xref: /freebsd-13-stable/sys/compat/ndis/subr_ntoskrnl.c (revision 3bc80996974a61a4223eae4c1ccd47b6ee32a48a)
1 /*-
2  * SPDX-License-Identifier: BSD-4-Clause
3  *
4  * Copyright (c) 2003
5  *	Bill Paul <wpaul@windriver.com>.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by Bill Paul.
18  * 4. Neither the name of the author nor the names of any co-contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY Bill Paul AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL Bill Paul OR THE VOICES IN HIS HEAD
26  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
27  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
28  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
29  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
30  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
31  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
32  * THE POSSIBILITY OF SUCH DAMAGE.
33  */
34 
35 #include <sys/cdefs.h>
36 #include <sys/ctype.h>
37 #include <sys/unistd.h>
38 #include <sys/param.h>
39 #include <sys/types.h>
40 #include <sys/errno.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
43 #include <sys/lock.h>
44 #include <sys/mutex.h>
45 
46 #include <sys/callout.h>
47 #include <sys/kdb.h>
48 #include <sys/kernel.h>
49 #include <sys/proc.h>
50 #include <sys/condvar.h>
51 #include <sys/kthread.h>
52 #include <sys/module.h>
53 #include <sys/smp.h>
54 #include <sys/sched.h>
55 #include <sys/sysctl.h>
56 
57 #include <machine/atomic.h>
58 #include <machine/bus.h>
59 #include <machine/stdarg.h>
60 #include <machine/resource.h>
61 
62 #include <sys/bus.h>
63 #include <sys/rman.h>
64 
65 #include <vm/vm.h>
66 #include <vm/vm_param.h>
67 #include <vm/pmap.h>
68 #include <vm/uma.h>
69 #include <vm/vm_kern.h>
70 #include <vm/vm_map.h>
71 #include <vm/vm_extern.h>
72 
73 #include <compat/ndis/pe_var.h>
74 #include <compat/ndis/cfg_var.h>
75 #include <compat/ndis/resource_var.h>
76 #include <compat/ndis/ntoskrnl_var.h>
77 #include <compat/ndis/hal_var.h>
78 #include <compat/ndis/ndis_var.h>
79 
80 #ifdef NTOSKRNL_DEBUG_TIMERS
81 static int sysctl_show_timers(SYSCTL_HANDLER_ARGS);
82 
83 SYSCTL_PROC(_debug, OID_AUTO, ntoskrnl_timers,
84     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, NULL, 0,
85     sysctl_show_timers, "I",
86     "Show ntoskrnl timer stats");
87 #endif
88 
89 struct kdpc_queue {
90 	list_entry		kq_disp;
91 	struct thread		*kq_td;
92 	int			kq_cpu;
93 	int			kq_exit;
94 	int			kq_running;
95 	kspin_lock		kq_lock;
96 	nt_kevent		kq_proc;
97 	nt_kevent		kq_done;
98 };
99 
100 typedef struct kdpc_queue kdpc_queue;
101 
102 struct wb_ext {
103 	struct cv		we_cv;
104 	struct thread		*we_td;
105 };
106 
107 typedef struct wb_ext wb_ext;
108 
109 #define NTOSKRNL_TIMEOUTS	256
110 #ifdef NTOSKRNL_DEBUG_TIMERS
111 static uint64_t ntoskrnl_timer_fires;
112 static uint64_t ntoskrnl_timer_sets;
113 static uint64_t ntoskrnl_timer_reloads;
114 static uint64_t ntoskrnl_timer_cancels;
115 #endif
116 
117 struct callout_entry {
118 	struct callout		ce_callout;
119 	list_entry		ce_list;
120 };
121 
122 typedef struct callout_entry callout_entry;
123 
124 static struct list_entry ntoskrnl_calllist;
125 static struct mtx ntoskrnl_calllock;
126 struct kuser_shared_data kuser_shared_data;
127 
128 static struct list_entry ntoskrnl_intlist;
129 static kspin_lock ntoskrnl_intlock;
130 
131 static uint8_t RtlEqualUnicodeString(unicode_string *,
132 	unicode_string *, uint8_t);
133 static void RtlCopyString(ansi_string *, const ansi_string *);
134 static void RtlCopyUnicodeString(unicode_string *,
135 	unicode_string *);
136 static irp *IoBuildSynchronousFsdRequest(uint32_t, device_object *,
137 	 void *, uint32_t, uint64_t *, nt_kevent *, io_status_block *);
138 static irp *IoBuildAsynchronousFsdRequest(uint32_t,
139 	device_object *, void *, uint32_t, uint64_t *, io_status_block *);
140 static irp *IoBuildDeviceIoControlRequest(uint32_t,
141 	device_object *, void *, uint32_t, void *, uint32_t,
142 	uint8_t, nt_kevent *, io_status_block *);
143 static irp *IoAllocateIrp(uint8_t, uint8_t);
144 static void IoReuseIrp(irp *, uint32_t);
145 static void IoFreeIrp(irp *);
146 static void IoInitializeIrp(irp *, uint16_t, uint8_t);
147 static irp *IoMakeAssociatedIrp(irp *, uint8_t);
148 static uint32_t KeWaitForMultipleObjects(uint32_t,
149 	nt_dispatch_header **, uint32_t, uint32_t, uint32_t, uint8_t,
150 	int64_t *, wait_block *);
151 static void ntoskrnl_waittest(nt_dispatch_header *, uint32_t);
152 static void ntoskrnl_satisfy_wait(nt_dispatch_header *, struct thread *);
153 static void ntoskrnl_satisfy_multiple_waits(wait_block *);
154 static int ntoskrnl_is_signalled(nt_dispatch_header *, struct thread *);
155 static void ntoskrnl_insert_timer(ktimer *, int);
156 static void ntoskrnl_remove_timer(ktimer *);
157 #ifdef NTOSKRNL_DEBUG_TIMERS
158 static void ntoskrnl_show_timers(void);
159 #endif
160 static void ntoskrnl_timercall(void *);
161 static void ntoskrnl_dpc_thread(void *);
162 static void ntoskrnl_destroy_dpc_threads(void);
163 static void ntoskrnl_destroy_workitem_threads(void);
164 static void ntoskrnl_workitem_thread(void *);
165 static void ntoskrnl_workitem(device_object *, void *);
166 static void ntoskrnl_unicode_to_ascii(uint16_t *, char *, int);
167 static void ntoskrnl_ascii_to_unicode(char *, uint16_t *, int);
168 static uint8_t ntoskrnl_insert_dpc(list_entry *, kdpc *);
169 static void WRITE_REGISTER_USHORT(uint16_t *, uint16_t);
170 static uint16_t READ_REGISTER_USHORT(uint16_t *);
171 static void WRITE_REGISTER_ULONG(uint32_t *, uint32_t);
172 static uint32_t READ_REGISTER_ULONG(uint32_t *);
173 static void WRITE_REGISTER_UCHAR(uint8_t *, uint8_t);
174 static uint8_t READ_REGISTER_UCHAR(uint8_t *);
175 static int64_t _allmul(int64_t, int64_t);
176 static int64_t _alldiv(int64_t, int64_t);
177 static int64_t _allrem(int64_t, int64_t);
178 static int64_t _allshr(int64_t, uint8_t);
179 static int64_t _allshl(int64_t, uint8_t);
180 static uint64_t _aullmul(uint64_t, uint64_t);
181 static uint64_t _aulldiv(uint64_t, uint64_t);
182 static uint64_t _aullrem(uint64_t, uint64_t);
183 static uint64_t _aullshr(uint64_t, uint8_t);
184 static uint64_t _aullshl(uint64_t, uint8_t);
185 static slist_entry *ntoskrnl_pushsl(slist_header *, slist_entry *);
186 static void InitializeSListHead(slist_header *);
187 static slist_entry *ntoskrnl_popsl(slist_header *);
188 static void ExFreePoolWithTag(void *, uint32_t);
189 static void ExInitializePagedLookasideList(paged_lookaside_list *,
190 	lookaside_alloc_func *, lookaside_free_func *,
191 	uint32_t, size_t, uint32_t, uint16_t);
192 static void ExDeletePagedLookasideList(paged_lookaside_list *);
193 static void ExInitializeNPagedLookasideList(npaged_lookaside_list *,
194 	lookaside_alloc_func *, lookaside_free_func *,
195 	uint32_t, size_t, uint32_t, uint16_t);
196 static void ExDeleteNPagedLookasideList(npaged_lookaside_list *);
197 static slist_entry
198 	*ExInterlockedPushEntrySList(slist_header *,
199 	slist_entry *, kspin_lock *);
200 static slist_entry
201 	*ExInterlockedPopEntrySList(slist_header *, kspin_lock *);
202 static uint32_t InterlockedIncrement(volatile uint32_t *);
203 static uint32_t InterlockedDecrement(volatile uint32_t *);
204 static void ExInterlockedAddLargeStatistic(uint64_t *, uint32_t);
205 static void *MmAllocateContiguousMemory(uint32_t, uint64_t);
206 static void *MmAllocateContiguousMemorySpecifyCache(uint32_t,
207 	uint64_t, uint64_t, uint64_t, enum nt_caching_type);
208 static void MmFreeContiguousMemory(void *);
209 static void MmFreeContiguousMemorySpecifyCache(void *, uint32_t,
210 	enum nt_caching_type);
211 static uint32_t MmSizeOfMdl(void *, size_t);
212 static void *MmMapLockedPages(mdl *, uint8_t);
213 static void *MmMapLockedPagesSpecifyCache(mdl *,
214 	uint8_t, uint32_t, void *, uint32_t, uint32_t);
215 static void MmUnmapLockedPages(void *, mdl *);
216 static device_t ntoskrnl_finddev(device_t, uint64_t, struct resource **);
217 static void RtlZeroMemory(void *, size_t);
218 static void RtlSecureZeroMemory(void *, size_t);
219 static void RtlFillMemory(void *, size_t, uint8_t);
220 static void RtlMoveMemory(void *, const void *, size_t);
221 static ndis_status RtlCharToInteger(const char *, uint32_t, uint32_t *);
222 static void RtlCopyMemory(void *, const void *, size_t);
223 static size_t RtlCompareMemory(const void *, const void *, size_t);
224 static ndis_status RtlUnicodeStringToInteger(unicode_string *,
225 	uint32_t, uint32_t *);
226 static int atoi (const char *);
227 static long atol (const char *);
228 static int rand(void);
229 static void srand(unsigned int);
230 static void KeQuerySystemTime(uint64_t *);
231 static uint32_t KeTickCount(void);
232 static uint8_t IoIsWdmVersionAvailable(uint8_t, uint8_t);
233 static int32_t IoOpenDeviceRegistryKey(struct device_object *, uint32_t,
234     uint32_t, void **);
235 static void ntoskrnl_thrfunc(void *);
236 static ndis_status PsCreateSystemThread(ndis_handle *,
237 	uint32_t, void *, ndis_handle, void *, void *, void *);
238 static ndis_status PsTerminateSystemThread(ndis_status);
239 static ndis_status IoGetDeviceObjectPointer(unicode_string *,
240 	uint32_t, void *, device_object *);
241 static ndis_status IoGetDeviceProperty(device_object *, uint32_t,
242 	uint32_t, void *, uint32_t *);
243 static void KeInitializeMutex(kmutant *, uint32_t);
244 static uint32_t KeReleaseMutex(kmutant *, uint8_t);
245 static uint32_t KeReadStateMutex(kmutant *);
246 static ndis_status ObReferenceObjectByHandle(ndis_handle,
247 	uint32_t, void *, uint8_t, void **, void **);
248 static void ObfDereferenceObject(void *);
249 static uint32_t ZwClose(ndis_handle);
250 static uint32_t WmiQueryTraceInformation(uint32_t, void *, uint32_t,
251 	uint32_t, void *);
252 static uint32_t WmiTraceMessage(uint64_t, uint32_t, void *, uint16_t, ...);
253 static uint32_t IoWMIRegistrationControl(device_object *, uint32_t);
254 static void *ntoskrnl_memset(void *, int, size_t);
255 static void *ntoskrnl_memmove(void *, void *, size_t);
256 static void *ntoskrnl_memchr(void *, unsigned char, size_t);
257 static char *ntoskrnl_strstr(char *, char *);
258 static char *ntoskrnl_strncat(char *, char *, size_t);
259 static int ntoskrnl_toupper(int);
260 static int ntoskrnl_tolower(int);
261 static funcptr ntoskrnl_findwrap(funcptr);
262 static uint32_t DbgPrint(char *, ...);
263 static void DbgBreakPoint(void);
264 static void KeBugCheckEx(uint32_t, u_long, u_long, u_long, u_long);
265 static int32_t KeDelayExecutionThread(uint8_t, uint8_t, int64_t *);
266 static int32_t KeSetPriorityThread(struct thread *, int32_t);
267 static void dummy(void);
268 
269 static struct mtx ntoskrnl_dispatchlock;
270 static struct mtx ntoskrnl_interlock;
271 static kspin_lock ntoskrnl_cancellock;
272 static int ntoskrnl_kth = 0;
273 static struct nt_objref_head ntoskrnl_reflist;
274 static uma_zone_t mdl_zone;
275 static uma_zone_t iw_zone;
276 static struct kdpc_queue *kq_queues;
277 static struct kdpc_queue *wq_queues;
278 static int wq_idx = 0;
279 
280 int
ntoskrnl_libinit()281 ntoskrnl_libinit()
282 {
283 	image_patch_table	*patch;
284 	int			error;
285 	struct proc		*p;
286 	kdpc_queue		*kq;
287 	callout_entry		*e;
288 	int			i;
289 
290 	mtx_init(&ntoskrnl_dispatchlock,
291 	    "ntoskrnl dispatch lock", MTX_NDIS_LOCK, MTX_DEF|MTX_RECURSE);
292 	mtx_init(&ntoskrnl_interlock, MTX_NTOSKRNL_SPIN_LOCK, NULL, MTX_SPIN);
293 	KeInitializeSpinLock(&ntoskrnl_cancellock);
294 	KeInitializeSpinLock(&ntoskrnl_intlock);
295 	TAILQ_INIT(&ntoskrnl_reflist);
296 
297 	InitializeListHead(&ntoskrnl_calllist);
298 	InitializeListHead(&ntoskrnl_intlist);
299 	mtx_init(&ntoskrnl_calllock, MTX_NTOSKRNL_SPIN_LOCK, NULL, MTX_SPIN);
300 
301 	kq_queues = ExAllocatePoolWithTag(NonPagedPool,
302 #ifdef NTOSKRNL_MULTIPLE_DPCS
303 	    sizeof(kdpc_queue) * mp_ncpus, 0);
304 #else
305 	    sizeof(kdpc_queue), 0);
306 #endif
307 
308 	if (kq_queues == NULL)
309 		return (ENOMEM);
310 
311 	wq_queues = ExAllocatePoolWithTag(NonPagedPool,
312 	    sizeof(kdpc_queue) * WORKITEM_THREADS, 0);
313 
314 	if (wq_queues == NULL)
315 		return (ENOMEM);
316 
317 #ifdef NTOSKRNL_MULTIPLE_DPCS
318 	bzero((char *)kq_queues, sizeof(kdpc_queue) * mp_ncpus);
319 #else
320 	bzero((char *)kq_queues, sizeof(kdpc_queue));
321 #endif
322 	bzero((char *)wq_queues, sizeof(kdpc_queue) * WORKITEM_THREADS);
323 
324 	/*
325 	 * Launch the DPC threads.
326 	 */
327 
328 #ifdef NTOSKRNL_MULTIPLE_DPCS
329 	for (i = 0; i < mp_ncpus; i++) {
330 #else
331 	for (i = 0; i < 1; i++) {
332 #endif
333 		kq = kq_queues + i;
334 		kq->kq_cpu = i;
335 		error = kproc_create(ntoskrnl_dpc_thread, kq, &p,
336 		    RFHIGHPID, NDIS_KSTACK_PAGES, "Windows DPC %d", i);
337 		if (error)
338 			panic("failed to launch DPC thread");
339 	}
340 
341 	/*
342 	 * Launch the workitem threads.
343 	 */
344 
345 	for (i = 0; i < WORKITEM_THREADS; i++) {
346 		kq = wq_queues + i;
347 		error = kproc_create(ntoskrnl_workitem_thread, kq, &p,
348 		    RFHIGHPID, NDIS_KSTACK_PAGES, "Windows Workitem %d", i);
349 		if (error)
350 			panic("failed to launch workitem thread");
351 	}
352 
353 	patch = ntoskrnl_functbl;
354 	while (patch->ipt_func != NULL) {
355 		windrv_wrap((funcptr)patch->ipt_func,
356 		    (funcptr *)&patch->ipt_wrap,
357 		    patch->ipt_argcnt, patch->ipt_ftype);
358 		patch++;
359 	}
360 
361 	for (i = 0; i < NTOSKRNL_TIMEOUTS; i++) {
362 		e = ExAllocatePoolWithTag(NonPagedPool,
363 		    sizeof(callout_entry), 0);
364 		if (e == NULL)
365 			panic("failed to allocate timeouts");
366 		mtx_lock_spin(&ntoskrnl_calllock);
367 		InsertHeadList((&ntoskrnl_calllist), (&e->ce_list));
368 		mtx_unlock_spin(&ntoskrnl_calllock);
369 	}
370 
371 	/*
372 	 * MDLs are supposed to be variable size (they describe
373 	 * buffers containing some number of pages, but we don't
374 	 * know ahead of time how many pages that will be). But
375 	 * always allocating them off the heap is very slow. As
376 	 * a compromise, we create an MDL UMA zone big enough to
377 	 * handle any buffer requiring up to 16 pages, and we
378 	 * use those for any MDLs for buffers of 16 pages or less
379 	 * in size. For buffers larger than that (which we assume
380 	 * will be few and far between, we allocate the MDLs off
381 	 * the heap.
382 	 */
383 
384 	mdl_zone = uma_zcreate("Windows MDL", MDL_ZONE_SIZE,
385 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
386 
387 	iw_zone = uma_zcreate("Windows WorkItem", sizeof(io_workitem),
388 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
389 
390 	return (0);
391 }
392 
393 int
394 ntoskrnl_libfini()
395 {
396 	image_patch_table	*patch;
397 	callout_entry		*e;
398 	list_entry		*l;
399 
400 	patch = ntoskrnl_functbl;
401 	while (patch->ipt_func != NULL) {
402 		windrv_unwrap(patch->ipt_wrap);
403 		patch++;
404 	}
405 
406 	/* Stop the workitem queues. */
407 	ntoskrnl_destroy_workitem_threads();
408 	/* Stop the DPC queues. */
409 	ntoskrnl_destroy_dpc_threads();
410 
411 	ExFreePool(kq_queues);
412 	ExFreePool(wq_queues);
413 
414 	uma_zdestroy(mdl_zone);
415 	uma_zdestroy(iw_zone);
416 
417 	mtx_lock_spin(&ntoskrnl_calllock);
418 	while(!IsListEmpty(&ntoskrnl_calllist)) {
419 		l = RemoveHeadList(&ntoskrnl_calllist);
420 		e = CONTAINING_RECORD(l, callout_entry, ce_list);
421 		mtx_unlock_spin(&ntoskrnl_calllock);
422 		ExFreePool(e);
423 		mtx_lock_spin(&ntoskrnl_calllock);
424 	}
425 	mtx_unlock_spin(&ntoskrnl_calllock);
426 
427 	mtx_destroy(&ntoskrnl_dispatchlock);
428 	mtx_destroy(&ntoskrnl_interlock);
429 	mtx_destroy(&ntoskrnl_calllock);
430 
431 	return (0);
432 }
433 
434 /*
435  * We need to be able to reference this externally from the wrapper;
436  * GCC only generates a local implementation of memset.
437  */
438 static void *
439 ntoskrnl_memset(buf, ch, size)
440 	void			*buf;
441 	int			ch;
442 	size_t			size;
443 {
444 	return (memset(buf, ch, size));
445 }
446 
447 static void *
448 ntoskrnl_memmove(dst, src, size)
449 	void			*src;
450 	void			*dst;
451 	size_t			size;
452 {
453 	bcopy(src, dst, size);
454 	return (dst);
455 }
456 
457 static void *
458 ntoskrnl_memchr(void *buf, unsigned char ch, size_t len)
459 {
460 	if (len != 0) {
461 		unsigned char *p = buf;
462 
463 		do {
464 			if (*p++ == ch)
465 				return (p - 1);
466 		} while (--len != 0);
467 	}
468 	return (NULL);
469 }
470 
471 static char *
472 ntoskrnl_strstr(s, find)
473 	char *s, *find;
474 {
475 	char c, sc;
476 	size_t len;
477 
478 	if ((c = *find++) != 0) {
479 		len = strlen(find);
480 		do {
481 			do {
482 				if ((sc = *s++) == 0)
483 					return (NULL);
484 			} while (sc != c);
485 		} while (strncmp(s, find, len) != 0);
486 		s--;
487 	}
488 	return ((char *)s);
489 }
490 
491 /* Taken from libc */
492 static char *
493 ntoskrnl_strncat(dst, src, n)
494 	char		*dst;
495 	char		*src;
496 	size_t		n;
497 {
498 	if (n != 0) {
499 		char *d = dst;
500 		const char *s = src;
501 
502 		while (*d != 0)
503 			d++;
504 		do {
505 			if ((*d = *s++) == 0)
506 				break;
507 			d++;
508 		} while (--n != 0);
509 		*d = 0;
510 	}
511 	return (dst);
512 }
513 
514 static int
515 ntoskrnl_toupper(c)
516 	int			c;
517 {
518 	return (toupper(c));
519 }
520 
521 static int
522 ntoskrnl_tolower(c)
523 	int			c;
524 {
525 	return (tolower(c));
526 }
527 
528 static uint8_t
529 RtlEqualUnicodeString(unicode_string *str1, unicode_string *str2,
530 	uint8_t caseinsensitive)
531 {
532 	int			i;
533 
534 	if (str1->us_len != str2->us_len)
535 		return (FALSE);
536 
537 	for (i = 0; i < str1->us_len; i++) {
538 		if (caseinsensitive == TRUE) {
539 			if (toupper((char)(str1->us_buf[i] & 0xFF)) !=
540 			    toupper((char)(str2->us_buf[i] & 0xFF)))
541 				return (FALSE);
542 		} else {
543 			if (str1->us_buf[i] != str2->us_buf[i])
544 				return (FALSE);
545 		}
546 	}
547 
548 	return (TRUE);
549 }
550 
551 static void
552 RtlCopyString(dst, src)
553 	ansi_string		*dst;
554 	const ansi_string	*src;
555 {
556 	if (src != NULL && src->as_buf != NULL && dst->as_buf != NULL) {
557 		dst->as_len = min(src->as_len, dst->as_maxlen);
558 		memcpy(dst->as_buf, src->as_buf, dst->as_len);
559 		if (dst->as_len < dst->as_maxlen)
560 			dst->as_buf[dst->as_len] = 0;
561 	} else
562 		dst->as_len = 0;
563 }
564 
565 static void
566 RtlCopyUnicodeString(dest, src)
567 	unicode_string		*dest;
568 	unicode_string		*src;
569 {
570 
571 	if (dest->us_maxlen >= src->us_len)
572 		dest->us_len = src->us_len;
573 	else
574 		dest->us_len = dest->us_maxlen;
575 	memcpy(dest->us_buf, src->us_buf, dest->us_len);
576 }
577 
578 static void
579 ntoskrnl_ascii_to_unicode(ascii, unicode, len)
580 	char			*ascii;
581 	uint16_t		*unicode;
582 	int			len;
583 {
584 	int			i;
585 	uint16_t		*ustr;
586 
587 	ustr = unicode;
588 	for (i = 0; i < len; i++) {
589 		*ustr = (uint16_t)ascii[i];
590 		ustr++;
591 	}
592 }
593 
594 static void
595 ntoskrnl_unicode_to_ascii(unicode, ascii, len)
596 	uint16_t		*unicode;
597 	char			*ascii;
598 	int			len;
599 {
600 	int			i;
601 	uint8_t			*astr;
602 
603 	astr = ascii;
604 	for (i = 0; i < len / 2; i++) {
605 		*astr = (uint8_t)unicode[i];
606 		astr++;
607 	}
608 }
609 
610 uint32_t
611 RtlUnicodeStringToAnsiString(ansi_string *dest, unicode_string *src, uint8_t allocate)
612 {
613 	if (dest == NULL || src == NULL)
614 		return (STATUS_INVALID_PARAMETER);
615 
616 	dest->as_len = src->us_len / 2;
617 	if (dest->as_maxlen < dest->as_len)
618 		dest->as_len = dest->as_maxlen;
619 
620 	if (allocate == TRUE) {
621 		dest->as_buf = ExAllocatePoolWithTag(NonPagedPool,
622 		    (src->us_len / 2) + 1, 0);
623 		if (dest->as_buf == NULL)
624 			return (STATUS_INSUFFICIENT_RESOURCES);
625 		dest->as_len = dest->as_maxlen = src->us_len / 2;
626 	} else {
627 		dest->as_len = src->us_len / 2; /* XXX */
628 		if (dest->as_maxlen < dest->as_len)
629 			dest->as_len = dest->as_maxlen;
630 	}
631 
632 	ntoskrnl_unicode_to_ascii(src->us_buf, dest->as_buf,
633 	    dest->as_len * 2);
634 
635 	return (STATUS_SUCCESS);
636 }
637 
638 uint32_t
639 RtlAnsiStringToUnicodeString(unicode_string *dest, ansi_string *src,
640 	uint8_t allocate)
641 {
642 	if (dest == NULL || src == NULL)
643 		return (STATUS_INVALID_PARAMETER);
644 
645 	if (allocate == TRUE) {
646 		dest->us_buf = ExAllocatePoolWithTag(NonPagedPool,
647 		    src->as_len * 2, 0);
648 		if (dest->us_buf == NULL)
649 			return (STATUS_INSUFFICIENT_RESOURCES);
650 		dest->us_len = dest->us_maxlen = strlen(src->as_buf) * 2;
651 	} else {
652 		dest->us_len = src->as_len * 2; /* XXX */
653 		if (dest->us_maxlen < dest->us_len)
654 			dest->us_len = dest->us_maxlen;
655 	}
656 
657 	ntoskrnl_ascii_to_unicode(src->as_buf, dest->us_buf,
658 	    dest->us_len / 2);
659 
660 	return (STATUS_SUCCESS);
661 }
662 
663 void *
664 ExAllocatePoolWithTag(pooltype, len, tag)
665 	uint32_t		pooltype;
666 	size_t			len;
667 	uint32_t		tag;
668 {
669 	void			*buf;
670 
671 	buf = malloc(len, M_DEVBUF, M_NOWAIT|M_ZERO);
672 	if (buf == NULL)
673 		return (NULL);
674 
675 	return (buf);
676 }
677 
678 static void
679 ExFreePoolWithTag(buf, tag)
680 	void		*buf;
681 	uint32_t	tag;
682 {
683 	ExFreePool(buf);
684 }
685 
686 void
687 ExFreePool(buf)
688 	void			*buf;
689 {
690 	free(buf, M_DEVBUF);
691 }
692 
693 uint32_t
694 IoAllocateDriverObjectExtension(drv, clid, extlen, ext)
695 	driver_object		*drv;
696 	void			*clid;
697 	uint32_t		extlen;
698 	void			**ext;
699 {
700 	custom_extension	*ce;
701 
702 	ce = ExAllocatePoolWithTag(NonPagedPool, sizeof(custom_extension)
703 	    + extlen, 0);
704 
705 	if (ce == NULL)
706 		return (STATUS_INSUFFICIENT_RESOURCES);
707 
708 	ce->ce_clid = clid;
709 	InsertTailList((&drv->dro_driverext->dre_usrext), (&ce->ce_list));
710 
711 	*ext = (void *)(ce + 1);
712 
713 	return (STATUS_SUCCESS);
714 }
715 
716 void *
717 IoGetDriverObjectExtension(drv, clid)
718 	driver_object		*drv;
719 	void			*clid;
720 {
721 	list_entry		*e;
722 	custom_extension	*ce;
723 
724 	/*
725 	 * Sanity check. Our dummy bus drivers don't have
726 	 * any driver extensions.
727 	 */
728 
729 	if (drv->dro_driverext == NULL)
730 		return (NULL);
731 
732 	e = drv->dro_driverext->dre_usrext.nle_flink;
733 	while (e != &drv->dro_driverext->dre_usrext) {
734 		ce = (custom_extension *)e;
735 		if (ce->ce_clid == clid)
736 			return ((void *)(ce + 1));
737 		e = e->nle_flink;
738 	}
739 
740 	return (NULL);
741 }
742 
743 uint32_t
744 IoCreateDevice(driver_object *drv, uint32_t devextlen, unicode_string *devname,
745 	uint32_t devtype, uint32_t devchars, uint8_t exclusive,
746 	device_object **newdev)
747 {
748 	device_object		*dev;
749 
750 	dev = ExAllocatePoolWithTag(NonPagedPool, sizeof(device_object), 0);
751 	if (dev == NULL)
752 		return (STATUS_INSUFFICIENT_RESOURCES);
753 
754 	dev->do_type = devtype;
755 	dev->do_drvobj = drv;
756 	dev->do_currirp = NULL;
757 	dev->do_flags = 0;
758 
759 	if (devextlen) {
760 		dev->do_devext = ExAllocatePoolWithTag(NonPagedPool,
761 		    devextlen, 0);
762 
763 		if (dev->do_devext == NULL) {
764 			ExFreePool(dev);
765 			return (STATUS_INSUFFICIENT_RESOURCES);
766 		}
767 
768 		bzero(dev->do_devext, devextlen);
769 	} else
770 		dev->do_devext = NULL;
771 
772 	dev->do_size = sizeof(device_object) + devextlen;
773 	dev->do_refcnt = 1;
774 	dev->do_attacheddev = NULL;
775 	dev->do_nextdev = NULL;
776 	dev->do_devtype = devtype;
777 	dev->do_stacksize = 1;
778 	dev->do_alignreq = 1;
779 	dev->do_characteristics = devchars;
780 	dev->do_iotimer = NULL;
781 	KeInitializeEvent(&dev->do_devlock, EVENT_TYPE_SYNC, TRUE);
782 
783 	/*
784 	 * Vpd is used for disk/tape devices,
785 	 * but we don't support those. (Yet.)
786 	 */
787 	dev->do_vpb = NULL;
788 
789 	dev->do_devobj_ext = ExAllocatePoolWithTag(NonPagedPool,
790 	    sizeof(devobj_extension), 0);
791 
792 	if (dev->do_devobj_ext == NULL) {
793 		if (dev->do_devext != NULL)
794 			ExFreePool(dev->do_devext);
795 		ExFreePool(dev);
796 		return (STATUS_INSUFFICIENT_RESOURCES);
797 	}
798 
799 	dev->do_devobj_ext->dve_type = 0;
800 	dev->do_devobj_ext->dve_size = sizeof(devobj_extension);
801 	dev->do_devobj_ext->dve_devobj = dev;
802 
803 	/*
804 	 * Attach this device to the driver object's list
805 	 * of devices. Note: this is not the same as attaching
806 	 * the device to the device stack. The driver's AddDevice
807 	 * routine must explicitly call IoAddDeviceToDeviceStack()
808 	 * to do that.
809 	 */
810 
811 	if (drv->dro_devobj == NULL) {
812 		drv->dro_devobj = dev;
813 		dev->do_nextdev = NULL;
814 	} else {
815 		dev->do_nextdev = drv->dro_devobj;
816 		drv->dro_devobj = dev;
817 	}
818 
819 	*newdev = dev;
820 
821 	return (STATUS_SUCCESS);
822 }
823 
824 void
825 IoDeleteDevice(dev)
826 	device_object		*dev;
827 {
828 	device_object		*prev;
829 
830 	if (dev == NULL)
831 		return;
832 
833 	if (dev->do_devobj_ext != NULL)
834 		ExFreePool(dev->do_devobj_ext);
835 
836 	if (dev->do_devext != NULL)
837 		ExFreePool(dev->do_devext);
838 
839 	/* Unlink the device from the driver's device list. */
840 
841 	prev = dev->do_drvobj->dro_devobj;
842 	if (prev == dev)
843 		dev->do_drvobj->dro_devobj = dev->do_nextdev;
844 	else {
845 		while (prev->do_nextdev != dev)
846 			prev = prev->do_nextdev;
847 		prev->do_nextdev = dev->do_nextdev;
848 	}
849 
850 	ExFreePool(dev);
851 }
852 
853 device_object *
854 IoGetAttachedDevice(dev)
855 	device_object		*dev;
856 {
857 	device_object		*d;
858 
859 	if (dev == NULL)
860 		return (NULL);
861 
862 	d = dev;
863 
864 	while (d->do_attacheddev != NULL)
865 		d = d->do_attacheddev;
866 
867 	return (d);
868 }
869 
870 static irp *
871 IoBuildSynchronousFsdRequest(func, dobj, buf, len, off, event, status)
872 	uint32_t		func;
873 	device_object		*dobj;
874 	void			*buf;
875 	uint32_t		len;
876 	uint64_t		*off;
877 	nt_kevent		*event;
878 	io_status_block		*status;
879 {
880 	irp			*ip;
881 
882 	ip = IoBuildAsynchronousFsdRequest(func, dobj, buf, len, off, status);
883 	if (ip == NULL)
884 		return (NULL);
885 	ip->irp_usrevent = event;
886 
887 	return (ip);
888 }
889 
890 static irp *
891 IoBuildAsynchronousFsdRequest(func, dobj, buf, len, off, status)
892 	uint32_t		func;
893 	device_object		*dobj;
894 	void			*buf;
895 	uint32_t		len;
896 	uint64_t		*off;
897 	io_status_block		*status;
898 {
899 	irp			*ip;
900 	io_stack_location	*sl;
901 
902 	ip = IoAllocateIrp(dobj->do_stacksize, TRUE);
903 	if (ip == NULL)
904 		return (NULL);
905 
906 	ip->irp_usriostat = status;
907 	ip->irp_tail.irp_overlay.irp_thread = NULL;
908 
909 	sl = IoGetNextIrpStackLocation(ip);
910 	sl->isl_major = func;
911 	sl->isl_minor = 0;
912 	sl->isl_flags = 0;
913 	sl->isl_ctl = 0;
914 	sl->isl_devobj = dobj;
915 	sl->isl_fileobj = NULL;
916 	sl->isl_completionfunc = NULL;
917 
918 	ip->irp_userbuf = buf;
919 
920 	if (dobj->do_flags & DO_BUFFERED_IO) {
921 		ip->irp_assoc.irp_sysbuf =
922 		    ExAllocatePoolWithTag(NonPagedPool, len, 0);
923 		if (ip->irp_assoc.irp_sysbuf == NULL) {
924 			IoFreeIrp(ip);
925 			return (NULL);
926 		}
927 		bcopy(buf, ip->irp_assoc.irp_sysbuf, len);
928 	}
929 
930 	if (dobj->do_flags & DO_DIRECT_IO) {
931 		ip->irp_mdl = IoAllocateMdl(buf, len, FALSE, FALSE, ip);
932 		if (ip->irp_mdl == NULL) {
933 			if (ip->irp_assoc.irp_sysbuf != NULL)
934 				ExFreePool(ip->irp_assoc.irp_sysbuf);
935 			IoFreeIrp(ip);
936 			return (NULL);
937 		}
938 		ip->irp_userbuf = NULL;
939 		ip->irp_assoc.irp_sysbuf = NULL;
940 	}
941 
942 	if (func == IRP_MJ_READ) {
943 		sl->isl_parameters.isl_read.isl_len = len;
944 		if (off != NULL)
945 			sl->isl_parameters.isl_read.isl_byteoff = *off;
946 		else
947 			sl->isl_parameters.isl_read.isl_byteoff = 0;
948 	}
949 
950 	if (func == IRP_MJ_WRITE) {
951 		sl->isl_parameters.isl_write.isl_len = len;
952 		if (off != NULL)
953 			sl->isl_parameters.isl_write.isl_byteoff = *off;
954 		else
955 			sl->isl_parameters.isl_write.isl_byteoff = 0;
956 	}
957 
958 	return (ip);
959 }
960 
961 static irp *
962 IoBuildDeviceIoControlRequest(uint32_t iocode, device_object *dobj, void *ibuf,
963 	uint32_t ilen, void *obuf, uint32_t olen, uint8_t isinternal,
964 	nt_kevent *event, io_status_block *status)
965 {
966 	irp			*ip;
967 	io_stack_location	*sl;
968 	uint32_t		buflen;
969 
970 	ip = IoAllocateIrp(dobj->do_stacksize, TRUE);
971 	if (ip == NULL)
972 		return (NULL);
973 	ip->irp_usrevent = event;
974 	ip->irp_usriostat = status;
975 	ip->irp_tail.irp_overlay.irp_thread = NULL;
976 
977 	sl = IoGetNextIrpStackLocation(ip);
978 	sl->isl_major = isinternal == TRUE ?
979 	    IRP_MJ_INTERNAL_DEVICE_CONTROL : IRP_MJ_DEVICE_CONTROL;
980 	sl->isl_minor = 0;
981 	sl->isl_flags = 0;
982 	sl->isl_ctl = 0;
983 	sl->isl_devobj = dobj;
984 	sl->isl_fileobj = NULL;
985 	sl->isl_completionfunc = NULL;
986 	sl->isl_parameters.isl_ioctl.isl_iocode = iocode;
987 	sl->isl_parameters.isl_ioctl.isl_ibuflen = ilen;
988 	sl->isl_parameters.isl_ioctl.isl_obuflen = olen;
989 
990 	switch(IO_METHOD(iocode)) {
991 	case METHOD_BUFFERED:
992 		if (ilen > olen)
993 			buflen = ilen;
994 		else
995 			buflen = olen;
996 		if (buflen) {
997 			ip->irp_assoc.irp_sysbuf =
998 			    ExAllocatePoolWithTag(NonPagedPool, buflen, 0);
999 			if (ip->irp_assoc.irp_sysbuf == NULL) {
1000 				IoFreeIrp(ip);
1001 				return (NULL);
1002 			}
1003 		}
1004 		if (ilen && ibuf != NULL) {
1005 			bcopy(ibuf, ip->irp_assoc.irp_sysbuf, ilen);
1006 			bzero((char *)ip->irp_assoc.irp_sysbuf + ilen,
1007 			    buflen - ilen);
1008 		} else
1009 			bzero(ip->irp_assoc.irp_sysbuf, ilen);
1010 		ip->irp_userbuf = obuf;
1011 		break;
1012 	case METHOD_IN_DIRECT:
1013 	case METHOD_OUT_DIRECT:
1014 		if (ilen && ibuf != NULL) {
1015 			ip->irp_assoc.irp_sysbuf =
1016 			    ExAllocatePoolWithTag(NonPagedPool, ilen, 0);
1017 			if (ip->irp_assoc.irp_sysbuf == NULL) {
1018 				IoFreeIrp(ip);
1019 				return (NULL);
1020 			}
1021 			bcopy(ibuf, ip->irp_assoc.irp_sysbuf, ilen);
1022 		}
1023 		if (olen && obuf != NULL) {
1024 			ip->irp_mdl = IoAllocateMdl(obuf, olen,
1025 			    FALSE, FALSE, ip);
1026 			/*
1027 			 * Normally we would MmProbeAndLockPages()
1028 			 * here, but we don't have to in our
1029 			 * imlementation.
1030 			 */
1031 		}
1032 		break;
1033 	case METHOD_NEITHER:
1034 		ip->irp_userbuf = obuf;
1035 		sl->isl_parameters.isl_ioctl.isl_type3ibuf = ibuf;
1036 		break;
1037 	default:
1038 		break;
1039 	}
1040 
1041 	/*
1042 	 * Ideally, we should associate this IRP with the calling
1043 	 * thread here.
1044 	 */
1045 
1046 	return (ip);
1047 }
1048 
1049 static irp *
1050 IoAllocateIrp(uint8_t stsize, uint8_t chargequota)
1051 {
1052 	irp			*i;
1053 
1054 	i = ExAllocatePoolWithTag(NonPagedPool, IoSizeOfIrp(stsize), 0);
1055 	if (i == NULL)
1056 		return (NULL);
1057 
1058 	IoInitializeIrp(i, IoSizeOfIrp(stsize), stsize);
1059 
1060 	return (i);
1061 }
1062 
1063 static irp *
1064 IoMakeAssociatedIrp(irp *ip, uint8_t stsize)
1065 {
1066 	irp			*associrp;
1067 
1068 	associrp = IoAllocateIrp(stsize, FALSE);
1069 	if (associrp == NULL)
1070 		return (NULL);
1071 
1072 	mtx_lock(&ntoskrnl_dispatchlock);
1073 	associrp->irp_flags |= IRP_ASSOCIATED_IRP;
1074 	associrp->irp_tail.irp_overlay.irp_thread =
1075 	    ip->irp_tail.irp_overlay.irp_thread;
1076 	associrp->irp_assoc.irp_master = ip;
1077 	mtx_unlock(&ntoskrnl_dispatchlock);
1078 
1079 	return (associrp);
1080 }
1081 
1082 static void
1083 IoFreeIrp(ip)
1084 	irp			*ip;
1085 {
1086 	ExFreePool(ip);
1087 }
1088 
1089 static void
1090 IoInitializeIrp(irp *io, uint16_t psize, uint8_t ssize)
1091 {
1092 	bzero((char *)io, IoSizeOfIrp(ssize));
1093 	io->irp_size = psize;
1094 	io->irp_stackcnt = ssize;
1095 	io->irp_currentstackloc = ssize;
1096 	InitializeListHead(&io->irp_thlist);
1097 	io->irp_tail.irp_overlay.irp_csl =
1098 	    (io_stack_location *)(io + 1) + ssize;
1099 }
1100 
1101 static void
1102 IoReuseIrp(ip, status)
1103 	irp			*ip;
1104 	uint32_t		status;
1105 {
1106 	uint8_t			allocflags;
1107 
1108 	allocflags = ip->irp_allocflags;
1109 	IoInitializeIrp(ip, ip->irp_size, ip->irp_stackcnt);
1110 	ip->irp_iostat.isb_status = status;
1111 	ip->irp_allocflags = allocflags;
1112 }
1113 
1114 void
1115 IoAcquireCancelSpinLock(uint8_t *irql)
1116 {
1117 	KeAcquireSpinLock(&ntoskrnl_cancellock, irql);
1118 }
1119 
1120 void
1121 IoReleaseCancelSpinLock(uint8_t irql)
1122 {
1123 	KeReleaseSpinLock(&ntoskrnl_cancellock, irql);
1124 }
1125 
1126 uint8_t
1127 IoCancelIrp(irp *ip)
1128 {
1129 	cancel_func		cfunc;
1130 	uint8_t			cancelirql;
1131 
1132 	IoAcquireCancelSpinLock(&cancelirql);
1133 	cfunc = IoSetCancelRoutine(ip, NULL);
1134 	ip->irp_cancel = TRUE;
1135 	if (cfunc == NULL) {
1136 		IoReleaseCancelSpinLock(cancelirql);
1137 		return (FALSE);
1138 	}
1139 	ip->irp_cancelirql = cancelirql;
1140 	MSCALL2(cfunc, IoGetCurrentIrpStackLocation(ip)->isl_devobj, ip);
1141 	return (uint8_t)IoSetCancelValue(ip, TRUE);
1142 }
1143 
1144 uint32_t
1145 IofCallDriver(dobj, ip)
1146 	device_object		*dobj;
1147 	irp			*ip;
1148 {
1149 	driver_object		*drvobj;
1150 	io_stack_location	*sl;
1151 	uint32_t		status;
1152 	driver_dispatch		disp;
1153 
1154 	drvobj = dobj->do_drvobj;
1155 
1156 	if (ip->irp_currentstackloc <= 0)
1157 		panic("IoCallDriver(): out of stack locations");
1158 
1159 	IoSetNextIrpStackLocation(ip);
1160 	sl = IoGetCurrentIrpStackLocation(ip);
1161 
1162 	sl->isl_devobj = dobj;
1163 
1164 	disp = drvobj->dro_dispatch[sl->isl_major];
1165 	status = MSCALL2(disp, dobj, ip);
1166 
1167 	return (status);
1168 }
1169 
1170 void
1171 IofCompleteRequest(irp *ip, uint8_t prioboost)
1172 {
1173 	uint32_t		status;
1174 	device_object		*dobj;
1175 	io_stack_location	*sl;
1176 	completion_func		cf;
1177 
1178 	KASSERT(ip->irp_iostat.isb_status != STATUS_PENDING,
1179 	    ("incorrect IRP(%p) status (STATUS_PENDING)", ip));
1180 
1181 	sl = IoGetCurrentIrpStackLocation(ip);
1182 	IoSkipCurrentIrpStackLocation(ip);
1183 
1184 	do {
1185 		if (sl->isl_ctl & SL_PENDING_RETURNED)
1186 			ip->irp_pendingreturned = TRUE;
1187 
1188 		if (ip->irp_currentstackloc != (ip->irp_stackcnt + 1))
1189 			dobj = IoGetCurrentIrpStackLocation(ip)->isl_devobj;
1190 		else
1191 			dobj = NULL;
1192 
1193 		if (sl->isl_completionfunc != NULL &&
1194 		    ((ip->irp_iostat.isb_status == STATUS_SUCCESS &&
1195 		    sl->isl_ctl & SL_INVOKE_ON_SUCCESS) ||
1196 		    (ip->irp_iostat.isb_status != STATUS_SUCCESS &&
1197 		    sl->isl_ctl & SL_INVOKE_ON_ERROR) ||
1198 		    (ip->irp_cancel == TRUE &&
1199 		    sl->isl_ctl & SL_INVOKE_ON_CANCEL))) {
1200 			cf = sl->isl_completionfunc;
1201 			status = MSCALL3(cf, dobj, ip, sl->isl_completionctx);
1202 			if (status == STATUS_MORE_PROCESSING_REQUIRED)
1203 				return;
1204 		} else {
1205 			if ((ip->irp_currentstackloc <= ip->irp_stackcnt) &&
1206 			    (ip->irp_pendingreturned == TRUE))
1207 				IoMarkIrpPending(ip);
1208 		}
1209 
1210 		/* move to the next.  */
1211 		IoSkipCurrentIrpStackLocation(ip);
1212 		sl++;
1213 	} while (ip->irp_currentstackloc <= (ip->irp_stackcnt + 1));
1214 
1215 	if (ip->irp_usriostat != NULL)
1216 		*ip->irp_usriostat = ip->irp_iostat;
1217 	if (ip->irp_usrevent != NULL)
1218 		KeSetEvent(ip->irp_usrevent, prioboost, FALSE);
1219 
1220 	/* Handle any associated IRPs. */
1221 
1222 	if (ip->irp_flags & IRP_ASSOCIATED_IRP) {
1223 		uint32_t		masterirpcnt;
1224 		irp			*masterirp;
1225 		mdl			*m;
1226 
1227 		masterirp = ip->irp_assoc.irp_master;
1228 		masterirpcnt =
1229 		    InterlockedDecrement(&masterirp->irp_assoc.irp_irpcnt);
1230 
1231 		while ((m = ip->irp_mdl) != NULL) {
1232 			ip->irp_mdl = m->mdl_next;
1233 			IoFreeMdl(m);
1234 		}
1235 		IoFreeIrp(ip);
1236 		if (masterirpcnt == 0)
1237 			IoCompleteRequest(masterirp, IO_NO_INCREMENT);
1238 		return;
1239 	}
1240 
1241 	/* With any luck, these conditions will never arise. */
1242 
1243 	if (ip->irp_flags & IRP_PAGING_IO) {
1244 		if (ip->irp_mdl != NULL)
1245 			IoFreeMdl(ip->irp_mdl);
1246 		IoFreeIrp(ip);
1247 	}
1248 }
1249 
1250 void
1251 ntoskrnl_intr(arg)
1252 	void			*arg;
1253 {
1254 	kinterrupt		*iobj;
1255 	uint8_t			irql;
1256 	uint8_t			claimed;
1257 	list_entry		*l;
1258 
1259 	KeAcquireSpinLock(&ntoskrnl_intlock, &irql);
1260 	l = ntoskrnl_intlist.nle_flink;
1261 	while (l != &ntoskrnl_intlist) {
1262 		iobj = CONTAINING_RECORD(l, kinterrupt, ki_list);
1263 		claimed = MSCALL2(iobj->ki_svcfunc, iobj, iobj->ki_svcctx);
1264 		if (claimed == TRUE)
1265 			break;
1266 		l = l->nle_flink;
1267 	}
1268 	KeReleaseSpinLock(&ntoskrnl_intlock, irql);
1269 }
1270 
1271 uint8_t
1272 KeAcquireInterruptSpinLock(iobj)
1273 	kinterrupt		*iobj;
1274 {
1275 	uint8_t			irql;
1276 	KeAcquireSpinLock(&ntoskrnl_intlock, &irql);
1277 	return (irql);
1278 }
1279 
1280 void
1281 KeReleaseInterruptSpinLock(kinterrupt *iobj, uint8_t irql)
1282 {
1283 	KeReleaseSpinLock(&ntoskrnl_intlock, irql);
1284 }
1285 
1286 uint8_t
1287 KeSynchronizeExecution(iobj, syncfunc, syncctx)
1288 	kinterrupt		*iobj;
1289 	void			*syncfunc;
1290 	void			*syncctx;
1291 {
1292 	uint8_t			irql;
1293 
1294 	KeAcquireSpinLock(&ntoskrnl_intlock, &irql);
1295 	MSCALL1(syncfunc, syncctx);
1296 	KeReleaseSpinLock(&ntoskrnl_intlock, irql);
1297 
1298 	return (TRUE);
1299 }
1300 
1301 /*
1302  * IoConnectInterrupt() is passed only the interrupt vector and
1303  * irql that a device wants to use, but no device-specific tag
1304  * of any kind. This conflicts rather badly with FreeBSD's
1305  * bus_setup_intr(), which needs the device_t for the device
1306  * requesting interrupt delivery. In order to bypass this
1307  * inconsistency, we implement a second level of interrupt
1308  * dispatching on top of bus_setup_intr(). All devices use
1309  * ntoskrnl_intr() as their ISR, and any device requesting
1310  * interrupts will be registered with ntoskrnl_intr()'s interrupt
1311  * dispatch list. When an interrupt arrives, we walk the list
1312  * and invoke all the registered ISRs. This effectively makes all
1313  * interrupts shared, but it's the only way to duplicate the
1314  * semantics of IoConnectInterrupt() and IoDisconnectInterrupt() properly.
1315  */
1316 
1317 uint32_t
1318 IoConnectInterrupt(kinterrupt **iobj, void *svcfunc, void *svcctx,
1319 	kspin_lock *lock, uint32_t vector, uint8_t irql, uint8_t syncirql,
1320 	uint8_t imode, uint8_t shared, uint32_t affinity, uint8_t savefloat)
1321 {
1322 	uint8_t			curirql;
1323 
1324 	*iobj = ExAllocatePoolWithTag(NonPagedPool, sizeof(kinterrupt), 0);
1325 	if (*iobj == NULL)
1326 		return (STATUS_INSUFFICIENT_RESOURCES);
1327 
1328 	(*iobj)->ki_svcfunc = svcfunc;
1329 	(*iobj)->ki_svcctx = svcctx;
1330 
1331 	if (lock == NULL) {
1332 		KeInitializeSpinLock(&(*iobj)->ki_lock_priv);
1333 		(*iobj)->ki_lock = &(*iobj)->ki_lock_priv;
1334 	} else
1335 		(*iobj)->ki_lock = lock;
1336 
1337 	KeAcquireSpinLock(&ntoskrnl_intlock, &curirql);
1338 	InsertHeadList((&ntoskrnl_intlist), (&(*iobj)->ki_list));
1339 	KeReleaseSpinLock(&ntoskrnl_intlock, curirql);
1340 
1341 	return (STATUS_SUCCESS);
1342 }
1343 
1344 void
1345 IoDisconnectInterrupt(iobj)
1346 	kinterrupt		*iobj;
1347 {
1348 	uint8_t			irql;
1349 
1350 	if (iobj == NULL)
1351 		return;
1352 
1353 	KeAcquireSpinLock(&ntoskrnl_intlock, &irql);
1354 	RemoveEntryList((&iobj->ki_list));
1355 	KeReleaseSpinLock(&ntoskrnl_intlock, irql);
1356 
1357 	ExFreePool(iobj);
1358 }
1359 
1360 device_object *
1361 IoAttachDeviceToDeviceStack(src, dst)
1362 	device_object		*src;
1363 	device_object		*dst;
1364 {
1365 	device_object		*attached;
1366 
1367 	mtx_lock(&ntoskrnl_dispatchlock);
1368 	attached = IoGetAttachedDevice(dst);
1369 	attached->do_attacheddev = src;
1370 	src->do_attacheddev = NULL;
1371 	src->do_stacksize = attached->do_stacksize + 1;
1372 	mtx_unlock(&ntoskrnl_dispatchlock);
1373 
1374 	return (attached);
1375 }
1376 
1377 void
1378 IoDetachDevice(topdev)
1379 	device_object		*topdev;
1380 {
1381 	device_object		*tail;
1382 
1383 	mtx_lock(&ntoskrnl_dispatchlock);
1384 
1385 	/* First, break the chain. */
1386 	tail = topdev->do_attacheddev;
1387 	if (tail == NULL) {
1388 		mtx_unlock(&ntoskrnl_dispatchlock);
1389 		return;
1390 	}
1391 	topdev->do_attacheddev = tail->do_attacheddev;
1392 	topdev->do_refcnt--;
1393 
1394 	/* Now reduce the stacksize count for the takm_il objects. */
1395 
1396 	tail = topdev->do_attacheddev;
1397 	while (tail != NULL) {
1398 		tail->do_stacksize--;
1399 		tail = tail->do_attacheddev;
1400 	}
1401 
1402 	mtx_unlock(&ntoskrnl_dispatchlock);
1403 }
1404 
1405 /*
1406  * For the most part, an object is considered signalled if
1407  * dh_sigstate == TRUE. The exception is for mutant objects
1408  * (mutexes), where the logic works like this:
1409  *
1410  * - If the thread already owns the object and sigstate is
1411  *   less than or equal to 0, then the object is considered
1412  *   signalled (recursive acquisition).
1413  * - If dh_sigstate == 1, the object is also considered
1414  *   signalled.
1415  */
1416 
1417 static int
1418 ntoskrnl_is_signalled(obj, td)
1419 	nt_dispatch_header	*obj;
1420 	struct thread		*td;
1421 {
1422 	kmutant			*km;
1423 
1424 	if (obj->dh_type == DISP_TYPE_MUTANT) {
1425 		km = (kmutant *)obj;
1426 		if ((obj->dh_sigstate <= 0 && km->km_ownerthread == td) ||
1427 		    obj->dh_sigstate == 1)
1428 			return (TRUE);
1429 		return (FALSE);
1430 	}
1431 
1432 	if (obj->dh_sigstate > 0)
1433 		return (TRUE);
1434 	return (FALSE);
1435 }
1436 
1437 static void
1438 ntoskrnl_satisfy_wait(obj, td)
1439 	nt_dispatch_header	*obj;
1440 	struct thread		*td;
1441 {
1442 	kmutant			*km;
1443 
1444 	switch (obj->dh_type) {
1445 	case DISP_TYPE_MUTANT:
1446 		km = (struct kmutant *)obj;
1447 		obj->dh_sigstate--;
1448 		/*
1449 		 * If sigstate reaches 0, the mutex is now
1450 		 * non-signalled (the new thread owns it).
1451 		 */
1452 		if (obj->dh_sigstate == 0) {
1453 			km->km_ownerthread = td;
1454 			if (km->km_abandoned == TRUE)
1455 				km->km_abandoned = FALSE;
1456 		}
1457 		break;
1458 	/* Synchronization objects get reset to unsignalled. */
1459 	case DISP_TYPE_SYNCHRONIZATION_EVENT:
1460 	case DISP_TYPE_SYNCHRONIZATION_TIMER:
1461 		obj->dh_sigstate = 0;
1462 		break;
1463 	case DISP_TYPE_SEMAPHORE:
1464 		obj->dh_sigstate--;
1465 		break;
1466 	default:
1467 		break;
1468 	}
1469 }
1470 
1471 static void
1472 ntoskrnl_satisfy_multiple_waits(wb)
1473 	wait_block		*wb;
1474 {
1475 	wait_block		*cur;
1476 	struct thread		*td;
1477 
1478 	cur = wb;
1479 	td = wb->wb_kthread;
1480 
1481 	do {
1482 		ntoskrnl_satisfy_wait(wb->wb_object, td);
1483 		cur->wb_awakened = TRUE;
1484 		cur = cur->wb_next;
1485 	} while (cur != wb);
1486 }
1487 
1488 /* Always called with dispatcher lock held. */
1489 static void
1490 ntoskrnl_waittest(obj, increment)
1491 	nt_dispatch_header	*obj;
1492 	uint32_t		increment;
1493 {
1494 	wait_block		*w, *next;
1495 	list_entry		*e;
1496 	struct thread		*td;
1497 	wb_ext			*we;
1498 	int			satisfied;
1499 
1500 	/*
1501 	 * Once an object has been signalled, we walk its list of
1502 	 * wait blocks. If a wait block can be awakened, then satisfy
1503 	 * waits as necessary and wake the thread.
1504 	 *
1505 	 * The rules work like this:
1506 	 *
1507 	 * If a wait block is marked as WAITTYPE_ANY, then
1508 	 * we can satisfy the wait conditions on the current
1509 	 * object and wake the thread right away. Satisfying
1510 	 * the wait also has the effect of breaking us out
1511 	 * of the search loop.
1512 	 *
1513 	 * If the object is marked as WAITTYLE_ALL, then the
1514 	 * wait block will be part of a circularly linked
1515 	 * list of wait blocks belonging to a waiting thread
1516 	 * that's sleeping in KeWaitForMultipleObjects(). In
1517 	 * order to wake the thread, all the objects in the
1518 	 * wait list must be in the signalled state. If they
1519 	 * are, we then satisfy all of them and wake the
1520 	 * thread.
1521 	 *
1522 	 */
1523 
1524 	e = obj->dh_waitlisthead.nle_flink;
1525 
1526 	while (e != &obj->dh_waitlisthead && obj->dh_sigstate > 0) {
1527 		w = CONTAINING_RECORD(e, wait_block, wb_waitlist);
1528 		we = w->wb_ext;
1529 		td = we->we_td;
1530 		satisfied = FALSE;
1531 		if (w->wb_waittype == WAITTYPE_ANY) {
1532 			/*
1533 			 * Thread can be awakened if
1534 			 * any wait is satisfied.
1535 			 */
1536 			ntoskrnl_satisfy_wait(obj, td);
1537 			satisfied = TRUE;
1538 			w->wb_awakened = TRUE;
1539 		} else {
1540 			/*
1541 			 * Thread can only be woken up
1542 			 * if all waits are satisfied.
1543 			 * If the thread is waiting on multiple
1544 			 * objects, they should all be linked
1545 			 * through the wb_next pointers in the
1546 			 * wait blocks.
1547 			 */
1548 			satisfied = TRUE;
1549 			next = w->wb_next;
1550 			while (next != w) {
1551 				if (ntoskrnl_is_signalled(obj, td) == FALSE) {
1552 					satisfied = FALSE;
1553 					break;
1554 				}
1555 				next = next->wb_next;
1556 			}
1557 			ntoskrnl_satisfy_multiple_waits(w);
1558 		}
1559 
1560 		if (satisfied == TRUE)
1561 			cv_broadcastpri(&we->we_cv,
1562 			    (w->wb_oldpri - (increment * 4)) > PRI_MIN_KERN ?
1563 			    w->wb_oldpri - (increment * 4) : PRI_MIN_KERN);
1564 
1565 		e = e->nle_flink;
1566 	}
1567 }
1568 
1569 /*
1570  * Return the number of 100 nanosecond intervals since
1571  * January 1, 1601. (?!?!)
1572  */
1573 void
1574 ntoskrnl_time(tval)
1575 	uint64_t                *tval;
1576 {
1577 	struct timespec		ts;
1578 
1579 	nanotime(&ts);
1580 	*tval = (uint64_t)ts.tv_nsec / 100 + (uint64_t)ts.tv_sec * 10000000 +
1581 	    11644473600 * 10000000; /* 100ns ticks from 1601 to 1970 */
1582 }
1583 
1584 static void
1585 KeQuerySystemTime(current_time)
1586 	uint64_t		*current_time;
1587 {
1588 	ntoskrnl_time(current_time);
1589 }
1590 
1591 static uint32_t
1592 KeTickCount(void)
1593 {
1594 	struct timeval tv;
1595 	getmicrouptime(&tv);
1596 	return tvtohz(&tv);
1597 }
1598 
1599 /*
1600  * KeWaitForSingleObject() is a tricky beast, because it can be used
1601  * with several different object types: semaphores, timers, events,
1602  * mutexes and threads. Semaphores don't appear very often, but the
1603  * other object types are quite common. KeWaitForSingleObject() is
1604  * what's normally used to acquire a mutex, and it can be used to
1605  * wait for a thread termination.
1606  *
1607  * The Windows NDIS API is implemented in terms of Windows kernel
1608  * primitives, and some of the object manipulation is duplicated in
1609  * NDIS. For example, NDIS has timers and events, which are actually
1610  * Windows kevents and ktimers. Now, you're supposed to only use the
1611  * NDIS variants of these objects within the confines of the NDIS API,
1612  * but there are some naughty developers out there who will use
1613  * KeWaitForSingleObject() on NDIS timer and event objects, so we
1614  * have to support that as well. Conseqently, our NDIS timer and event
1615  * code has to be closely tied into our ntoskrnl timer and event code,
1616  * just as it is in Windows.
1617  *
1618  * KeWaitForSingleObject() may do different things for different kinds
1619  * of objects:
1620  *
1621  * - For events, we check if the event has been signalled. If the
1622  *   event is already in the signalled state, we just return immediately,
1623  *   otherwise we wait for it to be set to the signalled state by someone
1624  *   else calling KeSetEvent(). Events can be either synchronization or
1625  *   notification events.
1626  *
1627  * - For timers, if the timer has already fired and the timer is in
1628  *   the signalled state, we just return, otherwise we wait on the
1629  *   timer. Unlike an event, timers get signalled automatically when
1630  *   they expire rather than someone having to trip them manually.
1631  *   Timers initialized with KeInitializeTimer() are always notification
1632  *   events: KeInitializeTimerEx() lets you initialize a timer as
1633  *   either a notification or synchronization event.
1634  *
1635  * - For mutexes, we try to acquire the mutex and if we can't, we wait
1636  *   on the mutex until it's available and then grab it. When a mutex is
1637  *   released, it enters the signalled state, which wakes up one of the
1638  *   threads waiting to acquire it. Mutexes are always synchronization
1639  *   events.
1640  *
1641  * - For threads, the only thing we do is wait until the thread object
1642  *   enters a signalled state, which occurs when the thread terminates.
1643  *   Threads are always notification events.
1644  *
1645  * A notification event wakes up all threads waiting on an object. A
1646  * synchronization event wakes up just one. Also, a synchronization event
1647  * is auto-clearing, which means we automatically set the event back to
1648  * the non-signalled state once the wakeup is done.
1649  */
1650 
1651 uint32_t
1652 KeWaitForSingleObject(void *arg, uint32_t reason, uint32_t mode,
1653     uint8_t alertable, int64_t *duetime)
1654 {
1655 	wait_block		w;
1656 	struct thread		*td = curthread;
1657 	struct timeval		tv;
1658 	int			error = 0;
1659 	uint64_t		curtime;
1660 	wb_ext			we;
1661 	nt_dispatch_header	*obj;
1662 
1663 	obj = arg;
1664 
1665 	if (obj == NULL)
1666 		return (STATUS_INVALID_PARAMETER);
1667 
1668 	mtx_lock(&ntoskrnl_dispatchlock);
1669 
1670 	cv_init(&we.we_cv, "KeWFS");
1671 	we.we_td = td;
1672 
1673 	/*
1674 	 * Check to see if this object is already signalled,
1675 	 * and just return without waiting if it is.
1676 	 */
1677 	if (ntoskrnl_is_signalled(obj, td) == TRUE) {
1678 		/* Sanity check the signal state value. */
1679 		if (obj->dh_sigstate != INT32_MIN) {
1680 			ntoskrnl_satisfy_wait(obj, curthread);
1681 			mtx_unlock(&ntoskrnl_dispatchlock);
1682 			return (STATUS_SUCCESS);
1683 		} else {
1684 			/*
1685 			 * There's a limit to how many times we can
1686 			 * recursively acquire a mutant. If we hit
1687 			 * the limit, something is very wrong.
1688 			 */
1689 			if (obj->dh_type == DISP_TYPE_MUTANT) {
1690 				mtx_unlock(&ntoskrnl_dispatchlock);
1691 				panic("mutant limit exceeded");
1692 			}
1693 		}
1694 	}
1695 
1696 	bzero((char *)&w, sizeof(wait_block));
1697 	w.wb_object = obj;
1698 	w.wb_ext = &we;
1699 	w.wb_waittype = WAITTYPE_ANY;
1700 	w.wb_next = &w;
1701 	w.wb_waitkey = 0;
1702 	w.wb_awakened = FALSE;
1703 	w.wb_oldpri = td->td_priority;
1704 
1705 	InsertTailList((&obj->dh_waitlisthead), (&w.wb_waitlist));
1706 
1707 	/*
1708 	 * The timeout value is specified in 100 nanosecond units
1709 	 * and can be a positive or negative number. If it's positive,
1710 	 * then the duetime is absolute, and we need to convert it
1711 	 * to an absolute offset relative to now in order to use it.
1712 	 * If it's negative, then the duetime is relative and we
1713 	 * just have to convert the units.
1714 	 */
1715 
1716 	if (duetime != NULL) {
1717 		if (*duetime < 0) {
1718 			tv.tv_sec = - (*duetime) / 10000000;
1719 			tv.tv_usec = (- (*duetime) / 10) -
1720 			    (tv.tv_sec * 1000000);
1721 		} else {
1722 			ntoskrnl_time(&curtime);
1723 			if (*duetime < curtime)
1724 				tv.tv_sec = tv.tv_usec = 0;
1725 			else {
1726 				tv.tv_sec = ((*duetime) - curtime) / 10000000;
1727 				tv.tv_usec = ((*duetime) - curtime) / 10 -
1728 				    (tv.tv_sec * 1000000);
1729 			}
1730 		}
1731 	}
1732 
1733 	if (duetime == NULL)
1734 		cv_wait(&we.we_cv, &ntoskrnl_dispatchlock);
1735 	else
1736 		error = cv_timedwait(&we.we_cv,
1737 		    &ntoskrnl_dispatchlock, tvtohz(&tv));
1738 
1739 	RemoveEntryList(&w.wb_waitlist);
1740 
1741 	cv_destroy(&we.we_cv);
1742 
1743 	/* We timed out. Leave the object alone and return status. */
1744 
1745 	if (error == EWOULDBLOCK) {
1746 		mtx_unlock(&ntoskrnl_dispatchlock);
1747 		return (STATUS_TIMEOUT);
1748 	}
1749 
1750 	mtx_unlock(&ntoskrnl_dispatchlock);
1751 
1752 	return (STATUS_SUCCESS);
1753 /*
1754 	return (KeWaitForMultipleObjects(1, &obj, WAITTYPE_ALL, reason,
1755 	    mode, alertable, duetime, &w));
1756 */
1757 }
1758 
1759 static uint32_t
1760 KeWaitForMultipleObjects(uint32_t cnt, nt_dispatch_header *obj[], uint32_t wtype,
1761 	uint32_t reason, uint32_t mode, uint8_t alertable, int64_t *duetime,
1762 	wait_block *wb_array)
1763 {
1764 	struct thread		*td = curthread;
1765 	wait_block		*whead, *w;
1766 	wait_block		_wb_array[MAX_WAIT_OBJECTS];
1767 	nt_dispatch_header	*cur;
1768 	struct timeval		tv;
1769 	int			i, wcnt = 0, error = 0;
1770 	uint64_t		curtime;
1771 	struct timespec		t1, t2;
1772 	uint32_t		status = STATUS_SUCCESS;
1773 	wb_ext			we;
1774 
1775 	if (cnt > MAX_WAIT_OBJECTS)
1776 		return (STATUS_INVALID_PARAMETER);
1777 	if (cnt > THREAD_WAIT_OBJECTS && wb_array == NULL)
1778 		return (STATUS_INVALID_PARAMETER);
1779 
1780 	mtx_lock(&ntoskrnl_dispatchlock);
1781 
1782 	cv_init(&we.we_cv, "KeWFM");
1783 	we.we_td = td;
1784 
1785 	if (wb_array == NULL)
1786 		whead = _wb_array;
1787 	else
1788 		whead = wb_array;
1789 
1790 	bzero((char *)whead, sizeof(wait_block) * cnt);
1791 
1792 	/* First pass: see if we can satisfy any waits immediately. */
1793 
1794 	wcnt = 0;
1795 	w = whead;
1796 
1797 	for (i = 0; i < cnt; i++) {
1798 		InsertTailList((&obj[i]->dh_waitlisthead),
1799 		    (&w->wb_waitlist));
1800 		w->wb_ext = &we;
1801 		w->wb_object = obj[i];
1802 		w->wb_waittype = wtype;
1803 		w->wb_waitkey = i;
1804 		w->wb_awakened = FALSE;
1805 		w->wb_oldpri = td->td_priority;
1806 		w->wb_next = w + 1;
1807 		w++;
1808 		wcnt++;
1809 		if (ntoskrnl_is_signalled(obj[i], td)) {
1810 			/*
1811 			 * There's a limit to how many times
1812 			 * we can recursively acquire a mutant.
1813 			 * If we hit the limit, something
1814 			 * is very wrong.
1815 			 */
1816 			if (obj[i]->dh_sigstate == INT32_MIN &&
1817 			    obj[i]->dh_type == DISP_TYPE_MUTANT) {
1818 				mtx_unlock(&ntoskrnl_dispatchlock);
1819 				panic("mutant limit exceeded");
1820 			}
1821 
1822 			/*
1823 			 * If this is a WAITTYPE_ANY wait, then
1824 			 * satisfy the waited object and exit
1825 			 * right now.
1826 			 */
1827 
1828 			if (wtype == WAITTYPE_ANY) {
1829 				ntoskrnl_satisfy_wait(obj[i], td);
1830 				status = STATUS_WAIT_0 + i;
1831 				goto wait_done;
1832 			} else {
1833 				w--;
1834 				wcnt--;
1835 				w->wb_object = NULL;
1836 				RemoveEntryList(&w->wb_waitlist);
1837 			}
1838 		}
1839 	}
1840 
1841 	/*
1842 	 * If this is a WAITTYPE_ALL wait and all objects are
1843 	 * already signalled, satisfy the waits and exit now.
1844 	 */
1845 
1846 	if (wtype == WAITTYPE_ALL && wcnt == 0) {
1847 		for (i = 0; i < cnt; i++)
1848 			ntoskrnl_satisfy_wait(obj[i], td);
1849 		status = STATUS_SUCCESS;
1850 		goto wait_done;
1851 	}
1852 
1853 	/*
1854 	 * Create a circular waitblock list. The waitcount
1855 	 * must always be non-zero when we get here.
1856 	 */
1857 
1858 	(w - 1)->wb_next = whead;
1859 
1860 	/* Wait on any objects that aren't yet signalled. */
1861 
1862 	/* Calculate timeout, if any. */
1863 
1864 	if (duetime != NULL) {
1865 		if (*duetime < 0) {
1866 			tv.tv_sec = - (*duetime) / 10000000;
1867 			tv.tv_usec = (- (*duetime) / 10) -
1868 			    (tv.tv_sec * 1000000);
1869 		} else {
1870 			ntoskrnl_time(&curtime);
1871 			if (*duetime < curtime)
1872 				tv.tv_sec = tv.tv_usec = 0;
1873 			else {
1874 				tv.tv_sec = ((*duetime) - curtime) / 10000000;
1875 				tv.tv_usec = ((*duetime) - curtime) / 10 -
1876 				    (tv.tv_sec * 1000000);
1877 			}
1878 		}
1879 	}
1880 
1881 	while (wcnt) {
1882 		nanotime(&t1);
1883 
1884 		if (duetime == NULL)
1885 			cv_wait(&we.we_cv, &ntoskrnl_dispatchlock);
1886 		else
1887 			error = cv_timedwait(&we.we_cv,
1888 			    &ntoskrnl_dispatchlock, tvtohz(&tv));
1889 
1890 		/* Wait with timeout expired. */
1891 
1892 		if (error) {
1893 			status = STATUS_TIMEOUT;
1894 			goto wait_done;
1895 		}
1896 
1897 		nanotime(&t2);
1898 
1899 		/* See what's been signalled. */
1900 
1901 		w = whead;
1902 		do {
1903 			cur = w->wb_object;
1904 			if (ntoskrnl_is_signalled(cur, td) == TRUE ||
1905 			    w->wb_awakened == TRUE) {
1906 				/* Sanity check the signal state value. */
1907 				if (cur->dh_sigstate == INT32_MIN &&
1908 				    cur->dh_type == DISP_TYPE_MUTANT) {
1909 					mtx_unlock(&ntoskrnl_dispatchlock);
1910 					panic("mutant limit exceeded");
1911 				}
1912 				wcnt--;
1913 				if (wtype == WAITTYPE_ANY) {
1914 					status = w->wb_waitkey &
1915 					    STATUS_WAIT_0;
1916 					goto wait_done;
1917 				}
1918 			}
1919 			w = w->wb_next;
1920 		} while (w != whead);
1921 
1922 		/*
1923 		 * If all objects have been signalled, or if this
1924 		 * is a WAITTYPE_ANY wait and we were woke up by
1925 		 * someone, we can bail.
1926 		 */
1927 
1928 		if (wcnt == 0) {
1929 			status = STATUS_SUCCESS;
1930 			goto wait_done;
1931 		}
1932 
1933 		/*
1934 		 * If this is WAITTYPE_ALL wait, and there's still
1935 		 * objects that haven't been signalled, deduct the
1936 		 * time that's elapsed so far from the timeout and
1937 		 * wait again (or continue waiting indefinitely if
1938 		 * there's no timeout).
1939 		 */
1940 
1941 		if (duetime != NULL) {
1942 			tv.tv_sec -= (t2.tv_sec - t1.tv_sec);
1943 			tv.tv_usec -= (t2.tv_nsec - t1.tv_nsec) / 1000;
1944 		}
1945 	}
1946 
1947 wait_done:
1948 
1949 	cv_destroy(&we.we_cv);
1950 
1951 	for (i = 0; i < cnt; i++) {
1952 		if (whead[i].wb_object != NULL)
1953 			RemoveEntryList(&whead[i].wb_waitlist);
1954 	}
1955 	mtx_unlock(&ntoskrnl_dispatchlock);
1956 
1957 	return (status);
1958 }
1959 
1960 static void
1961 WRITE_REGISTER_USHORT(uint16_t *reg, uint16_t val)
1962 {
1963 	bus_space_write_2(NDIS_BUS_SPACE_MEM, 0x0, (bus_size_t)reg, val);
1964 }
1965 
1966 static uint16_t
1967 READ_REGISTER_USHORT(reg)
1968 	uint16_t		*reg;
1969 {
1970 	return (bus_space_read_2(NDIS_BUS_SPACE_MEM, 0x0, (bus_size_t)reg));
1971 }
1972 
1973 static void
1974 WRITE_REGISTER_ULONG(reg, val)
1975 	uint32_t		*reg;
1976 	uint32_t		val;
1977 {
1978 	bus_space_write_4(NDIS_BUS_SPACE_MEM, 0x0, (bus_size_t)reg, val);
1979 }
1980 
1981 static uint32_t
1982 READ_REGISTER_ULONG(reg)
1983 	uint32_t		*reg;
1984 {
1985 	return (bus_space_read_4(NDIS_BUS_SPACE_MEM, 0x0, (bus_size_t)reg));
1986 }
1987 
1988 static uint8_t
1989 READ_REGISTER_UCHAR(uint8_t *reg)
1990 {
1991 	return (bus_space_read_1(NDIS_BUS_SPACE_MEM, 0x0, (bus_size_t)reg));
1992 }
1993 
1994 static void
1995 WRITE_REGISTER_UCHAR(uint8_t *reg, uint8_t val)
1996 {
1997 	bus_space_write_1(NDIS_BUS_SPACE_MEM, 0x0, (bus_size_t)reg, val);
1998 }
1999 
2000 static int64_t
2001 _allmul(a, b)
2002 	int64_t			a;
2003 	int64_t			b;
2004 {
2005 	return (a * b);
2006 }
2007 
2008 static int64_t
2009 _alldiv(a, b)
2010 	int64_t			a;
2011 	int64_t			b;
2012 {
2013 	return (a / b);
2014 }
2015 
2016 static int64_t
2017 _allrem(a, b)
2018 	int64_t			a;
2019 	int64_t			b;
2020 {
2021 	return (a % b);
2022 }
2023 
2024 static uint64_t
2025 _aullmul(a, b)
2026 	uint64_t		a;
2027 	uint64_t		b;
2028 {
2029 	return (a * b);
2030 }
2031 
2032 static uint64_t
2033 _aulldiv(a, b)
2034 	uint64_t		a;
2035 	uint64_t		b;
2036 {
2037 	return (a / b);
2038 }
2039 
2040 static uint64_t
2041 _aullrem(a, b)
2042 	uint64_t		a;
2043 	uint64_t		b;
2044 {
2045 	return (a % b);
2046 }
2047 
2048 static int64_t
2049 _allshl(int64_t a, uint8_t b)
2050 {
2051 	return (a << b);
2052 }
2053 
2054 static uint64_t
2055 _aullshl(uint64_t a, uint8_t b)
2056 {
2057 	return (a << b);
2058 }
2059 
2060 static int64_t
2061 _allshr(int64_t a, uint8_t b)
2062 {
2063 	return (a >> b);
2064 }
2065 
2066 static uint64_t
2067 _aullshr(uint64_t a, uint8_t b)
2068 {
2069 	return (a >> b);
2070 }
2071 
2072 static slist_entry *
2073 ntoskrnl_pushsl(head, entry)
2074 	slist_header		*head;
2075 	slist_entry		*entry;
2076 {
2077 	slist_entry		*oldhead;
2078 
2079 	oldhead = head->slh_list.slh_next;
2080 	entry->sl_next = head->slh_list.slh_next;
2081 	head->slh_list.slh_next = entry;
2082 	head->slh_list.slh_depth++;
2083 	head->slh_list.slh_seq++;
2084 
2085 	return (oldhead);
2086 }
2087 
2088 static void
2089 InitializeSListHead(head)
2090 	slist_header		*head;
2091 {
2092 	memset(head, 0, sizeof(*head));
2093 }
2094 
2095 static slist_entry *
2096 ntoskrnl_popsl(head)
2097 	slist_header		*head;
2098 {
2099 	slist_entry		*first;
2100 
2101 	first = head->slh_list.slh_next;
2102 	if (first != NULL) {
2103 		head->slh_list.slh_next = first->sl_next;
2104 		head->slh_list.slh_depth--;
2105 		head->slh_list.slh_seq++;
2106 	}
2107 
2108 	return (first);
2109 }
2110 
2111 /*
2112  * We need this to make lookaside lists work for amd64.
2113  * We pass a pointer to ExAllocatePoolWithTag() the lookaside
2114  * list structure. For amd64 to work right, this has to be a
2115  * pointer to the wrapped version of the routine, not the
2116  * original. Letting the Windows driver invoke the original
2117  * function directly will result in a convention calling
2118  * mismatch and a pretty crash. On x86, this effectively
2119  * becomes a no-op since ipt_func and ipt_wrap are the same.
2120  */
2121 
2122 static funcptr
2123 ntoskrnl_findwrap(func)
2124 	funcptr			func;
2125 {
2126 	image_patch_table	*patch;
2127 
2128 	patch = ntoskrnl_functbl;
2129 	while (patch->ipt_func != NULL) {
2130 		if ((funcptr)patch->ipt_func == func)
2131 			return ((funcptr)patch->ipt_wrap);
2132 		patch++;
2133 	}
2134 
2135 	return (NULL);
2136 }
2137 
2138 static void
2139 ExInitializePagedLookasideList(paged_lookaside_list *lookaside,
2140 	lookaside_alloc_func *allocfunc, lookaside_free_func *freefunc,
2141 	uint32_t flags, size_t size, uint32_t tag, uint16_t depth)
2142 {
2143 	bzero((char *)lookaside, sizeof(paged_lookaside_list));
2144 
2145 	if (size < sizeof(slist_entry))
2146 		lookaside->nll_l.gl_size = sizeof(slist_entry);
2147 	else
2148 		lookaside->nll_l.gl_size = size;
2149 	lookaside->nll_l.gl_tag = tag;
2150 	if (allocfunc == NULL)
2151 		lookaside->nll_l.gl_allocfunc =
2152 		    ntoskrnl_findwrap((funcptr)ExAllocatePoolWithTag);
2153 	else
2154 		lookaside->nll_l.gl_allocfunc = allocfunc;
2155 
2156 	if (freefunc == NULL)
2157 		lookaside->nll_l.gl_freefunc =
2158 		    ntoskrnl_findwrap((funcptr)ExFreePool);
2159 	else
2160 		lookaside->nll_l.gl_freefunc = freefunc;
2161 
2162 #ifdef __i386__
2163 	KeInitializeSpinLock(&lookaside->nll_obsoletelock);
2164 #endif
2165 
2166 	lookaside->nll_l.gl_type = NonPagedPool;
2167 	lookaside->nll_l.gl_depth = depth;
2168 	lookaside->nll_l.gl_maxdepth = LOOKASIDE_DEPTH;
2169 }
2170 
2171 static void
2172 ExDeletePagedLookasideList(lookaside)
2173 	paged_lookaside_list   *lookaside;
2174 {
2175 	void			*buf;
2176 	void		(*freefunc)(void *);
2177 
2178 	freefunc = lookaside->nll_l.gl_freefunc;
2179 	while((buf = ntoskrnl_popsl(&lookaside->nll_l.gl_listhead)) != NULL)
2180 		MSCALL1(freefunc, buf);
2181 }
2182 
2183 static void
2184 ExInitializeNPagedLookasideList(npaged_lookaside_list *lookaside,
2185 	lookaside_alloc_func *allocfunc, lookaside_free_func *freefunc,
2186 	uint32_t flags, size_t size, uint32_t tag, uint16_t depth)
2187 {
2188 	bzero((char *)lookaside, sizeof(npaged_lookaside_list));
2189 
2190 	if (size < sizeof(slist_entry))
2191 		lookaside->nll_l.gl_size = sizeof(slist_entry);
2192 	else
2193 		lookaside->nll_l.gl_size = size;
2194 	lookaside->nll_l.gl_tag = tag;
2195 	if (allocfunc == NULL)
2196 		lookaside->nll_l.gl_allocfunc =
2197 		    ntoskrnl_findwrap((funcptr)ExAllocatePoolWithTag);
2198 	else
2199 		lookaside->nll_l.gl_allocfunc = allocfunc;
2200 
2201 	if (freefunc == NULL)
2202 		lookaside->nll_l.gl_freefunc =
2203 		    ntoskrnl_findwrap((funcptr)ExFreePool);
2204 	else
2205 		lookaside->nll_l.gl_freefunc = freefunc;
2206 
2207 #ifdef __i386__
2208 	KeInitializeSpinLock(&lookaside->nll_obsoletelock);
2209 #endif
2210 
2211 	lookaside->nll_l.gl_type = NonPagedPool;
2212 	lookaside->nll_l.gl_depth = depth;
2213 	lookaside->nll_l.gl_maxdepth = LOOKASIDE_DEPTH;
2214 }
2215 
2216 static void
2217 ExDeleteNPagedLookasideList(lookaside)
2218 	npaged_lookaside_list   *lookaside;
2219 {
2220 	void			*buf;
2221 	void		(*freefunc)(void *);
2222 
2223 	freefunc = lookaside->nll_l.gl_freefunc;
2224 	while((buf = ntoskrnl_popsl(&lookaside->nll_l.gl_listhead)) != NULL)
2225 		MSCALL1(freefunc, buf);
2226 }
2227 
2228 slist_entry *
2229 InterlockedPushEntrySList(head, entry)
2230 	slist_header		*head;
2231 	slist_entry		*entry;
2232 {
2233 	slist_entry		*oldhead;
2234 
2235 	mtx_lock_spin(&ntoskrnl_interlock);
2236 	oldhead = ntoskrnl_pushsl(head, entry);
2237 	mtx_unlock_spin(&ntoskrnl_interlock);
2238 
2239 	return (oldhead);
2240 }
2241 
2242 slist_entry *
2243 InterlockedPopEntrySList(head)
2244 	slist_header		*head;
2245 {
2246 	slist_entry		*first;
2247 
2248 	mtx_lock_spin(&ntoskrnl_interlock);
2249 	first = ntoskrnl_popsl(head);
2250 	mtx_unlock_spin(&ntoskrnl_interlock);
2251 
2252 	return (first);
2253 }
2254 
2255 static slist_entry *
2256 ExInterlockedPushEntrySList(head, entry, lock)
2257 	slist_header		*head;
2258 	slist_entry		*entry;
2259 	kspin_lock		*lock;
2260 {
2261 	return (InterlockedPushEntrySList(head, entry));
2262 }
2263 
2264 static slist_entry *
2265 ExInterlockedPopEntrySList(head, lock)
2266 	slist_header		*head;
2267 	kspin_lock		*lock;
2268 {
2269 	return (InterlockedPopEntrySList(head));
2270 }
2271 
2272 uint16_t
2273 ExQueryDepthSList(head)
2274 	slist_header		*head;
2275 {
2276 	uint16_t		depth;
2277 
2278 	mtx_lock_spin(&ntoskrnl_interlock);
2279 	depth = head->slh_list.slh_depth;
2280 	mtx_unlock_spin(&ntoskrnl_interlock);
2281 
2282 	return (depth);
2283 }
2284 
2285 void
2286 KeInitializeSpinLock(lock)
2287 	kspin_lock		*lock;
2288 {
2289 	*lock = 0;
2290 }
2291 
2292 #ifdef __i386__
2293 void
2294 KefAcquireSpinLockAtDpcLevel(lock)
2295 	kspin_lock		*lock;
2296 {
2297 #ifdef NTOSKRNL_DEBUG_SPINLOCKS
2298 	int			i = 0;
2299 #endif
2300 
2301 	while (atomic_cmpset_acq_int((volatile u_int *)lock, 0, 1) == 0) {
2302 		/* sit and spin */;
2303 #ifdef NTOSKRNL_DEBUG_SPINLOCKS
2304 		i++;
2305 		if (i > 200000000)
2306 			panic("DEADLOCK!");
2307 #endif
2308 	}
2309 }
2310 
2311 void
2312 KefReleaseSpinLockFromDpcLevel(lock)
2313 	kspin_lock		*lock;
2314 {
2315 	atomic_store_rel_int((volatile u_int *)lock, 0);
2316 }
2317 
2318 uint8_t
2319 KeAcquireSpinLockRaiseToDpc(kspin_lock *lock)
2320 {
2321 	uint8_t			oldirql;
2322 
2323 	if (KeGetCurrentIrql() > DISPATCH_LEVEL)
2324 		panic("IRQL_NOT_LESS_THAN_OR_EQUAL");
2325 
2326 	KeRaiseIrql(DISPATCH_LEVEL, &oldirql);
2327 	KeAcquireSpinLockAtDpcLevel(lock);
2328 
2329 	return (oldirql);
2330 }
2331 #else
2332 void
2333 KeAcquireSpinLockAtDpcLevel(kspin_lock *lock)
2334 {
2335 	while (atomic_cmpset_acq_int((volatile u_int *)lock, 0, 1) == 0)
2336 		/* sit and spin */;
2337 }
2338 
2339 void
2340 KeReleaseSpinLockFromDpcLevel(kspin_lock *lock)
2341 {
2342 	atomic_store_rel_int((volatile u_int *)lock, 0);
2343 }
2344 #endif /* __i386__ */
2345 
2346 uintptr_t
2347 InterlockedExchange(dst, val)
2348 	volatile uint32_t	*dst;
2349 	uintptr_t		val;
2350 {
2351 	uintptr_t		r;
2352 
2353 	mtx_lock_spin(&ntoskrnl_interlock);
2354 	r = *dst;
2355 	*dst = val;
2356 	mtx_unlock_spin(&ntoskrnl_interlock);
2357 
2358 	return (r);
2359 }
2360 
2361 static uint32_t
2362 InterlockedIncrement(addend)
2363 	volatile uint32_t	*addend;
2364 {
2365 	atomic_add_long((volatile u_long *)addend, 1);
2366 	return (*addend);
2367 }
2368 
2369 static uint32_t
2370 InterlockedDecrement(addend)
2371 	volatile uint32_t	*addend;
2372 {
2373 	atomic_subtract_long((volatile u_long *)addend, 1);
2374 	return (*addend);
2375 }
2376 
2377 static void
2378 ExInterlockedAddLargeStatistic(addend, inc)
2379 	uint64_t		*addend;
2380 	uint32_t		inc;
2381 {
2382 	mtx_lock_spin(&ntoskrnl_interlock);
2383 	*addend += inc;
2384 	mtx_unlock_spin(&ntoskrnl_interlock);
2385 };
2386 
2387 mdl *
2388 IoAllocateMdl(void *vaddr, uint32_t len, uint8_t secondarybuf,
2389 	uint8_t chargequota, irp *iopkt)
2390 {
2391 	mdl			*m;
2392 	int			zone = 0;
2393 
2394 	if (MmSizeOfMdl(vaddr, len) > MDL_ZONE_SIZE)
2395 		m = ExAllocatePoolWithTag(NonPagedPool,
2396 		    MmSizeOfMdl(vaddr, len), 0);
2397 	else {
2398 		m = uma_zalloc(mdl_zone, M_NOWAIT | M_ZERO);
2399 		zone++;
2400 	}
2401 
2402 	if (m == NULL)
2403 		return (NULL);
2404 
2405 	MmInitializeMdl(m, vaddr, len);
2406 
2407 	/*
2408 	 * MmInitializMdl() clears the flags field, so we
2409 	 * have to set this here. If the MDL came from the
2410 	 * MDL UMA zone, tag it so we can release it to
2411 	 * the right place later.
2412 	 */
2413 	if (zone)
2414 		m->mdl_flags = MDL_ZONE_ALLOCED;
2415 
2416 	if (iopkt != NULL) {
2417 		if (secondarybuf == TRUE) {
2418 			mdl			*last;
2419 			last = iopkt->irp_mdl;
2420 			while (last->mdl_next != NULL)
2421 				last = last->mdl_next;
2422 			last->mdl_next = m;
2423 		} else {
2424 			if (iopkt->irp_mdl != NULL)
2425 				panic("leaking an MDL in IoAllocateMdl()");
2426 			iopkt->irp_mdl = m;
2427 		}
2428 	}
2429 
2430 	return (m);
2431 }
2432 
2433 void
2434 IoFreeMdl(m)
2435 	mdl			*m;
2436 {
2437 	if (m == NULL)
2438 		return;
2439 
2440 	if (m->mdl_flags & MDL_ZONE_ALLOCED)
2441 		uma_zfree(mdl_zone, m);
2442 	else
2443 		ExFreePool(m);
2444 }
2445 
2446 static void *
2447 MmAllocateContiguousMemory(size, highest)
2448 	uint32_t		size;
2449 	uint64_t		highest;
2450 {
2451 	void *addr;
2452 	size_t pagelength = roundup(size, PAGE_SIZE);
2453 
2454 	addr = ExAllocatePoolWithTag(NonPagedPool, pagelength, 0);
2455 
2456 	return (addr);
2457 }
2458 
2459 static void *
2460 MmAllocateContiguousMemorySpecifyCache(size, lowest, highest,
2461     boundary, cachetype)
2462 	uint32_t		size;
2463 	uint64_t		lowest;
2464 	uint64_t		highest;
2465 	uint64_t		boundary;
2466 	enum nt_caching_type	cachetype;
2467 {
2468 	vm_memattr_t		memattr;
2469 	void			*ret;
2470 
2471 	switch (cachetype) {
2472 	case MmNonCached:
2473 		memattr = VM_MEMATTR_UNCACHEABLE;
2474 		break;
2475 	case MmWriteCombined:
2476 		memattr = VM_MEMATTR_WRITE_COMBINING;
2477 		break;
2478 	case MmNonCachedUnordered:
2479 		memattr = VM_MEMATTR_UNCACHEABLE;
2480 		break;
2481 	case MmCached:
2482 	case MmHardwareCoherentCached:
2483 	case MmUSWCCached:
2484 	default:
2485 		memattr = VM_MEMATTR_DEFAULT;
2486 		break;
2487 	}
2488 
2489 	ret = (void *)kmem_alloc_contig(size, M_ZERO | M_NOWAIT, lowest,
2490 	    highest, PAGE_SIZE, boundary, memattr);
2491 	if (ret != NULL)
2492 		malloc_type_allocated(M_DEVBUF, round_page(size));
2493 	return (ret);
2494 }
2495 
2496 static void
2497 MmFreeContiguousMemory(base)
2498 	void			*base;
2499 {
2500 	ExFreePool(base);
2501 }
2502 
2503 static void
2504 MmFreeContiguousMemorySpecifyCache(base, size, cachetype)
2505 	void			*base;
2506 	uint32_t		size;
2507 	enum nt_caching_type	cachetype;
2508 {
2509 	contigfree(base, size, M_DEVBUF);
2510 }
2511 
2512 static uint32_t
2513 MmSizeOfMdl(vaddr, len)
2514 	void			*vaddr;
2515 	size_t			len;
2516 {
2517 	uint32_t		l;
2518 
2519 	l = sizeof(struct mdl) +
2520 	    (sizeof(vm_offset_t *) * SPAN_PAGES(vaddr, len));
2521 
2522 	return (l);
2523 }
2524 
2525 /*
2526  * The Microsoft documentation says this routine fills in the
2527  * page array of an MDL with the _physical_ page addresses that
2528  * comprise the buffer, but we don't really want to do that here.
2529  * Instead, we just fill in the page array with the kernel virtual
2530  * addresses of the buffers.
2531  */
2532 void
2533 MmBuildMdlForNonPagedPool(m)
2534 	mdl			*m;
2535 {
2536 	vm_offset_t		*mdl_pages;
2537 	int			pagecnt, i;
2538 
2539 	pagecnt = SPAN_PAGES(m->mdl_byteoffset, m->mdl_bytecount);
2540 
2541 	if (pagecnt > (m->mdl_size - sizeof(mdl)) / sizeof(vm_offset_t *))
2542 		panic("not enough pages in MDL to describe buffer");
2543 
2544 	mdl_pages = MmGetMdlPfnArray(m);
2545 
2546 	for (i = 0; i < pagecnt; i++)
2547 		*mdl_pages = (vm_offset_t)m->mdl_startva + (i * PAGE_SIZE);
2548 
2549 	m->mdl_flags |= MDL_SOURCE_IS_NONPAGED_POOL;
2550 	m->mdl_mappedsystemva = MmGetMdlVirtualAddress(m);
2551 }
2552 
2553 static void *
2554 MmMapLockedPages(mdl *buf, uint8_t accessmode)
2555 {
2556 	buf->mdl_flags |= MDL_MAPPED_TO_SYSTEM_VA;
2557 	return (MmGetMdlVirtualAddress(buf));
2558 }
2559 
2560 static void *
2561 MmMapLockedPagesSpecifyCache(mdl *buf, uint8_t accessmode, uint32_t cachetype,
2562 	void *vaddr, uint32_t bugcheck, uint32_t prio)
2563 {
2564 	return (MmMapLockedPages(buf, accessmode));
2565 }
2566 
2567 static void
2568 MmUnmapLockedPages(vaddr, buf)
2569 	void			*vaddr;
2570 	mdl			*buf;
2571 {
2572 	buf->mdl_flags &= ~MDL_MAPPED_TO_SYSTEM_VA;
2573 }
2574 
2575 /*
2576  * This function has a problem in that it will break if you
2577  * compile this module without PAE and try to use it on a PAE
2578  * kernel. Unfortunately, there's no way around this at the
2579  * moment. It's slightly less broken that using pmap_kextract().
2580  * You'd think the virtual memory subsystem would help us out
2581  * here, but it doesn't.
2582  */
2583 
2584 static uint64_t
2585 MmGetPhysicalAddress(void *base)
2586 {
2587 	return (pmap_extract(kernel_map->pmap, (vm_offset_t)base));
2588 }
2589 
2590 void *
2591 MmGetSystemRoutineAddress(ustr)
2592 	unicode_string		*ustr;
2593 {
2594 	ansi_string		astr;
2595 
2596 	if (RtlUnicodeStringToAnsiString(&astr, ustr, TRUE))
2597 		return (NULL);
2598 	return (ndis_get_routine_address(ntoskrnl_functbl, astr.as_buf));
2599 }
2600 
2601 uint8_t
2602 MmIsAddressValid(vaddr)
2603 	void			*vaddr;
2604 {
2605 	if (pmap_extract(kernel_map->pmap, (vm_offset_t)vaddr))
2606 		return (TRUE);
2607 
2608 	return (FALSE);
2609 }
2610 
2611 void *
2612 MmMapIoSpace(paddr, len, cachetype)
2613 	uint64_t		paddr;
2614 	uint32_t		len;
2615 	uint32_t		cachetype;
2616 {
2617 	devclass_t		nexus_class;
2618 	device_t		*nexus_devs, devp;
2619 	int			nexus_count = 0;
2620 	device_t		matching_dev = NULL;
2621 	struct resource		*res;
2622 	int			i;
2623 	vm_offset_t		v;
2624 
2625 	/* There will always be at least one nexus. */
2626 
2627 	nexus_class = devclass_find("nexus");
2628 	devclass_get_devices(nexus_class, &nexus_devs, &nexus_count);
2629 
2630 	for (i = 0; i < nexus_count; i++) {
2631 		devp = nexus_devs[i];
2632 		matching_dev = ntoskrnl_finddev(devp, paddr, &res);
2633 		if (matching_dev)
2634 			break;
2635 	}
2636 
2637 	free(nexus_devs, M_TEMP);
2638 
2639 	if (matching_dev == NULL)
2640 		return (NULL);
2641 
2642 	v = (vm_offset_t)rman_get_virtual(res);
2643 	if (paddr > rman_get_start(res))
2644 		v += paddr - rman_get_start(res);
2645 
2646 	return ((void *)v);
2647 }
2648 
2649 void
2650 MmUnmapIoSpace(vaddr, len)
2651 	void			*vaddr;
2652 	size_t			len;
2653 {
2654 }
2655 
2656 static device_t
2657 ntoskrnl_finddev(dev, paddr, res)
2658 	device_t		dev;
2659 	uint64_t		paddr;
2660 	struct resource		**res;
2661 {
2662 	device_t		*children = NULL;
2663 	device_t		matching_dev;
2664 	int			childcnt;
2665 	struct resource		*r;
2666 	struct resource_list	*rl;
2667 	struct resource_list_entry	*rle;
2668 	uint32_t		flags;
2669 	int			i;
2670 
2671 	/* We only want devices that have been successfully probed. */
2672 
2673 	if (device_is_alive(dev) == FALSE)
2674 		return (NULL);
2675 
2676 	rl = BUS_GET_RESOURCE_LIST(device_get_parent(dev), dev);
2677 	if (rl != NULL) {
2678 		STAILQ_FOREACH(rle, rl, link) {
2679 			r = rle->res;
2680 
2681 			if (r == NULL)
2682 				continue;
2683 
2684 			flags = rman_get_flags(r);
2685 
2686 			if (rle->type == SYS_RES_MEMORY &&
2687 			    paddr >= rman_get_start(r) &&
2688 			    paddr <= rman_get_end(r)) {
2689 				if (!(flags & RF_ACTIVE))
2690 					bus_activate_resource(dev,
2691 					    SYS_RES_MEMORY, 0, r);
2692 				*res = r;
2693 				return (dev);
2694 			}
2695 		}
2696 	}
2697 
2698 	/*
2699 	 * If this device has children, do another
2700 	 * level of recursion to inspect them.
2701 	 */
2702 
2703 	device_get_children(dev, &children, &childcnt);
2704 
2705 	for (i = 0; i < childcnt; i++) {
2706 		matching_dev = ntoskrnl_finddev(children[i], paddr, res);
2707 		if (matching_dev != NULL) {
2708 			free(children, M_TEMP);
2709 			return (matching_dev);
2710 		}
2711 	}
2712 
2713 	/* Won't somebody please think of the children! */
2714 
2715 	if (children != NULL)
2716 		free(children, M_TEMP);
2717 
2718 	return (NULL);
2719 }
2720 
2721 /*
2722  * Workitems are unlike DPCs, in that they run in a user-mode thread
2723  * context rather than at DISPATCH_LEVEL in kernel context. In our
2724  * case we run them in kernel context anyway.
2725  */
2726 static void
2727 ntoskrnl_workitem_thread(arg)
2728 	void			*arg;
2729 {
2730 	kdpc_queue		*kq;
2731 	list_entry		*l;
2732 	io_workitem		*iw;
2733 	uint8_t			irql;
2734 
2735 	kq = arg;
2736 
2737 	InitializeListHead(&kq->kq_disp);
2738 	kq->kq_td = curthread;
2739 	kq->kq_exit = 0;
2740 	KeInitializeSpinLock(&kq->kq_lock);
2741 	KeInitializeEvent(&kq->kq_proc, EVENT_TYPE_SYNC, FALSE);
2742 
2743 	while (1) {
2744 		KeWaitForSingleObject(&kq->kq_proc, 0, 0, TRUE, NULL);
2745 
2746 		KeAcquireSpinLock(&kq->kq_lock, &irql);
2747 
2748 		if (kq->kq_exit) {
2749 			kq->kq_exit = 0;
2750 			KeReleaseSpinLock(&kq->kq_lock, irql);
2751 			break;
2752 		}
2753 
2754 		while (!IsListEmpty(&kq->kq_disp)) {
2755 			l = RemoveHeadList(&kq->kq_disp);
2756 			iw = CONTAINING_RECORD(l,
2757 			    io_workitem, iw_listentry);
2758 			InitializeListHead((&iw->iw_listentry));
2759 			if (iw->iw_func == NULL)
2760 				continue;
2761 			KeReleaseSpinLock(&kq->kq_lock, irql);
2762 			MSCALL2(iw->iw_func, iw->iw_dobj, iw->iw_ctx);
2763 			KeAcquireSpinLock(&kq->kq_lock, &irql);
2764 		}
2765 
2766 		KeReleaseSpinLock(&kq->kq_lock, irql);
2767 	}
2768 
2769 	kproc_exit(0);
2770 	return; /* notreached */
2771 }
2772 
2773 static ndis_status
2774 RtlCharToInteger(src, base, val)
2775 	const char		*src;
2776 	uint32_t		base;
2777 	uint32_t		*val;
2778 {
2779 	int negative = 0;
2780 	uint32_t res;
2781 
2782 	if (!src || !val)
2783 		return (STATUS_ACCESS_VIOLATION);
2784 	while (*src != '\0' && *src <= ' ')
2785 		src++;
2786 	if (*src == '+')
2787 		src++;
2788 	else if (*src == '-') {
2789 		src++;
2790 		negative = 1;
2791 	}
2792 	if (base == 0) {
2793 		base = 10;
2794 		if (*src == '0') {
2795 			src++;
2796 			if (*src == 'b') {
2797 				base = 2;
2798 				src++;
2799 			} else if (*src == 'o') {
2800 				base = 8;
2801 				src++;
2802 			} else if (*src == 'x') {
2803 				base = 16;
2804 				src++;
2805 			}
2806 		}
2807 	} else if (!(base == 2 || base == 8 || base == 10 || base == 16))
2808 		return (STATUS_INVALID_PARAMETER);
2809 
2810 	for (res = 0; *src; src++) {
2811 		int v;
2812 		if (isdigit(*src))
2813 			v = *src - '0';
2814 		else if (isxdigit(*src))
2815 			v = tolower(*src) - 'a' + 10;
2816 		else
2817 			v = base;
2818 		if (v >= base)
2819 			return (STATUS_INVALID_PARAMETER);
2820 		res = res * base + v;
2821 	}
2822 	*val = negative ? -res : res;
2823 	return (STATUS_SUCCESS);
2824 }
2825 
2826 static void
2827 ntoskrnl_destroy_workitem_threads(void)
2828 {
2829 	kdpc_queue		*kq;
2830 	int			i;
2831 
2832 	for (i = 0; i < WORKITEM_THREADS; i++) {
2833 		kq = wq_queues + i;
2834 		kq->kq_exit = 1;
2835 		KeSetEvent(&kq->kq_proc, IO_NO_INCREMENT, FALSE);
2836 		while (kq->kq_exit)
2837 			tsleep(kq->kq_td->td_proc, PWAIT, "waitiw", hz/10);
2838 	}
2839 }
2840 
2841 io_workitem *
2842 IoAllocateWorkItem(dobj)
2843 	device_object		*dobj;
2844 {
2845 	io_workitem		*iw;
2846 
2847 	iw = uma_zalloc(iw_zone, M_NOWAIT);
2848 	if (iw == NULL)
2849 		return (NULL);
2850 
2851 	InitializeListHead(&iw->iw_listentry);
2852 	iw->iw_dobj = dobj;
2853 
2854 	mtx_lock(&ntoskrnl_dispatchlock);
2855 	iw->iw_idx = wq_idx;
2856 	WORKIDX_INC(wq_idx);
2857 	mtx_unlock(&ntoskrnl_dispatchlock);
2858 
2859 	return (iw);
2860 }
2861 
2862 void
2863 IoFreeWorkItem(iw)
2864 	io_workitem		*iw;
2865 {
2866 	uma_zfree(iw_zone, iw);
2867 }
2868 
2869 void
2870 IoQueueWorkItem(iw, iw_func, qtype, ctx)
2871 	io_workitem		*iw;
2872 	io_workitem_func	iw_func;
2873 	uint32_t		qtype;
2874 	void			*ctx;
2875 {
2876 	kdpc_queue		*kq;
2877 	list_entry		*l;
2878 	io_workitem		*cur;
2879 	uint8_t			irql;
2880 
2881 	kq = wq_queues + iw->iw_idx;
2882 
2883 	KeAcquireSpinLock(&kq->kq_lock, &irql);
2884 
2885 	/*
2886 	 * Traverse the list and make sure this workitem hasn't
2887 	 * already been inserted. Queuing the same workitem
2888 	 * twice will hose the list but good.
2889 	 */
2890 
2891 	l = kq->kq_disp.nle_flink;
2892 	while (l != &kq->kq_disp) {
2893 		cur = CONTAINING_RECORD(l, io_workitem, iw_listentry);
2894 		if (cur == iw) {
2895 			/* Already queued -- do nothing. */
2896 			KeReleaseSpinLock(&kq->kq_lock, irql);
2897 			return;
2898 		}
2899 		l = l->nle_flink;
2900 	}
2901 
2902 	iw->iw_func = iw_func;
2903 	iw->iw_ctx = ctx;
2904 
2905 	InsertTailList((&kq->kq_disp), (&iw->iw_listentry));
2906 	KeReleaseSpinLock(&kq->kq_lock, irql);
2907 
2908 	KeSetEvent(&kq->kq_proc, IO_NO_INCREMENT, FALSE);
2909 }
2910 
2911 static void
2912 ntoskrnl_workitem(dobj, arg)
2913 	device_object		*dobj;
2914 	void			*arg;
2915 {
2916 	io_workitem		*iw;
2917 	work_queue_item		*w;
2918 	work_item_func		f;
2919 
2920 	iw = arg;
2921 	w = (work_queue_item *)dobj;
2922 	f = (work_item_func)w->wqi_func;
2923 	uma_zfree(iw_zone, iw);
2924 	MSCALL2(f, w, w->wqi_ctx);
2925 }
2926 
2927 /*
2928  * The ExQueueWorkItem() API is deprecated in Windows XP. Microsoft
2929  * warns that it's unsafe and to use IoQueueWorkItem() instead. The
2930  * problem with ExQueueWorkItem() is that it can't guard against
2931  * the condition where a driver submits a job to the work queue and
2932  * is then unloaded before the job is able to run. IoQueueWorkItem()
2933  * acquires a reference to the device's device_object via the
2934  * object manager and retains it until after the job has completed,
2935  * which prevents the driver from being unloaded before the job
2936  * runs. (We don't currently support this behavior, though hopefully
2937  * that will change once the object manager API is fleshed out a bit.)
2938  *
2939  * Having said all that, the ExQueueWorkItem() API remains, because
2940  * there are still other parts of Windows that use it, including
2941  * NDIS itself: NdisScheduleWorkItem() calls ExQueueWorkItem().
2942  * We fake up the ExQueueWorkItem() API on top of our implementation
2943  * of IoQueueWorkItem(). Workitem thread #3 is reserved exclusively
2944  * for ExQueueWorkItem() jobs, and we pass a pointer to the work
2945  * queue item (provided by the caller) in to IoAllocateWorkItem()
2946  * instead of the device_object. We need to save this pointer so
2947  * we can apply a sanity check: as with the DPC queue and other
2948  * workitem queues, we can't allow the same work queue item to
2949  * be queued twice. If it's already pending, we silently return
2950  */
2951 
2952 void
2953 ExQueueWorkItem(w, qtype)
2954 	work_queue_item		*w;
2955 	uint32_t		qtype;
2956 {
2957 	io_workitem		*iw;
2958 	io_workitem_func	iwf;
2959 	kdpc_queue		*kq;
2960 	list_entry		*l;
2961 	io_workitem		*cur;
2962 	uint8_t			irql;
2963 
2964 	/*
2965 	 * We need to do a special sanity test to make sure
2966 	 * the ExQueueWorkItem() API isn't used to queue
2967 	 * the same workitem twice. Rather than checking the
2968 	 * io_workitem pointer itself, we test the attached
2969 	 * device object, which is really a pointer to the
2970 	 * legacy work queue item structure.
2971 	 */
2972 
2973 	kq = wq_queues + WORKITEM_LEGACY_THREAD;
2974 	KeAcquireSpinLock(&kq->kq_lock, &irql);
2975 	l = kq->kq_disp.nle_flink;
2976 	while (l != &kq->kq_disp) {
2977 		cur = CONTAINING_RECORD(l, io_workitem, iw_listentry);
2978 		if (cur->iw_dobj == (device_object *)w) {
2979 			/* Already queued -- do nothing. */
2980 			KeReleaseSpinLock(&kq->kq_lock, irql);
2981 			return;
2982 		}
2983 		l = l->nle_flink;
2984 	}
2985 	KeReleaseSpinLock(&kq->kq_lock, irql);
2986 
2987 	iw = IoAllocateWorkItem((device_object *)w);
2988 	if (iw == NULL)
2989 		return;
2990 
2991 	iw->iw_idx = WORKITEM_LEGACY_THREAD;
2992 	iwf = (io_workitem_func)ntoskrnl_findwrap((funcptr)ntoskrnl_workitem);
2993 	IoQueueWorkItem(iw, iwf, qtype, iw);
2994 }
2995 
2996 static void
2997 RtlZeroMemory(dst, len)
2998 	void			*dst;
2999 	size_t			len;
3000 {
3001 	bzero(dst, len);
3002 }
3003 
3004 static void
3005 RtlSecureZeroMemory(dst, len)
3006 	void			*dst;
3007 	size_t			len;
3008 {
3009 	memset(dst, 0, len);
3010 }
3011 
3012 static void
3013 RtlFillMemory(void *dst, size_t len, uint8_t c)
3014 {
3015 	memset(dst, c, len);
3016 }
3017 
3018 static void
3019 RtlMoveMemory(dst, src, len)
3020 	void			*dst;
3021 	const void		*src;
3022 	size_t			len;
3023 {
3024 	memmove(dst, src, len);
3025 }
3026 
3027 static void
3028 RtlCopyMemory(dst, src, len)
3029 	void			*dst;
3030 	const void		*src;
3031 	size_t			len;
3032 {
3033 	bcopy(src, dst, len);
3034 }
3035 
3036 static size_t
3037 RtlCompareMemory(s1, s2, len)
3038 	const void		*s1;
3039 	const void		*s2;
3040 	size_t			len;
3041 {
3042 	size_t			i;
3043 	uint8_t			*m1, *m2;
3044 
3045 	m1 = __DECONST(char *, s1);
3046 	m2 = __DECONST(char *, s2);
3047 
3048 	for (i = 0; i < len && m1[i] == m2[i]; i++);
3049 	return (i);
3050 }
3051 
3052 void
3053 RtlInitAnsiString(dst, src)
3054 	ansi_string		*dst;
3055 	char			*src;
3056 {
3057 	ansi_string		*a;
3058 
3059 	a = dst;
3060 	if (a == NULL)
3061 		return;
3062 	if (src == NULL) {
3063 		a->as_len = a->as_maxlen = 0;
3064 		a->as_buf = NULL;
3065 	} else {
3066 		a->as_buf = src;
3067 		a->as_len = a->as_maxlen = strlen(src);
3068 	}
3069 }
3070 
3071 void
3072 RtlInitUnicodeString(dst, src)
3073 	unicode_string		*dst;
3074 	uint16_t		*src;
3075 {
3076 	unicode_string		*u;
3077 	int			i;
3078 
3079 	u = dst;
3080 	if (u == NULL)
3081 		return;
3082 	if (src == NULL) {
3083 		u->us_len = u->us_maxlen = 0;
3084 		u->us_buf = NULL;
3085 	} else {
3086 		i = 0;
3087 		while(src[i] != 0)
3088 			i++;
3089 		u->us_buf = src;
3090 		u->us_len = u->us_maxlen = i * 2;
3091 	}
3092 }
3093 
3094 ndis_status
3095 RtlUnicodeStringToInteger(ustr, base, val)
3096 	unicode_string		*ustr;
3097 	uint32_t		base;
3098 	uint32_t		*val;
3099 {
3100 	uint16_t		*uchr;
3101 	int			len, neg = 0;
3102 	char			abuf[64];
3103 	char			*astr;
3104 
3105 	uchr = ustr->us_buf;
3106 	len = ustr->us_len;
3107 	bzero(abuf, sizeof(abuf));
3108 
3109 	if ((char)((*uchr) & 0xFF) == '-') {
3110 		neg = 1;
3111 		uchr++;
3112 		len -= 2;
3113 	} else if ((char)((*uchr) & 0xFF) == '+') {
3114 		neg = 0;
3115 		uchr++;
3116 		len -= 2;
3117 	}
3118 
3119 	if (base == 0) {
3120 		if ((char)((*uchr) & 0xFF) == 'b') {
3121 			base = 2;
3122 			uchr++;
3123 			len -= 2;
3124 		} else if ((char)((*uchr) & 0xFF) == 'o') {
3125 			base = 8;
3126 			uchr++;
3127 			len -= 2;
3128 		} else if ((char)((*uchr) & 0xFF) == 'x') {
3129 			base = 16;
3130 			uchr++;
3131 			len -= 2;
3132 		} else
3133 			base = 10;
3134 	}
3135 
3136 	astr = abuf;
3137 	if (neg) {
3138 		strcpy(astr, "-");
3139 		astr++;
3140 	}
3141 
3142 	ntoskrnl_unicode_to_ascii(uchr, astr, len);
3143 	*val = strtoul(abuf, NULL, base);
3144 
3145 	return (STATUS_SUCCESS);
3146 }
3147 
3148 void
3149 RtlFreeUnicodeString(ustr)
3150 	unicode_string		*ustr;
3151 {
3152 	if (ustr->us_buf == NULL)
3153 		return;
3154 	ExFreePool(ustr->us_buf);
3155 	ustr->us_buf = NULL;
3156 }
3157 
3158 void
3159 RtlFreeAnsiString(astr)
3160 	ansi_string		*astr;
3161 {
3162 	if (astr->as_buf == NULL)
3163 		return;
3164 	ExFreePool(astr->as_buf);
3165 	astr->as_buf = NULL;
3166 }
3167 
3168 static int
3169 atoi(str)
3170 	const char		*str;
3171 {
3172 	return (int)strtol(str, (char **)NULL, 10);
3173 }
3174 
3175 static long
3176 atol(str)
3177 	const char		*str;
3178 {
3179 	return strtol(str, (char **)NULL, 10);
3180 }
3181 
3182 static int
3183 rand(void)
3184 {
3185 
3186 	return (random());
3187 }
3188 
3189 static void
3190 srand(unsigned int seed __unused)
3191 {
3192 }
3193 
3194 static uint8_t
3195 IoIsWdmVersionAvailable(uint8_t major, uint8_t minor)
3196 {
3197 	if (major == WDM_MAJOR && minor == WDM_MINOR_WINXP)
3198 		return (TRUE);
3199 	return (FALSE);
3200 }
3201 
3202 static int32_t
3203 IoOpenDeviceRegistryKey(struct device_object *devobj, uint32_t type,
3204     uint32_t mask, void **key)
3205 {
3206 	return (NDIS_STATUS_INVALID_DEVICE_REQUEST);
3207 }
3208 
3209 static ndis_status
3210 IoGetDeviceObjectPointer(name, reqaccess, fileobj, devobj)
3211 	unicode_string		*name;
3212 	uint32_t		reqaccess;
3213 	void			*fileobj;
3214 	device_object		*devobj;
3215 {
3216 	return (STATUS_SUCCESS);
3217 }
3218 
3219 static ndis_status
3220 IoGetDeviceProperty(devobj, regprop, buflen, prop, reslen)
3221 	device_object		*devobj;
3222 	uint32_t		regprop;
3223 	uint32_t		buflen;
3224 	void			*prop;
3225 	uint32_t		*reslen;
3226 {
3227 	driver_object		*drv;
3228 	uint16_t		**name;
3229 
3230 	drv = devobj->do_drvobj;
3231 
3232 	switch (regprop) {
3233 	case DEVPROP_DRIVER_KEYNAME:
3234 		name = prop;
3235 		*name = drv->dro_drivername.us_buf;
3236 		*reslen = drv->dro_drivername.us_len;
3237 		break;
3238 	default:
3239 		return (STATUS_INVALID_PARAMETER_2);
3240 		break;
3241 	}
3242 
3243 	return (STATUS_SUCCESS);
3244 }
3245 
3246 static void
3247 KeInitializeMutex(kmutex, level)
3248 	kmutant			*kmutex;
3249 	uint32_t		level;
3250 {
3251 	InitializeListHead((&kmutex->km_header.dh_waitlisthead));
3252 	kmutex->km_abandoned = FALSE;
3253 	kmutex->km_apcdisable = 1;
3254 	kmutex->km_header.dh_sigstate = 1;
3255 	kmutex->km_header.dh_type = DISP_TYPE_MUTANT;
3256 	kmutex->km_header.dh_size = sizeof(kmutant) / sizeof(uint32_t);
3257 	kmutex->km_ownerthread = NULL;
3258 }
3259 
3260 static uint32_t
3261 KeReleaseMutex(kmutant *kmutex, uint8_t kwait)
3262 {
3263 	uint32_t		prevstate;
3264 
3265 	mtx_lock(&ntoskrnl_dispatchlock);
3266 	prevstate = kmutex->km_header.dh_sigstate;
3267 	if (kmutex->km_ownerthread != curthread) {
3268 		mtx_unlock(&ntoskrnl_dispatchlock);
3269 		return (STATUS_MUTANT_NOT_OWNED);
3270 	}
3271 
3272 	kmutex->km_header.dh_sigstate++;
3273 	kmutex->km_abandoned = FALSE;
3274 
3275 	if (kmutex->km_header.dh_sigstate == 1) {
3276 		kmutex->km_ownerthread = NULL;
3277 		ntoskrnl_waittest(&kmutex->km_header, IO_NO_INCREMENT);
3278 	}
3279 
3280 	mtx_unlock(&ntoskrnl_dispatchlock);
3281 
3282 	return (prevstate);
3283 }
3284 
3285 static uint32_t
3286 KeReadStateMutex(kmutex)
3287 	kmutant			*kmutex;
3288 {
3289 	return (kmutex->km_header.dh_sigstate);
3290 }
3291 
3292 void
3293 KeInitializeEvent(nt_kevent *kevent, uint32_t type, uint8_t state)
3294 {
3295 	InitializeListHead((&kevent->k_header.dh_waitlisthead));
3296 	kevent->k_header.dh_sigstate = state;
3297 	if (type == EVENT_TYPE_NOTIFY)
3298 		kevent->k_header.dh_type = DISP_TYPE_NOTIFICATION_EVENT;
3299 	else
3300 		kevent->k_header.dh_type = DISP_TYPE_SYNCHRONIZATION_EVENT;
3301 	kevent->k_header.dh_size = sizeof(nt_kevent) / sizeof(uint32_t);
3302 }
3303 
3304 uint32_t
3305 KeResetEvent(kevent)
3306 	nt_kevent		*kevent;
3307 {
3308 	uint32_t		prevstate;
3309 
3310 	mtx_lock(&ntoskrnl_dispatchlock);
3311 	prevstate = kevent->k_header.dh_sigstate;
3312 	kevent->k_header.dh_sigstate = FALSE;
3313 	mtx_unlock(&ntoskrnl_dispatchlock);
3314 
3315 	return (prevstate);
3316 }
3317 
3318 uint32_t
3319 KeSetEvent(nt_kevent *kevent, uint32_t increment, uint8_t kwait)
3320 {
3321 	uint32_t		prevstate;
3322 	wait_block		*w;
3323 	nt_dispatch_header	*dh;
3324 	struct thread		*td;
3325 	wb_ext			*we;
3326 
3327 	mtx_lock(&ntoskrnl_dispatchlock);
3328 	prevstate = kevent->k_header.dh_sigstate;
3329 	dh = &kevent->k_header;
3330 
3331 	if (IsListEmpty(&dh->dh_waitlisthead))
3332 		/*
3333 		 * If there's nobody in the waitlist, just set
3334 		 * the state to signalled.
3335 		 */
3336 		dh->dh_sigstate = 1;
3337 	else {
3338 		/*
3339 		 * Get the first waiter. If this is a synchronization
3340 		 * event, just wake up that one thread (don't bother
3341 		 * setting the state to signalled since we're supposed
3342 		 * to automatically clear synchronization events anyway).
3343 		 *
3344 		 * If it's a notification event, or the first
3345 		 * waiter is doing a WAITTYPE_ALL wait, go through
3346 		 * the full wait satisfaction process.
3347 		 */
3348 		w = CONTAINING_RECORD(dh->dh_waitlisthead.nle_flink,
3349 		    wait_block, wb_waitlist);
3350 		we = w->wb_ext;
3351 		td = we->we_td;
3352 		if (kevent->k_header.dh_type == DISP_TYPE_NOTIFICATION_EVENT ||
3353 		    w->wb_waittype == WAITTYPE_ALL) {
3354 			if (prevstate == 0) {
3355 				dh->dh_sigstate = 1;
3356 				ntoskrnl_waittest(dh, increment);
3357 			}
3358 		} else {
3359 			w->wb_awakened |= TRUE;
3360 			cv_broadcastpri(&we->we_cv,
3361 			    (w->wb_oldpri - (increment * 4)) > PRI_MIN_KERN ?
3362 			    w->wb_oldpri - (increment * 4) : PRI_MIN_KERN);
3363 		}
3364 	}
3365 
3366 	mtx_unlock(&ntoskrnl_dispatchlock);
3367 
3368 	return (prevstate);
3369 }
3370 
3371 void
3372 KeClearEvent(kevent)
3373 	nt_kevent		*kevent;
3374 {
3375 	kevent->k_header.dh_sigstate = FALSE;
3376 }
3377 
3378 uint32_t
3379 KeReadStateEvent(kevent)
3380 	nt_kevent		*kevent;
3381 {
3382 	return (kevent->k_header.dh_sigstate);
3383 }
3384 
3385 /*
3386  * The object manager in Windows is responsible for managing
3387  * references and access to various types of objects, including
3388  * device_objects, events, threads, timers and so on. However,
3389  * there's a difference in the way objects are handled in user
3390  * mode versus kernel mode.
3391  *
3392  * In user mode (i.e. Win32 applications), all objects are
3393  * managed by the object manager. For example, when you create
3394  * a timer or event object, you actually end up with an
3395  * object_header (for the object manager's bookkeeping
3396  * purposes) and an object body (which contains the actual object
3397  * structure, e.g. ktimer, kevent, etc...). This allows Windows
3398  * to manage resource quotas and to enforce access restrictions
3399  * on basically every kind of system object handled by the kernel.
3400  *
3401  * However, in kernel mode, you only end up using the object
3402  * manager some of the time. For example, in a driver, you create
3403  * a timer object by simply allocating the memory for a ktimer
3404  * structure and initializing it with KeInitializeTimer(). Hence,
3405  * the timer has no object_header and no reference counting or
3406  * security/resource checks are done on it. The assumption in
3407  * this case is that if you're running in kernel mode, you know
3408  * what you're doing, and you're already at an elevated privilege
3409  * anyway.
3410  *
3411  * There are some exceptions to this. The two most important ones
3412  * for our purposes are device_objects and threads. We need to use
3413  * the object manager to do reference counting on device_objects,
3414  * and for threads, you can only get a pointer to a thread's
3415  * dispatch header by using ObReferenceObjectByHandle() on the
3416  * handle returned by PsCreateSystemThread().
3417  */
3418 
3419 static ndis_status
3420 ObReferenceObjectByHandle(ndis_handle handle, uint32_t reqaccess, void *otype,
3421 	uint8_t accessmode, void **object, void **handleinfo)
3422 {
3423 	nt_objref		*nr;
3424 
3425 	nr = malloc(sizeof(nt_objref), M_DEVBUF, M_NOWAIT|M_ZERO);
3426 	if (nr == NULL)
3427 		return (STATUS_INSUFFICIENT_RESOURCES);
3428 
3429 	InitializeListHead((&nr->no_dh.dh_waitlisthead));
3430 	nr->no_obj = handle;
3431 	nr->no_dh.dh_type = DISP_TYPE_THREAD;
3432 	nr->no_dh.dh_sigstate = 0;
3433 	nr->no_dh.dh_size = (uint8_t)(sizeof(struct thread) /
3434 	    sizeof(uint32_t));
3435 	TAILQ_INSERT_TAIL(&ntoskrnl_reflist, nr, link);
3436 	*object = nr;
3437 
3438 	return (STATUS_SUCCESS);
3439 }
3440 
3441 static void
3442 ObfDereferenceObject(object)
3443 	void			*object;
3444 {
3445 	nt_objref		*nr;
3446 
3447 	nr = object;
3448 	TAILQ_REMOVE(&ntoskrnl_reflist, nr, link);
3449 	free(nr, M_DEVBUF);
3450 }
3451 
3452 static uint32_t
3453 ZwClose(handle)
3454 	ndis_handle		handle;
3455 {
3456 	return (STATUS_SUCCESS);
3457 }
3458 
3459 static uint32_t
3460 WmiQueryTraceInformation(traceclass, traceinfo, infolen, reqlen, buf)
3461 	uint32_t		traceclass;
3462 	void			*traceinfo;
3463 	uint32_t		infolen;
3464 	uint32_t		reqlen;
3465 	void			*buf;
3466 {
3467 	return (STATUS_NOT_FOUND);
3468 }
3469 
3470 static uint32_t
3471 WmiTraceMessage(uint64_t loghandle, uint32_t messageflags,
3472 	void *guid, uint16_t messagenum, ...)
3473 {
3474 	return (STATUS_SUCCESS);
3475 }
3476 
3477 static uint32_t
3478 IoWMIRegistrationControl(dobj, action)
3479 	device_object		*dobj;
3480 	uint32_t		action;
3481 {
3482 	return (STATUS_SUCCESS);
3483 }
3484 
3485 /*
3486  * This is here just in case the thread returns without calling
3487  * PsTerminateSystemThread().
3488  */
3489 static void
3490 ntoskrnl_thrfunc(arg)
3491 	void			*arg;
3492 {
3493 	thread_context		*thrctx;
3494 	uint32_t (*tfunc)(void *);
3495 	void			*tctx;
3496 	uint32_t		rval;
3497 
3498 	thrctx = arg;
3499 	tfunc = thrctx->tc_thrfunc;
3500 	tctx = thrctx->tc_thrctx;
3501 	free(thrctx, M_TEMP);
3502 
3503 	rval = MSCALL1(tfunc, tctx);
3504 
3505 	PsTerminateSystemThread(rval);
3506 	return; /* notreached */
3507 }
3508 
3509 static ndis_status
3510 PsCreateSystemThread(handle, reqaccess, objattrs, phandle,
3511 	clientid, thrfunc, thrctx)
3512 	ndis_handle		*handle;
3513 	uint32_t		reqaccess;
3514 	void			*objattrs;
3515 	ndis_handle		phandle;
3516 	void			*clientid;
3517 	void			*thrfunc;
3518 	void			*thrctx;
3519 {
3520 	int			error;
3521 	thread_context		*tc;
3522 	struct proc		*p;
3523 
3524 	tc = malloc(sizeof(thread_context), M_TEMP, M_NOWAIT);
3525 	if (tc == NULL)
3526 		return (STATUS_INSUFFICIENT_RESOURCES);
3527 
3528 	tc->tc_thrctx = thrctx;
3529 	tc->tc_thrfunc = thrfunc;
3530 
3531 	error = kproc_create(ntoskrnl_thrfunc, tc, &p,
3532 	    RFHIGHPID, NDIS_KSTACK_PAGES, "Windows Kthread %d", ntoskrnl_kth);
3533 
3534 	if (error) {
3535 		free(tc, M_TEMP);
3536 		return (STATUS_INSUFFICIENT_RESOURCES);
3537 	}
3538 
3539 	*handle = p;
3540 	ntoskrnl_kth++;
3541 
3542 	return (STATUS_SUCCESS);
3543 }
3544 
3545 /*
3546  * In Windows, the exit of a thread is an event that you're allowed
3547  * to wait on, assuming you've obtained a reference to the thread using
3548  * ObReferenceObjectByHandle(). Unfortunately, the only way we can
3549  * simulate this behavior is to register each thread we create in a
3550  * reference list, and if someone holds a reference to us, we poke
3551  * them.
3552  */
3553 static ndis_status
3554 PsTerminateSystemThread(status)
3555 	ndis_status		status;
3556 {
3557 	struct nt_objref	*nr;
3558 
3559 	mtx_lock(&ntoskrnl_dispatchlock);
3560 	TAILQ_FOREACH(nr, &ntoskrnl_reflist, link) {
3561 		if (nr->no_obj != curthread->td_proc)
3562 			continue;
3563 		nr->no_dh.dh_sigstate = 1;
3564 		ntoskrnl_waittest(&nr->no_dh, IO_NO_INCREMENT);
3565 		break;
3566 	}
3567 	mtx_unlock(&ntoskrnl_dispatchlock);
3568 
3569 	ntoskrnl_kth--;
3570 
3571 	kproc_exit(0);
3572 	return (0);	/* notreached */
3573 }
3574 
3575 static uint32_t
3576 DbgPrint(char *fmt, ...)
3577 {
3578 	va_list			ap;
3579 
3580 	if (bootverbose) {
3581 		va_start(ap, fmt);
3582 		vprintf(fmt, ap);
3583 		va_end(ap);
3584 	}
3585 
3586 	return (STATUS_SUCCESS);
3587 }
3588 
3589 static void
3590 DbgBreakPoint(void)
3591 {
3592 
3593 	kdb_enter(KDB_WHY_NDIS, "DbgBreakPoint(): breakpoint");
3594 }
3595 
3596 static void
3597 KeBugCheckEx(code, param1, param2, param3, param4)
3598     uint32_t			code;
3599     u_long			param1;
3600     u_long			param2;
3601     u_long			param3;
3602     u_long			param4;
3603 {
3604 	panic("KeBugCheckEx: STOP 0x%X", code);
3605 }
3606 
3607 static void
3608 ntoskrnl_timercall(arg)
3609 	void			*arg;
3610 {
3611 	ktimer			*timer;
3612 	struct timeval		tv;
3613 	kdpc			*dpc;
3614 
3615 	mtx_lock(&ntoskrnl_dispatchlock);
3616 
3617 	timer = arg;
3618 
3619 #ifdef NTOSKRNL_DEBUG_TIMERS
3620 	ntoskrnl_timer_fires++;
3621 #endif
3622 	ntoskrnl_remove_timer(timer);
3623 
3624 	/*
3625 	 * This should never happen, but complain
3626 	 * if it does.
3627 	 */
3628 
3629 	if (timer->k_header.dh_inserted == FALSE) {
3630 		mtx_unlock(&ntoskrnl_dispatchlock);
3631 		printf("NTOS: timer %p fired even though "
3632 		    "it was canceled\n", timer);
3633 		return;
3634 	}
3635 
3636 	/* Mark the timer as no longer being on the timer queue. */
3637 
3638 	timer->k_header.dh_inserted = FALSE;
3639 
3640 	/* Now signal the object and satisfy any waits on it. */
3641 
3642 	timer->k_header.dh_sigstate = 1;
3643 	ntoskrnl_waittest(&timer->k_header, IO_NO_INCREMENT);
3644 
3645 	/*
3646 	 * If this is a periodic timer, re-arm it
3647 	 * so it will fire again. We do this before
3648 	 * calling any deferred procedure calls because
3649 	 * it's possible the DPC might cancel the timer,
3650 	 * in which case it would be wrong for us to
3651 	 * re-arm it again afterwards.
3652 	 */
3653 
3654 	if (timer->k_period) {
3655 		tv.tv_sec = 0;
3656 		tv.tv_usec = timer->k_period * 1000;
3657 		timer->k_header.dh_inserted = TRUE;
3658 		ntoskrnl_insert_timer(timer, tvtohz(&tv));
3659 #ifdef NTOSKRNL_DEBUG_TIMERS
3660 		ntoskrnl_timer_reloads++;
3661 #endif
3662 	}
3663 
3664 	dpc = timer->k_dpc;
3665 
3666 	mtx_unlock(&ntoskrnl_dispatchlock);
3667 
3668 	/* If there's a DPC associated with the timer, queue it up. */
3669 
3670 	if (dpc != NULL)
3671 		KeInsertQueueDpc(dpc, NULL, NULL);
3672 }
3673 
3674 #ifdef NTOSKRNL_DEBUG_TIMERS
3675 static int
3676 sysctl_show_timers(SYSCTL_HANDLER_ARGS)
3677 {
3678 	int			ret;
3679 
3680 	ret = 0;
3681 	ntoskrnl_show_timers();
3682 	return (sysctl_handle_int(oidp, &ret, 0, req));
3683 }
3684 
3685 static void
3686 ntoskrnl_show_timers()
3687 {
3688 	int			i = 0;
3689 	list_entry		*l;
3690 
3691 	mtx_lock_spin(&ntoskrnl_calllock);
3692 	l = ntoskrnl_calllist.nle_flink;
3693 	while(l != &ntoskrnl_calllist) {
3694 		i++;
3695 		l = l->nle_flink;
3696 	}
3697 	mtx_unlock_spin(&ntoskrnl_calllock);
3698 
3699 	printf("\n");
3700 	printf("%d timers available (out of %d)\n", i, NTOSKRNL_TIMEOUTS);
3701 	printf("timer sets: %qu\n", ntoskrnl_timer_sets);
3702 	printf("timer reloads: %qu\n", ntoskrnl_timer_reloads);
3703 	printf("timer cancels: %qu\n", ntoskrnl_timer_cancels);
3704 	printf("timer fires: %qu\n", ntoskrnl_timer_fires);
3705 	printf("\n");
3706 }
3707 #endif
3708 
3709 /*
3710  * Must be called with dispatcher lock held.
3711  */
3712 
3713 static void
3714 ntoskrnl_insert_timer(timer, ticks)
3715 	ktimer			*timer;
3716 	int			ticks;
3717 {
3718 	callout_entry		*e;
3719 	list_entry		*l;
3720 	struct callout		*c;
3721 
3722 	/*
3723 	 * Try and allocate a timer.
3724 	 */
3725 	mtx_lock_spin(&ntoskrnl_calllock);
3726 	if (IsListEmpty(&ntoskrnl_calllist)) {
3727 		mtx_unlock_spin(&ntoskrnl_calllock);
3728 #ifdef NTOSKRNL_DEBUG_TIMERS
3729 		ntoskrnl_show_timers();
3730 #endif
3731 		panic("out of timers!");
3732 	}
3733 	l = RemoveHeadList(&ntoskrnl_calllist);
3734 	mtx_unlock_spin(&ntoskrnl_calllock);
3735 
3736 	e = CONTAINING_RECORD(l, callout_entry, ce_list);
3737 	c = &e->ce_callout;
3738 
3739 	timer->k_callout = c;
3740 
3741 	callout_init(c, 1);
3742 	callout_reset(c, ticks, ntoskrnl_timercall, timer);
3743 }
3744 
3745 static void
3746 ntoskrnl_remove_timer(timer)
3747 	ktimer			*timer;
3748 {
3749 	callout_entry		*e;
3750 
3751 	e = (callout_entry *)timer->k_callout;
3752 	callout_stop(timer->k_callout);
3753 
3754 	mtx_lock_spin(&ntoskrnl_calllock);
3755 	InsertHeadList((&ntoskrnl_calllist), (&e->ce_list));
3756 	mtx_unlock_spin(&ntoskrnl_calllock);
3757 }
3758 
3759 void
3760 KeInitializeTimer(timer)
3761 	ktimer			*timer;
3762 {
3763 	if (timer == NULL)
3764 		return;
3765 
3766 	KeInitializeTimerEx(timer,  EVENT_TYPE_NOTIFY);
3767 }
3768 
3769 void
3770 KeInitializeTimerEx(timer, type)
3771 	ktimer			*timer;
3772 	uint32_t		type;
3773 {
3774 	if (timer == NULL)
3775 		return;
3776 
3777 	bzero((char *)timer, sizeof(ktimer));
3778 	InitializeListHead((&timer->k_header.dh_waitlisthead));
3779 	timer->k_header.dh_sigstate = FALSE;
3780 	timer->k_header.dh_inserted = FALSE;
3781 	if (type == EVENT_TYPE_NOTIFY)
3782 		timer->k_header.dh_type = DISP_TYPE_NOTIFICATION_TIMER;
3783 	else
3784 		timer->k_header.dh_type = DISP_TYPE_SYNCHRONIZATION_TIMER;
3785 	timer->k_header.dh_size = sizeof(ktimer) / sizeof(uint32_t);
3786 }
3787 
3788 /*
3789  * DPC subsystem. A Windows Defered Procedure Call has the following
3790  * properties:
3791  * - It runs at DISPATCH_LEVEL.
3792  * - It can have one of 3 importance values that control when it
3793  *   runs relative to other DPCs in the queue.
3794  * - On SMP systems, it can be set to run on a specific processor.
3795  * In order to satisfy the last property, we create a DPC thread for
3796  * each CPU in the system and bind it to that CPU. Each thread
3797  * maintains three queues with different importance levels, which
3798  * will be processed in order from lowest to highest.
3799  *
3800  * In Windows, interrupt handlers run as DPCs. (Not to be confused
3801  * with ISRs, which run in interrupt context and can preempt DPCs.)
3802  * ISRs are given the highest importance so that they'll take
3803  * precedence over timers and other things.
3804  */
3805 
3806 static void
3807 ntoskrnl_dpc_thread(arg)
3808 	void			*arg;
3809 {
3810 	kdpc_queue		*kq;
3811 	kdpc			*d;
3812 	list_entry		*l;
3813 	uint8_t			irql;
3814 
3815 	kq = arg;
3816 
3817 	InitializeListHead(&kq->kq_disp);
3818 	kq->kq_td = curthread;
3819 	kq->kq_exit = 0;
3820 	kq->kq_running = FALSE;
3821 	KeInitializeSpinLock(&kq->kq_lock);
3822 	KeInitializeEvent(&kq->kq_proc, EVENT_TYPE_SYNC, FALSE);
3823 	KeInitializeEvent(&kq->kq_done, EVENT_TYPE_SYNC, FALSE);
3824 
3825 	/*
3826 	 * Elevate our priority. DPCs are used to run interrupt
3827 	 * handlers, and they should trigger as soon as possible
3828 	 * once scheduled by an ISR.
3829 	 */
3830 
3831 	thread_lock(curthread);
3832 #ifdef NTOSKRNL_MULTIPLE_DPCS
3833 	sched_bind(curthread, kq->kq_cpu);
3834 #endif
3835 	sched_prio(curthread, PRI_MIN_KERN);
3836 	thread_unlock(curthread);
3837 
3838 	while (1) {
3839 		KeWaitForSingleObject(&kq->kq_proc, 0, 0, TRUE, NULL);
3840 
3841 		KeAcquireSpinLock(&kq->kq_lock, &irql);
3842 
3843 		if (kq->kq_exit) {
3844 			kq->kq_exit = 0;
3845 			KeReleaseSpinLock(&kq->kq_lock, irql);
3846 			break;
3847 		}
3848 
3849 		kq->kq_running = TRUE;
3850 
3851 		while (!IsListEmpty(&kq->kq_disp)) {
3852 			l = RemoveHeadList((&kq->kq_disp));
3853 			d = CONTAINING_RECORD(l, kdpc, k_dpclistentry);
3854 			InitializeListHead((&d->k_dpclistentry));
3855 			KeReleaseSpinLockFromDpcLevel(&kq->kq_lock);
3856 			MSCALL4(d->k_deferedfunc, d, d->k_deferredctx,
3857 			    d->k_sysarg1, d->k_sysarg2);
3858 			KeAcquireSpinLockAtDpcLevel(&kq->kq_lock);
3859 		}
3860 
3861 		kq->kq_running = FALSE;
3862 
3863 		KeReleaseSpinLock(&kq->kq_lock, irql);
3864 
3865 		KeSetEvent(&kq->kq_done, IO_NO_INCREMENT, FALSE);
3866 	}
3867 
3868 	kproc_exit(0);
3869 	return; /* notreached */
3870 }
3871 
3872 static void
3873 ntoskrnl_destroy_dpc_threads(void)
3874 {
3875 	kdpc_queue		*kq;
3876 	kdpc			dpc;
3877 	int			i;
3878 
3879 	kq = kq_queues;
3880 #ifdef NTOSKRNL_MULTIPLE_DPCS
3881 	for (i = 0; i < mp_ncpus; i++) {
3882 #else
3883 	for (i = 0; i < 1; i++) {
3884 #endif
3885 		kq += i;
3886 
3887 		kq->kq_exit = 1;
3888 		KeInitializeDpc(&dpc, NULL, NULL);
3889 		KeSetTargetProcessorDpc(&dpc, i);
3890 		KeInsertQueueDpc(&dpc, NULL, NULL);
3891 		while (kq->kq_exit)
3892 			tsleep(kq->kq_td->td_proc, PWAIT, "dpcw", hz/10);
3893 	}
3894 }
3895 
3896 static uint8_t
3897 ntoskrnl_insert_dpc(head, dpc)
3898 	list_entry		*head;
3899 	kdpc			*dpc;
3900 {
3901 	list_entry		*l;
3902 	kdpc			*d;
3903 
3904 	l = head->nle_flink;
3905 	while (l != head) {
3906 		d = CONTAINING_RECORD(l, kdpc, k_dpclistentry);
3907 		if (d == dpc)
3908 			return (FALSE);
3909 		l = l->nle_flink;
3910 	}
3911 
3912 	if (dpc->k_importance == KDPC_IMPORTANCE_LOW)
3913 		InsertTailList((head), (&dpc->k_dpclistentry));
3914 	else
3915 		InsertHeadList((head), (&dpc->k_dpclistentry));
3916 
3917 	return (TRUE);
3918 }
3919 
3920 void
3921 KeInitializeDpc(dpc, dpcfunc, dpcctx)
3922 	kdpc			*dpc;
3923 	void			*dpcfunc;
3924 	void			*dpcctx;
3925 {
3926 
3927 	if (dpc == NULL)
3928 		return;
3929 
3930 	dpc->k_deferedfunc = dpcfunc;
3931 	dpc->k_deferredctx = dpcctx;
3932 	dpc->k_num = KDPC_CPU_DEFAULT;
3933 	dpc->k_importance = KDPC_IMPORTANCE_MEDIUM;
3934 	InitializeListHead((&dpc->k_dpclistentry));
3935 }
3936 
3937 uint8_t
3938 KeInsertQueueDpc(dpc, sysarg1, sysarg2)
3939 	kdpc			*dpc;
3940 	void			*sysarg1;
3941 	void			*sysarg2;
3942 {
3943 	kdpc_queue		*kq;
3944 	uint8_t			r;
3945 	uint8_t			irql;
3946 
3947 	if (dpc == NULL)
3948 		return (FALSE);
3949 
3950 	kq = kq_queues;
3951 
3952 #ifdef NTOSKRNL_MULTIPLE_DPCS
3953 	KeRaiseIrql(DISPATCH_LEVEL, &irql);
3954 
3955 	/*
3956 	 * By default, the DPC is queued to run on the same CPU
3957 	 * that scheduled it.
3958 	 */
3959 
3960 	if (dpc->k_num == KDPC_CPU_DEFAULT)
3961 		kq += curthread->td_oncpu;
3962 	else
3963 		kq += dpc->k_num;
3964 	KeAcquireSpinLockAtDpcLevel(&kq->kq_lock);
3965 #else
3966 	KeAcquireSpinLock(&kq->kq_lock, &irql);
3967 #endif
3968 
3969 	r = ntoskrnl_insert_dpc(&kq->kq_disp, dpc);
3970 	if (r == TRUE) {
3971 		dpc->k_sysarg1 = sysarg1;
3972 		dpc->k_sysarg2 = sysarg2;
3973 	}
3974 	KeReleaseSpinLock(&kq->kq_lock, irql);
3975 
3976 	if (r == FALSE)
3977 		return (r);
3978 
3979 	KeSetEvent(&kq->kq_proc, IO_NO_INCREMENT, FALSE);
3980 
3981 	return (r);
3982 }
3983 
3984 uint8_t
3985 KeRemoveQueueDpc(dpc)
3986 	kdpc			*dpc;
3987 {
3988 	kdpc_queue		*kq;
3989 	uint8_t			irql;
3990 
3991 	if (dpc == NULL)
3992 		return (FALSE);
3993 
3994 #ifdef NTOSKRNL_MULTIPLE_DPCS
3995 	KeRaiseIrql(DISPATCH_LEVEL, &irql);
3996 
3997 	kq = kq_queues + dpc->k_num;
3998 
3999 	KeAcquireSpinLockAtDpcLevel(&kq->kq_lock);
4000 #else
4001 	kq = kq_queues;
4002 	KeAcquireSpinLock(&kq->kq_lock, &irql);
4003 #endif
4004 
4005 	if (dpc->k_dpclistentry.nle_flink == &dpc->k_dpclistentry) {
4006 		KeReleaseSpinLockFromDpcLevel(&kq->kq_lock);
4007 		KeLowerIrql(irql);
4008 		return (FALSE);
4009 	}
4010 
4011 	RemoveEntryList((&dpc->k_dpclistentry));
4012 	InitializeListHead((&dpc->k_dpclistentry));
4013 
4014 	KeReleaseSpinLock(&kq->kq_lock, irql);
4015 
4016 	return (TRUE);
4017 }
4018 
4019 void
4020 KeSetImportanceDpc(dpc, imp)
4021 	kdpc			*dpc;
4022 	uint32_t		imp;
4023 {
4024 	if (imp != KDPC_IMPORTANCE_LOW &&
4025 	    imp != KDPC_IMPORTANCE_MEDIUM &&
4026 	    imp != KDPC_IMPORTANCE_HIGH)
4027 		return;
4028 
4029 	dpc->k_importance = (uint8_t)imp;
4030 }
4031 
4032 void
4033 KeSetTargetProcessorDpc(kdpc *dpc, uint8_t cpu)
4034 {
4035 	if (cpu > mp_ncpus)
4036 		return;
4037 
4038 	dpc->k_num = cpu;
4039 }
4040 
4041 void
4042 KeFlushQueuedDpcs(void)
4043 {
4044 	kdpc_queue		*kq;
4045 	int			i;
4046 
4047 	/*
4048 	 * Poke each DPC queue and wait
4049 	 * for them to drain.
4050 	 */
4051 
4052 #ifdef NTOSKRNL_MULTIPLE_DPCS
4053 	for (i = 0; i < mp_ncpus; i++) {
4054 #else
4055 	for (i = 0; i < 1; i++) {
4056 #endif
4057 		kq = kq_queues + i;
4058 		KeSetEvent(&kq->kq_proc, IO_NO_INCREMENT, FALSE);
4059 		KeWaitForSingleObject(&kq->kq_done, 0, 0, TRUE, NULL);
4060 	}
4061 }
4062 
4063 uint32_t
4064 KeGetCurrentProcessorNumber(void)
4065 {
4066 	return ((uint32_t)curthread->td_oncpu);
4067 }
4068 
4069 uint8_t
4070 KeSetTimerEx(timer, duetime, period, dpc)
4071 	ktimer			*timer;
4072 	int64_t			duetime;
4073 	uint32_t		period;
4074 	kdpc			*dpc;
4075 {
4076 	struct timeval		tv;
4077 	uint64_t		curtime;
4078 	uint8_t			pending;
4079 
4080 	if (timer == NULL)
4081 		return (FALSE);
4082 
4083 	mtx_lock(&ntoskrnl_dispatchlock);
4084 
4085 	if (timer->k_header.dh_inserted == TRUE) {
4086 		ntoskrnl_remove_timer(timer);
4087 #ifdef NTOSKRNL_DEBUG_TIMERS
4088 		ntoskrnl_timer_cancels++;
4089 #endif
4090 		timer->k_header.dh_inserted = FALSE;
4091 		pending = TRUE;
4092 	} else
4093 		pending = FALSE;
4094 
4095 	timer->k_duetime = duetime;
4096 	timer->k_period = period;
4097 	timer->k_header.dh_sigstate = FALSE;
4098 	timer->k_dpc = dpc;
4099 
4100 	if (duetime < 0) {
4101 		tv.tv_sec = - (duetime) / 10000000;
4102 		tv.tv_usec = (- (duetime) / 10) -
4103 		    (tv.tv_sec * 1000000);
4104 	} else {
4105 		ntoskrnl_time(&curtime);
4106 		if (duetime < curtime)
4107 			tv.tv_sec = tv.tv_usec = 0;
4108 		else {
4109 			tv.tv_sec = ((duetime) - curtime) / 10000000;
4110 			tv.tv_usec = ((duetime) - curtime) / 10 -
4111 			    (tv.tv_sec * 1000000);
4112 		}
4113 	}
4114 
4115 	timer->k_header.dh_inserted = TRUE;
4116 	ntoskrnl_insert_timer(timer, tvtohz(&tv));
4117 #ifdef NTOSKRNL_DEBUG_TIMERS
4118 	ntoskrnl_timer_sets++;
4119 #endif
4120 
4121 	mtx_unlock(&ntoskrnl_dispatchlock);
4122 
4123 	return (pending);
4124 }
4125 
4126 uint8_t
4127 KeSetTimer(timer, duetime, dpc)
4128 	ktimer			*timer;
4129 	int64_t			duetime;
4130 	kdpc			*dpc;
4131 {
4132 	return (KeSetTimerEx(timer, duetime, 0, dpc));
4133 }
4134 
4135 /*
4136  * The Windows DDK documentation seems to say that cancelling
4137  * a timer that has a DPC will result in the DPC also being
4138  * cancelled, but this isn't really the case.
4139  */
4140 
4141 uint8_t
4142 KeCancelTimer(timer)
4143 	ktimer			*timer;
4144 {
4145 	uint8_t			pending;
4146 
4147 	if (timer == NULL)
4148 		return (FALSE);
4149 
4150 	mtx_lock(&ntoskrnl_dispatchlock);
4151 
4152 	pending = timer->k_header.dh_inserted;
4153 
4154 	if (timer->k_header.dh_inserted == TRUE) {
4155 		timer->k_header.dh_inserted = FALSE;
4156 		ntoskrnl_remove_timer(timer);
4157 #ifdef NTOSKRNL_DEBUG_TIMERS
4158 		ntoskrnl_timer_cancels++;
4159 #endif
4160 	}
4161 
4162 	mtx_unlock(&ntoskrnl_dispatchlock);
4163 
4164 	return (pending);
4165 }
4166 
4167 uint8_t
4168 KeReadStateTimer(timer)
4169 	ktimer			*timer;
4170 {
4171 	return (timer->k_header.dh_sigstate);
4172 }
4173 
4174 static int32_t
4175 KeDelayExecutionThread(uint8_t wait_mode, uint8_t alertable, int64_t *interval)
4176 {
4177 	ktimer                  timer;
4178 
4179 	if (wait_mode != 0)
4180 		panic("invalid wait_mode %d", wait_mode);
4181 
4182 	KeInitializeTimer(&timer);
4183 	KeSetTimer(&timer, *interval, NULL);
4184 	KeWaitForSingleObject(&timer, 0, 0, alertable, NULL);
4185 
4186 	return STATUS_SUCCESS;
4187 }
4188 
4189 static uint64_t
4190 KeQueryInterruptTime(void)
4191 {
4192 	int ticks;
4193 	struct timeval tv;
4194 
4195 	getmicrouptime(&tv);
4196 
4197 	ticks = tvtohz(&tv);
4198 
4199 	return ticks * howmany(10000000, hz);
4200 }
4201 
4202 static struct thread *
4203 KeGetCurrentThread(void)
4204 {
4205 
4206 	return curthread;
4207 }
4208 
4209 static int32_t
4210 KeSetPriorityThread(td, pri)
4211 	struct thread	*td;
4212 	int32_t		pri;
4213 {
4214 	int32_t old;
4215 
4216 	if (td == NULL)
4217 		return LOW_REALTIME_PRIORITY;
4218 
4219 	if (td->td_priority <= PRI_MIN_KERN)
4220 		old = HIGH_PRIORITY;
4221 	else if (td->td_priority >= PRI_MAX_KERN)
4222 		old = LOW_PRIORITY;
4223 	else
4224 		old = LOW_REALTIME_PRIORITY;
4225 
4226 	thread_lock(td);
4227 	if (pri == HIGH_PRIORITY)
4228 		sched_prio(td, PRI_MIN_KERN);
4229 	if (pri == LOW_REALTIME_PRIORITY)
4230 		sched_prio(td, PRI_MIN_KERN + (PRI_MAX_KERN - PRI_MIN_KERN) / 2);
4231 	if (pri == LOW_PRIORITY)
4232 		sched_prio(td, PRI_MAX_KERN);
4233 	thread_unlock(td);
4234 
4235 	return old;
4236 }
4237 
4238 static void
4239 dummy()
4240 {
4241 	printf("ntoskrnl dummy called...\n");
4242 }
4243 
4244 image_patch_table ntoskrnl_functbl[] = {
4245 	IMPORT_SFUNC(RtlZeroMemory, 2),
4246 	IMPORT_SFUNC(RtlSecureZeroMemory, 2),
4247 	IMPORT_SFUNC(RtlFillMemory, 3),
4248 	IMPORT_SFUNC(RtlMoveMemory, 3),
4249 	IMPORT_SFUNC(RtlCharToInteger, 3),
4250 	IMPORT_SFUNC(RtlCopyMemory, 3),
4251 	IMPORT_SFUNC(RtlCopyString, 2),
4252 	IMPORT_SFUNC(RtlCompareMemory, 3),
4253 	IMPORT_SFUNC(RtlEqualUnicodeString, 3),
4254 	IMPORT_SFUNC(RtlCopyUnicodeString, 2),
4255 	IMPORT_SFUNC(RtlUnicodeStringToAnsiString, 3),
4256 	IMPORT_SFUNC(RtlAnsiStringToUnicodeString, 3),
4257 	IMPORT_SFUNC(RtlInitAnsiString, 2),
4258 	IMPORT_SFUNC_MAP(RtlInitString, RtlInitAnsiString, 2),
4259 	IMPORT_SFUNC(RtlInitUnicodeString, 2),
4260 	IMPORT_SFUNC(RtlFreeAnsiString, 1),
4261 	IMPORT_SFUNC(RtlFreeUnicodeString, 1),
4262 	IMPORT_SFUNC(RtlUnicodeStringToInteger, 3),
4263 	IMPORT_CFUNC(sprintf, 0),
4264 	IMPORT_CFUNC(vsprintf, 0),
4265 	IMPORT_CFUNC_MAP(_snprintf, snprintf, 0),
4266 	IMPORT_CFUNC_MAP(_vsnprintf, vsnprintf, 0),
4267 	IMPORT_CFUNC(DbgPrint, 0),
4268 	IMPORT_SFUNC(DbgBreakPoint, 0),
4269 	IMPORT_SFUNC(KeBugCheckEx, 5),
4270 	IMPORT_CFUNC(strncmp, 0),
4271 	IMPORT_CFUNC(strcmp, 0),
4272 	IMPORT_CFUNC_MAP(stricmp, strcasecmp, 0),
4273 	IMPORT_CFUNC(strncpy, 0),
4274 	IMPORT_CFUNC(strcpy, 0),
4275 	IMPORT_CFUNC(strlen, 0),
4276 	IMPORT_CFUNC_MAP(toupper, ntoskrnl_toupper, 0),
4277 	IMPORT_CFUNC_MAP(tolower, ntoskrnl_tolower, 0),
4278 	IMPORT_CFUNC_MAP(strstr, ntoskrnl_strstr, 0),
4279 	IMPORT_CFUNC_MAP(strncat, ntoskrnl_strncat, 0),
4280 	IMPORT_CFUNC_MAP(strchr, index, 0),
4281 	IMPORT_CFUNC_MAP(strrchr, rindex, 0),
4282 	IMPORT_CFUNC(memcpy, 0),
4283 	IMPORT_CFUNC_MAP(memmove, ntoskrnl_memmove, 0),
4284 	IMPORT_CFUNC_MAP(memset, ntoskrnl_memset, 0),
4285 	IMPORT_CFUNC_MAP(memchr, ntoskrnl_memchr, 0),
4286 	IMPORT_SFUNC(IoAllocateDriverObjectExtension, 4),
4287 	IMPORT_SFUNC(IoGetDriverObjectExtension, 2),
4288 	IMPORT_FFUNC(IofCallDriver, 2),
4289 	IMPORT_FFUNC(IofCompleteRequest, 2),
4290 	IMPORT_SFUNC(IoAcquireCancelSpinLock, 1),
4291 	IMPORT_SFUNC(IoReleaseCancelSpinLock, 1),
4292 	IMPORT_SFUNC(IoCancelIrp, 1),
4293 	IMPORT_SFUNC(IoConnectInterrupt, 11),
4294 	IMPORT_SFUNC(IoDisconnectInterrupt, 1),
4295 	IMPORT_SFUNC(IoCreateDevice, 7),
4296 	IMPORT_SFUNC(IoDeleteDevice, 1),
4297 	IMPORT_SFUNC(IoGetAttachedDevice, 1),
4298 	IMPORT_SFUNC(IoAttachDeviceToDeviceStack, 2),
4299 	IMPORT_SFUNC(IoDetachDevice, 1),
4300 	IMPORT_SFUNC(IoBuildSynchronousFsdRequest, 7),
4301 	IMPORT_SFUNC(IoBuildAsynchronousFsdRequest, 6),
4302 	IMPORT_SFUNC(IoBuildDeviceIoControlRequest, 9),
4303 	IMPORT_SFUNC(IoAllocateIrp, 2),
4304 	IMPORT_SFUNC(IoReuseIrp, 2),
4305 	IMPORT_SFUNC(IoMakeAssociatedIrp, 2),
4306 	IMPORT_SFUNC(IoFreeIrp, 1),
4307 	IMPORT_SFUNC(IoInitializeIrp, 3),
4308 	IMPORT_SFUNC(KeAcquireInterruptSpinLock, 1),
4309 	IMPORT_SFUNC(KeReleaseInterruptSpinLock, 2),
4310 	IMPORT_SFUNC(KeSynchronizeExecution, 3),
4311 	IMPORT_SFUNC(KeWaitForSingleObject, 5),
4312 	IMPORT_SFUNC(KeWaitForMultipleObjects, 8),
4313 	IMPORT_SFUNC(_allmul, 4),
4314 	IMPORT_SFUNC(_alldiv, 4),
4315 	IMPORT_SFUNC(_allrem, 4),
4316 	IMPORT_RFUNC(_allshr, 0),
4317 	IMPORT_RFUNC(_allshl, 0),
4318 	IMPORT_SFUNC(_aullmul, 4),
4319 	IMPORT_SFUNC(_aulldiv, 4),
4320 	IMPORT_SFUNC(_aullrem, 4),
4321 	IMPORT_RFUNC(_aullshr, 0),
4322 	IMPORT_RFUNC(_aullshl, 0),
4323 	IMPORT_CFUNC(atoi, 0),
4324 	IMPORT_CFUNC(atol, 0),
4325 	IMPORT_CFUNC(rand, 0),
4326 	IMPORT_CFUNC(srand, 0),
4327 	IMPORT_SFUNC(WRITE_REGISTER_USHORT, 2),
4328 	IMPORT_SFUNC(READ_REGISTER_USHORT, 1),
4329 	IMPORT_SFUNC(WRITE_REGISTER_ULONG, 2),
4330 	IMPORT_SFUNC(READ_REGISTER_ULONG, 1),
4331 	IMPORT_SFUNC(READ_REGISTER_UCHAR, 1),
4332 	IMPORT_SFUNC(WRITE_REGISTER_UCHAR, 2),
4333 	IMPORT_SFUNC(ExInitializePagedLookasideList, 7),
4334 	IMPORT_SFUNC(ExDeletePagedLookasideList, 1),
4335 	IMPORT_SFUNC(ExInitializeNPagedLookasideList, 7),
4336 	IMPORT_SFUNC(ExDeleteNPagedLookasideList, 1),
4337 	IMPORT_FFUNC(InterlockedPopEntrySList, 1),
4338 	IMPORT_FFUNC(InitializeSListHead, 1),
4339 	IMPORT_FFUNC(InterlockedPushEntrySList, 2),
4340 	IMPORT_SFUNC(ExQueryDepthSList, 1),
4341 	IMPORT_FFUNC_MAP(ExpInterlockedPopEntrySList,
4342 		InterlockedPopEntrySList, 1),
4343 	IMPORT_FFUNC_MAP(ExpInterlockedPushEntrySList,
4344 		InterlockedPushEntrySList, 2),
4345 	IMPORT_FFUNC(ExInterlockedPopEntrySList, 2),
4346 	IMPORT_FFUNC(ExInterlockedPushEntrySList, 3),
4347 	IMPORT_SFUNC(ExAllocatePoolWithTag, 3),
4348 	IMPORT_SFUNC(ExFreePoolWithTag, 2),
4349 	IMPORT_SFUNC(ExFreePool, 1),
4350 #ifdef __i386__
4351 	IMPORT_FFUNC(KefAcquireSpinLockAtDpcLevel, 1),
4352 	IMPORT_FFUNC(KefReleaseSpinLockFromDpcLevel,1),
4353 	IMPORT_FFUNC(KeAcquireSpinLockRaiseToDpc, 1),
4354 #else
4355 	/*
4356 	 * For AMD64, we can get away with just mapping
4357 	 * KeAcquireSpinLockRaiseToDpc() directly to KfAcquireSpinLock()
4358 	 * because the calling conventions end up being the same.
4359 	 * On i386, we have to be careful because KfAcquireSpinLock()
4360 	 * is _fastcall but KeAcquireSpinLockRaiseToDpc() isn't.
4361 	 */
4362 	IMPORT_SFUNC(KeAcquireSpinLockAtDpcLevel, 1),
4363 	IMPORT_SFUNC(KeReleaseSpinLockFromDpcLevel, 1),
4364 	IMPORT_SFUNC_MAP(KeAcquireSpinLockRaiseToDpc, KfAcquireSpinLock, 1),
4365 #endif
4366 	IMPORT_SFUNC_MAP(KeReleaseSpinLock, KfReleaseSpinLock, 1),
4367 	IMPORT_FFUNC(InterlockedIncrement, 1),
4368 	IMPORT_FFUNC(InterlockedDecrement, 1),
4369 	IMPORT_FFUNC(InterlockedExchange, 2),
4370 	IMPORT_FFUNC(ExInterlockedAddLargeStatistic, 2),
4371 	IMPORT_SFUNC(IoAllocateMdl, 5),
4372 	IMPORT_SFUNC(IoFreeMdl, 1),
4373 	IMPORT_SFUNC(MmAllocateContiguousMemory, 2 + 1),
4374 	IMPORT_SFUNC(MmAllocateContiguousMemorySpecifyCache, 5 + 3),
4375 	IMPORT_SFUNC(MmFreeContiguousMemory, 1),
4376 	IMPORT_SFUNC(MmFreeContiguousMemorySpecifyCache, 3),
4377 	IMPORT_SFUNC(MmSizeOfMdl, 1),
4378 	IMPORT_SFUNC(MmMapLockedPages, 2),
4379 	IMPORT_SFUNC(MmMapLockedPagesSpecifyCache, 6),
4380 	IMPORT_SFUNC(MmUnmapLockedPages, 2),
4381 	IMPORT_SFUNC(MmBuildMdlForNonPagedPool, 1),
4382 	IMPORT_SFUNC(MmGetPhysicalAddress, 1),
4383 	IMPORT_SFUNC(MmGetSystemRoutineAddress, 1),
4384 	IMPORT_SFUNC(MmIsAddressValid, 1),
4385 	IMPORT_SFUNC(MmMapIoSpace, 3 + 1),
4386 	IMPORT_SFUNC(MmUnmapIoSpace, 2),
4387 	IMPORT_SFUNC(KeInitializeSpinLock, 1),
4388 	IMPORT_SFUNC(IoIsWdmVersionAvailable, 2),
4389 	IMPORT_SFUNC(IoOpenDeviceRegistryKey, 4),
4390 	IMPORT_SFUNC(IoGetDeviceObjectPointer, 4),
4391 	IMPORT_SFUNC(IoGetDeviceProperty, 5),
4392 	IMPORT_SFUNC(IoAllocateWorkItem, 1),
4393 	IMPORT_SFUNC(IoFreeWorkItem, 1),
4394 	IMPORT_SFUNC(IoQueueWorkItem, 4),
4395 	IMPORT_SFUNC(ExQueueWorkItem, 2),
4396 	IMPORT_SFUNC(ntoskrnl_workitem, 2),
4397 	IMPORT_SFUNC(KeInitializeMutex, 2),
4398 	IMPORT_SFUNC(KeReleaseMutex, 2),
4399 	IMPORT_SFUNC(KeReadStateMutex, 1),
4400 	IMPORT_SFUNC(KeInitializeEvent, 3),
4401 	IMPORT_SFUNC(KeSetEvent, 3),
4402 	IMPORT_SFUNC(KeResetEvent, 1),
4403 	IMPORT_SFUNC(KeClearEvent, 1),
4404 	IMPORT_SFUNC(KeReadStateEvent, 1),
4405 	IMPORT_SFUNC(KeInitializeTimer, 1),
4406 	IMPORT_SFUNC(KeInitializeTimerEx, 2),
4407 	IMPORT_SFUNC(KeSetTimer, 3),
4408 	IMPORT_SFUNC(KeSetTimerEx, 4),
4409 	IMPORT_SFUNC(KeCancelTimer, 1),
4410 	IMPORT_SFUNC(KeReadStateTimer, 1),
4411 	IMPORT_SFUNC(KeInitializeDpc, 3),
4412 	IMPORT_SFUNC(KeInsertQueueDpc, 3),
4413 	IMPORT_SFUNC(KeRemoveQueueDpc, 1),
4414 	IMPORT_SFUNC(KeSetImportanceDpc, 2),
4415 	IMPORT_SFUNC(KeSetTargetProcessorDpc, 2),
4416 	IMPORT_SFUNC(KeFlushQueuedDpcs, 0),
4417 	IMPORT_SFUNC(KeGetCurrentProcessorNumber, 1),
4418 	IMPORT_SFUNC(ObReferenceObjectByHandle, 6),
4419 	IMPORT_FFUNC(ObfDereferenceObject, 1),
4420 	IMPORT_SFUNC(ZwClose, 1),
4421 	IMPORT_SFUNC(PsCreateSystemThread, 7),
4422 	IMPORT_SFUNC(PsTerminateSystemThread, 1),
4423 	IMPORT_SFUNC(IoWMIRegistrationControl, 2),
4424 	IMPORT_SFUNC(WmiQueryTraceInformation, 5),
4425 	IMPORT_CFUNC(WmiTraceMessage, 0),
4426 	IMPORT_SFUNC(KeQuerySystemTime, 1),
4427 	IMPORT_CFUNC(KeTickCount, 0),
4428 	IMPORT_SFUNC(KeDelayExecutionThread, 3),
4429 	IMPORT_SFUNC(KeQueryInterruptTime, 0),
4430 	IMPORT_SFUNC(KeGetCurrentThread, 0),
4431 	IMPORT_SFUNC(KeSetPriorityThread, 2),
4432 
4433 	/*
4434 	 * This last entry is a catch-all for any function we haven't
4435 	 * implemented yet. The PE import list patching routine will
4436 	 * use it for any function that doesn't have an explicit match
4437 	 * in this table.
4438 	 */
4439 
4440 	{ NULL, (FUNC)dummy, NULL, 0, WINDRV_WRAP_STDCALL },
4441 
4442 	/* End of list. */
4443 	{ NULL, NULL, NULL }
4444 };
4445