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 <linux/linkage.h>
  17#include <asm/hw_irq.h>
  18#include <asm/exception-64s.h>
  19#include <asm/ptrace.h>
  20#include <asm/cpuidle.h>
  21#include <asm/head-64.h>
  22#include <asm/feature-fixups.h>
  23#include <asm/kup.h>
  24
  25/*
  26 * Following are fixed section helper macros.
  27 *
  28 * EXC_REAL_BEGIN/END  - real, unrelocated exception vectors
  29 * EXC_VIRT_BEGIN/END  - virt (AIL), unrelocated exception vectors
  30 * TRAMP_REAL_BEGIN    - real, unrelocated helpers (virt may call these)
  31 * TRAMP_VIRT_BEGIN    - virt, unreloc helpers (in practice, real can use)
  32 * EXC_COMMON          - After switching to virtual, relocated mode.
  33 */
  34
  35#define EXC_REAL_BEGIN(name, start, size)			\
  36	FIXED_SECTION_ENTRY_BEGIN_LOCATION(real_vectors, exc_real_##start##_##name, start, size)
  37
  38#define EXC_REAL_END(name, start, size)				\
  39	FIXED_SECTION_ENTRY_END_LOCATION(real_vectors, exc_real_##start##_##name, start, size)
  40
  41#define EXC_VIRT_BEGIN(name, start, size)			\
  42	FIXED_SECTION_ENTRY_BEGIN_LOCATION(virt_vectors, exc_virt_##start##_##name, start, size)
  43
  44#define EXC_VIRT_END(name, start, size)				\
  45	FIXED_SECTION_ENTRY_END_LOCATION(virt_vectors, exc_virt_##start##_##name, start, size)
  46
  47#define EXC_COMMON_BEGIN(name)					\
  48	USE_TEXT_SECTION();					\
  49	.balign IFETCH_ALIGN_BYTES;				\
  50	.global name;						\
  51	_ASM_NOKPROBE_SYMBOL(name);				\
  52	DEFINE_FIXED_SYMBOL(name, text);			\
  53name:
  54
  55#define TRAMP_REAL_BEGIN(name)					\
  56	FIXED_SECTION_ENTRY_BEGIN(real_trampolines, name)
  57
  58#define TRAMP_VIRT_BEGIN(name)					\
  59	FIXED_SECTION_ENTRY_BEGIN(virt_trampolines, name)
  60
  61#define EXC_REAL_NONE(start, size)				\
  62	FIXED_SECTION_ENTRY_BEGIN_LOCATION(real_vectors, exc_real_##start##_##unused, start, size); \
  63	FIXED_SECTION_ENTRY_END_LOCATION(real_vectors, exc_real_##start##_##unused, start, size)
  64
  65#define EXC_VIRT_NONE(start, size)				\
  66	FIXED_SECTION_ENTRY_BEGIN_LOCATION(virt_vectors, exc_virt_##start##_##unused, start, size); \
  67	FIXED_SECTION_ENTRY_END_LOCATION(virt_vectors, exc_virt_##start##_##unused, start, size)
  68
  69/*
  70 * We're short on space and time in the exception prolog, so we can't
  71 * use the normal LOAD_REG_IMMEDIATE macro to load the address of label.
  72 * Instead we get the base of the kernel from paca->kernelbase and or in the low
  73 * part of label. This requires that the label be within 64KB of kernelbase, and
  74 * that kernelbase be 64K aligned.
  75 */
  76#define LOAD_HANDLER(reg, label)					\
  77	ld	reg,PACAKBASE(r13);	/* get high part of &label */	\
  78	ori	reg,reg,FIXED_SYMBOL_ABS_ADDR(label)
  79
  80#define __LOAD_HANDLER(reg, label, section)					\
  81	ld	reg,PACAKBASE(r13);					\
  82	ori	reg,reg,(ABS_ADDR(label, section))@l
  83
  84/*
  85 * Branches from unrelocated code (e.g., interrupts) to labels outside
  86 * head-y require >64K offsets.
  87 */
  88#define __LOAD_FAR_HANDLER(reg, label, section)					\
  89	ld	reg,PACAKBASE(r13);					\
  90	ori	reg,reg,(ABS_ADDR(label, section))@l;				\
  91	addis	reg,reg,(ABS_ADDR(label, section))@h
  92
  93/*
  94 * Interrupt code generation macros
  95 */
  96#define IVEC		.L_IVEC_\name\()	/* Interrupt vector address */
  97#define IHSRR		.L_IHSRR_\name\()	/* Sets SRR or HSRR registers */
  98#define IHSRR_IF_HVMODE	.L_IHSRR_IF_HVMODE_\name\() /* HSRR if HV else SRR */
  99#define IAREA		.L_IAREA_\name\()	/* PACA save area */
 100#define IVIRT		.L_IVIRT_\name\()	/* Has virt mode entry point */
 101#define IISIDE		.L_IISIDE_\name\()	/* Uses SRR0/1 not DAR/DSISR */
 102#define ICFAR		.L_ICFAR_\name\()	/* Uses CFAR */
 103#define ICFAR_IF_HVMODE	.L_ICFAR_IF_HVMODE_\name\() /* Uses CFAR if HV */
 104#define IDAR		.L_IDAR_\name\()	/* Uses DAR (or SRR0) */
 105#define IDSISR		.L_IDSISR_\name\()	/* Uses DSISR (or SRR1) */
 106#define IBRANCH_TO_COMMON	.L_IBRANCH_TO_COMMON_\name\() /* ENTRY branch to common */
 107#define IREALMODE_COMMON	.L_IREALMODE_COMMON_\name\() /* Common runs in realmode */
 108#define IMASK		.L_IMASK_\name\()	/* IRQ soft-mask bit */
 109#define IKVM_REAL	.L_IKVM_REAL_\name\()	/* Real entry tests KVM */
 110#define __IKVM_REAL(name)	.L_IKVM_REAL_ ## name
 111#define IKVM_VIRT	.L_IKVM_VIRT_\name\()	/* Virt entry tests KVM */
 112#define ISTACK		.L_ISTACK_\name\()	/* Set regular kernel stack */
 113#define __ISTACK(name)	.L_ISTACK_ ## name
 114#define IKUAP		.L_IKUAP_\name\()	/* Do KUAP lock */
 115#define IMSR_R12	.L_IMSR_R12_\name\()	/* Assumes MSR saved to r12 */
 116
 117#define INT_DEFINE_BEGIN(n)						\
 118.macro int_define_ ## n name
 119
 120#define INT_DEFINE_END(n)						\
 121.endm ;									\
 122int_define_ ## n n ;							\
 123do_define_int n
 124
 125.macro do_define_int name
 126	.ifndef IVEC
 127		.error "IVEC not defined"
 128	.endif
 129	.ifndef IHSRR
 130		IHSRR=0
 131	.endif
 132	.ifndef IHSRR_IF_HVMODE
 133		IHSRR_IF_HVMODE=0
 134	.endif
 135	.ifndef IAREA
 136		IAREA=PACA_EXGEN
 137	.endif
 138	.ifndef IVIRT
 139		IVIRT=1
 140	.endif
 141	.ifndef IISIDE
 142		IISIDE=0
 143	.endif
 144	.ifndef ICFAR
 145		ICFAR=1
 146	.endif
 147	.ifndef ICFAR_IF_HVMODE
 148		ICFAR_IF_HVMODE=0
 149	.endif
 150	.ifndef IDAR
 151		IDAR=0
 152	.endif
 153	.ifndef IDSISR
 154		IDSISR=0
 155	.endif
 156	.ifndef IBRANCH_TO_COMMON
 157		IBRANCH_TO_COMMON=1
 158	.endif
 159	.ifndef IREALMODE_COMMON
 160		IREALMODE_COMMON=0
 161	.else
 162		.if ! IBRANCH_TO_COMMON
 163			.error "IREALMODE_COMMON=1 but IBRANCH_TO_COMMON=0"
 164		.endif
 165	.endif
 166	.ifndef IMASK
 167		IMASK=0
 168	.endif
 169	.ifndef IKVM_REAL
 170		IKVM_REAL=0
 171	.endif
 172	.ifndef IKVM_VIRT
 173		IKVM_VIRT=0
 174	.endif
 175	.ifndef ISTACK
 176		ISTACK=1
 177	.endif
 178	.ifndef IKUAP
 179		IKUAP=1
 180	.endif
 181	.ifndef IMSR_R12
 182		IMSR_R12=0
 183	.endif
 184.endm
 185
 186/*
 187 * All interrupts which set HSRR registers, as well as SRESET and MCE and
 188 * syscall when invoked with "sc 1" switch to MSR[HV]=1 (HVMODE) to be taken,
 189 * so they all generally need to test whether they were taken in guest context.
 190 *
 191 * Note: SRESET and MCE may also be sent to the guest by the hypervisor, and be
 192 * taken with MSR[HV]=0.
 193 *
 194 * Interrupts which set SRR registers (with the above exceptions) do not
 195 * elevate to MSR[HV]=1 mode, though most can be taken when running with
 196 * MSR[HV]=1  (e.g., bare metal kernel and userspace). So these interrupts do
 197 * not need to test whether a guest is running because they get delivered to
 198 * the guest directly, including nested HV KVM guests.
 199 *
 200 * The exception is PR KVM, where the guest runs with MSR[PR]=1 and the host
 201 * runs with MSR[HV]=0, so the host takes all interrupts on behalf of the
 202 * guest. PR KVM runs with LPCR[AIL]=0 which causes interrupts to always be
 203 * delivered to the real-mode entry point, therefore such interrupts only test
 204 * KVM in their real mode handlers, and only when PR KVM is possible.
 205 *
 206 * Interrupts that are taken in MSR[HV]=0 and escalate to MSR[HV]=1 are always
 207 * delivered in real-mode when the MMU is in hash mode because the MMU
 208 * registers are not set appropriately to translate host addresses. In nested
 209 * radix mode these can be delivered in virt-mode as the host translations are
 210 * used implicitly (see: effective LPID, effective PID).
 211 */
 212
 213/*
 214 * If an interrupt is taken while a guest is running, it is immediately routed
 215 * to KVM to handle.
 216 */
 217
 218.macro KVMTEST name handler
 219#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
 220	lbz	r10,HSTATE_IN_GUEST(r13)
 221	cmpwi	r10,0
 222	/* HSRR variants have the 0x2 bit added to their trap number */
 223	.if IHSRR_IF_HVMODE
 224	BEGIN_FTR_SECTION
 225	li	r10,(IVEC + 0x2)
 226	FTR_SECTION_ELSE
 227	li	r10,(IVEC)
 228	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
 229	.elseif IHSRR
 230	li	r10,(IVEC + 0x2)
 231	.else
 232	li	r10,(IVEC)
 233	.endif
 234	bne	\handler
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 235#endif
 236.endm
 237
 238/*
 239 * This is the BOOK3S interrupt entry code macro.
 240 *
 241 * This can result in one of several things happening:
 242 * - Branch to the _common handler, relocated, in virtual mode.
 243 *   These are normal interrupts (synchronous and asynchronous) handled by
 244 *   the kernel.
 245 * - Branch to KVM, relocated but real mode interrupts remain in real mode.
 246 *   These occur when HSTATE_IN_GUEST is set. The interrupt may be caused by
 247 *   / intended for host or guest kernel, but KVM must always be involved
 248 *   because the machine state is set for guest execution.
 249 * - Branch to the masked handler, unrelocated.
 250 *   These occur when maskable asynchronous interrupts are taken with the
 251 *   irq_soft_mask set.
 252 * - Branch to an "early" handler in real mode but relocated.
 253 *   This is done if early=1. MCE and HMI use these to handle errors in real
 254 *   mode.
 255 * - Fall through and continue executing in real, unrelocated mode.
 256 *   This is done if early=2.
 257 */
 
 
 
 258
 259.macro GEN_BRANCH_TO_COMMON name, virt
 260	.if IREALMODE_COMMON
 261	LOAD_HANDLER(r10, \name\()_common)
 262	mtctr	r10
 263	bctr
 264	.else
 265	.if \virt
 266#ifndef CONFIG_RELOCATABLE
 267	b	\name\()_common_virt
 268#else
 269	LOAD_HANDLER(r10, \name\()_common_virt)
 270	mtctr	r10
 271	bctr
 272#endif
 273	.else
 274	LOAD_HANDLER(r10, \name\()_common_real)
 275	mtctr	r10
 276	bctr
 277	.endif
 278	.endif
 279.endm
 280
 281.macro GEN_INT_ENTRY name, virt, ool=0
 282	SET_SCRATCH0(r13)			/* save r13 */
 283	GET_PACA(r13)
 284	std	r9,IAREA+EX_R9(r13)		/* save r9 */
 285BEGIN_FTR_SECTION
 286	mfspr	r9,SPRN_PPR
 287END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
 288	HMT_MEDIUM
 289	std	r10,IAREA+EX_R10(r13)		/* save r10 */
 290	.if ICFAR
 291BEGIN_FTR_SECTION
 292	mfspr	r10,SPRN_CFAR
 293END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
 294	.elseif ICFAR_IF_HVMODE
 295BEGIN_FTR_SECTION
 296  BEGIN_FTR_SECTION_NESTED(69)
 297	mfspr	r10,SPRN_CFAR
 298  END_FTR_SECTION_NESTED(CPU_FTR_CFAR, CPU_FTR_CFAR, 69)
 299FTR_SECTION_ELSE
 300  BEGIN_FTR_SECTION_NESTED(69)
 301	li	r10,0
 302  END_FTR_SECTION_NESTED(CPU_FTR_CFAR, CPU_FTR_CFAR, 69)
 303ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
 304	.endif
 305	.if \ool
 306	.if !\virt
 307	b	tramp_real_\name
 308	.pushsection .text
 309	TRAMP_REAL_BEGIN(tramp_real_\name)
 310	.else
 311	b	tramp_virt_\name
 312	.pushsection .text
 313	TRAMP_VIRT_BEGIN(tramp_virt_\name)
 314	.endif
 315	.endif
 316
 317BEGIN_FTR_SECTION
 318	std	r9,IAREA+EX_PPR(r13)
 319END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
 320	.if ICFAR || ICFAR_IF_HVMODE
 321BEGIN_FTR_SECTION
 322	std	r10,IAREA+EX_CFAR(r13)
 323END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
 324	.endif
 325	INTERRUPT_TO_KERNEL
 326	mfctr	r10
 327	std	r10,IAREA+EX_CTR(r13)
 328	mfcr	r9
 329	std	r11,IAREA+EX_R11(r13)		/* save r11 - r12 */
 330	std	r12,IAREA+EX_R12(r13)
 331
 332	/*
 333	 * DAR/DSISR, SCRATCH0 must be read before setting MSR[RI],
 334	 * because a d-side MCE will clobber those registers so is
 335	 * not recoverable if they are live.
 336	 */
 337	GET_SCRATCH0(r10)
 338	std	r10,IAREA+EX_R13(r13)
 339	.if IDAR && !IISIDE
 340	.if IHSRR
 341	mfspr	r10,SPRN_HDAR
 342	.else
 343	mfspr	r10,SPRN_DAR
 344	.endif
 345	std	r10,IAREA+EX_DAR(r13)
 346	.endif
 347	.if IDSISR && !IISIDE
 348	.if IHSRR
 349	mfspr	r10,SPRN_HDSISR
 350	.else
 351	mfspr	r10,SPRN_DSISR
 352	.endif
 353	stw	r10,IAREA+EX_DSISR(r13)
 354	.endif
 355
 356	.if IHSRR_IF_HVMODE
 357	BEGIN_FTR_SECTION
 358	mfspr	r11,SPRN_HSRR0		/* save HSRR0 */
 359	mfspr	r12,SPRN_HSRR1		/* and HSRR1 */
 360	FTR_SECTION_ELSE
 361	mfspr	r11,SPRN_SRR0		/* save SRR0 */
 362	mfspr	r12,SPRN_SRR1		/* and SRR1 */
 363	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
 364	.elseif IHSRR
 365	mfspr	r11,SPRN_HSRR0		/* save HSRR0 */
 366	mfspr	r12,SPRN_HSRR1		/* and HSRR1 */
 367	.else
 368	mfspr	r11,SPRN_SRR0		/* save SRR0 */
 369	mfspr	r12,SPRN_SRR1		/* and SRR1 */
 370	.endif
 371
 372	.if IBRANCH_TO_COMMON
 373	GEN_BRANCH_TO_COMMON \name \virt
 374	.endif
 375
 376	.if \ool
 377	.popsection
 378	.endif
 379.endm
 380
 381/*
 382 * __GEN_COMMON_ENTRY is required to receive the branch from interrupt
 383 * entry, except in the case of the real-mode handlers which require
 384 * __GEN_REALMODE_COMMON_ENTRY.
 385 *
 386 * This switches to virtual mode and sets MSR[RI].
 387 */
 388.macro __GEN_COMMON_ENTRY name
 389DEFINE_FIXED_SYMBOL(\name\()_common_real, text)
 390\name\()_common_real:
 391	.if IKVM_REAL
 392		KVMTEST \name kvm_interrupt
 393	.endif
 394
 395	ld	r10,PACAKMSR(r13)	/* get MSR value for kernel */
 396	/* MSR[RI] is clear iff using SRR regs */
 397	.if IHSRR_IF_HVMODE
 398	BEGIN_FTR_SECTION
 399	xori	r10,r10,MSR_RI
 400	END_FTR_SECTION_IFCLR(CPU_FTR_HVMODE)
 401	.elseif ! IHSRR
 402	xori	r10,r10,MSR_RI
 403	.endif
 404	mtmsrd	r10
 405
 406	.if IVIRT
 407	.if IKVM_VIRT
 408	b	1f /* skip the virt test coming from real */
 409	.endif
 410
 411	.balign IFETCH_ALIGN_BYTES
 412DEFINE_FIXED_SYMBOL(\name\()_common_virt, text)
 413\name\()_common_virt:
 414	.if IKVM_VIRT
 415		KVMTEST \name kvm_interrupt
 4161:
 417	.endif
 418	.endif /* IVIRT */
 419.endm
 420
 421/*
 422 * Don't switch to virt mode. Used for early MCE and HMI handlers that
 423 * want to run in real mode.
 424 */
 425.macro __GEN_REALMODE_COMMON_ENTRY name
 426DEFINE_FIXED_SYMBOL(\name\()_common_real, text)
 427\name\()_common_real:
 428	.if IKVM_REAL
 429		KVMTEST \name kvm_interrupt
 430	.endif
 431.endm
 432
 433.macro __GEN_COMMON_BODY name
 434	.if IMASK
 435		.if ! ISTACK
 436		.error "No support for masked interrupt to use custom stack"
 437		.endif
 438
 439		/* If coming from user, skip soft-mask tests. */
 440		andi.	r10,r12,MSR_PR
 441		bne	3f
 442
 443		/*
 444		 * Kernel code running below __end_soft_masked may be
 445		 * implicitly soft-masked if it is within the regions
 446		 * in the soft mask table.
 447		 */
 448		LOAD_HANDLER(r10, __end_soft_masked)
 449		cmpld	r11,r10
 450		bge+	1f
 451
 452		/* SEARCH_SOFT_MASK_TABLE clobbers r9,r10,r12 */
 453		mtctr	r12
 454		stw	r9,PACA_EXGEN+EX_CCR(r13)
 455		SEARCH_SOFT_MASK_TABLE
 456		cmpdi	r12,0
 457		mfctr	r12		/* Restore r12 to SRR1 */
 458		lwz	r9,PACA_EXGEN+EX_CCR(r13)
 459		beq	1f		/* Not in soft-mask table */
 460		li	r10,IMASK
 461		b	2f		/* In soft-mask table, always mask */
 462
 463		/* Test the soft mask state against our interrupt's bit */
 4641:		lbz	r10,PACAIRQSOFTMASK(r13)
 4652:		andi.	r10,r10,IMASK
 466		/* Associate vector numbers with bits in paca->irq_happened */
 467		.if IVEC == 0x500 || IVEC == 0xea0
 468		li	r10,PACA_IRQ_EE
 469		.elseif IVEC == 0x900
 470		li	r10,PACA_IRQ_DEC
 471		.elseif IVEC == 0xa00 || IVEC == 0xe80
 472		li	r10,PACA_IRQ_DBELL
 473		.elseif IVEC == 0xe60
 474		li	r10,PACA_IRQ_HMI
 475		.elseif IVEC == 0xf00
 476		li	r10,PACA_IRQ_PMI
 477		.else
 478		.abort "Bad maskable vector"
 479		.endif
 480
 481		.if IHSRR_IF_HVMODE
 482		BEGIN_FTR_SECTION
 483		bne	masked_Hinterrupt
 484		FTR_SECTION_ELSE
 485		bne	masked_interrupt
 486		ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
 487		.elseif IHSRR
 488		bne	masked_Hinterrupt
 489		.else
 490		bne	masked_interrupt
 491		.endif
 492	.endif
 493
 494	.if ISTACK
 495	andi.	r10,r12,MSR_PR		/* See if coming from user	*/
 4963:	mr	r10,r1			/* Save r1			*/
 497	subi	r1,r1,INT_FRAME_SIZE	/* alloc frame on kernel stack	*/
 498	beq-	100f
 499	ld	r1,PACAKSAVE(r13)	/* kernel stack to use		*/
 500100:	tdgei	r1,-INT_FRAME_SIZE	/* trap if r1 is in userspace	*/
 501	EMIT_BUG_ENTRY 100b,__FILE__,__LINE__,0
 502	.endif
 503
 504	std	r9,_CCR(r1)		/* save CR in stackframe	*/
 505	std	r11,_NIP(r1)		/* save SRR0 in stackframe	*/
 506	std	r12,_MSR(r1)		/* save SRR1 in stackframe	*/
 507	std	r10,0(r1)		/* make stack chain pointer	*/
 508	std	r0,GPR0(r1)		/* save r0 in stackframe	*/
 509	std	r10,GPR1(r1)		/* save r1 in stackframe	*/
 510	SANITIZE_GPR(0)
 
 
 511
 512	/* Mark our [H]SRRs valid for return */
 513	li	r10,1
 514	.if IHSRR_IF_HVMODE
 515	BEGIN_FTR_SECTION
 516	stb	r10,PACAHSRR_VALID(r13)
 517	FTR_SECTION_ELSE
 518	stb	r10,PACASRR_VALID(r13)
 519	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
 520	.elseif IHSRR
 521	stb	r10,PACAHSRR_VALID(r13)
 522	.else
 523	stb	r10,PACASRR_VALID(r13)
 524	.endif
 525
 526	.if ISTACK
 527	.if IKUAP
 528	kuap_save_amr_and_lock r9, r10, cr1, cr0
 529	.endif
 530	beq	101f			/* if from kernel mode		*/
 531BEGIN_FTR_SECTION
 532	ld	r9,IAREA+EX_PPR(r13)	/* Read PPR from paca		*/
 533	std	r9,_PPR(r1)
 534END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
 535101:
 536	.else
 537	.if IKUAP
 538	kuap_save_amr_and_lock r9, r10, cr1
 539	.endif
 540	.endif
 541
 542	/* Save original regs values from save area to stack frame. */
 543	ld	r9,IAREA+EX_R9(r13)	/* move r9, r10 to stackframe	*/
 544	ld	r10,IAREA+EX_R10(r13)
 545	std	r9,GPR9(r1)
 546	std	r10,GPR10(r1)
 547	ld	r9,IAREA+EX_R11(r13)	/* move r11 - r13 to stackframe	*/
 548	ld	r10,IAREA+EX_R12(r13)
 549	ld	r11,IAREA+EX_R13(r13)
 550	std	r9,GPR11(r1)
 551	std	r10,GPR12(r1)
 552	std	r11,GPR13(r1)
 553	.if !IMSR_R12
 554	SANITIZE_GPRS(9, 12)
 555	.else
 556	SANITIZE_GPRS(9, 11)
 557	.endif
 558
 559	SAVE_NVGPRS(r1)
 560	SANITIZE_NVGPRS()
 561
 562	.if IDAR
 563	.if IISIDE
 564	ld	r10,_NIP(r1)
 565	.else
 566	ld	r10,IAREA+EX_DAR(r13)
 567	.endif
 568	std	r10,_DAR(r1)
 569	.endif
 570
 571	.if IDSISR
 572	.if IISIDE
 573	ld	r10,_MSR(r1)
 574	lis	r11,DSISR_SRR1_MATCH_64S@h
 575	and	r10,r10,r11
 576	.else
 577	lwz	r10,IAREA+EX_DSISR(r13)
 578	.endif
 579	std	r10,_DSISR(r1)
 580	.endif
 581
 582BEGIN_FTR_SECTION
 583	.if ICFAR || ICFAR_IF_HVMODE
 584	ld	r10,IAREA+EX_CFAR(r13)
 585	.else
 586	li	r10,0
 587	.endif
 588	std	r10,ORIG_GPR3(r1)
 589END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
 590	ld	r10,IAREA+EX_CTR(r13)
 591	std	r10,_CTR(r1)
 592	SAVE_GPRS(2, 8, r1)		/* save r2 - r8 in stackframe   */
 593	SANITIZE_GPRS(2, 8)
 594	mflr	r9			/* Get LR, later save to stack	*/
 595	LOAD_PACA_TOC()			/* get kernel TOC into r2	*/
 596	std	r9,_LINK(r1)
 597	lbz	r10,PACAIRQSOFTMASK(r13)
 598	mfspr	r11,SPRN_XER		/* save XER in stackframe	*/
 599	std	r10,SOFTE(r1)
 600	std	r11,_XER(r1)
 601	li	r9,IVEC
 602	std	r9,_TRAP(r1)		/* set trap number		*/
 603	li	r10,0
 604	LOAD_REG_IMMEDIATE(r11, STACK_FRAME_REGS_MARKER)
 605	std	r10,RESULT(r1)		/* clear regs->result		*/
 606	std	r11,STACK_INT_FRAME_MARKER(r1) /* mark the frame	*/
 607.endm
 608
 609/*
 610 * On entry r13 points to the paca, r9-r13 are saved in the paca,
 611 * r9 contains the saved CR, r11 and r12 contain the saved SRR0 and
 612 * SRR1, and relocation is on.
 613 *
 614 * If stack=0, then the stack is already set in r1, and r1 is saved in r10.
 615 * PPR save and CPU accounting is not done for the !stack case (XXX why not?)
 616 */
 617.macro GEN_COMMON name
 618	__GEN_COMMON_ENTRY \name
 619	__GEN_COMMON_BODY \name
 620.endm
 621
 622.macro SEARCH_RESTART_TABLE
 623#ifdef CONFIG_RELOCATABLE
 624	mr	r12,r2
 625	LOAD_PACA_TOC()
 626	LOAD_REG_ADDR(r9, __start___restart_table)
 627	LOAD_REG_ADDR(r10, __stop___restart_table)
 628	mr	r2,r12
 629#else
 630	LOAD_REG_IMMEDIATE_SYM(r9, r12, __start___restart_table)
 631	LOAD_REG_IMMEDIATE_SYM(r10, r12, __stop___restart_table)
 632#endif
 633300:
 634	cmpd	r9,r10
 635	beq	302f
 636	ld	r12,0(r9)
 637	cmpld	r11,r12
 638	blt	301f
 639	ld	r12,8(r9)
 640	cmpld	r11,r12
 641	bge	301f
 642	ld	r12,16(r9)
 643	b	303f
 644301:
 645	addi	r9,r9,24
 646	b	300b
 647302:
 648	li	r12,0
 649303:
 650.endm
 651
 652.macro SEARCH_SOFT_MASK_TABLE
 653#ifdef CONFIG_RELOCATABLE
 654	mr	r12,r2
 655	LOAD_PACA_TOC()
 656	LOAD_REG_ADDR(r9, __start___soft_mask_table)
 657	LOAD_REG_ADDR(r10, __stop___soft_mask_table)
 658	mr	r2,r12
 
 
 
 
 
 
 
 
 
 659#else
 660	LOAD_REG_IMMEDIATE_SYM(r9, r12, __start___soft_mask_table)
 661	LOAD_REG_IMMEDIATE_SYM(r10, r12, __stop___soft_mask_table)
 662#endif
 663300:
 664	cmpd	r9,r10
 665	beq	302f
 666	ld	r12,0(r9)
 667	cmpld	r11,r12
 668	blt	301f
 669	ld	r12,8(r9)
 670	cmpld	r11,r12
 671	bge	301f
 672	li	r12,1
 673	b	303f
 674301:
 675	addi	r9,r9,16
 676	b	300b
 677302:
 678	li	r12,0
 679303:
 680.endm
 681
 682/*
 683 * Restore all registers including H/SRR0/1 saved in a stack frame of a
 684 * standard exception.
 685 */
 686.macro EXCEPTION_RESTORE_REGS hsrr=0
 687	/* Move original SRR0 and SRR1 into the respective regs */
 688	ld	r9,_MSR(r1)
 689	li	r10,0
 690	.if \hsrr
 691	mtspr	SPRN_HSRR1,r9
 692	stb	r10,PACAHSRR_VALID(r13)
 693	.else
 694	mtspr	SPRN_SRR1,r9
 695	stb	r10,PACASRR_VALID(r13)
 696	.endif
 697	ld	r9,_NIP(r1)
 698	.if \hsrr
 699	mtspr	SPRN_HSRR0,r9
 700	.else
 701	mtspr	SPRN_SRR0,r9
 702	.endif
 703	ld	r9,_CTR(r1)
 704	mtctr	r9
 705	ld	r9,_XER(r1)
 706	mtxer	r9
 707	ld	r9,_LINK(r1)
 708	mtlr	r9
 709	ld	r9,_CCR(r1)
 710	mtcr	r9
 711	SANITIZE_RESTORE_NVGPRS()
 712	REST_GPRS(2, 13, r1)
 713	REST_GPR(0, r1)
 714	/* restore original r1. */
 715	ld	r1,GPR1(r1)
 716.endm
 717
 718/*
 719 * EARLY_BOOT_FIXUP - Fix real-mode interrupt with wrong endian in early boot.
 720 *
 721 * There's a short window during boot where although the kernel is running
 722 * little endian, any exceptions will cause the CPU to switch back to big
 723 * endian. For example a WARN() boils down to a trap instruction, which will
 724 * cause a program check, and we end up here but with the CPU in big endian
 725 * mode. The first instruction of the program check handler (in GEN_INT_ENTRY
 726 * below) is an mtsprg, which when executed in the wrong endian is an lhzu with
 727 * a ~3GB displacement from r3. The content of r3 is random, so that is a load
 728 * from some random location, and depending on the system can easily lead to a
 729 * checkstop, or an infinitely recursive page fault.
 730 *
 731 * So to handle that case we have a trampoline here that can detect we are in
 732 * the wrong endian and flip us back to the correct endian. We can't flip
 733 * MSR[LE] using mtmsr, so we have to use rfid. That requires backing up SRR0/1
 734 * as well as a GPR. To do that we use SPRG0/2/3, as SPRG1 is already used for
 735 * the paca. SPRG3 is user readable, but this trampoline is only active very
 736 * early in boot, and SPRG3 will be reinitialised in vdso_getcpu_init() before
 737 * userspace starts.
 738 */
 739.macro EARLY_BOOT_FIXUP
 740BEGIN_FTR_SECTION
 741#ifdef CONFIG_CPU_LITTLE_ENDIAN
 742	tdi   0,0,0x48    // Trap never, or in reverse endian: b . + 8
 743	b     2f          // Skip trampoline if endian is correct
 744	.long 0xa643707d  // mtsprg  0, r11      Backup r11
 745	.long 0xa6027a7d  // mfsrr0  r11
 746	.long 0xa643727d  // mtsprg  2, r11      Backup SRR0 in SPRG2
 747	.long 0xa6027b7d  // mfsrr1  r11
 748	.long 0xa643737d  // mtsprg  3, r11      Backup SRR1 in SPRG3
 749	.long 0xa600607d  // mfmsr   r11
 750	.long 0x01006b69  // xori    r11, r11, 1 Invert MSR[LE]
 751	.long 0xa6037b7d  // mtsrr1  r11
 752	/*
 753	 * This is 'li  r11,1f' where 1f is the absolute address of that
 754	 * label, byteswapped into the SI field of the instruction.
 755	 */
 756	.long 0x00006039 | \
 757		((ABS_ADDR(1f, real_vectors) & 0x00ff) << 24) | \
 758		((ABS_ADDR(1f, real_vectors) & 0xff00) << 8)
 759	.long 0xa6037a7d  // mtsrr0  r11
 760	.long 0x2400004c  // rfid
 7611:
 762	mfsprg r11, 3
 763	mtsrr1 r11        // Restore SRR1
 764	mfsprg r11, 2
 765	mtsrr0 r11        // Restore SRR0
 766	mfsprg r11, 0     // Restore r11
 7672:
 768#endif
 769	/*
 770	 * program check could hit at any time, and pseries can not block
 771	 * MSR[ME] in early boot. So check if there is anything useful in r13
 772	 * yet, and spin forever if not.
 773	 */
 774	mtsprg	0, r11
 775	mfcr	r11
 776	cmpdi	r13, 0
 777	beq	.
 778	mtcr	r11
 779	mfsprg	r11, 0
 780END_FTR_SECTION(0, 1)     // nop out after boot
 781.endm
 782
 783/*
 784 * There are a few constraints to be concerned with.
 785 * - Real mode exceptions code/data must be located at their physical location.
 786 * - Virtual mode exceptions must be mapped at their 0xc000... location.
 787 * - Fixed location code must not call directly beyond the __end_interrupts
 788 *   area when built with CONFIG_RELOCATABLE. LOAD_HANDLER / bctr sequence
 789 *   must be used.
 790 * - LOAD_HANDLER targets must be within first 64K of physical 0 /
 791 *   virtual 0xc00...
 792 * - Conditional branch targets must be within +/-32K of caller.
 793 *
 794 * "Virtual exceptions" run with relocation on (MSR_IR=1, MSR_DR=1), and
 795 * therefore don't have to run in physically located code or rfid to
 796 * virtual mode kernel code. However on relocatable kernels they do have
 797 * to branch to KERNELBASE offset because the rest of the kernel (outside
 798 * the exception vectors) may be located elsewhere.
 799 *
 800 * Virtual exceptions correspond with physical, except their entry points
 801 * are offset by 0xc000000000000000 and also tend to get an added 0x4000
 802 * offset applied. Virtual exceptions are enabled with the Alternate
 803 * Interrupt Location (AIL) bit set in the LPCR. However this does not
 804 * guarantee they will be delivered virtually. Some conditions (see the ISA)
 805 * cause exceptions to be delivered in real mode.
 806 *
 807 * The scv instructions are a special case. They get a 0x3000 offset applied.
 808 * scv exceptions have unique reentrancy properties, see below.
 809 *
 810 * It's impossible to receive interrupts below 0x300 via AIL.
 811 *
 812 * KVM: None of the virtual exceptions are from the guest. Anything that
 813 * escalated to HV=1 from HV=0 is delivered via real mode handlers.
 814 *
 815 *
 816 * We layout physical memory as follows:
 817 * 0x0000 - 0x00ff : Secondary processor spin code
 818 * 0x0100 - 0x18ff : Real mode pSeries interrupt vectors
 819 * 0x1900 - 0x2fff : Real mode trampolines
 820 * 0x3000 - 0x58ff : Relon (IR=1,DR=1) mode pSeries interrupt vectors
 821 * 0x5900 - 0x6fff : Relon mode trampolines
 822 * 0x7000 - 0x7fff : FWNMI data area
 823 * 0x8000 -   .... : Common interrupt handlers, remaining early
 824 *                   setup code, rest of kernel.
 825 *
 826 * We could reclaim 0x4000-0x42ff for real mode trampolines if the space
 827 * is necessary. Until then it's more consistent to explicitly put VIRT_NONE
 828 * vectors there.
 829 */
 830OPEN_FIXED_SECTION(real_vectors,        0x0100, 0x1900)
 831OPEN_FIXED_SECTION(real_trampolines,    0x1900, 0x3000)
 832OPEN_FIXED_SECTION(virt_vectors,        0x3000, 0x5900)
 833OPEN_FIXED_SECTION(virt_trampolines,    0x5900, 0x7000)
 834
 835#ifdef CONFIG_PPC_POWERNV
 836	.globl start_real_trampolines
 837	.globl end_real_trampolines
 838	.globl start_virt_trampolines
 839	.globl end_virt_trampolines
 840#endif
 841
 842#if defined(CONFIG_PPC_PSERIES) || defined(CONFIG_PPC_POWERNV)
 843/*
 844 * Data area reserved for FWNMI option.
 845 * This address (0x7000) is fixed by the RPA.
 846 * pseries and powernv need to keep the whole page from
 847 * 0x7000 to 0x8000 free for use by the firmware
 848 */
 849ZERO_FIXED_SECTION(fwnmi_page,          0x7000, 0x8000)
 850OPEN_TEXT_SECTION(0x8000)
 851#else
 852OPEN_TEXT_SECTION(0x7000)
 853#endif
 854
 855USE_FIXED_SECTION(real_vectors)
 856
 857/*
 858 * This is the start of the interrupt handlers for pSeries
 859 * This code runs with relocation off.
 860 * Code from here to __end_interrupts gets copied down to real
 861 * address 0x100 when we are running a relocatable kernel.
 862 * Therefore any relative branches in this section must only
 863 * branch to labels in this section.
 864 */
 865	.globl __start_interrupts
 866__start_interrupts:
 867
 868/**
 869 * Interrupt 0x3000 - System Call Vectored Interrupt (syscall).
 870 * This is a synchronous interrupt invoked with the "scv" instruction. The
 871 * system call does not alter the HV bit, so it is directed to the OS.
 872 *
 873 * Handling:
 874 * scv instructions enter the kernel without changing EE, RI, ME, or HV.
 875 * In particular, this means we can take a maskable interrupt at any point
 876 * in the scv handler, which is unlike any other interrupt. This is solved
 877 * by treating the instruction addresses in the handler as being soft-masked,
 878 * by adding a SOFT_MASK_TABLE entry for them.
 879 *
 880 * AIL-0 mode scv exceptions go to 0x17000-0x17fff, but we set AIL-3 and
 881 * ensure scv is never executed with relocation off, which means AIL-0
 882 * should never happen.
 883 *
 884 * Before leaving the following inside-__end_soft_masked text, at least of the
 885 * following must be true:
 886 * - MSR[PR]=1 (i.e., return to userspace)
 887 * - MSR_EE|MSR_RI is clear (no reentrant exceptions)
 888 * - Standard kernel environment is set up (stack, paca, etc)
 889 *
 890 * KVM:
 891 * These interrupts do not elevate HV 0->1, so HV is not involved. PR KVM
 892 * ensures that FSCR[SCV] is disabled whenever it has to force AIL off.
 893 *
 894 * Call convention:
 895 *
 896 * syscall register convention is in Documentation/arch/powerpc/syscall64-abi.rst
 897 */
 898EXC_VIRT_BEGIN(system_call_vectored, 0x3000, 0x1000)
 899	/* SCV 0 */
 900	mr	r9,r13
 901	GET_PACA(r13)
 902	mflr	r11
 903	mfctr	r12
 904	li	r10,IRQS_ALL_DISABLED
 905	stb	r10,PACAIRQSOFTMASK(r13)
 906#ifdef CONFIG_RELOCATABLE
 907	b	system_call_vectored_tramp
 908#else
 909	b	system_call_vectored_common
 
 
 
 
 910#endif
 911	nop
 912
 913	/* SCV 1 - 127 */
 914	.rept	127
 915	mr	r9,r13
 916	GET_PACA(r13)
 917	mflr	r11
 918	mfctr	r12
 919	li	r10,IRQS_ALL_DISABLED
 920	stb	r10,PACAIRQSOFTMASK(r13)
 921	li	r0,-1 /* cause failure */
 922#ifdef CONFIG_RELOCATABLE
 923	b	system_call_vectored_sigill_tramp
 924#else
 925	b	system_call_vectored_sigill
 926#endif
 927	.endr
 928EXC_VIRT_END(system_call_vectored, 0x3000, 0x1000)
 
 929
 930// Treat scv vectors as soft-masked, see comment above.
 931// Use absolute values rather than labels here, so they don't get relocated,
 932// because this code runs unrelocated.
 933SOFT_MASK_TABLE(0xc000000000003000, 0xc000000000004000)
 934
 935#ifdef CONFIG_RELOCATABLE
 936TRAMP_VIRT_BEGIN(system_call_vectored_tramp)
 937	__LOAD_HANDLER(r10, system_call_vectored_common, virt_trampolines)
 938	mtctr	r10
 939	bctr
 940
 941TRAMP_VIRT_BEGIN(system_call_vectored_sigill_tramp)
 942	__LOAD_HANDLER(r10, system_call_vectored_sigill, virt_trampolines)
 943	mtctr	r10
 944	bctr
 945#endif
 946
 
 
 
 
 947
 948/* No virt vectors corresponding with 0x0..0x100 */
 949EXC_VIRT_NONE(0x4000, 0x100)
 950
 
 
 951
 952/**
 953 * Interrupt 0x100 - System Reset Interrupt (SRESET aka NMI).
 954 * This is a non-maskable, asynchronous interrupt always taken in real-mode.
 955 * It is caused by:
 956 * - Wake from power-saving state, on powernv.
 957 * - An NMI from another CPU, triggered by firmware or hypercall.
 958 * - As crash/debug signal injected from BMC, firmware or hypervisor.
 959 *
 960 * Handling:
 961 * Power-save wakeup is the only performance critical path, so this is
 962 * determined quickly as possible first. In this case volatile registers
 963 * can be discarded and SPRs like CFAR don't need to be read.
 964 *
 965 * If not a powersave wakeup, then it's run as a regular interrupt, however
 966 * it uses its own stack and PACA save area to preserve the regular kernel
 967 * environment for debugging.
 968 *
 969 * This interrupt is not maskable, so triggering it when MSR[RI] is clear,
 970 * or SCRATCH0 is in use, etc. may cause a crash. It's also not entirely
 971 * correct to switch to virtual mode to run the regular interrupt handler
 972 * because it might be interrupted when the MMU is in a bad state (e.g., SLB
 973 * is clear).
 974 *
 975 * FWNMI:
 976 * PAPR specifies a "fwnmi" facility which sends the sreset to a different
 977 * entry point with a different register set up. Some hypervisors will
 978 * send the sreset to 0x100 in the guest if it is not fwnmi capable.
 979 *
 980 * KVM:
 981 * Unlike most SRR interrupts, this may be taken by the host while executing
 982 * in a guest, so a KVM test is required. KVM will pull the CPU out of guest
 983 * mode and then raise the sreset.
 984 */
 985INT_DEFINE_BEGIN(system_reset)
 986	IVEC=0x100
 987	IAREA=PACA_EXNMI
 988	IVIRT=0 /* no virt entry point */
 989	ISTACK=0
 990	IKVM_REAL=1
 991INT_DEFINE_END(system_reset)
 992
 993EXC_REAL_BEGIN(system_reset, 0x100, 0x100)
 994#ifdef CONFIG_PPC_P7_NAP
 995	/*
 996	 * If running native on arch 2.06 or later, check if we are waking up
 997	 * from nap/sleep/winkle, and branch to idle handler. This tests SRR1
 998	 * bits 46:47. A non-0 value indicates that we are coming from a power
 999	 * saving state. The idle wakeup handler initially runs in real mode,
1000	 * but we branch to the 0xc000... address so we can turn on relocation
1001	 * with mtmsrd later, after SPRs are restored.
1002	 *
1003	 * Careful to minimise cost for the fast path (idle wakeup) while
1004	 * also avoiding clobbering CFAR for the debug path (non-idle).
1005	 *
1006	 * For the idle wake case volatile registers can be clobbered, which
1007	 * is why we use those initially. If it turns out to not be an idle
1008	 * wake, carefully put everything back the way it was, so we can use
1009	 * common exception macros to handle it.
1010	 */
1011BEGIN_FTR_SECTION
1012	SET_SCRATCH0(r13)
1013	GET_PACA(r13)
1014	std	r3,PACA_EXNMI+0*8(r13)
1015	std	r4,PACA_EXNMI+1*8(r13)
1016	std	r5,PACA_EXNMI+2*8(r13)
1017	mfspr	r3,SPRN_SRR1
1018	mfocrf	r4,0x80
1019	rlwinm.	r5,r3,47-31,30,31
1020	bne+	system_reset_idle_wake
1021	/* Not powersave wakeup. Restore regs for regular interrupt handler. */
1022	mtocrf	0x80,r4
1023	ld	r3,PACA_EXNMI+0*8(r13)
1024	ld	r4,PACA_EXNMI+1*8(r13)
1025	ld	r5,PACA_EXNMI+2*8(r13)
1026	GET_SCRATCH0(r13)
1027END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
 
1028#endif
 
 
 
 
1029
1030	GEN_INT_ENTRY system_reset, virt=0
1031	/*
1032	 * In theory, we should not enable relocation here if it was disabled
1033	 * in SRR1, because the MMU may not be configured to support it (e.g.,
1034	 * SLB may have been cleared). In practice, there should only be a few
1035	 * small windows where that's the case, and sreset is considered to
1036	 * be dangerous anyway.
1037	 */
1038EXC_REAL_END(system_reset, 0x100, 0x100)
1039EXC_VIRT_NONE(0x4100, 0x100)
 
 
 
1040
1041#ifdef CONFIG_PPC_P7_NAP
1042TRAMP_REAL_BEGIN(system_reset_idle_wake)
1043	/* We are waking up from idle, so may clobber any volatile register */
1044	cmpwi	cr1,r5,2
1045	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
1046	__LOAD_FAR_HANDLER(r12, DOTSYM(idle_return_gpr_loss), real_trampolines)
1047	mtctr	r12
1048	bctr
1049#endif
1050
1051#ifdef CONFIG_PPC_PSERIES
1052/*
1053 * Vectors for the FWNMI option.  Share common code.
1054 */
1055TRAMP_REAL_BEGIN(system_reset_fwnmi)
1056	GEN_INT_ENTRY system_reset, virt=0
1057
1058#endif /* CONFIG_PPC_PSERIES */
 
 
 
 
1059
1060EXC_COMMON_BEGIN(system_reset_common)
1061	__GEN_COMMON_ENTRY system_reset
1062	/*
1063	 * Increment paca->in_nmi. When the interrupt entry wrapper later
1064	 * enable MSR_RI, then SLB or MCE will be able to recover, but a nested
1065	 * NMI will notice in_nmi and not recover because of the use of the NMI
1066	 * stack. in_nmi reentrancy is tested in system_reset_exception.
 
 
 
1067	 */
1068	lhz	r10,PACA_IN_NMI(r13)
1069	addi	r10,r10,1
1070	sth	r10,PACA_IN_NMI(r13)
 
 
1071
1072	mr	r10,r1
1073	ld	r1,PACA_NMI_EMERG_SP(r13)
1074	subi	r1,r1,INT_FRAME_SIZE
1075	__GEN_COMMON_BODY system_reset
 
1076
1077	addi	r3,r1,STACK_INT_FRAME_REGS
1078	bl	CFUNC(system_reset_exception)
 
 
 
1079
1080	/* Clear MSR_RI before setting SRR0 and SRR1. */
1081	li	r9,0
1082	mtmsrd	r9,1
 
 
1083
1084	/*
1085	 * MSR_RI is clear, now we can decrement paca->in_nmi.
1086	 */
1087	lhz	r10,PACA_IN_NMI(r13)
1088	subi	r10,r10,1
1089	sth	r10,PACA_IN_NMI(r13)
1090
1091	kuap_kernel_restore r9, r10
1092	EXCEPTION_RESTORE_REGS
1093	RFI_TO_USER_OR_KERNEL
 
 
 
 
 
 
 
 
 
 
 
1094
 
 
 
 
 
 
 
1095
1096/**
1097 * Interrupt 0x200 - Machine Check Interrupt (MCE).
1098 * This is a non-maskable interrupt always taken in real-mode. It can be
1099 * synchronous or asynchronous, caused by hardware or software, and it may be
1100 * taken in a power-saving state.
1101 *
1102 * Handling:
1103 * Similarly to system reset, this uses its own stack and PACA save area,
1104 * the difference is re-entrancy is allowed on the machine check stack.
1105 *
1106 * machine_check_early is run in real mode, and carefully decodes the
1107 * machine check and tries to handle it (e.g., flush the SLB if there was an
1108 * error detected there), determines if it was recoverable and logs the
1109 * event.
1110 *
1111 * This early code does not "reconcile" irq soft-mask state like SRESET or
1112 * regular interrupts do, so irqs_disabled() among other things may not work
1113 * properly (irq disable/enable already doesn't work because irq tracing can
1114 * not work in real mode).
1115 *
1116 * Then, depending on the execution context when the interrupt is taken, there
1117 * are 3 main actions:
1118 * - Executing in kernel mode. The event is queued with irq_work, which means
1119 *   it is handled when it is next safe to do so (i.e., the kernel has enabled
1120 *   interrupts), which could be immediately when the interrupt returns. This
1121 *   avoids nasty issues like switching to virtual mode when the MMU is in a
1122 *   bad state, or when executing OPAL code. (SRESET is exposed to such issues,
1123 *   but it has different priorities). Check to see if the CPU was in power
1124 *   save, and return via the wake up code if it was.
1125 *
1126 * - Executing in user mode. machine_check_exception is run like a normal
1127 *   interrupt handler, which processes the data generated by the early handler.
1128 *
1129 * - Executing in guest mode. The interrupt is run with its KVM test, and
1130 *   branches to KVM to deal with. KVM may queue the event for the host
1131 *   to report later.
1132 *
1133 * This interrupt is not maskable, so if it triggers when MSR[RI] is clear,
1134 * or SCRATCH0 is in use, it may cause a crash.
1135 *
1136 * KVM:
1137 * See SRESET.
1138 */
1139INT_DEFINE_BEGIN(machine_check_early)
1140	IVEC=0x200
1141	IAREA=PACA_EXMC
1142	IVIRT=0 /* no virt entry point */
1143	IREALMODE_COMMON=1
1144	ISTACK=0
1145	IDAR=1
1146	IDSISR=1
1147	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
1148INT_DEFINE_END(machine_check_early)
1149
1150INT_DEFINE_BEGIN(machine_check)
1151	IVEC=0x200
1152	IAREA=PACA_EXMC
1153	IVIRT=0 /* no virt entry point */
1154	IDAR=1
1155	IDSISR=1
1156	IKVM_REAL=1
1157INT_DEFINE_END(machine_check)
1158
1159EXC_REAL_BEGIN(machine_check, 0x200, 0x100)
1160	EARLY_BOOT_FIXUP
1161	GEN_INT_ENTRY machine_check_early, virt=0
1162EXC_REAL_END(machine_check, 0x200, 0x100)
1163EXC_VIRT_NONE(0x4200, 0x100)
1164
1165#ifdef CONFIG_PPC_PSERIES
1166TRAMP_REAL_BEGIN(machine_check_fwnmi)
1167	/* See comment at machine_check exception, don't turn on RI */
1168	GEN_INT_ENTRY machine_check_early, virt=0
1169#endif
1170
1171#define MACHINE_CHECK_HANDLER_WINDUP			\
1172	/* Clear MSR_RI before setting SRR0 and SRR1. */\
1173	li	r9,0;					\
1174	mtmsrd	r9,1;		/* Clear MSR_RI */	\
1175	/* Decrement paca->in_mce now RI is clear. */	\
1176	lhz	r12,PACA_IN_MCE(r13);			\
1177	subi	r12,r12,1;				\
1178	sth	r12,PACA_IN_MCE(r13);			\
1179	EXCEPTION_RESTORE_REGS
1180
1181EXC_COMMON_BEGIN(machine_check_early_common)
1182	__GEN_REALMODE_COMMON_ENTRY machine_check_early
1183
 
 
 
 
 
 
 
1184	/*
 
 
 
 
 
1185	 * Switch to mc_emergency stack and handle re-entrancy (we limit
1186	 * the nested MCE upto level 4 to avoid stack overflow).
1187	 * Save MCE registers srr1, srr0, dar and dsisr and then set ME=1
1188	 *
1189	 * We use paca->in_mce to check whether this is the first entry or
1190	 * nested machine check. We increment paca->in_mce to track nested
1191	 * machine checks.
1192	 *
1193	 * If this is the first entry then set stack pointer to
1194	 * paca->mc_emergency_sp, otherwise r1 is already pointing to
1195	 * stack frame on mc_emergency stack.
1196	 *
1197	 * NOTE: We are here with MSR_ME=0 (off), which means we risk a
1198	 * checkstop if we get another machine check exception before we do
1199	 * rfid with MSR_ME=1.
1200	 *
1201	 * This interrupt can wake directly from idle. If that is the case,
1202	 * the machine check is handled then the idle wakeup code is called
1203	 * to restore state.
1204	 */
 
1205	lhz	r10,PACA_IN_MCE(r13)
1206	cmpwi	r10,0			/* Are we in nested machine check */
1207	cmpwi	cr1,r10,MAX_MCE_DEPTH	/* Are we at maximum nesting */
1208	addi	r10,r10,1		/* increment paca->in_mce */
1209	sth	r10,PACA_IN_MCE(r13)
1210
1211	mr	r10,r1			/* Save r1 */
1212	bne	1f
1213	/* First machine check entry */
1214	ld	r1,PACAMCEMERGSP(r13)	/* Use MC emergency stack */
12151:	/* Limit nested MCE to level 4 to avoid stack overflow */
1216	bgt	cr1,unrecoverable_mce	/* Check if we hit limit of 4 */
1217	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame */
1218
1219	__GEN_COMMON_BODY machine_check_early
1220
1221BEGIN_FTR_SECTION
1222	bl	enable_machine_check
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1223END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1224	addi	r3,r1,STACK_INT_FRAME_REGS
1225BEGIN_FTR_SECTION
1226	bl	CFUNC(machine_check_early_boot)
1227END_FTR_SECTION(0, 1)     // nop out after boot
1228	bl	CFUNC(machine_check_early)
1229	std	r3,RESULT(r1)	/* Save result */
1230	ld	r12,_MSR(r1)
1231
1232#ifdef CONFIG_PPC_P7_NAP
1233	/*
1234	 * Check if thread was in power saving mode. We come here when any
1235	 * of the following is true:
1236	 * a. thread wasn't in power saving mode
1237	 * b. thread was in power saving mode with no state loss,
1238	 *    supervisor state loss or hypervisor state loss.
1239	 *
1240	 * Go back to nap/sleep/winkle mode again if (b) is true.
1241	 */
1242BEGIN_FTR_SECTION
1243	rlwinm.	r11,r12,47-31,30,31
1244	bne	machine_check_idle_common
1245END_FTR_SECTION_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
1246#endif
1247
1248#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1249	/*
1250	 * Check if we are coming from guest. If yes, then run the normal
1251	 * exception handler which will take the
1252	 * machine_check_kvm->kvm_interrupt branch to deliver the MC event
1253	 * to guest.
1254	 */
1255	lbz	r11,HSTATE_IN_GUEST(r13)
1256	cmpwi	r11,0			/* Check if coming from guest */
1257	bne	mce_deliver		/* continue if we are. */
1258#endif
 
1259
1260	/*
1261	 * Check if we are coming from userspace. If yes, then run the normal
1262	 * exception handler which will deliver the MC event to this kernel.
1263	 */
1264	andi.	r11,r12,MSR_PR		/* See if coming from user. */
1265	bne	mce_deliver		/* continue in V mode if we are. */
1266
1267	/*
1268	 * At this point we are coming from kernel context.
1269	 * Queue up the MCE event and return from the interrupt.
1270	 * But before that, check if this is an un-recoverable exception.
1271	 * If yes, then stay on emergency stack and panic.
1272	 */
1273	andi.	r11,r12,MSR_RI
1274	beq	unrecoverable_mce
 
1275
1276	/*
1277	 * Check if we have successfully handled/recovered from error, if not
1278	 * then stay on emergency stack and panic.
1279	 */
1280	ld	r3,RESULT(r1)	/* Load result */
1281	cmpdi	r3,0		/* see if we handled MCE successfully */
1282	beq	unrecoverable_mce /* if !handled then panic */
1283
1284	/*
1285	 * Return from MC interrupt.
1286	 * Queue up the MCE event so that we can log it later, while
1287	 * returning from kernel or opal call.
1288	 */
1289	bl	CFUNC(machine_check_queue_event)
1290	MACHINE_CHECK_HANDLER_WINDUP
1291	RFI_TO_KERNEL
1292
1293mce_deliver:
1294	/*
1295	 * This is a host user or guest MCE. Restore all registers, then
1296	 * run the "late" handler. For host user, this will run the
1297	 * machine_check_exception handler in virtual mode like a normal
1298	 * interrupt handler. For guest, this will trigger the KVM test
1299	 * and branch to the KVM interrupt similarly to other interrupts.
1300	 */
1301BEGIN_FTR_SECTION
1302	ld	r10,ORIG_GPR3(r1)
 
 
 
 
 
 
 
 
 
 
 
 
 
1303	mtspr	SPRN_CFAR,r10
1304END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
1305	MACHINE_CHECK_HANDLER_WINDUP
1306	GEN_INT_ENTRY machine_check, virt=0
1307
1308EXC_COMMON_BEGIN(machine_check_common)
1309	/*
1310	 * Machine check is different because we use a different
1311	 * save area: PACA_EXMC instead of PACA_EXGEN.
1312	 */
1313	GEN_COMMON machine_check
1314	addi	r3,r1,STACK_INT_FRAME_REGS
1315	bl	CFUNC(machine_check_exception_async)
1316	b	interrupt_return_srr
1317
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1318
1319#ifdef CONFIG_PPC_P7_NAP
1320/*
1321 * This is an idle wakeup. Low level machine check has already been
1322 * done. Queue the event then call the idle code to do the wake up.
 
 
 
 
 
 
1323 */
1324EXC_COMMON_BEGIN(machine_check_idle_common)
1325	bl	CFUNC(machine_check_queue_event)
1326
1327	/*
1328	 * GPR-loss wakeups are relatively straightforward, because the
1329	 * idle sleep code has saved all non-volatile registers on its
1330	 * own stack, and r1 in PACAR1.
1331	 *
1332	 * For no-loss wakeups the r1 and lr registers used by the
1333	 * early machine check handler have to be restored first. r2 is
1334	 * the kernel TOC, so no need to restore it.
1335	 *
1336	 * Then decrement MCE nesting after finishing with the stack.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1337	 */
1338	ld	r3,_MSR(r1)
1339	ld	r4,_LINK(r1)
1340	ld	r1,GPR1(r1)
 
 
 
 
 
 
 
 
 
 
 
 
 
1341
1342	lhz	r11,PACA_IN_MCE(r13)
1343	subi	r11,r11,1
1344	sth	r11,PACA_IN_MCE(r13)
1345
1346	mtlr	r4
1347	rlwinm	r10,r3,47-31,30,31
1348	cmpwi	cr1,r10,2
1349	bltlr	cr1	/* no state loss, return to idle caller with r3=SRR1 */
1350	b	idle_return_gpr_loss
1351#endif
1352
1353EXC_COMMON_BEGIN(unrecoverable_mce)
1354	/*
1355	 * We are going down. But there are chances that we might get hit by
1356	 * another MCE during panic path and we may run into unstable state
1357	 * with no way out. Hence, turn ME bit off while going down, so that
1358	 * when another MCE is hit during panic path, system will checkstop
1359	 * and hypervisor will get restarted cleanly by SP.
1360	 */
1361BEGIN_FTR_SECTION
1362	li	r10,0 /* clear MSR_RI */
1363	mtmsrd	r10,1
1364	bl	CFUNC(disable_machine_check)
1365END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
1366	ld	r10,PACAKMSR(r13)
1367	li	r3,MSR_ME
1368	andc	r10,r10,r3
1369	mtmsrd	r10
1370
1371	lhz	r12,PACA_IN_MCE(r13)
1372	subi	r12,r12,1
1373	sth	r12,PACA_IN_MCE(r13)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1374
 
 
1375	/*
1376	 * Invoke machine_check_exception to print MCE event and panic.
1377	 * This is the NMI version of the handler because we are called from
1378	 * the early handler which is a true NMI.
1379	 */
1380	addi	r3,r1,STACK_INT_FRAME_REGS
1381	bl	CFUNC(machine_check_exception)
 
 
 
 
1382
 
1383	/*
1384	 * We will not reach here. Even if we did, there is no way out.
1385	 * Call unrecoverable_exception and die.
1386	 */
1387	addi	r3,r1,STACK_INT_FRAME_REGS
1388	bl	CFUNC(unrecoverable_exception)
 
 
 
1389	b	.
 
1390
1391
1392/**
1393 * Interrupt 0x300 - Data Storage Interrupt (DSI).
1394 * This is a synchronous interrupt generated due to a data access exception,
1395 * e.g., a load orstore which does not have a valid page table entry with
1396 * permissions. DAWR matches also fault here, as do RC updates, and minor misc
1397 * errors e.g., copy/paste, AMO, certain invalid CI accesses, etc.
1398 *
1399 * Handling:
1400 * - Hash MMU
1401 *   Go to do_hash_fault, which attempts to fill the HPT from an entry in the
1402 *   Linux page table. Hash faults can hit in kernel mode in a fairly
1403 *   arbitrary state (e.g., interrupts disabled, locks held) when accessing
1404 *   "non-bolted" regions, e.g., vmalloc space. However these should always be
1405 *   backed by Linux page table entries.
1406 *
1407 *   If no entry is found the Linux page fault handler is invoked (by
1408 *   do_hash_fault). Linux page faults can happen in kernel mode due to user
1409 *   copy operations of course.
1410 *
1411 *   KVM: The KVM HDSI handler may perform a load with MSR[DR]=1 in guest
1412 *   MMU context, which may cause a DSI in the host, which must go to the
1413 *   KVM handler. MSR[IR] is not enabled, so the real-mode handler will
1414 *   always be used regardless of AIL setting.
1415 *
1416 * - Radix MMU
1417 *   The hardware loads from the Linux page table directly, so a fault goes
1418 *   immediately to Linux page fault.
1419 *
1420 * Conditions like DAWR match are handled on the way in to Linux page fault.
1421 */
1422INT_DEFINE_BEGIN(data_access)
1423	IVEC=0x300
1424	IDAR=1
1425	IDSISR=1
1426	IKVM_REAL=1
1427INT_DEFINE_END(data_access)
1428
1429EXC_REAL_BEGIN(data_access, 0x300, 0x80)
1430	GEN_INT_ENTRY data_access, virt=0
1431EXC_REAL_END(data_access, 0x300, 0x80)
1432EXC_VIRT_BEGIN(data_access, 0x4300, 0x80)
1433	GEN_INT_ENTRY data_access, virt=1
1434EXC_VIRT_END(data_access, 0x4300, 0x80)
1435EXC_COMMON_BEGIN(data_access_common)
1436	GEN_COMMON data_access
1437	ld	r4,_DSISR(r1)
1438	addi	r3,r1,STACK_INT_FRAME_REGS
1439	andis.	r0,r4,DSISR_DABRMATCH@h
1440	bne-	1f
1441#ifdef CONFIG_PPC_64S_HASH_MMU
1442BEGIN_MMU_FTR_SECTION
1443	bl	CFUNC(do_hash_fault)
1444MMU_FTR_SECTION_ELSE
1445	bl	CFUNC(do_page_fault)
1446ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1447#else
1448	bl	CFUNC(do_page_fault)
1449#endif
1450	b	interrupt_return_srr
1451
14521:	bl	CFUNC(do_break)
1453	/*
1454	 * do_break() may have changed the NV GPRS while handling a breakpoint.
1455	 * If so, we need to restore them with their updated values.
1456	 */
1457	HANDLER_RESTORE_NVGPRS()
1458	b	interrupt_return_srr
1459
 
1460
1461/**
1462 * Interrupt 0x380 - Data Segment Interrupt (DSLB).
1463 * This is a synchronous interrupt in response to an MMU fault missing SLB
1464 * entry for HPT, or an address outside RPT translation range.
1465 *
1466 * Handling:
1467 * - HPT:
1468 *   This refills the SLB, or reports an access fault similarly to a bad page
1469 *   fault. When coming from user-mode, the SLB handler may access any kernel
1470 *   data, though it may itself take a DSLB. When coming from kernel mode,
1471 *   recursive faults must be avoided so access is restricted to the kernel
1472 *   image text/data, kernel stack, and any data allocated below
1473 *   ppc64_bolted_size (first segment). The kernel handler must avoid stomping
1474 *   on user-handler data structures.
1475 *
1476 *   KVM: Same as 0x300, DSLB must test for KVM guest.
1477 */
1478INT_DEFINE_BEGIN(data_access_slb)
1479	IVEC=0x380
1480	IDAR=1
1481	IKVM_REAL=1
1482INT_DEFINE_END(data_access_slb)
1483
1484EXC_REAL_BEGIN(data_access_slb, 0x380, 0x80)
1485	GEN_INT_ENTRY data_access_slb, virt=0
1486EXC_REAL_END(data_access_slb, 0x380, 0x80)
1487EXC_VIRT_BEGIN(data_access_slb, 0x4380, 0x80)
1488	GEN_INT_ENTRY data_access_slb, virt=1
1489EXC_VIRT_END(data_access_slb, 0x4380, 0x80)
1490EXC_COMMON_BEGIN(data_access_slb_common)
1491	GEN_COMMON data_access_slb
1492#ifdef CONFIG_PPC_64S_HASH_MMU
1493BEGIN_MMU_FTR_SECTION
1494	/* HPT case, do SLB fault */
1495	addi	r3,r1,STACK_INT_FRAME_REGS
1496	bl	CFUNC(do_slb_fault)
1497	cmpdi	r3,0
1498	bne-	1f
1499	b	fast_interrupt_return_srr
15001:	/* Error case */
1501MMU_FTR_SECTION_ELSE
1502	/* Radix case, access is outside page table range */
1503	li	r3,-EFAULT
1504ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1505#else
1506	li	r3,-EFAULT
1507#endif
1508	std	r3,RESULT(r1)
1509	addi	r3,r1,STACK_INT_FRAME_REGS
1510	bl	CFUNC(do_bad_segment_interrupt)
1511	b	interrupt_return_srr
1512
1513
1514/**
1515 * Interrupt 0x400 - Instruction Storage Interrupt (ISI).
1516 * This is a synchronous interrupt in response to an MMU fault due to an
1517 * instruction fetch.
1518 *
1519 * Handling:
1520 * Similar to DSI, though in response to fetch. The faulting address is found
1521 * in SRR0 (rather than DAR), and status in SRR1 (rather than DSISR).
1522 */
1523INT_DEFINE_BEGIN(instruction_access)
1524	IVEC=0x400
1525	IISIDE=1
1526	IDAR=1
1527	IDSISR=1
1528#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1529	IKVM_REAL=1
1530#endif
1531INT_DEFINE_END(instruction_access)
1532
1533EXC_REAL_BEGIN(instruction_access, 0x400, 0x80)
1534	GEN_INT_ENTRY instruction_access, virt=0
1535EXC_REAL_END(instruction_access, 0x400, 0x80)
1536EXC_VIRT_BEGIN(instruction_access, 0x4400, 0x80)
1537	GEN_INT_ENTRY instruction_access, virt=1
1538EXC_VIRT_END(instruction_access, 0x4400, 0x80)
1539EXC_COMMON_BEGIN(instruction_access_common)
1540	GEN_COMMON instruction_access
1541	addi	r3,r1,STACK_INT_FRAME_REGS
1542#ifdef CONFIG_PPC_64S_HASH_MMU
1543BEGIN_MMU_FTR_SECTION
1544	bl	CFUNC(do_hash_fault)
1545MMU_FTR_SECTION_ELSE
1546	bl	CFUNC(do_page_fault)
1547ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1548#else
1549	bl	CFUNC(do_page_fault)
1550#endif
1551	b	interrupt_return_srr
1552
1553
1554/**
1555 * Interrupt 0x480 - Instruction Segment Interrupt (ISLB).
1556 * This is a synchronous interrupt in response to an MMU fault due to an
1557 * instruction fetch.
1558 *
1559 * Handling:
1560 * Similar to DSLB, though in response to fetch. The faulting address is found
1561 * in SRR0 (rather than DAR).
1562 */
1563INT_DEFINE_BEGIN(instruction_access_slb)
1564	IVEC=0x480
1565	IISIDE=1
1566	IDAR=1
1567#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1568	IKVM_REAL=1
1569#endif
1570INT_DEFINE_END(instruction_access_slb)
1571
1572EXC_REAL_BEGIN(instruction_access_slb, 0x480, 0x80)
1573	GEN_INT_ENTRY instruction_access_slb, virt=0
1574EXC_REAL_END(instruction_access_slb, 0x480, 0x80)
1575EXC_VIRT_BEGIN(instruction_access_slb, 0x4480, 0x80)
1576	GEN_INT_ENTRY instruction_access_slb, virt=1
1577EXC_VIRT_END(instruction_access_slb, 0x4480, 0x80)
1578EXC_COMMON_BEGIN(instruction_access_slb_common)
1579	GEN_COMMON instruction_access_slb
1580#ifdef CONFIG_PPC_64S_HASH_MMU
1581BEGIN_MMU_FTR_SECTION
1582	/* HPT case, do SLB fault */
1583	addi	r3,r1,STACK_INT_FRAME_REGS
1584	bl	CFUNC(do_slb_fault)
1585	cmpdi	r3,0
1586	bne-	1f
1587	b	fast_interrupt_return_srr
15881:	/* Error case */
1589MMU_FTR_SECTION_ELSE
1590	/* Radix case, access is outside page table range */
1591	li	r3,-EFAULT
1592ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_TYPE_RADIX)
1593#else
1594	li	r3,-EFAULT
1595#endif
1596	std	r3,RESULT(r1)
1597	addi	r3,r1,STACK_INT_FRAME_REGS
1598	bl	CFUNC(do_bad_segment_interrupt)
1599	b	interrupt_return_srr
 
 
 
 
 
 
 
 
 
 
 
1600
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1601
1602/**
1603 * Interrupt 0x500 - External Interrupt.
1604 * This is an asynchronous maskable interrupt in response to an "external
1605 * exception" from the interrupt controller or hypervisor (e.g., device
1606 * interrupt). It is maskable in hardware by clearing MSR[EE], and
1607 * soft-maskable with IRQS_DISABLED mask (i.e., local_irq_disable()).
1608 *
1609 * When running in HV mode, Linux sets up the LPCR[LPES] bit such that
1610 * interrupts are delivered with HSRR registers, guests use SRRs, which
1611 * reqiures IHSRR_IF_HVMODE.
1612 *
1613 * On bare metal POWER9 and later, Linux sets the LPCR[HVICE] bit such that
1614 * external interrupts are delivered as Hypervisor Virtualization Interrupts
1615 * rather than External Interrupts.
1616 *
1617 * Handling:
1618 * This calls into Linux IRQ handler. NVGPRs are not saved to reduce overhead,
1619 * because registers at the time of the interrupt are not so important as it is
1620 * asynchronous.
1621 *
1622 * If soft masked, the masked handler will note the pending interrupt for
1623 * replay, and clear MSR[EE] in the interrupted context.
1624 *
1625 * CFAR is not required because this is an asynchronous interrupt that in
1626 * general won't have much bearing on the state of the CPU, with the possible
1627 * exception of crash/debug IPIs, but those are generally moving to use SRESET
1628 * IPIs. Unless this is an HV interrupt and KVM HV is possible, in which case
1629 * it may be exiting the guest and need CFAR to be saved.
1630 */
1631INT_DEFINE_BEGIN(hardware_interrupt)
1632	IVEC=0x500
1633	IHSRR_IF_HVMODE=1
1634	IMASK=IRQS_DISABLED
1635	IKVM_REAL=1
1636	IKVM_VIRT=1
1637	ICFAR=0
1638#ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
1639	ICFAR_IF_HVMODE=1
1640#endif
1641INT_DEFINE_END(hardware_interrupt)
1642
1643EXC_REAL_BEGIN(hardware_interrupt, 0x500, 0x100)
1644	GEN_INT_ENTRY hardware_interrupt, virt=0
1645EXC_REAL_END(hardware_interrupt, 0x500, 0x100)
1646EXC_VIRT_BEGIN(hardware_interrupt, 0x4500, 0x100)
1647	GEN_INT_ENTRY hardware_interrupt, virt=1
1648EXC_VIRT_END(hardware_interrupt, 0x4500, 0x100)
1649EXC_COMMON_BEGIN(hardware_interrupt_common)
1650	GEN_COMMON hardware_interrupt
1651	addi	r3,r1,STACK_INT_FRAME_REGS
1652	bl	CFUNC(do_IRQ)
1653	BEGIN_FTR_SECTION
1654	b	interrupt_return_hsrr
1655	FTR_SECTION_ELSE
1656	b	interrupt_return_srr
1657	ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1658
 
1659
1660/**
1661 * Interrupt 0x600 - Alignment Interrupt
1662 * This is a synchronous interrupt in response to data alignment fault.
1663 */
1664INT_DEFINE_BEGIN(alignment)
1665	IVEC=0x600
1666	IDAR=1
1667	IDSISR=1
1668#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1669	IKVM_REAL=1
1670#endif
1671INT_DEFINE_END(alignment)
1672
1673EXC_REAL_BEGIN(alignment, 0x600, 0x100)
1674	GEN_INT_ENTRY alignment, virt=0
1675EXC_REAL_END(alignment, 0x600, 0x100)
1676EXC_VIRT_BEGIN(alignment, 0x4600, 0x100)
1677	GEN_INT_ENTRY alignment, virt=1
1678EXC_VIRT_END(alignment, 0x4600, 0x100)
1679EXC_COMMON_BEGIN(alignment_common)
1680	GEN_COMMON alignment
1681	addi	r3,r1,STACK_INT_FRAME_REGS
1682	bl	CFUNC(alignment_exception)
1683	HANDLER_RESTORE_NVGPRS() /* instruction emulation may change GPRs */
1684	b	interrupt_return_srr
1685
1686
1687/**
1688 * Interrupt 0x700 - Program Interrupt (program check).
1689 * This is a synchronous interrupt in response to various instruction faults:
1690 * traps, privilege errors, TM errors, floating point exceptions.
1691 *
1692 * Handling:
1693 * This interrupt may use the "emergency stack" in some cases when being taken
1694 * from kernel context, which complicates handling.
1695 */
1696INT_DEFINE_BEGIN(program_check)
1697	IVEC=0x700
1698#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1699	IKVM_REAL=1
1700#endif
1701INT_DEFINE_END(program_check)
1702
1703EXC_REAL_BEGIN(program_check, 0x700, 0x100)
1704	EARLY_BOOT_FIXUP
1705	GEN_INT_ENTRY program_check, virt=0
1706EXC_REAL_END(program_check, 0x700, 0x100)
1707EXC_VIRT_BEGIN(program_check, 0x4700, 0x100)
1708	GEN_INT_ENTRY program_check, virt=1
1709EXC_VIRT_END(program_check, 0x4700, 0x100)
1710EXC_COMMON_BEGIN(program_check_common)
1711	__GEN_COMMON_ENTRY program_check
1712
1713	/*
1714	 * It's possible to receive a TM Bad Thing type program check with
1715	 * userspace register values (in particular r1), but with SRR1 reporting
1716	 * that we came from the kernel. Normally that would confuse the bad
1717	 * stack logic, and we would report a bad kernel stack pointer. Instead
1718	 * we switch to the emergency stack if we're taking a TM Bad Thing from
1719	 * the kernel.
1720	 */
1721
1722	andi.	r10,r12,MSR_PR
1723	bne	.Lnormal_stack		/* If userspace, go normal path */
1724
1725	andis.	r10,r12,(SRR1_PROGTM)@h
1726	bne	.Lemergency_stack	/* If TM, emergency		*/
1727
1728	cmpdi	r1,-INT_FRAME_SIZE	/* check if r1 is in userspace	*/
1729	blt	.Lnormal_stack		/* normal path if not		*/
1730
1731	/* Use the emergency stack					*/
1732.Lemergency_stack:
1733	andi.	r10,r12,MSR_PR		/* Set CR0 correctly for label	*/
1734					/* 3 in EXCEPTION_PROLOG_COMMON	*/
1735	mr	r10,r1			/* Save r1			*/
1736	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack		*/
1737	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
1738	__ISTACK(program_check)=0
1739	__GEN_COMMON_BODY program_check
1740	b .Ldo_program_check
1741
1742.Lnormal_stack:
1743	__ISTACK(program_check)=1
1744	__GEN_COMMON_BODY program_check
1745
1746.Ldo_program_check:
1747	addi	r3,r1,STACK_INT_FRAME_REGS
1748	bl	CFUNC(program_check_exception)
1749	HANDLER_RESTORE_NVGPRS() /* instruction emulation may change GPRs */
1750	b	interrupt_return_srr
1751
 
 
1752
1753/*
1754 * Interrupt 0x800 - Floating-Point Unavailable Interrupt.
1755 * This is a synchronous interrupt in response to executing an fp instruction
1756 * with MSR[FP]=0.
1757 *
1758 * Handling:
1759 * This will load FP registers and enable the FP bit if coming from userspace,
1760 * otherwise report a bad kernel use of FP.
1761 */
1762INT_DEFINE_BEGIN(fp_unavailable)
1763	IVEC=0x800
1764#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1765	IKVM_REAL=1
1766#endif
1767	IMSR_R12=1
1768INT_DEFINE_END(fp_unavailable)
1769
1770EXC_REAL_BEGIN(fp_unavailable, 0x800, 0x100)
1771	GEN_INT_ENTRY fp_unavailable, virt=0
1772EXC_REAL_END(fp_unavailable, 0x800, 0x100)
1773EXC_VIRT_BEGIN(fp_unavailable, 0x4800, 0x100)
1774	GEN_INT_ENTRY fp_unavailable, virt=1
1775EXC_VIRT_END(fp_unavailable, 0x4800, 0x100)
1776EXC_COMMON_BEGIN(fp_unavailable_common)
1777	GEN_COMMON fp_unavailable
1778	bne	1f			/* if from user, just load it up */
1779	addi	r3,r1,STACK_INT_FRAME_REGS
1780	bl	CFUNC(kernel_fp_unavailable_exception)
17810:	trap
1782	EMIT_BUG_ENTRY 0b, __FILE__, __LINE__, 0
17831:
1784#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1785BEGIN_FTR_SECTION
1786	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
1787	 * transaction), go do TM stuff
1788	 */
1789	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
1790	bne-	2f
1791END_FTR_SECTION_IFSET(CPU_FTR_TM)
1792#endif
1793	bl	CFUNC(load_up_fpu)
1794	b	fast_interrupt_return_srr
1795#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
17962:	/* User process was in a transaction */
1797	addi	r3,r1,STACK_INT_FRAME_REGS
1798	bl	CFUNC(fp_unavailable_tm)
1799	b	interrupt_return_srr
1800#endif
1801
 
 
1802
1803/**
1804 * Interrupt 0x900 - Decrementer Interrupt.
1805 * This is an asynchronous interrupt in response to a decrementer exception
1806 * (e.g., DEC has wrapped below zero). It is maskable in hardware by clearing
1807 * MSR[EE], and soft-maskable with IRQS_DISABLED mask (i.e.,
1808 * local_irq_disable()).
1809 *
1810 * Handling:
1811 * This calls into Linux timer handler. NVGPRs are not saved (see 0x500).
1812 *
1813 * If soft masked, the masked handler will note the pending interrupt for
1814 * replay, and bump the decrementer to a high value, leaving MSR[EE] enabled
1815 * in the interrupted context.
1816 * If PPC_WATCHDOG is configured, the soft masked handler will actually set
1817 * things back up to run soft_nmi_interrupt as a regular interrupt handler
1818 * on the emergency stack.
1819 *
1820 * CFAR is not required because this is asynchronous (see hardware_interrupt).
1821 * A watchdog interrupt may like to have CFAR, but usually the interesting
1822 * branch is long gone by that point (e.g., infinite loop).
1823 */
1824INT_DEFINE_BEGIN(decrementer)
1825	IVEC=0x900
1826	IMASK=IRQS_DISABLED
1827#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1828	IKVM_REAL=1
1829#endif
1830	ICFAR=0
1831INT_DEFINE_END(decrementer)
1832
1833EXC_REAL_BEGIN(decrementer, 0x900, 0x80)
1834	GEN_INT_ENTRY decrementer, virt=0
1835EXC_REAL_END(decrementer, 0x900, 0x80)
1836EXC_VIRT_BEGIN(decrementer, 0x4900, 0x80)
1837	GEN_INT_ENTRY decrementer, virt=1
1838EXC_VIRT_END(decrementer, 0x4900, 0x80)
1839EXC_COMMON_BEGIN(decrementer_common)
1840	GEN_COMMON decrementer
1841	addi	r3,r1,STACK_INT_FRAME_REGS
1842	bl	CFUNC(timer_interrupt)
1843	b	interrupt_return_srr
1844
1845
1846/**
1847 * Interrupt 0x980 - Hypervisor Decrementer Interrupt.
1848 * This is an asynchronous interrupt, similar to 0x900 but for the HDEC
1849 * register.
1850 *
1851 * Handling:
1852 * Linux does not use this outside KVM where it's used to keep a host timer
1853 * while the guest is given control of DEC. It should normally be caught by
1854 * the KVM test and routed there.
1855 */
1856INT_DEFINE_BEGIN(hdecrementer)
1857	IVEC=0x980
1858	IHSRR=1
1859	ISTACK=0
1860	IKVM_REAL=1
1861	IKVM_VIRT=1
1862INT_DEFINE_END(hdecrementer)
1863
1864EXC_REAL_BEGIN(hdecrementer, 0x980, 0x80)
1865	GEN_INT_ENTRY hdecrementer, virt=0
1866EXC_REAL_END(hdecrementer, 0x980, 0x80)
1867EXC_VIRT_BEGIN(hdecrementer, 0x4980, 0x80)
1868	GEN_INT_ENTRY hdecrementer, virt=1
1869EXC_VIRT_END(hdecrementer, 0x4980, 0x80)
1870EXC_COMMON_BEGIN(hdecrementer_common)
1871	__GEN_COMMON_ENTRY hdecrementer
1872	/*
1873	 * Hypervisor decrementer interrupts not caught by the KVM test
1874	 * shouldn't occur but are sometimes left pending on exit from a KVM
1875	 * guest.  We don't need to do anything to clear them, as they are
1876	 * edge-triggered.
1877	 *
1878	 * Be careful to avoid touching the kernel stack.
1879	 */
1880	li	r10,0
1881	stb	r10,PACAHSRR_VALID(r13)
1882	ld	r10,PACA_EXGEN+EX_CTR(r13)
1883	mtctr	r10
1884	mtcrf	0x80,r9
1885	ld	r9,PACA_EXGEN+EX_R9(r13)
1886	ld	r10,PACA_EXGEN+EX_R10(r13)
1887	ld	r11,PACA_EXGEN+EX_R11(r13)
1888	ld	r12,PACA_EXGEN+EX_R12(r13)
1889	ld	r13,PACA_EXGEN+EX_R13(r13)
1890	HRFI_TO_KERNEL
1891
1892
1893/**
1894 * Interrupt 0xa00 - Directed Privileged Doorbell Interrupt.
1895 * This is an asynchronous interrupt in response to a msgsndp doorbell.
1896 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
1897 * IRQS_DISABLED mask (i.e., local_irq_disable()).
1898 *
1899 * Handling:
1900 * Guests may use this for IPIs between threads in a core if the
1901 * hypervisor supports it. NVGPRS are not saved (see 0x500).
1902 *
1903 * If soft masked, the masked handler will note the pending interrupt for
1904 * replay, leaving MSR[EE] enabled in the interrupted context because the
1905 * doorbells are edge triggered.
1906 *
1907 * CFAR is not required, similarly to hardware_interrupt.
1908 */
1909INT_DEFINE_BEGIN(doorbell_super)
1910	IVEC=0xa00
1911	IMASK=IRQS_DISABLED
1912#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1913	IKVM_REAL=1
1914#endif
1915	ICFAR=0
1916INT_DEFINE_END(doorbell_super)
1917
1918EXC_REAL_BEGIN(doorbell_super, 0xa00, 0x100)
1919	GEN_INT_ENTRY doorbell_super, virt=0
1920EXC_REAL_END(doorbell_super, 0xa00, 0x100)
1921EXC_VIRT_BEGIN(doorbell_super, 0x4a00, 0x100)
1922	GEN_INT_ENTRY doorbell_super, virt=1
1923EXC_VIRT_END(doorbell_super, 0x4a00, 0x100)
1924EXC_COMMON_BEGIN(doorbell_super_common)
1925	GEN_COMMON doorbell_super
1926	addi	r3,r1,STACK_INT_FRAME_REGS
1927#ifdef CONFIG_PPC_DOORBELL
1928	bl	CFUNC(doorbell_exception)
1929#else
1930	bl	CFUNC(unknown_async_exception)
1931#endif
1932	b	interrupt_return_srr
1933
 
 
 
 
 
1934
1935EXC_REAL_NONE(0xb00, 0x100)
1936EXC_VIRT_NONE(0x4b00, 0x100)
 
 
 
1937
1938/**
1939 * Interrupt 0xc00 - System Call Interrupt (syscall, hcall).
1940 * This is a synchronous interrupt invoked with the "sc" instruction. The
1941 * system call is invoked with "sc 0" and does not alter the HV bit, so it
1942 * is directed to the currently running OS. The hypercall is invoked with
1943 * "sc 1" and it sets HV=1, so it elevates to hypervisor.
1944 *
1945 * In HPT, sc 1 always goes to 0xc00 real mode. In RADIX, sc 1 can go to
1946 * 0x4c00 virtual mode.
1947 *
1948 * Handling:
1949 * If the KVM test fires then it was due to a hypercall and is accordingly
1950 * routed to KVM. Otherwise this executes a normal Linux system call.
1951 *
1952 * Call convention:
1953 *
1954 * syscall and hypercalls register conventions are documented in
1955 * Documentation/arch/powerpc/syscall64-abi.rst and
1956 * Documentation/arch/powerpc/papr_hcalls.rst respectively.
1957 *
1958 * The intersection of volatile registers that don't contain possible
1959 * inputs is: cr0, xer, ctr. We may use these as scratch regs upon entry
1960 * without saving, though xer is not a good idea to use, as hardware may
1961 * interpret some bits so it may be costly to change them.
1962 */
1963INT_DEFINE_BEGIN(system_call)
1964	IVEC=0xc00
1965	IKVM_REAL=1
1966	IKVM_VIRT=1
1967	ICFAR=0
1968INT_DEFINE_END(system_call)
1969
1970.macro SYSTEM_CALL virt
1971#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1972	/*
1973	 * There is a little bit of juggling to get syscall and hcall
1974	 * working well. Save r13 in ctr to avoid using SPRG scratch
1975	 * register.
1976	 *
1977	 * Userspace syscalls have already saved the PPR, hcalls must save
1978	 * it before setting HMT_MEDIUM.
1979	 */
1980	mtctr	r13
1981	GET_PACA(r13)
1982	std	r10,PACA_EXGEN+EX_R10(r13)
1983	INTERRUPT_TO_KERNEL
1984	KVMTEST system_call kvm_hcall /* uses r10, branch to kvm_hcall */
1985	mfctr	r9
1986#else
1987	mr	r9,r13
1988	GET_PACA(r13)
1989	INTERRUPT_TO_KERNEL
1990#endif
1991
1992#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
1993BEGIN_FTR_SECTION
1994	cmpdi	r0,0x1ebe
1995	beq-	1f
1996END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE)
1997#endif
 
1998
1999	/* We reach here with PACA in r13, r13 in r9. */
2000	mfspr	r11,SPRN_SRR0
2001	mfspr	r12,SPRN_SRR1
2002
2003	HMT_MEDIUM
2004
2005	.if ! \virt
2006	__LOAD_HANDLER(r10, system_call_common_real, real_vectors)
2007	mtctr	r10
2008	bctr
2009	.else
2010#ifdef CONFIG_RELOCATABLE
2011	__LOAD_HANDLER(r10, system_call_common, virt_vectors)
2012	mtctr	r10
2013	bctr
2014#else
2015	b	system_call_common
2016#endif
2017	.endif
2018
2019#ifdef CONFIG_PPC_FAST_ENDIAN_SWITCH
2020	/* Fast LE/BE switch system call */
20211:	mfspr	r12,SPRN_SRR1
2022	xori	r12,r12,MSR_LE
2023	mtspr	SPRN_SRR1,r12
2024	mr	r13,r9
2025	RFI_TO_USER	/* return to userspace */
2026	b	.	/* prevent speculative execution */
2027#endif
2028.endm
2029
2030EXC_REAL_BEGIN(system_call, 0xc00, 0x100)
2031	SYSTEM_CALL 0
2032EXC_REAL_END(system_call, 0xc00, 0x100)
2033EXC_VIRT_BEGIN(system_call, 0x4c00, 0x100)
2034	SYSTEM_CALL 1
2035EXC_VIRT_END(system_call, 0x4c00, 0x100)
2036
2037#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
2038TRAMP_REAL_BEGIN(kvm_hcall)
2039	std	r9,PACA_EXGEN+EX_R9(r13)
2040	std	r11,PACA_EXGEN+EX_R11(r13)
2041	std	r12,PACA_EXGEN+EX_R12(r13)
2042	mfcr	r9
2043	mfctr	r10
2044	std	r10,PACA_EXGEN+EX_R13(r13)
2045	li	r10,0
2046	std	r10,PACA_EXGEN+EX_CFAR(r13)
2047	std	r10,PACA_EXGEN+EX_CTR(r13)
2048	 /*
2049	  * Save the PPR (on systems that support it) before changing to
2050	  * HMT_MEDIUM. That allows the KVM code to save that value into the
2051	  * guest state (it is the guest's PPR value).
2052	  */
2053BEGIN_FTR_SECTION
2054	mfspr	r10,SPRN_PPR
2055	std	r10,PACA_EXGEN+EX_PPR(r13)
2056END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
 
 
 
 
 
 
 
 
 
 
 
 
 
2057
2058	HMT_MEDIUM
 
 
 
 
 
 
 
 
 
2059
2060#ifdef CONFIG_RELOCATABLE
2061	/*
2062	 * Requires __LOAD_FAR_HANDLER beause kvmppc_hcall lives
2063	 * outside the head section.
2064	 */
2065	__LOAD_FAR_HANDLER(r10, kvmppc_hcall, real_trampolines)
2066	mtctr   r10
2067	bctr
2068#else
2069	b       kvmppc_hcall
2070#endif
2071#endif
2072
2073/**
2074 * Interrupt 0xd00 - Trace Interrupt.
2075 * This is a synchronous interrupt in response to instruction step or
2076 * breakpoint faults.
2077 */
2078INT_DEFINE_BEGIN(single_step)
2079	IVEC=0xd00
2080#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2081	IKVM_REAL=1
2082#endif
2083INT_DEFINE_END(single_step)
2084
2085EXC_REAL_BEGIN(single_step, 0xd00, 0x100)
2086	GEN_INT_ENTRY single_step, virt=0
2087EXC_REAL_END(single_step, 0xd00, 0x100)
2088EXC_VIRT_BEGIN(single_step, 0x4d00, 0x100)
2089	GEN_INT_ENTRY single_step, virt=1
2090EXC_VIRT_END(single_step, 0x4d00, 0x100)
2091EXC_COMMON_BEGIN(single_step_common)
2092	GEN_COMMON single_step
2093	addi	r3,r1,STACK_INT_FRAME_REGS
2094	bl	CFUNC(single_step_exception)
2095	b	interrupt_return_srr
2096
2097
2098/**
2099 * Interrupt 0xe00 - Hypervisor Data Storage Interrupt (HDSI).
2100 * This is a synchronous interrupt in response to an MMU fault caused by a
2101 * guest data access.
2102 *
2103 * Handling:
2104 * This should always get routed to KVM. In radix MMU mode, this is caused
2105 * by a guest nested radix access that can't be performed due to the
2106 * partition scope page table. In hash mode, this can be caused by guests
2107 * running with translation disabled (virtual real mode) or with VPM enabled.
2108 * KVM will update the page table structures or disallow the access.
2109 */
2110INT_DEFINE_BEGIN(h_data_storage)
2111	IVEC=0xe00
2112	IHSRR=1
2113	IDAR=1
2114	IDSISR=1
2115	IKVM_REAL=1
2116	IKVM_VIRT=1
2117INT_DEFINE_END(h_data_storage)
2118
2119EXC_REAL_BEGIN(h_data_storage, 0xe00, 0x20)
2120	GEN_INT_ENTRY h_data_storage, virt=0, ool=1
2121EXC_REAL_END(h_data_storage, 0xe00, 0x20)
2122EXC_VIRT_BEGIN(h_data_storage, 0x4e00, 0x20)
2123	GEN_INT_ENTRY h_data_storage, virt=1, ool=1
2124EXC_VIRT_END(h_data_storage, 0x4e00, 0x20)
2125EXC_COMMON_BEGIN(h_data_storage_common)
2126	GEN_COMMON h_data_storage
2127	addi    r3,r1,STACK_INT_FRAME_REGS
2128BEGIN_MMU_FTR_SECTION
2129	bl	CFUNC(do_bad_page_fault_segv)
2130MMU_FTR_SECTION_ELSE
2131	bl	CFUNC(unknown_exception)
2132ALT_MMU_FTR_SECTION_END_IFSET(MMU_FTR_TYPE_RADIX)
2133	b       interrupt_return_hsrr
2134
2135
2136/**
2137 * Interrupt 0xe20 - Hypervisor Instruction Storage Interrupt (HISI).
2138 * This is a synchronous interrupt in response to an MMU fault caused by a
2139 * guest instruction fetch, similar to HDSI.
2140 */
2141INT_DEFINE_BEGIN(h_instr_storage)
2142	IVEC=0xe20
2143	IHSRR=1
2144	IKVM_REAL=1
2145	IKVM_VIRT=1
2146INT_DEFINE_END(h_instr_storage)
2147
2148EXC_REAL_BEGIN(h_instr_storage, 0xe20, 0x20)
2149	GEN_INT_ENTRY h_instr_storage, virt=0, ool=1
2150EXC_REAL_END(h_instr_storage, 0xe20, 0x20)
2151EXC_VIRT_BEGIN(h_instr_storage, 0x4e20, 0x20)
2152	GEN_INT_ENTRY h_instr_storage, virt=1, ool=1
2153EXC_VIRT_END(h_instr_storage, 0x4e20, 0x20)
2154EXC_COMMON_BEGIN(h_instr_storage_common)
2155	GEN_COMMON h_instr_storage
2156	addi	r3,r1,STACK_INT_FRAME_REGS
2157	bl	CFUNC(unknown_exception)
2158	b	interrupt_return_hsrr
2159
2160
2161/**
2162 * Interrupt 0xe40 - Hypervisor Emulation Assistance Interrupt.
2163 */
2164INT_DEFINE_BEGIN(emulation_assist)
2165	IVEC=0xe40
2166	IHSRR=1
2167	IKVM_REAL=1
2168	IKVM_VIRT=1
2169INT_DEFINE_END(emulation_assist)
2170
2171EXC_REAL_BEGIN(emulation_assist, 0xe40, 0x20)
2172	GEN_INT_ENTRY emulation_assist, virt=0, ool=1
2173EXC_REAL_END(emulation_assist, 0xe40, 0x20)
2174EXC_VIRT_BEGIN(emulation_assist, 0x4e40, 0x20)
2175	GEN_INT_ENTRY emulation_assist, virt=1, ool=1
2176EXC_VIRT_END(emulation_assist, 0x4e40, 0x20)
2177EXC_COMMON_BEGIN(emulation_assist_common)
2178	GEN_COMMON emulation_assist
2179	addi	r3,r1,STACK_INT_FRAME_REGS
2180	bl	CFUNC(emulation_assist_interrupt)
2181	HANDLER_RESTORE_NVGPRS() /* instruction emulation may change GPRs */
2182	b	interrupt_return_hsrr
2183
2184
2185/**
2186 * Interrupt 0xe60 - Hypervisor Maintenance Interrupt (HMI).
2187 * This is an asynchronous interrupt caused by a Hypervisor Maintenance
2188 * Exception. It is always taken in real mode but uses HSRR registers
2189 * unlike SRESET and MCE.
2190 *
2191 * It is maskable in hardware by clearing MSR[EE], and partially soft-maskable
2192 * with IRQS_DISABLED mask (i.e., local_irq_disable()).
2193 *
2194 * Handling:
2195 * This is a special case, this is handled similarly to machine checks, with an
2196 * initial real mode handler that is not soft-masked, which attempts to fix the
2197 * problem. Then a regular handler which is soft-maskable and reports the
2198 * problem.
2199 *
2200 * The emergency stack is used for the early real mode handler.
2201 *
2202 * XXX: unclear why MCE and HMI schemes could not be made common, e.g.,
2203 * either use soft-masking for the MCE, or use irq_work for the HMI.
2204 *
2205 * KVM:
2206 * Unlike MCE, this calls into KVM without calling the real mode handler
2207 * first.
2208 */
2209INT_DEFINE_BEGIN(hmi_exception_early)
2210	IVEC=0xe60
2211	IHSRR=1
2212	IREALMODE_COMMON=1
2213	ISTACK=0
2214	IKUAP=0 /* We don't touch AMR here, we never go to virtual mode */
2215	IKVM_REAL=1
2216INT_DEFINE_END(hmi_exception_early)
2217
2218INT_DEFINE_BEGIN(hmi_exception)
2219	IVEC=0xe60
2220	IHSRR=1
2221	IMASK=IRQS_DISABLED
2222	IKVM_REAL=1
2223INT_DEFINE_END(hmi_exception)
2224
2225EXC_REAL_BEGIN(hmi_exception, 0xe60, 0x20)
2226	GEN_INT_ENTRY hmi_exception_early, virt=0, ool=1
2227EXC_REAL_END(hmi_exception, 0xe60, 0x20)
2228EXC_VIRT_NONE(0x4e60, 0x20)
2229
2230EXC_COMMON_BEGIN(hmi_exception_early_common)
2231	__GEN_REALMODE_COMMON_ENTRY hmi_exception_early
 
2232
2233	mr	r10,r1			/* Save r1 */
2234	ld	r1,PACAEMERGSP(r13)	/* Use emergency stack for realmode */
2235	subi	r1,r1,INT_FRAME_SIZE	/* alloc stack frame		*/
 
2236
2237	__GEN_COMMON_BODY hmi_exception_early
 
2238
2239	addi	r3,r1,STACK_INT_FRAME_REGS
2240	bl	CFUNC(hmi_exception_realmode)
2241	cmpdi	cr0,r3,0
2242	bne	1f
2243
2244	EXCEPTION_RESTORE_REGS hsrr=1
2245	HRFI_TO_USER_OR_KERNEL
 
 
 
 
 
2246
22471:
2248	/*
2249	 * Go to virtual mode and pull the HMI event information from
2250	 * firmware.
2251	 */
2252	EXCEPTION_RESTORE_REGS hsrr=1
2253	GEN_INT_ENTRY hmi_exception, virt=0
 
 
 
 
 
 
 
 
2254
2255EXC_COMMON_BEGIN(hmi_exception_common)
2256	GEN_COMMON hmi_exception
2257	addi	r3,r1,STACK_INT_FRAME_REGS
2258	bl	CFUNC(handle_hmi_exception)
2259	b	interrupt_return_hsrr
2260
2261
2262/**
2263 * Interrupt 0xe80 - Directed Hypervisor Doorbell Interrupt.
2264 * This is an asynchronous interrupt in response to a msgsnd doorbell.
2265 * Similar to the 0xa00 doorbell but for host rather than guest.
2266 *
2267 * CFAR is not required (similar to doorbell_interrupt), unless KVM HV
2268 * is enabled, in which case it may be a guest exit. Most PowerNV kernels
2269 * include KVM support so it would be nice if this could be dynamically
2270 * patched out if KVM was not currently running any guests.
2271 */
2272INT_DEFINE_BEGIN(h_doorbell)
2273	IVEC=0xe80
2274	IHSRR=1
2275	IMASK=IRQS_DISABLED
2276	IKVM_REAL=1
2277	IKVM_VIRT=1
2278#ifndef CONFIG_KVM_BOOK3S_HV_POSSIBLE
2279	ICFAR=0
2280#endif
2281INT_DEFINE_END(h_doorbell)
2282
2283EXC_REAL_BEGIN(h_doorbell, 0xe80, 0x20)
2284	GEN_INT_ENTRY h_doorbell, virt=0, ool=1
2285EXC_REAL_END(h_doorbell, 0xe80, 0x20)
2286EXC_VIRT_BEGIN(h_doorbell, 0x4e80, 0x20)
2287	GEN_INT_ENTRY h_doorbell, virt=1, ool=1
2288EXC_VIRT_END(h_doorbell, 0x4e80, 0x20)
2289EXC_COMMON_BEGIN(h_doorbell_common)
2290	GEN_COMMON h_doorbell
2291	addi	r3,r1,STACK_INT_FRAME_REGS
2292#ifdef CONFIG_PPC_DOORBELL
2293	bl	CFUNC(doorbell_exception)
2294#else
2295	bl	CFUNC(unknown_async_exception)
2296#endif
2297	b	interrupt_return_hsrr
2298
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2299
2300/**
2301 * Interrupt 0xea0 - Hypervisor Virtualization Interrupt.
2302 * This is an asynchronous interrupt in response to an "external exception".
2303 * Similar to 0x500 but for host only.
2304 *
2305 * Like h_doorbell, CFAR is only required for KVM HV because this can be
2306 * a guest exit.
2307 */
2308INT_DEFINE_BEGIN(h_virt_irq)
2309	IVEC=0xea0
2310	IHSRR=1
2311	IMASK=IRQS_DISABLED
2312	IKVM_REAL=1
2313	IKVM_VIRT=1
2314#ifndef CONFIG_KVM_BOOK3S_HV_POSSIBLE
2315	ICFAR=0
2316#endif
2317INT_DEFINE_END(h_virt_irq)
2318
2319EXC_REAL_BEGIN(h_virt_irq, 0xea0, 0x20)
2320	GEN_INT_ENTRY h_virt_irq, virt=0, ool=1
2321EXC_REAL_END(h_virt_irq, 0xea0, 0x20)
2322EXC_VIRT_BEGIN(h_virt_irq, 0x4ea0, 0x20)
2323	GEN_INT_ENTRY h_virt_irq, virt=1, ool=1
2324EXC_VIRT_END(h_virt_irq, 0x4ea0, 0x20)
2325EXC_COMMON_BEGIN(h_virt_irq_common)
2326	GEN_COMMON h_virt_irq
2327	addi	r3,r1,STACK_INT_FRAME_REGS
2328	bl	CFUNC(do_IRQ)
2329	b	interrupt_return_hsrr
2330
2331
2332EXC_REAL_NONE(0xec0, 0x20)
2333EXC_VIRT_NONE(0x4ec0, 0x20)
2334EXC_REAL_NONE(0xee0, 0x20)
2335EXC_VIRT_NONE(0x4ee0, 0x20)
2336
 
 
 
 
 
 
 
 
 
2337
2338/*
2339 * Interrupt 0xf00 - Performance Monitor Interrupt (PMI, PMU).
2340 * This is an asynchronous interrupt in response to a PMU exception.
2341 * It is maskable in hardware by clearing MSR[EE], and soft-maskable with
2342 * IRQS_PMI_DISABLED mask (NOTE: NOT local_irq_disable()).
2343 *
2344 * Handling:
2345 * This calls into the perf subsystem.
2346 *
2347 * Like the watchdog soft-nmi, it appears an NMI interrupt to Linux, in that it
2348 * runs under local_irq_disable. However it may be soft-masked in
2349 * powerpc-specific code.
2350 *
2351 * If soft masked, the masked handler will note the pending interrupt for
2352 * replay, and clear MSR[EE] in the interrupted context.
2353 *
2354 * CFAR is not used by perf interrupts so not required.
2355 */
2356INT_DEFINE_BEGIN(performance_monitor)
2357	IVEC=0xf00
2358	IMASK=IRQS_PMI_DISABLED
2359#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2360	IKVM_REAL=1
2361#endif
2362	ICFAR=0
2363INT_DEFINE_END(performance_monitor)
2364
2365EXC_REAL_BEGIN(performance_monitor, 0xf00, 0x20)
2366	GEN_INT_ENTRY performance_monitor, virt=0, ool=1
2367EXC_REAL_END(performance_monitor, 0xf00, 0x20)
2368EXC_VIRT_BEGIN(performance_monitor, 0x4f00, 0x20)
2369	GEN_INT_ENTRY performance_monitor, virt=1, ool=1
2370EXC_VIRT_END(performance_monitor, 0x4f00, 0x20)
2371EXC_COMMON_BEGIN(performance_monitor_common)
2372	GEN_COMMON performance_monitor
2373	addi	r3,r1,STACK_INT_FRAME_REGS
2374	lbz	r4,PACAIRQSOFTMASK(r13)
2375	cmpdi	r4,IRQS_ENABLED
2376	bne	1f
2377	bl	CFUNC(performance_monitor_exception_async)
2378	b	interrupt_return_srr
23791:
2380	bl	CFUNC(performance_monitor_exception_nmi)
2381	/* Clear MSR_RI before setting SRR0 and SRR1. */
2382	li	r9,0
2383	mtmsrd	r9,1
2384
2385	kuap_kernel_restore r9, r10
2386
2387	EXCEPTION_RESTORE_REGS hsrr=0
2388	RFI_TO_KERNEL
2389
2390/**
2391 * Interrupt 0xf20 - Vector Unavailable Interrupt.
2392 * This is a synchronous interrupt in response to
2393 * executing a vector (or altivec) instruction with MSR[VEC]=0.
2394 * Similar to FP unavailable.
2395 */
2396INT_DEFINE_BEGIN(altivec_unavailable)
2397	IVEC=0xf20
2398#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2399	IKVM_REAL=1
2400#endif
2401	IMSR_R12=1
2402INT_DEFINE_END(altivec_unavailable)
2403
2404EXC_REAL_BEGIN(altivec_unavailable, 0xf20, 0x20)
2405	GEN_INT_ENTRY altivec_unavailable, virt=0, ool=1
2406EXC_REAL_END(altivec_unavailable, 0xf20, 0x20)
2407EXC_VIRT_BEGIN(altivec_unavailable, 0x4f20, 0x20)
2408	GEN_INT_ENTRY altivec_unavailable, virt=1, ool=1
2409EXC_VIRT_END(altivec_unavailable, 0x4f20, 0x20)
2410EXC_COMMON_BEGIN(altivec_unavailable_common)
2411	GEN_COMMON altivec_unavailable
2412#ifdef CONFIG_ALTIVEC
2413BEGIN_FTR_SECTION
2414	beq	1f
2415#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2416  BEGIN_FTR_SECTION_NESTED(69)
2417	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2418	 * transaction), go do TM stuff
2419	 */
2420	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2421	bne-	2f
2422  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2423#endif
2424	bl	CFUNC(load_up_altivec)
2425	b	fast_interrupt_return_srr
2426#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
24272:	/* User process was in a transaction */
2428	addi	r3,r1,STACK_INT_FRAME_REGS
2429	bl	CFUNC(altivec_unavailable_tm)
2430	b	interrupt_return_srr
 
 
2431#endif
24321:
2433END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2434#endif
2435	addi	r3,r1,STACK_INT_FRAME_REGS
2436	bl	CFUNC(altivec_unavailable_exception)
2437	b	interrupt_return_srr
2438
 
2439
2440/**
2441 * Interrupt 0xf40 - VSX Unavailable Interrupt.
2442 * This is a synchronous interrupt in response to
2443 * executing a VSX instruction with MSR[VSX]=0.
2444 * Similar to FP unavailable.
2445 */
2446INT_DEFINE_BEGIN(vsx_unavailable)
2447	IVEC=0xf40
2448#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2449	IKVM_REAL=1
2450#endif
2451	IMSR_R12=1
2452INT_DEFINE_END(vsx_unavailable)
2453
2454EXC_REAL_BEGIN(vsx_unavailable, 0xf40, 0x20)
2455	GEN_INT_ENTRY vsx_unavailable, virt=0, ool=1
2456EXC_REAL_END(vsx_unavailable, 0xf40, 0x20)
2457EXC_VIRT_BEGIN(vsx_unavailable, 0x4f40, 0x20)
2458	GEN_INT_ENTRY vsx_unavailable, virt=1, ool=1
2459EXC_VIRT_END(vsx_unavailable, 0x4f40, 0x20)
2460EXC_COMMON_BEGIN(vsx_unavailable_common)
2461	GEN_COMMON vsx_unavailable
2462#ifdef CONFIG_VSX
2463BEGIN_FTR_SECTION
2464	beq	1f
2465#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2466  BEGIN_FTR_SECTION_NESTED(69)
2467	/* Test if 2 TM state bits are zero.  If non-zero (ie. userspace was in
2468	 * transaction), go do TM stuff
2469	 */
2470	rldicl.	r0, r12, (64-MSR_TS_LG), (64-2)
2471	bne-	2f
2472  END_FTR_SECTION_NESTED(CPU_FTR_TM, CPU_FTR_TM, 69)
2473#endif
2474	b	load_up_vsx
2475#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
24762:	/* User process was in a transaction */
2477	addi	r3,r1,STACK_INT_FRAME_REGS
2478	bl	CFUNC(vsx_unavailable_tm)
2479	b	interrupt_return_srr
 
 
2480#endif
24811:
2482END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2483#endif
2484	addi	r3,r1,STACK_INT_FRAME_REGS
2485	bl	CFUNC(vsx_unavailable_exception)
2486	b	interrupt_return_srr
2487
2488
2489/**
2490 * Interrupt 0xf60 - Facility Unavailable Interrupt.
2491 * This is a synchronous interrupt in response to
2492 * executing an instruction without access to the facility that can be
2493 * resolved by the OS (e.g., FSCR, MSR).
2494 * Similar to FP unavailable.
2495 */
2496INT_DEFINE_BEGIN(facility_unavailable)
2497	IVEC=0xf60
2498#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2499	IKVM_REAL=1
2500#endif
2501INT_DEFINE_END(facility_unavailable)
2502
2503EXC_REAL_BEGIN(facility_unavailable, 0xf60, 0x20)
2504	GEN_INT_ENTRY facility_unavailable, virt=0, ool=1
2505EXC_REAL_END(facility_unavailable, 0xf60, 0x20)
2506EXC_VIRT_BEGIN(facility_unavailable, 0x4f60, 0x20)
2507	GEN_INT_ENTRY facility_unavailable, virt=1, ool=1
2508EXC_VIRT_END(facility_unavailable, 0x4f60, 0x20)
2509EXC_COMMON_BEGIN(facility_unavailable_common)
2510	GEN_COMMON facility_unavailable
2511	addi	r3,r1,STACK_INT_FRAME_REGS
2512	bl	CFUNC(facility_unavailable_exception)
2513	HANDLER_RESTORE_NVGPRS() /* instruction emulation may change GPRs */
2514	b	interrupt_return_srr
2515
2516
2517/**
2518 * Interrupt 0xf60 - Hypervisor Facility Unavailable Interrupt.
2519 * This is a synchronous interrupt in response to
2520 * executing an instruction without access to the facility that can only
2521 * be resolved in HV mode (e.g., HFSCR).
2522 * Similar to FP unavailable.
2523 */
2524INT_DEFINE_BEGIN(h_facility_unavailable)
2525	IVEC=0xf80
2526	IHSRR=1
2527	IKVM_REAL=1
2528	IKVM_VIRT=1
2529INT_DEFINE_END(h_facility_unavailable)
2530
2531EXC_REAL_BEGIN(h_facility_unavailable, 0xf80, 0x20)
2532	GEN_INT_ENTRY h_facility_unavailable, virt=0, ool=1
2533EXC_REAL_END(h_facility_unavailable, 0xf80, 0x20)
2534EXC_VIRT_BEGIN(h_facility_unavailable, 0x4f80, 0x20)
2535	GEN_INT_ENTRY h_facility_unavailable, virt=1, ool=1
2536EXC_VIRT_END(h_facility_unavailable, 0x4f80, 0x20)
2537EXC_COMMON_BEGIN(h_facility_unavailable_common)
2538	GEN_COMMON h_facility_unavailable
2539	addi	r3,r1,STACK_INT_FRAME_REGS
2540	bl	CFUNC(facility_unavailable_exception)
2541	/* XXX Shouldn't be necessary in practice */
2542	HANDLER_RESTORE_NVGPRS()
2543	b	interrupt_return_hsrr
2544
2545
2546EXC_REAL_NONE(0xfa0, 0x20)
2547EXC_VIRT_NONE(0x4fa0, 0x20)
2548EXC_REAL_NONE(0xfc0, 0x20)
2549EXC_VIRT_NONE(0x4fc0, 0x20)
2550EXC_REAL_NONE(0xfe0, 0x20)
2551EXC_VIRT_NONE(0x4fe0, 0x20)
2552
2553EXC_REAL_NONE(0x1000, 0x100)
2554EXC_VIRT_NONE(0x5000, 0x100)
2555EXC_REAL_NONE(0x1100, 0x100)
2556EXC_VIRT_NONE(0x5100, 0x100)
2557
2558#ifdef CONFIG_CBE_RAS
2559INT_DEFINE_BEGIN(cbe_system_error)
2560	IVEC=0x1200
2561	IHSRR=1
2562INT_DEFINE_END(cbe_system_error)
2563
2564EXC_REAL_BEGIN(cbe_system_error, 0x1200, 0x100)
2565	GEN_INT_ENTRY cbe_system_error, virt=0
2566EXC_REAL_END(cbe_system_error, 0x1200, 0x100)
2567EXC_VIRT_NONE(0x5200, 0x100)
2568EXC_COMMON_BEGIN(cbe_system_error_common)
2569	GEN_COMMON cbe_system_error
2570	addi	r3,r1,STACK_INT_FRAME_REGS
2571	bl	CFUNC(cbe_system_error_exception)
2572	b	interrupt_return_hsrr
2573
2574#else /* CONFIG_CBE_RAS */
2575EXC_REAL_NONE(0x1200, 0x100)
2576EXC_VIRT_NONE(0x5200, 0x100)
2577#endif
2578
2579/**
2580 * Interrupt 0x1300 - Instruction Address Breakpoint Interrupt.
2581 * This has been removed from the ISA before 2.01, which is the earliest
2582 * 64-bit BookS ISA supported, however the G5 / 970 implements this
2583 * interrupt with a non-architected feature available through the support
2584 * processor interface.
2585 */
2586INT_DEFINE_BEGIN(instruction_breakpoint)
2587	IVEC=0x1300
2588#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2589	IKVM_REAL=1
2590#endif
2591INT_DEFINE_END(instruction_breakpoint)
2592
2593EXC_REAL_BEGIN(instruction_breakpoint, 0x1300, 0x100)
2594	GEN_INT_ENTRY instruction_breakpoint, virt=0
2595EXC_REAL_END(instruction_breakpoint, 0x1300, 0x100)
2596EXC_VIRT_BEGIN(instruction_breakpoint, 0x5300, 0x100)
2597	GEN_INT_ENTRY instruction_breakpoint, virt=1
2598EXC_VIRT_END(instruction_breakpoint, 0x5300, 0x100)
2599EXC_COMMON_BEGIN(instruction_breakpoint_common)
2600	GEN_COMMON instruction_breakpoint
2601	addi	r3,r1,STACK_INT_FRAME_REGS
2602	bl	CFUNC(instruction_breakpoint_exception)
2603	b	interrupt_return_srr
2604
2605
2606EXC_REAL_NONE(0x1400, 0x100)
2607EXC_VIRT_NONE(0x5400, 0x100)
2608
2609/**
2610 * Interrupt 0x1500 - Soft Patch Interrupt
2611 *
2612 * Handling:
2613 * This is an implementation specific interrupt which can be used for a
2614 * range of exceptions.
2615 *
2616 * This interrupt handler is unique in that it runs the denormal assist
2617 * code even for guests (and even in guest context) without going to KVM,
2618 * for speed. POWER9 does not raise denorm exceptions, so this special case
2619 * could be phased out in future to reduce special cases.
2620 */
2621INT_DEFINE_BEGIN(denorm_exception)
2622	IVEC=0x1500
2623	IHSRR=1
2624	IBRANCH_TO_COMMON=0
2625	IKVM_REAL=1
2626INT_DEFINE_END(denorm_exception)
2627
2628EXC_REAL_BEGIN(denorm_exception, 0x1500, 0x100)
2629	GEN_INT_ENTRY denorm_exception, virt=0
2630#ifdef CONFIG_PPC_DENORMALISATION
2631	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2632	bne+	denorm_assist
2633#endif
2634	GEN_BRANCH_TO_COMMON denorm_exception, virt=0
2635EXC_REAL_END(denorm_exception, 0x1500, 0x100)
2636#ifdef CONFIG_PPC_DENORMALISATION
2637EXC_VIRT_BEGIN(denorm_exception, 0x5500, 0x100)
2638	GEN_INT_ENTRY denorm_exception, virt=1
2639	andis.	r10,r12,(HSRR1_DENORM)@h /* denorm? */
2640	bne+	denorm_assist
2641	GEN_BRANCH_TO_COMMON denorm_exception, virt=1
2642EXC_VIRT_END(denorm_exception, 0x5500, 0x100)
2643#else
2644EXC_VIRT_NONE(0x5500, 0x100)
2645#endif
2646
2647#ifdef CONFIG_PPC_DENORMALISATION
2648TRAMP_REAL_BEGIN(denorm_assist)
2649BEGIN_FTR_SECTION
2650/*
2651 * To denormalise we need to move a copy of the register to itself.
2652 * For POWER6 do that here for all FP regs.
2653 */
2654	mfmsr	r10
2655	ori	r10,r10,(MSR_FP|MSR_FE0|MSR_FE1)
2656	xori	r10,r10,(MSR_FE0|MSR_FE1)
2657	mtmsrd	r10
2658	sync
2659
2660	.Lreg=0
2661	.rept 32
2662	fmr	.Lreg,.Lreg
2663	.Lreg=.Lreg+1
2664	.endr
2665
2666FTR_SECTION_ELSE
2667/*
2668 * To denormalise we need to move a copy of the register to itself.
2669 * For POWER7 do that here for the first 32 VSX registers only.
2670 */
2671	mfmsr	r10
2672	oris	r10,r10,MSR_VSX@h
2673	mtmsrd	r10
2674	sync
2675
2676	.Lreg=0
2677	.rept 32
2678	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2679	.Lreg=.Lreg+1
2680	.endr
2681
2682ALT_FTR_SECTION_END_IFCLR(CPU_FTR_ARCH_206)
 
 
 
 
 
 
 
 
2683
2684BEGIN_FTR_SECTION
2685	b	denorm_done
2686END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
2687/*
2688 * To denormalise we need to move a copy of the register to itself.
2689 * For POWER8 we need to do that for all 64 VSX registers
2690 */
2691	.Lreg=32
2692	.rept 32
2693	XVCPSGNDP(.Lreg,.Lreg,.Lreg)
2694	.Lreg=.Lreg+1
2695	.endr
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2696
2697denorm_done:
2698	mfspr	r11,SPRN_HSRR0
2699	subi	r11,r11,4
2700	mtspr	SPRN_HSRR0,r11
2701	mtcrf	0x80,r9
2702	ld	r9,PACA_EXGEN+EX_R9(r13)
2703BEGIN_FTR_SECTION
2704	ld	r10,PACA_EXGEN+EX_PPR(r13)
2705	mtspr	SPRN_PPR,r10
2706END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
2707BEGIN_FTR_SECTION
2708	ld	r10,PACA_EXGEN+EX_CFAR(r13)
2709	mtspr	SPRN_CFAR,r10
2710END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
2711	li	r10,0
2712	stb	r10,PACAHSRR_VALID(r13)
2713	ld	r10,PACA_EXGEN+EX_R10(r13)
2714	ld	r11,PACA_EXGEN+EX_R11(r13)
2715	ld	r12,PACA_EXGEN+EX_R12(r13)
2716	ld	r13,PACA_EXGEN+EX_R13(r13)
2717	HRFI_TO_UNKNOWN
2718	b	.
2719#endif
2720
2721EXC_COMMON_BEGIN(denorm_exception_common)
2722	GEN_COMMON denorm_exception
2723	addi	r3,r1,STACK_INT_FRAME_REGS
2724	bl	CFUNC(unknown_exception)
2725	b	interrupt_return_hsrr
2726
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2727
2728#ifdef CONFIG_CBE_RAS
2729INT_DEFINE_BEGIN(cbe_maintenance)
2730	IVEC=0x1600
2731	IHSRR=1
2732INT_DEFINE_END(cbe_maintenance)
2733
2734EXC_REAL_BEGIN(cbe_maintenance, 0x1600, 0x100)
2735	GEN_INT_ENTRY cbe_maintenance, virt=0
2736EXC_REAL_END(cbe_maintenance, 0x1600, 0x100)
2737EXC_VIRT_NONE(0x5600, 0x100)
2738EXC_COMMON_BEGIN(cbe_maintenance_common)
2739	GEN_COMMON cbe_maintenance
2740	addi	r3,r1,STACK_INT_FRAME_REGS
2741	bl	CFUNC(cbe_maintenance_exception)
2742	b	interrupt_return_hsrr
2743
2744#else /* CONFIG_CBE_RAS */
2745EXC_REAL_NONE(0x1600, 0x100)
2746EXC_VIRT_NONE(0x5600, 0x100)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2747#endif
 
 
 
 
 
 
 
 
2748
2749
2750INT_DEFINE_BEGIN(altivec_assist)
2751	IVEC=0x1700
2752#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
2753	IKVM_REAL=1
2754#endif
2755INT_DEFINE_END(altivec_assist)
2756
2757EXC_REAL_BEGIN(altivec_assist, 0x1700, 0x100)
2758	GEN_INT_ENTRY altivec_assist, virt=0
2759EXC_REAL_END(altivec_assist, 0x1700, 0x100)
2760EXC_VIRT_BEGIN(altivec_assist, 0x5700, 0x100)
2761	GEN_INT_ENTRY altivec_assist, virt=1
2762EXC_VIRT_END(altivec_assist, 0x5700, 0x100)
2763EXC_COMMON_BEGIN(altivec_assist_common)
2764	GEN_COMMON altivec_assist
2765	addi	r3,r1,STACK_INT_FRAME_REGS
2766#ifdef CONFIG_ALTIVEC
2767	bl	CFUNC(altivec_assist_exception)
2768	HANDLER_RESTORE_NVGPRS() /* instruction emulation may change GPRs */
2769#else
2770	bl	CFUNC(unknown_exception)
2771#endif
2772	b	interrupt_return_srr
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2773
 
 
 
 
 
 
 
 
 
 
 
 
 
2774
2775#ifdef CONFIG_CBE_RAS
2776INT_DEFINE_BEGIN(cbe_thermal)
2777	IVEC=0x1800
2778	IHSRR=1
2779INT_DEFINE_END(cbe_thermal)
2780
2781EXC_REAL_BEGIN(cbe_thermal, 0x1800, 0x100)
2782	GEN_INT_ENTRY cbe_thermal, virt=0
2783EXC_REAL_END(cbe_thermal, 0x1800, 0x100)
2784EXC_VIRT_NONE(0x5800, 0x100)
2785EXC_COMMON_BEGIN(cbe_thermal_common)
2786	GEN_COMMON cbe_thermal
2787	addi	r3,r1,STACK_INT_FRAME_REGS
2788	bl	CFUNC(cbe_thermal_exception)
2789	b	interrupt_return_hsrr
2790
2791#else /* CONFIG_CBE_RAS */
2792EXC_REAL_NONE(0x1800, 0x100)
2793EXC_VIRT_NONE(0x5800, 0x100)
 
 
 
 
2794#endif
2795
 
 
2796
2797#ifdef CONFIG_PPC_WATCHDOG
2798
2799INT_DEFINE_BEGIN(soft_nmi)
2800	IVEC=0x900
2801	ISTACK=0
2802	ICFAR=0
2803INT_DEFINE_END(soft_nmi)
2804
2805/*
2806 * Branch to soft_nmi_interrupt using the emergency stack. The emergency
2807 * stack is one that is usable by maskable interrupts so long as MSR_EE
2808 * remains off. It is used for recovery when something has corrupted the
2809 * normal kernel stack, for example. The "soft NMI" must not use the process
2810 * stack because we want irq disabled sections to avoid touching the stack
2811 * at all (other than PMU interrupts), so use the emergency stack for this,
2812 * and run it entirely with interrupts hard disabled.
2813 */
2814EXC_COMMON_BEGIN(soft_nmi_common)
2815	mr	r10,r1
2816	ld	r1,PACAEMERGSP(r13)
2817	subi	r1,r1,INT_FRAME_SIZE
2818	__GEN_COMMON_BODY soft_nmi
2819
2820	addi	r3,r1,STACK_INT_FRAME_REGS
2821	bl	CFUNC(soft_nmi_interrupt)
2822
2823	/* Clear MSR_RI before setting SRR0 and SRR1. */
2824	li	r9,0
2825	mtmsrd	r9,1
2826
2827	kuap_kernel_restore r9, r10
 
 
2828
2829	EXCEPTION_RESTORE_REGS hsrr=0
2830	RFI_TO_KERNEL
2831
2832#endif /* CONFIG_PPC_WATCHDOG */
 
2833
2834/*
2835 * An interrupt came in while soft-disabled. We set paca->irq_happened, then:
2836 * - If it was a decrementer interrupt, we bump the dec to max and return.
2837 * - If it was a doorbell we return immediately since doorbells are edge
2838 *   triggered and won't automatically refire.
2839 * - If it was a HMI we return immediately since we handled it in realmode
2840 *   and it won't refire.
2841 * - Else it is one of PACA_IRQ_MUST_HARD_MASK, so hard disable and return.
2842 * This is called with r10 containing the value to OR to the paca field.
2843 */
2844.macro MASKED_INTERRUPT hsrr=0
2845	.if \hsrr
2846masked_Hinterrupt:
2847	.else
2848masked_interrupt:
2849	.endif
2850	stw	r9,PACA_EXGEN+EX_CCR(r13)
2851#ifdef CONFIG_PPC_IRQ_SOFT_MASK_DEBUG
2852	/*
2853	 * Ensure there was no previous MUST_HARD_MASK interrupt or
2854	 * HARD_DIS setting. If this does fire, the interrupt is still
2855	 * masked and MSR[EE] will be cleared on return, so no need to
2856	 * panic, but somebody probably enabled MSR[EE] under
2857	 * PACA_IRQ_HARD_DIS, mtmsr(mfmsr() | MSR_x) being a common
2858	 * cause.
2859	 */
2860	lbz	r9,PACAIRQHAPPENED(r13)
2861	andi.	r9,r9,(PACA_IRQ_MUST_HARD_MASK|PACA_IRQ_HARD_DIS)
28620:	tdnei	r9,0
2863	EMIT_WARN_ENTRY 0b,__FILE__,__LINE__,(BUGFLAG_WARNING | BUGFLAG_ONCE)
2864#endif
2865	lbz	r9,PACAIRQHAPPENED(r13)
2866	or	r9,r9,r10
2867	stb	r9,PACAIRQHAPPENED(r13)
2868
2869	.if ! \hsrr
2870	cmpwi	r10,PACA_IRQ_DEC
2871	bne	1f
2872	LOAD_REG_IMMEDIATE(r9, 0x7fffffff)
2873	mtspr	SPRN_DEC,r9
2874#ifdef CONFIG_PPC_WATCHDOG
2875	lwz	r9,PACA_EXGEN+EX_CCR(r13)
2876	b	soft_nmi_common
2877#else
2878	b	2f
2879#endif
2880	.endif
2881
28821:	andi.	r10,r10,PACA_IRQ_MUST_HARD_MASK
2883	beq	2f
2884	xori	r12,r12,MSR_EE	/* clear MSR_EE */
2885	.if \hsrr
2886	mtspr	SPRN_HSRR1,r12
2887	.else
2888	mtspr	SPRN_SRR1,r12
2889	.endif
2890	ori	r9,r9,PACA_IRQ_HARD_DIS
2891	stb	r9,PACAIRQHAPPENED(r13)
28922:	/* done */
2893	li	r9,0
2894	.if \hsrr
2895	stb	r9,PACAHSRR_VALID(r13)
2896	.else
2897	stb	r9,PACASRR_VALID(r13)
2898	.endif
2899
2900	SEARCH_RESTART_TABLE
2901	cmpdi	r12,0
2902	beq	3f
2903	.if \hsrr
2904	mtspr	SPRN_HSRR0,r12
2905	.else
2906	mtspr	SPRN_SRR0,r12
2907	.endif
29083:
2909
2910	ld	r9,PACA_EXGEN+EX_CTR(r13)
2911	mtctr	r9
2912	lwz	r9,PACA_EXGEN+EX_CCR(r13)
2913	mtcrf	0x80,r9
2914	std	r1,PACAR1(r13)
2915	ld	r9,PACA_EXGEN+EX_R9(r13)
2916	ld	r10,PACA_EXGEN+EX_R10(r13)
2917	ld	r11,PACA_EXGEN+EX_R11(r13)
2918	ld	r12,PACA_EXGEN+EX_R12(r13)
2919	ld	r13,PACA_EXGEN+EX_R13(r13)
2920	/* May return to masked low address where r13 is not set up */
2921	.if \hsrr
2922	HRFI_TO_KERNEL
2923	.else
2924	RFI_TO_KERNEL
2925	.endif
2926	b	.
2927.endm
2928
2929TRAMP_REAL_BEGIN(stf_barrier_fallback)
2930	std	r9,PACA_EXRFI+EX_R9(r13)
2931	std	r10,PACA_EXRFI+EX_R10(r13)
2932	sync
2933	ld	r9,PACA_EXRFI+EX_R9(r13)
2934	ld	r10,PACA_EXRFI+EX_R10(r13)
2935	ori	31,31,0
2936	.rept 14
2937	b	1f
29381:
2939	.endr
2940	blr
2941
2942/* Clobbers r10, r11, ctr */
2943.macro L1D_DISPLACEMENT_FLUSH
2944	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
2945	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
2946	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
2947	mtctr	r11
2948	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
2949
2950	/* order ld/st prior to dcbt stop all streams with flushing */
2951	sync
2952
2953	/*
2954	 * The load addresses are at staggered offsets within cachelines,
2955	 * which suits some pipelines better (on others it should not
2956	 * hurt).
2957	 */
29581:
2959	ld	r11,(0x80 + 8)*0(r10)
2960	ld	r11,(0x80 + 8)*1(r10)
2961	ld	r11,(0x80 + 8)*2(r10)
2962	ld	r11,(0x80 + 8)*3(r10)
2963	ld	r11,(0x80 + 8)*4(r10)
2964	ld	r11,(0x80 + 8)*5(r10)
2965	ld	r11,(0x80 + 8)*6(r10)
2966	ld	r11,(0x80 + 8)*7(r10)
2967	addi	r10,r10,0x80*8
2968	bdnz	1b
2969.endm
2970
2971TRAMP_REAL_BEGIN(entry_flush_fallback)
2972	std	r9,PACA_EXRFI+EX_R9(r13)
2973	std	r10,PACA_EXRFI+EX_R10(r13)
2974	std	r11,PACA_EXRFI+EX_R11(r13)
2975	mfctr	r9
2976	L1D_DISPLACEMENT_FLUSH
2977	mtctr	r9
2978	ld	r9,PACA_EXRFI+EX_R9(r13)
2979	ld	r10,PACA_EXRFI+EX_R10(r13)
2980	ld	r11,PACA_EXRFI+EX_R11(r13)
2981	blr
 
2982
2983/*
2984 * The SCV entry flush happens with interrupts enabled, so it must disable
2985 * to prevent EXRFI being clobbered by NMIs (e.g., soft_nmi_common). r10
2986 * (containing LR) does not need to be preserved here because scv entry
2987 * puts 0 in the pt_regs, CTR can be clobbered for the same reason.
2988 */
2989TRAMP_REAL_BEGIN(scv_entry_flush_fallback)
2990	li	r10,0
2991	mtmsrd	r10,1
2992	lbz	r10,PACAIRQHAPPENED(r13)
2993	ori	r10,r10,PACA_IRQ_HARD_DIS
2994	stb	r10,PACAIRQHAPPENED(r13)
2995	std	r11,PACA_EXRFI+EX_R11(r13)
2996	L1D_DISPLACEMENT_FLUSH
2997	ld	r11,PACA_EXRFI+EX_R11(r13)
2998	li	r10,MSR_RI
2999	mtmsrd	r10,1
3000	blr
3001
3002TRAMP_REAL_BEGIN(rfi_flush_fallback)
3003	SET_SCRATCH0(r13);
3004	GET_PACA(r13);
3005	std	r1,PACA_EXRFI+EX_R12(r13)
3006	ld	r1,PACAKSAVE(r13)
3007	std	r9,PACA_EXRFI+EX_R9(r13)
3008	std	r10,PACA_EXRFI+EX_R10(r13)
3009	std	r11,PACA_EXRFI+EX_R11(r13)
3010	mfctr	r9
3011	L1D_DISPLACEMENT_FLUSH
3012	mtctr	r9
3013	ld	r9,PACA_EXRFI+EX_R9(r13)
3014	ld	r10,PACA_EXRFI+EX_R10(r13)
3015	ld	r11,PACA_EXRFI+EX_R11(r13)
3016	ld	r1,PACA_EXRFI+EX_R12(r13)
3017	GET_SCRATCH0(r13);
3018	rfid
3019
3020TRAMP_REAL_BEGIN(hrfi_flush_fallback)
3021	SET_SCRATCH0(r13);
3022	GET_PACA(r13);
3023	std	r1,PACA_EXRFI+EX_R12(r13)
3024	ld	r1,PACAKSAVE(r13)
3025	std	r9,PACA_EXRFI+EX_R9(r13)
3026	std	r10,PACA_EXRFI+EX_R10(r13)
3027	std	r11,PACA_EXRFI+EX_R11(r13)
3028	mfctr	r9
3029	L1D_DISPLACEMENT_FLUSH
3030	mtctr	r9
3031	ld	r9,PACA_EXRFI+EX_R9(r13)
3032	ld	r10,PACA_EXRFI+EX_R10(r13)
3033	ld	r11,PACA_EXRFI+EX_R11(r13)
3034	ld	r1,PACA_EXRFI+EX_R12(r13)
3035	GET_SCRATCH0(r13);
3036	hrfid
3037
3038TRAMP_REAL_BEGIN(rfscv_flush_fallback)
3039	/* system call volatile */
3040	mr	r7,r13
3041	GET_PACA(r13);
3042	mr	r8,r1
3043	ld	r1,PACAKSAVE(r13)
3044	mfctr	r9
3045	ld	r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
3046	ld	r11,PACA_L1D_FLUSH_SIZE(r13)
3047	srdi	r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
3048	mtctr	r11
3049	DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
3050
3051	/* order ld/st prior to dcbt stop all streams with flushing */
3052	sync
3053
3054	/*
3055	 * The load adresses are at staggered offsets within cachelines,
3056	 * which suits some pipelines better (on others it should not
3057	 * hurt).
3058	 */
30591:
3060	ld	r11,(0x80 + 8)*0(r10)
3061	ld	r11,(0x80 + 8)*1(r10)
3062	ld	r11,(0x80 + 8)*2(r10)
3063	ld	r11,(0x80 + 8)*3(r10)
3064	ld	r11,(0x80 + 8)*4(r10)
3065	ld	r11,(0x80 + 8)*5(r10)
3066	ld	r11,(0x80 + 8)*6(r10)
3067	ld	r11,(0x80 + 8)*7(r10)
3068	addi	r10,r10,0x80*8
3069	bdnz	1b
3070
3071	mtctr	r9
3072	li	r9,0
3073	li	r10,0
3074	li	r11,0
3075	mr	r1,r8
3076	mr	r13,r7
3077	RFSCV
3078
3079USE_TEXT_SECTION()
3080
3081#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
3082kvm_interrupt:
3083	/*
3084	 * The conditional branch in KVMTEST can't reach all the way,
3085	 * make a stub.
3086	 */
3087	b	kvmppc_interrupt
3088#endif
3089
3090_GLOBAL(do_uaccess_flush)
3091	UACCESS_FLUSH_FIXUP_SECTION
3092	nop
3093	nop
3094	nop
3095	blr
3096	L1D_DISPLACEMENT_FLUSH
3097	blr
3098_ASM_NOKPROBE_SYMBOL(do_uaccess_flush)
3099EXPORT_SYMBOL(do_uaccess_flush)
3100
3101
3102MASKED_INTERRUPT
3103MASKED_INTERRUPT hsrr=1
3104
3105USE_FIXED_SECTION(virt_trampolines)
3106	/*
3107	 * All code below __end_soft_masked is treated as soft-masked. If
3108	 * any code runs here with MSR[EE]=1, it must then cope with pending
3109	 * soft interrupt being raised (i.e., by ensuring it is replayed).
3110	 *
3111	 * The __end_interrupts marker must be past the out-of-line (OOL)
3112	 * handlers, so that they are copied to real address 0x100 when running
3113	 * a relocatable kernel. This ensures they can be reached from the short
3114	 * trampoline handlers (like 0x4f00, 0x4f20, etc.) which branch
3115	 * directly, without using LOAD_HANDLER().
3116	 */
3117	.align	7
3118	.globl	__end_interrupts
3119__end_interrupts:
3120DEFINE_FIXED_SYMBOL(__end_interrupts, virt_trampolines)
3121
3122CLOSE_FIXED_SECTION(real_vectors);
3123CLOSE_FIXED_SECTION(real_trampolines);
3124CLOSE_FIXED_SECTION(virt_vectors);
3125CLOSE_FIXED_SECTION(virt_trampolines);
3126
3127USE_TEXT_SECTION()
3128
3129/* MSR[RI] should be clear because this uses SRR[01] */
3130_GLOBAL(enable_machine_check)
3131	mflr	r0
3132	bcl	20,31,$+4
31330:	mflr	r3
3134	addi	r3,r3,(1f - 0b)
3135	mtspr	SPRN_SRR0,r3
3136	mfmsr	r3
3137	ori	r3,r3,MSR_ME
3138	mtspr	SPRN_SRR1,r3
3139	RFI_TO_KERNEL
31401:	mtlr	r0
3141	blr
3142
3143/* MSR[RI] should be clear because this uses SRR[01] */
3144SYM_FUNC_START_LOCAL(disable_machine_check)
3145	mflr	r0
3146	bcl	20,31,$+4
31470:	mflr	r3
3148	addi	r3,r3,(1f - 0b)
3149	mtspr	SPRN_SRR0,r3
3150	mfmsr	r3
3151	li	r4,MSR_ME
3152	andc	r3,r3,r4
3153	mtspr	SPRN_SRR1,r3
3154	RFI_TO_KERNEL
31551:	mtlr	r0
3156	blr
3157SYM_FUNC_END(disable_machine_check)