xref: /freebsd/sys/compat/linuxkpi/common/src/linux_compat.c (revision fe2494903422ba3b924eba82cb63a6a9188fad7a)
1 /*-
2  * Copyright (c) 2010 Isilon Systems, Inc.
3  * Copyright (c) 2010 iX Systems, Inc.
4  * Copyright (c) 2010 Panasas, Inc.
5  * Copyright (c) 2013-2018 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 "opt_stack.h"
34 
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/malloc.h>
38 #include <sys/kernel.h>
39 #include <sys/sysctl.h>
40 #include <sys/proc.h>
41 #include <sys/sglist.h>
42 #include <sys/sleepqueue.h>
43 #include <sys/lock.h>
44 #include <sys/mutex.h>
45 #include <sys/bus.h>
46 #include <sys/fcntl.h>
47 #include <sys/file.h>
48 #include <sys/filio.h>
49 #include <sys/rwlock.h>
50 #include <sys/mman.h>
51 #include <sys/stack.h>
52 
53 #include <vm/vm.h>
54 #include <vm/pmap.h>
55 #include <vm/vm_object.h>
56 #include <vm/vm_page.h>
57 #include <vm/vm_pager.h>
58 
59 #include <machine/stdarg.h>
60 
61 #if defined(__i386__) || defined(__amd64__)
62 #include <machine/md_var.h>
63 #endif
64 
65 #include <linux/kobject.h>
66 #include <linux/device.h>
67 #include <linux/slab.h>
68 #include <linux/module.h>
69 #include <linux/moduleparam.h>
70 #include <linux/cdev.h>
71 #include <linux/file.h>
72 #include <linux/sysfs.h>
73 #include <linux/mm.h>
74 #include <linux/io.h>
75 #include <linux/vmalloc.h>
76 #include <linux/netdevice.h>
77 #include <linux/timer.h>
78 #include <linux/interrupt.h>
79 #include <linux/uaccess.h>
80 #include <linux/list.h>
81 #include <linux/kthread.h>
82 #include <linux/kernel.h>
83 #include <linux/compat.h>
84 #include <linux/poll.h>
85 #include <linux/smp.h>
86 
87 #if defined(__i386__) || defined(__amd64__)
88 #include <asm/smp.h>
89 #endif
90 
91 SYSCTL_NODE(_compat, OID_AUTO, linuxkpi, CTLFLAG_RW, 0, "LinuxKPI parameters");
92 
93 MALLOC_DEFINE(M_KMALLOC, "linux", "Linux kmalloc compat");
94 
95 #include <linux/rbtree.h>
96 /* Undo Linux compat changes. */
97 #undef RB_ROOT
98 #undef file
99 #undef cdev
100 #define	RB_ROOT(head)	(head)->rbh_root
101 
102 static struct vm_area_struct *linux_cdev_handle_find(void *handle);
103 
104 struct kobject linux_class_root;
105 struct device linux_root_device;
106 struct class linux_class_misc;
107 struct list_head pci_drivers;
108 struct list_head pci_devices;
109 spinlock_t pci_lock;
110 
111 unsigned long linux_timer_hz_mask;
112 
113 int
114 panic_cmp(struct rb_node *one, struct rb_node *two)
115 {
116 	panic("no cmp");
117 }
118 
119 RB_GENERATE(linux_root, rb_node, __entry, panic_cmp);
120 
121 int
122 kobject_set_name_vargs(struct kobject *kobj, const char *fmt, va_list args)
123 {
124 	va_list tmp_va;
125 	int len;
126 	char *old;
127 	char *name;
128 	char dummy;
129 
130 	old = kobj->name;
131 
132 	if (old && fmt == NULL)
133 		return (0);
134 
135 	/* compute length of string */
136 	va_copy(tmp_va, args);
137 	len = vsnprintf(&dummy, 0, fmt, tmp_va);
138 	va_end(tmp_va);
139 
140 	/* account for zero termination */
141 	len++;
142 
143 	/* check for error */
144 	if (len < 1)
145 		return (-EINVAL);
146 
147 	/* allocate memory for string */
148 	name = kzalloc(len, GFP_KERNEL);
149 	if (name == NULL)
150 		return (-ENOMEM);
151 	vsnprintf(name, len, fmt, args);
152 	kobj->name = name;
153 
154 	/* free old string */
155 	kfree(old);
156 
157 	/* filter new string */
158 	for (; *name != '\0'; name++)
159 		if (*name == '/')
160 			*name = '!';
161 	return (0);
162 }
163 
164 int
165 kobject_set_name(struct kobject *kobj, const char *fmt, ...)
166 {
167 	va_list args;
168 	int error;
169 
170 	va_start(args, fmt);
171 	error = kobject_set_name_vargs(kobj, fmt, args);
172 	va_end(args);
173 
174 	return (error);
175 }
176 
177 static int
178 kobject_add_complete(struct kobject *kobj, struct kobject *parent)
179 {
180 	const struct kobj_type *t;
181 	int error;
182 
183 	kobj->parent = parent;
184 	error = sysfs_create_dir(kobj);
185 	if (error == 0 && kobj->ktype && kobj->ktype->default_attrs) {
186 		struct attribute **attr;
187 		t = kobj->ktype;
188 
189 		for (attr = t->default_attrs; *attr != NULL; attr++) {
190 			error = sysfs_create_file(kobj, *attr);
191 			if (error)
192 				break;
193 		}
194 		if (error)
195 			sysfs_remove_dir(kobj);
196 
197 	}
198 	return (error);
199 }
200 
201 int
202 kobject_add(struct kobject *kobj, struct kobject *parent, const char *fmt, ...)
203 {
204 	va_list args;
205 	int error;
206 
207 	va_start(args, fmt);
208 	error = kobject_set_name_vargs(kobj, fmt, args);
209 	va_end(args);
210 	if (error)
211 		return (error);
212 
213 	return kobject_add_complete(kobj, parent);
214 }
215 
216 void
217 linux_kobject_release(struct kref *kref)
218 {
219 	struct kobject *kobj;
220 	char *name;
221 
222 	kobj = container_of(kref, struct kobject, kref);
223 	sysfs_remove_dir(kobj);
224 	name = kobj->name;
225 	if (kobj->ktype && kobj->ktype->release)
226 		kobj->ktype->release(kobj);
227 	kfree(name);
228 }
229 
230 static void
231 linux_kobject_kfree(struct kobject *kobj)
232 {
233 	kfree(kobj);
234 }
235 
236 static void
237 linux_kobject_kfree_name(struct kobject *kobj)
238 {
239 	if (kobj) {
240 		kfree(kobj->name);
241 	}
242 }
243 
244 const struct kobj_type linux_kfree_type = {
245 	.release = linux_kobject_kfree
246 };
247 
248 static void
249 linux_device_release(struct device *dev)
250 {
251 	pr_debug("linux_device_release: %s\n", dev_name(dev));
252 	kfree(dev);
253 }
254 
255 static ssize_t
256 linux_class_show(struct kobject *kobj, struct attribute *attr, char *buf)
257 {
258 	struct class_attribute *dattr;
259 	ssize_t error;
260 
261 	dattr = container_of(attr, struct class_attribute, attr);
262 	error = -EIO;
263 	if (dattr->show)
264 		error = dattr->show(container_of(kobj, struct class, kobj),
265 		    dattr, buf);
266 	return (error);
267 }
268 
269 static ssize_t
270 linux_class_store(struct kobject *kobj, struct attribute *attr, const char *buf,
271     size_t count)
272 {
273 	struct class_attribute *dattr;
274 	ssize_t error;
275 
276 	dattr = container_of(attr, struct class_attribute, attr);
277 	error = -EIO;
278 	if (dattr->store)
279 		error = dattr->store(container_of(kobj, struct class, kobj),
280 		    dattr, buf, count);
281 	return (error);
282 }
283 
284 static void
285 linux_class_release(struct kobject *kobj)
286 {
287 	struct class *class;
288 
289 	class = container_of(kobj, struct class, kobj);
290 	if (class->class_release)
291 		class->class_release(class);
292 }
293 
294 static const struct sysfs_ops linux_class_sysfs = {
295 	.show  = linux_class_show,
296 	.store = linux_class_store,
297 };
298 
299 const struct kobj_type linux_class_ktype = {
300 	.release = linux_class_release,
301 	.sysfs_ops = &linux_class_sysfs
302 };
303 
304 static void
305 linux_dev_release(struct kobject *kobj)
306 {
307 	struct device *dev;
308 
309 	dev = container_of(kobj, struct device, kobj);
310 	/* This is the precedence defined by linux. */
311 	if (dev->release)
312 		dev->release(dev);
313 	else if (dev->class && dev->class->dev_release)
314 		dev->class->dev_release(dev);
315 }
316 
317 static ssize_t
318 linux_dev_show(struct kobject *kobj, struct attribute *attr, char *buf)
319 {
320 	struct device_attribute *dattr;
321 	ssize_t error;
322 
323 	dattr = container_of(attr, struct device_attribute, attr);
324 	error = -EIO;
325 	if (dattr->show)
326 		error = dattr->show(container_of(kobj, struct device, kobj),
327 		    dattr, buf);
328 	return (error);
329 }
330 
331 static ssize_t
332 linux_dev_store(struct kobject *kobj, struct attribute *attr, const char *buf,
333     size_t count)
334 {
335 	struct device_attribute *dattr;
336 	ssize_t error;
337 
338 	dattr = container_of(attr, struct device_attribute, attr);
339 	error = -EIO;
340 	if (dattr->store)
341 		error = dattr->store(container_of(kobj, struct device, kobj),
342 		    dattr, buf, count);
343 	return (error);
344 }
345 
346 static const struct sysfs_ops linux_dev_sysfs = {
347 	.show  = linux_dev_show,
348 	.store = linux_dev_store,
349 };
350 
351 const struct kobj_type linux_dev_ktype = {
352 	.release = linux_dev_release,
353 	.sysfs_ops = &linux_dev_sysfs
354 };
355 
356 struct device *
357 device_create(struct class *class, struct device *parent, dev_t devt,
358     void *drvdata, const char *fmt, ...)
359 {
360 	struct device *dev;
361 	va_list args;
362 
363 	dev = kzalloc(sizeof(*dev), M_WAITOK);
364 	dev->parent = parent;
365 	dev->class = class;
366 	dev->devt = devt;
367 	dev->driver_data = drvdata;
368 	dev->release = linux_device_release;
369 	va_start(args, fmt);
370 	kobject_set_name_vargs(&dev->kobj, fmt, args);
371 	va_end(args);
372 	device_register(dev);
373 
374 	return (dev);
375 }
376 
377 int
378 kobject_init_and_add(struct kobject *kobj, const struct kobj_type *ktype,
379     struct kobject *parent, const char *fmt, ...)
380 {
381 	va_list args;
382 	int error;
383 
384 	kobject_init(kobj, ktype);
385 	kobj->ktype = ktype;
386 	kobj->parent = parent;
387 	kobj->name = NULL;
388 
389 	va_start(args, fmt);
390 	error = kobject_set_name_vargs(kobj, fmt, args);
391 	va_end(args);
392 	if (error)
393 		return (error);
394 	return kobject_add_complete(kobj, parent);
395 }
396 
397 static void
398 linux_kq_lock(void *arg)
399 {
400 	spinlock_t *s = arg;
401 
402 	spin_lock(s);
403 }
404 static void
405 linux_kq_unlock(void *arg)
406 {
407 	spinlock_t *s = arg;
408 
409 	spin_unlock(s);
410 }
411 
412 static void
413 linux_kq_lock_owned(void *arg)
414 {
415 #ifdef INVARIANTS
416 	spinlock_t *s = arg;
417 
418 	mtx_assert(&s->m, MA_OWNED);
419 #endif
420 }
421 
422 static void
423 linux_kq_lock_unowned(void *arg)
424 {
425 #ifdef INVARIANTS
426 	spinlock_t *s = arg;
427 
428 	mtx_assert(&s->m, MA_NOTOWNED);
429 #endif
430 }
431 
432 static void
433 linux_file_kqfilter_poll(struct linux_file *, int);
434 
435 struct linux_file *
436 linux_file_alloc(void)
437 {
438 	struct linux_file *filp;
439 
440 	filp = kzalloc(sizeof(*filp), GFP_KERNEL);
441 
442 	/* set initial refcount */
443 	filp->f_count = 1;
444 
445 	/* setup fields needed by kqueue support */
446 	spin_lock_init(&filp->f_kqlock);
447 	knlist_init(&filp->f_selinfo.si_note, &filp->f_kqlock,
448 	    linux_kq_lock, linux_kq_unlock,
449 	    linux_kq_lock_owned, linux_kq_lock_unowned);
450 
451 	return (filp);
452 }
453 
454 void
455 linux_file_free(struct linux_file *filp)
456 {
457 	if (filp->_file == NULL) {
458 		if (filp->f_shmem != NULL)
459 			vm_object_deallocate(filp->f_shmem);
460 		kfree(filp);
461 	} else {
462 		/*
463 		 * The close method of the character device or file
464 		 * will free the linux_file structure:
465 		 */
466 		_fdrop(filp->_file, curthread);
467 	}
468 }
469 
470 static int
471 linux_cdev_pager_fault(vm_object_t vm_obj, vm_ooffset_t offset, int prot,
472     vm_page_t *mres)
473 {
474 	struct vm_area_struct *vmap;
475 
476 	vmap = linux_cdev_handle_find(vm_obj->handle);
477 
478 	MPASS(vmap != NULL);
479 	MPASS(vmap->vm_private_data == vm_obj->handle);
480 
481 	if (likely(vmap->vm_ops != NULL && offset < vmap->vm_len)) {
482 		vm_paddr_t paddr = IDX_TO_OFF(vmap->vm_pfn) + offset;
483 		vm_page_t page;
484 
485 		if (((*mres)->flags & PG_FICTITIOUS) != 0) {
486 			/*
487 			 * If the passed in result page is a fake
488 			 * page, update it with the new physical
489 			 * address.
490 			 */
491 			page = *mres;
492 			vm_page_updatefake(page, paddr, vm_obj->memattr);
493 		} else {
494 			/*
495 			 * Replace the passed in "mres" page with our
496 			 * own fake page and free up the all of the
497 			 * original pages.
498 			 */
499 			VM_OBJECT_WUNLOCK(vm_obj);
500 			page = vm_page_getfake(paddr, vm_obj->memattr);
501 			VM_OBJECT_WLOCK(vm_obj);
502 
503 			vm_page_replace_checked(page, vm_obj,
504 			    (*mres)->pindex, *mres);
505 
506 			vm_page_lock(*mres);
507 			vm_page_free(*mres);
508 			vm_page_unlock(*mres);
509 			*mres = page;
510 		}
511 		page->valid = VM_PAGE_BITS_ALL;
512 		return (VM_PAGER_OK);
513 	}
514 	return (VM_PAGER_FAIL);
515 }
516 
517 static int
518 linux_cdev_pager_populate(vm_object_t vm_obj, vm_pindex_t pidx, int fault_type,
519     vm_prot_t max_prot, vm_pindex_t *first, vm_pindex_t *last)
520 {
521 	struct vm_area_struct *vmap;
522 	int err;
523 
524 	linux_set_current(curthread);
525 
526 	/* get VM area structure */
527 	vmap = linux_cdev_handle_find(vm_obj->handle);
528 	MPASS(vmap != NULL);
529 	MPASS(vmap->vm_private_data == vm_obj->handle);
530 
531 	VM_OBJECT_WUNLOCK(vm_obj);
532 
533 	down_write(&vmap->vm_mm->mmap_sem);
534 	if (unlikely(vmap->vm_ops == NULL)) {
535 		err = VM_FAULT_SIGBUS;
536 	} else {
537 		struct vm_fault vmf;
538 
539 		/* fill out VM fault structure */
540 		vmf.virtual_address = (void *)((uintptr_t)pidx << PAGE_SHIFT);
541 		vmf.flags = (fault_type & VM_PROT_WRITE) ? FAULT_FLAG_WRITE : 0;
542 		vmf.pgoff = 0;
543 		vmf.page = NULL;
544 		vmf.vma = vmap;
545 
546 		vmap->vm_pfn_count = 0;
547 		vmap->vm_pfn_pcount = &vmap->vm_pfn_count;
548 		vmap->vm_obj = vm_obj;
549 
550 		err = vmap->vm_ops->fault(vmap, &vmf);
551 
552 		while (vmap->vm_pfn_count == 0 && err == VM_FAULT_NOPAGE) {
553 			kern_yield(PRI_USER);
554 			err = vmap->vm_ops->fault(vmap, &vmf);
555 		}
556 	}
557 
558 	/* translate return code */
559 	switch (err) {
560 	case VM_FAULT_OOM:
561 		err = VM_PAGER_AGAIN;
562 		break;
563 	case VM_FAULT_SIGBUS:
564 		err = VM_PAGER_BAD;
565 		break;
566 	case VM_FAULT_NOPAGE:
567 		/*
568 		 * By contract the fault handler will return having
569 		 * busied all the pages itself. If pidx is already
570 		 * found in the object, it will simply xbusy the first
571 		 * page and return with vm_pfn_count set to 1.
572 		 */
573 		*first = vmap->vm_pfn_first;
574 		*last = *first + vmap->vm_pfn_count - 1;
575 		err = VM_PAGER_OK;
576 		break;
577 	default:
578 		err = VM_PAGER_ERROR;
579 		break;
580 	}
581 	up_write(&vmap->vm_mm->mmap_sem);
582 	VM_OBJECT_WLOCK(vm_obj);
583 	return (err);
584 }
585 
586 static struct rwlock linux_vma_lock;
587 static TAILQ_HEAD(, vm_area_struct) linux_vma_head =
588     TAILQ_HEAD_INITIALIZER(linux_vma_head);
589 
590 static void
591 linux_cdev_handle_free(struct vm_area_struct *vmap)
592 {
593 	/* Drop reference on vm_file */
594 	if (vmap->vm_file != NULL)
595 		fput(vmap->vm_file);
596 
597 	/* Drop reference on mm_struct */
598 	mmput(vmap->vm_mm);
599 
600 	kfree(vmap);
601 }
602 
603 static void
604 linux_cdev_handle_remove(struct vm_area_struct *vmap)
605 {
606 	rw_wlock(&linux_vma_lock);
607 	TAILQ_REMOVE(&linux_vma_head, vmap, vm_entry);
608 	rw_wunlock(&linux_vma_lock);
609 }
610 
611 static struct vm_area_struct *
612 linux_cdev_handle_find(void *handle)
613 {
614 	struct vm_area_struct *vmap;
615 
616 	rw_rlock(&linux_vma_lock);
617 	TAILQ_FOREACH(vmap, &linux_vma_head, vm_entry) {
618 		if (vmap->vm_private_data == handle)
619 			break;
620 	}
621 	rw_runlock(&linux_vma_lock);
622 	return (vmap);
623 }
624 
625 static int
626 linux_cdev_pager_ctor(void *handle, vm_ooffset_t size, vm_prot_t prot,
627 		      vm_ooffset_t foff, struct ucred *cred, u_short *color)
628 {
629 
630 	MPASS(linux_cdev_handle_find(handle) != NULL);
631 	*color = 0;
632 	return (0);
633 }
634 
635 static void
636 linux_cdev_pager_dtor(void *handle)
637 {
638 	const struct vm_operations_struct *vm_ops;
639 	struct vm_area_struct *vmap;
640 
641 	vmap = linux_cdev_handle_find(handle);
642 	MPASS(vmap != NULL);
643 
644 	/*
645 	 * Remove handle before calling close operation to prevent
646 	 * other threads from reusing the handle pointer.
647 	 */
648 	linux_cdev_handle_remove(vmap);
649 
650 	down_write(&vmap->vm_mm->mmap_sem);
651 	vm_ops = vmap->vm_ops;
652 	if (likely(vm_ops != NULL))
653 		vm_ops->close(vmap);
654 	up_write(&vmap->vm_mm->mmap_sem);
655 
656 	linux_cdev_handle_free(vmap);
657 }
658 
659 static struct cdev_pager_ops linux_cdev_pager_ops[2] = {
660   {
661 	/* OBJT_MGTDEVICE */
662 	.cdev_pg_populate	= linux_cdev_pager_populate,
663 	.cdev_pg_ctor	= linux_cdev_pager_ctor,
664 	.cdev_pg_dtor	= linux_cdev_pager_dtor
665   },
666   {
667 	/* OBJT_DEVICE */
668 	.cdev_pg_fault	= linux_cdev_pager_fault,
669 	.cdev_pg_ctor	= linux_cdev_pager_ctor,
670 	.cdev_pg_dtor	= linux_cdev_pager_dtor
671   },
672 };
673 
674 #define	OPW(fp,td,code) ({			\
675 	struct file *__fpop;			\
676 	__typeof(code) __retval;		\
677 						\
678 	__fpop = (td)->td_fpop;			\
679 	(td)->td_fpop = (fp);			\
680 	__retval = (code);			\
681 	(td)->td_fpop = __fpop;			\
682 	__retval;				\
683 })
684 
685 static int
686 linux_dev_fdopen(struct cdev *dev, int fflags, struct thread *td, struct file *file)
687 {
688 	struct linux_cdev *ldev;
689 	struct linux_file *filp;
690 	int error;
691 
692 	ldev = dev->si_drv1;
693 
694 	filp = linux_file_alloc();
695 	filp->f_dentry = &filp->f_dentry_store;
696 	filp->f_op = ldev->ops;
697 	filp->f_mode = file->f_flag;
698 	filp->f_flags = file->f_flag;
699 	filp->f_vnode = file->f_vnode;
700 	filp->_file = file;
701 
702 	linux_set_current(td);
703 
704 	if (filp->f_op->open) {
705 		error = -filp->f_op->open(file->f_vnode, filp);
706 		if (error) {
707 			kfree(filp);
708 			return (error);
709 		}
710 	}
711 
712 	/* hold on to the vnode - used for fstat() */
713 	vhold(filp->f_vnode);
714 
715 	/* release the file from devfs */
716 	finit(file, filp->f_mode, DTYPE_DEV, filp, &linuxfileops);
717 	return (ENXIO);
718 }
719 
720 #define	LINUX_IOCTL_MIN_PTR 0x10000UL
721 #define	LINUX_IOCTL_MAX_PTR (LINUX_IOCTL_MIN_PTR + IOCPARM_MAX)
722 
723 static inline int
724 linux_remap_address(void **uaddr, size_t len)
725 {
726 	uintptr_t uaddr_val = (uintptr_t)(*uaddr);
727 
728 	if (unlikely(uaddr_val >= LINUX_IOCTL_MIN_PTR &&
729 	    uaddr_val < LINUX_IOCTL_MAX_PTR)) {
730 		struct task_struct *pts = current;
731 		if (pts == NULL) {
732 			*uaddr = NULL;
733 			return (1);
734 		}
735 
736 		/* compute data offset */
737 		uaddr_val -= LINUX_IOCTL_MIN_PTR;
738 
739 		/* check that length is within bounds */
740 		if ((len > IOCPARM_MAX) ||
741 		    (uaddr_val + len) > pts->bsd_ioctl_len) {
742 			*uaddr = NULL;
743 			return (1);
744 		}
745 
746 		/* re-add kernel buffer address */
747 		uaddr_val += (uintptr_t)pts->bsd_ioctl_data;
748 
749 		/* update address location */
750 		*uaddr = (void *)uaddr_val;
751 		return (1);
752 	}
753 	return (0);
754 }
755 
756 int
757 linux_copyin(const void *uaddr, void *kaddr, size_t len)
758 {
759 	if (linux_remap_address(__DECONST(void **, &uaddr), len)) {
760 		if (uaddr == NULL)
761 			return (-EFAULT);
762 		memcpy(kaddr, uaddr, len);
763 		return (0);
764 	}
765 	return (-copyin(uaddr, kaddr, len));
766 }
767 
768 int
769 linux_copyout(const void *kaddr, void *uaddr, size_t len)
770 {
771 	if (linux_remap_address(&uaddr, len)) {
772 		if (uaddr == NULL)
773 			return (-EFAULT);
774 		memcpy(uaddr, kaddr, len);
775 		return (0);
776 	}
777 	return (-copyout(kaddr, uaddr, len));
778 }
779 
780 size_t
781 linux_clear_user(void *_uaddr, size_t _len)
782 {
783 	uint8_t *uaddr = _uaddr;
784 	size_t len = _len;
785 
786 	/* make sure uaddr is aligned before going into the fast loop */
787 	while (((uintptr_t)uaddr & 7) != 0 && len > 7) {
788 		if (subyte(uaddr, 0))
789 			return (_len);
790 		uaddr++;
791 		len--;
792 	}
793 
794 	/* zero 8 bytes at a time */
795 	while (len > 7) {
796 #ifdef __LP64__
797 		if (suword64(uaddr, 0))
798 			return (_len);
799 #else
800 		if (suword32(uaddr, 0))
801 			return (_len);
802 		if (suword32(uaddr + 4, 0))
803 			return (_len);
804 #endif
805 		uaddr += 8;
806 		len -= 8;
807 	}
808 
809 	/* zero fill end, if any */
810 	while (len > 0) {
811 		if (subyte(uaddr, 0))
812 			return (_len);
813 		uaddr++;
814 		len--;
815 	}
816 	return (0);
817 }
818 
819 int
820 linux_access_ok(int rw, const void *uaddr, size_t len)
821 {
822 	uintptr_t saddr;
823 	uintptr_t eaddr;
824 
825 	/* get start and end address */
826 	saddr = (uintptr_t)uaddr;
827 	eaddr = (uintptr_t)uaddr + len;
828 
829 	/* verify addresses are valid for userspace */
830 	return ((saddr == eaddr) ||
831 	    (eaddr > saddr && eaddr <= VM_MAXUSER_ADDRESS));
832 }
833 
834 /*
835  * This function should return either EINTR or ERESTART depending on
836  * the signal type sent to this thread:
837  */
838 static int
839 linux_get_error(struct task_struct *task, int error)
840 {
841 	/* check for signal type interrupt code */
842 	if (error == EINTR || error == ERESTARTSYS || error == ERESTART) {
843 		error = -linux_schedule_get_interrupt_value(task);
844 		if (error == 0)
845 			error = EINTR;
846 	}
847 	return (error);
848 }
849 
850 static int
851 linux_file_ioctl_sub(struct file *fp, struct linux_file *filp,
852     u_long cmd, caddr_t data, struct thread *td)
853 {
854 	struct task_struct *task = current;
855 	unsigned size;
856 	int error;
857 
858 	size = IOCPARM_LEN(cmd);
859 	/* refer to logic in sys_ioctl() */
860 	if (size > 0) {
861 		/*
862 		 * Setup hint for linux_copyin() and linux_copyout().
863 		 *
864 		 * Background: Linux code expects a user-space address
865 		 * while FreeBSD supplies a kernel-space address.
866 		 */
867 		task->bsd_ioctl_data = data;
868 		task->bsd_ioctl_len = size;
869 		data = (void *)LINUX_IOCTL_MIN_PTR;
870 	} else {
871 		/* fetch user-space pointer */
872 		data = *(void **)data;
873 	}
874 #if defined(__amd64__)
875 	if (td->td_proc->p_elf_machine == EM_386) {
876 		/* try the compat IOCTL handler first */
877 		if (filp->f_op->compat_ioctl != NULL)
878 			error = -OPW(fp, td, filp->f_op->compat_ioctl(filp, cmd, (u_long)data));
879 		else
880 			error = ENOTTY;
881 
882 		/* fallback to the regular IOCTL handler, if any */
883 		if (error == ENOTTY && filp->f_op->unlocked_ioctl != NULL)
884 			error = -OPW(fp, td, filp->f_op->unlocked_ioctl(filp, cmd, (u_long)data));
885 	} else
886 #endif
887 	if (filp->f_op->unlocked_ioctl != NULL)
888 		error = -OPW(fp, td, filp->f_op->unlocked_ioctl(filp, cmd, (u_long)data));
889 	else
890 		error = ENOTTY;
891 	if (size > 0) {
892 		task->bsd_ioctl_data = NULL;
893 		task->bsd_ioctl_len = 0;
894 	}
895 
896 	if (error == EWOULDBLOCK) {
897 		/* update kqfilter status, if any */
898 		linux_file_kqfilter_poll(filp,
899 		    LINUX_KQ_FLAG_HAS_READ | LINUX_KQ_FLAG_HAS_WRITE);
900 	} else {
901 		error = linux_get_error(task, error);
902 	}
903 	return (error);
904 }
905 
906 #define	LINUX_POLL_TABLE_NORMAL ((poll_table *)1)
907 
908 /*
909  * This function atomically updates the poll wakeup state and returns
910  * the previous state at the time of update.
911  */
912 static uint8_t
913 linux_poll_wakeup_state(atomic_t *v, const uint8_t *pstate)
914 {
915 	int c, old;
916 
917 	c = v->counter;
918 
919 	while ((old = atomic_cmpxchg(v, c, pstate[c])) != c)
920 		c = old;
921 
922 	return (c);
923 }
924 
925 
926 static int
927 linux_poll_wakeup_callback(wait_queue_t *wq, unsigned int wq_state, int flags, void *key)
928 {
929 	static const uint8_t state[LINUX_FWQ_STATE_MAX] = {
930 		[LINUX_FWQ_STATE_INIT] = LINUX_FWQ_STATE_INIT, /* NOP */
931 		[LINUX_FWQ_STATE_NOT_READY] = LINUX_FWQ_STATE_NOT_READY, /* NOP */
932 		[LINUX_FWQ_STATE_QUEUED] = LINUX_FWQ_STATE_READY,
933 		[LINUX_FWQ_STATE_READY] = LINUX_FWQ_STATE_READY, /* NOP */
934 	};
935 	struct linux_file *filp = container_of(wq, struct linux_file, f_wait_queue.wq);
936 
937 	switch (linux_poll_wakeup_state(&filp->f_wait_queue.state, state)) {
938 	case LINUX_FWQ_STATE_QUEUED:
939 		linux_poll_wakeup(filp);
940 		return (1);
941 	default:
942 		return (0);
943 	}
944 }
945 
946 void
947 linux_poll_wait(struct linux_file *filp, wait_queue_head_t *wqh, poll_table *p)
948 {
949 	static const uint8_t state[LINUX_FWQ_STATE_MAX] = {
950 		[LINUX_FWQ_STATE_INIT] = LINUX_FWQ_STATE_NOT_READY,
951 		[LINUX_FWQ_STATE_NOT_READY] = LINUX_FWQ_STATE_NOT_READY, /* NOP */
952 		[LINUX_FWQ_STATE_QUEUED] = LINUX_FWQ_STATE_QUEUED, /* NOP */
953 		[LINUX_FWQ_STATE_READY] = LINUX_FWQ_STATE_QUEUED,
954 	};
955 
956 	/* check if we are called inside the select system call */
957 	if (p == LINUX_POLL_TABLE_NORMAL)
958 		selrecord(curthread, &filp->f_selinfo);
959 
960 	switch (linux_poll_wakeup_state(&filp->f_wait_queue.state, state)) {
961 	case LINUX_FWQ_STATE_INIT:
962 		/* NOTE: file handles can only belong to one wait-queue */
963 		filp->f_wait_queue.wqh = wqh;
964 		filp->f_wait_queue.wq.func = &linux_poll_wakeup_callback;
965 		add_wait_queue(wqh, &filp->f_wait_queue.wq);
966 		atomic_set(&filp->f_wait_queue.state, LINUX_FWQ_STATE_QUEUED);
967 		break;
968 	default:
969 		break;
970 	}
971 }
972 
973 static void
974 linux_poll_wait_dequeue(struct linux_file *filp)
975 {
976 	static const uint8_t state[LINUX_FWQ_STATE_MAX] = {
977 		[LINUX_FWQ_STATE_INIT] = LINUX_FWQ_STATE_INIT,	/* NOP */
978 		[LINUX_FWQ_STATE_NOT_READY] = LINUX_FWQ_STATE_INIT,
979 		[LINUX_FWQ_STATE_QUEUED] = LINUX_FWQ_STATE_INIT,
980 		[LINUX_FWQ_STATE_READY] = LINUX_FWQ_STATE_INIT,
981 	};
982 
983 	seldrain(&filp->f_selinfo);
984 
985 	switch (linux_poll_wakeup_state(&filp->f_wait_queue.state, state)) {
986 	case LINUX_FWQ_STATE_NOT_READY:
987 	case LINUX_FWQ_STATE_QUEUED:
988 	case LINUX_FWQ_STATE_READY:
989 		remove_wait_queue(filp->f_wait_queue.wqh, &filp->f_wait_queue.wq);
990 		break;
991 	default:
992 		break;
993 	}
994 }
995 
996 void
997 linux_poll_wakeup(struct linux_file *filp)
998 {
999 	/* this function should be NULL-safe */
1000 	if (filp == NULL)
1001 		return;
1002 
1003 	selwakeup(&filp->f_selinfo);
1004 
1005 	spin_lock(&filp->f_kqlock);
1006 	filp->f_kqflags |= LINUX_KQ_FLAG_NEED_READ |
1007 	    LINUX_KQ_FLAG_NEED_WRITE;
1008 
1009 	/* make sure the "knote" gets woken up */
1010 	KNOTE_LOCKED(&filp->f_selinfo.si_note, 1);
1011 	spin_unlock(&filp->f_kqlock);
1012 }
1013 
1014 static void
1015 linux_file_kqfilter_detach(struct knote *kn)
1016 {
1017 	struct linux_file *filp = kn->kn_hook;
1018 
1019 	spin_lock(&filp->f_kqlock);
1020 	knlist_remove(&filp->f_selinfo.si_note, kn, 1);
1021 	spin_unlock(&filp->f_kqlock);
1022 }
1023 
1024 static int
1025 linux_file_kqfilter_read_event(struct knote *kn, long hint)
1026 {
1027 	struct linux_file *filp = kn->kn_hook;
1028 
1029 	mtx_assert(&filp->f_kqlock.m, MA_OWNED);
1030 
1031 	return ((filp->f_kqflags & LINUX_KQ_FLAG_NEED_READ) ? 1 : 0);
1032 }
1033 
1034 static int
1035 linux_file_kqfilter_write_event(struct knote *kn, long hint)
1036 {
1037 	struct linux_file *filp = kn->kn_hook;
1038 
1039 	mtx_assert(&filp->f_kqlock.m, MA_OWNED);
1040 
1041 	return ((filp->f_kqflags & LINUX_KQ_FLAG_NEED_WRITE) ? 1 : 0);
1042 }
1043 
1044 static struct filterops linux_dev_kqfiltops_read = {
1045 	.f_isfd = 1,
1046 	.f_detach = linux_file_kqfilter_detach,
1047 	.f_event = linux_file_kqfilter_read_event,
1048 };
1049 
1050 static struct filterops linux_dev_kqfiltops_write = {
1051 	.f_isfd = 1,
1052 	.f_detach = linux_file_kqfilter_detach,
1053 	.f_event = linux_file_kqfilter_write_event,
1054 };
1055 
1056 static void
1057 linux_file_kqfilter_poll(struct linux_file *filp, int kqflags)
1058 {
1059 	int temp;
1060 
1061 	if (filp->f_kqflags & kqflags) {
1062 		struct thread *td = curthread;
1063 
1064 		/* get the latest polling state */
1065 		temp = OPW(filp->_file, td, filp->f_op->poll(filp, NULL));
1066 
1067 		spin_lock(&filp->f_kqlock);
1068 		/* clear kqflags */
1069 		filp->f_kqflags &= ~(LINUX_KQ_FLAG_NEED_READ |
1070 		    LINUX_KQ_FLAG_NEED_WRITE);
1071 		/* update kqflags */
1072 		if (temp & (POLLIN | POLLOUT)) {
1073 			if (temp & POLLIN)
1074 				filp->f_kqflags |= LINUX_KQ_FLAG_NEED_READ;
1075 			if (temp & POLLOUT)
1076 				filp->f_kqflags |= LINUX_KQ_FLAG_NEED_WRITE;
1077 
1078 			/* make sure the "knote" gets woken up */
1079 			KNOTE_LOCKED(&filp->f_selinfo.si_note, 0);
1080 		}
1081 		spin_unlock(&filp->f_kqlock);
1082 	}
1083 }
1084 
1085 static int
1086 linux_file_kqfilter(struct file *file, struct knote *kn)
1087 {
1088 	struct linux_file *filp;
1089 	struct thread *td;
1090 	int error;
1091 
1092 	td = curthread;
1093 	filp = (struct linux_file *)file->f_data;
1094 	filp->f_flags = file->f_flag;
1095 	if (filp->f_op->poll == NULL)
1096 		return (EINVAL);
1097 
1098 	spin_lock(&filp->f_kqlock);
1099 	switch (kn->kn_filter) {
1100 	case EVFILT_READ:
1101 		filp->f_kqflags |= LINUX_KQ_FLAG_HAS_READ;
1102 		kn->kn_fop = &linux_dev_kqfiltops_read;
1103 		kn->kn_hook = filp;
1104 		knlist_add(&filp->f_selinfo.si_note, kn, 1);
1105 		error = 0;
1106 		break;
1107 	case EVFILT_WRITE:
1108 		filp->f_kqflags |= LINUX_KQ_FLAG_HAS_WRITE;
1109 		kn->kn_fop = &linux_dev_kqfiltops_write;
1110 		kn->kn_hook = filp;
1111 		knlist_add(&filp->f_selinfo.si_note, kn, 1);
1112 		error = 0;
1113 		break;
1114 	default:
1115 		error = EINVAL;
1116 		break;
1117 	}
1118 	spin_unlock(&filp->f_kqlock);
1119 
1120 	if (error == 0) {
1121 		linux_set_current(td);
1122 
1123 		/* update kqfilter status, if any */
1124 		linux_file_kqfilter_poll(filp,
1125 		    LINUX_KQ_FLAG_HAS_READ | LINUX_KQ_FLAG_HAS_WRITE);
1126 	}
1127 	return (error);
1128 }
1129 
1130 static int
1131 linux_file_mmap_single(struct file *fp, vm_ooffset_t *offset,
1132     vm_size_t size, struct vm_object **object, int nprot,
1133     struct thread *td)
1134 {
1135 	struct task_struct *task;
1136 	struct vm_area_struct *vmap;
1137 	struct mm_struct *mm;
1138 	struct linux_file *filp;
1139 	vm_memattr_t attr;
1140 	int error;
1141 
1142 	filp = (struct linux_file *)fp->f_data;
1143 	filp->f_flags = fp->f_flag;
1144 
1145 	if (filp->f_op->mmap == NULL)
1146 		return (EOPNOTSUPP);
1147 
1148 	linux_set_current(td);
1149 
1150 	/*
1151 	 * The same VM object might be shared by multiple processes
1152 	 * and the mm_struct is usually freed when a process exits.
1153 	 *
1154 	 * The atomic reference below makes sure the mm_struct is
1155 	 * available as long as the vmap is in the linux_vma_head.
1156 	 */
1157 	task = current;
1158 	mm = task->mm;
1159 	if (atomic_inc_not_zero(&mm->mm_users) == 0)
1160 		return (EINVAL);
1161 
1162 	vmap = kzalloc(sizeof(*vmap), GFP_KERNEL);
1163 	vmap->vm_start = 0;
1164 	vmap->vm_end = size;
1165 	vmap->vm_pgoff = *offset / PAGE_SIZE;
1166 	vmap->vm_pfn = 0;
1167 	vmap->vm_flags = vmap->vm_page_prot = (nprot & VM_PROT_ALL);
1168 	vmap->vm_ops = NULL;
1169 	vmap->vm_file = get_file(filp);
1170 	vmap->vm_mm = mm;
1171 
1172 	if (unlikely(down_write_killable(&vmap->vm_mm->mmap_sem))) {
1173 		error = linux_get_error(task, EINTR);
1174 	} else {
1175 		error = -OPW(fp, td, filp->f_op->mmap(filp, vmap));
1176 		error = linux_get_error(task, error);
1177 		up_write(&vmap->vm_mm->mmap_sem);
1178 	}
1179 
1180 	if (error != 0) {
1181 		linux_cdev_handle_free(vmap);
1182 		return (error);
1183 	}
1184 
1185 	attr = pgprot2cachemode(vmap->vm_page_prot);
1186 
1187 	if (vmap->vm_ops != NULL) {
1188 		struct vm_area_struct *ptr;
1189 		void *vm_private_data;
1190 		bool vm_no_fault;
1191 
1192 		if (vmap->vm_ops->open == NULL ||
1193 		    vmap->vm_ops->close == NULL ||
1194 		    vmap->vm_private_data == NULL) {
1195 			/* free allocated VM area struct */
1196 			linux_cdev_handle_free(vmap);
1197 			return (EINVAL);
1198 		}
1199 
1200 		vm_private_data = vmap->vm_private_data;
1201 
1202 		rw_wlock(&linux_vma_lock);
1203 		TAILQ_FOREACH(ptr, &linux_vma_head, vm_entry) {
1204 			if (ptr->vm_private_data == vm_private_data)
1205 				break;
1206 		}
1207 		/* check if there is an existing VM area struct */
1208 		if (ptr != NULL) {
1209 			/* check if the VM area structure is invalid */
1210 			if (ptr->vm_ops == NULL ||
1211 			    ptr->vm_ops->open == NULL ||
1212 			    ptr->vm_ops->close == NULL) {
1213 				error = ESTALE;
1214 				vm_no_fault = 1;
1215 			} else {
1216 				error = EEXIST;
1217 				vm_no_fault = (ptr->vm_ops->fault == NULL);
1218 			}
1219 		} else {
1220 			/* insert VM area structure into list */
1221 			TAILQ_INSERT_TAIL(&linux_vma_head, vmap, vm_entry);
1222 			error = 0;
1223 			vm_no_fault = (vmap->vm_ops->fault == NULL);
1224 		}
1225 		rw_wunlock(&linux_vma_lock);
1226 
1227 		if (error != 0) {
1228 			/* free allocated VM area struct */
1229 			linux_cdev_handle_free(vmap);
1230 			/* check for stale VM area struct */
1231 			if (error != EEXIST)
1232 				return (error);
1233 		}
1234 
1235 		/* check if there is no fault handler */
1236 		if (vm_no_fault) {
1237 			*object = cdev_pager_allocate(vm_private_data, OBJT_DEVICE,
1238 			    &linux_cdev_pager_ops[1], size, nprot, *offset,
1239 			    td->td_ucred);
1240 		} else {
1241 			*object = cdev_pager_allocate(vm_private_data, OBJT_MGTDEVICE,
1242 			    &linux_cdev_pager_ops[0], size, nprot, *offset,
1243 			    td->td_ucred);
1244 		}
1245 
1246 		/* check if allocating the VM object failed */
1247 		if (*object == NULL) {
1248 			if (error == 0) {
1249 				/* remove VM area struct from list */
1250 				linux_cdev_handle_remove(vmap);
1251 				/* free allocated VM area struct */
1252 				linux_cdev_handle_free(vmap);
1253 			}
1254 			return (EINVAL);
1255 		}
1256 	} else {
1257 		struct sglist *sg;
1258 
1259 		sg = sglist_alloc(1, M_WAITOK);
1260 		sglist_append_phys(sg,
1261 		    (vm_paddr_t)vmap->vm_pfn << PAGE_SHIFT, vmap->vm_len);
1262 
1263 		*object = vm_pager_allocate(OBJT_SG, sg, vmap->vm_len,
1264 		    nprot, 0, td->td_ucred);
1265 
1266 		linux_cdev_handle_free(vmap);
1267 
1268 		if (*object == NULL) {
1269 			sglist_free(sg);
1270 			return (EINVAL);
1271 		}
1272 	}
1273 
1274 	if (attr != VM_MEMATTR_DEFAULT) {
1275 		VM_OBJECT_WLOCK(*object);
1276 		vm_object_set_memattr(*object, attr);
1277 		VM_OBJECT_WUNLOCK(*object);
1278 	}
1279 	*offset = 0;
1280 	return (0);
1281 }
1282 
1283 struct cdevsw linuxcdevsw = {
1284 	.d_version = D_VERSION,
1285 	.d_fdopen = linux_dev_fdopen,
1286 	.d_name = "lkpidev",
1287 };
1288 
1289 static int
1290 linux_file_read(struct file *file, struct uio *uio, struct ucred *active_cred,
1291     int flags, struct thread *td)
1292 {
1293 	struct linux_file *filp;
1294 	ssize_t bytes;
1295 	int error;
1296 
1297 	error = 0;
1298 	filp = (struct linux_file *)file->f_data;
1299 	filp->f_flags = file->f_flag;
1300 	/* XXX no support for I/O vectors currently */
1301 	if (uio->uio_iovcnt != 1)
1302 		return (EOPNOTSUPP);
1303 	if (uio->uio_resid > DEVFS_IOSIZE_MAX)
1304 		return (EINVAL);
1305 	linux_set_current(td);
1306 	if (filp->f_op->read) {
1307 		bytes = OPW(file, td, filp->f_op->read(filp, uio->uio_iov->iov_base,
1308 		    uio->uio_iov->iov_len, &uio->uio_offset));
1309 		if (bytes >= 0) {
1310 			uio->uio_iov->iov_base =
1311 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
1312 			uio->uio_iov->iov_len -= bytes;
1313 			uio->uio_resid -= bytes;
1314 		} else {
1315 			error = linux_get_error(current, -bytes);
1316 		}
1317 	} else
1318 		error = ENXIO;
1319 
1320 	/* update kqfilter status, if any */
1321 	linux_file_kqfilter_poll(filp, LINUX_KQ_FLAG_HAS_READ);
1322 
1323 	return (error);
1324 }
1325 
1326 static int
1327 linux_file_write(struct file *file, struct uio *uio, struct ucred *active_cred,
1328     int flags, struct thread *td)
1329 {
1330 	struct linux_file *filp;
1331 	ssize_t bytes;
1332 	int error;
1333 
1334 	error = 0;
1335 	filp = (struct linux_file *)file->f_data;
1336 	filp->f_flags = file->f_flag;
1337 	/* XXX no support for I/O vectors currently */
1338 	if (uio->uio_iovcnt != 1)
1339 		return (EOPNOTSUPP);
1340 	if (uio->uio_resid > DEVFS_IOSIZE_MAX)
1341 		return (EINVAL);
1342 	linux_set_current(td);
1343 	if (filp->f_op->write) {
1344 		bytes = OPW(file, td, filp->f_op->write(filp, uio->uio_iov->iov_base,
1345 		    uio->uio_iov->iov_len, &uio->uio_offset));
1346 		if (bytes >= 0) {
1347 			uio->uio_iov->iov_base =
1348 			    ((uint8_t *)uio->uio_iov->iov_base) + bytes;
1349 			uio->uio_iov->iov_len -= bytes;
1350 			uio->uio_resid -= bytes;
1351 		} else {
1352 			error = linux_get_error(current, -bytes);
1353 		}
1354 	} else
1355 		error = ENXIO;
1356 
1357 	/* update kqfilter status, if any */
1358 	linux_file_kqfilter_poll(filp, LINUX_KQ_FLAG_HAS_WRITE);
1359 
1360 	return (error);
1361 }
1362 
1363 static int
1364 linux_file_poll(struct file *file, int events, struct ucred *active_cred,
1365     struct thread *td)
1366 {
1367 	struct linux_file *filp;
1368 	int revents;
1369 
1370 	filp = (struct linux_file *)file->f_data;
1371 	filp->f_flags = file->f_flag;
1372 	linux_set_current(td);
1373 	if (filp->f_op->poll != NULL)
1374 		revents = OPW(file, td, filp->f_op->poll(filp, LINUX_POLL_TABLE_NORMAL)) & events;
1375 	else
1376 		revents = 0;
1377 
1378 	return (revents);
1379 }
1380 
1381 static int
1382 linux_file_close(struct file *file, struct thread *td)
1383 {
1384 	struct linux_file *filp;
1385 	int error;
1386 
1387 	filp = (struct linux_file *)file->f_data;
1388 
1389 	KASSERT(file_count(filp) == 0, ("File refcount(%d) is not zero", file_count(filp)));
1390 
1391 	filp->f_flags = file->f_flag;
1392 	linux_set_current(td);
1393 	linux_poll_wait_dequeue(filp);
1394 	error = -OPW(file, td, filp->f_op->release(filp->f_vnode, filp));
1395 	funsetown(&filp->f_sigio);
1396 	if (filp->f_vnode != NULL)
1397 		vdrop(filp->f_vnode);
1398 	kfree(filp);
1399 
1400 	return (error);
1401 }
1402 
1403 static int
1404 linux_file_ioctl(struct file *fp, u_long cmd, void *data, struct ucred *cred,
1405     struct thread *td)
1406 {
1407 	struct linux_file *filp;
1408 	int error;
1409 
1410 	filp = (struct linux_file *)fp->f_data;
1411 	filp->f_flags = fp->f_flag;
1412 	error = 0;
1413 
1414 	linux_set_current(td);
1415 	switch (cmd) {
1416 	case FIONBIO:
1417 		break;
1418 	case FIOASYNC:
1419 		if (filp->f_op->fasync == NULL)
1420 			break;
1421 		error = -OPW(fp, td, filp->f_op->fasync(0, filp, fp->f_flag & FASYNC));
1422 		break;
1423 	case FIOSETOWN:
1424 		error = fsetown(*(int *)data, &filp->f_sigio);
1425 		if (error == 0) {
1426 			if (filp->f_op->fasync == NULL)
1427 				break;
1428 			error = -OPW(fp, td, filp->f_op->fasync(0, filp,
1429 			    fp->f_flag & FASYNC));
1430 		}
1431 		break;
1432 	case FIOGETOWN:
1433 		*(int *)data = fgetown(&filp->f_sigio);
1434 		break;
1435 	default:
1436 		error = linux_file_ioctl_sub(fp, filp, cmd, data, td);
1437 		break;
1438 	}
1439 	return (error);
1440 }
1441 
1442 static int
1443 linux_file_mmap_sub(struct thread *td, vm_size_t objsize, vm_prot_t prot,
1444     vm_prot_t *maxprotp, int *flagsp, struct file *fp,
1445     vm_ooffset_t *foff, vm_object_t *objp)
1446 {
1447 	/*
1448 	 * Character devices do not provide private mappings
1449 	 * of any kind:
1450 	 */
1451 	if ((*maxprotp & VM_PROT_WRITE) == 0 &&
1452 	    (prot & VM_PROT_WRITE) != 0)
1453 		return (EACCES);
1454 	if ((*flagsp & (MAP_PRIVATE | MAP_COPY)) != 0)
1455 		return (EINVAL);
1456 
1457 	return (linux_file_mmap_single(fp, foff, objsize, objp, (int)prot, td));
1458 }
1459 
1460 static int
1461 linux_file_mmap(struct file *fp, vm_map_t map, vm_offset_t *addr, vm_size_t size,
1462     vm_prot_t prot, vm_prot_t cap_maxprot, int flags, vm_ooffset_t foff,
1463     struct thread *td)
1464 {
1465 	struct linux_file *filp;
1466 	struct mount *mp;
1467 	struct vnode *vp;
1468 	vm_object_t object;
1469 	vm_prot_t maxprot;
1470 	int error;
1471 
1472 	filp = (struct linux_file *)fp->f_data;
1473 
1474 	vp = filp->f_vnode;
1475 	if (vp == NULL)
1476 		return (EOPNOTSUPP);
1477 
1478 	/*
1479 	 * Ensure that file and memory protections are
1480 	 * compatible.
1481 	 */
1482 	mp = vp->v_mount;
1483 	if (mp != NULL && (mp->mnt_flag & MNT_NOEXEC) != 0) {
1484 		maxprot = VM_PROT_NONE;
1485 		if ((prot & VM_PROT_EXECUTE) != 0)
1486 			return (EACCES);
1487 	} else
1488 		maxprot = VM_PROT_EXECUTE;
1489 	if ((fp->f_flag & FREAD) != 0)
1490 		maxprot |= VM_PROT_READ;
1491 	else if ((prot & VM_PROT_READ) != 0)
1492 		return (EACCES);
1493 
1494 	/*
1495 	 * If we are sharing potential changes via MAP_SHARED and we
1496 	 * are trying to get write permission although we opened it
1497 	 * without asking for it, bail out.
1498 	 *
1499 	 * Note that most character devices always share mappings.
1500 	 *
1501 	 * Rely on linux_file_mmap_sub() to fail invalid MAP_PRIVATE
1502 	 * requests rather than doing it here.
1503 	 */
1504 	if ((flags & MAP_SHARED) != 0) {
1505 		if ((fp->f_flag & FWRITE) != 0)
1506 			maxprot |= VM_PROT_WRITE;
1507 		else if ((prot & VM_PROT_WRITE) != 0)
1508 			return (EACCES);
1509 	}
1510 	maxprot &= cap_maxprot;
1511 
1512 	error = linux_file_mmap_sub(td, size, prot, &maxprot, &flags, fp, &foff,
1513 	    &object);
1514 	if (error != 0)
1515 		return (error);
1516 
1517 	error = vm_mmap_object(map, addr, size, prot, maxprot, flags, object,
1518 	    foff, FALSE, td);
1519 	if (error != 0)
1520 		vm_object_deallocate(object);
1521 	return (error);
1522 }
1523 
1524 static int
1525 linux_file_stat(struct file *fp, struct stat *sb, struct ucred *active_cred,
1526     struct thread *td)
1527 {
1528 	struct linux_file *filp;
1529 	struct vnode *vp;
1530 	int error;
1531 
1532 	filp = (struct linux_file *)fp->f_data;
1533 	if (filp->f_vnode == NULL)
1534 		return (EOPNOTSUPP);
1535 
1536 	vp = filp->f_vnode;
1537 
1538 	vn_lock(vp, LK_SHARED | LK_RETRY);
1539 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
1540 	VOP_UNLOCK(vp, 0);
1541 
1542 	return (error);
1543 }
1544 
1545 static int
1546 linux_file_fill_kinfo(struct file *fp, struct kinfo_file *kif,
1547     struct filedesc *fdp)
1548 {
1549 
1550 	return (0);
1551 }
1552 
1553 unsigned int
1554 linux_iminor(struct inode *inode)
1555 {
1556 	struct linux_cdev *ldev;
1557 
1558 	if (inode == NULL || inode->v_rdev == NULL ||
1559 	    inode->v_rdev->si_devsw != &linuxcdevsw)
1560 		return (-1U);
1561 	ldev = inode->v_rdev->si_drv1;
1562 	if (ldev == NULL)
1563 		return (-1U);
1564 
1565 	return (minor(ldev->dev));
1566 }
1567 
1568 struct fileops linuxfileops = {
1569 	.fo_read = linux_file_read,
1570 	.fo_write = linux_file_write,
1571 	.fo_truncate = invfo_truncate,
1572 	.fo_kqfilter = linux_file_kqfilter,
1573 	.fo_stat = linux_file_stat,
1574 	.fo_fill_kinfo = linux_file_fill_kinfo,
1575 	.fo_poll = linux_file_poll,
1576 	.fo_close = linux_file_close,
1577 	.fo_ioctl = linux_file_ioctl,
1578 	.fo_mmap = linux_file_mmap,
1579 	.fo_chmod = invfo_chmod,
1580 	.fo_chown = invfo_chown,
1581 	.fo_sendfile = invfo_sendfile,
1582 	.fo_flags = DFLAG_PASSABLE,
1583 };
1584 
1585 /*
1586  * Hash of vmmap addresses.  This is infrequently accessed and does not
1587  * need to be particularly large.  This is done because we must store the
1588  * caller's idea of the map size to properly unmap.
1589  */
1590 struct vmmap {
1591 	LIST_ENTRY(vmmap)	vm_next;
1592 	void 			*vm_addr;
1593 	unsigned long		vm_size;
1594 };
1595 
1596 struct vmmaphd {
1597 	struct vmmap *lh_first;
1598 };
1599 #define	VMMAP_HASH_SIZE	64
1600 #define	VMMAP_HASH_MASK	(VMMAP_HASH_SIZE - 1)
1601 #define	VM_HASH(addr)	((uintptr_t)(addr) >> PAGE_SHIFT) & VMMAP_HASH_MASK
1602 static struct vmmaphd vmmaphead[VMMAP_HASH_SIZE];
1603 static struct mtx vmmaplock;
1604 
1605 static void
1606 vmmap_add(void *addr, unsigned long size)
1607 {
1608 	struct vmmap *vmmap;
1609 
1610 	vmmap = kmalloc(sizeof(*vmmap), GFP_KERNEL);
1611 	mtx_lock(&vmmaplock);
1612 	vmmap->vm_size = size;
1613 	vmmap->vm_addr = addr;
1614 	LIST_INSERT_HEAD(&vmmaphead[VM_HASH(addr)], vmmap, vm_next);
1615 	mtx_unlock(&vmmaplock);
1616 }
1617 
1618 static struct vmmap *
1619 vmmap_remove(void *addr)
1620 {
1621 	struct vmmap *vmmap;
1622 
1623 	mtx_lock(&vmmaplock);
1624 	LIST_FOREACH(vmmap, &vmmaphead[VM_HASH(addr)], vm_next)
1625 		if (vmmap->vm_addr == addr)
1626 			break;
1627 	if (vmmap)
1628 		LIST_REMOVE(vmmap, vm_next);
1629 	mtx_unlock(&vmmaplock);
1630 
1631 	return (vmmap);
1632 }
1633 
1634 #if defined(__i386__) || defined(__amd64__) || defined(__powerpc__)
1635 void *
1636 _ioremap_attr(vm_paddr_t phys_addr, unsigned long size, int attr)
1637 {
1638 	void *addr;
1639 
1640 	addr = pmap_mapdev_attr(phys_addr, size, attr);
1641 	if (addr == NULL)
1642 		return (NULL);
1643 	vmmap_add(addr, size);
1644 
1645 	return (addr);
1646 }
1647 #endif
1648 
1649 void
1650 iounmap(void *addr)
1651 {
1652 	struct vmmap *vmmap;
1653 
1654 	vmmap = vmmap_remove(addr);
1655 	if (vmmap == NULL)
1656 		return;
1657 #if defined(__i386__) || defined(__amd64__) || defined(__powerpc__)
1658 	pmap_unmapdev((vm_offset_t)addr, vmmap->vm_size);
1659 #endif
1660 	kfree(vmmap);
1661 }
1662 
1663 
1664 void *
1665 vmap(struct page **pages, unsigned int count, unsigned long flags, int prot)
1666 {
1667 	vm_offset_t off;
1668 	size_t size;
1669 
1670 	size = count * PAGE_SIZE;
1671 	off = kva_alloc(size);
1672 	if (off == 0)
1673 		return (NULL);
1674 	vmmap_add((void *)off, size);
1675 	pmap_qenter(off, pages, count);
1676 
1677 	return ((void *)off);
1678 }
1679 
1680 void
1681 vunmap(void *addr)
1682 {
1683 	struct vmmap *vmmap;
1684 
1685 	vmmap = vmmap_remove(addr);
1686 	if (vmmap == NULL)
1687 		return;
1688 	pmap_qremove((vm_offset_t)addr, vmmap->vm_size / PAGE_SIZE);
1689 	kva_free((vm_offset_t)addr, vmmap->vm_size);
1690 	kfree(vmmap);
1691 }
1692 
1693 char *
1694 kvasprintf(gfp_t gfp, const char *fmt, va_list ap)
1695 {
1696 	unsigned int len;
1697 	char *p;
1698 	va_list aq;
1699 
1700 	va_copy(aq, ap);
1701 	len = vsnprintf(NULL, 0, fmt, aq);
1702 	va_end(aq);
1703 
1704 	p = kmalloc(len + 1, gfp);
1705 	if (p != NULL)
1706 		vsnprintf(p, len + 1, fmt, ap);
1707 
1708 	return (p);
1709 }
1710 
1711 char *
1712 kasprintf(gfp_t gfp, const char *fmt, ...)
1713 {
1714 	va_list ap;
1715 	char *p;
1716 
1717 	va_start(ap, fmt);
1718 	p = kvasprintf(gfp, fmt, ap);
1719 	va_end(ap);
1720 
1721 	return (p);
1722 }
1723 
1724 static void
1725 linux_timer_callback_wrapper(void *context)
1726 {
1727 	struct timer_list *timer;
1728 
1729 	linux_set_current(curthread);
1730 
1731 	timer = context;
1732 	timer->function(timer->data);
1733 }
1734 
1735 void
1736 mod_timer(struct timer_list *timer, int expires)
1737 {
1738 
1739 	timer->expires = expires;
1740 	callout_reset(&timer->callout,
1741 	    linux_timer_jiffies_until(expires),
1742 	    &linux_timer_callback_wrapper, timer);
1743 }
1744 
1745 void
1746 add_timer(struct timer_list *timer)
1747 {
1748 
1749 	callout_reset(&timer->callout,
1750 	    linux_timer_jiffies_until(timer->expires),
1751 	    &linux_timer_callback_wrapper, timer);
1752 }
1753 
1754 void
1755 add_timer_on(struct timer_list *timer, int cpu)
1756 {
1757 
1758 	callout_reset_on(&timer->callout,
1759 	    linux_timer_jiffies_until(timer->expires),
1760 	    &linux_timer_callback_wrapper, timer, cpu);
1761 }
1762 
1763 static void
1764 linux_timer_init(void *arg)
1765 {
1766 
1767 	/*
1768 	 * Compute an internal HZ value which can divide 2**32 to
1769 	 * avoid timer rounding problems when the tick value wraps
1770 	 * around 2**32:
1771 	 */
1772 	linux_timer_hz_mask = 1;
1773 	while (linux_timer_hz_mask < (unsigned long)hz)
1774 		linux_timer_hz_mask *= 2;
1775 	linux_timer_hz_mask--;
1776 }
1777 SYSINIT(linux_timer, SI_SUB_DRIVERS, SI_ORDER_FIRST, linux_timer_init, NULL);
1778 
1779 void
1780 linux_complete_common(struct completion *c, int all)
1781 {
1782 	int wakeup_swapper;
1783 
1784 	sleepq_lock(c);
1785 	if (all) {
1786 		c->done = UINT_MAX;
1787 		wakeup_swapper = sleepq_broadcast(c, SLEEPQ_SLEEP, 0, 0);
1788 	} else {
1789 		if (c->done != UINT_MAX)
1790 			c->done++;
1791 		wakeup_swapper = sleepq_signal(c, SLEEPQ_SLEEP, 0, 0);
1792 	}
1793 	sleepq_release(c);
1794 	if (wakeup_swapper)
1795 		kick_proc0();
1796 }
1797 
1798 /*
1799  * Indefinite wait for done != 0 with or without signals.
1800  */
1801 int
1802 linux_wait_for_common(struct completion *c, int flags)
1803 {
1804 	struct task_struct *task;
1805 	int error;
1806 
1807 	if (SCHEDULER_STOPPED())
1808 		return (0);
1809 
1810 	task = current;
1811 
1812 	if (flags != 0)
1813 		flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
1814 	else
1815 		flags = SLEEPQ_SLEEP;
1816 	error = 0;
1817 	for (;;) {
1818 		sleepq_lock(c);
1819 		if (c->done)
1820 			break;
1821 		sleepq_add(c, NULL, "completion", flags, 0);
1822 		if (flags & SLEEPQ_INTERRUPTIBLE) {
1823 			DROP_GIANT();
1824 			error = -sleepq_wait_sig(c, 0);
1825 			PICKUP_GIANT();
1826 			if (error != 0) {
1827 				linux_schedule_save_interrupt_value(task, error);
1828 				error = -ERESTARTSYS;
1829 				goto intr;
1830 			}
1831 		} else {
1832 			DROP_GIANT();
1833 			sleepq_wait(c, 0);
1834 			PICKUP_GIANT();
1835 		}
1836 	}
1837 	if (c->done != UINT_MAX)
1838 		c->done--;
1839 	sleepq_release(c);
1840 
1841 intr:
1842 	return (error);
1843 }
1844 
1845 /*
1846  * Time limited wait for done != 0 with or without signals.
1847  */
1848 int
1849 linux_wait_for_timeout_common(struct completion *c, int timeout, int flags)
1850 {
1851 	struct task_struct *task;
1852 	int end = jiffies + timeout;
1853 	int error;
1854 
1855 	if (SCHEDULER_STOPPED())
1856 		return (0);
1857 
1858 	task = current;
1859 
1860 	if (flags != 0)
1861 		flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
1862 	else
1863 		flags = SLEEPQ_SLEEP;
1864 
1865 	for (;;) {
1866 		sleepq_lock(c);
1867 		if (c->done)
1868 			break;
1869 		sleepq_add(c, NULL, "completion", flags, 0);
1870 		sleepq_set_timeout(c, linux_timer_jiffies_until(end));
1871 
1872 		DROP_GIANT();
1873 		if (flags & SLEEPQ_INTERRUPTIBLE)
1874 			error = -sleepq_timedwait_sig(c, 0);
1875 		else
1876 			error = -sleepq_timedwait(c, 0);
1877 		PICKUP_GIANT();
1878 
1879 		if (error != 0) {
1880 			/* check for timeout */
1881 			if (error == -EWOULDBLOCK) {
1882 				error = 0;	/* timeout */
1883 			} else {
1884 				/* signal happened */
1885 				linux_schedule_save_interrupt_value(task, error);
1886 				error = -ERESTARTSYS;
1887 			}
1888 			goto done;
1889 		}
1890 	}
1891 	if (c->done != UINT_MAX)
1892 		c->done--;
1893 	sleepq_release(c);
1894 
1895 	/* return how many jiffies are left */
1896 	error = linux_timer_jiffies_until(end);
1897 done:
1898 	return (error);
1899 }
1900 
1901 int
1902 linux_try_wait_for_completion(struct completion *c)
1903 {
1904 	int isdone;
1905 
1906 	sleepq_lock(c);
1907 	isdone = (c->done != 0);
1908 	if (c->done != 0 && c->done != UINT_MAX)
1909 		c->done--;
1910 	sleepq_release(c);
1911 	return (isdone);
1912 }
1913 
1914 int
1915 linux_completion_done(struct completion *c)
1916 {
1917 	int isdone;
1918 
1919 	sleepq_lock(c);
1920 	isdone = (c->done != 0);
1921 	sleepq_release(c);
1922 	return (isdone);
1923 }
1924 
1925 static void
1926 linux_cdev_release(struct kobject *kobj)
1927 {
1928 	struct linux_cdev *cdev;
1929 	struct kobject *parent;
1930 
1931 	cdev = container_of(kobj, struct linux_cdev, kobj);
1932 	parent = kobj->parent;
1933 	if (cdev->cdev)
1934 		destroy_dev(cdev->cdev);
1935 	kfree(cdev);
1936 	kobject_put(parent);
1937 }
1938 
1939 static void
1940 linux_cdev_static_release(struct kobject *kobj)
1941 {
1942 	struct linux_cdev *cdev;
1943 	struct kobject *parent;
1944 
1945 	cdev = container_of(kobj, struct linux_cdev, kobj);
1946 	parent = kobj->parent;
1947 	if (cdev->cdev)
1948 		destroy_dev(cdev->cdev);
1949 	kobject_put(parent);
1950 }
1951 
1952 const struct kobj_type linux_cdev_ktype = {
1953 	.release = linux_cdev_release,
1954 };
1955 
1956 const struct kobj_type linux_cdev_static_ktype = {
1957 	.release = linux_cdev_static_release,
1958 };
1959 
1960 static void
1961 linux_handle_ifnet_link_event(void *arg, struct ifnet *ifp, int linkstate)
1962 {
1963 	struct notifier_block *nb;
1964 
1965 	nb = arg;
1966 	if (linkstate == LINK_STATE_UP)
1967 		nb->notifier_call(nb, NETDEV_UP, ifp);
1968 	else
1969 		nb->notifier_call(nb, NETDEV_DOWN, ifp);
1970 }
1971 
1972 static void
1973 linux_handle_ifnet_arrival_event(void *arg, struct ifnet *ifp)
1974 {
1975 	struct notifier_block *nb;
1976 
1977 	nb = arg;
1978 	nb->notifier_call(nb, NETDEV_REGISTER, ifp);
1979 }
1980 
1981 static void
1982 linux_handle_ifnet_departure_event(void *arg, struct ifnet *ifp)
1983 {
1984 	struct notifier_block *nb;
1985 
1986 	nb = arg;
1987 	nb->notifier_call(nb, NETDEV_UNREGISTER, ifp);
1988 }
1989 
1990 static void
1991 linux_handle_iflladdr_event(void *arg, struct ifnet *ifp)
1992 {
1993 	struct notifier_block *nb;
1994 
1995 	nb = arg;
1996 	nb->notifier_call(nb, NETDEV_CHANGEADDR, ifp);
1997 }
1998 
1999 static void
2000 linux_handle_ifaddr_event(void *arg, struct ifnet *ifp)
2001 {
2002 	struct notifier_block *nb;
2003 
2004 	nb = arg;
2005 	nb->notifier_call(nb, NETDEV_CHANGEIFADDR, ifp);
2006 }
2007 
2008 int
2009 register_netdevice_notifier(struct notifier_block *nb)
2010 {
2011 
2012 	nb->tags[NETDEV_UP] = EVENTHANDLER_REGISTER(
2013 	    ifnet_link_event, linux_handle_ifnet_link_event, nb, 0);
2014 	nb->tags[NETDEV_REGISTER] = EVENTHANDLER_REGISTER(
2015 	    ifnet_arrival_event, linux_handle_ifnet_arrival_event, nb, 0);
2016 	nb->tags[NETDEV_UNREGISTER] = EVENTHANDLER_REGISTER(
2017 	    ifnet_departure_event, linux_handle_ifnet_departure_event, nb, 0);
2018 	nb->tags[NETDEV_CHANGEADDR] = EVENTHANDLER_REGISTER(
2019 	    iflladdr_event, linux_handle_iflladdr_event, nb, 0);
2020 
2021 	return (0);
2022 }
2023 
2024 int
2025 register_inetaddr_notifier(struct notifier_block *nb)
2026 {
2027 
2028 	nb->tags[NETDEV_CHANGEIFADDR] = EVENTHANDLER_REGISTER(
2029 	    ifaddr_event, linux_handle_ifaddr_event, nb, 0);
2030 	return (0);
2031 }
2032 
2033 int
2034 unregister_netdevice_notifier(struct notifier_block *nb)
2035 {
2036 
2037 	EVENTHANDLER_DEREGISTER(ifnet_link_event,
2038 	    nb->tags[NETDEV_UP]);
2039 	EVENTHANDLER_DEREGISTER(ifnet_arrival_event,
2040 	    nb->tags[NETDEV_REGISTER]);
2041 	EVENTHANDLER_DEREGISTER(ifnet_departure_event,
2042 	    nb->tags[NETDEV_UNREGISTER]);
2043 	EVENTHANDLER_DEREGISTER(iflladdr_event,
2044 	    nb->tags[NETDEV_CHANGEADDR]);
2045 
2046 	return (0);
2047 }
2048 
2049 int
2050 unregister_inetaddr_notifier(struct notifier_block *nb)
2051 {
2052 
2053 	EVENTHANDLER_DEREGISTER(ifaddr_event,
2054 	    nb->tags[NETDEV_CHANGEIFADDR]);
2055 
2056 	return (0);
2057 }
2058 
2059 struct list_sort_thunk {
2060 	int (*cmp)(void *, struct list_head *, struct list_head *);
2061 	void *priv;
2062 };
2063 
2064 static inline int
2065 linux_le_cmp(void *priv, const void *d1, const void *d2)
2066 {
2067 	struct list_head *le1, *le2;
2068 	struct list_sort_thunk *thunk;
2069 
2070 	thunk = priv;
2071 	le1 = *(__DECONST(struct list_head **, d1));
2072 	le2 = *(__DECONST(struct list_head **, d2));
2073 	return ((thunk->cmp)(thunk->priv, le1, le2));
2074 }
2075 
2076 void
2077 list_sort(void *priv, struct list_head *head, int (*cmp)(void *priv,
2078     struct list_head *a, struct list_head *b))
2079 {
2080 	struct list_sort_thunk thunk;
2081 	struct list_head **ar, *le;
2082 	size_t count, i;
2083 
2084 	count = 0;
2085 	list_for_each(le, head)
2086 		count++;
2087 	ar = malloc(sizeof(struct list_head *) * count, M_KMALLOC, M_WAITOK);
2088 	i = 0;
2089 	list_for_each(le, head)
2090 		ar[i++] = le;
2091 	thunk.cmp = cmp;
2092 	thunk.priv = priv;
2093 	qsort_r(ar, count, sizeof(struct list_head *), &thunk, linux_le_cmp);
2094 	INIT_LIST_HEAD(head);
2095 	for (i = 0; i < count; i++)
2096 		list_add_tail(ar[i], head);
2097 	free(ar, M_KMALLOC);
2098 }
2099 
2100 void
2101 linux_irq_handler(void *ent)
2102 {
2103 	struct irq_ent *irqe;
2104 
2105 	linux_set_current(curthread);
2106 
2107 	irqe = ent;
2108 	irqe->handler(irqe->irq, irqe->arg);
2109 }
2110 
2111 #if defined(__i386__) || defined(__amd64__)
2112 int
2113 linux_wbinvd_on_all_cpus(void)
2114 {
2115 
2116 	pmap_invalidate_cache();
2117 	return (0);
2118 }
2119 #endif
2120 
2121 int
2122 linux_on_each_cpu(void callback(void *), void *data)
2123 {
2124 
2125 	smp_rendezvous(smp_no_rendezvous_barrier, callback,
2126 	    smp_no_rendezvous_barrier, data);
2127 	return (0);
2128 }
2129 
2130 int
2131 linux_in_atomic(void)
2132 {
2133 
2134 	return ((curthread->td_pflags & TDP_NOFAULTING) != 0);
2135 }
2136 
2137 struct linux_cdev *
2138 linux_find_cdev(const char *name, unsigned major, unsigned minor)
2139 {
2140 	dev_t dev = MKDEV(major, minor);
2141 	struct cdev *cdev;
2142 
2143 	dev_lock();
2144 	LIST_FOREACH(cdev, &linuxcdevsw.d_devs, si_list) {
2145 		struct linux_cdev *ldev = cdev->si_drv1;
2146 		if (ldev->dev == dev &&
2147 		    strcmp(kobject_name(&ldev->kobj), name) == 0) {
2148 			break;
2149 		}
2150 	}
2151 	dev_unlock();
2152 
2153 	return (cdev != NULL ? cdev->si_drv1 : NULL);
2154 }
2155 
2156 int
2157 __register_chrdev(unsigned int major, unsigned int baseminor,
2158     unsigned int count, const char *name,
2159     const struct file_operations *fops)
2160 {
2161 	struct linux_cdev *cdev;
2162 	int ret = 0;
2163 	int i;
2164 
2165 	for (i = baseminor; i < baseminor + count; i++) {
2166 		cdev = cdev_alloc();
2167 		cdev_init(cdev, fops);
2168 		kobject_set_name(&cdev->kobj, name);
2169 
2170 		ret = cdev_add(cdev, makedev(major, i), 1);
2171 		if (ret != 0)
2172 			break;
2173 	}
2174 	return (ret);
2175 }
2176 
2177 int
2178 __register_chrdev_p(unsigned int major, unsigned int baseminor,
2179     unsigned int count, const char *name,
2180     const struct file_operations *fops, uid_t uid,
2181     gid_t gid, int mode)
2182 {
2183 	struct linux_cdev *cdev;
2184 	int ret = 0;
2185 	int i;
2186 
2187 	for (i = baseminor; i < baseminor + count; i++) {
2188 		cdev = cdev_alloc();
2189 		cdev_init(cdev, fops);
2190 		kobject_set_name(&cdev->kobj, name);
2191 
2192 		ret = cdev_add_ext(cdev, makedev(major, i), uid, gid, mode);
2193 		if (ret != 0)
2194 			break;
2195 	}
2196 	return (ret);
2197 }
2198 
2199 void
2200 __unregister_chrdev(unsigned int major, unsigned int baseminor,
2201     unsigned int count, const char *name)
2202 {
2203 	struct linux_cdev *cdevp;
2204 	int i;
2205 
2206 	for (i = baseminor; i < baseminor + count; i++) {
2207 		cdevp = linux_find_cdev(name, major, i);
2208 		if (cdevp != NULL)
2209 			cdev_del(cdevp);
2210 	}
2211 }
2212 
2213 void
2214 linux_dump_stack(void)
2215 {
2216 #ifdef STACK
2217 	struct stack st;
2218 
2219 	stack_zero(&st);
2220 	stack_save(&st);
2221 	stack_print(&st);
2222 #endif
2223 }
2224 
2225 #if defined(__i386__) || defined(__amd64__)
2226 bool linux_cpu_has_clflush;
2227 #endif
2228 
2229 static void
2230 linux_compat_init(void *arg)
2231 {
2232 	struct sysctl_oid *rootoid;
2233 	int i;
2234 
2235 #if defined(__i386__) || defined(__amd64__)
2236 	linux_cpu_has_clflush = (cpu_feature & CPUID_CLFSH);
2237 #endif
2238 	rw_init(&linux_vma_lock, "lkpi-vma-lock");
2239 
2240 	rootoid = SYSCTL_ADD_ROOT_NODE(NULL,
2241 	    OID_AUTO, "sys", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "sys");
2242 	kobject_init(&linux_class_root, &linux_class_ktype);
2243 	kobject_set_name(&linux_class_root, "class");
2244 	linux_class_root.oidp = SYSCTL_ADD_NODE(NULL, SYSCTL_CHILDREN(rootoid),
2245 	    OID_AUTO, "class", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "class");
2246 	kobject_init(&linux_root_device.kobj, &linux_dev_ktype);
2247 	kobject_set_name(&linux_root_device.kobj, "device");
2248 	linux_root_device.kobj.oidp = SYSCTL_ADD_NODE(NULL,
2249 	    SYSCTL_CHILDREN(rootoid), OID_AUTO, "device", CTLFLAG_RD, NULL,
2250 	    "device");
2251 	linux_root_device.bsddev = root_bus;
2252 	linux_class_misc.name = "misc";
2253 	class_register(&linux_class_misc);
2254 	INIT_LIST_HEAD(&pci_drivers);
2255 	INIT_LIST_HEAD(&pci_devices);
2256 	spin_lock_init(&pci_lock);
2257 	mtx_init(&vmmaplock, "IO Map lock", NULL, MTX_DEF);
2258 	for (i = 0; i < VMMAP_HASH_SIZE; i++)
2259 		LIST_INIT(&vmmaphead[i]);
2260 }
2261 SYSINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_init, NULL);
2262 
2263 static void
2264 linux_compat_uninit(void *arg)
2265 {
2266 	linux_kobject_kfree_name(&linux_class_root);
2267 	linux_kobject_kfree_name(&linux_root_device.kobj);
2268 	linux_kobject_kfree_name(&linux_class_misc.kobj);
2269 
2270 	mtx_destroy(&vmmaplock);
2271 	spin_lock_destroy(&pci_lock);
2272 	rw_destroy(&linux_vma_lock);
2273 }
2274 SYSUNINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_uninit, NULL);
2275 
2276 /*
2277  * NOTE: Linux frequently uses "unsigned long" for pointer to integer
2278  * conversion and vice versa, where in FreeBSD "uintptr_t" would be
2279  * used. Assert these types have the same size, else some parts of the
2280  * LinuxKPI may not work like expected:
2281  */
2282 CTASSERT(sizeof(unsigned long) == sizeof(uintptr_t));
2283