xref: /titanic_41/usr/src/uts/sun4v/ml/trap_table.s (revision bbb1277b6ec1b0daad4e3ed1a2b891d3e2ece2eb)
1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22/*
23 * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
24 * Use is subject to license terms.
25 */
26
27#if !defined(lint)
28#include "assym.h"
29#endif /* !lint */
30#include <sys/asm_linkage.h>
31#include <sys/privregs.h>
32#include <sys/sun4asi.h>
33#include <sys/machasi.h>
34#include <sys/hypervisor_api.h>
35#include <sys/machtrap.h>
36#include <sys/machthread.h>
37#include <sys/machbrand.h>
38#include <sys/pcb.h>
39#include <sys/pte.h>
40#include <sys/mmu.h>
41#include <sys/machpcb.h>
42#include <sys/async.h>
43#include <sys/intreg.h>
44#include <sys/scb.h>
45#include <sys/psr_compat.h>
46#include <sys/syscall.h>
47#include <sys/machparam.h>
48#include <sys/traptrace.h>
49#include <vm/hat_sfmmu.h>
50#include <sys/archsystm.h>
51#include <sys/utrap.h>
52#include <sys/clock.h>
53#include <sys/intr.h>
54#include <sys/fpu/fpu_simulator.h>
55#include <vm/seg_spt.h>
56
57/*
58 * WARNING: If you add a fast trap handler which can be invoked by a
59 * non-privileged user, you may have to use the FAST_TRAP_DONE macro
60 * instead of "done" instruction to return back to the user mode. See
61 * comments for the "fast_trap_done" entry point for more information.
62 *
63 * An alternate FAST_TRAP_DONE_CHK_INTR macro should be used for the
64 * cases where you always want to process any pending interrupts before
65 * returning back to the user mode.
66 */
67#define	FAST_TRAP_DONE		\
68	ba,a	fast_trap_done
69
70#define	FAST_TRAP_DONE_CHK_INTR	\
71	ba,a	fast_trap_done_chk_intr
72
73/*
74 * SPARC V9 Trap Table
75 *
76 * Most of the trap handlers are made from common building
77 * blocks, and some are instantiated multiple times within
78 * the trap table. So, I build a bunch of macros, then
79 * populate the table using only the macros.
80 *
81 * Many macros branch to sys_trap.  Its calling convention is:
82 *	%g1		kernel trap handler
83 *	%g2, %g3	args for above
84 *	%g4		desire %pil
85 */
86
87#ifdef	TRAPTRACE
88
89/*
90 * Tracing macro. Adds two instructions if TRAPTRACE is defined.
91 */
92#define	TT_TRACE(label)		\
93	ba	label		;\
94	rd	%pc, %g7
95#define	TT_TRACE_INS	2
96
97#define	TT_TRACE_L(label)	\
98	ba	label		;\
99	rd	%pc, %l4	;\
100	clr	%l4
101#define	TT_TRACE_L_INS	3
102
103#else
104
105#define	TT_TRACE(label)
106#define	TT_TRACE_INS	0
107
108#define	TT_TRACE_L(label)
109#define	TT_TRACE_L_INS	0
110
111#endif
112
113/*
114 * This first set are funneled to trap() with %tt as the type.
115 * Trap will then either panic or send the user a signal.
116 */
117/*
118 * NOT is used for traps that just shouldn't happen.
119 * It comes in both single and quadruple flavors.
120 */
121#if !defined(lint)
122	.global	trap
123#endif /* !lint */
124#define	NOT			\
125	TT_TRACE(trace_gen)	;\
126	set	trap, %g1	;\
127	rdpr	%tt, %g3	;\
128	ba,pt	%xcc, sys_trap	;\
129	sub	%g0, 1, %g4	;\
130	.align	32
131#define	NOT4	NOT; NOT; NOT; NOT
132
133#define	NOTP				\
134	TT_TRACE(trace_gen)		;\
135	ba,pt	%xcc, ptl1_panic	;\
136	  mov	PTL1_BAD_TRAP, %g1	;\
137	.align	32
138#define	NOTP4	NOTP; NOTP; NOTP; NOTP
139
140
141/*
142 * BAD is used for trap vectors we don't have a kernel
143 * handler for.
144 * It also comes in single and quadruple versions.
145 */
146#define	BAD	NOT
147#define	BAD4	NOT4
148
149#define	DONE			\
150	done;			\
151	.align	32
152
153/*
154 * TRAP vectors to the trap() function.
155 * It's main use is for user errors.
156 */
157#if !defined(lint)
158	.global	trap
159#endif /* !lint */
160#define	TRAP(arg)		\
161	TT_TRACE(trace_gen)	;\
162	set	trap, %g1	;\
163	mov	arg, %g3	;\
164	ba,pt	%xcc, sys_trap	;\
165	sub	%g0, 1, %g4	;\
166	.align	32
167
168/*
169 * SYSCALL is used for unsupported syscall interfaces (with 'which'
170 * set to 'nosys') and legacy support of old SunOS 4.x syscalls (with
171 * 'which' set to 'syscall_trap32').
172 *
173 * The SYSCALL_TRAP* macros are used for syscall entry points.
174 * SYSCALL_TRAP is used to support LP64 syscalls and SYSCALL_TRAP32
175 * is used to support ILP32.  Each macro can only be used once
176 * since they each define a symbol.  The symbols are used as hot patch
177 * points by the brand infrastructure to dynamically enable and disable
178 * brand syscall interposition.  See the comments around BRAND_CALLBACK
179 * and brand_plat_interposition_enable() for more information.
180 */
181#define	SYSCALL_NOTT(which)		\
182	set	(which), %g1		;\
183	ba,pt	%xcc, user_trap		;\
184	sub	%g0, 1, %g4		;\
185	.align	32
186
187#define	SYSCALL(which)			\
188	TT_TRACE(trace_gen)		;\
189	SYSCALL_NOTT(which)
190
191#define	SYSCALL_TRAP32				\
192	TT_TRACE(trace_gen)			;\
193	ALTENTRY(syscall_trap32_patch_point)	\
194	SYSCALL_NOTT(syscall_trap32)
195
196#define	SYSCALL_TRAP				\
197	TT_TRACE(trace_gen)			;\
198	ALTENTRY(syscall_trap_patch_point)	\
199	SYSCALL_NOTT(syscall_trap)
200
201/*
202 * GOTO just jumps to a label.
203 * It's used for things that can be fixed without going thru sys_trap.
204 */
205#define	GOTO(label)		\
206	.global	label		;\
207	ba,a	label		;\
208	.empty			;\
209	.align	32
210
211/*
212 * GOTO_TT just jumps to a label.
213 * correctable ECC error traps at  level 0 and 1 will use this macro.
214 * It's used for things that can be fixed without going thru sys_trap.
215 */
216#define	GOTO_TT(label, ttlabel)		\
217	.global	label		;\
218	TT_TRACE(ttlabel)	;\
219	ba,a	label		;\
220	.empty			;\
221	.align	32
222
223/*
224 * Privileged traps
225 * Takes breakpoint if privileged, calls trap() if not.
226 */
227#define	PRIV(label)			\
228	rdpr	%tstate, %g1		;\
229	btst	TSTATE_PRIV, %g1	;\
230	bnz	label			;\
231	rdpr	%tt, %g3		;\
232	set	trap, %g1		;\
233	ba,pt	%xcc, sys_trap		;\
234	sub	%g0, 1, %g4		;\
235	.align	32
236
237
238/*
239 * DTrace traps.
240 */
241#define	DTRACE_PID			\
242	.global dtrace_pid_probe				;\
243	set	dtrace_pid_probe, %g1				;\
244	ba,pt	%xcc, user_trap					;\
245	sub	%g0, 1, %g4					;\
246	.align	32
247
248#define	DTRACE_RETURN			\
249	.global dtrace_return_probe				;\
250	set	dtrace_return_probe, %g1			;\
251	ba,pt	%xcc, user_trap					;\
252	sub	%g0, 1, %g4					;\
253	.align	32
254
255/*
256 * REGISTER WINDOW MANAGEMENT MACROS
257 */
258
259/*
260 * various convenient units of padding
261 */
262#define	SKIP(n)	.skip 4*(n)
263
264/*
265 * CLEAN_WINDOW is the simple handler for cleaning a register window.
266 */
267#define	CLEAN_WINDOW						\
268	TT_TRACE_L(trace_win)					;\
269	rdpr %cleanwin, %l0; inc %l0; wrpr %l0, %cleanwin	;\
270	clr %l0; clr %l1; clr %l2; clr %l3			;\
271	clr %l4; clr %l5; clr %l6; clr %l7			;\
272	clr %o0; clr %o1; clr %o2; clr %o3			;\
273	clr %o4; clr %o5; clr %o6; clr %o7			;\
274	retry; .align 128
275
276#if !defined(lint)
277
278/*
279 * If we get an unresolved tlb miss while in a window handler, the fault
280 * handler will resume execution at the last instruction of the window
281 * hander, instead of delivering the fault to the kernel.  Spill handlers
282 * use this to spill windows into the wbuf.
283 *
284 * The mixed handler works by checking %sp, and branching to the correct
285 * handler.  This is done by branching back to label 1: for 32b frames,
286 * or label 2: for 64b frames; which implies the handler order is: 32b,
287 * 64b, mixed.  The 1: and 2: labels are offset into the routines to
288 * allow the branchs' delay slots to contain useful instructions.
289 */
290
291/*
292 * SPILL_32bit spills a 32-bit-wide kernel register window.  It
293 * assumes that the kernel context and the nucleus context are the
294 * same.  The stack pointer is required to be eight-byte aligned even
295 * though this code only needs it to be four-byte aligned.
296 */
297#define	SPILL_32bit(tail)					\
298	srl	%sp, 0, %sp					;\
2991:	st	%l0, [%sp + 0]					;\
300	st	%l1, [%sp + 4]					;\
301	st	%l2, [%sp + 8]					;\
302	st	%l3, [%sp + 12]					;\
303	st	%l4, [%sp + 16]					;\
304	st	%l5, [%sp + 20]					;\
305	st	%l6, [%sp + 24]					;\
306	st	%l7, [%sp + 28]					;\
307	st	%i0, [%sp + 32]					;\
308	st	%i1, [%sp + 36]					;\
309	st	%i2, [%sp + 40]					;\
310	st	%i3, [%sp + 44]					;\
311	st	%i4, [%sp + 48]					;\
312	st	%i5, [%sp + 52]					;\
313	st	%i6, [%sp + 56]					;\
314	st	%i7, [%sp + 60]					;\
315	TT_TRACE_L(trace_win)					;\
316	saved							;\
317	retry							;\
318	SKIP(31-19-TT_TRACE_L_INS)				;\
319	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
320	.empty
321
322/*
323 * SPILL_32bit_asi spills a 32-bit-wide register window into a 32-bit
324 * wide address space via the designated asi.  It is used to spill
325 * non-kernel windows.  The stack pointer is required to be eight-byte
326 * aligned even though this code only needs it to be four-byte
327 * aligned.
328 */
329#define	SPILL_32bit_asi(asi_num, tail)				\
330	srl	%sp, 0, %sp					;\
3311:	sta	%l0, [%sp + %g0]asi_num				;\
332	mov	4, %g1						;\
333	sta	%l1, [%sp + %g1]asi_num				;\
334	mov	8, %g2						;\
335	sta	%l2, [%sp + %g2]asi_num				;\
336	mov	12, %g3						;\
337	sta	%l3, [%sp + %g3]asi_num				;\
338	add	%sp, 16, %g4					;\
339	sta	%l4, [%g4 + %g0]asi_num				;\
340	sta	%l5, [%g4 + %g1]asi_num				;\
341	sta	%l6, [%g4 + %g2]asi_num				;\
342	sta	%l7, [%g4 + %g3]asi_num				;\
343	add	%g4, 16, %g4					;\
344	sta	%i0, [%g4 + %g0]asi_num				;\
345	sta	%i1, [%g4 + %g1]asi_num				;\
346	sta	%i2, [%g4 + %g2]asi_num				;\
347	sta	%i3, [%g4 + %g3]asi_num				;\
348	add	%g4, 16, %g4					;\
349	sta	%i4, [%g4 + %g0]asi_num				;\
350	sta	%i5, [%g4 + %g1]asi_num				;\
351	sta	%i6, [%g4 + %g2]asi_num				;\
352	sta	%i7, [%g4 + %g3]asi_num				;\
353	TT_TRACE_L(trace_win)					;\
354	saved							;\
355	retry							;\
356	SKIP(31-25-TT_TRACE_L_INS)				;\
357	ba,a,pt %xcc, fault_32bit_/**/tail			;\
358	.empty
359
360#define	SPILL_32bit_tt1(asi_num, tail)				\
361	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
362	.empty							;\
363	.align 128
364
365
366/*
367 * FILL_32bit fills a 32-bit-wide kernel register window.  It assumes
368 * that the kernel context and the nucleus context are the same.  The
369 * stack pointer is required to be eight-byte aligned even though this
370 * code only needs it to be four-byte aligned.
371 */
372#define	FILL_32bit(tail)					\
373	srl	%sp, 0, %sp					;\
3741:	TT_TRACE_L(trace_win)					;\
375	ld	[%sp + 0], %l0					;\
376	ld	[%sp + 4], %l1					;\
377	ld	[%sp + 8], %l2					;\
378	ld	[%sp + 12], %l3					;\
379	ld	[%sp + 16], %l4					;\
380	ld	[%sp + 20], %l5					;\
381	ld	[%sp + 24], %l6					;\
382	ld	[%sp + 28], %l7					;\
383	ld	[%sp + 32], %i0					;\
384	ld	[%sp + 36], %i1					;\
385	ld	[%sp + 40], %i2					;\
386	ld	[%sp + 44], %i3					;\
387	ld	[%sp + 48], %i4					;\
388	ld	[%sp + 52], %i5					;\
389	ld	[%sp + 56], %i6					;\
390	ld	[%sp + 60], %i7					;\
391	restored						;\
392	retry							;\
393	SKIP(31-19-TT_TRACE_L_INS)				;\
394	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
395	.empty
396
397/*
398 * FILL_32bit_asi fills a 32-bit-wide register window from a 32-bit
399 * wide address space via the designated asi.  It is used to fill
400 * non-kernel windows.  The stack pointer is required to be eight-byte
401 * aligned even though this code only needs it to be four-byte
402 * aligned.
403 */
404#define	FILL_32bit_asi(asi_num, tail)				\
405	srl	%sp, 0, %sp					;\
4061:	TT_TRACE_L(trace_win)					;\
407	mov	4, %g1						;\
408	lda	[%sp + %g0]asi_num, %l0				;\
409	mov	8, %g2						;\
410	lda	[%sp + %g1]asi_num, %l1				;\
411	mov	12, %g3						;\
412	lda	[%sp + %g2]asi_num, %l2				;\
413	lda	[%sp + %g3]asi_num, %l3				;\
414	add	%sp, 16, %g4					;\
415	lda	[%g4 + %g0]asi_num, %l4				;\
416	lda	[%g4 + %g1]asi_num, %l5				;\
417	lda	[%g4 + %g2]asi_num, %l6				;\
418	lda	[%g4 + %g3]asi_num, %l7				;\
419	add	%g4, 16, %g4					;\
420	lda	[%g4 + %g0]asi_num, %i0				;\
421	lda	[%g4 + %g1]asi_num, %i1				;\
422	lda	[%g4 + %g2]asi_num, %i2				;\
423	lda	[%g4 + %g3]asi_num, %i3				;\
424	add	%g4, 16, %g4					;\
425	lda	[%g4 + %g0]asi_num, %i4				;\
426	lda	[%g4 + %g1]asi_num, %i5				;\
427	lda	[%g4 + %g2]asi_num, %i6				;\
428	lda	[%g4 + %g3]asi_num, %i7				;\
429	restored						;\
430	retry							;\
431	SKIP(31-25-TT_TRACE_L_INS)				;\
432	ba,a,pt %xcc, fault_32bit_/**/tail			;\
433	.empty
434
435
436/*
437 * SPILL_64bit spills a 64-bit-wide kernel register window.  It
438 * assumes that the kernel context and the nucleus context are the
439 * same.  The stack pointer is required to be eight-byte aligned.
440 */
441#define	SPILL_64bit(tail)					\
4422:	stx	%l0, [%sp + V9BIAS64 + 0]			;\
443	stx	%l1, [%sp + V9BIAS64 + 8]			;\
444	stx	%l2, [%sp + V9BIAS64 + 16]			;\
445	stx	%l3, [%sp + V9BIAS64 + 24]			;\
446	stx	%l4, [%sp + V9BIAS64 + 32]			;\
447	stx	%l5, [%sp + V9BIAS64 + 40]			;\
448	stx	%l6, [%sp + V9BIAS64 + 48]			;\
449	stx	%l7, [%sp + V9BIAS64 + 56]			;\
450	stx	%i0, [%sp + V9BIAS64 + 64]			;\
451	stx	%i1, [%sp + V9BIAS64 + 72]			;\
452	stx	%i2, [%sp + V9BIAS64 + 80]			;\
453	stx	%i3, [%sp + V9BIAS64 + 88]			;\
454	stx	%i4, [%sp + V9BIAS64 + 96]			;\
455	stx	%i5, [%sp + V9BIAS64 + 104]			;\
456	stx	%i6, [%sp + V9BIAS64 + 112]			;\
457	stx	%i7, [%sp + V9BIAS64 + 120]			;\
458	TT_TRACE_L(trace_win)					;\
459	saved							;\
460	retry							;\
461	SKIP(31-18-TT_TRACE_L_INS)				;\
462	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
463	.empty
464
465#define	SPILL_64bit_ktt1(tail)				\
466	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
467	.empty							;\
468	.align 128
469
470#define	SPILL_mixed_ktt1(tail)				\
471	btst	1, %sp						;\
472	bz,a,pt	%xcc, fault_32bit_/**/tail			;\
473	srl	%sp, 0, %sp					;\
474	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
475	.empty							;\
476	.align 128
477
478/*
479 * SPILL_64bit_asi spills a 64-bit-wide register window into a 64-bit
480 * wide address space via the designated asi.  It is used to spill
481 * non-kernel windows.  The stack pointer is required to be eight-byte
482 * aligned.
483 */
484#define	SPILL_64bit_asi(asi_num, tail)				\
485	mov	0 + V9BIAS64, %g1				;\
4862:	stxa	%l0, [%sp + %g1]asi_num				;\
487	mov	8 + V9BIAS64, %g2				;\
488	stxa	%l1, [%sp + %g2]asi_num				;\
489	mov	16 + V9BIAS64, %g3				;\
490	stxa	%l2, [%sp + %g3]asi_num				;\
491	mov	24 + V9BIAS64, %g4				;\
492	stxa	%l3, [%sp + %g4]asi_num				;\
493	add	%sp, 32, %g5					;\
494	stxa	%l4, [%g5 + %g1]asi_num				;\
495	stxa	%l5, [%g5 + %g2]asi_num				;\
496	stxa	%l6, [%g5 + %g3]asi_num				;\
497	stxa	%l7, [%g5 + %g4]asi_num				;\
498	add	%g5, 32, %g5					;\
499	stxa	%i0, [%g5 + %g1]asi_num				;\
500	stxa	%i1, [%g5 + %g2]asi_num				;\
501	stxa	%i2, [%g5 + %g3]asi_num				;\
502	stxa	%i3, [%g5 + %g4]asi_num				;\
503	add	%g5, 32, %g5					;\
504	stxa	%i4, [%g5 + %g1]asi_num				;\
505	stxa	%i5, [%g5 + %g2]asi_num				;\
506	stxa	%i6, [%g5 + %g3]asi_num				;\
507	stxa	%i7, [%g5 + %g4]asi_num				;\
508	TT_TRACE_L(trace_win)					;\
509	saved							;\
510	retry							;\
511	SKIP(31-25-TT_TRACE_L_INS)				;\
512	ba,a,pt %xcc, fault_64bit_/**/tail			;\
513	.empty
514
515#define	SPILL_64bit_tt1(asi_num, tail)				\
516	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
517	.empty							;\
518	.align 128
519
520/*
521 * FILL_64bit fills a 64-bit-wide kernel register window.  It assumes
522 * that the kernel context and the nucleus context are the same.  The
523 * stack pointer is required to be eight-byte aligned.
524 */
525#define	FILL_64bit(tail)					\
5262:	TT_TRACE_L(trace_win)					;\
527	ldx	[%sp + V9BIAS64 + 0], %l0			;\
528	ldx	[%sp + V9BIAS64 + 8], %l1			;\
529	ldx	[%sp + V9BIAS64 + 16], %l2			;\
530	ldx	[%sp + V9BIAS64 + 24], %l3			;\
531	ldx	[%sp + V9BIAS64 + 32], %l4			;\
532	ldx	[%sp + V9BIAS64 + 40], %l5			;\
533	ldx	[%sp + V9BIAS64 + 48], %l6			;\
534	ldx	[%sp + V9BIAS64 + 56], %l7			;\
535	ldx	[%sp + V9BIAS64 + 64], %i0			;\
536	ldx	[%sp + V9BIAS64 + 72], %i1			;\
537	ldx	[%sp + V9BIAS64 + 80], %i2			;\
538	ldx	[%sp + V9BIAS64 + 88], %i3			;\
539	ldx	[%sp + V9BIAS64 + 96], %i4			;\
540	ldx	[%sp + V9BIAS64 + 104], %i5			;\
541	ldx	[%sp + V9BIAS64 + 112], %i6			;\
542	ldx	[%sp + V9BIAS64 + 120], %i7			;\
543	restored						;\
544	retry							;\
545	SKIP(31-18-TT_TRACE_L_INS)				;\
546	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
547	.empty
548
549/*
550 * FILL_64bit_asi fills a 64-bit-wide register window from a 64-bit
551 * wide address space via the designated asi.  It is used to fill
552 * non-kernel windows.  The stack pointer is required to be eight-byte
553 * aligned.
554 */
555#define	FILL_64bit_asi(asi_num, tail)				\
556	mov	V9BIAS64 + 0, %g1				;\
5572:	TT_TRACE_L(trace_win)					;\
558	ldxa	[%sp + %g1]asi_num, %l0				;\
559	mov	V9BIAS64 + 8, %g2				;\
560	ldxa	[%sp + %g2]asi_num, %l1				;\
561	mov	V9BIAS64 + 16, %g3				;\
562	ldxa	[%sp + %g3]asi_num, %l2				;\
563	mov	V9BIAS64 + 24, %g4				;\
564	ldxa	[%sp + %g4]asi_num, %l3				;\
565	add	%sp, 32, %g5					;\
566	ldxa	[%g5 + %g1]asi_num, %l4				;\
567	ldxa	[%g5 + %g2]asi_num, %l5				;\
568	ldxa	[%g5 + %g3]asi_num, %l6				;\
569	ldxa	[%g5 + %g4]asi_num, %l7				;\
570	add	%g5, 32, %g5					;\
571	ldxa	[%g5 + %g1]asi_num, %i0				;\
572	ldxa	[%g5 + %g2]asi_num, %i1				;\
573	ldxa	[%g5 + %g3]asi_num, %i2				;\
574	ldxa	[%g5 + %g4]asi_num, %i3				;\
575	add	%g5, 32, %g5					;\
576	ldxa	[%g5 + %g1]asi_num, %i4				;\
577	ldxa	[%g5 + %g2]asi_num, %i5				;\
578	ldxa	[%g5 + %g3]asi_num, %i6				;\
579	ldxa	[%g5 + %g4]asi_num, %i7				;\
580	restored						;\
581	retry							;\
582	SKIP(31-25-TT_TRACE_L_INS)				;\
583	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
584	.empty
585
586
587#endif /* !lint */
588
589/*
590 * SPILL_mixed spills either size window, depending on
591 * whether %sp is even or odd, to a 32-bit address space.
592 * This may only be used in conjunction with SPILL_32bit/
593 * FILL_64bit.
594 * Clear upper 32 bits of %sp if it is odd.
595 * We won't need to clear them in 64 bit kernel.
596 */
597#define	SPILL_mixed						\
598	btst	1, %sp						;\
599	bz,a,pt	%xcc, 1b					;\
600	srl	%sp, 0, %sp					;\
601	ba,pt	%xcc, 2b					;\
602	nop							;\
603	.align	128
604
605/*
606 * FILL_mixed(ASI) fills either size window, depending on
607 * whether %sp is even or odd, from a 32-bit address space.
608 * This may only be used in conjunction with FILL_32bit/
609 * FILL_64bit. New versions of FILL_mixed_{tt1,asi} would be
610 * needed for use with FILL_{32,64}bit_{tt1,asi}. Particular
611 * attention should be paid to the instructions that belong
612 * in the delay slots of the branches depending on the type
613 * of fill handler being branched to.
614 * Clear upper 32 bits of %sp if it is odd.
615 * We won't need to clear them in 64 bit kernel.
616 */
617#define	FILL_mixed						\
618	btst	1, %sp						;\
619	bz,a,pt	%xcc, 1b					;\
620	srl	%sp, 0, %sp					;\
621	ba,pt	%xcc, 2b					;\
622	nop							;\
623	.align	128
624
625
626/*
627 * SPILL_32clean/SPILL_64clean spill 32-bit and 64-bit register windows,
628 * respectively, into the address space via the designated asi.  The
629 * unbiased stack pointer is required to be eight-byte aligned (even for
630 * the 32-bit case even though this code does not require such strict
631 * alignment).
632 *
633 * With SPARC v9 the spill trap takes precedence over the cleanwin trap
634 * so when cansave == 0, canrestore == 6, and cleanwin == 6 the next save
635 * will cause cwp + 2 to be spilled but will not clean cwp + 1.  That
636 * window may contain kernel data so in user_rtt we set wstate to call
637 * these spill handlers on the first user spill trap.  These handler then
638 * spill the appropriate window but also back up a window and clean the
639 * window that didn't get a cleanwin trap.
640 */
641#define	SPILL_32clean(asi_num, tail)				\
642	srl	%sp, 0, %sp					;\
643	sta	%l0, [%sp + %g0]asi_num				;\
644	mov	4, %g1						;\
645	sta	%l1, [%sp + %g1]asi_num				;\
646	mov	8, %g2						;\
647	sta	%l2, [%sp + %g2]asi_num				;\
648	mov	12, %g3						;\
649	sta	%l3, [%sp + %g3]asi_num				;\
650	add	%sp, 16, %g4					;\
651	sta	%l4, [%g4 + %g0]asi_num				;\
652	sta	%l5, [%g4 + %g1]asi_num				;\
653	sta	%l6, [%g4 + %g2]asi_num				;\
654	sta	%l7, [%g4 + %g3]asi_num				;\
655	add	%g4, 16, %g4					;\
656	sta	%i0, [%g4 + %g0]asi_num				;\
657	sta	%i1, [%g4 + %g1]asi_num				;\
658	sta	%i2, [%g4 + %g2]asi_num				;\
659	sta	%i3, [%g4 + %g3]asi_num				;\
660	add	%g4, 16, %g4					;\
661	sta	%i4, [%g4 + %g0]asi_num				;\
662	sta	%i5, [%g4 + %g1]asi_num				;\
663	sta	%i6, [%g4 + %g2]asi_num				;\
664	sta	%i7, [%g4 + %g3]asi_num				;\
665	TT_TRACE_L(trace_win)					;\
666	b	.spill_clean					;\
667	  mov	WSTATE_USER32, %g7				;\
668	SKIP(31-25-TT_TRACE_L_INS)				;\
669	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
670	.empty
671
672#define	SPILL_64clean(asi_num, tail)				\
673	mov	0 + V9BIAS64, %g1				;\
674	stxa	%l0, [%sp + %g1]asi_num				;\
675	mov	8 + V9BIAS64, %g2				;\
676	stxa	%l1, [%sp + %g2]asi_num				;\
677	mov	16 + V9BIAS64, %g3				;\
678	stxa	%l2, [%sp + %g3]asi_num				;\
679	mov	24 + V9BIAS64, %g4				;\
680	stxa	%l3, [%sp + %g4]asi_num				;\
681	add	%sp, 32, %g5					;\
682	stxa	%l4, [%g5 + %g1]asi_num				;\
683	stxa	%l5, [%g5 + %g2]asi_num				;\
684	stxa	%l6, [%g5 + %g3]asi_num				;\
685	stxa	%l7, [%g5 + %g4]asi_num				;\
686	add	%g5, 32, %g5					;\
687	stxa	%i0, [%g5 + %g1]asi_num				;\
688	stxa	%i1, [%g5 + %g2]asi_num				;\
689	stxa	%i2, [%g5 + %g3]asi_num				;\
690	stxa	%i3, [%g5 + %g4]asi_num				;\
691	add	%g5, 32, %g5					;\
692	stxa	%i4, [%g5 + %g1]asi_num				;\
693	stxa	%i5, [%g5 + %g2]asi_num				;\
694	stxa	%i6, [%g5 + %g3]asi_num				;\
695	stxa	%i7, [%g5 + %g4]asi_num				;\
696	TT_TRACE_L(trace_win)					;\
697	b	.spill_clean					;\
698	  mov	WSTATE_USER64, %g7				;\
699	SKIP(31-25-TT_TRACE_L_INS)				;\
700	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
701	.empty
702
703
704/*
705 * Floating point disabled.
706 */
707#define	FP_DISABLED_TRAP		\
708	TT_TRACE(trace_gen)		;\
709	ba,pt	%xcc,.fp_disabled	;\
710	nop				;\
711	.align	32
712
713/*
714 * Floating point exceptions.
715 */
716#define	FP_IEEE_TRAP			\
717	TT_TRACE(trace_gen)		;\
718	ba,pt	%xcc,.fp_ieee_exception	;\
719	nop				;\
720	.align	32
721
722#define	FP_TRAP				\
723	TT_TRACE(trace_gen)		;\
724	ba,pt	%xcc,.fp_exception	;\
725	nop				;\
726	.align	32
727
728#if !defined(lint)
729
730/*
731 * ECACHE_ECC error traps at level 0 and level 1
732 */
733#define	ECACHE_ECC(table_name)		\
734	.global	table_name		;\
735table_name:				;\
736	membar	#Sync			;\
737	set	trap, %g1		;\
738	rdpr	%tt, %g3		;\
739	ba,pt	%xcc, sys_trap		;\
740	sub	%g0, 1, %g4		;\
741	.align	32
742
743#endif /* !lint */
744
745/*
746 * illegal instruction trap
747 */
748#define	ILLTRAP_INSTR			  \
749	membar	#Sync			  ;\
750	TT_TRACE(trace_gen)		  ;\
751	or	%g0, P_UTRAP4, %g2	  ;\
752	or	%g0, T_UNIMP_INSTR, %g3   ;\
753	sethi	%hi(.check_v9utrap), %g4  ;\
754	jmp	%g4 + %lo(.check_v9utrap) ;\
755	nop				  ;\
756	.align	32
757
758/*
759 * tag overflow trap
760 */
761#define	TAG_OVERFLOW			  \
762	TT_TRACE(trace_gen)		  ;\
763	or	%g0, P_UTRAP10, %g2	  ;\
764	or	%g0, T_TAG_OVERFLOW, %g3  ;\
765	sethi	%hi(.check_v9utrap), %g4  ;\
766	jmp	%g4 + %lo(.check_v9utrap) ;\
767	nop				  ;\
768	.align	32
769
770/*
771 * divide by zero trap
772 */
773#define	DIV_BY_ZERO			  \
774	TT_TRACE(trace_gen)		  ;\
775	or	%g0, P_UTRAP11, %g2	  ;\
776	or	%g0, T_IDIV0, %g3	  ;\
777	sethi	%hi(.check_v9utrap), %g4  ;\
778	jmp	%g4 + %lo(.check_v9utrap) ;\
779	nop				  ;\
780	.align	32
781
782/*
783 * trap instruction for V9 user trap handlers
784 */
785#define	TRAP_INSTR			  \
786	TT_TRACE(trace_gen)		  ;\
787	or	%g0, T_SOFTWARE_TRAP, %g3 ;\
788	sethi	%hi(.check_v9utrap), %g4  ;\
789	jmp	%g4 + %lo(.check_v9utrap) ;\
790	nop				  ;\
791	.align	32
792#define	TRP4	TRAP_INSTR; TRAP_INSTR; TRAP_INSTR; TRAP_INSTR
793
794/*
795 * LEVEL_INTERRUPT is for level N interrupts.
796 * VECTOR_INTERRUPT is for the vector trap.
797 */
798#define	LEVEL_INTERRUPT(level)		\
799	.global	tt_pil/**/level		;\
800tt_pil/**/level:			;\
801	ba,pt	%xcc, pil_interrupt	;\
802	mov	level, %g4		;\
803	.align	32
804
805#define	LEVEL14_INTERRUPT			\
806	ba	pil14_interrupt			;\
807	mov	PIL_14, %g4			;\
808	.align	32
809
810#define        LEVEL15_INTERRUPT                       \
811       ba      pil15_interrupt                 ;\
812       mov     PIL_15, %g4                     ;\
813       .align  32
814
815#define CPU_MONDO			\
816	ba,a,pt	%xcc, cpu_mondo		;\
817	.align	32
818
819#define DEV_MONDO			\
820	ba,a,pt	%xcc, dev_mondo		;\
821	.align	32
822
823/*
824 * We take over the rtba after we set our trap table and
825 * fault status area. The watchdog reset trap is now handled by the OS.
826 */
827#define WATCHDOG_RESET			\
828	mov	PTL1_BAD_WATCHDOG, %g1	;\
829	ba,a,pt	%xcc, .watchdog_trap	;\
830	.align	32
831
832/*
833 * RED is for traps that use the red mode handler.
834 * We should never see these either.
835 */
836#define RED			\
837	mov	PTL1_BAD_RED, %g1	;\
838	ba,a,pt	%xcc, .watchdog_trap	;\
839	.align	32
840
841
842/*
843 * MMU Trap Handlers.
844 */
845
846/*
847 * synthesize for trap(): SFSR in %g3
848 */
849#define	IMMU_EXCEPTION							\
850	MMU_FAULT_STATUS_AREA(%g3)					;\
851	rdpr	%tpc, %g2						;\
852	ldx	[%g3 + MMFSA_I_TYPE], %g1				;\
853	ldx	[%g3 + MMFSA_I_CTX], %g3				;\
854	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
855	or	%g3, %g1, %g3						;\
856	ba,pt	%xcc, .mmu_exception_end				;\
857	mov	T_INSTR_EXCEPTION, %g1					;\
858	.align	32
859
860/*
861 * synthesize for trap(): TAG_ACCESS in %g2, SFSR in %g3
862 */
863#define	DMMU_EXCEPTION							\
864	ba,a,pt	%xcc, .dmmu_exception					;\
865	.align	32
866
867/*
868 * synthesize for trap(): SFAR in %g2, SFSR in %g3
869 */
870#define	DMMU_EXC_AG_PRIV						\
871	MMU_FAULT_STATUS_AREA(%g3)					;\
872	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
873	/* Fault type not available in MMU fault status area */		;\
874	mov	MMFSA_F_PRVACT, %g1					;\
875	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
876	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
877	ba,pt	%xcc, .mmu_priv_exception				;\
878	or	%g3, %g1, %g3						;\
879	.align	32
880
881/*
882 * synthesize for trap(): SFAR in %g2, SFSR in %g3
883 */
884#define	DMMU_EXC_AG_NOT_ALIGNED						\
885	MMU_FAULT_STATUS_AREA(%g3)					;\
886	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
887	/* Fault type not available in MMU fault status area */		;\
888	mov	MMFSA_F_UNALIGN, %g1					;\
889	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
890	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
891	ba,pt	%xcc, .mmu_exception_not_aligned			;\
892	or	%g3, %g1, %g3			/* SFSR */		;\
893	.align	32
894/*
895 * SPARC V9 IMPL. DEP. #109(1) and (2) and #110(1) and (2)
896 */
897
898/*
899 * synthesize for trap(): SFAR in %g2, SFSR in %g3
900 */
901#define	DMMU_EXC_LDDF_NOT_ALIGNED					\
902	ba,a,pt	%xcc, .dmmu_exc_lddf_not_aligned			;\
903	.align	32
904/*
905 * synthesize for trap(): SFAR in %g2, SFSR in %g3
906 */
907#define	DMMU_EXC_STDF_NOT_ALIGNED					\
908	ba,a,pt	%xcc, .dmmu_exc_stdf_not_aligned			;\
909	.align	32
910
911#if defined(cscope)
912/*
913 * Define labels to direct cscope quickly to labels that
914 * are generated by macro expansion of DTLB_MISS().
915 */
916	.global	tt0_dtlbmiss
917tt0_dtlbmiss:
918	.global	tt1_dtlbmiss
919tt1_dtlbmiss:
920	nop
921#endif
922
923/*
924 * Data miss handler (must be exactly 32 instructions)
925 *
926 * This handler is invoked only if the hypervisor has been instructed
927 * not to do any TSB walk.
928 *
929 * Kernel and invalid context cases are handled by the sfmmu_kdtlb_miss
930 * handler.
931 *
932 * User TLB miss handling depends upon whether a user process has one or
933 * two TSBs. User TSB information (physical base and size code) is kept
934 * in two dedicated scratchpad registers. Absence of a user TSB (primarily
935 * second TSB) is indicated by a negative value (-1) in that register.
936 */
937
938/*
939 * synthesize for miss handler: pseudo-tag access in %g2 (with context "type"
940 * (0=kernel, 1=invalid, or 2=user) rather than context ID)
941 */
942#define	DTLB_MISS(table_name)						;\
943	.global	table_name/**/_dtlbmiss					;\
944table_name/**/_dtlbmiss:						;\
945	GET_MMU_D_PTAGACC_CTXTYPE(%g2, %g3)	/* 8 instr */		;\
946	cmp	%g3, INVALID_CONTEXT					;\
947	ble,pn	%xcc, sfmmu_kdtlb_miss					;\
948	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
949	mov	SCRATCHPAD_UTSBREG2, %g1				;\
950	ldxa	[%g1]ASI_SCRATCHPAD, %g1	/* get 2nd tsbreg */	;\
951	brgez,pn %g1, sfmmu_udtlb_slowpath	/* branch if 2 TSBs */	;\
952	  nop								;\
953	GET_1ST_TSBE_PTR(%g2, %g1, %g4, %g5)	/* 11 instr */		;\
954	ba,pt	%xcc, sfmmu_udtlb_fastpath	/* no 4M TSB, miss */	;\
955	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
956	.align 128
957
958
959#if defined(cscope)
960/*
961 * Define labels to direct cscope quickly to labels that
962 * are generated by macro expansion of ITLB_MISS().
963 */
964	.global	tt0_itlbmiss
965tt0_itlbmiss:
966	.global	tt1_itlbmiss
967tt1_itlbmiss:
968	nop
969#endif
970
971/*
972 * Instruction miss handler.
973 *
974 * This handler is invoked only if the hypervisor has been instructed
975 * not to do any TSB walk.
976 *
977 * ldda instructions will have their ASI patched
978 * by sfmmu_patch_ktsb at runtime.
979 * MUST be EXACTLY 32 instructions or we'll break.
980 */
981
982/*
983 * synthesize for miss handler: TAG_ACCESS in %g2 (with context "type"
984 * (0=kernel, 1=invalid, or 2=user) rather than context ID)
985 */
986#define	ITLB_MISS(table_name)						 \
987	.global	table_name/**/_itlbmiss					;\
988table_name/**/_itlbmiss:						;\
989	GET_MMU_I_PTAGACC_CTXTYPE(%g2, %g3)	/* 8 instr */		;\
990	cmp	%g3, INVALID_CONTEXT					;\
991	ble,pn	%xcc, sfmmu_kitlb_miss					;\
992	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
993	mov	SCRATCHPAD_UTSBREG2, %g1				;\
994	ldxa	[%g1]ASI_SCRATCHPAD, %g1	/* get 2nd tsbreg */	;\
995	brgez,pn %g1, sfmmu_uitlb_slowpath	/* branch if 2 TSBs */	;\
996	  nop								;\
997	GET_1ST_TSBE_PTR(%g2, %g1, %g4, %g5)	/* 11 instr */		;\
998	ba,pt	%xcc, sfmmu_uitlb_fastpath	/* no 4M TSB, miss */	;\
999	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
1000	.align 128
1001
1002#define	DTSB_MISS \
1003	GOTO_TT(sfmmu_slow_dmmu_miss,trace_dmmu)
1004
1005#define	ITSB_MISS \
1006	GOTO_TT(sfmmu_slow_immu_miss,trace_immu)
1007
1008/*
1009 * This macro is the first level handler for fast protection faults.
1010 * It first demaps the tlb entry which generated the fault and then
1011 * attempts to set the modify bit on the hash.  It needs to be
1012 * exactly 32 instructions.
1013 */
1014/*
1015 * synthesize for miss handler: TAG_ACCESS in %g2 (with context "type"
1016 * (0=kernel, 1=invalid, or 2=user) rather than context ID)
1017 */
1018#define	DTLB_PROT							 \
1019	GET_MMU_D_PTAGACC_CTXTYPE(%g2, %g3)	/* 8 instr */		;\
1020	/*								;\
1021	 *   g2 = pseudo-tag access register (ctx type rather than ctx ID) ;\
1022	 *   g3 = ctx type (0, 1, or 2)					;\
1023	 */								;\
1024	TT_TRACE(trace_dataprot)	/* 2 instr ifdef TRAPTRACE */	;\
1025					/* clobbers g1 and g6 XXXQ? */	;\
1026	brnz,pt %g3, sfmmu_uprot_trap		/* user trap */		;\
1027	  nop								;\
1028	ba,a,pt	%xcc, sfmmu_kprot_trap		/* kernel trap */	;\
1029	.align 128
1030
1031#define	DMMU_EXCEPTION_TL1						;\
1032	ba,a,pt	%xcc, mmu_trap_tl1					;\
1033	.align 32
1034
1035#define	MISALIGN_ADDR_TL1						;\
1036	ba,a,pt	%xcc, mmu_trap_tl1					;\
1037	.align 32
1038
1039/*
1040 * Trace a tsb hit
1041 * g1 = tsbe pointer (in/clobbered)
1042 * g2 = tag access register (in)
1043 * g3 - g4 = scratch (clobbered)
1044 * g5 = tsbe data (in)
1045 * g6 = scratch (clobbered)
1046 * g7 = pc we jumped here from (in)
1047 * ttextra = value to OR in to trap type (%tt) (in)
1048 */
1049#ifdef TRAPTRACE
1050#define TRACE_TSBHIT(ttextra)						 \
1051	membar	#Sync							;\
1052	sethi	%hi(FLUSH_ADDR), %g6					;\
1053	flush	%g6							;\
1054	TRACE_PTR(%g3, %g6)						;\
1055	GET_TRACE_TICK(%g6, %g4)					;\
1056	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi				;\
1057	stna	%g2, [%g3 + TRAP_ENT_SP]%asi	/* tag access */	;\
1058	stna	%g5, [%g3 + TRAP_ENT_F1]%asi	/* tsb data */		;\
1059	rdpr	%tnpc, %g6						;\
1060	stna	%g6, [%g3 + TRAP_ENT_F2]%asi				;\
1061	stna	%g1, [%g3 + TRAP_ENT_F3]%asi	/* tsb pointer */	;\
1062	stna	%g0, [%g3 + TRAP_ENT_F4]%asi				;\
1063	rdpr	%tpc, %g6						;\
1064	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi				;\
1065	TRACE_SAVE_TL_GL_REGS(%g3, %g6)					;\
1066	rdpr	%tt, %g6						;\
1067	or	%g6, (ttextra), %g1					;\
1068	stha	%g1, [%g3 + TRAP_ENT_TT]%asi				;\
1069	MMU_FAULT_STATUS_AREA(%g4)					;\
1070	mov	MMFSA_D_ADDR, %g1					;\
1071	cmp	%g6, FAST_IMMU_MISS_TT					;\
1072	move	%xcc, MMFSA_I_ADDR, %g1					;\
1073	cmp	%g6, T_INSTR_MMU_MISS					;\
1074	move	%xcc, MMFSA_I_ADDR, %g1					;\
1075	ldx	[%g4 + %g1], %g1					;\
1076	stxa	%g1, [%g3 + TRAP_ENT_TSTATE]%asi /* fault addr */	;\
1077	mov	MMFSA_D_CTX, %g1					;\
1078	cmp	%g6, FAST_IMMU_MISS_TT					;\
1079	move	%xcc, MMFSA_I_CTX, %g1					;\
1080	cmp	%g6, T_INSTR_MMU_MISS					;\
1081	move	%xcc, MMFSA_I_CTX, %g1					;\
1082	ldx	[%g4 + %g1], %g1					;\
1083	stna	%g1, [%g3 + TRAP_ENT_TR]%asi				;\
1084	TRACE_NEXT(%g3, %g4, %g6)
1085#else
1086#define TRACE_TSBHIT(ttextra)
1087#endif
1088
1089
1090#if defined(lint)
1091
1092struct scb	trap_table;
1093struct scb	scb;		/* trap_table/scb are the same object */
1094
1095#else /* lint */
1096
1097/*
1098 * =======================================================================
1099 *		SPARC V9 TRAP TABLE
1100 *
1101 * The trap table is divided into two halves: the first half is used when
1102 * taking traps when TL=0; the second half is used when taking traps from
1103 * TL>0. Note that handlers in the second half of the table might not be able
1104 * to make the same assumptions as handlers in the first half of the table.
1105 *
1106 * Worst case trap nesting so far:
1107 *
1108 *	at TL=0 client issues software trap requesting service
1109 *	at TL=1 nucleus wants a register window
1110 *	at TL=2 register window clean/spill/fill takes a TLB miss
1111 *	at TL=3 processing TLB miss
1112 *	at TL=4 handle asynchronous error
1113 *
1114 * Note that a trap from TL=4 to TL=5 places Spitfire in "RED mode".
1115 *
1116 * =======================================================================
1117 */
1118	.section ".text"
1119	.align	4
1120	.global trap_table, scb, trap_table0, trap_table1, etrap_table
1121	.type	trap_table, #object
1122	.type	trap_table0, #object
1123	.type	trap_table1, #object
1124	.type	scb, #object
1125trap_table:
1126scb:
1127trap_table0:
1128	/* hardware traps */
1129	NOT;				/* 000	reserved */
1130	RED;				/* 001	power on reset */
1131	WATCHDOG_RESET;			/* 002	watchdog reset */
1132	RED;				/* 003	externally initiated reset */
1133	RED;				/* 004	software initiated reset */
1134	RED;				/* 005	red mode exception */
1135	NOT; NOT;			/* 006 - 007 reserved */
1136	IMMU_EXCEPTION;			/* 008	instruction access exception */
1137	ITSB_MISS;			/* 009	instruction access MMU miss */
1138 	NOT;				/* 00A  reserved */
1139	NOT; NOT4;			/* 00B - 00F reserved */
1140	ILLTRAP_INSTR;			/* 010	illegal instruction */
1141	TRAP(T_PRIV_INSTR);		/* 011	privileged opcode */
1142	TRAP(T_UNIMP_LDD);		/* 012	unimplemented LDD */
1143	TRAP(T_UNIMP_STD);		/* 013	unimplemented STD */
1144	NOT4; NOT4; NOT4;		/* 014 - 01F reserved */
1145	FP_DISABLED_TRAP;		/* 020	fp disabled */
1146	FP_IEEE_TRAP;			/* 021	fp exception ieee 754 */
1147	FP_TRAP;			/* 022	fp exception other */
1148	TAG_OVERFLOW;			/* 023	tag overflow */
1149	CLEAN_WINDOW;			/* 024 - 027 clean window */
1150	DIV_BY_ZERO;			/* 028	division by zero */
1151	NOT;				/* 029	internal processor error */
1152	NOT; NOT; NOT4;			/* 02A - 02F reserved */
1153	DMMU_EXCEPTION;			/* 030	data access exception */
1154	DTSB_MISS;			/* 031	data access MMU miss */
1155	NOT;				/* 032  reserved */
1156	NOT;				/* 033	data access protection */
1157	DMMU_EXC_AG_NOT_ALIGNED;	/* 034	mem address not aligned */
1158	DMMU_EXC_LDDF_NOT_ALIGNED;	/* 035	LDDF mem address not aligned */
1159	DMMU_EXC_STDF_NOT_ALIGNED;	/* 036	STDF mem address not aligned */
1160	DMMU_EXC_AG_PRIV;		/* 037	privileged action */
1161	NOT;				/* 038	LDQF mem address not aligned */
1162	NOT;				/* 039	STQF mem address not aligned */
1163	NOT; NOT; NOT4;			/* 03A - 03F reserved */
1164	NOT;				/* 040	async data error */
1165	LEVEL_INTERRUPT(1);		/* 041	interrupt level 1 */
1166	LEVEL_INTERRUPT(2);		/* 042	interrupt level 2 */
1167	LEVEL_INTERRUPT(3);		/* 043	interrupt level 3 */
1168	LEVEL_INTERRUPT(4);		/* 044	interrupt level 4 */
1169	LEVEL_INTERRUPT(5);		/* 045	interrupt level 5 */
1170	LEVEL_INTERRUPT(6);		/* 046	interrupt level 6 */
1171	LEVEL_INTERRUPT(7);		/* 047	interrupt level 7 */
1172	LEVEL_INTERRUPT(8);		/* 048	interrupt level 8 */
1173	LEVEL_INTERRUPT(9);		/* 049	interrupt level 9 */
1174	LEVEL_INTERRUPT(10);		/* 04A	interrupt level 10 */
1175	LEVEL_INTERRUPT(11);		/* 04B	interrupt level 11 */
1176	LEVEL_INTERRUPT(12);		/* 04C	interrupt level 12 */
1177	LEVEL_INTERRUPT(13);		/* 04D	interrupt level 13 */
1178	LEVEL14_INTERRUPT;		/* 04E	interrupt level 14 */
1179	LEVEL15_INTERRUPT;		/* 04F	interrupt level 15 */
1180	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F reserved */
1181	NOT;				/* 060	interrupt vector */
1182	GOTO(kmdb_trap);		/* 061	PA watchpoint */
1183	GOTO(kmdb_trap);		/* 062	VA watchpoint */
1184	NOT;				/* 063	reserved */
1185	ITLB_MISS(tt0);			/* 064	instruction access MMU miss */
1186	DTLB_MISS(tt0);			/* 068	data access MMU miss */
1187	DTLB_PROT;			/* 06C	data access protection */
1188	NOT;				/* 070  reserved */
1189	NOT;				/* 071  reserved */
1190	NOT;				/* 072  reserved */
1191	NOT;				/* 073  reserved */
1192	NOT4; NOT4			/* 074 - 07B reserved */
1193	CPU_MONDO;			/* 07C	cpu_mondo */
1194	DEV_MONDO;			/* 07D	dev_mondo */
1195	GOTO_TT(resumable_error, trace_gen);	/* 07E  resumable error */
1196	GOTO_TT(nonresumable_error, trace_gen);	/* 07F  non-reasumable error */
1197	NOT4;				/* 080	spill 0 normal */
1198	SPILL_32bit_asi(ASI_AIUP,sn0);	/* 084	spill 1 normal */
1199	SPILL_64bit_asi(ASI_AIUP,sn0);	/* 088	spill 2 normal */
1200	SPILL_32clean(ASI_AIUP,sn0);	/* 08C	spill 3 normal */
1201	SPILL_64clean(ASI_AIUP,sn0);	/* 090	spill 4 normal */
1202	SPILL_32bit(not);		/* 094	spill 5 normal */
1203	SPILL_64bit(not);		/* 098	spill 6 normal */
1204	SPILL_mixed;			/* 09C	spill 7 normal */
1205	NOT4;				/* 0A0	spill 0 other */
1206	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0A4	spill 1 other */
1207	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0A8	spill 2 other */
1208	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0AC	spill 3 other */
1209	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0B0	spill 4 other */
1210	NOT4;				/* 0B4	spill 5 other */
1211	NOT4;				/* 0B8	spill 6 other */
1212	NOT4;				/* 0BC	spill 7 other */
1213	NOT4;				/* 0C0	fill 0 normal */
1214	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0C4	fill 1 normal */
1215	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0C8	fill 2 normal */
1216	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0CC	fill 3 normal */
1217	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0D0	fill 4 normal */
1218	FILL_32bit(not);		/* 0D4	fill 5 normal */
1219	FILL_64bit(not);		/* 0D8	fill 6 normal */
1220	FILL_mixed;			/* 0DC	fill 7 normal */
1221	NOT4;				/* 0E0	fill 0 other */
1222	NOT4;				/* 0E4	fill 1 other */
1223	NOT4;				/* 0E8	fill 2 other */
1224	NOT4;				/* 0EC	fill 3 other */
1225	NOT4;				/* 0F0	fill 4 other */
1226	NOT4;				/* 0F4	fill 5 other */
1227	NOT4;				/* 0F8	fill 6 other */
1228	NOT4;				/* 0FC	fill 7 other */
1229	/* user traps */
1230	GOTO(syscall_trap_4x);		/* 100	old system call */
1231	TRAP(T_BREAKPOINT);		/* 101	user breakpoint */
1232	TRAP(T_DIV0);			/* 102	user divide by zero */
1233	GOTO(.flushw);			/* 103	flush windows */
1234	GOTO(.clean_windows);		/* 104	clean windows */
1235	BAD;				/* 105	range check ?? */
1236	GOTO(.fix_alignment);		/* 106	do unaligned references */
1237	BAD;				/* 107	unused */
1238	SYSCALL_TRAP32;			/* 108	ILP32 system call on LP64 */
1239	GOTO(set_trap0_addr);		/* 109	set trap0 address */
1240	BAD; BAD; BAD4;			/* 10A - 10F unused */
1241	TRP4; TRP4; TRP4; TRP4;		/* 110 - 11F V9 user trap handlers */
1242	GOTO(.getcc);			/* 120	get condition codes */
1243	GOTO(.setcc);			/* 121	set condition codes */
1244	GOTO(.getpsr);			/* 122	get psr */
1245	GOTO(.setpsr);			/* 123	set psr (some fields) */
1246	GOTO(get_timestamp);		/* 124	get timestamp */
1247	GOTO(get_virtime);		/* 125	get lwp virtual time */
1248	PRIV(self_xcall);		/* 126	self xcall */
1249	GOTO(get_hrestime);		/* 127	get hrestime */
1250	BAD;				/* 128	ST_SETV9STACK */
1251	GOTO(.getlgrp);			/* 129  get lgrpid */
1252	BAD; BAD; BAD4;			/* 12A - 12F unused */
1253	BAD4; BAD4; 			/* 130 - 137 unused */
1254	DTRACE_PID;			/* 138  dtrace pid tracing provider */
1255	BAD;				/* 139  unused */
1256	DTRACE_RETURN;			/* 13A	dtrace pid return probe */
1257	BAD; BAD4;			/* 13B - 13F unused */
1258	SYSCALL_TRAP;			/* 140  LP64 system call */
1259	SYSCALL(nosys);			/* 141  unused system call trap */
1260#ifdef DEBUG_USER_TRAPTRACECTL
1261	GOTO(.traptrace_freeze);	/* 142  freeze traptrace */
1262	GOTO(.traptrace_unfreeze);	/* 143  unfreeze traptrace */
1263#else
1264	SYSCALL(nosys);			/* 142  unused system call trap */
1265	SYSCALL(nosys);			/* 143  unused system call trap */
1266#endif
1267	BAD4; BAD4; BAD4;		/* 144 - 14F unused */
1268	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1269	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1270	BAD;				/* 170 - unused */
1271	BAD;				/* 171 - unused */
1272	BAD; BAD;			/* 172 - 173 unused */
1273	BAD4; BAD4;			/* 174 - 17B unused */
1274#ifdef	PTL1_PANIC_DEBUG
1275	mov PTL1_BAD_DEBUG, %g1; GOTO(ptl1_panic);
1276					/* 17C	test ptl1_panic */
1277#else
1278	BAD;				/* 17C  unused */
1279#endif	/* PTL1_PANIC_DEBUG */
1280	PRIV(kmdb_trap);		/* 17D	kmdb enter (L1-A) */
1281	PRIV(kmdb_trap);		/* 17E	kmdb breakpoint */
1282	PRIV(obp_bpt);			/* 17F	obp breakpoint */
1283	/* reserved */
1284	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1285	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1286	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1287	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1288	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1289	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1290	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1291	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1292	.size	trap_table0, (.-trap_table0)
1293trap_table1:
1294	NOT4; NOT4;			/* 000 - 007 unused */
1295	NOT;				/* 008	instruction access exception */
1296	ITSB_MISS;			/* 009	instruction access MMU miss */
1297 	NOT;				/* 00A  reserved */
1298	NOT; NOT4;			/* 00B - 00F unused */
1299	NOT4; NOT4; NOT4; NOT4;		/* 010 - 01F unused */
1300	NOT4;				/* 020 - 023 unused */
1301	CLEAN_WINDOW;			/* 024 - 027 clean window */
1302	NOT4; NOT4;			/* 028 - 02F unused */
1303	DMMU_EXCEPTION_TL1;		/* 030 	data access exception */
1304	DTSB_MISS;			/* 031  data access MMU miss */
1305	NOT;				/* 032  reserved */
1306	NOT;				/* 033	unused */
1307	MISALIGN_ADDR_TL1;		/* 034	mem address not aligned */
1308	NOT; NOT; NOT; NOT4; NOT4	/* 035 - 03F unused */
1309	NOT4; NOT4; NOT4; NOT4;		/* 040 - 04F unused */
1310	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F unused */
1311	NOT;				/* 060	unused */
1312	GOTO(kmdb_trap_tl1);		/* 061	PA watchpoint */
1313	GOTO(kmdb_trap_tl1);		/* 062	VA watchpoint */
1314	NOT;				/* 063	reserved */
1315	ITLB_MISS(tt1);			/* 064	instruction access MMU miss */
1316	DTLB_MISS(tt1);			/* 068	data access MMU miss */
1317	DTLB_PROT;			/* 06C	data access protection */
1318	NOT;				/* 070  reserved */
1319	NOT;				/* 071  reserved */
1320	NOT;				/* 072  reserved */
1321	NOT;				/* 073  reserved */
1322	NOT4; NOT4;			/* 074 - 07B reserved */
1323	NOT;				/* 07C  reserved */
1324	NOT;				/* 07D  reserved */
1325	NOT;				/* 07E  resumable error */
1326	GOTO_TT(nonresumable_error, trace_gen);	/* 07F  nonresumable error */
1327	NOTP4;				/* 080	spill 0 normal */
1328	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 084	spill 1 normal */
1329	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 088	spill 2 normal */
1330	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 08C	spill 3 normal */
1331	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 090	spill 4 normal */
1332	NOTP4;				/* 094	spill 5 normal */
1333	SPILL_64bit_ktt1(sk);		/* 098	spill 6 normal */
1334	SPILL_mixed_ktt1(sk);		/* 09C	spill 7 normal */
1335	NOTP4;				/* 0A0	spill 0 other */
1336	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0A4  spill 1 other */
1337	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0A8	spill 2 other */
1338	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0AC	spill 3 other */
1339	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0B0  spill 4 other */
1340	NOTP4;				/* 0B4  spill 5 other */
1341	NOTP4;				/* 0B8  spill 6 other */
1342	NOTP4;				/* 0BC  spill 7 other */
1343	NOT4;				/* 0C0	fill 0 normal */
1344	NOT4;				/* 0C4	fill 1 normal */
1345	NOT4;				/* 0C8	fill 2 normal */
1346	NOT4;				/* 0CC	fill 3 normal */
1347	NOT4;				/* 0D0	fill 4 normal */
1348	NOT4;				/* 0D4	fill 5 normal */
1349	NOT4;				/* 0D8	fill 6 normal */
1350	NOT4;				/* 0DC	fill 7 normal */
1351	NOT4; NOT4; NOT4; NOT4;		/* 0E0 - 0EF unused */
1352	NOT4; NOT4; NOT4; NOT4;		/* 0F0 - 0FF unused */
1353/*
1354 * Code running at TL>0 does not use soft traps, so
1355 * we can truncate the table here.
1356 * However:
1357 * sun4v uses (hypervisor) ta instructions at TL > 0, so
1358 * provide a safety net for now.
1359 */
1360	/* soft traps */
1361	BAD4; BAD4; BAD4; BAD4;		/* 100 - 10F unused */
1362	BAD4; BAD4; BAD4; BAD4;		/* 110 - 11F unused */
1363	BAD4; BAD4; BAD4; BAD4;		/* 120 - 12F unused */
1364	BAD4; BAD4; BAD4; BAD4;		/* 130 - 13F unused */
1365	BAD4; BAD4; BAD4; BAD4;		/* 140 - 14F unused */
1366	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1367	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1368	BAD4; BAD4; BAD4; BAD4;		/* 170 - 17F unused */
1369	/* reserved */
1370	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1371	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1372	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1373	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1374	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1375	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1376	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1377	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1378etrap_table:
1379	.size	trap_table1, (.-trap_table1)
1380	.size	trap_table, (.-trap_table)
1381	.size	scb, (.-scb)
1382
1383/*
1384 * We get to exec_fault in the case of an instruction miss and tte
1385 * has no execute bit set.  We go to tl0 to handle it.
1386 *
1387 * g1 = tsbe pointer (in/clobbered)
1388 * g2 = tag access register (in)
1389 * g3 - g4 = scratch (clobbered)
1390 * g5 = tsbe data (in)
1391 * g6 = scratch (clobbered)
1392 * g7 = pc we jumped here from (in)
1393 */
1394/*
1395 * synthesize for miss handler: TAG_ACCESS in %g2 (with context "type"
1396 * (0=kernel, 1=invalid, or 2=user) rather than context ID)
1397 */
1398	ALTENTRY(exec_fault)
1399	TRACE_TSBHIT(TT_MMU_EXEC)
1400	MMU_FAULT_STATUS_AREA(%g4)
1401	ldx	[%g4 + MMFSA_I_ADDR], %g2	/* g2 = address */
1402	ldx	[%g4 + MMFSA_I_CTX], %g3	/* g3 = ctx */
1403	srlx	%g2, MMU_PAGESHIFT, %g2		! align address to page boundry
1404	cmp	%g3, USER_CONTEXT_TYPE
1405	sllx	%g2, MMU_PAGESHIFT, %g2
1406	movgu	%icc, USER_CONTEXT_TYPE, %g3
1407	or	%g2, %g3, %g2			/* TAG_ACCESS */
1408	mov	T_INSTR_MMU_MISS, %g3		! arg2 = traptype
1409	set	trap, %g1
1410	ba,pt	%xcc, sys_trap
1411	  mov	-1, %g4
1412
1413.mmu_exception_not_aligned:
1414	/* %g2 = sfar, %g3 = sfsr */
1415	rdpr	%tstate, %g1
1416	btst	TSTATE_PRIV, %g1
1417	bnz,pn	%icc, 2f
1418	nop
1419	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1420	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1421	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1422	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1423	brz,pt	%g5, 2f
1424	nop
1425	ldn	[%g5 + P_UTRAP15], %g5			! unaligned utrap?
1426	brz,pn	%g5, 2f
1427	nop
1428	btst	1, %sp
1429	bz,pt	%xcc, 1f				! 32 bit user program
1430	nop
1431	ba,pt	%xcc, .setup_v9utrap			! 64 bit user program
1432	nop
14331:
1434	ba,pt	%xcc, .setup_utrap
1435	or	%g2, %g0, %g7
14362:
1437	ba,pt	%xcc, .mmu_exception_end
1438	mov	T_ALIGNMENT, %g1
1439
1440.mmu_priv_exception:
1441	rdpr	%tstate, %g1
1442	btst	TSTATE_PRIV, %g1
1443	bnz,pn	%icc, 1f
1444	nop
1445	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1446	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1447	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1448	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1449	brz,pt	%g5, 1f
1450	nop
1451	ldn	[%g5 + P_UTRAP16], %g5
1452	brnz,pt	%g5, .setup_v9utrap
1453	nop
14541:
1455	mov	T_PRIV_INSTR, %g1
1456
1457.mmu_exception_end:
1458	CPU_INDEX(%g4, %g5)
1459	set	cpu_core, %g5
1460	sllx	%g4, CPU_CORE_SHIFT, %g4
1461	add	%g4, %g5, %g4
1462	lduh	[%g4 + CPUC_DTRACE_FLAGS], %g5
1463	andcc	%g5, CPU_DTRACE_NOFAULT, %g0
1464	bz	1f
1465	or	%g5, CPU_DTRACE_BADADDR, %g5
1466	stuh	%g5, [%g4 + CPUC_DTRACE_FLAGS]
1467	done
1468
14691:
1470	sllx	%g3, 32, %g3
1471	or	%g3, %g1, %g3
1472	set	trap, %g1
1473	ba,pt	%xcc, sys_trap
1474	sub	%g0, 1, %g4
1475
1476.fp_disabled:
1477	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1478	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1479	rdpr	%tstate, %g4
1480	btst	TSTATE_PRIV, %g4
1481	bnz,a,pn %icc, ptl1_panic
1482	  mov	PTL1_BAD_FPTRAP, %g1
1483
1484	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1485	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1486	brz,a,pt %g5, 2f
1487	  nop
1488	ldn	[%g5 + P_UTRAP7], %g5			! fp_disabled utrap?
1489	brz,a,pn %g5, 2f
1490	  nop
1491	btst	1, %sp
1492	bz,a,pt	%xcc, 1f				! 32 bit user program
1493	  nop
1494	ba,a,pt	%xcc, .setup_v9utrap			! 64 bit user program
1495	  nop
14961:
1497	ba,pt	%xcc, .setup_utrap
1498	  or	%g0, %g0, %g7
14992:
1500	set	fp_disabled, %g1
1501	ba,pt	%xcc, sys_trap
1502	  sub	%g0, 1, %g4
1503
1504.fp_ieee_exception:
1505	rdpr	%tstate, %g1
1506	btst	TSTATE_PRIV, %g1
1507	bnz,a,pn %icc, ptl1_panic
1508	  mov	PTL1_BAD_FPTRAP, %g1
1509	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1510	stx	%fsr, [%g1 + CPU_TMP1]
1511	ldx	[%g1 + CPU_TMP1], %g2
1512	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1513	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1514	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1515	brz,a,pt %g5, 1f
1516	  nop
1517	ldn	[%g5 + P_UTRAP8], %g5
1518	brnz,a,pt %g5, .setup_v9utrap
1519	  nop
15201:
1521	set	_fp_ieee_exception, %g1
1522	ba,pt	%xcc, sys_trap
1523	  sub	%g0, 1, %g4
1524
1525/*
1526 * Register Inputs:
1527 *	%g5		user trap handler
1528 *	%g7		misaligned addr - for alignment traps only
1529 */
1530.setup_utrap:
1531	set	trap, %g1			! setup in case we go
1532	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1533	sub	%g0, 1, %g4			! the save instruction below
1534
1535	/*
1536	 * If the DTrace pid provider is single stepping a copied-out
1537	 * instruction, t->t_dtrace_step will be set. In that case we need
1538	 * to abort the single-stepping (since execution of the instruction
1539	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1540	 */
1541	save	%sp, -SA(MINFRAME32), %sp	! window for trap handler
1542	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1543	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1544	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1545	rdpr	%tnpc, %l2			! arg1 == tnpc
1546	brz,pt	%g2, 1f
1547	rdpr	%tpc, %l1			! arg0 == tpc
1548
1549	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1550	ldn	[%g1 + T_DTRACE_NPC], %l2	! arg1 = t->t_dtrace_npc (step)
1551	brz,pt	%g2, 1f
1552	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1553	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
15541:
1555	mov	%g7, %l3			! arg2 == misaligned address
1556
1557	rdpr	%tstate, %g1			! cwp for trap handler
1558	rdpr	%cwp, %g4
1559	bclr	TSTATE_CWP_MASK, %g1
1560	wrpr	%g1, %g4, %tstate
1561	wrpr	%g0, %g5, %tnpc			! trap handler address
1562	FAST_TRAP_DONE
1563	/* NOTREACHED */
1564
1565.check_v9utrap:
1566	rdpr	%tstate, %g1
1567	btst	TSTATE_PRIV, %g1
1568	bnz,a,pn %icc, 3f
1569	  nop
1570	CPU_ADDR(%g4, %g1)				! load CPU struct addr
1571	ldn	[%g4 + CPU_THREAD], %g5			! load thread pointer
1572	ldn	[%g5 + T_PROCP], %g5			! load proc pointer
1573	ldn	[%g5 + P_UTRAPS], %g5			! are there utraps?
1574
1575	cmp	%g3, T_SOFTWARE_TRAP
1576	bne,a,pt %icc, 1f
1577	  nop
1578
1579	brz,pt %g5, 3f			! if p_utraps == NULL goto trap()
1580	  rdpr	%tt, %g3		! delay - get actual hw trap type
1581
1582	sub	%g3, 254, %g1		! UT_TRAP_INSTRUCTION_16 = p_utraps[18]
1583	ba,pt	%icc, 2f
1584	  smul	%g1, CPTRSIZE, %g2
15851:
1586	brz,a,pt %g5, 3f		! if p_utraps == NULL goto trap()
1587	  nop
1588
1589	cmp	%g3, T_UNIMP_INSTR
1590	bne,a,pt %icc, 2f
1591	  nop
1592
1593	mov	1, %g1
1594	st	%g1, [%g4 + CPU_TL1_HDLR] ! set CPU_TL1_HDLR
1595	rdpr	%tpc, %g1		! ld trapping instruction using
1596	lduwa	[%g1]ASI_AIUP, %g1	! "AS IF USER" ASI which could fault
1597	st	%g0, [%g4 + CPU_TL1_HDLR] ! clr CPU_TL1_HDLR
1598
1599	sethi	%hi(0xc1c00000), %g4	! setup mask for illtrap instruction
1600	andcc	%g1, %g4, %g4		! and instruction with mask
1601	bnz,a,pt %icc, 3f		! if %g4 == zero, %g1 is an ILLTRAP
1602	  nop				! fall thru to setup
16032:
1604	ldn	[%g5 + %g2], %g5
1605	brnz,a,pt %g5, .setup_v9utrap
1606	  nop
16073:
1608	set	trap, %g1
1609	ba,pt	%xcc, sys_trap
1610	  sub	%g0, 1, %g4
1611	/* NOTREACHED */
1612
1613/*
1614 * Register Inputs:
1615 *	%g5		user trap handler
1616 */
1617.setup_v9utrap:
1618	set	trap, %g1			! setup in case we go
1619	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1620	sub	%g0, 1, %g4			! the save instruction below
1621
1622	/*
1623	 * If the DTrace pid provider is single stepping a copied-out
1624	 * instruction, t->t_dtrace_step will be set. In that case we need
1625	 * to abort the single-stepping (since execution of the instruction
1626	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1627	 */
1628	save	%sp, -SA(MINFRAME64), %sp	! window for trap handler
1629	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1630	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1631	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1632	rdpr	%tnpc, %l7			! arg1 == tnpc
1633	brz,pt	%g2, 1f
1634	rdpr	%tpc, %l6			! arg0 == tpc
1635
1636	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1637	ldn	[%g1 + T_DTRACE_NPC], %l7	! arg1 == t->t_dtrace_npc (step)
1638	brz,pt	%g2, 1f
1639	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1640	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
16411:
1642	rdpr	%tstate, %g2			! cwp for trap handler
1643	rdpr	%cwp, %g4
1644	bclr	TSTATE_CWP_MASK, %g2
1645	wrpr	%g2, %g4, %tstate
1646
1647	ldn	[%g1 + T_PROCP], %g4		! load proc pointer
1648	ldn	[%g4 + P_AS], %g4		! load as pointer
1649	ldn	[%g4 + A_USERLIMIT], %g4	! load as userlimit
1650	cmp	%l7, %g4			! check for single-step set
1651	bne,pt	%xcc, 4f
1652	  nop
1653	ldn	[%g1 + T_LWP], %g1		! load klwp pointer
1654	ld	[%g1 + PCB_STEP], %g4		! load single-step flag
1655	cmp	%g4, STEP_ACTIVE		! step flags set in pcb?
1656	bne,pt	%icc, 4f
1657	  nop
1658	stn	%g5, [%g1 + PCB_TRACEPC]	! save trap handler addr in pcb
1659	mov	%l7, %g4			! on entry to precise user trap
1660	add	%l6, 4, %l7			! handler, %l6 == pc, %l7 == npc
1661						! at time of trap
1662	wrpr	%g0, %g4, %tnpc			! generate FLTBOUNDS,
1663						! %g4 == userlimit
1664	FAST_TRAP_DONE
1665	/* NOTREACHED */
16664:
1667	wrpr	%g0, %g5, %tnpc			! trap handler address
1668	FAST_TRAP_DONE_CHK_INTR
1669	/* NOTREACHED */
1670
1671.fp_exception:
1672	CPU_ADDR(%g1, %g4)
1673	stx	%fsr, [%g1 + CPU_TMP1]
1674	ldx	[%g1 + CPU_TMP1], %g2
1675
1676	/*
1677	 * Cheetah takes unfinished_FPop trap for certain range of operands
1678	 * to the "fitos" instruction. Instead of going through the slow
1679	 * software emulation path, we try to simulate the "fitos" instruction
1680	 * via "fitod" and "fdtos" provided the following conditions are met:
1681	 *
1682	 *	fpu_exists is set (if DEBUG)
1683	 *	not in privileged mode
1684	 *	ftt is unfinished_FPop
1685	 *	NXM IEEE trap is not enabled
1686	 *	instruction at %tpc is "fitos"
1687	 *
1688	 *  Usage:
1689	 *	%g1	per cpu address
1690	 *	%g2	%fsr
1691	 *	%g6	user instruction
1692	 *
1693	 * Note that we can take a memory access related trap while trying
1694	 * to fetch the user instruction. Therefore, we set CPU_TL1_HDLR
1695	 * flag to catch those traps and let the SFMMU code deal with page
1696	 * fault and data access exception.
1697	 */
1698#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
1699	sethi	%hi(fpu_exists), %g7
1700	ld	[%g7 + %lo(fpu_exists)], %g7
1701	brz,pn %g7, .fp_exception_cont
1702	  nop
1703#endif
1704	rdpr	%tstate, %g7			! branch if in privileged mode
1705	btst	TSTATE_PRIV, %g7
1706	bnz,pn	%xcc, .fp_exception_cont
1707	srl	%g2, FSR_FTT_SHIFT, %g7		! extract ftt from %fsr
1708	and	%g7, (FSR_FTT>>FSR_FTT_SHIFT), %g7
1709	cmp	%g7, FTT_UNFIN
1710	set	FSR_TEM_NX, %g5
1711	bne,pn	%xcc, .fp_exception_cont	! branch if NOT unfinished_FPop
1712	  andcc	%g2, %g5, %g0
1713	bne,pn	%xcc, .fp_exception_cont	! branch if FSR_TEM_NX enabled
1714	  rdpr	%tpc, %g5			! get faulting PC
1715
1716	or	%g0, 1, %g7
1717	st	%g7, [%g1 + CPU_TL1_HDLR]	! set tl1_hdlr flag
1718	lda	[%g5]ASI_USER, %g6		! get user's instruction
1719	st	%g0, [%g1 + CPU_TL1_HDLR]	! clear tl1_hdlr flag
1720
1721	set	FITOS_INSTR_MASK, %g7
1722	and	%g6, %g7, %g7
1723	set	FITOS_INSTR, %g5
1724	cmp	%g7, %g5
1725	bne,pn	%xcc, .fp_exception_cont	! branch if not FITOS_INSTR
1726	 nop
1727
1728	/*
1729	 * This is unfinished FPops trap for "fitos" instruction. We
1730	 * need to simulate "fitos" via "fitod" and "fdtos" instruction
1731	 * sequence.
1732	 *
1733	 * We need a temporary FP register to do the conversion. Since
1734	 * both source and destination operands for the "fitos" instruction
1735	 * have to be within %f0-%f31, we use an FP register from the upper
1736	 * half to guarantee that it won't collide with the source or the
1737	 * dest operand. However, we do have to save and restore its value.
1738	 *
1739	 * We use %d62 as a temporary FP register for the conversion and
1740	 * branch to appropriate instruction within the conversion tables
1741	 * based upon the rs2 and rd values.
1742	 */
1743
1744	std	%d62, [%g1 + CPU_TMP1]		! save original value
1745
1746	srl	%g6, FITOS_RS2_SHIFT, %g7
1747	and	%g7, FITOS_REG_MASK, %g7
1748	set	_fitos_fitod_table, %g4
1749	sllx	%g7, 2, %g7
1750	jmp	%g4 + %g7
1751	  ba,pt	%xcc, _fitos_fitod_done
1752	.empty
1753
1754_fitos_fitod_table:
1755	  fitod	%f0, %d62
1756	  fitod	%f1, %d62
1757	  fitod	%f2, %d62
1758	  fitod	%f3, %d62
1759	  fitod	%f4, %d62
1760	  fitod	%f5, %d62
1761	  fitod	%f6, %d62
1762	  fitod	%f7, %d62
1763	  fitod	%f8, %d62
1764	  fitod	%f9, %d62
1765	  fitod	%f10, %d62
1766	  fitod	%f11, %d62
1767	  fitod	%f12, %d62
1768	  fitod	%f13, %d62
1769	  fitod	%f14, %d62
1770	  fitod	%f15, %d62
1771	  fitod	%f16, %d62
1772	  fitod	%f17, %d62
1773	  fitod	%f18, %d62
1774	  fitod	%f19, %d62
1775	  fitod	%f20, %d62
1776	  fitod	%f21, %d62
1777	  fitod	%f22, %d62
1778	  fitod	%f23, %d62
1779	  fitod	%f24, %d62
1780	  fitod	%f25, %d62
1781	  fitod	%f26, %d62
1782	  fitod	%f27, %d62
1783	  fitod	%f28, %d62
1784	  fitod	%f29, %d62
1785	  fitod	%f30, %d62
1786	  fitod	%f31, %d62
1787_fitos_fitod_done:
1788
1789	/*
1790	 * Now convert data back into single precision
1791	 */
1792	srl	%g6, FITOS_RD_SHIFT, %g7
1793	and	%g7, FITOS_REG_MASK, %g7
1794	set	_fitos_fdtos_table, %g4
1795	sllx	%g7, 2, %g7
1796	jmp	%g4 + %g7
1797	  ba,pt	%xcc, _fitos_fdtos_done
1798	.empty
1799
1800_fitos_fdtos_table:
1801	  fdtos	%d62, %f0
1802	  fdtos	%d62, %f1
1803	  fdtos	%d62, %f2
1804	  fdtos	%d62, %f3
1805	  fdtos	%d62, %f4
1806	  fdtos	%d62, %f5
1807	  fdtos	%d62, %f6
1808	  fdtos	%d62, %f7
1809	  fdtos	%d62, %f8
1810	  fdtos	%d62, %f9
1811	  fdtos	%d62, %f10
1812	  fdtos	%d62, %f11
1813	  fdtos	%d62, %f12
1814	  fdtos	%d62, %f13
1815	  fdtos	%d62, %f14
1816	  fdtos	%d62, %f15
1817	  fdtos	%d62, %f16
1818	  fdtos	%d62, %f17
1819	  fdtos	%d62, %f18
1820	  fdtos	%d62, %f19
1821	  fdtos	%d62, %f20
1822	  fdtos	%d62, %f21
1823	  fdtos	%d62, %f22
1824	  fdtos	%d62, %f23
1825	  fdtos	%d62, %f24
1826	  fdtos	%d62, %f25
1827	  fdtos	%d62, %f26
1828	  fdtos	%d62, %f27
1829	  fdtos	%d62, %f28
1830	  fdtos	%d62, %f29
1831	  fdtos	%d62, %f30
1832	  fdtos	%d62, %f31
1833_fitos_fdtos_done:
1834
1835	ldd	[%g1 + CPU_TMP1], %d62		! restore %d62
1836
1837#if DEBUG
1838	/*
1839	 * Update FPop_unfinished trap kstat
1840	 */
1841	set	fpustat+FPUSTAT_UNFIN_KSTAT, %g7
1842	ldx	[%g7], %g5
18431:
1844	add	%g5, 1, %g6
1845
1846	casxa	[%g7] ASI_N, %g5, %g6
1847	cmp	%g5, %g6
1848	bne,a,pn %xcc, 1b
1849	  or	%g0, %g6, %g5
1850
1851	/*
1852	 * Update fpu_sim_fitos kstat
1853	 */
1854	set	fpuinfo+FPUINFO_FITOS_KSTAT, %g7
1855	ldx	[%g7], %g5
18561:
1857	add	%g5, 1, %g6
1858
1859	casxa	[%g7] ASI_N, %g5, %g6
1860	cmp	%g5, %g6
1861	bne,a,pn %xcc, 1b
1862	  or	%g0, %g6, %g5
1863#endif /* DEBUG */
1864
1865	FAST_TRAP_DONE
1866
1867.fp_exception_cont:
1868	/*
1869	 * Let _fp_exception deal with simulating FPop instruction.
1870	 * Note that we need to pass %fsr in %g2 (already read above).
1871	 */
1872
1873	set	_fp_exception, %g1
1874	ba,pt	%xcc, sys_trap
1875	sub	%g0, 1, %g4
1876
1877
1878/*
1879 * Register windows
1880 */
1881.flushw:
1882.clean_windows:
1883	rdpr	%tnpc, %g1
1884	wrpr	%g1, %tpc
1885	add	%g1, 4, %g1
1886	wrpr	%g1, %tnpc
1887	set	trap, %g1
1888	mov	T_FLUSH_PCB, %g3
1889	ba,pt	%xcc, sys_trap
1890	sub	%g0, 1, %g4
1891
1892/*
1893 * .spill_clean: clean the previous window, restore the wstate, and
1894 * "done".
1895 *
1896 * Entry: %g7 contains new wstate
1897 */
1898.spill_clean:
1899	sethi	%hi(nwin_minus_one), %g5
1900	ld	[%g5 + %lo(nwin_minus_one)], %g5 ! %g5 = nwin - 1
1901	rdpr	%cwp, %g6			! %g6 = %cwp
1902	deccc	%g6				! %g6--
1903	movneg	%xcc, %g5, %g6			! if (%g6<0) %g6 = nwin-1
1904	wrpr	%g6, %cwp
1905	TT_TRACE_L(trace_win)
1906	clr	%l0
1907	clr	%l1
1908	clr	%l2
1909	clr	%l3
1910	clr	%l4
1911	clr	%l5
1912	clr	%l6
1913	clr	%l7
1914	wrpr	%g0, %g7, %wstate
1915	saved
1916	retry			! restores correct %cwp
1917
1918.fix_alignment:
1919	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
1920	ldn	[%g1 + CPU_THREAD], %g1	! load thread pointer
1921	ldn	[%g1 + T_PROCP], %g1
1922	mov	1, %g2
1923	stb	%g2, [%g1 + P_FIXALIGNMENT]
1924	FAST_TRAP_DONE
1925
1926#define	STDF_REG(REG, ADDR, TMP)		\
1927	sll	REG, 3, REG			;\
1928mark1:	set	start1, TMP			;\
1929	jmp	REG + TMP			;\
1930	  nop					;\
1931start1:	ba,pt	%xcc, done1			;\
1932	  std	%f0, [ADDR + CPU_TMP1]		;\
1933	ba,pt	%xcc, done1			;\
1934	  std	%f32, [ADDR + CPU_TMP1]		;\
1935	ba,pt	%xcc, done1			;\
1936	  std	%f2, [ADDR + CPU_TMP1]		;\
1937	ba,pt	%xcc, done1			;\
1938	  std	%f34, [ADDR + CPU_TMP1]		;\
1939	ba,pt	%xcc, done1			;\
1940	  std	%f4, [ADDR + CPU_TMP1]		;\
1941	ba,pt	%xcc, done1			;\
1942	  std	%f36, [ADDR + CPU_TMP1]		;\
1943	ba,pt	%xcc, done1			;\
1944	  std	%f6, [ADDR + CPU_TMP1]		;\
1945	ba,pt	%xcc, done1			;\
1946	  std	%f38, [ADDR + CPU_TMP1]		;\
1947	ba,pt	%xcc, done1			;\
1948	  std	%f8, [ADDR + CPU_TMP1]		;\
1949	ba,pt	%xcc, done1			;\
1950	  std	%f40, [ADDR + CPU_TMP1]		;\
1951	ba,pt	%xcc, done1			;\
1952	  std	%f10, [ADDR + CPU_TMP1]		;\
1953	ba,pt	%xcc, done1			;\
1954	  std	%f42, [ADDR + CPU_TMP1]		;\
1955	ba,pt	%xcc, done1			;\
1956	  std	%f12, [ADDR + CPU_TMP1]		;\
1957	ba,pt	%xcc, done1			;\
1958	  std	%f44, [ADDR + CPU_TMP1]		;\
1959	ba,pt	%xcc, done1			;\
1960	  std	%f14, [ADDR + CPU_TMP1]		;\
1961	ba,pt	%xcc, done1			;\
1962	  std	%f46, [ADDR + CPU_TMP1]		;\
1963	ba,pt	%xcc, done1			;\
1964	  std	%f16, [ADDR + CPU_TMP1]		;\
1965	ba,pt	%xcc, done1			;\
1966	  std	%f48, [ADDR + CPU_TMP1]		;\
1967	ba,pt	%xcc, done1			;\
1968	  std	%f18, [ADDR + CPU_TMP1]		;\
1969	ba,pt	%xcc, done1			;\
1970	  std	%f50, [ADDR + CPU_TMP1]		;\
1971	ba,pt	%xcc, done1			;\
1972	  std	%f20, [ADDR + CPU_TMP1]		;\
1973	ba,pt	%xcc, done1			;\
1974	  std	%f52, [ADDR + CPU_TMP1]		;\
1975	ba,pt	%xcc, done1			;\
1976	  std	%f22, [ADDR + CPU_TMP1]		;\
1977	ba,pt	%xcc, done1			;\
1978	  std	%f54, [ADDR + CPU_TMP1]		;\
1979	ba,pt	%xcc, done1			;\
1980	  std	%f24, [ADDR + CPU_TMP1]		;\
1981	ba,pt	%xcc, done1			;\
1982	  std	%f56, [ADDR + CPU_TMP1]		;\
1983	ba,pt	%xcc, done1			;\
1984	  std	%f26, [ADDR + CPU_TMP1]		;\
1985	ba,pt	%xcc, done1			;\
1986	  std	%f58, [ADDR + CPU_TMP1]		;\
1987	ba,pt	%xcc, done1			;\
1988	  std	%f28, [ADDR + CPU_TMP1]		;\
1989	ba,pt	%xcc, done1			;\
1990	  std	%f60, [ADDR + CPU_TMP1]		;\
1991	ba,pt	%xcc, done1			;\
1992	  std	%f30, [ADDR + CPU_TMP1]		;\
1993	ba,pt	%xcc, done1			;\
1994	  std	%f62, [ADDR + CPU_TMP1]		;\
1995done1:
1996
1997#define	LDDF_REG(REG, ADDR, TMP)		\
1998	sll	REG, 3, REG			;\
1999mark2:	set	start2, TMP			;\
2000	jmp	REG + TMP			;\
2001	  nop					;\
2002start2:	ba,pt	%xcc, done2			;\
2003	  ldd	[ADDR + CPU_TMP1], %f0		;\
2004	ba,pt	%xcc, done2			;\
2005	  ldd	[ADDR + CPU_TMP1], %f32		;\
2006	ba,pt	%xcc, done2			;\
2007	  ldd	[ADDR + CPU_TMP1], %f2		;\
2008	ba,pt	%xcc, done2			;\
2009	  ldd	[ADDR + CPU_TMP1], %f34		;\
2010	ba,pt	%xcc, done2			;\
2011	  ldd	[ADDR + CPU_TMP1], %f4		;\
2012	ba,pt	%xcc, done2			;\
2013	  ldd	[ADDR + CPU_TMP1], %f36		;\
2014	ba,pt	%xcc, done2			;\
2015	  ldd	[ADDR + CPU_TMP1], %f6		;\
2016	ba,pt	%xcc, done2			;\
2017	  ldd	[ADDR + CPU_TMP1], %f38		;\
2018	ba,pt	%xcc, done2			;\
2019	  ldd	[ADDR + CPU_TMP1], %f8		;\
2020	ba,pt	%xcc, done2			;\
2021	  ldd	[ADDR + CPU_TMP1], %f40		;\
2022	ba,pt	%xcc, done2			;\
2023	  ldd	[ADDR + CPU_TMP1], %f10		;\
2024	ba,pt	%xcc, done2			;\
2025	  ldd	[ADDR + CPU_TMP1], %f42		;\
2026	ba,pt	%xcc, done2			;\
2027	  ldd	[ADDR + CPU_TMP1], %f12		;\
2028	ba,pt	%xcc, done2			;\
2029	  ldd	[ADDR + CPU_TMP1], %f44		;\
2030	ba,pt	%xcc, done2			;\
2031	  ldd	[ADDR + CPU_TMP1], %f14		;\
2032	ba,pt	%xcc, done2			;\
2033	  ldd	[ADDR + CPU_TMP1], %f46		;\
2034	ba,pt	%xcc, done2			;\
2035	  ldd	[ADDR + CPU_TMP1], %f16		;\
2036	ba,pt	%xcc, done2			;\
2037	  ldd	[ADDR + CPU_TMP1], %f48		;\
2038	ba,pt	%xcc, done2			;\
2039	  ldd	[ADDR + CPU_TMP1], %f18		;\
2040	ba,pt	%xcc, done2			;\
2041	  ldd	[ADDR + CPU_TMP1], %f50		;\
2042	ba,pt	%xcc, done2			;\
2043	  ldd	[ADDR + CPU_TMP1], %f20		;\
2044	ba,pt	%xcc, done2			;\
2045	  ldd	[ADDR + CPU_TMP1], %f52		;\
2046	ba,pt	%xcc, done2			;\
2047	  ldd	[ADDR + CPU_TMP1], %f22		;\
2048	ba,pt	%xcc, done2			;\
2049	  ldd	[ADDR + CPU_TMP1], %f54		;\
2050	ba,pt	%xcc, done2			;\
2051	  ldd	[ADDR + CPU_TMP1], %f24		;\
2052	ba,pt	%xcc, done2			;\
2053	  ldd	[ADDR + CPU_TMP1], %f56		;\
2054	ba,pt	%xcc, done2			;\
2055	  ldd	[ADDR + CPU_TMP1], %f26		;\
2056	ba,pt	%xcc, done2			;\
2057	  ldd	[ADDR + CPU_TMP1], %f58		;\
2058	ba,pt	%xcc, done2			;\
2059	  ldd	[ADDR + CPU_TMP1], %f28		;\
2060	ba,pt	%xcc, done2			;\
2061	  ldd	[ADDR + CPU_TMP1], %f60		;\
2062	ba,pt	%xcc, done2			;\
2063	  ldd	[ADDR + CPU_TMP1], %f30		;\
2064	ba,pt	%xcc, done2			;\
2065	  ldd	[ADDR + CPU_TMP1], %f62		;\
2066done2:
2067
2068.lddf_exception_not_aligned:
2069	/* %g2 = sfar, %g3 = sfsr */
2070	mov	%g2, %g5		! stash sfar
2071#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2072	sethi	%hi(fpu_exists), %g2	! check fpu_exists
2073	ld	[%g2 + %lo(fpu_exists)], %g2
2074	brz,a,pn %g2, 4f
2075	  nop
2076#endif
2077	CPU_ADDR(%g1, %g4)
2078	or	%g0, 1, %g4
2079	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2080
2081	rdpr	%tpc, %g2
2082	lda	[%g2]ASI_AIUP, %g6	! get the user's lddf instruction
2083	srl	%g6, 23, %g1		! using ldda or not?
2084	and	%g1, 1, %g1
2085	brz,a,pt %g1, 2f		! check for ldda instruction
2086	  nop
2087	srl	%g6, 13, %g1		! check immflag
2088	and	%g1, 1, %g1
2089	rdpr	%tstate, %g2		! %tstate in %g2
2090	brnz,a,pn %g1, 1f
2091	  srl	%g2, 31, %g1		! get asi from %tstate
2092	srl	%g6, 5, %g1		! get asi from instruction
2093	and	%g1, 0xFF, %g1		! imm_asi field
20941:
2095	cmp	%g1, ASI_P		! primary address space
2096	be,a,pt %icc, 2f
2097	  nop
2098	cmp	%g1, ASI_PNF		! primary no fault address space
2099	be,a,pt %icc, 2f
2100	  nop
2101	cmp	%g1, ASI_S		! secondary address space
2102	be,a,pt %icc, 2f
2103	  nop
2104	cmp	%g1, ASI_SNF		! secondary no fault address space
2105	bne,a,pn %icc, 3f
2106	  nop
21072:
2108	lduwa	[%g5]ASI_USER, %g7	! get first half of misaligned data
2109	add	%g5, 4, %g5		! increment misaligned data address
2110	lduwa	[%g5]ASI_USER, %g5	! get second half of misaligned data
2111
2112	sllx	%g7, 32, %g7
2113	or	%g5, %g7, %g5		! combine data
2114	CPU_ADDR(%g7, %g1)		! save data on a per-cpu basis
2115	stx	%g5, [%g7 + CPU_TMP1]	! save in cpu_tmp1
2116
2117	srl	%g6, 25, %g3		! %g6 has the instruction
2118	and	%g3, 0x1F, %g3		! %g3 has rd
2119	LDDF_REG(%g3, %g7, %g4)
2120
2121	CPU_ADDR(%g1, %g4)
2122	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2123	FAST_TRAP_DONE
21243:
2125	CPU_ADDR(%g1, %g4)
2126	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
21274:
2128	set	T_USER, %g3		! trap type in %g3
2129	or	%g3, T_LDDF_ALIGN, %g3
2130	mov	%g5, %g2		! misaligned vaddr in %g2
2131	set	fpu_trap, %g1		! goto C for the little and
2132	ba,pt	%xcc, sys_trap		! no fault little asi's
2133	  sub	%g0, 1, %g4
2134
2135.stdf_exception_not_aligned:
2136	/* %g2 = sfar, %g3 = sfsr */
2137	mov	%g2, %g5
2138
2139#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2140	sethi	%hi(fpu_exists), %g7		! check fpu_exists
2141	ld	[%g7 + %lo(fpu_exists)], %g3
2142	brz,a,pn %g3, 4f
2143	  nop
2144#endif
2145	CPU_ADDR(%g1, %g4)
2146	or	%g0, 1, %g4
2147	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2148
2149	rdpr	%tpc, %g2
2150	lda	[%g2]ASI_AIUP, %g6	! get the user's stdf instruction
2151
2152	srl	%g6, 23, %g1		! using stda or not?
2153	and	%g1, 1, %g1
2154	brz,a,pt %g1, 2f		! check for stda instruction
2155	  nop
2156	srl	%g6, 13, %g1		! check immflag
2157	and	%g1, 1, %g1
2158	rdpr	%tstate, %g2		! %tstate in %g2
2159	brnz,a,pn %g1, 1f
2160	  srl	%g2, 31, %g1		! get asi from %tstate
2161	srl	%g6, 5, %g1		! get asi from instruction
2162	and	%g1, 0xff, %g1		! imm_asi field
21631:
2164	cmp	%g1, ASI_P		! primary address space
2165	be,a,pt %icc, 2f
2166	  nop
2167	cmp	%g1, ASI_S		! secondary address space
2168	bne,a,pn %icc, 3f
2169	  nop
21702:
2171	srl	%g6, 25, %g6
2172	and	%g6, 0x1F, %g6		! %g6 has rd
2173	CPU_ADDR(%g7, %g1)
2174	STDF_REG(%g6, %g7, %g4)		! STDF_REG(REG, ADDR, TMP)
2175
2176	ldx	[%g7 + CPU_TMP1], %g6
2177	srlx	%g6, 32, %g7
2178	stuwa	%g7, [%g5]ASI_USER	! first half
2179	add	%g5, 4, %g5		! increment misaligned data address
2180	stuwa	%g6, [%g5]ASI_USER	! second half
2181
2182	CPU_ADDR(%g1, %g4)
2183	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2184	FAST_TRAP_DONE
21853:
2186	CPU_ADDR(%g1, %g4)
2187	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
21884:
2189	set	T_USER, %g3		! trap type in %g3
2190	or	%g3, T_STDF_ALIGN, %g3
2191	mov	%g5, %g2		! misaligned vaddr in %g2
2192	set	fpu_trap, %g1		! goto C for the little and
2193	ba,pt	%xcc, sys_trap		! nofault little asi's
2194	  sub	%g0, 1, %g4
2195
2196#ifdef DEBUG_USER_TRAPTRACECTL
2197
2198.traptrace_freeze:
2199	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2200	TT_TRACE_L(trace_win)
2201	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2202	set	trap_freeze, %g1
2203	mov	1, %g2
2204	st	%g2, [%g1]
2205	FAST_TRAP_DONE
2206
2207.traptrace_unfreeze:
2208	set	trap_freeze, %g1
2209	st	%g0, [%g1]
2210	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2211	TT_TRACE_L(trace_win)
2212	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2213	FAST_TRAP_DONE
2214
2215#endif /* DEBUG_USER_TRAPTRACECTL */
2216
2217.getcc:
2218	CPU_ADDR(%g1, %g2)
2219	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2220	rdpr	%tstate, %g3			! get tstate
2221	srlx	%g3, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2222	set	PSR_ICC, %g2
2223	and	%o0, %g2, %o0			! mask out the rest
2224	srl	%o0, PSR_ICC_SHIFT, %o0		! right justify
2225	wrpr	%g0, 0, %gl
2226	mov	%o0, %g1			! move ccr to normal %g1
2227	wrpr	%g0, 1, %gl
2228	! cannot assume globals retained their values after increasing %gl
2229	CPU_ADDR(%g1, %g2)
2230	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2231	FAST_TRAP_DONE
2232
2233.setcc:
2234	CPU_ADDR(%g1, %g2)
2235	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2236	wrpr	%g0, 0, %gl
2237	mov	%g1, %o0
2238	wrpr	%g0, 1, %gl
2239	! cannot assume globals retained their values after increasing %gl
2240	CPU_ADDR(%g1, %g2)
2241	sll	%o0, PSR_ICC_SHIFT, %g2
2242	set	PSR_ICC, %g3
2243	and	%g2, %g3, %g2			! mask out rest
2244	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g2
2245	rdpr	%tstate, %g3			! get tstate
2246	srl	%g3, 0, %g3			! clear upper word
2247	or	%g3, %g2, %g3			! or in new bits
2248	wrpr	%g3, %tstate
2249	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2250	FAST_TRAP_DONE
2251
2252/*
2253 * getpsr(void)
2254 * Note that the xcc part of the ccr is not provided.
2255 * The V8 code shows why the V9 trap is not faster:
2256 * #define GETPSR_TRAP() \
2257 *      mov %psr, %i0; jmp %l2; rett %l2+4; nop;
2258 */
2259
2260	.type	.getpsr, #function
2261.getpsr:
2262	rdpr	%tstate, %g1			! get tstate
2263	srlx	%g1, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2264	set	PSR_ICC, %g2
2265	and	%o0, %g2, %o0			! mask out the rest
2266
2267	rd	%fprs, %g1			! get fprs
2268	and	%g1, FPRS_FEF, %g2		! mask out dirty upper/lower
2269	sllx	%g2, PSR_FPRS_FEF_SHIFT, %g2	! shift fef to V8 psr.ef
2270	or	%o0, %g2, %o0			! or result into psr.ef
2271
2272	set	V9_PSR_IMPLVER, %g2		! SI assigned impl/ver: 0xef
2273	or	%o0, %g2, %o0			! or psr.impl/ver
2274	FAST_TRAP_DONE
2275	SET_SIZE(.getpsr)
2276
2277/*
2278 * setpsr(newpsr)
2279 * Note that there is no support for ccr.xcc in the V9 code.
2280 */
2281
2282	.type	.setpsr, #function
2283.setpsr:
2284	rdpr	%tstate, %g1			! get tstate
2285!	setx	TSTATE_V8_UBITS, %g2
2286	or 	%g0, CCR_ICC, %g3
2287	sllx	%g3, TSTATE_CCR_SHIFT, %g2
2288
2289	andn	%g1, %g2, %g1			! zero current user bits
2290	set	PSR_ICC, %g2
2291	and	%g2, %o0, %g2			! clear all but psr.icc bits
2292	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g3	! shift to tstate.ccr.icc
2293	wrpr	%g1, %g3, %tstate		! write tstate
2294
2295	set	PSR_EF, %g2
2296	and	%g2, %o0, %g2			! clear all but fp enable bit
2297	srlx	%g2, PSR_FPRS_FEF_SHIFT, %g4	! shift ef to V9 fprs.fef
2298	wr	%g0, %g4, %fprs			! write fprs
2299
2300	CPU_ADDR(%g1, %g2)			! load CPU struct addr to %g1
2301	ldn	[%g1 + CPU_THREAD], %g2		! load thread pointer
2302	ldn	[%g2 + T_LWP], %g3		! load klwp pointer
2303	ldn	[%g3 + LWP_FPU], %g2		! get lwp_fpu pointer
2304	stuw	%g4, [%g2 + FPU_FPRS]		! write fef value to fpu_fprs
2305	srlx	%g4, 2, %g4			! shift fef value to bit 0
2306	stub	%g4, [%g2 + FPU_EN]		! write fef value to fpu_en
2307	FAST_TRAP_DONE
2308	SET_SIZE(.setpsr)
2309
2310/*
2311 * getlgrp
2312 * get home lgrpid on which the calling thread is currently executing.
2313 */
2314	.type	.getlgrp, #function
2315.getlgrp:
2316	! Thanks for the incredibly helpful comments
2317	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2318	ld	[%g1 + CPU_ID], %o0	! load cpu_id
2319	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2320	ldn	[%g2 + T_LPL], %g2	! load lpl pointer
2321	ld	[%g2 + LPL_LGRPID], %g1	! load lpl_lgrpid
2322	sra	%g1, 0, %o1
2323	FAST_TRAP_DONE
2324	SET_SIZE(.getlgrp)
2325
2326/*
2327 * Entry for old 4.x trap (trap 0).
2328 */
2329	ENTRY_NP(syscall_trap_4x)
2330	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2331	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2332	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2333	ld	[%g2 + PCB_TRAP0], %g2	! lwp->lwp_pcb.pcb_trap0addr
2334	brz,pn	%g2, 1f			! has it been set?
2335	st	%l0, [%g1 + CPU_TMP1]	! delay - save some locals
2336	st	%l1, [%g1 + CPU_TMP2]
2337	rdpr	%tnpc, %l1		! save old tnpc
2338	wrpr	%g0, %g2, %tnpc		! setup tnpc
2339
2340	mov	%g1, %l0		! save CPU struct addr
2341	wrpr	%g0, 0, %gl
2342	mov	%l1, %g6		! pass tnpc to user code in %g6
2343	wrpr	%g0, 1, %gl
2344	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2345	ld	[%l0 + CPU_TMP1], %l0
2346	FAST_TRAP_DONE_CHK_INTR
23471:
2348	!
2349	! check for old syscall mmap which is the only different one which
2350	! must be the same.  Others are handled in the compatibility library.
2351	!
2352	mov	%g1, %l0		! save CPU struct addr
2353	wrpr	%g0, 0, %gl
2354	cmp	%g1, OSYS_mmap		! compare to old 4.x mmap
2355	movz	%icc, SYS_mmap, %g1
2356	wrpr	%g0, 1, %gl
2357	ld	[%l0 + CPU_TMP1], %l0
2358	SYSCALL(syscall_trap32)
2359	SET_SIZE(syscall_trap_4x)
2360
2361/*
2362 * Handler for software trap 9.
2363 * Set trap0 emulation address for old 4.x system call trap.
2364 * XXX - this should be a system call.
2365 */
2366	ENTRY_NP(set_trap0_addr)
2367	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2368	st	%l0, [%g1 + CPU_TMP1]	! save some locals
2369	st	%l1, [%g1 + CPU_TMP2]
2370	mov	%g1, %l0	! preserve CPU addr
2371	wrpr	%g0, 0, %gl
2372	mov	%g1, %l1
2373	wrpr	%g0, 1, %gl
2374	! cannot assume globals retained their values after increasing %gl
2375	ldn	[%l0 + CPU_THREAD], %g2	! load thread pointer
2376	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2377	andn	%l1, 3, %l1		! force alignment
2378	st	%l1, [%g2 + PCB_TRAP0]	! lwp->lwp_pcb.pcb_trap0addr
2379	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2380	ld	[%l0 + CPU_TMP1], %l0
2381	FAST_TRAP_DONE
2382	SET_SIZE(set_trap0_addr)
2383
2384/*
2385 * mmu_trap_tl1
2386 * trap handler for unexpected mmu traps.
2387 * simply checks if the trap was a user lddf/stdf alignment trap, in which
2388 * case we go to fpu_trap or a user trap from the window handler, in which
2389 * case we go save the state on the pcb.  Otherwise, we go to ptl1_panic.
2390 */
2391	.type	mmu_trap_tl1, #function
2392mmu_trap_tl1:
2393#ifdef	TRAPTRACE
2394	TRACE_PTR(%g5, %g6)
2395	GET_TRACE_TICK(%g6, %g7)
2396	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2397	TRACE_SAVE_TL_GL_REGS(%g5, %g6)
2398	rdpr	%tt, %g6
2399	stha	%g6, [%g5 + TRAP_ENT_TT]%asi
2400	rdpr	%tstate, %g6
2401	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi
2402	stna	%sp, [%g5 + TRAP_ENT_SP]%asi
2403	stna	%g0, [%g5 + TRAP_ENT_TR]%asi
2404	rdpr	%tpc, %g6
2405	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2406	MMU_FAULT_STATUS_AREA(%g6)
2407	ldx	[%g6 + MMFSA_D_ADDR], %g6
2408	stna	%g6, [%g5 + TRAP_ENT_F1]%asi !  MMU fault address
2409	CPU_PADDR(%g7, %g6);
2410	add	%g7, CPU_TL1_HDLR, %g7
2411	lda	[%g7]ASI_MEM, %g6
2412	stna	%g6, [%g5 + TRAP_ENT_F2]%asi
2413	MMU_FAULT_STATUS_AREA(%g6)
2414	ldx	[%g6 + MMFSA_D_TYPE], %g7 ! XXXQ should be a MMFSA_F_ constant?
2415	ldx	[%g6 + MMFSA_D_CTX], %g6
2416	sllx	%g6, SFSR_CTX_SHIFT, %g6
2417	or	%g6, %g7, %g6
2418	stna	%g6, [%g5 + TRAP_ENT_F3]%asi ! MMU context/type
2419	set	0xdeadbeef, %g6
2420	stna	%g6, [%g5 + TRAP_ENT_F4]%asi
2421	TRACE_NEXT(%g5, %g6, %g7)
2422#endif /* TRAPTRACE */
2423	CPU_PADDR(%g7, %g6);
2424	add     %g7, CPU_TL1_HDLR, %g7		! %g7 = &cpu_m.tl1_hdlr (PA)
2425	lda	[%g7]ASI_MEM, %g6
2426	brz,a,pt %g6, 1f
2427	  nop
2428	sta     %g0, [%g7]ASI_MEM
2429	! XXXQ need to setup registers for sfmmu_mmu_trap?
2430	ba,a,pt	%xcc, sfmmu_mmu_trap		! handle page faults
24311:
2432	rdpr	%tpc, %g7
2433	/* in user_rtt? */
2434	set	rtt_fill_start, %g6
2435	cmp	%g7, %g6
2436	blu,pn	%xcc, 6f
2437	 .empty
2438	set	rtt_fill_end, %g6
2439	cmp	%g7, %g6
2440	bgeu,pn %xcc, 6f
2441	 nop
2442	set	fault_rtt_fn1, %g7
2443	ba,a	7f
24446:
2445	! check to see if the trap pc is in a window spill/fill handling
2446	rdpr	%tpc, %g7
2447	/* tpc should be in the trap table */
2448	set	trap_table, %g6
2449	cmp	%g7, %g6
2450	blu,a,pn %xcc, ptl1_panic
2451	  mov	PTL1_BAD_MMUTRAP, %g1
2452	set	etrap_table, %g6
2453	cmp	%g7, %g6
2454	bgeu,a,pn %xcc, ptl1_panic
2455	  mov	PTL1_BAD_MMUTRAP, %g1
2456	! pc is inside the trap table, convert to trap type
2457	srl	%g7, 5, %g6		! XXXQ need #define
2458	and	%g6, 0x1ff, %g6		! XXXQ need #define
2459	! and check for a window trap type
2460	and	%g6, WTRAP_TTMASK, %g6
2461	cmp	%g6, WTRAP_TYPE
2462	bne,a,pn %xcc, ptl1_panic
2463	  mov	PTL1_BAD_MMUTRAP, %g1
2464	andn	%g7, WTRAP_ALIGN, %g7	/* 128 byte aligned */
2465	add	%g7, WTRAP_FAULTOFF, %g7
2466
24677:
2468	! Arguments are passed in the global set active after the
2469	! 'done' instruction. Before switching sets, must save
2470	! the calculated next pc
2471	wrpr	%g0, %g7, %tnpc
2472	wrpr	%g0, 1, %gl
2473	rdpr	%tt, %g5
2474	MMU_FAULT_STATUS_AREA(%g7)
2475	cmp	%g5, T_ALIGNMENT
2476	be,pn	%xcc, 1f
2477	ldx	[%g7 + MMFSA_D_ADDR], %g6
2478	ldx	[%g7 + MMFSA_D_CTX], %g7
2479	srlx	%g6, MMU_PAGESHIFT, %g6		/* align address */
2480	cmp	%g7, USER_CONTEXT_TYPE
2481	sllx	%g6, MMU_PAGESHIFT, %g6
2482	movgu	%icc, USER_CONTEXT_TYPE, %g7
2483	or	%g6, %g7, %g6			/* TAG_ACCESS */
24841:
2485	done
2486	SET_SIZE(mmu_trap_tl1)
2487
2488/*
2489 * Several traps use kmdb_trap and kmdb_trap_tl1 as their handlers.  These
2490 * traps are valid only when kmdb is loaded.  When the debugger is active,
2491 * the code below is rewritten to transfer control to the appropriate
2492 * debugger entry points.
2493 */
2494	.global	kmdb_trap
2495	.align	8
2496kmdb_trap:
2497	ba,a	trap_table0
2498	jmp	%g1 + 0
2499	nop
2500
2501	.global	kmdb_trap_tl1
2502	.align	8
2503kmdb_trap_tl1:
2504	ba,a	trap_table0
2505	jmp	%g1 + 0
2506	nop
2507
2508/*
2509 * This entry is copied from OBP's trap table during boot.
2510 */
2511	.global	obp_bpt
2512	.align	8
2513obp_bpt:
2514	NOT
2515
2516
2517
2518#ifdef	TRAPTRACE
2519/*
2520 * TRAPTRACE support.
2521 * labels here are branched to with "rd %pc, %g7" in the delay slot.
2522 * Return is done by "jmp %g7 + 4".
2523 */
2524
2525trace_dmmu:
2526	TRACE_PTR(%g3, %g6)
2527	GET_TRACE_TICK(%g6, %g5)
2528	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2529	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2530	rdpr	%tt, %g6
2531	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2532	rdpr	%tstate, %g6
2533	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2534	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2535	rdpr	%tpc, %g6
2536	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2537	MMU_FAULT_STATUS_AREA(%g6)
2538	ldx	[%g6 + MMFSA_D_ADDR], %g4
2539	stxa	%g4, [%g3 + TRAP_ENT_TR]%asi
2540	ldx	[%g6 + MMFSA_D_CTX], %g4
2541	stxa	%g4, [%g3 + TRAP_ENT_F1]%asi
2542	ldx	[%g6 + MMFSA_D_TYPE], %g4
2543	stxa	%g4, [%g3 + TRAP_ENT_F2]%asi
2544	stxa	%g6, [%g3 + TRAP_ENT_F3]%asi
2545	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2546	TRACE_NEXT(%g3, %g4, %g5)
2547	jmp	%g7 + 4
2548	nop
2549
2550trace_immu:
2551	TRACE_PTR(%g3, %g6)
2552	GET_TRACE_TICK(%g6, %g5)
2553	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2554	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2555	rdpr	%tt, %g6
2556	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2557	rdpr	%tstate, %g6
2558	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2559	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2560	rdpr	%tpc, %g6
2561	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2562	MMU_FAULT_STATUS_AREA(%g6)
2563	ldx	[%g6 + MMFSA_I_ADDR], %g4
2564	stxa	%g4, [%g3 + TRAP_ENT_TR]%asi
2565	ldx	[%g6 + MMFSA_I_CTX], %g4
2566	stxa	%g4, [%g3 + TRAP_ENT_F1]%asi
2567	ldx	[%g6 + MMFSA_I_TYPE], %g4
2568	stxa	%g4, [%g3 + TRAP_ENT_F2]%asi
2569	stxa	%g6, [%g3 + TRAP_ENT_F3]%asi
2570	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2571	TRACE_NEXT(%g3, %g4, %g5)
2572	jmp	%g7 + 4
2573	nop
2574
2575trace_gen:
2576	TRACE_PTR(%g3, %g6)
2577	GET_TRACE_TICK(%g6, %g5)
2578	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2579	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2580	rdpr	%tt, %g6
2581	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2582	rdpr	%tstate, %g6
2583	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2584	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2585	rdpr	%tpc, %g6
2586	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2587	stna	%g0, [%g3 + TRAP_ENT_TR]%asi
2588	stna	%g0, [%g3 + TRAP_ENT_F1]%asi
2589	stna	%g0, [%g3 + TRAP_ENT_F2]%asi
2590	stna	%g0, [%g3 + TRAP_ENT_F3]%asi
2591	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2592	TRACE_NEXT(%g3, %g4, %g5)
2593	jmp	%g7 + 4
2594	nop
2595
2596trace_win:
2597	TRACE_WIN_INFO(0, %l0, %l1, %l2)
2598	! Keep the locals as clean as possible, caller cleans %l4
2599	clr	%l2
2600	clr	%l1
2601	jmp	%l4 + 4
2602	  clr	%l0
2603
2604/*
2605 * Trace a tsb hit
2606 * g1 = tsbe pointer (in/clobbered)
2607 * g2 = tag access register (in)
2608 * g3 - g4 = scratch (clobbered)
2609 * g5 = tsbe data (in)
2610 * g6 = scratch (clobbered)
2611 * g7 = pc we jumped here from (in)
2612 */
2613
2614	! Do not disturb %g5, it will be used after the trace
2615	ALTENTRY(trace_tsbhit)
2616	TRACE_TSBHIT(0)
2617	jmp	%g7 + 4
2618	nop
2619
2620/*
2621 * Trace a TSB miss
2622 *
2623 * g1 = tsb8k pointer (in)
2624 * g2 = tag access register (in)
2625 * g3 = tsb4m pointer (in)
2626 * g4 = tsbe tag (in/clobbered)
2627 * g5 - g6 = scratch (clobbered)
2628 * g7 = pc we jumped here from (in)
2629 */
2630	.global	trace_tsbmiss
2631trace_tsbmiss:
2632	membar	#Sync
2633	sethi	%hi(FLUSH_ADDR), %g6
2634	flush	%g6
2635	TRACE_PTR(%g5, %g6)
2636	stna	%g2, [%g5 + TRAP_ENT_SP]%asi		! tag access
2637	stna	%g4, [%g5 + TRAP_ENT_F1]%asi		! XXX? tsb tag
2638	GET_TRACE_TICK(%g6, %g4)
2639	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2640	rdpr	%tnpc, %g6
2641	stna	%g6, [%g5 + TRAP_ENT_F2]%asi
2642	stna	%g1, [%g5 + TRAP_ENT_F3]%asi		! tsb8k pointer
2643	rdpr	%tpc, %g6
2644	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2645	TRACE_SAVE_TL_GL_REGS(%g5, %g6)
2646	rdpr	%tt, %g6
2647	or	%g6, TT_MMU_MISS, %g4
2648	stha	%g4, [%g5 + TRAP_ENT_TT]%asi
2649	mov	MMFSA_D_ADDR, %g4
2650	cmp	%g6, FAST_IMMU_MISS_TT
2651	move	%xcc, MMFSA_I_ADDR, %g4
2652	cmp	%g6, T_INSTR_MMU_MISS
2653	move	%xcc, MMFSA_I_ADDR, %g4
2654	MMU_FAULT_STATUS_AREA(%g6)
2655	ldx	[%g6 + %g4], %g6
2656	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi	! tag target
2657	cmp	%g4, MMFSA_D_ADDR
2658	move	%xcc, MMFSA_D_CTX, %g4
2659	movne	%xcc, MMFSA_I_CTX, %g4
2660	MMU_FAULT_STATUS_AREA(%g6)
2661	ldx	[%g6 + %g4], %g6
2662	stxa	%g6, [%g5 + TRAP_ENT_F4]%asi		! context ID
2663	stna	%g3, [%g5 + TRAP_ENT_TR]%asi		! tsb4m pointer
2664	TRACE_NEXT(%g5, %g4, %g6)
2665	jmp	%g7 + 4
2666	nop
2667
2668/*
2669 * g2 = tag access register (in)
2670 * g3 = ctx type (0, 1 or 2) (in) (not used)
2671 */
2672trace_dataprot:
2673	membar	#Sync
2674	sethi	%hi(FLUSH_ADDR), %g6
2675	flush	%g6
2676	TRACE_PTR(%g1, %g6)
2677	GET_TRACE_TICK(%g6, %g4)
2678	stxa	%g6, [%g1 + TRAP_ENT_TICK]%asi
2679	rdpr	%tpc, %g6
2680	stna	%g6, [%g1 + TRAP_ENT_TPC]%asi
2681	rdpr	%tstate, %g6
2682	stxa	%g6, [%g1 + TRAP_ENT_TSTATE]%asi
2683	stna	%g2, [%g1 + TRAP_ENT_SP]%asi		! tag access reg
2684	stna	%g0, [%g1 + TRAP_ENT_F1]%asi
2685	stna	%g0, [%g1 + TRAP_ENT_F2]%asi
2686	stna	%g0, [%g1 + TRAP_ENT_F3]%asi
2687	stna	%g0, [%g1 + TRAP_ENT_F4]%asi
2688	TRACE_SAVE_TL_GL_REGS(%g1, %g6)
2689	rdpr	%tt, %g6
2690	stha	%g6, [%g1 + TRAP_ENT_TT]%asi
2691	mov	MMFSA_D_CTX, %g4
2692	cmp	%g6, FAST_IMMU_MISS_TT
2693	move	%xcc, MMFSA_I_CTX, %g4
2694	cmp	%g6, T_INSTR_MMU_MISS
2695	move	%xcc, MMFSA_I_CTX, %g4
2696	MMU_FAULT_STATUS_AREA(%g6)
2697	ldx	[%g6 + %g4], %g6
2698	stxa	%g6, [%g1 + TRAP_ENT_TR]%asi	! context ID
2699	TRACE_NEXT(%g1, %g4, %g5)
2700	jmp	%g7 + 4
2701	nop
2702
2703#endif /* TRAPTRACE */
2704
2705/*
2706 * Handle watchdog reset trap. Enable the MMU using the MMU_ENABLE
2707 * HV service, which requires the return target to be specified as a VA
2708 * since we are enabling the MMU. We set the target to ptl1_panic.
2709 */
2710
2711	.type	.watchdog_trap, #function
2712.watchdog_trap:
2713	mov	1, %o0
2714	setx	ptl1_panic, %g2, %o1
2715	mov	MMU_ENABLE, %o5
2716	ta	FAST_TRAP
2717	done
2718	SET_SIZE(.watchdog_trap)
2719/*
2720 * synthesize for trap(): SFAR in %g2, SFSR in %g3
2721 */
2722	.type	.dmmu_exc_lddf_not_aligned, #function
2723.dmmu_exc_lddf_not_aligned:
2724	MMU_FAULT_STATUS_AREA(%g3)
2725	ldx	[%g3 + MMFSA_D_ADDR], %g2
2726	/* Fault type not available in MMU fault status area */
2727	mov	MMFSA_F_UNALIGN, %g1
2728	ldx	[%g3 + MMFSA_D_CTX], %g3
2729	sllx	%g3, SFSR_CTX_SHIFT, %g3
2730	btst	1, %sp
2731	bnz,pt	%xcc, .lddf_exception_not_aligned
2732	or	%g3, %g1, %g3			/* SFSR */
2733	ba,a,pt	%xcc, .mmu_exception_not_aligned
2734	SET_SIZE(.dmmu_exc_lddf_not_aligned)
2735
2736/*
2737 * synthesize for trap(): SFAR in %g2, SFSR in %g3
2738 */
2739	.type	.dmmu_exc_stdf_not_aligned, #function
2740.dmmu_exc_stdf_not_aligned:
2741	MMU_FAULT_STATUS_AREA(%g3)
2742	ldx	[%g3 + MMFSA_D_ADDR], %g2
2743	/* Fault type not available in MMU fault status area */
2744	mov	MMFSA_F_UNALIGN, %g1
2745	ldx	[%g3 + MMFSA_D_CTX], %g3
2746	sllx	%g3, SFSR_CTX_SHIFT, %g3
2747	btst	1, %sp
2748	bnz,pt	%xcc, .stdf_exception_not_aligned
2749	or	%g3, %g1, %g3			/* SFSR */
2750	ba,a,pt	%xcc, .mmu_exception_not_aligned
2751	SET_SIZE(.dmmu_exc_stdf_not_aligned)
2752
2753	.type	.dmmu_exception, #function
2754.dmmu_exception:
2755	MMU_FAULT_STATUS_AREA(%g3)
2756	ldx	[%g3 + MMFSA_D_ADDR], %g2
2757	ldx	[%g3 + MMFSA_D_TYPE], %g1
2758	ldx	[%g3 + MMFSA_D_CTX], %g4
2759	srlx	%g2, MMU_PAGESHIFT, %g2		/* align address */
2760	sllx	%g2, MMU_PAGESHIFT, %g2
2761	sllx	%g4, SFSR_CTX_SHIFT, %g3
2762	or	%g3, %g1, %g3			/* SFSR */
2763	cmp	%g4, USER_CONTEXT_TYPE
2764	movgeu	%icc, USER_CONTEXT_TYPE, %g4
2765	or	%g2, %g4, %g2			/* TAG_ACCESS */
2766	ba,pt	%xcc, .mmu_exception_end
2767	mov	T_DATA_EXCEPTION, %g1
2768	SET_SIZE(.dmmu_exception)
2769
2770	.align 32
2771	.global pil15_epilogue
2772pil15_epilogue:
2773	ba	pil_interrupt_common
2774	nop
2775	.align 32
2776
2777/*
2778 * fast_trap_done, fast_trap_done_chk_intr:
2779 *
2780 * Due to the design of UltraSPARC pipeline, pending interrupts are not
2781 * taken immediately after a RETRY or DONE instruction which causes IE to
2782 * go from 0 to 1. Instead, the instruction at %tpc or %tnpc is allowed
2783 * to execute first before taking any interrupts. If that instruction
2784 * results in other traps, and if the corresponding trap handler runs
2785 * entirely at TL=1 with interrupts disabled, then pending interrupts
2786 * won't be taken until after yet another instruction following the %tpc
2787 * or %tnpc.
2788 *
2789 * A malicious user program can use this feature to block out interrupts
2790 * for extended durations, which can result in send_mondo_timeout kernel
2791 * panic.
2792 *
2793 * This problem is addressed by servicing any pending interrupts via
2794 * sys_trap before returning back to the user mode from a fast trap
2795 * handler. The "done" instruction within a fast trap handler, which
2796 * runs entirely at TL=1 with interrupts disabled, is replaced with the
2797 * FAST_TRAP_DONE macro, which branches control to this fast_trap_done
2798 * entry point.
2799 *
2800 * We check for any pending interrupts here and force a sys_trap to
2801 * service those interrupts, if any. To minimize overhead, pending
2802 * interrupts are checked if the %tpc happens to be at 16K boundary,
2803 * which allows a malicious program to execute at most 4K consecutive
2804 * instructions before we service any pending interrupts. If a worst
2805 * case fast trap handler takes about 2 usec, then interrupts will be
2806 * blocked for at most 8 msec, less than a clock tick.
2807 *
2808 * For the cases where we don't know if the %tpc will cross a 16K
2809 * boundary, we can't use the above optimization and always process
2810 * any pending interrupts via fast_frap_done_chk_intr entry point.
2811 *
2812 * Entry Conditions:
2813 * 	%pstate		am:0 priv:1 ie:0
2814 * 			globals are AG (not normal globals)
2815 */
2816
2817	.global	fast_trap_done, fast_trap_done_chk_intr
2818fast_trap_done:
2819	rdpr	%tpc, %g5
2820	sethi	%hi(0xffffc000), %g6	! 1's complement of 0x3fff
2821	andncc	%g5, %g6, %g0		! check lower 14 bits of %tpc
2822	bz,pn	%icc, 1f		! branch if zero (lower 32 bits only)
2823	nop
2824	done
2825
2826fast_trap_done_chk_intr:
28271:	rd	SOFTINT, %g6
2828	brnz,pn	%g6, 2f		! branch if any pending intr
2829	nop
2830	done
2831
28322:
2833	/*
2834	 * We get here if there are any pending interrupts.
2835	 * Adjust %tpc/%tnpc as we'll be resuming via "retry"
2836	 * instruction.
2837	 */
2838	rdpr	%tnpc, %g5
2839	wrpr	%g0, %g5, %tpc
2840	add	%g5, 4, %g5
2841	wrpr	%g0, %g5, %tnpc
2842
2843	/*
2844	 * Force a dummy sys_trap call so that interrupts can be serviced.
2845	 */
2846	set	fast_trap_dummy_call, %g1
2847	ba,pt	%xcc, sys_trap
2848	  mov	-1, %g4
2849
2850fast_trap_dummy_call:
2851	retl
2852	nop
2853
2854/*
2855 * Currently the brand syscall interposition code is not enabled by
2856 * default.  Instead, when a branded zone is first booted the brand
2857 * infrastructure will patch the trap table so that the syscall
2858 * entry points are redirected to syscall_wrapper32 and syscall_wrapper
2859 * for ILP32 and LP64 syscalls respectively.  this is done in
2860 * brand_plat_interposition_enable().  Note that the syscall wrappers
2861 * below do not collect any trap trace data since the syscall hot patch
2862 * points are reached after trap trace data has already been collected.
2863 */
2864#define	BRAND_CALLBACK(callback_id)					    \
2865	CPU_ADDR(%g2, %g1)		/* load CPU struct addr to %g2	*/ ;\
2866	ldn	[%g2 + CPU_THREAD], %g3	/* load thread pointer		*/ ;\
2867	ldn	[%g3 + T_PROCP], %g3	/* get proc pointer		*/ ;\
2868	ldn	[%g3 + P_BRAND], %g3	/* get brand pointer		*/ ;\
2869	brz	%g3, 1f			/* No brand?  No callback. 	*/ ;\
2870	nop 								   ;\
2871	ldn	[%g3 + B_MACHOPS], %g3	/* get machops list		*/ ;\
2872	ldn	[%g3 + (callback_id << 3)], %g3 			   ;\
2873	brz	%g3, 1f							   ;\
2874	/*								    \
2875	 * This isn't pretty.  We want a low-latency way for the callback   \
2876	 * routine to decline to do anything.  We just pass in an address   \
2877	 * the routine can directly jmp back to, pretending that nothing    \
2878	 * has happened.						    \
2879	 * 								    \
2880	 * %g1: return address (where the brand handler jumps back to)	    \
2881	 * %g2: address of CPU structure				    \
2882	 * %g3: address of brand handler (where we will jump to)	    \
2883	 */								    \
2884	mov	%pc, %g1						   ;\
2885	add	%g1, 16, %g1						   ;\
2886	jmp	%g3							   ;\
2887	nop								   ;\
28881:
2889
2890	ENTRY_NP(syscall_wrapper32)
2891	BRAND_CALLBACK(BRAND_CB_SYSCALL32)
2892	SYSCALL_NOTT(syscall_trap32)
2893	SET_SIZE(syscall_wrapper32)
2894
2895	ENTRY_NP(syscall_wrapper)
2896	BRAND_CALLBACK(BRAND_CB_SYSCALL)
2897	SYSCALL_NOTT(syscall_trap)
2898	SET_SIZE(syscall_wrapper)
2899
2900#endif	/* lint */
2901