xref: /openbmc/linux/drivers/rpmsg/rpmsg_char.c (revision 046b212a)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (C) 2022, STMicroelectronics
4  * Copyright (c) 2016, Linaro Ltd.
5  * Copyright (c) 2012, Michal Simek <monstr@monstr.eu>
6  * Copyright (c) 2012, PetaLogix
7  * Copyright (c) 2011, Texas Instruments, Inc.
8  * Copyright (c) 2011, Google, Inc.
9  *
10  * Based on rpmsg performance statistics driver by Michal Simek, which in turn
11  * was based on TI & Google OMX rpmsg driver.
12  */
13 
14 #define pr_fmt(fmt)	KBUILD_MODNAME ": " fmt
15 
16 #include <linux/cdev.h>
17 #include <linux/device.h>
18 #include <linux/fs.h>
19 #include <linux/idr.h>
20 #include <linux/kernel.h>
21 #include <linux/module.h>
22 #include <linux/poll.h>
23 #include <linux/rpmsg.h>
24 #include <linux/skbuff.h>
25 #include <linux/slab.h>
26 #include <linux/uaccess.h>
27 #include <uapi/linux/rpmsg.h>
28 
29 #include "rpmsg_char.h"
30 #include "rpmsg_internal.h"
31 
32 #define RPMSG_DEV_MAX	(MINORMASK + 1)
33 
34 static dev_t rpmsg_major;
35 
36 static DEFINE_IDA(rpmsg_ept_ida);
37 static DEFINE_IDA(rpmsg_minor_ida);
38 
39 #define dev_to_eptdev(dev) container_of(dev, struct rpmsg_eptdev, dev)
40 #define cdev_to_eptdev(i_cdev) container_of(i_cdev, struct rpmsg_eptdev, cdev)
41 
42 /**
43  * struct rpmsg_eptdev - endpoint device context
44  * @dev:	endpoint device
45  * @cdev:	cdev for the endpoint device
46  * @rpdev:	underlaying rpmsg device
47  * @chinfo:	info used to open the endpoint
48  * @ept_lock:	synchronization of @ept modifications
49  * @ept:	rpmsg endpoint reference, when open
50  * @queue_lock:	synchronization of @queue operations
51  * @queue:	incoming message queue
52  * @readq:	wait object for incoming queue
53  * @default_ept: set to channel default endpoint if the default endpoint should be re-used
54  *              on device open to prevent endpoint address update.
55  * remote_flow_restricted: to indicate if the remote has requested for flow to be limited
56  * remote_flow_updated: to indicate if the flow control has been requested
57  */
58 struct rpmsg_eptdev {
59 	struct device dev;
60 	struct cdev cdev;
61 
62 	struct rpmsg_device *rpdev;
63 	struct rpmsg_channel_info chinfo;
64 
65 	struct mutex ept_lock;
66 	struct rpmsg_endpoint *ept;
67 	struct rpmsg_endpoint *default_ept;
68 
69 	spinlock_t queue_lock;
70 	struct sk_buff_head queue;
71 	wait_queue_head_t readq;
72 
73 	bool remote_flow_restricted;
74 	bool remote_flow_updated;
75 };
76 
77 int rpmsg_chrdev_eptdev_destroy(struct device *dev, void *data)
78 {
79 	struct rpmsg_eptdev *eptdev = dev_to_eptdev(dev);
80 
81 	mutex_lock(&eptdev->ept_lock);
82 	eptdev->rpdev = NULL;
83 	if (eptdev->ept) {
84 		/* The default endpoint is released by the rpmsg core */
85 		if (!eptdev->default_ept)
86 			rpmsg_destroy_ept(eptdev->ept);
87 		eptdev->ept = NULL;
88 	}
89 	mutex_unlock(&eptdev->ept_lock);
90 
91 	/* wake up any blocked readers */
92 	wake_up_interruptible(&eptdev->readq);
93 
94 	cdev_device_del(&eptdev->cdev, &eptdev->dev);
95 	put_device(&eptdev->dev);
96 
97 	return 0;
98 }
99 EXPORT_SYMBOL(rpmsg_chrdev_eptdev_destroy);
100 
101 static int rpmsg_ept_cb(struct rpmsg_device *rpdev, void *buf, int len,
102 			void *priv, u32 addr)
103 {
104 	struct rpmsg_eptdev *eptdev = priv;
105 	struct sk_buff *skb;
106 
107 	skb = alloc_skb(len, GFP_ATOMIC);
108 	if (!skb)
109 		return -ENOMEM;
110 
111 	skb_put_data(skb, buf, len);
112 
113 	spin_lock(&eptdev->queue_lock);
114 	skb_queue_tail(&eptdev->queue, skb);
115 	spin_unlock(&eptdev->queue_lock);
116 
117 	/* wake up any blocking processes, waiting for new data */
118 	wake_up_interruptible(&eptdev->readq);
119 
120 	return 0;
121 }
122 
123 static int rpmsg_ept_flow_cb(struct rpmsg_device *rpdev, void *priv, bool enable)
124 {
125 	struct rpmsg_eptdev *eptdev = priv;
126 
127 	eptdev->remote_flow_restricted = enable;
128 	eptdev->remote_flow_updated = true;
129 
130 	wake_up_interruptible(&eptdev->readq);
131 
132 	return 0;
133 }
134 
135 static int rpmsg_eptdev_open(struct inode *inode, struct file *filp)
136 {
137 	struct rpmsg_eptdev *eptdev = cdev_to_eptdev(inode->i_cdev);
138 	struct rpmsg_endpoint *ept;
139 	struct rpmsg_device *rpdev = eptdev->rpdev;
140 	struct device *dev = &eptdev->dev;
141 
142 	mutex_lock(&eptdev->ept_lock);
143 	if (eptdev->ept) {
144 		mutex_unlock(&eptdev->ept_lock);
145 		return -EBUSY;
146 	}
147 
148 	if (!eptdev->rpdev) {
149 		mutex_unlock(&eptdev->ept_lock);
150 		return -ENETRESET;
151 	}
152 
153 	get_device(dev);
154 
155 	/*
156 	 * If the default_ept is set, the rpmsg device default endpoint is used.
157 	 * Else a new endpoint is created on open that will be destroyed on release.
158 	 */
159 	if (eptdev->default_ept)
160 		ept = eptdev->default_ept;
161 	else
162 		ept = rpmsg_create_ept(rpdev, rpmsg_ept_cb, eptdev, eptdev->chinfo);
163 
164 	if (!ept) {
165 		dev_err(dev, "failed to open %s\n", eptdev->chinfo.name);
166 		put_device(dev);
167 		mutex_unlock(&eptdev->ept_lock);
168 		return -EINVAL;
169 	}
170 
171 	ept->flow_cb = rpmsg_ept_flow_cb;
172 	eptdev->ept = ept;
173 	filp->private_data = eptdev;
174 	mutex_unlock(&eptdev->ept_lock);
175 
176 	return 0;
177 }
178 
179 static int rpmsg_eptdev_release(struct inode *inode, struct file *filp)
180 {
181 	struct rpmsg_eptdev *eptdev = cdev_to_eptdev(inode->i_cdev);
182 	struct device *dev = &eptdev->dev;
183 
184 	/* Close the endpoint, if it's not already destroyed by the parent */
185 	mutex_lock(&eptdev->ept_lock);
186 	if (eptdev->ept) {
187 		if (!eptdev->default_ept)
188 			rpmsg_destroy_ept(eptdev->ept);
189 		eptdev->ept = NULL;
190 	}
191 	mutex_unlock(&eptdev->ept_lock);
192 	eptdev->remote_flow_updated = false;
193 
194 	/* Discard all SKBs */
195 	skb_queue_purge(&eptdev->queue);
196 
197 	put_device(dev);
198 
199 	return 0;
200 }
201 
202 static ssize_t rpmsg_eptdev_read_iter(struct kiocb *iocb, struct iov_iter *to)
203 {
204 	struct file *filp = iocb->ki_filp;
205 	struct rpmsg_eptdev *eptdev = filp->private_data;
206 	unsigned long flags;
207 	struct sk_buff *skb;
208 	int use;
209 
210 	if (!eptdev->ept)
211 		return -EPIPE;
212 
213 	spin_lock_irqsave(&eptdev->queue_lock, flags);
214 
215 	/* Wait for data in the queue */
216 	if (skb_queue_empty(&eptdev->queue)) {
217 		spin_unlock_irqrestore(&eptdev->queue_lock, flags);
218 
219 		if (filp->f_flags & O_NONBLOCK)
220 			return -EAGAIN;
221 
222 		/* Wait until we get data or the endpoint goes away */
223 		if (wait_event_interruptible(eptdev->readq,
224 					     !skb_queue_empty(&eptdev->queue) ||
225 					     !eptdev->ept))
226 			return -ERESTARTSYS;
227 
228 		/* We lost the endpoint while waiting */
229 		if (!eptdev->ept)
230 			return -EPIPE;
231 
232 		spin_lock_irqsave(&eptdev->queue_lock, flags);
233 	}
234 
235 	skb = skb_dequeue(&eptdev->queue);
236 	spin_unlock_irqrestore(&eptdev->queue_lock, flags);
237 	if (!skb)
238 		return -EFAULT;
239 
240 	use = min_t(size_t, iov_iter_count(to), skb->len);
241 	if (copy_to_iter(skb->data, use, to) != use)
242 		use = -EFAULT;
243 
244 	kfree_skb(skb);
245 
246 	return use;
247 }
248 
249 static ssize_t rpmsg_eptdev_write_iter(struct kiocb *iocb,
250 				       struct iov_iter *from)
251 {
252 	struct file *filp = iocb->ki_filp;
253 	struct rpmsg_eptdev *eptdev = filp->private_data;
254 	size_t len = iov_iter_count(from);
255 	void *kbuf;
256 	int ret;
257 
258 	kbuf = kzalloc(len, GFP_KERNEL);
259 	if (!kbuf)
260 		return -ENOMEM;
261 
262 	if (!copy_from_iter_full(kbuf, len, from)) {
263 		ret = -EFAULT;
264 		goto free_kbuf;
265 	}
266 
267 	if (mutex_lock_interruptible(&eptdev->ept_lock)) {
268 		ret = -ERESTARTSYS;
269 		goto free_kbuf;
270 	}
271 
272 	if (!eptdev->ept) {
273 		ret = -EPIPE;
274 		goto unlock_eptdev;
275 	}
276 
277 	if (filp->f_flags & O_NONBLOCK) {
278 		ret = rpmsg_trysendto(eptdev->ept, kbuf, len, eptdev->chinfo.dst);
279 		if (ret == -ENOMEM)
280 			ret = -EAGAIN;
281 	} else {
282 		ret = rpmsg_sendto(eptdev->ept, kbuf, len, eptdev->chinfo.dst);
283 	}
284 
285 unlock_eptdev:
286 	mutex_unlock(&eptdev->ept_lock);
287 
288 free_kbuf:
289 	kfree(kbuf);
290 	return ret < 0 ? ret : len;
291 }
292 
293 static __poll_t rpmsg_eptdev_poll(struct file *filp, poll_table *wait)
294 {
295 	struct rpmsg_eptdev *eptdev = filp->private_data;
296 	__poll_t mask = 0;
297 
298 	if (!eptdev->ept)
299 		return EPOLLERR;
300 
301 	poll_wait(filp, &eptdev->readq, wait);
302 
303 	if (!skb_queue_empty(&eptdev->queue))
304 		mask |= EPOLLIN | EPOLLRDNORM;
305 
306 	if (eptdev->remote_flow_updated)
307 		mask |= EPOLLPRI;
308 
309 	mutex_lock(&eptdev->ept_lock);
310 	mask |= rpmsg_poll(eptdev->ept, filp, wait);
311 	mutex_unlock(&eptdev->ept_lock);
312 
313 	return mask;
314 }
315 
316 static long rpmsg_eptdev_ioctl(struct file *fp, unsigned int cmd,
317 			       unsigned long arg)
318 {
319 	struct rpmsg_eptdev *eptdev = fp->private_data;
320 
321 	bool set;
322 	int ret;
323 
324 	switch (cmd) {
325 	case RPMSG_GET_OUTGOING_FLOWCONTROL:
326 		eptdev->remote_flow_updated = false;
327 		ret = put_user(eptdev->remote_flow_restricted, (int __user *)arg);
328 		break;
329 	case RPMSG_SET_INCOMING_FLOWCONTROL:
330 		if (arg > 1) {
331 			ret = -EINVAL;
332 			break;
333 		}
334 		set = !!arg;
335 		ret = rpmsg_set_flow_control(eptdev->ept, set, eptdev->chinfo.dst);
336 		break;
337 	case RPMSG_DESTROY_EPT_IOCTL:
338 		/* Don't allow to destroy a default endpoint. */
339 		if (eptdev->default_ept) {
340 			ret = -EINVAL;
341 			break;
342 		}
343 		ret = rpmsg_chrdev_eptdev_destroy(&eptdev->dev, NULL);
344 		break;
345 	default:
346 		ret = -EINVAL;
347 	}
348 
349 	return ret;
350 }
351 
352 static const struct file_operations rpmsg_eptdev_fops = {
353 	.owner = THIS_MODULE,
354 	.open = rpmsg_eptdev_open,
355 	.release = rpmsg_eptdev_release,
356 	.read_iter = rpmsg_eptdev_read_iter,
357 	.write_iter = rpmsg_eptdev_write_iter,
358 	.poll = rpmsg_eptdev_poll,
359 	.unlocked_ioctl = rpmsg_eptdev_ioctl,
360 	.compat_ioctl = compat_ptr_ioctl,
361 };
362 
363 static ssize_t name_show(struct device *dev, struct device_attribute *attr,
364 			 char *buf)
365 {
366 	struct rpmsg_eptdev *eptdev = dev_get_drvdata(dev);
367 
368 	return sprintf(buf, "%s\n", eptdev->chinfo.name);
369 }
370 static DEVICE_ATTR_RO(name);
371 
372 static ssize_t src_show(struct device *dev, struct device_attribute *attr,
373 			 char *buf)
374 {
375 	struct rpmsg_eptdev *eptdev = dev_get_drvdata(dev);
376 
377 	return sprintf(buf, "%d\n", eptdev->chinfo.src);
378 }
379 static DEVICE_ATTR_RO(src);
380 
381 static ssize_t dst_show(struct device *dev, struct device_attribute *attr,
382 			 char *buf)
383 {
384 	struct rpmsg_eptdev *eptdev = dev_get_drvdata(dev);
385 
386 	return sprintf(buf, "%d\n", eptdev->chinfo.dst);
387 }
388 static DEVICE_ATTR_RO(dst);
389 
390 static struct attribute *rpmsg_eptdev_attrs[] = {
391 	&dev_attr_name.attr,
392 	&dev_attr_src.attr,
393 	&dev_attr_dst.attr,
394 	NULL
395 };
396 ATTRIBUTE_GROUPS(rpmsg_eptdev);
397 
398 static void rpmsg_eptdev_release_device(struct device *dev)
399 {
400 	struct rpmsg_eptdev *eptdev = dev_to_eptdev(dev);
401 
402 	ida_simple_remove(&rpmsg_ept_ida, dev->id);
403 	ida_simple_remove(&rpmsg_minor_ida, MINOR(eptdev->dev.devt));
404 	kfree(eptdev);
405 }
406 
407 static struct rpmsg_eptdev *rpmsg_chrdev_eptdev_alloc(struct rpmsg_device *rpdev,
408 						      struct device *parent)
409 {
410 	struct rpmsg_eptdev *eptdev;
411 	struct device *dev;
412 
413 	eptdev = kzalloc(sizeof(*eptdev), GFP_KERNEL);
414 	if (!eptdev)
415 		return ERR_PTR(-ENOMEM);
416 
417 	dev = &eptdev->dev;
418 	eptdev->rpdev = rpdev;
419 
420 	mutex_init(&eptdev->ept_lock);
421 	spin_lock_init(&eptdev->queue_lock);
422 	skb_queue_head_init(&eptdev->queue);
423 	init_waitqueue_head(&eptdev->readq);
424 
425 	device_initialize(dev);
426 	dev->class = rpmsg_class;
427 	dev->parent = parent;
428 	dev->groups = rpmsg_eptdev_groups;
429 	dev_set_drvdata(dev, eptdev);
430 
431 	cdev_init(&eptdev->cdev, &rpmsg_eptdev_fops);
432 	eptdev->cdev.owner = THIS_MODULE;
433 
434 	return eptdev;
435 }
436 
437 static int rpmsg_chrdev_eptdev_add(struct rpmsg_eptdev *eptdev, struct rpmsg_channel_info chinfo)
438 {
439 	struct device *dev = &eptdev->dev;
440 	int ret;
441 
442 	eptdev->chinfo = chinfo;
443 
444 	ret = ida_simple_get(&rpmsg_minor_ida, 0, RPMSG_DEV_MAX, GFP_KERNEL);
445 	if (ret < 0)
446 		goto free_eptdev;
447 	dev->devt = MKDEV(MAJOR(rpmsg_major), ret);
448 
449 	ret = ida_simple_get(&rpmsg_ept_ida, 0, 0, GFP_KERNEL);
450 	if (ret < 0)
451 		goto free_minor_ida;
452 	dev->id = ret;
453 	dev_set_name(dev, "rpmsg%d", ret);
454 
455 	ret = cdev_device_add(&eptdev->cdev, &eptdev->dev);
456 	if (ret)
457 		goto free_ept_ida;
458 
459 	/* We can now rely on the release function for cleanup */
460 	dev->release = rpmsg_eptdev_release_device;
461 
462 	return ret;
463 
464 free_ept_ida:
465 	ida_simple_remove(&rpmsg_ept_ida, dev->id);
466 free_minor_ida:
467 	ida_simple_remove(&rpmsg_minor_ida, MINOR(dev->devt));
468 free_eptdev:
469 	put_device(dev);
470 	kfree(eptdev);
471 
472 	return ret;
473 }
474 
475 int rpmsg_chrdev_eptdev_create(struct rpmsg_device *rpdev, struct device *parent,
476 			       struct rpmsg_channel_info chinfo)
477 {
478 	struct rpmsg_eptdev *eptdev;
479 
480 	eptdev = rpmsg_chrdev_eptdev_alloc(rpdev, parent);
481 	if (IS_ERR(eptdev))
482 		return PTR_ERR(eptdev);
483 
484 	return rpmsg_chrdev_eptdev_add(eptdev, chinfo);
485 }
486 EXPORT_SYMBOL(rpmsg_chrdev_eptdev_create);
487 
488 static int rpmsg_chrdev_probe(struct rpmsg_device *rpdev)
489 {
490 	struct rpmsg_channel_info chinfo;
491 	struct rpmsg_eptdev *eptdev;
492 	struct device *dev = &rpdev->dev;
493 
494 	memcpy(chinfo.name, rpdev->id.name, RPMSG_NAME_SIZE);
495 	chinfo.src = rpdev->src;
496 	chinfo.dst = rpdev->dst;
497 
498 	eptdev = rpmsg_chrdev_eptdev_alloc(rpdev, dev);
499 	if (IS_ERR(eptdev))
500 		return PTR_ERR(eptdev);
501 
502 	/* Set the default_ept to the rpmsg device endpoint */
503 	eptdev->default_ept = rpdev->ept;
504 
505 	/*
506 	 * The rpmsg_ept_cb uses *priv parameter to get its rpmsg_eptdev context.
507 	 * Storedit in default_ept *priv field.
508 	 */
509 	eptdev->default_ept->priv = eptdev;
510 
511 	return rpmsg_chrdev_eptdev_add(eptdev, chinfo);
512 }
513 
514 static void rpmsg_chrdev_remove(struct rpmsg_device *rpdev)
515 {
516 	int ret;
517 
518 	ret = device_for_each_child(&rpdev->dev, NULL, rpmsg_chrdev_eptdev_destroy);
519 	if (ret)
520 		dev_warn(&rpdev->dev, "failed to destroy endpoints: %d\n", ret);
521 }
522 
523 static struct rpmsg_device_id rpmsg_chrdev_id_table[] = {
524 	{ .name	= "rpmsg-raw" },
525 	{ },
526 };
527 
528 static struct rpmsg_driver rpmsg_chrdev_driver = {
529 	.probe = rpmsg_chrdev_probe,
530 	.remove = rpmsg_chrdev_remove,
531 	.callback = rpmsg_ept_cb,
532 	.id_table = rpmsg_chrdev_id_table,
533 	.drv.name = "rpmsg_chrdev",
534 };
535 
536 static int rpmsg_chrdev_init(void)
537 {
538 	int ret;
539 
540 	ret = alloc_chrdev_region(&rpmsg_major, 0, RPMSG_DEV_MAX, "rpmsg_char");
541 	if (ret < 0) {
542 		pr_err("failed to allocate char dev region\n");
543 		return ret;
544 	}
545 
546 	ret = register_rpmsg_driver(&rpmsg_chrdev_driver);
547 	if (ret < 0) {
548 		pr_err("rpmsg: failed to register rpmsg raw driver\n");
549 		goto free_region;
550 	}
551 
552 	return 0;
553 
554 free_region:
555 	unregister_chrdev_region(rpmsg_major, RPMSG_DEV_MAX);
556 
557 	return ret;
558 }
559 postcore_initcall(rpmsg_chrdev_init);
560 
561 static void rpmsg_chrdev_exit(void)
562 {
563 	unregister_rpmsg_driver(&rpmsg_chrdev_driver);
564 	unregister_chrdev_region(rpmsg_major, RPMSG_DEV_MAX);
565 }
566 module_exit(rpmsg_chrdev_exit);
567 
568 MODULE_ALIAS("rpmsg:rpmsg_chrdev");
569 MODULE_LICENSE("GPL v2");
570