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