xref: /freebsd/sys/compat/linuxkpi/common/src/linux_compat.c (revision 8ef24a0d4b28fe230e20637f56869cc4148cd2ca)
1 /*-
2  * Copyright (c) 2010 Isilon Systems, Inc.
3  * Copyright (c) 2010 iX Systems, Inc.
4  * Copyright (c) 2010 Panasas, Inc.
5  * Copyright (c) 2013-2016 Mellanox Technologies, Ltd.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice unmodified, this list of conditions, and the following
13  *    disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
23  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
27  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/malloc.h>
36 #include <sys/kernel.h>
37 #include <sys/sysctl.h>
38 #include <sys/proc.h>
39 #include <sys/sglist.h>
40 #include <sys/sleepqueue.h>
41 #include <sys/lock.h>
42 #include <sys/mutex.h>
43 #include <sys/bus.h>
44 #include <sys/fcntl.h>
45 #include <sys/file.h>
46 #include <sys/filio.h>
47 #include <sys/rwlock.h>
48 
49 #include <vm/vm.h>
50 #include <vm/pmap.h>
51 
52 #include <machine/stdarg.h>
53 
54 #if defined(__i386__) || defined(__amd64__)
55 #include <machine/md_var.h>
56 #endif
57 
58 #include <linux/kobject.h>
59 #include <linux/device.h>
60 #include <linux/slab.h>
61 #include <linux/module.h>
62 #include <linux/cdev.h>
63 #include <linux/file.h>
64 #include <linux/sysfs.h>
65 #include <linux/mm.h>
66 #include <linux/io.h>
67 #include <linux/vmalloc.h>
68 #include <linux/netdevice.h>
69 #include <linux/timer.h>
70 #include <linux/workqueue.h>
71 #include <linux/rcupdate.h>
72 #include <linux/interrupt.h>
73 #include <linux/uaccess.h>
74 #include <linux/kernel.h>
75 
76 #include <vm/vm_pager.h>
77 
78 MALLOC_DEFINE(M_KMALLOC, "linux", "Linux kmalloc compat");
79 
80 #include <linux/rbtree.h>
81 /* Undo Linux compat changes. */
82 #undef RB_ROOT
83 #undef file
84 #undef cdev
85 #define	RB_ROOT(head)	(head)->rbh_root
86 
87 struct kobject linux_class_root;
88 struct device linux_root_device;
89 struct class linux_class_misc;
90 struct list_head pci_drivers;
91 struct list_head pci_devices;
92 struct net init_net;
93 spinlock_t pci_lock;
94 struct sx linux_global_rcu_lock;
95 
96 unsigned long linux_timer_hz_mask;
97 
98 int
99 panic_cmp(struct rb_node *one, struct rb_node *two)
100 {
101 	panic("no cmp");
102 }
103 
104 RB_GENERATE(linux_root, rb_node, __entry, panic_cmp);
105 
106 int
107 kobject_set_name_vargs(struct kobject *kobj, const char *fmt, va_list args)
108 {
109 	va_list tmp_va;
110 	int len;
111 	char *old;
112 	char *name;
113 	char dummy;
114 
115 	old = kobj->name;
116 
117 	if (old && fmt == NULL)
118 		return (0);
119 
120 	/* compute length of string */
121 	va_copy(tmp_va, args);
122 	len = vsnprintf(&dummy, 0, fmt, tmp_va);
123 	va_end(tmp_va);
124 
125 	/* account for zero termination */
126 	len++;
127 
128 	/* check for error */
129 	if (len < 1)
130 		return (-EINVAL);
131 
132 	/* allocate memory for string */
133 	name = kzalloc(len, GFP_KERNEL);
134 	if (name == NULL)
135 		return (-ENOMEM);
136 	vsnprintf(name, len, fmt, args);
137 	kobj->name = name;
138 
139 	/* free old string */
140 	kfree(old);
141 
142 	/* filter new string */
143 	for (; *name != '\0'; name++)
144 		if (*name == '/')
145 			*name = '!';
146 	return (0);
147 }
148 
149 int
150 kobject_set_name(struct kobject *kobj, const char *fmt, ...)
151 {
152 	va_list args;
153 	int error;
154 
155 	va_start(args, fmt);
156 	error = kobject_set_name_vargs(kobj, fmt, args);
157 	va_end(args);
158 
159 	return (error);
160 }
161 
162 static int
163 kobject_add_complete(struct kobject *kobj, struct kobject *parent)
164 {
165 	const struct kobj_type *t;
166 	int error;
167 
168 	kobj->parent = parent;
169 	error = sysfs_create_dir(kobj);
170 	if (error == 0 && kobj->ktype && kobj->ktype->default_attrs) {
171 		struct attribute **attr;
172 		t = kobj->ktype;
173 
174 		for (attr = t->default_attrs; *attr != NULL; attr++) {
175 			error = sysfs_create_file(kobj, *attr);
176 			if (error)
177 				break;
178 		}
179 		if (error)
180 			sysfs_remove_dir(kobj);
181 
182 	}
183 	return (error);
184 }
185 
186 int
187 kobject_add(struct kobject *kobj, struct kobject *parent, const char *fmt, ...)
188 {
189 	va_list args;
190 	int error;
191 
192 	va_start(args, fmt);
193 	error = kobject_set_name_vargs(kobj, fmt, args);
194 	va_end(args);
195 	if (error)
196 		return (error);
197 
198 	return kobject_add_complete(kobj, parent);
199 }
200 
201 void
202 linux_kobject_release(struct kref *kref)
203 {
204 	struct kobject *kobj;
205 	char *name;
206 
207 	kobj = container_of(kref, struct kobject, kref);
208 	sysfs_remove_dir(kobj);
209 	name = kobj->name;
210 	if (kobj->ktype && kobj->ktype->release)
211 		kobj->ktype->release(kobj);
212 	kfree(name);
213 }
214 
215 static void
216 linux_kobject_kfree(struct kobject *kobj)
217 {
218 	kfree(kobj);
219 }
220 
221 static void
222 linux_kobject_kfree_name(struct kobject *kobj)
223 {
224 	if (kobj) {
225 		kfree(kobj->name);
226 	}
227 }
228 
229 const struct kobj_type linux_kfree_type = {
230 	.release = linux_kobject_kfree
231 };
232 
233 static void
234 linux_device_release(struct device *dev)
235 {
236 	pr_debug("linux_device_release: %s\n", dev_name(dev));
237 	kfree(dev);
238 }
239 
240 static ssize_t
241 linux_class_show(struct kobject *kobj, struct attribute *attr, char *buf)
242 {
243 	struct class_attribute *dattr;
244 	ssize_t error;
245 
246 	dattr = container_of(attr, struct class_attribute, attr);
247 	error = -EIO;
248 	if (dattr->show)
249 		error = dattr->show(container_of(kobj, struct class, kobj),
250 		    dattr, buf);
251 	return (error);
252 }
253 
254 static ssize_t
255 linux_class_store(struct kobject *kobj, struct attribute *attr, const char *buf,
256     size_t count)
257 {
258 	struct class_attribute *dattr;
259 	ssize_t error;
260 
261 	dattr = container_of(attr, struct class_attribute, attr);
262 	error = -EIO;
263 	if (dattr->store)
264 		error = dattr->store(container_of(kobj, struct class, kobj),
265 		    dattr, buf, count);
266 	return (error);
267 }
268 
269 static void
270 linux_class_release(struct kobject *kobj)
271 {
272 	struct class *class;
273 
274 	class = container_of(kobj, struct class, kobj);
275 	if (class->class_release)
276 		class->class_release(class);
277 }
278 
279 static const struct sysfs_ops linux_class_sysfs = {
280 	.show  = linux_class_show,
281 	.store = linux_class_store,
282 };
283 
284 const struct kobj_type linux_class_ktype = {
285 	.release = linux_class_release,
286 	.sysfs_ops = &linux_class_sysfs
287 };
288 
289 static void
290 linux_dev_release(struct kobject *kobj)
291 {
292 	struct device *dev;
293 
294 	dev = container_of(kobj, struct device, kobj);
295 	/* This is the precedence defined by linux. */
296 	if (dev->release)
297 		dev->release(dev);
298 	else if (dev->class && dev->class->dev_release)
299 		dev->class->dev_release(dev);
300 }
301 
302 static ssize_t
303 linux_dev_show(struct kobject *kobj, struct attribute *attr, char *buf)
304 {
305 	struct device_attribute *dattr;
306 	ssize_t error;
307 
308 	dattr = container_of(attr, struct device_attribute, attr);
309 	error = -EIO;
310 	if (dattr->show)
311 		error = dattr->show(container_of(kobj, struct device, kobj),
312 		    dattr, buf);
313 	return (error);
314 }
315 
316 static ssize_t
317 linux_dev_store(struct kobject *kobj, struct attribute *attr, const char *buf,
318     size_t count)
319 {
320 	struct device_attribute *dattr;
321 	ssize_t error;
322 
323 	dattr = container_of(attr, struct device_attribute, attr);
324 	error = -EIO;
325 	if (dattr->store)
326 		error = dattr->store(container_of(kobj, struct device, kobj),
327 		    dattr, buf, count);
328 	return (error);
329 }
330 
331 static const struct sysfs_ops linux_dev_sysfs = {
332 	.show  = linux_dev_show,
333 	.store = linux_dev_store,
334 };
335 
336 const struct kobj_type linux_dev_ktype = {
337 	.release = linux_dev_release,
338 	.sysfs_ops = &linux_dev_sysfs
339 };
340 
341 struct device *
342 device_create(struct class *class, struct device *parent, dev_t devt,
343     void *drvdata, const char *fmt, ...)
344 {
345 	struct device *dev;
346 	va_list args;
347 
348 	dev = kzalloc(sizeof(*dev), M_WAITOK);
349 	dev->parent = parent;
350 	dev->class = class;
351 	dev->devt = devt;
352 	dev->driver_data = drvdata;
353 	dev->release = linux_device_release;
354 	va_start(args, fmt);
355 	kobject_set_name_vargs(&dev->kobj, fmt, args);
356 	va_end(args);
357 	device_register(dev);
358 
359 	return (dev);
360 }
361 
362 int
363 kobject_init_and_add(struct kobject *kobj, const struct kobj_type *ktype,
364     struct kobject *parent, const char *fmt, ...)
365 {
366 	va_list args;
367 	int error;
368 
369 	kobject_init(kobj, ktype);
370 	kobj->ktype = ktype;
371 	kobj->parent = parent;
372 	kobj->name = NULL;
373 
374 	va_start(args, fmt);
375 	error = kobject_set_name_vargs(kobj, fmt, args);
376 	va_end(args);
377 	if (error)
378 		return (error);
379 	return kobject_add_complete(kobj, parent);
380 }
381 
382 static void
383 linux_set_current(struct thread *td, struct task_struct *t)
384 {
385 	memset(t, 0, sizeof(*t));
386 	task_struct_fill(td, t);
387 	task_struct_set(td, t);
388 }
389 
390 static void
391 linux_clear_current(struct thread *td)
392 {
393 	task_struct_set(td, NULL);
394 }
395 
396 static void
397 linux_file_dtor(void *cdp)
398 {
399 	struct linux_file *filp;
400 	struct task_struct t;
401 	struct thread *td;
402 
403 	td = curthread;
404 	filp = cdp;
405 	linux_set_current(td, &t);
406 	filp->f_op->release(filp->f_vnode, filp);
407 	linux_clear_current(td);
408 	vdrop(filp->f_vnode);
409 	kfree(filp);
410 }
411 
412 static int
413 linux_dev_open(struct cdev *dev, int oflags, int devtype, struct thread *td)
414 {
415 	struct linux_cdev *ldev;
416 	struct linux_file *filp;
417 	struct task_struct t;
418 	struct file *file;
419 	int error;
420 
421 	file = td->td_fpop;
422 	ldev = dev->si_drv1;
423 	if (ldev == NULL)
424 		return (ENODEV);
425 	filp = kzalloc(sizeof(*filp), GFP_KERNEL);
426 	filp->f_dentry = &filp->f_dentry_store;
427 	filp->f_op = ldev->ops;
428 	filp->f_flags = file->f_flag;
429 	vhold(file->f_vnode);
430 	filp->f_vnode = file->f_vnode;
431 	linux_set_current(td, &t);
432 	if (filp->f_op->open) {
433 		error = -filp->f_op->open(file->f_vnode, filp);
434 		if (error) {
435 			kfree(filp);
436 			goto done;
437 		}
438 	}
439 	error = devfs_set_cdevpriv(filp, linux_file_dtor);
440 	if (error) {
441 		filp->f_op->release(file->f_vnode, filp);
442 		kfree(filp);
443 	}
444 done:
445 	linux_clear_current(td);
446 	return (error);
447 }
448 
449 static int
450 linux_dev_close(struct cdev *dev, int fflag, int devtype, struct thread *td)
451 {
452 	struct linux_cdev *ldev;
453 	struct linux_file *filp;
454 	struct file *file;
455 	int error;
456 
457 	file = td->td_fpop;
458 	ldev = dev->si_drv1;
459 	if (ldev == NULL)
460 		return (0);
461 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
462 		return (error);
463 	filp->f_flags = file->f_flag;
464         devfs_clear_cdevpriv();
465 
466 
467 	return (0);
468 }
469 
470 #define	LINUX_IOCTL_MIN_PTR 0x10000UL
471 #define	LINUX_IOCTL_MAX_PTR (LINUX_IOCTL_MIN_PTR + IOCPARM_MAX)
472 
473 static inline int
474 linux_remap_address(void **uaddr, size_t len)
475 {
476 	uintptr_t uaddr_val = (uintptr_t)(*uaddr);
477 
478 	if (unlikely(uaddr_val >= LINUX_IOCTL_MIN_PTR &&
479 	    uaddr_val < LINUX_IOCTL_MAX_PTR)) {
480 		struct task_struct *pts = current;
481 		if (pts == NULL) {
482 			*uaddr = NULL;
483 			return (1);
484 		}
485 
486 		/* compute data offset */
487 		uaddr_val -= LINUX_IOCTL_MIN_PTR;
488 
489 		/* check that length is within bounds */
490 		if ((len > IOCPARM_MAX) ||
491 		    (uaddr_val + len) > pts->bsd_ioctl_len) {
492 			*uaddr = NULL;
493 			return (1);
494 		}
495 
496 		/* re-add kernel buffer address */
497 		uaddr_val += (uintptr_t)pts->bsd_ioctl_data;
498 
499 		/* update address location */
500 		*uaddr = (void *)uaddr_val;
501 		return (1);
502 	}
503 	return (0);
504 }
505 
506 int
507 linux_copyin(const void *uaddr, void *kaddr, size_t len)
508 {
509 	if (linux_remap_address(__DECONST(void **, &uaddr), len)) {
510 		if (uaddr == NULL)
511 			return (-EFAULT);
512 		memcpy(kaddr, uaddr, len);
513 		return (0);
514 	}
515 	return (-copyin(uaddr, kaddr, len));
516 }
517 
518 int
519 linux_copyout(const void *kaddr, void *uaddr, size_t len)
520 {
521 	if (linux_remap_address(&uaddr, len)) {
522 		if (uaddr == NULL)
523 			return (-EFAULT);
524 		memcpy(uaddr, kaddr, len);
525 		return (0);
526 	}
527 	return (-copyout(kaddr, uaddr, len));
528 }
529 
530 static int
531 linux_dev_ioctl(struct cdev *dev, u_long cmd, caddr_t data, int fflag,
532     struct thread *td)
533 {
534 	struct linux_cdev *ldev;
535 	struct linux_file *filp;
536 	struct task_struct t;
537 	struct file *file;
538 	unsigned size;
539 	int error;
540 
541 	file = td->td_fpop;
542 	ldev = dev->si_drv1;
543 	if (ldev == NULL)
544 		return (0);
545 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
546 		return (error);
547 	filp->f_flags = file->f_flag;
548 	linux_set_current(td, &t);
549 	size = IOCPARM_LEN(cmd);
550 	/* refer to logic in sys_ioctl() */
551 	if (size > 0) {
552 		/*
553 		 * Setup hint for linux_copyin() and linux_copyout().
554 		 *
555 		 * Background: Linux code expects a user-space address
556 		 * while FreeBSD supplies a kernel-space address.
557 		 */
558 		t.bsd_ioctl_data = data;
559 		t.bsd_ioctl_len = size;
560 		data = (void *)LINUX_IOCTL_MIN_PTR;
561 	} else {
562 		/* fetch user-space pointer */
563 		data = *(void **)data;
564 	}
565 	if (filp->f_op->unlocked_ioctl)
566 		error = -filp->f_op->unlocked_ioctl(filp, cmd, (u_long)data);
567 	else
568 		error = ENOTTY;
569 	linux_clear_current(td);
570 
571 	return (error);
572 }
573 
574 static int
575 linux_dev_read(struct cdev *dev, struct uio *uio, int ioflag)
576 {
577 	struct linux_cdev *ldev;
578 	struct linux_file *filp;
579 	struct task_struct t;
580 	struct thread *td;
581 	struct file *file;
582 	ssize_t bytes;
583 	int error;
584 
585 	td = curthread;
586 	file = td->td_fpop;
587 	ldev = dev->si_drv1;
588 	if (ldev == NULL)
589 		return (0);
590 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
591 		return (error);
592 	filp->f_flags = file->f_flag;
593 	/* XXX no support for I/O vectors currently */
594 	if (uio->uio_iovcnt != 1)
595 		return (EOPNOTSUPP);
596 	linux_set_current(td, &t);
597 	if (filp->f_op->read) {
598 		bytes = filp->f_op->read(filp, uio->uio_iov->iov_base,
599 		    uio->uio_iov->iov_len, &uio->uio_offset);
600 		if (bytes >= 0) {
601 			uio->uio_iov->iov_base =
602 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
603 			uio->uio_iov->iov_len -= bytes;
604 			uio->uio_resid -= bytes;
605 		} else
606 			error = -bytes;
607 	} else
608 		error = ENXIO;
609 	linux_clear_current(td);
610 
611 	return (error);
612 }
613 
614 static int
615 linux_dev_write(struct cdev *dev, struct uio *uio, int ioflag)
616 {
617 	struct linux_cdev *ldev;
618 	struct linux_file *filp;
619 	struct task_struct t;
620 	struct thread *td;
621 	struct file *file;
622 	ssize_t bytes;
623 	int error;
624 
625 	td = curthread;
626 	file = td->td_fpop;
627 	ldev = dev->si_drv1;
628 	if (ldev == NULL)
629 		return (0);
630 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
631 		return (error);
632 	filp->f_flags = file->f_flag;
633 	/* XXX no support for I/O vectors currently */
634 	if (uio->uio_iovcnt != 1)
635 		return (EOPNOTSUPP);
636 	linux_set_current(td, &t);
637 	if (filp->f_op->write) {
638 		bytes = filp->f_op->write(filp, uio->uio_iov->iov_base,
639 		    uio->uio_iov->iov_len, &uio->uio_offset);
640 		if (bytes >= 0) {
641 			uio->uio_iov->iov_base =
642 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
643 			uio->uio_iov->iov_len -= bytes;
644 			uio->uio_resid -= bytes;
645 		} else
646 			error = -bytes;
647 	} else
648 		error = ENXIO;
649 	linux_clear_current(td);
650 
651 	return (error);
652 }
653 
654 static int
655 linux_dev_poll(struct cdev *dev, int events, struct thread *td)
656 {
657 	struct linux_cdev *ldev;
658 	struct linux_file *filp;
659 	struct task_struct t;
660 	struct file *file;
661 	int revents;
662 	int error;
663 
664 	file = td->td_fpop;
665 	ldev = dev->si_drv1;
666 	if (ldev == NULL)
667 		return (0);
668 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
669 		return (error);
670 	filp->f_flags = file->f_flag;
671 	linux_set_current(td, &t);
672 	if (filp->f_op->poll)
673 		revents = filp->f_op->poll(filp, NULL) & events;
674 	else
675 		revents = 0;
676 	linux_clear_current(td);
677 
678 	return (revents);
679 }
680 
681 static int
682 linux_dev_mmap_single(struct cdev *dev, vm_ooffset_t *offset,
683     vm_size_t size, struct vm_object **object, int nprot)
684 {
685 	struct linux_cdev *ldev;
686 	struct linux_file *filp;
687 	struct thread *td;
688 	struct task_struct t;
689 	struct file *file;
690 	struct vm_area_struct vma;
691 	int error;
692 
693 	td = curthread;
694 	file = td->td_fpop;
695 	ldev = dev->si_drv1;
696 	if (ldev == NULL)
697 		return (ENODEV);
698 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
699 		return (error);
700 	filp->f_flags = file->f_flag;
701 	linux_set_current(td, &t);
702 	vma.vm_start = 0;
703 	vma.vm_end = size;
704 	vma.vm_pgoff = *offset / PAGE_SIZE;
705 	vma.vm_pfn = 0;
706 	vma.vm_page_prot = VM_MEMATTR_DEFAULT;
707 	if (filp->f_op->mmap) {
708 		error = -filp->f_op->mmap(filp, &vma);
709 		if (error == 0) {
710 			struct sglist *sg;
711 
712 			sg = sglist_alloc(1, M_WAITOK);
713 			sglist_append_phys(sg,
714 			    (vm_paddr_t)vma.vm_pfn << PAGE_SHIFT, vma.vm_len);
715 			*object = vm_pager_allocate(OBJT_SG, sg, vma.vm_len,
716 			    nprot, 0, td->td_ucred);
717 		        if (*object == NULL) {
718 				sglist_free(sg);
719 				error = EINVAL;
720 				goto done;
721 			}
722 			*offset = 0;
723 			if (vma.vm_page_prot != VM_MEMATTR_DEFAULT) {
724 				VM_OBJECT_WLOCK(*object);
725 				vm_object_set_memattr(*object,
726 				    vma.vm_page_prot);
727 				VM_OBJECT_WUNLOCK(*object);
728 			}
729 		}
730 	} else
731 		error = ENODEV;
732 done:
733 	linux_clear_current(td);
734 	return (error);
735 }
736 
737 struct cdevsw linuxcdevsw = {
738 	.d_version = D_VERSION,
739 	.d_flags = D_TRACKCLOSE,
740 	.d_open = linux_dev_open,
741 	.d_close = linux_dev_close,
742 	.d_read = linux_dev_read,
743 	.d_write = linux_dev_write,
744 	.d_ioctl = linux_dev_ioctl,
745 	.d_mmap_single = linux_dev_mmap_single,
746 	.d_poll = linux_dev_poll,
747 };
748 
749 static int
750 linux_file_read(struct file *file, struct uio *uio, struct ucred *active_cred,
751     int flags, struct thread *td)
752 {
753 	struct linux_file *filp;
754 	struct task_struct t;
755 	ssize_t bytes;
756 	int error;
757 
758 	error = 0;
759 	filp = (struct linux_file *)file->f_data;
760 	filp->f_flags = file->f_flag;
761 	/* XXX no support for I/O vectors currently */
762 	if (uio->uio_iovcnt != 1)
763 		return (EOPNOTSUPP);
764 	linux_set_current(td, &t);
765 	if (filp->f_op->read) {
766 		bytes = filp->f_op->read(filp, uio->uio_iov->iov_base,
767 		    uio->uio_iov->iov_len, &uio->uio_offset);
768 		if (bytes >= 0) {
769 			uio->uio_iov->iov_base =
770 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
771 			uio->uio_iov->iov_len -= bytes;
772 			uio->uio_resid -= bytes;
773 		} else
774 			error = -bytes;
775 	} else
776 		error = ENXIO;
777 	linux_clear_current(td);
778 
779 	return (error);
780 }
781 
782 static int
783 linux_file_poll(struct file *file, int events, struct ucred *active_cred,
784     struct thread *td)
785 {
786 	struct linux_file *filp;
787 	struct task_struct t;
788 	int revents;
789 
790 	filp = (struct linux_file *)file->f_data;
791 	filp->f_flags = file->f_flag;
792 	linux_set_current(td, &t);
793 	if (filp->f_op->poll)
794 		revents = filp->f_op->poll(filp, NULL) & events;
795 	else
796 		revents = 0;
797 	linux_clear_current(td);
798 
799 	return (revents);
800 }
801 
802 static int
803 linux_file_close(struct file *file, struct thread *td)
804 {
805 	struct linux_file *filp;
806 	struct task_struct t;
807 	int error;
808 
809 	filp = (struct linux_file *)file->f_data;
810 	filp->f_flags = file->f_flag;
811 	linux_set_current(td, &t);
812 	error = -filp->f_op->release(NULL, filp);
813 	linux_clear_current(td);
814 	funsetown(&filp->f_sigio);
815 	kfree(filp);
816 
817 	return (error);
818 }
819 
820 static int
821 linux_file_ioctl(struct file *fp, u_long cmd, void *data, struct ucred *cred,
822     struct thread *td)
823 {
824 	struct linux_file *filp;
825 	struct task_struct t;
826 	int error;
827 
828 	filp = (struct linux_file *)fp->f_data;
829 	filp->f_flags = fp->f_flag;
830 	error = 0;
831 
832 	linux_set_current(td, &t);
833 	switch (cmd) {
834 	case FIONBIO:
835 		break;
836 	case FIOASYNC:
837 		if (filp->f_op->fasync == NULL)
838 			break;
839 		error = filp->f_op->fasync(0, filp, fp->f_flag & FASYNC);
840 		break;
841 	case FIOSETOWN:
842 		error = fsetown(*(int *)data, &filp->f_sigio);
843 		if (error == 0)
844 			error = filp->f_op->fasync(0, filp,
845 			    fp->f_flag & FASYNC);
846 		break;
847 	case FIOGETOWN:
848 		*(int *)data = fgetown(&filp->f_sigio);
849 		break;
850 	default:
851 		error = ENOTTY;
852 		break;
853 	}
854 	linux_clear_current(td);
855 	return (error);
856 }
857 
858 static int
859 linux_file_stat(struct file *fp, struct stat *sb, struct ucred *active_cred,
860     struct thread *td)
861 {
862 
863 	return (EOPNOTSUPP);
864 }
865 
866 static int
867 linux_file_fill_kinfo(struct file *fp, struct kinfo_file *kif,
868     struct filedesc *fdp)
869 {
870 
871 	return (0);
872 }
873 
874 struct fileops linuxfileops = {
875 	.fo_read = linux_file_read,
876 	.fo_write = invfo_rdwr,
877 	.fo_truncate = invfo_truncate,
878 	.fo_kqfilter = invfo_kqfilter,
879 	.fo_stat = linux_file_stat,
880 	.fo_fill_kinfo = linux_file_fill_kinfo,
881 	.fo_poll = linux_file_poll,
882 	.fo_close = linux_file_close,
883 	.fo_ioctl = linux_file_ioctl,
884 	.fo_chmod = invfo_chmod,
885 	.fo_chown = invfo_chown,
886 	.fo_sendfile = invfo_sendfile,
887 };
888 
889 /*
890  * Hash of vmmap addresses.  This is infrequently accessed and does not
891  * need to be particularly large.  This is done because we must store the
892  * caller's idea of the map size to properly unmap.
893  */
894 struct vmmap {
895 	LIST_ENTRY(vmmap)	vm_next;
896 	void 			*vm_addr;
897 	unsigned long		vm_size;
898 };
899 
900 struct vmmaphd {
901 	struct vmmap *lh_first;
902 };
903 #define	VMMAP_HASH_SIZE	64
904 #define	VMMAP_HASH_MASK	(VMMAP_HASH_SIZE - 1)
905 #define	VM_HASH(addr)	((uintptr_t)(addr) >> PAGE_SHIFT) & VMMAP_HASH_MASK
906 static struct vmmaphd vmmaphead[VMMAP_HASH_SIZE];
907 static struct mtx vmmaplock;
908 
909 static void
910 vmmap_add(void *addr, unsigned long size)
911 {
912 	struct vmmap *vmmap;
913 
914 	vmmap = kmalloc(sizeof(*vmmap), GFP_KERNEL);
915 	mtx_lock(&vmmaplock);
916 	vmmap->vm_size = size;
917 	vmmap->vm_addr = addr;
918 	LIST_INSERT_HEAD(&vmmaphead[VM_HASH(addr)], vmmap, vm_next);
919 	mtx_unlock(&vmmaplock);
920 }
921 
922 static struct vmmap *
923 vmmap_remove(void *addr)
924 {
925 	struct vmmap *vmmap;
926 
927 	mtx_lock(&vmmaplock);
928 	LIST_FOREACH(vmmap, &vmmaphead[VM_HASH(addr)], vm_next)
929 		if (vmmap->vm_addr == addr)
930 			break;
931 	if (vmmap)
932 		LIST_REMOVE(vmmap, vm_next);
933 	mtx_unlock(&vmmaplock);
934 
935 	return (vmmap);
936 }
937 
938 #if defined(__i386__) || defined(__amd64__)
939 void *
940 _ioremap_attr(vm_paddr_t phys_addr, unsigned long size, int attr)
941 {
942 	void *addr;
943 
944 	addr = pmap_mapdev_attr(phys_addr, size, attr);
945 	if (addr == NULL)
946 		return (NULL);
947 	vmmap_add(addr, size);
948 
949 	return (addr);
950 }
951 #endif
952 
953 void
954 iounmap(void *addr)
955 {
956 	struct vmmap *vmmap;
957 
958 	vmmap = vmmap_remove(addr);
959 	if (vmmap == NULL)
960 		return;
961 #if defined(__i386__) || defined(__amd64__)
962 	pmap_unmapdev((vm_offset_t)addr, vmmap->vm_size);
963 #endif
964 	kfree(vmmap);
965 }
966 
967 
968 void *
969 vmap(struct page **pages, unsigned int count, unsigned long flags, int prot)
970 {
971 	vm_offset_t off;
972 	size_t size;
973 
974 	size = count * PAGE_SIZE;
975 	off = kva_alloc(size);
976 	if (off == 0)
977 		return (NULL);
978 	vmmap_add((void *)off, size);
979 	pmap_qenter(off, pages, count);
980 
981 	return ((void *)off);
982 }
983 
984 void
985 vunmap(void *addr)
986 {
987 	struct vmmap *vmmap;
988 
989 	vmmap = vmmap_remove(addr);
990 	if (vmmap == NULL)
991 		return;
992 	pmap_qremove((vm_offset_t)addr, vmmap->vm_size / PAGE_SIZE);
993 	kva_free((vm_offset_t)addr, vmmap->vm_size);
994 	kfree(vmmap);
995 }
996 
997 char *
998 kvasprintf(gfp_t gfp, const char *fmt, va_list ap)
999 {
1000 	unsigned int len;
1001 	char *p;
1002 	va_list aq;
1003 
1004 	va_copy(aq, ap);
1005 	len = vsnprintf(NULL, 0, fmt, aq);
1006 	va_end(aq);
1007 
1008 	p = kmalloc(len + 1, gfp);
1009 	if (p != NULL)
1010 		vsnprintf(p, len + 1, fmt, ap);
1011 
1012 	return (p);
1013 }
1014 
1015 char *
1016 kasprintf(gfp_t gfp, const char *fmt, ...)
1017 {
1018 	va_list ap;
1019 	char *p;
1020 
1021 	va_start(ap, fmt);
1022 	p = kvasprintf(gfp, fmt, ap);
1023 	va_end(ap);
1024 
1025 	return (p);
1026 }
1027 
1028 static void
1029 linux_timer_callback_wrapper(void *context)
1030 {
1031 	struct timer_list *timer;
1032 
1033 	timer = context;
1034 	timer->function(timer->data);
1035 }
1036 
1037 void
1038 mod_timer(struct timer_list *timer, unsigned long expires)
1039 {
1040 
1041 	timer->expires = expires;
1042 	callout_reset(&timer->timer_callout,
1043 	    linux_timer_jiffies_until(expires),
1044 	    &linux_timer_callback_wrapper, timer);
1045 }
1046 
1047 void
1048 add_timer(struct timer_list *timer)
1049 {
1050 
1051 	callout_reset(&timer->timer_callout,
1052 	    linux_timer_jiffies_until(timer->expires),
1053 	    &linux_timer_callback_wrapper, timer);
1054 }
1055 
1056 static void
1057 linux_timer_init(void *arg)
1058 {
1059 
1060 	/*
1061 	 * Compute an internal HZ value which can divide 2**32 to
1062 	 * avoid timer rounding problems when the tick value wraps
1063 	 * around 2**32:
1064 	 */
1065 	linux_timer_hz_mask = 1;
1066 	while (linux_timer_hz_mask < (unsigned long)hz)
1067 		linux_timer_hz_mask *= 2;
1068 	linux_timer_hz_mask--;
1069 }
1070 SYSINIT(linux_timer, SI_SUB_DRIVERS, SI_ORDER_FIRST, linux_timer_init, NULL);
1071 
1072 void
1073 linux_complete_common(struct completion *c, int all)
1074 {
1075 	int wakeup_swapper;
1076 
1077 	sleepq_lock(c);
1078 	c->done++;
1079 	if (all)
1080 		wakeup_swapper = sleepq_broadcast(c, SLEEPQ_SLEEP, 0, 0);
1081 	else
1082 		wakeup_swapper = sleepq_signal(c, SLEEPQ_SLEEP, 0, 0);
1083 	sleepq_release(c);
1084 	if (wakeup_swapper)
1085 		kick_proc0();
1086 }
1087 
1088 /*
1089  * Indefinite wait for done != 0 with or without signals.
1090  */
1091 long
1092 linux_wait_for_common(struct completion *c, int flags)
1093 {
1094 
1095 	if (flags != 0)
1096 		flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
1097 	else
1098 		flags = SLEEPQ_SLEEP;
1099 	for (;;) {
1100 		sleepq_lock(c);
1101 		if (c->done)
1102 			break;
1103 		sleepq_add(c, NULL, "completion", flags, 0);
1104 		if (flags & SLEEPQ_INTERRUPTIBLE) {
1105 			if (sleepq_wait_sig(c, 0) != 0)
1106 				return (-ERESTARTSYS);
1107 		} else
1108 			sleepq_wait(c, 0);
1109 	}
1110 	c->done--;
1111 	sleepq_release(c);
1112 
1113 	return (0);
1114 }
1115 
1116 /*
1117  * Time limited wait for done != 0 with or without signals.
1118  */
1119 long
1120 linux_wait_for_timeout_common(struct completion *c, long timeout, int flags)
1121 {
1122 	long end = jiffies + timeout;
1123 
1124 	if (flags != 0)
1125 		flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
1126 	else
1127 		flags = SLEEPQ_SLEEP;
1128 	for (;;) {
1129 		int ret;
1130 
1131 		sleepq_lock(c);
1132 		if (c->done)
1133 			break;
1134 		sleepq_add(c, NULL, "completion", flags, 0);
1135 		sleepq_set_timeout(c, linux_timer_jiffies_until(end));
1136 		if (flags & SLEEPQ_INTERRUPTIBLE)
1137 			ret = sleepq_timedwait_sig(c, 0);
1138 		else
1139 			ret = sleepq_timedwait(c, 0);
1140 		if (ret != 0) {
1141 			/* check for timeout or signal */
1142 			if (ret == EWOULDBLOCK)
1143 				return (0);
1144 			else
1145 				return (-ERESTARTSYS);
1146 		}
1147 	}
1148 	c->done--;
1149 	sleepq_release(c);
1150 
1151 	/* return how many jiffies are left */
1152 	return (linux_timer_jiffies_until(end));
1153 }
1154 
1155 int
1156 linux_try_wait_for_completion(struct completion *c)
1157 {
1158 	int isdone;
1159 
1160 	isdone = 1;
1161 	sleepq_lock(c);
1162 	if (c->done)
1163 		c->done--;
1164 	else
1165 		isdone = 0;
1166 	sleepq_release(c);
1167 	return (isdone);
1168 }
1169 
1170 int
1171 linux_completion_done(struct completion *c)
1172 {
1173 	int isdone;
1174 
1175 	isdone = 1;
1176 	sleepq_lock(c);
1177 	if (c->done == 0)
1178 		isdone = 0;
1179 	sleepq_release(c);
1180 	return (isdone);
1181 }
1182 
1183 void
1184 linux_delayed_work_fn(void *arg)
1185 {
1186 	struct delayed_work *work;
1187 
1188 	work = arg;
1189 	taskqueue_enqueue(work->work.taskqueue, &work->work.work_task);
1190 }
1191 
1192 void
1193 linux_work_fn(void *context, int pending)
1194 {
1195 	struct work_struct *work;
1196 
1197 	work = context;
1198 	work->fn(work);
1199 }
1200 
1201 void
1202 linux_flush_fn(void *context, int pending)
1203 {
1204 }
1205 
1206 struct workqueue_struct *
1207 linux_create_workqueue_common(const char *name, int cpus)
1208 {
1209 	struct workqueue_struct *wq;
1210 
1211 	wq = kmalloc(sizeof(*wq), M_WAITOK);
1212 	wq->taskqueue = taskqueue_create(name, M_WAITOK,
1213 	    taskqueue_thread_enqueue,  &wq->taskqueue);
1214 	atomic_set(&wq->draining, 0);
1215 	taskqueue_start_threads(&wq->taskqueue, cpus, PWAIT, "%s", name);
1216 
1217 	return (wq);
1218 }
1219 
1220 void
1221 destroy_workqueue(struct workqueue_struct *wq)
1222 {
1223 	taskqueue_free(wq->taskqueue);
1224 	kfree(wq);
1225 }
1226 
1227 static void
1228 linux_cdev_release(struct kobject *kobj)
1229 {
1230 	struct linux_cdev *cdev;
1231 	struct kobject *parent;
1232 
1233 	cdev = container_of(kobj, struct linux_cdev, kobj);
1234 	parent = kobj->parent;
1235 	if (cdev->cdev)
1236 		destroy_dev(cdev->cdev);
1237 	kfree(cdev);
1238 	kobject_put(parent);
1239 }
1240 
1241 static void
1242 linux_cdev_static_release(struct kobject *kobj)
1243 {
1244 	struct linux_cdev *cdev;
1245 	struct kobject *parent;
1246 
1247 	cdev = container_of(kobj, struct linux_cdev, kobj);
1248 	parent = kobj->parent;
1249 	if (cdev->cdev)
1250 		destroy_dev(cdev->cdev);
1251 	kobject_put(parent);
1252 }
1253 
1254 const struct kobj_type linux_cdev_ktype = {
1255 	.release = linux_cdev_release,
1256 };
1257 
1258 const struct kobj_type linux_cdev_static_ktype = {
1259 	.release = linux_cdev_static_release,
1260 };
1261 
1262 static void
1263 linux_handle_ifnet_link_event(void *arg, struct ifnet *ifp, int linkstate)
1264 {
1265 	struct notifier_block *nb;
1266 
1267 	nb = arg;
1268 	if (linkstate == LINK_STATE_UP)
1269 		nb->notifier_call(nb, NETDEV_UP, ifp);
1270 	else
1271 		nb->notifier_call(nb, NETDEV_DOWN, ifp);
1272 }
1273 
1274 static void
1275 linux_handle_ifnet_arrival_event(void *arg, struct ifnet *ifp)
1276 {
1277 	struct notifier_block *nb;
1278 
1279 	nb = arg;
1280 	nb->notifier_call(nb, NETDEV_REGISTER, ifp);
1281 }
1282 
1283 static void
1284 linux_handle_ifnet_departure_event(void *arg, struct ifnet *ifp)
1285 {
1286 	struct notifier_block *nb;
1287 
1288 	nb = arg;
1289 	nb->notifier_call(nb, NETDEV_UNREGISTER, ifp);
1290 }
1291 
1292 static void
1293 linux_handle_iflladdr_event(void *arg, struct ifnet *ifp)
1294 {
1295 	struct notifier_block *nb;
1296 
1297 	nb = arg;
1298 	nb->notifier_call(nb, NETDEV_CHANGEADDR, ifp);
1299 }
1300 
1301 static void
1302 linux_handle_ifaddr_event(void *arg, struct ifnet *ifp)
1303 {
1304 	struct notifier_block *nb;
1305 
1306 	nb = arg;
1307 	nb->notifier_call(nb, NETDEV_CHANGEIFADDR, ifp);
1308 }
1309 
1310 int
1311 register_netdevice_notifier(struct notifier_block *nb)
1312 {
1313 
1314 	nb->tags[NETDEV_UP] = EVENTHANDLER_REGISTER(
1315 	    ifnet_link_event, linux_handle_ifnet_link_event, nb, 0);
1316 	nb->tags[NETDEV_REGISTER] = EVENTHANDLER_REGISTER(
1317 	    ifnet_arrival_event, linux_handle_ifnet_arrival_event, nb, 0);
1318 	nb->tags[NETDEV_UNREGISTER] = EVENTHANDLER_REGISTER(
1319 	    ifnet_departure_event, linux_handle_ifnet_departure_event, nb, 0);
1320 	nb->tags[NETDEV_CHANGEADDR] = EVENTHANDLER_REGISTER(
1321 	    iflladdr_event, linux_handle_iflladdr_event, nb, 0);
1322 
1323 	return (0);
1324 }
1325 
1326 int
1327 register_inetaddr_notifier(struct notifier_block *nb)
1328 {
1329 
1330         nb->tags[NETDEV_CHANGEIFADDR] = EVENTHANDLER_REGISTER(
1331             ifaddr_event, linux_handle_ifaddr_event, nb, 0);
1332         return (0);
1333 }
1334 
1335 int
1336 unregister_netdevice_notifier(struct notifier_block *nb)
1337 {
1338 
1339         EVENTHANDLER_DEREGISTER(ifnet_link_event,
1340 	    nb->tags[NETDEV_UP]);
1341         EVENTHANDLER_DEREGISTER(ifnet_arrival_event,
1342 	    nb->tags[NETDEV_REGISTER]);
1343         EVENTHANDLER_DEREGISTER(ifnet_departure_event,
1344 	    nb->tags[NETDEV_UNREGISTER]);
1345         EVENTHANDLER_DEREGISTER(iflladdr_event,
1346 	    nb->tags[NETDEV_CHANGEADDR]);
1347 
1348 	return (0);
1349 }
1350 
1351 int
1352 unregister_inetaddr_notifier(struct notifier_block *nb)
1353 {
1354 
1355         EVENTHANDLER_DEREGISTER(ifaddr_event,
1356             nb->tags[NETDEV_CHANGEIFADDR]);
1357 
1358         return (0);
1359 }
1360 
1361 void
1362 linux_irq_handler(void *ent)
1363 {
1364 	struct irq_ent *irqe;
1365 
1366 	irqe = ent;
1367 	irqe->handler(irqe->irq, irqe->arg);
1368 }
1369 
1370 #if defined(__i386__) || defined(__amd64__)
1371 bool linux_cpu_has_clflush;
1372 #endif
1373 
1374 static void
1375 linux_compat_init(void *arg)
1376 {
1377 	struct sysctl_oid *rootoid;
1378 	int i;
1379 
1380 #if defined(__i386__) || defined(__amd64__)
1381 	linux_cpu_has_clflush = (cpu_feature & CPUID_CLFSH);
1382 #endif
1383 	sx_init(&linux_global_rcu_lock, "LinuxGlobalRCU");
1384 
1385 	rootoid = SYSCTL_ADD_ROOT_NODE(NULL,
1386 	    OID_AUTO, "sys", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "sys");
1387 	kobject_init(&linux_class_root, &linux_class_ktype);
1388 	kobject_set_name(&linux_class_root, "class");
1389 	linux_class_root.oidp = SYSCTL_ADD_NODE(NULL, SYSCTL_CHILDREN(rootoid),
1390 	    OID_AUTO, "class", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "class");
1391 	kobject_init(&linux_root_device.kobj, &linux_dev_ktype);
1392 	kobject_set_name(&linux_root_device.kobj, "device");
1393 	linux_root_device.kobj.oidp = SYSCTL_ADD_NODE(NULL,
1394 	    SYSCTL_CHILDREN(rootoid), OID_AUTO, "device", CTLFLAG_RD, NULL,
1395 	    "device");
1396 	linux_root_device.bsddev = root_bus;
1397 	linux_class_misc.name = "misc";
1398 	class_register(&linux_class_misc);
1399 	INIT_LIST_HEAD(&pci_drivers);
1400 	INIT_LIST_HEAD(&pci_devices);
1401 	spin_lock_init(&pci_lock);
1402 	mtx_init(&vmmaplock, "IO Map lock", NULL, MTX_DEF);
1403 	for (i = 0; i < VMMAP_HASH_SIZE; i++)
1404 		LIST_INIT(&vmmaphead[i]);
1405 }
1406 SYSINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_init, NULL);
1407 
1408 static void
1409 linux_compat_uninit(void *arg)
1410 {
1411 	linux_kobject_kfree_name(&linux_class_root);
1412 	linux_kobject_kfree_name(&linux_root_device.kobj);
1413 	linux_kobject_kfree_name(&linux_class_misc.kobj);
1414 
1415 	synchronize_rcu();
1416 	sx_destroy(&linux_global_rcu_lock);
1417 }
1418 SYSUNINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_uninit, NULL);
1419 
1420 /*
1421  * NOTE: Linux frequently uses "unsigned long" for pointer to integer
1422  * conversion and vice versa, where in FreeBSD "uintptr_t" would be
1423  * used. Assert these types have the same size, else some parts of the
1424  * LinuxKPI may not work like expected:
1425  */
1426 CTASSERT(sizeof(unsigned long) == sizeof(uintptr_t));
1427