]>
Commit | Line | Data |
---|---|---|
9f27ee59 JF |
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> | |
907c3eb1 | 40 | #include <linux/blk-mq.h> |
597592d9 | 41 | #include <linux/hdreg.h> |
440a01a7 | 42 | #include <linux/cdrom.h> |
9f27ee59 | 43 | #include <linux/module.h> |
5a0e3ad6 | 44 | #include <linux/slab.h> |
2a48fc0a | 45 | #include <linux/mutex.h> |
9e973e64 | 46 | #include <linux/scatterlist.h> |
34ae2e47 | 47 | #include <linux/bitmap.h> |
155b7edb | 48 | #include <linux/list.h> |
9f27ee59 | 49 | |
1ccbf534 | 50 | #include <xen/xen.h> |
9f27ee59 JF |
51 | #include <xen/xenbus.h> |
52 | #include <xen/grant_table.h> | |
53 | #include <xen/events.h> | |
54 | #include <xen/page.h> | |
c1c5413a | 55 | #include <xen/platform_pci.h> |
9f27ee59 JF |
56 | |
57 | #include <xen/interface/grant_table.h> | |
58 | #include <xen/interface/io/blkif.h> | |
3e334239 | 59 | #include <xen/interface/io/protocols.h> |
9f27ee59 JF |
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 | ||
0a8704a5 RPM |
69 | struct grant { |
70 | grant_ref_t gref; | |
a7a6df22 | 71 | struct page *page; |
155b7edb | 72 | struct list_head node; |
0a8704a5 RPM |
73 | }; |
74 | ||
9f27ee59 JF |
75 | struct blk_shadow { |
76 | struct blkif_request req; | |
a945b980 | 77 | struct request *request; |
402b27f9 RPM |
78 | struct grant **grants_used; |
79 | struct grant **indirect_grants; | |
b7649158 | 80 | struct scatterlist *sg; |
c004a6fe | 81 | unsigned int num_sg; |
402b27f9 RPM |
82 | }; |
83 | ||
84 | struct split_bio { | |
85 | struct bio *bio; | |
86 | atomic_t pending; | |
9f27ee59 JF |
87 | }; |
88 | ||
2a48fc0a | 89 | static DEFINE_MUTEX(blkfront_mutex); |
83d5cde4 | 90 | static const struct block_device_operations xlvbd_block_fops; |
9f27ee59 | 91 | |
402b27f9 RPM |
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; | |
2d5dc3ba KRW |
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)"); | |
402b27f9 | 101 | |
86839c56 BL |
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 | ||
c004a6fe JG |
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 \ | |
9cce2914 | 114 | __CONST_RING_SIZE(blkif, XEN_PAGE_SIZE * XENBUS_MAX_RING_GRANTS) |
c004a6fe | 115 | |
86839c56 BL |
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) | |
9f27ee59 | 121 | |
81f35161 BL |
122 | /* |
123 | * Per-ring info. | |
124 | * Every blkfront device can associate with one or more blkfront_ring_info, | |
125 | * depending on how many hardware queues/rings to be used. | |
126 | */ | |
127 | struct blkfront_ring_info { | |
128 | struct blkif_front_ring ring; | |
129 | unsigned int ring_ref[XENBUS_MAX_RING_GRANTS]; | |
130 | unsigned int evtchn, irq; | |
131 | struct work_struct work; | |
132 | struct gnttab_free_callback callback; | |
133 | struct blk_shadow shadow[BLK_MAX_RING_SIZE]; | |
134 | struct list_head indirect_pages; | |
135 | unsigned long shadow_free; | |
136 | struct blkfront_info *dev_info; | |
137 | }; | |
138 | ||
9f27ee59 JF |
139 | /* |
140 | * We have one of these per vbd, whether ide, scsi or 'other'. They | |
141 | * hang in private_data off the gendisk structure. We may end up | |
142 | * putting all kinds of interesting stuff here :-) | |
143 | */ | |
144 | struct blkfront_info | |
145 | { | |
3467811e | 146 | spinlock_t io_lock; |
b70f5fa0 | 147 | struct mutex mutex; |
9f27ee59 | 148 | struct xenbus_device *xbdev; |
9f27ee59 JF |
149 | struct gendisk *gd; |
150 | int vdevice; | |
151 | blkif_vdev_t handle; | |
152 | enum blkif_state connected; | |
3df0e505 | 153 | /* Number of pages per ring buffer. */ |
86839c56 | 154 | unsigned int nr_ring_pages; |
9f27ee59 | 155 | struct request_queue *rq; |
bfe11d6d | 156 | struct list_head grants; |
0a8704a5 | 157 | unsigned int persistent_gnts_c; |
4913efe4 | 158 | unsigned int feature_flush; |
5ea42986 KRW |
159 | unsigned int feature_discard:1; |
160 | unsigned int feature_secdiscard:1; | |
ed30bf31 LD |
161 | unsigned int discard_granularity; |
162 | unsigned int discard_alignment; | |
0a8704a5 | 163 | unsigned int feature_persistent:1; |
c004a6fe | 164 | /* Number of 4KB segments handled */ |
402b27f9 | 165 | unsigned int max_indirect_segments; |
1d78d705 | 166 | int is_ready; |
907c3eb1 | 167 | struct blk_mq_tag_set tag_set; |
3df0e505 BL |
168 | struct blkfront_ring_info *rinfo; |
169 | unsigned int nr_rings; | |
9f27ee59 JF |
170 | }; |
171 | ||
0e345826 JB |
172 | static unsigned int nr_minors; |
173 | static unsigned long *minors; | |
174 | static DEFINE_SPINLOCK(minor_lock); | |
175 | ||
9f27ee59 JF |
176 | #define GRANT_INVALID_REF 0 |
177 | ||
178 | #define PARTS_PER_DISK 16 | |
9246b5f0 | 179 | #define PARTS_PER_EXT_DISK 256 |
9f27ee59 JF |
180 | |
181 | #define BLKIF_MAJOR(dev) ((dev)>>8) | |
182 | #define BLKIF_MINOR(dev) ((dev) & 0xff) | |
183 | ||
9246b5f0 CL |
184 | #define EXT_SHIFT 28 |
185 | #define EXTENDED (1<<EXT_SHIFT) | |
186 | #define VDEV_IS_EXTENDED(dev) ((dev)&(EXTENDED)) | |
187 | #define BLKIF_MINOR_EXT(dev) ((dev)&(~EXTENDED)) | |
c80a4209 SS |
188 | #define EMULATED_HD_DISK_MINOR_OFFSET (0) |
189 | #define EMULATED_HD_DISK_NAME_OFFSET (EMULATED_HD_DISK_MINOR_OFFSET / 256) | |
196cfe2a SB |
190 | #define EMULATED_SD_DISK_MINOR_OFFSET (0) |
191 | #define EMULATED_SD_DISK_NAME_OFFSET (EMULATED_SD_DISK_MINOR_OFFSET / 256) | |
9f27ee59 | 192 | |
9246b5f0 | 193 | #define DEV_NAME "xvd" /* name in /dev */ |
9f27ee59 | 194 | |
c004a6fe JG |
195 | /* |
196 | * Grants are always the same size as a Xen page (i.e 4KB). | |
197 | * A physical segment is always the same size as a Linux page. | |
198 | * Number of grants per physical segment | |
199 | */ | |
200 | #define GRANTS_PER_PSEG (PAGE_SIZE / XEN_PAGE_SIZE) | |
201 | ||
202 | #define GRANTS_PER_INDIRECT_FRAME \ | |
203 | (XEN_PAGE_SIZE / sizeof(struct blkif_request_segment)) | |
204 | ||
205 | #define PSEGS_PER_INDIRECT_FRAME \ | |
206 | (GRANTS_INDIRECT_FRAME / GRANTS_PSEGS) | |
207 | ||
208 | #define INDIRECT_GREFS(_grants) \ | |
209 | DIV_ROUND_UP(_grants, GRANTS_PER_INDIRECT_FRAME) | |
210 | ||
211 | #define GREFS(_psegs) ((_psegs) * GRANTS_PER_PSEG) | |
402b27f9 | 212 | |
81f35161 | 213 | static int blkfront_setup_indirect(struct blkfront_ring_info *rinfo); |
3df0e505 | 214 | static void blkfront_gather_backend_features(struct blkfront_info *info); |
402b27f9 | 215 | |
81f35161 | 216 | static int get_id_from_freelist(struct blkfront_ring_info *rinfo) |
9f27ee59 | 217 | { |
81f35161 BL |
218 | unsigned long free = rinfo->shadow_free; |
219 | ||
220 | BUG_ON(free >= BLK_RING_SIZE(rinfo->dev_info)); | |
221 | rinfo->shadow_free = rinfo->shadow[free].req.u.rw.id; | |
222 | rinfo->shadow[free].req.u.rw.id = 0x0fffffee; /* debug */ | |
9f27ee59 JF |
223 | return free; |
224 | } | |
225 | ||
81f35161 | 226 | static int add_id_to_freelist(struct blkfront_ring_info *rinfo, |
9f27ee59 JF |
227 | unsigned long id) |
228 | { | |
81f35161 | 229 | if (rinfo->shadow[id].req.u.rw.id != id) |
6878c32e | 230 | return -EINVAL; |
81f35161 | 231 | if (rinfo->shadow[id].request == NULL) |
6878c32e | 232 | return -EINVAL; |
81f35161 BL |
233 | rinfo->shadow[id].req.u.rw.id = rinfo->shadow_free; |
234 | rinfo->shadow[id].request = NULL; | |
235 | rinfo->shadow_free = id; | |
6878c32e | 236 | return 0; |
9f27ee59 JF |
237 | } |
238 | ||
81f35161 | 239 | static int fill_grant_buffer(struct blkfront_ring_info *rinfo, int num) |
9c1e050c | 240 | { |
81f35161 | 241 | struct blkfront_info *info = rinfo->dev_info; |
9c1e050c RPM |
242 | struct page *granted_page; |
243 | struct grant *gnt_list_entry, *n; | |
244 | int i = 0; | |
245 | ||
246 | while(i < num) { | |
247 | gnt_list_entry = kzalloc(sizeof(struct grant), GFP_NOIO); | |
248 | if (!gnt_list_entry) | |
249 | goto out_of_memory; | |
250 | ||
bfe11d6d RPM |
251 | if (info->feature_persistent) { |
252 | granted_page = alloc_page(GFP_NOIO); | |
253 | if (!granted_page) { | |
254 | kfree(gnt_list_entry); | |
255 | goto out_of_memory; | |
256 | } | |
a7a6df22 | 257 | gnt_list_entry->page = granted_page; |
9c1e050c RPM |
258 | } |
259 | ||
9c1e050c | 260 | gnt_list_entry->gref = GRANT_INVALID_REF; |
bfe11d6d | 261 | list_add(&gnt_list_entry->node, &info->grants); |
9c1e050c RPM |
262 | i++; |
263 | } | |
264 | ||
265 | return 0; | |
266 | ||
267 | out_of_memory: | |
268 | list_for_each_entry_safe(gnt_list_entry, n, | |
bfe11d6d | 269 | &info->grants, node) { |
9c1e050c | 270 | list_del(&gnt_list_entry->node); |
bfe11d6d | 271 | if (info->feature_persistent) |
a7a6df22 | 272 | __free_page(gnt_list_entry->page); |
9c1e050c RPM |
273 | kfree(gnt_list_entry); |
274 | i--; | |
275 | } | |
276 | BUG_ON(i != 0); | |
277 | return -ENOMEM; | |
278 | } | |
279 | ||
4f503fbd | 280 | static struct grant *get_free_grant(struct blkfront_info *info) |
9c1e050c RPM |
281 | { |
282 | struct grant *gnt_list_entry; | |
9c1e050c | 283 | |
bfe11d6d RPM |
284 | BUG_ON(list_empty(&info->grants)); |
285 | gnt_list_entry = list_first_entry(&info->grants, struct grant, | |
4f503fbd | 286 | node); |
9c1e050c RPM |
287 | list_del(&gnt_list_entry->node); |
288 | ||
4f503fbd | 289 | if (gnt_list_entry->gref != GRANT_INVALID_REF) |
9c1e050c | 290 | info->persistent_gnts_c--; |
4f503fbd JG |
291 | |
292 | return gnt_list_entry; | |
293 | } | |
294 | ||
295 | static inline void grant_foreign_access(const struct grant *gnt_list_entry, | |
296 | const struct blkfront_info *info) | |
297 | { | |
298 | gnttab_page_grant_foreign_access_ref_one(gnt_list_entry->gref, | |
299 | info->xbdev->otherend_id, | |
300 | gnt_list_entry->page, | |
301 | 0); | |
302 | } | |
303 | ||
304 | static struct grant *get_grant(grant_ref_t *gref_head, | |
305 | unsigned long gfn, | |
306 | struct blkfront_info *info) | |
307 | { | |
308 | struct grant *gnt_list_entry = get_free_grant(info); | |
309 | ||
310 | if (gnt_list_entry->gref != GRANT_INVALID_REF) | |
9c1e050c | 311 | return gnt_list_entry; |
4f503fbd JG |
312 | |
313 | /* Assign a gref to this page */ | |
314 | gnt_list_entry->gref = gnttab_claim_grant_reference(gref_head); | |
315 | BUG_ON(gnt_list_entry->gref == -ENOSPC); | |
316 | if (info->feature_persistent) | |
317 | grant_foreign_access(gnt_list_entry, info); | |
318 | else { | |
319 | /* Grant access to the GFN passed by the caller */ | |
320 | gnttab_grant_foreign_access_ref(gnt_list_entry->gref, | |
321 | info->xbdev->otherend_id, | |
322 | gfn, 0); | |
9c1e050c RPM |
323 | } |
324 | ||
4f503fbd JG |
325 | return gnt_list_entry; |
326 | } | |
327 | ||
328 | static struct grant *get_indirect_grant(grant_ref_t *gref_head, | |
329 | struct blkfront_info *info) | |
330 | { | |
331 | struct grant *gnt_list_entry = get_free_grant(info); | |
332 | ||
333 | if (gnt_list_entry->gref != GRANT_INVALID_REF) | |
334 | return gnt_list_entry; | |
335 | ||
9c1e050c RPM |
336 | /* Assign a gref to this page */ |
337 | gnt_list_entry->gref = gnttab_claim_grant_reference(gref_head); | |
338 | BUG_ON(gnt_list_entry->gref == -ENOSPC); | |
bfe11d6d | 339 | if (!info->feature_persistent) { |
4f503fbd JG |
340 | struct page *indirect_page; |
341 | ||
342 | /* Fetch a pre-allocated page to use for indirect grefs */ | |
3df0e505 BL |
343 | BUG_ON(list_empty(&info->rinfo->indirect_pages)); |
344 | indirect_page = list_first_entry(&info->rinfo->indirect_pages, | |
4f503fbd JG |
345 | struct page, lru); |
346 | list_del(&indirect_page->lru); | |
347 | gnt_list_entry->page = indirect_page; | |
bfe11d6d | 348 | } |
4f503fbd JG |
349 | grant_foreign_access(gnt_list_entry, info); |
350 | ||
9c1e050c RPM |
351 | return gnt_list_entry; |
352 | } | |
353 | ||
6878c32e KRW |
354 | static const char *op_name(int op) |
355 | { | |
356 | static const char *const names[] = { | |
357 | [BLKIF_OP_READ] = "read", | |
358 | [BLKIF_OP_WRITE] = "write", | |
359 | [BLKIF_OP_WRITE_BARRIER] = "barrier", | |
360 | [BLKIF_OP_FLUSH_DISKCACHE] = "flush", | |
361 | [BLKIF_OP_DISCARD] = "discard" }; | |
362 | ||
363 | if (op < 0 || op >= ARRAY_SIZE(names)) | |
364 | return "unknown"; | |
365 | ||
366 | if (!names[op]) | |
367 | return "reserved"; | |
368 | ||
369 | return names[op]; | |
370 | } | |
0e345826 JB |
371 | static int xlbd_reserve_minors(unsigned int minor, unsigned int nr) |
372 | { | |
373 | unsigned int end = minor + nr; | |
374 | int rc; | |
375 | ||
376 | if (end > nr_minors) { | |
377 | unsigned long *bitmap, *old; | |
378 | ||
f094148a | 379 | bitmap = kcalloc(BITS_TO_LONGS(end), sizeof(*bitmap), |
0e345826 JB |
380 | GFP_KERNEL); |
381 | if (bitmap == NULL) | |
382 | return -ENOMEM; | |
383 | ||
384 | spin_lock(&minor_lock); | |
385 | if (end > nr_minors) { | |
386 | old = minors; | |
387 | memcpy(bitmap, minors, | |
388 | BITS_TO_LONGS(nr_minors) * sizeof(*bitmap)); | |
389 | minors = bitmap; | |
390 | nr_minors = BITS_TO_LONGS(end) * BITS_PER_LONG; | |
391 | } else | |
392 | old = bitmap; | |
393 | spin_unlock(&minor_lock); | |
394 | kfree(old); | |
395 | } | |
396 | ||
397 | spin_lock(&minor_lock); | |
398 | if (find_next_bit(minors, end, minor) >= end) { | |
34ae2e47 | 399 | bitmap_set(minors, minor, nr); |
0e345826 JB |
400 | rc = 0; |
401 | } else | |
402 | rc = -EBUSY; | |
403 | spin_unlock(&minor_lock); | |
404 | ||
405 | return rc; | |
406 | } | |
407 | ||
408 | static void xlbd_release_minors(unsigned int minor, unsigned int nr) | |
409 | { | |
410 | unsigned int end = minor + nr; | |
411 | ||
412 | BUG_ON(end > nr_minors); | |
413 | spin_lock(&minor_lock); | |
34ae2e47 | 414 | bitmap_clear(minors, minor, nr); |
0e345826 JB |
415 | spin_unlock(&minor_lock); |
416 | } | |
417 | ||
9f27ee59 JF |
418 | static void blkif_restart_queue_callback(void *arg) |
419 | { | |
81f35161 BL |
420 | struct blkfront_ring_info *rinfo = (struct blkfront_ring_info *)arg; |
421 | schedule_work(&rinfo->work); | |
9f27ee59 JF |
422 | } |
423 | ||
afe42d7d | 424 | static int blkif_getgeo(struct block_device *bd, struct hd_geometry *hg) |
597592d9 IC |
425 | { |
426 | /* We don't have real geometry info, but let's at least return | |
427 | values consistent with the size of the device */ | |
428 | sector_t nsect = get_capacity(bd->bd_disk); | |
429 | sector_t cylinders = nsect; | |
430 | ||
431 | hg->heads = 0xff; | |
432 | hg->sectors = 0x3f; | |
433 | sector_div(cylinders, hg->heads * hg->sectors); | |
434 | hg->cylinders = cylinders; | |
435 | if ((sector_t)(hg->cylinders + 1) * hg->heads * hg->sectors < nsect) | |
436 | hg->cylinders = 0xffff; | |
437 | return 0; | |
438 | } | |
439 | ||
a63c848b | 440 | static int blkif_ioctl(struct block_device *bdev, fmode_t mode, |
62aa0054 | 441 | unsigned command, unsigned long argument) |
440a01a7 | 442 | { |
a63c848b | 443 | struct blkfront_info *info = bdev->bd_disk->private_data; |
440a01a7 CL |
444 | int i; |
445 | ||
446 | dev_dbg(&info->xbdev->dev, "command: 0x%x, argument: 0x%lx\n", | |
447 | command, (long)argument); | |
448 | ||
449 | switch (command) { | |
450 | case CDROMMULTISESSION: | |
451 | dev_dbg(&info->xbdev->dev, "FIXME: support multisession CDs later\n"); | |
452 | for (i = 0; i < sizeof(struct cdrom_multisession); i++) | |
453 | if (put_user(0, (char __user *)(argument + i))) | |
454 | return -EFAULT; | |
455 | return 0; | |
456 | ||
457 | case CDROM_GET_CAPABILITY: { | |
458 | struct gendisk *gd = info->gd; | |
459 | if (gd->flags & GENHD_FL_CD) | |
460 | return 0; | |
461 | return -EINVAL; | |
462 | } | |
463 | ||
464 | default: | |
465 | /*printk(KERN_ALERT "ioctl %08x not supported by Xen blkdev\n", | |
466 | command);*/ | |
467 | return -EINVAL; /* same return as native Linux */ | |
468 | } | |
469 | ||
470 | return 0; | |
471 | } | |
472 | ||
81f35161 | 473 | static int blkif_queue_discard_req(struct request *req, struct blkfront_ring_info *rinfo) |
9f27ee59 | 474 | { |
81f35161 | 475 | struct blkfront_info *info = rinfo->dev_info; |
9f27ee59 | 476 | struct blkif_request *ring_req; |
9f27ee59 | 477 | unsigned long id; |
33204663 JG |
478 | |
479 | /* Fill out a communications ring structure. */ | |
81f35161 BL |
480 | ring_req = RING_GET_REQUEST(&rinfo->ring, rinfo->ring.req_prod_pvt); |
481 | id = get_id_from_freelist(rinfo); | |
482 | rinfo->shadow[id].request = req; | |
33204663 JG |
483 | |
484 | ring_req->operation = BLKIF_OP_DISCARD; | |
485 | ring_req->u.discard.nr_sectors = blk_rq_sectors(req); | |
486 | ring_req->u.discard.id = id; | |
487 | ring_req->u.discard.sector_number = (blkif_sector_t)blk_rq_pos(req); | |
488 | if ((req->cmd_flags & REQ_SECURE) && info->feature_secdiscard) | |
489 | ring_req->u.discard.flag = BLKIF_DISCARD_SECURE; | |
490 | else | |
491 | ring_req->u.discard.flag = 0; | |
492 | ||
81f35161 | 493 | rinfo->ring.req_prod_pvt++; |
33204663 JG |
494 | |
495 | /* Keep a private copy so we can reissue requests when recovering. */ | |
81f35161 | 496 | rinfo->shadow[id].req = *ring_req; |
33204663 JG |
497 | |
498 | return 0; | |
499 | } | |
500 | ||
c004a6fe JG |
501 | struct setup_rw_req { |
502 | unsigned int grant_idx; | |
503 | struct blkif_request_segment *segments; | |
81f35161 | 504 | struct blkfront_ring_info *rinfo; |
c004a6fe JG |
505 | struct blkif_request *ring_req; |
506 | grant_ref_t gref_head; | |
507 | unsigned int id; | |
508 | /* Only used when persistent grant is used and it's a read request */ | |
509 | bool need_copy; | |
510 | unsigned int bvec_off; | |
511 | char *bvec_data; | |
512 | }; | |
513 | ||
514 | static void blkif_setup_rw_req_grant(unsigned long gfn, unsigned int offset, | |
515 | unsigned int len, void *data) | |
516 | { | |
517 | struct setup_rw_req *setup = data; | |
518 | int n, ref; | |
519 | struct grant *gnt_list_entry; | |
9f27ee59 | 520 | unsigned int fsect, lsect; |
c004a6fe JG |
521 | /* Convenient aliases */ |
522 | unsigned int grant_idx = setup->grant_idx; | |
523 | struct blkif_request *ring_req = setup->ring_req; | |
81f35161 BL |
524 | struct blkfront_ring_info *rinfo = setup->rinfo; |
525 | struct blkfront_info *info = rinfo->dev_info; | |
526 | struct blk_shadow *shadow = &rinfo->shadow[setup->id]; | |
c004a6fe JG |
527 | |
528 | if ((ring_req->operation == BLKIF_OP_INDIRECT) && | |
529 | (grant_idx % GRANTS_PER_INDIRECT_FRAME == 0)) { | |
530 | if (setup->segments) | |
531 | kunmap_atomic(setup->segments); | |
532 | ||
533 | n = grant_idx / GRANTS_PER_INDIRECT_FRAME; | |
534 | gnt_list_entry = get_indirect_grant(&setup->gref_head, info); | |
535 | shadow->indirect_grants[n] = gnt_list_entry; | |
536 | setup->segments = kmap_atomic(gnt_list_entry->page); | |
537 | ring_req->u.indirect.indirect_grefs[n] = gnt_list_entry->gref; | |
538 | } | |
539 | ||
540 | gnt_list_entry = get_grant(&setup->gref_head, gfn, info); | |
541 | ref = gnt_list_entry->gref; | |
542 | shadow->grants_used[grant_idx] = gnt_list_entry; | |
543 | ||
544 | if (setup->need_copy) { | |
545 | void *shared_data; | |
546 | ||
547 | shared_data = kmap_atomic(gnt_list_entry->page); | |
548 | /* | |
549 | * this does not wipe data stored outside the | |
550 | * range sg->offset..sg->offset+sg->length. | |
551 | * Therefore, blkback *could* see data from | |
552 | * previous requests. This is OK as long as | |
553 | * persistent grants are shared with just one | |
554 | * domain. It may need refactoring if this | |
555 | * changes | |
556 | */ | |
557 | memcpy(shared_data + offset, | |
558 | setup->bvec_data + setup->bvec_off, | |
559 | len); | |
560 | ||
561 | kunmap_atomic(shared_data); | |
562 | setup->bvec_off += len; | |
563 | } | |
564 | ||
565 | fsect = offset >> 9; | |
566 | lsect = fsect + (len >> 9) - 1; | |
567 | if (ring_req->operation != BLKIF_OP_INDIRECT) { | |
568 | ring_req->u.rw.seg[grant_idx] = | |
569 | (struct blkif_request_segment) { | |
570 | .gref = ref, | |
571 | .first_sect = fsect, | |
572 | .last_sect = lsect }; | |
573 | } else { | |
574 | setup->segments[grant_idx % GRANTS_PER_INDIRECT_FRAME] = | |
575 | (struct blkif_request_segment) { | |
576 | .gref = ref, | |
577 | .first_sect = fsect, | |
578 | .last_sect = lsect }; | |
579 | } | |
580 | ||
581 | (setup->grant_idx)++; | |
582 | } | |
583 | ||
81f35161 | 584 | static int blkif_queue_rw_req(struct request *req, struct blkfront_ring_info *rinfo) |
9f27ee59 | 585 | { |
81f35161 | 586 | struct blkfront_info *info = rinfo->dev_info; |
9f27ee59 | 587 | struct blkif_request *ring_req; |
9f27ee59 | 588 | unsigned long id; |
c004a6fe JG |
589 | int i; |
590 | struct setup_rw_req setup = { | |
591 | .grant_idx = 0, | |
592 | .segments = NULL, | |
81f35161 | 593 | .rinfo = rinfo, |
c004a6fe JG |
594 | .need_copy = rq_data_dir(req) && info->feature_persistent, |
595 | }; | |
0a8704a5 RPM |
596 | |
597 | /* | |
598 | * Used to store if we are able to queue the request by just using | |
599 | * existing persistent grants, or if we have to get new grants, | |
600 | * as there are not sufficiently many free. | |
601 | */ | |
9e973e64 | 602 | struct scatterlist *sg; |
c004a6fe | 603 | int num_sg, max_grefs, num_grant; |
9f27ee59 | 604 | |
c004a6fe | 605 | max_grefs = req->nr_phys_segments * GRANTS_PER_PSEG; |
c47206e2 RPM |
606 | if (max_grefs > BLKIF_MAX_SEGMENTS_PER_REQUEST) |
607 | /* | |
608 | * If we are using indirect segments we need to account | |
609 | * for the indirect grefs used in the request. | |
610 | */ | |
c004a6fe | 611 | max_grefs += INDIRECT_GREFS(max_grefs); |
402b27f9 | 612 | |
3df0e505 BL |
613 | /* |
614 | * We have to reserve 'max_grefs' grants because persistent | |
615 | * grants are shared by all rings. | |
616 | */ | |
617 | if (max_grefs > 0) | |
618 | if (gnttab_alloc_grant_references(max_grefs, &setup.gref_head) < 0) { | |
0a8704a5 | 619 | gnttab_request_free_callback( |
81f35161 | 620 | &rinfo->callback, |
0a8704a5 | 621 | blkif_restart_queue_callback, |
81f35161 | 622 | rinfo, |
402b27f9 | 623 | max_grefs); |
0a8704a5 RPM |
624 | return 1; |
625 | } | |
9f27ee59 JF |
626 | |
627 | /* Fill out a communications ring structure. */ | |
81f35161 BL |
628 | ring_req = RING_GET_REQUEST(&rinfo->ring, rinfo->ring.req_prod_pvt); |
629 | id = get_id_from_freelist(rinfo); | |
630 | rinfo->shadow[id].request = req; | |
9f27ee59 | 631 | |
33204663 | 632 | BUG_ON(info->max_indirect_segments == 0 && |
c004a6fe | 633 | GREFS(req->nr_phys_segments) > BLKIF_MAX_SEGMENTS_PER_REQUEST); |
33204663 | 634 | BUG_ON(info->max_indirect_segments && |
c004a6fe JG |
635 | GREFS(req->nr_phys_segments) > info->max_indirect_segments); |
636 | ||
81f35161 | 637 | num_sg = blk_rq_map_sg(req->q, req, rinfo->shadow[id].sg); |
c004a6fe JG |
638 | num_grant = 0; |
639 | /* Calculate the number of grant used */ | |
81f35161 | 640 | for_each_sg(rinfo->shadow[id].sg, sg, num_sg, i) |
c004a6fe JG |
641 | num_grant += gnttab_count_grant(sg->offset, sg->length); |
642 | ||
33204663 | 643 | ring_req->u.rw.id = id; |
81f35161 | 644 | rinfo->shadow[id].num_sg = num_sg; |
c004a6fe | 645 | if (num_grant > BLKIF_MAX_SEGMENTS_PER_REQUEST) { |
33204663 JG |
646 | /* |
647 | * The indirect operation can only be a BLKIF_OP_READ or | |
648 | * BLKIF_OP_WRITE | |
649 | */ | |
650 | BUG_ON(req->cmd_flags & (REQ_FLUSH | REQ_FUA)); | |
651 | ring_req->operation = BLKIF_OP_INDIRECT; | |
652 | ring_req->u.indirect.indirect_op = rq_data_dir(req) ? | |
653 | BLKIF_OP_WRITE : BLKIF_OP_READ; | |
654 | ring_req->u.indirect.sector_number = (blkif_sector_t)blk_rq_pos(req); | |
655 | ring_req->u.indirect.handle = info->handle; | |
c004a6fe | 656 | ring_req->u.indirect.nr_segments = num_grant; |
ed30bf31 | 657 | } else { |
33204663 JG |
658 | ring_req->u.rw.sector_number = (blkif_sector_t)blk_rq_pos(req); |
659 | ring_req->u.rw.handle = info->handle; | |
660 | ring_req->operation = rq_data_dir(req) ? | |
661 | BLKIF_OP_WRITE : BLKIF_OP_READ; | |
662 | if (req->cmd_flags & (REQ_FLUSH | REQ_FUA)) { | |
402b27f9 | 663 | /* |
33204663 JG |
664 | * Ideally we can do an unordered flush-to-disk. |
665 | * In case the backend onlysupports barriers, use that. | |
666 | * A barrier request a superset of FUA, so we can | |
667 | * implement it the same way. (It's also a FLUSH+FUA, | |
668 | * since it is guaranteed ordered WRT previous writes.) | |
402b27f9 | 669 | */ |
33204663 JG |
670 | switch (info->feature_flush & |
671 | ((REQ_FLUSH|REQ_FUA))) { | |
672 | case REQ_FLUSH|REQ_FUA: | |
673 | ring_req->operation = | |
674 | BLKIF_OP_WRITE_BARRIER; | |
675 | break; | |
676 | case REQ_FLUSH: | |
677 | ring_req->operation = | |
678 | BLKIF_OP_FLUSH_DISKCACHE; | |
679 | break; | |
680 | default: | |
681 | ring_req->operation = 0; | |
402b27f9 | 682 | } |
402b27f9 | 683 | } |
c004a6fe | 684 | ring_req->u.rw.nr_segments = num_grant; |
33204663 | 685 | } |
0a8704a5 | 686 | |
c004a6fe JG |
687 | setup.ring_req = ring_req; |
688 | setup.id = id; | |
81f35161 | 689 | for_each_sg(rinfo->shadow[id].sg, sg, num_sg, i) { |
c004a6fe | 690 | BUG_ON(sg->offset + sg->length > PAGE_SIZE); |
0a8704a5 | 691 | |
c004a6fe JG |
692 | if (setup.need_copy) { |
693 | setup.bvec_off = sg->offset; | |
694 | setup.bvec_data = kmap_atomic(sg_page(sg)); | |
695 | } | |
0a8704a5 | 696 | |
c004a6fe JG |
697 | gnttab_foreach_grant_in_range(sg_page(sg), |
698 | sg->offset, | |
699 | sg->length, | |
700 | blkif_setup_rw_req_grant, | |
701 | &setup); | |
0a8704a5 | 702 | |
c004a6fe JG |
703 | if (setup.need_copy) |
704 | kunmap_atomic(setup.bvec_data); | |
9f27ee59 | 705 | } |
c004a6fe JG |
706 | if (setup.segments) |
707 | kunmap_atomic(setup.segments); | |
9f27ee59 | 708 | |
81f35161 | 709 | rinfo->ring.req_prod_pvt++; |
9f27ee59 JF |
710 | |
711 | /* Keep a private copy so we can reissue requests when recovering. */ | |
81f35161 | 712 | rinfo->shadow[id].req = *ring_req; |
9f27ee59 | 713 | |
3df0e505 | 714 | if (max_grefs > 0) |
c004a6fe | 715 | gnttab_free_grant_references(setup.gref_head); |
9f27ee59 JF |
716 | |
717 | return 0; | |
718 | } | |
719 | ||
33204663 JG |
720 | /* |
721 | * Generate a Xen blkfront IO request from a blk layer request. Reads | |
722 | * and writes are handled as expected. | |
723 | * | |
724 | * @req: a request struct | |
725 | */ | |
81f35161 | 726 | static int blkif_queue_request(struct request *req, struct blkfront_ring_info *rinfo) |
33204663 | 727 | { |
81f35161 | 728 | if (unlikely(rinfo->dev_info->connected != BLKIF_STATE_CONNECTED)) |
33204663 JG |
729 | return 1; |
730 | ||
731 | if (unlikely(req->cmd_flags & (REQ_DISCARD | REQ_SECURE))) | |
81f35161 | 732 | return blkif_queue_discard_req(req, rinfo); |
33204663 | 733 | else |
81f35161 | 734 | return blkif_queue_rw_req(req, rinfo); |
33204663 | 735 | } |
9f27ee59 | 736 | |
81f35161 | 737 | static inline void flush_requests(struct blkfront_ring_info *rinfo) |
9f27ee59 JF |
738 | { |
739 | int notify; | |
740 | ||
81f35161 | 741 | RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&rinfo->ring, notify); |
9f27ee59 JF |
742 | |
743 | if (notify) | |
81f35161 | 744 | notify_remote_via_irq(rinfo->irq); |
9f27ee59 JF |
745 | } |
746 | ||
ad42d391 VK |
747 | static inline bool blkif_request_flush_invalid(struct request *req, |
748 | struct blkfront_info *info) | |
0f1ca65e AA |
749 | { |
750 | return ((req->cmd_type != REQ_TYPE_FS) || | |
ad42d391 VK |
751 | ((req->cmd_flags & REQ_FLUSH) && |
752 | !(info->feature_flush & REQ_FLUSH)) || | |
753 | ((req->cmd_flags & REQ_FUA) && | |
754 | !(info->feature_flush & REQ_FUA))); | |
0f1ca65e AA |
755 | } |
756 | ||
907c3eb1 BL |
757 | static int blkif_queue_rq(struct blk_mq_hw_ctx *hctx, |
758 | const struct blk_mq_queue_data *qd) | |
9f27ee59 | 759 | { |
81f35161 BL |
760 | struct blkfront_ring_info *rinfo = (struct blkfront_ring_info *)hctx->driver_data; |
761 | struct blkfront_info *info = rinfo->dev_info; | |
9f27ee59 | 762 | |
907c3eb1 BL |
763 | blk_mq_start_request(qd->rq); |
764 | spin_lock_irq(&info->io_lock); | |
81f35161 | 765 | if (RING_FULL(&rinfo->ring)) |
907c3eb1 | 766 | goto out_busy; |
9f27ee59 | 767 | |
81f35161 | 768 | if (blkif_request_flush_invalid(qd->rq, rinfo->dev_info)) |
907c3eb1 | 769 | goto out_err; |
296b2f6a | 770 | |
81f35161 | 771 | if (blkif_queue_request(qd->rq, rinfo)) |
907c3eb1 | 772 | goto out_busy; |
296b2f6a | 773 | |
81f35161 | 774 | flush_requests(rinfo); |
907c3eb1 BL |
775 | spin_unlock_irq(&info->io_lock); |
776 | return BLK_MQ_RQ_QUEUE_OK; | |
9f27ee59 | 777 | |
907c3eb1 BL |
778 | out_err: |
779 | spin_unlock_irq(&info->io_lock); | |
780 | return BLK_MQ_RQ_QUEUE_ERROR; | |
9f27ee59 | 781 | |
907c3eb1 BL |
782 | out_busy: |
783 | spin_unlock_irq(&info->io_lock); | |
784 | blk_mq_stop_hw_queue(hctx); | |
785 | return BLK_MQ_RQ_QUEUE_BUSY; | |
9f27ee59 JF |
786 | } |
787 | ||
81f35161 BL |
788 | static int blk_mq_init_hctx(struct blk_mq_hw_ctx *hctx, void *data, |
789 | unsigned int index) | |
790 | { | |
791 | struct blkfront_info *info = (struct blkfront_info *)data; | |
792 | ||
3df0e505 BL |
793 | BUG_ON(info->nr_rings <= index); |
794 | hctx->driver_data = &info->rinfo[index]; | |
81f35161 BL |
795 | return 0; |
796 | } | |
797 | ||
907c3eb1 BL |
798 | static struct blk_mq_ops blkfront_mq_ops = { |
799 | .queue_rq = blkif_queue_rq, | |
800 | .map_queue = blk_mq_map_queue, | |
81f35161 | 801 | .init_hctx = blk_mq_init_hctx, |
907c3eb1 BL |
802 | }; |
803 | ||
402b27f9 | 804 | static int xlvbd_init_blk_queue(struct gendisk *gd, u16 sector_size, |
7c4d7d71 | 805 | unsigned int physical_sector_size, |
402b27f9 | 806 | unsigned int segments) |
9f27ee59 | 807 | { |
165125e1 | 808 | struct request_queue *rq; |
ed30bf31 | 809 | struct blkfront_info *info = gd->private_data; |
9f27ee59 | 810 | |
907c3eb1 BL |
811 | memset(&info->tag_set, 0, sizeof(info->tag_set)); |
812 | info->tag_set.ops = &blkfront_mq_ops; | |
813 | info->tag_set.nr_hw_queues = 1; | |
814 | info->tag_set.queue_depth = BLK_RING_SIZE(info); | |
815 | info->tag_set.numa_node = NUMA_NO_NODE; | |
816 | info->tag_set.flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_SG_MERGE; | |
817 | info->tag_set.cmd_size = 0; | |
818 | info->tag_set.driver_data = info; | |
819 | ||
820 | if (blk_mq_alloc_tag_set(&info->tag_set)) | |
9f27ee59 | 821 | return -1; |
907c3eb1 BL |
822 | rq = blk_mq_init_queue(&info->tag_set); |
823 | if (IS_ERR(rq)) { | |
824 | blk_mq_free_tag_set(&info->tag_set); | |
825 | return -1; | |
826 | } | |
9f27ee59 | 827 | |
66d352e1 | 828 | queue_flag_set_unlocked(QUEUE_FLAG_VIRT, rq); |
9f27ee59 | 829 | |
ed30bf31 LD |
830 | if (info->feature_discard) { |
831 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, rq); | |
832 | blk_queue_max_discard_sectors(rq, get_capacity(gd)); | |
833 | rq->limits.discard_granularity = info->discard_granularity; | |
834 | rq->limits.discard_alignment = info->discard_alignment; | |
5ea42986 KRW |
835 | if (info->feature_secdiscard) |
836 | queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD, rq); | |
ed30bf31 LD |
837 | } |
838 | ||
9f27ee59 | 839 | /* Hard sector size and max sectors impersonate the equiv. hardware. */ |
e1defc4f | 840 | blk_queue_logical_block_size(rq, sector_size); |
7c4d7d71 | 841 | blk_queue_physical_block_size(rq, physical_sector_size); |
c004a6fe | 842 | blk_queue_max_hw_sectors(rq, (segments * XEN_PAGE_SIZE) / 512); |
9f27ee59 JF |
843 | |
844 | /* Each segment in a request is up to an aligned page in size. */ | |
845 | blk_queue_segment_boundary(rq, PAGE_SIZE - 1); | |
846 | blk_queue_max_segment_size(rq, PAGE_SIZE); | |
847 | ||
848 | /* Ensure a merged request will fit in a single I/O ring slot. */ | |
c004a6fe | 849 | blk_queue_max_segments(rq, segments / GRANTS_PER_PSEG); |
9f27ee59 JF |
850 | |
851 | /* Make sure buffer addresses are sector-aligned. */ | |
852 | blk_queue_dma_alignment(rq, 511); | |
853 | ||
1c91fe1a IC |
854 | /* Make sure we don't use bounce buffers. */ |
855 | blk_queue_bounce_limit(rq, BLK_BOUNCE_ANY); | |
856 | ||
9f27ee59 JF |
857 | gd->queue = rq; |
858 | ||
859 | return 0; | |
860 | } | |
861 | ||
fdf9b965 VK |
862 | static const char *flush_info(unsigned int feature_flush) |
863 | { | |
864 | switch (feature_flush & ((REQ_FLUSH | REQ_FUA))) { | |
865 | case REQ_FLUSH|REQ_FUA: | |
866 | return "barrier: enabled;"; | |
867 | case REQ_FLUSH: | |
868 | return "flush diskcache: enabled;"; | |
869 | default: | |
870 | return "barrier or flush: disabled;"; | |
871 | } | |
872 | } | |
9f27ee59 | 873 | |
4913efe4 | 874 | static void xlvbd_flush(struct blkfront_info *info) |
9f27ee59 | 875 | { |
4913efe4 | 876 | blk_queue_flush(info->rq, info->feature_flush); |
fdf9b965 VK |
877 | pr_info("blkfront: %s: %s %s %s %s %s\n", |
878 | info->gd->disk_name, flush_info(info->feature_flush), | |
879 | "persistent grants:", info->feature_persistent ? | |
880 | "enabled;" : "disabled;", "indirect descriptors:", | |
881 | info->max_indirect_segments ? "enabled;" : "disabled;"); | |
9f27ee59 JF |
882 | } |
883 | ||
c80a4209 SS |
884 | static int xen_translate_vdev(int vdevice, int *minor, unsigned int *offset) |
885 | { | |
886 | int major; | |
887 | major = BLKIF_MAJOR(vdevice); | |
888 | *minor = BLKIF_MINOR(vdevice); | |
889 | switch (major) { | |
890 | case XEN_IDE0_MAJOR: | |
891 | *offset = (*minor / 64) + EMULATED_HD_DISK_NAME_OFFSET; | |
892 | *minor = ((*minor / 64) * PARTS_PER_DISK) + | |
893 | EMULATED_HD_DISK_MINOR_OFFSET; | |
894 | break; | |
895 | case XEN_IDE1_MAJOR: | |
896 | *offset = (*minor / 64) + 2 + EMULATED_HD_DISK_NAME_OFFSET; | |
897 | *minor = (((*minor / 64) + 2) * PARTS_PER_DISK) + | |
898 | EMULATED_HD_DISK_MINOR_OFFSET; | |
899 | break; | |
900 | case XEN_SCSI_DISK0_MAJOR: | |
901 | *offset = (*minor / PARTS_PER_DISK) + EMULATED_SD_DISK_NAME_OFFSET; | |
902 | *minor = *minor + EMULATED_SD_DISK_MINOR_OFFSET; | |
903 | break; | |
904 | case XEN_SCSI_DISK1_MAJOR: | |
905 | case XEN_SCSI_DISK2_MAJOR: | |
906 | case XEN_SCSI_DISK3_MAJOR: | |
907 | case XEN_SCSI_DISK4_MAJOR: | |
908 | case XEN_SCSI_DISK5_MAJOR: | |
909 | case XEN_SCSI_DISK6_MAJOR: | |
910 | case XEN_SCSI_DISK7_MAJOR: | |
911 | *offset = (*minor / PARTS_PER_DISK) + | |
912 | ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16) + | |
913 | EMULATED_SD_DISK_NAME_OFFSET; | |
914 | *minor = *minor + | |
915 | ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16 * PARTS_PER_DISK) + | |
916 | EMULATED_SD_DISK_MINOR_OFFSET; | |
917 | break; | |
918 | case XEN_SCSI_DISK8_MAJOR: | |
919 | case XEN_SCSI_DISK9_MAJOR: | |
920 | case XEN_SCSI_DISK10_MAJOR: | |
921 | case XEN_SCSI_DISK11_MAJOR: | |
922 | case XEN_SCSI_DISK12_MAJOR: | |
923 | case XEN_SCSI_DISK13_MAJOR: | |
924 | case XEN_SCSI_DISK14_MAJOR: | |
925 | case XEN_SCSI_DISK15_MAJOR: | |
926 | *offset = (*minor / PARTS_PER_DISK) + | |
927 | ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16) + | |
928 | EMULATED_SD_DISK_NAME_OFFSET; | |
929 | *minor = *minor + | |
930 | ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16 * PARTS_PER_DISK) + | |
931 | EMULATED_SD_DISK_MINOR_OFFSET; | |
932 | break; | |
933 | case XENVBD_MAJOR: | |
934 | *offset = *minor / PARTS_PER_DISK; | |
935 | break; | |
936 | default: | |
937 | printk(KERN_WARNING "blkfront: your disk configuration is " | |
938 | "incorrect, please use an xvd device instead\n"); | |
939 | return -ENODEV; | |
940 | } | |
941 | return 0; | |
942 | } | |
9f27ee59 | 943 | |
e77c78c0 JB |
944 | static char *encode_disk_name(char *ptr, unsigned int n) |
945 | { | |
946 | if (n >= 26) | |
947 | ptr = encode_disk_name(ptr, n / 26 - 1); | |
948 | *ptr = 'a' + n % 26; | |
949 | return ptr + 1; | |
950 | } | |
951 | ||
9246b5f0 CL |
952 | static int xlvbd_alloc_gendisk(blkif_sector_t capacity, |
953 | struct blkfront_info *info, | |
7c4d7d71 SB |
954 | u16 vdisk_info, u16 sector_size, |
955 | unsigned int physical_sector_size) | |
9f27ee59 JF |
956 | { |
957 | struct gendisk *gd; | |
958 | int nr_minors = 1; | |
c80a4209 | 959 | int err; |
9246b5f0 CL |
960 | unsigned int offset; |
961 | int minor; | |
962 | int nr_parts; | |
e77c78c0 | 963 | char *ptr; |
9f27ee59 JF |
964 | |
965 | BUG_ON(info->gd != NULL); | |
966 | BUG_ON(info->rq != NULL); | |
967 | ||
9246b5f0 CL |
968 | if ((info->vdevice>>EXT_SHIFT) > 1) { |
969 | /* this is above the extended range; something is wrong */ | |
970 | printk(KERN_WARNING "blkfront: vdevice 0x%x is above the extended range; ignoring\n", info->vdevice); | |
971 | return -ENODEV; | |
972 | } | |
973 | ||
974 | if (!VDEV_IS_EXTENDED(info->vdevice)) { | |
c80a4209 SS |
975 | err = xen_translate_vdev(info->vdevice, &minor, &offset); |
976 | if (err) | |
977 | return err; | |
978 | nr_parts = PARTS_PER_DISK; | |
9246b5f0 CL |
979 | } else { |
980 | minor = BLKIF_MINOR_EXT(info->vdevice); | |
981 | nr_parts = PARTS_PER_EXT_DISK; | |
c80a4209 | 982 | offset = minor / nr_parts; |
89153b5c | 983 | if (xen_hvm_domain() && offset < EMULATED_HD_DISK_NAME_OFFSET + 4) |
c80a4209 SS |
984 | printk(KERN_WARNING "blkfront: vdevice 0x%x might conflict with " |
985 | "emulated IDE disks,\n\t choose an xvd device name" | |
986 | "from xvde on\n", info->vdevice); | |
9246b5f0 | 987 | } |
e77c78c0 JB |
988 | if (minor >> MINORBITS) { |
989 | pr_warn("blkfront: %#x's minor (%#x) out of range; ignoring\n", | |
990 | info->vdevice, minor); | |
991 | return -ENODEV; | |
992 | } | |
9246b5f0 CL |
993 | |
994 | if ((minor % nr_parts) == 0) | |
995 | nr_minors = nr_parts; | |
9f27ee59 | 996 | |
0e345826 JB |
997 | err = xlbd_reserve_minors(minor, nr_minors); |
998 | if (err) | |
999 | goto out; | |
1000 | err = -ENODEV; | |
1001 | ||
9f27ee59 JF |
1002 | gd = alloc_disk(nr_minors); |
1003 | if (gd == NULL) | |
0e345826 | 1004 | goto release; |
9f27ee59 | 1005 | |
e77c78c0 JB |
1006 | strcpy(gd->disk_name, DEV_NAME); |
1007 | ptr = encode_disk_name(gd->disk_name + sizeof(DEV_NAME) - 1, offset); | |
1008 | BUG_ON(ptr >= gd->disk_name + DISK_NAME_LEN); | |
1009 | if (nr_minors > 1) | |
1010 | *ptr = 0; | |
1011 | else | |
1012 | snprintf(ptr, gd->disk_name + DISK_NAME_LEN - ptr, | |
1013 | "%d", minor & (nr_parts - 1)); | |
9f27ee59 JF |
1014 | |
1015 | gd->major = XENVBD_MAJOR; | |
1016 | gd->first_minor = minor; | |
1017 | gd->fops = &xlvbd_block_fops; | |
1018 | gd->private_data = info; | |
1019 | gd->driverfs_dev = &(info->xbdev->dev); | |
1020 | set_capacity(gd, capacity); | |
1021 | ||
7c4d7d71 | 1022 | if (xlvbd_init_blk_queue(gd, sector_size, physical_sector_size, |
402b27f9 RPM |
1023 | info->max_indirect_segments ? : |
1024 | BLKIF_MAX_SEGMENTS_PER_REQUEST)) { | |
9f27ee59 | 1025 | del_gendisk(gd); |
0e345826 | 1026 | goto release; |
9f27ee59 JF |
1027 | } |
1028 | ||
1029 | info->rq = gd->queue; | |
1030 | info->gd = gd; | |
1031 | ||
4913efe4 | 1032 | xlvbd_flush(info); |
9f27ee59 JF |
1033 | |
1034 | if (vdisk_info & VDISK_READONLY) | |
1035 | set_disk_ro(gd, 1); | |
1036 | ||
1037 | if (vdisk_info & VDISK_REMOVABLE) | |
1038 | gd->flags |= GENHD_FL_REMOVABLE; | |
1039 | ||
1040 | if (vdisk_info & VDISK_CDROM) | |
1041 | gd->flags |= GENHD_FL_CD; | |
1042 | ||
1043 | return 0; | |
1044 | ||
0e345826 JB |
1045 | release: |
1046 | xlbd_release_minors(minor, nr_minors); | |
9f27ee59 JF |
1047 | out: |
1048 | return err; | |
1049 | } | |
1050 | ||
a66b5aeb DS |
1051 | static void xlvbd_release_gendisk(struct blkfront_info *info) |
1052 | { | |
3df0e505 | 1053 | unsigned int minor, nr_minors, i; |
a66b5aeb DS |
1054 | |
1055 | if (info->rq == NULL) | |
1056 | return; | |
1057 | ||
a66b5aeb | 1058 | /* No more blkif_request(). */ |
907c3eb1 | 1059 | blk_mq_stop_hw_queues(info->rq); |
a66b5aeb | 1060 | |
3df0e505 BL |
1061 | for (i = 0; i < info->nr_rings; i++) { |
1062 | struct blkfront_ring_info *rinfo = &info->rinfo[i]; | |
a66b5aeb | 1063 | |
3df0e505 BL |
1064 | /* No more gnttab callback work. */ |
1065 | gnttab_cancel_free_callback(&rinfo->callback); | |
1066 | ||
1067 | /* Flush gnttab callback work. Must be done with no locks held. */ | |
1068 | flush_work(&rinfo->work); | |
1069 | } | |
a66b5aeb DS |
1070 | |
1071 | del_gendisk(info->gd); | |
1072 | ||
1073 | minor = info->gd->first_minor; | |
1074 | nr_minors = info->gd->minors; | |
1075 | xlbd_release_minors(minor, nr_minors); | |
1076 | ||
1077 | blk_cleanup_queue(info->rq); | |
907c3eb1 | 1078 | blk_mq_free_tag_set(&info->tag_set); |
a66b5aeb DS |
1079 | info->rq = NULL; |
1080 | ||
1081 | put_disk(info->gd); | |
1082 | info->gd = NULL; | |
1083 | } | |
1084 | ||
907c3eb1 | 1085 | /* Must be called with io_lock holded */ |
81f35161 | 1086 | static void kick_pending_request_queues(struct blkfront_ring_info *rinfo) |
9f27ee59 | 1087 | { |
81f35161 BL |
1088 | if (!RING_FULL(&rinfo->ring)) |
1089 | blk_mq_start_stopped_hw_queues(rinfo->dev_info->rq, true); | |
9f27ee59 JF |
1090 | } |
1091 | ||
1092 | static void blkif_restart_queue(struct work_struct *work) | |
1093 | { | |
81f35161 | 1094 | struct blkfront_ring_info *rinfo = container_of(work, struct blkfront_ring_info, work); |
9f27ee59 | 1095 | |
81f35161 BL |
1096 | spin_lock_irq(&rinfo->dev_info->io_lock); |
1097 | if (rinfo->dev_info->connected == BLKIF_STATE_CONNECTED) | |
1098 | kick_pending_request_queues(rinfo); | |
1099 | spin_unlock_irq(&rinfo->dev_info->io_lock); | |
9f27ee59 JF |
1100 | } |
1101 | ||
3df0e505 | 1102 | static void blkif_free_ring(struct blkfront_ring_info *rinfo) |
9f27ee59 | 1103 | { |
155b7edb | 1104 | struct grant *persistent_gnt; |
3df0e505 | 1105 | struct blkfront_info *info = rinfo->dev_info; |
402b27f9 | 1106 | int i, j, segs; |
0a8704a5 | 1107 | |
bfe11d6d RPM |
1108 | /* |
1109 | * Remove indirect pages, this only happens when using indirect | |
1110 | * descriptors but not persistent grants | |
1111 | */ | |
81f35161 | 1112 | if (!list_empty(&rinfo->indirect_pages)) { |
bfe11d6d RPM |
1113 | struct page *indirect_page, *n; |
1114 | ||
1115 | BUG_ON(info->feature_persistent); | |
81f35161 | 1116 | list_for_each_entry_safe(indirect_page, n, &rinfo->indirect_pages, lru) { |
bfe11d6d RPM |
1117 | list_del(&indirect_page->lru); |
1118 | __free_page(indirect_page); | |
1119 | } | |
1120 | } | |
1121 | ||
86839c56 | 1122 | for (i = 0; i < BLK_RING_SIZE(info); i++) { |
402b27f9 RPM |
1123 | /* |
1124 | * Clear persistent grants present in requests already | |
1125 | * on the shared ring | |
1126 | */ | |
81f35161 | 1127 | if (!rinfo->shadow[i].request) |
402b27f9 RPM |
1128 | goto free_shadow; |
1129 | ||
81f35161 BL |
1130 | segs = rinfo->shadow[i].req.operation == BLKIF_OP_INDIRECT ? |
1131 | rinfo->shadow[i].req.u.indirect.nr_segments : | |
1132 | rinfo->shadow[i].req.u.rw.nr_segments; | |
402b27f9 | 1133 | for (j = 0; j < segs; j++) { |
81f35161 | 1134 | persistent_gnt = rinfo->shadow[i].grants_used[j]; |
402b27f9 | 1135 | gnttab_end_foreign_access(persistent_gnt->gref, 0, 0UL); |
bfe11d6d | 1136 | if (info->feature_persistent) |
a7a6df22 | 1137 | __free_page(persistent_gnt->page); |
402b27f9 RPM |
1138 | kfree(persistent_gnt); |
1139 | } | |
1140 | ||
81f35161 | 1141 | if (rinfo->shadow[i].req.operation != BLKIF_OP_INDIRECT) |
402b27f9 RPM |
1142 | /* |
1143 | * If this is not an indirect operation don't try to | |
1144 | * free indirect segments | |
1145 | */ | |
1146 | goto free_shadow; | |
1147 | ||
1148 | for (j = 0; j < INDIRECT_GREFS(segs); j++) { | |
81f35161 | 1149 | persistent_gnt = rinfo->shadow[i].indirect_grants[j]; |
402b27f9 | 1150 | gnttab_end_foreign_access(persistent_gnt->gref, 0, 0UL); |
a7a6df22 | 1151 | __free_page(persistent_gnt->page); |
402b27f9 RPM |
1152 | kfree(persistent_gnt); |
1153 | } | |
1154 | ||
1155 | free_shadow: | |
81f35161 BL |
1156 | kfree(rinfo->shadow[i].grants_used); |
1157 | rinfo->shadow[i].grants_used = NULL; | |
1158 | kfree(rinfo->shadow[i].indirect_grants); | |
1159 | rinfo->shadow[i].indirect_grants = NULL; | |
1160 | kfree(rinfo->shadow[i].sg); | |
1161 | rinfo->shadow[i].sg = NULL; | |
402b27f9 RPM |
1162 | } |
1163 | ||
9f27ee59 | 1164 | /* No more gnttab callback work. */ |
81f35161 | 1165 | gnttab_cancel_free_callback(&rinfo->callback); |
9f27ee59 JF |
1166 | |
1167 | /* Flush gnttab callback work. Must be done with no locks held. */ | |
81f35161 | 1168 | flush_work(&rinfo->work); |
9f27ee59 JF |
1169 | |
1170 | /* Free resources associated with old device channel. */ | |
86839c56 | 1171 | for (i = 0; i < info->nr_ring_pages; i++) { |
81f35161 BL |
1172 | if (rinfo->ring_ref[i] != GRANT_INVALID_REF) { |
1173 | gnttab_end_foreign_access(rinfo->ring_ref[i], 0, 0); | |
1174 | rinfo->ring_ref[i] = GRANT_INVALID_REF; | |
86839c56 | 1175 | } |
9f27ee59 | 1176 | } |
81f35161 BL |
1177 | free_pages((unsigned long)rinfo->ring.sring, get_order(info->nr_ring_pages * PAGE_SIZE)); |
1178 | rinfo->ring.sring = NULL; | |
86839c56 | 1179 | |
81f35161 BL |
1180 | if (rinfo->irq) |
1181 | unbind_from_irqhandler(rinfo->irq, rinfo); | |
1182 | rinfo->evtchn = rinfo->irq = 0; | |
3df0e505 | 1183 | } |
9f27ee59 | 1184 | |
3df0e505 BL |
1185 | static void blkif_free(struct blkfront_info *info, int suspend) |
1186 | { | |
1187 | struct grant *persistent_gnt, *n; | |
1188 | unsigned int i; | |
1189 | ||
1190 | /* Prevent new requests being issued until we fix things up. */ | |
1191 | spin_lock_irq(&info->io_lock); | |
1192 | info->connected = suspend ? | |
1193 | BLKIF_STATE_SUSPENDED : BLKIF_STATE_DISCONNECTED; | |
1194 | /* No more blkif_request(). */ | |
1195 | if (info->rq) | |
1196 | blk_mq_stop_hw_queues(info->rq); | |
1197 | ||
1198 | /* Remove all persistent grants */ | |
1199 | if (!list_empty(&info->grants)) { | |
1200 | list_for_each_entry_safe(persistent_gnt, n, | |
1201 | &info->grants, node) { | |
1202 | list_del(&persistent_gnt->node); | |
1203 | if (persistent_gnt->gref != GRANT_INVALID_REF) { | |
1204 | gnttab_end_foreign_access(persistent_gnt->gref, | |
1205 | 0, 0UL); | |
1206 | info->persistent_gnts_c--; | |
1207 | } | |
1208 | if (info->feature_persistent) | |
1209 | __free_page(persistent_gnt->page); | |
1210 | kfree(persistent_gnt); | |
1211 | } | |
1212 | } | |
1213 | BUG_ON(info->persistent_gnts_c != 0); | |
1214 | ||
1215 | for (i = 0; i < info->nr_rings; i++) | |
1216 | blkif_free_ring(&info->rinfo[i]); | |
1217 | ||
1218 | kfree(info->rinfo); | |
1219 | info->rinfo = NULL; | |
1220 | info->nr_rings = 0; | |
1221 | spin_unlock_irq(&info->io_lock); | |
9f27ee59 JF |
1222 | } |
1223 | ||
c004a6fe JG |
1224 | struct copy_from_grant { |
1225 | const struct blk_shadow *s; | |
1226 | unsigned int grant_idx; | |
1227 | unsigned int bvec_offset; | |
1228 | char *bvec_data; | |
1229 | }; | |
1230 | ||
1231 | static void blkif_copy_from_grant(unsigned long gfn, unsigned int offset, | |
1232 | unsigned int len, void *data) | |
1233 | { | |
1234 | struct copy_from_grant *info = data; | |
1235 | char *shared_data; | |
1236 | /* Convenient aliases */ | |
1237 | const struct blk_shadow *s = info->s; | |
1238 | ||
1239 | shared_data = kmap_atomic(s->grants_used[info->grant_idx]->page); | |
1240 | ||
1241 | memcpy(info->bvec_data + info->bvec_offset, | |
1242 | shared_data + offset, len); | |
1243 | ||
1244 | info->bvec_offset += len; | |
1245 | info->grant_idx++; | |
1246 | ||
1247 | kunmap_atomic(shared_data); | |
1248 | } | |
1249 | ||
81f35161 | 1250 | static void blkif_completion(struct blk_shadow *s, struct blkfront_ring_info *rinfo, |
0a8704a5 | 1251 | struct blkif_response *bret) |
9f27ee59 | 1252 | { |
d62f6918 | 1253 | int i = 0; |
b7649158 | 1254 | struct scatterlist *sg; |
c004a6fe | 1255 | int num_sg, num_grant; |
81f35161 | 1256 | struct blkfront_info *info = rinfo->dev_info; |
c004a6fe JG |
1257 | struct copy_from_grant data = { |
1258 | .s = s, | |
1259 | .grant_idx = 0, | |
1260 | }; | |
402b27f9 | 1261 | |
c004a6fe | 1262 | num_grant = s->req.operation == BLKIF_OP_INDIRECT ? |
402b27f9 | 1263 | s->req.u.indirect.nr_segments : s->req.u.rw.nr_segments; |
c004a6fe | 1264 | num_sg = s->num_sg; |
0a8704a5 | 1265 | |
bfe11d6d | 1266 | if (bret->operation == BLKIF_OP_READ && info->feature_persistent) { |
c004a6fe | 1267 | for_each_sg(s->sg, sg, num_sg, i) { |
b7649158 | 1268 | BUG_ON(sg->offset + sg->length > PAGE_SIZE); |
c004a6fe JG |
1269 | |
1270 | data.bvec_offset = sg->offset; | |
1271 | data.bvec_data = kmap_atomic(sg_page(sg)); | |
1272 | ||
1273 | gnttab_foreach_grant_in_range(sg_page(sg), | |
1274 | sg->offset, | |
1275 | sg->length, | |
1276 | blkif_copy_from_grant, | |
1277 | &data); | |
1278 | ||
1279 | kunmap_atomic(data.bvec_data); | |
0a8704a5 RPM |
1280 | } |
1281 | } | |
1282 | /* Add the persistent grant into the list of free grants */ | |
c004a6fe | 1283 | for (i = 0; i < num_grant; i++) { |
fbe363c4 RPM |
1284 | if (gnttab_query_foreign_access(s->grants_used[i]->gref)) { |
1285 | /* | |
1286 | * If the grant is still mapped by the backend (the | |
1287 | * backend has chosen to make this grant persistent) | |
1288 | * we add it at the head of the list, so it will be | |
1289 | * reused first. | |
1290 | */ | |
bfe11d6d RPM |
1291 | if (!info->feature_persistent) |
1292 | pr_alert_ratelimited("backed has not unmapped grant: %u\n", | |
1293 | s->grants_used[i]->gref); | |
1294 | list_add(&s->grants_used[i]->node, &info->grants); | |
fbe363c4 RPM |
1295 | info->persistent_gnts_c++; |
1296 | } else { | |
1297 | /* | |
1298 | * If the grant is not mapped by the backend we end the | |
1299 | * foreign access and add it to the tail of the list, | |
1300 | * so it will not be picked again unless we run out of | |
1301 | * persistent grants. | |
1302 | */ | |
1303 | gnttab_end_foreign_access(s->grants_used[i]->gref, 0, 0UL); | |
1304 | s->grants_used[i]->gref = GRANT_INVALID_REF; | |
bfe11d6d | 1305 | list_add_tail(&s->grants_used[i]->node, &info->grants); |
fbe363c4 | 1306 | } |
0a8704a5 | 1307 | } |
402b27f9 | 1308 | if (s->req.operation == BLKIF_OP_INDIRECT) { |
c004a6fe | 1309 | for (i = 0; i < INDIRECT_GREFS(num_grant); i++) { |
fbe363c4 | 1310 | if (gnttab_query_foreign_access(s->indirect_grants[i]->gref)) { |
bfe11d6d RPM |
1311 | if (!info->feature_persistent) |
1312 | pr_alert_ratelimited("backed has not unmapped grant: %u\n", | |
1313 | s->indirect_grants[i]->gref); | |
1314 | list_add(&s->indirect_grants[i]->node, &info->grants); | |
fbe363c4 RPM |
1315 | info->persistent_gnts_c++; |
1316 | } else { | |
bfe11d6d RPM |
1317 | struct page *indirect_page; |
1318 | ||
fbe363c4 | 1319 | gnttab_end_foreign_access(s->indirect_grants[i]->gref, 0, 0UL); |
bfe11d6d RPM |
1320 | /* |
1321 | * Add the used indirect page back to the list of | |
1322 | * available pages for indirect grefs. | |
1323 | */ | |
7b076750 | 1324 | if (!info->feature_persistent) { |
a7a6df22 | 1325 | indirect_page = s->indirect_grants[i]->page; |
81f35161 | 1326 | list_add(&indirect_page->lru, &rinfo->indirect_pages); |
7b076750 | 1327 | } |
fbe363c4 | 1328 | s->indirect_grants[i]->gref = GRANT_INVALID_REF; |
bfe11d6d | 1329 | list_add_tail(&s->indirect_grants[i]->node, &info->grants); |
fbe363c4 | 1330 | } |
402b27f9 RPM |
1331 | } |
1332 | } | |
9f27ee59 JF |
1333 | } |
1334 | ||
1335 | static irqreturn_t blkif_interrupt(int irq, void *dev_id) | |
1336 | { | |
1337 | struct request *req; | |
1338 | struct blkif_response *bret; | |
1339 | RING_IDX i, rp; | |
1340 | unsigned long flags; | |
81f35161 BL |
1341 | struct blkfront_ring_info *rinfo = (struct blkfront_ring_info *)dev_id; |
1342 | struct blkfront_info *info = rinfo->dev_info; | |
f4829a9b | 1343 | int error; |
9f27ee59 | 1344 | |
3467811e | 1345 | spin_lock_irqsave(&info->io_lock, flags); |
9f27ee59 JF |
1346 | |
1347 | if (unlikely(info->connected != BLKIF_STATE_CONNECTED)) { | |
3467811e | 1348 | spin_unlock_irqrestore(&info->io_lock, flags); |
9f27ee59 JF |
1349 | return IRQ_HANDLED; |
1350 | } | |
1351 | ||
1352 | again: | |
81f35161 | 1353 | rp = rinfo->ring.sring->rsp_prod; |
9f27ee59 JF |
1354 | rmb(); /* Ensure we see queued responses up to 'rp'. */ |
1355 | ||
81f35161 | 1356 | for (i = rinfo->ring.rsp_cons; i != rp; i++) { |
9f27ee59 | 1357 | unsigned long id; |
9f27ee59 | 1358 | |
81f35161 | 1359 | bret = RING_GET_RESPONSE(&rinfo->ring, i); |
9f27ee59 | 1360 | id = bret->id; |
6878c32e KRW |
1361 | /* |
1362 | * The backend has messed up and given us an id that we would | |
1363 | * never have given to it (we stamp it up to BLK_RING_SIZE - | |
1364 | * look in get_id_from_freelist. | |
1365 | */ | |
86839c56 | 1366 | if (id >= BLK_RING_SIZE(info)) { |
6878c32e KRW |
1367 | WARN(1, "%s: response to %s has incorrect id (%ld)\n", |
1368 | info->gd->disk_name, op_name(bret->operation), id); | |
1369 | /* We can't safely get the 'struct request' as | |
1370 | * the id is busted. */ | |
1371 | continue; | |
1372 | } | |
81f35161 | 1373 | req = rinfo->shadow[id].request; |
9f27ee59 | 1374 | |
5ea42986 | 1375 | if (bret->operation != BLKIF_OP_DISCARD) |
81f35161 | 1376 | blkif_completion(&rinfo->shadow[id], rinfo, bret); |
9f27ee59 | 1377 | |
81f35161 | 1378 | if (add_id_to_freelist(rinfo, id)) { |
6878c32e KRW |
1379 | WARN(1, "%s: response to %s (id %ld) couldn't be recycled!\n", |
1380 | info->gd->disk_name, op_name(bret->operation), id); | |
1381 | continue; | |
1382 | } | |
9f27ee59 | 1383 | |
f4829a9b | 1384 | error = (bret->status == BLKIF_RSP_OKAY) ? 0 : -EIO; |
9f27ee59 | 1385 | switch (bret->operation) { |
ed30bf31 LD |
1386 | case BLKIF_OP_DISCARD: |
1387 | if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) { | |
1388 | struct request_queue *rq = info->rq; | |
6878c32e KRW |
1389 | printk(KERN_WARNING "blkfront: %s: %s op failed\n", |
1390 | info->gd->disk_name, op_name(bret->operation)); | |
f4829a9b | 1391 | error = -EOPNOTSUPP; |
ed30bf31 | 1392 | info->feature_discard = 0; |
5ea42986 | 1393 | info->feature_secdiscard = 0; |
ed30bf31 | 1394 | queue_flag_clear(QUEUE_FLAG_DISCARD, rq); |
5ea42986 | 1395 | queue_flag_clear(QUEUE_FLAG_SECDISCARD, rq); |
ed30bf31 | 1396 | } |
f4829a9b | 1397 | blk_mq_complete_request(req, error); |
ed30bf31 | 1398 | break; |
edf6ef59 | 1399 | case BLKIF_OP_FLUSH_DISKCACHE: |
9f27ee59 JF |
1400 | case BLKIF_OP_WRITE_BARRIER: |
1401 | if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) { | |
6878c32e KRW |
1402 | printk(KERN_WARNING "blkfront: %s: %s op failed\n", |
1403 | info->gd->disk_name, op_name(bret->operation)); | |
f4829a9b | 1404 | error = -EOPNOTSUPP; |
dcb8baec JF |
1405 | } |
1406 | if (unlikely(bret->status == BLKIF_RSP_ERROR && | |
81f35161 | 1407 | rinfo->shadow[id].req.u.rw.nr_segments == 0)) { |
6878c32e KRW |
1408 | printk(KERN_WARNING "blkfront: %s: empty %s op failed\n", |
1409 | info->gd->disk_name, op_name(bret->operation)); | |
f4829a9b | 1410 | error = -EOPNOTSUPP; |
dcb8baec | 1411 | } |
f4829a9b CH |
1412 | if (unlikely(error)) { |
1413 | if (error == -EOPNOTSUPP) | |
1414 | error = 0; | |
4913efe4 TH |
1415 | info->feature_flush = 0; |
1416 | xlvbd_flush(info); | |
9f27ee59 JF |
1417 | } |
1418 | /* fall through */ | |
1419 | case BLKIF_OP_READ: | |
1420 | case BLKIF_OP_WRITE: | |
1421 | if (unlikely(bret->status != BLKIF_RSP_OKAY)) | |
1422 | dev_dbg(&info->xbdev->dev, "Bad return from blkdev data " | |
1423 | "request: %x\n", bret->status); | |
1424 | ||
f4829a9b | 1425 | blk_mq_complete_request(req, error); |
9f27ee59 JF |
1426 | break; |
1427 | default: | |
1428 | BUG(); | |
1429 | } | |
1430 | } | |
1431 | ||
81f35161 | 1432 | rinfo->ring.rsp_cons = i; |
9f27ee59 | 1433 | |
81f35161 | 1434 | if (i != rinfo->ring.req_prod_pvt) { |
9f27ee59 | 1435 | int more_to_do; |
81f35161 | 1436 | RING_FINAL_CHECK_FOR_RESPONSES(&rinfo->ring, more_to_do); |
9f27ee59 JF |
1437 | if (more_to_do) |
1438 | goto again; | |
1439 | } else | |
81f35161 | 1440 | rinfo->ring.sring->rsp_event = i + 1; |
9f27ee59 | 1441 | |
81f35161 | 1442 | kick_pending_request_queues(rinfo); |
9f27ee59 | 1443 | |
3467811e | 1444 | spin_unlock_irqrestore(&info->io_lock, flags); |
9f27ee59 JF |
1445 | |
1446 | return IRQ_HANDLED; | |
1447 | } | |
1448 | ||
1449 | ||
1450 | static int setup_blkring(struct xenbus_device *dev, | |
81f35161 | 1451 | struct blkfront_ring_info *rinfo) |
9f27ee59 JF |
1452 | { |
1453 | struct blkif_sring *sring; | |
86839c56 | 1454 | int err, i; |
81f35161 | 1455 | struct blkfront_info *info = rinfo->dev_info; |
c004a6fe | 1456 | unsigned long ring_size = info->nr_ring_pages * XEN_PAGE_SIZE; |
9cce2914 | 1457 | grant_ref_t gref[XENBUS_MAX_RING_GRANTS]; |
9f27ee59 | 1458 | |
86839c56 | 1459 | for (i = 0; i < info->nr_ring_pages; i++) |
81f35161 | 1460 | rinfo->ring_ref[i] = GRANT_INVALID_REF; |
9f27ee59 | 1461 | |
86839c56 BL |
1462 | sring = (struct blkif_sring *)__get_free_pages(GFP_NOIO | __GFP_HIGH, |
1463 | get_order(ring_size)); | |
9f27ee59 JF |
1464 | if (!sring) { |
1465 | xenbus_dev_fatal(dev, -ENOMEM, "allocating shared ring"); | |
1466 | return -ENOMEM; | |
1467 | } | |
1468 | SHARED_RING_INIT(sring); | |
81f35161 | 1469 | FRONT_RING_INIT(&rinfo->ring, sring, ring_size); |
9e973e64 | 1470 | |
81f35161 | 1471 | err = xenbus_grant_ring(dev, rinfo->ring.sring, info->nr_ring_pages, gref); |
9f27ee59 | 1472 | if (err < 0) { |
86839c56 | 1473 | free_pages((unsigned long)sring, get_order(ring_size)); |
81f35161 | 1474 | rinfo->ring.sring = NULL; |
9f27ee59 JF |
1475 | goto fail; |
1476 | } | |
86839c56 | 1477 | for (i = 0; i < info->nr_ring_pages; i++) |
81f35161 | 1478 | rinfo->ring_ref[i] = gref[i]; |
9f27ee59 | 1479 | |
81f35161 | 1480 | err = xenbus_alloc_evtchn(dev, &rinfo->evtchn); |
9f27ee59 JF |
1481 | if (err) |
1482 | goto fail; | |
1483 | ||
81f35161 BL |
1484 | err = bind_evtchn_to_irqhandler(rinfo->evtchn, blkif_interrupt, 0, |
1485 | "blkif", rinfo); | |
9f27ee59 JF |
1486 | if (err <= 0) { |
1487 | xenbus_dev_fatal(dev, err, | |
1488 | "bind_evtchn_to_irqhandler failed"); | |
1489 | goto fail; | |
1490 | } | |
81f35161 | 1491 | rinfo->irq = err; |
9f27ee59 JF |
1492 | |
1493 | return 0; | |
1494 | fail: | |
1495 | blkif_free(info, 0); | |
1496 | return err; | |
1497 | } | |
1498 | ||
1499 | ||
1500 | /* Common code used when first setting up, and when resuming. */ | |
203fd61f | 1501 | static int talk_to_blkback(struct xenbus_device *dev, |
9f27ee59 JF |
1502 | struct blkfront_info *info) |
1503 | { | |
1504 | const char *message = NULL; | |
1505 | struct xenbus_transaction xbt; | |
86839c56 BL |
1506 | int err, i; |
1507 | unsigned int max_page_order = 0; | |
1508 | unsigned int ring_page_order = 0; | |
3df0e505 | 1509 | struct blkfront_ring_info *rinfo; |
86839c56 BL |
1510 | |
1511 | err = xenbus_scanf(XBT_NIL, info->xbdev->otherend, | |
1512 | "max-ring-page-order", "%u", &max_page_order); | |
1513 | if (err != 1) | |
1514 | info->nr_ring_pages = 1; | |
1515 | else { | |
1516 | ring_page_order = min(xen_blkif_max_ring_order, max_page_order); | |
1517 | info->nr_ring_pages = 1 << ring_page_order; | |
1518 | } | |
9f27ee59 | 1519 | |
3df0e505 BL |
1520 | for (i = 0; i < info->nr_rings; i++) { |
1521 | rinfo = &info->rinfo[i]; | |
1522 | /* Create shared ring, alloc event channel. */ | |
1523 | err = setup_blkring(dev, rinfo); | |
1524 | if (err) | |
1525 | goto destroy_blkring; | |
1526 | } | |
9f27ee59 JF |
1527 | |
1528 | again: | |
1529 | err = xenbus_transaction_start(&xbt); | |
1530 | if (err) { | |
1531 | xenbus_dev_fatal(dev, err, "starting transaction"); | |
1532 | goto destroy_blkring; | |
1533 | } | |
1534 | ||
3df0e505 BL |
1535 | if (info->nr_rings == 1) { |
1536 | rinfo = &info->rinfo[0]; | |
1537 | if (info->nr_ring_pages == 1) { | |
1538 | err = xenbus_printf(xbt, dev->nodename, | |
1539 | "ring-ref", "%u", rinfo->ring_ref[0]); | |
86839c56 BL |
1540 | if (err) { |
1541 | message = "writing ring-ref"; | |
1542 | goto abort_transaction; | |
1543 | } | |
3df0e505 BL |
1544 | } else { |
1545 | err = xenbus_printf(xbt, dev->nodename, | |
1546 | "ring-page-order", "%u", ring_page_order); | |
1547 | if (err) { | |
1548 | message = "writing ring-page-order"; | |
1549 | goto abort_transaction; | |
1550 | } | |
1551 | ||
1552 | for (i = 0; i < info->nr_ring_pages; i++) { | |
1553 | char ring_ref_name[RINGREF_NAME_LEN]; | |
1554 | ||
1555 | snprintf(ring_ref_name, RINGREF_NAME_LEN, "ring-ref%u", i); | |
1556 | err = xenbus_printf(xbt, dev->nodename, ring_ref_name, | |
1557 | "%u", rinfo->ring_ref[i]); | |
1558 | if (err) { | |
1559 | message = "writing ring-ref"; | |
1560 | goto abort_transaction; | |
1561 | } | |
1562 | } | |
86839c56 | 1563 | } |
3df0e505 BL |
1564 | err = xenbus_printf(xbt, dev->nodename, |
1565 | "event-channel", "%u", rinfo->evtchn); | |
1566 | if (err) { | |
1567 | message = "writing event-channel"; | |
1568 | goto abort_transaction; | |
1569 | } | |
1570 | } else { | |
1571 | /* Not supported at this stage. */ | |
9f27ee59 JF |
1572 | goto abort_transaction; |
1573 | } | |
3e334239 MA |
1574 | err = xenbus_printf(xbt, dev->nodename, "protocol", "%s", |
1575 | XEN_IO_PROTO_ABI_NATIVE); | |
1576 | if (err) { | |
1577 | message = "writing protocol"; | |
1578 | goto abort_transaction; | |
1579 | } | |
0a8704a5 | 1580 | err = xenbus_printf(xbt, dev->nodename, |
cb5bd4d1 | 1581 | "feature-persistent", "%u", 1); |
0a8704a5 RPM |
1582 | if (err) |
1583 | dev_warn(&dev->dev, | |
1584 | "writing persistent grants feature to xenbus"); | |
9f27ee59 JF |
1585 | |
1586 | err = xenbus_transaction_end(xbt, 0); | |
1587 | if (err) { | |
1588 | if (err == -EAGAIN) | |
1589 | goto again; | |
1590 | xenbus_dev_fatal(dev, err, "completing transaction"); | |
1591 | goto destroy_blkring; | |
1592 | } | |
1593 | ||
3df0e505 BL |
1594 | for (i = 0; i < info->nr_rings; i++) { |
1595 | unsigned int j; | |
1596 | ||
1597 | rinfo = &info->rinfo[i]; | |
1598 | ||
1599 | for (j = 0; j < BLK_RING_SIZE(info); j++) | |
1600 | rinfo->shadow[j].req.u.rw.id = j + 1; | |
1601 | rinfo->shadow[BLK_RING_SIZE(info)-1].req.u.rw.id = 0x0fffffff; | |
1602 | } | |
9f27ee59 JF |
1603 | xenbus_switch_state(dev, XenbusStateInitialised); |
1604 | ||
1605 | return 0; | |
1606 | ||
1607 | abort_transaction: | |
1608 | xenbus_transaction_end(xbt, 1); | |
1609 | if (message) | |
1610 | xenbus_dev_fatal(dev, err, "%s", message); | |
1611 | destroy_blkring: | |
1612 | blkif_free(info, 0); | |
3df0e505 | 1613 | |
9f27ee59 JF |
1614 | return err; |
1615 | } | |
1616 | ||
9f27ee59 JF |
1617 | /** |
1618 | * Entry point to this code when a new device is created. Allocate the basic | |
1619 | * structures and the ring buffer for communication with the backend, and | |
1620 | * inform the backend of the appropriate details for those. Switch to | |
1621 | * Initialised state. | |
1622 | */ | |
1623 | static int blkfront_probe(struct xenbus_device *dev, | |
1624 | const struct xenbus_device_id *id) | |
1625 | { | |
86839c56 | 1626 | int err, vdevice; |
3df0e505 | 1627 | unsigned int r_index; |
9f27ee59 JF |
1628 | struct blkfront_info *info; |
1629 | ||
1630 | /* FIXME: Use dynamic device id if this is not set. */ | |
1631 | err = xenbus_scanf(XBT_NIL, dev->nodename, | |
1632 | "virtual-device", "%i", &vdevice); | |
1633 | if (err != 1) { | |
9246b5f0 CL |
1634 | /* go looking in the extended area instead */ |
1635 | err = xenbus_scanf(XBT_NIL, dev->nodename, "virtual-device-ext", | |
1636 | "%i", &vdevice); | |
1637 | if (err != 1) { | |
1638 | xenbus_dev_fatal(dev, err, "reading virtual-device"); | |
1639 | return err; | |
1640 | } | |
9f27ee59 JF |
1641 | } |
1642 | ||
b98a409b SS |
1643 | if (xen_hvm_domain()) { |
1644 | char *type; | |
1645 | int len; | |
1646 | /* no unplug has been done: do not hook devices != xen vbds */ | |
51c71a3b | 1647 | if (xen_has_pv_and_legacy_disk_devices()) { |
b98a409b SS |
1648 | int major; |
1649 | ||
1650 | if (!VDEV_IS_EXTENDED(vdevice)) | |
1651 | major = BLKIF_MAJOR(vdevice); | |
1652 | else | |
1653 | major = XENVBD_MAJOR; | |
1654 | ||
1655 | if (major != XENVBD_MAJOR) { | |
1656 | printk(KERN_INFO | |
1657 | "%s: HVM does not support vbd %d as xen block device\n", | |
02f1f217 | 1658 | __func__, vdevice); |
b98a409b SS |
1659 | return -ENODEV; |
1660 | } | |
1661 | } | |
1662 | /* do not create a PV cdrom device if we are an HVM guest */ | |
1663 | type = xenbus_read(XBT_NIL, dev->nodename, "device-type", &len); | |
1664 | if (IS_ERR(type)) | |
1665 | return -ENODEV; | |
1666 | if (strncmp(type, "cdrom", 5) == 0) { | |
1667 | kfree(type); | |
c1c5413a SS |
1668 | return -ENODEV; |
1669 | } | |
b98a409b | 1670 | kfree(type); |
c1c5413a | 1671 | } |
9f27ee59 JF |
1672 | info = kzalloc(sizeof(*info), GFP_KERNEL); |
1673 | if (!info) { | |
1674 | xenbus_dev_fatal(dev, -ENOMEM, "allocating info structure"); | |
1675 | return -ENOMEM; | |
1676 | } | |
1677 | ||
3df0e505 BL |
1678 | info->nr_rings = 1; |
1679 | info->rinfo = kzalloc(sizeof(struct blkfront_ring_info) * info->nr_rings, GFP_KERNEL); | |
1680 | if (!info->rinfo) { | |
1681 | xenbus_dev_fatal(dev, -ENOMEM, "allocating ring_info structure"); | |
1682 | kfree(info); | |
1683 | return -ENOMEM; | |
1684 | } | |
1685 | ||
1686 | for (r_index = 0; r_index < info->nr_rings; r_index++) { | |
1687 | struct blkfront_ring_info *rinfo; | |
1688 | ||
1689 | rinfo = &info->rinfo[r_index]; | |
1690 | INIT_LIST_HEAD(&rinfo->indirect_pages); | |
1691 | rinfo->dev_info = info; | |
1692 | INIT_WORK(&rinfo->work, blkif_restart_queue); | |
1693 | } | |
81f35161 | 1694 | |
b70f5fa0 | 1695 | mutex_init(&info->mutex); |
3467811e | 1696 | spin_lock_init(&info->io_lock); |
9f27ee59 JF |
1697 | info->xbdev = dev; |
1698 | info->vdevice = vdevice; | |
bfe11d6d | 1699 | INIT_LIST_HEAD(&info->grants); |
0a8704a5 | 1700 | info->persistent_gnts_c = 0; |
9f27ee59 | 1701 | info->connected = BLKIF_STATE_DISCONNECTED; |
9f27ee59 | 1702 | |
9f27ee59 JF |
1703 | /* Front end dir is a number, which is used as the id. */ |
1704 | info->handle = simple_strtoul(strrchr(dev->nodename, '/')+1, NULL, 0); | |
a1b4b12b | 1705 | dev_set_drvdata(&dev->dev, info); |
9f27ee59 | 1706 | |
9f27ee59 JF |
1707 | return 0; |
1708 | } | |
1709 | ||
4246a0b6 | 1710 | static void split_bio_end(struct bio *bio) |
402b27f9 RPM |
1711 | { |
1712 | struct split_bio *split_bio = bio->bi_private; | |
1713 | ||
402b27f9 RPM |
1714 | if (atomic_dec_and_test(&split_bio->pending)) { |
1715 | split_bio->bio->bi_phys_segments = 0; | |
4246a0b6 CH |
1716 | split_bio->bio->bi_error = bio->bi_error; |
1717 | bio_endio(split_bio->bio); | |
402b27f9 RPM |
1718 | kfree(split_bio); |
1719 | } | |
1720 | bio_put(bio); | |
1721 | } | |
9f27ee59 JF |
1722 | |
1723 | static int blkif_recover(struct blkfront_info *info) | |
1724 | { | |
3df0e505 | 1725 | unsigned int i, r_index; |
402b27f9 | 1726 | struct request *req, *n; |
9f27ee59 | 1727 | struct blk_shadow *copy; |
402b27f9 RPM |
1728 | int rc; |
1729 | struct bio *bio, *cloned_bio; | |
1730 | struct bio_list bio_list, merge_bio; | |
1731 | unsigned int segs, offset; | |
1732 | int pending, size; | |
1733 | struct split_bio *split_bio; | |
1734 | struct list_head requests; | |
402b27f9 | 1735 | |
3df0e505 | 1736 | blkfront_gather_backend_features(info); |
402b27f9 RPM |
1737 | segs = info->max_indirect_segments ? : BLKIF_MAX_SEGMENTS_PER_REQUEST; |
1738 | blk_queue_max_segments(info->rq, segs); | |
1739 | bio_list_init(&bio_list); | |
1740 | INIT_LIST_HEAD(&requests); | |
9f27ee59 | 1741 | |
3df0e505 BL |
1742 | for (r_index = 0; r_index < info->nr_rings; r_index++) { |
1743 | struct blkfront_ring_info *rinfo; | |
1744 | ||
1745 | rinfo = &info->rinfo[r_index]; | |
1746 | /* Stage 1: Make a safe copy of the shadow state. */ | |
1747 | copy = kmemdup(rinfo->shadow, sizeof(rinfo->shadow), | |
1748 | GFP_NOIO | __GFP_REPEAT | __GFP_HIGH); | |
1749 | if (!copy) | |
1750 | return -ENOMEM; | |
1751 | ||
1752 | /* Stage 2: Set up free list. */ | |
1753 | memset(&rinfo->shadow, 0, sizeof(rinfo->shadow)); | |
1754 | for (i = 0; i < BLK_RING_SIZE(info); i++) | |
1755 | rinfo->shadow[i].req.u.rw.id = i+1; | |
1756 | rinfo->shadow_free = rinfo->ring.req_prod_pvt; | |
1757 | rinfo->shadow[BLK_RING_SIZE(info)-1].req.u.rw.id = 0x0fffffff; | |
1758 | ||
1759 | rc = blkfront_setup_indirect(rinfo); | |
1760 | if (rc) { | |
1761 | kfree(copy); | |
1762 | return rc; | |
1763 | } | |
1764 | ||
1765 | for (i = 0; i < BLK_RING_SIZE(info); i++) { | |
1766 | /* Not in use? */ | |
1767 | if (!copy[i].request) | |
1768 | continue; | |
1769 | ||
402b27f9 | 1770 | /* |
3df0e505 | 1771 | * Get the bios in the request so we can re-queue them. |
402b27f9 | 1772 | */ |
3df0e505 BL |
1773 | if (copy[i].request->cmd_flags & |
1774 | (REQ_FLUSH | REQ_FUA | REQ_DISCARD | REQ_SECURE)) { | |
1775 | /* | |
1776 | * Flush operations don't contain bios, so | |
1777 | * we need to requeue the whole request | |
1778 | */ | |
1779 | list_add(©[i].request->queuelist, &requests); | |
1780 | continue; | |
1781 | } | |
1782 | merge_bio.head = copy[i].request->bio; | |
1783 | merge_bio.tail = copy[i].request->biotail; | |
1784 | bio_list_merge(&bio_list, &merge_bio); | |
1785 | copy[i].request->bio = NULL; | |
1786 | blk_end_request_all(copy[i].request, 0); | |
5ea42986 | 1787 | } |
9f27ee59 | 1788 | |
3df0e505 BL |
1789 | kfree(copy); |
1790 | } | |
9f27ee59 JF |
1791 | xenbus_switch_state(info->xbdev, XenbusStateConnected); |
1792 | ||
3467811e | 1793 | spin_lock_irq(&info->io_lock); |
9f27ee59 JF |
1794 | |
1795 | /* Now safe for us to use the shared ring */ | |
1796 | info->connected = BLKIF_STATE_CONNECTED; | |
1797 | ||
3df0e505 BL |
1798 | for (r_index = 0; r_index < info->nr_rings; r_index++) { |
1799 | struct blkfront_ring_info *rinfo; | |
1800 | ||
1801 | rinfo = &info->rinfo[r_index]; | |
1802 | /* Kick any other new requests queued since we resumed */ | |
1803 | kick_pending_request_queues(rinfo); | |
1804 | } | |
9f27ee59 | 1805 | |
402b27f9 RPM |
1806 | list_for_each_entry_safe(req, n, &requests, queuelist) { |
1807 | /* Requeue pending requests (flush or discard) */ | |
1808 | list_del_init(&req->queuelist); | |
1809 | BUG_ON(req->nr_phys_segments > segs); | |
907c3eb1 | 1810 | blk_mq_requeue_request(req); |
402b27f9 | 1811 | } |
3467811e | 1812 | spin_unlock_irq(&info->io_lock); |
907c3eb1 | 1813 | blk_mq_kick_requeue_list(info->rq); |
9f27ee59 | 1814 | |
402b27f9 RPM |
1815 | while ((bio = bio_list_pop(&bio_list)) != NULL) { |
1816 | /* Traverse the list of pending bios and re-queue them */ | |
1817 | if (bio_segments(bio) > segs) { | |
1818 | /* | |
1819 | * This bio has more segments than what we can | |
1820 | * handle, we have to split it. | |
1821 | */ | |
1822 | pending = (bio_segments(bio) + segs - 1) / segs; | |
1823 | split_bio = kzalloc(sizeof(*split_bio), GFP_NOIO); | |
1824 | BUG_ON(split_bio == NULL); | |
1825 | atomic_set(&split_bio->pending, pending); | |
1826 | split_bio->bio = bio; | |
1827 | for (i = 0; i < pending; i++) { | |
c004a6fe JG |
1828 | offset = (i * segs * XEN_PAGE_SIZE) >> 9; |
1829 | size = min((unsigned int)(segs * XEN_PAGE_SIZE) >> 9, | |
4f024f37 | 1830 | (unsigned int)bio_sectors(bio) - offset); |
402b27f9 RPM |
1831 | cloned_bio = bio_clone(bio, GFP_NOIO); |
1832 | BUG_ON(cloned_bio == NULL); | |
6678d83f | 1833 | bio_trim(cloned_bio, offset, size); |
402b27f9 RPM |
1834 | cloned_bio->bi_private = split_bio; |
1835 | cloned_bio->bi_end_io = split_bio_end; | |
1836 | submit_bio(cloned_bio->bi_rw, cloned_bio); | |
1837 | } | |
1838 | /* | |
1839 | * Now we have to wait for all those smaller bios to | |
1840 | * end, so we can also end the "parent" bio. | |
1841 | */ | |
1842 | continue; | |
1843 | } | |
1844 | /* We don't need to split this bio */ | |
1845 | submit_bio(bio->bi_rw, bio); | |
1846 | } | |
1847 | ||
9f27ee59 JF |
1848 | return 0; |
1849 | } | |
1850 | ||
1851 | /** | |
1852 | * We are reconnecting to the backend, due to a suspend/resume, or a backend | |
1853 | * driver restart. We tear down our blkif structure and recreate it, but | |
1854 | * leave the device-layer structures intact so that this is transparent to the | |
1855 | * rest of the kernel. | |
1856 | */ | |
1857 | static int blkfront_resume(struct xenbus_device *dev) | |
1858 | { | |
a1b4b12b | 1859 | struct blkfront_info *info = dev_get_drvdata(&dev->dev); |
9f27ee59 JF |
1860 | int err; |
1861 | ||
1862 | dev_dbg(&dev->dev, "blkfront_resume: %s\n", dev->nodename); | |
1863 | ||
1864 | blkif_free(info, info->connected == BLKIF_STATE_CONNECTED); | |
1865 | ||
203fd61f | 1866 | err = talk_to_blkback(dev, info); |
402b27f9 RPM |
1867 | |
1868 | /* | |
1869 | * We have to wait for the backend to switch to | |
1870 | * connected state, since we want to read which | |
1871 | * features it supports. | |
1872 | */ | |
9f27ee59 JF |
1873 | |
1874 | return err; | |
1875 | } | |
1876 | ||
b70f5fa0 DS |
1877 | static void |
1878 | blkfront_closing(struct blkfront_info *info) | |
1879 | { | |
1880 | struct xenbus_device *xbdev = info->xbdev; | |
1881 | struct block_device *bdev = NULL; | |
1882 | ||
1883 | mutex_lock(&info->mutex); | |
1884 | ||
1885 | if (xbdev->state == XenbusStateClosing) { | |
1886 | mutex_unlock(&info->mutex); | |
1887 | return; | |
1888 | } | |
1889 | ||
1890 | if (info->gd) | |
1891 | bdev = bdget_disk(info->gd, 0); | |
1892 | ||
1893 | mutex_unlock(&info->mutex); | |
1894 | ||
1895 | if (!bdev) { | |
1896 | xenbus_frontend_closed(xbdev); | |
1897 | return; | |
1898 | } | |
1899 | ||
1900 | mutex_lock(&bdev->bd_mutex); | |
1901 | ||
7b32d104 | 1902 | if (bdev->bd_openers) { |
b70f5fa0 DS |
1903 | xenbus_dev_error(xbdev, -EBUSY, |
1904 | "Device in use; refusing to close"); | |
1905 | xenbus_switch_state(xbdev, XenbusStateClosing); | |
1906 | } else { | |
1907 | xlvbd_release_gendisk(info); | |
1908 | xenbus_frontend_closed(xbdev); | |
1909 | } | |
1910 | ||
1911 | mutex_unlock(&bdev->bd_mutex); | |
1912 | bdput(bdev); | |
1913 | } | |
9f27ee59 | 1914 | |
ed30bf31 LD |
1915 | static void blkfront_setup_discard(struct blkfront_info *info) |
1916 | { | |
1917 | int err; | |
ed30bf31 LD |
1918 | unsigned int discard_granularity; |
1919 | unsigned int discard_alignment; | |
5ea42986 | 1920 | unsigned int discard_secure; |
ed30bf31 | 1921 | |
1c8cad6c OH |
1922 | info->feature_discard = 1; |
1923 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
1924 | "discard-granularity", "%u", &discard_granularity, | |
1925 | "discard-alignment", "%u", &discard_alignment, | |
1926 | NULL); | |
1927 | if (!err) { | |
1928 | info->discard_granularity = discard_granularity; | |
1929 | info->discard_alignment = discard_alignment; | |
1930 | } | |
1931 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
1932 | "discard-secure", "%d", &discard_secure, | |
1933 | NULL); | |
1934 | if (!err) | |
1935 | info->feature_secdiscard = !!discard_secure; | |
ed30bf31 LD |
1936 | } |
1937 | ||
81f35161 | 1938 | static int blkfront_setup_indirect(struct blkfront_ring_info *rinfo) |
402b27f9 | 1939 | { |
c004a6fe | 1940 | unsigned int psegs, grants; |
402b27f9 | 1941 | int err, i; |
81f35161 | 1942 | struct blkfront_info *info = rinfo->dev_info; |
402b27f9 | 1943 | |
d50babbe | 1944 | if (info->max_indirect_segments == 0) |
c004a6fe | 1945 | grants = BLKIF_MAX_SEGMENTS_PER_REQUEST; |
d50babbe | 1946 | else |
c004a6fe JG |
1947 | grants = info->max_indirect_segments; |
1948 | psegs = grants / GRANTS_PER_PSEG; | |
402b27f9 | 1949 | |
81f35161 | 1950 | err = fill_grant_buffer(rinfo, |
c004a6fe | 1951 | (grants + INDIRECT_GREFS(grants)) * BLK_RING_SIZE(info)); |
402b27f9 RPM |
1952 | if (err) |
1953 | goto out_of_memory; | |
1954 | ||
bfe11d6d RPM |
1955 | if (!info->feature_persistent && info->max_indirect_segments) { |
1956 | /* | |
1957 | * We are using indirect descriptors but not persistent | |
1958 | * grants, we need to allocate a set of pages that can be | |
1959 | * used for mapping indirect grefs | |
1960 | */ | |
c004a6fe | 1961 | int num = INDIRECT_GREFS(grants) * BLK_RING_SIZE(info); |
bfe11d6d | 1962 | |
81f35161 | 1963 | BUG_ON(!list_empty(&rinfo->indirect_pages)); |
bfe11d6d RPM |
1964 | for (i = 0; i < num; i++) { |
1965 | struct page *indirect_page = alloc_page(GFP_NOIO); | |
1966 | if (!indirect_page) | |
1967 | goto out_of_memory; | |
81f35161 | 1968 | list_add(&indirect_page->lru, &rinfo->indirect_pages); |
bfe11d6d RPM |
1969 | } |
1970 | } | |
1971 | ||
86839c56 | 1972 | for (i = 0; i < BLK_RING_SIZE(info); i++) { |
81f35161 BL |
1973 | rinfo->shadow[i].grants_used = kzalloc( |
1974 | sizeof(rinfo->shadow[i].grants_used[0]) * grants, | |
402b27f9 | 1975 | GFP_NOIO); |
81f35161 | 1976 | rinfo->shadow[i].sg = kzalloc(sizeof(rinfo->shadow[i].sg[0]) * psegs, GFP_NOIO); |
402b27f9 | 1977 | if (info->max_indirect_segments) |
81f35161 BL |
1978 | rinfo->shadow[i].indirect_grants = kzalloc( |
1979 | sizeof(rinfo->shadow[i].indirect_grants[0]) * | |
c004a6fe | 1980 | INDIRECT_GREFS(grants), |
402b27f9 | 1981 | GFP_NOIO); |
81f35161 BL |
1982 | if ((rinfo->shadow[i].grants_used == NULL) || |
1983 | (rinfo->shadow[i].sg == NULL) || | |
402b27f9 | 1984 | (info->max_indirect_segments && |
81f35161 | 1985 | (rinfo->shadow[i].indirect_grants == NULL))) |
402b27f9 | 1986 | goto out_of_memory; |
81f35161 | 1987 | sg_init_table(rinfo->shadow[i].sg, psegs); |
402b27f9 RPM |
1988 | } |
1989 | ||
1990 | ||
1991 | return 0; | |
1992 | ||
1993 | out_of_memory: | |
86839c56 | 1994 | for (i = 0; i < BLK_RING_SIZE(info); i++) { |
81f35161 BL |
1995 | kfree(rinfo->shadow[i].grants_used); |
1996 | rinfo->shadow[i].grants_used = NULL; | |
1997 | kfree(rinfo->shadow[i].sg); | |
1998 | rinfo->shadow[i].sg = NULL; | |
1999 | kfree(rinfo->shadow[i].indirect_grants); | |
2000 | rinfo->shadow[i].indirect_grants = NULL; | |
402b27f9 | 2001 | } |
81f35161 | 2002 | if (!list_empty(&rinfo->indirect_pages)) { |
bfe11d6d | 2003 | struct page *indirect_page, *n; |
81f35161 | 2004 | list_for_each_entry_safe(indirect_page, n, &rinfo->indirect_pages, lru) { |
bfe11d6d RPM |
2005 | list_del(&indirect_page->lru); |
2006 | __free_page(indirect_page); | |
2007 | } | |
2008 | } | |
402b27f9 RPM |
2009 | return -ENOMEM; |
2010 | } | |
2011 | ||
d50babbe BL |
2012 | /* |
2013 | * Gather all backend feature-* | |
2014 | */ | |
3df0e505 | 2015 | static void blkfront_gather_backend_features(struct blkfront_info *info) |
d50babbe BL |
2016 | { |
2017 | int err; | |
2018 | int barrier, flush, discard, persistent; | |
2019 | unsigned int indirect_segments; | |
2020 | ||
2021 | info->feature_flush = 0; | |
2022 | ||
2023 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2024 | "feature-barrier", "%d", &barrier, | |
2025 | NULL); | |
2026 | ||
2027 | /* | |
2028 | * If there's no "feature-barrier" defined, then it means | |
2029 | * we're dealing with a very old backend which writes | |
2030 | * synchronously; nothing to do. | |
2031 | * | |
2032 | * If there are barriers, then we use flush. | |
2033 | */ | |
2034 | if (!err && barrier) | |
2035 | info->feature_flush = REQ_FLUSH | REQ_FUA; | |
2036 | /* | |
2037 | * And if there is "feature-flush-cache" use that above | |
2038 | * barriers. | |
2039 | */ | |
2040 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2041 | "feature-flush-cache", "%d", &flush, | |
2042 | NULL); | |
2043 | ||
2044 | if (!err && flush) | |
2045 | info->feature_flush = REQ_FLUSH; | |
2046 | ||
2047 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2048 | "feature-discard", "%d", &discard, | |
2049 | NULL); | |
2050 | ||
2051 | if (!err && discard) | |
2052 | blkfront_setup_discard(info); | |
2053 | ||
2054 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2055 | "feature-persistent", "%u", &persistent, | |
2056 | NULL); | |
2057 | if (err) | |
2058 | info->feature_persistent = 0; | |
2059 | else | |
2060 | info->feature_persistent = persistent; | |
2061 | ||
2062 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2063 | "feature-max-indirect-segments", "%u", &indirect_segments, | |
2064 | NULL); | |
2065 | if (err) | |
2066 | info->max_indirect_segments = 0; | |
2067 | else | |
2068 | info->max_indirect_segments = min(indirect_segments, | |
2069 | xen_blkif_max_segments); | |
d50babbe BL |
2070 | } |
2071 | ||
9f27ee59 JF |
2072 | /* |
2073 | * Invoked when the backend is finally 'ready' (and has told produced | |
2074 | * the details about the physical device - #sectors, size, etc). | |
2075 | */ | |
2076 | static void blkfront_connect(struct blkfront_info *info) | |
2077 | { | |
2078 | unsigned long long sectors; | |
2079 | unsigned long sector_size; | |
7c4d7d71 | 2080 | unsigned int physical_sector_size; |
9f27ee59 | 2081 | unsigned int binfo; |
3df0e505 | 2082 | int err, i; |
9f27ee59 | 2083 | |
1fa73be6 S |
2084 | switch (info->connected) { |
2085 | case BLKIF_STATE_CONNECTED: | |
2086 | /* | |
2087 | * Potentially, the back-end may be signalling | |
2088 | * a capacity change; update the capacity. | |
2089 | */ | |
2090 | err = xenbus_scanf(XBT_NIL, info->xbdev->otherend, | |
2091 | "sectors", "%Lu", §ors); | |
2092 | if (XENBUS_EXIST_ERR(err)) | |
2093 | return; | |
2094 | printk(KERN_INFO "Setting capacity to %Lu\n", | |
2095 | sectors); | |
2096 | set_capacity(info->gd, sectors); | |
2def141e | 2097 | revalidate_disk(info->gd); |
1fa73be6 | 2098 | |
402b27f9 | 2099 | return; |
1fa73be6 | 2100 | case BLKIF_STATE_SUSPENDED: |
402b27f9 RPM |
2101 | /* |
2102 | * If we are recovering from suspension, we need to wait | |
2103 | * for the backend to announce it's features before | |
2104 | * reconnecting, at least we need to know if the backend | |
2105 | * supports indirect descriptors, and how many. | |
2106 | */ | |
2107 | blkif_recover(info); | |
9f27ee59 JF |
2108 | return; |
2109 | ||
b4dddb49 JF |
2110 | default: |
2111 | break; | |
1fa73be6 | 2112 | } |
9f27ee59 JF |
2113 | |
2114 | dev_dbg(&info->xbdev->dev, "%s:%s.\n", | |
2115 | __func__, info->xbdev->otherend); | |
2116 | ||
2117 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2118 | "sectors", "%llu", §ors, | |
2119 | "info", "%u", &binfo, | |
2120 | "sector-size", "%lu", §or_size, | |
2121 | NULL); | |
2122 | if (err) { | |
2123 | xenbus_dev_fatal(info->xbdev, err, | |
2124 | "reading backend fields at %s", | |
2125 | info->xbdev->otherend); | |
2126 | return; | |
2127 | } | |
2128 | ||
7c4d7d71 SB |
2129 | /* |
2130 | * physcial-sector-size is a newer field, so old backends may not | |
2131 | * provide this. Assume physical sector size to be the same as | |
2132 | * sector_size in that case. | |
2133 | */ | |
2134 | err = xenbus_scanf(XBT_NIL, info->xbdev->otherend, | |
2135 | "physical-sector-size", "%u", &physical_sector_size); | |
2136 | if (err != 1) | |
2137 | physical_sector_size = sector_size; | |
2138 | ||
3df0e505 BL |
2139 | blkfront_gather_backend_features(info); |
2140 | for (i = 0; i < info->nr_rings; i++) { | |
2141 | err = blkfront_setup_indirect(&info->rinfo[i]); | |
2142 | if (err) { | |
2143 | xenbus_dev_fatal(info->xbdev, err, "setup_indirect at %s", | |
2144 | info->xbdev->otherend); | |
2145 | blkif_free(info, 0); | |
2146 | break; | |
2147 | } | |
402b27f9 RPM |
2148 | } |
2149 | ||
7c4d7d71 SB |
2150 | err = xlvbd_alloc_gendisk(sectors, info, binfo, sector_size, |
2151 | physical_sector_size); | |
9f27ee59 JF |
2152 | if (err) { |
2153 | xenbus_dev_fatal(info->xbdev, err, "xlvbd_add at %s", | |
2154 | info->xbdev->otherend); | |
2155 | return; | |
2156 | } | |
2157 | ||
2158 | xenbus_switch_state(info->xbdev, XenbusStateConnected); | |
2159 | ||
2160 | /* Kick pending requests. */ | |
3467811e | 2161 | spin_lock_irq(&info->io_lock); |
9f27ee59 | 2162 | info->connected = BLKIF_STATE_CONNECTED; |
3df0e505 BL |
2163 | for (i = 0; i < info->nr_rings; i++) |
2164 | kick_pending_request_queues(&info->rinfo[i]); | |
3467811e | 2165 | spin_unlock_irq(&info->io_lock); |
9f27ee59 JF |
2166 | |
2167 | add_disk(info->gd); | |
1d78d705 CL |
2168 | |
2169 | info->is_ready = 1; | |
9f27ee59 JF |
2170 | } |
2171 | ||
9f27ee59 JF |
2172 | /** |
2173 | * Callback received when the backend's state changes. | |
2174 | */ | |
203fd61f | 2175 | static void blkback_changed(struct xenbus_device *dev, |
9f27ee59 JF |
2176 | enum xenbus_state backend_state) |
2177 | { | |
a1b4b12b | 2178 | struct blkfront_info *info = dev_get_drvdata(&dev->dev); |
9f27ee59 | 2179 | |
203fd61f | 2180 | dev_dbg(&dev->dev, "blkfront:blkback_changed to state %d.\n", backend_state); |
9f27ee59 JF |
2181 | |
2182 | switch (backend_state) { | |
9f27ee59 | 2183 | case XenbusStateInitWait: |
a9b54bb9 BL |
2184 | if (dev->state != XenbusStateInitialising) |
2185 | break; | |
8ab0144a BL |
2186 | if (talk_to_blkback(dev, info)) { |
2187 | kfree(info); | |
2188 | dev_set_drvdata(&dev->dev, NULL); | |
2189 | break; | |
2190 | } | |
2191 | case XenbusStateInitialising: | |
9f27ee59 | 2192 | case XenbusStateInitialised: |
b78c9512 NI |
2193 | case XenbusStateReconfiguring: |
2194 | case XenbusStateReconfigured: | |
9f27ee59 | 2195 | case XenbusStateUnknown: |
9f27ee59 JF |
2196 | break; |
2197 | ||
2198 | case XenbusStateConnected: | |
2199 | blkfront_connect(info); | |
2200 | break; | |
2201 | ||
36613717 DV |
2202 | case XenbusStateClosed: |
2203 | if (dev->state == XenbusStateClosed) | |
2204 | break; | |
2205 | /* Missed the backend's Closing state -- fallthrough */ | |
9f27ee59 | 2206 | case XenbusStateClosing: |
a54c8f0f CA |
2207 | if (info) |
2208 | blkfront_closing(info); | |
9f27ee59 JF |
2209 | break; |
2210 | } | |
2211 | } | |
2212 | ||
fa1bd359 | 2213 | static int blkfront_remove(struct xenbus_device *xbdev) |
9f27ee59 | 2214 | { |
fa1bd359 DS |
2215 | struct blkfront_info *info = dev_get_drvdata(&xbdev->dev); |
2216 | struct block_device *bdev = NULL; | |
2217 | struct gendisk *disk; | |
9f27ee59 | 2218 | |
fa1bd359 | 2219 | dev_dbg(&xbdev->dev, "%s removed", xbdev->nodename); |
9f27ee59 JF |
2220 | |
2221 | blkif_free(info, 0); | |
2222 | ||
fa1bd359 DS |
2223 | mutex_lock(&info->mutex); |
2224 | ||
2225 | disk = info->gd; | |
2226 | if (disk) | |
2227 | bdev = bdget_disk(disk, 0); | |
2228 | ||
2229 | info->xbdev = NULL; | |
2230 | mutex_unlock(&info->mutex); | |
2231 | ||
2232 | if (!bdev) { | |
2233 | kfree(info); | |
2234 | return 0; | |
2235 | } | |
2236 | ||
2237 | /* | |
2238 | * The xbdev was removed before we reached the Closed | |
2239 | * state. See if it's safe to remove the disk. If the bdev | |
2240 | * isn't closed yet, we let release take care of it. | |
2241 | */ | |
2242 | ||
2243 | mutex_lock(&bdev->bd_mutex); | |
2244 | info = disk->private_data; | |
2245 | ||
d54142c7 DS |
2246 | dev_warn(disk_to_dev(disk), |
2247 | "%s was hot-unplugged, %d stale handles\n", | |
2248 | xbdev->nodename, bdev->bd_openers); | |
2249 | ||
7b32d104 | 2250 | if (info && !bdev->bd_openers) { |
fa1bd359 DS |
2251 | xlvbd_release_gendisk(info); |
2252 | disk->private_data = NULL; | |
0e345826 | 2253 | kfree(info); |
fa1bd359 DS |
2254 | } |
2255 | ||
2256 | mutex_unlock(&bdev->bd_mutex); | |
2257 | bdput(bdev); | |
9f27ee59 JF |
2258 | |
2259 | return 0; | |
2260 | } | |
2261 | ||
1d78d705 CL |
2262 | static int blkfront_is_ready(struct xenbus_device *dev) |
2263 | { | |
a1b4b12b | 2264 | struct blkfront_info *info = dev_get_drvdata(&dev->dev); |
1d78d705 | 2265 | |
5d7ed20e | 2266 | return info->is_ready && info->xbdev; |
1d78d705 CL |
2267 | } |
2268 | ||
a63c848b | 2269 | static int blkif_open(struct block_device *bdev, fmode_t mode) |
9f27ee59 | 2270 | { |
13961743 DS |
2271 | struct gendisk *disk = bdev->bd_disk; |
2272 | struct blkfront_info *info; | |
2273 | int err = 0; | |
6e9624b8 | 2274 | |
2a48fc0a | 2275 | mutex_lock(&blkfront_mutex); |
6e9624b8 | 2276 | |
13961743 DS |
2277 | info = disk->private_data; |
2278 | if (!info) { | |
2279 | /* xbdev gone */ | |
2280 | err = -ERESTARTSYS; | |
2281 | goto out; | |
2282 | } | |
2283 | ||
2284 | mutex_lock(&info->mutex); | |
2285 | ||
2286 | if (!info->gd) | |
2287 | /* xbdev is closed */ | |
2288 | err = -ERESTARTSYS; | |
2289 | ||
2290 | mutex_unlock(&info->mutex); | |
2291 | ||
13961743 | 2292 | out: |
2a48fc0a | 2293 | mutex_unlock(&blkfront_mutex); |
13961743 | 2294 | return err; |
9f27ee59 JF |
2295 | } |
2296 | ||
db2a144b | 2297 | static void blkif_release(struct gendisk *disk, fmode_t mode) |
9f27ee59 | 2298 | { |
a63c848b | 2299 | struct blkfront_info *info = disk->private_data; |
7fd152f4 DS |
2300 | struct block_device *bdev; |
2301 | struct xenbus_device *xbdev; | |
2302 | ||
2a48fc0a | 2303 | mutex_lock(&blkfront_mutex); |
7fd152f4 DS |
2304 | |
2305 | bdev = bdget_disk(disk, 0); | |
7fd152f4 | 2306 | |
2f089cb8 FP |
2307 | if (!bdev) { |
2308 | WARN(1, "Block device %s yanked out from us!\n", disk->disk_name); | |
2309 | goto out_mutex; | |
2310 | } | |
acfca3c6 DS |
2311 | if (bdev->bd_openers) |
2312 | goto out; | |
2313 | ||
7fd152f4 DS |
2314 | /* |
2315 | * Check if we have been instructed to close. We will have | |
2316 | * deferred this request, because the bdev was still open. | |
2317 | */ | |
2318 | ||
2319 | mutex_lock(&info->mutex); | |
2320 | xbdev = info->xbdev; | |
2321 | ||
2322 | if (xbdev && xbdev->state == XenbusStateClosing) { | |
2323 | /* pending switch to state closed */ | |
d54142c7 | 2324 | dev_info(disk_to_dev(bdev->bd_disk), "releasing disk\n"); |
7fd152f4 DS |
2325 | xlvbd_release_gendisk(info); |
2326 | xenbus_frontend_closed(info->xbdev); | |
2327 | } | |
2328 | ||
2329 | mutex_unlock(&info->mutex); | |
2330 | ||
2331 | if (!xbdev) { | |
2332 | /* sudden device removal */ | |
d54142c7 | 2333 | dev_info(disk_to_dev(bdev->bd_disk), "releasing disk\n"); |
7fd152f4 DS |
2334 | xlvbd_release_gendisk(info); |
2335 | disk->private_data = NULL; | |
2336 | kfree(info); | |
9f27ee59 | 2337 | } |
7fd152f4 | 2338 | |
a4cc14ec | 2339 | out: |
dad5cf65 | 2340 | bdput(bdev); |
2f089cb8 | 2341 | out_mutex: |
2a48fc0a | 2342 | mutex_unlock(&blkfront_mutex); |
9f27ee59 JF |
2343 | } |
2344 | ||
83d5cde4 | 2345 | static const struct block_device_operations xlvbd_block_fops = |
9f27ee59 JF |
2346 | { |
2347 | .owner = THIS_MODULE, | |
a63c848b AV |
2348 | .open = blkif_open, |
2349 | .release = blkif_release, | |
597592d9 | 2350 | .getgeo = blkif_getgeo, |
8a6cfeb6 | 2351 | .ioctl = blkif_ioctl, |
9f27ee59 JF |
2352 | }; |
2353 | ||
2354 | ||
ec9c42ec | 2355 | static const struct xenbus_device_id blkfront_ids[] = { |
9f27ee59 JF |
2356 | { "vbd" }, |
2357 | { "" } | |
2358 | }; | |
2359 | ||
95afae48 DV |
2360 | static struct xenbus_driver blkfront_driver = { |
2361 | .ids = blkfront_ids, | |
9f27ee59 JF |
2362 | .probe = blkfront_probe, |
2363 | .remove = blkfront_remove, | |
2364 | .resume = blkfront_resume, | |
203fd61f | 2365 | .otherend_changed = blkback_changed, |
1d78d705 | 2366 | .is_ready = blkfront_is_ready, |
95afae48 | 2367 | }; |
9f27ee59 JF |
2368 | |
2369 | static int __init xlblk_init(void) | |
2370 | { | |
469738e6 LE |
2371 | int ret; |
2372 | ||
6e833587 | 2373 | if (!xen_domain()) |
9f27ee59 JF |
2374 | return -ENODEV; |
2375 | ||
9cce2914 | 2376 | if (xen_blkif_max_ring_order > XENBUS_MAX_RING_GRANT_ORDER) { |
86839c56 | 2377 | pr_info("Invalid max_ring_order (%d), will use default max: %d.\n", |
9cce2914 | 2378 | xen_blkif_max_ring_order, XENBUS_MAX_RING_GRANT_ORDER); |
86839c56 BL |
2379 | xen_blkif_max_ring_order = 0; |
2380 | } | |
2381 | ||
51c71a3b | 2382 | if (!xen_has_pv_disk_devices()) |
b9136d20 IM |
2383 | return -ENODEV; |
2384 | ||
9f27ee59 JF |
2385 | if (register_blkdev(XENVBD_MAJOR, DEV_NAME)) { |
2386 | printk(KERN_WARNING "xen_blk: can't get major %d with name %s\n", | |
2387 | XENVBD_MAJOR, DEV_NAME); | |
2388 | return -ENODEV; | |
2389 | } | |
2390 | ||
73db144b | 2391 | ret = xenbus_register_frontend(&blkfront_driver); |
469738e6 LE |
2392 | if (ret) { |
2393 | unregister_blkdev(XENVBD_MAJOR, DEV_NAME); | |
2394 | return ret; | |
2395 | } | |
2396 | ||
2397 | return 0; | |
9f27ee59 JF |
2398 | } |
2399 | module_init(xlblk_init); | |
2400 | ||
2401 | ||
5a60d0cd | 2402 | static void __exit xlblk_exit(void) |
9f27ee59 | 2403 | { |
8605067f JB |
2404 | xenbus_unregister_driver(&blkfront_driver); |
2405 | unregister_blkdev(XENVBD_MAJOR, DEV_NAME); | |
2406 | kfree(minors); | |
9f27ee59 JF |
2407 | } |
2408 | module_exit(xlblk_exit); | |
2409 | ||
2410 | MODULE_DESCRIPTION("Xen virtual block device frontend"); | |
2411 | MODULE_LICENSE("GPL"); | |
2412 | MODULE_ALIAS_BLOCKDEV_MAJOR(XENVBD_MAJOR); | |
d2f0c52b | 2413 | MODULE_ALIAS("xen:vbd"); |
4f93f09b | 2414 | MODULE_ALIAS("xenblk"); |