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
) {
44 kmem_free(efip
, sizeof(xfs_efi_log_item_t
) +
45 (nexts
- 1) * sizeof(xfs_extent_t
));
47 kmem_zone_free(xfs_efi_zone
, efip
);
52 * This returns the number of iovecs needed to log the given efi item.
53 * We only need 1 iovec for an efi item. It just logs the efi_log_format
58 xfs_efi_item_size(xfs_efi_log_item_t
*efip
)
64 * This is called to fill in the vector of log iovecs for the
65 * given efi log item. We use only 1 iovec, and we point that
66 * at the efi_log_format structure embedded in the efi item.
67 * It is at this point that we assert that all of the extent
68 * slots in the efi item have been filled.
71 xfs_efi_item_format(xfs_efi_log_item_t
*efip
,
72 xfs_log_iovec_t
*log_vector
)
76 ASSERT(efip
->efi_next_extent
== efip
->efi_format
.efi_nextents
);
78 efip
->efi_format
.efi_type
= XFS_LI_EFI
;
80 size
= sizeof(xfs_efi_log_format_t
);
81 size
+= (efip
->efi_format
.efi_nextents
- 1) * sizeof(xfs_extent_t
);
82 efip
->efi_format
.efi_size
= 1;
84 log_vector
->i_addr
= (xfs_caddr_t
)&(efip
->efi_format
);
85 log_vector
->i_len
= size
;
86 XLOG_VEC_SET_TYPE(log_vector
, XLOG_REG_TYPE_EFI_FORMAT
);
87 ASSERT(size
>= sizeof(xfs_efi_log_format_t
));
92 * Pinning has no meaning for an efi item, so just return.
96 xfs_efi_item_pin(xfs_efi_log_item_t
*efip
)
103 * While EFIs cannot really be pinned, the unpin operation is the
104 * last place at which the EFI is manipulated during a transaction.
105 * Here we coordinate with xfs_efi_cancel() to determine who gets to
110 xfs_efi_item_unpin(xfs_efi_log_item_t
*efip
, int stale
)
115 mp
= efip
->efi_item
.li_mountp
;
117 if (efip
->efi_flags
& XFS_EFI_CANCELED
) {
119 * xfs_trans_delete_ail() drops the AIL lock.
121 xfs_trans_delete_ail(mp
, (xfs_log_item_t
*)efip
, s
);
122 xfs_efi_item_free(efip
);
124 efip
->efi_flags
|= XFS_EFI_COMMITTED
;
130 * like unpin only we have to also clear the xaction descriptor
131 * pointing the log item if we free the item. This routine duplicates
132 * unpin because efi_flags is protected by the AIL lock. Freeing
133 * the descriptor and then calling unpin would force us to drop the AIL
134 * lock which would open up a race condition.
137 xfs_efi_item_unpin_remove(xfs_efi_log_item_t
*efip
, xfs_trans_t
*tp
)
140 xfs_log_item_desc_t
*lidp
;
143 mp
= efip
->efi_item
.li_mountp
;
145 if (efip
->efi_flags
& XFS_EFI_CANCELED
) {
147 * free the xaction descriptor pointing to this item
149 lidp
= xfs_trans_find_item(tp
, (xfs_log_item_t
*) efip
);
150 xfs_trans_free_item(tp
, lidp
);
152 * pull the item off the AIL.
153 * xfs_trans_delete_ail() drops the AIL lock.
155 xfs_trans_delete_ail(mp
, (xfs_log_item_t
*)efip
, s
);
156 xfs_efi_item_free(efip
);
158 efip
->efi_flags
|= XFS_EFI_COMMITTED
;
164 * Efi items have no locking or pushing. However, since EFIs are
165 * pulled from the AIL when their corresponding EFDs are committed
166 * to disk, their situation is very similar to being pinned. Return
167 * XFS_ITEM_PINNED so that the caller will eventually flush the log.
168 * This should help in getting the EFI out of the AIL.
172 xfs_efi_item_trylock(xfs_efi_log_item_t
*efip
)
174 return XFS_ITEM_PINNED
;
178 * Efi items have no locking, so just return.
182 xfs_efi_item_unlock(xfs_efi_log_item_t
*efip
)
184 if (efip
->efi_item
.li_flags
& XFS_LI_ABORTED
)
185 xfs_efi_item_free(efip
);
190 * The EFI is logged only once and cannot be moved in the log, so
191 * simply return the lsn at which it's been logged. The canceled
192 * flag is not paid any attention here. Checking for that is delayed
193 * until the EFI is unpinned.
197 xfs_efi_item_committed(xfs_efi_log_item_t
*efip
, xfs_lsn_t lsn
)
203 * There isn't much you can do to push on an efi item. It is simply
204 * stuck waiting for all of its corresponding efd items to be
209 xfs_efi_item_push(xfs_efi_log_item_t
*efip
)
215 * The EFI dependency tracking op doesn't do squat. It can't because
216 * it doesn't know where the free extent is coming from. The dependency
217 * tracking has to be handled by the "enclosing" metadata object. For
218 * example, for inodes, the inode is locked throughout the extent freeing
219 * so the dependency should be recorded there.
223 xfs_efi_item_committing(xfs_efi_log_item_t
*efip
, xfs_lsn_t lsn
)
229 * This is the ops vector shared by all efi log items.
231 static struct xfs_item_ops xfs_efi_item_ops
= {
232 .iop_size
= (uint(*)(xfs_log_item_t
*))xfs_efi_item_size
,
233 .iop_format
= (void(*)(xfs_log_item_t
*, xfs_log_iovec_t
*))
235 .iop_pin
= (void(*)(xfs_log_item_t
*))xfs_efi_item_pin
,
236 .iop_unpin
= (void(*)(xfs_log_item_t
*, int))xfs_efi_item_unpin
,
237 .iop_unpin_remove
= (void(*)(xfs_log_item_t
*, xfs_trans_t
*))
238 xfs_efi_item_unpin_remove
,
239 .iop_trylock
= (uint(*)(xfs_log_item_t
*))xfs_efi_item_trylock
,
240 .iop_unlock
= (void(*)(xfs_log_item_t
*))xfs_efi_item_unlock
,
241 .iop_committed
= (xfs_lsn_t(*)(xfs_log_item_t
*, xfs_lsn_t
))
242 xfs_efi_item_committed
,
243 .iop_push
= (void(*)(xfs_log_item_t
*))xfs_efi_item_push
,
245 .iop_committing
= (void(*)(xfs_log_item_t
*, xfs_lsn_t
))
246 xfs_efi_item_committing
251 * Allocate and initialize an efi item with the given number of extents.
254 xfs_efi_init(xfs_mount_t
*mp
,
258 xfs_efi_log_item_t
*efip
;
261 ASSERT(nextents
> 0);
262 if (nextents
> XFS_EFI_MAX_FAST_EXTENTS
) {
263 size
= (uint
)(sizeof(xfs_efi_log_item_t
) +
264 ((nextents
- 1) * sizeof(xfs_extent_t
)));
265 efip
= (xfs_efi_log_item_t
*)kmem_zalloc(size
, KM_SLEEP
);
267 efip
= (xfs_efi_log_item_t
*)kmem_zone_zalloc(xfs_efi_zone
,
271 efip
->efi_item
.li_type
= XFS_LI_EFI
;
272 efip
->efi_item
.li_ops
= &xfs_efi_item_ops
;
273 efip
->efi_item
.li_mountp
= mp
;
274 efip
->efi_format
.efi_nextents
= nextents
;
275 efip
->efi_format
.efi_id
= (__psint_t
)(void*)efip
;
281 * Copy an EFI format buffer from the given buf, and into the destination
282 * EFI format structure.
283 * The given buffer can be in 32 bit or 64 bit form (which has different padding),
284 * one of which will be the native format for this kernel.
285 * It will handle the conversion of formats if necessary.
288 xfs_efi_copy_format(xfs_log_iovec_t
*buf
, xfs_efi_log_format_t
*dst_efi_fmt
)
290 xfs_efi_log_format_t
*src_efi_fmt
= (xfs_efi_log_format_t
*)buf
->i_addr
;
292 uint len
= sizeof(xfs_efi_log_format_t
) +
293 (src_efi_fmt
->efi_nextents
- 1) * sizeof(xfs_extent_t
);
294 uint len32
= sizeof(xfs_efi_log_format_32_t
) +
295 (src_efi_fmt
->efi_nextents
- 1) * sizeof(xfs_extent_32_t
);
296 uint len64
= sizeof(xfs_efi_log_format_64_t
) +
297 (src_efi_fmt
->efi_nextents
- 1) * sizeof(xfs_extent_64_t
);
299 if (buf
->i_len
== len
) {
300 memcpy((char *)dst_efi_fmt
, (char*)src_efi_fmt
, len
);
302 } else if (buf
->i_len
== len32
) {
303 xfs_efi_log_format_32_t
*src_efi_fmt_32
=
304 (xfs_efi_log_format_32_t
*)buf
->i_addr
;
306 dst_efi_fmt
->efi_type
= src_efi_fmt_32
->efi_type
;
307 dst_efi_fmt
->efi_size
= src_efi_fmt_32
->efi_size
;
308 dst_efi_fmt
->efi_nextents
= src_efi_fmt_32
->efi_nextents
;
309 dst_efi_fmt
->efi_id
= src_efi_fmt_32
->efi_id
;
310 for (i
= 0; i
< dst_efi_fmt
->efi_nextents
; i
++) {
311 dst_efi_fmt
->efi_extents
[i
].ext_start
=
312 src_efi_fmt_32
->efi_extents
[i
].ext_start
;
313 dst_efi_fmt
->efi_extents
[i
].ext_len
=
314 src_efi_fmt_32
->efi_extents
[i
].ext_len
;
317 } else if (buf
->i_len
== len64
) {
318 xfs_efi_log_format_64_t
*src_efi_fmt_64
=
319 (xfs_efi_log_format_64_t
*)buf
->i_addr
;
321 dst_efi_fmt
->efi_type
= src_efi_fmt_64
->efi_type
;
322 dst_efi_fmt
->efi_size
= src_efi_fmt_64
->efi_size
;
323 dst_efi_fmt
->efi_nextents
= src_efi_fmt_64
->efi_nextents
;
324 dst_efi_fmt
->efi_id
= src_efi_fmt_64
->efi_id
;
325 for (i
= 0; i
< dst_efi_fmt
->efi_nextents
; i
++) {
326 dst_efi_fmt
->efi_extents
[i
].ext_start
=
327 src_efi_fmt_64
->efi_extents
[i
].ext_start
;
328 dst_efi_fmt
->efi_extents
[i
].ext_len
=
329 src_efi_fmt_64
->efi_extents
[i
].ext_len
;
337 * This is called by the efd item code below to release references to
338 * the given efi item. Each efd calls this with the number of
339 * extents that it has logged, and when the sum of these reaches
340 * the total number of extents logged by this efi item we can free
343 * Freeing the efi item requires that we remove it from the AIL.
344 * We'll use the AIL lock to protect our counters as well as
345 * the removal from the AIL.
348 xfs_efi_release(xfs_efi_log_item_t
*efip
,
355 mp
= efip
->efi_item
.li_mountp
;
356 ASSERT(efip
->efi_next_extent
> 0);
357 ASSERT(efip
->efi_flags
& XFS_EFI_COMMITTED
);
360 ASSERT(efip
->efi_next_extent
>= nextents
);
361 efip
->efi_next_extent
-= nextents
;
362 extents_left
= efip
->efi_next_extent
;
363 if (extents_left
== 0) {
365 * xfs_trans_delete_ail() drops the AIL lock.
367 xfs_trans_delete_ail(mp
, (xfs_log_item_t
*)efip
, s
);
368 xfs_efi_item_free(efip
);
375 xfs_efd_item_free(xfs_efd_log_item_t
*efdp
)
377 int nexts
= efdp
->efd_format
.efd_nextents
;
379 if (nexts
> XFS_EFD_MAX_FAST_EXTENTS
) {
380 kmem_free(efdp
, sizeof(xfs_efd_log_item_t
) +
381 (nexts
- 1) * sizeof(xfs_extent_t
));
383 kmem_zone_free(xfs_efd_zone
, efdp
);
388 * This returns the number of iovecs needed to log the given efd item.
389 * We only need 1 iovec for an efd item. It just logs the efd_log_format
394 xfs_efd_item_size(xfs_efd_log_item_t
*efdp
)
400 * This is called to fill in the vector of log iovecs for the
401 * given efd log item. We use only 1 iovec, and we point that
402 * at the efd_log_format structure embedded in the efd item.
403 * It is at this point that we assert that all of the extent
404 * slots in the efd item have been filled.
407 xfs_efd_item_format(xfs_efd_log_item_t
*efdp
,
408 xfs_log_iovec_t
*log_vector
)
412 ASSERT(efdp
->efd_next_extent
== efdp
->efd_format
.efd_nextents
);
414 efdp
->efd_format
.efd_type
= XFS_LI_EFD
;
416 size
= sizeof(xfs_efd_log_format_t
);
417 size
+= (efdp
->efd_format
.efd_nextents
- 1) * sizeof(xfs_extent_t
);
418 efdp
->efd_format
.efd_size
= 1;
420 log_vector
->i_addr
= (xfs_caddr_t
)&(efdp
->efd_format
);
421 log_vector
->i_len
= size
;
422 XLOG_VEC_SET_TYPE(log_vector
, XLOG_REG_TYPE_EFD_FORMAT
);
423 ASSERT(size
>= sizeof(xfs_efd_log_format_t
));
428 * Pinning has no meaning for an efd item, so just return.
432 xfs_efd_item_pin(xfs_efd_log_item_t
*efdp
)
439 * Since pinning has no meaning for an efd item, unpinning does
444 xfs_efd_item_unpin(xfs_efd_log_item_t
*efdp
, int stale
)
451 xfs_efd_item_unpin_remove(xfs_efd_log_item_t
*efdp
, xfs_trans_t
*tp
)
457 * Efd items have no locking, so just return success.
461 xfs_efd_item_trylock(xfs_efd_log_item_t
*efdp
)
463 return XFS_ITEM_LOCKED
;
467 * Efd items have no locking or pushing, so return failure
468 * so that the caller doesn't bother with us.
472 xfs_efd_item_unlock(xfs_efd_log_item_t
*efdp
)
474 if (efdp
->efd_item
.li_flags
& XFS_LI_ABORTED
)
475 xfs_efd_item_free(efdp
);
480 * When the efd item is committed to disk, all we need to do
481 * is delete our reference to our partner efi item and then
482 * free ourselves. Since we're freeing ourselves we must
483 * return -1 to keep the transaction code from further referencing
488 xfs_efd_item_committed(xfs_efd_log_item_t
*efdp
, xfs_lsn_t lsn
)
491 * If we got a log I/O error, it's always the case that the LR with the
492 * EFI got unpinned and freed before the EFD got aborted.
494 if ((efdp
->efd_item
.li_flags
& XFS_LI_ABORTED
) == 0)
495 xfs_efi_release(efdp
->efd_efip
, efdp
->efd_format
.efd_nextents
);
497 xfs_efd_item_free(efdp
);
498 return (xfs_lsn_t
)-1;
502 * There isn't much you can do to push on an efd item. It is simply
503 * stuck waiting for the log to be flushed to disk.
507 xfs_efd_item_push(xfs_efd_log_item_t
*efdp
)
513 * The EFD dependency tracking op doesn't do squat. It can't because
514 * it doesn't know where the free extent is coming from. The dependency
515 * tracking has to be handled by the "enclosing" metadata object. For
516 * example, for inodes, the inode is locked throughout the extent freeing
517 * so the dependency should be recorded there.
521 xfs_efd_item_committing(xfs_efd_log_item_t
*efip
, xfs_lsn_t lsn
)
527 * This is the ops vector shared by all efd log items.
529 static struct xfs_item_ops xfs_efd_item_ops
= {
530 .iop_size
= (uint(*)(xfs_log_item_t
*))xfs_efd_item_size
,
531 .iop_format
= (void(*)(xfs_log_item_t
*, xfs_log_iovec_t
*))
533 .iop_pin
= (void(*)(xfs_log_item_t
*))xfs_efd_item_pin
,
534 .iop_unpin
= (void(*)(xfs_log_item_t
*, int))xfs_efd_item_unpin
,
535 .iop_unpin_remove
= (void(*)(xfs_log_item_t
*, xfs_trans_t
*))
536 xfs_efd_item_unpin_remove
,
537 .iop_trylock
= (uint(*)(xfs_log_item_t
*))xfs_efd_item_trylock
,
538 .iop_unlock
= (void(*)(xfs_log_item_t
*))xfs_efd_item_unlock
,
539 .iop_committed
= (xfs_lsn_t(*)(xfs_log_item_t
*, xfs_lsn_t
))
540 xfs_efd_item_committed
,
541 .iop_push
= (void(*)(xfs_log_item_t
*))xfs_efd_item_push
,
543 .iop_committing
= (void(*)(xfs_log_item_t
*, xfs_lsn_t
))
544 xfs_efd_item_committing
549 * Allocate and initialize an efd item with the given number of extents.
552 xfs_efd_init(xfs_mount_t
*mp
,
553 xfs_efi_log_item_t
*efip
,
557 xfs_efd_log_item_t
*efdp
;
560 ASSERT(nextents
> 0);
561 if (nextents
> XFS_EFD_MAX_FAST_EXTENTS
) {
562 size
= (uint
)(sizeof(xfs_efd_log_item_t
) +
563 ((nextents
- 1) * sizeof(xfs_extent_t
)));
564 efdp
= (xfs_efd_log_item_t
*)kmem_zalloc(size
, KM_SLEEP
);
566 efdp
= (xfs_efd_log_item_t
*)kmem_zone_zalloc(xfs_efd_zone
,
570 efdp
->efd_item
.li_type
= XFS_LI_EFD
;
571 efdp
->efd_item
.li_ops
= &xfs_efd_item_ops
;
572 efdp
->efd_item
.li_mountp
= mp
;
573 efdp
->efd_efip
= efip
;
574 efdp
->efd_format
.efd_nextents
= nextents
;
575 efdp
->efd_format
.efd_efi_id
= efip
->efi_format
.efi_id
;