Commit | Line | Data |
---|---|---|
4d05a28d KRW |
1 | /* Xenbus code for blkif backend |
2 | Copyright (C) 2005 Rusty Russell <rusty@rustcorp.com.au> | |
3 | Copyright (C) 2005 XenSource Ltd | |
4 | ||
5 | This program is free software; you can redistribute it and/or modify | |
6 | it under the terms of the GNU General Public License as published by | |
7 | the Free Software Foundation; either version 2 of the License, or | |
8 | (at your option) any later version. | |
9 | ||
10 | This program is distributed in the hope that it will be useful, | |
11 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
12 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
13 | GNU General Public License for more details. | |
14 | ||
4d05a28d KRW |
15 | */ |
16 | ||
77387b82 TC |
17 | #define pr_fmt(fmt) "xen-blkback: " fmt |
18 | ||
4d05a28d KRW |
19 | #include <stdarg.h> |
20 | #include <linux/module.h> | |
21 | #include <linux/kthread.h> | |
ee9ff853 KRW |
22 | #include <xen/events.h> |
23 | #include <xen/grant_table.h> | |
4d05a28d KRW |
24 | #include "common.h" |
25 | ||
fa3184b8 | 26 | /* On the XenBus the max length of 'ring-ref%u'. */ |
86839c56 | 27 | #define RINGREF_NAME_LEN (20) |
1375590d | 28 | |
d6091b21 | 29 | struct backend_info { |
01f37f2d | 30 | struct xenbus_device *dev; |
51854322 | 31 | struct xen_blkif *blkif; |
01f37f2d KRW |
32 | struct xenbus_watch backend_watch; |
33 | unsigned major; | |
34 | unsigned minor; | |
35 | char *mode; | |
4d05a28d KRW |
36 | }; |
37 | ||
8b6bf747 | 38 | static struct kmem_cache *xen_blkif_cachep; |
4d05a28d KRW |
39 | static void connect(struct backend_info *); |
40 | static int connect_ring(struct backend_info *); | |
41 | static void backend_changed(struct xenbus_watch *, const char **, | |
42 | unsigned int); | |
814d04e7 VP |
43 | static void xen_blkif_free(struct xen_blkif *blkif); |
44 | static void xen_vbd_free(struct xen_vbd *vbd); | |
4d05a28d | 45 | |
8b6bf747 | 46 | struct xenbus_device *xen_blkbk_xenbus(struct backend_info *be) |
98e036a3 JF |
47 | { |
48 | return be->dev; | |
49 | } | |
50 | ||
814d04e7 VP |
51 | /* |
52 | * The last request could free the device from softirq context and | |
53 | * xen_blkif_free() can sleep. | |
54 | */ | |
55 | static void xen_blkif_deferred_free(struct work_struct *work) | |
56 | { | |
57 | struct xen_blkif *blkif; | |
58 | ||
59 | blkif = container_of(work, struct xen_blkif, free_work); | |
60 | xen_blkif_free(blkif); | |
61 | } | |
62 | ||
30fd1502 | 63 | static int blkback_name(struct xen_blkif *blkif, char *buf) |
4d05a28d KRW |
64 | { |
65 | char *devpath, *devname; | |
66 | struct xenbus_device *dev = blkif->be->dev; | |
67 | ||
68 | devpath = xenbus_read(XBT_NIL, dev->nodename, "dev", NULL); | |
69 | if (IS_ERR(devpath)) | |
70 | return PTR_ERR(devpath); | |
71 | ||
d6091b21 KRW |
72 | devname = strstr(devpath, "/dev/"); |
73 | if (devname != NULL) | |
4d05a28d KRW |
74 | devname += strlen("/dev/"); |
75 | else | |
76 | devname = devpath; | |
77 | ||
fa3184b8 | 78 | snprintf(buf, TASK_COMM_LEN, "%d.%s", blkif->domid, devname); |
4d05a28d KRW |
79 | kfree(devpath); |
80 | ||
81 | return 0; | |
82 | } | |
83 | ||
30fd1502 | 84 | static void xen_update_blkif_status(struct xen_blkif *blkif) |
4d05a28d KRW |
85 | { |
86 | int err; | |
fa3184b8 | 87 | char name[TASK_COMM_LEN]; |
2fb1ef4f KRW |
88 | struct xen_blkif_ring *ring; |
89 | int i; | |
4d05a28d KRW |
90 | |
91 | /* Not ready to connect? */ | |
2fb1ef4f | 92 | if (!blkif->rings || !blkif->rings[0].irq || !blkif->vbd.bdev) |
4d05a28d KRW |
93 | return; |
94 | ||
95 | /* Already connected? */ | |
96 | if (blkif->be->dev->state == XenbusStateConnected) | |
97 | return; | |
98 | ||
99 | /* Attempt to connect: exit if we fail to. */ | |
100 | connect(blkif->be); | |
101 | if (blkif->be->dev->state != XenbusStateConnected) | |
102 | return; | |
103 | ||
104 | err = blkback_name(blkif, name); | |
105 | if (err) { | |
106 | xenbus_dev_error(blkif->be->dev, err, "get blkback dev name"); | |
107 | return; | |
108 | } | |
109 | ||
cbf46290 CL |
110 | err = filemap_write_and_wait(blkif->vbd.bdev->bd_inode->i_mapping); |
111 | if (err) { | |
112 | xenbus_dev_error(blkif->be->dev, err, "block flush"); | |
113 | return; | |
114 | } | |
115 | invalidate_inode_pages2(blkif->vbd.bdev->bd_inode->i_mapping); | |
116 | ||
2fb1ef4f KRW |
117 | for (i = 0; i < blkif->nr_rings; i++) { |
118 | ring = &blkif->rings[i]; | |
119 | ring->xenblkd = kthread_run(xen_blkif_schedule, ring, "%s-%d", name, i); | |
120 | if (IS_ERR(ring->xenblkd)) { | |
121 | err = PTR_ERR(ring->xenblkd); | |
122 | ring->xenblkd = NULL; | |
123 | xenbus_dev_fatal(blkif->be->dev, err, | |
124 | "start %s-%d xenblkd", name, i); | |
125 | goto out; | |
126 | } | |
127 | } | |
128 | return; | |
129 | ||
130 | out: | |
131 | while (--i >= 0) { | |
132 | ring = &blkif->rings[i]; | |
133 | kthread_stop(ring->xenblkd); | |
4d05a28d | 134 | } |
2fb1ef4f KRW |
135 | return; |
136 | } | |
137 | ||
138 | static int xen_blkif_alloc_rings(struct xen_blkif *blkif) | |
139 | { | |
140 | unsigned int r; | |
141 | ||
142 | blkif->rings = kzalloc(blkif->nr_rings * sizeof(struct xen_blkif_ring), GFP_KERNEL); | |
143 | if (!blkif->rings) | |
144 | return -ENOMEM; | |
145 | ||
146 | for (r = 0; r < blkif->nr_rings; r++) { | |
147 | struct xen_blkif_ring *ring = &blkif->rings[r]; | |
148 | ||
149 | spin_lock_init(&ring->blk_ring_lock); | |
150 | init_waitqueue_head(&ring->wq); | |
151 | INIT_LIST_HEAD(&ring->pending_free); | |
d4bf0065 BL |
152 | INIT_LIST_HEAD(&ring->persistent_purge_list); |
153 | INIT_WORK(&ring->persistent_purge_work, xen_blkbk_unmap_purged_grants); | |
154 | spin_lock_init(&ring->free_pages_lock); | |
155 | INIT_LIST_HEAD(&ring->free_pages); | |
2fb1ef4f KRW |
156 | |
157 | spin_lock_init(&ring->pending_free_lock); | |
158 | init_waitqueue_head(&ring->pending_free_wq); | |
159 | init_waitqueue_head(&ring->shutdown_wq); | |
160 | ring->blkif = blkif; | |
db6fbc10 | 161 | ring->st_print = jiffies; |
2fb1ef4f KRW |
162 | xen_blkif_get(blkif); |
163 | } | |
164 | ||
165 | return 0; | |
4d05a28d KRW |
166 | } |
167 | ||
30fd1502 | 168 | static struct xen_blkif *xen_blkif_alloc(domid_t domid) |
ee9ff853 | 169 | { |
30fd1502 | 170 | struct xen_blkif *blkif; |
ee9ff853 | 171 | |
402b27f9 | 172 | BUILD_BUG_ON(MAX_INDIRECT_PAGES > BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST); |
ee9ff853 | 173 | |
654dbef2 | 174 | blkif = kmem_cache_zalloc(xen_blkif_cachep, GFP_KERNEL); |
ee9ff853 KRW |
175 | if (!blkif) |
176 | return ERR_PTR(-ENOMEM); | |
177 | ||
ee9ff853 | 178 | blkif->domid = domid; |
ee9ff853 | 179 | atomic_set(&blkif->refcnt, 1); |
29bde093 | 180 | init_completion(&blkif->drain_complete); |
59795700 | 181 | INIT_WORK(&blkif->free_work, xen_blkif_deferred_free); |
ee9ff853 KRW |
182 | |
183 | return blkif; | |
184 | } | |
185 | ||
59795700 | 186 | static int xen_blkif_map(struct xen_blkif_ring *ring, grant_ref_t *gref, |
86839c56 | 187 | unsigned int nr_grefs, unsigned int evtchn) |
ee9ff853 KRW |
188 | { |
189 | int err; | |
59795700 | 190 | struct xen_blkif *blkif = ring->blkif; |
ee9ff853 KRW |
191 | |
192 | /* Already connected through? */ | |
59795700 | 193 | if (ring->irq) |
ee9ff853 KRW |
194 | return 0; |
195 | ||
86839c56 | 196 | err = xenbus_map_ring_valloc(blkif->be->dev, gref, nr_grefs, |
59795700 | 197 | &ring->blk_ring); |
2d073846 | 198 | if (err < 0) |
ee9ff853 | 199 | return err; |
ee9ff853 KRW |
200 | |
201 | switch (blkif->blk_protocol) { | |
202 | case BLKIF_PROTOCOL_NATIVE: | |
203 | { | |
204 | struct blkif_sring *sring; | |
59795700 BL |
205 | sring = (struct blkif_sring *)ring->blk_ring; |
206 | BACK_RING_INIT(&ring->blk_rings.native, sring, | |
67de5dfb | 207 | XEN_PAGE_SIZE * nr_grefs); |
ee9ff853 KRW |
208 | break; |
209 | } | |
210 | case BLKIF_PROTOCOL_X86_32: | |
211 | { | |
212 | struct blkif_x86_32_sring *sring_x86_32; | |
59795700 BL |
213 | sring_x86_32 = (struct blkif_x86_32_sring *)ring->blk_ring; |
214 | BACK_RING_INIT(&ring->blk_rings.x86_32, sring_x86_32, | |
67de5dfb | 215 | XEN_PAGE_SIZE * nr_grefs); |
ee9ff853 KRW |
216 | break; |
217 | } | |
218 | case BLKIF_PROTOCOL_X86_64: | |
219 | { | |
220 | struct blkif_x86_64_sring *sring_x86_64; | |
59795700 BL |
221 | sring_x86_64 = (struct blkif_x86_64_sring *)ring->blk_ring; |
222 | BACK_RING_INIT(&ring->blk_rings.x86_64, sring_x86_64, | |
67de5dfb | 223 | XEN_PAGE_SIZE * nr_grefs); |
ee9ff853 KRW |
224 | break; |
225 | } | |
226 | default: | |
227 | BUG(); | |
228 | } | |
229 | ||
8b6bf747 KRW |
230 | err = bind_interdomain_evtchn_to_irqhandler(blkif->domid, evtchn, |
231 | xen_blkif_be_int, 0, | |
59795700 | 232 | "blkif-backend", ring); |
ee9ff853 | 233 | if (err < 0) { |
59795700 BL |
234 | xenbus_unmap_ring_vfree(blkif->be->dev, ring->blk_ring); |
235 | ring->blk_rings.common.sring = NULL; | |
ee9ff853 KRW |
236 | return err; |
237 | } | |
59795700 | 238 | ring->irq = err; |
ee9ff853 KRW |
239 | |
240 | return 0; | |
241 | } | |
242 | ||
814d04e7 | 243 | static int xen_blkif_disconnect(struct xen_blkif *blkif) |
ee9ff853 | 244 | { |
f929d42c | 245 | struct pending_req *req, *n; |
2fb1ef4f | 246 | unsigned int j, r; |
f929d42c | 247 | |
2fb1ef4f KRW |
248 | for (r = 0; r < blkif->nr_rings; r++) { |
249 | struct xen_blkif_ring *ring = &blkif->rings[r]; | |
250 | unsigned int i = 0; | |
ee9ff853 | 251 | |
2fb1ef4f KRW |
252 | if (ring->xenblkd) { |
253 | kthread_stop(ring->xenblkd); | |
254 | wake_up(&ring->shutdown_wq); | |
255 | ring->xenblkd = NULL; | |
256 | } | |
ee9ff853 | 257 | |
2fb1ef4f KRW |
258 | /* The above kthread_stop() guarantees that at this point we |
259 | * don't have any discard_io or other_io requests. So, checking | |
260 | * for inflight IO is enough. | |
261 | */ | |
262 | if (atomic_read(&ring->inflight) > 0) | |
263 | return -EBUSY; | |
ee9ff853 | 264 | |
2fb1ef4f KRW |
265 | if (ring->irq) { |
266 | unbind_from_irqhandler(ring->irq, ring); | |
267 | ring->irq = 0; | |
268 | } | |
814d04e7 | 269 | |
2fb1ef4f KRW |
270 | if (ring->blk_rings.common.sring) { |
271 | xenbus_unmap_ring_vfree(blkif->be->dev, ring->blk_ring); | |
272 | ring->blk_rings.common.sring = NULL; | |
273 | } | |
12ea7296 | 274 | |
2fb1ef4f KRW |
275 | /* Remove all persistent grants and the cache of ballooned pages. */ |
276 | xen_blkbk_free_caches(ring); | |
f929d42c | 277 | |
2fb1ef4f KRW |
278 | /* Check that there is no request in use */ |
279 | list_for_each_entry_safe(req, n, &ring->pending_free, free_list) { | |
280 | list_del(&req->free_list); | |
f929d42c | 281 | |
2fb1ef4f KRW |
282 | for (j = 0; j < MAX_INDIRECT_SEGMENTS; j++) |
283 | kfree(req->segments[j]); | |
f929d42c | 284 | |
2fb1ef4f KRW |
285 | for (j = 0; j < MAX_INDIRECT_PAGES; j++) |
286 | kfree(req->indirect_pages[j]); | |
f929d42c | 287 | |
2fb1ef4f KRW |
288 | kfree(req); |
289 | i++; | |
290 | } | |
291 | ||
d4bf0065 BL |
292 | BUG_ON(atomic_read(&ring->persistent_gnt_in_use) != 0); |
293 | BUG_ON(!list_empty(&ring->persistent_purge_list)); | |
294 | BUG_ON(!RB_EMPTY_ROOT(&ring->persistent_gnts)); | |
295 | BUG_ON(!list_empty(&ring->free_pages)); | |
296 | BUG_ON(ring->free_pages_num != 0); | |
297 | BUG_ON(ring->persistent_gnt_c != 0); | |
2fb1ef4f | 298 | WARN_ON(i != (XEN_BLKIF_REQS_PER_PAGE * blkif->nr_ring_pages)); |
93bb277f | 299 | xen_blkif_put(blkif); |
2fb1ef4f | 300 | } |
f929d42c | 301 | blkif->nr_ring_pages = 0; |
93bb277f BL |
302 | /* |
303 | * blkif->rings was allocated in connect_ring, so we should free it in | |
304 | * here. | |
305 | */ | |
306 | kfree(blkif->rings); | |
307 | blkif->rings = NULL; | |
308 | blkif->nr_rings = 0; | |
f929d42c | 309 | |
814d04e7 | 310 | return 0; |
ee9ff853 KRW |
311 | } |
312 | ||
2911758f | 313 | static void xen_blkif_free(struct xen_blkif *blkif) |
ee9ff853 | 314 | { |
bf0720c4 | 315 | |
814d04e7 VP |
316 | xen_blkif_disconnect(blkif); |
317 | xen_vbd_free(&blkif->vbd); | |
bf0720c4 | 318 | |
ef753411 | 319 | /* Make sure everything is drained before shutting down */ |
8b6bf747 | 320 | kmem_cache_free(xen_blkif_cachep, blkif); |
ee9ff853 KRW |
321 | } |
322 | ||
8b6bf747 | 323 | int __init xen_blkif_interface_init(void) |
ee9ff853 | 324 | { |
8b6bf747 | 325 | xen_blkif_cachep = kmem_cache_create("blkif_cache", |
30fd1502 | 326 | sizeof(struct xen_blkif), |
8b6bf747 KRW |
327 | 0, 0, NULL); |
328 | if (!xen_blkif_cachep) | |
ee9ff853 KRW |
329 | return -ENOMEM; |
330 | ||
331 | return 0; | |
332 | } | |
4d05a28d | 333 | |
a1397fa3 | 334 | /* |
4d05a28d KRW |
335 | * sysfs interface for VBD I/O requests |
336 | */ | |
337 | ||
db6fbc10 | 338 | #define VBD_SHOW_ALLRING(name, format) \ |
4d05a28d KRW |
339 | static ssize_t show_##name(struct device *_dev, \ |
340 | struct device_attribute *attr, \ | |
341 | char *buf) \ | |
342 | { \ | |
343 | struct xenbus_device *dev = to_xenbus_device(_dev); \ | |
5cf6e4f6 | 344 | struct backend_info *be = dev_get_drvdata(&dev->dev); \ |
db6fbc10 BL |
345 | struct xen_blkif *blkif = be->blkif; \ |
346 | unsigned int i; \ | |
347 | unsigned long long result = 0; \ | |
4d05a28d | 348 | \ |
db6fbc10 BL |
349 | if (!blkif->rings) \ |
350 | goto out; \ | |
351 | \ | |
352 | for (i = 0; i < blkif->nr_rings; i++) { \ | |
353 | struct xen_blkif_ring *ring = &blkif->rings[i]; \ | |
354 | \ | |
355 | result += ring->st_##name; \ | |
356 | } \ | |
357 | \ | |
358 | out: \ | |
359 | return sprintf(buf, format, result); \ | |
4d05a28d KRW |
360 | } \ |
361 | static DEVICE_ATTR(name, S_IRUGO, show_##name, NULL) | |
362 | ||
db6fbc10 BL |
363 | VBD_SHOW_ALLRING(oo_req, "%llu\n"); |
364 | VBD_SHOW_ALLRING(rd_req, "%llu\n"); | |
365 | VBD_SHOW_ALLRING(wr_req, "%llu\n"); | |
366 | VBD_SHOW_ALLRING(f_req, "%llu\n"); | |
367 | VBD_SHOW_ALLRING(ds_req, "%llu\n"); | |
368 | VBD_SHOW_ALLRING(rd_sect, "%llu\n"); | |
369 | VBD_SHOW_ALLRING(wr_sect, "%llu\n"); | |
4d05a28d | 370 | |
3d814731 | 371 | static struct attribute *xen_vbdstat_attrs[] = { |
4d05a28d KRW |
372 | &dev_attr_oo_req.attr, |
373 | &dev_attr_rd_req.attr, | |
374 | &dev_attr_wr_req.attr, | |
24f567f9 | 375 | &dev_attr_f_req.attr, |
b3cb0d6a | 376 | &dev_attr_ds_req.attr, |
4d05a28d KRW |
377 | &dev_attr_rd_sect.attr, |
378 | &dev_attr_wr_sect.attr, | |
379 | NULL | |
380 | }; | |
381 | ||
3d814731 | 382 | static struct attribute_group xen_vbdstat_group = { |
4d05a28d | 383 | .name = "statistics", |
3d814731 | 384 | .attrs = xen_vbdstat_attrs, |
4d05a28d KRW |
385 | }; |
386 | ||
db6fbc10 BL |
387 | #define VBD_SHOW(name, format, args...) \ |
388 | static ssize_t show_##name(struct device *_dev, \ | |
389 | struct device_attribute *attr, \ | |
390 | char *buf) \ | |
391 | { \ | |
392 | struct xenbus_device *dev = to_xenbus_device(_dev); \ | |
393 | struct backend_info *be = dev_get_drvdata(&dev->dev); \ | |
394 | \ | |
395 | return sprintf(buf, format, ##args); \ | |
396 | } \ | |
397 | static DEVICE_ATTR(name, S_IRUGO, show_##name, NULL) | |
398 | ||
4d05a28d KRW |
399 | VBD_SHOW(physical_device, "%x:%x\n", be->major, be->minor); |
400 | VBD_SHOW(mode, "%s\n", be->mode); | |
401 | ||
2911758f | 402 | static int xenvbd_sysfs_addif(struct xenbus_device *dev) |
4d05a28d KRW |
403 | { |
404 | int error; | |
405 | ||
406 | error = device_create_file(&dev->dev, &dev_attr_physical_device); | |
d6091b21 | 407 | if (error) |
4d05a28d KRW |
408 | goto fail1; |
409 | ||
410 | error = device_create_file(&dev->dev, &dev_attr_mode); | |
411 | if (error) | |
412 | goto fail2; | |
413 | ||
3d814731 | 414 | error = sysfs_create_group(&dev->dev.kobj, &xen_vbdstat_group); |
4d05a28d KRW |
415 | if (error) |
416 | goto fail3; | |
417 | ||
418 | return 0; | |
419 | ||
3d814731 | 420 | fail3: sysfs_remove_group(&dev->dev.kobj, &xen_vbdstat_group); |
4d05a28d KRW |
421 | fail2: device_remove_file(&dev->dev, &dev_attr_mode); |
422 | fail1: device_remove_file(&dev->dev, &dev_attr_physical_device); | |
423 | return error; | |
424 | } | |
425 | ||
2911758f | 426 | static void xenvbd_sysfs_delif(struct xenbus_device *dev) |
4d05a28d | 427 | { |
3d814731 | 428 | sysfs_remove_group(&dev->dev.kobj, &xen_vbdstat_group); |
4d05a28d KRW |
429 | device_remove_file(&dev->dev, &dev_attr_mode); |
430 | device_remove_file(&dev->dev, &dev_attr_physical_device); | |
431 | } | |
432 | ||
42c7841d | 433 | |
3d814731 | 434 | static void xen_vbd_free(struct xen_vbd *vbd) |
42c7841d KRW |
435 | { |
436 | if (vbd->bdev) | |
437 | blkdev_put(vbd->bdev, vbd->readonly ? FMODE_READ : FMODE_WRITE); | |
438 | vbd->bdev = NULL; | |
439 | } | |
440 | ||
3d814731 KRW |
441 | static int xen_vbd_create(struct xen_blkif *blkif, blkif_vdev_t handle, |
442 | unsigned major, unsigned minor, int readonly, | |
443 | int cdrom) | |
42c7841d | 444 | { |
3d814731 | 445 | struct xen_vbd *vbd; |
42c7841d | 446 | struct block_device *bdev; |
24f567f9 | 447 | struct request_queue *q; |
42c7841d KRW |
448 | |
449 | vbd = &blkif->vbd; | |
450 | vbd->handle = handle; | |
451 | vbd->readonly = readonly; | |
452 | vbd->type = 0; | |
453 | ||
454 | vbd->pdevice = MKDEV(major, minor); | |
455 | ||
456 | bdev = blkdev_get_by_dev(vbd->pdevice, vbd->readonly ? | |
457 | FMODE_READ : FMODE_WRITE, NULL); | |
458 | ||
459 | if (IS_ERR(bdev)) { | |
77387b82 | 460 | pr_warn("xen_vbd_create: device %08x could not be opened\n", |
42c7841d KRW |
461 | vbd->pdevice); |
462 | return -ENOENT; | |
463 | } | |
464 | ||
465 | vbd->bdev = bdev; | |
42c7841d | 466 | if (vbd->bdev->bd_disk == NULL) { |
77387b82 | 467 | pr_warn("xen_vbd_create: device %08x doesn't exist\n", |
42c7841d | 468 | vbd->pdevice); |
3d814731 | 469 | xen_vbd_free(vbd); |
42c7841d KRW |
470 | return -ENOENT; |
471 | } | |
6464920a | 472 | vbd->size = vbd_sz(vbd); |
42c7841d KRW |
473 | |
474 | if (vbd->bdev->bd_disk->flags & GENHD_FL_CD || cdrom) | |
475 | vbd->type |= VDISK_CDROM; | |
476 | if (vbd->bdev->bd_disk->flags & GENHD_FL_REMOVABLE) | |
477 | vbd->type |= VDISK_REMOVABLE; | |
478 | ||
24f567f9 KRW |
479 | q = bdev_get_queue(bdev); |
480 | if (q && q->flush_flags) | |
481 | vbd->flush_support = true; | |
482 | ||
5ea42986 KRW |
483 | if (q && blk_queue_secdiscard(q)) |
484 | vbd->discard_secure = true; | |
485 | ||
77387b82 | 486 | pr_debug("Successful creation of handle=%04x (dom=%u)\n", |
42c7841d KRW |
487 | handle, blkif->domid); |
488 | return 0; | |
489 | } | |
8b6bf747 | 490 | static int xen_blkbk_remove(struct xenbus_device *dev) |
4d05a28d | 491 | { |
5cf6e4f6 | 492 | struct backend_info *be = dev_get_drvdata(&dev->dev); |
4d05a28d | 493 | |
77387b82 | 494 | pr_debug("%s %p %d\n", __func__, dev, dev->otherend_id); |
4d05a28d KRW |
495 | |
496 | if (be->major || be->minor) | |
497 | xenvbd_sysfs_delif(dev); | |
498 | ||
499 | if (be->backend_watch.node) { | |
500 | unregister_xenbus_watch(&be->backend_watch); | |
501 | kfree(be->backend_watch.node); | |
502 | be->backend_watch.node = NULL; | |
503 | } | |
504 | ||
814d04e7 VP |
505 | dev_set_drvdata(&dev->dev, NULL); |
506 | ||
93bb277f | 507 | if (be->blkif) |
8b6bf747 | 508 | xen_blkif_disconnect(be->blkif); |
4d05a28d | 509 | |
93bb277f BL |
510 | /* Put the reference we set in xen_blkif_alloc(). */ |
511 | xen_blkif_put(be->blkif); | |
9d092603 | 512 | kfree(be->mode); |
4d05a28d | 513 | kfree(be); |
4d05a28d KRW |
514 | return 0; |
515 | } | |
516 | ||
24f567f9 KRW |
517 | int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt, |
518 | struct backend_info *be, int state) | |
4d05a28d KRW |
519 | { |
520 | struct xenbus_device *dev = be->dev; | |
521 | int err; | |
522 | ||
24f567f9 | 523 | err = xenbus_printf(xbt, dev->nodename, "feature-flush-cache", |
4d05a28d KRW |
524 | "%d", state); |
525 | if (err) | |
3389bb8b | 526 | dev_warn(&dev->dev, "writing feature-flush-cache (%d)", err); |
4d05a28d KRW |
527 | |
528 | return err; | |
529 | } | |
530 | ||
3389bb8b | 531 | static void xen_blkbk_discard(struct xenbus_transaction xbt, struct backend_info *be) |
b3cb0d6a LD |
532 | { |
533 | struct xenbus_device *dev = be->dev; | |
534 | struct xen_blkif *blkif = be->blkif; | |
b3cb0d6a | 535 | int err; |
c926b701 | 536 | int state = 0, discard_enable; |
4dae7670 KRW |
537 | struct block_device *bdev = be->blkif->vbd.bdev; |
538 | struct request_queue *q = bdev_get_queue(bdev); | |
539 | ||
c926b701 OH |
540 | err = xenbus_scanf(XBT_NIL, dev->nodename, "discard-enable", "%d", |
541 | &discard_enable); | |
542 | if (err == 1 && !discard_enable) | |
543 | return; | |
544 | ||
4dae7670 KRW |
545 | if (blk_queue_discard(q)) { |
546 | err = xenbus_printf(xbt, dev->nodename, | |
547 | "discard-granularity", "%u", | |
548 | q->limits.discard_granularity); | |
549 | if (err) { | |
3389bb8b KRW |
550 | dev_warn(&dev->dev, "writing discard-granularity (%d)", err); |
551 | return; | |
4dae7670 KRW |
552 | } |
553 | err = xenbus_printf(xbt, dev->nodename, | |
554 | "discard-alignment", "%u", | |
555 | q->limits.discard_alignment); | |
556 | if (err) { | |
3389bb8b KRW |
557 | dev_warn(&dev->dev, "writing discard-alignment (%d)", err); |
558 | return; | |
b3cb0d6a | 559 | } |
4dae7670 KRW |
560 | state = 1; |
561 | /* Optional. */ | |
562 | err = xenbus_printf(xbt, dev->nodename, | |
563 | "discard-secure", "%d", | |
564 | blkif->vbd.discard_secure); | |
565 | if (err) { | |
a71e23d9 | 566 | dev_warn(&dev->dev, "writing discard-secure (%d)", err); |
3389bb8b | 567 | return; |
b3cb0d6a | 568 | } |
b3cb0d6a | 569 | } |
b3cb0d6a LD |
570 | err = xenbus_printf(xbt, dev->nodename, "feature-discard", |
571 | "%d", state); | |
572 | if (err) | |
3389bb8b | 573 | dev_warn(&dev->dev, "writing feature-discard (%d)", err); |
b3cb0d6a | 574 | } |
29bde093 KRW |
575 | int xen_blkbk_barrier(struct xenbus_transaction xbt, |
576 | struct backend_info *be, int state) | |
577 | { | |
578 | struct xenbus_device *dev = be->dev; | |
579 | int err; | |
580 | ||
581 | err = xenbus_printf(xbt, dev->nodename, "feature-barrier", | |
582 | "%d", state); | |
583 | if (err) | |
3389bb8b | 584 | dev_warn(&dev->dev, "writing feature-barrier (%d)", err); |
29bde093 KRW |
585 | |
586 | return err; | |
587 | } | |
b3cb0d6a | 588 | |
01f37f2d | 589 | /* |
4d05a28d KRW |
590 | * Entry point to this code when a new device is created. Allocate the basic |
591 | * structures, and watch the store waiting for the hotplug scripts to tell us | |
592 | * the device's physical major and minor numbers. Switch to InitWait. | |
593 | */ | |
8b6bf747 KRW |
594 | static int xen_blkbk_probe(struct xenbus_device *dev, |
595 | const struct xenbus_device_id *id) | |
4d05a28d KRW |
596 | { |
597 | int err; | |
598 | struct backend_info *be = kzalloc(sizeof(struct backend_info), | |
599 | GFP_KERNEL); | |
77387b82 TC |
600 | |
601 | /* match the pr_debug in xen_blkbk_remove */ | |
602 | pr_debug("%s %p %d\n", __func__, dev, dev->otherend_id); | |
603 | ||
4d05a28d KRW |
604 | if (!be) { |
605 | xenbus_dev_fatal(dev, -ENOMEM, | |
606 | "allocating backend structure"); | |
607 | return -ENOMEM; | |
608 | } | |
609 | be->dev = dev; | |
5cf6e4f6 | 610 | dev_set_drvdata(&dev->dev, be); |
4d05a28d | 611 | |
8b6bf747 | 612 | be->blkif = xen_blkif_alloc(dev->otherend_id); |
4d05a28d KRW |
613 | if (IS_ERR(be->blkif)) { |
614 | err = PTR_ERR(be->blkif); | |
615 | be->blkif = NULL; | |
616 | xenbus_dev_fatal(dev, err, "creating block interface"); | |
617 | goto fail; | |
618 | } | |
619 | ||
5a705845 JB |
620 | err = xenbus_printf(XBT_NIL, dev->nodename, |
621 | "feature-max-indirect-segments", "%u", | |
622 | MAX_INDIRECT_SEGMENTS); | |
623 | if (err) | |
624 | dev_warn(&dev->dev, | |
625 | "writing %s/feature-max-indirect-segments (%d)", | |
626 | dev->nodename, err); | |
627 | ||
d62d8600 BL |
628 | /* Multi-queue: advertise how many queues are supported by us.*/ |
629 | err = xenbus_printf(XBT_NIL, dev->nodename, | |
630 | "multi-queue-max-queues", "%u", xenblk_max_queues); | |
631 | if (err) | |
632 | pr_warn("Error writing multi-queue-max-queues\n"); | |
633 | ||
4d05a28d KRW |
634 | /* setup back pointer */ |
635 | be->blkif->be = be; | |
636 | ||
88122933 JF |
637 | err = xenbus_watch_pathfmt(dev, &be->backend_watch, backend_changed, |
638 | "%s/%s", dev->nodename, "physical-device"); | |
4d05a28d KRW |
639 | if (err) |
640 | goto fail; | |
641 | ||
86839c56 BL |
642 | err = xenbus_printf(XBT_NIL, dev->nodename, "max-ring-page-order", "%u", |
643 | xen_blkif_max_ring_order); | |
644 | if (err) | |
645 | pr_warn("%s write out 'max-ring-page-order' failed\n", __func__); | |
646 | ||
4d05a28d KRW |
647 | err = xenbus_switch_state(dev, XenbusStateInitWait); |
648 | if (err) | |
649 | goto fail; | |
650 | ||
651 | return 0; | |
652 | ||
653 | fail: | |
77387b82 | 654 | pr_warn("%s failed\n", __func__); |
8b6bf747 | 655 | xen_blkbk_remove(dev); |
4d05a28d KRW |
656 | return err; |
657 | } | |
658 | ||
659 | ||
01f37f2d | 660 | /* |
4d05a28d KRW |
661 | * Callback received when the hotplug scripts have placed the physical-device |
662 | * node. Read it and the mode node, and create a vbd. If the frontend is | |
663 | * ready, connect. | |
664 | */ | |
665 | static void backend_changed(struct xenbus_watch *watch, | |
666 | const char **vec, unsigned int len) | |
667 | { | |
668 | int err; | |
669 | unsigned major; | |
670 | unsigned minor; | |
671 | struct backend_info *be | |
672 | = container_of(watch, struct backend_info, backend_watch); | |
673 | struct xenbus_device *dev = be->dev; | |
674 | int cdrom = 0; | |
9d092603 | 675 | unsigned long handle; |
4d05a28d KRW |
676 | char *device_type; |
677 | ||
77387b82 | 678 | pr_debug("%s %p %d\n", __func__, dev, dev->otherend_id); |
4d05a28d KRW |
679 | |
680 | err = xenbus_scanf(XBT_NIL, dev->nodename, "physical-device", "%x:%x", | |
681 | &major, &minor); | |
682 | if (XENBUS_EXIST_ERR(err)) { | |
01f37f2d KRW |
683 | /* |
684 | * Since this watch will fire once immediately after it is | |
685 | * registered, we expect this. Ignore it, and wait for the | |
686 | * hotplug scripts. | |
687 | */ | |
4d05a28d KRW |
688 | return; |
689 | } | |
690 | if (err != 2) { | |
691 | xenbus_dev_fatal(dev, err, "reading physical-device"); | |
692 | return; | |
693 | } | |
694 | ||
9d092603 JB |
695 | if (be->major | be->minor) { |
696 | if (be->major != major || be->minor != minor) | |
77387b82 | 697 | pr_warn("changing physical device (from %x:%x to %x:%x) not supported.\n", |
9d092603 | 698 | be->major, be->minor, major, minor); |
4d05a28d KRW |
699 | return; |
700 | } | |
701 | ||
702 | be->mode = xenbus_read(XBT_NIL, dev->nodename, "mode", NULL); | |
703 | if (IS_ERR(be->mode)) { | |
704 | err = PTR_ERR(be->mode); | |
705 | be->mode = NULL; | |
706 | xenbus_dev_fatal(dev, err, "reading mode"); | |
707 | return; | |
708 | } | |
709 | ||
710 | device_type = xenbus_read(XBT_NIL, dev->otherend, "device-type", NULL); | |
711 | if (!IS_ERR(device_type)) { | |
712 | cdrom = strcmp(device_type, "cdrom") == 0; | |
713 | kfree(device_type); | |
714 | } | |
715 | ||
9d092603 | 716 | /* Front end dir is a number, which is used as the handle. */ |
bb8e0e84 | 717 | err = kstrtoul(strrchr(dev->otherend, '/') + 1, 0, &handle); |
9d092603 JB |
718 | if (err) |
719 | return; | |
4d05a28d | 720 | |
9d092603 JB |
721 | be->major = major; |
722 | be->minor = minor; | |
4d05a28d | 723 | |
9d092603 JB |
724 | err = xen_vbd_create(be->blkif, handle, major, minor, |
725 | !strchr(be->mode, 'w'), cdrom); | |
4d05a28d | 726 | |
9d092603 JB |
727 | if (err) |
728 | xenbus_dev_fatal(dev, err, "creating vbd structure"); | |
729 | else { | |
4d05a28d KRW |
730 | err = xenvbd_sysfs_addif(dev); |
731 | if (err) { | |
3d814731 | 732 | xen_vbd_free(&be->blkif->vbd); |
4d05a28d | 733 | xenbus_dev_fatal(dev, err, "creating sysfs entries"); |
4d05a28d | 734 | } |
9d092603 | 735 | } |
4d05a28d | 736 | |
9d092603 JB |
737 | if (err) { |
738 | kfree(be->mode); | |
739 | be->mode = NULL; | |
740 | be->major = 0; | |
741 | be->minor = 0; | |
742 | } else { | |
4d05a28d | 743 | /* We're potentially connected now */ |
8b6bf747 | 744 | xen_update_blkif_status(be->blkif); |
4d05a28d KRW |
745 | } |
746 | } | |
747 | ||
748 | ||
01f37f2d | 749 | /* |
4d05a28d KRW |
750 | * Callback received when the frontend's state changes. |
751 | */ | |
752 | static void frontend_changed(struct xenbus_device *dev, | |
753 | enum xenbus_state frontend_state) | |
754 | { | |
5cf6e4f6 | 755 | struct backend_info *be = dev_get_drvdata(&dev->dev); |
4d05a28d KRW |
756 | int err; |
757 | ||
77387b82 | 758 | pr_debug("%s %p %s\n", __func__, dev, xenbus_strstate(frontend_state)); |
4d05a28d KRW |
759 | |
760 | switch (frontend_state) { | |
761 | case XenbusStateInitialising: | |
762 | if (dev->state == XenbusStateClosed) { | |
77387b82 | 763 | pr_info("%s: prepare for reconnect\n", dev->nodename); |
4d05a28d KRW |
764 | xenbus_switch_state(dev, XenbusStateInitWait); |
765 | } | |
766 | break; | |
767 | ||
768 | case XenbusStateInitialised: | |
769 | case XenbusStateConnected: | |
01f37f2d KRW |
770 | /* |
771 | * Ensure we connect even when two watches fire in | |
42b2aa86 | 772 | * close succession and we miss the intermediate value |
01f37f2d KRW |
773 | * of frontend_state. |
774 | */ | |
4d05a28d KRW |
775 | if (dev->state == XenbusStateConnected) |
776 | break; | |
777 | ||
01f37f2d KRW |
778 | /* |
779 | * Enforce precondition before potential leak point. | |
1bc05b0a | 780 | * xen_blkif_disconnect() is idempotent. |
313d7b00 | 781 | */ |
814d04e7 VP |
782 | err = xen_blkif_disconnect(be->blkif); |
783 | if (err) { | |
784 | xenbus_dev_fatal(dev, err, "pending I/O"); | |
785 | break; | |
786 | } | |
313d7b00 | 787 | |
4d05a28d | 788 | err = connect_ring(be); |
2d0382fa KRW |
789 | if (err) { |
790 | /* | |
791 | * Clean up so that memory resources can be used by | |
792 | * other devices. connect_ring reported already error. | |
793 | */ | |
794 | xen_blkif_disconnect(be->blkif); | |
4d05a28d | 795 | break; |
2d0382fa | 796 | } |
8b6bf747 | 797 | xen_update_blkif_status(be->blkif); |
4d05a28d KRW |
798 | break; |
799 | ||
800 | case XenbusStateClosing: | |
4d05a28d KRW |
801 | xenbus_switch_state(dev, XenbusStateClosing); |
802 | break; | |
803 | ||
804 | case XenbusStateClosed: | |
6f5986bc | 805 | xen_blkif_disconnect(be->blkif); |
4d05a28d KRW |
806 | xenbus_switch_state(dev, XenbusStateClosed); |
807 | if (xenbus_dev_is_online(dev)) | |
808 | break; | |
809 | /* fall through if not online */ | |
810 | case XenbusStateUnknown: | |
1bc05b0a | 811 | /* implies xen_blkif_disconnect() via xen_blkbk_remove() */ |
4d05a28d KRW |
812 | device_unregister(&dev->dev); |
813 | break; | |
814 | ||
815 | default: | |
816 | xenbus_dev_fatal(dev, -EINVAL, "saw state %d at frontend", | |
817 | frontend_state); | |
818 | break; | |
819 | } | |
820 | } | |
821 | ||
822 | ||
823 | /* ** Connection ** */ | |
824 | ||
825 | ||
01f37f2d | 826 | /* |
4d05a28d KRW |
827 | * Write the physical details regarding the block device to the store, and |
828 | * switch to Connected state. | |
829 | */ | |
830 | static void connect(struct backend_info *be) | |
831 | { | |
832 | struct xenbus_transaction xbt; | |
833 | int err; | |
834 | struct xenbus_device *dev = be->dev; | |
835 | ||
77387b82 | 836 | pr_debug("%s %s\n", __func__, dev->otherend); |
4d05a28d KRW |
837 | |
838 | /* Supply the information about the device the frontend needs */ | |
839 | again: | |
840 | err = xenbus_transaction_start(&xbt); | |
841 | if (err) { | |
842 | xenbus_dev_fatal(dev, err, "starting transaction"); | |
843 | return; | |
844 | } | |
845 | ||
3389bb8b KRW |
846 | /* If we can't advertise it is OK. */ |
847 | xen_blkbk_flush_diskcache(xbt, be, be->blkif->vbd.flush_support); | |
4d05a28d | 848 | |
3389bb8b | 849 | xen_blkbk_discard(xbt, be); |
b3cb0d6a | 850 | |
3389bb8b | 851 | xen_blkbk_barrier(xbt, be, be->blkif->vbd.flush_support); |
29bde093 | 852 | |
0a8704a5 RPM |
853 | err = xenbus_printf(xbt, dev->nodename, "feature-persistent", "%u", 1); |
854 | if (err) { | |
855 | xenbus_dev_fatal(dev, err, "writing %s/feature-persistent", | |
856 | dev->nodename); | |
857 | goto abort; | |
858 | } | |
859 | ||
4d05a28d | 860 | err = xenbus_printf(xbt, dev->nodename, "sectors", "%llu", |
42c7841d | 861 | (unsigned long long)vbd_sz(&be->blkif->vbd)); |
4d05a28d KRW |
862 | if (err) { |
863 | xenbus_dev_fatal(dev, err, "writing %s/sectors", | |
864 | dev->nodename); | |
865 | goto abort; | |
866 | } | |
867 | ||
868 | /* FIXME: use a typename instead */ | |
869 | err = xenbus_printf(xbt, dev->nodename, "info", "%u", | |
42c7841d KRW |
870 | be->blkif->vbd.type | |
871 | (be->blkif->vbd.readonly ? VDISK_READONLY : 0)); | |
4d05a28d KRW |
872 | if (err) { |
873 | xenbus_dev_fatal(dev, err, "writing %s/info", | |
874 | dev->nodename); | |
875 | goto abort; | |
876 | } | |
877 | err = xenbus_printf(xbt, dev->nodename, "sector-size", "%lu", | |
42c7841d KRW |
878 | (unsigned long) |
879 | bdev_logical_block_size(be->blkif->vbd.bdev)); | |
4d05a28d KRW |
880 | if (err) { |
881 | xenbus_dev_fatal(dev, err, "writing %s/sector-size", | |
882 | dev->nodename); | |
883 | goto abort; | |
884 | } | |
7c4d7d71 SB |
885 | err = xenbus_printf(xbt, dev->nodename, "physical-sector-size", "%u", |
886 | bdev_physical_block_size(be->blkif->vbd.bdev)); | |
887 | if (err) | |
888 | xenbus_dev_error(dev, err, "writing %s/physical-sector-size", | |
889 | dev->nodename); | |
4d05a28d KRW |
890 | |
891 | err = xenbus_transaction_end(xbt, 0); | |
892 | if (err == -EAGAIN) | |
893 | goto again; | |
894 | if (err) | |
895 | xenbus_dev_fatal(dev, err, "ending transaction"); | |
896 | ||
897 | err = xenbus_switch_state(dev, XenbusStateConnected); | |
898 | if (err) | |
08b8bfc1 | 899 | xenbus_dev_fatal(dev, err, "%s: switching to Connected state", |
4d05a28d KRW |
900 | dev->nodename); |
901 | ||
902 | return; | |
903 | abort: | |
904 | xenbus_transaction_end(xbt, 1); | |
905 | } | |
906 | ||
2fb1ef4f KRW |
907 | /* |
908 | * Each ring may have multi pages, depends on "ring-page-order". | |
909 | */ | |
910 | static int read_per_ring_refs(struct xen_blkif_ring *ring, const char *dir) | |
4d05a28d | 911 | { |
9cce2914 | 912 | unsigned int ring_ref[XENBUS_MAX_RING_GRANTS]; |
69b91ede BL |
913 | struct pending_req *req, *n; |
914 | int err, i, j; | |
2fb1ef4f KRW |
915 | struct xen_blkif *blkif = ring->blkif; |
916 | struct xenbus_device *dev = blkif->be->dev; | |
917 | unsigned int ring_page_order, nr_grefs, evtchn; | |
4d05a28d | 918 | |
2fb1ef4f | 919 | err = xenbus_scanf(XBT_NIL, dir, "event-channel", "%u", |
86839c56 BL |
920 | &evtchn); |
921 | if (err != 1) { | |
922 | err = -EINVAL; | |
2fb1ef4f | 923 | xenbus_dev_fatal(dev, err, "reading %s/event-channel", dir); |
4d05a28d KRW |
924 | return err; |
925 | } | |
86839c56 BL |
926 | |
927 | err = xenbus_scanf(XBT_NIL, dev->otherend, "ring-page-order", "%u", | |
928 | &ring_page_order); | |
929 | if (err != 1) { | |
2fb1ef4f | 930 | err = xenbus_scanf(XBT_NIL, dir, "ring-ref", "%u", &ring_ref[0]); |
86839c56 BL |
931 | if (err != 1) { |
932 | err = -EINVAL; | |
2fb1ef4f | 933 | xenbus_dev_fatal(dev, err, "reading %s/ring-ref", dir); |
86839c56 BL |
934 | return err; |
935 | } | |
936 | nr_grefs = 1; | |
86839c56 BL |
937 | } else { |
938 | unsigned int i; | |
939 | ||
940 | if (ring_page_order > xen_blkif_max_ring_order) { | |
941 | err = -EINVAL; | |
942 | xenbus_dev_fatal(dev, err, "%s/request %d ring page order exceed max:%d", | |
2fb1ef4f | 943 | dir, ring_page_order, |
86839c56 BL |
944 | xen_blkif_max_ring_order); |
945 | return err; | |
946 | } | |
947 | ||
948 | nr_grefs = 1 << ring_page_order; | |
949 | for (i = 0; i < nr_grefs; i++) { | |
950 | char ring_ref_name[RINGREF_NAME_LEN]; | |
951 | ||
952 | snprintf(ring_ref_name, RINGREF_NAME_LEN, "ring-ref%u", i); | |
2fb1ef4f | 953 | err = xenbus_scanf(XBT_NIL, dir, ring_ref_name, |
86839c56 BL |
954 | "%u", &ring_ref[i]); |
955 | if (err != 1) { | |
956 | err = -EINVAL; | |
957 | xenbus_dev_fatal(dev, err, "reading %s/%s", | |
2fb1ef4f | 958 | dir, ring_ref_name); |
86839c56 BL |
959 | return err; |
960 | } | |
86839c56 BL |
961 | } |
962 | } | |
2fb1ef4f | 963 | blkif->nr_ring_pages = nr_grefs; |
4d05a28d | 964 | |
86839c56 | 965 | for (i = 0; i < nr_grefs * XEN_BLKIF_REQS_PER_PAGE; i++) { |
69b91ede BL |
966 | req = kzalloc(sizeof(*req), GFP_KERNEL); |
967 | if (!req) | |
968 | goto fail; | |
59795700 | 969 | list_add_tail(&req->free_list, &ring->pending_free); |
69b91ede BL |
970 | for (j = 0; j < MAX_INDIRECT_SEGMENTS; j++) { |
971 | req->segments[j] = kzalloc(sizeof(*req->segments[0]), GFP_KERNEL); | |
972 | if (!req->segments[j]) | |
973 | goto fail; | |
974 | } | |
975 | for (j = 0; j < MAX_INDIRECT_PAGES; j++) { | |
976 | req->indirect_pages[j] = kzalloc(sizeof(*req->indirect_pages[0]), | |
977 | GFP_KERNEL); | |
978 | if (!req->indirect_pages[j]) | |
979 | goto fail; | |
980 | } | |
981 | } | |
982 | ||
4d05a28d | 983 | /* Map the shared frame, irq etc. */ |
59795700 | 984 | err = xen_blkif_map(ring, ring_ref, nr_grefs, evtchn); |
4d05a28d | 985 | if (err) { |
86839c56 | 986 | xenbus_dev_fatal(dev, err, "mapping ring-ref port %u", evtchn); |
4d05a28d KRW |
987 | return err; |
988 | } | |
989 | ||
990 | return 0; | |
69b91ede BL |
991 | |
992 | fail: | |
59795700 | 993 | list_for_each_entry_safe(req, n, &ring->pending_free, free_list) { |
69b91ede BL |
994 | list_del(&req->free_list); |
995 | for (j = 0; j < MAX_INDIRECT_SEGMENTS; j++) { | |
996 | if (!req->segments[j]) | |
997 | break; | |
998 | kfree(req->segments[j]); | |
999 | } | |
1000 | for (j = 0; j < MAX_INDIRECT_PAGES; j++) { | |
1001 | if (!req->indirect_pages[j]) | |
1002 | break; | |
1003 | kfree(req->indirect_pages[j]); | |
1004 | } | |
1005 | kfree(req); | |
1006 | } | |
1007 | return -ENOMEM; | |
2fb1ef4f KRW |
1008 | |
1009 | } | |
1010 | ||
1011 | static int connect_ring(struct backend_info *be) | |
1012 | { | |
1013 | struct xenbus_device *dev = be->dev; | |
1014 | unsigned int pers_grants; | |
1015 | char protocol[64] = ""; | |
1016 | int err, i; | |
1017 | char *xspath; | |
1018 | size_t xspathsize; | |
1019 | const size_t xenstore_path_ext_size = 11; /* sufficient for "/queue-NNN" */ | |
d62d8600 | 1020 | unsigned int requested_num_queues = 0; |
2fb1ef4f KRW |
1021 | |
1022 | pr_debug("%s %s\n", __func__, dev->otherend); | |
1023 | ||
1024 | be->blkif->blk_protocol = BLKIF_PROTOCOL_DEFAULT; | |
1025 | err = xenbus_gather(XBT_NIL, dev->otherend, "protocol", | |
1026 | "%63s", protocol, NULL); | |
1027 | if (err) | |
1028 | strcpy(protocol, "unspecified, assuming default"); | |
1029 | else if (0 == strcmp(protocol, XEN_IO_PROTO_ABI_NATIVE)) | |
1030 | be->blkif->blk_protocol = BLKIF_PROTOCOL_NATIVE; | |
1031 | else if (0 == strcmp(protocol, XEN_IO_PROTO_ABI_X86_32)) | |
1032 | be->blkif->blk_protocol = BLKIF_PROTOCOL_X86_32; | |
1033 | else if (0 == strcmp(protocol, XEN_IO_PROTO_ABI_X86_64)) | |
1034 | be->blkif->blk_protocol = BLKIF_PROTOCOL_X86_64; | |
1035 | else { | |
1036 | xenbus_dev_fatal(dev, err, "unknown fe protocol %s", protocol); | |
bde21f73 | 1037 | return -ENOSYS; |
2fb1ef4f KRW |
1038 | } |
1039 | err = xenbus_gather(XBT_NIL, dev->otherend, | |
1040 | "feature-persistent", "%u", | |
1041 | &pers_grants, NULL); | |
1042 | if (err) | |
1043 | pers_grants = 0; | |
1044 | ||
1045 | be->blkif->vbd.feature_gnt_persistent = pers_grants; | |
1046 | be->blkif->vbd.overflow_max_grants = 0; | |
1047 | ||
d62d8600 BL |
1048 | /* |
1049 | * Read the number of hardware queues from frontend. | |
1050 | */ | |
1051 | err = xenbus_scanf(XBT_NIL, dev->otherend, "multi-queue-num-queues", | |
1052 | "%u", &requested_num_queues); | |
1053 | if (err < 0) { | |
1054 | requested_num_queues = 1; | |
1055 | } else { | |
1056 | if (requested_num_queues > xenblk_max_queues | |
1057 | || requested_num_queues == 0) { | |
1058 | /* Buggy or malicious guest. */ | |
1059 | xenbus_dev_fatal(dev, err, | |
1060 | "guest requested %u queues, exceeding the maximum of %u.", | |
1061 | requested_num_queues, xenblk_max_queues); | |
1062 | return -ENOSYS; | |
1063 | } | |
1064 | } | |
1065 | be->blkif->nr_rings = requested_num_queues; | |
1066 | if (xen_blkif_alloc_rings(be->blkif)) | |
1067 | return -ENOMEM; | |
1068 | ||
2fb1ef4f KRW |
1069 | pr_info("%s: using %d queues, protocol %d (%s) %s\n", dev->nodename, |
1070 | be->blkif->nr_rings, be->blkif->blk_protocol, protocol, | |
1071 | pers_grants ? "persistent grants" : ""); | |
1072 | ||
1073 | if (be->blkif->nr_rings == 1) | |
1074 | return read_per_ring_refs(&be->blkif->rings[0], dev->otherend); | |
1075 | else { | |
1076 | xspathsize = strlen(dev->otherend) + xenstore_path_ext_size; | |
1077 | xspath = kmalloc(xspathsize, GFP_KERNEL); | |
1078 | if (!xspath) { | |
1079 | xenbus_dev_fatal(dev, -ENOMEM, "reading ring references"); | |
1080 | return -ENOMEM; | |
1081 | } | |
1082 | ||
1083 | for (i = 0; i < be->blkif->nr_rings; i++) { | |
1084 | memset(xspath, 0, xspathsize); | |
1085 | snprintf(xspath, xspathsize, "%s/queue-%u", dev->otherend, i); | |
1086 | err = read_per_ring_refs(&be->blkif->rings[i], xspath); | |
1087 | if (err) { | |
1088 | kfree(xspath); | |
1089 | return err; | |
1090 | } | |
1091 | } | |
1092 | kfree(xspath); | |
1093 | } | |
1094 | return 0; | |
4d05a28d KRW |
1095 | } |
1096 | ||
8b6bf747 | 1097 | static const struct xenbus_device_id xen_blkbk_ids[] = { |
4d05a28d KRW |
1098 | { "vbd" }, |
1099 | { "" } | |
1100 | }; | |
1101 | ||
95afae48 DV |
1102 | static struct xenbus_driver xen_blkbk_driver = { |
1103 | .ids = xen_blkbk_ids, | |
8b6bf747 KRW |
1104 | .probe = xen_blkbk_probe, |
1105 | .remove = xen_blkbk_remove, | |
4d05a28d | 1106 | .otherend_changed = frontend_changed |
95afae48 | 1107 | }; |
4d05a28d | 1108 | |
8b6bf747 | 1109 | int xen_blkif_xenbus_init(void) |
4d05a28d | 1110 | { |
73db144b | 1111 | return xenbus_register_backend(&xen_blkbk_driver); |
4d05a28d | 1112 | } |