1/* SPDX-License-Identifier: GPL-2.0-only */
2/*
3 *
4 * Copyright 2011 Paul Mackerras, IBM Corp. <paulus@au1.ibm.com>
5 *
6 * Derived from book3s_rmhandlers.S and other files, which are:
7 *
8 * Copyright SUSE Linux Products GmbH 2009
9 *
10 * Authors: Alexander Graf <agraf@suse.de>
11 */
12
13#include <asm/ppc_asm.h>
14#include <asm/kvm_asm.h>
15#include <asm/reg.h>
16#include <asm/mmu.h>
17#include <asm/page.h>
18#include <asm/ptrace.h>
19#include <asm/hvcall.h>
20#include <asm/asm-offsets.h>
21#include <asm/exception-64s.h>
22#include <asm/kvm_book3s_asm.h>
23#include <asm/book3s/64/mmu-hash.h>
24#include <asm/export.h>
25#include <asm/tm.h>
26#include <asm/opal.h>
27#include <asm/xive-regs.h>
28#include <asm/thread_info.h>
29#include <asm/asm-compat.h>
30#include <asm/feature-fixups.h>
31#include <asm/cpuidle.h>
32#include <asm/ultravisor-api.h>
33
34/* Sign-extend HDEC if not on POWER9 */
35#define EXTEND_HDEC(reg)			\
36BEGIN_FTR_SECTION;				\
37	extsw	reg, reg;			\
38END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_300)
39
40/* Values in HSTATE_NAPPING(r13) */
41#define NAPPING_CEDE	1
42#define NAPPING_NOVCPU	2
43#define NAPPING_UNSPLIT	3
44
45/* Stack frame offsets for kvmppc_hv_entry */
46#define SFS			208
47#define STACK_SLOT_TRAP		(SFS-4)
48#define STACK_SLOT_SHORT_PATH	(SFS-8)
49#define STACK_SLOT_TID		(SFS-16)
50#define STACK_SLOT_PSSCR	(SFS-24)
51#define STACK_SLOT_PID		(SFS-32)
52#define STACK_SLOT_IAMR		(SFS-40)
53#define STACK_SLOT_CIABR	(SFS-48)
54#define STACK_SLOT_DAWR		(SFS-56)
55#define STACK_SLOT_DAWRX	(SFS-64)
56#define STACK_SLOT_HFSCR	(SFS-72)
57#define STACK_SLOT_AMR		(SFS-80)
58#define STACK_SLOT_UAMOR	(SFS-88)
59/* the following is used by the P9 short path */
60#define STACK_SLOT_NVGPRS	(SFS-152)	/* 18 gprs */
61
62/*
63 * Call kvmppc_hv_entry in real mode.
64 * Must be called with interrupts hard-disabled.
65 *
66 * Input Registers:
67 *
68 * LR = return address to continue at after eventually re-enabling MMU
69 */
70_GLOBAL_TOC(kvmppc_hv_entry_trampoline)
71	mflr	r0
72	std	r0, PPC_LR_STKOFF(r1)
73	stdu	r1, -112(r1)
74	mfmsr	r10
75	std	r10, HSTATE_HOST_MSR(r13)
76	LOAD_REG_ADDR(r5, kvmppc_call_hv_entry)
77	li	r0,MSR_RI
78	andc	r0,r10,r0
79	li	r6,MSR_IR | MSR_DR
80	andc	r6,r10,r6
81	mtmsrd	r0,1		/* clear RI in MSR */
82	mtsrr0	r5
83	mtsrr1	r6
84	RFI_TO_KERNEL
85
86kvmppc_call_hv_entry:
87BEGIN_FTR_SECTION
88	/* On P9, do LPCR setting, if necessary */
89	ld	r3, HSTATE_SPLIT_MODE(r13)
90	cmpdi	r3, 0
91	beq	46f
92	lwz	r4, KVM_SPLIT_DO_SET(r3)
93	cmpwi	r4, 0
94	beq	46f
95	bl	kvmhv_p9_set_lpcr
96	nop
9746:
98END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
99
100	ld	r4, HSTATE_KVM_VCPU(r13)
101	bl	kvmppc_hv_entry
102
103	/* Back from guest - restore host state and return to caller */
104
105BEGIN_FTR_SECTION
106	/* Restore host DABR and DABRX */
107	ld	r5,HSTATE_DABR(r13)
108	li	r6,7
109	mtspr	SPRN_DABR,r5
110	mtspr	SPRN_DABRX,r6
111END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
112
113	/* Restore SPRG3 */
114	ld	r3,PACA_SPRG_VDSO(r13)
115	mtspr	SPRN_SPRG_VDSO_WRITE,r3
116
117	/* Reload the host's PMU registers */
118	bl	kvmhv_load_host_pmu
119
120	/*
121	 * Reload DEC.  HDEC interrupts were disabled when
122	 * we reloaded the host's LPCR value.
123	 */
124	ld	r3, HSTATE_DECEXP(r13)
125	mftb	r4
126	subf	r4, r4, r3
127	mtspr	SPRN_DEC, r4
128
129	/* hwthread_req may have got set by cede or no vcpu, so clear it */
130	li	r0, 0
131	stb	r0, HSTATE_HWTHREAD_REQ(r13)
132
133	/*
134	 * For external interrupts we need to call the Linux
135	 * handler to process the interrupt. We do that by jumping
136	 * to absolute address 0x500 for external interrupts.
137	 * The [h]rfid at the end of the handler will return to
138	 * the book3s_hv_interrupts.S code. For other interrupts
139	 * we do the rfid to get back to the book3s_hv_interrupts.S
140	 * code here.
141	 */
142	ld	r8, 112+PPC_LR_STKOFF(r1)
143	addi	r1, r1, 112
144	ld	r7, HSTATE_HOST_MSR(r13)
145
146	/* Return the trap number on this thread as the return value */
147	mr	r3, r12
148
149	/*
150	 * If we came back from the guest via a relocation-on interrupt,
151	 * we will be in virtual mode at this point, which makes it a
152	 * little easier to get back to the caller.
153	 */
154	mfmsr	r0
155	andi.	r0, r0, MSR_IR		/* in real mode? */
156	bne	.Lvirt_return
157
158	/* RFI into the highmem handler */
159	mfmsr	r6
160	li	r0, MSR_RI
161	andc	r6, r6, r0
162	mtmsrd	r6, 1			/* Clear RI in MSR */
163	mtsrr0	r8
164	mtsrr1	r7
165	RFI_TO_KERNEL
166
167	/* Virtual-mode return */
168.Lvirt_return:
169	mtlr	r8
170	blr
171
172kvmppc_primary_no_guest:
173	/* We handle this much like a ceded vcpu */
174	/* put the HDEC into the DEC, since HDEC interrupts don't wake us */
175	/* HDEC may be larger than DEC for arch >= v3.00, but since the */
176	/* HDEC value came from DEC in the first place, it will fit */
177	mfspr	r3, SPRN_HDEC
178	mtspr	SPRN_DEC, r3
179	/*
180	 * Make sure the primary has finished the MMU switch.
181	 * We should never get here on a secondary thread, but
182	 * check it for robustness' sake.
183	 */
184	ld	r5, HSTATE_KVM_VCORE(r13)
18565:	lbz	r0, VCORE_IN_GUEST(r5)
186	cmpwi	r0, 0
187	beq	65b
188	/* Set LPCR. */
189	ld	r8,VCORE_LPCR(r5)
190	mtspr	SPRN_LPCR,r8
191	isync
192	/* set our bit in napping_threads */
193	ld	r5, HSTATE_KVM_VCORE(r13)
194	lbz	r7, HSTATE_PTID(r13)
195	li	r0, 1
196	sld	r0, r0, r7
197	addi	r6, r5, VCORE_NAPPING_THREADS
1981:	lwarx	r3, 0, r6
199	or	r3, r3, r0
200	stwcx.	r3, 0, r6
201	bne	1b
202	/* order napping_threads update vs testing entry_exit_map */
203	isync
204	li	r12, 0
205	lwz	r7, VCORE_ENTRY_EXIT(r5)
206	cmpwi	r7, 0x100
207	bge	kvm_novcpu_exit	/* another thread already exiting */
208	li	r3, NAPPING_NOVCPU
209	stb	r3, HSTATE_NAPPING(r13)
210
211	li	r3, 0		/* Don't wake on privileged (OS) doorbell */
212	b	kvm_do_nap
213
214/*
215 * kvm_novcpu_wakeup
216 *	Entered from kvm_start_guest if kvm_hstate.napping is set
217 *	to NAPPING_NOVCPU
218 *		r2 = kernel TOC
219 *		r13 = paca
220 */
221kvm_novcpu_wakeup:
222	ld	r1, HSTATE_HOST_R1(r13)
223	ld	r5, HSTATE_KVM_VCORE(r13)
224	li	r0, 0
225	stb	r0, HSTATE_NAPPING(r13)
226
227	/* check the wake reason */
228	bl	kvmppc_check_wake_reason
229
230	/*
231	 * Restore volatile registers since we could have called
232	 * a C routine in kvmppc_check_wake_reason.
233	 *	r5 = VCORE
234	 */
235	ld	r5, HSTATE_KVM_VCORE(r13)
236
237	/* see if any other thread is already exiting */
238	lwz	r0, VCORE_ENTRY_EXIT(r5)
239	cmpwi	r0, 0x100
240	bge	kvm_novcpu_exit
241
242	/* clear our bit in napping_threads */
243	lbz	r7, HSTATE_PTID(r13)
244	li	r0, 1
245	sld	r0, r0, r7
246	addi	r6, r5, VCORE_NAPPING_THREADS
2474:	lwarx	r7, 0, r6
248	andc	r7, r7, r0
249	stwcx.	r7, 0, r6
250	bne	4b
251
252	/* See if the wake reason means we need to exit */
253	cmpdi	r3, 0
254	bge	kvm_novcpu_exit
255
256	/* See if our timeslice has expired (HDEC is negative) */
257	mfspr	r0, SPRN_HDEC
258	EXTEND_HDEC(r0)
259	li	r12, BOOK3S_INTERRUPT_HV_DECREMENTER
260	cmpdi	r0, 0
261	blt	kvm_novcpu_exit
262
263	/* Got an IPI but other vcpus aren't yet exiting, must be a latecomer */
264	ld	r4, HSTATE_KVM_VCPU(r13)
265	cmpdi	r4, 0
266	beq	kvmppc_primary_no_guest
267
268#ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
269	addi	r3, r4, VCPU_TB_RMENTRY
270	bl	kvmhv_start_timing
271#endif
272	b	kvmppc_got_guest
273
274kvm_novcpu_exit:
275#ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
276	ld	r4, HSTATE_KVM_VCPU(r13)
277	cmpdi	r4, 0
278	beq	13f
279	addi	r3, r4, VCPU_TB_RMEXIT
280	bl	kvmhv_accumulate_time
281#endif
28213:	mr	r3, r12
283	stw	r12, STACK_SLOT_TRAP(r1)
284	bl	kvmhv_commence_exit
285	nop
286	b	kvmhv_switch_to_host
287
288/*
289 * We come in here when wakened from Linux offline idle code.
290 * Relocation is off
291 * r3 contains the SRR1 wakeup value, SRR1 is trashed.
292 */
293_GLOBAL(idle_kvm_start_guest)
294	ld	r4,PACAEMERGSP(r13)
295	mfcr	r5
296	mflr	r0
297	std	r1,0(r4)
298	std	r5,8(r4)
299	std	r0,16(r4)
300	subi	r1,r4,STACK_FRAME_OVERHEAD
301	SAVE_NVGPRS(r1)
302
303	/*
304	 * Could avoid this and pass it through in r3. For now,
305	 * code expects it to be in SRR1.
306	 */
307	mtspr	SPRN_SRR1,r3
308
309	li	r0,0
310	stb	r0,PACA_FTRACE_ENABLED(r13)
311
312	li	r0,KVM_HWTHREAD_IN_KVM
313	stb	r0,HSTATE_HWTHREAD_STATE(r13)
314
315	/* kvm cede / napping does not come through here */
316	lbz	r0,HSTATE_NAPPING(r13)
317	twnei	r0,0
318
319	b	1f
320
321kvm_unsplit_wakeup:
322	li	r0, 0
323	stb	r0, HSTATE_NAPPING(r13)
324
3251:
326
327	/*
328	 * We weren't napping due to cede, so this must be a secondary
329	 * thread being woken up to run a guest, or being woken up due
330	 * to a stray IPI.  (Or due to some machine check or hypervisor
331	 * maintenance interrupt while the core is in KVM.)
332	 */
333
334	/* Check the wake reason in SRR1 to see why we got here */
335	bl	kvmppc_check_wake_reason
336	/*
337	 * kvmppc_check_wake_reason could invoke a C routine, but we
338	 * have no volatile registers to restore when we return.
339	 */
340
341	cmpdi	r3, 0
342	bge	kvm_no_guest
343
344	/* get vcore pointer, NULL if we have nothing to run */
345	ld	r5,HSTATE_KVM_VCORE(r13)
346	cmpdi	r5,0
347	/* if we have no vcore to run, go back to sleep */
348	beq	kvm_no_guest
349
350kvm_secondary_got_guest:
351
352	/* Set HSTATE_DSCR(r13) to something sensible */
353	ld	r6, PACA_DSCR_DEFAULT(r13)
354	std	r6, HSTATE_DSCR(r13)
355
356	/* On thread 0 of a subcore, set HDEC to max */
357	lbz	r4, HSTATE_PTID(r13)
358	cmpwi	r4, 0
359	bne	63f
360	LOAD_REG_ADDR(r6, decrementer_max)
361	ld	r6, 0(r6)
362	mtspr	SPRN_HDEC, r6
363	/* and set per-LPAR registers, if doing dynamic micro-threading */
364	ld	r6, HSTATE_SPLIT_MODE(r13)
365	cmpdi	r6, 0
366	beq	63f
367BEGIN_FTR_SECTION
368	ld	r0, KVM_SPLIT_RPR(r6)
369	mtspr	SPRN_RPR, r0
370	ld	r0, KVM_SPLIT_PMMAR(r6)
371	mtspr	SPRN_PMMAR, r0
372	ld	r0, KVM_SPLIT_LDBAR(r6)
373	mtspr	SPRN_LDBAR, r0
374	isync
375FTR_SECTION_ELSE
376	/* On P9 we use the split_info for coordinating LPCR changes */
377	lwz	r4, KVM_SPLIT_DO_SET(r6)
378	cmpwi	r4, 0
379	beq	1f
380	mr	r3, r6
381	bl	kvmhv_p9_set_lpcr
382	nop
3831:
384ALT_FTR_SECTION_END_IFCLR(CPU_FTR_ARCH_300)
38563:
386	/* Order load of vcpu after load of vcore */
387	lwsync
388	ld	r4, HSTATE_KVM_VCPU(r13)
389	bl	kvmppc_hv_entry
390
391	/* Back from the guest, go back to nap */
392	/* Clear our vcpu and vcore pointers so we don't come back in early */
393	li	r0, 0
394	std	r0, HSTATE_KVM_VCPU(r13)
395	/*
396	 * Once we clear HSTATE_KVM_VCORE(r13), the code in
397	 * kvmppc_run_core() is going to assume that all our vcpu
398	 * state is visible in memory.  This lwsync makes sure
399	 * that that is true.
400	 */
401	lwsync
402	std	r0, HSTATE_KVM_VCORE(r13)
403
404	/*
405	 * All secondaries exiting guest will fall through this path.
406	 * Before proceeding, just check for HMI interrupt and
407	 * invoke opal hmi handler. By now we are sure that the
408	 * primary thread on this core/subcore has already made partition
409	 * switch/TB resync and we are good to call opal hmi handler.
410	 */
411	cmpwi	r12, BOOK3S_INTERRUPT_HMI
412	bne	kvm_no_guest
413
414	li	r3,0			/* NULL argument */
415	bl	hmi_exception_realmode
416/*
417 * At this point we have finished executing in the guest.
418 * We need to wait for hwthread_req to become zero, since
419 * we may not turn on the MMU while hwthread_req is non-zero.
420 * While waiting we also need to check if we get given a vcpu to run.
421 */
422kvm_no_guest:
423	lbz	r3, HSTATE_HWTHREAD_REQ(r13)
424	cmpwi	r3, 0
425	bne	53f
426	HMT_MEDIUM
427	li	r0, KVM_HWTHREAD_IN_KERNEL
428	stb	r0, HSTATE_HWTHREAD_STATE(r13)
429	/* need to recheck hwthread_req after a barrier, to avoid race */
430	sync
431	lbz	r3, HSTATE_HWTHREAD_REQ(r13)
432	cmpwi	r3, 0
433	bne	54f
434
435	/*
436	 * Jump to idle_return_gpr_loss, which returns to the
437	 * idle_kvm_start_guest caller.
438	 */
439	li	r3, LPCR_PECE0
440	mfspr	r4, SPRN_LPCR
441	rlwimi	r4, r3, 0, LPCR_PECE0 | LPCR_PECE1
442	mtspr	SPRN_LPCR, r4
443	/* set up r3 for return */
444	mfspr	r3,SPRN_SRR1
445	REST_NVGPRS(r1)
446	addi	r1, r1, STACK_FRAME_OVERHEAD
447	ld	r0, 16(r1)
448	ld	r5, 8(r1)
449	ld	r1, 0(r1)
450	mtlr	r0
451	mtcr	r5
452	blr
453
45453:	HMT_LOW
455	ld	r5, HSTATE_KVM_VCORE(r13)
456	cmpdi	r5, 0
457	bne	60f
458	ld	r3, HSTATE_SPLIT_MODE(r13)
459	cmpdi	r3, 0
460	beq	kvm_no_guest
461	lwz	r0, KVM_SPLIT_DO_SET(r3)
462	cmpwi	r0, 0
463	bne	kvmhv_do_set
464	lwz	r0, KVM_SPLIT_DO_RESTORE(r3)
465	cmpwi	r0, 0
466	bne	kvmhv_do_restore
467	lbz	r0, KVM_SPLIT_DO_NAP(r3)
468	cmpwi	r0, 0
469	beq	kvm_no_guest
470	HMT_MEDIUM
471	b	kvm_unsplit_nap
47260:	HMT_MEDIUM
473	b	kvm_secondary_got_guest
474
47554:	li	r0, KVM_HWTHREAD_IN_KVM
476	stb	r0, HSTATE_HWTHREAD_STATE(r13)
477	b	kvm_no_guest
478
479kvmhv_do_set:
480	/* Set LPCR, LPIDR etc. on P9 */
481	HMT_MEDIUM
482	bl	kvmhv_p9_set_lpcr
483	nop
484	b	kvm_no_guest
485
486kvmhv_do_restore:
487	HMT_MEDIUM
488	bl	kvmhv_p9_restore_lpcr
489	nop
490	b	kvm_no_guest
491
492/*
493 * Here the primary thread is trying to return the core to
494 * whole-core mode, so we need to nap.
495 */
496kvm_unsplit_nap:
497	/*
498	 * When secondaries are napping in kvm_unsplit_nap() with
499	 * hwthread_req = 1, HMI goes ignored even though subcores are
500	 * already exited the guest. Hence HMI keeps waking up secondaries
501	 * from nap in a loop and secondaries always go back to nap since
502	 * no vcore is assigned to them. This makes impossible for primary
503	 * thread to get hold of secondary threads resulting into a soft
504	 * lockup in KVM path.
505	 *
506	 * Let us check if HMI is pending and handle it before we go to nap.
507	 */
508	cmpwi	r12, BOOK3S_INTERRUPT_HMI
509	bne	55f
510	li	r3, 0			/* NULL argument */
511	bl	hmi_exception_realmode
51255:
513	/*
514	 * Ensure that secondary doesn't nap when it has
515	 * its vcore pointer set.
516	 */
517	sync		/* matches smp_mb() before setting split_info.do_nap */
518	ld	r0, HSTATE_KVM_VCORE(r13)
519	cmpdi	r0, 0
520	bne	kvm_no_guest
521	/* clear any pending message */
522BEGIN_FTR_SECTION
523	lis	r6, (PPC_DBELL_SERVER << (63-36))@h
524	PPC_MSGCLR(6)
525END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
526	/* Set kvm_split_mode.napped[tid] = 1 */
527	ld	r3, HSTATE_SPLIT_MODE(r13)
528	li	r0, 1
529	lbz	r4, HSTATE_TID(r13)
530	addi	r4, r4, KVM_SPLIT_NAPPED
531	stbx	r0, r3, r4
532	/* Check the do_nap flag again after setting napped[] */
533	sync
534	lbz	r0, KVM_SPLIT_DO_NAP(r3)
535	cmpwi	r0, 0
536	beq	57f
537	li	r3, NAPPING_UNSPLIT
538	stb	r3, HSTATE_NAPPING(r13)
539	li	r3, (LPCR_PECEDH | LPCR_PECE0) >> 4
540	mfspr	r5, SPRN_LPCR
541	rlwimi	r5, r3, 4, (LPCR_PECEDP | LPCR_PECEDH | LPCR_PECE0 | LPCR_PECE1)
542	b	kvm_nap_sequence
543
54457:	li	r0, 0
545	stbx	r0, r3, r4
546	b	kvm_no_guest
547
548/******************************************************************************
549 *                                                                            *
550 *                               Entry code                                   *
551 *                                                                            *
552 *****************************************************************************/
553
554.global kvmppc_hv_entry
555kvmppc_hv_entry:
556
557	/* Required state:
558	 *
559	 * R4 = vcpu pointer (or NULL)
560	 * MSR = ~IR|DR
561	 * R13 = PACA
562	 * R1 = host R1
563	 * R2 = TOC
564	 * all other volatile GPRS = free
565	 * Does not preserve non-volatile GPRs or CR fields
566	 */
567	mflr	r0
568	std	r0, PPC_LR_STKOFF(r1)
569	stdu	r1, -SFS(r1)
570
571	/* Save R1 in the PACA */
572	std	r1, HSTATE_HOST_R1(r13)
573
574	li	r6, KVM_GUEST_MODE_HOST_HV
575	stb	r6, HSTATE_IN_GUEST(r13)
576
577#ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
578	/* Store initial timestamp */
579	cmpdi	r4, 0
580	beq	1f
581	addi	r3, r4, VCPU_TB_RMENTRY
582	bl	kvmhv_start_timing
5831:
584#endif
585
586	ld	r5, HSTATE_KVM_VCORE(r13)
587	ld	r9, VCORE_KVM(r5)	/* pointer to struct kvm */
588
589	/*
590	 * POWER7/POWER8 host -> guest partition switch code.
591	 * We don't have to lock against concurrent tlbies,
592	 * but we do have to coordinate across hardware threads.
593	 */
594	/* Set bit in entry map iff exit map is zero. */
595	li	r7, 1
596	lbz	r6, HSTATE_PTID(r13)
597	sld	r7, r7, r6
598	addi	r8, r5, VCORE_ENTRY_EXIT
59921:	lwarx	r3, 0, r8
600	cmpwi	r3, 0x100		/* any threads starting to exit? */
601	bge	secondary_too_late	/* if so we're too late to the party */
602	or	r3, r3, r7
603	stwcx.	r3, 0, r8
604	bne	21b
605
606	/* Primary thread switches to guest partition. */
607	cmpwi	r6,0
608	bne	10f
609
610	lwz	r7,KVM_LPID(r9)
611BEGIN_FTR_SECTION
612	ld	r6,KVM_SDR1(r9)
613	li	r0,LPID_RSVD		/* switch to reserved LPID */
614	mtspr	SPRN_LPID,r0
615	ptesync
616	mtspr	SPRN_SDR1,r6		/* switch to partition page table */
617END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_300)
618	mtspr	SPRN_LPID,r7
619	isync
620
621	/* See if we need to flush the TLB. */
622	mr	r3, r9			/* kvm pointer */
623	lhz	r4, PACAPACAINDEX(r13)	/* physical cpu number */
624	li	r5, 0			/* nested vcpu pointer */
625	bl	kvmppc_check_need_tlb_flush
626	nop
627	ld	r5, HSTATE_KVM_VCORE(r13)
628
629	/* Add timebase offset onto timebase */
63022:	ld	r8,VCORE_TB_OFFSET(r5)
631	cmpdi	r8,0
632	beq	37f
633	std	r8, VCORE_TB_OFFSET_APPL(r5)
634	mftb	r6		/* current host timebase */
635	add	r8,r8,r6
636	mtspr	SPRN_TBU40,r8	/* update upper 40 bits */
637	mftb	r7		/* check if lower 24 bits overflowed */
638	clrldi	r6,r6,40
639	clrldi	r7,r7,40
640	cmpld	r7,r6
641	bge	37f
642	addis	r8,r8,0x100	/* if so, increment upper 40 bits */
643	mtspr	SPRN_TBU40,r8
644
645	/* Load guest PCR value to select appropriate compat mode */
64637:	ld	r7, VCORE_PCR(r5)
647	cmpdi	r7, 0
648	beq	38f
649	mtspr	SPRN_PCR, r7
65038:
651
652BEGIN_FTR_SECTION
653	/* DPDES and VTB are shared between threads */
654	ld	r8, VCORE_DPDES(r5)
655	ld	r7, VCORE_VTB(r5)
656	mtspr	SPRN_DPDES, r8
657	mtspr	SPRN_VTB, r7
658END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
659
660	/* Mark the subcore state as inside guest */
661	bl	kvmppc_subcore_enter_guest
662	nop
663	ld	r5, HSTATE_KVM_VCORE(r13)
664	ld	r4, HSTATE_KVM_VCPU(r13)
665	li	r0,1
666	stb	r0,VCORE_IN_GUEST(r5)	/* signal secondaries to continue */
667
668	/* Do we have a guest vcpu to run? */
66910:	cmpdi	r4, 0
670	beq	kvmppc_primary_no_guest
671kvmppc_got_guest:
672	/* Increment yield count if they have a VPA */
673	ld	r3, VCPU_VPA(r4)
674	cmpdi	r3, 0
675	beq	25f
676	li	r6, LPPACA_YIELDCOUNT
677	LWZX_BE	r5, r3, r6
678	addi	r5, r5, 1
679	STWX_BE	r5, r3, r6
680	li	r6, 1
681	stb	r6, VCPU_VPA_DIRTY(r4)
68225:
683
684	/* Save purr/spurr */
685	mfspr	r5,SPRN_PURR
686	mfspr	r6,SPRN_SPURR
687	std	r5,HSTATE_PURR(r13)
688	std	r6,HSTATE_SPURR(r13)
689	ld	r7,VCPU_PURR(r4)
690	ld	r8,VCPU_SPURR(r4)
691	mtspr	SPRN_PURR,r7
692	mtspr	SPRN_SPURR,r8
693
694	/* Save host values of some registers */
695BEGIN_FTR_SECTION
696	mfspr	r5, SPRN_TIDR
697	mfspr	r6, SPRN_PSSCR
698	mfspr	r7, SPRN_PID
699	std	r5, STACK_SLOT_TID(r1)
700	std	r6, STACK_SLOT_PSSCR(r1)
701	std	r7, STACK_SLOT_PID(r1)
702	mfspr	r5, SPRN_HFSCR
703	std	r5, STACK_SLOT_HFSCR(r1)
704END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
705BEGIN_FTR_SECTION
706	mfspr	r5, SPRN_CIABR
707	mfspr	r6, SPRN_DAWR
708	mfspr	r7, SPRN_DAWRX
709	mfspr	r8, SPRN_IAMR
710	std	r5, STACK_SLOT_CIABR(r1)
711	std	r6, STACK_SLOT_DAWR(r1)
712	std	r7, STACK_SLOT_DAWRX(r1)
713	std	r8, STACK_SLOT_IAMR(r1)
714END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
715
716	mfspr	r5, SPRN_AMR
717	std	r5, STACK_SLOT_AMR(r1)
718	mfspr	r6, SPRN_UAMOR
719	std	r6, STACK_SLOT_UAMOR(r1)
720
721BEGIN_FTR_SECTION
722	/* Set partition DABR */
723	/* Do this before re-enabling PMU to avoid P7 DABR corruption bug */
724	lwz	r5,VCPU_DABRX(r4)
725	ld	r6,VCPU_DABR(r4)
726	mtspr	SPRN_DABRX,r5
727	mtspr	SPRN_DABR,r6
728	isync
729END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
730
731#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
732/*
733 * Branch around the call if both CPU_FTR_TM and
734 * CPU_FTR_P9_TM_HV_ASSIST are off.
735 */
736BEGIN_FTR_SECTION
737	b	91f
738END_FTR_SECTION(CPU_FTR_TM | CPU_FTR_P9_TM_HV_ASSIST, 0)
739	/*
740	 * NOTE THAT THIS TRASHES ALL NON-VOLATILE REGISTERS (but not CR)
741	 */
742	mr      r3, r4
743	ld      r4, VCPU_MSR(r3)
744	li	r5, 0			/* don't preserve non-vol regs */
745	bl	kvmppc_restore_tm_hv
746	nop
747	ld	r4, HSTATE_KVM_VCPU(r13)
74891:
749#endif
750
751	/* Load guest PMU registers; r4 = vcpu pointer here */
752	mr	r3, r4
753	bl	kvmhv_load_guest_pmu
754
755	/* Load up FP, VMX and VSX registers */
756	ld	r4, HSTATE_KVM_VCPU(r13)
757	bl	kvmppc_load_fp
758
759	ld	r14, VCPU_GPR(R14)(r4)
760	ld	r15, VCPU_GPR(R15)(r4)
761	ld	r16, VCPU_GPR(R16)(r4)
762	ld	r17, VCPU_GPR(R17)(r4)
763	ld	r18, VCPU_GPR(R18)(r4)
764	ld	r19, VCPU_GPR(R19)(r4)
765	ld	r20, VCPU_GPR(R20)(r4)
766	ld	r21, VCPU_GPR(R21)(r4)
767	ld	r22, VCPU_GPR(R22)(r4)
768	ld	r23, VCPU_GPR(R23)(r4)
769	ld	r24, VCPU_GPR(R24)(r4)
770	ld	r25, VCPU_GPR(R25)(r4)
771	ld	r26, VCPU_GPR(R26)(r4)
772	ld	r27, VCPU_GPR(R27)(r4)
773	ld	r28, VCPU_GPR(R28)(r4)
774	ld	r29, VCPU_GPR(R29)(r4)
775	ld	r30, VCPU_GPR(R30)(r4)
776	ld	r31, VCPU_GPR(R31)(r4)
777
778	/* Switch DSCR to guest value */
779	ld	r5, VCPU_DSCR(r4)
780	mtspr	SPRN_DSCR, r5
781
782BEGIN_FTR_SECTION
783	/* Skip next section on POWER7 */
784	b	8f
785END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
786	/* Load up POWER8-specific registers */
787	ld	r5, VCPU_IAMR(r4)
788	lwz	r6, VCPU_PSPB(r4)
789	ld	r7, VCPU_FSCR(r4)
790	mtspr	SPRN_IAMR, r5
791	mtspr	SPRN_PSPB, r6
792	mtspr	SPRN_FSCR, r7
793	/*
794	 * Handle broken DAWR case by not writing it. This means we
795	 * can still store the DAWR register for migration.
796	 */
797	LOAD_REG_ADDR(r5, dawr_force_enable)
798	lbz	r5, 0(r5)
799	cmpdi	r5, 0
800	beq	1f
801	ld	r5, VCPU_DAWR(r4)
802	ld	r6, VCPU_DAWRX(r4)
803	mtspr	SPRN_DAWR, r5
804	mtspr	SPRN_DAWRX, r6
8051:
806	ld	r7, VCPU_CIABR(r4)
807	ld	r8, VCPU_TAR(r4)
808	mtspr	SPRN_CIABR, r7
809	mtspr	SPRN_TAR, r8
810	ld	r5, VCPU_IC(r4)
811	ld	r8, VCPU_EBBHR(r4)
812	mtspr	SPRN_IC, r5
813	mtspr	SPRN_EBBHR, r8
814	ld	r5, VCPU_EBBRR(r4)
815	ld	r6, VCPU_BESCR(r4)
816	lwz	r7, VCPU_GUEST_PID(r4)
817	ld	r8, VCPU_WORT(r4)
818	mtspr	SPRN_EBBRR, r5
819	mtspr	SPRN_BESCR, r6
820	mtspr	SPRN_PID, r7
821	mtspr	SPRN_WORT, r8
822BEGIN_FTR_SECTION
823	/* POWER8-only registers */
824	ld	r5, VCPU_TCSCR(r4)
825	ld	r6, VCPU_ACOP(r4)
826	ld	r7, VCPU_CSIGR(r4)
827	ld	r8, VCPU_TACR(r4)
828	mtspr	SPRN_TCSCR, r5
829	mtspr	SPRN_ACOP, r6
830	mtspr	SPRN_CSIGR, r7
831	mtspr	SPRN_TACR, r8
832	nop
833FTR_SECTION_ELSE
834	/* POWER9-only registers */
835	ld	r5, VCPU_TID(r4)
836	ld	r6, VCPU_PSSCR(r4)
837	lbz	r8, HSTATE_FAKE_SUSPEND(r13)
838	oris	r6, r6, PSSCR_EC@h	/* This makes stop trap to HV */
839	rldimi	r6, r8, PSSCR_FAKE_SUSPEND_LG, 63 - PSSCR_FAKE_SUSPEND_LG
840	ld	r7, VCPU_HFSCR(r4)
841	mtspr	SPRN_TIDR, r5
842	mtspr	SPRN_PSSCR, r6
843	mtspr	SPRN_HFSCR, r7
844ALT_FTR_SECTION_END_IFCLR(CPU_FTR_ARCH_300)
8458:
846
847	ld	r5, VCPU_SPRG0(r4)
848	ld	r6, VCPU_SPRG1(r4)
849	ld	r7, VCPU_SPRG2(r4)
850	ld	r8, VCPU_SPRG3(r4)
851	mtspr	SPRN_SPRG0, r5
852	mtspr	SPRN_SPRG1, r6
853	mtspr	SPRN_SPRG2, r7
854	mtspr	SPRN_SPRG3, r8
855
856	/* Load up DAR and DSISR */
857	ld	r5, VCPU_DAR(r4)
858	lwz	r6, VCPU_DSISR(r4)
859	mtspr	SPRN_DAR, r5
860	mtspr	SPRN_DSISR, r6
861
862	/* Restore AMR and UAMOR, set AMOR to all 1s */
863	ld	r5,VCPU_AMR(r4)
864	ld	r6,VCPU_UAMOR(r4)
865	li	r7,-1
866	mtspr	SPRN_AMR,r5
867	mtspr	SPRN_UAMOR,r6
868	mtspr	SPRN_AMOR,r7
869
870	/* Restore state of CTRL run bit; assume 1 on entry */
871	lwz	r5,VCPU_CTRL(r4)
872	andi.	r5,r5,1
873	bne	4f
874	mfspr	r6,SPRN_CTRLF
875	clrrdi	r6,r6,1
876	mtspr	SPRN_CTRLT,r6
8774:
878	/* Secondary threads wait for primary to have done partition switch */
879	ld	r5, HSTATE_KVM_VCORE(r13)
880	lbz	r6, HSTATE_PTID(r13)
881	cmpwi	r6, 0
882	beq	21f
883	lbz	r0, VCORE_IN_GUEST(r5)
884	cmpwi	r0, 0
885	bne	21f
886	HMT_LOW
88720:	lwz	r3, VCORE_ENTRY_EXIT(r5)
888	cmpwi	r3, 0x100
889	bge	no_switch_exit
890	lbz	r0, VCORE_IN_GUEST(r5)
891	cmpwi	r0, 0
892	beq	20b
893	HMT_MEDIUM
89421:
895	/* Set LPCR. */
896	ld	r8,VCORE_LPCR(r5)
897	mtspr	SPRN_LPCR,r8
898	isync
899
900	/*
901	 * Set the decrementer to the guest decrementer.
902	 */
903	ld	r8,VCPU_DEC_EXPIRES(r4)
904	/* r8 is a host timebase value here, convert to guest TB */
905	ld	r5,HSTATE_KVM_VCORE(r13)
906	ld	r6,VCORE_TB_OFFSET_APPL(r5)
907	add	r8,r8,r6
908	mftb	r7
909	subf	r3,r7,r8
910	mtspr	SPRN_DEC,r3
911
912	/* Check if HDEC expires soon */
913	mfspr	r3, SPRN_HDEC
914	EXTEND_HDEC(r3)
915	cmpdi	r3, 512		/* 1 microsecond */
916	blt	hdec_soon
917
918	/* For hash guest, clear out and reload the SLB */
919	ld	r6, VCPU_KVM(r4)
920	lbz	r0, KVM_RADIX(r6)
921	cmpwi	r0, 0
922	bne	9f
923	li	r6, 0
924	slbmte	r6, r6
925	slbia
926	ptesync
927
928	/* Load up guest SLB entries (N.B. slb_max will be 0 for radix) */
929	lwz	r5,VCPU_SLB_MAX(r4)
930	cmpwi	r5,0
931	beq	9f
932	mtctr	r5
933	addi	r6,r4,VCPU_SLB
9341:	ld	r8,VCPU_SLB_E(r6)
935	ld	r9,VCPU_SLB_V(r6)
936	slbmte	r9,r8
937	addi	r6,r6,VCPU_SLB_SIZE
938	bdnz	1b
9399:
940
941#ifdef CONFIG_KVM_XICS
942	/* We are entering the guest on that thread, push VCPU to XIVE */
943	ld	r11, VCPU_XIVE_SAVED_STATE(r4)
944	li	r9, TM_QW1_OS
945	lwz	r8, VCPU_XIVE_CAM_WORD(r4)
946	cmpwi	r8, 0
947	beq	no_xive
948	li	r7, TM_QW1_OS + TM_WORD2
949	mfmsr	r0
950	andi.	r0, r0, MSR_DR		/* in real mode? */
951	beq	2f
952	ld	r10, HSTATE_XIVE_TIMA_VIRT(r13)
953	cmpldi	cr1, r10, 0
954	beq     cr1, no_xive
955	eieio
956	stdx	r11,r9,r10
957	stwx	r8,r7,r10
958	b	3f
9592:	ld	r10, HSTATE_XIVE_TIMA_PHYS(r13)
960	cmpldi	cr1, r10, 0
961	beq	cr1, no_xive
962	eieio
963	stdcix	r11,r9,r10
964	stwcix	r8,r7,r10
9653:	li	r9, 1
966	stb	r9, VCPU_XIVE_PUSHED(r4)
967	eieio
968
969	/*
970	 * We clear the irq_pending flag. There is a small chance of a
971	 * race vs. the escalation interrupt happening on another
972	 * processor setting it again, but the only consequence is to
973	 * cause a spurrious wakeup on the next H_CEDE which is not an
974	 * issue.
975	 */
976	li	r0,0
977	stb	r0, VCPU_IRQ_PENDING(r4)
978
979	/*
980	 * In single escalation mode, if the escalation interrupt is
981	 * on, we mask it.
982	 */
983	lbz	r0, VCPU_XIVE_ESC_ON(r4)
984	cmpwi	cr1, r0,0
985	beq	cr1, 1f
986	li	r9, XIVE_ESB_SET_PQ_01
987	beq	4f			/* in real mode? */
988	ld	r10, VCPU_XIVE_ESC_VADDR(r4)
989	ldx	r0, r10, r9
990	b	5f
9914:	ld	r10, VCPU_XIVE_ESC_RADDR(r4)
992	ldcix	r0, r10, r9
9935:	sync
994
995	/* We have a possible subtle race here: The escalation interrupt might
996	 * have fired and be on its way to the host queue while we mask it,
997	 * and if we unmask it early enough (re-cede right away), there is
998	 * a theorical possibility that it fires again, thus landing in the
999	 * target queue more than once which is a big no-no.
1000	 *
1001	 * Fortunately, solving this is rather easy. If the above load setting
1002	 * PQ to 01 returns a previous value where P is set, then we know the
1003	 * escalation interrupt is somewhere on its way to the host. In that
1004	 * case we simply don't clear the xive_esc_on flag below. It will be
1005	 * eventually cleared by the handler for the escalation interrupt.
1006	 *
1007	 * Then, when doing a cede, we check that flag again before re-enabling
1008	 * the escalation interrupt, and if set, we abort the cede.
1009	 */
1010	andi.	r0, r0, XIVE_ESB_VAL_P
1011	bne-	1f
1012
1013	/* Now P is 0, we can clear the flag */
1014	li	r0, 0
1015	stb	r0, VCPU_XIVE_ESC_ON(r4)
10161:
1017no_xive:
1018#endif /* CONFIG_KVM_XICS */
1019
1020	li	r0, 0
1021	stw	r0, STACK_SLOT_SHORT_PATH(r1)
1022
1023deliver_guest_interrupt:	/* r4 = vcpu, r13 = paca */
1024	/* Check if we can deliver an external or decrementer interrupt now */
1025	ld	r0, VCPU_PENDING_EXC(r4)
1026BEGIN_FTR_SECTION
1027	/* On POWER9, also check for emulated doorbell interrupt */
1028	lbz	r3, VCPU_DBELL_REQ(r4)
1029	or	r0, r0, r3
1030END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
1031	cmpdi	r0, 0
1032	beq	71f
1033	mr	r3, r4
1034	bl	kvmppc_guest_entry_inject_int
1035	ld	r4, HSTATE_KVM_VCPU(r13)
103671:
1037	ld	r6, VCPU_SRR0(r4)
1038	ld	r7, VCPU_SRR1(r4)
1039	mtspr	SPRN_SRR0, r6
1040	mtspr	SPRN_SRR1, r7
1041
1042fast_guest_entry_c:
1043	ld	r10, VCPU_PC(r4)
1044	ld	r11, VCPU_MSR(r4)
1045	/* r11 = vcpu->arch.msr & ~MSR_HV */
1046	rldicl	r11, r11, 63 - MSR_HV_LG, 1
1047	rotldi	r11, r11, 1 + MSR_HV_LG
1048	ori	r11, r11, MSR_ME
1049
1050	ld	r6, VCPU_CTR(r4)
1051	ld	r7, VCPU_XER(r4)
1052	mtctr	r6
1053	mtxer	r7
1054
1055/*
1056 * Required state:
1057 * R4 = vcpu
1058 * R10: value for HSRR0
1059 * R11: value for HSRR1
1060 * R13 = PACA
1061 */
1062fast_guest_return:
1063	li	r0,0
1064	stb	r0,VCPU_CEDED(r4)	/* cancel cede */
1065	mtspr	SPRN_HSRR0,r10
1066	mtspr	SPRN_HSRR1,r11
1067
1068	/* Activate guest mode, so faults get handled by KVM */
1069	li	r9, KVM_GUEST_MODE_GUEST_HV
1070	stb	r9, HSTATE_IN_GUEST(r13)
1071
1072#ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
1073	/* Accumulate timing */
1074	addi	r3, r4, VCPU_TB_GUEST
1075	bl	kvmhv_accumulate_time
1076#endif
1077
1078	/* Enter guest */
1079
1080BEGIN_FTR_SECTION
1081	ld	r5, VCPU_CFAR(r4)
1082	mtspr	SPRN_CFAR, r5
1083END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
1084BEGIN_FTR_SECTION
1085	ld	r0, VCPU_PPR(r4)
1086END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
1087
1088	ld	r5, VCPU_LR(r4)
1089	mtlr	r5
1090
1091	ld	r1, VCPU_GPR(R1)(r4)
1092	ld	r5, VCPU_GPR(R5)(r4)
1093	ld	r8, VCPU_GPR(R8)(r4)
1094	ld	r9, VCPU_GPR(R9)(r4)
1095	ld	r10, VCPU_GPR(R10)(r4)
1096	ld	r11, VCPU_GPR(R11)(r4)
1097	ld	r12, VCPU_GPR(R12)(r4)
1098	ld	r13, VCPU_GPR(R13)(r4)
1099
1100BEGIN_FTR_SECTION
1101	mtspr	SPRN_PPR, r0
1102END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
1103
1104/* Move canary into DSISR to check for later */
1105BEGIN_FTR_SECTION
1106	li	r0, 0x7fff
1107	mtspr	SPRN_HDSISR, r0
1108END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
1109
1110	ld	r6, VCPU_KVM(r4)
1111	lbz	r7, KVM_SECURE_GUEST(r6)
1112	cmpdi	r7, 0
1113	ld	r6, VCPU_GPR(R6)(r4)
1114	ld	r7, VCPU_GPR(R7)(r4)
1115	bne	ret_to_ultra
1116
1117	lwz	r0, VCPU_CR(r4)
1118	mtcr	r0
1119
1120	ld	r0, VCPU_GPR(R0)(r4)
1121	ld	r2, VCPU_GPR(R2)(r4)
1122	ld	r3, VCPU_GPR(R3)(r4)
1123	ld	r4, VCPU_GPR(R4)(r4)
1124	HRFI_TO_GUEST
1125	b	.
1126/*
1127 * Use UV_RETURN ultracall to return control back to the Ultravisor after
1128 * processing an hypercall or interrupt that was forwarded (a.k.a. reflected)
1129 * to the Hypervisor.
1130 *
1131 * All registers have already been loaded, except:
1132 *   R0 = hcall result
1133 *   R2 = SRR1, so UV can detect a synthesized interrupt (if any)
1134 *   R3 = UV_RETURN
1135 */
1136ret_to_ultra:
1137	lwz	r0, VCPU_CR(r4)
1138	mtcr	r0
1139
1140	ld	r0, VCPU_GPR(R3)(r4)
1141	mfspr	r2, SPRN_SRR1
1142	li	r3, 0
1143	ori	r3, r3, UV_RETURN
1144	ld	r4, VCPU_GPR(R4)(r4)
1145	sc	2
1146
1147/*
1148 * Enter the guest on a P9 or later system where we have exactly
1149 * one vcpu per vcore and we don't need to go to real mode
1150 * (which implies that host and guest are both using radix MMU mode).
1151 * r3 = vcpu pointer
1152 * Most SPRs and all the VSRs have been loaded already.
1153 */
1154_GLOBAL(__kvmhv_vcpu_entry_p9)
1155EXPORT_SYMBOL_GPL(__kvmhv_vcpu_entry_p9)
1156	mflr	r0
1157	std	r0, PPC_LR_STKOFF(r1)
1158	stdu	r1, -SFS(r1)
1159
1160	li	r0, 1
1161	stw	r0, STACK_SLOT_SHORT_PATH(r1)
1162
1163	std	r3, HSTATE_KVM_VCPU(r13)
1164	mfcr	r4
1165	stw	r4, SFS+8(r1)
1166
1167	std	r1, HSTATE_HOST_R1(r13)
1168
1169	reg = 14
1170	.rept	18
1171	std	reg, STACK_SLOT_NVGPRS + ((reg - 14) * 8)(r1)
1172	reg = reg + 1
1173	.endr
1174
1175	reg = 14
1176	.rept	18
1177	ld	reg, __VCPU_GPR(reg)(r3)
1178	reg = reg + 1
1179	.endr
1180
1181	mfmsr	r10
1182	std	r10, HSTATE_HOST_MSR(r13)
1183
1184	mr	r4, r3
1185	b	fast_guest_entry_c
1186guest_exit_short_path:
1187
1188	li	r0, KVM_GUEST_MODE_NONE
1189	stb	r0, HSTATE_IN_GUEST(r13)
1190
1191	reg = 14
1192	.rept	18
1193	std	reg, __VCPU_GPR(reg)(r9)
1194	reg = reg + 1
1195	.endr
1196
1197	reg = 14
1198	.rept	18
1199	ld	reg, STACK_SLOT_NVGPRS + ((reg - 14) * 8)(r1)
1200	reg = reg + 1
1201	.endr
1202
1203	lwz	r4, SFS+8(r1)
1204	mtcr	r4
1205
1206	mr	r3, r12		/* trap number */
1207
1208	addi	r1, r1, SFS
1209	ld	r0, PPC_LR_STKOFF(r1)
1210	mtlr	r0
1211
1212	/* If we are in real mode, do a rfid to get back to the caller */
1213	mfmsr	r4
1214	andi.	r5, r4, MSR_IR
1215	bnelr
1216	rldicl	r5, r4, 64 - MSR_TS_S_LG, 62	/* extract TS field */
1217	mtspr	SPRN_SRR0, r0
1218	ld	r10, HSTATE_HOST_MSR(r13)
1219	rldimi	r10, r5, MSR_TS_S_LG, 63 - MSR_TS_T_LG
1220	mtspr	SPRN_SRR1, r10
1221	RFI_TO_KERNEL
1222	b	.
1223
1224secondary_too_late:
1225	li	r12, 0
1226	stw	r12, STACK_SLOT_TRAP(r1)
1227	cmpdi	r4, 0
1228	beq	11f
1229	stw	r12, VCPU_TRAP(r4)
1230#ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
1231	addi	r3, r4, VCPU_TB_RMEXIT
1232	bl	kvmhv_accumulate_time
1233#endif
123411:	b	kvmhv_switch_to_host
1235
1236no_switch_exit:
1237	HMT_MEDIUM
1238	li	r12, 0
1239	b	12f
1240hdec_soon:
1241	li	r12, BOOK3S_INTERRUPT_HV_DECREMENTER
124212:	stw	r12, VCPU_TRAP(r4)
1243	mr	r9, r4
1244#ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
1245	addi	r3, r4, VCPU_TB_RMEXIT
1246	bl	kvmhv_accumulate_time
1247#endif
1248	b	guest_bypass
1249
1250/******************************************************************************
1251 *                                                                            *
1252 *                               Exit code                                    *
1253 *                                                                            *
1254 *****************************************************************************/
1255
1256/*
1257 * We come here from the first-level interrupt handlers.
1258 */
1259	.globl	kvmppc_interrupt_hv
1260kvmppc_interrupt_hv:
1261	/*
1262	 * Register contents:
1263	 * R12		= (guest CR << 32) | interrupt vector
1264	 * R13		= PACA
1265	 * guest R12 saved in shadow VCPU SCRATCH0
1266	 * guest CTR saved in shadow VCPU SCRATCH1 if RELOCATABLE
1267	 * guest R13 saved in SPRN_SCRATCH0
1268	 */
1269	std	r9, HSTATE_SCRATCH2(r13)
1270	lbz	r9, HSTATE_IN_GUEST(r13)
1271	cmpwi	r9, KVM_GUEST_MODE_HOST_HV
1272	beq	kvmppc_bad_host_intr
1273#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1274	cmpwi	r9, KVM_GUEST_MODE_GUEST
1275	ld	r9, HSTATE_SCRATCH2(r13)
1276	beq	kvmppc_interrupt_pr
1277#endif
1278	/* We're now back in the host but in guest MMU context */
1279	li	r9, KVM_GUEST_MODE_HOST_HV
1280	stb	r9, HSTATE_IN_GUEST(r13)
1281
1282	ld	r9, HSTATE_KVM_VCPU(r13)
1283
1284	/* Save registers */
1285
1286	std	r0, VCPU_GPR(R0)(r9)
1287	std	r1, VCPU_GPR(R1)(r9)
1288	std	r2, VCPU_GPR(R2)(r9)
1289	std	r3, VCPU_GPR(R3)(r9)
1290	std	r4, VCPU_GPR(R4)(r9)
1291	std	r5, VCPU_GPR(R5)(r9)
1292	std	r6, VCPU_GPR(R6)(r9)
1293	std	r7, VCPU_GPR(R7)(r9)
1294	std	r8, VCPU_GPR(R8)(r9)
1295	ld	r0, HSTATE_SCRATCH2(r13)
1296	std	r0, VCPU_GPR(R9)(r9)
1297	std	r10, VCPU_GPR(R10)(r9)
1298	std	r11, VCPU_GPR(R11)(r9)
1299	ld	r3, HSTATE_SCRATCH0(r13)
1300	std	r3, VCPU_GPR(R12)(r9)
1301	/* CR is in the high half of r12 */
1302	srdi	r4, r12, 32
1303	std	r4, VCPU_CR(r9)
1304BEGIN_FTR_SECTION
1305	ld	r3, HSTATE_CFAR(r13)
1306	std	r3, VCPU_CFAR(r9)
1307END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
1308BEGIN_FTR_SECTION
1309	ld	r4, HSTATE_PPR(r13)
1310	std	r4, VCPU_PPR(r9)
1311END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
1312
1313	/* Restore R1/R2 so we can handle faults */
1314	ld	r1, HSTATE_HOST_R1(r13)
1315	ld	r2, PACATOC(r13)
1316
1317	mfspr	r10, SPRN_SRR0
1318	mfspr	r11, SPRN_SRR1
1319	std	r10, VCPU_SRR0(r9)
1320	std	r11, VCPU_SRR1(r9)
1321	/* trap is in the low half of r12, clear CR from the high half */
1322	clrldi	r12, r12, 32
1323	andi.	r0, r12, 2		/* need to read HSRR0/1? */
1324	beq	1f
1325	mfspr	r10, SPRN_HSRR0
1326	mfspr	r11, SPRN_HSRR1
1327	clrrdi	r12, r12, 2
13281:	std	r10, VCPU_PC(r9)
1329	std	r11, VCPU_MSR(r9)
1330
1331	GET_SCRATCH0(r3)
1332	mflr	r4
1333	std	r3, VCPU_GPR(R13)(r9)
1334	std	r4, VCPU_LR(r9)
1335
1336	stw	r12,VCPU_TRAP(r9)
1337
1338	/*
1339	 * Now that we have saved away SRR0/1 and HSRR0/1,
1340	 * interrupts are recoverable in principle, so set MSR_RI.
1341	 * This becomes important for relocation-on interrupts from
1342	 * the guest, which we can get in radix mode on POWER9.
1343	 */
1344	li	r0, MSR_RI
1345	mtmsrd	r0, 1
1346
1347#ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
1348	addi	r3, r9, VCPU_TB_RMINTR
1349	mr	r4, r9
1350	bl	kvmhv_accumulate_time
1351	ld	r5, VCPU_GPR(R5)(r9)
1352	ld	r6, VCPU_GPR(R6)(r9)
1353	ld	r7, VCPU_GPR(R7)(r9)
1354	ld	r8, VCPU_GPR(R8)(r9)
1355#endif
1356
1357	/* Save HEIR (HV emulation assist reg) in emul_inst
1358	   if this is an HEI (HV emulation interrupt, e40) */
1359	li	r3,KVM_INST_FETCH_FAILED
1360	stw	r3,VCPU_LAST_INST(r9)
1361	cmpwi	r12,BOOK3S_INTERRUPT_H_EMUL_ASSIST
1362	bne	11f
1363	mfspr	r3,SPRN_HEIR
136411:	stw	r3,VCPU_HEIR(r9)
1365
1366	/* these are volatile across C function calls */
1367#ifdef CONFIG_RELOCATABLE
1368	ld	r3, HSTATE_SCRATCH1(r13)
1369	mtctr	r3
1370#else
1371	mfctr	r3
1372#endif
1373	mfxer	r4
1374	std	r3, VCPU_CTR(r9)
1375	std	r4, VCPU_XER(r9)
1376
1377	/* Save more register state  */
1378	mfdar	r3
1379	mfdsisr	r4
1380	std	r3, VCPU_DAR(r9)
1381	stw	r4, VCPU_DSISR(r9)
1382
1383	/* If this is a page table miss then see if it's theirs or ours */
1384	cmpwi	r12, BOOK3S_INTERRUPT_H_DATA_STORAGE
1385	beq	kvmppc_hdsi
1386	std	r3, VCPU_FAULT_DAR(r9)
1387	stw	r4, VCPU_FAULT_DSISR(r9)
1388	cmpwi	r12, BOOK3S_INTERRUPT_H_INST_STORAGE
1389	beq	kvmppc_hisi
1390
1391#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1392	/* For softpatch interrupt, go off and do TM instruction emulation */
1393	cmpwi	r12, BOOK3S_INTERRUPT_HV_SOFTPATCH
1394	beq	kvmppc_tm_emul
1395#endif
1396
1397	/* See if this is a leftover HDEC interrupt */
1398	cmpwi	r12,BOOK3S_INTERRUPT_HV_DECREMENTER
1399	bne	2f
1400	mfspr	r3,SPRN_HDEC
1401	EXTEND_HDEC(r3)
1402	cmpdi	r3,0
1403	mr	r4,r9
1404	bge	fast_guest_return
14052:
1406	/* See if this is an hcall we can handle in real mode */
1407	cmpwi	r12,BOOK3S_INTERRUPT_SYSCALL
1408	beq	hcall_try_real_mode
1409
1410	/* Hypervisor doorbell - exit only if host IPI flag set */
1411	cmpwi	r12, BOOK3S_INTERRUPT_H_DOORBELL
1412	bne	3f
1413BEGIN_FTR_SECTION
1414	PPC_MSGSYNC
1415	lwsync
1416	/* always exit if we're running a nested guest */
1417	ld	r0, VCPU_NESTED(r9)
1418	cmpdi	r0, 0
1419	bne	guest_exit_cont
1420END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
1421	lbz	r0, HSTATE_HOST_IPI(r13)
1422	cmpwi	r0, 0
1423	beq	maybe_reenter_guest
1424	b	guest_exit_cont
14253:
1426	/* If it's a hypervisor facility unavailable interrupt, save HFSCR */
1427	cmpwi	r12, BOOK3S_INTERRUPT_H_FAC_UNAVAIL
1428	bne	14f
1429	mfspr	r3, SPRN_HFSCR
1430	std	r3, VCPU_HFSCR(r9)
1431	b	guest_exit_cont
143214:
1433	/* External interrupt ? */
1434	cmpwi	r12, BOOK3S_INTERRUPT_EXTERNAL
1435	beq	kvmppc_guest_external
1436	/* See if it is a machine check */
1437	cmpwi	r12, BOOK3S_INTERRUPT_MACHINE_CHECK
1438	beq	machine_check_realmode
1439	/* Or a hypervisor maintenance interrupt */
1440	cmpwi	r12, BOOK3S_INTERRUPT_HMI
1441	beq	hmi_realmode
1442
1443guest_exit_cont:		/* r9 = vcpu, r12 = trap, r13 = paca */
1444
1445#ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
1446	addi	r3, r9, VCPU_TB_RMEXIT
1447	mr	r4, r9
1448	bl	kvmhv_accumulate_time
1449#endif
1450#ifdef CONFIG_KVM_XICS
1451	/* We are exiting, pull the VP from the XIVE */
1452	lbz	r0, VCPU_XIVE_PUSHED(r9)
1453	cmpwi	cr0, r0, 0
1454	beq	1f
1455	li	r7, TM_SPC_PULL_OS_CTX
1456	li	r6, TM_QW1_OS
1457	mfmsr	r0
1458	andi.	r0, r0, MSR_DR		/* in real mode? */
1459	beq	2f
1460	ld	r10, HSTATE_XIVE_TIMA_VIRT(r13)
1461	cmpldi	cr0, r10, 0
1462	beq	1f
1463	/* First load to pull the context, we ignore the value */
1464	eieio
1465	lwzx	r11, r7, r10
1466	/* Second load to recover the context state (Words 0 and 1) */
1467	ldx	r11, r6, r10
1468	b	3f
14692:	ld	r10, HSTATE_XIVE_TIMA_PHYS(r13)
1470	cmpldi	cr0, r10, 0
1471	beq	1f
1472	/* First load to pull the context, we ignore the value */
1473	eieio
1474	lwzcix	r11, r7, r10
1475	/* Second load to recover the context state (Words 0 and 1) */
1476	ldcix	r11, r6, r10
14773:	std	r11, VCPU_XIVE_SAVED_STATE(r9)
1478	/* Fixup some of the state for the next load */
1479	li	r10, 0
1480	li	r0, 0xff
1481	stb	r10, VCPU_XIVE_PUSHED(r9)
1482	stb	r10, (VCPU_XIVE_SAVED_STATE+3)(r9)
1483	stb	r0, (VCPU_XIVE_SAVED_STATE+4)(r9)
1484	eieio
14851:
1486#endif /* CONFIG_KVM_XICS */
1487
1488	/* If we came in through the P9 short path, go back out to C now */
1489	lwz	r0, STACK_SLOT_SHORT_PATH(r1)
1490	cmpwi	r0, 0
1491	bne	guest_exit_short_path
1492
1493	/* For hash guest, read the guest SLB and save it away */
1494	ld	r5, VCPU_KVM(r9)
1495	lbz	r0, KVM_RADIX(r5)
1496	li	r5, 0
1497	cmpwi	r0, 0
1498	bne	3f			/* for radix, save 0 entries */
1499	lwz	r0,VCPU_SLB_NR(r9)	/* number of entries in SLB */
1500	mtctr	r0
1501	li	r6,0
1502	addi	r7,r9,VCPU_SLB
15031:	slbmfee	r8,r6
1504	andis.	r0,r8,SLB_ESID_V@h
1505	beq	2f
1506	add	r8,r8,r6		/* put index in */
1507	slbmfev	r3,r6
1508	std	r8,VCPU_SLB_E(r7)
1509	std	r3,VCPU_SLB_V(r7)
1510	addi	r7,r7,VCPU_SLB_SIZE
1511	addi	r5,r5,1
15122:	addi	r6,r6,1
1513	bdnz	1b
1514	/* Finally clear out the SLB */
1515	li	r0,0
1516	slbmte	r0,r0
1517	slbia
1518	ptesync
15193:	stw	r5,VCPU_SLB_MAX(r9)
1520
1521	/* load host SLB entries */
1522BEGIN_MMU_FTR_SECTION
1523	b	0f
1524END_MMU_FTR_SECTION_IFSET(MMU_FTR_TYPE_RADIX)
1525	ld	r8,PACA_SLBSHADOWPTR(r13)
1526
1527	.rept	SLB_NUM_BOLTED
1528	li	r3, SLBSHADOW_SAVEAREA
1529	LDX_BE	r5, r8, r3
1530	addi	r3, r3, 8
1531	LDX_BE	r6, r8, r3
1532	andis.	r7,r5,SLB_ESID_V@h
1533	beq	1f
1534	slbmte	r6,r5
15351:	addi	r8,r8,16
1536	.endr
15370:
1538
1539guest_bypass:
1540	stw	r12, STACK_SLOT_TRAP(r1)
1541
1542	/* Save DEC */
1543	/* Do this before kvmhv_commence_exit so we know TB is guest TB */
1544	ld	r3, HSTATE_KVM_VCORE(r13)
1545	mfspr	r5,SPRN_DEC
1546	mftb	r6
1547	/* On P9, if the guest has large decr enabled, don't sign extend */
1548BEGIN_FTR_SECTION
1549	ld	r4, VCORE_LPCR(r3)
1550	andis.	r4, r4, LPCR_LD@h
1551	bne	16f
1552END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
1553	extsw	r5,r5
155416:	add	r5,r5,r6
1555	/* r5 is a guest timebase value here, convert to host TB */
1556	ld	r4,VCORE_TB_OFFSET_APPL(r3)
1557	subf	r5,r4,r5
1558	std	r5,VCPU_DEC_EXPIRES(r9)
1559
1560	/* Increment exit count, poke other threads to exit */
1561	mr 	r3, r12
1562	bl	kvmhv_commence_exit
1563	nop
1564	ld	r9, HSTATE_KVM_VCPU(r13)
1565
1566	/* Stop others sending VCPU interrupts to this physical CPU */
1567	li	r0, -1
1568	stw	r0, VCPU_CPU(r9)
1569	stw	r0, VCPU_THREAD_CPU(r9)
1570
1571	/* Save guest CTRL register, set runlatch to 1 */
1572	mfspr	r6,SPRN_CTRLF
1573	stw	r6,VCPU_CTRL(r9)
1574	andi.	r0,r6,1
1575	bne	4f
1576	ori	r6,r6,1
1577	mtspr	SPRN_CTRLT,r6
15784:
1579	/*
1580	 * Save the guest PURR/SPURR
1581	 */
1582	mfspr	r5,SPRN_PURR
1583	mfspr	r6,SPRN_SPURR
1584	ld	r7,VCPU_PURR(r9)
1585	ld	r8,VCPU_SPURR(r9)
1586	std	r5,VCPU_PURR(r9)
1587	std	r6,VCPU_SPURR(r9)
1588	subf	r5,r7,r5
1589	subf	r6,r8,r6
1590
1591	/*
1592	 * Restore host PURR/SPURR and add guest times
1593	 * so that the time in the guest gets accounted.
1594	 */
1595	ld	r3,HSTATE_PURR(r13)
1596	ld	r4,HSTATE_SPURR(r13)
1597	add	r3,r3,r5
1598	add	r4,r4,r6
1599	mtspr	SPRN_PURR,r3
1600	mtspr	SPRN_SPURR,r4
1601
1602BEGIN_FTR_SECTION
1603	b	8f
1604END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
1605	/* Save POWER8-specific registers */
1606	mfspr	r5, SPRN_IAMR
1607	mfspr	r6, SPRN_PSPB
1608	mfspr	r7, SPRN_FSCR
1609	std	r5, VCPU_IAMR(r9)
1610	stw	r6, VCPU_PSPB(r9)
1611	std	r7, VCPU_FSCR(r9)
1612	mfspr	r5, SPRN_IC
1613	mfspr	r7, SPRN_TAR
1614	std	r5, VCPU_IC(r9)
1615	std	r7, VCPU_TAR(r9)
1616	mfspr	r8, SPRN_EBBHR
1617	std	r8, VCPU_EBBHR(r9)
1618	mfspr	r5, SPRN_EBBRR
1619	mfspr	r6, SPRN_BESCR
1620	mfspr	r7, SPRN_PID
1621	mfspr	r8, SPRN_WORT
1622	std	r5, VCPU_EBBRR(r9)
1623	std	r6, VCPU_BESCR(r9)
1624	stw	r7, VCPU_GUEST_PID(r9)
1625	std	r8, VCPU_WORT(r9)
1626BEGIN_FTR_SECTION
1627	mfspr	r5, SPRN_TCSCR
1628	mfspr	r6, SPRN_ACOP
1629	mfspr	r7, SPRN_CSIGR
1630	mfspr	r8, SPRN_TACR
1631	std	r5, VCPU_TCSCR(r9)
1632	std	r6, VCPU_ACOP(r9)
1633	std	r7, VCPU_CSIGR(r9)
1634	std	r8, VCPU_TACR(r9)
1635FTR_SECTION_ELSE
1636	mfspr	r5, SPRN_TIDR
1637	mfspr	r6, SPRN_PSSCR
1638	std	r5, VCPU_TID(r9)
1639	rldicl	r6, r6, 4, 50		/* r6 &= PSSCR_GUEST_VIS */
1640	rotldi	r6, r6, 60
1641	std	r6, VCPU_PSSCR(r9)
1642	/* Restore host HFSCR value */
1643	ld	r7, STACK_SLOT_HFSCR(r1)
1644	mtspr	SPRN_HFSCR, r7
1645ALT_FTR_SECTION_END_IFCLR(CPU_FTR_ARCH_300)
1646	/*
1647	 * Restore various registers to 0, where non-zero values
1648	 * set by the guest could disrupt the host.
1649	 */
1650	li	r0, 0
1651	mtspr	SPRN_PSPB, r0
1652	mtspr	SPRN_WORT, r0
1653BEGIN_FTR_SECTION
1654	mtspr	SPRN_TCSCR, r0
1655	/* Set MMCRS to 1<<31 to freeze and disable the SPMC counters */
1656	li	r0, 1
1657	sldi	r0, r0, 31
1658	mtspr	SPRN_MMCRS, r0
1659END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_300)
1660
1661	/* Save and restore AMR, IAMR and UAMOR before turning on the MMU */
1662	ld	r8, STACK_SLOT_IAMR(r1)
1663	mtspr	SPRN_IAMR, r8
1664
16658:	/* Power7 jumps back in here */
1666	mfspr	r5,SPRN_AMR
1667	mfspr	r6,SPRN_UAMOR
1668	std	r5,VCPU_AMR(r9)
1669	std	r6,VCPU_UAMOR(r9)
1670	ld	r5,STACK_SLOT_AMR(r1)
1671	ld	r6,STACK_SLOT_UAMOR(r1)
1672	mtspr	SPRN_AMR, r5
1673	mtspr	SPRN_UAMOR, r6
1674
1675	/* Switch DSCR back to host value */
1676	mfspr	r8, SPRN_DSCR
1677	ld	r7, HSTATE_DSCR(r13)
1678	std	r8, VCPU_DSCR(r9)
1679	mtspr	SPRN_DSCR, r7
1680
1681	/* Save non-volatile GPRs */
1682	std	r14, VCPU_GPR(R14)(r9)
1683	std	r15, VCPU_GPR(R15)(r9)
1684	std	r16, VCPU_GPR(R16)(r9)
1685	std	r17, VCPU_GPR(R17)(r9)
1686	std	r18, VCPU_GPR(R18)(r9)
1687	std	r19, VCPU_GPR(R19)(r9)
1688	std	r20, VCPU_GPR(R20)(r9)
1689	std	r21, VCPU_GPR(R21)(r9)
1690	std	r22, VCPU_GPR(R22)(r9)
1691	std	r23, VCPU_GPR(R23)(r9)
1692	std	r24, VCPU_GPR(R24)(r9)
1693	std	r25, VCPU_GPR(R25)(r9)
1694	std	r26, VCPU_GPR(R26)(r9)
1695	std	r27, VCPU_GPR(R27)(r9)
1696	std	r28, VCPU_GPR(R28)(r9)
1697	std	r29, VCPU_GPR(R29)(r9)
1698	std	r30, VCPU_GPR(R30)(r9)
1699	std	r31, VCPU_GPR(R31)(r9)
1700
1701	/* Save SPRGs */
1702	mfspr	r3, SPRN_SPRG0
1703	mfspr	r4, SPRN_SPRG1
1704	mfspr	r5, SPRN_SPRG2
1705	mfspr	r6, SPRN_SPRG3
1706	std	r3, VCPU_SPRG0(r9)
1707	std	r4, VCPU_SPRG1(r9)
1708	std	r5, VCPU_SPRG2(r9)
1709	std	r6, VCPU_SPRG3(r9)
1710
1711	/* save FP state */
1712	mr	r3, r9
1713	bl	kvmppc_save_fp
1714
1715#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1716/*
1717 * Branch around the call if both CPU_FTR_TM and
1718 * CPU_FTR_P9_TM_HV_ASSIST are off.
1719 */
1720BEGIN_FTR_SECTION
1721	b	91f
1722END_FTR_SECTION(CPU_FTR_TM | CPU_FTR_P9_TM_HV_ASSIST, 0)
1723	/*
1724	 * NOTE THAT THIS TRASHES ALL NON-VOLATILE REGISTERS (but not CR)
1725	 */
1726	mr      r3, r9
1727	ld      r4, VCPU_MSR(r3)
1728	li	r5, 0			/* don't preserve non-vol regs */
1729	bl	kvmppc_save_tm_hv
1730	nop
1731	ld	r9, HSTATE_KVM_VCPU(r13)
173291:
1733#endif
1734
1735	/* Increment yield count if they have a VPA */
1736	ld	r8, VCPU_VPA(r9)	/* do they have a VPA? */
1737	cmpdi	r8, 0
1738	beq	25f
1739	li	r4, LPPACA_YIELDCOUNT
1740	LWZX_BE	r3, r8, r4
1741	addi	r3, r3, 1
1742	STWX_BE	r3, r8, r4
1743	li	r3, 1
1744	stb	r3, VCPU_VPA_DIRTY(r9)
174525:
1746	/* Save PMU registers if requested */
1747	/* r8 and cr0.eq are live here */
1748	mr	r3, r9
1749	li	r4, 1
1750	beq	21f			/* if no VPA, save PMU stuff anyway */
1751	lbz	r4, LPPACA_PMCINUSE(r8)
175221:	bl	kvmhv_save_guest_pmu
1753	ld	r9, HSTATE_KVM_VCPU(r13)
1754
1755	/* Restore host values of some registers */
1756BEGIN_FTR_SECTION
1757	ld	r5, STACK_SLOT_CIABR(r1)
1758	ld	r6, STACK_SLOT_DAWR(r1)
1759	ld	r7, STACK_SLOT_DAWRX(r1)
1760	mtspr	SPRN_CIABR, r5
1761	/*
1762	 * If the DAWR doesn't work, it's ok to write these here as
1763	 * this value should always be zero
1764	*/
1765	mtspr	SPRN_DAWR, r6
1766	mtspr	SPRN_DAWRX, r7
1767END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
1768BEGIN_FTR_SECTION
1769	ld	r5, STACK_SLOT_TID(r1)
1770	ld	r6, STACK_SLOT_PSSCR(r1)
1771	ld	r7, STACK_SLOT_PID(r1)
1772	mtspr	SPRN_TIDR, r5
1773	mtspr	SPRN_PSSCR, r6
1774	mtspr	SPRN_PID, r7
1775END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
1776
1777#ifdef CONFIG_PPC_RADIX_MMU
1778	/*
1779	 * Are we running hash or radix ?
1780	 */
1781	ld	r5, VCPU_KVM(r9)
1782	lbz	r0, KVM_RADIX(r5)
1783	cmpwi	cr2, r0, 0
1784	beq	cr2, 2f
1785
1786	/*
1787	 * Radix: do eieio; tlbsync; ptesync sequence in case we
1788	 * interrupted the guest between a tlbie and a ptesync.
1789	 */
1790	eieio
1791	tlbsync
1792	ptesync
1793
1794	/* Radix: Handle the case where the guest used an illegal PID */
1795	LOAD_REG_ADDR(r4, mmu_base_pid)
1796	lwz	r3, VCPU_GUEST_PID(r9)
1797	lwz	r5, 0(r4)
1798	cmpw	cr0,r3,r5
1799	blt	2f
1800
1801	/*
1802	 * Illegal PID, the HW might have prefetched and cached in the TLB
1803	 * some translations for the  LPID 0 / guest PID combination which
1804	 * Linux doesn't know about, so we need to flush that PID out of
1805	 * the TLB. First we need to set LPIDR to 0 so tlbiel applies to
1806	 * the right context.
1807	*/
1808	li	r0,0
1809	mtspr	SPRN_LPID,r0
1810	isync
1811
1812	/* Then do a congruence class local flush */
1813	ld	r6,VCPU_KVM(r9)
1814	lwz	r0,KVM_TLB_SETS(r6)
1815	mtctr	r0
1816	li	r7,0x400		/* IS field = 0b01 */
1817	ptesync
1818	sldi	r0,r3,32		/* RS has PID */
18191:	PPC_TLBIEL(7,0,2,1,1)		/* RIC=2, PRS=1, R=1 */
1820	addi	r7,r7,0x1000
1821	bdnz	1b
1822	ptesync
1823
18242:
1825#endif /* CONFIG_PPC_RADIX_MMU */
1826
1827	/*
1828	 * POWER7/POWER8 guest -> host partition switch code.
1829	 * We don't have to lock against tlbies but we do
1830	 * have to coordinate the hardware threads.
1831	 * Here STACK_SLOT_TRAP(r1) contains the trap number.
1832	 */
1833kvmhv_switch_to_host:
1834	/* Secondary threads wait for primary to do partition switch */
1835	ld	r5,HSTATE_KVM_VCORE(r13)
1836	ld	r4,VCORE_KVM(r5)	/* pointer to struct kvm */
1837	lbz	r3,HSTATE_PTID(r13)
1838	cmpwi	r3,0
1839	beq	15f
1840	HMT_LOW
184113:	lbz	r3,VCORE_IN_GUEST(r5)
1842	cmpwi	r3,0
1843	bne	13b
1844	HMT_MEDIUM
1845	b	16f
1846
1847	/* Primary thread waits for all the secondaries to exit guest */
184815:	lwz	r3,VCORE_ENTRY_EXIT(r5)
1849	rlwinm	r0,r3,32-8,0xff
1850	clrldi	r3,r3,56
1851	cmpw	r3,r0
1852	bne	15b
1853	isync
1854
1855	/* Did we actually switch to the guest at all? */
1856	lbz	r6, VCORE_IN_GUEST(r5)
1857	cmpwi	r6, 0
1858	beq	19f
1859
1860	/* Primary thread switches back to host partition */
1861	lwz	r7,KVM_HOST_LPID(r4)
1862BEGIN_FTR_SECTION
1863	ld	r6,KVM_HOST_SDR1(r4)
1864	li	r8,LPID_RSVD		/* switch to reserved LPID */
1865	mtspr	SPRN_LPID,r8
1866	ptesync
1867	mtspr	SPRN_SDR1,r6		/* switch to host page table */
1868END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_300)
1869	mtspr	SPRN_LPID,r7
1870	isync
1871
1872BEGIN_FTR_SECTION
1873	/* DPDES and VTB are shared between threads */
1874	mfspr	r7, SPRN_DPDES
1875	mfspr	r8, SPRN_VTB
1876	std	r7, VCORE_DPDES(r5)
1877	std	r8, VCORE_VTB(r5)
1878	/* clear DPDES so we don't get guest doorbells in the host */
1879	li	r8, 0
1880	mtspr	SPRN_DPDES, r8
1881END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
1882
1883	/* Subtract timebase offset from timebase */
1884	ld	r8, VCORE_TB_OFFSET_APPL(r5)
1885	cmpdi	r8,0
1886	beq	17f
1887	li	r0, 0
1888	std	r0, VCORE_TB_OFFSET_APPL(r5)
1889	mftb	r6			/* current guest timebase */
1890	subf	r8,r8,r6
1891	mtspr	SPRN_TBU40,r8		/* update upper 40 bits */
1892	mftb	r7			/* check if lower 24 bits overflowed */
1893	clrldi	r6,r6,40
1894	clrldi	r7,r7,40
1895	cmpld	r7,r6
1896	bge	17f
1897	addis	r8,r8,0x100		/* if so, increment upper 40 bits */
1898	mtspr	SPRN_TBU40,r8
1899
190017:
1901	/*
1902	 * If this is an HMI, we called kvmppc_realmode_hmi_handler
1903	 * above, which may or may not have already called
1904	 * kvmppc_subcore_exit_guest.  Fortunately, all that
1905	 * kvmppc_subcore_exit_guest does is clear a flag, so calling
1906	 * it again here is benign even if kvmppc_realmode_hmi_handler
1907	 * has already called it.
1908	 */
1909	bl	kvmppc_subcore_exit_guest
1910	nop
191130:	ld	r5,HSTATE_KVM_VCORE(r13)
1912	ld	r4,VCORE_KVM(r5)	/* pointer to struct kvm */
1913
1914	/* Reset PCR */
1915	ld	r0, VCORE_PCR(r5)
1916	cmpdi	r0, 0
1917	beq	18f
1918	li	r0, 0
1919	mtspr	SPRN_PCR, r0
192018:
1921	/* Signal secondary CPUs to continue */
1922	stb	r0,VCORE_IN_GUEST(r5)
192319:	lis	r8,0x7fff		/* MAX_INT@h */
1924	mtspr	SPRN_HDEC,r8
1925
192616:
1927BEGIN_FTR_SECTION
1928	/* On POWER9 with HPT-on-radix we need to wait for all other threads */
1929	ld	r3, HSTATE_SPLIT_MODE(r13)
1930	cmpdi	r3, 0
1931	beq	47f
1932	lwz	r8, KVM_SPLIT_DO_RESTORE(r3)
1933	cmpwi	r8, 0
1934	beq	47f
1935	bl	kvmhv_p9_restore_lpcr
1936	nop
1937	b	48f
193847:
1939END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
1940	ld	r8,KVM_HOST_LPCR(r4)
1941	mtspr	SPRN_LPCR,r8
1942	isync
194348:
1944#ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
1945	/* Finish timing, if we have a vcpu */
1946	ld	r4, HSTATE_KVM_VCPU(r13)
1947	cmpdi	r4, 0
1948	li	r3, 0
1949	beq	2f
1950	bl	kvmhv_accumulate_time
19512:
1952#endif
1953	/* Unset guest mode */
1954	li	r0, KVM_GUEST_MODE_NONE
1955	stb	r0, HSTATE_IN_GUEST(r13)
1956
1957	lwz	r12, STACK_SLOT_TRAP(r1)	/* return trap # in r12 */
1958	ld	r0, SFS+PPC_LR_STKOFF(r1)
1959	addi	r1, r1, SFS
1960	mtlr	r0
1961	blr
1962
1963kvmppc_guest_external:
1964	/* External interrupt, first check for host_ipi. If this is
1965	 * set, we know the host wants us out so let's do it now
1966	 */
1967	bl	kvmppc_read_intr
1968
1969	/*
1970	 * Restore the active volatile registers after returning from
1971	 * a C function.
1972	 */
1973	ld	r9, HSTATE_KVM_VCPU(r13)
1974	li	r12, BOOK3S_INTERRUPT_EXTERNAL
1975
1976	/*
1977	 * kvmppc_read_intr return codes:
1978	 *
1979	 * Exit to host (r3 > 0)
1980	 *   1 An interrupt is pending that needs to be handled by the host
1981	 *     Exit guest and return to host by branching to guest_exit_cont
1982	 *
1983	 *   2 Passthrough that needs completion in the host
1984	 *     Exit guest and return to host by branching to guest_exit_cont
1985	 *     However, we also set r12 to BOOK3S_INTERRUPT_HV_RM_HARD
1986	 *     to indicate to the host to complete handling the interrupt
1987	 *
1988	 * Before returning to guest, we check if any CPU is heading out
1989	 * to the host and if so, we head out also. If no CPUs are heading
1990	 * check return values <= 0.
1991	 *
1992	 * Return to guest (r3 <= 0)
1993	 *  0 No external interrupt is pending
1994	 * -1 A guest wakeup IPI (which has now been cleared)
1995	 *    In either case, we return to guest to deliver any pending
1996	 *    guest interrupts.
1997	 *
1998	 * -2 A PCI passthrough external interrupt was handled
1999	 *    (interrupt was delivered directly to guest)
2000	 *    Return to guest to deliver any pending guest interrupts.
2001	 */
2002
2003	cmpdi	r3, 1
2004	ble	1f
2005
2006	/* Return code = 2 */
2007	li	r12, BOOK3S_INTERRUPT_HV_RM_HARD
2008	stw	r12, VCPU_TRAP(r9)
2009	b	guest_exit_cont
2010
20111:	/* Return code <= 1 */
2012	cmpdi	r3, 0
2013	bgt	guest_exit_cont
2014
2015	/* Return code <= 0 */
2016maybe_reenter_guest:
2017	ld	r5, HSTATE_KVM_VCORE(r13)
2018	lwz	r0, VCORE_ENTRY_EXIT(r5)
2019	cmpwi	r0, 0x100
2020	mr	r4, r9
2021	blt	deliver_guest_interrupt
2022	b	guest_exit_cont
2023
2024#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2025/*
2026 * Softpatch interrupt for transactional memory emulation cases
2027 * on POWER9 DD2.2.  This is early in the guest exit path - we
2028 * haven't saved registers or done a treclaim yet.
2029 */
2030kvmppc_tm_emul:
2031	/* Save instruction image in HEIR */
2032	mfspr	r3, SPRN_HEIR
2033	stw	r3, VCPU_HEIR(r9)
2034
2035	/*
2036	 * The cases we want to handle here are those where the guest
2037	 * is in real suspend mode and is trying to transition to
2038	 * transactional mode.
2039	 */
2040	lbz	r0, HSTATE_FAKE_SUSPEND(r13)
2041	cmpwi	r0, 0		/* keep exiting guest if in fake suspend */
2042	bne	guest_exit_cont
2043	rldicl	r3, r11, 64 - MSR_TS_S_LG, 62
2044	cmpwi	r3, 1		/* or if not in suspend state */
2045	bne	guest_exit_cont
2046
2047	/* Call C code to do the emulation */
2048	mr	r3, r9
2049	bl	kvmhv_p9_tm_emulation_early
2050	nop
2051	ld	r9, HSTATE_KVM_VCPU(r13)
2052	li	r12, BOOK3S_INTERRUPT_HV_SOFTPATCH
2053	cmpwi	r3, 0
2054	beq	guest_exit_cont		/* continue exiting if not handled */
2055	ld	r10, VCPU_PC(r9)
2056	ld	r11, VCPU_MSR(r9)
2057	b	fast_interrupt_c_return	/* go back to guest if handled */
2058#endif /* CONFIG_PPC_TRANSACTIONAL_MEM */
2059
2060/*
2061 * Check whether an HDSI is an HPTE not found fault or something else.
2062 * If it is an HPTE not found fault that is due to the guest accessing
2063 * a page that they have mapped but which we have paged out, then
2064 * we continue on with the guest exit path.  In all other cases,
2065 * reflect the HDSI to the guest as a DSI.
2066 */
2067kvmppc_hdsi:
2068	ld	r3, VCPU_KVM(r9)
2069	lbz	r0, KVM_RADIX(r3)
2070	mfspr	r4, SPRN_HDAR
2071	mfspr	r6, SPRN_HDSISR
2072BEGIN_FTR_SECTION
2073	/* Look for DSISR canary. If we find it, retry instruction */
2074	cmpdi	r6, 0x7fff
2075	beq	6f
2076END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
2077	cmpwi	r0, 0
2078	bne	.Lradix_hdsi		/* on radix, just save DAR/DSISR/ASDR */
2079	/* HPTE not found fault or protection fault? */
2080	andis.	r0, r6, (DSISR_NOHPTE | DSISR_PROTFAULT)@h
2081	beq	1f			/* if not, send it to the guest */
2082	andi.	r0, r11, MSR_DR		/* data relocation enabled? */
2083	beq	3f
2084BEGIN_FTR_SECTION
2085	mfspr	r5, SPRN_ASDR		/* on POWER9, use ASDR to get VSID */
2086	b	4f
2087END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
2088	clrrdi	r0, r4, 28
2089	PPC_SLBFEE_DOT(R5, R0)		/* if so, look up SLB */
2090	li	r0, BOOK3S_INTERRUPT_DATA_SEGMENT
2091	bne	7f			/* if no SLB entry found */
20924:	std	r4, VCPU_FAULT_DAR(r9)
2093	stw	r6, VCPU_FAULT_DSISR(r9)
2094
2095	/* Search the hash table. */
2096	mr	r3, r9			/* vcpu pointer */
2097	li	r7, 1			/* data fault */
2098	bl	kvmppc_hpte_hv_fault
2099	ld	r9, HSTATE_KVM_VCPU(r13)
2100	ld	r10, VCPU_PC(r9)
2101	ld	r11, VCPU_MSR(r9)
2102	li	r12, BOOK3S_INTERRUPT_H_DATA_STORAGE
2103	cmpdi	r3, 0			/* retry the instruction */
2104	beq	6f
2105	cmpdi	r3, -1			/* handle in kernel mode */
2106	beq	guest_exit_cont
2107	cmpdi	r3, -2			/* MMIO emulation; need instr word */
2108	beq	2f
2109
2110	/* Synthesize a DSI (or DSegI) for the guest */
2111	ld	r4, VCPU_FAULT_DAR(r9)
2112	mr	r6, r3
21131:	li	r0, BOOK3S_INTERRUPT_DATA_STORAGE
2114	mtspr	SPRN_DSISR, r6
21157:	mtspr	SPRN_DAR, r4
2116	mtspr	SPRN_SRR0, r10
2117	mtspr	SPRN_SRR1, r11
2118	mr	r10, r0
2119	bl	kvmppc_msr_interrupt
2120fast_interrupt_c_return:
21216:	ld	r7, VCPU_CTR(r9)
2122	ld	r8, VCPU_XER(r9)
2123	mtctr	r7
2124	mtxer	r8
2125	mr	r4, r9
2126	b	fast_guest_return
2127
21283:	ld	r5, VCPU_KVM(r9)	/* not relocated, use VRMA */
2129	ld	r5, KVM_VRMA_SLB_V(r5)
2130	b	4b
2131
2132	/* If this is for emulated MMIO, load the instruction word */
21332:	li	r8, KVM_INST_FETCH_FAILED	/* In case lwz faults */
2134
2135	/* Set guest mode to 'jump over instruction' so if lwz faults
2136	 * we'll just continue at the next IP. */
2137	li	r0, KVM_GUEST_MODE_SKIP
2138	stb	r0, HSTATE_IN_GUEST(r13)
2139
2140	/* Do the access with MSR:DR enabled */
2141	mfmsr	r3
2142	ori	r4, r3, MSR_DR		/* Enable paging for data */
2143	mtmsrd	r4
2144	lwz	r8, 0(r10)
2145	mtmsrd	r3
2146
2147	/* Store the result */
2148	stw	r8, VCPU_LAST_INST(r9)
2149
2150	/* Unset guest mode. */
2151	li	r0, KVM_GUEST_MODE_HOST_HV
2152	stb	r0, HSTATE_IN_GUEST(r13)
2153	b	guest_exit_cont
2154
2155.Lradix_hdsi:
2156	std	r4, VCPU_FAULT_DAR(r9)
2157	stw	r6, VCPU_FAULT_DSISR(r9)
2158.Lradix_hisi:
2159	mfspr	r5, SPRN_ASDR
2160	std	r5, VCPU_FAULT_GPA(r9)
2161	b	guest_exit_cont
2162
2163/*
2164 * Similarly for an HISI, reflect it to the guest as an ISI unless
2165 * it is an HPTE not found fault for a page that we have paged out.
2166 */
2167kvmppc_hisi:
2168	ld	r3, VCPU_KVM(r9)
2169	lbz	r0, KVM_RADIX(r3)
2170	cmpwi	r0, 0
2171	bne	.Lradix_hisi		/* for radix, just save ASDR */
2172	andis.	r0, r11, SRR1_ISI_NOPT@h
2173	beq	1f
2174	andi.	r0, r11, MSR_IR		/* instruction relocation enabled? */
2175	beq	3f
2176BEGIN_FTR_SECTION
2177	mfspr	r5, SPRN_ASDR		/* on POWER9, use ASDR to get VSID */
2178	b	4f
2179END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
2180	clrrdi	r0, r10, 28
2181	PPC_SLBFEE_DOT(R5, R0)		/* if so, look up SLB */
2182	li	r0, BOOK3S_INTERRUPT_INST_SEGMENT
2183	bne	7f			/* if no SLB entry found */
21844:
2185	/* Search the hash table. */
2186	mr	r3, r9			/* vcpu pointer */
2187	mr	r4, r10
2188	mr	r6, r11
2189	li	r7, 0			/* instruction fault */
2190	bl	kvmppc_hpte_hv_fault
2191	ld	r9, HSTATE_KVM_VCPU(r13)
2192	ld	r10, VCPU_PC(r9)
2193	ld	r11, VCPU_MSR(r9)
2194	li	r12, BOOK3S_INTERRUPT_H_INST_STORAGE
2195	cmpdi	r3, 0			/* retry the instruction */
2196	beq	fast_interrupt_c_return
2197	cmpdi	r3, -1			/* handle in kernel mode */
2198	beq	guest_exit_cont
2199
2200	/* Synthesize an ISI (or ISegI) for the guest */
2201	mr	r11, r3
22021:	li	r0, BOOK3S_INTERRUPT_INST_STORAGE
22037:	mtspr	SPRN_SRR0, r10
2204	mtspr	SPRN_SRR1, r11
2205	mr	r10, r0
2206	bl	kvmppc_msr_interrupt
2207	b	fast_interrupt_c_return
2208
22093:	ld	r6, VCPU_KVM(r9)	/* not relocated, use VRMA */
2210	ld	r5, KVM_VRMA_SLB_V(r6)
2211	b	4b
2212
2213/*
2214 * Try to handle an hcall in real mode.
2215 * Returns to the guest if we handle it, or continues on up to
2216 * the kernel if we can't (i.e. if we don't have a handler for
2217 * it, or if the handler returns H_TOO_HARD).
2218 *
2219 * r5 - r8 contain hcall args,
2220 * r9 = vcpu, r10 = pc, r11 = msr, r12 = trap, r13 = paca
2221 */
2222hcall_try_real_mode:
2223	ld	r3,VCPU_GPR(R3)(r9)
2224	andi.	r0,r11,MSR_PR
2225	/* sc 1 from userspace - reflect to guest syscall */
2226	bne	sc_1_fast_return
2227	/* sc 1 from nested guest - give it to L1 to handle */
2228	ld	r0, VCPU_NESTED(r9)
2229	cmpdi	r0, 0
2230	bne	guest_exit_cont
2231	clrrdi	r3,r3,2
2232	cmpldi	r3,hcall_real_table_end - hcall_real_table
2233	bge	guest_exit_cont
2234	/* See if this hcall is enabled for in-kernel handling */
2235	ld	r4, VCPU_KVM(r9)
2236	srdi	r0, r3, 8	/* r0 = (r3 / 4) >> 6 */
2237	sldi	r0, r0, 3	/* index into kvm->arch.enabled_hcalls[] */
2238	add	r4, r4, r0
2239	ld	r0, KVM_ENABLED_HCALLS(r4)
2240	rlwinm	r4, r3, 32-2, 0x3f	/* r4 = (r3 / 4) & 0x3f */
2241	srd	r0, r0, r4
2242	andi.	r0, r0, 1
2243	beq	guest_exit_cont
2244	/* Get pointer to handler, if any, and call it */
2245	LOAD_REG_ADDR(r4, hcall_real_table)
2246	lwax	r3,r3,r4
2247	cmpwi	r3,0
2248	beq	guest_exit_cont
2249	add	r12,r3,r4
2250	mtctr	r12
2251	mr	r3,r9		/* get vcpu pointer */
2252	ld	r4,VCPU_GPR(R4)(r9)
2253	bctrl
2254	cmpdi	r3,H_TOO_HARD
2255	beq	hcall_real_fallback
2256	ld	r4,HSTATE_KVM_VCPU(r13)
2257	std	r3,VCPU_GPR(R3)(r4)
2258	ld	r10,VCPU_PC(r4)
2259	ld	r11,VCPU_MSR(r4)
2260	b	fast_guest_return
2261
2262sc_1_fast_return:
2263	mtspr	SPRN_SRR0,r10
2264	mtspr	SPRN_SRR1,r11
2265	li	r10, BOOK3S_INTERRUPT_SYSCALL
2266	bl	kvmppc_msr_interrupt
2267	mr	r4,r9
2268	b	fast_guest_return
2269
2270	/* We've attempted a real mode hcall, but it's punted it back
2271	 * to userspace.  We need to restore some clobbered volatiles
2272	 * before resuming the pass-it-to-qemu path */
2273hcall_real_fallback:
2274	li	r12,BOOK3S_INTERRUPT_SYSCALL
2275	ld	r9, HSTATE_KVM_VCPU(r13)
2276
2277	b	guest_exit_cont
2278
2279	.globl	hcall_real_table
2280hcall_real_table:
2281	.long	0		/* 0 - unused */
2282	.long	DOTSYM(kvmppc_h_remove) - hcall_real_table
2283	.long	DOTSYM(kvmppc_h_enter) - hcall_real_table
2284	.long	DOTSYM(kvmppc_h_read) - hcall_real_table
2285	.long	DOTSYM(kvmppc_h_clear_mod) - hcall_real_table
2286	.long	DOTSYM(kvmppc_h_clear_ref) - hcall_real_table
2287	.long	DOTSYM(kvmppc_h_protect) - hcall_real_table
2288#ifdef CONFIG_SPAPR_TCE_IOMMU
2289	.long	DOTSYM(kvmppc_h_get_tce) - hcall_real_table
2290	.long	DOTSYM(kvmppc_rm_h_put_tce) - hcall_real_table
2291#else
2292	.long	0		/* 0x1c */
2293	.long	0		/* 0x20 */
2294#endif
2295	.long	0		/* 0x24 - H_SET_SPRG0 */
2296	.long	DOTSYM(kvmppc_h_set_dabr) - hcall_real_table
2297	.long	DOTSYM(kvmppc_rm_h_page_init) - hcall_real_table
2298	.long	0		/* 0x30 */
2299	.long	0		/* 0x34 */
2300	.long	0		/* 0x38 */
2301	.long	0		/* 0x3c */
2302	.long	0		/* 0x40 */
2303	.long	0		/* 0x44 */
2304	.long	0		/* 0x48 */
2305	.long	0		/* 0x4c */
2306	.long	0		/* 0x50 */
2307	.long	0		/* 0x54 */
2308	.long	0		/* 0x58 */
2309	.long	0		/* 0x5c */
2310	.long	0		/* 0x60 */
2311#ifdef CONFIG_KVM_XICS
2312	.long	DOTSYM(kvmppc_rm_h_eoi) - hcall_real_table
2313	.long	DOTSYM(kvmppc_rm_h_cppr) - hcall_real_table
2314	.long	DOTSYM(kvmppc_rm_h_ipi) - hcall_real_table
2315	.long	DOTSYM(kvmppc_rm_h_ipoll) - hcall_real_table
2316	.long	DOTSYM(kvmppc_rm_h_xirr) - hcall_real_table
2317#else
2318	.long	0		/* 0x64 - H_EOI */
2319	.long	0		/* 0x68 - H_CPPR */
2320	.long	0		/* 0x6c - H_IPI */
2321	.long	0		/* 0x70 - H_IPOLL */
2322	.long	0		/* 0x74 - H_XIRR */
2323#endif
2324	.long	0		/* 0x78 */
2325	.long	0		/* 0x7c */
2326	.long	0		/* 0x80 */
2327	.long	0		/* 0x84 */
2328	.long	0		/* 0x88 */
2329	.long	0		/* 0x8c */
2330	.long	0		/* 0x90 */
2331	.long	0		/* 0x94 */
2332	.long	0		/* 0x98 */
2333	.long	0		/* 0x9c */
2334	.long	0		/* 0xa0 */
2335	.long	0		/* 0xa4 */
2336	.long	0		/* 0xa8 */
2337	.long	0		/* 0xac */
2338	.long	0		/* 0xb0 */
2339	.long	0		/* 0xb4 */
2340	.long	0		/* 0xb8 */
2341	.long	0		/* 0xbc */
2342	.long	0		/* 0xc0 */
2343	.long	0		/* 0xc4 */
2344	.long	0		/* 0xc8 */
2345	.long	0		/* 0xcc */
2346	.long	0		/* 0xd0 */
2347	.long	0		/* 0xd4 */
2348	.long	0		/* 0xd8 */
2349	.long	0		/* 0xdc */
2350	.long	DOTSYM(kvmppc_h_cede) - hcall_real_table
2351	.long	DOTSYM(kvmppc_rm_h_confer) - hcall_real_table
2352	.long	0		/* 0xe8 */
2353	.long	0		/* 0xec */
2354	.long	0		/* 0xf0 */
2355	.long	0		/* 0xf4 */
2356	.long	0		/* 0xf8 */
2357	.long	0		/* 0xfc */
2358	.long	0		/* 0x100 */
2359	.long	0		/* 0x104 */
2360	.long	0		/* 0x108 */
2361	.long	0		/* 0x10c */
2362	.long	0		/* 0x110 */
2363	.long	0		/* 0x114 */
2364	.long	0		/* 0x118 */
2365	.long	0		/* 0x11c */
2366	.long	0		/* 0x120 */
2367	.long	DOTSYM(kvmppc_h_bulk_remove) - hcall_real_table
2368	.long	0		/* 0x128 */
2369	.long	0		/* 0x12c */
2370	.long	0		/* 0x130 */
2371	.long	DOTSYM(kvmppc_h_set_xdabr) - hcall_real_table
2372#ifdef CONFIG_SPAPR_TCE_IOMMU
2373	.long	DOTSYM(kvmppc_rm_h_stuff_tce) - hcall_real_table
2374	.long	DOTSYM(kvmppc_rm_h_put_tce_indirect) - hcall_real_table
2375#else
2376	.long	0		/* 0x138 */
2377	.long	0		/* 0x13c */
2378#endif
2379	.long	0		/* 0x140 */
2380	.long	0		/* 0x144 */
2381	.long	0		/* 0x148 */
2382	.long	0		/* 0x14c */
2383	.long	0		/* 0x150 */
2384	.long	0		/* 0x154 */
2385	.long	0		/* 0x158 */
2386	.long	0		/* 0x15c */
2387	.long	0		/* 0x160 */
2388	.long	0		/* 0x164 */
2389	.long	0		/* 0x168 */
2390	.long	0		/* 0x16c */
2391	.long	0		/* 0x170 */
2392	.long	0		/* 0x174 */
2393	.long	0		/* 0x178 */
2394	.long	0		/* 0x17c */
2395	.long	0		/* 0x180 */
2396	.long	0		/* 0x184 */
2397	.long	0		/* 0x188 */
2398	.long	0		/* 0x18c */
2399	.long	0		/* 0x190 */
2400	.long	0		/* 0x194 */
2401	.long	0		/* 0x198 */
2402	.long	0		/* 0x19c */
2403	.long	0		/* 0x1a0 */
2404	.long	0		/* 0x1a4 */
2405	.long	0		/* 0x1a8 */
2406	.long	0		/* 0x1ac */
2407	.long	0		/* 0x1b0 */
2408	.long	0		/* 0x1b4 */
2409	.long	0		/* 0x1b8 */
2410	.long	0		/* 0x1bc */
2411	.long	0		/* 0x1c0 */
2412	.long	0		/* 0x1c4 */
2413	.long	0		/* 0x1c8 */
2414	.long	0		/* 0x1cc */
2415	.long	0		/* 0x1d0 */
2416	.long	0		/* 0x1d4 */
2417	.long	0		/* 0x1d8 */
2418	.long	0		/* 0x1dc */
2419	.long	0		/* 0x1e0 */
2420	.long	0		/* 0x1e4 */
2421	.long	0		/* 0x1e8 */
2422	.long	0		/* 0x1ec */
2423	.long	0		/* 0x1f0 */
2424	.long	0		/* 0x1f4 */
2425	.long	0		/* 0x1f8 */
2426	.long	0		/* 0x1fc */
2427	.long	0		/* 0x200 */
2428	.long	0		/* 0x204 */
2429	.long	0		/* 0x208 */
2430	.long	0		/* 0x20c */
2431	.long	0		/* 0x210 */
2432	.long	0		/* 0x214 */
2433	.long	0		/* 0x218 */
2434	.long	0		/* 0x21c */
2435	.long	0		/* 0x220 */
2436	.long	0		/* 0x224 */
2437	.long	0		/* 0x228 */
2438	.long	0		/* 0x22c */
2439	.long	0		/* 0x230 */
2440	.long	0		/* 0x234 */
2441	.long	0		/* 0x238 */
2442	.long	0		/* 0x23c */
2443	.long	0		/* 0x240 */
2444	.long	0		/* 0x244 */
2445	.long	0		/* 0x248 */
2446	.long	0		/* 0x24c */
2447	.long	0		/* 0x250 */
2448	.long	0		/* 0x254 */
2449	.long	0		/* 0x258 */
2450	.long	0		/* 0x25c */
2451	.long	0		/* 0x260 */
2452	.long	0		/* 0x264 */
2453	.long	0		/* 0x268 */
2454	.long	0		/* 0x26c */
2455	.long	0		/* 0x270 */
2456	.long	0		/* 0x274 */
2457	.long	0		/* 0x278 */
2458	.long	0		/* 0x27c */
2459	.long	0		/* 0x280 */
2460	.long	0		/* 0x284 */
2461	.long	0		/* 0x288 */
2462	.long	0		/* 0x28c */
2463	.long	0		/* 0x290 */
2464	.long	0		/* 0x294 */
2465	.long	0		/* 0x298 */
2466	.long	0		/* 0x29c */
2467	.long	0		/* 0x2a0 */
2468	.long	0		/* 0x2a4 */
2469	.long	0		/* 0x2a8 */
2470	.long	0		/* 0x2ac */
2471	.long	0		/* 0x2b0 */
2472	.long	0		/* 0x2b4 */
2473	.long	0		/* 0x2b8 */
2474	.long	0		/* 0x2bc */
2475	.long	0		/* 0x2c0 */
2476	.long	0		/* 0x2c4 */
2477	.long	0		/* 0x2c8 */
2478	.long	0		/* 0x2cc */
2479	.long	0		/* 0x2d0 */
2480	.long	0		/* 0x2d4 */
2481	.long	0		/* 0x2d8 */
2482	.long	0		/* 0x2dc */
2483	.long	0		/* 0x2e0 */
2484	.long	0		/* 0x2e4 */
2485	.long	0		/* 0x2e8 */
2486	.long	0		/* 0x2ec */
2487	.long	0		/* 0x2f0 */
2488	.long	0		/* 0x2f4 */
2489	.long	0		/* 0x2f8 */
2490#ifdef CONFIG_KVM_XICS
2491	.long	DOTSYM(kvmppc_rm_h_xirr_x) - hcall_real_table
2492#else
2493	.long	0		/* 0x2fc - H_XIRR_X*/
2494#endif
2495	.long	DOTSYM(kvmppc_h_random) - hcall_real_table
2496	.globl	hcall_real_table_end
2497hcall_real_table_end:
2498
2499_GLOBAL(kvmppc_h_set_xdabr)
2500EXPORT_SYMBOL_GPL(kvmppc_h_set_xdabr)
2501	andi.	r0, r5, DABRX_USER | DABRX_KERNEL
2502	beq	6f
2503	li	r0, DABRX_USER | DABRX_KERNEL | DABRX_BTI
2504	andc.	r0, r5, r0
2505	beq	3f
25066:	li	r3, H_PARAMETER
2507	blr
2508
2509_GLOBAL(kvmppc_h_set_dabr)
2510EXPORT_SYMBOL_GPL(kvmppc_h_set_dabr)
2511	li	r5, DABRX_USER | DABRX_KERNEL
25123:
2513BEGIN_FTR_SECTION
2514	b	2f
2515END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
2516	std	r4,VCPU_DABR(r3)
2517	stw	r5, VCPU_DABRX(r3)
2518	mtspr	SPRN_DABRX, r5
2519	/* Work around P7 bug where DABR can get corrupted on mtspr */
25201:	mtspr	SPRN_DABR,r4
2521	mfspr	r5, SPRN_DABR
2522	cmpd	r4, r5
2523	bne	1b
2524	isync
2525	li	r3,0
2526	blr
2527
25282:
2529	LOAD_REG_ADDR(r11, dawr_force_enable)
2530	lbz	r11, 0(r11)
2531	cmpdi	r11, 0
2532	bne	3f
2533	li	r3, H_HARDWARE
2534	blr
25353:
2536	/* Emulate H_SET_DABR/X on P8 for the sake of compat mode guests */
2537	rlwimi	r5, r4, 5, DAWRX_DR | DAWRX_DW
2538	rlwimi	r5, r4, 2, DAWRX_WT
2539	clrrdi	r4, r4, 3
2540	std	r4, VCPU_DAWR(r3)
2541	std	r5, VCPU_DAWRX(r3)
2542	/*
2543	 * If came in through the real mode hcall handler then it is necessary
2544	 * to write the registers since the return path won't. Otherwise it is
2545	 * sufficient to store then in the vcpu struct as they will be loaded
2546	 * next time the vcpu is run.
2547	 */
2548	mfmsr	r6
2549	andi.	r6, r6, MSR_DR		/* in real mode? */
2550	bne	4f
2551	mtspr	SPRN_DAWR, r4
2552	mtspr	SPRN_DAWRX, r5
25534:	li	r3, 0
2554	blr
2555
2556_GLOBAL(kvmppc_h_cede)		/* r3 = vcpu pointer, r11 = msr, r13 = paca */
2557	ori	r11,r11,MSR_EE
2558	std	r11,VCPU_MSR(r3)
2559	li	r0,1
2560	stb	r0,VCPU_CEDED(r3)
2561	sync			/* order setting ceded vs. testing prodded */
2562	lbz	r5,VCPU_PRODDED(r3)
2563	cmpwi	r5,0
2564	bne	kvm_cede_prodded
2565	li	r12,0		/* set trap to 0 to say hcall is handled */
2566	stw	r12,VCPU_TRAP(r3)
2567	li	r0,H_SUCCESS
2568	std	r0,VCPU_GPR(R3)(r3)
2569
2570	/*
2571	 * Set our bit in the bitmask of napping threads unless all the
2572	 * other threads are already napping, in which case we send this
2573	 * up to the host.
2574	 */
2575	ld	r5,HSTATE_KVM_VCORE(r13)
2576	lbz	r6,HSTATE_PTID(r13)
2577	lwz	r8,VCORE_ENTRY_EXIT(r5)
2578	clrldi	r8,r8,56
2579	li	r0,1
2580	sld	r0,r0,r6
2581	addi	r6,r5,VCORE_NAPPING_THREADS
258231:	lwarx	r4,0,r6
2583	or	r4,r4,r0
2584	cmpw	r4,r8
2585	beq	kvm_cede_exit
2586	stwcx.	r4,0,r6
2587	bne	31b
2588	/* order napping_threads update vs testing entry_exit_map */
2589	isync
2590	li	r0,NAPPING_CEDE
2591	stb	r0,HSTATE_NAPPING(r13)
2592	lwz	r7,VCORE_ENTRY_EXIT(r5)
2593	cmpwi	r7,0x100
2594	bge	33f		/* another thread already exiting */
2595
2596/*
2597 * Although not specifically required by the architecture, POWER7
2598 * preserves the following registers in nap mode, even if an SMT mode
2599 * switch occurs: SLB entries, PURR, SPURR, AMOR, UAMOR, AMR, SPRG0-3,
2600 * DAR, DSISR, DABR, DABRX, DSCR, PMCx, MMCRx, SIAR, SDAR.
2601 */
2602	/* Save non-volatile GPRs */
2603	std	r14, VCPU_GPR(R14)(r3)
2604	std	r15, VCPU_GPR(R15)(r3)
2605	std	r16, VCPU_GPR(R16)(r3)
2606	std	r17, VCPU_GPR(R17)(r3)
2607	std	r18, VCPU_GPR(R18)(r3)
2608	std	r19, VCPU_GPR(R19)(r3)
2609	std	r20, VCPU_GPR(R20)(r3)
2610	std	r21, VCPU_GPR(R21)(r3)
2611	std	r22, VCPU_GPR(R22)(r3)
2612	std	r23, VCPU_GPR(R23)(r3)
2613	std	r24, VCPU_GPR(R24)(r3)
2614	std	r25, VCPU_GPR(R25)(r3)
2615	std	r26, VCPU_GPR(R26)(r3)
2616	std	r27, VCPU_GPR(R27)(r3)
2617	std	r28, VCPU_GPR(R28)(r3)
2618	std	r29, VCPU_GPR(R29)(r3)
2619	std	r30, VCPU_GPR(R30)(r3)
2620	std	r31, VCPU_GPR(R31)(r3)
2621
2622	/* save FP state */
2623	bl	kvmppc_save_fp
2624
2625#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2626/*
2627 * Branch around the call if both CPU_FTR_TM and
2628 * CPU_FTR_P9_TM_HV_ASSIST are off.
2629 */
2630BEGIN_FTR_SECTION
2631	b	91f
2632END_FTR_SECTION(CPU_FTR_TM | CPU_FTR_P9_TM_HV_ASSIST, 0)
2633	/*
2634	 * NOTE THAT THIS TRASHES ALL NON-VOLATILE REGISTERS (but not CR)
2635	 */
2636	ld	r3, HSTATE_KVM_VCPU(r13)
2637	ld      r4, VCPU_MSR(r3)
2638	li	r5, 0			/* don't preserve non-vol regs */
2639	bl	kvmppc_save_tm_hv
2640	nop
264191:
2642#endif
2643
2644	/*
2645	 * Set DEC to the smaller of DEC and HDEC, so that we wake
2646	 * no later than the end of our timeslice (HDEC interrupts
2647	 * don't wake us from nap).
2648	 */
2649	mfspr	r3, SPRN_DEC
2650	mfspr	r4, SPRN_HDEC
2651	mftb	r5
2652BEGIN_FTR_SECTION
2653	/* On P9 check whether the guest has large decrementer mode enabled */
2654	ld	r6, HSTATE_KVM_VCORE(r13)
2655	ld	r6, VCORE_LPCR(r6)
2656	andis.	r6, r6, LPCR_LD@h
2657	bne	68f
2658END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
2659	extsw	r3, r3
266068:	EXTEND_HDEC(r4)
2661	cmpd	r3, r4
2662	ble	67f
2663	mtspr	SPRN_DEC, r4
266467:
2665	/* save expiry time of guest decrementer */
2666	add	r3, r3, r5
2667	ld	r4, HSTATE_KVM_VCPU(r13)
2668	ld	r5, HSTATE_KVM_VCORE(r13)
2669	ld	r6, VCORE_TB_OFFSET_APPL(r5)
2670	subf	r3, r6, r3	/* convert to host TB value */
2671	std	r3, VCPU_DEC_EXPIRES(r4)
2672
2673#ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
2674	ld	r4, HSTATE_KVM_VCPU(r13)
2675	addi	r3, r4, VCPU_TB_CEDE
2676	bl	kvmhv_accumulate_time
2677#endif
2678
2679	lis	r3, LPCR_PECEDP@h	/* Do wake on privileged doorbell */
2680
2681	/* Go back to host stack */
2682	ld	r1, HSTATE_HOST_R1(r13)
2683
2684	/*
2685	 * Take a nap until a decrementer or external or doobell interrupt
2686	 * occurs, with PECE1 and PECE0 set in LPCR.
2687	 * On POWER8, set PECEDH, and if we are ceding, also set PECEDP.
2688	 * Also clear the runlatch bit before napping.
2689	 */
2690kvm_do_nap:
2691	mfspr	r0, SPRN_CTRLF
2692	clrrdi	r0, r0, 1
2693	mtspr	SPRN_CTRLT, r0
2694
2695	li	r0,1
2696	stb	r0,HSTATE_HWTHREAD_REQ(r13)
2697	mfspr	r5,SPRN_LPCR
2698	ori	r5,r5,LPCR_PECE0 | LPCR_PECE1
2699BEGIN_FTR_SECTION
2700	ori	r5, r5, LPCR_PECEDH
2701	rlwimi	r5, r3, 0, LPCR_PECEDP
2702END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
2703
2704kvm_nap_sequence:		/* desired LPCR value in r5 */
2705BEGIN_FTR_SECTION
2706	/*
2707	 * PSSCR bits:	exit criterion = 1 (wakeup based on LPCR at sreset)
2708	 *		enable state loss = 1 (allow SMT mode switch)
2709	 *		requested level = 0 (just stop dispatching)
2710	 */
2711	lis	r3, (PSSCR_EC | PSSCR_ESL)@h
2712	/* Set LPCR_PECE_HVEE bit to enable wakeup by HV interrupts */
2713	li	r4, LPCR_PECE_HVEE@higher
2714	sldi	r4, r4, 32
2715	or	r5, r5, r4
2716FTR_SECTION_ELSE
2717	li	r3, PNV_THREAD_NAP
2718ALT_FTR_SECTION_END_IFSET(CPU_FTR_ARCH_300)
2719	mtspr	SPRN_LPCR,r5
2720	isync
2721
2722BEGIN_FTR_SECTION
2723	bl	isa300_idle_stop_mayloss
2724FTR_SECTION_ELSE
2725	bl	isa206_idle_insn_mayloss
2726ALT_FTR_SECTION_END_IFSET(CPU_FTR_ARCH_300)
2727
2728	mfspr	r0, SPRN_CTRLF
2729	ori	r0, r0, 1
2730	mtspr	SPRN_CTRLT, r0
2731
2732	mtspr	SPRN_SRR1, r3
2733
2734	li	r0, 0
2735	stb	r0, PACA_FTRACE_ENABLED(r13)
2736
2737	li	r0, KVM_HWTHREAD_IN_KVM
2738	stb	r0, HSTATE_HWTHREAD_STATE(r13)
2739
2740	lbz	r0, HSTATE_NAPPING(r13)
2741	cmpwi	r0, NAPPING_CEDE
2742	beq	kvm_end_cede
2743	cmpwi	r0, NAPPING_NOVCPU
2744	beq	kvm_novcpu_wakeup
2745	cmpwi	r0, NAPPING_UNSPLIT
2746	beq	kvm_unsplit_wakeup
2747	twi	31,0,0 /* Nap state must not be zero */
2748
274933:	mr	r4, r3
2750	li	r3, 0
2751	li	r12, 0
2752	b	34f
2753
2754kvm_end_cede:
2755	/* Woken by external or decrementer interrupt */
2756
2757	/* get vcpu pointer */
2758	ld	r4, HSTATE_KVM_VCPU(r13)
2759
2760#ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
2761	addi	r3, r4, VCPU_TB_RMINTR
2762	bl	kvmhv_accumulate_time
2763#endif
2764
2765#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
2766/*
2767 * Branch around the call if both CPU_FTR_TM and
2768 * CPU_FTR_P9_TM_HV_ASSIST are off.
2769 */
2770BEGIN_FTR_SECTION
2771	b	91f
2772END_FTR_SECTION(CPU_FTR_TM | CPU_FTR_P9_TM_HV_ASSIST, 0)
2773	/*
2774	 * NOTE THAT THIS TRASHES ALL NON-VOLATILE REGISTERS (but not CR)
2775	 */
2776	mr      r3, r4
2777	ld      r4, VCPU_MSR(r3)
2778	li	r5, 0			/* don't preserve non-vol regs */
2779	bl	kvmppc_restore_tm_hv
2780	nop
2781	ld	r4, HSTATE_KVM_VCPU(r13)
278291:
2783#endif
2784
2785	/* load up FP state */
2786	bl	kvmppc_load_fp
2787
2788	/* Restore guest decrementer */
2789	ld	r3, VCPU_DEC_EXPIRES(r4)
2790	ld	r5, HSTATE_KVM_VCORE(r13)
2791	ld	r6, VCORE_TB_OFFSET_APPL(r5)
2792	add	r3, r3, r6	/* convert host TB to guest TB value */
2793	mftb	r7
2794	subf	r3, r7, r3
2795	mtspr	SPRN_DEC, r3
2796
2797	/* Load NV GPRS */
2798	ld	r14, VCPU_GPR(R14)(r4)
2799	ld	r15, VCPU_GPR(R15)(r4)
2800	ld	r16, VCPU_GPR(R16)(r4)
2801	ld	r17, VCPU_GPR(R17)(r4)
2802	ld	r18, VCPU_GPR(R18)(r4)
2803	ld	r19, VCPU_GPR(R19)(r4)
2804	ld	r20, VCPU_GPR(R20)(r4)
2805	ld	r21, VCPU_GPR(R21)(r4)
2806	ld	r22, VCPU_GPR(R22)(r4)
2807	ld	r23, VCPU_GPR(R23)(r4)
2808	ld	r24, VCPU_GPR(R24)(r4)
2809	ld	r25, VCPU_GPR(R25)(r4)
2810	ld	r26, VCPU_GPR(R26)(r4)
2811	ld	r27, VCPU_GPR(R27)(r4)
2812	ld	r28, VCPU_GPR(R28)(r4)
2813	ld	r29, VCPU_GPR(R29)(r4)
2814	ld	r30, VCPU_GPR(R30)(r4)
2815	ld	r31, VCPU_GPR(R31)(r4)
2816
2817	/* Check the wake reason in SRR1 to see why we got here */
2818	bl	kvmppc_check_wake_reason
2819
2820	/*
2821	 * Restore volatile registers since we could have called a
2822	 * C routine in kvmppc_check_wake_reason
2823	 *	r4 = VCPU
2824	 * r3 tells us whether we need to return to host or not
2825	 * WARNING: it gets checked further down:
2826	 * should not modify r3 until this check is done.
2827	 */
2828	ld	r4, HSTATE_KVM_VCPU(r13)
2829
2830	/* clear our bit in vcore->napping_threads */
283134:	ld	r5,HSTATE_KVM_VCORE(r13)
2832	lbz	r7,HSTATE_PTID(r13)
2833	li	r0,1
2834	sld	r0,r0,r7
2835	addi	r6,r5,VCORE_NAPPING_THREADS
283632:	lwarx	r7,0,r6
2837	andc	r7,r7,r0
2838	stwcx.	r7,0,r6
2839	bne	32b
2840	li	r0,0
2841	stb	r0,HSTATE_NAPPING(r13)
2842
2843	/* See if the wake reason saved in r3 means we need to exit */
2844	stw	r12, VCPU_TRAP(r4)
2845	mr	r9, r4
2846	cmpdi	r3, 0
2847	bgt	guest_exit_cont
2848	b	maybe_reenter_guest
2849
2850	/* cede when already previously prodded case */
2851kvm_cede_prodded:
2852	li	r0,0
2853	stb	r0,VCPU_PRODDED(r3)
2854	sync			/* order testing prodded vs. clearing ceded */
2855	stb	r0,VCPU_CEDED(r3)
2856	li	r3,H_SUCCESS
2857	blr
2858
2859	/* we've ceded but we want to give control to the host */
2860kvm_cede_exit:
2861	ld	r9, HSTATE_KVM_VCPU(r13)
2862#ifdef CONFIG_KVM_XICS
2863	/* are we using XIVE with single escalation? */
2864	ld	r10, VCPU_XIVE_ESC_VADDR(r9)
2865	cmpdi	r10, 0
2866	beq	3f
2867	li	r6, XIVE_ESB_SET_PQ_00
2868	/*
2869	 * If we still have a pending escalation, abort the cede,
2870	 * and we must set PQ to 10 rather than 00 so that we don't
2871	 * potentially end up with two entries for the escalation
2872	 * interrupt in the XIVE interrupt queue.  In that case
2873	 * we also don't want to set xive_esc_on to 1 here in
2874	 * case we race with xive_esc_irq().
2875	 */
2876	lbz	r5, VCPU_XIVE_ESC_ON(r9)
2877	cmpwi	r5, 0
2878	beq	4f
2879	li	r0, 0
2880	stb	r0, VCPU_CEDED(r9)
2881	li	r6, XIVE_ESB_SET_PQ_10
2882	b	5f
28834:	li	r0, 1
2884	stb	r0, VCPU_XIVE_ESC_ON(r9)
2885	/* make sure store to xive_esc_on is seen before xive_esc_irq runs */
2886	sync
28875:	/* Enable XIVE escalation */
2888	mfmsr	r0
2889	andi.	r0, r0, MSR_DR		/* in real mode? */
2890	beq	1f
2891	ldx	r0, r10, r6
2892	b	2f
28931:	ld	r10, VCPU_XIVE_ESC_RADDR(r9)
2894	ldcix	r0, r10, r6
28952:	sync
2896#endif /* CONFIG_KVM_XICS */
28973:	b	guest_exit_cont
2898
2899	/* Try to do machine check recovery in real mode */
2900machine_check_realmode:
2901	mr	r3, r9		/* get vcpu pointer */
2902	bl	kvmppc_realmode_machine_check
2903	nop
2904	/* all machine checks go to virtual mode for further handling */
2905	ld	r9, HSTATE_KVM_VCPU(r13)
2906	li	r12, BOOK3S_INTERRUPT_MACHINE_CHECK
2907	b	guest_exit_cont
2908
2909/*
2910 * Call C code to handle a HMI in real mode.
2911 * Only the primary thread does the call, secondary threads are handled
2912 * by calling hmi_exception_realmode() after kvmppc_hv_entry returns.
2913 * r9 points to the vcpu on entry
2914 */
2915hmi_realmode:
2916	lbz	r0, HSTATE_PTID(r13)
2917	cmpwi	r0, 0
2918	bne	guest_exit_cont
2919	bl	kvmppc_realmode_hmi_handler
2920	ld	r9, HSTATE_KVM_VCPU(r13)
2921	li	r12, BOOK3S_INTERRUPT_HMI
2922	b	guest_exit_cont
2923
2924/*
2925 * Check the reason we woke from nap, and take appropriate action.
2926 * Returns (in r3):
2927 *	0 if nothing needs to be done
2928 *	1 if something happened that needs to be handled by the host
2929 *	-1 if there was a guest wakeup (IPI or msgsnd)
2930 *	-2 if we handled a PCI passthrough interrupt (returned by
2931 *		kvmppc_read_intr only)
2932 *
2933 * Also sets r12 to the interrupt vector for any interrupt that needs
2934 * to be handled now by the host (0x500 for external interrupt), or zero.
2935 * Modifies all volatile registers (since it may call a C function).
2936 * This routine calls kvmppc_read_intr, a C function, if an external
2937 * interrupt is pending.
2938 */
2939kvmppc_check_wake_reason:
2940	mfspr	r6, SPRN_SRR1
2941BEGIN_FTR_SECTION
2942	rlwinm	r6, r6, 45-31, 0xf	/* extract wake reason field (P8) */
2943FTR_SECTION_ELSE
2944	rlwinm	r6, r6, 45-31, 0xe	/* P7 wake reason field is 3 bits */
2945ALT_FTR_SECTION_END_IFSET(CPU_FTR_ARCH_207S)
2946	cmpwi	r6, 8			/* was it an external interrupt? */
2947	beq	7f			/* if so, see what it was */
2948	li	r3, 0
2949	li	r12, 0
2950	cmpwi	r6, 6			/* was it the decrementer? */
2951	beq	0f
2952BEGIN_FTR_SECTION
2953	cmpwi	r6, 5			/* privileged doorbell? */
2954	beq	0f
2955	cmpwi	r6, 3			/* hypervisor doorbell? */
2956	beq	3f
2957END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
2958	cmpwi	r6, 0xa			/* Hypervisor maintenance ? */
2959	beq	4f
2960	li	r3, 1			/* anything else, return 1 */
29610:	blr
2962
2963	/* hypervisor doorbell */
29643:	li	r12, BOOK3S_INTERRUPT_H_DOORBELL
2965
2966	/*
2967	 * Clear the doorbell as we will invoke the handler
2968	 * explicitly in the guest exit path.
2969	 */
2970	lis	r6, (PPC_DBELL_SERVER << (63-36))@h
2971	PPC_MSGCLR(6)
2972	/* see if it's a host IPI */
2973	li	r3, 1
2974BEGIN_FTR_SECTION
2975	PPC_MSGSYNC
2976	lwsync
2977END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
2978	lbz	r0, HSTATE_HOST_IPI(r13)
2979	cmpwi	r0, 0
2980	bnelr
2981	/* if not, return -1 */
2982	li	r3, -1
2983	blr
2984
2985	/* Woken up due to Hypervisor maintenance interrupt */
29864:	li	r12, BOOK3S_INTERRUPT_HMI
2987	li	r3, 1
2988	blr
2989
2990	/* external interrupt - create a stack frame so we can call C */
29917:	mflr	r0
2992	std	r0, PPC_LR_STKOFF(r1)
2993	stdu	r1, -PPC_MIN_STKFRM(r1)
2994	bl	kvmppc_read_intr
2995	nop
2996	li	r12, BOOK3S_INTERRUPT_EXTERNAL
2997	cmpdi	r3, 1
2998	ble	1f
2999
3000	/*
3001	 * Return code of 2 means PCI passthrough interrupt, but
3002	 * we need to return back to host to complete handling the
3003	 * interrupt. Trap reason is expected in r12 by guest
3004	 * exit code.
3005	 */
3006	li	r12, BOOK3S_INTERRUPT_HV_RM_HARD
30071:
3008	ld	r0, PPC_MIN_STKFRM+PPC_LR_STKOFF(r1)
3009	addi	r1, r1, PPC_MIN_STKFRM
3010	mtlr	r0
3011	blr
3012
3013/*
3014 * Save away FP, VMX and VSX registers.
3015 * r3 = vcpu pointer
3016 * N.B. r30 and r31 are volatile across this function,
3017 * thus it is not callable from C.
3018 */
3019kvmppc_save_fp:
3020	mflr	r30
3021	mr	r31,r3
3022	mfmsr	r5
3023	ori	r8,r5,MSR_FP
3024#ifdef CONFIG_ALTIVEC
3025BEGIN_FTR_SECTION
3026	oris	r8,r8,MSR_VEC@h
3027END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
3028#endif
3029#ifdef CONFIG_VSX
3030BEGIN_FTR_SECTION
3031	oris	r8,r8,MSR_VSX@h
3032END_FTR_SECTION_IFSET(CPU_FTR_VSX)
3033#endif
3034	mtmsrd	r8
3035	addi	r3,r3,VCPU_FPRS
3036	bl	store_fp_state
3037#ifdef CONFIG_ALTIVEC
3038BEGIN_FTR_SECTION
3039	addi	r3,r31,VCPU_VRS
3040	bl	store_vr_state
3041END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
3042#endif
3043	mfspr	r6,SPRN_VRSAVE
3044	stw	r6,VCPU_VRSAVE(r31)
3045	mtlr	r30
3046	blr
3047
3048/*
3049 * Load up FP, VMX and VSX registers
3050 * r4 = vcpu pointer
3051 * N.B. r30 and r31 are volatile across this function,
3052 * thus it is not callable from C.
3053 */
3054kvmppc_load_fp:
3055	mflr	r30
3056	mr	r31,r4
3057	mfmsr	r9
3058	ori	r8,r9,MSR_FP
3059#ifdef CONFIG_ALTIVEC
3060BEGIN_FTR_SECTION
3061	oris	r8,r8,MSR_VEC@h
3062END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
3063#endif
3064#ifdef CONFIG_VSX
3065BEGIN_FTR_SECTION
3066	oris	r8,r8,MSR_VSX@h
3067END_FTR_SECTION_IFSET(CPU_FTR_VSX)
3068#endif
3069	mtmsrd	r8
3070	addi	r3,r4,VCPU_FPRS
3071	bl	load_fp_state
3072#ifdef CONFIG_ALTIVEC
3073BEGIN_FTR_SECTION
3074	addi	r3,r31,VCPU_VRS
3075	bl	load_vr_state
3076END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
3077#endif
3078	lwz	r7,VCPU_VRSAVE(r31)
3079	mtspr	SPRN_VRSAVE,r7
3080	mtlr	r30
3081	mr	r4,r31
3082	blr
3083
3084#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
3085/*
3086 * Save transactional state and TM-related registers.
3087 * Called with r3 pointing to the vcpu struct and r4 containing
3088 * the guest MSR value.
3089 * r5 is non-zero iff non-volatile register state needs to be maintained.
3090 * If r5 == 0, this can modify all checkpointed registers, but
3091 * restores r1 and r2 before exit.
3092 */
3093_GLOBAL_TOC(kvmppc_save_tm_hv)
3094EXPORT_SYMBOL_GPL(kvmppc_save_tm_hv)
3095	/* See if we need to handle fake suspend mode */
3096BEGIN_FTR_SECTION
3097	b	__kvmppc_save_tm
3098END_FTR_SECTION_IFCLR(CPU_FTR_P9_TM_HV_ASSIST)
3099
3100	lbz	r0, HSTATE_FAKE_SUSPEND(r13) /* Were we fake suspended? */
3101	cmpwi	r0, 0
3102	beq	__kvmppc_save_tm
3103
3104	/* The following code handles the fake_suspend = 1 case */
3105	mflr	r0
3106	std	r0, PPC_LR_STKOFF(r1)
3107	stdu	r1, -PPC_MIN_STKFRM(r1)
3108
3109	/* Turn on TM. */
3110	mfmsr	r8
3111	li	r0, 1
3112	rldimi	r8, r0, MSR_TM_LG, 63-MSR_TM_LG
3113	mtmsrd	r8
3114
3115	rldicl. r8, r8, 64 - MSR_TS_S_LG, 62 /* Did we actually hrfid? */
3116	beq	4f
3117BEGIN_FTR_SECTION
3118	bl	pnv_power9_force_smt4_catch
3119END_FTR_SECTION_IFSET(CPU_FTR_P9_TM_XER_SO_BUG)
3120	nop
3121
3122	/* We have to treclaim here because that's the only way to do S->N */
3123	li	r3, TM_CAUSE_KVM_RESCHED
3124	TRECLAIM(R3)
3125
3126	/*
3127	 * We were in fake suspend, so we are not going to save the
3128	 * register state as the guest checkpointed state (since
3129	 * we already have it), therefore we can now use any volatile GPR.
3130	 * In fact treclaim in fake suspend state doesn't modify
3131	 * any registers.
3132	 */
3133
3134BEGIN_FTR_SECTION
3135	bl	pnv_power9_force_smt4_release
3136END_FTR_SECTION_IFSET(CPU_FTR_P9_TM_XER_SO_BUG)
3137	nop
3138
31394:
3140	mfspr	r3, SPRN_PSSCR
3141	/* PSSCR_FAKE_SUSPEND is a write-only bit, but clear it anyway */
3142	li	r0, PSSCR_FAKE_SUSPEND
3143	andc	r3, r3, r0
3144	mtspr	SPRN_PSSCR, r3
3145
3146	/* Don't save TEXASR, use value from last exit in real suspend state */
3147	ld	r9, HSTATE_KVM_VCPU(r13)
3148	mfspr	r5, SPRN_TFHAR
3149	mfspr	r6, SPRN_TFIAR
3150	std	r5, VCPU_TFHAR(r9)
3151	std	r6, VCPU_TFIAR(r9)
3152
3153	addi	r1, r1, PPC_MIN_STKFRM
3154	ld	r0, PPC_LR_STKOFF(r1)
3155	mtlr	r0
3156	blr
3157
3158/*
3159 * Restore transactional state and TM-related registers.
3160 * Called with r3 pointing to the vcpu struct
3161 * and r4 containing the guest MSR value.
3162 * r5 is non-zero iff non-volatile register state needs to be maintained.
3163 * This potentially modifies all checkpointed registers.
3164 * It restores r1 and r2 from the PACA.
3165 */
3166_GLOBAL_TOC(kvmppc_restore_tm_hv)
3167EXPORT_SYMBOL_GPL(kvmppc_restore_tm_hv)
3168	/*
3169	 * If we are doing TM emulation for the guest on a POWER9 DD2,
3170	 * then we don't actually do a trechkpt -- we either set up
3171	 * fake-suspend mode, or emulate a TM rollback.
3172	 */
3173BEGIN_FTR_SECTION
3174	b	__kvmppc_restore_tm
3175END_FTR_SECTION_IFCLR(CPU_FTR_P9_TM_HV_ASSIST)
3176	mflr	r0
3177	std	r0, PPC_LR_STKOFF(r1)
3178
3179	li	r0, 0
3180	stb	r0, HSTATE_FAKE_SUSPEND(r13)
3181
3182	/* Turn on TM so we can restore TM SPRs */
3183	mfmsr	r5
3184	li	r0, 1
3185	rldimi	r5, r0, MSR_TM_LG, 63-MSR_TM_LG
3186	mtmsrd	r5
3187
3188	/*
3189	 * The user may change these outside of a transaction, so they must
3190	 * always be context switched.
3191	 */
3192	ld	r5, VCPU_TFHAR(r3)
3193	ld	r6, VCPU_TFIAR(r3)
3194	ld	r7, VCPU_TEXASR(r3)
3195	mtspr	SPRN_TFHAR, r5
3196	mtspr	SPRN_TFIAR, r6
3197	mtspr	SPRN_TEXASR, r7
3198
3199	rldicl. r5, r4, 64 - MSR_TS_S_LG, 62
3200	beqlr		/* TM not active in guest */
3201
3202	/* Make sure the failure summary is set */
3203	oris	r7, r7, (TEXASR_FS)@h
3204	mtspr	SPRN_TEXASR, r7
3205
3206	cmpwi	r5, 1		/* check for suspended state */
3207	bgt	10f
3208	stb	r5, HSTATE_FAKE_SUSPEND(r13)
3209	b	9f		/* and return */
321010:	stdu	r1, -PPC_MIN_STKFRM(r1)
3211	/* guest is in transactional state, so simulate rollback */
3212	bl	kvmhv_emulate_tm_rollback
3213	nop
3214	addi	r1, r1, PPC_MIN_STKFRM
32159:	ld	r0, PPC_LR_STKOFF(r1)
3216	mtlr	r0
3217	blr
3218#endif /* CONFIG_PPC_TRANSACTIONAL_MEM */
3219
3220/*
3221 * We come here if we get any exception or interrupt while we are
3222 * executing host real mode code while in guest MMU context.
3223 * r12 is (CR << 32) | vector
3224 * r13 points to our PACA
3225 * r12 is saved in HSTATE_SCRATCH0(r13)
3226 * ctr is saved in HSTATE_SCRATCH1(r13) if RELOCATABLE
3227 * r9 is saved in HSTATE_SCRATCH2(r13)
3228 * r13 is saved in HSPRG1
3229 * cfar is saved in HSTATE_CFAR(r13)
3230 * ppr is saved in HSTATE_PPR(r13)
3231 */
3232kvmppc_bad_host_intr:
3233	/*
3234	 * Switch to the emergency stack, but start half-way down in
3235	 * case we were already on it.
3236	 */
3237	mr	r9, r1
3238	std	r1, PACAR1(r13)
3239	ld	r1, PACAEMERGSP(r13)
3240	subi	r1, r1, THREAD_SIZE/2 + INT_FRAME_SIZE
3241	std	r9, 0(r1)
3242	std	r0, GPR0(r1)
3243	std	r9, GPR1(r1)
3244	std	r2, GPR2(r1)
3245	SAVE_4GPRS(3, r1)
3246	SAVE_2GPRS(7, r1)
3247	srdi	r0, r12, 32
3248	clrldi	r12, r12, 32
3249	std	r0, _CCR(r1)
3250	std	r12, _TRAP(r1)
3251	andi.	r0, r12, 2
3252	beq	1f
3253	mfspr	r3, SPRN_HSRR0
3254	mfspr	r4, SPRN_HSRR1
3255	mfspr	r5, SPRN_HDAR
3256	mfspr	r6, SPRN_HDSISR
3257	b	2f
32581:	mfspr	r3, SPRN_SRR0
3259	mfspr	r4, SPRN_SRR1
3260	mfspr	r5, SPRN_DAR
3261	mfspr	r6, SPRN_DSISR
32622:	std	r3, _NIP(r1)
3263	std	r4, _MSR(r1)
3264	std	r5, _DAR(r1)
3265	std	r6, _DSISR(r1)
3266	ld	r9, HSTATE_SCRATCH2(r13)
3267	ld	r12, HSTATE_SCRATCH0(r13)
3268	GET_SCRATCH0(r0)
3269	SAVE_4GPRS(9, r1)
3270	std	r0, GPR13(r1)
3271	SAVE_NVGPRS(r1)
3272	ld	r5, HSTATE_CFAR(r13)
3273	std	r5, ORIG_GPR3(r1)
3274	mflr	r3
3275#ifdef CONFIG_RELOCATABLE
3276	ld	r4, HSTATE_SCRATCH1(r13)
3277#else
3278	mfctr	r4
3279#endif
3280	mfxer	r5
3281	lbz	r6, PACAIRQSOFTMASK(r13)
3282	std	r3, _LINK(r1)
3283	std	r4, _CTR(r1)
3284	std	r5, _XER(r1)
3285	std	r6, SOFTE(r1)
3286	ld	r2, PACATOC(r13)
3287	LOAD_REG_IMMEDIATE(3, 0x7265677368657265)
3288	std	r3, STACK_FRAME_OVERHEAD-16(r1)
3289
3290	/*
3291	 * On POWER9 do a minimal restore of the MMU and call C code,
3292	 * which will print a message and panic.
3293	 * XXX On POWER7 and POWER8, we just spin here since we don't
3294	 * know what the other threads are doing (and we don't want to
3295	 * coordinate with them) - but at least we now have register state
3296	 * in memory that we might be able to look at from another CPU.
3297	 */
3298BEGIN_FTR_SECTION
3299	b	.
3300END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_300)
3301	ld	r9, HSTATE_KVM_VCPU(r13)
3302	ld	r10, VCPU_KVM(r9)
3303
3304	li	r0, 0
3305	mtspr	SPRN_AMR, r0
3306	mtspr	SPRN_IAMR, r0
3307	mtspr	SPRN_CIABR, r0
3308	mtspr	SPRN_DAWRX, r0
3309
3310BEGIN_MMU_FTR_SECTION
3311	b	4f
3312END_MMU_FTR_SECTION_IFSET(MMU_FTR_TYPE_RADIX)
3313
3314	slbmte	r0, r0
3315	slbia
3316	ptesync
3317	ld	r8, PACA_SLBSHADOWPTR(r13)
3318	.rept	SLB_NUM_BOLTED
3319	li	r3, SLBSHADOW_SAVEAREA
3320	LDX_BE	r5, r8, r3
3321	addi	r3, r3, 8
3322	LDX_BE	r6, r8, r3
3323	andis.	r7, r5, SLB_ESID_V@h
3324	beq	3f
3325	slbmte	r6, r5
33263:	addi	r8, r8, 16
3327	.endr
3328
33294:	lwz	r7, KVM_HOST_LPID(r10)
3330	mtspr	SPRN_LPID, r7
3331	mtspr	SPRN_PID, r0
3332	ld	r8, KVM_HOST_LPCR(r10)
3333	mtspr	SPRN_LPCR, r8
3334	isync
3335	li	r0, KVM_GUEST_MODE_NONE
3336	stb	r0, HSTATE_IN_GUEST(r13)
3337
3338	/*
3339	 * Turn on the MMU and jump to C code
3340	 */
3341	bcl	20, 31, .+4
33425:	mflr	r3
3343	addi	r3, r3, 9f - 5b
3344	li	r4, -1
3345	rldimi	r3, r4, 62, 0	/* ensure 0xc000000000000000 bits are set */
3346	ld	r4, PACAKMSR(r13)
3347	mtspr	SPRN_SRR0, r3
3348	mtspr	SPRN_SRR1, r4
3349	RFI_TO_KERNEL
33509:	addi	r3, r1, STACK_FRAME_OVERHEAD
3351	bl	kvmppc_bad_interrupt
3352	b	9b
3353
3354/*
3355 * This mimics the MSR transition on IRQ delivery.  The new guest MSR is taken
3356 * from VCPU_INTR_MSR and is modified based on the required TM state changes.
3357 *   r11 has the guest MSR value (in/out)
3358 *   r9 has a vcpu pointer (in)
3359 *   r0 is used as a scratch register
3360 */
3361kvmppc_msr_interrupt:
3362	rldicl	r0, r11, 64 - MSR_TS_S_LG, 62
3363	cmpwi	r0, 2 /* Check if we are in transactional state..  */
3364	ld	r11, VCPU_INTR_MSR(r9)
3365	bne	1f
3366	/* ... if transactional, change to suspended */
3367	li	r0, 1
33681:	rldimi	r11, r0, MSR_TS_S_LG, 63 - MSR_TS_T_LG
3369	blr
3370
3371/*
3372 * Load up guest PMU state.  R3 points to the vcpu struct.
3373 */
3374_GLOBAL(kvmhv_load_guest_pmu)
3375EXPORT_SYMBOL_GPL(kvmhv_load_guest_pmu)
3376	mr	r4, r3
3377	mflr	r0
3378	li	r3, 1
3379	sldi	r3, r3, 31		/* MMCR0_FC (freeze counters) bit */
3380	mtspr	SPRN_MMCR0, r3		/* freeze all counters, disable ints */
3381	isync
3382BEGIN_FTR_SECTION
3383	ld	r3, VCPU_MMCR(r4)
3384	andi.	r5, r3, MMCR0_PMAO_SYNC | MMCR0_PMAO
3385	cmpwi	r5, MMCR0_PMAO
3386	beql	kvmppc_fix_pmao
3387END_FTR_SECTION_IFSET(CPU_FTR_PMAO_BUG)
3388	lwz	r3, VCPU_PMC(r4)	/* always load up guest PMU registers */
3389	lwz	r5, VCPU_PMC + 4(r4)	/* to prevent information leak */
3390	lwz	r6, VCPU_PMC + 8(r4)
3391	lwz	r7, VCPU_PMC + 12(r4)
3392	lwz	r8, VCPU_PMC + 16(r4)
3393	lwz	r9, VCPU_PMC + 20(r4)
3394	mtspr	SPRN_PMC1, r3
3395	mtspr	SPRN_PMC2, r5
3396	mtspr	SPRN_PMC3, r6
3397	mtspr	SPRN_PMC4, r7
3398	mtspr	SPRN_PMC5, r8
3399	mtspr	SPRN_PMC6, r9
3400	ld	r3, VCPU_MMCR(r4)
3401	ld	r5, VCPU_MMCR + 8(r4)
3402	ld	r6, VCPU_MMCR + 16(r4)
3403	ld	r7, VCPU_SIAR(r4)
3404	ld	r8, VCPU_SDAR(r4)
3405	mtspr	SPRN_MMCR1, r5
3406	mtspr	SPRN_MMCRA, r6
3407	mtspr	SPRN_SIAR, r7
3408	mtspr	SPRN_SDAR, r8
3409BEGIN_FTR_SECTION
3410	ld	r5, VCPU_MMCR + 24(r4)
3411	ld	r6, VCPU_SIER(r4)
3412	mtspr	SPRN_MMCR2, r5
3413	mtspr	SPRN_SIER, r6
3414BEGIN_FTR_SECTION_NESTED(96)
3415	lwz	r7, VCPU_PMC + 24(r4)
3416	lwz	r8, VCPU_PMC + 28(r4)
3417	ld	r9, VCPU_MMCR + 32(r4)
3418	mtspr	SPRN_SPMC1, r7
3419	mtspr	SPRN_SPMC2, r8
3420	mtspr	SPRN_MMCRS, r9
3421END_FTR_SECTION_NESTED(CPU_FTR_ARCH_300, 0, 96)
3422END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
3423	mtspr	SPRN_MMCR0, r3
3424	isync
3425	mtlr	r0
3426	blr
3427
3428/*
3429 * Reload host PMU state saved in the PACA by kvmhv_save_host_pmu.
3430 */
3431_GLOBAL(kvmhv_load_host_pmu)
3432EXPORT_SYMBOL_GPL(kvmhv_load_host_pmu)
3433	mflr	r0
3434	lbz	r4, PACA_PMCINUSE(r13) /* is the host using the PMU? */
3435	cmpwi	r4, 0
3436	beq	23f			/* skip if not */
3437BEGIN_FTR_SECTION
3438	ld	r3, HSTATE_MMCR0(r13)
3439	andi.	r4, r3, MMCR0_PMAO_SYNC | MMCR0_PMAO
3440	cmpwi	r4, MMCR0_PMAO
3441	beql	kvmppc_fix_pmao
3442END_FTR_SECTION_IFSET(CPU_FTR_PMAO_BUG)
3443	lwz	r3, HSTATE_PMC1(r13)
3444	lwz	r4, HSTATE_PMC2(r13)
3445	lwz	r5, HSTATE_PMC3(r13)
3446	lwz	r6, HSTATE_PMC4(r13)
3447	lwz	r8, HSTATE_PMC5(r13)
3448	lwz	r9, HSTATE_PMC6(r13)
3449	mtspr	SPRN_PMC1, r3
3450	mtspr	SPRN_PMC2, r4
3451	mtspr	SPRN_PMC3, r5
3452	mtspr	SPRN_PMC4, r6
3453	mtspr	SPRN_PMC5, r8
3454	mtspr	SPRN_PMC6, r9
3455	ld	r3, HSTATE_MMCR0(r13)
3456	ld	r4, HSTATE_MMCR1(r13)
3457	ld	r5, HSTATE_MMCRA(r13)
3458	ld	r6, HSTATE_SIAR(r13)
3459	ld	r7, HSTATE_SDAR(r13)
3460	mtspr	SPRN_MMCR1, r4
3461	mtspr	SPRN_MMCRA, r5
3462	mtspr	SPRN_SIAR, r6
3463	mtspr	SPRN_SDAR, r7
3464BEGIN_FTR_SECTION
3465	ld	r8, HSTATE_MMCR2(r13)
3466	ld	r9, HSTATE_SIER(r13)
3467	mtspr	SPRN_MMCR2, r8
3468	mtspr	SPRN_SIER, r9
3469END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
3470	mtspr	SPRN_MMCR0, r3
3471	isync
3472	mtlr	r0
347323:	blr
3474
3475/*
3476 * Save guest PMU state into the vcpu struct.
3477 * r3 = vcpu, r4 = full save flag (PMU in use flag set in VPA)
3478 */
3479_GLOBAL(kvmhv_save_guest_pmu)
3480EXPORT_SYMBOL_GPL(kvmhv_save_guest_pmu)
3481	mr	r9, r3
3482	mr	r8, r4
3483BEGIN_FTR_SECTION
3484	/*
3485	 * POWER8 seems to have a hardware bug where setting
3486	 * MMCR0[PMAE] along with MMCR0[PMC1CE] and/or MMCR0[PMCjCE]
3487	 * when some counters are already negative doesn't seem
3488	 * to cause a performance monitor alert (and hence interrupt).
3489	 * The effect of this is that when saving the PMU state,
3490	 * if there is no PMU alert pending when we read MMCR0
3491	 * before freezing the counters, but one becomes pending
3492	 * before we read the counters, we lose it.
3493	 * To work around this, we need a way to freeze the counters
3494	 * before reading MMCR0.  Normally, freezing the counters
3495	 * is done by writing MMCR0 (to set MMCR0[FC]) which
3496	 * unavoidably writes MMCR0[PMA0] as well.  On POWER8,
3497	 * we can also freeze the counters using MMCR2, by writing
3498	 * 1s to all the counter freeze condition bits (there are
3499	 * 9 bits each for 6 counters).
3500	 */
3501	li	r3, -1			/* set all freeze bits */
3502	clrrdi	r3, r3, 10
3503	mfspr	r10, SPRN_MMCR2
3504	mtspr	SPRN_MMCR2, r3
3505	isync
3506END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
3507	li	r3, 1
3508	sldi	r3, r3, 31		/* MMCR0_FC (freeze counters) bit */
3509	mfspr	r4, SPRN_MMCR0		/* save MMCR0 */
3510	mtspr	SPRN_MMCR0, r3		/* freeze all counters, disable ints */
3511	mfspr	r6, SPRN_MMCRA
3512	/* Clear MMCRA in order to disable SDAR updates */
3513	li	r7, 0
3514	mtspr	SPRN_MMCRA, r7
3515	isync
3516	cmpwi	r8, 0			/* did they ask for PMU stuff to be saved? */
3517	bne	21f
3518	std	r3, VCPU_MMCR(r9)	/* if not, set saved MMCR0 to FC */
3519	b	22f
352021:	mfspr	r5, SPRN_MMCR1
3521	mfspr	r7, SPRN_SIAR
3522	mfspr	r8, SPRN_SDAR
3523	std	r4, VCPU_MMCR(r9)
3524	std	r5, VCPU_MMCR + 8(r9)
3525	std	r6, VCPU_MMCR + 16(r9)
3526BEGIN_FTR_SECTION
3527	std	r10, VCPU_MMCR + 24(r9)
3528END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
3529	std	r7, VCPU_SIAR(r9)
3530	std	r8, VCPU_SDAR(r9)
3531	mfspr	r3, SPRN_PMC1
3532	mfspr	r4, SPRN_PMC2
3533	mfspr	r5, SPRN_PMC3
3534	mfspr	r6, SPRN_PMC4
3535	mfspr	r7, SPRN_PMC5
3536	mfspr	r8, SPRN_PMC6
3537	stw	r3, VCPU_PMC(r9)
3538	stw	r4, VCPU_PMC + 4(r9)
3539	stw	r5, VCPU_PMC + 8(r9)
3540	stw	r6, VCPU_PMC + 12(r9)
3541	stw	r7, VCPU_PMC + 16(r9)
3542	stw	r8, VCPU_PMC + 20(r9)
3543BEGIN_FTR_SECTION
3544	mfspr	r5, SPRN_SIER
3545	std	r5, VCPU_SIER(r9)
3546BEGIN_FTR_SECTION_NESTED(96)
3547	mfspr	r6, SPRN_SPMC1
3548	mfspr	r7, SPRN_SPMC2
3549	mfspr	r8, SPRN_MMCRS
3550	stw	r6, VCPU_PMC + 24(r9)
3551	stw	r7, VCPU_PMC + 28(r9)
3552	std	r8, VCPU_MMCR + 32(r9)
3553	lis	r4, 0x8000
3554	mtspr	SPRN_MMCRS, r4
3555END_FTR_SECTION_NESTED(CPU_FTR_ARCH_300, 0, 96)
3556END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
355722:	blr
3558
3559/*
3560 * This works around a hardware bug on POWER8E processors, where
3561 * writing a 1 to the MMCR0[PMAO] bit doesn't generate a
3562 * performance monitor interrupt.  Instead, when we need to have
3563 * an interrupt pending, we have to arrange for a counter to overflow.
3564 */
3565kvmppc_fix_pmao:
3566	li	r3, 0
3567	mtspr	SPRN_MMCR2, r3
3568	lis	r3, (MMCR0_PMXE | MMCR0_FCECE)@h
3569	ori	r3, r3, MMCR0_PMCjCE | MMCR0_C56RUN
3570	mtspr	SPRN_MMCR0, r3
3571	lis	r3, 0x7fff
3572	ori	r3, r3, 0xffff
3573	mtspr	SPRN_PMC6, r3
3574	isync
3575	blr
3576
3577#ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
3578/*
3579 * Start timing an activity
3580 * r3 = pointer to time accumulation struct, r4 = vcpu
3581 */
3582kvmhv_start_timing:
3583	ld	r5, HSTATE_KVM_VCORE(r13)
3584	ld	r6, VCORE_TB_OFFSET_APPL(r5)
3585	mftb	r5
3586	subf	r5, r6, r5	/* subtract current timebase offset */
3587	std	r3, VCPU_CUR_ACTIVITY(r4)
3588	std	r5, VCPU_ACTIVITY_START(r4)
3589	blr
3590
3591/*
3592 * Accumulate time to one activity and start another.
3593 * r3 = pointer to new time accumulation struct, r4 = vcpu
3594 */
3595kvmhv_accumulate_time:
3596	ld	r5, HSTATE_KVM_VCORE(r13)
3597	ld	r8, VCORE_TB_OFFSET_APPL(r5)
3598	ld	r5, VCPU_CUR_ACTIVITY(r4)
3599	ld	r6, VCPU_ACTIVITY_START(r4)
3600	std	r3, VCPU_CUR_ACTIVITY(r4)
3601	mftb	r7
3602	subf	r7, r8, r7	/* subtract current timebase offset */
3603	std	r7, VCPU_ACTIVITY_START(r4)
3604	cmpdi	r5, 0
3605	beqlr
3606	subf	r3, r6, r7
3607	ld	r8, TAS_SEQCOUNT(r5)
3608	cmpdi	r8, 0
3609	addi	r8, r8, 1
3610	std	r8, TAS_SEQCOUNT(r5)
3611	lwsync
3612	ld	r7, TAS_TOTAL(r5)
3613	add	r7, r7, r3
3614	std	r7, TAS_TOTAL(r5)
3615	ld	r6, TAS_MIN(r5)
3616	ld	r7, TAS_MAX(r5)
3617	beq	3f
3618	cmpd	r3, r6
3619	bge	1f
36203:	std	r3, TAS_MIN(r5)
36211:	cmpd	r3, r7
3622	ble	2f
3623	std	r3, TAS_MAX(r5)
36242:	lwsync
3625	addi	r8, r8, 1
3626	std	r8, TAS_SEQCOUNT(r5)
3627	blr
3628#endif
3629