xref: /titanic_41/usr/src/uts/sun4u/ml/trap_table.s (revision 8700009e2cc8cb186241e1fdd74973da1121ee4c)
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 * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
23 * Use is subject to license terms.
24 */
25
26#pragma ident	"%Z%%M%	%I%	%E% SMI"
27
28#if !defined(lint)
29#include "assym.h"
30#endif /* !lint */
31#include <sys/asm_linkage.h>
32#include <sys/privregs.h>
33#include <sys/sun4asi.h>
34#include <sys/spitregs.h>
35#include <sys/cheetahregs.h>
36#include <sys/machtrap.h>
37#include <sys/machthread.h>
38#include <sys/machbrand.h>
39#include <sys/pcb.h>
40#include <sys/pte.h>
41#include <sys/mmu.h>
42#include <sys/machpcb.h>
43#include <sys/async.h>
44#include <sys/intreg.h>
45#include <sys/scb.h>
46#include <sys/psr_compat.h>
47#include <sys/syscall.h>
48#include <sys/machparam.h>
49#include <sys/traptrace.h>
50#include <vm/hat_sfmmu.h>
51#include <sys/archsystm.h>
52#include <sys/utrap.h>
53#include <sys/clock.h>
54#include <sys/intr.h>
55#include <sys/fpu/fpu_simulator.h>
56#include <vm/seg_spt.h>
57
58/*
59 * WARNING: If you add a fast trap handler which can be invoked by a
60 * non-privileged user, you may have to use the FAST_TRAP_DONE macro
61 * instead of "done" instruction to return back to the user mode. See
62 * comments for the "fast_trap_done" entry point for more information.
63 *
64 * An alternate FAST_TRAP_DONE_CHK_INTR macro should be used for the
65 * cases where you always want to process any pending interrupts before
66 * returning back to the user mode.
67 */
68#define	FAST_TRAP_DONE		\
69	ba,a	fast_trap_done
70
71#define	FAST_TRAP_DONE_CHK_INTR	\
72	ba,a	fast_trap_done_chk_intr
73
74/*
75 * SPARC V9 Trap Table
76 *
77 * Most of the trap handlers are made from common building
78 * blocks, and some are instantiated multiple times within
79 * the trap table. So, I build a bunch of macros, then
80 * populate the table using only the macros.
81 *
82 * Many macros branch to sys_trap.  Its calling convention is:
83 *	%g1		kernel trap handler
84 *	%g2, %g3	args for above
85 *	%g4		desire %pil
86 */
87
88#ifdef	TRAPTRACE
89
90/*
91 * Tracing macro. Adds two instructions if TRAPTRACE is defined.
92 */
93#define	TT_TRACE(label)		\
94	ba	label		;\
95	rd	%pc, %g7
96#define	TT_TRACE_INS	2
97
98#define	TT_TRACE_L(label)	\
99	ba	label		;\
100	rd	%pc, %l4	;\
101	clr	%l4
102#define	TT_TRACE_L_INS	3
103
104#else
105
106#define	TT_TRACE(label)
107#define	TT_TRACE_INS	0
108
109#define	TT_TRACE_L(label)
110#define	TT_TRACE_L_INS	0
111
112#endif
113
114/*
115 * This first set are funneled to trap() with %tt as the type.
116 * Trap will then either panic or send the user a signal.
117 */
118/*
119 * NOT is used for traps that just shouldn't happen.
120 * It comes in both single and quadruple flavors.
121 */
122#if !defined(lint)
123	.global	trap
124#endif /* !lint */
125#define	NOT			\
126	TT_TRACE(trace_gen)	;\
127	set	trap, %g1	;\
128	rdpr	%tt, %g3	;\
129	ba,pt	%xcc, sys_trap	;\
130	sub	%g0, 1, %g4	;\
131	.align	32
132#define	NOT4	NOT; NOT; NOT; NOT
133/*
134 * RED is for traps that use the red mode handler.
135 * We should never see these either.
136 */
137#define	RED	NOT
138/*
139 * BAD is used for trap vectors we don't have a kernel
140 * handler for.
141 * It also comes in single and quadruple versions.
142 */
143#define	BAD	NOT
144#define	BAD4	NOT4
145
146#define	DONE			\
147	done;			\
148	.align	32
149
150/*
151 * TRAP vectors to the trap() function.
152 * It's main use is for user errors.
153 */
154#if !defined(lint)
155	.global	trap
156#endif /* !lint */
157#define	TRAP(arg)		\
158	TT_TRACE(trace_gen)	;\
159	set	trap, %g1	;\
160	mov	arg, %g3	;\
161	ba,pt	%xcc, sys_trap	;\
162	sub	%g0, 1, %g4	;\
163	.align	32
164
165/*
166 * SYSCALL is used for unsupported syscall interfaces (with 'which'
167 * set to 'nosys') and legacy support of old SunOS 4.x syscalls (with
168 * 'which' set to 'syscall_trap32').
169 *
170 * The SYSCALL_TRAP* macros are used for syscall entry points.
171 * SYSCALL_TRAP is used to support LP64 syscalls and SYSCALL_TRAP32
172 * is used to support ILP32.  Each macro can only be used once
173 * since they each define a symbol.  The symbols are used as hot patch
174 * points by the brand infrastructure to dynamically enable and disable
175 * brand syscall interposition.  See the comments around BRAND_CALLBACK
176 * and brand_plat_interposition_enable() for more information.
177 */
178#define	SYSCALL_NOTT(which)		\
179	set	(which), %g1		;\
180	ba,pt	%xcc, sys_trap		;\
181	sub	%g0, 1, %g4		;\
182	.align	32
183
184#define	SYSCALL(which)			\
185	TT_TRACE(trace_gen)		;\
186	SYSCALL_NOTT(which)
187
188#define	SYSCALL_TRAP32				\
189	TT_TRACE(trace_gen)			;\
190	ALTENTRY(syscall_trap32_patch_point)	\
191	SYSCALL_NOTT(syscall_trap32)
192
193#define	SYSCALL_TRAP				\
194	TT_TRACE(trace_gen)			;\
195	ALTENTRY(syscall_trap_patch_point)	\
196	SYSCALL_NOTT(syscall_trap)
197
198#define	FLUSHW(h_name)			\
199	.global h_name			;\
200h_name:					;\
201	set	trap, %g1		;\
202	mov	T_FLUSHW, %g3		;\
203	sub	%g0, 1, %g4		;\
204	save				;\
205	flushw				;\
206	restore				;\
207	FAST_TRAP_DONE			;\
208	.align	32
209
210/*
211 * GOTO just jumps to a label.
212 * It's used for things that can be fixed without going thru sys_trap.
213 */
214#define	GOTO(label)		\
215	.global	label		;\
216	ba,a	label		;\
217	.empty			;\
218	.align	32
219
220/*
221 * GOTO_TT just jumps to a label.
222 * correctable ECC error traps at  level 0 and 1 will use this macro.
223 * It's used for things that can be fixed without going thru sys_trap.
224 */
225#define	GOTO_TT(label, ttlabel)		\
226	.global	label		;\
227	TT_TRACE(ttlabel)	;\
228	ba,a	label		;\
229	.empty			;\
230	.align	32
231
232/*
233 * Privileged traps
234 * Takes breakpoint if privileged, calls trap() if not.
235 */
236#define	PRIV(label)			\
237	rdpr	%tstate, %g1		;\
238	btst	TSTATE_PRIV, %g1	;\
239	bnz	label			;\
240	rdpr	%tt, %g3		;\
241	set	trap, %g1		;\
242	ba,pt	%xcc, sys_trap		;\
243	sub	%g0, 1, %g4		;\
244	.align	32
245
246
247/*
248 * DTrace traps.
249 */
250#define	DTRACE_PID			\
251	.global dtrace_pid_probe				;\
252	set	dtrace_pid_probe, %g1				;\
253	ba,pt	%xcc, user_trap					;\
254	sub	%g0, 1, %g4					;\
255	.align	32
256
257#define	DTRACE_RETURN			\
258	.global dtrace_return_probe				;\
259	set	dtrace_return_probe, %g1			;\
260	ba,pt	%xcc, user_trap					;\
261	sub	%g0, 1, %g4					;\
262	.align	32
263
264/*
265 * REGISTER WINDOW MANAGEMENT MACROS
266 */
267
268/*
269 * various convenient units of padding
270 */
271#define	SKIP(n)	.skip 4*(n)
272
273/*
274 * CLEAN_WINDOW is the simple handler for cleaning a register window.
275 */
276#define	CLEAN_WINDOW						\
277	TT_TRACE_L(trace_win)					;\
278	rdpr %cleanwin, %l0; inc %l0; wrpr %l0, %cleanwin	;\
279	clr %l0; clr %l1; clr %l2; clr %l3			;\
280	clr %l4; clr %l5; clr %l6; clr %l7			;\
281	clr %o0; clr %o1; clr %o2; clr %o3			;\
282	clr %o4; clr %o5; clr %o6; clr %o7			;\
283	retry; .align 128
284
285#if !defined(lint)
286
287/*
288 * If we get an unresolved tlb miss while in a window handler, the fault
289 * handler will resume execution at the last instruction of the window
290 * hander, instead of delivering the fault to the kernel.  Spill handlers
291 * use this to spill windows into the wbuf.
292 *
293 * The mixed handler works by checking %sp, and branching to the correct
294 * handler.  This is done by branching back to label 1: for 32b frames,
295 * or label 2: for 64b frames; which implies the handler order is: 32b,
296 * 64b, mixed.  The 1: and 2: labels are offset into the routines to
297 * allow the branchs' delay slots to contain useful instructions.
298 */
299
300/*
301 * SPILL_32bit spills a 32-bit-wide kernel register window.  It
302 * assumes that the kernel context and the nucleus context are the
303 * same.  The stack pointer is required to be eight-byte aligned even
304 * though this code only needs it to be four-byte aligned.
305 */
306#define	SPILL_32bit(tail)					\
307	srl	%sp, 0, %sp					;\
3081:	st	%l0, [%sp + 0]					;\
309	st	%l1, [%sp + 4]					;\
310	st	%l2, [%sp + 8]					;\
311	st	%l3, [%sp + 12]					;\
312	st	%l4, [%sp + 16]					;\
313	st	%l5, [%sp + 20]					;\
314	st	%l6, [%sp + 24]					;\
315	st	%l7, [%sp + 28]					;\
316	st	%i0, [%sp + 32]					;\
317	st	%i1, [%sp + 36]					;\
318	st	%i2, [%sp + 40]					;\
319	st	%i3, [%sp + 44]					;\
320	st	%i4, [%sp + 48]					;\
321	st	%i5, [%sp + 52]					;\
322	st	%i6, [%sp + 56]					;\
323	st	%i7, [%sp + 60]					;\
324	TT_TRACE_L(trace_win)					;\
325	saved							;\
326	retry							;\
327	SKIP(31-19-TT_TRACE_L_INS)				;\
328	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
329	.empty
330
331/*
332 * SPILL_32bit_asi spills a 32-bit-wide register window into a 32-bit
333 * wide address space via the designated asi.  It is used to spill
334 * non-kernel windows.  The stack pointer is required to be eight-byte
335 * aligned even though this code only needs it to be four-byte
336 * aligned.
337 */
338#define	SPILL_32bit_asi(asi_num, tail)				\
339	srl	%sp, 0, %sp					;\
3401:	sta	%l0, [%sp + %g0]asi_num				;\
341	mov	4, %g1						;\
342	sta	%l1, [%sp + %g1]asi_num				;\
343	mov	8, %g2						;\
344	sta	%l2, [%sp + %g2]asi_num				;\
345	mov	12, %g3						;\
346	sta	%l3, [%sp + %g3]asi_num				;\
347	add	%sp, 16, %g4					;\
348	sta	%l4, [%g4 + %g0]asi_num				;\
349	sta	%l5, [%g4 + %g1]asi_num				;\
350	sta	%l6, [%g4 + %g2]asi_num				;\
351	sta	%l7, [%g4 + %g3]asi_num				;\
352	add	%g4, 16, %g4					;\
353	sta	%i0, [%g4 + %g0]asi_num				;\
354	sta	%i1, [%g4 + %g1]asi_num				;\
355	sta	%i2, [%g4 + %g2]asi_num				;\
356	sta	%i3, [%g4 + %g3]asi_num				;\
357	add	%g4, 16, %g4					;\
358	sta	%i4, [%g4 + %g0]asi_num				;\
359	sta	%i5, [%g4 + %g1]asi_num				;\
360	sta	%i6, [%g4 + %g2]asi_num				;\
361	sta	%i7, [%g4 + %g3]asi_num				;\
362	TT_TRACE_L(trace_win)					;\
363	saved							;\
364	retry							;\
365	SKIP(31-25-TT_TRACE_L_INS)				;\
366	ba,a,pt %xcc, fault_32bit_/**/tail			;\
367	.empty
368
369/*
370 * SPILL_32bit_tt1 spills a 32-bit-wide register window into a 32-bit
371 * wide address space via the designated asi.  It is used to spill
372 * windows at tl>1 where performance isn't the primary concern and
373 * where we don't want to use unnecessary registers.  The stack
374 * pointer is required to be eight-byte aligned even though this code
375 * only needs it to be four-byte aligned.
376 */
377#define	SPILL_32bit_tt1(asi_num, tail)				\
378	mov	asi_num, %asi					;\
3791:	srl	%sp, 0, %sp					;\
380	sta	%l0, [%sp + 0]%asi				;\
381	sta	%l1, [%sp + 4]%asi				;\
382	sta	%l2, [%sp + 8]%asi				;\
383	sta	%l3, [%sp + 12]%asi				;\
384	sta	%l4, [%sp + 16]%asi				;\
385	sta	%l5, [%sp + 20]%asi				;\
386	sta	%l6, [%sp + 24]%asi				;\
387	sta	%l7, [%sp + 28]%asi				;\
388	sta	%i0, [%sp + 32]%asi				;\
389	sta	%i1, [%sp + 36]%asi				;\
390	sta	%i2, [%sp + 40]%asi				;\
391	sta	%i3, [%sp + 44]%asi				;\
392	sta	%i4, [%sp + 48]%asi				;\
393	sta	%i5, [%sp + 52]%asi				;\
394	sta	%i6, [%sp + 56]%asi				;\
395	sta	%i7, [%sp + 60]%asi				;\
396	TT_TRACE_L(trace_win)					;\
397	saved							;\
398	retry							;\
399	SKIP(31-20-TT_TRACE_L_INS)				;\
400	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
401	.empty
402
403
404/*
405 * FILL_32bit fills a 32-bit-wide kernel register window.  It assumes
406 * that the kernel context and the nucleus context are the same.  The
407 * stack pointer is required to be eight-byte aligned even though this
408 * code only needs it to be four-byte aligned.
409 */
410#define	FILL_32bit(tail)					\
411	srl	%sp, 0, %sp					;\
4121:	TT_TRACE_L(trace_win)					;\
413	ld	[%sp + 0], %l0					;\
414	ld	[%sp + 4], %l1					;\
415	ld	[%sp + 8], %l2					;\
416	ld	[%sp + 12], %l3					;\
417	ld	[%sp + 16], %l4					;\
418	ld	[%sp + 20], %l5					;\
419	ld	[%sp + 24], %l6					;\
420	ld	[%sp + 28], %l7					;\
421	ld	[%sp + 32], %i0					;\
422	ld	[%sp + 36], %i1					;\
423	ld	[%sp + 40], %i2					;\
424	ld	[%sp + 44], %i3					;\
425	ld	[%sp + 48], %i4					;\
426	ld	[%sp + 52], %i5					;\
427	ld	[%sp + 56], %i6					;\
428	ld	[%sp + 60], %i7					;\
429	restored						;\
430	retry							;\
431	SKIP(31-19-TT_TRACE_L_INS)				;\
432	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
433	.empty
434
435/*
436 * FILL_32bit_asi fills a 32-bit-wide register window from a 32-bit
437 * wide address space via the designated asi.  It is used to fill
438 * non-kernel windows.  The stack pointer is required to be eight-byte
439 * aligned even though this code only needs it to be four-byte
440 * aligned.
441 */
442#define	FILL_32bit_asi(asi_num, tail)				\
443	srl	%sp, 0, %sp					;\
4441:	TT_TRACE_L(trace_win)					;\
445	mov	4, %g1						;\
446	lda	[%sp + %g0]asi_num, %l0				;\
447	mov	8, %g2						;\
448	lda	[%sp + %g1]asi_num, %l1				;\
449	mov	12, %g3						;\
450	lda	[%sp + %g2]asi_num, %l2				;\
451	lda	[%sp + %g3]asi_num, %l3				;\
452	add	%sp, 16, %g4					;\
453	lda	[%g4 + %g0]asi_num, %l4				;\
454	lda	[%g4 + %g1]asi_num, %l5				;\
455	lda	[%g4 + %g2]asi_num, %l6				;\
456	lda	[%g4 + %g3]asi_num, %l7				;\
457	add	%g4, 16, %g4					;\
458	lda	[%g4 + %g0]asi_num, %i0				;\
459	lda	[%g4 + %g1]asi_num, %i1				;\
460	lda	[%g4 + %g2]asi_num, %i2				;\
461	lda	[%g4 + %g3]asi_num, %i3				;\
462	add	%g4, 16, %g4					;\
463	lda	[%g4 + %g0]asi_num, %i4				;\
464	lda	[%g4 + %g1]asi_num, %i5				;\
465	lda	[%g4 + %g2]asi_num, %i6				;\
466	lda	[%g4 + %g3]asi_num, %i7				;\
467	restored						;\
468	retry							;\
469	SKIP(31-25-TT_TRACE_L_INS)				;\
470	ba,a,pt %xcc, fault_32bit_/**/tail			;\
471	.empty
472
473/*
474 * FILL_32bit_tt1 fills a 32-bit-wide register window from a 32-bit
475 * wide address space via the designated asi.  It is used to fill
476 * windows at tl>1 where performance isn't the primary concern and
477 * where we don't want to use unnecessary registers.  The stack
478 * pointer is required to be eight-byte aligned even though this code
479 * only needs it to be four-byte aligned.
480 */
481#define	FILL_32bit_tt1(asi_num, tail)				\
482	mov	asi_num, %asi					;\
4831:	srl	%sp, 0, %sp					;\
484	TT_TRACE_L(trace_win)					;\
485	lda	[%sp + 0]%asi, %l0				;\
486	lda	[%sp + 4]%asi, %l1				;\
487	lda	[%sp + 8]%asi, %l2				;\
488	lda	[%sp + 12]%asi, %l3				;\
489	lda	[%sp + 16]%asi, %l4				;\
490	lda	[%sp + 20]%asi, %l5				;\
491	lda	[%sp + 24]%asi, %l6				;\
492	lda	[%sp + 28]%asi, %l7				;\
493	lda	[%sp + 32]%asi, %i0				;\
494	lda	[%sp + 36]%asi, %i1				;\
495	lda	[%sp + 40]%asi, %i2				;\
496	lda	[%sp + 44]%asi, %i3				;\
497	lda	[%sp + 48]%asi, %i4				;\
498	lda	[%sp + 52]%asi, %i5				;\
499	lda	[%sp + 56]%asi, %i6				;\
500	lda	[%sp + 60]%asi, %i7				;\
501	restored						;\
502	retry							;\
503	SKIP(31-20-TT_TRACE_L_INS)				;\
504	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
505	.empty
506
507
508/*
509 * SPILL_64bit spills a 64-bit-wide kernel register window.  It
510 * assumes that the kernel context and the nucleus context are the
511 * same.  The stack pointer is required to be eight-byte aligned.
512 */
513#define	SPILL_64bit(tail)					\
5142:	stx	%l0, [%sp + V9BIAS64 + 0]			;\
515	stx	%l1, [%sp + V9BIAS64 + 8]			;\
516	stx	%l2, [%sp + V9BIAS64 + 16]			;\
517	stx	%l3, [%sp + V9BIAS64 + 24]			;\
518	stx	%l4, [%sp + V9BIAS64 + 32]			;\
519	stx	%l5, [%sp + V9BIAS64 + 40]			;\
520	stx	%l6, [%sp + V9BIAS64 + 48]			;\
521	stx	%l7, [%sp + V9BIAS64 + 56]			;\
522	stx	%i0, [%sp + V9BIAS64 + 64]			;\
523	stx	%i1, [%sp + V9BIAS64 + 72]			;\
524	stx	%i2, [%sp + V9BIAS64 + 80]			;\
525	stx	%i3, [%sp + V9BIAS64 + 88]			;\
526	stx	%i4, [%sp + V9BIAS64 + 96]			;\
527	stx	%i5, [%sp + V9BIAS64 + 104]			;\
528	stx	%i6, [%sp + V9BIAS64 + 112]			;\
529	stx	%i7, [%sp + V9BIAS64 + 120]			;\
530	TT_TRACE_L(trace_win)					;\
531	saved							;\
532	retry							;\
533	SKIP(31-18-TT_TRACE_L_INS)				;\
534	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
535	.empty
536
537/*
538 * SPILL_64bit_asi spills a 64-bit-wide register window into a 64-bit
539 * wide address space via the designated asi.  It is used to spill
540 * non-kernel windows.  The stack pointer is required to be eight-byte
541 * aligned.
542 */
543#define	SPILL_64bit_asi(asi_num, tail)				\
544	mov	0 + V9BIAS64, %g1				;\
5452:	stxa	%l0, [%sp + %g1]asi_num				;\
546	mov	8 + V9BIAS64, %g2				;\
547	stxa	%l1, [%sp + %g2]asi_num				;\
548	mov	16 + V9BIAS64, %g3				;\
549	stxa	%l2, [%sp + %g3]asi_num				;\
550	mov	24 + V9BIAS64, %g4				;\
551	stxa	%l3, [%sp + %g4]asi_num				;\
552	add	%sp, 32, %g5					;\
553	stxa	%l4, [%g5 + %g1]asi_num				;\
554	stxa	%l5, [%g5 + %g2]asi_num				;\
555	stxa	%l6, [%g5 + %g3]asi_num				;\
556	stxa	%l7, [%g5 + %g4]asi_num				;\
557	add	%g5, 32, %g5					;\
558	stxa	%i0, [%g5 + %g1]asi_num				;\
559	stxa	%i1, [%g5 + %g2]asi_num				;\
560	stxa	%i2, [%g5 + %g3]asi_num				;\
561	stxa	%i3, [%g5 + %g4]asi_num				;\
562	add	%g5, 32, %g5					;\
563	stxa	%i4, [%g5 + %g1]asi_num				;\
564	stxa	%i5, [%g5 + %g2]asi_num				;\
565	stxa	%i6, [%g5 + %g3]asi_num				;\
566	stxa	%i7, [%g5 + %g4]asi_num				;\
567	TT_TRACE_L(trace_win)					;\
568	saved							;\
569	retry							;\
570	SKIP(31-25-TT_TRACE_L_INS)				;\
571	ba,a,pt %xcc, fault_64bit_/**/tail			;\
572	.empty
573
574/*
575 * SPILL_64bit_tt1 spills a 64-bit-wide register window into a 64-bit
576 * wide address space via the designated asi.  It is used to spill
577 * windows at tl>1 where performance isn't the primary concern and
578 * where we don't want to use unnecessary registers.  The stack
579 * pointer is required to be eight-byte aligned.
580 */
581#define	SPILL_64bit_tt1(asi_num, tail)				\
582	mov	asi_num, %asi					;\
5832:	stxa	%l0, [%sp + V9BIAS64 + 0]%asi			;\
584	stxa	%l1, [%sp + V9BIAS64 + 8]%asi			;\
585	stxa	%l2, [%sp + V9BIAS64 + 16]%asi			;\
586	stxa	%l3, [%sp + V9BIAS64 + 24]%asi			;\
587	stxa	%l4, [%sp + V9BIAS64 + 32]%asi			;\
588	stxa	%l5, [%sp + V9BIAS64 + 40]%asi			;\
589	stxa	%l6, [%sp + V9BIAS64 + 48]%asi			;\
590	stxa	%l7, [%sp + V9BIAS64 + 56]%asi			;\
591	stxa	%i0, [%sp + V9BIAS64 + 64]%asi			;\
592	stxa	%i1, [%sp + V9BIAS64 + 72]%asi			;\
593	stxa	%i2, [%sp + V9BIAS64 + 80]%asi			;\
594	stxa	%i3, [%sp + V9BIAS64 + 88]%asi			;\
595	stxa	%i4, [%sp + V9BIAS64 + 96]%asi			;\
596	stxa	%i5, [%sp + V9BIAS64 + 104]%asi			;\
597	stxa	%i6, [%sp + V9BIAS64 + 112]%asi			;\
598	stxa	%i7, [%sp + V9BIAS64 + 120]%asi			;\
599	TT_TRACE_L(trace_win)					;\
600	saved							;\
601	retry							;\
602	SKIP(31-19-TT_TRACE_L_INS)				;\
603	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
604	.empty
605
606
607/*
608 * FILL_64bit fills a 64-bit-wide kernel register window.  It assumes
609 * that the kernel context and the nucleus context are the same.  The
610 * stack pointer is required to be eight-byte aligned.
611 */
612#define	FILL_64bit(tail)					\
6132:	TT_TRACE_L(trace_win)					;\
614	ldx	[%sp + V9BIAS64 + 0], %l0			;\
615	ldx	[%sp + V9BIAS64 + 8], %l1			;\
616	ldx	[%sp + V9BIAS64 + 16], %l2			;\
617	ldx	[%sp + V9BIAS64 + 24], %l3			;\
618	ldx	[%sp + V9BIAS64 + 32], %l4			;\
619	ldx	[%sp + V9BIAS64 + 40], %l5			;\
620	ldx	[%sp + V9BIAS64 + 48], %l6			;\
621	ldx	[%sp + V9BIAS64 + 56], %l7			;\
622	ldx	[%sp + V9BIAS64 + 64], %i0			;\
623	ldx	[%sp + V9BIAS64 + 72], %i1			;\
624	ldx	[%sp + V9BIAS64 + 80], %i2			;\
625	ldx	[%sp + V9BIAS64 + 88], %i3			;\
626	ldx	[%sp + V9BIAS64 + 96], %i4			;\
627	ldx	[%sp + V9BIAS64 + 104], %i5			;\
628	ldx	[%sp + V9BIAS64 + 112], %i6			;\
629	ldx	[%sp + V9BIAS64 + 120], %i7			;\
630	restored						;\
631	retry							;\
632	SKIP(31-18-TT_TRACE_L_INS)				;\
633	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
634	.empty
635
636/*
637 * FILL_64bit_asi fills a 64-bit-wide register window from a 64-bit
638 * wide address space via the designated asi.  It is used to fill
639 * non-kernel windows.  The stack pointer is required to be eight-byte
640 * aligned.
641 */
642#define	FILL_64bit_asi(asi_num, tail)				\
643	mov	V9BIAS64 + 0, %g1				;\
6442:	TT_TRACE_L(trace_win)					;\
645	ldxa	[%sp + %g1]asi_num, %l0				;\
646	mov	V9BIAS64 + 8, %g2				;\
647	ldxa	[%sp + %g2]asi_num, %l1				;\
648	mov	V9BIAS64 + 16, %g3				;\
649	ldxa	[%sp + %g3]asi_num, %l2				;\
650	mov	V9BIAS64 + 24, %g4				;\
651	ldxa	[%sp + %g4]asi_num, %l3				;\
652	add	%sp, 32, %g5					;\
653	ldxa	[%g5 + %g1]asi_num, %l4				;\
654	ldxa	[%g5 + %g2]asi_num, %l5				;\
655	ldxa	[%g5 + %g3]asi_num, %l6				;\
656	ldxa	[%g5 + %g4]asi_num, %l7				;\
657	add	%g5, 32, %g5					;\
658	ldxa	[%g5 + %g1]asi_num, %i0				;\
659	ldxa	[%g5 + %g2]asi_num, %i1				;\
660	ldxa	[%g5 + %g3]asi_num, %i2				;\
661	ldxa	[%g5 + %g4]asi_num, %i3				;\
662	add	%g5, 32, %g5					;\
663	ldxa	[%g5 + %g1]asi_num, %i4				;\
664	ldxa	[%g5 + %g2]asi_num, %i5				;\
665	ldxa	[%g5 + %g3]asi_num, %i6				;\
666	ldxa	[%g5 + %g4]asi_num, %i7				;\
667	restored						;\
668	retry							;\
669	SKIP(31-25-TT_TRACE_L_INS)				;\
670	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
671	.empty
672
673/*
674 * FILL_64bit_tt1 fills a 64-bit-wide register window from a 64-bit
675 * wide address space via the designated asi.  It is used to fill
676 * windows at tl>1 where performance isn't the primary concern and
677 * where we don't want to use unnecessary registers.  The stack
678 * pointer is required to be eight-byte aligned.
679 */
680#define	FILL_64bit_tt1(asi_num, tail)				\
681	mov	asi_num, %asi					;\
682	TT_TRACE_L(trace_win)					;\
683	ldxa	[%sp + V9BIAS64 + 0]%asi, %l0			;\
684	ldxa	[%sp + V9BIAS64 + 8]%asi, %l1			;\
685	ldxa	[%sp + V9BIAS64 + 16]%asi, %l2			;\
686	ldxa	[%sp + V9BIAS64 + 24]%asi, %l3			;\
687	ldxa	[%sp + V9BIAS64 + 32]%asi, %l4			;\
688	ldxa	[%sp + V9BIAS64 + 40]%asi, %l5			;\
689	ldxa	[%sp + V9BIAS64 + 48]%asi, %l6			;\
690	ldxa	[%sp + V9BIAS64 + 56]%asi, %l7			;\
691	ldxa	[%sp + V9BIAS64 + 64]%asi, %i0			;\
692	ldxa	[%sp + V9BIAS64 + 72]%asi, %i1			;\
693	ldxa	[%sp + V9BIAS64 + 80]%asi, %i2			;\
694	ldxa	[%sp + V9BIAS64 + 88]%asi, %i3			;\
695	ldxa	[%sp + V9BIAS64 + 96]%asi, %i4			;\
696	ldxa	[%sp + V9BIAS64 + 104]%asi, %i5			;\
697	ldxa	[%sp + V9BIAS64 + 112]%asi, %i6			;\
698	ldxa	[%sp + V9BIAS64 + 120]%asi, %i7			;\
699	restored						;\
700	retry							;\
701	SKIP(31-19-TT_TRACE_L_INS)				;\
702	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
703	.empty
704
705#endif /* !lint */
706
707/*
708 * SPILL_mixed spills either size window, depending on
709 * whether %sp is even or odd, to a 32-bit address space.
710 * This may only be used in conjunction with SPILL_32bit/
711 * SPILL_64bit. New versions of SPILL_mixed_{tt1,asi} would be
712 * needed for use with SPILL_{32,64}bit_{tt1,asi}.  Particular
713 * attention should be paid to the instructions that belong
714 * in the delay slots of the branches depending on the type
715 * of spill handler being branched to.
716 * Clear upper 32 bits of %sp if it is odd.
717 * We won't need to clear them in 64 bit kernel.
718 */
719#define	SPILL_mixed						\
720	btst	1, %sp						;\
721	bz,a,pt	%xcc, 1b					;\
722	srl	%sp, 0, %sp					;\
723	ba,pt	%xcc, 2b					;\
724	nop							;\
725	.align	128
726
727/*
728 * FILL_mixed(ASI) fills either size window, depending on
729 * whether %sp is even or odd, from a 32-bit address space.
730 * This may only be used in conjunction with FILL_32bit/
731 * FILL_64bit. New versions of FILL_mixed_{tt1,asi} would be
732 * needed for use with FILL_{32,64}bit_{tt1,asi}. Particular
733 * attention should be paid to the instructions that belong
734 * in the delay slots of the branches depending on the type
735 * of fill handler being branched to.
736 * Clear upper 32 bits of %sp if it is odd.
737 * We won't need to clear them in 64 bit kernel.
738 */
739#define	FILL_mixed						\
740	btst	1, %sp						;\
741	bz,a,pt	%xcc, 1b					;\
742	srl	%sp, 0, %sp					;\
743	ba,pt	%xcc, 2b					;\
744	nop							;\
745	.align	128
746
747
748/*
749 * SPILL_32clean/SPILL_64clean spill 32-bit and 64-bit register windows,
750 * respectively, into the address space via the designated asi.  The
751 * unbiased stack pointer is required to be eight-byte aligned (even for
752 * the 32-bit case even though this code does not require such strict
753 * alignment).
754 *
755 * With SPARC v9 the spill trap takes precedence over the cleanwin trap
756 * so when cansave == 0, canrestore == 6, and cleanwin == 6 the next save
757 * will cause cwp + 2 to be spilled but will not clean cwp + 1.  That
758 * window may contain kernel data so in user_rtt we set wstate to call
759 * these spill handlers on the first user spill trap.  These handler then
760 * spill the appropriate window but also back up a window and clean the
761 * window that didn't get a cleanwin trap.
762 */
763#define	SPILL_32clean(asi_num, tail)				\
764	srl	%sp, 0, %sp					;\
765	sta	%l0, [%sp + %g0]asi_num				;\
766	mov	4, %g1						;\
767	sta	%l1, [%sp + %g1]asi_num				;\
768	mov	8, %g2						;\
769	sta	%l2, [%sp + %g2]asi_num				;\
770	mov	12, %g3						;\
771	sta	%l3, [%sp + %g3]asi_num				;\
772	add	%sp, 16, %g4					;\
773	sta	%l4, [%g4 + %g0]asi_num				;\
774	sta	%l5, [%g4 + %g1]asi_num				;\
775	sta	%l6, [%g4 + %g2]asi_num				;\
776	sta	%l7, [%g4 + %g3]asi_num				;\
777	add	%g4, 16, %g4					;\
778	sta	%i0, [%g4 + %g0]asi_num				;\
779	sta	%i1, [%g4 + %g1]asi_num				;\
780	sta	%i2, [%g4 + %g2]asi_num				;\
781	sta	%i3, [%g4 + %g3]asi_num				;\
782	add	%g4, 16, %g4					;\
783	sta	%i4, [%g4 + %g0]asi_num				;\
784	sta	%i5, [%g4 + %g1]asi_num				;\
785	sta	%i6, [%g4 + %g2]asi_num				;\
786	sta	%i7, [%g4 + %g3]asi_num				;\
787	TT_TRACE_L(trace_win)					;\
788	b	.spill_clean					;\
789	  mov	WSTATE_USER32, %g7				;\
790	SKIP(31-25-TT_TRACE_L_INS)				;\
791	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
792	.empty
793
794#define	SPILL_64clean(asi_num, tail)				\
795	mov	0 + V9BIAS64, %g1				;\
796	stxa	%l0, [%sp + %g1]asi_num				;\
797	mov	8 + V9BIAS64, %g2				;\
798	stxa	%l1, [%sp + %g2]asi_num				;\
799	mov	16 + V9BIAS64, %g3				;\
800	stxa	%l2, [%sp + %g3]asi_num				;\
801	mov	24 + V9BIAS64, %g4				;\
802	stxa	%l3, [%sp + %g4]asi_num				;\
803	add	%sp, 32, %g5					;\
804	stxa	%l4, [%g5 + %g1]asi_num				;\
805	stxa	%l5, [%g5 + %g2]asi_num				;\
806	stxa	%l6, [%g5 + %g3]asi_num				;\
807	stxa	%l7, [%g5 + %g4]asi_num				;\
808	add	%g5, 32, %g5					;\
809	stxa	%i0, [%g5 + %g1]asi_num				;\
810	stxa	%i1, [%g5 + %g2]asi_num				;\
811	stxa	%i2, [%g5 + %g3]asi_num				;\
812	stxa	%i3, [%g5 + %g4]asi_num				;\
813	add	%g5, 32, %g5					;\
814	stxa	%i4, [%g5 + %g1]asi_num				;\
815	stxa	%i5, [%g5 + %g2]asi_num				;\
816	stxa	%i6, [%g5 + %g3]asi_num				;\
817	stxa	%i7, [%g5 + %g4]asi_num				;\
818	TT_TRACE_L(trace_win)					;\
819	b	.spill_clean					;\
820	  mov	WSTATE_USER64, %g7				;\
821	SKIP(31-25-TT_TRACE_L_INS)				;\
822	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
823	.empty
824
825
826/*
827 * Floating point disabled.
828 */
829#define	FP_DISABLED_TRAP		\
830	TT_TRACE(trace_gen)		;\
831	ba,pt	%xcc,.fp_disabled	;\
832	nop				;\
833	.align	32
834
835/*
836 * Floating point exceptions.
837 */
838#define	FP_IEEE_TRAP			\
839	TT_TRACE(trace_gen)		;\
840	ba,pt	%xcc,.fp_ieee_exception	;\
841	nop				;\
842	.align	32
843
844#define	FP_TRAP				\
845	TT_TRACE(trace_gen)		;\
846	ba,pt	%xcc,.fp_exception	;\
847	nop				;\
848	.align	32
849
850#if !defined(lint)
851/*
852 * asynchronous traps at level 0 and level 1
853 *
854 * The first instruction must be a membar for UltraSPARC-III
855 * to stop RED state entry if the store queue has many
856 * pending bad stores (PRM, Chapter 11).
857 */
858#define ASYNC_TRAP(ttype, ttlabel, table_name)\
859	.global	table_name	;\
860table_name:			;\
861	membar	#Sync		;\
862	TT_TRACE(ttlabel)	;\
863	ba	async_err	;\
864	mov	ttype, %g5	;\
865	.align	32
866
867/*
868 * Defaults to BAD entry, but establishes label to be used for
869 * architecture-specific overwrite of trap table entry.
870 */
871#define	LABELED_BAD(table_name)		\
872	.global	table_name		;\
873table_name:				;\
874	BAD
875
876#endif /* !lint */
877
878/*
879 * illegal instruction trap
880 */
881#define	ILLTRAP_INSTR			  \
882	membar	#Sync			  ;\
883	TT_TRACE(trace_gen)		  ;\
884	or	%g0, P_UTRAP4, %g2	  ;\
885	or	%g0, T_UNIMP_INSTR, %g3   ;\
886	sethi	%hi(.check_v9utrap), %g4  ;\
887	jmp	%g4 + %lo(.check_v9utrap) ;\
888	nop				  ;\
889	.align	32
890
891/*
892 * tag overflow trap
893 */
894#define	TAG_OVERFLOW			  \
895	TT_TRACE(trace_gen)		  ;\
896	or	%g0, P_UTRAP10, %g2	  ;\
897	or	%g0, T_TAG_OVERFLOW, %g3  ;\
898	sethi	%hi(.check_v9utrap), %g4  ;\
899	jmp	%g4 + %lo(.check_v9utrap) ;\
900	nop				  ;\
901	.align	32
902
903/*
904 * divide by zero trap
905 */
906#define	DIV_BY_ZERO			  \
907	TT_TRACE(trace_gen)		  ;\
908	or	%g0, P_UTRAP11, %g2	  ;\
909	or	%g0, T_IDIV0, %g3	  ;\
910	sethi	%hi(.check_v9utrap), %g4  ;\
911	jmp	%g4 + %lo(.check_v9utrap) ;\
912	nop				  ;\
913	.align	32
914
915/*
916 * trap instruction for V9 user trap handlers
917 */
918#define	TRAP_INSTR			  \
919	TT_TRACE(trace_gen)		  ;\
920	or	%g0, T_SOFTWARE_TRAP, %g3 ;\
921	sethi	%hi(.check_v9utrap), %g4  ;\
922	jmp	%g4 + %lo(.check_v9utrap) ;\
923	nop				  ;\
924	.align	32
925#define	TRP4	TRAP_INSTR; TRAP_INSTR; TRAP_INSTR; TRAP_INSTR
926
927/*
928 * LEVEL_INTERRUPT is for level N interrupts.
929 * VECTOR_INTERRUPT is for the vector trap.
930 */
931#define	LEVEL_INTERRUPT(level)		\
932	.global	tt_pil/**/level		;\
933tt_pil/**/level:			;\
934	ba,pt	%xcc, pil_interrupt	;\
935	mov	level, %g4		;\
936	.align	32
937
938#define	LEVEL14_INTERRUPT			\
939	ba	pil14_interrupt			;\
940	mov	PIL_14, %g4			;\
941	.align	32
942
943#define	VECTOR_INTERRUPT				\
944	ldxa	[%g0]ASI_INTR_RECEIVE_STATUS, %g1	;\
945	btst	IRSR_BUSY, %g1				;\
946	bnz,pt	%xcc, vec_interrupt			;\
947	nop						;\
948	ba,a,pt	%xcc, vec_intr_spurious			;\
949	.empty						;\
950	.align	32
951
952/*
953 * MMU Trap Handlers.
954 */
955#define	SWITCH_GLOBALS	/* mmu->alt, alt->mmu */			\
956	rdpr	%pstate, %g5						;\
957	wrpr	%g5, PSTATE_MG | PSTATE_AG, %pstate
958
959#define	IMMU_EXCEPTION							\
960	membar	#Sync							;\
961	SWITCH_GLOBALS							;\
962	wr	%g0, ASI_IMMU, %asi					;\
963	rdpr	%tpc, %g2						;\
964	ldxa	[MMU_SFSR]%asi, %g3					;\
965	ba,pt	%xcc, .mmu_exception_end				;\
966	mov	T_INSTR_EXCEPTION, %g1					;\
967	.align	32
968
969#define	DMMU_EXCEPTION							\
970	SWITCH_GLOBALS							;\
971	wr	%g0, ASI_DMMU, %asi					;\
972	ldxa	[MMU_TAG_ACCESS]%asi, %g2				;\
973	ldxa	[MMU_SFSR]%asi, %g3					;\
974	ba,pt	%xcc, .mmu_exception_end				;\
975	mov	T_DATA_EXCEPTION, %g1					;\
976	.align	32
977
978#define	DMMU_EXC_AG_PRIV						\
979	wr	%g0, ASI_DMMU, %asi					;\
980	ldxa	[MMU_SFAR]%asi, %g2					;\
981	ba,pt	%xcc, .mmu_priv_exception				;\
982	ldxa	[MMU_SFSR]%asi, %g3					;\
983	.align	32
984
985#define	DMMU_EXC_AG_NOT_ALIGNED						\
986	wr	%g0, ASI_DMMU, %asi					;\
987	ldxa	[MMU_SFAR]%asi, %g2					;\
988	ba,pt	%xcc, .mmu_exception_not_aligned			;\
989	ldxa	[MMU_SFSR]%asi, %g3					;\
990	.align	32
991
992/*
993 * SPARC V9 IMPL. DEP. #109(1) and (2) and #110(1) and (2)
994 */
995#define	DMMU_EXC_LDDF_NOT_ALIGNED					\
996	btst	1, %sp							;\
997	bnz,pt	%xcc, .lddf_exception_not_aligned			;\
998	wr	%g0, ASI_DMMU, %asi					;\
999	ldxa	[MMU_SFAR]%asi, %g2					;\
1000	ba,pt	%xcc, .mmu_exception_not_aligned			;\
1001	ldxa	[MMU_SFSR]%asi, %g3					;\
1002	.align	32
1003
1004#define	DMMU_EXC_STDF_NOT_ALIGNED					\
1005	btst	1, %sp							;\
1006	bnz,pt	%xcc, .stdf_exception_not_aligned			;\
1007	wr	%g0, ASI_DMMU, %asi					;\
1008	ldxa	[MMU_SFAR]%asi, %g2					;\
1009	ba,pt	%xcc, .mmu_exception_not_aligned			;\
1010	ldxa	[MMU_SFSR]%asi, %g3					;\
1011	.align	32
1012
1013/*
1014 * Flush the TLB using either the primary, secondary, or nucleus flush
1015 * operation based on whether the ctx from the tag access register matches
1016 * the primary or secondary context (flush the nucleus if neither matches).
1017 *
1018 * Requires a membar #Sync before next ld/st.
1019 * exits with:
1020 * g2 = tag access register
1021 * g3 = ctx number
1022 */
1023#if TAGACC_CTX_MASK != CTXREG_CTX_MASK
1024#error "TAGACC_CTX_MASK != CTXREG_CTX_MASK"
1025#endif
1026#define	DTLB_DEMAP_ENTRY						\
1027	mov	MMU_TAG_ACCESS, %g1					;\
1028	mov	MMU_PCONTEXT, %g5					;\
1029	ldxa	[%g1]ASI_DMMU, %g2					;\
1030	sethi	%hi(TAGACC_CTX_MASK), %g4				;\
1031	or	%g4, %lo(TAGACC_CTX_MASK), %g4				;\
1032	and	%g2, %g4, %g3			/* g3 = ctx */		;\
1033	ldxa	[%g5]ASI_DMMU, %g6		/* g6 = primary ctx */	;\
1034	and	%g6, %g4, %g6			/* &= CTXREG_CTX_MASK */ ;\
1035	cmp	%g3, %g6						;\
1036	be,pt	%xcc, 1f						;\
1037	andn	%g2, %g4, %g1			/* ctx = primary */	;\
1038	mov	MMU_SCONTEXT, %g5					;\
1039	ldxa	[%g5]ASI_DMMU, %g6		/* g6 = secondary ctx */ ;\
1040	and	%g6, %g4, %g6			/* &= CTXREG_CTX_MASK */ ;\
1041	cmp	%g3, %g6						;\
1042	be,a,pt	%xcc, 1f						;\
1043	  or	%g1, DEMAP_SECOND, %g1					;\
1044	or	%g1, DEMAP_NUCLEUS, %g1					;\
10451:	stxa	%g0, [%g1]ASI_DTLB_DEMAP	/* MMU_DEMAP_PAGE */	;\
1046	membar	#Sync
1047
1048#if defined(cscope)
1049/*
1050 * Define labels to direct cscope quickly to labels that
1051 * are generated by macro expansion of DTLB_MISS().
1052 */
1053	.global	tt0_dtlbmiss
1054tt0_dtlbmiss:
1055	.global	tt1_dtlbmiss
1056tt1_dtlbmiss:
1057	nop
1058#endif
1059
1060/*
1061 * Needs to be exactly 32 instructions
1062 *
1063 * UTLB NOTE: If we don't hit on the 8k pointer then we branch
1064 * to a special 4M tsb handler. It would be nice if that handler
1065 * could live in this file but currently it seems better to allow
1066 * it to fall thru to sfmmu_tsb_miss.
1067 */
1068#ifdef UTSB_PHYS
1069#define	DTLB_MISS(table_name)						;\
1070	.global	table_name/**/_dtlbmiss					;\
1071table_name/**/_dtlbmiss:						;\
1072	mov	MMU_TAG_ACCESS, %g6		/* select tag acc */	;\
1073	ldxa	[%g0]ASI_DMMU_TSB_8K, %g1	/* g1 = tsbe ptr */	;\
1074	ldxa	[%g6]ASI_DMMU, %g2		/* g2 = tag access */	;\
1075	sllx	%g2, TAGACC_CTX_LSHIFT, %g3				;\
1076	srlx	%g3, TAGACC_CTX_LSHIFT, %g3	/* g3 = ctx */		;\
1077	cmp	%g3, INVALID_CONTEXT					;\
1078	ble,pn	%xcc, sfmmu_kdtlb_miss					;\
1079	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
1080	mov	SCRATCHPAD_UTSBREG, %g3				;\
1081	ldxa	[%g3]ASI_SCRATCHPAD, %g3	/* g3 = 2nd tsb reg */	;\
1082	brgez,pn %g3, sfmmu_udtlb_slowpath	/* branch if 2 TSBs */	;\
1083	  nop								;\
1084	ldda	[%g1]ASI_QUAD_LDD_PHYS, %g4	/* g4 = tag, %g5 data */;\
1085	cmp	%g4, %g7						;\
1086	bne,pn	%xcc, sfmmu_tsb_miss_tt		/* no 4M TSB, miss */	;\
1087	  mov	-1, %g3			/* set 4M tsbe ptr to -1 */	;\
1088	TT_TRACE(trace_tsbhit)		/* 2 instr ifdef TRAPTRACE */	;\
1089	stxa	%g5, [%g0]ASI_DTLB_IN	/* trapstat expects TTE */	;\
1090	retry				/* in %g5 */			;\
1091	unimp	0							;\
1092	unimp	0							;\
1093	unimp	0							;\
1094	unimp	0							;\
1095	unimp	0							;\
1096	unimp	0							;\
1097	unimp	0							;\
1098	unimp	0							;\
1099	unimp	0							;\
1100        unimp   0                                                       ;\
1101        unimp   0                                                       ;\
1102        unimp   0                                                       ;\
1103	.align 128
1104#else /* UTSB_PHYS */
1105#define	DTLB_MISS(table_name)						;\
1106	.global	table_name/**/_dtlbmiss					;\
1107table_name/**/_dtlbmiss:						;\
1108	mov	MMU_TAG_ACCESS, %g6		/* select tag acc */	;\
1109	ldxa	[%g0]ASI_DMMU_TSB_8K, %g1	/* g1 = tsbe ptr */	;\
1110	ldxa	[%g6]ASI_DMMU, %g2		/* g2 = tag access */	;\
1111	sllx	%g2, TAGACC_CTX_LSHIFT, %g3				;\
1112	srlx	%g3, TAGACC_CTX_LSHIFT, %g3	/* g3 = ctx */		;\
1113	cmp	%g3, INVALID_CONTEXT					;\
1114	ble,pn	%xcc, sfmmu_kdtlb_miss					;\
1115	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
1116	brlz,pn %g1, sfmmu_udtlb_slowpath				;\
1117	  nop								;\
1118	ldda	[%g1]ASI_NQUAD_LD, %g4	/* g4 = tag, %g5 data */	;\
1119	cmp	%g4, %g7						;\
1120	bne,pn	%xcc, sfmmu_tsb_miss_tt		/* no 4M TSB, miss */	;\
1121	  mov	-1, %g3		/* set 4M tsbe ptr to -1 */		;\
1122	TT_TRACE(trace_tsbhit)		/* 2 instr ifdef TRAPTRACE */	;\
1123	stxa	%g5, [%g0]ASI_DTLB_IN	/* trapstat expects TTE */	;\
1124	retry				/* in %g5 */			;\
1125	unimp	0							;\
1126	unimp	0							;\
1127	unimp	0							;\
1128	unimp	0							;\
1129	unimp	0							;\
1130        unimp   0                                                       ;\
1131        unimp   0                                                       ;\
1132        unimp   0                                                       ;\
1133        unimp   0                                                       ;\
1134        unimp   0                                                       ;\
1135        unimp   0                                                       ;\
1136        unimp   0                                                       ;\
1137        unimp   0                                                       ;\
1138        unimp   0                                                       ;\
1139	.align 128
1140#endif /* UTSB_PHYS */
1141
1142#if defined(cscope)
1143/*
1144 * Define labels to direct cscope quickly to labels that
1145 * are generated by macro expansion of ITLB_MISS().
1146 */
1147	.global	tt0_itlbmiss
1148tt0_itlbmiss:
1149	.global	tt1_itlbmiss
1150tt1_itlbmiss:
1151	nop
1152#endif
1153
1154/*
1155 * Instruction miss handler.
1156 * ldda instructions will have their ASI patched
1157 * by sfmmu_patch_ktsb at runtime.
1158 * MUST be EXACTLY 32 instructions or we'll break.
1159 */
1160#ifdef UTSB_PHYS
1161#define	ITLB_MISS(table_name)						 \
1162	.global	table_name/**/_itlbmiss					;\
1163table_name/**/_itlbmiss:						;\
1164	mov	MMU_TAG_ACCESS, %g6		/* select tag acc */	;\
1165	ldxa	[%g0]ASI_IMMU_TSB_8K, %g1	/* g1 = tsbe ptr */	;\
1166	ldxa	[%g6]ASI_IMMU, %g2		/* g2 = tag access */	;\
1167	sllx	%g2, TAGACC_CTX_LSHIFT, %g3				;\
1168	srlx	%g3, TAGACC_CTX_LSHIFT, %g3	/* g3 = ctx */		;\
1169	cmp	%g3, INVALID_CONTEXT					;\
1170	ble,pn	%xcc, sfmmu_kitlb_miss					;\
1171	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
1172	mov	SCRATCHPAD_UTSBREG, %g3				;\
1173	ldxa	[%g3]ASI_SCRATCHPAD, %g3	/* g3 = 2nd tsb reg */	;\
1174	brgez,pn %g3, sfmmu_uitlb_slowpath	/* branch if 2 TSBs */	;\
1175	  nop								;\
1176	ldda	[%g1]ASI_QUAD_LDD_PHYS, %g4 /* g4 = tag, g5 = data */	;\
1177	cmp	%g4, %g7						;\
1178	bne,pn	%xcc, sfmmu_tsb_miss_tt	/* br if 8k ptr miss */		;\
1179	  mov	-1, %g3		/* set 4M TSB ptr to -1 */		;\
1180	andcc	%g5, TTE_EXECPRM_INT, %g0 /* check execute bit */	;\
1181	bz,pn	%icc, exec_fault					;\
1182	  nop								;\
1183	TT_TRACE(trace_tsbhit)		/* 2 instr ifdef TRAPTRACE */	;\
1184	stxa	%g5, [%g0]ASI_ITLB_IN	/* trapstat expects %g5 */	;\
1185	retry								;\
1186	unimp	0							;\
1187	unimp	0							;\
1188	unimp	0							;\
1189	unimp	0							;\
1190	unimp	0							;\
1191	unimp	0							;\
1192        unimp   0                                                       ;\
1193        unimp   0                                                       ;\
1194        unimp   0                                                       ;\
1195	.align 128
1196#else /* UTSB_PHYS */
1197#define	ITLB_MISS(table_name)						 \
1198	.global	table_name/**/_itlbmiss					;\
1199table_name/**/_itlbmiss:						;\
1200	mov	MMU_TAG_ACCESS, %g6		/* select tag acc */	;\
1201	ldxa	[%g0]ASI_IMMU_TSB_8K, %g1	/* g1 = tsbe ptr */	;\
1202	ldxa	[%g6]ASI_IMMU, %g2		/* g2 = tag access */	;\
1203	sllx	%g2, TAGACC_CTX_LSHIFT, %g3				;\
1204	srlx	%g3, TAGACC_CTX_LSHIFT, %g3	/* g3 = ctx */		;\
1205	cmp	%g3, INVALID_CONTEXT					;\
1206	ble,pn	%xcc, sfmmu_kitlb_miss					;\
1207	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
1208	brlz,pn	%g1, sfmmu_uitlb_slowpath	/* if >1 TSB branch */	;\
1209	  nop								;\
1210	ldda	[%g1]ASI_NQUAD_LD, %g4	/* g4 = tag, g5 = data */	;\
1211	cmp	%g4, %g7						;\
1212	bne,pn	%xcc, sfmmu_tsb_miss_tt	/* br if 8k ptr miss */		;\
1213	  mov	-1, %g3		/* set 4M TSB ptr to -1 */		;\
1214	andcc	%g5, TTE_EXECPRM_INT, %g0 /* check execute bit */	;\
1215	bz,pn	%icc, exec_fault					;\
1216	  nop								;\
1217	TT_TRACE(trace_tsbhit)		/* 2 instr ifdef TRAPTRACE */	;\
1218	stxa	%g5, [%g0]ASI_ITLB_IN	/* trapstat expects %g5 */	;\
1219	retry								;\
1220	unimp	0							;\
1221	unimp	0							;\
1222	unimp	0							;\
1223	unimp	0							;\
1224	unimp	0							;\
1225        unimp   0                                                       ;\
1226        unimp   0                                                       ;\
1227        unimp   0                                                       ;\
1228        unimp   0                                                       ;\
1229        unimp   0                                                       ;\
1230        unimp   0                                                       ;\
1231	.align 128
1232#endif /* UTSB_PHYS */
1233
1234
1235/*
1236 * This macro is the first level handler for fast protection faults.
1237 * It first demaps the tlb entry which generated the fault and then
1238 * attempts to set the modify bit on the hash.  It needs to be
1239 * exactly 32 instructions.
1240 */
1241#define	DTLB_PROT							 \
1242	DTLB_DEMAP_ENTRY		/* 20 instructions */		;\
1243	/*								;\
1244	 * At this point:						;\
1245	 *   g1 = ????							;\
1246	 *   g2 = tag access register					;\
1247	 *   g3 = ctx number						;\
1248	 *   g4 = ????							;\
1249	 */								;\
1250	TT_TRACE(trace_dataprot)	/* 2 instr ifdef TRAPTRACE */	;\
1251					/* clobbers g1 and g6 */	;\
1252	ldxa	[%g0]ASI_DMMU_TSB_8K, %g1	/* g1 = tsbe ptr */	;\
1253	brnz,pt %g3, sfmmu_uprot_trap		/* user trap */		;\
1254	  nop								;\
1255	ba,a,pt	%xcc, sfmmu_kprot_trap		/* kernel trap */	;\
1256	unimp	0							;\
1257	unimp	0							;\
1258	unimp	0							;\
1259	unimp	0							;\
1260	unimp	0							;\
1261	unimp	0							;\
1262	.align 128
1263
1264#define	DMMU_EXCEPTION_TL1						;\
1265	SWITCH_GLOBALS							;\
1266	ba,a,pt	%xcc, mmu_trap_tl1					;\
1267	  nop								;\
1268	.align 32
1269
1270#define	MISALIGN_ADDR_TL1						;\
1271	ba,a,pt	%xcc, mmu_trap_tl1					;\
1272	  nop								;\
1273	.align 32
1274
1275/*
1276 * Trace a tsb hit
1277 * g1 = tsbe pointer (in/clobbered)
1278 * g2 = tag access register (in)
1279 * g3 - g4 = scratch (clobbered)
1280 * g5 = tsbe data (in)
1281 * g6 = scratch (clobbered)
1282 * g7 = pc we jumped here from (in)
1283 * ttextra = value to OR in to trap type (%tt) (in)
1284 */
1285#ifdef TRAPTRACE
1286#define TRACE_TSBHIT(ttextra)						 \
1287	membar	#Sync							;\
1288	sethi	%hi(FLUSH_ADDR), %g6					;\
1289	flush	%g6							;\
1290	TRACE_PTR(%g3, %g6)						;\
1291	GET_TRACE_TICK(%g6)						;\
1292	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi				;\
1293	stxa	%g2, [%g3 + TRAP_ENT_SP]%asi	/* tag access */	;\
1294	stxa	%g5, [%g3 + TRAP_ENT_F1]%asi	/* tsb data */		;\
1295	rdpr	%tnpc, %g6						;\
1296	stxa	%g6, [%g3 + TRAP_ENT_F2]%asi				;\
1297	stxa	%g1, [%g3 + TRAP_ENT_F3]%asi	/* tsb pointer */	;\
1298	stxa	%g0, [%g3 + TRAP_ENT_F4]%asi				;\
1299	rdpr	%tpc, %g6						;\
1300	stxa	%g6, [%g3 + TRAP_ENT_TPC]%asi				;\
1301	rdpr	%tl, %g6						;\
1302	stha	%g6, [%g3 + TRAP_ENT_TL]%asi				;\
1303	rdpr	%tt, %g6						;\
1304	or	%g6, (ttextra), %g6					;\
1305	stha	%g6, [%g3 + TRAP_ENT_TT]%asi				;\
1306	ldxa	[%g0]ASI_IMMU, %g1		/* tag target */	;\
1307	ldxa	[%g0]ASI_DMMU, %g4					;\
1308	cmp	%g6, FAST_IMMU_MISS_TT					;\
1309	movne	%icc, %g4, %g1						;\
1310	stxa	%g1, [%g3 + TRAP_ENT_TSTATE]%asi /* tsb tag */		;\
1311	stxa	%g0, [%g3 + TRAP_ENT_TR]%asi				;\
1312	TRACE_NEXT(%g3, %g4, %g6)
1313#else
1314#define TRACE_TSBHIT(ttextra)
1315#endif
1316
1317#if defined(lint)
1318
1319struct scb	trap_table;
1320struct scb	scb;		/* trap_table/scb are the same object */
1321
1322#else /* lint */
1323
1324/*
1325 * =======================================================================
1326 *		SPARC V9 TRAP TABLE
1327 *
1328 * The trap table is divided into two halves: the first half is used when
1329 * taking traps when TL=0; the second half is used when taking traps from
1330 * TL>0. Note that handlers in the second half of the table might not be able
1331 * to make the same assumptions as handlers in the first half of the table.
1332 *
1333 * Worst case trap nesting so far:
1334 *
1335 *	at TL=0 client issues software trap requesting service
1336 *	at TL=1 nucleus wants a register window
1337 *	at TL=2 register window clean/spill/fill takes a TLB miss
1338 *	at TL=3 processing TLB miss
1339 *	at TL=4 handle asynchronous error
1340 *
1341 * Note that a trap from TL=4 to TL=5 places Spitfire in "RED mode".
1342 *
1343 * =======================================================================
1344 */
1345	.section ".text"
1346	.align	4
1347	.global trap_table, scb, trap_table0, trap_table1, etrap_table
1348	.type	trap_table, #object
1349	.type	scb, #object
1350trap_table:
1351scb:
1352trap_table0:
1353	/* hardware traps */
1354	NOT;				/* 000	reserved */
1355	RED;				/* 001	power on reset */
1356	RED;				/* 002	watchdog reset */
1357	RED;				/* 003	externally initiated reset */
1358	RED;				/* 004	software initiated reset */
1359	RED;				/* 005	red mode exception */
1360	NOT; NOT;			/* 006 - 007 reserved */
1361	IMMU_EXCEPTION;			/* 008	instruction access exception */
1362	NOT;				/* 009	instruction access MMU miss */
1363	ASYNC_TRAP(T_INSTR_ERROR, trace_gen, tt0_iae);
1364					/* 00A	instruction access error */
1365	NOT; NOT4;			/* 00B - 00F reserved */
1366	ILLTRAP_INSTR;			/* 010	illegal instruction */
1367	TRAP(T_PRIV_INSTR);		/* 011	privileged opcode */
1368	NOT;				/* 012	unimplemented LDD */
1369	NOT;				/* 013	unimplemented STD */
1370	NOT4; NOT4; NOT4;		/* 014 - 01F reserved */
1371	FP_DISABLED_TRAP;		/* 020	fp disabled */
1372	FP_IEEE_TRAP;			/* 021	fp exception ieee 754 */
1373	FP_TRAP;			/* 022	fp exception other */
1374	TAG_OVERFLOW;			/* 023	tag overflow */
1375	CLEAN_WINDOW;			/* 024 - 027 clean window */
1376	DIV_BY_ZERO;			/* 028	division by zero */
1377	NOT;				/* 029	internal processor error */
1378	NOT; NOT; NOT4;			/* 02A - 02F reserved */
1379	DMMU_EXCEPTION;			/* 030	data access exception */
1380	NOT;				/* 031	data access MMU miss */
1381	ASYNC_TRAP(T_DATA_ERROR, trace_gen, tt0_dae);
1382					/* 032	data access error */
1383	NOT;				/* 033	data access protection */
1384	DMMU_EXC_AG_NOT_ALIGNED;	/* 034	mem address not aligned */
1385	DMMU_EXC_LDDF_NOT_ALIGNED;	/* 035	LDDF mem address not aligned */
1386	DMMU_EXC_STDF_NOT_ALIGNED;	/* 036	STDF mem address not aligned */
1387	DMMU_EXC_AG_PRIV;		/* 037	privileged action */
1388	NOT;				/* 038	LDQF mem address not aligned */
1389	NOT;				/* 039	STQF mem address not aligned */
1390	NOT; NOT; NOT4;			/* 03A - 03F reserved */
1391	LABELED_BAD(tt0_asdat);		/* 040	async data error */
1392	LEVEL_INTERRUPT(1);		/* 041	interrupt level 1 */
1393	LEVEL_INTERRUPT(2);		/* 042	interrupt level 2 */
1394	LEVEL_INTERRUPT(3);		/* 043	interrupt level 3 */
1395	LEVEL_INTERRUPT(4);		/* 044	interrupt level 4 */
1396	LEVEL_INTERRUPT(5);		/* 045	interrupt level 5 */
1397	LEVEL_INTERRUPT(6);		/* 046	interrupt level 6 */
1398	LEVEL_INTERRUPT(7);		/* 047	interrupt level 7 */
1399	LEVEL_INTERRUPT(8);		/* 048	interrupt level 8 */
1400	LEVEL_INTERRUPT(9);		/* 049	interrupt level 9 */
1401	LEVEL_INTERRUPT(10);		/* 04A	interrupt level 10 */
1402	LEVEL_INTERRUPT(11);		/* 04B	interrupt level 11 */
1403	LEVEL_INTERRUPT(12);		/* 04C	interrupt level 12 */
1404	LEVEL_INTERRUPT(13);		/* 04D	interrupt level 13 */
1405	LEVEL14_INTERRUPT;		/* 04E	interrupt level 14 */
1406	LEVEL_INTERRUPT(15);		/* 04F	interrupt level 15 */
1407	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F reserved */
1408	VECTOR_INTERRUPT;		/* 060	interrupt vector */
1409	GOTO(kmdb_trap);		/* 061	PA watchpoint */
1410	GOTO(kmdb_trap);		/* 062	VA watchpoint */
1411	GOTO_TT(ce_err, trace_gen);	/* 063	corrected ECC error */
1412	ITLB_MISS(tt0);			/* 064	instruction access MMU miss */
1413	DTLB_MISS(tt0);			/* 068	data access MMU miss */
1414	DTLB_PROT;			/* 06C	data access protection */
1415	LABELED_BAD(tt0_fecc);		/* 070  fast ecache ECC error */
1416	LABELED_BAD(tt0_dperr);		/* 071  Cheetah+ dcache parity error */
1417	LABELED_BAD(tt0_iperr);		/* 072  Cheetah+ icache parity error */
1418	NOT;				/* 073  reserved */
1419	NOT4; NOT4; NOT4;		/* 074 - 07F reserved */
1420	NOT4;				/* 080	spill 0 normal */
1421	SPILL_32bit_asi(ASI_AIUP,sn0);	/* 084	spill 1 normal */
1422	SPILL_64bit_asi(ASI_AIUP,sn0);	/* 088	spill 2 normal */
1423	SPILL_32clean(ASI_AIUP,sn0);	/* 08C	spill 3 normal */
1424	SPILL_64clean(ASI_AIUP,sn0);	/* 090	spill 4 normal */
1425	SPILL_32bit(not);		/* 094	spill 5 normal */
1426	SPILL_64bit(not);		/* 098	spill 6 normal */
1427	SPILL_mixed;			/* 09C	spill 7 normal */
1428	NOT4;				/* 0A0	spill 0 other */
1429	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0A4	spill 1 other */
1430	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0A8	spill 2 other */
1431	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0AC	spill 3 other */
1432	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0B0	spill 4 other */
1433	NOT4;				/* 0B4	spill 5 other */
1434	NOT4;				/* 0B8	spill 6 other */
1435	NOT4;				/* 0BC	spill 7 other */
1436	NOT4;				/* 0C0	fill 0 normal */
1437	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0C4	fill 1 normal */
1438	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0C8	fill 2 normal */
1439	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0CC	fill 3 normal */
1440	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0D0	fill 4 normal */
1441	FILL_32bit(not);		/* 0D4	fill 5 normal */
1442	FILL_64bit(not);		/* 0D8	fill 6 normal */
1443	FILL_mixed;			/* 0DC	fill 7 normal */
1444	NOT4;				/* 0E0	fill 0 other */
1445	NOT4;				/* 0E4	fill 1 other */
1446	NOT4;				/* 0E8	fill 2 other */
1447	NOT4;				/* 0EC	fill 3 other */
1448	NOT4;				/* 0F0	fill 4 other */
1449	NOT4;				/* 0F4	fill 5 other */
1450	NOT4;				/* 0F8	fill 6 other */
1451	NOT4;				/* 0FC	fill 7 other */
1452	/* user traps */
1453	GOTO(syscall_trap_4x);		/* 100	old system call */
1454	TRAP(T_BREAKPOINT);		/* 101	user breakpoint */
1455	TRAP(T_DIV0);			/* 102	user divide by zero */
1456	FLUSHW(tt0_flushw);		/* 103	flush windows */
1457	GOTO(.clean_windows);		/* 104	clean windows */
1458	BAD;				/* 105	range check ?? */
1459	GOTO(.fix_alignment);		/* 106	do unaligned references */
1460	BAD;				/* 107	unused */
1461	SYSCALL_TRAP32;			/* 108	ILP32 system call on LP64 */
1462	GOTO(set_trap0_addr);		/* 109	set trap0 address */
1463	BAD; BAD; BAD4;			/* 10A - 10F unused */
1464	TRP4; TRP4; TRP4; TRP4;		/* 110 - 11F V9 user trap handlers */
1465	GOTO(.getcc);			/* 120	get condition codes */
1466	GOTO(.setcc);			/* 121	set condition codes */
1467	GOTO(.getpsr);			/* 122	get psr */
1468	GOTO(.setpsr);			/* 123	set psr (some fields) */
1469	GOTO(get_timestamp);		/* 124	get timestamp */
1470	GOTO(get_virtime);		/* 125	get lwp virtual time */
1471	PRIV(self_xcall);		/* 126	self xcall */
1472	GOTO(get_hrestime);		/* 127	get hrestime */
1473	BAD;				/* 128	ST_SETV9STACK */
1474	GOTO(.getlgrp);			/* 129  get lgrpid */
1475	BAD; BAD; BAD4;			/* 12A - 12F unused */
1476	BAD4; BAD4; 			/* 130 - 137 unused */
1477	DTRACE_PID;			/* 138  dtrace pid tracing provider */
1478	BAD;				/* 139  unused */
1479	DTRACE_RETURN;			/* 13A	dtrace pid return probe */
1480	BAD; BAD4;			/* 13B - 13F unused */
1481	SYSCALL_TRAP;			/* 140  LP64 system call */
1482	SYSCALL(nosys);			/* 141  unused system call trap */
1483#ifdef DEBUG_USER_TRAPTRACECTL
1484	GOTO(.traptrace_freeze);	/* 142  freeze traptrace */
1485	GOTO(.traptrace_unfreeze);	/* 143  unfreeze traptrace */
1486#else
1487	SYSCALL(nosys);			/* 142  unused system call trap */
1488	SYSCALL(nosys);			/* 143  unused system call trap */
1489#endif
1490	BAD4; BAD4; BAD4;		/* 144 - 14F unused */
1491	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1492	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1493	BAD;				/* 170 - unused */
1494	BAD;				/* 171 - unused */
1495	BAD; BAD;			/* 172 - 173 unused */
1496	BAD4; BAD4;			/* 174 - 17B unused */
1497#ifdef	PTL1_PANIC_DEBUG
1498	mov PTL1_BAD_DEBUG, %g1; GOTO(ptl1_panic);
1499					/* 17C	test ptl1_panic */
1500#else
1501	BAD;				/* 17C  unused */
1502#endif	/* PTL1_PANIC_DEBUG */
1503	PRIV(kmdb_trap);		/* 17D	kmdb enter (L1-A) */
1504	PRIV(kmdb_trap);		/* 17E	kmdb breakpoint */
1505	PRIV(kctx_obp_bpt);		/* 17F	obp breakpoint */
1506	/* reserved */
1507	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1508	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1509	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1510	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1511	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1512	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1513	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1514	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1515trap_table1:
1516	NOT4; NOT4; NOT; NOT;		/* 000 - 009 unused */
1517	ASYNC_TRAP(T_INSTR_ERROR + T_TL1, trace_gen, tt1_iae);
1518					/* 00A	instruction access error */
1519	NOT; NOT4;			/* 00B - 00F unused */
1520	NOT4; NOT4; NOT4; NOT4;		/* 010 - 01F unused */
1521	NOT4;				/* 020 - 023 unused */
1522	CLEAN_WINDOW;			/* 024 - 027 clean window */
1523	NOT4; NOT4;			/* 028 - 02F unused */
1524	DMMU_EXCEPTION_TL1;		/* 030 	data access exception */
1525	NOT;				/* 031 unused */
1526	ASYNC_TRAP(T_DATA_ERROR + T_TL1, trace_gen, tt1_dae);
1527					/* 032	data access error */
1528	NOT;				/* 033	unused */
1529	MISALIGN_ADDR_TL1;		/* 034	mem address not aligned */
1530	NOT; NOT; NOT; NOT4; NOT4	/* 035 - 03F unused */
1531	LABELED_BAD(tt1_asdat);		/* 040	async data error */
1532	NOT; NOT; NOT;			/* 041 - 043 unused */
1533	NOT4; NOT4; NOT4;		/* 044 - 04F unused */
1534	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F unused */
1535	NOT;				/* 060	unused */
1536	GOTO(kmdb_trap_tl1);		/* 061	PA watchpoint */
1537	GOTO(kmdb_trap_tl1);		/* 062	VA watchpoint */
1538	GOTO_TT(ce_err_tl1, trace_gen);	/* 063	corrected ECC error */
1539	ITLB_MISS(tt1);			/* 064	instruction access MMU miss */
1540	DTLB_MISS(tt1);			/* 068	data access MMU miss */
1541	DTLB_PROT;			/* 06C	data access protection */
1542	LABELED_BAD(tt1_fecc);		/* 070  fast ecache ECC error */
1543	LABELED_BAD(tt1_dperr);		/* 071  Cheetah+ dcache parity error */
1544	LABELED_BAD(tt1_iperr);		/* 072  Cheetah+ icache parity error */
1545	NOT;				/* 073  reserved */
1546	NOT4; NOT4; NOT4;		/* 074 - 07F reserved */
1547	NOT4;				/* 080	spill 0 normal */
1548	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 084	spill 1 normal */
1549	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 088	spill 2 normal */
1550	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 08C	spill 3 normal */
1551	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 090	spill 4 normal */
1552	SPILL_32bit(not);		/* 094	spill 5 normal */
1553	SPILL_64bit(not);		/* 098	spill 6 normal */
1554	SPILL_mixed;			/* 09C	spill 7 normal */
1555	NOT4;				/* 0A0	spill 0 other */
1556	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0A4	spill 1 other */
1557	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0A8	spill 2 other */
1558	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0AC	spill 3 other */
1559	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0B0  spill 4 other */
1560	NOT4;				/* 0B4  spill 5 other */
1561	NOT4;				/* 0B8  spill 6 other */
1562	NOT4;				/* 0BC  spill 7 other */
1563	NOT4;				/* 0C0	fill 0 normal */
1564	FILL_32bit_tt1(ASI_AIUP,fn1);	/* 0C4	fill 1 normal */
1565	FILL_64bit_tt1(ASI_AIUP,fn1);	/* 0C8	fill 2 normal */
1566	FILL_32bit_tt1(ASI_AIUP,fn1);	/* 0CC	fill 3 normal */
1567	FILL_64bit_tt1(ASI_AIUP,fn1);	/* 0D0	fill 4 normal */
1568	FILL_32bit(not);		/* 0D4	fill 5 normal */
1569	FILL_64bit(not);		/* 0D8	fill 6 normal */
1570	FILL_mixed;			/* 0DC	fill 7 normal */
1571	NOT4; NOT4; NOT4; NOT4;		/* 0E0 - 0EF unused */
1572	NOT4; NOT4; NOT4; NOT4;		/* 0F0 - 0FF unused */
1573	LABELED_BAD(tt1_swtrap0);	/* 100  fast ecache ECC error (cont) */
1574	LABELED_BAD(tt1_swtrap1);	/* 101  Ch+ D$ parity error (cont) */
1575	LABELED_BAD(tt1_swtrap2);	/* 102  Ch+ I$ parity error (cont) */
1576	NOT;				/* 103  reserved */
1577/*
1578 * We only reserve the above four special case soft traps for code running
1579 * at TL>0, so we can truncate the trap table here.
1580 */
1581etrap_table:
1582	.size	trap_table, (.-trap_table)
1583	.size	scb, (.-scb)
1584
1585/*
1586 * We get to exec_fault in the case of an instruction miss and tte
1587 * has no execute bit set.  We go to tl0 to handle it.
1588 *
1589 * g1 = tsbe pointer (in/clobbered)
1590 * g2 = tag access register (in)
1591 * g3 - g4 = scratch (clobbered)
1592 * g5 = tsbe data (in)
1593 * g6 = scratch (clobbered)
1594 */
1595	ALTENTRY(exec_fault)
1596	TRACE_TSBHIT(0x200)
1597	SWITCH_GLOBALS
1598	mov	MMU_TAG_ACCESS, %g4
1599	ldxa	[%g4]ASI_IMMU, %g2			! arg1 = addr
1600	mov	T_INSTR_MMU_MISS, %g3			! arg2 = traptype
1601	set	trap, %g1
1602	ba,pt	%xcc, sys_trap
1603	  mov	-1, %g4
1604
1605.mmu_exception_not_aligned:
1606	rdpr	%tstate, %g1
1607	btst	TSTATE_PRIV, %g1
1608	bnz,pn	%icc, 2f
1609	nop
1610	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1611	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1612	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1613	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1614	brz,pt	%g5, 2f
1615	nop
1616	ldn	[%g5 + P_UTRAP15], %g5			! unaligned utrap?
1617	brz,pn	%g5, 2f
1618	nop
1619	btst	1, %sp
1620	bz,pt	%xcc, 1f				! 32 bit user program
1621	nop
1622	ba,pt	%xcc, .setup_v9utrap			! 64 bit user program
1623	nop
16241:
1625	ba,pt	%xcc, .setup_utrap
1626	or	%g2, %g0, %g7
16272:
1628	ba,pt	%xcc, .mmu_exception_end
1629	mov	T_ALIGNMENT, %g1
1630
1631.mmu_priv_exception:
1632	rdpr	%tstate, %g1
1633	btst	TSTATE_PRIV, %g1
1634	bnz,pn	%icc, 1f
1635	nop
1636	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1637	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1638	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1639	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1640	brz,pt	%g5, 1f
1641	nop
1642	ldn	[%g5 + P_UTRAP16], %g5
1643	brnz,pt	%g5, .setup_v9utrap
1644	nop
16451:
1646	mov	T_PRIV_INSTR, %g1
1647
1648.mmu_exception_end:
1649	CPU_INDEX(%g4, %g5)
1650	set	cpu_core, %g5
1651	sllx	%g4, CPU_CORE_SHIFT, %g4
1652	add	%g4, %g5, %g4
1653	lduh	[%g4 + CPUC_DTRACE_FLAGS], %g5
1654	andcc	%g5, CPU_DTRACE_NOFAULT, %g0
1655	bz	%xcc, .mmu_exception_tlb_chk
1656	or	%g5, CPU_DTRACE_BADADDR, %g5
1657	stuh	%g5, [%g4 + CPUC_DTRACE_FLAGS]
1658	done
1659
1660.mmu_exception_tlb_chk:
1661	GET_CPU_IMPL(%g5)			! check SFSR.FT to see if this
1662	cmp	%g5, PANTHER_IMPL		! is a TLB parity error. But
1663	bne	2f				! we only do this check while
1664	mov	1, %g4				! running on Panther CPUs
1665	sllx	%g4, PN_SFSR_PARITY_SHIFT, %g4	! since US-I/II use the same
1666	andcc	%g3, %g4, %g0			! bit for something else which
1667	bz	2f				! will be handled later.
1668	nop
1669.mmu_exception_is_tlb_parity:
1670	.weak itlb_parity_trap
1671	.weak dtlb_parity_trap
1672	set	itlb_parity_trap, %g4
1673	cmp	%g1, T_INSTR_EXCEPTION		! branch to the itlb or
1674	be	3f				! dtlb parity handler
1675	nop					! if this trap is due
1676	set	dtlb_parity_trap, %g4
1677	cmp	%g1, T_DATA_EXCEPTION		! to a IMMU exception
1678	be	3f				! or DMMU exception.
1679	nop
16802:
1681	sllx	%g3, 32, %g3
1682	or	%g3, %g1, %g3
1683	set	trap, %g1
1684	ba,pt	%xcc, sys_trap
1685	sub	%g0, 1, %g4
16863:
1687	jmp	%g4				! off to the appropriate
1688	nop					! TLB parity handler
1689
1690.fp_disabled:
1691	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1692	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1693#ifdef SF_ERRATA_30 /* call causes fp-disabled */
1694	brz,a,pn %g1, 2f
1695	  nop
1696#endif
1697	rdpr	%tstate, %g4
1698	btst	TSTATE_PRIV, %g4
1699#ifdef SF_ERRATA_30 /* call causes fp-disabled */
1700	bnz,pn %icc, 2f
1701	  nop
1702#else
1703	bnz,a,pn %icc, ptl1_panic
1704	  mov	PTL1_BAD_FPTRAP, %g1
1705#endif
1706	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1707	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1708	brz,a,pt %g5, 2f
1709	  nop
1710	ldn	[%g5 + P_UTRAP7], %g5			! fp_disabled utrap?
1711	brz,a,pn %g5, 2f
1712	  nop
1713	btst	1, %sp
1714	bz,a,pt	%xcc, 1f				! 32 bit user program
1715	  nop
1716	ba,a,pt	%xcc, .setup_v9utrap			! 64 bit user program
1717	  nop
17181:
1719	ba,pt	%xcc, .setup_utrap
1720	  or	%g0, %g0, %g7
17212:
1722	set	fp_disabled, %g1
1723	ba,pt	%xcc, sys_trap
1724	  sub	%g0, 1, %g4
1725
1726.fp_ieee_exception:
1727	rdpr	%tstate, %g1
1728	btst	TSTATE_PRIV, %g1
1729	bnz,a,pn %icc, ptl1_panic
1730	  mov	PTL1_BAD_FPTRAP, %g1
1731	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1732	stx	%fsr, [%g1 + CPU_TMP1]
1733	ldx	[%g1 + CPU_TMP1], %g2
1734	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1735	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1736	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1737	brz,a,pt %g5, 1f
1738	  nop
1739	ldn	[%g5 + P_UTRAP8], %g5
1740	brnz,a,pt %g5, .setup_v9utrap
1741	  nop
17421:
1743	set	_fp_ieee_exception, %g1
1744	ba,pt	%xcc, sys_trap
1745	  sub	%g0, 1, %g4
1746
1747/*
1748 * Register Inputs:
1749 *	%g5		user trap handler
1750 *	%g7		misaligned addr - for alignment traps only
1751 */
1752.setup_utrap:
1753	set	trap, %g1			! setup in case we go
1754	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1755	sub	%g0, 1, %g4			! the save instruction below
1756
1757	/*
1758	 * If the DTrace pid provider is single stepping a copied-out
1759	 * instruction, t->t_dtrace_step will be set. In that case we need
1760	 * to abort the single-stepping (since execution of the instruction
1761	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1762	 */
1763	save	%sp, -SA(MINFRAME32), %sp	! window for trap handler
1764	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1765	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1766	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1767	rdpr	%tnpc, %l2			! arg1 == tnpc
1768	brz,pt	%g2, 1f
1769	rdpr	%tpc, %l1			! arg0 == tpc
1770
1771	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1772	ldn	[%g1 + T_DTRACE_NPC], %l2	! arg1 = t->t_dtrace_npc (step)
1773	brz,pt	%g2, 1f
1774	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1775	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
17761:
1777	mov	%g7, %l3			! arg2 == misaligned address
1778
1779	rdpr	%tstate, %g1			! cwp for trap handler
1780	rdpr	%cwp, %g4
1781	bclr	TSTATE_CWP_MASK, %g1
1782	wrpr	%g1, %g4, %tstate
1783	wrpr	%g0, %g5, %tnpc			! trap handler address
1784	FAST_TRAP_DONE
1785	/* NOTREACHED */
1786
1787.check_v9utrap:
1788	rdpr	%tstate, %g1
1789	btst	TSTATE_PRIV, %g1
1790	bnz,a,pn %icc, 3f
1791	  nop
1792	CPU_ADDR(%g4, %g1)				! load CPU struct addr
1793	ldn	[%g4 + CPU_THREAD], %g5			! load thread pointer
1794	ldn	[%g5 + T_PROCP], %g5			! load proc pointer
1795	ldn	[%g5 + P_UTRAPS], %g5			! are there utraps?
1796
1797	cmp	%g3, T_SOFTWARE_TRAP
1798	bne,a,pt %icc, 1f
1799	  nop
1800
1801	brz,pt %g5, 3f			! if p_utraps == NULL goto trap()
1802	  rdpr	%tt, %g3		! delay - get actual hw trap type
1803
1804	sub	%g3, 254, %g1		! UT_TRAP_INSTRUCTION_16 = p_utraps[18]
1805	ba,pt	%icc, 2f
1806	  smul	%g1, CPTRSIZE, %g2
18071:
1808	brz,a,pt %g5, 3f		! if p_utraps == NULL goto trap()
1809	  nop
1810
1811	cmp	%g3, T_UNIMP_INSTR
1812	bne,a,pt %icc, 2f
1813	  nop
1814
1815	mov	1, %g1
1816	st	%g1, [%g4 + CPU_TL1_HDLR] ! set CPU_TL1_HDLR
1817	rdpr	%tpc, %g1		! ld trapping instruction using
1818	lduwa	[%g1]ASI_AIUP, %g1	! "AS IF USER" ASI which could fault
1819	st	%g0, [%g4 + CPU_TL1_HDLR] ! clr CPU_TL1_HDLR
1820
1821	sethi	%hi(0xc1c00000), %g4	! setup mask for illtrap instruction
1822	andcc	%g1, %g4, %g4		! and instruction with mask
1823	bnz,a,pt %icc, 3f		! if %g4 == zero, %g1 is an ILLTRAP
1824	  nop				! fall thru to setup
18252:
1826	ldn	[%g5 + %g2], %g5
1827	brnz,a,pt %g5, .setup_v9utrap
1828	  nop
18293:
1830	set	trap, %g1
1831	ba,pt	%xcc, sys_trap
1832	  sub	%g0, 1, %g4
1833	/* NOTREACHED */
1834
1835/*
1836 * Register Inputs:
1837 *	%g5		user trap handler
1838 */
1839.setup_v9utrap:
1840	set	trap, %g1			! setup in case we go
1841	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1842	sub	%g0, 1, %g4			! the save instruction below
1843
1844	/*
1845	 * If the DTrace pid provider is single stepping a copied-out
1846	 * instruction, t->t_dtrace_step will be set. In that case we need
1847	 * to abort the single-stepping (since execution of the instruction
1848	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1849	 */
1850	save	%sp, -SA(MINFRAME64), %sp	! window for trap handler
1851	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1852	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1853	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1854	rdpr	%tnpc, %l7			! arg1 == tnpc
1855	brz,pt	%g2, 1f
1856	rdpr	%tpc, %l6			! arg0 == tpc
1857
1858	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1859	ldn	[%g1 + T_DTRACE_NPC], %l7	! arg1 == t->t_dtrace_npc (step)
1860	brz,pt	%g2, 1f
1861	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1862	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
18631:
1864	rdpr	%tstate, %g2			! cwp for trap handler
1865	rdpr	%cwp, %g4
1866	bclr	TSTATE_CWP_MASK, %g2
1867	wrpr	%g2, %g4, %tstate
1868
1869	ldn	[%g1 + T_PROCP], %g4		! load proc pointer
1870	ldn	[%g4 + P_AS], %g4		! load as pointer
1871	ldn	[%g4 + A_USERLIMIT], %g4	! load as userlimit
1872	cmp	%l7, %g4			! check for single-step set
1873	bne,pt	%xcc, 4f
1874	  nop
1875	ldn	[%g1 + T_LWP], %g1		! load klwp pointer
1876	ld	[%g1 + PCB_STEP], %g4		! load single-step flag
1877	cmp	%g4, STEP_ACTIVE		! step flags set in pcb?
1878	bne,pt	%icc, 4f
1879	  nop
1880	stn	%g5, [%g1 + PCB_TRACEPC]	! save trap handler addr in pcb
1881	mov	%l7, %g4			! on entry to precise user trap
1882	add	%l6, 4, %l7			! handler, %l6 == pc, %l7 == npc
1883						! at time of trap
1884	wrpr	%g0, %g4, %tnpc			! generate FLTBOUNDS,
1885						! %g4 == userlimit
1886	FAST_TRAP_DONE
1887	/* NOTREACHED */
18884:
1889	wrpr	%g0, %g5, %tnpc			! trap handler address
1890	FAST_TRAP_DONE_CHK_INTR
1891	/* NOTREACHED */
1892
1893.fp_exception:
1894	CPU_ADDR(%g1, %g4)
1895	stx	%fsr, [%g1 + CPU_TMP1]
1896	ldx	[%g1 + CPU_TMP1], %g2
1897
1898	/*
1899	 * Cheetah takes unfinished_FPop trap for certain range of operands
1900	 * to the "fitos" instruction. Instead of going through the slow
1901	 * software emulation path, we try to simulate the "fitos" instruction
1902	 * via "fitod" and "fdtos" provided the following conditions are met:
1903	 *
1904	 *	fpu_exists is set (if DEBUG)
1905	 *	not in privileged mode
1906	 *	ftt is unfinished_FPop
1907	 *	NXM IEEE trap is not enabled
1908	 *	instruction at %tpc is "fitos"
1909	 *
1910	 *  Usage:
1911	 *	%g1	per cpu address
1912	 *	%g2	%fsr
1913	 *	%g6	user instruction
1914	 *
1915	 * Note that we can take a memory access related trap while trying
1916	 * to fetch the user instruction. Therefore, we set CPU_TL1_HDLR
1917	 * flag to catch those traps and let the SFMMU code deal with page
1918	 * fault and data access exception.
1919	 */
1920#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
1921	sethi	%hi(fpu_exists), %g7
1922	ld	[%g7 + %lo(fpu_exists)], %g7
1923	brz,pn %g7, .fp_exception_cont
1924	  nop
1925#endif
1926	rdpr	%tstate, %g7			! branch if in privileged mode
1927	btst	TSTATE_PRIV, %g7
1928	bnz,pn	%xcc, .fp_exception_cont
1929	srl	%g2, FSR_FTT_SHIFT, %g7		! extract ftt from %fsr
1930	and	%g7, (FSR_FTT>>FSR_FTT_SHIFT), %g7
1931	cmp	%g7, FTT_UNFIN
1932	set	FSR_TEM_NX, %g5
1933	bne,pn	%xcc, .fp_exception_cont	! branch if NOT unfinished_FPop
1934	  andcc	%g2, %g5, %g0
1935	bne,pn	%xcc, .fp_exception_cont	! branch if FSR_TEM_NX enabled
1936	  rdpr	%tpc, %g5			! get faulting PC
1937
1938	or	%g0, 1, %g7
1939	st	%g7, [%g1 + CPU_TL1_HDLR]	! set tl1_hdlr flag
1940	lda	[%g5]ASI_USER, %g6		! get user's instruction
1941	st	%g0, [%g1 + CPU_TL1_HDLR]	! clear tl1_hdlr flag
1942
1943	set	FITOS_INSTR_MASK, %g7
1944	and	%g6, %g7, %g7
1945	set	FITOS_INSTR, %g5
1946	cmp	%g7, %g5
1947	bne,pn	%xcc, .fp_exception_cont	! branch if not FITOS_INSTR
1948	 nop
1949
1950	/*
1951	 * This is unfinished FPops trap for "fitos" instruction. We
1952	 * need to simulate "fitos" via "fitod" and "fdtos" instruction
1953	 * sequence.
1954	 *
1955	 * We need a temporary FP register to do the conversion. Since
1956	 * both source and destination operands for the "fitos" instruction
1957	 * have to be within %f0-%f31, we use an FP register from the upper
1958	 * half to guarantee that it won't collide with the source or the
1959	 * dest operand. However, we do have to save and restore its value.
1960	 *
1961	 * We use %d62 as a temporary FP register for the conversion and
1962	 * branch to appropriate instruction within the conversion tables
1963	 * based upon the rs2 and rd values.
1964	 */
1965
1966	std	%d62, [%g1 + CPU_TMP1]		! save original value
1967
1968	srl	%g6, FITOS_RS2_SHIFT, %g7
1969	and	%g7, FITOS_REG_MASK, %g7
1970	set	_fitos_fitod_table, %g4
1971	sllx	%g7, 2, %g7
1972	jmp	%g4 + %g7
1973	  ba,pt	%xcc, _fitos_fitod_done
1974	.empty
1975
1976_fitos_fitod_table:
1977	  fitod	%f0, %d62
1978	  fitod	%f1, %d62
1979	  fitod	%f2, %d62
1980	  fitod	%f3, %d62
1981	  fitod	%f4, %d62
1982	  fitod	%f5, %d62
1983	  fitod	%f6, %d62
1984	  fitod	%f7, %d62
1985	  fitod	%f8, %d62
1986	  fitod	%f9, %d62
1987	  fitod	%f10, %d62
1988	  fitod	%f11, %d62
1989	  fitod	%f12, %d62
1990	  fitod	%f13, %d62
1991	  fitod	%f14, %d62
1992	  fitod	%f15, %d62
1993	  fitod	%f16, %d62
1994	  fitod	%f17, %d62
1995	  fitod	%f18, %d62
1996	  fitod	%f19, %d62
1997	  fitod	%f20, %d62
1998	  fitod	%f21, %d62
1999	  fitod	%f22, %d62
2000	  fitod	%f23, %d62
2001	  fitod	%f24, %d62
2002	  fitod	%f25, %d62
2003	  fitod	%f26, %d62
2004	  fitod	%f27, %d62
2005	  fitod	%f28, %d62
2006	  fitod	%f29, %d62
2007	  fitod	%f30, %d62
2008	  fitod	%f31, %d62
2009_fitos_fitod_done:
2010
2011	/*
2012	 * Now convert data back into single precision
2013	 */
2014	srl	%g6, FITOS_RD_SHIFT, %g7
2015	and	%g7, FITOS_REG_MASK, %g7
2016	set	_fitos_fdtos_table, %g4
2017	sllx	%g7, 2, %g7
2018	jmp	%g4 + %g7
2019	  ba,pt	%xcc, _fitos_fdtos_done
2020	.empty
2021
2022_fitos_fdtos_table:
2023	  fdtos	%d62, %f0
2024	  fdtos	%d62, %f1
2025	  fdtos	%d62, %f2
2026	  fdtos	%d62, %f3
2027	  fdtos	%d62, %f4
2028	  fdtos	%d62, %f5
2029	  fdtos	%d62, %f6
2030	  fdtos	%d62, %f7
2031	  fdtos	%d62, %f8
2032	  fdtos	%d62, %f9
2033	  fdtos	%d62, %f10
2034	  fdtos	%d62, %f11
2035	  fdtos	%d62, %f12
2036	  fdtos	%d62, %f13
2037	  fdtos	%d62, %f14
2038	  fdtos	%d62, %f15
2039	  fdtos	%d62, %f16
2040	  fdtos	%d62, %f17
2041	  fdtos	%d62, %f18
2042	  fdtos	%d62, %f19
2043	  fdtos	%d62, %f20
2044	  fdtos	%d62, %f21
2045	  fdtos	%d62, %f22
2046	  fdtos	%d62, %f23
2047	  fdtos	%d62, %f24
2048	  fdtos	%d62, %f25
2049	  fdtos	%d62, %f26
2050	  fdtos	%d62, %f27
2051	  fdtos	%d62, %f28
2052	  fdtos	%d62, %f29
2053	  fdtos	%d62, %f30
2054	  fdtos	%d62, %f31
2055_fitos_fdtos_done:
2056
2057	ldd	[%g1 + CPU_TMP1], %d62		! restore %d62
2058
2059#if DEBUG
2060	/*
2061	 * Update FPop_unfinished trap kstat
2062	 */
2063	set	fpustat+FPUSTAT_UNFIN_KSTAT, %g7
2064	ldx	[%g7], %g5
20651:
2066	add	%g5, 1, %g6
2067
2068	casxa	[%g7] ASI_N, %g5, %g6
2069	cmp	%g5, %g6
2070	bne,a,pn %xcc, 1b
2071	  or	%g0, %g6, %g5
2072
2073	/*
2074	 * Update fpu_sim_fitos kstat
2075	 */
2076	set	fpuinfo+FPUINFO_FITOS_KSTAT, %g7
2077	ldx	[%g7], %g5
20781:
2079	add	%g5, 1, %g6
2080
2081	casxa	[%g7] ASI_N, %g5, %g6
2082	cmp	%g5, %g6
2083	bne,a,pn %xcc, 1b
2084	  or	%g0, %g6, %g5
2085#endif /* DEBUG */
2086
2087	FAST_TRAP_DONE
2088
2089.fp_exception_cont:
2090	/*
2091	 * Let _fp_exception deal with simulating FPop instruction.
2092	 * Note that we need to pass %fsr in %g2 (already read above).
2093	 */
2094
2095	set	_fp_exception, %g1
2096	ba,pt	%xcc, sys_trap
2097	sub	%g0, 1, %g4
2098
2099	.global opl_cleanw_patch
2100opl_cleanw_patch:
2101.clean_windows:
2102	set	trap, %g1
2103	mov	T_FLUSH_PCB, %g3
2104	sub	%g0, 1, %g4
2105	save
2106	flushw
2107	restore
2108	wrpr	%g0, %g0, %cleanwin	! no clean windows
2109
2110	CPU_ADDR(%g4, %g5)
2111	ldn	[%g4 + CPU_MPCB], %g4
2112	brz,a,pn %g4, 1f
2113	  nop
2114	ld	[%g4 + MPCB_WSTATE], %g5
2115	add	%g5, WSTATE_CLEAN_OFFSET, %g5
2116	wrpr	%g0, %g5, %wstate
21171:	FAST_TRAP_DONE
2118
2119/*
2120 * .spill_clean: clean the previous window, restore the wstate, and
2121 * "done".
2122 *
2123 * Entry: %g7 contains new wstate
2124 */
2125.spill_clean:
2126	sethi	%hi(nwin_minus_one), %g5
2127	ld	[%g5 + %lo(nwin_minus_one)], %g5 ! %g5 = nwin - 1
2128	rdpr	%cwp, %g6			! %g6 = %cwp
2129	deccc	%g6				! %g6--
2130	movneg	%xcc, %g5, %g6			! if (%g6<0) %g6 = nwin-1
2131	wrpr	%g6, %cwp
2132	TT_TRACE_L(trace_win)
2133	clr	%l0
2134	clr	%l1
2135	clr	%l2
2136	clr	%l3
2137	clr	%l4
2138	clr	%l5
2139	clr	%l6
2140	clr	%l7
2141	wrpr	%g0, %g7, %wstate
2142	saved
2143	retry			! restores correct %cwp
2144
2145.fix_alignment:
2146	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2147	ldn	[%g1 + CPU_THREAD], %g1	! load thread pointer
2148	ldn	[%g1 + T_PROCP], %g1
2149	mov	1, %g2
2150	stb	%g2, [%g1 + P_FIXALIGNMENT]
2151	FAST_TRAP_DONE
2152
2153#define	STDF_REG(REG, ADDR, TMP)		\
2154	sll	REG, 3, REG			;\
2155mark1:	set	start1, TMP			;\
2156	jmp	REG + TMP			;\
2157	  nop					;\
2158start1:	ba,pt	%xcc, done1			;\
2159	  std	%f0, [ADDR + CPU_TMP1]		;\
2160	ba,pt	%xcc, done1			;\
2161	  std	%f32, [ADDR + CPU_TMP1]		;\
2162	ba,pt	%xcc, done1			;\
2163	  std	%f2, [ADDR + CPU_TMP1]		;\
2164	ba,pt	%xcc, done1			;\
2165	  std	%f34, [ADDR + CPU_TMP1]		;\
2166	ba,pt	%xcc, done1			;\
2167	  std	%f4, [ADDR + CPU_TMP1]		;\
2168	ba,pt	%xcc, done1			;\
2169	  std	%f36, [ADDR + CPU_TMP1]		;\
2170	ba,pt	%xcc, done1			;\
2171	  std	%f6, [ADDR + CPU_TMP1]		;\
2172	ba,pt	%xcc, done1			;\
2173	  std	%f38, [ADDR + CPU_TMP1]		;\
2174	ba,pt	%xcc, done1			;\
2175	  std	%f8, [ADDR + CPU_TMP1]		;\
2176	ba,pt	%xcc, done1			;\
2177	  std	%f40, [ADDR + CPU_TMP1]		;\
2178	ba,pt	%xcc, done1			;\
2179	  std	%f10, [ADDR + CPU_TMP1]		;\
2180	ba,pt	%xcc, done1			;\
2181	  std	%f42, [ADDR + CPU_TMP1]		;\
2182	ba,pt	%xcc, done1			;\
2183	  std	%f12, [ADDR + CPU_TMP1]		;\
2184	ba,pt	%xcc, done1			;\
2185	  std	%f44, [ADDR + CPU_TMP1]		;\
2186	ba,pt	%xcc, done1			;\
2187	  std	%f14, [ADDR + CPU_TMP1]		;\
2188	ba,pt	%xcc, done1			;\
2189	  std	%f46, [ADDR + CPU_TMP1]		;\
2190	ba,pt	%xcc, done1			;\
2191	  std	%f16, [ADDR + CPU_TMP1]		;\
2192	ba,pt	%xcc, done1			;\
2193	  std	%f48, [ADDR + CPU_TMP1]		;\
2194	ba,pt	%xcc, done1			;\
2195	  std	%f18, [ADDR + CPU_TMP1]		;\
2196	ba,pt	%xcc, done1			;\
2197	  std	%f50, [ADDR + CPU_TMP1]		;\
2198	ba,pt	%xcc, done1			;\
2199	  std	%f20, [ADDR + CPU_TMP1]		;\
2200	ba,pt	%xcc, done1			;\
2201	  std	%f52, [ADDR + CPU_TMP1]		;\
2202	ba,pt	%xcc, done1			;\
2203	  std	%f22, [ADDR + CPU_TMP1]		;\
2204	ba,pt	%xcc, done1			;\
2205	  std	%f54, [ADDR + CPU_TMP1]		;\
2206	ba,pt	%xcc, done1			;\
2207	  std	%f24, [ADDR + CPU_TMP1]		;\
2208	ba,pt	%xcc, done1			;\
2209	  std	%f56, [ADDR + CPU_TMP1]		;\
2210	ba,pt	%xcc, done1			;\
2211	  std	%f26, [ADDR + CPU_TMP1]		;\
2212	ba,pt	%xcc, done1			;\
2213	  std	%f58, [ADDR + CPU_TMP1]		;\
2214	ba,pt	%xcc, done1			;\
2215	  std	%f28, [ADDR + CPU_TMP1]		;\
2216	ba,pt	%xcc, done1			;\
2217	  std	%f60, [ADDR + CPU_TMP1]		;\
2218	ba,pt	%xcc, done1			;\
2219	  std	%f30, [ADDR + CPU_TMP1]		;\
2220	ba,pt	%xcc, done1			;\
2221	  std	%f62, [ADDR + CPU_TMP1]		;\
2222done1:
2223
2224#define	LDDF_REG(REG, ADDR, TMP)		\
2225	sll	REG, 3, REG			;\
2226mark2:	set	start2, TMP			;\
2227	jmp	REG + TMP			;\
2228	  nop					;\
2229start2:	ba,pt	%xcc, done2			;\
2230	  ldd	[ADDR + CPU_TMP1], %f0		;\
2231	ba,pt	%xcc, done2			;\
2232	  ldd	[ADDR + CPU_TMP1], %f32		;\
2233	ba,pt	%xcc, done2			;\
2234	  ldd	[ADDR + CPU_TMP1], %f2		;\
2235	ba,pt	%xcc, done2			;\
2236	  ldd	[ADDR + CPU_TMP1], %f34		;\
2237	ba,pt	%xcc, done2			;\
2238	  ldd	[ADDR + CPU_TMP1], %f4		;\
2239	ba,pt	%xcc, done2			;\
2240	  ldd	[ADDR + CPU_TMP1], %f36		;\
2241	ba,pt	%xcc, done2			;\
2242	  ldd	[ADDR + CPU_TMP1], %f6		;\
2243	ba,pt	%xcc, done2			;\
2244	  ldd	[ADDR + CPU_TMP1], %f38		;\
2245	ba,pt	%xcc, done2			;\
2246	  ldd	[ADDR + CPU_TMP1], %f8		;\
2247	ba,pt	%xcc, done2			;\
2248	  ldd	[ADDR + CPU_TMP1], %f40		;\
2249	ba,pt	%xcc, done2			;\
2250	  ldd	[ADDR + CPU_TMP1], %f10		;\
2251	ba,pt	%xcc, done2			;\
2252	  ldd	[ADDR + CPU_TMP1], %f42		;\
2253	ba,pt	%xcc, done2			;\
2254	  ldd	[ADDR + CPU_TMP1], %f12		;\
2255	ba,pt	%xcc, done2			;\
2256	  ldd	[ADDR + CPU_TMP1], %f44		;\
2257	ba,pt	%xcc, done2			;\
2258	  ldd	[ADDR + CPU_TMP1], %f14		;\
2259	ba,pt	%xcc, done2			;\
2260	  ldd	[ADDR + CPU_TMP1], %f46		;\
2261	ba,pt	%xcc, done2			;\
2262	  ldd	[ADDR + CPU_TMP1], %f16		;\
2263	ba,pt	%xcc, done2			;\
2264	  ldd	[ADDR + CPU_TMP1], %f48		;\
2265	ba,pt	%xcc, done2			;\
2266	  ldd	[ADDR + CPU_TMP1], %f18		;\
2267	ba,pt	%xcc, done2			;\
2268	  ldd	[ADDR + CPU_TMP1], %f50		;\
2269	ba,pt	%xcc, done2			;\
2270	  ldd	[ADDR + CPU_TMP1], %f20		;\
2271	ba,pt	%xcc, done2			;\
2272	  ldd	[ADDR + CPU_TMP1], %f52		;\
2273	ba,pt	%xcc, done2			;\
2274	  ldd	[ADDR + CPU_TMP1], %f22		;\
2275	ba,pt	%xcc, done2			;\
2276	  ldd	[ADDR + CPU_TMP1], %f54		;\
2277	ba,pt	%xcc, done2			;\
2278	  ldd	[ADDR + CPU_TMP1], %f24		;\
2279	ba,pt	%xcc, done2			;\
2280	  ldd	[ADDR + CPU_TMP1], %f56		;\
2281	ba,pt	%xcc, done2			;\
2282	  ldd	[ADDR + CPU_TMP1], %f26		;\
2283	ba,pt	%xcc, done2			;\
2284	  ldd	[ADDR + CPU_TMP1], %f58		;\
2285	ba,pt	%xcc, done2			;\
2286	  ldd	[ADDR + CPU_TMP1], %f28		;\
2287	ba,pt	%xcc, done2			;\
2288	  ldd	[ADDR + CPU_TMP1], %f60		;\
2289	ba,pt	%xcc, done2			;\
2290	  ldd	[ADDR + CPU_TMP1], %f30		;\
2291	ba,pt	%xcc, done2			;\
2292	  ldd	[ADDR + CPU_TMP1], %f62		;\
2293done2:
2294
2295.lddf_exception_not_aligned:
2296	/*
2297	 * Cheetah overwrites SFAR on a DTLB miss, hence read it now.
2298	 */
2299	ldxa	[MMU_SFAR]%asi, %g5	! misaligned vaddr in %g5
2300
2301#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2302	sethi	%hi(fpu_exists), %g2		! check fpu_exists
2303	ld	[%g2 + %lo(fpu_exists)], %g2
2304	brz,a,pn %g2, 4f
2305	  nop
2306#endif
2307	CPU_ADDR(%g1, %g4)
2308	or	%g0, 1, %g4
2309	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2310
2311	rdpr	%tpc, %g2
2312	lda	[%g2]ASI_AIUP, %g6	! get the user's lddf instruction
2313	srl	%g6, 23, %g1		! using ldda or not?
2314	and	%g1, 1, %g1
2315	brz,a,pt %g1, 2f		! check for ldda instruction
2316	  nop
2317	srl	%g6, 13, %g1		! check immflag
2318	and	%g1, 1, %g1
2319	rdpr	%tstate, %g2		! %tstate in %g2
2320	brnz,a,pn %g1, 1f
2321	  srl	%g2, 31, %g1		! get asi from %tstate
2322	srl	%g6, 5, %g1		! get asi from instruction
2323	and	%g1, 0xFF, %g1		! imm_asi field
23241:
2325	cmp	%g1, ASI_P		! primary address space
2326	be,a,pt %icc, 2f
2327	  nop
2328	cmp	%g1, ASI_PNF		! primary no fault address space
2329	be,a,pt %icc, 2f
2330	  nop
2331	cmp	%g1, ASI_S		! secondary address space
2332	be,a,pt %icc, 2f
2333	  nop
2334	cmp	%g1, ASI_SNF		! secondary no fault address space
2335	bne,a,pn %icc, 3f
2336	  nop
23372:
2338	lduwa	[%g5]ASI_USER, %g7	! get first half of misaligned data
2339	add	%g5, 4, %g5		! increment misaligned data address
2340	lduwa	[%g5]ASI_USER, %g5	! get second half of misaligned data
2341
2342	sllx	%g7, 32, %g7
2343	or	%g5, %g7, %g5		! combine data
2344	CPU_ADDR(%g7, %g1)		! save data on a per-cpu basis
2345	stx	%g5, [%g7 + CPU_TMP1]	! save in cpu_tmp1
2346
2347	srl	%g6, 25, %g3		! %g6 has the instruction
2348	and	%g3, 0x1F, %g3		! %g3 has rd
2349	LDDF_REG(%g3, %g7, %g4)
2350
2351	CPU_ADDR(%g1, %g4)
2352	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2353	FAST_TRAP_DONE
23543:
2355	CPU_ADDR(%g1, %g4)
2356	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
23574:
2358	set	T_USER, %g3		! trap type in %g3
2359	or	%g3, T_LDDF_ALIGN, %g3
2360	mov	%g5, %g2		! misaligned vaddr in %g2
2361	set	fpu_trap, %g1		! goto C for the little and
2362	ba,pt	%xcc, sys_trap		! no fault little asi's
2363	  sub	%g0, 1, %g4
2364
2365.stdf_exception_not_aligned:
2366	/*
2367	 * Cheetah overwrites SFAR on a DTLB miss, hence read it now.
2368	 */
2369	ldxa	[MMU_SFAR]%asi, %g5	! misaligned vaddr in %g5
2370
2371#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2372	sethi	%hi(fpu_exists), %g7		! check fpu_exists
2373	ld	[%g7 + %lo(fpu_exists)], %g3
2374	brz,a,pn %g3, 4f
2375	  nop
2376#endif
2377	CPU_ADDR(%g1, %g4)
2378	or	%g0, 1, %g4
2379	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2380
2381	rdpr	%tpc, %g2
2382	lda	[%g2]ASI_AIUP, %g6	! get the user's stdf instruction
2383
2384	srl	%g6, 23, %g1		! using stda or not?
2385	and	%g1, 1, %g1
2386	brz,a,pt %g1, 2f		! check for stda instruction
2387	  nop
2388	srl	%g6, 13, %g1		! check immflag
2389	and	%g1, 1, %g1
2390	rdpr	%tstate, %g2		! %tstate in %g2
2391	brnz,a,pn %g1, 1f
2392	  srl	%g2, 31, %g1		! get asi from %tstate
2393	srl	%g6, 5, %g1		! get asi from instruction
2394	and	%g1, 0xFF, %g1		! imm_asi field
23951:
2396	cmp	%g1, ASI_P		! primary address space
2397	be,a,pt %icc, 2f
2398	  nop
2399	cmp	%g1, ASI_S		! secondary address space
2400	bne,a,pn %icc, 3f
2401	  nop
24022:
2403	srl	%g6, 25, %g6
2404	and	%g6, 0x1F, %g6		! %g6 has rd
2405	CPU_ADDR(%g7, %g1)
2406	STDF_REG(%g6, %g7, %g4)		! STDF_REG(REG, ADDR, TMP)
2407
2408	ldx	[%g7 + CPU_TMP1], %g6
2409	srlx	%g6, 32, %g7
2410	stuwa	%g7, [%g5]ASI_USER	! first half
2411	add	%g5, 4, %g5		! increment misaligned data address
2412	stuwa	%g6, [%g5]ASI_USER	! second half
2413
2414	CPU_ADDR(%g1, %g4)
2415	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2416	FAST_TRAP_DONE
24173:
2418	CPU_ADDR(%g1, %g4)
2419	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
24204:
2421	set	T_USER, %g3		! trap type in %g3
2422	or	%g3, T_STDF_ALIGN, %g3
2423	mov	%g5, %g2		! misaligned vaddr in %g2
2424	set	fpu_trap, %g1		! goto C for the little and
2425	ba,pt	%xcc, sys_trap		! nofault little asi's
2426	  sub	%g0, 1, %g4
2427
2428#ifdef DEBUG_USER_TRAPTRACECTL
2429
2430.traptrace_freeze:
2431	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2432	TT_TRACE_L(trace_win)
2433	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2434	set	trap_freeze, %g1
2435	mov	1, %g2
2436	st	%g2, [%g1]
2437	FAST_TRAP_DONE
2438
2439.traptrace_unfreeze:
2440	set	trap_freeze, %g1
2441	st	%g0, [%g1]
2442	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2443	TT_TRACE_L(trace_win)
2444	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2445	FAST_TRAP_DONE
2446
2447#endif /* DEBUG_USER_TRAPTRACECTL */
2448
2449.getcc:
2450	CPU_ADDR(%g1, %g2)
2451	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2452	stx	%o1, [%g1 + CPU_TMP2]		! save %o1
2453	rdpr	%tstate, %g3			! get tstate
2454	srlx	%g3, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2455	set	PSR_ICC, %g2
2456	and	%o0, %g2, %o0			! mask out the rest
2457	srl	%o0, PSR_ICC_SHIFT, %o0		! right justify
2458	rdpr	%pstate, %o1
2459	wrpr	%o1, PSTATE_AG, %pstate		! get into normal globals
2460	mov	%o0, %g1			! move ccr to normal %g1
2461	wrpr	%g0, %o1, %pstate		! back into alternate globals
2462	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2463	ldx	[%g1 + CPU_TMP2], %o1		! restore %o1
2464	FAST_TRAP_DONE
2465
2466.setcc:
2467	CPU_ADDR(%g1, %g2)
2468	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2469	stx	%o1, [%g1 + CPU_TMP2]		! save %o1
2470	rdpr	%pstate, %o0
2471	wrpr	%o0, PSTATE_AG, %pstate		! get into normal globals
2472	mov	%g1, %o1
2473	wrpr	%g0, %o0, %pstate		! back to alternates
2474	sll	%o1, PSR_ICC_SHIFT, %g2
2475	set	PSR_ICC, %g3
2476	and	%g2, %g3, %g2			! mask out rest
2477	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g2
2478	rdpr	%tstate, %g3			! get tstate
2479	srl	%g3, 0, %g3			! clear upper word
2480	or	%g3, %g2, %g3			! or in new bits
2481	wrpr	%g3, %tstate
2482	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2483	ldx	[%g1 + CPU_TMP2], %o1		! restore %o1
2484	FAST_TRAP_DONE
2485
2486/*
2487 * getpsr(void)
2488 * Note that the xcc part of the ccr is not provided.
2489 * The V8 code shows why the V9 trap is not faster:
2490 * #define GETPSR_TRAP() \
2491 *      mov %psr, %i0; jmp %l2; rett %l2+4; nop;
2492 */
2493
2494	.type	.getpsr, #function
2495.getpsr:
2496	rdpr	%tstate, %g1			! get tstate
2497	srlx	%g1, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2498	set	PSR_ICC, %g2
2499	and	%o0, %g2, %o0			! mask out the rest
2500
2501	rd	%fprs, %g1			! get fprs
2502	and	%g1, FPRS_FEF, %g2		! mask out dirty upper/lower
2503	sllx	%g2, PSR_FPRS_FEF_SHIFT, %g2	! shift fef to V8 psr.ef
2504	or	%o0, %g2, %o0			! or result into psr.ef
2505
2506	set	V9_PSR_IMPLVER, %g2		! SI assigned impl/ver: 0xef
2507	or	%o0, %g2, %o0			! or psr.impl/ver
2508	FAST_TRAP_DONE
2509	SET_SIZE(.getpsr)
2510
2511/*
2512 * setpsr(newpsr)
2513 * Note that there is no support for ccr.xcc in the V9 code.
2514 */
2515
2516	.type	.setpsr, #function
2517.setpsr:
2518	rdpr	%tstate, %g1			! get tstate
2519!	setx	TSTATE_V8_UBITS, %g2
2520	or 	%g0, CCR_ICC, %g3
2521	sllx	%g3, TSTATE_CCR_SHIFT, %g2
2522
2523	andn	%g1, %g2, %g1			! zero current user bits
2524	set	PSR_ICC, %g2
2525	and	%g2, %o0, %g2			! clear all but psr.icc bits
2526	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g3	! shift to tstate.ccr.icc
2527	wrpr	%g1, %g3, %tstate		! write tstate
2528
2529	set	PSR_EF, %g2
2530	and	%g2, %o0, %g2			! clear all but fp enable bit
2531	srlx	%g2, PSR_FPRS_FEF_SHIFT, %g4	! shift ef to V9 fprs.fef
2532	wr	%g0, %g4, %fprs			! write fprs
2533
2534	CPU_ADDR(%g1, %g2)			! load CPU struct addr to %g1
2535	ldn	[%g1 + CPU_THREAD], %g2		! load thread pointer
2536	ldn	[%g2 + T_LWP], %g3		! load klwp pointer
2537	ldn	[%g3 + LWP_FPU], %g2		! get lwp_fpu pointer
2538	stuw	%g4, [%g2 + FPU_FPRS]		! write fef value to fpu_fprs
2539	srlx	%g4, 2, %g4			! shift fef value to bit 0
2540	stub	%g4, [%g2 + FPU_EN]		! write fef value to fpu_en
2541	FAST_TRAP_DONE
2542	SET_SIZE(.setpsr)
2543
2544/*
2545 * getlgrp
2546 * get home lgrpid on which the calling thread is currently executing.
2547 */
2548	.type	.getlgrp, #function
2549.getlgrp:
2550	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2551	ld	[%g1 + CPU_ID], %o0	! load cpu_id
2552	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2553	ldn	[%g2 + T_LPL], %g2	! load lpl pointer
2554	ld	[%g2 + LPL_LGRPID], %g1	! load lpl_lgrpid
2555	sra	%g1, 0, %o1
2556	FAST_TRAP_DONE
2557	SET_SIZE(.getlgrp)
2558
2559/*
2560 * Entry for old 4.x trap (trap 0).
2561 */
2562	ENTRY_NP(syscall_trap_4x)
2563	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2564	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2565	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2566	ld	[%g2 + PCB_TRAP0], %g2	! lwp->lwp_pcb.pcb_trap0addr
2567	brz,pn	%g2, 1f			! has it been set?
2568	st	%l0, [%g1 + CPU_TMP1]	! delay - save some locals
2569	st	%l1, [%g1 + CPU_TMP2]
2570	rdpr	%tnpc, %l1		! save old tnpc
2571	wrpr	%g0, %g2, %tnpc		! setup tnpc
2572
2573	rdpr	%pstate, %l0
2574	wrpr	%l0, PSTATE_AG, %pstate	! switch to normal globals
2575	mov	%l1, %g6		! pass tnpc to user code in %g6
2576	wrpr	%l0, %g0, %pstate	! switch back to alternate globals
2577
2578	! Note that %g1 still contains CPU struct addr
2579	ld	[%g1 + CPU_TMP2], %l1	! restore locals
2580	ld	[%g1 + CPU_TMP1], %l0
2581	FAST_TRAP_DONE_CHK_INTR
25821:
2583	mov	%g1, %l0
2584	st	%l1, [%g1 + CPU_TMP2]
2585	rdpr	%pstate, %l1
2586	wrpr	%l1, PSTATE_AG, %pstate
2587	!
2588	! check for old syscall mmap which is the only different one which
2589	! must be the same.  Others are handled in the compatibility library.
2590	!
2591	cmp	%g1, OSYS_mmap	! compare to old 4.x mmap
2592	movz	%icc, SYS_mmap, %g1
2593	wrpr	%g0, %l1, %pstate
2594	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2595	ld	[%l0 + CPU_TMP1], %l0
2596	SYSCALL(syscall_trap32)
2597	SET_SIZE(syscall_trap_4x)
2598
2599/*
2600 * Handler for software trap 9.
2601 * Set trap0 emulation address for old 4.x system call trap.
2602 * XXX - this should be a system call.
2603 */
2604	ENTRY_NP(set_trap0_addr)
2605	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2606	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2607	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2608	st	%l0, [%g1 + CPU_TMP1]	! save some locals
2609	st	%l1, [%g1 + CPU_TMP2]
2610	rdpr	%pstate, %l0
2611	wrpr	%l0, PSTATE_AG, %pstate
2612	mov	%g1, %l1
2613	wrpr	%g0, %l0, %pstate
2614	andn	%l1, 3, %l1		! force alignment
2615	st	%l1, [%g2 + PCB_TRAP0]	! lwp->lwp_pcb.pcb_trap0addr
2616	ld	[%g1 + CPU_TMP1], %l0	! restore locals
2617	ld	[%g1 + CPU_TMP2], %l1
2618	FAST_TRAP_DONE
2619	SET_SIZE(set_trap0_addr)
2620
2621/*
2622 * mmu_trap_tl1
2623 * trap handler for unexpected mmu traps.
2624 * simply checks if the trap was a user lddf/stdf alignment trap, in which
2625 * case we go to fpu_trap or a user trap from the window handler, in which
2626 * case we go save the state on the pcb.  Otherwise, we go to ptl1_panic.
2627 */
2628	.type	mmu_trap_tl1, #function
2629mmu_trap_tl1:
2630#ifdef	TRAPTRACE
2631	TRACE_PTR(%g5, %g6)
2632	GET_TRACE_TICK(%g6)
2633	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2634	rdpr	%tl, %g6
2635	stha	%g6, [%g5 + TRAP_ENT_TL]%asi
2636	rdpr	%tt, %g6
2637	stha	%g6, [%g5 + TRAP_ENT_TT]%asi
2638	rdpr	%tstate, %g6
2639	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi
2640	stna	%sp, [%g5 + TRAP_ENT_SP]%asi
2641	stna	%g0, [%g5 + TRAP_ENT_TR]%asi
2642	rdpr	%tpc, %g6
2643	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2644	set	MMU_SFAR, %g6
2645	ldxa	[%g6]ASI_DMMU, %g6
2646	stxa	%g6, [%g5 + TRAP_ENT_F1]%asi
2647	CPU_PADDR(%g7, %g6);
2648	add	%g7, CPU_TL1_HDLR, %g7
2649	lda	[%g7]ASI_MEM, %g6
2650	stxa	%g6, [%g5 + TRAP_ENT_F2]%asi
2651	set	0xdeadbeef, %g6
2652	stna	%g6, [%g5 + TRAP_ENT_F3]%asi
2653	stna	%g6, [%g5 + TRAP_ENT_F4]%asi
2654	TRACE_NEXT(%g5, %g6, %g7)
2655#endif /* TRAPTRACE */
2656
2657	GET_CPU_IMPL(%g5)
2658	cmp	%g5, PANTHER_IMPL
2659	bne	mmu_trap_tl1_4
2660	  nop
2661	rdpr	%tt, %g5
2662	cmp	%g5, T_DATA_EXCEPTION
2663	bne	mmu_trap_tl1_4
2664	  nop
2665	wr	%g0, ASI_DMMU, %asi
2666	ldxa	[MMU_SFSR]%asi, %g5
2667	mov	1, %g6
2668	sllx	%g6, PN_SFSR_PARITY_SHIFT, %g6
2669	andcc	%g5, %g6, %g0
2670	bz	mmu_trap_tl1_4
2671
2672	/*
2673	 * We are running on a Panther and have hit a DTLB parity error.
2674	 */
2675	ldxa	[MMU_TAG_ACCESS]%asi, %g2
2676	mov	%g5, %g3
2677	ba,pt	%xcc, .mmu_exception_is_tlb_parity
2678	mov	T_DATA_EXCEPTION, %g1
2679
2680mmu_trap_tl1_4:
2681	CPU_PADDR(%g7, %g6);
2682	add     %g7, CPU_TL1_HDLR, %g7		! %g7 = &cpu_m.tl1_hdlr (PA)
2683	/*
2684	 * AM is cleared on trap, so addresses are 64 bit
2685	 */
2686	lda     [%g7]ASI_MEM, %g6
2687	brz,a,pt %g6, 1f
2688	  nop
2689	/*
2690	 * We are going to update cpu_m.tl1_hdlr using physical address.
2691	 * Flush the D$ line, so that stale data won't be accessed later.
2692	 */
2693	CPU_ADDR(%g6, %g5)
2694	add     %g6, CPU_TL1_HDLR, %g6		! %g6 = &cpu_m.tl1_hdlr (VA)
2695	GET_CPU_IMPL(%g5)
2696	cmp	%g5, CHEETAH_IMPL
2697	bl,pt	%icc, 3f
2698	 cmp	%g5, SPITFIRE_IMPL
2699	stxa	%g0, [%g7]ASI_DC_INVAL
2700	membar	#Sync
2701	ba,pt	%xcc, 2f
2702	 nop
27033:
2704	bl,pt	%icc, 2f
2705	 sethi	%hi(dcache_line_mask), %g5
2706	ld	[%g5 + %lo(dcache_line_mask)], %g5
2707	and	%g6, %g5, %g5
2708	stxa	%g0, [%g5]ASI_DC_TAG
2709	membar	#Sync
27102:
2711	sta     %g0, [%g7]ASI_MEM
2712	SWITCH_GLOBALS				! back to mmu globals
2713	ba,a,pt	%xcc, sfmmu_mmu_trap		! handle page faults
27141:
2715	rdpr	%tt, %g5
2716	rdpr	%tl, %g7
2717	sub	%g7, 1, %g6
2718	wrpr	%g6, %tl
2719	rdpr	%tt, %g6
2720	wrpr	%g7, %tl
2721	and	%g6, WTRAP_TTMASK, %g6
2722	cmp	%g6, WTRAP_TYPE
2723	bne,a,pn %xcc, ptl1_panic
2724	mov	PTL1_BAD_MMUTRAP, %g1
2725	rdpr	%tpc, %g7
2726	/* tpc should be in the trap table */
2727	set	trap_table, %g6
2728	cmp	%g7, %g6
2729	blt,a,pn %xcc, ptl1_panic
2730	  mov	PTL1_BAD_MMUTRAP, %g1
2731	set	etrap_table, %g6
2732	cmp	%g7, %g6
2733	bge,a,pn %xcc, ptl1_panic
2734	  mov	PTL1_BAD_MMUTRAP, %g1
2735	cmp	%g5, T_ALIGNMENT
2736	move	%icc, MMU_SFAR, %g6
2737	movne	%icc, MMU_TAG_ACCESS, %g6
2738	ldxa	[%g6]ASI_DMMU, %g6
2739	andn	%g7, WTRAP_ALIGN, %g7	/* 128 byte aligned */
2740	add	%g7, WTRAP_FAULTOFF, %g7
2741	wrpr	%g0, %g7, %tnpc
2742	done
2743	SET_SIZE(mmu_trap_tl1)
2744
2745/*
2746 * Several traps use kmdb_trap and kmdb_trap_tl1 as their handlers.  These
2747 * traps are valid only when kmdb is loaded.  When the debugger is active,
2748 * the code below is rewritten to transfer control to the appropriate
2749 * debugger entry points.
2750 */
2751	.global	kmdb_trap
2752	.align	8
2753kmdb_trap:
2754	ba,a	trap_table0
2755	jmp	%g1 + 0
2756	nop
2757
2758	.global	kmdb_trap_tl1
2759	.align	8
2760kmdb_trap_tl1:
2761	ba,a	trap_table0
2762	jmp	%g1 + 0
2763	nop
2764
2765/*
2766 * This entry is copied from OBP's trap table during boot.
2767 */
2768	.global	obp_bpt
2769	.align	8
2770obp_bpt:
2771	NOT
2772
2773/*
2774 * if kernel, set PCONTEXT to 0 for debuggers
2775 * if user, clear nucleus page sizes
2776 */
2777	.global kctx_obp_bpt
2778kctx_obp_bpt:
2779	set	obp_bpt, %g2
27801:
2781#ifndef _OPL
2782	mov	MMU_PCONTEXT, %g1
2783	ldxa	[%g1]ASI_DMMU, %g1
2784	srlx	%g1, CTXREG_NEXT_SHIFT, %g3
2785	brz,pt	%g3, 3f			! nucleus pgsz is 0, no problem
2786	  sllx	%g3, CTXREG_NEXT_SHIFT, %g3
2787	set	CTXREG_CTX_MASK, %g4	! check Pcontext
2788	btst	%g4, %g1
2789	bz,a,pt	%xcc, 2f
2790	  clr	%g3			! kernel:  PCONTEXT=0
2791	xor	%g3, %g1, %g3		! user:	clr N_pgsz0/1 bits
27922:
2793	set	DEMAP_ALL_TYPE, %g1
2794	stxa	%g0, [%g1]ASI_DTLB_DEMAP
2795	stxa	%g0, [%g1]ASI_ITLB_DEMAP
2796	mov	MMU_PCONTEXT, %g1
2797	stxa	%g3, [%g1]ASI_DMMU
2798        membar  #Sync
2799	sethi	%hi(FLUSH_ADDR), %g1
2800	flush	%g1			! flush required by immu
2801#endif /* _OPL */
28023:
2803	jmp	%g2
2804	  nop
2805
2806
2807#ifdef	TRAPTRACE
2808/*
2809 * TRAPTRACE support.
2810 * labels here are branched to with "rd %pc, %g7" in the delay slot.
2811 * Return is done by "jmp %g7 + 4".
2812 */
2813
2814trace_gen:
2815	TRACE_PTR(%g3, %g6)
2816	GET_TRACE_TICK(%g6)
2817	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2818	rdpr	%tl, %g6
2819	stha	%g6, [%g3 + TRAP_ENT_TL]%asi
2820	rdpr	%tt, %g6
2821	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2822	rdpr	%tstate, %g6
2823	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2824	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2825	rdpr	%tpc, %g6
2826	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2827	TRACE_NEXT(%g3, %g4, %g5)
2828	jmp	%g7 + 4
2829	nop
2830
2831trace_win:
2832	TRACE_WIN_INFO(0, %l0, %l1, %l2)
2833	! Keep the locals as clean as possible, caller cleans %l4
2834	clr	%l2
2835	clr	%l1
2836	jmp	%l4 + 4
2837	  clr	%l0
2838
2839/*
2840 * Trace a tsb hit
2841 * g1 = tsbe pointer (in/clobbered)
2842 * g2 = tag access register (in)
2843 * g3 - g4 = scratch (clobbered)
2844 * g5 = tsbe data (in)
2845 * g6 = scratch (clobbered)
2846 * g7 = pc we jumped here from (in)
2847 */
2848
2849	! Do not disturb %g5, it will be used after the trace
2850	ALTENTRY(trace_tsbhit)
2851	TRACE_TSBHIT(0)
2852	jmp	%g7 + 4
2853	nop
2854
2855/*
2856 * Trace a TSB miss
2857 *
2858 * g1 = tsb8k pointer (in)
2859 * g2 = tag access register (in)
2860 * g3 = tsb4m pointer (in)
2861 * g4 = tsbe tag (in/clobbered)
2862 * g5 - g6 = scratch (clobbered)
2863 * g7 = pc we jumped here from (in)
2864 */
2865	.global	trace_tsbmiss
2866trace_tsbmiss:
2867	membar	#Sync
2868	sethi	%hi(FLUSH_ADDR), %g6
2869	flush	%g6
2870	TRACE_PTR(%g5, %g6)
2871	GET_TRACE_TICK(%g6)
2872	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2873	stxa	%g2, [%g5 + TRAP_ENT_SP]%asi		! tag access
2874	stxa	%g4, [%g5 + TRAP_ENT_F1]%asi		! tsb tag
2875	rdpr	%tnpc, %g6
2876	stxa	%g6, [%g5 + TRAP_ENT_F2]%asi
2877	stna	%g1, [%g5 + TRAP_ENT_F3]%asi		! tsb8k pointer
2878	srlx	%g1, 32, %g6
2879	stna	%g6, [%g5 + TRAP_ENT_F4]%asi		! huh?
2880	rdpr	%tpc, %g6
2881	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2882	rdpr	%tl, %g6
2883	stha	%g6, [%g5 + TRAP_ENT_TL]%asi
2884	rdpr	%tt, %g6
2885	or	%g6, TT_MMU_MISS, %g4
2886	stha	%g4, [%g5 + TRAP_ENT_TT]%asi
2887	cmp	%g6, FAST_IMMU_MISS_TT
2888	be,a	%icc, 1f
2889	  ldxa	[%g0]ASI_IMMU, %g6
2890	ldxa	[%g0]ASI_DMMU, %g6
28911:	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi	! tag target
2892	stxa	%g3, [%g5 + TRAP_ENT_TR]%asi		! tsb4m pointer
2893	TRACE_NEXT(%g5, %g4, %g6)
2894	jmp	%g7 + 4
2895	nop
2896
2897/*
2898 * g2 = tag access register (in)
2899 * g3 = ctx number (in)
2900 */
2901trace_dataprot:
2902	membar	#Sync
2903	sethi	%hi(FLUSH_ADDR), %g6
2904	flush	%g6
2905	TRACE_PTR(%g1, %g6)
2906	GET_TRACE_TICK(%g6)
2907	stxa	%g6, [%g1 + TRAP_ENT_TICK]%asi
2908	rdpr	%tpc, %g6
2909	stna	%g6, [%g1 + TRAP_ENT_TPC]%asi
2910	rdpr	%tstate, %g6
2911	stxa	%g6, [%g1 + TRAP_ENT_TSTATE]%asi
2912	stxa	%g2, [%g1 + TRAP_ENT_SP]%asi		! tag access reg
2913	stxa	%g0, [%g1 + TRAP_ENT_TR]%asi
2914	stxa	%g0, [%g1 + TRAP_ENT_F1]%asi
2915	stxa	%g0, [%g1 + TRAP_ENT_F2]%asi
2916	stxa	%g0, [%g1 + TRAP_ENT_F3]%asi
2917	stxa	%g0, [%g1 + TRAP_ENT_F4]%asi
2918	rdpr	%tl, %g6
2919	stha	%g6, [%g1 + TRAP_ENT_TL]%asi
2920	rdpr	%tt, %g6
2921	stha	%g6, [%g1 + TRAP_ENT_TT]%asi
2922	TRACE_NEXT(%g1, %g4, %g5)
2923	jmp	%g7 + 4
2924	nop
2925
2926#endif /* TRAPTRACE */
2927
2928/*
2929 * fast_trap_done, fast_trap_done_chk_intr:
2930 *
2931 * Due to the design of UltraSPARC pipeline, pending interrupts are not
2932 * taken immediately after a RETRY or DONE instruction which causes IE to
2933 * go from 0 to 1. Instead, the instruction at %tpc or %tnpc is allowed
2934 * to execute first before taking any interrupts. If that instruction
2935 * results in other traps, and if the corresponding trap handler runs
2936 * entirely at TL=1 with interrupts disabled, then pending interrupts
2937 * won't be taken until after yet another instruction following the %tpc
2938 * or %tnpc.
2939 *
2940 * A malicious user program can use this feature to block out interrupts
2941 * for extended durations, which can result in send_mondo_timeout kernel
2942 * panic.
2943 *
2944 * This problem is addressed by servicing any pending interrupts via
2945 * sys_trap before returning back to the user mode from a fast trap
2946 * handler. The "done" instruction within a fast trap handler, which
2947 * runs entirely at TL=1 with interrupts disabled, is replaced with the
2948 * FAST_TRAP_DONE macro, which branches control to this fast_trap_done
2949 * entry point.
2950 *
2951 * We check for any pending interrupts here and force a sys_trap to
2952 * service those interrupts, if any. To minimize overhead, pending
2953 * interrupts are checked if the %tpc happens to be at 16K boundary,
2954 * which allows a malicious program to execute at most 4K consecutive
2955 * instructions before we service any pending interrupts. If a worst
2956 * case fast trap handler takes about 2 usec, then interrupts will be
2957 * blocked for at most 8 msec, less than a clock tick.
2958 *
2959 * For the cases where we don't know if the %tpc will cross a 16K
2960 * boundary, we can't use the above optimization and always process
2961 * any pending interrupts via fast_frap_done_chk_intr entry point.
2962 *
2963 * Entry Conditions:
2964 * 	%pstate		am:0 priv:1 ie:0
2965 * 			globals are AG (not normal globals)
2966 */
2967
2968	.global	fast_trap_done, fast_trap_done_chk_intr
2969fast_trap_done:
2970	rdpr	%tpc, %g5
2971	sethi	%hi(0xffffc000), %g6	! 1's complement of 0x3fff
2972	andncc	%g5, %g6, %g0		! check lower 14 bits of %tpc
2973	bz,a,pn	%icc, 1f		! branch if zero (lower 32 bits only)
2974	  ldxa	[%g0]ASI_INTR_RECEIVE_STATUS, %g5
2975	done
2976
2977	ALTENTRY(fast_trap_done_check_interrupts)
2978fast_trap_done_chk_intr:
2979	ldxa	[%g0]ASI_INTR_RECEIVE_STATUS, %g5
2980
29811:	rd	SOFTINT, %g6
2982	and	%g5, IRSR_BUSY, %g5
2983	orcc	%g5, %g6, %g0
2984	bnz,pn	%xcc, 2f		! branch if any pending intr
2985	nop
2986	done
2987
29882:
2989	/*
2990	 * We get here if there are any pending interrupts.
2991	 * Adjust %tpc/%tnpc as we'll be resuming via "retry"
2992	 * instruction.
2993	 */
2994	rdpr	%tnpc, %g5
2995	wrpr	%g0, %g5, %tpc
2996	add	%g5, 4, %g5
2997	wrpr	%g0, %g5, %tnpc
2998
2999	/*
3000	 * Force a dummy sys_trap call so that interrupts can be serviced.
3001	 */
3002	set	fast_trap_dummy_call, %g1
3003	ba,pt	%xcc, sys_trap
3004	  mov	-1, %g4
3005
3006fast_trap_dummy_call:
3007	retl
3008	nop
3009
3010/*
3011 * Currently the brand syscall interposition code is not enabled by
3012 * default.  Instead, when a branded zone is first booted the brand
3013 * infrastructure will patch the trap table so that the syscall
3014 * entry points are redirected to syscall_wrapper32 and syscall_wrapper
3015 * for ILP32 and LP64 syscalls respectively.  this is done in
3016 * brand_plat_interposition_enable().  Note that the syscall wrappers
3017 * below do not collect any trap trace data since the syscall hot patch
3018 * points are reached after trap trace data has already been collected.
3019 */
3020#define	BRAND_CALLBACK(callback_id)					    \
3021	CPU_ADDR(%g2, %g1)		/* load CPU struct addr to %g2	*/ ;\
3022	ldn	[%g2 + CPU_THREAD], %g3	/* load thread pointer		*/ ;\
3023	ldn	[%g3 + T_PROCP], %g3	/* get proc pointer		*/ ;\
3024	ldn	[%g3 + P_BRAND], %g3	/* get brand pointer		*/ ;\
3025	brz	%g3, 1f			/* No brand?  No callback. 	*/ ;\
3026	nop 								   ;\
3027	ldn	[%g3 + B_MACHOPS], %g3	/* get machops list		*/ ;\
3028	ldn	[%g3 + (callback_id << 3)], %g3 			   ;\
3029	brz	%g3, 1f							   ;\
3030	/*								    \
3031	 * This isn't pretty.  We want a low-latency way for the callback   \
3032	 * routine to decline to do anything.  We just pass in an address   \
3033	 * the routine can directly jmp back to, pretending that nothing    \
3034	 * has happened.						    \
3035	 * 								    \
3036	 * %g1: return address (where the brand handler jumps back to)	    \
3037	 * %g2: address of CPU structure				    \
3038	 * %g3: address of brand handler (where we will jump to)	    \
3039	 */								    \
3040	mov	%pc, %g1						   ;\
3041	add	%g1, 16, %g1						   ;\
3042	jmp	%g3							   ;\
3043	nop								   ;\
30441:
3045
3046	ENTRY_NP(syscall_wrapper32)
3047	BRAND_CALLBACK(BRAND_CB_SYSCALL32)
3048	SYSCALL_NOTT(syscall_trap32)
3049	SET_SIZE(syscall_wrapper32)
3050
3051	ENTRY_NP(syscall_wrapper)
3052	BRAND_CALLBACK(BRAND_CB_SYSCALL)
3053	SYSCALL_NOTT(syscall_trap)
3054	SET_SIZE(syscall_wrapper)
3055
3056#endif	/* lint */
3057