xref: /freebsd/sys/compat/linuxkpi/common/src/linux_compat.c (revision 99429157e8615dc3b7f11afbe3ed92de7476a5db)
1 /*-
2  * Copyright (c) 2010 Isilon Systems, Inc.
3  * Copyright (c) 2010 iX Systems, Inc.
4  * Copyright (c) 2010 Panasas, Inc.
5  * Copyright (c) 2013-2017 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 #include <vm/vm_object.h>
52 #include <vm/vm_page.h>
53 #include <vm/vm_pager.h>
54 
55 #include <machine/stdarg.h>
56 
57 #if defined(__i386__) || defined(__amd64__)
58 #include <machine/md_var.h>
59 #endif
60 
61 #include <linux/kobject.h>
62 #include <linux/device.h>
63 #include <linux/slab.h>
64 #include <linux/module.h>
65 #include <linux/moduleparam.h>
66 #include <linux/cdev.h>
67 #include <linux/file.h>
68 #include <linux/sysfs.h>
69 #include <linux/mm.h>
70 #include <linux/io.h>
71 #include <linux/vmalloc.h>
72 #include <linux/netdevice.h>
73 #include <linux/timer.h>
74 #include <linux/interrupt.h>
75 #include <linux/uaccess.h>
76 #include <linux/kernel.h>
77 #include <linux/list.h>
78 #include <linux/compat.h>
79 #include <linux/poll.h>
80 #include <linux/smp.h>
81 
82 #if defined(__i386__) || defined(__amd64__)
83 #include <asm/smp.h>
84 #endif
85 
86 SYSCTL_NODE(_compat, OID_AUTO, linuxkpi, CTLFLAG_RW, 0, "LinuxKPI parameters");
87 
88 MALLOC_DEFINE(M_KMALLOC, "linux", "Linux kmalloc compat");
89 
90 #include <linux/rbtree.h>
91 /* Undo Linux compat changes. */
92 #undef RB_ROOT
93 #undef file
94 #undef cdev
95 #define	RB_ROOT(head)	(head)->rbh_root
96 
97 static struct vm_area_struct *linux_cdev_handle_find(void *handle);
98 
99 struct kobject linux_class_root;
100 struct device linux_root_device;
101 struct class linux_class_misc;
102 struct list_head pci_drivers;
103 struct list_head pci_devices;
104 spinlock_t pci_lock;
105 
106 unsigned long linux_timer_hz_mask;
107 
108 int
109 panic_cmp(struct rb_node *one, struct rb_node *two)
110 {
111 	panic("no cmp");
112 }
113 
114 RB_GENERATE(linux_root, rb_node, __entry, panic_cmp);
115 
116 int
117 kobject_set_name_vargs(struct kobject *kobj, const char *fmt, va_list args)
118 {
119 	va_list tmp_va;
120 	int len;
121 	char *old;
122 	char *name;
123 	char dummy;
124 
125 	old = kobj->name;
126 
127 	if (old && fmt == NULL)
128 		return (0);
129 
130 	/* compute length of string */
131 	va_copy(tmp_va, args);
132 	len = vsnprintf(&dummy, 0, fmt, tmp_va);
133 	va_end(tmp_va);
134 
135 	/* account for zero termination */
136 	len++;
137 
138 	/* check for error */
139 	if (len < 1)
140 		return (-EINVAL);
141 
142 	/* allocate memory for string */
143 	name = kzalloc(len, GFP_KERNEL);
144 	if (name == NULL)
145 		return (-ENOMEM);
146 	vsnprintf(name, len, fmt, args);
147 	kobj->name = name;
148 
149 	/* free old string */
150 	kfree(old);
151 
152 	/* filter new string */
153 	for (; *name != '\0'; name++)
154 		if (*name == '/')
155 			*name = '!';
156 	return (0);
157 }
158 
159 int
160 kobject_set_name(struct kobject *kobj, const char *fmt, ...)
161 {
162 	va_list args;
163 	int error;
164 
165 	va_start(args, fmt);
166 	error = kobject_set_name_vargs(kobj, fmt, args);
167 	va_end(args);
168 
169 	return (error);
170 }
171 
172 static int
173 kobject_add_complete(struct kobject *kobj, struct kobject *parent)
174 {
175 	const struct kobj_type *t;
176 	int error;
177 
178 	kobj->parent = parent;
179 	error = sysfs_create_dir(kobj);
180 	if (error == 0 && kobj->ktype && kobj->ktype->default_attrs) {
181 		struct attribute **attr;
182 		t = kobj->ktype;
183 
184 		for (attr = t->default_attrs; *attr != NULL; attr++) {
185 			error = sysfs_create_file(kobj, *attr);
186 			if (error)
187 				break;
188 		}
189 		if (error)
190 			sysfs_remove_dir(kobj);
191 
192 	}
193 	return (error);
194 }
195 
196 int
197 kobject_add(struct kobject *kobj, struct kobject *parent, const char *fmt, ...)
198 {
199 	va_list args;
200 	int error;
201 
202 	va_start(args, fmt);
203 	error = kobject_set_name_vargs(kobj, fmt, args);
204 	va_end(args);
205 	if (error)
206 		return (error);
207 
208 	return kobject_add_complete(kobj, parent);
209 }
210 
211 void
212 linux_kobject_release(struct kref *kref)
213 {
214 	struct kobject *kobj;
215 	char *name;
216 
217 	kobj = container_of(kref, struct kobject, kref);
218 	sysfs_remove_dir(kobj);
219 	name = kobj->name;
220 	if (kobj->ktype && kobj->ktype->release)
221 		kobj->ktype->release(kobj);
222 	kfree(name);
223 }
224 
225 static void
226 linux_kobject_kfree(struct kobject *kobj)
227 {
228 	kfree(kobj);
229 }
230 
231 static void
232 linux_kobject_kfree_name(struct kobject *kobj)
233 {
234 	if (kobj) {
235 		kfree(kobj->name);
236 	}
237 }
238 
239 const struct kobj_type linux_kfree_type = {
240 	.release = linux_kobject_kfree
241 };
242 
243 static void
244 linux_device_release(struct device *dev)
245 {
246 	pr_debug("linux_device_release: %s\n", dev_name(dev));
247 	kfree(dev);
248 }
249 
250 static ssize_t
251 linux_class_show(struct kobject *kobj, struct attribute *attr, char *buf)
252 {
253 	struct class_attribute *dattr;
254 	ssize_t error;
255 
256 	dattr = container_of(attr, struct class_attribute, attr);
257 	error = -EIO;
258 	if (dattr->show)
259 		error = dattr->show(container_of(kobj, struct class, kobj),
260 		    dattr, buf);
261 	return (error);
262 }
263 
264 static ssize_t
265 linux_class_store(struct kobject *kobj, struct attribute *attr, const char *buf,
266     size_t count)
267 {
268 	struct class_attribute *dattr;
269 	ssize_t error;
270 
271 	dattr = container_of(attr, struct class_attribute, attr);
272 	error = -EIO;
273 	if (dattr->store)
274 		error = dattr->store(container_of(kobj, struct class, kobj),
275 		    dattr, buf, count);
276 	return (error);
277 }
278 
279 static void
280 linux_class_release(struct kobject *kobj)
281 {
282 	struct class *class;
283 
284 	class = container_of(kobj, struct class, kobj);
285 	if (class->class_release)
286 		class->class_release(class);
287 }
288 
289 static const struct sysfs_ops linux_class_sysfs = {
290 	.show  = linux_class_show,
291 	.store = linux_class_store,
292 };
293 
294 const struct kobj_type linux_class_ktype = {
295 	.release = linux_class_release,
296 	.sysfs_ops = &linux_class_sysfs
297 };
298 
299 static void
300 linux_dev_release(struct kobject *kobj)
301 {
302 	struct device *dev;
303 
304 	dev = container_of(kobj, struct device, kobj);
305 	/* This is the precedence defined by linux. */
306 	if (dev->release)
307 		dev->release(dev);
308 	else if (dev->class && dev->class->dev_release)
309 		dev->class->dev_release(dev);
310 }
311 
312 static ssize_t
313 linux_dev_show(struct kobject *kobj, struct attribute *attr, char *buf)
314 {
315 	struct device_attribute *dattr;
316 	ssize_t error;
317 
318 	dattr = container_of(attr, struct device_attribute, attr);
319 	error = -EIO;
320 	if (dattr->show)
321 		error = dattr->show(container_of(kobj, struct device, kobj),
322 		    dattr, buf);
323 	return (error);
324 }
325 
326 static ssize_t
327 linux_dev_store(struct kobject *kobj, struct attribute *attr, const char *buf,
328     size_t count)
329 {
330 	struct device_attribute *dattr;
331 	ssize_t error;
332 
333 	dattr = container_of(attr, struct device_attribute, attr);
334 	error = -EIO;
335 	if (dattr->store)
336 		error = dattr->store(container_of(kobj, struct device, kobj),
337 		    dattr, buf, count);
338 	return (error);
339 }
340 
341 static const struct sysfs_ops linux_dev_sysfs = {
342 	.show  = linux_dev_show,
343 	.store = linux_dev_store,
344 };
345 
346 const struct kobj_type linux_dev_ktype = {
347 	.release = linux_dev_release,
348 	.sysfs_ops = &linux_dev_sysfs
349 };
350 
351 struct device *
352 device_create(struct class *class, struct device *parent, dev_t devt,
353     void *drvdata, const char *fmt, ...)
354 {
355 	struct device *dev;
356 	va_list args;
357 
358 	dev = kzalloc(sizeof(*dev), M_WAITOK);
359 	dev->parent = parent;
360 	dev->class = class;
361 	dev->devt = devt;
362 	dev->driver_data = drvdata;
363 	dev->release = linux_device_release;
364 	va_start(args, fmt);
365 	kobject_set_name_vargs(&dev->kobj, fmt, args);
366 	va_end(args);
367 	device_register(dev);
368 
369 	return (dev);
370 }
371 
372 int
373 kobject_init_and_add(struct kobject *kobj, const struct kobj_type *ktype,
374     struct kobject *parent, const char *fmt, ...)
375 {
376 	va_list args;
377 	int error;
378 
379 	kobject_init(kobj, ktype);
380 	kobj->ktype = ktype;
381 	kobj->parent = parent;
382 	kobj->name = NULL;
383 
384 	va_start(args, fmt);
385 	error = kobject_set_name_vargs(kobj, fmt, args);
386 	va_end(args);
387 	if (error)
388 		return (error);
389 	return kobject_add_complete(kobj, parent);
390 }
391 
392 static void
393 linux_file_dtor(void *cdp)
394 {
395 	struct linux_file *filp;
396 
397 	linux_set_current(curthread);
398 	filp = cdp;
399 	filp->f_op->release(filp->f_vnode, filp);
400 	vdrop(filp->f_vnode);
401 	kfree(filp);
402 }
403 
404 static int
405 linux_cdev_pager_populate(vm_object_t vm_obj, vm_pindex_t pidx, int fault_type,
406     vm_prot_t max_prot, vm_pindex_t *first, vm_pindex_t *last)
407 {
408 	struct vm_area_struct *vmap;
409 	struct vm_fault vmf;
410 	int err;
411 
412 	linux_set_current(curthread);
413 
414 	/* get VM area structure */
415 	vmap = linux_cdev_handle_find(vm_obj->handle);
416 	MPASS(vmap != NULL);
417 	MPASS(vmap->vm_private_data == vm_obj->handle);
418 
419 	/* fill out VM fault structure */
420 	vmf.virtual_address = (void *)((uintptr_t)pidx << PAGE_SHIFT);
421 	vmf.flags = (fault_type & VM_PROT_WRITE) ? FAULT_FLAG_WRITE : 0;
422 	vmf.pgoff = 0;
423 	vmf.page = NULL;
424 
425 	VM_OBJECT_WUNLOCK(vm_obj);
426 
427 	down_write(&vmap->vm_mm->mmap_sem);
428 	if (unlikely(vmap->vm_ops == NULL)) {
429 		err = VM_FAULT_SIGBUS;
430 	} else {
431 		vmap->vm_pfn_count = 0;
432 		vmap->vm_pfn_pcount = &vmap->vm_pfn_count;
433 		vmap->vm_obj = vm_obj;
434 
435 		err = vmap->vm_ops->fault(vmap, &vmf);
436 
437 		while (vmap->vm_pfn_count == 0 && err == VM_FAULT_NOPAGE) {
438 			kern_yield(PRI_USER);
439 			err = vmap->vm_ops->fault(vmap, &vmf);
440 		}
441 	}
442 
443 	/* translate return code */
444 	switch (err) {
445 	case VM_FAULT_OOM:
446 		err = VM_PAGER_AGAIN;
447 		break;
448 	case VM_FAULT_SIGBUS:
449 		err = VM_PAGER_BAD;
450 		break;
451 	case VM_FAULT_NOPAGE:
452 		/*
453 		 * By contract the fault handler will return having
454 		 * busied all the pages itself. If pidx is already
455 		 * found in the object, it will simply xbusy the first
456 		 * page and return with vm_pfn_count set to 1.
457 		 */
458 		*first = vmap->vm_pfn_first;
459 		*last = *first + vmap->vm_pfn_count - 1;
460 		err = VM_PAGER_OK;
461 		break;
462 	default:
463 		err = VM_PAGER_ERROR;
464 		break;
465 	}
466 	up_write(&vmap->vm_mm->mmap_sem);
467 	VM_OBJECT_WLOCK(vm_obj);
468 	return (err);
469 }
470 
471 static struct rwlock linux_vma_lock;
472 static TAILQ_HEAD(, vm_area_struct) linux_vma_head =
473     TAILQ_HEAD_INITIALIZER(linux_vma_head);
474 
475 static struct vm_area_struct *
476 linux_cdev_handle_insert(void *handle, struct vm_area_struct *vmap)
477 {
478 	struct vm_area_struct *ptr;
479 
480 	rw_wlock(&linux_vma_lock);
481 	TAILQ_FOREACH(ptr, &linux_vma_head, vm_entry) {
482 		if (ptr->vm_private_data == handle) {
483 			rw_wunlock(&linux_vma_lock);
484 			kfree(vmap);
485 			return (NULL);
486 		}
487 	}
488 	/*
489 	 * The same VM object might be shared by multiple processes
490 	 * and the mm_struct is usually freed when a process exits.
491 	 *
492 	 * The atomic reference below makes sure the mm_struct is
493 	 * available as long as the vmap is in the linux_vma_head.
494 	 */
495 	if (atomic_inc_not_zero(&vmap->vm_mm->mm_users) == 0)
496 		panic("linuxkpi: mm_users is zero\n");
497 
498 	TAILQ_INSERT_TAIL(&linux_vma_head, vmap, vm_entry);
499 	rw_wunlock(&linux_vma_lock);
500 	return (vmap);
501 }
502 
503 static void
504 linux_cdev_handle_remove(struct vm_area_struct *vmap)
505 {
506 	if (vmap == NULL)
507 		return;
508 
509 	rw_wlock(&linux_vma_lock);
510 	TAILQ_REMOVE(&linux_vma_head, vmap, vm_entry);
511 	rw_wunlock(&linux_vma_lock);
512 
513 	/* Drop reference on mm_struct */
514 	mmput(vmap->vm_mm);
515 	kfree(vmap);
516 }
517 
518 static struct vm_area_struct *
519 linux_cdev_handle_find(void *handle)
520 {
521 	struct vm_area_struct *vmap;
522 
523 	rw_rlock(&linux_vma_lock);
524 	TAILQ_FOREACH(vmap, &linux_vma_head, vm_entry) {
525 		if (vmap->vm_private_data == handle)
526 			break;
527 	}
528 	rw_runlock(&linux_vma_lock);
529 	return (vmap);
530 }
531 
532 static int
533 linux_cdev_pager_ctor(void *handle, vm_ooffset_t size, vm_prot_t prot,
534 		      vm_ooffset_t foff, struct ucred *cred, u_short *color)
535 {
536 	const struct vm_operations_struct *vm_ops;
537 	struct vm_area_struct *vmap;
538 
539 	vmap = linux_cdev_handle_find(handle);
540 	MPASS(vmap != NULL);
541 
542 	*color = 0;
543 
544 	down_write(&vmap->vm_mm->mmap_sem);
545 	vm_ops = vmap->vm_ops;
546 	if (likely(vm_ops != NULL))
547 		vm_ops->open(vmap);
548 	up_write(&vmap->vm_mm->mmap_sem);
549 
550 	return (0);
551 }
552 
553 static void
554 linux_cdev_pager_dtor(void *handle)
555 {
556 	const struct vm_operations_struct *vm_ops;
557 	struct vm_area_struct *vmap;
558 
559 	vmap = linux_cdev_handle_find(handle);
560 	MPASS(vmap != NULL);
561 
562 	down_write(&vmap->vm_mm->mmap_sem);
563 	vm_ops = vmap->vm_ops;
564 	if (likely(vm_ops != NULL))
565 		vm_ops->close(vmap);
566 	up_write(&vmap->vm_mm->mmap_sem);
567 
568 	linux_cdev_handle_remove(vmap);
569 }
570 
571 static struct cdev_pager_ops linux_cdev_pager_ops = {
572 	.cdev_pg_populate	= linux_cdev_pager_populate,
573 	.cdev_pg_ctor	= linux_cdev_pager_ctor,
574 	.cdev_pg_dtor	= linux_cdev_pager_dtor
575 };
576 
577 static int
578 linux_dev_open(struct cdev *dev, int oflags, int devtype, struct thread *td)
579 {
580 	struct linux_cdev *ldev;
581 	struct linux_file *filp;
582 	struct file *file;
583 	int error;
584 
585 	file = td->td_fpop;
586 	ldev = dev->si_drv1;
587 	if (ldev == NULL)
588 		return (ENODEV);
589 	filp = kzalloc(sizeof(*filp), GFP_KERNEL);
590 	filp->f_dentry = &filp->f_dentry_store;
591 	filp->f_op = ldev->ops;
592 	filp->f_flags = file->f_flag;
593 	vhold(file->f_vnode);
594 	filp->f_vnode = file->f_vnode;
595 	linux_set_current(td);
596 	if (filp->f_op->open) {
597 		error = -filp->f_op->open(file->f_vnode, filp);
598 		if (error) {
599 			kfree(filp);
600 			goto done;
601 		}
602 	}
603 	error = devfs_set_cdevpriv(filp, linux_file_dtor);
604 	if (error) {
605 		filp->f_op->release(file->f_vnode, filp);
606 		kfree(filp);
607 	}
608 done:
609 	return (error);
610 }
611 
612 static int
613 linux_dev_close(struct cdev *dev, int fflag, int devtype, struct thread *td)
614 {
615 	struct linux_file *filp;
616 	struct file *file;
617 	int error;
618 
619 	file = td->td_fpop;
620 	if (dev->si_drv1 == NULL)
621 		return (0);
622 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
623 		return (error);
624 	filp->f_flags = file->f_flag;
625         devfs_clear_cdevpriv();
626 
627 
628 	return (0);
629 }
630 
631 #define	LINUX_IOCTL_MIN_PTR 0x10000UL
632 #define	LINUX_IOCTL_MAX_PTR (LINUX_IOCTL_MIN_PTR + IOCPARM_MAX)
633 
634 static inline int
635 linux_remap_address(void **uaddr, size_t len)
636 {
637 	uintptr_t uaddr_val = (uintptr_t)(*uaddr);
638 
639 	if (unlikely(uaddr_val >= LINUX_IOCTL_MIN_PTR &&
640 	    uaddr_val < LINUX_IOCTL_MAX_PTR)) {
641 		struct task_struct *pts = current;
642 		if (pts == NULL) {
643 			*uaddr = NULL;
644 			return (1);
645 		}
646 
647 		/* compute data offset */
648 		uaddr_val -= LINUX_IOCTL_MIN_PTR;
649 
650 		/* check that length is within bounds */
651 		if ((len > IOCPARM_MAX) ||
652 		    (uaddr_val + len) > pts->bsd_ioctl_len) {
653 			*uaddr = NULL;
654 			return (1);
655 		}
656 
657 		/* re-add kernel buffer address */
658 		uaddr_val += (uintptr_t)pts->bsd_ioctl_data;
659 
660 		/* update address location */
661 		*uaddr = (void *)uaddr_val;
662 		return (1);
663 	}
664 	return (0);
665 }
666 
667 int
668 linux_copyin(const void *uaddr, void *kaddr, size_t len)
669 {
670 	if (linux_remap_address(__DECONST(void **, &uaddr), len)) {
671 		if (uaddr == NULL)
672 			return (-EFAULT);
673 		memcpy(kaddr, uaddr, len);
674 		return (0);
675 	}
676 	return (-copyin(uaddr, kaddr, len));
677 }
678 
679 int
680 linux_copyout(const void *kaddr, void *uaddr, size_t len)
681 {
682 	if (linux_remap_address(&uaddr, len)) {
683 		if (uaddr == NULL)
684 			return (-EFAULT);
685 		memcpy(uaddr, kaddr, len);
686 		return (0);
687 	}
688 	return (-copyout(kaddr, uaddr, len));
689 }
690 
691 size_t
692 linux_clear_user(void *_uaddr, size_t _len)
693 {
694 	uint8_t *uaddr = _uaddr;
695 	size_t len = _len;
696 
697 	/* make sure uaddr is aligned before going into the fast loop */
698 	while (((uintptr_t)uaddr & 7) != 0 && len > 7) {
699 		if (subyte(uaddr, 0))
700 			return (_len);
701 		uaddr++;
702 		len--;
703 	}
704 
705 	/* zero 8 bytes at a time */
706 	while (len > 7) {
707 #ifdef __LP64__
708 		if (suword64(uaddr, 0))
709 			return (_len);
710 #else
711 		if (suword32(uaddr, 0))
712 			return (_len);
713 		if (suword32(uaddr + 4, 0))
714 			return (_len);
715 #endif
716 		uaddr += 8;
717 		len -= 8;
718 	}
719 
720 	/* zero fill end, if any */
721 	while (len > 0) {
722 		if (subyte(uaddr, 0))
723 			return (_len);
724 		uaddr++;
725 		len--;
726 	}
727 	return (0);
728 }
729 
730 int
731 linux_access_ok(int rw, const void *uaddr, size_t len)
732 {
733 	uintptr_t saddr;
734 	uintptr_t eaddr;
735 
736 	/* get start and end address */
737 	saddr = (uintptr_t)uaddr;
738 	eaddr = (uintptr_t)uaddr + len;
739 
740 	/* verify addresses are valid for userspace */
741 	return ((saddr == eaddr) ||
742 	    (eaddr > saddr && eaddr <= VM_MAXUSER_ADDRESS));
743 }
744 
745 static int
746 linux_dev_ioctl(struct cdev *dev, u_long cmd, caddr_t data, int fflag,
747     struct thread *td)
748 {
749 	struct linux_file *filp;
750 	struct file *file;
751 	unsigned size;
752 	int error;
753 
754 	file = td->td_fpop;
755 	if (dev->si_drv1 == NULL)
756 		return (ENXIO);
757 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
758 		return (error);
759 	filp->f_flags = file->f_flag;
760 
761 	linux_set_current(td);
762 	size = IOCPARM_LEN(cmd);
763 	/* refer to logic in sys_ioctl() */
764 	if (size > 0) {
765 		/*
766 		 * Setup hint for linux_copyin() and linux_copyout().
767 		 *
768 		 * Background: Linux code expects a user-space address
769 		 * while FreeBSD supplies a kernel-space address.
770 		 */
771 		current->bsd_ioctl_data = data;
772 		current->bsd_ioctl_len = size;
773 		data = (void *)LINUX_IOCTL_MIN_PTR;
774 	} else {
775 		/* fetch user-space pointer */
776 		data = *(void **)data;
777 	}
778 	if (filp->f_op->unlocked_ioctl)
779 		error = -filp->f_op->unlocked_ioctl(filp, cmd, (u_long)data);
780 	else
781 		error = ENOTTY;
782 	if (size > 0) {
783 		current->bsd_ioctl_data = NULL;
784 		current->bsd_ioctl_len = 0;
785 	}
786 
787 	return (error);
788 }
789 
790 static int
791 linux_dev_read(struct cdev *dev, struct uio *uio, int ioflag)
792 {
793 	struct linux_file *filp;
794 	struct thread *td;
795 	struct file *file;
796 	ssize_t bytes;
797 	int error;
798 
799 	td = curthread;
800 	file = td->td_fpop;
801 	if (dev->si_drv1 == NULL)
802 		return (ENXIO);
803 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
804 		return (error);
805 	filp->f_flags = file->f_flag;
806 	/* XXX no support for I/O vectors currently */
807 	if (uio->uio_iovcnt != 1)
808 		return (EOPNOTSUPP);
809 	linux_set_current(td);
810 	if (filp->f_op->read) {
811 		bytes = filp->f_op->read(filp, uio->uio_iov->iov_base,
812 		    uio->uio_iov->iov_len, &uio->uio_offset);
813 		if (bytes >= 0) {
814 			uio->uio_iov->iov_base =
815 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
816 			uio->uio_iov->iov_len -= bytes;
817 			uio->uio_resid -= bytes;
818 		} else
819 			error = -bytes;
820 	} else
821 		error = ENXIO;
822 
823 	return (error);
824 }
825 
826 static int
827 linux_dev_write(struct cdev *dev, struct uio *uio, int ioflag)
828 {
829 	struct linux_file *filp;
830 	struct thread *td;
831 	struct file *file;
832 	ssize_t bytes;
833 	int error;
834 
835 	td = curthread;
836 	file = td->td_fpop;
837 	if (dev->si_drv1 == NULL)
838 		return (ENXIO);
839 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
840 		return (error);
841 	filp->f_flags = file->f_flag;
842 	/* XXX no support for I/O vectors currently */
843 	if (uio->uio_iovcnt != 1)
844 		return (EOPNOTSUPP);
845 	linux_set_current(td);
846 	if (filp->f_op->write) {
847 		bytes = filp->f_op->write(filp, uio->uio_iov->iov_base,
848 		    uio->uio_iov->iov_len, &uio->uio_offset);
849 		if (bytes >= 0) {
850 			uio->uio_iov->iov_base =
851 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
852 			uio->uio_iov->iov_len -= bytes;
853 			uio->uio_resid -= bytes;
854 		} else
855 			error = -bytes;
856 	} else
857 		error = ENXIO;
858 
859 	return (error);
860 }
861 
862 static int
863 linux_dev_poll(struct cdev *dev, int events, struct thread *td)
864 {
865 	struct linux_file *filp;
866 	struct file *file;
867 	int revents;
868 
869 	if (dev->si_drv1 == NULL)
870 		goto error;
871 	if (devfs_get_cdevpriv((void **)&filp) != 0)
872 		goto error;
873 
874 	file = td->td_fpop;
875 	filp->f_flags = file->f_flag;
876 	linux_set_current(td);
877 	if (filp->f_op->poll)
878 		revents = filp->f_op->poll(filp, NULL) & events;
879 	else
880 		revents = 0;
881 
882 	return (revents);
883 error:
884 	return (events & (POLLHUP|POLLIN|POLLRDNORM|POLLOUT|POLLWRNORM));
885 }
886 
887 static int
888 linux_dev_mmap_single(struct cdev *dev, vm_ooffset_t *offset,
889     vm_size_t size, struct vm_object **object, int nprot)
890 {
891 	struct vm_area_struct *vmap;
892 	struct linux_file *filp;
893 	struct thread *td;
894 	struct file *file;
895 	vm_memattr_t attr;
896 	int error;
897 
898 	td = curthread;
899 	file = td->td_fpop;
900 	if (dev->si_drv1 == NULL)
901 		return (ENODEV);
902 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
903 		return (error);
904 	filp->f_flags = file->f_flag;
905 
906 	if (filp->f_op->mmap == NULL)
907 		return (ENODEV);
908 
909 	linux_set_current(td);
910 
911 	vmap = kzalloc(sizeof(*vmap), GFP_KERNEL);
912 	vmap->vm_start = 0;
913 	vmap->vm_end = size;
914 	vmap->vm_pgoff = *offset / PAGE_SIZE;
915 	vmap->vm_pfn = 0;
916 	vmap->vm_flags = vmap->vm_page_prot = nprot;
917 	vmap->vm_ops = NULL;
918 	vmap->vm_file = filp;
919 	vmap->vm_mm = current->mm;
920 
921 	if (unlikely(down_write_killable(&vmap->vm_mm->mmap_sem))) {
922 		error = EINTR;
923 	} else {
924 		error = -filp->f_op->mmap(filp, vmap);
925 		up_write(&vmap->vm_mm->mmap_sem);
926 	}
927 
928 	if (error != 0) {
929 		kfree(vmap);
930 		return (error);
931 	}
932 
933 	attr = pgprot2cachemode(vmap->vm_page_prot);
934 
935 	if (vmap->vm_ops != NULL) {
936 		void *vm_private_data;
937 
938 		if (vmap->vm_ops->fault == NULL ||
939 		    vmap->vm_ops->open == NULL ||
940 		    vmap->vm_ops->close == NULL ||
941 		    vmap->vm_private_data == NULL) {
942 			kfree(vmap);
943 			return (EINVAL);
944 		}
945 
946 		vm_private_data = vmap->vm_private_data;
947 
948 		vmap = linux_cdev_handle_insert(vm_private_data, vmap);
949 
950 		*object = cdev_pager_allocate(vm_private_data, OBJT_MGTDEVICE,
951 		    &linux_cdev_pager_ops, size, nprot, *offset, curthread->td_ucred);
952 
953 		if (*object == NULL) {
954 			linux_cdev_handle_remove(vmap);
955 			return (EINVAL);
956 		}
957 	} else {
958 		struct sglist *sg;
959 
960 		sg = sglist_alloc(1, M_WAITOK);
961 		sglist_append_phys(sg, (vm_paddr_t)vmap->vm_pfn << PAGE_SHIFT, vmap->vm_len);
962 
963 		*object = vm_pager_allocate(OBJT_SG, sg, vmap->vm_len,
964 		    nprot, 0, curthread->td_ucred);
965 
966 		kfree(vmap);
967 
968 		if (*object == NULL) {
969 			sglist_free(sg);
970 			return (EINVAL);
971 		}
972 	}
973 
974 	if (attr != VM_MEMATTR_DEFAULT) {
975 		VM_OBJECT_WLOCK(*object);
976 		vm_object_set_memattr(*object, attr);
977 		VM_OBJECT_WUNLOCK(*object);
978 	}
979 	*offset = 0;
980 	return (0);
981 }
982 
983 struct cdevsw linuxcdevsw = {
984 	.d_version = D_VERSION,
985 	.d_flags = D_TRACKCLOSE,
986 	.d_open = linux_dev_open,
987 	.d_close = linux_dev_close,
988 	.d_read = linux_dev_read,
989 	.d_write = linux_dev_write,
990 	.d_ioctl = linux_dev_ioctl,
991 	.d_mmap_single = linux_dev_mmap_single,
992 	.d_poll = linux_dev_poll,
993 };
994 
995 static int
996 linux_file_read(struct file *file, struct uio *uio, struct ucred *active_cred,
997     int flags, struct thread *td)
998 {
999 	struct linux_file *filp;
1000 	ssize_t bytes;
1001 	int error;
1002 
1003 	error = 0;
1004 	filp = (struct linux_file *)file->f_data;
1005 	filp->f_flags = file->f_flag;
1006 	/* XXX no support for I/O vectors currently */
1007 	if (uio->uio_iovcnt != 1)
1008 		return (EOPNOTSUPP);
1009 	linux_set_current(td);
1010 	if (filp->f_op->read) {
1011 		bytes = filp->f_op->read(filp, uio->uio_iov->iov_base,
1012 		    uio->uio_iov->iov_len, &uio->uio_offset);
1013 		if (bytes >= 0) {
1014 			uio->uio_iov->iov_base =
1015 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
1016 			uio->uio_iov->iov_len -= bytes;
1017 			uio->uio_resid -= bytes;
1018 		} else
1019 			error = -bytes;
1020 	} else
1021 		error = ENXIO;
1022 
1023 	return (error);
1024 }
1025 
1026 static int
1027 linux_file_poll(struct file *file, int events, struct ucred *active_cred,
1028     struct thread *td)
1029 {
1030 	struct linux_file *filp;
1031 	int revents;
1032 
1033 	filp = (struct linux_file *)file->f_data;
1034 	filp->f_flags = file->f_flag;
1035 	linux_set_current(td);
1036 	if (filp->f_op->poll)
1037 		revents = filp->f_op->poll(filp, NULL) & events;
1038 	else
1039 		revents = 0;
1040 
1041 	return (revents);
1042 }
1043 
1044 static int
1045 linux_file_close(struct file *file, struct thread *td)
1046 {
1047 	struct linux_file *filp;
1048 	int error;
1049 
1050 	filp = (struct linux_file *)file->f_data;
1051 	filp->f_flags = file->f_flag;
1052 	linux_set_current(td);
1053 	error = -filp->f_op->release(NULL, filp);
1054 	funsetown(&filp->f_sigio);
1055 	kfree(filp);
1056 
1057 	return (error);
1058 }
1059 
1060 static int
1061 linux_file_ioctl(struct file *fp, u_long cmd, void *data, struct ucred *cred,
1062     struct thread *td)
1063 {
1064 	struct linux_file *filp;
1065 	int error;
1066 
1067 	filp = (struct linux_file *)fp->f_data;
1068 	filp->f_flags = fp->f_flag;
1069 	error = 0;
1070 
1071 	linux_set_current(td);
1072 	switch (cmd) {
1073 	case FIONBIO:
1074 		break;
1075 	case FIOASYNC:
1076 		if (filp->f_op->fasync == NULL)
1077 			break;
1078 		error = filp->f_op->fasync(0, filp, fp->f_flag & FASYNC);
1079 		break;
1080 	case FIOSETOWN:
1081 		error = fsetown(*(int *)data, &filp->f_sigio);
1082 		if (error == 0)
1083 			error = filp->f_op->fasync(0, filp,
1084 			    fp->f_flag & FASYNC);
1085 		break;
1086 	case FIOGETOWN:
1087 		*(int *)data = fgetown(&filp->f_sigio);
1088 		break;
1089 	default:
1090 		error = ENOTTY;
1091 		break;
1092 	}
1093 	return (error);
1094 }
1095 
1096 static int
1097 linux_file_stat(struct file *fp, struct stat *sb, struct ucred *active_cred,
1098     struct thread *td)
1099 {
1100 
1101 	return (EOPNOTSUPP);
1102 }
1103 
1104 static int
1105 linux_file_fill_kinfo(struct file *fp, struct kinfo_file *kif,
1106     struct filedesc *fdp)
1107 {
1108 
1109 	return (0);
1110 }
1111 
1112 struct fileops linuxfileops = {
1113 	.fo_read = linux_file_read,
1114 	.fo_write = invfo_rdwr,
1115 	.fo_truncate = invfo_truncate,
1116 	.fo_kqfilter = invfo_kqfilter,
1117 	.fo_stat = linux_file_stat,
1118 	.fo_fill_kinfo = linux_file_fill_kinfo,
1119 	.fo_poll = linux_file_poll,
1120 	.fo_close = linux_file_close,
1121 	.fo_ioctl = linux_file_ioctl,
1122 	.fo_chmod = invfo_chmod,
1123 	.fo_chown = invfo_chown,
1124 	.fo_sendfile = invfo_sendfile,
1125 };
1126 
1127 /*
1128  * Hash of vmmap addresses.  This is infrequently accessed and does not
1129  * need to be particularly large.  This is done because we must store the
1130  * caller's idea of the map size to properly unmap.
1131  */
1132 struct vmmap {
1133 	LIST_ENTRY(vmmap)	vm_next;
1134 	void 			*vm_addr;
1135 	unsigned long		vm_size;
1136 };
1137 
1138 struct vmmaphd {
1139 	struct vmmap *lh_first;
1140 };
1141 #define	VMMAP_HASH_SIZE	64
1142 #define	VMMAP_HASH_MASK	(VMMAP_HASH_SIZE - 1)
1143 #define	VM_HASH(addr)	((uintptr_t)(addr) >> PAGE_SHIFT) & VMMAP_HASH_MASK
1144 static struct vmmaphd vmmaphead[VMMAP_HASH_SIZE];
1145 static struct mtx vmmaplock;
1146 
1147 static void
1148 vmmap_add(void *addr, unsigned long size)
1149 {
1150 	struct vmmap *vmmap;
1151 
1152 	vmmap = kmalloc(sizeof(*vmmap), GFP_KERNEL);
1153 	mtx_lock(&vmmaplock);
1154 	vmmap->vm_size = size;
1155 	vmmap->vm_addr = addr;
1156 	LIST_INSERT_HEAD(&vmmaphead[VM_HASH(addr)], vmmap, vm_next);
1157 	mtx_unlock(&vmmaplock);
1158 }
1159 
1160 static struct vmmap *
1161 vmmap_remove(void *addr)
1162 {
1163 	struct vmmap *vmmap;
1164 
1165 	mtx_lock(&vmmaplock);
1166 	LIST_FOREACH(vmmap, &vmmaphead[VM_HASH(addr)], vm_next)
1167 		if (vmmap->vm_addr == addr)
1168 			break;
1169 	if (vmmap)
1170 		LIST_REMOVE(vmmap, vm_next);
1171 	mtx_unlock(&vmmaplock);
1172 
1173 	return (vmmap);
1174 }
1175 
1176 #if defined(__i386__) || defined(__amd64__)
1177 void *
1178 _ioremap_attr(vm_paddr_t phys_addr, unsigned long size, int attr)
1179 {
1180 	void *addr;
1181 
1182 	addr = pmap_mapdev_attr(phys_addr, size, attr);
1183 	if (addr == NULL)
1184 		return (NULL);
1185 	vmmap_add(addr, size);
1186 
1187 	return (addr);
1188 }
1189 #endif
1190 
1191 void
1192 iounmap(void *addr)
1193 {
1194 	struct vmmap *vmmap;
1195 
1196 	vmmap = vmmap_remove(addr);
1197 	if (vmmap == NULL)
1198 		return;
1199 #if defined(__i386__) || defined(__amd64__)
1200 	pmap_unmapdev((vm_offset_t)addr, vmmap->vm_size);
1201 #endif
1202 	kfree(vmmap);
1203 }
1204 
1205 
1206 void *
1207 vmap(struct page **pages, unsigned int count, unsigned long flags, int prot)
1208 {
1209 	vm_offset_t off;
1210 	size_t size;
1211 
1212 	size = count * PAGE_SIZE;
1213 	off = kva_alloc(size);
1214 	if (off == 0)
1215 		return (NULL);
1216 	vmmap_add((void *)off, size);
1217 	pmap_qenter(off, pages, count);
1218 
1219 	return ((void *)off);
1220 }
1221 
1222 void
1223 vunmap(void *addr)
1224 {
1225 	struct vmmap *vmmap;
1226 
1227 	vmmap = vmmap_remove(addr);
1228 	if (vmmap == NULL)
1229 		return;
1230 	pmap_qremove((vm_offset_t)addr, vmmap->vm_size / PAGE_SIZE);
1231 	kva_free((vm_offset_t)addr, vmmap->vm_size);
1232 	kfree(vmmap);
1233 }
1234 
1235 char *
1236 kvasprintf(gfp_t gfp, const char *fmt, va_list ap)
1237 {
1238 	unsigned int len;
1239 	char *p;
1240 	va_list aq;
1241 
1242 	va_copy(aq, ap);
1243 	len = vsnprintf(NULL, 0, fmt, aq);
1244 	va_end(aq);
1245 
1246 	p = kmalloc(len + 1, gfp);
1247 	if (p != NULL)
1248 		vsnprintf(p, len + 1, fmt, ap);
1249 
1250 	return (p);
1251 }
1252 
1253 char *
1254 kasprintf(gfp_t gfp, const char *fmt, ...)
1255 {
1256 	va_list ap;
1257 	char *p;
1258 
1259 	va_start(ap, fmt);
1260 	p = kvasprintf(gfp, fmt, ap);
1261 	va_end(ap);
1262 
1263 	return (p);
1264 }
1265 
1266 static void
1267 linux_timer_callback_wrapper(void *context)
1268 {
1269 	struct timer_list *timer;
1270 
1271 	linux_set_current(curthread);
1272 
1273 	timer = context;
1274 	timer->function(timer->data);
1275 }
1276 
1277 void
1278 mod_timer(struct timer_list *timer, unsigned long expires)
1279 {
1280 
1281 	timer->expires = expires;
1282 	callout_reset(&timer->timer_callout,
1283 	    linux_timer_jiffies_until(expires),
1284 	    &linux_timer_callback_wrapper, timer);
1285 }
1286 
1287 void
1288 add_timer(struct timer_list *timer)
1289 {
1290 
1291 	callout_reset(&timer->timer_callout,
1292 	    linux_timer_jiffies_until(timer->expires),
1293 	    &linux_timer_callback_wrapper, timer);
1294 }
1295 
1296 void
1297 add_timer_on(struct timer_list *timer, int cpu)
1298 {
1299 
1300 	callout_reset_on(&timer->timer_callout,
1301 	    linux_timer_jiffies_until(timer->expires),
1302 	    &linux_timer_callback_wrapper, timer, cpu);
1303 }
1304 
1305 static void
1306 linux_timer_init(void *arg)
1307 {
1308 
1309 	/*
1310 	 * Compute an internal HZ value which can divide 2**32 to
1311 	 * avoid timer rounding problems when the tick value wraps
1312 	 * around 2**32:
1313 	 */
1314 	linux_timer_hz_mask = 1;
1315 	while (linux_timer_hz_mask < (unsigned long)hz)
1316 		linux_timer_hz_mask *= 2;
1317 	linux_timer_hz_mask--;
1318 }
1319 SYSINIT(linux_timer, SI_SUB_DRIVERS, SI_ORDER_FIRST, linux_timer_init, NULL);
1320 
1321 void
1322 linux_complete_common(struct completion *c, int all)
1323 {
1324 	int wakeup_swapper;
1325 
1326 	sleepq_lock(c);
1327 	c->done++;
1328 	if (all)
1329 		wakeup_swapper = sleepq_broadcast(c, SLEEPQ_SLEEP, 0, 0);
1330 	else
1331 		wakeup_swapper = sleepq_signal(c, SLEEPQ_SLEEP, 0, 0);
1332 	sleepq_release(c);
1333 	if (wakeup_swapper)
1334 		kick_proc0();
1335 }
1336 
1337 /*
1338  * Indefinite wait for done != 0 with or without signals.
1339  */
1340 long
1341 linux_wait_for_common(struct completion *c, int flags)
1342 {
1343 	long error;
1344 
1345 	if (SCHEDULER_STOPPED())
1346 		return (0);
1347 
1348 	DROP_GIANT();
1349 
1350 	if (flags != 0)
1351 		flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
1352 	else
1353 		flags = SLEEPQ_SLEEP;
1354 	error = 0;
1355 	for (;;) {
1356 		sleepq_lock(c);
1357 		if (c->done)
1358 			break;
1359 		sleepq_add(c, NULL, "completion", flags, 0);
1360 		if (flags & SLEEPQ_INTERRUPTIBLE) {
1361 			if (sleepq_wait_sig(c, 0) != 0) {
1362 				error = -ERESTARTSYS;
1363 				goto intr;
1364 			}
1365 		} else
1366 			sleepq_wait(c, 0);
1367 	}
1368 	c->done--;
1369 	sleepq_release(c);
1370 
1371 intr:
1372 	PICKUP_GIANT();
1373 
1374 	return (error);
1375 }
1376 
1377 /*
1378  * Time limited wait for done != 0 with or without signals.
1379  */
1380 long
1381 linux_wait_for_timeout_common(struct completion *c, long timeout, int flags)
1382 {
1383 	long end = jiffies + timeout, error;
1384 	int ret;
1385 
1386 	if (SCHEDULER_STOPPED())
1387 		return (0);
1388 
1389 	DROP_GIANT();
1390 
1391 	if (flags != 0)
1392 		flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
1393 	else
1394 		flags = SLEEPQ_SLEEP;
1395 
1396 	error = 0;
1397 	ret = 0;
1398 	for (;;) {
1399 		sleepq_lock(c);
1400 		if (c->done)
1401 			break;
1402 		sleepq_add(c, NULL, "completion", flags, 0);
1403 		sleepq_set_timeout(c, linux_timer_jiffies_until(end));
1404 		if (flags & SLEEPQ_INTERRUPTIBLE)
1405 			ret = sleepq_timedwait_sig(c, 0);
1406 		else
1407 			ret = sleepq_timedwait(c, 0);
1408 		if (ret != 0) {
1409 			/* check for timeout or signal */
1410 			if (ret == EWOULDBLOCK)
1411 				error = 0;
1412 			else
1413 				error = -ERESTARTSYS;
1414 			goto intr;
1415 		}
1416 	}
1417 	c->done--;
1418 	sleepq_release(c);
1419 
1420 intr:
1421 	PICKUP_GIANT();
1422 
1423 	/* return how many jiffies are left */
1424 	return (ret != 0 ? error : linux_timer_jiffies_until(end));
1425 }
1426 
1427 int
1428 linux_try_wait_for_completion(struct completion *c)
1429 {
1430 	int isdone;
1431 
1432 	isdone = 1;
1433 	sleepq_lock(c);
1434 	if (c->done)
1435 		c->done--;
1436 	else
1437 		isdone = 0;
1438 	sleepq_release(c);
1439 	return (isdone);
1440 }
1441 
1442 int
1443 linux_completion_done(struct completion *c)
1444 {
1445 	int isdone;
1446 
1447 	isdone = 1;
1448 	sleepq_lock(c);
1449 	if (c->done == 0)
1450 		isdone = 0;
1451 	sleepq_release(c);
1452 	return (isdone);
1453 }
1454 
1455 static void
1456 linux_cdev_release(struct kobject *kobj)
1457 {
1458 	struct linux_cdev *cdev;
1459 	struct kobject *parent;
1460 
1461 	cdev = container_of(kobj, struct linux_cdev, kobj);
1462 	parent = kobj->parent;
1463 	if (cdev->cdev)
1464 		destroy_dev(cdev->cdev);
1465 	kfree(cdev);
1466 	kobject_put(parent);
1467 }
1468 
1469 static void
1470 linux_cdev_static_release(struct kobject *kobj)
1471 {
1472 	struct linux_cdev *cdev;
1473 	struct kobject *parent;
1474 
1475 	cdev = container_of(kobj, struct linux_cdev, kobj);
1476 	parent = kobj->parent;
1477 	if (cdev->cdev)
1478 		destroy_dev(cdev->cdev);
1479 	kobject_put(parent);
1480 }
1481 
1482 const struct kobj_type linux_cdev_ktype = {
1483 	.release = linux_cdev_release,
1484 };
1485 
1486 const struct kobj_type linux_cdev_static_ktype = {
1487 	.release = linux_cdev_static_release,
1488 };
1489 
1490 static void
1491 linux_handle_ifnet_link_event(void *arg, struct ifnet *ifp, int linkstate)
1492 {
1493 	struct notifier_block *nb;
1494 
1495 	nb = arg;
1496 	if (linkstate == LINK_STATE_UP)
1497 		nb->notifier_call(nb, NETDEV_UP, ifp);
1498 	else
1499 		nb->notifier_call(nb, NETDEV_DOWN, ifp);
1500 }
1501 
1502 static void
1503 linux_handle_ifnet_arrival_event(void *arg, struct ifnet *ifp)
1504 {
1505 	struct notifier_block *nb;
1506 
1507 	nb = arg;
1508 	nb->notifier_call(nb, NETDEV_REGISTER, ifp);
1509 }
1510 
1511 static void
1512 linux_handle_ifnet_departure_event(void *arg, struct ifnet *ifp)
1513 {
1514 	struct notifier_block *nb;
1515 
1516 	nb = arg;
1517 	nb->notifier_call(nb, NETDEV_UNREGISTER, ifp);
1518 }
1519 
1520 static void
1521 linux_handle_iflladdr_event(void *arg, struct ifnet *ifp)
1522 {
1523 	struct notifier_block *nb;
1524 
1525 	nb = arg;
1526 	nb->notifier_call(nb, NETDEV_CHANGEADDR, ifp);
1527 }
1528 
1529 static void
1530 linux_handle_ifaddr_event(void *arg, struct ifnet *ifp)
1531 {
1532 	struct notifier_block *nb;
1533 
1534 	nb = arg;
1535 	nb->notifier_call(nb, NETDEV_CHANGEIFADDR, ifp);
1536 }
1537 
1538 int
1539 register_netdevice_notifier(struct notifier_block *nb)
1540 {
1541 
1542 	nb->tags[NETDEV_UP] = EVENTHANDLER_REGISTER(
1543 	    ifnet_link_event, linux_handle_ifnet_link_event, nb, 0);
1544 	nb->tags[NETDEV_REGISTER] = EVENTHANDLER_REGISTER(
1545 	    ifnet_arrival_event, linux_handle_ifnet_arrival_event, nb, 0);
1546 	nb->tags[NETDEV_UNREGISTER] = EVENTHANDLER_REGISTER(
1547 	    ifnet_departure_event, linux_handle_ifnet_departure_event, nb, 0);
1548 	nb->tags[NETDEV_CHANGEADDR] = EVENTHANDLER_REGISTER(
1549 	    iflladdr_event, linux_handle_iflladdr_event, nb, 0);
1550 
1551 	return (0);
1552 }
1553 
1554 int
1555 register_inetaddr_notifier(struct notifier_block *nb)
1556 {
1557 
1558         nb->tags[NETDEV_CHANGEIFADDR] = EVENTHANDLER_REGISTER(
1559             ifaddr_event, linux_handle_ifaddr_event, nb, 0);
1560         return (0);
1561 }
1562 
1563 int
1564 unregister_netdevice_notifier(struct notifier_block *nb)
1565 {
1566 
1567         EVENTHANDLER_DEREGISTER(ifnet_link_event,
1568 	    nb->tags[NETDEV_UP]);
1569         EVENTHANDLER_DEREGISTER(ifnet_arrival_event,
1570 	    nb->tags[NETDEV_REGISTER]);
1571         EVENTHANDLER_DEREGISTER(ifnet_departure_event,
1572 	    nb->tags[NETDEV_UNREGISTER]);
1573         EVENTHANDLER_DEREGISTER(iflladdr_event,
1574 	    nb->tags[NETDEV_CHANGEADDR]);
1575 
1576 	return (0);
1577 }
1578 
1579 int
1580 unregister_inetaddr_notifier(struct notifier_block *nb)
1581 {
1582 
1583         EVENTHANDLER_DEREGISTER(ifaddr_event,
1584             nb->tags[NETDEV_CHANGEIFADDR]);
1585 
1586         return (0);
1587 }
1588 
1589 struct list_sort_thunk {
1590 	int (*cmp)(void *, struct list_head *, struct list_head *);
1591 	void *priv;
1592 };
1593 
1594 static inline int
1595 linux_le_cmp(void *priv, const void *d1, const void *d2)
1596 {
1597 	struct list_head *le1, *le2;
1598 	struct list_sort_thunk *thunk;
1599 
1600 	thunk = priv;
1601 	le1 = *(__DECONST(struct list_head **, d1));
1602 	le2 = *(__DECONST(struct list_head **, d2));
1603 	return ((thunk->cmp)(thunk->priv, le1, le2));
1604 }
1605 
1606 void
1607 list_sort(void *priv, struct list_head *head, int (*cmp)(void *priv,
1608     struct list_head *a, struct list_head *b))
1609 {
1610 	struct list_sort_thunk thunk;
1611 	struct list_head **ar, *le;
1612 	size_t count, i;
1613 
1614 	count = 0;
1615 	list_for_each(le, head)
1616 		count++;
1617 	ar = malloc(sizeof(struct list_head *) * count, M_KMALLOC, M_WAITOK);
1618 	i = 0;
1619 	list_for_each(le, head)
1620 		ar[i++] = le;
1621 	thunk.cmp = cmp;
1622 	thunk.priv = priv;
1623 	qsort_r(ar, count, sizeof(struct list_head *), &thunk, linux_le_cmp);
1624 	INIT_LIST_HEAD(head);
1625 	for (i = 0; i < count; i++)
1626 		list_add_tail(ar[i], head);
1627 	free(ar, M_KMALLOC);
1628 }
1629 
1630 void
1631 linux_irq_handler(void *ent)
1632 {
1633 	struct irq_ent *irqe;
1634 
1635 	linux_set_current(curthread);
1636 
1637 	irqe = ent;
1638 	irqe->handler(irqe->irq, irqe->arg);
1639 }
1640 
1641 #if defined(__i386__) || defined(__amd64__)
1642 int
1643 linux_wbinvd_on_all_cpus(void)
1644 {
1645 
1646 	pmap_invalidate_cache();
1647 	return (0);
1648 }
1649 #endif
1650 
1651 int
1652 linux_on_each_cpu(void callback(void *), void *data)
1653 {
1654 
1655 	smp_rendezvous(smp_no_rendezvous_barrier, callback,
1656 	    smp_no_rendezvous_barrier, data);
1657 	return (0);
1658 }
1659 
1660 struct linux_cdev *
1661 linux_find_cdev(const char *name, unsigned major, unsigned minor)
1662 {
1663 	int unit = MKDEV(major, minor);
1664 	struct cdev *cdev;
1665 
1666 	dev_lock();
1667 	LIST_FOREACH(cdev, &linuxcdevsw.d_devs, si_list) {
1668 		struct linux_cdev *ldev = cdev->si_drv1;
1669 		if (dev2unit(cdev) == unit &&
1670 		    strcmp(kobject_name(&ldev->kobj), name) == 0) {
1671 			break;
1672 		}
1673 	}
1674 	dev_unlock();
1675 
1676 	return (cdev != NULL ? cdev->si_drv1 : NULL);
1677 }
1678 
1679 int
1680 __register_chrdev(unsigned int major, unsigned int baseminor,
1681     unsigned int count, const char *name,
1682     const struct file_operations *fops)
1683 {
1684 	struct linux_cdev *cdev;
1685 	int ret = 0;
1686 	int i;
1687 
1688 	for (i = baseminor; i < baseminor + count; i++) {
1689 		cdev = cdev_alloc();
1690 		cdev_init(cdev, fops);
1691 		kobject_set_name(&cdev->kobj, name);
1692 
1693 		ret = cdev_add(cdev, makedev(major, i), 1);
1694 		if (ret != 0)
1695 			break;
1696 	}
1697 	return (ret);
1698 }
1699 
1700 int
1701 __register_chrdev_p(unsigned int major, unsigned int baseminor,
1702     unsigned int count, const char *name,
1703     const struct file_operations *fops, uid_t uid,
1704     gid_t gid, int mode)
1705 {
1706 	struct linux_cdev *cdev;
1707 	int ret = 0;
1708 	int i;
1709 
1710 	for (i = baseminor; i < baseminor + count; i++) {
1711 		cdev = cdev_alloc();
1712 		cdev_init(cdev, fops);
1713 		kobject_set_name(&cdev->kobj, name);
1714 
1715 		ret = cdev_add_ext(cdev, makedev(major, i), uid, gid, mode);
1716 		if (ret != 0)
1717 			break;
1718 	}
1719 	return (ret);
1720 }
1721 
1722 void
1723 __unregister_chrdev(unsigned int major, unsigned int baseminor,
1724     unsigned int count, const char *name)
1725 {
1726 	struct linux_cdev *cdevp;
1727 	int i;
1728 
1729 	for (i = baseminor; i < baseminor + count; i++) {
1730 		cdevp = linux_find_cdev(name, major, i);
1731 		if (cdevp != NULL)
1732 			cdev_del(cdevp);
1733 	}
1734 }
1735 
1736 #if defined(__i386__) || defined(__amd64__)
1737 bool linux_cpu_has_clflush;
1738 #endif
1739 
1740 static void
1741 linux_compat_init(void *arg)
1742 {
1743 	struct sysctl_oid *rootoid;
1744 	int i;
1745 
1746 #if defined(__i386__) || defined(__amd64__)
1747 	linux_cpu_has_clflush = (cpu_feature & CPUID_CLFSH);
1748 #endif
1749 	rw_init(&linux_vma_lock, "lkpi-vma-lock");
1750 
1751 	rootoid = SYSCTL_ADD_ROOT_NODE(NULL,
1752 	    OID_AUTO, "sys", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "sys");
1753 	kobject_init(&linux_class_root, &linux_class_ktype);
1754 	kobject_set_name(&linux_class_root, "class");
1755 	linux_class_root.oidp = SYSCTL_ADD_NODE(NULL, SYSCTL_CHILDREN(rootoid),
1756 	    OID_AUTO, "class", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "class");
1757 	kobject_init(&linux_root_device.kobj, &linux_dev_ktype);
1758 	kobject_set_name(&linux_root_device.kobj, "device");
1759 	linux_root_device.kobj.oidp = SYSCTL_ADD_NODE(NULL,
1760 	    SYSCTL_CHILDREN(rootoid), OID_AUTO, "device", CTLFLAG_RD, NULL,
1761 	    "device");
1762 	linux_root_device.bsddev = root_bus;
1763 	linux_class_misc.name = "misc";
1764 	class_register(&linux_class_misc);
1765 	INIT_LIST_HEAD(&pci_drivers);
1766 	INIT_LIST_HEAD(&pci_devices);
1767 	spin_lock_init(&pci_lock);
1768 	mtx_init(&vmmaplock, "IO Map lock", NULL, MTX_DEF);
1769 	for (i = 0; i < VMMAP_HASH_SIZE; i++)
1770 		LIST_INIT(&vmmaphead[i]);
1771 }
1772 SYSINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_init, NULL);
1773 
1774 static void
1775 linux_compat_uninit(void *arg)
1776 {
1777 	linux_kobject_kfree_name(&linux_class_root);
1778 	linux_kobject_kfree_name(&linux_root_device.kobj);
1779 	linux_kobject_kfree_name(&linux_class_misc.kobj);
1780 
1781 	rw_destroy(&linux_vma_lock);
1782 }
1783 SYSUNINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_uninit, NULL);
1784 
1785 /*
1786  * NOTE: Linux frequently uses "unsigned long" for pointer to integer
1787  * conversion and vice versa, where in FreeBSD "uintptr_t" would be
1788  * used. Assert these types have the same size, else some parts of the
1789  * LinuxKPI may not work like expected:
1790  */
1791 CTASSERT(sizeof(unsigned long) == sizeof(uintptr_t));
1792