]>
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 | ||
17 | #include <stdarg.h> | |
18 | #include <linux/module.h> | |
19 | #include <linux/kthread.h> | |
ee9ff853 KRW |
20 | #include <xen/events.h> |
21 | #include <xen/grant_table.h> | |
4d05a28d KRW |
22 | #include "common.h" |
23 | ||
d6091b21 | 24 | struct backend_info { |
01f37f2d | 25 | struct xenbus_device *dev; |
51854322 | 26 | struct xen_blkif *blkif; |
01f37f2d KRW |
27 | struct xenbus_watch backend_watch; |
28 | unsigned major; | |
29 | unsigned minor; | |
30 | char *mode; | |
4d05a28d KRW |
31 | }; |
32 | ||
8b6bf747 | 33 | static struct kmem_cache *xen_blkif_cachep; |
4d05a28d KRW |
34 | static void connect(struct backend_info *); |
35 | static int connect_ring(struct backend_info *); | |
36 | static void backend_changed(struct xenbus_watch *, const char **, | |
37 | unsigned int); | |
814d04e7 VP |
38 | static void xen_blkif_free(struct xen_blkif *blkif); |
39 | static void xen_vbd_free(struct xen_vbd *vbd); | |
4d05a28d | 40 | |
8b6bf747 | 41 | struct xenbus_device *xen_blkbk_xenbus(struct backend_info *be) |
98e036a3 JF |
42 | { |
43 | return be->dev; | |
44 | } | |
45 | ||
814d04e7 VP |
46 | /* |
47 | * The last request could free the device from softirq context and | |
48 | * xen_blkif_free() can sleep. | |
49 | */ | |
50 | static void xen_blkif_deferred_free(struct work_struct *work) | |
51 | { | |
52 | struct xen_blkif *blkif; | |
53 | ||
54 | blkif = container_of(work, struct xen_blkif, free_work); | |
55 | xen_blkif_free(blkif); | |
56 | } | |
57 | ||
30fd1502 | 58 | static int blkback_name(struct xen_blkif *blkif, char *buf) |
4d05a28d KRW |
59 | { |
60 | char *devpath, *devname; | |
61 | struct xenbus_device *dev = blkif->be->dev; | |
62 | ||
63 | devpath = xenbus_read(XBT_NIL, dev->nodename, "dev", NULL); | |
64 | if (IS_ERR(devpath)) | |
65 | return PTR_ERR(devpath); | |
66 | ||
d6091b21 KRW |
67 | devname = strstr(devpath, "/dev/"); |
68 | if (devname != NULL) | |
4d05a28d KRW |
69 | devname += strlen("/dev/"); |
70 | else | |
71 | devname = devpath; | |
72 | ||
73 | snprintf(buf, TASK_COMM_LEN, "blkback.%d.%s", blkif->domid, devname); | |
74 | kfree(devpath); | |
75 | ||
76 | return 0; | |
77 | } | |
78 | ||
30fd1502 | 79 | static void xen_update_blkif_status(struct xen_blkif *blkif) |
4d05a28d KRW |
80 | { |
81 | int err; | |
82 | char name[TASK_COMM_LEN]; | |
83 | ||
84 | /* Not ready to connect? */ | |
85 | if (!blkif->irq || !blkif->vbd.bdev) | |
86 | return; | |
87 | ||
88 | /* Already connected? */ | |
89 | if (blkif->be->dev->state == XenbusStateConnected) | |
90 | return; | |
91 | ||
92 | /* Attempt to connect: exit if we fail to. */ | |
93 | connect(blkif->be); | |
94 | if (blkif->be->dev->state != XenbusStateConnected) | |
95 | return; | |
96 | ||
97 | err = blkback_name(blkif, name); | |
98 | if (err) { | |
99 | xenbus_dev_error(blkif->be->dev, err, "get blkback dev name"); | |
100 | return; | |
101 | } | |
102 | ||
cbf46290 CL |
103 | err = filemap_write_and_wait(blkif->vbd.bdev->bd_inode->i_mapping); |
104 | if (err) { | |
105 | xenbus_dev_error(blkif->be->dev, err, "block flush"); | |
106 | return; | |
107 | } | |
108 | invalidate_inode_pages2(blkif->vbd.bdev->bd_inode->i_mapping); | |
109 | ||
f170168b | 110 | blkif->xenblkd = kthread_run(xen_blkif_schedule, blkif, "%s", name); |
4d05a28d KRW |
111 | if (IS_ERR(blkif->xenblkd)) { |
112 | err = PTR_ERR(blkif->xenblkd); | |
113 | blkif->xenblkd = NULL; | |
114 | xenbus_dev_error(blkif->be->dev, err, "start xenblkd"); | |
3f3aad5e | 115 | return; |
4d05a28d KRW |
116 | } |
117 | } | |
118 | ||
30fd1502 | 119 | static struct xen_blkif *xen_blkif_alloc(domid_t domid) |
ee9ff853 | 120 | { |
30fd1502 | 121 | struct xen_blkif *blkif; |
bb642e83 RPM |
122 | struct pending_req *req, *n; |
123 | int i, j; | |
ee9ff853 | 124 | |
402b27f9 | 125 | BUILD_BUG_ON(MAX_INDIRECT_PAGES > BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST); |
ee9ff853 | 126 | |
654dbef2 | 127 | blkif = kmem_cache_zalloc(xen_blkif_cachep, GFP_KERNEL); |
ee9ff853 KRW |
128 | if (!blkif) |
129 | return ERR_PTR(-ENOMEM); | |
130 | ||
ee9ff853 KRW |
131 | blkif->domid = domid; |
132 | spin_lock_init(&blkif->blk_ring_lock); | |
133 | atomic_set(&blkif->refcnt, 1); | |
134 | init_waitqueue_head(&blkif->wq); | |
29bde093 KRW |
135 | init_completion(&blkif->drain_complete); |
136 | atomic_set(&blkif->drain, 0); | |
ee9ff853 | 137 | blkif->st_print = jiffies; |
0a8704a5 | 138 | blkif->persistent_gnts.rb_node = NULL; |
c6cc142d RPM |
139 | spin_lock_init(&blkif->free_pages_lock); |
140 | INIT_LIST_HEAD(&blkif->free_pages); | |
ef753411 | 141 | INIT_LIST_HEAD(&blkif->persistent_purge_list); |
c6cc142d | 142 | blkif->free_pages_num = 0; |
3f3aad5e | 143 | atomic_set(&blkif->persistent_gnt_in_use, 0); |
c05f3e3c | 144 | atomic_set(&blkif->inflight, 0); |
abb97b8c | 145 | INIT_WORK(&blkif->persistent_purge_work, xen_blkbk_unmap_purged_grants); |
ee9ff853 | 146 | |
bf0720c4 | 147 | INIT_LIST_HEAD(&blkif->pending_free); |
814d04e7 | 148 | INIT_WORK(&blkif->free_work, xen_blkif_deferred_free); |
bb642e83 RPM |
149 | |
150 | for (i = 0; i < XEN_BLKIF_REQS; i++) { | |
151 | req = kzalloc(sizeof(*req), GFP_KERNEL); | |
152 | if (!req) | |
153 | goto fail; | |
154 | list_add_tail(&req->free_list, | |
155 | &blkif->pending_free); | |
156 | for (j = 0; j < MAX_INDIRECT_SEGMENTS; j++) { | |
157 | req->segments[j] = kzalloc(sizeof(*req->segments[0]), | |
158 | GFP_KERNEL); | |
159 | if (!req->segments[j]) | |
160 | goto fail; | |
161 | } | |
162 | for (j = 0; j < MAX_INDIRECT_PAGES; j++) { | |
163 | req->indirect_pages[j] = kzalloc(sizeof(*req->indirect_pages[0]), | |
164 | GFP_KERNEL); | |
165 | if (!req->indirect_pages[j]) | |
166 | goto fail; | |
167 | } | |
168 | } | |
bf0720c4 RPM |
169 | spin_lock_init(&blkif->pending_free_lock); |
170 | init_waitqueue_head(&blkif->pending_free_wq); | |
8e3f8755 | 171 | init_waitqueue_head(&blkif->shutdown_wq); |
ee9ff853 KRW |
172 | |
173 | return blkif; | |
bb642e83 RPM |
174 | |
175 | fail: | |
176 | list_for_each_entry_safe(req, n, &blkif->pending_free, free_list) { | |
177 | list_del(&req->free_list); | |
178 | for (j = 0; j < MAX_INDIRECT_SEGMENTS; j++) { | |
179 | if (!req->segments[j]) | |
180 | break; | |
181 | kfree(req->segments[j]); | |
182 | } | |
183 | for (j = 0; j < MAX_INDIRECT_PAGES; j++) { | |
184 | if (!req->indirect_pages[j]) | |
185 | break; | |
186 | kfree(req->indirect_pages[j]); | |
187 | } | |
188 | kfree(req); | |
189 | } | |
190 | ||
191 | kmem_cache_free(xen_blkif_cachep, blkif); | |
192 | ||
193 | return ERR_PTR(-ENOMEM); | |
ee9ff853 KRW |
194 | } |
195 | ||
ccc9d90a | 196 | static int xen_blkif_map(struct xen_blkif *blkif, grant_ref_t gref, |
8b6bf747 | 197 | unsigned int evtchn) |
ee9ff853 KRW |
198 | { |
199 | int err; | |
200 | ||
201 | /* Already connected through? */ | |
202 | if (blkif->irq) | |
203 | return 0; | |
204 | ||
ccc9d90a WL |
205 | err = xenbus_map_ring_valloc(blkif->be->dev, &gref, 1, |
206 | &blkif->blk_ring); | |
2d073846 | 207 | if (err < 0) |
ee9ff853 | 208 | return err; |
ee9ff853 KRW |
209 | |
210 | switch (blkif->blk_protocol) { | |
211 | case BLKIF_PROTOCOL_NATIVE: | |
212 | { | |
213 | struct blkif_sring *sring; | |
2d073846 | 214 | sring = (struct blkif_sring *)blkif->blk_ring; |
ee9ff853 KRW |
215 | BACK_RING_INIT(&blkif->blk_rings.native, sring, PAGE_SIZE); |
216 | break; | |
217 | } | |
218 | case BLKIF_PROTOCOL_X86_32: | |
219 | { | |
220 | struct blkif_x86_32_sring *sring_x86_32; | |
2d073846 | 221 | sring_x86_32 = (struct blkif_x86_32_sring *)blkif->blk_ring; |
ee9ff853 KRW |
222 | BACK_RING_INIT(&blkif->blk_rings.x86_32, sring_x86_32, PAGE_SIZE); |
223 | break; | |
224 | } | |
225 | case BLKIF_PROTOCOL_X86_64: | |
226 | { | |
227 | struct blkif_x86_64_sring *sring_x86_64; | |
2d073846 | 228 | sring_x86_64 = (struct blkif_x86_64_sring *)blkif->blk_ring; |
ee9ff853 KRW |
229 | BACK_RING_INIT(&blkif->blk_rings.x86_64, sring_x86_64, PAGE_SIZE); |
230 | break; | |
231 | } | |
232 | default: | |
233 | BUG(); | |
234 | } | |
235 | ||
8b6bf747 KRW |
236 | err = bind_interdomain_evtchn_to_irqhandler(blkif->domid, evtchn, |
237 | xen_blkif_be_int, 0, | |
238 | "blkif-backend", blkif); | |
ee9ff853 | 239 | if (err < 0) { |
2d073846 | 240 | xenbus_unmap_ring_vfree(blkif->be->dev, blkif->blk_ring); |
ee9ff853 KRW |
241 | blkif->blk_rings.common.sring = NULL; |
242 | return err; | |
243 | } | |
244 | blkif->irq = err; | |
245 | ||
246 | return 0; | |
247 | } | |
248 | ||
814d04e7 | 249 | static int xen_blkif_disconnect(struct xen_blkif *blkif) |
ee9ff853 KRW |
250 | { |
251 | if (blkif->xenblkd) { | |
252 | kthread_stop(blkif->xenblkd); | |
8e3f8755 | 253 | wake_up(&blkif->shutdown_wq); |
ee9ff853 KRW |
254 | blkif->xenblkd = NULL; |
255 | } | |
256 | ||
814d04e7 VP |
257 | /* The above kthread_stop() guarantees that at this point we |
258 | * don't have any discard_io or other_io requests. So, checking | |
259 | * for inflight IO is enough. | |
260 | */ | |
261 | if (atomic_read(&blkif->inflight) > 0) | |
262 | return -EBUSY; | |
ee9ff853 KRW |
263 | |
264 | if (blkif->irq) { | |
265 | unbind_from_irqhandler(blkif->irq, blkif); | |
266 | blkif->irq = 0; | |
267 | } | |
268 | ||
269 | if (blkif->blk_rings.common.sring) { | |
2d073846 | 270 | xenbus_unmap_ring_vfree(blkif->be->dev, blkif->blk_ring); |
ee9ff853 KRW |
271 | blkif->blk_rings.common.sring = NULL; |
272 | } | |
814d04e7 | 273 | |
12ea7296 VK |
274 | /* Remove all persistent grants and the cache of ballooned pages. */ |
275 | xen_blkbk_free_caches(blkif); | |
276 | ||
814d04e7 | 277 | return 0; |
ee9ff853 KRW |
278 | } |
279 | ||
2911758f | 280 | static void xen_blkif_free(struct xen_blkif *blkif) |
ee9ff853 | 281 | { |
bb642e83 RPM |
282 | struct pending_req *req, *n; |
283 | int i = 0, j; | |
bf0720c4 | 284 | |
814d04e7 VP |
285 | xen_blkif_disconnect(blkif); |
286 | xen_vbd_free(&blkif->vbd); | |
bf0720c4 | 287 | |
ef753411 RPM |
288 | /* Make sure everything is drained before shutting down */ |
289 | BUG_ON(blkif->persistent_gnt_c != 0); | |
290 | BUG_ON(atomic_read(&blkif->persistent_gnt_in_use) != 0); | |
291 | BUG_ON(blkif->free_pages_num != 0); | |
292 | BUG_ON(!list_empty(&blkif->persistent_purge_list)); | |
293 | BUG_ON(!list_empty(&blkif->free_pages)); | |
294 | BUG_ON(!RB_EMPTY_ROOT(&blkif->persistent_gnts)); | |
295 | ||
bf0720c4 | 296 | /* Check that there is no request in use */ |
bb642e83 RPM |
297 | list_for_each_entry_safe(req, n, &blkif->pending_free, free_list) { |
298 | list_del(&req->free_list); | |
299 | ||
300 | for (j = 0; j < MAX_INDIRECT_SEGMENTS; j++) | |
301 | kfree(req->segments[j]); | |
302 | ||
303 | for (j = 0; j < MAX_INDIRECT_PAGES; j++) | |
304 | kfree(req->indirect_pages[j]); | |
305 | ||
306 | kfree(req); | |
bf0720c4 | 307 | i++; |
bb642e83 RPM |
308 | } |
309 | ||
310 | WARN_ON(i != XEN_BLKIF_REQS); | |
bf0720c4 | 311 | |
8b6bf747 | 312 | kmem_cache_free(xen_blkif_cachep, blkif); |
ee9ff853 KRW |
313 | } |
314 | ||
8b6bf747 | 315 | int __init xen_blkif_interface_init(void) |
ee9ff853 | 316 | { |
8b6bf747 | 317 | xen_blkif_cachep = kmem_cache_create("blkif_cache", |
30fd1502 | 318 | sizeof(struct xen_blkif), |
8b6bf747 KRW |
319 | 0, 0, NULL); |
320 | if (!xen_blkif_cachep) | |
ee9ff853 KRW |
321 | return -ENOMEM; |
322 | ||
323 | return 0; | |
324 | } | |
4d05a28d | 325 | |
a1397fa3 | 326 | /* |
4d05a28d KRW |
327 | * sysfs interface for VBD I/O requests |
328 | */ | |
329 | ||
330 | #define VBD_SHOW(name, format, args...) \ | |
331 | static ssize_t show_##name(struct device *_dev, \ | |
332 | struct device_attribute *attr, \ | |
333 | char *buf) \ | |
334 | { \ | |
335 | struct xenbus_device *dev = to_xenbus_device(_dev); \ | |
5cf6e4f6 | 336 | struct backend_info *be = dev_get_drvdata(&dev->dev); \ |
4d05a28d KRW |
337 | \ |
338 | return sprintf(buf, format, ##args); \ | |
339 | } \ | |
340 | static DEVICE_ATTR(name, S_IRUGO, show_##name, NULL) | |
341 | ||
986cacbd ZK |
342 | VBD_SHOW(oo_req, "%llu\n", be->blkif->st_oo_req); |
343 | VBD_SHOW(rd_req, "%llu\n", be->blkif->st_rd_req); | |
344 | VBD_SHOW(wr_req, "%llu\n", be->blkif->st_wr_req); | |
345 | VBD_SHOW(f_req, "%llu\n", be->blkif->st_f_req); | |
346 | VBD_SHOW(ds_req, "%llu\n", be->blkif->st_ds_req); | |
347 | VBD_SHOW(rd_sect, "%llu\n", be->blkif->st_rd_sect); | |
348 | VBD_SHOW(wr_sect, "%llu\n", be->blkif->st_wr_sect); | |
4d05a28d | 349 | |
3d814731 | 350 | static struct attribute *xen_vbdstat_attrs[] = { |
4d05a28d KRW |
351 | &dev_attr_oo_req.attr, |
352 | &dev_attr_rd_req.attr, | |
353 | &dev_attr_wr_req.attr, | |
24f567f9 | 354 | &dev_attr_f_req.attr, |
b3cb0d6a | 355 | &dev_attr_ds_req.attr, |
4d05a28d KRW |
356 | &dev_attr_rd_sect.attr, |
357 | &dev_attr_wr_sect.attr, | |
358 | NULL | |
359 | }; | |
360 | ||
3d814731 | 361 | static struct attribute_group xen_vbdstat_group = { |
4d05a28d | 362 | .name = "statistics", |
3d814731 | 363 | .attrs = xen_vbdstat_attrs, |
4d05a28d KRW |
364 | }; |
365 | ||
366 | VBD_SHOW(physical_device, "%x:%x\n", be->major, be->minor); | |
367 | VBD_SHOW(mode, "%s\n", be->mode); | |
368 | ||
2911758f | 369 | static int xenvbd_sysfs_addif(struct xenbus_device *dev) |
4d05a28d KRW |
370 | { |
371 | int error; | |
372 | ||
373 | error = device_create_file(&dev->dev, &dev_attr_physical_device); | |
d6091b21 | 374 | if (error) |
4d05a28d KRW |
375 | goto fail1; |
376 | ||
377 | error = device_create_file(&dev->dev, &dev_attr_mode); | |
378 | if (error) | |
379 | goto fail2; | |
380 | ||
3d814731 | 381 | error = sysfs_create_group(&dev->dev.kobj, &xen_vbdstat_group); |
4d05a28d KRW |
382 | if (error) |
383 | goto fail3; | |
384 | ||
385 | return 0; | |
386 | ||
3d814731 | 387 | fail3: sysfs_remove_group(&dev->dev.kobj, &xen_vbdstat_group); |
4d05a28d KRW |
388 | fail2: device_remove_file(&dev->dev, &dev_attr_mode); |
389 | fail1: device_remove_file(&dev->dev, &dev_attr_physical_device); | |
390 | return error; | |
391 | } | |
392 | ||
2911758f | 393 | static void xenvbd_sysfs_delif(struct xenbus_device *dev) |
4d05a28d | 394 | { |
3d814731 | 395 | sysfs_remove_group(&dev->dev.kobj, &xen_vbdstat_group); |
4d05a28d KRW |
396 | device_remove_file(&dev->dev, &dev_attr_mode); |
397 | device_remove_file(&dev->dev, &dev_attr_physical_device); | |
398 | } | |
399 | ||
42c7841d | 400 | |
3d814731 | 401 | static void xen_vbd_free(struct xen_vbd *vbd) |
42c7841d KRW |
402 | { |
403 | if (vbd->bdev) | |
404 | blkdev_put(vbd->bdev, vbd->readonly ? FMODE_READ : FMODE_WRITE); | |
405 | vbd->bdev = NULL; | |
406 | } | |
407 | ||
3d814731 KRW |
408 | static int xen_vbd_create(struct xen_blkif *blkif, blkif_vdev_t handle, |
409 | unsigned major, unsigned minor, int readonly, | |
410 | int cdrom) | |
42c7841d | 411 | { |
3d814731 | 412 | struct xen_vbd *vbd; |
42c7841d | 413 | struct block_device *bdev; |
24f567f9 | 414 | struct request_queue *q; |
42c7841d KRW |
415 | |
416 | vbd = &blkif->vbd; | |
417 | vbd->handle = handle; | |
418 | vbd->readonly = readonly; | |
419 | vbd->type = 0; | |
420 | ||
421 | vbd->pdevice = MKDEV(major, minor); | |
422 | ||
423 | bdev = blkdev_get_by_dev(vbd->pdevice, vbd->readonly ? | |
424 | FMODE_READ : FMODE_WRITE, NULL); | |
425 | ||
426 | if (IS_ERR(bdev)) { | |
3d814731 | 427 | DPRINTK("xen_vbd_create: device %08x could not be opened.\n", |
42c7841d KRW |
428 | vbd->pdevice); |
429 | return -ENOENT; | |
430 | } | |
431 | ||
432 | vbd->bdev = bdev; | |
42c7841d | 433 | if (vbd->bdev->bd_disk == NULL) { |
3d814731 | 434 | DPRINTK("xen_vbd_create: device %08x doesn't exist.\n", |
42c7841d | 435 | vbd->pdevice); |
3d814731 | 436 | xen_vbd_free(vbd); |
42c7841d KRW |
437 | return -ENOENT; |
438 | } | |
6464920a | 439 | vbd->size = vbd_sz(vbd); |
42c7841d KRW |
440 | |
441 | if (vbd->bdev->bd_disk->flags & GENHD_FL_CD || cdrom) | |
442 | vbd->type |= VDISK_CDROM; | |
443 | if (vbd->bdev->bd_disk->flags & GENHD_FL_REMOVABLE) | |
444 | vbd->type |= VDISK_REMOVABLE; | |
445 | ||
24f567f9 KRW |
446 | q = bdev_get_queue(bdev); |
447 | if (q && q->flush_flags) | |
448 | vbd->flush_support = true; | |
449 | ||
5ea42986 KRW |
450 | if (q && blk_queue_secdiscard(q)) |
451 | vbd->discard_secure = true; | |
452 | ||
42c7841d KRW |
453 | DPRINTK("Successful creation of handle=%04x (dom=%u)\n", |
454 | handle, blkif->domid); | |
455 | return 0; | |
456 | } | |
8b6bf747 | 457 | static int xen_blkbk_remove(struct xenbus_device *dev) |
4d05a28d | 458 | { |
5cf6e4f6 | 459 | struct backend_info *be = dev_get_drvdata(&dev->dev); |
4d05a28d KRW |
460 | |
461 | DPRINTK(""); | |
462 | ||
463 | if (be->major || be->minor) | |
464 | xenvbd_sysfs_delif(dev); | |
465 | ||
466 | if (be->backend_watch.node) { | |
467 | unregister_xenbus_watch(&be->backend_watch); | |
468 | kfree(be->backend_watch.node); | |
469 | be->backend_watch.node = NULL; | |
470 | } | |
471 | ||
814d04e7 VP |
472 | dev_set_drvdata(&dev->dev, NULL); |
473 | ||
4d05a28d | 474 | if (be->blkif) { |
8b6bf747 | 475 | xen_blkif_disconnect(be->blkif); |
814d04e7 | 476 | xen_blkif_put(be->blkif); |
4d05a28d KRW |
477 | } |
478 | ||
9d092603 | 479 | kfree(be->mode); |
4d05a28d | 480 | kfree(be); |
4d05a28d KRW |
481 | return 0; |
482 | } | |
483 | ||
24f567f9 KRW |
484 | int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt, |
485 | struct backend_info *be, int state) | |
4d05a28d KRW |
486 | { |
487 | struct xenbus_device *dev = be->dev; | |
488 | int err; | |
489 | ||
24f567f9 | 490 | err = xenbus_printf(xbt, dev->nodename, "feature-flush-cache", |
4d05a28d KRW |
491 | "%d", state); |
492 | if (err) | |
3389bb8b | 493 | dev_warn(&dev->dev, "writing feature-flush-cache (%d)", err); |
4d05a28d KRW |
494 | |
495 | return err; | |
496 | } | |
497 | ||
3389bb8b | 498 | static void xen_blkbk_discard(struct xenbus_transaction xbt, struct backend_info *be) |
b3cb0d6a LD |
499 | { |
500 | struct xenbus_device *dev = be->dev; | |
501 | struct xen_blkif *blkif = be->blkif; | |
b3cb0d6a | 502 | int err; |
c926b701 | 503 | int state = 0, discard_enable; |
4dae7670 KRW |
504 | struct block_device *bdev = be->blkif->vbd.bdev; |
505 | struct request_queue *q = bdev_get_queue(bdev); | |
506 | ||
c926b701 OH |
507 | err = xenbus_scanf(XBT_NIL, dev->nodename, "discard-enable", "%d", |
508 | &discard_enable); | |
509 | if (err == 1 && !discard_enable) | |
510 | return; | |
511 | ||
4dae7670 KRW |
512 | if (blk_queue_discard(q)) { |
513 | err = xenbus_printf(xbt, dev->nodename, | |
514 | "discard-granularity", "%u", | |
515 | q->limits.discard_granularity); | |
516 | if (err) { | |
3389bb8b KRW |
517 | dev_warn(&dev->dev, "writing discard-granularity (%d)", err); |
518 | return; | |
4dae7670 KRW |
519 | } |
520 | err = xenbus_printf(xbt, dev->nodename, | |
521 | "discard-alignment", "%u", | |
522 | q->limits.discard_alignment); | |
523 | if (err) { | |
3389bb8b KRW |
524 | dev_warn(&dev->dev, "writing discard-alignment (%d)", err); |
525 | return; | |
b3cb0d6a | 526 | } |
4dae7670 KRW |
527 | state = 1; |
528 | /* Optional. */ | |
529 | err = xenbus_printf(xbt, dev->nodename, | |
530 | "discard-secure", "%d", | |
531 | blkif->vbd.discard_secure); | |
532 | if (err) { | |
a71e23d9 | 533 | dev_warn(&dev->dev, "writing discard-secure (%d)", err); |
3389bb8b | 534 | return; |
b3cb0d6a | 535 | } |
b3cb0d6a | 536 | } |
b3cb0d6a LD |
537 | err = xenbus_printf(xbt, dev->nodename, "feature-discard", |
538 | "%d", state); | |
539 | if (err) | |
3389bb8b | 540 | dev_warn(&dev->dev, "writing feature-discard (%d)", err); |
b3cb0d6a | 541 | } |
29bde093 KRW |
542 | int xen_blkbk_barrier(struct xenbus_transaction xbt, |
543 | struct backend_info *be, int state) | |
544 | { | |
545 | struct xenbus_device *dev = be->dev; | |
546 | int err; | |
547 | ||
548 | err = xenbus_printf(xbt, dev->nodename, "feature-barrier", | |
549 | "%d", state); | |
550 | if (err) | |
3389bb8b | 551 | dev_warn(&dev->dev, "writing feature-barrier (%d)", err); |
29bde093 KRW |
552 | |
553 | return err; | |
554 | } | |
b3cb0d6a | 555 | |
01f37f2d | 556 | /* |
4d05a28d KRW |
557 | * Entry point to this code when a new device is created. Allocate the basic |
558 | * structures, and watch the store waiting for the hotplug scripts to tell us | |
559 | * the device's physical major and minor numbers. Switch to InitWait. | |
560 | */ | |
8b6bf747 KRW |
561 | static int xen_blkbk_probe(struct xenbus_device *dev, |
562 | const struct xenbus_device_id *id) | |
4d05a28d KRW |
563 | { |
564 | int err; | |
565 | struct backend_info *be = kzalloc(sizeof(struct backend_info), | |
566 | GFP_KERNEL); | |
567 | if (!be) { | |
568 | xenbus_dev_fatal(dev, -ENOMEM, | |
569 | "allocating backend structure"); | |
570 | return -ENOMEM; | |
571 | } | |
572 | be->dev = dev; | |
5cf6e4f6 | 573 | dev_set_drvdata(&dev->dev, be); |
4d05a28d | 574 | |
8b6bf747 | 575 | be->blkif = xen_blkif_alloc(dev->otherend_id); |
4d05a28d KRW |
576 | if (IS_ERR(be->blkif)) { |
577 | err = PTR_ERR(be->blkif); | |
578 | be->blkif = NULL; | |
579 | xenbus_dev_fatal(dev, err, "creating block interface"); | |
580 | goto fail; | |
581 | } | |
582 | ||
583 | /* setup back pointer */ | |
584 | be->blkif->be = be; | |
585 | ||
88122933 JF |
586 | err = xenbus_watch_pathfmt(dev, &be->backend_watch, backend_changed, |
587 | "%s/%s", dev->nodename, "physical-device"); | |
4d05a28d KRW |
588 | if (err) |
589 | goto fail; | |
590 | ||
591 | err = xenbus_switch_state(dev, XenbusStateInitWait); | |
592 | if (err) | |
593 | goto fail; | |
594 | ||
595 | return 0; | |
596 | ||
597 | fail: | |
598 | DPRINTK("failed"); | |
8b6bf747 | 599 | xen_blkbk_remove(dev); |
4d05a28d KRW |
600 | return err; |
601 | } | |
602 | ||
603 | ||
01f37f2d | 604 | /* |
4d05a28d KRW |
605 | * Callback received when the hotplug scripts have placed the physical-device |
606 | * node. Read it and the mode node, and create a vbd. If the frontend is | |
607 | * ready, connect. | |
608 | */ | |
609 | static void backend_changed(struct xenbus_watch *watch, | |
610 | const char **vec, unsigned int len) | |
611 | { | |
612 | int err; | |
613 | unsigned major; | |
614 | unsigned minor; | |
615 | struct backend_info *be | |
616 | = container_of(watch, struct backend_info, backend_watch); | |
617 | struct xenbus_device *dev = be->dev; | |
618 | int cdrom = 0; | |
9d092603 | 619 | unsigned long handle; |
4d05a28d KRW |
620 | char *device_type; |
621 | ||
622 | DPRINTK(""); | |
623 | ||
624 | err = xenbus_scanf(XBT_NIL, dev->nodename, "physical-device", "%x:%x", | |
625 | &major, &minor); | |
626 | if (XENBUS_EXIST_ERR(err)) { | |
01f37f2d KRW |
627 | /* |
628 | * Since this watch will fire once immediately after it is | |
629 | * registered, we expect this. Ignore it, and wait for the | |
630 | * hotplug scripts. | |
631 | */ | |
4d05a28d KRW |
632 | return; |
633 | } | |
634 | if (err != 2) { | |
635 | xenbus_dev_fatal(dev, err, "reading physical-device"); | |
636 | return; | |
637 | } | |
638 | ||
9d092603 JB |
639 | if (be->major | be->minor) { |
640 | if (be->major != major || be->minor != minor) | |
641 | pr_warn(DRV_PFX "changing physical device (from %x:%x to %x:%x) not supported.\n", | |
642 | be->major, be->minor, major, minor); | |
4d05a28d KRW |
643 | return; |
644 | } | |
645 | ||
646 | be->mode = xenbus_read(XBT_NIL, dev->nodename, "mode", NULL); | |
647 | if (IS_ERR(be->mode)) { | |
648 | err = PTR_ERR(be->mode); | |
649 | be->mode = NULL; | |
650 | xenbus_dev_fatal(dev, err, "reading mode"); | |
651 | return; | |
652 | } | |
653 | ||
654 | device_type = xenbus_read(XBT_NIL, dev->otherend, "device-type", NULL); | |
655 | if (!IS_ERR(device_type)) { | |
656 | cdrom = strcmp(device_type, "cdrom") == 0; | |
657 | kfree(device_type); | |
658 | } | |
659 | ||
9d092603 | 660 | /* Front end dir is a number, which is used as the handle. */ |
bb8e0e84 | 661 | err = kstrtoul(strrchr(dev->otherend, '/') + 1, 0, &handle); |
9d092603 JB |
662 | if (err) |
663 | return; | |
4d05a28d | 664 | |
9d092603 JB |
665 | be->major = major; |
666 | be->minor = minor; | |
4d05a28d | 667 | |
9d092603 JB |
668 | err = xen_vbd_create(be->blkif, handle, major, minor, |
669 | !strchr(be->mode, 'w'), cdrom); | |
4d05a28d | 670 | |
9d092603 JB |
671 | if (err) |
672 | xenbus_dev_fatal(dev, err, "creating vbd structure"); | |
673 | else { | |
4d05a28d KRW |
674 | err = xenvbd_sysfs_addif(dev); |
675 | if (err) { | |
3d814731 | 676 | xen_vbd_free(&be->blkif->vbd); |
4d05a28d | 677 | xenbus_dev_fatal(dev, err, "creating sysfs entries"); |
4d05a28d | 678 | } |
9d092603 | 679 | } |
4d05a28d | 680 | |
9d092603 JB |
681 | if (err) { |
682 | kfree(be->mode); | |
683 | be->mode = NULL; | |
684 | be->major = 0; | |
685 | be->minor = 0; | |
686 | } else { | |
4d05a28d | 687 | /* We're potentially connected now */ |
8b6bf747 | 688 | xen_update_blkif_status(be->blkif); |
4d05a28d KRW |
689 | } |
690 | } | |
691 | ||
692 | ||
01f37f2d | 693 | /* |
4d05a28d KRW |
694 | * Callback received when the frontend's state changes. |
695 | */ | |
696 | static void frontend_changed(struct xenbus_device *dev, | |
697 | enum xenbus_state frontend_state) | |
698 | { | |
5cf6e4f6 | 699 | struct backend_info *be = dev_get_drvdata(&dev->dev); |
4d05a28d KRW |
700 | int err; |
701 | ||
702 | DPRINTK("%s", xenbus_strstate(frontend_state)); | |
703 | ||
704 | switch (frontend_state) { | |
705 | case XenbusStateInitialising: | |
706 | if (dev->state == XenbusStateClosed) { | |
22b20f2d | 707 | pr_info(DRV_PFX "%s: prepare for reconnect\n", |
ebe81906 | 708 | dev->nodename); |
4d05a28d KRW |
709 | xenbus_switch_state(dev, XenbusStateInitWait); |
710 | } | |
711 | break; | |
712 | ||
713 | case XenbusStateInitialised: | |
714 | case XenbusStateConnected: | |
01f37f2d KRW |
715 | /* |
716 | * Ensure we connect even when two watches fire in | |
42b2aa86 | 717 | * close succession and we miss the intermediate value |
01f37f2d KRW |
718 | * of frontend_state. |
719 | */ | |
4d05a28d KRW |
720 | if (dev->state == XenbusStateConnected) |
721 | break; | |
722 | ||
01f37f2d KRW |
723 | /* |
724 | * Enforce precondition before potential leak point. | |
1bc05b0a | 725 | * xen_blkif_disconnect() is idempotent. |
313d7b00 | 726 | */ |
814d04e7 VP |
727 | err = xen_blkif_disconnect(be->blkif); |
728 | if (err) { | |
729 | xenbus_dev_fatal(dev, err, "pending I/O"); | |
730 | break; | |
731 | } | |
313d7b00 | 732 | |
4d05a28d KRW |
733 | err = connect_ring(be); |
734 | if (err) | |
735 | break; | |
8b6bf747 | 736 | xen_update_blkif_status(be->blkif); |
4d05a28d KRW |
737 | break; |
738 | ||
739 | case XenbusStateClosing: | |
4d05a28d KRW |
740 | xenbus_switch_state(dev, XenbusStateClosing); |
741 | break; | |
742 | ||
743 | case XenbusStateClosed: | |
6f5986bc | 744 | xen_blkif_disconnect(be->blkif); |
4d05a28d KRW |
745 | xenbus_switch_state(dev, XenbusStateClosed); |
746 | if (xenbus_dev_is_online(dev)) | |
747 | break; | |
748 | /* fall through if not online */ | |
749 | case XenbusStateUnknown: | |
1bc05b0a | 750 | /* implies xen_blkif_disconnect() via xen_blkbk_remove() */ |
4d05a28d KRW |
751 | device_unregister(&dev->dev); |
752 | break; | |
753 | ||
754 | default: | |
755 | xenbus_dev_fatal(dev, -EINVAL, "saw state %d at frontend", | |
756 | frontend_state); | |
757 | break; | |
758 | } | |
759 | } | |
760 | ||
761 | ||
762 | /* ** Connection ** */ | |
763 | ||
764 | ||
01f37f2d | 765 | /* |
4d05a28d KRW |
766 | * Write the physical details regarding the block device to the store, and |
767 | * switch to Connected state. | |
768 | */ | |
769 | static void connect(struct backend_info *be) | |
770 | { | |
771 | struct xenbus_transaction xbt; | |
772 | int err; | |
773 | struct xenbus_device *dev = be->dev; | |
774 | ||
775 | DPRINTK("%s", dev->otherend); | |
776 | ||
777 | /* Supply the information about the device the frontend needs */ | |
778 | again: | |
779 | err = xenbus_transaction_start(&xbt); | |
780 | if (err) { | |
781 | xenbus_dev_fatal(dev, err, "starting transaction"); | |
782 | return; | |
783 | } | |
784 | ||
3389bb8b KRW |
785 | /* If we can't advertise it is OK. */ |
786 | xen_blkbk_flush_diskcache(xbt, be, be->blkif->vbd.flush_support); | |
4d05a28d | 787 | |
3389bb8b | 788 | xen_blkbk_discard(xbt, be); |
b3cb0d6a | 789 | |
3389bb8b | 790 | xen_blkbk_barrier(xbt, be, be->blkif->vbd.flush_support); |
29bde093 | 791 | |
0a8704a5 RPM |
792 | err = xenbus_printf(xbt, dev->nodename, "feature-persistent", "%u", 1); |
793 | if (err) { | |
794 | xenbus_dev_fatal(dev, err, "writing %s/feature-persistent", | |
795 | dev->nodename); | |
796 | goto abort; | |
797 | } | |
402b27f9 RPM |
798 | err = xenbus_printf(xbt, dev->nodename, "feature-max-indirect-segments", "%u", |
799 | MAX_INDIRECT_SEGMENTS); | |
800 | if (err) | |
801 | dev_warn(&dev->dev, "writing %s/feature-max-indirect-segments (%d)", | |
802 | dev->nodename, err); | |
0a8704a5 | 803 | |
4d05a28d | 804 | err = xenbus_printf(xbt, dev->nodename, "sectors", "%llu", |
42c7841d | 805 | (unsigned long long)vbd_sz(&be->blkif->vbd)); |
4d05a28d KRW |
806 | if (err) { |
807 | xenbus_dev_fatal(dev, err, "writing %s/sectors", | |
808 | dev->nodename); | |
809 | goto abort; | |
810 | } | |
811 | ||
812 | /* FIXME: use a typename instead */ | |
813 | err = xenbus_printf(xbt, dev->nodename, "info", "%u", | |
42c7841d KRW |
814 | be->blkif->vbd.type | |
815 | (be->blkif->vbd.readonly ? VDISK_READONLY : 0)); | |
4d05a28d KRW |
816 | if (err) { |
817 | xenbus_dev_fatal(dev, err, "writing %s/info", | |
818 | dev->nodename); | |
819 | goto abort; | |
820 | } | |
821 | err = xenbus_printf(xbt, dev->nodename, "sector-size", "%lu", | |
42c7841d KRW |
822 | (unsigned long) |
823 | bdev_logical_block_size(be->blkif->vbd.bdev)); | |
4d05a28d KRW |
824 | if (err) { |
825 | xenbus_dev_fatal(dev, err, "writing %s/sector-size", | |
826 | dev->nodename); | |
827 | goto abort; | |
828 | } | |
7c4d7d71 SB |
829 | err = xenbus_printf(xbt, dev->nodename, "physical-sector-size", "%u", |
830 | bdev_physical_block_size(be->blkif->vbd.bdev)); | |
831 | if (err) | |
832 | xenbus_dev_error(dev, err, "writing %s/physical-sector-size", | |
833 | dev->nodename); | |
4d05a28d KRW |
834 | |
835 | err = xenbus_transaction_end(xbt, 0); | |
836 | if (err == -EAGAIN) | |
837 | goto again; | |
838 | if (err) | |
839 | xenbus_dev_fatal(dev, err, "ending transaction"); | |
840 | ||
841 | err = xenbus_switch_state(dev, XenbusStateConnected); | |
842 | if (err) | |
08b8bfc1 | 843 | xenbus_dev_fatal(dev, err, "%s: switching to Connected state", |
4d05a28d KRW |
844 | dev->nodename); |
845 | ||
846 | return; | |
847 | abort: | |
848 | xenbus_transaction_end(xbt, 1); | |
849 | } | |
850 | ||
851 | ||
852 | static int connect_ring(struct backend_info *be) | |
853 | { | |
854 | struct xenbus_device *dev = be->dev; | |
855 | unsigned long ring_ref; | |
856 | unsigned int evtchn; | |
0a8704a5 | 857 | unsigned int pers_grants; |
4d05a28d KRW |
858 | char protocol[64] = ""; |
859 | int err; | |
860 | ||
861 | DPRINTK("%s", dev->otherend); | |
862 | ||
d6091b21 KRW |
863 | err = xenbus_gather(XBT_NIL, dev->otherend, "ring-ref", "%lu", |
864 | &ring_ref, "event-channel", "%u", &evtchn, NULL); | |
4d05a28d KRW |
865 | if (err) { |
866 | xenbus_dev_fatal(dev, err, | |
867 | "reading %s/ring-ref and event-channel", | |
868 | dev->otherend); | |
869 | return err; | |
870 | } | |
871 | ||
b042a3ca | 872 | be->blkif->blk_protocol = BLKIF_PROTOCOL_DEFAULT; |
4d05a28d KRW |
873 | err = xenbus_gather(XBT_NIL, dev->otherend, "protocol", |
874 | "%63s", protocol, NULL); | |
875 | if (err) | |
b042a3ca | 876 | strcpy(protocol, "unspecified, assuming default"); |
4d05a28d KRW |
877 | else if (0 == strcmp(protocol, XEN_IO_PROTO_ABI_NATIVE)) |
878 | be->blkif->blk_protocol = BLKIF_PROTOCOL_NATIVE; | |
879 | else if (0 == strcmp(protocol, XEN_IO_PROTO_ABI_X86_32)) | |
880 | be->blkif->blk_protocol = BLKIF_PROTOCOL_X86_32; | |
881 | else if (0 == strcmp(protocol, XEN_IO_PROTO_ABI_X86_64)) | |
882 | be->blkif->blk_protocol = BLKIF_PROTOCOL_X86_64; | |
883 | else { | |
884 | xenbus_dev_fatal(dev, err, "unknown fe protocol %s", protocol); | |
885 | return -1; | |
886 | } | |
0a8704a5 | 887 | err = xenbus_gather(XBT_NIL, dev->otherend, |
cb5bd4d1 | 888 | "feature-persistent", "%u", |
0a8704a5 RPM |
889 | &pers_grants, NULL); |
890 | if (err) | |
891 | pers_grants = 0; | |
892 | ||
893 | be->blkif->vbd.feature_gnt_persistent = pers_grants; | |
894 | be->blkif->vbd.overflow_max_grants = 0; | |
895 | ||
896 | pr_info(DRV_PFX "ring-ref %ld, event-channel %d, protocol %d (%s) %s\n", | |
897 | ring_ref, evtchn, be->blkif->blk_protocol, protocol, | |
898 | pers_grants ? "persistent grants" : ""); | |
4d05a28d KRW |
899 | |
900 | /* Map the shared frame, irq etc. */ | |
8b6bf747 | 901 | err = xen_blkif_map(be->blkif, ring_ref, evtchn); |
4d05a28d KRW |
902 | if (err) { |
903 | xenbus_dev_fatal(dev, err, "mapping ring-ref %lu port %u", | |
904 | ring_ref, evtchn); | |
905 | return err; | |
906 | } | |
907 | ||
908 | return 0; | |
909 | } | |
910 | ||
8b6bf747 | 911 | static const struct xenbus_device_id xen_blkbk_ids[] = { |
4d05a28d KRW |
912 | { "vbd" }, |
913 | { "" } | |
914 | }; | |
915 | ||
95afae48 DV |
916 | static struct xenbus_driver xen_blkbk_driver = { |
917 | .ids = xen_blkbk_ids, | |
8b6bf747 KRW |
918 | .probe = xen_blkbk_probe, |
919 | .remove = xen_blkbk_remove, | |
4d05a28d | 920 | .otherend_changed = frontend_changed |
95afae48 | 921 | }; |
4d05a28d | 922 | |
8b6bf747 | 923 | int xen_blkif_xenbus_init(void) |
4d05a28d | 924 | { |
73db144b | 925 | return xenbus_register_backend(&xen_blkbk_driver); |
4d05a28d | 926 | } |