xref: /titanic_41/usr/src/uts/sun4v/ml/trap_table.s (revision 414388d7cb2ee98771e2ac7c2338c460abd44304)
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#define	SFSR_CTX_SHIFT	16
831
832#define	IMMU_EXCEPTION							\
833	MMU_FAULT_STATUS_AREA(%g3)					;\
834	rdpr	%tpc, %g2						;\
835	ldx	[%g3 + MMFSA_I_TYPE], %g1				;\
836	ldx	[%g3 + MMFSA_I_CTX], %g3				;\
837	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
838	or	%g3, %g1, %g3						;\
839	ba,pt	%xcc, .mmu_exception_end				;\
840	mov	T_INSTR_EXCEPTION, %g1					;\
841	.align	32
842
843#define	DMMU_EXCEPTION							\
844	MMU_FAULT_STATUS_AREA(%g3)					;\
845	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
846	ldx	[%g3 + MMFSA_D_TYPE], %g1				;\
847	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
848	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
849	or	%g3, %g1, %g3						;\
850	ba,pt	%xcc, .mmu_exception_end				;\
851	mov	T_DATA_EXCEPTION, %g1					;\
852	.align	32
853
854#define	DMMU_EXC_AG_PRIV						\
855	MMU_FAULT_STATUS_AREA(%g3)					;\
856	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
857	/* Fault type not available in MMU fault status area */		;\
858	mov	MMFSA_F_PRVACT, %g1					;\
859	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
860	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
861	ba,pt	%xcc, .mmu_priv_exception				;\
862	or	%g3, %g1, %g3						;\
863	.align	32
864
865#define	DMMU_EXC_AG_NOT_ALIGNED						\
866	MMU_FAULT_STATUS_AREA(%g3)					;\
867	ldx	[%g3 + MMFSA_D_ADDR], %g2				;\
868	/* Fault type not available in MMU fault status area */		;\
869	mov	MMFSA_F_UNALIGN, %g1					;\
870	ldx	[%g3 + MMFSA_D_CTX], %g3				;\
871	sllx	%g3, SFSR_CTX_SHIFT, %g3				;\
872	ba,pt	%xcc, .mmu_exception_not_aligned			;\
873	or	%g3, %g1, %g3						;\
874	.align	32
875/*
876 * SPARC V9 IMPL. DEP. #109(1) and (2) and #110(1) and (2)
877 */
878
879#define	DMMU_EXC_LDDF_NOT_ALIGNED					\
880	ba,a,pt	%xcc, .dmmu_exc_lddf_not_aligned			;\
881	.align	32
882#define	DMMU_EXC_STDF_NOT_ALIGNED					\
883	ba,a,pt	%xcc, .dmmu_exc_stdf_not_aligned			;\
884	.align	32
885
886#if TAGACC_CTX_MASK != CTXREG_CTX_MASK
887#error "TAGACC_CTX_MASK != CTXREG_CTX_MASK"
888#endif
889
890#if defined(cscope)
891/*
892 * Define labels to direct cscope quickly to labels that
893 * are generated by macro expansion of DTLB_MISS().
894 */
895	.global	tt0_dtlbmiss
896tt0_dtlbmiss:
897	.global	tt1_dtlbmiss
898tt1_dtlbmiss:
899	nop
900#endif
901
902/*
903 * Data miss handler (must be exactly 32 instructions)
904 *
905 * This handler is invoked only if the hypervisor has been instructed
906 * not to do any TSB walk.
907 *
908 * Kernel and invalid context cases are handled by the sfmmu_kdtlb_miss
909 * handler.
910 *
911 * User TLB miss handling depends upon whether a user process has one or
912 * two TSBs. User TSB information (physical base and size code) is kept
913 * in two dedicated scratchpad registers. Absence of a user TSB (primarily
914 * second TSB) is indicated by a negative value (-1) in that register.
915 */
916
917#define	DTLB_MISS(table_name)						;\
918	.global	table_name/**/_dtlbmiss					;\
919table_name/**/_dtlbmiss:						;\
920	HAT_PERCPU_DBSTAT(TSBMISS_DTLBMISS) /* 3 instr ifdef DEBUG */	;\
921	MMU_FAULT_STATUS_AREA(%g7)					;\
922	ldx	[%g7 + MMFSA_D_ADDR], %g2	/* address */		;\
923	ldx	[%g7 + MMFSA_D_CTX], %g3	/* g3 = ctx */		;\
924	or	%g2, %g3, %g2			/* XXXQ temp */		;\
925	cmp	%g3, INVALID_CONTEXT					;\
926	ble,pn	%xcc, sfmmu_kdtlb_miss					;\
927	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
928	mov	SCRATCHPAD_UTSBREG2, %g1				;\
929	ldxa	[%g1]ASI_SCRATCHPAD, %g1	/* get 2nd tsbreg */	;\
930	brgez,pn %g1, sfmmu_udtlb_slowpath	/* brnach if 2 TSBs */	;\
931	  nop								;\
932	GET_1ST_TSBE_PTR(%g2, %g1, %g4, %g5)	/* 11 instr */		;\
933	ba,pt	%xcc, sfmmu_udtlb_fastpath	/* no 4M TSB, miss */	;\
934	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
935	.align 128
936
937
938#if defined(cscope)
939/*
940 * Define labels to direct cscope quickly to labels that
941 * are generated by macro expansion of ITLB_MISS().
942 */
943	.global	tt0_itlbmiss
944tt0_itlbmiss:
945	.global	tt1_itlbmiss
946tt1_itlbmiss:
947	nop
948#endif
949
950/*
951 * Instruction miss handler.
952 *
953 * This handler is invoked only if the hypervisor has been instructed
954 * not to do any TSB walk.
955 *
956 * ldda instructions will have their ASI patched
957 * by sfmmu_patch_ktsb at runtime.
958 * MUST be EXACTLY 32 instructions or we'll break.
959 */
960
961#define	ITLB_MISS(table_name)						 \
962	.global	table_name/**/_itlbmiss					;\
963table_name/**/_itlbmiss:						;\
964	HAT_PERCPU_DBSTAT(TSBMISS_ITLBMISS) /* 3 instr ifdef DEBUG */	;\
965	MMU_FAULT_STATUS_AREA(%g7)					;\
966	ldx	[%g7 + MMFSA_I_ADDR], %g2	/* g2 = address */	;\
967	ldx	[%g7 + MMFSA_I_CTX], %g3	/* g3 = ctx */		;\
968	or	%g2, %g3, %g2			/* XXXQ temp */		;\
969	cmp	%g3, INVALID_CONTEXT					;\
970	ble,pn	%xcc, sfmmu_kitlb_miss					;\
971	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
972	mov	SCRATCHPAD_UTSBREG2, %g1				;\
973	ldxa	[%g1]ASI_SCRATCHPAD, %g1	/* get 2nd tsbreg */	;\
974	brgez,pn %g1, sfmmu_uitlb_slowpath	/* branch if 2 TSBS */	;\
975	  nop								;\
976	GET_1ST_TSBE_PTR(%g2, %g1, %g4, %g5)	/* 11 instr */		;\
977	ba,pt	%xcc, sfmmu_uitlb_fastpath	/* no 4M TSB, miss */	;\
978	  srlx	%g2, TAG_VALO_SHIFT, %g7	/* g7 = tsb tag */	;\
979	.align 128
980
981#define	DTSB_MISS \
982	GOTO_TT(sfmmu_slow_dmmu_miss,trace_dmmu)
983
984#define	ITSB_MISS \
985	GOTO_TT(sfmmu_slow_immu_miss,trace_immu)
986
987/*
988 * This macro is the first level handler for fast protection faults.
989 * It first demaps the tlb entry which generated the fault and then
990 * attempts to set the modify bit on the hash.  It needs to be
991 * exactly 32 instructions.
992 */
993#define	DTLB_PROT							 \
994	MMU_FAULT_STATUS_AREA(%g7)					;\
995	ldx	[%g7 + MMFSA_D_ADDR], %g2	/* address */		;\
996	ldx	[%g7 + MMFSA_D_CTX], %g3	/* %g3 = ctx */		;\
997	or	%g2, %g3, %g2			/* XXXQ temp */		;\
998	/*								;\
999	 *   g2 = tag access register					;\
1000	 *   g3 = ctx number						;\
1001	 */								;\
1002	TT_TRACE(trace_dataprot)	/* 2 instr ifdef TRAPTRACE */	;\
1003					/* clobbers g1 and g6 XXXQ? */	;\
1004	brnz,pt %g3, sfmmu_uprot_trap		/* user trap */		;\
1005	  nop								;\
1006	ba,a,pt	%xcc, sfmmu_kprot_trap		/* kernel trap */	;\
1007	.align 128
1008
1009#define	DMMU_EXCEPTION_TL1						;\
1010	ba,a,pt	%xcc, mmu_trap_tl1					;\
1011	.align 32
1012
1013#define	MISALIGN_ADDR_TL1						;\
1014	ba,a,pt	%xcc, mmu_trap_tl1					;\
1015	.align 32
1016
1017/*
1018 * Trace a tsb hit
1019 * g1 = tsbe pointer (in/clobbered)
1020 * g2 = tag access register (in)
1021 * g3 - g4 = scratch (clobbered)
1022 * g5 = tsbe data (in)
1023 * g6 = scratch (clobbered)
1024 * g7 = pc we jumped here from (in)
1025 * ttextra = value to OR in to trap type (%tt) (in)
1026 */
1027#ifdef TRAPTRACE
1028#define TRACE_TSBHIT(ttextra)						 \
1029	membar	#Sync							;\
1030	sethi	%hi(FLUSH_ADDR), %g6					;\
1031	flush	%g6							;\
1032	TRACE_PTR(%g3, %g6)						;\
1033	GET_TRACE_TICK(%g6)						;\
1034	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi				;\
1035	stna	%g2, [%g3 + TRAP_ENT_SP]%asi	/* tag access */	;\
1036	stna	%g5, [%g3 + TRAP_ENT_F1]%asi	/* tsb data */		;\
1037	rdpr	%tnpc, %g6						;\
1038	stna	%g6, [%g3 + TRAP_ENT_F2]%asi				;\
1039	stna	%g1, [%g3 + TRAP_ENT_F3]%asi	/* tsb pointer */	;\
1040	stna	%g0, [%g3 + TRAP_ENT_F4]%asi				;\
1041	rdpr	%tpc, %g6						;\
1042	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi				;\
1043	TRACE_SAVE_TL_GL_REGS(%g3, %g6)					;\
1044	rdpr	%tt, %g6						;\
1045	or	%g6, (ttextra), %g1					;\
1046	stha	%g1, [%g3 + TRAP_ENT_TT]%asi				;\
1047	MMU_FAULT_STATUS_AREA(%g4)					;\
1048	ldx	[%g4 + MMFSA_I_ADDR], %g1				;\
1049	ldx	[%g4 + MMFSA_D_ADDR], %g4				;\
1050	cmp	%g6, FAST_IMMU_MISS_TT					;\
1051	move	%icc, %g1, %g4						;\
1052	cmp	%g6, T_INSTR_MMU_MISS					;\
1053	move	%icc, %g1, %g4						;\
1054	stxa	%g4, [%g3 + TRAP_ENT_TSTATE]%asi /* fault addr */	;\
1055	MMU_FAULT_STATUS_AREA(%g4)					;\
1056	ldx	[%g4 + MMFSA_I_CTX], %g1				;\
1057	ldx	[%g4 + MMFSA_D_CTX], %g4				;\
1058	cmp	%g6, FAST_IMMU_MISS_TT					;\
1059	move	%icc, %g1, %g4						;\
1060	cmp	%g6, T_INSTR_MMU_MISS					;\
1061	move	%icc, %g1, %g4						;\
1062	stna	%g4, [%g3 + TRAP_ENT_TR]%asi				;\
1063	TRACE_NEXT(%g3, %g4, %g6)
1064#else
1065#define TRACE_TSBHIT(ttextra)
1066#endif
1067
1068
1069#if defined(lint)
1070
1071struct scb	trap_table;
1072struct scb	scb;		/* trap_table/scb are the same object */
1073
1074#else /* lint */
1075
1076/*
1077 * =======================================================================
1078 *		SPARC V9 TRAP TABLE
1079 *
1080 * The trap table is divided into two halves: the first half is used when
1081 * taking traps when TL=0; the second half is used when taking traps from
1082 * TL>0. Note that handlers in the second half of the table might not be able
1083 * to make the same assumptions as handlers in the first half of the table.
1084 *
1085 * Worst case trap nesting so far:
1086 *
1087 *	at TL=0 client issues software trap requesting service
1088 *	at TL=1 nucleus wants a register window
1089 *	at TL=2 register window clean/spill/fill takes a TLB miss
1090 *	at TL=3 processing TLB miss
1091 *	at TL=4 handle asynchronous error
1092 *
1093 * Note that a trap from TL=4 to TL=5 places Spitfire in "RED mode".
1094 *
1095 * =======================================================================
1096 */
1097	.section ".text"
1098	.align	4
1099	.global trap_table, scb, trap_table0, trap_table1, etrap_table
1100	.type	trap_table, #function
1101	.type	trap_table0, #function
1102	.type	trap_table1, #function
1103	.type	scb, #function
1104trap_table:
1105scb:
1106trap_table0:
1107	/* hardware traps */
1108	NOT;				/* 000	reserved */
1109	RED;				/* 001	power on reset */
1110	RED;				/* 002	watchdog reset */
1111	RED;				/* 003	externally initiated reset */
1112	RED;				/* 004	software initiated reset */
1113	RED;				/* 005	red mode exception */
1114	NOT; NOT;			/* 006 - 007 reserved */
1115	IMMU_EXCEPTION;			/* 008	instruction access exception */
1116	ITSB_MISS;			/* 009	instruction access MMU miss */
1117 	NOT;				/* 00A  reserved */
1118	NOT; NOT4;			/* 00B - 00F reserved */
1119	ILLTRAP_INSTR;			/* 010	illegal instruction */
1120	TRAP(T_PRIV_INSTR);		/* 011	privileged opcode */
1121	NOT;				/* 012	unimplemented LDD */
1122	NOT;				/* 013	unimplemented STD */
1123	NOT4; NOT4; NOT4;		/* 014 - 01F reserved */
1124	FP_DISABLED_TRAP;		/* 020	fp disabled */
1125	FP_IEEE_TRAP;			/* 021	fp exception ieee 754 */
1126	FP_TRAP;			/* 022	fp exception other */
1127	TAG_OVERFLOW;			/* 023	tag overflow */
1128	CLEAN_WINDOW;			/* 024 - 027 clean window */
1129	DIV_BY_ZERO;			/* 028	division by zero */
1130	NOT;				/* 029	internal processor error */
1131	NOT; NOT; NOT4;			/* 02A - 02F reserved */
1132	DMMU_EXCEPTION;			/* 030	data access exception */
1133	DTSB_MISS;			/* 031	data access MMU miss */
1134	NOT;				/* 032  reserved */
1135	NOT;				/* 033	data access protection */
1136	DMMU_EXC_AG_NOT_ALIGNED;	/* 034	mem address not aligned */
1137	DMMU_EXC_LDDF_NOT_ALIGNED;	/* 035	LDDF mem address not aligned */
1138	DMMU_EXC_STDF_NOT_ALIGNED;	/* 036	STDF mem address not aligned */
1139	DMMU_EXC_AG_PRIV;		/* 037	privileged action */
1140	NOT;				/* 038	LDQF mem address not aligned */
1141	NOT;				/* 039	STQF mem address not aligned */
1142	NOT; NOT; NOT4;			/* 03A - 03F reserved */
1143	NOT;				/* 040	async data error */
1144	LEVEL_INTERRUPT(1);		/* 041	interrupt level 1 */
1145	LEVEL_INTERRUPT(2);		/* 042	interrupt level 2 */
1146	LEVEL_INTERRUPT(3);		/* 043	interrupt level 3 */
1147	LEVEL_INTERRUPT(4);		/* 044	interrupt level 4 */
1148	LEVEL_INTERRUPT(5);		/* 045	interrupt level 5 */
1149	LEVEL_INTERRUPT(6);		/* 046	interrupt level 6 */
1150	LEVEL_INTERRUPT(7);		/* 047	interrupt level 7 */
1151	LEVEL_INTERRUPT(8);		/* 048	interrupt level 8 */
1152	LEVEL_INTERRUPT(9);		/* 049	interrupt level 9 */
1153	LEVEL_INTERRUPT(10);		/* 04A	interrupt level 10 */
1154	LEVEL_INTERRUPT(11);		/* 04B	interrupt level 11 */
1155	LEVEL_INTERRUPT(12);		/* 04C	interrupt level 12 */
1156	LEVEL_INTERRUPT(13);		/* 04D	interrupt level 13 */
1157	LEVEL14_INTERRUPT;		/* 04E	interrupt level 14 */
1158	LEVEL_INTERRUPT(15);		/* 04F	interrupt level 15 */
1159	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F reserved */
1160	NOT;				/* 060	interrupt vector */
1161	GOTO(kmdb_trap);		/* 061	PA watchpoint */
1162	GOTO(kmdb_trap);		/* 062	VA watchpoint */
1163	NOT;				/* 063	reserved */
1164	ITLB_MISS(tt0);			/* 064	instruction access MMU miss */
1165	DTLB_MISS(tt0);			/* 068	data access MMU miss */
1166	DTLB_PROT;			/* 06C	data access protection */
1167	NOT;				/* 070  reserved */
1168	NOT;				/* 071  reserved */
1169	NOT;				/* 072  reserved */
1170	NOT;				/* 073  reserved */
1171	NOT4; NOT4			/* 074 - 07B reserved */
1172	CPU_MONDO;			/* 07C	cpu_mondo */
1173	DEV_MONDO;			/* 07D	dev_mondo */
1174	GOTO_TT(resumable_error, trace_gen);	/* 07E  resumable error */
1175	GOTO_TT(nonresumable_error, trace_gen);	/* 07F  non-reasumable error */
1176	NOT4;				/* 080	spill 0 normal */
1177	SPILL_32bit_asi(ASI_AIUP,sn0);	/* 084	spill 1 normal */
1178	SPILL_64bit_asi(ASI_AIUP,sn0);	/* 088	spill 2 normal */
1179	SPILL_32clean(ASI_AIUP,sn0);	/* 08C	spill 3 normal */
1180	SPILL_64clean(ASI_AIUP,sn0);	/* 090	spill 4 normal */
1181	SPILL_32bit(not);		/* 094	spill 5 normal */
1182	SPILL_64bit(not);		/* 098	spill 6 normal */
1183	SPILL_mixed;			/* 09C	spill 7 normal */
1184	NOT4;				/* 0A0	spill 0 other */
1185	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0A4	spill 1 other */
1186	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0A8	spill 2 other */
1187	SPILL_32bit_asi(ASI_AIUS,so0);	/* 0AC	spill 3 other */
1188	SPILL_64bit_asi(ASI_AIUS,so0);	/* 0B0	spill 4 other */
1189	NOT4;				/* 0B4	spill 5 other */
1190	NOT4;				/* 0B8	spill 6 other */
1191	NOT4;				/* 0BC	spill 7 other */
1192	NOT4;				/* 0C0	fill 0 normal */
1193	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0C4	fill 1 normal */
1194	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0C8	fill 2 normal */
1195	FILL_32bit_asi(ASI_AIUP,fn0);	/* 0CC	fill 3 normal */
1196	FILL_64bit_asi(ASI_AIUP,fn0);	/* 0D0	fill 4 normal */
1197	FILL_32bit(not);		/* 0D4	fill 5 normal */
1198	FILL_64bit(not);		/* 0D8	fill 6 normal */
1199	FILL_mixed;			/* 0DC	fill 7 normal */
1200	NOT4;				/* 0E0	fill 0 other */
1201	NOT4;				/* 0E4	fill 1 other */
1202	NOT4;				/* 0E8	fill 2 other */
1203	NOT4;				/* 0EC	fill 3 other */
1204	NOT4;				/* 0F0	fill 4 other */
1205	NOT4;				/* 0F4	fill 5 other */
1206	NOT4;				/* 0F8	fill 6 other */
1207	NOT4;				/* 0FC	fill 7 other */
1208	/* user traps */
1209	GOTO(syscall_trap_4x);		/* 100	old system call */
1210	TRAP(T_BREAKPOINT);		/* 101	user breakpoint */
1211	TRAP(T_DIV0);			/* 102	user divide by zero */
1212	GOTO(.flushw);			/* 103	flush windows */
1213	GOTO(.clean_windows);		/* 104	clean windows */
1214	BAD;				/* 105	range check ?? */
1215	GOTO(.fix_alignment);		/* 106	do unaligned references */
1216	BAD;				/* 107	unused */
1217	SYSCALL(syscall_trap32);	/* 108	ILP32 system call on LP64 */
1218	GOTO(set_trap0_addr);		/* 109	set trap0 address */
1219	BAD; BAD; BAD4;			/* 10A - 10F unused */
1220	TRP4; TRP4; TRP4; TRP4;		/* 110 - 11F V9 user trap handlers */
1221	GOTO(.getcc);			/* 120	get condition codes */
1222	GOTO(.setcc);			/* 121	set condition codes */
1223	GOTO(.getpsr);			/* 122	get psr */
1224	GOTO(.setpsr);			/* 123	set psr (some fields) */
1225	GOTO(get_timestamp);		/* 124	get timestamp */
1226	GOTO(get_virtime);		/* 125	get lwp virtual time */
1227	PRIV(self_xcall);		/* 126	self xcall */
1228	GOTO(get_hrestime);		/* 127	get hrestime */
1229	BAD;				/* 128	ST_SETV9STACK */
1230	GOTO(.getlgrp);			/* 129  get lgrpid */
1231	BAD; BAD; BAD4;			/* 12A - 12F unused */
1232	BAD4; BAD4; 			/* 130 - 137 unused */
1233	DTRACE_PID;			/* 138  dtrace pid tracing provider */
1234	DTRACE_FASTTRAP;		/* 139  dtrace fasttrap provider */
1235	DTRACE_RETURN;			/* 13A	dtrace pid return probe */
1236	BAD; BAD4;			/* 13B - 13F unused */
1237	SYSCALL(syscall_trap)		/* 140  LP64 system call */
1238	BAD;				/* 141  unused */
1239#ifdef DEBUG_USER_TRAPTRACECTL
1240	GOTO(.traptrace_freeze);	/* 142  freeze traptrace */
1241	GOTO(.traptrace_unfreeze);	/* 143  unfreeze traptrace */
1242#else
1243	BAD; BAD;			/* 142 - 143 unused */
1244#endif
1245	BAD4; BAD4; BAD4;		/* 144 - 14F unused */
1246	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1247	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1248	BAD;				/* 170 - unused */
1249	BAD;				/* 171 - unused */
1250	BAD; BAD;			/* 172 - 173 unused */
1251	BAD4; BAD4;			/* 174 - 17B unused */
1252#ifdef	PTL1_PANIC_DEBUG
1253	mov PTL1_BAD_DEBUG, %g1; GOTO(ptl1_panic);
1254					/* 17C	test ptl1_panic */
1255#else
1256	BAD;				/* 17C  unused */
1257#endif	/* PTL1_PANIC_DEBUG */
1258	PRIV(kmdb_trap);		/* 17D	kmdb enter (L1-A) */
1259	PRIV(kmdb_trap);		/* 17E	kmdb breakpoint */
1260	PRIV(obp_bpt);			/* 17F	obp breakpoint */
1261	/* reserved */
1262	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1263	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1264	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1265	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1266	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1267	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1268	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1269	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1270	.size	trap_table0, (.-trap_table0)
1271trap_table1:
1272	NOT4; NOT4;			/* 000 - 007 unused */
1273	NOT;				/* 008	instruction access exception */
1274	ITSB_MISS;			/* 009	instruction access MMU miss */
1275 	NOT;				/* 00A  reserved */
1276	NOT; NOT4;			/* 00B - 00F unused */
1277	NOT4; NOT4; NOT4; NOT4;		/* 010 - 01F unused */
1278	NOT4;				/* 020 - 023 unused */
1279	CLEAN_WINDOW;			/* 024 - 027 clean window */
1280	NOT4; NOT4;			/* 028 - 02F unused */
1281	DMMU_EXCEPTION_TL1;		/* 030 	data access exception */
1282	DTSB_MISS;			/* 031  data access MMU miss */
1283	NOT;				/* 032  reserved */
1284	NOT;				/* 033	unused */
1285	MISALIGN_ADDR_TL1;		/* 034	mem address not aligned */
1286	NOT; NOT; NOT; NOT4; NOT4	/* 035 - 03F unused */
1287	NOT4; NOT4; NOT4; NOT4;		/* 040 - 04F unused */
1288	NOT4; NOT4; NOT4; NOT4;		/* 050 - 05F unused */
1289	NOT;				/* 060	unused */
1290	GOTO(kmdb_trap_tl1);		/* 061	PA watchpoint */
1291	GOTO(kmdb_trap_tl1);		/* 062	VA watchpoint */
1292	NOT;				/* 063	reserved */
1293	ITLB_MISS(tt1);			/* 064	instruction access MMU miss */
1294	DTLB_MISS(tt1);			/* 068	data access MMU miss */
1295	DTLB_PROT;			/* 06C	data access protection */
1296	NOT;				/* 070  reserved */
1297	NOT;				/* 071  reserved */
1298	NOT;				/* 072  reserved */
1299	NOT;				/* 073  reserved */
1300	NOT4; NOT4; NOT4;		/* 074 - 07F reserved */
1301	NOTP4;				/* 080	spill 0 normal */
1302	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 084	spill 1 normal */
1303	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 088	spill 2 normal */
1304	SPILL_32bit_tt1(ASI_AIUP,sn1);	/* 08C	spill 3 normal */
1305	SPILL_64bit_tt1(ASI_AIUP,sn1);	/* 090	spill 4 normal */
1306	NOTP4;				/* 094	spill 5 normal */
1307	SPILL_64bit_ktt1(sk);		/* 098	spill 6 normal */
1308	SPILL_mixed_ktt1(sk);		/* 09C	spill 7 normal */
1309	NOTP4;				/* 0A0	spill 0 other */
1310	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0A4  spill 1 other */
1311	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0A8	spill 2 other */
1312	SPILL_32bit_tt1(ASI_AIUS,so1);	/* 0AC	spill 3 other */
1313	SPILL_64bit_tt1(ASI_AIUS,so1);	/* 0B0  spill 4 other */
1314	NOTP4;				/* 0B4  spill 5 other */
1315	NOTP4;				/* 0B8  spill 6 other */
1316	NOTP4;				/* 0BC  spill 7 other */
1317	NOT4;				/* 0C0	fill 0 normal */
1318	NOT4;				/* 0C4	fill 1 normal */
1319	NOT4;				/* 0C8	fill 2 normal */
1320	NOT4;				/* 0CC	fill 3 normal */
1321	NOT4;				/* 0D0	fill 4 normal */
1322	NOT4;				/* 0D4	fill 5 normal */
1323	NOT4;				/* 0D8	fill 6 normal */
1324	NOT4;				/* 0DC	fill 7 normal */
1325	NOT4; NOT4; NOT4; NOT4;		/* 0E0 - 0EF unused */
1326	NOT4; NOT4; NOT4; NOT4;		/* 0F0 - 0FF unused */
1327/*
1328 * Code running at TL>0 does not use soft traps, so
1329 * we can truncate the table here.
1330 * However:
1331 * sun4v uses (hypervisor) ta instructions at TL > 0, so
1332 * provide a safety net for now.
1333 */
1334	/* soft traps */
1335	BAD4; BAD4; BAD4; BAD4;		/* 100 - 10F unused */
1336	BAD4; BAD4; BAD4; BAD4;		/* 110 - 11F unused */
1337	BAD4; BAD4; BAD4; BAD4;		/* 120 - 12F unused */
1338	BAD4; BAD4; BAD4; BAD4;		/* 130 - 13F unused */
1339	BAD4; BAD4; BAD4; BAD4;		/* 140 - 14F unused */
1340	BAD4; BAD4; BAD4; BAD4;		/* 150 - 15F unused */
1341	BAD4; BAD4; BAD4; BAD4;		/* 160 - 16F unused */
1342	BAD4; BAD4; BAD4; BAD4;		/* 170 - 17F unused */
1343	/* reserved */
1344	NOT4; NOT4; NOT4; NOT4;		/* 180 - 18F reserved */
1345	NOT4; NOT4; NOT4; NOT4;		/* 190 - 19F reserved */
1346	NOT4; NOT4; NOT4; NOT4;		/* 1A0 - 1AF reserved */
1347	NOT4; NOT4; NOT4; NOT4;		/* 1B0 - 1BF reserved */
1348	NOT4; NOT4; NOT4; NOT4;		/* 1C0 - 1CF reserved */
1349	NOT4; NOT4; NOT4; NOT4;		/* 1D0 - 1DF reserved */
1350	NOT4; NOT4; NOT4; NOT4;		/* 1E0 - 1EF reserved */
1351	NOT4; NOT4; NOT4; NOT4;		/* 1F0 - 1FF reserved */
1352etrap_table:
1353	.size	trap_table1, (.-trap_table1)
1354	.size	trap_table, (.-trap_table)
1355	.size	scb, (.-scb)
1356
1357/*
1358 * We get to exec_fault in the case of an instruction miss and tte
1359 * has no execute bit set.  We go to tl0 to handle it.
1360 *
1361 * g1 = tsbe pointer (in/clobbered)
1362 * g2 = tag access register (in)
1363 * g3 - g4 = scratch (clobbered)
1364 * g5 = tsbe data (in)
1365 * g6 = scratch (clobbered)
1366 * g7 = pc we jumped here from (in)
1367 */
1368	ALTENTRY(exec_fault)
1369	TRACE_TSBHIT(TT_MMU_EXEC)
1370	MMU_FAULT_STATUS_AREA(%g4)
1371	ldx	[%g4 + MMFSA_I_ADDR], %g2	/* g2 = address */
1372	ldx	[%g4 + MMFSA_I_CTX], %g3	/* g3 = ctx */
1373	srlx	%g2, MMU_PAGESHIFT, %g2		! align address to page boundry
1374	sllx	%g2, MMU_PAGESHIFT, %g2
1375	or	%g2, %g3, %g2			/* XXXQ temp */
1376	mov	T_INSTR_MMU_MISS, %g3		! arg2 = traptype
1377	set	trap, %g1
1378	ba,pt	%xcc, sys_trap
1379	  mov	-1, %g4
1380
1381.mmu_exception_not_aligned:
1382	/* %g2 = sfar, %g3 = sfsr */
1383	rdpr	%tstate, %g1
1384	btst	TSTATE_PRIV, %g1
1385	bnz,pn	%icc, 2f
1386	nop
1387	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1388	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1389	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1390	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1391	brz,pt	%g5, 2f
1392	nop
1393	ldn	[%g5 + P_UTRAP15], %g5			! unaligned utrap?
1394	brz,pn	%g5, 2f
1395	nop
1396	btst	1, %sp
1397	bz,pt	%xcc, 1f				! 32 bit user program
1398	nop
1399	ba,pt	%xcc, .setup_v9utrap			! 64 bit user program
1400	nop
14011:
1402	ba,pt	%xcc, .setup_utrap
1403	or	%g2, %g0, %g7
14042:
1405	ba,pt	%xcc, .mmu_exception_end
1406	mov	T_ALIGNMENT, %g1
1407
1408.mmu_priv_exception:
1409	rdpr	%tstate, %g1
1410	btst	TSTATE_PRIV, %g1
1411	bnz,pn	%icc, 1f
1412	nop
1413	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1414	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1415	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1416	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1417	brz,pt	%g5, 1f
1418	nop
1419	ldn	[%g5 + P_UTRAP16], %g5
1420	brnz,pt	%g5, .setup_v9utrap
1421	nop
14221:
1423	mov	T_PRIV_INSTR, %g1
1424
1425.mmu_exception_end:
1426	CPU_INDEX(%g4, %g5)
1427	set	cpu_core, %g5
1428	sllx	%g4, CPU_CORE_SHIFT, %g4
1429	add	%g4, %g5, %g4
1430	lduh	[%g4 + CPUC_DTRACE_FLAGS], %g5
1431	andcc	%g5, CPU_DTRACE_NOFAULT, %g0
1432	bz	1f
1433	or	%g5, CPU_DTRACE_BADADDR, %g5
1434	stuh	%g5, [%g4 + CPUC_DTRACE_FLAGS]
1435	done
1436
14371:
1438	sllx	%g3, 32, %g3
1439	or	%g3, %g1, %g3
1440	set	trap, %g1
1441	ba,pt	%xcc, sys_trap
1442	sub	%g0, 1, %g4
1443
1444.fp_disabled:
1445	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1446	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1447	rdpr	%tstate, %g4
1448	btst	TSTATE_PRIV, %g4
1449	bnz,a,pn %icc, ptl1_panic
1450	  mov	PTL1_BAD_FPTRAP, %g1
1451
1452	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1453	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1454	brz,a,pt %g5, 2f
1455	  nop
1456	ldn	[%g5 + P_UTRAP7], %g5			! fp_disabled utrap?
1457	brz,a,pn %g5, 2f
1458	  nop
1459	btst	1, %sp
1460	bz,a,pt	%xcc, 1f				! 32 bit user program
1461	  nop
1462	ba,a,pt	%xcc, .setup_v9utrap			! 64 bit user program
1463	  nop
14641:
1465	ba,pt	%xcc, .setup_utrap
1466	  or	%g0, %g0, %g7
14672:
1468	set	fp_disabled, %g1
1469	ba,pt	%xcc, sys_trap
1470	  sub	%g0, 1, %g4
1471
1472.fp_ieee_exception:
1473	rdpr	%tstate, %g1
1474	btst	TSTATE_PRIV, %g1
1475	bnz,a,pn %icc, ptl1_panic
1476	  mov	PTL1_BAD_FPTRAP, %g1
1477	CPU_ADDR(%g1, %g4)				! load CPU struct addr
1478	stx	%fsr, [%g1 + CPU_TMP1]
1479	ldx	[%g1 + CPU_TMP1], %g2
1480	ldn	[%g1 + CPU_THREAD], %g1			! load thread pointer
1481	ldn	[%g1 + T_PROCP], %g1			! load proc pointer
1482	ldn	[%g1 + P_UTRAPS], %g5			! are there utraps?
1483	brz,a,pt %g5, 1f
1484	  nop
1485	ldn	[%g5 + P_UTRAP8], %g5
1486	brnz,a,pt %g5, .setup_v9utrap
1487	  nop
14881:
1489	set	_fp_ieee_exception, %g1
1490	ba,pt	%xcc, sys_trap
1491	  sub	%g0, 1, %g4
1492
1493/*
1494 * Register Inputs:
1495 *	%g5		user trap handler
1496 *	%g7		misaligned addr - for alignment traps only
1497 */
1498.setup_utrap:
1499	set	trap, %g1			! setup in case we go
1500	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1501	sub	%g0, 1, %g4			! the save instruction below
1502
1503	/*
1504	 * If the DTrace pid provider is single stepping a copied-out
1505	 * instruction, t->t_dtrace_step will be set. In that case we need
1506	 * to abort the single-stepping (since execution of the instruction
1507	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1508	 */
1509	save	%sp, -SA(MINFRAME32), %sp	! window for trap handler
1510	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1511	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1512	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1513	rdpr	%tnpc, %l2			! arg1 == tnpc
1514	brz,pt	%g2, 1f
1515	rdpr	%tpc, %l1			! arg0 == tpc
1516
1517	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1518	ldn	[%g1 + T_DTRACE_NPC], %l2	! arg1 = t->t_dtrace_npc (step)
1519	brz,pt	%g2, 1f
1520	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1521	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
15221:
1523	mov	%g7, %l3			! arg2 == misaligned address
1524
1525	rdpr	%tstate, %g1			! cwp for trap handler
1526	rdpr	%cwp, %g4
1527	bclr	TSTATE_CWP_MASK, %g1
1528	wrpr	%g1, %g4, %tstate
1529	wrpr	%g0, %g5, %tnpc			! trap handler address
1530	FAST_TRAP_DONE
1531	/* NOTREACHED */
1532
1533.check_v9utrap:
1534	rdpr	%tstate, %g1
1535	btst	TSTATE_PRIV, %g1
1536	bnz,a,pn %icc, 3f
1537	  nop
1538	CPU_ADDR(%g4, %g1)				! load CPU struct addr
1539	ldn	[%g4 + CPU_THREAD], %g5			! load thread pointer
1540	ldn	[%g5 + T_PROCP], %g5			! load proc pointer
1541	ldn	[%g5 + P_UTRAPS], %g5			! are there utraps?
1542
1543	cmp	%g3, T_SOFTWARE_TRAP
1544	bne,a,pt %icc, 1f
1545	  nop
1546
1547	brz,pt %g5, 3f			! if p_utraps == NULL goto trap()
1548	  rdpr	%tt, %g3		! delay - get actual hw trap type
1549
1550	sub	%g3, 254, %g1		! UT_TRAP_INSTRUCTION_16 = p_utraps[18]
1551	ba,pt	%icc, 2f
1552	  smul	%g1, CPTRSIZE, %g2
15531:
1554	brz,a,pt %g5, 3f		! if p_utraps == NULL goto trap()
1555	  nop
1556
1557	cmp	%g3, T_UNIMP_INSTR
1558	bne,a,pt %icc, 2f
1559	  nop
1560
1561	mov	1, %g1
1562	st	%g1, [%g4 + CPU_TL1_HDLR] ! set CPU_TL1_HDLR
1563	rdpr	%tpc, %g1		! ld trapping instruction using
1564	lduwa	[%g1]ASI_AIUP, %g1	! "AS IF USER" ASI which could fault
1565	st	%g0, [%g4 + CPU_TL1_HDLR] ! clr CPU_TL1_HDLR
1566
1567	sethi	%hi(0xc1c00000), %g4	! setup mask for illtrap instruction
1568	andcc	%g1, %g4, %g4		! and instruction with mask
1569	bnz,a,pt %icc, 3f		! if %g4 == zero, %g1 is an ILLTRAP
1570	  nop				! fall thru to setup
15712:
1572	ldn	[%g5 + %g2], %g5
1573	brnz,a,pt %g5, .setup_v9utrap
1574	  nop
15753:
1576	set	trap, %g1
1577	ba,pt	%xcc, sys_trap
1578	  sub	%g0, 1, %g4
1579	/* NOTREACHED */
1580
1581/*
1582 * Register Inputs:
1583 *	%g5		user trap handler
1584 */
1585.setup_v9utrap:
1586	set	trap, %g1			! setup in case we go
1587	mov	T_FLUSH_PCB, %g3		! through sys_trap on
1588	sub	%g0, 1, %g4			! the save instruction below
1589
1590	/*
1591	 * If the DTrace pid provider is single stepping a copied-out
1592	 * instruction, t->t_dtrace_step will be set. In that case we need
1593	 * to abort the single-stepping (since execution of the instruction
1594	 * was interrupted) and use the value of t->t_dtrace_npc as the %npc.
1595	 */
1596	save	%sp, -SA(MINFRAME64), %sp	! window for trap handler
1597	CPU_ADDR(%g1, %g4)			! load CPU struct addr
1598	ldn	[%g1 + CPU_THREAD], %g1		! load thread pointer
1599	ldub	[%g1 + T_DTRACE_STEP], %g2	! load t->t_dtrace_step
1600	rdpr	%tnpc, %l7			! arg1 == tnpc
1601	brz,pt	%g2, 1f
1602	rdpr	%tpc, %l6			! arg0 == tpc
1603
1604	ldub	[%g1 + T_DTRACE_AST], %g2	! load t->t_dtrace_ast
1605	ldn	[%g1 + T_DTRACE_NPC], %l7	! arg1 == t->t_dtrace_npc (step)
1606	brz,pt	%g2, 1f
1607	st	%g0, [%g1 + T_DTRACE_FT]	! zero all pid provider flags
1608	stub	%g2, [%g1 + T_ASTFLAG]		! aston(t) if t->t_dtrace_ast
16091:
1610	rdpr	%tstate, %g2			! cwp for trap handler
1611	rdpr	%cwp, %g4
1612	bclr	TSTATE_CWP_MASK, %g2
1613	wrpr	%g2, %g4, %tstate
1614
1615	ldn	[%g1 + T_PROCP], %g4		! load proc pointer
1616	ldn	[%g4 + P_AS], %g4		! load as pointer
1617	ldn	[%g4 + A_USERLIMIT], %g4	! load as userlimit
1618	cmp	%l7, %g4			! check for single-step set
1619	bne,pt	%xcc, 4f
1620	  nop
1621	ldn	[%g1 + T_LWP], %g1		! load klwp pointer
1622	ld	[%g1 + PCB_STEP], %g4		! load single-step flag
1623	cmp	%g4, STEP_ACTIVE		! step flags set in pcb?
1624	bne,pt	%icc, 4f
1625	  nop
1626	stn	%g5, [%g1 + PCB_TRACEPC]	! save trap handler addr in pcb
1627	mov	%l7, %g4			! on entry to precise user trap
1628	add	%l6, 4, %l7			! handler, %l6 == pc, %l7 == npc
1629						! at time of trap
1630	wrpr	%g0, %g4, %tnpc			! generate FLTBOUNDS,
1631						! %g4 == userlimit
1632	FAST_TRAP_DONE
1633	/* NOTREACHED */
16344:
1635	wrpr	%g0, %g5, %tnpc			! trap handler address
1636	FAST_TRAP_DONE_CHK_INTR
1637	/* NOTREACHED */
1638
1639.fp_exception:
1640	CPU_ADDR(%g1, %g4)
1641	stx	%fsr, [%g1 + CPU_TMP1]
1642	ldx	[%g1 + CPU_TMP1], %g2
1643
1644	/*
1645	 * Cheetah takes unfinished_FPop trap for certain range of operands
1646	 * to the "fitos" instruction. Instead of going through the slow
1647	 * software emulation path, we try to simulate the "fitos" instruction
1648	 * via "fitod" and "fdtos" provided the following conditions are met:
1649	 *
1650	 *	fpu_exists is set (if DEBUG)
1651	 *	not in privileged mode
1652	 *	ftt is unfinished_FPop
1653	 *	NXM IEEE trap is not enabled
1654	 *	instruction at %tpc is "fitos"
1655	 *
1656	 *  Usage:
1657	 *	%g1	per cpu address
1658	 *	%g2	%fsr
1659	 *	%g6	user instruction
1660	 *
1661	 * Note that we can take a memory access related trap while trying
1662	 * to fetch the user instruction. Therefore, we set CPU_TL1_HDLR
1663	 * flag to catch those traps and let the SFMMU code deal with page
1664	 * fault and data access exception.
1665	 */
1666#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
1667	sethi	%hi(fpu_exists), %g7
1668	ld	[%g7 + %lo(fpu_exists)], %g7
1669	brz,pn %g7, .fp_exception_cont
1670	  nop
1671#endif
1672	rdpr	%tstate, %g7			! branch if in privileged mode
1673	btst	TSTATE_PRIV, %g7
1674	bnz,pn	%xcc, .fp_exception_cont
1675	srl	%g2, FSR_FTT_SHIFT, %g7		! extract ftt from %fsr
1676	and	%g7, (FSR_FTT>>FSR_FTT_SHIFT), %g7
1677	cmp	%g7, FTT_UNFIN
1678	set	FSR_TEM_NX, %g5
1679	bne,pn	%xcc, .fp_exception_cont	! branch if NOT unfinished_FPop
1680	  andcc	%g2, %g5, %g0
1681	bne,pn	%xcc, .fp_exception_cont	! branch if FSR_TEM_NX enabled
1682	  rdpr	%tpc, %g5			! get faulting PC
1683
1684	or	%g0, 1, %g7
1685	st	%g7, [%g1 + CPU_TL1_HDLR]	! set tl1_hdlr flag
1686	lda	[%g5]ASI_USER, %g6		! get user's instruction
1687	st	%g0, [%g1 + CPU_TL1_HDLR]	! clear tl1_hdlr flag
1688
1689	set	FITOS_INSTR_MASK, %g7
1690	and	%g6, %g7, %g7
1691	set	FITOS_INSTR, %g5
1692	cmp	%g7, %g5
1693	bne,pn	%xcc, .fp_exception_cont	! branch if not FITOS_INSTR
1694	 nop
1695
1696	/*
1697	 * This is unfinished FPops trap for "fitos" instruction. We
1698	 * need to simulate "fitos" via "fitod" and "fdtos" instruction
1699	 * sequence.
1700	 *
1701	 * We need a temporary FP register to do the conversion. Since
1702	 * both source and destination operands for the "fitos" instruction
1703	 * have to be within %f0-%f31, we use an FP register from the upper
1704	 * half to guarantee that it won't collide with the source or the
1705	 * dest operand. However, we do have to save and restore its value.
1706	 *
1707	 * We use %d62 as a temporary FP register for the conversion and
1708	 * branch to appropriate instruction within the conversion tables
1709	 * based upon the rs2 and rd values.
1710	 */
1711
1712	std	%d62, [%g1 + CPU_TMP1]		! save original value
1713
1714	srl	%g6, FITOS_RS2_SHIFT, %g7
1715	and	%g7, FITOS_REG_MASK, %g7
1716	set	_fitos_fitod_table, %g4
1717	sllx	%g7, 2, %g7
1718	jmp	%g4 + %g7
1719	  ba,pt	%xcc, _fitos_fitod_done
1720	.empty
1721
1722_fitos_fitod_table:
1723	  fitod	%f0, %d62
1724	  fitod	%f1, %d62
1725	  fitod	%f2, %d62
1726	  fitod	%f3, %d62
1727	  fitod	%f4, %d62
1728	  fitod	%f5, %d62
1729	  fitod	%f6, %d62
1730	  fitod	%f7, %d62
1731	  fitod	%f8, %d62
1732	  fitod	%f9, %d62
1733	  fitod	%f10, %d62
1734	  fitod	%f11, %d62
1735	  fitod	%f12, %d62
1736	  fitod	%f13, %d62
1737	  fitod	%f14, %d62
1738	  fitod	%f15, %d62
1739	  fitod	%f16, %d62
1740	  fitod	%f17, %d62
1741	  fitod	%f18, %d62
1742	  fitod	%f19, %d62
1743	  fitod	%f20, %d62
1744	  fitod	%f21, %d62
1745	  fitod	%f22, %d62
1746	  fitod	%f23, %d62
1747	  fitod	%f24, %d62
1748	  fitod	%f25, %d62
1749	  fitod	%f26, %d62
1750	  fitod	%f27, %d62
1751	  fitod	%f28, %d62
1752	  fitod	%f29, %d62
1753	  fitod	%f30, %d62
1754	  fitod	%f31, %d62
1755_fitos_fitod_done:
1756
1757	/*
1758	 * Now convert data back into single precision
1759	 */
1760	srl	%g6, FITOS_RD_SHIFT, %g7
1761	and	%g7, FITOS_REG_MASK, %g7
1762	set	_fitos_fdtos_table, %g4
1763	sllx	%g7, 2, %g7
1764	jmp	%g4 + %g7
1765	  ba,pt	%xcc, _fitos_fdtos_done
1766	.empty
1767
1768_fitos_fdtos_table:
1769	  fdtos	%d62, %f0
1770	  fdtos	%d62, %f1
1771	  fdtos	%d62, %f2
1772	  fdtos	%d62, %f3
1773	  fdtos	%d62, %f4
1774	  fdtos	%d62, %f5
1775	  fdtos	%d62, %f6
1776	  fdtos	%d62, %f7
1777	  fdtos	%d62, %f8
1778	  fdtos	%d62, %f9
1779	  fdtos	%d62, %f10
1780	  fdtos	%d62, %f11
1781	  fdtos	%d62, %f12
1782	  fdtos	%d62, %f13
1783	  fdtos	%d62, %f14
1784	  fdtos	%d62, %f15
1785	  fdtos	%d62, %f16
1786	  fdtos	%d62, %f17
1787	  fdtos	%d62, %f18
1788	  fdtos	%d62, %f19
1789	  fdtos	%d62, %f20
1790	  fdtos	%d62, %f21
1791	  fdtos	%d62, %f22
1792	  fdtos	%d62, %f23
1793	  fdtos	%d62, %f24
1794	  fdtos	%d62, %f25
1795	  fdtos	%d62, %f26
1796	  fdtos	%d62, %f27
1797	  fdtos	%d62, %f28
1798	  fdtos	%d62, %f29
1799	  fdtos	%d62, %f30
1800	  fdtos	%d62, %f31
1801_fitos_fdtos_done:
1802
1803	ldd	[%g1 + CPU_TMP1], %d62		! restore %d62
1804
1805#if DEBUG
1806	/*
1807	 * Update FPop_unfinished trap kstat
1808	 */
1809	set	fpustat+FPUSTAT_UNFIN_KSTAT, %g7
1810	ldx	[%g7], %g5
18111:
1812	add	%g5, 1, %g6
1813
1814	casxa	[%g7] ASI_N, %g5, %g6
1815	cmp	%g5, %g6
1816	bne,a,pn %xcc, 1b
1817	  or	%g0, %g6, %g5
1818
1819	/*
1820	 * Update fpu_sim_fitos kstat
1821	 */
1822	set	fpuinfo+FPUINFO_FITOS_KSTAT, %g7
1823	ldx	[%g7], %g5
18241:
1825	add	%g5, 1, %g6
1826
1827	casxa	[%g7] ASI_N, %g5, %g6
1828	cmp	%g5, %g6
1829	bne,a,pn %xcc, 1b
1830	  or	%g0, %g6, %g5
1831#endif /* DEBUG */
1832
1833	FAST_TRAP_DONE
1834
1835.fp_exception_cont:
1836	/*
1837	 * Let _fp_exception deal with simulating FPop instruction.
1838	 * Note that we need to pass %fsr in %g2 (already read above).
1839	 */
1840
1841	set	_fp_exception, %g1
1842	ba,pt	%xcc, sys_trap
1843	sub	%g0, 1, %g4
1844
1845
1846/*
1847 * Register windows
1848 */
1849
1850/*
1851 * FILL_32bit_flushw/FILL_64bit_flushw fills a 32/64-bit-wide register window
1852 * from a 32/64-bit * wide address space via the designated asi.
1853 * It is used to fill windows in user_flushw to avoid going above TL 2.
1854 */
1855/* TODO: Use the faster FILL based on FILL_32bit_asi/FILL_64bit_asi */
1856#define	FILL_32bit_flushw(asi_num)				\
1857	mov	asi_num, %asi					;\
1858	rdpr	%cwp, %g2					;\
1859	sub	%g2, 1, %g2					;\
1860	wrpr	%g2, %cwp					;\
18611:	srl	%sp, 0, %sp					;\
1862	lda	[%sp + 0]%asi, %l0				;\
1863	lda	[%sp + 4]%asi, %l1				;\
1864	lda	[%sp + 8]%asi, %l2				;\
1865	lda	[%sp + 12]%asi, %l3				;\
1866	lda	[%sp + 16]%asi, %l4				;\
1867	lda	[%sp + 20]%asi, %l5				;\
1868	lda	[%sp + 24]%asi, %l6				;\
1869	lda	[%sp + 28]%asi, %l7				;\
1870	lda	[%sp + 32]%asi, %i0				;\
1871	lda	[%sp + 36]%asi, %i1				;\
1872	lda	[%sp + 40]%asi, %i2				;\
1873	lda	[%sp + 44]%asi, %i3				;\
1874	lda	[%sp + 48]%asi, %i4				;\
1875	lda	[%sp + 52]%asi, %i5				;\
1876	lda	[%sp + 56]%asi, %i6				;\
1877	lda	[%sp + 60]%asi, %i7				;\
1878	restored						;\
1879	add	%g2, 1, %g2					;\
1880	wrpr	%g2, %cwp
1881
1882#define	FILL_64bit_flushw(asi_num)				\
1883	mov	asi_num, %asi					;\
1884	rdpr	%cwp, %g2					;\
1885	sub	%g2, 1, %g2					;\
1886	wrpr	%g2, %cwp					;\
1887	ldxa	[%sp + V9BIAS64 + 0]%asi, %l0			;\
1888	ldxa	[%sp + V9BIAS64 + 8]%asi, %l1			;\
1889	ldxa	[%sp + V9BIAS64 + 16]%asi, %l2			;\
1890	ldxa	[%sp + V9BIAS64 + 24]%asi, %l3			;\
1891	ldxa	[%sp + V9BIAS64 + 32]%asi, %l4			;\
1892	ldxa	[%sp + V9BIAS64 + 40]%asi, %l5			;\
1893	ldxa	[%sp + V9BIAS64 + 48]%asi, %l6			;\
1894	ldxa	[%sp + V9BIAS64 + 56]%asi, %l7			;\
1895	ldxa	[%sp + V9BIAS64 + 64]%asi, %i0			;\
1896	ldxa	[%sp + V9BIAS64 + 72]%asi, %i1			;\
1897	ldxa	[%sp + V9BIAS64 + 80]%asi, %i2			;\
1898	ldxa	[%sp + V9BIAS64 + 88]%asi, %i3			;\
1899	ldxa	[%sp + V9BIAS64 + 96]%asi, %i4			;\
1900	ldxa	[%sp + V9BIAS64 + 104]%asi, %i5			;\
1901	ldxa	[%sp + V9BIAS64 + 112]%asi, %i6			;\
1902	ldxa	[%sp + V9BIAS64 + 120]%asi, %i7			;\
1903	restored						;\
1904	add	%g2, 1, %g2					;\
1905	wrpr	%g2, %cwp
1906
1907.flushw:
1908	rdpr	%tnpc, %g1
1909	wrpr	%g1, %tpc
1910	add	%g1, 4, %g1
1911	wrpr	%g1, %tnpc
1912	set	trap, %g1
1913	mov	T_FLUSH_PCB, %g3
1914	ba,pt	%xcc, sys_trap
1915	sub	%g0, 1, %g4
1916
1917.clean_windows:
1918	set	trap, %g1
1919	mov	T_FLUSH_PCB, %g3
1920	sub	%g0, 1, %g4
1921	save
1922	flushw
1923	rdpr	%canrestore, %g2
1924	brnz	%g2, 1f
1925	nop
1926	rdpr	%wstate, %g2
1927	btst	1, %g2
1928	beq	2f
1929	nop
1930	FILL_32bit_flushw(ASI_AIUP)
1931	ba,a	1f
1932	 .empty
19332:
1934	FILL_64bit_flushw(ASI_AIUP)
19351:
1936	restore
1937	wrpr	%g0, %g0, %cleanwin	! no clean windows
1938
1939	CPU_ADDR(%g4, %g5)
1940	ldn	[%g4 + CPU_MPCB], %g4
1941	brz,a,pn %g4, 1f
1942	  nop
1943	ld	[%g4 + MPCB_WSTATE], %g5
1944	add	%g5, WSTATE_CLEAN_OFFSET, %g5
1945	wrpr	%g0, %g5, %wstate
19461:	FAST_TRAP_DONE
1947
1948/*
1949 * .spill_clean: clean the previous window, restore the wstate, and
1950 * "done".
1951 *
1952 * Entry: %g7 contains new wstate
1953 */
1954.spill_clean:
1955	sethi	%hi(nwin_minus_one), %g5
1956	ld	[%g5 + %lo(nwin_minus_one)], %g5 ! %g5 = nwin - 1
1957	rdpr	%cwp, %g6			! %g6 = %cwp
1958	deccc	%g6				! %g6--
1959	movneg	%xcc, %g5, %g6			! if (%g6<0) %g6 = nwin-1
1960	wrpr	%g6, %cwp
1961	TT_TRACE_L(trace_win)
1962	clr	%l0
1963	clr	%l1
1964	clr	%l2
1965	clr	%l3
1966	clr	%l4
1967	clr	%l5
1968	clr	%l6
1969	clr	%l7
1970	wrpr	%g0, %g7, %wstate
1971	saved
1972	retry			! restores correct %cwp
1973
1974.fix_alignment:
1975	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
1976	ldn	[%g1 + CPU_THREAD], %g1	! load thread pointer
1977	ldn	[%g1 + T_PROCP], %g1
1978	mov	1, %g2
1979	stb	%g2, [%g1 + P_FIXALIGNMENT]
1980	FAST_TRAP_DONE
1981
1982#define	STDF_REG(REG, ADDR, TMP)		\
1983	sll	REG, 3, REG			;\
1984mark1:	set	start1, TMP			;\
1985	jmp	REG + TMP			;\
1986	  nop					;\
1987start1:	ba,pt	%xcc, done1			;\
1988	  std	%f0, [ADDR + CPU_TMP1]		;\
1989	ba,pt	%xcc, done1			;\
1990	  std	%f32, [ADDR + CPU_TMP1]		;\
1991	ba,pt	%xcc, done1			;\
1992	  std	%f2, [ADDR + CPU_TMP1]		;\
1993	ba,pt	%xcc, done1			;\
1994	  std	%f34, [ADDR + CPU_TMP1]		;\
1995	ba,pt	%xcc, done1			;\
1996	  std	%f4, [ADDR + CPU_TMP1]		;\
1997	ba,pt	%xcc, done1			;\
1998	  std	%f36, [ADDR + CPU_TMP1]		;\
1999	ba,pt	%xcc, done1			;\
2000	  std	%f6, [ADDR + CPU_TMP1]		;\
2001	ba,pt	%xcc, done1			;\
2002	  std	%f38, [ADDR + CPU_TMP1]		;\
2003	ba,pt	%xcc, done1			;\
2004	  std	%f8, [ADDR + CPU_TMP1]		;\
2005	ba,pt	%xcc, done1			;\
2006	  std	%f40, [ADDR + CPU_TMP1]		;\
2007	ba,pt	%xcc, done1			;\
2008	  std	%f10, [ADDR + CPU_TMP1]		;\
2009	ba,pt	%xcc, done1			;\
2010	  std	%f42, [ADDR + CPU_TMP1]		;\
2011	ba,pt	%xcc, done1			;\
2012	  std	%f12, [ADDR + CPU_TMP1]		;\
2013	ba,pt	%xcc, done1			;\
2014	  std	%f44, [ADDR + CPU_TMP1]		;\
2015	ba,pt	%xcc, done1			;\
2016	  std	%f14, [ADDR + CPU_TMP1]		;\
2017	ba,pt	%xcc, done1			;\
2018	  std	%f46, [ADDR + CPU_TMP1]		;\
2019	ba,pt	%xcc, done1			;\
2020	  std	%f16, [ADDR + CPU_TMP1]		;\
2021	ba,pt	%xcc, done1			;\
2022	  std	%f48, [ADDR + CPU_TMP1]		;\
2023	ba,pt	%xcc, done1			;\
2024	  std	%f18, [ADDR + CPU_TMP1]		;\
2025	ba,pt	%xcc, done1			;\
2026	  std	%f50, [ADDR + CPU_TMP1]		;\
2027	ba,pt	%xcc, done1			;\
2028	  std	%f20, [ADDR + CPU_TMP1]		;\
2029	ba,pt	%xcc, done1			;\
2030	  std	%f52, [ADDR + CPU_TMP1]		;\
2031	ba,pt	%xcc, done1			;\
2032	  std	%f22, [ADDR + CPU_TMP1]		;\
2033	ba,pt	%xcc, done1			;\
2034	  std	%f54, [ADDR + CPU_TMP1]		;\
2035	ba,pt	%xcc, done1			;\
2036	  std	%f24, [ADDR + CPU_TMP1]		;\
2037	ba,pt	%xcc, done1			;\
2038	  std	%f56, [ADDR + CPU_TMP1]		;\
2039	ba,pt	%xcc, done1			;\
2040	  std	%f26, [ADDR + CPU_TMP1]		;\
2041	ba,pt	%xcc, done1			;\
2042	  std	%f58, [ADDR + CPU_TMP1]		;\
2043	ba,pt	%xcc, done1			;\
2044	  std	%f28, [ADDR + CPU_TMP1]		;\
2045	ba,pt	%xcc, done1			;\
2046	  std	%f60, [ADDR + CPU_TMP1]		;\
2047	ba,pt	%xcc, done1			;\
2048	  std	%f30, [ADDR + CPU_TMP1]		;\
2049	ba,pt	%xcc, done1			;\
2050	  std	%f62, [ADDR + CPU_TMP1]		;\
2051done1:
2052
2053#define	LDDF_REG(REG, ADDR, TMP)		\
2054	sll	REG, 3, REG			;\
2055mark2:	set	start2, TMP			;\
2056	jmp	REG + TMP			;\
2057	  nop					;\
2058start2:	ba,pt	%xcc, done2			;\
2059	  ldd	[ADDR + CPU_TMP1], %f0		;\
2060	ba,pt	%xcc, done2			;\
2061	  ldd	[ADDR + CPU_TMP1], %f32		;\
2062	ba,pt	%xcc, done2			;\
2063	  ldd	[ADDR + CPU_TMP1], %f2		;\
2064	ba,pt	%xcc, done2			;\
2065	  ldd	[ADDR + CPU_TMP1], %f34		;\
2066	ba,pt	%xcc, done2			;\
2067	  ldd	[ADDR + CPU_TMP1], %f4		;\
2068	ba,pt	%xcc, done2			;\
2069	  ldd	[ADDR + CPU_TMP1], %f36		;\
2070	ba,pt	%xcc, done2			;\
2071	  ldd	[ADDR + CPU_TMP1], %f6		;\
2072	ba,pt	%xcc, done2			;\
2073	  ldd	[ADDR + CPU_TMP1], %f38		;\
2074	ba,pt	%xcc, done2			;\
2075	  ldd	[ADDR + CPU_TMP1], %f8		;\
2076	ba,pt	%xcc, done2			;\
2077	  ldd	[ADDR + CPU_TMP1], %f40		;\
2078	ba,pt	%xcc, done2			;\
2079	  ldd	[ADDR + CPU_TMP1], %f10		;\
2080	ba,pt	%xcc, done2			;\
2081	  ldd	[ADDR + CPU_TMP1], %f42		;\
2082	ba,pt	%xcc, done2			;\
2083	  ldd	[ADDR + CPU_TMP1], %f12		;\
2084	ba,pt	%xcc, done2			;\
2085	  ldd	[ADDR + CPU_TMP1], %f44		;\
2086	ba,pt	%xcc, done2			;\
2087	  ldd	[ADDR + CPU_TMP1], %f14		;\
2088	ba,pt	%xcc, done2			;\
2089	  ldd	[ADDR + CPU_TMP1], %f46		;\
2090	ba,pt	%xcc, done2			;\
2091	  ldd	[ADDR + CPU_TMP1], %f16		;\
2092	ba,pt	%xcc, done2			;\
2093	  ldd	[ADDR + CPU_TMP1], %f48		;\
2094	ba,pt	%xcc, done2			;\
2095	  ldd	[ADDR + CPU_TMP1], %f18		;\
2096	ba,pt	%xcc, done2			;\
2097	  ldd	[ADDR + CPU_TMP1], %f50		;\
2098	ba,pt	%xcc, done2			;\
2099	  ldd	[ADDR + CPU_TMP1], %f20		;\
2100	ba,pt	%xcc, done2			;\
2101	  ldd	[ADDR + CPU_TMP1], %f52		;\
2102	ba,pt	%xcc, done2			;\
2103	  ldd	[ADDR + CPU_TMP1], %f22		;\
2104	ba,pt	%xcc, done2			;\
2105	  ldd	[ADDR + CPU_TMP1], %f54		;\
2106	ba,pt	%xcc, done2			;\
2107	  ldd	[ADDR + CPU_TMP1], %f24		;\
2108	ba,pt	%xcc, done2			;\
2109	  ldd	[ADDR + CPU_TMP1], %f56		;\
2110	ba,pt	%xcc, done2			;\
2111	  ldd	[ADDR + CPU_TMP1], %f26		;\
2112	ba,pt	%xcc, done2			;\
2113	  ldd	[ADDR + CPU_TMP1], %f58		;\
2114	ba,pt	%xcc, done2			;\
2115	  ldd	[ADDR + CPU_TMP1], %f28		;\
2116	ba,pt	%xcc, done2			;\
2117	  ldd	[ADDR + CPU_TMP1], %f60		;\
2118	ba,pt	%xcc, done2			;\
2119	  ldd	[ADDR + CPU_TMP1], %f30		;\
2120	ba,pt	%xcc, done2			;\
2121	  ldd	[ADDR + CPU_TMP1], %f62		;\
2122done2:
2123
2124.lddf_exception_not_aligned:
2125	/* %g2 = sfar, %g3 = sfsr */
2126	mov	%g2, %g5		! stash sfar
2127#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2128	sethi	%hi(fpu_exists), %g2	! check fpu_exists
2129	ld	[%g2 + %lo(fpu_exists)], %g2
2130	brz,a,pn %g2, 4f
2131	  nop
2132#endif
2133	CPU_ADDR(%g1, %g4)
2134	or	%g0, 1, %g4
2135	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2136
2137	rdpr	%tpc, %g2
2138	lda	[%g2]ASI_AIUP, %g6	! get the user's lddf instruction
2139	srl	%g6, 23, %g1		! using ldda or not?
2140	and	%g1, 1, %g1
2141	brz,a,pt %g1, 2f		! check for ldda instruction
2142	  nop
2143	srl	%g6, 13, %g1		! check immflag
2144	and	%g1, 1, %g1
2145	rdpr	%tstate, %g2		! %tstate in %g2
2146	brnz,a,pn %g1, 1f
2147	  srl	%g2, 31, %g1		! get asi from %tstate
2148	srl	%g6, 5, %g1		! get asi from instruction
2149	and	%g1, 0xFF, %g1		! imm_asi field
21501:
2151	cmp	%g1, ASI_P		! primary address space
2152	be,a,pt %icc, 2f
2153	  nop
2154	cmp	%g1, ASI_PNF		! primary no fault address space
2155	be,a,pt %icc, 2f
2156	  nop
2157	cmp	%g1, ASI_S		! secondary address space
2158	be,a,pt %icc, 2f
2159	  nop
2160	cmp	%g1, ASI_SNF		! secondary no fault address space
2161	bne,a,pn %icc, 3f
2162	  nop
21632:
2164	lduwa	[%g5]ASI_USER, %g7	! get first half of misaligned data
2165	add	%g5, 4, %g5		! increment misaligned data address
2166	lduwa	[%g5]ASI_USER, %g5	! get second half of misaligned data
2167
2168	sllx	%g7, 32, %g7
2169	or	%g5, %g7, %g5		! combine data
2170	CPU_ADDR(%g7, %g1)		! save data on a per-cpu basis
2171	stx	%g5, [%g7 + CPU_TMP1]	! save in cpu_tmp1
2172
2173	srl	%g6, 25, %g3		! %g6 has the instruction
2174	and	%g3, 0x1F, %g3		! %g3 has rd
2175	LDDF_REG(%g3, %g7, %g4)
2176
2177	CPU_ADDR(%g1, %g4)
2178	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2179	FAST_TRAP_DONE
21803:
2181	CPU_ADDR(%g1, %g4)
2182	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
21834:
2184	set	T_USER, %g3		! trap type in %g3
2185	or	%g3, T_LDDF_ALIGN, %g3
2186	mov	%g5, %g2		! misaligned vaddr in %g2
2187	set	fpu_trap, %g1		! goto C for the little and
2188	ba,pt	%xcc, sys_trap		! no fault little asi's
2189	  sub	%g0, 1, %g4
2190
2191.stdf_exception_not_aligned:
2192	/* %g2 = sfar, %g3 = sfsr */
2193	mov	%g2, %g5
2194
2195#if defined(DEBUG) || defined(NEED_FPU_EXISTS)
2196	sethi	%hi(fpu_exists), %g7		! check fpu_exists
2197	ld	[%g7 + %lo(fpu_exists)], %g3
2198	brz,a,pn %g3, 4f
2199	  nop
2200#endif
2201	CPU_ADDR(%g1, %g4)
2202	or	%g0, 1, %g4
2203	st	%g4, [%g1 + CPU_TL1_HDLR] ! set tl1_hdlr flag
2204
2205	rdpr	%tpc, %g2
2206	lda	[%g2]ASI_AIUP, %g6	! get the user's stdf instruction
2207
2208	srl	%g6, 23, %g1		! using stda or not?
2209	and	%g1, 1, %g1
2210	brz,a,pt %g1, 2f		! check for stda instruction
2211	  nop
2212	srl	%g6, 13, %g1		! check immflag
2213	and	%g1, 1, %g1
2214	rdpr	%tstate, %g2		! %tstate in %g2
2215	brnz,a,pn %g1, 1f
2216	  srl	%g2, 31, %g1		! get asi from %tstate
2217	srl	%g6, 5, %g1		! get asi from instruction
2218	and	%g1, 0xff, %g1		! imm_asi field
22191:
2220	cmp	%g1, ASI_P		! primary address space
2221	be,a,pt %icc, 2f
2222	  nop
2223	cmp	%g1, ASI_S		! secondary address space
2224	bne,a,pn %icc, 3f
2225	  nop
22262:
2227	srl	%g6, 25, %g6
2228	and	%g6, 0x1F, %g6		! %g6 has rd
2229	CPU_ADDR(%g7, %g1)
2230	STDF_REG(%g6, %g7, %g4)		! STDF_REG(REG, ADDR, TMP)
2231
2232	ldx	[%g7 + CPU_TMP1], %g6
2233	srlx	%g6, 32, %g7
2234	stuwa	%g7, [%g5]ASI_USER	! first half
2235	add	%g5, 4, %g5		! increment misaligned data address
2236	stuwa	%g6, [%g5]ASI_USER	! second half
2237
2238	CPU_ADDR(%g1, %g4)
2239	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
2240	FAST_TRAP_DONE
22413:
2242	CPU_ADDR(%g1, %g4)
2243	st	%g0, [%g1 + CPU_TL1_HDLR] ! clear tl1_hdlr flag
22444:
2245	set	T_USER, %g3		! trap type in %g3
2246	or	%g3, T_STDF_ALIGN, %g3
2247	mov	%g5, %g2		! misaligned vaddr in %g2
2248	set	fpu_trap, %g1		! goto C for the little and
2249	ba,pt	%xcc, sys_trap		! nofault little asi's
2250	  sub	%g0, 1, %g4
2251
2252#ifdef DEBUG_USER_TRAPTRACECTL
2253
2254.traptrace_freeze:
2255	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2256	TT_TRACE_L(trace_win)
2257	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2258	set	trap_freeze, %g1
2259	mov	1, %g2
2260	st	%g2, [%g1]
2261	FAST_TRAP_DONE
2262
2263.traptrace_unfreeze:
2264	set	trap_freeze, %g1
2265	st	%g0, [%g1]
2266	mov	%l0, %g1 ; mov	%l1, %g2 ; mov	%l2, %g3 ; mov	%l4, %g4
2267	TT_TRACE_L(trace_win)
2268	mov	%g4, %l4 ; mov	%g3, %l2 ; mov	%g2, %l1 ; mov	%g1, %l0
2269	FAST_TRAP_DONE
2270
2271#endif /* DEBUG_USER_TRAPTRACECTL */
2272
2273.getcc:
2274	CPU_ADDR(%g1, %g2)
2275	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2276	rdpr	%tstate, %g3			! get tstate
2277	srlx	%g3, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2278	set	PSR_ICC, %g2
2279	and	%o0, %g2, %o0			! mask out the rest
2280	srl	%o0, PSR_ICC_SHIFT, %o0		! right justify
2281	wrpr	%g0, 0, %gl
2282	mov	%o0, %g1			! move ccr to normal %g1
2283	wrpr	%g0, 1, %gl
2284	! cannot assume globals retained their values after increasing %gl
2285	CPU_ADDR(%g1, %g2)
2286	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2287	FAST_TRAP_DONE
2288
2289.setcc:
2290	CPU_ADDR(%g1, %g2)
2291	stx	%o0, [%g1 + CPU_TMP1]		! save %o0
2292	wrpr	%g0, 0, %gl
2293	mov	%g1, %o0
2294	wrpr	%g0, 1, %gl
2295	! cannot assume globals retained their values after increasing %gl
2296	CPU_ADDR(%g1, %g2)
2297	sll	%o0, PSR_ICC_SHIFT, %g2
2298	set	PSR_ICC, %g3
2299	and	%g2, %g3, %g2			! mask out rest
2300	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g2
2301	rdpr	%tstate, %g3			! get tstate
2302	srl	%g3, 0, %g3			! clear upper word
2303	or	%g3, %g2, %g3			! or in new bits
2304	wrpr	%g3, %tstate
2305	ldx	[%g1 + CPU_TMP1], %o0		! restore %o0
2306	FAST_TRAP_DONE
2307
2308/*
2309 * getpsr(void)
2310 * Note that the xcc part of the ccr is not provided.
2311 * The V8 code shows why the V9 trap is not faster:
2312 * #define GETPSR_TRAP() \
2313 *      mov %psr, %i0; jmp %l2; rett %l2+4; nop;
2314 */
2315
2316	.type	.getpsr, #function
2317.getpsr:
2318	rdpr	%tstate, %g1			! get tstate
2319	srlx	%g1, PSR_TSTATE_CC_SHIFT, %o0	! shift ccr to V8 psr
2320	set	PSR_ICC, %g2
2321	and	%o0, %g2, %o0			! mask out the rest
2322
2323	rd	%fprs, %g1			! get fprs
2324	and	%g1, FPRS_FEF, %g2		! mask out dirty upper/lower
2325	sllx	%g2, PSR_FPRS_FEF_SHIFT, %g2	! shift fef to V8 psr.ef
2326	or	%o0, %g2, %o0			! or result into psr.ef
2327
2328	set	V9_PSR_IMPLVER, %g2		! SI assigned impl/ver: 0xef
2329	or	%o0, %g2, %o0			! or psr.impl/ver
2330	FAST_TRAP_DONE
2331	SET_SIZE(.getpsr)
2332
2333/*
2334 * setpsr(newpsr)
2335 * Note that there is no support for ccr.xcc in the V9 code.
2336 */
2337
2338	.type	.setpsr, #function
2339.setpsr:
2340	rdpr	%tstate, %g1			! get tstate
2341!	setx	TSTATE_V8_UBITS, %g2
2342	or 	%g0, CCR_ICC, %g3
2343	sllx	%g3, TSTATE_CCR_SHIFT, %g2
2344
2345	andn	%g1, %g2, %g1			! zero current user bits
2346	set	PSR_ICC, %g2
2347	and	%g2, %o0, %g2			! clear all but psr.icc bits
2348	sllx	%g2, PSR_TSTATE_CC_SHIFT, %g3	! shift to tstate.ccr.icc
2349	wrpr	%g1, %g3, %tstate		! write tstate
2350
2351	set	PSR_EF, %g2
2352	and	%g2, %o0, %g2			! clear all but fp enable bit
2353	srlx	%g2, PSR_FPRS_FEF_SHIFT, %g4	! shift ef to V9 fprs.fef
2354	wr	%g0, %g4, %fprs			! write fprs
2355
2356	CPU_ADDR(%g1, %g2)			! load CPU struct addr to %g1
2357	ldn	[%g1 + CPU_THREAD], %g2		! load thread pointer
2358	ldn	[%g2 + T_LWP], %g3		! load klwp pointer
2359	ldn	[%g3 + LWP_FPU], %g2		! get lwp_fpu pointer
2360	stuw	%g4, [%g2 + FPU_FPRS]		! write fef value to fpu_fprs
2361	srlx	%g4, 2, %g4			! shift fef value to bit 0
2362	stub	%g4, [%g2 + FPU_EN]		! write fef value to fpu_en
2363	FAST_TRAP_DONE
2364	SET_SIZE(.setpsr)
2365
2366/*
2367 * getlgrp
2368 * get home lgrpid on which the calling thread is currently executing.
2369 */
2370	.type	.getlgrp, #function
2371.getlgrp:
2372	! Thanks for the incredibly helpful comments
2373	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2374	ld	[%g1 + CPU_ID], %o0	! load cpu_id
2375	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2376	ldn	[%g2 + T_LPL], %g2	! load lpl pointer
2377	ld	[%g2 + LPL_LGRPID], %g1	! load lpl_lgrpid
2378	sra	%g1, 0, %o1
2379	FAST_TRAP_DONE
2380	SET_SIZE(.getlgrp)
2381
2382/*
2383 * Entry for old 4.x trap (trap 0).
2384 */
2385	ENTRY_NP(syscall_trap_4x)
2386	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2387	ldn	[%g1 + CPU_THREAD], %g2	! load thread pointer
2388	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2389	ld	[%g2 + PCB_TRAP0], %g2	! lwp->lwp_pcb.pcb_trap0addr
2390	brz,pn	%g2, 1f			! has it been set?
2391	st	%l0, [%g1 + CPU_TMP1]	! delay - save some locals
2392	st	%l1, [%g1 + CPU_TMP2]
2393	rdpr	%tnpc, %l1		! save old tnpc
2394	wrpr	%g0, %g2, %tnpc		! setup tnpc
2395
2396	mov	%g1, %l0		! save CPU struct addr
2397	wrpr	%g0, 0, %gl
2398	mov	%l1, %g6		! pass tnpc to user code in %g6
2399	wrpr	%g0, 1, %gl
2400	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2401	ld	[%l0 + CPU_TMP1], %l0
2402	FAST_TRAP_DONE_CHK_INTR
24031:
2404	!
2405	! check for old syscall mmap which is the only different one which
2406	! must be the same.  Others are handled in the compatibility library.
2407	!
2408	mov	%g1, %l0		! save CPU struct addr
2409	wrpr	%g0, 0, %gl
2410	cmp	%g1, OSYS_mmap		! compare to old 4.x mmap
2411	movz	%icc, SYS_mmap, %g1
2412	wrpr	%g0, 1, %gl
2413	ld	[%l0 + CPU_TMP1], %l0
2414	SYSCALL(syscall_trap32)
2415	SET_SIZE(syscall_trap_4x)
2416
2417/*
2418 * Handler for software trap 9.
2419 * Set trap0 emulation address for old 4.x system call trap.
2420 * XXX - this should be a system call.
2421 */
2422	ENTRY_NP(set_trap0_addr)
2423	CPU_ADDR(%g1, %g2)		! load CPU struct addr to %g1 using %g2
2424	st	%l0, [%g1 + CPU_TMP1]	! save some locals
2425	st	%l1, [%g1 + CPU_TMP2]
2426	mov	%g1, %l0	! preserve CPU addr
2427	wrpr	%g0, 0, %gl
2428	mov	%g1, %l1
2429	wrpr	%g0, 1, %gl
2430	! cannot assume globals retained their values after increasing %gl
2431	ldn	[%l0 + CPU_THREAD], %g2	! load thread pointer
2432	ldn	[%g2 + T_LWP], %g2	! load klwp pointer
2433	andn	%l1, 3, %l1		! force alignment
2434	st	%l1, [%g2 + PCB_TRAP0]	! lwp->lwp_pcb.pcb_trap0addr
2435	ld	[%l0 + CPU_TMP2], %l1	! restore locals
2436	ld	[%l0 + CPU_TMP1], %l0
2437	FAST_TRAP_DONE
2438	SET_SIZE(set_trap0_addr)
2439
2440/*
2441 * mmu_trap_tl1
2442 * trap handler for unexpected mmu traps.
2443 * simply checks if the trap was a user lddf/stdf alignment trap, in which
2444 * case we go to fpu_trap or a user trap from the window handler, in which
2445 * case we go save the state on the pcb.  Otherwise, we go to ptl1_panic.
2446 */
2447	.type	mmu_trap_tl1, #function
2448mmu_trap_tl1:
2449#ifdef	TRAPTRACE
2450	TRACE_PTR(%g5, %g6)
2451	GET_TRACE_TICK(%g6)
2452	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2453	TRACE_SAVE_TL_GL_REGS(%g5, %g6)
2454	rdpr	%tt, %g6
2455	stha	%g6, [%g5 + TRAP_ENT_TT]%asi
2456	rdpr	%tstate, %g6
2457	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi
2458	stna	%sp, [%g5 + TRAP_ENT_SP]%asi
2459	stna	%g0, [%g5 + TRAP_ENT_TR]%asi
2460	rdpr	%tpc, %g6
2461	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2462	MMU_FAULT_STATUS_AREA(%g6)
2463	ldx	[%g6 + MMFSA_D_ADDR], %g6
2464	stna	%g6, [%g5 + TRAP_ENT_F1]%asi !  MMU fault address
2465	CPU_PADDR(%g7, %g6);
2466	add	%g7, CPU_TL1_HDLR, %g7
2467	lda	[%g7]ASI_MEM, %g6
2468	stna	%g6, [%g5 + TRAP_ENT_F2]%asi
2469	MMU_FAULT_STATUS_AREA(%g6)
2470	ldx	[%g6 + MMFSA_D_TYPE], %g7 ! XXXQ should be a MMFSA_F_ constant?
2471	ldx	[%g6 + MMFSA_D_CTX], %g6
2472	sllx	%g6, SFSR_CTX_SHIFT, %g6
2473	or	%g6, %g7, %g6
2474	stna	%g6, [%g5 + TRAP_ENT_F3]%asi ! MMU context/type
2475	set	0xdeadbeef, %g6
2476	stna	%g6, [%g5 + TRAP_ENT_F4]%asi
2477	TRACE_NEXT(%g5, %g6, %g7)
2478#endif /* TRAPTRACE */
2479	CPU_PADDR(%g7, %g6);
2480	add     %g7, CPU_TL1_HDLR, %g7		! %g7 = &cpu_m.tl1_hdlr (PA)
2481	lda	[%g7]ASI_MEM, %g6
2482	brz,a,pt %g6, 1f
2483	  nop
2484	sta     %g0, [%g7]ASI_MEM
2485	! XXXQ need to setup registers for sfmmu_mmu_trap?
2486	ba,a,pt	%xcc, sfmmu_mmu_trap		! handle page faults
24871:
2488	rdpr	%tpc, %g7
2489	/* in user_rtt? */
2490	set	rtt_fill_start, %g6
2491	cmp	%g7, %g6
2492	blu,pn	%xcc, 6f
2493	 .empty
2494	set	rtt_fill_end, %g6
2495	cmp	%g7, %g6
2496	bgeu,pn %xcc, 6f
2497	 nop
2498	set	fault_rtt_fn1, %g7
2499	ba,a	7f
25006:
2501	! check to see if the trap pc is in a window spill/fill handling
2502	rdpr	%tpc, %g7
2503	/* tpc should be in the trap table */
2504	set	trap_table, %g6
2505	cmp	%g7, %g6
2506	blu,a,pn %xcc, ptl1_panic
2507	  mov	PTL1_BAD_MMUTRAP, %g1
2508	set	etrap_table, %g6
2509	cmp	%g7, %g6
2510	bgeu,a,pn %xcc, ptl1_panic
2511	  mov	PTL1_BAD_MMUTRAP, %g1
2512	! pc is inside the trap table, convert to trap type
2513	srl	%g7, 5, %g6		! XXXQ need #define
2514	and	%g6, 0x1ff, %g6		! XXXQ need #define
2515	! and check for a window trap type
2516	and	%g6, WTRAP_TTMASK, %g6
2517	cmp	%g6, WTRAP_TYPE
2518	bne,a,pn %xcc, ptl1_panic
2519	  mov	PTL1_BAD_MMUTRAP, %g1
2520	andn	%g7, WTRAP_ALIGN, %g7	/* 128 byte aligned */
2521	add	%g7, WTRAP_FAULTOFF, %g7
2522
25237:
2524	! Arguments are passed in the global set active after the
2525	! 'done' instruction. Before switching sets, must save
2526	! the calculated next pc
2527	wrpr	%g0, %g7, %tnpc
2528	wrpr	%g0, 1, %gl
2529	rdpr	%tt, %g5
2530	cmp	%g5, T_ALIGNMENT
2531	MMU_FAULT_STATUS_AREA(%g4)
2532	ldx	[%g4 + MMFSA_D_ADDR], %g6
2533	done
2534	SET_SIZE(mmu_trap_tl1)
2535
2536/*
2537 * Several traps use kmdb_trap and kmdb_trap_tl1 as their handlers.  These
2538 * traps are valid only when kmdb is loaded.  When the debugger is active,
2539 * the code below is rewritten to transfer control to the appropriate
2540 * debugger entry points.
2541 */
2542	.global	kmdb_trap
2543	.align	8
2544kmdb_trap:
2545	ba,a	trap_table0
2546	jmp	%g1 + 0
2547	nop
2548
2549	.global	kmdb_trap_tl1
2550	.align	8
2551kmdb_trap_tl1:
2552	ba,a	trap_table0
2553	jmp	%g1 + 0
2554	nop
2555
2556/*
2557 * This entry is copied from OBP's trap table during boot.
2558 */
2559	.global	obp_bpt
2560	.align	8
2561obp_bpt:
2562	NOT
2563
2564
2565
2566#ifdef	TRAPTRACE
2567/*
2568 * TRAPTRACE support.
2569 * labels here are branched to with "rd %pc, %g7" in the delay slot.
2570 * Return is done by "jmp %g7 + 4".
2571 */
2572
2573trace_dmmu:
2574	TRACE_PTR(%g3, %g6)
2575	GET_TRACE_TICK(%g6)
2576	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2577	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2578	rdpr	%tt, %g6
2579	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2580	rdpr	%tstate, %g6
2581	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2582	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2583	rdpr	%tpc, %g6
2584	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2585	MMU_FAULT_STATUS_AREA(%g6)
2586	ldx	[%g6 + MMFSA_D_ADDR], %g4
2587	stxa	%g4, [%g3 + TRAP_ENT_TR]%asi
2588	ldx	[%g6 + MMFSA_D_CTX], %g4
2589	stxa	%g4, [%g3 + TRAP_ENT_F1]%asi
2590	ldx	[%g6 + MMFSA_D_TYPE], %g4
2591	stxa	%g4, [%g3 + TRAP_ENT_F2]%asi
2592	stxa	%g6, [%g3 + TRAP_ENT_F3]%asi
2593	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2594	TRACE_NEXT(%g3, %g4, %g5)
2595	jmp	%g7 + 4
2596	nop
2597
2598trace_immu:
2599	TRACE_PTR(%g3, %g6)
2600	GET_TRACE_TICK(%g6)
2601	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2602	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2603	rdpr	%tt, %g6
2604	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2605	rdpr	%tstate, %g6
2606	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2607	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2608	rdpr	%tpc, %g6
2609	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2610	MMU_FAULT_STATUS_AREA(%g6)
2611	ldx	[%g6 + MMFSA_I_ADDR], %g4
2612	stxa	%g4, [%g3 + TRAP_ENT_TR]%asi
2613	ldx	[%g6 + MMFSA_I_CTX], %g4
2614	stxa	%g4, [%g3 + TRAP_ENT_F1]%asi
2615	ldx	[%g6 + MMFSA_I_TYPE], %g4
2616	stxa	%g4, [%g3 + TRAP_ENT_F2]%asi
2617	stxa	%g6, [%g3 + TRAP_ENT_F3]%asi
2618	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2619	TRACE_NEXT(%g3, %g4, %g5)
2620	jmp	%g7 + 4
2621	nop
2622
2623trace_gen:
2624	TRACE_PTR(%g3, %g6)
2625	GET_TRACE_TICK(%g6)
2626	stxa	%g6, [%g3 + TRAP_ENT_TICK]%asi
2627	TRACE_SAVE_TL_GL_REGS(%g3, %g6)
2628	rdpr	%tt, %g6
2629	stha	%g6, [%g3 + TRAP_ENT_TT]%asi
2630	rdpr	%tstate, %g6
2631	stxa	%g6, [%g3 + TRAP_ENT_TSTATE]%asi
2632	stna	%sp, [%g3 + TRAP_ENT_SP]%asi
2633	rdpr	%tpc, %g6
2634	stna	%g6, [%g3 + TRAP_ENT_TPC]%asi
2635	stna	%g0, [%g3 + TRAP_ENT_TR]%asi
2636	stna	%g0, [%g3 + TRAP_ENT_F1]%asi
2637	stna	%g0, [%g3 + TRAP_ENT_F2]%asi
2638	stna	%g0, [%g3 + TRAP_ENT_F3]%asi
2639	stna	%g0, [%g3 + TRAP_ENT_F4]%asi
2640	TRACE_NEXT(%g3, %g4, %g5)
2641	jmp	%g7 + 4
2642	nop
2643
2644trace_win:
2645	TRACE_WIN_INFO(0, %l0, %l1, %l2)
2646	! Keep the locals as clean as possible, caller cleans %l4
2647	clr	%l2
2648	clr	%l1
2649	jmp	%l4 + 4
2650	  clr	%l0
2651
2652/*
2653 * Trace a tsb hit
2654 * g1 = tsbe pointer (in/clobbered)
2655 * g2 = tag access register (in)
2656 * g3 - g4 = scratch (clobbered)
2657 * g5 = tsbe data (in)
2658 * g6 = scratch (clobbered)
2659 * g7 = pc we jumped here from (in)
2660 */
2661
2662	! Do not disturb %g5, it will be used after the trace
2663	ALTENTRY(trace_tsbhit)
2664	TRACE_TSBHIT(0)
2665	jmp	%g7 + 4
2666	nop
2667
2668/*
2669 * Trace a TSB miss
2670 *
2671 * g1 = tsb8k pointer (in)
2672 * g2 = tag access register (in)
2673 * g3 = tsb4m pointer (in)
2674 * g4 = tsbe tag (in/clobbered)
2675 * g5 - g6 = scratch (clobbered)
2676 * g7 = pc we jumped here from (in)
2677 */
2678	.global	trace_tsbmiss
2679trace_tsbmiss:
2680	membar	#Sync
2681	sethi	%hi(FLUSH_ADDR), %g6
2682	flush	%g6
2683	TRACE_PTR(%g5, %g6)
2684	GET_TRACE_TICK(%g6)
2685	stxa	%g6, [%g5 + TRAP_ENT_TICK]%asi
2686	stna	%g2, [%g5 + TRAP_ENT_SP]%asi		! tag access
2687	stna	%g4, [%g5 + TRAP_ENT_F1]%asi		! XXX? tsb tag
2688	rdpr	%tnpc, %g6
2689	stna	%g6, [%g5 + TRAP_ENT_F2]%asi
2690	stna	%g1, [%g5 + TRAP_ENT_F3]%asi		! tsb8k pointer
2691	srlx	%g1, 32, %g6
2692	stna	%g6, [%g5 + TRAP_ENT_F4]%asi		! huh?
2693	rdpr	%tpc, %g6
2694	stna	%g6, [%g5 + TRAP_ENT_TPC]%asi
2695	TRACE_SAVE_TL_GL_REGS(%g5, %g6)
2696	rdpr	%tt, %g6
2697	or	%g6, TT_MMU_MISS, %g4
2698	stha	%g4, [%g5 + TRAP_ENT_TT]%asi
2699	MMU_FAULT_STATUS_AREA(%g4)
2700	cmp	%g6, FAST_IMMU_MISS_TT
2701	be,a	%icc, 1f
2702	  ldx	[%g4 + MMFSA_I_ADDR], %g6
2703	cmp	%g6, T_INSTR_MMU_MISS
2704	be,a	%icc, 1f
2705	  ldx	[%g4 + MMFSA_I_ADDR], %g6
2706	ldx	[%g4 + MMFSA_D_ADDR], %g6
27071:	stxa	%g6, [%g5 + TRAP_ENT_TSTATE]%asi	! tag target
2708	stna	%g3, [%g5 + TRAP_ENT_TR]%asi		! tsb4m pointer
2709	TRACE_NEXT(%g5, %g4, %g6)
2710	jmp	%g7 + 4
2711	nop
2712
2713/*
2714 * g2 = tag access register (in)
2715 * g3 = ctx number (in)
2716 */
2717trace_dataprot:
2718	membar	#Sync
2719	sethi	%hi(FLUSH_ADDR), %g6
2720	flush	%g6
2721	TRACE_PTR(%g1, %g6)
2722	GET_TRACE_TICK(%g6)
2723	stxa	%g6, [%g1 + TRAP_ENT_TICK]%asi
2724	rdpr	%tpc, %g6
2725	stna	%g6, [%g1 + TRAP_ENT_TPC]%asi
2726	rdpr	%tstate, %g6
2727	stxa	%g6, [%g1 + TRAP_ENT_TSTATE]%asi
2728	stna	%g2, [%g1 + TRAP_ENT_SP]%asi		! tag access reg
2729	stna	%g0, [%g1 + TRAP_ENT_TR]%asi
2730	stna	%g0, [%g1 + TRAP_ENT_F1]%asi
2731	stna	%g0, [%g1 + TRAP_ENT_F2]%asi
2732	stna	%g0, [%g1 + TRAP_ENT_F3]%asi
2733	stna	%g0, [%g1 + TRAP_ENT_F4]%asi
2734	TRACE_SAVE_TL_GL_REGS(%g1, %g6)
2735	rdpr	%tt, %g6
2736	stha	%g6, [%g1 + TRAP_ENT_TT]%asi
2737	TRACE_NEXT(%g1, %g4, %g5)
2738	jmp	%g7 + 4
2739	nop
2740
2741#endif /* TRAPTRACE */
2742
2743.dmmu_exc_lddf_not_aligned:
2744	MMU_FAULT_STATUS_AREA(%g3)
2745	ldx	[%g3 + MMFSA_D_ADDR], %g2
2746	/* Fault type not available in MMU fault status area */
2747	mov	MMFSA_F_UNALIGN, %g1
2748	ldx	[%g3 + MMFSA_D_CTX], %g3
2749	sllx	%g3, SFSR_CTX_SHIFT, %g3
2750	btst	1, %sp
2751	bnz,pt	%xcc, .lddf_exception_not_aligned
2752	or	%g3, %g1, %g3
2753	ba,a,pt	%xcc, .mmu_exception_not_aligned
2754
2755.dmmu_exc_stdf_not_aligned:
2756	MMU_FAULT_STATUS_AREA(%g3)
2757	ldx	[%g3 + MMFSA_D_ADDR], %g2
2758	/* Fault type not available in MMU fault status area */
2759	mov	MMFSA_F_UNALIGN, %g1
2760	ldx	[%g3 + MMFSA_D_CTX], %g3
2761	sllx	%g3, SFSR_CTX_SHIFT, %g3
2762	btst	1, %sp
2763	bnz,pt	%xcc, .stdf_exception_not_aligned
2764	or	%g3, %g1, %g3
2765	ba,a,pt	%xcc, .mmu_exception_not_aligned
2766
2767/*
2768 * expects offset into tsbmiss area in %g1 and return pc in %g7
2769 */
2770stat_mmu:
2771	CPU_INDEX(%g5, %g6)
2772	sethi	%hi(tsbmiss_area), %g6
2773	sllx	%g5, TSBMISS_SHIFT, %g5
2774	or	%g6, %lo(tsbmiss_area), %g6
2775	add	%g6, %g5, %g6		/* g6 = tsbmiss area */
2776	ld	[%g6 + %g1], %g5
2777	add	%g5, 1, %g5
2778	jmp	%g7 + 4
2779	st	%g5, [%g6 + %g1]
2780
2781
2782/*
2783 * fast_trap_done, fast_trap_done_chk_intr:
2784 *
2785 * Due to the design of UltraSPARC pipeline, pending interrupts are not
2786 * taken immediately after a RETRY or DONE instruction which causes IE to
2787 * go from 0 to 1. Instead, the instruction at %tpc or %tnpc is allowed
2788 * to execute first before taking any interrupts. If that instruction
2789 * results in other traps, and if the corresponding trap handler runs
2790 * entirely at TL=1 with interrupts disabled, then pending interrupts
2791 * won't be taken until after yet another instruction following the %tpc
2792 * or %tnpc.
2793 *
2794 * A malicious user program can use this feature to block out interrupts
2795 * for extended durations, which can result in send_mondo_timeout kernel
2796 * panic.
2797 *
2798 * This problem is addressed by servicing any pending interrupts via
2799 * sys_trap before returning back to the user mode from a fast trap
2800 * handler. The "done" instruction within a fast trap handler, which
2801 * runs entirely at TL=1 with interrupts disabled, is replaced with the
2802 * FAST_TRAP_DONE macro, which branches control to this fast_trap_done
2803 * entry point.
2804 *
2805 * We check for any pending interrupts here and force a sys_trap to
2806 * service those interrupts, if any. To minimize overhead, pending
2807 * interrupts are checked if the %tpc happens to be at 16K boundary,
2808 * which allows a malicious program to execute at most 4K consecutive
2809 * instructions before we service any pending interrupts. If a worst
2810 * case fast trap handler takes about 2 usec, then interrupts will be
2811 * blocked for at most 8 msec, less than a clock tick.
2812 *
2813 * For the cases where we don't know if the %tpc will cross a 16K
2814 * boundary, we can't use the above optimization and always process
2815 * any pending interrupts via fast_frap_done_chk_intr entry point.
2816 *
2817 * Entry Conditions:
2818 * 	%pstate		am:0 priv:1 ie:0
2819 * 			globals are AG (not normal globals)
2820 */
2821
2822	.global	fast_trap_done, fast_trap_done_chk_intr
2823fast_trap_done:
2824	rdpr	%tpc, %g5
2825	sethi	%hi(0xffffc000), %g6	! 1's complement of 0x3fff
2826	andncc	%g5, %g6, %g0		! check lower 14 bits of %tpc
2827	bz,pn	%icc, 1f		! branch if zero (lower 32 bits only)
2828	nop
2829	done
2830
2831fast_trap_done_chk_intr:
28321:	rd	SOFTINT, %g6
2833	brnz,pn	%g6, 2f		! branch if any pending intr
2834	nop
2835	done
2836
28372:
2838	/*
2839	 * We get here if there are any pending interrupts.
2840	 * Adjust %tpc/%tnpc as we'll be resuming via "retry"
2841	 * instruction.
2842	 */
2843	rdpr	%tnpc, %g5
2844	wrpr	%g0, %g5, %tpc
2845	add	%g5, 4, %g5
2846	wrpr	%g0, %g5, %tnpc
2847
2848	/*
2849	 * Force a dummy sys_trap call so that interrupts can be serviced.
2850	 */
2851	set	fast_trap_dummy_call, %g1
2852	ba,pt	%xcc, sys_trap
2853	  mov	-1, %g4
2854
2855fast_trap_dummy_call:
2856	retl
2857	nop
2858
2859#endif	/* lint */
2860