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 | ||
6cc56833 JG |
63 | /* |
64 | * The minimal size of segment supported by the block framework is PAGE_SIZE. | |
65 | * When Linux is using a different page size than Xen, it may not be possible | |
66 | * to put all the data in a single segment. | |
67 | * This can happen when the backend doesn't support indirect descriptor and | |
68 | * therefore the maximum amount of data that a request can carry is | |
69 | * BLKIF_MAX_SEGMENTS_PER_REQUEST * XEN_PAGE_SIZE = 44KB | |
70 | * | |
71 | * Note that we only support one extra request. So the Linux page size | |
72 | * should be <= ( 2 * BLKIF_MAX_SEGMENTS_PER_REQUEST * XEN_PAGE_SIZE) = | |
73 | * 88KB. | |
74 | */ | |
75 | #define HAS_EXTRA_REQ (BLKIF_MAX_SEGMENTS_PER_REQUEST < XEN_PFN_PER_PAGE) | |
76 | ||
9f27ee59 JF |
77 | enum blkif_state { |
78 | BLKIF_STATE_DISCONNECTED, | |
79 | BLKIF_STATE_CONNECTED, | |
80 | BLKIF_STATE_SUSPENDED, | |
81 | }; | |
82 | ||
0a8704a5 RPM |
83 | struct grant { |
84 | grant_ref_t gref; | |
a7a6df22 | 85 | struct page *page; |
155b7edb | 86 | struct list_head node; |
0a8704a5 RPM |
87 | }; |
88 | ||
6cc56833 JG |
89 | enum blk_req_status { |
90 | REQ_WAITING, | |
91 | REQ_DONE, | |
92 | REQ_ERROR, | |
93 | REQ_EOPNOTSUPP, | |
94 | }; | |
95 | ||
9f27ee59 JF |
96 | struct blk_shadow { |
97 | struct blkif_request req; | |
a945b980 | 98 | struct request *request; |
402b27f9 RPM |
99 | struct grant **grants_used; |
100 | struct grant **indirect_grants; | |
b7649158 | 101 | struct scatterlist *sg; |
c004a6fe | 102 | unsigned int num_sg; |
6cc56833 JG |
103 | enum blk_req_status status; |
104 | ||
105 | #define NO_ASSOCIATED_ID ~0UL | |
106 | /* | |
107 | * Id of the sibling if we ever need 2 requests when handling a | |
108 | * block I/O request | |
109 | */ | |
110 | unsigned long associated_id; | |
402b27f9 RPM |
111 | }; |
112 | ||
113 | struct split_bio { | |
114 | struct bio *bio; | |
115 | atomic_t pending; | |
9f27ee59 JF |
116 | }; |
117 | ||
2a48fc0a | 118 | static DEFINE_MUTEX(blkfront_mutex); |
83d5cde4 | 119 | static const struct block_device_operations xlvbd_block_fops; |
9f27ee59 | 120 | |
402b27f9 RPM |
121 | /* |
122 | * Maximum number of segments in indirect requests, the actual value used by | |
123 | * the frontend driver is the minimum of this value and the value provided | |
124 | * by the backend driver. | |
125 | */ | |
126 | ||
127 | static unsigned int xen_blkif_max_segments = 32; | |
14e710fe JB |
128 | module_param_named(max_indirect_segments, xen_blkif_max_segments, uint, |
129 | S_IRUGO); | |
130 | MODULE_PARM_DESC(max_indirect_segments, | |
131 | "Maximum amount of segments in indirect requests (default is 32)"); | |
402b27f9 | 132 | |
28d949bc BL |
133 | static unsigned int xen_blkif_max_queues = 4; |
134 | module_param_named(max_queues, xen_blkif_max_queues, uint, S_IRUGO); | |
135 | MODULE_PARM_DESC(max_queues, "Maximum number of hardware queues/rings used per virtual disk"); | |
136 | ||
86839c56 BL |
137 | /* |
138 | * Maximum order of pages to be used for the shared ring between front and | |
139 | * backend, 4KB page granularity is used. | |
140 | */ | |
141 | static unsigned int xen_blkif_max_ring_order; | |
142 | module_param_named(max_ring_page_order, xen_blkif_max_ring_order, int, S_IRUGO); | |
143 | MODULE_PARM_DESC(max_ring_page_order, "Maximum order of pages to be used for the shared ring"); | |
144 | ||
c004a6fe JG |
145 | #define BLK_RING_SIZE(info) \ |
146 | __CONST_RING_SIZE(blkif, XEN_PAGE_SIZE * (info)->nr_ring_pages) | |
147 | ||
148 | #define BLK_MAX_RING_SIZE \ | |
9cce2914 | 149 | __CONST_RING_SIZE(blkif, XEN_PAGE_SIZE * XENBUS_MAX_RING_GRANTS) |
c004a6fe | 150 | |
86839c56 | 151 | /* |
6f03a7ff KRW |
152 | * ring-ref%u i=(-1UL) would take 11 characters + 'ring-ref' is 8, so 19 |
153 | * characters are enough. Define to 20 to keep consistent with backend. | |
86839c56 BL |
154 | */ |
155 | #define RINGREF_NAME_LEN (20) | |
28d949bc BL |
156 | /* |
157 | * queue-%u would take 7 + 10(UINT_MAX) = 17 characters. | |
158 | */ | |
159 | #define QUEUE_NAME_LEN (17) | |
9f27ee59 | 160 | |
81f35161 BL |
161 | /* |
162 | * Per-ring info. | |
163 | * Every blkfront device can associate with one or more blkfront_ring_info, | |
164 | * depending on how many hardware queues/rings to be used. | |
165 | */ | |
166 | struct blkfront_ring_info { | |
11659569 BL |
167 | /* Lock to protect data in every ring buffer. */ |
168 | spinlock_t ring_lock; | |
81f35161 BL |
169 | struct blkif_front_ring ring; |
170 | unsigned int ring_ref[XENBUS_MAX_RING_GRANTS]; | |
171 | unsigned int evtchn, irq; | |
172 | struct work_struct work; | |
173 | struct gnttab_free_callback callback; | |
174 | struct blk_shadow shadow[BLK_MAX_RING_SIZE]; | |
175 | struct list_head indirect_pages; | |
73716df7 BL |
176 | struct list_head grants; |
177 | unsigned int persistent_gnts_c; | |
81f35161 BL |
178 | unsigned long shadow_free; |
179 | struct blkfront_info *dev_info; | |
180 | }; | |
181 | ||
9f27ee59 JF |
182 | /* |
183 | * We have one of these per vbd, whether ide, scsi or 'other'. They | |
184 | * hang in private_data off the gendisk structure. We may end up | |
185 | * putting all kinds of interesting stuff here :-) | |
186 | */ | |
187 | struct blkfront_info | |
188 | { | |
b70f5fa0 | 189 | struct mutex mutex; |
9f27ee59 | 190 | struct xenbus_device *xbdev; |
9f27ee59 JF |
191 | struct gendisk *gd; |
192 | int vdevice; | |
193 | blkif_vdev_t handle; | |
194 | enum blkif_state connected; | |
3df0e505 | 195 | /* Number of pages per ring buffer. */ |
86839c56 | 196 | unsigned int nr_ring_pages; |
9f27ee59 | 197 | struct request_queue *rq; |
4913efe4 | 198 | unsigned int feature_flush; |
5ea42986 KRW |
199 | unsigned int feature_discard:1; |
200 | unsigned int feature_secdiscard:1; | |
ed30bf31 LD |
201 | unsigned int discard_granularity; |
202 | unsigned int discard_alignment; | |
0a8704a5 | 203 | unsigned int feature_persistent:1; |
c004a6fe | 204 | /* Number of 4KB segments handled */ |
402b27f9 | 205 | unsigned int max_indirect_segments; |
1d78d705 | 206 | int is_ready; |
907c3eb1 | 207 | struct blk_mq_tag_set tag_set; |
3df0e505 BL |
208 | struct blkfront_ring_info *rinfo; |
209 | unsigned int nr_rings; | |
7b427a59 BL |
210 | /* Save uncomplete reqs and bios for migration. */ |
211 | struct list_head requests; | |
212 | struct bio_list bio_list; | |
9f27ee59 JF |
213 | }; |
214 | ||
0e345826 JB |
215 | static unsigned int nr_minors; |
216 | static unsigned long *minors; | |
217 | static DEFINE_SPINLOCK(minor_lock); | |
218 | ||
9f27ee59 JF |
219 | #define GRANT_INVALID_REF 0 |
220 | ||
221 | #define PARTS_PER_DISK 16 | |
9246b5f0 | 222 | #define PARTS_PER_EXT_DISK 256 |
9f27ee59 JF |
223 | |
224 | #define BLKIF_MAJOR(dev) ((dev)>>8) | |
225 | #define BLKIF_MINOR(dev) ((dev) & 0xff) | |
226 | ||
9246b5f0 CL |
227 | #define EXT_SHIFT 28 |
228 | #define EXTENDED (1<<EXT_SHIFT) | |
229 | #define VDEV_IS_EXTENDED(dev) ((dev)&(EXTENDED)) | |
230 | #define BLKIF_MINOR_EXT(dev) ((dev)&(~EXTENDED)) | |
c80a4209 SS |
231 | #define EMULATED_HD_DISK_MINOR_OFFSET (0) |
232 | #define EMULATED_HD_DISK_NAME_OFFSET (EMULATED_HD_DISK_MINOR_OFFSET / 256) | |
196cfe2a SB |
233 | #define EMULATED_SD_DISK_MINOR_OFFSET (0) |
234 | #define EMULATED_SD_DISK_NAME_OFFSET (EMULATED_SD_DISK_MINOR_OFFSET / 256) | |
9f27ee59 | 235 | |
9246b5f0 | 236 | #define DEV_NAME "xvd" /* name in /dev */ |
9f27ee59 | 237 | |
c004a6fe JG |
238 | /* |
239 | * Grants are always the same size as a Xen page (i.e 4KB). | |
240 | * A physical segment is always the same size as a Linux page. | |
241 | * Number of grants per physical segment | |
242 | */ | |
243 | #define GRANTS_PER_PSEG (PAGE_SIZE / XEN_PAGE_SIZE) | |
244 | ||
245 | #define GRANTS_PER_INDIRECT_FRAME \ | |
246 | (XEN_PAGE_SIZE / sizeof(struct blkif_request_segment)) | |
247 | ||
248 | #define PSEGS_PER_INDIRECT_FRAME \ | |
249 | (GRANTS_INDIRECT_FRAME / GRANTS_PSEGS) | |
250 | ||
251 | #define INDIRECT_GREFS(_grants) \ | |
252 | DIV_ROUND_UP(_grants, GRANTS_PER_INDIRECT_FRAME) | |
253 | ||
254 | #define GREFS(_psegs) ((_psegs) * GRANTS_PER_PSEG) | |
402b27f9 | 255 | |
81f35161 | 256 | static int blkfront_setup_indirect(struct blkfront_ring_info *rinfo); |
3df0e505 | 257 | static void blkfront_gather_backend_features(struct blkfront_info *info); |
402b27f9 | 258 | |
81f35161 | 259 | static int get_id_from_freelist(struct blkfront_ring_info *rinfo) |
9f27ee59 | 260 | { |
81f35161 BL |
261 | unsigned long free = rinfo->shadow_free; |
262 | ||
263 | BUG_ON(free >= BLK_RING_SIZE(rinfo->dev_info)); | |
264 | rinfo->shadow_free = rinfo->shadow[free].req.u.rw.id; | |
265 | rinfo->shadow[free].req.u.rw.id = 0x0fffffee; /* debug */ | |
9f27ee59 JF |
266 | return free; |
267 | } | |
268 | ||
81f35161 | 269 | static int add_id_to_freelist(struct blkfront_ring_info *rinfo, |
6f03a7ff | 270 | unsigned long id) |
9f27ee59 | 271 | { |
81f35161 | 272 | if (rinfo->shadow[id].req.u.rw.id != id) |
6878c32e | 273 | return -EINVAL; |
81f35161 | 274 | if (rinfo->shadow[id].request == NULL) |
6878c32e | 275 | return -EINVAL; |
81f35161 BL |
276 | rinfo->shadow[id].req.u.rw.id = rinfo->shadow_free; |
277 | rinfo->shadow[id].request = NULL; | |
278 | rinfo->shadow_free = id; | |
6878c32e | 279 | return 0; |
9f27ee59 JF |
280 | } |
281 | ||
81f35161 | 282 | static int fill_grant_buffer(struct blkfront_ring_info *rinfo, int num) |
9c1e050c | 283 | { |
81f35161 | 284 | struct blkfront_info *info = rinfo->dev_info; |
9c1e050c RPM |
285 | struct page *granted_page; |
286 | struct grant *gnt_list_entry, *n; | |
287 | int i = 0; | |
288 | ||
6f03a7ff | 289 | while (i < num) { |
9c1e050c RPM |
290 | gnt_list_entry = kzalloc(sizeof(struct grant), GFP_NOIO); |
291 | if (!gnt_list_entry) | |
292 | goto out_of_memory; | |
293 | ||
bfe11d6d RPM |
294 | if (info->feature_persistent) { |
295 | granted_page = alloc_page(GFP_NOIO); | |
296 | if (!granted_page) { | |
297 | kfree(gnt_list_entry); | |
298 | goto out_of_memory; | |
299 | } | |
a7a6df22 | 300 | gnt_list_entry->page = granted_page; |
9c1e050c RPM |
301 | } |
302 | ||
9c1e050c | 303 | gnt_list_entry->gref = GRANT_INVALID_REF; |
73716df7 | 304 | list_add(&gnt_list_entry->node, &rinfo->grants); |
9c1e050c RPM |
305 | i++; |
306 | } | |
307 | ||
308 | return 0; | |
309 | ||
310 | out_of_memory: | |
311 | list_for_each_entry_safe(gnt_list_entry, n, | |
73716df7 | 312 | &rinfo->grants, node) { |
9c1e050c | 313 | list_del(&gnt_list_entry->node); |
bfe11d6d | 314 | if (info->feature_persistent) |
a7a6df22 | 315 | __free_page(gnt_list_entry->page); |
9c1e050c RPM |
316 | kfree(gnt_list_entry); |
317 | i--; | |
318 | } | |
319 | BUG_ON(i != 0); | |
320 | return -ENOMEM; | |
321 | } | |
322 | ||
73716df7 | 323 | static struct grant *get_free_grant(struct blkfront_ring_info *rinfo) |
9c1e050c RPM |
324 | { |
325 | struct grant *gnt_list_entry; | |
9c1e050c | 326 | |
73716df7 BL |
327 | BUG_ON(list_empty(&rinfo->grants)); |
328 | gnt_list_entry = list_first_entry(&rinfo->grants, struct grant, | |
4f503fbd | 329 | node); |
9c1e050c RPM |
330 | list_del(&gnt_list_entry->node); |
331 | ||
4f503fbd | 332 | if (gnt_list_entry->gref != GRANT_INVALID_REF) |
73716df7 | 333 | rinfo->persistent_gnts_c--; |
4f503fbd JG |
334 | |
335 | return gnt_list_entry; | |
336 | } | |
337 | ||
338 | static inline void grant_foreign_access(const struct grant *gnt_list_entry, | |
339 | const struct blkfront_info *info) | |
340 | { | |
341 | gnttab_page_grant_foreign_access_ref_one(gnt_list_entry->gref, | |
342 | info->xbdev->otherend_id, | |
343 | gnt_list_entry->page, | |
344 | 0); | |
345 | } | |
346 | ||
347 | static struct grant *get_grant(grant_ref_t *gref_head, | |
348 | unsigned long gfn, | |
73716df7 | 349 | struct blkfront_ring_info *rinfo) |
4f503fbd | 350 | { |
73716df7 BL |
351 | struct grant *gnt_list_entry = get_free_grant(rinfo); |
352 | struct blkfront_info *info = rinfo->dev_info; | |
4f503fbd JG |
353 | |
354 | if (gnt_list_entry->gref != GRANT_INVALID_REF) | |
9c1e050c | 355 | return gnt_list_entry; |
4f503fbd JG |
356 | |
357 | /* Assign a gref to this page */ | |
358 | gnt_list_entry->gref = gnttab_claim_grant_reference(gref_head); | |
359 | BUG_ON(gnt_list_entry->gref == -ENOSPC); | |
360 | if (info->feature_persistent) | |
361 | grant_foreign_access(gnt_list_entry, info); | |
362 | else { | |
363 | /* Grant access to the GFN passed by the caller */ | |
364 | gnttab_grant_foreign_access_ref(gnt_list_entry->gref, | |
365 | info->xbdev->otherend_id, | |
366 | gfn, 0); | |
9c1e050c RPM |
367 | } |
368 | ||
4f503fbd JG |
369 | return gnt_list_entry; |
370 | } | |
371 | ||
372 | static struct grant *get_indirect_grant(grant_ref_t *gref_head, | |
73716df7 | 373 | struct blkfront_ring_info *rinfo) |
4f503fbd | 374 | { |
73716df7 BL |
375 | struct grant *gnt_list_entry = get_free_grant(rinfo); |
376 | struct blkfront_info *info = rinfo->dev_info; | |
4f503fbd JG |
377 | |
378 | if (gnt_list_entry->gref != GRANT_INVALID_REF) | |
379 | return gnt_list_entry; | |
380 | ||
9c1e050c RPM |
381 | /* Assign a gref to this page */ |
382 | gnt_list_entry->gref = gnttab_claim_grant_reference(gref_head); | |
383 | BUG_ON(gnt_list_entry->gref == -ENOSPC); | |
bfe11d6d | 384 | if (!info->feature_persistent) { |
4f503fbd JG |
385 | struct page *indirect_page; |
386 | ||
387 | /* Fetch a pre-allocated page to use for indirect grefs */ | |
73716df7 BL |
388 | BUG_ON(list_empty(&rinfo->indirect_pages)); |
389 | indirect_page = list_first_entry(&rinfo->indirect_pages, | |
4f503fbd JG |
390 | struct page, lru); |
391 | list_del(&indirect_page->lru); | |
392 | gnt_list_entry->page = indirect_page; | |
bfe11d6d | 393 | } |
4f503fbd JG |
394 | grant_foreign_access(gnt_list_entry, info); |
395 | ||
9c1e050c RPM |
396 | return gnt_list_entry; |
397 | } | |
398 | ||
6878c32e KRW |
399 | static const char *op_name(int op) |
400 | { | |
401 | static const char *const names[] = { | |
402 | [BLKIF_OP_READ] = "read", | |
403 | [BLKIF_OP_WRITE] = "write", | |
404 | [BLKIF_OP_WRITE_BARRIER] = "barrier", | |
405 | [BLKIF_OP_FLUSH_DISKCACHE] = "flush", | |
406 | [BLKIF_OP_DISCARD] = "discard" }; | |
407 | ||
408 | if (op < 0 || op >= ARRAY_SIZE(names)) | |
409 | return "unknown"; | |
410 | ||
411 | if (!names[op]) | |
412 | return "reserved"; | |
413 | ||
414 | return names[op]; | |
415 | } | |
0e345826 JB |
416 | static int xlbd_reserve_minors(unsigned int minor, unsigned int nr) |
417 | { | |
418 | unsigned int end = minor + nr; | |
419 | int rc; | |
420 | ||
421 | if (end > nr_minors) { | |
422 | unsigned long *bitmap, *old; | |
423 | ||
f094148a | 424 | bitmap = kcalloc(BITS_TO_LONGS(end), sizeof(*bitmap), |
0e345826 JB |
425 | GFP_KERNEL); |
426 | if (bitmap == NULL) | |
427 | return -ENOMEM; | |
428 | ||
429 | spin_lock(&minor_lock); | |
430 | if (end > nr_minors) { | |
431 | old = minors; | |
432 | memcpy(bitmap, minors, | |
433 | BITS_TO_LONGS(nr_minors) * sizeof(*bitmap)); | |
434 | minors = bitmap; | |
435 | nr_minors = BITS_TO_LONGS(end) * BITS_PER_LONG; | |
436 | } else | |
437 | old = bitmap; | |
438 | spin_unlock(&minor_lock); | |
439 | kfree(old); | |
440 | } | |
441 | ||
442 | spin_lock(&minor_lock); | |
443 | if (find_next_bit(minors, end, minor) >= end) { | |
34ae2e47 | 444 | bitmap_set(minors, minor, nr); |
0e345826 JB |
445 | rc = 0; |
446 | } else | |
447 | rc = -EBUSY; | |
448 | spin_unlock(&minor_lock); | |
449 | ||
450 | return rc; | |
451 | } | |
452 | ||
453 | static void xlbd_release_minors(unsigned int minor, unsigned int nr) | |
454 | { | |
455 | unsigned int end = minor + nr; | |
456 | ||
457 | BUG_ON(end > nr_minors); | |
458 | spin_lock(&minor_lock); | |
34ae2e47 | 459 | bitmap_clear(minors, minor, nr); |
0e345826 JB |
460 | spin_unlock(&minor_lock); |
461 | } | |
462 | ||
9f27ee59 JF |
463 | static void blkif_restart_queue_callback(void *arg) |
464 | { | |
81f35161 BL |
465 | struct blkfront_ring_info *rinfo = (struct blkfront_ring_info *)arg; |
466 | schedule_work(&rinfo->work); | |
9f27ee59 JF |
467 | } |
468 | ||
afe42d7d | 469 | static int blkif_getgeo(struct block_device *bd, struct hd_geometry *hg) |
597592d9 IC |
470 | { |
471 | /* We don't have real geometry info, but let's at least return | |
472 | values consistent with the size of the device */ | |
473 | sector_t nsect = get_capacity(bd->bd_disk); | |
474 | sector_t cylinders = nsect; | |
475 | ||
476 | hg->heads = 0xff; | |
477 | hg->sectors = 0x3f; | |
478 | sector_div(cylinders, hg->heads * hg->sectors); | |
479 | hg->cylinders = cylinders; | |
480 | if ((sector_t)(hg->cylinders + 1) * hg->heads * hg->sectors < nsect) | |
481 | hg->cylinders = 0xffff; | |
482 | return 0; | |
483 | } | |
484 | ||
a63c848b | 485 | static int blkif_ioctl(struct block_device *bdev, fmode_t mode, |
62aa0054 | 486 | unsigned command, unsigned long argument) |
440a01a7 | 487 | { |
a63c848b | 488 | struct blkfront_info *info = bdev->bd_disk->private_data; |
440a01a7 CL |
489 | int i; |
490 | ||
491 | dev_dbg(&info->xbdev->dev, "command: 0x%x, argument: 0x%lx\n", | |
492 | command, (long)argument); | |
493 | ||
494 | switch (command) { | |
495 | case CDROMMULTISESSION: | |
496 | dev_dbg(&info->xbdev->dev, "FIXME: support multisession CDs later\n"); | |
497 | for (i = 0; i < sizeof(struct cdrom_multisession); i++) | |
498 | if (put_user(0, (char __user *)(argument + i))) | |
499 | return -EFAULT; | |
500 | return 0; | |
501 | ||
502 | case CDROM_GET_CAPABILITY: { | |
503 | struct gendisk *gd = info->gd; | |
504 | if (gd->flags & GENHD_FL_CD) | |
505 | return 0; | |
506 | return -EINVAL; | |
507 | } | |
508 | ||
509 | default: | |
510 | /*printk(KERN_ALERT "ioctl %08x not supported by Xen blkdev\n", | |
511 | command);*/ | |
512 | return -EINVAL; /* same return as native Linux */ | |
513 | } | |
514 | ||
515 | return 0; | |
516 | } | |
517 | ||
2e073969 JG |
518 | static unsigned long blkif_ring_get_request(struct blkfront_ring_info *rinfo, |
519 | struct request *req, | |
520 | struct blkif_request **ring_req) | |
521 | { | |
522 | unsigned long id; | |
523 | ||
524 | *ring_req = RING_GET_REQUEST(&rinfo->ring, rinfo->ring.req_prod_pvt); | |
525 | rinfo->ring.req_prod_pvt++; | |
526 | ||
527 | id = get_id_from_freelist(rinfo); | |
528 | rinfo->shadow[id].request = req; | |
6cc56833 JG |
529 | rinfo->shadow[id].status = REQ_WAITING; |
530 | rinfo->shadow[id].associated_id = NO_ASSOCIATED_ID; | |
2e073969 JG |
531 | |
532 | (*ring_req)->u.rw.id = id; | |
533 | ||
534 | return id; | |
535 | } | |
536 | ||
81f35161 | 537 | static int blkif_queue_discard_req(struct request *req, struct blkfront_ring_info *rinfo) |
9f27ee59 | 538 | { |
81f35161 | 539 | struct blkfront_info *info = rinfo->dev_info; |
9f27ee59 | 540 | struct blkif_request *ring_req; |
9f27ee59 | 541 | unsigned long id; |
33204663 JG |
542 | |
543 | /* Fill out a communications ring structure. */ | |
2e073969 | 544 | id = blkif_ring_get_request(rinfo, req, &ring_req); |
33204663 JG |
545 | |
546 | ring_req->operation = BLKIF_OP_DISCARD; | |
547 | ring_req->u.discard.nr_sectors = blk_rq_sectors(req); | |
548 | ring_req->u.discard.id = id; | |
549 | ring_req->u.discard.sector_number = (blkif_sector_t)blk_rq_pos(req); | |
550 | if ((req->cmd_flags & REQ_SECURE) && info->feature_secdiscard) | |
551 | ring_req->u.discard.flag = BLKIF_DISCARD_SECURE; | |
552 | else | |
553 | ring_req->u.discard.flag = 0; | |
554 | ||
33204663 | 555 | /* Keep a private copy so we can reissue requests when recovering. */ |
81f35161 | 556 | rinfo->shadow[id].req = *ring_req; |
33204663 JG |
557 | |
558 | return 0; | |
559 | } | |
560 | ||
c004a6fe JG |
561 | struct setup_rw_req { |
562 | unsigned int grant_idx; | |
563 | struct blkif_request_segment *segments; | |
81f35161 | 564 | struct blkfront_ring_info *rinfo; |
c004a6fe JG |
565 | struct blkif_request *ring_req; |
566 | grant_ref_t gref_head; | |
567 | unsigned int id; | |
568 | /* Only used when persistent grant is used and it's a read request */ | |
569 | bool need_copy; | |
570 | unsigned int bvec_off; | |
571 | char *bvec_data; | |
6cc56833 JG |
572 | |
573 | bool require_extra_req; | |
574 | struct blkif_request *extra_ring_req; | |
c004a6fe JG |
575 | }; |
576 | ||
577 | static void blkif_setup_rw_req_grant(unsigned long gfn, unsigned int offset, | |
578 | unsigned int len, void *data) | |
579 | { | |
580 | struct setup_rw_req *setup = data; | |
581 | int n, ref; | |
582 | struct grant *gnt_list_entry; | |
9f27ee59 | 583 | unsigned int fsect, lsect; |
c004a6fe JG |
584 | /* Convenient aliases */ |
585 | unsigned int grant_idx = setup->grant_idx; | |
586 | struct blkif_request *ring_req = setup->ring_req; | |
81f35161 | 587 | struct blkfront_ring_info *rinfo = setup->rinfo; |
6cc56833 JG |
588 | /* |
589 | * We always use the shadow of the first request to store the list | |
590 | * of grant associated to the block I/O request. This made the | |
591 | * completion more easy to handle even if the block I/O request is | |
592 | * split. | |
593 | */ | |
81f35161 | 594 | struct blk_shadow *shadow = &rinfo->shadow[setup->id]; |
c004a6fe | 595 | |
6cc56833 JG |
596 | if (unlikely(setup->require_extra_req && |
597 | grant_idx >= BLKIF_MAX_SEGMENTS_PER_REQUEST)) { | |
598 | /* | |
599 | * We are using the second request, setup grant_idx | |
600 | * to be the index of the segment array. | |
601 | */ | |
602 | grant_idx -= BLKIF_MAX_SEGMENTS_PER_REQUEST; | |
603 | ring_req = setup->extra_ring_req; | |
604 | } | |
605 | ||
c004a6fe JG |
606 | if ((ring_req->operation == BLKIF_OP_INDIRECT) && |
607 | (grant_idx % GRANTS_PER_INDIRECT_FRAME == 0)) { | |
608 | if (setup->segments) | |
609 | kunmap_atomic(setup->segments); | |
610 | ||
611 | n = grant_idx / GRANTS_PER_INDIRECT_FRAME; | |
73716df7 | 612 | gnt_list_entry = get_indirect_grant(&setup->gref_head, rinfo); |
c004a6fe JG |
613 | shadow->indirect_grants[n] = gnt_list_entry; |
614 | setup->segments = kmap_atomic(gnt_list_entry->page); | |
615 | ring_req->u.indirect.indirect_grefs[n] = gnt_list_entry->gref; | |
616 | } | |
617 | ||
73716df7 | 618 | gnt_list_entry = get_grant(&setup->gref_head, gfn, rinfo); |
c004a6fe | 619 | ref = gnt_list_entry->gref; |
6cc56833 JG |
620 | /* |
621 | * All the grants are stored in the shadow of the first | |
622 | * request. Therefore we have to use the global index. | |
623 | */ | |
624 | shadow->grants_used[setup->grant_idx] = gnt_list_entry; | |
c004a6fe JG |
625 | |
626 | if (setup->need_copy) { | |
627 | void *shared_data; | |
628 | ||
629 | shared_data = kmap_atomic(gnt_list_entry->page); | |
630 | /* | |
631 | * this does not wipe data stored outside the | |
632 | * range sg->offset..sg->offset+sg->length. | |
633 | * Therefore, blkback *could* see data from | |
634 | * previous requests. This is OK as long as | |
635 | * persistent grants are shared with just one | |
636 | * domain. It may need refactoring if this | |
637 | * changes | |
638 | */ | |
639 | memcpy(shared_data + offset, | |
640 | setup->bvec_data + setup->bvec_off, | |
641 | len); | |
642 | ||
643 | kunmap_atomic(shared_data); | |
644 | setup->bvec_off += len; | |
645 | } | |
646 | ||
647 | fsect = offset >> 9; | |
648 | lsect = fsect + (len >> 9) - 1; | |
649 | if (ring_req->operation != BLKIF_OP_INDIRECT) { | |
650 | ring_req->u.rw.seg[grant_idx] = | |
651 | (struct blkif_request_segment) { | |
652 | .gref = ref, | |
653 | .first_sect = fsect, | |
654 | .last_sect = lsect }; | |
655 | } else { | |
656 | setup->segments[grant_idx % GRANTS_PER_INDIRECT_FRAME] = | |
657 | (struct blkif_request_segment) { | |
658 | .gref = ref, | |
659 | .first_sect = fsect, | |
660 | .last_sect = lsect }; | |
661 | } | |
662 | ||
663 | (setup->grant_idx)++; | |
664 | } | |
665 | ||
6cc56833 JG |
666 | static void blkif_setup_extra_req(struct blkif_request *first, |
667 | struct blkif_request *second) | |
668 | { | |
669 | uint16_t nr_segments = first->u.rw.nr_segments; | |
670 | ||
671 | /* | |
672 | * The second request is only present when the first request uses | |
673 | * all its segments. It's always the continuity of the first one. | |
674 | */ | |
675 | first->u.rw.nr_segments = BLKIF_MAX_SEGMENTS_PER_REQUEST; | |
676 | ||
677 | second->u.rw.nr_segments = nr_segments - BLKIF_MAX_SEGMENTS_PER_REQUEST; | |
678 | second->u.rw.sector_number = first->u.rw.sector_number + | |
679 | (BLKIF_MAX_SEGMENTS_PER_REQUEST * XEN_PAGE_SIZE) / 512; | |
680 | ||
681 | second->u.rw.handle = first->u.rw.handle; | |
682 | second->operation = first->operation; | |
683 | } | |
684 | ||
81f35161 | 685 | static int blkif_queue_rw_req(struct request *req, struct blkfront_ring_info *rinfo) |
9f27ee59 | 686 | { |
81f35161 | 687 | struct blkfront_info *info = rinfo->dev_info; |
6cc56833 JG |
688 | struct blkif_request *ring_req, *extra_ring_req = NULL; |
689 | unsigned long id, extra_id = NO_ASSOCIATED_ID; | |
690 | bool require_extra_req = false; | |
c004a6fe JG |
691 | int i; |
692 | struct setup_rw_req setup = { | |
693 | .grant_idx = 0, | |
694 | .segments = NULL, | |
81f35161 | 695 | .rinfo = rinfo, |
c004a6fe JG |
696 | .need_copy = rq_data_dir(req) && info->feature_persistent, |
697 | }; | |
0a8704a5 RPM |
698 | |
699 | /* | |
700 | * Used to store if we are able to queue the request by just using | |
701 | * existing persistent grants, or if we have to get new grants, | |
702 | * as there are not sufficiently many free. | |
703 | */ | |
9e973e64 | 704 | struct scatterlist *sg; |
c004a6fe | 705 | int num_sg, max_grefs, num_grant; |
9f27ee59 | 706 | |
c004a6fe | 707 | max_grefs = req->nr_phys_segments * GRANTS_PER_PSEG; |
c47206e2 RPM |
708 | if (max_grefs > BLKIF_MAX_SEGMENTS_PER_REQUEST) |
709 | /* | |
710 | * If we are using indirect segments we need to account | |
711 | * for the indirect grefs used in the request. | |
712 | */ | |
c004a6fe | 713 | max_grefs += INDIRECT_GREFS(max_grefs); |
402b27f9 | 714 | |
3df0e505 BL |
715 | /* |
716 | * We have to reserve 'max_grefs' grants because persistent | |
717 | * grants are shared by all rings. | |
718 | */ | |
719 | if (max_grefs > 0) | |
720 | if (gnttab_alloc_grant_references(max_grefs, &setup.gref_head) < 0) { | |
0a8704a5 | 721 | gnttab_request_free_callback( |
81f35161 | 722 | &rinfo->callback, |
0a8704a5 | 723 | blkif_restart_queue_callback, |
81f35161 | 724 | rinfo, |
402b27f9 | 725 | max_grefs); |
0a8704a5 RPM |
726 | return 1; |
727 | } | |
9f27ee59 JF |
728 | |
729 | /* Fill out a communications ring structure. */ | |
2e073969 | 730 | id = blkif_ring_get_request(rinfo, req, &ring_req); |
9f27ee59 | 731 | |
81f35161 | 732 | num_sg = blk_rq_map_sg(req->q, req, rinfo->shadow[id].sg); |
c004a6fe JG |
733 | num_grant = 0; |
734 | /* Calculate the number of grant used */ | |
81f35161 | 735 | for_each_sg(rinfo->shadow[id].sg, sg, num_sg, i) |
c004a6fe JG |
736 | num_grant += gnttab_count_grant(sg->offset, sg->length); |
737 | ||
6cc56833 JG |
738 | require_extra_req = info->max_indirect_segments == 0 && |
739 | num_grant > BLKIF_MAX_SEGMENTS_PER_REQUEST; | |
740 | BUG_ON(!HAS_EXTRA_REQ && require_extra_req); | |
741 | ||
81f35161 | 742 | rinfo->shadow[id].num_sg = num_sg; |
6cc56833 JG |
743 | if (num_grant > BLKIF_MAX_SEGMENTS_PER_REQUEST && |
744 | likely(!require_extra_req)) { | |
33204663 JG |
745 | /* |
746 | * The indirect operation can only be a BLKIF_OP_READ or | |
747 | * BLKIF_OP_WRITE | |
748 | */ | |
749 | BUG_ON(req->cmd_flags & (REQ_FLUSH | REQ_FUA)); | |
750 | ring_req->operation = BLKIF_OP_INDIRECT; | |
751 | ring_req->u.indirect.indirect_op = rq_data_dir(req) ? | |
752 | BLKIF_OP_WRITE : BLKIF_OP_READ; | |
753 | ring_req->u.indirect.sector_number = (blkif_sector_t)blk_rq_pos(req); | |
754 | ring_req->u.indirect.handle = info->handle; | |
c004a6fe | 755 | ring_req->u.indirect.nr_segments = num_grant; |
ed30bf31 | 756 | } else { |
33204663 JG |
757 | ring_req->u.rw.sector_number = (blkif_sector_t)blk_rq_pos(req); |
758 | ring_req->u.rw.handle = info->handle; | |
759 | ring_req->operation = rq_data_dir(req) ? | |
760 | BLKIF_OP_WRITE : BLKIF_OP_READ; | |
761 | if (req->cmd_flags & (REQ_FLUSH | REQ_FUA)) { | |
402b27f9 | 762 | /* |
33204663 JG |
763 | * Ideally we can do an unordered flush-to-disk. |
764 | * In case the backend onlysupports barriers, use that. | |
765 | * A barrier request a superset of FUA, so we can | |
766 | * implement it the same way. (It's also a FLUSH+FUA, | |
767 | * since it is guaranteed ordered WRT previous writes.) | |
402b27f9 | 768 | */ |
33204663 JG |
769 | switch (info->feature_flush & |
770 | ((REQ_FLUSH|REQ_FUA))) { | |
771 | case REQ_FLUSH|REQ_FUA: | |
772 | ring_req->operation = | |
773 | BLKIF_OP_WRITE_BARRIER; | |
774 | break; | |
775 | case REQ_FLUSH: | |
776 | ring_req->operation = | |
777 | BLKIF_OP_FLUSH_DISKCACHE; | |
778 | break; | |
779 | default: | |
780 | ring_req->operation = 0; | |
402b27f9 | 781 | } |
402b27f9 | 782 | } |
c004a6fe | 783 | ring_req->u.rw.nr_segments = num_grant; |
6cc56833 JG |
784 | if (unlikely(require_extra_req)) { |
785 | extra_id = blkif_ring_get_request(rinfo, req, | |
786 | &extra_ring_req); | |
787 | /* | |
788 | * Only the first request contains the scatter-gather | |
789 | * list. | |
790 | */ | |
791 | rinfo->shadow[extra_id].num_sg = 0; | |
792 | ||
793 | blkif_setup_extra_req(ring_req, extra_ring_req); | |
794 | ||
795 | /* Link the 2 requests together */ | |
796 | rinfo->shadow[extra_id].associated_id = id; | |
797 | rinfo->shadow[id].associated_id = extra_id; | |
798 | } | |
33204663 | 799 | } |
0a8704a5 | 800 | |
c004a6fe JG |
801 | setup.ring_req = ring_req; |
802 | setup.id = id; | |
6cc56833 JG |
803 | |
804 | setup.require_extra_req = require_extra_req; | |
805 | if (unlikely(require_extra_req)) | |
806 | setup.extra_ring_req = extra_ring_req; | |
807 | ||
81f35161 | 808 | for_each_sg(rinfo->shadow[id].sg, sg, num_sg, i) { |
c004a6fe | 809 | BUG_ON(sg->offset + sg->length > PAGE_SIZE); |
0a8704a5 | 810 | |
c004a6fe JG |
811 | if (setup.need_copy) { |
812 | setup.bvec_off = sg->offset; | |
813 | setup.bvec_data = kmap_atomic(sg_page(sg)); | |
814 | } | |
0a8704a5 | 815 | |
c004a6fe JG |
816 | gnttab_foreach_grant_in_range(sg_page(sg), |
817 | sg->offset, | |
818 | sg->length, | |
819 | blkif_setup_rw_req_grant, | |
820 | &setup); | |
0a8704a5 | 821 | |
c004a6fe JG |
822 | if (setup.need_copy) |
823 | kunmap_atomic(setup.bvec_data); | |
9f27ee59 | 824 | } |
c004a6fe JG |
825 | if (setup.segments) |
826 | kunmap_atomic(setup.segments); | |
9f27ee59 | 827 | |
9f27ee59 | 828 | /* Keep a private copy so we can reissue requests when recovering. */ |
81f35161 | 829 | rinfo->shadow[id].req = *ring_req; |
6cc56833 JG |
830 | if (unlikely(require_extra_req)) |
831 | rinfo->shadow[extra_id].req = *extra_ring_req; | |
9f27ee59 | 832 | |
3df0e505 | 833 | if (max_grefs > 0) |
c004a6fe | 834 | gnttab_free_grant_references(setup.gref_head); |
9f27ee59 JF |
835 | |
836 | return 0; | |
837 | } | |
838 | ||
33204663 JG |
839 | /* |
840 | * Generate a Xen blkfront IO request from a blk layer request. Reads | |
841 | * and writes are handled as expected. | |
842 | * | |
843 | * @req: a request struct | |
844 | */ | |
81f35161 | 845 | static int blkif_queue_request(struct request *req, struct blkfront_ring_info *rinfo) |
33204663 | 846 | { |
81f35161 | 847 | if (unlikely(rinfo->dev_info->connected != BLKIF_STATE_CONNECTED)) |
33204663 JG |
848 | return 1; |
849 | ||
850 | if (unlikely(req->cmd_flags & (REQ_DISCARD | REQ_SECURE))) | |
81f35161 | 851 | return blkif_queue_discard_req(req, rinfo); |
33204663 | 852 | else |
81f35161 | 853 | return blkif_queue_rw_req(req, rinfo); |
33204663 | 854 | } |
9f27ee59 | 855 | |
81f35161 | 856 | static inline void flush_requests(struct blkfront_ring_info *rinfo) |
9f27ee59 JF |
857 | { |
858 | int notify; | |
859 | ||
81f35161 | 860 | RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&rinfo->ring, notify); |
9f27ee59 JF |
861 | |
862 | if (notify) | |
81f35161 | 863 | notify_remote_via_irq(rinfo->irq); |
9f27ee59 JF |
864 | } |
865 | ||
ad42d391 VK |
866 | static inline bool blkif_request_flush_invalid(struct request *req, |
867 | struct blkfront_info *info) | |
0f1ca65e AA |
868 | { |
869 | return ((req->cmd_type != REQ_TYPE_FS) || | |
ad42d391 VK |
870 | ((req->cmd_flags & REQ_FLUSH) && |
871 | !(info->feature_flush & REQ_FLUSH)) || | |
872 | ((req->cmd_flags & REQ_FUA) && | |
873 | !(info->feature_flush & REQ_FUA))); | |
0f1ca65e AA |
874 | } |
875 | ||
907c3eb1 | 876 | static int blkif_queue_rq(struct blk_mq_hw_ctx *hctx, |
6f03a7ff | 877 | const struct blk_mq_queue_data *qd) |
9f27ee59 | 878 | { |
11659569 | 879 | unsigned long flags; |
2a6f71ad BL |
880 | int qid = hctx->queue_num; |
881 | struct blkfront_info *info = hctx->queue->queuedata; | |
882 | struct blkfront_ring_info *rinfo = NULL; | |
9f27ee59 | 883 | |
2a6f71ad BL |
884 | BUG_ON(info->nr_rings <= qid); |
885 | rinfo = &info->rinfo[qid]; | |
907c3eb1 | 886 | blk_mq_start_request(qd->rq); |
11659569 | 887 | spin_lock_irqsave(&rinfo->ring_lock, flags); |
81f35161 | 888 | if (RING_FULL(&rinfo->ring)) |
907c3eb1 | 889 | goto out_busy; |
9f27ee59 | 890 | |
81f35161 | 891 | if (blkif_request_flush_invalid(qd->rq, rinfo->dev_info)) |
907c3eb1 | 892 | goto out_err; |
296b2f6a | 893 | |
81f35161 | 894 | if (blkif_queue_request(qd->rq, rinfo)) |
907c3eb1 | 895 | goto out_busy; |
296b2f6a | 896 | |
81f35161 | 897 | flush_requests(rinfo); |
11659569 | 898 | spin_unlock_irqrestore(&rinfo->ring_lock, flags); |
907c3eb1 | 899 | return BLK_MQ_RQ_QUEUE_OK; |
9f27ee59 | 900 | |
907c3eb1 | 901 | out_err: |
11659569 | 902 | spin_unlock_irqrestore(&rinfo->ring_lock, flags); |
907c3eb1 | 903 | return BLK_MQ_RQ_QUEUE_ERROR; |
9f27ee59 | 904 | |
907c3eb1 | 905 | out_busy: |
11659569 | 906 | spin_unlock_irqrestore(&rinfo->ring_lock, flags); |
907c3eb1 BL |
907 | blk_mq_stop_hw_queue(hctx); |
908 | return BLK_MQ_RQ_QUEUE_BUSY; | |
9f27ee59 JF |
909 | } |
910 | ||
907c3eb1 BL |
911 | static struct blk_mq_ops blkfront_mq_ops = { |
912 | .queue_rq = blkif_queue_rq, | |
913 | .map_queue = blk_mq_map_queue, | |
914 | }; | |
915 | ||
402b27f9 | 916 | static int xlvbd_init_blk_queue(struct gendisk *gd, u16 sector_size, |
7c4d7d71 | 917 | unsigned int physical_sector_size, |
402b27f9 | 918 | unsigned int segments) |
9f27ee59 | 919 | { |
165125e1 | 920 | struct request_queue *rq; |
ed30bf31 | 921 | struct blkfront_info *info = gd->private_data; |
9f27ee59 | 922 | |
907c3eb1 BL |
923 | memset(&info->tag_set, 0, sizeof(info->tag_set)); |
924 | info->tag_set.ops = &blkfront_mq_ops; | |
28d949bc | 925 | info->tag_set.nr_hw_queues = info->nr_rings; |
6cc56833 JG |
926 | if (HAS_EXTRA_REQ && info->max_indirect_segments == 0) { |
927 | /* | |
928 | * When indirect descriptior is not supported, the I/O request | |
929 | * will be split between multiple request in the ring. | |
930 | * To avoid problems when sending the request, divide by | |
931 | * 2 the depth of the queue. | |
932 | */ | |
933 | info->tag_set.queue_depth = BLK_RING_SIZE(info) / 2; | |
934 | } else | |
935 | info->tag_set.queue_depth = BLK_RING_SIZE(info); | |
907c3eb1 BL |
936 | info->tag_set.numa_node = NUMA_NO_NODE; |
937 | info->tag_set.flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_SG_MERGE; | |
938 | info->tag_set.cmd_size = 0; | |
939 | info->tag_set.driver_data = info; | |
940 | ||
941 | if (blk_mq_alloc_tag_set(&info->tag_set)) | |
bde21f73 | 942 | return -EINVAL; |
907c3eb1 BL |
943 | rq = blk_mq_init_queue(&info->tag_set); |
944 | if (IS_ERR(rq)) { | |
945 | blk_mq_free_tag_set(&info->tag_set); | |
bde21f73 | 946 | return PTR_ERR(rq); |
907c3eb1 | 947 | } |
9f27ee59 | 948 | |
2a6f71ad | 949 | rq->queuedata = info; |
66d352e1 | 950 | queue_flag_set_unlocked(QUEUE_FLAG_VIRT, rq); |
9f27ee59 | 951 | |
ed30bf31 LD |
952 | if (info->feature_discard) { |
953 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, rq); | |
954 | blk_queue_max_discard_sectors(rq, get_capacity(gd)); | |
955 | rq->limits.discard_granularity = info->discard_granularity; | |
956 | rq->limits.discard_alignment = info->discard_alignment; | |
5ea42986 KRW |
957 | if (info->feature_secdiscard) |
958 | queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD, rq); | |
ed30bf31 LD |
959 | } |
960 | ||
9f27ee59 | 961 | /* Hard sector size and max sectors impersonate the equiv. hardware. */ |
e1defc4f | 962 | blk_queue_logical_block_size(rq, sector_size); |
7c4d7d71 | 963 | blk_queue_physical_block_size(rq, physical_sector_size); |
c004a6fe | 964 | blk_queue_max_hw_sectors(rq, (segments * XEN_PAGE_SIZE) / 512); |
9f27ee59 JF |
965 | |
966 | /* Each segment in a request is up to an aligned page in size. */ | |
967 | blk_queue_segment_boundary(rq, PAGE_SIZE - 1); | |
968 | blk_queue_max_segment_size(rq, PAGE_SIZE); | |
969 | ||
970 | /* Ensure a merged request will fit in a single I/O ring slot. */ | |
c004a6fe | 971 | blk_queue_max_segments(rq, segments / GRANTS_PER_PSEG); |
9f27ee59 JF |
972 | |
973 | /* Make sure buffer addresses are sector-aligned. */ | |
974 | blk_queue_dma_alignment(rq, 511); | |
975 | ||
1c91fe1a IC |
976 | /* Make sure we don't use bounce buffers. */ |
977 | blk_queue_bounce_limit(rq, BLK_BOUNCE_ANY); | |
978 | ||
9f27ee59 JF |
979 | gd->queue = rq; |
980 | ||
981 | return 0; | |
982 | } | |
983 | ||
fdf9b965 VK |
984 | static const char *flush_info(unsigned int feature_flush) |
985 | { | |
986 | switch (feature_flush & ((REQ_FLUSH | REQ_FUA))) { | |
987 | case REQ_FLUSH|REQ_FUA: | |
988 | return "barrier: enabled;"; | |
989 | case REQ_FLUSH: | |
990 | return "flush diskcache: enabled;"; | |
991 | default: | |
992 | return "barrier or flush: disabled;"; | |
993 | } | |
994 | } | |
9f27ee59 | 995 | |
4913efe4 | 996 | static void xlvbd_flush(struct blkfront_info *info) |
9f27ee59 | 997 | { |
bfd230ac JA |
998 | blk_queue_write_cache(info->rq, info->feature_flush & REQ_FLUSH, |
999 | info->feature_flush & REQ_FUA); | |
fdf9b965 VK |
1000 | pr_info("blkfront: %s: %s %s %s %s %s\n", |
1001 | info->gd->disk_name, flush_info(info->feature_flush), | |
1002 | "persistent grants:", info->feature_persistent ? | |
1003 | "enabled;" : "disabled;", "indirect descriptors:", | |
1004 | info->max_indirect_segments ? "enabled;" : "disabled;"); | |
9f27ee59 JF |
1005 | } |
1006 | ||
c80a4209 SS |
1007 | static int xen_translate_vdev(int vdevice, int *minor, unsigned int *offset) |
1008 | { | |
1009 | int major; | |
1010 | major = BLKIF_MAJOR(vdevice); | |
1011 | *minor = BLKIF_MINOR(vdevice); | |
1012 | switch (major) { | |
1013 | case XEN_IDE0_MAJOR: | |
1014 | *offset = (*minor / 64) + EMULATED_HD_DISK_NAME_OFFSET; | |
1015 | *minor = ((*minor / 64) * PARTS_PER_DISK) + | |
1016 | EMULATED_HD_DISK_MINOR_OFFSET; | |
1017 | break; | |
1018 | case XEN_IDE1_MAJOR: | |
1019 | *offset = (*minor / 64) + 2 + EMULATED_HD_DISK_NAME_OFFSET; | |
1020 | *minor = (((*minor / 64) + 2) * PARTS_PER_DISK) + | |
1021 | EMULATED_HD_DISK_MINOR_OFFSET; | |
1022 | break; | |
1023 | case XEN_SCSI_DISK0_MAJOR: | |
1024 | *offset = (*minor / PARTS_PER_DISK) + EMULATED_SD_DISK_NAME_OFFSET; | |
1025 | *minor = *minor + EMULATED_SD_DISK_MINOR_OFFSET; | |
1026 | break; | |
1027 | case XEN_SCSI_DISK1_MAJOR: | |
1028 | case XEN_SCSI_DISK2_MAJOR: | |
1029 | case XEN_SCSI_DISK3_MAJOR: | |
1030 | case XEN_SCSI_DISK4_MAJOR: | |
1031 | case XEN_SCSI_DISK5_MAJOR: | |
1032 | case XEN_SCSI_DISK6_MAJOR: | |
1033 | case XEN_SCSI_DISK7_MAJOR: | |
1034 | *offset = (*minor / PARTS_PER_DISK) + | |
1035 | ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16) + | |
1036 | EMULATED_SD_DISK_NAME_OFFSET; | |
1037 | *minor = *minor + | |
1038 | ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16 * PARTS_PER_DISK) + | |
1039 | EMULATED_SD_DISK_MINOR_OFFSET; | |
1040 | break; | |
1041 | case XEN_SCSI_DISK8_MAJOR: | |
1042 | case XEN_SCSI_DISK9_MAJOR: | |
1043 | case XEN_SCSI_DISK10_MAJOR: | |
1044 | case XEN_SCSI_DISK11_MAJOR: | |
1045 | case XEN_SCSI_DISK12_MAJOR: | |
1046 | case XEN_SCSI_DISK13_MAJOR: | |
1047 | case XEN_SCSI_DISK14_MAJOR: | |
1048 | case XEN_SCSI_DISK15_MAJOR: | |
1049 | *offset = (*minor / PARTS_PER_DISK) + | |
1050 | ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16) + | |
1051 | EMULATED_SD_DISK_NAME_OFFSET; | |
1052 | *minor = *minor + | |
1053 | ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16 * PARTS_PER_DISK) + | |
1054 | EMULATED_SD_DISK_MINOR_OFFSET; | |
1055 | break; | |
1056 | case XENVBD_MAJOR: | |
1057 | *offset = *minor / PARTS_PER_DISK; | |
1058 | break; | |
1059 | default: | |
1060 | printk(KERN_WARNING "blkfront: your disk configuration is " | |
1061 | "incorrect, please use an xvd device instead\n"); | |
1062 | return -ENODEV; | |
1063 | } | |
1064 | return 0; | |
1065 | } | |
9f27ee59 | 1066 | |
e77c78c0 JB |
1067 | static char *encode_disk_name(char *ptr, unsigned int n) |
1068 | { | |
1069 | if (n >= 26) | |
1070 | ptr = encode_disk_name(ptr, n / 26 - 1); | |
1071 | *ptr = 'a' + n % 26; | |
1072 | return ptr + 1; | |
1073 | } | |
1074 | ||
9246b5f0 CL |
1075 | static int xlvbd_alloc_gendisk(blkif_sector_t capacity, |
1076 | struct blkfront_info *info, | |
7c4d7d71 SB |
1077 | u16 vdisk_info, u16 sector_size, |
1078 | unsigned int physical_sector_size) | |
9f27ee59 JF |
1079 | { |
1080 | struct gendisk *gd; | |
1081 | int nr_minors = 1; | |
c80a4209 | 1082 | int err; |
9246b5f0 CL |
1083 | unsigned int offset; |
1084 | int minor; | |
1085 | int nr_parts; | |
e77c78c0 | 1086 | char *ptr; |
9f27ee59 JF |
1087 | |
1088 | BUG_ON(info->gd != NULL); | |
1089 | BUG_ON(info->rq != NULL); | |
1090 | ||
9246b5f0 CL |
1091 | if ((info->vdevice>>EXT_SHIFT) > 1) { |
1092 | /* this is above the extended range; something is wrong */ | |
1093 | printk(KERN_WARNING "blkfront: vdevice 0x%x is above the extended range; ignoring\n", info->vdevice); | |
1094 | return -ENODEV; | |
1095 | } | |
1096 | ||
1097 | if (!VDEV_IS_EXTENDED(info->vdevice)) { | |
c80a4209 SS |
1098 | err = xen_translate_vdev(info->vdevice, &minor, &offset); |
1099 | if (err) | |
1100 | return err; | |
1101 | nr_parts = PARTS_PER_DISK; | |
9246b5f0 CL |
1102 | } else { |
1103 | minor = BLKIF_MINOR_EXT(info->vdevice); | |
1104 | nr_parts = PARTS_PER_EXT_DISK; | |
c80a4209 | 1105 | offset = minor / nr_parts; |
89153b5c | 1106 | if (xen_hvm_domain() && offset < EMULATED_HD_DISK_NAME_OFFSET + 4) |
c80a4209 SS |
1107 | printk(KERN_WARNING "blkfront: vdevice 0x%x might conflict with " |
1108 | "emulated IDE disks,\n\t choose an xvd device name" | |
1109 | "from xvde on\n", info->vdevice); | |
9246b5f0 | 1110 | } |
e77c78c0 JB |
1111 | if (minor >> MINORBITS) { |
1112 | pr_warn("blkfront: %#x's minor (%#x) out of range; ignoring\n", | |
1113 | info->vdevice, minor); | |
1114 | return -ENODEV; | |
1115 | } | |
9246b5f0 CL |
1116 | |
1117 | if ((minor % nr_parts) == 0) | |
1118 | nr_minors = nr_parts; | |
9f27ee59 | 1119 | |
0e345826 JB |
1120 | err = xlbd_reserve_minors(minor, nr_minors); |
1121 | if (err) | |
1122 | goto out; | |
1123 | err = -ENODEV; | |
1124 | ||
9f27ee59 JF |
1125 | gd = alloc_disk(nr_minors); |
1126 | if (gd == NULL) | |
0e345826 | 1127 | goto release; |
9f27ee59 | 1128 | |
e77c78c0 JB |
1129 | strcpy(gd->disk_name, DEV_NAME); |
1130 | ptr = encode_disk_name(gd->disk_name + sizeof(DEV_NAME) - 1, offset); | |
1131 | BUG_ON(ptr >= gd->disk_name + DISK_NAME_LEN); | |
1132 | if (nr_minors > 1) | |
1133 | *ptr = 0; | |
1134 | else | |
1135 | snprintf(ptr, gd->disk_name + DISK_NAME_LEN - ptr, | |
1136 | "%d", minor & (nr_parts - 1)); | |
9f27ee59 JF |
1137 | |
1138 | gd->major = XENVBD_MAJOR; | |
1139 | gd->first_minor = minor; | |
1140 | gd->fops = &xlvbd_block_fops; | |
1141 | gd->private_data = info; | |
1142 | gd->driverfs_dev = &(info->xbdev->dev); | |
1143 | set_capacity(gd, capacity); | |
1144 | ||
7c4d7d71 | 1145 | if (xlvbd_init_blk_queue(gd, sector_size, physical_sector_size, |
402b27f9 RPM |
1146 | info->max_indirect_segments ? : |
1147 | BLKIF_MAX_SEGMENTS_PER_REQUEST)) { | |
9f27ee59 | 1148 | del_gendisk(gd); |
0e345826 | 1149 | goto release; |
9f27ee59 JF |
1150 | } |
1151 | ||
1152 | info->rq = gd->queue; | |
1153 | info->gd = gd; | |
1154 | ||
4913efe4 | 1155 | xlvbd_flush(info); |
9f27ee59 JF |
1156 | |
1157 | if (vdisk_info & VDISK_READONLY) | |
1158 | set_disk_ro(gd, 1); | |
1159 | ||
1160 | if (vdisk_info & VDISK_REMOVABLE) | |
1161 | gd->flags |= GENHD_FL_REMOVABLE; | |
1162 | ||
1163 | if (vdisk_info & VDISK_CDROM) | |
1164 | gd->flags |= GENHD_FL_CD; | |
1165 | ||
1166 | return 0; | |
1167 | ||
0e345826 JB |
1168 | release: |
1169 | xlbd_release_minors(minor, nr_minors); | |
9f27ee59 JF |
1170 | out: |
1171 | return err; | |
1172 | } | |
1173 | ||
a66b5aeb DS |
1174 | static void xlvbd_release_gendisk(struct blkfront_info *info) |
1175 | { | |
3df0e505 | 1176 | unsigned int minor, nr_minors, i; |
a66b5aeb DS |
1177 | |
1178 | if (info->rq == NULL) | |
1179 | return; | |
1180 | ||
a66b5aeb | 1181 | /* No more blkif_request(). */ |
907c3eb1 | 1182 | blk_mq_stop_hw_queues(info->rq); |
a66b5aeb | 1183 | |
3df0e505 BL |
1184 | for (i = 0; i < info->nr_rings; i++) { |
1185 | struct blkfront_ring_info *rinfo = &info->rinfo[i]; | |
a66b5aeb | 1186 | |
3df0e505 BL |
1187 | /* No more gnttab callback work. */ |
1188 | gnttab_cancel_free_callback(&rinfo->callback); | |
1189 | ||
1190 | /* Flush gnttab callback work. Must be done with no locks held. */ | |
1191 | flush_work(&rinfo->work); | |
1192 | } | |
a66b5aeb DS |
1193 | |
1194 | del_gendisk(info->gd); | |
1195 | ||
1196 | minor = info->gd->first_minor; | |
1197 | nr_minors = info->gd->minors; | |
1198 | xlbd_release_minors(minor, nr_minors); | |
1199 | ||
1200 | blk_cleanup_queue(info->rq); | |
907c3eb1 | 1201 | blk_mq_free_tag_set(&info->tag_set); |
a66b5aeb DS |
1202 | info->rq = NULL; |
1203 | ||
1204 | put_disk(info->gd); | |
1205 | info->gd = NULL; | |
1206 | } | |
1207 | ||
11659569 BL |
1208 | /* Already hold rinfo->ring_lock. */ |
1209 | static inline void kick_pending_request_queues_locked(struct blkfront_ring_info *rinfo) | |
9f27ee59 | 1210 | { |
81f35161 BL |
1211 | if (!RING_FULL(&rinfo->ring)) |
1212 | blk_mq_start_stopped_hw_queues(rinfo->dev_info->rq, true); | |
9f27ee59 JF |
1213 | } |
1214 | ||
11659569 BL |
1215 | static void kick_pending_request_queues(struct blkfront_ring_info *rinfo) |
1216 | { | |
1217 | unsigned long flags; | |
1218 | ||
1219 | spin_lock_irqsave(&rinfo->ring_lock, flags); | |
1220 | kick_pending_request_queues_locked(rinfo); | |
1221 | spin_unlock_irqrestore(&rinfo->ring_lock, flags); | |
1222 | } | |
1223 | ||
9f27ee59 JF |
1224 | static void blkif_restart_queue(struct work_struct *work) |
1225 | { | |
81f35161 | 1226 | struct blkfront_ring_info *rinfo = container_of(work, struct blkfront_ring_info, work); |
9f27ee59 | 1227 | |
81f35161 BL |
1228 | if (rinfo->dev_info->connected == BLKIF_STATE_CONNECTED) |
1229 | kick_pending_request_queues(rinfo); | |
9f27ee59 JF |
1230 | } |
1231 | ||
3df0e505 | 1232 | static void blkif_free_ring(struct blkfront_ring_info *rinfo) |
9f27ee59 | 1233 | { |
73716df7 | 1234 | struct grant *persistent_gnt, *n; |
3df0e505 | 1235 | struct blkfront_info *info = rinfo->dev_info; |
402b27f9 | 1236 | int i, j, segs; |
0a8704a5 | 1237 | |
bfe11d6d RPM |
1238 | /* |
1239 | * Remove indirect pages, this only happens when using indirect | |
1240 | * descriptors but not persistent grants | |
1241 | */ | |
81f35161 | 1242 | if (!list_empty(&rinfo->indirect_pages)) { |
bfe11d6d RPM |
1243 | struct page *indirect_page, *n; |
1244 | ||
1245 | BUG_ON(info->feature_persistent); | |
81f35161 | 1246 | list_for_each_entry_safe(indirect_page, n, &rinfo->indirect_pages, lru) { |
bfe11d6d RPM |
1247 | list_del(&indirect_page->lru); |
1248 | __free_page(indirect_page); | |
1249 | } | |
1250 | } | |
1251 | ||
73716df7 BL |
1252 | /* Remove all persistent grants. */ |
1253 | if (!list_empty(&rinfo->grants)) { | |
1254 | list_for_each_entry_safe(persistent_gnt, n, | |
1255 | &rinfo->grants, node) { | |
1256 | list_del(&persistent_gnt->node); | |
1257 | if (persistent_gnt->gref != GRANT_INVALID_REF) { | |
1258 | gnttab_end_foreign_access(persistent_gnt->gref, | |
1259 | 0, 0UL); | |
1260 | rinfo->persistent_gnts_c--; | |
1261 | } | |
1262 | if (info->feature_persistent) | |
1263 | __free_page(persistent_gnt->page); | |
1264 | kfree(persistent_gnt); | |
1265 | } | |
1266 | } | |
1267 | BUG_ON(rinfo->persistent_gnts_c != 0); | |
1268 | ||
86839c56 | 1269 | for (i = 0; i < BLK_RING_SIZE(info); i++) { |
402b27f9 RPM |
1270 | /* |
1271 | * Clear persistent grants present in requests already | |
1272 | * on the shared ring | |
1273 | */ | |
81f35161 | 1274 | if (!rinfo->shadow[i].request) |
402b27f9 RPM |
1275 | goto free_shadow; |
1276 | ||
81f35161 BL |
1277 | segs = rinfo->shadow[i].req.operation == BLKIF_OP_INDIRECT ? |
1278 | rinfo->shadow[i].req.u.indirect.nr_segments : | |
1279 | rinfo->shadow[i].req.u.rw.nr_segments; | |
402b27f9 | 1280 | for (j = 0; j < segs; j++) { |
81f35161 | 1281 | persistent_gnt = rinfo->shadow[i].grants_used[j]; |
402b27f9 | 1282 | gnttab_end_foreign_access(persistent_gnt->gref, 0, 0UL); |
bfe11d6d | 1283 | if (info->feature_persistent) |
a7a6df22 | 1284 | __free_page(persistent_gnt->page); |
402b27f9 RPM |
1285 | kfree(persistent_gnt); |
1286 | } | |
1287 | ||
81f35161 | 1288 | if (rinfo->shadow[i].req.operation != BLKIF_OP_INDIRECT) |
402b27f9 RPM |
1289 | /* |
1290 | * If this is not an indirect operation don't try to | |
1291 | * free indirect segments | |
1292 | */ | |
1293 | goto free_shadow; | |
1294 | ||
1295 | for (j = 0; j < INDIRECT_GREFS(segs); j++) { | |
81f35161 | 1296 | persistent_gnt = rinfo->shadow[i].indirect_grants[j]; |
402b27f9 | 1297 | gnttab_end_foreign_access(persistent_gnt->gref, 0, 0UL); |
a7a6df22 | 1298 | __free_page(persistent_gnt->page); |
402b27f9 RPM |
1299 | kfree(persistent_gnt); |
1300 | } | |
1301 | ||
1302 | free_shadow: | |
81f35161 BL |
1303 | kfree(rinfo->shadow[i].grants_used); |
1304 | rinfo->shadow[i].grants_used = NULL; | |
1305 | kfree(rinfo->shadow[i].indirect_grants); | |
1306 | rinfo->shadow[i].indirect_grants = NULL; | |
1307 | kfree(rinfo->shadow[i].sg); | |
1308 | rinfo->shadow[i].sg = NULL; | |
402b27f9 RPM |
1309 | } |
1310 | ||
9f27ee59 | 1311 | /* No more gnttab callback work. */ |
81f35161 | 1312 | gnttab_cancel_free_callback(&rinfo->callback); |
9f27ee59 JF |
1313 | |
1314 | /* Flush gnttab callback work. Must be done with no locks held. */ | |
81f35161 | 1315 | flush_work(&rinfo->work); |
9f27ee59 JF |
1316 | |
1317 | /* Free resources associated with old device channel. */ | |
86839c56 | 1318 | for (i = 0; i < info->nr_ring_pages; i++) { |
81f35161 BL |
1319 | if (rinfo->ring_ref[i] != GRANT_INVALID_REF) { |
1320 | gnttab_end_foreign_access(rinfo->ring_ref[i], 0, 0); | |
1321 | rinfo->ring_ref[i] = GRANT_INVALID_REF; | |
86839c56 | 1322 | } |
9f27ee59 | 1323 | } |
81f35161 BL |
1324 | free_pages((unsigned long)rinfo->ring.sring, get_order(info->nr_ring_pages * PAGE_SIZE)); |
1325 | rinfo->ring.sring = NULL; | |
86839c56 | 1326 | |
81f35161 BL |
1327 | if (rinfo->irq) |
1328 | unbind_from_irqhandler(rinfo->irq, rinfo); | |
1329 | rinfo->evtchn = rinfo->irq = 0; | |
3df0e505 | 1330 | } |
9f27ee59 | 1331 | |
3df0e505 BL |
1332 | static void blkif_free(struct blkfront_info *info, int suspend) |
1333 | { | |
3df0e505 BL |
1334 | unsigned int i; |
1335 | ||
1336 | /* Prevent new requests being issued until we fix things up. */ | |
3df0e505 BL |
1337 | info->connected = suspend ? |
1338 | BLKIF_STATE_SUSPENDED : BLKIF_STATE_DISCONNECTED; | |
1339 | /* No more blkif_request(). */ | |
1340 | if (info->rq) | |
1341 | blk_mq_stop_hw_queues(info->rq); | |
1342 | ||
3df0e505 BL |
1343 | for (i = 0; i < info->nr_rings; i++) |
1344 | blkif_free_ring(&info->rinfo[i]); | |
1345 | ||
1346 | kfree(info->rinfo); | |
1347 | info->rinfo = NULL; | |
1348 | info->nr_rings = 0; | |
9f27ee59 JF |
1349 | } |
1350 | ||
c004a6fe JG |
1351 | struct copy_from_grant { |
1352 | const struct blk_shadow *s; | |
1353 | unsigned int grant_idx; | |
1354 | unsigned int bvec_offset; | |
1355 | char *bvec_data; | |
1356 | }; | |
1357 | ||
1358 | static void blkif_copy_from_grant(unsigned long gfn, unsigned int offset, | |
1359 | unsigned int len, void *data) | |
1360 | { | |
1361 | struct copy_from_grant *info = data; | |
1362 | char *shared_data; | |
1363 | /* Convenient aliases */ | |
1364 | const struct blk_shadow *s = info->s; | |
1365 | ||
1366 | shared_data = kmap_atomic(s->grants_used[info->grant_idx]->page); | |
1367 | ||
1368 | memcpy(info->bvec_data + info->bvec_offset, | |
1369 | shared_data + offset, len); | |
1370 | ||
1371 | info->bvec_offset += len; | |
1372 | info->grant_idx++; | |
1373 | ||
1374 | kunmap_atomic(shared_data); | |
1375 | } | |
1376 | ||
6cc56833 JG |
1377 | static enum blk_req_status blkif_rsp_to_req_status(int rsp) |
1378 | { | |
1379 | switch (rsp) | |
1380 | { | |
1381 | case BLKIF_RSP_OKAY: | |
1382 | return REQ_DONE; | |
1383 | case BLKIF_RSP_EOPNOTSUPP: | |
1384 | return REQ_EOPNOTSUPP; | |
1385 | case BLKIF_RSP_ERROR: | |
1386 | /* Fallthrough. */ | |
1387 | default: | |
1388 | return REQ_ERROR; | |
1389 | } | |
1390 | } | |
1391 | ||
1392 | /* | |
1393 | * Get the final status of the block request based on two ring response | |
1394 | */ | |
1395 | static int blkif_get_final_status(enum blk_req_status s1, | |
1396 | enum blk_req_status s2) | |
1397 | { | |
1398 | BUG_ON(s1 == REQ_WAITING); | |
1399 | BUG_ON(s2 == REQ_WAITING); | |
1400 | ||
1401 | if (s1 == REQ_ERROR || s2 == REQ_ERROR) | |
1402 | return BLKIF_RSP_ERROR; | |
1403 | else if (s1 == REQ_EOPNOTSUPP || s2 == REQ_EOPNOTSUPP) | |
1404 | return BLKIF_RSP_EOPNOTSUPP; | |
1405 | return BLKIF_RSP_OKAY; | |
1406 | } | |
1407 | ||
1408 | static bool blkif_completion(unsigned long *id, | |
1409 | struct blkfront_ring_info *rinfo, | |
0a8704a5 | 1410 | struct blkif_response *bret) |
9f27ee59 | 1411 | { |
d62f6918 | 1412 | int i = 0; |
b7649158 | 1413 | struct scatterlist *sg; |
c004a6fe | 1414 | int num_sg, num_grant; |
81f35161 | 1415 | struct blkfront_info *info = rinfo->dev_info; |
6cc56833 | 1416 | struct blk_shadow *s = &rinfo->shadow[*id]; |
c004a6fe | 1417 | struct copy_from_grant data = { |
c004a6fe JG |
1418 | .grant_idx = 0, |
1419 | }; | |
402b27f9 | 1420 | |
c004a6fe | 1421 | num_grant = s->req.operation == BLKIF_OP_INDIRECT ? |
402b27f9 | 1422 | s->req.u.indirect.nr_segments : s->req.u.rw.nr_segments; |
6cc56833 JG |
1423 | |
1424 | /* The I/O request may be split in two. */ | |
1425 | if (unlikely(s->associated_id != NO_ASSOCIATED_ID)) { | |
1426 | struct blk_shadow *s2 = &rinfo->shadow[s->associated_id]; | |
1427 | ||
1428 | /* Keep the status of the current response in shadow. */ | |
1429 | s->status = blkif_rsp_to_req_status(bret->status); | |
1430 | ||
1431 | /* Wait the second response if not yet here. */ | |
1432 | if (s2->status == REQ_WAITING) | |
1433 | return 0; | |
1434 | ||
1435 | bret->status = blkif_get_final_status(s->status, | |
1436 | s2->status); | |
1437 | ||
1438 | /* | |
1439 | * All the grants is stored in the first shadow in order | |
1440 | * to make the completion code simpler. | |
1441 | */ | |
1442 | num_grant += s2->req.u.rw.nr_segments; | |
1443 | ||
1444 | /* | |
1445 | * The two responses may not come in order. Only the | |
1446 | * first request will store the scatter-gather list. | |
1447 | */ | |
1448 | if (s2->num_sg != 0) { | |
1449 | /* Update "id" with the ID of the first response. */ | |
1450 | *id = s->associated_id; | |
1451 | s = s2; | |
1452 | } | |
1453 | ||
1454 | /* | |
1455 | * We don't need anymore the second request, so recycling | |
1456 | * it now. | |
1457 | */ | |
1458 | if (add_id_to_freelist(rinfo, s->associated_id)) | |
1459 | WARN(1, "%s: can't recycle the second part (id = %ld) of the request\n", | |
1460 | info->gd->disk_name, s->associated_id); | |
1461 | } | |
1462 | ||
1463 | data.s = s; | |
c004a6fe | 1464 | num_sg = s->num_sg; |
0a8704a5 | 1465 | |
bfe11d6d | 1466 | if (bret->operation == BLKIF_OP_READ && info->feature_persistent) { |
c004a6fe | 1467 | for_each_sg(s->sg, sg, num_sg, i) { |
b7649158 | 1468 | BUG_ON(sg->offset + sg->length > PAGE_SIZE); |
c004a6fe JG |
1469 | |
1470 | data.bvec_offset = sg->offset; | |
1471 | data.bvec_data = kmap_atomic(sg_page(sg)); | |
1472 | ||
1473 | gnttab_foreach_grant_in_range(sg_page(sg), | |
1474 | sg->offset, | |
1475 | sg->length, | |
1476 | blkif_copy_from_grant, | |
1477 | &data); | |
1478 | ||
1479 | kunmap_atomic(data.bvec_data); | |
0a8704a5 RPM |
1480 | } |
1481 | } | |
1482 | /* Add the persistent grant into the list of free grants */ | |
c004a6fe | 1483 | for (i = 0; i < num_grant; i++) { |
fbe363c4 RPM |
1484 | if (gnttab_query_foreign_access(s->grants_used[i]->gref)) { |
1485 | /* | |
1486 | * If the grant is still mapped by the backend (the | |
1487 | * backend has chosen to make this grant persistent) | |
1488 | * we add it at the head of the list, so it will be | |
1489 | * reused first. | |
1490 | */ | |
bfe11d6d RPM |
1491 | if (!info->feature_persistent) |
1492 | pr_alert_ratelimited("backed has not unmapped grant: %u\n", | |
1493 | s->grants_used[i]->gref); | |
73716df7 BL |
1494 | list_add(&s->grants_used[i]->node, &rinfo->grants); |
1495 | rinfo->persistent_gnts_c++; | |
fbe363c4 RPM |
1496 | } else { |
1497 | /* | |
1498 | * If the grant is not mapped by the backend we end the | |
1499 | * foreign access and add it to the tail of the list, | |
1500 | * so it will not be picked again unless we run out of | |
1501 | * persistent grants. | |
1502 | */ | |
1503 | gnttab_end_foreign_access(s->grants_used[i]->gref, 0, 0UL); | |
1504 | s->grants_used[i]->gref = GRANT_INVALID_REF; | |
73716df7 | 1505 | list_add_tail(&s->grants_used[i]->node, &rinfo->grants); |
fbe363c4 | 1506 | } |
0a8704a5 | 1507 | } |
402b27f9 | 1508 | if (s->req.operation == BLKIF_OP_INDIRECT) { |
c004a6fe | 1509 | for (i = 0; i < INDIRECT_GREFS(num_grant); i++) { |
fbe363c4 | 1510 | if (gnttab_query_foreign_access(s->indirect_grants[i]->gref)) { |
bfe11d6d RPM |
1511 | if (!info->feature_persistent) |
1512 | pr_alert_ratelimited("backed has not unmapped grant: %u\n", | |
1513 | s->indirect_grants[i]->gref); | |
73716df7 BL |
1514 | list_add(&s->indirect_grants[i]->node, &rinfo->grants); |
1515 | rinfo->persistent_gnts_c++; | |
fbe363c4 | 1516 | } else { |
bfe11d6d RPM |
1517 | struct page *indirect_page; |
1518 | ||
fbe363c4 | 1519 | gnttab_end_foreign_access(s->indirect_grants[i]->gref, 0, 0UL); |
bfe11d6d RPM |
1520 | /* |
1521 | * Add the used indirect page back to the list of | |
1522 | * available pages for indirect grefs. | |
1523 | */ | |
7b076750 | 1524 | if (!info->feature_persistent) { |
a7a6df22 | 1525 | indirect_page = s->indirect_grants[i]->page; |
81f35161 | 1526 | list_add(&indirect_page->lru, &rinfo->indirect_pages); |
7b076750 | 1527 | } |
fbe363c4 | 1528 | s->indirect_grants[i]->gref = GRANT_INVALID_REF; |
73716df7 | 1529 | list_add_tail(&s->indirect_grants[i]->node, &rinfo->grants); |
fbe363c4 | 1530 | } |
402b27f9 RPM |
1531 | } |
1532 | } | |
6cc56833 JG |
1533 | |
1534 | return 1; | |
9f27ee59 JF |
1535 | } |
1536 | ||
1537 | static irqreturn_t blkif_interrupt(int irq, void *dev_id) | |
1538 | { | |
1539 | struct request *req; | |
1540 | struct blkif_response *bret; | |
1541 | RING_IDX i, rp; | |
1542 | unsigned long flags; | |
81f35161 BL |
1543 | struct blkfront_ring_info *rinfo = (struct blkfront_ring_info *)dev_id; |
1544 | struct blkfront_info *info = rinfo->dev_info; | |
f4829a9b | 1545 | int error; |
9f27ee59 | 1546 | |
11659569 | 1547 | if (unlikely(info->connected != BLKIF_STATE_CONNECTED)) |
9f27ee59 | 1548 | return IRQ_HANDLED; |
9f27ee59 | 1549 | |
11659569 | 1550 | spin_lock_irqsave(&rinfo->ring_lock, flags); |
9f27ee59 | 1551 | again: |
81f35161 | 1552 | rp = rinfo->ring.sring->rsp_prod; |
9f27ee59 JF |
1553 | rmb(); /* Ensure we see queued responses up to 'rp'. */ |
1554 | ||
81f35161 | 1555 | for (i = rinfo->ring.rsp_cons; i != rp; i++) { |
9f27ee59 | 1556 | unsigned long id; |
9f27ee59 | 1557 | |
81f35161 | 1558 | bret = RING_GET_RESPONSE(&rinfo->ring, i); |
9f27ee59 | 1559 | id = bret->id; |
6878c32e KRW |
1560 | /* |
1561 | * The backend has messed up and given us an id that we would | |
1562 | * never have given to it (we stamp it up to BLK_RING_SIZE - | |
1563 | * look in get_id_from_freelist. | |
1564 | */ | |
86839c56 | 1565 | if (id >= BLK_RING_SIZE(info)) { |
6878c32e KRW |
1566 | WARN(1, "%s: response to %s has incorrect id (%ld)\n", |
1567 | info->gd->disk_name, op_name(bret->operation), id); | |
1568 | /* We can't safely get the 'struct request' as | |
1569 | * the id is busted. */ | |
1570 | continue; | |
1571 | } | |
81f35161 | 1572 | req = rinfo->shadow[id].request; |
9f27ee59 | 1573 | |
6cc56833 JG |
1574 | if (bret->operation != BLKIF_OP_DISCARD) { |
1575 | /* | |
1576 | * We may need to wait for an extra response if the | |
1577 | * I/O request is split in 2 | |
1578 | */ | |
1579 | if (!blkif_completion(&id, rinfo, bret)) | |
1580 | continue; | |
1581 | } | |
9f27ee59 | 1582 | |
81f35161 | 1583 | if (add_id_to_freelist(rinfo, id)) { |
6878c32e KRW |
1584 | WARN(1, "%s: response to %s (id %ld) couldn't be recycled!\n", |
1585 | info->gd->disk_name, op_name(bret->operation), id); | |
1586 | continue; | |
1587 | } | |
9f27ee59 | 1588 | |
f4829a9b | 1589 | error = (bret->status == BLKIF_RSP_OKAY) ? 0 : -EIO; |
9f27ee59 | 1590 | switch (bret->operation) { |
ed30bf31 LD |
1591 | case BLKIF_OP_DISCARD: |
1592 | if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) { | |
1593 | struct request_queue *rq = info->rq; | |
6878c32e KRW |
1594 | printk(KERN_WARNING "blkfront: %s: %s op failed\n", |
1595 | info->gd->disk_name, op_name(bret->operation)); | |
f4829a9b | 1596 | error = -EOPNOTSUPP; |
ed30bf31 | 1597 | info->feature_discard = 0; |
5ea42986 | 1598 | info->feature_secdiscard = 0; |
ed30bf31 | 1599 | queue_flag_clear(QUEUE_FLAG_DISCARD, rq); |
5ea42986 | 1600 | queue_flag_clear(QUEUE_FLAG_SECDISCARD, rq); |
ed30bf31 | 1601 | } |
f4829a9b | 1602 | blk_mq_complete_request(req, error); |
ed30bf31 | 1603 | break; |
edf6ef59 | 1604 | case BLKIF_OP_FLUSH_DISKCACHE: |
9f27ee59 JF |
1605 | case BLKIF_OP_WRITE_BARRIER: |
1606 | if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) { | |
6878c32e KRW |
1607 | printk(KERN_WARNING "blkfront: %s: %s op failed\n", |
1608 | info->gd->disk_name, op_name(bret->operation)); | |
f4829a9b | 1609 | error = -EOPNOTSUPP; |
dcb8baec JF |
1610 | } |
1611 | if (unlikely(bret->status == BLKIF_RSP_ERROR && | |
81f35161 | 1612 | rinfo->shadow[id].req.u.rw.nr_segments == 0)) { |
6878c32e KRW |
1613 | printk(KERN_WARNING "blkfront: %s: empty %s op failed\n", |
1614 | info->gd->disk_name, op_name(bret->operation)); | |
f4829a9b | 1615 | error = -EOPNOTSUPP; |
dcb8baec | 1616 | } |
f4829a9b CH |
1617 | if (unlikely(error)) { |
1618 | if (error == -EOPNOTSUPP) | |
1619 | error = 0; | |
4913efe4 TH |
1620 | info->feature_flush = 0; |
1621 | xlvbd_flush(info); | |
9f27ee59 JF |
1622 | } |
1623 | /* fall through */ | |
1624 | case BLKIF_OP_READ: | |
1625 | case BLKIF_OP_WRITE: | |
1626 | if (unlikely(bret->status != BLKIF_RSP_OKAY)) | |
1627 | dev_dbg(&info->xbdev->dev, "Bad return from blkdev data " | |
1628 | "request: %x\n", bret->status); | |
1629 | ||
f4829a9b | 1630 | blk_mq_complete_request(req, error); |
9f27ee59 JF |
1631 | break; |
1632 | default: | |
1633 | BUG(); | |
1634 | } | |
1635 | } | |
1636 | ||
81f35161 | 1637 | rinfo->ring.rsp_cons = i; |
9f27ee59 | 1638 | |
81f35161 | 1639 | if (i != rinfo->ring.req_prod_pvt) { |
9f27ee59 | 1640 | int more_to_do; |
81f35161 | 1641 | RING_FINAL_CHECK_FOR_RESPONSES(&rinfo->ring, more_to_do); |
9f27ee59 JF |
1642 | if (more_to_do) |
1643 | goto again; | |
1644 | } else | |
81f35161 | 1645 | rinfo->ring.sring->rsp_event = i + 1; |
9f27ee59 | 1646 | |
11659569 | 1647 | kick_pending_request_queues_locked(rinfo); |
9f27ee59 | 1648 | |
11659569 | 1649 | spin_unlock_irqrestore(&rinfo->ring_lock, flags); |
9f27ee59 JF |
1650 | |
1651 | return IRQ_HANDLED; | |
1652 | } | |
1653 | ||
1654 | ||
1655 | static int setup_blkring(struct xenbus_device *dev, | |
81f35161 | 1656 | struct blkfront_ring_info *rinfo) |
9f27ee59 JF |
1657 | { |
1658 | struct blkif_sring *sring; | |
86839c56 | 1659 | int err, i; |
81f35161 | 1660 | struct blkfront_info *info = rinfo->dev_info; |
c004a6fe | 1661 | unsigned long ring_size = info->nr_ring_pages * XEN_PAGE_SIZE; |
9cce2914 | 1662 | grant_ref_t gref[XENBUS_MAX_RING_GRANTS]; |
9f27ee59 | 1663 | |
86839c56 | 1664 | for (i = 0; i < info->nr_ring_pages; i++) |
81f35161 | 1665 | rinfo->ring_ref[i] = GRANT_INVALID_REF; |
9f27ee59 | 1666 | |
86839c56 BL |
1667 | sring = (struct blkif_sring *)__get_free_pages(GFP_NOIO | __GFP_HIGH, |
1668 | get_order(ring_size)); | |
9f27ee59 JF |
1669 | if (!sring) { |
1670 | xenbus_dev_fatal(dev, -ENOMEM, "allocating shared ring"); | |
1671 | return -ENOMEM; | |
1672 | } | |
1673 | SHARED_RING_INIT(sring); | |
81f35161 | 1674 | FRONT_RING_INIT(&rinfo->ring, sring, ring_size); |
9e973e64 | 1675 | |
81f35161 | 1676 | err = xenbus_grant_ring(dev, rinfo->ring.sring, info->nr_ring_pages, gref); |
9f27ee59 | 1677 | if (err < 0) { |
86839c56 | 1678 | free_pages((unsigned long)sring, get_order(ring_size)); |
81f35161 | 1679 | rinfo->ring.sring = NULL; |
9f27ee59 JF |
1680 | goto fail; |
1681 | } | |
86839c56 | 1682 | for (i = 0; i < info->nr_ring_pages; i++) |
81f35161 | 1683 | rinfo->ring_ref[i] = gref[i]; |
9f27ee59 | 1684 | |
81f35161 | 1685 | err = xenbus_alloc_evtchn(dev, &rinfo->evtchn); |
9f27ee59 JF |
1686 | if (err) |
1687 | goto fail; | |
1688 | ||
81f35161 BL |
1689 | err = bind_evtchn_to_irqhandler(rinfo->evtchn, blkif_interrupt, 0, |
1690 | "blkif", rinfo); | |
9f27ee59 JF |
1691 | if (err <= 0) { |
1692 | xenbus_dev_fatal(dev, err, | |
1693 | "bind_evtchn_to_irqhandler failed"); | |
1694 | goto fail; | |
1695 | } | |
81f35161 | 1696 | rinfo->irq = err; |
9f27ee59 JF |
1697 | |
1698 | return 0; | |
1699 | fail: | |
1700 | blkif_free(info, 0); | |
1701 | return err; | |
1702 | } | |
1703 | ||
28d949bc BL |
1704 | /* |
1705 | * Write out per-ring/queue nodes including ring-ref and event-channel, and each | |
1706 | * ring buffer may have multi pages depending on ->nr_ring_pages. | |
1707 | */ | |
1708 | static int write_per_ring_nodes(struct xenbus_transaction xbt, | |
1709 | struct blkfront_ring_info *rinfo, const char *dir) | |
1710 | { | |
1711 | int err; | |
1712 | unsigned int i; | |
1713 | const char *message = NULL; | |
1714 | struct blkfront_info *info = rinfo->dev_info; | |
1715 | ||
1716 | if (info->nr_ring_pages == 1) { | |
1717 | err = xenbus_printf(xbt, dir, "ring-ref", "%u", rinfo->ring_ref[0]); | |
1718 | if (err) { | |
1719 | message = "writing ring-ref"; | |
1720 | goto abort_transaction; | |
1721 | } | |
1722 | } else { | |
1723 | for (i = 0; i < info->nr_ring_pages; i++) { | |
1724 | char ring_ref_name[RINGREF_NAME_LEN]; | |
1725 | ||
1726 | snprintf(ring_ref_name, RINGREF_NAME_LEN, "ring-ref%u", i); | |
1727 | err = xenbus_printf(xbt, dir, ring_ref_name, | |
1728 | "%u", rinfo->ring_ref[i]); | |
1729 | if (err) { | |
1730 | message = "writing ring-ref"; | |
1731 | goto abort_transaction; | |
1732 | } | |
1733 | } | |
1734 | } | |
1735 | ||
1736 | err = xenbus_printf(xbt, dir, "event-channel", "%u", rinfo->evtchn); | |
1737 | if (err) { | |
1738 | message = "writing event-channel"; | |
1739 | goto abort_transaction; | |
1740 | } | |
1741 | ||
1742 | return 0; | |
1743 | ||
1744 | abort_transaction: | |
1745 | xenbus_transaction_end(xbt, 1); | |
1746 | if (message) | |
1747 | xenbus_dev_fatal(info->xbdev, err, "%s", message); | |
1748 | ||
1749 | return err; | |
1750 | } | |
9f27ee59 JF |
1751 | |
1752 | /* Common code used when first setting up, and when resuming. */ | |
203fd61f | 1753 | static int talk_to_blkback(struct xenbus_device *dev, |
9f27ee59 JF |
1754 | struct blkfront_info *info) |
1755 | { | |
1756 | const char *message = NULL; | |
1757 | struct xenbus_transaction xbt; | |
28d949bc BL |
1758 | int err; |
1759 | unsigned int i, max_page_order = 0; | |
86839c56 BL |
1760 | unsigned int ring_page_order = 0; |
1761 | ||
1762 | err = xenbus_scanf(XBT_NIL, info->xbdev->otherend, | |
1763 | "max-ring-page-order", "%u", &max_page_order); | |
1764 | if (err != 1) | |
1765 | info->nr_ring_pages = 1; | |
1766 | else { | |
1767 | ring_page_order = min(xen_blkif_max_ring_order, max_page_order); | |
1768 | info->nr_ring_pages = 1 << ring_page_order; | |
1769 | } | |
9f27ee59 | 1770 | |
3df0e505 | 1771 | for (i = 0; i < info->nr_rings; i++) { |
28d949bc BL |
1772 | struct blkfront_ring_info *rinfo = &info->rinfo[i]; |
1773 | ||
3df0e505 BL |
1774 | /* Create shared ring, alloc event channel. */ |
1775 | err = setup_blkring(dev, rinfo); | |
1776 | if (err) | |
1777 | goto destroy_blkring; | |
1778 | } | |
9f27ee59 JF |
1779 | |
1780 | again: | |
1781 | err = xenbus_transaction_start(&xbt); | |
1782 | if (err) { | |
1783 | xenbus_dev_fatal(dev, err, "starting transaction"); | |
1784 | goto destroy_blkring; | |
1785 | } | |
1786 | ||
28d949bc BL |
1787 | if (info->nr_ring_pages > 1) { |
1788 | err = xenbus_printf(xbt, dev->nodename, "ring-page-order", "%u", | |
1789 | ring_page_order); | |
1790 | if (err) { | |
1791 | message = "writing ring-page-order"; | |
1792 | goto abort_transaction; | |
1793 | } | |
1794 | } | |
3df0e505 | 1795 | |
28d949bc BL |
1796 | /* We already got the number of queues/rings in _probe */ |
1797 | if (info->nr_rings == 1) { | |
1798 | err = write_per_ring_nodes(xbt, &info->rinfo[0], dev->nodename); | |
1799 | if (err) | |
1800 | goto destroy_blkring; | |
1801 | } else { | |
1802 | char *path; | |
1803 | size_t pathsize; | |
3df0e505 | 1804 | |
28d949bc BL |
1805 | err = xenbus_printf(xbt, dev->nodename, "multi-queue-num-queues", "%u", |
1806 | info->nr_rings); | |
3df0e505 | 1807 | if (err) { |
28d949bc | 1808 | message = "writing multi-queue-num-queues"; |
3df0e505 BL |
1809 | goto abort_transaction; |
1810 | } | |
28d949bc BL |
1811 | |
1812 | pathsize = strlen(dev->nodename) + QUEUE_NAME_LEN; | |
1813 | path = kmalloc(pathsize, GFP_KERNEL); | |
1814 | if (!path) { | |
1815 | err = -ENOMEM; | |
1816 | message = "ENOMEM while writing ring references"; | |
1817 | goto abort_transaction; | |
1818 | } | |
1819 | ||
1820 | for (i = 0; i < info->nr_rings; i++) { | |
1821 | memset(path, 0, pathsize); | |
1822 | snprintf(path, pathsize, "%s/queue-%u", dev->nodename, i); | |
1823 | err = write_per_ring_nodes(xbt, &info->rinfo[i], path); | |
1824 | if (err) { | |
1825 | kfree(path); | |
1826 | goto destroy_blkring; | |
1827 | } | |
1828 | } | |
1829 | kfree(path); | |
9f27ee59 | 1830 | } |
3e334239 MA |
1831 | err = xenbus_printf(xbt, dev->nodename, "protocol", "%s", |
1832 | XEN_IO_PROTO_ABI_NATIVE); | |
1833 | if (err) { | |
1834 | message = "writing protocol"; | |
1835 | goto abort_transaction; | |
1836 | } | |
0a8704a5 | 1837 | err = xenbus_printf(xbt, dev->nodename, |
cb5bd4d1 | 1838 | "feature-persistent", "%u", 1); |
0a8704a5 RPM |
1839 | if (err) |
1840 | dev_warn(&dev->dev, | |
1841 | "writing persistent grants feature to xenbus"); | |
9f27ee59 JF |
1842 | |
1843 | err = xenbus_transaction_end(xbt, 0); | |
1844 | if (err) { | |
1845 | if (err == -EAGAIN) | |
1846 | goto again; | |
1847 | xenbus_dev_fatal(dev, err, "completing transaction"); | |
1848 | goto destroy_blkring; | |
1849 | } | |
1850 | ||
3df0e505 BL |
1851 | for (i = 0; i < info->nr_rings; i++) { |
1852 | unsigned int j; | |
28d949bc | 1853 | struct blkfront_ring_info *rinfo = &info->rinfo[i]; |
3df0e505 BL |
1854 | |
1855 | for (j = 0; j < BLK_RING_SIZE(info); j++) | |
1856 | rinfo->shadow[j].req.u.rw.id = j + 1; | |
1857 | rinfo->shadow[BLK_RING_SIZE(info)-1].req.u.rw.id = 0x0fffffff; | |
1858 | } | |
9f27ee59 JF |
1859 | xenbus_switch_state(dev, XenbusStateInitialised); |
1860 | ||
1861 | return 0; | |
1862 | ||
1863 | abort_transaction: | |
1864 | xenbus_transaction_end(xbt, 1); | |
1865 | if (message) | |
1866 | xenbus_dev_fatal(dev, err, "%s", message); | |
1867 | destroy_blkring: | |
1868 | blkif_free(info, 0); | |
3df0e505 | 1869 | |
c31ecf6c KRW |
1870 | kfree(info); |
1871 | dev_set_drvdata(&dev->dev, NULL); | |
1872 | ||
9f27ee59 JF |
1873 | return err; |
1874 | } | |
1875 | ||
3db70a85 BL |
1876 | static int negotiate_mq(struct blkfront_info *info) |
1877 | { | |
1878 | unsigned int backend_max_queues = 0; | |
1879 | int err; | |
1880 | unsigned int i; | |
1881 | ||
1882 | BUG_ON(info->nr_rings); | |
1883 | ||
1884 | /* Check if backend supports multiple queues. */ | |
1885 | err = xenbus_scanf(XBT_NIL, info->xbdev->otherend, | |
1886 | "multi-queue-max-queues", "%u", &backend_max_queues); | |
1887 | if (err < 0) | |
1888 | backend_max_queues = 1; | |
1889 | ||
1890 | info->nr_rings = min(backend_max_queues, xen_blkif_max_queues); | |
1891 | /* We need at least one ring. */ | |
1892 | if (!info->nr_rings) | |
1893 | info->nr_rings = 1; | |
1894 | ||
1895 | info->rinfo = kzalloc(sizeof(struct blkfront_ring_info) * info->nr_rings, GFP_KERNEL); | |
1896 | if (!info->rinfo) { | |
1897 | xenbus_dev_fatal(info->xbdev, -ENOMEM, "allocating ring_info structure"); | |
1898 | return -ENOMEM; | |
1899 | } | |
1900 | ||
1901 | for (i = 0; i < info->nr_rings; i++) { | |
1902 | struct blkfront_ring_info *rinfo; | |
1903 | ||
1904 | rinfo = &info->rinfo[i]; | |
1905 | INIT_LIST_HEAD(&rinfo->indirect_pages); | |
1906 | INIT_LIST_HEAD(&rinfo->grants); | |
1907 | rinfo->dev_info = info; | |
1908 | INIT_WORK(&rinfo->work, blkif_restart_queue); | |
1909 | spin_lock_init(&rinfo->ring_lock); | |
1910 | } | |
1911 | return 0; | |
1912 | } | |
9f27ee59 JF |
1913 | /** |
1914 | * Entry point to this code when a new device is created. Allocate the basic | |
1915 | * structures and the ring buffer for communication with the backend, and | |
1916 | * inform the backend of the appropriate details for those. Switch to | |
1917 | * Initialised state. | |
1918 | */ | |
1919 | static int blkfront_probe(struct xenbus_device *dev, | |
1920 | const struct xenbus_device_id *id) | |
1921 | { | |
86839c56 | 1922 | int err, vdevice; |
9f27ee59 JF |
1923 | struct blkfront_info *info; |
1924 | ||
1925 | /* FIXME: Use dynamic device id if this is not set. */ | |
1926 | err = xenbus_scanf(XBT_NIL, dev->nodename, | |
1927 | "virtual-device", "%i", &vdevice); | |
1928 | if (err != 1) { | |
9246b5f0 CL |
1929 | /* go looking in the extended area instead */ |
1930 | err = xenbus_scanf(XBT_NIL, dev->nodename, "virtual-device-ext", | |
1931 | "%i", &vdevice); | |
1932 | if (err != 1) { | |
1933 | xenbus_dev_fatal(dev, err, "reading virtual-device"); | |
1934 | return err; | |
1935 | } | |
9f27ee59 JF |
1936 | } |
1937 | ||
b98a409b SS |
1938 | if (xen_hvm_domain()) { |
1939 | char *type; | |
1940 | int len; | |
1941 | /* no unplug has been done: do not hook devices != xen vbds */ | |
51c71a3b | 1942 | if (xen_has_pv_and_legacy_disk_devices()) { |
b98a409b SS |
1943 | int major; |
1944 | ||
1945 | if (!VDEV_IS_EXTENDED(vdevice)) | |
1946 | major = BLKIF_MAJOR(vdevice); | |
1947 | else | |
1948 | major = XENVBD_MAJOR; | |
1949 | ||
1950 | if (major != XENVBD_MAJOR) { | |
1951 | printk(KERN_INFO | |
1952 | "%s: HVM does not support vbd %d as xen block device\n", | |
02f1f217 | 1953 | __func__, vdevice); |
b98a409b SS |
1954 | return -ENODEV; |
1955 | } | |
1956 | } | |
1957 | /* do not create a PV cdrom device if we are an HVM guest */ | |
1958 | type = xenbus_read(XBT_NIL, dev->nodename, "device-type", &len); | |
1959 | if (IS_ERR(type)) | |
1960 | return -ENODEV; | |
1961 | if (strncmp(type, "cdrom", 5) == 0) { | |
1962 | kfree(type); | |
c1c5413a SS |
1963 | return -ENODEV; |
1964 | } | |
b98a409b | 1965 | kfree(type); |
c1c5413a | 1966 | } |
9f27ee59 JF |
1967 | info = kzalloc(sizeof(*info), GFP_KERNEL); |
1968 | if (!info) { | |
1969 | xenbus_dev_fatal(dev, -ENOMEM, "allocating info structure"); | |
1970 | return -ENOMEM; | |
1971 | } | |
1972 | ||
28d949bc | 1973 | info->xbdev = dev; |
3db70a85 BL |
1974 | err = negotiate_mq(info); |
1975 | if (err) { | |
3df0e505 | 1976 | kfree(info); |
3db70a85 | 1977 | return err; |
3df0e505 | 1978 | } |
81f35161 | 1979 | |
b70f5fa0 | 1980 | mutex_init(&info->mutex); |
9f27ee59 JF |
1981 | info->vdevice = vdevice; |
1982 | info->connected = BLKIF_STATE_DISCONNECTED; | |
9f27ee59 | 1983 | |
9f27ee59 JF |
1984 | /* Front end dir is a number, which is used as the id. */ |
1985 | info->handle = simple_strtoul(strrchr(dev->nodename, '/')+1, NULL, 0); | |
a1b4b12b | 1986 | dev_set_drvdata(&dev->dev, info); |
9f27ee59 | 1987 | |
9f27ee59 JF |
1988 | return 0; |
1989 | } | |
1990 | ||
4246a0b6 | 1991 | static void split_bio_end(struct bio *bio) |
402b27f9 RPM |
1992 | { |
1993 | struct split_bio *split_bio = bio->bi_private; | |
1994 | ||
402b27f9 RPM |
1995 | if (atomic_dec_and_test(&split_bio->pending)) { |
1996 | split_bio->bio->bi_phys_segments = 0; | |
4246a0b6 CH |
1997 | split_bio->bio->bi_error = bio->bi_error; |
1998 | bio_endio(split_bio->bio); | |
402b27f9 RPM |
1999 | kfree(split_bio); |
2000 | } | |
2001 | bio_put(bio); | |
2002 | } | |
9f27ee59 JF |
2003 | |
2004 | static int blkif_recover(struct blkfront_info *info) | |
2005 | { | |
3df0e505 | 2006 | unsigned int i, r_index; |
402b27f9 | 2007 | struct request *req, *n; |
402b27f9 RPM |
2008 | int rc; |
2009 | struct bio *bio, *cloned_bio; | |
402b27f9 RPM |
2010 | unsigned int segs, offset; |
2011 | int pending, size; | |
2012 | struct split_bio *split_bio; | |
402b27f9 | 2013 | |
3df0e505 | 2014 | blkfront_gather_backend_features(info); |
402b27f9 RPM |
2015 | segs = info->max_indirect_segments ? : BLKIF_MAX_SEGMENTS_PER_REQUEST; |
2016 | blk_queue_max_segments(info->rq, segs); | |
9f27ee59 | 2017 | |
3df0e505 | 2018 | for (r_index = 0; r_index < info->nr_rings; r_index++) { |
7b427a59 | 2019 | struct blkfront_ring_info *rinfo = &info->rinfo[r_index]; |
3df0e505 BL |
2020 | |
2021 | rc = blkfront_setup_indirect(rinfo); | |
7b427a59 | 2022 | if (rc) |
3df0e505 | 2023 | return rc; |
3df0e505 | 2024 | } |
9f27ee59 JF |
2025 | xenbus_switch_state(info->xbdev, XenbusStateConnected); |
2026 | ||
9f27ee59 JF |
2027 | /* Now safe for us to use the shared ring */ |
2028 | info->connected = BLKIF_STATE_CONNECTED; | |
2029 | ||
3df0e505 BL |
2030 | for (r_index = 0; r_index < info->nr_rings; r_index++) { |
2031 | struct blkfront_ring_info *rinfo; | |
2032 | ||
2033 | rinfo = &info->rinfo[r_index]; | |
2034 | /* Kick any other new requests queued since we resumed */ | |
2035 | kick_pending_request_queues(rinfo); | |
2036 | } | |
9f27ee59 | 2037 | |
7b427a59 | 2038 | list_for_each_entry_safe(req, n, &info->requests, queuelist) { |
402b27f9 RPM |
2039 | /* Requeue pending requests (flush or discard) */ |
2040 | list_del_init(&req->queuelist); | |
2041 | BUG_ON(req->nr_phys_segments > segs); | |
907c3eb1 | 2042 | blk_mq_requeue_request(req); |
402b27f9 | 2043 | } |
907c3eb1 | 2044 | blk_mq_kick_requeue_list(info->rq); |
9f27ee59 | 2045 | |
7b427a59 | 2046 | while ((bio = bio_list_pop(&info->bio_list)) != NULL) { |
402b27f9 RPM |
2047 | /* Traverse the list of pending bios and re-queue them */ |
2048 | if (bio_segments(bio) > segs) { | |
2049 | /* | |
2050 | * This bio has more segments than what we can | |
2051 | * handle, we have to split it. | |
2052 | */ | |
2053 | pending = (bio_segments(bio) + segs - 1) / segs; | |
2054 | split_bio = kzalloc(sizeof(*split_bio), GFP_NOIO); | |
2055 | BUG_ON(split_bio == NULL); | |
2056 | atomic_set(&split_bio->pending, pending); | |
2057 | split_bio->bio = bio; | |
2058 | for (i = 0; i < pending; i++) { | |
c004a6fe JG |
2059 | offset = (i * segs * XEN_PAGE_SIZE) >> 9; |
2060 | size = min((unsigned int)(segs * XEN_PAGE_SIZE) >> 9, | |
4f024f37 | 2061 | (unsigned int)bio_sectors(bio) - offset); |
402b27f9 RPM |
2062 | cloned_bio = bio_clone(bio, GFP_NOIO); |
2063 | BUG_ON(cloned_bio == NULL); | |
6678d83f | 2064 | bio_trim(cloned_bio, offset, size); |
402b27f9 RPM |
2065 | cloned_bio->bi_private = split_bio; |
2066 | cloned_bio->bi_end_io = split_bio_end; | |
2067 | submit_bio(cloned_bio->bi_rw, cloned_bio); | |
2068 | } | |
2069 | /* | |
2070 | * Now we have to wait for all those smaller bios to | |
2071 | * end, so we can also end the "parent" bio. | |
2072 | */ | |
2073 | continue; | |
2074 | } | |
2075 | /* We don't need to split this bio */ | |
2076 | submit_bio(bio->bi_rw, bio); | |
2077 | } | |
2078 | ||
9f27ee59 JF |
2079 | return 0; |
2080 | } | |
2081 | ||
2082 | /** | |
2083 | * We are reconnecting to the backend, due to a suspend/resume, or a backend | |
2084 | * driver restart. We tear down our blkif structure and recreate it, but | |
2085 | * leave the device-layer structures intact so that this is transparent to the | |
2086 | * rest of the kernel. | |
2087 | */ | |
2088 | static int blkfront_resume(struct xenbus_device *dev) | |
2089 | { | |
a1b4b12b | 2090 | struct blkfront_info *info = dev_get_drvdata(&dev->dev); |
3db70a85 | 2091 | int err = 0; |
7b427a59 | 2092 | unsigned int i, j; |
9f27ee59 JF |
2093 | |
2094 | dev_dbg(&dev->dev, "blkfront_resume: %s\n", dev->nodename); | |
2095 | ||
7b427a59 BL |
2096 | bio_list_init(&info->bio_list); |
2097 | INIT_LIST_HEAD(&info->requests); | |
2098 | for (i = 0; i < info->nr_rings; i++) { | |
2099 | struct blkfront_ring_info *rinfo = &info->rinfo[i]; | |
2100 | struct bio_list merge_bio; | |
2101 | struct blk_shadow *shadow = rinfo->shadow; | |
2102 | ||
2103 | for (j = 0; j < BLK_RING_SIZE(info); j++) { | |
2104 | /* Not in use? */ | |
2105 | if (!shadow[j].request) | |
2106 | continue; | |
2107 | ||
2108 | /* | |
2109 | * Get the bios in the request so we can re-queue them. | |
2110 | */ | |
2111 | if (shadow[j].request->cmd_flags & | |
2112 | (REQ_FLUSH | REQ_FUA | REQ_DISCARD | REQ_SECURE)) { | |
2113 | /* | |
2114 | * Flush operations don't contain bios, so | |
2115 | * we need to requeue the whole request | |
2116 | */ | |
2117 | list_add(&shadow[j].request->queuelist, &info->requests); | |
2118 | continue; | |
2119 | } | |
2120 | merge_bio.head = shadow[j].request->bio; | |
2121 | merge_bio.tail = shadow[j].request->biotail; | |
2122 | bio_list_merge(&info->bio_list, &merge_bio); | |
2123 | shadow[j].request->bio = NULL; | |
2124 | blk_mq_end_request(shadow[j].request, 0); | |
2125 | } | |
2126 | } | |
2127 | ||
9f27ee59 JF |
2128 | blkif_free(info, info->connected == BLKIF_STATE_CONNECTED); |
2129 | ||
3db70a85 BL |
2130 | err = negotiate_mq(info); |
2131 | if (err) | |
2132 | return err; | |
2133 | ||
203fd61f | 2134 | err = talk_to_blkback(dev, info); |
2a6f71ad BL |
2135 | if (!err) |
2136 | blk_mq_update_nr_hw_queues(&info->tag_set, info->nr_rings); | |
402b27f9 RPM |
2137 | |
2138 | /* | |
2139 | * We have to wait for the backend to switch to | |
2140 | * connected state, since we want to read which | |
2141 | * features it supports. | |
2142 | */ | |
9f27ee59 JF |
2143 | |
2144 | return err; | |
2145 | } | |
2146 | ||
6f03a7ff | 2147 | static void blkfront_closing(struct blkfront_info *info) |
b70f5fa0 DS |
2148 | { |
2149 | struct xenbus_device *xbdev = info->xbdev; | |
2150 | struct block_device *bdev = NULL; | |
2151 | ||
2152 | mutex_lock(&info->mutex); | |
2153 | ||
2154 | if (xbdev->state == XenbusStateClosing) { | |
2155 | mutex_unlock(&info->mutex); | |
2156 | return; | |
2157 | } | |
2158 | ||
2159 | if (info->gd) | |
2160 | bdev = bdget_disk(info->gd, 0); | |
2161 | ||
2162 | mutex_unlock(&info->mutex); | |
2163 | ||
2164 | if (!bdev) { | |
2165 | xenbus_frontend_closed(xbdev); | |
2166 | return; | |
2167 | } | |
2168 | ||
2169 | mutex_lock(&bdev->bd_mutex); | |
2170 | ||
7b32d104 | 2171 | if (bdev->bd_openers) { |
b70f5fa0 DS |
2172 | xenbus_dev_error(xbdev, -EBUSY, |
2173 | "Device in use; refusing to close"); | |
2174 | xenbus_switch_state(xbdev, XenbusStateClosing); | |
2175 | } else { | |
2176 | xlvbd_release_gendisk(info); | |
2177 | xenbus_frontend_closed(xbdev); | |
2178 | } | |
2179 | ||
2180 | mutex_unlock(&bdev->bd_mutex); | |
2181 | bdput(bdev); | |
2182 | } | |
9f27ee59 | 2183 | |
ed30bf31 LD |
2184 | static void blkfront_setup_discard(struct blkfront_info *info) |
2185 | { | |
2186 | int err; | |
ed30bf31 LD |
2187 | unsigned int discard_granularity; |
2188 | unsigned int discard_alignment; | |
5ea42986 | 2189 | unsigned int discard_secure; |
ed30bf31 | 2190 | |
1c8cad6c OH |
2191 | info->feature_discard = 1; |
2192 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2193 | "discard-granularity", "%u", &discard_granularity, | |
2194 | "discard-alignment", "%u", &discard_alignment, | |
2195 | NULL); | |
2196 | if (!err) { | |
2197 | info->discard_granularity = discard_granularity; | |
2198 | info->discard_alignment = discard_alignment; | |
2199 | } | |
2200 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2201 | "discard-secure", "%d", &discard_secure, | |
2202 | NULL); | |
2203 | if (!err) | |
2204 | info->feature_secdiscard = !!discard_secure; | |
ed30bf31 LD |
2205 | } |
2206 | ||
81f35161 | 2207 | static int blkfront_setup_indirect(struct blkfront_ring_info *rinfo) |
402b27f9 | 2208 | { |
c004a6fe | 2209 | unsigned int psegs, grants; |
402b27f9 | 2210 | int err, i; |
81f35161 | 2211 | struct blkfront_info *info = rinfo->dev_info; |
402b27f9 | 2212 | |
6cc56833 JG |
2213 | if (info->max_indirect_segments == 0) { |
2214 | if (!HAS_EXTRA_REQ) | |
2215 | grants = BLKIF_MAX_SEGMENTS_PER_REQUEST; | |
2216 | else { | |
2217 | /* | |
2218 | * When an extra req is required, the maximum | |
2219 | * grants supported is related to the size of the | |
2220 | * Linux block segment. | |
2221 | */ | |
2222 | grants = GRANTS_PER_PSEG; | |
2223 | } | |
2224 | } | |
d50babbe | 2225 | else |
c004a6fe JG |
2226 | grants = info->max_indirect_segments; |
2227 | psegs = grants / GRANTS_PER_PSEG; | |
402b27f9 | 2228 | |
81f35161 | 2229 | err = fill_grant_buffer(rinfo, |
c004a6fe | 2230 | (grants + INDIRECT_GREFS(grants)) * BLK_RING_SIZE(info)); |
402b27f9 RPM |
2231 | if (err) |
2232 | goto out_of_memory; | |
2233 | ||
bfe11d6d RPM |
2234 | if (!info->feature_persistent && info->max_indirect_segments) { |
2235 | /* | |
2236 | * We are using indirect descriptors but not persistent | |
2237 | * grants, we need to allocate a set of pages that can be | |
2238 | * used for mapping indirect grefs | |
2239 | */ | |
c004a6fe | 2240 | int num = INDIRECT_GREFS(grants) * BLK_RING_SIZE(info); |
bfe11d6d | 2241 | |
81f35161 | 2242 | BUG_ON(!list_empty(&rinfo->indirect_pages)); |
bfe11d6d RPM |
2243 | for (i = 0; i < num; i++) { |
2244 | struct page *indirect_page = alloc_page(GFP_NOIO); | |
2245 | if (!indirect_page) | |
2246 | goto out_of_memory; | |
81f35161 | 2247 | list_add(&indirect_page->lru, &rinfo->indirect_pages); |
bfe11d6d RPM |
2248 | } |
2249 | } | |
2250 | ||
86839c56 | 2251 | for (i = 0; i < BLK_RING_SIZE(info); i++) { |
81f35161 BL |
2252 | rinfo->shadow[i].grants_used = kzalloc( |
2253 | sizeof(rinfo->shadow[i].grants_used[0]) * grants, | |
402b27f9 | 2254 | GFP_NOIO); |
81f35161 | 2255 | rinfo->shadow[i].sg = kzalloc(sizeof(rinfo->shadow[i].sg[0]) * psegs, GFP_NOIO); |
402b27f9 | 2256 | if (info->max_indirect_segments) |
81f35161 BL |
2257 | rinfo->shadow[i].indirect_grants = kzalloc( |
2258 | sizeof(rinfo->shadow[i].indirect_grants[0]) * | |
c004a6fe | 2259 | INDIRECT_GREFS(grants), |
402b27f9 | 2260 | GFP_NOIO); |
81f35161 BL |
2261 | if ((rinfo->shadow[i].grants_used == NULL) || |
2262 | (rinfo->shadow[i].sg == NULL) || | |
402b27f9 | 2263 | (info->max_indirect_segments && |
81f35161 | 2264 | (rinfo->shadow[i].indirect_grants == NULL))) |
402b27f9 | 2265 | goto out_of_memory; |
81f35161 | 2266 | sg_init_table(rinfo->shadow[i].sg, psegs); |
402b27f9 RPM |
2267 | } |
2268 | ||
2269 | ||
2270 | return 0; | |
2271 | ||
2272 | out_of_memory: | |
86839c56 | 2273 | for (i = 0; i < BLK_RING_SIZE(info); i++) { |
81f35161 BL |
2274 | kfree(rinfo->shadow[i].grants_used); |
2275 | rinfo->shadow[i].grants_used = NULL; | |
2276 | kfree(rinfo->shadow[i].sg); | |
2277 | rinfo->shadow[i].sg = NULL; | |
2278 | kfree(rinfo->shadow[i].indirect_grants); | |
2279 | rinfo->shadow[i].indirect_grants = NULL; | |
402b27f9 | 2280 | } |
81f35161 | 2281 | if (!list_empty(&rinfo->indirect_pages)) { |
bfe11d6d | 2282 | struct page *indirect_page, *n; |
81f35161 | 2283 | list_for_each_entry_safe(indirect_page, n, &rinfo->indirect_pages, lru) { |
bfe11d6d RPM |
2284 | list_del(&indirect_page->lru); |
2285 | __free_page(indirect_page); | |
2286 | } | |
2287 | } | |
402b27f9 RPM |
2288 | return -ENOMEM; |
2289 | } | |
2290 | ||
d50babbe BL |
2291 | /* |
2292 | * Gather all backend feature-* | |
2293 | */ | |
3df0e505 | 2294 | static void blkfront_gather_backend_features(struct blkfront_info *info) |
d50babbe BL |
2295 | { |
2296 | int err; | |
2297 | int barrier, flush, discard, persistent; | |
2298 | unsigned int indirect_segments; | |
2299 | ||
2300 | info->feature_flush = 0; | |
2301 | ||
2302 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2303 | "feature-barrier", "%d", &barrier, | |
2304 | NULL); | |
2305 | ||
2306 | /* | |
2307 | * If there's no "feature-barrier" defined, then it means | |
2308 | * we're dealing with a very old backend which writes | |
2309 | * synchronously; nothing to do. | |
2310 | * | |
2311 | * If there are barriers, then we use flush. | |
2312 | */ | |
2313 | if (!err && barrier) | |
2314 | info->feature_flush = REQ_FLUSH | REQ_FUA; | |
2315 | /* | |
2316 | * And if there is "feature-flush-cache" use that above | |
2317 | * barriers. | |
2318 | */ | |
2319 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2320 | "feature-flush-cache", "%d", &flush, | |
2321 | NULL); | |
2322 | ||
2323 | if (!err && flush) | |
2324 | info->feature_flush = REQ_FLUSH; | |
2325 | ||
2326 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2327 | "feature-discard", "%d", &discard, | |
2328 | NULL); | |
2329 | ||
2330 | if (!err && discard) | |
2331 | blkfront_setup_discard(info); | |
2332 | ||
2333 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2334 | "feature-persistent", "%u", &persistent, | |
2335 | NULL); | |
2336 | if (err) | |
2337 | info->feature_persistent = 0; | |
2338 | else | |
2339 | info->feature_persistent = persistent; | |
2340 | ||
2341 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2342 | "feature-max-indirect-segments", "%u", &indirect_segments, | |
2343 | NULL); | |
2344 | if (err) | |
2345 | info->max_indirect_segments = 0; | |
2346 | else | |
2347 | info->max_indirect_segments = min(indirect_segments, | |
2348 | xen_blkif_max_segments); | |
d50babbe BL |
2349 | } |
2350 | ||
9f27ee59 JF |
2351 | /* |
2352 | * Invoked when the backend is finally 'ready' (and has told produced | |
2353 | * the details about the physical device - #sectors, size, etc). | |
2354 | */ | |
2355 | static void blkfront_connect(struct blkfront_info *info) | |
2356 | { | |
2357 | unsigned long long sectors; | |
2358 | unsigned long sector_size; | |
7c4d7d71 | 2359 | unsigned int physical_sector_size; |
9f27ee59 | 2360 | unsigned int binfo; |
3df0e505 | 2361 | int err, i; |
9f27ee59 | 2362 | |
1fa73be6 S |
2363 | switch (info->connected) { |
2364 | case BLKIF_STATE_CONNECTED: | |
2365 | /* | |
2366 | * Potentially, the back-end may be signalling | |
2367 | * a capacity change; update the capacity. | |
2368 | */ | |
2369 | err = xenbus_scanf(XBT_NIL, info->xbdev->otherend, | |
2370 | "sectors", "%Lu", §ors); | |
2371 | if (XENBUS_EXIST_ERR(err)) | |
2372 | return; | |
2373 | printk(KERN_INFO "Setting capacity to %Lu\n", | |
2374 | sectors); | |
2375 | set_capacity(info->gd, sectors); | |
2def141e | 2376 | revalidate_disk(info->gd); |
1fa73be6 | 2377 | |
402b27f9 | 2378 | return; |
1fa73be6 | 2379 | case BLKIF_STATE_SUSPENDED: |
402b27f9 RPM |
2380 | /* |
2381 | * If we are recovering from suspension, we need to wait | |
2382 | * for the backend to announce it's features before | |
2383 | * reconnecting, at least we need to know if the backend | |
2384 | * supports indirect descriptors, and how many. | |
2385 | */ | |
2386 | blkif_recover(info); | |
9f27ee59 JF |
2387 | return; |
2388 | ||
b4dddb49 JF |
2389 | default: |
2390 | break; | |
1fa73be6 | 2391 | } |
9f27ee59 JF |
2392 | |
2393 | dev_dbg(&info->xbdev->dev, "%s:%s.\n", | |
2394 | __func__, info->xbdev->otherend); | |
2395 | ||
2396 | err = xenbus_gather(XBT_NIL, info->xbdev->otherend, | |
2397 | "sectors", "%llu", §ors, | |
2398 | "info", "%u", &binfo, | |
2399 | "sector-size", "%lu", §or_size, | |
2400 | NULL); | |
2401 | if (err) { | |
2402 | xenbus_dev_fatal(info->xbdev, err, | |
2403 | "reading backend fields at %s", | |
2404 | info->xbdev->otherend); | |
2405 | return; | |
2406 | } | |
2407 | ||
7c4d7d71 SB |
2408 | /* |
2409 | * physcial-sector-size is a newer field, so old backends may not | |
2410 | * provide this. Assume physical sector size to be the same as | |
2411 | * sector_size in that case. | |
2412 | */ | |
2413 | err = xenbus_scanf(XBT_NIL, info->xbdev->otherend, | |
2414 | "physical-sector-size", "%u", &physical_sector_size); | |
2415 | if (err != 1) | |
2416 | physical_sector_size = sector_size; | |
2417 | ||
3df0e505 BL |
2418 | blkfront_gather_backend_features(info); |
2419 | for (i = 0; i < info->nr_rings; i++) { | |
2420 | err = blkfront_setup_indirect(&info->rinfo[i]); | |
2421 | if (err) { | |
2422 | xenbus_dev_fatal(info->xbdev, err, "setup_indirect at %s", | |
2423 | info->xbdev->otherend); | |
2424 | blkif_free(info, 0); | |
2425 | break; | |
2426 | } | |
402b27f9 RPM |
2427 | } |
2428 | ||
7c4d7d71 SB |
2429 | err = xlvbd_alloc_gendisk(sectors, info, binfo, sector_size, |
2430 | physical_sector_size); | |
9f27ee59 JF |
2431 | if (err) { |
2432 | xenbus_dev_fatal(info->xbdev, err, "xlvbd_add at %s", | |
2433 | info->xbdev->otherend); | |
2434 | return; | |
2435 | } | |
2436 | ||
2437 | xenbus_switch_state(info->xbdev, XenbusStateConnected); | |
2438 | ||
2439 | /* Kick pending requests. */ | |
9f27ee59 | 2440 | info->connected = BLKIF_STATE_CONNECTED; |
3df0e505 BL |
2441 | for (i = 0; i < info->nr_rings; i++) |
2442 | kick_pending_request_queues(&info->rinfo[i]); | |
9f27ee59 JF |
2443 | |
2444 | add_disk(info->gd); | |
1d78d705 CL |
2445 | |
2446 | info->is_ready = 1; | |
9f27ee59 JF |
2447 | } |
2448 | ||
9f27ee59 JF |
2449 | /** |
2450 | * Callback received when the backend's state changes. | |
2451 | */ | |
203fd61f | 2452 | static void blkback_changed(struct xenbus_device *dev, |
9f27ee59 JF |
2453 | enum xenbus_state backend_state) |
2454 | { | |
a1b4b12b | 2455 | struct blkfront_info *info = dev_get_drvdata(&dev->dev); |
9f27ee59 | 2456 | |
203fd61f | 2457 | dev_dbg(&dev->dev, "blkfront:blkback_changed to state %d.\n", backend_state); |
9f27ee59 JF |
2458 | |
2459 | switch (backend_state) { | |
9f27ee59 | 2460 | case XenbusStateInitWait: |
a9b54bb9 BL |
2461 | if (dev->state != XenbusStateInitialising) |
2462 | break; | |
c31ecf6c | 2463 | if (talk_to_blkback(dev, info)) |
8ab0144a | 2464 | break; |
8ab0144a | 2465 | case XenbusStateInitialising: |
9f27ee59 | 2466 | case XenbusStateInitialised: |
b78c9512 NI |
2467 | case XenbusStateReconfiguring: |
2468 | case XenbusStateReconfigured: | |
9f27ee59 | 2469 | case XenbusStateUnknown: |
9f27ee59 JF |
2470 | break; |
2471 | ||
2472 | case XenbusStateConnected: | |
efd15352 BL |
2473 | /* |
2474 | * talk_to_blkback sets state to XenbusStateInitialised | |
2475 | * and blkfront_connect sets it to XenbusStateConnected | |
2476 | * (if connection went OK). | |
2477 | * | |
2478 | * If the backend (or toolstack) decides to poke at backend | |
2479 | * state (and re-trigger the watch by setting the state repeatedly | |
2480 | * to XenbusStateConnected (4)) we need to deal with this. | |
2481 | * This is allowed as this is used to communicate to the guest | |
2482 | * that the size of disk has changed! | |
2483 | */ | |
2484 | if ((dev->state != XenbusStateInitialised) && | |
2485 | (dev->state != XenbusStateConnected)) { | |
c31ecf6c KRW |
2486 | if (talk_to_blkback(dev, info)) |
2487 | break; | |
2488 | } | |
efd15352 | 2489 | |
9f27ee59 JF |
2490 | blkfront_connect(info); |
2491 | break; | |
2492 | ||
36613717 DV |
2493 | case XenbusStateClosed: |
2494 | if (dev->state == XenbusStateClosed) | |
2495 | break; | |
2496 | /* Missed the backend's Closing state -- fallthrough */ | |
9f27ee59 | 2497 | case XenbusStateClosing: |
a54c8f0f CA |
2498 | if (info) |
2499 | blkfront_closing(info); | |
9f27ee59 JF |
2500 | break; |
2501 | } | |
2502 | } | |
2503 | ||
fa1bd359 | 2504 | static int blkfront_remove(struct xenbus_device *xbdev) |
9f27ee59 | 2505 | { |
fa1bd359 DS |
2506 | struct blkfront_info *info = dev_get_drvdata(&xbdev->dev); |
2507 | struct block_device *bdev = NULL; | |
2508 | struct gendisk *disk; | |
9f27ee59 | 2509 | |
fa1bd359 | 2510 | dev_dbg(&xbdev->dev, "%s removed", xbdev->nodename); |
9f27ee59 JF |
2511 | |
2512 | blkif_free(info, 0); | |
2513 | ||
fa1bd359 DS |
2514 | mutex_lock(&info->mutex); |
2515 | ||
2516 | disk = info->gd; | |
2517 | if (disk) | |
2518 | bdev = bdget_disk(disk, 0); | |
2519 | ||
2520 | info->xbdev = NULL; | |
2521 | mutex_unlock(&info->mutex); | |
2522 | ||
2523 | if (!bdev) { | |
2524 | kfree(info); | |
2525 | return 0; | |
2526 | } | |
2527 | ||
2528 | /* | |
2529 | * The xbdev was removed before we reached the Closed | |
2530 | * state. See if it's safe to remove the disk. If the bdev | |
2531 | * isn't closed yet, we let release take care of it. | |
2532 | */ | |
2533 | ||
2534 | mutex_lock(&bdev->bd_mutex); | |
2535 | info = disk->private_data; | |
2536 | ||
d54142c7 DS |
2537 | dev_warn(disk_to_dev(disk), |
2538 | "%s was hot-unplugged, %d stale handles\n", | |
2539 | xbdev->nodename, bdev->bd_openers); | |
2540 | ||
7b32d104 | 2541 | if (info && !bdev->bd_openers) { |
fa1bd359 DS |
2542 | xlvbd_release_gendisk(info); |
2543 | disk->private_data = NULL; | |
0e345826 | 2544 | kfree(info); |
fa1bd359 DS |
2545 | } |
2546 | ||
2547 | mutex_unlock(&bdev->bd_mutex); | |
2548 | bdput(bdev); | |
9f27ee59 JF |
2549 | |
2550 | return 0; | |
2551 | } | |
2552 | ||
1d78d705 CL |
2553 | static int blkfront_is_ready(struct xenbus_device *dev) |
2554 | { | |
a1b4b12b | 2555 | struct blkfront_info *info = dev_get_drvdata(&dev->dev); |
1d78d705 | 2556 | |
5d7ed20e | 2557 | return info->is_ready && info->xbdev; |
1d78d705 CL |
2558 | } |
2559 | ||
a63c848b | 2560 | static int blkif_open(struct block_device *bdev, fmode_t mode) |
9f27ee59 | 2561 | { |
13961743 DS |
2562 | struct gendisk *disk = bdev->bd_disk; |
2563 | struct blkfront_info *info; | |
2564 | int err = 0; | |
6e9624b8 | 2565 | |
2a48fc0a | 2566 | mutex_lock(&blkfront_mutex); |
6e9624b8 | 2567 | |
13961743 DS |
2568 | info = disk->private_data; |
2569 | if (!info) { | |
2570 | /* xbdev gone */ | |
2571 | err = -ERESTARTSYS; | |
2572 | goto out; | |
2573 | } | |
2574 | ||
2575 | mutex_lock(&info->mutex); | |
2576 | ||
2577 | if (!info->gd) | |
2578 | /* xbdev is closed */ | |
2579 | err = -ERESTARTSYS; | |
2580 | ||
2581 | mutex_unlock(&info->mutex); | |
2582 | ||
13961743 | 2583 | out: |
2a48fc0a | 2584 | mutex_unlock(&blkfront_mutex); |
13961743 | 2585 | return err; |
9f27ee59 JF |
2586 | } |
2587 | ||
db2a144b | 2588 | static void blkif_release(struct gendisk *disk, fmode_t mode) |
9f27ee59 | 2589 | { |
a63c848b | 2590 | struct blkfront_info *info = disk->private_data; |
7fd152f4 DS |
2591 | struct block_device *bdev; |
2592 | struct xenbus_device *xbdev; | |
2593 | ||
2a48fc0a | 2594 | mutex_lock(&blkfront_mutex); |
7fd152f4 DS |
2595 | |
2596 | bdev = bdget_disk(disk, 0); | |
7fd152f4 | 2597 | |
2f089cb8 FP |
2598 | if (!bdev) { |
2599 | WARN(1, "Block device %s yanked out from us!\n", disk->disk_name); | |
2600 | goto out_mutex; | |
2601 | } | |
acfca3c6 DS |
2602 | if (bdev->bd_openers) |
2603 | goto out; | |
2604 | ||
7fd152f4 DS |
2605 | /* |
2606 | * Check if we have been instructed to close. We will have | |
2607 | * deferred this request, because the bdev was still open. | |
2608 | */ | |
2609 | ||
2610 | mutex_lock(&info->mutex); | |
2611 | xbdev = info->xbdev; | |
2612 | ||
2613 | if (xbdev && xbdev->state == XenbusStateClosing) { | |
2614 | /* pending switch to state closed */ | |
d54142c7 | 2615 | dev_info(disk_to_dev(bdev->bd_disk), "releasing disk\n"); |
7fd152f4 DS |
2616 | xlvbd_release_gendisk(info); |
2617 | xenbus_frontend_closed(info->xbdev); | |
2618 | } | |
2619 | ||
2620 | mutex_unlock(&info->mutex); | |
2621 | ||
2622 | if (!xbdev) { | |
2623 | /* sudden device removal */ | |
d54142c7 | 2624 | dev_info(disk_to_dev(bdev->bd_disk), "releasing disk\n"); |
7fd152f4 DS |
2625 | xlvbd_release_gendisk(info); |
2626 | disk->private_data = NULL; | |
2627 | kfree(info); | |
9f27ee59 | 2628 | } |
7fd152f4 | 2629 | |
a4cc14ec | 2630 | out: |
dad5cf65 | 2631 | bdput(bdev); |
2f089cb8 | 2632 | out_mutex: |
2a48fc0a | 2633 | mutex_unlock(&blkfront_mutex); |
9f27ee59 JF |
2634 | } |
2635 | ||
83d5cde4 | 2636 | static const struct block_device_operations xlvbd_block_fops = |
9f27ee59 JF |
2637 | { |
2638 | .owner = THIS_MODULE, | |
a63c848b AV |
2639 | .open = blkif_open, |
2640 | .release = blkif_release, | |
597592d9 | 2641 | .getgeo = blkif_getgeo, |
8a6cfeb6 | 2642 | .ioctl = blkif_ioctl, |
9f27ee59 JF |
2643 | }; |
2644 | ||
2645 | ||
ec9c42ec | 2646 | static const struct xenbus_device_id blkfront_ids[] = { |
9f27ee59 JF |
2647 | { "vbd" }, |
2648 | { "" } | |
2649 | }; | |
2650 | ||
95afae48 DV |
2651 | static struct xenbus_driver blkfront_driver = { |
2652 | .ids = blkfront_ids, | |
9f27ee59 JF |
2653 | .probe = blkfront_probe, |
2654 | .remove = blkfront_remove, | |
2655 | .resume = blkfront_resume, | |
203fd61f | 2656 | .otherend_changed = blkback_changed, |
1d78d705 | 2657 | .is_ready = blkfront_is_ready, |
95afae48 | 2658 | }; |
9f27ee59 JF |
2659 | |
2660 | static int __init xlblk_init(void) | |
2661 | { | |
469738e6 | 2662 | int ret; |
28d949bc | 2663 | int nr_cpus = num_online_cpus(); |
469738e6 | 2664 | |
6e833587 | 2665 | if (!xen_domain()) |
9f27ee59 JF |
2666 | return -ENODEV; |
2667 | ||
9cce2914 | 2668 | if (xen_blkif_max_ring_order > XENBUS_MAX_RING_GRANT_ORDER) { |
86839c56 | 2669 | pr_info("Invalid max_ring_order (%d), will use default max: %d.\n", |
9cce2914 | 2670 | xen_blkif_max_ring_order, XENBUS_MAX_RING_GRANT_ORDER); |
45fc8264 | 2671 | xen_blkif_max_ring_order = XENBUS_MAX_RING_GRANT_ORDER; |
86839c56 BL |
2672 | } |
2673 | ||
28d949bc BL |
2674 | if (xen_blkif_max_queues > nr_cpus) { |
2675 | pr_info("Invalid max_queues (%d), will use default max: %d.\n", | |
2676 | xen_blkif_max_queues, nr_cpus); | |
2677 | xen_blkif_max_queues = nr_cpus; | |
2678 | } | |
2679 | ||
51c71a3b | 2680 | if (!xen_has_pv_disk_devices()) |
b9136d20 IM |
2681 | return -ENODEV; |
2682 | ||
9f27ee59 JF |
2683 | if (register_blkdev(XENVBD_MAJOR, DEV_NAME)) { |
2684 | printk(KERN_WARNING "xen_blk: can't get major %d with name %s\n", | |
2685 | XENVBD_MAJOR, DEV_NAME); | |
2686 | return -ENODEV; | |
2687 | } | |
2688 | ||
73db144b | 2689 | ret = xenbus_register_frontend(&blkfront_driver); |
469738e6 LE |
2690 | if (ret) { |
2691 | unregister_blkdev(XENVBD_MAJOR, DEV_NAME); | |
2692 | return ret; | |
2693 | } | |
2694 | ||
2695 | return 0; | |
9f27ee59 JF |
2696 | } |
2697 | module_init(xlblk_init); | |
2698 | ||
2699 | ||
5a60d0cd | 2700 | static void __exit xlblk_exit(void) |
9f27ee59 | 2701 | { |
8605067f JB |
2702 | xenbus_unregister_driver(&blkfront_driver); |
2703 | unregister_blkdev(XENVBD_MAJOR, DEV_NAME); | |
2704 | kfree(minors); | |
9f27ee59 JF |
2705 | } |
2706 | module_exit(xlblk_exit); | |
2707 | ||
2708 | MODULE_DESCRIPTION("Xen virtual block device frontend"); | |
2709 | MODULE_LICENSE("GPL"); | |
2710 | MODULE_ALIAS_BLOCKDEV_MAJOR(XENVBD_MAJOR); | |
d2f0c52b | 2711 | MODULE_ALIAS("xen:vbd"); |
4f93f09b | 2712 | MODULE_ALIAS("xenblk"); |