xref: /openbmc/linux/fs/xfs/xfs_extent_busy.c (revision 95e9fd10)
1 /*
2  * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
3  * Copyright (c) 2010 David Chinner.
4  * Copyright (c) 2011 Christoph Hellwig.
5  * All Rights Reserved.
6  *
7  * This program is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU General Public License as
9  * published by the Free Software Foundation.
10  *
11  * This program is distributed in the hope that it would be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14  * GNU General Public License for more details.
15  *
16  * You should have received a copy of the GNU General Public License
17  * along with this program; if not, write the Free Software Foundation,
18  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
19  */
20 #include "xfs.h"
21 #include "xfs_fs.h"
22 #include "xfs_types.h"
23 #include "xfs_log.h"
24 #include "xfs_trans.h"
25 #include "xfs_sb.h"
26 #include "xfs_ag.h"
27 #include "xfs_mount.h"
28 #include "xfs_bmap_btree.h"
29 #include "xfs_alloc.h"
30 #include "xfs_inode.h"
31 #include "xfs_extent_busy.h"
32 #include "xfs_trace.h"
33 
34 void
35 xfs_extent_busy_insert(
36 	struct xfs_trans	*tp,
37 	xfs_agnumber_t		agno,
38 	xfs_agblock_t		bno,
39 	xfs_extlen_t		len,
40 	unsigned int		flags)
41 {
42 	struct xfs_extent_busy	*new;
43 	struct xfs_extent_busy	*busyp;
44 	struct xfs_perag	*pag;
45 	struct rb_node		**rbp;
46 	struct rb_node		*parent = NULL;
47 
48 	new = kmem_zalloc(sizeof(struct xfs_extent_busy), KM_MAYFAIL);
49 	if (!new) {
50 		/*
51 		 * No Memory!  Since it is now not possible to track the free
52 		 * block, make this a synchronous transaction to insure that
53 		 * the block is not reused before this transaction commits.
54 		 */
55 		trace_xfs_extent_busy_enomem(tp->t_mountp, agno, bno, len);
56 		xfs_trans_set_sync(tp);
57 		return;
58 	}
59 
60 	new->agno = agno;
61 	new->bno = bno;
62 	new->length = len;
63 	INIT_LIST_HEAD(&new->list);
64 	new->flags = flags;
65 
66 	/* trace before insert to be able to see failed inserts */
67 	trace_xfs_extent_busy(tp->t_mountp, agno, bno, len);
68 
69 	pag = xfs_perag_get(tp->t_mountp, new->agno);
70 	spin_lock(&pag->pagb_lock);
71 	rbp = &pag->pagb_tree.rb_node;
72 	while (*rbp) {
73 		parent = *rbp;
74 		busyp = rb_entry(parent, struct xfs_extent_busy, rb_node);
75 
76 		if (new->bno < busyp->bno) {
77 			rbp = &(*rbp)->rb_left;
78 			ASSERT(new->bno + new->length <= busyp->bno);
79 		} else if (new->bno > busyp->bno) {
80 			rbp = &(*rbp)->rb_right;
81 			ASSERT(bno >= busyp->bno + busyp->length);
82 		} else {
83 			ASSERT(0);
84 		}
85 	}
86 
87 	rb_link_node(&new->rb_node, parent, rbp);
88 	rb_insert_color(&new->rb_node, &pag->pagb_tree);
89 
90 	list_add(&new->list, &tp->t_busy);
91 	spin_unlock(&pag->pagb_lock);
92 	xfs_perag_put(pag);
93 }
94 
95 /*
96  * Search for a busy extent within the range of the extent we are about to
97  * allocate.  You need to be holding the busy extent tree lock when calling
98  * xfs_extent_busy_search(). This function returns 0 for no overlapping busy
99  * extent, -1 for an overlapping but not exact busy extent, and 1 for an exact
100  * match. This is done so that a non-zero return indicates an overlap that
101  * will require a synchronous transaction, but it can still be
102  * used to distinguish between a partial or exact match.
103  */
104 int
105 xfs_extent_busy_search(
106 	struct xfs_mount	*mp,
107 	xfs_agnumber_t		agno,
108 	xfs_agblock_t		bno,
109 	xfs_extlen_t		len)
110 {
111 	struct xfs_perag	*pag;
112 	struct rb_node		*rbp;
113 	struct xfs_extent_busy	*busyp;
114 	int			match = 0;
115 
116 	pag = xfs_perag_get(mp, agno);
117 	spin_lock(&pag->pagb_lock);
118 
119 	rbp = pag->pagb_tree.rb_node;
120 
121 	/* find closest start bno overlap */
122 	while (rbp) {
123 		busyp = rb_entry(rbp, struct xfs_extent_busy, rb_node);
124 		if (bno < busyp->bno) {
125 			/* may overlap, but exact start block is lower */
126 			if (bno + len > busyp->bno)
127 				match = -1;
128 			rbp = rbp->rb_left;
129 		} else if (bno > busyp->bno) {
130 			/* may overlap, but exact start block is higher */
131 			if (bno < busyp->bno + busyp->length)
132 				match = -1;
133 			rbp = rbp->rb_right;
134 		} else {
135 			/* bno matches busyp, length determines exact match */
136 			match = (busyp->length == len) ? 1 : -1;
137 			break;
138 		}
139 	}
140 	spin_unlock(&pag->pagb_lock);
141 	xfs_perag_put(pag);
142 	return match;
143 }
144 
145 /*
146  * The found free extent [fbno, fend] overlaps part or all of the given busy
147  * extent.  If the overlap covers the beginning, the end, or all of the busy
148  * extent, the overlapping portion can be made unbusy and used for the
149  * allocation.  We can't split a busy extent because we can't modify a
150  * transaction/CIL context busy list, but we can update an entries block
151  * number or length.
152  *
153  * Returns true if the extent can safely be reused, or false if the search
154  * needs to be restarted.
155  */
156 STATIC bool
157 xfs_extent_busy_update_extent(
158 	struct xfs_mount	*mp,
159 	struct xfs_perag	*pag,
160 	struct xfs_extent_busy	*busyp,
161 	xfs_agblock_t		fbno,
162 	xfs_extlen_t		flen,
163 	bool			userdata)
164 {
165 	xfs_agblock_t		fend = fbno + flen;
166 	xfs_agblock_t		bbno = busyp->bno;
167 	xfs_agblock_t		bend = bbno + busyp->length;
168 
169 	/*
170 	 * This extent is currently being discarded.  Give the thread
171 	 * performing the discard a chance to mark the extent unbusy
172 	 * and retry.
173 	 */
174 	if (busyp->flags & XFS_EXTENT_BUSY_DISCARDED) {
175 		spin_unlock(&pag->pagb_lock);
176 		delay(1);
177 		spin_lock(&pag->pagb_lock);
178 		return false;
179 	}
180 
181 	/*
182 	 * If there is a busy extent overlapping a user allocation, we have
183 	 * no choice but to force the log and retry the search.
184 	 *
185 	 * Fortunately this does not happen during normal operation, but
186 	 * only if the filesystem is very low on space and has to dip into
187 	 * the AGFL for normal allocations.
188 	 */
189 	if (userdata)
190 		goto out_force_log;
191 
192 	if (bbno < fbno && bend > fend) {
193 		/*
194 		 * Case 1:
195 		 *    bbno           bend
196 		 *    +BBBBBBBBBBBBBBBBB+
197 		 *        +---------+
198 		 *        fbno   fend
199 		 */
200 
201 		/*
202 		 * We would have to split the busy extent to be able to track
203 		 * it correct, which we cannot do because we would have to
204 		 * modify the list of busy extents attached to the transaction
205 		 * or CIL context, which is immutable.
206 		 *
207 		 * Force out the log to clear the busy extent and retry the
208 		 * search.
209 		 */
210 		goto out_force_log;
211 	} else if (bbno >= fbno && bend <= fend) {
212 		/*
213 		 * Case 2:
214 		 *    bbno           bend
215 		 *    +BBBBBBBBBBBBBBBBB+
216 		 *    +-----------------+
217 		 *    fbno           fend
218 		 *
219 		 * Case 3:
220 		 *    bbno           bend
221 		 *    +BBBBBBBBBBBBBBBBB+
222 		 *    +--------------------------+
223 		 *    fbno                    fend
224 		 *
225 		 * Case 4:
226 		 *             bbno           bend
227 		 *             +BBBBBBBBBBBBBBBBB+
228 		 *    +--------------------------+
229 		 *    fbno                    fend
230 		 *
231 		 * Case 5:
232 		 *             bbno           bend
233 		 *             +BBBBBBBBBBBBBBBBB+
234 		 *    +-----------------------------------+
235 		 *    fbno                             fend
236 		 *
237 		 */
238 
239 		/*
240 		 * The busy extent is fully covered by the extent we are
241 		 * allocating, and can simply be removed from the rbtree.
242 		 * However we cannot remove it from the immutable list
243 		 * tracking busy extents in the transaction or CIL context,
244 		 * so set the length to zero to mark it invalid.
245 		 *
246 		 * We also need to restart the busy extent search from the
247 		 * tree root, because erasing the node can rearrange the
248 		 * tree topology.
249 		 */
250 		rb_erase(&busyp->rb_node, &pag->pagb_tree);
251 		busyp->length = 0;
252 		return false;
253 	} else if (fend < bend) {
254 		/*
255 		 * Case 6:
256 		 *              bbno           bend
257 		 *             +BBBBBBBBBBBBBBBBB+
258 		 *             +---------+
259 		 *             fbno   fend
260 		 *
261 		 * Case 7:
262 		 *             bbno           bend
263 		 *             +BBBBBBBBBBBBBBBBB+
264 		 *    +------------------+
265 		 *    fbno            fend
266 		 *
267 		 */
268 		busyp->bno = fend;
269 	} else if (bbno < fbno) {
270 		/*
271 		 * Case 8:
272 		 *    bbno           bend
273 		 *    +BBBBBBBBBBBBBBBBB+
274 		 *        +-------------+
275 		 *        fbno       fend
276 		 *
277 		 * Case 9:
278 		 *    bbno           bend
279 		 *    +BBBBBBBBBBBBBBBBB+
280 		 *        +----------------------+
281 		 *        fbno                fend
282 		 */
283 		busyp->length = fbno - busyp->bno;
284 	} else {
285 		ASSERT(0);
286 	}
287 
288 	trace_xfs_extent_busy_reuse(mp, pag->pag_agno, fbno, flen);
289 	return true;
290 
291 out_force_log:
292 	spin_unlock(&pag->pagb_lock);
293 	xfs_log_force(mp, XFS_LOG_SYNC);
294 	trace_xfs_extent_busy_force(mp, pag->pag_agno, fbno, flen);
295 	spin_lock(&pag->pagb_lock);
296 	return false;
297 }
298 
299 
300 /*
301  * For a given extent [fbno, flen], make sure we can reuse it safely.
302  */
303 void
304 xfs_extent_busy_reuse(
305 	struct xfs_mount	*mp,
306 	xfs_agnumber_t		agno,
307 	xfs_agblock_t		fbno,
308 	xfs_extlen_t		flen,
309 	bool			userdata)
310 {
311 	struct xfs_perag	*pag;
312 	struct rb_node		*rbp;
313 
314 	ASSERT(flen > 0);
315 
316 	pag = xfs_perag_get(mp, agno);
317 	spin_lock(&pag->pagb_lock);
318 restart:
319 	rbp = pag->pagb_tree.rb_node;
320 	while (rbp) {
321 		struct xfs_extent_busy *busyp =
322 			rb_entry(rbp, struct xfs_extent_busy, rb_node);
323 		xfs_agblock_t	bbno = busyp->bno;
324 		xfs_agblock_t	bend = bbno + busyp->length;
325 
326 		if (fbno + flen <= bbno) {
327 			rbp = rbp->rb_left;
328 			continue;
329 		} else if (fbno >= bend) {
330 			rbp = rbp->rb_right;
331 			continue;
332 		}
333 
334 		if (!xfs_extent_busy_update_extent(mp, pag, busyp, fbno, flen,
335 						  userdata))
336 			goto restart;
337 	}
338 	spin_unlock(&pag->pagb_lock);
339 	xfs_perag_put(pag);
340 }
341 
342 /*
343  * For a given extent [fbno, flen], search the busy extent list to find a
344  * subset of the extent that is not busy.  If *rlen is smaller than
345  * args->minlen no suitable extent could be found, and the higher level
346  * code needs to force out the log and retry the allocation.
347  */
348 void
349 xfs_extent_busy_trim(
350 	struct xfs_alloc_arg	*args,
351 	xfs_agblock_t		bno,
352 	xfs_extlen_t		len,
353 	xfs_agblock_t		*rbno,
354 	xfs_extlen_t		*rlen)
355 {
356 	xfs_agblock_t		fbno;
357 	xfs_extlen_t		flen;
358 	struct rb_node		*rbp;
359 
360 	ASSERT(len > 0);
361 
362 	spin_lock(&args->pag->pagb_lock);
363 restart:
364 	fbno = bno;
365 	flen = len;
366 	rbp = args->pag->pagb_tree.rb_node;
367 	while (rbp && flen >= args->minlen) {
368 		struct xfs_extent_busy *busyp =
369 			rb_entry(rbp, struct xfs_extent_busy, rb_node);
370 		xfs_agblock_t	fend = fbno + flen;
371 		xfs_agblock_t	bbno = busyp->bno;
372 		xfs_agblock_t	bend = bbno + busyp->length;
373 
374 		if (fend <= bbno) {
375 			rbp = rbp->rb_left;
376 			continue;
377 		} else if (fbno >= bend) {
378 			rbp = rbp->rb_right;
379 			continue;
380 		}
381 
382 		/*
383 		 * If this is a metadata allocation, try to reuse the busy
384 		 * extent instead of trimming the allocation.
385 		 */
386 		if (!args->userdata &&
387 		    !(busyp->flags & XFS_EXTENT_BUSY_DISCARDED)) {
388 			if (!xfs_extent_busy_update_extent(args->mp, args->pag,
389 							  busyp, fbno, flen,
390 							  false))
391 				goto restart;
392 			continue;
393 		}
394 
395 		if (bbno <= fbno) {
396 			/* start overlap */
397 
398 			/*
399 			 * Case 1:
400 			 *    bbno           bend
401 			 *    +BBBBBBBBBBBBBBBBB+
402 			 *        +---------+
403 			 *        fbno   fend
404 			 *
405 			 * Case 2:
406 			 *    bbno           bend
407 			 *    +BBBBBBBBBBBBBBBBB+
408 			 *    +-------------+
409 			 *    fbno       fend
410 			 *
411 			 * Case 3:
412 			 *    bbno           bend
413 			 *    +BBBBBBBBBBBBBBBBB+
414 			 *        +-------------+
415 			 *        fbno       fend
416 			 *
417 			 * Case 4:
418 			 *    bbno           bend
419 			 *    +BBBBBBBBBBBBBBBBB+
420 			 *    +-----------------+
421 			 *    fbno           fend
422 			 *
423 			 * No unbusy region in extent, return failure.
424 			 */
425 			if (fend <= bend)
426 				goto fail;
427 
428 			/*
429 			 * Case 5:
430 			 *    bbno           bend
431 			 *    +BBBBBBBBBBBBBBBBB+
432 			 *        +----------------------+
433 			 *        fbno                fend
434 			 *
435 			 * Case 6:
436 			 *    bbno           bend
437 			 *    +BBBBBBBBBBBBBBBBB+
438 			 *    +--------------------------+
439 			 *    fbno                    fend
440 			 *
441 			 * Needs to be trimmed to:
442 			 *                       +-------+
443 			 *                       fbno fend
444 			 */
445 			fbno = bend;
446 		} else if (bend >= fend) {
447 			/* end overlap */
448 
449 			/*
450 			 * Case 7:
451 			 *             bbno           bend
452 			 *             +BBBBBBBBBBBBBBBBB+
453 			 *    +------------------+
454 			 *    fbno            fend
455 			 *
456 			 * Case 8:
457 			 *             bbno           bend
458 			 *             +BBBBBBBBBBBBBBBBB+
459 			 *    +--------------------------+
460 			 *    fbno                    fend
461 			 *
462 			 * Needs to be trimmed to:
463 			 *    +-------+
464 			 *    fbno fend
465 			 */
466 			fend = bbno;
467 		} else {
468 			/* middle overlap */
469 
470 			/*
471 			 * Case 9:
472 			 *             bbno           bend
473 			 *             +BBBBBBBBBBBBBBBBB+
474 			 *    +-----------------------------------+
475 			 *    fbno                             fend
476 			 *
477 			 * Can be trimmed to:
478 			 *    +-------+        OR         +-------+
479 			 *    fbno fend                   fbno fend
480 			 *
481 			 * Backward allocation leads to significant
482 			 * fragmentation of directories, which degrades
483 			 * directory performance, therefore we always want to
484 			 * choose the option that produces forward allocation
485 			 * patterns.
486 			 * Preferring the lower bno extent will make the next
487 			 * request use "fend" as the start of the next
488 			 * allocation;  if the segment is no longer busy at
489 			 * that point, we'll get a contiguous allocation, but
490 			 * even if it is still busy, we will get a forward
491 			 * allocation.
492 			 * We try to avoid choosing the segment at "bend",
493 			 * because that can lead to the next allocation
494 			 * taking the segment at "fbno", which would be a
495 			 * backward allocation.  We only use the segment at
496 			 * "fbno" if it is much larger than the current
497 			 * requested size, because in that case there's a
498 			 * good chance subsequent allocations will be
499 			 * contiguous.
500 			 */
501 			if (bbno - fbno >= args->maxlen) {
502 				/* left candidate fits perfect */
503 				fend = bbno;
504 			} else if (fend - bend >= args->maxlen * 4) {
505 				/* right candidate has enough free space */
506 				fbno = bend;
507 			} else if (bbno - fbno >= args->minlen) {
508 				/* left candidate fits minimum requirement */
509 				fend = bbno;
510 			} else {
511 				goto fail;
512 			}
513 		}
514 
515 		flen = fend - fbno;
516 	}
517 	spin_unlock(&args->pag->pagb_lock);
518 
519 	if (fbno != bno || flen != len) {
520 		trace_xfs_extent_busy_trim(args->mp, args->agno, bno, len,
521 					  fbno, flen);
522 	}
523 	*rbno = fbno;
524 	*rlen = flen;
525 	return;
526 fail:
527 	/*
528 	 * Return a zero extent length as failure indications.  All callers
529 	 * re-check if the trimmed extent satisfies the minlen requirement.
530 	 */
531 	spin_unlock(&args->pag->pagb_lock);
532 	trace_xfs_extent_busy_trim(args->mp, args->agno, bno, len, fbno, 0);
533 	*rbno = fbno;
534 	*rlen = 0;
535 }
536 
537 STATIC void
538 xfs_extent_busy_clear_one(
539 	struct xfs_mount	*mp,
540 	struct xfs_perag	*pag,
541 	struct xfs_extent_busy	*busyp)
542 {
543 	if (busyp->length) {
544 		trace_xfs_extent_busy_clear(mp, busyp->agno, busyp->bno,
545 						busyp->length);
546 		rb_erase(&busyp->rb_node, &pag->pagb_tree);
547 	}
548 
549 	list_del_init(&busyp->list);
550 	kmem_free(busyp);
551 }
552 
553 /*
554  * Remove all extents on the passed in list from the busy extents tree.
555  * If do_discard is set skip extents that need to be discarded, and mark
556  * these as undergoing a discard operation instead.
557  */
558 void
559 xfs_extent_busy_clear(
560 	struct xfs_mount	*mp,
561 	struct list_head	*list,
562 	bool			do_discard)
563 {
564 	struct xfs_extent_busy	*busyp, *n;
565 	struct xfs_perag	*pag = NULL;
566 	xfs_agnumber_t		agno = NULLAGNUMBER;
567 
568 	list_for_each_entry_safe(busyp, n, list, list) {
569 		if (busyp->agno != agno) {
570 			if (pag) {
571 				spin_unlock(&pag->pagb_lock);
572 				xfs_perag_put(pag);
573 			}
574 			pag = xfs_perag_get(mp, busyp->agno);
575 			spin_lock(&pag->pagb_lock);
576 			agno = busyp->agno;
577 		}
578 
579 		if (do_discard && busyp->length &&
580 		    !(busyp->flags & XFS_EXTENT_BUSY_SKIP_DISCARD))
581 			busyp->flags = XFS_EXTENT_BUSY_DISCARDED;
582 		else
583 			xfs_extent_busy_clear_one(mp, pag, busyp);
584 	}
585 
586 	if (pag) {
587 		spin_unlock(&pag->pagb_lock);
588 		xfs_perag_put(pag);
589 	}
590 }
591 
592 /*
593  * Callback for list_sort to sort busy extents by the AG they reside in.
594  */
595 int
596 xfs_extent_busy_ag_cmp(
597 	void			*priv,
598 	struct list_head	*a,
599 	struct list_head	*b)
600 {
601 	return container_of(a, struct xfs_extent_busy, list)->agno -
602 		container_of(b, struct xfs_extent_busy, list)->agno;
603 }
604