summaryrefslogtreecommitdiffstats
path: root/sys/kern/subr_disk.c
blob: 23915402e30d89c1a509d68bf150cd3d3e4d552f (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
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
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
/*-
 * ----------------------------------------------------------------------------
 * "THE BEER-WARE LICENSE" (Revision 42):
 * <phk@FreeBSD.ORG> wrote this file.  As long as you retain this notice you
 * can do whatever you want with this stuff. If we meet some day, and you think
 * this stuff is worth it, you can buy me a beer in return.   Poul-Henning Kamp
 * ----------------------------------------------------------------------------
 *
 * The bioq_disksort() (and the specification of the bioq API)
 * have been written by Luigi Rizzo and Fabio Checconi under the same
 * license as above.
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");

#include "opt_geom.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/bio.h>
#include <sys/conf.h>
#include <sys/disk.h>
#include <geom/geom_disk.h>

/*-
 * Disk error is the preface to plaintive error messages
 * about failing disk transfers.  It prints messages of the form
 * 	"hp0g: BLABLABLA cmd=read fsbn 12345 of 12344-12347"
 * blkdone should be -1 if the position of the error is unknown.
 * The message is printed with printf.
 */
void
disk_err(struct bio *bp, const char *what, int blkdone, int nl)
{
	daddr_t sn;

	if (bp->bio_dev != NULL)
		printf("%s: %s ", devtoname(bp->bio_dev), what);
	else if (bp->bio_disk != NULL)
		printf("%s%d: %s ",
		    bp->bio_disk->d_name, bp->bio_disk->d_unit, what);
	else
		printf("disk??: %s ", what);
	switch(bp->bio_cmd) {
	case BIO_READ:		printf("cmd=read "); break;
	case BIO_WRITE:		printf("cmd=write "); break;
	case BIO_DELETE:	printf("cmd=delete "); break;
	case BIO_GETATTR:	printf("cmd=getattr "); break;
	case BIO_FLUSH:		printf("cmd=flush "); break;
	default:		printf("cmd=%x ", bp->bio_cmd); break;
	}
	sn = bp->bio_pblkno;
	if (bp->bio_bcount <= DEV_BSIZE) {
		printf("fsbn %jd%s", (intmax_t)sn, nl ? "\n" : "");
		return;
	}
	if (blkdone >= 0) {
		sn += blkdone;
		printf("fsbn %jd of ", (intmax_t)sn);
	}
	printf("%jd-%jd", (intmax_t)bp->bio_pblkno,
	    (intmax_t)(bp->bio_pblkno + (bp->bio_bcount - 1) / DEV_BSIZE));
	if (nl)
		printf("\n");
}

/*
 * BIO queue implementation
 *
 * Please read carefully the description below before making any change
 * to the code, or you might change the behaviour of the data structure
 * in undesirable ways.
 *
 * A bioq stores disk I/O request (bio), normally sorted according to
 * the distance of the requested position (bio->bio_offset) from the
 * current head position (bioq->last_offset) in the scan direction, i.e.
 *
 * 	(uoff_t)(bio_offset - last_offset)
 *
 * Note that the cast to unsigned (uoff_t) is fundamental to insure
 * that the distance is computed in the scan direction.
 *
 * The main methods for manipulating the bioq are:
 *
 *   bioq_disksort()	performs an ordered insertion;
 *
 *   bioq_first()	return the head of the queue, without removing;
 *
 *   bioq_takefirst()	return and remove the head of the queue,
 *		updating the 'current head position' as
 *		bioq->last_offset = bio->bio_offset + bio->bio_length;
 *
 * When updating the 'current head position', we assume that the result of
 * bioq_takefirst() is dispatched to the device, so bioq->last_offset
 * represents the head position once the request is complete.
 *
 * If the bioq is manipulated using only the above calls, it starts
 * with a sorted sequence of requests with bio_offset >= last_offset,
 * possibly followed by another sorted sequence of requests with
 * 0 <= bio_offset < bioq->last_offset 
 *
 * NOTE: historical behaviour was to ignore bio->bio_length in the
 *	update, but its use tracks the head position in a better way.
 *	Historical behaviour was also to update the head position when
 *	the request under service is complete, rather than when the
 *	request is extracted from the queue. However, the current API
 *	has no method to update the head position; secondly, once
 *	a request has been submitted to the disk, we have no idea of
 *	the actual head position, so the final one is our best guess.
 *
 * --- Direct queue manipulation ---
 *
 * A bioq uses an underlying TAILQ to store requests, so we also
 * export methods to manipulate the TAILQ, in particular:
 *
 * bioq_insert_tail()	insert an entry at the end.
 *		It also creates a 'barrier' so all subsequent
 *		insertions through bioq_disksort() will end up
 *		after this entry;
 *
 * bioq_insert_head()	insert an entry at the head, update
 *		bioq->last_offset = bio->bio_offset so that
 *		all subsequent insertions through bioq_disksort()
 *		will end up after this entry;
 *
 * bioq_remove()	remove a generic element from the queue, act as
 *		bioq_takefirst() if invoked on the head of the queue.
 *
 * The semantic of these methods is the same as the operations
 * on the underlying TAILQ, but with additional guarantees on
 * subsequent bioq_disksort() calls. E.g. bioq_insert_tail()
 * can be useful for making sure that all previous ops are flushed
 * to disk before continuing.
 *
 * Updating bioq->last_offset on a bioq_insert_head() guarantees
 * that the bio inserted with the last bioq_insert_head() will stay
 * at the head of the queue even after subsequent bioq_disksort().
 *
 * Note that when the direct queue manipulation functions are used,
 * the queue may contain multiple inversion points (i.e. more than
 * two sorted sequences of requests).
 *
 */

void
bioq_init(struct bio_queue_head *head)
{

	TAILQ_INIT(&head->queue);
	head->last_offset = 0;
	head->insert_point = NULL;
}

void
bioq_remove(struct bio_queue_head *head, struct bio *bp)
{

	if (head->insert_point == NULL) {
		if (bp == TAILQ_FIRST(&head->queue))
			head->last_offset = bp->bio_offset + bp->bio_length;
	} else if (bp == head->insert_point)
		head->insert_point = NULL;

	TAILQ_REMOVE(&head->queue, bp, bio_queue);
}

void
bioq_flush(struct bio_queue_head *head, struct devstat *stp, int error)
{
	struct bio *bp;

	while ((bp = bioq_takefirst(head)) != NULL)
		biofinish(bp, stp, error);
}

void
bioq_insert_head(struct bio_queue_head *head, struct bio *bp)
{

	if (head->insert_point == NULL)
		head->last_offset = bp->bio_offset;
	TAILQ_INSERT_HEAD(&head->queue, bp, bio_queue);
}

void
bioq_insert_tail(struct bio_queue_head *head, struct bio *bp)
{

	TAILQ_INSERT_TAIL(&head->queue, bp, bio_queue);
	head->insert_point = bp;
	head->last_offset = bp->bio_offset;
}

struct bio *
bioq_first(struct bio_queue_head *head)
{

	return (TAILQ_FIRST(&head->queue));
}

struct bio *
bioq_takefirst(struct bio_queue_head *head)
{
	struct bio *bp;

	bp = TAILQ_FIRST(&head->queue);
	if (bp != NULL)
		bioq_remove(head, bp);
	return (bp);
}

/*
 * Compute the sorting key. The cast to unsigned is
 * fundamental for correctness, see the description
 * near the beginning of the file.
 */
static inline uoff_t
bioq_bio_key(struct bio_queue_head *head, struct bio *bp)
{

	return ((uoff_t)(bp->bio_offset - head->last_offset));
}

/*
 * Seek sort for disks.
 *
 * Sort all requests in a single queue while keeping
 * track of the current position of the disk with last_offset.
 * See above for details.
 */
void
bioq_disksort(struct bio_queue_head *head, struct bio *bp)
{
	struct bio *cur, *prev;
	uoff_t key;

	if ((bp->bio_flags & BIO_ORDERED) != 0) {
		/*
		 * Ordered transactions can only be dispatched
		 * after any currently queued transactions.  They
		 * also have barrier semantics - no transactions
		 * queued in the future can pass them.
		 */
		bioq_insert_tail(head, bp);
		return;
	}

	prev = NULL;
	key = bioq_bio_key(head, bp);
	cur = TAILQ_FIRST(&head->queue);

	if (head->insert_point) {
		prev = head->insert_point;
		cur = TAILQ_NEXT(head->insert_point, bio_queue);
	}

	while (cur != NULL && key >= bioq_bio_key(head, cur)) {
		prev = cur;
		cur = TAILQ_NEXT(cur, bio_queue);
	}

	if (prev == NULL)
		TAILQ_INSERT_HEAD(&head->queue, bp, bio_queue);
	else
		TAILQ_INSERT_AFTER(&head->queue, prev, bp, bio_queue);
}
OpenPOWER on IntegriCloud