xref: /freebsd/sys/compat/linuxkpi/common/src/linux_compat.c (revision 4ca50eab86aec8001978a612a42f9ae8388eceab)
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 
52 #include <machine/stdarg.h>
53 
54 #if defined(__i386__) || defined(__amd64__)
55 #include <machine/md_var.h>
56 #endif
57 
58 #include <linux/kobject.h>
59 #include <linux/device.h>
60 #include <linux/slab.h>
61 #include <linux/module.h>
62 #include <linux/moduleparam.h>
63 #include <linux/cdev.h>
64 #include <linux/file.h>
65 #include <linux/sysfs.h>
66 #include <linux/mm.h>
67 #include <linux/io.h>
68 #include <linux/vmalloc.h>
69 #include <linux/netdevice.h>
70 #include <linux/timer.h>
71 #include <linux/interrupt.h>
72 #include <linux/uaccess.h>
73 #include <linux/kernel.h>
74 #include <linux/list.h>
75 #include <linux/compat.h>
76 #include <linux/poll.h>
77 
78 #include <vm/vm_pager.h>
79 
80 SYSCTL_NODE(_compat, OID_AUTO, linuxkpi, CTLFLAG_RW, 0, "LinuxKPI parameters");
81 
82 MALLOC_DEFINE(M_KMALLOC, "linux", "Linux kmalloc compat");
83 
84 #include <linux/rbtree.h>
85 /* Undo Linux compat changes. */
86 #undef RB_ROOT
87 #undef file
88 #undef cdev
89 #define	RB_ROOT(head)	(head)->rbh_root
90 
91 static struct vm_area_struct *linux_cdev_handle_find(void *handle);
92 
93 struct kobject linux_class_root;
94 struct device linux_root_device;
95 struct class linux_class_misc;
96 struct list_head pci_drivers;
97 struct list_head pci_devices;
98 spinlock_t pci_lock;
99 
100 unsigned long linux_timer_hz_mask;
101 
102 int
103 panic_cmp(struct rb_node *one, struct rb_node *two)
104 {
105 	panic("no cmp");
106 }
107 
108 RB_GENERATE(linux_root, rb_node, __entry, panic_cmp);
109 
110 int
111 kobject_set_name_vargs(struct kobject *kobj, const char *fmt, va_list args)
112 {
113 	va_list tmp_va;
114 	int len;
115 	char *old;
116 	char *name;
117 	char dummy;
118 
119 	old = kobj->name;
120 
121 	if (old && fmt == NULL)
122 		return (0);
123 
124 	/* compute length of string */
125 	va_copy(tmp_va, args);
126 	len = vsnprintf(&dummy, 0, fmt, tmp_va);
127 	va_end(tmp_va);
128 
129 	/* account for zero termination */
130 	len++;
131 
132 	/* check for error */
133 	if (len < 1)
134 		return (-EINVAL);
135 
136 	/* allocate memory for string */
137 	name = kzalloc(len, GFP_KERNEL);
138 	if (name == NULL)
139 		return (-ENOMEM);
140 	vsnprintf(name, len, fmt, args);
141 	kobj->name = name;
142 
143 	/* free old string */
144 	kfree(old);
145 
146 	/* filter new string */
147 	for (; *name != '\0'; name++)
148 		if (*name == '/')
149 			*name = '!';
150 	return (0);
151 }
152 
153 int
154 kobject_set_name(struct kobject *kobj, const char *fmt, ...)
155 {
156 	va_list args;
157 	int error;
158 
159 	va_start(args, fmt);
160 	error = kobject_set_name_vargs(kobj, fmt, args);
161 	va_end(args);
162 
163 	return (error);
164 }
165 
166 static int
167 kobject_add_complete(struct kobject *kobj, struct kobject *parent)
168 {
169 	const struct kobj_type *t;
170 	int error;
171 
172 	kobj->parent = parent;
173 	error = sysfs_create_dir(kobj);
174 	if (error == 0 && kobj->ktype && kobj->ktype->default_attrs) {
175 		struct attribute **attr;
176 		t = kobj->ktype;
177 
178 		for (attr = t->default_attrs; *attr != NULL; attr++) {
179 			error = sysfs_create_file(kobj, *attr);
180 			if (error)
181 				break;
182 		}
183 		if (error)
184 			sysfs_remove_dir(kobj);
185 
186 	}
187 	return (error);
188 }
189 
190 int
191 kobject_add(struct kobject *kobj, struct kobject *parent, const char *fmt, ...)
192 {
193 	va_list args;
194 	int error;
195 
196 	va_start(args, fmt);
197 	error = kobject_set_name_vargs(kobj, fmt, args);
198 	va_end(args);
199 	if (error)
200 		return (error);
201 
202 	return kobject_add_complete(kobj, parent);
203 }
204 
205 void
206 linux_kobject_release(struct kref *kref)
207 {
208 	struct kobject *kobj;
209 	char *name;
210 
211 	kobj = container_of(kref, struct kobject, kref);
212 	sysfs_remove_dir(kobj);
213 	name = kobj->name;
214 	if (kobj->ktype && kobj->ktype->release)
215 		kobj->ktype->release(kobj);
216 	kfree(name);
217 }
218 
219 static void
220 linux_kobject_kfree(struct kobject *kobj)
221 {
222 	kfree(kobj);
223 }
224 
225 static void
226 linux_kobject_kfree_name(struct kobject *kobj)
227 {
228 	if (kobj) {
229 		kfree(kobj->name);
230 	}
231 }
232 
233 const struct kobj_type linux_kfree_type = {
234 	.release = linux_kobject_kfree
235 };
236 
237 static void
238 linux_device_release(struct device *dev)
239 {
240 	pr_debug("linux_device_release: %s\n", dev_name(dev));
241 	kfree(dev);
242 }
243 
244 static ssize_t
245 linux_class_show(struct kobject *kobj, struct attribute *attr, char *buf)
246 {
247 	struct class_attribute *dattr;
248 	ssize_t error;
249 
250 	dattr = container_of(attr, struct class_attribute, attr);
251 	error = -EIO;
252 	if (dattr->show)
253 		error = dattr->show(container_of(kobj, struct class, kobj),
254 		    dattr, buf);
255 	return (error);
256 }
257 
258 static ssize_t
259 linux_class_store(struct kobject *kobj, struct attribute *attr, const char *buf,
260     size_t count)
261 {
262 	struct class_attribute *dattr;
263 	ssize_t error;
264 
265 	dattr = container_of(attr, struct class_attribute, attr);
266 	error = -EIO;
267 	if (dattr->store)
268 		error = dattr->store(container_of(kobj, struct class, kobj),
269 		    dattr, buf, count);
270 	return (error);
271 }
272 
273 static void
274 linux_class_release(struct kobject *kobj)
275 {
276 	struct class *class;
277 
278 	class = container_of(kobj, struct class, kobj);
279 	if (class->class_release)
280 		class->class_release(class);
281 }
282 
283 static const struct sysfs_ops linux_class_sysfs = {
284 	.show  = linux_class_show,
285 	.store = linux_class_store,
286 };
287 
288 const struct kobj_type linux_class_ktype = {
289 	.release = linux_class_release,
290 	.sysfs_ops = &linux_class_sysfs
291 };
292 
293 static void
294 linux_dev_release(struct kobject *kobj)
295 {
296 	struct device *dev;
297 
298 	dev = container_of(kobj, struct device, kobj);
299 	/* This is the precedence defined by linux. */
300 	if (dev->release)
301 		dev->release(dev);
302 	else if (dev->class && dev->class->dev_release)
303 		dev->class->dev_release(dev);
304 }
305 
306 static ssize_t
307 linux_dev_show(struct kobject *kobj, struct attribute *attr, char *buf)
308 {
309 	struct device_attribute *dattr;
310 	ssize_t error;
311 
312 	dattr = container_of(attr, struct device_attribute, attr);
313 	error = -EIO;
314 	if (dattr->show)
315 		error = dattr->show(container_of(kobj, struct device, kobj),
316 		    dattr, buf);
317 	return (error);
318 }
319 
320 static ssize_t
321 linux_dev_store(struct kobject *kobj, struct attribute *attr, const char *buf,
322     size_t count)
323 {
324 	struct device_attribute *dattr;
325 	ssize_t error;
326 
327 	dattr = container_of(attr, struct device_attribute, attr);
328 	error = -EIO;
329 	if (dattr->store)
330 		error = dattr->store(container_of(kobj, struct device, kobj),
331 		    dattr, buf, count);
332 	return (error);
333 }
334 
335 static const struct sysfs_ops linux_dev_sysfs = {
336 	.show  = linux_dev_show,
337 	.store = linux_dev_store,
338 };
339 
340 const struct kobj_type linux_dev_ktype = {
341 	.release = linux_dev_release,
342 	.sysfs_ops = &linux_dev_sysfs
343 };
344 
345 struct device *
346 device_create(struct class *class, struct device *parent, dev_t devt,
347     void *drvdata, const char *fmt, ...)
348 {
349 	struct device *dev;
350 	va_list args;
351 
352 	dev = kzalloc(sizeof(*dev), M_WAITOK);
353 	dev->parent = parent;
354 	dev->class = class;
355 	dev->devt = devt;
356 	dev->driver_data = drvdata;
357 	dev->release = linux_device_release;
358 	va_start(args, fmt);
359 	kobject_set_name_vargs(&dev->kobj, fmt, args);
360 	va_end(args);
361 	device_register(dev);
362 
363 	return (dev);
364 }
365 
366 int
367 kobject_init_and_add(struct kobject *kobj, const struct kobj_type *ktype,
368     struct kobject *parent, const char *fmt, ...)
369 {
370 	va_list args;
371 	int error;
372 
373 	kobject_init(kobj, ktype);
374 	kobj->ktype = ktype;
375 	kobj->parent = parent;
376 	kobj->name = NULL;
377 
378 	va_start(args, fmt);
379 	error = kobject_set_name_vargs(kobj, fmt, args);
380 	va_end(args);
381 	if (error)
382 		return (error);
383 	return kobject_add_complete(kobj, parent);
384 }
385 
386 static void
387 linux_file_dtor(void *cdp)
388 {
389 	struct linux_file *filp;
390 
391 	linux_set_current(curthread);
392 	filp = cdp;
393 	filp->f_op->release(filp->f_vnode, filp);
394 	vdrop(filp->f_vnode);
395 	kfree(filp);
396 }
397 
398 static int
399 linux_cdev_pager_populate(vm_object_t vm_obj, vm_pindex_t pidx, int fault_type,
400     vm_prot_t max_prot, vm_pindex_t *first, vm_pindex_t *last)
401 {
402 	struct vm_area_struct *vmap;
403 	struct vm_fault vmf;
404 	int err;
405 
406 	linux_set_current(curthread);
407 
408 	/* get VM area structure */
409 	vmap = linux_cdev_handle_find(vm_obj->handle);
410 	MPASS(vmap != NULL);
411 	MPASS(vmap->vm_private_data == vm_obj->handle);
412 
413 	/* fill out VM fault structure */
414 	vmf.virtual_address = (void *)((uintptr_t)pidx << PAGE_SHIFT);
415 	vmf.flags = (fault_type & VM_PROT_WRITE) ? FAULT_FLAG_WRITE : 0;
416 	vmf.pgoff = 0;
417 	vmf.page = NULL;
418 
419 	VM_OBJECT_WUNLOCK(vm_obj);
420 
421 	down_write(&vmap->vm_mm->mmap_sem);
422 	if (unlikely(vmap->vm_ops == NULL)) {
423 		err = VM_FAULT_SIGBUS;
424 	} else {
425 		vmap->vm_pfn_count = 0;
426 		vmap->vm_pfn_pcount = &vmap->vm_pfn_count;
427 		vmap->vm_obj = vm_obj;
428 
429 		err = vmap->vm_ops->fault(vmap, &vmf);
430 
431 		while (vmap->vm_pfn_count == 0 && err == VM_FAULT_NOPAGE) {
432 			kern_yield(0);
433 			err = vmap->vm_ops->fault(vmap, &vmf);
434 		}
435 	}
436 
437 	/* translate return code */
438 	switch (err) {
439 	case VM_FAULT_OOM:
440 		err = VM_PAGER_AGAIN;
441 		break;
442 	case VM_FAULT_SIGBUS:
443 		err = VM_PAGER_BAD;
444 		break;
445 	case VM_FAULT_NOPAGE:
446 		/*
447 		 * By contract the fault handler will return having
448 		 * busied all the pages itself. If pidx is already
449 		 * found in the object, it will simply xbusy the first
450 		 * page and return with vm_pfn_count set to 1.
451 		 */
452 		*first = vmap->vm_pfn_first;
453 		*last = *first + vmap->vm_pfn_count - 1;
454 		err = VM_PAGER_OK;
455 		break;
456 	default:
457 		err = VM_PAGER_ERROR;
458 		break;
459 	}
460 	up_write(&vmap->vm_mm->mmap_sem);
461 	VM_OBJECT_WLOCK(vm_obj);
462 	return (err);
463 }
464 
465 static struct rwlock linux_vma_lock;
466 static TAILQ_HEAD(, vm_area_struct) linux_vma_head =
467     TAILQ_HEAD_INITIALIZER(linux_vma_head);
468 
469 static struct vm_area_struct *
470 linux_cdev_handle_insert(void *handle, struct vm_area_struct *vmap)
471 {
472 	struct vm_area_struct *ptr;
473 
474 	rw_wlock(&linux_vma_lock);
475 	TAILQ_FOREACH(ptr, &linux_vma_head, vm_entry) {
476 		if (ptr->vm_private_data == handle) {
477 			rw_wunlock(&linux_vma_lock);
478 			kfree(vmap);
479 			return (NULL);
480 		}
481 	}
482 	TAILQ_INSERT_TAIL(&linux_vma_head, vmap, vm_entry);
483 	rw_wunlock(&linux_vma_lock);
484 	return (vmap);
485 }
486 
487 static void
488 linux_cdev_handle_remove(struct vm_area_struct *vmap)
489 {
490 	if (vmap == NULL)
491 		return;
492 
493 	rw_wlock(&linux_vma_lock);
494 	TAILQ_REMOVE(&linux_vma_head, vmap, vm_entry);
495 	rw_wunlock(&linux_vma_lock);
496 	kfree(vmap);
497 }
498 
499 static struct vm_area_struct *
500 linux_cdev_handle_find(void *handle)
501 {
502 	struct vm_area_struct *vmap;
503 
504 	rw_rlock(&linux_vma_lock);
505 	TAILQ_FOREACH(vmap, &linux_vma_head, vm_entry) {
506 		if (vmap->vm_private_data == handle)
507 			break;
508 	}
509 	rw_runlock(&linux_vma_lock);
510 	return (vmap);
511 }
512 
513 static int
514 linux_cdev_pager_ctor(void *handle, vm_ooffset_t size, vm_prot_t prot,
515 		      vm_ooffset_t foff, struct ucred *cred, u_short *color)
516 {
517 	const struct vm_operations_struct *vm_ops;
518 	struct vm_area_struct *vmap;
519 
520 	vmap = linux_cdev_handle_find(handle);
521 	MPASS(vmap != NULL);
522 
523 	*color = 0;
524 
525 	down_write(&vmap->vm_mm->mmap_sem);
526 	vm_ops = vmap->vm_ops;
527 	if (likely(vm_ops != NULL))
528 		vm_ops->open(vmap);
529 	up_write(&vmap->vm_mm->mmap_sem);
530 
531 	return (0);
532 }
533 
534 static void
535 linux_cdev_pager_dtor(void *handle)
536 {
537 	const struct vm_operations_struct *vm_ops;
538 	struct vm_area_struct *vmap;
539 
540 	vmap = linux_cdev_handle_find(handle);
541 	MPASS(vmap != NULL);
542 
543 	down_write(&vmap->vm_mm->mmap_sem);
544 	vm_ops = vmap->vm_ops;
545 	if (likely(vm_ops != NULL))
546 		vm_ops->close(vmap);
547 	up_write(&vmap->vm_mm->mmap_sem);
548 
549 	linux_cdev_handle_remove(vmap);
550 }
551 
552 static struct cdev_pager_ops linux_cdev_pager_ops = {
553 	.cdev_pg_populate	= linux_cdev_pager_populate,
554 	.cdev_pg_ctor	= linux_cdev_pager_ctor,
555 	.cdev_pg_dtor	= linux_cdev_pager_dtor
556 };
557 
558 static int
559 linux_dev_open(struct cdev *dev, int oflags, int devtype, struct thread *td)
560 {
561 	struct linux_cdev *ldev;
562 	struct linux_file *filp;
563 	struct file *file;
564 	int error;
565 
566 	file = td->td_fpop;
567 	ldev = dev->si_drv1;
568 	if (ldev == NULL)
569 		return (ENODEV);
570 	filp = kzalloc(sizeof(*filp), GFP_KERNEL);
571 	filp->f_dentry = &filp->f_dentry_store;
572 	filp->f_op = ldev->ops;
573 	filp->f_flags = file->f_flag;
574 	vhold(file->f_vnode);
575 	filp->f_vnode = file->f_vnode;
576 	linux_set_current(td);
577 	if (filp->f_op->open) {
578 		error = -filp->f_op->open(file->f_vnode, filp);
579 		if (error) {
580 			kfree(filp);
581 			goto done;
582 		}
583 	}
584 	error = devfs_set_cdevpriv(filp, linux_file_dtor);
585 	if (error) {
586 		filp->f_op->release(file->f_vnode, filp);
587 		kfree(filp);
588 	}
589 done:
590 	return (error);
591 }
592 
593 static int
594 linux_dev_close(struct cdev *dev, int fflag, int devtype, struct thread *td)
595 {
596 	struct linux_file *filp;
597 	struct file *file;
598 	int error;
599 
600 	file = td->td_fpop;
601 	if (dev->si_drv1 == NULL)
602 		return (0);
603 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
604 		return (error);
605 	filp->f_flags = file->f_flag;
606         devfs_clear_cdevpriv();
607 
608 
609 	return (0);
610 }
611 
612 #define	LINUX_IOCTL_MIN_PTR 0x10000UL
613 #define	LINUX_IOCTL_MAX_PTR (LINUX_IOCTL_MIN_PTR + IOCPARM_MAX)
614 
615 static inline int
616 linux_remap_address(void **uaddr, size_t len)
617 {
618 	uintptr_t uaddr_val = (uintptr_t)(*uaddr);
619 
620 	if (unlikely(uaddr_val >= LINUX_IOCTL_MIN_PTR &&
621 	    uaddr_val < LINUX_IOCTL_MAX_PTR)) {
622 		struct task_struct *pts = current;
623 		if (pts == NULL) {
624 			*uaddr = NULL;
625 			return (1);
626 		}
627 
628 		/* compute data offset */
629 		uaddr_val -= LINUX_IOCTL_MIN_PTR;
630 
631 		/* check that length is within bounds */
632 		if ((len > IOCPARM_MAX) ||
633 		    (uaddr_val + len) > pts->bsd_ioctl_len) {
634 			*uaddr = NULL;
635 			return (1);
636 		}
637 
638 		/* re-add kernel buffer address */
639 		uaddr_val += (uintptr_t)pts->bsd_ioctl_data;
640 
641 		/* update address location */
642 		*uaddr = (void *)uaddr_val;
643 		return (1);
644 	}
645 	return (0);
646 }
647 
648 int
649 linux_copyin(const void *uaddr, void *kaddr, size_t len)
650 {
651 	if (linux_remap_address(__DECONST(void **, &uaddr), len)) {
652 		if (uaddr == NULL)
653 			return (-EFAULT);
654 		memcpy(kaddr, uaddr, len);
655 		return (0);
656 	}
657 	return (-copyin(uaddr, kaddr, len));
658 }
659 
660 int
661 linux_copyout(const void *kaddr, void *uaddr, size_t len)
662 {
663 	if (linux_remap_address(&uaddr, len)) {
664 		if (uaddr == NULL)
665 			return (-EFAULT);
666 		memcpy(uaddr, kaddr, len);
667 		return (0);
668 	}
669 	return (-copyout(kaddr, uaddr, len));
670 }
671 
672 size_t
673 linux_clear_user(void *_uaddr, size_t _len)
674 {
675 	uint8_t *uaddr = _uaddr;
676 	size_t len = _len;
677 
678 	/* make sure uaddr is aligned before going into the fast loop */
679 	while (((uintptr_t)uaddr & 7) != 0 && len > 7) {
680 		if (subyte(uaddr, 0))
681 			return (_len);
682 		uaddr++;
683 		len--;
684 	}
685 
686 	/* zero 8 bytes at a time */
687 	while (len > 7) {
688 #ifdef __LP64__
689 		if (suword64(uaddr, 0))
690 			return (_len);
691 #else
692 		if (suword32(uaddr, 0))
693 			return (_len);
694 		if (suword32(uaddr + 4, 0))
695 			return (_len);
696 #endif
697 		uaddr += 8;
698 		len -= 8;
699 	}
700 
701 	/* zero fill end, if any */
702 	while (len > 0) {
703 		if (subyte(uaddr, 0))
704 			return (_len);
705 		uaddr++;
706 		len--;
707 	}
708 	return (0);
709 }
710 
711 int
712 linux_access_ok(int rw, const void *uaddr, size_t len)
713 {
714 	uintptr_t saddr;
715 	uintptr_t eaddr;
716 
717 	/* get start and end address */
718 	saddr = (uintptr_t)uaddr;
719 	eaddr = (uintptr_t)uaddr + len;
720 
721 	/* verify addresses are valid for userspace */
722 	return ((saddr == eaddr) ||
723 	    (eaddr > saddr && eaddr <= VM_MAXUSER_ADDRESS));
724 }
725 
726 static int
727 linux_dev_ioctl(struct cdev *dev, u_long cmd, caddr_t data, int fflag,
728     struct thread *td)
729 {
730 	struct linux_file *filp;
731 	struct file *file;
732 	unsigned size;
733 	int error;
734 
735 	file = td->td_fpop;
736 	if (dev->si_drv1 == NULL)
737 		return (ENXIO);
738 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
739 		return (error);
740 	filp->f_flags = file->f_flag;
741 
742 	linux_set_current(td);
743 	size = IOCPARM_LEN(cmd);
744 	/* refer to logic in sys_ioctl() */
745 	if (size > 0) {
746 		/*
747 		 * Setup hint for linux_copyin() and linux_copyout().
748 		 *
749 		 * Background: Linux code expects a user-space address
750 		 * while FreeBSD supplies a kernel-space address.
751 		 */
752 		current->bsd_ioctl_data = data;
753 		current->bsd_ioctl_len = size;
754 		data = (void *)LINUX_IOCTL_MIN_PTR;
755 	} else {
756 		/* fetch user-space pointer */
757 		data = *(void **)data;
758 	}
759 	if (filp->f_op->unlocked_ioctl)
760 		error = -filp->f_op->unlocked_ioctl(filp, cmd, (u_long)data);
761 	else
762 		error = ENOTTY;
763 	if (size > 0) {
764 		current->bsd_ioctl_data = NULL;
765 		current->bsd_ioctl_len = 0;
766 	}
767 
768 	return (error);
769 }
770 
771 static int
772 linux_dev_read(struct cdev *dev, struct uio *uio, int ioflag)
773 {
774 	struct linux_file *filp;
775 	struct thread *td;
776 	struct file *file;
777 	ssize_t bytes;
778 	int error;
779 
780 	td = curthread;
781 	file = td->td_fpop;
782 	if (dev->si_drv1 == NULL)
783 		return (ENXIO);
784 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
785 		return (error);
786 	filp->f_flags = file->f_flag;
787 	/* XXX no support for I/O vectors currently */
788 	if (uio->uio_iovcnt != 1)
789 		return (EOPNOTSUPP);
790 	linux_set_current(td);
791 	if (filp->f_op->read) {
792 		bytes = filp->f_op->read(filp, uio->uio_iov->iov_base,
793 		    uio->uio_iov->iov_len, &uio->uio_offset);
794 		if (bytes >= 0) {
795 			uio->uio_iov->iov_base =
796 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
797 			uio->uio_iov->iov_len -= bytes;
798 			uio->uio_resid -= bytes;
799 		} else
800 			error = -bytes;
801 	} else
802 		error = ENXIO;
803 
804 	return (error);
805 }
806 
807 static int
808 linux_dev_write(struct cdev *dev, struct uio *uio, int ioflag)
809 {
810 	struct linux_file *filp;
811 	struct thread *td;
812 	struct file *file;
813 	ssize_t bytes;
814 	int error;
815 
816 	td = curthread;
817 	file = td->td_fpop;
818 	if (dev->si_drv1 == NULL)
819 		return (ENXIO);
820 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
821 		return (error);
822 	filp->f_flags = file->f_flag;
823 	/* XXX no support for I/O vectors currently */
824 	if (uio->uio_iovcnt != 1)
825 		return (EOPNOTSUPP);
826 	linux_set_current(td);
827 	if (filp->f_op->write) {
828 		bytes = filp->f_op->write(filp, uio->uio_iov->iov_base,
829 		    uio->uio_iov->iov_len, &uio->uio_offset);
830 		if (bytes >= 0) {
831 			uio->uio_iov->iov_base =
832 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
833 			uio->uio_iov->iov_len -= bytes;
834 			uio->uio_resid -= bytes;
835 		} else
836 			error = -bytes;
837 	} else
838 		error = ENXIO;
839 
840 	return (error);
841 }
842 
843 static int
844 linux_dev_poll(struct cdev *dev, int events, struct thread *td)
845 {
846 	struct linux_file *filp;
847 	struct file *file;
848 	int revents;
849 
850 	if (dev->si_drv1 == NULL)
851 		goto error;
852 	if (devfs_get_cdevpriv((void **)&filp) != 0)
853 		goto error;
854 
855 	file = td->td_fpop;
856 	filp->f_flags = file->f_flag;
857 	linux_set_current(td);
858 	if (filp->f_op->poll)
859 		revents = filp->f_op->poll(filp, NULL) & events;
860 	else
861 		revents = 0;
862 
863 	return (revents);
864 error:
865 	return (events & (POLLHUP|POLLIN|POLLRDNORM|POLLOUT|POLLWRNORM));
866 }
867 
868 static int
869 linux_dev_mmap_single(struct cdev *dev, vm_ooffset_t *offset,
870     vm_size_t size, struct vm_object **object, int nprot)
871 {
872 	struct vm_area_struct *vmap;
873 	struct linux_file *filp;
874 	struct thread *td;
875 	struct file *file;
876 	vm_memattr_t attr;
877 	int error;
878 
879 	td = curthread;
880 	file = td->td_fpop;
881 	if (dev->si_drv1 == NULL)
882 		return (ENODEV);
883 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
884 		return (error);
885 	filp->f_flags = file->f_flag;
886 
887 	if (filp->f_op->mmap == NULL)
888 		return (ENODEV);
889 
890 	linux_set_current(td);
891 
892 	vmap = kzalloc(sizeof(*vmap), GFP_KERNEL);
893 	vmap->vm_start = 0;
894 	vmap->vm_end = size;
895 	vmap->vm_pgoff = *offset / PAGE_SIZE;
896 	vmap->vm_pfn = 0;
897 	vmap->vm_flags = vmap->vm_page_prot = nprot;
898 	vmap->vm_ops = NULL;
899 	vmap->vm_file = filp;
900 	vmap->vm_mm = current->mm;
901 
902 	if (unlikely(down_write_killable(&vmap->vm_mm->mmap_sem))) {
903 		error = EINTR;
904 	} else {
905 		error = -filp->f_op->mmap(filp, vmap);
906 		up_write(&vmap->vm_mm->mmap_sem);
907 	}
908 
909 	if (error != 0) {
910 		kfree(vmap);
911 		return (error);
912 	}
913 
914 	attr = pgprot2cachemode(vmap->vm_page_prot);
915 
916 	if (vmap->vm_ops != NULL) {
917 		void *vm_private_data;
918 
919 		if (vmap->vm_ops->fault == NULL ||
920 		    vmap->vm_ops->open == NULL ||
921 		    vmap->vm_ops->close == NULL ||
922 		    vmap->vm_private_data == NULL) {
923 			kfree(vmap);
924 			return (EINVAL);
925 		}
926 
927 		vm_private_data = vmap->vm_private_data;
928 
929 		vmap = linux_cdev_handle_insert(vm_private_data, vmap);
930 
931 		*object = cdev_pager_allocate(vm_private_data, OBJT_MGTDEVICE,
932 		    &linux_cdev_pager_ops, size, nprot, *offset, curthread->td_ucred);
933 
934 		if (*object == NULL) {
935 			linux_cdev_handle_remove(vmap);
936 			return (EINVAL);
937 		}
938 	} else {
939 		struct sglist *sg;
940 
941 		sg = sglist_alloc(1, M_WAITOK);
942 		sglist_append_phys(sg, (vm_paddr_t)vmap->vm_pfn << PAGE_SHIFT, vmap->vm_len);
943 
944 		*object = vm_pager_allocate(OBJT_SG, sg, vmap->vm_len,
945 		    nprot, 0, curthread->td_ucred);
946 
947 		kfree(vmap);
948 
949 		if (*object == NULL) {
950 			sglist_free(sg);
951 			return (EINVAL);
952 		}
953 	}
954 
955 	if (attr != VM_MEMATTR_DEFAULT) {
956 		VM_OBJECT_WLOCK(*object);
957 		vm_object_set_memattr(*object, attr);
958 		VM_OBJECT_WUNLOCK(*object);
959 	}
960 	*offset = 0;
961 	return (0);
962 }
963 
964 struct cdevsw linuxcdevsw = {
965 	.d_version = D_VERSION,
966 	.d_flags = D_TRACKCLOSE,
967 	.d_open = linux_dev_open,
968 	.d_close = linux_dev_close,
969 	.d_read = linux_dev_read,
970 	.d_write = linux_dev_write,
971 	.d_ioctl = linux_dev_ioctl,
972 	.d_mmap_single = linux_dev_mmap_single,
973 	.d_poll = linux_dev_poll,
974 };
975 
976 static int
977 linux_file_read(struct file *file, struct uio *uio, struct ucred *active_cred,
978     int flags, struct thread *td)
979 {
980 	struct linux_file *filp;
981 	ssize_t bytes;
982 	int error;
983 
984 	error = 0;
985 	filp = (struct linux_file *)file->f_data;
986 	filp->f_flags = file->f_flag;
987 	/* XXX no support for I/O vectors currently */
988 	if (uio->uio_iovcnt != 1)
989 		return (EOPNOTSUPP);
990 	linux_set_current(td);
991 	if (filp->f_op->read) {
992 		bytes = filp->f_op->read(filp, uio->uio_iov->iov_base,
993 		    uio->uio_iov->iov_len, &uio->uio_offset);
994 		if (bytes >= 0) {
995 			uio->uio_iov->iov_base =
996 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
997 			uio->uio_iov->iov_len -= bytes;
998 			uio->uio_resid -= bytes;
999 		} else
1000 			error = -bytes;
1001 	} else
1002 		error = ENXIO;
1003 
1004 	return (error);
1005 }
1006 
1007 static int
1008 linux_file_poll(struct file *file, int events, struct ucred *active_cred,
1009     struct thread *td)
1010 {
1011 	struct linux_file *filp;
1012 	int revents;
1013 
1014 	filp = (struct linux_file *)file->f_data;
1015 	filp->f_flags = file->f_flag;
1016 	linux_set_current(td);
1017 	if (filp->f_op->poll)
1018 		revents = filp->f_op->poll(filp, NULL) & events;
1019 	else
1020 		revents = 0;
1021 
1022 	return (revents);
1023 }
1024 
1025 static int
1026 linux_file_close(struct file *file, struct thread *td)
1027 {
1028 	struct linux_file *filp;
1029 	int error;
1030 
1031 	filp = (struct linux_file *)file->f_data;
1032 	filp->f_flags = file->f_flag;
1033 	linux_set_current(td);
1034 	error = -filp->f_op->release(NULL, filp);
1035 	funsetown(&filp->f_sigio);
1036 	kfree(filp);
1037 
1038 	return (error);
1039 }
1040 
1041 static int
1042 linux_file_ioctl(struct file *fp, u_long cmd, void *data, struct ucred *cred,
1043     struct thread *td)
1044 {
1045 	struct linux_file *filp;
1046 	int error;
1047 
1048 	filp = (struct linux_file *)fp->f_data;
1049 	filp->f_flags = fp->f_flag;
1050 	error = 0;
1051 
1052 	linux_set_current(td);
1053 	switch (cmd) {
1054 	case FIONBIO:
1055 		break;
1056 	case FIOASYNC:
1057 		if (filp->f_op->fasync == NULL)
1058 			break;
1059 		error = filp->f_op->fasync(0, filp, fp->f_flag & FASYNC);
1060 		break;
1061 	case FIOSETOWN:
1062 		error = fsetown(*(int *)data, &filp->f_sigio);
1063 		if (error == 0)
1064 			error = filp->f_op->fasync(0, filp,
1065 			    fp->f_flag & FASYNC);
1066 		break;
1067 	case FIOGETOWN:
1068 		*(int *)data = fgetown(&filp->f_sigio);
1069 		break;
1070 	default:
1071 		error = ENOTTY;
1072 		break;
1073 	}
1074 	return (error);
1075 }
1076 
1077 static int
1078 linux_file_stat(struct file *fp, struct stat *sb, struct ucred *active_cred,
1079     struct thread *td)
1080 {
1081 
1082 	return (EOPNOTSUPP);
1083 }
1084 
1085 static int
1086 linux_file_fill_kinfo(struct file *fp, struct kinfo_file *kif,
1087     struct filedesc *fdp)
1088 {
1089 
1090 	return (0);
1091 }
1092 
1093 struct fileops linuxfileops = {
1094 	.fo_read = linux_file_read,
1095 	.fo_write = invfo_rdwr,
1096 	.fo_truncate = invfo_truncate,
1097 	.fo_kqfilter = invfo_kqfilter,
1098 	.fo_stat = linux_file_stat,
1099 	.fo_fill_kinfo = linux_file_fill_kinfo,
1100 	.fo_poll = linux_file_poll,
1101 	.fo_close = linux_file_close,
1102 	.fo_ioctl = linux_file_ioctl,
1103 	.fo_chmod = invfo_chmod,
1104 	.fo_chown = invfo_chown,
1105 	.fo_sendfile = invfo_sendfile,
1106 };
1107 
1108 /*
1109  * Hash of vmmap addresses.  This is infrequently accessed and does not
1110  * need to be particularly large.  This is done because we must store the
1111  * caller's idea of the map size to properly unmap.
1112  */
1113 struct vmmap {
1114 	LIST_ENTRY(vmmap)	vm_next;
1115 	void 			*vm_addr;
1116 	unsigned long		vm_size;
1117 };
1118 
1119 struct vmmaphd {
1120 	struct vmmap *lh_first;
1121 };
1122 #define	VMMAP_HASH_SIZE	64
1123 #define	VMMAP_HASH_MASK	(VMMAP_HASH_SIZE - 1)
1124 #define	VM_HASH(addr)	((uintptr_t)(addr) >> PAGE_SHIFT) & VMMAP_HASH_MASK
1125 static struct vmmaphd vmmaphead[VMMAP_HASH_SIZE];
1126 static struct mtx vmmaplock;
1127 
1128 static void
1129 vmmap_add(void *addr, unsigned long size)
1130 {
1131 	struct vmmap *vmmap;
1132 
1133 	vmmap = kmalloc(sizeof(*vmmap), GFP_KERNEL);
1134 	mtx_lock(&vmmaplock);
1135 	vmmap->vm_size = size;
1136 	vmmap->vm_addr = addr;
1137 	LIST_INSERT_HEAD(&vmmaphead[VM_HASH(addr)], vmmap, vm_next);
1138 	mtx_unlock(&vmmaplock);
1139 }
1140 
1141 static struct vmmap *
1142 vmmap_remove(void *addr)
1143 {
1144 	struct vmmap *vmmap;
1145 
1146 	mtx_lock(&vmmaplock);
1147 	LIST_FOREACH(vmmap, &vmmaphead[VM_HASH(addr)], vm_next)
1148 		if (vmmap->vm_addr == addr)
1149 			break;
1150 	if (vmmap)
1151 		LIST_REMOVE(vmmap, vm_next);
1152 	mtx_unlock(&vmmaplock);
1153 
1154 	return (vmmap);
1155 }
1156 
1157 #if defined(__i386__) || defined(__amd64__)
1158 void *
1159 _ioremap_attr(vm_paddr_t phys_addr, unsigned long size, int attr)
1160 {
1161 	void *addr;
1162 
1163 	addr = pmap_mapdev_attr(phys_addr, size, attr);
1164 	if (addr == NULL)
1165 		return (NULL);
1166 	vmmap_add(addr, size);
1167 
1168 	return (addr);
1169 }
1170 #endif
1171 
1172 void
1173 iounmap(void *addr)
1174 {
1175 	struct vmmap *vmmap;
1176 
1177 	vmmap = vmmap_remove(addr);
1178 	if (vmmap == NULL)
1179 		return;
1180 #if defined(__i386__) || defined(__amd64__)
1181 	pmap_unmapdev((vm_offset_t)addr, vmmap->vm_size);
1182 #endif
1183 	kfree(vmmap);
1184 }
1185 
1186 
1187 void *
1188 vmap(struct page **pages, unsigned int count, unsigned long flags, int prot)
1189 {
1190 	vm_offset_t off;
1191 	size_t size;
1192 
1193 	size = count * PAGE_SIZE;
1194 	off = kva_alloc(size);
1195 	if (off == 0)
1196 		return (NULL);
1197 	vmmap_add((void *)off, size);
1198 	pmap_qenter(off, pages, count);
1199 
1200 	return ((void *)off);
1201 }
1202 
1203 void
1204 vunmap(void *addr)
1205 {
1206 	struct vmmap *vmmap;
1207 
1208 	vmmap = vmmap_remove(addr);
1209 	if (vmmap == NULL)
1210 		return;
1211 	pmap_qremove((vm_offset_t)addr, vmmap->vm_size / PAGE_SIZE);
1212 	kva_free((vm_offset_t)addr, vmmap->vm_size);
1213 	kfree(vmmap);
1214 }
1215 
1216 char *
1217 kvasprintf(gfp_t gfp, const char *fmt, va_list ap)
1218 {
1219 	unsigned int len;
1220 	char *p;
1221 	va_list aq;
1222 
1223 	va_copy(aq, ap);
1224 	len = vsnprintf(NULL, 0, fmt, aq);
1225 	va_end(aq);
1226 
1227 	p = kmalloc(len + 1, gfp);
1228 	if (p != NULL)
1229 		vsnprintf(p, len + 1, fmt, ap);
1230 
1231 	return (p);
1232 }
1233 
1234 char *
1235 kasprintf(gfp_t gfp, const char *fmt, ...)
1236 {
1237 	va_list ap;
1238 	char *p;
1239 
1240 	va_start(ap, fmt);
1241 	p = kvasprintf(gfp, fmt, ap);
1242 	va_end(ap);
1243 
1244 	return (p);
1245 }
1246 
1247 static void
1248 linux_timer_callback_wrapper(void *context)
1249 {
1250 	struct timer_list *timer;
1251 
1252 	linux_set_current(curthread);
1253 
1254 	timer = context;
1255 	timer->function(timer->data);
1256 }
1257 
1258 void
1259 mod_timer(struct timer_list *timer, unsigned long expires)
1260 {
1261 
1262 	timer->expires = expires;
1263 	callout_reset(&timer->timer_callout,
1264 	    linux_timer_jiffies_until(expires),
1265 	    &linux_timer_callback_wrapper, timer);
1266 }
1267 
1268 void
1269 add_timer(struct timer_list *timer)
1270 {
1271 
1272 	callout_reset(&timer->timer_callout,
1273 	    linux_timer_jiffies_until(timer->expires),
1274 	    &linux_timer_callback_wrapper, timer);
1275 }
1276 
1277 void
1278 add_timer_on(struct timer_list *timer, int cpu)
1279 {
1280 
1281 	callout_reset_on(&timer->timer_callout,
1282 	    linux_timer_jiffies_until(timer->expires),
1283 	    &linux_timer_callback_wrapper, timer, cpu);
1284 }
1285 
1286 static void
1287 linux_timer_init(void *arg)
1288 {
1289 
1290 	/*
1291 	 * Compute an internal HZ value which can divide 2**32 to
1292 	 * avoid timer rounding problems when the tick value wraps
1293 	 * around 2**32:
1294 	 */
1295 	linux_timer_hz_mask = 1;
1296 	while (linux_timer_hz_mask < (unsigned long)hz)
1297 		linux_timer_hz_mask *= 2;
1298 	linux_timer_hz_mask--;
1299 }
1300 SYSINIT(linux_timer, SI_SUB_DRIVERS, SI_ORDER_FIRST, linux_timer_init, NULL);
1301 
1302 void
1303 linux_complete_common(struct completion *c, int all)
1304 {
1305 	int wakeup_swapper;
1306 
1307 	sleepq_lock(c);
1308 	c->done++;
1309 	if (all)
1310 		wakeup_swapper = sleepq_broadcast(c, SLEEPQ_SLEEP, 0, 0);
1311 	else
1312 		wakeup_swapper = sleepq_signal(c, SLEEPQ_SLEEP, 0, 0);
1313 	sleepq_release(c);
1314 	if (wakeup_swapper)
1315 		kick_proc0();
1316 }
1317 
1318 /*
1319  * Indefinite wait for done != 0 with or without signals.
1320  */
1321 long
1322 linux_wait_for_common(struct completion *c, int flags)
1323 {
1324 	if (SCHEDULER_STOPPED())
1325 		return (0);
1326 
1327 	if (flags != 0)
1328 		flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
1329 	else
1330 		flags = SLEEPQ_SLEEP;
1331 	for (;;) {
1332 		sleepq_lock(c);
1333 		if (c->done)
1334 			break;
1335 		sleepq_add(c, NULL, "completion", flags, 0);
1336 		if (flags & SLEEPQ_INTERRUPTIBLE) {
1337 			if (sleepq_wait_sig(c, 0) != 0)
1338 				return (-ERESTARTSYS);
1339 		} else
1340 			sleepq_wait(c, 0);
1341 	}
1342 	c->done--;
1343 	sleepq_release(c);
1344 
1345 	return (0);
1346 }
1347 
1348 /*
1349  * Time limited wait for done != 0 with or without signals.
1350  */
1351 long
1352 linux_wait_for_timeout_common(struct completion *c, long timeout, int flags)
1353 {
1354 	long end = jiffies + timeout;
1355 
1356 	if (SCHEDULER_STOPPED())
1357 		return (0);
1358 
1359 	if (flags != 0)
1360 		flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
1361 	else
1362 		flags = SLEEPQ_SLEEP;
1363 	for (;;) {
1364 		int ret;
1365 
1366 		sleepq_lock(c);
1367 		if (c->done)
1368 			break;
1369 		sleepq_add(c, NULL, "completion", flags, 0);
1370 		sleepq_set_timeout(c, linux_timer_jiffies_until(end));
1371 		if (flags & SLEEPQ_INTERRUPTIBLE)
1372 			ret = sleepq_timedwait_sig(c, 0);
1373 		else
1374 			ret = sleepq_timedwait(c, 0);
1375 		if (ret != 0) {
1376 			/* check for timeout or signal */
1377 			if (ret == EWOULDBLOCK)
1378 				return (0);
1379 			else
1380 				return (-ERESTARTSYS);
1381 		}
1382 	}
1383 	c->done--;
1384 	sleepq_release(c);
1385 
1386 	/* return how many jiffies are left */
1387 	return (linux_timer_jiffies_until(end));
1388 }
1389 
1390 int
1391 linux_try_wait_for_completion(struct completion *c)
1392 {
1393 	int isdone;
1394 
1395 	isdone = 1;
1396 	sleepq_lock(c);
1397 	if (c->done)
1398 		c->done--;
1399 	else
1400 		isdone = 0;
1401 	sleepq_release(c);
1402 	return (isdone);
1403 }
1404 
1405 int
1406 linux_completion_done(struct completion *c)
1407 {
1408 	int isdone;
1409 
1410 	isdone = 1;
1411 	sleepq_lock(c);
1412 	if (c->done == 0)
1413 		isdone = 0;
1414 	sleepq_release(c);
1415 	return (isdone);
1416 }
1417 
1418 static void
1419 linux_cdev_release(struct kobject *kobj)
1420 {
1421 	struct linux_cdev *cdev;
1422 	struct kobject *parent;
1423 
1424 	cdev = container_of(kobj, struct linux_cdev, kobj);
1425 	parent = kobj->parent;
1426 	if (cdev->cdev)
1427 		destroy_dev(cdev->cdev);
1428 	kfree(cdev);
1429 	kobject_put(parent);
1430 }
1431 
1432 static void
1433 linux_cdev_static_release(struct kobject *kobj)
1434 {
1435 	struct linux_cdev *cdev;
1436 	struct kobject *parent;
1437 
1438 	cdev = container_of(kobj, struct linux_cdev, kobj);
1439 	parent = kobj->parent;
1440 	if (cdev->cdev)
1441 		destroy_dev(cdev->cdev);
1442 	kobject_put(parent);
1443 }
1444 
1445 const struct kobj_type linux_cdev_ktype = {
1446 	.release = linux_cdev_release,
1447 };
1448 
1449 const struct kobj_type linux_cdev_static_ktype = {
1450 	.release = linux_cdev_static_release,
1451 };
1452 
1453 static void
1454 linux_handle_ifnet_link_event(void *arg, struct ifnet *ifp, int linkstate)
1455 {
1456 	struct notifier_block *nb;
1457 
1458 	nb = arg;
1459 	if (linkstate == LINK_STATE_UP)
1460 		nb->notifier_call(nb, NETDEV_UP, ifp);
1461 	else
1462 		nb->notifier_call(nb, NETDEV_DOWN, ifp);
1463 }
1464 
1465 static void
1466 linux_handle_ifnet_arrival_event(void *arg, struct ifnet *ifp)
1467 {
1468 	struct notifier_block *nb;
1469 
1470 	nb = arg;
1471 	nb->notifier_call(nb, NETDEV_REGISTER, ifp);
1472 }
1473 
1474 static void
1475 linux_handle_ifnet_departure_event(void *arg, struct ifnet *ifp)
1476 {
1477 	struct notifier_block *nb;
1478 
1479 	nb = arg;
1480 	nb->notifier_call(nb, NETDEV_UNREGISTER, ifp);
1481 }
1482 
1483 static void
1484 linux_handle_iflladdr_event(void *arg, struct ifnet *ifp)
1485 {
1486 	struct notifier_block *nb;
1487 
1488 	nb = arg;
1489 	nb->notifier_call(nb, NETDEV_CHANGEADDR, ifp);
1490 }
1491 
1492 static void
1493 linux_handle_ifaddr_event(void *arg, struct ifnet *ifp)
1494 {
1495 	struct notifier_block *nb;
1496 
1497 	nb = arg;
1498 	nb->notifier_call(nb, NETDEV_CHANGEIFADDR, ifp);
1499 }
1500 
1501 int
1502 register_netdevice_notifier(struct notifier_block *nb)
1503 {
1504 
1505 	nb->tags[NETDEV_UP] = EVENTHANDLER_REGISTER(
1506 	    ifnet_link_event, linux_handle_ifnet_link_event, nb, 0);
1507 	nb->tags[NETDEV_REGISTER] = EVENTHANDLER_REGISTER(
1508 	    ifnet_arrival_event, linux_handle_ifnet_arrival_event, nb, 0);
1509 	nb->tags[NETDEV_UNREGISTER] = EVENTHANDLER_REGISTER(
1510 	    ifnet_departure_event, linux_handle_ifnet_departure_event, nb, 0);
1511 	nb->tags[NETDEV_CHANGEADDR] = EVENTHANDLER_REGISTER(
1512 	    iflladdr_event, linux_handle_iflladdr_event, nb, 0);
1513 
1514 	return (0);
1515 }
1516 
1517 int
1518 register_inetaddr_notifier(struct notifier_block *nb)
1519 {
1520 
1521         nb->tags[NETDEV_CHANGEIFADDR] = EVENTHANDLER_REGISTER(
1522             ifaddr_event, linux_handle_ifaddr_event, nb, 0);
1523         return (0);
1524 }
1525 
1526 int
1527 unregister_netdevice_notifier(struct notifier_block *nb)
1528 {
1529 
1530         EVENTHANDLER_DEREGISTER(ifnet_link_event,
1531 	    nb->tags[NETDEV_UP]);
1532         EVENTHANDLER_DEREGISTER(ifnet_arrival_event,
1533 	    nb->tags[NETDEV_REGISTER]);
1534         EVENTHANDLER_DEREGISTER(ifnet_departure_event,
1535 	    nb->tags[NETDEV_UNREGISTER]);
1536         EVENTHANDLER_DEREGISTER(iflladdr_event,
1537 	    nb->tags[NETDEV_CHANGEADDR]);
1538 
1539 	return (0);
1540 }
1541 
1542 int
1543 unregister_inetaddr_notifier(struct notifier_block *nb)
1544 {
1545 
1546         EVENTHANDLER_DEREGISTER(ifaddr_event,
1547             nb->tags[NETDEV_CHANGEIFADDR]);
1548 
1549         return (0);
1550 }
1551 
1552 struct list_sort_thunk {
1553 	int (*cmp)(void *, struct list_head *, struct list_head *);
1554 	void *priv;
1555 };
1556 
1557 static inline int
1558 linux_le_cmp(void *priv, const void *d1, const void *d2)
1559 {
1560 	struct list_head *le1, *le2;
1561 	struct list_sort_thunk *thunk;
1562 
1563 	thunk = priv;
1564 	le1 = *(__DECONST(struct list_head **, d1));
1565 	le2 = *(__DECONST(struct list_head **, d2));
1566 	return ((thunk->cmp)(thunk->priv, le1, le2));
1567 }
1568 
1569 void
1570 list_sort(void *priv, struct list_head *head, int (*cmp)(void *priv,
1571     struct list_head *a, struct list_head *b))
1572 {
1573 	struct list_sort_thunk thunk;
1574 	struct list_head **ar, *le;
1575 	size_t count, i;
1576 
1577 	count = 0;
1578 	list_for_each(le, head)
1579 		count++;
1580 	ar = malloc(sizeof(struct list_head *) * count, M_KMALLOC, M_WAITOK);
1581 	i = 0;
1582 	list_for_each(le, head)
1583 		ar[i++] = le;
1584 	thunk.cmp = cmp;
1585 	thunk.priv = priv;
1586 	qsort_r(ar, count, sizeof(struct list_head *), &thunk, linux_le_cmp);
1587 	INIT_LIST_HEAD(head);
1588 	for (i = 0; i < count; i++)
1589 		list_add_tail(ar[i], head);
1590 	free(ar, M_KMALLOC);
1591 }
1592 
1593 void
1594 linux_irq_handler(void *ent)
1595 {
1596 	struct irq_ent *irqe;
1597 
1598 	linux_set_current(curthread);
1599 
1600 	irqe = ent;
1601 	irqe->handler(irqe->irq, irqe->arg);
1602 }
1603 
1604 struct linux_cdev *
1605 linux_find_cdev(const char *name, unsigned major, unsigned minor)
1606 {
1607 	int unit = MKDEV(major, minor);
1608 	struct cdev *cdev;
1609 
1610 	dev_lock();
1611 	LIST_FOREACH(cdev, &linuxcdevsw.d_devs, si_list) {
1612 		struct linux_cdev *ldev = cdev->si_drv1;
1613 		if (dev2unit(cdev) == unit &&
1614 		    strcmp(kobject_name(&ldev->kobj), name) == 0) {
1615 			break;
1616 		}
1617 	}
1618 	dev_unlock();
1619 
1620 	return (cdev != NULL ? cdev->si_drv1 : NULL);
1621 }
1622 
1623 int
1624 __register_chrdev(unsigned int major, unsigned int baseminor,
1625     unsigned int count, const char *name,
1626     const struct file_operations *fops)
1627 {
1628 	struct linux_cdev *cdev;
1629 	int ret = 0;
1630 	int i;
1631 
1632 	for (i = baseminor; i < baseminor + count; i++) {
1633 		cdev = cdev_alloc();
1634 		cdev_init(cdev, fops);
1635 		kobject_set_name(&cdev->kobj, name);
1636 
1637 		ret = cdev_add(cdev, makedev(major, i), 1);
1638 		if (ret != 0)
1639 			break;
1640 	}
1641 	return (ret);
1642 }
1643 
1644 int
1645 __register_chrdev_p(unsigned int major, unsigned int baseminor,
1646     unsigned int count, const char *name,
1647     const struct file_operations *fops, uid_t uid,
1648     gid_t gid, int mode)
1649 {
1650 	struct linux_cdev *cdev;
1651 	int ret = 0;
1652 	int i;
1653 
1654 	for (i = baseminor; i < baseminor + count; i++) {
1655 		cdev = cdev_alloc();
1656 		cdev_init(cdev, fops);
1657 		kobject_set_name(&cdev->kobj, name);
1658 
1659 		ret = cdev_add_ext(cdev, makedev(major, i), uid, gid, mode);
1660 		if (ret != 0)
1661 			break;
1662 	}
1663 	return (ret);
1664 }
1665 
1666 void
1667 __unregister_chrdev(unsigned int major, unsigned int baseminor,
1668     unsigned int count, const char *name)
1669 {
1670 	struct linux_cdev *cdevp;
1671 	int i;
1672 
1673 	for (i = baseminor; i < baseminor + count; i++) {
1674 		cdevp = linux_find_cdev(name, major, i);
1675 		if (cdevp != NULL)
1676 			cdev_del(cdevp);
1677 	}
1678 }
1679 
1680 #if defined(__i386__) || defined(__amd64__)
1681 bool linux_cpu_has_clflush;
1682 #endif
1683 
1684 static void
1685 linux_compat_init(void *arg)
1686 {
1687 	struct sysctl_oid *rootoid;
1688 	int i;
1689 
1690 #if defined(__i386__) || defined(__amd64__)
1691 	linux_cpu_has_clflush = (cpu_feature & CPUID_CLFSH);
1692 #endif
1693 	rw_init(&linux_vma_lock, "lkpi-vma-lock");
1694 
1695 	rootoid = SYSCTL_ADD_ROOT_NODE(NULL,
1696 	    OID_AUTO, "sys", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "sys");
1697 	kobject_init(&linux_class_root, &linux_class_ktype);
1698 	kobject_set_name(&linux_class_root, "class");
1699 	linux_class_root.oidp = SYSCTL_ADD_NODE(NULL, SYSCTL_CHILDREN(rootoid),
1700 	    OID_AUTO, "class", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "class");
1701 	kobject_init(&linux_root_device.kobj, &linux_dev_ktype);
1702 	kobject_set_name(&linux_root_device.kobj, "device");
1703 	linux_root_device.kobj.oidp = SYSCTL_ADD_NODE(NULL,
1704 	    SYSCTL_CHILDREN(rootoid), OID_AUTO, "device", CTLFLAG_RD, NULL,
1705 	    "device");
1706 	linux_root_device.bsddev = root_bus;
1707 	linux_class_misc.name = "misc";
1708 	class_register(&linux_class_misc);
1709 	INIT_LIST_HEAD(&pci_drivers);
1710 	INIT_LIST_HEAD(&pci_devices);
1711 	spin_lock_init(&pci_lock);
1712 	mtx_init(&vmmaplock, "IO Map lock", NULL, MTX_DEF);
1713 	for (i = 0; i < VMMAP_HASH_SIZE; i++)
1714 		LIST_INIT(&vmmaphead[i]);
1715 }
1716 SYSINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_init, NULL);
1717 
1718 static void
1719 linux_compat_uninit(void *arg)
1720 {
1721 	linux_kobject_kfree_name(&linux_class_root);
1722 	linux_kobject_kfree_name(&linux_root_device.kobj);
1723 	linux_kobject_kfree_name(&linux_class_misc.kobj);
1724 
1725 	rw_destroy(&linux_vma_lock);
1726 }
1727 SYSUNINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_uninit, NULL);
1728 
1729 /*
1730  * NOTE: Linux frequently uses "unsigned long" for pointer to integer
1731  * conversion and vice versa, where in FreeBSD "uintptr_t" would be
1732  * used. Assert these types have the same size, else some parts of the
1733  * LinuxKPI may not work like expected:
1734  */
1735 CTASSERT(sizeof(unsigned long) == sizeof(uintptr_t));
1736