xref: /openbmc/linux/drivers/xen/evtchn.c (revision 82003e04) 
1 /******************************************************************************
2  * evtchn.c
3  *
4  * Driver for receiving and demuxing event-channel signals.
5  *
6  * Copyright (c) 2004-2005, K A Fraser
7  * Multi-process extensions Copyright (c) 2004, Steven Smith
8  *
9  * This program is free software; you can redistribute it and/or
10  * modify it under the terms of the GNU General Public License version 2
11  * as published by the Free Software Foundation; or, when distributed
12  * separately from the Linux kernel or incorporated into other
13  * software packages, subject to the following license:
14  *
15  * Permission is hereby granted, free of charge, to any person obtaining a copy
16  * of this source file (the "Software"), to deal in the Software without
17  * restriction, including without limitation the rights to use, copy, modify,
18  * merge, publish, distribute, sublicense, and/or sell copies of the Software,
19  * and to permit persons to whom the Software is furnished to do so, subject to
20  * the following conditions:
21  *
22  * The above copyright notice and this permission notice shall be included in
23  * all copies or substantial portions of the Software.
24  *
25  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
26  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
27  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
28  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
29  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
30  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
31  * IN THE SOFTWARE.
32  */
33 
34 #define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt
35 
36 #include <linux/module.h>
37 #include <linux/kernel.h>
38 #include <linux/sched.h>
39 #include <linux/slab.h>
40 #include <linux/string.h>
41 #include <linux/errno.h>
42 #include <linux/fs.h>
43 #include <linux/miscdevice.h>
44 #include <linux/major.h>
45 #include <linux/proc_fs.h>
46 #include <linux/stat.h>
47 #include <linux/poll.h>
48 #include <linux/irq.h>
49 #include <linux/init.h>
50 #include <linux/mutex.h>
51 #include <linux/cpu.h>
52 #include <linux/mm.h>
53 #include <linux/vmalloc.h>
54 
55 #include <xen/xen.h>
56 #include <xen/events.h>
57 #include <xen/evtchn.h>
58 #include <xen/xen-ops.h>
59 #include <asm/xen/hypervisor.h>
60 
61 struct per_user_data {
62 	struct mutex bind_mutex; /* serialize bind/unbind operations */
63 	struct rb_root evtchns;
64 	unsigned int nr_evtchns;
65 
66 	/* Notification ring, accessed via /dev/xen/evtchn. */
67 	unsigned int ring_size;
68 	evtchn_port_t *ring;
69 	unsigned int ring_cons, ring_prod, ring_overflow;
70 	struct mutex ring_cons_mutex; /* protect against concurrent readers */
71 	spinlock_t ring_prod_lock; /* product against concurrent interrupts */
72 
73 	/* Processes wait on this queue when ring is empty. */
74 	wait_queue_head_t evtchn_wait;
75 	struct fasync_struct *evtchn_async_queue;
76 	const char *name;
77 
78 	domid_t restrict_domid;
79 };
80 
81 #define UNRESTRICTED_DOMID ((domid_t)-1)
82 
83 struct user_evtchn {
84 	struct rb_node node;
85 	struct per_user_data *user;
86 	unsigned port;
87 	bool enabled;
88 };
89 
90 static evtchn_port_t *evtchn_alloc_ring(unsigned int size)
91 {
92 	evtchn_port_t *ring;
93 	size_t s = size * sizeof(*ring);
94 
95 	ring = kmalloc(s, GFP_KERNEL);
96 	if (!ring)
97 		ring = vmalloc(s);
98 
99 	return ring;
100 }
101 
102 static void evtchn_free_ring(evtchn_port_t *ring)
103 {
104 	kvfree(ring);
105 }
106 
107 static unsigned int evtchn_ring_offset(struct per_user_data *u,
108 				       unsigned int idx)
109 {
110 	return idx & (u->ring_size - 1);
111 }
112 
113 static evtchn_port_t *evtchn_ring_entry(struct per_user_data *u,
114 					unsigned int idx)
115 {
116 	return u->ring + evtchn_ring_offset(u, idx);
117 }
118 
119 static int add_evtchn(struct per_user_data *u, struct user_evtchn *evtchn)
120 {
121 	struct rb_node **new = &(u->evtchns.rb_node), *parent = NULL;
122 
123 	u->nr_evtchns++;
124 
125 	while (*new) {
126 		struct user_evtchn *this;
127 
128 		this = container_of(*new, struct user_evtchn, node);
129 
130 		parent = *new;
131 		if (this->port < evtchn->port)
132 			new = &((*new)->rb_left);
133 		else if (this->port > evtchn->port)
134 			new = &((*new)->rb_right);
135 		else
136 			return -EEXIST;
137 	}
138 
139 	/* Add new node and rebalance tree. */
140 	rb_link_node(&evtchn->node, parent, new);
141 	rb_insert_color(&evtchn->node, &u->evtchns);
142 
143 	return 0;
144 }
145 
146 static void del_evtchn(struct per_user_data *u, struct user_evtchn *evtchn)
147 {
148 	u->nr_evtchns--;
149 	rb_erase(&evtchn->node, &u->evtchns);
150 	kfree(evtchn);
151 }
152 
153 static struct user_evtchn *find_evtchn(struct per_user_data *u, unsigned port)
154 {
155 	struct rb_node *node = u->evtchns.rb_node;
156 
157 	while (node) {
158 		struct user_evtchn *evtchn;
159 
160 		evtchn = container_of(node, struct user_evtchn, node);
161 
162 		if (evtchn->port < port)
163 			node = node->rb_left;
164 		else if (evtchn->port > port)
165 			node = node->rb_right;
166 		else
167 			return evtchn;
168 	}
169 	return NULL;
170 }
171 
172 static irqreturn_t evtchn_interrupt(int irq, void *data)
173 {
174 	struct user_evtchn *evtchn = data;
175 	struct per_user_data *u = evtchn->user;
176 
177 	WARN(!evtchn->enabled,
178 	     "Interrupt for port %d, but apparently not enabled; per-user %p\n",
179 	     evtchn->port, u);
180 
181 	disable_irq_nosync(irq);
182 	evtchn->enabled = false;
183 
184 	spin_lock(&u->ring_prod_lock);
185 
186 	if ((u->ring_prod - u->ring_cons) < u->ring_size) {
187 		*evtchn_ring_entry(u, u->ring_prod) = evtchn->port;
188 		wmb(); /* Ensure ring contents visible */
189 		if (u->ring_cons == u->ring_prod++) {
190 			wake_up_interruptible(&u->evtchn_wait);
191 			kill_fasync(&u->evtchn_async_queue,
192 				    SIGIO, POLL_IN);
193 		}
194 	} else
195 		u->ring_overflow = 1;
196 
197 	spin_unlock(&u->ring_prod_lock);
198 
199 	return IRQ_HANDLED;
200 }
201 
202 static ssize_t evtchn_read(struct file *file, char __user *buf,
203 			   size_t count, loff_t *ppos)
204 {
205 	int rc;
206 	unsigned int c, p, bytes1 = 0, bytes2 = 0;
207 	struct per_user_data *u = file->private_data;
208 
209 	/* Whole number of ports. */
210 	count &= ~(sizeof(evtchn_port_t)-1);
211 
212 	if (count == 0)
213 		return 0;
214 
215 	if (count > PAGE_SIZE)
216 		count = PAGE_SIZE;
217 
218 	for (;;) {
219 		mutex_lock(&u->ring_cons_mutex);
220 
221 		rc = -EFBIG;
222 		if (u->ring_overflow)
223 			goto unlock_out;
224 
225 		c = u->ring_cons;
226 		p = u->ring_prod;
227 		if (c != p)
228 			break;
229 
230 		mutex_unlock(&u->ring_cons_mutex);
231 
232 		if (file->f_flags & O_NONBLOCK)
233 			return -EAGAIN;
234 
235 		rc = wait_event_interruptible(u->evtchn_wait,
236 					      u->ring_cons != u->ring_prod);
237 		if (rc)
238 			return rc;
239 	}
240 
241 	/* Byte lengths of two chunks. Chunk split (if any) is at ring wrap. */
242 	if (((c ^ p) & u->ring_size) != 0) {
243 		bytes1 = (u->ring_size - evtchn_ring_offset(u, c)) *
244 			sizeof(evtchn_port_t);
245 		bytes2 = evtchn_ring_offset(u, p) * sizeof(evtchn_port_t);
246 	} else {
247 		bytes1 = (p - c) * sizeof(evtchn_port_t);
248 		bytes2 = 0;
249 	}
250 
251 	/* Truncate chunks according to caller's maximum byte count. */
252 	if (bytes1 > count) {
253 		bytes1 = count;
254 		bytes2 = 0;
255 	} else if ((bytes1 + bytes2) > count) {
256 		bytes2 = count - bytes1;
257 	}
258 
259 	rc = -EFAULT;
260 	rmb(); /* Ensure that we see the port before we copy it. */
261 	if (copy_to_user(buf, evtchn_ring_entry(u, c), bytes1) ||
262 	    ((bytes2 != 0) &&
263 	     copy_to_user(&buf[bytes1], &u->ring[0], bytes2)))
264 		goto unlock_out;
265 
266 	u->ring_cons += (bytes1 + bytes2) / sizeof(evtchn_port_t);
267 	rc = bytes1 + bytes2;
268 
269  unlock_out:
270 	mutex_unlock(&u->ring_cons_mutex);
271 	return rc;
272 }
273 
274 static ssize_t evtchn_write(struct file *file, const char __user *buf,
275 			    size_t count, loff_t *ppos)
276 {
277 	int rc, i;
278 	evtchn_port_t *kbuf = (evtchn_port_t *)__get_free_page(GFP_KERNEL);
279 	struct per_user_data *u = file->private_data;
280 
281 	if (kbuf == NULL)
282 		return -ENOMEM;
283 
284 	/* Whole number of ports. */
285 	count &= ~(sizeof(evtchn_port_t)-1);
286 
287 	rc = 0;
288 	if (count == 0)
289 		goto out;
290 
291 	if (count > PAGE_SIZE)
292 		count = PAGE_SIZE;
293 
294 	rc = -EFAULT;
295 	if (copy_from_user(kbuf, buf, count) != 0)
296 		goto out;
297 
298 	mutex_lock(&u->bind_mutex);
299 
300 	for (i = 0; i < (count/sizeof(evtchn_port_t)); i++) {
301 		unsigned port = kbuf[i];
302 		struct user_evtchn *evtchn;
303 
304 		evtchn = find_evtchn(u, port);
305 		if (evtchn && !evtchn->enabled) {
306 			evtchn->enabled = true;
307 			enable_irq(irq_from_evtchn(port));
308 		}
309 	}
310 
311 	mutex_unlock(&u->bind_mutex);
312 
313 	rc = count;
314 
315  out:
316 	free_page((unsigned long)kbuf);
317 	return rc;
318 }
319 
320 static int evtchn_resize_ring(struct per_user_data *u)
321 {
322 	unsigned int new_size;
323 	evtchn_port_t *new_ring, *old_ring;
324 
325 	/*
326 	 * Ensure the ring is large enough to capture all possible
327 	 * events. i.e., one free slot for each bound event.
328 	 */
329 	if (u->nr_evtchns <= u->ring_size)
330 		return 0;
331 
332 	if (u->ring_size == 0)
333 		new_size = 64;
334 	else
335 		new_size = 2 * u->ring_size;
336 
337 	new_ring = evtchn_alloc_ring(new_size);
338 	if (!new_ring)
339 		return -ENOMEM;
340 
341 	old_ring = u->ring;
342 
343 	/*
344 	 * Access to the ring contents is serialized by either the
345 	 * prod /or/ cons lock so take both when resizing.
346 	 */
347 	mutex_lock(&u->ring_cons_mutex);
348 	spin_lock_irq(&u->ring_prod_lock);
349 
350 	/*
351 	 * Copy the old ring contents to the new ring.
352 	 *
353 	 * To take care of wrapping, a full ring, and the new index
354 	 * pointing into the second half, simply copy the old contents
355 	 * twice.
356 	 *
357 	 * +---------+    +------------------+
358 	 * |34567  12| -> |34567  1234567  12|
359 	 * +-----p-c-+    +-------c------p---+
360 	 */
361 	memcpy(new_ring, old_ring, u->ring_size * sizeof(*u->ring));
362 	memcpy(new_ring + u->ring_size, old_ring,
363 	       u->ring_size * sizeof(*u->ring));
364 
365 	u->ring = new_ring;
366 	u->ring_size = new_size;
367 
368 	spin_unlock_irq(&u->ring_prod_lock);
369 	mutex_unlock(&u->ring_cons_mutex);
370 
371 	evtchn_free_ring(old_ring);
372 
373 	return 0;
374 }
375 
376 static int evtchn_bind_to_user(struct per_user_data *u, int port)
377 {
378 	struct user_evtchn *evtchn;
379 	struct evtchn_close close;
380 	int rc = 0;
381 
382 	/*
383 	 * Ports are never reused, so every caller should pass in a
384 	 * unique port.
385 	 *
386 	 * (Locking not necessary because we haven't registered the
387 	 * interrupt handler yet, and our caller has already
388 	 * serialized bind operations.)
389 	 */
390 
391 	evtchn = kzalloc(sizeof(*evtchn), GFP_KERNEL);
392 	if (!evtchn)
393 		return -ENOMEM;
394 
395 	evtchn->user = u;
396 	evtchn->port = port;
397 	evtchn->enabled = true; /* start enabled */
398 
399 	rc = add_evtchn(u, evtchn);
400 	if (rc < 0)
401 		goto err;
402 
403 	rc = evtchn_resize_ring(u);
404 	if (rc < 0)
405 		goto err;
406 
407 	rc = bind_evtchn_to_irqhandler(port, evtchn_interrupt, 0,
408 				       u->name, evtchn);
409 	if (rc < 0)
410 		goto err;
411 
412 	rc = evtchn_make_refcounted(port);
413 	return rc;
414 
415 err:
416 	/* bind failed, should close the port now */
417 	close.port = port;
418 	if (HYPERVISOR_event_channel_op(EVTCHNOP_close, &close) != 0)
419 		BUG();
420 	del_evtchn(u, evtchn);
421 	return rc;
422 }
423 
424 static void evtchn_unbind_from_user(struct per_user_data *u,
425 				    struct user_evtchn *evtchn)
426 {
427 	int irq = irq_from_evtchn(evtchn->port);
428 
429 	BUG_ON(irq < 0);
430 
431 	unbind_from_irqhandler(irq, evtchn);
432 
433 	del_evtchn(u, evtchn);
434 }
435 
436 static long evtchn_ioctl(struct file *file,
437 			 unsigned int cmd, unsigned long arg)
438 {
439 	int rc;
440 	struct per_user_data *u = file->private_data;
441 	void __user *uarg = (void __user *) arg;
442 
443 	/* Prevent bind from racing with unbind */
444 	mutex_lock(&u->bind_mutex);
445 
446 	switch (cmd) {
447 	case IOCTL_EVTCHN_BIND_VIRQ: {
448 		struct ioctl_evtchn_bind_virq bind;
449 		struct evtchn_bind_virq bind_virq;
450 
451 		rc = -EACCES;
452 		if (u->restrict_domid != UNRESTRICTED_DOMID)
453 			break;
454 
455 		rc = -EFAULT;
456 		if (copy_from_user(&bind, uarg, sizeof(bind)))
457 			break;
458 
459 		bind_virq.virq = bind.virq;
460 		bind_virq.vcpu = xen_vcpu_nr(0);
461 		rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
462 						 &bind_virq);
463 		if (rc != 0)
464 			break;
465 
466 		rc = evtchn_bind_to_user(u, bind_virq.port);
467 		if (rc == 0)
468 			rc = bind_virq.port;
469 		break;
470 	}
471 
472 	case IOCTL_EVTCHN_BIND_INTERDOMAIN: {
473 		struct ioctl_evtchn_bind_interdomain bind;
474 		struct evtchn_bind_interdomain bind_interdomain;
475 
476 		rc = -EFAULT;
477 		if (copy_from_user(&bind, uarg, sizeof(bind)))
478 			break;
479 
480 		rc = -EACCES;
481 		if (u->restrict_domid != UNRESTRICTED_DOMID &&
482 		    u->restrict_domid != bind.remote_domain)
483 			break;
484 
485 		bind_interdomain.remote_dom  = bind.remote_domain;
486 		bind_interdomain.remote_port = bind.remote_port;
487 		rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_interdomain,
488 						 &bind_interdomain);
489 		if (rc != 0)
490 			break;
491 
492 		rc = evtchn_bind_to_user(u, bind_interdomain.local_port);
493 		if (rc == 0)
494 			rc = bind_interdomain.local_port;
495 		break;
496 	}
497 
498 	case IOCTL_EVTCHN_BIND_UNBOUND_PORT: {
499 		struct ioctl_evtchn_bind_unbound_port bind;
500 		struct evtchn_alloc_unbound alloc_unbound;
501 
502 		rc = -EACCES;
503 		if (u->restrict_domid != UNRESTRICTED_DOMID)
504 			break;
505 
506 		rc = -EFAULT;
507 		if (copy_from_user(&bind, uarg, sizeof(bind)))
508 			break;
509 
510 		alloc_unbound.dom        = DOMID_SELF;
511 		alloc_unbound.remote_dom = bind.remote_domain;
512 		rc = HYPERVISOR_event_channel_op(EVTCHNOP_alloc_unbound,
513 						 &alloc_unbound);
514 		if (rc != 0)
515 			break;
516 
517 		rc = evtchn_bind_to_user(u, alloc_unbound.port);
518 		if (rc == 0)
519 			rc = alloc_unbound.port;
520 		break;
521 	}
522 
523 	case IOCTL_EVTCHN_UNBIND: {
524 		struct ioctl_evtchn_unbind unbind;
525 		struct user_evtchn *evtchn;
526 
527 		rc = -EFAULT;
528 		if (copy_from_user(&unbind, uarg, sizeof(unbind)))
529 			break;
530 
531 		rc = -EINVAL;
532 		if (unbind.port >= xen_evtchn_nr_channels())
533 			break;
534 
535 		rc = -ENOTCONN;
536 		evtchn = find_evtchn(u, unbind.port);
537 		if (!evtchn)
538 			break;
539 
540 		disable_irq(irq_from_evtchn(unbind.port));
541 		evtchn_unbind_from_user(u, evtchn);
542 		rc = 0;
543 		break;
544 	}
545 
546 	case IOCTL_EVTCHN_NOTIFY: {
547 		struct ioctl_evtchn_notify notify;
548 		struct user_evtchn *evtchn;
549 
550 		rc = -EFAULT;
551 		if (copy_from_user(&notify, uarg, sizeof(notify)))
552 			break;
553 
554 		rc = -ENOTCONN;
555 		evtchn = find_evtchn(u, notify.port);
556 		if (evtchn) {
557 			notify_remote_via_evtchn(notify.port);
558 			rc = 0;
559 		}
560 		break;
561 	}
562 
563 	case IOCTL_EVTCHN_RESET: {
564 		/* Initialise the ring to empty. Clear errors. */
565 		mutex_lock(&u->ring_cons_mutex);
566 		spin_lock_irq(&u->ring_prod_lock);
567 		u->ring_cons = u->ring_prod = u->ring_overflow = 0;
568 		spin_unlock_irq(&u->ring_prod_lock);
569 		mutex_unlock(&u->ring_cons_mutex);
570 		rc = 0;
571 		break;
572 	}
573 
574 	case IOCTL_EVTCHN_RESTRICT_DOMID: {
575 		struct ioctl_evtchn_restrict_domid ierd;
576 
577 		rc = -EACCES;
578 		if (u->restrict_domid != UNRESTRICTED_DOMID)
579 			break;
580 
581 		rc = -EFAULT;
582 		if (copy_from_user(&ierd, uarg, sizeof(ierd)))
583 		    break;
584 
585 		rc = -EINVAL;
586 		if (ierd.domid == 0 || ierd.domid >= DOMID_FIRST_RESERVED)
587 			break;
588 
589 		u->restrict_domid = ierd.domid;
590 		rc = 0;
591 
592 		break;
593 	}
594 
595 	default:
596 		rc = -ENOSYS;
597 		break;
598 	}
599 	mutex_unlock(&u->bind_mutex);
600 
601 	return rc;
602 }
603 
604 static unsigned int evtchn_poll(struct file *file, poll_table *wait)
605 {
606 	unsigned int mask = POLLOUT | POLLWRNORM;
607 	struct per_user_data *u = file->private_data;
608 
609 	poll_wait(file, &u->evtchn_wait, wait);
610 	if (u->ring_cons != u->ring_prod)
611 		mask |= POLLIN | POLLRDNORM;
612 	if (u->ring_overflow)
613 		mask = POLLERR;
614 	return mask;
615 }
616 
617 static int evtchn_fasync(int fd, struct file *filp, int on)
618 {
619 	struct per_user_data *u = filp->private_data;
620 	return fasync_helper(fd, filp, on, &u->evtchn_async_queue);
621 }
622 
623 static int evtchn_open(struct inode *inode, struct file *filp)
624 {
625 	struct per_user_data *u;
626 
627 	u = kzalloc(sizeof(*u), GFP_KERNEL);
628 	if (u == NULL)
629 		return -ENOMEM;
630 
631 	u->name = kasprintf(GFP_KERNEL, "evtchn:%s", current->comm);
632 	if (u->name == NULL) {
633 		kfree(u);
634 		return -ENOMEM;
635 	}
636 
637 	init_waitqueue_head(&u->evtchn_wait);
638 
639 	mutex_init(&u->bind_mutex);
640 	mutex_init(&u->ring_cons_mutex);
641 	spin_lock_init(&u->ring_prod_lock);
642 
643 	u->restrict_domid = UNRESTRICTED_DOMID;
644 
645 	filp->private_data = u;
646 
647 	return nonseekable_open(inode, filp);
648 }
649 
650 static int evtchn_release(struct inode *inode, struct file *filp)
651 {
652 	struct per_user_data *u = filp->private_data;
653 	struct rb_node *node;
654 
655 	while ((node = u->evtchns.rb_node)) {
656 		struct user_evtchn *evtchn;
657 
658 		evtchn = rb_entry(node, struct user_evtchn, node);
659 		disable_irq(irq_from_evtchn(evtchn->port));
660 		evtchn_unbind_from_user(u, evtchn);
661 	}
662 
663 	evtchn_free_ring(u->ring);
664 	kfree(u->name);
665 	kfree(u);
666 
667 	return 0;
668 }
669 
670 static const struct file_operations evtchn_fops = {
671 	.owner   = THIS_MODULE,
672 	.read    = evtchn_read,
673 	.write   = evtchn_write,
674 	.unlocked_ioctl = evtchn_ioctl,
675 	.poll    = evtchn_poll,
676 	.fasync  = evtchn_fasync,
677 	.open    = evtchn_open,
678 	.release = evtchn_release,
679 	.llseek	 = no_llseek,
680 };
681 
682 static struct miscdevice evtchn_miscdev = {
683 	.minor        = MISC_DYNAMIC_MINOR,
684 	.name         = "xen/evtchn",
685 	.fops         = &evtchn_fops,
686 };
687 static int __init evtchn_init(void)
688 {
689 	int err;
690 
691 	if (!xen_domain())
692 		return -ENODEV;
693 
694 	/* Create '/dev/xen/evtchn'. */
695 	err = misc_register(&evtchn_miscdev);
696 	if (err != 0) {
697 		pr_err("Could not register /dev/xen/evtchn\n");
698 		return err;
699 	}
700 
701 	pr_info("Event-channel device installed\n");
702 
703 	return 0;
704 }
705 
706 static void __exit evtchn_cleanup(void)
707 {
708 	misc_deregister(&evtchn_miscdev);
709 }
710 
711 module_init(evtchn_init);
712 module_exit(evtchn_cleanup);
713 
714 MODULE_LICENSE("GPL");
715