ld_ataraid.c revision 1.1 1 /* $NetBSD: ld_ataraid.c,v 1.1 2003/01/27 18:21:29 thorpej Exp $ */
2
3 /*
4 * Copyright (c) 2003 Wasabi Systems, Inc.
5 * All rights reserved.
6 *
7 * Written by Jason R. Thorpe for Wasabi Systems, Inc.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed for the NetBSD Project by
20 * Wasabi Systems, Inc.
21 * 4. The name of Wasabi Systems, Inc. may not be used to endorse
22 * or promote products derived from this software without specific prior
23 * written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
27 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
28 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC
29 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 * POSSIBILITY OF SUCH DAMAGE.
36 */
37
38 /*
39 * Support for ATA RAID logical disks.
40 *
41 * Note that all the RAID happens in software here; the ATA RAID
42 * controllers we're dealing with (Promise, etc.) only support
43 * configuration data on the component disks, with the BIOS supporting
44 * booting from the RAID volumes.
45 */
46
47 #include "rnd.h"
48
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/conf.h>
52 #include <sys/kernel.h>
53 #include <sys/device.h>
54 #include <sys/buf.h>
55 #include <sys/dkio.h>
56 #include <sys/disk.h>
57 #include <sys/disklabel.h>
58 #include <sys/fcntl.h>
59 #include <sys/malloc.h>
60 #include <sys/vnode.h>
61 #if NRND > 0
62 #include <sys/rnd.h>
63 #endif
64
65 #include <miscfs/specfs/specdev.h>
66
67 #include <dev/ldvar.h>
68
69 #include <dev/ata/ata_raidvar.h>
70
71 struct ld_ataraid_softc {
72 struct ld_softc sc_ld;
73
74 struct ataraid_array_info *sc_aai;
75 struct iinfo *sc_itable;
76 struct vnode *sc_vnodes[ATA_RAID_MAX_DISKS];
77
78 void (*sc_iodone)(struct buf *);
79 };
80
81 static int ld_ataraid_match(struct device *, struct cfdata *, void *);
82 static void ld_ataraid_attach(struct device *, struct device *, void *);
83
84 static int ld_ataraid_dump(struct ld_softc *, void *, int, int);
85
86 static int ld_ataraid_start_span(struct ld_softc *, struct buf *);
87
88 static int ld_ataraid_start_raid0(struct ld_softc *, struct buf *);
89 static void ld_ataraid_iodone_raid0(struct buf *);
90
91 CFATTACH_DECL(ld_ataraid, sizeof(struct ld_ataraid_softc),
92 ld_ataraid_match, ld_ataraid_attach, NULL, NULL);
93
94 static int ld_ataraid_initialized;
95 static struct pool ld_ataraid_cbufpl;
96
97 struct cbuf {
98 struct buf cb_buf; /* new I/O buf */
99 struct buf *cb_obp; /* ptr. to original I/O buf */
100 struct ld_ataraid_softc *cb_sc; /* pointer to ld softc */
101 u_int cb_comp; /* target component */
102 SIMPLEQ_ENTRY(cbuf) cb_q; /* fifo of component buffers */
103 };
104
105 #define CBUF_GET() pool_get(&ld_ataraid_cbufpl, PR_NOWAIT);
106 #define CBUF_PUT(cbp) pool_put(&ld_ataraid_cbufpl, (cbp))
107
108 static int
109 ld_ataraid_match(struct device *parent, struct cfdata *match, void *aux)
110 {
111
112 return (1);
113 }
114
115 static void
116 ld_ataraid_attach(struct device *parent, struct device *self, void *aux)
117 {
118 struct ld_ataraid_softc *sc = (void *) self;
119 struct ld_softc *ld = &sc->sc_ld;
120 struct ataraid_array_info *aai = aux;
121 const char *level;
122 struct vnode *vp;
123 char unklev[32];
124 u_int i;
125
126 if (ld_ataraid_initialized == 0) {
127 ld_ataraid_initialized = 1;
128 pool_init(&ld_ataraid_cbufpl, sizeof(struct cbuf), 0,
129 0, 0, "ldcbuf", NULL);
130 }
131
132 sc->sc_aai = aai; /* this data persists */
133
134 ld->sc_flags = LDF_ENABLED;
135 ld->sc_maxxfer = MAXPHYS * aai->aai_width; /* XXX */
136 ld->sc_secperunit = aai->aai_capacity;
137 ld->sc_secsize = 512; /* XXX */
138 ld->sc_maxqueuecnt = 128; /* XXX */
139 ld->sc_dump = ld_ataraid_dump;
140
141 switch (aai->aai_level) {
142 case AAI_L_SPAN:
143 level = "SPAN";
144 ld->sc_start = ld_ataraid_start_span;
145 sc->sc_iodone = ld_ataraid_iodone_raid0;
146 break;
147
148 case AAI_L_RAID0:
149 level = "RAID0";
150 ld->sc_start = ld_ataraid_start_raid0;
151 sc->sc_iodone = ld_ataraid_iodone_raid0;
152 break;
153
154 case AAI_L_RAID1:
155 level = "RAID1";
156 break;
157
158 case AAI_L_RAID0 | AAI_L_RAID1:
159 level = "RAID0+1";
160 break;
161
162 default:
163 sprintf(unklev, "<unknown level 0x%x>", aai->aai_level);
164 level = unklev;
165 }
166
167 aprint_naive(": ATA %s array\n", level);
168 aprint_normal(": %s ATA %s array\n",
169 ata_raid_type_name(aai->aai_type), level);
170
171 if (ld->sc_start == NULL) {
172 aprint_error("%s: unsupported array type\n",
173 ld->sc_dv.dv_xname);
174 return;
175 }
176
177 /*
178 * We get a geometry from the device; use it.
179 */
180 ld->sc_nheads = aai->aai_heads;
181 ld->sc_nsectors = aai->aai_sectors;
182 ld->sc_ncylinders = aai->aai_cylinders;
183
184 /*
185 * Configure all the component disks.
186 */
187 for (i = 0; i < aai->aai_ndisks; i++) {
188 struct ataraid_disk_info *adi = &aai->aai_disks[i];
189 int bmajor, error;
190 dev_t dev;
191
192 bmajor = devsw_name2blk(adi->adi_dev->dv_xname, NULL, 0);
193 dev = MAKEDISKDEV(bmajor, adi->adi_dev->dv_unit, RAW_PART);
194 error = bdevvp(dev, &vp);
195 if (error)
196 goto bad;
197 error = VOP_OPEN(vp, FREAD|FWRITE, NOCRED, 0);
198 if (error) {
199 vput(vp);
200 /*
201 * XXX This is bogus. We should just mark the
202 * XXX component as FAILED, and write-back new
203 * XXX config blocks.
204 */
205 goto bad;
206 }
207
208 VOP_UNLOCK(vp, 0);
209 sc->sc_vnodes[i] = vp;
210 }
211
212 ldattach(ld);
213 return;
214 bad:
215 for (i = 0; i < aai->aai_ndisks; i++) {
216 vp = sc->sc_vnodes[i];
217 sc->sc_vnodes[i] = NULL;
218 (void) vn_close(vp, FREAD|FWRITE, NOCRED, curproc);
219 }
220 }
221
222 static struct cbuf *
223 ld_ataraid_make_cbuf(struct ld_ataraid_softc *sc, struct buf *bp,
224 u_int comp, daddr_t bn, caddr_t addr, long bcount)
225 {
226 struct cbuf *cbp;
227
228 cbp = CBUF_GET();
229 if (cbp == NULL)
230 return (NULL);
231 cbp->cb_buf.b_flags = bp->b_flags | B_CALL;
232 cbp->cb_buf.b_iodone = sc->sc_iodone;
233 cbp->cb_buf.b_proc = bp->b_proc;
234 cbp->cb_buf.b_vp = sc->sc_vnodes[comp];
235 cbp->cb_buf.b_dev = sc->sc_vnodes[comp]->v_rdev;
236 cbp->cb_buf.b_blkno = bn + sc->sc_aai->aai_offset;
237 cbp->cb_buf.b_data = addr;
238 LIST_INIT(&cbp->cb_buf.b_dep);
239 cbp->cb_buf.b_bcount = bcount;
240
241 /* Context for iodone */
242 cbp->cb_obp = bp;
243 cbp->cb_sc = sc;
244 cbp->cb_comp = comp;
245
246 return (cbp);
247 }
248
249 static int
250 ld_ataraid_start_span(struct ld_softc *ld, struct buf *bp)
251 {
252 struct ld_ataraid_softc *sc = (void *) ld;
253 struct ataraid_array_info *aai = sc->sc_aai;
254 struct ataraid_disk_info *adi;
255 SIMPLEQ_HEAD(, cbuf) cbufq;
256 struct cbuf *cbp;
257 caddr_t addr;
258 daddr_t bn;
259 long bcount, rcount;
260 u_int comp;
261 int s;
262
263 /* Allocate component buffers. */
264 SIMPLEQ_INIT(&cbufq);
265 addr = bp->b_data;
266
267 /* Find the first component. */
268 comp = 0;
269 adi = &aai->aai_disks[comp];
270 bn = bp->b_rawblkno;
271 while (bn >= adi->adi_compsize) {
272 bn -= adi->adi_compsize;
273 adi = &aai->aai_disks[++comp];
274 }
275
276 s = splbio(); /* XXX big hammer */
277
278 bp->b_resid = bp->b_bcount;
279
280 for (bcount = bp->b_bcount; bcount > 0; bcount -= rcount) {
281 rcount = bp->b_bcount;
282 if ((adi->adi_compsize - bn) < btodb(rcount))
283 rcount = dbtob(adi->adi_compsize - bn);
284
285 cbp = ld_ataraid_make_cbuf(sc, bp, comp, bn, addr, rcount);
286 if (cbp == NULL) {
287 /* Free the already allocated component buffers. */
288 while ((cbp = SIMPLEQ_FIRST(&cbufq)) != NULL) {
289 SIMPLEQ_REMOVE_HEAD(&cbufq, cb_q);
290 CBUF_PUT(cbp);
291 }
292
293 splx(s);
294
295 /* Notify the upper layer that we are out of memory. */
296 return (ENOMEM);
297 }
298
299 /*
300 * For a span, we always know we advance to the next disk,
301 * and always start at offset 0 on that disk.
302 */
303 adi = &aai->aai_disks[++comp];
304 bn = 0;
305
306 SIMPLEQ_INSERT_TAIL(&cbufq, cbp, cb_q);
307 addr += rcount;
308 }
309
310 /* Now fire off the requests. */
311 while ((cbp = SIMPLEQ_FIRST(&cbufq)) != NULL) {
312 SIMPLEQ_REMOVE_HEAD(&cbufq, cb_q);
313 if ((cbp->cb_buf.b_flags & B_READ) == 0)
314 cbp->cb_buf.b_vp->v_numoutput++;
315 VOP_STRATEGY(&cbp->cb_buf);
316 }
317
318 splx(s);
319
320 return (0);
321 }
322
323 static int
324 ld_ataraid_start_raid0(struct ld_softc *ld, struct buf *bp)
325 {
326 struct ld_ataraid_softc *sc = (void *) ld;
327 struct ataraid_array_info *aai = sc->sc_aai;
328 SIMPLEQ_HEAD(, cbuf) cbufq;
329 struct cbuf *cbp;
330 caddr_t addr;
331 daddr_t bn, cbn, tbn, off;
332 long bcount, rcount;
333 u_int comp;
334 int s;
335
336 /* Allocate component buffers. */
337 SIMPLEQ_INIT(&cbufq);
338 addr = bp->b_data;
339 bn = bp->b_rawblkno;
340
341 s = splbio(); /* XXX big hammer */
342
343 bp->b_resid = bp->b_bcount;
344
345 for (bcount = bp->b_bcount; bcount > 0; bcount -= rcount) {
346 tbn = bn / aai->aai_interleave;
347 off = bn % aai->aai_interleave;
348
349 if (__predict_false(tbn == aai->aai_capacity /
350 aai->aai_interleave)) {
351 /* Last stripe. */
352 daddr_t sz = (aai->aai_capacity -
353 (tbn * aai->aai_interleave)) /
354 aai->aai_width;
355 comp = off / sz;
356 cbn = ((tbn / aai->aai_width) * aai->aai_interleave) +
357 (off % sz);
358 rcount = min(bcount, dbtob(sz));
359 } else {
360 comp = tbn % aai->aai_width;
361 cbn = ((tbn / aai->aai_width) * aai->aai_interleave) +
362 off;
363 rcount = min(bcount, dbtob(aai->aai_interleave - off));
364 }
365
366 cbp = ld_ataraid_make_cbuf(sc, bp, comp, cbn, addr, rcount);
367 if (cbp == NULL) {
368 /* Free the already allocated component buffers. */
369 while ((cbp = SIMPLEQ_FIRST(&cbufq)) != NULL) {
370 SIMPLEQ_REMOVE_HEAD(&cbufq, cb_q);
371 CBUF_PUT(cbp);
372 }
373
374 splx(s);
375
376 /* Notify the upper layer that we are out of memory. */
377 return (ENOMEM);
378 }
379 SIMPLEQ_INSERT_TAIL(&cbufq, cbp, cb_q);
380 bn += btodb(rcount);
381 addr += rcount;
382 }
383
384 /* Now fire off the requests. */
385 while ((cbp = SIMPLEQ_FIRST(&cbufq)) != NULL) {
386 SIMPLEQ_REMOVE_HEAD(&cbufq, cb_q);
387 if ((cbp->cb_buf.b_flags & B_READ) == 0)
388 cbp->cb_buf.b_vp->v_numoutput++;
389 VOP_STRATEGY(&cbp->cb_buf);
390 }
391
392 splx(s);
393
394 return (0);
395 }
396
397 /*
398 * Called at interrupt time. Mark the component as done and if all
399 * components are done, take an "interrupt".
400 */
401 static void
402 ld_ataraid_iodone_raid0(struct buf *vbp)
403 {
404 struct cbuf *cbp = (struct cbuf *) vbp;
405 struct buf *bp = cbp->cb_obp;
406 struct ld_ataraid_softc *sc = cbp->cb_sc;
407 long count;
408 int s;
409
410 s = splbio();
411
412 if (cbp->cb_buf.b_flags & B_ERROR) {
413 bp->b_flags |= B_ERROR;
414 bp->b_error = cbp->cb_buf.b_error ?
415 cbp->cb_buf.b_error : EIO;
416
417 /* XXX Update component config blocks. */
418
419 printf("%s: error %d on component %d\n",
420 sc->sc_ld.sc_dv.dv_xname, bp->b_error, cbp->cb_comp);
421 }
422 count = cbp->cb_buf.b_bcount;
423 CBUF_PUT(cbp);
424
425 /* If all done, "interrupt". */
426 bp->b_resid -= count;
427 if (bp->b_resid < 0)
428 panic("ld_ataraid_iodone_raid0: count");
429 if (bp->b_resid == 0)
430 lddone(&sc->sc_ld, bp);
431 splx(s);
432 }
433
434 static int
435 ld_ataraid_dump(struct ld_softc *sc, void *data, int blkno, int blkcnt)
436 {
437
438 return (EIO);
439 }
440