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, 2014 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/param.h>
31 #include <sys/systm.h>
32 #include <sys/malloc.h>
33 #include <sys/kernel.h>
34 #include <sys/sysctl.h>
35 #include <sys/proc.h>
36 #include <sys/sglist.h>
37 #include <sys/sleepqueue.h>
38 #include <sys/lock.h>
39 #include <sys/mutex.h>
40 #include <sys/bus.h>
41 #include <sys/fcntl.h>
42 #include <sys/file.h>
43 #include <sys/filio.h>
44 
45 #include <vm/vm.h>
46 #include <vm/pmap.h>
47 
48 #include <machine/stdarg.h>
49 #include <machine/pmap.h>
50 
51 #include <linux/kobject.h>
52 #include <linux/device.h>
53 #include <linux/slab.h>
54 #include <linux/module.h>
55 #include <linux/cdev.h>
56 #include <linux/file.h>
57 #include <linux/sysfs.h>
58 #include <linux/mm.h>
59 #include <linux/io.h>
60 #include <linux/vmalloc.h>
61 #include <linux/timer.h>
62 #include <linux/netdevice.h>
63 
64 #include <vm/vm_pager.h>
65 
66 MALLOC_DEFINE(M_KMALLOC, "linux", "Linux kmalloc compat");
67 
68 #include <linux/rbtree.h>
69 /* Undo Linux compat changes. */
70 #undef RB_ROOT
71 #undef file
72 #undef cdev
73 #define	RB_ROOT(head)	(head)->rbh_root
74 
75 struct kobject class_root;
76 struct device linux_rootdev;
77 struct class miscclass;
78 struct list_head pci_drivers;
79 struct list_head pci_devices;
80 struct net init_net;
81 spinlock_t pci_lock;
82 
83 unsigned long linux_timer_hz_mask;
84 
85 int
panic_cmp(struct rb_node * one,struct rb_node * two)86 panic_cmp(struct rb_node *one, struct rb_node *two)
87 {
88 	panic("no cmp");
89 }
90 
91 RB_GENERATE(linux_root, rb_node, __entry, panic_cmp);
92 
93 int
kobject_set_name(struct kobject * kobj,const char * fmt,...)94 kobject_set_name(struct kobject *kobj, const char *fmt, ...)
95 {
96 	va_list args;
97 	int error;
98 
99 	va_start(args, fmt);
100 	error = kobject_set_name_vargs(kobj, fmt, args);
101 	va_end(args);
102 
103 	return (error);
104 }
105 
106 static inline int
kobject_add_complete(struct kobject * kobj,struct kobject * parent)107 kobject_add_complete(struct kobject *kobj, struct kobject *parent)
108 {
109 	struct kobj_type *t;
110 	int error;
111 
112 	kobj->parent = kobject_get(parent);
113 	error = sysfs_create_dir(kobj);
114 	if (error == 0 && kobj->ktype && kobj->ktype->default_attrs) {
115 		struct attribute **attr;
116 		t = kobj->ktype;
117 
118 		for (attr = t->default_attrs; *attr != NULL; attr++) {
119 			error = sysfs_create_file(kobj, *attr);
120 			if (error)
121 				break;
122 		}
123 		if (error)
124 			sysfs_remove_dir(kobj);
125 
126 	}
127 	return (error);
128 }
129 
130 int
kobject_add(struct kobject * kobj,struct kobject * parent,const char * fmt,...)131 kobject_add(struct kobject *kobj, struct kobject *parent, const char *fmt, ...)
132 {
133 	va_list args;
134 	int error;
135 
136 	va_start(args, fmt);
137 	error = kobject_set_name_vargs(kobj, fmt, args);
138 	va_end(args);
139 	if (error)
140 		return (error);
141 
142 	return kobject_add_complete(kobj, parent);
143 }
144 
145 void
kobject_release(struct kref * kref)146 kobject_release(struct kref *kref)
147 {
148 	struct kobject *kobj;
149 	char *name;
150 
151 	kobj = container_of(kref, struct kobject, kref);
152 	sysfs_remove_dir(kobj);
153 	if (kobj->parent)
154 		kobject_put(kobj->parent);
155 	kobj->parent = NULL;
156 	name = kobj->name;
157 	if (kobj->ktype && kobj->ktype->release)
158 		kobj->ktype->release(kobj);
159 	kfree(name);
160 }
161 
162 static void
kobject_kfree(struct kobject * kobj)163 kobject_kfree(struct kobject *kobj)
164 {
165 	kfree(kobj);
166 }
167 
168 static void
kobject_kfree_name(struct kobject * kobj)169 kobject_kfree_name(struct kobject *kobj)
170 {
171 	if (kobj) {
172 		kfree(kobj->name);
173 	}
174 }
175 
176 struct kobj_type kfree_type = { .release = kobject_kfree };
177 
178 static void
dev_release(struct device * dev)179 dev_release(struct device *dev)
180 {
181 	pr_debug("dev_release: %s\n", dev_name(dev));
182 	kfree(dev);
183 }
184 
185 struct device *
device_create(struct class * class,struct device * parent,dev_t devt,void * drvdata,const char * fmt,...)186 device_create(struct class *class, struct device *parent, dev_t devt,
187     void *drvdata, const char *fmt, ...)
188 {
189 	struct device *dev;
190 	va_list args;
191 
192 	dev = kzalloc(sizeof(*dev), M_WAITOK);
193 	dev->parent = parent;
194 	dev->class = class;
195 	dev->devt = devt;
196 	dev->driver_data = drvdata;
197 	dev->release = dev_release;
198 	va_start(args, fmt);
199 	kobject_set_name_vargs(&dev->kobj, fmt, args);
200 	va_end(args);
201 	device_register(dev);
202 
203 	return (dev);
204 }
205 
206 int
kobject_init_and_add(struct kobject * kobj,struct kobj_type * ktype,struct kobject * parent,const char * fmt,...)207 kobject_init_and_add(struct kobject *kobj, struct kobj_type *ktype,
208     struct kobject *parent, const char *fmt, ...)
209 {
210 	va_list args;
211 	int error;
212 
213 	kobject_init(kobj, ktype);
214 	kobj->ktype = ktype;
215 	kobj->parent = parent;
216 	kobj->name = NULL;
217 
218 	va_start(args, fmt);
219 	error = kobject_set_name_vargs(kobj, fmt, args);
220 	va_end(args);
221 	if (error)
222 		return (error);
223 	return kobject_add_complete(kobj, parent);
224 }
225 
226 static void
linux_file_dtor(void * cdp)227 linux_file_dtor(void *cdp)
228 {
229 	struct linux_file *filp;
230 
231 	filp = cdp;
232 	filp->f_op->release(filp->f_vnode, filp);
233 	vdrop(filp->f_vnode);
234 	kfree(filp);
235 }
236 
237 static int
linux_dev_open(struct cdev * dev,int oflags,int devtype,struct thread * td)238 linux_dev_open(struct cdev *dev, int oflags, int devtype, struct thread *td)
239 {
240 	struct linux_cdev *ldev;
241 	struct linux_file *filp;
242 	struct file *file;
243 	int error;
244 
245 	file = curthread->td_fpop;
246 	ldev = dev->si_drv1;
247 	if (ldev == NULL)
248 		return (ENODEV);
249 	filp = kzalloc(sizeof(*filp), GFP_KERNEL);
250 	filp->f_dentry = &filp->f_dentry_store;
251 	filp->f_op = ldev->ops;
252 	filp->f_flags = file->f_flag;
253 	vhold(file->f_vnode);
254 	filp->f_vnode = file->f_vnode;
255 	if (filp->f_op->open) {
256 		error = -filp->f_op->open(file->f_vnode, filp);
257 		if (error) {
258 			kfree(filp);
259 			return (error);
260 		}
261 	}
262 	error = devfs_set_cdevpriv(filp, linux_file_dtor);
263 	if (error) {
264 		filp->f_op->release(file->f_vnode, filp);
265 		kfree(filp);
266 		return (error);
267 	}
268 
269 	return 0;
270 }
271 
272 static int
linux_dev_close(struct cdev * dev,int fflag,int devtype,struct thread * td)273 linux_dev_close(struct cdev *dev, int fflag, int devtype, struct thread *td)
274 {
275 	struct linux_cdev *ldev;
276 	struct linux_file *filp;
277 	struct file *file;
278 	int error;
279 
280 	file = curthread->td_fpop;
281 	ldev = dev->si_drv1;
282 	if (ldev == NULL)
283 		return (0);
284 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
285 		return (error);
286 	filp->f_flags = file->f_flag;
287         devfs_clear_cdevpriv();
288 
289 
290 	return (0);
291 }
292 
293 static int
linux_dev_ioctl(struct cdev * dev,u_long cmd,caddr_t data,int fflag,struct thread * td)294 linux_dev_ioctl(struct cdev *dev, u_long cmd, caddr_t data, int fflag,
295     struct thread *td)
296 {
297 	struct linux_cdev *ldev;
298 	struct linux_file *filp;
299 	struct file *file;
300 	int error;
301 
302 	file = curthread->td_fpop;
303 	ldev = dev->si_drv1;
304 	if (ldev == NULL)
305 		return (0);
306 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
307 		return (error);
308 	filp->f_flags = file->f_flag;
309 	/*
310 	 * Linux does not have a generic ioctl copyin/copyout layer.  All
311 	 * linux ioctls must be converted to void ioctls which pass a
312 	 * pointer to the address of the data.  We want the actual user
313 	 * address so we dereference here.
314 	 */
315 	data = *(void **)data;
316 	if (filp->f_op->unlocked_ioctl)
317 		error = -filp->f_op->unlocked_ioctl(filp, cmd, (u_long)data);
318 	else
319 		error = ENOTTY;
320 
321 	return (error);
322 }
323 
324 static int
linux_dev_read(struct cdev * dev,struct uio * uio,int ioflag)325 linux_dev_read(struct cdev *dev, struct uio *uio, int ioflag)
326 {
327 	struct linux_cdev *ldev;
328 	struct linux_file *filp;
329 	struct file *file;
330 	ssize_t bytes;
331 	int error;
332 
333 	file = curthread->td_fpop;
334 	ldev = dev->si_drv1;
335 	if (ldev == NULL)
336 		return (0);
337 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
338 		return (error);
339 	filp->f_flags = file->f_flag;
340 	if (uio->uio_iovcnt != 1)
341 		panic("linux_dev_read: uio %p iovcnt %d",
342 		    uio, uio->uio_iovcnt);
343 	if (filp->f_op->read) {
344 		bytes = filp->f_op->read(filp, uio->uio_iov->iov_base,
345 		    uio->uio_iov->iov_len, &uio->uio_offset);
346 		if (bytes >= 0) {
347 			uio->uio_iov->iov_base += bytes;
348 			uio->uio_iov->iov_len -= bytes;
349 			uio->uio_resid -= bytes;
350 		} else
351 			error = -bytes;
352 	} else
353 		error = ENXIO;
354 
355 	return (error);
356 }
357 
358 static int
linux_dev_write(struct cdev * dev,struct uio * uio,int ioflag)359 linux_dev_write(struct cdev *dev, struct uio *uio, int ioflag)
360 {
361 	struct linux_cdev *ldev;
362 	struct linux_file *filp;
363 	struct file *file;
364 	ssize_t bytes;
365 	int error;
366 
367 	file = curthread->td_fpop;
368 	ldev = dev->si_drv1;
369 	if (ldev == NULL)
370 		return (0);
371 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
372 		return (error);
373 	filp->f_flags = file->f_flag;
374 	if (uio->uio_iovcnt != 1)
375 		panic("linux_dev_write: uio %p iovcnt %d",
376 		    uio, uio->uio_iovcnt);
377 	if (filp->f_op->write) {
378 		bytes = filp->f_op->write(filp, uio->uio_iov->iov_base,
379 		    uio->uio_iov->iov_len, &uio->uio_offset);
380 		if (bytes >= 0) {
381 			uio->uio_iov->iov_base += bytes;
382 			uio->uio_iov->iov_len -= bytes;
383 			uio->uio_resid -= bytes;
384 		} else
385 			error = -bytes;
386 	} else
387 		error = ENXIO;
388 
389 	return (error);
390 }
391 
392 static int
linux_dev_poll(struct cdev * dev,int events,struct thread * td)393 linux_dev_poll(struct cdev *dev, int events, struct thread *td)
394 {
395 	struct linux_cdev *ldev;
396 	struct linux_file *filp;
397 	struct file *file;
398 	int revents;
399 	int error;
400 
401 	file = curthread->td_fpop;
402 	ldev = dev->si_drv1;
403 	if (ldev == NULL)
404 		return (0);
405 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
406 		return (error);
407 	filp->f_flags = file->f_flag;
408 	if (filp->f_op->poll)
409 		revents = filp->f_op->poll(filp, NULL) & events;
410 	else
411 		revents = 0;
412 
413 	return (revents);
414 }
415 
416 static int
linux_dev_mmap_single(struct cdev * dev,vm_ooffset_t * offset,vm_size_t size,struct vm_object ** object,int nprot)417 linux_dev_mmap_single(struct cdev *dev, vm_ooffset_t *offset,
418     vm_size_t size, struct vm_object **object, int nprot)
419 {
420 	struct linux_cdev *ldev;
421 	struct linux_file *filp;
422 	struct file *file;
423 	struct vm_area_struct vma;
424 	int error;
425 
426 	file = curthread->td_fpop;
427 	ldev = dev->si_drv1;
428 	if (ldev == NULL)
429 		return (ENODEV);
430 	if ((error = devfs_get_cdevpriv((void **)&filp)) != 0)
431 		return (error);
432 	filp->f_flags = file->f_flag;
433 	vma.vm_start = 0;
434 	vma.vm_end = size;
435 	vma.vm_pgoff = *offset / PAGE_SIZE;
436 	vma.vm_pfn = 0;
437 	vma.vm_page_prot = VM_MEMATTR_DEFAULT;
438 	if (filp->f_op->mmap) {
439 		error = -filp->f_op->mmap(filp, &vma);
440 		if (error == 0) {
441 			struct sglist *sg;
442 
443 			sg = sglist_alloc(1, M_WAITOK);
444 			sglist_append_phys(sg,
445 			    (vm_paddr_t)vma.vm_pfn << PAGE_SHIFT, vma.vm_len);
446 			*object = vm_pager_allocate(OBJT_SG, sg, vma.vm_len,
447 			    nprot, 0, curthread->td_ucred);
448 		        if (*object == NULL) {
449 				sglist_free(sg);
450 				return (EINVAL);
451 			}
452 			*offset = 0;
453 			if (vma.vm_page_prot != VM_MEMATTR_DEFAULT) {
454 				VM_OBJECT_LOCK(*object);
455 				vm_object_set_memattr(*object,
456 				    vma.vm_page_prot);
457 				VM_OBJECT_UNLOCK(*object);
458 			}
459 		}
460 	} else
461 		error = ENODEV;
462 
463 	return (error);
464 }
465 
466 struct cdevsw linuxcdevsw = {
467 	.d_version = D_VERSION,
468 	.d_flags = D_TRACKCLOSE,
469 	.d_open = linux_dev_open,
470 	.d_close = linux_dev_close,
471 	.d_read = linux_dev_read,
472 	.d_write = linux_dev_write,
473 	.d_ioctl = linux_dev_ioctl,
474 	.d_mmap_single = linux_dev_mmap_single,
475 	.d_poll = linux_dev_poll,
476 };
477 
478 static int
linux_file_read(struct file * file,struct uio * uio,struct ucred * active_cred,int flags,struct thread * td)479 linux_file_read(struct file *file, struct uio *uio, struct ucred *active_cred,
480     int flags, struct thread *td)
481 {
482 	struct linux_file *filp;
483 	ssize_t bytes;
484 	int error;
485 
486 	error = 0;
487 	filp = (struct linux_file *)file->f_data;
488 	filp->f_flags = file->f_flag;
489 	if (uio->uio_iovcnt != 1)
490 		panic("linux_file_read: uio %p iovcnt %d",
491 		    uio, uio->uio_iovcnt);
492 	if (filp->f_op->read) {
493 		bytes = filp->f_op->read(filp, uio->uio_iov->iov_base,
494 		    uio->uio_iov->iov_len, &uio->uio_offset);
495 		if (bytes >= 0) {
496 			uio->uio_iov->iov_base += bytes;
497 			uio->uio_iov->iov_len -= bytes;
498 			uio->uio_resid -= bytes;
499 		} else
500 			error = -bytes;
501 	} else
502 		error = ENXIO;
503 
504 	return (error);
505 }
506 
507 static int
linux_file_poll(struct file * file,int events,struct ucred * active_cred,struct thread * td)508 linux_file_poll(struct file *file, int events, struct ucred *active_cred,
509     struct thread *td)
510 {
511 	struct linux_file *filp;
512 	int revents;
513 
514 	filp = (struct linux_file *)file->f_data;
515 	filp->f_flags = file->f_flag;
516 	if (filp->f_op->poll)
517 		revents = filp->f_op->poll(filp, NULL) & events;
518 	else
519 		revents = 0;
520 
521 	return (0);
522 }
523 
524 static int
linux_file_close(struct file * file,struct thread * td)525 linux_file_close(struct file *file, struct thread *td)
526 {
527 	struct linux_file *filp;
528 	int error;
529 
530 	filp = (struct linux_file *)file->f_data;
531 	filp->f_flags = file->f_flag;
532 	error = -filp->f_op->release(NULL, filp);
533 	funsetown(&filp->f_sigio);
534 	kfree(filp);
535 
536 	return (error);
537 }
538 
539 static int
linux_file_ioctl(struct file * fp,u_long cmd,void * data,struct ucred * cred,struct thread * td)540 linux_file_ioctl(struct file *fp, u_long cmd, void *data, struct ucred *cred,
541     struct thread *td)
542 {
543 	struct linux_file *filp;
544 	int error;
545 
546 	filp = (struct linux_file *)fp->f_data;
547 	filp->f_flags = fp->f_flag;
548 	error = 0;
549 
550 	switch (cmd) {
551 	case FIONBIO:
552 		break;
553 	case FIOASYNC:
554 		if (filp->f_op->fasync == NULL)
555 			break;
556 		error = filp->f_op->fasync(0, filp, fp->f_flag & FASYNC);
557 		break;
558 	case FIOSETOWN:
559 		error = fsetown(*(int *)data, &filp->f_sigio);
560 		if (error == 0)
561 			error = filp->f_op->fasync(0, filp,
562 			    fp->f_flag & FASYNC);
563 		break;
564 	case FIOGETOWN:
565 		*(int *)data = fgetown(&filp->f_sigio);
566 		break;
567 	default:
568 		error = ENOTTY;
569 		break;
570 	}
571 	return (error);
572 }
573 
574 struct fileops linuxfileops = {
575 	.fo_read = linux_file_read,
576 	.fo_poll = linux_file_poll,
577 	.fo_close = linux_file_close,
578 	.fo_ioctl = linux_file_ioctl,
579 	.fo_chmod = invfo_chmod,
580 	.fo_chown = invfo_chown,
581 };
582 
583 /*
584  * Hash of vmmap addresses.  This is infrequently accessed and does not
585  * need to be particularly large.  This is done because we must store the
586  * caller's idea of the map size to properly unmap.
587  */
588 struct vmmap {
589 	LIST_ENTRY(vmmap)	vm_next;
590 	void 			*vm_addr;
591 	unsigned long		vm_size;
592 };
593 
594 struct vmmaphd {
595 	struct vmmap *lh_first;
596 };
597 #define	VMMAP_HASH_SIZE	64
598 #define	VMMAP_HASH_MASK	(VMMAP_HASH_SIZE - 1)
599 #define	VM_HASH(addr)	((uintptr_t)(addr) >> PAGE_SHIFT) & VMMAP_HASH_MASK
600 static struct vmmaphd vmmaphead[VMMAP_HASH_SIZE];
601 static struct mtx vmmaplock;
602 
603 static void
vmmap_add(void * addr,unsigned long size)604 vmmap_add(void *addr, unsigned long size)
605 {
606 	struct vmmap *vmmap;
607 
608 	vmmap = kmalloc(sizeof(*vmmap), GFP_KERNEL);
609 	mtx_lock(&vmmaplock);
610 	vmmap->vm_size = size;
611 	vmmap->vm_addr = addr;
612 	LIST_INSERT_HEAD(&vmmaphead[VM_HASH(addr)], vmmap, vm_next);
613 	mtx_unlock(&vmmaplock);
614 }
615 
616 static struct vmmap *
vmmap_remove(void * addr)617 vmmap_remove(void *addr)
618 {
619 	struct vmmap *vmmap;
620 
621 	mtx_lock(&vmmaplock);
622 	LIST_FOREACH(vmmap, &vmmaphead[VM_HASH(addr)], vm_next)
623 		if (vmmap->vm_addr == addr)
624 			break;
625 	if (vmmap)
626 		LIST_REMOVE(vmmap, vm_next);
627 	mtx_unlock(&vmmaplock);
628 
629 	return (vmmap);
630 }
631 
632 void *
_ioremap_attr(vm_paddr_t phys_addr,unsigned long size,int attr)633 _ioremap_attr(vm_paddr_t phys_addr, unsigned long size, int attr)
634 {
635 	void *addr;
636 
637 	addr = pmap_mapdev_attr(phys_addr, size, attr);
638 	if (addr == NULL)
639 		return (NULL);
640 	vmmap_add(addr, size);
641 
642 	return (addr);
643 }
644 
645 void
iounmap(void * addr)646 iounmap(void *addr)
647 {
648 	struct vmmap *vmmap;
649 
650 	vmmap = vmmap_remove(addr);
651 	if (vmmap == NULL)
652 		return;
653 	pmap_unmapdev((vm_offset_t)addr, vmmap->vm_size);
654 	kfree(vmmap);
655 }
656 
657 
658 void *
vmap(struct page ** pages,unsigned int count,unsigned long flags,int prot)659 vmap(struct page **pages, unsigned int count, unsigned long flags, int prot)
660 {
661 	vm_offset_t off;
662 	size_t size;
663 
664 	size = count * PAGE_SIZE;
665 	off = kmem_alloc_nofault(kernel_map, size);
666 	if (off == 0)
667 		return (NULL);
668 	vmmap_add((void *)off, size);
669 	pmap_qenter(off, pages, count);
670 
671 	return ((void *)off);
672 }
673 
674 void
vunmap(void * addr)675 vunmap(void *addr)
676 {
677 	struct vmmap *vmmap;
678 
679 	vmmap = vmmap_remove(addr);
680 	if (vmmap == NULL)
681 		return;
682 	pmap_qremove((vm_offset_t)addr, vmmap->vm_size / PAGE_SIZE);
683 	kmem_free(kernel_map, (vm_offset_t)addr, vmmap->vm_size);
684 	kfree(vmmap);
685 }
686 
687 char *
kvasprintf(gfp_t gfp,const char * fmt,va_list ap)688 kvasprintf(gfp_t gfp, const char *fmt, va_list ap)
689 {
690 	unsigned int len;
691 	char *p;
692 	va_list aq;
693 
694 	va_copy(aq, ap);
695 	len = vsnprintf(NULL, 0, fmt, aq);
696 	va_end(aq);
697 
698 	p = kmalloc(len + 1, gfp);
699 	if (p != NULL)
700 		vsnprintf(p, len + 1, fmt, ap);
701 
702 	return (p);
703 }
704 
705 char *
kasprintf(gfp_t gfp,const char * fmt,...)706 kasprintf(gfp_t gfp, const char *fmt, ...)
707 {
708 	va_list ap;
709 	char *p;
710 
711 	va_start(ap, fmt);
712 	p = kvasprintf(gfp, fmt, ap);
713 	va_end(ap);
714 
715 	return (p);
716 }
717 
718 static int
linux_timer_jiffies_until(unsigned long expires)719 linux_timer_jiffies_until(unsigned long expires)
720 {
721 	int delta = expires - jiffies;
722 	/* guard against already expired values */
723 	if (delta < 1)
724 		delta = 1;
725 	return (delta);
726 }
727 
728 static void
linux_timer_callback_wrapper(void * context)729 linux_timer_callback_wrapper(void *context)
730 {
731 	struct timer_list *timer;
732 
733 	timer = context;
734 	timer->function(timer->data);
735 }
736 
737 void
mod_timer(struct timer_list * timer,unsigned long expires)738 mod_timer(struct timer_list *timer, unsigned long expires)
739 {
740 
741 	timer->expires = expires;
742 	callout_reset(&timer->timer_callout,
743 	    linux_timer_jiffies_until(expires),
744 	    &linux_timer_callback_wrapper, timer);
745 }
746 
747 void
add_timer(struct timer_list * timer)748 add_timer(struct timer_list *timer)
749 {
750 
751 	callout_reset(&timer->timer_callout,
752 	    linux_timer_jiffies_until(timer->expires),
753 	    &linux_timer_callback_wrapper, timer);
754 }
755 
756 static void
linux_timer_init(void * arg)757 linux_timer_init(void *arg)
758 {
759 
760 	/*
761 	 * Compute an internal HZ value which can divide 2**32 to
762 	 * avoid timer rounding problems when the tick value wraps
763 	 * around 2**32:
764 	 */
765 	linux_timer_hz_mask = 1;
766 	while (linux_timer_hz_mask < (unsigned long)hz)
767 		linux_timer_hz_mask *= 2;
768 	linux_timer_hz_mask--;
769 }
770 SYSINIT(linux_timer, SI_SUB_DRIVERS, SI_ORDER_FIRST, linux_timer_init, NULL);
771 
772 void
linux_complete_common(struct completion * c,int all)773 linux_complete_common(struct completion *c, int all)
774 {
775 	int wakeup_swapper;
776 
777 	sleepq_lock(c);
778 	c->done++;
779 	if (all)
780 		wakeup_swapper = sleepq_broadcast(c, SLEEPQ_SLEEP, 0, 0);
781 	else
782 		wakeup_swapper = sleepq_signal(c, SLEEPQ_SLEEP, 0, 0);
783 	sleepq_release(c);
784 	if (wakeup_swapper)
785 		kick_proc0();
786 }
787 
788 /*
789  * Indefinite wait for done != 0 with or without signals.
790  */
791 long
linux_wait_for_common(struct completion * c,int flags)792 linux_wait_for_common(struct completion *c, int flags)
793 {
794 
795 	if (flags != 0)
796 		flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
797 	else
798 		flags = SLEEPQ_SLEEP;
799 	for (;;) {
800 		sleepq_lock(c);
801 		if (c->done)
802 			break;
803 		sleepq_add(c, NULL, "completion", flags, 0);
804 		if (flags & SLEEPQ_INTERRUPTIBLE) {
805 			if (sleepq_wait_sig(c, 0) != 0)
806 				return (-ERESTARTSYS);
807 		} else
808 			sleepq_wait(c, 0);
809 	}
810 	c->done--;
811 	sleepq_release(c);
812 
813 	return (0);
814 }
815 
816 /*
817  * Time limited wait for done != 0 with or without signals.
818  */
819 long
linux_wait_for_timeout_common(struct completion * c,long timeout,int flags)820 linux_wait_for_timeout_common(struct completion *c, long timeout, int flags)
821 {
822 	long end = jiffies + timeout;
823 
824 	if (flags != 0)
825 		flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
826 	else
827 		flags = SLEEPQ_SLEEP;
828 	for (;;) {
829 		int ret;
830 
831 		sleepq_lock(c);
832 		if (c->done)
833 			break;
834 		sleepq_add(c, NULL, "completion", flags, 0);
835 		sleepq_set_timeout(c, linux_timer_jiffies_until(end));
836 		if (flags & SLEEPQ_INTERRUPTIBLE)
837 			ret = sleepq_timedwait_sig(c, 0);
838 		else
839 			ret = sleepq_timedwait(c, 0);
840 		if (ret != 0) {
841 			/* check for timeout or signal */
842 			if (ret == EWOULDBLOCK)
843 				return (0);
844 			else
845 				return (-ERESTARTSYS);
846 		}
847 	}
848 	c->done--;
849 	sleepq_release(c);
850 
851 	/* return how many jiffies are left */
852 	return (linux_timer_jiffies_until(end));
853 }
854 
855 int
linux_try_wait_for_completion(struct completion * c)856 linux_try_wait_for_completion(struct completion *c)
857 {
858 	int isdone;
859 
860 	isdone = 1;
861 	sleepq_lock(c);
862 	if (c->done)
863 		c->done--;
864 	else
865 		isdone = 0;
866 	sleepq_release(c);
867 	return (isdone);
868 }
869 
870 int
linux_completion_done(struct completion * c)871 linux_completion_done(struct completion *c)
872 {
873 	int isdone;
874 
875 	isdone = 1;
876 	sleepq_lock(c);
877 	if (c->done == 0)
878 		isdone = 0;
879 	sleepq_release(c);
880 	return (isdone);
881 }
882 
883 static void
linux_compat_init(void * arg)884 linux_compat_init(void *arg)
885 {
886 	struct sysctl_oid *rootoid;
887 	int i;
888 
889 	rootoid = SYSCTL_ADD_NODE(NULL, SYSCTL_STATIC_CHILDREN(),
890 	    OID_AUTO, "sys", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "sys");
891 	kobject_init(&class_root, &class_ktype);
892 	kobject_set_name(&class_root, "class");
893 	class_root.oidp = SYSCTL_ADD_NODE(NULL, SYSCTL_CHILDREN(rootoid),
894 	    OID_AUTO, "class", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "class");
895 	kobject_init(&linux_rootdev.kobj, &dev_ktype);
896 	kobject_set_name(&linux_rootdev.kobj, "device");
897 	linux_rootdev.kobj.oidp = SYSCTL_ADD_NODE(NULL,
898 	    SYSCTL_CHILDREN(rootoid), OID_AUTO, "device", CTLFLAG_RD, NULL,
899 	    "device");
900 	linux_rootdev.bsddev = root_bus;
901 	miscclass.name = "misc";
902 	class_register(&miscclass);
903 	INIT_LIST_HEAD(&pci_drivers);
904 	INIT_LIST_HEAD(&pci_devices);
905 	spin_lock_init(&pci_lock);
906 	mtx_init(&vmmaplock, "IO Map lock", NULL, MTX_DEF);
907 	for (i = 0; i < VMMAP_HASH_SIZE; i++)
908 		LIST_INIT(&vmmaphead[i]);
909 }
910 
911 SYSINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_init, NULL);
912 
913 static void
linux_compat_uninit(void * arg)914 linux_compat_uninit(void *arg)
915 {
916 	kobject_kfree_name(&class_root);
917 	kobject_kfree_name(&linux_rootdev.kobj);
918 	kobject_kfree_name(&miscclass.kobj);
919 }
920 SYSUNINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_uninit, NULL);
921