inode), -1, si->si_blocks,
153 si->si_bh, OCFS2_BH_IGNORE_CACHE, NULL);
155 spin_lock(&osb->osb_lock);
156 ocfs2_update_slot_info(si);
157 spin_unlock(&osb->osb_lock);
163 /* post the our slot info stuff into it's destination bh and write it
165 static void ocfs2_update_disk_slot_extended(struct ocfs2_slot_info *si,
167 struct buffer_head **bh)
169 int blkind = slot_num / si->si_slots_per_block;
170 int slotno = slot_num % si->si_slots_per_block;
171 struct ocfs2_slot_map_extended *se;
173 BUG_ON(blkind >= si->si_blocks);
175 se = (struct ocfs2_slot_map_extended *)si->si_bh[blkind]->b_data;
176 se->se_slots[slotno].es_valid = si->si_slots[slot_num].sl_valid;
177 if (si->si_slots[slot_num].sl_valid)
178 se->se_slots[slotno].es_node_num =
179 cpu_to_le32(si->si_slots[slot_num].sl_node_num);
180 *bh = si->si_bh[blkind];
183 static void ocfs2_update_disk_slot_old(struct ocfs2_slot_info *si,
185 struct buffer_head **bh)
188 struct ocfs2_slot_map *sm;
190 sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data;
191 for (i = 0; i < si->si_num_slots; i++) {
192 if (si->si_slots[i].sl_valid)
194 cpu_to_le16(si->si_slots[i].sl_node_num);
196 sm->sm_slots[i] = cpu_to_le16(OCFS2_INVALID_SLOT);
201 static int ocfs2_update_disk_slot(struct ocfs2_super *osb,
202 struct ocfs2_slot_info *si,
206 struct buffer_head *bh;
208 spin_lock(&osb->osb_lock);
210 ocfs2_update_disk_slot_extended(si, slot_num, &bh);
212 ocfs2_update_disk_slot_old(si, slot_num, &bh);
213 spin_unlock(&osb->osb_lock);
215 status = ocfs2_write_block(osb, bh, INODE_CACHE(si->si_inode));
223 * Calculate how many bytes are needed by the slot map. Returns
224 * an error if the slot map file is too small.
226 static int ocfs2_slot_map_physical_size(struct ocfs2_super *osb,
228 unsigned long long *bytes)
230 unsigned long long bytes_needed;
232 if (ocfs2_uses_extended_slot_map(osb)) {
233 bytes_needed = osb->max_slots *
234 sizeof(struct ocfs2_extended_slot);
236 bytes_needed = osb->max_slots * sizeof(__le16);
238 if (bytes_needed > i_size_read(inode)) {
240 "Slot map file is too small! (size %llu, needed %llu)\n",
241 i_size_read(inode), bytes_needed);
245 *bytes = bytes_needed;
249 /* try to find global node in the slot info. Returns -ENOENT
250 * if nothing is found. */
251 static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
252 unsigned int node_num)
254 int i, ret = -ENOENT;
256 for(i = 0; i < si->si_num_slots; i++) {
257 if (si->si_slots[i].sl_valid &&
258 (node_num == si->si_slots[i].sl_node_num)) {
267 static int __ocfs2_find_empty_slot(struct ocfs2_slot_info *si,
270 int i, ret = -ENOSPC;
272 if ((preferred >= 0) && (preferred < si->si_num_slots)) {
273 if (!si->si_slots[preferred].sl_valid) {
279 for(i = 0; i < si->si_num_slots; i++) {
280 if (!si->si_slots[i].sl_valid) {
289 int ocfs2_node_num_to_slot(struct ocfs2_super *osb, unsigned int node_num)
292 struct ocfs2_slot_info *si = osb->slot_info;
294 spin_lock(&osb->osb_lock);
295 slot = __ocfs2_node_num_to_slot(si, node_num);
296 spin_unlock(&osb->osb_lock);
301 int ocfs2_slot_to_node_num_locked(struct ocfs2_super *osb, int slot_num,
302 unsigned int *node_num)
304 struct ocfs2_slot_info *si = osb->slot_info;
306 assert_spin_locked(&osb->osb_lock);
308 BUG_ON(slot_num < 0);
309 BUG_ON(slot_num >= osb->max_slots);
311 if (!si->si_slots[slot_num].sl_valid)
314 *node_num = si->si_slots[slot_num].sl_node_num;
318 static void __ocfs2_free_slot_info(struct ocfs2_slot_info *si)
327 for (i = 0; i < si->si_blocks; i++) {
329 brelse(si->si_bh[i]);
339 int ocfs2_clear_slot(struct ocfs2_super *osb, int slot_num)
341 struct ocfs2_slot_info *si = osb->slot_info;
346 spin_lock(&osb->osb_lock);
347 ocfs2_invalidate_slot(si, slot_num);
348 spin_unlock(&osb->osb_lock);
350 return ocfs2_update_disk_slot(osb, osb->slot_info, slot_num);
353 static int ocfs2_map_slot_buffers(struct ocfs2_super *osb,
354 struct ocfs2_slot_info *si)
358 unsigned long long blocks, bytes = 0;
360 struct buffer_head *bh;
362 status = ocfs2_slot_map_physical_size(osb, si->si_inode, &bytes);
366 blocks = ocfs2_blocks_for_bytes(si->si_inode->i_sb, bytes);
367 BUG_ON(blocks > UINT_MAX);
368 si->si_blocks = blocks;
373 si->si_slots_per_block =
374 (osb->sb->s_blocksize /
375 sizeof(struct ocfs2_extended_slot));
377 si->si_slots_per_block = osb->sb->s_blocksize / sizeof(__le16);
379 /* The size checks above should ensure this */
380 BUG_ON((osb->max_slots / si->si_slots_per_block) > blocks);
382 trace_ocfs2_map_slot_buffers(bytes, si->si_blocks);
384 si->si_bh = kcalloc(si->si_blocks, sizeof(struct buffer_head *),
392 for (i = 0; i < si->si_blocks; i++) {
393 status = ocfs2_extent_map_get_blocks(si->si_inode, i,
400 trace_ocfs2_map_slot_buffers_block((unsigned long long)blkno, i);
402 bh = NULL; /* Acquire a fresh bh */
403 status = ocfs2_read_blocks(INODE_CACHE(si->si_inode), blkno,
404 1, &bh, OCFS2_BH_IGNORE_CACHE, NULL);
417 int ocfs2_init_slot_info(struct ocfs2_super *osb)
420 struct inode *inode = NULL;
421 struct ocfs2_slot_info *si;
423 si = kzalloc(struct_size(si, si_slots, osb->max_slots), GFP_KERNEL);
430 si->si_extended = ocfs2_uses_extended_slot_map(osb);
431 si->si_num_slots = osb->max_slots;
433 inode = ocfs2_get_system_file_inode(osb, SLOT_MAP_SYSTEM_INODE,
441 si->si_inode = inode;
442 status = ocfs2_map_slot_buffers(osb, si);
448 osb->slot_info = (struct ocfs2_slot_info *)si;
451 __ocfs2_free_slot_info(si);
456 void ocfs2_free_slot_info(struct ocfs2_super *osb)
458 struct ocfs2_slot_info *si = osb->slot_info;
460 osb->slot_info = NULL;
461 __ocfs2_free_slot_info(si);
464 int ocfs2_find_slot(struct ocfs2_super *osb)
468 struct ocfs2_slot_info *si;
472 spin_lock(&osb->osb_lock);
473 ocfs2_update_slot_info(si);
475 /* search for ourselves first and take the slot if it already
476 * exists. Perhaps we need to mark this in a variable for our
477 * own journal recovery? Possibly not, though we certainly
478 * need to warn to the user */
479 slot = __ocfs2_node_num_to_slot(si, osb->node_num);
481 /* if no slot yet, then just take 1st available
483 slot = __ocfs2_find_empty_slot(si, osb->preferred_slot);
485 spin_unlock(&osb->osb_lock);
486 mlog(ML_ERROR, "no free slots available!\n");
491 printk(KERN_INFO "ocfs2: Slot %d on device (%s) was already "
492 "allocated to this node!\n", slot, osb->dev_str);
494 ocfs2_set_slot(si, slot, osb->node_num);
495 osb->slot_num = slot;
496 spin_unlock(&osb->osb_lock);
498 trace_ocfs2_find_slot(osb->slot_num);
500 status = ocfs2_update_disk_slot(osb, si, osb->slot_num);
504 * if write block failed, invalidate slot to avoid overwrite
505 * slot during dismount in case another node rightly has mounted
507 spin_lock(&osb->osb_lock);
508 ocfs2_invalidate_slot(si, osb->slot_num);
509 osb->slot_num = OCFS2_INVALID_SLOT;
510 spin_unlock(&osb->osb_lock);
517 void ocfs2_put_slot(struct ocfs2_super *osb)
519 int status, slot_num;
520 struct ocfs2_slot_info *si = osb->slot_info;
525 spin_lock(&osb->osb_lock);
526 ocfs2_update_slot_info(si);
528 slot_num = osb->slot_num;
529 ocfs2_invalidate_slot(si, osb->slot_num);
530 osb->slot_num = OCFS2_INVALID_SLOT;
531 spin_unlock(&osb->osb_lock);
533 status = ocfs2_update_disk_slot(osb, si, slot_num);
537 ocfs2_free_slot_info(osb);