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,PACA_EXGEN+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,PACA_EXGEN+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		lbz	r10,PACAIRQSOFTMASK(r13)
512		andi.	r10,r10,IMASK
513		/* Associate vector numbers with bits in paca->irq_happened */
514		.if IVEC == 0x500 || IVEC == 0xea0
515		li	r10,PACA_IRQ_EE
516		.elseif IVEC == 0x900
517		li	r10,PACA_IRQ_DEC
518		.elseif IVEC == 0xa00 || IVEC == 0xe80
519		li	r10,PACA_IRQ_DBELL
520		.elseif IVEC == 0xe60
521		li	r10,PACA_IRQ_HMI
522		.elseif IVEC == 0xf00
523		li	r10,PACA_IRQ_PMI
524		.else
525		.abort "Bad maskable vector"
526		.endif
527
528		.if IHSRR_IF_HVMODE
529		BEGIN_FTR_SECTION
530		bne	masked_Hinterrupt
531		FTR_SECTION_ELSE
532		bne	masked_interrupt
533		ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
534		.elseif IHSRR
535		bne	masked_Hinterrupt
536		.else
537		bne	masked_interrupt
538		.endif
539	.endif
540
541	.if ISTACK
542	andi.	r10,r12,MSR_PR		/* See if coming from user	*/
543	mr	r10,r1			/* Save r1			*/
544	subi	r1,r1,INT_FRAME_SIZE	/* alloc frame on kernel stack	*/
545	beq-	100f
546	ld	r1,PACAKSAVE(r13)	/* kernel stack to use		*/
547100:	tdgei	r1,-INT_FRAME_SIZE	/* trap if r1 is in userspace	*/
548	EMIT_BUG_ENTRY 100b,__FILE__,__LINE__,0
549	.endif
550
551	std	r9,_CCR(r1)		/* save CR in stackframe	*/
552	std	r11,_NIP(r1)		/* save SRR0 in stackframe	*/
553	std	r12,_MSR(r1)		/* save SRR1 in stackframe	*/
554	std	r10,0(r1)		/* make stack chain pointer	*/
555	std	r0,GPR0(r1)		/* save r0 in stackframe	*/
556	std	r10,GPR1(r1)		/* save r1 in stackframe	*/
557
558	.if ISET_RI
559	li	r10,MSR_RI
560	mtmsrd	r10,1			/* Set MSR_RI */
561	.endif
562
563	.if ISTACK
564	.if IKUAP
565	kuap_save_amr_and_lock r9, r10, cr1, cr0
566	.endif
567	beq	101f			/* if from kernel mode		*/
568	ACCOUNT_CPU_USER_ENTRY(r13, r9, r10)
569BEGIN_FTR_SECTION
570	ld	r9,IAREA+EX_PPR(r13)	/* Read PPR from paca		*/
571	std	r9,_PPR(r1)
572END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
573101:
574	.else
575	.if IKUAP
576	kuap_save_amr_and_lock r9, r10, cr1
577	.endif
578	.endif
579
580	/* Save original regs values from save area to stack frame. */
581	ld	r9,IAREA+EX_R9(r13)	/* move r9, r10 to stackframe	*/
582	ld	r10,IAREA+EX_R10(r13)
583	std	r9,GPR9(r1)
584	std	r10,GPR10(r1)
585	ld	r9,IAREA+EX_R11(r13)	/* move r11 - r13 to stackframe	*/
586	ld	r10,IAREA+EX_R12(r13)
587	ld	r11,IAREA+EX_R13(r13)
588	std	r9,GPR11(r1)
589	std	r10,GPR12(r1)
590	std	r11,GPR13(r1)
591
592	SAVE_NVGPRS(r1)
593
594	.if IDAR
595	.if IISIDE
596	ld	r10,_NIP(r1)
597	.else
598	ld	r10,IAREA+EX_DAR(r13)
599	.endif
600	std	r10,_DAR(r1)
601	.endif
602
603	.if IDSISR
604	.if IISIDE
605	ld	r10,_MSR(r1)
606	lis	r11,DSISR_SRR1_MATCH_64S@h
607	and	r10,r10,r11
608	.else
609	lwz	r10,IAREA+EX_DSISR(r13)
610	.endif
611	std	r10,_DSISR(r1)
612	.endif
613
614BEGIN_FTR_SECTION
615	ld	r10,IAREA+EX_CFAR(r13)
616	std	r10,ORIG_GPR3(r1)
617END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
618	ld	r10,IAREA+EX_CTR(r13)
619	std	r10,_CTR(r1)
620	std	r2,GPR2(r1)		/* save r2 in stackframe	*/
621	SAVE_4GPRS(3, r1)		/* save r3 - r6 in stackframe   */
622	SAVE_2GPRS(7, r1)		/* save r7, r8 in stackframe	*/
623	mflr	r9			/* Get LR, later save to stack	*/
624	ld	r2,PACATOC(r13)		/* get kernel TOC into r2	*/
625	std	r9,_LINK(r1)
626	lbz	r10,PACAIRQSOFTMASK(r13)
627	mfspr	r11,SPRN_XER		/* save XER in stackframe	*/
628	std	r10,SOFTE(r1)
629	std	r11,_XER(r1)
630	li	r9,IVEC
631	std	r9,_TRAP(r1)		/* set trap number		*/
632	li	r10,0
633	ld	r11,exception_marker@toc(r2)
634	std	r10,RESULT(r1)		/* clear regs->result		*/
635	std	r11,STACK_FRAME_OVERHEAD-16(r1) /* mark the frame	*/
636
637	.if ISTACK
638	ACCOUNT_STOLEN_TIME
639	.endif
640
641	.if IRECONCILE
642	RECONCILE_IRQ_STATE(r10, r11)
643	.endif
644.endm
645
646/*
647 * On entry r13 points to the paca, r9-r13 are saved in the paca,
648 * r9 contains the saved CR, r11 and r12 contain the saved SRR0 and
649 * SRR1, and relocation is on.
650 *
651 * If stack=0, then the stack is already set in r1, and r1 is saved in r10.
652 * PPR save and CPU accounting is not done for the !stack case (XXX why not?)
653 */
654.macro GEN_COMMON name
655	__GEN_COMMON_ENTRY \name
656	__GEN_COMMON_BODY \name
657.endm
658
659/*
660 * Restore all registers including H/SRR0/1 saved in a stack frame of a
661 * standard exception.
662 */
663.macro EXCEPTION_RESTORE_REGS hsrr=0
664	/* Move original SRR0 and SRR1 into the respective regs */
665	ld	r9,_MSR(r1)
666	.if \hsrr
667	mtspr	SPRN_HSRR1,r9
668	.else
669	mtspr	SPRN_SRR1,r9
670	.endif
671	ld	r9,_NIP(r1)
672	.if \hsrr
673	mtspr	SPRN_HSRR0,r9
674	.else
675	mtspr	SPRN_SRR0,r9
676	.endif
677	ld	r9,_CTR(r1)
678	mtctr	r9
679	ld	r9,_XER(r1)
680	mtxer	r9
681	ld	r9,_LINK(r1)
682	mtlr	r9
683	ld	r9,_CCR(r1)
684	mtcr	r9
685	REST_8GPRS(2, r1)
686	REST_4GPRS(10, r1)
687	REST_GPR(0, r1)
688	/* restore original r1. */
689	ld	r1,GPR1(r1)
690.endm
691
692#define RUNLATCH_ON				\
693BEGIN_FTR_SECTION				\
694	ld	r3, PACA_THREAD_INFO(r13);	\
695	ld	r4,TI_LOCAL_FLAGS(r3);		\
696	andi.	r0,r4,_TLF_RUNLATCH;		\
697	beql	ppc64_runlatch_on_trampoline;	\
698END_FTR_SECTION_IFSET(CPU_FTR_CTRL)
699
700/*
701 * When the idle code in power4_idle puts the CPU into NAP mode,
702 * it has to do so in a loop, and relies on the external interrupt
703 * and decrementer interrupt entry code to get it out of the loop.
704 * It sets the _TLF_NAPPING bit in current_thread_info()->local_flags
705 * to signal that it is in the loop and needs help to get out.
706 */
707#ifdef CONFIG_PPC_970_NAP
708#define FINISH_NAP				\
709BEGIN_FTR_SECTION				\
710	ld	r11, PACA_THREAD_INFO(r13);	\
711	ld	r9,TI_LOCAL_FLAGS(r11);		\
712	andi.	r10,r9,_TLF_NAPPING;		\
713	bnel	power4_fixup_nap;		\
714END_FTR_SECTION_IFSET(CPU_FTR_CAN_NAP)
715#else
716#define FINISH_NAP
717#endif
718
719/*
720 * There are a few constraints to be concerned with.
721 * - Real mode exceptions code/data must be located at their physical location.
722 * - Virtual mode exceptions must be mapped at their 0xc000... location.
723 * - Fixed location code must not call directly beyond the __end_interrupts
724 *   area when built with CONFIG_RELOCATABLE. LOAD_HANDLER / bctr sequence
725 *   must be used.
726 * - LOAD_HANDLER targets must be within first 64K of physical 0 /
727 *   virtual 0xc00...
728 * - Conditional branch targets must be within +/-32K of caller.
729 *
730 * "Virtual exceptions" run with relocation on (MSR_IR=1, MSR_DR=1), and
731 * therefore don't have to run in physically located code or rfid to
732 * virtual mode kernel code. However on relocatable kernels they do have
733 * to branch to KERNELBASE offset because the rest of the kernel (outside
734 * the exception vectors) may be located elsewhere.
735 *
736 * Virtual exceptions correspond with physical, except their entry points
737 * are offset by 0xc000000000000000 and also tend to get an added 0x4000
738 * offset applied. Virtual exceptions are enabled with the Alternate
739 * Interrupt Location (AIL) bit set in the LPCR. However this does not
740 * guarantee they will be delivered virtually. Some conditions (see the ISA)
741 * cause exceptions to be delivered in real mode.
742 *
743 * It's impossible to receive interrupts below 0x300 via AIL.
744 *
745 * KVM: None of the virtual exceptions are from the guest. Anything that
746 * escalated to HV=1 from HV=0 is delivered via real mode handlers.
747 *
748 *
749 * We layout physical memory as follows:
750 * 0x0000 - 0x00ff : Secondary processor spin code
751 * 0x0100 - 0x18ff : Real mode pSeries interrupt vectors
752 * 0x1900 - 0x3fff : Real mode trampolines
753 * 0x4000 - 0x58ff : Relon (IR=1,DR=1) mode pSeries interrupt vectors
754 * 0x5900 - 0x6fff : Relon mode trampolines
755 * 0x7000 - 0x7fff : FWNMI data area
756 * 0x8000 -   .... : Common interrupt handlers, remaining early
757 *                   setup code, rest of kernel.
758 *
759 * We could reclaim 0x4000-0x42ff for real mode trampolines if the space
760 * is necessary. Until then it's more consistent to explicitly put VIRT_NONE
761 * vectors there.
762 */
763OPEN_FIXED_SECTION(real_vectors,        0x0100, 0x1900)
764OPEN_FIXED_SECTION(real_trampolines,    0x1900, 0x4000)
765OPEN_FIXED_SECTION(virt_vectors,        0x4000, 0x5900)
766OPEN_FIXED_SECTION(virt_trampolines,    0x5900, 0x7000)
767
768#ifdef CONFIG_PPC_POWERNV
769	.globl start_real_trampolines
770	.globl end_real_trampolines
771	.globl start_virt_trampolines
772	.globl end_virt_trampolines
773#endif
774
775#if defined(CONFIG_PPC_PSERIES) || defined(CONFIG_PPC_POWERNV)
776/*
777 * Data area reserved for FWNMI option.
778 * This address (0x7000) is fixed by the RPA.
779 * pseries and powernv need to keep the whole page from
780 * 0x7000 to 0x8000 free for use by the firmware
781 */
782ZERO_FIXED_SECTION(fwnmi_page,          0x7000, 0x8000)
783OPEN_TEXT_SECTION(0x8000)
784#else
785OPEN_TEXT_SECTION(0x7000)
786#endif
787
788USE_FIXED_SECTION(real_vectors)
789
790/*
791 * This is the start of the interrupt handlers for pSeries
792 * This code runs with relocation off.
793 * Code from here to __end_interrupts gets copied down to real
794 * address 0x100 when we are running a relocatable kernel.
795 * Therefore any relative branches in this section must only
796 * branch to labels in this section.
797 */
798	.globl __start_interrupts
799__start_interrupts:
800
801/* No virt vectors corresponding with 0x0..0x100 */
802EXC_VIRT_NONE(0x4000, 0x100)
803
804
805/**
806 * Interrupt 0x100 - System Reset Interrupt (SRESET aka NMI).
807 * This is a non-maskable, asynchronous interrupt always taken in real-mode.
808 * It is caused by:
809 * - Wake from power-saving state, on powernv.
810 * - An NMI from another CPU, triggered by firmware or hypercall.
811 * - As crash/debug signal injected from BMC, firmware or hypervisor.
812 *
813 * Handling:
814 * Power-save wakeup is the only performance critical path, so this is
815 * determined quickly as possible first. In this case volatile registers
816 * can be discarded and SPRs like CFAR don't need to be read.
817 *
818 * If not a powersave wakeup, then it's run as a regular interrupt, however
819 * it uses its own stack and PACA save area to preserve the regular kernel
820 * environment for debugging.
821 *
822 * This interrupt is not maskable, so triggering it when MSR[RI] is clear,
823 * or SCRATCH0 is in use, etc. may cause a crash. It's also not entirely
824 * correct to switch to virtual mode to run the regular interrupt handler
825 * because it might be interrupted when the MMU is in a bad state (e.g., SLB
826 * is clear).
827 *
828 * FWNMI:
829 * PAPR specifies a "fwnmi" facility which sends the sreset to a different
830 * entry point with a different register set up. Some hypervisors will
831 * send the sreset to 0x100 in the guest if it is not fwnmi capable.
832 *
833 * KVM:
834 * Unlike most SRR interrupts, this may be taken by the host while executing
835 * in a guest, so a KVM test is required. KVM will pull the CPU out of guest
836 * mode and then raise the sreset.
837 */
838INT_DEFINE_BEGIN(system_reset)
839	IVEC=0x100
840	IAREA=PACA_EXNMI
841	IVIRT=0 /* no virt entry point */
842	/*
843	 * MSR_RI is not enabled, because PACA_EXNMI and nmi stack is
844	 * being used, so a nested NMI exception would corrupt it.
845	 */
846	ISET_RI=0
847	ISTACK=0
848	IRECONCILE=0
849	IKVM_REAL=1
850INT_DEFINE_END(system_reset)
851
852EXC_REAL_BEGIN(system_reset, 0x100, 0x100)
853#ifdef CONFIG_PPC_P7_NAP
854	/*
855	 * If running native on arch 2.06 or later, check if we are waking up
856	 * from nap/sleep/winkle, and branch to idle handler. This tests SRR1
857	 * bits 46:47. A non-0 value indicates that we are coming from a power
858	 * saving state. The idle wakeup handler initially runs in real mode,
859	 * but we branch to the 0xc000... address so we can turn on relocation
860	 * with mtmsrd later, after SPRs are restored.
861	 *
862	 * Careful to minimise cost for the fast path (idle wakeup) while
863	 * also avoiding clobbering CFAR for the debug path (non-idle).
864	 *
865	 * For the idle wake case volatile registers can be clobbered, which
866	 * is why we use those initially. If it turns out to not be an idle
867	 * wake, carefully put everything back the way it was, so we can use
868	 * common exception macros to handle it.
869	 */
870BEGIN_FTR_SECTION
871	SET_SCRATCH0(r13)
872	GET_PACA(r13)
873	std	r3,PACA_EXNMI+0*8(r13)
874	std	r4,PACA_EXNMI+1*8(r13)
875	std	r5,PACA_EXNMI+2*8(r13)
876	mfspr	r3,SPRN_SRR1
877	mfocrf	r4,0x80
878	rlwinm.	r5,r3,47-31,30,31
879	bne+	system_reset_idle_wake
880	/* Not powersave wakeup. Restore regs for regular interrupt handler. */
881	mtocrf	0x80,r4
882	ld	r3,PACA_EXNMI+0*8(r13)
883	ld	r4,PACA_EXNMI+1*8(r13)
884	ld	r5,PACA_EXNMI+2*8(r13)
885	GET_SCRATCH0(r13)
886END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
887#endif
888
889	GEN_INT_ENTRY system_reset, virt=0
890	/*
891	 * In theory, we should not enable relocation here if it was disabled
892	 * in SRR1, because the MMU may not be configured to support it (e.g.,
893	 * SLB may have been cleared). In practice, there should only be a few
894	 * small windows where that's the case, and sreset is considered to
895	 * be dangerous anyway.
896	 */
897EXC_REAL_END(system_reset, 0x100, 0x100)
898EXC_VIRT_NONE(0x4100, 0x100)
899
900#ifdef CONFIG_PPC_P7_NAP
901TRAMP_REAL_BEGIN(system_reset_idle_wake)
902	/* We are waking up from idle, so may clobber any volatile register */
903	cmpwi	cr1,r5,2
904	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
905	BRANCH_TO_C000(r12, DOTSYM(idle_return_gpr_loss))
906#endif
907
908#ifdef CONFIG_PPC_PSERIES
909/*
910 * Vectors for the FWNMI option.  Share common code.
911 */
912TRAMP_REAL_BEGIN(system_reset_fwnmi)
913	/* XXX: fwnmi guest could run a nested/PR guest, so why no test?  */
914	__IKVM_REAL(system_reset)=0
915	GEN_INT_ENTRY system_reset, virt=0
916
917#endif /* CONFIG_PPC_PSERIES */
918
919EXC_COMMON_BEGIN(system_reset_common)
920	__GEN_COMMON_ENTRY system_reset
921	/*
922	 * Increment paca->in_nmi then enable MSR_RI. SLB or MCE will be able
923	 * to recover, but nested NMI will notice in_nmi and not recover
924	 * because of the use of the NMI stack. in_nmi reentrancy is tested in
925	 * system_reset_exception.
926	 */
927	lhz	r10,PACA_IN_NMI(r13)
928	addi	r10,r10,1
929	sth	r10,PACA_IN_NMI(r13)
930	li	r10,MSR_RI
931	mtmsrd 	r10,1
932
933	mr	r10,r1
934	ld	r1,PACA_NMI_EMERG_SP(r13)
935	subi	r1,r1,INT_FRAME_SIZE
936	__GEN_COMMON_BODY system_reset
937	/*
938	 * Set IRQS_ALL_DISABLED unconditionally so irqs_disabled() does
939	 * the right thing. We do not want to reconcile because that goes
940	 * through irq tracing which we don't want in NMI.
941	 *
942	 * Save PACAIRQHAPPENED to _DAR (otherwise unused), and set HARD_DIS
943	 * as we are running with MSR[EE]=0.
944	 */
945	li	r10,IRQS_ALL_DISABLED
946	stb	r10,PACAIRQSOFTMASK(r13)
947	lbz	r10,PACAIRQHAPPENED(r13)
948	std	r10,_DAR(r1)
949	ori	r10,r10,PACA_IRQ_HARD_DIS
950	stb	r10,PACAIRQHAPPENED(r13)
951
952	addi	r3,r1,STACK_FRAME_OVERHEAD
953	bl	system_reset_exception
954
955	/* Clear MSR_RI before setting SRR0 and SRR1. */
956	li	r9,0
957	mtmsrd	r9,1
958
959	/*
960	 * MSR_RI is clear, now we can decrement paca->in_nmi.
961	 */
962	lhz	r10,PACA_IN_NMI(r13)
963	subi	r10,r10,1
964	sth	r10,PACA_IN_NMI(r13)
965
966	/*
967	 * Restore soft mask settings.
968	 */
969	ld	r10,_DAR(r1)
970	stb	r10,PACAIRQHAPPENED(r13)
971	ld	r10,SOFTE(r1)
972	stb	r10,PACAIRQSOFTMASK(r13)
973
974	EXCEPTION_RESTORE_REGS
975	RFI_TO_USER_OR_KERNEL
976
977	GEN_KVM system_reset
978
979
980/**
981 * Interrupt 0x200 - Machine Check Interrupt (MCE).
982 * This is a non-maskable interrupt always taken in real-mode. It can be
983 * synchronous or asynchronous, caused by hardware or software, and it may be
984 * taken in a power-saving state.
985 *
986 * Handling:
987 * Similarly to system reset, this uses its own stack and PACA save area,
988 * the difference is re-entrancy is allowed on the machine check stack.
989 *
990 * machine_check_early is run in real mode, and carefully decodes the
991 * machine check and tries to handle it (e.g., flush the SLB if there was an
992 * error detected there), determines if it was recoverable and logs the
993 * event.
994 *
995 * This early code does not "reconcile" irq soft-mask state like SRESET or
996 * regular interrupts do, so irqs_disabled() among other things may not work
997 * properly (irq disable/enable already doesn't work because irq tracing can
998 * not work in real mode).
999 *
1000 * Then, depending on the execution context when the interrupt is taken, there
1001 * are 3 main actions:
1002 * - Executing in kernel mode. The event is queued with irq_work, which means
1003 *   it is handled when it is next safe to do so (i.e., the kernel has enabled
1004 *   interrupts), which could be immediately when the interrupt returns. This
1005 *   avoids nasty issues like switching to virtual mode when the MMU is in a
1006 *   bad state, or when executing OPAL code. (SRESET is exposed to such issues,
1007 *   but it has different priorities). Check to see if the CPU was in power
1008 *   save, and return via the wake up code if it was.
1009 *
1010 * - Executing in user mode. machine_check_exception is run like a normal
1011 *   interrupt handler, which processes the data generated by the early handler.
1012 *
1013 * - Executing in guest mode. The interrupt is run with its KVM test, and
1014 *   branches to KVM to deal with. KVM may queue the event for the host
1015 *   to report later.
1016 *
1017 * This interrupt is not maskable, so if it triggers when MSR[RI] is clear,
1018 * or SCRATCH0 is in use, it may cause a crash.
1019 *
1020 * KVM:
1021 * See SRESET.
1022 */
1023INT_DEFINE_BEGIN(machine_check_early)
1024	IVEC=0x200
1025	IAREA=PACA_EXMC
1026	IVIRT=0 /* no virt entry point */
1027	IREALMODE_COMMON=1
1028	/*
1029	 * MSR_RI is not enabled, because PACA_EXMC is being used, so a
1030	 * nested machine check corrupts it. machine_check_common enables
1031	 * MSR_RI.
1032	 */
1033	ISET_RI=0
1034	ISTACK=0
1035	IDAR=1
1036	IDSISR=1
1037	IRECONCILE=0
1038	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
1039INT_DEFINE_END(machine_check_early)
1040
1041INT_DEFINE_BEGIN(machine_check)
1042	IVEC=0x200
1043	IAREA=PACA_EXMC
1044	IVIRT=0 /* no virt entry point */
1045	ISET_RI=0
1046	IDAR=1
1047	IDSISR=1
1048	IKVM_SKIP=1
1049	IKVM_REAL=1
1050INT_DEFINE_END(machine_check)
1051
1052EXC_REAL_BEGIN(machine_check, 0x200, 0x100)
1053	GEN_INT_ENTRY machine_check_early, virt=0
1054EXC_REAL_END(machine_check, 0x200, 0x100)
1055EXC_VIRT_NONE(0x4200, 0x100)
1056
1057#ifdef CONFIG_PPC_PSERIES
1058TRAMP_REAL_BEGIN(machine_check_fwnmi)
1059	/* See comment at machine_check exception, don't turn on RI */
1060	GEN_INT_ENTRY machine_check_early, virt=0
1061#endif
1062
1063#define MACHINE_CHECK_HANDLER_WINDUP			\
1064	/* Clear MSR_RI before setting SRR0 and SRR1. */\
1065	li	r9,0;					\
1066	mtmsrd	r9,1;		/* Clear MSR_RI */	\
1067	/* Decrement paca->in_mce now RI is clear. */	\
1068	lhz	r12,PACA_IN_MCE(r13);			\
1069	subi	r12,r12,1;				\
1070	sth	r12,PACA_IN_MCE(r13);			\
1071	EXCEPTION_RESTORE_REGS
1072
1073EXC_COMMON_BEGIN(machine_check_early_common)
1074	__GEN_REALMODE_COMMON_ENTRY machine_check_early
1075
1076	/*
1077	 * Switch to mc_emergency stack and handle re-entrancy (we limit
1078	 * the nested MCE upto level 4 to avoid stack overflow).
1079	 * Save MCE registers srr1, srr0, dar and dsisr and then set ME=1
1080	 *
1081	 * We use paca->in_mce to check whether this is the first entry or
1082	 * nested machine check. We increment paca->in_mce to track nested
1083	 * machine checks.
1084	 *
1085	 * If this is the first entry then set stack pointer to
1086	 * paca->mc_emergency_sp, otherwise r1 is already pointing to
1087	 * stack frame on mc_emergency stack.
1088	 *
1089	 * NOTE: We are here with MSR_ME=0 (off), which means we risk a
1090	 * checkstop if we get another machine check exception before we do
1091	 * rfid with MSR_ME=1.
1092	 *
1093	 * This interrupt can wake directly from idle. If that is the case,
1094	 * the machine check is handled then the idle wakeup code is called
1095	 * to restore state.
1096	 */
1097	lhz	r10,PACA_IN_MCE(r13)
1098	cmpwi	r10,0			/* Are we in nested machine check */
1099	cmpwi	cr1,r10,MAX_MCE_DEPTH	/* Are we at maximum nesting */
1100	addi	r10,r10,1		/* increment paca->in_mce */
1101	sth	r10,PACA_IN_MCE(r13)
1102
1103	mr	r10,r1			/* Save r1 */
1104	bne	1f
1105	/* First machine check entry */
1106	ld	r1,PACAMCEMERGSP(r13)	/* Use MC emergency stack */
11071:	/* Limit nested MCE to level 4 to avoid stack overflow */
1108	bgt	cr1,unrecoverable_mce	/* Check if we hit limit of 4 */
1109	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame */
1110
1111	__GEN_COMMON_BODY machine_check_early
1112
1113BEGIN_FTR_SECTION
1114	bl	enable_machine_check
1115END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1116	li	r10,MSR_RI
1117	mtmsrd	r10,1
1118
1119	addi	r3,r1,STACK_FRAME_OVERHEAD
1120	bl	machine_check_early
1121	std	r3,RESULT(r1)	/* Save result */
1122	ld	r12,_MSR(r1)
1123
1124#ifdef CONFIG_PPC_P7_NAP
1125	/*
1126	 * Check if thread was in power saving mode. We come here when any
1127	 * of the following is true:
1128	 * a. thread wasn't in power saving mode
1129	 * b. thread was in power saving mode with no state loss,
1130	 *    supervisor state loss or hypervisor state loss.
1131	 *
1132	 * Go back to nap/sleep/winkle mode again if (b) is true.
1133	 */
1134BEGIN_FTR_SECTION
1135	rlwinm.	r11,r12,47-31,30,31
1136	bne	machine_check_idle_common
1137END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
1138#endif
1139
1140#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1141	/*
1142	 * Check if we are coming from guest. If yes, then run the normal
1143	 * exception handler which will take the
1144	 * machine_check_kvm->kvmppc_interrupt branch to deliver the MC event
1145	 * to guest.
1146	 */
1147	lbz	r11,HSTATE_IN_GUEST(r13)
1148	cmpwi	r11,0			/* Check if coming from guest */
1149	bne	mce_deliver		/* continue if we are. */
1150#endif
1151
1152	/*
1153	 * Check if we are coming from userspace. If yes, then run the normal
1154	 * exception handler which will deliver the MC event to this kernel.
1155	 */
1156	andi.	r11,r12,MSR_PR		/* See if coming from user. */
1157	bne	mce_deliver		/* continue in V mode if we are. */
1158
1159	/*
1160	 * At this point we are coming from kernel context.
1161	 * Queue up the MCE event and return from the interrupt.
1162	 * But before that, check if this is an un-recoverable exception.
1163	 * If yes, then stay on emergency stack and panic.
1164	 */
1165	andi.	r11,r12,MSR_RI
1166	beq	unrecoverable_mce
1167
1168	/*
1169	 * Check if we have successfully handled/recovered from error, if not
1170	 * then stay on emergency stack and panic.
1171	 */
1172	ld	r3,RESULT(r1)	/* Load result */
1173	cmpdi	r3,0		/* see if we handled MCE successfully */
1174	beq	unrecoverable_mce /* if !handled then panic */
1175
1176	/*
1177	 * Return from MC interrupt.
1178	 * Queue up the MCE event so that we can log it later, while
1179	 * returning from kernel or opal call.
1180	 */
1181	bl	machine_check_queue_event
1182	MACHINE_CHECK_HANDLER_WINDUP
1183	RFI_TO_KERNEL
1184
1185mce_deliver:
1186	/*
1187	 * This is a host user or guest MCE. Restore all registers, then
1188	 * run the "late" handler. For host user, this will run the
1189	 * machine_check_exception handler in virtual mode like a normal
1190	 * interrupt handler. For guest, this will trigger the KVM test
1191	 * and branch to the KVM interrupt similarly to other interrupts.
1192	 */
1193BEGIN_FTR_SECTION
1194	ld	r10,ORIG_GPR3(r1)
1195	mtspr	SPRN_CFAR,r10
1196END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
1197	MACHINE_CHECK_HANDLER_WINDUP
1198	GEN_INT_ENTRY machine_check, virt=0
1199
1200EXC_COMMON_BEGIN(machine_check_common)
1201	/*
1202	 * Machine check is different because we use a different
1203	 * save area: PACA_EXMC instead of PACA_EXGEN.
1204	 */
1205	GEN_COMMON machine_check
1206
1207	FINISH_NAP
1208	/* Enable MSR_RI when finished with PACA_EXMC */
1209	li	r10,MSR_RI
1210	mtmsrd 	r10,1
1211	addi	r3,r1,STACK_FRAME_OVERHEAD
1212	bl	machine_check_exception
1213	b	interrupt_return
1214
1215	GEN_KVM machine_check
1216
1217
1218#ifdef CONFIG_PPC_P7_NAP
1219/*
1220 * This is an idle wakeup. Low level machine check has already been
1221 * done. Queue the event then call the idle code to do the wake up.
1222 */
1223EXC_COMMON_BEGIN(machine_check_idle_common)
1224	bl	machine_check_queue_event
1225
1226	/*
1227	 * We have not used any non-volatile GPRs here, and as a rule
1228	 * most exception code including machine check does not.
1229	 * Therefore PACA_NAPSTATELOST does not need to be set. Idle
1230	 * wakeup will restore volatile registers.
1231	 *
1232	 * Load the original SRR1 into r3 for pnv_powersave_wakeup_mce.
1233	 *
1234	 * Then decrement MCE nesting after finishing with the stack.
1235	 */
1236	ld	r3,_MSR(r1)
1237	ld	r4,_LINK(r1)
1238
1239	lhz	r11,PACA_IN_MCE(r13)
1240	subi	r11,r11,1
1241	sth	r11,PACA_IN_MCE(r13)
1242
1243	mtlr	r4
1244	rlwinm	r10,r3,47-31,30,31
1245	cmpwi	cr1,r10,2
1246	bltlr	cr1	/* no state loss, return to idle caller */
1247	b	idle_return_gpr_loss
1248#endif
1249
1250EXC_COMMON_BEGIN(unrecoverable_mce)
1251	/*
1252	 * We are going down. But there are chances that we might get hit by
1253	 * another MCE during panic path and we may run into unstable state
1254	 * with no way out. Hence, turn ME bit off while going down, so that
1255	 * when another MCE is hit during panic path, system will checkstop
1256	 * and hypervisor will get restarted cleanly by SP.
1257	 */
1258BEGIN_FTR_SECTION
1259	li	r10,0 /* clear MSR_RI */
1260	mtmsrd	r10,1
1261	bl	disable_machine_check
1262END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1263	ld	r10,PACAKMSR(r13)
1264	li	r3,MSR_ME
1265	andc	r10,r10,r3
1266	mtmsrd	r10
1267
1268	/* Invoke machine_check_exception to print MCE event and panic. */
1269	addi	r3,r1,STACK_FRAME_OVERHEAD
1270	bl	machine_check_exception
1271
1272	/*
1273	 * We will not reach here. Even if we did, there is no way out.
1274	 * Call unrecoverable_exception and die.
1275	 */
1276	addi	r3,r1,STACK_FRAME_OVERHEAD
1277	bl	unrecoverable_exception
1278	b	.
1279
1280
1281/**
1282 * Interrupt 0x300 - Data Storage Interrupt (DSI).
1283 * This is a synchronous interrupt generated due to a data access exception,
1284 * e.g., a load orstore which does not have a valid page table entry with
1285 * permissions. DAWR matches also fault here, as do RC updates, and minor misc
1286 * errors e.g., copy/paste, AMO, certain invalid CI accesses, etc.
1287 *
1288 * Handling:
1289 * - Hash MMU
1290 *   Go to do_hash_page first to see if the HPT can be filled from an entry in
1291 *   the Linux page table. Hash faults can hit in kernel mode in a fairly
1292 *   arbitrary state (e.g., interrupts disabled, locks held) when accessing
1293 *   "non-bolted" regions, e.g., vmalloc space. However these should always be
1294 *   backed by Linux page tables.
1295 *
1296 *   If none is found, do a Linux page fault. Linux page faults can happen in
1297 *   kernel mode due to user copy operations of course.
1298 *
1299 * - Radix MMU
1300 *   The hardware loads from the Linux page table directly, so a fault goes
1301 *   immediately to Linux page fault.
1302 *
1303 * Conditions like DAWR match are handled on the way in to Linux page fault.
1304 */
1305INT_DEFINE_BEGIN(data_access)
1306	IVEC=0x300
1307	IDAR=1
1308	IDSISR=1
1309#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1310	IKVM_SKIP=1
1311	IKVM_REAL=1
1312#endif
1313INT_DEFINE_END(data_access)
1314
1315EXC_REAL_BEGIN(data_access, 0x300, 0x80)
1316	GEN_INT_ENTRY data_access, virt=0
1317EXC_REAL_END(data_access, 0x300, 0x80)
1318EXC_VIRT_BEGIN(data_access, 0x4300, 0x80)
1319	GEN_INT_ENTRY data_access, virt=1
1320EXC_VIRT_END(data_access, 0x4300, 0x80)
1321EXC_COMMON_BEGIN(data_access_common)
1322	GEN_COMMON data_access
1323	ld	r4,_DAR(r1)
1324	ld	r5,_DSISR(r1)
1325BEGIN_MMU_FTR_SECTION
1326	ld	r6,_MSR(r1)
1327	li	r3,0x300
1328	b	do_hash_page		/* Try to handle as hpte fault */
1329MMU_FTR_SECTION_ELSE
1330	b	handle_page_fault
1331ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1332
1333	GEN_KVM data_access
1334
1335
1336/**
1337 * Interrupt 0x380 - Data Segment Interrupt (DSLB).
1338 * This is a synchronous interrupt in response to an MMU fault missing SLB
1339 * entry for HPT, or an address outside RPT translation range.
1340 *
1341 * Handling:
1342 * - HPT:
1343 *   This refills the SLB, or reports an access fault similarly to a bad page
1344 *   fault. When coming from user-mode, the SLB handler may access any kernel
1345 *   data, though it may itself take a DSLB. When coming from kernel mode,
1346 *   recursive faults must be avoided so access is restricted to the kernel
1347 *   image text/data, kernel stack, and any data allocated below
1348 *   ppc64_bolted_size (first segment). The kernel handler must avoid stomping
1349 *   on user-handler data structures.
1350 *
1351 * A dedicated save area EXSLB is used (XXX: but it actually need not be
1352 * these days, we could use EXGEN).
1353 */
1354INT_DEFINE_BEGIN(data_access_slb)
1355	IVEC=0x380
1356	IAREA=PACA_EXSLB
1357	IRECONCILE=0
1358	IDAR=1
1359#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1360	IKVM_SKIP=1
1361	IKVM_REAL=1
1362#endif
1363INT_DEFINE_END(data_access_slb)
1364
1365EXC_REAL_BEGIN(data_access_slb, 0x380, 0x80)
1366	GEN_INT_ENTRY data_access_slb, virt=0
1367EXC_REAL_END(data_access_slb, 0x380, 0x80)
1368EXC_VIRT_BEGIN(data_access_slb, 0x4380, 0x80)
1369	GEN_INT_ENTRY data_access_slb, virt=1
1370EXC_VIRT_END(data_access_slb, 0x4380, 0x80)
1371EXC_COMMON_BEGIN(data_access_slb_common)
1372	GEN_COMMON data_access_slb
1373	ld	r4,_DAR(r1)
1374	addi	r3,r1,STACK_FRAME_OVERHEAD
1375BEGIN_MMU_FTR_SECTION
1376	/* HPT case, do SLB fault */
1377	bl	do_slb_fault
1378	cmpdi	r3,0
1379	bne-	1f
1380	b	fast_interrupt_return
13811:	/* Error case */
1382MMU_FTR_SECTION_ELSE
1383	/* Radix case, access is outside page table range */
1384	li	r3,-EFAULT
1385ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1386	std	r3,RESULT(r1)
1387	RECONCILE_IRQ_STATE(r10, r11)
1388	ld	r4,_DAR(r1)
1389	ld	r5,RESULT(r1)
1390	addi	r3,r1,STACK_FRAME_OVERHEAD
1391	bl	do_bad_slb_fault
1392	b	interrupt_return
1393
1394	GEN_KVM data_access_slb
1395
1396
1397/**
1398 * Interrupt 0x400 - Instruction Storage Interrupt (ISI).
1399 * This is a synchronous interrupt in response to an MMU fault due to an
1400 * instruction fetch.
1401 *
1402 * Handling:
1403 * Similar to DSI, though in response to fetch. The faulting address is found
1404 * in SRR0 (rather than DAR), and status in SRR1 (rather than DSISR).
1405 */
1406INT_DEFINE_BEGIN(instruction_access)
1407	IVEC=0x400
1408	IISIDE=1
1409	IDAR=1
1410	IDSISR=1
1411#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1412	IKVM_REAL=1
1413#endif
1414INT_DEFINE_END(instruction_access)
1415
1416EXC_REAL_BEGIN(instruction_access, 0x400, 0x80)
1417	GEN_INT_ENTRY instruction_access, virt=0
1418EXC_REAL_END(instruction_access, 0x400, 0x80)
1419EXC_VIRT_BEGIN(instruction_access, 0x4400, 0x80)
1420	GEN_INT_ENTRY instruction_access, virt=1
1421EXC_VIRT_END(instruction_access, 0x4400, 0x80)
1422EXC_COMMON_BEGIN(instruction_access_common)
1423	GEN_COMMON instruction_access
1424	ld	r4,_DAR(r1)
1425	ld	r5,_DSISR(r1)
1426BEGIN_MMU_FTR_SECTION
1427	ld      r6,_MSR(r1)
1428	li	r3,0x400
1429	b	do_hash_page		/* Try to handle as hpte fault */
1430MMU_FTR_SECTION_ELSE
1431	b	handle_page_fault
1432ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1433
1434	GEN_KVM instruction_access
1435
1436
1437/**
1438 * Interrupt 0x480 - Instruction Segment Interrupt (ISLB).
1439 * This is a synchronous interrupt in response to an MMU fault due to an
1440 * instruction fetch.
1441 *
1442 * Handling:
1443 * Similar to DSLB, though in response to fetch. The faulting address is found
1444 * in SRR0 (rather than DAR).
1445 */
1446INT_DEFINE_BEGIN(instruction_access_slb)
1447	IVEC=0x480
1448	IAREA=PACA_EXSLB
1449	IRECONCILE=0
1450	IISIDE=1
1451	IDAR=1
1452#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1453	IKVM_REAL=1
1454#endif
1455INT_DEFINE_END(instruction_access_slb)
1456
1457EXC_REAL_BEGIN(instruction_access_slb, 0x480, 0x80)
1458	GEN_INT_ENTRY instruction_access_slb, virt=0
1459EXC_REAL_END(instruction_access_slb, 0x480, 0x80)
1460EXC_VIRT_BEGIN(instruction_access_slb, 0x4480, 0x80)
1461	GEN_INT_ENTRY instruction_access_slb, virt=1
1462EXC_VIRT_END(instruction_access_slb, 0x4480, 0x80)
1463EXC_COMMON_BEGIN(instruction_access_slb_common)
1464	GEN_COMMON instruction_access_slb
1465	ld	r4,_DAR(r1)
1466	addi	r3,r1,STACK_FRAME_OVERHEAD
1467BEGIN_MMU_FTR_SECTION
1468	/* HPT case, do SLB fault */
1469	bl	do_slb_fault
1470	cmpdi	r3,0
1471	bne-	1f
1472	b	fast_interrupt_return
14731:	/* Error case */
1474MMU_FTR_SECTION_ELSE
1475	/* Radix case, access is outside page table range */
1476	li	r3,-EFAULT
1477ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1478	std	r3,RESULT(r1)
1479	RECONCILE_IRQ_STATE(r10, r11)
1480	ld	r4,_DAR(r1)
1481	ld	r5,RESULT(r1)
1482	addi	r3,r1,STACK_FRAME_OVERHEAD
1483	bl	do_bad_slb_fault
1484	b	interrupt_return
1485
1486	GEN_KVM instruction_access_slb
1487
1488
1489/**
1490 * Interrupt 0x500 - External Interrupt.
1491 * This is an asynchronous maskable interrupt in response to an "external
1492 * exception" from the interrupt controller or hypervisor (e.g., device
1493 * interrupt). It is maskable in hardware by clearing MSR[EE], and
1494 * soft-maskable with IRQS_DISABLED mask (i.e., local_irq_disable()).
1495 *
1496 * When running in HV mode, Linux sets up the LPCR[LPES] bit such that
1497 * interrupts are delivered with HSRR registers, guests use SRRs, which
1498 * reqiures IHSRR_IF_HVMODE.
1499 *
1500 * On bare metal POWER9 and later, Linux sets the LPCR[HVICE] bit such that
1501 * external interrupts are delivered as Hypervisor Virtualization Interrupts
1502 * rather than External Interrupts.
1503 *
1504 * Handling:
1505 * This calls into Linux IRQ handler. NVGPRs are not saved to reduce overhead,
1506 * because registers at the time of the interrupt are not so important as it is
1507 * asynchronous.
1508 *
1509 * If soft masked, the masked handler will note the pending interrupt for
1510 * replay, and clear MSR[EE] in the interrupted context.
1511 */
1512INT_DEFINE_BEGIN(hardware_interrupt)
1513	IVEC=0x500
1514	IHSRR_IF_HVMODE=1
1515	IMASK=IRQS_DISABLED
1516	IKVM_REAL=1
1517	IKVM_VIRT=1
1518INT_DEFINE_END(hardware_interrupt)
1519
1520EXC_REAL_BEGIN(hardware_interrupt, 0x500, 0x100)
1521	GEN_INT_ENTRY hardware_interrupt, virt=0
1522EXC_REAL_END(hardware_interrupt, 0x500, 0x100)
1523EXC_VIRT_BEGIN(hardware_interrupt, 0x4500, 0x100)
1524	GEN_INT_ENTRY hardware_interrupt, virt=1
1525EXC_VIRT_END(hardware_interrupt, 0x4500, 0x100)
1526EXC_COMMON_BEGIN(hardware_interrupt_common)
1527	GEN_COMMON hardware_interrupt
1528	FINISH_NAP
1529	RUNLATCH_ON
1530	addi	r3,r1,STACK_FRAME_OVERHEAD
1531	bl	do_IRQ
1532	b	interrupt_return
1533
1534	GEN_KVM hardware_interrupt
1535
1536
1537/**
1538 * Interrupt 0x600 - Alignment Interrupt
1539 * This is a synchronous interrupt in response to data alignment fault.
1540 */
1541INT_DEFINE_BEGIN(alignment)
1542	IVEC=0x600
1543	IDAR=1
1544	IDSISR=1
1545#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1546	IKVM_REAL=1
1547#endif
1548INT_DEFINE_END(alignment)
1549
1550EXC_REAL_BEGIN(alignment, 0x600, 0x100)
1551	GEN_INT_ENTRY alignment, virt=0
1552EXC_REAL_END(alignment, 0x600, 0x100)
1553EXC_VIRT_BEGIN(alignment, 0x4600, 0x100)
1554	GEN_INT_ENTRY alignment, virt=1
1555EXC_VIRT_END(alignment, 0x4600, 0x100)
1556EXC_COMMON_BEGIN(alignment_common)
1557	GEN_COMMON alignment
1558	addi	r3,r1,STACK_FRAME_OVERHEAD
1559	bl	alignment_exception
1560	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
1561	b	interrupt_return
1562
1563	GEN_KVM alignment
1564
1565
1566/**
1567 * Interrupt 0x700 - Program Interrupt (program check).
1568 * This is a synchronous interrupt in response to various instruction faults:
1569 * traps, privilege errors, TM errors, floating point exceptions.
1570 *
1571 * Handling:
1572 * This interrupt may use the "emergency stack" in some cases when being taken
1573 * from kernel context, which complicates handling.
1574 */
1575INT_DEFINE_BEGIN(program_check)
1576	IVEC=0x700
1577#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1578	IKVM_REAL=1
1579#endif
1580INT_DEFINE_END(program_check)
1581
1582EXC_REAL_BEGIN(program_check, 0x700, 0x100)
1583	GEN_INT_ENTRY program_check, virt=0
1584EXC_REAL_END(program_check, 0x700, 0x100)
1585EXC_VIRT_BEGIN(program_check, 0x4700, 0x100)
1586	GEN_INT_ENTRY program_check, virt=1
1587EXC_VIRT_END(program_check, 0x4700, 0x100)
1588EXC_COMMON_BEGIN(program_check_common)
1589	__GEN_COMMON_ENTRY program_check
1590
1591	/*
1592	 * It's possible to receive a TM Bad Thing type program check with
1593	 * userspace register values (in particular r1), but with SRR1 reporting
1594	 * that we came from the kernel. Normally that would confuse the bad
1595	 * stack logic, and we would report a bad kernel stack pointer. Instead
1596	 * we switch to the emergency stack if we're taking a TM Bad Thing from
1597	 * the kernel.
1598	 */
1599
1600	andi.	r10,r12,MSR_PR
1601	bne	2f			/* If userspace, go normal path */
1602
1603	andis.	r10,r12,(SRR1_PROGTM)@h
1604	bne	1f			/* If TM, emergency		*/
1605
1606	cmpdi	r1,-INT_FRAME_SIZE	/* check if r1 is in userspace	*/
1607	blt	2f			/* normal path if not		*/
1608
1609	/* Use the emergency stack					*/
16101:	andi.	r10,r12,MSR_PR		/* Set CR0 correctly for label	*/
1611					/* 3 in EXCEPTION_PROLOG_COMMON	*/
1612	mr	r10,r1			/* Save r1			*/
1613	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack		*/
1614	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
1615	__ISTACK(program_check)=0
1616	__GEN_COMMON_BODY program_check
1617	b 3f
16182:
1619	__ISTACK(program_check)=1
1620	__GEN_COMMON_BODY program_check
16213:
1622	addi	r3,r1,STACK_FRAME_OVERHEAD
1623	bl	program_check_exception
1624	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
1625	b	interrupt_return
1626
1627	GEN_KVM program_check
1628
1629
1630/*
1631 * Interrupt 0x800 - Floating-Point Unavailable Interrupt.
1632 * This is a synchronous interrupt in response to executing an fp instruction
1633 * with MSR[FP]=0.
1634 *
1635 * Handling:
1636 * This will load FP registers and enable the FP bit if coming from userspace,
1637 * otherwise report a bad kernel use of FP.
1638 */
1639INT_DEFINE_BEGIN(fp_unavailable)
1640	IVEC=0x800
1641	IRECONCILE=0
1642#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1643	IKVM_REAL=1
1644#endif
1645INT_DEFINE_END(fp_unavailable)
1646
1647EXC_REAL_BEGIN(fp_unavailable, 0x800, 0x100)
1648	GEN_INT_ENTRY fp_unavailable, virt=0
1649EXC_REAL_END(fp_unavailable, 0x800, 0x100)
1650EXC_VIRT_BEGIN(fp_unavailable, 0x4800, 0x100)
1651	GEN_INT_ENTRY fp_unavailable, virt=1
1652EXC_VIRT_END(fp_unavailable, 0x4800, 0x100)
1653EXC_COMMON_BEGIN(fp_unavailable_common)
1654	GEN_COMMON fp_unavailable
1655	bne	1f			/* if from user, just load it up */
1656	RECONCILE_IRQ_STATE(r10, r11)
1657	addi	r3,r1,STACK_FRAME_OVERHEAD
1658	bl	kernel_fp_unavailable_exception
16590:	trap
1660	EMIT_BUG_ENTRY 0b, __FILE__, __LINE__, 0
16611:
1662#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1663BEGIN_FTR_SECTION
1664	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
1665	 * transaction), go do TM stuff
1666	 */
1667	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
1668	bne-	2f
1669END_FTR_SECTION_IFSET(CPU_FTR_TM)
1670#endif
1671	bl	load_up_fpu
1672	b	fast_interrupt_return
1673#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
16742:	/* User process was in a transaction */
1675	RECONCILE_IRQ_STATE(r10, r11)
1676	addi	r3,r1,STACK_FRAME_OVERHEAD
1677	bl	fp_unavailable_tm
1678	b	interrupt_return
1679#endif
1680
1681	GEN_KVM fp_unavailable
1682
1683
1684/**
1685 * Interrupt 0x900 - Decrementer Interrupt.
1686 * This is an asynchronous interrupt in response to a decrementer exception
1687 * (e.g., DEC has wrapped below zero). It is maskable in hardware by clearing
1688 * MSR[EE], and soft-maskable with IRQS_DISABLED mask (i.e.,
1689 * local_irq_disable()).
1690 *
1691 * Handling:
1692 * This calls into Linux timer handler. NVGPRs are not saved (see 0x500).
1693 *
1694 * If soft masked, the masked handler will note the pending interrupt for
1695 * replay, and bump the decrementer to a high value, leaving MSR[EE] enabled
1696 * in the interrupted context.
1697 * If PPC_WATCHDOG is configured, the soft masked handler will actually set
1698 * things back up to run soft_nmi_interrupt as a regular interrupt handler
1699 * on the emergency stack.
1700 */
1701INT_DEFINE_BEGIN(decrementer)
1702	IVEC=0x900
1703	IMASK=IRQS_DISABLED
1704#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1705	IKVM_REAL=1
1706#endif
1707INT_DEFINE_END(decrementer)
1708
1709EXC_REAL_BEGIN(decrementer, 0x900, 0x80)
1710	GEN_INT_ENTRY decrementer, virt=0
1711EXC_REAL_END(decrementer, 0x900, 0x80)
1712EXC_VIRT_BEGIN(decrementer, 0x4900, 0x80)
1713	GEN_INT_ENTRY decrementer, virt=1
1714EXC_VIRT_END(decrementer, 0x4900, 0x80)
1715EXC_COMMON_BEGIN(decrementer_common)
1716	GEN_COMMON decrementer
1717	FINISH_NAP
1718	RUNLATCH_ON
1719	addi	r3,r1,STACK_FRAME_OVERHEAD
1720	bl	timer_interrupt
1721	b	interrupt_return
1722
1723	GEN_KVM decrementer
1724
1725
1726/**
1727 * Interrupt 0x980 - Hypervisor Decrementer Interrupt.
1728 * This is an asynchronous interrupt, similar to 0x900 but for the HDEC
1729 * register.
1730 *
1731 * Handling:
1732 * Linux does not use this outside KVM where it's used to keep a host timer
1733 * while the guest is given control of DEC. It should normally be caught by
1734 * the KVM test and routed there.
1735 */
1736INT_DEFINE_BEGIN(hdecrementer)
1737	IVEC=0x980
1738	IHSRR=1
1739	ISTACK=0
1740	IRECONCILE=0
1741	IKVM_REAL=1
1742	IKVM_VIRT=1
1743INT_DEFINE_END(hdecrementer)
1744
1745EXC_REAL_BEGIN(hdecrementer, 0x980, 0x80)
1746	GEN_INT_ENTRY hdecrementer, virt=0
1747EXC_REAL_END(hdecrementer, 0x980, 0x80)
1748EXC_VIRT_BEGIN(hdecrementer, 0x4980, 0x80)
1749	GEN_INT_ENTRY hdecrementer, virt=1
1750EXC_VIRT_END(hdecrementer, 0x4980, 0x80)
1751EXC_COMMON_BEGIN(hdecrementer_common)
1752	__GEN_COMMON_ENTRY hdecrementer
1753	/*
1754	 * Hypervisor decrementer interrupts not caught by the KVM test
1755	 * shouldn't occur but are sometimes left pending on exit from a KVM
1756	 * guest.  We don't need to do anything to clear them, as they are
1757	 * edge-triggered.
1758	 *
1759	 * Be careful to avoid touching the kernel stack.
1760	 */
1761	ld	r10,PACA_EXGEN+EX_CTR(r13)
1762	mtctr	r10
1763	mtcrf	0x80,r9
1764	ld	r9,PACA_EXGEN+EX_R9(r13)
1765	ld	r10,PACA_EXGEN+EX_R10(r13)
1766	ld	r11,PACA_EXGEN+EX_R11(r13)
1767	ld	r12,PACA_EXGEN+EX_R12(r13)
1768	ld	r13,PACA_EXGEN+EX_R13(r13)
1769	HRFI_TO_KERNEL
1770
1771	GEN_KVM hdecrementer
1772
1773
1774/**
1775 * Interrupt 0xa00 - Directed Privileged Doorbell Interrupt.
1776 * This is an asynchronous interrupt in response to a msgsndp doorbell.
1777 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
1778 * IRQS_DISABLED mask (i.e., local_irq_disable()).
1779 *
1780 * Handling:
1781 * Guests may use this for IPIs between threads in a core if the
1782 * hypervisor supports it. NVGPRS are not saved (see 0x500).
1783 *
1784 * If soft masked, the masked handler will note the pending interrupt for
1785 * replay, leaving MSR[EE] enabled in the interrupted context because the
1786 * doorbells are edge triggered.
1787 */
1788INT_DEFINE_BEGIN(doorbell_super)
1789	IVEC=0xa00
1790	IMASK=IRQS_DISABLED
1791#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1792	IKVM_REAL=1
1793#endif
1794INT_DEFINE_END(doorbell_super)
1795
1796EXC_REAL_BEGIN(doorbell_super, 0xa00, 0x100)
1797	GEN_INT_ENTRY doorbell_super, virt=0
1798EXC_REAL_END(doorbell_super, 0xa00, 0x100)
1799EXC_VIRT_BEGIN(doorbell_super, 0x4a00, 0x100)
1800	GEN_INT_ENTRY doorbell_super, virt=1
1801EXC_VIRT_END(doorbell_super, 0x4a00, 0x100)
1802EXC_COMMON_BEGIN(doorbell_super_common)
1803	GEN_COMMON doorbell_super
1804	FINISH_NAP
1805	RUNLATCH_ON
1806	addi	r3,r1,STACK_FRAME_OVERHEAD
1807#ifdef CONFIG_PPC_DOORBELL
1808	bl	doorbell_exception
1809#else
1810	bl	unknown_exception
1811#endif
1812	b	interrupt_return
1813
1814	GEN_KVM doorbell_super
1815
1816
1817EXC_REAL_NONE(0xb00, 0x100)
1818EXC_VIRT_NONE(0x4b00, 0x100)
1819
1820/**
1821 * Interrupt 0xc00 - System Call Interrupt (syscall, hcall).
1822 * This is a synchronous interrupt invoked with the "sc" instruction. The
1823 * system call is invoked with "sc 0" and does not alter the HV bit, so it
1824 * is directed to the currently running OS. The hypercall is invoked with
1825 * "sc 1" and it sets HV=1, so it elevates to hypervisor.
1826 *
1827 * In HPT, sc 1 always goes to 0xc00 real mode. In RADIX, sc 1 can go to
1828 * 0x4c00 virtual mode.
1829 *
1830 * Handling:
1831 * If the KVM test fires then it was due to a hypercall and is accordingly
1832 * routed to KVM. Otherwise this executes a normal Linux system call.
1833 *
1834 * Call convention:
1835 *
1836 * syscall and hypercalls register conventions are documented in
1837 * Documentation/powerpc/syscall64-abi.rst and
1838 * Documentation/powerpc/papr_hcalls.rst respectively.
1839 *
1840 * The intersection of volatile registers that don't contain possible
1841 * inputs is: cr0, xer, ctr. We may use these as scratch regs upon entry
1842 * without saving, though xer is not a good idea to use, as hardware may
1843 * interpret some bits so it may be costly to change them.
1844 */
1845INT_DEFINE_BEGIN(system_call)
1846	IVEC=0xc00
1847	IKVM_REAL=1
1848	IKVM_VIRT=1
1849INT_DEFINE_END(system_call)
1850
1851.macro SYSTEM_CALL virt
1852#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1853	/*
1854	 * There is a little bit of juggling to get syscall and hcall
1855	 * working well. Save r13 in ctr to avoid using SPRG scratch
1856	 * register.
1857	 *
1858	 * Userspace syscalls have already saved the PPR, hcalls must save
1859	 * it before setting HMT_MEDIUM.
1860	 */
1861	mtctr	r13
1862	GET_PACA(r13)
1863	std	r10,PACA_EXGEN+EX_R10(r13)
1864	INTERRUPT_TO_KERNEL
1865	KVMTEST system_call /* uses r10, branch to system_call_kvm */
1866	mfctr	r9
1867#else
1868	mr	r9,r13
1869	GET_PACA(r13)
1870	INTERRUPT_TO_KERNEL
1871#endif
1872
1873#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
1874BEGIN_FTR_SECTION
1875	cmpdi	r0,0x1ebe
1876	beq-	1f
1877END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE)
1878#endif
1879
1880	/* We reach here with PACA in r13, r13 in r9. */
1881	mfspr	r11,SPRN_SRR0
1882	mfspr	r12,SPRN_SRR1
1883
1884	HMT_MEDIUM
1885
1886	.if ! \virt
1887	__LOAD_HANDLER(r10, system_call_common)
1888	mtspr	SPRN_SRR0,r10
1889	ld	r10,PACAKMSR(r13)
1890	mtspr	SPRN_SRR1,r10
1891	RFI_TO_KERNEL
1892	b	.	/* prevent speculative execution */
1893	.else
1894	li	r10,MSR_RI
1895	mtmsrd 	r10,1			/* Set RI (EE=0) */
1896#ifdef CONFIG_RELOCATABLE
1897	__LOAD_HANDLER(r10, system_call_common)
1898	mtctr	r10
1899	bctr
1900#else
1901	b	system_call_common
1902#endif
1903	.endif
1904
1905#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
1906	/* Fast LE/BE switch system call */
19071:	mfspr	r12,SPRN_SRR1
1908	xori	r12,r12,MSR_LE
1909	mtspr	SPRN_SRR1,r12
1910	mr	r13,r9
1911	RFI_TO_USER	/* return to userspace */
1912	b	.	/* prevent speculative execution */
1913#endif
1914.endm
1915
1916EXC_REAL_BEGIN(system_call, 0xc00, 0x100)
1917	SYSTEM_CALL 0
1918EXC_REAL_END(system_call, 0xc00, 0x100)
1919EXC_VIRT_BEGIN(system_call, 0x4c00, 0x100)
1920	SYSTEM_CALL 1
1921EXC_VIRT_END(system_call, 0x4c00, 0x100)
1922
1923#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1924TRAMP_REAL_BEGIN(system_call_kvm)
1925	/*
1926	 * This is a hcall, so register convention is as above, with these
1927	 * differences:
1928	 * r13 = PACA
1929	 * ctr = orig r13
1930	 * orig r10 saved in PACA
1931	 */
1932	 /*
1933	  * Save the PPR (on systems that support it) before changing to
1934	  * HMT_MEDIUM. That allows the KVM code to save that value into the
1935	  * guest state (it is the guest's PPR value).
1936	  */
1937BEGIN_FTR_SECTION
1938	mfspr	r10,SPRN_PPR
1939	std	r10,HSTATE_PPR(r13)
1940END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
1941	HMT_MEDIUM
1942	mfctr	r10
1943	SET_SCRATCH0(r10)
1944	mfcr	r10
1945	std	r12,HSTATE_SCRATCH0(r13)
1946	sldi	r12,r10,32
1947	ori	r12,r12,0xc00
1948#ifdef CONFIG_RELOCATABLE
1949	/*
1950	 * Requires __LOAD_FAR_HANDLER beause kvmppc_interrupt lives
1951	 * outside the head section.
1952	 */
1953	__LOAD_FAR_HANDLER(r10, kvmppc_interrupt)
1954	mtctr   r10
1955	ld	r10,PACA_EXGEN+EX_R10(r13)
1956	bctr
1957#else
1958	ld	r10,PACA_EXGEN+EX_R10(r13)
1959	b       kvmppc_interrupt
1960#endif
1961#endif
1962
1963
1964/**
1965 * Interrupt 0xd00 - Trace Interrupt.
1966 * This is a synchronous interrupt in response to instruction step or
1967 * breakpoint faults.
1968 */
1969INT_DEFINE_BEGIN(single_step)
1970	IVEC=0xd00
1971#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1972	IKVM_REAL=1
1973#endif
1974INT_DEFINE_END(single_step)
1975
1976EXC_REAL_BEGIN(single_step, 0xd00, 0x100)
1977	GEN_INT_ENTRY single_step, virt=0
1978EXC_REAL_END(single_step, 0xd00, 0x100)
1979EXC_VIRT_BEGIN(single_step, 0x4d00, 0x100)
1980	GEN_INT_ENTRY single_step, virt=1
1981EXC_VIRT_END(single_step, 0x4d00, 0x100)
1982EXC_COMMON_BEGIN(single_step_common)
1983	GEN_COMMON single_step
1984	addi	r3,r1,STACK_FRAME_OVERHEAD
1985	bl	single_step_exception
1986	b	interrupt_return
1987
1988	GEN_KVM single_step
1989
1990
1991/**
1992 * Interrupt 0xe00 - Hypervisor Data Storage Interrupt (HDSI).
1993 * This is a synchronous interrupt in response to an MMU fault caused by a
1994 * guest data access.
1995 *
1996 * Handling:
1997 * This should always get routed to KVM. In radix MMU mode, this is caused
1998 * by a guest nested radix access that can't be performed due to the
1999 * partition scope page table. In hash mode, this can be caused by guests
2000 * running with translation disabled (virtual real mode) or with VPM enabled.
2001 * KVM will update the page table structures or disallow the access.
2002 */
2003INT_DEFINE_BEGIN(h_data_storage)
2004	IVEC=0xe00
2005	IHSRR=1
2006	IDAR=1
2007	IDSISR=1
2008	IKVM_SKIP=1
2009	IKVM_REAL=1
2010	IKVM_VIRT=1
2011INT_DEFINE_END(h_data_storage)
2012
2013EXC_REAL_BEGIN(h_data_storage, 0xe00, 0x20)
2014	GEN_INT_ENTRY h_data_storage, virt=0, ool=1
2015EXC_REAL_END(h_data_storage, 0xe00, 0x20)
2016EXC_VIRT_BEGIN(h_data_storage, 0x4e00, 0x20)
2017	GEN_INT_ENTRY h_data_storage, virt=1, ool=1
2018EXC_VIRT_END(h_data_storage, 0x4e00, 0x20)
2019EXC_COMMON_BEGIN(h_data_storage_common)
2020	GEN_COMMON h_data_storage
2021	addi    r3,r1,STACK_FRAME_OVERHEAD
2022BEGIN_MMU_FTR_SECTION
2023	ld	r4,_DAR(r1)
2024	li	r5,SIGSEGV
2025	bl      bad_page_fault
2026MMU_FTR_SECTION_ELSE
2027	bl      unknown_exception
2028ALT_MMU_FTR_SECTION_END_IFSET(MMU_FTR_TYPE_RADIX)
2029	b       interrupt_return
2030
2031	GEN_KVM h_data_storage
2032
2033
2034/**
2035 * Interrupt 0xe20 - Hypervisor Instruction Storage Interrupt (HISI).
2036 * This is a synchronous interrupt in response to an MMU fault caused by a
2037 * guest instruction fetch, similar to HDSI.
2038 */
2039INT_DEFINE_BEGIN(h_instr_storage)
2040	IVEC=0xe20
2041	IHSRR=1
2042	IKVM_REAL=1
2043	IKVM_VIRT=1
2044INT_DEFINE_END(h_instr_storage)
2045
2046EXC_REAL_BEGIN(h_instr_storage, 0xe20, 0x20)
2047	GEN_INT_ENTRY h_instr_storage, virt=0, ool=1
2048EXC_REAL_END(h_instr_storage, 0xe20, 0x20)
2049EXC_VIRT_BEGIN(h_instr_storage, 0x4e20, 0x20)
2050	GEN_INT_ENTRY h_instr_storage, virt=1, ool=1
2051EXC_VIRT_END(h_instr_storage, 0x4e20, 0x20)
2052EXC_COMMON_BEGIN(h_instr_storage_common)
2053	GEN_COMMON h_instr_storage
2054	addi	r3,r1,STACK_FRAME_OVERHEAD
2055	bl	unknown_exception
2056	b	interrupt_return
2057
2058	GEN_KVM h_instr_storage
2059
2060
2061/**
2062 * Interrupt 0xe40 - Hypervisor Emulation Assistance Interrupt.
2063 */
2064INT_DEFINE_BEGIN(emulation_assist)
2065	IVEC=0xe40
2066	IHSRR=1
2067	IKVM_REAL=1
2068	IKVM_VIRT=1
2069INT_DEFINE_END(emulation_assist)
2070
2071EXC_REAL_BEGIN(emulation_assist, 0xe40, 0x20)
2072	GEN_INT_ENTRY emulation_assist, virt=0, ool=1
2073EXC_REAL_END(emulation_assist, 0xe40, 0x20)
2074EXC_VIRT_BEGIN(emulation_assist, 0x4e40, 0x20)
2075	GEN_INT_ENTRY emulation_assist, virt=1, ool=1
2076EXC_VIRT_END(emulation_assist, 0x4e40, 0x20)
2077EXC_COMMON_BEGIN(emulation_assist_common)
2078	GEN_COMMON emulation_assist
2079	addi	r3,r1,STACK_FRAME_OVERHEAD
2080	bl	emulation_assist_interrupt
2081	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2082	b	interrupt_return
2083
2084	GEN_KVM emulation_assist
2085
2086
2087/**
2088 * Interrupt 0xe60 - Hypervisor Maintenance Interrupt (HMI).
2089 * This is an asynchronous interrupt caused by a Hypervisor Maintenance
2090 * Exception. It is always taken in real mode but uses HSRR registers
2091 * unlike SRESET and MCE.
2092 *
2093 * It is maskable in hardware by clearing MSR[EE], and partially soft-maskable
2094 * with IRQS_DISABLED mask (i.e., local_irq_disable()).
2095 *
2096 * Handling:
2097 * This is a special case, this is handled similarly to machine checks, with an
2098 * initial real mode handler that is not soft-masked, which attempts to fix the
2099 * problem. Then a regular handler which is soft-maskable and reports the
2100 * problem.
2101 *
2102 * The emergency stack is used for the early real mode handler.
2103 *
2104 * XXX: unclear why MCE and HMI schemes could not be made common, e.g.,
2105 * either use soft-masking for the MCE, or use irq_work for the HMI.
2106 *
2107 * KVM:
2108 * Unlike MCE, this calls into KVM without calling the real mode handler
2109 * first.
2110 */
2111INT_DEFINE_BEGIN(hmi_exception_early)
2112	IVEC=0xe60
2113	IHSRR=1
2114	IREALMODE_COMMON=1
2115	ISTACK=0
2116	IRECONCILE=0
2117	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
2118	IKVM_REAL=1
2119INT_DEFINE_END(hmi_exception_early)
2120
2121INT_DEFINE_BEGIN(hmi_exception)
2122	IVEC=0xe60
2123	IHSRR=1
2124	IMASK=IRQS_DISABLED
2125	IKVM_REAL=1
2126INT_DEFINE_END(hmi_exception)
2127
2128EXC_REAL_BEGIN(hmi_exception, 0xe60, 0x20)
2129	GEN_INT_ENTRY hmi_exception_early, virt=0, ool=1
2130EXC_REAL_END(hmi_exception, 0xe60, 0x20)
2131EXC_VIRT_NONE(0x4e60, 0x20)
2132
2133EXC_COMMON_BEGIN(hmi_exception_early_common)
2134	__GEN_REALMODE_COMMON_ENTRY hmi_exception_early
2135
2136	mr	r10,r1			/* Save r1 */
2137	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack for realmode */
2138	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
2139
2140	__GEN_COMMON_BODY hmi_exception_early
2141
2142	addi	r3,r1,STACK_FRAME_OVERHEAD
2143	bl	hmi_exception_realmode
2144	cmpdi	cr0,r3,0
2145	bne	1f
2146
2147	EXCEPTION_RESTORE_REGS hsrr=1
2148	HRFI_TO_USER_OR_KERNEL
2149
21501:
2151	/*
2152	 * Go to virtual mode and pull the HMI event information from
2153	 * firmware.
2154	 */
2155	EXCEPTION_RESTORE_REGS hsrr=1
2156	GEN_INT_ENTRY hmi_exception, virt=0
2157
2158	GEN_KVM hmi_exception_early
2159
2160EXC_COMMON_BEGIN(hmi_exception_common)
2161	GEN_COMMON hmi_exception
2162	FINISH_NAP
2163	RUNLATCH_ON
2164	addi	r3,r1,STACK_FRAME_OVERHEAD
2165	bl	handle_hmi_exception
2166	b	interrupt_return
2167
2168	GEN_KVM hmi_exception
2169
2170
2171/**
2172 * Interrupt 0xe80 - Directed Hypervisor Doorbell Interrupt.
2173 * This is an asynchronous interrupt in response to a msgsnd doorbell.
2174 * Similar to the 0xa00 doorbell but for host rather than guest.
2175 */
2176INT_DEFINE_BEGIN(h_doorbell)
2177	IVEC=0xe80
2178	IHSRR=1
2179	IMASK=IRQS_DISABLED
2180	IKVM_REAL=1
2181	IKVM_VIRT=1
2182INT_DEFINE_END(h_doorbell)
2183
2184EXC_REAL_BEGIN(h_doorbell, 0xe80, 0x20)
2185	GEN_INT_ENTRY h_doorbell, virt=0, ool=1
2186EXC_REAL_END(h_doorbell, 0xe80, 0x20)
2187EXC_VIRT_BEGIN(h_doorbell, 0x4e80, 0x20)
2188	GEN_INT_ENTRY h_doorbell, virt=1, ool=1
2189EXC_VIRT_END(h_doorbell, 0x4e80, 0x20)
2190EXC_COMMON_BEGIN(h_doorbell_common)
2191	GEN_COMMON h_doorbell
2192	FINISH_NAP
2193	RUNLATCH_ON
2194	addi	r3,r1,STACK_FRAME_OVERHEAD
2195#ifdef CONFIG_PPC_DOORBELL
2196	bl	doorbell_exception
2197#else
2198	bl	unknown_exception
2199#endif
2200	b	interrupt_return
2201
2202	GEN_KVM h_doorbell
2203
2204
2205/**
2206 * Interrupt 0xea0 - Hypervisor Virtualization Interrupt.
2207 * This is an asynchronous interrupt in response to an "external exception".
2208 * Similar to 0x500 but for host only.
2209 */
2210INT_DEFINE_BEGIN(h_virt_irq)
2211	IVEC=0xea0
2212	IHSRR=1
2213	IMASK=IRQS_DISABLED
2214	IKVM_REAL=1
2215	IKVM_VIRT=1
2216INT_DEFINE_END(h_virt_irq)
2217
2218EXC_REAL_BEGIN(h_virt_irq, 0xea0, 0x20)
2219	GEN_INT_ENTRY h_virt_irq, virt=0, ool=1
2220EXC_REAL_END(h_virt_irq, 0xea0, 0x20)
2221EXC_VIRT_BEGIN(h_virt_irq, 0x4ea0, 0x20)
2222	GEN_INT_ENTRY h_virt_irq, virt=1, ool=1
2223EXC_VIRT_END(h_virt_irq, 0x4ea0, 0x20)
2224EXC_COMMON_BEGIN(h_virt_irq_common)
2225	GEN_COMMON h_virt_irq
2226	FINISH_NAP
2227	RUNLATCH_ON
2228	addi	r3,r1,STACK_FRAME_OVERHEAD
2229	bl	do_IRQ
2230	b	interrupt_return
2231
2232	GEN_KVM h_virt_irq
2233
2234
2235EXC_REAL_NONE(0xec0, 0x20)
2236EXC_VIRT_NONE(0x4ec0, 0x20)
2237EXC_REAL_NONE(0xee0, 0x20)
2238EXC_VIRT_NONE(0x4ee0, 0x20)
2239
2240
2241/*
2242 * Interrupt 0xf00 - Performance Monitor Interrupt (PMI, PMU).
2243 * This is an asynchronous interrupt in response to a PMU exception.
2244 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
2245 * IRQS_PMI_DISABLED mask (NOTE: NOT local_irq_disable()).
2246 *
2247 * Handling:
2248 * This calls into the perf subsystem.
2249 *
2250 * Like the watchdog soft-nmi, it appears an NMI interrupt to Linux, in that it
2251 * runs under local_irq_disable. However it may be soft-masked in
2252 * powerpc-specific code.
2253 *
2254 * If soft masked, the masked handler will note the pending interrupt for
2255 * replay, and clear MSR[EE] in the interrupted context.
2256 */
2257INT_DEFINE_BEGIN(performance_monitor)
2258	IVEC=0xf00
2259	IMASK=IRQS_PMI_DISABLED
2260#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2261	IKVM_REAL=1
2262#endif
2263INT_DEFINE_END(performance_monitor)
2264
2265EXC_REAL_BEGIN(performance_monitor, 0xf00, 0x20)
2266	GEN_INT_ENTRY performance_monitor, virt=0, ool=1
2267EXC_REAL_END(performance_monitor, 0xf00, 0x20)
2268EXC_VIRT_BEGIN(performance_monitor, 0x4f00, 0x20)
2269	GEN_INT_ENTRY performance_monitor, virt=1, ool=1
2270EXC_VIRT_END(performance_monitor, 0x4f00, 0x20)
2271EXC_COMMON_BEGIN(performance_monitor_common)
2272	GEN_COMMON performance_monitor
2273	FINISH_NAP
2274	RUNLATCH_ON
2275	addi	r3,r1,STACK_FRAME_OVERHEAD
2276	bl	performance_monitor_exception
2277	b	interrupt_return
2278
2279	GEN_KVM performance_monitor
2280
2281
2282/**
2283 * Interrupt 0xf20 - Vector Unavailable Interrupt.
2284 * This is a synchronous interrupt in response to
2285 * executing a vector (or altivec) instruction with MSR[VEC]=0.
2286 * Similar to FP unavailable.
2287 */
2288INT_DEFINE_BEGIN(altivec_unavailable)
2289	IVEC=0xf20
2290	IRECONCILE=0
2291#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2292	IKVM_REAL=1
2293#endif
2294INT_DEFINE_END(altivec_unavailable)
2295
2296EXC_REAL_BEGIN(altivec_unavailable, 0xf20, 0x20)
2297	GEN_INT_ENTRY altivec_unavailable, virt=0, ool=1
2298EXC_REAL_END(altivec_unavailable, 0xf20, 0x20)
2299EXC_VIRT_BEGIN(altivec_unavailable, 0x4f20, 0x20)
2300	GEN_INT_ENTRY altivec_unavailable, virt=1, ool=1
2301EXC_VIRT_END(altivec_unavailable, 0x4f20, 0x20)
2302EXC_COMMON_BEGIN(altivec_unavailable_common)
2303	GEN_COMMON altivec_unavailable
2304#ifdef CONFIG_ALTIVEC
2305BEGIN_FTR_SECTION
2306	beq	1f
2307#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2308  BEGIN_FTR_SECTION_NESTED(69)
2309	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2310	 * transaction), go do TM stuff
2311	 */
2312	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2313	bne-	2f
2314  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2315#endif
2316	bl	load_up_altivec
2317	b	fast_interrupt_return
2318#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
23192:	/* User process was in a transaction */
2320	RECONCILE_IRQ_STATE(r10, r11)
2321	addi	r3,r1,STACK_FRAME_OVERHEAD
2322	bl	altivec_unavailable_tm
2323	b	interrupt_return
2324#endif
23251:
2326END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2327#endif
2328	RECONCILE_IRQ_STATE(r10, r11)
2329	addi	r3,r1,STACK_FRAME_OVERHEAD
2330	bl	altivec_unavailable_exception
2331	b	interrupt_return
2332
2333	GEN_KVM altivec_unavailable
2334
2335
2336/**
2337 * Interrupt 0xf40 - VSX Unavailable Interrupt.
2338 * This is a synchronous interrupt in response to
2339 * executing a VSX instruction with MSR[VSX]=0.
2340 * Similar to FP unavailable.
2341 */
2342INT_DEFINE_BEGIN(vsx_unavailable)
2343	IVEC=0xf40
2344	IRECONCILE=0
2345#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2346	IKVM_REAL=1
2347#endif
2348INT_DEFINE_END(vsx_unavailable)
2349
2350EXC_REAL_BEGIN(vsx_unavailable, 0xf40, 0x20)
2351	GEN_INT_ENTRY vsx_unavailable, virt=0, ool=1
2352EXC_REAL_END(vsx_unavailable, 0xf40, 0x20)
2353EXC_VIRT_BEGIN(vsx_unavailable, 0x4f40, 0x20)
2354	GEN_INT_ENTRY vsx_unavailable, virt=1, ool=1
2355EXC_VIRT_END(vsx_unavailable, 0x4f40, 0x20)
2356EXC_COMMON_BEGIN(vsx_unavailable_common)
2357	GEN_COMMON vsx_unavailable
2358#ifdef CONFIG_VSX
2359BEGIN_FTR_SECTION
2360	beq	1f
2361#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2362  BEGIN_FTR_SECTION_NESTED(69)
2363	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2364	 * transaction), go do TM stuff
2365	 */
2366	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2367	bne-	2f
2368  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2369#endif
2370	b	load_up_vsx
2371#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
23722:	/* User process was in a transaction */
2373	RECONCILE_IRQ_STATE(r10, r11)
2374	addi	r3,r1,STACK_FRAME_OVERHEAD
2375	bl	vsx_unavailable_tm
2376	b	interrupt_return
2377#endif
23781:
2379END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2380#endif
2381	RECONCILE_IRQ_STATE(r10, r11)
2382	addi	r3,r1,STACK_FRAME_OVERHEAD
2383	bl	vsx_unavailable_exception
2384	b	interrupt_return
2385
2386	GEN_KVM vsx_unavailable
2387
2388
2389/**
2390 * Interrupt 0xf60 - Facility Unavailable Interrupt.
2391 * This is a synchronous interrupt in response to
2392 * executing an instruction without access to the facility that can be
2393 * resolved by the OS (e.g., FSCR, MSR).
2394 * Similar to FP unavailable.
2395 */
2396INT_DEFINE_BEGIN(facility_unavailable)
2397	IVEC=0xf60
2398#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2399	IKVM_REAL=1
2400#endif
2401INT_DEFINE_END(facility_unavailable)
2402
2403EXC_REAL_BEGIN(facility_unavailable, 0xf60, 0x20)
2404	GEN_INT_ENTRY facility_unavailable, virt=0, ool=1
2405EXC_REAL_END(facility_unavailable, 0xf60, 0x20)
2406EXC_VIRT_BEGIN(facility_unavailable, 0x4f60, 0x20)
2407	GEN_INT_ENTRY facility_unavailable, virt=1, ool=1
2408EXC_VIRT_END(facility_unavailable, 0x4f60, 0x20)
2409EXC_COMMON_BEGIN(facility_unavailable_common)
2410	GEN_COMMON facility_unavailable
2411	addi	r3,r1,STACK_FRAME_OVERHEAD
2412	bl	facility_unavailable_exception
2413	b	interrupt_return
2414
2415	GEN_KVM facility_unavailable
2416
2417
2418/**
2419 * Interrupt 0xf60 - Hypervisor Facility Unavailable Interrupt.
2420 * This is a synchronous interrupt in response to
2421 * executing an instruction without access to the facility that can only
2422 * be resolved in HV mode (e.g., HFSCR).
2423 * Similar to FP unavailable.
2424 */
2425INT_DEFINE_BEGIN(h_facility_unavailable)
2426	IVEC=0xf80
2427	IHSRR=1
2428	IKVM_REAL=1
2429	IKVM_VIRT=1
2430INT_DEFINE_END(h_facility_unavailable)
2431
2432EXC_REAL_BEGIN(h_facility_unavailable, 0xf80, 0x20)
2433	GEN_INT_ENTRY h_facility_unavailable, virt=0, ool=1
2434EXC_REAL_END(h_facility_unavailable, 0xf80, 0x20)
2435EXC_VIRT_BEGIN(h_facility_unavailable, 0x4f80, 0x20)
2436	GEN_INT_ENTRY h_facility_unavailable, virt=1, ool=1
2437EXC_VIRT_END(h_facility_unavailable, 0x4f80, 0x20)
2438EXC_COMMON_BEGIN(h_facility_unavailable_common)
2439	GEN_COMMON h_facility_unavailable
2440	addi	r3,r1,STACK_FRAME_OVERHEAD
2441	bl	facility_unavailable_exception
2442	b	interrupt_return
2443
2444	GEN_KVM h_facility_unavailable
2445
2446
2447EXC_REAL_NONE(0xfa0, 0x20)
2448EXC_VIRT_NONE(0x4fa0, 0x20)
2449EXC_REAL_NONE(0xfc0, 0x20)
2450EXC_VIRT_NONE(0x4fc0, 0x20)
2451EXC_REAL_NONE(0xfe0, 0x20)
2452EXC_VIRT_NONE(0x4fe0, 0x20)
2453
2454EXC_REAL_NONE(0x1000, 0x100)
2455EXC_VIRT_NONE(0x5000, 0x100)
2456EXC_REAL_NONE(0x1100, 0x100)
2457EXC_VIRT_NONE(0x5100, 0x100)
2458
2459#ifdef CONFIG_CBE_RAS
2460INT_DEFINE_BEGIN(cbe_system_error)
2461	IVEC=0x1200
2462	IHSRR=1
2463	IKVM_SKIP=1
2464	IKVM_REAL=1
2465INT_DEFINE_END(cbe_system_error)
2466
2467EXC_REAL_BEGIN(cbe_system_error, 0x1200, 0x100)
2468	GEN_INT_ENTRY cbe_system_error, virt=0
2469EXC_REAL_END(cbe_system_error, 0x1200, 0x100)
2470EXC_VIRT_NONE(0x5200, 0x100)
2471EXC_COMMON_BEGIN(cbe_system_error_common)
2472	GEN_COMMON cbe_system_error
2473	addi	r3,r1,STACK_FRAME_OVERHEAD
2474	bl	cbe_system_error_exception
2475	b	interrupt_return
2476
2477	GEN_KVM cbe_system_error
2478
2479#else /* CONFIG_CBE_RAS */
2480EXC_REAL_NONE(0x1200, 0x100)
2481EXC_VIRT_NONE(0x5200, 0x100)
2482#endif
2483
2484
2485INT_DEFINE_BEGIN(instruction_breakpoint)
2486	IVEC=0x1300
2487#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2488	IKVM_SKIP=1
2489	IKVM_REAL=1
2490#endif
2491INT_DEFINE_END(instruction_breakpoint)
2492
2493EXC_REAL_BEGIN(instruction_breakpoint, 0x1300, 0x100)
2494	GEN_INT_ENTRY instruction_breakpoint, virt=0
2495EXC_REAL_END(instruction_breakpoint, 0x1300, 0x100)
2496EXC_VIRT_BEGIN(instruction_breakpoint, 0x5300, 0x100)
2497	GEN_INT_ENTRY instruction_breakpoint, virt=1
2498EXC_VIRT_END(instruction_breakpoint, 0x5300, 0x100)
2499EXC_COMMON_BEGIN(instruction_breakpoint_common)
2500	GEN_COMMON instruction_breakpoint
2501	addi	r3,r1,STACK_FRAME_OVERHEAD
2502	bl	instruction_breakpoint_exception
2503	b	interrupt_return
2504
2505	GEN_KVM instruction_breakpoint
2506
2507
2508EXC_REAL_NONE(0x1400, 0x100)
2509EXC_VIRT_NONE(0x5400, 0x100)
2510
2511/**
2512 * Interrupt 0x1500 - Soft Patch Interrupt
2513 *
2514 * Handling:
2515 * This is an implementation specific interrupt which can be used for a
2516 * range of exceptions.
2517 *
2518 * This interrupt handler is unique in that it runs the denormal assist
2519 * code even for guests (and even in guest context) without going to KVM,
2520 * for speed. POWER9 does not raise denorm exceptions, so this special case
2521 * could be phased out in future to reduce special cases.
2522 */
2523INT_DEFINE_BEGIN(denorm_exception)
2524	IVEC=0x1500
2525	IHSRR=1
2526	IBRANCH_COMMON=0
2527	IKVM_REAL=1
2528INT_DEFINE_END(denorm_exception)
2529
2530EXC_REAL_BEGIN(denorm_exception, 0x1500, 0x100)
2531	GEN_INT_ENTRY denorm_exception, virt=0
2532#ifdef CONFIG_PPC_DENORMALISATION
2533	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2534	bne+	denorm_assist
2535#endif
2536	GEN_BRANCH_TO_COMMON denorm_exception, virt=0
2537EXC_REAL_END(denorm_exception, 0x1500, 0x100)
2538#ifdef CONFIG_PPC_DENORMALISATION
2539EXC_VIRT_BEGIN(denorm_exception, 0x5500, 0x100)
2540	GEN_INT_ENTRY denorm_exception, virt=1
2541	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2542	bne+	denorm_assist
2543	GEN_BRANCH_TO_COMMON denorm_exception, virt=1
2544EXC_VIRT_END(denorm_exception, 0x5500, 0x100)
2545#else
2546EXC_VIRT_NONE(0x5500, 0x100)
2547#endif
2548
2549#ifdef CONFIG_PPC_DENORMALISATION
2550TRAMP_REAL_BEGIN(denorm_assist)
2551BEGIN_FTR_SECTION
2552/*
2553 * To denormalise we need to move a copy of the register to itself.
2554 * For POWER6 do that here for all FP regs.
2555 */
2556	mfmsr	r10
2557	ori	r10,r10,(MSR_FP|MSR_FE0|MSR_FE1)
2558	xori	r10,r10,(MSR_FE0|MSR_FE1)
2559	mtmsrd	r10
2560	sync
2561
2562	.Lreg=0
2563	.rept 32
2564	fmr	.Lreg,.Lreg
2565	.Lreg=.Lreg+1
2566	.endr
2567
2568FTR_SECTION_ELSE
2569/*
2570 * To denormalise we need to move a copy of the register to itself.
2571 * For POWER7 do that here for the first 32 VSX registers only.
2572 */
2573	mfmsr	r10
2574	oris	r10,r10,MSR_VSX@h
2575	mtmsrd	r10
2576	sync
2577
2578	.Lreg=0
2579	.rept 32
2580	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2581	.Lreg=.Lreg+1
2582	.endr
2583
2584ALT_FTR_SECTION_END_IFCLR(CPU_FTR_ARCH_206)
2585
2586BEGIN_FTR_SECTION
2587	b	denorm_done
2588END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
2589/*
2590 * To denormalise we need to move a copy of the register to itself.
2591 * For POWER8 we need to do that for all 64 VSX registers
2592 */
2593	.Lreg=32
2594	.rept 32
2595	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2596	.Lreg=.Lreg+1
2597	.endr
2598
2599denorm_done:
2600	mfspr	r11,SPRN_HSRR0
2601	subi	r11,r11,4
2602	mtspr	SPRN_HSRR0,r11
2603	mtcrf	0x80,r9
2604	ld	r9,PACA_EXGEN+EX_R9(r13)
2605BEGIN_FTR_SECTION
2606	ld	r10,PACA_EXGEN+EX_PPR(r13)
2607	mtspr	SPRN_PPR,r10
2608END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
2609BEGIN_FTR_SECTION
2610	ld	r10,PACA_EXGEN+EX_CFAR(r13)
2611	mtspr	SPRN_CFAR,r10
2612END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
2613	ld	r10,PACA_EXGEN+EX_R10(r13)
2614	ld	r11,PACA_EXGEN+EX_R11(r13)
2615	ld	r12,PACA_EXGEN+EX_R12(r13)
2616	ld	r13,PACA_EXGEN+EX_R13(r13)
2617	HRFI_TO_UNKNOWN
2618	b	.
2619#endif
2620
2621EXC_COMMON_BEGIN(denorm_exception_common)
2622	GEN_COMMON denorm_exception
2623	addi	r3,r1,STACK_FRAME_OVERHEAD
2624	bl	unknown_exception
2625	b	interrupt_return
2626
2627	GEN_KVM denorm_exception
2628
2629
2630#ifdef CONFIG_CBE_RAS
2631INT_DEFINE_BEGIN(cbe_maintenance)
2632	IVEC=0x1600
2633	IHSRR=1
2634	IKVM_SKIP=1
2635	IKVM_REAL=1
2636INT_DEFINE_END(cbe_maintenance)
2637
2638EXC_REAL_BEGIN(cbe_maintenance, 0x1600, 0x100)
2639	GEN_INT_ENTRY cbe_maintenance, virt=0
2640EXC_REAL_END(cbe_maintenance, 0x1600, 0x100)
2641EXC_VIRT_NONE(0x5600, 0x100)
2642EXC_COMMON_BEGIN(cbe_maintenance_common)
2643	GEN_COMMON cbe_maintenance
2644	addi	r3,r1,STACK_FRAME_OVERHEAD
2645	bl	cbe_maintenance_exception
2646	b	interrupt_return
2647
2648	GEN_KVM cbe_maintenance
2649
2650#else /* CONFIG_CBE_RAS */
2651EXC_REAL_NONE(0x1600, 0x100)
2652EXC_VIRT_NONE(0x5600, 0x100)
2653#endif
2654
2655
2656INT_DEFINE_BEGIN(altivec_assist)
2657	IVEC=0x1700
2658#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2659	IKVM_REAL=1
2660#endif
2661INT_DEFINE_END(altivec_assist)
2662
2663EXC_REAL_BEGIN(altivec_assist, 0x1700, 0x100)
2664	GEN_INT_ENTRY altivec_assist, virt=0
2665EXC_REAL_END(altivec_assist, 0x1700, 0x100)
2666EXC_VIRT_BEGIN(altivec_assist, 0x5700, 0x100)
2667	GEN_INT_ENTRY altivec_assist, virt=1
2668EXC_VIRT_END(altivec_assist, 0x5700, 0x100)
2669EXC_COMMON_BEGIN(altivec_assist_common)
2670	GEN_COMMON altivec_assist
2671	addi	r3,r1,STACK_FRAME_OVERHEAD
2672#ifdef CONFIG_ALTIVEC
2673	bl	altivec_assist_exception
2674	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2675#else
2676	bl	unknown_exception
2677#endif
2678	b	interrupt_return
2679
2680	GEN_KVM altivec_assist
2681
2682
2683#ifdef CONFIG_CBE_RAS
2684INT_DEFINE_BEGIN(cbe_thermal)
2685	IVEC=0x1800
2686	IHSRR=1
2687	IKVM_SKIP=1
2688	IKVM_REAL=1
2689INT_DEFINE_END(cbe_thermal)
2690
2691EXC_REAL_BEGIN(cbe_thermal, 0x1800, 0x100)
2692	GEN_INT_ENTRY cbe_thermal, virt=0
2693EXC_REAL_END(cbe_thermal, 0x1800, 0x100)
2694EXC_VIRT_NONE(0x5800, 0x100)
2695EXC_COMMON_BEGIN(cbe_thermal_common)
2696	GEN_COMMON cbe_thermal
2697	addi	r3,r1,STACK_FRAME_OVERHEAD
2698	bl	cbe_thermal_exception
2699	b	interrupt_return
2700
2701	GEN_KVM cbe_thermal
2702
2703#else /* CONFIG_CBE_RAS */
2704EXC_REAL_NONE(0x1800, 0x100)
2705EXC_VIRT_NONE(0x5800, 0x100)
2706#endif
2707
2708
2709#ifdef CONFIG_PPC_WATCHDOG
2710
2711INT_DEFINE_BEGIN(soft_nmi)
2712	IVEC=0x900
2713	ISTACK=0
2714	IRECONCILE=0	/* Soft-NMI may fire under local_irq_disable */
2715INT_DEFINE_END(soft_nmi)
2716
2717/*
2718 * Branch to soft_nmi_interrupt using the emergency stack. The emergency
2719 * stack is one that is usable by maskable interrupts so long as MSR_EE
2720 * remains off. It is used for recovery when something has corrupted the
2721 * normal kernel stack, for example. The "soft NMI" must not use the process
2722 * stack because we want irq disabled sections to avoid touching the stack
2723 * at all (other than PMU interrupts), so use the emergency stack for this,
2724 * and run it entirely with interrupts hard disabled.
2725 */
2726EXC_COMMON_BEGIN(soft_nmi_common)
2727	mfspr	r11,SPRN_SRR0
2728	mr	r10,r1
2729	ld	r1,PACAEMERGSP(r13)
2730	subi	r1,r1,INT_FRAME_SIZE
2731	__GEN_COMMON_BODY soft_nmi
2732
2733	/*
2734	 * Set IRQS_ALL_DISABLED and save PACAIRQHAPPENED (see
2735	 * system_reset_common)
2736	 */
2737	li	r10,IRQS_ALL_DISABLED
2738	stb	r10,PACAIRQSOFTMASK(r13)
2739	lbz	r10,PACAIRQHAPPENED(r13)
2740	std	r10,_DAR(r1)
2741	ori	r10,r10,PACA_IRQ_HARD_DIS
2742	stb	r10,PACAIRQHAPPENED(r13)
2743
2744	addi	r3,r1,STACK_FRAME_OVERHEAD
2745	bl	soft_nmi_interrupt
2746
2747	/* Clear MSR_RI before setting SRR0 and SRR1. */
2748	li	r9,0
2749	mtmsrd	r9,1
2750
2751	/*
2752	 * Restore soft mask settings.
2753	 */
2754	ld	r10,_DAR(r1)
2755	stb	r10,PACAIRQHAPPENED(r13)
2756	ld	r10,SOFTE(r1)
2757	stb	r10,PACAIRQSOFTMASK(r13)
2758
2759	kuap_restore_amr r10
2760	EXCEPTION_RESTORE_REGS hsrr=0
2761	RFI_TO_KERNEL
2762
2763#endif /* CONFIG_PPC_WATCHDOG */
2764
2765/*
2766 * An interrupt came in while soft-disabled. We set paca->irq_happened, then:
2767 * - If it was a decrementer interrupt, we bump the dec to max and and return.
2768 * - If it was a doorbell we return immediately since doorbells are edge
2769 *   triggered and won't automatically refire.
2770 * - If it was a HMI we return immediately since we handled it in realmode
2771 *   and it won't refire.
2772 * - Else it is one of PACA_IRQ_MUST_HARD_MASK, so hard disable and return.
2773 * This is called with r10 containing the value to OR to the paca field.
2774 */
2775.macro MASKED_INTERRUPT hsrr=0
2776	.if \hsrr
2777masked_Hinterrupt:
2778	.else
2779masked_interrupt:
2780	.endif
2781	lbz	r11,PACAIRQHAPPENED(r13)
2782	or	r11,r11,r10
2783	stb	r11,PACAIRQHAPPENED(r13)
2784	cmpwi	r10,PACA_IRQ_DEC
2785	bne	1f
2786	lis	r10,0x7fff
2787	ori	r10,r10,0xffff
2788	mtspr	SPRN_DEC,r10
2789#ifdef CONFIG_PPC_WATCHDOG
2790	b	soft_nmi_common
2791#else
2792	b	2f
2793#endif
27941:	andi.	r10,r10,PACA_IRQ_MUST_HARD_MASK
2795	beq	2f
2796	xori	r12,r12,MSR_EE	/* clear MSR_EE */
2797	.if \hsrr
2798	mtspr	SPRN_HSRR1,r12
2799	.else
2800	mtspr	SPRN_SRR1,r12
2801	.endif
2802	ori	r11,r11,PACA_IRQ_HARD_DIS
2803	stb	r11,PACAIRQHAPPENED(r13)
28042:	/* done */
2805	ld	r10,PACA_EXGEN+EX_CTR(r13)
2806	mtctr	r10
2807	mtcrf	0x80,r9
2808	std	r1,PACAR1(r13)
2809	ld	r9,PACA_EXGEN+EX_R9(r13)
2810	ld	r10,PACA_EXGEN+EX_R10(r13)
2811	ld	r11,PACA_EXGEN+EX_R11(r13)
2812	ld	r12,PACA_EXGEN+EX_R12(r13)
2813	/* returns to kernel where r13 must be set up, so don't restore it */
2814	.if \hsrr
2815	HRFI_TO_KERNEL
2816	.else
2817	RFI_TO_KERNEL
2818	.endif
2819	b	.
2820.endm
2821
2822TRAMP_REAL_BEGIN(stf_barrier_fallback)
2823	std	r9,PACA_EXRFI+EX_R9(r13)
2824	std	r10,PACA_EXRFI+EX_R10(r13)
2825	sync
2826	ld	r9,PACA_EXRFI+EX_R9(r13)
2827	ld	r10,PACA_EXRFI+EX_R10(r13)
2828	ori	31,31,0
2829	.rept 14
2830	b	1f
28311:
2832	.endr
2833	blr
2834
2835TRAMP_REAL_BEGIN(rfi_flush_fallback)
2836	SET_SCRATCH0(r13);
2837	GET_PACA(r13);
2838	std	r1,PACA_EXRFI+EX_R12(r13)
2839	ld	r1,PACAKSAVE(r13)
2840	std	r9,PACA_EXRFI+EX_R9(r13)
2841	std	r10,PACA_EXRFI+EX_R10(r13)
2842	std	r11,PACA_EXRFI+EX_R11(r13)
2843	mfctr	r9
2844	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
2845	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
2846	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
2847	mtctr	r11
2848	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
2849
2850	/* order ld/st prior to dcbt stop all streams with flushing */
2851	sync
2852
2853	/*
2854	 * The load adresses are at staggered offsets within cachelines,
2855	 * which suits some pipelines better (on others it should not
2856	 * hurt).
2857	 */
28581:
2859	ld	r11,(0x80 + 8)*0(r10)
2860	ld	r11,(0x80 + 8)*1(r10)
2861	ld	r11,(0x80 + 8)*2(r10)
2862	ld	r11,(0x80 + 8)*3(r10)
2863	ld	r11,(0x80 + 8)*4(r10)
2864	ld	r11,(0x80 + 8)*5(r10)
2865	ld	r11,(0x80 + 8)*6(r10)
2866	ld	r11,(0x80 + 8)*7(r10)
2867	addi	r10,r10,0x80*8
2868	bdnz	1b
2869
2870	mtctr	r9
2871	ld	r9,PACA_EXRFI+EX_R9(r13)
2872	ld	r10,PACA_EXRFI+EX_R10(r13)
2873	ld	r11,PACA_EXRFI+EX_R11(r13)
2874	ld	r1,PACA_EXRFI+EX_R12(r13)
2875	GET_SCRATCH0(r13);
2876	rfid
2877
2878TRAMP_REAL_BEGIN(hrfi_flush_fallback)
2879	SET_SCRATCH0(r13);
2880	GET_PACA(r13);
2881	std	r1,PACA_EXRFI+EX_R12(r13)
2882	ld	r1,PACAKSAVE(r13)
2883	std	r9,PACA_EXRFI+EX_R9(r13)
2884	std	r10,PACA_EXRFI+EX_R10(r13)
2885	std	r11,PACA_EXRFI+EX_R11(r13)
2886	mfctr	r9
2887	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
2888	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
2889	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
2890	mtctr	r11
2891	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
2892
2893	/* order ld/st prior to dcbt stop all streams with flushing */
2894	sync
2895
2896	/*
2897	 * The load adresses are at staggered offsets within cachelines,
2898	 * which suits some pipelines better (on others it should not
2899	 * hurt).
2900	 */
29011:
2902	ld	r11,(0x80 + 8)*0(r10)
2903	ld	r11,(0x80 + 8)*1(r10)
2904	ld	r11,(0x80 + 8)*2(r10)
2905	ld	r11,(0x80 + 8)*3(r10)
2906	ld	r11,(0x80 + 8)*4(r10)
2907	ld	r11,(0x80 + 8)*5(r10)
2908	ld	r11,(0x80 + 8)*6(r10)
2909	ld	r11,(0x80 + 8)*7(r10)
2910	addi	r10,r10,0x80*8
2911	bdnz	1b
2912
2913	mtctr	r9
2914	ld	r9,PACA_EXRFI+EX_R9(r13)
2915	ld	r10,PACA_EXRFI+EX_R10(r13)
2916	ld	r11,PACA_EXRFI+EX_R11(r13)
2917	ld	r1,PACA_EXRFI+EX_R12(r13)
2918	GET_SCRATCH0(r13);
2919	hrfid
2920
2921USE_TEXT_SECTION()
2922	MASKED_INTERRUPT
2923	MASKED_INTERRUPT hsrr=1
2924
2925#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
2926kvmppc_skip_interrupt:
2927	/*
2928	 * Here all GPRs are unchanged from when the interrupt happened
2929	 * except for r13, which is saved in SPRG_SCRATCH0.
2930	 */
2931	mfspr	r13, SPRN_SRR0
2932	addi	r13, r13, 4
2933	mtspr	SPRN_SRR0, r13
2934	GET_SCRATCH0(r13)
2935	RFI_TO_KERNEL
2936	b	.
2937
2938kvmppc_skip_Hinterrupt:
2939	/*
2940	 * Here all GPRs are unchanged from when the interrupt happened
2941	 * except for r13, which is saved in SPRG_SCRATCH0.
2942	 */
2943	mfspr	r13, SPRN_HSRR0
2944	addi	r13, r13, 4
2945	mtspr	SPRN_HSRR0, r13
2946	GET_SCRATCH0(r13)
2947	HRFI_TO_KERNEL
2948	b	.
2949#endif
2950
2951	/*
2952	 * Relocation-on interrupts: A subset of the interrupts can be delivered
2953	 * with IR=1/DR=1, if AIL==2 and MSR.HV won't be changed by delivering
2954	 * it.  Addresses are the same as the original interrupt addresses, but
2955	 * offset by 0xc000000000004000.
2956	 * It's impossible to receive interrupts below 0x300 via this mechanism.
2957	 * KVM: None of these traps are from the guest ; anything that escalated
2958	 * to HV=1 from HV=0 is delivered via real mode handlers.
2959	 */
2960
2961	/*
2962	 * This uses the standard macro, since the original 0x300 vector
2963	 * only has extra guff for STAB-based processors -- which never
2964	 * come here.
2965	 */
2966
2967EXC_COMMON_BEGIN(ppc64_runlatch_on_trampoline)
2968	b	__ppc64_runlatch_on
2969
2970USE_FIXED_SECTION(virt_trampolines)
2971	/*
2972	 * The __end_interrupts marker must be past the out-of-line (OOL)
2973	 * handlers, so that they are copied to real address 0x100 when running
2974	 * a relocatable kernel. This ensures they can be reached from the short
2975	 * trampoline handlers (like 0x4f00, 0x4f20, etc.) which branch
2976	 * directly, without using LOAD_HANDLER().
2977	 */
2978	.align	7
2979	.globl	__end_interrupts
2980__end_interrupts:
2981DEFINE_FIXED_SYMBOL(__end_interrupts)
2982
2983#ifdef CONFIG_PPC_970_NAP
2984	/*
2985	 * Called by exception entry code if _TLF_NAPPING was set, this clears
2986	 * the NAPPING flag, and redirects the exception exit to
2987	 * power4_fixup_nap_return.
2988	 */
2989	.globl power4_fixup_nap
2990EXC_COMMON_BEGIN(power4_fixup_nap)
2991	andc	r9,r9,r10
2992	std	r9,TI_LOCAL_FLAGS(r11)
2993	LOAD_REG_ADDR(r10, power4_idle_nap_return)
2994	std	r10,_NIP(r1)
2995	blr
2996
2997power4_idle_nap_return:
2998	blr
2999#endif
3000
3001CLOSE_FIXED_SECTION(real_vectors);
3002CLOSE_FIXED_SECTION(real_trampolines);
3003CLOSE_FIXED_SECTION(virt_vectors);
3004CLOSE_FIXED_SECTION(virt_trampolines);
3005
3006USE_TEXT_SECTION()
3007
3008/* MSR[RI] should be clear because this uses SRR[01] */
3009enable_machine_check:
3010	mflr	r0
3011	bcl	20,31,$+4
30120:	mflr	r3
3013	addi	r3,r3,(1f - 0b)
3014	mtspr	SPRN_SRR0,r3
3015	mfmsr	r3
3016	ori	r3,r3,MSR_ME
3017	mtspr	SPRN_SRR1,r3
3018	RFI_TO_KERNEL
30191:	mtlr	r0
3020	blr
3021
3022/* MSR[RI] should be clear because this uses SRR[01] */
3023disable_machine_check:
3024	mflr	r0
3025	bcl	20,31,$+4
30260:	mflr	r3
3027	addi	r3,r3,(1f - 0b)
3028	mtspr	SPRN_SRR0,r3
3029	mfmsr	r3
3030	li	r4,MSR_ME
3031	andc	r3,r3,r4
3032	mtspr	SPRN_SRR1,r3
3033	RFI_TO_KERNEL
30341:	mtlr	r0
3035	blr
3036
3037/*
3038 * Hash table stuff
3039 */
3040	.balign	IFETCH_ALIGN_BYTES
3041do_hash_page:
3042#ifdef CONFIG_PPC_BOOK3S_64
3043	lis	r0,(DSISR_BAD_FAULT_64S | DSISR_DABRMATCH | DSISR_KEYFAULT)@h
3044	ori	r0,r0,DSISR_BAD_FAULT_64S@l
3045	and.	r0,r5,r0		/* weird error? */
3046	bne-	handle_page_fault	/* if not, try to insert a HPTE */
3047	ld	r11, PACA_THREAD_INFO(r13)
3048	lwz	r0,TI_PREEMPT(r11)	/* If we're in an "NMI" */
3049	andis.	r0,r0,NMI_MASK@h	/* (i.e. an irq when soft-disabled) */
3050	bne	77f			/* then don't call hash_page now */
3051
3052	/*
3053	 * r3 contains the trap number
3054	 * r4 contains the faulting address
3055	 * r5 contains dsisr
3056	 * r6 msr
3057	 *
3058	 * at return r3 = 0 for success, 1 for page fault, negative for error
3059	 */
3060	bl	__hash_page		/* build HPTE if possible */
3061        cmpdi	r3,0			/* see if __hash_page succeeded */
3062
3063	/* Success */
3064	beq	interrupt_return	/* Return from exception on success */
3065
3066	/* Error */
3067	blt-	13f
3068
3069	/* Reload DAR/DSISR into r4/r5 for the DABR check below */
3070	ld	r4,_DAR(r1)
3071	ld      r5,_DSISR(r1)
3072#endif /* CONFIG_PPC_BOOK3S_64 */
3073
3074/* Here we have a page fault that hash_page can't handle. */
3075handle_page_fault:
307611:	andis.  r0,r5,DSISR_DABRMATCH@h
3077	bne-    handle_dabr_fault
3078	addi	r3,r1,STACK_FRAME_OVERHEAD
3079	bl	do_page_fault
3080	cmpdi	r3,0
3081	beq+	interrupt_return
3082	mr	r5,r3
3083	addi	r3,r1,STACK_FRAME_OVERHEAD
3084	ld	r4,_DAR(r1)
3085	bl	bad_page_fault
3086	b	interrupt_return
3087
3088/* We have a data breakpoint exception - handle it */
3089handle_dabr_fault:
3090	ld      r4,_DAR(r1)
3091	ld      r5,_DSISR(r1)
3092	addi    r3,r1,STACK_FRAME_OVERHEAD
3093	bl      do_break
3094	/*
3095	 * do_break() may have changed the NV GPRS while handling a breakpoint.
3096	 * If so, we need to restore them with their updated values.
3097	 */
3098	REST_NVGPRS(r1)
3099	b       interrupt_return
3100
3101
3102#ifdef CONFIG_PPC_BOOK3S_64
3103/* We have a page fault that hash_page could handle but HV refused
3104 * the PTE insertion
3105 */
310613:	mr	r5,r3
3107	addi	r3,r1,STACK_FRAME_OVERHEAD
3108	ld	r4,_DAR(r1)
3109	bl	low_hash_fault
3110	b	interrupt_return
3111#endif
3112
3113/*
3114 * We come here as a result of a DSI at a point where we don't want
3115 * to call hash_page, such as when we are accessing memory (possibly
3116 * user memory) inside a PMU interrupt that occurred while interrupts
3117 * were soft-disabled.  We want to invoke the exception handler for
3118 * the access, or panic if there isn't a handler.
3119 */
312077:	addi	r3,r1,STACK_FRAME_OVERHEAD
3121	li	r5,SIGSEGV
3122	bl	bad_page_fault
3123	b	interrupt_return
3124