]>
git.ipfire.org Git - people/ms/linux.git/blob - fs/ocfs2/slot_map.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
5 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
8 #include <linux/types.h>
9 #include <linux/slab.h>
10 #include <linux/highmem.h>
12 #include <cluster/masklog.h>
17 #include "extent_map.h"
18 #include "heartbeat.h"
23 #include "ocfs2_trace.h"
25 #include "buffer_head_io.h"
30 unsigned int sl_node_num
;
33 struct ocfs2_slot_info
{
35 int si_slots_per_block
;
36 struct inode
*si_inode
;
37 unsigned int si_blocks
;
38 struct buffer_head
**si_bh
;
39 unsigned int si_num_slots
;
40 struct ocfs2_slot si_slots
[];
44 static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info
*si
,
45 unsigned int node_num
);
47 static void ocfs2_invalidate_slot(struct ocfs2_slot_info
*si
,
50 BUG_ON((slot_num
< 0) || (slot_num
>= si
->si_num_slots
));
51 si
->si_slots
[slot_num
].sl_valid
= 0;
54 static void ocfs2_set_slot(struct ocfs2_slot_info
*si
,
55 int slot_num
, unsigned int node_num
)
57 BUG_ON((slot_num
< 0) || (slot_num
>= si
->si_num_slots
));
59 si
->si_slots
[slot_num
].sl_valid
= 1;
60 si
->si_slots
[slot_num
].sl_node_num
= node_num
;
63 /* This version is for the extended slot map */
64 static void ocfs2_update_slot_info_extended(struct ocfs2_slot_info
*si
)
67 struct ocfs2_slot_map_extended
*se
;
70 for (b
= 0; b
< si
->si_blocks
; b
++) {
71 se
= (struct ocfs2_slot_map_extended
*)si
->si_bh
[b
]->b_data
;
73 (i
< si
->si_slots_per_block
) &&
74 (slotno
< si
->si_num_slots
);
76 if (se
->se_slots
[i
].es_valid
)
77 ocfs2_set_slot(si
, slotno
,
78 le32_to_cpu(se
->se_slots
[i
].es_node_num
));
80 ocfs2_invalidate_slot(si
, slotno
);
86 * Post the slot information on disk into our slot_info struct.
87 * Must be protected by osb_lock.
89 static void ocfs2_update_slot_info_old(struct ocfs2_slot_info
*si
)
92 struct ocfs2_slot_map
*sm
;
94 sm
= (struct ocfs2_slot_map
*)si
->si_bh
[0]->b_data
;
96 for (i
= 0; i
< si
->si_num_slots
; i
++) {
97 if (le16_to_cpu(sm
->sm_slots
[i
]) == (u16
)OCFS2_INVALID_SLOT
)
98 ocfs2_invalidate_slot(si
, i
);
100 ocfs2_set_slot(si
, i
, le16_to_cpu(sm
->sm_slots
[i
]));
104 static void ocfs2_update_slot_info(struct ocfs2_slot_info
*si
)
107 * The slot data will have been refreshed when ocfs2_super_lock
111 ocfs2_update_slot_info_extended(si
);
113 ocfs2_update_slot_info_old(si
);
116 int ocfs2_refresh_slot_info(struct ocfs2_super
*osb
)
119 struct ocfs2_slot_info
*si
= osb
->slot_info
;
124 BUG_ON(si
->si_blocks
== 0);
125 BUG_ON(si
->si_bh
== NULL
);
127 trace_ocfs2_refresh_slot_info(si
->si_blocks
);
130 * We pass -1 as blocknr because we expect all of si->si_bh to
131 * be !NULL. Thus, ocfs2_read_blocks() will ignore blocknr. If
132 * this is not true, the read of -1 (UINT64_MAX) will fail.
134 ret
= ocfs2_read_blocks(INODE_CACHE(si
->si_inode
), -1, si
->si_blocks
,
135 si
->si_bh
, OCFS2_BH_IGNORE_CACHE
, NULL
);
137 spin_lock(&osb
->osb_lock
);
138 ocfs2_update_slot_info(si
);
139 spin_unlock(&osb
->osb_lock
);
145 /* post the our slot info stuff into it's destination bh and write it
147 static void ocfs2_update_disk_slot_extended(struct ocfs2_slot_info
*si
,
149 struct buffer_head
**bh
)
151 int blkind
= slot_num
/ si
->si_slots_per_block
;
152 int slotno
= slot_num
% si
->si_slots_per_block
;
153 struct ocfs2_slot_map_extended
*se
;
155 BUG_ON(blkind
>= si
->si_blocks
);
157 se
= (struct ocfs2_slot_map_extended
*)si
->si_bh
[blkind
]->b_data
;
158 se
->se_slots
[slotno
].es_valid
= si
->si_slots
[slot_num
].sl_valid
;
159 if (si
->si_slots
[slot_num
].sl_valid
)
160 se
->se_slots
[slotno
].es_node_num
=
161 cpu_to_le32(si
->si_slots
[slot_num
].sl_node_num
);
162 *bh
= si
->si_bh
[blkind
];
165 static void ocfs2_update_disk_slot_old(struct ocfs2_slot_info
*si
,
167 struct buffer_head
**bh
)
170 struct ocfs2_slot_map
*sm
;
172 sm
= (struct ocfs2_slot_map
*)si
->si_bh
[0]->b_data
;
173 for (i
= 0; i
< si
->si_num_slots
; i
++) {
174 if (si
->si_slots
[i
].sl_valid
)
176 cpu_to_le16(si
->si_slots
[i
].sl_node_num
);
178 sm
->sm_slots
[i
] = cpu_to_le16(OCFS2_INVALID_SLOT
);
183 static int ocfs2_update_disk_slot(struct ocfs2_super
*osb
,
184 struct ocfs2_slot_info
*si
,
188 struct buffer_head
*bh
;
190 spin_lock(&osb
->osb_lock
);
192 ocfs2_update_disk_slot_extended(si
, slot_num
, &bh
);
194 ocfs2_update_disk_slot_old(si
, slot_num
, &bh
);
195 spin_unlock(&osb
->osb_lock
);
197 status
= ocfs2_write_block(osb
, bh
, INODE_CACHE(si
->si_inode
));
205 * Calculate how many bytes are needed by the slot map. Returns
206 * an error if the slot map file is too small.
208 static int ocfs2_slot_map_physical_size(struct ocfs2_super
*osb
,
210 unsigned long long *bytes
)
212 unsigned long long bytes_needed
;
214 if (ocfs2_uses_extended_slot_map(osb
)) {
215 bytes_needed
= osb
->max_slots
*
216 sizeof(struct ocfs2_extended_slot
);
218 bytes_needed
= osb
->max_slots
* sizeof(__le16
);
220 if (bytes_needed
> i_size_read(inode
)) {
222 "Slot map file is too small! (size %llu, needed %llu)\n",
223 i_size_read(inode
), bytes_needed
);
227 *bytes
= bytes_needed
;
231 /* try to find global node in the slot info. Returns -ENOENT
232 * if nothing is found. */
233 static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info
*si
,
234 unsigned int node_num
)
236 int i
, ret
= -ENOENT
;
238 for(i
= 0; i
< si
->si_num_slots
; i
++) {
239 if (si
->si_slots
[i
].sl_valid
&&
240 (node_num
== si
->si_slots
[i
].sl_node_num
)) {
249 static int __ocfs2_find_empty_slot(struct ocfs2_slot_info
*si
,
252 int i
, ret
= -ENOSPC
;
254 if ((preferred
>= 0) && (preferred
< si
->si_num_slots
)) {
255 if (!si
->si_slots
[preferred
].sl_valid
) {
261 for(i
= 0; i
< si
->si_num_slots
; i
++) {
262 if (!si
->si_slots
[i
].sl_valid
) {
271 int ocfs2_node_num_to_slot(struct ocfs2_super
*osb
, unsigned int node_num
)
274 struct ocfs2_slot_info
*si
= osb
->slot_info
;
276 spin_lock(&osb
->osb_lock
);
277 slot
= __ocfs2_node_num_to_slot(si
, node_num
);
278 spin_unlock(&osb
->osb_lock
);
283 int ocfs2_slot_to_node_num_locked(struct ocfs2_super
*osb
, int slot_num
,
284 unsigned int *node_num
)
286 struct ocfs2_slot_info
*si
= osb
->slot_info
;
288 assert_spin_locked(&osb
->osb_lock
);
290 BUG_ON(slot_num
< 0);
291 BUG_ON(slot_num
>= osb
->max_slots
);
293 if (!si
->si_slots
[slot_num
].sl_valid
)
296 *node_num
= si
->si_slots
[slot_num
].sl_node_num
;
300 static void __ocfs2_free_slot_info(struct ocfs2_slot_info
*si
)
309 for (i
= 0; i
< si
->si_blocks
; i
++) {
311 brelse(si
->si_bh
[i
]);
321 int ocfs2_clear_slot(struct ocfs2_super
*osb
, int slot_num
)
323 struct ocfs2_slot_info
*si
= osb
->slot_info
;
328 spin_lock(&osb
->osb_lock
);
329 ocfs2_invalidate_slot(si
, slot_num
);
330 spin_unlock(&osb
->osb_lock
);
332 return ocfs2_update_disk_slot(osb
, osb
->slot_info
, slot_num
);
335 static int ocfs2_map_slot_buffers(struct ocfs2_super
*osb
,
336 struct ocfs2_slot_info
*si
)
340 unsigned long long blocks
, bytes
= 0;
342 struct buffer_head
*bh
;
344 status
= ocfs2_slot_map_physical_size(osb
, si
->si_inode
, &bytes
);
348 blocks
= ocfs2_blocks_for_bytes(si
->si_inode
->i_sb
, bytes
);
349 BUG_ON(blocks
> UINT_MAX
);
350 si
->si_blocks
= blocks
;
355 si
->si_slots_per_block
=
356 (osb
->sb
->s_blocksize
/
357 sizeof(struct ocfs2_extended_slot
));
359 si
->si_slots_per_block
= osb
->sb
->s_blocksize
/ sizeof(__le16
);
361 /* The size checks above should ensure this */
362 BUG_ON((osb
->max_slots
/ si
->si_slots_per_block
) > blocks
);
364 trace_ocfs2_map_slot_buffers(bytes
, si
->si_blocks
);
366 si
->si_bh
= kcalloc(si
->si_blocks
, sizeof(struct buffer_head
*),
374 for (i
= 0; i
< si
->si_blocks
; i
++) {
375 status
= ocfs2_extent_map_get_blocks(si
->si_inode
, i
,
382 trace_ocfs2_map_slot_buffers_block((unsigned long long)blkno
, i
);
384 bh
= NULL
; /* Acquire a fresh bh */
385 status
= ocfs2_read_blocks(INODE_CACHE(si
->si_inode
), blkno
,
386 1, &bh
, OCFS2_BH_IGNORE_CACHE
, NULL
);
399 int ocfs2_init_slot_info(struct ocfs2_super
*osb
)
402 struct inode
*inode
= NULL
;
403 struct ocfs2_slot_info
*si
;
405 si
= kzalloc(struct_size(si
, si_slots
, osb
->max_slots
), GFP_KERNEL
);
412 si
->si_extended
= ocfs2_uses_extended_slot_map(osb
);
413 si
->si_num_slots
= osb
->max_slots
;
415 inode
= ocfs2_get_system_file_inode(osb
, SLOT_MAP_SYSTEM_INODE
,
423 si
->si_inode
= inode
;
424 status
= ocfs2_map_slot_buffers(osb
, si
);
430 osb
->slot_info
= (struct ocfs2_slot_info
*)si
;
433 __ocfs2_free_slot_info(si
);
438 void ocfs2_free_slot_info(struct ocfs2_super
*osb
)
440 struct ocfs2_slot_info
*si
= osb
->slot_info
;
442 osb
->slot_info
= NULL
;
443 __ocfs2_free_slot_info(si
);
446 int ocfs2_find_slot(struct ocfs2_super
*osb
)
450 struct ocfs2_slot_info
*si
;
454 spin_lock(&osb
->osb_lock
);
455 ocfs2_update_slot_info(si
);
457 /* search for ourselves first and take the slot if it already
458 * exists. Perhaps we need to mark this in a variable for our
459 * own journal recovery? Possibly not, though we certainly
460 * need to warn to the user */
461 slot
= __ocfs2_node_num_to_slot(si
, osb
->node_num
);
463 /* if no slot yet, then just take 1st available
465 slot
= __ocfs2_find_empty_slot(si
, osb
->preferred_slot
);
467 spin_unlock(&osb
->osb_lock
);
468 mlog(ML_ERROR
, "no free slots available!\n");
473 printk(KERN_INFO
"ocfs2: Slot %d on device (%s) was already "
474 "allocated to this node!\n", slot
, osb
->dev_str
);
476 ocfs2_set_slot(si
, slot
, osb
->node_num
);
477 osb
->slot_num
= slot
;
478 spin_unlock(&osb
->osb_lock
);
480 trace_ocfs2_find_slot(osb
->slot_num
);
482 status
= ocfs2_update_disk_slot(osb
, si
, osb
->slot_num
);
486 * if write block failed, invalidate slot to avoid overwrite
487 * slot during dismount in case another node rightly has mounted
489 spin_lock(&osb
->osb_lock
);
490 ocfs2_invalidate_slot(si
, osb
->slot_num
);
491 osb
->slot_num
= OCFS2_INVALID_SLOT
;
492 spin_unlock(&osb
->osb_lock
);
499 void ocfs2_put_slot(struct ocfs2_super
*osb
)
501 int status
, slot_num
;
502 struct ocfs2_slot_info
*si
= osb
->slot_info
;
507 spin_lock(&osb
->osb_lock
);
508 ocfs2_update_slot_info(si
);
510 slot_num
= osb
->slot_num
;
511 ocfs2_invalidate_slot(si
, osb
->slot_num
);
512 osb
->slot_num
= OCFS2_INVALID_SLOT
;
513 spin_unlock(&osb
->osb_lock
);
515 status
= ocfs2_update_disk_slot(osb
, si
, slot_num
);
519 ocfs2_free_slot_info(osb
);