xref: /titanic_50/usr/src/uts/sun4v/ml/trap_table.s (revision 94d05f6c7f329fdf908da99ab50b37d3d33f9fe5)
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 2008 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 CPU_MONDO			\
811	ba,a,pt	%xcc, cpu_mondo		;\
812	.align	32
813
814#define DEV_MONDO			\
815	ba,a,pt	%xcc, dev_mondo		;\
816	.align	32
817
818/*
819 * We take over the rtba after we set our trap table and
820 * fault status area. The watchdog reset trap is now handled by the OS.
821 */
822#define WATCHDOG_RESET			\
823	mov	PTL1_BAD_WATCHDOG, %g1	;\
824	ba,a,pt	%xcc, .watchdog_trap	;\
825	.align	32
826
827/*
828 * RED is for traps that use the red mode handler.
829 * We should never see these either.
830 */
831#define RED			\
832	mov	PTL1_BAD_RED, %g1	;\
833	ba,a,pt	%xcc, .watchdog_trap	;\
834	.align	32
835
836
837/*
838 * MMU Trap Handlers.
839 */
840
841/*
842 * synthesize for trap(): SFSR in %g3
843 */
844#define	IMMU_EXCEPTION							\
845	MMU_FAULT_STATUS_AREA(%g3)					;\
846	rdpr	%tpc, %g2						;\
847	ldx	[%g3 + MMFSA_I_TYPE], %g1				;\
848	ldx	[%g3 + MMFSA_I_CTX], %g3				;\
849	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
850	or	%g3, %g1, %g3						;\
851	ba,pt	%xcc, .mmu_exception_end				;\
852	mov	T_INSTR_EXCEPTION, %g1					;\
853	.align	32
854
855/*
856 * synthesize for trap(): TAG_ACCESS in %g2, SFSR in %g3
857 */
858#define	DMMU_EXCEPTION							\
859	ba,a,pt	%xcc, .dmmu_exception					;\
860	.align	32
861
862/*
863 * synthesize for trap(): SFAR in %g2, SFSR in %g3
864 */
865#define	DMMU_EXC_AG_PRIV						\
866	MMU_FAULT_STATUS_AREA(%g3)					;\
867	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
868	/* Fault type not available in MMU fault status area */		;\
869	mov	MMFSA_F_PRVACT, %g1					;\
870	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
871	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
872	ba,pt	%xcc, .mmu_priv_exception				;\
873	or	%g3, %g1, %g3						;\
874	.align	32
875
876/*
877 * synthesize for trap(): SFAR in %g2, SFSR in %g3
878 */
879#define	DMMU_EXC_AG_NOT_ALIGNED						\
880	MMU_FAULT_STATUS_AREA(%g3)					;\
881	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
882	/* Fault type not available in MMU fault status area */		;\
883	mov	MMFSA_F_UNALIGN, %g1					;\
884	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
885	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
886	ba,pt	%xcc, .mmu_exception_not_aligned			;\
887	or	%g3, %g1, %g3			/* SFSR */		;\
888	.align	32
889/*
890 * SPARC V9 IMPL. DEP. #109(1) and (2) and #110(1) and (2)
891 */
892
893/*
894 * synthesize for trap(): SFAR in %g2, SFSR in %g3
895 */
896#define	DMMU_EXC_LDDF_NOT_ALIGNED					\
897	ba,a,pt	%xcc, .dmmu_exc_lddf_not_aligned			;\
898	.align	32
899/*
900 * synthesize for trap(): SFAR in %g2, SFSR in %g3
901 */
902#define	DMMU_EXC_STDF_NOT_ALIGNED					\
903	ba,a,pt	%xcc, .dmmu_exc_stdf_not_aligned			;\
904	.align	32
905
906#if defined(cscope)
907/*
908 * Define labels to direct cscope quickly to labels that
909 * are generated by macro expansion of DTLB_MISS().
910 */
911	.global	tt0_dtlbmiss
912tt0_dtlbmiss:
913	.global	tt1_dtlbmiss
914tt1_dtlbmiss:
915	nop
916#endif
917
918/*
919 * Data miss handler (must be exactly 32 instructions)
920 *
921 * This handler is invoked only if the hypervisor has been instructed
922 * not to do any TSB walk.
923 *
924 * Kernel and invalid context cases are handled by the sfmmu_kdtlb_miss
925 * handler.
926 *
927 * User TLB miss handling depends upon whether a user process has one or
928 * two TSBs. User TSB information (physical base and size code) is kept
929 * in two dedicated scratchpad registers. Absence of a user TSB (primarily
930 * second TSB) is indicated by a negative value (-1) in that register.
931 */
932
933/*
934 * synthesize for miss handler: pseudo-tag access in %g2 (with context "type"
935 * (0=kernel, 1=invalid, or 2=user) rather than context ID)
936 */
937#define	DTLB_MISS(table_name)						;\
938	.global	table_name/**/_dtlbmiss					;\
939table_name/**/_dtlbmiss:						;\
940	GET_MMU_D_PTAGACC_CTXTYPE(%g2, %g3)	/* 8 instr */		;\
941	cmp	%g3, INVALID_CONTEXT					;\
942	ble,pn	%xcc, sfmmu_kdtlb_miss					;\
943	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
944	mov	SCRATCHPAD_UTSBREG2, %g1				;\
945	ldxa	[%g1]ASI_SCRATCHPAD, %g1	/* get 2nd tsbreg */	;\
946	brgez,pn %g1, sfmmu_udtlb_slowpath	/* branch if 2 TSBs */	;\
947	  nop								;\
948	GET_1ST_TSBE_PTR(%g2, %g1, %g4, %g5)	/* 11 instr */		;\
949	ba,pt	%xcc, sfmmu_udtlb_fastpath	/* no 4M TSB, miss */	;\
950	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
951	.align 128
952
953
954#if defined(cscope)
955/*
956 * Define labels to direct cscope quickly to labels that
957 * are generated by macro expansion of ITLB_MISS().
958 */
959	.global	tt0_itlbmiss
960tt0_itlbmiss:
961	.global	tt1_itlbmiss
962tt1_itlbmiss:
963	nop
964#endif
965
966/*
967 * Instruction miss handler.
968 *
969 * This handler is invoked only if the hypervisor has been instructed
970 * not to do any TSB walk.
971 *
972 * ldda instructions will have their ASI patched
973 * by sfmmu_patch_ktsb at runtime.
974 * MUST be EXACTLY 32 instructions or we'll break.
975 */
976
977/*
978 * synthesize for miss handler: TAG_ACCESS in %g2 (with context "type"
979 * (0=kernel, 1=invalid, or 2=user) rather than context ID)
980 */
981#define	ITLB_MISS(table_name)						 \
982	.global	table_name/**/_itlbmiss					;\
983table_name/**/_itlbmiss:						;\
984	GET_MMU_I_PTAGACC_CTXTYPE(%g2, %g3)	/* 8 instr */		;\
985	cmp	%g3, INVALID_CONTEXT					;\
986	ble,pn	%xcc, sfmmu_kitlb_miss					;\
987	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
988	mov	SCRATCHPAD_UTSBREG2, %g1				;\
989	ldxa	[%g1]ASI_SCRATCHPAD, %g1	/* get 2nd tsbreg */	;\
990	brgez,pn %g1, sfmmu_uitlb_slowpath	/* branch if 2 TSBs */	;\
991	  nop								;\
992	GET_1ST_TSBE_PTR(%g2, %g1, %g4, %g5)	/* 11 instr */		;\
993	ba,pt	%xcc, sfmmu_uitlb_fastpath	/* no 4M TSB, miss */	;\
994	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
995	.align 128
996
997#define	DTSB_MISS \
998	GOTO_TT(sfmmu_slow_dmmu_miss,trace_dmmu)
999
1000#define	ITSB_MISS \
1001	GOTO_TT(sfmmu_slow_immu_miss,trace_immu)
1002
1003/*
1004 * This macro is the first level handler for fast protection faults.
1005 * It first demaps the tlb entry which generated the fault and then
1006 * attempts to set the modify bit on the hash.  It needs to be
1007 * exactly 32 instructions.
1008 */
1009/*
1010 * synthesize for miss handler: TAG_ACCESS in %g2 (with context "type"
1011 * (0=kernel, 1=invalid, or 2=user) rather than context ID)
1012 */
1013#define	DTLB_PROT							 \
1014	GET_MMU_D_PTAGACC_CTXTYPE(%g2, %g3)	/* 8 instr */		;\
1015	/*								;\
1016	 *   g2 = pseudo-tag access register (ctx type rather than ctx ID) ;\
1017	 *   g3 = ctx type (0, 1, or 2)					;\
1018	 */								;\
1019	TT_TRACE(trace_dataprot)	/* 2 instr ifdef TRAPTRACE */	;\
1020					/* clobbers g1 and g6 XXXQ? */	;\
1021	brnz,pt %g3, sfmmu_uprot_trap		/* user trap */		;\
1022	  nop								;\
1023	ba,a,pt	%xcc, sfmmu_kprot_trap		/* kernel trap */	;\
1024	.align 128
1025
1026#define	DMMU_EXCEPTION_TL1						;\
1027	ba,a,pt	%xcc, mmu_trap_tl1					;\
1028	.align 32
1029
1030#define	MISALIGN_ADDR_TL1						;\
1031	ba,a,pt	%xcc, mmu_trap_tl1					;\
1032	.align 32
1033
1034/*
1035 * Trace a tsb hit
1036 * g1 = tsbe pointer (in/clobbered)
1037 * g2 = tag access register (in)
1038 * g3 - g4 = scratch (clobbered)
1039 * g5 = tsbe data (in)
1040 * g6 = scratch (clobbered)
1041 * g7 = pc we jumped here from (in)
1042 * ttextra = value to OR in to trap type (%tt) (in)
1043 */
1044#ifdef TRAPTRACE
1045#define TRACE_TSBHIT(ttextra)						 \
1046	membar	#Sync							;\
1047	sethi	%hi(FLUSH_ADDR), %g6					;\
1048	flush	%g6							;\
1049	TRACE_PTR(%g3, %g6)						;\
1050	GET_TRACE_TICK(%g6)						;\
1051	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi				;\
1052	stna	%g2, [%g3 + TRAP_ENT_SP]%asi	/* tag access */	;\
1053	stna	%g5, [%g3 + TRAP_ENT_F1]%asi	/* tsb data */		;\
1054	rdpr	%tnpc, %g6						;\
1055	stna	%g6, [%g3 + TRAP_ENT_F2]%asi				;\
1056	stna	%g1, [%g3 + TRAP_ENT_F3]%asi	/* tsb pointer */	;\
1057	stna	%g0, [%g3 + TRAP_ENT_F4]%asi				;\
1058	rdpr	%tpc, %g6						;\
1059	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi				;\
1060	TRACE_SAVE_TL_GL_REGS(%g3, %g6)					;\
1061	rdpr	%tt, %g6						;\
1062	or	%g6, (ttextra), %g1					;\
1063	stha	%g1, [%g3 + TRAP_ENT_TT]%asi				;\
1064	MMU_FAULT_STATUS_AREA(%g4)					;\
1065	mov	MMFSA_D_ADDR, %g1					;\
1066	cmp	%g6, FAST_IMMU_MISS_TT					;\
1067	move	%xcc, MMFSA_I_ADDR, %g1					;\
1068	cmp	%g6, T_INSTR_MMU_MISS					;\
1069	move	%xcc, MMFSA_I_ADDR, %g1					;\
1070	ldx	[%g4 + %g1], %g1					;\
1071	stxa	%g1, [%g3 + TRAP_ENT_TSTATE]%asi /* fault addr */	;\
1072	mov	MMFSA_D_CTX, %g1					;\
1073	cmp	%g6, FAST_IMMU_MISS_TT					;\
1074	move	%xcc, MMFSA_I_CTX, %g1					;\
1075	cmp	%g6, T_INSTR_MMU_MISS					;\
1076	move	%xcc, MMFSA_I_CTX, %g1					;\
1077	ldx	[%g4 + %g1], %g1					;\
1078	stna	%g1, [%g3 + TRAP_ENT_TR]%asi				;\
1079	TRACE_NEXT(%g3, %g4, %g6)
1080#else
1081#define TRACE_TSBHIT(ttextra)
1082#endif
1083
1084
1085#if defined(lint)
1086
1087struct scb	trap_table;
1088struct scb	scb;		/* trap_table/scb are the same object */
1089
1090#else /* lint */
1091
1092/*
1093 * =======================================================================
1094 *		SPARC V9 TRAP TABLE
1095 *
1096 * The trap table is divided into two halves: the first half is used when
1097 * taking traps when TL=0; the second half is used when taking traps from
1098 * TL>0. Note that handlers in the second half of the table might not be able
1099 * to make the same assumptions as handlers in the first half of the table.
1100 *
1101 * Worst case trap nesting so far:
1102 *
1103 *	at TL=0 client issues software trap requesting service
1104 *	at TL=1 nucleus wants a register window
1105 *	at TL=2 register window clean/spill/fill takes a TLB miss
1106 *	at TL=3 processing TLB miss
1107 *	at TL=4 handle asynchronous error
1108 *
1109 * Note that a trap from TL=4 to TL=5 places Spitfire in "RED mode".
1110 *
1111 * =======================================================================
1112 */
1113	.section ".text"
1114	.align	4
1115	.global trap_table, scb, trap_table0, trap_table1, etrap_table
1116	.type	trap_table, #object
1117	.type	trap_table0, #object
1118	.type	trap_table1, #object
1119	.type	scb, #object
1120trap_table:
1121scb:
1122trap_table0:
1123	/* hardware traps */
1124	NOT;				/* 000	reserved */
1125	RED;				/* 001	power on reset */
1126	WATCHDOG_RESET;			/* 002	watchdog reset */
1127	RED;				/* 003	externally initiated reset */
1128	RED;				/* 004	software initiated reset */
1129	RED;				/* 005	red mode exception */
1130	NOT; NOT;			/* 006 - 007 reserved */
1131	IMMU_EXCEPTION;			/* 008	instruction access exception */
1132	ITSB_MISS;			/* 009	instruction access MMU miss */
1133 	NOT;				/* 00A  reserved */
1134	NOT; NOT4;			/* 00B - 00F reserved */
1135	ILLTRAP_INSTR;			/* 010	illegal instruction */
1136	TRAP(T_PRIV_INSTR);		/* 011	privileged opcode */
1137	TRAP(T_UNIMP_LDD);		/* 012	unimplemented LDD */
1138	TRAP(T_UNIMP_STD);		/* 013	unimplemented STD */
1139	NOT4; NOT4; NOT4;		/* 014 - 01F reserved */
1140	FP_DISABLED_TRAP;		/* 020	fp disabled */
1141	FP_IEEE_TRAP;			/* 021	fp exception ieee 754 */
1142	FP_TRAP;			/* 022	fp exception other */
1143	TAG_OVERFLOW;			/* 023	tag overflow */
1144	CLEAN_WINDOW;			/* 024 - 027 clean window */
1145	DIV_BY_ZERO;			/* 028	division by zero */
1146	NOT;				/* 029	internal processor error */
1147	NOT; NOT; NOT4;			/* 02A - 02F reserved */
1148	DMMU_EXCEPTION;			/* 030	data access exception */
1149	DTSB_MISS;			/* 031	data access MMU miss */
1150	NOT;				/* 032  reserved */
1151	NOT;				/* 033	data access protection */
1152	DMMU_EXC_AG_NOT_ALIGNED;	/* 034	mem address not aligned */
1153	DMMU_EXC_LDDF_NOT_ALIGNED;	/* 035	LDDF mem address not aligned */
1154	DMMU_EXC_STDF_NOT_ALIGNED;	/* 036	STDF mem address not aligned */
1155	DMMU_EXC_AG_PRIV;		/* 037	privileged action */
1156	NOT;				/* 038	LDQF mem address not aligned */
1157	NOT;				/* 039	STQF mem address not aligned */
1158	NOT; NOT; NOT4;			/* 03A - 03F reserved */
1159	NOT;				/* 040	async data error */
1160	LEVEL_INTERRUPT(1);		/* 041	interrupt level 1 */
1161	LEVEL_INTERRUPT(2);		/* 042	interrupt level 2 */
1162	LEVEL_INTERRUPT(3);		/* 043	interrupt level 3 */
1163	LEVEL_INTERRUPT(4);		/* 044	interrupt level 4 */
1164	LEVEL_INTERRUPT(5);		/* 045	interrupt level 5 */
1165	LEVEL_INTERRUPT(6);		/* 046	interrupt level 6 */
1166	LEVEL_INTERRUPT(7);		/* 047	interrupt level 7 */
1167	LEVEL_INTERRUPT(8);		/* 048	interrupt level 8 */
1168	LEVEL_INTERRUPT(9);		/* 049	interrupt level 9 */
1169	LEVEL_INTERRUPT(10);		/* 04A	interrupt level 10 */
1170	LEVEL_INTERRUPT(11);		/* 04B	interrupt level 11 */
1171	LEVEL_INTERRUPT(12);		/* 04C	interrupt level 12 */
1172	LEVEL_INTERRUPT(13);		/* 04D	interrupt level 13 */
1173	LEVEL14_INTERRUPT;		/* 04E	interrupt level 14 */
1174	LEVEL_INTERRUPT(15);		/* 04F	interrupt level 15 */
1175	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F reserved */
1176	NOT;				/* 060	interrupt vector */
1177	GOTO(kmdb_trap);		/* 061	PA watchpoint */
1178	GOTO(kmdb_trap);		/* 062	VA watchpoint */
1179	NOT;				/* 063	reserved */
1180	ITLB_MISS(tt0);			/* 064	instruction access MMU miss */
1181	DTLB_MISS(tt0);			/* 068	data access MMU miss */
1182	DTLB_PROT;			/* 06C	data access protection */
1183	NOT;				/* 070  reserved */
1184	NOT;				/* 071  reserved */
1185	NOT;				/* 072  reserved */
1186	NOT;				/* 073  reserved */
1187	NOT4; NOT4			/* 074 - 07B reserved */
1188	CPU_MONDO;			/* 07C	cpu_mondo */
1189	DEV_MONDO;			/* 07D	dev_mondo */
1190	GOTO_TT(resumable_error, trace_gen);	/* 07E  resumable error */
1191	GOTO_TT(nonresumable_error, trace_gen);	/* 07F  non-reasumable error */
1192	NOT4;				/* 080	spill 0 normal */
1193	SPILL_32bit_asi(ASI_AIUP,sn0);	/* 084	spill 1 normal */
1194	SPILL_64bit_asi(ASI_AIUP,sn0);	/* 088	spill 2 normal */
1195	SPILL_32clean(ASI_AIUP,sn0);	/* 08C	spill 3 normal */
1196	SPILL_64clean(ASI_AIUP,sn0);	/* 090	spill 4 normal */
1197	SPILL_32bit(not);		/* 094	spill 5 normal */
1198	SPILL_64bit(not);		/* 098	spill 6 normal */
1199	SPILL_mixed;			/* 09C	spill 7 normal */
1200	NOT4;				/* 0A0	spill 0 other */
1201	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0A4	spill 1 other */
1202	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0A8	spill 2 other */
1203	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0AC	spill 3 other */
1204	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0B0	spill 4 other */
1205	NOT4;				/* 0B4	spill 5 other */
1206	NOT4;				/* 0B8	spill 6 other */
1207	NOT4;				/* 0BC	spill 7 other */
1208	NOT4;				/* 0C0	fill 0 normal */
1209	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0C4	fill 1 normal */
1210	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0C8	fill 2 normal */
1211	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0CC	fill 3 normal */
1212	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0D0	fill 4 normal */
1213	FILL_32bit(not);		/* 0D4	fill 5 normal */
1214	FILL_64bit(not);		/* 0D8	fill 6 normal */
1215	FILL_mixed;			/* 0DC	fill 7 normal */
1216	NOT4;				/* 0E0	fill 0 other */
1217	NOT4;				/* 0E4	fill 1 other */
1218	NOT4;				/* 0E8	fill 2 other */
1219	NOT4;				/* 0EC	fill 3 other */
1220	NOT4;				/* 0F0	fill 4 other */
1221	NOT4;				/* 0F4	fill 5 other */
1222	NOT4;				/* 0F8	fill 6 other */
1223	NOT4;				/* 0FC	fill 7 other */
1224	/* user traps */
1225	GOTO(syscall_trap_4x);		/* 100	old system call */
1226	TRAP(T_BREAKPOINT);		/* 101	user breakpoint */
1227	TRAP(T_DIV0);			/* 102	user divide by zero */
1228	GOTO(.flushw);			/* 103	flush windows */
1229	GOTO(.clean_windows);		/* 104	clean windows */
1230	BAD;				/* 105	range check ?? */
1231	GOTO(.fix_alignment);		/* 106	do unaligned references */
1232	BAD;				/* 107	unused */
1233	SYSCALL_TRAP32;			/* 108	ILP32 system call on LP64 */
1234	GOTO(set_trap0_addr);		/* 109	set trap0 address */
1235	BAD; BAD; BAD4;			/* 10A - 10F unused */
1236	TRP4; TRP4; TRP4; TRP4;		/* 110 - 11F V9 user trap handlers */
1237	GOTO(.getcc);			/* 120	get condition codes */
1238	GOTO(.setcc);			/* 121	set condition codes */
1239	GOTO(.getpsr);			/* 122	get psr */
1240	GOTO(.setpsr);			/* 123	set psr (some fields) */
1241	GOTO(get_timestamp);		/* 124	get timestamp */
1242	GOTO(get_virtime);		/* 125	get lwp virtual time */
1243	PRIV(self_xcall);		/* 126	self xcall */
1244	GOTO(get_hrestime);		/* 127	get hrestime */
1245	BAD;				/* 128	ST_SETV9STACK */
1246	GOTO(.getlgrp);			/* 129  get lgrpid */
1247	BAD; BAD; BAD4;			/* 12A - 12F unused */
1248	BAD4; BAD4; 			/* 130 - 137 unused */
1249	DTRACE_PID;			/* 138  dtrace pid tracing provider */
1250	BAD;				/* 139  unused */
1251	DTRACE_RETURN;			/* 13A	dtrace pid return probe */
1252	BAD; BAD4;			/* 13B - 13F unused */
1253	SYSCALL_TRAP;			/* 140  LP64 system call */
1254	SYSCALL(nosys);			/* 141  unused system call trap */
1255#ifdef DEBUG_USER_TRAPTRACECTL
1256	GOTO(.traptrace_freeze);	/* 142  freeze traptrace */
1257	GOTO(.traptrace_unfreeze);	/* 143  unfreeze traptrace */
1258#else
1259	SYSCALL(nosys);			/* 142  unused system call trap */
1260	SYSCALL(nosys);			/* 143  unused system call trap */
1261#endif
1262	BAD4; BAD4; BAD4;		/* 144 - 14F unused */
1263	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1264	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1265	BAD;				/* 170 - unused */
1266	BAD;				/* 171 - unused */
1267	BAD; BAD;			/* 172 - 173 unused */
1268	BAD4; BAD4;			/* 174 - 17B unused */
1269#ifdef	PTL1_PANIC_DEBUG
1270	mov PTL1_BAD_DEBUG, %g1; GOTO(ptl1_panic);
1271					/* 17C	test ptl1_panic */
1272#else
1273	BAD;				/* 17C  unused */
1274#endif	/* PTL1_PANIC_DEBUG */
1275	PRIV(kmdb_trap);		/* 17D	kmdb enter (L1-A) */
1276	PRIV(kmdb_trap);		/* 17E	kmdb breakpoint */
1277	PRIV(obp_bpt);			/* 17F	obp breakpoint */
1278	/* reserved */
1279	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1280	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1281	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1282	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1283	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1284	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1285	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1286	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1287	.size	trap_table0, (.-trap_table0)
1288trap_table1:
1289	NOT4; NOT4;			/* 000 - 007 unused */
1290	NOT;				/* 008	instruction access exception */
1291	ITSB_MISS;			/* 009	instruction access MMU miss */
1292 	NOT;				/* 00A  reserved */
1293	NOT; NOT4;			/* 00B - 00F unused */
1294	NOT4; NOT4; NOT4; NOT4;		/* 010 - 01F unused */
1295	NOT4;				/* 020 - 023 unused */
1296	CLEAN_WINDOW;			/* 024 - 027 clean window */
1297	NOT4; NOT4;			/* 028 - 02F unused */
1298	DMMU_EXCEPTION_TL1;		/* 030 	data access exception */
1299	DTSB_MISS;			/* 031  data access MMU miss */
1300	NOT;				/* 032  reserved */
1301	NOT;				/* 033	unused */
1302	MISALIGN_ADDR_TL1;		/* 034	mem address not aligned */
1303	NOT; NOT; NOT; NOT4; NOT4	/* 035 - 03F unused */
1304	NOT4; NOT4; NOT4; NOT4;		/* 040 - 04F unused */
1305	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F unused */
1306	NOT;				/* 060	unused */
1307	GOTO(kmdb_trap_tl1);		/* 061	PA watchpoint */
1308	GOTO(kmdb_trap_tl1);		/* 062	VA watchpoint */
1309	NOT;				/* 063	reserved */
1310	ITLB_MISS(tt1);			/* 064	instruction access MMU miss */
1311	DTLB_MISS(tt1);			/* 068	data access MMU miss */
1312	DTLB_PROT;			/* 06C	data access protection */
1313	NOT;				/* 070  reserved */
1314	NOT;				/* 071  reserved */
1315	NOT;				/* 072  reserved */
1316	NOT;				/* 073  reserved */
1317	NOT4; NOT4;			/* 074 - 07B reserved */
1318	NOT;				/* 07C  reserved */
1319	NOT;				/* 07D  reserved */
1320	NOT;				/* 07E  resumable error */
1321	GOTO_TT(nonresumable_error, trace_gen);	/* 07F  nonresumable error */
1322	NOTP4;				/* 080	spill 0 normal */
1323	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 084	spill 1 normal */
1324	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 088	spill 2 normal */
1325	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 08C	spill 3 normal */
1326	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 090	spill 4 normal */
1327	NOTP4;				/* 094	spill 5 normal */
1328	SPILL_64bit_ktt1(sk);		/* 098	spill 6 normal */
1329	SPILL_mixed_ktt1(sk);		/* 09C	spill 7 normal */
1330	NOTP4;				/* 0A0	spill 0 other */
1331	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0A4  spill 1 other */
1332	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0A8	spill 2 other */
1333	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0AC	spill 3 other */
1334	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0B0  spill 4 other */
1335	NOTP4;				/* 0B4  spill 5 other */
1336	NOTP4;				/* 0B8  spill 6 other */
1337	NOTP4;				/* 0BC  spill 7 other */
1338	NOT4;				/* 0C0	fill 0 normal */
1339	NOT4;				/* 0C4	fill 1 normal */
1340	NOT4;				/* 0C8	fill 2 normal */
1341	NOT4;				/* 0CC	fill 3 normal */
1342	NOT4;				/* 0D0	fill 4 normal */
1343	NOT4;				/* 0D4	fill 5 normal */
1344	NOT4;				/* 0D8	fill 6 normal */
1345	NOT4;				/* 0DC	fill 7 normal */
1346	NOT4; NOT4; NOT4; NOT4;		/* 0E0 - 0EF unused */
1347	NOT4; NOT4; NOT4; NOT4;		/* 0F0 - 0FF unused */
1348/*
1349 * Code running at TL>0 does not use soft traps, so
1350 * we can truncate the table here.
1351 * However:
1352 * sun4v uses (hypervisor) ta instructions at TL > 0, so
1353 * provide a safety net for now.
1354 */
1355	/* soft traps */
1356	BAD4; BAD4; BAD4; BAD4;		/* 100 - 10F unused */
1357	BAD4; BAD4; BAD4; BAD4;		/* 110 - 11F unused */
1358	BAD4; BAD4; BAD4; BAD4;		/* 120 - 12F unused */
1359	BAD4; BAD4; BAD4; BAD4;		/* 130 - 13F unused */
1360	BAD4; BAD4; BAD4; BAD4;		/* 140 - 14F unused */
1361	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1362	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1363	BAD4; BAD4; BAD4; BAD4;		/* 170 - 17F unused */
1364	/* reserved */
1365	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1366	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1367	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1368	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1369	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1370	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1371	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1372	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1373etrap_table:
1374	.size	trap_table1, (.-trap_table1)
1375	.size	trap_table, (.-trap_table)
1376	.size	scb, (.-scb)
1377
1378/*
1379 * We get to exec_fault in the case of an instruction miss and tte
1380 * has no execute bit set.  We go to tl0 to handle it.
1381 *
1382 * g1 = tsbe pointer (in/clobbered)
1383 * g2 = tag access register (in)
1384 * g3 - g4 = scratch (clobbered)
1385 * g5 = tsbe data (in)
1386 * g6 = scratch (clobbered)
1387 * g7 = pc we jumped here from (in)
1388 */
1389/*
1390 * synthesize for miss handler: TAG_ACCESS in %g2 (with context "type"
1391 * (0=kernel, 1=invalid, or 2=user) rather than context ID)
1392 */
1393	ALTENTRY(exec_fault)
1394	set	icache_is_coherent, %g6		/* check soft exec mode */
1395	ld	[%g6], %g6
1396	brz,pn	%g6, sfmmu_slow_immu_miss
1397	  nop
1398	TRACE_TSBHIT(TT_MMU_EXEC)
1399	MMU_FAULT_STATUS_AREA(%g4)
1400	ldx	[%g4 + MMFSA_I_ADDR], %g2	/* g2 = address */
1401	ldx	[%g4 + MMFSA_I_CTX], %g3	/* g3 = ctx */
1402	srlx	%g2, MMU_PAGESHIFT, %g2		! align address to page boundry
1403	cmp	%g3, USER_CONTEXT_TYPE
1404	sllx	%g2, MMU_PAGESHIFT, %g2
1405	movgu	%icc, USER_CONTEXT_TYPE, %g3
1406	or	%g2, %g3, %g2			/* TAG_ACCESS */
1407	mov	T_INSTR_MMU_MISS, %g3		! arg2 = traptype
1408	set	trap, %g1
1409	ba,pt	%xcc, sys_trap
1410	  mov	-1, %g4
1411
1412.mmu_exception_not_aligned:
1413	/* %g2 = sfar, %g3 = sfsr */
1414	rdpr	%tstate, %g1
1415	btst	TSTATE_PRIV, %g1
1416	bnz,pn	%icc, 2f
1417	nop
1418	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1419	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1420	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1421	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1422	brz,pt	%g5, 2f
1423	nop
1424	ldn	[%g5 + P_UTRAP15], %g5			! unaligned utrap?
1425	brz,pn	%g5, 2f
1426	nop
1427	btst	1, %sp
1428	bz,pt	%xcc, 1f				! 32 bit user program
1429	nop
1430	ba,pt	%xcc, .setup_v9utrap			! 64 bit user program
1431	nop
14321:
1433	ba,pt	%xcc, .setup_utrap
1434	or	%g2, %g0, %g7
14352:
1436	ba,pt	%xcc, .mmu_exception_end
1437	mov	T_ALIGNMENT, %g1
1438
1439.mmu_priv_exception:
1440	rdpr	%tstate, %g1
1441	btst	TSTATE_PRIV, %g1
1442	bnz,pn	%icc, 1f
1443	nop
1444	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1445	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1446	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1447	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1448	brz,pt	%g5, 1f
1449	nop
1450	ldn	[%g5 + P_UTRAP16], %g5
1451	brnz,pt	%g5, .setup_v9utrap
1452	nop
14531:
1454	mov	T_PRIV_INSTR, %g1
1455
1456.mmu_exception_end:
1457	CPU_INDEX(%g4, %g5)
1458	set	cpu_core, %g5
1459	sllx	%g4, CPU_CORE_SHIFT, %g4
1460	add	%g4, %g5, %g4
1461	lduh	[%g4 + CPUC_DTRACE_FLAGS], %g5
1462	andcc	%g5, CPU_DTRACE_NOFAULT, %g0
1463	bz	1f
1464	or	%g5, CPU_DTRACE_BADADDR, %g5
1465	stuh	%g5, [%g4 + CPUC_DTRACE_FLAGS]
1466	done
1467
14681:
1469	sllx	%g3, 32, %g3
1470	or	%g3, %g1, %g3
1471	set	trap, %g1
1472	ba,pt	%xcc, sys_trap
1473	sub	%g0, 1, %g4
1474
1475.fp_disabled:
1476	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1477	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1478	rdpr	%tstate, %g4
1479	btst	TSTATE_PRIV, %g4
1480	bnz,a,pn %icc, ptl1_panic
1481	  mov	PTL1_BAD_FPTRAP, %g1
1482
1483	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1484	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1485	brz,a,pt %g5, 2f
1486	  nop
1487	ldn	[%g5 + P_UTRAP7], %g5			! fp_disabled utrap?
1488	brz,a,pn %g5, 2f
1489	  nop
1490	btst	1, %sp
1491	bz,a,pt	%xcc, 1f				! 32 bit user program
1492	  nop
1493	ba,a,pt	%xcc, .setup_v9utrap			! 64 bit user program
1494	  nop
14951:
1496	ba,pt	%xcc, .setup_utrap
1497	  or	%g0, %g0, %g7
14982:
1499	set	fp_disabled, %g1
1500	ba,pt	%xcc, sys_trap
1501	  sub	%g0, 1, %g4
1502
1503.fp_ieee_exception:
1504	rdpr	%tstate, %g1
1505	btst	TSTATE_PRIV, %g1
1506	bnz,a,pn %icc, ptl1_panic
1507	  mov	PTL1_BAD_FPTRAP, %g1
1508	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1509	stx	%fsr, [%g1 + CPU_TMP1]
1510	ldx	[%g1 + CPU_TMP1], %g2
1511	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1512	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1513	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1514	brz,a,pt %g5, 1f
1515	  nop
1516	ldn	[%g5 + P_UTRAP8], %g5
1517	brnz,a,pt %g5, .setup_v9utrap
1518	  nop
15191:
1520	set	_fp_ieee_exception, %g1
1521	ba,pt	%xcc, sys_trap
1522	  sub	%g0, 1, %g4
1523
1524/*
1525 * Register Inputs:
1526 *	%g5		user trap handler
1527 *	%g7		misaligned addr - for alignment traps only
1528 */
1529.setup_utrap:
1530	set	trap, %g1			! setup in case we go
1531	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1532	sub	%g0, 1, %g4			! the save instruction below
1533
1534	/*
1535	 * If the DTrace pid provider is single stepping a copied-out
1536	 * instruction, t->t_dtrace_step will be set. In that case we need
1537	 * to abort the single-stepping (since execution of the instruction
1538	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1539	 */
1540	save	%sp, -SA(MINFRAME32), %sp	! window for trap handler
1541	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1542	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1543	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1544	rdpr	%tnpc, %l2			! arg1 == tnpc
1545	brz,pt	%g2, 1f
1546	rdpr	%tpc, %l1			! arg0 == tpc
1547
1548	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1549	ldn	[%g1 + T_DTRACE_NPC], %l2	! arg1 = t->t_dtrace_npc (step)
1550	brz,pt	%g2, 1f
1551	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1552	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
15531:
1554	mov	%g7, %l3			! arg2 == misaligned address
1555
1556	rdpr	%tstate, %g1			! cwp for trap handler
1557	rdpr	%cwp, %g4
1558	bclr	TSTATE_CWP_MASK, %g1
1559	wrpr	%g1, %g4, %tstate
1560	wrpr	%g0, %g5, %tnpc			! trap handler address
1561	FAST_TRAP_DONE
1562	/* NOTREACHED */
1563
1564.check_v9utrap:
1565	rdpr	%tstate, %g1
1566	btst	TSTATE_PRIV, %g1
1567	bnz,a,pn %icc, 3f
1568	  nop
1569	CPU_ADDR(%g4, %g1)				! load CPU struct addr
1570	ldn	[%g4 + CPU_THREAD], %g5			! load thread pointer
1571	ldn	[%g5 + T_PROCP], %g5			! load proc pointer
1572	ldn	[%g5 + P_UTRAPS], %g5			! are there utraps?
1573
1574	cmp	%g3, T_SOFTWARE_TRAP
1575	bne,a,pt %icc, 1f
1576	  nop
1577
1578	brz,pt %g5, 3f			! if p_utraps == NULL goto trap()
1579	  rdpr	%tt, %g3		! delay - get actual hw trap type
1580
1581	sub	%g3, 254, %g1		! UT_TRAP_INSTRUCTION_16 = p_utraps[18]
1582	ba,pt	%icc, 2f
1583	  smul	%g1, CPTRSIZE, %g2
15841:
1585	brz,a,pt %g5, 3f		! if p_utraps == NULL goto trap()
1586	  nop
1587
1588	cmp	%g3, T_UNIMP_INSTR
1589	bne,a,pt %icc, 2f
1590	  nop
1591
1592	mov	1, %g1
1593	st	%g1, [%g4 + CPU_TL1_HDLR] ! set CPU_TL1_HDLR
1594	rdpr	%tpc, %g1		! ld trapping instruction using
1595	lduwa	[%g1]ASI_AIUP, %g1	! "AS IF USER" ASI which could fault
1596	st	%g0, [%g4 + CPU_TL1_HDLR] ! clr CPU_TL1_HDLR
1597
1598	sethi	%hi(0xc1c00000), %g4	! setup mask for illtrap instruction
1599	andcc	%g1, %g4, %g4		! and instruction with mask
1600	bnz,a,pt %icc, 3f		! if %g4 == zero, %g1 is an ILLTRAP
1601	  nop				! fall thru to setup
16022:
1603	ldn	[%g5 + %g2], %g5
1604	brnz,a,pt %g5, .setup_v9utrap
1605	  nop
16063:
1607	set	trap, %g1
1608	ba,pt	%xcc, sys_trap
1609	  sub	%g0, 1, %g4
1610	/* NOTREACHED */
1611
1612/*
1613 * Register Inputs:
1614 *	%g5		user trap handler
1615 */
1616.setup_v9utrap:
1617	set	trap, %g1			! setup in case we go
1618	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1619	sub	%g0, 1, %g4			! the save instruction below
1620
1621	/*
1622	 * If the DTrace pid provider is single stepping a copied-out
1623	 * instruction, t->t_dtrace_step will be set. In that case we need
1624	 * to abort the single-stepping (since execution of the instruction
1625	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1626	 */
1627	save	%sp, -SA(MINFRAME64), %sp	! window for trap handler
1628	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1629	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1630	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1631	rdpr	%tnpc, %l7			! arg1 == tnpc
1632	brz,pt	%g2, 1f
1633	rdpr	%tpc, %l6			! arg0 == tpc
1634
1635	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1636	ldn	[%g1 + T_DTRACE_NPC], %l7	! arg1 == t->t_dtrace_npc (step)
1637	brz,pt	%g2, 1f
1638	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1639	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
16401:
1641	rdpr	%tstate, %g2			! cwp for trap handler
1642	rdpr	%cwp, %g4
1643	bclr	TSTATE_CWP_MASK, %g2
1644	wrpr	%g2, %g4, %tstate
1645
1646	ldn	[%g1 + T_PROCP], %g4		! load proc pointer
1647	ldn	[%g4 + P_AS], %g4		! load as pointer
1648	ldn	[%g4 + A_USERLIMIT], %g4	! load as userlimit
1649	cmp	%l7, %g4			! check for single-step set
1650	bne,pt	%xcc, 4f
1651	  nop
1652	ldn	[%g1 + T_LWP], %g1		! load klwp pointer
1653	ld	[%g1 + PCB_STEP], %g4		! load single-step flag
1654	cmp	%g4, STEP_ACTIVE		! step flags set in pcb?
1655	bne,pt	%icc, 4f
1656	  nop
1657	stn	%g5, [%g1 + PCB_TRACEPC]	! save trap handler addr in pcb
1658	mov	%l7, %g4			! on entry to precise user trap
1659	add	%l6, 4, %l7			! handler, %l6 == pc, %l7 == npc
1660						! at time of trap
1661	wrpr	%g0, %g4, %tnpc			! generate FLTBOUNDS,
1662						! %g4 == userlimit
1663	FAST_TRAP_DONE
1664	/* NOTREACHED */
16654:
1666	wrpr	%g0, %g5, %tnpc			! trap handler address
1667	FAST_TRAP_DONE_CHK_INTR
1668	/* NOTREACHED */
1669
1670.fp_exception:
1671	CPU_ADDR(%g1, %g4)
1672	stx	%fsr, [%g1 + CPU_TMP1]
1673	ldx	[%g1 + CPU_TMP1], %g2
1674
1675	/*
1676	 * Cheetah takes unfinished_FPop trap for certain range of operands
1677	 * to the "fitos" instruction. Instead of going through the slow
1678	 * software emulation path, we try to simulate the "fitos" instruction
1679	 * via "fitod" and "fdtos" provided the following conditions are met:
1680	 *
1681	 *	fpu_exists is set (if DEBUG)
1682	 *	not in privileged mode
1683	 *	ftt is unfinished_FPop
1684	 *	NXM IEEE trap is not enabled
1685	 *	instruction at %tpc is "fitos"
1686	 *
1687	 *  Usage:
1688	 *	%g1	per cpu address
1689	 *	%g2	%fsr
1690	 *	%g6	user instruction
1691	 *
1692	 * Note that we can take a memory access related trap while trying
1693	 * to fetch the user instruction. Therefore, we set CPU_TL1_HDLR
1694	 * flag to catch those traps and let the SFMMU code deal with page
1695	 * fault and data access exception.
1696	 */
1697#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
1698	sethi	%hi(fpu_exists), %g7
1699	ld	[%g7 + %lo(fpu_exists)], %g7
1700	brz,pn %g7, .fp_exception_cont
1701	  nop
1702#endif
1703	rdpr	%tstate, %g7			! branch if in privileged mode
1704	btst	TSTATE_PRIV, %g7
1705	bnz,pn	%xcc, .fp_exception_cont
1706	srl	%g2, FSR_FTT_SHIFT, %g7		! extract ftt from %fsr
1707	and	%g7, (FSR_FTT>>FSR_FTT_SHIFT), %g7
1708	cmp	%g7, FTT_UNFIN
1709	set	FSR_TEM_NX, %g5
1710	bne,pn	%xcc, .fp_exception_cont	! branch if NOT unfinished_FPop
1711	  andcc	%g2, %g5, %g0
1712	bne,pn	%xcc, .fp_exception_cont	! branch if FSR_TEM_NX enabled
1713	  rdpr	%tpc, %g5			! get faulting PC
1714
1715	or	%g0, 1, %g7
1716	st	%g7, [%g1 + CPU_TL1_HDLR]	! set tl1_hdlr flag
1717	lda	[%g5]ASI_USER, %g6		! get user's instruction
1718	st	%g0, [%g1 + CPU_TL1_HDLR]	! clear tl1_hdlr flag
1719
1720	set	FITOS_INSTR_MASK, %g7
1721	and	%g6, %g7, %g7
1722	set	FITOS_INSTR, %g5
1723	cmp	%g7, %g5
1724	bne,pn	%xcc, .fp_exception_cont	! branch if not FITOS_INSTR
1725	 nop
1726
1727	/*
1728	 * This is unfinished FPops trap for "fitos" instruction. We
1729	 * need to simulate "fitos" via "fitod" and "fdtos" instruction
1730	 * sequence.
1731	 *
1732	 * We need a temporary FP register to do the conversion. Since
1733	 * both source and destination operands for the "fitos" instruction
1734	 * have to be within %f0-%f31, we use an FP register from the upper
1735	 * half to guarantee that it won't collide with the source or the
1736	 * dest operand. However, we do have to save and restore its value.
1737	 *
1738	 * We use %d62 as a temporary FP register for the conversion and
1739	 * branch to appropriate instruction within the conversion tables
1740	 * based upon the rs2 and rd values.
1741	 */
1742
1743	std	%d62, [%g1 + CPU_TMP1]		! save original value
1744
1745	srl	%g6, FITOS_RS2_SHIFT, %g7
1746	and	%g7, FITOS_REG_MASK, %g7
1747	set	_fitos_fitod_table, %g4
1748	sllx	%g7, 2, %g7
1749	jmp	%g4 + %g7
1750	  ba,pt	%xcc, _fitos_fitod_done
1751	.empty
1752
1753_fitos_fitod_table:
1754	  fitod	%f0, %d62
1755	  fitod	%f1, %d62
1756	  fitod	%f2, %d62
1757	  fitod	%f3, %d62
1758	  fitod	%f4, %d62
1759	  fitod	%f5, %d62
1760	  fitod	%f6, %d62
1761	  fitod	%f7, %d62
1762	  fitod	%f8, %d62
1763	  fitod	%f9, %d62
1764	  fitod	%f10, %d62
1765	  fitod	%f11, %d62
1766	  fitod	%f12, %d62
1767	  fitod	%f13, %d62
1768	  fitod	%f14, %d62
1769	  fitod	%f15, %d62
1770	  fitod	%f16, %d62
1771	  fitod	%f17, %d62
1772	  fitod	%f18, %d62
1773	  fitod	%f19, %d62
1774	  fitod	%f20, %d62
1775	  fitod	%f21, %d62
1776	  fitod	%f22, %d62
1777	  fitod	%f23, %d62
1778	  fitod	%f24, %d62
1779	  fitod	%f25, %d62
1780	  fitod	%f26, %d62
1781	  fitod	%f27, %d62
1782	  fitod	%f28, %d62
1783	  fitod	%f29, %d62
1784	  fitod	%f30, %d62
1785	  fitod	%f31, %d62
1786_fitos_fitod_done:
1787
1788	/*
1789	 * Now convert data back into single precision
1790	 */
1791	srl	%g6, FITOS_RD_SHIFT, %g7
1792	and	%g7, FITOS_REG_MASK, %g7
1793	set	_fitos_fdtos_table, %g4
1794	sllx	%g7, 2, %g7
1795	jmp	%g4 + %g7
1796	  ba,pt	%xcc, _fitos_fdtos_done
1797	.empty
1798
1799_fitos_fdtos_table:
1800	  fdtos	%d62, %f0
1801	  fdtos	%d62, %f1
1802	  fdtos	%d62, %f2
1803	  fdtos	%d62, %f3
1804	  fdtos	%d62, %f4
1805	  fdtos	%d62, %f5
1806	  fdtos	%d62, %f6
1807	  fdtos	%d62, %f7
1808	  fdtos	%d62, %f8
1809	  fdtos	%d62, %f9
1810	  fdtos	%d62, %f10
1811	  fdtos	%d62, %f11
1812	  fdtos	%d62, %f12
1813	  fdtos	%d62, %f13
1814	  fdtos	%d62, %f14
1815	  fdtos	%d62, %f15
1816	  fdtos	%d62, %f16
1817	  fdtos	%d62, %f17
1818	  fdtos	%d62, %f18
1819	  fdtos	%d62, %f19
1820	  fdtos	%d62, %f20
1821	  fdtos	%d62, %f21
1822	  fdtos	%d62, %f22
1823	  fdtos	%d62, %f23
1824	  fdtos	%d62, %f24
1825	  fdtos	%d62, %f25
1826	  fdtos	%d62, %f26
1827	  fdtos	%d62, %f27
1828	  fdtos	%d62, %f28
1829	  fdtos	%d62, %f29
1830	  fdtos	%d62, %f30
1831	  fdtos	%d62, %f31
1832_fitos_fdtos_done:
1833
1834	ldd	[%g1 + CPU_TMP1], %d62		! restore %d62
1835
1836#if DEBUG
1837	/*
1838	 * Update FPop_unfinished trap kstat
1839	 */
1840	set	fpustat+FPUSTAT_UNFIN_KSTAT, %g7
1841	ldx	[%g7], %g5
18421:
1843	add	%g5, 1, %g6
1844
1845	casxa	[%g7] ASI_N, %g5, %g6
1846	cmp	%g5, %g6
1847	bne,a,pn %xcc, 1b
1848	  or	%g0, %g6, %g5
1849
1850	/*
1851	 * Update fpu_sim_fitos kstat
1852	 */
1853	set	fpuinfo+FPUINFO_FITOS_KSTAT, %g7
1854	ldx	[%g7], %g5
18551:
1856	add	%g5, 1, %g6
1857
1858	casxa	[%g7] ASI_N, %g5, %g6
1859	cmp	%g5, %g6
1860	bne,a,pn %xcc, 1b
1861	  or	%g0, %g6, %g5
1862#endif /* DEBUG */
1863
1864	FAST_TRAP_DONE
1865
1866.fp_exception_cont:
1867	/*
1868	 * Let _fp_exception deal with simulating FPop instruction.
1869	 * Note that we need to pass %fsr in %g2 (already read above).
1870	 */
1871
1872	set	_fp_exception, %g1
1873	ba,pt	%xcc, sys_trap
1874	sub	%g0, 1, %g4
1875
1876
1877/*
1878 * Register windows
1879 */
1880.flushw:
1881.clean_windows:
1882	rdpr	%tnpc, %g1
1883	wrpr	%g1, %tpc
1884	add	%g1, 4, %g1
1885	wrpr	%g1, %tnpc
1886	set	trap, %g1
1887	mov	T_FLUSH_PCB, %g3
1888	ba,pt	%xcc, sys_trap
1889	sub	%g0, 1, %g4
1890
1891/*
1892 * .spill_clean: clean the previous window, restore the wstate, and
1893 * "done".
1894 *
1895 * Entry: %g7 contains new wstate
1896 */
1897.spill_clean:
1898	sethi	%hi(nwin_minus_one), %g5
1899	ld	[%g5 + %lo(nwin_minus_one)], %g5 ! %g5 = nwin - 1
1900	rdpr	%cwp, %g6			! %g6 = %cwp
1901	deccc	%g6				! %g6--
1902	movneg	%xcc, %g5, %g6			! if (%g6<0) %g6 = nwin-1
1903	wrpr	%g6, %cwp
1904	TT_TRACE_L(trace_win)
1905	clr	%l0
1906	clr	%l1
1907	clr	%l2
1908	clr	%l3
1909	clr	%l4
1910	clr	%l5
1911	clr	%l6
1912	clr	%l7
1913	wrpr	%g0, %g7, %wstate
1914	saved
1915	retry			! restores correct %cwp
1916
1917.fix_alignment:
1918	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
1919	ldn	[%g1 + CPU_THREAD], %g1	! load thread pointer
1920	ldn	[%g1 + T_PROCP], %g1
1921	mov	1, %g2
1922	stb	%g2, [%g1 + P_FIXALIGNMENT]
1923	FAST_TRAP_DONE
1924
1925#define	STDF_REG(REG, ADDR, TMP)		\
1926	sll	REG, 3, REG			;\
1927mark1:	set	start1, TMP			;\
1928	jmp	REG + TMP			;\
1929	  nop					;\
1930start1:	ba,pt	%xcc, done1			;\
1931	  std	%f0, [ADDR + CPU_TMP1]		;\
1932	ba,pt	%xcc, done1			;\
1933	  std	%f32, [ADDR + CPU_TMP1]		;\
1934	ba,pt	%xcc, done1			;\
1935	  std	%f2, [ADDR + CPU_TMP1]		;\
1936	ba,pt	%xcc, done1			;\
1937	  std	%f34, [ADDR + CPU_TMP1]		;\
1938	ba,pt	%xcc, done1			;\
1939	  std	%f4, [ADDR + CPU_TMP1]		;\
1940	ba,pt	%xcc, done1			;\
1941	  std	%f36, [ADDR + CPU_TMP1]		;\
1942	ba,pt	%xcc, done1			;\
1943	  std	%f6, [ADDR + CPU_TMP1]		;\
1944	ba,pt	%xcc, done1			;\
1945	  std	%f38, [ADDR + CPU_TMP1]		;\
1946	ba,pt	%xcc, done1			;\
1947	  std	%f8, [ADDR + CPU_TMP1]		;\
1948	ba,pt	%xcc, done1			;\
1949	  std	%f40, [ADDR + CPU_TMP1]		;\
1950	ba,pt	%xcc, done1			;\
1951	  std	%f10, [ADDR + CPU_TMP1]		;\
1952	ba,pt	%xcc, done1			;\
1953	  std	%f42, [ADDR + CPU_TMP1]		;\
1954	ba,pt	%xcc, done1			;\
1955	  std	%f12, [ADDR + CPU_TMP1]		;\
1956	ba,pt	%xcc, done1			;\
1957	  std	%f44, [ADDR + CPU_TMP1]		;\
1958	ba,pt	%xcc, done1			;\
1959	  std	%f14, [ADDR + CPU_TMP1]		;\
1960	ba,pt	%xcc, done1			;\
1961	  std	%f46, [ADDR + CPU_TMP1]		;\
1962	ba,pt	%xcc, done1			;\
1963	  std	%f16, [ADDR + CPU_TMP1]		;\
1964	ba,pt	%xcc, done1			;\
1965	  std	%f48, [ADDR + CPU_TMP1]		;\
1966	ba,pt	%xcc, done1			;\
1967	  std	%f18, [ADDR + CPU_TMP1]		;\
1968	ba,pt	%xcc, done1			;\
1969	  std	%f50, [ADDR + CPU_TMP1]		;\
1970	ba,pt	%xcc, done1			;\
1971	  std	%f20, [ADDR + CPU_TMP1]		;\
1972	ba,pt	%xcc, done1			;\
1973	  std	%f52, [ADDR + CPU_TMP1]		;\
1974	ba,pt	%xcc, done1			;\
1975	  std	%f22, [ADDR + CPU_TMP1]		;\
1976	ba,pt	%xcc, done1			;\
1977	  std	%f54, [ADDR + CPU_TMP1]		;\
1978	ba,pt	%xcc, done1			;\
1979	  std	%f24, [ADDR + CPU_TMP1]		;\
1980	ba,pt	%xcc, done1			;\
1981	  std	%f56, [ADDR + CPU_TMP1]		;\
1982	ba,pt	%xcc, done1			;\
1983	  std	%f26, [ADDR + CPU_TMP1]		;\
1984	ba,pt	%xcc, done1			;\
1985	  std	%f58, [ADDR + CPU_TMP1]		;\
1986	ba,pt	%xcc, done1			;\
1987	  std	%f28, [ADDR + CPU_TMP1]		;\
1988	ba,pt	%xcc, done1			;\
1989	  std	%f60, [ADDR + CPU_TMP1]		;\
1990	ba,pt	%xcc, done1			;\
1991	  std	%f30, [ADDR + CPU_TMP1]		;\
1992	ba,pt	%xcc, done1			;\
1993	  std	%f62, [ADDR + CPU_TMP1]		;\
1994done1:
1995
1996#define	LDDF_REG(REG, ADDR, TMP)		\
1997	sll	REG, 3, REG			;\
1998mark2:	set	start2, TMP			;\
1999	jmp	REG + TMP			;\
2000	  nop					;\
2001start2:	ba,pt	%xcc, done2			;\
2002	  ldd	[ADDR + CPU_TMP1], %f0		;\
2003	ba,pt	%xcc, done2			;\
2004	  ldd	[ADDR + CPU_TMP1], %f32		;\
2005	ba,pt	%xcc, done2			;\
2006	  ldd	[ADDR + CPU_TMP1], %f2		;\
2007	ba,pt	%xcc, done2			;\
2008	  ldd	[ADDR + CPU_TMP1], %f34		;\
2009	ba,pt	%xcc, done2			;\
2010	  ldd	[ADDR + CPU_TMP1], %f4		;\
2011	ba,pt	%xcc, done2			;\
2012	  ldd	[ADDR + CPU_TMP1], %f36		;\
2013	ba,pt	%xcc, done2			;\
2014	  ldd	[ADDR + CPU_TMP1], %f6		;\
2015	ba,pt	%xcc, done2			;\
2016	  ldd	[ADDR + CPU_TMP1], %f38		;\
2017	ba,pt	%xcc, done2			;\
2018	  ldd	[ADDR + CPU_TMP1], %f8		;\
2019	ba,pt	%xcc, done2			;\
2020	  ldd	[ADDR + CPU_TMP1], %f40		;\
2021	ba,pt	%xcc, done2			;\
2022	  ldd	[ADDR + CPU_TMP1], %f10		;\
2023	ba,pt	%xcc, done2			;\
2024	  ldd	[ADDR + CPU_TMP1], %f42		;\
2025	ba,pt	%xcc, done2			;\
2026	  ldd	[ADDR + CPU_TMP1], %f12		;\
2027	ba,pt	%xcc, done2			;\
2028	  ldd	[ADDR + CPU_TMP1], %f44		;\
2029	ba,pt	%xcc, done2			;\
2030	  ldd	[ADDR + CPU_TMP1], %f14		;\
2031	ba,pt	%xcc, done2			;\
2032	  ldd	[ADDR + CPU_TMP1], %f46		;\
2033	ba,pt	%xcc, done2			;\
2034	  ldd	[ADDR + CPU_TMP1], %f16		;\
2035	ba,pt	%xcc, done2			;\
2036	  ldd	[ADDR + CPU_TMP1], %f48		;\
2037	ba,pt	%xcc, done2			;\
2038	  ldd	[ADDR + CPU_TMP1], %f18		;\
2039	ba,pt	%xcc, done2			;\
2040	  ldd	[ADDR + CPU_TMP1], %f50		;\
2041	ba,pt	%xcc, done2			;\
2042	  ldd	[ADDR + CPU_TMP1], %f20		;\
2043	ba,pt	%xcc, done2			;\
2044	  ldd	[ADDR + CPU_TMP1], %f52		;\
2045	ba,pt	%xcc, done2			;\
2046	  ldd	[ADDR + CPU_TMP1], %f22		;\
2047	ba,pt	%xcc, done2			;\
2048	  ldd	[ADDR + CPU_TMP1], %f54		;\
2049	ba,pt	%xcc, done2			;\
2050	  ldd	[ADDR + CPU_TMP1], %f24		;\
2051	ba,pt	%xcc, done2			;\
2052	  ldd	[ADDR + CPU_TMP1], %f56		;\
2053	ba,pt	%xcc, done2			;\
2054	  ldd	[ADDR + CPU_TMP1], %f26		;\
2055	ba,pt	%xcc, done2			;\
2056	  ldd	[ADDR + CPU_TMP1], %f58		;\
2057	ba,pt	%xcc, done2			;\
2058	  ldd	[ADDR + CPU_TMP1], %f28		;\
2059	ba,pt	%xcc, done2			;\
2060	  ldd	[ADDR + CPU_TMP1], %f60		;\
2061	ba,pt	%xcc, done2			;\
2062	  ldd	[ADDR + CPU_TMP1], %f30		;\
2063	ba,pt	%xcc, done2			;\
2064	  ldd	[ADDR + CPU_TMP1], %f62		;\
2065done2:
2066
2067.lddf_exception_not_aligned:
2068	/* %g2 = sfar, %g3 = sfsr */
2069	mov	%g2, %g5		! stash sfar
2070#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2071	sethi	%hi(fpu_exists), %g2	! check fpu_exists
2072	ld	[%g2 + %lo(fpu_exists)], %g2
2073	brz,a,pn %g2, 4f
2074	  nop
2075#endif
2076	CPU_ADDR(%g1, %g4)
2077	or	%g0, 1, %g4
2078	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2079
2080	rdpr	%tpc, %g2
2081	lda	[%g2]ASI_AIUP, %g6	! get the user's lddf instruction
2082	srl	%g6, 23, %g1		! using ldda or not?
2083	and	%g1, 1, %g1
2084	brz,a,pt %g1, 2f		! check for ldda instruction
2085	  nop
2086	srl	%g6, 13, %g1		! check immflag
2087	and	%g1, 1, %g1
2088	rdpr	%tstate, %g2		! %tstate in %g2
2089	brnz,a,pn %g1, 1f
2090	  srl	%g2, 31, %g1		! get asi from %tstate
2091	srl	%g6, 5, %g1		! get asi from instruction
2092	and	%g1, 0xFF, %g1		! imm_asi field
20931:
2094	cmp	%g1, ASI_P		! primary address space
2095	be,a,pt %icc, 2f
2096	  nop
2097	cmp	%g1, ASI_PNF		! primary no fault address space
2098	be,a,pt %icc, 2f
2099	  nop
2100	cmp	%g1, ASI_S		! secondary address space
2101	be,a,pt %icc, 2f
2102	  nop
2103	cmp	%g1, ASI_SNF		! secondary no fault address space
2104	bne,a,pn %icc, 3f
2105	  nop
21062:
2107	lduwa	[%g5]ASI_USER, %g7	! get first half of misaligned data
2108	add	%g5, 4, %g5		! increment misaligned data address
2109	lduwa	[%g5]ASI_USER, %g5	! get second half of misaligned data
2110
2111	sllx	%g7, 32, %g7
2112	or	%g5, %g7, %g5		! combine data
2113	CPU_ADDR(%g7, %g1)		! save data on a per-cpu basis
2114	stx	%g5, [%g7 + CPU_TMP1]	! save in cpu_tmp1
2115
2116	srl	%g6, 25, %g3		! %g6 has the instruction
2117	and	%g3, 0x1F, %g3		! %g3 has rd
2118	LDDF_REG(%g3, %g7, %g4)
2119
2120	CPU_ADDR(%g1, %g4)
2121	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2122	FAST_TRAP_DONE
21233:
2124	CPU_ADDR(%g1, %g4)
2125	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
21264:
2127	set	T_USER, %g3		! trap type in %g3
2128	or	%g3, T_LDDF_ALIGN, %g3
2129	mov	%g5, %g2		! misaligned vaddr in %g2
2130	set	fpu_trap, %g1		! goto C for the little and
2131	ba,pt	%xcc, sys_trap		! no fault little asi's
2132	  sub	%g0, 1, %g4
2133
2134.stdf_exception_not_aligned:
2135	/* %g2 = sfar, %g3 = sfsr */
2136	mov	%g2, %g5
2137
2138#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2139	sethi	%hi(fpu_exists), %g7		! check fpu_exists
2140	ld	[%g7 + %lo(fpu_exists)], %g3
2141	brz,a,pn %g3, 4f
2142	  nop
2143#endif
2144	CPU_ADDR(%g1, %g4)
2145	or	%g0, 1, %g4
2146	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2147
2148	rdpr	%tpc, %g2
2149	lda	[%g2]ASI_AIUP, %g6	! get the user's stdf instruction
2150
2151	srl	%g6, 23, %g1		! using stda or not?
2152	and	%g1, 1, %g1
2153	brz,a,pt %g1, 2f		! check for stda instruction
2154	  nop
2155	srl	%g6, 13, %g1		! check immflag
2156	and	%g1, 1, %g1
2157	rdpr	%tstate, %g2		! %tstate in %g2
2158	brnz,a,pn %g1, 1f
2159	  srl	%g2, 31, %g1		! get asi from %tstate
2160	srl	%g6, 5, %g1		! get asi from instruction
2161	and	%g1, 0xff, %g1		! imm_asi field
21621:
2163	cmp	%g1, ASI_P		! primary address space
2164	be,a,pt %icc, 2f
2165	  nop
2166	cmp	%g1, ASI_S		! secondary address space
2167	bne,a,pn %icc, 3f
2168	  nop
21692:
2170	srl	%g6, 25, %g6
2171	and	%g6, 0x1F, %g6		! %g6 has rd
2172	CPU_ADDR(%g7, %g1)
2173	STDF_REG(%g6, %g7, %g4)		! STDF_REG(REG, ADDR, TMP)
2174
2175	ldx	[%g7 + CPU_TMP1], %g6
2176	srlx	%g6, 32, %g7
2177	stuwa	%g7, [%g5]ASI_USER	! first half
2178	add	%g5, 4, %g5		! increment misaligned data address
2179	stuwa	%g6, [%g5]ASI_USER	! second half
2180
2181	CPU_ADDR(%g1, %g4)
2182	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2183	FAST_TRAP_DONE
21843:
2185	CPU_ADDR(%g1, %g4)
2186	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
21874:
2188	set	T_USER, %g3		! trap type in %g3
2189	or	%g3, T_STDF_ALIGN, %g3
2190	mov	%g5, %g2		! misaligned vaddr in %g2
2191	set	fpu_trap, %g1		! goto C for the little and
2192	ba,pt	%xcc, sys_trap		! nofault little asi's
2193	  sub	%g0, 1, %g4
2194
2195#ifdef DEBUG_USER_TRAPTRACECTL
2196
2197.traptrace_freeze:
2198	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2199	TT_TRACE_L(trace_win)
2200	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2201	set	trap_freeze, %g1
2202	mov	1, %g2
2203	st	%g2, [%g1]
2204	FAST_TRAP_DONE
2205
2206.traptrace_unfreeze:
2207	set	trap_freeze, %g1
2208	st	%g0, [%g1]
2209	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2210	TT_TRACE_L(trace_win)
2211	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2212	FAST_TRAP_DONE
2213
2214#endif /* DEBUG_USER_TRAPTRACECTL */
2215
2216.getcc:
2217	CPU_ADDR(%g1, %g2)
2218	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2219	rdpr	%tstate, %g3			! get tstate
2220	srlx	%g3, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2221	set	PSR_ICC, %g2
2222	and	%o0, %g2, %o0			! mask out the rest
2223	srl	%o0, PSR_ICC_SHIFT, %o0		! right justify
2224	wrpr	%g0, 0, %gl
2225	mov	%o0, %g1			! move ccr to normal %g1
2226	wrpr	%g0, 1, %gl
2227	! cannot assume globals retained their values after increasing %gl
2228	CPU_ADDR(%g1, %g2)
2229	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2230	FAST_TRAP_DONE
2231
2232.setcc:
2233	CPU_ADDR(%g1, %g2)
2234	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2235	wrpr	%g0, 0, %gl
2236	mov	%g1, %o0
2237	wrpr	%g0, 1, %gl
2238	! cannot assume globals retained their values after increasing %gl
2239	CPU_ADDR(%g1, %g2)
2240	sll	%o0, PSR_ICC_SHIFT, %g2
2241	set	PSR_ICC, %g3
2242	and	%g2, %g3, %g2			! mask out rest
2243	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g2
2244	rdpr	%tstate, %g3			! get tstate
2245	srl	%g3, 0, %g3			! clear upper word
2246	or	%g3, %g2, %g3			! or in new bits
2247	wrpr	%g3, %tstate
2248	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2249	FAST_TRAP_DONE
2250
2251/*
2252 * getpsr(void)
2253 * Note that the xcc part of the ccr is not provided.
2254 * The V8 code shows why the V9 trap is not faster:
2255 * #define GETPSR_TRAP() \
2256 *      mov %psr, %i0; jmp %l2; rett %l2+4; nop;
2257 */
2258
2259	.type	.getpsr, #function
2260.getpsr:
2261	rdpr	%tstate, %g1			! get tstate
2262	srlx	%g1, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2263	set	PSR_ICC, %g2
2264	and	%o0, %g2, %o0			! mask out the rest
2265
2266	rd	%fprs, %g1			! get fprs
2267	and	%g1, FPRS_FEF, %g2		! mask out dirty upper/lower
2268	sllx	%g2, PSR_FPRS_FEF_SHIFT, %g2	! shift fef to V8 psr.ef
2269	or	%o0, %g2, %o0			! or result into psr.ef
2270
2271	set	V9_PSR_IMPLVER, %g2		! SI assigned impl/ver: 0xef
2272	or	%o0, %g2, %o0			! or psr.impl/ver
2273	FAST_TRAP_DONE
2274	SET_SIZE(.getpsr)
2275
2276/*
2277 * setpsr(newpsr)
2278 * Note that there is no support for ccr.xcc in the V9 code.
2279 */
2280
2281	.type	.setpsr, #function
2282.setpsr:
2283	rdpr	%tstate, %g1			! get tstate
2284!	setx	TSTATE_V8_UBITS, %g2
2285	or 	%g0, CCR_ICC, %g3
2286	sllx	%g3, TSTATE_CCR_SHIFT, %g2
2287
2288	andn	%g1, %g2, %g1			! zero current user bits
2289	set	PSR_ICC, %g2
2290	and	%g2, %o0, %g2			! clear all but psr.icc bits
2291	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g3	! shift to tstate.ccr.icc
2292	wrpr	%g1, %g3, %tstate		! write tstate
2293
2294	set	PSR_EF, %g2
2295	and	%g2, %o0, %g2			! clear all but fp enable bit
2296	srlx	%g2, PSR_FPRS_FEF_SHIFT, %g4	! shift ef to V9 fprs.fef
2297	wr	%g0, %g4, %fprs			! write fprs
2298
2299	CPU_ADDR(%g1, %g2)			! load CPU struct addr to %g1
2300	ldn	[%g1 + CPU_THREAD], %g2		! load thread pointer
2301	ldn	[%g2 + T_LWP], %g3		! load klwp pointer
2302	ldn	[%g3 + LWP_FPU], %g2		! get lwp_fpu pointer
2303	stuw	%g4, [%g2 + FPU_FPRS]		! write fef value to fpu_fprs
2304	srlx	%g4, 2, %g4			! shift fef value to bit 0
2305	stub	%g4, [%g2 + FPU_EN]		! write fef value to fpu_en
2306	FAST_TRAP_DONE
2307	SET_SIZE(.setpsr)
2308
2309/*
2310 * getlgrp
2311 * get home lgrpid on which the calling thread is currently executing.
2312 */
2313	.type	.getlgrp, #function
2314.getlgrp:
2315	! Thanks for the incredibly helpful comments
2316	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2317	ld	[%g1 + CPU_ID], %o0	! load cpu_id
2318	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2319	ldn	[%g2 + T_LPL], %g2	! load lpl pointer
2320	ld	[%g2 + LPL_LGRPID], %g1	! load lpl_lgrpid
2321	sra	%g1, 0, %o1
2322	FAST_TRAP_DONE
2323	SET_SIZE(.getlgrp)
2324
2325/*
2326 * Entry for old 4.x trap (trap 0).
2327 */
2328	ENTRY_NP(syscall_trap_4x)
2329	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2330	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2331	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2332	ld	[%g2 + PCB_TRAP0], %g2	! lwp->lwp_pcb.pcb_trap0addr
2333	brz,pn	%g2, 1f			! has it been set?
2334	st	%l0, [%g1 + CPU_TMP1]	! delay - save some locals
2335	st	%l1, [%g1 + CPU_TMP2]
2336	rdpr	%tnpc, %l1		! save old tnpc
2337	wrpr	%g0, %g2, %tnpc		! setup tnpc
2338
2339	mov	%g1, %l0		! save CPU struct addr
2340	wrpr	%g0, 0, %gl
2341	mov	%l1, %g6		! pass tnpc to user code in %g6
2342	wrpr	%g0, 1, %gl
2343	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2344	ld	[%l0 + CPU_TMP1], %l0
2345	FAST_TRAP_DONE_CHK_INTR
23461:
2347	!
2348	! check for old syscall mmap which is the only different one which
2349	! must be the same.  Others are handled in the compatibility library.
2350	!
2351	mov	%g1, %l0		! save CPU struct addr
2352	wrpr	%g0, 0, %gl
2353	cmp	%g1, OSYS_mmap		! compare to old 4.x mmap
2354	movz	%icc, SYS_mmap, %g1
2355	wrpr	%g0, 1, %gl
2356	ld	[%l0 + CPU_TMP1], %l0
2357	SYSCALL(syscall_trap32)
2358	SET_SIZE(syscall_trap_4x)
2359
2360/*
2361 * Handler for software trap 9.
2362 * Set trap0 emulation address for old 4.x system call trap.
2363 * XXX - this should be a system call.
2364 */
2365	ENTRY_NP(set_trap0_addr)
2366	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2367	st	%l0, [%g1 + CPU_TMP1]	! save some locals
2368	st	%l1, [%g1 + CPU_TMP2]
2369	mov	%g1, %l0	! preserve CPU addr
2370	wrpr	%g0, 0, %gl
2371	mov	%g1, %l1
2372	wrpr	%g0, 1, %gl
2373	! cannot assume globals retained their values after increasing %gl
2374	ldn	[%l0 + CPU_THREAD], %g2	! load thread pointer
2375	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2376	andn	%l1, 3, %l1		! force alignment
2377	st	%l1, [%g2 + PCB_TRAP0]	! lwp->lwp_pcb.pcb_trap0addr
2378	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2379	ld	[%l0 + CPU_TMP1], %l0
2380	FAST_TRAP_DONE
2381	SET_SIZE(set_trap0_addr)
2382
2383/*
2384 * mmu_trap_tl1
2385 * trap handler for unexpected mmu traps.
2386 * simply checks if the trap was a user lddf/stdf alignment trap, in which
2387 * case we go to fpu_trap or a user trap from the window handler, in which
2388 * case we go save the state on the pcb.  Otherwise, we go to ptl1_panic.
2389 */
2390	.type	mmu_trap_tl1, #function
2391mmu_trap_tl1:
2392#ifdef	TRAPTRACE
2393	TRACE_PTR(%g5, %g6)
2394	GET_TRACE_TICK(%g6)
2395	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2396	TRACE_SAVE_TL_GL_REGS(%g5, %g6)
2397	rdpr	%tt, %g6
2398	stha	%g6, [%g5 + TRAP_ENT_TT]%asi
2399	rdpr	%tstate, %g6
2400	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi
2401	stna	%sp, [%g5 + TRAP_ENT_SP]%asi
2402	stna	%g0, [%g5 + TRAP_ENT_TR]%asi
2403	rdpr	%tpc, %g6
2404	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2405	MMU_FAULT_STATUS_AREA(%g6)
2406	ldx	[%g6 + MMFSA_D_ADDR], %g6
2407	stna	%g6, [%g5 + TRAP_ENT_F1]%asi !  MMU fault address
2408	CPU_PADDR(%g7, %g6);
2409	add	%g7, CPU_TL1_HDLR, %g7
2410	lda	[%g7]ASI_MEM, %g6
2411	stna	%g6, [%g5 + TRAP_ENT_F2]%asi
2412	MMU_FAULT_STATUS_AREA(%g6)
2413	ldx	[%g6 + MMFSA_D_TYPE], %g7 ! XXXQ should be a MMFSA_F_ constant?
2414	ldx	[%g6 + MMFSA_D_CTX], %g6
2415	sllx	%g6, SFSR_CTX_SHIFT, %g6
2416	or	%g6, %g7, %g6
2417	stna	%g6, [%g5 + TRAP_ENT_F3]%asi ! MMU context/type
2418	set	0xdeadbeef, %g6
2419	stna	%g6, [%g5 + TRAP_ENT_F4]%asi
2420	TRACE_NEXT(%g5, %g6, %g7)
2421#endif /* TRAPTRACE */
2422	CPU_PADDR(%g7, %g6);
2423	add     %g7, CPU_TL1_HDLR, %g7		! %g7 = &cpu_m.tl1_hdlr (PA)
2424	lda	[%g7]ASI_MEM, %g6
2425	brz,a,pt %g6, 1f
2426	  nop
2427	sta     %g0, [%g7]ASI_MEM
2428	! XXXQ need to setup registers for sfmmu_mmu_trap?
2429	ba,a,pt	%xcc, sfmmu_mmu_trap		! handle page faults
24301:
2431	rdpr	%tpc, %g7
2432	/* in user_rtt? */
2433	set	rtt_fill_start, %g6
2434	cmp	%g7, %g6
2435	blu,pn	%xcc, 6f
2436	 .empty
2437	set	rtt_fill_end, %g6
2438	cmp	%g7, %g6
2439	bgeu,pn %xcc, 6f
2440	 nop
2441	set	fault_rtt_fn1, %g7
2442	ba,a	7f
24436:
2444	! check to see if the trap pc is in a window spill/fill handling
2445	rdpr	%tpc, %g7
2446	/* tpc should be in the trap table */
2447	set	trap_table, %g6
2448	cmp	%g7, %g6
2449	blu,a,pn %xcc, ptl1_panic
2450	  mov	PTL1_BAD_MMUTRAP, %g1
2451	set	etrap_table, %g6
2452	cmp	%g7, %g6
2453	bgeu,a,pn %xcc, ptl1_panic
2454	  mov	PTL1_BAD_MMUTRAP, %g1
2455	! pc is inside the trap table, convert to trap type
2456	srl	%g7, 5, %g6		! XXXQ need #define
2457	and	%g6, 0x1ff, %g6		! XXXQ need #define
2458	! and check for a window trap type
2459	and	%g6, WTRAP_TTMASK, %g6
2460	cmp	%g6, WTRAP_TYPE
2461	bne,a,pn %xcc, ptl1_panic
2462	  mov	PTL1_BAD_MMUTRAP, %g1
2463	andn	%g7, WTRAP_ALIGN, %g7	/* 128 byte aligned */
2464	add	%g7, WTRAP_FAULTOFF, %g7
2465
24667:
2467	! Arguments are passed in the global set active after the
2468	! 'done' instruction. Before switching sets, must save
2469	! the calculated next pc
2470	wrpr	%g0, %g7, %tnpc
2471	wrpr	%g0, 1, %gl
2472	rdpr	%tt, %g5
2473	MMU_FAULT_STATUS_AREA(%g7)
2474	cmp	%g5, T_ALIGNMENT
2475	be,pn	%xcc, 1f
2476	ldx	[%g7 + MMFSA_D_ADDR], %g6
2477	ldx	[%g7 + MMFSA_D_CTX], %g7
2478	srlx	%g6, MMU_PAGESHIFT, %g6		/* align address */
2479	cmp	%g7, USER_CONTEXT_TYPE
2480	sllx	%g6, MMU_PAGESHIFT, %g6
2481	movgu	%icc, USER_CONTEXT_TYPE, %g7
2482	or	%g6, %g7, %g6			/* TAG_ACCESS */
24831:
2484	done
2485	SET_SIZE(mmu_trap_tl1)
2486
2487/*
2488 * Several traps use kmdb_trap and kmdb_trap_tl1 as their handlers.  These
2489 * traps are valid only when kmdb is loaded.  When the debugger is active,
2490 * the code below is rewritten to transfer control to the appropriate
2491 * debugger entry points.
2492 */
2493	.global	kmdb_trap
2494	.align	8
2495kmdb_trap:
2496	ba,a	trap_table0
2497	jmp	%g1 + 0
2498	nop
2499
2500	.global	kmdb_trap_tl1
2501	.align	8
2502kmdb_trap_tl1:
2503	ba,a	trap_table0
2504	jmp	%g1 + 0
2505	nop
2506
2507/*
2508 * This entry is copied from OBP's trap table during boot.
2509 */
2510	.global	obp_bpt
2511	.align	8
2512obp_bpt:
2513	NOT
2514
2515
2516
2517#ifdef	TRAPTRACE
2518/*
2519 * TRAPTRACE support.
2520 * labels here are branched to with "rd %pc, %g7" in the delay slot.
2521 * Return is done by "jmp %g7 + 4".
2522 */
2523
2524trace_dmmu:
2525	TRACE_PTR(%g3, %g6)
2526	GET_TRACE_TICK(%g6)
2527	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2528	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2529	rdpr	%tt, %g6
2530	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2531	rdpr	%tstate, %g6
2532	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2533	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2534	rdpr	%tpc, %g6
2535	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2536	MMU_FAULT_STATUS_AREA(%g6)
2537	ldx	[%g6 + MMFSA_D_ADDR], %g4
2538	stxa	%g4, [%g3 + TRAP_ENT_TR]%asi
2539	ldx	[%g6 + MMFSA_D_CTX], %g4
2540	stxa	%g4, [%g3 + TRAP_ENT_F1]%asi
2541	ldx	[%g6 + MMFSA_D_TYPE], %g4
2542	stxa	%g4, [%g3 + TRAP_ENT_F2]%asi
2543	stxa	%g6, [%g3 + TRAP_ENT_F3]%asi
2544	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2545	TRACE_NEXT(%g3, %g4, %g5)
2546	jmp	%g7 + 4
2547	nop
2548
2549trace_immu:
2550	TRACE_PTR(%g3, %g6)
2551	GET_TRACE_TICK(%g6)
2552	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2553	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2554	rdpr	%tt, %g6
2555	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2556	rdpr	%tstate, %g6
2557	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2558	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2559	rdpr	%tpc, %g6
2560	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2561	MMU_FAULT_STATUS_AREA(%g6)
2562	ldx	[%g6 + MMFSA_I_ADDR], %g4
2563	stxa	%g4, [%g3 + TRAP_ENT_TR]%asi
2564	ldx	[%g6 + MMFSA_I_CTX], %g4
2565	stxa	%g4, [%g3 + TRAP_ENT_F1]%asi
2566	ldx	[%g6 + MMFSA_I_TYPE], %g4
2567	stxa	%g4, [%g3 + TRAP_ENT_F2]%asi
2568	stxa	%g6, [%g3 + TRAP_ENT_F3]%asi
2569	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2570	TRACE_NEXT(%g3, %g4, %g5)
2571	jmp	%g7 + 4
2572	nop
2573
2574trace_gen:
2575	TRACE_PTR(%g3, %g6)
2576	GET_TRACE_TICK(%g6)
2577	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2578	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2579	rdpr	%tt, %g6
2580	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2581	rdpr	%tstate, %g6
2582	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2583	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2584	rdpr	%tpc, %g6
2585	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2586	stna	%g0, [%g3 + TRAP_ENT_TR]%asi
2587	stna	%g0, [%g3 + TRAP_ENT_F1]%asi
2588	stna	%g0, [%g3 + TRAP_ENT_F2]%asi
2589	stna	%g0, [%g3 + TRAP_ENT_F3]%asi
2590	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2591	TRACE_NEXT(%g3, %g4, %g5)
2592	jmp	%g7 + 4
2593	nop
2594
2595trace_win:
2596	TRACE_WIN_INFO(0, %l0, %l1, %l2)
2597	! Keep the locals as clean as possible, caller cleans %l4
2598	clr	%l2
2599	clr	%l1
2600	jmp	%l4 + 4
2601	  clr	%l0
2602
2603/*
2604 * Trace a tsb hit
2605 * g1 = tsbe pointer (in/clobbered)
2606 * g2 = tag access register (in)
2607 * g3 - g4 = scratch (clobbered)
2608 * g5 = tsbe data (in)
2609 * g6 = scratch (clobbered)
2610 * g7 = pc we jumped here from (in)
2611 */
2612
2613	! Do not disturb %g5, it will be used after the trace
2614	ALTENTRY(trace_tsbhit)
2615	TRACE_TSBHIT(0)
2616	jmp	%g7 + 4
2617	nop
2618
2619/*
2620 * Trace a TSB miss
2621 *
2622 * g1 = tsb8k pointer (in)
2623 * g2 = tag access register (in)
2624 * g3 = tsb4m pointer (in)
2625 * g4 = tsbe tag (in/clobbered)
2626 * g5 - g6 = scratch (clobbered)
2627 * g7 = pc we jumped here from (in)
2628 */
2629	.global	trace_tsbmiss
2630trace_tsbmiss:
2631	membar	#Sync
2632	sethi	%hi(FLUSH_ADDR), %g6
2633	flush	%g6
2634	TRACE_PTR(%g5, %g6)
2635	GET_TRACE_TICK(%g6)
2636	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2637	stna	%g2, [%g5 + TRAP_ENT_SP]%asi		! tag access
2638	stna	%g4, [%g5 + TRAP_ENT_F1]%asi		! XXX? tsb tag
2639	rdpr	%tnpc, %g6
2640	stna	%g6, [%g5 + TRAP_ENT_F2]%asi
2641	stna	%g1, [%g5 + TRAP_ENT_F3]%asi		! tsb8k pointer
2642	rdpr	%tpc, %g6
2643	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2644	TRACE_SAVE_TL_GL_REGS(%g5, %g6)
2645	rdpr	%tt, %g6
2646	or	%g6, TT_MMU_MISS, %g4
2647	stha	%g4, [%g5 + TRAP_ENT_TT]%asi
2648	mov	MMFSA_D_ADDR, %g4
2649	cmp	%g6, FAST_IMMU_MISS_TT
2650	move	%xcc, MMFSA_I_ADDR, %g4
2651	cmp	%g6, T_INSTR_MMU_MISS
2652	move	%xcc, MMFSA_I_ADDR, %g4
2653	MMU_FAULT_STATUS_AREA(%g6)
2654	ldx	[%g6 + %g4], %g6
2655	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi	! tag target
2656	cmp	%g4, MMFSA_D_ADDR
2657	move	%xcc, MMFSA_D_CTX, %g4
2658	movne	%xcc, MMFSA_I_CTX, %g4
2659	MMU_FAULT_STATUS_AREA(%g6)
2660	ldx	[%g6 + %g4], %g6
2661	stxa	%g6, [%g5 + TRAP_ENT_F4]%asi		! context ID
2662	stna	%g3, [%g5 + TRAP_ENT_TR]%asi		! tsb4m pointer
2663	TRACE_NEXT(%g5, %g4, %g6)
2664	jmp	%g7 + 4
2665	nop
2666
2667/*
2668 * g2 = tag access register (in)
2669 * g3 = ctx type (0, 1 or 2) (in) (not used)
2670 */
2671trace_dataprot:
2672	membar	#Sync
2673	sethi	%hi(FLUSH_ADDR), %g6
2674	flush	%g6
2675	TRACE_PTR(%g1, %g6)
2676	GET_TRACE_TICK(%g6)
2677	stxa	%g6, [%g1 + TRAP_ENT_TICK]%asi
2678	rdpr	%tpc, %g6
2679	stna	%g6, [%g1 + TRAP_ENT_TPC]%asi
2680	rdpr	%tstate, %g6
2681	stxa	%g6, [%g1 + TRAP_ENT_TSTATE]%asi
2682	stna	%g2, [%g1 + TRAP_ENT_SP]%asi		! tag access reg
2683	stna	%g0, [%g1 + TRAP_ENT_F1]%asi
2684	stna	%g0, [%g1 + TRAP_ENT_F2]%asi
2685	stna	%g0, [%g1 + TRAP_ENT_F3]%asi
2686	stna	%g0, [%g1 + TRAP_ENT_F4]%asi
2687	TRACE_SAVE_TL_GL_REGS(%g1, %g6)
2688	rdpr	%tt, %g6
2689	stha	%g6, [%g1 + TRAP_ENT_TT]%asi
2690	mov	MMFSA_D_CTX, %g4
2691	cmp	%g6, FAST_IMMU_MISS_TT
2692	move	%xcc, MMFSA_I_CTX, %g4
2693	cmp	%g6, T_INSTR_MMU_MISS
2694	move	%xcc, MMFSA_I_CTX, %g4
2695	MMU_FAULT_STATUS_AREA(%g6)
2696	ldx	[%g6 + %g4], %g6
2697	stxa	%g6, [%g1 + TRAP_ENT_TR]%asi	! context ID
2698	TRACE_NEXT(%g1, %g4, %g5)
2699	jmp	%g7 + 4
2700	nop
2701
2702#endif /* TRAPTRACE */
2703
2704/*
2705 * Handle watchdog reset trap. Enable the MMU using the MMU_ENABLE
2706 * HV service, which requires the return target to be specified as a VA
2707 * since we are enabling the MMU. We set the target to ptl1_panic.
2708 */
2709
2710	.type	.watchdog_trap, #function
2711.watchdog_trap:
2712	mov	1, %o0
2713	setx	ptl1_panic, %g2, %o1
2714	mov	MMU_ENABLE, %o5
2715	ta	FAST_TRAP
2716	done
2717	SET_SIZE(.watchdog_trap)
2718/*
2719 * synthesize for trap(): SFAR in %g2, SFSR in %g3
2720 */
2721	.type	.dmmu_exc_lddf_not_aligned, #function
2722.dmmu_exc_lddf_not_aligned:
2723	MMU_FAULT_STATUS_AREA(%g3)
2724	ldx	[%g3 + MMFSA_D_ADDR], %g2
2725	/* Fault type not available in MMU fault status area */
2726	mov	MMFSA_F_UNALIGN, %g1
2727	ldx	[%g3 + MMFSA_D_CTX], %g3
2728	sllx	%g3, SFSR_CTX_SHIFT, %g3
2729	btst	1, %sp
2730	bnz,pt	%xcc, .lddf_exception_not_aligned
2731	or	%g3, %g1, %g3			/* SFSR */
2732	ba,a,pt	%xcc, .mmu_exception_not_aligned
2733	SET_SIZE(.dmmu_exc_lddf_not_aligned)
2734
2735/*
2736 * synthesize for trap(): SFAR in %g2, SFSR in %g3
2737 */
2738	.type	.dmmu_exc_stdf_not_aligned, #function
2739.dmmu_exc_stdf_not_aligned:
2740	MMU_FAULT_STATUS_AREA(%g3)
2741	ldx	[%g3 + MMFSA_D_ADDR], %g2
2742	/* Fault type not available in MMU fault status area */
2743	mov	MMFSA_F_UNALIGN, %g1
2744	ldx	[%g3 + MMFSA_D_CTX], %g3
2745	sllx	%g3, SFSR_CTX_SHIFT, %g3
2746	btst	1, %sp
2747	bnz,pt	%xcc, .stdf_exception_not_aligned
2748	or	%g3, %g1, %g3			/* SFSR */
2749	ba,a,pt	%xcc, .mmu_exception_not_aligned
2750	SET_SIZE(.dmmu_exc_stdf_not_aligned)
2751
2752	.type	.dmmu_exception, #function
2753.dmmu_exception:
2754	MMU_FAULT_STATUS_AREA(%g3)
2755	ldx	[%g3 + MMFSA_D_ADDR], %g2
2756	ldx	[%g3 + MMFSA_D_TYPE], %g1
2757	ldx	[%g3 + MMFSA_D_CTX], %g4
2758	srlx	%g2, MMU_PAGESHIFT, %g2		/* align address */
2759	sllx	%g2, MMU_PAGESHIFT, %g2
2760	sllx	%g4, SFSR_CTX_SHIFT, %g3
2761	or	%g3, %g1, %g3			/* SFSR */
2762	cmp	%g4, USER_CONTEXT_TYPE
2763	movgeu	%icc, USER_CONTEXT_TYPE, %g4
2764	or	%g2, %g4, %g2			/* TAG_ACCESS */
2765	ba,pt	%xcc, .mmu_exception_end
2766	mov	T_DATA_EXCEPTION, %g1
2767	SET_SIZE(.dmmu_exception)
2768
2769/*
2770 * fast_trap_done, fast_trap_done_chk_intr:
2771 *
2772 * Due to the design of UltraSPARC pipeline, pending interrupts are not
2773 * taken immediately after a RETRY or DONE instruction which causes IE to
2774 * go from 0 to 1. Instead, the instruction at %tpc or %tnpc is allowed
2775 * to execute first before taking any interrupts. If that instruction
2776 * results in other traps, and if the corresponding trap handler runs
2777 * entirely at TL=1 with interrupts disabled, then pending interrupts
2778 * won't be taken until after yet another instruction following the %tpc
2779 * or %tnpc.
2780 *
2781 * A malicious user program can use this feature to block out interrupts
2782 * for extended durations, which can result in send_mondo_timeout kernel
2783 * panic.
2784 *
2785 * This problem is addressed by servicing any pending interrupts via
2786 * sys_trap before returning back to the user mode from a fast trap
2787 * handler. The "done" instruction within a fast trap handler, which
2788 * runs entirely at TL=1 with interrupts disabled, is replaced with the
2789 * FAST_TRAP_DONE macro, which branches control to this fast_trap_done
2790 * entry point.
2791 *
2792 * We check for any pending interrupts here and force a sys_trap to
2793 * service those interrupts, if any. To minimize overhead, pending
2794 * interrupts are checked if the %tpc happens to be at 16K boundary,
2795 * which allows a malicious program to execute at most 4K consecutive
2796 * instructions before we service any pending interrupts. If a worst
2797 * case fast trap handler takes about 2 usec, then interrupts will be
2798 * blocked for at most 8 msec, less than a clock tick.
2799 *
2800 * For the cases where we don't know if the %tpc will cross a 16K
2801 * boundary, we can't use the above optimization and always process
2802 * any pending interrupts via fast_frap_done_chk_intr entry point.
2803 *
2804 * Entry Conditions:
2805 * 	%pstate		am:0 priv:1 ie:0
2806 * 			globals are AG (not normal globals)
2807 */
2808
2809	.global	fast_trap_done, fast_trap_done_chk_intr
2810fast_trap_done:
2811	rdpr	%tpc, %g5
2812	sethi	%hi(0xffffc000), %g6	! 1's complement of 0x3fff
2813	andncc	%g5, %g6, %g0		! check lower 14 bits of %tpc
2814	bz,pn	%icc, 1f		! branch if zero (lower 32 bits only)
2815	nop
2816	done
2817
2818fast_trap_done_chk_intr:
28191:	rd	SOFTINT, %g6
2820	brnz,pn	%g6, 2f		! branch if any pending intr
2821	nop
2822	done
2823
28242:
2825	/*
2826	 * We get here if there are any pending interrupts.
2827	 * Adjust %tpc/%tnpc as we'll be resuming via "retry"
2828	 * instruction.
2829	 */
2830	rdpr	%tnpc, %g5
2831	wrpr	%g0, %g5, %tpc
2832	add	%g5, 4, %g5
2833	wrpr	%g0, %g5, %tnpc
2834
2835	/*
2836	 * Force a dummy sys_trap call so that interrupts can be serviced.
2837	 */
2838	set	fast_trap_dummy_call, %g1
2839	ba,pt	%xcc, sys_trap
2840	  mov	-1, %g4
2841
2842fast_trap_dummy_call:
2843	retl
2844	nop
2845
2846/*
2847 * Currently the brand syscall interposition code is not enabled by
2848 * default.  Instead, when a branded zone is first booted the brand
2849 * infrastructure will patch the trap table so that the syscall
2850 * entry points are redirected to syscall_wrapper32 and syscall_wrapper
2851 * for ILP32 and LP64 syscalls respectively.  this is done in
2852 * brand_plat_interposition_enable().  Note that the syscall wrappers
2853 * below do not collect any trap trace data since the syscall hot patch
2854 * points are reached after trap trace data has already been collected.
2855 */
2856#define	BRAND_CALLBACK(callback_id)					    \
2857	CPU_ADDR(%g2, %g1)		/* load CPU struct addr to %g2	*/ ;\
2858	ldn	[%g2 + CPU_THREAD], %g3	/* load thread pointer		*/ ;\
2859	ldn	[%g3 + T_PROCP], %g3	/* get proc pointer		*/ ;\
2860	ldn	[%g3 + P_BRAND], %g3	/* get brand pointer		*/ ;\
2861	brz	%g3, 1f			/* No brand?  No callback. 	*/ ;\
2862	nop 								   ;\
2863	ldn	[%g3 + B_MACHOPS], %g3	/* get machops list		*/ ;\
2864	ldn	[%g3 + (callback_id << 3)], %g3 			   ;\
2865	brz	%g3, 1f							   ;\
2866	/*								    \
2867	 * This isn't pretty.  We want a low-latency way for the callback   \
2868	 * routine to decline to do anything.  We just pass in an address   \
2869	 * the routine can directly jmp back to, pretending that nothing    \
2870	 * has happened.						    \
2871	 * 								    \
2872	 * %g1: return address (where the brand handler jumps back to)	    \
2873	 * %g2: address of CPU structure				    \
2874	 * %g3: address of brand handler (where we will jump to)	    \
2875	 */								    \
2876	mov	%pc, %g1						   ;\
2877	add	%g1, 16, %g1						   ;\
2878	jmp	%g3							   ;\
2879	nop								   ;\
28801:
2881
2882	ENTRY_NP(syscall_wrapper32)
2883	BRAND_CALLBACK(BRAND_CB_SYSCALL32)
2884	SYSCALL_NOTT(syscall_trap32)
2885	SET_SIZE(syscall_wrapper32)
2886
2887	ENTRY_NP(syscall_wrapper)
2888	BRAND_CALLBACK(BRAND_CB_SYSCALL)
2889	SYSCALL_NOTT(syscall_trap)
2890	SET_SIZE(syscall_wrapper)
2891
2892#endif	/* lint */
2893