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_4GPRS(3, r1)		/* save r3 - r6 in stackframe   */
578	SAVE_2GPRS(7, r1)		/* save r7, r8 in stackframe	*/
579	mflr	r9			/* Get LR, later save to stack	*/
580	ld	r2,PACATOC(r13)		/* get kernel TOC into r2	*/
581	std	r9,_LINK(r1)
582	lbz	r10,PACAIRQSOFTMASK(r13)
583	mfspr	r11,SPRN_XER		/* save XER in stackframe	*/
584	std	r10,SOFTE(r1)
585	std	r11,_XER(r1)
586	li	r9,IVEC
587	std	r9,_TRAP(r1)		/* set trap number		*/
588	li	r10,0
589	ld	r11,exception_marker@toc(r2)
590	std	r10,RESULT(r1)		/* clear regs->result		*/
591	std	r11,STACK_FRAME_OVERHEAD-16(r1) /* mark the frame	*/
592.endm
593
594/*
595 * On entry r13 points to the paca, r9-r13 are saved in the paca,
596 * r9 contains the saved CR, r11 and r12 contain the saved SRR0 and
597 * SRR1, and relocation is on.
598 *
599 * If stack=0, then the stack is already set in r1, and r1 is saved in r10.
600 * PPR save and CPU accounting is not done for the !stack case (XXX why not?)
601 */
602.macro GEN_COMMON name
603	__GEN_COMMON_ENTRY \name
604	__GEN_COMMON_BODY \name
605.endm
606
607.macro SEARCH_RESTART_TABLE
608#ifdef CONFIG_RELOCATABLE
609	mr	r12,r2
610	ld	r2,PACATOC(r13)
611	LOAD_REG_ADDR(r9, __start___restart_table)
612	LOAD_REG_ADDR(r10, __stop___restart_table)
613	mr	r2,r12
614#else
615	LOAD_REG_IMMEDIATE_SYM(r9, r12, __start___restart_table)
616	LOAD_REG_IMMEDIATE_SYM(r10, r12, __stop___restart_table)
617#endif
618300:
619	cmpd	r9,r10
620	beq	302f
621	ld	r12,0(r9)
622	cmpld	r11,r12
623	blt	301f
624	ld	r12,8(r9)
625	cmpld	r11,r12
626	bge	301f
627	ld	r12,16(r9)
628	b	303f
629301:
630	addi	r9,r9,24
631	b	300b
632302:
633	li	r12,0
634303:
635.endm
636
637.macro SEARCH_SOFT_MASK_TABLE
638#ifdef CONFIG_RELOCATABLE
639	mr	r12,r2
640	ld	r2,PACATOC(r13)
641	LOAD_REG_ADDR(r9, __start___soft_mask_table)
642	LOAD_REG_ADDR(r10, __stop___soft_mask_table)
643	mr	r2,r12
644#else
645	LOAD_REG_IMMEDIATE_SYM(r9, r12, __start___soft_mask_table)
646	LOAD_REG_IMMEDIATE_SYM(r10, r12, __stop___soft_mask_table)
647#endif
648300:
649	cmpd	r9,r10
650	beq	302f
651	ld	r12,0(r9)
652	cmpld	r11,r12
653	blt	301f
654	ld	r12,8(r9)
655	cmpld	r11,r12
656	bge	301f
657	li	r12,1
658	b	303f
659301:
660	addi	r9,r9,16
661	b	300b
662302:
663	li	r12,0
664303:
665.endm
666
667/*
668 * Restore all registers including H/SRR0/1 saved in a stack frame of a
669 * standard exception.
670 */
671.macro EXCEPTION_RESTORE_REGS hsrr=0
672	/* Move original SRR0 and SRR1 into the respective regs */
673	ld	r9,_MSR(r1)
674	li	r10,0
675	.if \hsrr
676	mtspr	SPRN_HSRR1,r9
677	stb	r10,PACAHSRR_VALID(r13)
678	.else
679	mtspr	SPRN_SRR1,r9
680	stb	r10,PACASRR_VALID(r13)
681	.endif
682	ld	r9,_NIP(r1)
683	.if \hsrr
684	mtspr	SPRN_HSRR0,r9
685	.else
686	mtspr	SPRN_SRR0,r9
687	.endif
688	ld	r9,_CTR(r1)
689	mtctr	r9
690	ld	r9,_XER(r1)
691	mtxer	r9
692	ld	r9,_LINK(r1)
693	mtlr	r9
694	ld	r9,_CCR(r1)
695	mtcr	r9
696	REST_8GPRS(2, r1)
697	REST_4GPRS(10, r1)
698	REST_GPR(0, r1)
699	/* restore original r1. */
700	ld	r1,GPR1(r1)
701.endm
702
703/*
704 * There are a few constraints to be concerned with.
705 * - Real mode exceptions code/data must be located at their physical location.
706 * - Virtual mode exceptions must be mapped at their 0xc000... location.
707 * - Fixed location code must not call directly beyond the __end_interrupts
708 *   area when built with CONFIG_RELOCATABLE. LOAD_HANDLER / bctr sequence
709 *   must be used.
710 * - LOAD_HANDLER targets must be within first 64K of physical 0 /
711 *   virtual 0xc00...
712 * - Conditional branch targets must be within +/-32K of caller.
713 *
714 * "Virtual exceptions" run with relocation on (MSR_IR=1, MSR_DR=1), and
715 * therefore don't have to run in physically located code or rfid to
716 * virtual mode kernel code. However on relocatable kernels they do have
717 * to branch to KERNELBASE offset because the rest of the kernel (outside
718 * the exception vectors) may be located elsewhere.
719 *
720 * Virtual exceptions correspond with physical, except their entry points
721 * are offset by 0xc000000000000000 and also tend to get an added 0x4000
722 * offset applied. Virtual exceptions are enabled with the Alternate
723 * Interrupt Location (AIL) bit set in the LPCR. However this does not
724 * guarantee they will be delivered virtually. Some conditions (see the ISA)
725 * cause exceptions to be delivered in real mode.
726 *
727 * The scv instructions are a special case. They get a 0x3000 offset applied.
728 * scv exceptions have unique reentrancy properties, see below.
729 *
730 * It's impossible to receive interrupts below 0x300 via AIL.
731 *
732 * KVM: None of the virtual exceptions are from the guest. Anything that
733 * escalated to HV=1 from HV=0 is delivered via real mode handlers.
734 *
735 *
736 * We layout physical memory as follows:
737 * 0x0000 - 0x00ff : Secondary processor spin code
738 * 0x0100 - 0x18ff : Real mode pSeries interrupt vectors
739 * 0x1900 - 0x2fff : Real mode trampolines
740 * 0x3000 - 0x58ff : Relon (IR=1,DR=1) mode pSeries interrupt vectors
741 * 0x5900 - 0x6fff : Relon mode trampolines
742 * 0x7000 - 0x7fff : FWNMI data area
743 * 0x8000 -   .... : Common interrupt handlers, remaining early
744 *                   setup code, rest of kernel.
745 *
746 * We could reclaim 0x4000-0x42ff for real mode trampolines if the space
747 * is necessary. Until then it's more consistent to explicitly put VIRT_NONE
748 * vectors there.
749 */
750OPEN_FIXED_SECTION(real_vectors,        0x0100, 0x1900)
751OPEN_FIXED_SECTION(real_trampolines,    0x1900, 0x3000)
752OPEN_FIXED_SECTION(virt_vectors,        0x3000, 0x5900)
753OPEN_FIXED_SECTION(virt_trampolines,    0x5900, 0x7000)
754
755#ifdef CONFIG_PPC_POWERNV
756	.globl start_real_trampolines
757	.globl end_real_trampolines
758	.globl start_virt_trampolines
759	.globl end_virt_trampolines
760#endif
761
762#if defined(CONFIG_PPC_PSERIES) || defined(CONFIG_PPC_POWERNV)
763/*
764 * Data area reserved for FWNMI option.
765 * This address (0x7000) is fixed by the RPA.
766 * pseries and powernv need to keep the whole page from
767 * 0x7000 to 0x8000 free for use by the firmware
768 */
769ZERO_FIXED_SECTION(fwnmi_page,          0x7000, 0x8000)
770OPEN_TEXT_SECTION(0x8000)
771#else
772OPEN_TEXT_SECTION(0x7000)
773#endif
774
775USE_FIXED_SECTION(real_vectors)
776
777/*
778 * This is the start of the interrupt handlers for pSeries
779 * This code runs with relocation off.
780 * Code from here to __end_interrupts gets copied down to real
781 * address 0x100 when we are running a relocatable kernel.
782 * Therefore any relative branches in this section must only
783 * branch to labels in this section.
784 */
785	.globl __start_interrupts
786__start_interrupts:
787
788/**
789 * Interrupt 0x3000 - System Call Vectored Interrupt (syscall).
790 * This is a synchronous interrupt invoked with the "scv" instruction. The
791 * system call does not alter the HV bit, so it is directed to the OS.
792 *
793 * Handling:
794 * scv instructions enter the kernel without changing EE, RI, ME, or HV.
795 * In particular, this means we can take a maskable interrupt at any point
796 * in the scv handler, which is unlike any other interrupt. This is solved
797 * by treating the instruction addresses in the handler as being soft-masked,
798 * by adding a SOFT_MASK_TABLE entry for them.
799 *
800 * AIL-0 mode scv exceptions go to 0x17000-0x17fff, but we set AIL-3 and
801 * ensure scv is never executed with relocation off, which means AIL-0
802 * should never happen.
803 *
804 * Before leaving the following inside-__end_soft_masked text, at least of the
805 * following must be true:
806 * - MSR[PR]=1 (i.e., return to userspace)
807 * - MSR_EE|MSR_RI is clear (no reentrant exceptions)
808 * - Standard kernel environment is set up (stack, paca, etc)
809 *
810 * Call convention:
811 *
812 * syscall register convention is in Documentation/powerpc/syscall64-abi.rst
813 */
814EXC_VIRT_BEGIN(system_call_vectored, 0x3000, 0x1000)
8151:
816	/* SCV 0 */
817	mr	r9,r13
818	GET_PACA(r13)
819	mflr	r11
820	mfctr	r12
821	li	r10,IRQS_ALL_DISABLED
822	stb	r10,PACAIRQSOFTMASK(r13)
823#ifdef CONFIG_RELOCATABLE
824	b	system_call_vectored_tramp
825#else
826	b	system_call_vectored_common
827#endif
828	nop
829
830	/* SCV 1 - 127 */
831	.rept	127
832	mr	r9,r13
833	GET_PACA(r13)
834	mflr	r11
835	mfctr	r12
836	li	r10,IRQS_ALL_DISABLED
837	stb	r10,PACAIRQSOFTMASK(r13)
838	li	r0,-1 /* cause failure */
839#ifdef CONFIG_RELOCATABLE
840	b	system_call_vectored_sigill_tramp
841#else
842	b	system_call_vectored_sigill
843#endif
844	.endr
8452:
846EXC_VIRT_END(system_call_vectored, 0x3000, 0x1000)
847
848SOFT_MASK_TABLE(1b, 2b) // Treat scv vectors as soft-masked, see comment above.
849
850#ifdef CONFIG_RELOCATABLE
851TRAMP_VIRT_BEGIN(system_call_vectored_tramp)
852	__LOAD_HANDLER(r10, system_call_vectored_common)
853	mtctr	r10
854	bctr
855
856TRAMP_VIRT_BEGIN(system_call_vectored_sigill_tramp)
857	__LOAD_HANDLER(r10, system_call_vectored_sigill)
858	mtctr	r10
859	bctr
860#endif
861
862
863/* No virt vectors corresponding with 0x0..0x100 */
864EXC_VIRT_NONE(0x4000, 0x100)
865
866
867/**
868 * Interrupt 0x100 - System Reset Interrupt (SRESET aka NMI).
869 * This is a non-maskable, asynchronous interrupt always taken in real-mode.
870 * It is caused by:
871 * - Wake from power-saving state, on powernv.
872 * - An NMI from another CPU, triggered by firmware or hypercall.
873 * - As crash/debug signal injected from BMC, firmware or hypervisor.
874 *
875 * Handling:
876 * Power-save wakeup is the only performance critical path, so this is
877 * determined quickly as possible first. In this case volatile registers
878 * can be discarded and SPRs like CFAR don't need to be read.
879 *
880 * If not a powersave wakeup, then it's run as a regular interrupt, however
881 * it uses its own stack and PACA save area to preserve the regular kernel
882 * environment for debugging.
883 *
884 * This interrupt is not maskable, so triggering it when MSR[RI] is clear,
885 * or SCRATCH0 is in use, etc. may cause a crash. It's also not entirely
886 * correct to switch to virtual mode to run the regular interrupt handler
887 * because it might be interrupted when the MMU is in a bad state (e.g., SLB
888 * is clear).
889 *
890 * FWNMI:
891 * PAPR specifies a "fwnmi" facility which sends the sreset to a different
892 * entry point with a different register set up. Some hypervisors will
893 * send the sreset to 0x100 in the guest if it is not fwnmi capable.
894 *
895 * KVM:
896 * Unlike most SRR interrupts, this may be taken by the host while executing
897 * in a guest, so a KVM test is required. KVM will pull the CPU out of guest
898 * mode and then raise the sreset.
899 */
900INT_DEFINE_BEGIN(system_reset)
901	IVEC=0x100
902	IAREA=PACA_EXNMI
903	IVIRT=0 /* no virt entry point */
904	/*
905	 * MSR_RI is not enabled, because PACA_EXNMI and nmi stack is
906	 * being used, so a nested NMI exception would corrupt it.
907	 */
908	ISET_RI=0
909	ISTACK=0
910	IKVM_REAL=1
911INT_DEFINE_END(system_reset)
912
913EXC_REAL_BEGIN(system_reset, 0x100, 0x100)
914#ifdef CONFIG_PPC_P7_NAP
915	/*
916	 * If running native on arch 2.06 or later, check if we are waking up
917	 * from nap/sleep/winkle, and branch to idle handler. This tests SRR1
918	 * bits 46:47. A non-0 value indicates that we are coming from a power
919	 * saving state. The idle wakeup handler initially runs in real mode,
920	 * but we branch to the 0xc000... address so we can turn on relocation
921	 * with mtmsrd later, after SPRs are restored.
922	 *
923	 * Careful to minimise cost for the fast path (idle wakeup) while
924	 * also avoiding clobbering CFAR for the debug path (non-idle).
925	 *
926	 * For the idle wake case volatile registers can be clobbered, which
927	 * is why we use those initially. If it turns out to not be an idle
928	 * wake, carefully put everything back the way it was, so we can use
929	 * common exception macros to handle it.
930	 */
931BEGIN_FTR_SECTION
932	SET_SCRATCH0(r13)
933	GET_PACA(r13)
934	std	r3,PACA_EXNMI+0*8(r13)
935	std	r4,PACA_EXNMI+1*8(r13)
936	std	r5,PACA_EXNMI+2*8(r13)
937	mfspr	r3,SPRN_SRR1
938	mfocrf	r4,0x80
939	rlwinm.	r5,r3,47-31,30,31
940	bne+	system_reset_idle_wake
941	/* Not powersave wakeup. Restore regs for regular interrupt handler. */
942	mtocrf	0x80,r4
943	ld	r3,PACA_EXNMI+0*8(r13)
944	ld	r4,PACA_EXNMI+1*8(r13)
945	ld	r5,PACA_EXNMI+2*8(r13)
946	GET_SCRATCH0(r13)
947END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
948#endif
949
950	GEN_INT_ENTRY system_reset, virt=0
951	/*
952	 * In theory, we should not enable relocation here if it was disabled
953	 * in SRR1, because the MMU may not be configured to support it (e.g.,
954	 * SLB may have been cleared). In practice, there should only be a few
955	 * small windows where that's the case, and sreset is considered to
956	 * be dangerous anyway.
957	 */
958EXC_REAL_END(system_reset, 0x100, 0x100)
959EXC_VIRT_NONE(0x4100, 0x100)
960
961#ifdef CONFIG_PPC_P7_NAP
962TRAMP_REAL_BEGIN(system_reset_idle_wake)
963	/* We are waking up from idle, so may clobber any volatile register */
964	cmpwi	cr1,r5,2
965	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
966	BRANCH_TO_C000(r12, DOTSYM(idle_return_gpr_loss))
967#endif
968
969#ifdef CONFIG_PPC_PSERIES
970/*
971 * Vectors for the FWNMI option.  Share common code.
972 */
973TRAMP_REAL_BEGIN(system_reset_fwnmi)
974	GEN_INT_ENTRY system_reset, virt=0
975
976#endif /* CONFIG_PPC_PSERIES */
977
978EXC_COMMON_BEGIN(system_reset_common)
979	__GEN_COMMON_ENTRY system_reset
980	/*
981	 * Increment paca->in_nmi then enable MSR_RI. SLB or MCE will be able
982	 * to recover, but nested NMI will notice in_nmi and not recover
983	 * because of the use of the NMI stack. in_nmi reentrancy is tested in
984	 * system_reset_exception.
985	 */
986	lhz	r10,PACA_IN_NMI(r13)
987	addi	r10,r10,1
988	sth	r10,PACA_IN_NMI(r13)
989	li	r10,MSR_RI
990	mtmsrd 	r10,1
991
992	mr	r10,r1
993	ld	r1,PACA_NMI_EMERG_SP(r13)
994	subi	r1,r1,INT_FRAME_SIZE
995	__GEN_COMMON_BODY system_reset
996
997	addi	r3,r1,STACK_FRAME_OVERHEAD
998	bl	system_reset_exception
999
1000	/* Clear MSR_RI before setting SRR0 and SRR1. */
1001	li	r9,0
1002	mtmsrd	r9,1
1003
1004	/*
1005	 * MSR_RI is clear, now we can decrement paca->in_nmi.
1006	 */
1007	lhz	r10,PACA_IN_NMI(r13)
1008	subi	r10,r10,1
1009	sth	r10,PACA_IN_NMI(r13)
1010
1011	kuap_kernel_restore r9, r10
1012	EXCEPTION_RESTORE_REGS
1013	RFI_TO_USER_OR_KERNEL
1014
1015
1016/**
1017 * Interrupt 0x200 - Machine Check Interrupt (MCE).
1018 * This is a non-maskable interrupt always taken in real-mode. It can be
1019 * synchronous or asynchronous, caused by hardware or software, and it may be
1020 * taken in a power-saving state.
1021 *
1022 * Handling:
1023 * Similarly to system reset, this uses its own stack and PACA save area,
1024 * the difference is re-entrancy is allowed on the machine check stack.
1025 *
1026 * machine_check_early is run in real mode, and carefully decodes the
1027 * machine check and tries to handle it (e.g., flush the SLB if there was an
1028 * error detected there), determines if it was recoverable and logs the
1029 * event.
1030 *
1031 * This early code does not "reconcile" irq soft-mask state like SRESET or
1032 * regular interrupts do, so irqs_disabled() among other things may not work
1033 * properly (irq disable/enable already doesn't work because irq tracing can
1034 * not work in real mode).
1035 *
1036 * Then, depending on the execution context when the interrupt is taken, there
1037 * are 3 main actions:
1038 * - Executing in kernel mode. The event is queued with irq_work, which means
1039 *   it is handled when it is next safe to do so (i.e., the kernel has enabled
1040 *   interrupts), which could be immediately when the interrupt returns. This
1041 *   avoids nasty issues like switching to virtual mode when the MMU is in a
1042 *   bad state, or when executing OPAL code. (SRESET is exposed to such issues,
1043 *   but it has different priorities). Check to see if the CPU was in power
1044 *   save, and return via the wake up code if it was.
1045 *
1046 * - Executing in user mode. machine_check_exception is run like a normal
1047 *   interrupt handler, which processes the data generated by the early handler.
1048 *
1049 * - Executing in guest mode. The interrupt is run with its KVM test, and
1050 *   branches to KVM to deal with. KVM may queue the event for the host
1051 *   to report later.
1052 *
1053 * This interrupt is not maskable, so if it triggers when MSR[RI] is clear,
1054 * or SCRATCH0 is in use, it may cause a crash.
1055 *
1056 * KVM:
1057 * See SRESET.
1058 */
1059INT_DEFINE_BEGIN(machine_check_early)
1060	IVEC=0x200
1061	IAREA=PACA_EXMC
1062	IVIRT=0 /* no virt entry point */
1063	IREALMODE_COMMON=1
1064	/*
1065	 * MSR_RI is not enabled, because PACA_EXMC is being used, so a
1066	 * nested machine check corrupts it. machine_check_common enables
1067	 * MSR_RI.
1068	 */
1069	ISET_RI=0
1070	ISTACK=0
1071	IDAR=1
1072	IDSISR=1
1073	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
1074INT_DEFINE_END(machine_check_early)
1075
1076INT_DEFINE_BEGIN(machine_check)
1077	IVEC=0x200
1078	IAREA=PACA_EXMC
1079	IVIRT=0 /* no virt entry point */
1080	ISET_RI=0
1081	IDAR=1
1082	IDSISR=1
1083	IKVM_REAL=1
1084INT_DEFINE_END(machine_check)
1085
1086EXC_REAL_BEGIN(machine_check, 0x200, 0x100)
1087	GEN_INT_ENTRY machine_check_early, virt=0
1088EXC_REAL_END(machine_check, 0x200, 0x100)
1089EXC_VIRT_NONE(0x4200, 0x100)
1090
1091#ifdef CONFIG_PPC_PSERIES
1092TRAMP_REAL_BEGIN(machine_check_fwnmi)
1093	/* See comment at machine_check exception, don't turn on RI */
1094	GEN_INT_ENTRY machine_check_early, virt=0
1095#endif
1096
1097#define MACHINE_CHECK_HANDLER_WINDUP			\
1098	/* Clear MSR_RI before setting SRR0 and SRR1. */\
1099	li	r9,0;					\
1100	mtmsrd	r9,1;		/* Clear MSR_RI */	\
1101	/* Decrement paca->in_mce now RI is clear. */	\
1102	lhz	r12,PACA_IN_MCE(r13);			\
1103	subi	r12,r12,1;				\
1104	sth	r12,PACA_IN_MCE(r13);			\
1105	EXCEPTION_RESTORE_REGS
1106
1107EXC_COMMON_BEGIN(machine_check_early_common)
1108	__GEN_REALMODE_COMMON_ENTRY machine_check_early
1109
1110	/*
1111	 * Switch to mc_emergency stack and handle re-entrancy (we limit
1112	 * the nested MCE upto level 4 to avoid stack overflow).
1113	 * Save MCE registers srr1, srr0, dar and dsisr and then set ME=1
1114	 *
1115	 * We use paca->in_mce to check whether this is the first entry or
1116	 * nested machine check. We increment paca->in_mce to track nested
1117	 * machine checks.
1118	 *
1119	 * If this is the first entry then set stack pointer to
1120	 * paca->mc_emergency_sp, otherwise r1 is already pointing to
1121	 * stack frame on mc_emergency stack.
1122	 *
1123	 * NOTE: We are here with MSR_ME=0 (off), which means we risk a
1124	 * checkstop if we get another machine check exception before we do
1125	 * rfid with MSR_ME=1.
1126	 *
1127	 * This interrupt can wake directly from idle. If that is the case,
1128	 * the machine check is handled then the idle wakeup code is called
1129	 * to restore state.
1130	 */
1131	lhz	r10,PACA_IN_MCE(r13)
1132	cmpwi	r10,0			/* Are we in nested machine check */
1133	cmpwi	cr1,r10,MAX_MCE_DEPTH	/* Are we at maximum nesting */
1134	addi	r10,r10,1		/* increment paca->in_mce */
1135	sth	r10,PACA_IN_MCE(r13)
1136
1137	mr	r10,r1			/* Save r1 */
1138	bne	1f
1139	/* First machine check entry */
1140	ld	r1,PACAMCEMERGSP(r13)	/* Use MC emergency stack */
11411:	/* Limit nested MCE to level 4 to avoid stack overflow */
1142	bgt	cr1,unrecoverable_mce	/* Check if we hit limit of 4 */
1143	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame */
1144
1145	__GEN_COMMON_BODY machine_check_early
1146
1147BEGIN_FTR_SECTION
1148	bl	enable_machine_check
1149END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1150	li	r10,MSR_RI
1151	mtmsrd	r10,1
1152
1153	addi	r3,r1,STACK_FRAME_OVERHEAD
1154	bl	machine_check_early
1155	std	r3,RESULT(r1)	/* Save result */
1156	ld	r12,_MSR(r1)
1157
1158#ifdef CONFIG_PPC_P7_NAP
1159	/*
1160	 * Check if thread was in power saving mode. We come here when any
1161	 * of the following is true:
1162	 * a. thread wasn't in power saving mode
1163	 * b. thread was in power saving mode with no state loss,
1164	 *    supervisor state loss or hypervisor state loss.
1165	 *
1166	 * Go back to nap/sleep/winkle mode again if (b) is true.
1167	 */
1168BEGIN_FTR_SECTION
1169	rlwinm.	r11,r12,47-31,30,31
1170	bne	machine_check_idle_common
1171END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
1172#endif
1173
1174#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1175	/*
1176	 * Check if we are coming from guest. If yes, then run the normal
1177	 * exception handler which will take the
1178	 * machine_check_kvm->kvm_interrupt branch to deliver the MC event
1179	 * to guest.
1180	 */
1181	lbz	r11,HSTATE_IN_GUEST(r13)
1182	cmpwi	r11,0			/* Check if coming from guest */
1183	bne	mce_deliver		/* continue if we are. */
1184#endif
1185
1186	/*
1187	 * Check if we are coming from userspace. If yes, then run the normal
1188	 * exception handler which will deliver the MC event to this kernel.
1189	 */
1190	andi.	r11,r12,MSR_PR		/* See if coming from user. */
1191	bne	mce_deliver		/* continue in V mode if we are. */
1192
1193	/*
1194	 * At this point we are coming from kernel context.
1195	 * Queue up the MCE event and return from the interrupt.
1196	 * But before that, check if this is an un-recoverable exception.
1197	 * If yes, then stay on emergency stack and panic.
1198	 */
1199	andi.	r11,r12,MSR_RI
1200	beq	unrecoverable_mce
1201
1202	/*
1203	 * Check if we have successfully handled/recovered from error, if not
1204	 * then stay on emergency stack and panic.
1205	 */
1206	ld	r3,RESULT(r1)	/* Load result */
1207	cmpdi	r3,0		/* see if we handled MCE successfully */
1208	beq	unrecoverable_mce /* if !handled then panic */
1209
1210	/*
1211	 * Return from MC interrupt.
1212	 * Queue up the MCE event so that we can log it later, while
1213	 * returning from kernel or opal call.
1214	 */
1215	bl	machine_check_queue_event
1216	MACHINE_CHECK_HANDLER_WINDUP
1217	RFI_TO_KERNEL
1218
1219mce_deliver:
1220	/*
1221	 * This is a host user or guest MCE. Restore all registers, then
1222	 * run the "late" handler. For host user, this will run the
1223	 * machine_check_exception handler in virtual mode like a normal
1224	 * interrupt handler. For guest, this will trigger the KVM test
1225	 * and branch to the KVM interrupt similarly to other interrupts.
1226	 */
1227BEGIN_FTR_SECTION
1228	ld	r10,ORIG_GPR3(r1)
1229	mtspr	SPRN_CFAR,r10
1230END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
1231	MACHINE_CHECK_HANDLER_WINDUP
1232	GEN_INT_ENTRY machine_check, virt=0
1233
1234EXC_COMMON_BEGIN(machine_check_common)
1235	/*
1236	 * Machine check is different because we use a different
1237	 * save area: PACA_EXMC instead of PACA_EXGEN.
1238	 */
1239	GEN_COMMON machine_check
1240
1241	/* Enable MSR_RI when finished with PACA_EXMC */
1242	li	r10,MSR_RI
1243	mtmsrd 	r10,1
1244	addi	r3,r1,STACK_FRAME_OVERHEAD
1245	bl	machine_check_exception
1246	b	interrupt_return_srr
1247
1248
1249#ifdef CONFIG_PPC_P7_NAP
1250/*
1251 * This is an idle wakeup. Low level machine check has already been
1252 * done. Queue the event then call the idle code to do the wake up.
1253 */
1254EXC_COMMON_BEGIN(machine_check_idle_common)
1255	bl	machine_check_queue_event
1256
1257	/*
1258	 * GPR-loss wakeups are relatively straightforward, because the
1259	 * idle sleep code has saved all non-volatile registers on its
1260	 * own stack, and r1 in PACAR1.
1261	 *
1262	 * For no-loss wakeups the r1 and lr registers used by the
1263	 * early machine check handler have to be restored first. r2 is
1264	 * the kernel TOC, so no need to restore it.
1265	 *
1266	 * Then decrement MCE nesting after finishing with the stack.
1267	 */
1268	ld	r3,_MSR(r1)
1269	ld	r4,_LINK(r1)
1270	ld	r1,GPR1(r1)
1271
1272	lhz	r11,PACA_IN_MCE(r13)
1273	subi	r11,r11,1
1274	sth	r11,PACA_IN_MCE(r13)
1275
1276	mtlr	r4
1277	rlwinm	r10,r3,47-31,30,31
1278	cmpwi	cr1,r10,2
1279	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
1280	b	idle_return_gpr_loss
1281#endif
1282
1283EXC_COMMON_BEGIN(unrecoverable_mce)
1284	/*
1285	 * We are going down. But there are chances that we might get hit by
1286	 * another MCE during panic path and we may run into unstable state
1287	 * with no way out. Hence, turn ME bit off while going down, so that
1288	 * when another MCE is hit during panic path, system will checkstop
1289	 * and hypervisor will get restarted cleanly by SP.
1290	 */
1291BEGIN_FTR_SECTION
1292	li	r10,0 /* clear MSR_RI */
1293	mtmsrd	r10,1
1294	bl	disable_machine_check
1295END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1296	ld	r10,PACAKMSR(r13)
1297	li	r3,MSR_ME
1298	andc	r10,r10,r3
1299	mtmsrd	r10
1300
1301	lhz	r12,PACA_IN_MCE(r13)
1302	subi	r12,r12,1
1303	sth	r12,PACA_IN_MCE(r13)
1304
1305	/* Invoke machine_check_exception to print MCE event and panic. */
1306	addi	r3,r1,STACK_FRAME_OVERHEAD
1307	bl	machine_check_exception
1308
1309	/*
1310	 * We will not reach here. Even if we did, there is no way out.
1311	 * Call unrecoverable_exception and die.
1312	 */
1313	addi	r3,r1,STACK_FRAME_OVERHEAD
1314	bl	unrecoverable_exception
1315	b	.
1316
1317
1318/**
1319 * Interrupt 0x300 - Data Storage Interrupt (DSI).
1320 * This is a synchronous interrupt generated due to a data access exception,
1321 * e.g., a load orstore which does not have a valid page table entry with
1322 * permissions. DAWR matches also fault here, as do RC updates, and minor misc
1323 * errors e.g., copy/paste, AMO, certain invalid CI accesses, etc.
1324 *
1325 * Handling:
1326 * - Hash MMU
1327 *   Go to do_hash_fault, which attempts to fill the HPT from an entry in the
1328 *   Linux page table. Hash faults can hit in kernel mode in a fairly
1329 *   arbitrary state (e.g., interrupts disabled, locks held) when accessing
1330 *   "non-bolted" regions, e.g., vmalloc space. However these should always be
1331 *   backed by Linux page table entries.
1332 *
1333 *   If no entry is found the Linux page fault handler is invoked (by
1334 *   do_hash_fault). Linux page faults can happen in kernel mode due to user
1335 *   copy operations of course.
1336 *
1337 *   KVM: The KVM HDSI handler may perform a load with MSR[DR]=1 in guest
1338 *   MMU context, which may cause a DSI in the host, which must go to the
1339 *   KVM handler. MSR[IR] is not enabled, so the real-mode handler will
1340 *   always be used regardless of AIL setting.
1341 *
1342 * - Radix MMU
1343 *   The hardware loads from the Linux page table directly, so a fault goes
1344 *   immediately to Linux page fault.
1345 *
1346 * Conditions like DAWR match are handled on the way in to Linux page fault.
1347 */
1348INT_DEFINE_BEGIN(data_access)
1349	IVEC=0x300
1350	IDAR=1
1351	IDSISR=1
1352	IKVM_REAL=1
1353INT_DEFINE_END(data_access)
1354
1355EXC_REAL_BEGIN(data_access, 0x300, 0x80)
1356	GEN_INT_ENTRY data_access, virt=0
1357EXC_REAL_END(data_access, 0x300, 0x80)
1358EXC_VIRT_BEGIN(data_access, 0x4300, 0x80)
1359	GEN_INT_ENTRY data_access, virt=1
1360EXC_VIRT_END(data_access, 0x4300, 0x80)
1361EXC_COMMON_BEGIN(data_access_common)
1362	GEN_COMMON data_access
1363	ld	r4,_DSISR(r1)
1364	addi	r3,r1,STACK_FRAME_OVERHEAD
1365	andis.	r0,r4,DSISR_DABRMATCH@h
1366	bne-	1f
1367BEGIN_MMU_FTR_SECTION
1368	bl	do_hash_fault
1369MMU_FTR_SECTION_ELSE
1370	bl	do_page_fault
1371ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1372	b	interrupt_return_srr
1373
13741:	bl	do_break
1375	/*
1376	 * do_break() may have changed the NV GPRS while handling a breakpoint.
1377	 * If so, we need to restore them with their updated values.
1378	 */
1379	REST_NVGPRS(r1)
1380	b	interrupt_return_srr
1381
1382
1383/**
1384 * Interrupt 0x380 - Data Segment Interrupt (DSLB).
1385 * This is a synchronous interrupt in response to an MMU fault missing SLB
1386 * entry for HPT, or an address outside RPT translation range.
1387 *
1388 * Handling:
1389 * - HPT:
1390 *   This refills the SLB, or reports an access fault similarly to a bad page
1391 *   fault. When coming from user-mode, the SLB handler may access any kernel
1392 *   data, though it may itself take a DSLB. When coming from kernel mode,
1393 *   recursive faults must be avoided so access is restricted to the kernel
1394 *   image text/data, kernel stack, and any data allocated below
1395 *   ppc64_bolted_size (first segment). The kernel handler must avoid stomping
1396 *   on user-handler data structures.
1397 *
1398 *   KVM: Same as 0x300, DSLB must test for KVM guest.
1399 */
1400INT_DEFINE_BEGIN(data_access_slb)
1401	IVEC=0x380
1402	IDAR=1
1403	IKVM_REAL=1
1404INT_DEFINE_END(data_access_slb)
1405
1406EXC_REAL_BEGIN(data_access_slb, 0x380, 0x80)
1407	GEN_INT_ENTRY data_access_slb, virt=0
1408EXC_REAL_END(data_access_slb, 0x380, 0x80)
1409EXC_VIRT_BEGIN(data_access_slb, 0x4380, 0x80)
1410	GEN_INT_ENTRY data_access_slb, virt=1
1411EXC_VIRT_END(data_access_slb, 0x4380, 0x80)
1412EXC_COMMON_BEGIN(data_access_slb_common)
1413	GEN_COMMON data_access_slb
1414BEGIN_MMU_FTR_SECTION
1415	/* HPT case, do SLB fault */
1416	addi	r3,r1,STACK_FRAME_OVERHEAD
1417	bl	do_slb_fault
1418	cmpdi	r3,0
1419	bne-	1f
1420	b	fast_interrupt_return_srr
14211:	/* Error case */
1422MMU_FTR_SECTION_ELSE
1423	/* Radix case, access is outside page table range */
1424	li	r3,-EFAULT
1425ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1426	std	r3,RESULT(r1)
1427	addi	r3,r1,STACK_FRAME_OVERHEAD
1428	bl	do_bad_slb_fault
1429	b	interrupt_return_srr
1430
1431
1432/**
1433 * Interrupt 0x400 - Instruction Storage Interrupt (ISI).
1434 * This is a synchronous interrupt in response to an MMU fault due to an
1435 * instruction fetch.
1436 *
1437 * Handling:
1438 * Similar to DSI, though in response to fetch. The faulting address is found
1439 * in SRR0 (rather than DAR), and status in SRR1 (rather than DSISR).
1440 */
1441INT_DEFINE_BEGIN(instruction_access)
1442	IVEC=0x400
1443	IISIDE=1
1444	IDAR=1
1445	IDSISR=1
1446#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1447	IKVM_REAL=1
1448#endif
1449INT_DEFINE_END(instruction_access)
1450
1451EXC_REAL_BEGIN(instruction_access, 0x400, 0x80)
1452	GEN_INT_ENTRY instruction_access, virt=0
1453EXC_REAL_END(instruction_access, 0x400, 0x80)
1454EXC_VIRT_BEGIN(instruction_access, 0x4400, 0x80)
1455	GEN_INT_ENTRY instruction_access, virt=1
1456EXC_VIRT_END(instruction_access, 0x4400, 0x80)
1457EXC_COMMON_BEGIN(instruction_access_common)
1458	GEN_COMMON instruction_access
1459	addi	r3,r1,STACK_FRAME_OVERHEAD
1460BEGIN_MMU_FTR_SECTION
1461	bl	do_hash_fault
1462MMU_FTR_SECTION_ELSE
1463	bl	do_page_fault
1464ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1465	b	interrupt_return_srr
1466
1467
1468/**
1469 * Interrupt 0x480 - Instruction Segment Interrupt (ISLB).
1470 * This is a synchronous interrupt in response to an MMU fault due to an
1471 * instruction fetch.
1472 *
1473 * Handling:
1474 * Similar to DSLB, though in response to fetch. The faulting address is found
1475 * in SRR0 (rather than DAR).
1476 */
1477INT_DEFINE_BEGIN(instruction_access_slb)
1478	IVEC=0x480
1479	IISIDE=1
1480	IDAR=1
1481#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1482	IKVM_REAL=1
1483#endif
1484INT_DEFINE_END(instruction_access_slb)
1485
1486EXC_REAL_BEGIN(instruction_access_slb, 0x480, 0x80)
1487	GEN_INT_ENTRY instruction_access_slb, virt=0
1488EXC_REAL_END(instruction_access_slb, 0x480, 0x80)
1489EXC_VIRT_BEGIN(instruction_access_slb, 0x4480, 0x80)
1490	GEN_INT_ENTRY instruction_access_slb, virt=1
1491EXC_VIRT_END(instruction_access_slb, 0x4480, 0x80)
1492EXC_COMMON_BEGIN(instruction_access_slb_common)
1493	GEN_COMMON instruction_access_slb
1494BEGIN_MMU_FTR_SECTION
1495	/* HPT case, do SLB fault */
1496	addi	r3,r1,STACK_FRAME_OVERHEAD
1497	bl	do_slb_fault
1498	cmpdi	r3,0
1499	bne-	1f
1500	b	fast_interrupt_return_srr
15011:	/* Error case */
1502MMU_FTR_SECTION_ELSE
1503	/* Radix case, access is outside page table range */
1504	li	r3,-EFAULT
1505ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1506	std	r3,RESULT(r1)
1507	addi	r3,r1,STACK_FRAME_OVERHEAD
1508	bl	do_bad_slb_fault
1509	b	interrupt_return_srr
1510
1511
1512/**
1513 * Interrupt 0x500 - External Interrupt.
1514 * This is an asynchronous maskable interrupt in response to an "external
1515 * exception" from the interrupt controller or hypervisor (e.g., device
1516 * interrupt). It is maskable in hardware by clearing MSR[EE], and
1517 * soft-maskable with IRQS_DISABLED mask (i.e., local_irq_disable()).
1518 *
1519 * When running in HV mode, Linux sets up the LPCR[LPES] bit such that
1520 * interrupts are delivered with HSRR registers, guests use SRRs, which
1521 * reqiures IHSRR_IF_HVMODE.
1522 *
1523 * On bare metal POWER9 and later, Linux sets the LPCR[HVICE] bit such that
1524 * external interrupts are delivered as Hypervisor Virtualization Interrupts
1525 * rather than External Interrupts.
1526 *
1527 * Handling:
1528 * This calls into Linux IRQ handler. NVGPRs are not saved to reduce overhead,
1529 * because registers at the time of the interrupt are not so important as it is
1530 * asynchronous.
1531 *
1532 * If soft masked, the masked handler will note the pending interrupt for
1533 * replay, and clear MSR[EE] in the interrupted context.
1534 */
1535INT_DEFINE_BEGIN(hardware_interrupt)
1536	IVEC=0x500
1537	IHSRR_IF_HVMODE=1
1538	IMASK=IRQS_DISABLED
1539	IKVM_REAL=1
1540	IKVM_VIRT=1
1541INT_DEFINE_END(hardware_interrupt)
1542
1543EXC_REAL_BEGIN(hardware_interrupt, 0x500, 0x100)
1544	GEN_INT_ENTRY hardware_interrupt, virt=0
1545EXC_REAL_END(hardware_interrupt, 0x500, 0x100)
1546EXC_VIRT_BEGIN(hardware_interrupt, 0x4500, 0x100)
1547	GEN_INT_ENTRY hardware_interrupt, virt=1
1548EXC_VIRT_END(hardware_interrupt, 0x4500, 0x100)
1549EXC_COMMON_BEGIN(hardware_interrupt_common)
1550	GEN_COMMON hardware_interrupt
1551	addi	r3,r1,STACK_FRAME_OVERHEAD
1552	bl	do_IRQ
1553	BEGIN_FTR_SECTION
1554	b	interrupt_return_hsrr
1555	FTR_SECTION_ELSE
1556	b	interrupt_return_srr
1557	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
1558
1559
1560/**
1561 * Interrupt 0x600 - Alignment Interrupt
1562 * This is a synchronous interrupt in response to data alignment fault.
1563 */
1564INT_DEFINE_BEGIN(alignment)
1565	IVEC=0x600
1566	IDAR=1
1567	IDSISR=1
1568#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1569	IKVM_REAL=1
1570#endif
1571INT_DEFINE_END(alignment)
1572
1573EXC_REAL_BEGIN(alignment, 0x600, 0x100)
1574	GEN_INT_ENTRY alignment, virt=0
1575EXC_REAL_END(alignment, 0x600, 0x100)
1576EXC_VIRT_BEGIN(alignment, 0x4600, 0x100)
1577	GEN_INT_ENTRY alignment, virt=1
1578EXC_VIRT_END(alignment, 0x4600, 0x100)
1579EXC_COMMON_BEGIN(alignment_common)
1580	GEN_COMMON alignment
1581	addi	r3,r1,STACK_FRAME_OVERHEAD
1582	bl	alignment_exception
1583	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
1584	b	interrupt_return_srr
1585
1586
1587/**
1588 * Interrupt 0x700 - Program Interrupt (program check).
1589 * This is a synchronous interrupt in response to various instruction faults:
1590 * traps, privilege errors, TM errors, floating point exceptions.
1591 *
1592 * Handling:
1593 * This interrupt may use the "emergency stack" in some cases when being taken
1594 * from kernel context, which complicates handling.
1595 */
1596INT_DEFINE_BEGIN(program_check)
1597	IVEC=0x700
1598#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1599	IKVM_REAL=1
1600#endif
1601INT_DEFINE_END(program_check)
1602
1603EXC_REAL_BEGIN(program_check, 0x700, 0x100)
1604
1605#ifdef CONFIG_CPU_LITTLE_ENDIAN
1606	/*
1607	 * There's a short window during boot where although the kernel is
1608	 * running little endian, any exceptions will cause the CPU to switch
1609	 * back to big endian. For example a WARN() boils down to a trap
1610	 * instruction, which will cause a program check, and we end up here but
1611	 * with the CPU in big endian mode. The first instruction of the program
1612	 * check handler (in GEN_INT_ENTRY below) is an mtsprg, which when
1613	 * executed in the wrong endian is an lhzu with a ~3GB displacement from
1614	 * r3. The content of r3 is random, so that is a load from some random
1615	 * location, and depending on the system can easily lead to a checkstop,
1616	 * or an infinitely recursive page fault.
1617	 *
1618	 * So to handle that case we have a trampoline here that can detect we
1619	 * are in the wrong endian and flip us back to the correct endian. We
1620	 * can't flip MSR[LE] using mtmsr, so we have to use rfid. That requires
1621	 * backing up SRR0/1 as well as a GPR. To do that we use SPRG0/2/3, as
1622	 * SPRG1 is already used for the paca. SPRG3 is user readable, but this
1623	 * trampoline is only active very early in boot, and SPRG3 will be
1624	 * reinitialised in vdso_getcpu_init() before userspace starts.
1625	 */
1626BEGIN_FTR_SECTION
1627	tdi   0,0,0x48    // Trap never, or in reverse endian: b . + 8
1628	b     1f          // Skip trampoline if endian is correct
1629	.long 0xa643707d  // mtsprg  0, r11      Backup r11
1630	.long 0xa6027a7d  // mfsrr0  r11
1631	.long 0xa643727d  // mtsprg  2, r11      Backup SRR0 in SPRG2
1632	.long 0xa6027b7d  // mfsrr1  r11
1633	.long 0xa643737d  // mtsprg  3, r11      Backup SRR1 in SPRG3
1634	.long 0xa600607d  // mfmsr   r11
1635	.long 0x01006b69  // xori    r11, r11, 1 Invert MSR[LE]
1636	.long 0xa6037b7d  // mtsrr1  r11
1637	.long 0x34076039  // li      r11, 0x734
1638	.long 0xa6037a7d  // mtsrr0  r11
1639	.long 0x2400004c  // rfid
1640	mfsprg r11, 3
1641	mtsrr1 r11        // Restore SRR1
1642	mfsprg r11, 2
1643	mtsrr0 r11        // Restore SRR0
1644	mfsprg r11, 0     // Restore r11
16451:
1646END_FTR_SECTION(0, 1)     // nop out after boot
1647#endif /* CONFIG_CPU_LITTLE_ENDIAN */
1648
1649	GEN_INT_ENTRY program_check, virt=0
1650EXC_REAL_END(program_check, 0x700, 0x100)
1651EXC_VIRT_BEGIN(program_check, 0x4700, 0x100)
1652	GEN_INT_ENTRY program_check, virt=1
1653EXC_VIRT_END(program_check, 0x4700, 0x100)
1654EXC_COMMON_BEGIN(program_check_common)
1655	__GEN_COMMON_ENTRY program_check
1656
1657	/*
1658	 * It's possible to receive a TM Bad Thing type program check with
1659	 * userspace register values (in particular r1), but with SRR1 reporting
1660	 * that we came from the kernel. Normally that would confuse the bad
1661	 * stack logic, and we would report a bad kernel stack pointer. Instead
1662	 * we switch to the emergency stack if we're taking a TM Bad Thing from
1663	 * the kernel.
1664	 */
1665
1666	andi.	r10,r12,MSR_PR
1667	bne	2f			/* If userspace, go normal path */
1668
1669	andis.	r10,r12,(SRR1_PROGTM)@h
1670	bne	1f			/* If TM, emergency		*/
1671
1672	cmpdi	r1,-INT_FRAME_SIZE	/* check if r1 is in userspace	*/
1673	blt	2f			/* normal path if not		*/
1674
1675	/* Use the emergency stack					*/
16761:	andi.	r10,r12,MSR_PR		/* Set CR0 correctly for label	*/
1677					/* 3 in EXCEPTION_PROLOG_COMMON	*/
1678	mr	r10,r1			/* Save r1			*/
1679	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack		*/
1680	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
1681	__ISTACK(program_check)=0
1682	__GEN_COMMON_BODY program_check
1683	b 3f
16842:
1685	__ISTACK(program_check)=1
1686	__GEN_COMMON_BODY program_check
16873:
1688	addi	r3,r1,STACK_FRAME_OVERHEAD
1689	bl	program_check_exception
1690	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
1691	b	interrupt_return_srr
1692
1693
1694/*
1695 * Interrupt 0x800 - Floating-Point Unavailable Interrupt.
1696 * This is a synchronous interrupt in response to executing an fp instruction
1697 * with MSR[FP]=0.
1698 *
1699 * Handling:
1700 * This will load FP registers and enable the FP bit if coming from userspace,
1701 * otherwise report a bad kernel use of FP.
1702 */
1703INT_DEFINE_BEGIN(fp_unavailable)
1704	IVEC=0x800
1705#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1706	IKVM_REAL=1
1707#endif
1708INT_DEFINE_END(fp_unavailable)
1709
1710EXC_REAL_BEGIN(fp_unavailable, 0x800, 0x100)
1711	GEN_INT_ENTRY fp_unavailable, virt=0
1712EXC_REAL_END(fp_unavailable, 0x800, 0x100)
1713EXC_VIRT_BEGIN(fp_unavailable, 0x4800, 0x100)
1714	GEN_INT_ENTRY fp_unavailable, virt=1
1715EXC_VIRT_END(fp_unavailable, 0x4800, 0x100)
1716EXC_COMMON_BEGIN(fp_unavailable_common)
1717	GEN_COMMON fp_unavailable
1718	bne	1f			/* if from user, just load it up */
1719	addi	r3,r1,STACK_FRAME_OVERHEAD
1720	bl	kernel_fp_unavailable_exception
17210:	trap
1722	EMIT_BUG_ENTRY 0b, __FILE__, __LINE__, 0
17231:
1724#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1725BEGIN_FTR_SECTION
1726	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
1727	 * transaction), go do TM stuff
1728	 */
1729	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
1730	bne-	2f
1731END_FTR_SECTION_IFSET(CPU_FTR_TM)
1732#endif
1733	bl	load_up_fpu
1734	b	fast_interrupt_return_srr
1735#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
17362:	/* User process was in a transaction */
1737	addi	r3,r1,STACK_FRAME_OVERHEAD
1738	bl	fp_unavailable_tm
1739	b	interrupt_return_srr
1740#endif
1741
1742
1743/**
1744 * Interrupt 0x900 - Decrementer Interrupt.
1745 * This is an asynchronous interrupt in response to a decrementer exception
1746 * (e.g., DEC has wrapped below zero). It is maskable in hardware by clearing
1747 * MSR[EE], and soft-maskable with IRQS_DISABLED mask (i.e.,
1748 * local_irq_disable()).
1749 *
1750 * Handling:
1751 * This calls into Linux timer handler. NVGPRs are not saved (see 0x500).
1752 *
1753 * If soft masked, the masked handler will note the pending interrupt for
1754 * replay, and bump the decrementer to a high value, leaving MSR[EE] enabled
1755 * in the interrupted context.
1756 * If PPC_WATCHDOG is configured, the soft masked handler will actually set
1757 * things back up to run soft_nmi_interrupt as a regular interrupt handler
1758 * on the emergency stack.
1759 */
1760INT_DEFINE_BEGIN(decrementer)
1761	IVEC=0x900
1762	IMASK=IRQS_DISABLED
1763#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1764	IKVM_REAL=1
1765#endif
1766INT_DEFINE_END(decrementer)
1767
1768EXC_REAL_BEGIN(decrementer, 0x900, 0x80)
1769	GEN_INT_ENTRY decrementer, virt=0
1770EXC_REAL_END(decrementer, 0x900, 0x80)
1771EXC_VIRT_BEGIN(decrementer, 0x4900, 0x80)
1772	GEN_INT_ENTRY decrementer, virt=1
1773EXC_VIRT_END(decrementer, 0x4900, 0x80)
1774EXC_COMMON_BEGIN(decrementer_common)
1775	GEN_COMMON decrementer
1776	addi	r3,r1,STACK_FRAME_OVERHEAD
1777	bl	timer_interrupt
1778	b	interrupt_return_srr
1779
1780
1781/**
1782 * Interrupt 0x980 - Hypervisor Decrementer Interrupt.
1783 * This is an asynchronous interrupt, similar to 0x900 but for the HDEC
1784 * register.
1785 *
1786 * Handling:
1787 * Linux does not use this outside KVM where it's used to keep a host timer
1788 * while the guest is given control of DEC. It should normally be caught by
1789 * the KVM test and routed there.
1790 */
1791INT_DEFINE_BEGIN(hdecrementer)
1792	IVEC=0x980
1793	IHSRR=1
1794	ISTACK=0
1795	IKVM_REAL=1
1796	IKVM_VIRT=1
1797INT_DEFINE_END(hdecrementer)
1798
1799EXC_REAL_BEGIN(hdecrementer, 0x980, 0x80)
1800	GEN_INT_ENTRY hdecrementer, virt=0
1801EXC_REAL_END(hdecrementer, 0x980, 0x80)
1802EXC_VIRT_BEGIN(hdecrementer, 0x4980, 0x80)
1803	GEN_INT_ENTRY hdecrementer, virt=1
1804EXC_VIRT_END(hdecrementer, 0x4980, 0x80)
1805EXC_COMMON_BEGIN(hdecrementer_common)
1806	__GEN_COMMON_ENTRY hdecrementer
1807	/*
1808	 * Hypervisor decrementer interrupts not caught by the KVM test
1809	 * shouldn't occur but are sometimes left pending on exit from a KVM
1810	 * guest.  We don't need to do anything to clear them, as they are
1811	 * edge-triggered.
1812	 *
1813	 * Be careful to avoid touching the kernel stack.
1814	 */
1815	li	r10,0
1816	stb	r10,PACAHSRR_VALID(r13)
1817	ld	r10,PACA_EXGEN+EX_CTR(r13)
1818	mtctr	r10
1819	mtcrf	0x80,r9
1820	ld	r9,PACA_EXGEN+EX_R9(r13)
1821	ld	r10,PACA_EXGEN+EX_R10(r13)
1822	ld	r11,PACA_EXGEN+EX_R11(r13)
1823	ld	r12,PACA_EXGEN+EX_R12(r13)
1824	ld	r13,PACA_EXGEN+EX_R13(r13)
1825	HRFI_TO_KERNEL
1826
1827
1828/**
1829 * Interrupt 0xa00 - Directed Privileged Doorbell Interrupt.
1830 * This is an asynchronous interrupt in response to a msgsndp doorbell.
1831 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
1832 * IRQS_DISABLED mask (i.e., local_irq_disable()).
1833 *
1834 * Handling:
1835 * Guests may use this for IPIs between threads in a core if the
1836 * hypervisor supports it. NVGPRS are not saved (see 0x500).
1837 *
1838 * If soft masked, the masked handler will note the pending interrupt for
1839 * replay, leaving MSR[EE] enabled in the interrupted context because the
1840 * doorbells are edge triggered.
1841 */
1842INT_DEFINE_BEGIN(doorbell_super)
1843	IVEC=0xa00
1844	IMASK=IRQS_DISABLED
1845#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1846	IKVM_REAL=1
1847#endif
1848INT_DEFINE_END(doorbell_super)
1849
1850EXC_REAL_BEGIN(doorbell_super, 0xa00, 0x100)
1851	GEN_INT_ENTRY doorbell_super, virt=0
1852EXC_REAL_END(doorbell_super, 0xa00, 0x100)
1853EXC_VIRT_BEGIN(doorbell_super, 0x4a00, 0x100)
1854	GEN_INT_ENTRY doorbell_super, virt=1
1855EXC_VIRT_END(doorbell_super, 0x4a00, 0x100)
1856EXC_COMMON_BEGIN(doorbell_super_common)
1857	GEN_COMMON doorbell_super
1858	addi	r3,r1,STACK_FRAME_OVERHEAD
1859#ifdef CONFIG_PPC_DOORBELL
1860	bl	doorbell_exception
1861#else
1862	bl	unknown_async_exception
1863#endif
1864	b	interrupt_return_srr
1865
1866
1867EXC_REAL_NONE(0xb00, 0x100)
1868EXC_VIRT_NONE(0x4b00, 0x100)
1869
1870/**
1871 * Interrupt 0xc00 - System Call Interrupt (syscall, hcall).
1872 * This is a synchronous interrupt invoked with the "sc" instruction. The
1873 * system call is invoked with "sc 0" and does not alter the HV bit, so it
1874 * is directed to the currently running OS. The hypercall is invoked with
1875 * "sc 1" and it sets HV=1, so it elevates to hypervisor.
1876 *
1877 * In HPT, sc 1 always goes to 0xc00 real mode. In RADIX, sc 1 can go to
1878 * 0x4c00 virtual mode.
1879 *
1880 * Handling:
1881 * If the KVM test fires then it was due to a hypercall and is accordingly
1882 * routed to KVM. Otherwise this executes a normal Linux system call.
1883 *
1884 * Call convention:
1885 *
1886 * syscall and hypercalls register conventions are documented in
1887 * Documentation/powerpc/syscall64-abi.rst and
1888 * Documentation/powerpc/papr_hcalls.rst respectively.
1889 *
1890 * The intersection of volatile registers that don't contain possible
1891 * inputs is: cr0, xer, ctr. We may use these as scratch regs upon entry
1892 * without saving, though xer is not a good idea to use, as hardware may
1893 * interpret some bits so it may be costly to change them.
1894 */
1895INT_DEFINE_BEGIN(system_call)
1896	IVEC=0xc00
1897	IKVM_REAL=1
1898	IKVM_VIRT=1
1899INT_DEFINE_END(system_call)
1900
1901.macro SYSTEM_CALL virt
1902#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1903	/*
1904	 * There is a little bit of juggling to get syscall and hcall
1905	 * working well. Save r13 in ctr to avoid using SPRG scratch
1906	 * register.
1907	 *
1908	 * Userspace syscalls have already saved the PPR, hcalls must save
1909	 * it before setting HMT_MEDIUM.
1910	 */
1911	mtctr	r13
1912	GET_PACA(r13)
1913	std	r10,PACA_EXGEN+EX_R10(r13)
1914	INTERRUPT_TO_KERNEL
1915	KVMTEST system_call kvm_hcall /* uses r10, branch to kvm_hcall */
1916	mfctr	r9
1917#else
1918	mr	r9,r13
1919	GET_PACA(r13)
1920	INTERRUPT_TO_KERNEL
1921#endif
1922
1923#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
1924BEGIN_FTR_SECTION
1925	cmpdi	r0,0x1ebe
1926	beq-	1f
1927END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE)
1928#endif
1929
1930	/* We reach here with PACA in r13, r13 in r9. */
1931	mfspr	r11,SPRN_SRR0
1932	mfspr	r12,SPRN_SRR1
1933
1934	HMT_MEDIUM
1935
1936	.if ! \virt
1937	__LOAD_HANDLER(r10, system_call_common_real)
1938	mtctr	r10
1939	bctr
1940	.else
1941#ifdef CONFIG_RELOCATABLE
1942	__LOAD_HANDLER(r10, system_call_common)
1943	mtctr	r10
1944	bctr
1945#else
1946	b	system_call_common
1947#endif
1948	.endif
1949
1950#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
1951	/* Fast LE/BE switch system call */
19521:	mfspr	r12,SPRN_SRR1
1953	xori	r12,r12,MSR_LE
1954	mtspr	SPRN_SRR1,r12
1955	mr	r13,r9
1956	RFI_TO_USER	/* return to userspace */
1957	b	.	/* prevent speculative execution */
1958#endif
1959.endm
1960
1961EXC_REAL_BEGIN(system_call, 0xc00, 0x100)
1962	SYSTEM_CALL 0
1963EXC_REAL_END(system_call, 0xc00, 0x100)
1964EXC_VIRT_BEGIN(system_call, 0x4c00, 0x100)
1965	SYSTEM_CALL 1
1966EXC_VIRT_END(system_call, 0x4c00, 0x100)
1967
1968#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1969TRAMP_REAL_BEGIN(kvm_hcall)
1970	std	r9,PACA_EXGEN+EX_R9(r13)
1971	std	r11,PACA_EXGEN+EX_R11(r13)
1972	std	r12,PACA_EXGEN+EX_R12(r13)
1973	mfcr	r9
1974	mfctr	r10
1975	std	r10,PACA_EXGEN+EX_R13(r13)
1976	li	r10,0
1977	std	r10,PACA_EXGEN+EX_CFAR(r13)
1978	std	r10,PACA_EXGEN+EX_CTR(r13)
1979	 /*
1980	  * Save the PPR (on systems that support it) before changing to
1981	  * HMT_MEDIUM. That allows the KVM code to save that value into the
1982	  * guest state (it is the guest's PPR value).
1983	  */
1984BEGIN_FTR_SECTION
1985	mfspr	r10,SPRN_PPR
1986	std	r10,PACA_EXGEN+EX_PPR(r13)
1987END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
1988
1989	HMT_MEDIUM
1990
1991#ifdef CONFIG_RELOCATABLE
1992	/*
1993	 * Requires __LOAD_FAR_HANDLER beause kvmppc_hcall lives
1994	 * outside the head section.
1995	 */
1996	__LOAD_FAR_HANDLER(r10, kvmppc_hcall)
1997	mtctr   r10
1998	bctr
1999#else
2000	b       kvmppc_hcall
2001#endif
2002#endif
2003
2004/**
2005 * Interrupt 0xd00 - Trace Interrupt.
2006 * This is a synchronous interrupt in response to instruction step or
2007 * breakpoint faults.
2008 */
2009INT_DEFINE_BEGIN(single_step)
2010	IVEC=0xd00
2011#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2012	IKVM_REAL=1
2013#endif
2014INT_DEFINE_END(single_step)
2015
2016EXC_REAL_BEGIN(single_step, 0xd00, 0x100)
2017	GEN_INT_ENTRY single_step, virt=0
2018EXC_REAL_END(single_step, 0xd00, 0x100)
2019EXC_VIRT_BEGIN(single_step, 0x4d00, 0x100)
2020	GEN_INT_ENTRY single_step, virt=1
2021EXC_VIRT_END(single_step, 0x4d00, 0x100)
2022EXC_COMMON_BEGIN(single_step_common)
2023	GEN_COMMON single_step
2024	addi	r3,r1,STACK_FRAME_OVERHEAD
2025	bl	single_step_exception
2026	b	interrupt_return_srr
2027
2028
2029/**
2030 * Interrupt 0xe00 - Hypervisor Data Storage Interrupt (HDSI).
2031 * This is a synchronous interrupt in response to an MMU fault caused by a
2032 * guest data access.
2033 *
2034 * Handling:
2035 * This should always get routed to KVM. In radix MMU mode, this is caused
2036 * by a guest nested radix access that can't be performed due to the
2037 * partition scope page table. In hash mode, this can be caused by guests
2038 * running with translation disabled (virtual real mode) or with VPM enabled.
2039 * KVM will update the page table structures or disallow the access.
2040 */
2041INT_DEFINE_BEGIN(h_data_storage)
2042	IVEC=0xe00
2043	IHSRR=1
2044	IDAR=1
2045	IDSISR=1
2046	IKVM_REAL=1
2047	IKVM_VIRT=1
2048INT_DEFINE_END(h_data_storage)
2049
2050EXC_REAL_BEGIN(h_data_storage, 0xe00, 0x20)
2051	GEN_INT_ENTRY h_data_storage, virt=0, ool=1
2052EXC_REAL_END(h_data_storage, 0xe00, 0x20)
2053EXC_VIRT_BEGIN(h_data_storage, 0x4e00, 0x20)
2054	GEN_INT_ENTRY h_data_storage, virt=1, ool=1
2055EXC_VIRT_END(h_data_storage, 0x4e00, 0x20)
2056EXC_COMMON_BEGIN(h_data_storage_common)
2057	GEN_COMMON h_data_storage
2058	addi    r3,r1,STACK_FRAME_OVERHEAD
2059BEGIN_MMU_FTR_SECTION
2060	bl      do_bad_page_fault_segv
2061MMU_FTR_SECTION_ELSE
2062	bl      unknown_exception
2063ALT_MMU_FTR_SECTION_END_IFSET(MMU_FTR_TYPE_RADIX)
2064	b       interrupt_return_hsrr
2065
2066
2067/**
2068 * Interrupt 0xe20 - Hypervisor Instruction Storage Interrupt (HISI).
2069 * This is a synchronous interrupt in response to an MMU fault caused by a
2070 * guest instruction fetch, similar to HDSI.
2071 */
2072INT_DEFINE_BEGIN(h_instr_storage)
2073	IVEC=0xe20
2074	IHSRR=1
2075	IKVM_REAL=1
2076	IKVM_VIRT=1
2077INT_DEFINE_END(h_instr_storage)
2078
2079EXC_REAL_BEGIN(h_instr_storage, 0xe20, 0x20)
2080	GEN_INT_ENTRY h_instr_storage, virt=0, ool=1
2081EXC_REAL_END(h_instr_storage, 0xe20, 0x20)
2082EXC_VIRT_BEGIN(h_instr_storage, 0x4e20, 0x20)
2083	GEN_INT_ENTRY h_instr_storage, virt=1, ool=1
2084EXC_VIRT_END(h_instr_storage, 0x4e20, 0x20)
2085EXC_COMMON_BEGIN(h_instr_storage_common)
2086	GEN_COMMON h_instr_storage
2087	addi	r3,r1,STACK_FRAME_OVERHEAD
2088	bl	unknown_exception
2089	b	interrupt_return_hsrr
2090
2091
2092/**
2093 * Interrupt 0xe40 - Hypervisor Emulation Assistance Interrupt.
2094 */
2095INT_DEFINE_BEGIN(emulation_assist)
2096	IVEC=0xe40
2097	IHSRR=1
2098	IKVM_REAL=1
2099	IKVM_VIRT=1
2100INT_DEFINE_END(emulation_assist)
2101
2102EXC_REAL_BEGIN(emulation_assist, 0xe40, 0x20)
2103	GEN_INT_ENTRY emulation_assist, virt=0, ool=1
2104EXC_REAL_END(emulation_assist, 0xe40, 0x20)
2105EXC_VIRT_BEGIN(emulation_assist, 0x4e40, 0x20)
2106	GEN_INT_ENTRY emulation_assist, virt=1, ool=1
2107EXC_VIRT_END(emulation_assist, 0x4e40, 0x20)
2108EXC_COMMON_BEGIN(emulation_assist_common)
2109	GEN_COMMON emulation_assist
2110	addi	r3,r1,STACK_FRAME_OVERHEAD
2111	bl	emulation_assist_interrupt
2112	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2113	b	interrupt_return_hsrr
2114
2115
2116/**
2117 * Interrupt 0xe60 - Hypervisor Maintenance Interrupt (HMI).
2118 * This is an asynchronous interrupt caused by a Hypervisor Maintenance
2119 * Exception. It is always taken in real mode but uses HSRR registers
2120 * unlike SRESET and MCE.
2121 *
2122 * It is maskable in hardware by clearing MSR[EE], and partially soft-maskable
2123 * with IRQS_DISABLED mask (i.e., local_irq_disable()).
2124 *
2125 * Handling:
2126 * This is a special case, this is handled similarly to machine checks, with an
2127 * initial real mode handler that is not soft-masked, which attempts to fix the
2128 * problem. Then a regular handler which is soft-maskable and reports the
2129 * problem.
2130 *
2131 * The emergency stack is used for the early real mode handler.
2132 *
2133 * XXX: unclear why MCE and HMI schemes could not be made common, e.g.,
2134 * either use soft-masking for the MCE, or use irq_work for the HMI.
2135 *
2136 * KVM:
2137 * Unlike MCE, this calls into KVM without calling the real mode handler
2138 * first.
2139 */
2140INT_DEFINE_BEGIN(hmi_exception_early)
2141	IVEC=0xe60
2142	IHSRR=1
2143	IREALMODE_COMMON=1
2144	ISTACK=0
2145	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
2146	IKVM_REAL=1
2147INT_DEFINE_END(hmi_exception_early)
2148
2149INT_DEFINE_BEGIN(hmi_exception)
2150	IVEC=0xe60
2151	IHSRR=1
2152	IMASK=IRQS_DISABLED
2153	IKVM_REAL=1
2154INT_DEFINE_END(hmi_exception)
2155
2156EXC_REAL_BEGIN(hmi_exception, 0xe60, 0x20)
2157	GEN_INT_ENTRY hmi_exception_early, virt=0, ool=1
2158EXC_REAL_END(hmi_exception, 0xe60, 0x20)
2159EXC_VIRT_NONE(0x4e60, 0x20)
2160
2161EXC_COMMON_BEGIN(hmi_exception_early_common)
2162	__GEN_REALMODE_COMMON_ENTRY hmi_exception_early
2163
2164	mr	r10,r1			/* Save r1 */
2165	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack for realmode */
2166	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
2167
2168	__GEN_COMMON_BODY hmi_exception_early
2169
2170	addi	r3,r1,STACK_FRAME_OVERHEAD
2171	bl	hmi_exception_realmode
2172	cmpdi	cr0,r3,0
2173	bne	1f
2174
2175	EXCEPTION_RESTORE_REGS hsrr=1
2176	HRFI_TO_USER_OR_KERNEL
2177
21781:
2179	/*
2180	 * Go to virtual mode and pull the HMI event information from
2181	 * firmware.
2182	 */
2183	EXCEPTION_RESTORE_REGS hsrr=1
2184	GEN_INT_ENTRY hmi_exception, virt=0
2185
2186EXC_COMMON_BEGIN(hmi_exception_common)
2187	GEN_COMMON hmi_exception
2188	addi	r3,r1,STACK_FRAME_OVERHEAD
2189	bl	handle_hmi_exception
2190	b	interrupt_return_hsrr
2191
2192
2193/**
2194 * Interrupt 0xe80 - Directed Hypervisor Doorbell Interrupt.
2195 * This is an asynchronous interrupt in response to a msgsnd doorbell.
2196 * Similar to the 0xa00 doorbell but for host rather than guest.
2197 */
2198INT_DEFINE_BEGIN(h_doorbell)
2199	IVEC=0xe80
2200	IHSRR=1
2201	IMASK=IRQS_DISABLED
2202	IKVM_REAL=1
2203	IKVM_VIRT=1
2204INT_DEFINE_END(h_doorbell)
2205
2206EXC_REAL_BEGIN(h_doorbell, 0xe80, 0x20)
2207	GEN_INT_ENTRY h_doorbell, virt=0, ool=1
2208EXC_REAL_END(h_doorbell, 0xe80, 0x20)
2209EXC_VIRT_BEGIN(h_doorbell, 0x4e80, 0x20)
2210	GEN_INT_ENTRY h_doorbell, virt=1, ool=1
2211EXC_VIRT_END(h_doorbell, 0x4e80, 0x20)
2212EXC_COMMON_BEGIN(h_doorbell_common)
2213	GEN_COMMON h_doorbell
2214	addi	r3,r1,STACK_FRAME_OVERHEAD
2215#ifdef CONFIG_PPC_DOORBELL
2216	bl	doorbell_exception
2217#else
2218	bl	unknown_async_exception
2219#endif
2220	b	interrupt_return_hsrr
2221
2222
2223/**
2224 * Interrupt 0xea0 - Hypervisor Virtualization Interrupt.
2225 * This is an asynchronous interrupt in response to an "external exception".
2226 * Similar to 0x500 but for host only.
2227 */
2228INT_DEFINE_BEGIN(h_virt_irq)
2229	IVEC=0xea0
2230	IHSRR=1
2231	IMASK=IRQS_DISABLED
2232	IKVM_REAL=1
2233	IKVM_VIRT=1
2234INT_DEFINE_END(h_virt_irq)
2235
2236EXC_REAL_BEGIN(h_virt_irq, 0xea0, 0x20)
2237	GEN_INT_ENTRY h_virt_irq, virt=0, ool=1
2238EXC_REAL_END(h_virt_irq, 0xea0, 0x20)
2239EXC_VIRT_BEGIN(h_virt_irq, 0x4ea0, 0x20)
2240	GEN_INT_ENTRY h_virt_irq, virt=1, ool=1
2241EXC_VIRT_END(h_virt_irq, 0x4ea0, 0x20)
2242EXC_COMMON_BEGIN(h_virt_irq_common)
2243	GEN_COMMON h_virt_irq
2244	addi	r3,r1,STACK_FRAME_OVERHEAD
2245	bl	do_IRQ
2246	b	interrupt_return_hsrr
2247
2248
2249EXC_REAL_NONE(0xec0, 0x20)
2250EXC_VIRT_NONE(0x4ec0, 0x20)
2251EXC_REAL_NONE(0xee0, 0x20)
2252EXC_VIRT_NONE(0x4ee0, 0x20)
2253
2254
2255/*
2256 * Interrupt 0xf00 - Performance Monitor Interrupt (PMI, PMU).
2257 * This is an asynchronous interrupt in response to a PMU exception.
2258 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
2259 * IRQS_PMI_DISABLED mask (NOTE: NOT local_irq_disable()).
2260 *
2261 * Handling:
2262 * This calls into the perf subsystem.
2263 *
2264 * Like the watchdog soft-nmi, it appears an NMI interrupt to Linux, in that it
2265 * runs under local_irq_disable. However it may be soft-masked in
2266 * powerpc-specific code.
2267 *
2268 * If soft masked, the masked handler will note the pending interrupt for
2269 * replay, and clear MSR[EE] in the interrupted context.
2270 */
2271INT_DEFINE_BEGIN(performance_monitor)
2272	IVEC=0xf00
2273	IMASK=IRQS_PMI_DISABLED
2274#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2275	IKVM_REAL=1
2276#endif
2277INT_DEFINE_END(performance_monitor)
2278
2279EXC_REAL_BEGIN(performance_monitor, 0xf00, 0x20)
2280	GEN_INT_ENTRY performance_monitor, virt=0, ool=1
2281EXC_REAL_END(performance_monitor, 0xf00, 0x20)
2282EXC_VIRT_BEGIN(performance_monitor, 0x4f00, 0x20)
2283	GEN_INT_ENTRY performance_monitor, virt=1, ool=1
2284EXC_VIRT_END(performance_monitor, 0x4f00, 0x20)
2285EXC_COMMON_BEGIN(performance_monitor_common)
2286	GEN_COMMON performance_monitor
2287	addi	r3,r1,STACK_FRAME_OVERHEAD
2288	bl	performance_monitor_exception
2289	b	interrupt_return_srr
2290
2291
2292/**
2293 * Interrupt 0xf20 - Vector Unavailable Interrupt.
2294 * This is a synchronous interrupt in response to
2295 * executing a vector (or altivec) instruction with MSR[VEC]=0.
2296 * Similar to FP unavailable.
2297 */
2298INT_DEFINE_BEGIN(altivec_unavailable)
2299	IVEC=0xf20
2300#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2301	IKVM_REAL=1
2302#endif
2303INT_DEFINE_END(altivec_unavailable)
2304
2305EXC_REAL_BEGIN(altivec_unavailable, 0xf20, 0x20)
2306	GEN_INT_ENTRY altivec_unavailable, virt=0, ool=1
2307EXC_REAL_END(altivec_unavailable, 0xf20, 0x20)
2308EXC_VIRT_BEGIN(altivec_unavailable, 0x4f20, 0x20)
2309	GEN_INT_ENTRY altivec_unavailable, virt=1, ool=1
2310EXC_VIRT_END(altivec_unavailable, 0x4f20, 0x20)
2311EXC_COMMON_BEGIN(altivec_unavailable_common)
2312	GEN_COMMON altivec_unavailable
2313#ifdef CONFIG_ALTIVEC
2314BEGIN_FTR_SECTION
2315	beq	1f
2316#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2317  BEGIN_FTR_SECTION_NESTED(69)
2318	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2319	 * transaction), go do TM stuff
2320	 */
2321	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2322	bne-	2f
2323  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2324#endif
2325	bl	load_up_altivec
2326	b	fast_interrupt_return_srr
2327#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
23282:	/* User process was in a transaction */
2329	addi	r3,r1,STACK_FRAME_OVERHEAD
2330	bl	altivec_unavailable_tm
2331	b	interrupt_return_srr
2332#endif
23331:
2334END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2335#endif
2336	addi	r3,r1,STACK_FRAME_OVERHEAD
2337	bl	altivec_unavailable_exception
2338	b	interrupt_return_srr
2339
2340
2341/**
2342 * Interrupt 0xf40 - VSX Unavailable Interrupt.
2343 * This is a synchronous interrupt in response to
2344 * executing a VSX instruction with MSR[VSX]=0.
2345 * Similar to FP unavailable.
2346 */
2347INT_DEFINE_BEGIN(vsx_unavailable)
2348	IVEC=0xf40
2349#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2350	IKVM_REAL=1
2351#endif
2352INT_DEFINE_END(vsx_unavailable)
2353
2354EXC_REAL_BEGIN(vsx_unavailable, 0xf40, 0x20)
2355	GEN_INT_ENTRY vsx_unavailable, virt=0, ool=1
2356EXC_REAL_END(vsx_unavailable, 0xf40, 0x20)
2357EXC_VIRT_BEGIN(vsx_unavailable, 0x4f40, 0x20)
2358	GEN_INT_ENTRY vsx_unavailable, virt=1, ool=1
2359EXC_VIRT_END(vsx_unavailable, 0x4f40, 0x20)
2360EXC_COMMON_BEGIN(vsx_unavailable_common)
2361	GEN_COMMON vsx_unavailable
2362#ifdef CONFIG_VSX
2363BEGIN_FTR_SECTION
2364	beq	1f
2365#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2366  BEGIN_FTR_SECTION_NESTED(69)
2367	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2368	 * transaction), go do TM stuff
2369	 */
2370	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2371	bne-	2f
2372  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2373#endif
2374	b	load_up_vsx
2375#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
23762:	/* User process was in a transaction */
2377	addi	r3,r1,STACK_FRAME_OVERHEAD
2378	bl	vsx_unavailable_tm
2379	b	interrupt_return_srr
2380#endif
23811:
2382END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2383#endif
2384	addi	r3,r1,STACK_FRAME_OVERHEAD
2385	bl	vsx_unavailable_exception
2386	b	interrupt_return_srr
2387
2388
2389/**
2390 * Interrupt 0xf60 - Facility Unavailable Interrupt.
2391 * This is a synchronous interrupt in response to
2392 * executing an instruction without access to the facility that can be
2393 * resolved by the OS (e.g., FSCR, MSR).
2394 * Similar to FP unavailable.
2395 */
2396INT_DEFINE_BEGIN(facility_unavailable)
2397	IVEC=0xf60
2398#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2399	IKVM_REAL=1
2400#endif
2401INT_DEFINE_END(facility_unavailable)
2402
2403EXC_REAL_BEGIN(facility_unavailable, 0xf60, 0x20)
2404	GEN_INT_ENTRY facility_unavailable, virt=0, ool=1
2405EXC_REAL_END(facility_unavailable, 0xf60, 0x20)
2406EXC_VIRT_BEGIN(facility_unavailable, 0x4f60, 0x20)
2407	GEN_INT_ENTRY facility_unavailable, virt=1, ool=1
2408EXC_VIRT_END(facility_unavailable, 0x4f60, 0x20)
2409EXC_COMMON_BEGIN(facility_unavailable_common)
2410	GEN_COMMON facility_unavailable
2411	addi	r3,r1,STACK_FRAME_OVERHEAD
2412	bl	facility_unavailable_exception
2413	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2414	b	interrupt_return_srr
2415
2416
2417/**
2418 * Interrupt 0xf60 - Hypervisor Facility Unavailable Interrupt.
2419 * This is a synchronous interrupt in response to
2420 * executing an instruction without access to the facility that can only
2421 * be resolved in HV mode (e.g., HFSCR).
2422 * Similar to FP unavailable.
2423 */
2424INT_DEFINE_BEGIN(h_facility_unavailable)
2425	IVEC=0xf80
2426	IHSRR=1
2427	IKVM_REAL=1
2428	IKVM_VIRT=1
2429INT_DEFINE_END(h_facility_unavailable)
2430
2431EXC_REAL_BEGIN(h_facility_unavailable, 0xf80, 0x20)
2432	GEN_INT_ENTRY h_facility_unavailable, virt=0, ool=1
2433EXC_REAL_END(h_facility_unavailable, 0xf80, 0x20)
2434EXC_VIRT_BEGIN(h_facility_unavailable, 0x4f80, 0x20)
2435	GEN_INT_ENTRY h_facility_unavailable, virt=1, ool=1
2436EXC_VIRT_END(h_facility_unavailable, 0x4f80, 0x20)
2437EXC_COMMON_BEGIN(h_facility_unavailable_common)
2438	GEN_COMMON h_facility_unavailable
2439	addi	r3,r1,STACK_FRAME_OVERHEAD
2440	bl	facility_unavailable_exception
2441	REST_NVGPRS(r1) /* XXX Shouldn't be necessary in practice */
2442	b	interrupt_return_hsrr
2443
2444
2445EXC_REAL_NONE(0xfa0, 0x20)
2446EXC_VIRT_NONE(0x4fa0, 0x20)
2447EXC_REAL_NONE(0xfc0, 0x20)
2448EXC_VIRT_NONE(0x4fc0, 0x20)
2449EXC_REAL_NONE(0xfe0, 0x20)
2450EXC_VIRT_NONE(0x4fe0, 0x20)
2451
2452EXC_REAL_NONE(0x1000, 0x100)
2453EXC_VIRT_NONE(0x5000, 0x100)
2454EXC_REAL_NONE(0x1100, 0x100)
2455EXC_VIRT_NONE(0x5100, 0x100)
2456
2457#ifdef CONFIG_CBE_RAS
2458INT_DEFINE_BEGIN(cbe_system_error)
2459	IVEC=0x1200
2460	IHSRR=1
2461INT_DEFINE_END(cbe_system_error)
2462
2463EXC_REAL_BEGIN(cbe_system_error, 0x1200, 0x100)
2464	GEN_INT_ENTRY cbe_system_error, virt=0
2465EXC_REAL_END(cbe_system_error, 0x1200, 0x100)
2466EXC_VIRT_NONE(0x5200, 0x100)
2467EXC_COMMON_BEGIN(cbe_system_error_common)
2468	GEN_COMMON cbe_system_error
2469	addi	r3,r1,STACK_FRAME_OVERHEAD
2470	bl	cbe_system_error_exception
2471	b	interrupt_return_hsrr
2472
2473#else /* CONFIG_CBE_RAS */
2474EXC_REAL_NONE(0x1200, 0x100)
2475EXC_VIRT_NONE(0x5200, 0x100)
2476#endif
2477
2478/**
2479 * Interrupt 0x1300 - Instruction Address Breakpoint Interrupt.
2480 * This has been removed from the ISA before 2.01, which is the earliest
2481 * 64-bit BookS ISA supported, however the G5 / 970 implements this
2482 * interrupt with a non-architected feature available through the support
2483 * processor interface.
2484 */
2485INT_DEFINE_BEGIN(instruction_breakpoint)
2486	IVEC=0x1300
2487#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2488	IKVM_REAL=1
2489#endif
2490INT_DEFINE_END(instruction_breakpoint)
2491
2492EXC_REAL_BEGIN(instruction_breakpoint, 0x1300, 0x100)
2493	GEN_INT_ENTRY instruction_breakpoint, virt=0
2494EXC_REAL_END(instruction_breakpoint, 0x1300, 0x100)
2495EXC_VIRT_BEGIN(instruction_breakpoint, 0x5300, 0x100)
2496	GEN_INT_ENTRY instruction_breakpoint, virt=1
2497EXC_VIRT_END(instruction_breakpoint, 0x5300, 0x100)
2498EXC_COMMON_BEGIN(instruction_breakpoint_common)
2499	GEN_COMMON instruction_breakpoint
2500	addi	r3,r1,STACK_FRAME_OVERHEAD
2501	bl	instruction_breakpoint_exception
2502	b	interrupt_return_srr
2503
2504
2505EXC_REAL_NONE(0x1400, 0x100)
2506EXC_VIRT_NONE(0x5400, 0x100)
2507
2508/**
2509 * Interrupt 0x1500 - Soft Patch Interrupt
2510 *
2511 * Handling:
2512 * This is an implementation specific interrupt which can be used for a
2513 * range of exceptions.
2514 *
2515 * This interrupt handler is unique in that it runs the denormal assist
2516 * code even for guests (and even in guest context) without going to KVM,
2517 * for speed. POWER9 does not raise denorm exceptions, so this special case
2518 * could be phased out in future to reduce special cases.
2519 */
2520INT_DEFINE_BEGIN(denorm_exception)
2521	IVEC=0x1500
2522	IHSRR=1
2523	IBRANCH_TO_COMMON=0
2524	IKVM_REAL=1
2525INT_DEFINE_END(denorm_exception)
2526
2527EXC_REAL_BEGIN(denorm_exception, 0x1500, 0x100)
2528	GEN_INT_ENTRY denorm_exception, virt=0
2529#ifdef CONFIG_PPC_DENORMALISATION
2530	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2531	bne+	denorm_assist
2532#endif
2533	GEN_BRANCH_TO_COMMON denorm_exception, virt=0
2534EXC_REAL_END(denorm_exception, 0x1500, 0x100)
2535#ifdef CONFIG_PPC_DENORMALISATION
2536EXC_VIRT_BEGIN(denorm_exception, 0x5500, 0x100)
2537	GEN_INT_ENTRY denorm_exception, virt=1
2538	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2539	bne+	denorm_assist
2540	GEN_BRANCH_TO_COMMON denorm_exception, virt=1
2541EXC_VIRT_END(denorm_exception, 0x5500, 0x100)
2542#else
2543EXC_VIRT_NONE(0x5500, 0x100)
2544#endif
2545
2546#ifdef CONFIG_PPC_DENORMALISATION
2547TRAMP_REAL_BEGIN(denorm_assist)
2548BEGIN_FTR_SECTION
2549/*
2550 * To denormalise we need to move a copy of the register to itself.
2551 * For POWER6 do that here for all FP regs.
2552 */
2553	mfmsr	r10
2554	ori	r10,r10,(MSR_FP|MSR_FE0|MSR_FE1)
2555	xori	r10,r10,(MSR_FE0|MSR_FE1)
2556	mtmsrd	r10
2557	sync
2558
2559	.Lreg=0
2560	.rept 32
2561	fmr	.Lreg,.Lreg
2562	.Lreg=.Lreg+1
2563	.endr
2564
2565FTR_SECTION_ELSE
2566/*
2567 * To denormalise we need to move a copy of the register to itself.
2568 * For POWER7 do that here for the first 32 VSX registers only.
2569 */
2570	mfmsr	r10
2571	oris	r10,r10,MSR_VSX@h
2572	mtmsrd	r10
2573	sync
2574
2575	.Lreg=0
2576	.rept 32
2577	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2578	.Lreg=.Lreg+1
2579	.endr
2580
2581ALT_FTR_SECTION_END_IFCLR(CPU_FTR_ARCH_206)
2582
2583BEGIN_FTR_SECTION
2584	b	denorm_done
2585END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
2586/*
2587 * To denormalise we need to move a copy of the register to itself.
2588 * For POWER8 we need to do that for all 64 VSX registers
2589 */
2590	.Lreg=32
2591	.rept 32
2592	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2593	.Lreg=.Lreg+1
2594	.endr
2595
2596denorm_done:
2597	mfspr	r11,SPRN_HSRR0
2598	subi	r11,r11,4
2599	mtspr	SPRN_HSRR0,r11
2600	mtcrf	0x80,r9
2601	ld	r9,PACA_EXGEN+EX_R9(r13)
2602BEGIN_FTR_SECTION
2603	ld	r10,PACA_EXGEN+EX_PPR(r13)
2604	mtspr	SPRN_PPR,r10
2605END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
2606BEGIN_FTR_SECTION
2607	ld	r10,PACA_EXGEN+EX_CFAR(r13)
2608	mtspr	SPRN_CFAR,r10
2609END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
2610	li	r10,0
2611	stb	r10,PACAHSRR_VALID(r13)
2612	ld	r10,PACA_EXGEN+EX_R10(r13)
2613	ld	r11,PACA_EXGEN+EX_R11(r13)
2614	ld	r12,PACA_EXGEN+EX_R12(r13)
2615	ld	r13,PACA_EXGEN+EX_R13(r13)
2616	HRFI_TO_UNKNOWN
2617	b	.
2618#endif
2619
2620EXC_COMMON_BEGIN(denorm_exception_common)
2621	GEN_COMMON denorm_exception
2622	addi	r3,r1,STACK_FRAME_OVERHEAD
2623	bl	unknown_exception
2624	b	interrupt_return_hsrr
2625
2626
2627#ifdef CONFIG_CBE_RAS
2628INT_DEFINE_BEGIN(cbe_maintenance)
2629	IVEC=0x1600
2630	IHSRR=1
2631INT_DEFINE_END(cbe_maintenance)
2632
2633EXC_REAL_BEGIN(cbe_maintenance, 0x1600, 0x100)
2634	GEN_INT_ENTRY cbe_maintenance, virt=0
2635EXC_REAL_END(cbe_maintenance, 0x1600, 0x100)
2636EXC_VIRT_NONE(0x5600, 0x100)
2637EXC_COMMON_BEGIN(cbe_maintenance_common)
2638	GEN_COMMON cbe_maintenance
2639	addi	r3,r1,STACK_FRAME_OVERHEAD
2640	bl	cbe_maintenance_exception
2641	b	interrupt_return_hsrr
2642
2643#else /* CONFIG_CBE_RAS */
2644EXC_REAL_NONE(0x1600, 0x100)
2645EXC_VIRT_NONE(0x5600, 0x100)
2646#endif
2647
2648
2649INT_DEFINE_BEGIN(altivec_assist)
2650	IVEC=0x1700
2651#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2652	IKVM_REAL=1
2653#endif
2654INT_DEFINE_END(altivec_assist)
2655
2656EXC_REAL_BEGIN(altivec_assist, 0x1700, 0x100)
2657	GEN_INT_ENTRY altivec_assist, virt=0
2658EXC_REAL_END(altivec_assist, 0x1700, 0x100)
2659EXC_VIRT_BEGIN(altivec_assist, 0x5700, 0x100)
2660	GEN_INT_ENTRY altivec_assist, virt=1
2661EXC_VIRT_END(altivec_assist, 0x5700, 0x100)
2662EXC_COMMON_BEGIN(altivec_assist_common)
2663	GEN_COMMON altivec_assist
2664	addi	r3,r1,STACK_FRAME_OVERHEAD
2665#ifdef CONFIG_ALTIVEC
2666	bl	altivec_assist_exception
2667	REST_NVGPRS(r1) /* instruction emulation may change GPRs */
2668#else
2669	bl	unknown_exception
2670#endif
2671	b	interrupt_return_srr
2672
2673
2674#ifdef CONFIG_CBE_RAS
2675INT_DEFINE_BEGIN(cbe_thermal)
2676	IVEC=0x1800
2677	IHSRR=1
2678INT_DEFINE_END(cbe_thermal)
2679
2680EXC_REAL_BEGIN(cbe_thermal, 0x1800, 0x100)
2681	GEN_INT_ENTRY cbe_thermal, virt=0
2682EXC_REAL_END(cbe_thermal, 0x1800, 0x100)
2683EXC_VIRT_NONE(0x5800, 0x100)
2684EXC_COMMON_BEGIN(cbe_thermal_common)
2685	GEN_COMMON cbe_thermal
2686	addi	r3,r1,STACK_FRAME_OVERHEAD
2687	bl	cbe_thermal_exception
2688	b	interrupt_return_hsrr
2689
2690#else /* CONFIG_CBE_RAS */
2691EXC_REAL_NONE(0x1800, 0x100)
2692EXC_VIRT_NONE(0x5800, 0x100)
2693#endif
2694
2695
2696#ifdef CONFIG_PPC_WATCHDOG
2697
2698INT_DEFINE_BEGIN(soft_nmi)
2699	IVEC=0x900
2700	ISTACK=0
2701INT_DEFINE_END(soft_nmi)
2702
2703/*
2704 * Branch to soft_nmi_interrupt using the emergency stack. The emergency
2705 * stack is one that is usable by maskable interrupts so long as MSR_EE
2706 * remains off. It is used for recovery when something has corrupted the
2707 * normal kernel stack, for example. The "soft NMI" must not use the process
2708 * stack because we want irq disabled sections to avoid touching the stack
2709 * at all (other than PMU interrupts), so use the emergency stack for this,
2710 * and run it entirely with interrupts hard disabled.
2711 */
2712EXC_COMMON_BEGIN(soft_nmi_common)
2713	mr	r10,r1
2714	ld	r1,PACAEMERGSP(r13)
2715	subi	r1,r1,INT_FRAME_SIZE
2716	__GEN_COMMON_BODY soft_nmi
2717
2718	addi	r3,r1,STACK_FRAME_OVERHEAD
2719	bl	soft_nmi_interrupt
2720
2721	/* Clear MSR_RI before setting SRR0 and SRR1. */
2722	li	r9,0
2723	mtmsrd	r9,1
2724
2725	kuap_kernel_restore r9, r10
2726
2727	EXCEPTION_RESTORE_REGS hsrr=0
2728	RFI_TO_KERNEL
2729
2730#endif /* CONFIG_PPC_WATCHDOG */
2731
2732/*
2733 * An interrupt came in while soft-disabled. We set paca->irq_happened, then:
2734 * - If it was a decrementer interrupt, we bump the dec to max and and return.
2735 * - If it was a doorbell we return immediately since doorbells are edge
2736 *   triggered and won't automatically refire.
2737 * - If it was a HMI we return immediately since we handled it in realmode
2738 *   and it won't refire.
2739 * - Else it is one of PACA_IRQ_MUST_HARD_MASK, so hard disable and return.
2740 * This is called with r10 containing the value to OR to the paca field.
2741 */
2742.macro MASKED_INTERRUPT hsrr=0
2743	.if \hsrr
2744masked_Hinterrupt:
2745	.else
2746masked_interrupt:
2747	.endif
2748	stw	r9,PACA_EXGEN+EX_CCR(r13)
2749	lbz	r9,PACAIRQHAPPENED(r13)
2750	or	r9,r9,r10
2751	stb	r9,PACAIRQHAPPENED(r13)
2752
2753	.if ! \hsrr
2754	cmpwi	r10,PACA_IRQ_DEC
2755	bne	1f
2756	LOAD_REG_IMMEDIATE(r9, 0x7fffffff)
2757	mtspr	SPRN_DEC,r9
2758#ifdef CONFIG_PPC_WATCHDOG
2759	lwz	r9,PACA_EXGEN+EX_CCR(r13)
2760	b	soft_nmi_common
2761#else
2762	b	2f
2763#endif
2764	.endif
2765
27661:	andi.	r10,r10,PACA_IRQ_MUST_HARD_MASK
2767	beq	2f
2768	xori	r12,r12,MSR_EE	/* clear MSR_EE */
2769	.if \hsrr
2770	mtspr	SPRN_HSRR1,r12
2771	.else
2772	mtspr	SPRN_SRR1,r12
2773	.endif
2774	ori	r9,r9,PACA_IRQ_HARD_DIS
2775	stb	r9,PACAIRQHAPPENED(r13)
27762:	/* done */
2777	li	r9,0
2778	.if \hsrr
2779	stb	r9,PACAHSRR_VALID(r13)
2780	.else
2781	stb	r9,PACASRR_VALID(r13)
2782	.endif
2783
2784	SEARCH_RESTART_TABLE
2785	cmpdi	r12,0
2786	beq	3f
2787	.if \hsrr
2788	mtspr	SPRN_HSRR0,r12
2789	.else
2790	mtspr	SPRN_SRR0,r12
2791	.endif
27923:
2793
2794	ld	r9,PACA_EXGEN+EX_CTR(r13)
2795	mtctr	r9
2796	lwz	r9,PACA_EXGEN+EX_CCR(r13)
2797	mtcrf	0x80,r9
2798	std	r1,PACAR1(r13)
2799	ld	r9,PACA_EXGEN+EX_R9(r13)
2800	ld	r10,PACA_EXGEN+EX_R10(r13)
2801	ld	r11,PACA_EXGEN+EX_R11(r13)
2802	ld	r12,PACA_EXGEN+EX_R12(r13)
2803	ld	r13,PACA_EXGEN+EX_R13(r13)
2804	/* May return to masked low address where r13 is not set up */
2805	.if \hsrr
2806	HRFI_TO_KERNEL
2807	.else
2808	RFI_TO_KERNEL
2809	.endif
2810	b	.
2811.endm
2812
2813TRAMP_REAL_BEGIN(stf_barrier_fallback)
2814	std	r9,PACA_EXRFI+EX_R9(r13)
2815	std	r10,PACA_EXRFI+EX_R10(r13)
2816	sync
2817	ld	r9,PACA_EXRFI+EX_R9(r13)
2818	ld	r10,PACA_EXRFI+EX_R10(r13)
2819	ori	31,31,0
2820	.rept 14
2821	b	1f
28221:
2823	.endr
2824	blr
2825
2826/* Clobbers r10, r11, ctr */
2827.macro L1D_DISPLACEMENT_FLUSH
2828	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
2829	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
2830	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
2831	mtctr	r11
2832	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
2833
2834	/* order ld/st prior to dcbt stop all streams with flushing */
2835	sync
2836
2837	/*
2838	 * The load addresses are at staggered offsets within cachelines,
2839	 * which suits some pipelines better (on others it should not
2840	 * hurt).
2841	 */
28421:
2843	ld	r11,(0x80 + 8)*0(r10)
2844	ld	r11,(0x80 + 8)*1(r10)
2845	ld	r11,(0x80 + 8)*2(r10)
2846	ld	r11,(0x80 + 8)*3(r10)
2847	ld	r11,(0x80 + 8)*4(r10)
2848	ld	r11,(0x80 + 8)*5(r10)
2849	ld	r11,(0x80 + 8)*6(r10)
2850	ld	r11,(0x80 + 8)*7(r10)
2851	addi	r10,r10,0x80*8
2852	bdnz	1b
2853.endm
2854
2855TRAMP_REAL_BEGIN(entry_flush_fallback)
2856	std	r9,PACA_EXRFI+EX_R9(r13)
2857	std	r10,PACA_EXRFI+EX_R10(r13)
2858	std	r11,PACA_EXRFI+EX_R11(r13)
2859	mfctr	r9
2860	L1D_DISPLACEMENT_FLUSH
2861	mtctr	r9
2862	ld	r9,PACA_EXRFI+EX_R9(r13)
2863	ld	r10,PACA_EXRFI+EX_R10(r13)
2864	ld	r11,PACA_EXRFI+EX_R11(r13)
2865	blr
2866
2867/*
2868 * The SCV entry flush happens with interrupts enabled, so it must disable
2869 * to prevent EXRFI being clobbered by NMIs (e.g., soft_nmi_common). r10
2870 * (containing LR) does not need to be preserved here because scv entry
2871 * puts 0 in the pt_regs, CTR can be clobbered for the same reason.
2872 */
2873TRAMP_REAL_BEGIN(scv_entry_flush_fallback)
2874	li	r10,0
2875	mtmsrd	r10,1
2876	lbz	r10,PACAIRQHAPPENED(r13)
2877	ori	r10,r10,PACA_IRQ_HARD_DIS
2878	stb	r10,PACAIRQHAPPENED(r13)
2879	std	r11,PACA_EXRFI+EX_R11(r13)
2880	L1D_DISPLACEMENT_FLUSH
2881	ld	r11,PACA_EXRFI+EX_R11(r13)
2882	li	r10,MSR_RI
2883	mtmsrd	r10,1
2884	blr
2885
2886TRAMP_REAL_BEGIN(rfi_flush_fallback)
2887	SET_SCRATCH0(r13);
2888	GET_PACA(r13);
2889	std	r1,PACA_EXRFI+EX_R12(r13)
2890	ld	r1,PACAKSAVE(r13)
2891	std	r9,PACA_EXRFI+EX_R9(r13)
2892	std	r10,PACA_EXRFI+EX_R10(r13)
2893	std	r11,PACA_EXRFI+EX_R11(r13)
2894	mfctr	r9
2895	L1D_DISPLACEMENT_FLUSH
2896	mtctr	r9
2897	ld	r9,PACA_EXRFI+EX_R9(r13)
2898	ld	r10,PACA_EXRFI+EX_R10(r13)
2899	ld	r11,PACA_EXRFI+EX_R11(r13)
2900	ld	r1,PACA_EXRFI+EX_R12(r13)
2901	GET_SCRATCH0(r13);
2902	rfid
2903
2904TRAMP_REAL_BEGIN(hrfi_flush_fallback)
2905	SET_SCRATCH0(r13);
2906	GET_PACA(r13);
2907	std	r1,PACA_EXRFI+EX_R12(r13)
2908	ld	r1,PACAKSAVE(r13)
2909	std	r9,PACA_EXRFI+EX_R9(r13)
2910	std	r10,PACA_EXRFI+EX_R10(r13)
2911	std	r11,PACA_EXRFI+EX_R11(r13)
2912	mfctr	r9
2913	L1D_DISPLACEMENT_FLUSH
2914	mtctr	r9
2915	ld	r9,PACA_EXRFI+EX_R9(r13)
2916	ld	r10,PACA_EXRFI+EX_R10(r13)
2917	ld	r11,PACA_EXRFI+EX_R11(r13)
2918	ld	r1,PACA_EXRFI+EX_R12(r13)
2919	GET_SCRATCH0(r13);
2920	hrfid
2921
2922TRAMP_REAL_BEGIN(rfscv_flush_fallback)
2923	/* system call volatile */
2924	mr	r7,r13
2925	GET_PACA(r13);
2926	mr	r8,r1
2927	ld	r1,PACAKSAVE(r13)
2928	mfctr	r9
2929	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
2930	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
2931	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
2932	mtctr	r11
2933	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
2934
2935	/* order ld/st prior to dcbt stop all streams with flushing */
2936	sync
2937
2938	/*
2939	 * The load adresses are at staggered offsets within cachelines,
2940	 * which suits some pipelines better (on others it should not
2941	 * hurt).
2942	 */
29431:
2944	ld	r11,(0x80 + 8)*0(r10)
2945	ld	r11,(0x80 + 8)*1(r10)
2946	ld	r11,(0x80 + 8)*2(r10)
2947	ld	r11,(0x80 + 8)*3(r10)
2948	ld	r11,(0x80 + 8)*4(r10)
2949	ld	r11,(0x80 + 8)*5(r10)
2950	ld	r11,(0x80 + 8)*6(r10)
2951	ld	r11,(0x80 + 8)*7(r10)
2952	addi	r10,r10,0x80*8
2953	bdnz	1b
2954
2955	mtctr	r9
2956	li	r9,0
2957	li	r10,0
2958	li	r11,0
2959	mr	r1,r8
2960	mr	r13,r7
2961	RFSCV
2962
2963USE_TEXT_SECTION()
2964
2965#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
2966kvm_interrupt:
2967	/*
2968	 * The conditional branch in KVMTEST can't reach all the way,
2969	 * make a stub.
2970	 */
2971	b	kvmppc_interrupt
2972#endif
2973
2974_GLOBAL(do_uaccess_flush)
2975	UACCESS_FLUSH_FIXUP_SECTION
2976	nop
2977	nop
2978	nop
2979	blr
2980	L1D_DISPLACEMENT_FLUSH
2981	blr
2982_ASM_NOKPROBE_SYMBOL(do_uaccess_flush)
2983EXPORT_SYMBOL(do_uaccess_flush)
2984
2985
2986MASKED_INTERRUPT
2987MASKED_INTERRUPT hsrr=1
2988
2989	/*
2990	 * Relocation-on interrupts: A subset of the interrupts can be delivered
2991	 * with IR=1/DR=1, if AIL==2 and MSR.HV won't be changed by delivering
2992	 * it.  Addresses are the same as the original interrupt addresses, but
2993	 * offset by 0xc000000000004000.
2994	 * It's impossible to receive interrupts below 0x300 via this mechanism.
2995	 * KVM: None of these traps are from the guest ; anything that escalated
2996	 * to HV=1 from HV=0 is delivered via real mode handlers.
2997	 */
2998
2999	/*
3000	 * This uses the standard macro, since the original 0x300 vector
3001	 * only has extra guff for STAB-based processors -- which never
3002	 * come here.
3003	 */
3004
3005USE_FIXED_SECTION(virt_trampolines)
3006	/*
3007	 * All code below __end_soft_masked is treated as soft-masked. If
3008	 * any code runs here with MSR[EE]=1, it must then cope with pending
3009	 * soft interrupt being raised (i.e., by ensuring it is replayed).
3010	 *
3011	 * The __end_interrupts marker must be past the out-of-line (OOL)
3012	 * handlers, so that they are copied to real address 0x100 when running
3013	 * a relocatable kernel. This ensures they can be reached from the short
3014	 * trampoline handlers (like 0x4f00, 0x4f20, etc.) which branch
3015	 * directly, without using LOAD_HANDLER().
3016	 */
3017	.align	7
3018	.globl	__end_interrupts
3019__end_interrupts:
3020DEFINE_FIXED_SYMBOL(__end_interrupts)
3021
3022CLOSE_FIXED_SECTION(real_vectors);
3023CLOSE_FIXED_SECTION(real_trampolines);
3024CLOSE_FIXED_SECTION(virt_vectors);
3025CLOSE_FIXED_SECTION(virt_trampolines);
3026
3027USE_TEXT_SECTION()
3028
3029/* MSR[RI] should be clear because this uses SRR[01] */
3030enable_machine_check:
3031	mflr	r0
3032	bcl	20,31,$+4
30330:	mflr	r3
3034	addi	r3,r3,(1f - 0b)
3035	mtspr	SPRN_SRR0,r3
3036	mfmsr	r3
3037	ori	r3,r3,MSR_ME
3038	mtspr	SPRN_SRR1,r3
3039	RFI_TO_KERNEL
30401:	mtlr	r0
3041	blr
3042
3043/* MSR[RI] should be clear because this uses SRR[01] */
3044disable_machine_check:
3045	mflr	r0
3046	bcl	20,31,$+4
30470:	mflr	r3
3048	addi	r3,r3,(1f - 0b)
3049	mtspr	SPRN_SRR0,r3
3050	mfmsr	r3
3051	li	r4,MSR_ME
3052	andc	r3,r3,r4
3053	mtspr	SPRN_SRR1,r3
3054	RFI_TO_KERNEL
30551:	mtlr	r0
3056	blr
3057