xref: /titanic_44/usr/src/uts/intel/kdi/amd64/kdi_asm.s (revision 3289e1bdec732dbdf04517907516fd0d00f6a6f8)
1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22/*
23 * Copyright 2007 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/*
30 * Debugger entry for both master and slave CPUs
31 */
32
33#if defined(__lint)
34#include <sys/types.h>
35#endif
36
37#include <sys/segments.h>
38#include <sys/asm_linkage.h>
39#include <sys/controlregs.h>
40#include <sys/x86_archext.h>
41#include <sys/privregs.h>
42#include <sys/machprivregs.h>
43#include <sys/kdi_regs.h>
44#include <sys/psw.h>
45#include <sys/uadmin.h>
46#ifdef __xpv
47#include <sys/hypervisor.h>
48#endif
49
50#ifdef _ASM
51
52#include <kdi_assym.h>
53#include <assym.h>
54
55/* clobbers %rdx, %rcx, returns addr in %rax, CPU ID in %rbx */
56#define	GET_CPUSAVE_ADDR \
57	movzbq	%gs:CPU_ID, %rbx;		\
58	movq	%rbx, %rax;			\
59	movq	$KRS_SIZE, %rcx;		\
60	mulq	%rcx;				\
61	movq	$kdi_cpusave, %rdx;		\
62	/*CSTYLED*/				\
63	addq	(%rdx), %rax
64
65/*
66 * Save copies of the IDT and GDT descriptors.  Note that we only save the IDT
67 * and GDT if the IDT isn't ours, as we may be legitimately re-entering the
68 * debugger through the trap handler.  We don't want to clobber the saved IDT
69 * in the process, as we'd end up resuming the world on our IDT.
70 */
71#define	SAVE_IDTGDT				\
72	movq	%gs:CPU_IDT, %r11;		\
73	leaq    kdi_idt(%rip), %rsi;		\
74	cmpq	%rsi, %r11;			\
75	je	1f;				\
76	movq	%r11, KRS_IDT(%rax);		\
77	movq	%gs:CPU_GDT, %r11;		\
78	movq	%r11, KRS_GDT(%rax);		\
791:
80
81#ifdef __xpv
82
83#define	SAVE_GSBASE(reg) /* nothing */
84#define	RESTORE_GSBASE(reg) /* nothing */
85
86#else
87
88#define	SAVE_GSBASE(base)				\
89	movl	$MSR_AMD_GSBASE, %ecx;			\
90	rdmsr;						\
91	shlq	$32, %rdx;				\
92	orq	%rax, %rdx;				\
93	movq	%rdx, REG_OFF(KDIREG_GSBASE)(base)
94
95#define	RESTORE_GSBASE(base)				\
96	movq	REG_OFF(KDIREG_GSBASE)(base), %rdx;	\
97	movq	%rdx, %rax;				\
98	shrq	$32, %rdx;				\
99	movl	$MSR_AMD_GSBASE, %ecx;			\
100	wrmsr
101
102#endif /* __xpv */
103
104/*
105 * %ss, %rsp, %rflags, %cs, %rip, %err, %trapno are already on the stack.  Note
106 * that on the hypervisor, we skip the save/restore of GSBASE: it's slow, and
107 * unnecessary.
108 */
109#define	KDI_SAVE_REGS(base) \
110	movq	%rdi, REG_OFF(KDIREG_RDI)(base);	\
111	movq	%rsi, REG_OFF(KDIREG_RSI)(base);	\
112	movq	%rdx, REG_OFF(KDIREG_RDX)(base);	\
113	movq	%rcx, REG_OFF(KDIREG_RCX)(base);	\
114	movq	%r8, REG_OFF(KDIREG_R8)(base);		\
115	movq	%r9, REG_OFF(KDIREG_R9)(base);		\
116	movq	%rax, REG_OFF(KDIREG_RAX)(base);	\
117	movq	%rbx, REG_OFF(KDIREG_RBX)(base);	\
118	movq	%rbp, REG_OFF(KDIREG_RBP)(base);	\
119	movq	%r10, REG_OFF(KDIREG_R10)(base);	\
120	movq	%r11, REG_OFF(KDIREG_R11)(base);	\
121	movq	%r12, REG_OFF(KDIREG_R12)(base);	\
122	movq	%r13, REG_OFF(KDIREG_R13)(base);	\
123	movq	%r14, REG_OFF(KDIREG_R14)(base);	\
124	movq	%r15, REG_OFF(KDIREG_R15)(base);	\
125	movq	%rbp, REG_OFF(KDIREG_SAVFP)(base);	\
126	movq	REG_OFF(KDIREG_RIP)(base), %rax;	\
127	movq	%rax, REG_OFF(KDIREG_SAVPC)(base);	\
128	clrq	%rax;					\
129	movw	%ds, %ax;				\
130	movq	%rax, REG_OFF(KDIREG_DS)(base);		\
131	movw	%es, %ax;				\
132	movq	%rax, REG_OFF(KDIREG_ES)(base);		\
133	movw	%fs, %ax;				\
134	movq	%rax, REG_OFF(KDIREG_FS)(base);		\
135	movw	%gs, %ax;				\
136	movq	%rax, REG_OFF(KDIREG_GS)(base);		\
137	SAVE_GSBASE(base)
138
139#define	KDI_RESTORE_REGS(base) \
140	movq	base, %rdi;				\
141	RESTORE_GSBASE(%rdi);				\
142	movq	REG_OFF(KDIREG_ES)(%rdi), %rax;		\
143	movw	%ax, %es;				\
144	movq	REG_OFF(KDIREG_DS)(%rdi), %rax;		\
145	movw	%ax, %ds;				\
146	movq	REG_OFF(KDIREG_R15)(%rdi), %r15;	\
147	movq	REG_OFF(KDIREG_R14)(%rdi), %r14;	\
148	movq	REG_OFF(KDIREG_R13)(%rdi), %r13;	\
149	movq	REG_OFF(KDIREG_R12)(%rdi), %r12;	\
150	movq	REG_OFF(KDIREG_R11)(%rdi), %r11;	\
151	movq	REG_OFF(KDIREG_R10)(%rdi), %r10;	\
152	movq	REG_OFF(KDIREG_RBP)(%rdi), %rbp;	\
153	movq	REG_OFF(KDIREG_RBX)(%rdi), %rbx;	\
154	movq	REG_OFF(KDIREG_RAX)(%rdi), %rax;	\
155	movq	REG_OFF(KDIREG_R9)(%rdi), %r9;		\
156	movq	REG_OFF(KDIREG_R8)(%rdi), %r8;		\
157	movq	REG_OFF(KDIREG_RCX)(%rdi), %rcx;	\
158	movq	REG_OFF(KDIREG_RDX)(%rdi), %rdx;	\
159	movq	REG_OFF(KDIREG_RSI)(%rdi), %rsi;	\
160	movq	REG_OFF(KDIREG_RDI)(%rdi), %rdi
161
162/*
163 * Given the address of the current CPU's cpusave area in %rax, the following
164 * macro restores the debugging state to said CPU.  Restored state includes
165 * the debug registers from the global %dr variables, and debugging MSRs from
166 * the CPU save area.  This code would be in a separate routine, but for the
167 * fact that some of the MSRs are jump-sensitive.  As such, we need to minimize
168 * the number of jumps taken subsequent to the update of said MSRs.  We can
169 * remove one jump (the ret) by using a macro instead of a function for the
170 * debugging state restoration code.
171 *
172 * Takes the cpusave area in %rdi as a parameter, clobbers %rax-%rdx
173 */
174#define	KDI_RESTORE_DEBUGGING_STATE \
175	pushq	%rdi;						\
176	leaq	kdi_drreg(%rip), %r15;				\
177	movl	$7, %edi;					\
178	movq	DR_CTL(%r15), %rsi;				\
179	call	kdi_dreg_set;					\
180								\
181	movl	$6, %edi;					\
182	movq	$KDIREG_DRSTAT_RESERVED, %rsi;			\
183	call	kdi_dreg_set;					\
184								\
185	movl	$0, %edi;					\
186	movq	DRADDR_OFF(0)(%r15), %rsi;			\
187	call	kdi_dreg_set;					\
188	movl	$1, %edi;					\
189	movq	DRADDR_OFF(1)(%r15), %rsi;			\
190	call	kdi_dreg_set;					\
191	movl	$2, %edi;					\
192	movq	DRADDR_OFF(2)(%r15), %rsi;			\
193	call	kdi_dreg_set;					\
194	movl	$3, %edi;					\
195	movq	DRADDR_OFF(3)(%r15), %rsi;			\
196	call	kdi_dreg_set;					\
197	popq	%rdi;						\
198								\
199	/*							\
200	 * Write any requested MSRs.				\
201	 */							\
202	movq	KRS_MSR(%rdi), %rbx;				\
203	cmpq	$0, %rbx;					\
204	je	3f;						\
2051:								\
206	movl	MSR_NUM(%rbx), %ecx;				\
207	cmpl	$0, %ecx;					\
208	je	3f;						\
209								\
210	movl	MSR_TYPE(%rbx), %edx;				\
211	cmpl	$KDI_MSR_WRITE, %edx;				\
212	jne	2f;						\
213								\
214	movq	MSR_VALP(%rbx), %rdx;				\
215	movl	0(%rdx), %eax;					\
216	movl	4(%rdx), %edx;					\
217	wrmsr;							\
2182:								\
219	addq	$MSR_SIZE, %rbx;				\
220	jmp	1b;						\
2213:								\
222	/*							\
223	 * We must not branch after re-enabling LBR.  If	\
224	 * kdi_wsr_wrexit_msr is set, it contains the number	\
225	 * of the MSR that controls LBR.  kdi_wsr_wrexit_valp	\
226	 * contains the value that is to be written to enable	\
227	 * LBR.							\
228	 */							\
229	leaq	kdi_msr_wrexit_msr(%rip), %rcx;			\
230	movl	(%rcx), %ecx;					\
231	cmpl	$0, %ecx;					\
232	je	1f;						\
233								\
234	leaq	kdi_msr_wrexit_valp(%rip), %rdx;		\
235	movq	(%rdx), %rdx;					\
236	movl	0(%rdx), %eax;					\
237	movl	4(%rdx), %edx;					\
238								\
239	wrmsr;							\
2401:
241
242/*
243 * Each cpusave buffer has an area set aside for a ring buffer of breadcrumbs.
244 * The following macros manage the buffer.
245 */
246
247/* Advance the ring buffer */
248#define	ADVANCE_CRUMB_POINTER(cpusave, tmp1, tmp2) \
249	movq	KRS_CURCRUMBIDX(cpusave), tmp1;	\
250	cmpq	$[KDI_NCRUMBS - 1], tmp1;	\
251	jge	1f;				\
252	/* Advance the pointer and index */	\
253	addq	$1, tmp1;			\
254	movq	tmp1, KRS_CURCRUMBIDX(cpusave);	\
255	movq	KRS_CURCRUMB(cpusave), tmp1;	\
256	addq	$KRM_SIZE, tmp1;		\
257	jmp	2f;				\
2581:	/* Reset the pointer and index */	\
259	movq	$0, KRS_CURCRUMBIDX(cpusave);	\
260	leaq	KRS_CRUMBS(cpusave), tmp1;	\
2612:	movq	tmp1, KRS_CURCRUMB(cpusave);	\
262	/* Clear the new crumb */		\
263	movq	$KDI_NCRUMBS, tmp2;		\
2643:	movq	$0, -4(tmp1, tmp2, 4);		\
265	decq	tmp2;				\
266	jnz	3b
267
268/* Set a value in the current breadcrumb buffer */
269#define	ADD_CRUMB(cpusave, offset, value, tmp) \
270	movq	KRS_CURCRUMB(cpusave), tmp;	\
271	movq	value, offset(tmp)
272
273#endif	/* _ASM */
274
275#if defined(__lint)
276void
277kdi_cmnint(void)
278{
279}
280#else	/* __lint */
281
282	/* XXX implement me */
283	ENTRY_NP(kdi_nmiint)
284	clrq	%rcx
285	movq	(%rcx), %rcx
286	SET_SIZE(kdi_nmiint)
287
288	/*
289	 * The main entry point for master CPUs.  It also serves as the trap
290	 * handler for all traps and interrupts taken during single-step.
291	 */
292	ENTRY_NP(kdi_cmnint)
293	ALTENTRY(kdi_master_entry)
294
295	pushq	%rax
296	CLI(%rax)
297	popq	%rax
298
299	/* Save current register state */
300	subq	$REG_OFF(KDIREG_TRAPNO), %rsp
301	KDI_SAVE_REGS(%rsp)
302
303#ifdef __xpv
304	/*
305	 * Clear saved_upcall_mask in unused byte of cs slot on stack.
306	 * It can only confuse things.
307	 */
308	movb	$0, REG_OFF(KDIREG_CS)+4(%rsp)
309#endif
310
311#if !defined(__xpv)
312	/*
313	 * Switch to the kernel's GSBASE.  Neither GSBASE nor the ill-named
314	 * KGSBASE can be trusted, as the kernel may or may not have already
315	 * done a swapgs.  All is not lost, as the kernel can divine the correct
316	 * value for us.  Note that the previous GSBASE is saved in the
317	 * KDI_SAVE_REGS macro to prevent a usermode process's GSBASE from being
318	 * blown away.  On the hypervisor, we don't need to do this, since it's
319	 * ensured we're on our requested kernel GSBASE already.
320	 */
321	subq	$10, %rsp
322	sgdt	(%rsp)
323	movq	2(%rsp), %rdi	/* gdt base now in %rdi */
324	addq	$10, %rsp
325	call	kdi_gdt2gsbase	/* returns kernel's GSBASE in %rax */
326
327	movq	%rax, %rdx
328	shrq	$32, %rdx
329	movl	$MSR_AMD_GSBASE, %ecx
330	wrmsr
331#endif	/* __xpv */
332
333	GET_CPUSAVE_ADDR	/* %rax = cpusave, %rbx = CPU ID */
334
335	ADVANCE_CRUMB_POINTER(%rax, %rcx, %rdx)
336
337	ADD_CRUMB(%rax, KRM_CPU_STATE, $KDI_CPU_STATE_MASTER, %rdx)
338
339	movq	REG_OFF(KDIREG_RIP)(%rsp), %rcx
340	ADD_CRUMB(%rax, KRM_PC, %rcx, %rdx)
341	ADD_CRUMB(%rax, KRM_SP, %rsp, %rdx)
342	movq	REG_OFF(KDIREG_TRAPNO)(%rsp), %rcx
343	ADD_CRUMB(%rax, KRM_TRAPNO, %rcx, %rdx)
344
345	movq	%rsp, %rbp
346	pushq	%rax
347
348	/*
349	 * Were we in the debugger when we took the trap (i.e. was %esp in one
350	 * of the debugger's memory ranges)?
351	 */
352	leaq	kdi_memranges, %rcx
353	movl	kdi_nmemranges, %edx
3541:	cmpq	MR_BASE(%rcx), %rsp
355	jl	2f		/* below this range -- try the next one */
356	cmpq	MR_LIM(%rcx), %rsp
357	jg	2f		/* above this range -- try the next one */
358	jmp	3f		/* matched within this range */
359
3602:	decl	%edx
361	jz	kdi_save_common_state	/* %rsp not within debugger memory */
362	addq	$MR_SIZE, %rcx
363	jmp	1b
364
3653:	/*
366	 * The master is still set.  That should only happen if we hit a trap
367	 * while running in the debugger.  Note that it may be an intentional
368	 * fault.  kmdb_dpi_handle_fault will sort it all out.
369	 */
370
371	movq	REG_OFF(KDIREG_TRAPNO)(%rbp), %rdi
372	movq	REG_OFF(KDIREG_RIP)(%rbp), %rsi
373	movq	REG_OFF(KDIREG_RSP)(%rbp), %rdx
374	movq	%rbx, %rcx		/* cpuid */
375
376	call	kdi_dvec_handle_fault
377
378	/*
379	 * If we're here, we ran into a debugger problem, and the user
380	 * elected to solve it by having the debugger debug itself.  The
381	 * state we're about to save is that of the debugger when it took
382	 * the fault.
383	 */
384
385	jmp	kdi_save_common_state
386
387	SET_SIZE(kdi_master_entry)
388	SET_SIZE(kdi_cmnint)
389
390#endif	/* __lint */
391
392/*
393 * The cross-call handler for slave CPUs.
394 *
395 * The debugger is single-threaded, so only one CPU, called the master, may be
396 * running it at any given time.  The other CPUs, known as slaves, spin in a
397 * busy loop until there's something for them to do.  This is the entry point
398 * for the slaves - they'll be sent here in response to a cross-call sent by the
399 * master.
400 */
401
402#if defined(__lint)
403char kdi_slave_entry_patch;
404
405void
406kdi_slave_entry(void)
407{
408}
409#else /* __lint */
410	.globl	kdi_slave_entry_patch;
411
412	ENTRY_NP(kdi_slave_entry)
413
414	/* kdi_msr_add_clrentry knows where this is */
415kdi_slave_entry_patch:
416	KDI_MSR_PATCH;
417
418	/*
419	 * Cross calls are implemented as function calls, so our stack currently
420	 * looks like one you'd get from a zero-argument function call.  That
421	 * is, there's the return %rip at %rsp, and that's about it.  We need
422	 * to make it look like an interrupt stack.  When we first save, we'll
423	 * reverse the saved %ss and %rip, which we'll fix back up when we've
424	 * freed up some general-purpose registers.  We'll also need to fix up
425	 * the saved %rsp.
426	 */
427
428	pushq	%rsp		/* pushed value off by 8 */
429	pushfq
430	CLI(%rax)
431	pushq	$KCS_SEL
432	clrq	%rax
433	movw	%ss, %ax
434	pushq	%rax		/* rip should be here */
435	pushq	$-1		/* phony trap error code */
436	pushq	$-1		/* phony trap number */
437
438	subq	$REG_OFF(KDIREG_TRAPNO), %rsp
439	KDI_SAVE_REGS(%rsp)
440
441	movq	REG_OFF(KDIREG_SS)(%rsp), %rax
442	xchgq	REG_OFF(KDIREG_RIP)(%rsp), %rax
443	movq	%rax, REG_OFF(KDIREG_SS)(%rsp)
444
445	movq	REG_OFF(KDIREG_RSP)(%rsp), %rax
446	addq	$8, %rax
447	movq	%rax, REG_OFF(KDIREG_RSP)(%rsp)
448
449	/*
450	 * We've saved all of the general-purpose registers, and have a stack
451	 * that is irettable (after we strip down to the error code)
452	 */
453
454	GET_CPUSAVE_ADDR	/* %rax = cpusave, %rbx = CPU ID */
455
456	ADVANCE_CRUMB_POINTER(%rax, %rcx, %rdx)
457
458	ADD_CRUMB(%rax, KRM_CPU_STATE, $KDI_CPU_STATE_SLAVE, %rdx)
459
460	movq	REG_OFF(KDIREG_RIP)(%rsp), %rcx
461	ADD_CRUMB(%rax, KRM_PC, %rcx, %rdx)
462
463	pushq	%rax
464	jmp	kdi_save_common_state
465
466	SET_SIZE(kdi_slave_entry)
467
468#endif	/* __lint */
469
470/*
471 * The state of the world:
472 *
473 * The stack has a complete set of saved registers and segment
474 * selectors, arranged in the kdi_regs.h order.  It also has a pointer
475 * to our cpusave area.
476 *
477 * We need to save, into the cpusave area, a pointer to these saved
478 * registers.  First we check whether we should jump straight back to
479 * the kernel.  If not, we save a few more registers, ready the
480 * machine for debugger entry, and enter the debugger.
481 */
482
483#if !defined(__lint)
484
485	ENTRY_NP(kdi_save_common_state)
486
487	popq	%rdi			/* the cpusave area */
488	movq	%rsp, KRS_GREGS(%rdi)	/* save ptr to current saved regs */
489
490	pushq	%rdi
491	call	kdi_trap_pass
492	cmpq	$1, %rax
493	je	kdi_pass_to_kernel
494	popq	%rax /* cpusave in %rax */
495
496	SAVE_IDTGDT
497
498#if !defined(__xpv)
499	/* Save off %cr0, and clear write protect */
500	movq	%cr0, %rcx
501	movq	%rcx, KRS_CR0(%rax)
502	andq	$_BITNOT(CR0_WP), %rcx
503	movq	%rcx, %cr0
504#endif
505
506	/* Save the debug registers and disable any active watchpoints */
507
508	movq	%rax, %r15		/* save cpusave area ptr */
509	movl	$7, %edi
510	call	kdi_dreg_get
511	movq	%rax, KRS_DRCTL(%r15)
512
513	andq	$_BITNOT(KDIREG_DRCTL_WPALLEN_MASK), %rax
514	movq	%rax, %rsi
515	movl	$7, %edi
516	call	kdi_dreg_set
517
518	movl	$6, %edi
519	call	kdi_dreg_get
520	movq	%rax, KRS_DRSTAT(%r15)
521
522	movl	$0, %edi
523	call	kdi_dreg_get
524	movq	%rax, KRS_DROFF(0)(%r15)
525
526	movl	$1, %edi
527	call	kdi_dreg_get
528	movq	%rax, KRS_DROFF(1)(%r15)
529
530	movl	$2, %edi
531	call	kdi_dreg_get
532	movq	%rax, KRS_DROFF(2)(%r15)
533
534	movl	$3, %edi
535	call	kdi_dreg_get
536	movq	%rax, KRS_DROFF(3)(%r15)
537
538	movq	%r15, %rax	/* restore cpu save area to rax */
539
540	/*
541	 * Save any requested MSRs.
542	 */
543	movq	KRS_MSR(%rax), %rcx
544	cmpq	$0, %rcx
545	je	no_msr
546
547	pushq	%rax		/* rdmsr clobbers %eax */
548	movq	%rcx, %rbx
549
5501:
551	movl	MSR_NUM(%rbx), %ecx
552	cmpl	$0, %ecx
553	je	msr_done
554
555	movl	MSR_TYPE(%rbx), %edx
556	cmpl	$KDI_MSR_READ, %edx
557	jne	msr_next
558
559	rdmsr			/* addr in %ecx, value into %edx:%eax */
560	movl	%eax, MSR_VAL(%rbx)
561	movl	%edx, _CONST(MSR_VAL + 4)(%rbx)
562
563msr_next:
564	addq	$MSR_SIZE, %rbx
565	jmp	1b
566
567msr_done:
568	popq	%rax
569
570no_msr:
571	clrq	%rbp		/* stack traces should end here */
572
573	pushq	%rax
574	movq	%rax, %rdi	/* cpusave */
575
576	call	kdi_debugger_entry
577
578	/* Pass cpusave to kdi_resume */
579	popq	%rdi
580
581	jmp	kdi_resume
582
583	SET_SIZE(kdi_save_common_state)
584
585#endif	/* !__lint */
586
587/*
588 * Resume the world.  The code that calls kdi_resume has already
589 * decided whether or not to restore the IDT.
590 */
591#if defined(__lint)
592void
593kdi_resume(void)
594{
595}
596#else	/* __lint */
597
598	/* cpusave in %rdi */
599	ENTRY_NP(kdi_resume)
600
601	/*
602	 * Send this CPU back into the world
603	 */
604#if !defined(__xpv)
605	movq	KRS_CR0(%rdi), %rdx
606	movq	%rdx, %cr0
607#endif
608
609	KDI_RESTORE_DEBUGGING_STATE
610
611	movq	KRS_GREGS(%rdi), %rsp
612	KDI_RESTORE_REGS(%rsp)
613	addq	$REG_OFF(KDIREG_RIP), %rsp	/* Discard state, trapno, err */
614	IRET
615	/*NOTREACHED*/
616	SET_SIZE(kdi_resume)
617
618#endif	/* __lint */
619
620#if !defined(__lint)
621
622	ENTRY_NP(kdi_pass_to_kernel)
623
624	popq	%rdi /* cpusave */
625
626	movq	$KDI_CPU_STATE_NONE, KRS_CPU_STATE(%rdi)
627
628	/*
629	 * Find the trap and vector off the right kernel handler.  The trap
630	 * handler will expect the stack to be in trap order, with %rip being
631	 * the last entry, so we'll need to restore all our regs.  On i86xpv
632	 * we'll need to compensate for XPV_TRAP_POP.
633	 *
634	 * We're hard-coding the three cases where KMDB has installed permanent
635	 * handlers, since after we KDI_RESTORE_REGS(), we don't have registers
636	 * to work with; we can't use a global since other CPUs can easily pass
637	 * through here at the same time.
638	 *
639	 * Note that we handle T_DBGENTR since userspace might have tried it.
640	 */
641	movq	KRS_GREGS(%rdi), %rsp
642	movq	REG_OFF(KDIREG_TRAPNO)(%rsp), %rdi
643	cmpq	$T_SGLSTP, %rdi
644	je	1f
645	cmpq	$T_BPTFLT, %rdi
646	je	2f
647	cmpq	$T_DBGENTR, %rdi
648	je	3f
649	/*
650	 * Hmm, unknown handler.  Somebody forgot to update this when they
651	 * added a new trap interposition... try to drop back into kmdb.
652	 */
653	int	$T_DBGENTR
654
655#define	CALL_TRAP_HANDLER(name) \
656	KDI_RESTORE_REGS(%rsp); \
657	/* Discard state, trapno, err */ \
658	addq	$REG_OFF(KDIREG_RIP), %rsp; \
659	XPV_TRAP_PUSH; \
660	jmp	%cs:name
661
6621:
663	CALL_TRAP_HANDLER(dbgtrap)
664	/*NOTREACHED*/
6652:
666	CALL_TRAP_HANDLER(brktrap)
667	/*NOTREACHED*/
6683:
669	CALL_TRAP_HANDLER(invaltrap)
670	/*NOTREACHED*/
671
672	SET_SIZE(kdi_pass_to_kernel)
673
674	/*
675	 * A minimal version of mdboot(), to be used by the master CPU only.
676	 */
677	ENTRY_NP(kdi_reboot)
678
679	movl	$AD_BOOT, %edi
680	movl	$A_SHUTDOWN, %esi
681	call	*psm_shutdownf
682#if defined(__xpv)
683	movl	$SHUTDOWN_reboot, %edi
684	call	HYPERVISOR_shutdown
685#else
686	call	reset
687#endif
688	/*NOTREACHED*/
689
690	SET_SIZE(kdi_reboot)
691
692#endif	/* !__lint */
693
694#if defined(__lint)
695/*ARGSUSED*/
696void
697kdi_cpu_debug_init(kdi_cpusave_t *save)
698{
699}
700#else	/* __lint */
701
702	ENTRY_NP(kdi_cpu_debug_init)
703	pushq	%rbp
704	movq	%rsp, %rbp
705
706	pushq	%rbx		/* macro will clobber %rbx */
707	KDI_RESTORE_DEBUGGING_STATE
708	popq	%rbx
709
710	leave
711	ret
712
713	SET_SIZE(kdi_cpu_debug_init)
714#endif	/* !__lint */
715
716