xref: /titanic_41/usr/src/uts/sun4v/ml/trap_table.s (revision 74e20cfe817b82802b16fac8690dadcda76f54f5)
1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22/*
23 * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
24 * Use is subject to license terms.
25 */
26
27#pragma ident	"%Z%%M%	%I%	%E% SMI"
28
29#if !defined(lint)
30#include "assym.h"
31#endif /* !lint */
32#include <sys/asm_linkage.h>
33#include <sys/privregs.h>
34#include <sys/sun4asi.h>
35#include <sys/machasi.h>
36#include <sys/hypervisor_api.h>
37#include <sys/machtrap.h>
38#include <sys/machthread.h>
39#include <sys/pcb.h>
40#include <sys/pte.h>
41#include <sys/mmu.h>
42#include <sys/machpcb.h>
43#include <sys/async.h>
44#include <sys/intreg.h>
45#include <sys/scb.h>
46#include <sys/psr_compat.h>
47#include <sys/syscall.h>
48#include <sys/machparam.h>
49#include <sys/traptrace.h>
50#include <vm/hat_sfmmu.h>
51#include <sys/archsystm.h>
52#include <sys/utrap.h>
53#include <sys/clock.h>
54#include <sys/intr.h>
55#include <sys/fpu/fpu_simulator.h>
56#include <vm/seg_spt.h>
57
58/*
59 * WARNING: If you add a fast trap handler which can be invoked by a
60 * non-privileged user, you may have to use the FAST_TRAP_DONE macro
61 * instead of "done" instruction to return back to the user mode. See
62 * comments for the "fast_trap_done" entry point for more information.
63 *
64 * An alternate FAST_TRAP_DONE_CHK_INTR macro should be used for the
65 * cases where you always want to process any pending interrupts before
66 * returning back to the user mode.
67 */
68#define	FAST_TRAP_DONE		\
69	ba,a	fast_trap_done
70
71#define	FAST_TRAP_DONE_CHK_INTR	\
72	ba,a	fast_trap_done_chk_intr
73
74/*
75 * SPARC V9 Trap Table
76 *
77 * Most of the trap handlers are made from common building
78 * blocks, and some are instantiated multiple times within
79 * the trap table. So, I build a bunch of macros, then
80 * populate the table using only the macros.
81 *
82 * Many macros branch to sys_trap.  Its calling convention is:
83 *	%g1		kernel trap handler
84 *	%g2, %g3	args for above
85 *	%g4		desire %pil
86 */
87
88#ifdef	TRAPTRACE
89
90/*
91 * Tracing macro. Adds two instructions if TRAPTRACE is defined.
92 */
93#define	TT_TRACE(label)		\
94	ba	label		;\
95	rd	%pc, %g7
96#define	TT_TRACE_INS	2
97
98#define	TT_TRACE_L(label)	\
99	ba	label		;\
100	rd	%pc, %l4	;\
101	clr	%l4
102#define	TT_TRACE_L_INS	3
103
104#else
105
106#define	TT_TRACE(label)
107#define	TT_TRACE_INS	0
108
109#define	TT_TRACE_L(label)
110#define	TT_TRACE_L_INS	0
111
112#endif
113
114/*
115 * This macro is used to update per cpu mmu stats in perf critical
116 * paths. It is only enabled in debug kernels or if SFMMU_STAT_GATHER
117 * is defined.
118 */
119#if defined(DEBUG) || defined(SFMMU_STAT_GATHER)
120#define	HAT_PERCPU_DBSTAT(stat)			\
121	mov	stat, %g1			;\
122	ba	stat_mmu			;\
123	rd	%pc, %g7
124#else
125#define	HAT_PERCPU_DBSTAT(stat)
126#endif /* DEBUG || SFMMU_STAT_GATHER */
127
128/*
129 * This first set are funneled to trap() with %tt as the type.
130 * Trap will then either panic or send the user a signal.
131 */
132/*
133 * NOT is used for traps that just shouldn't happen.
134 * It comes in both single and quadruple flavors.
135 */
136#if !defined(lint)
137	.global	trap
138#endif /* !lint */
139#define	NOT			\
140	TT_TRACE(trace_gen)	;\
141	set	trap, %g1	;\
142	rdpr	%tt, %g3	;\
143	ba,pt	%xcc, sys_trap	;\
144	sub	%g0, 1, %g4	;\
145	.align	32
146#define	NOT4	NOT; NOT; NOT; NOT
147
148#define	NOTP				\
149	TT_TRACE(trace_gen)		;\
150	ba,pt	%xcc, ptl1_panic	;\
151	  mov	PTL1_BAD_TRAP, %g1	;\
152	.align	32
153#define	NOTP4	NOTP; NOTP; NOTP; NOTP
154
155
156/*
157 * BAD is used for trap vectors we don't have a kernel
158 * handler for.
159 * It also comes in single and quadruple versions.
160 */
161#define	BAD	NOT
162#define	BAD4	NOT4
163
164#define	DONE			\
165	done;			\
166	.align	32
167
168/*
169 * TRAP vectors to the trap() function.
170 * It's main use is for user errors.
171 */
172#if !defined(lint)
173	.global	trap
174#endif /* !lint */
175#define	TRAP(arg)		\
176	TT_TRACE(trace_gen)	;\
177	set	trap, %g1	;\
178	mov	arg, %g3	;\
179	ba,pt	%xcc, sys_trap	;\
180	sub	%g0, 1, %g4	;\
181	.align	32
182
183/*
184 * SYSCALL is used for system calls on both ILP32 and LP64 kernels
185 * depending on the "which" parameter (should be syscall_trap,
186 * syscall_trap32, or nosys for unused system call traps).
187 */
188#define	SYSCALL(which)			\
189	TT_TRACE(trace_gen)		;\
190	set	(which), %g1		;\
191	ba,pt	%xcc, sys_trap		;\
192	sub	%g0, 1, %g4		;\
193	.align	32
194
195/*
196 * GOTO just jumps to a label.
197 * It's used for things that can be fixed without going thru sys_trap.
198 */
199#define	GOTO(label)		\
200	.global	label		;\
201	ba,a	label		;\
202	.empty			;\
203	.align	32
204
205/*
206 * GOTO_TT just jumps to a label.
207 * correctable ECC error traps at  level 0 and 1 will use this macro.
208 * It's used for things that can be fixed without going thru sys_trap.
209 */
210#define	GOTO_TT(label, ttlabel)		\
211	.global	label		;\
212	TT_TRACE(ttlabel)	;\
213	ba,a	label		;\
214	.empty			;\
215	.align	32
216
217/*
218 * Privileged traps
219 * Takes breakpoint if privileged, calls trap() if not.
220 */
221#define	PRIV(label)			\
222	rdpr	%tstate, %g1		;\
223	btst	TSTATE_PRIV, %g1	;\
224	bnz	label			;\
225	rdpr	%tt, %g3		;\
226	set	trap, %g1		;\
227	ba,pt	%xcc, sys_trap		;\
228	sub	%g0, 1, %g4		;\
229	.align	32
230
231
232/*
233 * DTrace traps.
234 */
235#define	DTRACE_PID			\
236	.global dtrace_pid_probe				;\
237	set	dtrace_pid_probe, %g1				;\
238	ba,pt	%xcc, user_trap					;\
239	sub	%g0, 1, %g4					;\
240	.align	32
241
242#define	DTRACE_RETURN			\
243	.global dtrace_return_probe				;\
244	set	dtrace_return_probe, %g1			;\
245	ba,pt	%xcc, user_trap					;\
246	sub	%g0, 1, %g4					;\
247	.align	32
248
249/*
250 * REGISTER WINDOW MANAGEMENT MACROS
251 */
252
253/*
254 * various convenient units of padding
255 */
256#define	SKIP(n)	.skip 4*(n)
257
258/*
259 * CLEAN_WINDOW is the simple handler for cleaning a register window.
260 */
261#define	CLEAN_WINDOW						\
262	TT_TRACE_L(trace_win)					;\
263	rdpr %cleanwin, %l0; inc %l0; wrpr %l0, %cleanwin	;\
264	clr %l0; clr %l1; clr %l2; clr %l3			;\
265	clr %l4; clr %l5; clr %l6; clr %l7			;\
266	clr %o0; clr %o1; clr %o2; clr %o3			;\
267	clr %o4; clr %o5; clr %o6; clr %o7			;\
268	retry; .align 128
269
270#if !defined(lint)
271
272/*
273 * If we get an unresolved tlb miss while in a window handler, the fault
274 * handler will resume execution at the last instruction of the window
275 * hander, instead of delivering the fault to the kernel.  Spill handlers
276 * use this to spill windows into the wbuf.
277 *
278 * The mixed handler works by checking %sp, and branching to the correct
279 * handler.  This is done by branching back to label 1: for 32b frames,
280 * or label 2: for 64b frames; which implies the handler order is: 32b,
281 * 64b, mixed.  The 1: and 2: labels are offset into the routines to
282 * allow the branchs' delay slots to contain useful instructions.
283 */
284
285/*
286 * SPILL_32bit spills a 32-bit-wide kernel register window.  It
287 * assumes that the kernel context and the nucleus context are the
288 * same.  The stack pointer is required to be eight-byte aligned even
289 * though this code only needs it to be four-byte aligned.
290 */
291#define	SPILL_32bit(tail)					\
292	srl	%sp, 0, %sp					;\
2931:	st	%l0, [%sp + 0]					;\
294	st	%l1, [%sp + 4]					;\
295	st	%l2, [%sp + 8]					;\
296	st	%l3, [%sp + 12]					;\
297	st	%l4, [%sp + 16]					;\
298	st	%l5, [%sp + 20]					;\
299	st	%l6, [%sp + 24]					;\
300	st	%l7, [%sp + 28]					;\
301	st	%i0, [%sp + 32]					;\
302	st	%i1, [%sp + 36]					;\
303	st	%i2, [%sp + 40]					;\
304	st	%i3, [%sp + 44]					;\
305	st	%i4, [%sp + 48]					;\
306	st	%i5, [%sp + 52]					;\
307	st	%i6, [%sp + 56]					;\
308	st	%i7, [%sp + 60]					;\
309	TT_TRACE_L(trace_win)					;\
310	saved							;\
311	retry							;\
312	SKIP(31-19-TT_TRACE_L_INS)				;\
313	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
314	.empty
315
316/*
317 * SPILL_32bit_asi spills a 32-bit-wide register window into a 32-bit
318 * wide address space via the designated asi.  It is used to spill
319 * non-kernel windows.  The stack pointer is required to be eight-byte
320 * aligned even though this code only needs it to be four-byte
321 * aligned.
322 */
323#define	SPILL_32bit_asi(asi_num, tail)				\
324	srl	%sp, 0, %sp					;\
3251:	sta	%l0, [%sp + %g0]asi_num				;\
326	mov	4, %g1						;\
327	sta	%l1, [%sp + %g1]asi_num				;\
328	mov	8, %g2						;\
329	sta	%l2, [%sp + %g2]asi_num				;\
330	mov	12, %g3						;\
331	sta	%l3, [%sp + %g3]asi_num				;\
332	add	%sp, 16, %g4					;\
333	sta	%l4, [%g4 + %g0]asi_num				;\
334	sta	%l5, [%g4 + %g1]asi_num				;\
335	sta	%l6, [%g4 + %g2]asi_num				;\
336	sta	%l7, [%g4 + %g3]asi_num				;\
337	add	%g4, 16, %g4					;\
338	sta	%i0, [%g4 + %g0]asi_num				;\
339	sta	%i1, [%g4 + %g1]asi_num				;\
340	sta	%i2, [%g4 + %g2]asi_num				;\
341	sta	%i3, [%g4 + %g3]asi_num				;\
342	add	%g4, 16, %g4					;\
343	sta	%i4, [%g4 + %g0]asi_num				;\
344	sta	%i5, [%g4 + %g1]asi_num				;\
345	sta	%i6, [%g4 + %g2]asi_num				;\
346	sta	%i7, [%g4 + %g3]asi_num				;\
347	TT_TRACE_L(trace_win)					;\
348	saved							;\
349	retry							;\
350	SKIP(31-25-TT_TRACE_L_INS)				;\
351	ba,a,pt %xcc, fault_32bit_/**/tail			;\
352	.empty
353
354#define	SPILL_32bit_tt1(asi_num, tail)				\
355	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
356	.empty							;\
357	.align 128
358
359
360/*
361 * FILL_32bit fills a 32-bit-wide kernel register window.  It assumes
362 * that the kernel context and the nucleus context are the same.  The
363 * stack pointer is required to be eight-byte aligned even though this
364 * code only needs it to be four-byte aligned.
365 */
366#define	FILL_32bit(tail)					\
367	srl	%sp, 0, %sp					;\
3681:	TT_TRACE_L(trace_win)					;\
369	ld	[%sp + 0], %l0					;\
370	ld	[%sp + 4], %l1					;\
371	ld	[%sp + 8], %l2					;\
372	ld	[%sp + 12], %l3					;\
373	ld	[%sp + 16], %l4					;\
374	ld	[%sp + 20], %l5					;\
375	ld	[%sp + 24], %l6					;\
376	ld	[%sp + 28], %l7					;\
377	ld	[%sp + 32], %i0					;\
378	ld	[%sp + 36], %i1					;\
379	ld	[%sp + 40], %i2					;\
380	ld	[%sp + 44], %i3					;\
381	ld	[%sp + 48], %i4					;\
382	ld	[%sp + 52], %i5					;\
383	ld	[%sp + 56], %i6					;\
384	ld	[%sp + 60], %i7					;\
385	restored						;\
386	retry							;\
387	SKIP(31-19-TT_TRACE_L_INS)				;\
388	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
389	.empty
390
391/*
392 * FILL_32bit_asi fills a 32-bit-wide register window from a 32-bit
393 * wide address space via the designated asi.  It is used to fill
394 * non-kernel windows.  The stack pointer is required to be eight-byte
395 * aligned even though this code only needs it to be four-byte
396 * aligned.
397 */
398#define	FILL_32bit_asi(asi_num, tail)				\
399	srl	%sp, 0, %sp					;\
4001:	TT_TRACE_L(trace_win)					;\
401	mov	4, %g1						;\
402	lda	[%sp + %g0]asi_num, %l0				;\
403	mov	8, %g2						;\
404	lda	[%sp + %g1]asi_num, %l1				;\
405	mov	12, %g3						;\
406	lda	[%sp + %g2]asi_num, %l2				;\
407	lda	[%sp + %g3]asi_num, %l3				;\
408	add	%sp, 16, %g4					;\
409	lda	[%g4 + %g0]asi_num, %l4				;\
410	lda	[%g4 + %g1]asi_num, %l5				;\
411	lda	[%g4 + %g2]asi_num, %l6				;\
412	lda	[%g4 + %g3]asi_num, %l7				;\
413	add	%g4, 16, %g4					;\
414	lda	[%g4 + %g0]asi_num, %i0				;\
415	lda	[%g4 + %g1]asi_num, %i1				;\
416	lda	[%g4 + %g2]asi_num, %i2				;\
417	lda	[%g4 + %g3]asi_num, %i3				;\
418	add	%g4, 16, %g4					;\
419	lda	[%g4 + %g0]asi_num, %i4				;\
420	lda	[%g4 + %g1]asi_num, %i5				;\
421	lda	[%g4 + %g2]asi_num, %i6				;\
422	lda	[%g4 + %g3]asi_num, %i7				;\
423	restored						;\
424	retry							;\
425	SKIP(31-25-TT_TRACE_L_INS)				;\
426	ba,a,pt %xcc, fault_32bit_/**/tail			;\
427	.empty
428
429
430/*
431 * SPILL_64bit spills a 64-bit-wide kernel register window.  It
432 * assumes that the kernel context and the nucleus context are the
433 * same.  The stack pointer is required to be eight-byte aligned.
434 */
435#define	SPILL_64bit(tail)					\
4362:	stx	%l0, [%sp + V9BIAS64 + 0]			;\
437	stx	%l1, [%sp + V9BIAS64 + 8]			;\
438	stx	%l2, [%sp + V9BIAS64 + 16]			;\
439	stx	%l3, [%sp + V9BIAS64 + 24]			;\
440	stx	%l4, [%sp + V9BIAS64 + 32]			;\
441	stx	%l5, [%sp + V9BIAS64 + 40]			;\
442	stx	%l6, [%sp + V9BIAS64 + 48]			;\
443	stx	%l7, [%sp + V9BIAS64 + 56]			;\
444	stx	%i0, [%sp + V9BIAS64 + 64]			;\
445	stx	%i1, [%sp + V9BIAS64 + 72]			;\
446	stx	%i2, [%sp + V9BIAS64 + 80]			;\
447	stx	%i3, [%sp + V9BIAS64 + 88]			;\
448	stx	%i4, [%sp + V9BIAS64 + 96]			;\
449	stx	%i5, [%sp + V9BIAS64 + 104]			;\
450	stx	%i6, [%sp + V9BIAS64 + 112]			;\
451	stx	%i7, [%sp + V9BIAS64 + 120]			;\
452	TT_TRACE_L(trace_win)					;\
453	saved							;\
454	retry							;\
455	SKIP(31-18-TT_TRACE_L_INS)				;\
456	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
457	.empty
458
459#define	SPILL_64bit_ktt1(tail)				\
460	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
461	.empty							;\
462	.align 128
463
464#define	SPILL_mixed_ktt1(tail)				\
465	btst	1, %sp						;\
466	bz,a,pt	%xcc, fault_32bit_/**/tail			;\
467	srl	%sp, 0, %sp					;\
468	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
469	.empty							;\
470	.align 128
471
472/*
473 * SPILL_64bit_asi spills a 64-bit-wide register window into a 64-bit
474 * wide address space via the designated asi.  It is used to spill
475 * non-kernel windows.  The stack pointer is required to be eight-byte
476 * aligned.
477 */
478#define	SPILL_64bit_asi(asi_num, tail)				\
479	mov	0 + V9BIAS64, %g1				;\
4802:	stxa	%l0, [%sp + %g1]asi_num				;\
481	mov	8 + V9BIAS64, %g2				;\
482	stxa	%l1, [%sp + %g2]asi_num				;\
483	mov	16 + V9BIAS64, %g3				;\
484	stxa	%l2, [%sp + %g3]asi_num				;\
485	mov	24 + V9BIAS64, %g4				;\
486	stxa	%l3, [%sp + %g4]asi_num				;\
487	add	%sp, 32, %g5					;\
488	stxa	%l4, [%g5 + %g1]asi_num				;\
489	stxa	%l5, [%g5 + %g2]asi_num				;\
490	stxa	%l6, [%g5 + %g3]asi_num				;\
491	stxa	%l7, [%g5 + %g4]asi_num				;\
492	add	%g5, 32, %g5					;\
493	stxa	%i0, [%g5 + %g1]asi_num				;\
494	stxa	%i1, [%g5 + %g2]asi_num				;\
495	stxa	%i2, [%g5 + %g3]asi_num				;\
496	stxa	%i3, [%g5 + %g4]asi_num				;\
497	add	%g5, 32, %g5					;\
498	stxa	%i4, [%g5 + %g1]asi_num				;\
499	stxa	%i5, [%g5 + %g2]asi_num				;\
500	stxa	%i6, [%g5 + %g3]asi_num				;\
501	stxa	%i7, [%g5 + %g4]asi_num				;\
502	TT_TRACE_L(trace_win)					;\
503	saved							;\
504	retry							;\
505	SKIP(31-25-TT_TRACE_L_INS)				;\
506	ba,a,pt %xcc, fault_64bit_/**/tail			;\
507	.empty
508
509#define	SPILL_64bit_tt1(asi_num, tail)				\
510	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
511	.empty							;\
512	.align 128
513
514/*
515 * FILL_64bit fills a 64-bit-wide kernel register window.  It assumes
516 * that the kernel context and the nucleus context are the same.  The
517 * stack pointer is required to be eight-byte aligned.
518 */
519#define	FILL_64bit(tail)					\
5202:	TT_TRACE_L(trace_win)					;\
521	ldx	[%sp + V9BIAS64 + 0], %l0			;\
522	ldx	[%sp + V9BIAS64 + 8], %l1			;\
523	ldx	[%sp + V9BIAS64 + 16], %l2			;\
524	ldx	[%sp + V9BIAS64 + 24], %l3			;\
525	ldx	[%sp + V9BIAS64 + 32], %l4			;\
526	ldx	[%sp + V9BIAS64 + 40], %l5			;\
527	ldx	[%sp + V9BIAS64 + 48], %l6			;\
528	ldx	[%sp + V9BIAS64 + 56], %l7			;\
529	ldx	[%sp + V9BIAS64 + 64], %i0			;\
530	ldx	[%sp + V9BIAS64 + 72], %i1			;\
531	ldx	[%sp + V9BIAS64 + 80], %i2			;\
532	ldx	[%sp + V9BIAS64 + 88], %i3			;\
533	ldx	[%sp + V9BIAS64 + 96], %i4			;\
534	ldx	[%sp + V9BIAS64 + 104], %i5			;\
535	ldx	[%sp + V9BIAS64 + 112], %i6			;\
536	ldx	[%sp + V9BIAS64 + 120], %i7			;\
537	restored						;\
538	retry							;\
539	SKIP(31-18-TT_TRACE_L_INS)				;\
540	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
541	.empty
542
543/*
544 * FILL_64bit_asi fills a 64-bit-wide register window from a 64-bit
545 * wide address space via the designated asi.  It is used to fill
546 * non-kernel windows.  The stack pointer is required to be eight-byte
547 * aligned.
548 */
549#define	FILL_64bit_asi(asi_num, tail)				\
550	mov	V9BIAS64 + 0, %g1				;\
5512:	TT_TRACE_L(trace_win)					;\
552	ldxa	[%sp + %g1]asi_num, %l0				;\
553	mov	V9BIAS64 + 8, %g2				;\
554	ldxa	[%sp + %g2]asi_num, %l1				;\
555	mov	V9BIAS64 + 16, %g3				;\
556	ldxa	[%sp + %g3]asi_num, %l2				;\
557	mov	V9BIAS64 + 24, %g4				;\
558	ldxa	[%sp + %g4]asi_num, %l3				;\
559	add	%sp, 32, %g5					;\
560	ldxa	[%g5 + %g1]asi_num, %l4				;\
561	ldxa	[%g5 + %g2]asi_num, %l5				;\
562	ldxa	[%g5 + %g3]asi_num, %l6				;\
563	ldxa	[%g5 + %g4]asi_num, %l7				;\
564	add	%g5, 32, %g5					;\
565	ldxa	[%g5 + %g1]asi_num, %i0				;\
566	ldxa	[%g5 + %g2]asi_num, %i1				;\
567	ldxa	[%g5 + %g3]asi_num, %i2				;\
568	ldxa	[%g5 + %g4]asi_num, %i3				;\
569	add	%g5, 32, %g5					;\
570	ldxa	[%g5 + %g1]asi_num, %i4				;\
571	ldxa	[%g5 + %g2]asi_num, %i5				;\
572	ldxa	[%g5 + %g3]asi_num, %i6				;\
573	ldxa	[%g5 + %g4]asi_num, %i7				;\
574	restored						;\
575	retry							;\
576	SKIP(31-25-TT_TRACE_L_INS)				;\
577	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
578	.empty
579
580
581#endif /* !lint */
582
583/*
584 * SPILL_mixed spills either size window, depending on
585 * whether %sp is even or odd, to a 32-bit address space.
586 * This may only be used in conjunction with SPILL_32bit/
587 * FILL_64bit.
588 * Clear upper 32 bits of %sp if it is odd.
589 * We won't need to clear them in 64 bit kernel.
590 */
591#define	SPILL_mixed						\
592	btst	1, %sp						;\
593	bz,a,pt	%xcc, 1b					;\
594	srl	%sp, 0, %sp					;\
595	ba,pt	%xcc, 2b					;\
596	nop							;\
597	.align	128
598
599/*
600 * FILL_mixed(ASI) fills either size window, depending on
601 * whether %sp is even or odd, from a 32-bit address space.
602 * This may only be used in conjunction with FILL_32bit/
603 * FILL_64bit. New versions of FILL_mixed_{tt1,asi} would be
604 * needed for use with FILL_{32,64}bit_{tt1,asi}. Particular
605 * attention should be paid to the instructions that belong
606 * in the delay slots of the branches depending on the type
607 * of fill handler being branched to.
608 * Clear upper 32 bits of %sp if it is odd.
609 * We won't need to clear them in 64 bit kernel.
610 */
611#define	FILL_mixed						\
612	btst	1, %sp						;\
613	bz,a,pt	%xcc, 1b					;\
614	srl	%sp, 0, %sp					;\
615	ba,pt	%xcc, 2b					;\
616	nop							;\
617	.align	128
618
619
620/*
621 * SPILL_32clean/SPILL_64clean spill 32-bit and 64-bit register windows,
622 * respectively, into the address space via the designated asi.  The
623 * unbiased stack pointer is required to be eight-byte aligned (even for
624 * the 32-bit case even though this code does not require such strict
625 * alignment).
626 *
627 * With SPARC v9 the spill trap takes precedence over the cleanwin trap
628 * so when cansave == 0, canrestore == 6, and cleanwin == 6 the next save
629 * will cause cwp + 2 to be spilled but will not clean cwp + 1.  That
630 * window may contain kernel data so in user_rtt we set wstate to call
631 * these spill handlers on the first user spill trap.  These handler then
632 * spill the appropriate window but also back up a window and clean the
633 * window that didn't get a cleanwin trap.
634 */
635#define	SPILL_32clean(asi_num, tail)				\
636	srl	%sp, 0, %sp					;\
637	sta	%l0, [%sp + %g0]asi_num				;\
638	mov	4, %g1						;\
639	sta	%l1, [%sp + %g1]asi_num				;\
640	mov	8, %g2						;\
641	sta	%l2, [%sp + %g2]asi_num				;\
642	mov	12, %g3						;\
643	sta	%l3, [%sp + %g3]asi_num				;\
644	add	%sp, 16, %g4					;\
645	sta	%l4, [%g4 + %g0]asi_num				;\
646	sta	%l5, [%g4 + %g1]asi_num				;\
647	sta	%l6, [%g4 + %g2]asi_num				;\
648	sta	%l7, [%g4 + %g3]asi_num				;\
649	add	%g4, 16, %g4					;\
650	sta	%i0, [%g4 + %g0]asi_num				;\
651	sta	%i1, [%g4 + %g1]asi_num				;\
652	sta	%i2, [%g4 + %g2]asi_num				;\
653	sta	%i3, [%g4 + %g3]asi_num				;\
654	add	%g4, 16, %g4					;\
655	sta	%i4, [%g4 + %g0]asi_num				;\
656	sta	%i5, [%g4 + %g1]asi_num				;\
657	sta	%i6, [%g4 + %g2]asi_num				;\
658	sta	%i7, [%g4 + %g3]asi_num				;\
659	TT_TRACE_L(trace_win)					;\
660	b	.spill_clean					;\
661	  mov	WSTATE_USER32, %g7				;\
662	SKIP(31-25-TT_TRACE_L_INS)				;\
663	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
664	.empty
665
666#define	SPILL_64clean(asi_num, tail)				\
667	mov	0 + V9BIAS64, %g1				;\
668	stxa	%l0, [%sp + %g1]asi_num				;\
669	mov	8 + V9BIAS64, %g2				;\
670	stxa	%l1, [%sp + %g2]asi_num				;\
671	mov	16 + V9BIAS64, %g3				;\
672	stxa	%l2, [%sp + %g3]asi_num				;\
673	mov	24 + V9BIAS64, %g4				;\
674	stxa	%l3, [%sp + %g4]asi_num				;\
675	add	%sp, 32, %g5					;\
676	stxa	%l4, [%g5 + %g1]asi_num				;\
677	stxa	%l5, [%g5 + %g2]asi_num				;\
678	stxa	%l6, [%g5 + %g3]asi_num				;\
679	stxa	%l7, [%g5 + %g4]asi_num				;\
680	add	%g5, 32, %g5					;\
681	stxa	%i0, [%g5 + %g1]asi_num				;\
682	stxa	%i1, [%g5 + %g2]asi_num				;\
683	stxa	%i2, [%g5 + %g3]asi_num				;\
684	stxa	%i3, [%g5 + %g4]asi_num				;\
685	add	%g5, 32, %g5					;\
686	stxa	%i4, [%g5 + %g1]asi_num				;\
687	stxa	%i5, [%g5 + %g2]asi_num				;\
688	stxa	%i6, [%g5 + %g3]asi_num				;\
689	stxa	%i7, [%g5 + %g4]asi_num				;\
690	TT_TRACE_L(trace_win)					;\
691	b	.spill_clean					;\
692	  mov	WSTATE_USER64, %g7				;\
693	SKIP(31-25-TT_TRACE_L_INS)				;\
694	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
695	.empty
696
697
698/*
699 * Floating point disabled.
700 */
701#define	FP_DISABLED_TRAP		\
702	TT_TRACE(trace_gen)		;\
703	ba,pt	%xcc,.fp_disabled	;\
704	nop				;\
705	.align	32
706
707/*
708 * Floating point exceptions.
709 */
710#define	FP_IEEE_TRAP			\
711	TT_TRACE(trace_gen)		;\
712	ba,pt	%xcc,.fp_ieee_exception	;\
713	nop				;\
714	.align	32
715
716#define	FP_TRAP				\
717	TT_TRACE(trace_gen)		;\
718	ba,pt	%xcc,.fp_exception	;\
719	nop				;\
720	.align	32
721
722#if !defined(lint)
723
724/*
725 * ECACHE_ECC error traps at level 0 and level 1
726 */
727#define	ECACHE_ECC(table_name)		\
728	.global	table_name		;\
729table_name:				;\
730	membar	#Sync			;\
731	set	trap, %g1		;\
732	rdpr	%tt, %g3		;\
733	ba,pt	%xcc, sys_trap		;\
734	sub	%g0, 1, %g4		;\
735	.align	32
736
737#endif /* !lint */
738
739/*
740 * illegal instruction trap
741 */
742#define	ILLTRAP_INSTR			  \
743	membar	#Sync			  ;\
744	TT_TRACE(trace_gen)		  ;\
745	or	%g0, P_UTRAP4, %g2	  ;\
746	or	%g0, T_UNIMP_INSTR, %g3   ;\
747	sethi	%hi(.check_v9utrap), %g4  ;\
748	jmp	%g4 + %lo(.check_v9utrap) ;\
749	nop				  ;\
750	.align	32
751
752/*
753 * tag overflow trap
754 */
755#define	TAG_OVERFLOW			  \
756	TT_TRACE(trace_gen)		  ;\
757	or	%g0, P_UTRAP10, %g2	  ;\
758	or	%g0, T_TAG_OVERFLOW, %g3  ;\
759	sethi	%hi(.check_v9utrap), %g4  ;\
760	jmp	%g4 + %lo(.check_v9utrap) ;\
761	nop				  ;\
762	.align	32
763
764/*
765 * divide by zero trap
766 */
767#define	DIV_BY_ZERO			  \
768	TT_TRACE(trace_gen)		  ;\
769	or	%g0, P_UTRAP11, %g2	  ;\
770	or	%g0, T_IDIV0, %g3	  ;\
771	sethi	%hi(.check_v9utrap), %g4  ;\
772	jmp	%g4 + %lo(.check_v9utrap) ;\
773	nop				  ;\
774	.align	32
775
776/*
777 * trap instruction for V9 user trap handlers
778 */
779#define	TRAP_INSTR			  \
780	TT_TRACE(trace_gen)		  ;\
781	or	%g0, T_SOFTWARE_TRAP, %g3 ;\
782	sethi	%hi(.check_v9utrap), %g4  ;\
783	jmp	%g4 + %lo(.check_v9utrap) ;\
784	nop				  ;\
785	.align	32
786#define	TRP4	TRAP_INSTR; TRAP_INSTR; TRAP_INSTR; TRAP_INSTR
787
788/*
789 * LEVEL_INTERRUPT is for level N interrupts.
790 * VECTOR_INTERRUPT is for the vector trap.
791 */
792#define	LEVEL_INTERRUPT(level)		\
793	.global	tt_pil/**/level		;\
794tt_pil/**/level:			;\
795	ba,pt	%xcc, pil_interrupt	;\
796	mov	level, %g4		;\
797	.align	32
798
799#define	LEVEL14_INTERRUPT			\
800	ba	pil14_interrupt			;\
801	mov	PIL_14, %g4			;\
802	.align	32
803
804#define CPU_MONDO			\
805	ba,a,pt	%xcc, cpu_mondo		;\
806	.align	32
807
808#define DEV_MONDO			\
809	ba,a,pt	%xcc, dev_mondo		;\
810	.align	32
811
812/*
813 * We take over the rtba after we set our trap table and
814 * fault status area. The watchdog reset trap is now handled by the OS.
815 */
816#define WATCHDOG_RESET			\
817	mov	PTL1_BAD_WATCHDOG, %g1	;\
818	ba,a,pt	%xcc, .watchdog_trap	;\
819	.align	32
820
821/*
822 * RED is for traps that use the red mode handler.
823 * We should never see these either.
824 */
825#define RED			\
826	mov	PTL1_BAD_RED, %g1	;\
827	ba,a,pt	%xcc, .watchdog_trap	;\
828	.align	32
829
830
831/*
832 * MMU Trap Handlers.
833 */
834
835/*
836 * synthesize for trap(): SFSR in %g3
837 */
838#define	IMMU_EXCEPTION							\
839	MMU_FAULT_STATUS_AREA(%g3)					;\
840	rdpr	%tpc, %g2						;\
841	ldx	[%g3 + MMFSA_I_TYPE], %g1				;\
842	ldx	[%g3 + MMFSA_I_CTX], %g3				;\
843	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
844	or	%g3, %g1, %g3						;\
845	ba,pt	%xcc, .mmu_exception_end				;\
846	mov	T_INSTR_EXCEPTION, %g1					;\
847	.align	32
848
849/*
850 * synthesize for trap(): TAG_ACCESS in %g2, SFSR in %g3
851 */
852#define	DMMU_EXCEPTION							\
853	ba,a,pt	%xcc, .dmmu_exception					;\
854	.align	32
855
856/*
857 * synthesize for trap(): SFAR in %g2, SFSR in %g3
858 */
859#define	DMMU_EXC_AG_PRIV						\
860	MMU_FAULT_STATUS_AREA(%g3)					;\
861	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
862	/* Fault type not available in MMU fault status area */		;\
863	mov	MMFSA_F_PRVACT, %g1					;\
864	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
865	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
866	ba,pt	%xcc, .mmu_priv_exception				;\
867	or	%g3, %g1, %g3						;\
868	.align	32
869
870/*
871 * synthesize for trap(): SFAR in %g2, SFSR in %g3
872 */
873#define	DMMU_EXC_AG_NOT_ALIGNED						\
874	MMU_FAULT_STATUS_AREA(%g3)					;\
875	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
876	/* Fault type not available in MMU fault status area */		;\
877	mov	MMFSA_F_UNALIGN, %g1					;\
878	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
879	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
880	ba,pt	%xcc, .mmu_exception_not_aligned			;\
881	or	%g3, %g1, %g3			/* SFSR */		;\
882	.align	32
883/*
884 * SPARC V9 IMPL. DEP. #109(1) and (2) and #110(1) and (2)
885 */
886
887/*
888 * synthesize for trap(): SFAR in %g2, SFSR in %g3
889 */
890#define	DMMU_EXC_LDDF_NOT_ALIGNED					\
891	ba,a,pt	%xcc, .dmmu_exc_lddf_not_aligned			;\
892	.align	32
893/*
894 * synthesize for trap(): SFAR in %g2, SFSR in %g3
895 */
896#define	DMMU_EXC_STDF_NOT_ALIGNED					\
897	ba,a,pt	%xcc, .dmmu_exc_stdf_not_aligned			;\
898	.align	32
899
900#if TAGACC_CTX_MASK != CTXREG_CTX_MASK
901#error "TAGACC_CTX_MASK != CTXREG_CTX_MASK"
902#endif
903
904#if defined(cscope)
905/*
906 * Define labels to direct cscope quickly to labels that
907 * are generated by macro expansion of DTLB_MISS().
908 */
909	.global	tt0_dtlbmiss
910tt0_dtlbmiss:
911	.global	tt1_dtlbmiss
912tt1_dtlbmiss:
913	nop
914#endif
915
916/*
917 * Data miss handler (must be exactly 32 instructions)
918 *
919 * This handler is invoked only if the hypervisor has been instructed
920 * not to do any TSB walk.
921 *
922 * Kernel and invalid context cases are handled by the sfmmu_kdtlb_miss
923 * handler.
924 *
925 * User TLB miss handling depends upon whether a user process has one or
926 * two TSBs. User TSB information (physical base and size code) is kept
927 * in two dedicated scratchpad registers. Absence of a user TSB (primarily
928 * second TSB) is indicated by a negative value (-1) in that register.
929 */
930
931/*
932 * synthesize for miss handler: TAG_ACCESS in %g2
933 */
934#define	DTLB_MISS(table_name)						;\
935	.global	table_name/**/_dtlbmiss					;\
936table_name/**/_dtlbmiss:						;\
937	HAT_PERCPU_DBSTAT(TSBMISS_DTLBMISS) /* 3 instr ifdef DEBUG */	;\
938	MMU_FAULT_STATUS_AREA(%g7)					;\
939	ldx	[%g7 + MMFSA_D_ADDR], %g2	/* address */		;\
940	ldx	[%g7 + MMFSA_D_CTX], %g3	/* g3 = ctx */		;\
941	srlx	%g2, MMU_PAGESHIFT, %g2		/* align address */	;\
942	sllx	%g2, MMU_PAGESHIFT, %g2					;\
943	or	%g2, %g3, %g2			/* TAG_ACCESS */	;\
944	cmp	%g3, INVALID_CONTEXT					;\
945	ble,pn	%xcc, sfmmu_kdtlb_miss					;\
946	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
947	mov	SCRATCHPAD_UTSBREG2, %g1				;\
948	ldxa	[%g1]ASI_SCRATCHPAD, %g1	/* get 2nd tsbreg */	;\
949	brgez,pn %g1, sfmmu_udtlb_slowpath	/* brnach if 2 TSBs */	;\
950	  nop								;\
951	GET_1ST_TSBE_PTR(%g2, %g1, %g4, %g5)	/* 11 instr */		;\
952	ba,pt	%xcc, sfmmu_udtlb_fastpath	/* no 4M TSB, miss */	;\
953	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
954	.align 128
955
956
957#if defined(cscope)
958/*
959 * Define labels to direct cscope quickly to labels that
960 * are generated by macro expansion of ITLB_MISS().
961 */
962	.global	tt0_itlbmiss
963tt0_itlbmiss:
964	.global	tt1_itlbmiss
965tt1_itlbmiss:
966	nop
967#endif
968
969/*
970 * Instruction miss handler.
971 *
972 * This handler is invoked only if the hypervisor has been instructed
973 * not to do any TSB walk.
974 *
975 * ldda instructions will have their ASI patched
976 * by sfmmu_patch_ktsb at runtime.
977 * MUST be EXACTLY 32 instructions or we'll break.
978 */
979
980/*
981 * synthesize for miss handler: TAG_ACCESS in %g2
982 */
983#define	ITLB_MISS(table_name)						 \
984	.global	table_name/**/_itlbmiss					;\
985table_name/**/_itlbmiss:						;\
986	HAT_PERCPU_DBSTAT(TSBMISS_ITLBMISS) /* 3 instr ifdef DEBUG */	;\
987	MMU_FAULT_STATUS_AREA(%g7)					;\
988	ldx	[%g7 + MMFSA_I_ADDR], %g2	/* g2 = address */	;\
989	ldx	[%g7 + MMFSA_I_CTX], %g3	/* g3 = ctx */		;\
990	srlx	%g2, MMU_PAGESHIFT, %g2		/* align address */	;\
991	sllx	%g2, MMU_PAGESHIFT, %g2					;\
992	or	%g2, %g3, %g2			/* TAG_ACCESS */	;\
993	cmp	%g3, INVALID_CONTEXT					;\
994	ble,pn	%xcc, sfmmu_kitlb_miss					;\
995	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
996	mov	SCRATCHPAD_UTSBREG2, %g1				;\
997	ldxa	[%g1]ASI_SCRATCHPAD, %g1	/* get 2nd tsbreg */	;\
998	brgez,pn %g1, sfmmu_uitlb_slowpath	/* branch if 2 TSBS */	;\
999	  nop								;\
1000	GET_1ST_TSBE_PTR(%g2, %g1, %g4, %g5)	/* 11 instr */		;\
1001	ba,pt	%xcc, sfmmu_uitlb_fastpath	/* no 4M TSB, miss */	;\
1002	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
1003	.align 128
1004
1005#define	DTSB_MISS \
1006	GOTO_TT(sfmmu_slow_dmmu_miss,trace_dmmu)
1007
1008#define	ITSB_MISS \
1009	GOTO_TT(sfmmu_slow_immu_miss,trace_immu)
1010
1011/*
1012 * This macro is the first level handler for fast protection faults.
1013 * It first demaps the tlb entry which generated the fault and then
1014 * attempts to set the modify bit on the hash.  It needs to be
1015 * exactly 32 instructions.
1016 */
1017/*
1018 * synthesize for miss handler: TAG_ACCESS in %g2
1019 */
1020#define	DTLB_PROT							 \
1021	MMU_FAULT_STATUS_AREA(%g7)					;\
1022	ldx	[%g7 + MMFSA_D_ADDR], %g2	/* address */		;\
1023	ldx	[%g7 + MMFSA_D_CTX], %g3	/* %g3 = ctx */		;\
1024	srlx	%g2, MMU_PAGESHIFT, %g2		/* align address */	;\
1025	sllx	%g2, MMU_PAGESHIFT, %g2					;\
1026	or	%g2, %g3, %g2			/* TAG_ACCESS */	;\
1027	/*								;\
1028	 *   g2 = tag access register					;\
1029	 *   g3 = ctx number						;\
1030	 */								;\
1031	TT_TRACE(trace_dataprot)	/* 2 instr ifdef TRAPTRACE */	;\
1032					/* clobbers g1 and g6 XXXQ? */	;\
1033	brnz,pt %g3, sfmmu_uprot_trap		/* user trap */		;\
1034	  nop								;\
1035	ba,a,pt	%xcc, sfmmu_kprot_trap		/* kernel trap */	;\
1036	.align 128
1037
1038#define	DMMU_EXCEPTION_TL1						;\
1039	ba,a,pt	%xcc, mmu_trap_tl1					;\
1040	.align 32
1041
1042#define	MISALIGN_ADDR_TL1						;\
1043	ba,a,pt	%xcc, mmu_trap_tl1					;\
1044	.align 32
1045
1046/*
1047 * Trace a tsb hit
1048 * g1 = tsbe pointer (in/clobbered)
1049 * g2 = tag access register (in)
1050 * g3 - g4 = scratch (clobbered)
1051 * g5 = tsbe data (in)
1052 * g6 = scratch (clobbered)
1053 * g7 = pc we jumped here from (in)
1054 * ttextra = value to OR in to trap type (%tt) (in)
1055 */
1056#ifdef TRAPTRACE
1057#define TRACE_TSBHIT(ttextra)						 \
1058	membar	#Sync							;\
1059	sethi	%hi(FLUSH_ADDR), %g6					;\
1060	flush	%g6							;\
1061	TRACE_PTR(%g3, %g6)						;\
1062	GET_TRACE_TICK(%g6)						;\
1063	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi				;\
1064	stna	%g2, [%g3 + TRAP_ENT_SP]%asi	/* tag access */	;\
1065	stna	%g5, [%g3 + TRAP_ENT_F1]%asi	/* tsb data */		;\
1066	rdpr	%tnpc, %g6						;\
1067	stna	%g6, [%g3 + TRAP_ENT_F2]%asi				;\
1068	stna	%g1, [%g3 + TRAP_ENT_F3]%asi	/* tsb pointer */	;\
1069	stna	%g0, [%g3 + TRAP_ENT_F4]%asi				;\
1070	rdpr	%tpc, %g6						;\
1071	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi				;\
1072	TRACE_SAVE_TL_GL_REGS(%g3, %g6)					;\
1073	rdpr	%tt, %g6						;\
1074	or	%g6, (ttextra), %g1					;\
1075	stha	%g1, [%g3 + TRAP_ENT_TT]%asi				;\
1076	MMU_FAULT_STATUS_AREA(%g4)					;\
1077	mov	MMFSA_D_ADDR, %g1					;\
1078	cmp	%g6, FAST_IMMU_MISS_TT					;\
1079	move	%xcc, MMFSA_I_ADDR, %g1					;\
1080	cmp	%g6, T_INSTR_MMU_MISS					;\
1081	move	%xcc, MMFSA_I_ADDR, %g1					;\
1082	ldx	[%g4 + %g1], %g1					;\
1083	stxa	%g1, [%g3 + TRAP_ENT_TSTATE]%asi /* fault addr */	;\
1084	mov	MMFSA_D_CTX, %g1					;\
1085	cmp	%g6, FAST_IMMU_MISS_TT					;\
1086	move	%xcc, MMFSA_I_CTX, %g1					;\
1087	cmp	%g6, T_INSTR_MMU_MISS					;\
1088	move	%xcc, MMFSA_I_CTX, %g1					;\
1089	ldx	[%g4 + %g1], %g1					;\
1090	stna	%g1, [%g3 + TRAP_ENT_TR]%asi				;\
1091	TRACE_NEXT(%g3, %g4, %g6)
1092#else
1093#define TRACE_TSBHIT(ttextra)
1094#endif
1095
1096
1097#if defined(lint)
1098
1099struct scb	trap_table;
1100struct scb	scb;		/* trap_table/scb are the same object */
1101
1102#else /* lint */
1103
1104/*
1105 * =======================================================================
1106 *		SPARC V9 TRAP TABLE
1107 *
1108 * The trap table is divided into two halves: the first half is used when
1109 * taking traps when TL=0; the second half is used when taking traps from
1110 * TL>0. Note that handlers in the second half of the table might not be able
1111 * to make the same assumptions as handlers in the first half of the table.
1112 *
1113 * Worst case trap nesting so far:
1114 *
1115 *	at TL=0 client issues software trap requesting service
1116 *	at TL=1 nucleus wants a register window
1117 *	at TL=2 register window clean/spill/fill takes a TLB miss
1118 *	at TL=3 processing TLB miss
1119 *	at TL=4 handle asynchronous error
1120 *
1121 * Note that a trap from TL=4 to TL=5 places Spitfire in "RED mode".
1122 *
1123 * =======================================================================
1124 */
1125	.section ".text"
1126	.align	4
1127	.global trap_table, scb, trap_table0, trap_table1, etrap_table
1128	.type	trap_table, #object
1129	.type	trap_table0, #object
1130	.type	trap_table1, #object
1131	.type	scb, #object
1132trap_table:
1133scb:
1134trap_table0:
1135	/* hardware traps */
1136	NOT;				/* 000	reserved */
1137	RED;				/* 001	power on reset */
1138	WATCHDOG_RESET;			/* 002	watchdog reset */
1139	RED;				/* 003	externally initiated reset */
1140	RED;				/* 004	software initiated reset */
1141	RED;				/* 005	red mode exception */
1142	NOT; NOT;			/* 006 - 007 reserved */
1143	IMMU_EXCEPTION;			/* 008	instruction access exception */
1144	ITSB_MISS;			/* 009	instruction access MMU miss */
1145 	NOT;				/* 00A  reserved */
1146	NOT; NOT4;			/* 00B - 00F reserved */
1147	ILLTRAP_INSTR;			/* 010	illegal instruction */
1148	TRAP(T_PRIV_INSTR);		/* 011	privileged opcode */
1149	TRAP(T_UNIMP_LDD);		/* 012	unimplemented LDD */
1150	TRAP(T_UNIMP_STD);		/* 013	unimplemented STD */
1151	NOT4; NOT4; NOT4;		/* 014 - 01F reserved */
1152	FP_DISABLED_TRAP;		/* 020	fp disabled */
1153	FP_IEEE_TRAP;			/* 021	fp exception ieee 754 */
1154	FP_TRAP;			/* 022	fp exception other */
1155	TAG_OVERFLOW;			/* 023	tag overflow */
1156	CLEAN_WINDOW;			/* 024 - 027 clean window */
1157	DIV_BY_ZERO;			/* 028	division by zero */
1158	NOT;				/* 029	internal processor error */
1159	NOT; NOT; NOT4;			/* 02A - 02F reserved */
1160	DMMU_EXCEPTION;			/* 030	data access exception */
1161	DTSB_MISS;			/* 031	data access MMU miss */
1162	NOT;				/* 032  reserved */
1163	NOT;				/* 033	data access protection */
1164	DMMU_EXC_AG_NOT_ALIGNED;	/* 034	mem address not aligned */
1165	DMMU_EXC_LDDF_NOT_ALIGNED;	/* 035	LDDF mem address not aligned */
1166	DMMU_EXC_STDF_NOT_ALIGNED;	/* 036	STDF mem address not aligned */
1167	DMMU_EXC_AG_PRIV;		/* 037	privileged action */
1168	NOT;				/* 038	LDQF mem address not aligned */
1169	NOT;				/* 039	STQF mem address not aligned */
1170	NOT; NOT; NOT4;			/* 03A - 03F reserved */
1171	NOT;				/* 040	async data error */
1172	LEVEL_INTERRUPT(1);		/* 041	interrupt level 1 */
1173	LEVEL_INTERRUPT(2);		/* 042	interrupt level 2 */
1174	LEVEL_INTERRUPT(3);		/* 043	interrupt level 3 */
1175	LEVEL_INTERRUPT(4);		/* 044	interrupt level 4 */
1176	LEVEL_INTERRUPT(5);		/* 045	interrupt level 5 */
1177	LEVEL_INTERRUPT(6);		/* 046	interrupt level 6 */
1178	LEVEL_INTERRUPT(7);		/* 047	interrupt level 7 */
1179	LEVEL_INTERRUPT(8);		/* 048	interrupt level 8 */
1180	LEVEL_INTERRUPT(9);		/* 049	interrupt level 9 */
1181	LEVEL_INTERRUPT(10);		/* 04A	interrupt level 10 */
1182	LEVEL_INTERRUPT(11);		/* 04B	interrupt level 11 */
1183	LEVEL_INTERRUPT(12);		/* 04C	interrupt level 12 */
1184	LEVEL_INTERRUPT(13);		/* 04D	interrupt level 13 */
1185	LEVEL14_INTERRUPT;		/* 04E	interrupt level 14 */
1186	LEVEL_INTERRUPT(15);		/* 04F	interrupt level 15 */
1187	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F reserved */
1188	NOT;				/* 060	interrupt vector */
1189	GOTO(kmdb_trap);		/* 061	PA watchpoint */
1190	GOTO(kmdb_trap);		/* 062	VA watchpoint */
1191	NOT;				/* 063	reserved */
1192	ITLB_MISS(tt0);			/* 064	instruction access MMU miss */
1193	DTLB_MISS(tt0);			/* 068	data access MMU miss */
1194	DTLB_PROT;			/* 06C	data access protection */
1195	NOT;				/* 070  reserved */
1196	NOT;				/* 071  reserved */
1197	NOT;				/* 072  reserved */
1198	NOT;				/* 073  reserved */
1199	NOT4; NOT4			/* 074 - 07B reserved */
1200	CPU_MONDO;			/* 07C	cpu_mondo */
1201	DEV_MONDO;			/* 07D	dev_mondo */
1202	GOTO_TT(resumable_error, trace_gen);	/* 07E  resumable error */
1203	GOTO_TT(nonresumable_error, trace_gen);	/* 07F  non-reasumable error */
1204	NOT4;				/* 080	spill 0 normal */
1205	SPILL_32bit_asi(ASI_AIUP,sn0);	/* 084	spill 1 normal */
1206	SPILL_64bit_asi(ASI_AIUP,sn0);	/* 088	spill 2 normal */
1207	SPILL_32clean(ASI_AIUP,sn0);	/* 08C	spill 3 normal */
1208	SPILL_64clean(ASI_AIUP,sn0);	/* 090	spill 4 normal */
1209	SPILL_32bit(not);		/* 094	spill 5 normal */
1210	SPILL_64bit(not);		/* 098	spill 6 normal */
1211	SPILL_mixed;			/* 09C	spill 7 normal */
1212	NOT4;				/* 0A0	spill 0 other */
1213	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0A4	spill 1 other */
1214	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0A8	spill 2 other */
1215	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0AC	spill 3 other */
1216	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0B0	spill 4 other */
1217	NOT4;				/* 0B4	spill 5 other */
1218	NOT4;				/* 0B8	spill 6 other */
1219	NOT4;				/* 0BC	spill 7 other */
1220	NOT4;				/* 0C0	fill 0 normal */
1221	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0C4	fill 1 normal */
1222	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0C8	fill 2 normal */
1223	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0CC	fill 3 normal */
1224	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0D0	fill 4 normal */
1225	FILL_32bit(not);		/* 0D4	fill 5 normal */
1226	FILL_64bit(not);		/* 0D8	fill 6 normal */
1227	FILL_mixed;			/* 0DC	fill 7 normal */
1228	NOT4;				/* 0E0	fill 0 other */
1229	NOT4;				/* 0E4	fill 1 other */
1230	NOT4;				/* 0E8	fill 2 other */
1231	NOT4;				/* 0EC	fill 3 other */
1232	NOT4;				/* 0F0	fill 4 other */
1233	NOT4;				/* 0F4	fill 5 other */
1234	NOT4;				/* 0F8	fill 6 other */
1235	NOT4;				/* 0FC	fill 7 other */
1236	/* user traps */
1237	GOTO(syscall_trap_4x);		/* 100	old system call */
1238	TRAP(T_BREAKPOINT);		/* 101	user breakpoint */
1239	TRAP(T_DIV0);			/* 102	user divide by zero */
1240	GOTO(.flushw);			/* 103	flush windows */
1241	GOTO(.clean_windows);		/* 104	clean windows */
1242	BAD;				/* 105	range check ?? */
1243	GOTO(.fix_alignment);		/* 106	do unaligned references */
1244	BAD;				/* 107	unused */
1245	SYSCALL(syscall_trap32);	/* 108	ILP32 system call on LP64 */
1246	GOTO(set_trap0_addr);		/* 109	set trap0 address */
1247	BAD; BAD; BAD4;			/* 10A - 10F unused */
1248	TRP4; TRP4; TRP4; TRP4;		/* 110 - 11F V9 user trap handlers */
1249	GOTO(.getcc);			/* 120	get condition codes */
1250	GOTO(.setcc);			/* 121	set condition codes */
1251	GOTO(.getpsr);			/* 122	get psr */
1252	GOTO(.setpsr);			/* 123	set psr (some fields) */
1253	GOTO(get_timestamp);		/* 124	get timestamp */
1254	GOTO(get_virtime);		/* 125	get lwp virtual time */
1255	PRIV(self_xcall);		/* 126	self xcall */
1256	GOTO(get_hrestime);		/* 127	get hrestime */
1257	BAD;				/* 128	ST_SETV9STACK */
1258	GOTO(.getlgrp);			/* 129  get lgrpid */
1259	BAD; BAD; BAD4;			/* 12A - 12F unused */
1260	BAD4; BAD4; 			/* 130 - 137 unused */
1261	DTRACE_PID;			/* 138  dtrace pid tracing provider */
1262	BAD;				/* 139  unused */
1263	DTRACE_RETURN;			/* 13A	dtrace pid return probe */
1264	BAD; BAD4;			/* 13B - 13F unused */
1265	SYSCALL(syscall_trap)		/* 140  LP64 system call */
1266	SYSCALL(nosys);			/* 141  unused system call trap */
1267#ifdef DEBUG_USER_TRAPTRACECTL
1268	GOTO(.traptrace_freeze);	/* 142  freeze traptrace */
1269	GOTO(.traptrace_unfreeze);	/* 143  unfreeze traptrace */
1270#else
1271	SYSCALL(nosys);			/* 142  unused system call trap */
1272	SYSCALL(nosys);			/* 143  unused system call trap */
1273#endif
1274	BAD4; BAD4; BAD4;		/* 144 - 14F unused */
1275	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1276	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1277	BAD;				/* 170 - unused */
1278	BAD;				/* 171 - unused */
1279	BAD; BAD;			/* 172 - 173 unused */
1280	BAD4; BAD4;			/* 174 - 17B unused */
1281#ifdef	PTL1_PANIC_DEBUG
1282	mov PTL1_BAD_DEBUG, %g1; GOTO(ptl1_panic);
1283					/* 17C	test ptl1_panic */
1284#else
1285	BAD;				/* 17C  unused */
1286#endif	/* PTL1_PANIC_DEBUG */
1287	PRIV(kmdb_trap);		/* 17D	kmdb enter (L1-A) */
1288	PRIV(kmdb_trap);		/* 17E	kmdb breakpoint */
1289	PRIV(obp_bpt);			/* 17F	obp breakpoint */
1290	/* reserved */
1291	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1292	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1293	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1294	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1295	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1296	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1297	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1298	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1299	.size	trap_table0, (.-trap_table0)
1300trap_table1:
1301	NOT4; NOT4;			/* 000 - 007 unused */
1302	NOT;				/* 008	instruction access exception */
1303	ITSB_MISS;			/* 009	instruction access MMU miss */
1304 	NOT;				/* 00A  reserved */
1305	NOT; NOT4;			/* 00B - 00F unused */
1306	NOT4; NOT4; NOT4; NOT4;		/* 010 - 01F unused */
1307	NOT4;				/* 020 - 023 unused */
1308	CLEAN_WINDOW;			/* 024 - 027 clean window */
1309	NOT4; NOT4;			/* 028 - 02F unused */
1310	DMMU_EXCEPTION_TL1;		/* 030 	data access exception */
1311	DTSB_MISS;			/* 031  data access MMU miss */
1312	NOT;				/* 032  reserved */
1313	NOT;				/* 033	unused */
1314	MISALIGN_ADDR_TL1;		/* 034	mem address not aligned */
1315	NOT; NOT; NOT; NOT4; NOT4	/* 035 - 03F unused */
1316	NOT4; NOT4; NOT4; NOT4;		/* 040 - 04F unused */
1317	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F unused */
1318	NOT;				/* 060	unused */
1319	GOTO(kmdb_trap_tl1);		/* 061	PA watchpoint */
1320	GOTO(kmdb_trap_tl1);		/* 062	VA watchpoint */
1321	NOT;				/* 063	reserved */
1322	ITLB_MISS(tt1);			/* 064	instruction access MMU miss */
1323	DTLB_MISS(tt1);			/* 068	data access MMU miss */
1324	DTLB_PROT;			/* 06C	data access protection */
1325	NOT;				/* 070  reserved */
1326	NOT;				/* 071  reserved */
1327	NOT;				/* 072  reserved */
1328	NOT;				/* 073  reserved */
1329	NOT4; NOT4;			/* 074 - 07B reserved */
1330	NOT;				/* 07C  reserved */
1331	NOT;				/* 07D  reserved */
1332	NOT;				/* 07E  resumable error */
1333	GOTO_TT(nonresumable_error, trace_gen);	/* 07F  nonresumable error */
1334	NOTP4;				/* 080	spill 0 normal */
1335	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 084	spill 1 normal */
1336	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 088	spill 2 normal */
1337	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 08C	spill 3 normal */
1338	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 090	spill 4 normal */
1339	NOTP4;				/* 094	spill 5 normal */
1340	SPILL_64bit_ktt1(sk);		/* 098	spill 6 normal */
1341	SPILL_mixed_ktt1(sk);		/* 09C	spill 7 normal */
1342	NOTP4;				/* 0A0	spill 0 other */
1343	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0A4  spill 1 other */
1344	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0A8	spill 2 other */
1345	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0AC	spill 3 other */
1346	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0B0  spill 4 other */
1347	NOTP4;				/* 0B4  spill 5 other */
1348	NOTP4;				/* 0B8  spill 6 other */
1349	NOTP4;				/* 0BC  spill 7 other */
1350	NOT4;				/* 0C0	fill 0 normal */
1351	NOT4;				/* 0C4	fill 1 normal */
1352	NOT4;				/* 0C8	fill 2 normal */
1353	NOT4;				/* 0CC	fill 3 normal */
1354	NOT4;				/* 0D0	fill 4 normal */
1355	NOT4;				/* 0D4	fill 5 normal */
1356	NOT4;				/* 0D8	fill 6 normal */
1357	NOT4;				/* 0DC	fill 7 normal */
1358	NOT4; NOT4; NOT4; NOT4;		/* 0E0 - 0EF unused */
1359	NOT4; NOT4; NOT4; NOT4;		/* 0F0 - 0FF unused */
1360/*
1361 * Code running at TL>0 does not use soft traps, so
1362 * we can truncate the table here.
1363 * However:
1364 * sun4v uses (hypervisor) ta instructions at TL > 0, so
1365 * provide a safety net for now.
1366 */
1367	/* soft traps */
1368	BAD4; BAD4; BAD4; BAD4;		/* 100 - 10F unused */
1369	BAD4; BAD4; BAD4; BAD4;		/* 110 - 11F unused */
1370	BAD4; BAD4; BAD4; BAD4;		/* 120 - 12F unused */
1371	BAD4; BAD4; BAD4; BAD4;		/* 130 - 13F unused */
1372	BAD4; BAD4; BAD4; BAD4;		/* 140 - 14F unused */
1373	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1374	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1375	BAD4; BAD4; BAD4; BAD4;		/* 170 - 17F unused */
1376	/* reserved */
1377	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1378	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1379	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1380	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1381	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1382	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1383	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1384	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1385etrap_table:
1386	.size	trap_table1, (.-trap_table1)
1387	.size	trap_table, (.-trap_table)
1388	.size	scb, (.-scb)
1389
1390/*
1391 * We get to exec_fault in the case of an instruction miss and tte
1392 * has no execute bit set.  We go to tl0 to handle it.
1393 *
1394 * g1 = tsbe pointer (in/clobbered)
1395 * g2 = tag access register (in)
1396 * g3 - g4 = scratch (clobbered)
1397 * g5 = tsbe data (in)
1398 * g6 = scratch (clobbered)
1399 * g7 = pc we jumped here from (in)
1400 */
1401/*
1402 * synthesize for trap(): TAG_ACCESS in %g2
1403 */
1404	ALTENTRY(exec_fault)
1405	TRACE_TSBHIT(TT_MMU_EXEC)
1406	MMU_FAULT_STATUS_AREA(%g4)
1407	ldx	[%g4 + MMFSA_I_ADDR], %g2	/* g2 = address */
1408	ldx	[%g4 + MMFSA_I_CTX], %g3	/* g3 = ctx */
1409	srlx	%g2, MMU_PAGESHIFT, %g2		! align address to page boundry
1410	sllx	%g2, MMU_PAGESHIFT, %g2
1411	or	%g2, %g3, %g2			/* TAG_ACCESS */
1412	mov	T_INSTR_MMU_MISS, %g3		! arg2 = traptype
1413	set	trap, %g1
1414	ba,pt	%xcc, sys_trap
1415	  mov	-1, %g4
1416
1417.mmu_exception_not_aligned:
1418	/* %g2 = sfar, %g3 = sfsr */
1419	rdpr	%tstate, %g1
1420	btst	TSTATE_PRIV, %g1
1421	bnz,pn	%icc, 2f
1422	nop
1423	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1424	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1425	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1426	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1427	brz,pt	%g5, 2f
1428	nop
1429	ldn	[%g5 + P_UTRAP15], %g5			! unaligned utrap?
1430	brz,pn	%g5, 2f
1431	nop
1432	btst	1, %sp
1433	bz,pt	%xcc, 1f				! 32 bit user program
1434	nop
1435	ba,pt	%xcc, .setup_v9utrap			! 64 bit user program
1436	nop
14371:
1438	ba,pt	%xcc, .setup_utrap
1439	or	%g2, %g0, %g7
14402:
1441	ba,pt	%xcc, .mmu_exception_end
1442	mov	T_ALIGNMENT, %g1
1443
1444.mmu_priv_exception:
1445	rdpr	%tstate, %g1
1446	btst	TSTATE_PRIV, %g1
1447	bnz,pn	%icc, 1f
1448	nop
1449	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1450	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1451	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1452	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1453	brz,pt	%g5, 1f
1454	nop
1455	ldn	[%g5 + P_UTRAP16], %g5
1456	brnz,pt	%g5, .setup_v9utrap
1457	nop
14581:
1459	mov	T_PRIV_INSTR, %g1
1460
1461.mmu_exception_end:
1462	CPU_INDEX(%g4, %g5)
1463	set	cpu_core, %g5
1464	sllx	%g4, CPU_CORE_SHIFT, %g4
1465	add	%g4, %g5, %g4
1466	lduh	[%g4 + CPUC_DTRACE_FLAGS], %g5
1467	andcc	%g5, CPU_DTRACE_NOFAULT, %g0
1468	bz	1f
1469	or	%g5, CPU_DTRACE_BADADDR, %g5
1470	stuh	%g5, [%g4 + CPUC_DTRACE_FLAGS]
1471	done
1472
14731:
1474	sllx	%g3, 32, %g3
1475	or	%g3, %g1, %g3
1476	set	trap, %g1
1477	ba,pt	%xcc, sys_trap
1478	sub	%g0, 1, %g4
1479
1480.fp_disabled:
1481	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1482	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1483	rdpr	%tstate, %g4
1484	btst	TSTATE_PRIV, %g4
1485	bnz,a,pn %icc, ptl1_panic
1486	  mov	PTL1_BAD_FPTRAP, %g1
1487
1488	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1489	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1490	brz,a,pt %g5, 2f
1491	  nop
1492	ldn	[%g5 + P_UTRAP7], %g5			! fp_disabled utrap?
1493	brz,a,pn %g5, 2f
1494	  nop
1495	btst	1, %sp
1496	bz,a,pt	%xcc, 1f				! 32 bit user program
1497	  nop
1498	ba,a,pt	%xcc, .setup_v9utrap			! 64 bit user program
1499	  nop
15001:
1501	ba,pt	%xcc, .setup_utrap
1502	  or	%g0, %g0, %g7
15032:
1504	set	fp_disabled, %g1
1505	ba,pt	%xcc, sys_trap
1506	  sub	%g0, 1, %g4
1507
1508.fp_ieee_exception:
1509	rdpr	%tstate, %g1
1510	btst	TSTATE_PRIV, %g1
1511	bnz,a,pn %icc, ptl1_panic
1512	  mov	PTL1_BAD_FPTRAP, %g1
1513	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1514	stx	%fsr, [%g1 + CPU_TMP1]
1515	ldx	[%g1 + CPU_TMP1], %g2
1516	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1517	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1518	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1519	brz,a,pt %g5, 1f
1520	  nop
1521	ldn	[%g5 + P_UTRAP8], %g5
1522	brnz,a,pt %g5, .setup_v9utrap
1523	  nop
15241:
1525	set	_fp_ieee_exception, %g1
1526	ba,pt	%xcc, sys_trap
1527	  sub	%g0, 1, %g4
1528
1529/*
1530 * Register Inputs:
1531 *	%g5		user trap handler
1532 *	%g7		misaligned addr - for alignment traps only
1533 */
1534.setup_utrap:
1535	set	trap, %g1			! setup in case we go
1536	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1537	sub	%g0, 1, %g4			! the save instruction below
1538
1539	/*
1540	 * If the DTrace pid provider is single stepping a copied-out
1541	 * instruction, t->t_dtrace_step will be set. In that case we need
1542	 * to abort the single-stepping (since execution of the instruction
1543	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1544	 */
1545	save	%sp, -SA(MINFRAME32), %sp	! window for trap handler
1546	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1547	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1548	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1549	rdpr	%tnpc, %l2			! arg1 == tnpc
1550	brz,pt	%g2, 1f
1551	rdpr	%tpc, %l1			! arg0 == tpc
1552
1553	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1554	ldn	[%g1 + T_DTRACE_NPC], %l2	! arg1 = t->t_dtrace_npc (step)
1555	brz,pt	%g2, 1f
1556	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1557	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
15581:
1559	mov	%g7, %l3			! arg2 == misaligned address
1560
1561	rdpr	%tstate, %g1			! cwp for trap handler
1562	rdpr	%cwp, %g4
1563	bclr	TSTATE_CWP_MASK, %g1
1564	wrpr	%g1, %g4, %tstate
1565	wrpr	%g0, %g5, %tnpc			! trap handler address
1566	FAST_TRAP_DONE
1567	/* NOTREACHED */
1568
1569.check_v9utrap:
1570	rdpr	%tstate, %g1
1571	btst	TSTATE_PRIV, %g1
1572	bnz,a,pn %icc, 3f
1573	  nop
1574	CPU_ADDR(%g4, %g1)				! load CPU struct addr
1575	ldn	[%g4 + CPU_THREAD], %g5			! load thread pointer
1576	ldn	[%g5 + T_PROCP], %g5			! load proc pointer
1577	ldn	[%g5 + P_UTRAPS], %g5			! are there utraps?
1578
1579	cmp	%g3, T_SOFTWARE_TRAP
1580	bne,a,pt %icc, 1f
1581	  nop
1582
1583	brz,pt %g5, 3f			! if p_utraps == NULL goto trap()
1584	  rdpr	%tt, %g3		! delay - get actual hw trap type
1585
1586	sub	%g3, 254, %g1		! UT_TRAP_INSTRUCTION_16 = p_utraps[18]
1587	ba,pt	%icc, 2f
1588	  smul	%g1, CPTRSIZE, %g2
15891:
1590	brz,a,pt %g5, 3f		! if p_utraps == NULL goto trap()
1591	  nop
1592
1593	cmp	%g3, T_UNIMP_INSTR
1594	bne,a,pt %icc, 2f
1595	  nop
1596
1597	mov	1, %g1
1598	st	%g1, [%g4 + CPU_TL1_HDLR] ! set CPU_TL1_HDLR
1599	rdpr	%tpc, %g1		! ld trapping instruction using
1600	lduwa	[%g1]ASI_AIUP, %g1	! "AS IF USER" ASI which could fault
1601	st	%g0, [%g4 + CPU_TL1_HDLR] ! clr CPU_TL1_HDLR
1602
1603	sethi	%hi(0xc1c00000), %g4	! setup mask for illtrap instruction
1604	andcc	%g1, %g4, %g4		! and instruction with mask
1605	bnz,a,pt %icc, 3f		! if %g4 == zero, %g1 is an ILLTRAP
1606	  nop				! fall thru to setup
16072:
1608	ldn	[%g5 + %g2], %g5
1609	brnz,a,pt %g5, .setup_v9utrap
1610	  nop
16113:
1612	set	trap, %g1
1613	ba,pt	%xcc, sys_trap
1614	  sub	%g0, 1, %g4
1615	/* NOTREACHED */
1616
1617/*
1618 * Register Inputs:
1619 *	%g5		user trap handler
1620 */
1621.setup_v9utrap:
1622	set	trap, %g1			! setup in case we go
1623	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1624	sub	%g0, 1, %g4			! the save instruction below
1625
1626	/*
1627	 * If the DTrace pid provider is single stepping a copied-out
1628	 * instruction, t->t_dtrace_step will be set. In that case we need
1629	 * to abort the single-stepping (since execution of the instruction
1630	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1631	 */
1632	save	%sp, -SA(MINFRAME64), %sp	! window for trap handler
1633	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1634	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1635	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1636	rdpr	%tnpc, %l7			! arg1 == tnpc
1637	brz,pt	%g2, 1f
1638	rdpr	%tpc, %l6			! arg0 == tpc
1639
1640	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1641	ldn	[%g1 + T_DTRACE_NPC], %l7	! arg1 == t->t_dtrace_npc (step)
1642	brz,pt	%g2, 1f
1643	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1644	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
16451:
1646	rdpr	%tstate, %g2			! cwp for trap handler
1647	rdpr	%cwp, %g4
1648	bclr	TSTATE_CWP_MASK, %g2
1649	wrpr	%g2, %g4, %tstate
1650
1651	ldn	[%g1 + T_PROCP], %g4		! load proc pointer
1652	ldn	[%g4 + P_AS], %g4		! load as pointer
1653	ldn	[%g4 + A_USERLIMIT], %g4	! load as userlimit
1654	cmp	%l7, %g4			! check for single-step set
1655	bne,pt	%xcc, 4f
1656	  nop
1657	ldn	[%g1 + T_LWP], %g1		! load klwp pointer
1658	ld	[%g1 + PCB_STEP], %g4		! load single-step flag
1659	cmp	%g4, STEP_ACTIVE		! step flags set in pcb?
1660	bne,pt	%icc, 4f
1661	  nop
1662	stn	%g5, [%g1 + PCB_TRACEPC]	! save trap handler addr in pcb
1663	mov	%l7, %g4			! on entry to precise user trap
1664	add	%l6, 4, %l7			! handler, %l6 == pc, %l7 == npc
1665						! at time of trap
1666	wrpr	%g0, %g4, %tnpc			! generate FLTBOUNDS,
1667						! %g4 == userlimit
1668	FAST_TRAP_DONE
1669	/* NOTREACHED */
16704:
1671	wrpr	%g0, %g5, %tnpc			! trap handler address
1672	FAST_TRAP_DONE_CHK_INTR
1673	/* NOTREACHED */
1674
1675.fp_exception:
1676	CPU_ADDR(%g1, %g4)
1677	stx	%fsr, [%g1 + CPU_TMP1]
1678	ldx	[%g1 + CPU_TMP1], %g2
1679
1680	/*
1681	 * Cheetah takes unfinished_FPop trap for certain range of operands
1682	 * to the "fitos" instruction. Instead of going through the slow
1683	 * software emulation path, we try to simulate the "fitos" instruction
1684	 * via "fitod" and "fdtos" provided the following conditions are met:
1685	 *
1686	 *	fpu_exists is set (if DEBUG)
1687	 *	not in privileged mode
1688	 *	ftt is unfinished_FPop
1689	 *	NXM IEEE trap is not enabled
1690	 *	instruction at %tpc is "fitos"
1691	 *
1692	 *  Usage:
1693	 *	%g1	per cpu address
1694	 *	%g2	%fsr
1695	 *	%g6	user instruction
1696	 *
1697	 * Note that we can take a memory access related trap while trying
1698	 * to fetch the user instruction. Therefore, we set CPU_TL1_HDLR
1699	 * flag to catch those traps and let the SFMMU code deal with page
1700	 * fault and data access exception.
1701	 */
1702#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
1703	sethi	%hi(fpu_exists), %g7
1704	ld	[%g7 + %lo(fpu_exists)], %g7
1705	brz,pn %g7, .fp_exception_cont
1706	  nop
1707#endif
1708	rdpr	%tstate, %g7			! branch if in privileged mode
1709	btst	TSTATE_PRIV, %g7
1710	bnz,pn	%xcc, .fp_exception_cont
1711	srl	%g2, FSR_FTT_SHIFT, %g7		! extract ftt from %fsr
1712	and	%g7, (FSR_FTT>>FSR_FTT_SHIFT), %g7
1713	cmp	%g7, FTT_UNFIN
1714	set	FSR_TEM_NX, %g5
1715	bne,pn	%xcc, .fp_exception_cont	! branch if NOT unfinished_FPop
1716	  andcc	%g2, %g5, %g0
1717	bne,pn	%xcc, .fp_exception_cont	! branch if FSR_TEM_NX enabled
1718	  rdpr	%tpc, %g5			! get faulting PC
1719
1720	or	%g0, 1, %g7
1721	st	%g7, [%g1 + CPU_TL1_HDLR]	! set tl1_hdlr flag
1722	lda	[%g5]ASI_USER, %g6		! get user's instruction
1723	st	%g0, [%g1 + CPU_TL1_HDLR]	! clear tl1_hdlr flag
1724
1725	set	FITOS_INSTR_MASK, %g7
1726	and	%g6, %g7, %g7
1727	set	FITOS_INSTR, %g5
1728	cmp	%g7, %g5
1729	bne,pn	%xcc, .fp_exception_cont	! branch if not FITOS_INSTR
1730	 nop
1731
1732	/*
1733	 * This is unfinished FPops trap for "fitos" instruction. We
1734	 * need to simulate "fitos" via "fitod" and "fdtos" instruction
1735	 * sequence.
1736	 *
1737	 * We need a temporary FP register to do the conversion. Since
1738	 * both source and destination operands for the "fitos" instruction
1739	 * have to be within %f0-%f31, we use an FP register from the upper
1740	 * half to guarantee that it won't collide with the source or the
1741	 * dest operand. However, we do have to save and restore its value.
1742	 *
1743	 * We use %d62 as a temporary FP register for the conversion and
1744	 * branch to appropriate instruction within the conversion tables
1745	 * based upon the rs2 and rd values.
1746	 */
1747
1748	std	%d62, [%g1 + CPU_TMP1]		! save original value
1749
1750	srl	%g6, FITOS_RS2_SHIFT, %g7
1751	and	%g7, FITOS_REG_MASK, %g7
1752	set	_fitos_fitod_table, %g4
1753	sllx	%g7, 2, %g7
1754	jmp	%g4 + %g7
1755	  ba,pt	%xcc, _fitos_fitod_done
1756	.empty
1757
1758_fitos_fitod_table:
1759	  fitod	%f0, %d62
1760	  fitod	%f1, %d62
1761	  fitod	%f2, %d62
1762	  fitod	%f3, %d62
1763	  fitod	%f4, %d62
1764	  fitod	%f5, %d62
1765	  fitod	%f6, %d62
1766	  fitod	%f7, %d62
1767	  fitod	%f8, %d62
1768	  fitod	%f9, %d62
1769	  fitod	%f10, %d62
1770	  fitod	%f11, %d62
1771	  fitod	%f12, %d62
1772	  fitod	%f13, %d62
1773	  fitod	%f14, %d62
1774	  fitod	%f15, %d62
1775	  fitod	%f16, %d62
1776	  fitod	%f17, %d62
1777	  fitod	%f18, %d62
1778	  fitod	%f19, %d62
1779	  fitod	%f20, %d62
1780	  fitod	%f21, %d62
1781	  fitod	%f22, %d62
1782	  fitod	%f23, %d62
1783	  fitod	%f24, %d62
1784	  fitod	%f25, %d62
1785	  fitod	%f26, %d62
1786	  fitod	%f27, %d62
1787	  fitod	%f28, %d62
1788	  fitod	%f29, %d62
1789	  fitod	%f30, %d62
1790	  fitod	%f31, %d62
1791_fitos_fitod_done:
1792
1793	/*
1794	 * Now convert data back into single precision
1795	 */
1796	srl	%g6, FITOS_RD_SHIFT, %g7
1797	and	%g7, FITOS_REG_MASK, %g7
1798	set	_fitos_fdtos_table, %g4
1799	sllx	%g7, 2, %g7
1800	jmp	%g4 + %g7
1801	  ba,pt	%xcc, _fitos_fdtos_done
1802	.empty
1803
1804_fitos_fdtos_table:
1805	  fdtos	%d62, %f0
1806	  fdtos	%d62, %f1
1807	  fdtos	%d62, %f2
1808	  fdtos	%d62, %f3
1809	  fdtos	%d62, %f4
1810	  fdtos	%d62, %f5
1811	  fdtos	%d62, %f6
1812	  fdtos	%d62, %f7
1813	  fdtos	%d62, %f8
1814	  fdtos	%d62, %f9
1815	  fdtos	%d62, %f10
1816	  fdtos	%d62, %f11
1817	  fdtos	%d62, %f12
1818	  fdtos	%d62, %f13
1819	  fdtos	%d62, %f14
1820	  fdtos	%d62, %f15
1821	  fdtos	%d62, %f16
1822	  fdtos	%d62, %f17
1823	  fdtos	%d62, %f18
1824	  fdtos	%d62, %f19
1825	  fdtos	%d62, %f20
1826	  fdtos	%d62, %f21
1827	  fdtos	%d62, %f22
1828	  fdtos	%d62, %f23
1829	  fdtos	%d62, %f24
1830	  fdtos	%d62, %f25
1831	  fdtos	%d62, %f26
1832	  fdtos	%d62, %f27
1833	  fdtos	%d62, %f28
1834	  fdtos	%d62, %f29
1835	  fdtos	%d62, %f30
1836	  fdtos	%d62, %f31
1837_fitos_fdtos_done:
1838
1839	ldd	[%g1 + CPU_TMP1], %d62		! restore %d62
1840
1841#if DEBUG
1842	/*
1843	 * Update FPop_unfinished trap kstat
1844	 */
1845	set	fpustat+FPUSTAT_UNFIN_KSTAT, %g7
1846	ldx	[%g7], %g5
18471:
1848	add	%g5, 1, %g6
1849
1850	casxa	[%g7] ASI_N, %g5, %g6
1851	cmp	%g5, %g6
1852	bne,a,pn %xcc, 1b
1853	  or	%g0, %g6, %g5
1854
1855	/*
1856	 * Update fpu_sim_fitos kstat
1857	 */
1858	set	fpuinfo+FPUINFO_FITOS_KSTAT, %g7
1859	ldx	[%g7], %g5
18601:
1861	add	%g5, 1, %g6
1862
1863	casxa	[%g7] ASI_N, %g5, %g6
1864	cmp	%g5, %g6
1865	bne,a,pn %xcc, 1b
1866	  or	%g0, %g6, %g5
1867#endif /* DEBUG */
1868
1869	FAST_TRAP_DONE
1870
1871.fp_exception_cont:
1872	/*
1873	 * Let _fp_exception deal with simulating FPop instruction.
1874	 * Note that we need to pass %fsr in %g2 (already read above).
1875	 */
1876
1877	set	_fp_exception, %g1
1878	ba,pt	%xcc, sys_trap
1879	sub	%g0, 1, %g4
1880
1881
1882/*
1883 * Register windows
1884 */
1885.flushw:
1886.clean_windows:
1887	rdpr	%tnpc, %g1
1888	wrpr	%g1, %tpc
1889	add	%g1, 4, %g1
1890	wrpr	%g1, %tnpc
1891	set	trap, %g1
1892	mov	T_FLUSH_PCB, %g3
1893	ba,pt	%xcc, sys_trap
1894	sub	%g0, 1, %g4
1895
1896/*
1897 * .spill_clean: clean the previous window, restore the wstate, and
1898 * "done".
1899 *
1900 * Entry: %g7 contains new wstate
1901 */
1902.spill_clean:
1903	sethi	%hi(nwin_minus_one), %g5
1904	ld	[%g5 + %lo(nwin_minus_one)], %g5 ! %g5 = nwin - 1
1905	rdpr	%cwp, %g6			! %g6 = %cwp
1906	deccc	%g6				! %g6--
1907	movneg	%xcc, %g5, %g6			! if (%g6<0) %g6 = nwin-1
1908	wrpr	%g6, %cwp
1909	TT_TRACE_L(trace_win)
1910	clr	%l0
1911	clr	%l1
1912	clr	%l2
1913	clr	%l3
1914	clr	%l4
1915	clr	%l5
1916	clr	%l6
1917	clr	%l7
1918	wrpr	%g0, %g7, %wstate
1919	saved
1920	retry			! restores correct %cwp
1921
1922.fix_alignment:
1923	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
1924	ldn	[%g1 + CPU_THREAD], %g1	! load thread pointer
1925	ldn	[%g1 + T_PROCP], %g1
1926	mov	1, %g2
1927	stb	%g2, [%g1 + P_FIXALIGNMENT]
1928	FAST_TRAP_DONE
1929
1930#define	STDF_REG(REG, ADDR, TMP)		\
1931	sll	REG, 3, REG			;\
1932mark1:	set	start1, TMP			;\
1933	jmp	REG + TMP			;\
1934	  nop					;\
1935start1:	ba,pt	%xcc, done1			;\
1936	  std	%f0, [ADDR + CPU_TMP1]		;\
1937	ba,pt	%xcc, done1			;\
1938	  std	%f32, [ADDR + CPU_TMP1]		;\
1939	ba,pt	%xcc, done1			;\
1940	  std	%f2, [ADDR + CPU_TMP1]		;\
1941	ba,pt	%xcc, done1			;\
1942	  std	%f34, [ADDR + CPU_TMP1]		;\
1943	ba,pt	%xcc, done1			;\
1944	  std	%f4, [ADDR + CPU_TMP1]		;\
1945	ba,pt	%xcc, done1			;\
1946	  std	%f36, [ADDR + CPU_TMP1]		;\
1947	ba,pt	%xcc, done1			;\
1948	  std	%f6, [ADDR + CPU_TMP1]		;\
1949	ba,pt	%xcc, done1			;\
1950	  std	%f38, [ADDR + CPU_TMP1]		;\
1951	ba,pt	%xcc, done1			;\
1952	  std	%f8, [ADDR + CPU_TMP1]		;\
1953	ba,pt	%xcc, done1			;\
1954	  std	%f40, [ADDR + CPU_TMP1]		;\
1955	ba,pt	%xcc, done1			;\
1956	  std	%f10, [ADDR + CPU_TMP1]		;\
1957	ba,pt	%xcc, done1			;\
1958	  std	%f42, [ADDR + CPU_TMP1]		;\
1959	ba,pt	%xcc, done1			;\
1960	  std	%f12, [ADDR + CPU_TMP1]		;\
1961	ba,pt	%xcc, done1			;\
1962	  std	%f44, [ADDR + CPU_TMP1]		;\
1963	ba,pt	%xcc, done1			;\
1964	  std	%f14, [ADDR + CPU_TMP1]		;\
1965	ba,pt	%xcc, done1			;\
1966	  std	%f46, [ADDR + CPU_TMP1]		;\
1967	ba,pt	%xcc, done1			;\
1968	  std	%f16, [ADDR + CPU_TMP1]		;\
1969	ba,pt	%xcc, done1			;\
1970	  std	%f48, [ADDR + CPU_TMP1]		;\
1971	ba,pt	%xcc, done1			;\
1972	  std	%f18, [ADDR + CPU_TMP1]		;\
1973	ba,pt	%xcc, done1			;\
1974	  std	%f50, [ADDR + CPU_TMP1]		;\
1975	ba,pt	%xcc, done1			;\
1976	  std	%f20, [ADDR + CPU_TMP1]		;\
1977	ba,pt	%xcc, done1			;\
1978	  std	%f52, [ADDR + CPU_TMP1]		;\
1979	ba,pt	%xcc, done1			;\
1980	  std	%f22, [ADDR + CPU_TMP1]		;\
1981	ba,pt	%xcc, done1			;\
1982	  std	%f54, [ADDR + CPU_TMP1]		;\
1983	ba,pt	%xcc, done1			;\
1984	  std	%f24, [ADDR + CPU_TMP1]		;\
1985	ba,pt	%xcc, done1			;\
1986	  std	%f56, [ADDR + CPU_TMP1]		;\
1987	ba,pt	%xcc, done1			;\
1988	  std	%f26, [ADDR + CPU_TMP1]		;\
1989	ba,pt	%xcc, done1			;\
1990	  std	%f58, [ADDR + CPU_TMP1]		;\
1991	ba,pt	%xcc, done1			;\
1992	  std	%f28, [ADDR + CPU_TMP1]		;\
1993	ba,pt	%xcc, done1			;\
1994	  std	%f60, [ADDR + CPU_TMP1]		;\
1995	ba,pt	%xcc, done1			;\
1996	  std	%f30, [ADDR + CPU_TMP1]		;\
1997	ba,pt	%xcc, done1			;\
1998	  std	%f62, [ADDR + CPU_TMP1]		;\
1999done1:
2000
2001#define	LDDF_REG(REG, ADDR, TMP)		\
2002	sll	REG, 3, REG			;\
2003mark2:	set	start2, TMP			;\
2004	jmp	REG + TMP			;\
2005	  nop					;\
2006start2:	ba,pt	%xcc, done2			;\
2007	  ldd	[ADDR + CPU_TMP1], %f0		;\
2008	ba,pt	%xcc, done2			;\
2009	  ldd	[ADDR + CPU_TMP1], %f32		;\
2010	ba,pt	%xcc, done2			;\
2011	  ldd	[ADDR + CPU_TMP1], %f2		;\
2012	ba,pt	%xcc, done2			;\
2013	  ldd	[ADDR + CPU_TMP1], %f34		;\
2014	ba,pt	%xcc, done2			;\
2015	  ldd	[ADDR + CPU_TMP1], %f4		;\
2016	ba,pt	%xcc, done2			;\
2017	  ldd	[ADDR + CPU_TMP1], %f36		;\
2018	ba,pt	%xcc, done2			;\
2019	  ldd	[ADDR + CPU_TMP1], %f6		;\
2020	ba,pt	%xcc, done2			;\
2021	  ldd	[ADDR + CPU_TMP1], %f38		;\
2022	ba,pt	%xcc, done2			;\
2023	  ldd	[ADDR + CPU_TMP1], %f8		;\
2024	ba,pt	%xcc, done2			;\
2025	  ldd	[ADDR + CPU_TMP1], %f40		;\
2026	ba,pt	%xcc, done2			;\
2027	  ldd	[ADDR + CPU_TMP1], %f10		;\
2028	ba,pt	%xcc, done2			;\
2029	  ldd	[ADDR + CPU_TMP1], %f42		;\
2030	ba,pt	%xcc, done2			;\
2031	  ldd	[ADDR + CPU_TMP1], %f12		;\
2032	ba,pt	%xcc, done2			;\
2033	  ldd	[ADDR + CPU_TMP1], %f44		;\
2034	ba,pt	%xcc, done2			;\
2035	  ldd	[ADDR + CPU_TMP1], %f14		;\
2036	ba,pt	%xcc, done2			;\
2037	  ldd	[ADDR + CPU_TMP1], %f46		;\
2038	ba,pt	%xcc, done2			;\
2039	  ldd	[ADDR + CPU_TMP1], %f16		;\
2040	ba,pt	%xcc, done2			;\
2041	  ldd	[ADDR + CPU_TMP1], %f48		;\
2042	ba,pt	%xcc, done2			;\
2043	  ldd	[ADDR + CPU_TMP1], %f18		;\
2044	ba,pt	%xcc, done2			;\
2045	  ldd	[ADDR + CPU_TMP1], %f50		;\
2046	ba,pt	%xcc, done2			;\
2047	  ldd	[ADDR + CPU_TMP1], %f20		;\
2048	ba,pt	%xcc, done2			;\
2049	  ldd	[ADDR + CPU_TMP1], %f52		;\
2050	ba,pt	%xcc, done2			;\
2051	  ldd	[ADDR + CPU_TMP1], %f22		;\
2052	ba,pt	%xcc, done2			;\
2053	  ldd	[ADDR + CPU_TMP1], %f54		;\
2054	ba,pt	%xcc, done2			;\
2055	  ldd	[ADDR + CPU_TMP1], %f24		;\
2056	ba,pt	%xcc, done2			;\
2057	  ldd	[ADDR + CPU_TMP1], %f56		;\
2058	ba,pt	%xcc, done2			;\
2059	  ldd	[ADDR + CPU_TMP1], %f26		;\
2060	ba,pt	%xcc, done2			;\
2061	  ldd	[ADDR + CPU_TMP1], %f58		;\
2062	ba,pt	%xcc, done2			;\
2063	  ldd	[ADDR + CPU_TMP1], %f28		;\
2064	ba,pt	%xcc, done2			;\
2065	  ldd	[ADDR + CPU_TMP1], %f60		;\
2066	ba,pt	%xcc, done2			;\
2067	  ldd	[ADDR + CPU_TMP1], %f30		;\
2068	ba,pt	%xcc, done2			;\
2069	  ldd	[ADDR + CPU_TMP1], %f62		;\
2070done2:
2071
2072.lddf_exception_not_aligned:
2073	/* %g2 = sfar, %g3 = sfsr */
2074	mov	%g2, %g5		! stash sfar
2075#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2076	sethi	%hi(fpu_exists), %g2	! check fpu_exists
2077	ld	[%g2 + %lo(fpu_exists)], %g2
2078	brz,a,pn %g2, 4f
2079	  nop
2080#endif
2081	CPU_ADDR(%g1, %g4)
2082	or	%g0, 1, %g4
2083	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2084
2085	rdpr	%tpc, %g2
2086	lda	[%g2]ASI_AIUP, %g6	! get the user's lddf instruction
2087	srl	%g6, 23, %g1		! using ldda or not?
2088	and	%g1, 1, %g1
2089	brz,a,pt %g1, 2f		! check for ldda instruction
2090	  nop
2091	srl	%g6, 13, %g1		! check immflag
2092	and	%g1, 1, %g1
2093	rdpr	%tstate, %g2		! %tstate in %g2
2094	brnz,a,pn %g1, 1f
2095	  srl	%g2, 31, %g1		! get asi from %tstate
2096	srl	%g6, 5, %g1		! get asi from instruction
2097	and	%g1, 0xFF, %g1		! imm_asi field
20981:
2099	cmp	%g1, ASI_P		! primary address space
2100	be,a,pt %icc, 2f
2101	  nop
2102	cmp	%g1, ASI_PNF		! primary no fault address space
2103	be,a,pt %icc, 2f
2104	  nop
2105	cmp	%g1, ASI_S		! secondary address space
2106	be,a,pt %icc, 2f
2107	  nop
2108	cmp	%g1, ASI_SNF		! secondary no fault address space
2109	bne,a,pn %icc, 3f
2110	  nop
21112:
2112	lduwa	[%g5]ASI_USER, %g7	! get first half of misaligned data
2113	add	%g5, 4, %g5		! increment misaligned data address
2114	lduwa	[%g5]ASI_USER, %g5	! get second half of misaligned data
2115
2116	sllx	%g7, 32, %g7
2117	or	%g5, %g7, %g5		! combine data
2118	CPU_ADDR(%g7, %g1)		! save data on a per-cpu basis
2119	stx	%g5, [%g7 + CPU_TMP1]	! save in cpu_tmp1
2120
2121	srl	%g6, 25, %g3		! %g6 has the instruction
2122	and	%g3, 0x1F, %g3		! %g3 has rd
2123	LDDF_REG(%g3, %g7, %g4)
2124
2125	CPU_ADDR(%g1, %g4)
2126	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2127	FAST_TRAP_DONE
21283:
2129	CPU_ADDR(%g1, %g4)
2130	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
21314:
2132	set	T_USER, %g3		! trap type in %g3
2133	or	%g3, T_LDDF_ALIGN, %g3
2134	mov	%g5, %g2		! misaligned vaddr in %g2
2135	set	fpu_trap, %g1		! goto C for the little and
2136	ba,pt	%xcc, sys_trap		! no fault little asi's
2137	  sub	%g0, 1, %g4
2138
2139.stdf_exception_not_aligned:
2140	/* %g2 = sfar, %g3 = sfsr */
2141	mov	%g2, %g5
2142
2143#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2144	sethi	%hi(fpu_exists), %g7		! check fpu_exists
2145	ld	[%g7 + %lo(fpu_exists)], %g3
2146	brz,a,pn %g3, 4f
2147	  nop
2148#endif
2149	CPU_ADDR(%g1, %g4)
2150	or	%g0, 1, %g4
2151	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2152
2153	rdpr	%tpc, %g2
2154	lda	[%g2]ASI_AIUP, %g6	! get the user's stdf instruction
2155
2156	srl	%g6, 23, %g1		! using stda or not?
2157	and	%g1, 1, %g1
2158	brz,a,pt %g1, 2f		! check for stda instruction
2159	  nop
2160	srl	%g6, 13, %g1		! check immflag
2161	and	%g1, 1, %g1
2162	rdpr	%tstate, %g2		! %tstate in %g2
2163	brnz,a,pn %g1, 1f
2164	  srl	%g2, 31, %g1		! get asi from %tstate
2165	srl	%g6, 5, %g1		! get asi from instruction
2166	and	%g1, 0xff, %g1		! imm_asi field
21671:
2168	cmp	%g1, ASI_P		! primary address space
2169	be,a,pt %icc, 2f
2170	  nop
2171	cmp	%g1, ASI_S		! secondary address space
2172	bne,a,pn %icc, 3f
2173	  nop
21742:
2175	srl	%g6, 25, %g6
2176	and	%g6, 0x1F, %g6		! %g6 has rd
2177	CPU_ADDR(%g7, %g1)
2178	STDF_REG(%g6, %g7, %g4)		! STDF_REG(REG, ADDR, TMP)
2179
2180	ldx	[%g7 + CPU_TMP1], %g6
2181	srlx	%g6, 32, %g7
2182	stuwa	%g7, [%g5]ASI_USER	! first half
2183	add	%g5, 4, %g5		! increment misaligned data address
2184	stuwa	%g6, [%g5]ASI_USER	! second half
2185
2186	CPU_ADDR(%g1, %g4)
2187	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2188	FAST_TRAP_DONE
21893:
2190	CPU_ADDR(%g1, %g4)
2191	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
21924:
2193	set	T_USER, %g3		! trap type in %g3
2194	or	%g3, T_STDF_ALIGN, %g3
2195	mov	%g5, %g2		! misaligned vaddr in %g2
2196	set	fpu_trap, %g1		! goto C for the little and
2197	ba,pt	%xcc, sys_trap		! nofault little asi's
2198	  sub	%g0, 1, %g4
2199
2200#ifdef DEBUG_USER_TRAPTRACECTL
2201
2202.traptrace_freeze:
2203	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2204	TT_TRACE_L(trace_win)
2205	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2206	set	trap_freeze, %g1
2207	mov	1, %g2
2208	st	%g2, [%g1]
2209	FAST_TRAP_DONE
2210
2211.traptrace_unfreeze:
2212	set	trap_freeze, %g1
2213	st	%g0, [%g1]
2214	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2215	TT_TRACE_L(trace_win)
2216	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2217	FAST_TRAP_DONE
2218
2219#endif /* DEBUG_USER_TRAPTRACECTL */
2220
2221.getcc:
2222	CPU_ADDR(%g1, %g2)
2223	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2224	rdpr	%tstate, %g3			! get tstate
2225	srlx	%g3, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2226	set	PSR_ICC, %g2
2227	and	%o0, %g2, %o0			! mask out the rest
2228	srl	%o0, PSR_ICC_SHIFT, %o0		! right justify
2229	wrpr	%g0, 0, %gl
2230	mov	%o0, %g1			! move ccr to normal %g1
2231	wrpr	%g0, 1, %gl
2232	! cannot assume globals retained their values after increasing %gl
2233	CPU_ADDR(%g1, %g2)
2234	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2235	FAST_TRAP_DONE
2236
2237.setcc:
2238	CPU_ADDR(%g1, %g2)
2239	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2240	wrpr	%g0, 0, %gl
2241	mov	%g1, %o0
2242	wrpr	%g0, 1, %gl
2243	! cannot assume globals retained their values after increasing %gl
2244	CPU_ADDR(%g1, %g2)
2245	sll	%o0, PSR_ICC_SHIFT, %g2
2246	set	PSR_ICC, %g3
2247	and	%g2, %g3, %g2			! mask out rest
2248	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g2
2249	rdpr	%tstate, %g3			! get tstate
2250	srl	%g3, 0, %g3			! clear upper word
2251	or	%g3, %g2, %g3			! or in new bits
2252	wrpr	%g3, %tstate
2253	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2254	FAST_TRAP_DONE
2255
2256/*
2257 * getpsr(void)
2258 * Note that the xcc part of the ccr is not provided.
2259 * The V8 code shows why the V9 trap is not faster:
2260 * #define GETPSR_TRAP() \
2261 *      mov %psr, %i0; jmp %l2; rett %l2+4; nop;
2262 */
2263
2264	.type	.getpsr, #function
2265.getpsr:
2266	rdpr	%tstate, %g1			! get tstate
2267	srlx	%g1, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2268	set	PSR_ICC, %g2
2269	and	%o0, %g2, %o0			! mask out the rest
2270
2271	rd	%fprs, %g1			! get fprs
2272	and	%g1, FPRS_FEF, %g2		! mask out dirty upper/lower
2273	sllx	%g2, PSR_FPRS_FEF_SHIFT, %g2	! shift fef to V8 psr.ef
2274	or	%o0, %g2, %o0			! or result into psr.ef
2275
2276	set	V9_PSR_IMPLVER, %g2		! SI assigned impl/ver: 0xef
2277	or	%o0, %g2, %o0			! or psr.impl/ver
2278	FAST_TRAP_DONE
2279	SET_SIZE(.getpsr)
2280
2281/*
2282 * setpsr(newpsr)
2283 * Note that there is no support for ccr.xcc in the V9 code.
2284 */
2285
2286	.type	.setpsr, #function
2287.setpsr:
2288	rdpr	%tstate, %g1			! get tstate
2289!	setx	TSTATE_V8_UBITS, %g2
2290	or 	%g0, CCR_ICC, %g3
2291	sllx	%g3, TSTATE_CCR_SHIFT, %g2
2292
2293	andn	%g1, %g2, %g1			! zero current user bits
2294	set	PSR_ICC, %g2
2295	and	%g2, %o0, %g2			! clear all but psr.icc bits
2296	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g3	! shift to tstate.ccr.icc
2297	wrpr	%g1, %g3, %tstate		! write tstate
2298
2299	set	PSR_EF, %g2
2300	and	%g2, %o0, %g2			! clear all but fp enable bit
2301	srlx	%g2, PSR_FPRS_FEF_SHIFT, %g4	! shift ef to V9 fprs.fef
2302	wr	%g0, %g4, %fprs			! write fprs
2303
2304	CPU_ADDR(%g1, %g2)			! load CPU struct addr to %g1
2305	ldn	[%g1 + CPU_THREAD], %g2		! load thread pointer
2306	ldn	[%g2 + T_LWP], %g3		! load klwp pointer
2307	ldn	[%g3 + LWP_FPU], %g2		! get lwp_fpu pointer
2308	stuw	%g4, [%g2 + FPU_FPRS]		! write fef value to fpu_fprs
2309	srlx	%g4, 2, %g4			! shift fef value to bit 0
2310	stub	%g4, [%g2 + FPU_EN]		! write fef value to fpu_en
2311	FAST_TRAP_DONE
2312	SET_SIZE(.setpsr)
2313
2314/*
2315 * getlgrp
2316 * get home lgrpid on which the calling thread is currently executing.
2317 */
2318	.type	.getlgrp, #function
2319.getlgrp:
2320	! Thanks for the incredibly helpful comments
2321	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2322	ld	[%g1 + CPU_ID], %o0	! load cpu_id
2323	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2324	ldn	[%g2 + T_LPL], %g2	! load lpl pointer
2325	ld	[%g2 + LPL_LGRPID], %g1	! load lpl_lgrpid
2326	sra	%g1, 0, %o1
2327	FAST_TRAP_DONE
2328	SET_SIZE(.getlgrp)
2329
2330/*
2331 * Entry for old 4.x trap (trap 0).
2332 */
2333	ENTRY_NP(syscall_trap_4x)
2334	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2335	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2336	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2337	ld	[%g2 + PCB_TRAP0], %g2	! lwp->lwp_pcb.pcb_trap0addr
2338	brz,pn	%g2, 1f			! has it been set?
2339	st	%l0, [%g1 + CPU_TMP1]	! delay - save some locals
2340	st	%l1, [%g1 + CPU_TMP2]
2341	rdpr	%tnpc, %l1		! save old tnpc
2342	wrpr	%g0, %g2, %tnpc		! setup tnpc
2343
2344	mov	%g1, %l0		! save CPU struct addr
2345	wrpr	%g0, 0, %gl
2346	mov	%l1, %g6		! pass tnpc to user code in %g6
2347	wrpr	%g0, 1, %gl
2348	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2349	ld	[%l0 + CPU_TMP1], %l0
2350	FAST_TRAP_DONE_CHK_INTR
23511:
2352	!
2353	! check for old syscall mmap which is the only different one which
2354	! must be the same.  Others are handled in the compatibility library.
2355	!
2356	mov	%g1, %l0		! save CPU struct addr
2357	wrpr	%g0, 0, %gl
2358	cmp	%g1, OSYS_mmap		! compare to old 4.x mmap
2359	movz	%icc, SYS_mmap, %g1
2360	wrpr	%g0, 1, %gl
2361	ld	[%l0 + CPU_TMP1], %l0
2362	SYSCALL(syscall_trap32)
2363	SET_SIZE(syscall_trap_4x)
2364
2365/*
2366 * Handler for software trap 9.
2367 * Set trap0 emulation address for old 4.x system call trap.
2368 * XXX - this should be a system call.
2369 */
2370	ENTRY_NP(set_trap0_addr)
2371	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2372	st	%l0, [%g1 + CPU_TMP1]	! save some locals
2373	st	%l1, [%g1 + CPU_TMP2]
2374	mov	%g1, %l0	! preserve CPU addr
2375	wrpr	%g0, 0, %gl
2376	mov	%g1, %l1
2377	wrpr	%g0, 1, %gl
2378	! cannot assume globals retained their values after increasing %gl
2379	ldn	[%l0 + CPU_THREAD], %g2	! load thread pointer
2380	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2381	andn	%l1, 3, %l1		! force alignment
2382	st	%l1, [%g2 + PCB_TRAP0]	! lwp->lwp_pcb.pcb_trap0addr
2383	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2384	ld	[%l0 + CPU_TMP1], %l0
2385	FAST_TRAP_DONE
2386	SET_SIZE(set_trap0_addr)
2387
2388/*
2389 * mmu_trap_tl1
2390 * trap handler for unexpected mmu traps.
2391 * simply checks if the trap was a user lddf/stdf alignment trap, in which
2392 * case we go to fpu_trap or a user trap from the window handler, in which
2393 * case we go save the state on the pcb.  Otherwise, we go to ptl1_panic.
2394 */
2395	.type	mmu_trap_tl1, #function
2396mmu_trap_tl1:
2397#ifdef	TRAPTRACE
2398	TRACE_PTR(%g5, %g6)
2399	GET_TRACE_TICK(%g6)
2400	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2401	TRACE_SAVE_TL_GL_REGS(%g5, %g6)
2402	rdpr	%tt, %g6
2403	stha	%g6, [%g5 + TRAP_ENT_TT]%asi
2404	rdpr	%tstate, %g6
2405	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi
2406	stna	%sp, [%g5 + TRAP_ENT_SP]%asi
2407	stna	%g0, [%g5 + TRAP_ENT_TR]%asi
2408	rdpr	%tpc, %g6
2409	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2410	MMU_FAULT_STATUS_AREA(%g6)
2411	ldx	[%g6 + MMFSA_D_ADDR], %g6
2412	stna	%g6, [%g5 + TRAP_ENT_F1]%asi !  MMU fault address
2413	CPU_PADDR(%g7, %g6);
2414	add	%g7, CPU_TL1_HDLR, %g7
2415	lda	[%g7]ASI_MEM, %g6
2416	stna	%g6, [%g5 + TRAP_ENT_F2]%asi
2417	MMU_FAULT_STATUS_AREA(%g6)
2418	ldx	[%g6 + MMFSA_D_TYPE], %g7 ! XXXQ should be a MMFSA_F_ constant?
2419	ldx	[%g6 + MMFSA_D_CTX], %g6
2420	sllx	%g6, SFSR_CTX_SHIFT, %g6
2421	or	%g6, %g7, %g6
2422	stna	%g6, [%g5 + TRAP_ENT_F3]%asi ! MMU context/type
2423	set	0xdeadbeef, %g6
2424	stna	%g6, [%g5 + TRAP_ENT_F4]%asi
2425	TRACE_NEXT(%g5, %g6, %g7)
2426#endif /* TRAPTRACE */
2427	CPU_PADDR(%g7, %g6);
2428	add     %g7, CPU_TL1_HDLR, %g7		! %g7 = &cpu_m.tl1_hdlr (PA)
2429	lda	[%g7]ASI_MEM, %g6
2430	brz,a,pt %g6, 1f
2431	  nop
2432	sta     %g0, [%g7]ASI_MEM
2433	! XXXQ need to setup registers for sfmmu_mmu_trap?
2434	ba,a,pt	%xcc, sfmmu_mmu_trap		! handle page faults
24351:
2436	rdpr	%tpc, %g7
2437	/* in user_rtt? */
2438	set	rtt_fill_start, %g6
2439	cmp	%g7, %g6
2440	blu,pn	%xcc, 6f
2441	 .empty
2442	set	rtt_fill_end, %g6
2443	cmp	%g7, %g6
2444	bgeu,pn %xcc, 6f
2445	 nop
2446	set	fault_rtt_fn1, %g7
2447	ba,a	7f
24486:
2449	! check to see if the trap pc is in a window spill/fill handling
2450	rdpr	%tpc, %g7
2451	/* tpc should be in the trap table */
2452	set	trap_table, %g6
2453	cmp	%g7, %g6
2454	blu,a,pn %xcc, ptl1_panic
2455	  mov	PTL1_BAD_MMUTRAP, %g1
2456	set	etrap_table, %g6
2457	cmp	%g7, %g6
2458	bgeu,a,pn %xcc, ptl1_panic
2459	  mov	PTL1_BAD_MMUTRAP, %g1
2460	! pc is inside the trap table, convert to trap type
2461	srl	%g7, 5, %g6		! XXXQ need #define
2462	and	%g6, 0x1ff, %g6		! XXXQ need #define
2463	! and check for a window trap type
2464	and	%g6, WTRAP_TTMASK, %g6
2465	cmp	%g6, WTRAP_TYPE
2466	bne,a,pn %xcc, ptl1_panic
2467	  mov	PTL1_BAD_MMUTRAP, %g1
2468	andn	%g7, WTRAP_ALIGN, %g7	/* 128 byte aligned */
2469	add	%g7, WTRAP_FAULTOFF, %g7
2470
24717:
2472	! Arguments are passed in the global set active after the
2473	! 'done' instruction. Before switching sets, must save
2474	! the calculated next pc
2475	wrpr	%g0, %g7, %tnpc
2476	wrpr	%g0, 1, %gl
2477	rdpr	%tt, %g5
2478	MMU_FAULT_STATUS_AREA(%g7)
2479	cmp	%g5, T_ALIGNMENT
2480	be,pn	%xcc, 1f
2481	ldx	[%g7 + MMFSA_D_ADDR], %g6
2482	ldx	[%g7 + MMFSA_D_CTX], %g7
2483	srlx	%g6, MMU_PAGESHIFT, %g6		/* align address */
2484	sllx	%g6, MMU_PAGESHIFT, %g6
2485	or	%g6, %g7, %g6			/* TAG_ACCESS */
24861:
2487	done
2488	SET_SIZE(mmu_trap_tl1)
2489
2490/*
2491 * Several traps use kmdb_trap and kmdb_trap_tl1 as their handlers.  These
2492 * traps are valid only when kmdb is loaded.  When the debugger is active,
2493 * the code below is rewritten to transfer control to the appropriate
2494 * debugger entry points.
2495 */
2496	.global	kmdb_trap
2497	.align	8
2498kmdb_trap:
2499	ba,a	trap_table0
2500	jmp	%g1 + 0
2501	nop
2502
2503	.global	kmdb_trap_tl1
2504	.align	8
2505kmdb_trap_tl1:
2506	ba,a	trap_table0
2507	jmp	%g1 + 0
2508	nop
2509
2510/*
2511 * This entry is copied from OBP's trap table during boot.
2512 */
2513	.global	obp_bpt
2514	.align	8
2515obp_bpt:
2516	NOT
2517
2518
2519
2520#ifdef	TRAPTRACE
2521/*
2522 * TRAPTRACE support.
2523 * labels here are branched to with "rd %pc, %g7" in the delay slot.
2524 * Return is done by "jmp %g7 + 4".
2525 */
2526
2527trace_dmmu:
2528	TRACE_PTR(%g3, %g6)
2529	GET_TRACE_TICK(%g6)
2530	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2531	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2532	rdpr	%tt, %g6
2533	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2534	rdpr	%tstate, %g6
2535	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2536	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2537	rdpr	%tpc, %g6
2538	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2539	MMU_FAULT_STATUS_AREA(%g6)
2540	ldx	[%g6 + MMFSA_D_ADDR], %g4
2541	stxa	%g4, [%g3 + TRAP_ENT_TR]%asi
2542	ldx	[%g6 + MMFSA_D_CTX], %g4
2543	stxa	%g4, [%g3 + TRAP_ENT_F1]%asi
2544	ldx	[%g6 + MMFSA_D_TYPE], %g4
2545	stxa	%g4, [%g3 + TRAP_ENT_F2]%asi
2546	stxa	%g6, [%g3 + TRAP_ENT_F3]%asi
2547	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2548	TRACE_NEXT(%g3, %g4, %g5)
2549	jmp	%g7 + 4
2550	nop
2551
2552trace_immu:
2553	TRACE_PTR(%g3, %g6)
2554	GET_TRACE_TICK(%g6)
2555	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2556	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2557	rdpr	%tt, %g6
2558	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2559	rdpr	%tstate, %g6
2560	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2561	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2562	rdpr	%tpc, %g6
2563	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2564	MMU_FAULT_STATUS_AREA(%g6)
2565	ldx	[%g6 + MMFSA_I_ADDR], %g4
2566	stxa	%g4, [%g3 + TRAP_ENT_TR]%asi
2567	ldx	[%g6 + MMFSA_I_CTX], %g4
2568	stxa	%g4, [%g3 + TRAP_ENT_F1]%asi
2569	ldx	[%g6 + MMFSA_I_TYPE], %g4
2570	stxa	%g4, [%g3 + TRAP_ENT_F2]%asi
2571	stxa	%g6, [%g3 + TRAP_ENT_F3]%asi
2572	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2573	TRACE_NEXT(%g3, %g4, %g5)
2574	jmp	%g7 + 4
2575	nop
2576
2577trace_gen:
2578	TRACE_PTR(%g3, %g6)
2579	GET_TRACE_TICK(%g6)
2580	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2581	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2582	rdpr	%tt, %g6
2583	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2584	rdpr	%tstate, %g6
2585	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2586	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2587	rdpr	%tpc, %g6
2588	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2589	stna	%g0, [%g3 + TRAP_ENT_TR]%asi
2590	stna	%g0, [%g3 + TRAP_ENT_F1]%asi
2591	stna	%g0, [%g3 + TRAP_ENT_F2]%asi
2592	stna	%g0, [%g3 + TRAP_ENT_F3]%asi
2593	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2594	TRACE_NEXT(%g3, %g4, %g5)
2595	jmp	%g7 + 4
2596	nop
2597
2598trace_win:
2599	TRACE_WIN_INFO(0, %l0, %l1, %l2)
2600	! Keep the locals as clean as possible, caller cleans %l4
2601	clr	%l2
2602	clr	%l1
2603	jmp	%l4 + 4
2604	  clr	%l0
2605
2606/*
2607 * Trace a tsb hit
2608 * g1 = tsbe pointer (in/clobbered)
2609 * g2 = tag access register (in)
2610 * g3 - g4 = scratch (clobbered)
2611 * g5 = tsbe data (in)
2612 * g6 = scratch (clobbered)
2613 * g7 = pc we jumped here from (in)
2614 */
2615
2616	! Do not disturb %g5, it will be used after the trace
2617	ALTENTRY(trace_tsbhit)
2618	TRACE_TSBHIT(0)
2619	jmp	%g7 + 4
2620	nop
2621
2622/*
2623 * Trace a TSB miss
2624 *
2625 * g1 = tsb8k pointer (in)
2626 * g2 = tag access register (in)
2627 * g3 = tsb4m pointer (in)
2628 * g4 = tsbe tag (in/clobbered)
2629 * g5 - g6 = scratch (clobbered)
2630 * g7 = pc we jumped here from (in)
2631 */
2632	.global	trace_tsbmiss
2633trace_tsbmiss:
2634	membar	#Sync
2635	sethi	%hi(FLUSH_ADDR), %g6
2636	flush	%g6
2637	TRACE_PTR(%g5, %g6)
2638	GET_TRACE_TICK(%g6)
2639	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2640	stna	%g2, [%g5 + TRAP_ENT_SP]%asi		! tag access
2641	stna	%g4, [%g5 + TRAP_ENT_F1]%asi		! XXX? tsb tag
2642	rdpr	%tnpc, %g6
2643	stna	%g6, [%g5 + TRAP_ENT_F2]%asi
2644	stna	%g1, [%g5 + TRAP_ENT_F3]%asi		! tsb8k pointer
2645	srlx	%g1, 32, %g6
2646	stna	%g6, [%g5 + TRAP_ENT_F4]%asi		! huh?
2647	rdpr	%tpc, %g6
2648	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2649	TRACE_SAVE_TL_GL_REGS(%g5, %g6)
2650	rdpr	%tt, %g6
2651	or	%g6, TT_MMU_MISS, %g4
2652	stha	%g4, [%g5 + TRAP_ENT_TT]%asi
2653	mov	MMFSA_D_ADDR, %g4
2654	cmp	%g6, FAST_IMMU_MISS_TT
2655	move	%xcc, MMFSA_I_ADDR, %g4
2656	cmp	%g6, T_INSTR_MMU_MISS
2657	move	%xcc, MMFSA_I_ADDR, %g4
2658	MMU_FAULT_STATUS_AREA(%g6)
2659	ldx	[%g6 + %g4], %g6
2660	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi	! tag target
2661	stna	%g3, [%g5 + TRAP_ENT_TR]%asi		! tsb4m pointer
2662	TRACE_NEXT(%g5, %g4, %g6)
2663	jmp	%g7 + 4
2664	nop
2665
2666/*
2667 * g2 = tag access register (in)
2668 * g3 = ctx number (in)
2669 */
2670trace_dataprot:
2671	membar	#Sync
2672	sethi	%hi(FLUSH_ADDR), %g6
2673	flush	%g6
2674	TRACE_PTR(%g1, %g6)
2675	GET_TRACE_TICK(%g6)
2676	stxa	%g6, [%g1 + TRAP_ENT_TICK]%asi
2677	rdpr	%tpc, %g6
2678	stna	%g6, [%g1 + TRAP_ENT_TPC]%asi
2679	rdpr	%tstate, %g6
2680	stxa	%g6, [%g1 + TRAP_ENT_TSTATE]%asi
2681	stna	%g2, [%g1 + TRAP_ENT_SP]%asi		! tag access reg
2682	stna	%g0, [%g1 + TRAP_ENT_TR]%asi
2683	stna	%g0, [%g1 + TRAP_ENT_F1]%asi
2684	stna	%g0, [%g1 + TRAP_ENT_F2]%asi
2685	stna	%g0, [%g1 + TRAP_ENT_F3]%asi
2686	stna	%g0, [%g1 + TRAP_ENT_F4]%asi
2687	TRACE_SAVE_TL_GL_REGS(%g1, %g6)
2688	rdpr	%tt, %g6
2689	stha	%g6, [%g1 + TRAP_ENT_TT]%asi
2690	TRACE_NEXT(%g1, %g4, %g5)
2691	jmp	%g7 + 4
2692	nop
2693
2694#endif /* TRAPTRACE */
2695
2696/*
2697 * Handle watchdog reset trap. Enable the MMU using the MMU_ENABLE
2698 * HV service, which requires the return target to be specified as a VA
2699 * since we are enabling the MMU. We set the target to ptl1_panic.
2700 */
2701
2702	.type	.watchdog_trap, #function
2703.watchdog_trap:
2704	mov	1, %o0
2705	setx	ptl1_panic, %g2, %o1
2706	mov	MMU_ENABLE, %o5
2707	ta	FAST_TRAP
2708	done
2709	SET_SIZE(.watchdog_trap)
2710/*
2711 * synthesize for trap(): SFAR in %g2, SFSR in %g3
2712 */
2713	.type	.dmmu_exc_lddf_not_aligned, #function
2714.dmmu_exc_lddf_not_aligned:
2715	MMU_FAULT_STATUS_AREA(%g3)
2716	ldx	[%g3 + MMFSA_D_ADDR], %g2
2717	/* Fault type not available in MMU fault status area */
2718	mov	MMFSA_F_UNALIGN, %g1
2719	ldx	[%g3 + MMFSA_D_CTX], %g3
2720	sllx	%g3, SFSR_CTX_SHIFT, %g3
2721	btst	1, %sp
2722	bnz,pt	%xcc, .lddf_exception_not_aligned
2723	or	%g3, %g1, %g3			/* SFSR */
2724	ba,a,pt	%xcc, .mmu_exception_not_aligned
2725	SET_SIZE(.dmmu_exc_lddf_not_aligned)
2726
2727/*
2728 * synthesize for trap(): SFAR in %g2, SFSR in %g3
2729 */
2730	.type	.dmmu_exc_stdf_not_aligned, #function
2731.dmmu_exc_stdf_not_aligned:
2732	MMU_FAULT_STATUS_AREA(%g3)
2733	ldx	[%g3 + MMFSA_D_ADDR], %g2
2734	/* Fault type not available in MMU fault status area */
2735	mov	MMFSA_F_UNALIGN, %g1
2736	ldx	[%g3 + MMFSA_D_CTX], %g3
2737	sllx	%g3, SFSR_CTX_SHIFT, %g3
2738	btst	1, %sp
2739	bnz,pt	%xcc, .stdf_exception_not_aligned
2740	or	%g3, %g1, %g3			/* SFSR */
2741	ba,a,pt	%xcc, .mmu_exception_not_aligned
2742	SET_SIZE(.dmmu_exc_stdf_not_aligned)
2743
2744	.type	.dmmu_exception, #function
2745.dmmu_exception:
2746	MMU_FAULT_STATUS_AREA(%g3)
2747	ldx	[%g3 + MMFSA_D_ADDR], %g2
2748	ldx	[%g3 + MMFSA_D_TYPE], %g1
2749	ldx	[%g3 + MMFSA_D_CTX], %g3
2750	srlx	%g2, MMU_PAGESHIFT, %g2		/* align address */
2751	sllx	%g2, MMU_PAGESHIFT, %g2
2752	or	%g2, %g3, %g2			/* TAG_ACCESS */
2753	sllx	%g3, SFSR_CTX_SHIFT, %g3
2754	or	%g3, %g1, %g3			/* SFSR */
2755	ba,pt	%xcc, .mmu_exception_end
2756	mov	T_DATA_EXCEPTION, %g1
2757	SET_SIZE(.dmmu_exception)
2758/*
2759 * expects offset into tsbmiss area in %g1 and return pc in %g7
2760 */
2761stat_mmu:
2762	CPU_INDEX(%g5, %g6)
2763	sethi	%hi(tsbmiss_area), %g6
2764	sllx	%g5, TSBMISS_SHIFT, %g5
2765	or	%g6, %lo(tsbmiss_area), %g6
2766	add	%g6, %g5, %g6		/* g6 = tsbmiss area */
2767	ld	[%g6 + %g1], %g5
2768	add	%g5, 1, %g5
2769	jmp	%g7 + 4
2770	st	%g5, [%g6 + %g1]
2771
2772
2773/*
2774 * fast_trap_done, fast_trap_done_chk_intr:
2775 *
2776 * Due to the design of UltraSPARC pipeline, pending interrupts are not
2777 * taken immediately after a RETRY or DONE instruction which causes IE to
2778 * go from 0 to 1. Instead, the instruction at %tpc or %tnpc is allowed
2779 * to execute first before taking any interrupts. If that instruction
2780 * results in other traps, and if the corresponding trap handler runs
2781 * entirely at TL=1 with interrupts disabled, then pending interrupts
2782 * won't be taken until after yet another instruction following the %tpc
2783 * or %tnpc.
2784 *
2785 * A malicious user program can use this feature to block out interrupts
2786 * for extended durations, which can result in send_mondo_timeout kernel
2787 * panic.
2788 *
2789 * This problem is addressed by servicing any pending interrupts via
2790 * sys_trap before returning back to the user mode from a fast trap
2791 * handler. The "done" instruction within a fast trap handler, which
2792 * runs entirely at TL=1 with interrupts disabled, is replaced with the
2793 * FAST_TRAP_DONE macro, which branches control to this fast_trap_done
2794 * entry point.
2795 *
2796 * We check for any pending interrupts here and force a sys_trap to
2797 * service those interrupts, if any. To minimize overhead, pending
2798 * interrupts are checked if the %tpc happens to be at 16K boundary,
2799 * which allows a malicious program to execute at most 4K consecutive
2800 * instructions before we service any pending interrupts. If a worst
2801 * case fast trap handler takes about 2 usec, then interrupts will be
2802 * blocked for at most 8 msec, less than a clock tick.
2803 *
2804 * For the cases where we don't know if the %tpc will cross a 16K
2805 * boundary, we can't use the above optimization and always process
2806 * any pending interrupts via fast_frap_done_chk_intr entry point.
2807 *
2808 * Entry Conditions:
2809 * 	%pstate		am:0 priv:1 ie:0
2810 * 			globals are AG (not normal globals)
2811 */
2812
2813	.global	fast_trap_done, fast_trap_done_chk_intr
2814fast_trap_done:
2815	rdpr	%tpc, %g5
2816	sethi	%hi(0xffffc000), %g6	! 1's complement of 0x3fff
2817	andncc	%g5, %g6, %g0		! check lower 14 bits of %tpc
2818	bz,pn	%icc, 1f		! branch if zero (lower 32 bits only)
2819	nop
2820	done
2821
2822fast_trap_done_chk_intr:
28231:	rd	SOFTINT, %g6
2824	brnz,pn	%g6, 2f		! branch if any pending intr
2825	nop
2826	done
2827
28282:
2829	/*
2830	 * We get here if there are any pending interrupts.
2831	 * Adjust %tpc/%tnpc as we'll be resuming via "retry"
2832	 * instruction.
2833	 */
2834	rdpr	%tnpc, %g5
2835	wrpr	%g0, %g5, %tpc
2836	add	%g5, 4, %g5
2837	wrpr	%g0, %g5, %tnpc
2838
2839	/*
2840	 * Force a dummy sys_trap call so that interrupts can be serviced.
2841	 */
2842	set	fast_trap_dummy_call, %g1
2843	ba,pt	%xcc, sys_trap
2844	  mov	-1, %g4
2845
2846fast_trap_dummy_call:
2847	retl
2848	nop
2849
2850#endif	/* lint */
2851