Blame view

block/blk-merge.c 9.89 KB
d6d481969   Jens Axboe   block: ll_rw_blk....
1
2
3
4
5
6
7
8
9
10
11
12
13
  /*
   * Functions related to segment and merge handling
   */
  #include <linux/kernel.h>
  #include <linux/module.h>
  #include <linux/bio.h>
  #include <linux/blkdev.h>
  #include <linux/scatterlist.h>
  
  #include "blk.h"
  
  void blk_recalc_rq_sectors(struct request *rq, int nsect)
  {
e17fc0a1c   David Woodhouse   Allow elevators t...
14
  	if (blk_fs_request(rq) || blk_discard_rq(rq)) {
d6d481969   Jens Axboe   block: ll_rw_blk....
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
  		rq->hard_sector += nsect;
  		rq->hard_nr_sectors -= nsect;
  
  		/*
  		 * Move the I/O submission pointers ahead if required.
  		 */
  		if ((rq->nr_sectors >= rq->hard_nr_sectors) &&
  		    (rq->sector <= rq->hard_sector)) {
  			rq->sector = rq->hard_sector;
  			rq->nr_sectors = rq->hard_nr_sectors;
  			rq->hard_cur_sectors = bio_cur_sectors(rq->bio);
  			rq->current_nr_sectors = rq->hard_cur_sectors;
  			rq->buffer = bio_data(rq->bio);
  		}
  
  		/*
  		 * if total number of sectors is less than the first segment
  		 * size, something has gone terribly wrong
  		 */
  		if (rq->nr_sectors < rq->current_nr_sectors) {
6728cb0e6   Jens Axboe   block: make core ...
35
36
  			printk(KERN_ERR "blk: request botched
  ");
d6d481969   Jens Axboe   block: ll_rw_blk....
37
38
39
40
41
42
43
44
  			rq->nr_sectors = rq->current_nr_sectors;
  		}
  	}
  }
  
  void blk_recalc_rq_segments(struct request *rq)
  {
  	int nr_phys_segs;
d6d481969   Jens Axboe   block: ll_rw_blk....
45
  	unsigned int phys_size;
d6d481969   Jens Axboe   block: ll_rw_blk....
46
47
  	struct bio_vec *bv, *bvprv = NULL;
  	int seg_size;
d6d481969   Jens Axboe   block: ll_rw_blk....
48
49
50
51
52
53
54
  	int cluster;
  	struct req_iterator iter;
  	int high, highprv = 1;
  	struct request_queue *q = rq->q;
  
  	if (!rq->bio)
  		return;
75ad23bc0   Nick Piggin   block: make queue...
55
  	cluster = test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags);
5df97b91b   Mikulas Patocka   drop vmerge accou...
56
57
  	seg_size = 0;
  	phys_size = nr_phys_segs = 0;
d6d481969   Jens Axboe   block: ll_rw_blk....
58
59
60
61
62
63
64
65
  	rq_for_each_segment(bv, rq, iter) {
  		/*
  		 * the trick here is making sure that a high page is never
  		 * considered part of another segment, since that might
  		 * change with the bounce page.
  		 */
  		high = page_to_pfn(bv->bv_page) > q->bounce_pfn;
  		if (high || highprv)
b8b3e16cf   Mikulas Patocka   block: drop virtu...
66
  			goto new_segment;
d6d481969   Jens Axboe   block: ll_rw_blk....
67
68
69
70
71
72
73
  		if (cluster) {
  			if (seg_size + bv->bv_len > q->max_segment_size)
  				goto new_segment;
  			if (!BIOVEC_PHYS_MERGEABLE(bvprv, bv))
  				goto new_segment;
  			if (!BIOVEC_SEG_BOUNDARY(q, bvprv, bv))
  				goto new_segment;
d6d481969   Jens Axboe   block: ll_rw_blk....
74
75
  
  			seg_size += bv->bv_len;
d6d481969   Jens Axboe   block: ll_rw_blk....
76
77
78
79
  			bvprv = bv;
  			continue;
  		}
  new_segment:
867714271   FUJITA Tomonori   block: fix nr_phy...
80
81
  		if (nr_phys_segs == 1 && seg_size > rq->bio->bi_seg_front_size)
  			rq->bio->bi_seg_front_size = seg_size;
d6d481969   Jens Axboe   block: ll_rw_blk....
82
83
84
85
86
  		nr_phys_segs++;
  		bvprv = bv;
  		seg_size = bv->bv_len;
  		highprv = high;
  	}
867714271   FUJITA Tomonori   block: fix nr_phy...
87
88
89
90
  	if (nr_phys_segs == 1 && seg_size > rq->bio->bi_seg_front_size)
  		rq->bio->bi_seg_front_size = seg_size;
  	if (seg_size > rq->biotail->bi_seg_back_size)
  		rq->biotail->bi_seg_back_size = seg_size;
d6d481969   Jens Axboe   block: ll_rw_blk....
91
  	rq->nr_phys_segments = nr_phys_segs;
d6d481969   Jens Axboe   block: ll_rw_blk....
92
93
94
95
96
97
98
99
100
101
102
103
  }
  
  void blk_recount_segments(struct request_queue *q, struct bio *bio)
  {
  	struct request rq;
  	struct bio *nxt = bio->bi_next;
  	rq.q = q;
  	rq.bio = rq.biotail = bio;
  	bio->bi_next = NULL;
  	blk_recalc_rq_segments(&rq);
  	bio->bi_next = nxt;
  	bio->bi_phys_segments = rq.nr_phys_segments;
d6d481969   Jens Axboe   block: ll_rw_blk....
104
105
106
107
108
109
110
  	bio->bi_flags |= (1 << BIO_SEG_VALID);
  }
  EXPORT_SYMBOL(blk_recount_segments);
  
  static int blk_phys_contig_segment(struct request_queue *q, struct bio *bio,
  				   struct bio *nxt)
  {
75ad23bc0   Nick Piggin   block: make queue...
111
  	if (!test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags))
d6d481969   Jens Axboe   block: ll_rw_blk....
112
  		return 0;
867714271   FUJITA Tomonori   block: fix nr_phy...
113
114
  	if (bio->bi_seg_back_size + nxt->bi_seg_front_size >
  	    q->max_segment_size)
d6d481969   Jens Axboe   block: ll_rw_blk....
115
  		return 0;
e17fc0a1c   David Woodhouse   Allow elevators t...
116
117
118
119
120
  	if (!bio_has_data(bio))
  		return 1;
  
  	if (!BIOVEC_PHYS_MERGEABLE(__BVEC_END(bio), __BVEC_START(nxt)))
  		return 0;
d6d481969   Jens Axboe   block: ll_rw_blk....
121
  	/*
e17fc0a1c   David Woodhouse   Allow elevators t...
122
  	 * bio and nxt are contiguous in memory; check if the queue allows
d6d481969   Jens Axboe   block: ll_rw_blk....
123
124
125
126
127
128
129
  	 * these two to be merged into one
  	 */
  	if (BIO_SEG_BOUNDARY(q, bio, nxt))
  		return 1;
  
  	return 0;
  }
d6d481969   Jens Axboe   block: ll_rw_blk....
130
131
132
133
134
135
136
137
138
139
140
141
142
  /*
   * map a request to scatterlist, return number of sg entries setup. Caller
   * must make sure sg can hold rq->nr_phys_segments entries
   */
  int blk_rq_map_sg(struct request_queue *q, struct request *rq,
  		  struct scatterlist *sglist)
  {
  	struct bio_vec *bvec, *bvprv;
  	struct req_iterator iter;
  	struct scatterlist *sg;
  	int nsegs, cluster;
  
  	nsegs = 0;
75ad23bc0   Nick Piggin   block: make queue...
143
  	cluster = test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags);
d6d481969   Jens Axboe   block: ll_rw_blk....
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
  
  	/*
  	 * for each bio in rq
  	 */
  	bvprv = NULL;
  	sg = NULL;
  	rq_for_each_segment(bvec, rq, iter) {
  		int nbytes = bvec->bv_len;
  
  		if (bvprv && cluster) {
  			if (sg->length + nbytes > q->max_segment_size)
  				goto new_segment;
  
  			if (!BIOVEC_PHYS_MERGEABLE(bvprv, bvec))
  				goto new_segment;
  			if (!BIOVEC_SEG_BOUNDARY(q, bvprv, bvec))
  				goto new_segment;
  
  			sg->length += nbytes;
  		} else {
  new_segment:
  			if (!sg)
  				sg = sglist;
  			else {
  				/*
  				 * If the driver previously mapped a shorter
  				 * list, we could see a termination bit
  				 * prematurely unless it fully inits the sg
  				 * table on each mapping. We KNOW that there
  				 * must be more entries here or the driver
  				 * would be buggy, so force clear the
  				 * termination bit to avoid doing a full
  				 * sg_init_table() in drivers for each command.
  				 */
  				sg->page_link &= ~0x02;
  				sg = sg_next(sg);
  			}
  
  			sg_set_page(sg, bvec->bv_page, nbytes, bvec->bv_offset);
  			nsegs++;
  		}
  		bvprv = bvec;
  	} /* segments in rq */
f18573abc   FUJITA Tomonori   block: move the p...
187
188
189
190
191
192
193
194
  
  	if (unlikely(rq->cmd_flags & REQ_COPY_USER) &&
  	    (rq->data_len & q->dma_pad_mask)) {
  		unsigned int pad_len = (q->dma_pad_mask & ~rq->data_len) + 1;
  
  		sg->length += pad_len;
  		rq->extra_len += pad_len;
  	}
2fb98e841   Tejun Heo   block: implement ...
195
  	if (q->dma_drain_size && q->dma_drain_needed(rq)) {
db0a2e009   Tejun Heo   block: clear drai...
196
197
  		if (rq->cmd_flags & REQ_RW)
  			memset(q->dma_drain_buffer, 0, q->dma_drain_size);
d6d481969   Jens Axboe   block: ll_rw_blk....
198
199
200
201
202
203
204
  		sg->page_link &= ~0x02;
  		sg = sg_next(sg);
  		sg_set_page(sg, virt_to_page(q->dma_drain_buffer),
  			    q->dma_drain_size,
  			    ((unsigned long)q->dma_drain_buffer) &
  			    (PAGE_SIZE - 1));
  		nsegs++;
7a85f8896   FUJITA Tomonori   block: restore th...
205
  		rq->extra_len += q->dma_drain_size;
d6d481969   Jens Axboe   block: ll_rw_blk....
206
207
208
209
210
211
212
  	}
  
  	if (sg)
  		sg_mark_end(sg);
  
  	return nsegs;
  }
d6d481969   Jens Axboe   block: ll_rw_blk....
213
  EXPORT_SYMBOL(blk_rq_map_sg);
d6d481969   Jens Axboe   block: ll_rw_blk....
214
215
216
217
  static inline int ll_new_hw_segment(struct request_queue *q,
  				    struct request *req,
  				    struct bio *bio)
  {
d6d481969   Jens Axboe   block: ll_rw_blk....
218
  	int nr_phys_segs = bio_phys_segments(q, bio);
5df97b91b   Mikulas Patocka   drop vmerge accou...
219
  	if (req->nr_phys_segments + nr_phys_segs > q->max_hw_segments
d6d481969   Jens Axboe   block: ll_rw_blk....
220
221
222
223
224
225
226
227
228
229
230
  	    || req->nr_phys_segments + nr_phys_segs > q->max_phys_segments) {
  		req->cmd_flags |= REQ_NOMERGE;
  		if (req == q->last_merge)
  			q->last_merge = NULL;
  		return 0;
  	}
  
  	/*
  	 * This will form the start of a new hw segment.  Bump both
  	 * counters.
  	 */
d6d481969   Jens Axboe   block: ll_rw_blk....
231
232
233
234
235
236
237
238
  	req->nr_phys_segments += nr_phys_segs;
  	return 1;
  }
  
  int ll_back_merge_fn(struct request_queue *q, struct request *req,
  		     struct bio *bio)
  {
  	unsigned short max_sectors;
d6d481969   Jens Axboe   block: ll_rw_blk....
239
240
241
242
243
244
245
246
247
248
249
250
  
  	if (unlikely(blk_pc_request(req)))
  		max_sectors = q->max_hw_sectors;
  	else
  		max_sectors = q->max_sectors;
  
  	if (req->nr_sectors + bio_sectors(bio) > max_sectors) {
  		req->cmd_flags |= REQ_NOMERGE;
  		if (req == q->last_merge)
  			q->last_merge = NULL;
  		return 0;
  	}
2cdf79caf   Jens Axboe   block: get rid of...
251
  	if (!bio_flagged(req->biotail, BIO_SEG_VALID))
d6d481969   Jens Axboe   block: ll_rw_blk....
252
  		blk_recount_segments(q, req->biotail);
2cdf79caf   Jens Axboe   block: get rid of...
253
  	if (!bio_flagged(bio, BIO_SEG_VALID))
d6d481969   Jens Axboe   block: ll_rw_blk....
254
  		blk_recount_segments(q, bio);
d6d481969   Jens Axboe   block: ll_rw_blk....
255
256
257
  
  	return ll_new_hw_segment(q, req, bio);
  }
6728cb0e6   Jens Axboe   block: make core ...
258
  int ll_front_merge_fn(struct request_queue *q, struct request *req,
d6d481969   Jens Axboe   block: ll_rw_blk....
259
260
261
  		      struct bio *bio)
  {
  	unsigned short max_sectors;
d6d481969   Jens Axboe   block: ll_rw_blk....
262
263
264
265
266
267
268
269
270
271
272
273
274
  
  	if (unlikely(blk_pc_request(req)))
  		max_sectors = q->max_hw_sectors;
  	else
  		max_sectors = q->max_sectors;
  
  
  	if (req->nr_sectors + bio_sectors(bio) > max_sectors) {
  		req->cmd_flags |= REQ_NOMERGE;
  		if (req == q->last_merge)
  			q->last_merge = NULL;
  		return 0;
  	}
2cdf79caf   Jens Axboe   block: get rid of...
275
  	if (!bio_flagged(bio, BIO_SEG_VALID))
d6d481969   Jens Axboe   block: ll_rw_blk....
276
  		blk_recount_segments(q, bio);
2cdf79caf   Jens Axboe   block: get rid of...
277
  	if (!bio_flagged(req->bio, BIO_SEG_VALID))
d6d481969   Jens Axboe   block: ll_rw_blk....
278
  		blk_recount_segments(q, req->bio);
d6d481969   Jens Axboe   block: ll_rw_blk....
279
280
281
282
283
284
285
286
  
  	return ll_new_hw_segment(q, req, bio);
  }
  
  static int ll_merge_requests_fn(struct request_queue *q, struct request *req,
  				struct request *next)
  {
  	int total_phys_segments;
867714271   FUJITA Tomonori   block: fix nr_phy...
287
288
  	unsigned int seg_size =
  		req->biotail->bi_seg_back_size + next->bio->bi_seg_front_size;
d6d481969   Jens Axboe   block: ll_rw_blk....
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
  
  	/*
  	 * First check if the either of the requests are re-queued
  	 * requests.  Can't merge them if they are.
  	 */
  	if (req->special || next->special)
  		return 0;
  
  	/*
  	 * Will it become too large?
  	 */
  	if ((req->nr_sectors + next->nr_sectors) > q->max_sectors)
  		return 0;
  
  	total_phys_segments = req->nr_phys_segments + next->nr_phys_segments;
867714271   FUJITA Tomonori   block: fix nr_phy...
304
305
306
307
308
  	if (blk_phys_contig_segment(q, req->biotail, next->bio)) {
  		if (req->nr_phys_segments == 1)
  			req->bio->bi_seg_front_size = seg_size;
  		if (next->nr_phys_segments == 1)
  			next->biotail->bi_seg_back_size = seg_size;
d6d481969   Jens Axboe   block: ll_rw_blk....
309
  		total_phys_segments--;
867714271   FUJITA Tomonori   block: fix nr_phy...
310
  	}
d6d481969   Jens Axboe   block: ll_rw_blk....
311
312
313
  
  	if (total_phys_segments > q->max_phys_segments)
  		return 0;
5df97b91b   Mikulas Patocka   drop vmerge accou...
314
  	if (total_phys_segments > q->max_hw_segments)
d6d481969   Jens Axboe   block: ll_rw_blk....
315
316
317
318
  		return 0;
  
  	/* Merge is OK... */
  	req->nr_phys_segments = total_phys_segments;
d6d481969   Jens Axboe   block: ll_rw_blk....
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
  	return 1;
  }
  
  /*
   * Has to be called with the request spinlock acquired
   */
  static int attempt_merge(struct request_queue *q, struct request *req,
  			  struct request *next)
  {
  	if (!rq_mergeable(req) || !rq_mergeable(next))
  		return 0;
  
  	/*
  	 * not contiguous
  	 */
  	if (req->sector + req->nr_sectors != next->sector)
  		return 0;
  
  	if (rq_data_dir(req) != rq_data_dir(next)
  	    || req->rq_disk != next->rq_disk
  	    || next->special)
  		return 0;
7ba1ba12e   Martin K. Petersen   block: Block laye...
341
342
  	if (blk_integrity_rq(req) != blk_integrity_rq(next))
  		return 0;
d6d481969   Jens Axboe   block: ll_rw_blk....
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
  	/*
  	 * If we are allowed to merge, then append bio list
  	 * from next to rq and release next. merge_requests_fn
  	 * will have updated segment counts, update sector
  	 * counts here.
  	 */
  	if (!ll_merge_requests_fn(q, req, next))
  		return 0;
  
  	/*
  	 * At this point we have either done a back merge
  	 * or front merge. We need the smaller start_time of
  	 * the merged requests to be the current request
  	 * for accounting purposes.
  	 */
  	if (time_after(req->start_time, next->start_time))
  		req->start_time = next->start_time;
  
  	req->biotail->bi_next = next->bio;
  	req->biotail = next->biotail;
  
  	req->nr_sectors = req->hard_nr_sectors += next->hard_nr_sectors;
  
  	elv_merge_requests(q, req, next);
  
  	if (req->rq_disk) {
e71bf0d0e   Tejun Heo   block: fix disk->...
369
  		struct hd_struct *part;
c99590591   Tejun Heo   block: fix diskst...
370
  		int cpu;
e71bf0d0e   Tejun Heo   block: fix disk->...
371

074a7aca7   Tejun Heo   block: move stats...
372
  		cpu = part_stat_lock();
e71bf0d0e   Tejun Heo   block: fix disk->...
373
  		part = disk_map_sector_rcu(req->rq_disk, req->sector);
c99590591   Tejun Heo   block: fix diskst...
374

074a7aca7   Tejun Heo   block: move stats...
375
376
  		part_round_stats(cpu, part);
  		part_dec_in_flight(part);
e71bf0d0e   Tejun Heo   block: fix disk->...
377

074a7aca7   Tejun Heo   block: move stats...
378
  		part_stat_unlock();
d6d481969   Jens Axboe   block: ll_rw_blk....
379
380
381
  	}
  
  	req->ioprio = ioprio_best(req->ioprio, next->ioprio);
ab780f1ec   Jens Axboe   block: inherit CP...
382
383
  	if (blk_rq_cpu_valid(next))
  		req->cpu = next->cpu;
d6d481969   Jens Axboe   block: ll_rw_blk....
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
  
  	__blk_put_request(q, next);
  	return 1;
  }
  
  int attempt_back_merge(struct request_queue *q, struct request *rq)
  {
  	struct request *next = elv_latter_request(q, rq);
  
  	if (next)
  		return attempt_merge(q, rq, next);
  
  	return 0;
  }
  
  int attempt_front_merge(struct request_queue *q, struct request *rq)
  {
  	struct request *prev = elv_former_request(q, rq);
  
  	if (prev)
  		return attempt_merge(q, prev, rq);
  
  	return 0;
  }