udf_strat_direct.c revision 1.1.8.2 1 /* $NetBSD: udf_strat_direct.c,v 1.1.8.2 2008/07/31 04:51:02 simonb Exp $ */
2
3 /*
4 * Copyright (c) 2006, 2008 Reinoud Zandijk
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26 *
27 */
28
29 #include <sys/cdefs.h>
30 #ifndef lint
31 __KERNEL_RCSID(0, "$NetBSD: udf_strat_direct.c,v 1.1.8.2 2008/07/31 04:51:02 simonb Exp $");
32 #endif /* not lint */
33
34
35 #if defined(_KERNEL_OPT)
36 #include "opt_quota.h"
37 #include "opt_compat_netbsd.h"
38 #endif
39
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/sysctl.h>
43 #include <sys/namei.h>
44 #include <sys/proc.h>
45 #include <sys/kernel.h>
46 #include <sys/vnode.h>
47 #include <miscfs/genfs/genfs_node.h>
48 #include <sys/mount.h>
49 #include <sys/buf.h>
50 #include <sys/file.h>
51 #include <sys/device.h>
52 #include <sys/disklabel.h>
53 #include <sys/ioctl.h>
54 #include <sys/malloc.h>
55 #include <sys/dirent.h>
56 #include <sys/stat.h>
57 #include <sys/conf.h>
58 #include <sys/kauth.h>
59 #include <sys/kthread.h>
60 #include <dev/clock_subr.h>
61
62 #include <fs/udf/ecma167-udf.h>
63 #include <fs/udf/udf_mount.h>
64
65 #if defined(_KERNEL_OPT)
66 #include "opt_udf.h"
67 #endif
68
69 #include "udf.h"
70 #include "udf_subr.h"
71 #include "udf_bswap.h"
72
73
74 #define VTOI(vnode) ((struct udf_node *) vnode->v_data)
75 #define PRIV(ump) ((struct strat_private *) ump->strategy_private)
76
77 /* --------------------------------------------------------------------- */
78
79 /* BUFQ's */
80 #define UDF_SHED_MAX 3
81
82 #define UDF_SHED_READING 0
83 #define UDF_SHED_WRITING 1
84 #define UDF_SHED_SEQWRITING 2
85
86
87 struct strat_private {
88 struct pool desc_pool; /* node descriptors */
89 };
90
91 /* --------------------------------------------------------------------- */
92
93 static void
94 udf_wr_nodedscr_callback(struct buf *buf)
95 {
96 struct udf_node *udf_node;
97
98 KASSERT(buf);
99 KASSERT(buf->b_data);
100
101 /* called when write action is done */
102 DPRINTF(WRITE, ("udf_wr_nodedscr_callback(): node written out\n"));
103
104 udf_node = VTOI(buf->b_vp);
105 if (udf_node == NULL) {
106 putiobuf(buf);
107 printf("udf_wr_node_callback: NULL node?\n");
108 return;
109 }
110
111 /* XXX right flags to mark dirty again on error? */
112 if (buf->b_error) {
113 /* write error on `defect free' media??? how to solve? */
114 /* XXX lookup UDF standard for unallocatable space */
115 udf_node->i_flags |= IN_MODIFIED | IN_ACCESSED;
116 }
117
118 /* decrement outstanding_nodedscr */
119 KASSERT(udf_node->outstanding_nodedscr >= 1);
120 udf_node->outstanding_nodedscr--;
121 if (udf_node->outstanding_nodedscr == 0) {
122 /* unlock the node */
123 KASSERT(udf_node->i_flags & IN_CALLBACK_ULK);
124 UDF_UNLOCK_NODE(udf_node, IN_CALLBACK_ULK);
125
126 wakeup(&udf_node->outstanding_nodedscr);
127 }
128 /* unreference the vnode so it can be recycled */
129 holdrele(udf_node->vnode);
130
131 putiobuf(buf);
132 }
133
134 /* --------------------------------------------------------------------- */
135
136 static int
137 udf_getblank_nodedscr_direct(struct udf_strat_args *args)
138 {
139 union dscrptr **dscrptr = &args->dscr;
140 struct udf_mount *ump = args->ump;
141 struct strat_private *priv = PRIV(ump);
142 uint32_t lb_size;
143
144 lb_size = udf_rw32(ump->logical_vol->lb_size);
145 *dscrptr = pool_get(&priv->desc_pool, PR_WAITOK);
146 memset(*dscrptr, 0, lb_size);
147
148 return 0;
149 }
150
151
152 static void
153 udf_free_nodedscr_direct(struct udf_strat_args *args)
154 {
155 union dscrptr *dscr = args->dscr;
156 struct udf_mount *ump = args->ump;
157 struct strat_private *priv = PRIV(ump);
158
159 pool_put(&priv->desc_pool, dscr);
160 }
161
162
163 static int
164 udf_read_nodedscr_direct(struct udf_strat_args *args)
165 {
166 union dscrptr **dscrptr = &args->dscr;
167 union dscrptr *tmpdscr;
168 struct udf_mount *ump = args->ump;
169 struct long_ad *icb = args->icb;
170 struct strat_private *priv = PRIV(ump);
171 uint32_t lb_size;
172 uint32_t sector, dummy;
173 int error;
174
175 lb_size = udf_rw32(ump->logical_vol->lb_size);
176
177 error = udf_translate_vtop(ump, icb, §or, &dummy);
178 if (error)
179 return error;
180
181 /* try to read in fe/efe */
182 error = udf_read_phys_dscr(ump, sector, M_UDFTEMP, &tmpdscr);
183 if (error)
184 return error;
185
186 *dscrptr = pool_get(&priv->desc_pool, PR_WAITOK);
187 memcpy(*dscrptr, tmpdscr, lb_size);
188 free(tmpdscr, M_UDFTEMP);
189
190 return 0;
191 }
192
193
194 static int
195 udf_write_nodedscr_direct(struct udf_strat_args *args)
196 {
197 struct udf_mount *ump = args->ump;
198 struct udf_node *udf_node = args->udf_node;
199 union dscrptr *dscr = args->dscr;
200 struct long_ad *icb = args->icb;
201 int waitfor = args->waitfor;
202 uint32_t logsector, sector, dummy;
203 int error, vpart;
204
205 /*
206 * we have to decide if we write it out sequential or at its fixed
207 * position by examining the partition its (to be) written on.
208 */
209 vpart = udf_rw16(udf_node->loc.loc.part_num);
210 logsector = udf_rw32(icb->loc.lb_num);
211 KASSERT(ump->vtop_tp[vpart] != UDF_VTOP_TYPE_VIRT);
212
213 sector = 0;
214 error = udf_translate_vtop(ump, icb, §or, &dummy);
215 if (error)
216 goto out;
217
218 /* add reference to the vnode to prevent recycling */
219 vhold(udf_node->vnode);
220
221 if (waitfor) {
222 DPRINTF(WRITE, ("udf_write_nodedscr: sync write\n"));
223
224 error = udf_write_phys_dscr_sync(ump, udf_node, UDF_C_NODE,
225 dscr, sector, logsector);
226 } else {
227 DPRINTF(WRITE, ("udf_write_nodedscr: no wait, async write\n"));
228
229 error = udf_write_phys_dscr_async(ump, udf_node, UDF_C_NODE,
230 dscr, sector, logsector, udf_wr_nodedscr_callback);
231 /* will be UNLOCKED in call back */
232 return error;
233 }
234
235 holdrele(udf_node->vnode);
236 out:
237 udf_node->outstanding_nodedscr--;
238 if (udf_node->outstanding_nodedscr == 0) {
239 UDF_UNLOCK_NODE(udf_node, 0);
240 wakeup(&udf_node->outstanding_nodedscr);
241 }
242
243 return error;
244 }
245
246 /* --------------------------------------------------------------------- */
247
248 static void
249 udf_queue_buf_direct(struct udf_strat_args *args)
250 {
251 struct udf_mount *ump = args->ump;
252 struct buf *buf = args->nestbuf;
253 struct buf *nestbuf;
254 struct desc_tag *tag;
255 struct long_ad *node_ad_cpy;
256 uint64_t *lmapping, *pmapping, *lmappos, blknr, run_start;
257 uint32_t our_sectornr, sectornr;
258 uint32_t lb_size, buf_offset, rbuflen, bpos;
259 uint16_t vpart_num;
260 uint8_t *fidblk;
261 off_t rblk;
262 int sector_size = ump->discinfo.sector_size;
263 int blks = sector_size / DEV_BSIZE;
264 int len, buf_len, sector, sectors, run_length;
265 int what, class, queue;
266
267 KASSERT(ump);
268 KASSERT(buf);
269 KASSERT(buf->b_iodone == nestiobuf_iodone);
270
271 what = buf->b_udf_c_type;
272 queue = UDF_SHED_READING;
273 if ((buf->b_flags & B_READ) == 0) {
274 /* writing */
275 queue = UDF_SHED_SEQWRITING;
276 if (what == UDF_C_DSCR)
277 queue = UDF_SHED_WRITING;
278 if (what == UDF_C_NODE)
279 queue = UDF_SHED_WRITING;
280 }
281
282 /* use disc sheduler */
283 class = ump->discinfo.mmc_class;
284 KASSERT((class == MMC_CLASS_UNKN) || (class == MMC_CLASS_DISC) ||
285 (ump->discinfo.mmc_cur & MMC_CAP_HW_DEFECTFREE));
286
287 if (queue == UDF_SHED_READING) {
288 DPRINTF(SHEDULE, ("\nudf_issue_buf READ %p : sector %d type %d,"
289 "b_resid %d, b_bcount %d, b_bufsize %d\n",
290 buf, (uint32_t) buf->b_blkno / blks, buf->b_udf_c_type,
291 buf->b_resid, buf->b_bcount, buf->b_bufsize));
292 VOP_STRATEGY(ump->devvp, buf);
293 return;
294 }
295
296 /* (sectorsize == lb_size) for UDF */
297 lb_size = udf_rw32(ump->logical_vol->lb_size);
298 blknr = buf->b_blkno;
299 our_sectornr = blknr / blks;
300
301 if (queue == UDF_SHED_WRITING) {
302 DPRINTF(SHEDULE, ("\nudf_issue_buf WRITE %p : sector %d "
303 "type %d, b_resid %d, b_bcount %d, b_bufsize %d\n",
304 buf, (uint32_t) buf->b_blkno / blks, buf->b_udf_c_type,
305 buf->b_resid, buf->b_bcount, buf->b_bufsize));
306 /* if we have FIDs fixup using buffer's sector number(s) */
307 if (buf->b_udf_c_type == UDF_C_FIDS) {
308 panic("UDF_C_FIDS in SHED_WRITING!\n");
309 buf_len = buf->b_bcount;
310 sectornr = our_sectornr;
311 bpos = 0;
312 while (buf_len) {
313 len = MIN(buf_len, sector_size);
314 fidblk = (uint8_t *) buf->b_data + bpos;
315 udf_fixup_fid_block(fidblk, sector_size,
316 0, len, sectornr);
317 sectornr++;
318 bpos += len;
319 buf_len -= len;
320 }
321 }
322 udf_fixup_node_internals(ump, buf->b_data, buf->b_udf_c_type);
323 VOP_STRATEGY(ump->devvp, buf);
324 return;
325 }
326
327 /* UDF_SHED_SEQWRITING */
328 KASSERT(queue == UDF_SHED_SEQWRITING);
329 DPRINTF(SHEDULE, ("\nudf_issue_buf SEQWRITE %p : sector XXXX "
330 "type %d, b_resid %d, b_bcount %d, b_bufsize %d\n",
331 buf, buf->b_udf_c_type, buf->b_resid, buf->b_bcount,
332 buf->b_bufsize));
333
334 /*
335 * Buffers should not have been allocated to disc addresses yet on
336 * this queue. Note that a buffer can get multiple extents allocated.
337 *
338 * lmapping contains lb_num relative to base partition.
339 */
340 lmapping = ump->la_lmapping;
341 node_ad_cpy = ump->la_node_ad_cpy;
342
343 /* logically allocate buf and map it in the file */
344 udf_late_allocate_buf(ump, buf, lmapping, node_ad_cpy, &vpart_num);
345
346 /* if we have FIDs, fixup using the new allocation table */
347 if (buf->b_udf_c_type == UDF_C_FIDS) {
348 buf_len = buf->b_bcount;
349 bpos = 0;
350 lmappos = lmapping;
351 while (buf_len) {
352 sectornr = *lmappos++;
353 len = MIN(buf_len, sector_size);
354 fidblk = (uint8_t *) buf->b_data + bpos;
355 udf_fixup_fid_block(fidblk, sector_size,
356 0, len, sectornr);
357 bpos += len;
358 buf_len -= len;
359 }
360 }
361 if (buf->b_udf_c_type == UDF_C_METADATA_SBM) {
362 if (buf->b_lblkno == 0) {
363 /* update the tag location inside */
364 tag = (struct desc_tag *) buf->b_data;
365 tag->tag_loc = udf_rw32(buf->b_blkno);
366 udf_validate_tag_and_crc_sums(buf->b_data);
367 }
368 }
369 udf_fixup_node_internals(ump, buf->b_data, buf->b_udf_c_type);
370
371 /*
372 * Translate new mappings in lmapping to pmappings and try to
373 * conglomerate extents to reduce the number of writes.
374 *
375 * pmapping to contain lb_nums as used for disc adressing.
376 */
377 pmapping = ump->la_pmapping;
378 sectors = (buf->b_bcount + sector_size -1) / sector_size;
379 udf_translate_vtop_list(ump, sectors, vpart_num, lmapping, pmapping);
380
381 for (sector = 0; sector < sectors; sector++) {
382 buf_offset = sector * sector_size;
383 DPRINTF(WRITE, ("\tprocessing rel sector %d\n", sector));
384
385 DPRINTF(WRITE, ("\tissue write sector %"PRIu64"\n",
386 pmapping[sector]));
387
388 run_start = pmapping[sector];
389 run_length = 1;
390 while (sector < sectors-1) {
391 if (pmapping[sector+1] != pmapping[sector]+1)
392 break;
393 run_length++;
394 sector++;
395 }
396
397 /* nest an iobuf for the extent */
398 rbuflen = run_length * sector_size;
399 rblk = run_start * (sector_size/DEV_BSIZE);
400
401 nestbuf = getiobuf(NULL, true);
402 nestiobuf_setup(buf, nestbuf, buf_offset, rbuflen);
403 /* nestbuf is B_ASYNC */
404
405 /* identify this nestbuf */
406 nestbuf->b_lblkno = sector;
407 assert(nestbuf->b_vp == buf->b_vp);
408
409 /* CD shedules on raw blkno */
410 nestbuf->b_blkno = rblk;
411 nestbuf->b_proc = NULL;
412 nestbuf->b_rawblkno = rblk;
413 nestbuf->b_udf_c_type = UDF_C_PROCESSED;
414
415 VOP_STRATEGY(ump->devvp, nestbuf);
416 }
417 }
418
419
420 static void
421 udf_discstrat_init_direct(struct udf_strat_args *args)
422 {
423 struct udf_mount *ump = args->ump;
424 struct strat_private *priv = PRIV(ump);
425 uint32_t lb_size;
426
427 KASSERT(priv == NULL);
428 ump->strategy_private = malloc(sizeof(struct strat_private),
429 M_UDFTEMP, M_WAITOK);
430 priv = ump->strategy_private;
431 memset(priv, 0 , sizeof(struct strat_private));
432
433 /*
434 * Initialise pool for descriptors associated with nodes. This is done
435 * in lb_size units though currently lb_size is dictated to be
436 * sector_size.
437 */
438 memset(&priv->desc_pool, 0, sizeof(struct pool));
439
440 lb_size = udf_rw32(ump->logical_vol->lb_size);
441 pool_init(&priv->desc_pool, lb_size, 0, 0, 0, "udf_desc_pool", NULL,
442 IPL_NONE);
443 }
444
445
446 static void
447 udf_discstrat_finish_direct(struct udf_strat_args *args)
448 {
449 struct udf_mount *ump = args->ump;
450 struct strat_private *priv = PRIV(ump);
451
452 /* destroy our pool */
453 pool_destroy(&priv->desc_pool);
454
455 /* free our private space */
456 free(ump->strategy_private, M_UDFTEMP);
457 ump->strategy_private = NULL;
458 }
459
460 /* --------------------------------------------------------------------- */
461
462 struct udf_strategy udf_strat_direct =
463 {
464 udf_getblank_nodedscr_direct,
465 udf_free_nodedscr_direct,
466 udf_read_nodedscr_direct,
467 udf_write_nodedscr_direct,
468 udf_queue_buf_direct,
469 udf_discstrat_init_direct,
470 udf_discstrat_finish_direct
471 };
472
473