xref: /titanic_41/usr/src/uts/sun4v/ml/trap_table.s (revision 2449e17f82f6097fd2c665b64723e31ceecbeca6)
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 2007 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/machbrand.h>
40#include <sys/pcb.h>
41#include <sys/pte.h>
42#include <sys/mmu.h>
43#include <sys/machpcb.h>
44#include <sys/async.h>
45#include <sys/intreg.h>
46#include <sys/scb.h>
47#include <sys/psr_compat.h>
48#include <sys/syscall.h>
49#include <sys/machparam.h>
50#include <sys/traptrace.h>
51#include <vm/hat_sfmmu.h>
52#include <sys/archsystm.h>
53#include <sys/utrap.h>
54#include <sys/clock.h>
55#include <sys/intr.h>
56#include <sys/fpu/fpu_simulator.h>
57#include <vm/seg_spt.h>
58
59/*
60 * WARNING: If you add a fast trap handler which can be invoked by a
61 * non-privileged user, you may have to use the FAST_TRAP_DONE macro
62 * instead of "done" instruction to return back to the user mode. See
63 * comments for the "fast_trap_done" entry point for more information.
64 *
65 * An alternate FAST_TRAP_DONE_CHK_INTR macro should be used for the
66 * cases where you always want to process any pending interrupts before
67 * returning back to the user mode.
68 */
69#define	FAST_TRAP_DONE		\
70	ba,a	fast_trap_done
71
72#define	FAST_TRAP_DONE_CHK_INTR	\
73	ba,a	fast_trap_done_chk_intr
74
75/*
76 * SPARC V9 Trap Table
77 *
78 * Most of the trap handlers are made from common building
79 * blocks, and some are instantiated multiple times within
80 * the trap table. So, I build a bunch of macros, then
81 * populate the table using only the macros.
82 *
83 * Many macros branch to sys_trap.  Its calling convention is:
84 *	%g1		kernel trap handler
85 *	%g2, %g3	args for above
86 *	%g4		desire %pil
87 */
88
89#ifdef	TRAPTRACE
90
91/*
92 * Tracing macro. Adds two instructions if TRAPTRACE is defined.
93 */
94#define	TT_TRACE(label)		\
95	ba	label		;\
96	rd	%pc, %g7
97#define	TT_TRACE_INS	2
98
99#define	TT_TRACE_L(label)	\
100	ba	label		;\
101	rd	%pc, %l4	;\
102	clr	%l4
103#define	TT_TRACE_L_INS	3
104
105#else
106
107#define	TT_TRACE(label)
108#define	TT_TRACE_INS	0
109
110#define	TT_TRACE_L(label)
111#define	TT_TRACE_L_INS	0
112
113#endif
114
115/*
116 * This first set are funneled to trap() with %tt as the type.
117 * Trap will then either panic or send the user a signal.
118 */
119/*
120 * NOT is used for traps that just shouldn't happen.
121 * It comes in both single and quadruple flavors.
122 */
123#if !defined(lint)
124	.global	trap
125#endif /* !lint */
126#define	NOT			\
127	TT_TRACE(trace_gen)	;\
128	set	trap, %g1	;\
129	rdpr	%tt, %g3	;\
130	ba,pt	%xcc, sys_trap	;\
131	sub	%g0, 1, %g4	;\
132	.align	32
133#define	NOT4	NOT; NOT; NOT; NOT
134
135#define	NOTP				\
136	TT_TRACE(trace_gen)		;\
137	ba,pt	%xcc, ptl1_panic	;\
138	  mov	PTL1_BAD_TRAP, %g1	;\
139	.align	32
140#define	NOTP4	NOTP; NOTP; NOTP; NOTP
141
142
143/*
144 * BAD is used for trap vectors we don't have a kernel
145 * handler for.
146 * It also comes in single and quadruple versions.
147 */
148#define	BAD	NOT
149#define	BAD4	NOT4
150
151#define	DONE			\
152	done;			\
153	.align	32
154
155/*
156 * TRAP vectors to the trap() function.
157 * It's main use is for user errors.
158 */
159#if !defined(lint)
160	.global	trap
161#endif /* !lint */
162#define	TRAP(arg)		\
163	TT_TRACE(trace_gen)	;\
164	set	trap, %g1	;\
165	mov	arg, %g3	;\
166	ba,pt	%xcc, sys_trap	;\
167	sub	%g0, 1, %g4	;\
168	.align	32
169
170/*
171 * SYSCALL is used for unsupported syscall interfaces (with 'which'
172 * set to 'nosys') and legacy support of old SunOS 4.x syscalls (with
173 * 'which' set to 'syscall_trap32').
174 *
175 * The SYSCALL_TRAP* macros are used for syscall entry points.
176 * SYSCALL_TRAP is used to support LP64 syscalls and SYSCALL_TRAP32
177 * is used to support ILP32.  Each macro can only be used once
178 * since they each define a symbol.  The symbols are used as hot patch
179 * points by the brand infrastructure to dynamically enable and disable
180 * brand syscall interposition.  See the comments around BRAND_CALLBACK
181 * and brand_plat_interposition_enable() for more information.
182 */
183#define	SYSCALL_NOTT(which)		\
184	set	(which), %g1		;\
185	ba,pt	%xcc, sys_trap		;\
186	sub	%g0, 1, %g4		;\
187	.align	32
188
189#define	SYSCALL(which)			\
190	TT_TRACE(trace_gen)		;\
191	SYSCALL_NOTT(which)
192
193#define	SYSCALL_TRAP32				\
194	TT_TRACE(trace_gen)			;\
195	ALTENTRY(syscall_trap32_patch_point)	\
196	SYSCALL_NOTT(syscall_trap32)
197
198#define	SYSCALL_TRAP				\
199	TT_TRACE(trace_gen)			;\
200	ALTENTRY(syscall_trap_patch_point)	\
201	SYSCALL_NOTT(syscall_trap)
202
203/*
204 * GOTO just jumps to a label.
205 * It's used for things that can be fixed without going thru sys_trap.
206 */
207#define	GOTO(label)		\
208	.global	label		;\
209	ba,a	label		;\
210	.empty			;\
211	.align	32
212
213/*
214 * GOTO_TT just jumps to a label.
215 * correctable ECC error traps at  level 0 and 1 will use this macro.
216 * It's used for things that can be fixed without going thru sys_trap.
217 */
218#define	GOTO_TT(label, ttlabel)		\
219	.global	label		;\
220	TT_TRACE(ttlabel)	;\
221	ba,a	label		;\
222	.empty			;\
223	.align	32
224
225/*
226 * Privileged traps
227 * Takes breakpoint if privileged, calls trap() if not.
228 */
229#define	PRIV(label)			\
230	rdpr	%tstate, %g1		;\
231	btst	TSTATE_PRIV, %g1	;\
232	bnz	label			;\
233	rdpr	%tt, %g3		;\
234	set	trap, %g1		;\
235	ba,pt	%xcc, sys_trap		;\
236	sub	%g0, 1, %g4		;\
237	.align	32
238
239
240/*
241 * DTrace traps.
242 */
243#define	DTRACE_PID			\
244	.global dtrace_pid_probe				;\
245	set	dtrace_pid_probe, %g1				;\
246	ba,pt	%xcc, user_trap					;\
247	sub	%g0, 1, %g4					;\
248	.align	32
249
250#define	DTRACE_RETURN			\
251	.global dtrace_return_probe				;\
252	set	dtrace_return_probe, %g1			;\
253	ba,pt	%xcc, user_trap					;\
254	sub	%g0, 1, %g4					;\
255	.align	32
256
257/*
258 * REGISTER WINDOW MANAGEMENT MACROS
259 */
260
261/*
262 * various convenient units of padding
263 */
264#define	SKIP(n)	.skip 4*(n)
265
266/*
267 * CLEAN_WINDOW is the simple handler for cleaning a register window.
268 */
269#define	CLEAN_WINDOW						\
270	TT_TRACE_L(trace_win)					;\
271	rdpr %cleanwin, %l0; inc %l0; wrpr %l0, %cleanwin	;\
272	clr %l0; clr %l1; clr %l2; clr %l3			;\
273	clr %l4; clr %l5; clr %l6; clr %l7			;\
274	clr %o0; clr %o1; clr %o2; clr %o3			;\
275	clr %o4; clr %o5; clr %o6; clr %o7			;\
276	retry; .align 128
277
278#if !defined(lint)
279
280/*
281 * If we get an unresolved tlb miss while in a window handler, the fault
282 * handler will resume execution at the last instruction of the window
283 * hander, instead of delivering the fault to the kernel.  Spill handlers
284 * use this to spill windows into the wbuf.
285 *
286 * The mixed handler works by checking %sp, and branching to the correct
287 * handler.  This is done by branching back to label 1: for 32b frames,
288 * or label 2: for 64b frames; which implies the handler order is: 32b,
289 * 64b, mixed.  The 1: and 2: labels are offset into the routines to
290 * allow the branchs' delay slots to contain useful instructions.
291 */
292
293/*
294 * SPILL_32bit spills a 32-bit-wide kernel register window.  It
295 * assumes that the kernel context and the nucleus context are the
296 * same.  The stack pointer is required to be eight-byte aligned even
297 * though this code only needs it to be four-byte aligned.
298 */
299#define	SPILL_32bit(tail)					\
300	srl	%sp, 0, %sp					;\
3011:	st	%l0, [%sp + 0]					;\
302	st	%l1, [%sp + 4]					;\
303	st	%l2, [%sp + 8]					;\
304	st	%l3, [%sp + 12]					;\
305	st	%l4, [%sp + 16]					;\
306	st	%l5, [%sp + 20]					;\
307	st	%l6, [%sp + 24]					;\
308	st	%l7, [%sp + 28]					;\
309	st	%i0, [%sp + 32]					;\
310	st	%i1, [%sp + 36]					;\
311	st	%i2, [%sp + 40]					;\
312	st	%i3, [%sp + 44]					;\
313	st	%i4, [%sp + 48]					;\
314	st	%i5, [%sp + 52]					;\
315	st	%i6, [%sp + 56]					;\
316	st	%i7, [%sp + 60]					;\
317	TT_TRACE_L(trace_win)					;\
318	saved							;\
319	retry							;\
320	SKIP(31-19-TT_TRACE_L_INS)				;\
321	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
322	.empty
323
324/*
325 * SPILL_32bit_asi spills a 32-bit-wide register window into a 32-bit
326 * wide address space via the designated asi.  It is used to spill
327 * non-kernel windows.  The stack pointer is required to be eight-byte
328 * aligned even though this code only needs it to be four-byte
329 * aligned.
330 */
331#define	SPILL_32bit_asi(asi_num, tail)				\
332	srl	%sp, 0, %sp					;\
3331:	sta	%l0, [%sp + %g0]asi_num				;\
334	mov	4, %g1						;\
335	sta	%l1, [%sp + %g1]asi_num				;\
336	mov	8, %g2						;\
337	sta	%l2, [%sp + %g2]asi_num				;\
338	mov	12, %g3						;\
339	sta	%l3, [%sp + %g3]asi_num				;\
340	add	%sp, 16, %g4					;\
341	sta	%l4, [%g4 + %g0]asi_num				;\
342	sta	%l5, [%g4 + %g1]asi_num				;\
343	sta	%l6, [%g4 + %g2]asi_num				;\
344	sta	%l7, [%g4 + %g3]asi_num				;\
345	add	%g4, 16, %g4					;\
346	sta	%i0, [%g4 + %g0]asi_num				;\
347	sta	%i1, [%g4 + %g1]asi_num				;\
348	sta	%i2, [%g4 + %g2]asi_num				;\
349	sta	%i3, [%g4 + %g3]asi_num				;\
350	add	%g4, 16, %g4					;\
351	sta	%i4, [%g4 + %g0]asi_num				;\
352	sta	%i5, [%g4 + %g1]asi_num				;\
353	sta	%i6, [%g4 + %g2]asi_num				;\
354	sta	%i7, [%g4 + %g3]asi_num				;\
355	TT_TRACE_L(trace_win)					;\
356	saved							;\
357	retry							;\
358	SKIP(31-25-TT_TRACE_L_INS)				;\
359	ba,a,pt %xcc, fault_32bit_/**/tail			;\
360	.empty
361
362#define	SPILL_32bit_tt1(asi_num, tail)				\
363	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
364	.empty							;\
365	.align 128
366
367
368/*
369 * FILL_32bit fills a 32-bit-wide kernel register window.  It assumes
370 * that the kernel context and the nucleus context are the same.  The
371 * stack pointer is required to be eight-byte aligned even though this
372 * code only needs it to be four-byte aligned.
373 */
374#define	FILL_32bit(tail)					\
375	srl	%sp, 0, %sp					;\
3761:	TT_TRACE_L(trace_win)					;\
377	ld	[%sp + 0], %l0					;\
378	ld	[%sp + 4], %l1					;\
379	ld	[%sp + 8], %l2					;\
380	ld	[%sp + 12], %l3					;\
381	ld	[%sp + 16], %l4					;\
382	ld	[%sp + 20], %l5					;\
383	ld	[%sp + 24], %l6					;\
384	ld	[%sp + 28], %l7					;\
385	ld	[%sp + 32], %i0					;\
386	ld	[%sp + 36], %i1					;\
387	ld	[%sp + 40], %i2					;\
388	ld	[%sp + 44], %i3					;\
389	ld	[%sp + 48], %i4					;\
390	ld	[%sp + 52], %i5					;\
391	ld	[%sp + 56], %i6					;\
392	ld	[%sp + 60], %i7					;\
393	restored						;\
394	retry							;\
395	SKIP(31-19-TT_TRACE_L_INS)				;\
396	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
397	.empty
398
399/*
400 * FILL_32bit_asi fills a 32-bit-wide register window from a 32-bit
401 * wide address space via the designated asi.  It is used to fill
402 * non-kernel windows.  The stack pointer is required to be eight-byte
403 * aligned even though this code only needs it to be four-byte
404 * aligned.
405 */
406#define	FILL_32bit_asi(asi_num, tail)				\
407	srl	%sp, 0, %sp					;\
4081:	TT_TRACE_L(trace_win)					;\
409	mov	4, %g1						;\
410	lda	[%sp + %g0]asi_num, %l0				;\
411	mov	8, %g2						;\
412	lda	[%sp + %g1]asi_num, %l1				;\
413	mov	12, %g3						;\
414	lda	[%sp + %g2]asi_num, %l2				;\
415	lda	[%sp + %g3]asi_num, %l3				;\
416	add	%sp, 16, %g4					;\
417	lda	[%g4 + %g0]asi_num, %l4				;\
418	lda	[%g4 + %g1]asi_num, %l5				;\
419	lda	[%g4 + %g2]asi_num, %l6				;\
420	lda	[%g4 + %g3]asi_num, %l7				;\
421	add	%g4, 16, %g4					;\
422	lda	[%g4 + %g0]asi_num, %i0				;\
423	lda	[%g4 + %g1]asi_num, %i1				;\
424	lda	[%g4 + %g2]asi_num, %i2				;\
425	lda	[%g4 + %g3]asi_num, %i3				;\
426	add	%g4, 16, %g4					;\
427	lda	[%g4 + %g0]asi_num, %i4				;\
428	lda	[%g4 + %g1]asi_num, %i5				;\
429	lda	[%g4 + %g2]asi_num, %i6				;\
430	lda	[%g4 + %g3]asi_num, %i7				;\
431	restored						;\
432	retry							;\
433	SKIP(31-25-TT_TRACE_L_INS)				;\
434	ba,a,pt %xcc, fault_32bit_/**/tail			;\
435	.empty
436
437
438/*
439 * SPILL_64bit spills a 64-bit-wide kernel register window.  It
440 * assumes that the kernel context and the nucleus context are the
441 * same.  The stack pointer is required to be eight-byte aligned.
442 */
443#define	SPILL_64bit(tail)					\
4442:	stx	%l0, [%sp + V9BIAS64 + 0]			;\
445	stx	%l1, [%sp + V9BIAS64 + 8]			;\
446	stx	%l2, [%sp + V9BIAS64 + 16]			;\
447	stx	%l3, [%sp + V9BIAS64 + 24]			;\
448	stx	%l4, [%sp + V9BIAS64 + 32]			;\
449	stx	%l5, [%sp + V9BIAS64 + 40]			;\
450	stx	%l6, [%sp + V9BIAS64 + 48]			;\
451	stx	%l7, [%sp + V9BIAS64 + 56]			;\
452	stx	%i0, [%sp + V9BIAS64 + 64]			;\
453	stx	%i1, [%sp + V9BIAS64 + 72]			;\
454	stx	%i2, [%sp + V9BIAS64 + 80]			;\
455	stx	%i3, [%sp + V9BIAS64 + 88]			;\
456	stx	%i4, [%sp + V9BIAS64 + 96]			;\
457	stx	%i5, [%sp + V9BIAS64 + 104]			;\
458	stx	%i6, [%sp + V9BIAS64 + 112]			;\
459	stx	%i7, [%sp + V9BIAS64 + 120]			;\
460	TT_TRACE_L(trace_win)					;\
461	saved							;\
462	retry							;\
463	SKIP(31-18-TT_TRACE_L_INS)				;\
464	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
465	.empty
466
467#define	SPILL_64bit_ktt1(tail)				\
468	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
469	.empty							;\
470	.align 128
471
472#define	SPILL_mixed_ktt1(tail)				\
473	btst	1, %sp						;\
474	bz,a,pt	%xcc, fault_32bit_/**/tail			;\
475	srl	%sp, 0, %sp					;\
476	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
477	.empty							;\
478	.align 128
479
480/*
481 * SPILL_64bit_asi spills a 64-bit-wide register window into a 64-bit
482 * wide address space via the designated asi.  It is used to spill
483 * non-kernel windows.  The stack pointer is required to be eight-byte
484 * aligned.
485 */
486#define	SPILL_64bit_asi(asi_num, tail)				\
487	mov	0 + V9BIAS64, %g1				;\
4882:	stxa	%l0, [%sp + %g1]asi_num				;\
489	mov	8 + V9BIAS64, %g2				;\
490	stxa	%l1, [%sp + %g2]asi_num				;\
491	mov	16 + V9BIAS64, %g3				;\
492	stxa	%l2, [%sp + %g3]asi_num				;\
493	mov	24 + V9BIAS64, %g4				;\
494	stxa	%l3, [%sp + %g4]asi_num				;\
495	add	%sp, 32, %g5					;\
496	stxa	%l4, [%g5 + %g1]asi_num				;\
497	stxa	%l5, [%g5 + %g2]asi_num				;\
498	stxa	%l6, [%g5 + %g3]asi_num				;\
499	stxa	%l7, [%g5 + %g4]asi_num				;\
500	add	%g5, 32, %g5					;\
501	stxa	%i0, [%g5 + %g1]asi_num				;\
502	stxa	%i1, [%g5 + %g2]asi_num				;\
503	stxa	%i2, [%g5 + %g3]asi_num				;\
504	stxa	%i3, [%g5 + %g4]asi_num				;\
505	add	%g5, 32, %g5					;\
506	stxa	%i4, [%g5 + %g1]asi_num				;\
507	stxa	%i5, [%g5 + %g2]asi_num				;\
508	stxa	%i6, [%g5 + %g3]asi_num				;\
509	stxa	%i7, [%g5 + %g4]asi_num				;\
510	TT_TRACE_L(trace_win)					;\
511	saved							;\
512	retry							;\
513	SKIP(31-25-TT_TRACE_L_INS)				;\
514	ba,a,pt %xcc, fault_64bit_/**/tail			;\
515	.empty
516
517#define	SPILL_64bit_tt1(asi_num, tail)				\
518	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
519	.empty							;\
520	.align 128
521
522/*
523 * FILL_64bit fills a 64-bit-wide kernel register window.  It assumes
524 * that the kernel context and the nucleus context are the same.  The
525 * stack pointer is required to be eight-byte aligned.
526 */
527#define	FILL_64bit(tail)					\
5282:	TT_TRACE_L(trace_win)					;\
529	ldx	[%sp + V9BIAS64 + 0], %l0			;\
530	ldx	[%sp + V9BIAS64 + 8], %l1			;\
531	ldx	[%sp + V9BIAS64 + 16], %l2			;\
532	ldx	[%sp + V9BIAS64 + 24], %l3			;\
533	ldx	[%sp + V9BIAS64 + 32], %l4			;\
534	ldx	[%sp + V9BIAS64 + 40], %l5			;\
535	ldx	[%sp + V9BIAS64 + 48], %l6			;\
536	ldx	[%sp + V9BIAS64 + 56], %l7			;\
537	ldx	[%sp + V9BIAS64 + 64], %i0			;\
538	ldx	[%sp + V9BIAS64 + 72], %i1			;\
539	ldx	[%sp + V9BIAS64 + 80], %i2			;\
540	ldx	[%sp + V9BIAS64 + 88], %i3			;\
541	ldx	[%sp + V9BIAS64 + 96], %i4			;\
542	ldx	[%sp + V9BIAS64 + 104], %i5			;\
543	ldx	[%sp + V9BIAS64 + 112], %i6			;\
544	ldx	[%sp + V9BIAS64 + 120], %i7			;\
545	restored						;\
546	retry							;\
547	SKIP(31-18-TT_TRACE_L_INS)				;\
548	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
549	.empty
550
551/*
552 * FILL_64bit_asi fills a 64-bit-wide register window from a 64-bit
553 * wide address space via the designated asi.  It is used to fill
554 * non-kernel windows.  The stack pointer is required to be eight-byte
555 * aligned.
556 */
557#define	FILL_64bit_asi(asi_num, tail)				\
558	mov	V9BIAS64 + 0, %g1				;\
5592:	TT_TRACE_L(trace_win)					;\
560	ldxa	[%sp + %g1]asi_num, %l0				;\
561	mov	V9BIAS64 + 8, %g2				;\
562	ldxa	[%sp + %g2]asi_num, %l1				;\
563	mov	V9BIAS64 + 16, %g3				;\
564	ldxa	[%sp + %g3]asi_num, %l2				;\
565	mov	V9BIAS64 + 24, %g4				;\
566	ldxa	[%sp + %g4]asi_num, %l3				;\
567	add	%sp, 32, %g5					;\
568	ldxa	[%g5 + %g1]asi_num, %l4				;\
569	ldxa	[%g5 + %g2]asi_num, %l5				;\
570	ldxa	[%g5 + %g3]asi_num, %l6				;\
571	ldxa	[%g5 + %g4]asi_num, %l7				;\
572	add	%g5, 32, %g5					;\
573	ldxa	[%g5 + %g1]asi_num, %i0				;\
574	ldxa	[%g5 + %g2]asi_num, %i1				;\
575	ldxa	[%g5 + %g3]asi_num, %i2				;\
576	ldxa	[%g5 + %g4]asi_num, %i3				;\
577	add	%g5, 32, %g5					;\
578	ldxa	[%g5 + %g1]asi_num, %i4				;\
579	ldxa	[%g5 + %g2]asi_num, %i5				;\
580	ldxa	[%g5 + %g3]asi_num, %i6				;\
581	ldxa	[%g5 + %g4]asi_num, %i7				;\
582	restored						;\
583	retry							;\
584	SKIP(31-25-TT_TRACE_L_INS)				;\
585	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
586	.empty
587
588
589#endif /* !lint */
590
591/*
592 * SPILL_mixed spills either size window, depending on
593 * whether %sp is even or odd, to a 32-bit address space.
594 * This may only be used in conjunction with SPILL_32bit/
595 * FILL_64bit.
596 * Clear upper 32 bits of %sp if it is odd.
597 * We won't need to clear them in 64 bit kernel.
598 */
599#define	SPILL_mixed						\
600	btst	1, %sp						;\
601	bz,a,pt	%xcc, 1b					;\
602	srl	%sp, 0, %sp					;\
603	ba,pt	%xcc, 2b					;\
604	nop							;\
605	.align	128
606
607/*
608 * FILL_mixed(ASI) fills either size window, depending on
609 * whether %sp is even or odd, from a 32-bit address space.
610 * This may only be used in conjunction with FILL_32bit/
611 * FILL_64bit. New versions of FILL_mixed_{tt1,asi} would be
612 * needed for use with FILL_{32,64}bit_{tt1,asi}. Particular
613 * attention should be paid to the instructions that belong
614 * in the delay slots of the branches depending on the type
615 * of fill handler being branched to.
616 * Clear upper 32 bits of %sp if it is odd.
617 * We won't need to clear them in 64 bit kernel.
618 */
619#define	FILL_mixed						\
620	btst	1, %sp						;\
621	bz,a,pt	%xcc, 1b					;\
622	srl	%sp, 0, %sp					;\
623	ba,pt	%xcc, 2b					;\
624	nop							;\
625	.align	128
626
627
628/*
629 * SPILL_32clean/SPILL_64clean spill 32-bit and 64-bit register windows,
630 * respectively, into the address space via the designated asi.  The
631 * unbiased stack pointer is required to be eight-byte aligned (even for
632 * the 32-bit case even though this code does not require such strict
633 * alignment).
634 *
635 * With SPARC v9 the spill trap takes precedence over the cleanwin trap
636 * so when cansave == 0, canrestore == 6, and cleanwin == 6 the next save
637 * will cause cwp + 2 to be spilled but will not clean cwp + 1.  That
638 * window may contain kernel data so in user_rtt we set wstate to call
639 * these spill handlers on the first user spill trap.  These handler then
640 * spill the appropriate window but also back up a window and clean the
641 * window that didn't get a cleanwin trap.
642 */
643#define	SPILL_32clean(asi_num, tail)				\
644	srl	%sp, 0, %sp					;\
645	sta	%l0, [%sp + %g0]asi_num				;\
646	mov	4, %g1						;\
647	sta	%l1, [%sp + %g1]asi_num				;\
648	mov	8, %g2						;\
649	sta	%l2, [%sp + %g2]asi_num				;\
650	mov	12, %g3						;\
651	sta	%l3, [%sp + %g3]asi_num				;\
652	add	%sp, 16, %g4					;\
653	sta	%l4, [%g4 + %g0]asi_num				;\
654	sta	%l5, [%g4 + %g1]asi_num				;\
655	sta	%l6, [%g4 + %g2]asi_num				;\
656	sta	%l7, [%g4 + %g3]asi_num				;\
657	add	%g4, 16, %g4					;\
658	sta	%i0, [%g4 + %g0]asi_num				;\
659	sta	%i1, [%g4 + %g1]asi_num				;\
660	sta	%i2, [%g4 + %g2]asi_num				;\
661	sta	%i3, [%g4 + %g3]asi_num				;\
662	add	%g4, 16, %g4					;\
663	sta	%i4, [%g4 + %g0]asi_num				;\
664	sta	%i5, [%g4 + %g1]asi_num				;\
665	sta	%i6, [%g4 + %g2]asi_num				;\
666	sta	%i7, [%g4 + %g3]asi_num				;\
667	TT_TRACE_L(trace_win)					;\
668	b	.spill_clean					;\
669	  mov	WSTATE_USER32, %g7				;\
670	SKIP(31-25-TT_TRACE_L_INS)				;\
671	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
672	.empty
673
674#define	SPILL_64clean(asi_num, tail)				\
675	mov	0 + V9BIAS64, %g1				;\
676	stxa	%l0, [%sp + %g1]asi_num				;\
677	mov	8 + V9BIAS64, %g2				;\
678	stxa	%l1, [%sp + %g2]asi_num				;\
679	mov	16 + V9BIAS64, %g3				;\
680	stxa	%l2, [%sp + %g3]asi_num				;\
681	mov	24 + V9BIAS64, %g4				;\
682	stxa	%l3, [%sp + %g4]asi_num				;\
683	add	%sp, 32, %g5					;\
684	stxa	%l4, [%g5 + %g1]asi_num				;\
685	stxa	%l5, [%g5 + %g2]asi_num				;\
686	stxa	%l6, [%g5 + %g3]asi_num				;\
687	stxa	%l7, [%g5 + %g4]asi_num				;\
688	add	%g5, 32, %g5					;\
689	stxa	%i0, [%g5 + %g1]asi_num				;\
690	stxa	%i1, [%g5 + %g2]asi_num				;\
691	stxa	%i2, [%g5 + %g3]asi_num				;\
692	stxa	%i3, [%g5 + %g4]asi_num				;\
693	add	%g5, 32, %g5					;\
694	stxa	%i4, [%g5 + %g1]asi_num				;\
695	stxa	%i5, [%g5 + %g2]asi_num				;\
696	stxa	%i6, [%g5 + %g3]asi_num				;\
697	stxa	%i7, [%g5 + %g4]asi_num				;\
698	TT_TRACE_L(trace_win)					;\
699	b	.spill_clean					;\
700	  mov	WSTATE_USER64, %g7				;\
701	SKIP(31-25-TT_TRACE_L_INS)				;\
702	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
703	.empty
704
705
706/*
707 * Floating point disabled.
708 */
709#define	FP_DISABLED_TRAP		\
710	TT_TRACE(trace_gen)		;\
711	ba,pt	%xcc,.fp_disabled	;\
712	nop				;\
713	.align	32
714
715/*
716 * Floating point exceptions.
717 */
718#define	FP_IEEE_TRAP			\
719	TT_TRACE(trace_gen)		;\
720	ba,pt	%xcc,.fp_ieee_exception	;\
721	nop				;\
722	.align	32
723
724#define	FP_TRAP				\
725	TT_TRACE(trace_gen)		;\
726	ba,pt	%xcc,.fp_exception	;\
727	nop				;\
728	.align	32
729
730#if !defined(lint)
731
732/*
733 * ECACHE_ECC error traps at level 0 and level 1
734 */
735#define	ECACHE_ECC(table_name)		\
736	.global	table_name		;\
737table_name:				;\
738	membar	#Sync			;\
739	set	trap, %g1		;\
740	rdpr	%tt, %g3		;\
741	ba,pt	%xcc, sys_trap		;\
742	sub	%g0, 1, %g4		;\
743	.align	32
744
745#endif /* !lint */
746
747/*
748 * illegal instruction trap
749 */
750#define	ILLTRAP_INSTR			  \
751	membar	#Sync			  ;\
752	TT_TRACE(trace_gen)		  ;\
753	or	%g0, P_UTRAP4, %g2	  ;\
754	or	%g0, T_UNIMP_INSTR, %g3   ;\
755	sethi	%hi(.check_v9utrap), %g4  ;\
756	jmp	%g4 + %lo(.check_v9utrap) ;\
757	nop				  ;\
758	.align	32
759
760/*
761 * tag overflow trap
762 */
763#define	TAG_OVERFLOW			  \
764	TT_TRACE(trace_gen)		  ;\
765	or	%g0, P_UTRAP10, %g2	  ;\
766	or	%g0, T_TAG_OVERFLOW, %g3  ;\
767	sethi	%hi(.check_v9utrap), %g4  ;\
768	jmp	%g4 + %lo(.check_v9utrap) ;\
769	nop				  ;\
770	.align	32
771
772/*
773 * divide by zero trap
774 */
775#define	DIV_BY_ZERO			  \
776	TT_TRACE(trace_gen)		  ;\
777	or	%g0, P_UTRAP11, %g2	  ;\
778	or	%g0, T_IDIV0, %g3	  ;\
779	sethi	%hi(.check_v9utrap), %g4  ;\
780	jmp	%g4 + %lo(.check_v9utrap) ;\
781	nop				  ;\
782	.align	32
783
784/*
785 * trap instruction for V9 user trap handlers
786 */
787#define	TRAP_INSTR			  \
788	TT_TRACE(trace_gen)		  ;\
789	or	%g0, T_SOFTWARE_TRAP, %g3 ;\
790	sethi	%hi(.check_v9utrap), %g4  ;\
791	jmp	%g4 + %lo(.check_v9utrap) ;\
792	nop				  ;\
793	.align	32
794#define	TRP4	TRAP_INSTR; TRAP_INSTR; TRAP_INSTR; TRAP_INSTR
795
796/*
797 * LEVEL_INTERRUPT is for level N interrupts.
798 * VECTOR_INTERRUPT is for the vector trap.
799 */
800#define	LEVEL_INTERRUPT(level)		\
801	.global	tt_pil/**/level		;\
802tt_pil/**/level:			;\
803	ba,pt	%xcc, pil_interrupt	;\
804	mov	level, %g4		;\
805	.align	32
806
807#define	LEVEL14_INTERRUPT			\
808	ba	pil14_interrupt			;\
809	mov	PIL_14, %g4			;\
810	.align	32
811
812#define CPU_MONDO			\
813	ba,a,pt	%xcc, cpu_mondo		;\
814	.align	32
815
816#define DEV_MONDO			\
817	ba,a,pt	%xcc, dev_mondo		;\
818	.align	32
819
820/*
821 * We take over the rtba after we set our trap table and
822 * fault status area. The watchdog reset trap is now handled by the OS.
823 */
824#define WATCHDOG_RESET			\
825	mov	PTL1_BAD_WATCHDOG, %g1	;\
826	ba,a,pt	%xcc, .watchdog_trap	;\
827	.align	32
828
829/*
830 * RED is for traps that use the red mode handler.
831 * We should never see these either.
832 */
833#define RED			\
834	mov	PTL1_BAD_RED, %g1	;\
835	ba,a,pt	%xcc, .watchdog_trap	;\
836	.align	32
837
838
839/*
840 * MMU Trap Handlers.
841 */
842
843/*
844 * synthesize for trap(): SFSR in %g3
845 */
846#define	IMMU_EXCEPTION							\
847	MMU_FAULT_STATUS_AREA(%g3)					;\
848	rdpr	%tpc, %g2						;\
849	ldx	[%g3 + MMFSA_I_TYPE], %g1				;\
850	ldx	[%g3 + MMFSA_I_CTX], %g3				;\
851	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
852	or	%g3, %g1, %g3						;\
853	ba,pt	%xcc, .mmu_exception_end				;\
854	mov	T_INSTR_EXCEPTION, %g1					;\
855	.align	32
856
857/*
858 * synthesize for trap(): TAG_ACCESS in %g2, SFSR in %g3
859 */
860#define	DMMU_EXCEPTION							\
861	ba,a,pt	%xcc, .dmmu_exception					;\
862	.align	32
863
864/*
865 * synthesize for trap(): SFAR in %g2, SFSR in %g3
866 */
867#define	DMMU_EXC_AG_PRIV						\
868	MMU_FAULT_STATUS_AREA(%g3)					;\
869	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
870	/* Fault type not available in MMU fault status area */		;\
871	mov	MMFSA_F_PRVACT, %g1					;\
872	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
873	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
874	ba,pt	%xcc, .mmu_priv_exception				;\
875	or	%g3, %g1, %g3						;\
876	.align	32
877
878/*
879 * synthesize for trap(): SFAR in %g2, SFSR in %g3
880 */
881#define	DMMU_EXC_AG_NOT_ALIGNED						\
882	MMU_FAULT_STATUS_AREA(%g3)					;\
883	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
884	/* Fault type not available in MMU fault status area */		;\
885	mov	MMFSA_F_UNALIGN, %g1					;\
886	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
887	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
888	ba,pt	%xcc, .mmu_exception_not_aligned			;\
889	or	%g3, %g1, %g3			/* SFSR */		;\
890	.align	32
891/*
892 * SPARC V9 IMPL. DEP. #109(1) and (2) and #110(1) and (2)
893 */
894
895/*
896 * synthesize for trap(): SFAR in %g2, SFSR in %g3
897 */
898#define	DMMU_EXC_LDDF_NOT_ALIGNED					\
899	ba,a,pt	%xcc, .dmmu_exc_lddf_not_aligned			;\
900	.align	32
901/*
902 * synthesize for trap(): SFAR in %g2, SFSR in %g3
903 */
904#define	DMMU_EXC_STDF_NOT_ALIGNED					\
905	ba,a,pt	%xcc, .dmmu_exc_stdf_not_aligned			;\
906	.align	32
907
908#if defined(cscope)
909/*
910 * Define labels to direct cscope quickly to labels that
911 * are generated by macro expansion of DTLB_MISS().
912 */
913	.global	tt0_dtlbmiss
914tt0_dtlbmiss:
915	.global	tt1_dtlbmiss
916tt1_dtlbmiss:
917	nop
918#endif
919
920/*
921 * Data miss handler (must be exactly 32 instructions)
922 *
923 * This handler is invoked only if the hypervisor has been instructed
924 * not to do any TSB walk.
925 *
926 * Kernel and invalid context cases are handled by the sfmmu_kdtlb_miss
927 * handler.
928 *
929 * User TLB miss handling depends upon whether a user process has one or
930 * two TSBs. User TSB information (physical base and size code) is kept
931 * in two dedicated scratchpad registers. Absence of a user TSB (primarily
932 * second TSB) is indicated by a negative value (-1) in that register.
933 */
934
935/*
936 * synthesize for miss handler: pseudo-tag access in %g2 (with context "type"
937 * (0=kernel, 1=invalid, or 2=user) rather than context ID)
938 */
939#define	DTLB_MISS(table_name)						;\
940	.global	table_name/**/_dtlbmiss					;\
941table_name/**/_dtlbmiss:						;\
942	GET_MMU_D_PTAGACC_CTXTYPE(%g2, %g3)	/* 8 instr */		;\
943	cmp	%g3, INVALID_CONTEXT					;\
944	ble,pn	%xcc, sfmmu_kdtlb_miss					;\
945	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
946	mov	SCRATCHPAD_UTSBREG2, %g1				;\
947	ldxa	[%g1]ASI_SCRATCHPAD, %g1	/* get 2nd tsbreg */	;\
948	brgez,pn %g1, sfmmu_udtlb_slowpath	/* branch if 2 TSBs */	;\
949	  nop								;\
950	GET_1ST_TSBE_PTR(%g2, %g1, %g4, %g5)	/* 11 instr */		;\
951	ba,pt	%xcc, sfmmu_udtlb_fastpath	/* no 4M TSB, miss */	;\
952	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
953	.align 128
954
955
956#if defined(cscope)
957/*
958 * Define labels to direct cscope quickly to labels that
959 * are generated by macro expansion of ITLB_MISS().
960 */
961	.global	tt0_itlbmiss
962tt0_itlbmiss:
963	.global	tt1_itlbmiss
964tt1_itlbmiss:
965	nop
966#endif
967
968/*
969 * Instruction miss handler.
970 *
971 * This handler is invoked only if the hypervisor has been instructed
972 * not to do any TSB walk.
973 *
974 * ldda instructions will have their ASI patched
975 * by sfmmu_patch_ktsb at runtime.
976 * MUST be EXACTLY 32 instructions or we'll break.
977 */
978
979/*
980 * synthesize for miss handler: TAG_ACCESS in %g2 (with context "type"
981 * (0=kernel, 1=invalid, or 2=user) rather than context ID)
982 */
983#define	ITLB_MISS(table_name)						 \
984	.global	table_name/**/_itlbmiss					;\
985table_name/**/_itlbmiss:						;\
986	GET_MMU_I_PTAGACC_CTXTYPE(%g2, %g3)	/* 8 instr */		;\
987	cmp	%g3, INVALID_CONTEXT					;\
988	ble,pn	%xcc, sfmmu_kitlb_miss					;\
989	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
990	mov	SCRATCHPAD_UTSBREG2, %g1				;\
991	ldxa	[%g1]ASI_SCRATCHPAD, %g1	/* get 2nd tsbreg */	;\
992	brgez,pn %g1, sfmmu_uitlb_slowpath	/* branch if 2 TSBs */	;\
993	  nop								;\
994	GET_1ST_TSBE_PTR(%g2, %g1, %g4, %g5)	/* 11 instr */		;\
995	ba,pt	%xcc, sfmmu_uitlb_fastpath	/* no 4M TSB, miss */	;\
996	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
997	.align 128
998
999#define	DTSB_MISS \
1000	GOTO_TT(sfmmu_slow_dmmu_miss,trace_dmmu)
1001
1002#define	ITSB_MISS \
1003	GOTO_TT(sfmmu_slow_immu_miss,trace_immu)
1004
1005/*
1006 * This macro is the first level handler for fast protection faults.
1007 * It first demaps the tlb entry which generated the fault and then
1008 * attempts to set the modify bit on the hash.  It needs to be
1009 * exactly 32 instructions.
1010 */
1011/*
1012 * synthesize for miss handler: TAG_ACCESS in %g2 (with context "type"
1013 * (0=kernel, 1=invalid, or 2=user) rather than context ID)
1014 */
1015#define	DTLB_PROT							 \
1016	GET_MMU_D_PTAGACC_CTXTYPE(%g2, %g3)	/* 8 instr */		;\
1017	/*								;\
1018	 *   g2 = pseudo-tag access register (ctx type rather than ctx ID) ;\
1019	 *   g3 = ctx type (0, 1, or 2)					;\
1020	 */								;\
1021	TT_TRACE(trace_dataprot)	/* 2 instr ifdef TRAPTRACE */	;\
1022					/* clobbers g1 and g6 XXXQ? */	;\
1023	brnz,pt %g3, sfmmu_uprot_trap		/* user trap */		;\
1024	  nop								;\
1025	ba,a,pt	%xcc, sfmmu_kprot_trap		/* kernel trap */	;\
1026	.align 128
1027
1028#define	DMMU_EXCEPTION_TL1						;\
1029	ba,a,pt	%xcc, mmu_trap_tl1					;\
1030	.align 32
1031
1032#define	MISALIGN_ADDR_TL1						;\
1033	ba,a,pt	%xcc, mmu_trap_tl1					;\
1034	.align 32
1035
1036/*
1037 * Trace a tsb hit
1038 * g1 = tsbe pointer (in/clobbered)
1039 * g2 = tag access register (in)
1040 * g3 - g4 = scratch (clobbered)
1041 * g5 = tsbe data (in)
1042 * g6 = scratch (clobbered)
1043 * g7 = pc we jumped here from (in)
1044 * ttextra = value to OR in to trap type (%tt) (in)
1045 */
1046#ifdef TRAPTRACE
1047#define TRACE_TSBHIT(ttextra)						 \
1048	membar	#Sync							;\
1049	sethi	%hi(FLUSH_ADDR), %g6					;\
1050	flush	%g6							;\
1051	TRACE_PTR(%g3, %g6)						;\
1052	GET_TRACE_TICK(%g6)						;\
1053	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi				;\
1054	stna	%g2, [%g3 + TRAP_ENT_SP]%asi	/* tag access */	;\
1055	stna	%g5, [%g3 + TRAP_ENT_F1]%asi	/* tsb data */		;\
1056	rdpr	%tnpc, %g6						;\
1057	stna	%g6, [%g3 + TRAP_ENT_F2]%asi				;\
1058	stna	%g1, [%g3 + TRAP_ENT_F3]%asi	/* tsb pointer */	;\
1059	stna	%g0, [%g3 + TRAP_ENT_F4]%asi				;\
1060	rdpr	%tpc, %g6						;\
1061	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi				;\
1062	TRACE_SAVE_TL_GL_REGS(%g3, %g6)					;\
1063	rdpr	%tt, %g6						;\
1064	or	%g6, (ttextra), %g1					;\
1065	stha	%g1, [%g3 + TRAP_ENT_TT]%asi				;\
1066	MMU_FAULT_STATUS_AREA(%g4)					;\
1067	mov	MMFSA_D_ADDR, %g1					;\
1068	cmp	%g6, FAST_IMMU_MISS_TT					;\
1069	move	%xcc, MMFSA_I_ADDR, %g1					;\
1070	cmp	%g6, T_INSTR_MMU_MISS					;\
1071	move	%xcc, MMFSA_I_ADDR, %g1					;\
1072	ldx	[%g4 + %g1], %g1					;\
1073	stxa	%g1, [%g3 + TRAP_ENT_TSTATE]%asi /* fault addr */	;\
1074	mov	MMFSA_D_CTX, %g1					;\
1075	cmp	%g6, FAST_IMMU_MISS_TT					;\
1076	move	%xcc, MMFSA_I_CTX, %g1					;\
1077	cmp	%g6, T_INSTR_MMU_MISS					;\
1078	move	%xcc, MMFSA_I_CTX, %g1					;\
1079	ldx	[%g4 + %g1], %g1					;\
1080	stna	%g1, [%g3 + TRAP_ENT_TR]%asi				;\
1081	TRACE_NEXT(%g3, %g4, %g6)
1082#else
1083#define TRACE_TSBHIT(ttextra)
1084#endif
1085
1086
1087#if defined(lint)
1088
1089struct scb	trap_table;
1090struct scb	scb;		/* trap_table/scb are the same object */
1091
1092#else /* lint */
1093
1094/*
1095 * =======================================================================
1096 *		SPARC V9 TRAP TABLE
1097 *
1098 * The trap table is divided into two halves: the first half is used when
1099 * taking traps when TL=0; the second half is used when taking traps from
1100 * TL>0. Note that handlers in the second half of the table might not be able
1101 * to make the same assumptions as handlers in the first half of the table.
1102 *
1103 * Worst case trap nesting so far:
1104 *
1105 *	at TL=0 client issues software trap requesting service
1106 *	at TL=1 nucleus wants a register window
1107 *	at TL=2 register window clean/spill/fill takes a TLB miss
1108 *	at TL=3 processing TLB miss
1109 *	at TL=4 handle asynchronous error
1110 *
1111 * Note that a trap from TL=4 to TL=5 places Spitfire in "RED mode".
1112 *
1113 * =======================================================================
1114 */
1115	.section ".text"
1116	.align	4
1117	.global trap_table, scb, trap_table0, trap_table1, etrap_table
1118	.type	trap_table, #object
1119	.type	trap_table0, #object
1120	.type	trap_table1, #object
1121	.type	scb, #object
1122trap_table:
1123scb:
1124trap_table0:
1125	/* hardware traps */
1126	NOT;				/* 000	reserved */
1127	RED;				/* 001	power on reset */
1128	WATCHDOG_RESET;			/* 002	watchdog reset */
1129	RED;				/* 003	externally initiated reset */
1130	RED;				/* 004	software initiated reset */
1131	RED;				/* 005	red mode exception */
1132	NOT; NOT;			/* 006 - 007 reserved */
1133	IMMU_EXCEPTION;			/* 008	instruction access exception */
1134	ITSB_MISS;			/* 009	instruction access MMU miss */
1135 	NOT;				/* 00A  reserved */
1136	NOT; NOT4;			/* 00B - 00F reserved */
1137	ILLTRAP_INSTR;			/* 010	illegal instruction */
1138	TRAP(T_PRIV_INSTR);		/* 011	privileged opcode */
1139	TRAP(T_UNIMP_LDD);		/* 012	unimplemented LDD */
1140	TRAP(T_UNIMP_STD);		/* 013	unimplemented STD */
1141	NOT4; NOT4; NOT4;		/* 014 - 01F reserved */
1142	FP_DISABLED_TRAP;		/* 020	fp disabled */
1143	FP_IEEE_TRAP;			/* 021	fp exception ieee 754 */
1144	FP_TRAP;			/* 022	fp exception other */
1145	TAG_OVERFLOW;			/* 023	tag overflow */
1146	CLEAN_WINDOW;			/* 024 - 027 clean window */
1147	DIV_BY_ZERO;			/* 028	division by zero */
1148	NOT;				/* 029	internal processor error */
1149	NOT; NOT; NOT4;			/* 02A - 02F reserved */
1150	DMMU_EXCEPTION;			/* 030	data access exception */
1151	DTSB_MISS;			/* 031	data access MMU miss */
1152	NOT;				/* 032  reserved */
1153	NOT;				/* 033	data access protection */
1154	DMMU_EXC_AG_NOT_ALIGNED;	/* 034	mem address not aligned */
1155	DMMU_EXC_LDDF_NOT_ALIGNED;	/* 035	LDDF mem address not aligned */
1156	DMMU_EXC_STDF_NOT_ALIGNED;	/* 036	STDF mem address not aligned */
1157	DMMU_EXC_AG_PRIV;		/* 037	privileged action */
1158	NOT;				/* 038	LDQF mem address not aligned */
1159	NOT;				/* 039	STQF mem address not aligned */
1160	NOT; NOT; NOT4;			/* 03A - 03F reserved */
1161	NOT;				/* 040	async data error */
1162	LEVEL_INTERRUPT(1);		/* 041	interrupt level 1 */
1163	LEVEL_INTERRUPT(2);		/* 042	interrupt level 2 */
1164	LEVEL_INTERRUPT(3);		/* 043	interrupt level 3 */
1165	LEVEL_INTERRUPT(4);		/* 044	interrupt level 4 */
1166	LEVEL_INTERRUPT(5);		/* 045	interrupt level 5 */
1167	LEVEL_INTERRUPT(6);		/* 046	interrupt level 6 */
1168	LEVEL_INTERRUPT(7);		/* 047	interrupt level 7 */
1169	LEVEL_INTERRUPT(8);		/* 048	interrupt level 8 */
1170	LEVEL_INTERRUPT(9);		/* 049	interrupt level 9 */
1171	LEVEL_INTERRUPT(10);		/* 04A	interrupt level 10 */
1172	LEVEL_INTERRUPT(11);		/* 04B	interrupt level 11 */
1173	LEVEL_INTERRUPT(12);		/* 04C	interrupt level 12 */
1174	LEVEL_INTERRUPT(13);		/* 04D	interrupt level 13 */
1175	LEVEL14_INTERRUPT;		/* 04E	interrupt level 14 */
1176	LEVEL_INTERRUPT(15);		/* 04F	interrupt level 15 */
1177	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F reserved */
1178	NOT;				/* 060	interrupt vector */
1179	GOTO(kmdb_trap);		/* 061	PA watchpoint */
1180	GOTO(kmdb_trap);		/* 062	VA watchpoint */
1181	NOT;				/* 063	reserved */
1182	ITLB_MISS(tt0);			/* 064	instruction access MMU miss */
1183	DTLB_MISS(tt0);			/* 068	data access MMU miss */
1184	DTLB_PROT;			/* 06C	data access protection */
1185	NOT;				/* 070  reserved */
1186	NOT;				/* 071  reserved */
1187	NOT;				/* 072  reserved */
1188	NOT;				/* 073  reserved */
1189	NOT4; NOT4			/* 074 - 07B reserved */
1190	CPU_MONDO;			/* 07C	cpu_mondo */
1191	DEV_MONDO;			/* 07D	dev_mondo */
1192	GOTO_TT(resumable_error, trace_gen);	/* 07E  resumable error */
1193	GOTO_TT(nonresumable_error, trace_gen);	/* 07F  non-reasumable error */
1194	NOT4;				/* 080	spill 0 normal */
1195	SPILL_32bit_asi(ASI_AIUP,sn0);	/* 084	spill 1 normal */
1196	SPILL_64bit_asi(ASI_AIUP,sn0);	/* 088	spill 2 normal */
1197	SPILL_32clean(ASI_AIUP,sn0);	/* 08C	spill 3 normal */
1198	SPILL_64clean(ASI_AIUP,sn0);	/* 090	spill 4 normal */
1199	SPILL_32bit(not);		/* 094	spill 5 normal */
1200	SPILL_64bit(not);		/* 098	spill 6 normal */
1201	SPILL_mixed;			/* 09C	spill 7 normal */
1202	NOT4;				/* 0A0	spill 0 other */
1203	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0A4	spill 1 other */
1204	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0A8	spill 2 other */
1205	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0AC	spill 3 other */
1206	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0B0	spill 4 other */
1207	NOT4;				/* 0B4	spill 5 other */
1208	NOT4;				/* 0B8	spill 6 other */
1209	NOT4;				/* 0BC	spill 7 other */
1210	NOT4;				/* 0C0	fill 0 normal */
1211	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0C4	fill 1 normal */
1212	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0C8	fill 2 normal */
1213	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0CC	fill 3 normal */
1214	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0D0	fill 4 normal */
1215	FILL_32bit(not);		/* 0D4	fill 5 normal */
1216	FILL_64bit(not);		/* 0D8	fill 6 normal */
1217	FILL_mixed;			/* 0DC	fill 7 normal */
1218	NOT4;				/* 0E0	fill 0 other */
1219	NOT4;				/* 0E4	fill 1 other */
1220	NOT4;				/* 0E8	fill 2 other */
1221	NOT4;				/* 0EC	fill 3 other */
1222	NOT4;				/* 0F0	fill 4 other */
1223	NOT4;				/* 0F4	fill 5 other */
1224	NOT4;				/* 0F8	fill 6 other */
1225	NOT4;				/* 0FC	fill 7 other */
1226	/* user traps */
1227	GOTO(syscall_trap_4x);		/* 100	old system call */
1228	TRAP(T_BREAKPOINT);		/* 101	user breakpoint */
1229	TRAP(T_DIV0);			/* 102	user divide by zero */
1230	GOTO(.flushw);			/* 103	flush windows */
1231	GOTO(.clean_windows);		/* 104	clean windows */
1232	BAD;				/* 105	range check ?? */
1233	GOTO(.fix_alignment);		/* 106	do unaligned references */
1234	BAD;				/* 107	unused */
1235	SYSCALL_TRAP32;			/* 108	ILP32 system call on LP64 */
1236	GOTO(set_trap0_addr);		/* 109	set trap0 address */
1237	BAD; BAD; BAD4;			/* 10A - 10F unused */
1238	TRP4; TRP4; TRP4; TRP4;		/* 110 - 11F V9 user trap handlers */
1239	GOTO(.getcc);			/* 120	get condition codes */
1240	GOTO(.setcc);			/* 121	set condition codes */
1241	GOTO(.getpsr);			/* 122	get psr */
1242	GOTO(.setpsr);			/* 123	set psr (some fields) */
1243	GOTO(get_timestamp);		/* 124	get timestamp */
1244	GOTO(get_virtime);		/* 125	get lwp virtual time */
1245	PRIV(self_xcall);		/* 126	self xcall */
1246	GOTO(get_hrestime);		/* 127	get hrestime */
1247	BAD;				/* 128	ST_SETV9STACK */
1248	GOTO(.getlgrp);			/* 129  get lgrpid */
1249	BAD; BAD; BAD4;			/* 12A - 12F unused */
1250	BAD4; BAD4; 			/* 130 - 137 unused */
1251	DTRACE_PID;			/* 138  dtrace pid tracing provider */
1252	BAD;				/* 139  unused */
1253	DTRACE_RETURN;			/* 13A	dtrace pid return probe */
1254	BAD; BAD4;			/* 13B - 13F unused */
1255	SYSCALL_TRAP;			/* 140  LP64 system call */
1256	SYSCALL(nosys);			/* 141  unused system call trap */
1257#ifdef DEBUG_USER_TRAPTRACECTL
1258	GOTO(.traptrace_freeze);	/* 142  freeze traptrace */
1259	GOTO(.traptrace_unfreeze);	/* 143  unfreeze traptrace */
1260#else
1261	SYSCALL(nosys);			/* 142  unused system call trap */
1262	SYSCALL(nosys);			/* 143  unused system call trap */
1263#endif
1264	BAD4; BAD4; BAD4;		/* 144 - 14F unused */
1265	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1266	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1267	BAD;				/* 170 - unused */
1268	BAD;				/* 171 - unused */
1269	BAD; BAD;			/* 172 - 173 unused */
1270	BAD4; BAD4;			/* 174 - 17B unused */
1271#ifdef	PTL1_PANIC_DEBUG
1272	mov PTL1_BAD_DEBUG, %g1; GOTO(ptl1_panic);
1273					/* 17C	test ptl1_panic */
1274#else
1275	BAD;				/* 17C  unused */
1276#endif	/* PTL1_PANIC_DEBUG */
1277	PRIV(kmdb_trap);		/* 17D	kmdb enter (L1-A) */
1278	PRIV(kmdb_trap);		/* 17E	kmdb breakpoint */
1279	PRIV(obp_bpt);			/* 17F	obp breakpoint */
1280	/* reserved */
1281	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1282	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1283	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1284	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1285	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1286	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1287	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1288	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1289	.size	trap_table0, (.-trap_table0)
1290trap_table1:
1291	NOT4; NOT4;			/* 000 - 007 unused */
1292	NOT;				/* 008	instruction access exception */
1293	ITSB_MISS;			/* 009	instruction access MMU miss */
1294 	NOT;				/* 00A  reserved */
1295	NOT; NOT4;			/* 00B - 00F unused */
1296	NOT4; NOT4; NOT4; NOT4;		/* 010 - 01F unused */
1297	NOT4;				/* 020 - 023 unused */
1298	CLEAN_WINDOW;			/* 024 - 027 clean window */
1299	NOT4; NOT4;			/* 028 - 02F unused */
1300	DMMU_EXCEPTION_TL1;		/* 030 	data access exception */
1301	DTSB_MISS;			/* 031  data access MMU miss */
1302	NOT;				/* 032  reserved */
1303	NOT;				/* 033	unused */
1304	MISALIGN_ADDR_TL1;		/* 034	mem address not aligned */
1305	NOT; NOT; NOT; NOT4; NOT4	/* 035 - 03F unused */
1306	NOT4; NOT4; NOT4; NOT4;		/* 040 - 04F unused */
1307	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F unused */
1308	NOT;				/* 060	unused */
1309	GOTO(kmdb_trap_tl1);		/* 061	PA watchpoint */
1310	GOTO(kmdb_trap_tl1);		/* 062	VA watchpoint */
1311	NOT;				/* 063	reserved */
1312	ITLB_MISS(tt1);			/* 064	instruction access MMU miss */
1313	DTLB_MISS(tt1);			/* 068	data access MMU miss */
1314	DTLB_PROT;			/* 06C	data access protection */
1315	NOT;				/* 070  reserved */
1316	NOT;				/* 071  reserved */
1317	NOT;				/* 072  reserved */
1318	NOT;				/* 073  reserved */
1319	NOT4; NOT4;			/* 074 - 07B reserved */
1320	NOT;				/* 07C  reserved */
1321	NOT;				/* 07D  reserved */
1322	NOT;				/* 07E  resumable error */
1323	GOTO_TT(nonresumable_error, trace_gen);	/* 07F  nonresumable error */
1324	NOTP4;				/* 080	spill 0 normal */
1325	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 084	spill 1 normal */
1326	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 088	spill 2 normal */
1327	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 08C	spill 3 normal */
1328	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 090	spill 4 normal */
1329	NOTP4;				/* 094	spill 5 normal */
1330	SPILL_64bit_ktt1(sk);		/* 098	spill 6 normal */
1331	SPILL_mixed_ktt1(sk);		/* 09C	spill 7 normal */
1332	NOTP4;				/* 0A0	spill 0 other */
1333	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0A4  spill 1 other */
1334	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0A8	spill 2 other */
1335	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0AC	spill 3 other */
1336	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0B0  spill 4 other */
1337	NOTP4;				/* 0B4  spill 5 other */
1338	NOTP4;				/* 0B8  spill 6 other */
1339	NOTP4;				/* 0BC  spill 7 other */
1340	NOT4;				/* 0C0	fill 0 normal */
1341	NOT4;				/* 0C4	fill 1 normal */
1342	NOT4;				/* 0C8	fill 2 normal */
1343	NOT4;				/* 0CC	fill 3 normal */
1344	NOT4;				/* 0D0	fill 4 normal */
1345	NOT4;				/* 0D4	fill 5 normal */
1346	NOT4;				/* 0D8	fill 6 normal */
1347	NOT4;				/* 0DC	fill 7 normal */
1348	NOT4; NOT4; NOT4; NOT4;		/* 0E0 - 0EF unused */
1349	NOT4; NOT4; NOT4; NOT4;		/* 0F0 - 0FF unused */
1350/*
1351 * Code running at TL>0 does not use soft traps, so
1352 * we can truncate the table here.
1353 * However:
1354 * sun4v uses (hypervisor) ta instructions at TL > 0, so
1355 * provide a safety net for now.
1356 */
1357	/* soft traps */
1358	BAD4; BAD4; BAD4; BAD4;		/* 100 - 10F unused */
1359	BAD4; BAD4; BAD4; BAD4;		/* 110 - 11F unused */
1360	BAD4; BAD4; BAD4; BAD4;		/* 120 - 12F unused */
1361	BAD4; BAD4; BAD4; BAD4;		/* 130 - 13F unused */
1362	BAD4; BAD4; BAD4; BAD4;		/* 140 - 14F unused */
1363	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1364	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1365	BAD4; BAD4; BAD4; BAD4;		/* 170 - 17F unused */
1366	/* reserved */
1367	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1368	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1369	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1370	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1371	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1372	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1373	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1374	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1375etrap_table:
1376	.size	trap_table1, (.-trap_table1)
1377	.size	trap_table, (.-trap_table)
1378	.size	scb, (.-scb)
1379
1380/*
1381 * We get to exec_fault in the case of an instruction miss and tte
1382 * has no execute bit set.  We go to tl0 to handle it.
1383 *
1384 * g1 = tsbe pointer (in/clobbered)
1385 * g2 = tag access register (in)
1386 * g3 - g4 = scratch (clobbered)
1387 * g5 = tsbe data (in)
1388 * g6 = scratch (clobbered)
1389 * g7 = pc we jumped here from (in)
1390 */
1391/*
1392 * synthesize for miss handler: TAG_ACCESS in %g2 (with context "type"
1393 * (0=kernel, 1=invalid, or 2=user) rather than context ID)
1394 */
1395	ALTENTRY(exec_fault)
1396	TRACE_TSBHIT(TT_MMU_EXEC)
1397	MMU_FAULT_STATUS_AREA(%g4)
1398	ldx	[%g4 + MMFSA_I_ADDR], %g2	/* g2 = address */
1399	ldx	[%g4 + MMFSA_I_CTX], %g3	/* g3 = ctx */
1400	srlx	%g2, MMU_PAGESHIFT, %g2		! align address to page boundry
1401	cmp	%g3, USER_CONTEXT_TYPE
1402	sllx	%g2, MMU_PAGESHIFT, %g2
1403	movgu	%icc, USER_CONTEXT_TYPE, %g3
1404	or	%g2, %g3, %g2			/* TAG_ACCESS */
1405	mov	T_INSTR_MMU_MISS, %g3		! arg2 = traptype
1406	set	trap, %g1
1407	ba,pt	%xcc, sys_trap
1408	  mov	-1, %g4
1409
1410.mmu_exception_not_aligned:
1411	/* %g2 = sfar, %g3 = sfsr */
1412	rdpr	%tstate, %g1
1413	btst	TSTATE_PRIV, %g1
1414	bnz,pn	%icc, 2f
1415	nop
1416	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1417	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1418	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1419	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1420	brz,pt	%g5, 2f
1421	nop
1422	ldn	[%g5 + P_UTRAP15], %g5			! unaligned utrap?
1423	brz,pn	%g5, 2f
1424	nop
1425	btst	1, %sp
1426	bz,pt	%xcc, 1f				! 32 bit user program
1427	nop
1428	ba,pt	%xcc, .setup_v9utrap			! 64 bit user program
1429	nop
14301:
1431	ba,pt	%xcc, .setup_utrap
1432	or	%g2, %g0, %g7
14332:
1434	ba,pt	%xcc, .mmu_exception_end
1435	mov	T_ALIGNMENT, %g1
1436
1437.mmu_priv_exception:
1438	rdpr	%tstate, %g1
1439	btst	TSTATE_PRIV, %g1
1440	bnz,pn	%icc, 1f
1441	nop
1442	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1443	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1444	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1445	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1446	brz,pt	%g5, 1f
1447	nop
1448	ldn	[%g5 + P_UTRAP16], %g5
1449	brnz,pt	%g5, .setup_v9utrap
1450	nop
14511:
1452	mov	T_PRIV_INSTR, %g1
1453
1454.mmu_exception_end:
1455	CPU_INDEX(%g4, %g5)
1456	set	cpu_core, %g5
1457	sllx	%g4, CPU_CORE_SHIFT, %g4
1458	add	%g4, %g5, %g4
1459	lduh	[%g4 + CPUC_DTRACE_FLAGS], %g5
1460	andcc	%g5, CPU_DTRACE_NOFAULT, %g0
1461	bz	1f
1462	or	%g5, CPU_DTRACE_BADADDR, %g5
1463	stuh	%g5, [%g4 + CPUC_DTRACE_FLAGS]
1464	done
1465
14661:
1467	sllx	%g3, 32, %g3
1468	or	%g3, %g1, %g3
1469	set	trap, %g1
1470	ba,pt	%xcc, sys_trap
1471	sub	%g0, 1, %g4
1472
1473.fp_disabled:
1474	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1475	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1476	rdpr	%tstate, %g4
1477	btst	TSTATE_PRIV, %g4
1478	bnz,a,pn %icc, ptl1_panic
1479	  mov	PTL1_BAD_FPTRAP, %g1
1480
1481	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1482	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1483	brz,a,pt %g5, 2f
1484	  nop
1485	ldn	[%g5 + P_UTRAP7], %g5			! fp_disabled utrap?
1486	brz,a,pn %g5, 2f
1487	  nop
1488	btst	1, %sp
1489	bz,a,pt	%xcc, 1f				! 32 bit user program
1490	  nop
1491	ba,a,pt	%xcc, .setup_v9utrap			! 64 bit user program
1492	  nop
14931:
1494	ba,pt	%xcc, .setup_utrap
1495	  or	%g0, %g0, %g7
14962:
1497	set	fp_disabled, %g1
1498	ba,pt	%xcc, sys_trap
1499	  sub	%g0, 1, %g4
1500
1501.fp_ieee_exception:
1502	rdpr	%tstate, %g1
1503	btst	TSTATE_PRIV, %g1
1504	bnz,a,pn %icc, ptl1_panic
1505	  mov	PTL1_BAD_FPTRAP, %g1
1506	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1507	stx	%fsr, [%g1 + CPU_TMP1]
1508	ldx	[%g1 + CPU_TMP1], %g2
1509	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1510	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1511	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1512	brz,a,pt %g5, 1f
1513	  nop
1514	ldn	[%g5 + P_UTRAP8], %g5
1515	brnz,a,pt %g5, .setup_v9utrap
1516	  nop
15171:
1518	set	_fp_ieee_exception, %g1
1519	ba,pt	%xcc, sys_trap
1520	  sub	%g0, 1, %g4
1521
1522/*
1523 * Register Inputs:
1524 *	%g5		user trap handler
1525 *	%g7		misaligned addr - for alignment traps only
1526 */
1527.setup_utrap:
1528	set	trap, %g1			! setup in case we go
1529	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1530	sub	%g0, 1, %g4			! the save instruction below
1531
1532	/*
1533	 * If the DTrace pid provider is single stepping a copied-out
1534	 * instruction, t->t_dtrace_step will be set. In that case we need
1535	 * to abort the single-stepping (since execution of the instruction
1536	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1537	 */
1538	save	%sp, -SA(MINFRAME32), %sp	! window for trap handler
1539	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1540	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1541	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1542	rdpr	%tnpc, %l2			! arg1 == tnpc
1543	brz,pt	%g2, 1f
1544	rdpr	%tpc, %l1			! arg0 == tpc
1545
1546	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1547	ldn	[%g1 + T_DTRACE_NPC], %l2	! arg1 = t->t_dtrace_npc (step)
1548	brz,pt	%g2, 1f
1549	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1550	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
15511:
1552	mov	%g7, %l3			! arg2 == misaligned address
1553
1554	rdpr	%tstate, %g1			! cwp for trap handler
1555	rdpr	%cwp, %g4
1556	bclr	TSTATE_CWP_MASK, %g1
1557	wrpr	%g1, %g4, %tstate
1558	wrpr	%g0, %g5, %tnpc			! trap handler address
1559	FAST_TRAP_DONE
1560	/* NOTREACHED */
1561
1562.check_v9utrap:
1563	rdpr	%tstate, %g1
1564	btst	TSTATE_PRIV, %g1
1565	bnz,a,pn %icc, 3f
1566	  nop
1567	CPU_ADDR(%g4, %g1)				! load CPU struct addr
1568	ldn	[%g4 + CPU_THREAD], %g5			! load thread pointer
1569	ldn	[%g5 + T_PROCP], %g5			! load proc pointer
1570	ldn	[%g5 + P_UTRAPS], %g5			! are there utraps?
1571
1572	cmp	%g3, T_SOFTWARE_TRAP
1573	bne,a,pt %icc, 1f
1574	  nop
1575
1576	brz,pt %g5, 3f			! if p_utraps == NULL goto trap()
1577	  rdpr	%tt, %g3		! delay - get actual hw trap type
1578
1579	sub	%g3, 254, %g1		! UT_TRAP_INSTRUCTION_16 = p_utraps[18]
1580	ba,pt	%icc, 2f
1581	  smul	%g1, CPTRSIZE, %g2
15821:
1583	brz,a,pt %g5, 3f		! if p_utraps == NULL goto trap()
1584	  nop
1585
1586	cmp	%g3, T_UNIMP_INSTR
1587	bne,a,pt %icc, 2f
1588	  nop
1589
1590	mov	1, %g1
1591	st	%g1, [%g4 + CPU_TL1_HDLR] ! set CPU_TL1_HDLR
1592	rdpr	%tpc, %g1		! ld trapping instruction using
1593	lduwa	[%g1]ASI_AIUP, %g1	! "AS IF USER" ASI which could fault
1594	st	%g0, [%g4 + CPU_TL1_HDLR] ! clr CPU_TL1_HDLR
1595
1596	sethi	%hi(0xc1c00000), %g4	! setup mask for illtrap instruction
1597	andcc	%g1, %g4, %g4		! and instruction with mask
1598	bnz,a,pt %icc, 3f		! if %g4 == zero, %g1 is an ILLTRAP
1599	  nop				! fall thru to setup
16002:
1601	ldn	[%g5 + %g2], %g5
1602	brnz,a,pt %g5, .setup_v9utrap
1603	  nop
16043:
1605	set	trap, %g1
1606	ba,pt	%xcc, sys_trap
1607	  sub	%g0, 1, %g4
1608	/* NOTREACHED */
1609
1610/*
1611 * Register Inputs:
1612 *	%g5		user trap handler
1613 */
1614.setup_v9utrap:
1615	set	trap, %g1			! setup in case we go
1616	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1617	sub	%g0, 1, %g4			! the save instruction below
1618
1619	/*
1620	 * If the DTrace pid provider is single stepping a copied-out
1621	 * instruction, t->t_dtrace_step will be set. In that case we need
1622	 * to abort the single-stepping (since execution of the instruction
1623	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1624	 */
1625	save	%sp, -SA(MINFRAME64), %sp	! window for trap handler
1626	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1627	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1628	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1629	rdpr	%tnpc, %l7			! arg1 == tnpc
1630	brz,pt	%g2, 1f
1631	rdpr	%tpc, %l6			! arg0 == tpc
1632
1633	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1634	ldn	[%g1 + T_DTRACE_NPC], %l7	! arg1 == t->t_dtrace_npc (step)
1635	brz,pt	%g2, 1f
1636	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1637	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
16381:
1639	rdpr	%tstate, %g2			! cwp for trap handler
1640	rdpr	%cwp, %g4
1641	bclr	TSTATE_CWP_MASK, %g2
1642	wrpr	%g2, %g4, %tstate
1643
1644	ldn	[%g1 + T_PROCP], %g4		! load proc pointer
1645	ldn	[%g4 + P_AS], %g4		! load as pointer
1646	ldn	[%g4 + A_USERLIMIT], %g4	! load as userlimit
1647	cmp	%l7, %g4			! check for single-step set
1648	bne,pt	%xcc, 4f
1649	  nop
1650	ldn	[%g1 + T_LWP], %g1		! load klwp pointer
1651	ld	[%g1 + PCB_STEP], %g4		! load single-step flag
1652	cmp	%g4, STEP_ACTIVE		! step flags set in pcb?
1653	bne,pt	%icc, 4f
1654	  nop
1655	stn	%g5, [%g1 + PCB_TRACEPC]	! save trap handler addr in pcb
1656	mov	%l7, %g4			! on entry to precise user trap
1657	add	%l6, 4, %l7			! handler, %l6 == pc, %l7 == npc
1658						! at time of trap
1659	wrpr	%g0, %g4, %tnpc			! generate FLTBOUNDS,
1660						! %g4 == userlimit
1661	FAST_TRAP_DONE
1662	/* NOTREACHED */
16634:
1664	wrpr	%g0, %g5, %tnpc			! trap handler address
1665	FAST_TRAP_DONE_CHK_INTR
1666	/* NOTREACHED */
1667
1668.fp_exception:
1669	CPU_ADDR(%g1, %g4)
1670	stx	%fsr, [%g1 + CPU_TMP1]
1671	ldx	[%g1 + CPU_TMP1], %g2
1672
1673	/*
1674	 * Cheetah takes unfinished_FPop trap for certain range of operands
1675	 * to the "fitos" instruction. Instead of going through the slow
1676	 * software emulation path, we try to simulate the "fitos" instruction
1677	 * via "fitod" and "fdtos" provided the following conditions are met:
1678	 *
1679	 *	fpu_exists is set (if DEBUG)
1680	 *	not in privileged mode
1681	 *	ftt is unfinished_FPop
1682	 *	NXM IEEE trap is not enabled
1683	 *	instruction at %tpc is "fitos"
1684	 *
1685	 *  Usage:
1686	 *	%g1	per cpu address
1687	 *	%g2	%fsr
1688	 *	%g6	user instruction
1689	 *
1690	 * Note that we can take a memory access related trap while trying
1691	 * to fetch the user instruction. Therefore, we set CPU_TL1_HDLR
1692	 * flag to catch those traps and let the SFMMU code deal with page
1693	 * fault and data access exception.
1694	 */
1695#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
1696	sethi	%hi(fpu_exists), %g7
1697	ld	[%g7 + %lo(fpu_exists)], %g7
1698	brz,pn %g7, .fp_exception_cont
1699	  nop
1700#endif
1701	rdpr	%tstate, %g7			! branch if in privileged mode
1702	btst	TSTATE_PRIV, %g7
1703	bnz,pn	%xcc, .fp_exception_cont
1704	srl	%g2, FSR_FTT_SHIFT, %g7		! extract ftt from %fsr
1705	and	%g7, (FSR_FTT>>FSR_FTT_SHIFT), %g7
1706	cmp	%g7, FTT_UNFIN
1707	set	FSR_TEM_NX, %g5
1708	bne,pn	%xcc, .fp_exception_cont	! branch if NOT unfinished_FPop
1709	  andcc	%g2, %g5, %g0
1710	bne,pn	%xcc, .fp_exception_cont	! branch if FSR_TEM_NX enabled
1711	  rdpr	%tpc, %g5			! get faulting PC
1712
1713	or	%g0, 1, %g7
1714	st	%g7, [%g1 + CPU_TL1_HDLR]	! set tl1_hdlr flag
1715	lda	[%g5]ASI_USER, %g6		! get user's instruction
1716	st	%g0, [%g1 + CPU_TL1_HDLR]	! clear tl1_hdlr flag
1717
1718	set	FITOS_INSTR_MASK, %g7
1719	and	%g6, %g7, %g7
1720	set	FITOS_INSTR, %g5
1721	cmp	%g7, %g5
1722	bne,pn	%xcc, .fp_exception_cont	! branch if not FITOS_INSTR
1723	 nop
1724
1725	/*
1726	 * This is unfinished FPops trap for "fitos" instruction. We
1727	 * need to simulate "fitos" via "fitod" and "fdtos" instruction
1728	 * sequence.
1729	 *
1730	 * We need a temporary FP register to do the conversion. Since
1731	 * both source and destination operands for the "fitos" instruction
1732	 * have to be within %f0-%f31, we use an FP register from the upper
1733	 * half to guarantee that it won't collide with the source or the
1734	 * dest operand. However, we do have to save and restore its value.
1735	 *
1736	 * We use %d62 as a temporary FP register for the conversion and
1737	 * branch to appropriate instruction within the conversion tables
1738	 * based upon the rs2 and rd values.
1739	 */
1740
1741	std	%d62, [%g1 + CPU_TMP1]		! save original value
1742
1743	srl	%g6, FITOS_RS2_SHIFT, %g7
1744	and	%g7, FITOS_REG_MASK, %g7
1745	set	_fitos_fitod_table, %g4
1746	sllx	%g7, 2, %g7
1747	jmp	%g4 + %g7
1748	  ba,pt	%xcc, _fitos_fitod_done
1749	.empty
1750
1751_fitos_fitod_table:
1752	  fitod	%f0, %d62
1753	  fitod	%f1, %d62
1754	  fitod	%f2, %d62
1755	  fitod	%f3, %d62
1756	  fitod	%f4, %d62
1757	  fitod	%f5, %d62
1758	  fitod	%f6, %d62
1759	  fitod	%f7, %d62
1760	  fitod	%f8, %d62
1761	  fitod	%f9, %d62
1762	  fitod	%f10, %d62
1763	  fitod	%f11, %d62
1764	  fitod	%f12, %d62
1765	  fitod	%f13, %d62
1766	  fitod	%f14, %d62
1767	  fitod	%f15, %d62
1768	  fitod	%f16, %d62
1769	  fitod	%f17, %d62
1770	  fitod	%f18, %d62
1771	  fitod	%f19, %d62
1772	  fitod	%f20, %d62
1773	  fitod	%f21, %d62
1774	  fitod	%f22, %d62
1775	  fitod	%f23, %d62
1776	  fitod	%f24, %d62
1777	  fitod	%f25, %d62
1778	  fitod	%f26, %d62
1779	  fitod	%f27, %d62
1780	  fitod	%f28, %d62
1781	  fitod	%f29, %d62
1782	  fitod	%f30, %d62
1783	  fitod	%f31, %d62
1784_fitos_fitod_done:
1785
1786	/*
1787	 * Now convert data back into single precision
1788	 */
1789	srl	%g6, FITOS_RD_SHIFT, %g7
1790	and	%g7, FITOS_REG_MASK, %g7
1791	set	_fitos_fdtos_table, %g4
1792	sllx	%g7, 2, %g7
1793	jmp	%g4 + %g7
1794	  ba,pt	%xcc, _fitos_fdtos_done
1795	.empty
1796
1797_fitos_fdtos_table:
1798	  fdtos	%d62, %f0
1799	  fdtos	%d62, %f1
1800	  fdtos	%d62, %f2
1801	  fdtos	%d62, %f3
1802	  fdtos	%d62, %f4
1803	  fdtos	%d62, %f5
1804	  fdtos	%d62, %f6
1805	  fdtos	%d62, %f7
1806	  fdtos	%d62, %f8
1807	  fdtos	%d62, %f9
1808	  fdtos	%d62, %f10
1809	  fdtos	%d62, %f11
1810	  fdtos	%d62, %f12
1811	  fdtos	%d62, %f13
1812	  fdtos	%d62, %f14
1813	  fdtos	%d62, %f15
1814	  fdtos	%d62, %f16
1815	  fdtos	%d62, %f17
1816	  fdtos	%d62, %f18
1817	  fdtos	%d62, %f19
1818	  fdtos	%d62, %f20
1819	  fdtos	%d62, %f21
1820	  fdtos	%d62, %f22
1821	  fdtos	%d62, %f23
1822	  fdtos	%d62, %f24
1823	  fdtos	%d62, %f25
1824	  fdtos	%d62, %f26
1825	  fdtos	%d62, %f27
1826	  fdtos	%d62, %f28
1827	  fdtos	%d62, %f29
1828	  fdtos	%d62, %f30
1829	  fdtos	%d62, %f31
1830_fitos_fdtos_done:
1831
1832	ldd	[%g1 + CPU_TMP1], %d62		! restore %d62
1833
1834#if DEBUG
1835	/*
1836	 * Update FPop_unfinished trap kstat
1837	 */
1838	set	fpustat+FPUSTAT_UNFIN_KSTAT, %g7
1839	ldx	[%g7], %g5
18401:
1841	add	%g5, 1, %g6
1842
1843	casxa	[%g7] ASI_N, %g5, %g6
1844	cmp	%g5, %g6
1845	bne,a,pn %xcc, 1b
1846	  or	%g0, %g6, %g5
1847
1848	/*
1849	 * Update fpu_sim_fitos kstat
1850	 */
1851	set	fpuinfo+FPUINFO_FITOS_KSTAT, %g7
1852	ldx	[%g7], %g5
18531:
1854	add	%g5, 1, %g6
1855
1856	casxa	[%g7] ASI_N, %g5, %g6
1857	cmp	%g5, %g6
1858	bne,a,pn %xcc, 1b
1859	  or	%g0, %g6, %g5
1860#endif /* DEBUG */
1861
1862	FAST_TRAP_DONE
1863
1864.fp_exception_cont:
1865	/*
1866	 * Let _fp_exception deal with simulating FPop instruction.
1867	 * Note that we need to pass %fsr in %g2 (already read above).
1868	 */
1869
1870	set	_fp_exception, %g1
1871	ba,pt	%xcc, sys_trap
1872	sub	%g0, 1, %g4
1873
1874
1875/*
1876 * Register windows
1877 */
1878.flushw:
1879.clean_windows:
1880	rdpr	%tnpc, %g1
1881	wrpr	%g1, %tpc
1882	add	%g1, 4, %g1
1883	wrpr	%g1, %tnpc
1884	set	trap, %g1
1885	mov	T_FLUSH_PCB, %g3
1886	ba,pt	%xcc, sys_trap
1887	sub	%g0, 1, %g4
1888
1889/*
1890 * .spill_clean: clean the previous window, restore the wstate, and
1891 * "done".
1892 *
1893 * Entry: %g7 contains new wstate
1894 */
1895.spill_clean:
1896	sethi	%hi(nwin_minus_one), %g5
1897	ld	[%g5 + %lo(nwin_minus_one)], %g5 ! %g5 = nwin - 1
1898	rdpr	%cwp, %g6			! %g6 = %cwp
1899	deccc	%g6				! %g6--
1900	movneg	%xcc, %g5, %g6			! if (%g6<0) %g6 = nwin-1
1901	wrpr	%g6, %cwp
1902	TT_TRACE_L(trace_win)
1903	clr	%l0
1904	clr	%l1
1905	clr	%l2
1906	clr	%l3
1907	clr	%l4
1908	clr	%l5
1909	clr	%l6
1910	clr	%l7
1911	wrpr	%g0, %g7, %wstate
1912	saved
1913	retry			! restores correct %cwp
1914
1915.fix_alignment:
1916	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
1917	ldn	[%g1 + CPU_THREAD], %g1	! load thread pointer
1918	ldn	[%g1 + T_PROCP], %g1
1919	mov	1, %g2
1920	stb	%g2, [%g1 + P_FIXALIGNMENT]
1921	FAST_TRAP_DONE
1922
1923#define	STDF_REG(REG, ADDR, TMP)		\
1924	sll	REG, 3, REG			;\
1925mark1:	set	start1, TMP			;\
1926	jmp	REG + TMP			;\
1927	  nop					;\
1928start1:	ba,pt	%xcc, done1			;\
1929	  std	%f0, [ADDR + CPU_TMP1]		;\
1930	ba,pt	%xcc, done1			;\
1931	  std	%f32, [ADDR + CPU_TMP1]		;\
1932	ba,pt	%xcc, done1			;\
1933	  std	%f2, [ADDR + CPU_TMP1]		;\
1934	ba,pt	%xcc, done1			;\
1935	  std	%f34, [ADDR + CPU_TMP1]		;\
1936	ba,pt	%xcc, done1			;\
1937	  std	%f4, [ADDR + CPU_TMP1]		;\
1938	ba,pt	%xcc, done1			;\
1939	  std	%f36, [ADDR + CPU_TMP1]		;\
1940	ba,pt	%xcc, done1			;\
1941	  std	%f6, [ADDR + CPU_TMP1]		;\
1942	ba,pt	%xcc, done1			;\
1943	  std	%f38, [ADDR + CPU_TMP1]		;\
1944	ba,pt	%xcc, done1			;\
1945	  std	%f8, [ADDR + CPU_TMP1]		;\
1946	ba,pt	%xcc, done1			;\
1947	  std	%f40, [ADDR + CPU_TMP1]		;\
1948	ba,pt	%xcc, done1			;\
1949	  std	%f10, [ADDR + CPU_TMP1]		;\
1950	ba,pt	%xcc, done1			;\
1951	  std	%f42, [ADDR + CPU_TMP1]		;\
1952	ba,pt	%xcc, done1			;\
1953	  std	%f12, [ADDR + CPU_TMP1]		;\
1954	ba,pt	%xcc, done1			;\
1955	  std	%f44, [ADDR + CPU_TMP1]		;\
1956	ba,pt	%xcc, done1			;\
1957	  std	%f14, [ADDR + CPU_TMP1]		;\
1958	ba,pt	%xcc, done1			;\
1959	  std	%f46, [ADDR + CPU_TMP1]		;\
1960	ba,pt	%xcc, done1			;\
1961	  std	%f16, [ADDR + CPU_TMP1]		;\
1962	ba,pt	%xcc, done1			;\
1963	  std	%f48, [ADDR + CPU_TMP1]		;\
1964	ba,pt	%xcc, done1			;\
1965	  std	%f18, [ADDR + CPU_TMP1]		;\
1966	ba,pt	%xcc, done1			;\
1967	  std	%f50, [ADDR + CPU_TMP1]		;\
1968	ba,pt	%xcc, done1			;\
1969	  std	%f20, [ADDR + CPU_TMP1]		;\
1970	ba,pt	%xcc, done1			;\
1971	  std	%f52, [ADDR + CPU_TMP1]		;\
1972	ba,pt	%xcc, done1			;\
1973	  std	%f22, [ADDR + CPU_TMP1]		;\
1974	ba,pt	%xcc, done1			;\
1975	  std	%f54, [ADDR + CPU_TMP1]		;\
1976	ba,pt	%xcc, done1			;\
1977	  std	%f24, [ADDR + CPU_TMP1]		;\
1978	ba,pt	%xcc, done1			;\
1979	  std	%f56, [ADDR + CPU_TMP1]		;\
1980	ba,pt	%xcc, done1			;\
1981	  std	%f26, [ADDR + CPU_TMP1]		;\
1982	ba,pt	%xcc, done1			;\
1983	  std	%f58, [ADDR + CPU_TMP1]		;\
1984	ba,pt	%xcc, done1			;\
1985	  std	%f28, [ADDR + CPU_TMP1]		;\
1986	ba,pt	%xcc, done1			;\
1987	  std	%f60, [ADDR + CPU_TMP1]		;\
1988	ba,pt	%xcc, done1			;\
1989	  std	%f30, [ADDR + CPU_TMP1]		;\
1990	ba,pt	%xcc, done1			;\
1991	  std	%f62, [ADDR + CPU_TMP1]		;\
1992done1:
1993
1994#define	LDDF_REG(REG, ADDR, TMP)		\
1995	sll	REG, 3, REG			;\
1996mark2:	set	start2, TMP			;\
1997	jmp	REG + TMP			;\
1998	  nop					;\
1999start2:	ba,pt	%xcc, done2			;\
2000	  ldd	[ADDR + CPU_TMP1], %f0		;\
2001	ba,pt	%xcc, done2			;\
2002	  ldd	[ADDR + CPU_TMP1], %f32		;\
2003	ba,pt	%xcc, done2			;\
2004	  ldd	[ADDR + CPU_TMP1], %f2		;\
2005	ba,pt	%xcc, done2			;\
2006	  ldd	[ADDR + CPU_TMP1], %f34		;\
2007	ba,pt	%xcc, done2			;\
2008	  ldd	[ADDR + CPU_TMP1], %f4		;\
2009	ba,pt	%xcc, done2			;\
2010	  ldd	[ADDR + CPU_TMP1], %f36		;\
2011	ba,pt	%xcc, done2			;\
2012	  ldd	[ADDR + CPU_TMP1], %f6		;\
2013	ba,pt	%xcc, done2			;\
2014	  ldd	[ADDR + CPU_TMP1], %f38		;\
2015	ba,pt	%xcc, done2			;\
2016	  ldd	[ADDR + CPU_TMP1], %f8		;\
2017	ba,pt	%xcc, done2			;\
2018	  ldd	[ADDR + CPU_TMP1], %f40		;\
2019	ba,pt	%xcc, done2			;\
2020	  ldd	[ADDR + CPU_TMP1], %f10		;\
2021	ba,pt	%xcc, done2			;\
2022	  ldd	[ADDR + CPU_TMP1], %f42		;\
2023	ba,pt	%xcc, done2			;\
2024	  ldd	[ADDR + CPU_TMP1], %f12		;\
2025	ba,pt	%xcc, done2			;\
2026	  ldd	[ADDR + CPU_TMP1], %f44		;\
2027	ba,pt	%xcc, done2			;\
2028	  ldd	[ADDR + CPU_TMP1], %f14		;\
2029	ba,pt	%xcc, done2			;\
2030	  ldd	[ADDR + CPU_TMP1], %f46		;\
2031	ba,pt	%xcc, done2			;\
2032	  ldd	[ADDR + CPU_TMP1], %f16		;\
2033	ba,pt	%xcc, done2			;\
2034	  ldd	[ADDR + CPU_TMP1], %f48		;\
2035	ba,pt	%xcc, done2			;\
2036	  ldd	[ADDR + CPU_TMP1], %f18		;\
2037	ba,pt	%xcc, done2			;\
2038	  ldd	[ADDR + CPU_TMP1], %f50		;\
2039	ba,pt	%xcc, done2			;\
2040	  ldd	[ADDR + CPU_TMP1], %f20		;\
2041	ba,pt	%xcc, done2			;\
2042	  ldd	[ADDR + CPU_TMP1], %f52		;\
2043	ba,pt	%xcc, done2			;\
2044	  ldd	[ADDR + CPU_TMP1], %f22		;\
2045	ba,pt	%xcc, done2			;\
2046	  ldd	[ADDR + CPU_TMP1], %f54		;\
2047	ba,pt	%xcc, done2			;\
2048	  ldd	[ADDR + CPU_TMP1], %f24		;\
2049	ba,pt	%xcc, done2			;\
2050	  ldd	[ADDR + CPU_TMP1], %f56		;\
2051	ba,pt	%xcc, done2			;\
2052	  ldd	[ADDR + CPU_TMP1], %f26		;\
2053	ba,pt	%xcc, done2			;\
2054	  ldd	[ADDR + CPU_TMP1], %f58		;\
2055	ba,pt	%xcc, done2			;\
2056	  ldd	[ADDR + CPU_TMP1], %f28		;\
2057	ba,pt	%xcc, done2			;\
2058	  ldd	[ADDR + CPU_TMP1], %f60		;\
2059	ba,pt	%xcc, done2			;\
2060	  ldd	[ADDR + CPU_TMP1], %f30		;\
2061	ba,pt	%xcc, done2			;\
2062	  ldd	[ADDR + CPU_TMP1], %f62		;\
2063done2:
2064
2065.lddf_exception_not_aligned:
2066	/* %g2 = sfar, %g3 = sfsr */
2067	mov	%g2, %g5		! stash sfar
2068#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2069	sethi	%hi(fpu_exists), %g2	! check fpu_exists
2070	ld	[%g2 + %lo(fpu_exists)], %g2
2071	brz,a,pn %g2, 4f
2072	  nop
2073#endif
2074	CPU_ADDR(%g1, %g4)
2075	or	%g0, 1, %g4
2076	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2077
2078	rdpr	%tpc, %g2
2079	lda	[%g2]ASI_AIUP, %g6	! get the user's lddf instruction
2080	srl	%g6, 23, %g1		! using ldda or not?
2081	and	%g1, 1, %g1
2082	brz,a,pt %g1, 2f		! check for ldda instruction
2083	  nop
2084	srl	%g6, 13, %g1		! check immflag
2085	and	%g1, 1, %g1
2086	rdpr	%tstate, %g2		! %tstate in %g2
2087	brnz,a,pn %g1, 1f
2088	  srl	%g2, 31, %g1		! get asi from %tstate
2089	srl	%g6, 5, %g1		! get asi from instruction
2090	and	%g1, 0xFF, %g1		! imm_asi field
20911:
2092	cmp	%g1, ASI_P		! primary address space
2093	be,a,pt %icc, 2f
2094	  nop
2095	cmp	%g1, ASI_PNF		! primary no fault address space
2096	be,a,pt %icc, 2f
2097	  nop
2098	cmp	%g1, ASI_S		! secondary address space
2099	be,a,pt %icc, 2f
2100	  nop
2101	cmp	%g1, ASI_SNF		! secondary no fault address space
2102	bne,a,pn %icc, 3f
2103	  nop
21042:
2105	lduwa	[%g5]ASI_USER, %g7	! get first half of misaligned data
2106	add	%g5, 4, %g5		! increment misaligned data address
2107	lduwa	[%g5]ASI_USER, %g5	! get second half of misaligned data
2108
2109	sllx	%g7, 32, %g7
2110	or	%g5, %g7, %g5		! combine data
2111	CPU_ADDR(%g7, %g1)		! save data on a per-cpu basis
2112	stx	%g5, [%g7 + CPU_TMP1]	! save in cpu_tmp1
2113
2114	srl	%g6, 25, %g3		! %g6 has the instruction
2115	and	%g3, 0x1F, %g3		! %g3 has rd
2116	LDDF_REG(%g3, %g7, %g4)
2117
2118	CPU_ADDR(%g1, %g4)
2119	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2120	FAST_TRAP_DONE
21213:
2122	CPU_ADDR(%g1, %g4)
2123	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
21244:
2125	set	T_USER, %g3		! trap type in %g3
2126	or	%g3, T_LDDF_ALIGN, %g3
2127	mov	%g5, %g2		! misaligned vaddr in %g2
2128	set	fpu_trap, %g1		! goto C for the little and
2129	ba,pt	%xcc, sys_trap		! no fault little asi's
2130	  sub	%g0, 1, %g4
2131
2132.stdf_exception_not_aligned:
2133	/* %g2 = sfar, %g3 = sfsr */
2134	mov	%g2, %g5
2135
2136#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2137	sethi	%hi(fpu_exists), %g7		! check fpu_exists
2138	ld	[%g7 + %lo(fpu_exists)], %g3
2139	brz,a,pn %g3, 4f
2140	  nop
2141#endif
2142	CPU_ADDR(%g1, %g4)
2143	or	%g0, 1, %g4
2144	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2145
2146	rdpr	%tpc, %g2
2147	lda	[%g2]ASI_AIUP, %g6	! get the user's stdf instruction
2148
2149	srl	%g6, 23, %g1		! using stda or not?
2150	and	%g1, 1, %g1
2151	brz,a,pt %g1, 2f		! check for stda instruction
2152	  nop
2153	srl	%g6, 13, %g1		! check immflag
2154	and	%g1, 1, %g1
2155	rdpr	%tstate, %g2		! %tstate in %g2
2156	brnz,a,pn %g1, 1f
2157	  srl	%g2, 31, %g1		! get asi from %tstate
2158	srl	%g6, 5, %g1		! get asi from instruction
2159	and	%g1, 0xff, %g1		! imm_asi field
21601:
2161	cmp	%g1, ASI_P		! primary address space
2162	be,a,pt %icc, 2f
2163	  nop
2164	cmp	%g1, ASI_S		! secondary address space
2165	bne,a,pn %icc, 3f
2166	  nop
21672:
2168	srl	%g6, 25, %g6
2169	and	%g6, 0x1F, %g6		! %g6 has rd
2170	CPU_ADDR(%g7, %g1)
2171	STDF_REG(%g6, %g7, %g4)		! STDF_REG(REG, ADDR, TMP)
2172
2173	ldx	[%g7 + CPU_TMP1], %g6
2174	srlx	%g6, 32, %g7
2175	stuwa	%g7, [%g5]ASI_USER	! first half
2176	add	%g5, 4, %g5		! increment misaligned data address
2177	stuwa	%g6, [%g5]ASI_USER	! second half
2178
2179	CPU_ADDR(%g1, %g4)
2180	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2181	FAST_TRAP_DONE
21823:
2183	CPU_ADDR(%g1, %g4)
2184	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
21854:
2186	set	T_USER, %g3		! trap type in %g3
2187	or	%g3, T_STDF_ALIGN, %g3
2188	mov	%g5, %g2		! misaligned vaddr in %g2
2189	set	fpu_trap, %g1		! goto C for the little and
2190	ba,pt	%xcc, sys_trap		! nofault little asi's
2191	  sub	%g0, 1, %g4
2192
2193#ifdef DEBUG_USER_TRAPTRACECTL
2194
2195.traptrace_freeze:
2196	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2197	TT_TRACE_L(trace_win)
2198	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2199	set	trap_freeze, %g1
2200	mov	1, %g2
2201	st	%g2, [%g1]
2202	FAST_TRAP_DONE
2203
2204.traptrace_unfreeze:
2205	set	trap_freeze, %g1
2206	st	%g0, [%g1]
2207	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2208	TT_TRACE_L(trace_win)
2209	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2210	FAST_TRAP_DONE
2211
2212#endif /* DEBUG_USER_TRAPTRACECTL */
2213
2214.getcc:
2215	CPU_ADDR(%g1, %g2)
2216	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2217	rdpr	%tstate, %g3			! get tstate
2218	srlx	%g3, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2219	set	PSR_ICC, %g2
2220	and	%o0, %g2, %o0			! mask out the rest
2221	srl	%o0, PSR_ICC_SHIFT, %o0		! right justify
2222	wrpr	%g0, 0, %gl
2223	mov	%o0, %g1			! move ccr to normal %g1
2224	wrpr	%g0, 1, %gl
2225	! cannot assume globals retained their values after increasing %gl
2226	CPU_ADDR(%g1, %g2)
2227	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2228	FAST_TRAP_DONE
2229
2230.setcc:
2231	CPU_ADDR(%g1, %g2)
2232	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2233	wrpr	%g0, 0, %gl
2234	mov	%g1, %o0
2235	wrpr	%g0, 1, %gl
2236	! cannot assume globals retained their values after increasing %gl
2237	CPU_ADDR(%g1, %g2)
2238	sll	%o0, PSR_ICC_SHIFT, %g2
2239	set	PSR_ICC, %g3
2240	and	%g2, %g3, %g2			! mask out rest
2241	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g2
2242	rdpr	%tstate, %g3			! get tstate
2243	srl	%g3, 0, %g3			! clear upper word
2244	or	%g3, %g2, %g3			! or in new bits
2245	wrpr	%g3, %tstate
2246	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2247	FAST_TRAP_DONE
2248
2249/*
2250 * getpsr(void)
2251 * Note that the xcc part of the ccr is not provided.
2252 * The V8 code shows why the V9 trap is not faster:
2253 * #define GETPSR_TRAP() \
2254 *      mov %psr, %i0; jmp %l2; rett %l2+4; nop;
2255 */
2256
2257	.type	.getpsr, #function
2258.getpsr:
2259	rdpr	%tstate, %g1			! get tstate
2260	srlx	%g1, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2261	set	PSR_ICC, %g2
2262	and	%o0, %g2, %o0			! mask out the rest
2263
2264	rd	%fprs, %g1			! get fprs
2265	and	%g1, FPRS_FEF, %g2		! mask out dirty upper/lower
2266	sllx	%g2, PSR_FPRS_FEF_SHIFT, %g2	! shift fef to V8 psr.ef
2267	or	%o0, %g2, %o0			! or result into psr.ef
2268
2269	set	V9_PSR_IMPLVER, %g2		! SI assigned impl/ver: 0xef
2270	or	%o0, %g2, %o0			! or psr.impl/ver
2271	FAST_TRAP_DONE
2272	SET_SIZE(.getpsr)
2273
2274/*
2275 * setpsr(newpsr)
2276 * Note that there is no support for ccr.xcc in the V9 code.
2277 */
2278
2279	.type	.setpsr, #function
2280.setpsr:
2281	rdpr	%tstate, %g1			! get tstate
2282!	setx	TSTATE_V8_UBITS, %g2
2283	or 	%g0, CCR_ICC, %g3
2284	sllx	%g3, TSTATE_CCR_SHIFT, %g2
2285
2286	andn	%g1, %g2, %g1			! zero current user bits
2287	set	PSR_ICC, %g2
2288	and	%g2, %o0, %g2			! clear all but psr.icc bits
2289	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g3	! shift to tstate.ccr.icc
2290	wrpr	%g1, %g3, %tstate		! write tstate
2291
2292	set	PSR_EF, %g2
2293	and	%g2, %o0, %g2			! clear all but fp enable bit
2294	srlx	%g2, PSR_FPRS_FEF_SHIFT, %g4	! shift ef to V9 fprs.fef
2295	wr	%g0, %g4, %fprs			! write fprs
2296
2297	CPU_ADDR(%g1, %g2)			! load CPU struct addr to %g1
2298	ldn	[%g1 + CPU_THREAD], %g2		! load thread pointer
2299	ldn	[%g2 + T_LWP], %g3		! load klwp pointer
2300	ldn	[%g3 + LWP_FPU], %g2		! get lwp_fpu pointer
2301	stuw	%g4, [%g2 + FPU_FPRS]		! write fef value to fpu_fprs
2302	srlx	%g4, 2, %g4			! shift fef value to bit 0
2303	stub	%g4, [%g2 + FPU_EN]		! write fef value to fpu_en
2304	FAST_TRAP_DONE
2305	SET_SIZE(.setpsr)
2306
2307/*
2308 * getlgrp
2309 * get home lgrpid on which the calling thread is currently executing.
2310 */
2311	.type	.getlgrp, #function
2312.getlgrp:
2313	! Thanks for the incredibly helpful comments
2314	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2315	ld	[%g1 + CPU_ID], %o0	! load cpu_id
2316	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2317	ldn	[%g2 + T_LPL], %g2	! load lpl pointer
2318	ld	[%g2 + LPL_LGRPID], %g1	! load lpl_lgrpid
2319	sra	%g1, 0, %o1
2320	FAST_TRAP_DONE
2321	SET_SIZE(.getlgrp)
2322
2323/*
2324 * Entry for old 4.x trap (trap 0).
2325 */
2326	ENTRY_NP(syscall_trap_4x)
2327	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2328	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2329	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2330	ld	[%g2 + PCB_TRAP0], %g2	! lwp->lwp_pcb.pcb_trap0addr
2331	brz,pn	%g2, 1f			! has it been set?
2332	st	%l0, [%g1 + CPU_TMP1]	! delay - save some locals
2333	st	%l1, [%g1 + CPU_TMP2]
2334	rdpr	%tnpc, %l1		! save old tnpc
2335	wrpr	%g0, %g2, %tnpc		! setup tnpc
2336
2337	mov	%g1, %l0		! save CPU struct addr
2338	wrpr	%g0, 0, %gl
2339	mov	%l1, %g6		! pass tnpc to user code in %g6
2340	wrpr	%g0, 1, %gl
2341	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2342	ld	[%l0 + CPU_TMP1], %l0
2343	FAST_TRAP_DONE_CHK_INTR
23441:
2345	!
2346	! check for old syscall mmap which is the only different one which
2347	! must be the same.  Others are handled in the compatibility library.
2348	!
2349	mov	%g1, %l0		! save CPU struct addr
2350	wrpr	%g0, 0, %gl
2351	cmp	%g1, OSYS_mmap		! compare to old 4.x mmap
2352	movz	%icc, SYS_mmap, %g1
2353	wrpr	%g0, 1, %gl
2354	ld	[%l0 + CPU_TMP1], %l0
2355	SYSCALL(syscall_trap32)
2356	SET_SIZE(syscall_trap_4x)
2357
2358/*
2359 * Handler for software trap 9.
2360 * Set trap0 emulation address for old 4.x system call trap.
2361 * XXX - this should be a system call.
2362 */
2363	ENTRY_NP(set_trap0_addr)
2364	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2365	st	%l0, [%g1 + CPU_TMP1]	! save some locals
2366	st	%l1, [%g1 + CPU_TMP2]
2367	mov	%g1, %l0	! preserve CPU addr
2368	wrpr	%g0, 0, %gl
2369	mov	%g1, %l1
2370	wrpr	%g0, 1, %gl
2371	! cannot assume globals retained their values after increasing %gl
2372	ldn	[%l0 + CPU_THREAD], %g2	! load thread pointer
2373	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2374	andn	%l1, 3, %l1		! force alignment
2375	st	%l1, [%g2 + PCB_TRAP0]	! lwp->lwp_pcb.pcb_trap0addr
2376	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2377	ld	[%l0 + CPU_TMP1], %l0
2378	FAST_TRAP_DONE
2379	SET_SIZE(set_trap0_addr)
2380
2381/*
2382 * mmu_trap_tl1
2383 * trap handler for unexpected mmu traps.
2384 * simply checks if the trap was a user lddf/stdf alignment trap, in which
2385 * case we go to fpu_trap or a user trap from the window handler, in which
2386 * case we go save the state on the pcb.  Otherwise, we go to ptl1_panic.
2387 */
2388	.type	mmu_trap_tl1, #function
2389mmu_trap_tl1:
2390#ifdef	TRAPTRACE
2391	TRACE_PTR(%g5, %g6)
2392	GET_TRACE_TICK(%g6)
2393	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2394	TRACE_SAVE_TL_GL_REGS(%g5, %g6)
2395	rdpr	%tt, %g6
2396	stha	%g6, [%g5 + TRAP_ENT_TT]%asi
2397	rdpr	%tstate, %g6
2398	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi
2399	stna	%sp, [%g5 + TRAP_ENT_SP]%asi
2400	stna	%g0, [%g5 + TRAP_ENT_TR]%asi
2401	rdpr	%tpc, %g6
2402	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2403	MMU_FAULT_STATUS_AREA(%g6)
2404	ldx	[%g6 + MMFSA_D_ADDR], %g6
2405	stna	%g6, [%g5 + TRAP_ENT_F1]%asi !  MMU fault address
2406	CPU_PADDR(%g7, %g6);
2407	add	%g7, CPU_TL1_HDLR, %g7
2408	lda	[%g7]ASI_MEM, %g6
2409	stna	%g6, [%g5 + TRAP_ENT_F2]%asi
2410	MMU_FAULT_STATUS_AREA(%g6)
2411	ldx	[%g6 + MMFSA_D_TYPE], %g7 ! XXXQ should be a MMFSA_F_ constant?
2412	ldx	[%g6 + MMFSA_D_CTX], %g6
2413	sllx	%g6, SFSR_CTX_SHIFT, %g6
2414	or	%g6, %g7, %g6
2415	stna	%g6, [%g5 + TRAP_ENT_F3]%asi ! MMU context/type
2416	set	0xdeadbeef, %g6
2417	stna	%g6, [%g5 + TRAP_ENT_F4]%asi
2418	TRACE_NEXT(%g5, %g6, %g7)
2419#endif /* TRAPTRACE */
2420	CPU_PADDR(%g7, %g6);
2421	add     %g7, CPU_TL1_HDLR, %g7		! %g7 = &cpu_m.tl1_hdlr (PA)
2422	lda	[%g7]ASI_MEM, %g6
2423	brz,a,pt %g6, 1f
2424	  nop
2425	sta     %g0, [%g7]ASI_MEM
2426	! XXXQ need to setup registers for sfmmu_mmu_trap?
2427	ba,a,pt	%xcc, sfmmu_mmu_trap		! handle page faults
24281:
2429	rdpr	%tpc, %g7
2430	/* in user_rtt? */
2431	set	rtt_fill_start, %g6
2432	cmp	%g7, %g6
2433	blu,pn	%xcc, 6f
2434	 .empty
2435	set	rtt_fill_end, %g6
2436	cmp	%g7, %g6
2437	bgeu,pn %xcc, 6f
2438	 nop
2439	set	fault_rtt_fn1, %g7
2440	ba,a	7f
24416:
2442	! check to see if the trap pc is in a window spill/fill handling
2443	rdpr	%tpc, %g7
2444	/* tpc should be in the trap table */
2445	set	trap_table, %g6
2446	cmp	%g7, %g6
2447	blu,a,pn %xcc, ptl1_panic
2448	  mov	PTL1_BAD_MMUTRAP, %g1
2449	set	etrap_table, %g6
2450	cmp	%g7, %g6
2451	bgeu,a,pn %xcc, ptl1_panic
2452	  mov	PTL1_BAD_MMUTRAP, %g1
2453	! pc is inside the trap table, convert to trap type
2454	srl	%g7, 5, %g6		! XXXQ need #define
2455	and	%g6, 0x1ff, %g6		! XXXQ need #define
2456	! and check for a window trap type
2457	and	%g6, WTRAP_TTMASK, %g6
2458	cmp	%g6, WTRAP_TYPE
2459	bne,a,pn %xcc, ptl1_panic
2460	  mov	PTL1_BAD_MMUTRAP, %g1
2461	andn	%g7, WTRAP_ALIGN, %g7	/* 128 byte aligned */
2462	add	%g7, WTRAP_FAULTOFF, %g7
2463
24647:
2465	! Arguments are passed in the global set active after the
2466	! 'done' instruction. Before switching sets, must save
2467	! the calculated next pc
2468	wrpr	%g0, %g7, %tnpc
2469	wrpr	%g0, 1, %gl
2470	rdpr	%tt, %g5
2471	MMU_FAULT_STATUS_AREA(%g7)
2472	cmp	%g5, T_ALIGNMENT
2473	be,pn	%xcc, 1f
2474	ldx	[%g7 + MMFSA_D_ADDR], %g6
2475	ldx	[%g7 + MMFSA_D_CTX], %g7
2476	srlx	%g6, MMU_PAGESHIFT, %g6		/* align address */
2477	cmp	%g7, USER_CONTEXT_TYPE
2478	sllx	%g6, MMU_PAGESHIFT, %g6
2479	movgu	%icc, USER_CONTEXT_TYPE, %g7
2480	or	%g6, %g7, %g6			/* TAG_ACCESS */
24811:
2482	done
2483	SET_SIZE(mmu_trap_tl1)
2484
2485/*
2486 * Several traps use kmdb_trap and kmdb_trap_tl1 as their handlers.  These
2487 * traps are valid only when kmdb is loaded.  When the debugger is active,
2488 * the code below is rewritten to transfer control to the appropriate
2489 * debugger entry points.
2490 */
2491	.global	kmdb_trap
2492	.align	8
2493kmdb_trap:
2494	ba,a	trap_table0
2495	jmp	%g1 + 0
2496	nop
2497
2498	.global	kmdb_trap_tl1
2499	.align	8
2500kmdb_trap_tl1:
2501	ba,a	trap_table0
2502	jmp	%g1 + 0
2503	nop
2504
2505/*
2506 * This entry is copied from OBP's trap table during boot.
2507 */
2508	.global	obp_bpt
2509	.align	8
2510obp_bpt:
2511	NOT
2512
2513
2514
2515#ifdef	TRAPTRACE
2516/*
2517 * TRAPTRACE support.
2518 * labels here are branched to with "rd %pc, %g7" in the delay slot.
2519 * Return is done by "jmp %g7 + 4".
2520 */
2521
2522trace_dmmu:
2523	TRACE_PTR(%g3, %g6)
2524	GET_TRACE_TICK(%g6)
2525	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2526	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2527	rdpr	%tt, %g6
2528	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2529	rdpr	%tstate, %g6
2530	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2531	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2532	rdpr	%tpc, %g6
2533	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2534	MMU_FAULT_STATUS_AREA(%g6)
2535	ldx	[%g6 + MMFSA_D_ADDR], %g4
2536	stxa	%g4, [%g3 + TRAP_ENT_TR]%asi
2537	ldx	[%g6 + MMFSA_D_CTX], %g4
2538	stxa	%g4, [%g3 + TRAP_ENT_F1]%asi
2539	ldx	[%g6 + MMFSA_D_TYPE], %g4
2540	stxa	%g4, [%g3 + TRAP_ENT_F2]%asi
2541	stxa	%g6, [%g3 + TRAP_ENT_F3]%asi
2542	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2543	TRACE_NEXT(%g3, %g4, %g5)
2544	jmp	%g7 + 4
2545	nop
2546
2547trace_immu:
2548	TRACE_PTR(%g3, %g6)
2549	GET_TRACE_TICK(%g6)
2550	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2551	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2552	rdpr	%tt, %g6
2553	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2554	rdpr	%tstate, %g6
2555	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2556	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2557	rdpr	%tpc, %g6
2558	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2559	MMU_FAULT_STATUS_AREA(%g6)
2560	ldx	[%g6 + MMFSA_I_ADDR], %g4
2561	stxa	%g4, [%g3 + TRAP_ENT_TR]%asi
2562	ldx	[%g6 + MMFSA_I_CTX], %g4
2563	stxa	%g4, [%g3 + TRAP_ENT_F1]%asi
2564	ldx	[%g6 + MMFSA_I_TYPE], %g4
2565	stxa	%g4, [%g3 + TRAP_ENT_F2]%asi
2566	stxa	%g6, [%g3 + TRAP_ENT_F3]%asi
2567	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2568	TRACE_NEXT(%g3, %g4, %g5)
2569	jmp	%g7 + 4
2570	nop
2571
2572trace_gen:
2573	TRACE_PTR(%g3, %g6)
2574	GET_TRACE_TICK(%g6)
2575	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2576	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2577	rdpr	%tt, %g6
2578	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2579	rdpr	%tstate, %g6
2580	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2581	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2582	rdpr	%tpc, %g6
2583	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2584	stna	%g0, [%g3 + TRAP_ENT_TR]%asi
2585	stna	%g0, [%g3 + TRAP_ENT_F1]%asi
2586	stna	%g0, [%g3 + TRAP_ENT_F2]%asi
2587	stna	%g0, [%g3 + TRAP_ENT_F3]%asi
2588	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2589	TRACE_NEXT(%g3, %g4, %g5)
2590	jmp	%g7 + 4
2591	nop
2592
2593trace_win:
2594	TRACE_WIN_INFO(0, %l0, %l1, %l2)
2595	! Keep the locals as clean as possible, caller cleans %l4
2596	clr	%l2
2597	clr	%l1
2598	jmp	%l4 + 4
2599	  clr	%l0
2600
2601/*
2602 * Trace a tsb hit
2603 * g1 = tsbe pointer (in/clobbered)
2604 * g2 = tag access register (in)
2605 * g3 - g4 = scratch (clobbered)
2606 * g5 = tsbe data (in)
2607 * g6 = scratch (clobbered)
2608 * g7 = pc we jumped here from (in)
2609 */
2610
2611	! Do not disturb %g5, it will be used after the trace
2612	ALTENTRY(trace_tsbhit)
2613	TRACE_TSBHIT(0)
2614	jmp	%g7 + 4
2615	nop
2616
2617/*
2618 * Trace a TSB miss
2619 *
2620 * g1 = tsb8k pointer (in)
2621 * g2 = tag access register (in)
2622 * g3 = tsb4m pointer (in)
2623 * g4 = tsbe tag (in/clobbered)
2624 * g5 - g6 = scratch (clobbered)
2625 * g7 = pc we jumped here from (in)
2626 */
2627	.global	trace_tsbmiss
2628trace_tsbmiss:
2629	membar	#Sync
2630	sethi	%hi(FLUSH_ADDR), %g6
2631	flush	%g6
2632	TRACE_PTR(%g5, %g6)
2633	GET_TRACE_TICK(%g6)
2634	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2635	stna	%g2, [%g5 + TRAP_ENT_SP]%asi		! tag access
2636	stna	%g4, [%g5 + TRAP_ENT_F1]%asi		! XXX? tsb tag
2637	rdpr	%tnpc, %g6
2638	stna	%g6, [%g5 + TRAP_ENT_F2]%asi
2639	stna	%g1, [%g5 + TRAP_ENT_F3]%asi		! tsb8k pointer
2640	rdpr	%tpc, %g6
2641	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2642	TRACE_SAVE_TL_GL_REGS(%g5, %g6)
2643	rdpr	%tt, %g6
2644	or	%g6, TT_MMU_MISS, %g4
2645	stha	%g4, [%g5 + TRAP_ENT_TT]%asi
2646	mov	MMFSA_D_ADDR, %g4
2647	cmp	%g6, FAST_IMMU_MISS_TT
2648	move	%xcc, MMFSA_I_ADDR, %g4
2649	cmp	%g6, T_INSTR_MMU_MISS
2650	move	%xcc, MMFSA_I_ADDR, %g4
2651	MMU_FAULT_STATUS_AREA(%g6)
2652	ldx	[%g6 + %g4], %g6
2653	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi	! tag target
2654	cmp	%g4, MMFSA_D_ADDR
2655	move	%xcc, MMFSA_D_CTX, %g4
2656	movne	%xcc, MMFSA_I_CTX, %g4
2657	MMU_FAULT_STATUS_AREA(%g6)
2658	ldx	[%g6 + %g4], %g6
2659	stxa	%g6, [%g5 + TRAP_ENT_F4]%asi		! context ID
2660	stna	%g3, [%g5 + TRAP_ENT_TR]%asi		! tsb4m pointer
2661	TRACE_NEXT(%g5, %g4, %g6)
2662	jmp	%g7 + 4
2663	nop
2664
2665/*
2666 * g2 = tag access register (in)
2667 * g3 = ctx type (0, 1 or 2) (in) (not used)
2668 */
2669trace_dataprot:
2670	membar	#Sync
2671	sethi	%hi(FLUSH_ADDR), %g6
2672	flush	%g6
2673	TRACE_PTR(%g1, %g6)
2674	GET_TRACE_TICK(%g6)
2675	stxa	%g6, [%g1 + TRAP_ENT_TICK]%asi
2676	rdpr	%tpc, %g6
2677	stna	%g6, [%g1 + TRAP_ENT_TPC]%asi
2678	rdpr	%tstate, %g6
2679	stxa	%g6, [%g1 + TRAP_ENT_TSTATE]%asi
2680	stna	%g2, [%g1 + TRAP_ENT_SP]%asi		! tag access reg
2681	stna	%g0, [%g1 + TRAP_ENT_F1]%asi
2682	stna	%g0, [%g1 + TRAP_ENT_F2]%asi
2683	stna	%g0, [%g1 + TRAP_ENT_F3]%asi
2684	stna	%g0, [%g1 + TRAP_ENT_F4]%asi
2685	TRACE_SAVE_TL_GL_REGS(%g1, %g6)
2686	rdpr	%tt, %g6
2687	stha	%g6, [%g1 + TRAP_ENT_TT]%asi
2688	mov	MMFSA_D_CTX, %g4
2689	cmp	%g6, FAST_IMMU_MISS_TT
2690	move	%xcc, MMFSA_I_CTX, %g4
2691	cmp	%g6, T_INSTR_MMU_MISS
2692	move	%xcc, MMFSA_I_CTX, %g4
2693	MMU_FAULT_STATUS_AREA(%g6)
2694	ldx	[%g6 + %g4], %g6
2695	stxa	%g6, [%g1 + TRAP_ENT_TR]%asi	! context ID
2696	TRACE_NEXT(%g1, %g4, %g5)
2697	jmp	%g7 + 4
2698	nop
2699
2700#endif /* TRAPTRACE */
2701
2702/*
2703 * Handle watchdog reset trap. Enable the MMU using the MMU_ENABLE
2704 * HV service, which requires the return target to be specified as a VA
2705 * since we are enabling the MMU. We set the target to ptl1_panic.
2706 */
2707
2708	.type	.watchdog_trap, #function
2709.watchdog_trap:
2710	mov	1, %o0
2711	setx	ptl1_panic, %g2, %o1
2712	mov	MMU_ENABLE, %o5
2713	ta	FAST_TRAP
2714	done
2715	SET_SIZE(.watchdog_trap)
2716/*
2717 * synthesize for trap(): SFAR in %g2, SFSR in %g3
2718 */
2719	.type	.dmmu_exc_lddf_not_aligned, #function
2720.dmmu_exc_lddf_not_aligned:
2721	MMU_FAULT_STATUS_AREA(%g3)
2722	ldx	[%g3 + MMFSA_D_ADDR], %g2
2723	/* Fault type not available in MMU fault status area */
2724	mov	MMFSA_F_UNALIGN, %g1
2725	ldx	[%g3 + MMFSA_D_CTX], %g3
2726	sllx	%g3, SFSR_CTX_SHIFT, %g3
2727	btst	1, %sp
2728	bnz,pt	%xcc, .lddf_exception_not_aligned
2729	or	%g3, %g1, %g3			/* SFSR */
2730	ba,a,pt	%xcc, .mmu_exception_not_aligned
2731	SET_SIZE(.dmmu_exc_lddf_not_aligned)
2732
2733/*
2734 * synthesize for trap(): SFAR in %g2, SFSR in %g3
2735 */
2736	.type	.dmmu_exc_stdf_not_aligned, #function
2737.dmmu_exc_stdf_not_aligned:
2738	MMU_FAULT_STATUS_AREA(%g3)
2739	ldx	[%g3 + MMFSA_D_ADDR], %g2
2740	/* Fault type not available in MMU fault status area */
2741	mov	MMFSA_F_UNALIGN, %g1
2742	ldx	[%g3 + MMFSA_D_CTX], %g3
2743	sllx	%g3, SFSR_CTX_SHIFT, %g3
2744	btst	1, %sp
2745	bnz,pt	%xcc, .stdf_exception_not_aligned
2746	or	%g3, %g1, %g3			/* SFSR */
2747	ba,a,pt	%xcc, .mmu_exception_not_aligned
2748	SET_SIZE(.dmmu_exc_stdf_not_aligned)
2749
2750	.type	.dmmu_exception, #function
2751.dmmu_exception:
2752	MMU_FAULT_STATUS_AREA(%g3)
2753	ldx	[%g3 + MMFSA_D_ADDR], %g2
2754	ldx	[%g3 + MMFSA_D_TYPE], %g1
2755	ldx	[%g3 + MMFSA_D_CTX], %g4
2756	srlx	%g2, MMU_PAGESHIFT, %g2		/* align address */
2757	sllx	%g2, MMU_PAGESHIFT, %g2
2758	sllx	%g4, SFSR_CTX_SHIFT, %g3
2759	or	%g3, %g1, %g3			/* SFSR */
2760	cmp	%g4, USER_CONTEXT_TYPE
2761	movgeu	%icc, USER_CONTEXT_TYPE, %g4
2762	or	%g2, %g4, %g2			/* TAG_ACCESS */
2763	ba,pt	%xcc, .mmu_exception_end
2764	mov	T_DATA_EXCEPTION, %g1
2765	SET_SIZE(.dmmu_exception)
2766
2767/*
2768 * fast_trap_done, fast_trap_done_chk_intr:
2769 *
2770 * Due to the design of UltraSPARC pipeline, pending interrupts are not
2771 * taken immediately after a RETRY or DONE instruction which causes IE to
2772 * go from 0 to 1. Instead, the instruction at %tpc or %tnpc is allowed
2773 * to execute first before taking any interrupts. If that instruction
2774 * results in other traps, and if the corresponding trap handler runs
2775 * entirely at TL=1 with interrupts disabled, then pending interrupts
2776 * won't be taken until after yet another instruction following the %tpc
2777 * or %tnpc.
2778 *
2779 * A malicious user program can use this feature to block out interrupts
2780 * for extended durations, which can result in send_mondo_timeout kernel
2781 * panic.
2782 *
2783 * This problem is addressed by servicing any pending interrupts via
2784 * sys_trap before returning back to the user mode from a fast trap
2785 * handler. The "done" instruction within a fast trap handler, which
2786 * runs entirely at TL=1 with interrupts disabled, is replaced with the
2787 * FAST_TRAP_DONE macro, which branches control to this fast_trap_done
2788 * entry point.
2789 *
2790 * We check for any pending interrupts here and force a sys_trap to
2791 * service those interrupts, if any. To minimize overhead, pending
2792 * interrupts are checked if the %tpc happens to be at 16K boundary,
2793 * which allows a malicious program to execute at most 4K consecutive
2794 * instructions before we service any pending interrupts. If a worst
2795 * case fast trap handler takes about 2 usec, then interrupts will be
2796 * blocked for at most 8 msec, less than a clock tick.
2797 *
2798 * For the cases where we don't know if the %tpc will cross a 16K
2799 * boundary, we can't use the above optimization and always process
2800 * any pending interrupts via fast_frap_done_chk_intr entry point.
2801 *
2802 * Entry Conditions:
2803 * 	%pstate		am:0 priv:1 ie:0
2804 * 			globals are AG (not normal globals)
2805 */
2806
2807	.global	fast_trap_done, fast_trap_done_chk_intr
2808fast_trap_done:
2809	rdpr	%tpc, %g5
2810	sethi	%hi(0xffffc000), %g6	! 1's complement of 0x3fff
2811	andncc	%g5, %g6, %g0		! check lower 14 bits of %tpc
2812	bz,pn	%icc, 1f		! branch if zero (lower 32 bits only)
2813	nop
2814	done
2815
2816fast_trap_done_chk_intr:
28171:	rd	SOFTINT, %g6
2818	brnz,pn	%g6, 2f		! branch if any pending intr
2819	nop
2820	done
2821
28222:
2823	/*
2824	 * We get here if there are any pending interrupts.
2825	 * Adjust %tpc/%tnpc as we'll be resuming via "retry"
2826	 * instruction.
2827	 */
2828	rdpr	%tnpc, %g5
2829	wrpr	%g0, %g5, %tpc
2830	add	%g5, 4, %g5
2831	wrpr	%g0, %g5, %tnpc
2832
2833	/*
2834	 * Force a dummy sys_trap call so that interrupts can be serviced.
2835	 */
2836	set	fast_trap_dummy_call, %g1
2837	ba,pt	%xcc, sys_trap
2838	  mov	-1, %g4
2839
2840fast_trap_dummy_call:
2841	retl
2842	nop
2843
2844/*
2845 * Currently the brand syscall interposition code is not enabled by
2846 * default.  Instead, when a branded zone is first booted the brand
2847 * infrastructure will patch the trap table so that the syscall
2848 * entry points are redirected to syscall_wrapper32 and syscall_wrapper
2849 * for ILP32 and LP64 syscalls respectively.  this is done in
2850 * brand_plat_interposition_enable().  Note that the syscall wrappers
2851 * below do not collect any trap trace data since the syscall hot patch
2852 * points are reached after trap trace data has already been collected.
2853 */
2854#define	BRAND_CALLBACK(callback_id)					    \
2855	CPU_ADDR(%g2, %g1)		/* load CPU struct addr to %g2	*/ ;\
2856	ldn	[%g2 + CPU_THREAD], %g3	/* load thread pointer		*/ ;\
2857	ldn	[%g3 + T_PROCP], %g3	/* get proc pointer		*/ ;\
2858	ldn	[%g3 + P_BRAND], %g3	/* get brand pointer		*/ ;\
2859	brz	%g3, 1f			/* No brand?  No callback. 	*/ ;\
2860	nop 								   ;\
2861	ldn	[%g3 + B_MACHOPS], %g3	/* get machops list		*/ ;\
2862	ldn	[%g3 + (callback_id << 3)], %g3 			   ;\
2863	brz	%g3, 1f							   ;\
2864	/*								    \
2865	 * This isn't pretty.  We want a low-latency way for the callback   \
2866	 * routine to decline to do anything.  We just pass in an address   \
2867	 * the routine can directly jmp back to, pretending that nothing    \
2868	 * has happened.						    \
2869	 * 								    \
2870	 * %g1: return address (where the brand handler jumps back to)	    \
2871	 * %g2: address of CPU structure				    \
2872	 * %g3: address of brand handler (where we will jump to)	    \
2873	 */								    \
2874	mov	%pc, %g1						   ;\
2875	add	%g1, 16, %g1						   ;\
2876	jmp	%g3							   ;\
2877	nop								   ;\
28781:
2879
2880	ENTRY_NP(syscall_wrapper32)
2881	BRAND_CALLBACK(BRAND_CB_SYSCALL32)
2882	SYSCALL_NOTT(syscall_trap32)
2883	SET_SIZE(syscall_wrapper32)
2884
2885	ENTRY_NP(syscall_wrapper)
2886	BRAND_CALLBACK(BRAND_CB_SYSCALL)
2887	SYSCALL_NOTT(syscall_trap)
2888	SET_SIZE(syscall_wrapper)
2889
2890#endif	/* lint */
2891