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
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
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
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
|
/* vinuminterrupt.c: bottom half of the driver */
/*-
* Copyright (c) 1997, 1998, 1999
* Nan Yang Computer Services Limited. All rights reserved.
*
* Parts copyright (c) 1997, 1998 Cybernet Corporation, NetMAX project.
*
* Written by Greg Lehey
*
* This software is distributed under the so-called ``Berkeley
* License'':
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Nan Yang Computer
* Services Limited.
* 4. Neither the name of the Company nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* This software is provided ``as is'', and any express or implied
* warranties, including, but not limited to, the implied warranties of
* merchantability and fitness for a particular purpose are disclaimed.
* In no event shall the company or contributors be liable for any
* direct, indirect, incidental, special, exemplary, or consequential
* damages (including, but not limited to, procurement of substitute
* goods or services; loss of use, data, or profits; or business
* interruption) however caused and on any theory of liability, whether
* in contract, strict liability, or tort (including negligence or
* otherwise) arising in any way out of the use of this software, even if
* advised of the possibility of such damage.
*
* $Id: vinuminterrupt.c,v 1.9 2000/02/16 01:59:02 grog Exp grog $
* $FreeBSD$
*/
#include <dev/vinum/vinumhdr.h>
#include <dev/vinum/request.h>
#include <sys/resourcevar.h>
void complete_raid5_write(struct rqelement *);
void complete_rqe(struct buf *bp);
void sdio_done(struct buf *bp);
/*
* Take a completed buffer, transfer the data back if
* it's a read, and complete the high-level request
* if this is the last subrequest.
*
* The bp parameter is in fact a struct rqelement, which
* includes a couple of extras at the end.
*/
void
complete_rqe(struct buf *bp)
{
struct rqelement *rqe;
struct request *rq;
struct rqgroup *rqg;
struct buf *ubp; /* user buffer */
struct drive *drive;
rqe = (struct rqelement *) bp; /* point to the element element that completed */
rqg = rqe->rqg; /* and the request group */
rq = rqg->rq; /* and the complete request */
ubp = rq->bp; /* user buffer */
#ifdef VINUMDEBUG
if (debug & DEBUG_LASTREQS)
logrq(loginfo_iodone, (union rqinfou) rqe, ubp);
#endif
drive = &DRIVE[rqe->driveno];
drive->active--; /* one less outstanding I/O on this drive */
vinum_conf.active--; /* one less outstanding I/O globally */
if ((drive->active == (DRIVE_MAXACTIVE - 1)) /* we were at the drive limit */
||(vinum_conf.active == VINUM_MAXACTIVE)) /* or the global limit */
wakeup(&launch_requests); /* let another one at it */
if ((bp->b_flags & B_ERROR) != 0) { /* transfer in error */
if (bp->b_error != 0) /* did it return a number? */
rq->error = bp->b_error; /* yes, put it in. */
else if (rq->error == 0) /* no: do we have one already? */
rq->error = EIO; /* no: catchall "I/O error" */
SD[rqe->sdno].lasterror = rq->error;
if (bp->b_iocmd == BIO_READ) {
log(LOG_ERR, "%s: fatal read I/O error\n", SD[rqe->sdno].name);
set_sd_state(rqe->sdno, sd_crashed, setstate_force); /* subdisk is crashed */
} else { /* write operation */
log(LOG_ERR, "%s: fatal write I/O error\n", SD[rqe->sdno].name);
set_sd_state(rqe->sdno, sd_stale, setstate_force); /* subdisk is stale */
}
if (rq->error == ENXIO) { /* the drive's down too */
log(LOG_ERR, "%s: fatal drive I/O error\n", DRIVE[rqe->driveno].label.name);
DRIVE[rqe->driveno].lasterror = rq->error;
set_drive_state(rqe->driveno, /* take the drive down */
drive_down,
setstate_force);
}
}
/* Now update the statistics */
if (bp->b_iocmd == BIO_READ) { /* read operation */
DRIVE[rqe->driveno].reads++;
DRIVE[rqe->driveno].bytes_read += bp->b_bcount;
SD[rqe->sdno].reads++;
SD[rqe->sdno].bytes_read += bp->b_bcount;
PLEX[rqe->rqg->plexno].reads++;
PLEX[rqe->rqg->plexno].bytes_read += bp->b_bcount;
if (PLEX[rqe->rqg->plexno].volno >= 0)
VOL[PLEX[rqe->rqg->plexno].volno].bytes_read += bp->b_bcount;
} else { /* write operation */
DRIVE[rqe->driveno].writes++;
DRIVE[rqe->driveno].bytes_written += bp->b_bcount;
SD[rqe->sdno].writes++;
SD[rqe->sdno].bytes_written += bp->b_bcount;
PLEX[rqe->rqg->plexno].writes++;
PLEX[rqe->rqg->plexno].bytes_written += bp->b_bcount;
if (PLEX[rqe->rqg->plexno].volno >= 0)
VOL[PLEX[rqe->rqg->plexno].volno].bytes_written += bp->b_bcount;
}
rqg->active--; /* one less request active */
if (rqg->flags & XFR_RECOVERY_READ) { /* recovery read, */
int *sdata; /* source */
int *data; /* and group data */
int length; /* and count involved */
int count; /* loop counter */
struct rqelement *urqe = &rqg->rqe[rqg->badsdno]; /* rqe of the bad subdisk */
/* XOR destination is the user data */
sdata = (int *) &rqe->b.b_data[rqe->groupoffset << DEV_BSHIFT]; /* old data contents */
data = (int *) &urqe->b.b_data[urqe->groupoffset << DEV_BSHIFT]; /* destination */
length = urqe->grouplen << (DEV_BSHIFT - 2); /* and count involved */
for (count = 0; count < length; count++)
data[count] ^= sdata[count];
/*
* In a normal read, we will normally read directly
* into the user buffer. This doesn't work if
* we're also doing a recovery, so we have to
* copy it
*/
if (rqe->flags & XFR_NORMAL_READ) { /* normal read as well, */
char *src = &rqe->b.b_data[rqe->dataoffset << DEV_BSHIFT]; /* read data is here */
char *dst;
dst = (char *) ubp->b_data + (rqe->useroffset << DEV_BSHIFT); /* where to put it in user buffer */
length = rqe->datalen << DEV_BSHIFT; /* and count involved */
bcopy(src, dst, length); /* move it */
}
} else if ((rqg->flags & (XFR_NORMAL_WRITE | XFR_DEGRADED_WRITE)) /* RAID 4/5 group write operation */
&&(rqg->active == 0)) /* and we've finished phase 1 */
complete_raid5_write(rqe);
if (rqg->active == 0) { /* request group finished, */
rq->active--; /* one less */
if (rqg->lock) { /* got a lock? */
unlockrange(rqg->plexno, rqg->lock); /* yes, free it */
rqg->lock = 0;
}
}
if (rq->active == 0) { /* request finished, */
#if VINUMDEBUG
if (debug & DEBUG_RESID) {
if (ubp->b_resid != 0) /* still something to transfer? */
Debugger("resid");
}
#endif
if (rq->error) { /* did we have an error? */
if (rq->isplex) { /* plex operation, */
ubp->b_flags |= B_ERROR; /* yes, propagate to user */
ubp->b_error = rq->error;
} else /* try to recover */
queue_daemon_request(daemonrq_ioerror, (union daemoninfo) rq); /* let the daemon complete */
} else {
ubp->b_resid = 0; /* completed our transfer */
if (rq->isplex == 0) /* volume request, */
VOL[rq->volplex.volno].active--; /* another request finished */
biodone(ubp); /* top level buffer completed */
freerq(rq); /* return the request storage */
}
}
}
/* Free a request block and anything hanging off it */
void
freerq(struct request *rq)
{
struct rqgroup *rqg;
struct rqgroup *nrqg; /* next in chain */
int rqno;
for (rqg = rq->rqg; rqg != NULL; rqg = nrqg) { /* through the whole request chain */
if (rqg->lock) /* got a lock? */
unlockrange(rqg->plexno, rqg->lock); /* yes, free it */
for (rqno = 0; rqno < rqg->count; rqno++)
if ((rqg->rqe[rqno].flags & XFR_MALLOCED) /* data buffer was malloced, */
&&rqg->rqe[rqno].b.b_data) /* and the allocation succeeded */
Free(rqg->rqe[rqno].b.b_data); /* free it */
nrqg = rqg->next; /* note the next one */
Free(rqg); /* and free this one */
}
Free(rq); /* free the request itself */
}
/* I/O on subdisk completed */
void
sdio_done(struct buf *bp)
{
struct sdbuf *sbp;
sbp = (struct sdbuf *) bp;
if (sbp->b.b_flags & B_ERROR) { /* had an error */
sbp->bp->b_flags |= B_ERROR; /* propagate upwards */
sbp->bp->b_error = sbp->b.b_error;
}
#ifdef VINUMDEBUG
if (debug & DEBUG_LASTREQS)
logrq(loginfo_sdiodone, (union rqinfou) bp, bp);
#endif
sbp->bp->b_resid = sbp->b.b_resid; /* copy the resid field */
/* Now update the statistics */
if (bp->b_iocmd == BIO_READ) { /* read operation */
DRIVE[sbp->driveno].reads++;
DRIVE[sbp->driveno].bytes_read += sbp->b.b_bcount;
SD[sbp->sdno].reads++;
SD[sbp->sdno].bytes_read += sbp->b.b_bcount;
} else { /* write operation */
DRIVE[sbp->driveno].writes++;
DRIVE[sbp->driveno].bytes_written += sbp->b.b_bcount;
SD[sbp->sdno].writes++;
SD[sbp->sdno].bytes_written += sbp->b.b_bcount;
}
biodone(sbp->bp); /* complete the caller's I/O */
Free(sbp);
}
/* Start the second phase of a RAID-4 or RAID-5 group write operation. */
void
complete_raid5_write(struct rqelement *rqe)
{
int *sdata; /* source */
int *pdata; /* and parity block data */
int length; /* and count involved */
int count; /* loop counter */
int rqno; /* request index */
int rqoffset; /* offset of request data from parity data */
struct buf *bp; /* user buffer header */
struct request *rq; /* pointer to our request */
struct rqgroup *rqg; /* and to the request group */
struct rqelement *prqe; /* point to the parity block */
struct drive *drive; /* drive to access */
rqg = rqe->rqg; /* and to our request group */
rq = rqg->rq; /* point to our request */
bp = rq->bp; /* user's buffer header */
prqe = &rqg->rqe[0]; /* point to the parity block */
/*
* If we get to this function, we have normal or
* degraded writes, or a combination of both. We do
* the same thing in each case: we perform an
* exclusive or to the parity block. The only
* difference is the origin of the data and the
* address range.
*/
if (rqe->flags & XFR_DEGRADED_WRITE) { /* do the degraded write stuff */
pdata = (int *) (&prqe->b.b_data[(prqe->groupoffset) << DEV_BSHIFT]); /* parity data pointer */
bzero(pdata, prqe->grouplen << DEV_BSHIFT); /* start with nothing in the parity block */
/* Now get what data we need from each block */
for (rqno = 1; rqno < rqg->count; rqno++) { /* for all the data blocks */
/*
* This can do with improvement. If we're doing
* both a degraded and a normal write, we don't
* need to xor (nor to read) the part of the block
* that we're going to overwrite. FIXME XXX
*/
rqe = &rqg->rqe[rqno]; /* this request */
sdata = (int *) (&rqe->b.b_data[rqe->groupoffset << DEV_BSHIFT]); /* old data */
length = rqe->grouplen << (DEV_BSHIFT - 2); /* and count involved */
/*
* add the data block to the parity block. Before
* we started the request, we zeroed the parity
* block, so the result of adding all the other
* blocks and the block we want to write will be
* the correct parity block.
*/
for (count = 0; count < length; count++)
pdata[count] ^= sdata[count];
if ((rqe->flags & XFR_MALLOCED) /* the buffer was malloced, */
&&((rqg->flags & XFR_NORMAL_WRITE) == 0)) { /* and we have no normal write, */
Free(rqe->b.b_data); /* free it now */
rqe->flags &= ~XFR_MALLOCED;
}
}
}
if (rqg->flags & XFR_NORMAL_WRITE) { /* do normal write stuff */
/* Get what data we need from each block */
for (rqno = 1; rqno < rqg->count; rqno++) { /* for all the data blocks */
rqe = &rqg->rqe[rqno]; /* this request */
if ((rqe->flags & (XFR_DATA_BLOCK | XFR_BAD_SUBDISK | XFR_NORMAL_WRITE))
== (XFR_DATA_BLOCK | XFR_NORMAL_WRITE)) { /* good data block to write */
sdata = (int *) &rqe->b.b_data[rqe->dataoffset << DEV_BSHIFT]; /* old data contents */
rqoffset = rqe->dataoffset + rqe->sdoffset - prqe->sdoffset; /* corresponding parity block offset */
pdata = (int *) (&prqe->b.b_data[rqoffset << DEV_BSHIFT]); /* parity data pointer */
length = rqe->datalen << (DEV_BSHIFT - 2); /* and count involved */
/*
* "remove" the old data block
* from the parity block
*/
if ((pdata < ((int *) prqe->b.b_data))
|| (&pdata[length] > ((int *) (prqe->b.b_data + prqe->b.b_bcount)))
|| (sdata < ((int *) rqe->b.b_data))
|| (&sdata[length] > ((int *) (rqe->b.b_data + rqe->b.b_bcount))))
panic("complete_raid5_write: bounds overflow");
for (count = 0; count < length; count++)
pdata[count] ^= sdata[count];
/* "add" the new data block */
sdata = (int *) (&bp->b_data[rqe->useroffset << DEV_BSHIFT]); /* new data */
if ((sdata < ((int *) bp->b_data))
|| (&sdata[length] > ((int *) (bp->b_data + bp->b_bcount))))
panic("complete_raid5_write: bounds overflow");
for (count = 0; count < length; count++)
pdata[count] ^= sdata[count];
/* Free the malloced buffer */
if (rqe->flags & XFR_MALLOCED) { /* the buffer was malloced, */
Free(rqe->b.b_data); /* free it */
rqe->flags &= ~XFR_MALLOCED;
} else
panic("complete_raid5_write: malloc conflict");
if ((rqe->b.b_iocmd == BIO_READ) /* this was a read */
&&((rqe->flags & XFR_BAD_SUBDISK) == 0)) { /* and we can write this block */
rqe->b.b_flags &= ~B_DONE; /* we're writing now */
rqe->b.b_iodone = complete_rqe; /* call us here when done */
rqe->flags &= ~XFR_PARITYOP; /* reset flags that brought us here */
rqe->b.b_data = &bp->b_data[rqe->useroffset << DEV_BSHIFT]; /* point to the user data */
rqe->b.b_bcount = rqe->datalen << DEV_BSHIFT; /* length to write */
rqe->b.b_bufsize = rqe->b.b_bcount; /* don't claim more */
rqe->b.b_resid = rqe->b.b_bcount; /* nothing transferred */
rqe->b.b_blkno += rqe->dataoffset; /* point to the correct block */
rqg->active++; /* another active request */
drive = &DRIVE[rqe->driveno]; /* drive to access */
/* We can't sleep here, so we just increment the counters. */
drive->active++;
if (drive->active >= drive->maxactive)
drive->maxactive = drive->active;
vinum_conf.active++;
if (vinum_conf.active >= vinum_conf.maxactive)
vinum_conf.maxactive = vinum_conf.active;
#if VINUMDEBUG
if (debug & DEBUG_ADDRESSES)
log(LOG_DEBUG,
" %s dev %d.%d, sd %d, offset 0x%x, devoffset 0x%x, length %ld\n",
rqe->b.b_iocmd == BIO_READ ? "Read" : "Write",
major(rqe->b.b_dev),
minor(rqe->b.b_dev),
rqe->sdno,
(u_int) (rqe->b.b_blkno - SD[rqe->sdno].driveoffset),
rqe->b.b_blkno,
rqe->b.b_bcount);
if (debug & DEBUG_LASTREQS)
logrq(loginfo_raid5_data, (union rqinfou) rqe, bp);
#endif
BUF_STRATEGY(&rqe->b, 0);
}
}
}
}
/* Finally, write the parity block */
rqe = &rqg->rqe[0];
rqe->b.b_flags &= ~B_DONE; /* we're writing now */
rqe->b.b_iodone = complete_rqe; /* call us here when done */
rqg->flags &= ~XFR_PARITYOP; /* reset flags that brought us here */
rqe->b.b_bcount = rqe->buflen << DEV_BSHIFT; /* length to write */
rqe->b.b_bufsize = rqe->b.b_bcount; /* don't claim we have more */
rqe->b.b_resid = rqe->b.b_bcount; /* nothing transferred */
rqg->active++; /* another active request */
drive = &DRIVE[rqe->driveno]; /* drive to access */
/* We can't sleep here, so we just increment the counters. */
drive->active++;
if (drive->active >= drive->maxactive)
drive->maxactive = drive->active;
vinum_conf.active++;
if (vinum_conf.active >= vinum_conf.maxactive)
vinum_conf.maxactive = vinum_conf.active;
#if VINUMDEBUG
if (debug & DEBUG_ADDRESSES)
log(LOG_DEBUG,
" %s dev %d.%d, sd %d, offset 0x%x, devoffset 0x%x, length %ld\n",
rqe->b.b_iocmd == BIO_READ ? "Read" : "Write",
major(rqe->b.b_dev),
minor(rqe->b.b_dev),
rqe->sdno,
(u_int) (rqe->b.b_blkno - SD[rqe->sdno].driveoffset),
rqe->b.b_blkno,
rqe->b.b_bcount);
if (debug & DEBUG_LASTREQS)
logrq(loginfo_raid5_parity, (union rqinfou) rqe, bp);
#endif
BUF_STRATEGY(&rqe->b, 0);
}
|