1/* SPDX-License-Identifier: GPL-2.0 */
2/*
3 * This file contains the 64-bit "server" PowerPC variant
4 * of the low level exception handling including exception
5 * vectors, exception return, part of the slb and stab
6 * handling and other fixed offset specific things.
7 *
8 * This file is meant to be #included from head_64.S due to
9 * position dependent assembly.
10 *
11 * Most of this originates from head_64.S and thus has the same
12 * copyright history.
13 *
14 */
15
16#include <asm/hw_irq.h>
17#include <asm/exception-64s.h>
18#include <asm/ptrace.h>
19#include <asm/cpuidle.h>
20#include <asm/head-64.h>
21#include <asm/feature-fixups.h>
22#include <asm/kup.h>
23
24/* PACA save area offsets (exgen, exmc, etc) */
25#define EX_R9		0
26#define EX_R10		8
27#define EX_R11		16
28#define EX_R12		24
29#define EX_R13		32
30#define EX_DAR		40
31#define EX_DSISR	48
32#define EX_CCR		52
33#define EX_CFAR		56
34#define EX_PPR		64
35#define EX_CTR		72
36.if EX_SIZE != 10
37	.error "EX_SIZE is wrong"
38.endif
39
40/*
41 * Following are fixed section helper macros.
42 *
43 * EXC_REAL_BEGIN/END  - real, unrelocated exception vectors
44 * EXC_VIRT_BEGIN/END  - virt (AIL), unrelocated exception vectors
45 * TRAMP_REAL_BEGIN    - real, unrelocated helpers (virt may call these)
46 * TRAMP_VIRT_BEGIN    - virt, unreloc helpers (in practice, real can use)
47 * EXC_COMMON          - After switching to virtual, relocated mode.
48 */
49
50#define EXC_REAL_BEGIN(name, start, size)			\
51	FIXED_SECTION_ENTRY_BEGIN_LOCATION(real_vectors, exc_real_##start##_##name, start, size)
52
53#define EXC_REAL_END(name, start, size)				\
54	FIXED_SECTION_ENTRY_END_LOCATION(real_vectors, exc_real_##start##_##name, start, size)
55
56#define EXC_VIRT_BEGIN(name, start, size)			\
57	FIXED_SECTION_ENTRY_BEGIN_LOCATION(virt_vectors, exc_virt_##start##_##name, start, size)
58
59#define EXC_VIRT_END(name, start, size)				\
60	FIXED_SECTION_ENTRY_END_LOCATION(virt_vectors, exc_virt_##start##_##name, start, size)
61
62#define EXC_COMMON_BEGIN(name)					\
63	USE_TEXT_SECTION();					\
64	.balign IFETCH_ALIGN_BYTES;				\
65	.global name;						\
66	_ASM_NOKPROBE_SYMBOL(name);				\
67	DEFINE_FIXED_SYMBOL(name);				\
68name:
69
70#define TRAMP_REAL_BEGIN(name)					\
71	FIXED_SECTION_ENTRY_BEGIN(real_trampolines, name)
72
73#define TRAMP_VIRT_BEGIN(name)					\
74	FIXED_SECTION_ENTRY_BEGIN(virt_trampolines, name)
75
76#define EXC_REAL_NONE(start, size)				\
77	FIXED_SECTION_ENTRY_BEGIN_LOCATION(real_vectors, exc_real_##start##_##unused, start, size); \
78	FIXED_SECTION_ENTRY_END_LOCATION(real_vectors, exc_real_##start##_##unused, start, size)
79
80#define EXC_VIRT_NONE(start, size)				\
81	FIXED_SECTION_ENTRY_BEGIN_LOCATION(virt_vectors, exc_virt_##start##_##unused, start, size); \
82	FIXED_SECTION_ENTRY_END_LOCATION(virt_vectors, exc_virt_##start##_##unused, start, size)
83
84/*
85 * We're short on space and time in the exception prolog, so we can't
86 * use the normal LOAD_REG_IMMEDIATE macro to load the address of label.
87 * Instead we get the base of the kernel from paca->kernelbase and or in the low
88 * part of label. This requires that the label be within 64KB of kernelbase, and
89 * that kernelbase be 64K aligned.
90 */
91#define LOAD_HANDLER(reg, label)					\
92	ld	reg,PACAKBASE(r13);	/* get high part of &label */	\
93	ori	reg,reg,FIXED_SYMBOL_ABS_ADDR(label)
94
95#define __LOAD_HANDLER(reg, label)					\
96	ld	reg,PACAKBASE(r13);					\
97	ori	reg,reg,(ABS_ADDR(label))@l
98
99/*
100 * Branches from unrelocated code (e.g., interrupts) to labels outside
101 * head-y require >64K offsets.
102 */
103#define __LOAD_FAR_HANDLER(reg, label)					\
104	ld	reg,PACAKBASE(r13);					\
105	ori	reg,reg,(ABS_ADDR(label))@l;				\
106	addis	reg,reg,(ABS_ADDR(label))@h
107
108/*
109 * Branch to label using its 0xC000 address. This results in instruction
110 * address suitable for MSR[IR]=0 or 1, which allows relocation to be turned
111 * on using mtmsr rather than rfid.
112 *
113 * This could set the 0xc bits for !RELOCATABLE as an immediate, rather than
114 * load KBASE for a slight optimisation.
115 */
116#define BRANCH_TO_C000(reg, label)					\
117	__LOAD_FAR_HANDLER(reg, label);					\
118	mtctr	reg;							\
119	bctr
120
121/*
122 * Interrupt code generation macros
123 */
124#define IVEC		.L_IVEC_\name\()	/* Interrupt vector address */
125#define IHSRR		.L_IHSRR_\name\()	/* Sets SRR or HSRR registers */
126#define IHSRR_IF_HVMODE	.L_IHSRR_IF_HVMODE_\name\() /* HSRR if HV else SRR */
127#define IAREA		.L_IAREA_\name\()	/* PACA save area */
128#define IVIRT		.L_IVIRT_\name\()	/* Has virt mode entry point */
129#define IISIDE		.L_IISIDE_\name\()	/* Uses SRR0/1 not DAR/DSISR */
130#define IDAR		.L_IDAR_\name\()	/* Uses DAR (or SRR0) */
131#define IDSISR		.L_IDSISR_\name\()	/* Uses DSISR (or SRR1) */
132#define ISET_RI		.L_ISET_RI_\name\()	/* Run common code w/ MSR[RI]=1 */
133#define IBRANCH_TO_COMMON	.L_IBRANCH_TO_COMMON_\name\() /* ENTRY branch to common */
134#define IREALMODE_COMMON	.L_IREALMODE_COMMON_\name\() /* Common runs in realmode */
135#define IMASK		.L_IMASK_\name\()	/* IRQ soft-mask bit */
136#define IKVM_SKIP	.L_IKVM_SKIP_\name\()	/* Generate KVM skip handler */
137#define IKVM_REAL	.L_IKVM_REAL_\name\()	/* Real entry tests KVM */
138#define __IKVM_REAL(name)	.L_IKVM_REAL_ ## name
139#define IKVM_VIRT	.L_IKVM_VIRT_\name\()	/* Virt entry tests KVM */
140#define ISTACK		.L_ISTACK_\name\()	/* Set regular kernel stack */
141#define __ISTACK(name)	.L_ISTACK_ ## name
142#define IRECONCILE	.L_IRECONCILE_\name\()	/* Do RECONCILE_IRQ_STATE */
143#define IKUAP		.L_IKUAP_\name\()	/* Do KUAP lock */
144
145#define INT_DEFINE_BEGIN(n)						\
146.macro int_define_ ## n name
147
148#define INT_DEFINE_END(n)						\
149.endm ;									\
150int_define_ ## n n ;							\
151do_define_int n
152
153.macro do_define_int name
154	.ifndef IVEC
155		.error "IVEC not defined"
156	.endif
157	.ifndef IHSRR
158		IHSRR=0
159	.endif
160	.ifndef IHSRR_IF_HVMODE
161		IHSRR_IF_HVMODE=0
162	.endif
163	.ifndef IAREA
164		IAREA=PACA_EXGEN
165	.endif
166	.ifndef IVIRT
167		IVIRT=1
168	.endif
169	.ifndef IISIDE
170		IISIDE=0
171	.endif
172	.ifndef IDAR
173		IDAR=0
174	.endif
175	.ifndef IDSISR
176		IDSISR=0
177	.endif
178	.ifndef ISET_RI
179		ISET_RI=1
180	.endif
181	.ifndef IBRANCH_TO_COMMON
182		IBRANCH_TO_COMMON=1
183	.endif
184	.ifndef IREALMODE_COMMON
185		IREALMODE_COMMON=0
186	.else
187		.if ! IBRANCH_TO_COMMON
188			.error "IREALMODE_COMMON=1 but IBRANCH_TO_COMMON=0"
189		.endif
190	.endif
191	.ifndef IMASK
192		IMASK=0
193	.endif
194	.ifndef IKVM_SKIP
195		IKVM_SKIP=0
196	.endif
197	.ifndef IKVM_REAL
198		IKVM_REAL=0
199	.endif
200	.ifndef IKVM_VIRT
201		IKVM_VIRT=0
202	.endif
203	.ifndef ISTACK
204		ISTACK=1
205	.endif
206	.ifndef IRECONCILE
207		IRECONCILE=1
208	.endif
209	.ifndef IKUAP
210		IKUAP=1
211	.endif
212.endm
213
214#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
215#ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
216/*
217 * All interrupts which set HSRR registers, as well as SRESET and MCE and
218 * syscall when invoked with "sc 1" switch to MSR[HV]=1 (HVMODE) to be taken,
219 * so they all generally need to test whether they were taken in guest context.
220 *
221 * Note: SRESET and MCE may also be sent to the guest by the hypervisor, and be
222 * taken with MSR[HV]=0.
223 *
224 * Interrupts which set SRR registers (with the above exceptions) do not
225 * elevate to MSR[HV]=1 mode, though most can be taken when running with
226 * MSR[HV]=1  (e.g., bare metal kernel and userspace). So these interrupts do
227 * not need to test whether a guest is running because they get delivered to
228 * the guest directly, including nested HV KVM guests.
229 *
230 * The exception is PR KVM, where the guest runs with MSR[PR]=1 and the host
231 * runs with MSR[HV]=0, so the host takes all interrupts on behalf of the
232 * guest. PR KVM runs with LPCR[AIL]=0 which causes interrupts to always be
233 * delivered to the real-mode entry point, therefore such interrupts only test
234 * KVM in their real mode handlers, and only when PR KVM is possible.
235 *
236 * Interrupts that are taken in MSR[HV]=0 and escalate to MSR[HV]=1 are always
237 * delivered in real-mode when the MMU is in hash mode because the MMU
238 * registers are not set appropriately to translate host addresses. In nested
239 * radix mode these can be delivered in virt-mode as the host translations are
240 * used implicitly (see: effective LPID, effective PID).
241 */
242
243/*
244 * If an interrupt is taken while a guest is running, it is immediately routed
245 * to KVM to handle. If both HV and PR KVM arepossible, KVM interrupts go first
246 * to kvmppc_interrupt_hv, which handles the PR guest case.
247 */
248#define kvmppc_interrupt kvmppc_interrupt_hv
249#else
250#define kvmppc_interrupt kvmppc_interrupt_pr
251#endif
252
253.macro KVMTEST name
254	lbz	r10,HSTATE_IN_GUEST(r13)
255	cmpwi	r10,0
256	bne	\name\()_kvm
257.endm
258
259.macro GEN_KVM name
260	.balign IFETCH_ALIGN_BYTES
261\name\()_kvm:
262
263	.if IKVM_SKIP
264	cmpwi	r10,KVM_GUEST_MODE_SKIP
265	beq	89f
266	.else
267BEGIN_FTR_SECTION
268	ld	r10,IAREA+EX_CFAR(r13)
269	std	r10,HSTATE_CFAR(r13)
270END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
271	.endif
272
273	ld	r10,IAREA+EX_CTR(r13)
274	mtctr	r10
275BEGIN_FTR_SECTION
276	ld	r10,IAREA+EX_PPR(r13)
277	std	r10,HSTATE_PPR(r13)
278END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
279	ld	r11,IAREA+EX_R11(r13)
280	ld	r12,IAREA+EX_R12(r13)
281	std	r12,HSTATE_SCRATCH0(r13)
282	sldi	r12,r9,32
283	ld	r9,IAREA+EX_R9(r13)
284	ld	r10,IAREA+EX_R10(r13)
285	/* HSRR variants have the 0x2 bit added to their trap number */
286	.if IHSRR_IF_HVMODE
287	BEGIN_FTR_SECTION
288	ori	r12,r12,(IVEC + 0x2)
289	FTR_SECTION_ELSE
290	ori	r12,r12,(IVEC)
291	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
292	.elseif IHSRR
293	ori	r12,r12,(IVEC+ 0x2)
294	.else
295	ori	r12,r12,(IVEC)
296	.endif
297	b	kvmppc_interrupt
298
299	.if IKVM_SKIP
30089:	mtocrf	0x80,r9
301	ld	r10,IAREA+EX_CTR(r13)
302	mtctr	r10
303	ld	r9,IAREA+EX_R9(r13)
304	ld	r10,IAREA+EX_R10(r13)
305	ld	r11,IAREA+EX_R11(r13)
306	ld	r12,IAREA+EX_R12(r13)
307	.if IHSRR_IF_HVMODE
308	BEGIN_FTR_SECTION
309	b	kvmppc_skip_Hinterrupt
310	FTR_SECTION_ELSE
311	b	kvmppc_skip_interrupt
312	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
313	.elseif IHSRR
314	b	kvmppc_skip_Hinterrupt
315	.else
316	b	kvmppc_skip_interrupt
317	.endif
318	.endif
319.endm
320
321#else
322.macro KVMTEST name
323.endm
324.macro GEN_KVM name
325.endm
326#endif
327
328/*
329 * This is the BOOK3S interrupt entry code macro.
330 *
331 * This can result in one of several things happening:
332 * - Branch to the _common handler, relocated, in virtual mode.
333 *   These are normal interrupts (synchronous and asynchronous) handled by
334 *   the kernel.
335 * - Branch to KVM, relocated but real mode interrupts remain in real mode.
336 *   These occur when HSTATE_IN_GUEST is set. The interrupt may be caused by
337 *   / intended for host or guest kernel, but KVM must always be involved
338 *   because the machine state is set for guest execution.
339 * - Branch to the masked handler, unrelocated.
340 *   These occur when maskable asynchronous interrupts are taken with the
341 *   irq_soft_mask set.
342 * - Branch to an "early" handler in real mode but relocated.
343 *   This is done if early=1. MCE and HMI use these to handle errors in real
344 *   mode.
345 * - Fall through and continue executing in real, unrelocated mode.
346 *   This is done if early=2.
347 */
348
349.macro GEN_BRANCH_TO_COMMON name, virt
350	.if IREALMODE_COMMON
351	LOAD_HANDLER(r10, \name\()_common)
352	mtctr	r10
353	bctr
354	.else
355	.if \virt
356#ifndef CONFIG_RELOCATABLE
357	b	\name\()_common_virt
358#else
359	LOAD_HANDLER(r10, \name\()_common_virt)
360	mtctr	r10
361	bctr
362#endif
363	.else
364	LOAD_HANDLER(r10, \name\()_common_real)
365	mtctr	r10
366	bctr
367	.endif
368	.endif
369.endm
370
371.macro GEN_INT_ENTRY name, virt, ool=0
372	SET_SCRATCH0(r13)			/* save r13 */
373	GET_PACA(r13)
374	std	r9,IAREA+EX_R9(r13)		/* save r9 */
375BEGIN_FTR_SECTION
376	mfspr	r9,SPRN_PPR
377END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
378	HMT_MEDIUM
379	std	r10,IAREA+EX_R10(r13)		/* save r10 - r12 */
380BEGIN_FTR_SECTION
381	mfspr	r10,SPRN_CFAR
382END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
383	.if \ool
384	.if !\virt
385	b	tramp_real_\name
386	.pushsection .text
387	TRAMP_REAL_BEGIN(tramp_real_\name)
388	.else
389	b	tramp_virt_\name
390	.pushsection .text
391	TRAMP_VIRT_BEGIN(tramp_virt_\name)
392	.endif
393	.endif
394
395BEGIN_FTR_SECTION
396	std	r9,IAREA+EX_PPR(r13)
397END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
398BEGIN_FTR_SECTION
399	std	r10,IAREA+EX_CFAR(r13)
400END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
401	INTERRUPT_TO_KERNEL
402	mfctr	r10
403	std	r10,IAREA+EX_CTR(r13)
404	mfcr	r9
405	std	r11,IAREA+EX_R11(r13)
406	std	r12,IAREA+EX_R12(r13)
407
408	/*
409	 * DAR/DSISR, SCRATCH0 must be read before setting MSR[RI],
410	 * because a d-side MCE will clobber those registers so is
411	 * not recoverable if they are live.
412	 */
413	GET_SCRATCH0(r10)
414	std	r10,IAREA+EX_R13(r13)
415	.if IDAR && !IISIDE
416	.if IHSRR
417	mfspr	r10,SPRN_HDAR
418	.else
419	mfspr	r10,SPRN_DAR
420	.endif
421	std	r10,IAREA+EX_DAR(r13)
422	.endif
423	.if IDSISR && !IISIDE
424	.if IHSRR
425	mfspr	r10,SPRN_HDSISR
426	.else
427	mfspr	r10,SPRN_DSISR
428	.endif
429	stw	r10,IAREA+EX_DSISR(r13)
430	.endif
431
432	.if IHSRR_IF_HVMODE
433	BEGIN_FTR_SECTION
434	mfspr	r11,SPRN_HSRR0		/* save HSRR0 */
435	mfspr	r12,SPRN_HSRR1		/* and HSRR1 */
436	FTR_SECTION_ELSE
437	mfspr	r11,SPRN_SRR0		/* save SRR0 */
438	mfspr	r12,SPRN_SRR1		/* and SRR1 */
439	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
440	.elseif IHSRR
441	mfspr	r11,SPRN_HSRR0		/* save HSRR0 */
442	mfspr	r12,SPRN_HSRR1		/* and HSRR1 */
443	.else
444	mfspr	r11,SPRN_SRR0		/* save SRR0 */
445	mfspr	r12,SPRN_SRR1		/* and SRR1 */
446	.endif
447
448	.if IBRANCH_TO_COMMON
449	GEN_BRANCH_TO_COMMON \name \virt
450	.endif
451
452	.if \ool
453	.popsection
454	.endif
455.endm
456
457/*
458 * __GEN_COMMON_ENTRY is required to receive the branch from interrupt
459 * entry, except in the case of the real-mode handlers which require
460 * __GEN_REALMODE_COMMON_ENTRY.
461 *
462 * This switches to virtual mode and sets MSR[RI].
463 */
464.macro __GEN_COMMON_ENTRY name
465DEFINE_FIXED_SYMBOL(\name\()_common_real)
466\name\()_common_real:
467	.if IKVM_REAL
468		KVMTEST \name
469	.endif
470
471	ld	r10,PACAKMSR(r13)	/* get MSR value for kernel */
472	/* MSR[RI] is clear iff using SRR regs */
473	.if IHSRR == EXC_HV_OR_STD
474	BEGIN_FTR_SECTION
475	xori	r10,r10,MSR_RI
476	END_FTR_SECTION_IFCLR(CPU_FTR_HVMODE)
477	.elseif ! IHSRR
478	xori	r10,r10,MSR_RI
479	.endif
480	mtmsrd	r10
481
482	.if IVIRT
483	.if IKVM_VIRT
484	b	1f /* skip the virt test coming from real */
485	.endif
486
487	.balign IFETCH_ALIGN_BYTES
488DEFINE_FIXED_SYMBOL(\name\()_common_virt)
489\name\()_common_virt:
490	.if IKVM_VIRT
491		KVMTEST \name
4921:
493	.endif
494	.endif /* IVIRT */
495.endm
496
497/*
498 * Don't switch to virt mode. Used for early MCE and HMI handlers that
499 * want to run in real mode.
500 */
501.macro __GEN_REALMODE_COMMON_ENTRY name
502DEFINE_FIXED_SYMBOL(\name\()_common_real)
503\name\()_common_real:
504	.if IKVM_REAL
505		KVMTEST \name
506	.endif
507.endm
508
509.macro __GEN_COMMON_BODY name
510	.if IMASK
511		.if ! ISTACK
512		.error "No support for masked interrupt to use custom stack"
513		.endif
514
515		/* If coming from user, skip soft-mask tests. */
516		andi.	r10,r12,MSR_PR
517		bne	2f
518
519		/* Kernel code running below __end_interrupts is implicitly
520		 * soft-masked */
521		LOAD_HANDLER(r10, __end_interrupts)
522		cmpld	r11,r10
523		li	r10,IMASK
524		blt-	1f
525
526		/* Test the soft mask state against our interrupt's bit */
527		lbz	r10,PACAIRQSOFTMASK(r13)
5281:		andi.	r10,r10,IMASK
529		/* Associate vector numbers with bits in paca->irq_happened */
530		.if IVEC == 0x500 || IVEC == 0xea0
531		li	r10,PACA_IRQ_EE
532		.elseif IVEC == 0x900
533		li	r10,PACA_IRQ_DEC
534		.elseif IVEC == 0xa00 || IVEC == 0xe80
535		li	r10,PACA_IRQ_DBELL
536		.elseif IVEC == 0xe60
537		li	r10,PACA_IRQ_HMI
538		.elseif IVEC == 0xf00
539		li	r10,PACA_IRQ_PMI
540		.else
541		.abort "Bad maskable vector"
542		.endif
543
544		.if IHSRR_IF_HVMODE
545		BEGIN_FTR_SECTION
546		bne	masked_Hinterrupt
547		FTR_SECTION_ELSE
548		bne	masked_interrupt
549		ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
550		.elseif IHSRR
551		bne	masked_Hinterrupt
552		.else
553		bne	masked_interrupt
554		.endif
555	.endif
556
557	.if ISTACK
558	andi.	r10,r12,MSR_PR		/* See if coming from user	*/
5592:	mr	r10,r1			/* Save r1			*/
560	subi	r1,r1,INT_FRAME_SIZE	/* alloc frame on kernel stack	*/
561	beq-	100f
562	ld	r1,PACAKSAVE(r13)	/* kernel stack to use		*/
563100:	tdgei	r1,-INT_FRAME_SIZE	/* trap if r1 is in userspace	*/
564	EMIT_BUG_ENTRY 100b,__FILE__,__LINE__,0
565	.endif
566
567	std	r9,_CCR(r1)		/* save CR in stackframe	*/
568	std	r11,_NIP(r1)		/* save SRR0 in stackframe	*/
569	std	r12,_MSR(r1)		/* save SRR1 in stackframe	*/
570	std	r10,0(r1)		/* make stack chain pointer	*/
571	std	r0,GPR0(r1)		/* save r0 in stackframe	*/
572	std	r10,GPR1(r1)		/* save r1 in stackframe	*/
573
574	.if ISET_RI
575	li	r10,MSR_RI
576	mtmsrd	r10,1			/* Set MSR_RI */
577	.endif
578
579	.if ISTACK
580	.if IKUAP
581	kuap_save_amr_and_lock r9, r10, cr1, cr0
582	.endif
583	beq	101f			/* if from kernel mode		*/
584	ACCOUNT_CPU_USER_ENTRY(r13, r9, r10)
585BEGIN_FTR_SECTION
586	ld	r9,IAREA+EX_PPR(r13)	/* Read PPR from paca		*/
587	std	r9,_PPR(r1)
588END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
589101:
590	.else
591	.if IKUAP
592	kuap_save_amr_and_lock r9, r10, cr1
593	.endif
594	.endif
595
596	/* Save original regs values from save area to stack frame. */
597	ld	r9,IAREA+EX_R9(r13)	/* move r9, r10 to stackframe	*/
598	ld	r10,IAREA+EX_R10(r13)
599	std	r9,GPR9(r1)
600	std	r10,GPR10(r1)
601	ld	r9,IAREA+EX_R11(r13)	/* move r11 - r13 to stackframe	*/
602	ld	r10,IAREA+EX_R12(r13)
603	ld	r11,IAREA+EX_R13(r13)
604	std	r9,GPR11(r1)
605	std	r10,GPR12(r1)
606	std	r11,GPR13(r1)
607
608	SAVE_NVGPRS(r1)
609
610	.if IDAR
611	.if IISIDE
612	ld	r10,_NIP(r1)
613	.else
614	ld	r10,IAREA+EX_DAR(r13)
615	.endif
616	std	r10,_DAR(r1)
617	.endif
618
619	.if IDSISR
620	.if IISIDE
621	ld	r10,_MSR(r1)
622	lis	r11,DSISR_SRR1_MATCH_64S@h
623	and	r10,r10,r11
624	.else
625	lwz	r10,IAREA+EX_DSISR(r13)
626	.endif
627	std	r10,_DSISR(r1)
628	.endif
629
630BEGIN_FTR_SECTION
631	ld	r10,IAREA+EX_CFAR(r13)
632	std	r10,ORIG_GPR3(r1)
633END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
634	ld	r10,IAREA+EX_CTR(r13)
635	std	r10,_CTR(r1)
636	std	r2,GPR2(r1)		/* save r2 in stackframe	*/
637	SAVE_4GPRS(3, r1)		/* save r3 - r6 in stackframe   */
638	SAVE_2GPRS(7, r1)		/* save r7, r8 in stackframe	*/
639	mflr	r9			/* Get LR, later save to stack	*/
640	ld	r2,PACATOC(r13)		/* get kernel TOC into r2	*/
641	std	r9,_LINK(r1)
642	lbz	r10,PACAIRQSOFTMASK(r13)
643	mfspr	r11,SPRN_XER		/* save XER in stackframe	*/
644	std	r10,SOFTE(r1)
645	std	r11,_XER(r1)
646	li	r9,IVEC
647	std	r9,_TRAP(r1)		/* set trap number		*/
648	li	r10,0
649	ld	r11,exception_marker@toc(r2)
650	std	r10,RESULT(r1)		/* clear regs->result		*/
651	std	r11,STACK_FRAME_OVERHEAD-16(r1) /* mark the frame	*/
652
653	.if ISTACK
654	ACCOUNT_STOLEN_TIME
655	.endif
656
657	.if IRECONCILE
658	RECONCILE_IRQ_STATE(r10, r11)
659	.endif
660.endm
661
662/*
663 * On entry r13 points to the paca, r9-r13 are saved in the paca,
664 * r9 contains the saved CR, r11 and r12 contain the saved SRR0 and
665 * SRR1, and relocation is on.
666 *
667 * If stack=0, then the stack is already set in r1, and r1 is saved in r10.
668 * PPR save and CPU accounting is not done for the !stack case (XXX why not?)
669 */
670.macro GEN_COMMON name
671	__GEN_COMMON_ENTRY \name
672	__GEN_COMMON_BODY \name
673.endm
674
675/*
676 * Restore all registers including H/SRR0/1 saved in a stack frame of a
677 * standard exception.
678 */
679.macro EXCEPTION_RESTORE_REGS hsrr=0
680	/* Move original SRR0 and SRR1 into the respective regs */
681	ld	r9,_MSR(r1)
682	.if \hsrr
683	mtspr	SPRN_HSRR1,r9
684	.else
685	mtspr	SPRN_SRR1,r9
686	.endif
687	ld	r9,_NIP(r1)
688	.if \hsrr
689	mtspr	SPRN_HSRR0,r9
690	.else
691	mtspr	SPRN_SRR0,r9
692	.endif
693	ld	r9,_CTR(r1)
694	mtctr	r9
695	ld	r9,_XER(r1)
696	mtxer	r9
697	ld	r9,_LINK(r1)
698	mtlr	r9
699	ld	r9,_CCR(r1)
700	mtcr	r9
701	REST_8GPRS(2, r1)
702	REST_4GPRS(10, r1)
703	REST_GPR(0, r1)
704	/* restore original r1. */
705	ld	r1,GPR1(r1)
706.endm
707
708#define RUNLATCH_ON				\
709BEGIN_FTR_SECTION				\
710	ld	r3, PACA_THREAD_INFO(r13);	\
711	ld	r4,TI_LOCAL_FLAGS(r3);		\
712	andi.	r0,r4,_TLF_RUNLATCH;		\
713	beql	ppc64_runlatch_on_trampoline;	\
714END_FTR_SECTION_IFSET(CPU_FTR_CTRL)
715
716/*
717 * When the idle code in power4_idle puts the CPU into NAP mode,
718 * it has to do so in a loop, and relies on the external interrupt
719 * and decrementer interrupt entry code to get it out of the loop.
720 * It sets the _TLF_NAPPING bit in current_thread_info()->local_flags
721 * to signal that it is in the loop and needs help to get out.
722 */
723#ifdef CONFIG_PPC_970_NAP
724#define FINISH_NAP				\
725BEGIN_FTR_SECTION				\
726	ld	r11, PACA_THREAD_INFO(r13);	\
727	ld	r9,TI_LOCAL_FLAGS(r11);		\
728	andi.	r10,r9,_TLF_NAPPING;		\
729	bnel	power4_fixup_nap;		\
730END_FTR_SECTION_IFSET(CPU_FTR_CAN_NAP)
731#else
732#define FINISH_NAP
733#endif
734
735/*
736 * There are a few constraints to be concerned with.
737 * - Real mode exceptions code/data must be located at their physical location.
738 * - Virtual mode exceptions must be mapped at their 0xc000... location.
739 * - Fixed location code must not call directly beyond the __end_interrupts
740 *   area when built with CONFIG_RELOCATABLE. LOAD_HANDLER / bctr sequence
741 *   must be used.
742 * - LOAD_HANDLER targets must be within first 64K of physical 0 /
743 *   virtual 0xc00...
744 * - Conditional branch targets must be within +/-32K of caller.
745 *
746 * "Virtual exceptions" run with relocation on (MSR_IR=1, MSR_DR=1), and
747 * therefore don't have to run in physically located code or rfid to
748 * virtual mode kernel code. However on relocatable kernels they do have
749 * to branch to KERNELBASE offset because the rest of the kernel (outside
750 * the exception vectors) may be located elsewhere.
751 *
752 * Virtual exceptions correspond with physical, except their entry points
753 * are offset by 0xc000000000000000 and also tend to get an added 0x4000
754 * offset applied. Virtual exceptions are enabled with the Alternate
755 * Interrupt Location (AIL) bit set in the LPCR. However this does not
756 * guarantee they will be delivered virtually. Some conditions (see the ISA)
757 * cause exceptions to be delivered in real mode.
758 *
759 * The scv instructions are a special case. They get a 0x3000 offset applied.
760 * scv exceptions have unique reentrancy properties, see below.
761 *
762 * It's impossible to receive interrupts below 0x300 via AIL.
763 *
764 * KVM: None of the virtual exceptions are from the guest. Anything that
765 * escalated to HV=1 from HV=0 is delivered via real mode handlers.
766 *
767 *
768 * We layout physical memory as follows:
769 * 0x0000 - 0x00ff : Secondary processor spin code
770 * 0x0100 - 0x18ff : Real mode pSeries interrupt vectors
771 * 0x1900 - 0x2fff : Real mode trampolines
772 * 0x3000 - 0x58ff : Relon (IR=1,DR=1) mode pSeries interrupt vectors
773 * 0x5900 - 0x6fff : Relon mode trampolines
774 * 0x7000 - 0x7fff : FWNMI data area
775 * 0x8000 -   .... : Common interrupt handlers, remaining early
776 *                   setup code, rest of kernel.
777 *
778 * We could reclaim 0x4000-0x42ff for real mode trampolines if the space
779 * is necessary. Until then it's more consistent to explicitly put VIRT_NONE
780 * vectors there.
781 */
782OPEN_FIXED_SECTION(real_vectors,        0x0100, 0x1900)
783OPEN_FIXED_SECTION(real_trampolines,    0x1900, 0x3000)
784OPEN_FIXED_SECTION(virt_vectors,        0x3000, 0x5900)
785OPEN_FIXED_SECTION(virt_trampolines,    0x5900, 0x7000)
786
787#ifdef CONFIG_PPC_POWERNV
788	.globl start_real_trampolines
789	.globl end_real_trampolines
790	.globl start_virt_trampolines
791	.globl end_virt_trampolines
792#endif
793
794#if defined(CONFIG_PPC_PSERIES) || defined(CONFIG_PPC_POWERNV)
795/*
796 * Data area reserved for FWNMI option.
797 * This address (0x7000) is fixed by the RPA.
798 * pseries and powernv need to keep the whole page from
799 * 0x7000 to 0x8000 free for use by the firmware
800 */
801ZERO_FIXED_SECTION(fwnmi_page,          0x7000, 0x8000)
802OPEN_TEXT_SECTION(0x8000)
803#else
804OPEN_TEXT_SECTION(0x7000)
805#endif
806
807USE_FIXED_SECTION(real_vectors)
808
809/*
810 * This is the start of the interrupt handlers for pSeries
811 * This code runs with relocation off.
812 * Code from here to __end_interrupts gets copied down to real
813 * address 0x100 when we are running a relocatable kernel.
814 * Therefore any relative branches in this section must only
815 * branch to labels in this section.
816 */
817	.globl __start_interrupts
818__start_interrupts:
819
820/**
821 * Interrupt 0x3000 - System Call Vectored Interrupt (syscall).
822 * This is a synchronous interrupt invoked with the "scv" instruction. The
823 * system call does not alter the HV bit, so it is directed to the OS.
824 *
825 * Handling:
826 * scv instructions enter the kernel without changing EE, RI, ME, or HV.
827 * In particular, this means we can take a maskable interrupt at any point
828 * in the scv handler, which is unlike any other interrupt. This is solved
829 * by treating the instruction addresses below __end_interrupts as being
830 * soft-masked.
831 *
832 * AIL-0 mode scv exceptions go to 0x17000-0x17fff, but we set AIL-3 and
833 * ensure scv is never executed with relocation off, which means AIL-0
834 * should never happen.
835 *
836 * Before leaving the below __end_interrupts text, at least of the following
837 * must be true:
838 * - MSR[PR]=1 (i.e., return to userspace)
839 * - MSR_EE|MSR_RI is set (no reentrant exceptions)
840 * - Standard kernel environment is set up (stack, paca, etc)
841 *
842 * Call convention:
843 *
844 * syscall register convention is in Documentation/powerpc/syscall64-abi.rst
845 */
846EXC_VIRT_BEGIN(system_call_vectored, 0x3000, 0x1000)
847	/* SCV 0 */
848	mr	r9,r13
849	GET_PACA(r13)
850	mflr	r11
851	mfctr	r12
852	li	r10,IRQS_ALL_DISABLED
853	stb	r10,PACAIRQSOFTMASK(r13)
854#ifdef CONFIG_RELOCATABLE
855	b	system_call_vectored_tramp
856#else
857	b	system_call_vectored_common
858#endif
859	nop
860
861	/* SCV 1 - 127 */
862	.rept	127
863	mr	r9,r13
864	GET_PACA(r13)
865	mflr	r11
866	mfctr	r12
867	li	r10,IRQS_ALL_DISABLED
868	stb	r10,PACAIRQSOFTMASK(r13)
869	li	r0,-1 /* cause failure */
870#ifdef CONFIG_RELOCATABLE
871	b	system_call_vectored_sigill_tramp
872#else
873	b	system_call_vectored_sigill
874#endif
875	.endr
876EXC_VIRT_END(system_call_vectored, 0x3000, 0x1000)
877
878#ifdef CONFIG_RELOCATABLE
879TRAMP_VIRT_BEGIN(system_call_vectored_tramp)
880	__LOAD_HANDLER(r10, system_call_vectored_common)
881	mtctr	r10
882	bctr
883
884TRAMP_VIRT_BEGIN(system_call_vectored_sigill_tramp)
885	__LOAD_HANDLER(r10, system_call_vectored_sigill)
886	mtctr	r10
887	bctr
888#endif
889
890
891/* No virt vectors corresponding with 0x0..0x100 */
892EXC_VIRT_NONE(0x4000, 0x100)
893
894
895/**
896 * Interrupt 0x100 - System Reset Interrupt (SRESET aka NMI).
897 * This is a non-maskable, asynchronous interrupt always taken in real-mode.
898 * It is caused by:
899 * - Wake from power-saving state, on powernv.
900 * - An NMI from another CPU, triggered by firmware or hypercall.
901 * - As crash/debug signal injected from BMC, firmware or hypervisor.
902 *
903 * Handling:
904 * Power-save wakeup is the only performance critical path, so this is
905 * determined quickly as possible first. In this case volatile registers
906 * can be discarded and SPRs like CFAR don't need to be read.
907 *
908 * If not a powersave wakeup, then it's run as a regular interrupt, however
909 * it uses its own stack and PACA save area to preserve the regular kernel
910 * environment for debugging.
911 *
912 * This interrupt is not maskable, so triggering it when MSR[RI] is clear,
913 * or SCRATCH0 is in use, etc. may cause a crash. It's also not entirely
914 * correct to switch to virtual mode to run the regular interrupt handler
915 * because it might be interrupted when the MMU is in a bad state (e.g., SLB
916 * is clear).
917 *
918 * FWNMI:
919 * PAPR specifies a "fwnmi" facility which sends the sreset to a different
920 * entry point with a different register set up. Some hypervisors will
921 * send the sreset to 0x100 in the guest if it is not fwnmi capable.
922 *
923 * KVM:
924 * Unlike most SRR interrupts, this may be taken by the host while executing
925 * in a guest, so a KVM test is required. KVM will pull the CPU out of guest
926 * mode and then raise the sreset.
927 */
928INT_DEFINE_BEGIN(system_reset)
929	IVEC=0x100
930	IAREA=PACA_EXNMI
931	IVIRT=0 /* no virt entry point */
932	/*
933	 * MSR_RI is not enabled, because PACA_EXNMI and nmi stack is
934	 * being used, so a nested NMI exception would corrupt it.
935	 */
936	ISET_RI=0
937	ISTACK=0
938	IRECONCILE=0
939	IKVM_REAL=1
940INT_DEFINE_END(system_reset)
941
942EXC_REAL_BEGIN(system_reset, 0x100, 0x100)
943#ifdef CONFIG_PPC_P7_NAP
944	/*
945	 * If running native on arch 2.06 or later, check if we are waking up
946	 * from nap/sleep/winkle, and branch to idle handler. This tests SRR1
947	 * bits 46:47. A non-0 value indicates that we are coming from a power
948	 * saving state. The idle wakeup handler initially runs in real mode,
949	 * but we branch to the 0xc000... address so we can turn on relocation
950	 * with mtmsrd later, after SPRs are restored.
951	 *
952	 * Careful to minimise cost for the fast path (idle wakeup) while
953	 * also avoiding clobbering CFAR for the debug path (non-idle).
954	 *
955	 * For the idle wake case volatile registers can be clobbered, which
956	 * is why we use those initially. If it turns out to not be an idle
957	 * wake, carefully put everything back the way it was, so we can use
958	 * common exception macros to handle it.
959	 */
960BEGIN_FTR_SECTION
961	SET_SCRATCH0(r13)
962	GET_PACA(r13)
963	std	r3,PACA_EXNMI+0*8(r13)
964	std	r4,PACA_EXNMI+1*8(r13)
965	std	r5,PACA_EXNMI+2*8(r13)
966	mfspr	r3,SPRN_SRR1
967	mfocrf	r4,0x80
968	rlwinm.	r5,r3,47-31,30,31
969	bne+	system_reset_idle_wake
970	/* Not powersave wakeup. Restore regs for regular interrupt handler. */
971	mtocrf	0x80,r4
972	ld	r3,PACA_EXNMI+0*8(r13)
973	ld	r4,PACA_EXNMI+1*8(r13)
974	ld	r5,PACA_EXNMI+2*8(r13)
975	GET_SCRATCH0(r13)
976END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
977#endif
978
979	GEN_INT_ENTRY system_reset, virt=0
980	/*
981	 * In theory, we should not enable relocation here if it was disabled
982	 * in SRR1, because the MMU may not be configured to support it (e.g.,
983	 * SLB may have been cleared). In practice, there should only be a few
984	 * small windows where that's the case, and sreset is considered to
985	 * be dangerous anyway.
986	 */
987EXC_REAL_END(system_reset, 0x100, 0x100)
988EXC_VIRT_NONE(0x4100, 0x100)
989
990#ifdef CONFIG_PPC_P7_NAP
991TRAMP_REAL_BEGIN(system_reset_idle_wake)
992	/* We are waking up from idle, so may clobber any volatile register */
993	cmpwi	cr1,r5,2
994	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
995	BRANCH_TO_C000(r12, DOTSYM(idle_return_gpr_loss))
996#endif
997
998#ifdef CONFIG_PPC_PSERIES
999/*
1000 * Vectors for the FWNMI option.  Share common code.
1001 */
1002TRAMP_REAL_BEGIN(system_reset_fwnmi)
1003	/* XXX: fwnmi guest could run a nested/PR guest, so why no test?  */
1004	__IKVM_REAL(system_reset)=0
1005	GEN_INT_ENTRY system_reset, virt=0
1006
1007#endif /* CONFIG_PPC_PSERIES */
1008
1009EXC_COMMON_BEGIN(system_reset_common)
1010	__GEN_COMMON_ENTRY system_reset
1011	/*
1012	 * Increment paca->in_nmi then enable MSR_RI. SLB or MCE will be able
1013	 * to recover, but nested NMI will notice in_nmi and not recover
1014	 * because of the use of the NMI stack. in_nmi reentrancy is tested in
1015	 * system_reset_exception.
1016	 */
1017	lhz	r10,PACA_IN_NMI(r13)
1018	addi	r10,r10,1
1019	sth	r10,PACA_IN_NMI(r13)
1020	li	r10,MSR_RI
1021	mtmsrd 	r10,1
1022
1023	mr	r10,r1
1024	ld	r1,PACA_NMI_EMERG_SP(r13)
1025	subi	r1,r1,INT_FRAME_SIZE
1026	__GEN_COMMON_BODY system_reset
1027	/*
1028	 * Set IRQS_ALL_DISABLED unconditionally so irqs_disabled() does
1029	 * the right thing. We do not want to reconcile because that goes
1030	 * through irq tracing which we don't want in NMI.
1031	 *
1032	 * Save PACAIRQHAPPENED to RESULT (otherwise unused), and set HARD_DIS
1033	 * as we are running with MSR[EE]=0.
1034	 */
1035	li	r10,IRQS_ALL_DISABLED
1036	stb	r10,PACAIRQSOFTMASK(r13)
1037	lbz	r10,PACAIRQHAPPENED(r13)
1038	std	r10,RESULT(r1)
1039	ori	r10,r10,PACA_IRQ_HARD_DIS
1040	stb	r10,PACAIRQHAPPENED(r13)
1041
1042	addi	r3,r1,STACK_FRAME_OVERHEAD
1043	bl	system_reset_exception
1044
1045	/* Clear MSR_RI before setting SRR0 and SRR1. */
1046	li	r9,0
1047	mtmsrd	r9,1
1048
1049	/*
1050	 * MSR_RI is clear, now we can decrement paca->in_nmi.
1051	 */
1052	lhz	r10,PACA_IN_NMI(r13)
1053	subi	r10,r10,1
1054	sth	r10,PACA_IN_NMI(r13)
1055
1056	/*
1057	 * Restore soft mask settings.
1058	 */
1059	ld	r10,RESULT(r1)
1060	stb	r10,PACAIRQHAPPENED(r13)
1061	ld	r10,SOFTE(r1)
1062	stb	r10,PACAIRQSOFTMASK(r13)
1063
1064	kuap_restore_amr r9, r10
1065	EXCEPTION_RESTORE_REGS
1066	RFI_TO_USER_OR_KERNEL
1067
1068	GEN_KVM system_reset
1069
1070
1071/**
1072 * Interrupt 0x200 - Machine Check Interrupt (MCE).
1073 * This is a non-maskable interrupt always taken in real-mode. It can be
1074 * synchronous or asynchronous, caused by hardware or software, and it may be
1075 * taken in a power-saving state.
1076 *
1077 * Handling:
1078 * Similarly to system reset, this uses its own stack and PACA save area,
1079 * the difference is re-entrancy is allowed on the machine check stack.
1080 *
1081 * machine_check_early is run in real mode, and carefully decodes the
1082 * machine check and tries to handle it (e.g., flush the SLB if there was an
1083 * error detected there), determines if it was recoverable and logs the
1084 * event.
1085 *
1086 * This early code does not "reconcile" irq soft-mask state like SRESET or
1087 * regular interrupts do, so irqs_disabled() among other things may not work
1088 * properly (irq disable/enable already doesn't work because irq tracing can
1089 * not work in real mode).
1090 *
1091 * Then, depending on the execution context when the interrupt is taken, there
1092 * are 3 main actions:
1093 * - Executing in kernel mode. The event is queued with irq_work, which means
1094 *   it is handled when it is next safe to do so (i.e., the kernel has enabled
1095 *   interrupts), which could be immediately when the interrupt returns. This
1096 *   avoids nasty issues like switching to virtual mode when the MMU is in a
1097 *   bad state, or when executing OPAL code. (SRESET is exposed to such issues,
1098 *   but it has different priorities). Check to see if the CPU was in power
1099 *   save, and return via the wake up code if it was.
1100 *
1101 * - Executing in user mode. machine_check_exception is run like a normal
1102 *   interrupt handler, which processes the data generated by the early handler.
1103 *
1104 * - Executing in guest mode. The interrupt is run with its KVM test, and
1105 *   branches to KVM to deal with. KVM may queue the event for the host
1106 *   to report later.
1107 *
1108 * This interrupt is not maskable, so if it triggers when MSR[RI] is clear,
1109 * or SCRATCH0 is in use, it may cause a crash.
1110 *
1111 * KVM:
1112 * See SRESET.
1113 */
1114INT_DEFINE_BEGIN(machine_check_early)
1115	IVEC=0x200
1116	IAREA=PACA_EXMC
1117	IVIRT=0 /* no virt entry point */
1118	IREALMODE_COMMON=1
1119	/*
1120	 * MSR_RI is not enabled, because PACA_EXMC is being used, so a
1121	 * nested machine check corrupts it. machine_check_common enables
1122	 * MSR_RI.
1123	 */
1124	ISET_RI=0
1125	ISTACK=0
1126	IDAR=1
1127	IDSISR=1
1128	IRECONCILE=0
1129	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
1130INT_DEFINE_END(machine_check_early)
1131
1132INT_DEFINE_BEGIN(machine_check)
1133	IVEC=0x200
1134	IAREA=PACA_EXMC
1135	IVIRT=0 /* no virt entry point */
1136	ISET_RI=0
1137	IDAR=1
1138	IDSISR=1
1139	IKVM_SKIP=1
1140	IKVM_REAL=1
1141INT_DEFINE_END(machine_check)
1142
1143EXC_REAL_BEGIN(machine_check, 0x200, 0x100)
1144	GEN_INT_ENTRY machine_check_early, virt=0
1145EXC_REAL_END(machine_check, 0x200, 0x100)
1146EXC_VIRT_NONE(0x4200, 0x100)
1147
1148#ifdef CONFIG_PPC_PSERIES
1149TRAMP_REAL_BEGIN(machine_check_fwnmi)
1150	/* See comment at machine_check exception, don't turn on RI */
1151	GEN_INT_ENTRY machine_check_early, virt=0
1152#endif
1153
1154#define MACHINE_CHECK_HANDLER_WINDUP			\
1155	/* Clear MSR_RI before setting SRR0 and SRR1. */\
1156	li	r9,0;					\
1157	mtmsrd	r9,1;		/* Clear MSR_RI */	\
1158	/* Decrement paca->in_mce now RI is clear. */	\
1159	lhz	r12,PACA_IN_MCE(r13);			\
1160	subi	r12,r12,1;				\
1161	sth	r12,PACA_IN_MCE(r13);			\
1162	EXCEPTION_RESTORE_REGS
1163
1164EXC_COMMON_BEGIN(machine_check_early_common)
1165	__GEN_REALMODE_COMMON_ENTRY machine_check_early
1166
1167	/*
1168	 * Switch to mc_emergency stack and handle re-entrancy (we limit
1169	 * the nested MCE upto level 4 to avoid stack overflow).
1170	 * Save MCE registers srr1, srr0, dar and dsisr and then set ME=1
1171	 *
1172	 * We use paca->in_mce to check whether this is the first entry or
1173	 * nested machine check. We increment paca->in_mce to track nested
1174	 * machine checks.
1175	 *
1176	 * If this is the first entry then set stack pointer to
1177	 * paca->mc_emergency_sp, otherwise r1 is already pointing to
1178	 * stack frame on mc_emergency stack.
1179	 *
1180	 * NOTE: We are here with MSR_ME=0 (off), which means we risk a
1181	 * checkstop if we get another machine check exception before we do
1182	 * rfid with MSR_ME=1.
1183	 *
1184	 * This interrupt can wake directly from idle. If that is the case,
1185	 * the machine check is handled then the idle wakeup code is called
1186	 * to restore state.
1187	 */
1188	lhz	r10,PACA_IN_MCE(r13)
1189	cmpwi	r10,0			/* Are we in nested machine check */
1190	cmpwi	cr1,r10,MAX_MCE_DEPTH	/* Are we at maximum nesting */
1191	addi	r10,r10,1		/* increment paca->in_mce */
1192	sth	r10,PACA_IN_MCE(r13)
1193
1194	mr	r10,r1			/* Save r1 */
1195	bne	1f
1196	/* First machine check entry */
1197	ld	r1,PACAMCEMERGSP(r13)	/* Use MC emergency stack */
11981:	/* Limit nested MCE to level 4 to avoid stack overflow */
1199	bgt	cr1,unrecoverable_mce	/* Check if we hit limit of 4 */
1200	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame */
1201
1202	__GEN_COMMON_BODY machine_check_early
1203
1204BEGIN_FTR_SECTION
1205	bl	enable_machine_check
1206END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1207	li	r10,MSR_RI
1208	mtmsrd	r10,1
1209
1210	/*
1211	 * Set IRQS_ALL_DISABLED and save PACAIRQHAPPENED (see
1212	 * system_reset_common)
1213	 */
1214	li	r10,IRQS_ALL_DISABLED
1215	stb	r10,PACAIRQSOFTMASK(r13)
1216	lbz	r10,PACAIRQHAPPENED(r13)
1217	std	r10,RESULT(r1)
1218	ori	r10,r10,PACA_IRQ_HARD_DIS
1219	stb	r10,PACAIRQHAPPENED(r13)
1220
1221	addi	r3,r1,STACK_FRAME_OVERHEAD
1222	bl	machine_check_early
1223	std	r3,RESULT(r1)	/* Save result */
1224	ld	r12,_MSR(r1)
1225
1226	/*
1227	 * Restore soft mask settings.
1228	 */
1229	ld	r10,RESULT(r1)
1230	stb	r10,PACAIRQHAPPENED(r13)
1231	ld	r10,SOFTE(r1)
1232	stb	r10,PACAIRQSOFTMASK(r13)
1233
1234#ifdef CONFIG_PPC_P7_NAP
1235	/*
1236	 * Check if thread was in power saving mode. We come here when any
1237	 * of the following is true:
1238	 * a. thread wasn't in power saving mode
1239	 * b. thread was in power saving mode with no state loss,
1240	 *    supervisor state loss or hypervisor state loss.
1241	 *
1242	 * Go back to nap/sleep/winkle mode again if (b) is true.
1243	 */
1244BEGIN_FTR_SECTION
1245	rlwinm.	r11,r12,47-31,30,31
1246	bne	machine_check_idle_common
1247END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
1248#endif
1249
1250#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1251	/*
1252	 * Check if we are coming from guest. If yes, then run the normal
1253	 * exception handler which will take the
1254	 * machine_check_kvm->kvmppc_interrupt branch to deliver the MC event
1255	 * to guest.
1256	 */
1257	lbz	r11,HSTATE_IN_GUEST(r13)
1258	cmpwi	r11,0			/* Check if coming from guest */
1259	bne	mce_deliver		/* continue if we are. */
1260#endif
1261
1262	/*
1263	 * Check if we are coming from userspace. If yes, then run the normal
1264	 * exception handler which will deliver the MC event to this kernel.
1265	 */
1266	andi.	r11,r12,MSR_PR		/* See if coming from user. */
1267	bne	mce_deliver		/* continue in V mode if we are. */
1268
1269	/*
1270	 * At this point we are coming from kernel context.
1271	 * Queue up the MCE event and return from the interrupt.
1272	 * But before that, check if this is an un-recoverable exception.
1273	 * If yes, then stay on emergency stack and panic.
1274	 */
1275	andi.	r11,r12,MSR_RI
1276	beq	unrecoverable_mce
1277
1278	/*
1279	 * Check if we have successfully handled/recovered from error, if not
1280	 * then stay on emergency stack and panic.
1281	 */
1282	ld	r3,RESULT(r1)	/* Load result */
1283	cmpdi	r3,0		/* see if we handled MCE successfully */
1284	beq	unrecoverable_mce /* if !handled then panic */
1285
1286	/*
1287	 * Return from MC interrupt.
1288	 * Queue up the MCE event so that we can log it later, while
1289	 * returning from kernel or opal call.
1290	 */
1291	bl	machine_check_queue_event
1292	MACHINE_CHECK_HANDLER_WINDUP
1293	RFI_TO_KERNEL
1294
1295mce_deliver:
1296	/*
1297	 * This is a host user or guest MCE. Restore all registers, then
1298	 * run the "late" handler. For host user, this will run the
1299	 * machine_check_exception handler in virtual mode like a normal
1300	 * interrupt handler. For guest, this will trigger the KVM test
1301	 * and branch to the KVM interrupt similarly to other interrupts.
1302	 */
1303BEGIN_FTR_SECTION
1304	ld	r10,ORIG_GPR3(r1)
1305	mtspr	SPRN_CFAR,r10
1306END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
1307	MACHINE_CHECK_HANDLER_WINDUP
1308	GEN_INT_ENTRY machine_check, virt=0
1309
1310EXC_COMMON_BEGIN(machine_check_common)
1311	/*
1312	 * Machine check is different because we use a different
1313	 * save area: PACA_EXMC instead of PACA_EXGEN.
1314	 */
1315	GEN_COMMON machine_check
1316
1317	FINISH_NAP
1318	/* Enable MSR_RI when finished with PACA_EXMC */
1319	li	r10,MSR_RI
1320	mtmsrd 	r10,1
1321	addi	r3,r1,STACK_FRAME_OVERHEAD
1322	bl	machine_check_exception
1323	b	interrupt_return
1324
1325	GEN_KVM machine_check
1326
1327
1328#ifdef CONFIG_PPC_P7_NAP
1329/*
1330 * This is an idle wakeup. Low level machine check has already been
1331 * done. Queue the event then call the idle code to do the wake up.
1332 */
1333EXC_COMMON_BEGIN(machine_check_idle_common)
1334	bl	machine_check_queue_event
1335
1336	/*
1337	 * GPR-loss wakeups are relatively straightforward, because the
1338	 * idle sleep code has saved all non-volatile registers on its
1339	 * own stack, and r1 in PACAR1.
1340	 *
1341	 * For no-loss wakeups the r1 and lr registers used by the
1342	 * early machine check handler have to be restored first. r2 is
1343	 * the kernel TOC, so no need to restore it.
1344	 *
1345	 * Then decrement MCE nesting after finishing with the stack.
1346	 */
1347	ld	r3,_MSR(r1)
1348	ld	r4,_LINK(r1)
1349	ld	r1,GPR1(r1)
1350
1351	lhz	r11,PACA_IN_MCE(r13)
1352	subi	r11,r11,1
1353	sth	r11,PACA_IN_MCE(r13)
1354
1355	mtlr	r4
1356	rlwinm	r10,r3,47-31,30,31
1357	cmpwi	cr1,r10,2
1358	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
1359	b	idle_return_gpr_loss
1360#endif
1361
1362EXC_COMMON_BEGIN(unrecoverable_mce)
1363	/*
1364	 * We are going down. But there are chances that we might get hit by
1365	 * another MCE during panic path and we may run into unstable state
1366	 * with no way out. Hence, turn ME bit off while going down, so that
1367	 * when another MCE is hit during panic path, system will checkstop
1368	 * and hypervisor will get restarted cleanly by SP.
1369	 */
1370BEGIN_FTR_SECTION
1371	li	r10,0 /* clear MSR_RI */
1372	mtmsrd	r10,1
1373	bl	disable_machine_check
1374END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1375	ld	r10,PACAKMSR(r13)
1376	li	r3,MSR_ME
1377	andc	r10,r10,r3
1378	mtmsrd	r10
1379
1380	lhz	r12,PACA_IN_MCE(r13)
1381	subi	r12,r12,1
1382	sth	r12,PACA_IN_MCE(r13)
1383
1384	/* Invoke machine_check_exception to print MCE event and panic. */
1385	addi	r3,r1,STACK_FRAME_OVERHEAD
1386	bl	machine_check_exception
1387
1388	/*
1389	 * We will not reach here. Even if we did, there is no way out.
1390	 * Call unrecoverable_exception and die.
1391	 */
1392	addi	r3,r1,STACK_FRAME_OVERHEAD
1393	bl	unrecoverable_exception
1394	b	.
1395
1396
1397/**
1398 * Interrupt 0x300 - Data Storage Interrupt (DSI).
1399 * This is a synchronous interrupt generated due to a data access exception,
1400 * e.g., a load orstore which does not have a valid page table entry with
1401 * permissions. DAWR matches also fault here, as do RC updates, and minor misc
1402 * errors e.g., copy/paste, AMO, certain invalid CI accesses, etc.
1403 *
1404 * Handling:
1405 * - Hash MMU
1406 *   Go to do_hash_page first to see if the HPT can be filled from an entry in
1407 *   the Linux page table. Hash faults can hit in kernel mode in a fairly
1408 *   arbitrary state (e.g., interrupts disabled, locks held) when accessing
1409 *   "non-bolted" regions, e.g., vmalloc space. However these should always be
1410 *   backed by Linux page tables.
1411 *
1412 *   If none is found, do a Linux page fault. Linux page faults can happen in
1413 *   kernel mode due to user copy operations of course.
1414 *
1415 * - Radix MMU
1416 *   The hardware loads from the Linux page table directly, so a fault goes
1417 *   immediately to Linux page fault.
1418 *
1419 * Conditions like DAWR match are handled on the way in to Linux page fault.
1420 */
1421INT_DEFINE_BEGIN(data_access)
1422	IVEC=0x300
1423	IDAR=1
1424	IDSISR=1
1425#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1426	IKVM_SKIP=1
1427	IKVM_REAL=1
1428#endif
1429INT_DEFINE_END(data_access)
1430
1431EXC_REAL_BEGIN(data_access, 0x300, 0x80)
1432	GEN_INT_ENTRY data_access, virt=0
1433EXC_REAL_END(data_access, 0x300, 0x80)
1434EXC_VIRT_BEGIN(data_access, 0x4300, 0x80)
1435	GEN_INT_ENTRY data_access, virt=1
1436EXC_VIRT_END(data_access, 0x4300, 0x80)
1437EXC_COMMON_BEGIN(data_access_common)
1438	GEN_COMMON data_access
1439	ld	r4,_DAR(r1)
1440	ld	r5,_DSISR(r1)
1441BEGIN_MMU_FTR_SECTION
1442	ld	r6,_MSR(r1)
1443	li	r3,0x300
1444	b	do_hash_page		/* Try to handle as hpte fault */
1445MMU_FTR_SECTION_ELSE
1446	b	handle_page_fault
1447ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1448
1449	GEN_KVM data_access
1450
1451
1452/**
1453 * Interrupt 0x380 - Data Segment Interrupt (DSLB).
1454 * This is a synchronous interrupt in response to an MMU fault missing SLB
1455 * entry for HPT, or an address outside RPT translation range.
1456 *
1457 * Handling:
1458 * - HPT:
1459 *   This refills the SLB, or reports an access fault similarly to a bad page
1460 *   fault. When coming from user-mode, the SLB handler may access any kernel
1461 *   data, though it may itself take a DSLB. When coming from kernel mode,
1462 *   recursive faults must be avoided so access is restricted to the kernel
1463 *   image text/data, kernel stack, and any data allocated below
1464 *   ppc64_bolted_size (first segment). The kernel handler must avoid stomping
1465 *   on user-handler data structures.
1466 *
1467 * A dedicated save area EXSLB is used (XXX: but it actually need not be
1468 * these days, we could use EXGEN).
1469 */
1470INT_DEFINE_BEGIN(data_access_slb)
1471	IVEC=0x380
1472	IAREA=PACA_EXSLB
1473	IRECONCILE=0
1474	IDAR=1
1475#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1476	IKVM_SKIP=1
1477	IKVM_REAL=1
1478#endif
1479INT_DEFINE_END(data_access_slb)
1480
1481EXC_REAL_BEGIN(data_access_slb, 0x380, 0x80)
1482	GEN_INT_ENTRY data_access_slb, virt=0
1483EXC_REAL_END(data_access_slb, 0x380, 0x80)
1484EXC_VIRT_BEGIN(data_access_slb, 0x4380, 0x80)
1485	GEN_INT_ENTRY data_access_slb, virt=1
1486EXC_VIRT_END(data_access_slb, 0x4380, 0x80)
1487EXC_COMMON_BEGIN(data_access_slb_common)
1488	GEN_COMMON data_access_slb
1489	ld	r4,_DAR(r1)
1490	addi	r3,r1,STACK_FRAME_OVERHEAD
1491BEGIN_MMU_FTR_SECTION
1492	/* HPT case, do SLB fault */
1493	bl	do_slb_fault
1494	cmpdi	r3,0
1495	bne-	1f
1496	b	fast_interrupt_return
14971:	/* Error case */
1498MMU_FTR_SECTION_ELSE
1499	/* Radix case, access is outside page table range */
1500	li	r3,-EFAULT
1501ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1502	std	r3,RESULT(r1)
1503	RECONCILE_IRQ_STATE(r10, r11)
1504	ld	r4,_DAR(r1)
1505	ld	r5,RESULT(r1)
1506	addi	r3,r1,STACK_FRAME_OVERHEAD
1507	bl	do_bad_slb_fault
1508	b	interrupt_return
1509
1510	GEN_KVM data_access_slb
1511
1512
1513/**
1514 * Interrupt 0x400 - Instruction Storage Interrupt (ISI).
1515 * This is a synchronous interrupt in response to an MMU fault due to an
1516 * instruction fetch.
1517 *
1518 * Handling:
1519 * Similar to DSI, though in response to fetch. The faulting address is found
1520 * in SRR0 (rather than DAR), and status in SRR1 (rather than DSISR).
1521 */
1522INT_DEFINE_BEGIN(instruction_access)
1523	IVEC=0x400
1524	IISIDE=1
1525	IDAR=1
1526	IDSISR=1
1527#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1528	IKVM_REAL=1
1529#endif
1530INT_DEFINE_END(instruction_access)
1531
1532EXC_REAL_BEGIN(instruction_access, 0x400, 0x80)
1533	GEN_INT_ENTRY instruction_access, virt=0
1534EXC_REAL_END(instruction_access, 0x400, 0x80)
1535EXC_VIRT_BEGIN(instruction_access, 0x4400, 0x80)
1536	GEN_INT_ENTRY instruction_access, virt=1
1537EXC_VIRT_END(instruction_access, 0x4400, 0x80)
1538EXC_COMMON_BEGIN(instruction_access_common)
1539	GEN_COMMON instruction_access
1540	ld	r4,_DAR(r1)
1541	ld	r5,_DSISR(r1)
1542BEGIN_MMU_FTR_SECTION
1543	ld      r6,_MSR(r1)
1544	li	r3,0x400
1545	b	do_hash_page		/* Try to handle as hpte fault */
1546MMU_FTR_SECTION_ELSE
1547	b	handle_page_fault
1548ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1549
1550	GEN_KVM instruction_access
1551
1552
1553/**
1554 * Interrupt 0x480 - Instruction Segment Interrupt (ISLB).
1555 * This is a synchronous interrupt in response to an MMU fault due to an
1556 * instruction fetch.
1557 *
1558 * Handling:
1559 * Similar to DSLB, though in response to fetch. The faulting address is found
1560 * in SRR0 (rather than DAR).
1561 */
1562INT_DEFINE_BEGIN(instruction_access_slb)
1563	IVEC=0x480
1564	IAREA=PACA_EXSLB
1565	IRECONCILE=0
1566	IISIDE=1
1567	IDAR=1
1568#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1569	IKVM_REAL=1
1570#endif
1571INT_DEFINE_END(instruction_access_slb)
1572
1573EXC_REAL_BEGIN(instruction_access_slb, 0x480, 0x80)
1574	GEN_INT_ENTRY instruction_access_slb, virt=0
1575EXC_REAL_END(instruction_access_slb, 0x480, 0x80)
1576EXC_VIRT_BEGIN(instruction_access_slb, 0x4480, 0x80)
1577	GEN_INT_ENTRY instruction_access_slb, virt=1
1578EXC_VIRT_END(instruction_access_slb, 0x4480, 0x80)
1579EXC_COMMON_BEGIN(instruction_access_slb_common)
1580	GEN_COMMON instruction_access_slb
1581	ld	r4,_DAR(r1)
1582	addi	r3,r1,STACK_FRAME_OVERHEAD
1583BEGIN_MMU_FTR_SECTION
1584	/* HPT case, do SLB fault */
1585	bl	do_slb_fault
1586	cmpdi	r3,0
1587	bne-	1f
1588	b	fast_interrupt_return
15891:	/* Error case */
1590MMU_FTR_SECTION_ELSE
1591	/* Radix case, access is outside page table range */
1592	li	r3,-EFAULT
1593ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1594	std	r3,RESULT(r1)
1595	RECONCILE_IRQ_STATE(r10, r11)
1596	ld	r4,_DAR(r1)
1597	ld	r5,RESULT(r1)
1598	addi	r3,r1,STACK_FRAME_OVERHEAD
1599	bl	do_bad_slb_fault
1600	b	interrupt_return
1601
1602	GEN_KVM instruction_access_slb
1603
1604
1605/**
1606 * Interrupt 0x500 - External Interrupt.
1607 * This is an asynchronous maskable interrupt in response to an "external
1608 * exception" from the interrupt controller or hypervisor (e.g., device
1609 * interrupt). It is maskable in hardware by clearing MSR[EE], and
1610 * soft-maskable with IRQS_DISABLED mask (i.e., local_irq_disable()).
1611 *
1612 * When running in HV mode, Linux sets up the LPCR[LPES] bit such that
1613 * interrupts are delivered with HSRR registers, guests use SRRs, which
1614 * reqiures IHSRR_IF_HVMODE.
1615 *
1616 * On bare metal POWER9 and later, Linux sets the LPCR[HVICE] bit such that
1617 * external interrupts are delivered as Hypervisor Virtualization Interrupts
1618 * rather than External Interrupts.
1619 *
1620 * Handling:
1621 * This calls into Linux IRQ handler. NVGPRs are not saved to reduce overhead,
1622 * because registers at the time of the interrupt are not so important as it is
1623 * asynchronous.
1624 *
1625 * If soft masked, the masked handler will note the pending interrupt for
1626 * replay, and clear MSR[EE] in the interrupted context.
1627 */
1628INT_DEFINE_BEGIN(hardware_interrupt)
1629	IVEC=0x500
1630	IHSRR_IF_HVMODE=1
1631	IMASK=IRQS_DISABLED
1632	IKVM_REAL=1
1633	IKVM_VIRT=1
1634INT_DEFINE_END(hardware_interrupt)
1635
1636EXC_REAL_BEGIN(hardware_interrupt, 0x500, 0x100)
1637	GEN_INT_ENTRY hardware_interrupt, virt=0
1638EXC_REAL_END(hardware_interrupt, 0x500, 0x100)
1639EXC_VIRT_BEGIN(hardware_interrupt, 0x4500, 0x100)
1640	GEN_INT_ENTRY hardware_interrupt, virt=1
1641EXC_VIRT_END(hardware_interrupt, 0x4500, 0x100)
1642EXC_COMMON_BEGIN(hardware_interrupt_common)
1643	GEN_COMMON hardware_interrupt
1644	FINISH_NAP
1645	RUNLATCH_ON
1646	addi	r3,r1,STACK_FRAME_OVERHEAD
1647	bl	do_IRQ
1648	b	interrupt_return
1649
1650	GEN_KVM hardware_interrupt
1651
1652
1653/**
1654 * Interrupt 0x600 - Alignment Interrupt
1655 * This is a synchronous interrupt in response to data alignment fault.
1656 */
1657INT_DEFINE_BEGIN(alignment)
1658	IVEC=0x600
1659	IDAR=1
1660	IDSISR=1
1661#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1662	IKVM_REAL=1
1663#endif
1664INT_DEFINE_END(alignment)
1665
1666EXC_REAL_BEGIN(alignment, 0x600, 0x100)
1667	GEN_INT_ENTRY alignment, virt=0
1668EXC_REAL_END(alignment, 0x600, 0x100)
1669EXC_VIRT_BEGIN(alignment, 0x4600, 0x100)
1670	GEN_INT_ENTRY alignment, virt=1
1671EXC_VIRT_END(alignment, 0x4600, 0x100)
1672EXC_COMMON_BEGIN(alignment_common)
1673	GEN_COMMON alignment
1674	addi	r3,r1,STACK_FRAME_OVERHEAD
1675	bl	alignment_exception
1676	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
1677	b	interrupt_return
1678
1679	GEN_KVM alignment
1680
1681
1682/**
1683 * Interrupt 0x700 - Program Interrupt (program check).
1684 * This is a synchronous interrupt in response to various instruction faults:
1685 * traps, privilege errors, TM errors, floating point exceptions.
1686 *
1687 * Handling:
1688 * This interrupt may use the "emergency stack" in some cases when being taken
1689 * from kernel context, which complicates handling.
1690 */
1691INT_DEFINE_BEGIN(program_check)
1692	IVEC=0x700
1693#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1694	IKVM_REAL=1
1695#endif
1696INT_DEFINE_END(program_check)
1697
1698EXC_REAL_BEGIN(program_check, 0x700, 0x100)
1699	GEN_INT_ENTRY program_check, virt=0
1700EXC_REAL_END(program_check, 0x700, 0x100)
1701EXC_VIRT_BEGIN(program_check, 0x4700, 0x100)
1702	GEN_INT_ENTRY program_check, virt=1
1703EXC_VIRT_END(program_check, 0x4700, 0x100)
1704EXC_COMMON_BEGIN(program_check_common)
1705	__GEN_COMMON_ENTRY program_check
1706
1707	/*
1708	 * It's possible to receive a TM Bad Thing type program check with
1709	 * userspace register values (in particular r1), but with SRR1 reporting
1710	 * that we came from the kernel. Normally that would confuse the bad
1711	 * stack logic, and we would report a bad kernel stack pointer. Instead
1712	 * we switch to the emergency stack if we're taking a TM Bad Thing from
1713	 * the kernel.
1714	 */
1715
1716	andi.	r10,r12,MSR_PR
1717	bne	2f			/* If userspace, go normal path */
1718
1719	andis.	r10,r12,(SRR1_PROGTM)@h
1720	bne	1f			/* If TM, emergency		*/
1721
1722	cmpdi	r1,-INT_FRAME_SIZE	/* check if r1 is in userspace	*/
1723	blt	2f			/* normal path if not		*/
1724
1725	/* Use the emergency stack					*/
17261:	andi.	r10,r12,MSR_PR		/* Set CR0 correctly for label	*/
1727					/* 3 in EXCEPTION_PROLOG_COMMON	*/
1728	mr	r10,r1			/* Save r1			*/
1729	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack		*/
1730	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
1731	__ISTACK(program_check)=0
1732	__GEN_COMMON_BODY program_check
1733	b 3f
17342:
1735	__ISTACK(program_check)=1
1736	__GEN_COMMON_BODY program_check
17373:
1738	addi	r3,r1,STACK_FRAME_OVERHEAD
1739	bl	program_check_exception
1740	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
1741	b	interrupt_return
1742
1743	GEN_KVM program_check
1744
1745
1746/*
1747 * Interrupt 0x800 - Floating-Point Unavailable Interrupt.
1748 * This is a synchronous interrupt in response to executing an fp instruction
1749 * with MSR[FP]=0.
1750 *
1751 * Handling:
1752 * This will load FP registers and enable the FP bit if coming from userspace,
1753 * otherwise report a bad kernel use of FP.
1754 */
1755INT_DEFINE_BEGIN(fp_unavailable)
1756	IVEC=0x800
1757	IRECONCILE=0
1758#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1759	IKVM_REAL=1
1760#endif
1761INT_DEFINE_END(fp_unavailable)
1762
1763EXC_REAL_BEGIN(fp_unavailable, 0x800, 0x100)
1764	GEN_INT_ENTRY fp_unavailable, virt=0
1765EXC_REAL_END(fp_unavailable, 0x800, 0x100)
1766EXC_VIRT_BEGIN(fp_unavailable, 0x4800, 0x100)
1767	GEN_INT_ENTRY fp_unavailable, virt=1
1768EXC_VIRT_END(fp_unavailable, 0x4800, 0x100)
1769EXC_COMMON_BEGIN(fp_unavailable_common)
1770	GEN_COMMON fp_unavailable
1771	bne	1f			/* if from user, just load it up */
1772	RECONCILE_IRQ_STATE(r10, r11)
1773	addi	r3,r1,STACK_FRAME_OVERHEAD
1774	bl	kernel_fp_unavailable_exception
17750:	trap
1776	EMIT_BUG_ENTRY 0b, __FILE__, __LINE__, 0
17771:
1778#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1779BEGIN_FTR_SECTION
1780	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
1781	 * transaction), go do TM stuff
1782	 */
1783	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
1784	bne-	2f
1785END_FTR_SECTION_IFSET(CPU_FTR_TM)
1786#endif
1787	bl	load_up_fpu
1788	b	fast_interrupt_return
1789#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
17902:	/* User process was in a transaction */
1791	RECONCILE_IRQ_STATE(r10, r11)
1792	addi	r3,r1,STACK_FRAME_OVERHEAD
1793	bl	fp_unavailable_tm
1794	b	interrupt_return
1795#endif
1796
1797	GEN_KVM fp_unavailable
1798
1799
1800/**
1801 * Interrupt 0x900 - Decrementer Interrupt.
1802 * This is an asynchronous interrupt in response to a decrementer exception
1803 * (e.g., DEC has wrapped below zero). It is maskable in hardware by clearing
1804 * MSR[EE], and soft-maskable with IRQS_DISABLED mask (i.e.,
1805 * local_irq_disable()).
1806 *
1807 * Handling:
1808 * This calls into Linux timer handler. NVGPRs are not saved (see 0x500).
1809 *
1810 * If soft masked, the masked handler will note the pending interrupt for
1811 * replay, and bump the decrementer to a high value, leaving MSR[EE] enabled
1812 * in the interrupted context.
1813 * If PPC_WATCHDOG is configured, the soft masked handler will actually set
1814 * things back up to run soft_nmi_interrupt as a regular interrupt handler
1815 * on the emergency stack.
1816 */
1817INT_DEFINE_BEGIN(decrementer)
1818	IVEC=0x900
1819	IMASK=IRQS_DISABLED
1820#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1821	IKVM_REAL=1
1822#endif
1823INT_DEFINE_END(decrementer)
1824
1825EXC_REAL_BEGIN(decrementer, 0x900, 0x80)
1826	GEN_INT_ENTRY decrementer, virt=0
1827EXC_REAL_END(decrementer, 0x900, 0x80)
1828EXC_VIRT_BEGIN(decrementer, 0x4900, 0x80)
1829	GEN_INT_ENTRY decrementer, virt=1
1830EXC_VIRT_END(decrementer, 0x4900, 0x80)
1831EXC_COMMON_BEGIN(decrementer_common)
1832	GEN_COMMON decrementer
1833	FINISH_NAP
1834	RUNLATCH_ON
1835	addi	r3,r1,STACK_FRAME_OVERHEAD
1836	bl	timer_interrupt
1837	b	interrupt_return
1838
1839	GEN_KVM decrementer
1840
1841
1842/**
1843 * Interrupt 0x980 - Hypervisor Decrementer Interrupt.
1844 * This is an asynchronous interrupt, similar to 0x900 but for the HDEC
1845 * register.
1846 *
1847 * Handling:
1848 * Linux does not use this outside KVM where it's used to keep a host timer
1849 * while the guest is given control of DEC. It should normally be caught by
1850 * the KVM test and routed there.
1851 */
1852INT_DEFINE_BEGIN(hdecrementer)
1853	IVEC=0x980
1854	IHSRR=1
1855	ISTACK=0
1856	IRECONCILE=0
1857	IKVM_REAL=1
1858	IKVM_VIRT=1
1859INT_DEFINE_END(hdecrementer)
1860
1861EXC_REAL_BEGIN(hdecrementer, 0x980, 0x80)
1862	GEN_INT_ENTRY hdecrementer, virt=0
1863EXC_REAL_END(hdecrementer, 0x980, 0x80)
1864EXC_VIRT_BEGIN(hdecrementer, 0x4980, 0x80)
1865	GEN_INT_ENTRY hdecrementer, virt=1
1866EXC_VIRT_END(hdecrementer, 0x4980, 0x80)
1867EXC_COMMON_BEGIN(hdecrementer_common)
1868	__GEN_COMMON_ENTRY hdecrementer
1869	/*
1870	 * Hypervisor decrementer interrupts not caught by the KVM test
1871	 * shouldn't occur but are sometimes left pending on exit from a KVM
1872	 * guest.  We don't need to do anything to clear them, as they are
1873	 * edge-triggered.
1874	 *
1875	 * Be careful to avoid touching the kernel stack.
1876	 */
1877	ld	r10,PACA_EXGEN+EX_CTR(r13)
1878	mtctr	r10
1879	mtcrf	0x80,r9
1880	ld	r9,PACA_EXGEN+EX_R9(r13)
1881	ld	r10,PACA_EXGEN+EX_R10(r13)
1882	ld	r11,PACA_EXGEN+EX_R11(r13)
1883	ld	r12,PACA_EXGEN+EX_R12(r13)
1884	ld	r13,PACA_EXGEN+EX_R13(r13)
1885	HRFI_TO_KERNEL
1886
1887	GEN_KVM hdecrementer
1888
1889
1890/**
1891 * Interrupt 0xa00 - Directed Privileged Doorbell Interrupt.
1892 * This is an asynchronous interrupt in response to a msgsndp doorbell.
1893 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
1894 * IRQS_DISABLED mask (i.e., local_irq_disable()).
1895 *
1896 * Handling:
1897 * Guests may use this for IPIs between threads in a core if the
1898 * hypervisor supports it. NVGPRS are not saved (see 0x500).
1899 *
1900 * If soft masked, the masked handler will note the pending interrupt for
1901 * replay, leaving MSR[EE] enabled in the interrupted context because the
1902 * doorbells are edge triggered.
1903 */
1904INT_DEFINE_BEGIN(doorbell_super)
1905	IVEC=0xa00
1906	IMASK=IRQS_DISABLED
1907#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1908	IKVM_REAL=1
1909#endif
1910INT_DEFINE_END(doorbell_super)
1911
1912EXC_REAL_BEGIN(doorbell_super, 0xa00, 0x100)
1913	GEN_INT_ENTRY doorbell_super, virt=0
1914EXC_REAL_END(doorbell_super, 0xa00, 0x100)
1915EXC_VIRT_BEGIN(doorbell_super, 0x4a00, 0x100)
1916	GEN_INT_ENTRY doorbell_super, virt=1
1917EXC_VIRT_END(doorbell_super, 0x4a00, 0x100)
1918EXC_COMMON_BEGIN(doorbell_super_common)
1919	GEN_COMMON doorbell_super
1920	FINISH_NAP
1921	RUNLATCH_ON
1922	addi	r3,r1,STACK_FRAME_OVERHEAD
1923#ifdef CONFIG_PPC_DOORBELL
1924	bl	doorbell_exception
1925#else
1926	bl	unknown_exception
1927#endif
1928	b	interrupt_return
1929
1930	GEN_KVM doorbell_super
1931
1932
1933EXC_REAL_NONE(0xb00, 0x100)
1934EXC_VIRT_NONE(0x4b00, 0x100)
1935
1936/**
1937 * Interrupt 0xc00 - System Call Interrupt (syscall, hcall).
1938 * This is a synchronous interrupt invoked with the "sc" instruction. The
1939 * system call is invoked with "sc 0" and does not alter the HV bit, so it
1940 * is directed to the currently running OS. The hypercall is invoked with
1941 * "sc 1" and it sets HV=1, so it elevates to hypervisor.
1942 *
1943 * In HPT, sc 1 always goes to 0xc00 real mode. In RADIX, sc 1 can go to
1944 * 0x4c00 virtual mode.
1945 *
1946 * Handling:
1947 * If the KVM test fires then it was due to a hypercall and is accordingly
1948 * routed to KVM. Otherwise this executes a normal Linux system call.
1949 *
1950 * Call convention:
1951 *
1952 * syscall and hypercalls register conventions are documented in
1953 * Documentation/powerpc/syscall64-abi.rst and
1954 * Documentation/powerpc/papr_hcalls.rst respectively.
1955 *
1956 * The intersection of volatile registers that don't contain possible
1957 * inputs is: cr0, xer, ctr. We may use these as scratch regs upon entry
1958 * without saving, though xer is not a good idea to use, as hardware may
1959 * interpret some bits so it may be costly to change them.
1960 */
1961INT_DEFINE_BEGIN(system_call)
1962	IVEC=0xc00
1963	IKVM_REAL=1
1964	IKVM_VIRT=1
1965INT_DEFINE_END(system_call)
1966
1967.macro SYSTEM_CALL virt
1968#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1969	/*
1970	 * There is a little bit of juggling to get syscall and hcall
1971	 * working well. Save r13 in ctr to avoid using SPRG scratch
1972	 * register.
1973	 *
1974	 * Userspace syscalls have already saved the PPR, hcalls must save
1975	 * it before setting HMT_MEDIUM.
1976	 */
1977	mtctr	r13
1978	GET_PACA(r13)
1979	std	r10,PACA_EXGEN+EX_R10(r13)
1980	INTERRUPT_TO_KERNEL
1981	KVMTEST system_call /* uses r10, branch to system_call_kvm */
1982	mfctr	r9
1983#else
1984	mr	r9,r13
1985	GET_PACA(r13)
1986	INTERRUPT_TO_KERNEL
1987#endif
1988
1989#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
1990BEGIN_FTR_SECTION
1991	cmpdi	r0,0x1ebe
1992	beq-	1f
1993END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE)
1994#endif
1995
1996	/* We reach here with PACA in r13, r13 in r9. */
1997	mfspr	r11,SPRN_SRR0
1998	mfspr	r12,SPRN_SRR1
1999
2000	HMT_MEDIUM
2001
2002	.if ! \virt
2003	__LOAD_HANDLER(r10, system_call_common)
2004	mtspr	SPRN_SRR0,r10
2005	ld	r10,PACAKMSR(r13)
2006	mtspr	SPRN_SRR1,r10
2007	RFI_TO_KERNEL
2008	b	.	/* prevent speculative execution */
2009	.else
2010	li	r10,MSR_RI
2011	mtmsrd 	r10,1			/* Set RI (EE=0) */
2012#ifdef CONFIG_RELOCATABLE
2013	__LOAD_HANDLER(r10, system_call_common)
2014	mtctr	r10
2015	bctr
2016#else
2017	b	system_call_common
2018#endif
2019	.endif
2020
2021#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
2022	/* Fast LE/BE switch system call */
20231:	mfspr	r12,SPRN_SRR1
2024	xori	r12,r12,MSR_LE
2025	mtspr	SPRN_SRR1,r12
2026	mr	r13,r9
2027	RFI_TO_USER	/* return to userspace */
2028	b	.	/* prevent speculative execution */
2029#endif
2030.endm
2031
2032EXC_REAL_BEGIN(system_call, 0xc00, 0x100)
2033	SYSTEM_CALL 0
2034EXC_REAL_END(system_call, 0xc00, 0x100)
2035EXC_VIRT_BEGIN(system_call, 0x4c00, 0x100)
2036	SYSTEM_CALL 1
2037EXC_VIRT_END(system_call, 0x4c00, 0x100)
2038
2039#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
2040TRAMP_REAL_BEGIN(system_call_kvm)
2041	/*
2042	 * This is a hcall, so register convention is as above, with these
2043	 * differences:
2044	 * r13 = PACA
2045	 * ctr = orig r13
2046	 * orig r10 saved in PACA
2047	 */
2048	 /*
2049	  * Save the PPR (on systems that support it) before changing to
2050	  * HMT_MEDIUM. That allows the KVM code to save that value into the
2051	  * guest state (it is the guest's PPR value).
2052	  */
2053BEGIN_FTR_SECTION
2054	mfspr	r10,SPRN_PPR
2055	std	r10,HSTATE_PPR(r13)
2056END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
2057	HMT_MEDIUM
2058	mfctr	r10
2059	SET_SCRATCH0(r10)
2060	mfcr	r10
2061	std	r12,HSTATE_SCRATCH0(r13)
2062	sldi	r12,r10,32
2063	ori	r12,r12,0xc00
2064#ifdef CONFIG_RELOCATABLE
2065	/*
2066	 * Requires __LOAD_FAR_HANDLER beause kvmppc_interrupt lives
2067	 * outside the head section.
2068	 */
2069	__LOAD_FAR_HANDLER(r10, kvmppc_interrupt)
2070	mtctr   r10
2071	ld	r10,PACA_EXGEN+EX_R10(r13)
2072	bctr
2073#else
2074	ld	r10,PACA_EXGEN+EX_R10(r13)
2075	b       kvmppc_interrupt
2076#endif
2077#endif
2078
2079
2080/**
2081 * Interrupt 0xd00 - Trace Interrupt.
2082 * This is a synchronous interrupt in response to instruction step or
2083 * breakpoint faults.
2084 */
2085INT_DEFINE_BEGIN(single_step)
2086	IVEC=0xd00
2087#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2088	IKVM_REAL=1
2089#endif
2090INT_DEFINE_END(single_step)
2091
2092EXC_REAL_BEGIN(single_step, 0xd00, 0x100)
2093	GEN_INT_ENTRY single_step, virt=0
2094EXC_REAL_END(single_step, 0xd00, 0x100)
2095EXC_VIRT_BEGIN(single_step, 0x4d00, 0x100)
2096	GEN_INT_ENTRY single_step, virt=1
2097EXC_VIRT_END(single_step, 0x4d00, 0x100)
2098EXC_COMMON_BEGIN(single_step_common)
2099	GEN_COMMON single_step
2100	addi	r3,r1,STACK_FRAME_OVERHEAD
2101	bl	single_step_exception
2102	b	interrupt_return
2103
2104	GEN_KVM single_step
2105
2106
2107/**
2108 * Interrupt 0xe00 - Hypervisor Data Storage Interrupt (HDSI).
2109 * This is a synchronous interrupt in response to an MMU fault caused by a
2110 * guest data access.
2111 *
2112 * Handling:
2113 * This should always get routed to KVM. In radix MMU mode, this is caused
2114 * by a guest nested radix access that can't be performed due to the
2115 * partition scope page table. In hash mode, this can be caused by guests
2116 * running with translation disabled (virtual real mode) or with VPM enabled.
2117 * KVM will update the page table structures or disallow the access.
2118 */
2119INT_DEFINE_BEGIN(h_data_storage)
2120	IVEC=0xe00
2121	IHSRR=1
2122	IDAR=1
2123	IDSISR=1
2124	IKVM_SKIP=1
2125	IKVM_REAL=1
2126	IKVM_VIRT=1
2127INT_DEFINE_END(h_data_storage)
2128
2129EXC_REAL_BEGIN(h_data_storage, 0xe00, 0x20)
2130	GEN_INT_ENTRY h_data_storage, virt=0, ool=1
2131EXC_REAL_END(h_data_storage, 0xe00, 0x20)
2132EXC_VIRT_BEGIN(h_data_storage, 0x4e00, 0x20)
2133	GEN_INT_ENTRY h_data_storage, virt=1, ool=1
2134EXC_VIRT_END(h_data_storage, 0x4e00, 0x20)
2135EXC_COMMON_BEGIN(h_data_storage_common)
2136	GEN_COMMON h_data_storage
2137	addi    r3,r1,STACK_FRAME_OVERHEAD
2138BEGIN_MMU_FTR_SECTION
2139	ld	r4,_DAR(r1)
2140	li	r5,SIGSEGV
2141	bl      bad_page_fault
2142MMU_FTR_SECTION_ELSE
2143	bl      unknown_exception
2144ALT_MMU_FTR_SECTION_END_IFSET(MMU_FTR_TYPE_RADIX)
2145	b       interrupt_return
2146
2147	GEN_KVM h_data_storage
2148
2149
2150/**
2151 * Interrupt 0xe20 - Hypervisor Instruction Storage Interrupt (HISI).
2152 * This is a synchronous interrupt in response to an MMU fault caused by a
2153 * guest instruction fetch, similar to HDSI.
2154 */
2155INT_DEFINE_BEGIN(h_instr_storage)
2156	IVEC=0xe20
2157	IHSRR=1
2158	IKVM_REAL=1
2159	IKVM_VIRT=1
2160INT_DEFINE_END(h_instr_storage)
2161
2162EXC_REAL_BEGIN(h_instr_storage, 0xe20, 0x20)
2163	GEN_INT_ENTRY h_instr_storage, virt=0, ool=1
2164EXC_REAL_END(h_instr_storage, 0xe20, 0x20)
2165EXC_VIRT_BEGIN(h_instr_storage, 0x4e20, 0x20)
2166	GEN_INT_ENTRY h_instr_storage, virt=1, ool=1
2167EXC_VIRT_END(h_instr_storage, 0x4e20, 0x20)
2168EXC_COMMON_BEGIN(h_instr_storage_common)
2169	GEN_COMMON h_instr_storage
2170	addi	r3,r1,STACK_FRAME_OVERHEAD
2171	bl	unknown_exception
2172	b	interrupt_return
2173
2174	GEN_KVM h_instr_storage
2175
2176
2177/**
2178 * Interrupt 0xe40 - Hypervisor Emulation Assistance Interrupt.
2179 */
2180INT_DEFINE_BEGIN(emulation_assist)
2181	IVEC=0xe40
2182	IHSRR=1
2183	IKVM_REAL=1
2184	IKVM_VIRT=1
2185INT_DEFINE_END(emulation_assist)
2186
2187EXC_REAL_BEGIN(emulation_assist, 0xe40, 0x20)
2188	GEN_INT_ENTRY emulation_assist, virt=0, ool=1
2189EXC_REAL_END(emulation_assist, 0xe40, 0x20)
2190EXC_VIRT_BEGIN(emulation_assist, 0x4e40, 0x20)
2191	GEN_INT_ENTRY emulation_assist, virt=1, ool=1
2192EXC_VIRT_END(emulation_assist, 0x4e40, 0x20)
2193EXC_COMMON_BEGIN(emulation_assist_common)
2194	GEN_COMMON emulation_assist
2195	addi	r3,r1,STACK_FRAME_OVERHEAD
2196	bl	emulation_assist_interrupt
2197	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2198	b	interrupt_return
2199
2200	GEN_KVM emulation_assist
2201
2202
2203/**
2204 * Interrupt 0xe60 - Hypervisor Maintenance Interrupt (HMI).
2205 * This is an asynchronous interrupt caused by a Hypervisor Maintenance
2206 * Exception. It is always taken in real mode but uses HSRR registers
2207 * unlike SRESET and MCE.
2208 *
2209 * It is maskable in hardware by clearing MSR[EE], and partially soft-maskable
2210 * with IRQS_DISABLED mask (i.e., local_irq_disable()).
2211 *
2212 * Handling:
2213 * This is a special case, this is handled similarly to machine checks, with an
2214 * initial real mode handler that is not soft-masked, which attempts to fix the
2215 * problem. Then a regular handler which is soft-maskable and reports the
2216 * problem.
2217 *
2218 * The emergency stack is used for the early real mode handler.
2219 *
2220 * XXX: unclear why MCE and HMI schemes could not be made common, e.g.,
2221 * either use soft-masking for the MCE, or use irq_work for the HMI.
2222 *
2223 * KVM:
2224 * Unlike MCE, this calls into KVM without calling the real mode handler
2225 * first.
2226 */
2227INT_DEFINE_BEGIN(hmi_exception_early)
2228	IVEC=0xe60
2229	IHSRR=1
2230	IREALMODE_COMMON=1
2231	ISTACK=0
2232	IRECONCILE=0
2233	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
2234	IKVM_REAL=1
2235INT_DEFINE_END(hmi_exception_early)
2236
2237INT_DEFINE_BEGIN(hmi_exception)
2238	IVEC=0xe60
2239	IHSRR=1
2240	IMASK=IRQS_DISABLED
2241	IKVM_REAL=1
2242INT_DEFINE_END(hmi_exception)
2243
2244EXC_REAL_BEGIN(hmi_exception, 0xe60, 0x20)
2245	GEN_INT_ENTRY hmi_exception_early, virt=0, ool=1
2246EXC_REAL_END(hmi_exception, 0xe60, 0x20)
2247EXC_VIRT_NONE(0x4e60, 0x20)
2248
2249EXC_COMMON_BEGIN(hmi_exception_early_common)
2250	__GEN_REALMODE_COMMON_ENTRY hmi_exception_early
2251
2252	mr	r10,r1			/* Save r1 */
2253	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack for realmode */
2254	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
2255
2256	__GEN_COMMON_BODY hmi_exception_early
2257
2258	addi	r3,r1,STACK_FRAME_OVERHEAD
2259	bl	hmi_exception_realmode
2260	cmpdi	cr0,r3,0
2261	bne	1f
2262
2263	EXCEPTION_RESTORE_REGS hsrr=1
2264	HRFI_TO_USER_OR_KERNEL
2265
22661:
2267	/*
2268	 * Go to virtual mode and pull the HMI event information from
2269	 * firmware.
2270	 */
2271	EXCEPTION_RESTORE_REGS hsrr=1
2272	GEN_INT_ENTRY hmi_exception, virt=0
2273
2274	GEN_KVM hmi_exception_early
2275
2276EXC_COMMON_BEGIN(hmi_exception_common)
2277	GEN_COMMON hmi_exception
2278	FINISH_NAP
2279	RUNLATCH_ON
2280	addi	r3,r1,STACK_FRAME_OVERHEAD
2281	bl	handle_hmi_exception
2282	b	interrupt_return
2283
2284	GEN_KVM hmi_exception
2285
2286
2287/**
2288 * Interrupt 0xe80 - Directed Hypervisor Doorbell Interrupt.
2289 * This is an asynchronous interrupt in response to a msgsnd doorbell.
2290 * Similar to the 0xa00 doorbell but for host rather than guest.
2291 */
2292INT_DEFINE_BEGIN(h_doorbell)
2293	IVEC=0xe80
2294	IHSRR=1
2295	IMASK=IRQS_DISABLED
2296	IKVM_REAL=1
2297	IKVM_VIRT=1
2298INT_DEFINE_END(h_doorbell)
2299
2300EXC_REAL_BEGIN(h_doorbell, 0xe80, 0x20)
2301	GEN_INT_ENTRY h_doorbell, virt=0, ool=1
2302EXC_REAL_END(h_doorbell, 0xe80, 0x20)
2303EXC_VIRT_BEGIN(h_doorbell, 0x4e80, 0x20)
2304	GEN_INT_ENTRY h_doorbell, virt=1, ool=1
2305EXC_VIRT_END(h_doorbell, 0x4e80, 0x20)
2306EXC_COMMON_BEGIN(h_doorbell_common)
2307	GEN_COMMON h_doorbell
2308	FINISH_NAP
2309	RUNLATCH_ON
2310	addi	r3,r1,STACK_FRAME_OVERHEAD
2311#ifdef CONFIG_PPC_DOORBELL
2312	bl	doorbell_exception
2313#else
2314	bl	unknown_exception
2315#endif
2316	b	interrupt_return
2317
2318	GEN_KVM h_doorbell
2319
2320
2321/**
2322 * Interrupt 0xea0 - Hypervisor Virtualization Interrupt.
2323 * This is an asynchronous interrupt in response to an "external exception".
2324 * Similar to 0x500 but for host only.
2325 */
2326INT_DEFINE_BEGIN(h_virt_irq)
2327	IVEC=0xea0
2328	IHSRR=1
2329	IMASK=IRQS_DISABLED
2330	IKVM_REAL=1
2331	IKVM_VIRT=1
2332INT_DEFINE_END(h_virt_irq)
2333
2334EXC_REAL_BEGIN(h_virt_irq, 0xea0, 0x20)
2335	GEN_INT_ENTRY h_virt_irq, virt=0, ool=1
2336EXC_REAL_END(h_virt_irq, 0xea0, 0x20)
2337EXC_VIRT_BEGIN(h_virt_irq, 0x4ea0, 0x20)
2338	GEN_INT_ENTRY h_virt_irq, virt=1, ool=1
2339EXC_VIRT_END(h_virt_irq, 0x4ea0, 0x20)
2340EXC_COMMON_BEGIN(h_virt_irq_common)
2341	GEN_COMMON h_virt_irq
2342	FINISH_NAP
2343	RUNLATCH_ON
2344	addi	r3,r1,STACK_FRAME_OVERHEAD
2345	bl	do_IRQ
2346	b	interrupt_return
2347
2348	GEN_KVM h_virt_irq
2349
2350
2351EXC_REAL_NONE(0xec0, 0x20)
2352EXC_VIRT_NONE(0x4ec0, 0x20)
2353EXC_REAL_NONE(0xee0, 0x20)
2354EXC_VIRT_NONE(0x4ee0, 0x20)
2355
2356
2357/*
2358 * Interrupt 0xf00 - Performance Monitor Interrupt (PMI, PMU).
2359 * This is an asynchronous interrupt in response to a PMU exception.
2360 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
2361 * IRQS_PMI_DISABLED mask (NOTE: NOT local_irq_disable()).
2362 *
2363 * Handling:
2364 * This calls into the perf subsystem.
2365 *
2366 * Like the watchdog soft-nmi, it appears an NMI interrupt to Linux, in that it
2367 * runs under local_irq_disable. However it may be soft-masked in
2368 * powerpc-specific code.
2369 *
2370 * If soft masked, the masked handler will note the pending interrupt for
2371 * replay, and clear MSR[EE] in the interrupted context.
2372 */
2373INT_DEFINE_BEGIN(performance_monitor)
2374	IVEC=0xf00
2375	IMASK=IRQS_PMI_DISABLED
2376#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2377	IKVM_REAL=1
2378#endif
2379INT_DEFINE_END(performance_monitor)
2380
2381EXC_REAL_BEGIN(performance_monitor, 0xf00, 0x20)
2382	GEN_INT_ENTRY performance_monitor, virt=0, ool=1
2383EXC_REAL_END(performance_monitor, 0xf00, 0x20)
2384EXC_VIRT_BEGIN(performance_monitor, 0x4f00, 0x20)
2385	GEN_INT_ENTRY performance_monitor, virt=1, ool=1
2386EXC_VIRT_END(performance_monitor, 0x4f00, 0x20)
2387EXC_COMMON_BEGIN(performance_monitor_common)
2388	GEN_COMMON performance_monitor
2389	FINISH_NAP
2390	RUNLATCH_ON
2391	addi	r3,r1,STACK_FRAME_OVERHEAD
2392	bl	performance_monitor_exception
2393	b	interrupt_return
2394
2395	GEN_KVM performance_monitor
2396
2397
2398/**
2399 * Interrupt 0xf20 - Vector Unavailable Interrupt.
2400 * This is a synchronous interrupt in response to
2401 * executing a vector (or altivec) instruction with MSR[VEC]=0.
2402 * Similar to FP unavailable.
2403 */
2404INT_DEFINE_BEGIN(altivec_unavailable)
2405	IVEC=0xf20
2406	IRECONCILE=0
2407#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2408	IKVM_REAL=1
2409#endif
2410INT_DEFINE_END(altivec_unavailable)
2411
2412EXC_REAL_BEGIN(altivec_unavailable, 0xf20, 0x20)
2413	GEN_INT_ENTRY altivec_unavailable, virt=0, ool=1
2414EXC_REAL_END(altivec_unavailable, 0xf20, 0x20)
2415EXC_VIRT_BEGIN(altivec_unavailable, 0x4f20, 0x20)
2416	GEN_INT_ENTRY altivec_unavailable, virt=1, ool=1
2417EXC_VIRT_END(altivec_unavailable, 0x4f20, 0x20)
2418EXC_COMMON_BEGIN(altivec_unavailable_common)
2419	GEN_COMMON altivec_unavailable
2420#ifdef CONFIG_ALTIVEC
2421BEGIN_FTR_SECTION
2422	beq	1f
2423#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2424  BEGIN_FTR_SECTION_NESTED(69)
2425	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2426	 * transaction), go do TM stuff
2427	 */
2428	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2429	bne-	2f
2430  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2431#endif
2432	bl	load_up_altivec
2433	b	fast_interrupt_return
2434#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
24352:	/* User process was in a transaction */
2436	RECONCILE_IRQ_STATE(r10, r11)
2437	addi	r3,r1,STACK_FRAME_OVERHEAD
2438	bl	altivec_unavailable_tm
2439	b	interrupt_return
2440#endif
24411:
2442END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2443#endif
2444	RECONCILE_IRQ_STATE(r10, r11)
2445	addi	r3,r1,STACK_FRAME_OVERHEAD
2446	bl	altivec_unavailable_exception
2447	b	interrupt_return
2448
2449	GEN_KVM altivec_unavailable
2450
2451
2452/**
2453 * Interrupt 0xf40 - VSX Unavailable Interrupt.
2454 * This is a synchronous interrupt in response to
2455 * executing a VSX instruction with MSR[VSX]=0.
2456 * Similar to FP unavailable.
2457 */
2458INT_DEFINE_BEGIN(vsx_unavailable)
2459	IVEC=0xf40
2460	IRECONCILE=0
2461#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2462	IKVM_REAL=1
2463#endif
2464INT_DEFINE_END(vsx_unavailable)
2465
2466EXC_REAL_BEGIN(vsx_unavailable, 0xf40, 0x20)
2467	GEN_INT_ENTRY vsx_unavailable, virt=0, ool=1
2468EXC_REAL_END(vsx_unavailable, 0xf40, 0x20)
2469EXC_VIRT_BEGIN(vsx_unavailable, 0x4f40, 0x20)
2470	GEN_INT_ENTRY vsx_unavailable, virt=1, ool=1
2471EXC_VIRT_END(vsx_unavailable, 0x4f40, 0x20)
2472EXC_COMMON_BEGIN(vsx_unavailable_common)
2473	GEN_COMMON vsx_unavailable
2474#ifdef CONFIG_VSX
2475BEGIN_FTR_SECTION
2476	beq	1f
2477#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2478  BEGIN_FTR_SECTION_NESTED(69)
2479	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2480	 * transaction), go do TM stuff
2481	 */
2482	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2483	bne-	2f
2484  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2485#endif
2486	b	load_up_vsx
2487#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
24882:	/* User process was in a transaction */
2489	RECONCILE_IRQ_STATE(r10, r11)
2490	addi	r3,r1,STACK_FRAME_OVERHEAD
2491	bl	vsx_unavailable_tm
2492	b	interrupt_return
2493#endif
24941:
2495END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2496#endif
2497	RECONCILE_IRQ_STATE(r10, r11)
2498	addi	r3,r1,STACK_FRAME_OVERHEAD
2499	bl	vsx_unavailable_exception
2500	b	interrupt_return
2501
2502	GEN_KVM vsx_unavailable
2503
2504
2505/**
2506 * Interrupt 0xf60 - Facility Unavailable Interrupt.
2507 * This is a synchronous interrupt in response to
2508 * executing an instruction without access to the facility that can be
2509 * resolved by the OS (e.g., FSCR, MSR).
2510 * Similar to FP unavailable.
2511 */
2512INT_DEFINE_BEGIN(facility_unavailable)
2513	IVEC=0xf60
2514#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2515	IKVM_REAL=1
2516#endif
2517INT_DEFINE_END(facility_unavailable)
2518
2519EXC_REAL_BEGIN(facility_unavailable, 0xf60, 0x20)
2520	GEN_INT_ENTRY facility_unavailable, virt=0, ool=1
2521EXC_REAL_END(facility_unavailable, 0xf60, 0x20)
2522EXC_VIRT_BEGIN(facility_unavailable, 0x4f60, 0x20)
2523	GEN_INT_ENTRY facility_unavailable, virt=1, ool=1
2524EXC_VIRT_END(facility_unavailable, 0x4f60, 0x20)
2525EXC_COMMON_BEGIN(facility_unavailable_common)
2526	GEN_COMMON facility_unavailable
2527	addi	r3,r1,STACK_FRAME_OVERHEAD
2528	bl	facility_unavailable_exception
2529	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2530	b	interrupt_return
2531
2532	GEN_KVM facility_unavailable
2533
2534
2535/**
2536 * Interrupt 0xf60 - Hypervisor Facility Unavailable Interrupt.
2537 * This is a synchronous interrupt in response to
2538 * executing an instruction without access to the facility that can only
2539 * be resolved in HV mode (e.g., HFSCR).
2540 * Similar to FP unavailable.
2541 */
2542INT_DEFINE_BEGIN(h_facility_unavailable)
2543	IVEC=0xf80
2544	IHSRR=1
2545	IKVM_REAL=1
2546	IKVM_VIRT=1
2547INT_DEFINE_END(h_facility_unavailable)
2548
2549EXC_REAL_BEGIN(h_facility_unavailable, 0xf80, 0x20)
2550	GEN_INT_ENTRY h_facility_unavailable, virt=0, ool=1
2551EXC_REAL_END(h_facility_unavailable, 0xf80, 0x20)
2552EXC_VIRT_BEGIN(h_facility_unavailable, 0x4f80, 0x20)
2553	GEN_INT_ENTRY h_facility_unavailable, virt=1, ool=1
2554EXC_VIRT_END(h_facility_unavailable, 0x4f80, 0x20)
2555EXC_COMMON_BEGIN(h_facility_unavailable_common)
2556	GEN_COMMON h_facility_unavailable
2557	addi	r3,r1,STACK_FRAME_OVERHEAD
2558	bl	facility_unavailable_exception
2559	REST_NVGPRS(r1) /* XXX Shouldn't be necessary in practice */
2560	b	interrupt_return
2561
2562	GEN_KVM h_facility_unavailable
2563
2564
2565EXC_REAL_NONE(0xfa0, 0x20)
2566EXC_VIRT_NONE(0x4fa0, 0x20)
2567EXC_REAL_NONE(0xfc0, 0x20)
2568EXC_VIRT_NONE(0x4fc0, 0x20)
2569EXC_REAL_NONE(0xfe0, 0x20)
2570EXC_VIRT_NONE(0x4fe0, 0x20)
2571
2572EXC_REAL_NONE(0x1000, 0x100)
2573EXC_VIRT_NONE(0x5000, 0x100)
2574EXC_REAL_NONE(0x1100, 0x100)
2575EXC_VIRT_NONE(0x5100, 0x100)
2576
2577#ifdef CONFIG_CBE_RAS
2578INT_DEFINE_BEGIN(cbe_system_error)
2579	IVEC=0x1200
2580	IHSRR=1
2581	IKVM_SKIP=1
2582	IKVM_REAL=1
2583INT_DEFINE_END(cbe_system_error)
2584
2585EXC_REAL_BEGIN(cbe_system_error, 0x1200, 0x100)
2586	GEN_INT_ENTRY cbe_system_error, virt=0
2587EXC_REAL_END(cbe_system_error, 0x1200, 0x100)
2588EXC_VIRT_NONE(0x5200, 0x100)
2589EXC_COMMON_BEGIN(cbe_system_error_common)
2590	GEN_COMMON cbe_system_error
2591	addi	r3,r1,STACK_FRAME_OVERHEAD
2592	bl	cbe_system_error_exception
2593	b	interrupt_return
2594
2595	GEN_KVM cbe_system_error
2596
2597#else /* CONFIG_CBE_RAS */
2598EXC_REAL_NONE(0x1200, 0x100)
2599EXC_VIRT_NONE(0x5200, 0x100)
2600#endif
2601
2602
2603INT_DEFINE_BEGIN(instruction_breakpoint)
2604	IVEC=0x1300
2605#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2606	IKVM_SKIP=1
2607	IKVM_REAL=1
2608#endif
2609INT_DEFINE_END(instruction_breakpoint)
2610
2611EXC_REAL_BEGIN(instruction_breakpoint, 0x1300, 0x100)
2612	GEN_INT_ENTRY instruction_breakpoint, virt=0
2613EXC_REAL_END(instruction_breakpoint, 0x1300, 0x100)
2614EXC_VIRT_BEGIN(instruction_breakpoint, 0x5300, 0x100)
2615	GEN_INT_ENTRY instruction_breakpoint, virt=1
2616EXC_VIRT_END(instruction_breakpoint, 0x5300, 0x100)
2617EXC_COMMON_BEGIN(instruction_breakpoint_common)
2618	GEN_COMMON instruction_breakpoint
2619	addi	r3,r1,STACK_FRAME_OVERHEAD
2620	bl	instruction_breakpoint_exception
2621	b	interrupt_return
2622
2623	GEN_KVM instruction_breakpoint
2624
2625
2626EXC_REAL_NONE(0x1400, 0x100)
2627EXC_VIRT_NONE(0x5400, 0x100)
2628
2629/**
2630 * Interrupt 0x1500 - Soft Patch Interrupt
2631 *
2632 * Handling:
2633 * This is an implementation specific interrupt which can be used for a
2634 * range of exceptions.
2635 *
2636 * This interrupt handler is unique in that it runs the denormal assist
2637 * code even for guests (and even in guest context) without going to KVM,
2638 * for speed. POWER9 does not raise denorm exceptions, so this special case
2639 * could be phased out in future to reduce special cases.
2640 */
2641INT_DEFINE_BEGIN(denorm_exception)
2642	IVEC=0x1500
2643	IHSRR=1
2644	IBRANCH_TO_COMMON=0
2645	IKVM_REAL=1
2646INT_DEFINE_END(denorm_exception)
2647
2648EXC_REAL_BEGIN(denorm_exception, 0x1500, 0x100)
2649	GEN_INT_ENTRY denorm_exception, virt=0
2650#ifdef CONFIG_PPC_DENORMALISATION
2651	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2652	bne+	denorm_assist
2653#endif
2654	GEN_BRANCH_TO_COMMON denorm_exception, virt=0
2655EXC_REAL_END(denorm_exception, 0x1500, 0x100)
2656#ifdef CONFIG_PPC_DENORMALISATION
2657EXC_VIRT_BEGIN(denorm_exception, 0x5500, 0x100)
2658	GEN_INT_ENTRY denorm_exception, virt=1
2659	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2660	bne+	denorm_assist
2661	GEN_BRANCH_TO_COMMON denorm_exception, virt=1
2662EXC_VIRT_END(denorm_exception, 0x5500, 0x100)
2663#else
2664EXC_VIRT_NONE(0x5500, 0x100)
2665#endif
2666
2667#ifdef CONFIG_PPC_DENORMALISATION
2668TRAMP_REAL_BEGIN(denorm_assist)
2669BEGIN_FTR_SECTION
2670/*
2671 * To denormalise we need to move a copy of the register to itself.
2672 * For POWER6 do that here for all FP regs.
2673 */
2674	mfmsr	r10
2675	ori	r10,r10,(MSR_FP|MSR_FE0|MSR_FE1)
2676	xori	r10,r10,(MSR_FE0|MSR_FE1)
2677	mtmsrd	r10
2678	sync
2679
2680	.Lreg=0
2681	.rept 32
2682	fmr	.Lreg,.Lreg
2683	.Lreg=.Lreg+1
2684	.endr
2685
2686FTR_SECTION_ELSE
2687/*
2688 * To denormalise we need to move a copy of the register to itself.
2689 * For POWER7 do that here for the first 32 VSX registers only.
2690 */
2691	mfmsr	r10
2692	oris	r10,r10,MSR_VSX@h
2693	mtmsrd	r10
2694	sync
2695
2696	.Lreg=0
2697	.rept 32
2698	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2699	.Lreg=.Lreg+1
2700	.endr
2701
2702ALT_FTR_SECTION_END_IFCLR(CPU_FTR_ARCH_206)
2703
2704BEGIN_FTR_SECTION
2705	b	denorm_done
2706END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
2707/*
2708 * To denormalise we need to move a copy of the register to itself.
2709 * For POWER8 we need to do that for all 64 VSX registers
2710 */
2711	.Lreg=32
2712	.rept 32
2713	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2714	.Lreg=.Lreg+1
2715	.endr
2716
2717denorm_done:
2718	mfspr	r11,SPRN_HSRR0
2719	subi	r11,r11,4
2720	mtspr	SPRN_HSRR0,r11
2721	mtcrf	0x80,r9
2722	ld	r9,PACA_EXGEN+EX_R9(r13)
2723BEGIN_FTR_SECTION
2724	ld	r10,PACA_EXGEN+EX_PPR(r13)
2725	mtspr	SPRN_PPR,r10
2726END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
2727BEGIN_FTR_SECTION
2728	ld	r10,PACA_EXGEN+EX_CFAR(r13)
2729	mtspr	SPRN_CFAR,r10
2730END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
2731	ld	r10,PACA_EXGEN+EX_R10(r13)
2732	ld	r11,PACA_EXGEN+EX_R11(r13)
2733	ld	r12,PACA_EXGEN+EX_R12(r13)
2734	ld	r13,PACA_EXGEN+EX_R13(r13)
2735	HRFI_TO_UNKNOWN
2736	b	.
2737#endif
2738
2739EXC_COMMON_BEGIN(denorm_exception_common)
2740	GEN_COMMON denorm_exception
2741	addi	r3,r1,STACK_FRAME_OVERHEAD
2742	bl	unknown_exception
2743	b	interrupt_return
2744
2745	GEN_KVM denorm_exception
2746
2747
2748#ifdef CONFIG_CBE_RAS
2749INT_DEFINE_BEGIN(cbe_maintenance)
2750	IVEC=0x1600
2751	IHSRR=1
2752	IKVM_SKIP=1
2753	IKVM_REAL=1
2754INT_DEFINE_END(cbe_maintenance)
2755
2756EXC_REAL_BEGIN(cbe_maintenance, 0x1600, 0x100)
2757	GEN_INT_ENTRY cbe_maintenance, virt=0
2758EXC_REAL_END(cbe_maintenance, 0x1600, 0x100)
2759EXC_VIRT_NONE(0x5600, 0x100)
2760EXC_COMMON_BEGIN(cbe_maintenance_common)
2761	GEN_COMMON cbe_maintenance
2762	addi	r3,r1,STACK_FRAME_OVERHEAD
2763	bl	cbe_maintenance_exception
2764	b	interrupt_return
2765
2766	GEN_KVM cbe_maintenance
2767
2768#else /* CONFIG_CBE_RAS */
2769EXC_REAL_NONE(0x1600, 0x100)
2770EXC_VIRT_NONE(0x5600, 0x100)
2771#endif
2772
2773
2774INT_DEFINE_BEGIN(altivec_assist)
2775	IVEC=0x1700
2776#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2777	IKVM_REAL=1
2778#endif
2779INT_DEFINE_END(altivec_assist)
2780
2781EXC_REAL_BEGIN(altivec_assist, 0x1700, 0x100)
2782	GEN_INT_ENTRY altivec_assist, virt=0
2783EXC_REAL_END(altivec_assist, 0x1700, 0x100)
2784EXC_VIRT_BEGIN(altivec_assist, 0x5700, 0x100)
2785	GEN_INT_ENTRY altivec_assist, virt=1
2786EXC_VIRT_END(altivec_assist, 0x5700, 0x100)
2787EXC_COMMON_BEGIN(altivec_assist_common)
2788	GEN_COMMON altivec_assist
2789	addi	r3,r1,STACK_FRAME_OVERHEAD
2790#ifdef CONFIG_ALTIVEC
2791	bl	altivec_assist_exception
2792	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2793#else
2794	bl	unknown_exception
2795#endif
2796	b	interrupt_return
2797
2798	GEN_KVM altivec_assist
2799
2800
2801#ifdef CONFIG_CBE_RAS
2802INT_DEFINE_BEGIN(cbe_thermal)
2803	IVEC=0x1800
2804	IHSRR=1
2805	IKVM_SKIP=1
2806	IKVM_REAL=1
2807INT_DEFINE_END(cbe_thermal)
2808
2809EXC_REAL_BEGIN(cbe_thermal, 0x1800, 0x100)
2810	GEN_INT_ENTRY cbe_thermal, virt=0
2811EXC_REAL_END(cbe_thermal, 0x1800, 0x100)
2812EXC_VIRT_NONE(0x5800, 0x100)
2813EXC_COMMON_BEGIN(cbe_thermal_common)
2814	GEN_COMMON cbe_thermal
2815	addi	r3,r1,STACK_FRAME_OVERHEAD
2816	bl	cbe_thermal_exception
2817	b	interrupt_return
2818
2819	GEN_KVM cbe_thermal
2820
2821#else /* CONFIG_CBE_RAS */
2822EXC_REAL_NONE(0x1800, 0x100)
2823EXC_VIRT_NONE(0x5800, 0x100)
2824#endif
2825
2826
2827#ifdef CONFIG_PPC_WATCHDOG
2828
2829INT_DEFINE_BEGIN(soft_nmi)
2830	IVEC=0x900
2831	ISTACK=0
2832	IRECONCILE=0	/* Soft-NMI may fire under local_irq_disable */
2833INT_DEFINE_END(soft_nmi)
2834
2835/*
2836 * Branch to soft_nmi_interrupt using the emergency stack. The emergency
2837 * stack is one that is usable by maskable interrupts so long as MSR_EE
2838 * remains off. It is used for recovery when something has corrupted the
2839 * normal kernel stack, for example. The "soft NMI" must not use the process
2840 * stack because we want irq disabled sections to avoid touching the stack
2841 * at all (other than PMU interrupts), so use the emergency stack for this,
2842 * and run it entirely with interrupts hard disabled.
2843 */
2844EXC_COMMON_BEGIN(soft_nmi_common)
2845	mfspr	r11,SPRN_SRR0
2846	mr	r10,r1
2847	ld	r1,PACAEMERGSP(r13)
2848	subi	r1,r1,INT_FRAME_SIZE
2849	__GEN_COMMON_BODY soft_nmi
2850
2851	/*
2852	 * Set IRQS_ALL_DISABLED and save PACAIRQHAPPENED (see
2853	 * system_reset_common)
2854	 */
2855	li	r10,IRQS_ALL_DISABLED
2856	stb	r10,PACAIRQSOFTMASK(r13)
2857	lbz	r10,PACAIRQHAPPENED(r13)
2858	std	r10,RESULT(r1)
2859	ori	r10,r10,PACA_IRQ_HARD_DIS
2860	stb	r10,PACAIRQHAPPENED(r13)
2861
2862	addi	r3,r1,STACK_FRAME_OVERHEAD
2863	bl	soft_nmi_interrupt
2864
2865	/* Clear MSR_RI before setting SRR0 and SRR1. */
2866	li	r9,0
2867	mtmsrd	r9,1
2868
2869	/*
2870	 * Restore soft mask settings.
2871	 */
2872	ld	r10,RESULT(r1)
2873	stb	r10,PACAIRQHAPPENED(r13)
2874	ld	r10,SOFTE(r1)
2875	stb	r10,PACAIRQSOFTMASK(r13)
2876
2877	kuap_restore_amr r9, r10
2878	EXCEPTION_RESTORE_REGS hsrr=0
2879	RFI_TO_KERNEL
2880
2881#endif /* CONFIG_PPC_WATCHDOG */
2882
2883/*
2884 * An interrupt came in while soft-disabled. We set paca->irq_happened, then:
2885 * - If it was a decrementer interrupt, we bump the dec to max and and return.
2886 * - If it was a doorbell we return immediately since doorbells are edge
2887 *   triggered and won't automatically refire.
2888 * - If it was a HMI we return immediately since we handled it in realmode
2889 *   and it won't refire.
2890 * - Else it is one of PACA_IRQ_MUST_HARD_MASK, so hard disable and return.
2891 * This is called with r10 containing the value to OR to the paca field.
2892 */
2893.macro MASKED_INTERRUPT hsrr=0
2894	.if \hsrr
2895masked_Hinterrupt:
2896	.else
2897masked_interrupt:
2898	.endif
2899	lbz	r11,PACAIRQHAPPENED(r13)
2900	or	r11,r11,r10
2901	stb	r11,PACAIRQHAPPENED(r13)
2902	cmpwi	r10,PACA_IRQ_DEC
2903	bne	1f
2904	lis	r10,0x7fff
2905	ori	r10,r10,0xffff
2906	mtspr	SPRN_DEC,r10
2907#ifdef CONFIG_PPC_WATCHDOG
2908	b	soft_nmi_common
2909#else
2910	b	2f
2911#endif
29121:	andi.	r10,r10,PACA_IRQ_MUST_HARD_MASK
2913	beq	2f
2914	xori	r12,r12,MSR_EE	/* clear MSR_EE */
2915	.if \hsrr
2916	mtspr	SPRN_HSRR1,r12
2917	.else
2918	mtspr	SPRN_SRR1,r12
2919	.endif
2920	ori	r11,r11,PACA_IRQ_HARD_DIS
2921	stb	r11,PACAIRQHAPPENED(r13)
29222:	/* done */
2923	ld	r10,PACA_EXGEN+EX_CTR(r13)
2924	mtctr	r10
2925	mtcrf	0x80,r9
2926	std	r1,PACAR1(r13)
2927	ld	r9,PACA_EXGEN+EX_R9(r13)
2928	ld	r10,PACA_EXGEN+EX_R10(r13)
2929	ld	r11,PACA_EXGEN+EX_R11(r13)
2930	ld	r12,PACA_EXGEN+EX_R12(r13)
2931	ld	r13,PACA_EXGEN+EX_R13(r13)
2932	/* May return to masked low address where r13 is not set up */
2933	.if \hsrr
2934	HRFI_TO_KERNEL
2935	.else
2936	RFI_TO_KERNEL
2937	.endif
2938	b	.
2939.endm
2940
2941TRAMP_REAL_BEGIN(stf_barrier_fallback)
2942	std	r9,PACA_EXRFI+EX_R9(r13)
2943	std	r10,PACA_EXRFI+EX_R10(r13)
2944	sync
2945	ld	r9,PACA_EXRFI+EX_R9(r13)
2946	ld	r10,PACA_EXRFI+EX_R10(r13)
2947	ori	31,31,0
2948	.rept 14
2949	b	1f
29501:
2951	.endr
2952	blr
2953
2954TRAMP_REAL_BEGIN(rfi_flush_fallback)
2955	SET_SCRATCH0(r13);
2956	GET_PACA(r13);
2957	std	r1,PACA_EXRFI+EX_R12(r13)
2958	ld	r1,PACAKSAVE(r13)
2959	std	r9,PACA_EXRFI+EX_R9(r13)
2960	std	r10,PACA_EXRFI+EX_R10(r13)
2961	std	r11,PACA_EXRFI+EX_R11(r13)
2962	mfctr	r9
2963	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
2964	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
2965	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
2966	mtctr	r11
2967	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
2968
2969	/* order ld/st prior to dcbt stop all streams with flushing */
2970	sync
2971
2972	/*
2973	 * The load adresses are at staggered offsets within cachelines,
2974	 * which suits some pipelines better (on others it should not
2975	 * hurt).
2976	 */
29771:
2978	ld	r11,(0x80 + 8)*0(r10)
2979	ld	r11,(0x80 + 8)*1(r10)
2980	ld	r11,(0x80 + 8)*2(r10)
2981	ld	r11,(0x80 + 8)*3(r10)
2982	ld	r11,(0x80 + 8)*4(r10)
2983	ld	r11,(0x80 + 8)*5(r10)
2984	ld	r11,(0x80 + 8)*6(r10)
2985	ld	r11,(0x80 + 8)*7(r10)
2986	addi	r10,r10,0x80*8
2987	bdnz	1b
2988
2989	mtctr	r9
2990	ld	r9,PACA_EXRFI+EX_R9(r13)
2991	ld	r10,PACA_EXRFI+EX_R10(r13)
2992	ld	r11,PACA_EXRFI+EX_R11(r13)
2993	ld	r1,PACA_EXRFI+EX_R12(r13)
2994	GET_SCRATCH0(r13);
2995	rfid
2996
2997TRAMP_REAL_BEGIN(hrfi_flush_fallback)
2998	SET_SCRATCH0(r13);
2999	GET_PACA(r13);
3000	std	r1,PACA_EXRFI+EX_R12(r13)
3001	ld	r1,PACAKSAVE(r13)
3002	std	r9,PACA_EXRFI+EX_R9(r13)
3003	std	r10,PACA_EXRFI+EX_R10(r13)
3004	std	r11,PACA_EXRFI+EX_R11(r13)
3005	mfctr	r9
3006	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
3007	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
3008	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
3009	mtctr	r11
3010	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
3011
3012	/* order ld/st prior to dcbt stop all streams with flushing */
3013	sync
3014
3015	/*
3016	 * The load adresses are at staggered offsets within cachelines,
3017	 * which suits some pipelines better (on others it should not
3018	 * hurt).
3019	 */
30201:
3021	ld	r11,(0x80 + 8)*0(r10)
3022	ld	r11,(0x80 + 8)*1(r10)
3023	ld	r11,(0x80 + 8)*2(r10)
3024	ld	r11,(0x80 + 8)*3(r10)
3025	ld	r11,(0x80 + 8)*4(r10)
3026	ld	r11,(0x80 + 8)*5(r10)
3027	ld	r11,(0x80 + 8)*6(r10)
3028	ld	r11,(0x80 + 8)*7(r10)
3029	addi	r10,r10,0x80*8
3030	bdnz	1b
3031
3032	mtctr	r9
3033	ld	r9,PACA_EXRFI+EX_R9(r13)
3034	ld	r10,PACA_EXRFI+EX_R10(r13)
3035	ld	r11,PACA_EXRFI+EX_R11(r13)
3036	ld	r1,PACA_EXRFI+EX_R12(r13)
3037	GET_SCRATCH0(r13);
3038	hrfid
3039
3040TRAMP_REAL_BEGIN(rfscv_flush_fallback)
3041	/* system call volatile */
3042	mr	r7,r13
3043	GET_PACA(r13);
3044	mr	r8,r1
3045	ld	r1,PACAKSAVE(r13)
3046	mfctr	r9
3047	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
3048	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
3049	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
3050	mtctr	r11
3051	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
3052
3053	/* order ld/st prior to dcbt stop all streams with flushing */
3054	sync
3055
3056	/*
3057	 * The load adresses are at staggered offsets within cachelines,
3058	 * which suits some pipelines better (on others it should not
3059	 * hurt).
3060	 */
30611:
3062	ld	r11,(0x80 + 8)*0(r10)
3063	ld	r11,(0x80 + 8)*1(r10)
3064	ld	r11,(0x80 + 8)*2(r10)
3065	ld	r11,(0x80 + 8)*3(r10)
3066	ld	r11,(0x80 + 8)*4(r10)
3067	ld	r11,(0x80 + 8)*5(r10)
3068	ld	r11,(0x80 + 8)*6(r10)
3069	ld	r11,(0x80 + 8)*7(r10)
3070	addi	r10,r10,0x80*8
3071	bdnz	1b
3072
3073	mtctr	r9
3074	li	r9,0
3075	li	r10,0
3076	li	r11,0
3077	mr	r1,r8
3078	mr	r13,r7
3079	RFSCV
3080
3081USE_TEXT_SECTION()
3082	MASKED_INTERRUPT
3083	MASKED_INTERRUPT hsrr=1
3084
3085#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
3086kvmppc_skip_interrupt:
3087	/*
3088	 * Here all GPRs are unchanged from when the interrupt happened
3089	 * except for r13, which is saved in SPRG_SCRATCH0.
3090	 */
3091	mfspr	r13, SPRN_SRR0
3092	addi	r13, r13, 4
3093	mtspr	SPRN_SRR0, r13
3094	GET_SCRATCH0(r13)
3095	RFI_TO_KERNEL
3096	b	.
3097
3098kvmppc_skip_Hinterrupt:
3099	/*
3100	 * Here all GPRs are unchanged from when the interrupt happened
3101	 * except for r13, which is saved in SPRG_SCRATCH0.
3102	 */
3103	mfspr	r13, SPRN_HSRR0
3104	addi	r13, r13, 4
3105	mtspr	SPRN_HSRR0, r13
3106	GET_SCRATCH0(r13)
3107	HRFI_TO_KERNEL
3108	b	.
3109#endif
3110
3111	/*
3112	 * Relocation-on interrupts: A subset of the interrupts can be delivered
3113	 * with IR=1/DR=1, if AIL==2 and MSR.HV won't be changed by delivering
3114	 * it.  Addresses are the same as the original interrupt addresses, but
3115	 * offset by 0xc000000000004000.
3116	 * It's impossible to receive interrupts below 0x300 via this mechanism.
3117	 * KVM: None of these traps are from the guest ; anything that escalated
3118	 * to HV=1 from HV=0 is delivered via real mode handlers.
3119	 */
3120
3121	/*
3122	 * This uses the standard macro, since the original 0x300 vector
3123	 * only has extra guff for STAB-based processors -- which never
3124	 * come here.
3125	 */
3126
3127EXC_COMMON_BEGIN(ppc64_runlatch_on_trampoline)
3128	b	__ppc64_runlatch_on
3129
3130USE_FIXED_SECTION(virt_trampolines)
3131	/*
3132	 * All code below __end_interrupts is treated as soft-masked. If
3133	 * any code runs here with MSR[EE]=1, it must then cope with pending
3134	 * soft interrupt being raised (i.e., by ensuring it is replayed).
3135	 *
3136	 * The __end_interrupts marker must be past the out-of-line (OOL)
3137	 * handlers, so that they are copied to real address 0x100 when running
3138	 * a relocatable kernel. This ensures they can be reached from the short
3139	 * trampoline handlers (like 0x4f00, 0x4f20, etc.) which branch
3140	 * directly, without using LOAD_HANDLER().
3141	 */
3142	.align	7
3143	.globl	__end_interrupts
3144__end_interrupts:
3145DEFINE_FIXED_SYMBOL(__end_interrupts)
3146
3147#ifdef CONFIG_PPC_970_NAP
3148	/*
3149	 * Called by exception entry code if _TLF_NAPPING was set, this clears
3150	 * the NAPPING flag, and redirects the exception exit to
3151	 * power4_fixup_nap_return.
3152	 */
3153	.globl power4_fixup_nap
3154EXC_COMMON_BEGIN(power4_fixup_nap)
3155	andc	r9,r9,r10
3156	std	r9,TI_LOCAL_FLAGS(r11)
3157	LOAD_REG_ADDR(r10, power4_idle_nap_return)
3158	std	r10,_NIP(r1)
3159	blr
3160
3161power4_idle_nap_return:
3162	blr
3163#endif
3164
3165CLOSE_FIXED_SECTION(real_vectors);
3166CLOSE_FIXED_SECTION(real_trampolines);
3167CLOSE_FIXED_SECTION(virt_vectors);
3168CLOSE_FIXED_SECTION(virt_trampolines);
3169
3170USE_TEXT_SECTION()
3171
3172/* MSR[RI] should be clear because this uses SRR[01] */
3173enable_machine_check:
3174	mflr	r0
3175	bcl	20,31,$+4
31760:	mflr	r3
3177	addi	r3,r3,(1f - 0b)
3178	mtspr	SPRN_SRR0,r3
3179	mfmsr	r3
3180	ori	r3,r3,MSR_ME
3181	mtspr	SPRN_SRR1,r3
3182	RFI_TO_KERNEL
31831:	mtlr	r0
3184	blr
3185
3186/* MSR[RI] should be clear because this uses SRR[01] */
3187disable_machine_check:
3188	mflr	r0
3189	bcl	20,31,$+4
31900:	mflr	r3
3191	addi	r3,r3,(1f - 0b)
3192	mtspr	SPRN_SRR0,r3
3193	mfmsr	r3
3194	li	r4,MSR_ME
3195	andc	r3,r3,r4
3196	mtspr	SPRN_SRR1,r3
3197	RFI_TO_KERNEL
31981:	mtlr	r0
3199	blr
3200
3201/*
3202 * Hash table stuff
3203 */
3204	.balign	IFETCH_ALIGN_BYTES
3205do_hash_page:
3206#ifdef CONFIG_PPC_BOOK3S_64
3207	lis	r0,(DSISR_BAD_FAULT_64S | DSISR_DABRMATCH | DSISR_KEYFAULT)@h
3208	ori	r0,r0,DSISR_BAD_FAULT_64S@l
3209	and.	r0,r5,r0		/* weird error? */
3210	bne-	handle_page_fault	/* if not, try to insert a HPTE */
3211
3212	/*
3213	 * If we are in an "NMI" (e.g., an interrupt when soft-disabled), then
3214	 * don't call hash_page, just fail the fault. This is required to
3215	 * prevent re-entrancy problems in the hash code, namely perf
3216	 * interrupts hitting while something holds H_PAGE_BUSY, and taking a
3217	 * hash fault. See the comment in hash_preload().
3218	 */
3219	ld	r11, PACA_THREAD_INFO(r13)
3220	lwz	r0,TI_PREEMPT(r11)
3221	andis.	r0,r0,NMI_MASK@h
3222	bne	77f
3223
3224	/*
3225	 * r3 contains the trap number
3226	 * r4 contains the faulting address
3227	 * r5 contains dsisr
3228	 * r6 msr
3229	 *
3230	 * at return r3 = 0 for success, 1 for page fault, negative for error
3231	 */
3232	bl	__hash_page		/* build HPTE if possible */
3233        cmpdi	r3,0			/* see if __hash_page succeeded */
3234
3235	/* Success */
3236	beq	interrupt_return	/* Return from exception on success */
3237
3238	/* Error */
3239	blt-	13f
3240
3241	/* Reload DAR/DSISR into r4/r5 for the DABR check below */
3242	ld	r4,_DAR(r1)
3243	ld      r5,_DSISR(r1)
3244#endif /* CONFIG_PPC_BOOK3S_64 */
3245
3246/* Here we have a page fault that hash_page can't handle. */
3247handle_page_fault:
324811:	andis.  r0,r5,DSISR_DABRMATCH@h
3249	bne-    handle_dabr_fault
3250	addi	r3,r1,STACK_FRAME_OVERHEAD
3251	bl	do_page_fault
3252	cmpdi	r3,0
3253	beq+	interrupt_return
3254	mr	r5,r3
3255	addi	r3,r1,STACK_FRAME_OVERHEAD
3256	ld	r4,_DAR(r1)
3257	bl	bad_page_fault
3258	b	interrupt_return
3259
3260/* We have a data breakpoint exception - handle it */
3261handle_dabr_fault:
3262	ld      r4,_DAR(r1)
3263	ld      r5,_DSISR(r1)
3264	addi    r3,r1,STACK_FRAME_OVERHEAD
3265	bl      do_break
3266	/*
3267	 * do_break() may have changed the NV GPRS while handling a breakpoint.
3268	 * If so, we need to restore them with their updated values.
3269	 */
3270	REST_NVGPRS(r1)
3271	b       interrupt_return
3272
3273
3274#ifdef CONFIG_PPC_BOOK3S_64
3275/* We have a page fault that hash_page could handle but HV refused
3276 * the PTE insertion
3277 */
327813:	mr	r5,r3
3279	addi	r3,r1,STACK_FRAME_OVERHEAD
3280	ld	r4,_DAR(r1)
3281	bl	low_hash_fault
3282	b	interrupt_return
3283#endif
3284
3285/*
3286 * We come here as a result of a DSI at a point where we don't want
3287 * to call hash_page, such as when we are accessing memory (possibly
3288 * user memory) inside a PMU interrupt that occurred while interrupts
3289 * were soft-disabled.  We want to invoke the exception handler for
3290 * the access, or panic if there isn't a handler.
3291 */
329277:	addi	r3,r1,STACK_FRAME_OVERHEAD
3293	li	r5,SIGSEGV
3294	bl	bad_page_fault
3295	b	interrupt_return
3296