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_format.h"
21 #include "xfs_log_format.h"
22 #include "xfs_trans_resv.h"
23 #include "xfs_mount.h"
24 #include "xfs_trans.h"
25 #include "xfs_trans_priv.h"
26 #include "xfs_buf_item.h"
27 #include "xfs_extfree_item.h"
31 kmem_zone_t
*xfs_efi_zone
;
32 kmem_zone_t
*xfs_efd_zone
;
34 static inline struct xfs_efi_log_item
*EFI_ITEM(struct xfs_log_item
*lip
)
36 return container_of(lip
, struct xfs_efi_log_item
, efi_item
);
41 struct xfs_efi_log_item
*efip
)
43 if (efip
->efi_format
.efi_nextents
> XFS_EFI_MAX_FAST_EXTENTS
)
46 kmem_zone_free(xfs_efi_zone
, efip
);
50 * Freeing the efi requires that we remove it from the AIL if it has already
51 * been placed there. However, the EFI may not yet have been placed in the AIL
52 * when called by xfs_efi_release() from EFD processing due to the ordering of
53 * committed vs unpin operations in bulk insert operations. Hence the reference
54 * count to ensure only the last caller frees the EFI.
58 struct xfs_efi_log_item
*efip
)
60 struct xfs_ail
*ailp
= efip
->efi_item
.li_ailp
;
62 if (atomic_dec_and_test(&efip
->efi_refcount
)) {
63 spin_lock(&ailp
->xa_lock
);
64 /* xfs_trans_ail_delete() drops the AIL lock. */
65 xfs_trans_ail_delete(ailp
, &efip
->efi_item
,
66 SHUTDOWN_LOG_IO_ERROR
);
67 xfs_efi_item_free(efip
);
72 * This returns the number of iovecs needed to log the given efi item.
73 * We only need 1 iovec for an efi item. It just logs the efi_log_format
78 struct xfs_efi_log_item
*efip
)
80 return sizeof(struct xfs_efi_log_format
) +
81 (efip
->efi_format
.efi_nextents
- 1) * sizeof(xfs_extent_t
);
86 struct xfs_log_item
*lip
,
91 *nbytes
+= xfs_efi_item_sizeof(EFI_ITEM(lip
));
95 * This is called to fill in the vector of log iovecs for the
96 * given efi log item. We use only 1 iovec, and we point that
97 * at the efi_log_format structure embedded in the efi item.
98 * It is at this point that we assert that all of the extent
99 * slots in the efi item have been filled.
103 struct xfs_log_item
*lip
,
104 struct xfs_log_vec
*lv
)
106 struct xfs_efi_log_item
*efip
= EFI_ITEM(lip
);
107 struct xfs_log_iovec
*vecp
= NULL
;
109 ASSERT(atomic_read(&efip
->efi_next_extent
) ==
110 efip
->efi_format
.efi_nextents
);
112 efip
->efi_format
.efi_type
= XFS_LI_EFI
;
113 efip
->efi_format
.efi_size
= 1;
115 xlog_copy_iovec(lv
, &vecp
, XLOG_REG_TYPE_EFI_FORMAT
,
117 xfs_efi_item_sizeof(efip
));
122 * Pinning has no meaning for an efi item, so just return.
126 struct xfs_log_item
*lip
)
131 * While EFIs cannot really be pinned, the unpin operation is the last place at
132 * which the EFI is manipulated during a transaction. If we are being asked to
133 * remove the EFI it's because the transaction has been cancelled and by
134 * definition that means the EFI cannot be in the AIL so remove it from the
135 * transaction and free it. Otherwise coordinate with xfs_efi_release()
136 * to determine who gets to free the EFI.
140 struct xfs_log_item
*lip
,
143 struct xfs_efi_log_item
*efip
= EFI_ITEM(lip
);
146 ASSERT(!(lip
->li_flags
& XFS_LI_IN_AIL
));
148 xfs_trans_del_item(lip
);
149 xfs_efi_item_free(efip
);
152 __xfs_efi_release(efip
);
156 * Efi items have no locking or pushing. However, since EFIs are pulled from
157 * the AIL when their corresponding EFDs are committed to disk, their situation
158 * is very similar to being pinned. Return XFS_ITEM_PINNED so that the caller
159 * will eventually flush the log. This should help in getting the EFI out of
164 struct xfs_log_item
*lip
,
165 struct list_head
*buffer_list
)
167 return XFS_ITEM_PINNED
;
172 struct xfs_log_item
*lip
)
174 if (lip
->li_flags
& XFS_LI_ABORTED
)
175 xfs_efi_item_free(EFI_ITEM(lip
));
179 * The EFI is logged only once and cannot be moved in the log, so simply return
180 * the lsn at which it's been logged.
183 xfs_efi_item_committed(
184 struct xfs_log_item
*lip
,
191 * The EFI dependency tracking op doesn't do squat. It can't because
192 * it doesn't know where the free extent is coming from. The dependency
193 * tracking has to be handled by the "enclosing" metadata object. For
194 * example, for inodes, the inode is locked throughout the extent freeing
195 * so the dependency should be recorded there.
198 xfs_efi_item_committing(
199 struct xfs_log_item
*lip
,
205 * This is the ops vector shared by all efi log items.
207 static const struct xfs_item_ops xfs_efi_item_ops
= {
208 .iop_size
= xfs_efi_item_size
,
209 .iop_format
= xfs_efi_item_format
,
210 .iop_pin
= xfs_efi_item_pin
,
211 .iop_unpin
= xfs_efi_item_unpin
,
212 .iop_unlock
= xfs_efi_item_unlock
,
213 .iop_committed
= xfs_efi_item_committed
,
214 .iop_push
= xfs_efi_item_push
,
215 .iop_committing
= xfs_efi_item_committing
220 * Allocate and initialize an efi item with the given number of extents.
222 struct xfs_efi_log_item
*
224 struct xfs_mount
*mp
,
228 struct xfs_efi_log_item
*efip
;
231 ASSERT(nextents
> 0);
232 if (nextents
> XFS_EFI_MAX_FAST_EXTENTS
) {
233 size
= (uint
)(sizeof(xfs_efi_log_item_t
) +
234 ((nextents
- 1) * sizeof(xfs_extent_t
)));
235 efip
= kmem_zalloc(size
, KM_SLEEP
);
237 efip
= kmem_zone_zalloc(xfs_efi_zone
, KM_SLEEP
);
240 xfs_log_item_init(mp
, &efip
->efi_item
, XFS_LI_EFI
, &xfs_efi_item_ops
);
241 efip
->efi_format
.efi_nextents
= nextents
;
242 efip
->efi_format
.efi_id
= (uintptr_t)(void *)efip
;
243 atomic_set(&efip
->efi_next_extent
, 0);
244 atomic_set(&efip
->efi_refcount
, 2);
250 * Copy an EFI format buffer from the given buf, and into the destination
251 * EFI format structure.
252 * The given buffer can be in 32 bit or 64 bit form (which has different padding),
253 * one of which will be the native format for this kernel.
254 * It will handle the conversion of formats if necessary.
257 xfs_efi_copy_format(xfs_log_iovec_t
*buf
, xfs_efi_log_format_t
*dst_efi_fmt
)
259 xfs_efi_log_format_t
*src_efi_fmt
= buf
->i_addr
;
261 uint len
= sizeof(xfs_efi_log_format_t
) +
262 (src_efi_fmt
->efi_nextents
- 1) * sizeof(xfs_extent_t
);
263 uint len32
= sizeof(xfs_efi_log_format_32_t
) +
264 (src_efi_fmt
->efi_nextents
- 1) * sizeof(xfs_extent_32_t
);
265 uint len64
= sizeof(xfs_efi_log_format_64_t
) +
266 (src_efi_fmt
->efi_nextents
- 1) * sizeof(xfs_extent_64_t
);
268 if (buf
->i_len
== len
) {
269 memcpy((char *)dst_efi_fmt
, (char*)src_efi_fmt
, len
);
271 } else if (buf
->i_len
== len32
) {
272 xfs_efi_log_format_32_t
*src_efi_fmt_32
= buf
->i_addr
;
274 dst_efi_fmt
->efi_type
= src_efi_fmt_32
->efi_type
;
275 dst_efi_fmt
->efi_size
= src_efi_fmt_32
->efi_size
;
276 dst_efi_fmt
->efi_nextents
= src_efi_fmt_32
->efi_nextents
;
277 dst_efi_fmt
->efi_id
= src_efi_fmt_32
->efi_id
;
278 for (i
= 0; i
< dst_efi_fmt
->efi_nextents
; i
++) {
279 dst_efi_fmt
->efi_extents
[i
].ext_start
=
280 src_efi_fmt_32
->efi_extents
[i
].ext_start
;
281 dst_efi_fmt
->efi_extents
[i
].ext_len
=
282 src_efi_fmt_32
->efi_extents
[i
].ext_len
;
285 } else if (buf
->i_len
== len64
) {
286 xfs_efi_log_format_64_t
*src_efi_fmt_64
= buf
->i_addr
;
288 dst_efi_fmt
->efi_type
= src_efi_fmt_64
->efi_type
;
289 dst_efi_fmt
->efi_size
= src_efi_fmt_64
->efi_size
;
290 dst_efi_fmt
->efi_nextents
= src_efi_fmt_64
->efi_nextents
;
291 dst_efi_fmt
->efi_id
= src_efi_fmt_64
->efi_id
;
292 for (i
= 0; i
< dst_efi_fmt
->efi_nextents
; i
++) {
293 dst_efi_fmt
->efi_extents
[i
].ext_start
=
294 src_efi_fmt_64
->efi_extents
[i
].ext_start
;
295 dst_efi_fmt
->efi_extents
[i
].ext_len
=
296 src_efi_fmt_64
->efi_extents
[i
].ext_len
;
300 return -EFSCORRUPTED
;
304 * This is called by the efd item code below to release references to the given
305 * efi item. Each efd calls this with the number of extents that it has
306 * logged, and when the sum of these reaches the total number of extents logged
307 * by this efi item we can free the efi item.
310 xfs_efi_release(xfs_efi_log_item_t
*efip
,
313 ASSERT(atomic_read(&efip
->efi_next_extent
) >= nextents
);
314 if (atomic_sub_and_test(nextents
, &efip
->efi_next_extent
)) {
315 /* recovery needs us to drop the EFI reference, too */
316 if (test_bit(XFS_EFI_RECOVERED
, &efip
->efi_flags
))
317 __xfs_efi_release(efip
);
319 __xfs_efi_release(efip
);
320 /* efip may now have been freed, do not reference it again. */
324 static inline struct xfs_efd_log_item
*EFD_ITEM(struct xfs_log_item
*lip
)
326 return container_of(lip
, struct xfs_efd_log_item
, efd_item
);
330 xfs_efd_item_free(struct xfs_efd_log_item
*efdp
)
332 if (efdp
->efd_format
.efd_nextents
> XFS_EFD_MAX_FAST_EXTENTS
)
335 kmem_zone_free(xfs_efd_zone
, efdp
);
339 * This returns the number of iovecs needed to log the given efd item.
340 * We only need 1 iovec for an efd item. It just logs the efd_log_format
345 struct xfs_efd_log_item
*efdp
)
347 return sizeof(xfs_efd_log_format_t
) +
348 (efdp
->efd_format
.efd_nextents
- 1) * sizeof(xfs_extent_t
);
353 struct xfs_log_item
*lip
,
358 *nbytes
+= xfs_efd_item_sizeof(EFD_ITEM(lip
));
362 * This is called to fill in the vector of log iovecs for the
363 * given efd log item. We use only 1 iovec, and we point that
364 * at the efd_log_format structure embedded in the efd item.
365 * It is at this point that we assert that all of the extent
366 * slots in the efd item have been filled.
370 struct xfs_log_item
*lip
,
371 struct xfs_log_vec
*lv
)
373 struct xfs_efd_log_item
*efdp
= EFD_ITEM(lip
);
374 struct xfs_log_iovec
*vecp
= NULL
;
376 ASSERT(efdp
->efd_next_extent
== efdp
->efd_format
.efd_nextents
);
378 efdp
->efd_format
.efd_type
= XFS_LI_EFD
;
379 efdp
->efd_format
.efd_size
= 1;
381 xlog_copy_iovec(lv
, &vecp
, XLOG_REG_TYPE_EFD_FORMAT
,
383 xfs_efd_item_sizeof(efdp
));
387 * Pinning has no meaning for an efd item, so just return.
391 struct xfs_log_item
*lip
)
396 * Since pinning has no meaning for an efd item, unpinning does
401 struct xfs_log_item
*lip
,
407 * There isn't much you can do to push on an efd item. It is simply stuck
408 * waiting for the log to be flushed to disk.
412 struct xfs_log_item
*lip
,
413 struct list_head
*buffer_list
)
415 return XFS_ITEM_PINNED
;
420 struct xfs_log_item
*lip
)
422 if (lip
->li_flags
& XFS_LI_ABORTED
)
423 xfs_efd_item_free(EFD_ITEM(lip
));
427 * When the efd item is committed to disk, all we need to do
428 * is delete our reference to our partner efi item and then
429 * free ourselves. Since we're freeing ourselves we must
430 * return -1 to keep the transaction code from further referencing
434 xfs_efd_item_committed(
435 struct xfs_log_item
*lip
,
438 struct xfs_efd_log_item
*efdp
= EFD_ITEM(lip
);
441 * If we got a log I/O error, it's always the case that the LR with the
442 * EFI got unpinned and freed before the EFD got aborted.
444 if (!(lip
->li_flags
& XFS_LI_ABORTED
))
445 xfs_efi_release(efdp
->efd_efip
, efdp
->efd_format
.efd_nextents
);
447 xfs_efd_item_free(efdp
);
448 return (xfs_lsn_t
)-1;
452 * The EFD dependency tracking op doesn't do squat. It can't because
453 * it doesn't know where the free extent is coming from. The dependency
454 * tracking has to be handled by the "enclosing" metadata object. For
455 * example, for inodes, the inode is locked throughout the extent freeing
456 * so the dependency should be recorded there.
459 xfs_efd_item_committing(
460 struct xfs_log_item
*lip
,
466 * This is the ops vector shared by all efd log items.
468 static const struct xfs_item_ops xfs_efd_item_ops
= {
469 .iop_size
= xfs_efd_item_size
,
470 .iop_format
= xfs_efd_item_format
,
471 .iop_pin
= xfs_efd_item_pin
,
472 .iop_unpin
= xfs_efd_item_unpin
,
473 .iop_unlock
= xfs_efd_item_unlock
,
474 .iop_committed
= xfs_efd_item_committed
,
475 .iop_push
= xfs_efd_item_push
,
476 .iop_committing
= xfs_efd_item_committing
480 * Allocate and initialize an efd item with the given number of extents.
482 struct xfs_efd_log_item
*
484 struct xfs_mount
*mp
,
485 struct xfs_efi_log_item
*efip
,
489 struct xfs_efd_log_item
*efdp
;
492 ASSERT(nextents
> 0);
493 if (nextents
> XFS_EFD_MAX_FAST_EXTENTS
) {
494 size
= (uint
)(sizeof(xfs_efd_log_item_t
) +
495 ((nextents
- 1) * sizeof(xfs_extent_t
)));
496 efdp
= kmem_zalloc(size
, KM_SLEEP
);
498 efdp
= kmem_zone_zalloc(xfs_efd_zone
, KM_SLEEP
);
501 xfs_log_item_init(mp
, &efdp
->efd_item
, XFS_LI_EFD
, &xfs_efd_item_ops
);
502 efdp
->efd_efip
= efip
;
503 efdp
->efd_format
.efd_nextents
= nextents
;
504 efdp
->efd_format
.efd_efi_id
= efip
->efi_format
.efi_id
;