xref: /linux/drivers/xen/xenbus/xenbus_client.c (revision e58e871becec2d3b04ed91c0c16fe8deac9c9dfa)
1 /******************************************************************************
2  * Client-facing interface for the Xenbus driver.  In other words, the
3  * interface between the Xenbus and the device-specific code, be it the
4  * frontend or the backend of that driver.
5  *
6  * Copyright (C) 2005 XenSource Ltd
7  *
8  * This program is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU General Public License version 2
10  * as published by the Free Software Foundation; or, when distributed
11  * separately from the Linux kernel or incorporated into other
12  * software packages, subject to the following license:
13  *
14  * Permission is hereby granted, free of charge, to any person obtaining a copy
15  * of this source file (the "Software"), to deal in the Software without
16  * restriction, including without limitation the rights to use, copy, modify,
17  * merge, publish, distribute, sublicense, and/or sell copies of the Software,
18  * and to permit persons to whom the Software is furnished to do so, subject to
19  * the following conditions:
20  *
21  * The above copyright notice and this permission notice shall be included in
22  * all copies or substantial portions of the Software.
23  *
24  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
25  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
26  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
27  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
28  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
29  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
30  * IN THE SOFTWARE.
31  */
32 
33 #include <linux/mm.h>
34 #include <linux/slab.h>
35 #include <linux/types.h>
36 #include <linux/spinlock.h>
37 #include <linux/vmalloc.h>
38 #include <linux/export.h>
39 #include <asm/xen/hypervisor.h>
40 #include <xen/page.h>
41 #include <xen/interface/xen.h>
42 #include <xen/interface/event_channel.h>
43 #include <xen/balloon.h>
44 #include <xen/events.h>
45 #include <xen/grant_table.h>
46 #include <xen/xenbus.h>
47 #include <xen/xen.h>
48 #include <xen/features.h>
49 
50 #include "xenbus.h"
51 
52 #define XENBUS_PAGES(_grants)	(DIV_ROUND_UP(_grants, XEN_PFN_PER_PAGE))
53 
54 #define XENBUS_MAX_RING_PAGES	(XENBUS_PAGES(XENBUS_MAX_RING_GRANTS))
55 
56 struct xenbus_map_node {
57 	struct list_head next;
58 	union {
59 		struct {
60 			struct vm_struct *area;
61 		} pv;
62 		struct {
63 			struct page *pages[XENBUS_MAX_RING_PAGES];
64 			unsigned long addrs[XENBUS_MAX_RING_GRANTS];
65 			void *addr;
66 		} hvm;
67 	};
68 	grant_handle_t handles[XENBUS_MAX_RING_GRANTS];
69 	unsigned int   nr_handles;
70 };
71 
72 static DEFINE_SPINLOCK(xenbus_valloc_lock);
73 static LIST_HEAD(xenbus_valloc_pages);
74 
75 struct xenbus_ring_ops {
76 	int (*map)(struct xenbus_device *dev,
77 		   grant_ref_t *gnt_refs, unsigned int nr_grefs,
78 		   void **vaddr);
79 	int (*unmap)(struct xenbus_device *dev, void *vaddr);
80 };
81 
82 static const struct xenbus_ring_ops *ring_ops __read_mostly;
83 
84 const char *xenbus_strstate(enum xenbus_state state)
85 {
86 	static const char *const name[] = {
87 		[ XenbusStateUnknown      ] = "Unknown",
88 		[ XenbusStateInitialising ] = "Initialising",
89 		[ XenbusStateInitWait     ] = "InitWait",
90 		[ XenbusStateInitialised  ] = "Initialised",
91 		[ XenbusStateConnected    ] = "Connected",
92 		[ XenbusStateClosing      ] = "Closing",
93 		[ XenbusStateClosed	  ] = "Closed",
94 		[XenbusStateReconfiguring] = "Reconfiguring",
95 		[XenbusStateReconfigured] = "Reconfigured",
96 	};
97 	return (state < ARRAY_SIZE(name)) ? name[state] : "INVALID";
98 }
99 EXPORT_SYMBOL_GPL(xenbus_strstate);
100 
101 /**
102  * xenbus_watch_path - register a watch
103  * @dev: xenbus device
104  * @path: path to watch
105  * @watch: watch to register
106  * @callback: callback to register
107  *
108  * Register a @watch on the given path, using the given xenbus_watch structure
109  * for storage, and the given @callback function as the callback.  Return 0 on
110  * success, or -errno on error.  On success, the given @path will be saved as
111  * @watch->node, and remains the caller's to free.  On error, @watch->node will
112  * be NULL, the device will switch to %XenbusStateClosing, and the error will
113  * be saved in the store.
114  */
115 int xenbus_watch_path(struct xenbus_device *dev, const char *path,
116 		      struct xenbus_watch *watch,
117 		      void (*callback)(struct xenbus_watch *,
118 				       const char *, const char *))
119 {
120 	int err;
121 
122 	watch->node = path;
123 	watch->callback = callback;
124 
125 	err = register_xenbus_watch(watch);
126 
127 	if (err) {
128 		watch->node = NULL;
129 		watch->callback = NULL;
130 		xenbus_dev_fatal(dev, err, "adding watch on %s", path);
131 	}
132 
133 	return err;
134 }
135 EXPORT_SYMBOL_GPL(xenbus_watch_path);
136 
137 
138 /**
139  * xenbus_watch_pathfmt - register a watch on a sprintf-formatted path
140  * @dev: xenbus device
141  * @watch: watch to register
142  * @callback: callback to register
143  * @pathfmt: format of path to watch
144  *
145  * Register a watch on the given @path, using the given xenbus_watch
146  * structure for storage, and the given @callback function as the callback.
147  * Return 0 on success, or -errno on error.  On success, the watched path
148  * (@path/@path2) will be saved as @watch->node, and becomes the caller's to
149  * kfree().  On error, watch->node will be NULL, so the caller has nothing to
150  * free, the device will switch to %XenbusStateClosing, and the error will be
151  * saved in the store.
152  */
153 int xenbus_watch_pathfmt(struct xenbus_device *dev,
154 			 struct xenbus_watch *watch,
155 			 void (*callback)(struct xenbus_watch *,
156 					  const char *, const char *),
157 			 const char *pathfmt, ...)
158 {
159 	int err;
160 	va_list ap;
161 	char *path;
162 
163 	va_start(ap, pathfmt);
164 	path = kvasprintf(GFP_NOIO | __GFP_HIGH, pathfmt, ap);
165 	va_end(ap);
166 
167 	if (!path) {
168 		xenbus_dev_fatal(dev, -ENOMEM, "allocating path for watch");
169 		return -ENOMEM;
170 	}
171 	err = xenbus_watch_path(dev, path, watch, callback);
172 
173 	if (err)
174 		kfree(path);
175 	return err;
176 }
177 EXPORT_SYMBOL_GPL(xenbus_watch_pathfmt);
178 
179 static void xenbus_switch_fatal(struct xenbus_device *, int, int,
180 				const char *, ...);
181 
182 static int
183 __xenbus_switch_state(struct xenbus_device *dev,
184 		      enum xenbus_state state, int depth)
185 {
186 	/* We check whether the state is currently set to the given value, and
187 	   if not, then the state is set.  We don't want to unconditionally
188 	   write the given state, because we don't want to fire watches
189 	   unnecessarily.  Furthermore, if the node has gone, we don't write
190 	   to it, as the device will be tearing down, and we don't want to
191 	   resurrect that directory.
192 
193 	   Note that, because of this cached value of our state, this
194 	   function will not take a caller's Xenstore transaction
195 	   (something it was trying to in the past) because dev->state
196 	   would not get reset if the transaction was aborted.
197 	 */
198 
199 	struct xenbus_transaction xbt;
200 	int current_state;
201 	int err, abort;
202 
203 	if (state == dev->state)
204 		return 0;
205 
206 again:
207 	abort = 1;
208 
209 	err = xenbus_transaction_start(&xbt);
210 	if (err) {
211 		xenbus_switch_fatal(dev, depth, err, "starting transaction");
212 		return 0;
213 	}
214 
215 	err = xenbus_scanf(xbt, dev->nodename, "state", "%d", &current_state);
216 	if (err != 1)
217 		goto abort;
218 
219 	err = xenbus_printf(xbt, dev->nodename, "state", "%d", state);
220 	if (err) {
221 		xenbus_switch_fatal(dev, depth, err, "writing new state");
222 		goto abort;
223 	}
224 
225 	abort = 0;
226 abort:
227 	err = xenbus_transaction_end(xbt, abort);
228 	if (err) {
229 		if (err == -EAGAIN && !abort)
230 			goto again;
231 		xenbus_switch_fatal(dev, depth, err, "ending transaction");
232 	} else
233 		dev->state = state;
234 
235 	return 0;
236 }
237 
238 /**
239  * xenbus_switch_state
240  * @dev: xenbus device
241  * @state: new state
242  *
243  * Advertise in the store a change of the given driver to the given new_state.
244  * Return 0 on success, or -errno on error.  On error, the device will switch
245  * to XenbusStateClosing, and the error will be saved in the store.
246  */
247 int xenbus_switch_state(struct xenbus_device *dev, enum xenbus_state state)
248 {
249 	return __xenbus_switch_state(dev, state, 0);
250 }
251 
252 EXPORT_SYMBOL_GPL(xenbus_switch_state);
253 
254 int xenbus_frontend_closed(struct xenbus_device *dev)
255 {
256 	xenbus_switch_state(dev, XenbusStateClosed);
257 	complete(&dev->down);
258 	return 0;
259 }
260 EXPORT_SYMBOL_GPL(xenbus_frontend_closed);
261 
262 static void xenbus_va_dev_error(struct xenbus_device *dev, int err,
263 				const char *fmt, va_list ap)
264 {
265 	unsigned int len;
266 	char *printf_buffer;
267 	char *path_buffer;
268 
269 #define PRINTF_BUFFER_SIZE 4096
270 
271 	printf_buffer = kmalloc(PRINTF_BUFFER_SIZE, GFP_KERNEL);
272 	if (!printf_buffer)
273 		return;
274 
275 	len = sprintf(printf_buffer, "%i ", -err);
276 	vsnprintf(printf_buffer + len, PRINTF_BUFFER_SIZE - len, fmt, ap);
277 
278 	dev_err(&dev->dev, "%s\n", printf_buffer);
279 
280 	path_buffer = kasprintf(GFP_KERNEL, "error/%s", dev->nodename);
281 	if (!path_buffer ||
282 	    xenbus_write(XBT_NIL, path_buffer, "error", printf_buffer))
283 		dev_err(&dev->dev, "failed to write error node for %s (%s)\n",
284 			dev->nodename, printf_buffer);
285 
286 	kfree(printf_buffer);
287 	kfree(path_buffer);
288 }
289 
290 /**
291  * xenbus_dev_error
292  * @dev: xenbus device
293  * @err: error to report
294  * @fmt: error message format
295  *
296  * Report the given negative errno into the store, along with the given
297  * formatted message.
298  */
299 void xenbus_dev_error(struct xenbus_device *dev, int err, const char *fmt, ...)
300 {
301 	va_list ap;
302 
303 	va_start(ap, fmt);
304 	xenbus_va_dev_error(dev, err, fmt, ap);
305 	va_end(ap);
306 }
307 EXPORT_SYMBOL_GPL(xenbus_dev_error);
308 
309 /**
310  * xenbus_dev_fatal
311  * @dev: xenbus device
312  * @err: error to report
313  * @fmt: error message format
314  *
315  * Equivalent to xenbus_dev_error(dev, err, fmt, args), followed by
316  * xenbus_switch_state(dev, XenbusStateClosing) to schedule an orderly
317  * closedown of this driver and its peer.
318  */
319 
320 void xenbus_dev_fatal(struct xenbus_device *dev, int err, const char *fmt, ...)
321 {
322 	va_list ap;
323 
324 	va_start(ap, fmt);
325 	xenbus_va_dev_error(dev, err, fmt, ap);
326 	va_end(ap);
327 
328 	xenbus_switch_state(dev, XenbusStateClosing);
329 }
330 EXPORT_SYMBOL_GPL(xenbus_dev_fatal);
331 
332 /**
333  * Equivalent to xenbus_dev_fatal(dev, err, fmt, args), but helps
334  * avoiding recursion within xenbus_switch_state.
335  */
336 static void xenbus_switch_fatal(struct xenbus_device *dev, int depth, int err,
337 				const char *fmt, ...)
338 {
339 	va_list ap;
340 
341 	va_start(ap, fmt);
342 	xenbus_va_dev_error(dev, err, fmt, ap);
343 	va_end(ap);
344 
345 	if (!depth)
346 		__xenbus_switch_state(dev, XenbusStateClosing, 1);
347 }
348 
349 /**
350  * xenbus_grant_ring
351  * @dev: xenbus device
352  * @vaddr: starting virtual address of the ring
353  * @nr_pages: number of pages to be granted
354  * @grefs: grant reference array to be filled in
355  *
356  * Grant access to the given @vaddr to the peer of the given device.
357  * Then fill in @grefs with grant references.  Return 0 on success, or
358  * -errno on error.  On error, the device will switch to
359  * XenbusStateClosing, and the error will be saved in the store.
360  */
361 int xenbus_grant_ring(struct xenbus_device *dev, void *vaddr,
362 		      unsigned int nr_pages, grant_ref_t *grefs)
363 {
364 	int err;
365 	int i, j;
366 
367 	for (i = 0; i < nr_pages; i++) {
368 		err = gnttab_grant_foreign_access(dev->otherend_id,
369 						  virt_to_gfn(vaddr), 0);
370 		if (err < 0) {
371 			xenbus_dev_fatal(dev, err,
372 					 "granting access to ring page");
373 			goto fail;
374 		}
375 		grefs[i] = err;
376 
377 		vaddr = vaddr + XEN_PAGE_SIZE;
378 	}
379 
380 	return 0;
381 
382 fail:
383 	for (j = 0; j < i; j++)
384 		gnttab_end_foreign_access_ref(grefs[j], 0);
385 	return err;
386 }
387 EXPORT_SYMBOL_GPL(xenbus_grant_ring);
388 
389 
390 /**
391  * Allocate an event channel for the given xenbus_device, assigning the newly
392  * created local port to *port.  Return 0 on success, or -errno on error.  On
393  * error, the device will switch to XenbusStateClosing, and the error will be
394  * saved in the store.
395  */
396 int xenbus_alloc_evtchn(struct xenbus_device *dev, int *port)
397 {
398 	struct evtchn_alloc_unbound alloc_unbound;
399 	int err;
400 
401 	alloc_unbound.dom = DOMID_SELF;
402 	alloc_unbound.remote_dom = dev->otherend_id;
403 
404 	err = HYPERVISOR_event_channel_op(EVTCHNOP_alloc_unbound,
405 					  &alloc_unbound);
406 	if (err)
407 		xenbus_dev_fatal(dev, err, "allocating event channel");
408 	else
409 		*port = alloc_unbound.port;
410 
411 	return err;
412 }
413 EXPORT_SYMBOL_GPL(xenbus_alloc_evtchn);
414 
415 
416 /**
417  * Free an existing event channel. Returns 0 on success or -errno on error.
418  */
419 int xenbus_free_evtchn(struct xenbus_device *dev, int port)
420 {
421 	struct evtchn_close close;
422 	int err;
423 
424 	close.port = port;
425 
426 	err = HYPERVISOR_event_channel_op(EVTCHNOP_close, &close);
427 	if (err)
428 		xenbus_dev_error(dev, err, "freeing event channel %d", port);
429 
430 	return err;
431 }
432 EXPORT_SYMBOL_GPL(xenbus_free_evtchn);
433 
434 
435 /**
436  * xenbus_map_ring_valloc
437  * @dev: xenbus device
438  * @gnt_refs: grant reference array
439  * @nr_grefs: number of grant references
440  * @vaddr: pointer to address to be filled out by mapping
441  *
442  * Map @nr_grefs pages of memory into this domain from another
443  * domain's grant table.  xenbus_map_ring_valloc allocates @nr_grefs
444  * pages of virtual address space, maps the pages to that address, and
445  * sets *vaddr to that address.  Returns 0 on success, and GNTST_*
446  * (see xen/include/interface/grant_table.h) or -ENOMEM / -EINVAL on
447  * error. If an error is returned, device will switch to
448  * XenbusStateClosing and the error message will be saved in XenStore.
449  */
450 int xenbus_map_ring_valloc(struct xenbus_device *dev, grant_ref_t *gnt_refs,
451 			   unsigned int nr_grefs, void **vaddr)
452 {
453 	return ring_ops->map(dev, gnt_refs, nr_grefs, vaddr);
454 }
455 EXPORT_SYMBOL_GPL(xenbus_map_ring_valloc);
456 
457 /* N.B. sizeof(phys_addr_t) doesn't always equal to sizeof(unsigned
458  * long), e.g. 32-on-64.  Caller is responsible for preparing the
459  * right array to feed into this function */
460 static int __xenbus_map_ring(struct xenbus_device *dev,
461 			     grant_ref_t *gnt_refs,
462 			     unsigned int nr_grefs,
463 			     grant_handle_t *handles,
464 			     phys_addr_t *addrs,
465 			     unsigned int flags,
466 			     bool *leaked)
467 {
468 	struct gnttab_map_grant_ref map[XENBUS_MAX_RING_GRANTS];
469 	struct gnttab_unmap_grant_ref unmap[XENBUS_MAX_RING_GRANTS];
470 	int i, j;
471 	int err = GNTST_okay;
472 
473 	if (nr_grefs > XENBUS_MAX_RING_GRANTS)
474 		return -EINVAL;
475 
476 	for (i = 0; i < nr_grefs; i++) {
477 		memset(&map[i], 0, sizeof(map[i]));
478 		gnttab_set_map_op(&map[i], addrs[i], flags, gnt_refs[i],
479 				  dev->otherend_id);
480 		handles[i] = INVALID_GRANT_HANDLE;
481 	}
482 
483 	gnttab_batch_map(map, i);
484 
485 	for (i = 0; i < nr_grefs; i++) {
486 		if (map[i].status != GNTST_okay) {
487 			err = map[i].status;
488 			xenbus_dev_fatal(dev, map[i].status,
489 					 "mapping in shared page %d from domain %d",
490 					 gnt_refs[i], dev->otherend_id);
491 			goto fail;
492 		} else
493 			handles[i] = map[i].handle;
494 	}
495 
496 	return GNTST_okay;
497 
498  fail:
499 	for (i = j = 0; i < nr_grefs; i++) {
500 		if (handles[i] != INVALID_GRANT_HANDLE) {
501 			memset(&unmap[j], 0, sizeof(unmap[j]));
502 			gnttab_set_unmap_op(&unmap[j], (phys_addr_t)addrs[i],
503 					    GNTMAP_host_map, handles[i]);
504 			j++;
505 		}
506 	}
507 
508 	if (HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, unmap, j))
509 		BUG();
510 
511 	*leaked = false;
512 	for (i = 0; i < j; i++) {
513 		if (unmap[i].status != GNTST_okay) {
514 			*leaked = true;
515 			break;
516 		}
517 	}
518 
519 	return err;
520 }
521 
522 static int xenbus_map_ring_valloc_pv(struct xenbus_device *dev,
523 				     grant_ref_t *gnt_refs,
524 				     unsigned int nr_grefs,
525 				     void **vaddr)
526 {
527 	struct xenbus_map_node *node;
528 	struct vm_struct *area;
529 	pte_t *ptes[XENBUS_MAX_RING_GRANTS];
530 	phys_addr_t phys_addrs[XENBUS_MAX_RING_GRANTS];
531 	int err = GNTST_okay;
532 	int i;
533 	bool leaked;
534 
535 	*vaddr = NULL;
536 
537 	if (nr_grefs > XENBUS_MAX_RING_GRANTS)
538 		return -EINVAL;
539 
540 	node = kzalloc(sizeof(*node), GFP_KERNEL);
541 	if (!node)
542 		return -ENOMEM;
543 
544 	area = alloc_vm_area(XEN_PAGE_SIZE * nr_grefs, ptes);
545 	if (!area) {
546 		kfree(node);
547 		return -ENOMEM;
548 	}
549 
550 	for (i = 0; i < nr_grefs; i++)
551 		phys_addrs[i] = arbitrary_virt_to_machine(ptes[i]).maddr;
552 
553 	err = __xenbus_map_ring(dev, gnt_refs, nr_grefs, node->handles,
554 				phys_addrs,
555 				GNTMAP_host_map | GNTMAP_contains_pte,
556 				&leaked);
557 	if (err)
558 		goto failed;
559 
560 	node->nr_handles = nr_grefs;
561 	node->pv.area = area;
562 
563 	spin_lock(&xenbus_valloc_lock);
564 	list_add(&node->next, &xenbus_valloc_pages);
565 	spin_unlock(&xenbus_valloc_lock);
566 
567 	*vaddr = area->addr;
568 	return 0;
569 
570 failed:
571 	if (!leaked)
572 		free_vm_area(area);
573 	else
574 		pr_alert("leaking VM area %p size %u page(s)", area, nr_grefs);
575 
576 	kfree(node);
577 	return err;
578 }
579 
580 struct map_ring_valloc_hvm
581 {
582 	unsigned int idx;
583 
584 	/* Why do we need two arrays? See comment of __xenbus_map_ring */
585 	phys_addr_t phys_addrs[XENBUS_MAX_RING_GRANTS];
586 	unsigned long addrs[XENBUS_MAX_RING_GRANTS];
587 };
588 
589 static void xenbus_map_ring_setup_grant_hvm(unsigned long gfn,
590 					    unsigned int goffset,
591 					    unsigned int len,
592 					    void *data)
593 {
594 	struct map_ring_valloc_hvm *info = data;
595 	unsigned long vaddr = (unsigned long)gfn_to_virt(gfn);
596 
597 	info->phys_addrs[info->idx] = vaddr;
598 	info->addrs[info->idx] = vaddr;
599 
600 	info->idx++;
601 }
602 
603 static int xenbus_map_ring_valloc_hvm(struct xenbus_device *dev,
604 				      grant_ref_t *gnt_ref,
605 				      unsigned int nr_grefs,
606 				      void **vaddr)
607 {
608 	struct xenbus_map_node *node;
609 	int err;
610 	void *addr;
611 	bool leaked = false;
612 	struct map_ring_valloc_hvm info = {
613 		.idx = 0,
614 	};
615 	unsigned int nr_pages = XENBUS_PAGES(nr_grefs);
616 
617 	if (nr_grefs > XENBUS_MAX_RING_GRANTS)
618 		return -EINVAL;
619 
620 	*vaddr = NULL;
621 
622 	node = kzalloc(sizeof(*node), GFP_KERNEL);
623 	if (!node)
624 		return -ENOMEM;
625 
626 	err = alloc_xenballooned_pages(nr_pages, node->hvm.pages);
627 	if (err)
628 		goto out_err;
629 
630 	gnttab_foreach_grant(node->hvm.pages, nr_grefs,
631 			     xenbus_map_ring_setup_grant_hvm,
632 			     &info);
633 
634 	err = __xenbus_map_ring(dev, gnt_ref, nr_grefs, node->handles,
635 				info.phys_addrs, GNTMAP_host_map, &leaked);
636 	node->nr_handles = nr_grefs;
637 
638 	if (err)
639 		goto out_free_ballooned_pages;
640 
641 	addr = vmap(node->hvm.pages, nr_pages, VM_MAP | VM_IOREMAP,
642 		    PAGE_KERNEL);
643 	if (!addr) {
644 		err = -ENOMEM;
645 		goto out_xenbus_unmap_ring;
646 	}
647 
648 	node->hvm.addr = addr;
649 
650 	spin_lock(&xenbus_valloc_lock);
651 	list_add(&node->next, &xenbus_valloc_pages);
652 	spin_unlock(&xenbus_valloc_lock);
653 
654 	*vaddr = addr;
655 	return 0;
656 
657  out_xenbus_unmap_ring:
658 	if (!leaked)
659 		xenbus_unmap_ring(dev, node->handles, nr_grefs, info.addrs);
660 	else
661 		pr_alert("leaking %p size %u page(s)",
662 			 addr, nr_pages);
663  out_free_ballooned_pages:
664 	if (!leaked)
665 		free_xenballooned_pages(nr_pages, node->hvm.pages);
666  out_err:
667 	kfree(node);
668 	return err;
669 }
670 
671 
672 /**
673  * xenbus_map_ring
674  * @dev: xenbus device
675  * @gnt_refs: grant reference array
676  * @nr_grefs: number of grant reference
677  * @handles: pointer to grant handle to be filled
678  * @vaddrs: addresses to be mapped to
679  * @leaked: fail to clean up a failed map, caller should not free vaddr
680  *
681  * Map pages of memory into this domain from another domain's grant table.
682  * xenbus_map_ring does not allocate the virtual address space (you must do
683  * this yourself!). It only maps in the pages to the specified address.
684  * Returns 0 on success, and GNTST_* (see xen/include/interface/grant_table.h)
685  * or -ENOMEM / -EINVAL on error. If an error is returned, device will switch to
686  * XenbusStateClosing and the first error message will be saved in XenStore.
687  * Further more if we fail to map the ring, caller should check @leaked.
688  * If @leaked is not zero it means xenbus_map_ring fails to clean up, caller
689  * should not free the address space of @vaddr.
690  */
691 int xenbus_map_ring(struct xenbus_device *dev, grant_ref_t *gnt_refs,
692 		    unsigned int nr_grefs, grant_handle_t *handles,
693 		    unsigned long *vaddrs, bool *leaked)
694 {
695 	phys_addr_t phys_addrs[XENBUS_MAX_RING_GRANTS];
696 	int i;
697 
698 	if (nr_grefs > XENBUS_MAX_RING_GRANTS)
699 		return -EINVAL;
700 
701 	for (i = 0; i < nr_grefs; i++)
702 		phys_addrs[i] = (unsigned long)vaddrs[i];
703 
704 	return __xenbus_map_ring(dev, gnt_refs, nr_grefs, handles,
705 				 phys_addrs, GNTMAP_host_map, leaked);
706 }
707 EXPORT_SYMBOL_GPL(xenbus_map_ring);
708 
709 
710 /**
711  * xenbus_unmap_ring_vfree
712  * @dev: xenbus device
713  * @vaddr: addr to unmap
714  *
715  * Based on Rusty Russell's skeleton driver's unmap_page.
716  * Unmap a page of memory in this domain that was imported from another domain.
717  * Use xenbus_unmap_ring_vfree if you mapped in your memory with
718  * xenbus_map_ring_valloc (it will free the virtual address space).
719  * Returns 0 on success and returns GNTST_* on error
720  * (see xen/include/interface/grant_table.h).
721  */
722 int xenbus_unmap_ring_vfree(struct xenbus_device *dev, void *vaddr)
723 {
724 	return ring_ops->unmap(dev, vaddr);
725 }
726 EXPORT_SYMBOL_GPL(xenbus_unmap_ring_vfree);
727 
728 static int xenbus_unmap_ring_vfree_pv(struct xenbus_device *dev, void *vaddr)
729 {
730 	struct xenbus_map_node *node;
731 	struct gnttab_unmap_grant_ref unmap[XENBUS_MAX_RING_GRANTS];
732 	unsigned int level;
733 	int i;
734 	bool leaked = false;
735 	int err;
736 
737 	spin_lock(&xenbus_valloc_lock);
738 	list_for_each_entry(node, &xenbus_valloc_pages, next) {
739 		if (node->pv.area->addr == vaddr) {
740 			list_del(&node->next);
741 			goto found;
742 		}
743 	}
744 	node = NULL;
745  found:
746 	spin_unlock(&xenbus_valloc_lock);
747 
748 	if (!node) {
749 		xenbus_dev_error(dev, -ENOENT,
750 				 "can't find mapped virtual address %p", vaddr);
751 		return GNTST_bad_virt_addr;
752 	}
753 
754 	for (i = 0; i < node->nr_handles; i++) {
755 		unsigned long addr;
756 
757 		memset(&unmap[i], 0, sizeof(unmap[i]));
758 		addr = (unsigned long)vaddr + (XEN_PAGE_SIZE * i);
759 		unmap[i].host_addr = arbitrary_virt_to_machine(
760 			lookup_address(addr, &level)).maddr;
761 		unmap[i].dev_bus_addr = 0;
762 		unmap[i].handle = node->handles[i];
763 	}
764 
765 	if (HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, unmap, i))
766 		BUG();
767 
768 	err = GNTST_okay;
769 	leaked = false;
770 	for (i = 0; i < node->nr_handles; i++) {
771 		if (unmap[i].status != GNTST_okay) {
772 			leaked = true;
773 			xenbus_dev_error(dev, unmap[i].status,
774 					 "unmapping page at handle %d error %d",
775 					 node->handles[i], unmap[i].status);
776 			err = unmap[i].status;
777 			break;
778 		}
779 	}
780 
781 	if (!leaked)
782 		free_vm_area(node->pv.area);
783 	else
784 		pr_alert("leaking VM area %p size %u page(s)",
785 			 node->pv.area, node->nr_handles);
786 
787 	kfree(node);
788 	return err;
789 }
790 
791 struct unmap_ring_vfree_hvm
792 {
793 	unsigned int idx;
794 	unsigned long addrs[XENBUS_MAX_RING_GRANTS];
795 };
796 
797 static void xenbus_unmap_ring_setup_grant_hvm(unsigned long gfn,
798 					      unsigned int goffset,
799 					      unsigned int len,
800 					      void *data)
801 {
802 	struct unmap_ring_vfree_hvm *info = data;
803 
804 	info->addrs[info->idx] = (unsigned long)gfn_to_virt(gfn);
805 
806 	info->idx++;
807 }
808 
809 static int xenbus_unmap_ring_vfree_hvm(struct xenbus_device *dev, void *vaddr)
810 {
811 	int rv;
812 	struct xenbus_map_node *node;
813 	void *addr;
814 	struct unmap_ring_vfree_hvm info = {
815 		.idx = 0,
816 	};
817 	unsigned int nr_pages;
818 
819 	spin_lock(&xenbus_valloc_lock);
820 	list_for_each_entry(node, &xenbus_valloc_pages, next) {
821 		addr = node->hvm.addr;
822 		if (addr == vaddr) {
823 			list_del(&node->next);
824 			goto found;
825 		}
826 	}
827 	node = addr = NULL;
828  found:
829 	spin_unlock(&xenbus_valloc_lock);
830 
831 	if (!node) {
832 		xenbus_dev_error(dev, -ENOENT,
833 				 "can't find mapped virtual address %p", vaddr);
834 		return GNTST_bad_virt_addr;
835 	}
836 
837 	nr_pages = XENBUS_PAGES(node->nr_handles);
838 
839 	gnttab_foreach_grant(node->hvm.pages, node->nr_handles,
840 			     xenbus_unmap_ring_setup_grant_hvm,
841 			     &info);
842 
843 	rv = xenbus_unmap_ring(dev, node->handles, node->nr_handles,
844 			       info.addrs);
845 	if (!rv) {
846 		vunmap(vaddr);
847 		free_xenballooned_pages(nr_pages, node->hvm.pages);
848 	}
849 	else
850 		WARN(1, "Leaking %p, size %u page(s)\n", vaddr, nr_pages);
851 
852 	kfree(node);
853 	return rv;
854 }
855 
856 /**
857  * xenbus_unmap_ring
858  * @dev: xenbus device
859  * @handles: grant handle array
860  * @nr_handles: number of handles in the array
861  * @vaddrs: addresses to unmap
862  *
863  * Unmap memory in this domain that was imported from another domain.
864  * Returns 0 on success and returns GNTST_* on error
865  * (see xen/include/interface/grant_table.h).
866  */
867 int xenbus_unmap_ring(struct xenbus_device *dev,
868 		      grant_handle_t *handles, unsigned int nr_handles,
869 		      unsigned long *vaddrs)
870 {
871 	struct gnttab_unmap_grant_ref unmap[XENBUS_MAX_RING_GRANTS];
872 	int i;
873 	int err;
874 
875 	if (nr_handles > XENBUS_MAX_RING_GRANTS)
876 		return -EINVAL;
877 
878 	for (i = 0; i < nr_handles; i++)
879 		gnttab_set_unmap_op(&unmap[i], vaddrs[i],
880 				    GNTMAP_host_map, handles[i]);
881 
882 	if (HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, unmap, i))
883 		BUG();
884 
885 	err = GNTST_okay;
886 	for (i = 0; i < nr_handles; i++) {
887 		if (unmap[i].status != GNTST_okay) {
888 			xenbus_dev_error(dev, unmap[i].status,
889 					 "unmapping page at handle %d error %d",
890 					 handles[i], unmap[i].status);
891 			err = unmap[i].status;
892 			break;
893 		}
894 	}
895 
896 	return err;
897 }
898 EXPORT_SYMBOL_GPL(xenbus_unmap_ring);
899 
900 
901 /**
902  * xenbus_read_driver_state
903  * @path: path for driver
904  *
905  * Return the state of the driver rooted at the given store path, or
906  * XenbusStateUnknown if no state can be read.
907  */
908 enum xenbus_state xenbus_read_driver_state(const char *path)
909 {
910 	enum xenbus_state result;
911 	int err = xenbus_gather(XBT_NIL, path, "state", "%d", &result, NULL);
912 	if (err)
913 		result = XenbusStateUnknown;
914 
915 	return result;
916 }
917 EXPORT_SYMBOL_GPL(xenbus_read_driver_state);
918 
919 static const struct xenbus_ring_ops ring_ops_pv = {
920 	.map = xenbus_map_ring_valloc_pv,
921 	.unmap = xenbus_unmap_ring_vfree_pv,
922 };
923 
924 static const struct xenbus_ring_ops ring_ops_hvm = {
925 	.map = xenbus_map_ring_valloc_hvm,
926 	.unmap = xenbus_unmap_ring_vfree_hvm,
927 };
928 
929 void __init xenbus_ring_ops_init(void)
930 {
931 	if (!xen_feature(XENFEAT_auto_translated_physmap))
932 		ring_ops = &ring_ops_pv;
933 	else
934 		ring_ops = &ring_ops_hvm;
935 }
936