xref: /linux/arch/powerpc/kernel/exceptions-64s.S (revision 95298d63c67673c654c08952672d016212b26054)
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		lbz	r10,PACAIRQSOFTMASK(r13)
512		andi.	r10,r10,IMASK
513		/* Associate vector numbers with bits in paca->irq_happened */
514		.if IVEC == 0x500 || IVEC == 0xea0
515		li	r10,PACA_IRQ_EE
516		.elseif IVEC == 0x900
517		li	r10,PACA_IRQ_DEC
518		.elseif IVEC == 0xa00 || IVEC == 0xe80
519		li	r10,PACA_IRQ_DBELL
520		.elseif IVEC == 0xe60
521		li	r10,PACA_IRQ_HMI
522		.elseif IVEC == 0xf00
523		li	r10,PACA_IRQ_PMI
524		.else
525		.abort "Bad maskable vector"
526		.endif
527
528		.if IHSRR_IF_HVMODE
529		BEGIN_FTR_SECTION
530		bne	masked_Hinterrupt
531		FTR_SECTION_ELSE
532		bne	masked_interrupt
533		ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
534		.elseif IHSRR
535		bne	masked_Hinterrupt
536		.else
537		bne	masked_interrupt
538		.endif
539	.endif
540
541	.if ISTACK
542	andi.	r10,r12,MSR_PR		/* See if coming from user	*/
543	mr	r10,r1			/* Save r1			*/
544	subi	r1,r1,INT_FRAME_SIZE	/* alloc frame on kernel stack	*/
545	beq-	100f
546	ld	r1,PACAKSAVE(r13)	/* kernel stack to use		*/
547100:	tdgei	r1,-INT_FRAME_SIZE	/* trap if r1 is in userspace	*/
548	EMIT_BUG_ENTRY 100b,__FILE__,__LINE__,0
549	.endif
550
551	std	r9,_CCR(r1)		/* save CR in stackframe	*/
552	std	r11,_NIP(r1)		/* save SRR0 in stackframe	*/
553	std	r12,_MSR(r1)		/* save SRR1 in stackframe	*/
554	std	r10,0(r1)		/* make stack chain pointer	*/
555	std	r0,GPR0(r1)		/* save r0 in stackframe	*/
556	std	r10,GPR1(r1)		/* save r1 in stackframe	*/
557
558	.if ISET_RI
559	li	r10,MSR_RI
560	mtmsrd	r10,1			/* Set MSR_RI */
561	.endif
562
563	.if ISTACK
564	.if IKUAP
565	kuap_save_amr_and_lock r9, r10, cr1, cr0
566	.endif
567	beq	101f			/* if from kernel mode		*/
568	ACCOUNT_CPU_USER_ENTRY(r13, r9, r10)
569BEGIN_FTR_SECTION
570	ld	r9,IAREA+EX_PPR(r13)	/* Read PPR from paca		*/
571	std	r9,_PPR(r1)
572END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
573101:
574	.else
575	.if IKUAP
576	kuap_save_amr_and_lock r9, r10, cr1
577	.endif
578	.endif
579
580	/* Save original regs values from save area to stack frame. */
581	ld	r9,IAREA+EX_R9(r13)	/* move r9, r10 to stackframe	*/
582	ld	r10,IAREA+EX_R10(r13)
583	std	r9,GPR9(r1)
584	std	r10,GPR10(r1)
585	ld	r9,IAREA+EX_R11(r13)	/* move r11 - r13 to stackframe	*/
586	ld	r10,IAREA+EX_R12(r13)
587	ld	r11,IAREA+EX_R13(r13)
588	std	r9,GPR11(r1)
589	std	r10,GPR12(r1)
590	std	r11,GPR13(r1)
591
592	SAVE_NVGPRS(r1)
593
594	.if IDAR
595	.if IISIDE
596	ld	r10,_NIP(r1)
597	.else
598	ld	r10,IAREA+EX_DAR(r13)
599	.endif
600	std	r10,_DAR(r1)
601	.endif
602
603	.if IDSISR
604	.if IISIDE
605	ld	r10,_MSR(r1)
606	lis	r11,DSISR_SRR1_MATCH_64S@h
607	and	r10,r10,r11
608	.else
609	lwz	r10,IAREA+EX_DSISR(r13)
610	.endif
611	std	r10,_DSISR(r1)
612	.endif
613
614BEGIN_FTR_SECTION
615	ld	r10,IAREA+EX_CFAR(r13)
616	std	r10,ORIG_GPR3(r1)
617END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
618	ld	r10,IAREA+EX_CTR(r13)
619	std	r10,_CTR(r1)
620	std	r2,GPR2(r1)		/* save r2 in stackframe	*/
621	SAVE_4GPRS(3, r1)		/* save r3 - r6 in stackframe   */
622	SAVE_2GPRS(7, r1)		/* save r7, r8 in stackframe	*/
623	mflr	r9			/* Get LR, later save to stack	*/
624	ld	r2,PACATOC(r13)		/* get kernel TOC into r2	*/
625	std	r9,_LINK(r1)
626	lbz	r10,PACAIRQSOFTMASK(r13)
627	mfspr	r11,SPRN_XER		/* save XER in stackframe	*/
628	std	r10,SOFTE(r1)
629	std	r11,_XER(r1)
630	li	r9,IVEC
631	std	r9,_TRAP(r1)		/* set trap number		*/
632	li	r10,0
633	ld	r11,exception_marker@toc(r2)
634	std	r10,RESULT(r1)		/* clear regs->result		*/
635	std	r11,STACK_FRAME_OVERHEAD-16(r1) /* mark the frame	*/
636
637	.if ISTACK
638	ACCOUNT_STOLEN_TIME
639	.endif
640
641	.if IRECONCILE
642	RECONCILE_IRQ_STATE(r10, r11)
643	.endif
644.endm
645
646/*
647 * On entry r13 points to the paca, r9-r13 are saved in the paca,
648 * r9 contains the saved CR, r11 and r12 contain the saved SRR0 and
649 * SRR1, and relocation is on.
650 *
651 * If stack=0, then the stack is already set in r1, and r1 is saved in r10.
652 * PPR save and CPU accounting is not done for the !stack case (XXX why not?)
653 */
654.macro GEN_COMMON name
655	__GEN_COMMON_ENTRY \name
656	__GEN_COMMON_BODY \name
657.endm
658
659/*
660 * Restore all registers including H/SRR0/1 saved in a stack frame of a
661 * standard exception.
662 */
663.macro EXCEPTION_RESTORE_REGS hsrr=0
664	/* Move original SRR0 and SRR1 into the respective regs */
665	ld	r9,_MSR(r1)
666	.if \hsrr
667	mtspr	SPRN_HSRR1,r9
668	.else
669	mtspr	SPRN_SRR1,r9
670	.endif
671	ld	r9,_NIP(r1)
672	.if \hsrr
673	mtspr	SPRN_HSRR0,r9
674	.else
675	mtspr	SPRN_SRR0,r9
676	.endif
677	ld	r9,_CTR(r1)
678	mtctr	r9
679	ld	r9,_XER(r1)
680	mtxer	r9
681	ld	r9,_LINK(r1)
682	mtlr	r9
683	ld	r9,_CCR(r1)
684	mtcr	r9
685	REST_8GPRS(2, r1)
686	REST_4GPRS(10, r1)
687	REST_GPR(0, r1)
688	/* restore original r1. */
689	ld	r1,GPR1(r1)
690.endm
691
692#define RUNLATCH_ON				\
693BEGIN_FTR_SECTION				\
694	ld	r3, PACA_THREAD_INFO(r13);	\
695	ld	r4,TI_LOCAL_FLAGS(r3);		\
696	andi.	r0,r4,_TLF_RUNLATCH;		\
697	beql	ppc64_runlatch_on_trampoline;	\
698END_FTR_SECTION_IFSET(CPU_FTR_CTRL)
699
700/*
701 * When the idle code in power4_idle puts the CPU into NAP mode,
702 * it has to do so in a loop, and relies on the external interrupt
703 * and decrementer interrupt entry code to get it out of the loop.
704 * It sets the _TLF_NAPPING bit in current_thread_info()->local_flags
705 * to signal that it is in the loop and needs help to get out.
706 */
707#ifdef CONFIG_PPC_970_NAP
708#define FINISH_NAP				\
709BEGIN_FTR_SECTION				\
710	ld	r11, PACA_THREAD_INFO(r13);	\
711	ld	r9,TI_LOCAL_FLAGS(r11);		\
712	andi.	r10,r9,_TLF_NAPPING;		\
713	bnel	power4_fixup_nap;		\
714END_FTR_SECTION_IFSET(CPU_FTR_CAN_NAP)
715#else
716#define FINISH_NAP
717#endif
718
719/*
720 * There are a few constraints to be concerned with.
721 * - Real mode exceptions code/data must be located at their physical location.
722 * - Virtual mode exceptions must be mapped at their 0xc000... location.
723 * - Fixed location code must not call directly beyond the __end_interrupts
724 *   area when built with CONFIG_RELOCATABLE. LOAD_HANDLER / bctr sequence
725 *   must be used.
726 * - LOAD_HANDLER targets must be within first 64K of physical 0 /
727 *   virtual 0xc00...
728 * - Conditional branch targets must be within +/-32K of caller.
729 *
730 * "Virtual exceptions" run with relocation on (MSR_IR=1, MSR_DR=1), and
731 * therefore don't have to run in physically located code or rfid to
732 * virtual mode kernel code. However on relocatable kernels they do have
733 * to branch to KERNELBASE offset because the rest of the kernel (outside
734 * the exception vectors) may be located elsewhere.
735 *
736 * Virtual exceptions correspond with physical, except their entry points
737 * are offset by 0xc000000000000000 and also tend to get an added 0x4000
738 * offset applied. Virtual exceptions are enabled with the Alternate
739 * Interrupt Location (AIL) bit set in the LPCR. However this does not
740 * guarantee they will be delivered virtually. Some conditions (see the ISA)
741 * cause exceptions to be delivered in real mode.
742 *
743 * It's impossible to receive interrupts below 0x300 via AIL.
744 *
745 * KVM: None of the virtual exceptions are from the guest. Anything that
746 * escalated to HV=1 from HV=0 is delivered via real mode handlers.
747 *
748 *
749 * We layout physical memory as follows:
750 * 0x0000 - 0x00ff : Secondary processor spin code
751 * 0x0100 - 0x18ff : Real mode pSeries interrupt vectors
752 * 0x1900 - 0x3fff : Real mode trampolines
753 * 0x4000 - 0x58ff : Relon (IR=1,DR=1) mode pSeries interrupt vectors
754 * 0x5900 - 0x6fff : Relon mode trampolines
755 * 0x7000 - 0x7fff : FWNMI data area
756 * 0x8000 -   .... : Common interrupt handlers, remaining early
757 *                   setup code, rest of kernel.
758 *
759 * We could reclaim 0x4000-0x42ff for real mode trampolines if the space
760 * is necessary. Until then it's more consistent to explicitly put VIRT_NONE
761 * vectors there.
762 */
763OPEN_FIXED_SECTION(real_vectors,        0x0100, 0x1900)
764OPEN_FIXED_SECTION(real_trampolines,    0x1900, 0x4000)
765OPEN_FIXED_SECTION(virt_vectors,        0x4000, 0x5900)
766OPEN_FIXED_SECTION(virt_trampolines,    0x5900, 0x7000)
767
768#ifdef CONFIG_PPC_POWERNV
769	.globl start_real_trampolines
770	.globl end_real_trampolines
771	.globl start_virt_trampolines
772	.globl end_virt_trampolines
773#endif
774
775#if defined(CONFIG_PPC_PSERIES) || defined(CONFIG_PPC_POWERNV)
776/*
777 * Data area reserved for FWNMI option.
778 * This address (0x7000) is fixed by the RPA.
779 * pseries and powernv need to keep the whole page from
780 * 0x7000 to 0x8000 free for use by the firmware
781 */
782ZERO_FIXED_SECTION(fwnmi_page,          0x7000, 0x8000)
783OPEN_TEXT_SECTION(0x8000)
784#else
785OPEN_TEXT_SECTION(0x7000)
786#endif
787
788USE_FIXED_SECTION(real_vectors)
789
790/*
791 * This is the start of the interrupt handlers for pSeries
792 * This code runs with relocation off.
793 * Code from here to __end_interrupts gets copied down to real
794 * address 0x100 when we are running a relocatable kernel.
795 * Therefore any relative branches in this section must only
796 * branch to labels in this section.
797 */
798	.globl __start_interrupts
799__start_interrupts:
800
801/* No virt vectors corresponding with 0x0..0x100 */
802EXC_VIRT_NONE(0x4000, 0x100)
803
804
805/**
806 * Interrupt 0x100 - System Reset Interrupt (SRESET aka NMI).
807 * This is a non-maskable, asynchronous interrupt always taken in real-mode.
808 * It is caused by:
809 * - Wake from power-saving state, on powernv.
810 * - An NMI from another CPU, triggered by firmware or hypercall.
811 * - As crash/debug signal injected from BMC, firmware or hypervisor.
812 *
813 * Handling:
814 * Power-save wakeup is the only performance critical path, so this is
815 * determined quickly as possible first. In this case volatile registers
816 * can be discarded and SPRs like CFAR don't need to be read.
817 *
818 * If not a powersave wakeup, then it's run as a regular interrupt, however
819 * it uses its own stack and PACA save area to preserve the regular kernel
820 * environment for debugging.
821 *
822 * This interrupt is not maskable, so triggering it when MSR[RI] is clear,
823 * or SCRATCH0 is in use, etc. may cause a crash. It's also not entirely
824 * correct to switch to virtual mode to run the regular interrupt handler
825 * because it might be interrupted when the MMU is in a bad state (e.g., SLB
826 * is clear).
827 *
828 * FWNMI:
829 * PAPR specifies a "fwnmi" facility which sends the sreset to a different
830 * entry point with a different register set up. Some hypervisors will
831 * send the sreset to 0x100 in the guest if it is not fwnmi capable.
832 *
833 * KVM:
834 * Unlike most SRR interrupts, this may be taken by the host while executing
835 * in a guest, so a KVM test is required. KVM will pull the CPU out of guest
836 * mode and then raise the sreset.
837 */
838INT_DEFINE_BEGIN(system_reset)
839	IVEC=0x100
840	IAREA=PACA_EXNMI
841	IVIRT=0 /* no virt entry point */
842	/*
843	 * MSR_RI is not enabled, because PACA_EXNMI and nmi stack is
844	 * being used, so a nested NMI exception would corrupt it.
845	 */
846	ISET_RI=0
847	ISTACK=0
848	IRECONCILE=0
849	IKVM_REAL=1
850INT_DEFINE_END(system_reset)
851
852EXC_REAL_BEGIN(system_reset, 0x100, 0x100)
853#ifdef CONFIG_PPC_P7_NAP
854	/*
855	 * If running native on arch 2.06 or later, check if we are waking up
856	 * from nap/sleep/winkle, and branch to idle handler. This tests SRR1
857	 * bits 46:47. A non-0 value indicates that we are coming from a power
858	 * saving state. The idle wakeup handler initially runs in real mode,
859	 * but we branch to the 0xc000... address so we can turn on relocation
860	 * with mtmsrd later, after SPRs are restored.
861	 *
862	 * Careful to minimise cost for the fast path (idle wakeup) while
863	 * also avoiding clobbering CFAR for the debug path (non-idle).
864	 *
865	 * For the idle wake case volatile registers can be clobbered, which
866	 * is why we use those initially. If it turns out to not be an idle
867	 * wake, carefully put everything back the way it was, so we can use
868	 * common exception macros to handle it.
869	 */
870BEGIN_FTR_SECTION
871	SET_SCRATCH0(r13)
872	GET_PACA(r13)
873	std	r3,PACA_EXNMI+0*8(r13)
874	std	r4,PACA_EXNMI+1*8(r13)
875	std	r5,PACA_EXNMI+2*8(r13)
876	mfspr	r3,SPRN_SRR1
877	mfocrf	r4,0x80
878	rlwinm.	r5,r3,47-31,30,31
879	bne+	system_reset_idle_wake
880	/* Not powersave wakeup. Restore regs for regular interrupt handler. */
881	mtocrf	0x80,r4
882	ld	r3,PACA_EXNMI+0*8(r13)
883	ld	r4,PACA_EXNMI+1*8(r13)
884	ld	r5,PACA_EXNMI+2*8(r13)
885	GET_SCRATCH0(r13)
886END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
887#endif
888
889	GEN_INT_ENTRY system_reset, virt=0
890	/*
891	 * In theory, we should not enable relocation here if it was disabled
892	 * in SRR1, because the MMU may not be configured to support it (e.g.,
893	 * SLB may have been cleared). In practice, there should only be a few
894	 * small windows where that's the case, and sreset is considered to
895	 * be dangerous anyway.
896	 */
897EXC_REAL_END(system_reset, 0x100, 0x100)
898EXC_VIRT_NONE(0x4100, 0x100)
899
900#ifdef CONFIG_PPC_P7_NAP
901TRAMP_REAL_BEGIN(system_reset_idle_wake)
902	/* We are waking up from idle, so may clobber any volatile register */
903	cmpwi	cr1,r5,2
904	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
905	BRANCH_TO_C000(r12, DOTSYM(idle_return_gpr_loss))
906#endif
907
908#ifdef CONFIG_PPC_PSERIES
909/*
910 * Vectors for the FWNMI option.  Share common code.
911 */
912TRAMP_REAL_BEGIN(system_reset_fwnmi)
913	/* XXX: fwnmi guest could run a nested/PR guest, so why no test?  */
914	__IKVM_REAL(system_reset)=0
915	GEN_INT_ENTRY system_reset, virt=0
916
917#endif /* CONFIG_PPC_PSERIES */
918
919EXC_COMMON_BEGIN(system_reset_common)
920	__GEN_COMMON_ENTRY system_reset
921	/*
922	 * Increment paca->in_nmi then enable MSR_RI. SLB or MCE will be able
923	 * to recover, but nested NMI will notice in_nmi and not recover
924	 * because of the use of the NMI stack. in_nmi reentrancy is tested in
925	 * system_reset_exception.
926	 */
927	lhz	r10,PACA_IN_NMI(r13)
928	addi	r10,r10,1
929	sth	r10,PACA_IN_NMI(r13)
930	li	r10,MSR_RI
931	mtmsrd 	r10,1
932
933	mr	r10,r1
934	ld	r1,PACA_NMI_EMERG_SP(r13)
935	subi	r1,r1,INT_FRAME_SIZE
936	__GEN_COMMON_BODY system_reset
937	/*
938	 * Set IRQS_ALL_DISABLED unconditionally so irqs_disabled() does
939	 * the right thing. We do not want to reconcile because that goes
940	 * through irq tracing which we don't want in NMI.
941	 *
942	 * Save PACAIRQHAPPENED to RESULT (otherwise unused), and set HARD_DIS
943	 * as we are running with MSR[EE]=0.
944	 */
945	li	r10,IRQS_ALL_DISABLED
946	stb	r10,PACAIRQSOFTMASK(r13)
947	lbz	r10,PACAIRQHAPPENED(r13)
948	std	r10,RESULT(r1)
949	ori	r10,r10,PACA_IRQ_HARD_DIS
950	stb	r10,PACAIRQHAPPENED(r13)
951
952	addi	r3,r1,STACK_FRAME_OVERHEAD
953	bl	system_reset_exception
954
955	/* Clear MSR_RI before setting SRR0 and SRR1. */
956	li	r9,0
957	mtmsrd	r9,1
958
959	/*
960	 * MSR_RI is clear, now we can decrement paca->in_nmi.
961	 */
962	lhz	r10,PACA_IN_NMI(r13)
963	subi	r10,r10,1
964	sth	r10,PACA_IN_NMI(r13)
965
966	/*
967	 * Restore soft mask settings.
968	 */
969	ld	r10,RESULT(r1)
970	stb	r10,PACAIRQHAPPENED(r13)
971	ld	r10,SOFTE(r1)
972	stb	r10,PACAIRQSOFTMASK(r13)
973
974	kuap_restore_amr r9, r10
975	EXCEPTION_RESTORE_REGS
976	RFI_TO_USER_OR_KERNEL
977
978	GEN_KVM system_reset
979
980
981/**
982 * Interrupt 0x200 - Machine Check Interrupt (MCE).
983 * This is a non-maskable interrupt always taken in real-mode. It can be
984 * synchronous or asynchronous, caused by hardware or software, and it may be
985 * taken in a power-saving state.
986 *
987 * Handling:
988 * Similarly to system reset, this uses its own stack and PACA save area,
989 * the difference is re-entrancy is allowed on the machine check stack.
990 *
991 * machine_check_early is run in real mode, and carefully decodes the
992 * machine check and tries to handle it (e.g., flush the SLB if there was an
993 * error detected there), determines if it was recoverable and logs the
994 * event.
995 *
996 * This early code does not "reconcile" irq soft-mask state like SRESET or
997 * regular interrupts do, so irqs_disabled() among other things may not work
998 * properly (irq disable/enable already doesn't work because irq tracing can
999 * not work in real mode).
1000 *
1001 * Then, depending on the execution context when the interrupt is taken, there
1002 * are 3 main actions:
1003 * - Executing in kernel mode. The event is queued with irq_work, which means
1004 *   it is handled when it is next safe to do so (i.e., the kernel has enabled
1005 *   interrupts), which could be immediately when the interrupt returns. This
1006 *   avoids nasty issues like switching to virtual mode when the MMU is in a
1007 *   bad state, or when executing OPAL code. (SRESET is exposed to such issues,
1008 *   but it has different priorities). Check to see if the CPU was in power
1009 *   save, and return via the wake up code if it was.
1010 *
1011 * - Executing in user mode. machine_check_exception is run like a normal
1012 *   interrupt handler, which processes the data generated by the early handler.
1013 *
1014 * - Executing in guest mode. The interrupt is run with its KVM test, and
1015 *   branches to KVM to deal with. KVM may queue the event for the host
1016 *   to report later.
1017 *
1018 * This interrupt is not maskable, so if it triggers when MSR[RI] is clear,
1019 * or SCRATCH0 is in use, it may cause a crash.
1020 *
1021 * KVM:
1022 * See SRESET.
1023 */
1024INT_DEFINE_BEGIN(machine_check_early)
1025	IVEC=0x200
1026	IAREA=PACA_EXMC
1027	IVIRT=0 /* no virt entry point */
1028	IREALMODE_COMMON=1
1029	/*
1030	 * MSR_RI is not enabled, because PACA_EXMC is being used, so a
1031	 * nested machine check corrupts it. machine_check_common enables
1032	 * MSR_RI.
1033	 */
1034	ISET_RI=0
1035	ISTACK=0
1036	IDAR=1
1037	IDSISR=1
1038	IRECONCILE=0
1039	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
1040INT_DEFINE_END(machine_check_early)
1041
1042INT_DEFINE_BEGIN(machine_check)
1043	IVEC=0x200
1044	IAREA=PACA_EXMC
1045	IVIRT=0 /* no virt entry point */
1046	ISET_RI=0
1047	IDAR=1
1048	IDSISR=1
1049	IKVM_SKIP=1
1050	IKVM_REAL=1
1051INT_DEFINE_END(machine_check)
1052
1053EXC_REAL_BEGIN(machine_check, 0x200, 0x100)
1054	GEN_INT_ENTRY machine_check_early, virt=0
1055EXC_REAL_END(machine_check, 0x200, 0x100)
1056EXC_VIRT_NONE(0x4200, 0x100)
1057
1058#ifdef CONFIG_PPC_PSERIES
1059TRAMP_REAL_BEGIN(machine_check_fwnmi)
1060	/* See comment at machine_check exception, don't turn on RI */
1061	GEN_INT_ENTRY machine_check_early, virt=0
1062#endif
1063
1064#define MACHINE_CHECK_HANDLER_WINDUP			\
1065	/* Clear MSR_RI before setting SRR0 and SRR1. */\
1066	li	r9,0;					\
1067	mtmsrd	r9,1;		/* Clear MSR_RI */	\
1068	/* Decrement paca->in_mce now RI is clear. */	\
1069	lhz	r12,PACA_IN_MCE(r13);			\
1070	subi	r12,r12,1;				\
1071	sth	r12,PACA_IN_MCE(r13);			\
1072	EXCEPTION_RESTORE_REGS
1073
1074EXC_COMMON_BEGIN(machine_check_early_common)
1075	__GEN_REALMODE_COMMON_ENTRY machine_check_early
1076
1077	/*
1078	 * Switch to mc_emergency stack and handle re-entrancy (we limit
1079	 * the nested MCE upto level 4 to avoid stack overflow).
1080	 * Save MCE registers srr1, srr0, dar and dsisr and then set ME=1
1081	 *
1082	 * We use paca->in_mce to check whether this is the first entry or
1083	 * nested machine check. We increment paca->in_mce to track nested
1084	 * machine checks.
1085	 *
1086	 * If this is the first entry then set stack pointer to
1087	 * paca->mc_emergency_sp, otherwise r1 is already pointing to
1088	 * stack frame on mc_emergency stack.
1089	 *
1090	 * NOTE: We are here with MSR_ME=0 (off), which means we risk a
1091	 * checkstop if we get another machine check exception before we do
1092	 * rfid with MSR_ME=1.
1093	 *
1094	 * This interrupt can wake directly from idle. If that is the case,
1095	 * the machine check is handled then the idle wakeup code is called
1096	 * to restore state.
1097	 */
1098	lhz	r10,PACA_IN_MCE(r13)
1099	cmpwi	r10,0			/* Are we in nested machine check */
1100	cmpwi	cr1,r10,MAX_MCE_DEPTH	/* Are we at maximum nesting */
1101	addi	r10,r10,1		/* increment paca->in_mce */
1102	sth	r10,PACA_IN_MCE(r13)
1103
1104	mr	r10,r1			/* Save r1 */
1105	bne	1f
1106	/* First machine check entry */
1107	ld	r1,PACAMCEMERGSP(r13)	/* Use MC emergency stack */
11081:	/* Limit nested MCE to level 4 to avoid stack overflow */
1109	bgt	cr1,unrecoverable_mce	/* Check if we hit limit of 4 */
1110	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame */
1111
1112	__GEN_COMMON_BODY machine_check_early
1113
1114BEGIN_FTR_SECTION
1115	bl	enable_machine_check
1116END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1117	li	r10,MSR_RI
1118	mtmsrd	r10,1
1119
1120	/*
1121	 * Set IRQS_ALL_DISABLED and save PACAIRQHAPPENED (see
1122	 * system_reset_common)
1123	 */
1124	li	r10,IRQS_ALL_DISABLED
1125	stb	r10,PACAIRQSOFTMASK(r13)
1126	lbz	r10,PACAIRQHAPPENED(r13)
1127	std	r10,RESULT(r1)
1128	ori	r10,r10,PACA_IRQ_HARD_DIS
1129	stb	r10,PACAIRQHAPPENED(r13)
1130
1131	addi	r3,r1,STACK_FRAME_OVERHEAD
1132	bl	machine_check_early
1133	std	r3,RESULT(r1)	/* Save result */
1134	ld	r12,_MSR(r1)
1135
1136	/*
1137	 * Restore soft mask settings.
1138	 */
1139	ld	r10,RESULT(r1)
1140	stb	r10,PACAIRQHAPPENED(r13)
1141	ld	r10,SOFTE(r1)
1142	stb	r10,PACAIRQSOFTMASK(r13)
1143
1144#ifdef CONFIG_PPC_P7_NAP
1145	/*
1146	 * Check if thread was in power saving mode. We come here when any
1147	 * of the following is true:
1148	 * a. thread wasn't in power saving mode
1149	 * b. thread was in power saving mode with no state loss,
1150	 *    supervisor state loss or hypervisor state loss.
1151	 *
1152	 * Go back to nap/sleep/winkle mode again if (b) is true.
1153	 */
1154BEGIN_FTR_SECTION
1155	rlwinm.	r11,r12,47-31,30,31
1156	bne	machine_check_idle_common
1157END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
1158#endif
1159
1160#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1161	/*
1162	 * Check if we are coming from guest. If yes, then run the normal
1163	 * exception handler which will take the
1164	 * machine_check_kvm->kvmppc_interrupt branch to deliver the MC event
1165	 * to guest.
1166	 */
1167	lbz	r11,HSTATE_IN_GUEST(r13)
1168	cmpwi	r11,0			/* Check if coming from guest */
1169	bne	mce_deliver		/* continue if we are. */
1170#endif
1171
1172	/*
1173	 * Check if we are coming from userspace. If yes, then run the normal
1174	 * exception handler which will deliver the MC event to this kernel.
1175	 */
1176	andi.	r11,r12,MSR_PR		/* See if coming from user. */
1177	bne	mce_deliver		/* continue in V mode if we are. */
1178
1179	/*
1180	 * At this point we are coming from kernel context.
1181	 * Queue up the MCE event and return from the interrupt.
1182	 * But before that, check if this is an un-recoverable exception.
1183	 * If yes, then stay on emergency stack and panic.
1184	 */
1185	andi.	r11,r12,MSR_RI
1186	beq	unrecoverable_mce
1187
1188	/*
1189	 * Check if we have successfully handled/recovered from error, if not
1190	 * then stay on emergency stack and panic.
1191	 */
1192	ld	r3,RESULT(r1)	/* Load result */
1193	cmpdi	r3,0		/* see if we handled MCE successfully */
1194	beq	unrecoverable_mce /* if !handled then panic */
1195
1196	/*
1197	 * Return from MC interrupt.
1198	 * Queue up the MCE event so that we can log it later, while
1199	 * returning from kernel or opal call.
1200	 */
1201	bl	machine_check_queue_event
1202	MACHINE_CHECK_HANDLER_WINDUP
1203	RFI_TO_KERNEL
1204
1205mce_deliver:
1206	/*
1207	 * This is a host user or guest MCE. Restore all registers, then
1208	 * run the "late" handler. For host user, this will run the
1209	 * machine_check_exception handler in virtual mode like a normal
1210	 * interrupt handler. For guest, this will trigger the KVM test
1211	 * and branch to the KVM interrupt similarly to other interrupts.
1212	 */
1213BEGIN_FTR_SECTION
1214	ld	r10,ORIG_GPR3(r1)
1215	mtspr	SPRN_CFAR,r10
1216END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
1217	MACHINE_CHECK_HANDLER_WINDUP
1218	GEN_INT_ENTRY machine_check, virt=0
1219
1220EXC_COMMON_BEGIN(machine_check_common)
1221	/*
1222	 * Machine check is different because we use a different
1223	 * save area: PACA_EXMC instead of PACA_EXGEN.
1224	 */
1225	GEN_COMMON machine_check
1226
1227	FINISH_NAP
1228	/* Enable MSR_RI when finished with PACA_EXMC */
1229	li	r10,MSR_RI
1230	mtmsrd 	r10,1
1231	addi	r3,r1,STACK_FRAME_OVERHEAD
1232	bl	machine_check_exception
1233	b	interrupt_return
1234
1235	GEN_KVM machine_check
1236
1237
1238#ifdef CONFIG_PPC_P7_NAP
1239/*
1240 * This is an idle wakeup. Low level machine check has already been
1241 * done. Queue the event then call the idle code to do the wake up.
1242 */
1243EXC_COMMON_BEGIN(machine_check_idle_common)
1244	bl	machine_check_queue_event
1245
1246	/*
1247	 * GPR-loss wakeups are relatively straightforward, because the
1248	 * idle sleep code has saved all non-volatile registers on its
1249	 * own stack, and r1 in PACAR1.
1250	 *
1251	 * For no-loss wakeups the r1 and lr registers used by the
1252	 * early machine check handler have to be restored first. r2 is
1253	 * the kernel TOC, so no need to restore it.
1254	 *
1255	 * Then decrement MCE nesting after finishing with the stack.
1256	 */
1257	ld	r3,_MSR(r1)
1258	ld	r4,_LINK(r1)
1259	ld	r1,GPR1(r1)
1260
1261	lhz	r11,PACA_IN_MCE(r13)
1262	subi	r11,r11,1
1263	sth	r11,PACA_IN_MCE(r13)
1264
1265	mtlr	r4
1266	rlwinm	r10,r3,47-31,30,31
1267	cmpwi	cr1,r10,2
1268	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
1269	b	idle_return_gpr_loss
1270#endif
1271
1272EXC_COMMON_BEGIN(unrecoverable_mce)
1273	/*
1274	 * We are going down. But there are chances that we might get hit by
1275	 * another MCE during panic path and we may run into unstable state
1276	 * with no way out. Hence, turn ME bit off while going down, so that
1277	 * when another MCE is hit during panic path, system will checkstop
1278	 * and hypervisor will get restarted cleanly by SP.
1279	 */
1280BEGIN_FTR_SECTION
1281	li	r10,0 /* clear MSR_RI */
1282	mtmsrd	r10,1
1283	bl	disable_machine_check
1284END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1285	ld	r10,PACAKMSR(r13)
1286	li	r3,MSR_ME
1287	andc	r10,r10,r3
1288	mtmsrd	r10
1289
1290	lhz	r12,PACA_IN_MCE(r13)
1291	subi	r12,r12,1
1292	sth	r12,PACA_IN_MCE(r13)
1293
1294	/* Invoke machine_check_exception to print MCE event and panic. */
1295	addi	r3,r1,STACK_FRAME_OVERHEAD
1296	bl	machine_check_exception
1297
1298	/*
1299	 * We will not reach here. Even if we did, there is no way out.
1300	 * Call unrecoverable_exception and die.
1301	 */
1302	addi	r3,r1,STACK_FRAME_OVERHEAD
1303	bl	unrecoverable_exception
1304	b	.
1305
1306
1307/**
1308 * Interrupt 0x300 - Data Storage Interrupt (DSI).
1309 * This is a synchronous interrupt generated due to a data access exception,
1310 * e.g., a load orstore which does not have a valid page table entry with
1311 * permissions. DAWR matches also fault here, as do RC updates, and minor misc
1312 * errors e.g., copy/paste, AMO, certain invalid CI accesses, etc.
1313 *
1314 * Handling:
1315 * - Hash MMU
1316 *   Go to do_hash_page first to see if the HPT can be filled from an entry in
1317 *   the Linux page table. Hash faults can hit in kernel mode in a fairly
1318 *   arbitrary state (e.g., interrupts disabled, locks held) when accessing
1319 *   "non-bolted" regions, e.g., vmalloc space. However these should always be
1320 *   backed by Linux page tables.
1321 *
1322 *   If none is found, do a Linux page fault. Linux page faults can happen in
1323 *   kernel mode due to user copy operations of course.
1324 *
1325 * - Radix MMU
1326 *   The hardware loads from the Linux page table directly, so a fault goes
1327 *   immediately to Linux page fault.
1328 *
1329 * Conditions like DAWR match are handled on the way in to Linux page fault.
1330 */
1331INT_DEFINE_BEGIN(data_access)
1332	IVEC=0x300
1333	IDAR=1
1334	IDSISR=1
1335#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1336	IKVM_SKIP=1
1337	IKVM_REAL=1
1338#endif
1339INT_DEFINE_END(data_access)
1340
1341EXC_REAL_BEGIN(data_access, 0x300, 0x80)
1342	GEN_INT_ENTRY data_access, virt=0
1343EXC_REAL_END(data_access, 0x300, 0x80)
1344EXC_VIRT_BEGIN(data_access, 0x4300, 0x80)
1345	GEN_INT_ENTRY data_access, virt=1
1346EXC_VIRT_END(data_access, 0x4300, 0x80)
1347EXC_COMMON_BEGIN(data_access_common)
1348	GEN_COMMON data_access
1349	ld	r4,_DAR(r1)
1350	ld	r5,_DSISR(r1)
1351BEGIN_MMU_FTR_SECTION
1352	ld	r6,_MSR(r1)
1353	li	r3,0x300
1354	b	do_hash_page		/* Try to handle as hpte fault */
1355MMU_FTR_SECTION_ELSE
1356	b	handle_page_fault
1357ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1358
1359	GEN_KVM data_access
1360
1361
1362/**
1363 * Interrupt 0x380 - Data Segment Interrupt (DSLB).
1364 * This is a synchronous interrupt in response to an MMU fault missing SLB
1365 * entry for HPT, or an address outside RPT translation range.
1366 *
1367 * Handling:
1368 * - HPT:
1369 *   This refills the SLB, or reports an access fault similarly to a bad page
1370 *   fault. When coming from user-mode, the SLB handler may access any kernel
1371 *   data, though it may itself take a DSLB. When coming from kernel mode,
1372 *   recursive faults must be avoided so access is restricted to the kernel
1373 *   image text/data, kernel stack, and any data allocated below
1374 *   ppc64_bolted_size (first segment). The kernel handler must avoid stomping
1375 *   on user-handler data structures.
1376 *
1377 * A dedicated save area EXSLB is used (XXX: but it actually need not be
1378 * these days, we could use EXGEN).
1379 */
1380INT_DEFINE_BEGIN(data_access_slb)
1381	IVEC=0x380
1382	IAREA=PACA_EXSLB
1383	IRECONCILE=0
1384	IDAR=1
1385#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1386	IKVM_SKIP=1
1387	IKVM_REAL=1
1388#endif
1389INT_DEFINE_END(data_access_slb)
1390
1391EXC_REAL_BEGIN(data_access_slb, 0x380, 0x80)
1392	GEN_INT_ENTRY data_access_slb, virt=0
1393EXC_REAL_END(data_access_slb, 0x380, 0x80)
1394EXC_VIRT_BEGIN(data_access_slb, 0x4380, 0x80)
1395	GEN_INT_ENTRY data_access_slb, virt=1
1396EXC_VIRT_END(data_access_slb, 0x4380, 0x80)
1397EXC_COMMON_BEGIN(data_access_slb_common)
1398	GEN_COMMON data_access_slb
1399	ld	r4,_DAR(r1)
1400	addi	r3,r1,STACK_FRAME_OVERHEAD
1401BEGIN_MMU_FTR_SECTION
1402	/* HPT case, do SLB fault */
1403	bl	do_slb_fault
1404	cmpdi	r3,0
1405	bne-	1f
1406	b	fast_interrupt_return
14071:	/* Error case */
1408MMU_FTR_SECTION_ELSE
1409	/* Radix case, access is outside page table range */
1410	li	r3,-EFAULT
1411ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1412	std	r3,RESULT(r1)
1413	RECONCILE_IRQ_STATE(r10, r11)
1414	ld	r4,_DAR(r1)
1415	ld	r5,RESULT(r1)
1416	addi	r3,r1,STACK_FRAME_OVERHEAD
1417	bl	do_bad_slb_fault
1418	b	interrupt_return
1419
1420	GEN_KVM data_access_slb
1421
1422
1423/**
1424 * Interrupt 0x400 - Instruction Storage Interrupt (ISI).
1425 * This is a synchronous interrupt in response to an MMU fault due to an
1426 * instruction fetch.
1427 *
1428 * Handling:
1429 * Similar to DSI, though in response to fetch. The faulting address is found
1430 * in SRR0 (rather than DAR), and status in SRR1 (rather than DSISR).
1431 */
1432INT_DEFINE_BEGIN(instruction_access)
1433	IVEC=0x400
1434	IISIDE=1
1435	IDAR=1
1436	IDSISR=1
1437#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1438	IKVM_REAL=1
1439#endif
1440INT_DEFINE_END(instruction_access)
1441
1442EXC_REAL_BEGIN(instruction_access, 0x400, 0x80)
1443	GEN_INT_ENTRY instruction_access, virt=0
1444EXC_REAL_END(instruction_access, 0x400, 0x80)
1445EXC_VIRT_BEGIN(instruction_access, 0x4400, 0x80)
1446	GEN_INT_ENTRY instruction_access, virt=1
1447EXC_VIRT_END(instruction_access, 0x4400, 0x80)
1448EXC_COMMON_BEGIN(instruction_access_common)
1449	GEN_COMMON instruction_access
1450	ld	r4,_DAR(r1)
1451	ld	r5,_DSISR(r1)
1452BEGIN_MMU_FTR_SECTION
1453	ld      r6,_MSR(r1)
1454	li	r3,0x400
1455	b	do_hash_page		/* Try to handle as hpte fault */
1456MMU_FTR_SECTION_ELSE
1457	b	handle_page_fault
1458ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1459
1460	GEN_KVM instruction_access
1461
1462
1463/**
1464 * Interrupt 0x480 - Instruction Segment Interrupt (ISLB).
1465 * This is a synchronous interrupt in response to an MMU fault due to an
1466 * instruction fetch.
1467 *
1468 * Handling:
1469 * Similar to DSLB, though in response to fetch. The faulting address is found
1470 * in SRR0 (rather than DAR).
1471 */
1472INT_DEFINE_BEGIN(instruction_access_slb)
1473	IVEC=0x480
1474	IAREA=PACA_EXSLB
1475	IRECONCILE=0
1476	IISIDE=1
1477	IDAR=1
1478#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1479	IKVM_REAL=1
1480#endif
1481INT_DEFINE_END(instruction_access_slb)
1482
1483EXC_REAL_BEGIN(instruction_access_slb, 0x480, 0x80)
1484	GEN_INT_ENTRY instruction_access_slb, virt=0
1485EXC_REAL_END(instruction_access_slb, 0x480, 0x80)
1486EXC_VIRT_BEGIN(instruction_access_slb, 0x4480, 0x80)
1487	GEN_INT_ENTRY instruction_access_slb, virt=1
1488EXC_VIRT_END(instruction_access_slb, 0x4480, 0x80)
1489EXC_COMMON_BEGIN(instruction_access_slb_common)
1490	GEN_COMMON instruction_access_slb
1491	ld	r4,_DAR(r1)
1492	addi	r3,r1,STACK_FRAME_OVERHEAD
1493BEGIN_MMU_FTR_SECTION
1494	/* HPT case, do SLB fault */
1495	bl	do_slb_fault
1496	cmpdi	r3,0
1497	bne-	1f
1498	b	fast_interrupt_return
14991:	/* Error case */
1500MMU_FTR_SECTION_ELSE
1501	/* Radix case, access is outside page table range */
1502	li	r3,-EFAULT
1503ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1504	std	r3,RESULT(r1)
1505	RECONCILE_IRQ_STATE(r10, r11)
1506	ld	r4,_DAR(r1)
1507	ld	r5,RESULT(r1)
1508	addi	r3,r1,STACK_FRAME_OVERHEAD
1509	bl	do_bad_slb_fault
1510	b	interrupt_return
1511
1512	GEN_KVM instruction_access_slb
1513
1514
1515/**
1516 * Interrupt 0x500 - External Interrupt.
1517 * This is an asynchronous maskable interrupt in response to an "external
1518 * exception" from the interrupt controller or hypervisor (e.g., device
1519 * interrupt). It is maskable in hardware by clearing MSR[EE], and
1520 * soft-maskable with IRQS_DISABLED mask (i.e., local_irq_disable()).
1521 *
1522 * When running in HV mode, Linux sets up the LPCR[LPES] bit such that
1523 * interrupts are delivered with HSRR registers, guests use SRRs, which
1524 * reqiures IHSRR_IF_HVMODE.
1525 *
1526 * On bare metal POWER9 and later, Linux sets the LPCR[HVICE] bit such that
1527 * external interrupts are delivered as Hypervisor Virtualization Interrupts
1528 * rather than External Interrupts.
1529 *
1530 * Handling:
1531 * This calls into Linux IRQ handler. NVGPRs are not saved to reduce overhead,
1532 * because registers at the time of the interrupt are not so important as it is
1533 * asynchronous.
1534 *
1535 * If soft masked, the masked handler will note the pending interrupt for
1536 * replay, and clear MSR[EE] in the interrupted context.
1537 */
1538INT_DEFINE_BEGIN(hardware_interrupt)
1539	IVEC=0x500
1540	IHSRR_IF_HVMODE=1
1541	IMASK=IRQS_DISABLED
1542	IKVM_REAL=1
1543	IKVM_VIRT=1
1544INT_DEFINE_END(hardware_interrupt)
1545
1546EXC_REAL_BEGIN(hardware_interrupt, 0x500, 0x100)
1547	GEN_INT_ENTRY hardware_interrupt, virt=0
1548EXC_REAL_END(hardware_interrupt, 0x500, 0x100)
1549EXC_VIRT_BEGIN(hardware_interrupt, 0x4500, 0x100)
1550	GEN_INT_ENTRY hardware_interrupt, virt=1
1551EXC_VIRT_END(hardware_interrupt, 0x4500, 0x100)
1552EXC_COMMON_BEGIN(hardware_interrupt_common)
1553	GEN_COMMON hardware_interrupt
1554	FINISH_NAP
1555	RUNLATCH_ON
1556	addi	r3,r1,STACK_FRAME_OVERHEAD
1557	bl	do_IRQ
1558	b	interrupt_return
1559
1560	GEN_KVM hardware_interrupt
1561
1562
1563/**
1564 * Interrupt 0x600 - Alignment Interrupt
1565 * This is a synchronous interrupt in response to data alignment fault.
1566 */
1567INT_DEFINE_BEGIN(alignment)
1568	IVEC=0x600
1569	IDAR=1
1570	IDSISR=1
1571#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1572	IKVM_REAL=1
1573#endif
1574INT_DEFINE_END(alignment)
1575
1576EXC_REAL_BEGIN(alignment, 0x600, 0x100)
1577	GEN_INT_ENTRY alignment, virt=0
1578EXC_REAL_END(alignment, 0x600, 0x100)
1579EXC_VIRT_BEGIN(alignment, 0x4600, 0x100)
1580	GEN_INT_ENTRY alignment, virt=1
1581EXC_VIRT_END(alignment, 0x4600, 0x100)
1582EXC_COMMON_BEGIN(alignment_common)
1583	GEN_COMMON alignment
1584	addi	r3,r1,STACK_FRAME_OVERHEAD
1585	bl	alignment_exception
1586	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
1587	b	interrupt_return
1588
1589	GEN_KVM alignment
1590
1591
1592/**
1593 * Interrupt 0x700 - Program Interrupt (program check).
1594 * This is a synchronous interrupt in response to various instruction faults:
1595 * traps, privilege errors, TM errors, floating point exceptions.
1596 *
1597 * Handling:
1598 * This interrupt may use the "emergency stack" in some cases when being taken
1599 * from kernel context, which complicates handling.
1600 */
1601INT_DEFINE_BEGIN(program_check)
1602	IVEC=0x700
1603#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1604	IKVM_REAL=1
1605#endif
1606INT_DEFINE_END(program_check)
1607
1608EXC_REAL_BEGIN(program_check, 0x700, 0x100)
1609	GEN_INT_ENTRY program_check, virt=0
1610EXC_REAL_END(program_check, 0x700, 0x100)
1611EXC_VIRT_BEGIN(program_check, 0x4700, 0x100)
1612	GEN_INT_ENTRY program_check, virt=1
1613EXC_VIRT_END(program_check, 0x4700, 0x100)
1614EXC_COMMON_BEGIN(program_check_common)
1615	__GEN_COMMON_ENTRY program_check
1616
1617	/*
1618	 * It's possible to receive a TM Bad Thing type program check with
1619	 * userspace register values (in particular r1), but with SRR1 reporting
1620	 * that we came from the kernel. Normally that would confuse the bad
1621	 * stack logic, and we would report a bad kernel stack pointer. Instead
1622	 * we switch to the emergency stack if we're taking a TM Bad Thing from
1623	 * the kernel.
1624	 */
1625
1626	andi.	r10,r12,MSR_PR
1627	bne	2f			/* If userspace, go normal path */
1628
1629	andis.	r10,r12,(SRR1_PROGTM)@h
1630	bne	1f			/* If TM, emergency		*/
1631
1632	cmpdi	r1,-INT_FRAME_SIZE	/* check if r1 is in userspace	*/
1633	blt	2f			/* normal path if not		*/
1634
1635	/* Use the emergency stack					*/
16361:	andi.	r10,r12,MSR_PR		/* Set CR0 correctly for label	*/
1637					/* 3 in EXCEPTION_PROLOG_COMMON	*/
1638	mr	r10,r1			/* Save r1			*/
1639	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack		*/
1640	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
1641	__ISTACK(program_check)=0
1642	__GEN_COMMON_BODY program_check
1643	b 3f
16442:
1645	__ISTACK(program_check)=1
1646	__GEN_COMMON_BODY program_check
16473:
1648	addi	r3,r1,STACK_FRAME_OVERHEAD
1649	bl	program_check_exception
1650	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
1651	b	interrupt_return
1652
1653	GEN_KVM program_check
1654
1655
1656/*
1657 * Interrupt 0x800 - Floating-Point Unavailable Interrupt.
1658 * This is a synchronous interrupt in response to executing an fp instruction
1659 * with MSR[FP]=0.
1660 *
1661 * Handling:
1662 * This will load FP registers and enable the FP bit if coming from userspace,
1663 * otherwise report a bad kernel use of FP.
1664 */
1665INT_DEFINE_BEGIN(fp_unavailable)
1666	IVEC=0x800
1667	IRECONCILE=0
1668#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1669	IKVM_REAL=1
1670#endif
1671INT_DEFINE_END(fp_unavailable)
1672
1673EXC_REAL_BEGIN(fp_unavailable, 0x800, 0x100)
1674	GEN_INT_ENTRY fp_unavailable, virt=0
1675EXC_REAL_END(fp_unavailable, 0x800, 0x100)
1676EXC_VIRT_BEGIN(fp_unavailable, 0x4800, 0x100)
1677	GEN_INT_ENTRY fp_unavailable, virt=1
1678EXC_VIRT_END(fp_unavailable, 0x4800, 0x100)
1679EXC_COMMON_BEGIN(fp_unavailable_common)
1680	GEN_COMMON fp_unavailable
1681	bne	1f			/* if from user, just load it up */
1682	RECONCILE_IRQ_STATE(r10, r11)
1683	addi	r3,r1,STACK_FRAME_OVERHEAD
1684	bl	kernel_fp_unavailable_exception
16850:	trap
1686	EMIT_BUG_ENTRY 0b, __FILE__, __LINE__, 0
16871:
1688#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1689BEGIN_FTR_SECTION
1690	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
1691	 * transaction), go do TM stuff
1692	 */
1693	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
1694	bne-	2f
1695END_FTR_SECTION_IFSET(CPU_FTR_TM)
1696#endif
1697	bl	load_up_fpu
1698	b	fast_interrupt_return
1699#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
17002:	/* User process was in a transaction */
1701	RECONCILE_IRQ_STATE(r10, r11)
1702	addi	r3,r1,STACK_FRAME_OVERHEAD
1703	bl	fp_unavailable_tm
1704	b	interrupt_return
1705#endif
1706
1707	GEN_KVM fp_unavailable
1708
1709
1710/**
1711 * Interrupt 0x900 - Decrementer Interrupt.
1712 * This is an asynchronous interrupt in response to a decrementer exception
1713 * (e.g., DEC has wrapped below zero). It is maskable in hardware by clearing
1714 * MSR[EE], and soft-maskable with IRQS_DISABLED mask (i.e.,
1715 * local_irq_disable()).
1716 *
1717 * Handling:
1718 * This calls into Linux timer handler. NVGPRs are not saved (see 0x500).
1719 *
1720 * If soft masked, the masked handler will note the pending interrupt for
1721 * replay, and bump the decrementer to a high value, leaving MSR[EE] enabled
1722 * in the interrupted context.
1723 * If PPC_WATCHDOG is configured, the soft masked handler will actually set
1724 * things back up to run soft_nmi_interrupt as a regular interrupt handler
1725 * on the emergency stack.
1726 */
1727INT_DEFINE_BEGIN(decrementer)
1728	IVEC=0x900
1729	IMASK=IRQS_DISABLED
1730#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1731	IKVM_REAL=1
1732#endif
1733INT_DEFINE_END(decrementer)
1734
1735EXC_REAL_BEGIN(decrementer, 0x900, 0x80)
1736	GEN_INT_ENTRY decrementer, virt=0
1737EXC_REAL_END(decrementer, 0x900, 0x80)
1738EXC_VIRT_BEGIN(decrementer, 0x4900, 0x80)
1739	GEN_INT_ENTRY decrementer, virt=1
1740EXC_VIRT_END(decrementer, 0x4900, 0x80)
1741EXC_COMMON_BEGIN(decrementer_common)
1742	GEN_COMMON decrementer
1743	FINISH_NAP
1744	RUNLATCH_ON
1745	addi	r3,r1,STACK_FRAME_OVERHEAD
1746	bl	timer_interrupt
1747	b	interrupt_return
1748
1749	GEN_KVM decrementer
1750
1751
1752/**
1753 * Interrupt 0x980 - Hypervisor Decrementer Interrupt.
1754 * This is an asynchronous interrupt, similar to 0x900 but for the HDEC
1755 * register.
1756 *
1757 * Handling:
1758 * Linux does not use this outside KVM where it's used to keep a host timer
1759 * while the guest is given control of DEC. It should normally be caught by
1760 * the KVM test and routed there.
1761 */
1762INT_DEFINE_BEGIN(hdecrementer)
1763	IVEC=0x980
1764	IHSRR=1
1765	ISTACK=0
1766	IRECONCILE=0
1767	IKVM_REAL=1
1768	IKVM_VIRT=1
1769INT_DEFINE_END(hdecrementer)
1770
1771EXC_REAL_BEGIN(hdecrementer, 0x980, 0x80)
1772	GEN_INT_ENTRY hdecrementer, virt=0
1773EXC_REAL_END(hdecrementer, 0x980, 0x80)
1774EXC_VIRT_BEGIN(hdecrementer, 0x4980, 0x80)
1775	GEN_INT_ENTRY hdecrementer, virt=1
1776EXC_VIRT_END(hdecrementer, 0x4980, 0x80)
1777EXC_COMMON_BEGIN(hdecrementer_common)
1778	__GEN_COMMON_ENTRY hdecrementer
1779	/*
1780	 * Hypervisor decrementer interrupts not caught by the KVM test
1781	 * shouldn't occur but are sometimes left pending on exit from a KVM
1782	 * guest.  We don't need to do anything to clear them, as they are
1783	 * edge-triggered.
1784	 *
1785	 * Be careful to avoid touching the kernel stack.
1786	 */
1787	ld	r10,PACA_EXGEN+EX_CTR(r13)
1788	mtctr	r10
1789	mtcrf	0x80,r9
1790	ld	r9,PACA_EXGEN+EX_R9(r13)
1791	ld	r10,PACA_EXGEN+EX_R10(r13)
1792	ld	r11,PACA_EXGEN+EX_R11(r13)
1793	ld	r12,PACA_EXGEN+EX_R12(r13)
1794	ld	r13,PACA_EXGEN+EX_R13(r13)
1795	HRFI_TO_KERNEL
1796
1797	GEN_KVM hdecrementer
1798
1799
1800/**
1801 * Interrupt 0xa00 - Directed Privileged Doorbell Interrupt.
1802 * This is an asynchronous interrupt in response to a msgsndp doorbell.
1803 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
1804 * IRQS_DISABLED mask (i.e., local_irq_disable()).
1805 *
1806 * Handling:
1807 * Guests may use this for IPIs between threads in a core if the
1808 * hypervisor supports it. NVGPRS are not saved (see 0x500).
1809 *
1810 * If soft masked, the masked handler will note the pending interrupt for
1811 * replay, leaving MSR[EE] enabled in the interrupted context because the
1812 * doorbells are edge triggered.
1813 */
1814INT_DEFINE_BEGIN(doorbell_super)
1815	IVEC=0xa00
1816	IMASK=IRQS_DISABLED
1817#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1818	IKVM_REAL=1
1819#endif
1820INT_DEFINE_END(doorbell_super)
1821
1822EXC_REAL_BEGIN(doorbell_super, 0xa00, 0x100)
1823	GEN_INT_ENTRY doorbell_super, virt=0
1824EXC_REAL_END(doorbell_super, 0xa00, 0x100)
1825EXC_VIRT_BEGIN(doorbell_super, 0x4a00, 0x100)
1826	GEN_INT_ENTRY doorbell_super, virt=1
1827EXC_VIRT_END(doorbell_super, 0x4a00, 0x100)
1828EXC_COMMON_BEGIN(doorbell_super_common)
1829	GEN_COMMON doorbell_super
1830	FINISH_NAP
1831	RUNLATCH_ON
1832	addi	r3,r1,STACK_FRAME_OVERHEAD
1833#ifdef CONFIG_PPC_DOORBELL
1834	bl	doorbell_exception
1835#else
1836	bl	unknown_exception
1837#endif
1838	b	interrupt_return
1839
1840	GEN_KVM doorbell_super
1841
1842
1843EXC_REAL_NONE(0xb00, 0x100)
1844EXC_VIRT_NONE(0x4b00, 0x100)
1845
1846/**
1847 * Interrupt 0xc00 - System Call Interrupt (syscall, hcall).
1848 * This is a synchronous interrupt invoked with the "sc" instruction. The
1849 * system call is invoked with "sc 0" and does not alter the HV bit, so it
1850 * is directed to the currently running OS. The hypercall is invoked with
1851 * "sc 1" and it sets HV=1, so it elevates to hypervisor.
1852 *
1853 * In HPT, sc 1 always goes to 0xc00 real mode. In RADIX, sc 1 can go to
1854 * 0x4c00 virtual mode.
1855 *
1856 * Handling:
1857 * If the KVM test fires then it was due to a hypercall and is accordingly
1858 * routed to KVM. Otherwise this executes a normal Linux system call.
1859 *
1860 * Call convention:
1861 *
1862 * syscall and hypercalls register conventions are documented in
1863 * Documentation/powerpc/syscall64-abi.rst and
1864 * Documentation/powerpc/papr_hcalls.rst respectively.
1865 *
1866 * The intersection of volatile registers that don't contain possible
1867 * inputs is: cr0, xer, ctr. We may use these as scratch regs upon entry
1868 * without saving, though xer is not a good idea to use, as hardware may
1869 * interpret some bits so it may be costly to change them.
1870 */
1871INT_DEFINE_BEGIN(system_call)
1872	IVEC=0xc00
1873	IKVM_REAL=1
1874	IKVM_VIRT=1
1875INT_DEFINE_END(system_call)
1876
1877.macro SYSTEM_CALL virt
1878#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1879	/*
1880	 * There is a little bit of juggling to get syscall and hcall
1881	 * working well. Save r13 in ctr to avoid using SPRG scratch
1882	 * register.
1883	 *
1884	 * Userspace syscalls have already saved the PPR, hcalls must save
1885	 * it before setting HMT_MEDIUM.
1886	 */
1887	mtctr	r13
1888	GET_PACA(r13)
1889	std	r10,PACA_EXGEN+EX_R10(r13)
1890	INTERRUPT_TO_KERNEL
1891	KVMTEST system_call /* uses r10, branch to system_call_kvm */
1892	mfctr	r9
1893#else
1894	mr	r9,r13
1895	GET_PACA(r13)
1896	INTERRUPT_TO_KERNEL
1897#endif
1898
1899#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
1900BEGIN_FTR_SECTION
1901	cmpdi	r0,0x1ebe
1902	beq-	1f
1903END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE)
1904#endif
1905
1906	/* We reach here with PACA in r13, r13 in r9. */
1907	mfspr	r11,SPRN_SRR0
1908	mfspr	r12,SPRN_SRR1
1909
1910	HMT_MEDIUM
1911
1912	.if ! \virt
1913	__LOAD_HANDLER(r10, system_call_common)
1914	mtspr	SPRN_SRR0,r10
1915	ld	r10,PACAKMSR(r13)
1916	mtspr	SPRN_SRR1,r10
1917	RFI_TO_KERNEL
1918	b	.	/* prevent speculative execution */
1919	.else
1920	li	r10,MSR_RI
1921	mtmsrd 	r10,1			/* Set RI (EE=0) */
1922#ifdef CONFIG_RELOCATABLE
1923	__LOAD_HANDLER(r10, system_call_common)
1924	mtctr	r10
1925	bctr
1926#else
1927	b	system_call_common
1928#endif
1929	.endif
1930
1931#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
1932	/* Fast LE/BE switch system call */
19331:	mfspr	r12,SPRN_SRR1
1934	xori	r12,r12,MSR_LE
1935	mtspr	SPRN_SRR1,r12
1936	mr	r13,r9
1937	RFI_TO_USER	/* return to userspace */
1938	b	.	/* prevent speculative execution */
1939#endif
1940.endm
1941
1942EXC_REAL_BEGIN(system_call, 0xc00, 0x100)
1943	SYSTEM_CALL 0
1944EXC_REAL_END(system_call, 0xc00, 0x100)
1945EXC_VIRT_BEGIN(system_call, 0x4c00, 0x100)
1946	SYSTEM_CALL 1
1947EXC_VIRT_END(system_call, 0x4c00, 0x100)
1948
1949#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1950TRAMP_REAL_BEGIN(system_call_kvm)
1951	/*
1952	 * This is a hcall, so register convention is as above, with these
1953	 * differences:
1954	 * r13 = PACA
1955	 * ctr = orig r13
1956	 * orig r10 saved in PACA
1957	 */
1958	 /*
1959	  * Save the PPR (on systems that support it) before changing to
1960	  * HMT_MEDIUM. That allows the KVM code to save that value into the
1961	  * guest state (it is the guest's PPR value).
1962	  */
1963BEGIN_FTR_SECTION
1964	mfspr	r10,SPRN_PPR
1965	std	r10,HSTATE_PPR(r13)
1966END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
1967	HMT_MEDIUM
1968	mfctr	r10
1969	SET_SCRATCH0(r10)
1970	mfcr	r10
1971	std	r12,HSTATE_SCRATCH0(r13)
1972	sldi	r12,r10,32
1973	ori	r12,r12,0xc00
1974#ifdef CONFIG_RELOCATABLE
1975	/*
1976	 * Requires __LOAD_FAR_HANDLER beause kvmppc_interrupt lives
1977	 * outside the head section.
1978	 */
1979	__LOAD_FAR_HANDLER(r10, kvmppc_interrupt)
1980	mtctr   r10
1981	ld	r10,PACA_EXGEN+EX_R10(r13)
1982	bctr
1983#else
1984	ld	r10,PACA_EXGEN+EX_R10(r13)
1985	b       kvmppc_interrupt
1986#endif
1987#endif
1988
1989
1990/**
1991 * Interrupt 0xd00 - Trace Interrupt.
1992 * This is a synchronous interrupt in response to instruction step or
1993 * breakpoint faults.
1994 */
1995INT_DEFINE_BEGIN(single_step)
1996	IVEC=0xd00
1997#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1998	IKVM_REAL=1
1999#endif
2000INT_DEFINE_END(single_step)
2001
2002EXC_REAL_BEGIN(single_step, 0xd00, 0x100)
2003	GEN_INT_ENTRY single_step, virt=0
2004EXC_REAL_END(single_step, 0xd00, 0x100)
2005EXC_VIRT_BEGIN(single_step, 0x4d00, 0x100)
2006	GEN_INT_ENTRY single_step, virt=1
2007EXC_VIRT_END(single_step, 0x4d00, 0x100)
2008EXC_COMMON_BEGIN(single_step_common)
2009	GEN_COMMON single_step
2010	addi	r3,r1,STACK_FRAME_OVERHEAD
2011	bl	single_step_exception
2012	b	interrupt_return
2013
2014	GEN_KVM single_step
2015
2016
2017/**
2018 * Interrupt 0xe00 - Hypervisor Data Storage Interrupt (HDSI).
2019 * This is a synchronous interrupt in response to an MMU fault caused by a
2020 * guest data access.
2021 *
2022 * Handling:
2023 * This should always get routed to KVM. In radix MMU mode, this is caused
2024 * by a guest nested radix access that can't be performed due to the
2025 * partition scope page table. In hash mode, this can be caused by guests
2026 * running with translation disabled (virtual real mode) or with VPM enabled.
2027 * KVM will update the page table structures or disallow the access.
2028 */
2029INT_DEFINE_BEGIN(h_data_storage)
2030	IVEC=0xe00
2031	IHSRR=1
2032	IDAR=1
2033	IDSISR=1
2034	IKVM_SKIP=1
2035	IKVM_REAL=1
2036	IKVM_VIRT=1
2037INT_DEFINE_END(h_data_storage)
2038
2039EXC_REAL_BEGIN(h_data_storage, 0xe00, 0x20)
2040	GEN_INT_ENTRY h_data_storage, virt=0, ool=1
2041EXC_REAL_END(h_data_storage, 0xe00, 0x20)
2042EXC_VIRT_BEGIN(h_data_storage, 0x4e00, 0x20)
2043	GEN_INT_ENTRY h_data_storage, virt=1, ool=1
2044EXC_VIRT_END(h_data_storage, 0x4e00, 0x20)
2045EXC_COMMON_BEGIN(h_data_storage_common)
2046	GEN_COMMON h_data_storage
2047	addi    r3,r1,STACK_FRAME_OVERHEAD
2048BEGIN_MMU_FTR_SECTION
2049	ld	r4,_DAR(r1)
2050	li	r5,SIGSEGV
2051	bl      bad_page_fault
2052MMU_FTR_SECTION_ELSE
2053	bl      unknown_exception
2054ALT_MMU_FTR_SECTION_END_IFSET(MMU_FTR_TYPE_RADIX)
2055	b       interrupt_return
2056
2057	GEN_KVM h_data_storage
2058
2059
2060/**
2061 * Interrupt 0xe20 - Hypervisor Instruction Storage Interrupt (HISI).
2062 * This is a synchronous interrupt in response to an MMU fault caused by a
2063 * guest instruction fetch, similar to HDSI.
2064 */
2065INT_DEFINE_BEGIN(h_instr_storage)
2066	IVEC=0xe20
2067	IHSRR=1
2068	IKVM_REAL=1
2069	IKVM_VIRT=1
2070INT_DEFINE_END(h_instr_storage)
2071
2072EXC_REAL_BEGIN(h_instr_storage, 0xe20, 0x20)
2073	GEN_INT_ENTRY h_instr_storage, virt=0, ool=1
2074EXC_REAL_END(h_instr_storage, 0xe20, 0x20)
2075EXC_VIRT_BEGIN(h_instr_storage, 0x4e20, 0x20)
2076	GEN_INT_ENTRY h_instr_storage, virt=1, ool=1
2077EXC_VIRT_END(h_instr_storage, 0x4e20, 0x20)
2078EXC_COMMON_BEGIN(h_instr_storage_common)
2079	GEN_COMMON h_instr_storage
2080	addi	r3,r1,STACK_FRAME_OVERHEAD
2081	bl	unknown_exception
2082	b	interrupt_return
2083
2084	GEN_KVM h_instr_storage
2085
2086
2087/**
2088 * Interrupt 0xe40 - Hypervisor Emulation Assistance Interrupt.
2089 */
2090INT_DEFINE_BEGIN(emulation_assist)
2091	IVEC=0xe40
2092	IHSRR=1
2093	IKVM_REAL=1
2094	IKVM_VIRT=1
2095INT_DEFINE_END(emulation_assist)
2096
2097EXC_REAL_BEGIN(emulation_assist, 0xe40, 0x20)
2098	GEN_INT_ENTRY emulation_assist, virt=0, ool=1
2099EXC_REAL_END(emulation_assist, 0xe40, 0x20)
2100EXC_VIRT_BEGIN(emulation_assist, 0x4e40, 0x20)
2101	GEN_INT_ENTRY emulation_assist, virt=1, ool=1
2102EXC_VIRT_END(emulation_assist, 0x4e40, 0x20)
2103EXC_COMMON_BEGIN(emulation_assist_common)
2104	GEN_COMMON emulation_assist
2105	addi	r3,r1,STACK_FRAME_OVERHEAD
2106	bl	emulation_assist_interrupt
2107	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2108	b	interrupt_return
2109
2110	GEN_KVM emulation_assist
2111
2112
2113/**
2114 * Interrupt 0xe60 - Hypervisor Maintenance Interrupt (HMI).
2115 * This is an asynchronous interrupt caused by a Hypervisor Maintenance
2116 * Exception. It is always taken in real mode but uses HSRR registers
2117 * unlike SRESET and MCE.
2118 *
2119 * It is maskable in hardware by clearing MSR[EE], and partially soft-maskable
2120 * with IRQS_DISABLED mask (i.e., local_irq_disable()).
2121 *
2122 * Handling:
2123 * This is a special case, this is handled similarly to machine checks, with an
2124 * initial real mode handler that is not soft-masked, which attempts to fix the
2125 * problem. Then a regular handler which is soft-maskable and reports the
2126 * problem.
2127 *
2128 * The emergency stack is used for the early real mode handler.
2129 *
2130 * XXX: unclear why MCE and HMI schemes could not be made common, e.g.,
2131 * either use soft-masking for the MCE, or use irq_work for the HMI.
2132 *
2133 * KVM:
2134 * Unlike MCE, this calls into KVM without calling the real mode handler
2135 * first.
2136 */
2137INT_DEFINE_BEGIN(hmi_exception_early)
2138	IVEC=0xe60
2139	IHSRR=1
2140	IREALMODE_COMMON=1
2141	ISTACK=0
2142	IRECONCILE=0
2143	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
2144	IKVM_REAL=1
2145INT_DEFINE_END(hmi_exception_early)
2146
2147INT_DEFINE_BEGIN(hmi_exception)
2148	IVEC=0xe60
2149	IHSRR=1
2150	IMASK=IRQS_DISABLED
2151	IKVM_REAL=1
2152INT_DEFINE_END(hmi_exception)
2153
2154EXC_REAL_BEGIN(hmi_exception, 0xe60, 0x20)
2155	GEN_INT_ENTRY hmi_exception_early, virt=0, ool=1
2156EXC_REAL_END(hmi_exception, 0xe60, 0x20)
2157EXC_VIRT_NONE(0x4e60, 0x20)
2158
2159EXC_COMMON_BEGIN(hmi_exception_early_common)
2160	__GEN_REALMODE_COMMON_ENTRY hmi_exception_early
2161
2162	mr	r10,r1			/* Save r1 */
2163	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack for realmode */
2164	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
2165
2166	__GEN_COMMON_BODY hmi_exception_early
2167
2168	addi	r3,r1,STACK_FRAME_OVERHEAD
2169	bl	hmi_exception_realmode
2170	cmpdi	cr0,r3,0
2171	bne	1f
2172
2173	EXCEPTION_RESTORE_REGS hsrr=1
2174	HRFI_TO_USER_OR_KERNEL
2175
21761:
2177	/*
2178	 * Go to virtual mode and pull the HMI event information from
2179	 * firmware.
2180	 */
2181	EXCEPTION_RESTORE_REGS hsrr=1
2182	GEN_INT_ENTRY hmi_exception, virt=0
2183
2184	GEN_KVM hmi_exception_early
2185
2186EXC_COMMON_BEGIN(hmi_exception_common)
2187	GEN_COMMON hmi_exception
2188	FINISH_NAP
2189	RUNLATCH_ON
2190	addi	r3,r1,STACK_FRAME_OVERHEAD
2191	bl	handle_hmi_exception
2192	b	interrupt_return
2193
2194	GEN_KVM hmi_exception
2195
2196
2197/**
2198 * Interrupt 0xe80 - Directed Hypervisor Doorbell Interrupt.
2199 * This is an asynchronous interrupt in response to a msgsnd doorbell.
2200 * Similar to the 0xa00 doorbell but for host rather than guest.
2201 */
2202INT_DEFINE_BEGIN(h_doorbell)
2203	IVEC=0xe80
2204	IHSRR=1
2205	IMASK=IRQS_DISABLED
2206	IKVM_REAL=1
2207	IKVM_VIRT=1
2208INT_DEFINE_END(h_doorbell)
2209
2210EXC_REAL_BEGIN(h_doorbell, 0xe80, 0x20)
2211	GEN_INT_ENTRY h_doorbell, virt=0, ool=1
2212EXC_REAL_END(h_doorbell, 0xe80, 0x20)
2213EXC_VIRT_BEGIN(h_doorbell, 0x4e80, 0x20)
2214	GEN_INT_ENTRY h_doorbell, virt=1, ool=1
2215EXC_VIRT_END(h_doorbell, 0x4e80, 0x20)
2216EXC_COMMON_BEGIN(h_doorbell_common)
2217	GEN_COMMON h_doorbell
2218	FINISH_NAP
2219	RUNLATCH_ON
2220	addi	r3,r1,STACK_FRAME_OVERHEAD
2221#ifdef CONFIG_PPC_DOORBELL
2222	bl	doorbell_exception
2223#else
2224	bl	unknown_exception
2225#endif
2226	b	interrupt_return
2227
2228	GEN_KVM h_doorbell
2229
2230
2231/**
2232 * Interrupt 0xea0 - Hypervisor Virtualization Interrupt.
2233 * This is an asynchronous interrupt in response to an "external exception".
2234 * Similar to 0x500 but for host only.
2235 */
2236INT_DEFINE_BEGIN(h_virt_irq)
2237	IVEC=0xea0
2238	IHSRR=1
2239	IMASK=IRQS_DISABLED
2240	IKVM_REAL=1
2241	IKVM_VIRT=1
2242INT_DEFINE_END(h_virt_irq)
2243
2244EXC_REAL_BEGIN(h_virt_irq, 0xea0, 0x20)
2245	GEN_INT_ENTRY h_virt_irq, virt=0, ool=1
2246EXC_REAL_END(h_virt_irq, 0xea0, 0x20)
2247EXC_VIRT_BEGIN(h_virt_irq, 0x4ea0, 0x20)
2248	GEN_INT_ENTRY h_virt_irq, virt=1, ool=1
2249EXC_VIRT_END(h_virt_irq, 0x4ea0, 0x20)
2250EXC_COMMON_BEGIN(h_virt_irq_common)
2251	GEN_COMMON h_virt_irq
2252	FINISH_NAP
2253	RUNLATCH_ON
2254	addi	r3,r1,STACK_FRAME_OVERHEAD
2255	bl	do_IRQ
2256	b	interrupt_return
2257
2258	GEN_KVM h_virt_irq
2259
2260
2261EXC_REAL_NONE(0xec0, 0x20)
2262EXC_VIRT_NONE(0x4ec0, 0x20)
2263EXC_REAL_NONE(0xee0, 0x20)
2264EXC_VIRT_NONE(0x4ee0, 0x20)
2265
2266
2267/*
2268 * Interrupt 0xf00 - Performance Monitor Interrupt (PMI, PMU).
2269 * This is an asynchronous interrupt in response to a PMU exception.
2270 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
2271 * IRQS_PMI_DISABLED mask (NOTE: NOT local_irq_disable()).
2272 *
2273 * Handling:
2274 * This calls into the perf subsystem.
2275 *
2276 * Like the watchdog soft-nmi, it appears an NMI interrupt to Linux, in that it
2277 * runs under local_irq_disable. However it may be soft-masked in
2278 * powerpc-specific code.
2279 *
2280 * If soft masked, the masked handler will note the pending interrupt for
2281 * replay, and clear MSR[EE] in the interrupted context.
2282 */
2283INT_DEFINE_BEGIN(performance_monitor)
2284	IVEC=0xf00
2285	IMASK=IRQS_PMI_DISABLED
2286#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2287	IKVM_REAL=1
2288#endif
2289INT_DEFINE_END(performance_monitor)
2290
2291EXC_REAL_BEGIN(performance_monitor, 0xf00, 0x20)
2292	GEN_INT_ENTRY performance_monitor, virt=0, ool=1
2293EXC_REAL_END(performance_monitor, 0xf00, 0x20)
2294EXC_VIRT_BEGIN(performance_monitor, 0x4f00, 0x20)
2295	GEN_INT_ENTRY performance_monitor, virt=1, ool=1
2296EXC_VIRT_END(performance_monitor, 0x4f00, 0x20)
2297EXC_COMMON_BEGIN(performance_monitor_common)
2298	GEN_COMMON performance_monitor
2299	FINISH_NAP
2300	RUNLATCH_ON
2301	addi	r3,r1,STACK_FRAME_OVERHEAD
2302	bl	performance_monitor_exception
2303	b	interrupt_return
2304
2305	GEN_KVM performance_monitor
2306
2307
2308/**
2309 * Interrupt 0xf20 - Vector Unavailable Interrupt.
2310 * This is a synchronous interrupt in response to
2311 * executing a vector (or altivec) instruction with MSR[VEC]=0.
2312 * Similar to FP unavailable.
2313 */
2314INT_DEFINE_BEGIN(altivec_unavailable)
2315	IVEC=0xf20
2316	IRECONCILE=0
2317#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2318	IKVM_REAL=1
2319#endif
2320INT_DEFINE_END(altivec_unavailable)
2321
2322EXC_REAL_BEGIN(altivec_unavailable, 0xf20, 0x20)
2323	GEN_INT_ENTRY altivec_unavailable, virt=0, ool=1
2324EXC_REAL_END(altivec_unavailable, 0xf20, 0x20)
2325EXC_VIRT_BEGIN(altivec_unavailable, 0x4f20, 0x20)
2326	GEN_INT_ENTRY altivec_unavailable, virt=1, ool=1
2327EXC_VIRT_END(altivec_unavailable, 0x4f20, 0x20)
2328EXC_COMMON_BEGIN(altivec_unavailable_common)
2329	GEN_COMMON altivec_unavailable
2330#ifdef CONFIG_ALTIVEC
2331BEGIN_FTR_SECTION
2332	beq	1f
2333#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2334  BEGIN_FTR_SECTION_NESTED(69)
2335	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2336	 * transaction), go do TM stuff
2337	 */
2338	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2339	bne-	2f
2340  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2341#endif
2342	bl	load_up_altivec
2343	b	fast_interrupt_return
2344#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
23452:	/* User process was in a transaction */
2346	RECONCILE_IRQ_STATE(r10, r11)
2347	addi	r3,r1,STACK_FRAME_OVERHEAD
2348	bl	altivec_unavailable_tm
2349	b	interrupt_return
2350#endif
23511:
2352END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2353#endif
2354	RECONCILE_IRQ_STATE(r10, r11)
2355	addi	r3,r1,STACK_FRAME_OVERHEAD
2356	bl	altivec_unavailable_exception
2357	b	interrupt_return
2358
2359	GEN_KVM altivec_unavailable
2360
2361
2362/**
2363 * Interrupt 0xf40 - VSX Unavailable Interrupt.
2364 * This is a synchronous interrupt in response to
2365 * executing a VSX instruction with MSR[VSX]=0.
2366 * Similar to FP unavailable.
2367 */
2368INT_DEFINE_BEGIN(vsx_unavailable)
2369	IVEC=0xf40
2370	IRECONCILE=0
2371#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2372	IKVM_REAL=1
2373#endif
2374INT_DEFINE_END(vsx_unavailable)
2375
2376EXC_REAL_BEGIN(vsx_unavailable, 0xf40, 0x20)
2377	GEN_INT_ENTRY vsx_unavailable, virt=0, ool=1
2378EXC_REAL_END(vsx_unavailable, 0xf40, 0x20)
2379EXC_VIRT_BEGIN(vsx_unavailable, 0x4f40, 0x20)
2380	GEN_INT_ENTRY vsx_unavailable, virt=1, ool=1
2381EXC_VIRT_END(vsx_unavailable, 0x4f40, 0x20)
2382EXC_COMMON_BEGIN(vsx_unavailable_common)
2383	GEN_COMMON vsx_unavailable
2384#ifdef CONFIG_VSX
2385BEGIN_FTR_SECTION
2386	beq	1f
2387#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2388  BEGIN_FTR_SECTION_NESTED(69)
2389	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2390	 * transaction), go do TM stuff
2391	 */
2392	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2393	bne-	2f
2394  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2395#endif
2396	b	load_up_vsx
2397#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
23982:	/* User process was in a transaction */
2399	RECONCILE_IRQ_STATE(r10, r11)
2400	addi	r3,r1,STACK_FRAME_OVERHEAD
2401	bl	vsx_unavailable_tm
2402	b	interrupt_return
2403#endif
24041:
2405END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2406#endif
2407	RECONCILE_IRQ_STATE(r10, r11)
2408	addi	r3,r1,STACK_FRAME_OVERHEAD
2409	bl	vsx_unavailable_exception
2410	b	interrupt_return
2411
2412	GEN_KVM vsx_unavailable
2413
2414
2415/**
2416 * Interrupt 0xf60 - Facility Unavailable Interrupt.
2417 * This is a synchronous interrupt in response to
2418 * executing an instruction without access to the facility that can be
2419 * resolved by the OS (e.g., FSCR, MSR).
2420 * Similar to FP unavailable.
2421 */
2422INT_DEFINE_BEGIN(facility_unavailable)
2423	IVEC=0xf60
2424#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2425	IKVM_REAL=1
2426#endif
2427INT_DEFINE_END(facility_unavailable)
2428
2429EXC_REAL_BEGIN(facility_unavailable, 0xf60, 0x20)
2430	GEN_INT_ENTRY facility_unavailable, virt=0, ool=1
2431EXC_REAL_END(facility_unavailable, 0xf60, 0x20)
2432EXC_VIRT_BEGIN(facility_unavailable, 0x4f60, 0x20)
2433	GEN_INT_ENTRY facility_unavailable, virt=1, ool=1
2434EXC_VIRT_END(facility_unavailable, 0x4f60, 0x20)
2435EXC_COMMON_BEGIN(facility_unavailable_common)
2436	GEN_COMMON facility_unavailable
2437	addi	r3,r1,STACK_FRAME_OVERHEAD
2438	bl	facility_unavailable_exception
2439	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2440	b	interrupt_return
2441
2442	GEN_KVM facility_unavailable
2443
2444
2445/**
2446 * Interrupt 0xf60 - Hypervisor Facility Unavailable Interrupt.
2447 * This is a synchronous interrupt in response to
2448 * executing an instruction without access to the facility that can only
2449 * be resolved in HV mode (e.g., HFSCR).
2450 * Similar to FP unavailable.
2451 */
2452INT_DEFINE_BEGIN(h_facility_unavailable)
2453	IVEC=0xf80
2454	IHSRR=1
2455	IKVM_REAL=1
2456	IKVM_VIRT=1
2457INT_DEFINE_END(h_facility_unavailable)
2458
2459EXC_REAL_BEGIN(h_facility_unavailable, 0xf80, 0x20)
2460	GEN_INT_ENTRY h_facility_unavailable, virt=0, ool=1
2461EXC_REAL_END(h_facility_unavailable, 0xf80, 0x20)
2462EXC_VIRT_BEGIN(h_facility_unavailable, 0x4f80, 0x20)
2463	GEN_INT_ENTRY h_facility_unavailable, virt=1, ool=1
2464EXC_VIRT_END(h_facility_unavailable, 0x4f80, 0x20)
2465EXC_COMMON_BEGIN(h_facility_unavailable_common)
2466	GEN_COMMON h_facility_unavailable
2467	addi	r3,r1,STACK_FRAME_OVERHEAD
2468	bl	facility_unavailable_exception
2469	REST_NVGPRS(r1) /* XXX Shouldn't be necessary in practice */
2470	b	interrupt_return
2471
2472	GEN_KVM h_facility_unavailable
2473
2474
2475EXC_REAL_NONE(0xfa0, 0x20)
2476EXC_VIRT_NONE(0x4fa0, 0x20)
2477EXC_REAL_NONE(0xfc0, 0x20)
2478EXC_VIRT_NONE(0x4fc0, 0x20)
2479EXC_REAL_NONE(0xfe0, 0x20)
2480EXC_VIRT_NONE(0x4fe0, 0x20)
2481
2482EXC_REAL_NONE(0x1000, 0x100)
2483EXC_VIRT_NONE(0x5000, 0x100)
2484EXC_REAL_NONE(0x1100, 0x100)
2485EXC_VIRT_NONE(0x5100, 0x100)
2486
2487#ifdef CONFIG_CBE_RAS
2488INT_DEFINE_BEGIN(cbe_system_error)
2489	IVEC=0x1200
2490	IHSRR=1
2491	IKVM_SKIP=1
2492	IKVM_REAL=1
2493INT_DEFINE_END(cbe_system_error)
2494
2495EXC_REAL_BEGIN(cbe_system_error, 0x1200, 0x100)
2496	GEN_INT_ENTRY cbe_system_error, virt=0
2497EXC_REAL_END(cbe_system_error, 0x1200, 0x100)
2498EXC_VIRT_NONE(0x5200, 0x100)
2499EXC_COMMON_BEGIN(cbe_system_error_common)
2500	GEN_COMMON cbe_system_error
2501	addi	r3,r1,STACK_FRAME_OVERHEAD
2502	bl	cbe_system_error_exception
2503	b	interrupt_return
2504
2505	GEN_KVM cbe_system_error
2506
2507#else /* CONFIG_CBE_RAS */
2508EXC_REAL_NONE(0x1200, 0x100)
2509EXC_VIRT_NONE(0x5200, 0x100)
2510#endif
2511
2512
2513INT_DEFINE_BEGIN(instruction_breakpoint)
2514	IVEC=0x1300
2515#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2516	IKVM_SKIP=1
2517	IKVM_REAL=1
2518#endif
2519INT_DEFINE_END(instruction_breakpoint)
2520
2521EXC_REAL_BEGIN(instruction_breakpoint, 0x1300, 0x100)
2522	GEN_INT_ENTRY instruction_breakpoint, virt=0
2523EXC_REAL_END(instruction_breakpoint, 0x1300, 0x100)
2524EXC_VIRT_BEGIN(instruction_breakpoint, 0x5300, 0x100)
2525	GEN_INT_ENTRY instruction_breakpoint, virt=1
2526EXC_VIRT_END(instruction_breakpoint, 0x5300, 0x100)
2527EXC_COMMON_BEGIN(instruction_breakpoint_common)
2528	GEN_COMMON instruction_breakpoint
2529	addi	r3,r1,STACK_FRAME_OVERHEAD
2530	bl	instruction_breakpoint_exception
2531	b	interrupt_return
2532
2533	GEN_KVM instruction_breakpoint
2534
2535
2536EXC_REAL_NONE(0x1400, 0x100)
2537EXC_VIRT_NONE(0x5400, 0x100)
2538
2539/**
2540 * Interrupt 0x1500 - Soft Patch Interrupt
2541 *
2542 * Handling:
2543 * This is an implementation specific interrupt which can be used for a
2544 * range of exceptions.
2545 *
2546 * This interrupt handler is unique in that it runs the denormal assist
2547 * code even for guests (and even in guest context) without going to KVM,
2548 * for speed. POWER9 does not raise denorm exceptions, so this special case
2549 * could be phased out in future to reduce special cases.
2550 */
2551INT_DEFINE_BEGIN(denorm_exception)
2552	IVEC=0x1500
2553	IHSRR=1
2554	IBRANCH_COMMON=0
2555	IKVM_REAL=1
2556INT_DEFINE_END(denorm_exception)
2557
2558EXC_REAL_BEGIN(denorm_exception, 0x1500, 0x100)
2559	GEN_INT_ENTRY denorm_exception, virt=0
2560#ifdef CONFIG_PPC_DENORMALISATION
2561	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2562	bne+	denorm_assist
2563#endif
2564	GEN_BRANCH_TO_COMMON denorm_exception, virt=0
2565EXC_REAL_END(denorm_exception, 0x1500, 0x100)
2566#ifdef CONFIG_PPC_DENORMALISATION
2567EXC_VIRT_BEGIN(denorm_exception, 0x5500, 0x100)
2568	GEN_INT_ENTRY denorm_exception, virt=1
2569	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2570	bne+	denorm_assist
2571	GEN_BRANCH_TO_COMMON denorm_exception, virt=1
2572EXC_VIRT_END(denorm_exception, 0x5500, 0x100)
2573#else
2574EXC_VIRT_NONE(0x5500, 0x100)
2575#endif
2576
2577#ifdef CONFIG_PPC_DENORMALISATION
2578TRAMP_REAL_BEGIN(denorm_assist)
2579BEGIN_FTR_SECTION
2580/*
2581 * To denormalise we need to move a copy of the register to itself.
2582 * For POWER6 do that here for all FP regs.
2583 */
2584	mfmsr	r10
2585	ori	r10,r10,(MSR_FP|MSR_FE0|MSR_FE1)
2586	xori	r10,r10,(MSR_FE0|MSR_FE1)
2587	mtmsrd	r10
2588	sync
2589
2590	.Lreg=0
2591	.rept 32
2592	fmr	.Lreg,.Lreg
2593	.Lreg=.Lreg+1
2594	.endr
2595
2596FTR_SECTION_ELSE
2597/*
2598 * To denormalise we need to move a copy of the register to itself.
2599 * For POWER7 do that here for the first 32 VSX registers only.
2600 */
2601	mfmsr	r10
2602	oris	r10,r10,MSR_VSX@h
2603	mtmsrd	r10
2604	sync
2605
2606	.Lreg=0
2607	.rept 32
2608	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2609	.Lreg=.Lreg+1
2610	.endr
2611
2612ALT_FTR_SECTION_END_IFCLR(CPU_FTR_ARCH_206)
2613
2614BEGIN_FTR_SECTION
2615	b	denorm_done
2616END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
2617/*
2618 * To denormalise we need to move a copy of the register to itself.
2619 * For POWER8 we need to do that for all 64 VSX registers
2620 */
2621	.Lreg=32
2622	.rept 32
2623	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2624	.Lreg=.Lreg+1
2625	.endr
2626
2627denorm_done:
2628	mfspr	r11,SPRN_HSRR0
2629	subi	r11,r11,4
2630	mtspr	SPRN_HSRR0,r11
2631	mtcrf	0x80,r9
2632	ld	r9,PACA_EXGEN+EX_R9(r13)
2633BEGIN_FTR_SECTION
2634	ld	r10,PACA_EXGEN+EX_PPR(r13)
2635	mtspr	SPRN_PPR,r10
2636END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
2637BEGIN_FTR_SECTION
2638	ld	r10,PACA_EXGEN+EX_CFAR(r13)
2639	mtspr	SPRN_CFAR,r10
2640END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
2641	ld	r10,PACA_EXGEN+EX_R10(r13)
2642	ld	r11,PACA_EXGEN+EX_R11(r13)
2643	ld	r12,PACA_EXGEN+EX_R12(r13)
2644	ld	r13,PACA_EXGEN+EX_R13(r13)
2645	HRFI_TO_UNKNOWN
2646	b	.
2647#endif
2648
2649EXC_COMMON_BEGIN(denorm_exception_common)
2650	GEN_COMMON denorm_exception
2651	addi	r3,r1,STACK_FRAME_OVERHEAD
2652	bl	unknown_exception
2653	b	interrupt_return
2654
2655	GEN_KVM denorm_exception
2656
2657
2658#ifdef CONFIG_CBE_RAS
2659INT_DEFINE_BEGIN(cbe_maintenance)
2660	IVEC=0x1600
2661	IHSRR=1
2662	IKVM_SKIP=1
2663	IKVM_REAL=1
2664INT_DEFINE_END(cbe_maintenance)
2665
2666EXC_REAL_BEGIN(cbe_maintenance, 0x1600, 0x100)
2667	GEN_INT_ENTRY cbe_maintenance, virt=0
2668EXC_REAL_END(cbe_maintenance, 0x1600, 0x100)
2669EXC_VIRT_NONE(0x5600, 0x100)
2670EXC_COMMON_BEGIN(cbe_maintenance_common)
2671	GEN_COMMON cbe_maintenance
2672	addi	r3,r1,STACK_FRAME_OVERHEAD
2673	bl	cbe_maintenance_exception
2674	b	interrupt_return
2675
2676	GEN_KVM cbe_maintenance
2677
2678#else /* CONFIG_CBE_RAS */
2679EXC_REAL_NONE(0x1600, 0x100)
2680EXC_VIRT_NONE(0x5600, 0x100)
2681#endif
2682
2683
2684INT_DEFINE_BEGIN(altivec_assist)
2685	IVEC=0x1700
2686#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2687	IKVM_REAL=1
2688#endif
2689INT_DEFINE_END(altivec_assist)
2690
2691EXC_REAL_BEGIN(altivec_assist, 0x1700, 0x100)
2692	GEN_INT_ENTRY altivec_assist, virt=0
2693EXC_REAL_END(altivec_assist, 0x1700, 0x100)
2694EXC_VIRT_BEGIN(altivec_assist, 0x5700, 0x100)
2695	GEN_INT_ENTRY altivec_assist, virt=1
2696EXC_VIRT_END(altivec_assist, 0x5700, 0x100)
2697EXC_COMMON_BEGIN(altivec_assist_common)
2698	GEN_COMMON altivec_assist
2699	addi	r3,r1,STACK_FRAME_OVERHEAD
2700#ifdef CONFIG_ALTIVEC
2701	bl	altivec_assist_exception
2702	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2703#else
2704	bl	unknown_exception
2705#endif
2706	b	interrupt_return
2707
2708	GEN_KVM altivec_assist
2709
2710
2711#ifdef CONFIG_CBE_RAS
2712INT_DEFINE_BEGIN(cbe_thermal)
2713	IVEC=0x1800
2714	IHSRR=1
2715	IKVM_SKIP=1
2716	IKVM_REAL=1
2717INT_DEFINE_END(cbe_thermal)
2718
2719EXC_REAL_BEGIN(cbe_thermal, 0x1800, 0x100)
2720	GEN_INT_ENTRY cbe_thermal, virt=0
2721EXC_REAL_END(cbe_thermal, 0x1800, 0x100)
2722EXC_VIRT_NONE(0x5800, 0x100)
2723EXC_COMMON_BEGIN(cbe_thermal_common)
2724	GEN_COMMON cbe_thermal
2725	addi	r3,r1,STACK_FRAME_OVERHEAD
2726	bl	cbe_thermal_exception
2727	b	interrupt_return
2728
2729	GEN_KVM cbe_thermal
2730
2731#else /* CONFIG_CBE_RAS */
2732EXC_REAL_NONE(0x1800, 0x100)
2733EXC_VIRT_NONE(0x5800, 0x100)
2734#endif
2735
2736
2737#ifdef CONFIG_PPC_WATCHDOG
2738
2739INT_DEFINE_BEGIN(soft_nmi)
2740	IVEC=0x900
2741	ISTACK=0
2742	IRECONCILE=0	/* Soft-NMI may fire under local_irq_disable */
2743INT_DEFINE_END(soft_nmi)
2744
2745/*
2746 * Branch to soft_nmi_interrupt using the emergency stack. The emergency
2747 * stack is one that is usable by maskable interrupts so long as MSR_EE
2748 * remains off. It is used for recovery when something has corrupted the
2749 * normal kernel stack, for example. The "soft NMI" must not use the process
2750 * stack because we want irq disabled sections to avoid touching the stack
2751 * at all (other than PMU interrupts), so use the emergency stack for this,
2752 * and run it entirely with interrupts hard disabled.
2753 */
2754EXC_COMMON_BEGIN(soft_nmi_common)
2755	mfspr	r11,SPRN_SRR0
2756	mr	r10,r1
2757	ld	r1,PACAEMERGSP(r13)
2758	subi	r1,r1,INT_FRAME_SIZE
2759	__GEN_COMMON_BODY soft_nmi
2760
2761	/*
2762	 * Set IRQS_ALL_DISABLED and save PACAIRQHAPPENED (see
2763	 * system_reset_common)
2764	 */
2765	li	r10,IRQS_ALL_DISABLED
2766	stb	r10,PACAIRQSOFTMASK(r13)
2767	lbz	r10,PACAIRQHAPPENED(r13)
2768	std	r10,RESULT(r1)
2769	ori	r10,r10,PACA_IRQ_HARD_DIS
2770	stb	r10,PACAIRQHAPPENED(r13)
2771
2772	addi	r3,r1,STACK_FRAME_OVERHEAD
2773	bl	soft_nmi_interrupt
2774
2775	/* Clear MSR_RI before setting SRR0 and SRR1. */
2776	li	r9,0
2777	mtmsrd	r9,1
2778
2779	/*
2780	 * Restore soft mask settings.
2781	 */
2782	ld	r10,RESULT(r1)
2783	stb	r10,PACAIRQHAPPENED(r13)
2784	ld	r10,SOFTE(r1)
2785	stb	r10,PACAIRQSOFTMASK(r13)
2786
2787	kuap_restore_amr r9, r10
2788	EXCEPTION_RESTORE_REGS hsrr=0
2789	RFI_TO_KERNEL
2790
2791#endif /* CONFIG_PPC_WATCHDOG */
2792
2793/*
2794 * An interrupt came in while soft-disabled. We set paca->irq_happened, then:
2795 * - If it was a decrementer interrupt, we bump the dec to max and and return.
2796 * - If it was a doorbell we return immediately since doorbells are edge
2797 *   triggered and won't automatically refire.
2798 * - If it was a HMI we return immediately since we handled it in realmode
2799 *   and it won't refire.
2800 * - Else it is one of PACA_IRQ_MUST_HARD_MASK, so hard disable and return.
2801 * This is called with r10 containing the value to OR to the paca field.
2802 */
2803.macro MASKED_INTERRUPT hsrr=0
2804	.if \hsrr
2805masked_Hinterrupt:
2806	.else
2807masked_interrupt:
2808	.endif
2809	lbz	r11,PACAIRQHAPPENED(r13)
2810	or	r11,r11,r10
2811	stb	r11,PACAIRQHAPPENED(r13)
2812	cmpwi	r10,PACA_IRQ_DEC
2813	bne	1f
2814	lis	r10,0x7fff
2815	ori	r10,r10,0xffff
2816	mtspr	SPRN_DEC,r10
2817#ifdef CONFIG_PPC_WATCHDOG
2818	b	soft_nmi_common
2819#else
2820	b	2f
2821#endif
28221:	andi.	r10,r10,PACA_IRQ_MUST_HARD_MASK
2823	beq	2f
2824	xori	r12,r12,MSR_EE	/* clear MSR_EE */
2825	.if \hsrr
2826	mtspr	SPRN_HSRR1,r12
2827	.else
2828	mtspr	SPRN_SRR1,r12
2829	.endif
2830	ori	r11,r11,PACA_IRQ_HARD_DIS
2831	stb	r11,PACAIRQHAPPENED(r13)
28322:	/* done */
2833	ld	r10,PACA_EXGEN+EX_CTR(r13)
2834	mtctr	r10
2835	mtcrf	0x80,r9
2836	std	r1,PACAR1(r13)
2837	ld	r9,PACA_EXGEN+EX_R9(r13)
2838	ld	r10,PACA_EXGEN+EX_R10(r13)
2839	ld	r11,PACA_EXGEN+EX_R11(r13)
2840	ld	r12,PACA_EXGEN+EX_R12(r13)
2841	/* returns to kernel where r13 must be set up, so don't restore it */
2842	.if \hsrr
2843	HRFI_TO_KERNEL
2844	.else
2845	RFI_TO_KERNEL
2846	.endif
2847	b	.
2848.endm
2849
2850TRAMP_REAL_BEGIN(stf_barrier_fallback)
2851	std	r9,PACA_EXRFI+EX_R9(r13)
2852	std	r10,PACA_EXRFI+EX_R10(r13)
2853	sync
2854	ld	r9,PACA_EXRFI+EX_R9(r13)
2855	ld	r10,PACA_EXRFI+EX_R10(r13)
2856	ori	31,31,0
2857	.rept 14
2858	b	1f
28591:
2860	.endr
2861	blr
2862
2863TRAMP_REAL_BEGIN(rfi_flush_fallback)
2864	SET_SCRATCH0(r13);
2865	GET_PACA(r13);
2866	std	r1,PACA_EXRFI+EX_R12(r13)
2867	ld	r1,PACAKSAVE(r13)
2868	std	r9,PACA_EXRFI+EX_R9(r13)
2869	std	r10,PACA_EXRFI+EX_R10(r13)
2870	std	r11,PACA_EXRFI+EX_R11(r13)
2871	mfctr	r9
2872	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
2873	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
2874	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
2875	mtctr	r11
2876	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
2877
2878	/* order ld/st prior to dcbt stop all streams with flushing */
2879	sync
2880
2881	/*
2882	 * The load adresses are at staggered offsets within cachelines,
2883	 * which suits some pipelines better (on others it should not
2884	 * hurt).
2885	 */
28861:
2887	ld	r11,(0x80 + 8)*0(r10)
2888	ld	r11,(0x80 + 8)*1(r10)
2889	ld	r11,(0x80 + 8)*2(r10)
2890	ld	r11,(0x80 + 8)*3(r10)
2891	ld	r11,(0x80 + 8)*4(r10)
2892	ld	r11,(0x80 + 8)*5(r10)
2893	ld	r11,(0x80 + 8)*6(r10)
2894	ld	r11,(0x80 + 8)*7(r10)
2895	addi	r10,r10,0x80*8
2896	bdnz	1b
2897
2898	mtctr	r9
2899	ld	r9,PACA_EXRFI+EX_R9(r13)
2900	ld	r10,PACA_EXRFI+EX_R10(r13)
2901	ld	r11,PACA_EXRFI+EX_R11(r13)
2902	ld	r1,PACA_EXRFI+EX_R12(r13)
2903	GET_SCRATCH0(r13);
2904	rfid
2905
2906TRAMP_REAL_BEGIN(hrfi_flush_fallback)
2907	SET_SCRATCH0(r13);
2908	GET_PACA(r13);
2909	std	r1,PACA_EXRFI+EX_R12(r13)
2910	ld	r1,PACAKSAVE(r13)
2911	std	r9,PACA_EXRFI+EX_R9(r13)
2912	std	r10,PACA_EXRFI+EX_R10(r13)
2913	std	r11,PACA_EXRFI+EX_R11(r13)
2914	mfctr	r9
2915	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
2916	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
2917	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
2918	mtctr	r11
2919	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
2920
2921	/* order ld/st prior to dcbt stop all streams with flushing */
2922	sync
2923
2924	/*
2925	 * The load adresses are at staggered offsets within cachelines,
2926	 * which suits some pipelines better (on others it should not
2927	 * hurt).
2928	 */
29291:
2930	ld	r11,(0x80 + 8)*0(r10)
2931	ld	r11,(0x80 + 8)*1(r10)
2932	ld	r11,(0x80 + 8)*2(r10)
2933	ld	r11,(0x80 + 8)*3(r10)
2934	ld	r11,(0x80 + 8)*4(r10)
2935	ld	r11,(0x80 + 8)*5(r10)
2936	ld	r11,(0x80 + 8)*6(r10)
2937	ld	r11,(0x80 + 8)*7(r10)
2938	addi	r10,r10,0x80*8
2939	bdnz	1b
2940
2941	mtctr	r9
2942	ld	r9,PACA_EXRFI+EX_R9(r13)
2943	ld	r10,PACA_EXRFI+EX_R10(r13)
2944	ld	r11,PACA_EXRFI+EX_R11(r13)
2945	ld	r1,PACA_EXRFI+EX_R12(r13)
2946	GET_SCRATCH0(r13);
2947	hrfid
2948
2949USE_TEXT_SECTION()
2950	MASKED_INTERRUPT
2951	MASKED_INTERRUPT hsrr=1
2952
2953#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
2954kvmppc_skip_interrupt:
2955	/*
2956	 * Here all GPRs are unchanged from when the interrupt happened
2957	 * except for r13, which is saved in SPRG_SCRATCH0.
2958	 */
2959	mfspr	r13, SPRN_SRR0
2960	addi	r13, r13, 4
2961	mtspr	SPRN_SRR0, r13
2962	GET_SCRATCH0(r13)
2963	RFI_TO_KERNEL
2964	b	.
2965
2966kvmppc_skip_Hinterrupt:
2967	/*
2968	 * Here all GPRs are unchanged from when the interrupt happened
2969	 * except for r13, which is saved in SPRG_SCRATCH0.
2970	 */
2971	mfspr	r13, SPRN_HSRR0
2972	addi	r13, r13, 4
2973	mtspr	SPRN_HSRR0, r13
2974	GET_SCRATCH0(r13)
2975	HRFI_TO_KERNEL
2976	b	.
2977#endif
2978
2979	/*
2980	 * Relocation-on interrupts: A subset of the interrupts can be delivered
2981	 * with IR=1/DR=1, if AIL==2 and MSR.HV won't be changed by delivering
2982	 * it.  Addresses are the same as the original interrupt addresses, but
2983	 * offset by 0xc000000000004000.
2984	 * It's impossible to receive interrupts below 0x300 via this mechanism.
2985	 * KVM: None of these traps are from the guest ; anything that escalated
2986	 * to HV=1 from HV=0 is delivered via real mode handlers.
2987	 */
2988
2989	/*
2990	 * This uses the standard macro, since the original 0x300 vector
2991	 * only has extra guff for STAB-based processors -- which never
2992	 * come here.
2993	 */
2994
2995EXC_COMMON_BEGIN(ppc64_runlatch_on_trampoline)
2996	b	__ppc64_runlatch_on
2997
2998USE_FIXED_SECTION(virt_trampolines)
2999	/*
3000	 * The __end_interrupts marker must be past the out-of-line (OOL)
3001	 * handlers, so that they are copied to real address 0x100 when running
3002	 * a relocatable kernel. This ensures they can be reached from the short
3003	 * trampoline handlers (like 0x4f00, 0x4f20, etc.) which branch
3004	 * directly, without using LOAD_HANDLER().
3005	 */
3006	.align	7
3007	.globl	__end_interrupts
3008__end_interrupts:
3009DEFINE_FIXED_SYMBOL(__end_interrupts)
3010
3011#ifdef CONFIG_PPC_970_NAP
3012	/*
3013	 * Called by exception entry code if _TLF_NAPPING was set, this clears
3014	 * the NAPPING flag, and redirects the exception exit to
3015	 * power4_fixup_nap_return.
3016	 */
3017	.globl power4_fixup_nap
3018EXC_COMMON_BEGIN(power4_fixup_nap)
3019	andc	r9,r9,r10
3020	std	r9,TI_LOCAL_FLAGS(r11)
3021	LOAD_REG_ADDR(r10, power4_idle_nap_return)
3022	std	r10,_NIP(r1)
3023	blr
3024
3025power4_idle_nap_return:
3026	blr
3027#endif
3028
3029CLOSE_FIXED_SECTION(real_vectors);
3030CLOSE_FIXED_SECTION(real_trampolines);
3031CLOSE_FIXED_SECTION(virt_vectors);
3032CLOSE_FIXED_SECTION(virt_trampolines);
3033
3034USE_TEXT_SECTION()
3035
3036/* MSR[RI] should be clear because this uses SRR[01] */
3037enable_machine_check:
3038	mflr	r0
3039	bcl	20,31,$+4
30400:	mflr	r3
3041	addi	r3,r3,(1f - 0b)
3042	mtspr	SPRN_SRR0,r3
3043	mfmsr	r3
3044	ori	r3,r3,MSR_ME
3045	mtspr	SPRN_SRR1,r3
3046	RFI_TO_KERNEL
30471:	mtlr	r0
3048	blr
3049
3050/* MSR[RI] should be clear because this uses SRR[01] */
3051disable_machine_check:
3052	mflr	r0
3053	bcl	20,31,$+4
30540:	mflr	r3
3055	addi	r3,r3,(1f - 0b)
3056	mtspr	SPRN_SRR0,r3
3057	mfmsr	r3
3058	li	r4,MSR_ME
3059	andc	r3,r3,r4
3060	mtspr	SPRN_SRR1,r3
3061	RFI_TO_KERNEL
30621:	mtlr	r0
3063	blr
3064
3065/*
3066 * Hash table stuff
3067 */
3068	.balign	IFETCH_ALIGN_BYTES
3069do_hash_page:
3070#ifdef CONFIG_PPC_BOOK3S_64
3071	lis	r0,(DSISR_BAD_FAULT_64S | DSISR_DABRMATCH | DSISR_KEYFAULT)@h
3072	ori	r0,r0,DSISR_BAD_FAULT_64S@l
3073	and.	r0,r5,r0		/* weird error? */
3074	bne-	handle_page_fault	/* if not, try to insert a HPTE */
3075	ld	r11, PACA_THREAD_INFO(r13)
3076	lwz	r0,TI_PREEMPT(r11)	/* If we're in an "NMI" */
3077	andis.	r0,r0,NMI_MASK@h	/* (i.e. an irq when soft-disabled) */
3078	bne	77f			/* then don't call hash_page now */
3079
3080	/*
3081	 * r3 contains the trap number
3082	 * r4 contains the faulting address
3083	 * r5 contains dsisr
3084	 * r6 msr
3085	 *
3086	 * at return r3 = 0 for success, 1 for page fault, negative for error
3087	 */
3088	bl	__hash_page		/* build HPTE if possible */
3089        cmpdi	r3,0			/* see if __hash_page succeeded */
3090
3091	/* Success */
3092	beq	interrupt_return	/* Return from exception on success */
3093
3094	/* Error */
3095	blt-	13f
3096
3097	/* Reload DAR/DSISR into r4/r5 for the DABR check below */
3098	ld	r4,_DAR(r1)
3099	ld      r5,_DSISR(r1)
3100#endif /* CONFIG_PPC_BOOK3S_64 */
3101
3102/* Here we have a page fault that hash_page can't handle. */
3103handle_page_fault:
310411:	andis.  r0,r5,DSISR_DABRMATCH@h
3105	bne-    handle_dabr_fault
3106	addi	r3,r1,STACK_FRAME_OVERHEAD
3107	bl	do_page_fault
3108	cmpdi	r3,0
3109	beq+	interrupt_return
3110	mr	r5,r3
3111	addi	r3,r1,STACK_FRAME_OVERHEAD
3112	ld	r4,_DAR(r1)
3113	bl	bad_page_fault
3114	b	interrupt_return
3115
3116/* We have a data breakpoint exception - handle it */
3117handle_dabr_fault:
3118	ld      r4,_DAR(r1)
3119	ld      r5,_DSISR(r1)
3120	addi    r3,r1,STACK_FRAME_OVERHEAD
3121	bl      do_break
3122	/*
3123	 * do_break() may have changed the NV GPRS while handling a breakpoint.
3124	 * If so, we need to restore them with their updated values.
3125	 */
3126	REST_NVGPRS(r1)
3127	b       interrupt_return
3128
3129
3130#ifdef CONFIG_PPC_BOOK3S_64
3131/* We have a page fault that hash_page could handle but HV refused
3132 * the PTE insertion
3133 */
313413:	mr	r5,r3
3135	addi	r3,r1,STACK_FRAME_OVERHEAD
3136	ld	r4,_DAR(r1)
3137	bl	low_hash_fault
3138	b	interrupt_return
3139#endif
3140
3141/*
3142 * We come here as a result of a DSI at a point where we don't want
3143 * to call hash_page, such as when we are accessing memory (possibly
3144 * user memory) inside a PMU interrupt that occurred while interrupts
3145 * were soft-disabled.  We want to invoke the exception handler for
3146 * the access, or panic if there isn't a handler.
3147 */
314877:	addi	r3,r1,STACK_FRAME_OVERHEAD
3149	li	r5,SIGSEGV
3150	bl	bad_page_fault
3151	b	interrupt_return
3152