1 /* $NetBSD: ld_ataraid.c,v 1.10 2004/01/25 18:06:48 hannken 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 <sys/cdefs.h>
48 __KERNEL_RCSID(0, "$NetBSD: ld_ataraid.c,v 1.10 2004/01/25 18:06:48 hannken Exp $");
49
50 #include "rnd.h"
51
52 #include <sys/param.h>
53 #include <sys/systm.h>
54 #include <sys/conf.h>
55 #include <sys/kernel.h>
56 #include <sys/device.h>
57 #include <sys/buf.h>
58 #include <sys/dkio.h>
59 #include <sys/disk.h>
60 #include <sys/disklabel.h>
61 #include <sys/fcntl.h>
62 #include <sys/malloc.h>
63 #include <sys/vnode.h>
64 #if NRND > 0
65 #include <sys/rnd.h>
66 #endif
67
68 #include <miscfs/specfs/specdev.h>
69
70 #include <dev/ldvar.h>
71
72 #include <dev/ata/ata_raidvar.h>
73
74 struct ld_ataraid_softc {
75 struct ld_softc sc_ld;
76
77 struct ataraid_array_info *sc_aai;
78 struct vnode *sc_vnodes[ATA_RAID_MAX_DISKS];
79
80 void (*sc_iodone)(struct buf *);
81 };
82
83 static int ld_ataraid_match(struct device *, struct cfdata *, void *);
84 static void ld_ataraid_attach(struct device *, struct device *, void *);
85
86 static int ld_ataraid_dump(struct ld_softc *, void *, int, int);
87
88 static int ld_ataraid_start_span(struct ld_softc *, struct buf *);
89
90 static int ld_ataraid_start_raid0(struct ld_softc *, struct buf *);
91 static void ld_ataraid_iodone_raid0(struct buf *);
92
93 CFATTACH_DECL(ld_ataraid, sizeof(struct ld_ataraid_softc),
94 ld_ataraid_match, ld_ataraid_attach, NULL, NULL);
95
96 static int ld_ataraid_initialized;
97 static struct pool ld_ataraid_cbufpl;
98
99 struct cbuf {
100 struct buf cb_buf; /* new I/O buf */
101 struct buf *cb_obp; /* ptr. to original I/O buf */
102 struct ld_ataraid_softc *cb_sc; /* pointer to ld softc */
103 u_int cb_comp; /* target component */
104 SIMPLEQ_ENTRY(cbuf) cb_q; /* fifo of component buffers */
105 };
106
107 #define CBUF_GET() pool_get(&ld_ataraid_cbufpl, PR_NOWAIT);
108 #define CBUF_PUT(cbp) pool_put(&ld_ataraid_cbufpl, (cbp))
109
110 static int
111 ld_ataraid_match(struct device *parent, struct cfdata *match, void *aux)
112 {
113
114 return (1);
115 }
116
117 static void
118 ld_ataraid_attach(struct device *parent, struct device *self, void *aux)
119 {
120 struct ld_ataraid_softc *sc = (void *) self;
121 struct ld_softc *ld = &sc->sc_ld;
122 struct ataraid_array_info *aai = aux;
123 const char *level;
124 struct vnode *vp;
125 char unklev[32];
126 u_int i;
127
128 if (ld_ataraid_initialized == 0) {
129 ld_ataraid_initialized = 1;
130 pool_init(&ld_ataraid_cbufpl, sizeof(struct cbuf), 0,
131 0, 0, "ldcbuf", NULL);
132 }
133
134 sc->sc_aai = aai; /* this data persists */
135
136 ld->sc_maxxfer = MAXPHYS * aai->aai_width; /* XXX */
137 ld->sc_secperunit = aai->aai_capacity;
138 ld->sc_secsize = 512; /* XXX */
139 ld->sc_maxqueuecnt = 128; /* XXX */
140 ld->sc_dump = ld_ataraid_dump;
141
142 switch (aai->aai_level) {
143 case AAI_L_SPAN:
144 level = "SPAN";
145 ld->sc_start = ld_ataraid_start_span;
146 sc->sc_iodone = ld_ataraid_iodone_raid0;
147 break;
148
149 case AAI_L_RAID0:
150 level = "RAID-0";
151 ld->sc_start = ld_ataraid_start_raid0;
152 sc->sc_iodone = ld_ataraid_iodone_raid0;
153 break;
154
155 case AAI_L_RAID1:
156 level = "RAID-1";
157 break;
158
159 case AAI_L_RAID0 | AAI_L_RAID1:
160 level = "RAID-10";
161 break;
162
163 default:
164 sprintf(unklev, "<unknown level 0x%x>", aai->aai_level);
165 level = unklev;
166 }
167
168 aprint_naive(": ATA %s array\n", level);
169 aprint_normal(": %s ATA %s array\n",
170 ata_raid_type_name(aai->aai_type), level);
171
172 if (ld->sc_start == NULL) {
173 aprint_error("%s: unsupported array type\n",
174 ld->sc_dv.dv_xname);
175 return;
176 }
177
178 /*
179 * We get a geometry from the device; use it.
180 */
181 ld->sc_nheads = aai->aai_heads;
182 ld->sc_nsectors = aai->aai_sectors;
183 ld->sc_ncylinders = aai->aai_cylinders;
184
185 /*
186 * Configure all the component disks.
187 */
188 for (i = 0; i < aai->aai_ndisks; i++) {
189 struct ataraid_disk_info *adi = &aai->aai_disks[i];
190 int bmajor, error;
191 dev_t dev;
192
193 bmajor = devsw_name2blk(adi->adi_dev->dv_xname, NULL, 0);
194 dev = MAKEDISKDEV(bmajor, adi->adi_dev->dv_unit, RAW_PART);
195 error = bdevvp(dev, &vp);
196 if (error)
197 break;
198 error = VOP_OPEN(vp, FREAD|FWRITE, NOCRED, 0);
199 if (error) {
200 vput(vp);
201 /*
202 * XXX This is bogus. We should just mark the
203 * XXX component as FAILED, and write-back new
204 * XXX config blocks.
205 */
206 break;
207 }
208
209 VOP_UNLOCK(vp, 0);
210 sc->sc_vnodes[i] = vp;
211 }
212 if (i == aai->aai_ndisks) {
213 ld->sc_flags = LDF_ENABLED;
214 goto finish;
215 }
216
217 for (i = 0; i < aai->aai_ndisks; i++) {
218 vp = sc->sc_vnodes[i];
219 sc->sc_vnodes[i] = NULL;
220 if (vp != NULL)
221 (void) vn_close(vp, FREAD|FWRITE, NOCRED, curproc);
222 }
223
224 finish:
225 ldattach(ld);
226 }
227
228 static struct cbuf *
229 ld_ataraid_make_cbuf(struct ld_ataraid_softc *sc, struct buf *bp,
230 u_int comp, daddr_t bn, caddr_t addr, long bcount)
231 {
232 struct cbuf *cbp;
233
234 cbp = CBUF_GET();
235 if (cbp == NULL)
236 return (NULL);
237 BUF_INIT(&cbp->cb_buf);
238 cbp->cb_buf.b_flags = bp->b_flags | B_CALL;
239 cbp->cb_buf.b_iodone = sc->sc_iodone;
240 cbp->cb_buf.b_proc = bp->b_proc;
241 cbp->cb_buf.b_vp = sc->sc_vnodes[comp];
242 cbp->cb_buf.b_blkno = bn + sc->sc_aai->aai_offset;
243 cbp->cb_buf.b_data = addr;
244 cbp->cb_buf.b_bcount = bcount;
245
246 /* Context for iodone */
247 cbp->cb_obp = bp;
248 cbp->cb_sc = sc;
249 cbp->cb_comp = comp;
250
251 return (cbp);
252 }
253
254 static int
255 ld_ataraid_start_span(struct ld_softc *ld, struct buf *bp)
256 {
257 struct ld_ataraid_softc *sc = (void *) ld;
258 struct ataraid_array_info *aai = sc->sc_aai;
259 struct ataraid_disk_info *adi;
260 SIMPLEQ_HEAD(, cbuf) cbufq;
261 struct cbuf *cbp;
262 caddr_t addr;
263 daddr_t bn;
264 long bcount, rcount;
265 u_int comp;
266
267 /* Allocate component buffers. */
268 SIMPLEQ_INIT(&cbufq);
269 addr = bp->b_data;
270
271 /* Find the first component. */
272 comp = 0;
273 adi = &aai->aai_disks[comp];
274 bn = bp->b_rawblkno;
275 while (bn >= adi->adi_compsize) {
276 bn -= adi->adi_compsize;
277 adi = &aai->aai_disks[++comp];
278 }
279
280 bp->b_resid = bp->b_bcount;
281
282 for (bcount = bp->b_bcount; bcount > 0; bcount -= rcount) {
283 rcount = bp->b_bcount;
284 if ((adi->adi_compsize - bn) < btodb(rcount))
285 rcount = dbtob(adi->adi_compsize - bn);
286
287 cbp = ld_ataraid_make_cbuf(sc, bp, comp, bn, addr, rcount);
288 if (cbp == NULL) {
289 /* Free the already allocated component buffers. */
290 while ((cbp = SIMPLEQ_FIRST(&cbufq)) != NULL) {
291 SIMPLEQ_REMOVE_HEAD(&cbufq, cb_q);
292 CBUF_PUT(cbp);
293 }
294 return (EAGAIN);
295 }
296
297 /*
298 * For a span, we always know we advance to the next disk,
299 * and always start at offset 0 on that disk.
300 */
301 adi = &aai->aai_disks[++comp];
302 bn = 0;
303
304 SIMPLEQ_INSERT_TAIL(&cbufq, cbp, cb_q);
305 addr += rcount;
306 }
307
308 /* Now fire off the requests. */
309 while ((cbp = SIMPLEQ_FIRST(&cbufq)) != NULL) {
310 SIMPLEQ_REMOVE_HEAD(&cbufq, cb_q);
311 if ((cbp->cb_buf.b_flags & B_READ) == 0)
312 cbp->cb_buf.b_vp->v_numoutput++;
313 VOP_STRATEGY(cbp->cb_buf.b_vp, &cbp->cb_buf);
314 }
315
316 return (0);
317 }
318
319 static int
320 ld_ataraid_start_raid0(struct ld_softc *ld, struct buf *bp)
321 {
322 struct ld_ataraid_softc *sc = (void *) ld;
323 struct ataraid_array_info *aai = sc->sc_aai;
324 SIMPLEQ_HEAD(, cbuf) cbufq;
325 struct cbuf *cbp;
326 caddr_t addr;
327 daddr_t bn, cbn, tbn, off;
328 long bcount, rcount;
329 u_int comp;
330
331 /* Allocate component buffers. */
332 SIMPLEQ_INIT(&cbufq);
333 addr = bp->b_data;
334 bn = bp->b_rawblkno;
335
336 bp->b_resid = bp->b_bcount;
337
338 for (bcount = bp->b_bcount; bcount > 0; bcount -= rcount) {
339 tbn = bn / aai->aai_interleave;
340 off = bn % aai->aai_interleave;
341
342 if (__predict_false(tbn == aai->aai_capacity /
343 aai->aai_interleave)) {
344 /* Last stripe. */
345 daddr_t sz = (aai->aai_capacity -
346 (tbn * aai->aai_interleave)) /
347 aai->aai_width;
348 comp = off / sz;
349 cbn = ((tbn / aai->aai_width) * aai->aai_interleave) +
350 (off % sz);
351 rcount = min(bcount, dbtob(sz));
352 } else {
353 comp = tbn % aai->aai_width;
354 cbn = ((tbn / aai->aai_width) * aai->aai_interleave) +
355 off;
356 rcount = min(bcount, dbtob(aai->aai_interleave - off));
357 }
358
359 cbp = ld_ataraid_make_cbuf(sc, bp, comp, cbn, addr, rcount);
360 if (cbp == NULL) {
361 /* Free the already allocated component buffers. */
362 while ((cbp = SIMPLEQ_FIRST(&cbufq)) != NULL) {
363 SIMPLEQ_REMOVE_HEAD(&cbufq, cb_q);
364 CBUF_PUT(cbp);
365 }
366 return (EAGAIN);
367 }
368 SIMPLEQ_INSERT_TAIL(&cbufq, cbp, cb_q);
369 bn += btodb(rcount);
370 addr += rcount;
371 }
372
373 /* Now fire off the requests. */
374 while ((cbp = SIMPLEQ_FIRST(&cbufq)) != NULL) {
375 SIMPLEQ_REMOVE_HEAD(&cbufq, cb_q);
376 if ((cbp->cb_buf.b_flags & B_READ) == 0)
377 cbp->cb_buf.b_vp->v_numoutput++;
378 VOP_STRATEGY(cbp->cb_buf.b_vp, &cbp->cb_buf);
379 }
380
381 return (0);
382 }
383
384 /*
385 * Called at interrupt time. Mark the component as done and if all
386 * components are done, take an "interrupt".
387 */
388 static void
389 ld_ataraid_iodone_raid0(struct buf *vbp)
390 {
391 struct cbuf *cbp = (struct cbuf *) vbp;
392 struct buf *bp = cbp->cb_obp;
393 struct ld_ataraid_softc *sc = cbp->cb_sc;
394 long count;
395 int s;
396
397 s = splbio();
398
399 if (cbp->cb_buf.b_flags & B_ERROR) {
400 bp->b_flags |= B_ERROR;
401 bp->b_error = cbp->cb_buf.b_error ?
402 cbp->cb_buf.b_error : EIO;
403
404 /* XXX Update component config blocks. */
405
406 printf("%s: error %d on component %d\n",
407 sc->sc_ld.sc_dv.dv_xname, bp->b_error, cbp->cb_comp);
408 }
409 count = cbp->cb_buf.b_bcount;
410 CBUF_PUT(cbp);
411
412 /* If all done, "interrupt". */
413 bp->b_resid -= count;
414 if (bp->b_resid < 0)
415 panic("ld_ataraid_iodone_raid0: count");
416 if (bp->b_resid == 0)
417 lddone(&sc->sc_ld, bp);
418 splx(s);
419 }
420
421 static int
422 ld_ataraid_dump(struct ld_softc *sc, void *data, int blkno, int blkcnt)
423 {
424
425 return (EIO);
426 }
Cache object: 4ae5a11afea4599dc9517c522cc9720b
|