xref: /freebsd/sys/compat/linuxkpi/common/src/linux_compat.c (revision a8fc61d51a53345b7267b1a5e83077c52d6a9b59)
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(0);
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 	TAILQ_INSERT_TAIL(&linux_vma_head, vmap, vm_entry);
489 	rw_wunlock(&linux_vma_lock);
490 	return (vmap);
491 }
492 
493 static void
494 linux_cdev_handle_remove(struct vm_area_struct *vmap)
495 {
496 	if (vmap == NULL)
497 		return;
498 
499 	rw_wlock(&linux_vma_lock);
500 	TAILQ_REMOVE(&linux_vma_head, vmap, vm_entry);
501 	rw_wunlock(&linux_vma_lock);
502 	kfree(vmap);
503 }
504 
505 static struct vm_area_struct *
506 linux_cdev_handle_find(void *handle)
507 {
508 	struct vm_area_struct *vmap;
509 
510 	rw_rlock(&linux_vma_lock);
511 	TAILQ_FOREACH(vmap, &linux_vma_head, vm_entry) {
512 		if (vmap->vm_private_data == handle)
513 			break;
514 	}
515 	rw_runlock(&linux_vma_lock);
516 	return (vmap);
517 }
518 
519 static int
520 linux_cdev_pager_ctor(void *handle, vm_ooffset_t size, vm_prot_t prot,
521 		      vm_ooffset_t foff, struct ucred *cred, u_short *color)
522 {
523 	const struct vm_operations_struct *vm_ops;
524 	struct vm_area_struct *vmap;
525 
526 	vmap = linux_cdev_handle_find(handle);
527 	MPASS(vmap != NULL);
528 
529 	*color = 0;
530 
531 	down_write(&vmap->vm_mm->mmap_sem);
532 	vm_ops = vmap->vm_ops;
533 	if (likely(vm_ops != NULL))
534 		vm_ops->open(vmap);
535 	up_write(&vmap->vm_mm->mmap_sem);
536 
537 	return (0);
538 }
539 
540 static void
541 linux_cdev_pager_dtor(void *handle)
542 {
543 	const struct vm_operations_struct *vm_ops;
544 	struct vm_area_struct *vmap;
545 
546 	vmap = linux_cdev_handle_find(handle);
547 	MPASS(vmap != NULL);
548 
549 	down_write(&vmap->vm_mm->mmap_sem);
550 	vm_ops = vmap->vm_ops;
551 	if (likely(vm_ops != NULL))
552 		vm_ops->close(vmap);
553 	up_write(&vmap->vm_mm->mmap_sem);
554 
555 	linux_cdev_handle_remove(vmap);
556 }
557 
558 static struct cdev_pager_ops linux_cdev_pager_ops = {
559 	.cdev_pg_populate	= linux_cdev_pager_populate,
560 	.cdev_pg_ctor	= linux_cdev_pager_ctor,
561 	.cdev_pg_dtor	= linux_cdev_pager_dtor
562 };
563 
564 static int
565 linux_dev_open(struct cdev *dev, int oflags, int devtype, struct thread *td)
566 {
567 	struct linux_cdev *ldev;
568 	struct linux_file *filp;
569 	struct file *file;
570 	int error;
571 
572 	file = td->td_fpop;
573 	ldev = dev->si_drv1;
574 	if (ldev == NULL)
575 		return (ENODEV);
576 	filp = kzalloc(sizeof(*filp), GFP_KERNEL);
577 	filp->f_dentry = &filp->f_dentry_store;
578 	filp->f_op = ldev->ops;
579 	filp->f_flags = file->f_flag;
580 	vhold(file->f_vnode);
581 	filp->f_vnode = file->f_vnode;
582 	linux_set_current(td);
583 	if (filp->f_op->open) {
584 		error = -filp->f_op->open(file->f_vnode, filp);
585 		if (error) {
586 			kfree(filp);
587 			goto done;
588 		}
589 	}
590 	error = devfs_set_cdevpriv(filp, linux_file_dtor);
591 	if (error) {
592 		filp->f_op->release(file->f_vnode, filp);
593 		kfree(filp);
594 	}
595 done:
596 	return (error);
597 }
598 
599 static int
600 linux_dev_close(struct cdev *dev, int fflag, int devtype, struct thread *td)
601 {
602 	struct linux_file *filp;
603 	struct file *file;
604 	int error;
605 
606 	file = td->td_fpop;
607 	if (dev->si_drv1 == NULL)
608 		return (0);
609 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
610 		return (error);
611 	filp->f_flags = file->f_flag;
612         devfs_clear_cdevpriv();
613 
614 
615 	return (0);
616 }
617 
618 #define	LINUX_IOCTL_MIN_PTR 0x10000UL
619 #define	LINUX_IOCTL_MAX_PTR (LINUX_IOCTL_MIN_PTR + IOCPARM_MAX)
620 
621 static inline int
622 linux_remap_address(void **uaddr, size_t len)
623 {
624 	uintptr_t uaddr_val = (uintptr_t)(*uaddr);
625 
626 	if (unlikely(uaddr_val >= LINUX_IOCTL_MIN_PTR &&
627 	    uaddr_val < LINUX_IOCTL_MAX_PTR)) {
628 		struct task_struct *pts = current;
629 		if (pts == NULL) {
630 			*uaddr = NULL;
631 			return (1);
632 		}
633 
634 		/* compute data offset */
635 		uaddr_val -= LINUX_IOCTL_MIN_PTR;
636 
637 		/* check that length is within bounds */
638 		if ((len > IOCPARM_MAX) ||
639 		    (uaddr_val + len) > pts->bsd_ioctl_len) {
640 			*uaddr = NULL;
641 			return (1);
642 		}
643 
644 		/* re-add kernel buffer address */
645 		uaddr_val += (uintptr_t)pts->bsd_ioctl_data;
646 
647 		/* update address location */
648 		*uaddr = (void *)uaddr_val;
649 		return (1);
650 	}
651 	return (0);
652 }
653 
654 int
655 linux_copyin(const void *uaddr, void *kaddr, size_t len)
656 {
657 	if (linux_remap_address(__DECONST(void **, &uaddr), len)) {
658 		if (uaddr == NULL)
659 			return (-EFAULT);
660 		memcpy(kaddr, uaddr, len);
661 		return (0);
662 	}
663 	return (-copyin(uaddr, kaddr, len));
664 }
665 
666 int
667 linux_copyout(const void *kaddr, void *uaddr, size_t len)
668 {
669 	if (linux_remap_address(&uaddr, len)) {
670 		if (uaddr == NULL)
671 			return (-EFAULT);
672 		memcpy(uaddr, kaddr, len);
673 		return (0);
674 	}
675 	return (-copyout(kaddr, uaddr, len));
676 }
677 
678 size_t
679 linux_clear_user(void *_uaddr, size_t _len)
680 {
681 	uint8_t *uaddr = _uaddr;
682 	size_t len = _len;
683 
684 	/* make sure uaddr is aligned before going into the fast loop */
685 	while (((uintptr_t)uaddr & 7) != 0 && len > 7) {
686 		if (subyte(uaddr, 0))
687 			return (_len);
688 		uaddr++;
689 		len--;
690 	}
691 
692 	/* zero 8 bytes at a time */
693 	while (len > 7) {
694 #ifdef __LP64__
695 		if (suword64(uaddr, 0))
696 			return (_len);
697 #else
698 		if (suword32(uaddr, 0))
699 			return (_len);
700 		if (suword32(uaddr + 4, 0))
701 			return (_len);
702 #endif
703 		uaddr += 8;
704 		len -= 8;
705 	}
706 
707 	/* zero fill end, if any */
708 	while (len > 0) {
709 		if (subyte(uaddr, 0))
710 			return (_len);
711 		uaddr++;
712 		len--;
713 	}
714 	return (0);
715 }
716 
717 int
718 linux_access_ok(int rw, const void *uaddr, size_t len)
719 {
720 	uintptr_t saddr;
721 	uintptr_t eaddr;
722 
723 	/* get start and end address */
724 	saddr = (uintptr_t)uaddr;
725 	eaddr = (uintptr_t)uaddr + len;
726 
727 	/* verify addresses are valid for userspace */
728 	return ((saddr == eaddr) ||
729 	    (eaddr > saddr && eaddr <= VM_MAXUSER_ADDRESS));
730 }
731 
732 static int
733 linux_dev_ioctl(struct cdev *dev, u_long cmd, caddr_t data, int fflag,
734     struct thread *td)
735 {
736 	struct linux_file *filp;
737 	struct file *file;
738 	unsigned size;
739 	int error;
740 
741 	file = td->td_fpop;
742 	if (dev->si_drv1 == NULL)
743 		return (ENXIO);
744 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
745 		return (error);
746 	filp->f_flags = file->f_flag;
747 
748 	linux_set_current(td);
749 	size = IOCPARM_LEN(cmd);
750 	/* refer to logic in sys_ioctl() */
751 	if (size > 0) {
752 		/*
753 		 * Setup hint for linux_copyin() and linux_copyout().
754 		 *
755 		 * Background: Linux code expects a user-space address
756 		 * while FreeBSD supplies a kernel-space address.
757 		 */
758 		current->bsd_ioctl_data = data;
759 		current->bsd_ioctl_len = size;
760 		data = (void *)LINUX_IOCTL_MIN_PTR;
761 	} else {
762 		/* fetch user-space pointer */
763 		data = *(void **)data;
764 	}
765 	if (filp->f_op->unlocked_ioctl)
766 		error = -filp->f_op->unlocked_ioctl(filp, cmd, (u_long)data);
767 	else
768 		error = ENOTTY;
769 	if (size > 0) {
770 		current->bsd_ioctl_data = NULL;
771 		current->bsd_ioctl_len = 0;
772 	}
773 
774 	return (error);
775 }
776 
777 static int
778 linux_dev_read(struct cdev *dev, struct uio *uio, int ioflag)
779 {
780 	struct linux_file *filp;
781 	struct thread *td;
782 	struct file *file;
783 	ssize_t bytes;
784 	int error;
785 
786 	td = curthread;
787 	file = td->td_fpop;
788 	if (dev->si_drv1 == NULL)
789 		return (ENXIO);
790 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
791 		return (error);
792 	filp->f_flags = file->f_flag;
793 	/* XXX no support for I/O vectors currently */
794 	if (uio->uio_iovcnt != 1)
795 		return (EOPNOTSUPP);
796 	linux_set_current(td);
797 	if (filp->f_op->read) {
798 		bytes = filp->f_op->read(filp, uio->uio_iov->iov_base,
799 		    uio->uio_iov->iov_len, &uio->uio_offset);
800 		if (bytes >= 0) {
801 			uio->uio_iov->iov_base =
802 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
803 			uio->uio_iov->iov_len -= bytes;
804 			uio->uio_resid -= bytes;
805 		} else
806 			error = -bytes;
807 	} else
808 		error = ENXIO;
809 
810 	return (error);
811 }
812 
813 static int
814 linux_dev_write(struct cdev *dev, struct uio *uio, int ioflag)
815 {
816 	struct linux_file *filp;
817 	struct thread *td;
818 	struct file *file;
819 	ssize_t bytes;
820 	int error;
821 
822 	td = curthread;
823 	file = td->td_fpop;
824 	if (dev->si_drv1 == NULL)
825 		return (ENXIO);
826 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
827 		return (error);
828 	filp->f_flags = file->f_flag;
829 	/* XXX no support for I/O vectors currently */
830 	if (uio->uio_iovcnt != 1)
831 		return (EOPNOTSUPP);
832 	linux_set_current(td);
833 	if (filp->f_op->write) {
834 		bytes = filp->f_op->write(filp, uio->uio_iov->iov_base,
835 		    uio->uio_iov->iov_len, &uio->uio_offset);
836 		if (bytes >= 0) {
837 			uio->uio_iov->iov_base =
838 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
839 			uio->uio_iov->iov_len -= bytes;
840 			uio->uio_resid -= bytes;
841 		} else
842 			error = -bytes;
843 	} else
844 		error = ENXIO;
845 
846 	return (error);
847 }
848 
849 static int
850 linux_dev_poll(struct cdev *dev, int events, struct thread *td)
851 {
852 	struct linux_file *filp;
853 	struct file *file;
854 	int revents;
855 
856 	if (dev->si_drv1 == NULL)
857 		goto error;
858 	if (devfs_get_cdevpriv((void **)&filp) != 0)
859 		goto error;
860 
861 	file = td->td_fpop;
862 	filp->f_flags = file->f_flag;
863 	linux_set_current(td);
864 	if (filp->f_op->poll)
865 		revents = filp->f_op->poll(filp, NULL) & events;
866 	else
867 		revents = 0;
868 
869 	return (revents);
870 error:
871 	return (events & (POLLHUP|POLLIN|POLLRDNORM|POLLOUT|POLLWRNORM));
872 }
873 
874 static int
875 linux_dev_mmap_single(struct cdev *dev, vm_ooffset_t *offset,
876     vm_size_t size, struct vm_object **object, int nprot)
877 {
878 	struct vm_area_struct *vmap;
879 	struct linux_file *filp;
880 	struct thread *td;
881 	struct file *file;
882 	vm_memattr_t attr;
883 	int error;
884 
885 	td = curthread;
886 	file = td->td_fpop;
887 	if (dev->si_drv1 == NULL)
888 		return (ENODEV);
889 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
890 		return (error);
891 	filp->f_flags = file->f_flag;
892 
893 	if (filp->f_op->mmap == NULL)
894 		return (ENODEV);
895 
896 	linux_set_current(td);
897 
898 	vmap = kzalloc(sizeof(*vmap), GFP_KERNEL);
899 	vmap->vm_start = 0;
900 	vmap->vm_end = size;
901 	vmap->vm_pgoff = *offset / PAGE_SIZE;
902 	vmap->vm_pfn = 0;
903 	vmap->vm_flags = vmap->vm_page_prot = nprot;
904 	vmap->vm_ops = NULL;
905 	vmap->vm_file = filp;
906 	vmap->vm_mm = current->mm;
907 
908 	if (unlikely(down_write_killable(&vmap->vm_mm->mmap_sem))) {
909 		error = EINTR;
910 	} else {
911 		error = -filp->f_op->mmap(filp, vmap);
912 		up_write(&vmap->vm_mm->mmap_sem);
913 	}
914 
915 	if (error != 0) {
916 		kfree(vmap);
917 		return (error);
918 	}
919 
920 	attr = pgprot2cachemode(vmap->vm_page_prot);
921 
922 	if (vmap->vm_ops != NULL) {
923 		void *vm_private_data;
924 
925 		if (vmap->vm_ops->fault == NULL ||
926 		    vmap->vm_ops->open == NULL ||
927 		    vmap->vm_ops->close == NULL ||
928 		    vmap->vm_private_data == NULL) {
929 			kfree(vmap);
930 			return (EINVAL);
931 		}
932 
933 		vm_private_data = vmap->vm_private_data;
934 
935 		vmap = linux_cdev_handle_insert(vm_private_data, vmap);
936 
937 		*object = cdev_pager_allocate(vm_private_data, OBJT_MGTDEVICE,
938 		    &linux_cdev_pager_ops, size, nprot, *offset, curthread->td_ucred);
939 
940 		if (*object == NULL) {
941 			linux_cdev_handle_remove(vmap);
942 			return (EINVAL);
943 		}
944 	} else {
945 		struct sglist *sg;
946 
947 		sg = sglist_alloc(1, M_WAITOK);
948 		sglist_append_phys(sg, (vm_paddr_t)vmap->vm_pfn << PAGE_SHIFT, vmap->vm_len);
949 
950 		*object = vm_pager_allocate(OBJT_SG, sg, vmap->vm_len,
951 		    nprot, 0, curthread->td_ucred);
952 
953 		kfree(vmap);
954 
955 		if (*object == NULL) {
956 			sglist_free(sg);
957 			return (EINVAL);
958 		}
959 	}
960 
961 	if (attr != VM_MEMATTR_DEFAULT) {
962 		VM_OBJECT_WLOCK(*object);
963 		vm_object_set_memattr(*object, attr);
964 		VM_OBJECT_WUNLOCK(*object);
965 	}
966 	*offset = 0;
967 	return (0);
968 }
969 
970 struct cdevsw linuxcdevsw = {
971 	.d_version = D_VERSION,
972 	.d_flags = D_TRACKCLOSE,
973 	.d_open = linux_dev_open,
974 	.d_close = linux_dev_close,
975 	.d_read = linux_dev_read,
976 	.d_write = linux_dev_write,
977 	.d_ioctl = linux_dev_ioctl,
978 	.d_mmap_single = linux_dev_mmap_single,
979 	.d_poll = linux_dev_poll,
980 };
981 
982 static int
983 linux_file_read(struct file *file, struct uio *uio, struct ucred *active_cred,
984     int flags, struct thread *td)
985 {
986 	struct linux_file *filp;
987 	ssize_t bytes;
988 	int error;
989 
990 	error = 0;
991 	filp = (struct linux_file *)file->f_data;
992 	filp->f_flags = file->f_flag;
993 	/* XXX no support for I/O vectors currently */
994 	if (uio->uio_iovcnt != 1)
995 		return (EOPNOTSUPP);
996 	linux_set_current(td);
997 	if (filp->f_op->read) {
998 		bytes = filp->f_op->read(filp, uio->uio_iov->iov_base,
999 		    uio->uio_iov->iov_len, &uio->uio_offset);
1000 		if (bytes >= 0) {
1001 			uio->uio_iov->iov_base =
1002 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
1003 			uio->uio_iov->iov_len -= bytes;
1004 			uio->uio_resid -= bytes;
1005 		} else
1006 			error = -bytes;
1007 	} else
1008 		error = ENXIO;
1009 
1010 	return (error);
1011 }
1012 
1013 static int
1014 linux_file_poll(struct file *file, int events, struct ucred *active_cred,
1015     struct thread *td)
1016 {
1017 	struct linux_file *filp;
1018 	int revents;
1019 
1020 	filp = (struct linux_file *)file->f_data;
1021 	filp->f_flags = file->f_flag;
1022 	linux_set_current(td);
1023 	if (filp->f_op->poll)
1024 		revents = filp->f_op->poll(filp, NULL) & events;
1025 	else
1026 		revents = 0;
1027 
1028 	return (revents);
1029 }
1030 
1031 static int
1032 linux_file_close(struct file *file, struct thread *td)
1033 {
1034 	struct linux_file *filp;
1035 	int error;
1036 
1037 	filp = (struct linux_file *)file->f_data;
1038 	filp->f_flags = file->f_flag;
1039 	linux_set_current(td);
1040 	error = -filp->f_op->release(NULL, filp);
1041 	funsetown(&filp->f_sigio);
1042 	kfree(filp);
1043 
1044 	return (error);
1045 }
1046 
1047 static int
1048 linux_file_ioctl(struct file *fp, u_long cmd, void *data, struct ucred *cred,
1049     struct thread *td)
1050 {
1051 	struct linux_file *filp;
1052 	int error;
1053 
1054 	filp = (struct linux_file *)fp->f_data;
1055 	filp->f_flags = fp->f_flag;
1056 	error = 0;
1057 
1058 	linux_set_current(td);
1059 	switch (cmd) {
1060 	case FIONBIO:
1061 		break;
1062 	case FIOASYNC:
1063 		if (filp->f_op->fasync == NULL)
1064 			break;
1065 		error = filp->f_op->fasync(0, filp, fp->f_flag & FASYNC);
1066 		break;
1067 	case FIOSETOWN:
1068 		error = fsetown(*(int *)data, &filp->f_sigio);
1069 		if (error == 0)
1070 			error = filp->f_op->fasync(0, filp,
1071 			    fp->f_flag & FASYNC);
1072 		break;
1073 	case FIOGETOWN:
1074 		*(int *)data = fgetown(&filp->f_sigio);
1075 		break;
1076 	default:
1077 		error = ENOTTY;
1078 		break;
1079 	}
1080 	return (error);
1081 }
1082 
1083 static int
1084 linux_file_stat(struct file *fp, struct stat *sb, struct ucred *active_cred,
1085     struct thread *td)
1086 {
1087 
1088 	return (EOPNOTSUPP);
1089 }
1090 
1091 static int
1092 linux_file_fill_kinfo(struct file *fp, struct kinfo_file *kif,
1093     struct filedesc *fdp)
1094 {
1095 
1096 	return (0);
1097 }
1098 
1099 struct fileops linuxfileops = {
1100 	.fo_read = linux_file_read,
1101 	.fo_write = invfo_rdwr,
1102 	.fo_truncate = invfo_truncate,
1103 	.fo_kqfilter = invfo_kqfilter,
1104 	.fo_stat = linux_file_stat,
1105 	.fo_fill_kinfo = linux_file_fill_kinfo,
1106 	.fo_poll = linux_file_poll,
1107 	.fo_close = linux_file_close,
1108 	.fo_ioctl = linux_file_ioctl,
1109 	.fo_chmod = invfo_chmod,
1110 	.fo_chown = invfo_chown,
1111 	.fo_sendfile = invfo_sendfile,
1112 };
1113 
1114 /*
1115  * Hash of vmmap addresses.  This is infrequently accessed and does not
1116  * need to be particularly large.  This is done because we must store the
1117  * caller's idea of the map size to properly unmap.
1118  */
1119 struct vmmap {
1120 	LIST_ENTRY(vmmap)	vm_next;
1121 	void 			*vm_addr;
1122 	unsigned long		vm_size;
1123 };
1124 
1125 struct vmmaphd {
1126 	struct vmmap *lh_first;
1127 };
1128 #define	VMMAP_HASH_SIZE	64
1129 #define	VMMAP_HASH_MASK	(VMMAP_HASH_SIZE - 1)
1130 #define	VM_HASH(addr)	((uintptr_t)(addr) >> PAGE_SHIFT) & VMMAP_HASH_MASK
1131 static struct vmmaphd vmmaphead[VMMAP_HASH_SIZE];
1132 static struct mtx vmmaplock;
1133 
1134 static void
1135 vmmap_add(void *addr, unsigned long size)
1136 {
1137 	struct vmmap *vmmap;
1138 
1139 	vmmap = kmalloc(sizeof(*vmmap), GFP_KERNEL);
1140 	mtx_lock(&vmmaplock);
1141 	vmmap->vm_size = size;
1142 	vmmap->vm_addr = addr;
1143 	LIST_INSERT_HEAD(&vmmaphead[VM_HASH(addr)], vmmap, vm_next);
1144 	mtx_unlock(&vmmaplock);
1145 }
1146 
1147 static struct vmmap *
1148 vmmap_remove(void *addr)
1149 {
1150 	struct vmmap *vmmap;
1151 
1152 	mtx_lock(&vmmaplock);
1153 	LIST_FOREACH(vmmap, &vmmaphead[VM_HASH(addr)], vm_next)
1154 		if (vmmap->vm_addr == addr)
1155 			break;
1156 	if (vmmap)
1157 		LIST_REMOVE(vmmap, vm_next);
1158 	mtx_unlock(&vmmaplock);
1159 
1160 	return (vmmap);
1161 }
1162 
1163 #if defined(__i386__) || defined(__amd64__)
1164 void *
1165 _ioremap_attr(vm_paddr_t phys_addr, unsigned long size, int attr)
1166 {
1167 	void *addr;
1168 
1169 	addr = pmap_mapdev_attr(phys_addr, size, attr);
1170 	if (addr == NULL)
1171 		return (NULL);
1172 	vmmap_add(addr, size);
1173 
1174 	return (addr);
1175 }
1176 #endif
1177 
1178 void
1179 iounmap(void *addr)
1180 {
1181 	struct vmmap *vmmap;
1182 
1183 	vmmap = vmmap_remove(addr);
1184 	if (vmmap == NULL)
1185 		return;
1186 #if defined(__i386__) || defined(__amd64__)
1187 	pmap_unmapdev((vm_offset_t)addr, vmmap->vm_size);
1188 #endif
1189 	kfree(vmmap);
1190 }
1191 
1192 
1193 void *
1194 vmap(struct page **pages, unsigned int count, unsigned long flags, int prot)
1195 {
1196 	vm_offset_t off;
1197 	size_t size;
1198 
1199 	size = count * PAGE_SIZE;
1200 	off = kva_alloc(size);
1201 	if (off == 0)
1202 		return (NULL);
1203 	vmmap_add((void *)off, size);
1204 	pmap_qenter(off, pages, count);
1205 
1206 	return ((void *)off);
1207 }
1208 
1209 void
1210 vunmap(void *addr)
1211 {
1212 	struct vmmap *vmmap;
1213 
1214 	vmmap = vmmap_remove(addr);
1215 	if (vmmap == NULL)
1216 		return;
1217 	pmap_qremove((vm_offset_t)addr, vmmap->vm_size / PAGE_SIZE);
1218 	kva_free((vm_offset_t)addr, vmmap->vm_size);
1219 	kfree(vmmap);
1220 }
1221 
1222 char *
1223 kvasprintf(gfp_t gfp, const char *fmt, va_list ap)
1224 {
1225 	unsigned int len;
1226 	char *p;
1227 	va_list aq;
1228 
1229 	va_copy(aq, ap);
1230 	len = vsnprintf(NULL, 0, fmt, aq);
1231 	va_end(aq);
1232 
1233 	p = kmalloc(len + 1, gfp);
1234 	if (p != NULL)
1235 		vsnprintf(p, len + 1, fmt, ap);
1236 
1237 	return (p);
1238 }
1239 
1240 char *
1241 kasprintf(gfp_t gfp, const char *fmt, ...)
1242 {
1243 	va_list ap;
1244 	char *p;
1245 
1246 	va_start(ap, fmt);
1247 	p = kvasprintf(gfp, fmt, ap);
1248 	va_end(ap);
1249 
1250 	return (p);
1251 }
1252 
1253 static void
1254 linux_timer_callback_wrapper(void *context)
1255 {
1256 	struct timer_list *timer;
1257 
1258 	linux_set_current(curthread);
1259 
1260 	timer = context;
1261 	timer->function(timer->data);
1262 }
1263 
1264 void
1265 mod_timer(struct timer_list *timer, unsigned long expires)
1266 {
1267 
1268 	timer->expires = expires;
1269 	callout_reset(&timer->timer_callout,
1270 	    linux_timer_jiffies_until(expires),
1271 	    &linux_timer_callback_wrapper, timer);
1272 }
1273 
1274 void
1275 add_timer(struct timer_list *timer)
1276 {
1277 
1278 	callout_reset(&timer->timer_callout,
1279 	    linux_timer_jiffies_until(timer->expires),
1280 	    &linux_timer_callback_wrapper, timer);
1281 }
1282 
1283 void
1284 add_timer_on(struct timer_list *timer, int cpu)
1285 {
1286 
1287 	callout_reset_on(&timer->timer_callout,
1288 	    linux_timer_jiffies_until(timer->expires),
1289 	    &linux_timer_callback_wrapper, timer, cpu);
1290 }
1291 
1292 static void
1293 linux_timer_init(void *arg)
1294 {
1295 
1296 	/*
1297 	 * Compute an internal HZ value which can divide 2**32 to
1298 	 * avoid timer rounding problems when the tick value wraps
1299 	 * around 2**32:
1300 	 */
1301 	linux_timer_hz_mask = 1;
1302 	while (linux_timer_hz_mask < (unsigned long)hz)
1303 		linux_timer_hz_mask *= 2;
1304 	linux_timer_hz_mask--;
1305 }
1306 SYSINIT(linux_timer, SI_SUB_DRIVERS, SI_ORDER_FIRST, linux_timer_init, NULL);
1307 
1308 void
1309 linux_complete_common(struct completion *c, int all)
1310 {
1311 	int wakeup_swapper;
1312 
1313 	sleepq_lock(c);
1314 	c->done++;
1315 	if (all)
1316 		wakeup_swapper = sleepq_broadcast(c, SLEEPQ_SLEEP, 0, 0);
1317 	else
1318 		wakeup_swapper = sleepq_signal(c, SLEEPQ_SLEEP, 0, 0);
1319 	sleepq_release(c);
1320 	if (wakeup_swapper)
1321 		kick_proc0();
1322 }
1323 
1324 /*
1325  * Indefinite wait for done != 0 with or without signals.
1326  */
1327 long
1328 linux_wait_for_common(struct completion *c, int flags)
1329 {
1330 	long error;
1331 
1332 	if (SCHEDULER_STOPPED())
1333 		return (0);
1334 
1335 	DROP_GIANT();
1336 
1337 	if (flags != 0)
1338 		flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
1339 	else
1340 		flags = SLEEPQ_SLEEP;
1341 	error = 0;
1342 	for (;;) {
1343 		sleepq_lock(c);
1344 		if (c->done)
1345 			break;
1346 		sleepq_add(c, NULL, "completion", flags, 0);
1347 		if (flags & SLEEPQ_INTERRUPTIBLE) {
1348 			if (sleepq_wait_sig(c, 0) != 0) {
1349 				error = -ERESTARTSYS;
1350 				goto intr;
1351 			}
1352 		} else
1353 			sleepq_wait(c, 0);
1354 	}
1355 	c->done--;
1356 	sleepq_release(c);
1357 
1358 intr:
1359 	PICKUP_GIANT();
1360 
1361 	return (error);
1362 }
1363 
1364 /*
1365  * Time limited wait for done != 0 with or without signals.
1366  */
1367 long
1368 linux_wait_for_timeout_common(struct completion *c, long timeout, int flags)
1369 {
1370 	long end = jiffies + timeout, error;
1371 	int ret;
1372 
1373 	if (SCHEDULER_STOPPED())
1374 		return (0);
1375 
1376 	DROP_GIANT();
1377 
1378 	if (flags != 0)
1379 		flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
1380 	else
1381 		flags = SLEEPQ_SLEEP;
1382 
1383 	error = 0;
1384 	ret = 0;
1385 	for (;;) {
1386 		sleepq_lock(c);
1387 		if (c->done)
1388 			break;
1389 		sleepq_add(c, NULL, "completion", flags, 0);
1390 		sleepq_set_timeout(c, linux_timer_jiffies_until(end));
1391 		if (flags & SLEEPQ_INTERRUPTIBLE)
1392 			ret = sleepq_timedwait_sig(c, 0);
1393 		else
1394 			ret = sleepq_timedwait(c, 0);
1395 		if (ret != 0) {
1396 			/* check for timeout or signal */
1397 			if (ret == EWOULDBLOCK)
1398 				error = 0;
1399 			else
1400 				error = -ERESTARTSYS;
1401 			goto intr;
1402 		}
1403 	}
1404 	c->done--;
1405 	sleepq_release(c);
1406 
1407 intr:
1408 	PICKUP_GIANT();
1409 
1410 	/* return how many jiffies are left */
1411 	return (ret != 0 ? error : linux_timer_jiffies_until(end));
1412 }
1413 
1414 int
1415 linux_try_wait_for_completion(struct completion *c)
1416 {
1417 	int isdone;
1418 
1419 	isdone = 1;
1420 	sleepq_lock(c);
1421 	if (c->done)
1422 		c->done--;
1423 	else
1424 		isdone = 0;
1425 	sleepq_release(c);
1426 	return (isdone);
1427 }
1428 
1429 int
1430 linux_completion_done(struct completion *c)
1431 {
1432 	int isdone;
1433 
1434 	isdone = 1;
1435 	sleepq_lock(c);
1436 	if (c->done == 0)
1437 		isdone = 0;
1438 	sleepq_release(c);
1439 	return (isdone);
1440 }
1441 
1442 static void
1443 linux_cdev_release(struct kobject *kobj)
1444 {
1445 	struct linux_cdev *cdev;
1446 	struct kobject *parent;
1447 
1448 	cdev = container_of(kobj, struct linux_cdev, kobj);
1449 	parent = kobj->parent;
1450 	if (cdev->cdev)
1451 		destroy_dev(cdev->cdev);
1452 	kfree(cdev);
1453 	kobject_put(parent);
1454 }
1455 
1456 static void
1457 linux_cdev_static_release(struct kobject *kobj)
1458 {
1459 	struct linux_cdev *cdev;
1460 	struct kobject *parent;
1461 
1462 	cdev = container_of(kobj, struct linux_cdev, kobj);
1463 	parent = kobj->parent;
1464 	if (cdev->cdev)
1465 		destroy_dev(cdev->cdev);
1466 	kobject_put(parent);
1467 }
1468 
1469 const struct kobj_type linux_cdev_ktype = {
1470 	.release = linux_cdev_release,
1471 };
1472 
1473 const struct kobj_type linux_cdev_static_ktype = {
1474 	.release = linux_cdev_static_release,
1475 };
1476 
1477 static void
1478 linux_handle_ifnet_link_event(void *arg, struct ifnet *ifp, int linkstate)
1479 {
1480 	struct notifier_block *nb;
1481 
1482 	nb = arg;
1483 	if (linkstate == LINK_STATE_UP)
1484 		nb->notifier_call(nb, NETDEV_UP, ifp);
1485 	else
1486 		nb->notifier_call(nb, NETDEV_DOWN, ifp);
1487 }
1488 
1489 static void
1490 linux_handle_ifnet_arrival_event(void *arg, struct ifnet *ifp)
1491 {
1492 	struct notifier_block *nb;
1493 
1494 	nb = arg;
1495 	nb->notifier_call(nb, NETDEV_REGISTER, ifp);
1496 }
1497 
1498 static void
1499 linux_handle_ifnet_departure_event(void *arg, struct ifnet *ifp)
1500 {
1501 	struct notifier_block *nb;
1502 
1503 	nb = arg;
1504 	nb->notifier_call(nb, NETDEV_UNREGISTER, ifp);
1505 }
1506 
1507 static void
1508 linux_handle_iflladdr_event(void *arg, struct ifnet *ifp)
1509 {
1510 	struct notifier_block *nb;
1511 
1512 	nb = arg;
1513 	nb->notifier_call(nb, NETDEV_CHANGEADDR, ifp);
1514 }
1515 
1516 static void
1517 linux_handle_ifaddr_event(void *arg, struct ifnet *ifp)
1518 {
1519 	struct notifier_block *nb;
1520 
1521 	nb = arg;
1522 	nb->notifier_call(nb, NETDEV_CHANGEIFADDR, ifp);
1523 }
1524 
1525 int
1526 register_netdevice_notifier(struct notifier_block *nb)
1527 {
1528 
1529 	nb->tags[NETDEV_UP] = EVENTHANDLER_REGISTER(
1530 	    ifnet_link_event, linux_handle_ifnet_link_event, nb, 0);
1531 	nb->tags[NETDEV_REGISTER] = EVENTHANDLER_REGISTER(
1532 	    ifnet_arrival_event, linux_handle_ifnet_arrival_event, nb, 0);
1533 	nb->tags[NETDEV_UNREGISTER] = EVENTHANDLER_REGISTER(
1534 	    ifnet_departure_event, linux_handle_ifnet_departure_event, nb, 0);
1535 	nb->tags[NETDEV_CHANGEADDR] = EVENTHANDLER_REGISTER(
1536 	    iflladdr_event, linux_handle_iflladdr_event, nb, 0);
1537 
1538 	return (0);
1539 }
1540 
1541 int
1542 register_inetaddr_notifier(struct notifier_block *nb)
1543 {
1544 
1545         nb->tags[NETDEV_CHANGEIFADDR] = EVENTHANDLER_REGISTER(
1546             ifaddr_event, linux_handle_ifaddr_event, nb, 0);
1547         return (0);
1548 }
1549 
1550 int
1551 unregister_netdevice_notifier(struct notifier_block *nb)
1552 {
1553 
1554         EVENTHANDLER_DEREGISTER(ifnet_link_event,
1555 	    nb->tags[NETDEV_UP]);
1556         EVENTHANDLER_DEREGISTER(ifnet_arrival_event,
1557 	    nb->tags[NETDEV_REGISTER]);
1558         EVENTHANDLER_DEREGISTER(ifnet_departure_event,
1559 	    nb->tags[NETDEV_UNREGISTER]);
1560         EVENTHANDLER_DEREGISTER(iflladdr_event,
1561 	    nb->tags[NETDEV_CHANGEADDR]);
1562 
1563 	return (0);
1564 }
1565 
1566 int
1567 unregister_inetaddr_notifier(struct notifier_block *nb)
1568 {
1569 
1570         EVENTHANDLER_DEREGISTER(ifaddr_event,
1571             nb->tags[NETDEV_CHANGEIFADDR]);
1572 
1573         return (0);
1574 }
1575 
1576 struct list_sort_thunk {
1577 	int (*cmp)(void *, struct list_head *, struct list_head *);
1578 	void *priv;
1579 };
1580 
1581 static inline int
1582 linux_le_cmp(void *priv, const void *d1, const void *d2)
1583 {
1584 	struct list_head *le1, *le2;
1585 	struct list_sort_thunk *thunk;
1586 
1587 	thunk = priv;
1588 	le1 = *(__DECONST(struct list_head **, d1));
1589 	le2 = *(__DECONST(struct list_head **, d2));
1590 	return ((thunk->cmp)(thunk->priv, le1, le2));
1591 }
1592 
1593 void
1594 list_sort(void *priv, struct list_head *head, int (*cmp)(void *priv,
1595     struct list_head *a, struct list_head *b))
1596 {
1597 	struct list_sort_thunk thunk;
1598 	struct list_head **ar, *le;
1599 	size_t count, i;
1600 
1601 	count = 0;
1602 	list_for_each(le, head)
1603 		count++;
1604 	ar = malloc(sizeof(struct list_head *) * count, M_KMALLOC, M_WAITOK);
1605 	i = 0;
1606 	list_for_each(le, head)
1607 		ar[i++] = le;
1608 	thunk.cmp = cmp;
1609 	thunk.priv = priv;
1610 	qsort_r(ar, count, sizeof(struct list_head *), &thunk, linux_le_cmp);
1611 	INIT_LIST_HEAD(head);
1612 	for (i = 0; i < count; i++)
1613 		list_add_tail(ar[i], head);
1614 	free(ar, M_KMALLOC);
1615 }
1616 
1617 void
1618 linux_irq_handler(void *ent)
1619 {
1620 	struct irq_ent *irqe;
1621 
1622 	linux_set_current(curthread);
1623 
1624 	irqe = ent;
1625 	irqe->handler(irqe->irq, irqe->arg);
1626 }
1627 
1628 #if defined(__i386__) || defined(__amd64__)
1629 static void
1630 wbinvd_cb(void *arg __unused)
1631 {
1632 
1633 	wbinvd();
1634 }
1635 
1636 int
1637 linux_wbinvd_on_all_cpus(void)
1638 {
1639 
1640 	return (linux_on_each_cpu(wbinvd_cb, NULL));
1641 }
1642 #endif
1643 
1644 int
1645 linux_on_each_cpu(void callback(void *), void *data)
1646 {
1647 
1648 	smp_rendezvous(smp_no_rendezvous_barrier, callback,
1649 	    smp_no_rendezvous_barrier, data);
1650 	return (0);
1651 }
1652 
1653 struct linux_cdev *
1654 linux_find_cdev(const char *name, unsigned major, unsigned minor)
1655 {
1656 	int unit = MKDEV(major, minor);
1657 	struct cdev *cdev;
1658 
1659 	dev_lock();
1660 	LIST_FOREACH(cdev, &linuxcdevsw.d_devs, si_list) {
1661 		struct linux_cdev *ldev = cdev->si_drv1;
1662 		if (dev2unit(cdev) == unit &&
1663 		    strcmp(kobject_name(&ldev->kobj), name) == 0) {
1664 			break;
1665 		}
1666 	}
1667 	dev_unlock();
1668 
1669 	return (cdev != NULL ? cdev->si_drv1 : NULL);
1670 }
1671 
1672 int
1673 __register_chrdev(unsigned int major, unsigned int baseminor,
1674     unsigned int count, const char *name,
1675     const struct file_operations *fops)
1676 {
1677 	struct linux_cdev *cdev;
1678 	int ret = 0;
1679 	int i;
1680 
1681 	for (i = baseminor; i < baseminor + count; i++) {
1682 		cdev = cdev_alloc();
1683 		cdev_init(cdev, fops);
1684 		kobject_set_name(&cdev->kobj, name);
1685 
1686 		ret = cdev_add(cdev, makedev(major, i), 1);
1687 		if (ret != 0)
1688 			break;
1689 	}
1690 	return (ret);
1691 }
1692 
1693 int
1694 __register_chrdev_p(unsigned int major, unsigned int baseminor,
1695     unsigned int count, const char *name,
1696     const struct file_operations *fops, uid_t uid,
1697     gid_t gid, int mode)
1698 {
1699 	struct linux_cdev *cdev;
1700 	int ret = 0;
1701 	int i;
1702 
1703 	for (i = baseminor; i < baseminor + count; i++) {
1704 		cdev = cdev_alloc();
1705 		cdev_init(cdev, fops);
1706 		kobject_set_name(&cdev->kobj, name);
1707 
1708 		ret = cdev_add_ext(cdev, makedev(major, i), uid, gid, mode);
1709 		if (ret != 0)
1710 			break;
1711 	}
1712 	return (ret);
1713 }
1714 
1715 void
1716 __unregister_chrdev(unsigned int major, unsigned int baseminor,
1717     unsigned int count, const char *name)
1718 {
1719 	struct linux_cdev *cdevp;
1720 	int i;
1721 
1722 	for (i = baseminor; i < baseminor + count; i++) {
1723 		cdevp = linux_find_cdev(name, major, i);
1724 		if (cdevp != NULL)
1725 			cdev_del(cdevp);
1726 	}
1727 }
1728 
1729 #if defined(__i386__) || defined(__amd64__)
1730 bool linux_cpu_has_clflush;
1731 #endif
1732 
1733 static void
1734 linux_compat_init(void *arg)
1735 {
1736 	struct sysctl_oid *rootoid;
1737 	int i;
1738 
1739 #if defined(__i386__) || defined(__amd64__)
1740 	linux_cpu_has_clflush = (cpu_feature & CPUID_CLFSH);
1741 #endif
1742 	rw_init(&linux_vma_lock, "lkpi-vma-lock");
1743 
1744 	rootoid = SYSCTL_ADD_ROOT_NODE(NULL,
1745 	    OID_AUTO, "sys", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "sys");
1746 	kobject_init(&linux_class_root, &linux_class_ktype);
1747 	kobject_set_name(&linux_class_root, "class");
1748 	linux_class_root.oidp = SYSCTL_ADD_NODE(NULL, SYSCTL_CHILDREN(rootoid),
1749 	    OID_AUTO, "class", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "class");
1750 	kobject_init(&linux_root_device.kobj, &linux_dev_ktype);
1751 	kobject_set_name(&linux_root_device.kobj, "device");
1752 	linux_root_device.kobj.oidp = SYSCTL_ADD_NODE(NULL,
1753 	    SYSCTL_CHILDREN(rootoid), OID_AUTO, "device", CTLFLAG_RD, NULL,
1754 	    "device");
1755 	linux_root_device.bsddev = root_bus;
1756 	linux_class_misc.name = "misc";
1757 	class_register(&linux_class_misc);
1758 	INIT_LIST_HEAD(&pci_drivers);
1759 	INIT_LIST_HEAD(&pci_devices);
1760 	spin_lock_init(&pci_lock);
1761 	mtx_init(&vmmaplock, "IO Map lock", NULL, MTX_DEF);
1762 	for (i = 0; i < VMMAP_HASH_SIZE; i++)
1763 		LIST_INIT(&vmmaphead[i]);
1764 }
1765 SYSINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_init, NULL);
1766 
1767 static void
1768 linux_compat_uninit(void *arg)
1769 {
1770 	linux_kobject_kfree_name(&linux_class_root);
1771 	linux_kobject_kfree_name(&linux_root_device.kobj);
1772 	linux_kobject_kfree_name(&linux_class_misc.kobj);
1773 
1774 	rw_destroy(&linux_vma_lock);
1775 }
1776 SYSUNINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_uninit, NULL);
1777 
1778 /*
1779  * NOTE: Linux frequently uses "unsigned long" for pointer to integer
1780  * conversion and vice versa, where in FreeBSD "uintptr_t" would be
1781  * used. Assert these types have the same size, else some parts of the
1782  * LinuxKPI may not work like expected:
1783  */
1784 CTASSERT(sizeof(unsigned long) == sizeof(uintptr_t));
1785