xref: /linux/fs/xfs/xfs_extent_busy.c (revision 3f0a50f345f78183f6e9b39c2f45ca5dcaa511ca)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
4  * Copyright (c) 2010 David Chinner.
5  * Copyright (c) 2011 Christoph Hellwig.
6  * All Rights Reserved.
7  */
8 #include "xfs.h"
9 #include "xfs_fs.h"
10 #include "xfs_format.h"
11 #include "xfs_log_format.h"
12 #include "xfs_shared.h"
13 #include "xfs_trans_resv.h"
14 #include "xfs_mount.h"
15 #include "xfs_alloc.h"
16 #include "xfs_extent_busy.h"
17 #include "xfs_trace.h"
18 #include "xfs_trans.h"
19 #include "xfs_log.h"
20 #include "xfs_ag.h"
21 
22 void
23 xfs_extent_busy_insert(
24 	struct xfs_trans	*tp,
25 	struct xfs_perag	*pag,
26 	xfs_agblock_t		bno,
27 	xfs_extlen_t		len,
28 	unsigned int		flags)
29 {
30 	struct xfs_extent_busy	*new;
31 	struct xfs_extent_busy	*busyp;
32 	struct rb_node		**rbp;
33 	struct rb_node		*parent = NULL;
34 
35 	new = kmem_zalloc(sizeof(struct xfs_extent_busy), 0);
36 	new->agno = pag->pag_agno;
37 	new->bno = bno;
38 	new->length = len;
39 	INIT_LIST_HEAD(&new->list);
40 	new->flags = flags;
41 
42 	/* trace before insert to be able to see failed inserts */
43 	trace_xfs_extent_busy(tp->t_mountp, pag->pag_agno, bno, len);
44 
45 	spin_lock(&pag->pagb_lock);
46 	rbp = &pag->pagb_tree.rb_node;
47 	while (*rbp) {
48 		parent = *rbp;
49 		busyp = rb_entry(parent, struct xfs_extent_busy, rb_node);
50 
51 		if (new->bno < busyp->bno) {
52 			rbp = &(*rbp)->rb_left;
53 			ASSERT(new->bno + new->length <= busyp->bno);
54 		} else if (new->bno > busyp->bno) {
55 			rbp = &(*rbp)->rb_right;
56 			ASSERT(bno >= busyp->bno + busyp->length);
57 		} else {
58 			ASSERT(0);
59 		}
60 	}
61 
62 	rb_link_node(&new->rb_node, parent, rbp);
63 	rb_insert_color(&new->rb_node, &pag->pagb_tree);
64 
65 	list_add(&new->list, &tp->t_busy);
66 	spin_unlock(&pag->pagb_lock);
67 }
68 
69 /*
70  * Search for a busy extent within the range of the extent we are about to
71  * allocate.  You need to be holding the busy extent tree lock when calling
72  * xfs_extent_busy_search(). This function returns 0 for no overlapping busy
73  * extent, -1 for an overlapping but not exact busy extent, and 1 for an exact
74  * match. This is done so that a non-zero return indicates an overlap that
75  * will require a synchronous transaction, but it can still be
76  * used to distinguish between a partial or exact match.
77  */
78 int
79 xfs_extent_busy_search(
80 	struct xfs_mount	*mp,
81 	struct xfs_perag	*pag,
82 	xfs_agblock_t		bno,
83 	xfs_extlen_t		len)
84 {
85 	struct rb_node		*rbp;
86 	struct xfs_extent_busy	*busyp;
87 	int			match = 0;
88 
89 	/* find closest start bno overlap */
90 	spin_lock(&pag->pagb_lock);
91 	rbp = pag->pagb_tree.rb_node;
92 	while (rbp) {
93 		busyp = rb_entry(rbp, struct xfs_extent_busy, rb_node);
94 		if (bno < busyp->bno) {
95 			/* may overlap, but exact start block is lower */
96 			if (bno + len > busyp->bno)
97 				match = -1;
98 			rbp = rbp->rb_left;
99 		} else if (bno > busyp->bno) {
100 			/* may overlap, but exact start block is higher */
101 			if (bno < busyp->bno + busyp->length)
102 				match = -1;
103 			rbp = rbp->rb_right;
104 		} else {
105 			/* bno matches busyp, length determines exact match */
106 			match = (busyp->length == len) ? 1 : -1;
107 			break;
108 		}
109 	}
110 	spin_unlock(&pag->pagb_lock);
111 	return match;
112 }
113 
114 /*
115  * The found free extent [fbno, fend] overlaps part or all of the given busy
116  * extent.  If the overlap covers the beginning, the end, or all of the busy
117  * extent, the overlapping portion can be made unbusy and used for the
118  * allocation.  We can't split a busy extent because we can't modify a
119  * transaction/CIL context busy list, but we can update an entry's block
120  * number or length.
121  *
122  * Returns true if the extent can safely be reused, or false if the search
123  * needs to be restarted.
124  */
125 STATIC bool
126 xfs_extent_busy_update_extent(
127 	struct xfs_mount	*mp,
128 	struct xfs_perag	*pag,
129 	struct xfs_extent_busy	*busyp,
130 	xfs_agblock_t		fbno,
131 	xfs_extlen_t		flen,
132 	bool			userdata) __releases(&pag->pagb_lock)
133 					  __acquires(&pag->pagb_lock)
134 {
135 	xfs_agblock_t		fend = fbno + flen;
136 	xfs_agblock_t		bbno = busyp->bno;
137 	xfs_agblock_t		bend = bbno + busyp->length;
138 
139 	/*
140 	 * This extent is currently being discarded.  Give the thread
141 	 * performing the discard a chance to mark the extent unbusy
142 	 * and retry.
143 	 */
144 	if (busyp->flags & XFS_EXTENT_BUSY_DISCARDED) {
145 		spin_unlock(&pag->pagb_lock);
146 		delay(1);
147 		spin_lock(&pag->pagb_lock);
148 		return false;
149 	}
150 
151 	/*
152 	 * If there is a busy extent overlapping a user allocation, we have
153 	 * no choice but to force the log and retry the search.
154 	 *
155 	 * Fortunately this does not happen during normal operation, but
156 	 * only if the filesystem is very low on space and has to dip into
157 	 * the AGFL for normal allocations.
158 	 */
159 	if (userdata)
160 		goto out_force_log;
161 
162 	if (bbno < fbno && bend > fend) {
163 		/*
164 		 * Case 1:
165 		 *    bbno           bend
166 		 *    +BBBBBBBBBBBBBBBBB+
167 		 *        +---------+
168 		 *        fbno   fend
169 		 */
170 
171 		/*
172 		 * We would have to split the busy extent to be able to track
173 		 * it correct, which we cannot do because we would have to
174 		 * modify the list of busy extents attached to the transaction
175 		 * or CIL context, which is immutable.
176 		 *
177 		 * Force out the log to clear the busy extent and retry the
178 		 * search.
179 		 */
180 		goto out_force_log;
181 	} else if (bbno >= fbno && bend <= fend) {
182 		/*
183 		 * Case 2:
184 		 *    bbno           bend
185 		 *    +BBBBBBBBBBBBBBBBB+
186 		 *    +-----------------+
187 		 *    fbno           fend
188 		 *
189 		 * Case 3:
190 		 *    bbno           bend
191 		 *    +BBBBBBBBBBBBBBBBB+
192 		 *    +--------------------------+
193 		 *    fbno                    fend
194 		 *
195 		 * Case 4:
196 		 *             bbno           bend
197 		 *             +BBBBBBBBBBBBBBBBB+
198 		 *    +--------------------------+
199 		 *    fbno                    fend
200 		 *
201 		 * Case 5:
202 		 *             bbno           bend
203 		 *             +BBBBBBBBBBBBBBBBB+
204 		 *    +-----------------------------------+
205 		 *    fbno                             fend
206 		 *
207 		 */
208 
209 		/*
210 		 * The busy extent is fully covered by the extent we are
211 		 * allocating, and can simply be removed from the rbtree.
212 		 * However we cannot remove it from the immutable list
213 		 * tracking busy extents in the transaction or CIL context,
214 		 * so set the length to zero to mark it invalid.
215 		 *
216 		 * We also need to restart the busy extent search from the
217 		 * tree root, because erasing the node can rearrange the
218 		 * tree topology.
219 		 */
220 		rb_erase(&busyp->rb_node, &pag->pagb_tree);
221 		busyp->length = 0;
222 		return false;
223 	} else if (fend < bend) {
224 		/*
225 		 * Case 6:
226 		 *              bbno           bend
227 		 *             +BBBBBBBBBBBBBBBBB+
228 		 *             +---------+
229 		 *             fbno   fend
230 		 *
231 		 * Case 7:
232 		 *             bbno           bend
233 		 *             +BBBBBBBBBBBBBBBBB+
234 		 *    +------------------+
235 		 *    fbno            fend
236 		 *
237 		 */
238 		busyp->bno = fend;
239 	} else if (bbno < fbno) {
240 		/*
241 		 * Case 8:
242 		 *    bbno           bend
243 		 *    +BBBBBBBBBBBBBBBBB+
244 		 *        +-------------+
245 		 *        fbno       fend
246 		 *
247 		 * Case 9:
248 		 *    bbno           bend
249 		 *    +BBBBBBBBBBBBBBBBB+
250 		 *        +----------------------+
251 		 *        fbno                fend
252 		 */
253 		busyp->length = fbno - busyp->bno;
254 	} else {
255 		ASSERT(0);
256 	}
257 
258 	trace_xfs_extent_busy_reuse(mp, pag->pag_agno, fbno, flen);
259 	return true;
260 
261 out_force_log:
262 	spin_unlock(&pag->pagb_lock);
263 	xfs_log_force(mp, XFS_LOG_SYNC);
264 	trace_xfs_extent_busy_force(mp, pag->pag_agno, fbno, flen);
265 	spin_lock(&pag->pagb_lock);
266 	return false;
267 }
268 
269 
270 /*
271  * For a given extent [fbno, flen], make sure we can reuse it safely.
272  */
273 void
274 xfs_extent_busy_reuse(
275 	struct xfs_mount	*mp,
276 	struct xfs_perag	*pag,
277 	xfs_agblock_t		fbno,
278 	xfs_extlen_t		flen,
279 	bool			userdata)
280 {
281 	struct rb_node		*rbp;
282 
283 	ASSERT(flen > 0);
284 	spin_lock(&pag->pagb_lock);
285 restart:
286 	rbp = pag->pagb_tree.rb_node;
287 	while (rbp) {
288 		struct xfs_extent_busy *busyp =
289 			rb_entry(rbp, struct xfs_extent_busy, rb_node);
290 		xfs_agblock_t	bbno = busyp->bno;
291 		xfs_agblock_t	bend = bbno + busyp->length;
292 
293 		if (fbno + flen <= bbno) {
294 			rbp = rbp->rb_left;
295 			continue;
296 		} else if (fbno >= bend) {
297 			rbp = rbp->rb_right;
298 			continue;
299 		}
300 
301 		if (!xfs_extent_busy_update_extent(mp, pag, busyp, fbno, flen,
302 						  userdata))
303 			goto restart;
304 	}
305 	spin_unlock(&pag->pagb_lock);
306 }
307 
308 /*
309  * For a given extent [fbno, flen], search the busy extent list to find a
310  * subset of the extent that is not busy.  If *rlen is smaller than
311  * args->minlen no suitable extent could be found, and the higher level
312  * code needs to force out the log and retry the allocation.
313  *
314  * Return the current busy generation for the AG if the extent is busy. This
315  * value can be used to wait for at least one of the currently busy extents
316  * to be cleared. Note that the busy list is not guaranteed to be empty after
317  * the gen is woken. The state of a specific extent must always be confirmed
318  * with another call to xfs_extent_busy_trim() before it can be used.
319  */
320 bool
321 xfs_extent_busy_trim(
322 	struct xfs_alloc_arg	*args,
323 	xfs_agblock_t		*bno,
324 	xfs_extlen_t		*len,
325 	unsigned		*busy_gen)
326 {
327 	xfs_agblock_t		fbno;
328 	xfs_extlen_t		flen;
329 	struct rb_node		*rbp;
330 	bool			ret = false;
331 
332 	ASSERT(*len > 0);
333 
334 	spin_lock(&args->pag->pagb_lock);
335 	fbno = *bno;
336 	flen = *len;
337 	rbp = args->pag->pagb_tree.rb_node;
338 	while (rbp && flen >= args->minlen) {
339 		struct xfs_extent_busy *busyp =
340 			rb_entry(rbp, struct xfs_extent_busy, rb_node);
341 		xfs_agblock_t	fend = fbno + flen;
342 		xfs_agblock_t	bbno = busyp->bno;
343 		xfs_agblock_t	bend = bbno + busyp->length;
344 
345 		if (fend <= bbno) {
346 			rbp = rbp->rb_left;
347 			continue;
348 		} else if (fbno >= bend) {
349 			rbp = rbp->rb_right;
350 			continue;
351 		}
352 
353 		if (bbno <= fbno) {
354 			/* start overlap */
355 
356 			/*
357 			 * Case 1:
358 			 *    bbno           bend
359 			 *    +BBBBBBBBBBBBBBBBB+
360 			 *        +---------+
361 			 *        fbno   fend
362 			 *
363 			 * Case 2:
364 			 *    bbno           bend
365 			 *    +BBBBBBBBBBBBBBBBB+
366 			 *    +-------------+
367 			 *    fbno       fend
368 			 *
369 			 * Case 3:
370 			 *    bbno           bend
371 			 *    +BBBBBBBBBBBBBBBBB+
372 			 *        +-------------+
373 			 *        fbno       fend
374 			 *
375 			 * Case 4:
376 			 *    bbno           bend
377 			 *    +BBBBBBBBBBBBBBBBB+
378 			 *    +-----------------+
379 			 *    fbno           fend
380 			 *
381 			 * No unbusy region in extent, return failure.
382 			 */
383 			if (fend <= bend)
384 				goto fail;
385 
386 			/*
387 			 * Case 5:
388 			 *    bbno           bend
389 			 *    +BBBBBBBBBBBBBBBBB+
390 			 *        +----------------------+
391 			 *        fbno                fend
392 			 *
393 			 * Case 6:
394 			 *    bbno           bend
395 			 *    +BBBBBBBBBBBBBBBBB+
396 			 *    +--------------------------+
397 			 *    fbno                    fend
398 			 *
399 			 * Needs to be trimmed to:
400 			 *                       +-------+
401 			 *                       fbno fend
402 			 */
403 			fbno = bend;
404 		} else if (bend >= fend) {
405 			/* end overlap */
406 
407 			/*
408 			 * Case 7:
409 			 *             bbno           bend
410 			 *             +BBBBBBBBBBBBBBBBB+
411 			 *    +------------------+
412 			 *    fbno            fend
413 			 *
414 			 * Case 8:
415 			 *             bbno           bend
416 			 *             +BBBBBBBBBBBBBBBBB+
417 			 *    +--------------------------+
418 			 *    fbno                    fend
419 			 *
420 			 * Needs to be trimmed to:
421 			 *    +-------+
422 			 *    fbno fend
423 			 */
424 			fend = bbno;
425 		} else {
426 			/* middle overlap */
427 
428 			/*
429 			 * Case 9:
430 			 *             bbno           bend
431 			 *             +BBBBBBBBBBBBBBBBB+
432 			 *    +-----------------------------------+
433 			 *    fbno                             fend
434 			 *
435 			 * Can be trimmed to:
436 			 *    +-------+        OR         +-------+
437 			 *    fbno fend                   fbno fend
438 			 *
439 			 * Backward allocation leads to significant
440 			 * fragmentation of directories, which degrades
441 			 * directory performance, therefore we always want to
442 			 * choose the option that produces forward allocation
443 			 * patterns.
444 			 * Preferring the lower bno extent will make the next
445 			 * request use "fend" as the start of the next
446 			 * allocation;  if the segment is no longer busy at
447 			 * that point, we'll get a contiguous allocation, but
448 			 * even if it is still busy, we will get a forward
449 			 * allocation.
450 			 * We try to avoid choosing the segment at "bend",
451 			 * because that can lead to the next allocation
452 			 * taking the segment at "fbno", which would be a
453 			 * backward allocation.  We only use the segment at
454 			 * "fbno" if it is much larger than the current
455 			 * requested size, because in that case there's a
456 			 * good chance subsequent allocations will be
457 			 * contiguous.
458 			 */
459 			if (bbno - fbno >= args->maxlen) {
460 				/* left candidate fits perfect */
461 				fend = bbno;
462 			} else if (fend - bend >= args->maxlen * 4) {
463 				/* right candidate has enough free space */
464 				fbno = bend;
465 			} else if (bbno - fbno >= args->minlen) {
466 				/* left candidate fits minimum requirement */
467 				fend = bbno;
468 			} else {
469 				goto fail;
470 			}
471 		}
472 
473 		flen = fend - fbno;
474 	}
475 out:
476 
477 	if (fbno != *bno || flen != *len) {
478 		trace_xfs_extent_busy_trim(args->mp, args->agno, *bno, *len,
479 					  fbno, flen);
480 		*bno = fbno;
481 		*len = flen;
482 		*busy_gen = args->pag->pagb_gen;
483 		ret = true;
484 	}
485 	spin_unlock(&args->pag->pagb_lock);
486 	return ret;
487 fail:
488 	/*
489 	 * Return a zero extent length as failure indications.  All callers
490 	 * re-check if the trimmed extent satisfies the minlen requirement.
491 	 */
492 	flen = 0;
493 	goto out;
494 }
495 
496 STATIC void
497 xfs_extent_busy_clear_one(
498 	struct xfs_mount	*mp,
499 	struct xfs_perag	*pag,
500 	struct xfs_extent_busy	*busyp)
501 {
502 	if (busyp->length) {
503 		trace_xfs_extent_busy_clear(mp, busyp->agno, busyp->bno,
504 						busyp->length);
505 		rb_erase(&busyp->rb_node, &pag->pagb_tree);
506 	}
507 
508 	list_del_init(&busyp->list);
509 	kmem_free(busyp);
510 }
511 
512 static void
513 xfs_extent_busy_put_pag(
514 	struct xfs_perag	*pag,
515 	bool			wakeup)
516 		__releases(pag->pagb_lock)
517 {
518 	if (wakeup) {
519 		pag->pagb_gen++;
520 		wake_up_all(&pag->pagb_wait);
521 	}
522 
523 	spin_unlock(&pag->pagb_lock);
524 	xfs_perag_put(pag);
525 }
526 
527 /*
528  * Remove all extents on the passed in list from the busy extents tree.
529  * If do_discard is set skip extents that need to be discarded, and mark
530  * these as undergoing a discard operation instead.
531  */
532 void
533 xfs_extent_busy_clear(
534 	struct xfs_mount	*mp,
535 	struct list_head	*list,
536 	bool			do_discard)
537 {
538 	struct xfs_extent_busy	*busyp, *n;
539 	struct xfs_perag	*pag = NULL;
540 	xfs_agnumber_t		agno = NULLAGNUMBER;
541 	bool			wakeup = false;
542 
543 	list_for_each_entry_safe(busyp, n, list, list) {
544 		if (busyp->agno != agno) {
545 			if (pag)
546 				xfs_extent_busy_put_pag(pag, wakeup);
547 			agno = busyp->agno;
548 			pag = xfs_perag_get(mp, agno);
549 			spin_lock(&pag->pagb_lock);
550 			wakeup = false;
551 		}
552 
553 		if (do_discard && busyp->length &&
554 		    !(busyp->flags & XFS_EXTENT_BUSY_SKIP_DISCARD)) {
555 			busyp->flags = XFS_EXTENT_BUSY_DISCARDED;
556 		} else {
557 			xfs_extent_busy_clear_one(mp, pag, busyp);
558 			wakeup = true;
559 		}
560 	}
561 
562 	if (pag)
563 		xfs_extent_busy_put_pag(pag, wakeup);
564 }
565 
566 /*
567  * Flush out all busy extents for this AG.
568  */
569 void
570 xfs_extent_busy_flush(
571 	struct xfs_mount	*mp,
572 	struct xfs_perag	*pag,
573 	unsigned		busy_gen)
574 {
575 	DEFINE_WAIT		(wait);
576 	int			error;
577 
578 	error = xfs_log_force(mp, XFS_LOG_SYNC);
579 	if (error)
580 		return;
581 
582 	do {
583 		prepare_to_wait(&pag->pagb_wait, &wait, TASK_KILLABLE);
584 		if  (busy_gen != READ_ONCE(pag->pagb_gen))
585 			break;
586 		schedule();
587 	} while (1);
588 
589 	finish_wait(&pag->pagb_wait, &wait);
590 }
591 
592 void
593 xfs_extent_busy_wait_all(
594 	struct xfs_mount	*mp)
595 {
596 	struct xfs_perag	*pag;
597 	DEFINE_WAIT		(wait);
598 	xfs_agnumber_t		agno;
599 
600 	for_each_perag(mp, agno, pag) {
601 		do {
602 			prepare_to_wait(&pag->pagb_wait, &wait, TASK_KILLABLE);
603 			if  (RB_EMPTY_ROOT(&pag->pagb_tree))
604 				break;
605 			schedule();
606 		} while (1);
607 		finish_wait(&pag->pagb_wait, &wait);
608 	}
609 }
610 
611 /*
612  * Callback for list_sort to sort busy extents by the AG they reside in.
613  */
614 int
615 xfs_extent_busy_ag_cmp(
616 	void			*priv,
617 	const struct list_head	*l1,
618 	const struct list_head	*l2)
619 {
620 	struct xfs_extent_busy	*b1 =
621 		container_of(l1, struct xfs_extent_busy, list);
622 	struct xfs_extent_busy	*b2 =
623 		container_of(l2, struct xfs_extent_busy, list);
624 	s32 diff;
625 
626 	diff = b1->agno - b2->agno;
627 	if (!diff)
628 		diff = b1->bno - b2->bno;
629 	return diff;
630 }
631