xref: /linux/arch/powerpc/kernel/exceptions-64s.S (revision 26fbb4c8c7c3ee9a4c3b4de555a8587b5a19154e)
1/* SPDX-License-Identifier: GPL-2.0 */
2/*
3 * This file contains the 64-bit "server" PowerPC variant
4 * of the low level exception handling including exception
5 * vectors, exception return, part of the slb and stab
6 * handling and other fixed offset specific things.
7 *
8 * This file is meant to be #included from head_64.S due to
9 * position dependent assembly.
10 *
11 * Most of this originates from head_64.S and thus has the same
12 * copyright history.
13 *
14 */
15
16#include <asm/hw_irq.h>
17#include <asm/exception-64s.h>
18#include <asm/ptrace.h>
19#include <asm/cpuidle.h>
20#include <asm/head-64.h>
21#include <asm/feature-fixups.h>
22#include <asm/kup.h>
23
24/* PACA save area offsets (exgen, exmc, etc) */
25#define EX_R9		0
26#define EX_R10		8
27#define EX_R11		16
28#define EX_R12		24
29#define EX_R13		32
30#define EX_DAR		40
31#define EX_DSISR	48
32#define EX_CCR		52
33#define EX_CFAR		56
34#define EX_PPR		64
35#define EX_CTR		72
36.if EX_SIZE != 10
37	.error "EX_SIZE is wrong"
38.endif
39
40/*
41 * Following are fixed section helper macros.
42 *
43 * EXC_REAL_BEGIN/END  - real, unrelocated exception vectors
44 * EXC_VIRT_BEGIN/END  - virt (AIL), unrelocated exception vectors
45 * TRAMP_REAL_BEGIN    - real, unrelocated helpers (virt may call these)
46 * TRAMP_VIRT_BEGIN    - virt, unreloc helpers (in practice, real can use)
47 * EXC_COMMON          - After switching to virtual, relocated mode.
48 */
49
50#define EXC_REAL_BEGIN(name, start, size)			\
51	FIXED_SECTION_ENTRY_BEGIN_LOCATION(real_vectors, exc_real_##start##_##name, start, size)
52
53#define EXC_REAL_END(name, start, size)				\
54	FIXED_SECTION_ENTRY_END_LOCATION(real_vectors, exc_real_##start##_##name, start, size)
55
56#define EXC_VIRT_BEGIN(name, start, size)			\
57	FIXED_SECTION_ENTRY_BEGIN_LOCATION(virt_vectors, exc_virt_##start##_##name, start, size)
58
59#define EXC_VIRT_END(name, start, size)				\
60	FIXED_SECTION_ENTRY_END_LOCATION(virt_vectors, exc_virt_##start##_##name, start, size)
61
62#define EXC_COMMON_BEGIN(name)					\
63	USE_TEXT_SECTION();					\
64	.balign IFETCH_ALIGN_BYTES;				\
65	.global name;						\
66	_ASM_NOKPROBE_SYMBOL(name);				\
67	DEFINE_FIXED_SYMBOL(name);				\
68name:
69
70#define TRAMP_REAL_BEGIN(name)					\
71	FIXED_SECTION_ENTRY_BEGIN(real_trampolines, name)
72
73#define TRAMP_VIRT_BEGIN(name)					\
74	FIXED_SECTION_ENTRY_BEGIN(virt_trampolines, name)
75
76#define EXC_REAL_NONE(start, size)				\
77	FIXED_SECTION_ENTRY_BEGIN_LOCATION(real_vectors, exc_real_##start##_##unused, start, size); \
78	FIXED_SECTION_ENTRY_END_LOCATION(real_vectors, exc_real_##start##_##unused, start, size)
79
80#define EXC_VIRT_NONE(start, size)				\
81	FIXED_SECTION_ENTRY_BEGIN_LOCATION(virt_vectors, exc_virt_##start##_##unused, start, size); \
82	FIXED_SECTION_ENTRY_END_LOCATION(virt_vectors, exc_virt_##start##_##unused, start, size)
83
84/*
85 * We're short on space and time in the exception prolog, so we can't
86 * use the normal LOAD_REG_IMMEDIATE macro to load the address of label.
87 * Instead we get the base of the kernel from paca->kernelbase and or in the low
88 * part of label. This requires that the label be within 64KB of kernelbase, and
89 * that kernelbase be 64K aligned.
90 */
91#define LOAD_HANDLER(reg, label)					\
92	ld	reg,PACAKBASE(r13);	/* get high part of &label */	\
93	ori	reg,reg,FIXED_SYMBOL_ABS_ADDR(label)
94
95#define __LOAD_HANDLER(reg, label)					\
96	ld	reg,PACAKBASE(r13);					\
97	ori	reg,reg,(ABS_ADDR(label))@l
98
99/*
100 * Branches from unrelocated code (e.g., interrupts) to labels outside
101 * head-y require >64K offsets.
102 */
103#define __LOAD_FAR_HANDLER(reg, label)					\
104	ld	reg,PACAKBASE(r13);					\
105	ori	reg,reg,(ABS_ADDR(label))@l;				\
106	addis	reg,reg,(ABS_ADDR(label))@h
107
108/*
109 * Branch to label using its 0xC000 address. This results in instruction
110 * address suitable for MSR[IR]=0 or 1, which allows relocation to be turned
111 * on using mtmsr rather than rfid.
112 *
113 * This could set the 0xc bits for !RELOCATABLE as an immediate, rather than
114 * load KBASE for a slight optimisation.
115 */
116#define BRANCH_TO_C000(reg, label)					\
117	__LOAD_FAR_HANDLER(reg, label);					\
118	mtctr	reg;							\
119	bctr
120
121/*
122 * Interrupt code generation macros
123 */
124#define IVEC		.L_IVEC_\name\()	/* Interrupt vector address */
125#define IHSRR		.L_IHSRR_\name\()	/* Sets SRR or HSRR registers */
126#define IHSRR_IF_HVMODE	.L_IHSRR_IF_HVMODE_\name\() /* HSRR if HV else SRR */
127#define IAREA		.L_IAREA_\name\()	/* PACA save area */
128#define IVIRT		.L_IVIRT_\name\()	/* Has virt mode entry point */
129#define IISIDE		.L_IISIDE_\name\()	/* Uses SRR0/1 not DAR/DSISR */
130#define IDAR		.L_IDAR_\name\()	/* Uses DAR (or SRR0) */
131#define IDSISR		.L_IDSISR_\name\()	/* Uses DSISR (or SRR1) */
132#define ISET_RI		.L_ISET_RI_\name\()	/* Run common code w/ MSR[RI]=1 */
133#define IBRANCH_TO_COMMON	.L_IBRANCH_TO_COMMON_\name\() /* ENTRY branch to common */
134#define IREALMODE_COMMON	.L_IREALMODE_COMMON_\name\() /* Common runs in realmode */
135#define IMASK		.L_IMASK_\name\()	/* IRQ soft-mask bit */
136#define IKVM_SKIP	.L_IKVM_SKIP_\name\()	/* Generate KVM skip handler */
137#define IKVM_REAL	.L_IKVM_REAL_\name\()	/* Real entry tests KVM */
138#define __IKVM_REAL(name)	.L_IKVM_REAL_ ## name
139#define IKVM_VIRT	.L_IKVM_VIRT_\name\()	/* Virt entry tests KVM */
140#define ISTACK		.L_ISTACK_\name\()	/* Set regular kernel stack */
141#define __ISTACK(name)	.L_ISTACK_ ## name
142#define IRECONCILE	.L_IRECONCILE_\name\()	/* Do RECONCILE_IRQ_STATE */
143#define IKUAP		.L_IKUAP_\name\()	/* Do KUAP lock */
144
145#define INT_DEFINE_BEGIN(n)						\
146.macro int_define_ ## n name
147
148#define INT_DEFINE_END(n)						\
149.endm ;									\
150int_define_ ## n n ;							\
151do_define_int n
152
153.macro do_define_int name
154	.ifndef IVEC
155		.error "IVEC not defined"
156	.endif
157	.ifndef IHSRR
158		IHSRR=0
159	.endif
160	.ifndef IHSRR_IF_HVMODE
161		IHSRR_IF_HVMODE=0
162	.endif
163	.ifndef IAREA
164		IAREA=PACA_EXGEN
165	.endif
166	.ifndef IVIRT
167		IVIRT=1
168	.endif
169	.ifndef IISIDE
170		IISIDE=0
171	.endif
172	.ifndef IDAR
173		IDAR=0
174	.endif
175	.ifndef IDSISR
176		IDSISR=0
177	.endif
178	.ifndef ISET_RI
179		ISET_RI=1
180	.endif
181	.ifndef IBRANCH_TO_COMMON
182		IBRANCH_TO_COMMON=1
183	.endif
184	.ifndef IREALMODE_COMMON
185		IREALMODE_COMMON=0
186	.else
187		.if ! IBRANCH_TO_COMMON
188			.error "IREALMODE_COMMON=1 but IBRANCH_TO_COMMON=0"
189		.endif
190	.endif
191	.ifndef IMASK
192		IMASK=0
193	.endif
194	.ifndef IKVM_SKIP
195		IKVM_SKIP=0
196	.endif
197	.ifndef IKVM_REAL
198		IKVM_REAL=0
199	.endif
200	.ifndef IKVM_VIRT
201		IKVM_VIRT=0
202	.endif
203	.ifndef ISTACK
204		ISTACK=1
205	.endif
206	.ifndef IRECONCILE
207		IRECONCILE=1
208	.endif
209	.ifndef IKUAP
210		IKUAP=1
211	.endif
212.endm
213
214#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
215#ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
216/*
217 * All interrupts which set HSRR registers, as well as SRESET and MCE and
218 * syscall when invoked with "sc 1" switch to MSR[HV]=1 (HVMODE) to be taken,
219 * so they all generally need to test whether they were taken in guest context.
220 *
221 * Note: SRESET and MCE may also be sent to the guest by the hypervisor, and be
222 * taken with MSR[HV]=0.
223 *
224 * Interrupts which set SRR registers (with the above exceptions) do not
225 * elevate to MSR[HV]=1 mode, though most can be taken when running with
226 * MSR[HV]=1  (e.g., bare metal kernel and userspace). So these interrupts do
227 * not need to test whether a guest is running because they get delivered to
228 * the guest directly, including nested HV KVM guests.
229 *
230 * The exception is PR KVM, where the guest runs with MSR[PR]=1 and the host
231 * runs with MSR[HV]=0, so the host takes all interrupts on behalf of the
232 * guest. PR KVM runs with LPCR[AIL]=0 which causes interrupts to always be
233 * delivered to the real-mode entry point, therefore such interrupts only test
234 * KVM in their real mode handlers, and only when PR KVM is possible.
235 *
236 * Interrupts that are taken in MSR[HV]=0 and escalate to MSR[HV]=1 are always
237 * delivered in real-mode when the MMU is in hash mode because the MMU
238 * registers are not set appropriately to translate host addresses. In nested
239 * radix mode these can be delivered in virt-mode as the host translations are
240 * used implicitly (see: effective LPID, effective PID).
241 */
242
243/*
244 * If an interrupt is taken while a guest is running, it is immediately routed
245 * to KVM to handle. If both HV and PR KVM arepossible, KVM interrupts go first
246 * to kvmppc_interrupt_hv, which handles the PR guest case.
247 */
248#define kvmppc_interrupt kvmppc_interrupt_hv
249#else
250#define kvmppc_interrupt kvmppc_interrupt_pr
251#endif
252
253.macro KVMTEST name
254	lbz	r10,HSTATE_IN_GUEST(r13)
255	cmpwi	r10,0
256	bne	\name\()_kvm
257.endm
258
259.macro GEN_KVM name
260	.balign IFETCH_ALIGN_BYTES
261\name\()_kvm:
262
263	.if IKVM_SKIP
264	cmpwi	r10,KVM_GUEST_MODE_SKIP
265	beq	89f
266	.else
267BEGIN_FTR_SECTION
268	ld	r10,IAREA+EX_CFAR(r13)
269	std	r10,HSTATE_CFAR(r13)
270END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
271	.endif
272
273	ld	r10,IAREA+EX_CTR(r13)
274	mtctr	r10
275BEGIN_FTR_SECTION
276	ld	r10,IAREA+EX_PPR(r13)
277	std	r10,HSTATE_PPR(r13)
278END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
279	ld	r11,IAREA+EX_R11(r13)
280	ld	r12,IAREA+EX_R12(r13)
281	std	r12,HSTATE_SCRATCH0(r13)
282	sldi	r12,r9,32
283	ld	r9,IAREA+EX_R9(r13)
284	ld	r10,IAREA+EX_R10(r13)
285	/* HSRR variants have the 0x2 bit added to their trap number */
286	.if IHSRR_IF_HVMODE
287	BEGIN_FTR_SECTION
288	ori	r12,r12,(IVEC + 0x2)
289	FTR_SECTION_ELSE
290	ori	r12,r12,(IVEC)
291	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
292	.elseif IHSRR
293	ori	r12,r12,(IVEC+ 0x2)
294	.else
295	ori	r12,r12,(IVEC)
296	.endif
297	b	kvmppc_interrupt
298
299	.if IKVM_SKIP
30089:	mtocrf	0x80,r9
301	ld	r10,IAREA+EX_CTR(r13)
302	mtctr	r10
303	ld	r9,IAREA+EX_R9(r13)
304	ld	r10,IAREA+EX_R10(r13)
305	ld	r11,IAREA+EX_R11(r13)
306	ld	r12,IAREA+EX_R12(r13)
307	.if IHSRR_IF_HVMODE
308	BEGIN_FTR_SECTION
309	b	kvmppc_skip_Hinterrupt
310	FTR_SECTION_ELSE
311	b	kvmppc_skip_interrupt
312	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
313	.elseif IHSRR
314	b	kvmppc_skip_Hinterrupt
315	.else
316	b	kvmppc_skip_interrupt
317	.endif
318	.endif
319.endm
320
321#else
322.macro KVMTEST name
323.endm
324.macro GEN_KVM name
325.endm
326#endif
327
328/*
329 * This is the BOOK3S interrupt entry code macro.
330 *
331 * This can result in one of several things happening:
332 * - Branch to the _common handler, relocated, in virtual mode.
333 *   These are normal interrupts (synchronous and asynchronous) handled by
334 *   the kernel.
335 * - Branch to KVM, relocated but real mode interrupts remain in real mode.
336 *   These occur when HSTATE_IN_GUEST is set. The interrupt may be caused by
337 *   / intended for host or guest kernel, but KVM must always be involved
338 *   because the machine state is set for guest execution.
339 * - Branch to the masked handler, unrelocated.
340 *   These occur when maskable asynchronous interrupts are taken with the
341 *   irq_soft_mask set.
342 * - Branch to an "early" handler in real mode but relocated.
343 *   This is done if early=1. MCE and HMI use these to handle errors in real
344 *   mode.
345 * - Fall through and continue executing in real, unrelocated mode.
346 *   This is done if early=2.
347 */
348
349.macro GEN_BRANCH_TO_COMMON name, virt
350	.if IREALMODE_COMMON
351	LOAD_HANDLER(r10, \name\()_common)
352	mtctr	r10
353	bctr
354	.else
355	.if \virt
356#ifndef CONFIG_RELOCATABLE
357	b	\name\()_common_virt
358#else
359	LOAD_HANDLER(r10, \name\()_common_virt)
360	mtctr	r10
361	bctr
362#endif
363	.else
364	LOAD_HANDLER(r10, \name\()_common_real)
365	mtctr	r10
366	bctr
367	.endif
368	.endif
369.endm
370
371.macro GEN_INT_ENTRY name, virt, ool=0
372	SET_SCRATCH0(r13)			/* save r13 */
373	GET_PACA(r13)
374	std	r9,IAREA+EX_R9(r13)		/* save r9 */
375BEGIN_FTR_SECTION
376	mfspr	r9,SPRN_PPR
377END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
378	HMT_MEDIUM
379	std	r10,IAREA+EX_R10(r13)		/* save r10 - r12 */
380BEGIN_FTR_SECTION
381	mfspr	r10,SPRN_CFAR
382END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
383	.if \ool
384	.if !\virt
385	b	tramp_real_\name
386	.pushsection .text
387	TRAMP_REAL_BEGIN(tramp_real_\name)
388	.else
389	b	tramp_virt_\name
390	.pushsection .text
391	TRAMP_VIRT_BEGIN(tramp_virt_\name)
392	.endif
393	.endif
394
395BEGIN_FTR_SECTION
396	std	r9,IAREA+EX_PPR(r13)
397END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
398BEGIN_FTR_SECTION
399	std	r10,IAREA+EX_CFAR(r13)
400END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
401	INTERRUPT_TO_KERNEL
402	mfctr	r10
403	std	r10,IAREA+EX_CTR(r13)
404	mfcr	r9
405	std	r11,IAREA+EX_R11(r13)
406	std	r12,IAREA+EX_R12(r13)
407
408	/*
409	 * DAR/DSISR, SCRATCH0 must be read before setting MSR[RI],
410	 * because a d-side MCE will clobber those registers so is
411	 * not recoverable if they are live.
412	 */
413	GET_SCRATCH0(r10)
414	std	r10,IAREA+EX_R13(r13)
415	.if IDAR && !IISIDE
416	.if IHSRR
417	mfspr	r10,SPRN_HDAR
418	.else
419	mfspr	r10,SPRN_DAR
420	.endif
421	std	r10,IAREA+EX_DAR(r13)
422	.endif
423	.if IDSISR && !IISIDE
424	.if IHSRR
425	mfspr	r10,SPRN_HDSISR
426	.else
427	mfspr	r10,SPRN_DSISR
428	.endif
429	stw	r10,IAREA+EX_DSISR(r13)
430	.endif
431
432	.if IHSRR_IF_HVMODE
433	BEGIN_FTR_SECTION
434	mfspr	r11,SPRN_HSRR0		/* save HSRR0 */
435	mfspr	r12,SPRN_HSRR1		/* and HSRR1 */
436	FTR_SECTION_ELSE
437	mfspr	r11,SPRN_SRR0		/* save SRR0 */
438	mfspr	r12,SPRN_SRR1		/* and SRR1 */
439	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
440	.elseif IHSRR
441	mfspr	r11,SPRN_HSRR0		/* save HSRR0 */
442	mfspr	r12,SPRN_HSRR1		/* and HSRR1 */
443	.else
444	mfspr	r11,SPRN_SRR0		/* save SRR0 */
445	mfspr	r12,SPRN_SRR1		/* and SRR1 */
446	.endif
447
448	.if IBRANCH_TO_COMMON
449	GEN_BRANCH_TO_COMMON \name \virt
450	.endif
451
452	.if \ool
453	.popsection
454	.endif
455.endm
456
457/*
458 * __GEN_COMMON_ENTRY is required to receive the branch from interrupt
459 * entry, except in the case of the real-mode handlers which require
460 * __GEN_REALMODE_COMMON_ENTRY.
461 *
462 * This switches to virtual mode and sets MSR[RI].
463 */
464.macro __GEN_COMMON_ENTRY name
465DEFINE_FIXED_SYMBOL(\name\()_common_real)
466\name\()_common_real:
467	.if IKVM_REAL
468		KVMTEST \name
469	.endif
470
471	ld	r10,PACAKMSR(r13)	/* get MSR value for kernel */
472	/* MSR[RI] is clear iff using SRR regs */
473	.if IHSRR == EXC_HV_OR_STD
474	BEGIN_FTR_SECTION
475	xori	r10,r10,MSR_RI
476	END_FTR_SECTION_IFCLR(CPU_FTR_HVMODE)
477	.elseif ! IHSRR
478	xori	r10,r10,MSR_RI
479	.endif
480	mtmsrd	r10
481
482	.if IVIRT
483	.if IKVM_VIRT
484	b	1f /* skip the virt test coming from real */
485	.endif
486
487	.balign IFETCH_ALIGN_BYTES
488DEFINE_FIXED_SYMBOL(\name\()_common_virt)
489\name\()_common_virt:
490	.if IKVM_VIRT
491		KVMTEST \name
4921:
493	.endif
494	.endif /* IVIRT */
495.endm
496
497/*
498 * Don't switch to virt mode. Used for early MCE and HMI handlers that
499 * want to run in real mode.
500 */
501.macro __GEN_REALMODE_COMMON_ENTRY name
502DEFINE_FIXED_SYMBOL(\name\()_common_real)
503\name\()_common_real:
504	.if IKVM_REAL
505		KVMTEST \name
506	.endif
507.endm
508
509.macro __GEN_COMMON_BODY name
510	.if IMASK
511		.if ! ISTACK
512		.error "No support for masked interrupt to use custom stack"
513		.endif
514
515		/* If coming from user, skip soft-mask tests. */
516		andi.	r10,r12,MSR_PR
517		bne	2f
518
519		/* Kernel code running below __end_interrupts is implicitly
520		 * soft-masked */
521		LOAD_HANDLER(r10, __end_interrupts)
522		cmpld	r11,r10
523		li	r10,IMASK
524		blt-	1f
525
526		/* Test the soft mask state against our interrupt's bit */
527		lbz	r10,PACAIRQSOFTMASK(r13)
5281:		andi.	r10,r10,IMASK
529		/* Associate vector numbers with bits in paca->irq_happened */
530		.if IVEC == 0x500 || IVEC == 0xea0
531		li	r10,PACA_IRQ_EE
532		.elseif IVEC == 0x900
533		li	r10,PACA_IRQ_DEC
534		.elseif IVEC == 0xa00 || IVEC == 0xe80
535		li	r10,PACA_IRQ_DBELL
536		.elseif IVEC == 0xe60
537		li	r10,PACA_IRQ_HMI
538		.elseif IVEC == 0xf00
539		li	r10,PACA_IRQ_PMI
540		.else
541		.abort "Bad maskable vector"
542		.endif
543
544		.if IHSRR_IF_HVMODE
545		BEGIN_FTR_SECTION
546		bne	masked_Hinterrupt
547		FTR_SECTION_ELSE
548		bne	masked_interrupt
549		ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
550		.elseif IHSRR
551		bne	masked_Hinterrupt
552		.else
553		bne	masked_interrupt
554		.endif
555	.endif
556
557	.if ISTACK
558	andi.	r10,r12,MSR_PR		/* See if coming from user	*/
5592:	mr	r10,r1			/* Save r1			*/
560	subi	r1,r1,INT_FRAME_SIZE	/* alloc frame on kernel stack	*/
561	beq-	100f
562	ld	r1,PACAKSAVE(r13)	/* kernel stack to use		*/
563100:	tdgei	r1,-INT_FRAME_SIZE	/* trap if r1 is in userspace	*/
564	EMIT_BUG_ENTRY 100b,__FILE__,__LINE__,0
565	.endif
566
567	std	r9,_CCR(r1)		/* save CR in stackframe	*/
568	std	r11,_NIP(r1)		/* save SRR0 in stackframe	*/
569	std	r12,_MSR(r1)		/* save SRR1 in stackframe	*/
570	std	r10,0(r1)		/* make stack chain pointer	*/
571	std	r0,GPR0(r1)		/* save r0 in stackframe	*/
572	std	r10,GPR1(r1)		/* save r1 in stackframe	*/
573
574	.if ISET_RI
575	li	r10,MSR_RI
576	mtmsrd	r10,1			/* Set MSR_RI */
577	.endif
578
579	.if ISTACK
580	.if IKUAP
581	kuap_save_amr_and_lock r9, r10, cr1, cr0
582	.endif
583	beq	101f			/* if from kernel mode		*/
584	ACCOUNT_CPU_USER_ENTRY(r13, r9, r10)
585BEGIN_FTR_SECTION
586	ld	r9,IAREA+EX_PPR(r13)	/* Read PPR from paca		*/
587	std	r9,_PPR(r1)
588END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
589101:
590	.else
591	.if IKUAP
592	kuap_save_amr_and_lock r9, r10, cr1
593	.endif
594	.endif
595
596	/* Save original regs values from save area to stack frame. */
597	ld	r9,IAREA+EX_R9(r13)	/* move r9, r10 to stackframe	*/
598	ld	r10,IAREA+EX_R10(r13)
599	std	r9,GPR9(r1)
600	std	r10,GPR10(r1)
601	ld	r9,IAREA+EX_R11(r13)	/* move r11 - r13 to stackframe	*/
602	ld	r10,IAREA+EX_R12(r13)
603	ld	r11,IAREA+EX_R13(r13)
604	std	r9,GPR11(r1)
605	std	r10,GPR12(r1)
606	std	r11,GPR13(r1)
607
608	SAVE_NVGPRS(r1)
609
610	.if IDAR
611	.if IISIDE
612	ld	r10,_NIP(r1)
613	.else
614	ld	r10,IAREA+EX_DAR(r13)
615	.endif
616	std	r10,_DAR(r1)
617	.endif
618
619	.if IDSISR
620	.if IISIDE
621	ld	r10,_MSR(r1)
622	lis	r11,DSISR_SRR1_MATCH_64S@h
623	and	r10,r10,r11
624	.else
625	lwz	r10,IAREA+EX_DSISR(r13)
626	.endif
627	std	r10,_DSISR(r1)
628	.endif
629
630BEGIN_FTR_SECTION
631	ld	r10,IAREA+EX_CFAR(r13)
632	std	r10,ORIG_GPR3(r1)
633END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
634	ld	r10,IAREA+EX_CTR(r13)
635	std	r10,_CTR(r1)
636	std	r2,GPR2(r1)		/* save r2 in stackframe	*/
637	SAVE_4GPRS(3, r1)		/* save r3 - r6 in stackframe   */
638	SAVE_2GPRS(7, r1)		/* save r7, r8 in stackframe	*/
639	mflr	r9			/* Get LR, later save to stack	*/
640	ld	r2,PACATOC(r13)		/* get kernel TOC into r2	*/
641	std	r9,_LINK(r1)
642	lbz	r10,PACAIRQSOFTMASK(r13)
643	mfspr	r11,SPRN_XER		/* save XER in stackframe	*/
644	std	r10,SOFTE(r1)
645	std	r11,_XER(r1)
646	li	r9,IVEC
647	std	r9,_TRAP(r1)		/* set trap number		*/
648	li	r10,0
649	ld	r11,exception_marker@toc(r2)
650	std	r10,RESULT(r1)		/* clear regs->result		*/
651	std	r11,STACK_FRAME_OVERHEAD-16(r1) /* mark the frame	*/
652
653	.if ISTACK
654	ACCOUNT_STOLEN_TIME
655	.endif
656
657	.if IRECONCILE
658	RECONCILE_IRQ_STATE(r10, r11)
659	.endif
660.endm
661
662/*
663 * On entry r13 points to the paca, r9-r13 are saved in the paca,
664 * r9 contains the saved CR, r11 and r12 contain the saved SRR0 and
665 * SRR1, and relocation is on.
666 *
667 * If stack=0, then the stack is already set in r1, and r1 is saved in r10.
668 * PPR save and CPU accounting is not done for the !stack case (XXX why not?)
669 */
670.macro GEN_COMMON name
671	__GEN_COMMON_ENTRY \name
672	__GEN_COMMON_BODY \name
673.endm
674
675/*
676 * Restore all registers including H/SRR0/1 saved in a stack frame of a
677 * standard exception.
678 */
679.macro EXCEPTION_RESTORE_REGS hsrr=0
680	/* Move original SRR0 and SRR1 into the respective regs */
681	ld	r9,_MSR(r1)
682	.if \hsrr
683	mtspr	SPRN_HSRR1,r9
684	.else
685	mtspr	SPRN_SRR1,r9
686	.endif
687	ld	r9,_NIP(r1)
688	.if \hsrr
689	mtspr	SPRN_HSRR0,r9
690	.else
691	mtspr	SPRN_SRR0,r9
692	.endif
693	ld	r9,_CTR(r1)
694	mtctr	r9
695	ld	r9,_XER(r1)
696	mtxer	r9
697	ld	r9,_LINK(r1)
698	mtlr	r9
699	ld	r9,_CCR(r1)
700	mtcr	r9
701	REST_8GPRS(2, r1)
702	REST_4GPRS(10, r1)
703	REST_GPR(0, r1)
704	/* restore original r1. */
705	ld	r1,GPR1(r1)
706.endm
707
708#define RUNLATCH_ON				\
709BEGIN_FTR_SECTION				\
710	ld	r3, PACA_THREAD_INFO(r13);	\
711	ld	r4,TI_LOCAL_FLAGS(r3);		\
712	andi.	r0,r4,_TLF_RUNLATCH;		\
713	beql	ppc64_runlatch_on_trampoline;	\
714END_FTR_SECTION_IFSET(CPU_FTR_CTRL)
715
716/*
717 * When the idle code in power4_idle puts the CPU into NAP mode,
718 * it has to do so in a loop, and relies on the external interrupt
719 * and decrementer interrupt entry code to get it out of the loop.
720 * It sets the _TLF_NAPPING bit in current_thread_info()->local_flags
721 * to signal that it is in the loop and needs help to get out.
722 */
723#ifdef CONFIG_PPC_970_NAP
724#define FINISH_NAP				\
725BEGIN_FTR_SECTION				\
726	ld	r11, PACA_THREAD_INFO(r13);	\
727	ld	r9,TI_LOCAL_FLAGS(r11);		\
728	andi.	r10,r9,_TLF_NAPPING;		\
729	bnel	power4_fixup_nap;		\
730END_FTR_SECTION_IFSET(CPU_FTR_CAN_NAP)
731#else
732#define FINISH_NAP
733#endif
734
735/*
736 * There are a few constraints to be concerned with.
737 * - Real mode exceptions code/data must be located at their physical location.
738 * - Virtual mode exceptions must be mapped at their 0xc000... location.
739 * - Fixed location code must not call directly beyond the __end_interrupts
740 *   area when built with CONFIG_RELOCATABLE. LOAD_HANDLER / bctr sequence
741 *   must be used.
742 * - LOAD_HANDLER targets must be within first 64K of physical 0 /
743 *   virtual 0xc00...
744 * - Conditional branch targets must be within +/-32K of caller.
745 *
746 * "Virtual exceptions" run with relocation on (MSR_IR=1, MSR_DR=1), and
747 * therefore don't have to run in physically located code or rfid to
748 * virtual mode kernel code. However on relocatable kernels they do have
749 * to branch to KERNELBASE offset because the rest of the kernel (outside
750 * the exception vectors) may be located elsewhere.
751 *
752 * Virtual exceptions correspond with physical, except their entry points
753 * are offset by 0xc000000000000000 and also tend to get an added 0x4000
754 * offset applied. Virtual exceptions are enabled with the Alternate
755 * Interrupt Location (AIL) bit set in the LPCR. However this does not
756 * guarantee they will be delivered virtually. Some conditions (see the ISA)
757 * cause exceptions to be delivered in real mode.
758 *
759 * The scv instructions are a special case. They get a 0x3000 offset applied.
760 * scv exceptions have unique reentrancy properties, see below.
761 *
762 * It's impossible to receive interrupts below 0x300 via AIL.
763 *
764 * KVM: None of the virtual exceptions are from the guest. Anything that
765 * escalated to HV=1 from HV=0 is delivered via real mode handlers.
766 *
767 *
768 * We layout physical memory as follows:
769 * 0x0000 - 0x00ff : Secondary processor spin code
770 * 0x0100 - 0x18ff : Real mode pSeries interrupt vectors
771 * 0x1900 - 0x2fff : Real mode trampolines
772 * 0x3000 - 0x58ff : Relon (IR=1,DR=1) mode pSeries interrupt vectors
773 * 0x5900 - 0x6fff : Relon mode trampolines
774 * 0x7000 - 0x7fff : FWNMI data area
775 * 0x8000 -   .... : Common interrupt handlers, remaining early
776 *                   setup code, rest of kernel.
777 *
778 * We could reclaim 0x4000-0x42ff for real mode trampolines if the space
779 * is necessary. Until then it's more consistent to explicitly put VIRT_NONE
780 * vectors there.
781 */
782OPEN_FIXED_SECTION(real_vectors,        0x0100, 0x1900)
783OPEN_FIXED_SECTION(real_trampolines,    0x1900, 0x3000)
784OPEN_FIXED_SECTION(virt_vectors,        0x3000, 0x5900)
785OPEN_FIXED_SECTION(virt_trampolines,    0x5900, 0x7000)
786
787#ifdef CONFIG_PPC_POWERNV
788	.globl start_real_trampolines
789	.globl end_real_trampolines
790	.globl start_virt_trampolines
791	.globl end_virt_trampolines
792#endif
793
794#if defined(CONFIG_PPC_PSERIES) || defined(CONFIG_PPC_POWERNV)
795/*
796 * Data area reserved for FWNMI option.
797 * This address (0x7000) is fixed by the RPA.
798 * pseries and powernv need to keep the whole page from
799 * 0x7000 to 0x8000 free for use by the firmware
800 */
801ZERO_FIXED_SECTION(fwnmi_page,          0x7000, 0x8000)
802OPEN_TEXT_SECTION(0x8000)
803#else
804OPEN_TEXT_SECTION(0x7000)
805#endif
806
807USE_FIXED_SECTION(real_vectors)
808
809/*
810 * This is the start of the interrupt handlers for pSeries
811 * This code runs with relocation off.
812 * Code from here to __end_interrupts gets copied down to real
813 * address 0x100 when we are running a relocatable kernel.
814 * Therefore any relative branches in this section must only
815 * branch to labels in this section.
816 */
817	.globl __start_interrupts
818__start_interrupts:
819
820/**
821 * Interrupt 0x3000 - System Call Vectored Interrupt (syscall).
822 * This is a synchronous interrupt invoked with the "scv" instruction. The
823 * system call does not alter the HV bit, so it is directed to the OS.
824 *
825 * Handling:
826 * scv instructions enter the kernel without changing EE, RI, ME, or HV.
827 * In particular, this means we can take a maskable interrupt at any point
828 * in the scv handler, which is unlike any other interrupt. This is solved
829 * by treating the instruction addresses below __end_interrupts as being
830 * soft-masked.
831 *
832 * AIL-0 mode scv exceptions go to 0x17000-0x17fff, but we set AIL-3 and
833 * ensure scv is never executed with relocation off, which means AIL-0
834 * should never happen.
835 *
836 * Before leaving the below __end_interrupts text, at least of the following
837 * must be true:
838 * - MSR[PR]=1 (i.e., return to userspace)
839 * - MSR_EE|MSR_RI is set (no reentrant exceptions)
840 * - Standard kernel environment is set up (stack, paca, etc)
841 *
842 * Call convention:
843 *
844 * syscall register convention is in Documentation/powerpc/syscall64-abi.rst
845 */
846EXC_VIRT_BEGIN(system_call_vectored, 0x3000, 0x1000)
847	/* SCV 0 */
848	mr	r9,r13
849	GET_PACA(r13)
850	mflr	r11
851	mfctr	r12
852	li	r10,IRQS_ALL_DISABLED
853	stb	r10,PACAIRQSOFTMASK(r13)
854#ifdef CONFIG_RELOCATABLE
855	b	system_call_vectored_tramp
856#else
857	b	system_call_vectored_common
858#endif
859	nop
860
861	/* SCV 1 - 127 */
862	.rept	127
863	mr	r9,r13
864	GET_PACA(r13)
865	mflr	r11
866	mfctr	r12
867	li	r10,IRQS_ALL_DISABLED
868	stb	r10,PACAIRQSOFTMASK(r13)
869	li	r0,-1 /* cause failure */
870#ifdef CONFIG_RELOCATABLE
871	b	system_call_vectored_sigill_tramp
872#else
873	b	system_call_vectored_sigill
874#endif
875	.endr
876EXC_VIRT_END(system_call_vectored, 0x3000, 0x1000)
877
878#ifdef CONFIG_RELOCATABLE
879TRAMP_VIRT_BEGIN(system_call_vectored_tramp)
880	__LOAD_HANDLER(r10, system_call_vectored_common)
881	mtctr	r10
882	bctr
883
884TRAMP_VIRT_BEGIN(system_call_vectored_sigill_tramp)
885	__LOAD_HANDLER(r10, system_call_vectored_sigill)
886	mtctr	r10
887	bctr
888#endif
889
890
891/* No virt vectors corresponding with 0x0..0x100 */
892EXC_VIRT_NONE(0x4000, 0x100)
893
894
895/**
896 * Interrupt 0x100 - System Reset Interrupt (SRESET aka NMI).
897 * This is a non-maskable, asynchronous interrupt always taken in real-mode.
898 * It is caused by:
899 * - Wake from power-saving state, on powernv.
900 * - An NMI from another CPU, triggered by firmware or hypercall.
901 * - As crash/debug signal injected from BMC, firmware or hypervisor.
902 *
903 * Handling:
904 * Power-save wakeup is the only performance critical path, so this is
905 * determined quickly as possible first. In this case volatile registers
906 * can be discarded and SPRs like CFAR don't need to be read.
907 *
908 * If not a powersave wakeup, then it's run as a regular interrupt, however
909 * it uses its own stack and PACA save area to preserve the regular kernel
910 * environment for debugging.
911 *
912 * This interrupt is not maskable, so triggering it when MSR[RI] is clear,
913 * or SCRATCH0 is in use, etc. may cause a crash. It's also not entirely
914 * correct to switch to virtual mode to run the regular interrupt handler
915 * because it might be interrupted when the MMU is in a bad state (e.g., SLB
916 * is clear).
917 *
918 * FWNMI:
919 * PAPR specifies a "fwnmi" facility which sends the sreset to a different
920 * entry point with a different register set up. Some hypervisors will
921 * send the sreset to 0x100 in the guest if it is not fwnmi capable.
922 *
923 * KVM:
924 * Unlike most SRR interrupts, this may be taken by the host while executing
925 * in a guest, so a KVM test is required. KVM will pull the CPU out of guest
926 * mode and then raise the sreset.
927 */
928INT_DEFINE_BEGIN(system_reset)
929	IVEC=0x100
930	IAREA=PACA_EXNMI
931	IVIRT=0 /* no virt entry point */
932	/*
933	 * MSR_RI is not enabled, because PACA_EXNMI and nmi stack is
934	 * being used, so a nested NMI exception would corrupt it.
935	 */
936	ISET_RI=0
937	ISTACK=0
938	IRECONCILE=0
939	IKVM_REAL=1
940INT_DEFINE_END(system_reset)
941
942EXC_REAL_BEGIN(system_reset, 0x100, 0x100)
943#ifdef CONFIG_PPC_P7_NAP
944	/*
945	 * If running native on arch 2.06 or later, check if we are waking up
946	 * from nap/sleep/winkle, and branch to idle handler. This tests SRR1
947	 * bits 46:47. A non-0 value indicates that we are coming from a power
948	 * saving state. The idle wakeup handler initially runs in real mode,
949	 * but we branch to the 0xc000... address so we can turn on relocation
950	 * with mtmsrd later, after SPRs are restored.
951	 *
952	 * Careful to minimise cost for the fast path (idle wakeup) while
953	 * also avoiding clobbering CFAR for the debug path (non-idle).
954	 *
955	 * For the idle wake case volatile registers can be clobbered, which
956	 * is why we use those initially. If it turns out to not be an idle
957	 * wake, carefully put everything back the way it was, so we can use
958	 * common exception macros to handle it.
959	 */
960BEGIN_FTR_SECTION
961	SET_SCRATCH0(r13)
962	GET_PACA(r13)
963	std	r3,PACA_EXNMI+0*8(r13)
964	std	r4,PACA_EXNMI+1*8(r13)
965	std	r5,PACA_EXNMI+2*8(r13)
966	mfspr	r3,SPRN_SRR1
967	mfocrf	r4,0x80
968	rlwinm.	r5,r3,47-31,30,31
969	bne+	system_reset_idle_wake
970	/* Not powersave wakeup. Restore regs for regular interrupt handler. */
971	mtocrf	0x80,r4
972	ld	r3,PACA_EXNMI+0*8(r13)
973	ld	r4,PACA_EXNMI+1*8(r13)
974	ld	r5,PACA_EXNMI+2*8(r13)
975	GET_SCRATCH0(r13)
976END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
977#endif
978
979	GEN_INT_ENTRY system_reset, virt=0
980	/*
981	 * In theory, we should not enable relocation here if it was disabled
982	 * in SRR1, because the MMU may not be configured to support it (e.g.,
983	 * SLB may have been cleared). In practice, there should only be a few
984	 * small windows where that's the case, and sreset is considered to
985	 * be dangerous anyway.
986	 */
987EXC_REAL_END(system_reset, 0x100, 0x100)
988EXC_VIRT_NONE(0x4100, 0x100)
989
990#ifdef CONFIG_PPC_P7_NAP
991TRAMP_REAL_BEGIN(system_reset_idle_wake)
992	/* We are waking up from idle, so may clobber any volatile register */
993	cmpwi	cr1,r5,2
994	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
995	BRANCH_TO_C000(r12, DOTSYM(idle_return_gpr_loss))
996#endif
997
998#ifdef CONFIG_PPC_PSERIES
999/*
1000 * Vectors for the FWNMI option.  Share common code.
1001 */
1002TRAMP_REAL_BEGIN(system_reset_fwnmi)
1003	GEN_INT_ENTRY system_reset, virt=0
1004
1005#endif /* CONFIG_PPC_PSERIES */
1006
1007EXC_COMMON_BEGIN(system_reset_common)
1008	__GEN_COMMON_ENTRY system_reset
1009	/*
1010	 * Increment paca->in_nmi then enable MSR_RI. SLB or MCE will be able
1011	 * to recover, but nested NMI will notice in_nmi and not recover
1012	 * because of the use of the NMI stack. in_nmi reentrancy is tested in
1013	 * system_reset_exception.
1014	 */
1015	lhz	r10,PACA_IN_NMI(r13)
1016	addi	r10,r10,1
1017	sth	r10,PACA_IN_NMI(r13)
1018	li	r10,MSR_RI
1019	mtmsrd 	r10,1
1020
1021	mr	r10,r1
1022	ld	r1,PACA_NMI_EMERG_SP(r13)
1023	subi	r1,r1,INT_FRAME_SIZE
1024	__GEN_COMMON_BODY system_reset
1025	/*
1026	 * Set IRQS_ALL_DISABLED unconditionally so irqs_disabled() does
1027	 * the right thing. We do not want to reconcile because that goes
1028	 * through irq tracing which we don't want in NMI.
1029	 *
1030	 * Save PACAIRQHAPPENED to RESULT (otherwise unused), and set HARD_DIS
1031	 * as we are running with MSR[EE]=0.
1032	 */
1033	li	r10,IRQS_ALL_DISABLED
1034	stb	r10,PACAIRQSOFTMASK(r13)
1035	lbz	r10,PACAIRQHAPPENED(r13)
1036	std	r10,RESULT(r1)
1037	ori	r10,r10,PACA_IRQ_HARD_DIS
1038	stb	r10,PACAIRQHAPPENED(r13)
1039
1040	addi	r3,r1,STACK_FRAME_OVERHEAD
1041	bl	system_reset_exception
1042
1043	/* Clear MSR_RI before setting SRR0 and SRR1. */
1044	li	r9,0
1045	mtmsrd	r9,1
1046
1047	/*
1048	 * MSR_RI is clear, now we can decrement paca->in_nmi.
1049	 */
1050	lhz	r10,PACA_IN_NMI(r13)
1051	subi	r10,r10,1
1052	sth	r10,PACA_IN_NMI(r13)
1053
1054	/*
1055	 * Restore soft mask settings.
1056	 */
1057	ld	r10,RESULT(r1)
1058	stb	r10,PACAIRQHAPPENED(r13)
1059	ld	r10,SOFTE(r1)
1060	stb	r10,PACAIRQSOFTMASK(r13)
1061
1062	kuap_kernel_restore r9, r10
1063	EXCEPTION_RESTORE_REGS
1064	RFI_TO_USER_OR_KERNEL
1065
1066	GEN_KVM system_reset
1067
1068
1069/**
1070 * Interrupt 0x200 - Machine Check Interrupt (MCE).
1071 * This is a non-maskable interrupt always taken in real-mode. It can be
1072 * synchronous or asynchronous, caused by hardware or software, and it may be
1073 * taken in a power-saving state.
1074 *
1075 * Handling:
1076 * Similarly to system reset, this uses its own stack and PACA save area,
1077 * the difference is re-entrancy is allowed on the machine check stack.
1078 *
1079 * machine_check_early is run in real mode, and carefully decodes the
1080 * machine check and tries to handle it (e.g., flush the SLB if there was an
1081 * error detected there), determines if it was recoverable and logs the
1082 * event.
1083 *
1084 * This early code does not "reconcile" irq soft-mask state like SRESET or
1085 * regular interrupts do, so irqs_disabled() among other things may not work
1086 * properly (irq disable/enable already doesn't work because irq tracing can
1087 * not work in real mode).
1088 *
1089 * Then, depending on the execution context when the interrupt is taken, there
1090 * are 3 main actions:
1091 * - Executing in kernel mode. The event is queued with irq_work, which means
1092 *   it is handled when it is next safe to do so (i.e., the kernel has enabled
1093 *   interrupts), which could be immediately when the interrupt returns. This
1094 *   avoids nasty issues like switching to virtual mode when the MMU is in a
1095 *   bad state, or when executing OPAL code. (SRESET is exposed to such issues,
1096 *   but it has different priorities). Check to see if the CPU was in power
1097 *   save, and return via the wake up code if it was.
1098 *
1099 * - Executing in user mode. machine_check_exception is run like a normal
1100 *   interrupt handler, which processes the data generated by the early handler.
1101 *
1102 * - Executing in guest mode. The interrupt is run with its KVM test, and
1103 *   branches to KVM to deal with. KVM may queue the event for the host
1104 *   to report later.
1105 *
1106 * This interrupt is not maskable, so if it triggers when MSR[RI] is clear,
1107 * or SCRATCH0 is in use, it may cause a crash.
1108 *
1109 * KVM:
1110 * See SRESET.
1111 */
1112INT_DEFINE_BEGIN(machine_check_early)
1113	IVEC=0x200
1114	IAREA=PACA_EXMC
1115	IVIRT=0 /* no virt entry point */
1116	IREALMODE_COMMON=1
1117	/*
1118	 * MSR_RI is not enabled, because PACA_EXMC is being used, so a
1119	 * nested machine check corrupts it. machine_check_common enables
1120	 * MSR_RI.
1121	 */
1122	ISET_RI=0
1123	ISTACK=0
1124	IDAR=1
1125	IDSISR=1
1126	IRECONCILE=0
1127	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
1128INT_DEFINE_END(machine_check_early)
1129
1130INT_DEFINE_BEGIN(machine_check)
1131	IVEC=0x200
1132	IAREA=PACA_EXMC
1133	IVIRT=0 /* no virt entry point */
1134	ISET_RI=0
1135	IDAR=1
1136	IDSISR=1
1137	IKVM_SKIP=1
1138	IKVM_REAL=1
1139INT_DEFINE_END(machine_check)
1140
1141EXC_REAL_BEGIN(machine_check, 0x200, 0x100)
1142	GEN_INT_ENTRY machine_check_early, virt=0
1143EXC_REAL_END(machine_check, 0x200, 0x100)
1144EXC_VIRT_NONE(0x4200, 0x100)
1145
1146#ifdef CONFIG_PPC_PSERIES
1147TRAMP_REAL_BEGIN(machine_check_fwnmi)
1148	/* See comment at machine_check exception, don't turn on RI */
1149	GEN_INT_ENTRY machine_check_early, virt=0
1150#endif
1151
1152#define MACHINE_CHECK_HANDLER_WINDUP			\
1153	/* Clear MSR_RI before setting SRR0 and SRR1. */\
1154	li	r9,0;					\
1155	mtmsrd	r9,1;		/* Clear MSR_RI */	\
1156	/* Decrement paca->in_mce now RI is clear. */	\
1157	lhz	r12,PACA_IN_MCE(r13);			\
1158	subi	r12,r12,1;				\
1159	sth	r12,PACA_IN_MCE(r13);			\
1160	EXCEPTION_RESTORE_REGS
1161
1162EXC_COMMON_BEGIN(machine_check_early_common)
1163	__GEN_REALMODE_COMMON_ENTRY machine_check_early
1164
1165	/*
1166	 * Switch to mc_emergency stack and handle re-entrancy (we limit
1167	 * the nested MCE upto level 4 to avoid stack overflow).
1168	 * Save MCE registers srr1, srr0, dar and dsisr and then set ME=1
1169	 *
1170	 * We use paca->in_mce to check whether this is the first entry or
1171	 * nested machine check. We increment paca->in_mce to track nested
1172	 * machine checks.
1173	 *
1174	 * If this is the first entry then set stack pointer to
1175	 * paca->mc_emergency_sp, otherwise r1 is already pointing to
1176	 * stack frame on mc_emergency stack.
1177	 *
1178	 * NOTE: We are here with MSR_ME=0 (off), which means we risk a
1179	 * checkstop if we get another machine check exception before we do
1180	 * rfid with MSR_ME=1.
1181	 *
1182	 * This interrupt can wake directly from idle. If that is the case,
1183	 * the machine check is handled then the idle wakeup code is called
1184	 * to restore state.
1185	 */
1186	lhz	r10,PACA_IN_MCE(r13)
1187	cmpwi	r10,0			/* Are we in nested machine check */
1188	cmpwi	cr1,r10,MAX_MCE_DEPTH	/* Are we at maximum nesting */
1189	addi	r10,r10,1		/* increment paca->in_mce */
1190	sth	r10,PACA_IN_MCE(r13)
1191
1192	mr	r10,r1			/* Save r1 */
1193	bne	1f
1194	/* First machine check entry */
1195	ld	r1,PACAMCEMERGSP(r13)	/* Use MC emergency stack */
11961:	/* Limit nested MCE to level 4 to avoid stack overflow */
1197	bgt	cr1,unrecoverable_mce	/* Check if we hit limit of 4 */
1198	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame */
1199
1200	__GEN_COMMON_BODY machine_check_early
1201
1202BEGIN_FTR_SECTION
1203	bl	enable_machine_check
1204END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1205	li	r10,MSR_RI
1206	mtmsrd	r10,1
1207
1208	/*
1209	 * Set IRQS_ALL_DISABLED and save PACAIRQHAPPENED (see
1210	 * system_reset_common)
1211	 */
1212	li	r10,IRQS_ALL_DISABLED
1213	stb	r10,PACAIRQSOFTMASK(r13)
1214	lbz	r10,PACAIRQHAPPENED(r13)
1215	std	r10,RESULT(r1)
1216	ori	r10,r10,PACA_IRQ_HARD_DIS
1217	stb	r10,PACAIRQHAPPENED(r13)
1218
1219	addi	r3,r1,STACK_FRAME_OVERHEAD
1220	bl	machine_check_early
1221	std	r3,RESULT(r1)	/* Save result */
1222	ld	r12,_MSR(r1)
1223
1224	/*
1225	 * Restore soft mask settings.
1226	 */
1227	ld	r10,RESULT(r1)
1228	stb	r10,PACAIRQHAPPENED(r13)
1229	ld	r10,SOFTE(r1)
1230	stb	r10,PACAIRQSOFTMASK(r13)
1231
1232#ifdef CONFIG_PPC_P7_NAP
1233	/*
1234	 * Check if thread was in power saving mode. We come here when any
1235	 * of the following is true:
1236	 * a. thread wasn't in power saving mode
1237	 * b. thread was in power saving mode with no state loss,
1238	 *    supervisor state loss or hypervisor state loss.
1239	 *
1240	 * Go back to nap/sleep/winkle mode again if (b) is true.
1241	 */
1242BEGIN_FTR_SECTION
1243	rlwinm.	r11,r12,47-31,30,31
1244	bne	machine_check_idle_common
1245END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
1246#endif
1247
1248#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1249	/*
1250	 * Check if we are coming from guest. If yes, then run the normal
1251	 * exception handler which will take the
1252	 * machine_check_kvm->kvmppc_interrupt branch to deliver the MC event
1253	 * to guest.
1254	 */
1255	lbz	r11,HSTATE_IN_GUEST(r13)
1256	cmpwi	r11,0			/* Check if coming from guest */
1257	bne	mce_deliver		/* continue if we are. */
1258#endif
1259
1260	/*
1261	 * Check if we are coming from userspace. If yes, then run the normal
1262	 * exception handler which will deliver the MC event to this kernel.
1263	 */
1264	andi.	r11,r12,MSR_PR		/* See if coming from user. */
1265	bne	mce_deliver		/* continue in V mode if we are. */
1266
1267	/*
1268	 * At this point we are coming from kernel context.
1269	 * Queue up the MCE event and return from the interrupt.
1270	 * But before that, check if this is an un-recoverable exception.
1271	 * If yes, then stay on emergency stack and panic.
1272	 */
1273	andi.	r11,r12,MSR_RI
1274	beq	unrecoverable_mce
1275
1276	/*
1277	 * Check if we have successfully handled/recovered from error, if not
1278	 * then stay on emergency stack and panic.
1279	 */
1280	ld	r3,RESULT(r1)	/* Load result */
1281	cmpdi	r3,0		/* see if we handled MCE successfully */
1282	beq	unrecoverable_mce /* if !handled then panic */
1283
1284	/*
1285	 * Return from MC interrupt.
1286	 * Queue up the MCE event so that we can log it later, while
1287	 * returning from kernel or opal call.
1288	 */
1289	bl	machine_check_queue_event
1290	MACHINE_CHECK_HANDLER_WINDUP
1291	RFI_TO_KERNEL
1292
1293mce_deliver:
1294	/*
1295	 * This is a host user or guest MCE. Restore all registers, then
1296	 * run the "late" handler. For host user, this will run the
1297	 * machine_check_exception handler in virtual mode like a normal
1298	 * interrupt handler. For guest, this will trigger the KVM test
1299	 * and branch to the KVM interrupt similarly to other interrupts.
1300	 */
1301BEGIN_FTR_SECTION
1302	ld	r10,ORIG_GPR3(r1)
1303	mtspr	SPRN_CFAR,r10
1304END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
1305	MACHINE_CHECK_HANDLER_WINDUP
1306	GEN_INT_ENTRY machine_check, virt=0
1307
1308EXC_COMMON_BEGIN(machine_check_common)
1309	/*
1310	 * Machine check is different because we use a different
1311	 * save area: PACA_EXMC instead of PACA_EXGEN.
1312	 */
1313	GEN_COMMON machine_check
1314
1315	FINISH_NAP
1316	/* Enable MSR_RI when finished with PACA_EXMC */
1317	li	r10,MSR_RI
1318	mtmsrd 	r10,1
1319	addi	r3,r1,STACK_FRAME_OVERHEAD
1320	bl	machine_check_exception
1321	b	interrupt_return
1322
1323	GEN_KVM machine_check
1324
1325
1326#ifdef CONFIG_PPC_P7_NAP
1327/*
1328 * This is an idle wakeup. Low level machine check has already been
1329 * done. Queue the event then call the idle code to do the wake up.
1330 */
1331EXC_COMMON_BEGIN(machine_check_idle_common)
1332	bl	machine_check_queue_event
1333
1334	/*
1335	 * GPR-loss wakeups are relatively straightforward, because the
1336	 * idle sleep code has saved all non-volatile registers on its
1337	 * own stack, and r1 in PACAR1.
1338	 *
1339	 * For no-loss wakeups the r1 and lr registers used by the
1340	 * early machine check handler have to be restored first. r2 is
1341	 * the kernel TOC, so no need to restore it.
1342	 *
1343	 * Then decrement MCE nesting after finishing with the stack.
1344	 */
1345	ld	r3,_MSR(r1)
1346	ld	r4,_LINK(r1)
1347	ld	r1,GPR1(r1)
1348
1349	lhz	r11,PACA_IN_MCE(r13)
1350	subi	r11,r11,1
1351	sth	r11,PACA_IN_MCE(r13)
1352
1353	mtlr	r4
1354	rlwinm	r10,r3,47-31,30,31
1355	cmpwi	cr1,r10,2
1356	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
1357	b	idle_return_gpr_loss
1358#endif
1359
1360EXC_COMMON_BEGIN(unrecoverable_mce)
1361	/*
1362	 * We are going down. But there are chances that we might get hit by
1363	 * another MCE during panic path and we may run into unstable state
1364	 * with no way out. Hence, turn ME bit off while going down, so that
1365	 * when another MCE is hit during panic path, system will checkstop
1366	 * and hypervisor will get restarted cleanly by SP.
1367	 */
1368BEGIN_FTR_SECTION
1369	li	r10,0 /* clear MSR_RI */
1370	mtmsrd	r10,1
1371	bl	disable_machine_check
1372END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1373	ld	r10,PACAKMSR(r13)
1374	li	r3,MSR_ME
1375	andc	r10,r10,r3
1376	mtmsrd	r10
1377
1378	lhz	r12,PACA_IN_MCE(r13)
1379	subi	r12,r12,1
1380	sth	r12,PACA_IN_MCE(r13)
1381
1382	/* Invoke machine_check_exception to print MCE event and panic. */
1383	addi	r3,r1,STACK_FRAME_OVERHEAD
1384	bl	machine_check_exception
1385
1386	/*
1387	 * We will not reach here. Even if we did, there is no way out.
1388	 * Call unrecoverable_exception and die.
1389	 */
1390	addi	r3,r1,STACK_FRAME_OVERHEAD
1391	bl	unrecoverable_exception
1392	b	.
1393
1394
1395/**
1396 * Interrupt 0x300 - Data Storage Interrupt (DSI).
1397 * This is a synchronous interrupt generated due to a data access exception,
1398 * e.g., a load orstore which does not have a valid page table entry with
1399 * permissions. DAWR matches also fault here, as do RC updates, and minor misc
1400 * errors e.g., copy/paste, AMO, certain invalid CI accesses, etc.
1401 *
1402 * Handling:
1403 * - Hash MMU
1404 *   Go to do_hash_page first to see if the HPT can be filled from an entry in
1405 *   the Linux page table. Hash faults can hit in kernel mode in a fairly
1406 *   arbitrary state (e.g., interrupts disabled, locks held) when accessing
1407 *   "non-bolted" regions, e.g., vmalloc space. However these should always be
1408 *   backed by Linux page tables.
1409 *
1410 *   If none is found, do a Linux page fault. Linux page faults can happen in
1411 *   kernel mode due to user copy operations of course.
1412 *
1413 *   KVM: The KVM HDSI handler may perform a load with MSR[DR]=1 in guest
1414 *   MMU context, which may cause a DSI in the host, which must go to the
1415 *   KVM handler. MSR[IR] is not enabled, so the real-mode handler will
1416 *   always be used regardless of AIL setting.
1417 *
1418 * - Radix MMU
1419 *   The hardware loads from the Linux page table directly, so a fault goes
1420 *   immediately to Linux page fault.
1421 *
1422 * Conditions like DAWR match are handled on the way in to Linux page fault.
1423 */
1424INT_DEFINE_BEGIN(data_access)
1425	IVEC=0x300
1426	IDAR=1
1427	IDSISR=1
1428	IKVM_SKIP=1
1429	IKVM_REAL=1
1430INT_DEFINE_END(data_access)
1431
1432EXC_REAL_BEGIN(data_access, 0x300, 0x80)
1433	GEN_INT_ENTRY data_access, virt=0
1434EXC_REAL_END(data_access, 0x300, 0x80)
1435EXC_VIRT_BEGIN(data_access, 0x4300, 0x80)
1436	GEN_INT_ENTRY data_access, virt=1
1437EXC_VIRT_END(data_access, 0x4300, 0x80)
1438EXC_COMMON_BEGIN(data_access_common)
1439	GEN_COMMON data_access
1440	ld	r4,_DAR(r1)
1441	ld	r5,_DSISR(r1)
1442BEGIN_MMU_FTR_SECTION
1443	ld	r6,_MSR(r1)
1444	li	r3,0x300
1445	b	do_hash_page		/* Try to handle as hpte fault */
1446MMU_FTR_SECTION_ELSE
1447	b	handle_page_fault
1448ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1449
1450	GEN_KVM data_access
1451
1452
1453/**
1454 * Interrupt 0x380 - Data Segment Interrupt (DSLB).
1455 * This is a synchronous interrupt in response to an MMU fault missing SLB
1456 * entry for HPT, or an address outside RPT translation range.
1457 *
1458 * Handling:
1459 * - HPT:
1460 *   This refills the SLB, or reports an access fault similarly to a bad page
1461 *   fault. When coming from user-mode, the SLB handler may access any kernel
1462 *   data, though it may itself take a DSLB. When coming from kernel mode,
1463 *   recursive faults must be avoided so access is restricted to the kernel
1464 *   image text/data, kernel stack, and any data allocated below
1465 *   ppc64_bolted_size (first segment). The kernel handler must avoid stomping
1466 *   on user-handler data structures.
1467 *
1468 *   KVM: Same as 0x300, DSLB must test for KVM guest.
1469 *
1470 * A dedicated save area EXSLB is used (XXX: but it actually need not be
1471 * these days, we could use EXGEN).
1472 */
1473INT_DEFINE_BEGIN(data_access_slb)
1474	IVEC=0x380
1475	IAREA=PACA_EXSLB
1476	IRECONCILE=0
1477	IDAR=1
1478	IKVM_SKIP=1
1479	IKVM_REAL=1
1480INT_DEFINE_END(data_access_slb)
1481
1482EXC_REAL_BEGIN(data_access_slb, 0x380, 0x80)
1483	GEN_INT_ENTRY data_access_slb, virt=0
1484EXC_REAL_END(data_access_slb, 0x380, 0x80)
1485EXC_VIRT_BEGIN(data_access_slb, 0x4380, 0x80)
1486	GEN_INT_ENTRY data_access_slb, virt=1
1487EXC_VIRT_END(data_access_slb, 0x4380, 0x80)
1488EXC_COMMON_BEGIN(data_access_slb_common)
1489	GEN_COMMON data_access_slb
1490	ld	r4,_DAR(r1)
1491	addi	r3,r1,STACK_FRAME_OVERHEAD
1492BEGIN_MMU_FTR_SECTION
1493	/* HPT case, do SLB fault */
1494	bl	do_slb_fault
1495	cmpdi	r3,0
1496	bne-	1f
1497	b	fast_interrupt_return
14981:	/* Error case */
1499MMU_FTR_SECTION_ELSE
1500	/* Radix case, access is outside page table range */
1501	li	r3,-EFAULT
1502ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1503	std	r3,RESULT(r1)
1504	RECONCILE_IRQ_STATE(r10, r11)
1505	ld	r4,_DAR(r1)
1506	ld	r5,RESULT(r1)
1507	addi	r3,r1,STACK_FRAME_OVERHEAD
1508	bl	do_bad_slb_fault
1509	b	interrupt_return
1510
1511	GEN_KVM data_access_slb
1512
1513
1514/**
1515 * Interrupt 0x400 - Instruction Storage Interrupt (ISI).
1516 * This is a synchronous interrupt in response to an MMU fault due to an
1517 * instruction fetch.
1518 *
1519 * Handling:
1520 * Similar to DSI, though in response to fetch. The faulting address is found
1521 * in SRR0 (rather than DAR), and status in SRR1 (rather than DSISR).
1522 */
1523INT_DEFINE_BEGIN(instruction_access)
1524	IVEC=0x400
1525	IISIDE=1
1526	IDAR=1
1527	IDSISR=1
1528#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1529	IKVM_REAL=1
1530#endif
1531INT_DEFINE_END(instruction_access)
1532
1533EXC_REAL_BEGIN(instruction_access, 0x400, 0x80)
1534	GEN_INT_ENTRY instruction_access, virt=0
1535EXC_REAL_END(instruction_access, 0x400, 0x80)
1536EXC_VIRT_BEGIN(instruction_access, 0x4400, 0x80)
1537	GEN_INT_ENTRY instruction_access, virt=1
1538EXC_VIRT_END(instruction_access, 0x4400, 0x80)
1539EXC_COMMON_BEGIN(instruction_access_common)
1540	GEN_COMMON instruction_access
1541	ld	r4,_DAR(r1)
1542	ld	r5,_DSISR(r1)
1543BEGIN_MMU_FTR_SECTION
1544	ld      r6,_MSR(r1)
1545	li	r3,0x400
1546	b	do_hash_page		/* Try to handle as hpte fault */
1547MMU_FTR_SECTION_ELSE
1548	b	handle_page_fault
1549ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1550
1551	GEN_KVM instruction_access
1552
1553
1554/**
1555 * Interrupt 0x480 - Instruction Segment Interrupt (ISLB).
1556 * This is a synchronous interrupt in response to an MMU fault due to an
1557 * instruction fetch.
1558 *
1559 * Handling:
1560 * Similar to DSLB, though in response to fetch. The faulting address is found
1561 * in SRR0 (rather than DAR).
1562 */
1563INT_DEFINE_BEGIN(instruction_access_slb)
1564	IVEC=0x480
1565	IAREA=PACA_EXSLB
1566	IRECONCILE=0
1567	IISIDE=1
1568	IDAR=1
1569#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1570	IKVM_REAL=1
1571#endif
1572INT_DEFINE_END(instruction_access_slb)
1573
1574EXC_REAL_BEGIN(instruction_access_slb, 0x480, 0x80)
1575	GEN_INT_ENTRY instruction_access_slb, virt=0
1576EXC_REAL_END(instruction_access_slb, 0x480, 0x80)
1577EXC_VIRT_BEGIN(instruction_access_slb, 0x4480, 0x80)
1578	GEN_INT_ENTRY instruction_access_slb, virt=1
1579EXC_VIRT_END(instruction_access_slb, 0x4480, 0x80)
1580EXC_COMMON_BEGIN(instruction_access_slb_common)
1581	GEN_COMMON instruction_access_slb
1582	ld	r4,_DAR(r1)
1583	addi	r3,r1,STACK_FRAME_OVERHEAD
1584BEGIN_MMU_FTR_SECTION
1585	/* HPT case, do SLB fault */
1586	bl	do_slb_fault
1587	cmpdi	r3,0
1588	bne-	1f
1589	b	fast_interrupt_return
15901:	/* Error case */
1591MMU_FTR_SECTION_ELSE
1592	/* Radix case, access is outside page table range */
1593	li	r3,-EFAULT
1594ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1595	std	r3,RESULT(r1)
1596	RECONCILE_IRQ_STATE(r10, r11)
1597	ld	r4,_DAR(r1)
1598	ld	r5,RESULT(r1)
1599	addi	r3,r1,STACK_FRAME_OVERHEAD
1600	bl	do_bad_slb_fault
1601	b	interrupt_return
1602
1603	GEN_KVM instruction_access_slb
1604
1605
1606/**
1607 * Interrupt 0x500 - External Interrupt.
1608 * This is an asynchronous maskable interrupt in response to an "external
1609 * exception" from the interrupt controller or hypervisor (e.g., device
1610 * interrupt). It is maskable in hardware by clearing MSR[EE], and
1611 * soft-maskable with IRQS_DISABLED mask (i.e., local_irq_disable()).
1612 *
1613 * When running in HV mode, Linux sets up the LPCR[LPES] bit such that
1614 * interrupts are delivered with HSRR registers, guests use SRRs, which
1615 * reqiures IHSRR_IF_HVMODE.
1616 *
1617 * On bare metal POWER9 and later, Linux sets the LPCR[HVICE] bit such that
1618 * external interrupts are delivered as Hypervisor Virtualization Interrupts
1619 * rather than External Interrupts.
1620 *
1621 * Handling:
1622 * This calls into Linux IRQ handler. NVGPRs are not saved to reduce overhead,
1623 * because registers at the time of the interrupt are not so important as it is
1624 * asynchronous.
1625 *
1626 * If soft masked, the masked handler will note the pending interrupt for
1627 * replay, and clear MSR[EE] in the interrupted context.
1628 */
1629INT_DEFINE_BEGIN(hardware_interrupt)
1630	IVEC=0x500
1631	IHSRR_IF_HVMODE=1
1632	IMASK=IRQS_DISABLED
1633	IKVM_REAL=1
1634	IKVM_VIRT=1
1635INT_DEFINE_END(hardware_interrupt)
1636
1637EXC_REAL_BEGIN(hardware_interrupt, 0x500, 0x100)
1638	GEN_INT_ENTRY hardware_interrupt, virt=0
1639EXC_REAL_END(hardware_interrupt, 0x500, 0x100)
1640EXC_VIRT_BEGIN(hardware_interrupt, 0x4500, 0x100)
1641	GEN_INT_ENTRY hardware_interrupt, virt=1
1642EXC_VIRT_END(hardware_interrupt, 0x4500, 0x100)
1643EXC_COMMON_BEGIN(hardware_interrupt_common)
1644	GEN_COMMON hardware_interrupt
1645	FINISH_NAP
1646	RUNLATCH_ON
1647	addi	r3,r1,STACK_FRAME_OVERHEAD
1648	bl	do_IRQ
1649	b	interrupt_return
1650
1651	GEN_KVM hardware_interrupt
1652
1653
1654/**
1655 * Interrupt 0x600 - Alignment Interrupt
1656 * This is a synchronous interrupt in response to data alignment fault.
1657 */
1658INT_DEFINE_BEGIN(alignment)
1659	IVEC=0x600
1660	IDAR=1
1661	IDSISR=1
1662#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1663	IKVM_REAL=1
1664#endif
1665INT_DEFINE_END(alignment)
1666
1667EXC_REAL_BEGIN(alignment, 0x600, 0x100)
1668	GEN_INT_ENTRY alignment, virt=0
1669EXC_REAL_END(alignment, 0x600, 0x100)
1670EXC_VIRT_BEGIN(alignment, 0x4600, 0x100)
1671	GEN_INT_ENTRY alignment, virt=1
1672EXC_VIRT_END(alignment, 0x4600, 0x100)
1673EXC_COMMON_BEGIN(alignment_common)
1674	GEN_COMMON alignment
1675	addi	r3,r1,STACK_FRAME_OVERHEAD
1676	bl	alignment_exception
1677	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
1678	b	interrupt_return
1679
1680	GEN_KVM alignment
1681
1682
1683/**
1684 * Interrupt 0x700 - Program Interrupt (program check).
1685 * This is a synchronous interrupt in response to various instruction faults:
1686 * traps, privilege errors, TM errors, floating point exceptions.
1687 *
1688 * Handling:
1689 * This interrupt may use the "emergency stack" in some cases when being taken
1690 * from kernel context, which complicates handling.
1691 */
1692INT_DEFINE_BEGIN(program_check)
1693	IVEC=0x700
1694#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1695	IKVM_REAL=1
1696#endif
1697INT_DEFINE_END(program_check)
1698
1699EXC_REAL_BEGIN(program_check, 0x700, 0x100)
1700	GEN_INT_ENTRY program_check, virt=0
1701EXC_REAL_END(program_check, 0x700, 0x100)
1702EXC_VIRT_BEGIN(program_check, 0x4700, 0x100)
1703	GEN_INT_ENTRY program_check, virt=1
1704EXC_VIRT_END(program_check, 0x4700, 0x100)
1705EXC_COMMON_BEGIN(program_check_common)
1706	__GEN_COMMON_ENTRY program_check
1707
1708	/*
1709	 * It's possible to receive a TM Bad Thing type program check with
1710	 * userspace register values (in particular r1), but with SRR1 reporting
1711	 * that we came from the kernel. Normally that would confuse the bad
1712	 * stack logic, and we would report a bad kernel stack pointer. Instead
1713	 * we switch to the emergency stack if we're taking a TM Bad Thing from
1714	 * the kernel.
1715	 */
1716
1717	andi.	r10,r12,MSR_PR
1718	bne	2f			/* If userspace, go normal path */
1719
1720	andis.	r10,r12,(SRR1_PROGTM)@h
1721	bne	1f			/* If TM, emergency		*/
1722
1723	cmpdi	r1,-INT_FRAME_SIZE	/* check if r1 is in userspace	*/
1724	blt	2f			/* normal path if not		*/
1725
1726	/* Use the emergency stack					*/
17271:	andi.	r10,r12,MSR_PR		/* Set CR0 correctly for label	*/
1728					/* 3 in EXCEPTION_PROLOG_COMMON	*/
1729	mr	r10,r1			/* Save r1			*/
1730	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack		*/
1731	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
1732	__ISTACK(program_check)=0
1733	__GEN_COMMON_BODY program_check
1734	b 3f
17352:
1736	__ISTACK(program_check)=1
1737	__GEN_COMMON_BODY program_check
17383:
1739	addi	r3,r1,STACK_FRAME_OVERHEAD
1740	bl	program_check_exception
1741	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
1742	b	interrupt_return
1743
1744	GEN_KVM program_check
1745
1746
1747/*
1748 * Interrupt 0x800 - Floating-Point Unavailable Interrupt.
1749 * This is a synchronous interrupt in response to executing an fp instruction
1750 * with MSR[FP]=0.
1751 *
1752 * Handling:
1753 * This will load FP registers and enable the FP bit if coming from userspace,
1754 * otherwise report a bad kernel use of FP.
1755 */
1756INT_DEFINE_BEGIN(fp_unavailable)
1757	IVEC=0x800
1758	IRECONCILE=0
1759#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1760	IKVM_REAL=1
1761#endif
1762INT_DEFINE_END(fp_unavailable)
1763
1764EXC_REAL_BEGIN(fp_unavailable, 0x800, 0x100)
1765	GEN_INT_ENTRY fp_unavailable, virt=0
1766EXC_REAL_END(fp_unavailable, 0x800, 0x100)
1767EXC_VIRT_BEGIN(fp_unavailable, 0x4800, 0x100)
1768	GEN_INT_ENTRY fp_unavailable, virt=1
1769EXC_VIRT_END(fp_unavailable, 0x4800, 0x100)
1770EXC_COMMON_BEGIN(fp_unavailable_common)
1771	GEN_COMMON fp_unavailable
1772	bne	1f			/* if from user, just load it up */
1773	RECONCILE_IRQ_STATE(r10, r11)
1774	addi	r3,r1,STACK_FRAME_OVERHEAD
1775	bl	kernel_fp_unavailable_exception
17760:	trap
1777	EMIT_BUG_ENTRY 0b, __FILE__, __LINE__, 0
17781:
1779#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1780BEGIN_FTR_SECTION
1781	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
1782	 * transaction), go do TM stuff
1783	 */
1784	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
1785	bne-	2f
1786END_FTR_SECTION_IFSET(CPU_FTR_TM)
1787#endif
1788	bl	load_up_fpu
1789	b	fast_interrupt_return
1790#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
17912:	/* User process was in a transaction */
1792	RECONCILE_IRQ_STATE(r10, r11)
1793	addi	r3,r1,STACK_FRAME_OVERHEAD
1794	bl	fp_unavailable_tm
1795	b	interrupt_return
1796#endif
1797
1798	GEN_KVM fp_unavailable
1799
1800
1801/**
1802 * Interrupt 0x900 - Decrementer Interrupt.
1803 * This is an asynchronous interrupt in response to a decrementer exception
1804 * (e.g., DEC has wrapped below zero). It is maskable in hardware by clearing
1805 * MSR[EE], and soft-maskable with IRQS_DISABLED mask (i.e.,
1806 * local_irq_disable()).
1807 *
1808 * Handling:
1809 * This calls into Linux timer handler. NVGPRs are not saved (see 0x500).
1810 *
1811 * If soft masked, the masked handler will note the pending interrupt for
1812 * replay, and bump the decrementer to a high value, leaving MSR[EE] enabled
1813 * in the interrupted context.
1814 * If PPC_WATCHDOG is configured, the soft masked handler will actually set
1815 * things back up to run soft_nmi_interrupt as a regular interrupt handler
1816 * on the emergency stack.
1817 */
1818INT_DEFINE_BEGIN(decrementer)
1819	IVEC=0x900
1820	IMASK=IRQS_DISABLED
1821#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1822	IKVM_REAL=1
1823#endif
1824INT_DEFINE_END(decrementer)
1825
1826EXC_REAL_BEGIN(decrementer, 0x900, 0x80)
1827	GEN_INT_ENTRY decrementer, virt=0
1828EXC_REAL_END(decrementer, 0x900, 0x80)
1829EXC_VIRT_BEGIN(decrementer, 0x4900, 0x80)
1830	GEN_INT_ENTRY decrementer, virt=1
1831EXC_VIRT_END(decrementer, 0x4900, 0x80)
1832EXC_COMMON_BEGIN(decrementer_common)
1833	GEN_COMMON decrementer
1834	FINISH_NAP
1835	RUNLATCH_ON
1836	addi	r3,r1,STACK_FRAME_OVERHEAD
1837	bl	timer_interrupt
1838	b	interrupt_return
1839
1840	GEN_KVM decrementer
1841
1842
1843/**
1844 * Interrupt 0x980 - Hypervisor Decrementer Interrupt.
1845 * This is an asynchronous interrupt, similar to 0x900 but for the HDEC
1846 * register.
1847 *
1848 * Handling:
1849 * Linux does not use this outside KVM where it's used to keep a host timer
1850 * while the guest is given control of DEC. It should normally be caught by
1851 * the KVM test and routed there.
1852 */
1853INT_DEFINE_BEGIN(hdecrementer)
1854	IVEC=0x980
1855	IHSRR=1
1856	ISTACK=0
1857	IRECONCILE=0
1858	IKVM_REAL=1
1859	IKVM_VIRT=1
1860INT_DEFINE_END(hdecrementer)
1861
1862EXC_REAL_BEGIN(hdecrementer, 0x980, 0x80)
1863	GEN_INT_ENTRY hdecrementer, virt=0
1864EXC_REAL_END(hdecrementer, 0x980, 0x80)
1865EXC_VIRT_BEGIN(hdecrementer, 0x4980, 0x80)
1866	GEN_INT_ENTRY hdecrementer, virt=1
1867EXC_VIRT_END(hdecrementer, 0x4980, 0x80)
1868EXC_COMMON_BEGIN(hdecrementer_common)
1869	__GEN_COMMON_ENTRY hdecrementer
1870	/*
1871	 * Hypervisor decrementer interrupts not caught by the KVM test
1872	 * shouldn't occur but are sometimes left pending on exit from a KVM
1873	 * guest.  We don't need to do anything to clear them, as they are
1874	 * edge-triggered.
1875	 *
1876	 * Be careful to avoid touching the kernel stack.
1877	 */
1878	ld	r10,PACA_EXGEN+EX_CTR(r13)
1879	mtctr	r10
1880	mtcrf	0x80,r9
1881	ld	r9,PACA_EXGEN+EX_R9(r13)
1882	ld	r10,PACA_EXGEN+EX_R10(r13)
1883	ld	r11,PACA_EXGEN+EX_R11(r13)
1884	ld	r12,PACA_EXGEN+EX_R12(r13)
1885	ld	r13,PACA_EXGEN+EX_R13(r13)
1886	HRFI_TO_KERNEL
1887
1888	GEN_KVM hdecrementer
1889
1890
1891/**
1892 * Interrupt 0xa00 - Directed Privileged Doorbell Interrupt.
1893 * This is an asynchronous interrupt in response to a msgsndp doorbell.
1894 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
1895 * IRQS_DISABLED mask (i.e., local_irq_disable()).
1896 *
1897 * Handling:
1898 * Guests may use this for IPIs between threads in a core if the
1899 * hypervisor supports it. NVGPRS are not saved (see 0x500).
1900 *
1901 * If soft masked, the masked handler will note the pending interrupt for
1902 * replay, leaving MSR[EE] enabled in the interrupted context because the
1903 * doorbells are edge triggered.
1904 */
1905INT_DEFINE_BEGIN(doorbell_super)
1906	IVEC=0xa00
1907	IMASK=IRQS_DISABLED
1908#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1909	IKVM_REAL=1
1910#endif
1911INT_DEFINE_END(doorbell_super)
1912
1913EXC_REAL_BEGIN(doorbell_super, 0xa00, 0x100)
1914	GEN_INT_ENTRY doorbell_super, virt=0
1915EXC_REAL_END(doorbell_super, 0xa00, 0x100)
1916EXC_VIRT_BEGIN(doorbell_super, 0x4a00, 0x100)
1917	GEN_INT_ENTRY doorbell_super, virt=1
1918EXC_VIRT_END(doorbell_super, 0x4a00, 0x100)
1919EXC_COMMON_BEGIN(doorbell_super_common)
1920	GEN_COMMON doorbell_super
1921	FINISH_NAP
1922	RUNLATCH_ON
1923	addi	r3,r1,STACK_FRAME_OVERHEAD
1924#ifdef CONFIG_PPC_DOORBELL
1925	bl	doorbell_exception
1926#else
1927	bl	unknown_exception
1928#endif
1929	b	interrupt_return
1930
1931	GEN_KVM doorbell_super
1932
1933
1934EXC_REAL_NONE(0xb00, 0x100)
1935EXC_VIRT_NONE(0x4b00, 0x100)
1936
1937/**
1938 * Interrupt 0xc00 - System Call Interrupt (syscall, hcall).
1939 * This is a synchronous interrupt invoked with the "sc" instruction. The
1940 * system call is invoked with "sc 0" and does not alter the HV bit, so it
1941 * is directed to the currently running OS. The hypercall is invoked with
1942 * "sc 1" and it sets HV=1, so it elevates to hypervisor.
1943 *
1944 * In HPT, sc 1 always goes to 0xc00 real mode. In RADIX, sc 1 can go to
1945 * 0x4c00 virtual mode.
1946 *
1947 * Handling:
1948 * If the KVM test fires then it was due to a hypercall and is accordingly
1949 * routed to KVM. Otherwise this executes a normal Linux system call.
1950 *
1951 * Call convention:
1952 *
1953 * syscall and hypercalls register conventions are documented in
1954 * Documentation/powerpc/syscall64-abi.rst and
1955 * Documentation/powerpc/papr_hcalls.rst respectively.
1956 *
1957 * The intersection of volatile registers that don't contain possible
1958 * inputs is: cr0, xer, ctr. We may use these as scratch regs upon entry
1959 * without saving, though xer is not a good idea to use, as hardware may
1960 * interpret some bits so it may be costly to change them.
1961 */
1962INT_DEFINE_BEGIN(system_call)
1963	IVEC=0xc00
1964	IKVM_REAL=1
1965	IKVM_VIRT=1
1966INT_DEFINE_END(system_call)
1967
1968.macro SYSTEM_CALL virt
1969#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1970	/*
1971	 * There is a little bit of juggling to get syscall and hcall
1972	 * working well. Save r13 in ctr to avoid using SPRG scratch
1973	 * register.
1974	 *
1975	 * Userspace syscalls have already saved the PPR, hcalls must save
1976	 * it before setting HMT_MEDIUM.
1977	 */
1978	mtctr	r13
1979	GET_PACA(r13)
1980	std	r10,PACA_EXGEN+EX_R10(r13)
1981	INTERRUPT_TO_KERNEL
1982	KVMTEST system_call /* uses r10, branch to system_call_kvm */
1983	mfctr	r9
1984#else
1985	mr	r9,r13
1986	GET_PACA(r13)
1987	INTERRUPT_TO_KERNEL
1988#endif
1989
1990#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
1991BEGIN_FTR_SECTION
1992	cmpdi	r0,0x1ebe
1993	beq-	1f
1994END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE)
1995#endif
1996
1997	/* We reach here with PACA in r13, r13 in r9. */
1998	mfspr	r11,SPRN_SRR0
1999	mfspr	r12,SPRN_SRR1
2000
2001	HMT_MEDIUM
2002
2003	.if ! \virt
2004	__LOAD_HANDLER(r10, system_call_common)
2005	mtspr	SPRN_SRR0,r10
2006	ld	r10,PACAKMSR(r13)
2007	mtspr	SPRN_SRR1,r10
2008	RFI_TO_KERNEL
2009	b	.	/* prevent speculative execution */
2010	.else
2011	li	r10,MSR_RI
2012	mtmsrd 	r10,1			/* Set RI (EE=0) */
2013#ifdef CONFIG_RELOCATABLE
2014	__LOAD_HANDLER(r10, system_call_common)
2015	mtctr	r10
2016	bctr
2017#else
2018	b	system_call_common
2019#endif
2020	.endif
2021
2022#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
2023	/* Fast LE/BE switch system call */
20241:	mfspr	r12,SPRN_SRR1
2025	xori	r12,r12,MSR_LE
2026	mtspr	SPRN_SRR1,r12
2027	mr	r13,r9
2028	RFI_TO_USER	/* return to userspace */
2029	b	.	/* prevent speculative execution */
2030#endif
2031.endm
2032
2033EXC_REAL_BEGIN(system_call, 0xc00, 0x100)
2034	SYSTEM_CALL 0
2035EXC_REAL_END(system_call, 0xc00, 0x100)
2036EXC_VIRT_BEGIN(system_call, 0x4c00, 0x100)
2037	SYSTEM_CALL 1
2038EXC_VIRT_END(system_call, 0x4c00, 0x100)
2039
2040#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
2041TRAMP_REAL_BEGIN(system_call_kvm)
2042	/*
2043	 * This is a hcall, so register convention is as above, with these
2044	 * differences:
2045	 * r13 = PACA
2046	 * ctr = orig r13
2047	 * orig r10 saved in PACA
2048	 */
2049	 /*
2050	  * Save the PPR (on systems that support it) before changing to
2051	  * HMT_MEDIUM. That allows the KVM code to save that value into the
2052	  * guest state (it is the guest's PPR value).
2053	  */
2054BEGIN_FTR_SECTION
2055	mfspr	r10,SPRN_PPR
2056	std	r10,HSTATE_PPR(r13)
2057END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
2058	HMT_MEDIUM
2059	mfctr	r10
2060	SET_SCRATCH0(r10)
2061	mfcr	r10
2062	std	r12,HSTATE_SCRATCH0(r13)
2063	sldi	r12,r10,32
2064	ori	r12,r12,0xc00
2065#ifdef CONFIG_RELOCATABLE
2066	/*
2067	 * Requires __LOAD_FAR_HANDLER beause kvmppc_interrupt lives
2068	 * outside the head section.
2069	 */
2070	__LOAD_FAR_HANDLER(r10, kvmppc_interrupt)
2071	mtctr   r10
2072	ld	r10,PACA_EXGEN+EX_R10(r13)
2073	bctr
2074#else
2075	ld	r10,PACA_EXGEN+EX_R10(r13)
2076	b       kvmppc_interrupt
2077#endif
2078#endif
2079
2080
2081/**
2082 * Interrupt 0xd00 - Trace Interrupt.
2083 * This is a synchronous interrupt in response to instruction step or
2084 * breakpoint faults.
2085 */
2086INT_DEFINE_BEGIN(single_step)
2087	IVEC=0xd00
2088#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2089	IKVM_REAL=1
2090#endif
2091INT_DEFINE_END(single_step)
2092
2093EXC_REAL_BEGIN(single_step, 0xd00, 0x100)
2094	GEN_INT_ENTRY single_step, virt=0
2095EXC_REAL_END(single_step, 0xd00, 0x100)
2096EXC_VIRT_BEGIN(single_step, 0x4d00, 0x100)
2097	GEN_INT_ENTRY single_step, virt=1
2098EXC_VIRT_END(single_step, 0x4d00, 0x100)
2099EXC_COMMON_BEGIN(single_step_common)
2100	GEN_COMMON single_step
2101	addi	r3,r1,STACK_FRAME_OVERHEAD
2102	bl	single_step_exception
2103	b	interrupt_return
2104
2105	GEN_KVM single_step
2106
2107
2108/**
2109 * Interrupt 0xe00 - Hypervisor Data Storage Interrupt (HDSI).
2110 * This is a synchronous interrupt in response to an MMU fault caused by a
2111 * guest data access.
2112 *
2113 * Handling:
2114 * This should always get routed to KVM. In radix MMU mode, this is caused
2115 * by a guest nested radix access that can't be performed due to the
2116 * partition scope page table. In hash mode, this can be caused by guests
2117 * running with translation disabled (virtual real mode) or with VPM enabled.
2118 * KVM will update the page table structures or disallow the access.
2119 */
2120INT_DEFINE_BEGIN(h_data_storage)
2121	IVEC=0xe00
2122	IHSRR=1
2123	IDAR=1
2124	IDSISR=1
2125	IKVM_SKIP=1
2126	IKVM_REAL=1
2127	IKVM_VIRT=1
2128INT_DEFINE_END(h_data_storage)
2129
2130EXC_REAL_BEGIN(h_data_storage, 0xe00, 0x20)
2131	GEN_INT_ENTRY h_data_storage, virt=0, ool=1
2132EXC_REAL_END(h_data_storage, 0xe00, 0x20)
2133EXC_VIRT_BEGIN(h_data_storage, 0x4e00, 0x20)
2134	GEN_INT_ENTRY h_data_storage, virt=1, ool=1
2135EXC_VIRT_END(h_data_storage, 0x4e00, 0x20)
2136EXC_COMMON_BEGIN(h_data_storage_common)
2137	GEN_COMMON h_data_storage
2138	addi    r3,r1,STACK_FRAME_OVERHEAD
2139BEGIN_MMU_FTR_SECTION
2140	ld	r4,_DAR(r1)
2141	li	r5,SIGSEGV
2142	bl      bad_page_fault
2143MMU_FTR_SECTION_ELSE
2144	bl      unknown_exception
2145ALT_MMU_FTR_SECTION_END_IFSET(MMU_FTR_TYPE_RADIX)
2146	b       interrupt_return
2147
2148	GEN_KVM h_data_storage
2149
2150
2151/**
2152 * Interrupt 0xe20 - Hypervisor Instruction Storage Interrupt (HISI).
2153 * This is a synchronous interrupt in response to an MMU fault caused by a
2154 * guest instruction fetch, similar to HDSI.
2155 */
2156INT_DEFINE_BEGIN(h_instr_storage)
2157	IVEC=0xe20
2158	IHSRR=1
2159	IKVM_REAL=1
2160	IKVM_VIRT=1
2161INT_DEFINE_END(h_instr_storage)
2162
2163EXC_REAL_BEGIN(h_instr_storage, 0xe20, 0x20)
2164	GEN_INT_ENTRY h_instr_storage, virt=0, ool=1
2165EXC_REAL_END(h_instr_storage, 0xe20, 0x20)
2166EXC_VIRT_BEGIN(h_instr_storage, 0x4e20, 0x20)
2167	GEN_INT_ENTRY h_instr_storage, virt=1, ool=1
2168EXC_VIRT_END(h_instr_storage, 0x4e20, 0x20)
2169EXC_COMMON_BEGIN(h_instr_storage_common)
2170	GEN_COMMON h_instr_storage
2171	addi	r3,r1,STACK_FRAME_OVERHEAD
2172	bl	unknown_exception
2173	b	interrupt_return
2174
2175	GEN_KVM h_instr_storage
2176
2177
2178/**
2179 * Interrupt 0xe40 - Hypervisor Emulation Assistance Interrupt.
2180 */
2181INT_DEFINE_BEGIN(emulation_assist)
2182	IVEC=0xe40
2183	IHSRR=1
2184	IKVM_REAL=1
2185	IKVM_VIRT=1
2186INT_DEFINE_END(emulation_assist)
2187
2188EXC_REAL_BEGIN(emulation_assist, 0xe40, 0x20)
2189	GEN_INT_ENTRY emulation_assist, virt=0, ool=1
2190EXC_REAL_END(emulation_assist, 0xe40, 0x20)
2191EXC_VIRT_BEGIN(emulation_assist, 0x4e40, 0x20)
2192	GEN_INT_ENTRY emulation_assist, virt=1, ool=1
2193EXC_VIRT_END(emulation_assist, 0x4e40, 0x20)
2194EXC_COMMON_BEGIN(emulation_assist_common)
2195	GEN_COMMON emulation_assist
2196	addi	r3,r1,STACK_FRAME_OVERHEAD
2197	bl	emulation_assist_interrupt
2198	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2199	b	interrupt_return
2200
2201	GEN_KVM emulation_assist
2202
2203
2204/**
2205 * Interrupt 0xe60 - Hypervisor Maintenance Interrupt (HMI).
2206 * This is an asynchronous interrupt caused by a Hypervisor Maintenance
2207 * Exception. It is always taken in real mode but uses HSRR registers
2208 * unlike SRESET and MCE.
2209 *
2210 * It is maskable in hardware by clearing MSR[EE], and partially soft-maskable
2211 * with IRQS_DISABLED mask (i.e., local_irq_disable()).
2212 *
2213 * Handling:
2214 * This is a special case, this is handled similarly to machine checks, with an
2215 * initial real mode handler that is not soft-masked, which attempts to fix the
2216 * problem. Then a regular handler which is soft-maskable and reports the
2217 * problem.
2218 *
2219 * The emergency stack is used for the early real mode handler.
2220 *
2221 * XXX: unclear why MCE and HMI schemes could not be made common, e.g.,
2222 * either use soft-masking for the MCE, or use irq_work for the HMI.
2223 *
2224 * KVM:
2225 * Unlike MCE, this calls into KVM without calling the real mode handler
2226 * first.
2227 */
2228INT_DEFINE_BEGIN(hmi_exception_early)
2229	IVEC=0xe60
2230	IHSRR=1
2231	IREALMODE_COMMON=1
2232	ISTACK=0
2233	IRECONCILE=0
2234	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
2235	IKVM_REAL=1
2236INT_DEFINE_END(hmi_exception_early)
2237
2238INT_DEFINE_BEGIN(hmi_exception)
2239	IVEC=0xe60
2240	IHSRR=1
2241	IMASK=IRQS_DISABLED
2242	IKVM_REAL=1
2243INT_DEFINE_END(hmi_exception)
2244
2245EXC_REAL_BEGIN(hmi_exception, 0xe60, 0x20)
2246	GEN_INT_ENTRY hmi_exception_early, virt=0, ool=1
2247EXC_REAL_END(hmi_exception, 0xe60, 0x20)
2248EXC_VIRT_NONE(0x4e60, 0x20)
2249
2250EXC_COMMON_BEGIN(hmi_exception_early_common)
2251	__GEN_REALMODE_COMMON_ENTRY hmi_exception_early
2252
2253	mr	r10,r1			/* Save r1 */
2254	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack for realmode */
2255	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
2256
2257	__GEN_COMMON_BODY hmi_exception_early
2258
2259	addi	r3,r1,STACK_FRAME_OVERHEAD
2260	bl	hmi_exception_realmode
2261	cmpdi	cr0,r3,0
2262	bne	1f
2263
2264	EXCEPTION_RESTORE_REGS hsrr=1
2265	HRFI_TO_USER_OR_KERNEL
2266
22671:
2268	/*
2269	 * Go to virtual mode and pull the HMI event information from
2270	 * firmware.
2271	 */
2272	EXCEPTION_RESTORE_REGS hsrr=1
2273	GEN_INT_ENTRY hmi_exception, virt=0
2274
2275	GEN_KVM hmi_exception_early
2276
2277EXC_COMMON_BEGIN(hmi_exception_common)
2278	GEN_COMMON hmi_exception
2279	FINISH_NAP
2280	RUNLATCH_ON
2281	addi	r3,r1,STACK_FRAME_OVERHEAD
2282	bl	handle_hmi_exception
2283	b	interrupt_return
2284
2285	GEN_KVM hmi_exception
2286
2287
2288/**
2289 * Interrupt 0xe80 - Directed Hypervisor Doorbell Interrupt.
2290 * This is an asynchronous interrupt in response to a msgsnd doorbell.
2291 * Similar to the 0xa00 doorbell but for host rather than guest.
2292 */
2293INT_DEFINE_BEGIN(h_doorbell)
2294	IVEC=0xe80
2295	IHSRR=1
2296	IMASK=IRQS_DISABLED
2297	IKVM_REAL=1
2298	IKVM_VIRT=1
2299INT_DEFINE_END(h_doorbell)
2300
2301EXC_REAL_BEGIN(h_doorbell, 0xe80, 0x20)
2302	GEN_INT_ENTRY h_doorbell, virt=0, ool=1
2303EXC_REAL_END(h_doorbell, 0xe80, 0x20)
2304EXC_VIRT_BEGIN(h_doorbell, 0x4e80, 0x20)
2305	GEN_INT_ENTRY h_doorbell, virt=1, ool=1
2306EXC_VIRT_END(h_doorbell, 0x4e80, 0x20)
2307EXC_COMMON_BEGIN(h_doorbell_common)
2308	GEN_COMMON h_doorbell
2309	FINISH_NAP
2310	RUNLATCH_ON
2311	addi	r3,r1,STACK_FRAME_OVERHEAD
2312#ifdef CONFIG_PPC_DOORBELL
2313	bl	doorbell_exception
2314#else
2315	bl	unknown_exception
2316#endif
2317	b	interrupt_return
2318
2319	GEN_KVM h_doorbell
2320
2321
2322/**
2323 * Interrupt 0xea0 - Hypervisor Virtualization Interrupt.
2324 * This is an asynchronous interrupt in response to an "external exception".
2325 * Similar to 0x500 but for host only.
2326 */
2327INT_DEFINE_BEGIN(h_virt_irq)
2328	IVEC=0xea0
2329	IHSRR=1
2330	IMASK=IRQS_DISABLED
2331	IKVM_REAL=1
2332	IKVM_VIRT=1
2333INT_DEFINE_END(h_virt_irq)
2334
2335EXC_REAL_BEGIN(h_virt_irq, 0xea0, 0x20)
2336	GEN_INT_ENTRY h_virt_irq, virt=0, ool=1
2337EXC_REAL_END(h_virt_irq, 0xea0, 0x20)
2338EXC_VIRT_BEGIN(h_virt_irq, 0x4ea0, 0x20)
2339	GEN_INT_ENTRY h_virt_irq, virt=1, ool=1
2340EXC_VIRT_END(h_virt_irq, 0x4ea0, 0x20)
2341EXC_COMMON_BEGIN(h_virt_irq_common)
2342	GEN_COMMON h_virt_irq
2343	FINISH_NAP
2344	RUNLATCH_ON
2345	addi	r3,r1,STACK_FRAME_OVERHEAD
2346	bl	do_IRQ
2347	b	interrupt_return
2348
2349	GEN_KVM h_virt_irq
2350
2351
2352EXC_REAL_NONE(0xec0, 0x20)
2353EXC_VIRT_NONE(0x4ec0, 0x20)
2354EXC_REAL_NONE(0xee0, 0x20)
2355EXC_VIRT_NONE(0x4ee0, 0x20)
2356
2357
2358/*
2359 * Interrupt 0xf00 - Performance Monitor Interrupt (PMI, PMU).
2360 * This is an asynchronous interrupt in response to a PMU exception.
2361 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
2362 * IRQS_PMI_DISABLED mask (NOTE: NOT local_irq_disable()).
2363 *
2364 * Handling:
2365 * This calls into the perf subsystem.
2366 *
2367 * Like the watchdog soft-nmi, it appears an NMI interrupt to Linux, in that it
2368 * runs under local_irq_disable. However it may be soft-masked in
2369 * powerpc-specific code.
2370 *
2371 * If soft masked, the masked handler will note the pending interrupt for
2372 * replay, and clear MSR[EE] in the interrupted context.
2373 */
2374INT_DEFINE_BEGIN(performance_monitor)
2375	IVEC=0xf00
2376	IMASK=IRQS_PMI_DISABLED
2377#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2378	IKVM_REAL=1
2379#endif
2380INT_DEFINE_END(performance_monitor)
2381
2382EXC_REAL_BEGIN(performance_monitor, 0xf00, 0x20)
2383	GEN_INT_ENTRY performance_monitor, virt=0, ool=1
2384EXC_REAL_END(performance_monitor, 0xf00, 0x20)
2385EXC_VIRT_BEGIN(performance_monitor, 0x4f00, 0x20)
2386	GEN_INT_ENTRY performance_monitor, virt=1, ool=1
2387EXC_VIRT_END(performance_monitor, 0x4f00, 0x20)
2388EXC_COMMON_BEGIN(performance_monitor_common)
2389	GEN_COMMON performance_monitor
2390	FINISH_NAP
2391	RUNLATCH_ON
2392	addi	r3,r1,STACK_FRAME_OVERHEAD
2393	bl	performance_monitor_exception
2394	b	interrupt_return
2395
2396	GEN_KVM performance_monitor
2397
2398
2399/**
2400 * Interrupt 0xf20 - Vector Unavailable Interrupt.
2401 * This is a synchronous interrupt in response to
2402 * executing a vector (or altivec) instruction with MSR[VEC]=0.
2403 * Similar to FP unavailable.
2404 */
2405INT_DEFINE_BEGIN(altivec_unavailable)
2406	IVEC=0xf20
2407	IRECONCILE=0
2408#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2409	IKVM_REAL=1
2410#endif
2411INT_DEFINE_END(altivec_unavailable)
2412
2413EXC_REAL_BEGIN(altivec_unavailable, 0xf20, 0x20)
2414	GEN_INT_ENTRY altivec_unavailable, virt=0, ool=1
2415EXC_REAL_END(altivec_unavailable, 0xf20, 0x20)
2416EXC_VIRT_BEGIN(altivec_unavailable, 0x4f20, 0x20)
2417	GEN_INT_ENTRY altivec_unavailable, virt=1, ool=1
2418EXC_VIRT_END(altivec_unavailable, 0x4f20, 0x20)
2419EXC_COMMON_BEGIN(altivec_unavailable_common)
2420	GEN_COMMON altivec_unavailable
2421#ifdef CONFIG_ALTIVEC
2422BEGIN_FTR_SECTION
2423	beq	1f
2424#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2425  BEGIN_FTR_SECTION_NESTED(69)
2426	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2427	 * transaction), go do TM stuff
2428	 */
2429	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2430	bne-	2f
2431  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2432#endif
2433	bl	load_up_altivec
2434	b	fast_interrupt_return
2435#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
24362:	/* User process was in a transaction */
2437	RECONCILE_IRQ_STATE(r10, r11)
2438	addi	r3,r1,STACK_FRAME_OVERHEAD
2439	bl	altivec_unavailable_tm
2440	b	interrupt_return
2441#endif
24421:
2443END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2444#endif
2445	RECONCILE_IRQ_STATE(r10, r11)
2446	addi	r3,r1,STACK_FRAME_OVERHEAD
2447	bl	altivec_unavailable_exception
2448	b	interrupt_return
2449
2450	GEN_KVM altivec_unavailable
2451
2452
2453/**
2454 * Interrupt 0xf40 - VSX Unavailable Interrupt.
2455 * This is a synchronous interrupt in response to
2456 * executing a VSX instruction with MSR[VSX]=0.
2457 * Similar to FP unavailable.
2458 */
2459INT_DEFINE_BEGIN(vsx_unavailable)
2460	IVEC=0xf40
2461	IRECONCILE=0
2462#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2463	IKVM_REAL=1
2464#endif
2465INT_DEFINE_END(vsx_unavailable)
2466
2467EXC_REAL_BEGIN(vsx_unavailable, 0xf40, 0x20)
2468	GEN_INT_ENTRY vsx_unavailable, virt=0, ool=1
2469EXC_REAL_END(vsx_unavailable, 0xf40, 0x20)
2470EXC_VIRT_BEGIN(vsx_unavailable, 0x4f40, 0x20)
2471	GEN_INT_ENTRY vsx_unavailable, virt=1, ool=1
2472EXC_VIRT_END(vsx_unavailable, 0x4f40, 0x20)
2473EXC_COMMON_BEGIN(vsx_unavailable_common)
2474	GEN_COMMON vsx_unavailable
2475#ifdef CONFIG_VSX
2476BEGIN_FTR_SECTION
2477	beq	1f
2478#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2479  BEGIN_FTR_SECTION_NESTED(69)
2480	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2481	 * transaction), go do TM stuff
2482	 */
2483	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2484	bne-	2f
2485  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2486#endif
2487	b	load_up_vsx
2488#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
24892:	/* User process was in a transaction */
2490	RECONCILE_IRQ_STATE(r10, r11)
2491	addi	r3,r1,STACK_FRAME_OVERHEAD
2492	bl	vsx_unavailable_tm
2493	b	interrupt_return
2494#endif
24951:
2496END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2497#endif
2498	RECONCILE_IRQ_STATE(r10, r11)
2499	addi	r3,r1,STACK_FRAME_OVERHEAD
2500	bl	vsx_unavailable_exception
2501	b	interrupt_return
2502
2503	GEN_KVM vsx_unavailable
2504
2505
2506/**
2507 * Interrupt 0xf60 - Facility Unavailable Interrupt.
2508 * This is a synchronous interrupt in response to
2509 * executing an instruction without access to the facility that can be
2510 * resolved by the OS (e.g., FSCR, MSR).
2511 * Similar to FP unavailable.
2512 */
2513INT_DEFINE_BEGIN(facility_unavailable)
2514	IVEC=0xf60
2515#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2516	IKVM_REAL=1
2517#endif
2518INT_DEFINE_END(facility_unavailable)
2519
2520EXC_REAL_BEGIN(facility_unavailable, 0xf60, 0x20)
2521	GEN_INT_ENTRY facility_unavailable, virt=0, ool=1
2522EXC_REAL_END(facility_unavailable, 0xf60, 0x20)
2523EXC_VIRT_BEGIN(facility_unavailable, 0x4f60, 0x20)
2524	GEN_INT_ENTRY facility_unavailable, virt=1, ool=1
2525EXC_VIRT_END(facility_unavailable, 0x4f60, 0x20)
2526EXC_COMMON_BEGIN(facility_unavailable_common)
2527	GEN_COMMON facility_unavailable
2528	addi	r3,r1,STACK_FRAME_OVERHEAD
2529	bl	facility_unavailable_exception
2530	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2531	b	interrupt_return
2532
2533	GEN_KVM facility_unavailable
2534
2535
2536/**
2537 * Interrupt 0xf60 - Hypervisor Facility Unavailable Interrupt.
2538 * This is a synchronous interrupt in response to
2539 * executing an instruction without access to the facility that can only
2540 * be resolved in HV mode (e.g., HFSCR).
2541 * Similar to FP unavailable.
2542 */
2543INT_DEFINE_BEGIN(h_facility_unavailable)
2544	IVEC=0xf80
2545	IHSRR=1
2546	IKVM_REAL=1
2547	IKVM_VIRT=1
2548INT_DEFINE_END(h_facility_unavailable)
2549
2550EXC_REAL_BEGIN(h_facility_unavailable, 0xf80, 0x20)
2551	GEN_INT_ENTRY h_facility_unavailable, virt=0, ool=1
2552EXC_REAL_END(h_facility_unavailable, 0xf80, 0x20)
2553EXC_VIRT_BEGIN(h_facility_unavailable, 0x4f80, 0x20)
2554	GEN_INT_ENTRY h_facility_unavailable, virt=1, ool=1
2555EXC_VIRT_END(h_facility_unavailable, 0x4f80, 0x20)
2556EXC_COMMON_BEGIN(h_facility_unavailable_common)
2557	GEN_COMMON h_facility_unavailable
2558	addi	r3,r1,STACK_FRAME_OVERHEAD
2559	bl	facility_unavailable_exception
2560	REST_NVGPRS(r1) /* XXX Shouldn't be necessary in practice */
2561	b	interrupt_return
2562
2563	GEN_KVM h_facility_unavailable
2564
2565
2566EXC_REAL_NONE(0xfa0, 0x20)
2567EXC_VIRT_NONE(0x4fa0, 0x20)
2568EXC_REAL_NONE(0xfc0, 0x20)
2569EXC_VIRT_NONE(0x4fc0, 0x20)
2570EXC_REAL_NONE(0xfe0, 0x20)
2571EXC_VIRT_NONE(0x4fe0, 0x20)
2572
2573EXC_REAL_NONE(0x1000, 0x100)
2574EXC_VIRT_NONE(0x5000, 0x100)
2575EXC_REAL_NONE(0x1100, 0x100)
2576EXC_VIRT_NONE(0x5100, 0x100)
2577
2578#ifdef CONFIG_CBE_RAS
2579INT_DEFINE_BEGIN(cbe_system_error)
2580	IVEC=0x1200
2581	IHSRR=1
2582	IKVM_SKIP=1
2583	IKVM_REAL=1
2584INT_DEFINE_END(cbe_system_error)
2585
2586EXC_REAL_BEGIN(cbe_system_error, 0x1200, 0x100)
2587	GEN_INT_ENTRY cbe_system_error, virt=0
2588EXC_REAL_END(cbe_system_error, 0x1200, 0x100)
2589EXC_VIRT_NONE(0x5200, 0x100)
2590EXC_COMMON_BEGIN(cbe_system_error_common)
2591	GEN_COMMON cbe_system_error
2592	addi	r3,r1,STACK_FRAME_OVERHEAD
2593	bl	cbe_system_error_exception
2594	b	interrupt_return
2595
2596	GEN_KVM cbe_system_error
2597
2598#else /* CONFIG_CBE_RAS */
2599EXC_REAL_NONE(0x1200, 0x100)
2600EXC_VIRT_NONE(0x5200, 0x100)
2601#endif
2602
2603
2604INT_DEFINE_BEGIN(instruction_breakpoint)
2605	IVEC=0x1300
2606#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2607	IKVM_SKIP=1
2608	IKVM_REAL=1
2609#endif
2610INT_DEFINE_END(instruction_breakpoint)
2611
2612EXC_REAL_BEGIN(instruction_breakpoint, 0x1300, 0x100)
2613	GEN_INT_ENTRY instruction_breakpoint, virt=0
2614EXC_REAL_END(instruction_breakpoint, 0x1300, 0x100)
2615EXC_VIRT_BEGIN(instruction_breakpoint, 0x5300, 0x100)
2616	GEN_INT_ENTRY instruction_breakpoint, virt=1
2617EXC_VIRT_END(instruction_breakpoint, 0x5300, 0x100)
2618EXC_COMMON_BEGIN(instruction_breakpoint_common)
2619	GEN_COMMON instruction_breakpoint
2620	addi	r3,r1,STACK_FRAME_OVERHEAD
2621	bl	instruction_breakpoint_exception
2622	b	interrupt_return
2623
2624	GEN_KVM instruction_breakpoint
2625
2626
2627EXC_REAL_NONE(0x1400, 0x100)
2628EXC_VIRT_NONE(0x5400, 0x100)
2629
2630/**
2631 * Interrupt 0x1500 - Soft Patch Interrupt
2632 *
2633 * Handling:
2634 * This is an implementation specific interrupt which can be used for a
2635 * range of exceptions.
2636 *
2637 * This interrupt handler is unique in that it runs the denormal assist
2638 * code even for guests (and even in guest context) without going to KVM,
2639 * for speed. POWER9 does not raise denorm exceptions, so this special case
2640 * could be phased out in future to reduce special cases.
2641 */
2642INT_DEFINE_BEGIN(denorm_exception)
2643	IVEC=0x1500
2644	IHSRR=1
2645	IBRANCH_TO_COMMON=0
2646	IKVM_REAL=1
2647INT_DEFINE_END(denorm_exception)
2648
2649EXC_REAL_BEGIN(denorm_exception, 0x1500, 0x100)
2650	GEN_INT_ENTRY denorm_exception, virt=0
2651#ifdef CONFIG_PPC_DENORMALISATION
2652	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2653	bne+	denorm_assist
2654#endif
2655	GEN_BRANCH_TO_COMMON denorm_exception, virt=0
2656EXC_REAL_END(denorm_exception, 0x1500, 0x100)
2657#ifdef CONFIG_PPC_DENORMALISATION
2658EXC_VIRT_BEGIN(denorm_exception, 0x5500, 0x100)
2659	GEN_INT_ENTRY denorm_exception, virt=1
2660	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2661	bne+	denorm_assist
2662	GEN_BRANCH_TO_COMMON denorm_exception, virt=1
2663EXC_VIRT_END(denorm_exception, 0x5500, 0x100)
2664#else
2665EXC_VIRT_NONE(0x5500, 0x100)
2666#endif
2667
2668#ifdef CONFIG_PPC_DENORMALISATION
2669TRAMP_REAL_BEGIN(denorm_assist)
2670BEGIN_FTR_SECTION
2671/*
2672 * To denormalise we need to move a copy of the register to itself.
2673 * For POWER6 do that here for all FP regs.
2674 */
2675	mfmsr	r10
2676	ori	r10,r10,(MSR_FP|MSR_FE0|MSR_FE1)
2677	xori	r10,r10,(MSR_FE0|MSR_FE1)
2678	mtmsrd	r10
2679	sync
2680
2681	.Lreg=0
2682	.rept 32
2683	fmr	.Lreg,.Lreg
2684	.Lreg=.Lreg+1
2685	.endr
2686
2687FTR_SECTION_ELSE
2688/*
2689 * To denormalise we need to move a copy of the register to itself.
2690 * For POWER7 do that here for the first 32 VSX registers only.
2691 */
2692	mfmsr	r10
2693	oris	r10,r10,MSR_VSX@h
2694	mtmsrd	r10
2695	sync
2696
2697	.Lreg=0
2698	.rept 32
2699	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2700	.Lreg=.Lreg+1
2701	.endr
2702
2703ALT_FTR_SECTION_END_IFCLR(CPU_FTR_ARCH_206)
2704
2705BEGIN_FTR_SECTION
2706	b	denorm_done
2707END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
2708/*
2709 * To denormalise we need to move a copy of the register to itself.
2710 * For POWER8 we need to do that for all 64 VSX registers
2711 */
2712	.Lreg=32
2713	.rept 32
2714	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2715	.Lreg=.Lreg+1
2716	.endr
2717
2718denorm_done:
2719	mfspr	r11,SPRN_HSRR0
2720	subi	r11,r11,4
2721	mtspr	SPRN_HSRR0,r11
2722	mtcrf	0x80,r9
2723	ld	r9,PACA_EXGEN+EX_R9(r13)
2724BEGIN_FTR_SECTION
2725	ld	r10,PACA_EXGEN+EX_PPR(r13)
2726	mtspr	SPRN_PPR,r10
2727END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
2728BEGIN_FTR_SECTION
2729	ld	r10,PACA_EXGEN+EX_CFAR(r13)
2730	mtspr	SPRN_CFAR,r10
2731END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
2732	ld	r10,PACA_EXGEN+EX_R10(r13)
2733	ld	r11,PACA_EXGEN+EX_R11(r13)
2734	ld	r12,PACA_EXGEN+EX_R12(r13)
2735	ld	r13,PACA_EXGEN+EX_R13(r13)
2736	HRFI_TO_UNKNOWN
2737	b	.
2738#endif
2739
2740EXC_COMMON_BEGIN(denorm_exception_common)
2741	GEN_COMMON denorm_exception
2742	addi	r3,r1,STACK_FRAME_OVERHEAD
2743	bl	unknown_exception
2744	b	interrupt_return
2745
2746	GEN_KVM denorm_exception
2747
2748
2749#ifdef CONFIG_CBE_RAS
2750INT_DEFINE_BEGIN(cbe_maintenance)
2751	IVEC=0x1600
2752	IHSRR=1
2753	IKVM_SKIP=1
2754	IKVM_REAL=1
2755INT_DEFINE_END(cbe_maintenance)
2756
2757EXC_REAL_BEGIN(cbe_maintenance, 0x1600, 0x100)
2758	GEN_INT_ENTRY cbe_maintenance, virt=0
2759EXC_REAL_END(cbe_maintenance, 0x1600, 0x100)
2760EXC_VIRT_NONE(0x5600, 0x100)
2761EXC_COMMON_BEGIN(cbe_maintenance_common)
2762	GEN_COMMON cbe_maintenance
2763	addi	r3,r1,STACK_FRAME_OVERHEAD
2764	bl	cbe_maintenance_exception
2765	b	interrupt_return
2766
2767	GEN_KVM cbe_maintenance
2768
2769#else /* CONFIG_CBE_RAS */
2770EXC_REAL_NONE(0x1600, 0x100)
2771EXC_VIRT_NONE(0x5600, 0x100)
2772#endif
2773
2774
2775INT_DEFINE_BEGIN(altivec_assist)
2776	IVEC=0x1700
2777#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2778	IKVM_REAL=1
2779#endif
2780INT_DEFINE_END(altivec_assist)
2781
2782EXC_REAL_BEGIN(altivec_assist, 0x1700, 0x100)
2783	GEN_INT_ENTRY altivec_assist, virt=0
2784EXC_REAL_END(altivec_assist, 0x1700, 0x100)
2785EXC_VIRT_BEGIN(altivec_assist, 0x5700, 0x100)
2786	GEN_INT_ENTRY altivec_assist, virt=1
2787EXC_VIRT_END(altivec_assist, 0x5700, 0x100)
2788EXC_COMMON_BEGIN(altivec_assist_common)
2789	GEN_COMMON altivec_assist
2790	addi	r3,r1,STACK_FRAME_OVERHEAD
2791#ifdef CONFIG_ALTIVEC
2792	bl	altivec_assist_exception
2793	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2794#else
2795	bl	unknown_exception
2796#endif
2797	b	interrupt_return
2798
2799	GEN_KVM altivec_assist
2800
2801
2802#ifdef CONFIG_CBE_RAS
2803INT_DEFINE_BEGIN(cbe_thermal)
2804	IVEC=0x1800
2805	IHSRR=1
2806	IKVM_SKIP=1
2807	IKVM_REAL=1
2808INT_DEFINE_END(cbe_thermal)
2809
2810EXC_REAL_BEGIN(cbe_thermal, 0x1800, 0x100)
2811	GEN_INT_ENTRY cbe_thermal, virt=0
2812EXC_REAL_END(cbe_thermal, 0x1800, 0x100)
2813EXC_VIRT_NONE(0x5800, 0x100)
2814EXC_COMMON_BEGIN(cbe_thermal_common)
2815	GEN_COMMON cbe_thermal
2816	addi	r3,r1,STACK_FRAME_OVERHEAD
2817	bl	cbe_thermal_exception
2818	b	interrupt_return
2819
2820	GEN_KVM cbe_thermal
2821
2822#else /* CONFIG_CBE_RAS */
2823EXC_REAL_NONE(0x1800, 0x100)
2824EXC_VIRT_NONE(0x5800, 0x100)
2825#endif
2826
2827
2828#ifdef CONFIG_PPC_WATCHDOG
2829
2830INT_DEFINE_BEGIN(soft_nmi)
2831	IVEC=0x900
2832	ISTACK=0
2833	IRECONCILE=0	/* Soft-NMI may fire under local_irq_disable */
2834INT_DEFINE_END(soft_nmi)
2835
2836/*
2837 * Branch to soft_nmi_interrupt using the emergency stack. The emergency
2838 * stack is one that is usable by maskable interrupts so long as MSR_EE
2839 * remains off. It is used for recovery when something has corrupted the
2840 * normal kernel stack, for example. The "soft NMI" must not use the process
2841 * stack because we want irq disabled sections to avoid touching the stack
2842 * at all (other than PMU interrupts), so use the emergency stack for this,
2843 * and run it entirely with interrupts hard disabled.
2844 */
2845EXC_COMMON_BEGIN(soft_nmi_common)
2846	mfspr	r11,SPRN_SRR0
2847	mr	r10,r1
2848	ld	r1,PACAEMERGSP(r13)
2849	subi	r1,r1,INT_FRAME_SIZE
2850	__GEN_COMMON_BODY soft_nmi
2851
2852	/*
2853	 * Set IRQS_ALL_DISABLED and save PACAIRQHAPPENED (see
2854	 * system_reset_common)
2855	 */
2856	li	r10,IRQS_ALL_DISABLED
2857	stb	r10,PACAIRQSOFTMASK(r13)
2858	lbz	r10,PACAIRQHAPPENED(r13)
2859	std	r10,RESULT(r1)
2860	ori	r10,r10,PACA_IRQ_HARD_DIS
2861	stb	r10,PACAIRQHAPPENED(r13)
2862
2863	addi	r3,r1,STACK_FRAME_OVERHEAD
2864	bl	soft_nmi_interrupt
2865
2866	/* Clear MSR_RI before setting SRR0 and SRR1. */
2867	li	r9,0
2868	mtmsrd	r9,1
2869
2870	/*
2871	 * Restore soft mask settings.
2872	 */
2873	ld	r10,RESULT(r1)
2874	stb	r10,PACAIRQHAPPENED(r13)
2875	ld	r10,SOFTE(r1)
2876	stb	r10,PACAIRQSOFTMASK(r13)
2877
2878	kuap_kernel_restore r9, r10
2879	EXCEPTION_RESTORE_REGS hsrr=0
2880	RFI_TO_KERNEL
2881
2882#endif /* CONFIG_PPC_WATCHDOG */
2883
2884/*
2885 * An interrupt came in while soft-disabled. We set paca->irq_happened, then:
2886 * - If it was a decrementer interrupt, we bump the dec to max and and return.
2887 * - If it was a doorbell we return immediately since doorbells are edge
2888 *   triggered and won't automatically refire.
2889 * - If it was a HMI we return immediately since we handled it in realmode
2890 *   and it won't refire.
2891 * - Else it is one of PACA_IRQ_MUST_HARD_MASK, so hard disable and return.
2892 * This is called with r10 containing the value to OR to the paca field.
2893 */
2894.macro MASKED_INTERRUPT hsrr=0
2895	.if \hsrr
2896masked_Hinterrupt:
2897	.else
2898masked_interrupt:
2899	.endif
2900	lbz	r11,PACAIRQHAPPENED(r13)
2901	or	r11,r11,r10
2902	stb	r11,PACAIRQHAPPENED(r13)
2903	cmpwi	r10,PACA_IRQ_DEC
2904	bne	1f
2905	lis	r10,0x7fff
2906	ori	r10,r10,0xffff
2907	mtspr	SPRN_DEC,r10
2908#ifdef CONFIG_PPC_WATCHDOG
2909	b	soft_nmi_common
2910#else
2911	b	2f
2912#endif
29131:	andi.	r10,r10,PACA_IRQ_MUST_HARD_MASK
2914	beq	2f
2915	xori	r12,r12,MSR_EE	/* clear MSR_EE */
2916	.if \hsrr
2917	mtspr	SPRN_HSRR1,r12
2918	.else
2919	mtspr	SPRN_SRR1,r12
2920	.endif
2921	ori	r11,r11,PACA_IRQ_HARD_DIS
2922	stb	r11,PACAIRQHAPPENED(r13)
29232:	/* done */
2924	ld	r10,PACA_EXGEN+EX_CTR(r13)
2925	mtctr	r10
2926	mtcrf	0x80,r9
2927	std	r1,PACAR1(r13)
2928	ld	r9,PACA_EXGEN+EX_R9(r13)
2929	ld	r10,PACA_EXGEN+EX_R10(r13)
2930	ld	r11,PACA_EXGEN+EX_R11(r13)
2931	ld	r12,PACA_EXGEN+EX_R12(r13)
2932	ld	r13,PACA_EXGEN+EX_R13(r13)
2933	/* May return to masked low address where r13 is not set up */
2934	.if \hsrr
2935	HRFI_TO_KERNEL
2936	.else
2937	RFI_TO_KERNEL
2938	.endif
2939	b	.
2940.endm
2941
2942TRAMP_REAL_BEGIN(stf_barrier_fallback)
2943	std	r9,PACA_EXRFI+EX_R9(r13)
2944	std	r10,PACA_EXRFI+EX_R10(r13)
2945	sync
2946	ld	r9,PACA_EXRFI+EX_R9(r13)
2947	ld	r10,PACA_EXRFI+EX_R10(r13)
2948	ori	31,31,0
2949	.rept 14
2950	b	1f
29511:
2952	.endr
2953	blr
2954
2955/* Clobbers r10, r11, ctr */
2956.macro L1D_DISPLACEMENT_FLUSH
2957	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
2958	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
2959	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
2960	mtctr	r11
2961	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
2962
2963	/* order ld/st prior to dcbt stop all streams with flushing */
2964	sync
2965
2966	/*
2967	 * The load addresses are at staggered offsets within cachelines,
2968	 * which suits some pipelines better (on others it should not
2969	 * hurt).
2970	 */
29711:
2972	ld	r11,(0x80 + 8)*0(r10)
2973	ld	r11,(0x80 + 8)*1(r10)
2974	ld	r11,(0x80 + 8)*2(r10)
2975	ld	r11,(0x80 + 8)*3(r10)
2976	ld	r11,(0x80 + 8)*4(r10)
2977	ld	r11,(0x80 + 8)*5(r10)
2978	ld	r11,(0x80 + 8)*6(r10)
2979	ld	r11,(0x80 + 8)*7(r10)
2980	addi	r10,r10,0x80*8
2981	bdnz	1b
2982.endm
2983
2984TRAMP_REAL_BEGIN(entry_flush_fallback)
2985	std	r9,PACA_EXRFI+EX_R9(r13)
2986	std	r10,PACA_EXRFI+EX_R10(r13)
2987	std	r11,PACA_EXRFI+EX_R11(r13)
2988	mfctr	r9
2989	L1D_DISPLACEMENT_FLUSH
2990	mtctr	r9
2991	ld	r9,PACA_EXRFI+EX_R9(r13)
2992	ld	r10,PACA_EXRFI+EX_R10(r13)
2993	ld	r11,PACA_EXRFI+EX_R11(r13)
2994	blr
2995
2996/*
2997 * The SCV entry flush happens with interrupts enabled, so it must disable
2998 * to prevent EXRFI being clobbered by NMIs (e.g., soft_nmi_common). r10
2999 * (containing LR) does not need to be preserved here because scv entry
3000 * puts 0 in the pt_regs, CTR can be clobbered for the same reason.
3001 */
3002TRAMP_REAL_BEGIN(scv_entry_flush_fallback)
3003	li	r10,0
3004	mtmsrd	r10,1
3005	lbz	r10,PACAIRQHAPPENED(r13)
3006	ori	r10,r10,PACA_IRQ_HARD_DIS
3007	stb	r10,PACAIRQHAPPENED(r13)
3008	std	r11,PACA_EXRFI+EX_R11(r13)
3009	L1D_DISPLACEMENT_FLUSH
3010	ld	r11,PACA_EXRFI+EX_R11(r13)
3011	li	r10,MSR_RI
3012	mtmsrd	r10,1
3013	blr
3014
3015TRAMP_REAL_BEGIN(rfi_flush_fallback)
3016	SET_SCRATCH0(r13);
3017	GET_PACA(r13);
3018	std	r1,PACA_EXRFI+EX_R12(r13)
3019	ld	r1,PACAKSAVE(r13)
3020	std	r9,PACA_EXRFI+EX_R9(r13)
3021	std	r10,PACA_EXRFI+EX_R10(r13)
3022	std	r11,PACA_EXRFI+EX_R11(r13)
3023	mfctr	r9
3024	L1D_DISPLACEMENT_FLUSH
3025	mtctr	r9
3026	ld	r9,PACA_EXRFI+EX_R9(r13)
3027	ld	r10,PACA_EXRFI+EX_R10(r13)
3028	ld	r11,PACA_EXRFI+EX_R11(r13)
3029	ld	r1,PACA_EXRFI+EX_R12(r13)
3030	GET_SCRATCH0(r13);
3031	rfid
3032
3033TRAMP_REAL_BEGIN(hrfi_flush_fallback)
3034	SET_SCRATCH0(r13);
3035	GET_PACA(r13);
3036	std	r1,PACA_EXRFI+EX_R12(r13)
3037	ld	r1,PACAKSAVE(r13)
3038	std	r9,PACA_EXRFI+EX_R9(r13)
3039	std	r10,PACA_EXRFI+EX_R10(r13)
3040	std	r11,PACA_EXRFI+EX_R11(r13)
3041	mfctr	r9
3042	L1D_DISPLACEMENT_FLUSH
3043	mtctr	r9
3044	ld	r9,PACA_EXRFI+EX_R9(r13)
3045	ld	r10,PACA_EXRFI+EX_R10(r13)
3046	ld	r11,PACA_EXRFI+EX_R11(r13)
3047	ld	r1,PACA_EXRFI+EX_R12(r13)
3048	GET_SCRATCH0(r13);
3049	hrfid
3050
3051TRAMP_REAL_BEGIN(rfscv_flush_fallback)
3052	/* system call volatile */
3053	mr	r7,r13
3054	GET_PACA(r13);
3055	mr	r8,r1
3056	ld	r1,PACAKSAVE(r13)
3057	mfctr	r9
3058	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
3059	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
3060	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
3061	mtctr	r11
3062	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
3063
3064	/* order ld/st prior to dcbt stop all streams with flushing */
3065	sync
3066
3067	/*
3068	 * The load adresses are at staggered offsets within cachelines,
3069	 * which suits some pipelines better (on others it should not
3070	 * hurt).
3071	 */
30721:
3073	ld	r11,(0x80 + 8)*0(r10)
3074	ld	r11,(0x80 + 8)*1(r10)
3075	ld	r11,(0x80 + 8)*2(r10)
3076	ld	r11,(0x80 + 8)*3(r10)
3077	ld	r11,(0x80 + 8)*4(r10)
3078	ld	r11,(0x80 + 8)*5(r10)
3079	ld	r11,(0x80 + 8)*6(r10)
3080	ld	r11,(0x80 + 8)*7(r10)
3081	addi	r10,r10,0x80*8
3082	bdnz	1b
3083
3084	mtctr	r9
3085	li	r9,0
3086	li	r10,0
3087	li	r11,0
3088	mr	r1,r8
3089	mr	r13,r7
3090	RFSCV
3091
3092USE_TEXT_SECTION()
3093
3094_GLOBAL(do_uaccess_flush)
3095	UACCESS_FLUSH_FIXUP_SECTION
3096	nop
3097	nop
3098	nop
3099	blr
3100	L1D_DISPLACEMENT_FLUSH
3101	blr
3102_ASM_NOKPROBE_SYMBOL(do_uaccess_flush)
3103EXPORT_SYMBOL(do_uaccess_flush)
3104
3105
3106MASKED_INTERRUPT
3107MASKED_INTERRUPT hsrr=1
3108
3109#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
3110kvmppc_skip_interrupt:
3111	/*
3112	 * Here all GPRs are unchanged from when the interrupt happened
3113	 * except for r13, which is saved in SPRG_SCRATCH0.
3114	 */
3115	mfspr	r13, SPRN_SRR0
3116	addi	r13, r13, 4
3117	mtspr	SPRN_SRR0, r13
3118	GET_SCRATCH0(r13)
3119	RFI_TO_KERNEL
3120	b	.
3121
3122kvmppc_skip_Hinterrupt:
3123	/*
3124	 * Here all GPRs are unchanged from when the interrupt happened
3125	 * except for r13, which is saved in SPRG_SCRATCH0.
3126	 */
3127	mfspr	r13, SPRN_HSRR0
3128	addi	r13, r13, 4
3129	mtspr	SPRN_HSRR0, r13
3130	GET_SCRATCH0(r13)
3131	HRFI_TO_KERNEL
3132	b	.
3133#endif
3134
3135	/*
3136	 * Relocation-on interrupts: A subset of the interrupts can be delivered
3137	 * with IR=1/DR=1, if AIL==2 and MSR.HV won't be changed by delivering
3138	 * it.  Addresses are the same as the original interrupt addresses, but
3139	 * offset by 0xc000000000004000.
3140	 * It's impossible to receive interrupts below 0x300 via this mechanism.
3141	 * KVM: None of these traps are from the guest ; anything that escalated
3142	 * to HV=1 from HV=0 is delivered via real mode handlers.
3143	 */
3144
3145	/*
3146	 * This uses the standard macro, since the original 0x300 vector
3147	 * only has extra guff for STAB-based processors -- which never
3148	 * come here.
3149	 */
3150
3151EXC_COMMON_BEGIN(ppc64_runlatch_on_trampoline)
3152	b	__ppc64_runlatch_on
3153
3154USE_FIXED_SECTION(virt_trampolines)
3155	/*
3156	 * All code below __end_interrupts is treated as soft-masked. If
3157	 * any code runs here with MSR[EE]=1, it must then cope with pending
3158	 * soft interrupt being raised (i.e., by ensuring it is replayed).
3159	 *
3160	 * The __end_interrupts marker must be past the out-of-line (OOL)
3161	 * handlers, so that they are copied to real address 0x100 when running
3162	 * a relocatable kernel. This ensures they can be reached from the short
3163	 * trampoline handlers (like 0x4f00, 0x4f20, etc.) which branch
3164	 * directly, without using LOAD_HANDLER().
3165	 */
3166	.align	7
3167	.globl	__end_interrupts
3168__end_interrupts:
3169DEFINE_FIXED_SYMBOL(__end_interrupts)
3170
3171#ifdef CONFIG_PPC_970_NAP
3172	/*
3173	 * Called by exception entry code if _TLF_NAPPING was set, this clears
3174	 * the NAPPING flag, and redirects the exception exit to
3175	 * power4_fixup_nap_return.
3176	 */
3177	.globl power4_fixup_nap
3178EXC_COMMON_BEGIN(power4_fixup_nap)
3179	andc	r9,r9,r10
3180	std	r9,TI_LOCAL_FLAGS(r11)
3181	LOAD_REG_ADDR(r10, power4_idle_nap_return)
3182	std	r10,_NIP(r1)
3183	blr
3184
3185power4_idle_nap_return:
3186	blr
3187#endif
3188
3189CLOSE_FIXED_SECTION(real_vectors);
3190CLOSE_FIXED_SECTION(real_trampolines);
3191CLOSE_FIXED_SECTION(virt_vectors);
3192CLOSE_FIXED_SECTION(virt_trampolines);
3193
3194USE_TEXT_SECTION()
3195
3196/* MSR[RI] should be clear because this uses SRR[01] */
3197enable_machine_check:
3198	mflr	r0
3199	bcl	20,31,$+4
32000:	mflr	r3
3201	addi	r3,r3,(1f - 0b)
3202	mtspr	SPRN_SRR0,r3
3203	mfmsr	r3
3204	ori	r3,r3,MSR_ME
3205	mtspr	SPRN_SRR1,r3
3206	RFI_TO_KERNEL
32071:	mtlr	r0
3208	blr
3209
3210/* MSR[RI] should be clear because this uses SRR[01] */
3211disable_machine_check:
3212	mflr	r0
3213	bcl	20,31,$+4
32140:	mflr	r3
3215	addi	r3,r3,(1f - 0b)
3216	mtspr	SPRN_SRR0,r3
3217	mfmsr	r3
3218	li	r4,MSR_ME
3219	andc	r3,r3,r4
3220	mtspr	SPRN_SRR1,r3
3221	RFI_TO_KERNEL
32221:	mtlr	r0
3223	blr
3224
3225/*
3226 * Hash table stuff
3227 */
3228	.balign	IFETCH_ALIGN_BYTES
3229do_hash_page:
3230#ifdef CONFIG_PPC_BOOK3S_64
3231	lis	r0,(DSISR_BAD_FAULT_64S | DSISR_DABRMATCH | DSISR_KEYFAULT)@h
3232	ori	r0,r0,DSISR_BAD_FAULT_64S@l
3233	and.	r0,r5,r0		/* weird error? */
3234	bne-	handle_page_fault	/* if not, try to insert a HPTE */
3235
3236	/*
3237	 * If we are in an "NMI" (e.g., an interrupt when soft-disabled), then
3238	 * don't call hash_page, just fail the fault. This is required to
3239	 * prevent re-entrancy problems in the hash code, namely perf
3240	 * interrupts hitting while something holds H_PAGE_BUSY, and taking a
3241	 * hash fault. See the comment in hash_preload().
3242	 */
3243	ld	r11, PACA_THREAD_INFO(r13)
3244	lwz	r0,TI_PREEMPT(r11)
3245	andis.	r0,r0,NMI_MASK@h
3246	bne	77f
3247
3248	/*
3249	 * r3 contains the trap number
3250	 * r4 contains the faulting address
3251	 * r5 contains dsisr
3252	 * r6 msr
3253	 *
3254	 * at return r3 = 0 for success, 1 for page fault, negative for error
3255	 */
3256	bl	__hash_page		/* build HPTE if possible */
3257        cmpdi	r3,0			/* see if __hash_page succeeded */
3258
3259	/* Success */
3260	beq	interrupt_return	/* Return from exception on success */
3261
3262	/* Error */
3263	blt-	13f
3264
3265	/* Reload DAR/DSISR into r4/r5 for the DABR check below */
3266	ld	r4,_DAR(r1)
3267	ld      r5,_DSISR(r1)
3268#endif /* CONFIG_PPC_BOOK3S_64 */
3269
3270/* Here we have a page fault that hash_page can't handle. */
3271handle_page_fault:
327211:	andis.  r0,r5,DSISR_DABRMATCH@h
3273	bne-    handle_dabr_fault
3274	addi	r3,r1,STACK_FRAME_OVERHEAD
3275	bl	do_page_fault
3276	cmpdi	r3,0
3277	beq+	interrupt_return
3278	mr	r5,r3
3279	addi	r3,r1,STACK_FRAME_OVERHEAD
3280	ld	r4,_DAR(r1)
3281	bl	__bad_page_fault
3282	b	interrupt_return
3283
3284/* We have a data breakpoint exception - handle it */
3285handle_dabr_fault:
3286	ld      r4,_DAR(r1)
3287	ld      r5,_DSISR(r1)
3288	addi    r3,r1,STACK_FRAME_OVERHEAD
3289	bl      do_break
3290	/*
3291	 * do_break() may have changed the NV GPRS while handling a breakpoint.
3292	 * If so, we need to restore them with their updated values.
3293	 */
3294	REST_NVGPRS(r1)
3295	b       interrupt_return
3296
3297
3298#ifdef CONFIG_PPC_BOOK3S_64
3299/* We have a page fault that hash_page could handle but HV refused
3300 * the PTE insertion
3301 */
330213:	mr	r5,r3
3303	addi	r3,r1,STACK_FRAME_OVERHEAD
3304	ld	r4,_DAR(r1)
3305	bl	low_hash_fault
3306	b	interrupt_return
3307#endif
3308
3309/*
3310 * We come here as a result of a DSI at a point where we don't want
3311 * to call hash_page, such as when we are accessing memory (possibly
3312 * user memory) inside a PMU interrupt that occurred while interrupts
3313 * were soft-disabled.  We want to invoke the exception handler for
3314 * the access, or panic if there isn't a handler.
3315 */
331677:	addi	r3,r1,STACK_FRAME_OVERHEAD
3317	li	r5,SIGSEGV
3318	bl	bad_page_fault
3319	b	interrupt_return
3320