xref: /linux/drivers/rpmsg/rpmsg_char.c (revision a93fbb002310ef04fce504dbf1510f6eb8265188)
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  */
56 struct rpmsg_eptdev {
57 	struct device dev;
58 	struct cdev cdev;
59 
60 	struct rpmsg_device *rpdev;
61 	struct rpmsg_channel_info chinfo;
62 
63 	struct mutex ept_lock;
64 	struct rpmsg_endpoint *ept;
65 	struct rpmsg_endpoint *default_ept;
66 
67 	spinlock_t queue_lock;
68 	struct sk_buff_head queue;
69 	wait_queue_head_t readq;
70 
71 };
72 
73 int rpmsg_chrdev_eptdev_destroy(struct device *dev, void *data)
74 {
75 	struct rpmsg_eptdev *eptdev = dev_to_eptdev(dev);
76 
77 	mutex_lock(&eptdev->ept_lock);
78 	if (eptdev->ept) {
79 		rpmsg_destroy_ept(eptdev->ept);
80 		eptdev->ept = NULL;
81 	}
82 	mutex_unlock(&eptdev->ept_lock);
83 
84 	/* wake up any blocked readers */
85 	wake_up_interruptible(&eptdev->readq);
86 
87 	cdev_device_del(&eptdev->cdev, &eptdev->dev);
88 	put_device(&eptdev->dev);
89 
90 	return 0;
91 }
92 EXPORT_SYMBOL(rpmsg_chrdev_eptdev_destroy);
93 
94 static int rpmsg_ept_cb(struct rpmsg_device *rpdev, void *buf, int len,
95 			void *priv, u32 addr)
96 {
97 	struct rpmsg_eptdev *eptdev = priv;
98 	struct sk_buff *skb;
99 
100 	skb = alloc_skb(len, GFP_ATOMIC);
101 	if (!skb)
102 		return -ENOMEM;
103 
104 	skb_put_data(skb, buf, len);
105 
106 	spin_lock(&eptdev->queue_lock);
107 	skb_queue_tail(&eptdev->queue, skb);
108 	spin_unlock(&eptdev->queue_lock);
109 
110 	/* wake up any blocking processes, waiting for new data */
111 	wake_up_interruptible(&eptdev->readq);
112 
113 	return 0;
114 }
115 
116 static int rpmsg_eptdev_open(struct inode *inode, struct file *filp)
117 {
118 	struct rpmsg_eptdev *eptdev = cdev_to_eptdev(inode->i_cdev);
119 	struct rpmsg_endpoint *ept;
120 	struct rpmsg_device *rpdev = eptdev->rpdev;
121 	struct device *dev = &eptdev->dev;
122 
123 	if (eptdev->ept)
124 		return -EBUSY;
125 
126 	get_device(dev);
127 
128 	/*
129 	 * If the default_ept is set, the rpmsg device default endpoint is used.
130 	 * Else a new endpoint is created on open that will be destroyed on release.
131 	 */
132 	if (eptdev->default_ept)
133 		ept = eptdev->default_ept;
134 	else
135 		ept = rpmsg_create_ept(rpdev, rpmsg_ept_cb, eptdev, eptdev->chinfo);
136 
137 	if (!ept) {
138 		dev_err(dev, "failed to open %s\n", eptdev->chinfo.name);
139 		put_device(dev);
140 		return -EINVAL;
141 	}
142 
143 	eptdev->ept = ept;
144 	filp->private_data = eptdev;
145 
146 	return 0;
147 }
148 
149 static int rpmsg_eptdev_release(struct inode *inode, struct file *filp)
150 {
151 	struct rpmsg_eptdev *eptdev = cdev_to_eptdev(inode->i_cdev);
152 	struct device *dev = &eptdev->dev;
153 
154 	/* Close the endpoint, if it's not already destroyed by the parent */
155 	mutex_lock(&eptdev->ept_lock);
156 	if (eptdev->ept) {
157 		if (!eptdev->default_ept)
158 			rpmsg_destroy_ept(eptdev->ept);
159 		eptdev->ept = NULL;
160 	}
161 	mutex_unlock(&eptdev->ept_lock);
162 
163 	/* Discard all SKBs */
164 	skb_queue_purge(&eptdev->queue);
165 
166 	put_device(dev);
167 
168 	return 0;
169 }
170 
171 static ssize_t rpmsg_eptdev_read_iter(struct kiocb *iocb, struct iov_iter *to)
172 {
173 	struct file *filp = iocb->ki_filp;
174 	struct rpmsg_eptdev *eptdev = filp->private_data;
175 	unsigned long flags;
176 	struct sk_buff *skb;
177 	int use;
178 
179 	if (!eptdev->ept)
180 		return -EPIPE;
181 
182 	spin_lock_irqsave(&eptdev->queue_lock, flags);
183 
184 	/* Wait for data in the queue */
185 	if (skb_queue_empty(&eptdev->queue)) {
186 		spin_unlock_irqrestore(&eptdev->queue_lock, flags);
187 
188 		if (filp->f_flags & O_NONBLOCK)
189 			return -EAGAIN;
190 
191 		/* Wait until we get data or the endpoint goes away */
192 		if (wait_event_interruptible(eptdev->readq,
193 					     !skb_queue_empty(&eptdev->queue) ||
194 					     !eptdev->ept))
195 			return -ERESTARTSYS;
196 
197 		/* We lost the endpoint while waiting */
198 		if (!eptdev->ept)
199 			return -EPIPE;
200 
201 		spin_lock_irqsave(&eptdev->queue_lock, flags);
202 	}
203 
204 	skb = skb_dequeue(&eptdev->queue);
205 	spin_unlock_irqrestore(&eptdev->queue_lock, flags);
206 	if (!skb)
207 		return -EFAULT;
208 
209 	use = min_t(size_t, iov_iter_count(to), skb->len);
210 	if (copy_to_iter(skb->data, use, to) != use)
211 		use = -EFAULT;
212 
213 	kfree_skb(skb);
214 
215 	return use;
216 }
217 
218 static ssize_t rpmsg_eptdev_write_iter(struct kiocb *iocb,
219 				       struct iov_iter *from)
220 {
221 	struct file *filp = iocb->ki_filp;
222 	struct rpmsg_eptdev *eptdev = filp->private_data;
223 	size_t len = iov_iter_count(from);
224 	void *kbuf;
225 	int ret;
226 
227 	kbuf = kzalloc(len, GFP_KERNEL);
228 	if (!kbuf)
229 		return -ENOMEM;
230 
231 	if (!copy_from_iter_full(kbuf, len, from)) {
232 		ret = -EFAULT;
233 		goto free_kbuf;
234 	}
235 
236 	if (mutex_lock_interruptible(&eptdev->ept_lock)) {
237 		ret = -ERESTARTSYS;
238 		goto free_kbuf;
239 	}
240 
241 	if (!eptdev->ept) {
242 		ret = -EPIPE;
243 		goto unlock_eptdev;
244 	}
245 
246 	if (filp->f_flags & O_NONBLOCK) {
247 		ret = rpmsg_trysendto(eptdev->ept, kbuf, len, eptdev->chinfo.dst);
248 		if (ret == -ENOMEM)
249 			ret = -EAGAIN;
250 	} else {
251 		ret = rpmsg_sendto(eptdev->ept, kbuf, len, eptdev->chinfo.dst);
252 	}
253 
254 unlock_eptdev:
255 	mutex_unlock(&eptdev->ept_lock);
256 
257 free_kbuf:
258 	kfree(kbuf);
259 	return ret < 0 ? ret : len;
260 }
261 
262 static __poll_t rpmsg_eptdev_poll(struct file *filp, poll_table *wait)
263 {
264 	struct rpmsg_eptdev *eptdev = filp->private_data;
265 	__poll_t mask = 0;
266 
267 	if (!eptdev->ept)
268 		return EPOLLERR;
269 
270 	poll_wait(filp, &eptdev->readq, wait);
271 
272 	if (!skb_queue_empty(&eptdev->queue))
273 		mask |= EPOLLIN | EPOLLRDNORM;
274 
275 	mask |= rpmsg_poll(eptdev->ept, filp, wait);
276 
277 	return mask;
278 }
279 
280 static long rpmsg_eptdev_ioctl(struct file *fp, unsigned int cmd,
281 			       unsigned long arg)
282 {
283 	struct rpmsg_eptdev *eptdev = fp->private_data;
284 
285 	if (cmd != RPMSG_DESTROY_EPT_IOCTL)
286 		return -EINVAL;
287 
288 	/* Don't allow to destroy a default endpoint. */
289 	if (eptdev->default_ept)
290 		return -EINVAL;
291 
292 	return rpmsg_chrdev_eptdev_destroy(&eptdev->dev, NULL);
293 }
294 
295 static const struct file_operations rpmsg_eptdev_fops = {
296 	.owner = THIS_MODULE,
297 	.open = rpmsg_eptdev_open,
298 	.release = rpmsg_eptdev_release,
299 	.read_iter = rpmsg_eptdev_read_iter,
300 	.write_iter = rpmsg_eptdev_write_iter,
301 	.poll = rpmsg_eptdev_poll,
302 	.unlocked_ioctl = rpmsg_eptdev_ioctl,
303 	.compat_ioctl = compat_ptr_ioctl,
304 };
305 
306 static ssize_t name_show(struct device *dev, struct device_attribute *attr,
307 			 char *buf)
308 {
309 	struct rpmsg_eptdev *eptdev = dev_get_drvdata(dev);
310 
311 	return sprintf(buf, "%s\n", eptdev->chinfo.name);
312 }
313 static DEVICE_ATTR_RO(name);
314 
315 static ssize_t src_show(struct device *dev, struct device_attribute *attr,
316 			 char *buf)
317 {
318 	struct rpmsg_eptdev *eptdev = dev_get_drvdata(dev);
319 
320 	return sprintf(buf, "%d\n", eptdev->chinfo.src);
321 }
322 static DEVICE_ATTR_RO(src);
323 
324 static ssize_t dst_show(struct device *dev, struct device_attribute *attr,
325 			 char *buf)
326 {
327 	struct rpmsg_eptdev *eptdev = dev_get_drvdata(dev);
328 
329 	return sprintf(buf, "%d\n", eptdev->chinfo.dst);
330 }
331 static DEVICE_ATTR_RO(dst);
332 
333 static struct attribute *rpmsg_eptdev_attrs[] = {
334 	&dev_attr_name.attr,
335 	&dev_attr_src.attr,
336 	&dev_attr_dst.attr,
337 	NULL
338 };
339 ATTRIBUTE_GROUPS(rpmsg_eptdev);
340 
341 static void rpmsg_eptdev_release_device(struct device *dev)
342 {
343 	struct rpmsg_eptdev *eptdev = dev_to_eptdev(dev);
344 
345 	ida_simple_remove(&rpmsg_ept_ida, dev->id);
346 	ida_simple_remove(&rpmsg_minor_ida, MINOR(eptdev->dev.devt));
347 	kfree(eptdev);
348 }
349 
350 static struct rpmsg_eptdev *rpmsg_chrdev_eptdev_alloc(struct rpmsg_device *rpdev,
351 						      struct device *parent)
352 {
353 	struct rpmsg_eptdev *eptdev;
354 	struct device *dev;
355 
356 	eptdev = kzalloc(sizeof(*eptdev), GFP_KERNEL);
357 	if (!eptdev)
358 		return ERR_PTR(-ENOMEM);
359 
360 	dev = &eptdev->dev;
361 	eptdev->rpdev = rpdev;
362 
363 	mutex_init(&eptdev->ept_lock);
364 	spin_lock_init(&eptdev->queue_lock);
365 	skb_queue_head_init(&eptdev->queue);
366 	init_waitqueue_head(&eptdev->readq);
367 
368 	device_initialize(dev);
369 	dev->class = rpmsg_class;
370 	dev->parent = parent;
371 	dev->groups = rpmsg_eptdev_groups;
372 	dev_set_drvdata(dev, eptdev);
373 
374 	cdev_init(&eptdev->cdev, &rpmsg_eptdev_fops);
375 	eptdev->cdev.owner = THIS_MODULE;
376 
377 	return eptdev;
378 }
379 
380 static int rpmsg_chrdev_eptdev_add(struct rpmsg_eptdev *eptdev, struct rpmsg_channel_info chinfo)
381 {
382 	struct device *dev = &eptdev->dev;
383 	int ret;
384 
385 	eptdev->chinfo = chinfo;
386 
387 	ret = ida_simple_get(&rpmsg_minor_ida, 0, RPMSG_DEV_MAX, GFP_KERNEL);
388 	if (ret < 0)
389 		goto free_eptdev;
390 	dev->devt = MKDEV(MAJOR(rpmsg_major), ret);
391 
392 	ret = ida_simple_get(&rpmsg_ept_ida, 0, 0, GFP_KERNEL);
393 	if (ret < 0)
394 		goto free_minor_ida;
395 	dev->id = ret;
396 	dev_set_name(dev, "rpmsg%d", ret);
397 
398 	ret = cdev_device_add(&eptdev->cdev, &eptdev->dev);
399 	if (ret)
400 		goto free_ept_ida;
401 
402 	/* We can now rely on the release function for cleanup */
403 	dev->release = rpmsg_eptdev_release_device;
404 
405 	return ret;
406 
407 free_ept_ida:
408 	ida_simple_remove(&rpmsg_ept_ida, dev->id);
409 free_minor_ida:
410 	ida_simple_remove(&rpmsg_minor_ida, MINOR(dev->devt));
411 free_eptdev:
412 	put_device(dev);
413 	kfree(eptdev);
414 
415 	return ret;
416 }
417 
418 int rpmsg_chrdev_eptdev_create(struct rpmsg_device *rpdev, struct device *parent,
419 			       struct rpmsg_channel_info chinfo)
420 {
421 	struct rpmsg_eptdev *eptdev;
422 	int ret;
423 
424 	eptdev = rpmsg_chrdev_eptdev_alloc(rpdev, parent);
425 	if (IS_ERR(eptdev))
426 		return PTR_ERR(eptdev);
427 
428 	ret = rpmsg_chrdev_eptdev_add(eptdev, chinfo);
429 
430 	return ret;
431 }
432 EXPORT_SYMBOL(rpmsg_chrdev_eptdev_create);
433 
434 static int rpmsg_chrdev_probe(struct rpmsg_device *rpdev)
435 {
436 	struct rpmsg_channel_info chinfo;
437 	struct rpmsg_eptdev *eptdev;
438 	struct device *dev = &rpdev->dev;
439 
440 	memcpy(chinfo.name, rpdev->id.name, RPMSG_NAME_SIZE);
441 	chinfo.src = rpdev->src;
442 	chinfo.dst = rpdev->dst;
443 
444 	eptdev = rpmsg_chrdev_eptdev_alloc(rpdev, dev);
445 	if (IS_ERR(eptdev))
446 		return PTR_ERR(eptdev);
447 
448 	/* Set the default_ept to the rpmsg device endpoint */
449 	eptdev->default_ept = rpdev->ept;
450 
451 	/*
452 	 * The rpmsg_ept_cb uses *priv parameter to get its rpmsg_eptdev context.
453 	 * Storedit in default_ept *priv field.
454 	 */
455 	eptdev->default_ept->priv = eptdev;
456 
457 	return rpmsg_chrdev_eptdev_add(eptdev, chinfo);
458 }
459 
460 static void rpmsg_chrdev_remove(struct rpmsg_device *rpdev)
461 {
462 	int ret;
463 
464 	ret = device_for_each_child(&rpdev->dev, NULL, rpmsg_chrdev_eptdev_destroy);
465 	if (ret)
466 		dev_warn(&rpdev->dev, "failed to destroy endpoints: %d\n", ret);
467 }
468 
469 static struct rpmsg_device_id rpmsg_chrdev_id_table[] = {
470 	{ .name	= "rpmsg-raw" },
471 	{ },
472 };
473 
474 static struct rpmsg_driver rpmsg_chrdev_driver = {
475 	.probe = rpmsg_chrdev_probe,
476 	.remove = rpmsg_chrdev_remove,
477 	.callback = rpmsg_ept_cb,
478 	.id_table = rpmsg_chrdev_id_table,
479 	.drv.name = "rpmsg_chrdev",
480 };
481 
482 static int rpmsg_chrdev_init(void)
483 {
484 	int ret;
485 
486 	ret = alloc_chrdev_region(&rpmsg_major, 0, RPMSG_DEV_MAX, "rpmsg_char");
487 	if (ret < 0) {
488 		pr_err("failed to allocate char dev region\n");
489 		return ret;
490 	}
491 
492 	ret = register_rpmsg_driver(&rpmsg_chrdev_driver);
493 	if (ret < 0) {
494 		pr_err("rpmsg: failed to register rpmsg raw driver\n");
495 		goto free_region;
496 	}
497 
498 	return 0;
499 
500 free_region:
501 	unregister_chrdev_region(rpmsg_major, RPMSG_DEV_MAX);
502 
503 	return ret;
504 }
505 postcore_initcall(rpmsg_chrdev_init);
506 
507 static void rpmsg_chrdev_exit(void)
508 {
509 	unregister_rpmsg_driver(&rpmsg_chrdev_driver);
510 	unregister_chrdev_region(rpmsg_major, RPMSG_DEV_MAX);
511 }
512 module_exit(rpmsg_chrdev_exit);
513 
514 MODULE_ALIAS("rpmsg:rpmsg_chrdev");
515 MODULE_LICENSE("GPL v2");
516