?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
drm/omap: Fix error handling path in 'omap_dmm_probe()'
[linux/fpc-iii.git] / drivers / block / xen-blkfront.c
blob2fee2eef988d2992480363ea2578ddb9edf17240
1 /*
2 * blkfront.c
3 *
4 * XenLinux virtual block device driver.
5 *
6 * Copyright (c) 2003-2004, Keir Fraser & Steve Hand
7 * Modifications by Mark A. Williamson are (c) Intel Research Cambridge
8 * Copyright (c) 2004, Christian Limpach
9 * Copyright (c) 2004, Andrew Warfield
10 * Copyright (c) 2005, Christopher Clark
11 * Copyright (c) 2005, XenSource Ltd
12 *
13 * This program is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU General Public License version 2
15 * as published by the Free Software Foundation; or, when distributed
16 * separately from the Linux kernel or incorporated into other
17 * software packages, subject to the following license:
18 *
19 * Permission is hereby granted, free of charge, to any person obtaining a copy
20 * of this source file (the "Software"), to deal in the Software without
21 * restriction, including without limitation the rights to use, copy, modify,
22 * merge, publish, distribute, sublicense, and/or sell copies of the Software,
23 * and to permit persons to whom the Software is furnished to do so, subject to
24 * the following conditions:
25 *
26 * The above copyright notice and this permission notice shall be included in
27 * all copies or substantial portions of the Software.
28 *
29 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
30 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
31 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
32 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
33 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
34 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
35 * IN THE SOFTWARE.
36 */
37
38 #include <linux/interrupt.h>
39 #include <linux/blkdev.h>
40 #include <linux/blk-mq.h>
41 #include <linux/hdreg.h>
42 #include <linux/cdrom.h>
43 #include <linux/module.h>
44 #include <linux/slab.h>
45 #include <linux/mutex.h>
46 #include <linux/scatterlist.h>
47 #include <linux/bitmap.h>
48 #include <linux/list.h>
49
50 #include <xen/xen.h>
51 #include <xen/xenbus.h>
52 #include <xen/grant_table.h>
53 #include <xen/events.h>
54 #include <xen/page.h>
55 #include <xen/platform_pci.h>
56
57 #include <xen/interface/grant_table.h>
58 #include <xen/interface/io/blkif.h>
59 #include <xen/interface/io/protocols.h>
60
61 #include <asm/xen/hypervisor.h>
62
63 enum blkif_state {
64 BLKIF_STATE_DISCONNECTED,
65 BLKIF_STATE_CONNECTED,
66 BLKIF_STATE_SUSPENDED,
67 };
68
69 struct grant {
70 grant_ref_t gref;
71 struct page *page;
72 struct list_head node;
73 };
74
75 struct blk_shadow {
76 struct blkif_request req;
77 struct request *request;
78 struct grant **grants_used;
79 struct grant **indirect_grants;
80 struct scatterlist *sg;
81 unsigned int num_sg;
82 };
83
84 struct split_bio {
85 struct bio *bio;
86 atomic_t pending;
87 };
88
89 static DEFINE_MUTEX(blkfront_mutex);
90 static const struct block_device_operations xlvbd_block_fops;
91
92 /*
93 * Maximum number of segments in indirect requests, the actual value used by
94 * the frontend driver is the minimum of this value and the value provided
95 * by the backend driver.
96 */
97
98 static unsigned int xen_blkif_max_segments = 32;
99 module_param_named(max, xen_blkif_max_segments, int, S_IRUGO);
100 MODULE_PARM_DESC(max, "Maximum amount of segments in indirect requests (default is 32)");
101
102 /*
103 * Maximum order of pages to be used for the shared ring between front and
104 * backend, 4KB page granularity is used.
105 */
106 static unsigned int xen_blkif_max_ring_order;
107 module_param_named(max_ring_page_order, xen_blkif_max_ring_order, int, S_IRUGO);
108 MODULE_PARM_DESC(max_ring_page_order, "Maximum order of pages to be used for the shared ring");
109
110 #define BLK_RING_SIZE(info) \
111 __CONST_RING_SIZE(blkif, XEN_PAGE_SIZE * (info)->nr_ring_pages)
112
113 #define BLK_MAX_RING_SIZE \
114 __CONST_RING_SIZE(blkif, XEN_PAGE_SIZE * XENBUS_MAX_RING_GRANTS)
115
116 /*
117 * ring-ref%i i=(-1UL) would take 11 characters + 'ring-ref' is 8, so 19
118 * characters are enough. Define to 20 to keep consist with backend.
119 */
120 #define RINGREF_NAME_LEN (20)
121
122 /*
123 * We have one of these per vbd, whether ide, scsi or 'other'. They
124 * hang in private_data off the gendisk structure. We may end up
125 * putting all kinds of interesting stuff here :-)
126 */
127 struct blkfront_info
128 {
129 spinlock_t io_lock;
130 struct mutex mutex;
131 struct xenbus_device *xbdev;
132 struct gendisk *gd;
133 int vdevice;
134 blkif_vdev_t handle;
135 enum blkif_state connected;
136 int ring_ref[XENBUS_MAX_RING_GRANTS];
137 unsigned int nr_ring_pages;
138 struct blkif_front_ring ring;
139 unsigned int evtchn, irq;
140 struct request_queue *rq;
141 struct work_struct work;
142 struct gnttab_free_callback callback;
143 struct blk_shadow shadow[BLK_MAX_RING_SIZE];
144 struct list_head grants;
145 struct list_head indirect_pages;
146 unsigned int persistent_gnts_c;
147 unsigned long shadow_free;
148 unsigned int feature_flush;
149 unsigned int feature_discard:1;
150 unsigned int feature_secdiscard:1;
151 unsigned int discard_granularity;
152 unsigned int discard_alignment;
153 unsigned int feature_persistent:1;
154 /* Number of 4KB segments handled */
155 unsigned int max_indirect_segments;
156 int is_ready;
157 struct blk_mq_tag_set tag_set;
158 };
159
160 static unsigned int nr_minors;
161 static unsigned long *minors;
162 static DEFINE_SPINLOCK(minor_lock);
163
164 #define GRANT_INVALID_REF 0
165
166 #define PARTS_PER_DISK 16
167 #define PARTS_PER_EXT_DISK 256
168
169 #define BLKIF_MAJOR(dev) ((dev)>>8)
170 #define BLKIF_MINOR(dev) ((dev) & 0xff)
171
172 #define EXT_SHIFT 28
173 #define EXTENDED (1<<EXT_SHIFT)
174 #define VDEV_IS_EXTENDED(dev) ((dev)&(EXTENDED))
175 #define BLKIF_MINOR_EXT(dev) ((dev)&(~EXTENDED))
176 #define EMULATED_HD_DISK_MINOR_OFFSET (0)
177 #define EMULATED_HD_DISK_NAME_OFFSET (EMULATED_HD_DISK_MINOR_OFFSET / 256)
178 #define EMULATED_SD_DISK_MINOR_OFFSET (0)
179 #define EMULATED_SD_DISK_NAME_OFFSET (EMULATED_SD_DISK_MINOR_OFFSET / 256)
180
181 #define DEV_NAME "xvd" /* name in /dev */
182
183 /*
184 * Grants are always the same size as a Xen page (i.e 4KB).
185 * A physical segment is always the same size as a Linux page.
186 * Number of grants per physical segment
187 */
188 #define GRANTS_PER_PSEG (PAGE_SIZE / XEN_PAGE_SIZE)
189
190 #define GRANTS_PER_INDIRECT_FRAME \
191 (XEN_PAGE_SIZE / sizeof(struct blkif_request_segment))
192
193 #define PSEGS_PER_INDIRECT_FRAME \
194 (GRANTS_INDIRECT_FRAME / GRANTS_PSEGS)
195
196 #define INDIRECT_GREFS(_grants) \
197 DIV_ROUND_UP(_grants, GRANTS_PER_INDIRECT_FRAME)
198
199 #define GREFS(_psegs) ((_psegs) * GRANTS_PER_PSEG)
200
201 static int blkfront_setup_indirect(struct blkfront_info *info);
202 static int blkfront_gather_backend_features(struct blkfront_info *info);
203
204 static int get_id_from_freelist(struct blkfront_info *info)
205 {
206 unsigned long free = info->shadow_free;
207 BUG_ON(free >= BLK_RING_SIZE(info));
208 info->shadow_free = info->shadow[free].req.u.rw.id;
209 info->shadow[free].req.u.rw.id = 0x0fffffee; /* debug */
210 return free;
211 }
212
213 static int add_id_to_freelist(struct blkfront_info *info,
214 unsigned long id)
215 {
216 if (info->shadow[id].req.u.rw.id != id)
217 return -EINVAL;
218 if (info->shadow[id].request == NULL)
219 return -EINVAL;
220 info->shadow[id].req.u.rw.id = info->shadow_free;
221 info->shadow[id].request = NULL;
222 info->shadow_free = id;
223 return 0;
224 }
225
226 static int fill_grant_buffer(struct blkfront_info *info, int num)
227 {
228 struct page *granted_page;
229 struct grant *gnt_list_entry, *n;
230 int i = 0;
231
232 while(i < num) {
233 gnt_list_entry = kzalloc(sizeof(struct grant), GFP_NOIO);
234 if (!gnt_list_entry)
235 goto out_of_memory;
236
237 if (info->feature_persistent) {
238 granted_page = alloc_page(GFP_NOIO);
239 if (!granted_page) {
240 kfree(gnt_list_entry);
241 goto out_of_memory;
242 }
243 gnt_list_entry->page = granted_page;
244 }
245
246 gnt_list_entry->gref = GRANT_INVALID_REF;
247 list_add(&gnt_list_entry->node, &info->grants);
248 i++;
249 }
250
251 return 0;
252
253 out_of_memory:
254 list_for_each_entry_safe(gnt_list_entry, n,
255 &info->grants, node) {
256 list_del(&gnt_list_entry->node);
257 if (info->feature_persistent)
258 __free_page(gnt_list_entry->page);
259 kfree(gnt_list_entry);
260 i--;
261 }
262 BUG_ON(i != 0);
263 return -ENOMEM;
264 }
265
266 static struct grant *get_free_grant(struct blkfront_info *info)
267 {
268 struct grant *gnt_list_entry;
269
270 BUG_ON(list_empty(&info->grants));
271 gnt_list_entry = list_first_entry(&info->grants, struct grant,
272 node);
273 list_del(&gnt_list_entry->node);
274
275 if (gnt_list_entry->gref != GRANT_INVALID_REF)
276 info->persistent_gnts_c--;
277
278 return gnt_list_entry;
279 }
280
281 static inline void grant_foreign_access(const struct grant *gnt_list_entry,
282 const struct blkfront_info *info)
283 {
284 gnttab_page_grant_foreign_access_ref_one(gnt_list_entry->gref,
285 info->xbdev->otherend_id,
286 gnt_list_entry->page,
287 0);
288 }
289
290 static struct grant *get_grant(grant_ref_t *gref_head,
291 unsigned long gfn,
292 struct blkfront_info *info)
293 {
294 struct grant *gnt_list_entry = get_free_grant(info);
295
296 if (gnt_list_entry->gref != GRANT_INVALID_REF)
297 return gnt_list_entry;
298
299 /* Assign a gref to this page */
300 gnt_list_entry->gref = gnttab_claim_grant_reference(gref_head);
301 BUG_ON(gnt_list_entry->gref == -ENOSPC);
302 if (info->feature_persistent)
303 grant_foreign_access(gnt_list_entry, info);
304 else {
305 /* Grant access to the GFN passed by the caller */
306 gnttab_grant_foreign_access_ref(gnt_list_entry->gref,
307 info->xbdev->otherend_id,
308 gfn, 0);
309 }
310
311 return gnt_list_entry;
312 }
313
314 static struct grant *get_indirect_grant(grant_ref_t *gref_head,
315 struct blkfront_info *info)
316 {
317 struct grant *gnt_list_entry = get_free_grant(info);
318
319 if (gnt_list_entry->gref != GRANT_INVALID_REF)
320 return gnt_list_entry;
321
322 /* Assign a gref to this page */
323 gnt_list_entry->gref = gnttab_claim_grant_reference(gref_head);
324 BUG_ON(gnt_list_entry->gref == -ENOSPC);
325 if (!info->feature_persistent) {
326 struct page *indirect_page;
327
328 /* Fetch a pre-allocated page to use for indirect grefs */
329 BUG_ON(list_empty(&info->indirect_pages));
330 indirect_page = list_first_entry(&info->indirect_pages,
331 struct page, lru);
332 list_del(&indirect_page->lru);
333 gnt_list_entry->page = indirect_page;
334 }
335 grant_foreign_access(gnt_list_entry, info);
336
337 return gnt_list_entry;
338 }
339
340 static const char *op_name(int op)
341 {
342 static const char *const names[] = {
343 [BLKIF_OP_READ] = "read",
344 [BLKIF_OP_WRITE] = "write",
345 [BLKIF_OP_WRITE_BARRIER] = "barrier",
346 [BLKIF_OP_FLUSH_DISKCACHE] = "flush",
347 [BLKIF_OP_DISCARD] = "discard" };
348
349 if (op < 0 || op >= ARRAY_SIZE(names))
350 return "unknown";
351
352 if (!names[op])
353 return "reserved";
354
355 return names[op];
356 }
357 static int xlbd_reserve_minors(unsigned int minor, unsigned int nr)
358 {
359 unsigned int end = minor + nr;
360 int rc;
361
362 if (end > nr_minors) {
363 unsigned long *bitmap, *old;
364
365 bitmap = kcalloc(BITS_TO_LONGS(end), sizeof(*bitmap),
366 GFP_KERNEL);
367 if (bitmap == NULL)
368 return -ENOMEM;
369
370 spin_lock(&minor_lock);
371 if (end > nr_minors) {
372 old = minors;
373 memcpy(bitmap, minors,
374 BITS_TO_LONGS(nr_minors) * sizeof(*bitmap));
375 minors = bitmap;
376 nr_minors = BITS_TO_LONGS(end) * BITS_PER_LONG;
377 } else
378 old = bitmap;
379 spin_unlock(&minor_lock);
380 kfree(old);
381 }
382
383 spin_lock(&minor_lock);
384 if (find_next_bit(minors, end, minor) >= end) {
385 bitmap_set(minors, minor, nr);
386 rc = 0;
387 } else
388 rc = -EBUSY;
389 spin_unlock(&minor_lock);
390
391 return rc;
392 }
393
394 static void xlbd_release_minors(unsigned int minor, unsigned int nr)
395 {
396 unsigned int end = minor + nr;
397
398 BUG_ON(end > nr_minors);
399 spin_lock(&minor_lock);
400 bitmap_clear(minors, minor, nr);
401 spin_unlock(&minor_lock);
402 }
403
404 static void blkif_restart_queue_callback(void *arg)
405 {
406 struct blkfront_info *info = (struct blkfront_info *)arg;
407 schedule_work(&info->work);
408 }
409
410 static int blkif_getgeo(struct block_device *bd, struct hd_geometry *hg)
411 {
412 /* We don't have real geometry info, but let's at least return
413 values consistent with the size of the device */
414 sector_t nsect = get_capacity(bd->bd_disk);
415 sector_t cylinders = nsect;
416
417 hg->heads = 0xff;
418 hg->sectors = 0x3f;
419 sector_div(cylinders, hg->heads * hg->sectors);
420 hg->cylinders = cylinders;
421 if ((sector_t)(hg->cylinders + 1) * hg->heads * hg->sectors < nsect)
422 hg->cylinders = 0xffff;
423 return 0;
424 }
425
426 static int blkif_ioctl(struct block_device *bdev, fmode_t mode,
427 unsigned command, unsigned long argument)
428 {
429 struct blkfront_info *info = bdev->bd_disk->private_data;
430 int i;
431
432 dev_dbg(&info->xbdev->dev, "command: 0x%x, argument: 0x%lx\n",
433 command, (long)argument);
434
435 switch (command) {
436 case CDROMMULTISESSION:
437 dev_dbg(&info->xbdev->dev, "FIXME: support multisession CDs later\n");
438 for (i = 0; i < sizeof(struct cdrom_multisession); i++)
439 if (put_user(0, (char __user *)(argument + i)))
440 return -EFAULT;
441 return 0;
442
443 case CDROM_GET_CAPABILITY: {
444 struct gendisk *gd = info->gd;
445 if (gd->flags & GENHD_FL_CD)
446 return 0;
447 return -EINVAL;
448 }
449
450 default:
451 /*printk(KERN_ALERT "ioctl %08x not supported by Xen blkdev\n",
452 command);*/
453 return -EINVAL; /* same return as native Linux */
454 }
455
456 return 0;
457 }
458
459 static int blkif_queue_discard_req(struct request *req)
460 {
461 struct blkfront_info *info = req->rq_disk->private_data;
462 struct blkif_request *ring_req;
463 unsigned long id;
464
465 /* Fill out a communications ring structure. */
466 ring_req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt);
467 id = get_id_from_freelist(info);
468 info->shadow[id].request = req;
469
470 ring_req->operation = BLKIF_OP_DISCARD;
471 ring_req->u.discard.nr_sectors = blk_rq_sectors(req);
472 ring_req->u.discard.id = id;
473 ring_req->u.discard.sector_number = (blkif_sector_t)blk_rq_pos(req);
474 if ((req->cmd_flags & REQ_SECURE) && info->feature_secdiscard)
475 ring_req->u.discard.flag = BLKIF_DISCARD_SECURE;
476 else
477 ring_req->u.discard.flag = 0;
478
479 info->ring.req_prod_pvt++;
480
481 /* Keep a private copy so we can reissue requests when recovering. */
482 info->shadow[id].req = *ring_req;
483
484 return 0;
485 }
486
487 struct setup_rw_req {
488 unsigned int grant_idx;
489 struct blkif_request_segment *segments;
490 struct blkfront_info *info;
491 struct blkif_request *ring_req;
492 grant_ref_t gref_head;
493 unsigned int id;
494 /* Only used when persistent grant is used and it's a read request */
495 bool need_copy;
496 unsigned int bvec_off;
497 char *bvec_data;
498 };
499
500 static void blkif_setup_rw_req_grant(unsigned long gfn, unsigned int offset,
501 unsigned int len, void *data)
502 {
503 struct setup_rw_req *setup = data;
504 int n, ref;
505 struct grant *gnt_list_entry;
506 unsigned int fsect, lsect;
507 /* Convenient aliases */
508 unsigned int grant_idx = setup->grant_idx;
509 struct blkif_request *ring_req = setup->ring_req;
510 struct blkfront_info *info = setup->info;
511 struct blk_shadow *shadow = &info->shadow[setup->id];
512
513 if ((ring_req->operation == BLKIF_OP_INDIRECT) &&
514 (grant_idx % GRANTS_PER_INDIRECT_FRAME == 0)) {
515 if (setup->segments)
516 kunmap_atomic(setup->segments);
517
518 n = grant_idx / GRANTS_PER_INDIRECT_FRAME;
519 gnt_list_entry = get_indirect_grant(&setup->gref_head, info);
520 shadow->indirect_grants[n] = gnt_list_entry;
521 setup->segments = kmap_atomic(gnt_list_entry->page);
522 ring_req->u.indirect.indirect_grefs[n] = gnt_list_entry->gref;
523 }
524
525 gnt_list_entry = get_grant(&setup->gref_head, gfn, info);
526 ref = gnt_list_entry->gref;
527 shadow->grants_used[grant_idx] = gnt_list_entry;
528
529 if (setup->need_copy) {
530 void *shared_data;
531
532 shared_data = kmap_atomic(gnt_list_entry->page);
533 /*
534 * this does not wipe data stored outside the
535 * range sg->offset..sg->offset+sg->length.
536 * Therefore, blkback *could* see data from
537 * previous requests. This is OK as long as
538 * persistent grants are shared with just one
539 * domain. It may need refactoring if this
540 * changes
541 */
542 memcpy(shared_data + offset,
543 setup->bvec_data + setup->bvec_off,
544 len);
545
546 kunmap_atomic(shared_data);
547 setup->bvec_off += len;
548 }
549
550 fsect = offset >> 9;
551 lsect = fsect + (len >> 9) - 1;
552 if (ring_req->operation != BLKIF_OP_INDIRECT) {
553 ring_req->u.rw.seg[grant_idx] =
554 (struct blkif_request_segment) {
555 .gref = ref,
556 .first_sect = fsect,
557 .last_sect = lsect };
558 } else {
559 setup->segments[grant_idx % GRANTS_PER_INDIRECT_FRAME] =
560 (struct blkif_request_segment) {
561 .gref = ref,
562 .first_sect = fsect,
563 .last_sect = lsect };
564 }
565
566 (setup->grant_idx)++;
567 }
568
569 static int blkif_queue_rw_req(struct request *req)
570 {
571 struct blkfront_info *info = req->rq_disk->private_data;
572 struct blkif_request *ring_req;
573 unsigned long id;
574 int i;
575 struct setup_rw_req setup = {
576 .grant_idx = 0,
577 .segments = NULL,
578 .info = info,
579 .need_copy = rq_data_dir(req) && info->feature_persistent,
580 };
581
582 /*
583 * Used to store if we are able to queue the request by just using
584 * existing persistent grants, or if we have to get new grants,
585 * as there are not sufficiently many free.
586 */
587 bool new_persistent_gnts;
588 struct scatterlist *sg;
589 int num_sg, max_grefs, num_grant;
590
591 max_grefs = req->nr_phys_segments * GRANTS_PER_PSEG;
592 if (max_grefs > BLKIF_MAX_SEGMENTS_PER_REQUEST)
593 /*
594 * If we are using indirect segments we need to account
595 * for the indirect grefs used in the request.
596 */
597 max_grefs += INDIRECT_GREFS(max_grefs);
598
599 /* Check if we have enough grants to allocate a requests */
600 if (info->persistent_gnts_c < max_grefs) {
601 new_persistent_gnts = 1;
602 if (gnttab_alloc_grant_references(
603 max_grefs - info->persistent_gnts_c,
604 &setup.gref_head) < 0) {
605 gnttab_request_free_callback(
606 &info->callback,
607 blkif_restart_queue_callback,
608 info,
609 max_grefs);
610 return 1;
611 }
612 } else
613 new_persistent_gnts = 0;
614
615 /* Fill out a communications ring structure. */
616 ring_req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt);
617 id = get_id_from_freelist(info);
618 info->shadow[id].request = req;
619
620 BUG_ON(info->max_indirect_segments == 0 &&
621 GREFS(req->nr_phys_segments) > BLKIF_MAX_SEGMENTS_PER_REQUEST);
622 BUG_ON(info->max_indirect_segments &&
623 GREFS(req->nr_phys_segments) > info->max_indirect_segments);
624
625 num_sg = blk_rq_map_sg(req->q, req, info->shadow[id].sg);
626 num_grant = 0;
627 /* Calculate the number of grant used */
628 for_each_sg(info->shadow[id].sg, sg, num_sg, i)
629 num_grant += gnttab_count_grant(sg->offset, sg->length);
630
631 ring_req->u.rw.id = id;
632 info->shadow[id].num_sg = num_sg;
633 if (num_grant > BLKIF_MAX_SEGMENTS_PER_REQUEST) {
634 /*
635 * The indirect operation can only be a BLKIF_OP_READ or
636 * BLKIF_OP_WRITE
637 */
638 BUG_ON(req->cmd_flags & (REQ_FLUSH | REQ_FUA));
639 ring_req->operation = BLKIF_OP_INDIRECT;
640 ring_req->u.indirect.indirect_op = rq_data_dir(req) ?
641 BLKIF_OP_WRITE : BLKIF_OP_READ;
642 ring_req->u.indirect.sector_number = (blkif_sector_t)blk_rq_pos(req);
643 ring_req->u.indirect.handle = info->handle;
644 ring_req->u.indirect.nr_segments = num_grant;
645 } else {
646 ring_req->u.rw.sector_number = (blkif_sector_t)blk_rq_pos(req);
647 ring_req->u.rw.handle = info->handle;
648 ring_req->operation = rq_data_dir(req) ?
649 BLKIF_OP_WRITE : BLKIF_OP_READ;
650 if (req->cmd_flags & (REQ_FLUSH | REQ_FUA)) {
651 /*
652 * Ideally we can do an unordered flush-to-disk.
653 * In case the backend onlysupports barriers, use that.
654 * A barrier request a superset of FUA, so we can
655 * implement it the same way. (It's also a FLUSH+FUA,
656 * since it is guaranteed ordered WRT previous writes.)
657 */
658 switch (info->feature_flush &
659 ((REQ_FLUSH|REQ_FUA))) {
660 case REQ_FLUSH|REQ_FUA:
661 ring_req->operation =
662 BLKIF_OP_WRITE_BARRIER;
663 break;
664 case REQ_FLUSH:
665 ring_req->operation =
666 BLKIF_OP_FLUSH_DISKCACHE;
667 break;
668 default:
669 ring_req->operation = 0;
670 }
671 }
672 ring_req->u.rw.nr_segments = num_grant;
673 }
674
675 setup.ring_req = ring_req;
676 setup.id = id;
677 for_each_sg(info->shadow[id].sg, sg, num_sg, i) {
678 BUG_ON(sg->offset + sg->length > PAGE_SIZE);
679
680 if (setup.need_copy) {
681 setup.bvec_off = sg->offset;
682 setup.bvec_data = kmap_atomic(sg_page(sg));
683 }
684
685 gnttab_foreach_grant_in_range(sg_page(sg),
686 sg->offset,
687 sg->length,
688 blkif_setup_rw_req_grant,
689 &setup);
690
691 if (setup.need_copy)
692 kunmap_atomic(setup.bvec_data);
693 }
694 if (setup.segments)
695 kunmap_atomic(setup.segments);
696
697 info->ring.req_prod_pvt++;
698
699 /* Keep a private copy so we can reissue requests when recovering. */
700 info->shadow[id].req = *ring_req;
701
702 if (new_persistent_gnts)
703 gnttab_free_grant_references(setup.gref_head);
704
705 return 0;
706 }
707
708 /*
709 * Generate a Xen blkfront IO request from a blk layer request. Reads
710 * and writes are handled as expected.
711 *
712 * @req: a request struct
713 */
714 static int blkif_queue_request(struct request *req)
715 {
716 struct blkfront_info *info = req->rq_disk->private_data;
717
718 if (unlikely(info->connected != BLKIF_STATE_CONNECTED))
719 return 1;
720
721 if (unlikely(req->cmd_flags & (REQ_DISCARD | REQ_SECURE)))
722 return blkif_queue_discard_req(req);
723 else
724 return blkif_queue_rw_req(req);
725 }
726
727 static inline void flush_requests(struct blkfront_info *info)
728 {
729 int notify;
730
731 RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&info->ring, notify);
732
733 if (notify)
734 notify_remote_via_irq(info->irq);
735 }
736
737 static inline bool blkif_request_flush_invalid(struct request *req,
738 struct blkfront_info *info)
739 {
740 return ((req->cmd_type != REQ_TYPE_FS) ||
741 ((req->cmd_flags & REQ_FLUSH) &&
742 !(info->feature_flush & REQ_FLUSH)) ||
743 ((req->cmd_flags & REQ_FUA) &&
744 !(info->feature_flush & REQ_FUA)));
745 }
746
747 static int blkif_queue_rq(struct blk_mq_hw_ctx *hctx,
748 const struct blk_mq_queue_data *qd)
749 {
750 struct blkfront_info *info = qd->rq->rq_disk->private_data;
751
752 blk_mq_start_request(qd->rq);
753 spin_lock_irq(&info->io_lock);
754 if (RING_FULL(&info->ring))
755 goto out_busy;
756
757 if (blkif_request_flush_invalid(qd->rq, info))
758 goto out_err;
759
760 if (blkif_queue_request(qd->rq))
761 goto out_busy;
762
763 flush_requests(info);
764 spin_unlock_irq(&info->io_lock);
765 return BLK_MQ_RQ_QUEUE_OK;
766
767 out_err:
768 spin_unlock_irq(&info->io_lock);
769 return BLK_MQ_RQ_QUEUE_ERROR;
770
771 out_busy:
772 spin_unlock_irq(&info->io_lock);
773 blk_mq_stop_hw_queue(hctx);
774 return BLK_MQ_RQ_QUEUE_BUSY;
775 }
776
777 static struct blk_mq_ops blkfront_mq_ops = {
778 .queue_rq = blkif_queue_rq,
779 .map_queue = blk_mq_map_queue,
780 };
781
782 static int xlvbd_init_blk_queue(struct gendisk *gd, u16 sector_size,
783 unsigned int physical_sector_size,
784 unsigned int segments)
785 {
786 struct request_queue *rq;
787 struct blkfront_info *info = gd->private_data;
788
789 memset(&info->tag_set, 0, sizeof(info->tag_set));
790 info->tag_set.ops = &blkfront_mq_ops;
791 info->tag_set.nr_hw_queues = 1;
792 info->tag_set.queue_depth = BLK_RING_SIZE(info);
793 info->tag_set.numa_node = NUMA_NO_NODE;
794 info->tag_set.flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_SG_MERGE;
795 info->tag_set.cmd_size = 0;
796 info->tag_set.driver_data = info;
797
798 if (blk_mq_alloc_tag_set(&info->tag_set))
799 return -1;
800 rq = blk_mq_init_queue(&info->tag_set);
801 if (IS_ERR(rq)) {
802 blk_mq_free_tag_set(&info->tag_set);
803 return -1;
804 }
805
806 queue_flag_set_unlocked(QUEUE_FLAG_VIRT, rq);
807
808 if (info->feature_discard) {
809 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, rq);
810 blk_queue_max_discard_sectors(rq, get_capacity(gd));
811 rq->limits.discard_granularity = info->discard_granularity;
812 rq->limits.discard_alignment = info->discard_alignment;
813 if (info->feature_secdiscard)
814 queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD, rq);
815 }
816
817 /* Hard sector size and max sectors impersonate the equiv. hardware. */
818 blk_queue_logical_block_size(rq, sector_size);
819 blk_queue_physical_block_size(rq, physical_sector_size);
820 blk_queue_max_hw_sectors(rq, (segments * XEN_PAGE_SIZE) / 512);
821
822 /* Each segment in a request is up to an aligned page in size. */
823 blk_queue_segment_boundary(rq, PAGE_SIZE - 1);
824 blk_queue_max_segment_size(rq, PAGE_SIZE);
825
826 /* Ensure a merged request will fit in a single I/O ring slot. */
827 blk_queue_max_segments(rq, segments / GRANTS_PER_PSEG);
828
829 /* Make sure buffer addresses are sector-aligned. */
830 blk_queue_dma_alignment(rq, 511);
831
832 /* Make sure we don't use bounce buffers. */
833 blk_queue_bounce_limit(rq, BLK_BOUNCE_ANY);
834
835 gd->queue = rq;
836
837 return 0;
838 }
839
840 static const char *flush_info(unsigned int feature_flush)
841 {
842 switch (feature_flush & ((REQ_FLUSH | REQ_FUA))) {
843 case REQ_FLUSH|REQ_FUA:
844 return "barrier: enabled;";
845 case REQ_FLUSH:
846 return "flush diskcache: enabled;";
847 default:
848 return "barrier or flush: disabled;";
849 }
850 }
851
852 static void xlvbd_flush(struct blkfront_info *info)
853 {
854 blk_queue_flush(info->rq, info->feature_flush);
855 pr_info("blkfront: %s: %s %s %s %s %s\n",
856 info->gd->disk_name, flush_info(info->feature_flush),
857 "persistent grants:", info->feature_persistent ?
858 "enabled;" : "disabled;", "indirect descriptors:",
859 info->max_indirect_segments ? "enabled;" : "disabled;");
860 }
861
862 static int xen_translate_vdev(int vdevice, int *minor, unsigned int *offset)
863 {
864 int major;
865 major = BLKIF_MAJOR(vdevice);
866 *minor = BLKIF_MINOR(vdevice);
867 switch (major) {
868 case XEN_IDE0_MAJOR:
869 *offset = (*minor / 64) + EMULATED_HD_DISK_NAME_OFFSET;
870 *minor = ((*minor / 64) * PARTS_PER_DISK) +
871 EMULATED_HD_DISK_MINOR_OFFSET;
872 break;
873 case XEN_IDE1_MAJOR:
874 *offset = (*minor / 64) + 2 + EMULATED_HD_DISK_NAME_OFFSET;
875 *minor = (((*minor / 64) + 2) * PARTS_PER_DISK) +
876 EMULATED_HD_DISK_MINOR_OFFSET;
877 break;
878 case XEN_SCSI_DISK0_MAJOR:
879 *offset = (*minor / PARTS_PER_DISK) + EMULATED_SD_DISK_NAME_OFFSET;
880 *minor = *minor + EMULATED_SD_DISK_MINOR_OFFSET;
881 break;
882 case XEN_SCSI_DISK1_MAJOR:
883 case XEN_SCSI_DISK2_MAJOR:
884 case XEN_SCSI_DISK3_MAJOR:
885 case XEN_SCSI_DISK4_MAJOR:
886 case XEN_SCSI_DISK5_MAJOR:
887 case XEN_SCSI_DISK6_MAJOR:
888 case XEN_SCSI_DISK7_MAJOR:
889 *offset = (*minor / PARTS_PER_DISK) +
890 ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16) +
891 EMULATED_SD_DISK_NAME_OFFSET;
892 *minor = *minor +
893 ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16 * PARTS_PER_DISK) +
894 EMULATED_SD_DISK_MINOR_OFFSET;
895 break;
896 case XEN_SCSI_DISK8_MAJOR:
897 case XEN_SCSI_DISK9_MAJOR:
898 case XEN_SCSI_DISK10_MAJOR:
899 case XEN_SCSI_DISK11_MAJOR:
900 case XEN_SCSI_DISK12_MAJOR:
901 case XEN_SCSI_DISK13_MAJOR:
902 case XEN_SCSI_DISK14_MAJOR:
903 case XEN_SCSI_DISK15_MAJOR:
904 *offset = (*minor / PARTS_PER_DISK) +
905 ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16) +
906 EMULATED_SD_DISK_NAME_OFFSET;
907 *minor = *minor +
908 ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16 * PARTS_PER_DISK) +
909 EMULATED_SD_DISK_MINOR_OFFSET;
910 break;
911 case XENVBD_MAJOR:
912 *offset = *minor / PARTS_PER_DISK;
913 break;
914 default:
915 printk(KERN_WARNING "blkfront: your disk configuration is "
916 "incorrect, please use an xvd device instead\n");
917 return -ENODEV;
918 }
919 return 0;
920 }
921
922 static char *encode_disk_name(char *ptr, unsigned int n)
923 {
924 if (n >= 26)
925 ptr = encode_disk_name(ptr, n / 26 - 1);
926 *ptr = 'a' + n % 26;
927 return ptr + 1;
928 }
929
930 static int xlvbd_alloc_gendisk(blkif_sector_t capacity,
931 struct blkfront_info *info,
932 u16 vdisk_info, u16 sector_size,
933 unsigned int physical_sector_size)
934 {
935 struct gendisk *gd;
936 int nr_minors = 1;
937 int err;
938 unsigned int offset;
939 int minor;
940 int nr_parts;
941 char *ptr;
942
943 BUG_ON(info->gd != NULL);
944 BUG_ON(info->rq != NULL);
945
946 if ((info->vdevice>>EXT_SHIFT) > 1) {
947 /* this is above the extended range; something is wrong */
948 printk(KERN_WARNING "blkfront: vdevice 0x%x is above the extended range; ignoring\n", info->vdevice);
949 return -ENODEV;
950 }
951
952 if (!VDEV_IS_EXTENDED(info->vdevice)) {
953 err = xen_translate_vdev(info->vdevice, &minor, &offset);
954 if (err)
955 return err;
956 nr_parts = PARTS_PER_DISK;
957 } else {
958 minor = BLKIF_MINOR_EXT(info->vdevice);
959 nr_parts = PARTS_PER_EXT_DISK;
960 offset = minor / nr_parts;
961 if (xen_hvm_domain() && offset < EMULATED_HD_DISK_NAME_OFFSET + 4)
962 printk(KERN_WARNING "blkfront: vdevice 0x%x might conflict with "
963 "emulated IDE disks,\n\t choose an xvd device name"
964 "from xvde on\n", info->vdevice);
965 }
966 if (minor >> MINORBITS) {
967 pr_warn("blkfront: %#x's minor (%#x) out of range; ignoring\n",
968 info->vdevice, minor);
969 return -ENODEV;
970 }
971
972 if ((minor % nr_parts) == 0)
973 nr_minors = nr_parts;
974
975 err = xlbd_reserve_minors(minor, nr_minors);
976 if (err)
977 goto out;
978 err = -ENODEV;
979
980 gd = alloc_disk(nr_minors);
981 if (gd == NULL)
982 goto release;
983
984 strcpy(gd->disk_name, DEV_NAME);
985 ptr = encode_disk_name(gd->disk_name + sizeof(DEV_NAME) - 1, offset);
986 BUG_ON(ptr >= gd->disk_name + DISK_NAME_LEN);
987 if (nr_minors > 1)
988 *ptr = 0;
989 else
990 snprintf(ptr, gd->disk_name + DISK_NAME_LEN - ptr,
991 "%d", minor & (nr_parts - 1));
992
993 gd->major = XENVBD_MAJOR;
994 gd->first_minor = minor;
995 gd->fops = &xlvbd_block_fops;
996 gd->private_data = info;
997 gd->driverfs_dev = &(info->xbdev->dev);
998 set_capacity(gd, capacity);
999
1000 if (xlvbd_init_blk_queue(gd, sector_size, physical_sector_size,
1001 info->max_indirect_segments ? :
1002 BLKIF_MAX_SEGMENTS_PER_REQUEST)) {
1003 del_gendisk(gd);
1004 goto release;
1005 }
1006
1007 info->rq = gd->queue;
1008 info->gd = gd;
1009
1010 xlvbd_flush(info);
1011
1012 if (vdisk_info & VDISK_READONLY)
1013 set_disk_ro(gd, 1);
1014
1015 if (vdisk_info & VDISK_REMOVABLE)
1016 gd->flags |= GENHD_FL_REMOVABLE;
1017
1018 if (vdisk_info & VDISK_CDROM)
1019 gd->flags |= GENHD_FL_CD;
1020
1021 return 0;
1022
1023 release:
1024 xlbd_release_minors(minor, nr_minors);
1025 out:
1026 return err;
1027 }
1028
1029 static void xlvbd_release_gendisk(struct blkfront_info *info)
1030 {
1031 unsigned int minor, nr_minors;
1032
1033 if (info->rq == NULL)
1034 return;
1035
1036 /* No more blkif_request(). */
1037 blk_mq_stop_hw_queues(info->rq);
1038
1039 /* No more gnttab callback work. */
1040 gnttab_cancel_free_callback(&info->callback);
1041
1042 /* Flush gnttab callback work. Must be done with no locks held. */
1043 flush_work(&info->work);
1044
1045 del_gendisk(info->gd);
1046
1047 minor = info->gd->first_minor;
1048 nr_minors = info->gd->minors;
1049 xlbd_release_minors(minor, nr_minors);
1050
1051 blk_cleanup_queue(info->rq);
1052 blk_mq_free_tag_set(&info->tag_set);
1053 info->rq = NULL;
1054
1055 put_disk(info->gd);
1056 info->gd = NULL;
1057 }
1058
1059 /* Must be called with io_lock holded */
1060 static void kick_pending_request_queues(struct blkfront_info *info)
1061 {
1062 if (!RING_FULL(&info->ring))
1063 blk_mq_start_stopped_hw_queues(info->rq, true);
1064 }
1065
1066 static void blkif_restart_queue(struct work_struct *work)
1067 {
1068 struct blkfront_info *info = container_of(work, struct blkfront_info, work);
1069
1070 spin_lock_irq(&info->io_lock);
1071 if (info->connected == BLKIF_STATE_CONNECTED)
1072 kick_pending_request_queues(info);
1073 spin_unlock_irq(&info->io_lock);
1074 }
1075
1076 static void blkif_free(struct blkfront_info *info, int suspend)
1077 {
1078 struct grant *persistent_gnt;
1079 struct grant *n;
1080 int i, j, segs;
1081
1082 /* Prevent new requests being issued until we fix things up. */
1083 spin_lock_irq(&info->io_lock);
1084 info->connected = suspend ?
1085 BLKIF_STATE_SUSPENDED : BLKIF_STATE_DISCONNECTED;
1086 /* No more blkif_request(). */
1087 if (info->rq)
1088 blk_mq_stop_hw_queues(info->rq);
1089
1090 /* Remove all persistent grants */
1091 if (!list_empty(&info->grants)) {
1092 list_for_each_entry_safe(persistent_gnt, n,
1093 &info->grants, node) {
1094 list_del(&persistent_gnt->node);
1095 if (persistent_gnt->gref != GRANT_INVALID_REF) {
1096 gnttab_end_foreign_access(persistent_gnt->gref,
1097 0, 0UL);
1098 info->persistent_gnts_c--;
1099 }
1100 if (info->feature_persistent)
1101 __free_page(persistent_gnt->page);
1102 kfree(persistent_gnt);
1103 }
1104 }
1105 BUG_ON(info->persistent_gnts_c != 0);
1106
1107 /*
1108 * Remove indirect pages, this only happens when using indirect
1109 * descriptors but not persistent grants
1110 */
1111 if (!list_empty(&info->indirect_pages)) {
1112 struct page *indirect_page, *n;
1113
1114 BUG_ON(info->feature_persistent);
1115 list_for_each_entry_safe(indirect_page, n, &info->indirect_pages, lru) {
1116 list_del(&indirect_page->lru);
1117 __free_page(indirect_page);
1118 }
1119 }
1120
1121 for (i = 0; i < BLK_RING_SIZE(info); i++) {
1122 /*
1123 * Clear persistent grants present in requests already
1124 * on the shared ring
1125 */
1126 if (!info->shadow[i].request)
1127 goto free_shadow;
1128
1129 segs = info->shadow[i].req.operation == BLKIF_OP_INDIRECT ?
1130 info->shadow[i].req.u.indirect.nr_segments :
1131 info->shadow[i].req.u.rw.nr_segments;
1132 for (j = 0; j < segs; j++) {
1133 persistent_gnt = info->shadow[i].grants_used[j];
1134 gnttab_end_foreign_access(persistent_gnt->gref, 0, 0UL);
1135 if (info->feature_persistent)
1136 __free_page(persistent_gnt->page);
1137 kfree(persistent_gnt);
1138 }
1139
1140 if (info->shadow[i].req.operation != BLKIF_OP_INDIRECT)
1141 /*
1142 * If this is not an indirect operation don't try to
1143 * free indirect segments
1144 */
1145 goto free_shadow;
1146
1147 for (j = 0; j < INDIRECT_GREFS(segs); j++) {
1148 persistent_gnt = info->shadow[i].indirect_grants[j];
1149 gnttab_end_foreign_access(persistent_gnt->gref, 0, 0UL);
1150 __free_page(persistent_gnt->page);
1151 kfree(persistent_gnt);
1152 }
1153
1154 free_shadow:
1155 kfree(info->shadow[i].grants_used);
1156 info->shadow[i].grants_used = NULL;
1157 kfree(info->shadow[i].indirect_grants);
1158 info->shadow[i].indirect_grants = NULL;
1159 kfree(info->shadow[i].sg);
1160 info->shadow[i].sg = NULL;
1161 }
1162
1163 /* No more gnttab callback work. */
1164 gnttab_cancel_free_callback(&info->callback);
1165 spin_unlock_irq(&info->io_lock);
1166
1167 /* Flush gnttab callback work. Must be done with no locks held. */
1168 flush_work(&info->work);
1169
1170 /* Free resources associated with old device channel. */
1171 for (i = 0; i < info->nr_ring_pages; i++) {
1172 if (info->ring_ref[i] != GRANT_INVALID_REF) {
1173 gnttab_end_foreign_access(info->ring_ref[i], 0, 0);
1174 info->ring_ref[i] = GRANT_INVALID_REF;
1175 }
1176 }
1177 free_pages((unsigned long)info->ring.sring, get_order(info->nr_ring_pages * PAGE_SIZE));
1178 info->ring.sring = NULL;
1179
1180 if (info->irq)
1181 unbind_from_irqhandler(info->irq, info);
1182 info->evtchn = info->irq = 0;
1183
1184 }
1185
1186 struct copy_from_grant {
1187 const struct blk_shadow *s;
1188 unsigned int grant_idx;
1189 unsigned int bvec_offset;
1190 char *bvec_data;
1191 };
1192
1193 static void blkif_copy_from_grant(unsigned long gfn, unsigned int offset,
1194 unsigned int len, void *data)
1195 {
1196 struct copy_from_grant *info = data;
1197 char *shared_data;
1198 /* Convenient aliases */
1199 const struct blk_shadow *s = info->s;
1200
1201 shared_data = kmap_atomic(s->grants_used[info->grant_idx]->page);
1202
1203 memcpy(info->bvec_data + info->bvec_offset,
1204 shared_data + offset, len);
1205
1206 info->bvec_offset += len;
1207 info->grant_idx++;
1208
1209 kunmap_atomic(shared_data);
1210 }
1211
1212 static void blkif_completion(struct blk_shadow *s, struct blkfront_info *info,
1213 struct blkif_response *bret)
1214 {
1215 int i = 0;
1216 struct scatterlist *sg;
1217 int num_sg, num_grant;
1218 struct copy_from_grant data = {
1219 .s = s,
1220 .grant_idx = 0,
1221 };
1222
1223 num_grant = s->req.operation == BLKIF_OP_INDIRECT ?
1224 s->req.u.indirect.nr_segments : s->req.u.rw.nr_segments;
1225 num_sg = s->num_sg;
1226
1227 if (bret->operation == BLKIF_OP_READ && info->feature_persistent) {
1228 for_each_sg(s->sg, sg, num_sg, i) {
1229 BUG_ON(sg->offset + sg->length > PAGE_SIZE);
1230
1231 data.bvec_offset = sg->offset;
1232 data.bvec_data = kmap_atomic(sg_page(sg));
1233
1234 gnttab_foreach_grant_in_range(sg_page(sg),
1235 sg->offset,
1236 sg->length,
1237 blkif_copy_from_grant,
1238 &data);
1239
1240 kunmap_atomic(data.bvec_data);
1241 }
1242 }
1243 /* Add the persistent grant into the list of free grants */
1244 for (i = 0; i < num_grant; i++) {
1245 if (gnttab_query_foreign_access(s->grants_used[i]->gref)) {
1246 /*
1247 * If the grant is still mapped by the backend (the
1248 * backend has chosen to make this grant persistent)
1249 * we add it at the head of the list, so it will be
1250 * reused first.
1251 */
1252 if (!info->feature_persistent)
1253 pr_alert_ratelimited("backed has not unmapped grant: %u\n",
1254 s->grants_used[i]->gref);
1255 list_add(&s->grants_used[i]->node, &info->grants);
1256 info->persistent_gnts_c++;
1257 } else {
1258 /*
1259 * If the grant is not mapped by the backend we end the
1260 * foreign access and add it to the tail of the list,
1261 * so it will not be picked again unless we run out of
1262 * persistent grants.
1263 */
1264 gnttab_end_foreign_access(s->grants_used[i]->gref, 0, 0UL);
1265 s->grants_used[i]->gref = GRANT_INVALID_REF;
1266 list_add_tail(&s->grants_used[i]->node, &info->grants);
1267 }
1268 }
1269 if (s->req.operation == BLKIF_OP_INDIRECT) {
1270 for (i = 0; i < INDIRECT_GREFS(num_grant); i++) {
1271 if (gnttab_query_foreign_access(s->indirect_grants[i]->gref)) {
1272 if (!info->feature_persistent)
1273 pr_alert_ratelimited("backed has not unmapped grant: %u\n",
1274 s->indirect_grants[i]->gref);
1275 list_add(&s->indirect_grants[i]->node, &info->grants);
1276 info->persistent_gnts_c++;
1277 } else {
1278 struct page *indirect_page;
1279
1280 gnttab_end_foreign_access(s->indirect_grants[i]->gref, 0, 0UL);
1281 /*
1282 * Add the used indirect page back to the list of
1283 * available pages for indirect grefs.
1284 */
1285 if (!info->feature_persistent) {
1286 indirect_page = s->indirect_grants[i]->page;
1287 list_add(&indirect_page->lru, &info->indirect_pages);
1288 }
1289 s->indirect_grants[i]->gref = GRANT_INVALID_REF;
1290 list_add_tail(&s->indirect_grants[i]->node, &info->grants);
1291 }
1292 }
1293 }
1294 }
1295
1296 static irqreturn_t blkif_interrupt(int irq, void *dev_id)
1297 {
1298 struct request *req;
1299 struct blkif_response *bret;
1300 RING_IDX i, rp;
1301 unsigned long flags;
1302 struct blkfront_info *info = (struct blkfront_info *)dev_id;
1303 int error;
1304
1305 spin_lock_irqsave(&info->io_lock, flags);
1306
1307 if (unlikely(info->connected != BLKIF_STATE_CONNECTED)) {
1308 spin_unlock_irqrestore(&info->io_lock, flags);
1309 return IRQ_HANDLED;
1310 }
1311
1312 again:
1313 rp = info->ring.sring->rsp_prod;
1314 rmb(); /* Ensure we see queued responses up to 'rp'. */
1315
1316 for (i = info->ring.rsp_cons; i != rp; i++) {
1317 unsigned long id;
1318
1319 bret = RING_GET_RESPONSE(&info->ring, i);
1320 id = bret->id;
1321 /*
1322 * The backend has messed up and given us an id that we would
1323 * never have given to it (we stamp it up to BLK_RING_SIZE -
1324 * look in get_id_from_freelist.
1325 */
1326 if (id >= BLK_RING_SIZE(info)) {
1327 WARN(1, "%s: response to %s has incorrect id (%ld)\n",
1328 info->gd->disk_name, op_name(bret->operation), id);
1329 /* We can't safely get the 'struct request' as
1330 * the id is busted. */
1331 continue;
1332 }
1333 req = info->shadow[id].request;
1334
1335 if (bret->operation != BLKIF_OP_DISCARD)
1336 blkif_completion(&info->shadow[id], info, bret);
1337
1338 if (add_id_to_freelist(info, id)) {
1339 WARN(1, "%s: response to %s (id %ld) couldn't be recycled!\n",
1340 info->gd->disk_name, op_name(bret->operation), id);
1341 continue;
1342 }
1343
1344 error = (bret->status == BLKIF_RSP_OKAY) ? 0 : -EIO;
1345 switch (bret->operation) {
1346 case BLKIF_OP_DISCARD:
1347 if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) {
1348 struct request_queue *rq = info->rq;
1349 printk(KERN_WARNING "blkfront: %s: %s op failed\n",
1350 info->gd->disk_name, op_name(bret->operation));
1351 error = -EOPNOTSUPP;
1352 info->feature_discard = 0;
1353 info->feature_secdiscard = 0;
1354 queue_flag_clear(QUEUE_FLAG_DISCARD, rq);
1355 queue_flag_clear(QUEUE_FLAG_SECDISCARD, rq);
1356 }
1357 blk_mq_complete_request(req, error);
1358 break;
1359 case BLKIF_OP_FLUSH_DISKCACHE:
1360 case BLKIF_OP_WRITE_BARRIER:
1361 if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) {
1362 printk(KERN_WARNING "blkfront: %s: %s op failed\n",
1363 info->gd->disk_name, op_name(bret->operation));
1364 error = -EOPNOTSUPP;
1365 }
1366 if (unlikely(bret->status == BLKIF_RSP_ERROR &&
1367 info->shadow[id].req.u.rw.nr_segments == 0)) {
1368 printk(KERN_WARNING "blkfront: %s: empty %s op failed\n",
1369 info->gd->disk_name, op_name(bret->operation));
1370 error = -EOPNOTSUPP;
1371 }
1372 if (unlikely(error)) {
1373 if (error == -EOPNOTSUPP)
1374 error = 0;
1375 info->feature_flush = 0;
1376 xlvbd_flush(info);
1377 }
1378 /* fall through */
1379 case BLKIF_OP_READ:
1380 case BLKIF_OP_WRITE:
1381 if (unlikely(bret->status != BLKIF_RSP_OKAY))
1382 dev_dbg(&info->xbdev->dev, "Bad return from blkdev data "
1383 "request: %x\n", bret->status);
1384
1385 blk_mq_complete_request(req, error);
1386 break;
1387 default:
1388 BUG();
1389 }
1390 }
1391
1392 info->ring.rsp_cons = i;
1393
1394 if (i != info->ring.req_prod_pvt) {
1395 int more_to_do;
1396 RING_FINAL_CHECK_FOR_RESPONSES(&info->ring, more_to_do);
1397 if (more_to_do)
1398 goto again;
1399 } else
1400 info->ring.sring->rsp_event = i + 1;
1401
1402 kick_pending_request_queues(info);
1403
1404 spin_unlock_irqrestore(&info->io_lock, flags);
1405
1406 return IRQ_HANDLED;
1407 }
1408
1409
1410 static int setup_blkring(struct xenbus_device *dev,
1411 struct blkfront_info *info)
1412 {
1413 struct blkif_sring *sring;
1414 int err, i;
1415 unsigned long ring_size = info->nr_ring_pages * XEN_PAGE_SIZE;
1416 grant_ref_t gref[XENBUS_MAX_RING_GRANTS];
1417
1418 for (i = 0; i < info->nr_ring_pages; i++)
1419 info->ring_ref[i] = GRANT_INVALID_REF;
1420
1421 sring = (struct blkif_sring *)__get_free_pages(GFP_NOIO | __GFP_HIGH,
1422 get_order(ring_size));
1423 if (!sring) {
1424 xenbus_dev_fatal(dev, -ENOMEM, "allocating shared ring");
1425 return -ENOMEM;
1426 }
1427 SHARED_RING_INIT(sring);
1428 FRONT_RING_INIT(&info->ring, sring, ring_size);
1429
1430 err = xenbus_grant_ring(dev, info->ring.sring, info->nr_ring_pages, gref);
1431 if (err < 0) {
1432 free_pages((unsigned long)sring, get_order(ring_size));
1433 info->ring.sring = NULL;
1434 goto fail;
1435 }
1436 for (i = 0; i < info->nr_ring_pages; i++)
1437 info->ring_ref[i] = gref[i];
1438
1439 err = xenbus_alloc_evtchn(dev, &info->evtchn);
1440 if (err)
1441 goto fail;
1442
1443 err = bind_evtchn_to_irqhandler(info->evtchn, blkif_interrupt, 0,
1444 "blkif", info);
1445 if (err <= 0) {
1446 xenbus_dev_fatal(dev, err,
1447 "bind_evtchn_to_irqhandler failed");
1448 goto fail;
1449 }
1450 info->irq = err;
1451
1452 return 0;
1453 fail:
1454 blkif_free(info, 0);
1455 return err;
1456 }
1457
1458
1459 /* Common code used when first setting up, and when resuming. */
1460 static int talk_to_blkback(struct xenbus_device *dev,
1461 struct blkfront_info *info)
1462 {
1463 const char *message = NULL;
1464 struct xenbus_transaction xbt;
1465 int err, i;
1466 unsigned int max_page_order = 0;
1467 unsigned int ring_page_order = 0;
1468
1469 err = xenbus_scanf(XBT_NIL, info->xbdev->otherend,
1470 "max-ring-page-order", "%u", &max_page_order);
1471 if (err != 1)
1472 info->nr_ring_pages = 1;
1473 else {
1474 ring_page_order = min(xen_blkif_max_ring_order, max_page_order);
1475 info->nr_ring_pages = 1 << ring_page_order;
1476 }
1477
1478 /* Create shared ring, alloc event channel. */
1479 err = setup_blkring(dev, info);
1480 if (err)
1481 goto out;
1482
1483 again:
1484 err = xenbus_transaction_start(&xbt);
1485 if (err) {
1486 xenbus_dev_fatal(dev, err, "starting transaction");
1487 goto destroy_blkring;
1488 }
1489
1490 if (info->nr_ring_pages == 1) {
1491 err = xenbus_printf(xbt, dev->nodename,
1492 "ring-ref", "%u", info->ring_ref[0]);
1493 if (err) {
1494 message = "writing ring-ref";
1495 goto abort_transaction;
1496 }
1497 } else {
1498 err = xenbus_printf(xbt, dev->nodename,
1499 "ring-page-order", "%u", ring_page_order);
1500 if (err) {
1501 message = "writing ring-page-order";
1502 goto abort_transaction;
1503 }
1504
1505 for (i = 0; i < info->nr_ring_pages; i++) {
1506 char ring_ref_name[RINGREF_NAME_LEN];
1507
1508 snprintf(ring_ref_name, RINGREF_NAME_LEN, "ring-ref%u", i);
1509 err = xenbus_printf(xbt, dev->nodename, ring_ref_name,
1510 "%u", info->ring_ref[i]);
1511 if (err) {
1512 message = "writing ring-ref";
1513 goto abort_transaction;
1514 }
1515 }
1516 }
1517 err = xenbus_printf(xbt, dev->nodename,
1518 "event-channel", "%u", info->evtchn);
1519 if (err) {
1520 message = "writing event-channel";
1521 goto abort_transaction;
1522 }
1523 err = xenbus_printf(xbt, dev->nodename, "protocol", "%s",
1524 XEN_IO_PROTO_ABI_NATIVE);
1525 if (err) {
1526 message = "writing protocol";
1527 goto abort_transaction;
1528 }
1529 err = xenbus_printf(xbt, dev->nodename,
1530 "feature-persistent", "%u", 1);
1531 if (err)
1532 dev_warn(&dev->dev,
1533 "writing persistent grants feature to xenbus");
1534
1535 err = xenbus_transaction_end(xbt, 0);
1536 if (err) {
1537 if (err == -EAGAIN)
1538 goto again;
1539 xenbus_dev_fatal(dev, err, "completing transaction");
1540 goto destroy_blkring;
1541 }
1542
1543 for (i = 0; i < BLK_RING_SIZE(info); i++)
1544 info->shadow[i].req.u.rw.id = i+1;
1545 info->shadow[BLK_RING_SIZE(info)-1].req.u.rw.id = 0x0fffffff;
1546 xenbus_switch_state(dev, XenbusStateInitialised);
1547
1548 return 0;
1549
1550 abort_transaction:
1551 xenbus_transaction_end(xbt, 1);
1552 if (message)
1553 xenbus_dev_fatal(dev, err, "%s", message);
1554 destroy_blkring:
1555 blkif_free(info, 0);
1556 out:
1557 return err;
1558 }
1559
1560 /**
1561 * Entry point to this code when a new device is created. Allocate the basic
1562 * structures and the ring buffer for communication with the backend, and
1563 * inform the backend of the appropriate details for those. Switch to
1564 * Initialised state.
1565 */
1566 static int blkfront_probe(struct xenbus_device *dev,
1567 const struct xenbus_device_id *id)
1568 {
1569 int err, vdevice;
1570 struct blkfront_info *info;
1571
1572 /* FIXME: Use dynamic device id if this is not set. */
1573 err = xenbus_scanf(XBT_NIL, dev->nodename,
1574 "virtual-device", "%i", &vdevice);
1575 if (err != 1) {
1576 /* go looking in the extended area instead */
1577 err = xenbus_scanf(XBT_NIL, dev->nodename, "virtual-device-ext",
1578 "%i", &vdevice);
1579 if (err != 1) {
1580 xenbus_dev_fatal(dev, err, "reading virtual-device");
1581 return err;
1582 }
1583 }
1584
1585 if (xen_hvm_domain()) {
1586 char *type;
1587 int len;
1588 /* no unplug has been done: do not hook devices != xen vbds */
1589 if (xen_has_pv_and_legacy_disk_devices()) {
1590 int major;
1591
1592 if (!VDEV_IS_EXTENDED(vdevice))
1593 major = BLKIF_MAJOR(vdevice);
1594 else
1595 major = XENVBD_MAJOR;
1596
1597 if (major != XENVBD_MAJOR) {
1598 printk(KERN_INFO
1599 "%s: HVM does not support vbd %d as xen block device\n",
1600 __func__, vdevice);
1601 return -ENODEV;
1602 }
1603 }
1604 /* do not create a PV cdrom device if we are an HVM guest */
1605 type = xenbus_read(XBT_NIL, dev->nodename, "device-type", &len);
1606 if (IS_ERR(type))
1607 return -ENODEV;
1608 if (strncmp(type, "cdrom", 5) == 0) {
1609 kfree(type);
1610 return -ENODEV;
1611 }
1612 kfree(type);
1613 }
1614 info = kzalloc(sizeof(*info), GFP_KERNEL);
1615 if (!info) {
1616 xenbus_dev_fatal(dev, -ENOMEM, "allocating info structure");
1617 return -ENOMEM;
1618 }
1619
1620 mutex_init(&info->mutex);
1621 spin_lock_init(&info->io_lock);
1622 info->xbdev = dev;
1623 info->vdevice = vdevice;
1624 INIT_LIST_HEAD(&info->grants);
1625 INIT_LIST_HEAD(&info->indirect_pages);
1626 info->persistent_gnts_c = 0;
1627 info->connected = BLKIF_STATE_DISCONNECTED;
1628 INIT_WORK(&info->work, blkif_restart_queue);
1629
1630 /* Front end dir is a number, which is used as the id. */
1631 info->handle = simple_strtoul(strrchr(dev->nodename, '/')+1, NULL, 0);
1632 dev_set_drvdata(&dev->dev, info);
1633
1634 return 0;
1635 }
1636
1637 static void split_bio_end(struct bio *bio)
1638 {
1639 struct split_bio *split_bio = bio->bi_private;
1640
1641 if (atomic_dec_and_test(&split_bio->pending)) {
1642 split_bio->bio->bi_phys_segments = 0;
1643 split_bio->bio->bi_error = bio->bi_error;
1644 bio_endio(split_bio->bio);
1645 kfree(split_bio);
1646 }
1647 bio_put(bio);
1648 }
1649
1650 static int blkif_recover(struct blkfront_info *info)
1651 {
1652 int i;
1653 struct request *req, *n;
1654 struct blk_shadow *copy;
1655 int rc;
1656 struct bio *bio, *cloned_bio;
1657 struct bio_list bio_list, merge_bio;
1658 unsigned int segs, offset;
1659 int pending, size;
1660 struct split_bio *split_bio;
1661 struct list_head requests;
1662
1663 /* Stage 1: Make a safe copy of the shadow state. */
1664 copy = kmemdup(info->shadow, sizeof(info->shadow),
1665 GFP_NOIO | __GFP_REPEAT | __GFP_HIGH);
1666 if (!copy)
1667 return -ENOMEM;
1668
1669 /* Stage 2: Set up free list. */
1670 memset(&info->shadow, 0, sizeof(info->shadow));
1671 for (i = 0; i < BLK_RING_SIZE(info); i++)
1672 info->shadow[i].req.u.rw.id = i+1;
1673 info->shadow_free = info->ring.req_prod_pvt;
1674 info->shadow[BLK_RING_SIZE(info)-1].req.u.rw.id = 0x0fffffff;
1675
1676 rc = blkfront_gather_backend_features(info);
1677 if (rc) {
1678 kfree(copy);
1679 return rc;
1680 }
1681
1682 segs = info->max_indirect_segments ? : BLKIF_MAX_SEGMENTS_PER_REQUEST;
1683 blk_queue_max_segments(info->rq, segs);
1684 bio_list_init(&bio_list);
1685 INIT_LIST_HEAD(&requests);
1686 for (i = 0; i < BLK_RING_SIZE(info); i++) {
1687 /* Not in use? */
1688 if (!copy[i].request)
1689 continue;
1690
1691 /*
1692 * Get the bios in the request so we can re-queue them.
1693 */
1694 if (copy[i].request->cmd_flags &
1695 (REQ_FLUSH | REQ_FUA | REQ_DISCARD | REQ_SECURE)) {
1696 /*
1697 * Flush operations don't contain bios, so
1698 * we need to requeue the whole request
1699 */
1700 list_add(&copy[i].request->queuelist, &requests);
1701 continue;
1702 }
1703 merge_bio.head = copy[i].request->bio;
1704 merge_bio.tail = copy[i].request->biotail;
1705 bio_list_merge(&bio_list, &merge_bio);
1706 copy[i].request->bio = NULL;
1707 blk_end_request_all(copy[i].request, 0);
1708 }
1709
1710 kfree(copy);
1711
1712 xenbus_switch_state(info->xbdev, XenbusStateConnected);
1713
1714 spin_lock_irq(&info->io_lock);
1715
1716 /* Now safe for us to use the shared ring */
1717 info->connected = BLKIF_STATE_CONNECTED;
1718
1719 /* Kick any other new requests queued since we resumed */
1720 kick_pending_request_queues(info);
1721
1722 list_for_each_entry_safe(req, n, &requests, queuelist) {
1723 /* Requeue pending requests (flush or discard) */
1724 list_del_init(&req->queuelist);
1725 BUG_ON(req->nr_phys_segments > segs);
1726 blk_mq_requeue_request(req);
1727 }
1728 spin_unlock_irq(&info->io_lock);
1729 blk_mq_kick_requeue_list(info->rq);
1730
1731 while ((bio = bio_list_pop(&bio_list)) != NULL) {
1732 /* Traverse the list of pending bios and re-queue them */
1733 if (bio_segments(bio) > segs) {
1734 /*
1735 * This bio has more segments than what we can
1736 * handle, we have to split it.
1737 */
1738 pending = (bio_segments(bio) + segs - 1) / segs;
1739 split_bio = kzalloc(sizeof(*split_bio), GFP_NOIO);
1740 BUG_ON(split_bio == NULL);
1741 atomic_set(&split_bio->pending, pending);
1742 split_bio->bio = bio;
1743 for (i = 0; i < pending; i++) {
1744 offset = (i * segs * XEN_PAGE_SIZE) >> 9;
1745 size = min((unsigned int)(segs * XEN_PAGE_SIZE) >> 9,
1746 (unsigned int)bio_sectors(bio) - offset);
1747 cloned_bio = bio_clone(bio, GFP_NOIO);
1748 BUG_ON(cloned_bio == NULL);
1749 bio_trim(cloned_bio, offset, size);
1750 cloned_bio->bi_private = split_bio;
1751 cloned_bio->bi_end_io = split_bio_end;
1752 submit_bio(cloned_bio->bi_rw, cloned_bio);
1753 }
1754 /*
1755 * Now we have to wait for all those smaller bios to
1756 * end, so we can also end the "parent" bio.
1757 */
1758 continue;
1759 }
1760 /* We don't need to split this bio */
1761 submit_bio(bio->bi_rw, bio);
1762 }
1763
1764 return 0;
1765 }
1766
1767 /**
1768 * We are reconnecting to the backend, due to a suspend/resume, or a backend
1769 * driver restart. We tear down our blkif structure and recreate it, but
1770 * leave the device-layer structures intact so that this is transparent to the
1771 * rest of the kernel.
1772 */
1773 static int blkfront_resume(struct xenbus_device *dev)
1774 {
1775 struct blkfront_info *info = dev_get_drvdata(&dev->dev);
1776 int err;
1777
1778 dev_dbg(&dev->dev, "blkfront_resume: %s\n", dev->nodename);
1779
1780 blkif_free(info, info->connected == BLKIF_STATE_CONNECTED);
1781
1782 err = talk_to_blkback(dev, info);
1783
1784 /*
1785 * We have to wait for the backend to switch to
1786 * connected state, since we want to read which
1787 * features it supports.
1788 */
1789
1790 return err;
1791 }
1792
1793 static void
1794 blkfront_closing(struct blkfront_info *info)
1795 {
1796 struct xenbus_device *xbdev = info->xbdev;
1797 struct block_device *bdev = NULL;
1798
1799 mutex_lock(&info->mutex);
1800
1801 if (xbdev->state == XenbusStateClosing) {
1802 mutex_unlock(&info->mutex);
1803 return;
1804 }
1805
1806 if (info->gd)
1807 bdev = bdget_disk(info->gd, 0);
1808
1809 mutex_unlock(&info->mutex);
1810
1811 if (!bdev) {
1812 xenbus_frontend_closed(xbdev);
1813 return;
1814 }
1815
1816 mutex_lock(&bdev->bd_mutex);
1817
1818 if (bdev->bd_openers) {
1819 xenbus_dev_error(xbdev, -EBUSY,
1820 "Device in use; refusing to close");
1821 xenbus_switch_state(xbdev, XenbusStateClosing);
1822 } else {
1823 xlvbd_release_gendisk(info);
1824 xenbus_frontend_closed(xbdev);
1825 }
1826
1827 mutex_unlock(&bdev->bd_mutex);
1828 bdput(bdev);
1829 }
1830
1831 static void blkfront_setup_discard(struct blkfront_info *info)
1832 {
1833 int err;
1834 unsigned int discard_granularity;
1835 unsigned int discard_alignment;
1836 unsigned int discard_secure;
1837
1838 info->feature_discard = 1;
1839 err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1840 "discard-granularity", "%u", &discard_granularity,
1841 "discard-alignment", "%u", &discard_alignment,
1842 NULL);
1843 if (!err) {
1844 info->discard_granularity = discard_granularity;
1845 info->discard_alignment = discard_alignment;
1846 }
1847 err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1848 "discard-secure", "%d", &discard_secure,
1849 NULL);
1850 if (!err)
1851 info->feature_secdiscard = !!discard_secure;
1852 }
1853
1854 static int blkfront_setup_indirect(struct blkfront_info *info)
1855 {
1856 unsigned int psegs, grants;
1857 int err, i;
1858
1859 if (info->max_indirect_segments == 0)
1860 grants = BLKIF_MAX_SEGMENTS_PER_REQUEST;
1861 else
1862 grants = info->max_indirect_segments;
1863 psegs = grants / GRANTS_PER_PSEG;
1864
1865 err = fill_grant_buffer(info,
1866 (grants + INDIRECT_GREFS(grants)) * BLK_RING_SIZE(info));
1867 if (err)
1868 goto out_of_memory;
1869
1870 if (!info->feature_persistent && info->max_indirect_segments) {
1871 /*
1872 * We are using indirect descriptors but not persistent
1873 * grants, we need to allocate a set of pages that can be
1874 * used for mapping indirect grefs
1875 */
1876 int num = INDIRECT_GREFS(grants) * BLK_RING_SIZE(info);
1877
1878 BUG_ON(!list_empty(&info->indirect_pages));
1879 for (i = 0; i < num; i++) {
1880 struct page *indirect_page = alloc_page(GFP_NOIO);
1881 if (!indirect_page)
1882 goto out_of_memory;
1883 list_add(&indirect_page->lru, &info->indirect_pages);
1884 }
1885 }
1886
1887 for (i = 0; i < BLK_RING_SIZE(info); i++) {
1888 info->shadow[i].grants_used = kzalloc(
1889 sizeof(info->shadow[i].grants_used[0]) * grants,
1890 GFP_NOIO);
1891 info->shadow[i].sg = kzalloc(sizeof(info->shadow[i].sg[0]) * psegs, GFP_NOIO);
1892 if (info->max_indirect_segments)
1893 info->shadow[i].indirect_grants = kzalloc(
1894 sizeof(info->shadow[i].indirect_grants[0]) *
1895 INDIRECT_GREFS(grants),
1896 GFP_NOIO);
1897 if ((info->shadow[i].grants_used == NULL) ||
1898 (info->shadow[i].sg == NULL) ||
1899 (info->max_indirect_segments &&
1900 (info->shadow[i].indirect_grants == NULL)))
1901 goto out_of_memory;
1902 sg_init_table(info->shadow[i].sg, psegs);
1903 }
1904
1905
1906 return 0;
1907
1908 out_of_memory:
1909 for (i = 0; i < BLK_RING_SIZE(info); i++) {
1910 kfree(info->shadow[i].grants_used);
1911 info->shadow[i].grants_used = NULL;
1912 kfree(info->shadow[i].sg);
1913 info->shadow[i].sg = NULL;
1914 kfree(info->shadow[i].indirect_grants);
1915 info->shadow[i].indirect_grants = NULL;
1916 }
1917 if (!list_empty(&info->indirect_pages)) {
1918 struct page *indirect_page, *n;
1919 list_for_each_entry_safe(indirect_page, n, &info->indirect_pages, lru) {
1920 list_del(&indirect_page->lru);
1921 __free_page(indirect_page);
1922 }
1923 }
1924 return -ENOMEM;
1925 }
1926
1927 /*
1928 * Gather all backend feature-*
1929 */
1930 static int blkfront_gather_backend_features(struct blkfront_info *info)
1931 {
1932 int err;
1933 int barrier, flush, discard, persistent;
1934 unsigned int indirect_segments;
1935
1936 info->feature_flush = 0;
1937
1938 err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1939 "feature-barrier", "%d", &barrier,
1940 NULL);
1941
1942 /*
1943 * If there's no "feature-barrier" defined, then it means
1944 * we're dealing with a very old backend which writes
1945 * synchronously; nothing to do.
1946 *
1947 * If there are barriers, then we use flush.
1948 */
1949 if (!err && barrier)
1950 info->feature_flush = REQ_FLUSH | REQ_FUA;
1951 /*
1952 * And if there is "feature-flush-cache" use that above
1953 * barriers.
1954 */
1955 err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1956 "feature-flush-cache", "%d", &flush,
1957 NULL);
1958
1959 if (!err && flush)
1960 info->feature_flush = REQ_FLUSH;
1961
1962 err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1963 "feature-discard", "%d", &discard,
1964 NULL);
1965
1966 if (!err && discard)
1967 blkfront_setup_discard(info);
1968
1969 err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1970 "feature-persistent", "%u", &persistent,
1971 NULL);
1972 if (err)
1973 info->feature_persistent = 0;
1974 else
1975 info->feature_persistent = persistent;
1976
1977 err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1978 "feature-max-indirect-segments", "%u", &indirect_segments,
1979 NULL);
1980 if (err)
1981 info->max_indirect_segments = 0;
1982 else
1983 info->max_indirect_segments = min(indirect_segments,
1984 xen_blkif_max_segments);
1985
1986 return blkfront_setup_indirect(info);
1987 }
1988
1989 /*
1990 * Invoked when the backend is finally 'ready' (and has told produced
1991 * the details about the physical device - #sectors, size, etc).
1992 */
1993 static void blkfront_connect(struct blkfront_info *info)
1994 {
1995 unsigned long long sectors;
1996 unsigned long sector_size;
1997 unsigned int physical_sector_size;
1998 unsigned int binfo;
1999 int err;
2000
2001 switch (info->connected) {
2002 case BLKIF_STATE_CONNECTED:
2003 /*
2004 * Potentially, the back-end may be signalling
2005 * a capacity change; update the capacity.
2006 */
2007 err = xenbus_scanf(XBT_NIL, info->xbdev->otherend,
2008 "sectors", "%Lu", &sectors);
2009 if (XENBUS_EXIST_ERR(err))
2010 return;
2011 printk(KERN_INFO "Setting capacity to %Lu\n",
2012 sectors);
2013 set_capacity(info->gd, sectors);
2014 revalidate_disk(info->gd);
2015
2016 return;
2017 case BLKIF_STATE_SUSPENDED:
2018 /*
2019 * If we are recovering from suspension, we need to wait
2020 * for the backend to announce it's features before
2021 * reconnecting, at least we need to know if the backend
2022 * supports indirect descriptors, and how many.
2023 */
2024 blkif_recover(info);
2025 return;
2026
2027 default:
2028 break;
2029 }
2030
2031 dev_dbg(&info->xbdev->dev, "%s:%s.\n",
2032 __func__, info->xbdev->otherend);
2033
2034 err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
2035 "sectors", "%llu", &sectors,
2036 "info", "%u", &binfo,
2037 "sector-size", "%lu", &sector_size,
2038 NULL);
2039 if (err) {
2040 xenbus_dev_fatal(info->xbdev, err,
2041 "reading backend fields at %s",
2042 info->xbdev->otherend);
2043 return;
2044 }
2045
2046 /*
2047 * physcial-sector-size is a newer field, so old backends may not
2048 * provide this. Assume physical sector size to be the same as
2049 * sector_size in that case.
2050 */
2051 err = xenbus_scanf(XBT_NIL, info->xbdev->otherend,
2052 "physical-sector-size", "%u", &physical_sector_size);
2053 if (err != 1)
2054 physical_sector_size = sector_size;
2055
2056 err = blkfront_gather_backend_features(info);
2057 if (err) {
2058 xenbus_dev_fatal(info->xbdev, err, "setup_indirect at %s",
2059 info->xbdev->otherend);
2060 return;
2061 }
2062
2063 err = xlvbd_alloc_gendisk(sectors, info, binfo, sector_size,
2064 physical_sector_size);
2065 if (err) {
2066 xenbus_dev_fatal(info->xbdev, err, "xlvbd_add at %s",
2067 info->xbdev->otherend);
2068 return;
2069 }
2070
2071 xenbus_switch_state(info->xbdev, XenbusStateConnected);
2072
2073 /* Kick pending requests. */
2074 spin_lock_irq(&info->io_lock);
2075 info->connected = BLKIF_STATE_CONNECTED;
2076 kick_pending_request_queues(info);
2077 spin_unlock_irq(&info->io_lock);
2078
2079 add_disk(info->gd);
2080
2081 info->is_ready = 1;
2082 }
2083
2084 /**
2085 * Callback received when the backend's state changes.
2086 */
2087 static void blkback_changed(struct xenbus_device *dev,
2088 enum xenbus_state backend_state)
2089 {
2090 struct blkfront_info *info = dev_get_drvdata(&dev->dev);
2091
2092 dev_dbg(&dev->dev, "blkfront:blkback_changed to state %d.\n", backend_state);
2093
2094 switch (backend_state) {
2095 case XenbusStateInitWait:
2096 if (dev->state != XenbusStateInitialising)
2097 break;
2098 if (talk_to_blkback(dev, info)) {
2099 kfree(info);
2100 dev_set_drvdata(&dev->dev, NULL);
2101 break;
2102 }
2103 case XenbusStateInitialising:
2104 case XenbusStateInitialised:
2105 case XenbusStateReconfiguring:
2106 case XenbusStateReconfigured:
2107 case XenbusStateUnknown:
2108 break;
2109
2110 case XenbusStateConnected:
2111 blkfront_connect(info);
2112 break;
2113
2114 case XenbusStateClosed:
2115 if (dev->state == XenbusStateClosed)
2116 break;
2117 /* Missed the backend's Closing state -- fallthrough */
2118 case XenbusStateClosing:
2119 if (info)
2120 blkfront_closing(info);
2121 break;
2122 }
2123 }
2124
2125 static int blkfront_remove(struct xenbus_device *xbdev)
2126 {
2127 struct blkfront_info *info = dev_get_drvdata(&xbdev->dev);
2128 struct block_device *bdev = NULL;
2129 struct gendisk *disk;
2130
2131 dev_dbg(&xbdev->dev, "%s removed", xbdev->nodename);
2132
2133 blkif_free(info, 0);
2134
2135 mutex_lock(&info->mutex);
2136
2137 disk = info->gd;
2138 if (disk)
2139 bdev = bdget_disk(disk, 0);
2140
2141 info->xbdev = NULL;
2142 mutex_unlock(&info->mutex);
2143
2144 if (!bdev) {
2145 kfree(info);
2146 return 0;
2147 }
2148
2149 /*
2150 * The xbdev was removed before we reached the Closed
2151 * state. See if it's safe to remove the disk. If the bdev
2152 * isn't closed yet, we let release take care of it.
2153 */
2154
2155 mutex_lock(&bdev->bd_mutex);
2156 info = disk->private_data;
2157
2158 dev_warn(disk_to_dev(disk),
2159 "%s was hot-unplugged, %d stale handles\n",
2160 xbdev->nodename, bdev->bd_openers);
2161
2162 if (info && !bdev->bd_openers) {
2163 xlvbd_release_gendisk(info);
2164 disk->private_data = NULL;
2165 kfree(info);
2166 }
2167
2168 mutex_unlock(&bdev->bd_mutex);
2169 bdput(bdev);
2170
2171 return 0;
2172 }
2173
2174 static int blkfront_is_ready(struct xenbus_device *dev)
2175 {
2176 struct blkfront_info *info = dev_get_drvdata(&dev->dev);
2177
2178 return info->is_ready && info->xbdev;
2179 }
2180
2181 static int blkif_open(struct block_device *bdev, fmode_t mode)
2182 {
2183 struct gendisk *disk = bdev->bd_disk;
2184 struct blkfront_info *info;
2185 int err = 0;
2186
2187 mutex_lock(&blkfront_mutex);
2188
2189 info = disk->private_data;
2190 if (!info) {
2191 /* xbdev gone */
2192 err = -ERESTARTSYS;
2193 goto out;
2194 }
2195
2196 mutex_lock(&info->mutex);
2197
2198 if (!info->gd)
2199 /* xbdev is closed */
2200 err = -ERESTARTSYS;
2201
2202 mutex_unlock(&info->mutex);
2203
2204 out:
2205 mutex_unlock(&blkfront_mutex);
2206 return err;
2207 }
2208
2209 static void blkif_release(struct gendisk *disk, fmode_t mode)
2210 {
2211 struct blkfront_info *info = disk->private_data;
2212 struct block_device *bdev;
2213 struct xenbus_device *xbdev;
2214
2215 mutex_lock(&blkfront_mutex);
2216
2217 bdev = bdget_disk(disk, 0);
2218
2219 if (!bdev) {
2220 WARN(1, "Block device %s yanked out from us!\n", disk->disk_name);
2221 goto out_mutex;
2222 }
2223 if (bdev->bd_openers)
2224 goto out;
2225
2226 /*
2227 * Check if we have been instructed to close. We will have
2228 * deferred this request, because the bdev was still open.
2229 */
2230
2231 mutex_lock(&info->mutex);
2232 xbdev = info->xbdev;
2233
2234 if (xbdev && xbdev->state == XenbusStateClosing) {
2235 /* pending switch to state closed */
2236 dev_info(disk_to_dev(bdev->bd_disk), "releasing disk\n");
2237 xlvbd_release_gendisk(info);
2238 xenbus_frontend_closed(info->xbdev);
2239 }
2240
2241 mutex_unlock(&info->mutex);
2242
2243 if (!xbdev) {
2244 /* sudden device removal */
2245 dev_info(disk_to_dev(bdev->bd_disk), "releasing disk\n");
2246 xlvbd_release_gendisk(info);
2247 disk->private_data = NULL;
2248 kfree(info);
2249 }
2250
2251 out:
2252 bdput(bdev);
2253 out_mutex:
2254 mutex_unlock(&blkfront_mutex);
2255 }
2256
2257 static const struct block_device_operations xlvbd_block_fops =
2258 {
2259 .owner = THIS_MODULE,
2260 .open = blkif_open,
2261 .release = blkif_release,
2262 .getgeo = blkif_getgeo,
2263 .ioctl = blkif_ioctl,
2264 };
2265
2266
2267 static const struct xenbus_device_id blkfront_ids[] = {
2268 { "vbd" },
2269 { "" }
2270 };
2271
2272 static struct xenbus_driver blkfront_driver = {
2273 .ids = blkfront_ids,
2274 .probe = blkfront_probe,
2275 .remove = blkfront_remove,
2276 .resume = blkfront_resume,
2277 .otherend_changed = blkback_changed,
2278 .is_ready = blkfront_is_ready,
2279 };
2280
2281 static int __init xlblk_init(void)
2282 {
2283 int ret;
2284
2285 if (!xen_domain())
2286 return -ENODEV;
2287
2288 if (xen_blkif_max_ring_order > XENBUS_MAX_RING_GRANT_ORDER) {
2289 pr_info("Invalid max_ring_order (%d), will use default max: %d.\n",
2290 xen_blkif_max_ring_order, XENBUS_MAX_RING_GRANT_ORDER);
2291 xen_blkif_max_ring_order = 0;
2292 }
2293
2294 if (!xen_has_pv_disk_devices())
2295 return -ENODEV;
2296
2297 if (register_blkdev(XENVBD_MAJOR, DEV_NAME)) {
2298 printk(KERN_WARNING "xen_blk: can't get major %d with name %s\n",
2299 XENVBD_MAJOR, DEV_NAME);
2300 return -ENODEV;
2301 }
2302
2303 ret = xenbus_register_frontend(&blkfront_driver);
2304 if (ret) {
2305 unregister_blkdev(XENVBD_MAJOR, DEV_NAME);
2306 return ret;
2307 }
2308
2309 return 0;
2310 }
2311 module_init(xlblk_init);
2312
2313
2314 static void __exit xlblk_exit(void)
2315 {
2316 xenbus_unregister_driver(&blkfront_driver);
2317 unregister_blkdev(XENVBD_MAJOR, DEV_NAME);
2318 kfree(minors);
2319 }
2320 module_exit(xlblk_exit);
2321
2322 MODULE_DESCRIPTION("Xen virtual block device frontend");
2323 MODULE_LICENSE("GPL");
2324 MODULE_ALIAS_BLOCKDEV_MAJOR(XENVBD_MAJOR);
2325 MODULE_ALIAS("xen:vbd");
2326 MODULE_ALIAS("xenblk");
Linux with added FPC-III support