xref: /linux/arch/powerpc/kernel/exceptions-64s.S (revision 31284f703db2f1605b2dbc6bb0632b04d7be13e7)
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/*
25 * Following are fixed section helper macros.
26 *
27 * EXC_REAL_BEGIN/END  - real, unrelocated exception vectors
28 * EXC_VIRT_BEGIN/END  - virt (AIL), unrelocated exception vectors
29 * TRAMP_REAL_BEGIN    - real, unrelocated helpers (virt may call these)
30 * TRAMP_VIRT_BEGIN    - virt, unreloc helpers (in practice, real can use)
31 * EXC_COMMON          - After switching to virtual, relocated mode.
32 */
33
34#define EXC_REAL_BEGIN(name, start, size)			\
35	FIXED_SECTION_ENTRY_BEGIN_LOCATION(real_vectors, exc_real_##start##_##name, start, size)
36
37#define EXC_REAL_END(name, start, size)				\
38	FIXED_SECTION_ENTRY_END_LOCATION(real_vectors, exc_real_##start##_##name, start, size)
39
40#define EXC_VIRT_BEGIN(name, start, size)			\
41	FIXED_SECTION_ENTRY_BEGIN_LOCATION(virt_vectors, exc_virt_##start##_##name, start, size)
42
43#define EXC_VIRT_END(name, start, size)				\
44	FIXED_SECTION_ENTRY_END_LOCATION(virt_vectors, exc_virt_##start##_##name, start, size)
45
46#define EXC_COMMON_BEGIN(name)					\
47	USE_TEXT_SECTION();					\
48	.balign IFETCH_ALIGN_BYTES;				\
49	.global name;						\
50	_ASM_NOKPROBE_SYMBOL(name);				\
51	DEFINE_FIXED_SYMBOL(name);				\
52name:
53
54#define TRAMP_REAL_BEGIN(name)					\
55	FIXED_SECTION_ENTRY_BEGIN(real_trampolines, name)
56
57#define TRAMP_VIRT_BEGIN(name)					\
58	FIXED_SECTION_ENTRY_BEGIN(virt_trampolines, name)
59
60#define EXC_REAL_NONE(start, size)				\
61	FIXED_SECTION_ENTRY_BEGIN_LOCATION(real_vectors, exc_real_##start##_##unused, start, size); \
62	FIXED_SECTION_ENTRY_END_LOCATION(real_vectors, exc_real_##start##_##unused, start, size)
63
64#define EXC_VIRT_NONE(start, size)				\
65	FIXED_SECTION_ENTRY_BEGIN_LOCATION(virt_vectors, exc_virt_##start##_##unused, start, size); \
66	FIXED_SECTION_ENTRY_END_LOCATION(virt_vectors, exc_virt_##start##_##unused, start, size)
67
68/*
69 * We're short on space and time in the exception prolog, so we can't
70 * use the normal LOAD_REG_IMMEDIATE macro to load the address of label.
71 * Instead we get the base of the kernel from paca->kernelbase and or in the low
72 * part of label. This requires that the label be within 64KB of kernelbase, and
73 * that kernelbase be 64K aligned.
74 */
75#define LOAD_HANDLER(reg, label)					\
76	ld	reg,PACAKBASE(r13);	/* get high part of &label */	\
77	ori	reg,reg,FIXED_SYMBOL_ABS_ADDR(label)
78
79#define __LOAD_HANDLER(reg, label)					\
80	ld	reg,PACAKBASE(r13);					\
81	ori	reg,reg,(ABS_ADDR(label))@l
82
83/*
84 * Branches from unrelocated code (e.g., interrupts) to labels outside
85 * head-y require >64K offsets.
86 */
87#define __LOAD_FAR_HANDLER(reg, label)					\
88	ld	reg,PACAKBASE(r13);					\
89	ori	reg,reg,(ABS_ADDR(label))@l;				\
90	addis	reg,reg,(ABS_ADDR(label))@h
91
92/*
93 * Branch to label using its 0xC000 address. This results in instruction
94 * address suitable for MSR[IR]=0 or 1, which allows relocation to be turned
95 * on using mtmsr rather than rfid.
96 *
97 * This could set the 0xc bits for !RELOCATABLE as an immediate, rather than
98 * load KBASE for a slight optimisation.
99 */
100#define BRANCH_TO_C000(reg, label)					\
101	__LOAD_FAR_HANDLER(reg, label);					\
102	mtctr	reg;							\
103	bctr
104
105/*
106 * Interrupt code generation macros
107 */
108#define IVEC		.L_IVEC_\name\()	/* Interrupt vector address */
109#define IHSRR		.L_IHSRR_\name\()	/* Sets SRR or HSRR registers */
110#define IHSRR_IF_HVMODE	.L_IHSRR_IF_HVMODE_\name\() /* HSRR if HV else SRR */
111#define IAREA		.L_IAREA_\name\()	/* PACA save area */
112#define IVIRT		.L_IVIRT_\name\()	/* Has virt mode entry point */
113#define IISIDE		.L_IISIDE_\name\()	/* Uses SRR0/1 not DAR/DSISR */
114#define IDAR		.L_IDAR_\name\()	/* Uses DAR (or SRR0) */
115#define IDSISR		.L_IDSISR_\name\()	/* Uses DSISR (or SRR1) */
116#define ISET_RI		.L_ISET_RI_\name\()	/* Run common code w/ MSR[RI]=1 */
117#define IBRANCH_TO_COMMON	.L_IBRANCH_TO_COMMON_\name\() /* ENTRY branch to common */
118#define IREALMODE_COMMON	.L_IREALMODE_COMMON_\name\() /* Common runs in realmode */
119#define IMASK		.L_IMASK_\name\()	/* IRQ soft-mask bit */
120#define IKVM_REAL	.L_IKVM_REAL_\name\()	/* Real entry tests KVM */
121#define __IKVM_REAL(name)	.L_IKVM_REAL_ ## name
122#define IKVM_VIRT	.L_IKVM_VIRT_\name\()	/* Virt entry tests KVM */
123#define ISTACK		.L_ISTACK_\name\()	/* Set regular kernel stack */
124#define __ISTACK(name)	.L_ISTACK_ ## name
125#define IKUAP		.L_IKUAP_\name\()	/* Do KUAP lock */
126
127#define INT_DEFINE_BEGIN(n)						\
128.macro int_define_ ## n name
129
130#define INT_DEFINE_END(n)						\
131.endm ;									\
132int_define_ ## n n ;							\
133do_define_int n
134
135.macro do_define_int name
136	.ifndef IVEC
137		.error "IVEC not defined"
138	.endif
139	.ifndef IHSRR
140		IHSRR=0
141	.endif
142	.ifndef IHSRR_IF_HVMODE
143		IHSRR_IF_HVMODE=0
144	.endif
145	.ifndef IAREA
146		IAREA=PACA_EXGEN
147	.endif
148	.ifndef IVIRT
149		IVIRT=1
150	.endif
151	.ifndef IISIDE
152		IISIDE=0
153	.endif
154	.ifndef IDAR
155		IDAR=0
156	.endif
157	.ifndef IDSISR
158		IDSISR=0
159	.endif
160	.ifndef ISET_RI
161		ISET_RI=1
162	.endif
163	.ifndef IBRANCH_TO_COMMON
164		IBRANCH_TO_COMMON=1
165	.endif
166	.ifndef IREALMODE_COMMON
167		IREALMODE_COMMON=0
168	.else
169		.if ! IBRANCH_TO_COMMON
170			.error "IREALMODE_COMMON=1 but IBRANCH_TO_COMMON=0"
171		.endif
172	.endif
173	.ifndef IMASK
174		IMASK=0
175	.endif
176	.ifndef IKVM_REAL
177		IKVM_REAL=0
178	.endif
179	.ifndef IKVM_VIRT
180		IKVM_VIRT=0
181	.endif
182	.ifndef ISTACK
183		ISTACK=1
184	.endif
185	.ifndef IKUAP
186		IKUAP=1
187	.endif
188.endm
189
190/*
191 * All interrupts which set HSRR registers, as well as SRESET and MCE and
192 * syscall when invoked with "sc 1" switch to MSR[HV]=1 (HVMODE) to be taken,
193 * so they all generally need to test whether they were taken in guest context.
194 *
195 * Note: SRESET and MCE may also be sent to the guest by the hypervisor, and be
196 * taken with MSR[HV]=0.
197 *
198 * Interrupts which set SRR registers (with the above exceptions) do not
199 * elevate to MSR[HV]=1 mode, though most can be taken when running with
200 * MSR[HV]=1  (e.g., bare metal kernel and userspace). So these interrupts do
201 * not need to test whether a guest is running because they get delivered to
202 * the guest directly, including nested HV KVM guests.
203 *
204 * The exception is PR KVM, where the guest runs with MSR[PR]=1 and the host
205 * runs with MSR[HV]=0, so the host takes all interrupts on behalf of the
206 * guest. PR KVM runs with LPCR[AIL]=0 which causes interrupts to always be
207 * delivered to the real-mode entry point, therefore such interrupts only test
208 * KVM in their real mode handlers, and only when PR KVM is possible.
209 *
210 * Interrupts that are taken in MSR[HV]=0 and escalate to MSR[HV]=1 are always
211 * delivered in real-mode when the MMU is in hash mode because the MMU
212 * registers are not set appropriately to translate host addresses. In nested
213 * radix mode these can be delivered in virt-mode as the host translations are
214 * used implicitly (see: effective LPID, effective PID).
215 */
216
217/*
218 * If an interrupt is taken while a guest is running, it is immediately routed
219 * to KVM to handle.
220 */
221
222.macro KVMTEST name handler
223#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
224	lbz	r10,HSTATE_IN_GUEST(r13)
225	cmpwi	r10,0
226	/* HSRR variants have the 0x2 bit added to their trap number */
227	.if IHSRR_IF_HVMODE
228	BEGIN_FTR_SECTION
229	li	r10,(IVEC + 0x2)
230	FTR_SECTION_ELSE
231	li	r10,(IVEC)
232	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
233	.elseif IHSRR
234	li	r10,(IVEC + 0x2)
235	.else
236	li	r10,(IVEC)
237	.endif
238	bne	\handler
239#endif
240.endm
241
242/*
243 * This is the BOOK3S interrupt entry code macro.
244 *
245 * This can result in one of several things happening:
246 * - Branch to the _common handler, relocated, in virtual mode.
247 *   These are normal interrupts (synchronous and asynchronous) handled by
248 *   the kernel.
249 * - Branch to KVM, relocated but real mode interrupts remain in real mode.
250 *   These occur when HSTATE_IN_GUEST is set. The interrupt may be caused by
251 *   / intended for host or guest kernel, but KVM must always be involved
252 *   because the machine state is set for guest execution.
253 * - Branch to the masked handler, unrelocated.
254 *   These occur when maskable asynchronous interrupts are taken with the
255 *   irq_soft_mask set.
256 * - Branch to an "early" handler in real mode but relocated.
257 *   This is done if early=1. MCE and HMI use these to handle errors in real
258 *   mode.
259 * - Fall through and continue executing in real, unrelocated mode.
260 *   This is done if early=2.
261 */
262
263.macro GEN_BRANCH_TO_COMMON name, virt
264	.if IREALMODE_COMMON
265	LOAD_HANDLER(r10, \name\()_common)
266	mtctr	r10
267	bctr
268	.else
269	.if \virt
270#ifndef CONFIG_RELOCATABLE
271	b	\name\()_common_virt
272#else
273	LOAD_HANDLER(r10, \name\()_common_virt)
274	mtctr	r10
275	bctr
276#endif
277	.else
278	LOAD_HANDLER(r10, \name\()_common_real)
279	mtctr	r10
280	bctr
281	.endif
282	.endif
283.endm
284
285.macro GEN_INT_ENTRY name, virt, ool=0
286	SET_SCRATCH0(r13)			/* save r13 */
287	GET_PACA(r13)
288	std	r9,IAREA+EX_R9(r13)		/* save r9 */
289BEGIN_FTR_SECTION
290	mfspr	r9,SPRN_PPR
291END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
292	HMT_MEDIUM
293	std	r10,IAREA+EX_R10(r13)		/* save r10 - r12 */
294BEGIN_FTR_SECTION
295	mfspr	r10,SPRN_CFAR
296END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
297	.if \ool
298	.if !\virt
299	b	tramp_real_\name
300	.pushsection .text
301	TRAMP_REAL_BEGIN(tramp_real_\name)
302	.else
303	b	tramp_virt_\name
304	.pushsection .text
305	TRAMP_VIRT_BEGIN(tramp_virt_\name)
306	.endif
307	.endif
308
309BEGIN_FTR_SECTION
310	std	r9,IAREA+EX_PPR(r13)
311END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
312BEGIN_FTR_SECTION
313	std	r10,IAREA+EX_CFAR(r13)
314END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
315	INTERRUPT_TO_KERNEL
316	mfctr	r10
317	std	r10,IAREA+EX_CTR(r13)
318	mfcr	r9
319	std	r11,IAREA+EX_R11(r13)
320	std	r12,IAREA+EX_R12(r13)
321
322	/*
323	 * DAR/DSISR, SCRATCH0 must be read before setting MSR[RI],
324	 * because a d-side MCE will clobber those registers so is
325	 * not recoverable if they are live.
326	 */
327	GET_SCRATCH0(r10)
328	std	r10,IAREA+EX_R13(r13)
329	.if IDAR && !IISIDE
330	.if IHSRR
331	mfspr	r10,SPRN_HDAR
332	.else
333	mfspr	r10,SPRN_DAR
334	.endif
335	std	r10,IAREA+EX_DAR(r13)
336	.endif
337	.if IDSISR && !IISIDE
338	.if IHSRR
339	mfspr	r10,SPRN_HDSISR
340	.else
341	mfspr	r10,SPRN_DSISR
342	.endif
343	stw	r10,IAREA+EX_DSISR(r13)
344	.endif
345
346	.if IHSRR_IF_HVMODE
347	BEGIN_FTR_SECTION
348	mfspr	r11,SPRN_HSRR0		/* save HSRR0 */
349	mfspr	r12,SPRN_HSRR1		/* and HSRR1 */
350	FTR_SECTION_ELSE
351	mfspr	r11,SPRN_SRR0		/* save SRR0 */
352	mfspr	r12,SPRN_SRR1		/* and SRR1 */
353	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
354	.elseif IHSRR
355	mfspr	r11,SPRN_HSRR0		/* save HSRR0 */
356	mfspr	r12,SPRN_HSRR1		/* and HSRR1 */
357	.else
358	mfspr	r11,SPRN_SRR0		/* save SRR0 */
359	mfspr	r12,SPRN_SRR1		/* and SRR1 */
360	.endif
361
362	.if IBRANCH_TO_COMMON
363	GEN_BRANCH_TO_COMMON \name \virt
364	.endif
365
366	.if \ool
367	.popsection
368	.endif
369.endm
370
371/*
372 * __GEN_COMMON_ENTRY is required to receive the branch from interrupt
373 * entry, except in the case of the real-mode handlers which require
374 * __GEN_REALMODE_COMMON_ENTRY.
375 *
376 * This switches to virtual mode and sets MSR[RI].
377 */
378.macro __GEN_COMMON_ENTRY name
379DEFINE_FIXED_SYMBOL(\name\()_common_real)
380\name\()_common_real:
381	.if IKVM_REAL
382		KVMTEST \name kvm_interrupt
383	.endif
384
385	ld	r10,PACAKMSR(r13)	/* get MSR value for kernel */
386	/* MSR[RI] is clear iff using SRR regs */
387	.if IHSRR_IF_HVMODE
388	BEGIN_FTR_SECTION
389	xori	r10,r10,MSR_RI
390	END_FTR_SECTION_IFCLR(CPU_FTR_HVMODE)
391	.elseif ! IHSRR
392	xori	r10,r10,MSR_RI
393	.endif
394	mtmsrd	r10
395
396	.if IVIRT
397	.if IKVM_VIRT
398	b	1f /* skip the virt test coming from real */
399	.endif
400
401	.balign IFETCH_ALIGN_BYTES
402DEFINE_FIXED_SYMBOL(\name\()_common_virt)
403\name\()_common_virt:
404	.if IKVM_VIRT
405		KVMTEST \name kvm_interrupt
4061:
407	.endif
408	.endif /* IVIRT */
409.endm
410
411/*
412 * Don't switch to virt mode. Used for early MCE and HMI handlers that
413 * want to run in real mode.
414 */
415.macro __GEN_REALMODE_COMMON_ENTRY name
416DEFINE_FIXED_SYMBOL(\name\()_common_real)
417\name\()_common_real:
418	.if IKVM_REAL
419		KVMTEST \name kvm_interrupt
420	.endif
421.endm
422
423.macro __GEN_COMMON_BODY name
424	.if IMASK
425		.if ! ISTACK
426		.error "No support for masked interrupt to use custom stack"
427		.endif
428
429		/* If coming from user, skip soft-mask tests. */
430		andi.	r10,r12,MSR_PR
431		bne	3f
432
433		/*
434		 * Kernel code running below __end_soft_masked may be
435		 * implicitly soft-masked if it is within the regions
436		 * in the soft mask table.
437		 */
438		LOAD_HANDLER(r10, __end_soft_masked)
439		cmpld	r11,r10
440		bge+	1f
441
442		/* SEARCH_SOFT_MASK_TABLE clobbers r9,r10,r12 */
443		mtctr	r12
444		stw	r9,PACA_EXGEN+EX_CCR(r13)
445		SEARCH_SOFT_MASK_TABLE
446		cmpdi	r12,0
447		mfctr	r12		/* Restore r12 to SRR1 */
448		lwz	r9,PACA_EXGEN+EX_CCR(r13)
449		beq	1f		/* Not in soft-mask table */
450		li	r10,IMASK
451		b	2f		/* In soft-mask table, always mask */
452
453		/* Test the soft mask state against our interrupt's bit */
4541:		lbz	r10,PACAIRQSOFTMASK(r13)
4552:		andi.	r10,r10,IMASK
456		/* Associate vector numbers with bits in paca->irq_happened */
457		.if IVEC == 0x500 || IVEC == 0xea0
458		li	r10,PACA_IRQ_EE
459		.elseif IVEC == 0x900
460		li	r10,PACA_IRQ_DEC
461		.elseif IVEC == 0xa00 || IVEC == 0xe80
462		li	r10,PACA_IRQ_DBELL
463		.elseif IVEC == 0xe60
464		li	r10,PACA_IRQ_HMI
465		.elseif IVEC == 0xf00
466		li	r10,PACA_IRQ_PMI
467		.else
468		.abort "Bad maskable vector"
469		.endif
470
471		.if IHSRR_IF_HVMODE
472		BEGIN_FTR_SECTION
473		bne	masked_Hinterrupt
474		FTR_SECTION_ELSE
475		bne	masked_interrupt
476		ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
477		.elseif IHSRR
478		bne	masked_Hinterrupt
479		.else
480		bne	masked_interrupt
481		.endif
482	.endif
483
484	.if ISTACK
485	andi.	r10,r12,MSR_PR		/* See if coming from user	*/
4863:	mr	r10,r1			/* Save r1			*/
487	subi	r1,r1,INT_FRAME_SIZE	/* alloc frame on kernel stack	*/
488	beq-	100f
489	ld	r1,PACAKSAVE(r13)	/* kernel stack to use		*/
490100:	tdgei	r1,-INT_FRAME_SIZE	/* trap if r1 is in userspace	*/
491	EMIT_BUG_ENTRY 100b,__FILE__,__LINE__,0
492	.endif
493
494	std	r9,_CCR(r1)		/* save CR in stackframe	*/
495	std	r11,_NIP(r1)		/* save SRR0 in stackframe	*/
496	std	r12,_MSR(r1)		/* save SRR1 in stackframe	*/
497	std	r10,0(r1)		/* make stack chain pointer	*/
498	std	r0,GPR0(r1)		/* save r0 in stackframe	*/
499	std	r10,GPR1(r1)		/* save r1 in stackframe	*/
500
501	/* Mark our [H]SRRs valid for return */
502	li	r10,1
503	.if IHSRR_IF_HVMODE
504	BEGIN_FTR_SECTION
505	stb	r10,PACAHSRR_VALID(r13)
506	FTR_SECTION_ELSE
507	stb	r10,PACASRR_VALID(r13)
508	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
509	.elseif IHSRR
510	stb	r10,PACAHSRR_VALID(r13)
511	.else
512	stb	r10,PACASRR_VALID(r13)
513	.endif
514
515	.if ISET_RI
516	li	r10,MSR_RI
517	mtmsrd	r10,1			/* Set MSR_RI */
518	.endif
519
520	.if ISTACK
521	.if IKUAP
522	kuap_save_amr_and_lock r9, r10, cr1, cr0
523	.endif
524	beq	101f			/* if from kernel mode		*/
525BEGIN_FTR_SECTION
526	ld	r9,IAREA+EX_PPR(r13)	/* Read PPR from paca		*/
527	std	r9,_PPR(r1)
528END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
529101:
530	.else
531	.if IKUAP
532	kuap_save_amr_and_lock r9, r10, cr1
533	.endif
534	.endif
535
536	/* Save original regs values from save area to stack frame. */
537	ld	r9,IAREA+EX_R9(r13)	/* move r9, r10 to stackframe	*/
538	ld	r10,IAREA+EX_R10(r13)
539	std	r9,GPR9(r1)
540	std	r10,GPR10(r1)
541	ld	r9,IAREA+EX_R11(r13)	/* move r11 - r13 to stackframe	*/
542	ld	r10,IAREA+EX_R12(r13)
543	ld	r11,IAREA+EX_R13(r13)
544	std	r9,GPR11(r1)
545	std	r10,GPR12(r1)
546	std	r11,GPR13(r1)
547
548	SAVE_NVGPRS(r1)
549
550	.if IDAR
551	.if IISIDE
552	ld	r10,_NIP(r1)
553	.else
554	ld	r10,IAREA+EX_DAR(r13)
555	.endif
556	std	r10,_DAR(r1)
557	.endif
558
559	.if IDSISR
560	.if IISIDE
561	ld	r10,_MSR(r1)
562	lis	r11,DSISR_SRR1_MATCH_64S@h
563	and	r10,r10,r11
564	.else
565	lwz	r10,IAREA+EX_DSISR(r13)
566	.endif
567	std	r10,_DSISR(r1)
568	.endif
569
570BEGIN_FTR_SECTION
571	ld	r10,IAREA+EX_CFAR(r13)
572	std	r10,ORIG_GPR3(r1)
573END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
574	ld	r10,IAREA+EX_CTR(r13)
575	std	r10,_CTR(r1)
576	std	r2,GPR2(r1)		/* save r2 in stackframe	*/
577	SAVE_GPRS(3, 8, r1)		/* save r3 - r8 in stackframe   */
578	mflr	r9			/* Get LR, later save to stack	*/
579	ld	r2,PACATOC(r13)		/* get kernel TOC into r2	*/
580	std	r9,_LINK(r1)
581	lbz	r10,PACAIRQSOFTMASK(r13)
582	mfspr	r11,SPRN_XER		/* save XER in stackframe	*/
583	std	r10,SOFTE(r1)
584	std	r11,_XER(r1)
585	li	r9,IVEC
586	std	r9,_TRAP(r1)		/* set trap number		*/
587	li	r10,0
588	ld	r11,exception_marker@toc(r2)
589	std	r10,RESULT(r1)		/* clear regs->result		*/
590	std	r11,STACK_FRAME_OVERHEAD-16(r1) /* mark the frame	*/
591.endm
592
593/*
594 * On entry r13 points to the paca, r9-r13 are saved in the paca,
595 * r9 contains the saved CR, r11 and r12 contain the saved SRR0 and
596 * SRR1, and relocation is on.
597 *
598 * If stack=0, then the stack is already set in r1, and r1 is saved in r10.
599 * PPR save and CPU accounting is not done for the !stack case (XXX why not?)
600 */
601.macro GEN_COMMON name
602	__GEN_COMMON_ENTRY \name
603	__GEN_COMMON_BODY \name
604.endm
605
606.macro SEARCH_RESTART_TABLE
607#ifdef CONFIG_RELOCATABLE
608	mr	r12,r2
609	ld	r2,PACATOC(r13)
610	LOAD_REG_ADDR(r9, __start___restart_table)
611	LOAD_REG_ADDR(r10, __stop___restart_table)
612	mr	r2,r12
613#else
614	LOAD_REG_IMMEDIATE_SYM(r9, r12, __start___restart_table)
615	LOAD_REG_IMMEDIATE_SYM(r10, r12, __stop___restart_table)
616#endif
617300:
618	cmpd	r9,r10
619	beq	302f
620	ld	r12,0(r9)
621	cmpld	r11,r12
622	blt	301f
623	ld	r12,8(r9)
624	cmpld	r11,r12
625	bge	301f
626	ld	r12,16(r9)
627	b	303f
628301:
629	addi	r9,r9,24
630	b	300b
631302:
632	li	r12,0
633303:
634.endm
635
636.macro SEARCH_SOFT_MASK_TABLE
637#ifdef CONFIG_RELOCATABLE
638	mr	r12,r2
639	ld	r2,PACATOC(r13)
640	LOAD_REG_ADDR(r9, __start___soft_mask_table)
641	LOAD_REG_ADDR(r10, __stop___soft_mask_table)
642	mr	r2,r12
643#else
644	LOAD_REG_IMMEDIATE_SYM(r9, r12, __start___soft_mask_table)
645	LOAD_REG_IMMEDIATE_SYM(r10, r12, __stop___soft_mask_table)
646#endif
647300:
648	cmpd	r9,r10
649	beq	302f
650	ld	r12,0(r9)
651	cmpld	r11,r12
652	blt	301f
653	ld	r12,8(r9)
654	cmpld	r11,r12
655	bge	301f
656	li	r12,1
657	b	303f
658301:
659	addi	r9,r9,16
660	b	300b
661302:
662	li	r12,0
663303:
664.endm
665
666/*
667 * Restore all registers including H/SRR0/1 saved in a stack frame of a
668 * standard exception.
669 */
670.macro EXCEPTION_RESTORE_REGS hsrr=0
671	/* Move original SRR0 and SRR1 into the respective regs */
672	ld	r9,_MSR(r1)
673	li	r10,0
674	.if \hsrr
675	mtspr	SPRN_HSRR1,r9
676	stb	r10,PACAHSRR_VALID(r13)
677	.else
678	mtspr	SPRN_SRR1,r9
679	stb	r10,PACASRR_VALID(r13)
680	.endif
681	ld	r9,_NIP(r1)
682	.if \hsrr
683	mtspr	SPRN_HSRR0,r9
684	.else
685	mtspr	SPRN_SRR0,r9
686	.endif
687	ld	r9,_CTR(r1)
688	mtctr	r9
689	ld	r9,_XER(r1)
690	mtxer	r9
691	ld	r9,_LINK(r1)
692	mtlr	r9
693	ld	r9,_CCR(r1)
694	mtcr	r9
695	REST_GPRS(2, 13, r1)
696	REST_GPR(0, r1)
697	/* restore original r1. */
698	ld	r1,GPR1(r1)
699.endm
700
701/*
702 * There are a few constraints to be concerned with.
703 * - Real mode exceptions code/data must be located at their physical location.
704 * - Virtual mode exceptions must be mapped at their 0xc000... location.
705 * - Fixed location code must not call directly beyond the __end_interrupts
706 *   area when built with CONFIG_RELOCATABLE. LOAD_HANDLER / bctr sequence
707 *   must be used.
708 * - LOAD_HANDLER targets must be within first 64K of physical 0 /
709 *   virtual 0xc00...
710 * - Conditional branch targets must be within +/-32K of caller.
711 *
712 * "Virtual exceptions" run with relocation on (MSR_IR=1, MSR_DR=1), and
713 * therefore don't have to run in physically located code or rfid to
714 * virtual mode kernel code. However on relocatable kernels they do have
715 * to branch to KERNELBASE offset because the rest of the kernel (outside
716 * the exception vectors) may be located elsewhere.
717 *
718 * Virtual exceptions correspond with physical, except their entry points
719 * are offset by 0xc000000000000000 and also tend to get an added 0x4000
720 * offset applied. Virtual exceptions are enabled with the Alternate
721 * Interrupt Location (AIL) bit set in the LPCR. However this does not
722 * guarantee they will be delivered virtually. Some conditions (see the ISA)
723 * cause exceptions to be delivered in real mode.
724 *
725 * The scv instructions are a special case. They get a 0x3000 offset applied.
726 * scv exceptions have unique reentrancy properties, see below.
727 *
728 * It's impossible to receive interrupts below 0x300 via AIL.
729 *
730 * KVM: None of the virtual exceptions are from the guest. Anything that
731 * escalated to HV=1 from HV=0 is delivered via real mode handlers.
732 *
733 *
734 * We layout physical memory as follows:
735 * 0x0000 - 0x00ff : Secondary processor spin code
736 * 0x0100 - 0x18ff : Real mode pSeries interrupt vectors
737 * 0x1900 - 0x2fff : Real mode trampolines
738 * 0x3000 - 0x58ff : Relon (IR=1,DR=1) mode pSeries interrupt vectors
739 * 0x5900 - 0x6fff : Relon mode trampolines
740 * 0x7000 - 0x7fff : FWNMI data area
741 * 0x8000 -   .... : Common interrupt handlers, remaining early
742 *                   setup code, rest of kernel.
743 *
744 * We could reclaim 0x4000-0x42ff for real mode trampolines if the space
745 * is necessary. Until then it's more consistent to explicitly put VIRT_NONE
746 * vectors there.
747 */
748OPEN_FIXED_SECTION(real_vectors,        0x0100, 0x1900)
749OPEN_FIXED_SECTION(real_trampolines,    0x1900, 0x3000)
750OPEN_FIXED_SECTION(virt_vectors,        0x3000, 0x5900)
751OPEN_FIXED_SECTION(virt_trampolines,    0x5900, 0x7000)
752
753#ifdef CONFIG_PPC_POWERNV
754	.globl start_real_trampolines
755	.globl end_real_trampolines
756	.globl start_virt_trampolines
757	.globl end_virt_trampolines
758#endif
759
760#if defined(CONFIG_PPC_PSERIES) || defined(CONFIG_PPC_POWERNV)
761/*
762 * Data area reserved for FWNMI option.
763 * This address (0x7000) is fixed by the RPA.
764 * pseries and powernv need to keep the whole page from
765 * 0x7000 to 0x8000 free for use by the firmware
766 */
767ZERO_FIXED_SECTION(fwnmi_page,          0x7000, 0x8000)
768OPEN_TEXT_SECTION(0x8000)
769#else
770OPEN_TEXT_SECTION(0x7000)
771#endif
772
773USE_FIXED_SECTION(real_vectors)
774
775/*
776 * This is the start of the interrupt handlers for pSeries
777 * This code runs with relocation off.
778 * Code from here to __end_interrupts gets copied down to real
779 * address 0x100 when we are running a relocatable kernel.
780 * Therefore any relative branches in this section must only
781 * branch to labels in this section.
782 */
783	.globl __start_interrupts
784__start_interrupts:
785
786/**
787 * Interrupt 0x3000 - System Call Vectored Interrupt (syscall).
788 * This is a synchronous interrupt invoked with the "scv" instruction. The
789 * system call does not alter the HV bit, so it is directed to the OS.
790 *
791 * Handling:
792 * scv instructions enter the kernel without changing EE, RI, ME, or HV.
793 * In particular, this means we can take a maskable interrupt at any point
794 * in the scv handler, which is unlike any other interrupt. This is solved
795 * by treating the instruction addresses in the handler as being soft-masked,
796 * by adding a SOFT_MASK_TABLE entry for them.
797 *
798 * AIL-0 mode scv exceptions go to 0x17000-0x17fff, but we set AIL-3 and
799 * ensure scv is never executed with relocation off, which means AIL-0
800 * should never happen.
801 *
802 * Before leaving the following inside-__end_soft_masked text, at least of the
803 * following must be true:
804 * - MSR[PR]=1 (i.e., return to userspace)
805 * - MSR_EE|MSR_RI is clear (no reentrant exceptions)
806 * - Standard kernel environment is set up (stack, paca, etc)
807 *
808 * Call convention:
809 *
810 * syscall register convention is in Documentation/powerpc/syscall64-abi.rst
811 */
812EXC_VIRT_BEGIN(system_call_vectored, 0x3000, 0x1000)
813	/* SCV 0 */
814	mr	r9,r13
815	GET_PACA(r13)
816	mflr	r11
817	mfctr	r12
818	li	r10,IRQS_ALL_DISABLED
819	stb	r10,PACAIRQSOFTMASK(r13)
820#ifdef CONFIG_RELOCATABLE
821	b	system_call_vectored_tramp
822#else
823	b	system_call_vectored_common
824#endif
825	nop
826
827	/* SCV 1 - 127 */
828	.rept	127
829	mr	r9,r13
830	GET_PACA(r13)
831	mflr	r11
832	mfctr	r12
833	li	r10,IRQS_ALL_DISABLED
834	stb	r10,PACAIRQSOFTMASK(r13)
835	li	r0,-1 /* cause failure */
836#ifdef CONFIG_RELOCATABLE
837	b	system_call_vectored_sigill_tramp
838#else
839	b	system_call_vectored_sigill
840#endif
841	.endr
842EXC_VIRT_END(system_call_vectored, 0x3000, 0x1000)
843
844// Treat scv vectors as soft-masked, see comment above.
845// Use absolute values rather than labels here, so they don't get relocated,
846// because this code runs unrelocated.
847SOFT_MASK_TABLE(0xc000000000003000, 0xc000000000004000)
848
849#ifdef CONFIG_RELOCATABLE
850TRAMP_VIRT_BEGIN(system_call_vectored_tramp)
851	__LOAD_HANDLER(r10, system_call_vectored_common)
852	mtctr	r10
853	bctr
854
855TRAMP_VIRT_BEGIN(system_call_vectored_sigill_tramp)
856	__LOAD_HANDLER(r10, system_call_vectored_sigill)
857	mtctr	r10
858	bctr
859#endif
860
861
862/* No virt vectors corresponding with 0x0..0x100 */
863EXC_VIRT_NONE(0x4000, 0x100)
864
865
866/**
867 * Interrupt 0x100 - System Reset Interrupt (SRESET aka NMI).
868 * This is a non-maskable, asynchronous interrupt always taken in real-mode.
869 * It is caused by:
870 * - Wake from power-saving state, on powernv.
871 * - An NMI from another CPU, triggered by firmware or hypercall.
872 * - As crash/debug signal injected from BMC, firmware or hypervisor.
873 *
874 * Handling:
875 * Power-save wakeup is the only performance critical path, so this is
876 * determined quickly as possible first. In this case volatile registers
877 * can be discarded and SPRs like CFAR don't need to be read.
878 *
879 * If not a powersave wakeup, then it's run as a regular interrupt, however
880 * it uses its own stack and PACA save area to preserve the regular kernel
881 * environment for debugging.
882 *
883 * This interrupt is not maskable, so triggering it when MSR[RI] is clear,
884 * or SCRATCH0 is in use, etc. may cause a crash. It's also not entirely
885 * correct to switch to virtual mode to run the regular interrupt handler
886 * because it might be interrupted when the MMU is in a bad state (e.g., SLB
887 * is clear).
888 *
889 * FWNMI:
890 * PAPR specifies a "fwnmi" facility which sends the sreset to a different
891 * entry point with a different register set up. Some hypervisors will
892 * send the sreset to 0x100 in the guest if it is not fwnmi capable.
893 *
894 * KVM:
895 * Unlike most SRR interrupts, this may be taken by the host while executing
896 * in a guest, so a KVM test is required. KVM will pull the CPU out of guest
897 * mode and then raise the sreset.
898 */
899INT_DEFINE_BEGIN(system_reset)
900	IVEC=0x100
901	IAREA=PACA_EXNMI
902	IVIRT=0 /* no virt entry point */
903	/*
904	 * MSR_RI is not enabled, because PACA_EXNMI and nmi stack is
905	 * being used, so a nested NMI exception would corrupt it.
906	 */
907	ISET_RI=0
908	ISTACK=0
909	IKVM_REAL=1
910INT_DEFINE_END(system_reset)
911
912EXC_REAL_BEGIN(system_reset, 0x100, 0x100)
913#ifdef CONFIG_PPC_P7_NAP
914	/*
915	 * If running native on arch 2.06 or later, check if we are waking up
916	 * from nap/sleep/winkle, and branch to idle handler. This tests SRR1
917	 * bits 46:47. A non-0 value indicates that we are coming from a power
918	 * saving state. The idle wakeup handler initially runs in real mode,
919	 * but we branch to the 0xc000... address so we can turn on relocation
920	 * with mtmsrd later, after SPRs are restored.
921	 *
922	 * Careful to minimise cost for the fast path (idle wakeup) while
923	 * also avoiding clobbering CFAR for the debug path (non-idle).
924	 *
925	 * For the idle wake case volatile registers can be clobbered, which
926	 * is why we use those initially. If it turns out to not be an idle
927	 * wake, carefully put everything back the way it was, so we can use
928	 * common exception macros to handle it.
929	 */
930BEGIN_FTR_SECTION
931	SET_SCRATCH0(r13)
932	GET_PACA(r13)
933	std	r3,PACA_EXNMI+0*8(r13)
934	std	r4,PACA_EXNMI+1*8(r13)
935	std	r5,PACA_EXNMI+2*8(r13)
936	mfspr	r3,SPRN_SRR1
937	mfocrf	r4,0x80
938	rlwinm.	r5,r3,47-31,30,31
939	bne+	system_reset_idle_wake
940	/* Not powersave wakeup. Restore regs for regular interrupt handler. */
941	mtocrf	0x80,r4
942	ld	r3,PACA_EXNMI+0*8(r13)
943	ld	r4,PACA_EXNMI+1*8(r13)
944	ld	r5,PACA_EXNMI+2*8(r13)
945	GET_SCRATCH0(r13)
946END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
947#endif
948
949	GEN_INT_ENTRY system_reset, virt=0
950	/*
951	 * In theory, we should not enable relocation here if it was disabled
952	 * in SRR1, because the MMU may not be configured to support it (e.g.,
953	 * SLB may have been cleared). In practice, there should only be a few
954	 * small windows where that's the case, and sreset is considered to
955	 * be dangerous anyway.
956	 */
957EXC_REAL_END(system_reset, 0x100, 0x100)
958EXC_VIRT_NONE(0x4100, 0x100)
959
960#ifdef CONFIG_PPC_P7_NAP
961TRAMP_REAL_BEGIN(system_reset_idle_wake)
962	/* We are waking up from idle, so may clobber any volatile register */
963	cmpwi	cr1,r5,2
964	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
965	BRANCH_TO_C000(r12, DOTSYM(idle_return_gpr_loss))
966#endif
967
968#ifdef CONFIG_PPC_PSERIES
969/*
970 * Vectors for the FWNMI option.  Share common code.
971 */
972TRAMP_REAL_BEGIN(system_reset_fwnmi)
973	GEN_INT_ENTRY system_reset, virt=0
974
975#endif /* CONFIG_PPC_PSERIES */
976
977EXC_COMMON_BEGIN(system_reset_common)
978	__GEN_COMMON_ENTRY system_reset
979	/*
980	 * Increment paca->in_nmi then enable MSR_RI. SLB or MCE will be able
981	 * to recover, but nested NMI will notice in_nmi and not recover
982	 * because of the use of the NMI stack. in_nmi reentrancy is tested in
983	 * system_reset_exception.
984	 */
985	lhz	r10,PACA_IN_NMI(r13)
986	addi	r10,r10,1
987	sth	r10,PACA_IN_NMI(r13)
988	li	r10,MSR_RI
989	mtmsrd 	r10,1
990
991	mr	r10,r1
992	ld	r1,PACA_NMI_EMERG_SP(r13)
993	subi	r1,r1,INT_FRAME_SIZE
994	__GEN_COMMON_BODY system_reset
995
996	addi	r3,r1,STACK_FRAME_OVERHEAD
997	bl	system_reset_exception
998
999	/* Clear MSR_RI before setting SRR0 and SRR1. */
1000	li	r9,0
1001	mtmsrd	r9,1
1002
1003	/*
1004	 * MSR_RI is clear, now we can decrement paca->in_nmi.
1005	 */
1006	lhz	r10,PACA_IN_NMI(r13)
1007	subi	r10,r10,1
1008	sth	r10,PACA_IN_NMI(r13)
1009
1010	kuap_kernel_restore r9, r10
1011	EXCEPTION_RESTORE_REGS
1012	RFI_TO_USER_OR_KERNEL
1013
1014
1015/**
1016 * Interrupt 0x200 - Machine Check Interrupt (MCE).
1017 * This is a non-maskable interrupt always taken in real-mode. It can be
1018 * synchronous or asynchronous, caused by hardware or software, and it may be
1019 * taken in a power-saving state.
1020 *
1021 * Handling:
1022 * Similarly to system reset, this uses its own stack and PACA save area,
1023 * the difference is re-entrancy is allowed on the machine check stack.
1024 *
1025 * machine_check_early is run in real mode, and carefully decodes the
1026 * machine check and tries to handle it (e.g., flush the SLB if there was an
1027 * error detected there), determines if it was recoverable and logs the
1028 * event.
1029 *
1030 * This early code does not "reconcile" irq soft-mask state like SRESET or
1031 * regular interrupts do, so irqs_disabled() among other things may not work
1032 * properly (irq disable/enable already doesn't work because irq tracing can
1033 * not work in real mode).
1034 *
1035 * Then, depending on the execution context when the interrupt is taken, there
1036 * are 3 main actions:
1037 * - Executing in kernel mode. The event is queued with irq_work, which means
1038 *   it is handled when it is next safe to do so (i.e., the kernel has enabled
1039 *   interrupts), which could be immediately when the interrupt returns. This
1040 *   avoids nasty issues like switching to virtual mode when the MMU is in a
1041 *   bad state, or when executing OPAL code. (SRESET is exposed to such issues,
1042 *   but it has different priorities). Check to see if the CPU was in power
1043 *   save, and return via the wake up code if it was.
1044 *
1045 * - Executing in user mode. machine_check_exception is run like a normal
1046 *   interrupt handler, which processes the data generated by the early handler.
1047 *
1048 * - Executing in guest mode. The interrupt is run with its KVM test, and
1049 *   branches to KVM to deal with. KVM may queue the event for the host
1050 *   to report later.
1051 *
1052 * This interrupt is not maskable, so if it triggers when MSR[RI] is clear,
1053 * or SCRATCH0 is in use, it may cause a crash.
1054 *
1055 * KVM:
1056 * See SRESET.
1057 */
1058INT_DEFINE_BEGIN(machine_check_early)
1059	IVEC=0x200
1060	IAREA=PACA_EXMC
1061	IVIRT=0 /* no virt entry point */
1062	IREALMODE_COMMON=1
1063	/*
1064	 * MSR_RI is not enabled, because PACA_EXMC is being used, so a
1065	 * nested machine check corrupts it. machine_check_common enables
1066	 * MSR_RI.
1067	 */
1068	ISET_RI=0
1069	ISTACK=0
1070	IDAR=1
1071	IDSISR=1
1072	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
1073INT_DEFINE_END(machine_check_early)
1074
1075INT_DEFINE_BEGIN(machine_check)
1076	IVEC=0x200
1077	IAREA=PACA_EXMC
1078	IVIRT=0 /* no virt entry point */
1079	ISET_RI=0
1080	IDAR=1
1081	IDSISR=1
1082	IKVM_REAL=1
1083INT_DEFINE_END(machine_check)
1084
1085EXC_REAL_BEGIN(machine_check, 0x200, 0x100)
1086	GEN_INT_ENTRY machine_check_early, virt=0
1087EXC_REAL_END(machine_check, 0x200, 0x100)
1088EXC_VIRT_NONE(0x4200, 0x100)
1089
1090#ifdef CONFIG_PPC_PSERIES
1091TRAMP_REAL_BEGIN(machine_check_fwnmi)
1092	/* See comment at machine_check exception, don't turn on RI */
1093	GEN_INT_ENTRY machine_check_early, virt=0
1094#endif
1095
1096#define MACHINE_CHECK_HANDLER_WINDUP			\
1097	/* Clear MSR_RI before setting SRR0 and SRR1. */\
1098	li	r9,0;					\
1099	mtmsrd	r9,1;		/* Clear MSR_RI */	\
1100	/* Decrement paca->in_mce now RI is clear. */	\
1101	lhz	r12,PACA_IN_MCE(r13);			\
1102	subi	r12,r12,1;				\
1103	sth	r12,PACA_IN_MCE(r13);			\
1104	EXCEPTION_RESTORE_REGS
1105
1106EXC_COMMON_BEGIN(machine_check_early_common)
1107	__GEN_REALMODE_COMMON_ENTRY machine_check_early
1108
1109	/*
1110	 * Switch to mc_emergency stack and handle re-entrancy (we limit
1111	 * the nested MCE upto level 4 to avoid stack overflow).
1112	 * Save MCE registers srr1, srr0, dar and dsisr and then set ME=1
1113	 *
1114	 * We use paca->in_mce to check whether this is the first entry or
1115	 * nested machine check. We increment paca->in_mce to track nested
1116	 * machine checks.
1117	 *
1118	 * If this is the first entry then set stack pointer to
1119	 * paca->mc_emergency_sp, otherwise r1 is already pointing to
1120	 * stack frame on mc_emergency stack.
1121	 *
1122	 * NOTE: We are here with MSR_ME=0 (off), which means we risk a
1123	 * checkstop if we get another machine check exception before we do
1124	 * rfid with MSR_ME=1.
1125	 *
1126	 * This interrupt can wake directly from idle. If that is the case,
1127	 * the machine check is handled then the idle wakeup code is called
1128	 * to restore state.
1129	 */
1130	lhz	r10,PACA_IN_MCE(r13)
1131	cmpwi	r10,0			/* Are we in nested machine check */
1132	cmpwi	cr1,r10,MAX_MCE_DEPTH	/* Are we at maximum nesting */
1133	addi	r10,r10,1		/* increment paca->in_mce */
1134	sth	r10,PACA_IN_MCE(r13)
1135
1136	mr	r10,r1			/* Save r1 */
1137	bne	1f
1138	/* First machine check entry */
1139	ld	r1,PACAMCEMERGSP(r13)	/* Use MC emergency stack */
11401:	/* Limit nested MCE to level 4 to avoid stack overflow */
1141	bgt	cr1,unrecoverable_mce	/* Check if we hit limit of 4 */
1142	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame */
1143
1144	__GEN_COMMON_BODY machine_check_early
1145
1146BEGIN_FTR_SECTION
1147	bl	enable_machine_check
1148END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1149	li	r10,MSR_RI
1150	mtmsrd	r10,1
1151
1152	addi	r3,r1,STACK_FRAME_OVERHEAD
1153	bl	machine_check_early
1154	std	r3,RESULT(r1)	/* Save result */
1155	ld	r12,_MSR(r1)
1156
1157#ifdef CONFIG_PPC_P7_NAP
1158	/*
1159	 * Check if thread was in power saving mode. We come here when any
1160	 * of the following is true:
1161	 * a. thread wasn't in power saving mode
1162	 * b. thread was in power saving mode with no state loss,
1163	 *    supervisor state loss or hypervisor state loss.
1164	 *
1165	 * Go back to nap/sleep/winkle mode again if (b) is true.
1166	 */
1167BEGIN_FTR_SECTION
1168	rlwinm.	r11,r12,47-31,30,31
1169	bne	machine_check_idle_common
1170END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
1171#endif
1172
1173#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1174	/*
1175	 * Check if we are coming from guest. If yes, then run the normal
1176	 * exception handler which will take the
1177	 * machine_check_kvm->kvm_interrupt branch to deliver the MC event
1178	 * to guest.
1179	 */
1180	lbz	r11,HSTATE_IN_GUEST(r13)
1181	cmpwi	r11,0			/* Check if coming from guest */
1182	bne	mce_deliver		/* continue if we are. */
1183#endif
1184
1185	/*
1186	 * Check if we are coming from userspace. If yes, then run the normal
1187	 * exception handler which will deliver the MC event to this kernel.
1188	 */
1189	andi.	r11,r12,MSR_PR		/* See if coming from user. */
1190	bne	mce_deliver		/* continue in V mode if we are. */
1191
1192	/*
1193	 * At this point we are coming from kernel context.
1194	 * Queue up the MCE event and return from the interrupt.
1195	 * But before that, check if this is an un-recoverable exception.
1196	 * If yes, then stay on emergency stack and panic.
1197	 */
1198	andi.	r11,r12,MSR_RI
1199	beq	unrecoverable_mce
1200
1201	/*
1202	 * Check if we have successfully handled/recovered from error, if not
1203	 * then stay on emergency stack and panic.
1204	 */
1205	ld	r3,RESULT(r1)	/* Load result */
1206	cmpdi	r3,0		/* see if we handled MCE successfully */
1207	beq	unrecoverable_mce /* if !handled then panic */
1208
1209	/*
1210	 * Return from MC interrupt.
1211	 * Queue up the MCE event so that we can log it later, while
1212	 * returning from kernel or opal call.
1213	 */
1214	bl	machine_check_queue_event
1215	MACHINE_CHECK_HANDLER_WINDUP
1216	RFI_TO_KERNEL
1217
1218mce_deliver:
1219	/*
1220	 * This is a host user or guest MCE. Restore all registers, then
1221	 * run the "late" handler. For host user, this will run the
1222	 * machine_check_exception handler in virtual mode like a normal
1223	 * interrupt handler. For guest, this will trigger the KVM test
1224	 * and branch to the KVM interrupt similarly to other interrupts.
1225	 */
1226BEGIN_FTR_SECTION
1227	ld	r10,ORIG_GPR3(r1)
1228	mtspr	SPRN_CFAR,r10
1229END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
1230	MACHINE_CHECK_HANDLER_WINDUP
1231	GEN_INT_ENTRY machine_check, virt=0
1232
1233EXC_COMMON_BEGIN(machine_check_common)
1234	/*
1235	 * Machine check is different because we use a different
1236	 * save area: PACA_EXMC instead of PACA_EXGEN.
1237	 */
1238	GEN_COMMON machine_check
1239
1240	/* Enable MSR_RI when finished with PACA_EXMC */
1241	li	r10,MSR_RI
1242	mtmsrd 	r10,1
1243	addi	r3,r1,STACK_FRAME_OVERHEAD
1244	bl	machine_check_exception_async
1245	b	interrupt_return_srr
1246
1247
1248#ifdef CONFIG_PPC_P7_NAP
1249/*
1250 * This is an idle wakeup. Low level machine check has already been
1251 * done. Queue the event then call the idle code to do the wake up.
1252 */
1253EXC_COMMON_BEGIN(machine_check_idle_common)
1254	bl	machine_check_queue_event
1255
1256	/*
1257	 * GPR-loss wakeups are relatively straightforward, because the
1258	 * idle sleep code has saved all non-volatile registers on its
1259	 * own stack, and r1 in PACAR1.
1260	 *
1261	 * For no-loss wakeups the r1 and lr registers used by the
1262	 * early machine check handler have to be restored first. r2 is
1263	 * the kernel TOC, so no need to restore it.
1264	 *
1265	 * Then decrement MCE nesting after finishing with the stack.
1266	 */
1267	ld	r3,_MSR(r1)
1268	ld	r4,_LINK(r1)
1269	ld	r1,GPR1(r1)
1270
1271	lhz	r11,PACA_IN_MCE(r13)
1272	subi	r11,r11,1
1273	sth	r11,PACA_IN_MCE(r13)
1274
1275	mtlr	r4
1276	rlwinm	r10,r3,47-31,30,31
1277	cmpwi	cr1,r10,2
1278	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
1279	b	idle_return_gpr_loss
1280#endif
1281
1282EXC_COMMON_BEGIN(unrecoverable_mce)
1283	/*
1284	 * We are going down. But there are chances that we might get hit by
1285	 * another MCE during panic path and we may run into unstable state
1286	 * with no way out. Hence, turn ME bit off while going down, so that
1287	 * when another MCE is hit during panic path, system will checkstop
1288	 * and hypervisor will get restarted cleanly by SP.
1289	 */
1290BEGIN_FTR_SECTION
1291	li	r10,0 /* clear MSR_RI */
1292	mtmsrd	r10,1
1293	bl	disable_machine_check
1294END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1295	ld	r10,PACAKMSR(r13)
1296	li	r3,MSR_ME
1297	andc	r10,r10,r3
1298	mtmsrd	r10
1299
1300	lhz	r12,PACA_IN_MCE(r13)
1301	subi	r12,r12,1
1302	sth	r12,PACA_IN_MCE(r13)
1303
1304	/*
1305	 * Invoke machine_check_exception to print MCE event and panic.
1306	 * This is the NMI version of the handler because we are called from
1307	 * the early handler which is a true NMI.
1308	 */
1309	addi	r3,r1,STACK_FRAME_OVERHEAD
1310	bl	machine_check_exception
1311
1312	/*
1313	 * We will not reach here. Even if we did, there is no way out.
1314	 * Call unrecoverable_exception and die.
1315	 */
1316	addi	r3,r1,STACK_FRAME_OVERHEAD
1317	bl	unrecoverable_exception
1318	b	.
1319
1320
1321/**
1322 * Interrupt 0x300 - Data Storage Interrupt (DSI).
1323 * This is a synchronous interrupt generated due to a data access exception,
1324 * e.g., a load orstore which does not have a valid page table entry with
1325 * permissions. DAWR matches also fault here, as do RC updates, and minor misc
1326 * errors e.g., copy/paste, AMO, certain invalid CI accesses, etc.
1327 *
1328 * Handling:
1329 * - Hash MMU
1330 *   Go to do_hash_fault, which attempts to fill the HPT from an entry in the
1331 *   Linux page table. Hash faults can hit in kernel mode in a fairly
1332 *   arbitrary state (e.g., interrupts disabled, locks held) when accessing
1333 *   "non-bolted" regions, e.g., vmalloc space. However these should always be
1334 *   backed by Linux page table entries.
1335 *
1336 *   If no entry is found the Linux page fault handler is invoked (by
1337 *   do_hash_fault). Linux page faults can happen in kernel mode due to user
1338 *   copy operations of course.
1339 *
1340 *   KVM: The KVM HDSI handler may perform a load with MSR[DR]=1 in guest
1341 *   MMU context, which may cause a DSI in the host, which must go to the
1342 *   KVM handler. MSR[IR] is not enabled, so the real-mode handler will
1343 *   always be used regardless of AIL setting.
1344 *
1345 * - Radix MMU
1346 *   The hardware loads from the Linux page table directly, so a fault goes
1347 *   immediately to Linux page fault.
1348 *
1349 * Conditions like DAWR match are handled on the way in to Linux page fault.
1350 */
1351INT_DEFINE_BEGIN(data_access)
1352	IVEC=0x300
1353	IDAR=1
1354	IDSISR=1
1355	IKVM_REAL=1
1356INT_DEFINE_END(data_access)
1357
1358EXC_REAL_BEGIN(data_access, 0x300, 0x80)
1359	GEN_INT_ENTRY data_access, virt=0
1360EXC_REAL_END(data_access, 0x300, 0x80)
1361EXC_VIRT_BEGIN(data_access, 0x4300, 0x80)
1362	GEN_INT_ENTRY data_access, virt=1
1363EXC_VIRT_END(data_access, 0x4300, 0x80)
1364EXC_COMMON_BEGIN(data_access_common)
1365	GEN_COMMON data_access
1366	ld	r4,_DSISR(r1)
1367	addi	r3,r1,STACK_FRAME_OVERHEAD
1368	andis.	r0,r4,DSISR_DABRMATCH@h
1369	bne-	1f
1370#ifdef CONFIG_PPC_64S_HASH_MMU
1371BEGIN_MMU_FTR_SECTION
1372	bl	do_hash_fault
1373MMU_FTR_SECTION_ELSE
1374	bl	do_page_fault
1375ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1376#else
1377	bl	do_page_fault
1378#endif
1379	b	interrupt_return_srr
1380
13811:	bl	do_break
1382	/*
1383	 * do_break() may have changed the NV GPRS while handling a breakpoint.
1384	 * If so, we need to restore them with their updated values.
1385	 */
1386	REST_NVGPRS(r1)
1387	b	interrupt_return_srr
1388
1389
1390/**
1391 * Interrupt 0x380 - Data Segment Interrupt (DSLB).
1392 * This is a synchronous interrupt in response to an MMU fault missing SLB
1393 * entry for HPT, or an address outside RPT translation range.
1394 *
1395 * Handling:
1396 * - HPT:
1397 *   This refills the SLB, or reports an access fault similarly to a bad page
1398 *   fault. When coming from user-mode, the SLB handler may access any kernel
1399 *   data, though it may itself take a DSLB. When coming from kernel mode,
1400 *   recursive faults must be avoided so access is restricted to the kernel
1401 *   image text/data, kernel stack, and any data allocated below
1402 *   ppc64_bolted_size (first segment). The kernel handler must avoid stomping
1403 *   on user-handler data structures.
1404 *
1405 *   KVM: Same as 0x300, DSLB must test for KVM guest.
1406 */
1407INT_DEFINE_BEGIN(data_access_slb)
1408	IVEC=0x380
1409	IDAR=1
1410	IKVM_REAL=1
1411INT_DEFINE_END(data_access_slb)
1412
1413EXC_REAL_BEGIN(data_access_slb, 0x380, 0x80)
1414	GEN_INT_ENTRY data_access_slb, virt=0
1415EXC_REAL_END(data_access_slb, 0x380, 0x80)
1416EXC_VIRT_BEGIN(data_access_slb, 0x4380, 0x80)
1417	GEN_INT_ENTRY data_access_slb, virt=1
1418EXC_VIRT_END(data_access_slb, 0x4380, 0x80)
1419EXC_COMMON_BEGIN(data_access_slb_common)
1420	GEN_COMMON data_access_slb
1421#ifdef CONFIG_PPC_64S_HASH_MMU
1422BEGIN_MMU_FTR_SECTION
1423	/* HPT case, do SLB fault */
1424	addi	r3,r1,STACK_FRAME_OVERHEAD
1425	bl	do_slb_fault
1426	cmpdi	r3,0
1427	bne-	1f
1428	b	fast_interrupt_return_srr
14291:	/* Error case */
1430MMU_FTR_SECTION_ELSE
1431	/* Radix case, access is outside page table range */
1432	li	r3,-EFAULT
1433ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1434#else
1435	li	r3,-EFAULT
1436#endif
1437	std	r3,RESULT(r1)
1438	addi	r3,r1,STACK_FRAME_OVERHEAD
1439	bl	do_bad_segment_interrupt
1440	b	interrupt_return_srr
1441
1442
1443/**
1444 * Interrupt 0x400 - Instruction Storage Interrupt (ISI).
1445 * This is a synchronous interrupt in response to an MMU fault due to an
1446 * instruction fetch.
1447 *
1448 * Handling:
1449 * Similar to DSI, though in response to fetch. The faulting address is found
1450 * in SRR0 (rather than DAR), and status in SRR1 (rather than DSISR).
1451 */
1452INT_DEFINE_BEGIN(instruction_access)
1453	IVEC=0x400
1454	IISIDE=1
1455	IDAR=1
1456	IDSISR=1
1457#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1458	IKVM_REAL=1
1459#endif
1460INT_DEFINE_END(instruction_access)
1461
1462EXC_REAL_BEGIN(instruction_access, 0x400, 0x80)
1463	GEN_INT_ENTRY instruction_access, virt=0
1464EXC_REAL_END(instruction_access, 0x400, 0x80)
1465EXC_VIRT_BEGIN(instruction_access, 0x4400, 0x80)
1466	GEN_INT_ENTRY instruction_access, virt=1
1467EXC_VIRT_END(instruction_access, 0x4400, 0x80)
1468EXC_COMMON_BEGIN(instruction_access_common)
1469	GEN_COMMON instruction_access
1470	addi	r3,r1,STACK_FRAME_OVERHEAD
1471#ifdef CONFIG_PPC_64S_HASH_MMU
1472BEGIN_MMU_FTR_SECTION
1473	bl	do_hash_fault
1474MMU_FTR_SECTION_ELSE
1475	bl	do_page_fault
1476ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1477#else
1478	bl	do_page_fault
1479#endif
1480	b	interrupt_return_srr
1481
1482
1483/**
1484 * Interrupt 0x480 - Instruction Segment Interrupt (ISLB).
1485 * This is a synchronous interrupt in response to an MMU fault due to an
1486 * instruction fetch.
1487 *
1488 * Handling:
1489 * Similar to DSLB, though in response to fetch. The faulting address is found
1490 * in SRR0 (rather than DAR).
1491 */
1492INT_DEFINE_BEGIN(instruction_access_slb)
1493	IVEC=0x480
1494	IISIDE=1
1495	IDAR=1
1496#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1497	IKVM_REAL=1
1498#endif
1499INT_DEFINE_END(instruction_access_slb)
1500
1501EXC_REAL_BEGIN(instruction_access_slb, 0x480, 0x80)
1502	GEN_INT_ENTRY instruction_access_slb, virt=0
1503EXC_REAL_END(instruction_access_slb, 0x480, 0x80)
1504EXC_VIRT_BEGIN(instruction_access_slb, 0x4480, 0x80)
1505	GEN_INT_ENTRY instruction_access_slb, virt=1
1506EXC_VIRT_END(instruction_access_slb, 0x4480, 0x80)
1507EXC_COMMON_BEGIN(instruction_access_slb_common)
1508	GEN_COMMON instruction_access_slb
1509#ifdef CONFIG_PPC_64S_HASH_MMU
1510BEGIN_MMU_FTR_SECTION
1511	/* HPT case, do SLB fault */
1512	addi	r3,r1,STACK_FRAME_OVERHEAD
1513	bl	do_slb_fault
1514	cmpdi	r3,0
1515	bne-	1f
1516	b	fast_interrupt_return_srr
15171:	/* Error case */
1518MMU_FTR_SECTION_ELSE
1519	/* Radix case, access is outside page table range */
1520	li	r3,-EFAULT
1521ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1522#else
1523	li	r3,-EFAULT
1524#endif
1525	std	r3,RESULT(r1)
1526	addi	r3,r1,STACK_FRAME_OVERHEAD
1527	bl	do_bad_segment_interrupt
1528	b	interrupt_return_srr
1529
1530
1531/**
1532 * Interrupt 0x500 - External Interrupt.
1533 * This is an asynchronous maskable interrupt in response to an "external
1534 * exception" from the interrupt controller or hypervisor (e.g., device
1535 * interrupt). It is maskable in hardware by clearing MSR[EE], and
1536 * soft-maskable with IRQS_DISABLED mask (i.e., local_irq_disable()).
1537 *
1538 * When running in HV mode, Linux sets up the LPCR[LPES] bit such that
1539 * interrupts are delivered with HSRR registers, guests use SRRs, which
1540 * reqiures IHSRR_IF_HVMODE.
1541 *
1542 * On bare metal POWER9 and later, Linux sets the LPCR[HVICE] bit such that
1543 * external interrupts are delivered as Hypervisor Virtualization Interrupts
1544 * rather than External Interrupts.
1545 *
1546 * Handling:
1547 * This calls into Linux IRQ handler. NVGPRs are not saved to reduce overhead,
1548 * because registers at the time of the interrupt are not so important as it is
1549 * asynchronous.
1550 *
1551 * If soft masked, the masked handler will note the pending interrupt for
1552 * replay, and clear MSR[EE] in the interrupted context.
1553 */
1554INT_DEFINE_BEGIN(hardware_interrupt)
1555	IVEC=0x500
1556	IHSRR_IF_HVMODE=1
1557	IMASK=IRQS_DISABLED
1558	IKVM_REAL=1
1559	IKVM_VIRT=1
1560INT_DEFINE_END(hardware_interrupt)
1561
1562EXC_REAL_BEGIN(hardware_interrupt, 0x500, 0x100)
1563	GEN_INT_ENTRY hardware_interrupt, virt=0
1564EXC_REAL_END(hardware_interrupt, 0x500, 0x100)
1565EXC_VIRT_BEGIN(hardware_interrupt, 0x4500, 0x100)
1566	GEN_INT_ENTRY hardware_interrupt, virt=1
1567EXC_VIRT_END(hardware_interrupt, 0x4500, 0x100)
1568EXC_COMMON_BEGIN(hardware_interrupt_common)
1569	GEN_COMMON hardware_interrupt
1570	addi	r3,r1,STACK_FRAME_OVERHEAD
1571	bl	do_IRQ
1572	BEGIN_FTR_SECTION
1573	b	interrupt_return_hsrr
1574	FTR_SECTION_ELSE
1575	b	interrupt_return_srr
1576	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
1577
1578
1579/**
1580 * Interrupt 0x600 - Alignment Interrupt
1581 * This is a synchronous interrupt in response to data alignment fault.
1582 */
1583INT_DEFINE_BEGIN(alignment)
1584	IVEC=0x600
1585	IDAR=1
1586	IDSISR=1
1587#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1588	IKVM_REAL=1
1589#endif
1590INT_DEFINE_END(alignment)
1591
1592EXC_REAL_BEGIN(alignment, 0x600, 0x100)
1593	GEN_INT_ENTRY alignment, virt=0
1594EXC_REAL_END(alignment, 0x600, 0x100)
1595EXC_VIRT_BEGIN(alignment, 0x4600, 0x100)
1596	GEN_INT_ENTRY alignment, virt=1
1597EXC_VIRT_END(alignment, 0x4600, 0x100)
1598EXC_COMMON_BEGIN(alignment_common)
1599	GEN_COMMON alignment
1600	addi	r3,r1,STACK_FRAME_OVERHEAD
1601	bl	alignment_exception
1602	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
1603	b	interrupt_return_srr
1604
1605
1606/**
1607 * Interrupt 0x700 - Program Interrupt (program check).
1608 * This is a synchronous interrupt in response to various instruction faults:
1609 * traps, privilege errors, TM errors, floating point exceptions.
1610 *
1611 * Handling:
1612 * This interrupt may use the "emergency stack" in some cases when being taken
1613 * from kernel context, which complicates handling.
1614 */
1615INT_DEFINE_BEGIN(program_check)
1616	IVEC=0x700
1617#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1618	IKVM_REAL=1
1619#endif
1620INT_DEFINE_END(program_check)
1621
1622EXC_REAL_BEGIN(program_check, 0x700, 0x100)
1623
1624#ifdef CONFIG_CPU_LITTLE_ENDIAN
1625	/*
1626	 * There's a short window during boot where although the kernel is
1627	 * running little endian, any exceptions will cause the CPU to switch
1628	 * back to big endian. For example a WARN() boils down to a trap
1629	 * instruction, which will cause a program check, and we end up here but
1630	 * with the CPU in big endian mode. The first instruction of the program
1631	 * check handler (in GEN_INT_ENTRY below) is an mtsprg, which when
1632	 * executed in the wrong endian is an lhzu with a ~3GB displacement from
1633	 * r3. The content of r3 is random, so that is a load from some random
1634	 * location, and depending on the system can easily lead to a checkstop,
1635	 * or an infinitely recursive page fault.
1636	 *
1637	 * So to handle that case we have a trampoline here that can detect we
1638	 * are in the wrong endian and flip us back to the correct endian. We
1639	 * can't flip MSR[LE] using mtmsr, so we have to use rfid. That requires
1640	 * backing up SRR0/1 as well as a GPR. To do that we use SPRG0/2/3, as
1641	 * SPRG1 is already used for the paca. SPRG3 is user readable, but this
1642	 * trampoline is only active very early in boot, and SPRG3 will be
1643	 * reinitialised in vdso_getcpu_init() before userspace starts.
1644	 */
1645BEGIN_FTR_SECTION
1646	tdi   0,0,0x48    // Trap never, or in reverse endian: b . + 8
1647	b     1f          // Skip trampoline if endian is correct
1648	.long 0xa643707d  // mtsprg  0, r11      Backup r11
1649	.long 0xa6027a7d  // mfsrr0  r11
1650	.long 0xa643727d  // mtsprg  2, r11      Backup SRR0 in SPRG2
1651	.long 0xa6027b7d  // mfsrr1  r11
1652	.long 0xa643737d  // mtsprg  3, r11      Backup SRR1 in SPRG3
1653	.long 0xa600607d  // mfmsr   r11
1654	.long 0x01006b69  // xori    r11, r11, 1 Invert MSR[LE]
1655	.long 0xa6037b7d  // mtsrr1  r11
1656	.long 0x34076039  // li      r11, 0x734
1657	.long 0xa6037a7d  // mtsrr0  r11
1658	.long 0x2400004c  // rfid
1659	mfsprg r11, 3
1660	mtsrr1 r11        // Restore SRR1
1661	mfsprg r11, 2
1662	mtsrr0 r11        // Restore SRR0
1663	mfsprg r11, 0     // Restore r11
16641:
1665END_FTR_SECTION(0, 1)     // nop out after boot
1666#endif /* CONFIG_CPU_LITTLE_ENDIAN */
1667
1668	GEN_INT_ENTRY program_check, virt=0
1669EXC_REAL_END(program_check, 0x700, 0x100)
1670EXC_VIRT_BEGIN(program_check, 0x4700, 0x100)
1671	GEN_INT_ENTRY program_check, virt=1
1672EXC_VIRT_END(program_check, 0x4700, 0x100)
1673EXC_COMMON_BEGIN(program_check_common)
1674	__GEN_COMMON_ENTRY program_check
1675
1676	/*
1677	 * It's possible to receive a TM Bad Thing type program check with
1678	 * userspace register values (in particular r1), but with SRR1 reporting
1679	 * that we came from the kernel. Normally that would confuse the bad
1680	 * stack logic, and we would report a bad kernel stack pointer. Instead
1681	 * we switch to the emergency stack if we're taking a TM Bad Thing from
1682	 * the kernel.
1683	 */
1684
1685	andi.	r10,r12,MSR_PR
1686	bne	.Lnormal_stack		/* If userspace, go normal path */
1687
1688	andis.	r10,r12,(SRR1_PROGTM)@h
1689	bne	.Lemergency_stack	/* If TM, emergency		*/
1690
1691	cmpdi	r1,-INT_FRAME_SIZE	/* check if r1 is in userspace	*/
1692	blt	.Lnormal_stack		/* normal path if not		*/
1693
1694	/* Use the emergency stack					*/
1695.Lemergency_stack:
1696	andi.	r10,r12,MSR_PR		/* Set CR0 correctly for label	*/
1697					/* 3 in EXCEPTION_PROLOG_COMMON	*/
1698	mr	r10,r1			/* Save r1			*/
1699	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack		*/
1700	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
1701	__ISTACK(program_check)=0
1702	__GEN_COMMON_BODY program_check
1703	b .Ldo_program_check
1704
1705.Lnormal_stack:
1706	__ISTACK(program_check)=1
1707	__GEN_COMMON_BODY program_check
1708
1709.Ldo_program_check:
1710	addi	r3,r1,STACK_FRAME_OVERHEAD
1711	bl	program_check_exception
1712	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
1713	b	interrupt_return_srr
1714
1715
1716/*
1717 * Interrupt 0x800 - Floating-Point Unavailable Interrupt.
1718 * This is a synchronous interrupt in response to executing an fp instruction
1719 * with MSR[FP]=0.
1720 *
1721 * Handling:
1722 * This will load FP registers and enable the FP bit if coming from userspace,
1723 * otherwise report a bad kernel use of FP.
1724 */
1725INT_DEFINE_BEGIN(fp_unavailable)
1726	IVEC=0x800
1727#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1728	IKVM_REAL=1
1729#endif
1730INT_DEFINE_END(fp_unavailable)
1731
1732EXC_REAL_BEGIN(fp_unavailable, 0x800, 0x100)
1733	GEN_INT_ENTRY fp_unavailable, virt=0
1734EXC_REAL_END(fp_unavailable, 0x800, 0x100)
1735EXC_VIRT_BEGIN(fp_unavailable, 0x4800, 0x100)
1736	GEN_INT_ENTRY fp_unavailable, virt=1
1737EXC_VIRT_END(fp_unavailable, 0x4800, 0x100)
1738EXC_COMMON_BEGIN(fp_unavailable_common)
1739	GEN_COMMON fp_unavailable
1740	bne	1f			/* if from user, just load it up */
1741	addi	r3,r1,STACK_FRAME_OVERHEAD
1742	bl	kernel_fp_unavailable_exception
17430:	trap
1744	EMIT_BUG_ENTRY 0b, __FILE__, __LINE__, 0
17451:
1746#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1747BEGIN_FTR_SECTION
1748	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
1749	 * transaction), go do TM stuff
1750	 */
1751	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
1752	bne-	2f
1753END_FTR_SECTION_IFSET(CPU_FTR_TM)
1754#endif
1755	bl	load_up_fpu
1756	b	fast_interrupt_return_srr
1757#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
17582:	/* User process was in a transaction */
1759	addi	r3,r1,STACK_FRAME_OVERHEAD
1760	bl	fp_unavailable_tm
1761	b	interrupt_return_srr
1762#endif
1763
1764
1765/**
1766 * Interrupt 0x900 - Decrementer Interrupt.
1767 * This is an asynchronous interrupt in response to a decrementer exception
1768 * (e.g., DEC has wrapped below zero). It is maskable in hardware by clearing
1769 * MSR[EE], and soft-maskable with IRQS_DISABLED mask (i.e.,
1770 * local_irq_disable()).
1771 *
1772 * Handling:
1773 * This calls into Linux timer handler. NVGPRs are not saved (see 0x500).
1774 *
1775 * If soft masked, the masked handler will note the pending interrupt for
1776 * replay, and bump the decrementer to a high value, leaving MSR[EE] enabled
1777 * in the interrupted context.
1778 * If PPC_WATCHDOG is configured, the soft masked handler will actually set
1779 * things back up to run soft_nmi_interrupt as a regular interrupt handler
1780 * on the emergency stack.
1781 */
1782INT_DEFINE_BEGIN(decrementer)
1783	IVEC=0x900
1784	IMASK=IRQS_DISABLED
1785#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1786	IKVM_REAL=1
1787#endif
1788INT_DEFINE_END(decrementer)
1789
1790EXC_REAL_BEGIN(decrementer, 0x900, 0x80)
1791	GEN_INT_ENTRY decrementer, virt=0
1792EXC_REAL_END(decrementer, 0x900, 0x80)
1793EXC_VIRT_BEGIN(decrementer, 0x4900, 0x80)
1794	GEN_INT_ENTRY decrementer, virt=1
1795EXC_VIRT_END(decrementer, 0x4900, 0x80)
1796EXC_COMMON_BEGIN(decrementer_common)
1797	GEN_COMMON decrementer
1798	addi	r3,r1,STACK_FRAME_OVERHEAD
1799	bl	timer_interrupt
1800	b	interrupt_return_srr
1801
1802
1803/**
1804 * Interrupt 0x980 - Hypervisor Decrementer Interrupt.
1805 * This is an asynchronous interrupt, similar to 0x900 but for the HDEC
1806 * register.
1807 *
1808 * Handling:
1809 * Linux does not use this outside KVM where it's used to keep a host timer
1810 * while the guest is given control of DEC. It should normally be caught by
1811 * the KVM test and routed there.
1812 */
1813INT_DEFINE_BEGIN(hdecrementer)
1814	IVEC=0x980
1815	IHSRR=1
1816	ISTACK=0
1817	IKVM_REAL=1
1818	IKVM_VIRT=1
1819INT_DEFINE_END(hdecrementer)
1820
1821EXC_REAL_BEGIN(hdecrementer, 0x980, 0x80)
1822	GEN_INT_ENTRY hdecrementer, virt=0
1823EXC_REAL_END(hdecrementer, 0x980, 0x80)
1824EXC_VIRT_BEGIN(hdecrementer, 0x4980, 0x80)
1825	GEN_INT_ENTRY hdecrementer, virt=1
1826EXC_VIRT_END(hdecrementer, 0x4980, 0x80)
1827EXC_COMMON_BEGIN(hdecrementer_common)
1828	__GEN_COMMON_ENTRY hdecrementer
1829	/*
1830	 * Hypervisor decrementer interrupts not caught by the KVM test
1831	 * shouldn't occur but are sometimes left pending on exit from a KVM
1832	 * guest.  We don't need to do anything to clear them, as they are
1833	 * edge-triggered.
1834	 *
1835	 * Be careful to avoid touching the kernel stack.
1836	 */
1837	li	r10,0
1838	stb	r10,PACAHSRR_VALID(r13)
1839	ld	r10,PACA_EXGEN+EX_CTR(r13)
1840	mtctr	r10
1841	mtcrf	0x80,r9
1842	ld	r9,PACA_EXGEN+EX_R9(r13)
1843	ld	r10,PACA_EXGEN+EX_R10(r13)
1844	ld	r11,PACA_EXGEN+EX_R11(r13)
1845	ld	r12,PACA_EXGEN+EX_R12(r13)
1846	ld	r13,PACA_EXGEN+EX_R13(r13)
1847	HRFI_TO_KERNEL
1848
1849
1850/**
1851 * Interrupt 0xa00 - Directed Privileged Doorbell Interrupt.
1852 * This is an asynchronous interrupt in response to a msgsndp doorbell.
1853 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
1854 * IRQS_DISABLED mask (i.e., local_irq_disable()).
1855 *
1856 * Handling:
1857 * Guests may use this for IPIs between threads in a core if the
1858 * hypervisor supports it. NVGPRS are not saved (see 0x500).
1859 *
1860 * If soft masked, the masked handler will note the pending interrupt for
1861 * replay, leaving MSR[EE] enabled in the interrupted context because the
1862 * doorbells are edge triggered.
1863 */
1864INT_DEFINE_BEGIN(doorbell_super)
1865	IVEC=0xa00
1866	IMASK=IRQS_DISABLED
1867#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1868	IKVM_REAL=1
1869#endif
1870INT_DEFINE_END(doorbell_super)
1871
1872EXC_REAL_BEGIN(doorbell_super, 0xa00, 0x100)
1873	GEN_INT_ENTRY doorbell_super, virt=0
1874EXC_REAL_END(doorbell_super, 0xa00, 0x100)
1875EXC_VIRT_BEGIN(doorbell_super, 0x4a00, 0x100)
1876	GEN_INT_ENTRY doorbell_super, virt=1
1877EXC_VIRT_END(doorbell_super, 0x4a00, 0x100)
1878EXC_COMMON_BEGIN(doorbell_super_common)
1879	GEN_COMMON doorbell_super
1880	addi	r3,r1,STACK_FRAME_OVERHEAD
1881#ifdef CONFIG_PPC_DOORBELL
1882	bl	doorbell_exception
1883#else
1884	bl	unknown_async_exception
1885#endif
1886	b	interrupt_return_srr
1887
1888
1889EXC_REAL_NONE(0xb00, 0x100)
1890EXC_VIRT_NONE(0x4b00, 0x100)
1891
1892/**
1893 * Interrupt 0xc00 - System Call Interrupt (syscall, hcall).
1894 * This is a synchronous interrupt invoked with the "sc" instruction. The
1895 * system call is invoked with "sc 0" and does not alter the HV bit, so it
1896 * is directed to the currently running OS. The hypercall is invoked with
1897 * "sc 1" and it sets HV=1, so it elevates to hypervisor.
1898 *
1899 * In HPT, sc 1 always goes to 0xc00 real mode. In RADIX, sc 1 can go to
1900 * 0x4c00 virtual mode.
1901 *
1902 * Handling:
1903 * If the KVM test fires then it was due to a hypercall and is accordingly
1904 * routed to KVM. Otherwise this executes a normal Linux system call.
1905 *
1906 * Call convention:
1907 *
1908 * syscall and hypercalls register conventions are documented in
1909 * Documentation/powerpc/syscall64-abi.rst and
1910 * Documentation/powerpc/papr_hcalls.rst respectively.
1911 *
1912 * The intersection of volatile registers that don't contain possible
1913 * inputs is: cr0, xer, ctr. We may use these as scratch regs upon entry
1914 * without saving, though xer is not a good idea to use, as hardware may
1915 * interpret some bits so it may be costly to change them.
1916 */
1917INT_DEFINE_BEGIN(system_call)
1918	IVEC=0xc00
1919	IKVM_REAL=1
1920	IKVM_VIRT=1
1921INT_DEFINE_END(system_call)
1922
1923.macro SYSTEM_CALL virt
1924#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1925	/*
1926	 * There is a little bit of juggling to get syscall and hcall
1927	 * working well. Save r13 in ctr to avoid using SPRG scratch
1928	 * register.
1929	 *
1930	 * Userspace syscalls have already saved the PPR, hcalls must save
1931	 * it before setting HMT_MEDIUM.
1932	 */
1933	mtctr	r13
1934	GET_PACA(r13)
1935	std	r10,PACA_EXGEN+EX_R10(r13)
1936	INTERRUPT_TO_KERNEL
1937	KVMTEST system_call kvm_hcall /* uses r10, branch to kvm_hcall */
1938	mfctr	r9
1939#else
1940	mr	r9,r13
1941	GET_PACA(r13)
1942	INTERRUPT_TO_KERNEL
1943#endif
1944
1945#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
1946BEGIN_FTR_SECTION
1947	cmpdi	r0,0x1ebe
1948	beq-	1f
1949END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE)
1950#endif
1951
1952	/* We reach here with PACA in r13, r13 in r9. */
1953	mfspr	r11,SPRN_SRR0
1954	mfspr	r12,SPRN_SRR1
1955
1956	HMT_MEDIUM
1957
1958	.if ! \virt
1959	__LOAD_HANDLER(r10, system_call_common_real)
1960	mtctr	r10
1961	bctr
1962	.else
1963#ifdef CONFIG_RELOCATABLE
1964	__LOAD_HANDLER(r10, system_call_common)
1965	mtctr	r10
1966	bctr
1967#else
1968	b	system_call_common
1969#endif
1970	.endif
1971
1972#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
1973	/* Fast LE/BE switch system call */
19741:	mfspr	r12,SPRN_SRR1
1975	xori	r12,r12,MSR_LE
1976	mtspr	SPRN_SRR1,r12
1977	mr	r13,r9
1978	RFI_TO_USER	/* return to userspace */
1979	b	.	/* prevent speculative execution */
1980#endif
1981.endm
1982
1983EXC_REAL_BEGIN(system_call, 0xc00, 0x100)
1984	SYSTEM_CALL 0
1985EXC_REAL_END(system_call, 0xc00, 0x100)
1986EXC_VIRT_BEGIN(system_call, 0x4c00, 0x100)
1987	SYSTEM_CALL 1
1988EXC_VIRT_END(system_call, 0x4c00, 0x100)
1989
1990#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1991TRAMP_REAL_BEGIN(kvm_hcall)
1992	std	r9,PACA_EXGEN+EX_R9(r13)
1993	std	r11,PACA_EXGEN+EX_R11(r13)
1994	std	r12,PACA_EXGEN+EX_R12(r13)
1995	mfcr	r9
1996	mfctr	r10
1997	std	r10,PACA_EXGEN+EX_R13(r13)
1998	li	r10,0
1999	std	r10,PACA_EXGEN+EX_CFAR(r13)
2000	std	r10,PACA_EXGEN+EX_CTR(r13)
2001	 /*
2002	  * Save the PPR (on systems that support it) before changing to
2003	  * HMT_MEDIUM. That allows the KVM code to save that value into the
2004	  * guest state (it is the guest's PPR value).
2005	  */
2006BEGIN_FTR_SECTION
2007	mfspr	r10,SPRN_PPR
2008	std	r10,PACA_EXGEN+EX_PPR(r13)
2009END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
2010
2011	HMT_MEDIUM
2012
2013#ifdef CONFIG_RELOCATABLE
2014	/*
2015	 * Requires __LOAD_FAR_HANDLER beause kvmppc_hcall lives
2016	 * outside the head section.
2017	 */
2018	__LOAD_FAR_HANDLER(r10, kvmppc_hcall)
2019	mtctr   r10
2020	bctr
2021#else
2022	b       kvmppc_hcall
2023#endif
2024#endif
2025
2026/**
2027 * Interrupt 0xd00 - Trace Interrupt.
2028 * This is a synchronous interrupt in response to instruction step or
2029 * breakpoint faults.
2030 */
2031INT_DEFINE_BEGIN(single_step)
2032	IVEC=0xd00
2033#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2034	IKVM_REAL=1
2035#endif
2036INT_DEFINE_END(single_step)
2037
2038EXC_REAL_BEGIN(single_step, 0xd00, 0x100)
2039	GEN_INT_ENTRY single_step, virt=0
2040EXC_REAL_END(single_step, 0xd00, 0x100)
2041EXC_VIRT_BEGIN(single_step, 0x4d00, 0x100)
2042	GEN_INT_ENTRY single_step, virt=1
2043EXC_VIRT_END(single_step, 0x4d00, 0x100)
2044EXC_COMMON_BEGIN(single_step_common)
2045	GEN_COMMON single_step
2046	addi	r3,r1,STACK_FRAME_OVERHEAD
2047	bl	single_step_exception
2048	b	interrupt_return_srr
2049
2050
2051/**
2052 * Interrupt 0xe00 - Hypervisor Data Storage Interrupt (HDSI).
2053 * This is a synchronous interrupt in response to an MMU fault caused by a
2054 * guest data access.
2055 *
2056 * Handling:
2057 * This should always get routed to KVM. In radix MMU mode, this is caused
2058 * by a guest nested radix access that can't be performed due to the
2059 * partition scope page table. In hash mode, this can be caused by guests
2060 * running with translation disabled (virtual real mode) or with VPM enabled.
2061 * KVM will update the page table structures or disallow the access.
2062 */
2063INT_DEFINE_BEGIN(h_data_storage)
2064	IVEC=0xe00
2065	IHSRR=1
2066	IDAR=1
2067	IDSISR=1
2068	IKVM_REAL=1
2069	IKVM_VIRT=1
2070INT_DEFINE_END(h_data_storage)
2071
2072EXC_REAL_BEGIN(h_data_storage, 0xe00, 0x20)
2073	GEN_INT_ENTRY h_data_storage, virt=0, ool=1
2074EXC_REAL_END(h_data_storage, 0xe00, 0x20)
2075EXC_VIRT_BEGIN(h_data_storage, 0x4e00, 0x20)
2076	GEN_INT_ENTRY h_data_storage, virt=1, ool=1
2077EXC_VIRT_END(h_data_storage, 0x4e00, 0x20)
2078EXC_COMMON_BEGIN(h_data_storage_common)
2079	GEN_COMMON h_data_storage
2080	addi    r3,r1,STACK_FRAME_OVERHEAD
2081BEGIN_MMU_FTR_SECTION
2082	bl      do_bad_page_fault_segv
2083MMU_FTR_SECTION_ELSE
2084	bl      unknown_exception
2085ALT_MMU_FTR_SECTION_END_IFSET(MMU_FTR_TYPE_RADIX)
2086	b       interrupt_return_hsrr
2087
2088
2089/**
2090 * Interrupt 0xe20 - Hypervisor Instruction Storage Interrupt (HISI).
2091 * This is a synchronous interrupt in response to an MMU fault caused by a
2092 * guest instruction fetch, similar to HDSI.
2093 */
2094INT_DEFINE_BEGIN(h_instr_storage)
2095	IVEC=0xe20
2096	IHSRR=1
2097	IKVM_REAL=1
2098	IKVM_VIRT=1
2099INT_DEFINE_END(h_instr_storage)
2100
2101EXC_REAL_BEGIN(h_instr_storage, 0xe20, 0x20)
2102	GEN_INT_ENTRY h_instr_storage, virt=0, ool=1
2103EXC_REAL_END(h_instr_storage, 0xe20, 0x20)
2104EXC_VIRT_BEGIN(h_instr_storage, 0x4e20, 0x20)
2105	GEN_INT_ENTRY h_instr_storage, virt=1, ool=1
2106EXC_VIRT_END(h_instr_storage, 0x4e20, 0x20)
2107EXC_COMMON_BEGIN(h_instr_storage_common)
2108	GEN_COMMON h_instr_storage
2109	addi	r3,r1,STACK_FRAME_OVERHEAD
2110	bl	unknown_exception
2111	b	interrupt_return_hsrr
2112
2113
2114/**
2115 * Interrupt 0xe40 - Hypervisor Emulation Assistance Interrupt.
2116 */
2117INT_DEFINE_BEGIN(emulation_assist)
2118	IVEC=0xe40
2119	IHSRR=1
2120	IKVM_REAL=1
2121	IKVM_VIRT=1
2122INT_DEFINE_END(emulation_assist)
2123
2124EXC_REAL_BEGIN(emulation_assist, 0xe40, 0x20)
2125	GEN_INT_ENTRY emulation_assist, virt=0, ool=1
2126EXC_REAL_END(emulation_assist, 0xe40, 0x20)
2127EXC_VIRT_BEGIN(emulation_assist, 0x4e40, 0x20)
2128	GEN_INT_ENTRY emulation_assist, virt=1, ool=1
2129EXC_VIRT_END(emulation_assist, 0x4e40, 0x20)
2130EXC_COMMON_BEGIN(emulation_assist_common)
2131	GEN_COMMON emulation_assist
2132	addi	r3,r1,STACK_FRAME_OVERHEAD
2133	bl	emulation_assist_interrupt
2134	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2135	b	interrupt_return_hsrr
2136
2137
2138/**
2139 * Interrupt 0xe60 - Hypervisor Maintenance Interrupt (HMI).
2140 * This is an asynchronous interrupt caused by a Hypervisor Maintenance
2141 * Exception. It is always taken in real mode but uses HSRR registers
2142 * unlike SRESET and MCE.
2143 *
2144 * It is maskable in hardware by clearing MSR[EE], and partially soft-maskable
2145 * with IRQS_DISABLED mask (i.e., local_irq_disable()).
2146 *
2147 * Handling:
2148 * This is a special case, this is handled similarly to machine checks, with an
2149 * initial real mode handler that is not soft-masked, which attempts to fix the
2150 * problem. Then a regular handler which is soft-maskable and reports the
2151 * problem.
2152 *
2153 * The emergency stack is used for the early real mode handler.
2154 *
2155 * XXX: unclear why MCE and HMI schemes could not be made common, e.g.,
2156 * either use soft-masking for the MCE, or use irq_work for the HMI.
2157 *
2158 * KVM:
2159 * Unlike MCE, this calls into KVM without calling the real mode handler
2160 * first.
2161 */
2162INT_DEFINE_BEGIN(hmi_exception_early)
2163	IVEC=0xe60
2164	IHSRR=1
2165	IREALMODE_COMMON=1
2166	ISTACK=0
2167	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
2168	IKVM_REAL=1
2169INT_DEFINE_END(hmi_exception_early)
2170
2171INT_DEFINE_BEGIN(hmi_exception)
2172	IVEC=0xe60
2173	IHSRR=1
2174	IMASK=IRQS_DISABLED
2175	IKVM_REAL=1
2176INT_DEFINE_END(hmi_exception)
2177
2178EXC_REAL_BEGIN(hmi_exception, 0xe60, 0x20)
2179	GEN_INT_ENTRY hmi_exception_early, virt=0, ool=1
2180EXC_REAL_END(hmi_exception, 0xe60, 0x20)
2181EXC_VIRT_NONE(0x4e60, 0x20)
2182
2183EXC_COMMON_BEGIN(hmi_exception_early_common)
2184	__GEN_REALMODE_COMMON_ENTRY hmi_exception_early
2185
2186	mr	r10,r1			/* Save r1 */
2187	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack for realmode */
2188	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
2189
2190	__GEN_COMMON_BODY hmi_exception_early
2191
2192	addi	r3,r1,STACK_FRAME_OVERHEAD
2193	bl	hmi_exception_realmode
2194	cmpdi	cr0,r3,0
2195	bne	1f
2196
2197	EXCEPTION_RESTORE_REGS hsrr=1
2198	HRFI_TO_USER_OR_KERNEL
2199
22001:
2201	/*
2202	 * Go to virtual mode and pull the HMI event information from
2203	 * firmware.
2204	 */
2205	EXCEPTION_RESTORE_REGS hsrr=1
2206	GEN_INT_ENTRY hmi_exception, virt=0
2207
2208EXC_COMMON_BEGIN(hmi_exception_common)
2209	GEN_COMMON hmi_exception
2210	addi	r3,r1,STACK_FRAME_OVERHEAD
2211	bl	handle_hmi_exception
2212	b	interrupt_return_hsrr
2213
2214
2215/**
2216 * Interrupt 0xe80 - Directed Hypervisor Doorbell Interrupt.
2217 * This is an asynchronous interrupt in response to a msgsnd doorbell.
2218 * Similar to the 0xa00 doorbell but for host rather than guest.
2219 */
2220INT_DEFINE_BEGIN(h_doorbell)
2221	IVEC=0xe80
2222	IHSRR=1
2223	IMASK=IRQS_DISABLED
2224	IKVM_REAL=1
2225	IKVM_VIRT=1
2226INT_DEFINE_END(h_doorbell)
2227
2228EXC_REAL_BEGIN(h_doorbell, 0xe80, 0x20)
2229	GEN_INT_ENTRY h_doorbell, virt=0, ool=1
2230EXC_REAL_END(h_doorbell, 0xe80, 0x20)
2231EXC_VIRT_BEGIN(h_doorbell, 0x4e80, 0x20)
2232	GEN_INT_ENTRY h_doorbell, virt=1, ool=1
2233EXC_VIRT_END(h_doorbell, 0x4e80, 0x20)
2234EXC_COMMON_BEGIN(h_doorbell_common)
2235	GEN_COMMON h_doorbell
2236	addi	r3,r1,STACK_FRAME_OVERHEAD
2237#ifdef CONFIG_PPC_DOORBELL
2238	bl	doorbell_exception
2239#else
2240	bl	unknown_async_exception
2241#endif
2242	b	interrupt_return_hsrr
2243
2244
2245/**
2246 * Interrupt 0xea0 - Hypervisor Virtualization Interrupt.
2247 * This is an asynchronous interrupt in response to an "external exception".
2248 * Similar to 0x500 but for host only.
2249 */
2250INT_DEFINE_BEGIN(h_virt_irq)
2251	IVEC=0xea0
2252	IHSRR=1
2253	IMASK=IRQS_DISABLED
2254	IKVM_REAL=1
2255	IKVM_VIRT=1
2256INT_DEFINE_END(h_virt_irq)
2257
2258EXC_REAL_BEGIN(h_virt_irq, 0xea0, 0x20)
2259	GEN_INT_ENTRY h_virt_irq, virt=0, ool=1
2260EXC_REAL_END(h_virt_irq, 0xea0, 0x20)
2261EXC_VIRT_BEGIN(h_virt_irq, 0x4ea0, 0x20)
2262	GEN_INT_ENTRY h_virt_irq, virt=1, ool=1
2263EXC_VIRT_END(h_virt_irq, 0x4ea0, 0x20)
2264EXC_COMMON_BEGIN(h_virt_irq_common)
2265	GEN_COMMON h_virt_irq
2266	addi	r3,r1,STACK_FRAME_OVERHEAD
2267	bl	do_IRQ
2268	b	interrupt_return_hsrr
2269
2270
2271EXC_REAL_NONE(0xec0, 0x20)
2272EXC_VIRT_NONE(0x4ec0, 0x20)
2273EXC_REAL_NONE(0xee0, 0x20)
2274EXC_VIRT_NONE(0x4ee0, 0x20)
2275
2276
2277/*
2278 * Interrupt 0xf00 - Performance Monitor Interrupt (PMI, PMU).
2279 * This is an asynchronous interrupt in response to a PMU exception.
2280 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
2281 * IRQS_PMI_DISABLED mask (NOTE: NOT local_irq_disable()).
2282 *
2283 * Handling:
2284 * This calls into the perf subsystem.
2285 *
2286 * Like the watchdog soft-nmi, it appears an NMI interrupt to Linux, in that it
2287 * runs under local_irq_disable. However it may be soft-masked in
2288 * powerpc-specific code.
2289 *
2290 * If soft masked, the masked handler will note the pending interrupt for
2291 * replay, and clear MSR[EE] in the interrupted context.
2292 */
2293INT_DEFINE_BEGIN(performance_monitor)
2294	IVEC=0xf00
2295	IMASK=IRQS_PMI_DISABLED
2296#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2297	IKVM_REAL=1
2298#endif
2299INT_DEFINE_END(performance_monitor)
2300
2301EXC_REAL_BEGIN(performance_monitor, 0xf00, 0x20)
2302	GEN_INT_ENTRY performance_monitor, virt=0, ool=1
2303EXC_REAL_END(performance_monitor, 0xf00, 0x20)
2304EXC_VIRT_BEGIN(performance_monitor, 0x4f00, 0x20)
2305	GEN_INT_ENTRY performance_monitor, virt=1, ool=1
2306EXC_VIRT_END(performance_monitor, 0x4f00, 0x20)
2307EXC_COMMON_BEGIN(performance_monitor_common)
2308	GEN_COMMON performance_monitor
2309	addi	r3,r1,STACK_FRAME_OVERHEAD
2310	bl	performance_monitor_exception
2311	b	interrupt_return_srr
2312
2313
2314/**
2315 * Interrupt 0xf20 - Vector Unavailable Interrupt.
2316 * This is a synchronous interrupt in response to
2317 * executing a vector (or altivec) instruction with MSR[VEC]=0.
2318 * Similar to FP unavailable.
2319 */
2320INT_DEFINE_BEGIN(altivec_unavailable)
2321	IVEC=0xf20
2322#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2323	IKVM_REAL=1
2324#endif
2325INT_DEFINE_END(altivec_unavailable)
2326
2327EXC_REAL_BEGIN(altivec_unavailable, 0xf20, 0x20)
2328	GEN_INT_ENTRY altivec_unavailable, virt=0, ool=1
2329EXC_REAL_END(altivec_unavailable, 0xf20, 0x20)
2330EXC_VIRT_BEGIN(altivec_unavailable, 0x4f20, 0x20)
2331	GEN_INT_ENTRY altivec_unavailable, virt=1, ool=1
2332EXC_VIRT_END(altivec_unavailable, 0x4f20, 0x20)
2333EXC_COMMON_BEGIN(altivec_unavailable_common)
2334	GEN_COMMON altivec_unavailable
2335#ifdef CONFIG_ALTIVEC
2336BEGIN_FTR_SECTION
2337	beq	1f
2338#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2339  BEGIN_FTR_SECTION_NESTED(69)
2340	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2341	 * transaction), go do TM stuff
2342	 */
2343	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2344	bne-	2f
2345  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2346#endif
2347	bl	load_up_altivec
2348	b	fast_interrupt_return_srr
2349#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
23502:	/* User process was in a transaction */
2351	addi	r3,r1,STACK_FRAME_OVERHEAD
2352	bl	altivec_unavailable_tm
2353	b	interrupt_return_srr
2354#endif
23551:
2356END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2357#endif
2358	addi	r3,r1,STACK_FRAME_OVERHEAD
2359	bl	altivec_unavailable_exception
2360	b	interrupt_return_srr
2361
2362
2363/**
2364 * Interrupt 0xf40 - VSX Unavailable Interrupt.
2365 * This is a synchronous interrupt in response to
2366 * executing a VSX instruction with MSR[VSX]=0.
2367 * Similar to FP unavailable.
2368 */
2369INT_DEFINE_BEGIN(vsx_unavailable)
2370	IVEC=0xf40
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	addi	r3,r1,STACK_FRAME_OVERHEAD
2400	bl	vsx_unavailable_tm
2401	b	interrupt_return_srr
2402#endif
24031:
2404END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2405#endif
2406	addi	r3,r1,STACK_FRAME_OVERHEAD
2407	bl	vsx_unavailable_exception
2408	b	interrupt_return_srr
2409
2410
2411/**
2412 * Interrupt 0xf60 - Facility Unavailable Interrupt.
2413 * This is a synchronous interrupt in response to
2414 * executing an instruction without access to the facility that can be
2415 * resolved by the OS (e.g., FSCR, MSR).
2416 * Similar to FP unavailable.
2417 */
2418INT_DEFINE_BEGIN(facility_unavailable)
2419	IVEC=0xf60
2420#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2421	IKVM_REAL=1
2422#endif
2423INT_DEFINE_END(facility_unavailable)
2424
2425EXC_REAL_BEGIN(facility_unavailable, 0xf60, 0x20)
2426	GEN_INT_ENTRY facility_unavailable, virt=0, ool=1
2427EXC_REAL_END(facility_unavailable, 0xf60, 0x20)
2428EXC_VIRT_BEGIN(facility_unavailable, 0x4f60, 0x20)
2429	GEN_INT_ENTRY facility_unavailable, virt=1, ool=1
2430EXC_VIRT_END(facility_unavailable, 0x4f60, 0x20)
2431EXC_COMMON_BEGIN(facility_unavailable_common)
2432	GEN_COMMON facility_unavailable
2433	addi	r3,r1,STACK_FRAME_OVERHEAD
2434	bl	facility_unavailable_exception
2435	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2436	b	interrupt_return_srr
2437
2438
2439/**
2440 * Interrupt 0xf60 - Hypervisor Facility Unavailable Interrupt.
2441 * This is a synchronous interrupt in response to
2442 * executing an instruction without access to the facility that can only
2443 * be resolved in HV mode (e.g., HFSCR).
2444 * Similar to FP unavailable.
2445 */
2446INT_DEFINE_BEGIN(h_facility_unavailable)
2447	IVEC=0xf80
2448	IHSRR=1
2449	IKVM_REAL=1
2450	IKVM_VIRT=1
2451INT_DEFINE_END(h_facility_unavailable)
2452
2453EXC_REAL_BEGIN(h_facility_unavailable, 0xf80, 0x20)
2454	GEN_INT_ENTRY h_facility_unavailable, virt=0, ool=1
2455EXC_REAL_END(h_facility_unavailable, 0xf80, 0x20)
2456EXC_VIRT_BEGIN(h_facility_unavailable, 0x4f80, 0x20)
2457	GEN_INT_ENTRY h_facility_unavailable, virt=1, ool=1
2458EXC_VIRT_END(h_facility_unavailable, 0x4f80, 0x20)
2459EXC_COMMON_BEGIN(h_facility_unavailable_common)
2460	GEN_COMMON h_facility_unavailable
2461	addi	r3,r1,STACK_FRAME_OVERHEAD
2462	bl	facility_unavailable_exception
2463	REST_NVGPRS(r1) /* XXX Shouldn't be necessary in practice */
2464	b	interrupt_return_hsrr
2465
2466
2467EXC_REAL_NONE(0xfa0, 0x20)
2468EXC_VIRT_NONE(0x4fa0, 0x20)
2469EXC_REAL_NONE(0xfc0, 0x20)
2470EXC_VIRT_NONE(0x4fc0, 0x20)
2471EXC_REAL_NONE(0xfe0, 0x20)
2472EXC_VIRT_NONE(0x4fe0, 0x20)
2473
2474EXC_REAL_NONE(0x1000, 0x100)
2475EXC_VIRT_NONE(0x5000, 0x100)
2476EXC_REAL_NONE(0x1100, 0x100)
2477EXC_VIRT_NONE(0x5100, 0x100)
2478
2479#ifdef CONFIG_CBE_RAS
2480INT_DEFINE_BEGIN(cbe_system_error)
2481	IVEC=0x1200
2482	IHSRR=1
2483INT_DEFINE_END(cbe_system_error)
2484
2485EXC_REAL_BEGIN(cbe_system_error, 0x1200, 0x100)
2486	GEN_INT_ENTRY cbe_system_error, virt=0
2487EXC_REAL_END(cbe_system_error, 0x1200, 0x100)
2488EXC_VIRT_NONE(0x5200, 0x100)
2489EXC_COMMON_BEGIN(cbe_system_error_common)
2490	GEN_COMMON cbe_system_error
2491	addi	r3,r1,STACK_FRAME_OVERHEAD
2492	bl	cbe_system_error_exception
2493	b	interrupt_return_hsrr
2494
2495#else /* CONFIG_CBE_RAS */
2496EXC_REAL_NONE(0x1200, 0x100)
2497EXC_VIRT_NONE(0x5200, 0x100)
2498#endif
2499
2500/**
2501 * Interrupt 0x1300 - Instruction Address Breakpoint Interrupt.
2502 * This has been removed from the ISA before 2.01, which is the earliest
2503 * 64-bit BookS ISA supported, however the G5 / 970 implements this
2504 * interrupt with a non-architected feature available through the support
2505 * processor interface.
2506 */
2507INT_DEFINE_BEGIN(instruction_breakpoint)
2508	IVEC=0x1300
2509#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2510	IKVM_REAL=1
2511#endif
2512INT_DEFINE_END(instruction_breakpoint)
2513
2514EXC_REAL_BEGIN(instruction_breakpoint, 0x1300, 0x100)
2515	GEN_INT_ENTRY instruction_breakpoint, virt=0
2516EXC_REAL_END(instruction_breakpoint, 0x1300, 0x100)
2517EXC_VIRT_BEGIN(instruction_breakpoint, 0x5300, 0x100)
2518	GEN_INT_ENTRY instruction_breakpoint, virt=1
2519EXC_VIRT_END(instruction_breakpoint, 0x5300, 0x100)
2520EXC_COMMON_BEGIN(instruction_breakpoint_common)
2521	GEN_COMMON instruction_breakpoint
2522	addi	r3,r1,STACK_FRAME_OVERHEAD
2523	bl	instruction_breakpoint_exception
2524	b	interrupt_return_srr
2525
2526
2527EXC_REAL_NONE(0x1400, 0x100)
2528EXC_VIRT_NONE(0x5400, 0x100)
2529
2530/**
2531 * Interrupt 0x1500 - Soft Patch Interrupt
2532 *
2533 * Handling:
2534 * This is an implementation specific interrupt which can be used for a
2535 * range of exceptions.
2536 *
2537 * This interrupt handler is unique in that it runs the denormal assist
2538 * code even for guests (and even in guest context) without going to KVM,
2539 * for speed. POWER9 does not raise denorm exceptions, so this special case
2540 * could be phased out in future to reduce special cases.
2541 */
2542INT_DEFINE_BEGIN(denorm_exception)
2543	IVEC=0x1500
2544	IHSRR=1
2545	IBRANCH_TO_COMMON=0
2546	IKVM_REAL=1
2547INT_DEFINE_END(denorm_exception)
2548
2549EXC_REAL_BEGIN(denorm_exception, 0x1500, 0x100)
2550	GEN_INT_ENTRY denorm_exception, virt=0
2551#ifdef CONFIG_PPC_DENORMALISATION
2552	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2553	bne+	denorm_assist
2554#endif
2555	GEN_BRANCH_TO_COMMON denorm_exception, virt=0
2556EXC_REAL_END(denorm_exception, 0x1500, 0x100)
2557#ifdef CONFIG_PPC_DENORMALISATION
2558EXC_VIRT_BEGIN(denorm_exception, 0x5500, 0x100)
2559	GEN_INT_ENTRY denorm_exception, virt=1
2560	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2561	bne+	denorm_assist
2562	GEN_BRANCH_TO_COMMON denorm_exception, virt=1
2563EXC_VIRT_END(denorm_exception, 0x5500, 0x100)
2564#else
2565EXC_VIRT_NONE(0x5500, 0x100)
2566#endif
2567
2568#ifdef CONFIG_PPC_DENORMALISATION
2569TRAMP_REAL_BEGIN(denorm_assist)
2570BEGIN_FTR_SECTION
2571/*
2572 * To denormalise we need to move a copy of the register to itself.
2573 * For POWER6 do that here for all FP regs.
2574 */
2575	mfmsr	r10
2576	ori	r10,r10,(MSR_FP|MSR_FE0|MSR_FE1)
2577	xori	r10,r10,(MSR_FE0|MSR_FE1)
2578	mtmsrd	r10
2579	sync
2580
2581	.Lreg=0
2582	.rept 32
2583	fmr	.Lreg,.Lreg
2584	.Lreg=.Lreg+1
2585	.endr
2586
2587FTR_SECTION_ELSE
2588/*
2589 * To denormalise we need to move a copy of the register to itself.
2590 * For POWER7 do that here for the first 32 VSX registers only.
2591 */
2592	mfmsr	r10
2593	oris	r10,r10,MSR_VSX@h
2594	mtmsrd	r10
2595	sync
2596
2597	.Lreg=0
2598	.rept 32
2599	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2600	.Lreg=.Lreg+1
2601	.endr
2602
2603ALT_FTR_SECTION_END_IFCLR(CPU_FTR_ARCH_206)
2604
2605BEGIN_FTR_SECTION
2606	b	denorm_done
2607END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
2608/*
2609 * To denormalise we need to move a copy of the register to itself.
2610 * For POWER8 we need to do that for all 64 VSX registers
2611 */
2612	.Lreg=32
2613	.rept 32
2614	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2615	.Lreg=.Lreg+1
2616	.endr
2617
2618denorm_done:
2619	mfspr	r11,SPRN_HSRR0
2620	subi	r11,r11,4
2621	mtspr	SPRN_HSRR0,r11
2622	mtcrf	0x80,r9
2623	ld	r9,PACA_EXGEN+EX_R9(r13)
2624BEGIN_FTR_SECTION
2625	ld	r10,PACA_EXGEN+EX_PPR(r13)
2626	mtspr	SPRN_PPR,r10
2627END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
2628BEGIN_FTR_SECTION
2629	ld	r10,PACA_EXGEN+EX_CFAR(r13)
2630	mtspr	SPRN_CFAR,r10
2631END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
2632	li	r10,0
2633	stb	r10,PACAHSRR_VALID(r13)
2634	ld	r10,PACA_EXGEN+EX_R10(r13)
2635	ld	r11,PACA_EXGEN+EX_R11(r13)
2636	ld	r12,PACA_EXGEN+EX_R12(r13)
2637	ld	r13,PACA_EXGEN+EX_R13(r13)
2638	HRFI_TO_UNKNOWN
2639	b	.
2640#endif
2641
2642EXC_COMMON_BEGIN(denorm_exception_common)
2643	GEN_COMMON denorm_exception
2644	addi	r3,r1,STACK_FRAME_OVERHEAD
2645	bl	unknown_exception
2646	b	interrupt_return_hsrr
2647
2648
2649#ifdef CONFIG_CBE_RAS
2650INT_DEFINE_BEGIN(cbe_maintenance)
2651	IVEC=0x1600
2652	IHSRR=1
2653INT_DEFINE_END(cbe_maintenance)
2654
2655EXC_REAL_BEGIN(cbe_maintenance, 0x1600, 0x100)
2656	GEN_INT_ENTRY cbe_maintenance, virt=0
2657EXC_REAL_END(cbe_maintenance, 0x1600, 0x100)
2658EXC_VIRT_NONE(0x5600, 0x100)
2659EXC_COMMON_BEGIN(cbe_maintenance_common)
2660	GEN_COMMON cbe_maintenance
2661	addi	r3,r1,STACK_FRAME_OVERHEAD
2662	bl	cbe_maintenance_exception
2663	b	interrupt_return_hsrr
2664
2665#else /* CONFIG_CBE_RAS */
2666EXC_REAL_NONE(0x1600, 0x100)
2667EXC_VIRT_NONE(0x5600, 0x100)
2668#endif
2669
2670
2671INT_DEFINE_BEGIN(altivec_assist)
2672	IVEC=0x1700
2673#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2674	IKVM_REAL=1
2675#endif
2676INT_DEFINE_END(altivec_assist)
2677
2678EXC_REAL_BEGIN(altivec_assist, 0x1700, 0x100)
2679	GEN_INT_ENTRY altivec_assist, virt=0
2680EXC_REAL_END(altivec_assist, 0x1700, 0x100)
2681EXC_VIRT_BEGIN(altivec_assist, 0x5700, 0x100)
2682	GEN_INT_ENTRY altivec_assist, virt=1
2683EXC_VIRT_END(altivec_assist, 0x5700, 0x100)
2684EXC_COMMON_BEGIN(altivec_assist_common)
2685	GEN_COMMON altivec_assist
2686	addi	r3,r1,STACK_FRAME_OVERHEAD
2687#ifdef CONFIG_ALTIVEC
2688	bl	altivec_assist_exception
2689	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2690#else
2691	bl	unknown_exception
2692#endif
2693	b	interrupt_return_srr
2694
2695
2696#ifdef CONFIG_CBE_RAS
2697INT_DEFINE_BEGIN(cbe_thermal)
2698	IVEC=0x1800
2699	IHSRR=1
2700INT_DEFINE_END(cbe_thermal)
2701
2702EXC_REAL_BEGIN(cbe_thermal, 0x1800, 0x100)
2703	GEN_INT_ENTRY cbe_thermal, virt=0
2704EXC_REAL_END(cbe_thermal, 0x1800, 0x100)
2705EXC_VIRT_NONE(0x5800, 0x100)
2706EXC_COMMON_BEGIN(cbe_thermal_common)
2707	GEN_COMMON cbe_thermal
2708	addi	r3,r1,STACK_FRAME_OVERHEAD
2709	bl	cbe_thermal_exception
2710	b	interrupt_return_hsrr
2711
2712#else /* CONFIG_CBE_RAS */
2713EXC_REAL_NONE(0x1800, 0x100)
2714EXC_VIRT_NONE(0x5800, 0x100)
2715#endif
2716
2717
2718#ifdef CONFIG_PPC_WATCHDOG
2719
2720INT_DEFINE_BEGIN(soft_nmi)
2721	IVEC=0x900
2722	ISTACK=0
2723INT_DEFINE_END(soft_nmi)
2724
2725/*
2726 * Branch to soft_nmi_interrupt using the emergency stack. The emergency
2727 * stack is one that is usable by maskable interrupts so long as MSR_EE
2728 * remains off. It is used for recovery when something has corrupted the
2729 * normal kernel stack, for example. The "soft NMI" must not use the process
2730 * stack because we want irq disabled sections to avoid touching the stack
2731 * at all (other than PMU interrupts), so use the emergency stack for this,
2732 * and run it entirely with interrupts hard disabled.
2733 */
2734EXC_COMMON_BEGIN(soft_nmi_common)
2735	mr	r10,r1
2736	ld	r1,PACAEMERGSP(r13)
2737	subi	r1,r1,INT_FRAME_SIZE
2738	__GEN_COMMON_BODY soft_nmi
2739
2740	addi	r3,r1,STACK_FRAME_OVERHEAD
2741	bl	soft_nmi_interrupt
2742
2743	/* Clear MSR_RI before setting SRR0 and SRR1. */
2744	li	r9,0
2745	mtmsrd	r9,1
2746
2747	kuap_kernel_restore r9, r10
2748
2749	EXCEPTION_RESTORE_REGS hsrr=0
2750	RFI_TO_KERNEL
2751
2752#endif /* CONFIG_PPC_WATCHDOG */
2753
2754/*
2755 * An interrupt came in while soft-disabled. We set paca->irq_happened, then:
2756 * - If it was a decrementer interrupt, we bump the dec to max and and return.
2757 * - If it was a doorbell we return immediately since doorbells are edge
2758 *   triggered and won't automatically refire.
2759 * - If it was a HMI we return immediately since we handled it in realmode
2760 *   and it won't refire.
2761 * - Else it is one of PACA_IRQ_MUST_HARD_MASK, so hard disable and return.
2762 * This is called with r10 containing the value to OR to the paca field.
2763 */
2764.macro MASKED_INTERRUPT hsrr=0
2765	.if \hsrr
2766masked_Hinterrupt:
2767	.else
2768masked_interrupt:
2769	.endif
2770	stw	r9,PACA_EXGEN+EX_CCR(r13)
2771	lbz	r9,PACAIRQHAPPENED(r13)
2772	or	r9,r9,r10
2773	stb	r9,PACAIRQHAPPENED(r13)
2774
2775	.if ! \hsrr
2776	cmpwi	r10,PACA_IRQ_DEC
2777	bne	1f
2778	LOAD_REG_IMMEDIATE(r9, 0x7fffffff)
2779	mtspr	SPRN_DEC,r9
2780#ifdef CONFIG_PPC_WATCHDOG
2781	lwz	r9,PACA_EXGEN+EX_CCR(r13)
2782	b	soft_nmi_common
2783#else
2784	b	2f
2785#endif
2786	.endif
2787
27881:	andi.	r10,r10,PACA_IRQ_MUST_HARD_MASK
2789	beq	2f
2790	xori	r12,r12,MSR_EE	/* clear MSR_EE */
2791	.if \hsrr
2792	mtspr	SPRN_HSRR1,r12
2793	.else
2794	mtspr	SPRN_SRR1,r12
2795	.endif
2796	ori	r9,r9,PACA_IRQ_HARD_DIS
2797	stb	r9,PACAIRQHAPPENED(r13)
27982:	/* done */
2799	li	r9,0
2800	.if \hsrr
2801	stb	r9,PACAHSRR_VALID(r13)
2802	.else
2803	stb	r9,PACASRR_VALID(r13)
2804	.endif
2805
2806	SEARCH_RESTART_TABLE
2807	cmpdi	r12,0
2808	beq	3f
2809	.if \hsrr
2810	mtspr	SPRN_HSRR0,r12
2811	.else
2812	mtspr	SPRN_SRR0,r12
2813	.endif
28143:
2815
2816	ld	r9,PACA_EXGEN+EX_CTR(r13)
2817	mtctr	r9
2818	lwz	r9,PACA_EXGEN+EX_CCR(r13)
2819	mtcrf	0x80,r9
2820	std	r1,PACAR1(r13)
2821	ld	r9,PACA_EXGEN+EX_R9(r13)
2822	ld	r10,PACA_EXGEN+EX_R10(r13)
2823	ld	r11,PACA_EXGEN+EX_R11(r13)
2824	ld	r12,PACA_EXGEN+EX_R12(r13)
2825	ld	r13,PACA_EXGEN+EX_R13(r13)
2826	/* May return to masked low address where r13 is not set up */
2827	.if \hsrr
2828	HRFI_TO_KERNEL
2829	.else
2830	RFI_TO_KERNEL
2831	.endif
2832	b	.
2833.endm
2834
2835TRAMP_REAL_BEGIN(stf_barrier_fallback)
2836	std	r9,PACA_EXRFI+EX_R9(r13)
2837	std	r10,PACA_EXRFI+EX_R10(r13)
2838	sync
2839	ld	r9,PACA_EXRFI+EX_R9(r13)
2840	ld	r10,PACA_EXRFI+EX_R10(r13)
2841	ori	31,31,0
2842	.rept 14
2843	b	1f
28441:
2845	.endr
2846	blr
2847
2848/* Clobbers r10, r11, ctr */
2849.macro L1D_DISPLACEMENT_FLUSH
2850	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
2851	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
2852	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
2853	mtctr	r11
2854	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
2855
2856	/* order ld/st prior to dcbt stop all streams with flushing */
2857	sync
2858
2859	/*
2860	 * The load addresses are at staggered offsets within cachelines,
2861	 * which suits some pipelines better (on others it should not
2862	 * hurt).
2863	 */
28641:
2865	ld	r11,(0x80 + 8)*0(r10)
2866	ld	r11,(0x80 + 8)*1(r10)
2867	ld	r11,(0x80 + 8)*2(r10)
2868	ld	r11,(0x80 + 8)*3(r10)
2869	ld	r11,(0x80 + 8)*4(r10)
2870	ld	r11,(0x80 + 8)*5(r10)
2871	ld	r11,(0x80 + 8)*6(r10)
2872	ld	r11,(0x80 + 8)*7(r10)
2873	addi	r10,r10,0x80*8
2874	bdnz	1b
2875.endm
2876
2877TRAMP_REAL_BEGIN(entry_flush_fallback)
2878	std	r9,PACA_EXRFI+EX_R9(r13)
2879	std	r10,PACA_EXRFI+EX_R10(r13)
2880	std	r11,PACA_EXRFI+EX_R11(r13)
2881	mfctr	r9
2882	L1D_DISPLACEMENT_FLUSH
2883	mtctr	r9
2884	ld	r9,PACA_EXRFI+EX_R9(r13)
2885	ld	r10,PACA_EXRFI+EX_R10(r13)
2886	ld	r11,PACA_EXRFI+EX_R11(r13)
2887	blr
2888
2889/*
2890 * The SCV entry flush happens with interrupts enabled, so it must disable
2891 * to prevent EXRFI being clobbered by NMIs (e.g., soft_nmi_common). r10
2892 * (containing LR) does not need to be preserved here because scv entry
2893 * puts 0 in the pt_regs, CTR can be clobbered for the same reason.
2894 */
2895TRAMP_REAL_BEGIN(scv_entry_flush_fallback)
2896	li	r10,0
2897	mtmsrd	r10,1
2898	lbz	r10,PACAIRQHAPPENED(r13)
2899	ori	r10,r10,PACA_IRQ_HARD_DIS
2900	stb	r10,PACAIRQHAPPENED(r13)
2901	std	r11,PACA_EXRFI+EX_R11(r13)
2902	L1D_DISPLACEMENT_FLUSH
2903	ld	r11,PACA_EXRFI+EX_R11(r13)
2904	li	r10,MSR_RI
2905	mtmsrd	r10,1
2906	blr
2907
2908TRAMP_REAL_BEGIN(rfi_flush_fallback)
2909	SET_SCRATCH0(r13);
2910	GET_PACA(r13);
2911	std	r1,PACA_EXRFI+EX_R12(r13)
2912	ld	r1,PACAKSAVE(r13)
2913	std	r9,PACA_EXRFI+EX_R9(r13)
2914	std	r10,PACA_EXRFI+EX_R10(r13)
2915	std	r11,PACA_EXRFI+EX_R11(r13)
2916	mfctr	r9
2917	L1D_DISPLACEMENT_FLUSH
2918	mtctr	r9
2919	ld	r9,PACA_EXRFI+EX_R9(r13)
2920	ld	r10,PACA_EXRFI+EX_R10(r13)
2921	ld	r11,PACA_EXRFI+EX_R11(r13)
2922	ld	r1,PACA_EXRFI+EX_R12(r13)
2923	GET_SCRATCH0(r13);
2924	rfid
2925
2926TRAMP_REAL_BEGIN(hrfi_flush_fallback)
2927	SET_SCRATCH0(r13);
2928	GET_PACA(r13);
2929	std	r1,PACA_EXRFI+EX_R12(r13)
2930	ld	r1,PACAKSAVE(r13)
2931	std	r9,PACA_EXRFI+EX_R9(r13)
2932	std	r10,PACA_EXRFI+EX_R10(r13)
2933	std	r11,PACA_EXRFI+EX_R11(r13)
2934	mfctr	r9
2935	L1D_DISPLACEMENT_FLUSH
2936	mtctr	r9
2937	ld	r9,PACA_EXRFI+EX_R9(r13)
2938	ld	r10,PACA_EXRFI+EX_R10(r13)
2939	ld	r11,PACA_EXRFI+EX_R11(r13)
2940	ld	r1,PACA_EXRFI+EX_R12(r13)
2941	GET_SCRATCH0(r13);
2942	hrfid
2943
2944TRAMP_REAL_BEGIN(rfscv_flush_fallback)
2945	/* system call volatile */
2946	mr	r7,r13
2947	GET_PACA(r13);
2948	mr	r8,r1
2949	ld	r1,PACAKSAVE(r13)
2950	mfctr	r9
2951	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
2952	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
2953	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
2954	mtctr	r11
2955	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
2956
2957	/* order ld/st prior to dcbt stop all streams with flushing */
2958	sync
2959
2960	/*
2961	 * The load adresses are at staggered offsets within cachelines,
2962	 * which suits some pipelines better (on others it should not
2963	 * hurt).
2964	 */
29651:
2966	ld	r11,(0x80 + 8)*0(r10)
2967	ld	r11,(0x80 + 8)*1(r10)
2968	ld	r11,(0x80 + 8)*2(r10)
2969	ld	r11,(0x80 + 8)*3(r10)
2970	ld	r11,(0x80 + 8)*4(r10)
2971	ld	r11,(0x80 + 8)*5(r10)
2972	ld	r11,(0x80 + 8)*6(r10)
2973	ld	r11,(0x80 + 8)*7(r10)
2974	addi	r10,r10,0x80*8
2975	bdnz	1b
2976
2977	mtctr	r9
2978	li	r9,0
2979	li	r10,0
2980	li	r11,0
2981	mr	r1,r8
2982	mr	r13,r7
2983	RFSCV
2984
2985USE_TEXT_SECTION()
2986
2987#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
2988kvm_interrupt:
2989	/*
2990	 * The conditional branch in KVMTEST can't reach all the way,
2991	 * make a stub.
2992	 */
2993	b	kvmppc_interrupt
2994#endif
2995
2996_GLOBAL(do_uaccess_flush)
2997	UACCESS_FLUSH_FIXUP_SECTION
2998	nop
2999	nop
3000	nop
3001	blr
3002	L1D_DISPLACEMENT_FLUSH
3003	blr
3004_ASM_NOKPROBE_SYMBOL(do_uaccess_flush)
3005EXPORT_SYMBOL(do_uaccess_flush)
3006
3007
3008MASKED_INTERRUPT
3009MASKED_INTERRUPT hsrr=1
3010
3011	/*
3012	 * Relocation-on interrupts: A subset of the interrupts can be delivered
3013	 * with IR=1/DR=1, if AIL==2 and MSR.HV won't be changed by delivering
3014	 * it.  Addresses are the same as the original interrupt addresses, but
3015	 * offset by 0xc000000000004000.
3016	 * It's impossible to receive interrupts below 0x300 via this mechanism.
3017	 * KVM: None of these traps are from the guest ; anything that escalated
3018	 * to HV=1 from HV=0 is delivered via real mode handlers.
3019	 */
3020
3021	/*
3022	 * This uses the standard macro, since the original 0x300 vector
3023	 * only has extra guff for STAB-based processors -- which never
3024	 * come here.
3025	 */
3026
3027USE_FIXED_SECTION(virt_trampolines)
3028	/*
3029	 * All code below __end_soft_masked is treated as soft-masked. If
3030	 * any code runs here with MSR[EE]=1, it must then cope with pending
3031	 * soft interrupt being raised (i.e., by ensuring it is replayed).
3032	 *
3033	 * The __end_interrupts marker must be past the out-of-line (OOL)
3034	 * handlers, so that they are copied to real address 0x100 when running
3035	 * a relocatable kernel. This ensures they can be reached from the short
3036	 * trampoline handlers (like 0x4f00, 0x4f20, etc.) which branch
3037	 * directly, without using LOAD_HANDLER().
3038	 */
3039	.align	7
3040	.globl	__end_interrupts
3041__end_interrupts:
3042DEFINE_FIXED_SYMBOL(__end_interrupts)
3043
3044CLOSE_FIXED_SECTION(real_vectors);
3045CLOSE_FIXED_SECTION(real_trampolines);
3046CLOSE_FIXED_SECTION(virt_vectors);
3047CLOSE_FIXED_SECTION(virt_trampolines);
3048
3049USE_TEXT_SECTION()
3050
3051/* MSR[RI] should be clear because this uses SRR[01] */
3052enable_machine_check:
3053	mflr	r0
3054	bcl	20,31,$+4
30550:	mflr	r3
3056	addi	r3,r3,(1f - 0b)
3057	mtspr	SPRN_SRR0,r3
3058	mfmsr	r3
3059	ori	r3,r3,MSR_ME
3060	mtspr	SPRN_SRR1,r3
3061	RFI_TO_KERNEL
30621:	mtlr	r0
3063	blr
3064
3065/* MSR[RI] should be clear because this uses SRR[01] */
3066disable_machine_check:
3067	mflr	r0
3068	bcl	20,31,$+4
30690:	mflr	r3
3070	addi	r3,r3,(1f - 0b)
3071	mtspr	SPRN_SRR0,r3
3072	mfmsr	r3
3073	li	r4,MSR_ME
3074	andc	r3,r3,r4
3075	mtspr	SPRN_SRR1,r3
3076	RFI_TO_KERNEL
30771:	mtlr	r0
3078	blr
3079