2 * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 #include "xfs_types.h"
23 #include "xfs_trans.h"
24 #include "xfs_buf_item.h"
27 #include "xfs_dmapi.h"
28 #include "xfs_mount.h"
29 #include "xfs_trans_priv.h"
30 #include "xfs_extfree_item.h"
33 kmem_zone_t
*xfs_efi_zone
;
34 kmem_zone_t
*xfs_efd_zone
;
36 STATIC
void xfs_efi_item_unlock(xfs_efi_log_item_t
*);
39 xfs_efi_item_free(xfs_efi_log_item_t
*efip
)
41 int nexts
= efip
->efi_format
.efi_nextents
;
43 if (nexts
> XFS_EFI_MAX_FAST_EXTENTS
) {
46 kmem_zone_free(xfs_efi_zone
, efip
);
51 * This returns the number of iovecs needed to log the given efi item.
52 * We only need 1 iovec for an efi item. It just logs the efi_log_format
57 xfs_efi_item_size(xfs_efi_log_item_t
*efip
)
63 * This is called to fill in the vector of log iovecs for the
64 * given efi log item. We use only 1 iovec, and we point that
65 * at the efi_log_format structure embedded in the efi item.
66 * It is at this point that we assert that all of the extent
67 * slots in the efi item have been filled.
70 xfs_efi_item_format(xfs_efi_log_item_t
*efip
,
71 xfs_log_iovec_t
*log_vector
)
75 ASSERT(efip
->efi_next_extent
== efip
->efi_format
.efi_nextents
);
77 efip
->efi_format
.efi_type
= XFS_LI_EFI
;
79 size
= sizeof(xfs_efi_log_format_t
);
80 size
+= (efip
->efi_format
.efi_nextents
- 1) * sizeof(xfs_extent_t
);
81 efip
->efi_format
.efi_size
= 1;
83 log_vector
->i_addr
= (xfs_caddr_t
)&(efip
->efi_format
);
84 log_vector
->i_len
= size
;
85 log_vector
->i_type
= XLOG_REG_TYPE_EFI_FORMAT
;
86 ASSERT(size
>= sizeof(xfs_efi_log_format_t
));
91 * Pinning has no meaning for an efi item, so just return.
95 xfs_efi_item_pin(xfs_efi_log_item_t
*efip
)
102 * While EFIs cannot really be pinned, the unpin operation is the
103 * last place at which the EFI is manipulated during a transaction.
104 * Here we coordinate with xfs_efi_cancel() to determine who gets to
109 xfs_efi_item_unpin(xfs_efi_log_item_t
*efip
, int stale
)
111 struct xfs_ail
*ailp
= efip
->efi_item
.li_ailp
;
113 spin_lock(&ailp
->xa_lock
);
114 if (efip
->efi_flags
& XFS_EFI_CANCELED
) {
115 /* xfs_trans_ail_delete() drops the AIL lock. */
116 xfs_trans_ail_delete(ailp
, (xfs_log_item_t
*)efip
);
117 xfs_efi_item_free(efip
);
119 efip
->efi_flags
|= XFS_EFI_COMMITTED
;
120 spin_unlock(&ailp
->xa_lock
);
125 * like unpin only we have to also clear the xaction descriptor
126 * pointing the log item if we free the item. This routine duplicates
127 * unpin because efi_flags is protected by the AIL lock. Freeing
128 * the descriptor and then calling unpin would force us to drop the AIL
129 * lock which would open up a race condition.
132 xfs_efi_item_unpin_remove(xfs_efi_log_item_t
*efip
, xfs_trans_t
*tp
)
134 struct xfs_ail
*ailp
= efip
->efi_item
.li_ailp
;
135 xfs_log_item_desc_t
*lidp
;
137 spin_lock(&ailp
->xa_lock
);
138 if (efip
->efi_flags
& XFS_EFI_CANCELED
) {
140 * free the xaction descriptor pointing to this item
142 lidp
= xfs_trans_find_item(tp
, (xfs_log_item_t
*) efip
);
143 xfs_trans_free_item(tp
, lidp
);
145 /* xfs_trans_ail_delete() drops the AIL lock. */
146 xfs_trans_ail_delete(ailp
, (xfs_log_item_t
*)efip
);
147 xfs_efi_item_free(efip
);
149 efip
->efi_flags
|= XFS_EFI_COMMITTED
;
150 spin_unlock(&ailp
->xa_lock
);
155 * Efi items have no locking or pushing. However, since EFIs are
156 * pulled from the AIL when their corresponding EFDs are committed
157 * to disk, their situation is very similar to being pinned. Return
158 * XFS_ITEM_PINNED so that the caller will eventually flush the log.
159 * This should help in getting the EFI out of the AIL.
163 xfs_efi_item_trylock(xfs_efi_log_item_t
*efip
)
165 return XFS_ITEM_PINNED
;
169 * Efi items have no locking, so just return.
173 xfs_efi_item_unlock(xfs_efi_log_item_t
*efip
)
175 if (efip
->efi_item
.li_flags
& XFS_LI_ABORTED
)
176 xfs_efi_item_free(efip
);
181 * The EFI is logged only once and cannot be moved in the log, so
182 * simply return the lsn at which it's been logged. The canceled
183 * flag is not paid any attention here. Checking for that is delayed
184 * until the EFI is unpinned.
188 xfs_efi_item_committed(xfs_efi_log_item_t
*efip
, xfs_lsn_t lsn
)
194 * There isn't much you can do to push on an efi item. It is simply
195 * stuck waiting for all of its corresponding efd items to be
200 xfs_efi_item_push(xfs_efi_log_item_t
*efip
)
206 * The EFI dependency tracking op doesn't do squat. It can't because
207 * it doesn't know where the free extent is coming from. The dependency
208 * tracking has to be handled by the "enclosing" metadata object. For
209 * example, for inodes, the inode is locked throughout the extent freeing
210 * so the dependency should be recorded there.
214 xfs_efi_item_committing(xfs_efi_log_item_t
*efip
, xfs_lsn_t lsn
)
220 * This is the ops vector shared by all efi log items.
222 static struct xfs_item_ops xfs_efi_item_ops
= {
223 .iop_size
= (uint(*)(xfs_log_item_t
*))xfs_efi_item_size
,
224 .iop_format
= (void(*)(xfs_log_item_t
*, xfs_log_iovec_t
*))
226 .iop_pin
= (void(*)(xfs_log_item_t
*))xfs_efi_item_pin
,
227 .iop_unpin
= (void(*)(xfs_log_item_t
*, int))xfs_efi_item_unpin
,
228 .iop_unpin_remove
= (void(*)(xfs_log_item_t
*, xfs_trans_t
*))
229 xfs_efi_item_unpin_remove
,
230 .iop_trylock
= (uint(*)(xfs_log_item_t
*))xfs_efi_item_trylock
,
231 .iop_unlock
= (void(*)(xfs_log_item_t
*))xfs_efi_item_unlock
,
232 .iop_committed
= (xfs_lsn_t(*)(xfs_log_item_t
*, xfs_lsn_t
))
233 xfs_efi_item_committed
,
234 .iop_push
= (void(*)(xfs_log_item_t
*))xfs_efi_item_push
,
236 .iop_committing
= (void(*)(xfs_log_item_t
*, xfs_lsn_t
))
237 xfs_efi_item_committing
242 * Allocate and initialize an efi item with the given number of extents.
245 xfs_efi_init(xfs_mount_t
*mp
,
249 xfs_efi_log_item_t
*efip
;
252 ASSERT(nextents
> 0);
253 if (nextents
> XFS_EFI_MAX_FAST_EXTENTS
) {
254 size
= (uint
)(sizeof(xfs_efi_log_item_t
) +
255 ((nextents
- 1) * sizeof(xfs_extent_t
)));
256 efip
= (xfs_efi_log_item_t
*)kmem_zalloc(size
, KM_SLEEP
);
258 efip
= (xfs_efi_log_item_t
*)kmem_zone_zalloc(xfs_efi_zone
,
262 efip
->efi_item
.li_type
= XFS_LI_EFI
;
263 efip
->efi_item
.li_ops
= &xfs_efi_item_ops
;
264 efip
->efi_item
.li_mountp
= mp
;
265 efip
->efi_item
.li_ailp
= mp
->m_ail
;
266 efip
->efi_format
.efi_nextents
= nextents
;
267 efip
->efi_format
.efi_id
= (__psint_t
)(void*)efip
;
273 * Copy an EFI format buffer from the given buf, and into the destination
274 * EFI format structure.
275 * The given buffer can be in 32 bit or 64 bit form (which has different padding),
276 * one of which will be the native format for this kernel.
277 * It will handle the conversion of formats if necessary.
280 xfs_efi_copy_format(xfs_log_iovec_t
*buf
, xfs_efi_log_format_t
*dst_efi_fmt
)
282 xfs_efi_log_format_t
*src_efi_fmt
= (xfs_efi_log_format_t
*)buf
->i_addr
;
284 uint len
= sizeof(xfs_efi_log_format_t
) +
285 (src_efi_fmt
->efi_nextents
- 1) * sizeof(xfs_extent_t
);
286 uint len32
= sizeof(xfs_efi_log_format_32_t
) +
287 (src_efi_fmt
->efi_nextents
- 1) * sizeof(xfs_extent_32_t
);
288 uint len64
= sizeof(xfs_efi_log_format_64_t
) +
289 (src_efi_fmt
->efi_nextents
- 1) * sizeof(xfs_extent_64_t
);
291 if (buf
->i_len
== len
) {
292 memcpy((char *)dst_efi_fmt
, (char*)src_efi_fmt
, len
);
294 } else if (buf
->i_len
== len32
) {
295 xfs_efi_log_format_32_t
*src_efi_fmt_32
=
296 (xfs_efi_log_format_32_t
*)buf
->i_addr
;
298 dst_efi_fmt
->efi_type
= src_efi_fmt_32
->efi_type
;
299 dst_efi_fmt
->efi_size
= src_efi_fmt_32
->efi_size
;
300 dst_efi_fmt
->efi_nextents
= src_efi_fmt_32
->efi_nextents
;
301 dst_efi_fmt
->efi_id
= src_efi_fmt_32
->efi_id
;
302 for (i
= 0; i
< dst_efi_fmt
->efi_nextents
; i
++) {
303 dst_efi_fmt
->efi_extents
[i
].ext_start
=
304 src_efi_fmt_32
->efi_extents
[i
].ext_start
;
305 dst_efi_fmt
->efi_extents
[i
].ext_len
=
306 src_efi_fmt_32
->efi_extents
[i
].ext_len
;
309 } else if (buf
->i_len
== len64
) {
310 xfs_efi_log_format_64_t
*src_efi_fmt_64
=
311 (xfs_efi_log_format_64_t
*)buf
->i_addr
;
313 dst_efi_fmt
->efi_type
= src_efi_fmt_64
->efi_type
;
314 dst_efi_fmt
->efi_size
= src_efi_fmt_64
->efi_size
;
315 dst_efi_fmt
->efi_nextents
= src_efi_fmt_64
->efi_nextents
;
316 dst_efi_fmt
->efi_id
= src_efi_fmt_64
->efi_id
;
317 for (i
= 0; i
< dst_efi_fmt
->efi_nextents
; i
++) {
318 dst_efi_fmt
->efi_extents
[i
].ext_start
=
319 src_efi_fmt_64
->efi_extents
[i
].ext_start
;
320 dst_efi_fmt
->efi_extents
[i
].ext_len
=
321 src_efi_fmt_64
->efi_extents
[i
].ext_len
;
329 * This is called by the efd item code below to release references to
330 * the given efi item. Each efd calls this with the number of
331 * extents that it has logged, and when the sum of these reaches
332 * the total number of extents logged by this efi item we can free
335 * Freeing the efi item requires that we remove it from the AIL.
336 * We'll use the AIL lock to protect our counters as well as
337 * the removal from the AIL.
340 xfs_efi_release(xfs_efi_log_item_t
*efip
,
343 struct xfs_ail
*ailp
= efip
->efi_item
.li_ailp
;
346 ASSERT(efip
->efi_next_extent
> 0);
347 ASSERT(efip
->efi_flags
& XFS_EFI_COMMITTED
);
349 spin_lock(&ailp
->xa_lock
);
350 ASSERT(efip
->efi_next_extent
>= nextents
);
351 efip
->efi_next_extent
-= nextents
;
352 extents_left
= efip
->efi_next_extent
;
353 if (extents_left
== 0) {
354 /* xfs_trans_ail_delete() drops the AIL lock. */
355 xfs_trans_ail_delete(ailp
, (xfs_log_item_t
*)efip
);
356 xfs_efi_item_free(efip
);
358 spin_unlock(&ailp
->xa_lock
);
363 xfs_efd_item_free(xfs_efd_log_item_t
*efdp
)
365 int nexts
= efdp
->efd_format
.efd_nextents
;
367 if (nexts
> XFS_EFD_MAX_FAST_EXTENTS
) {
370 kmem_zone_free(xfs_efd_zone
, efdp
);
375 * This returns the number of iovecs needed to log the given efd item.
376 * We only need 1 iovec for an efd item. It just logs the efd_log_format
381 xfs_efd_item_size(xfs_efd_log_item_t
*efdp
)
387 * This is called to fill in the vector of log iovecs for the
388 * given efd log item. We use only 1 iovec, and we point that
389 * at the efd_log_format structure embedded in the efd item.
390 * It is at this point that we assert that all of the extent
391 * slots in the efd item have been filled.
394 xfs_efd_item_format(xfs_efd_log_item_t
*efdp
,
395 xfs_log_iovec_t
*log_vector
)
399 ASSERT(efdp
->efd_next_extent
== efdp
->efd_format
.efd_nextents
);
401 efdp
->efd_format
.efd_type
= XFS_LI_EFD
;
403 size
= sizeof(xfs_efd_log_format_t
);
404 size
+= (efdp
->efd_format
.efd_nextents
- 1) * sizeof(xfs_extent_t
);
405 efdp
->efd_format
.efd_size
= 1;
407 log_vector
->i_addr
= (xfs_caddr_t
)&(efdp
->efd_format
);
408 log_vector
->i_len
= size
;
409 log_vector
->i_type
= XLOG_REG_TYPE_EFD_FORMAT
;
410 ASSERT(size
>= sizeof(xfs_efd_log_format_t
));
415 * Pinning has no meaning for an efd item, so just return.
419 xfs_efd_item_pin(xfs_efd_log_item_t
*efdp
)
426 * Since pinning has no meaning for an efd item, unpinning does
431 xfs_efd_item_unpin(xfs_efd_log_item_t
*efdp
, int stale
)
438 xfs_efd_item_unpin_remove(xfs_efd_log_item_t
*efdp
, xfs_trans_t
*tp
)
444 * Efd items have no locking, so just return success.
448 xfs_efd_item_trylock(xfs_efd_log_item_t
*efdp
)
450 return XFS_ITEM_LOCKED
;
454 * Efd items have no locking or pushing, so return failure
455 * so that the caller doesn't bother with us.
459 xfs_efd_item_unlock(xfs_efd_log_item_t
*efdp
)
461 if (efdp
->efd_item
.li_flags
& XFS_LI_ABORTED
)
462 xfs_efd_item_free(efdp
);
467 * When the efd item is committed to disk, all we need to do
468 * is delete our reference to our partner efi item and then
469 * free ourselves. Since we're freeing ourselves we must
470 * return -1 to keep the transaction code from further referencing
475 xfs_efd_item_committed(xfs_efd_log_item_t
*efdp
, xfs_lsn_t lsn
)
478 * If we got a log I/O error, it's always the case that the LR with the
479 * EFI got unpinned and freed before the EFD got aborted.
481 if ((efdp
->efd_item
.li_flags
& XFS_LI_ABORTED
) == 0)
482 xfs_efi_release(efdp
->efd_efip
, efdp
->efd_format
.efd_nextents
);
484 xfs_efd_item_free(efdp
);
485 return (xfs_lsn_t
)-1;
489 * There isn't much you can do to push on an efd item. It is simply
490 * stuck waiting for the log to be flushed to disk.
494 xfs_efd_item_push(xfs_efd_log_item_t
*efdp
)
500 * The EFD dependency tracking op doesn't do squat. It can't because
501 * it doesn't know where the free extent is coming from. The dependency
502 * tracking has to be handled by the "enclosing" metadata object. For
503 * example, for inodes, the inode is locked throughout the extent freeing
504 * so the dependency should be recorded there.
508 xfs_efd_item_committing(xfs_efd_log_item_t
*efip
, xfs_lsn_t lsn
)
514 * This is the ops vector shared by all efd log items.
516 static struct xfs_item_ops xfs_efd_item_ops
= {
517 .iop_size
= (uint(*)(xfs_log_item_t
*))xfs_efd_item_size
,
518 .iop_format
= (void(*)(xfs_log_item_t
*, xfs_log_iovec_t
*))
520 .iop_pin
= (void(*)(xfs_log_item_t
*))xfs_efd_item_pin
,
521 .iop_unpin
= (void(*)(xfs_log_item_t
*, int))xfs_efd_item_unpin
,
522 .iop_unpin_remove
= (void(*)(xfs_log_item_t
*, xfs_trans_t
*))
523 xfs_efd_item_unpin_remove
,
524 .iop_trylock
= (uint(*)(xfs_log_item_t
*))xfs_efd_item_trylock
,
525 .iop_unlock
= (void(*)(xfs_log_item_t
*))xfs_efd_item_unlock
,
526 .iop_committed
= (xfs_lsn_t(*)(xfs_log_item_t
*, xfs_lsn_t
))
527 xfs_efd_item_committed
,
528 .iop_push
= (void(*)(xfs_log_item_t
*))xfs_efd_item_push
,
530 .iop_committing
= (void(*)(xfs_log_item_t
*, xfs_lsn_t
))
531 xfs_efd_item_committing
536 * Allocate and initialize an efd item with the given number of extents.
539 xfs_efd_init(xfs_mount_t
*mp
,
540 xfs_efi_log_item_t
*efip
,
544 xfs_efd_log_item_t
*efdp
;
547 ASSERT(nextents
> 0);
548 if (nextents
> XFS_EFD_MAX_FAST_EXTENTS
) {
549 size
= (uint
)(sizeof(xfs_efd_log_item_t
) +
550 ((nextents
- 1) * sizeof(xfs_extent_t
)));
551 efdp
= (xfs_efd_log_item_t
*)kmem_zalloc(size
, KM_SLEEP
);
553 efdp
= (xfs_efd_log_item_t
*)kmem_zone_zalloc(xfs_efd_zone
,
557 efdp
->efd_item
.li_type
= XFS_LI_EFD
;
558 efdp
->efd_item
.li_ops
= &xfs_efd_item_ops
;
559 efdp
->efd_item
.li_mountp
= mp
;
560 efdp
->efd_item
.li_ailp
= mp
->m_ail
;
561 efdp
->efd_efip
= efip
;
562 efdp
->efd_format
.efd_nextents
= nextents
;
563 efdp
->efd_format
.efd_efi_id
= efip
->efi_format
.efi_id
;