xref: /titanic_51/usr/src/uts/sun4v/ml/trap_table.s (revision 36fe4a92b52649b0979d6a13212f4cea730d19c7)
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, Version 1.0 only
6 * (the "License").  You may not use this file except in compliance
7 * with the License.
8 *
9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10 * or http://www.opensolaris.org/os/licensing.
11 * See the License for the specific language governing permissions
12 * and limitations under the License.
13 *
14 * When distributing Covered Code, include this CDDL HEADER in each
15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16 * If applicable, add the following below this CDDL HEADER, with the
17 * fields enclosed by brackets "[]" replaced with your own identifying
18 * information: Portions Copyright [yyyy] [name of copyright owner]
19 *
20 * CDDL HEADER END
21 */
22/*
23 * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
24 * Use is subject to license terms.
25 */
26
27#pragma ident	"%Z%%M%	%I%	%E% SMI"
28
29#if !defined(lint)
30#include "assym.h"
31#endif /* !lint */
32#include <sys/asm_linkage.h>
33#include <sys/privregs.h>
34#include <sys/sun4asi.h>
35#include <sys/machasi.h>
36#include <sys/hypervisor_api.h>
37#include <sys/machtrap.h>
38#include <sys/machthread.h>
39#include <sys/pcb.h>
40#include <sys/pte.h>
41#include <sys/mmu.h>
42#include <sys/machpcb.h>
43#include <sys/async.h>
44#include <sys/intreg.h>
45#include <sys/scb.h>
46#include <sys/psr_compat.h>
47#include <sys/syscall.h>
48#include <sys/machparam.h>
49#include <sys/traptrace.h>
50#include <vm/hat_sfmmu.h>
51#include <sys/archsystm.h>
52#include <sys/utrap.h>
53#include <sys/clock.h>
54#include <sys/intr.h>
55#include <sys/fpu/fpu_simulator.h>
56#include <vm/seg_spt.h>
57
58/*
59 * WARNING: If you add a fast trap handler which can be invoked by a
60 * non-privileged user, you may have to use the FAST_TRAP_DONE macro
61 * instead of "done" instruction to return back to the user mode. See
62 * comments for the "fast_trap_done" entry point for more information.
63 *
64 * An alternate FAST_TRAP_DONE_CHK_INTR macro should be used for the
65 * cases where you always want to process any pending interrupts before
66 * returning back to the user mode.
67 */
68#define	FAST_TRAP_DONE		\
69	ba,a	fast_trap_done
70
71#define	FAST_TRAP_DONE_CHK_INTR	\
72	ba,a	fast_trap_done_chk_intr
73
74/*
75 * SPARC V9 Trap Table
76 *
77 * Most of the trap handlers are made from common building
78 * blocks, and some are instantiated multiple times within
79 * the trap table. So, I build a bunch of macros, then
80 * populate the table using only the macros.
81 *
82 * Many macros branch to sys_trap.  Its calling convention is:
83 *	%g1		kernel trap handler
84 *	%g2, %g3	args for above
85 *	%g4		desire %pil
86 */
87
88#ifdef	TRAPTRACE
89
90/*
91 * Tracing macro. Adds two instructions if TRAPTRACE is defined.
92 */
93#define	TT_TRACE(label)		\
94	ba	label		;\
95	rd	%pc, %g7
96#define	TT_TRACE_INS	2
97
98#define	TT_TRACE_L(label)	\
99	ba	label		;\
100	rd	%pc, %l4	;\
101	clr	%l4
102#define	TT_TRACE_L_INS	3
103
104#else
105
106#define	TT_TRACE(label)
107#define	TT_TRACE_INS	0
108
109#define	TT_TRACE_L(label)
110#define	TT_TRACE_L_INS	0
111
112#endif
113
114/*
115 * This macro is used to update per cpu mmu stats in perf critical
116 * paths. It is only enabled in debug kernels or if SFMMU_STAT_GATHER
117 * is defined.
118 */
119#if defined(DEBUG) || defined(SFMMU_STAT_GATHER)
120#define	HAT_PERCPU_DBSTAT(stat)			\
121	mov	stat, %g1			;\
122	ba	stat_mmu			;\
123	rd	%pc, %g7
124#else
125#define	HAT_PERCPU_DBSTAT(stat)
126#endif /* DEBUG || SFMMU_STAT_GATHER */
127
128/*
129 * This first set are funneled to trap() with %tt as the type.
130 * Trap will then either panic or send the user a signal.
131 */
132/*
133 * NOT is used for traps that just shouldn't happen.
134 * It comes in both single and quadruple flavors.
135 */
136#if !defined(lint)
137	.global	trap
138#endif /* !lint */
139#define	NOT			\
140	TT_TRACE(trace_gen)	;\
141	set	trap, %g1	;\
142	rdpr	%tt, %g3	;\
143	ba,pt	%xcc, sys_trap	;\
144	sub	%g0, 1, %g4	;\
145	.align	32
146#define	NOT4	NOT; NOT; NOT; NOT
147
148#define	NOTP				\
149	TT_TRACE(trace_gen)		;\
150	ba,pt	%xcc, ptl1_panic	;\
151	  mov	PTL1_BAD_TRAP, %g1	;\
152	.align	32
153#define	NOTP4	NOTP; NOTP; NOTP; NOTP
154
155/*
156 * RED is for traps that use the red mode handler.
157 * We should never see these either.
158 */
159#define	RED	NOT
160/*
161 * BAD is used for trap vectors we don't have a kernel
162 * handler for.
163 * It also comes in single and quadruple versions.
164 */
165#define	BAD	NOT
166#define	BAD4	NOT4
167
168#define	DONE			\
169	done;			\
170	.align	32
171
172/*
173 * TRAP vectors to the trap() function.
174 * It's main use is for user errors.
175 */
176#if !defined(lint)
177	.global	trap
178#endif /* !lint */
179#define	TRAP(arg)		\
180	TT_TRACE(trace_gen)	;\
181	set	trap, %g1	;\
182	mov	arg, %g3	;\
183	ba,pt	%xcc, sys_trap	;\
184	sub	%g0, 1, %g4	;\
185	.align	32
186
187/*
188 * SYSCALL is used for system calls on both ILP32 and LP64 kernels
189 * depending on the "which" parameter (should be either syscall_trap
190 * or syscall_trap32).
191 */
192#define	SYSCALL(which)			\
193	TT_TRACE(trace_gen)		;\
194	set	(which), %g1		;\
195	ba,pt	%xcc, sys_trap		;\
196	sub	%g0, 1, %g4		;\
197	.align	32
198
199/*
200 * GOTO just jumps to a label.
201 * It's used for things that can be fixed without going thru sys_trap.
202 */
203#define	GOTO(label)		\
204	.global	label		;\
205	ba,a	label		;\
206	.empty			;\
207	.align	32
208
209/*
210 * GOTO_TT just jumps to a label.
211 * correctable ECC error traps at  level 0 and 1 will use this macro.
212 * It's used for things that can be fixed without going thru sys_trap.
213 */
214#define	GOTO_TT(label, ttlabel)		\
215	.global	label		;\
216	TT_TRACE(ttlabel)	;\
217	ba,a	label		;\
218	.empty			;\
219	.align	32
220
221/*
222 * Privileged traps
223 * Takes breakpoint if privileged, calls trap() if not.
224 */
225#define	PRIV(label)			\
226	rdpr	%tstate, %g1		;\
227	btst	TSTATE_PRIV, %g1	;\
228	bnz	label			;\
229	rdpr	%tt, %g3		;\
230	set	trap, %g1		;\
231	ba,pt	%xcc, sys_trap		;\
232	sub	%g0, 1, %g4		;\
233	.align	32
234
235
236/*
237 * DTrace traps.
238 */
239#define	DTRACE_FASTTRAP			\
240	.global dtrace_fasttrap_probe				;\
241	.global dtrace_fasttrap_probe_ptr			;\
242	sethi	%hi(dtrace_fasttrap_probe_ptr), %g4		;\
243	ldn	[%g4 + %lo(dtrace_fasttrap_probe_ptr)], %g4	;\
244	set	dtrace_fasttrap_probe, %g1			;\
245	brnz,pn	%g4, user_trap					;\
246	sub	%g0, 1, %g4					;\
247	FAST_TRAP_DONE						;\
248	.align	32
249
250#define	DTRACE_PID			\
251	.global dtrace_pid_probe				;\
252	set	dtrace_pid_probe, %g1				;\
253	ba,pt	%xcc, user_trap					;\
254	sub	%g0, 1, %g4					;\
255	.align	32
256
257#define	DTRACE_RETURN			\
258	.global dtrace_return_probe				;\
259	set	dtrace_return_probe, %g1			;\
260	ba,pt	%xcc, user_trap					;\
261	sub	%g0, 1, %g4					;\
262	.align	32
263
264/*
265 * REGISTER WINDOW MANAGEMENT MACROS
266 */
267
268/*
269 * various convenient units of padding
270 */
271#define	SKIP(n)	.skip 4*(n)
272
273/*
274 * CLEAN_WINDOW is the simple handler for cleaning a register window.
275 */
276#define	CLEAN_WINDOW						\
277	TT_TRACE_L(trace_win)					;\
278	rdpr %cleanwin, %l0; inc %l0; wrpr %l0, %cleanwin	;\
279	clr %l0; clr %l1; clr %l2; clr %l3			;\
280	clr %l4; clr %l5; clr %l6; clr %l7			;\
281	clr %o0; clr %o1; clr %o2; clr %o3			;\
282	clr %o4; clr %o5; clr %o6; clr %o7			;\
283	retry; .align 128
284
285#if !defined(lint)
286
287/*
288 * If we get an unresolved tlb miss while in a window handler, the fault
289 * handler will resume execution at the last instruction of the window
290 * hander, instead of delivering the fault to the kernel.  Spill handlers
291 * use this to spill windows into the wbuf.
292 *
293 * The mixed handler works by checking %sp, and branching to the correct
294 * handler.  This is done by branching back to label 1: for 32b frames,
295 * or label 2: for 64b frames; which implies the handler order is: 32b,
296 * 64b, mixed.  The 1: and 2: labels are offset into the routines to
297 * allow the branchs' delay slots to contain useful instructions.
298 */
299
300/*
301 * SPILL_32bit spills a 32-bit-wide kernel register window.  It
302 * assumes that the kernel context and the nucleus context are the
303 * same.  The stack pointer is required to be eight-byte aligned even
304 * though this code only needs it to be four-byte aligned.
305 */
306#define	SPILL_32bit(tail)					\
307	srl	%sp, 0, %sp					;\
3081:	st	%l0, [%sp + 0]					;\
309	st	%l1, [%sp + 4]					;\
310	st	%l2, [%sp + 8]					;\
311	st	%l3, [%sp + 12]					;\
312	st	%l4, [%sp + 16]					;\
313	st	%l5, [%sp + 20]					;\
314	st	%l6, [%sp + 24]					;\
315	st	%l7, [%sp + 28]					;\
316	st	%i0, [%sp + 32]					;\
317	st	%i1, [%sp + 36]					;\
318	st	%i2, [%sp + 40]					;\
319	st	%i3, [%sp + 44]					;\
320	st	%i4, [%sp + 48]					;\
321	st	%i5, [%sp + 52]					;\
322	st	%i6, [%sp + 56]					;\
323	st	%i7, [%sp + 60]					;\
324	TT_TRACE_L(trace_win)					;\
325	saved							;\
326	retry							;\
327	SKIP(31-19-TT_TRACE_L_INS)				;\
328	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
329	.empty
330
331/*
332 * SPILL_32bit_asi spills a 32-bit-wide register window into a 32-bit
333 * wide address space via the designated asi.  It is used to spill
334 * non-kernel windows.  The stack pointer is required to be eight-byte
335 * aligned even though this code only needs it to be four-byte
336 * aligned.
337 */
338#define	SPILL_32bit_asi(asi_num, tail)				\
339	srl	%sp, 0, %sp					;\
3401:	sta	%l0, [%sp + %g0]asi_num				;\
341	mov	4, %g1						;\
342	sta	%l1, [%sp + %g1]asi_num				;\
343	mov	8, %g2						;\
344	sta	%l2, [%sp + %g2]asi_num				;\
345	mov	12, %g3						;\
346	sta	%l3, [%sp + %g3]asi_num				;\
347	add	%sp, 16, %g4					;\
348	sta	%l4, [%g4 + %g0]asi_num				;\
349	sta	%l5, [%g4 + %g1]asi_num				;\
350	sta	%l6, [%g4 + %g2]asi_num				;\
351	sta	%l7, [%g4 + %g3]asi_num				;\
352	add	%g4, 16, %g4					;\
353	sta	%i0, [%g4 + %g0]asi_num				;\
354	sta	%i1, [%g4 + %g1]asi_num				;\
355	sta	%i2, [%g4 + %g2]asi_num				;\
356	sta	%i3, [%g4 + %g3]asi_num				;\
357	add	%g4, 16, %g4					;\
358	sta	%i4, [%g4 + %g0]asi_num				;\
359	sta	%i5, [%g4 + %g1]asi_num				;\
360	sta	%i6, [%g4 + %g2]asi_num				;\
361	sta	%i7, [%g4 + %g3]asi_num				;\
362	TT_TRACE_L(trace_win)					;\
363	saved							;\
364	retry							;\
365	SKIP(31-25-TT_TRACE_L_INS)				;\
366	ba,a,pt %xcc, fault_32bit_/**/tail			;\
367	.empty
368
369#define	SPILL_32bit_tt1(asi_num, tail)				\
370	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
371	.empty							;\
372	.align 128
373
374
375/*
376 * FILL_32bit fills a 32-bit-wide kernel register window.  It assumes
377 * that the kernel context and the nucleus context are the same.  The
378 * stack pointer is required to be eight-byte aligned even though this
379 * code only needs it to be four-byte aligned.
380 */
381#define	FILL_32bit(tail)					\
382	srl	%sp, 0, %sp					;\
3831:	TT_TRACE_L(trace_win)					;\
384	ld	[%sp + 0], %l0					;\
385	ld	[%sp + 4], %l1					;\
386	ld	[%sp + 8], %l2					;\
387	ld	[%sp + 12], %l3					;\
388	ld	[%sp + 16], %l4					;\
389	ld	[%sp + 20], %l5					;\
390	ld	[%sp + 24], %l6					;\
391	ld	[%sp + 28], %l7					;\
392	ld	[%sp + 32], %i0					;\
393	ld	[%sp + 36], %i1					;\
394	ld	[%sp + 40], %i2					;\
395	ld	[%sp + 44], %i3					;\
396	ld	[%sp + 48], %i4					;\
397	ld	[%sp + 52], %i5					;\
398	ld	[%sp + 56], %i6					;\
399	ld	[%sp + 60], %i7					;\
400	restored						;\
401	retry							;\
402	SKIP(31-19-TT_TRACE_L_INS)				;\
403	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
404	.empty
405
406/*
407 * FILL_32bit_asi fills a 32-bit-wide register window from a 32-bit
408 * wide address space via the designated asi.  It is used to fill
409 * non-kernel windows.  The stack pointer is required to be eight-byte
410 * aligned even though this code only needs it to be four-byte
411 * aligned.
412 */
413#define	FILL_32bit_asi(asi_num, tail)				\
414	srl	%sp, 0, %sp					;\
4151:	TT_TRACE_L(trace_win)					;\
416	mov	4, %g1						;\
417	lda	[%sp + %g0]asi_num, %l0				;\
418	mov	8, %g2						;\
419	lda	[%sp + %g1]asi_num, %l1				;\
420	mov	12, %g3						;\
421	lda	[%sp + %g2]asi_num, %l2				;\
422	lda	[%sp + %g3]asi_num, %l3				;\
423	add	%sp, 16, %g4					;\
424	lda	[%g4 + %g0]asi_num, %l4				;\
425	lda	[%g4 + %g1]asi_num, %l5				;\
426	lda	[%g4 + %g2]asi_num, %l6				;\
427	lda	[%g4 + %g3]asi_num, %l7				;\
428	add	%g4, 16, %g4					;\
429	lda	[%g4 + %g0]asi_num, %i0				;\
430	lda	[%g4 + %g1]asi_num, %i1				;\
431	lda	[%g4 + %g2]asi_num, %i2				;\
432	lda	[%g4 + %g3]asi_num, %i3				;\
433	add	%g4, 16, %g4					;\
434	lda	[%g4 + %g0]asi_num, %i4				;\
435	lda	[%g4 + %g1]asi_num, %i5				;\
436	lda	[%g4 + %g2]asi_num, %i6				;\
437	lda	[%g4 + %g3]asi_num, %i7				;\
438	restored						;\
439	retry							;\
440	SKIP(31-25-TT_TRACE_L_INS)				;\
441	ba,a,pt %xcc, fault_32bit_/**/tail			;\
442	.empty
443
444
445/*
446 * SPILL_64bit spills a 64-bit-wide kernel register window.  It
447 * assumes that the kernel context and the nucleus context are the
448 * same.  The stack pointer is required to be eight-byte aligned.
449 */
450#define	SPILL_64bit(tail)					\
4512:	stx	%l0, [%sp + V9BIAS64 + 0]			;\
452	stx	%l1, [%sp + V9BIAS64 + 8]			;\
453	stx	%l2, [%sp + V9BIAS64 + 16]			;\
454	stx	%l3, [%sp + V9BIAS64 + 24]			;\
455	stx	%l4, [%sp + V9BIAS64 + 32]			;\
456	stx	%l5, [%sp + V9BIAS64 + 40]			;\
457	stx	%l6, [%sp + V9BIAS64 + 48]			;\
458	stx	%l7, [%sp + V9BIAS64 + 56]			;\
459	stx	%i0, [%sp + V9BIAS64 + 64]			;\
460	stx	%i1, [%sp + V9BIAS64 + 72]			;\
461	stx	%i2, [%sp + V9BIAS64 + 80]			;\
462	stx	%i3, [%sp + V9BIAS64 + 88]			;\
463	stx	%i4, [%sp + V9BIAS64 + 96]			;\
464	stx	%i5, [%sp + V9BIAS64 + 104]			;\
465	stx	%i6, [%sp + V9BIAS64 + 112]			;\
466	stx	%i7, [%sp + V9BIAS64 + 120]			;\
467	TT_TRACE_L(trace_win)					;\
468	saved							;\
469	retry							;\
470	SKIP(31-18-TT_TRACE_L_INS)				;\
471	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
472	.empty
473
474#define	SPILL_64bit_ktt1(tail)				\
475	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
476	.empty							;\
477	.align 128
478
479#define	SPILL_mixed_ktt1(tail)				\
480	btst	1, %sp						;\
481	bz,a,pt	%xcc, fault_32bit_/**/tail			;\
482	srl	%sp, 0, %sp					;\
483	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
484	.empty							;\
485	.align 128
486
487/*
488 * SPILL_64bit_asi spills a 64-bit-wide register window into a 64-bit
489 * wide address space via the designated asi.  It is used to spill
490 * non-kernel windows.  The stack pointer is required to be eight-byte
491 * aligned.
492 */
493#define	SPILL_64bit_asi(asi_num, tail)				\
494	mov	0 + V9BIAS64, %g1				;\
4952:	stxa	%l0, [%sp + %g1]asi_num				;\
496	mov	8 + V9BIAS64, %g2				;\
497	stxa	%l1, [%sp + %g2]asi_num				;\
498	mov	16 + V9BIAS64, %g3				;\
499	stxa	%l2, [%sp + %g3]asi_num				;\
500	mov	24 + V9BIAS64, %g4				;\
501	stxa	%l3, [%sp + %g4]asi_num				;\
502	add	%sp, 32, %g5					;\
503	stxa	%l4, [%g5 + %g1]asi_num				;\
504	stxa	%l5, [%g5 + %g2]asi_num				;\
505	stxa	%l6, [%g5 + %g3]asi_num				;\
506	stxa	%l7, [%g5 + %g4]asi_num				;\
507	add	%g5, 32, %g5					;\
508	stxa	%i0, [%g5 + %g1]asi_num				;\
509	stxa	%i1, [%g5 + %g2]asi_num				;\
510	stxa	%i2, [%g5 + %g3]asi_num				;\
511	stxa	%i3, [%g5 + %g4]asi_num				;\
512	add	%g5, 32, %g5					;\
513	stxa	%i4, [%g5 + %g1]asi_num				;\
514	stxa	%i5, [%g5 + %g2]asi_num				;\
515	stxa	%i6, [%g5 + %g3]asi_num				;\
516	stxa	%i7, [%g5 + %g4]asi_num				;\
517	TT_TRACE_L(trace_win)					;\
518	saved							;\
519	retry							;\
520	SKIP(31-25-TT_TRACE_L_INS)				;\
521	ba,a,pt %xcc, fault_64bit_/**/tail			;\
522	.empty
523
524#define	SPILL_64bit_tt1(asi_num, tail)				\
525	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
526	.empty							;\
527	.align 128
528
529/*
530 * FILL_64bit fills a 64-bit-wide kernel register window.  It assumes
531 * that the kernel context and the nucleus context are the same.  The
532 * stack pointer is required to be eight-byte aligned.
533 */
534#define	FILL_64bit(tail)					\
5352:	TT_TRACE_L(trace_win)					;\
536	ldx	[%sp + V9BIAS64 + 0], %l0			;\
537	ldx	[%sp + V9BIAS64 + 8], %l1			;\
538	ldx	[%sp + V9BIAS64 + 16], %l2			;\
539	ldx	[%sp + V9BIAS64 + 24], %l3			;\
540	ldx	[%sp + V9BIAS64 + 32], %l4			;\
541	ldx	[%sp + V9BIAS64 + 40], %l5			;\
542	ldx	[%sp + V9BIAS64 + 48], %l6			;\
543	ldx	[%sp + V9BIAS64 + 56], %l7			;\
544	ldx	[%sp + V9BIAS64 + 64], %i0			;\
545	ldx	[%sp + V9BIAS64 + 72], %i1			;\
546	ldx	[%sp + V9BIAS64 + 80], %i2			;\
547	ldx	[%sp + V9BIAS64 + 88], %i3			;\
548	ldx	[%sp + V9BIAS64 + 96], %i4			;\
549	ldx	[%sp + V9BIAS64 + 104], %i5			;\
550	ldx	[%sp + V9BIAS64 + 112], %i6			;\
551	ldx	[%sp + V9BIAS64 + 120], %i7			;\
552	restored						;\
553	retry							;\
554	SKIP(31-18-TT_TRACE_L_INS)				;\
555	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
556	.empty
557
558/*
559 * FILL_64bit_asi fills a 64-bit-wide register window from a 64-bit
560 * wide address space via the designated asi.  It is used to fill
561 * non-kernel windows.  The stack pointer is required to be eight-byte
562 * aligned.
563 */
564#define	FILL_64bit_asi(asi_num, tail)				\
565	mov	V9BIAS64 + 0, %g1				;\
5662:	TT_TRACE_L(trace_win)					;\
567	ldxa	[%sp + %g1]asi_num, %l0				;\
568	mov	V9BIAS64 + 8, %g2				;\
569	ldxa	[%sp + %g2]asi_num, %l1				;\
570	mov	V9BIAS64 + 16, %g3				;\
571	ldxa	[%sp + %g3]asi_num, %l2				;\
572	mov	V9BIAS64 + 24, %g4				;\
573	ldxa	[%sp + %g4]asi_num, %l3				;\
574	add	%sp, 32, %g5					;\
575	ldxa	[%g5 + %g1]asi_num, %l4				;\
576	ldxa	[%g5 + %g2]asi_num, %l5				;\
577	ldxa	[%g5 + %g3]asi_num, %l6				;\
578	ldxa	[%g5 + %g4]asi_num, %l7				;\
579	add	%g5, 32, %g5					;\
580	ldxa	[%g5 + %g1]asi_num, %i0				;\
581	ldxa	[%g5 + %g2]asi_num, %i1				;\
582	ldxa	[%g5 + %g3]asi_num, %i2				;\
583	ldxa	[%g5 + %g4]asi_num, %i3				;\
584	add	%g5, 32, %g5					;\
585	ldxa	[%g5 + %g1]asi_num, %i4				;\
586	ldxa	[%g5 + %g2]asi_num, %i5				;\
587	ldxa	[%g5 + %g3]asi_num, %i6				;\
588	ldxa	[%g5 + %g4]asi_num, %i7				;\
589	restored						;\
590	retry							;\
591	SKIP(31-25-TT_TRACE_L_INS)				;\
592	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
593	.empty
594
595
596#endif /* !lint */
597
598/*
599 * SPILL_mixed spills either size window, depending on
600 * whether %sp is even or odd, to a 32-bit address space.
601 * This may only be used in conjunction with SPILL_32bit/
602 * FILL_64bit.
603 * Clear upper 32 bits of %sp if it is odd.
604 * We won't need to clear them in 64 bit kernel.
605 */
606#define	SPILL_mixed						\
607	btst	1, %sp						;\
608	bz,a,pt	%xcc, 1b					;\
609	srl	%sp, 0, %sp					;\
610	ba,pt	%xcc, 2b					;\
611	nop							;\
612	.align	128
613
614/*
615 * FILL_mixed(ASI) fills either size window, depending on
616 * whether %sp is even or odd, from a 32-bit address space.
617 * This may only be used in conjunction with FILL_32bit/
618 * FILL_64bit. New versions of FILL_mixed_{tt1,asi} would be
619 * needed for use with FILL_{32,64}bit_{tt1,asi}. Particular
620 * attention should be paid to the instructions that belong
621 * in the delay slots of the branches depending on the type
622 * of fill handler being branched to.
623 * Clear upper 32 bits of %sp if it is odd.
624 * We won't need to clear them in 64 bit kernel.
625 */
626#define	FILL_mixed						\
627	btst	1, %sp						;\
628	bz,a,pt	%xcc, 1b					;\
629	srl	%sp, 0, %sp					;\
630	ba,pt	%xcc, 2b					;\
631	nop							;\
632	.align	128
633
634
635/*
636 * SPILL_32clean/SPILL_64clean spill 32-bit and 64-bit register windows,
637 * respectively, into the address space via the designated asi.  The
638 * unbiased stack pointer is required to be eight-byte aligned (even for
639 * the 32-bit case even though this code does not require such strict
640 * alignment).
641 *
642 * With SPARC v9 the spill trap takes precedence over the cleanwin trap
643 * so when cansave == 0, canrestore == 6, and cleanwin == 6 the next save
644 * will cause cwp + 2 to be spilled but will not clean cwp + 1.  That
645 * window may contain kernel data so in user_rtt we set wstate to call
646 * these spill handlers on the first user spill trap.  These handler then
647 * spill the appropriate window but also back up a window and clean the
648 * window that didn't get a cleanwin trap.
649 */
650#define	SPILL_32clean(asi_num, tail)				\
651	srl	%sp, 0, %sp					;\
652	sta	%l0, [%sp + %g0]asi_num				;\
653	mov	4, %g1						;\
654	sta	%l1, [%sp + %g1]asi_num				;\
655	mov	8, %g2						;\
656	sta	%l2, [%sp + %g2]asi_num				;\
657	mov	12, %g3						;\
658	sta	%l3, [%sp + %g3]asi_num				;\
659	add	%sp, 16, %g4					;\
660	sta	%l4, [%g4 + %g0]asi_num				;\
661	sta	%l5, [%g4 + %g1]asi_num				;\
662	sta	%l6, [%g4 + %g2]asi_num				;\
663	sta	%l7, [%g4 + %g3]asi_num				;\
664	add	%g4, 16, %g4					;\
665	sta	%i0, [%g4 + %g0]asi_num				;\
666	sta	%i1, [%g4 + %g1]asi_num				;\
667	sta	%i2, [%g4 + %g2]asi_num				;\
668	sta	%i3, [%g4 + %g3]asi_num				;\
669	add	%g4, 16, %g4					;\
670	sta	%i4, [%g4 + %g0]asi_num				;\
671	sta	%i5, [%g4 + %g1]asi_num				;\
672	sta	%i6, [%g4 + %g2]asi_num				;\
673	sta	%i7, [%g4 + %g3]asi_num				;\
674	TT_TRACE_L(trace_win)					;\
675	b	.spill_clean					;\
676	  mov	WSTATE_USER32, %g7				;\
677	SKIP(31-25-TT_TRACE_L_INS)				;\
678	ba,a,pt	%xcc, fault_32bit_/**/tail			;\
679	.empty
680
681#define	SPILL_64clean(asi_num, tail)				\
682	mov	0 + V9BIAS64, %g1				;\
683	stxa	%l0, [%sp + %g1]asi_num				;\
684	mov	8 + V9BIAS64, %g2				;\
685	stxa	%l1, [%sp + %g2]asi_num				;\
686	mov	16 + V9BIAS64, %g3				;\
687	stxa	%l2, [%sp + %g3]asi_num				;\
688	mov	24 + V9BIAS64, %g4				;\
689	stxa	%l3, [%sp + %g4]asi_num				;\
690	add	%sp, 32, %g5					;\
691	stxa	%l4, [%g5 + %g1]asi_num				;\
692	stxa	%l5, [%g5 + %g2]asi_num				;\
693	stxa	%l6, [%g5 + %g3]asi_num				;\
694	stxa	%l7, [%g5 + %g4]asi_num				;\
695	add	%g5, 32, %g5					;\
696	stxa	%i0, [%g5 + %g1]asi_num				;\
697	stxa	%i1, [%g5 + %g2]asi_num				;\
698	stxa	%i2, [%g5 + %g3]asi_num				;\
699	stxa	%i3, [%g5 + %g4]asi_num				;\
700	add	%g5, 32, %g5					;\
701	stxa	%i4, [%g5 + %g1]asi_num				;\
702	stxa	%i5, [%g5 + %g2]asi_num				;\
703	stxa	%i6, [%g5 + %g3]asi_num				;\
704	stxa	%i7, [%g5 + %g4]asi_num				;\
705	TT_TRACE_L(trace_win)					;\
706	b	.spill_clean					;\
707	  mov	WSTATE_USER64, %g7				;\
708	SKIP(31-25-TT_TRACE_L_INS)				;\
709	ba,a,pt	%xcc, fault_64bit_/**/tail			;\
710	.empty
711
712
713/*
714 * Floating point disabled.
715 */
716#define	FP_DISABLED_TRAP		\
717	TT_TRACE(trace_gen)		;\
718	ba,pt	%xcc,.fp_disabled	;\
719	nop				;\
720	.align	32
721
722/*
723 * Floating point exceptions.
724 */
725#define	FP_IEEE_TRAP			\
726	TT_TRACE(trace_gen)		;\
727	ba,pt	%xcc,.fp_ieee_exception	;\
728	nop				;\
729	.align	32
730
731#define	FP_TRAP				\
732	TT_TRACE(trace_gen)		;\
733	ba,pt	%xcc,.fp_exception	;\
734	nop				;\
735	.align	32
736
737#if !defined(lint)
738
739/*
740 * ECACHE_ECC error traps at level 0 and level 1
741 */
742#define	ECACHE_ECC(table_name)		\
743	.global	table_name		;\
744table_name:				;\
745	membar	#Sync			;\
746	set	trap, %g1		;\
747	rdpr	%tt, %g3		;\
748	ba,pt	%xcc, sys_trap		;\
749	sub	%g0, 1, %g4		;\
750	.align	32
751
752#endif /* !lint */
753
754/*
755 * illegal instruction trap
756 */
757#define	ILLTRAP_INSTR			  \
758	membar	#Sync			  ;\
759	TT_TRACE(trace_gen)		  ;\
760	or	%g0, P_UTRAP4, %g2	  ;\
761	or	%g0, T_UNIMP_INSTR, %g3   ;\
762	sethi	%hi(.check_v9utrap), %g4  ;\
763	jmp	%g4 + %lo(.check_v9utrap) ;\
764	nop				  ;\
765	.align	32
766
767/*
768 * tag overflow trap
769 */
770#define	TAG_OVERFLOW			  \
771	TT_TRACE(trace_gen)		  ;\
772	or	%g0, P_UTRAP10, %g2	  ;\
773	or	%g0, T_TAG_OVERFLOW, %g3  ;\
774	sethi	%hi(.check_v9utrap), %g4  ;\
775	jmp	%g4 + %lo(.check_v9utrap) ;\
776	nop				  ;\
777	.align	32
778
779/*
780 * divide by zero trap
781 */
782#define	DIV_BY_ZERO			  \
783	TT_TRACE(trace_gen)		  ;\
784	or	%g0, P_UTRAP11, %g2	  ;\
785	or	%g0, T_IDIV0, %g3	  ;\
786	sethi	%hi(.check_v9utrap), %g4  ;\
787	jmp	%g4 + %lo(.check_v9utrap) ;\
788	nop				  ;\
789	.align	32
790
791/*
792 * trap instruction for V9 user trap handlers
793 */
794#define	TRAP_INSTR			  \
795	TT_TRACE(trace_gen)		  ;\
796	or	%g0, T_SOFTWARE_TRAP, %g3 ;\
797	sethi	%hi(.check_v9utrap), %g4  ;\
798	jmp	%g4 + %lo(.check_v9utrap) ;\
799	nop				  ;\
800	.align	32
801#define	TRP4	TRAP_INSTR; TRAP_INSTR; TRAP_INSTR; TRAP_INSTR
802
803/*
804 * LEVEL_INTERRUPT is for level N interrupts.
805 * VECTOR_INTERRUPT is for the vector trap.
806 */
807#define	LEVEL_INTERRUPT(level)		\
808	.global	tt_pil/**/level		;\
809tt_pil/**/level:			;\
810	ba,pt	%xcc, pil_interrupt	;\
811	mov	level, %g4		;\
812	.align	32
813
814#define	LEVEL14_INTERRUPT			\
815	ba	pil14_interrupt			;\
816	mov	PIL_14, %g4			;\
817	.align	32
818
819#define CPU_MONDO			\
820	ba,a,pt	%xcc, cpu_mondo		;\
821	.align	32
822
823#define DEV_MONDO			\
824	ba,a,pt	%xcc, dev_mondo		;\
825	.align	32
826
827/*
828 * MMU Trap Handlers.
829 */
830
831/*
832 * synthesize for trap(): SFSR in %g3
833 */
834#define	IMMU_EXCEPTION							\
835	MMU_FAULT_STATUS_AREA(%g3)					;\
836	rdpr	%tpc, %g2						;\
837	ldx	[%g3 + MMFSA_I_TYPE], %g1				;\
838	ldx	[%g3 + MMFSA_I_CTX], %g3				;\
839	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
840	or	%g3, %g1, %g3						;\
841	ba,pt	%xcc, .mmu_exception_end				;\
842	mov	T_INSTR_EXCEPTION, %g1					;\
843	.align	32
844
845/*
846 * synthesize for trap(): TAG_ACCESS in %g2, SFSR in %g3
847 */
848#define	DMMU_EXCEPTION							\
849	ba,a,pt	%xcc, .dmmu_exception					;\
850	.align	32
851
852/*
853 * synthesize for trap(): SFAR in %g2, SFSR in %g3
854 */
855#define	DMMU_EXC_AG_PRIV						\
856	MMU_FAULT_STATUS_AREA(%g3)					;\
857	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
858	/* Fault type not available in MMU fault status area */		;\
859	mov	MMFSA_F_PRVACT, %g1					;\
860	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
861	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
862	ba,pt	%xcc, .mmu_priv_exception				;\
863	or	%g3, %g1, %g3						;\
864	.align	32
865
866/*
867 * synthesize for trap(): SFAR in %g2, SFSR in %g3
868 */
869#define	DMMU_EXC_AG_NOT_ALIGNED						\
870	MMU_FAULT_STATUS_AREA(%g3)					;\
871	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
872	/* Fault type not available in MMU fault status area */		;\
873	mov	MMFSA_F_UNALIGN, %g1					;\
874	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
875	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
876	ba,pt	%xcc, .mmu_exception_not_aligned			;\
877	or	%g3, %g1, %g3			/* SFSR */		;\
878	.align	32
879/*
880 * SPARC V9 IMPL. DEP. #109(1) and (2) and #110(1) and (2)
881 */
882
883/*
884 * synthesize for trap(): SFAR in %g2, SFSR in %g3
885 */
886#define	DMMU_EXC_LDDF_NOT_ALIGNED					\
887	ba,a,pt	%xcc, .dmmu_exc_lddf_not_aligned			;\
888	.align	32
889/*
890 * synthesize for trap(): SFAR in %g2, SFSR in %g3
891 */
892#define	DMMU_EXC_STDF_NOT_ALIGNED					\
893	ba,a,pt	%xcc, .dmmu_exc_stdf_not_aligned			;\
894	.align	32
895
896#if TAGACC_CTX_MASK != CTXREG_CTX_MASK
897#error "TAGACC_CTX_MASK != CTXREG_CTX_MASK"
898#endif
899
900#if defined(cscope)
901/*
902 * Define labels to direct cscope quickly to labels that
903 * are generated by macro expansion of DTLB_MISS().
904 */
905	.global	tt0_dtlbmiss
906tt0_dtlbmiss:
907	.global	tt1_dtlbmiss
908tt1_dtlbmiss:
909	nop
910#endif
911
912/*
913 * Data miss handler (must be exactly 32 instructions)
914 *
915 * This handler is invoked only if the hypervisor has been instructed
916 * not to do any TSB walk.
917 *
918 * Kernel and invalid context cases are handled by the sfmmu_kdtlb_miss
919 * handler.
920 *
921 * User TLB miss handling depends upon whether a user process has one or
922 * two TSBs. User TSB information (physical base and size code) is kept
923 * in two dedicated scratchpad registers. Absence of a user TSB (primarily
924 * second TSB) is indicated by a negative value (-1) in that register.
925 */
926
927/*
928 * synthesize for miss handler: TAG_ACCESS in %g2
929 */
930#define	DTLB_MISS(table_name)						;\
931	.global	table_name/**/_dtlbmiss					;\
932table_name/**/_dtlbmiss:						;\
933	HAT_PERCPU_DBSTAT(TSBMISS_DTLBMISS) /* 3 instr ifdef DEBUG */	;\
934	MMU_FAULT_STATUS_AREA(%g7)					;\
935	ldx	[%g7 + MMFSA_D_ADDR], %g2	/* address */		;\
936	ldx	[%g7 + MMFSA_D_CTX], %g3	/* g3 = ctx */		;\
937	or	%g2, %g3, %g2			/* TAG_ACCESS */	;\
938	cmp	%g3, INVALID_CONTEXT					;\
939	ble,pn	%xcc, sfmmu_kdtlb_miss					;\
940	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
941	mov	SCRATCHPAD_UTSBREG2, %g1				;\
942	ldxa	[%g1]ASI_SCRATCHPAD, %g1	/* get 2nd tsbreg */	;\
943	brgez,pn %g1, sfmmu_udtlb_slowpath	/* brnach if 2 TSBs */	;\
944	  nop								;\
945	GET_1ST_TSBE_PTR(%g2, %g1, %g4, %g5)	/* 11 instr */		;\
946	ba,pt	%xcc, sfmmu_udtlb_fastpath	/* no 4M TSB, miss */	;\
947	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
948	.align 128
949
950
951#if defined(cscope)
952/*
953 * Define labels to direct cscope quickly to labels that
954 * are generated by macro expansion of ITLB_MISS().
955 */
956	.global	tt0_itlbmiss
957tt0_itlbmiss:
958	.global	tt1_itlbmiss
959tt1_itlbmiss:
960	nop
961#endif
962
963/*
964 * Instruction miss handler.
965 *
966 * This handler is invoked only if the hypervisor has been instructed
967 * not to do any TSB walk.
968 *
969 * ldda instructions will have their ASI patched
970 * by sfmmu_patch_ktsb at runtime.
971 * MUST be EXACTLY 32 instructions or we'll break.
972 */
973
974/*
975 * synthesize for miss handler: TAG_ACCESS in %g2
976 */
977#define	ITLB_MISS(table_name)						 \
978	.global	table_name/**/_itlbmiss					;\
979table_name/**/_itlbmiss:						;\
980	HAT_PERCPU_DBSTAT(TSBMISS_ITLBMISS) /* 3 instr ifdef DEBUG */	;\
981	MMU_FAULT_STATUS_AREA(%g7)					;\
982	ldx	[%g7 + MMFSA_I_ADDR], %g2	/* g2 = address */	;\
983	ldx	[%g7 + MMFSA_I_CTX], %g3	/* g3 = ctx */		;\
984	or	%g2, %g3, %g2			/* TAG_ACCESS */	;\
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
1011 */
1012#define	DTLB_PROT							 \
1013	MMU_FAULT_STATUS_AREA(%g7)					;\
1014	ldx	[%g7 + MMFSA_D_ADDR], %g2	/* address */		;\
1015	ldx	[%g7 + MMFSA_D_CTX], %g3	/* %g3 = ctx */		;\
1016	or	%g2, %g3, %g2			/* TAG_ACCESS */	;\
1017	/*								;\
1018	 *   g2 = tag access register					;\
1019	 *   g3 = ctx number						;\
1020	 */								;\
1021	TT_TRACE(trace_dataprot)	/* 2 instr ifdef TRAPTRACE */	;\
1022					/* clobbers g1 and g6 XXXQ? */	;\
1023	brnz,pt %g3, sfmmu_uprot_trap		/* user trap */		;\
1024	  nop								;\
1025	ba,a,pt	%xcc, sfmmu_kprot_trap		/* kernel trap */	;\
1026	.align 128
1027
1028#define	DMMU_EXCEPTION_TL1						;\
1029	ba,a,pt	%xcc, mmu_trap_tl1					;\
1030	.align 32
1031
1032#define	MISALIGN_ADDR_TL1						;\
1033	ba,a,pt	%xcc, mmu_trap_tl1					;\
1034	.align 32
1035
1036/*
1037 * Trace a tsb hit
1038 * g1 = tsbe pointer (in/clobbered)
1039 * g2 = tag access register (in)
1040 * g3 - g4 = scratch (clobbered)
1041 * g5 = tsbe data (in)
1042 * g6 = scratch (clobbered)
1043 * g7 = pc we jumped here from (in)
1044 * ttextra = value to OR in to trap type (%tt) (in)
1045 */
1046#ifdef TRAPTRACE
1047#define TRACE_TSBHIT(ttextra)						 \
1048	membar	#Sync							;\
1049	sethi	%hi(FLUSH_ADDR), %g6					;\
1050	flush	%g6							;\
1051	TRACE_PTR(%g3, %g6)						;\
1052	GET_TRACE_TICK(%g6)						;\
1053	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi				;\
1054	stna	%g2, [%g3 + TRAP_ENT_SP]%asi	/* tag access */	;\
1055	stna	%g5, [%g3 + TRAP_ENT_F1]%asi	/* tsb data */		;\
1056	rdpr	%tnpc, %g6						;\
1057	stna	%g6, [%g3 + TRAP_ENT_F2]%asi				;\
1058	stna	%g1, [%g3 + TRAP_ENT_F3]%asi	/* tsb pointer */	;\
1059	stna	%g0, [%g3 + TRAP_ENT_F4]%asi				;\
1060	rdpr	%tpc, %g6						;\
1061	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi				;\
1062	TRACE_SAVE_TL_GL_REGS(%g3, %g6)					;\
1063	rdpr	%tt, %g6						;\
1064	or	%g6, (ttextra), %g1					;\
1065	stha	%g1, [%g3 + TRAP_ENT_TT]%asi				;\
1066	MMU_FAULT_STATUS_AREA(%g4)					;\
1067	mov	MMFSA_D_ADDR, %g1					;\
1068	cmp	%g6, FAST_IMMU_MISS_TT					;\
1069	move	%xcc, MMFSA_I_ADDR, %g1					;\
1070	cmp	%g6, T_INSTR_MMU_MISS					;\
1071	move	%xcc, MMFSA_I_ADDR, %g1					;\
1072	ldx	[%g4 + %g1], %g1					;\
1073	stxa	%g1, [%g3 + TRAP_ENT_TSTATE]%asi /* fault addr */	;\
1074	mov	MMFSA_D_CTX, %g1					;\
1075	cmp	%g6, FAST_IMMU_MISS_TT					;\
1076	move	%xcc, MMFSA_I_CTX, %g1					;\
1077	cmp	%g6, T_INSTR_MMU_MISS					;\
1078	move	%xcc, MMFSA_I_CTX, %g1					;\
1079	ldx	[%g4 + %g1], %g1					;\
1080	stna	%g1, [%g3 + TRAP_ENT_TR]%asi				;\
1081	TRACE_NEXT(%g3, %g4, %g6)
1082#else
1083#define TRACE_TSBHIT(ttextra)
1084#endif
1085
1086
1087#if defined(lint)
1088
1089struct scb	trap_table;
1090struct scb	scb;		/* trap_table/scb are the same object */
1091
1092#else /* lint */
1093
1094/*
1095 * =======================================================================
1096 *		SPARC V9 TRAP TABLE
1097 *
1098 * The trap table is divided into two halves: the first half is used when
1099 * taking traps when TL=0; the second half is used when taking traps from
1100 * TL>0. Note that handlers in the second half of the table might not be able
1101 * to make the same assumptions as handlers in the first half of the table.
1102 *
1103 * Worst case trap nesting so far:
1104 *
1105 *	at TL=0 client issues software trap requesting service
1106 *	at TL=1 nucleus wants a register window
1107 *	at TL=2 register window clean/spill/fill takes a TLB miss
1108 *	at TL=3 processing TLB miss
1109 *	at TL=4 handle asynchronous error
1110 *
1111 * Note that a trap from TL=4 to TL=5 places Spitfire in "RED mode".
1112 *
1113 * =======================================================================
1114 */
1115	.section ".text"
1116	.align	4
1117	.global trap_table, scb, trap_table0, trap_table1, etrap_table
1118	.type	trap_table, #function
1119	.type	trap_table0, #function
1120	.type	trap_table1, #function
1121	.type	scb, #function
1122trap_table:
1123scb:
1124trap_table0:
1125	/* hardware traps */
1126	NOT;				/* 000	reserved */
1127	RED;				/* 001	power on reset */
1128	RED;				/* 002	watchdog reset */
1129	RED;				/* 003	externally initiated reset */
1130	RED;				/* 004	software initiated reset */
1131	RED;				/* 005	red mode exception */
1132	NOT; NOT;			/* 006 - 007 reserved */
1133	IMMU_EXCEPTION;			/* 008	instruction access exception */
1134	ITSB_MISS;			/* 009	instruction access MMU miss */
1135 	NOT;				/* 00A  reserved */
1136	NOT; NOT4;			/* 00B - 00F reserved */
1137	ILLTRAP_INSTR;			/* 010	illegal instruction */
1138	TRAP(T_PRIV_INSTR);		/* 011	privileged opcode */
1139	TRAP(T_UNIMP_LDD);		/* 012	unimplemented LDD */
1140	TRAP(T_UNIMP_STD);		/* 013	unimplemented STD */
1141	NOT4; NOT4; NOT4;		/* 014 - 01F reserved */
1142	FP_DISABLED_TRAP;		/* 020	fp disabled */
1143	FP_IEEE_TRAP;			/* 021	fp exception ieee 754 */
1144	FP_TRAP;			/* 022	fp exception other */
1145	TAG_OVERFLOW;			/* 023	tag overflow */
1146	CLEAN_WINDOW;			/* 024 - 027 clean window */
1147	DIV_BY_ZERO;			/* 028	division by zero */
1148	NOT;				/* 029	internal processor error */
1149	NOT; NOT; NOT4;			/* 02A - 02F reserved */
1150	DMMU_EXCEPTION;			/* 030	data access exception */
1151	DTSB_MISS;			/* 031	data access MMU miss */
1152	NOT;				/* 032  reserved */
1153	NOT;				/* 033	data access protection */
1154	DMMU_EXC_AG_NOT_ALIGNED;	/* 034	mem address not aligned */
1155	DMMU_EXC_LDDF_NOT_ALIGNED;	/* 035	LDDF mem address not aligned */
1156	DMMU_EXC_STDF_NOT_ALIGNED;	/* 036	STDF mem address not aligned */
1157	DMMU_EXC_AG_PRIV;		/* 037	privileged action */
1158	NOT;				/* 038	LDQF mem address not aligned */
1159	NOT;				/* 039	STQF mem address not aligned */
1160	NOT; NOT; NOT4;			/* 03A - 03F reserved */
1161	NOT;				/* 040	async data error */
1162	LEVEL_INTERRUPT(1);		/* 041	interrupt level 1 */
1163	LEVEL_INTERRUPT(2);		/* 042	interrupt level 2 */
1164	LEVEL_INTERRUPT(3);		/* 043	interrupt level 3 */
1165	LEVEL_INTERRUPT(4);		/* 044	interrupt level 4 */
1166	LEVEL_INTERRUPT(5);		/* 045	interrupt level 5 */
1167	LEVEL_INTERRUPT(6);		/* 046	interrupt level 6 */
1168	LEVEL_INTERRUPT(7);		/* 047	interrupt level 7 */
1169	LEVEL_INTERRUPT(8);		/* 048	interrupt level 8 */
1170	LEVEL_INTERRUPT(9);		/* 049	interrupt level 9 */
1171	LEVEL_INTERRUPT(10);		/* 04A	interrupt level 10 */
1172	LEVEL_INTERRUPT(11);		/* 04B	interrupt level 11 */
1173	LEVEL_INTERRUPT(12);		/* 04C	interrupt level 12 */
1174	LEVEL_INTERRUPT(13);		/* 04D	interrupt level 13 */
1175	LEVEL14_INTERRUPT;		/* 04E	interrupt level 14 */
1176	LEVEL_INTERRUPT(15);		/* 04F	interrupt level 15 */
1177	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F reserved */
1178	NOT;				/* 060	interrupt vector */
1179	GOTO(kmdb_trap);		/* 061	PA watchpoint */
1180	GOTO(kmdb_trap);		/* 062	VA watchpoint */
1181	NOT;				/* 063	reserved */
1182	ITLB_MISS(tt0);			/* 064	instruction access MMU miss */
1183	DTLB_MISS(tt0);			/* 068	data access MMU miss */
1184	DTLB_PROT;			/* 06C	data access protection */
1185	NOT;				/* 070  reserved */
1186	NOT;				/* 071  reserved */
1187	NOT;				/* 072  reserved */
1188	NOT;				/* 073  reserved */
1189	NOT4; NOT4			/* 074 - 07B reserved */
1190	CPU_MONDO;			/* 07C	cpu_mondo */
1191	DEV_MONDO;			/* 07D	dev_mondo */
1192	GOTO_TT(resumable_error, trace_gen);	/* 07E  resumable error */
1193	GOTO_TT(nonresumable_error, trace_gen);	/* 07F  non-reasumable error */
1194	NOT4;				/* 080	spill 0 normal */
1195	SPILL_32bit_asi(ASI_AIUP,sn0);	/* 084	spill 1 normal */
1196	SPILL_64bit_asi(ASI_AIUP,sn0);	/* 088	spill 2 normal */
1197	SPILL_32clean(ASI_AIUP,sn0);	/* 08C	spill 3 normal */
1198	SPILL_64clean(ASI_AIUP,sn0);	/* 090	spill 4 normal */
1199	SPILL_32bit(not);		/* 094	spill 5 normal */
1200	SPILL_64bit(not);		/* 098	spill 6 normal */
1201	SPILL_mixed;			/* 09C	spill 7 normal */
1202	NOT4;				/* 0A0	spill 0 other */
1203	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0A4	spill 1 other */
1204	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0A8	spill 2 other */
1205	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0AC	spill 3 other */
1206	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0B0	spill 4 other */
1207	NOT4;				/* 0B4	spill 5 other */
1208	NOT4;				/* 0B8	spill 6 other */
1209	NOT4;				/* 0BC	spill 7 other */
1210	NOT4;				/* 0C0	fill 0 normal */
1211	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0C4	fill 1 normal */
1212	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0C8	fill 2 normal */
1213	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0CC	fill 3 normal */
1214	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0D0	fill 4 normal */
1215	FILL_32bit(not);		/* 0D4	fill 5 normal */
1216	FILL_64bit(not);		/* 0D8	fill 6 normal */
1217	FILL_mixed;			/* 0DC	fill 7 normal */
1218	NOT4;				/* 0E0	fill 0 other */
1219	NOT4;				/* 0E4	fill 1 other */
1220	NOT4;				/* 0E8	fill 2 other */
1221	NOT4;				/* 0EC	fill 3 other */
1222	NOT4;				/* 0F0	fill 4 other */
1223	NOT4;				/* 0F4	fill 5 other */
1224	NOT4;				/* 0F8	fill 6 other */
1225	NOT4;				/* 0FC	fill 7 other */
1226	/* user traps */
1227	GOTO(syscall_trap_4x);		/* 100	old system call */
1228	TRAP(T_BREAKPOINT);		/* 101	user breakpoint */
1229	TRAP(T_DIV0);			/* 102	user divide by zero */
1230	GOTO(.flushw);			/* 103	flush windows */
1231	GOTO(.clean_windows);		/* 104	clean windows */
1232	BAD;				/* 105	range check ?? */
1233	GOTO(.fix_alignment);		/* 106	do unaligned references */
1234	BAD;				/* 107	unused */
1235	SYSCALL(syscall_trap32);	/* 108	ILP32 system call on LP64 */
1236	GOTO(set_trap0_addr);		/* 109	set trap0 address */
1237	BAD; BAD; BAD4;			/* 10A - 10F unused */
1238	TRP4; TRP4; TRP4; TRP4;		/* 110 - 11F V9 user trap handlers */
1239	GOTO(.getcc);			/* 120	get condition codes */
1240	GOTO(.setcc);			/* 121	set condition codes */
1241	GOTO(.getpsr);			/* 122	get psr */
1242	GOTO(.setpsr);			/* 123	set psr (some fields) */
1243	GOTO(get_timestamp);		/* 124	get timestamp */
1244	GOTO(get_virtime);		/* 125	get lwp virtual time */
1245	PRIV(self_xcall);		/* 126	self xcall */
1246	GOTO(get_hrestime);		/* 127	get hrestime */
1247	BAD;				/* 128	ST_SETV9STACK */
1248	GOTO(.getlgrp);			/* 129  get lgrpid */
1249	BAD; BAD; BAD4;			/* 12A - 12F unused */
1250	BAD4; BAD4; 			/* 130 - 137 unused */
1251	DTRACE_PID;			/* 138  dtrace pid tracing provider */
1252	DTRACE_FASTTRAP;		/* 139  dtrace fasttrap provider */
1253	DTRACE_RETURN;			/* 13A	dtrace pid return probe */
1254	BAD; BAD4;			/* 13B - 13F unused */
1255	SYSCALL(syscall_trap)		/* 140  LP64 system call */
1256	BAD;				/* 141  unused */
1257#ifdef DEBUG_USER_TRAPTRACECTL
1258	GOTO(.traptrace_freeze);	/* 142  freeze traptrace */
1259	GOTO(.traptrace_unfreeze);	/* 143  unfreeze traptrace */
1260#else
1261	BAD; BAD;			/* 142 - 143 unused */
1262#endif
1263	BAD4; BAD4; BAD4;		/* 144 - 14F unused */
1264	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1265	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1266	BAD;				/* 170 - unused */
1267	BAD;				/* 171 - unused */
1268	BAD; BAD;			/* 172 - 173 unused */
1269	BAD4; BAD4;			/* 174 - 17B unused */
1270#ifdef	PTL1_PANIC_DEBUG
1271	mov PTL1_BAD_DEBUG, %g1; GOTO(ptl1_panic);
1272					/* 17C	test ptl1_panic */
1273#else
1274	BAD;				/* 17C  unused */
1275#endif	/* PTL1_PANIC_DEBUG */
1276	PRIV(kmdb_trap);		/* 17D	kmdb enter (L1-A) */
1277	PRIV(kmdb_trap);		/* 17E	kmdb breakpoint */
1278	PRIV(obp_bpt);			/* 17F	obp breakpoint */
1279	/* reserved */
1280	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1281	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1282	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1283	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1284	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1285	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1286	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1287	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1288	.size	trap_table0, (.-trap_table0)
1289trap_table1:
1290	NOT4; NOT4;			/* 000 - 007 unused */
1291	NOT;				/* 008	instruction access exception */
1292	ITSB_MISS;			/* 009	instruction access MMU miss */
1293 	NOT;				/* 00A  reserved */
1294	NOT; NOT4;			/* 00B - 00F unused */
1295	NOT4; NOT4; NOT4; NOT4;		/* 010 - 01F unused */
1296	NOT4;				/* 020 - 023 unused */
1297	CLEAN_WINDOW;			/* 024 - 027 clean window */
1298	NOT4; NOT4;			/* 028 - 02F unused */
1299	DMMU_EXCEPTION_TL1;		/* 030 	data access exception */
1300	DTSB_MISS;			/* 031  data access MMU miss */
1301	NOT;				/* 032  reserved */
1302	NOT;				/* 033	unused */
1303	MISALIGN_ADDR_TL1;		/* 034	mem address not aligned */
1304	NOT; NOT; NOT; NOT4; NOT4	/* 035 - 03F unused */
1305	NOT4; NOT4; NOT4; NOT4;		/* 040 - 04F unused */
1306	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F unused */
1307	NOT;				/* 060	unused */
1308	GOTO(kmdb_trap_tl1);		/* 061	PA watchpoint */
1309	GOTO(kmdb_trap_tl1);		/* 062	VA watchpoint */
1310	NOT;				/* 063	reserved */
1311	ITLB_MISS(tt1);			/* 064	instruction access MMU miss */
1312	DTLB_MISS(tt1);			/* 068	data access MMU miss */
1313	DTLB_PROT;			/* 06C	data access protection */
1314	NOT;				/* 070  reserved */
1315	NOT;				/* 071  reserved */
1316	NOT;				/* 072  reserved */
1317	NOT;				/* 073  reserved */
1318	NOT4; NOT4;			/* 074 - 07B reserved */
1319	NOT;				/* 07C  reserved */
1320	NOT;				/* 07D  reserved */
1321	NOT;				/* 07E  resumable error */
1322	GOTO_TT(nonresumable_error, trace_gen);	/* 07F  nonresumable error */
1323	NOTP4;				/* 080	spill 0 normal */
1324	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 084	spill 1 normal */
1325	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 088	spill 2 normal */
1326	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 08C	spill 3 normal */
1327	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 090	spill 4 normal */
1328	NOTP4;				/* 094	spill 5 normal */
1329	SPILL_64bit_ktt1(sk);		/* 098	spill 6 normal */
1330	SPILL_mixed_ktt1(sk);		/* 09C	spill 7 normal */
1331	NOTP4;				/* 0A0	spill 0 other */
1332	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0A4  spill 1 other */
1333	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0A8	spill 2 other */
1334	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0AC	spill 3 other */
1335	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0B0  spill 4 other */
1336	NOTP4;				/* 0B4  spill 5 other */
1337	NOTP4;				/* 0B8  spill 6 other */
1338	NOTP4;				/* 0BC  spill 7 other */
1339	NOT4;				/* 0C0	fill 0 normal */
1340	NOT4;				/* 0C4	fill 1 normal */
1341	NOT4;				/* 0C8	fill 2 normal */
1342	NOT4;				/* 0CC	fill 3 normal */
1343	NOT4;				/* 0D0	fill 4 normal */
1344	NOT4;				/* 0D4	fill 5 normal */
1345	NOT4;				/* 0D8	fill 6 normal */
1346	NOT4;				/* 0DC	fill 7 normal */
1347	NOT4; NOT4; NOT4; NOT4;		/* 0E0 - 0EF unused */
1348	NOT4; NOT4; NOT4; NOT4;		/* 0F0 - 0FF unused */
1349/*
1350 * Code running at TL>0 does not use soft traps, so
1351 * we can truncate the table here.
1352 * However:
1353 * sun4v uses (hypervisor) ta instructions at TL > 0, so
1354 * provide a safety net for now.
1355 */
1356	/* soft traps */
1357	BAD4; BAD4; BAD4; BAD4;		/* 100 - 10F unused */
1358	BAD4; BAD4; BAD4; BAD4;		/* 110 - 11F unused */
1359	BAD4; BAD4; BAD4; BAD4;		/* 120 - 12F unused */
1360	BAD4; BAD4; BAD4; BAD4;		/* 130 - 13F unused */
1361	BAD4; BAD4; BAD4; BAD4;		/* 140 - 14F unused */
1362	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1363	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1364	BAD4; BAD4; BAD4; BAD4;		/* 170 - 17F unused */
1365	/* reserved */
1366	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1367	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1368	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1369	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1370	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1371	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1372	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1373	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1374etrap_table:
1375	.size	trap_table1, (.-trap_table1)
1376	.size	trap_table, (.-trap_table)
1377	.size	scb, (.-scb)
1378
1379/*
1380 * We get to exec_fault in the case of an instruction miss and tte
1381 * has no execute bit set.  We go to tl0 to handle it.
1382 *
1383 * g1 = tsbe pointer (in/clobbered)
1384 * g2 = tag access register (in)
1385 * g3 - g4 = scratch (clobbered)
1386 * g5 = tsbe data (in)
1387 * g6 = scratch (clobbered)
1388 * g7 = pc we jumped here from (in)
1389 */
1390/*
1391 * synthesize for trap(): TAG_ACCESS in %g2
1392 */
1393	ALTENTRY(exec_fault)
1394	TRACE_TSBHIT(TT_MMU_EXEC)
1395	MMU_FAULT_STATUS_AREA(%g4)
1396	ldx	[%g4 + MMFSA_I_ADDR], %g2	/* g2 = address */
1397	ldx	[%g4 + MMFSA_I_CTX], %g3	/* g3 = ctx */
1398	srlx	%g2, MMU_PAGESHIFT, %g2		! align address to page boundry
1399	sllx	%g2, MMU_PAGESHIFT, %g2
1400	or	%g2, %g3, %g2			/* TAG_ACCESS */
1401	mov	T_INSTR_MMU_MISS, %g3		! arg2 = traptype
1402	set	trap, %g1
1403	ba,pt	%xcc, sys_trap
1404	  mov	-1, %g4
1405
1406.mmu_exception_not_aligned:
1407	/* %g2 = sfar, %g3 = sfsr */
1408	rdpr	%tstate, %g1
1409	btst	TSTATE_PRIV, %g1
1410	bnz,pn	%icc, 2f
1411	nop
1412	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1413	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1414	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1415	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1416	brz,pt	%g5, 2f
1417	nop
1418	ldn	[%g5 + P_UTRAP15], %g5			! unaligned utrap?
1419	brz,pn	%g5, 2f
1420	nop
1421	btst	1, %sp
1422	bz,pt	%xcc, 1f				! 32 bit user program
1423	nop
1424	ba,pt	%xcc, .setup_v9utrap			! 64 bit user program
1425	nop
14261:
1427	ba,pt	%xcc, .setup_utrap
1428	or	%g2, %g0, %g7
14292:
1430	ba,pt	%xcc, .mmu_exception_end
1431	mov	T_ALIGNMENT, %g1
1432
1433.mmu_priv_exception:
1434	rdpr	%tstate, %g1
1435	btst	TSTATE_PRIV, %g1
1436	bnz,pn	%icc, 1f
1437	nop
1438	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1439	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1440	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1441	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1442	brz,pt	%g5, 1f
1443	nop
1444	ldn	[%g5 + P_UTRAP16], %g5
1445	brnz,pt	%g5, .setup_v9utrap
1446	nop
14471:
1448	mov	T_PRIV_INSTR, %g1
1449
1450.mmu_exception_end:
1451	CPU_INDEX(%g4, %g5)
1452	set	cpu_core, %g5
1453	sllx	%g4, CPU_CORE_SHIFT, %g4
1454	add	%g4, %g5, %g4
1455	lduh	[%g4 + CPUC_DTRACE_FLAGS], %g5
1456	andcc	%g5, CPU_DTRACE_NOFAULT, %g0
1457	bz	1f
1458	or	%g5, CPU_DTRACE_BADADDR, %g5
1459	stuh	%g5, [%g4 + CPUC_DTRACE_FLAGS]
1460	done
1461
14621:
1463	sllx	%g3, 32, %g3
1464	or	%g3, %g1, %g3
1465	set	trap, %g1
1466	ba,pt	%xcc, sys_trap
1467	sub	%g0, 1, %g4
1468
1469.fp_disabled:
1470	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1471	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1472	rdpr	%tstate, %g4
1473	btst	TSTATE_PRIV, %g4
1474	bnz,a,pn %icc, ptl1_panic
1475	  mov	PTL1_BAD_FPTRAP, %g1
1476
1477	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1478	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1479	brz,a,pt %g5, 2f
1480	  nop
1481	ldn	[%g5 + P_UTRAP7], %g5			! fp_disabled utrap?
1482	brz,a,pn %g5, 2f
1483	  nop
1484	btst	1, %sp
1485	bz,a,pt	%xcc, 1f				! 32 bit user program
1486	  nop
1487	ba,a,pt	%xcc, .setup_v9utrap			! 64 bit user program
1488	  nop
14891:
1490	ba,pt	%xcc, .setup_utrap
1491	  or	%g0, %g0, %g7
14922:
1493	set	fp_disabled, %g1
1494	ba,pt	%xcc, sys_trap
1495	  sub	%g0, 1, %g4
1496
1497.fp_ieee_exception:
1498	rdpr	%tstate, %g1
1499	btst	TSTATE_PRIV, %g1
1500	bnz,a,pn %icc, ptl1_panic
1501	  mov	PTL1_BAD_FPTRAP, %g1
1502	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1503	stx	%fsr, [%g1 + CPU_TMP1]
1504	ldx	[%g1 + CPU_TMP1], %g2
1505	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1506	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1507	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1508	brz,a,pt %g5, 1f
1509	  nop
1510	ldn	[%g5 + P_UTRAP8], %g5
1511	brnz,a,pt %g5, .setup_v9utrap
1512	  nop
15131:
1514	set	_fp_ieee_exception, %g1
1515	ba,pt	%xcc, sys_trap
1516	  sub	%g0, 1, %g4
1517
1518/*
1519 * Register Inputs:
1520 *	%g5		user trap handler
1521 *	%g7		misaligned addr - for alignment traps only
1522 */
1523.setup_utrap:
1524	set	trap, %g1			! setup in case we go
1525	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1526	sub	%g0, 1, %g4			! the save instruction below
1527
1528	/*
1529	 * If the DTrace pid provider is single stepping a copied-out
1530	 * instruction, t->t_dtrace_step will be set. In that case we need
1531	 * to abort the single-stepping (since execution of the instruction
1532	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1533	 */
1534	save	%sp, -SA(MINFRAME32), %sp	! window for trap handler
1535	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1536	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1537	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1538	rdpr	%tnpc, %l2			! arg1 == tnpc
1539	brz,pt	%g2, 1f
1540	rdpr	%tpc, %l1			! arg0 == tpc
1541
1542	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1543	ldn	[%g1 + T_DTRACE_NPC], %l2	! arg1 = t->t_dtrace_npc (step)
1544	brz,pt	%g2, 1f
1545	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1546	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
15471:
1548	mov	%g7, %l3			! arg2 == misaligned address
1549
1550	rdpr	%tstate, %g1			! cwp for trap handler
1551	rdpr	%cwp, %g4
1552	bclr	TSTATE_CWP_MASK, %g1
1553	wrpr	%g1, %g4, %tstate
1554	wrpr	%g0, %g5, %tnpc			! trap handler address
1555	FAST_TRAP_DONE
1556	/* NOTREACHED */
1557
1558.check_v9utrap:
1559	rdpr	%tstate, %g1
1560	btst	TSTATE_PRIV, %g1
1561	bnz,a,pn %icc, 3f
1562	  nop
1563	CPU_ADDR(%g4, %g1)				! load CPU struct addr
1564	ldn	[%g4 + CPU_THREAD], %g5			! load thread pointer
1565	ldn	[%g5 + T_PROCP], %g5			! load proc pointer
1566	ldn	[%g5 + P_UTRAPS], %g5			! are there utraps?
1567
1568	cmp	%g3, T_SOFTWARE_TRAP
1569	bne,a,pt %icc, 1f
1570	  nop
1571
1572	brz,pt %g5, 3f			! if p_utraps == NULL goto trap()
1573	  rdpr	%tt, %g3		! delay - get actual hw trap type
1574
1575	sub	%g3, 254, %g1		! UT_TRAP_INSTRUCTION_16 = p_utraps[18]
1576	ba,pt	%icc, 2f
1577	  smul	%g1, CPTRSIZE, %g2
15781:
1579	brz,a,pt %g5, 3f		! if p_utraps == NULL goto trap()
1580	  nop
1581
1582	cmp	%g3, T_UNIMP_INSTR
1583	bne,a,pt %icc, 2f
1584	  nop
1585
1586	mov	1, %g1
1587	st	%g1, [%g4 + CPU_TL1_HDLR] ! set CPU_TL1_HDLR
1588	rdpr	%tpc, %g1		! ld trapping instruction using
1589	lduwa	[%g1]ASI_AIUP, %g1	! "AS IF USER" ASI which could fault
1590	st	%g0, [%g4 + CPU_TL1_HDLR] ! clr CPU_TL1_HDLR
1591
1592	sethi	%hi(0xc1c00000), %g4	! setup mask for illtrap instruction
1593	andcc	%g1, %g4, %g4		! and instruction with mask
1594	bnz,a,pt %icc, 3f		! if %g4 == zero, %g1 is an ILLTRAP
1595	  nop				! fall thru to setup
15962:
1597	ldn	[%g5 + %g2], %g5
1598	brnz,a,pt %g5, .setup_v9utrap
1599	  nop
16003:
1601	set	trap, %g1
1602	ba,pt	%xcc, sys_trap
1603	  sub	%g0, 1, %g4
1604	/* NOTREACHED */
1605
1606/*
1607 * Register Inputs:
1608 *	%g5		user trap handler
1609 */
1610.setup_v9utrap:
1611	set	trap, %g1			! setup in case we go
1612	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1613	sub	%g0, 1, %g4			! the save instruction below
1614
1615	/*
1616	 * If the DTrace pid provider is single stepping a copied-out
1617	 * instruction, t->t_dtrace_step will be set. In that case we need
1618	 * to abort the single-stepping (since execution of the instruction
1619	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1620	 */
1621	save	%sp, -SA(MINFRAME64), %sp	! window for trap handler
1622	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1623	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1624	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1625	rdpr	%tnpc, %l7			! arg1 == tnpc
1626	brz,pt	%g2, 1f
1627	rdpr	%tpc, %l6			! arg0 == tpc
1628
1629	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1630	ldn	[%g1 + T_DTRACE_NPC], %l7	! arg1 == t->t_dtrace_npc (step)
1631	brz,pt	%g2, 1f
1632	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1633	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
16341:
1635	rdpr	%tstate, %g2			! cwp for trap handler
1636	rdpr	%cwp, %g4
1637	bclr	TSTATE_CWP_MASK, %g2
1638	wrpr	%g2, %g4, %tstate
1639
1640	ldn	[%g1 + T_PROCP], %g4		! load proc pointer
1641	ldn	[%g4 + P_AS], %g4		! load as pointer
1642	ldn	[%g4 + A_USERLIMIT], %g4	! load as userlimit
1643	cmp	%l7, %g4			! check for single-step set
1644	bne,pt	%xcc, 4f
1645	  nop
1646	ldn	[%g1 + T_LWP], %g1		! load klwp pointer
1647	ld	[%g1 + PCB_STEP], %g4		! load single-step flag
1648	cmp	%g4, STEP_ACTIVE		! step flags set in pcb?
1649	bne,pt	%icc, 4f
1650	  nop
1651	stn	%g5, [%g1 + PCB_TRACEPC]	! save trap handler addr in pcb
1652	mov	%l7, %g4			! on entry to precise user trap
1653	add	%l6, 4, %l7			! handler, %l6 == pc, %l7 == npc
1654						! at time of trap
1655	wrpr	%g0, %g4, %tnpc			! generate FLTBOUNDS,
1656						! %g4 == userlimit
1657	FAST_TRAP_DONE
1658	/* NOTREACHED */
16594:
1660	wrpr	%g0, %g5, %tnpc			! trap handler address
1661	FAST_TRAP_DONE_CHK_INTR
1662	/* NOTREACHED */
1663
1664.fp_exception:
1665	CPU_ADDR(%g1, %g4)
1666	stx	%fsr, [%g1 + CPU_TMP1]
1667	ldx	[%g1 + CPU_TMP1], %g2
1668
1669	/*
1670	 * Cheetah takes unfinished_FPop trap for certain range of operands
1671	 * to the "fitos" instruction. Instead of going through the slow
1672	 * software emulation path, we try to simulate the "fitos" instruction
1673	 * via "fitod" and "fdtos" provided the following conditions are met:
1674	 *
1675	 *	fpu_exists is set (if DEBUG)
1676	 *	not in privileged mode
1677	 *	ftt is unfinished_FPop
1678	 *	NXM IEEE trap is not enabled
1679	 *	instruction at %tpc is "fitos"
1680	 *
1681	 *  Usage:
1682	 *	%g1	per cpu address
1683	 *	%g2	%fsr
1684	 *	%g6	user instruction
1685	 *
1686	 * Note that we can take a memory access related trap while trying
1687	 * to fetch the user instruction. Therefore, we set CPU_TL1_HDLR
1688	 * flag to catch those traps and let the SFMMU code deal with page
1689	 * fault and data access exception.
1690	 */
1691#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
1692	sethi	%hi(fpu_exists), %g7
1693	ld	[%g7 + %lo(fpu_exists)], %g7
1694	brz,pn %g7, .fp_exception_cont
1695	  nop
1696#endif
1697	rdpr	%tstate, %g7			! branch if in privileged mode
1698	btst	TSTATE_PRIV, %g7
1699	bnz,pn	%xcc, .fp_exception_cont
1700	srl	%g2, FSR_FTT_SHIFT, %g7		! extract ftt from %fsr
1701	and	%g7, (FSR_FTT>>FSR_FTT_SHIFT), %g7
1702	cmp	%g7, FTT_UNFIN
1703	set	FSR_TEM_NX, %g5
1704	bne,pn	%xcc, .fp_exception_cont	! branch if NOT unfinished_FPop
1705	  andcc	%g2, %g5, %g0
1706	bne,pn	%xcc, .fp_exception_cont	! branch if FSR_TEM_NX enabled
1707	  rdpr	%tpc, %g5			! get faulting PC
1708
1709	or	%g0, 1, %g7
1710	st	%g7, [%g1 + CPU_TL1_HDLR]	! set tl1_hdlr flag
1711	lda	[%g5]ASI_USER, %g6		! get user's instruction
1712	st	%g0, [%g1 + CPU_TL1_HDLR]	! clear tl1_hdlr flag
1713
1714	set	FITOS_INSTR_MASK, %g7
1715	and	%g6, %g7, %g7
1716	set	FITOS_INSTR, %g5
1717	cmp	%g7, %g5
1718	bne,pn	%xcc, .fp_exception_cont	! branch if not FITOS_INSTR
1719	 nop
1720
1721	/*
1722	 * This is unfinished FPops trap for "fitos" instruction. We
1723	 * need to simulate "fitos" via "fitod" and "fdtos" instruction
1724	 * sequence.
1725	 *
1726	 * We need a temporary FP register to do the conversion. Since
1727	 * both source and destination operands for the "fitos" instruction
1728	 * have to be within %f0-%f31, we use an FP register from the upper
1729	 * half to guarantee that it won't collide with the source or the
1730	 * dest operand. However, we do have to save and restore its value.
1731	 *
1732	 * We use %d62 as a temporary FP register for the conversion and
1733	 * branch to appropriate instruction within the conversion tables
1734	 * based upon the rs2 and rd values.
1735	 */
1736
1737	std	%d62, [%g1 + CPU_TMP1]		! save original value
1738
1739	srl	%g6, FITOS_RS2_SHIFT, %g7
1740	and	%g7, FITOS_REG_MASK, %g7
1741	set	_fitos_fitod_table, %g4
1742	sllx	%g7, 2, %g7
1743	jmp	%g4 + %g7
1744	  ba,pt	%xcc, _fitos_fitod_done
1745	.empty
1746
1747_fitos_fitod_table:
1748	  fitod	%f0, %d62
1749	  fitod	%f1, %d62
1750	  fitod	%f2, %d62
1751	  fitod	%f3, %d62
1752	  fitod	%f4, %d62
1753	  fitod	%f5, %d62
1754	  fitod	%f6, %d62
1755	  fitod	%f7, %d62
1756	  fitod	%f8, %d62
1757	  fitod	%f9, %d62
1758	  fitod	%f10, %d62
1759	  fitod	%f11, %d62
1760	  fitod	%f12, %d62
1761	  fitod	%f13, %d62
1762	  fitod	%f14, %d62
1763	  fitod	%f15, %d62
1764	  fitod	%f16, %d62
1765	  fitod	%f17, %d62
1766	  fitod	%f18, %d62
1767	  fitod	%f19, %d62
1768	  fitod	%f20, %d62
1769	  fitod	%f21, %d62
1770	  fitod	%f22, %d62
1771	  fitod	%f23, %d62
1772	  fitod	%f24, %d62
1773	  fitod	%f25, %d62
1774	  fitod	%f26, %d62
1775	  fitod	%f27, %d62
1776	  fitod	%f28, %d62
1777	  fitod	%f29, %d62
1778	  fitod	%f30, %d62
1779	  fitod	%f31, %d62
1780_fitos_fitod_done:
1781
1782	/*
1783	 * Now convert data back into single precision
1784	 */
1785	srl	%g6, FITOS_RD_SHIFT, %g7
1786	and	%g7, FITOS_REG_MASK, %g7
1787	set	_fitos_fdtos_table, %g4
1788	sllx	%g7, 2, %g7
1789	jmp	%g4 + %g7
1790	  ba,pt	%xcc, _fitos_fdtos_done
1791	.empty
1792
1793_fitos_fdtos_table:
1794	  fdtos	%d62, %f0
1795	  fdtos	%d62, %f1
1796	  fdtos	%d62, %f2
1797	  fdtos	%d62, %f3
1798	  fdtos	%d62, %f4
1799	  fdtos	%d62, %f5
1800	  fdtos	%d62, %f6
1801	  fdtos	%d62, %f7
1802	  fdtos	%d62, %f8
1803	  fdtos	%d62, %f9
1804	  fdtos	%d62, %f10
1805	  fdtos	%d62, %f11
1806	  fdtos	%d62, %f12
1807	  fdtos	%d62, %f13
1808	  fdtos	%d62, %f14
1809	  fdtos	%d62, %f15
1810	  fdtos	%d62, %f16
1811	  fdtos	%d62, %f17
1812	  fdtos	%d62, %f18
1813	  fdtos	%d62, %f19
1814	  fdtos	%d62, %f20
1815	  fdtos	%d62, %f21
1816	  fdtos	%d62, %f22
1817	  fdtos	%d62, %f23
1818	  fdtos	%d62, %f24
1819	  fdtos	%d62, %f25
1820	  fdtos	%d62, %f26
1821	  fdtos	%d62, %f27
1822	  fdtos	%d62, %f28
1823	  fdtos	%d62, %f29
1824	  fdtos	%d62, %f30
1825	  fdtos	%d62, %f31
1826_fitos_fdtos_done:
1827
1828	ldd	[%g1 + CPU_TMP1], %d62		! restore %d62
1829
1830#if DEBUG
1831	/*
1832	 * Update FPop_unfinished trap kstat
1833	 */
1834	set	fpustat+FPUSTAT_UNFIN_KSTAT, %g7
1835	ldx	[%g7], %g5
18361:
1837	add	%g5, 1, %g6
1838
1839	casxa	[%g7] ASI_N, %g5, %g6
1840	cmp	%g5, %g6
1841	bne,a,pn %xcc, 1b
1842	  or	%g0, %g6, %g5
1843
1844	/*
1845	 * Update fpu_sim_fitos kstat
1846	 */
1847	set	fpuinfo+FPUINFO_FITOS_KSTAT, %g7
1848	ldx	[%g7], %g5
18491:
1850	add	%g5, 1, %g6
1851
1852	casxa	[%g7] ASI_N, %g5, %g6
1853	cmp	%g5, %g6
1854	bne,a,pn %xcc, 1b
1855	  or	%g0, %g6, %g5
1856#endif /* DEBUG */
1857
1858	FAST_TRAP_DONE
1859
1860.fp_exception_cont:
1861	/*
1862	 * Let _fp_exception deal with simulating FPop instruction.
1863	 * Note that we need to pass %fsr in %g2 (already read above).
1864	 */
1865
1866	set	_fp_exception, %g1
1867	ba,pt	%xcc, sys_trap
1868	sub	%g0, 1, %g4
1869
1870
1871/*
1872 * Register windows
1873 */
1874
1875/*
1876 * FILL_32bit_flushw/FILL_64bit_flushw fills a 32/64-bit-wide register window
1877 * from a 32/64-bit * wide address space via the designated asi.
1878 * It is used to fill windows in user_flushw to avoid going above TL 2.
1879 */
1880/* TODO: Use the faster FILL based on FILL_32bit_asi/FILL_64bit_asi */
1881#define	FILL_32bit_flushw(asi_num)				\
1882	mov	asi_num, %asi					;\
1883	rdpr	%cwp, %g2					;\
1884	sub	%g2, 1, %g2					;\
1885	wrpr	%g2, %cwp					;\
18861:	srl	%sp, 0, %sp					;\
1887	lda	[%sp + 0]%asi, %l0				;\
1888	lda	[%sp + 4]%asi, %l1				;\
1889	lda	[%sp + 8]%asi, %l2				;\
1890	lda	[%sp + 12]%asi, %l3				;\
1891	lda	[%sp + 16]%asi, %l4				;\
1892	lda	[%sp + 20]%asi, %l5				;\
1893	lda	[%sp + 24]%asi, %l6				;\
1894	lda	[%sp + 28]%asi, %l7				;\
1895	lda	[%sp + 32]%asi, %i0				;\
1896	lda	[%sp + 36]%asi, %i1				;\
1897	lda	[%sp + 40]%asi, %i2				;\
1898	lda	[%sp + 44]%asi, %i3				;\
1899	lda	[%sp + 48]%asi, %i4				;\
1900	lda	[%sp + 52]%asi, %i5				;\
1901	lda	[%sp + 56]%asi, %i6				;\
1902	lda	[%sp + 60]%asi, %i7				;\
1903	restored						;\
1904	add	%g2, 1, %g2					;\
1905	wrpr	%g2, %cwp
1906
1907#define	FILL_64bit_flushw(asi_num)				\
1908	mov	asi_num, %asi					;\
1909	rdpr	%cwp, %g2					;\
1910	sub	%g2, 1, %g2					;\
1911	wrpr	%g2, %cwp					;\
1912	ldxa	[%sp + V9BIAS64 + 0]%asi, %l0			;\
1913	ldxa	[%sp + V9BIAS64 + 8]%asi, %l1			;\
1914	ldxa	[%sp + V9BIAS64 + 16]%asi, %l2			;\
1915	ldxa	[%sp + V9BIAS64 + 24]%asi, %l3			;\
1916	ldxa	[%sp + V9BIAS64 + 32]%asi, %l4			;\
1917	ldxa	[%sp + V9BIAS64 + 40]%asi, %l5			;\
1918	ldxa	[%sp + V9BIAS64 + 48]%asi, %l6			;\
1919	ldxa	[%sp + V9BIAS64 + 56]%asi, %l7			;\
1920	ldxa	[%sp + V9BIAS64 + 64]%asi, %i0			;\
1921	ldxa	[%sp + V9BIAS64 + 72]%asi, %i1			;\
1922	ldxa	[%sp + V9BIAS64 + 80]%asi, %i2			;\
1923	ldxa	[%sp + V9BIAS64 + 88]%asi, %i3			;\
1924	ldxa	[%sp + V9BIAS64 + 96]%asi, %i4			;\
1925	ldxa	[%sp + V9BIAS64 + 104]%asi, %i5			;\
1926	ldxa	[%sp + V9BIAS64 + 112]%asi, %i6			;\
1927	ldxa	[%sp + V9BIAS64 + 120]%asi, %i7			;\
1928	restored						;\
1929	add	%g2, 1, %g2					;\
1930	wrpr	%g2, %cwp
1931
1932.flushw:
1933	rdpr	%tnpc, %g1
1934	wrpr	%g1, %tpc
1935	add	%g1, 4, %g1
1936	wrpr	%g1, %tnpc
1937	set	trap, %g1
1938	mov	T_FLUSH_PCB, %g3
1939	ba,pt	%xcc, sys_trap
1940	sub	%g0, 1, %g4
1941
1942.clean_windows:
1943	set	trap, %g1
1944	mov	T_FLUSH_PCB, %g3
1945	sub	%g0, 1, %g4
1946	save
1947	flushw
1948	rdpr	%canrestore, %g2
1949	brnz	%g2, 1f
1950	nop
1951	rdpr	%wstate, %g2
1952	btst	1, %g2
1953	beq	2f
1954	nop
1955	FILL_32bit_flushw(ASI_AIUP)
1956	ba,a	1f
1957	 .empty
19582:
1959	FILL_64bit_flushw(ASI_AIUP)
19601:
1961	restore
1962	wrpr	%g0, %g0, %cleanwin	! no clean windows
1963
1964	CPU_ADDR(%g4, %g5)
1965	ldn	[%g4 + CPU_MPCB], %g4
1966	brz,a,pn %g4, 1f
1967	  nop
1968	ld	[%g4 + MPCB_WSTATE], %g5
1969	add	%g5, WSTATE_CLEAN_OFFSET, %g5
1970	wrpr	%g0, %g5, %wstate
19711:	FAST_TRAP_DONE
1972
1973/*
1974 * .spill_clean: clean the previous window, restore the wstate, and
1975 * "done".
1976 *
1977 * Entry: %g7 contains new wstate
1978 */
1979.spill_clean:
1980	sethi	%hi(nwin_minus_one), %g5
1981	ld	[%g5 + %lo(nwin_minus_one)], %g5 ! %g5 = nwin - 1
1982	rdpr	%cwp, %g6			! %g6 = %cwp
1983	deccc	%g6				! %g6--
1984	movneg	%xcc, %g5, %g6			! if (%g6<0) %g6 = nwin-1
1985	wrpr	%g6, %cwp
1986	TT_TRACE_L(trace_win)
1987	clr	%l0
1988	clr	%l1
1989	clr	%l2
1990	clr	%l3
1991	clr	%l4
1992	clr	%l5
1993	clr	%l6
1994	clr	%l7
1995	wrpr	%g0, %g7, %wstate
1996	saved
1997	retry			! restores correct %cwp
1998
1999.fix_alignment:
2000	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2001	ldn	[%g1 + CPU_THREAD], %g1	! load thread pointer
2002	ldn	[%g1 + T_PROCP], %g1
2003	mov	1, %g2
2004	stb	%g2, [%g1 + P_FIXALIGNMENT]
2005	FAST_TRAP_DONE
2006
2007#define	STDF_REG(REG, ADDR, TMP)		\
2008	sll	REG, 3, REG			;\
2009mark1:	set	start1, TMP			;\
2010	jmp	REG + TMP			;\
2011	  nop					;\
2012start1:	ba,pt	%xcc, done1			;\
2013	  std	%f0, [ADDR + CPU_TMP1]		;\
2014	ba,pt	%xcc, done1			;\
2015	  std	%f32, [ADDR + CPU_TMP1]		;\
2016	ba,pt	%xcc, done1			;\
2017	  std	%f2, [ADDR + CPU_TMP1]		;\
2018	ba,pt	%xcc, done1			;\
2019	  std	%f34, [ADDR + CPU_TMP1]		;\
2020	ba,pt	%xcc, done1			;\
2021	  std	%f4, [ADDR + CPU_TMP1]		;\
2022	ba,pt	%xcc, done1			;\
2023	  std	%f36, [ADDR + CPU_TMP1]		;\
2024	ba,pt	%xcc, done1			;\
2025	  std	%f6, [ADDR + CPU_TMP1]		;\
2026	ba,pt	%xcc, done1			;\
2027	  std	%f38, [ADDR + CPU_TMP1]		;\
2028	ba,pt	%xcc, done1			;\
2029	  std	%f8, [ADDR + CPU_TMP1]		;\
2030	ba,pt	%xcc, done1			;\
2031	  std	%f40, [ADDR + CPU_TMP1]		;\
2032	ba,pt	%xcc, done1			;\
2033	  std	%f10, [ADDR + CPU_TMP1]		;\
2034	ba,pt	%xcc, done1			;\
2035	  std	%f42, [ADDR + CPU_TMP1]		;\
2036	ba,pt	%xcc, done1			;\
2037	  std	%f12, [ADDR + CPU_TMP1]		;\
2038	ba,pt	%xcc, done1			;\
2039	  std	%f44, [ADDR + CPU_TMP1]		;\
2040	ba,pt	%xcc, done1			;\
2041	  std	%f14, [ADDR + CPU_TMP1]		;\
2042	ba,pt	%xcc, done1			;\
2043	  std	%f46, [ADDR + CPU_TMP1]		;\
2044	ba,pt	%xcc, done1			;\
2045	  std	%f16, [ADDR + CPU_TMP1]		;\
2046	ba,pt	%xcc, done1			;\
2047	  std	%f48, [ADDR + CPU_TMP1]		;\
2048	ba,pt	%xcc, done1			;\
2049	  std	%f18, [ADDR + CPU_TMP1]		;\
2050	ba,pt	%xcc, done1			;\
2051	  std	%f50, [ADDR + CPU_TMP1]		;\
2052	ba,pt	%xcc, done1			;\
2053	  std	%f20, [ADDR + CPU_TMP1]		;\
2054	ba,pt	%xcc, done1			;\
2055	  std	%f52, [ADDR + CPU_TMP1]		;\
2056	ba,pt	%xcc, done1			;\
2057	  std	%f22, [ADDR + CPU_TMP1]		;\
2058	ba,pt	%xcc, done1			;\
2059	  std	%f54, [ADDR + CPU_TMP1]		;\
2060	ba,pt	%xcc, done1			;\
2061	  std	%f24, [ADDR + CPU_TMP1]		;\
2062	ba,pt	%xcc, done1			;\
2063	  std	%f56, [ADDR + CPU_TMP1]		;\
2064	ba,pt	%xcc, done1			;\
2065	  std	%f26, [ADDR + CPU_TMP1]		;\
2066	ba,pt	%xcc, done1			;\
2067	  std	%f58, [ADDR + CPU_TMP1]		;\
2068	ba,pt	%xcc, done1			;\
2069	  std	%f28, [ADDR + CPU_TMP1]		;\
2070	ba,pt	%xcc, done1			;\
2071	  std	%f60, [ADDR + CPU_TMP1]		;\
2072	ba,pt	%xcc, done1			;\
2073	  std	%f30, [ADDR + CPU_TMP1]		;\
2074	ba,pt	%xcc, done1			;\
2075	  std	%f62, [ADDR + CPU_TMP1]		;\
2076done1:
2077
2078#define	LDDF_REG(REG, ADDR, TMP)		\
2079	sll	REG, 3, REG			;\
2080mark2:	set	start2, TMP			;\
2081	jmp	REG + TMP			;\
2082	  nop					;\
2083start2:	ba,pt	%xcc, done2			;\
2084	  ldd	[ADDR + CPU_TMP1], %f0		;\
2085	ba,pt	%xcc, done2			;\
2086	  ldd	[ADDR + CPU_TMP1], %f32		;\
2087	ba,pt	%xcc, done2			;\
2088	  ldd	[ADDR + CPU_TMP1], %f2		;\
2089	ba,pt	%xcc, done2			;\
2090	  ldd	[ADDR + CPU_TMP1], %f34		;\
2091	ba,pt	%xcc, done2			;\
2092	  ldd	[ADDR + CPU_TMP1], %f4		;\
2093	ba,pt	%xcc, done2			;\
2094	  ldd	[ADDR + CPU_TMP1], %f36		;\
2095	ba,pt	%xcc, done2			;\
2096	  ldd	[ADDR + CPU_TMP1], %f6		;\
2097	ba,pt	%xcc, done2			;\
2098	  ldd	[ADDR + CPU_TMP1], %f38		;\
2099	ba,pt	%xcc, done2			;\
2100	  ldd	[ADDR + CPU_TMP1], %f8		;\
2101	ba,pt	%xcc, done2			;\
2102	  ldd	[ADDR + CPU_TMP1], %f40		;\
2103	ba,pt	%xcc, done2			;\
2104	  ldd	[ADDR + CPU_TMP1], %f10		;\
2105	ba,pt	%xcc, done2			;\
2106	  ldd	[ADDR + CPU_TMP1], %f42		;\
2107	ba,pt	%xcc, done2			;\
2108	  ldd	[ADDR + CPU_TMP1], %f12		;\
2109	ba,pt	%xcc, done2			;\
2110	  ldd	[ADDR + CPU_TMP1], %f44		;\
2111	ba,pt	%xcc, done2			;\
2112	  ldd	[ADDR + CPU_TMP1], %f14		;\
2113	ba,pt	%xcc, done2			;\
2114	  ldd	[ADDR + CPU_TMP1], %f46		;\
2115	ba,pt	%xcc, done2			;\
2116	  ldd	[ADDR + CPU_TMP1], %f16		;\
2117	ba,pt	%xcc, done2			;\
2118	  ldd	[ADDR + CPU_TMP1], %f48		;\
2119	ba,pt	%xcc, done2			;\
2120	  ldd	[ADDR + CPU_TMP1], %f18		;\
2121	ba,pt	%xcc, done2			;\
2122	  ldd	[ADDR + CPU_TMP1], %f50		;\
2123	ba,pt	%xcc, done2			;\
2124	  ldd	[ADDR + CPU_TMP1], %f20		;\
2125	ba,pt	%xcc, done2			;\
2126	  ldd	[ADDR + CPU_TMP1], %f52		;\
2127	ba,pt	%xcc, done2			;\
2128	  ldd	[ADDR + CPU_TMP1], %f22		;\
2129	ba,pt	%xcc, done2			;\
2130	  ldd	[ADDR + CPU_TMP1], %f54		;\
2131	ba,pt	%xcc, done2			;\
2132	  ldd	[ADDR + CPU_TMP1], %f24		;\
2133	ba,pt	%xcc, done2			;\
2134	  ldd	[ADDR + CPU_TMP1], %f56		;\
2135	ba,pt	%xcc, done2			;\
2136	  ldd	[ADDR + CPU_TMP1], %f26		;\
2137	ba,pt	%xcc, done2			;\
2138	  ldd	[ADDR + CPU_TMP1], %f58		;\
2139	ba,pt	%xcc, done2			;\
2140	  ldd	[ADDR + CPU_TMP1], %f28		;\
2141	ba,pt	%xcc, done2			;\
2142	  ldd	[ADDR + CPU_TMP1], %f60		;\
2143	ba,pt	%xcc, done2			;\
2144	  ldd	[ADDR + CPU_TMP1], %f30		;\
2145	ba,pt	%xcc, done2			;\
2146	  ldd	[ADDR + CPU_TMP1], %f62		;\
2147done2:
2148
2149.lddf_exception_not_aligned:
2150	/* %g2 = sfar, %g3 = sfsr */
2151	mov	%g2, %g5		! stash sfar
2152#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2153	sethi	%hi(fpu_exists), %g2	! check fpu_exists
2154	ld	[%g2 + %lo(fpu_exists)], %g2
2155	brz,a,pn %g2, 4f
2156	  nop
2157#endif
2158	CPU_ADDR(%g1, %g4)
2159	or	%g0, 1, %g4
2160	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2161
2162	rdpr	%tpc, %g2
2163	lda	[%g2]ASI_AIUP, %g6	! get the user's lddf instruction
2164	srl	%g6, 23, %g1		! using ldda or not?
2165	and	%g1, 1, %g1
2166	brz,a,pt %g1, 2f		! check for ldda instruction
2167	  nop
2168	srl	%g6, 13, %g1		! check immflag
2169	and	%g1, 1, %g1
2170	rdpr	%tstate, %g2		! %tstate in %g2
2171	brnz,a,pn %g1, 1f
2172	  srl	%g2, 31, %g1		! get asi from %tstate
2173	srl	%g6, 5, %g1		! get asi from instruction
2174	and	%g1, 0xFF, %g1		! imm_asi field
21751:
2176	cmp	%g1, ASI_P		! primary address space
2177	be,a,pt %icc, 2f
2178	  nop
2179	cmp	%g1, ASI_PNF		! primary no fault address space
2180	be,a,pt %icc, 2f
2181	  nop
2182	cmp	%g1, ASI_S		! secondary address space
2183	be,a,pt %icc, 2f
2184	  nop
2185	cmp	%g1, ASI_SNF		! secondary no fault address space
2186	bne,a,pn %icc, 3f
2187	  nop
21882:
2189	lduwa	[%g5]ASI_USER, %g7	! get first half of misaligned data
2190	add	%g5, 4, %g5		! increment misaligned data address
2191	lduwa	[%g5]ASI_USER, %g5	! get second half of misaligned data
2192
2193	sllx	%g7, 32, %g7
2194	or	%g5, %g7, %g5		! combine data
2195	CPU_ADDR(%g7, %g1)		! save data on a per-cpu basis
2196	stx	%g5, [%g7 + CPU_TMP1]	! save in cpu_tmp1
2197
2198	srl	%g6, 25, %g3		! %g6 has the instruction
2199	and	%g3, 0x1F, %g3		! %g3 has rd
2200	LDDF_REG(%g3, %g7, %g4)
2201
2202	CPU_ADDR(%g1, %g4)
2203	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2204	FAST_TRAP_DONE
22053:
2206	CPU_ADDR(%g1, %g4)
2207	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
22084:
2209	set	T_USER, %g3		! trap type in %g3
2210	or	%g3, T_LDDF_ALIGN, %g3
2211	mov	%g5, %g2		! misaligned vaddr in %g2
2212	set	fpu_trap, %g1		! goto C for the little and
2213	ba,pt	%xcc, sys_trap		! no fault little asi's
2214	  sub	%g0, 1, %g4
2215
2216.stdf_exception_not_aligned:
2217	/* %g2 = sfar, %g3 = sfsr */
2218	mov	%g2, %g5
2219
2220#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2221	sethi	%hi(fpu_exists), %g7		! check fpu_exists
2222	ld	[%g7 + %lo(fpu_exists)], %g3
2223	brz,a,pn %g3, 4f
2224	  nop
2225#endif
2226	CPU_ADDR(%g1, %g4)
2227	or	%g0, 1, %g4
2228	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2229
2230	rdpr	%tpc, %g2
2231	lda	[%g2]ASI_AIUP, %g6	! get the user's stdf instruction
2232
2233	srl	%g6, 23, %g1		! using stda or not?
2234	and	%g1, 1, %g1
2235	brz,a,pt %g1, 2f		! check for stda instruction
2236	  nop
2237	srl	%g6, 13, %g1		! check immflag
2238	and	%g1, 1, %g1
2239	rdpr	%tstate, %g2		! %tstate in %g2
2240	brnz,a,pn %g1, 1f
2241	  srl	%g2, 31, %g1		! get asi from %tstate
2242	srl	%g6, 5, %g1		! get asi from instruction
2243	and	%g1, 0xff, %g1		! imm_asi field
22441:
2245	cmp	%g1, ASI_P		! primary address space
2246	be,a,pt %icc, 2f
2247	  nop
2248	cmp	%g1, ASI_S		! secondary address space
2249	bne,a,pn %icc, 3f
2250	  nop
22512:
2252	srl	%g6, 25, %g6
2253	and	%g6, 0x1F, %g6		! %g6 has rd
2254	CPU_ADDR(%g7, %g1)
2255	STDF_REG(%g6, %g7, %g4)		! STDF_REG(REG, ADDR, TMP)
2256
2257	ldx	[%g7 + CPU_TMP1], %g6
2258	srlx	%g6, 32, %g7
2259	stuwa	%g7, [%g5]ASI_USER	! first half
2260	add	%g5, 4, %g5		! increment misaligned data address
2261	stuwa	%g6, [%g5]ASI_USER	! second half
2262
2263	CPU_ADDR(%g1, %g4)
2264	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2265	FAST_TRAP_DONE
22663:
2267	CPU_ADDR(%g1, %g4)
2268	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
22694:
2270	set	T_USER, %g3		! trap type in %g3
2271	or	%g3, T_STDF_ALIGN, %g3
2272	mov	%g5, %g2		! misaligned vaddr in %g2
2273	set	fpu_trap, %g1		! goto C for the little and
2274	ba,pt	%xcc, sys_trap		! nofault little asi's
2275	  sub	%g0, 1, %g4
2276
2277#ifdef DEBUG_USER_TRAPTRACECTL
2278
2279.traptrace_freeze:
2280	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2281	TT_TRACE_L(trace_win)
2282	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2283	set	trap_freeze, %g1
2284	mov	1, %g2
2285	st	%g2, [%g1]
2286	FAST_TRAP_DONE
2287
2288.traptrace_unfreeze:
2289	set	trap_freeze, %g1
2290	st	%g0, [%g1]
2291	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2292	TT_TRACE_L(trace_win)
2293	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2294	FAST_TRAP_DONE
2295
2296#endif /* DEBUG_USER_TRAPTRACECTL */
2297
2298.getcc:
2299	CPU_ADDR(%g1, %g2)
2300	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2301	rdpr	%tstate, %g3			! get tstate
2302	srlx	%g3, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2303	set	PSR_ICC, %g2
2304	and	%o0, %g2, %o0			! mask out the rest
2305	srl	%o0, PSR_ICC_SHIFT, %o0		! right justify
2306	wrpr	%g0, 0, %gl
2307	mov	%o0, %g1			! move ccr to normal %g1
2308	wrpr	%g0, 1, %gl
2309	! cannot assume globals retained their values after increasing %gl
2310	CPU_ADDR(%g1, %g2)
2311	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2312	FAST_TRAP_DONE
2313
2314.setcc:
2315	CPU_ADDR(%g1, %g2)
2316	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2317	wrpr	%g0, 0, %gl
2318	mov	%g1, %o0
2319	wrpr	%g0, 1, %gl
2320	! cannot assume globals retained their values after increasing %gl
2321	CPU_ADDR(%g1, %g2)
2322	sll	%o0, PSR_ICC_SHIFT, %g2
2323	set	PSR_ICC, %g3
2324	and	%g2, %g3, %g2			! mask out rest
2325	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g2
2326	rdpr	%tstate, %g3			! get tstate
2327	srl	%g3, 0, %g3			! clear upper word
2328	or	%g3, %g2, %g3			! or in new bits
2329	wrpr	%g3, %tstate
2330	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2331	FAST_TRAP_DONE
2332
2333/*
2334 * getpsr(void)
2335 * Note that the xcc part of the ccr is not provided.
2336 * The V8 code shows why the V9 trap is not faster:
2337 * #define GETPSR_TRAP() \
2338 *      mov %psr, %i0; jmp %l2; rett %l2+4; nop;
2339 */
2340
2341	.type	.getpsr, #function
2342.getpsr:
2343	rdpr	%tstate, %g1			! get tstate
2344	srlx	%g1, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2345	set	PSR_ICC, %g2
2346	and	%o0, %g2, %o0			! mask out the rest
2347
2348	rd	%fprs, %g1			! get fprs
2349	and	%g1, FPRS_FEF, %g2		! mask out dirty upper/lower
2350	sllx	%g2, PSR_FPRS_FEF_SHIFT, %g2	! shift fef to V8 psr.ef
2351	or	%o0, %g2, %o0			! or result into psr.ef
2352
2353	set	V9_PSR_IMPLVER, %g2		! SI assigned impl/ver: 0xef
2354	or	%o0, %g2, %o0			! or psr.impl/ver
2355	FAST_TRAP_DONE
2356	SET_SIZE(.getpsr)
2357
2358/*
2359 * setpsr(newpsr)
2360 * Note that there is no support for ccr.xcc in the V9 code.
2361 */
2362
2363	.type	.setpsr, #function
2364.setpsr:
2365	rdpr	%tstate, %g1			! get tstate
2366!	setx	TSTATE_V8_UBITS, %g2
2367	or 	%g0, CCR_ICC, %g3
2368	sllx	%g3, TSTATE_CCR_SHIFT, %g2
2369
2370	andn	%g1, %g2, %g1			! zero current user bits
2371	set	PSR_ICC, %g2
2372	and	%g2, %o0, %g2			! clear all but psr.icc bits
2373	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g3	! shift to tstate.ccr.icc
2374	wrpr	%g1, %g3, %tstate		! write tstate
2375
2376	set	PSR_EF, %g2
2377	and	%g2, %o0, %g2			! clear all but fp enable bit
2378	srlx	%g2, PSR_FPRS_FEF_SHIFT, %g4	! shift ef to V9 fprs.fef
2379	wr	%g0, %g4, %fprs			! write fprs
2380
2381	CPU_ADDR(%g1, %g2)			! load CPU struct addr to %g1
2382	ldn	[%g1 + CPU_THREAD], %g2		! load thread pointer
2383	ldn	[%g2 + T_LWP], %g3		! load klwp pointer
2384	ldn	[%g3 + LWP_FPU], %g2		! get lwp_fpu pointer
2385	stuw	%g4, [%g2 + FPU_FPRS]		! write fef value to fpu_fprs
2386	srlx	%g4, 2, %g4			! shift fef value to bit 0
2387	stub	%g4, [%g2 + FPU_EN]		! write fef value to fpu_en
2388	FAST_TRAP_DONE
2389	SET_SIZE(.setpsr)
2390
2391/*
2392 * getlgrp
2393 * get home lgrpid on which the calling thread is currently executing.
2394 */
2395	.type	.getlgrp, #function
2396.getlgrp:
2397	! Thanks for the incredibly helpful comments
2398	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2399	ld	[%g1 + CPU_ID], %o0	! load cpu_id
2400	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2401	ldn	[%g2 + T_LPL], %g2	! load lpl pointer
2402	ld	[%g2 + LPL_LGRPID], %g1	! load lpl_lgrpid
2403	sra	%g1, 0, %o1
2404	FAST_TRAP_DONE
2405	SET_SIZE(.getlgrp)
2406
2407/*
2408 * Entry for old 4.x trap (trap 0).
2409 */
2410	ENTRY_NP(syscall_trap_4x)
2411	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2412	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2413	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2414	ld	[%g2 + PCB_TRAP0], %g2	! lwp->lwp_pcb.pcb_trap0addr
2415	brz,pn	%g2, 1f			! has it been set?
2416	st	%l0, [%g1 + CPU_TMP1]	! delay - save some locals
2417	st	%l1, [%g1 + CPU_TMP2]
2418	rdpr	%tnpc, %l1		! save old tnpc
2419	wrpr	%g0, %g2, %tnpc		! setup tnpc
2420
2421	mov	%g1, %l0		! save CPU struct addr
2422	wrpr	%g0, 0, %gl
2423	mov	%l1, %g6		! pass tnpc to user code in %g6
2424	wrpr	%g0, 1, %gl
2425	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2426	ld	[%l0 + CPU_TMP1], %l0
2427	FAST_TRAP_DONE_CHK_INTR
24281:
2429	!
2430	! check for old syscall mmap which is the only different one which
2431	! must be the same.  Others are handled in the compatibility library.
2432	!
2433	mov	%g1, %l0		! save CPU struct addr
2434	wrpr	%g0, 0, %gl
2435	cmp	%g1, OSYS_mmap		! compare to old 4.x mmap
2436	movz	%icc, SYS_mmap, %g1
2437	wrpr	%g0, 1, %gl
2438	ld	[%l0 + CPU_TMP1], %l0
2439	SYSCALL(syscall_trap32)
2440	SET_SIZE(syscall_trap_4x)
2441
2442/*
2443 * Handler for software trap 9.
2444 * Set trap0 emulation address for old 4.x system call trap.
2445 * XXX - this should be a system call.
2446 */
2447	ENTRY_NP(set_trap0_addr)
2448	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2449	st	%l0, [%g1 + CPU_TMP1]	! save some locals
2450	st	%l1, [%g1 + CPU_TMP2]
2451	mov	%g1, %l0	! preserve CPU addr
2452	wrpr	%g0, 0, %gl
2453	mov	%g1, %l1
2454	wrpr	%g0, 1, %gl
2455	! cannot assume globals retained their values after increasing %gl
2456	ldn	[%l0 + CPU_THREAD], %g2	! load thread pointer
2457	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2458	andn	%l1, 3, %l1		! force alignment
2459	st	%l1, [%g2 + PCB_TRAP0]	! lwp->lwp_pcb.pcb_trap0addr
2460	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2461	ld	[%l0 + CPU_TMP1], %l0
2462	FAST_TRAP_DONE
2463	SET_SIZE(set_trap0_addr)
2464
2465/*
2466 * mmu_trap_tl1
2467 * trap handler for unexpected mmu traps.
2468 * simply checks if the trap was a user lddf/stdf alignment trap, in which
2469 * case we go to fpu_trap or a user trap from the window handler, in which
2470 * case we go save the state on the pcb.  Otherwise, we go to ptl1_panic.
2471 */
2472	.type	mmu_trap_tl1, #function
2473mmu_trap_tl1:
2474#ifdef	TRAPTRACE
2475	TRACE_PTR(%g5, %g6)
2476	GET_TRACE_TICK(%g6)
2477	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2478	TRACE_SAVE_TL_GL_REGS(%g5, %g6)
2479	rdpr	%tt, %g6
2480	stha	%g6, [%g5 + TRAP_ENT_TT]%asi
2481	rdpr	%tstate, %g6
2482	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi
2483	stna	%sp, [%g5 + TRAP_ENT_SP]%asi
2484	stna	%g0, [%g5 + TRAP_ENT_TR]%asi
2485	rdpr	%tpc, %g6
2486	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2487	MMU_FAULT_STATUS_AREA(%g6)
2488	ldx	[%g6 + MMFSA_D_ADDR], %g6
2489	stna	%g6, [%g5 + TRAP_ENT_F1]%asi !  MMU fault address
2490	CPU_PADDR(%g7, %g6);
2491	add	%g7, CPU_TL1_HDLR, %g7
2492	lda	[%g7]ASI_MEM, %g6
2493	stna	%g6, [%g5 + TRAP_ENT_F2]%asi
2494	MMU_FAULT_STATUS_AREA(%g6)
2495	ldx	[%g6 + MMFSA_D_TYPE], %g7 ! XXXQ should be a MMFSA_F_ constant?
2496	ldx	[%g6 + MMFSA_D_CTX], %g6
2497	sllx	%g6, SFSR_CTX_SHIFT, %g6
2498	or	%g6, %g7, %g6
2499	stna	%g6, [%g5 + TRAP_ENT_F3]%asi ! MMU context/type
2500	set	0xdeadbeef, %g6
2501	stna	%g6, [%g5 + TRAP_ENT_F4]%asi
2502	TRACE_NEXT(%g5, %g6, %g7)
2503#endif /* TRAPTRACE */
2504	CPU_PADDR(%g7, %g6);
2505	add     %g7, CPU_TL1_HDLR, %g7		! %g7 = &cpu_m.tl1_hdlr (PA)
2506	lda	[%g7]ASI_MEM, %g6
2507	brz,a,pt %g6, 1f
2508	  nop
2509	sta     %g0, [%g7]ASI_MEM
2510	! XXXQ need to setup registers for sfmmu_mmu_trap?
2511	ba,a,pt	%xcc, sfmmu_mmu_trap		! handle page faults
25121:
2513	rdpr	%tpc, %g7
2514	/* in user_rtt? */
2515	set	rtt_fill_start, %g6
2516	cmp	%g7, %g6
2517	blu,pn	%xcc, 6f
2518	 .empty
2519	set	rtt_fill_end, %g6
2520	cmp	%g7, %g6
2521	bgeu,pn %xcc, 6f
2522	 nop
2523	set	fault_rtt_fn1, %g7
2524	ba,a	7f
25256:
2526	! check to see if the trap pc is in a window spill/fill handling
2527	rdpr	%tpc, %g7
2528	/* tpc should be in the trap table */
2529	set	trap_table, %g6
2530	cmp	%g7, %g6
2531	blu,a,pn %xcc, ptl1_panic
2532	  mov	PTL1_BAD_MMUTRAP, %g1
2533	set	etrap_table, %g6
2534	cmp	%g7, %g6
2535	bgeu,a,pn %xcc, ptl1_panic
2536	  mov	PTL1_BAD_MMUTRAP, %g1
2537	! pc is inside the trap table, convert to trap type
2538	srl	%g7, 5, %g6		! XXXQ need #define
2539	and	%g6, 0x1ff, %g6		! XXXQ need #define
2540	! and check for a window trap type
2541	and	%g6, WTRAP_TTMASK, %g6
2542	cmp	%g6, WTRAP_TYPE
2543	bne,a,pn %xcc, ptl1_panic
2544	  mov	PTL1_BAD_MMUTRAP, %g1
2545	andn	%g7, WTRAP_ALIGN, %g7	/* 128 byte aligned */
2546	add	%g7, WTRAP_FAULTOFF, %g7
2547
25487:
2549	! Arguments are passed in the global set active after the
2550	! 'done' instruction. Before switching sets, must save
2551	! the calculated next pc
2552	wrpr	%g0, %g7, %tnpc
2553	wrpr	%g0, 1, %gl
2554	rdpr	%tt, %g5
2555	MMU_FAULT_STATUS_AREA(%g7)
2556	cmp	%g5, T_ALIGNMENT
2557	be,pn	%xcc, 1f
2558	ldx	[%g7 + MMFSA_D_ADDR], %g6
2559	ldx	[%g7 + MMFSA_D_CTX], %g7
2560	srlx	%g6, MMU_PAGESHIFT, %g6		/* align address */
2561	sllx	%g6, MMU_PAGESHIFT, %g6
2562	or	%g6, %g7, %g6			/* TAG_ACCESS */
25631:
2564	done
2565	SET_SIZE(mmu_trap_tl1)
2566
2567/*
2568 * Several traps use kmdb_trap and kmdb_trap_tl1 as their handlers.  These
2569 * traps are valid only when kmdb is loaded.  When the debugger is active,
2570 * the code below is rewritten to transfer control to the appropriate
2571 * debugger entry points.
2572 */
2573	.global	kmdb_trap
2574	.align	8
2575kmdb_trap:
2576	ba,a	trap_table0
2577	jmp	%g1 + 0
2578	nop
2579
2580	.global	kmdb_trap_tl1
2581	.align	8
2582kmdb_trap_tl1:
2583	ba,a	trap_table0
2584	jmp	%g1 + 0
2585	nop
2586
2587/*
2588 * This entry is copied from OBP's trap table during boot.
2589 */
2590	.global	obp_bpt
2591	.align	8
2592obp_bpt:
2593	NOT
2594
2595
2596
2597#ifdef	TRAPTRACE
2598/*
2599 * TRAPTRACE support.
2600 * labels here are branched to with "rd %pc, %g7" in the delay slot.
2601 * Return is done by "jmp %g7 + 4".
2602 */
2603
2604trace_dmmu:
2605	TRACE_PTR(%g3, %g6)
2606	GET_TRACE_TICK(%g6)
2607	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2608	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2609	rdpr	%tt, %g6
2610	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2611	rdpr	%tstate, %g6
2612	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2613	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2614	rdpr	%tpc, %g6
2615	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2616	MMU_FAULT_STATUS_AREA(%g6)
2617	ldx	[%g6 + MMFSA_D_ADDR], %g4
2618	stxa	%g4, [%g3 + TRAP_ENT_TR]%asi
2619	ldx	[%g6 + MMFSA_D_CTX], %g4
2620	stxa	%g4, [%g3 + TRAP_ENT_F1]%asi
2621	ldx	[%g6 + MMFSA_D_TYPE], %g4
2622	stxa	%g4, [%g3 + TRAP_ENT_F2]%asi
2623	stxa	%g6, [%g3 + TRAP_ENT_F3]%asi
2624	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2625	TRACE_NEXT(%g3, %g4, %g5)
2626	jmp	%g7 + 4
2627	nop
2628
2629trace_immu:
2630	TRACE_PTR(%g3, %g6)
2631	GET_TRACE_TICK(%g6)
2632	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2633	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2634	rdpr	%tt, %g6
2635	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2636	rdpr	%tstate, %g6
2637	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2638	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2639	rdpr	%tpc, %g6
2640	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2641	MMU_FAULT_STATUS_AREA(%g6)
2642	ldx	[%g6 + MMFSA_I_ADDR], %g4
2643	stxa	%g4, [%g3 + TRAP_ENT_TR]%asi
2644	ldx	[%g6 + MMFSA_I_CTX], %g4
2645	stxa	%g4, [%g3 + TRAP_ENT_F1]%asi
2646	ldx	[%g6 + MMFSA_I_TYPE], %g4
2647	stxa	%g4, [%g3 + TRAP_ENT_F2]%asi
2648	stxa	%g6, [%g3 + TRAP_ENT_F3]%asi
2649	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2650	TRACE_NEXT(%g3, %g4, %g5)
2651	jmp	%g7 + 4
2652	nop
2653
2654trace_gen:
2655	TRACE_PTR(%g3, %g6)
2656	GET_TRACE_TICK(%g6)
2657	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2658	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2659	rdpr	%tt, %g6
2660	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2661	rdpr	%tstate, %g6
2662	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2663	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2664	rdpr	%tpc, %g6
2665	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2666	stna	%g0, [%g3 + TRAP_ENT_TR]%asi
2667	stna	%g0, [%g3 + TRAP_ENT_F1]%asi
2668	stna	%g0, [%g3 + TRAP_ENT_F2]%asi
2669	stna	%g0, [%g3 + TRAP_ENT_F3]%asi
2670	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2671	TRACE_NEXT(%g3, %g4, %g5)
2672	jmp	%g7 + 4
2673	nop
2674
2675trace_win:
2676	TRACE_WIN_INFO(0, %l0, %l1, %l2)
2677	! Keep the locals as clean as possible, caller cleans %l4
2678	clr	%l2
2679	clr	%l1
2680	jmp	%l4 + 4
2681	  clr	%l0
2682
2683/*
2684 * Trace a tsb hit
2685 * g1 = tsbe pointer (in/clobbered)
2686 * g2 = tag access register (in)
2687 * g3 - g4 = scratch (clobbered)
2688 * g5 = tsbe data (in)
2689 * g6 = scratch (clobbered)
2690 * g7 = pc we jumped here from (in)
2691 */
2692
2693	! Do not disturb %g5, it will be used after the trace
2694	ALTENTRY(trace_tsbhit)
2695	TRACE_TSBHIT(0)
2696	jmp	%g7 + 4
2697	nop
2698
2699/*
2700 * Trace a TSB miss
2701 *
2702 * g1 = tsb8k pointer (in)
2703 * g2 = tag access register (in)
2704 * g3 = tsb4m pointer (in)
2705 * g4 = tsbe tag (in/clobbered)
2706 * g5 - g6 = scratch (clobbered)
2707 * g7 = pc we jumped here from (in)
2708 */
2709	.global	trace_tsbmiss
2710trace_tsbmiss:
2711	membar	#Sync
2712	sethi	%hi(FLUSH_ADDR), %g6
2713	flush	%g6
2714	TRACE_PTR(%g5, %g6)
2715	GET_TRACE_TICK(%g6)
2716	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2717	stna	%g2, [%g5 + TRAP_ENT_SP]%asi		! tag access
2718	stna	%g4, [%g5 + TRAP_ENT_F1]%asi		! XXX? tsb tag
2719	rdpr	%tnpc, %g6
2720	stna	%g6, [%g5 + TRAP_ENT_F2]%asi
2721	stna	%g1, [%g5 + TRAP_ENT_F3]%asi		! tsb8k pointer
2722	srlx	%g1, 32, %g6
2723	stna	%g6, [%g5 + TRAP_ENT_F4]%asi		! huh?
2724	rdpr	%tpc, %g6
2725	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2726	TRACE_SAVE_TL_GL_REGS(%g5, %g6)
2727	rdpr	%tt, %g6
2728	or	%g6, TT_MMU_MISS, %g4
2729	stha	%g4, [%g5 + TRAP_ENT_TT]%asi
2730	mov	MMFSA_D_ADDR, %g4
2731	cmp	%g6, FAST_IMMU_MISS_TT
2732	move	%xcc, MMFSA_I_ADDR, %g4
2733	cmp	%g6, T_INSTR_MMU_MISS
2734	move	%xcc, MMFSA_I_ADDR, %g4
2735	MMU_FAULT_STATUS_AREA(%g6)
2736	ldx	[%g6 + %g4], %g6
2737	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi	! tag target
2738	stna	%g3, [%g5 + TRAP_ENT_TR]%asi		! tsb4m pointer
2739	TRACE_NEXT(%g5, %g4, %g6)
2740	jmp	%g7 + 4
2741	nop
2742
2743/*
2744 * g2 = tag access register (in)
2745 * g3 = ctx number (in)
2746 */
2747trace_dataprot:
2748	membar	#Sync
2749	sethi	%hi(FLUSH_ADDR), %g6
2750	flush	%g6
2751	TRACE_PTR(%g1, %g6)
2752	GET_TRACE_TICK(%g6)
2753	stxa	%g6, [%g1 + TRAP_ENT_TICK]%asi
2754	rdpr	%tpc, %g6
2755	stna	%g6, [%g1 + TRAP_ENT_TPC]%asi
2756	rdpr	%tstate, %g6
2757	stxa	%g6, [%g1 + TRAP_ENT_TSTATE]%asi
2758	stna	%g2, [%g1 + TRAP_ENT_SP]%asi		! tag access reg
2759	stna	%g0, [%g1 + TRAP_ENT_TR]%asi
2760	stna	%g0, [%g1 + TRAP_ENT_F1]%asi
2761	stna	%g0, [%g1 + TRAP_ENT_F2]%asi
2762	stna	%g0, [%g1 + TRAP_ENT_F3]%asi
2763	stna	%g0, [%g1 + TRAP_ENT_F4]%asi
2764	TRACE_SAVE_TL_GL_REGS(%g1, %g6)
2765	rdpr	%tt, %g6
2766	stha	%g6, [%g1 + TRAP_ENT_TT]%asi
2767	TRACE_NEXT(%g1, %g4, %g5)
2768	jmp	%g7 + 4
2769	nop
2770
2771#endif /* TRAPTRACE */
2772
2773/*
2774 * synthesize for trap(): SFAR in %g2, SFSR in %g3
2775 */
2776	.type	.dmmu_exc_lddf_not_aligned, #function
2777.dmmu_exc_lddf_not_aligned:
2778	MMU_FAULT_STATUS_AREA(%g3)
2779	ldx	[%g3 + MMFSA_D_ADDR], %g2
2780	/* Fault type not available in MMU fault status area */
2781	mov	MMFSA_F_UNALIGN, %g1
2782	ldx	[%g3 + MMFSA_D_CTX], %g3
2783	sllx	%g3, SFSR_CTX_SHIFT, %g3
2784	btst	1, %sp
2785	bnz,pt	%xcc, .lddf_exception_not_aligned
2786	or	%g3, %g1, %g3			/* SFSR */
2787	ba,a,pt	%xcc, .mmu_exception_not_aligned
2788	SET_SIZE(.dmmu_exc_lddf_not_aligned)
2789
2790/*
2791 * synthesize for trap(): SFAR in %g2, SFSR in %g3
2792 */
2793	.type	.dmmu_exc_stdf_not_aligned, #function
2794.dmmu_exc_stdf_not_aligned:
2795	MMU_FAULT_STATUS_AREA(%g3)
2796	ldx	[%g3 + MMFSA_D_ADDR], %g2
2797	/* Fault type not available in MMU fault status area */
2798	mov	MMFSA_F_UNALIGN, %g1
2799	ldx	[%g3 + MMFSA_D_CTX], %g3
2800	sllx	%g3, SFSR_CTX_SHIFT, %g3
2801	btst	1, %sp
2802	bnz,pt	%xcc, .stdf_exception_not_aligned
2803	or	%g3, %g1, %g3			/* SFSR */
2804	ba,a,pt	%xcc, .mmu_exception_not_aligned
2805	SET_SIZE(.dmmu_exc_stdf_not_aligned)
2806
2807	.type	.dmmu_exception, #function
2808.dmmu_exception:
2809	MMU_FAULT_STATUS_AREA(%g3)
2810	ldx	[%g3 + MMFSA_D_ADDR], %g2
2811	ldx	[%g3 + MMFSA_D_TYPE], %g1
2812	ldx	[%g3 + MMFSA_D_CTX], %g3
2813	srlx	%g2, MMU_PAGESHIFT, %g2		/* align address */
2814	sllx	%g2, MMU_PAGESHIFT, %g2
2815	or	%g2, %g3, %g2			/* TAG_ACCESS */
2816	sllx	%g3, SFSR_CTX_SHIFT, %g3
2817	or	%g3, %g1, %g3			/* SFSR */
2818	ba,pt	%xcc, .mmu_exception_end
2819	mov	T_DATA_EXCEPTION, %g1
2820	SET_SIZE(.dmmu_exception)
2821/*
2822 * expects offset into tsbmiss area in %g1 and return pc in %g7
2823 */
2824stat_mmu:
2825	CPU_INDEX(%g5, %g6)
2826	sethi	%hi(tsbmiss_area), %g6
2827	sllx	%g5, TSBMISS_SHIFT, %g5
2828	or	%g6, %lo(tsbmiss_area), %g6
2829	add	%g6, %g5, %g6		/* g6 = tsbmiss area */
2830	ld	[%g6 + %g1], %g5
2831	add	%g5, 1, %g5
2832	jmp	%g7 + 4
2833	st	%g5, [%g6 + %g1]
2834
2835
2836/*
2837 * fast_trap_done, fast_trap_done_chk_intr:
2838 *
2839 * Due to the design of UltraSPARC pipeline, pending interrupts are not
2840 * taken immediately after a RETRY or DONE instruction which causes IE to
2841 * go from 0 to 1. Instead, the instruction at %tpc or %tnpc is allowed
2842 * to execute first before taking any interrupts. If that instruction
2843 * results in other traps, and if the corresponding trap handler runs
2844 * entirely at TL=1 with interrupts disabled, then pending interrupts
2845 * won't be taken until after yet another instruction following the %tpc
2846 * or %tnpc.
2847 *
2848 * A malicious user program can use this feature to block out interrupts
2849 * for extended durations, which can result in send_mondo_timeout kernel
2850 * panic.
2851 *
2852 * This problem is addressed by servicing any pending interrupts via
2853 * sys_trap before returning back to the user mode from a fast trap
2854 * handler. The "done" instruction within a fast trap handler, which
2855 * runs entirely at TL=1 with interrupts disabled, is replaced with the
2856 * FAST_TRAP_DONE macro, which branches control to this fast_trap_done
2857 * entry point.
2858 *
2859 * We check for any pending interrupts here and force a sys_trap to
2860 * service those interrupts, if any. To minimize overhead, pending
2861 * interrupts are checked if the %tpc happens to be at 16K boundary,
2862 * which allows a malicious program to execute at most 4K consecutive
2863 * instructions before we service any pending interrupts. If a worst
2864 * case fast trap handler takes about 2 usec, then interrupts will be
2865 * blocked for at most 8 msec, less than a clock tick.
2866 *
2867 * For the cases where we don't know if the %tpc will cross a 16K
2868 * boundary, we can't use the above optimization and always process
2869 * any pending interrupts via fast_frap_done_chk_intr entry point.
2870 *
2871 * Entry Conditions:
2872 * 	%pstate		am:0 priv:1 ie:0
2873 * 			globals are AG (not normal globals)
2874 */
2875
2876	.global	fast_trap_done, fast_trap_done_chk_intr
2877fast_trap_done:
2878	rdpr	%tpc, %g5
2879	sethi	%hi(0xffffc000), %g6	! 1's complement of 0x3fff
2880	andncc	%g5, %g6, %g0		! check lower 14 bits of %tpc
2881	bz,pn	%icc, 1f		! branch if zero (lower 32 bits only)
2882	nop
2883	done
2884
2885fast_trap_done_chk_intr:
28861:	rd	SOFTINT, %g6
2887	brnz,pn	%g6, 2f		! branch if any pending intr
2888	nop
2889	done
2890
28912:
2892	/*
2893	 * We get here if there are any pending interrupts.
2894	 * Adjust %tpc/%tnpc as we'll be resuming via "retry"
2895	 * instruction.
2896	 */
2897	rdpr	%tnpc, %g5
2898	wrpr	%g0, %g5, %tpc
2899	add	%g5, 4, %g5
2900	wrpr	%g0, %g5, %tnpc
2901
2902	/*
2903	 * Force a dummy sys_trap call so that interrupts can be serviced.
2904	 */
2905	set	fast_trap_dummy_call, %g1
2906	ba,pt	%xcc, sys_trap
2907	  mov	-1, %g4
2908
2909fast_trap_dummy_call:
2910	retl
2911	nop
2912
2913#endif	/* lint */
2914