2 * This program is free software; you can redistribute it and/or modify
3 * it under the terms of the GNU General Public License, version 2, as
4 * published by the Free Software Foundation.
6 * This program is distributed in the hope that it will be useful,
7 * but WITHOUT ANY WARRANTY; without even the implied warranty of
8 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9 * GNU General Public License for more details.
11 * Copyright 2011 Paul Mackerras, IBM Corp. <paulus@au1.ibm.com>
13 * Derived from book3s_rmhandlers.S and other files, which are:
15 * Copyright SUSE Linux Products GmbH 2009
17 * Authors: Alexander Graf <agraf@suse.de>
20 #include <asm/ppc_asm.h>
21 #include <asm/kvm_asm.h>
25 #include <asm/ptrace.h>
26 #include <asm/hvcall.h>
27 #include <asm/asm-offsets.h>
28 #include <asm/exception-64s.h>
29 #include <asm/kvm_book3s_asm.h>
30 #include <asm/mmu-hash64.h>
33 #define VCPU_GPRS_TM(reg) (((reg) * ULONG_SIZE) + VCPU_GPR_TM)
35 /* Values in HSTATE_NAPPING(r13) */
36 #define NAPPING_CEDE 1
37 #define NAPPING_NOVCPU 2
40 * Call kvmppc_hv_entry in real mode.
41 * Must be called with interrupts hard-disabled.
45 * LR = return address to continue at after eventually re-enabling MMU
47 _GLOBAL_TOC(kvmppc_hv_entry_trampoline)
49 std r0, PPC_LR_STKOFF(r1)
52 LOAD_REG_ADDR(r5, kvmppc_call_hv_entry)
57 mtmsrd r0,1 /* clear RI in MSR */
63 ld r4, HSTATE_KVM_VCPU(r13)
66 /* Back from guest - restore host state and return to caller */
69 /* Restore host DABR and DABRX */
70 ld r5,HSTATE_DABR(r13)
74 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
77 ld r3,PACA_SPRG_VDSO(r13)
78 mtspr SPRN_SPRG_VDSO_WRITE,r3
80 /* Reload the host's PMU registers */
81 ld r3, PACALPPACAPTR(r13) /* is the host using the PMU? */
82 lbz r4, LPPACA_PMCINUSE(r3)
84 beq 23f /* skip if not */
86 ld r3, HSTATE_MMCR(r13)
87 andi. r4, r3, MMCR0_PMAO_SYNC | MMCR0_PMAO
90 END_FTR_SECTION_IFSET(CPU_FTR_PMAO_BUG)
91 lwz r3, HSTATE_PMC(r13)
92 lwz r4, HSTATE_PMC + 4(r13)
93 lwz r5, HSTATE_PMC + 8(r13)
94 lwz r6, HSTATE_PMC + 12(r13)
95 lwz r8, HSTATE_PMC + 16(r13)
96 lwz r9, HSTATE_PMC + 20(r13)
98 lwz r10, HSTATE_PMC + 24(r13)
99 lwz r11, HSTATE_PMC + 28(r13)
100 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
110 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
111 ld r3, HSTATE_MMCR(r13)
112 ld r4, HSTATE_MMCR + 8(r13)
113 ld r5, HSTATE_MMCR + 16(r13)
114 ld r6, HSTATE_MMCR + 24(r13)
115 ld r7, HSTATE_MMCR + 32(r13)
121 ld r8, HSTATE_MMCR + 40(r13)
122 ld r9, HSTATE_MMCR + 48(r13)
125 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
131 * Reload DEC. HDEC interrupts were disabled when
132 * we reloaded the host's LPCR value.
134 ld r3, HSTATE_DECEXP(r13)
140 * For external and machine check interrupts, we need
141 * to call the Linux handler to process the interrupt.
142 * We do that by jumping to absolute address 0x500 for
143 * external interrupts, or the machine_check_fwnmi label
144 * for machine checks (since firmware might have patched
145 * the vector area at 0x200). The [h]rfid at the end of the
146 * handler will return to the book3s_hv_interrupts.S code.
147 * For other interrupts we do the rfid to get back
148 * to the book3s_hv_interrupts.S code here.
150 ld r8, 112+PPC_LR_STKOFF(r1)
152 ld r7, HSTATE_HOST_MSR(r13)
154 cmpwi cr1, r12, BOOK3S_INTERRUPT_MACHINE_CHECK
155 cmpwi r12, BOOK3S_INTERRUPT_EXTERNAL
158 cmpwi cr2, r12, BOOK3S_INTERRUPT_HMI
159 beq cr2, 14f /* HMI check */
160 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
162 /* RFI into the highmem handler, or branch to interrupt handler */
166 mtmsrd r6, 1 /* Clear RI in MSR */
169 beqa 0x500 /* external interrupt (PPC970) */
170 beq cr1, 13f /* machine check */
173 /* On POWER7, we have external interrupts set to use HSRR0/1 */
174 11: mtspr SPRN_HSRR0, r8
178 13: b machine_check_fwnmi
180 14: mtspr SPRN_HSRR0, r8
182 b hmi_exception_after_realmode
184 kvmppc_primary_no_guest:
185 /* We handle this much like a ceded vcpu */
186 /* set our bit in napping_threads */
187 ld r5, HSTATE_KVM_VCORE(r13)
188 lbz r7, HSTATE_PTID(r13)
191 addi r6, r5, VCORE_NAPPING_THREADS
196 /* order napping_threads update vs testing entry_exit_count */
199 lwz r7, VCORE_ENTRY_EXIT(r5)
201 bge kvm_novcpu_exit /* another thread already exiting */
202 li r3, NAPPING_NOVCPU
203 stb r3, HSTATE_NAPPING(r13)
205 stb r3, HSTATE_HWTHREAD_REQ(r13)
210 ld r1, HSTATE_HOST_R1(r13)
211 ld r5, HSTATE_KVM_VCORE(r13)
213 stb r0, HSTATE_NAPPING(r13)
214 stb r0, HSTATE_HWTHREAD_REQ(r13)
216 /* check the wake reason */
217 bl kvmppc_check_wake_reason
219 /* see if any other thread is already exiting */
220 lwz r0, VCORE_ENTRY_EXIT(r5)
224 /* clear our bit in napping_threads */
225 lbz r7, HSTATE_PTID(r13)
228 addi r6, r5, VCORE_NAPPING_THREADS
234 /* See if the wake reason means we need to exit */
238 /* Got an IPI but other vcpus aren't yet exiting, must be a latecomer */
239 ld r4, HSTATE_KVM_VCPU(r13)
247 * We come in here when wakened from nap mode.
248 * Relocation is off and most register values are lost.
249 * r13 points to the PACA.
251 .globl kvm_start_guest
254 /* Set runlatch bit the minute you wake up from nap */
261 li r0,KVM_HWTHREAD_IN_KVM
262 stb r0,HSTATE_HWTHREAD_STATE(r13)
264 /* NV GPR values from power7_idle() will no longer be valid */
266 stb r0,PACA_NAPSTATELOST(r13)
268 /* were we napping due to cede? */
269 lbz r0,HSTATE_NAPPING(r13)
270 cmpwi r0,NAPPING_CEDE
272 cmpwi r0,NAPPING_NOVCPU
273 beq kvm_novcpu_wakeup
275 ld r1,PACAEMERGSP(r13)
276 subi r1,r1,STACK_FRAME_OVERHEAD
279 * We weren't napping due to cede, so this must be a secondary
280 * thread being woken up to run a guest, or being woken up due
281 * to a stray IPI. (Or due to some machine check or hypervisor
282 * maintenance interrupt while the core is in KVM.)
285 /* Check the wake reason in SRR1 to see why we got here */
286 bl kvmppc_check_wake_reason
290 /* get vcpu pointer, NULL if we have no vcpu to run */
291 ld r4,HSTATE_KVM_VCPU(r13)
293 /* if we have no vcpu to run, go back to sleep */
296 /* Set HSTATE_DSCR(r13) to something sensible */
297 ld r6, PACA_DSCR(r13)
298 std r6, HSTATE_DSCR(r13)
302 /* Back from the guest, go back to nap */
303 /* Clear our vcpu pointer so we don't come back in early */
305 std r0, HSTATE_KVM_VCPU(r13)
307 * Make sure we clear HSTATE_KVM_VCPU(r13) before incrementing
308 * the nap_count, because once the increment to nap_count is
309 * visible we could be given another vcpu.
313 /* increment the nap count and then go to nap mode */
314 ld r4, HSTATE_KVM_VCORE(r13)
315 addi r4, r4, VCORE_NAP_COUNT
322 li r0, KVM_HWTHREAD_IN_NAP
323 stb r0, HSTATE_HWTHREAD_STATE(r13)
325 /* Clear the runlatch bit before napping */
332 rlwimi r4, r3, 0, LPCR_PECE0 | LPCR_PECE1
335 std r0, HSTATE_SCRATCH0(r13)
337 ld r0, HSTATE_SCRATCH0(r13)
343 /******************************************************************************
347 *****************************************************************************/
349 .global kvmppc_hv_entry
354 * R4 = vcpu pointer (or NULL)
358 * all other volatile GPRS = free
361 std r0, PPC_LR_STKOFF(r1)
364 /* Save R1 in the PACA */
365 std r1, HSTATE_HOST_R1(r13)
367 li r6, KVM_GUEST_MODE_HOST_HV
368 stb r6, HSTATE_IN_GUEST(r13)
378 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
380 * POWER7 host -> guest partition switch code.
381 * We don't have to lock against concurrent tlbies,
382 * but we do have to coordinate across hardware threads.
384 /* Increment entry count iff exit count is zero. */
385 ld r5,HSTATE_KVM_VCORE(r13)
386 addi r9,r5,VCORE_ENTRY_EXIT
388 cmpwi r3,0x100 /* any threads starting to exit? */
389 bge secondary_too_late /* if so we're too late to the party */
394 /* Primary thread switches to guest partition. */
395 ld r9,VCORE_KVM(r5) /* pointer to struct kvm */
396 lbz r6,HSTATE_PTID(r13)
401 li r0,LPID_RSVD /* switch to reserved LPID */
404 mtspr SPRN_SDR1,r6 /* switch to partition page table */
408 /* See if we need to flush the TLB */
409 lhz r6,PACAPACAINDEX(r13) /* test_bit(cpu, need_tlb_flush) */
410 clrldi r7,r6,64-6 /* extract bit number (6 bits) */
411 srdi r6,r6,6 /* doubleword number */
412 sldi r6,r6,3 /* address offset */
414 addi r6,r6,KVM_NEED_FLUSH /* dword in kvm->arch.need_tlb_flush */
420 23: ldarx r7,0,r6 /* if set, clear the bit */
424 /* Flush the TLB of any entries for this LPID */
425 /* use arch 2.07S as a proxy for POWER8 */
427 li r6,512 /* POWER8 has 512 sets */
429 li r6,128 /* POWER7 has 128 sets */
430 ALT_FTR_SECTION_END_IFSET(CPU_FTR_ARCH_207S)
432 li r7,0x800 /* IS field = 0b10 */
439 /* Add timebase offset onto timebase */
440 22: ld r8,VCORE_TB_OFFSET(r5)
443 mftb r6 /* current host timebase */
445 mtspr SPRN_TBU40,r8 /* update upper 40 bits */
446 mftb r7 /* check if lower 24 bits overflowed */
451 addis r8,r8,0x100 /* if so, increment upper 40 bits */
454 /* Load guest PCR value to select appropriate compat mode */
455 37: ld r7, VCORE_PCR(r5)
462 /* DPDES is shared between threads */
463 ld r8, VCORE_DPDES(r5)
465 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
468 stb r0,VCORE_IN_GUEST(r5) /* signal secondaries to continue */
471 /* Secondary threads wait for primary to have done partition switch */
472 20: lbz r0,VCORE_IN_GUEST(r5)
476 /* Set LPCR and RMOR. */
477 10: ld r8,VCORE_LPCR(r5)
483 /* Check if HDEC expires soon */
485 cmpwi r3,512 /* 1 microsecond */
486 li r12,BOOK3S_INTERRUPT_HV_DECREMENTER
491 * PPC970 host -> guest partition switch code.
492 * We have to lock against concurrent tlbies,
493 * using native_tlbie_lock to lock against host tlbies
494 * and kvm->arch.tlbie_lock to lock against guest tlbies.
495 * We also have to invalidate the TLB since its
496 * entries aren't tagged with the LPID.
498 30: ld r5,HSTATE_KVM_VCORE(r13)
499 ld r9,VCORE_KVM(r5) /* pointer to struct kvm */
501 /* first take native_tlbie_lock */
504 .tc native_tlbie_lock[TC],native_tlbie_lock
506 ld r3,toc_tlbie_lock@toc(2)
507 #ifdef __BIG_ENDIAN__
508 lwz r8,PACA_LOCK_TOKEN(r13)
510 lwz r8,PACAPACAINDEX(r13)
519 ld r5,HSTATE_KVM_VCORE(r13)
520 ld r7,VCORE_LPCR(r5) /* use vcore->lpcr to store HID4 */
522 rotldi r0,r0,HID4_LPID5_SH /* all lpid bits in HID4 = 1 */
526 mtspr SPRN_HID4,r0 /* switch to reserved LPID */
529 stw r0,0(r3) /* drop native_tlbie_lock */
531 /* invalidate the whole TLB */
540 /* Take the guest's tlbie_lock */
541 addi r3,r9,KVM_TLBIE_LOCK
549 mtspr SPRN_SDR1,r6 /* switch to partition page table */
551 /* Set up HID4 with the guest's LPID etc. */
556 /* drop the guest's tlbie_lock */
560 /* Check if HDEC expires soon */
563 li r12,BOOK3S_INTERRUPT_HV_DECREMENTER
566 /* Enable HDEC interrupts */
569 rldimi r0,r3, HID0_HDICE_SH, 64-HID0_HDICE_SH-1
579 /* Do we have a guest vcpu to run? */
581 beq kvmppc_primary_no_guest
584 /* Load up guest SLB entries */
585 lwz r5,VCPU_SLB_MAX(r4)
590 1: ld r8,VCPU_SLB_E(r6)
593 addi r6,r6,VCPU_SLB_SIZE
596 /* Increment yield count if they have a VPA */
600 li r6, LPPACA_YIELDCOUNT
605 stb r6, VCPU_VPA_DIRTY(r4)
609 /* Save purr/spurr */
612 std r5,HSTATE_PURR(r13)
613 std r6,HSTATE_SPURR(r13)
618 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
621 /* Set partition DABR */
622 /* Do this before re-enabling PMU to avoid P7 DABR corruption bug */
623 lwz r5,VCPU_DABRX(r4)
627 BEGIN_FTR_SECTION_NESTED(89)
629 END_FTR_SECTION_NESTED(CPU_FTR_ARCH_206, CPU_FTR_ARCH_206, 89)
630 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
632 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
635 END_FTR_SECTION_IFCLR(CPU_FTR_TM)
637 /* Turn on TM/FP/VSX/VMX so we can restore them. */
643 oris r5, r5, (MSR_VEC | MSR_VSX)@h
647 * The user may change these outside of a transaction, so they must
648 * always be context switched.
650 ld r5, VCPU_TFHAR(r4)
651 ld r6, VCPU_TFIAR(r4)
652 ld r7, VCPU_TEXASR(r4)
655 mtspr SPRN_TEXASR, r7
658 rldicl. r5, r5, 64 - MSR_TS_S_LG, 62
659 beq skip_tm /* TM not active in guest */
661 /* Make sure the failure summary is set, otherwise we'll program check
662 * when we trechkpt. It's possible that this might have been not set
663 * on a kvmppc_set_one_reg() call but we shouldn't let this crash the
666 oris r7, r7, (TEXASR_FS)@h
667 mtspr SPRN_TEXASR, r7
670 * We need to load up the checkpointed state for the guest.
671 * We need to do this early as it will blow away any GPRs, VSRs and
676 addi r3, r31, VCPU_FPRS_TM
678 addi r3, r31, VCPU_VRS_TM
681 lwz r7, VCPU_VRSAVE_TM(r4)
682 mtspr SPRN_VRSAVE, r7
684 ld r5, VCPU_LR_TM(r4)
685 lwz r6, VCPU_CR_TM(r4)
686 ld r7, VCPU_CTR_TM(r4)
687 ld r8, VCPU_AMR_TM(r4)
688 ld r9, VCPU_TAR_TM(r4)
696 * Load up PPR and DSCR values but don't put them in the actual SPRs
697 * till the last moment to avoid running with userspace PPR and DSCR for
700 ld r29, VCPU_DSCR_TM(r4)
701 ld r30, VCPU_PPR_TM(r4)
703 std r2, PACATMSCRATCH(r13) /* Save TOC */
705 /* Clear the MSR RI since r1, r13 are all going to be foobar. */
709 /* Load GPRs r0-r28 */
712 ld reg, VCPU_GPRS_TM(reg)(r31)
719 /* Load final GPRs */
720 ld 29, VCPU_GPRS_TM(29)(r31)
721 ld 30, VCPU_GPRS_TM(30)(r31)
722 ld 31, VCPU_GPRS_TM(31)(r31)
724 /* TM checkpointed state is now setup. All GPRs are now volatile. */
727 /* Now let's get back the state we need. */
730 ld r29, HSTATE_DSCR(r13)
732 ld r4, HSTATE_KVM_VCPU(r13)
733 ld r1, HSTATE_HOST_R1(r13)
734 ld r2, PACATMSCRATCH(r13)
736 /* Set the MSR RI since we have our registers back. */
742 /* Load guest PMU registers */
743 /* R4 is live here (vcpu pointer) */
745 sldi r3, r3, 31 /* MMCR0_FC (freeze counters) bit */
746 mtspr SPRN_MMCR0, r3 /* freeze all counters, disable ints */
750 andi. r5, r3, MMCR0_PMAO_SYNC | MMCR0_PMAO
753 END_FTR_SECTION_IFSET(CPU_FTR_PMAO_BUG)
754 lwz r3, VCPU_PMC(r4) /* always load up guest PMU registers */
755 lwz r5, VCPU_PMC + 4(r4) /* to prevent information leak */
756 lwz r6, VCPU_PMC + 8(r4)
757 lwz r7, VCPU_PMC + 12(r4)
758 lwz r8, VCPU_PMC + 16(r4)
759 lwz r9, VCPU_PMC + 20(r4)
761 lwz r10, VCPU_PMC + 24(r4)
762 lwz r11, VCPU_PMC + 28(r4)
763 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
773 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
775 ld r5, VCPU_MMCR + 8(r4)
776 ld r6, VCPU_MMCR + 16(r4)
784 ld r5, VCPU_MMCR + 24(r4)
786 lwz r7, VCPU_PMC + 24(r4)
787 lwz r8, VCPU_PMC + 28(r4)
788 ld r9, VCPU_MMCR + 32(r4)
794 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
798 /* Load up FP, VMX and VSX registers */
801 ld r14, VCPU_GPR(R14)(r4)
802 ld r15, VCPU_GPR(R15)(r4)
803 ld r16, VCPU_GPR(R16)(r4)
804 ld r17, VCPU_GPR(R17)(r4)
805 ld r18, VCPU_GPR(R18)(r4)
806 ld r19, VCPU_GPR(R19)(r4)
807 ld r20, VCPU_GPR(R20)(r4)
808 ld r21, VCPU_GPR(R21)(r4)
809 ld r22, VCPU_GPR(R22)(r4)
810 ld r23, VCPU_GPR(R23)(r4)
811 ld r24, VCPU_GPR(R24)(r4)
812 ld r25, VCPU_GPR(R25)(r4)
813 ld r26, VCPU_GPR(R26)(r4)
814 ld r27, VCPU_GPR(R27)(r4)
815 ld r28, VCPU_GPR(R28)(r4)
816 ld r29, VCPU_GPR(R29)(r4)
817 ld r30, VCPU_GPR(R30)(r4)
818 ld r31, VCPU_GPR(R31)(r4)
821 /* Switch DSCR to guest value */
824 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
827 /* Skip next section on POWER7 or PPC970 */
829 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
830 /* Turn on TM so we can access TFHAR/TFIAR/TEXASR */
833 rldimi r8, r0, MSR_TM_LG, 63-MSR_TM_LG
836 /* Load up POWER8-specific registers */
838 lwz r6, VCPU_PSPB(r4)
844 ld r6, VCPU_DAWRX(r4)
845 ld r7, VCPU_CIABR(r4)
855 ld r8, VCPU_EBBHR(r4)
857 ld r5, VCPU_EBBRR(r4)
858 ld r6, VCPU_BESCR(r4)
859 ld r7, VCPU_CSIGR(r4)
865 ld r5, VCPU_TCSCR(r4)
867 lwz r7, VCPU_GUEST_PID(r4)
876 * Set the decrementer to the guest decrementer.
878 ld r8,VCPU_DEC_EXPIRES(r4)
879 /* r8 is a host timebase value here, convert to guest TB */
880 ld r5,HSTATE_KVM_VCORE(r13)
881 ld r6,VCORE_TB_OFFSET(r5)
888 ld r5, VCPU_SPRG0(r4)
889 ld r6, VCPU_SPRG1(r4)
890 ld r7, VCPU_SPRG2(r4)
891 ld r8, VCPU_SPRG3(r4)
897 /* Load up DAR and DSISR */
899 lwz r6, VCPU_DSISR(r4)
904 /* Restore AMR and UAMOR, set AMOR to all 1s */
911 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
913 /* Restore state of CTRL run bit; assume 1 on entry */
927 kvmppc_cede_reentry: /* r4 = vcpu, r13 = paca */
935 deliver_guest_interrupt:
936 /* r11 = vcpu->arch.msr & ~MSR_HV */
937 rldicl r11, r11, 63 - MSR_HV_LG, 1
938 rotldi r11, r11, 1 + MSR_HV_LG
941 /* Check if we can deliver an external or decrementer interrupt now */
942 ld r0, VCPU_PENDING_EXC(r4)
943 rldicl r0, r0, 64 - BOOK3S_IRQPRIO_EXTERNAL_LEVEL, 63
945 andi. r8, r11, MSR_EE
948 /* Insert EXTERNAL_LEVEL bit into LPCR at the MER bit position */
949 rldimi r8, r0, LPCR_MER_SH, 63 - LPCR_MER_SH
952 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
954 li r0, BOOK3S_INTERRUPT_EXTERNAL
958 li r0, BOOK3S_INTERRUPT_DECREMENTER
961 12: mtspr SPRN_SRR0, r10
965 bl kvmppc_msr_interrupt
971 * R10: value for HSRR0
972 * R11: value for HSRR1
977 stb r0,VCPU_CEDED(r4) /* cancel cede */
981 /* Activate guest mode, so faults get handled by KVM */
982 li r9, KVM_GUEST_MODE_GUEST_HV
983 stb r9, HSTATE_IN_GUEST(r13)
990 END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
993 END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
1000 ld r1, VCPU_GPR(R1)(r4)
1001 ld r2, VCPU_GPR(R2)(r4)
1002 ld r3, VCPU_GPR(R3)(r4)
1003 ld r5, VCPU_GPR(R5)(r4)
1004 ld r6, VCPU_GPR(R6)(r4)
1005 ld r7, VCPU_GPR(R7)(r4)
1006 ld r8, VCPU_GPR(R8)(r4)
1007 ld r9, VCPU_GPR(R9)(r4)
1008 ld r10, VCPU_GPR(R10)(r4)
1009 ld r11, VCPU_GPR(R11)(r4)
1010 ld r12, VCPU_GPR(R12)(r4)
1011 ld r13, VCPU_GPR(R13)(r4)
1015 END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
1016 ld r0, VCPU_GPR(R0)(r4)
1017 ld r4, VCPU_GPR(R4)(r4)
1022 /******************************************************************************
1026 *****************************************************************************/
1029 * We come here from the first-level interrupt handlers.
1031 .globl kvmppc_interrupt_hv
1032 kvmppc_interrupt_hv:
1034 * Register contents:
1035 * R12 = interrupt vector
1037 * guest CR, R12 saved in shadow VCPU SCRATCH1/0
1038 * guest R13 saved in SPRN_SCRATCH0
1040 std r9, HSTATE_SCRATCH2(r13)
1042 lbz r9, HSTATE_IN_GUEST(r13)
1043 cmpwi r9, KVM_GUEST_MODE_HOST_HV
1044 beq kvmppc_bad_host_intr
1045 #ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1046 cmpwi r9, KVM_GUEST_MODE_GUEST
1047 ld r9, HSTATE_SCRATCH2(r13)
1048 beq kvmppc_interrupt_pr
1050 /* We're now back in the host but in guest MMU context */
1051 li r9, KVM_GUEST_MODE_HOST_HV
1052 stb r9, HSTATE_IN_GUEST(r13)
1054 ld r9, HSTATE_KVM_VCPU(r13)
1056 /* Save registers */
1058 std r0, VCPU_GPR(R0)(r9)
1059 std r1, VCPU_GPR(R1)(r9)
1060 std r2, VCPU_GPR(R2)(r9)
1061 std r3, VCPU_GPR(R3)(r9)
1062 std r4, VCPU_GPR(R4)(r9)
1063 std r5, VCPU_GPR(R5)(r9)
1064 std r6, VCPU_GPR(R6)(r9)
1065 std r7, VCPU_GPR(R7)(r9)
1066 std r8, VCPU_GPR(R8)(r9)
1067 ld r0, HSTATE_SCRATCH2(r13)
1068 std r0, VCPU_GPR(R9)(r9)
1069 std r10, VCPU_GPR(R10)(r9)
1070 std r11, VCPU_GPR(R11)(r9)
1071 ld r3, HSTATE_SCRATCH0(r13)
1072 lwz r4, HSTATE_SCRATCH1(r13)
1073 std r3, VCPU_GPR(R12)(r9)
1076 ld r3, HSTATE_CFAR(r13)
1077 std r3, VCPU_CFAR(r9)
1078 END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
1080 ld r4, HSTATE_PPR(r13)
1081 std r4, VCPU_PPR(r9)
1082 END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
1084 /* Restore R1/R2 so we can handle faults */
1085 ld r1, HSTATE_HOST_R1(r13)
1088 mfspr r10, SPRN_SRR0
1089 mfspr r11, SPRN_SRR1
1090 std r10, VCPU_SRR0(r9)
1091 std r11, VCPU_SRR1(r9)
1092 andi. r0, r12, 2 /* need to read HSRR0/1? */
1094 mfspr r10, SPRN_HSRR0
1095 mfspr r11, SPRN_HSRR1
1097 1: std r10, VCPU_PC(r9)
1098 std r11, VCPU_MSR(r9)
1102 std r3, VCPU_GPR(R13)(r9)
1105 stw r12,VCPU_TRAP(r9)
1107 /* Save HEIR (HV emulation assist reg) in last_inst
1108 if this is an HEI (HV emulation interrupt, e40) */
1109 li r3,KVM_INST_FETCH_FAILED
1111 cmpwi r12,BOOK3S_INTERRUPT_H_EMUL_ASSIST
1114 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1115 11: stw r3,VCPU_LAST_INST(r9)
1117 /* these are volatile across C function calls */
1120 std r3, VCPU_CTR(r9)
1121 stw r4, VCPU_XER(r9)
1124 /* If this is a page table miss then see if it's theirs or ours */
1125 cmpwi r12, BOOK3S_INTERRUPT_H_DATA_STORAGE
1127 cmpwi r12, BOOK3S_INTERRUPT_H_INST_STORAGE
1129 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1131 /* See if this is a leftover HDEC interrupt */
1132 cmpwi r12,BOOK3S_INTERRUPT_HV_DECREMENTER
1138 /* See if this is an hcall we can handle in real mode */
1139 cmpwi r12,BOOK3S_INTERRUPT_SYSCALL
1140 beq hcall_try_real_mode
1142 /* Only handle external interrupts here on arch 206 and later */
1144 b ext_interrupt_to_host
1145 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_206)
1147 /* External interrupt ? */
1148 cmpwi r12, BOOK3S_INTERRUPT_EXTERNAL
1149 bne+ ext_interrupt_to_host
1151 /* External interrupt, first check for host_ipi. If this is
1152 * set, we know the host wants us out so let's do it now
1156 bgt ext_interrupt_to_host
1158 /* Check if any CPU is heading out to the host, if so head out too */
1159 ld r5, HSTATE_KVM_VCORE(r13)
1160 lwz r0, VCORE_ENTRY_EXIT(r5)
1162 bge ext_interrupt_to_host
1164 /* Return to guest after delivering any pending interrupt */
1166 b deliver_guest_interrupt
1168 ext_interrupt_to_host:
1170 guest_exit_cont: /* r9 = vcpu, r12 = trap, r13 = paca */
1171 /* Save more register state */
1174 std r6, VCPU_DAR(r9)
1175 stw r7, VCPU_DSISR(r9)
1177 /* don't overwrite fault_dar/fault_dsisr if HDSI */
1178 cmpwi r12,BOOK3S_INTERRUPT_H_DATA_STORAGE
1180 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1181 std r6, VCPU_FAULT_DAR(r9)
1182 stw r7, VCPU_FAULT_DSISR(r9)
1184 /* See if it is a machine check */
1185 cmpwi r12, BOOK3S_INTERRUPT_MACHINE_CHECK
1186 beq machine_check_realmode
1189 /* Save guest CTRL register, set runlatch to 1 */
1190 6: mfspr r6,SPRN_CTRLF
1191 stw r6,VCPU_CTRL(r9)
1197 /* Read the guest SLB and save it away */
1198 lwz r0,VCPU_SLB_NR(r9) /* number of entries in SLB */
1204 andis. r0,r8,SLB_ESID_V@h
1206 add r8,r8,r6 /* put index in */
1208 std r8,VCPU_SLB_E(r7)
1209 std r3,VCPU_SLB_V(r7)
1210 addi r7,r7,VCPU_SLB_SIZE
1214 stw r5,VCPU_SLB_MAX(r9)
1217 * Save the guest PURR/SPURR
1223 ld r8,VCPU_SPURR(r9)
1224 std r5,VCPU_PURR(r9)
1225 std r6,VCPU_SPURR(r9)
1230 * Restore host PURR/SPURR and add guest times
1231 * so that the time in the guest gets accounted.
1233 ld r3,HSTATE_PURR(r13)
1234 ld r4,HSTATE_SPURR(r13)
1239 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_201)
1246 /* r5 is a guest timebase value here, convert to host TB */
1247 ld r3,HSTATE_KVM_VCORE(r13)
1248 ld r4,VCORE_TB_OFFSET(r3)
1250 std r5,VCPU_DEC_EXPIRES(r9)
1254 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
1255 /* Save POWER8-specific registers */
1259 std r5, VCPU_IAMR(r9)
1260 stw r6, VCPU_PSPB(r9)
1261 std r7, VCPU_FSCR(r9)
1266 std r6, VCPU_VTB(r9)
1267 std r7, VCPU_TAR(r9)
1268 mfspr r8, SPRN_EBBHR
1269 std r8, VCPU_EBBHR(r9)
1270 mfspr r5, SPRN_EBBRR
1271 mfspr r6, SPRN_BESCR
1272 mfspr r7, SPRN_CSIGR
1274 std r5, VCPU_EBBRR(r9)
1275 std r6, VCPU_BESCR(r9)
1276 std r7, VCPU_CSIGR(r9)
1277 std r8, VCPU_TACR(r9)
1278 mfspr r5, SPRN_TCSCR
1282 std r5, VCPU_TCSCR(r9)
1283 std r6, VCPU_ACOP(r9)
1284 stw r7, VCPU_GUEST_PID(r9)
1285 std r8, VCPU_WORT(r9)
1288 /* Save and reset AMR and UAMOR before turning on the MMU */
1293 std r6,VCPU_UAMOR(r9)
1296 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1298 /* Switch DSCR back to host value */
1301 ld r7, HSTATE_DSCR(r13)
1302 std r8, VCPU_DSCR(r9)
1304 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1306 /* Save non-volatile GPRs */
1307 std r14, VCPU_GPR(R14)(r9)
1308 std r15, VCPU_GPR(R15)(r9)
1309 std r16, VCPU_GPR(R16)(r9)
1310 std r17, VCPU_GPR(R17)(r9)
1311 std r18, VCPU_GPR(R18)(r9)
1312 std r19, VCPU_GPR(R19)(r9)
1313 std r20, VCPU_GPR(R20)(r9)
1314 std r21, VCPU_GPR(R21)(r9)
1315 std r22, VCPU_GPR(R22)(r9)
1316 std r23, VCPU_GPR(R23)(r9)
1317 std r24, VCPU_GPR(R24)(r9)
1318 std r25, VCPU_GPR(R25)(r9)
1319 std r26, VCPU_GPR(R26)(r9)
1320 std r27, VCPU_GPR(R27)(r9)
1321 std r28, VCPU_GPR(R28)(r9)
1322 std r29, VCPU_GPR(R29)(r9)
1323 std r30, VCPU_GPR(R30)(r9)
1324 std r31, VCPU_GPR(R31)(r9)
1327 mfspr r3, SPRN_SPRG0
1328 mfspr r4, SPRN_SPRG1
1329 mfspr r5, SPRN_SPRG2
1330 mfspr r6, SPRN_SPRG3
1331 std r3, VCPU_SPRG0(r9)
1332 std r4, VCPU_SPRG1(r9)
1333 std r5, VCPU_SPRG2(r9)
1334 std r6, VCPU_SPRG3(r9)
1340 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1343 END_FTR_SECTION_IFCLR(CPU_FTR_TM)
1347 rldimi r8, r0, MSR_TM_LG, 63-MSR_TM_LG
1351 rldicl. r5, r5, 64 - MSR_TS_S_LG, 62
1352 beq 1f /* TM not active in guest. */
1354 li r3, TM_CAUSE_KVM_RESCHED
1356 /* Clear the MSR RI since r1, r13 are all going to be foobar. */
1360 /* All GPRs are volatile at this point. */
1363 /* Temporarily store r13 and r9 so we have some regs to play with */
1366 std r9, PACATMSCRATCH(r13)
1367 ld r9, HSTATE_KVM_VCPU(r13)
1369 /* Get a few more GPRs free. */
1370 std r29, VCPU_GPRS_TM(29)(r9)
1371 std r30, VCPU_GPRS_TM(30)(r9)
1372 std r31, VCPU_GPRS_TM(31)(r9)
1374 /* Save away PPR and DSCR soon so don't run with user values. */
1377 mfspr r30, SPRN_DSCR
1378 ld r29, HSTATE_DSCR(r13)
1379 mtspr SPRN_DSCR, r29
1381 /* Save all but r9, r13 & r29-r31 */
1384 .if (reg != 9) && (reg != 13)
1385 std reg, VCPU_GPRS_TM(reg)(r9)
1389 /* ... now save r13 */
1391 std r4, VCPU_GPRS_TM(13)(r9)
1392 /* ... and save r9 */
1393 ld r4, PACATMSCRATCH(r13)
1394 std r4, VCPU_GPRS_TM(9)(r9)
1396 /* Reload stack pointer and TOC. */
1397 ld r1, HSTATE_HOST_R1(r13)
1400 /* Set MSR RI now we have r1 and r13 back. */
1404 /* Save away checkpinted SPRs. */
1405 std r31, VCPU_PPR_TM(r9)
1406 std r30, VCPU_DSCR_TM(r9)
1412 std r5, VCPU_LR_TM(r9)
1413 stw r6, VCPU_CR_TM(r9)
1414 std r7, VCPU_CTR_TM(r9)
1415 std r8, VCPU_AMR_TM(r9)
1416 std r10, VCPU_TAR_TM(r9)
1418 /* Restore r12 as trap number. */
1419 lwz r12, VCPU_TRAP(r9)
1422 addi r3, r9, VCPU_FPRS_TM
1424 addi r3, r9, VCPU_VRS_TM
1426 mfspr r6, SPRN_VRSAVE
1427 stw r6, VCPU_VRSAVE_TM(r9)
1430 * We need to save these SPRs after the treclaim so that the software
1431 * error code is recorded correctly in the TEXASR. Also the user may
1432 * change these outside of a transaction, so they must always be
1435 mfspr r5, SPRN_TFHAR
1436 mfspr r6, SPRN_TFIAR
1437 mfspr r7, SPRN_TEXASR
1438 std r5, VCPU_TFHAR(r9)
1439 std r6, VCPU_TFIAR(r9)
1440 std r7, VCPU_TEXASR(r9)
1444 /* Increment yield count if they have a VPA */
1445 ld r8, VCPU_VPA(r9) /* do they have a VPA? */
1448 li r4, LPPACA_YIELDCOUNT
1453 stb r3, VCPU_VPA_DIRTY(r9)
1455 /* Save PMU registers if requested */
1456 /* r8 and cr0.eq are live here */
1459 * POWER8 seems to have a hardware bug where setting
1460 * MMCR0[PMAE] along with MMCR0[PMC1CE] and/or MMCR0[PMCjCE]
1461 * when some counters are already negative doesn't seem
1462 * to cause a performance monitor alert (and hence interrupt).
1463 * The effect of this is that when saving the PMU state,
1464 * if there is no PMU alert pending when we read MMCR0
1465 * before freezing the counters, but one becomes pending
1466 * before we read the counters, we lose it.
1467 * To work around this, we need a way to freeze the counters
1468 * before reading MMCR0. Normally, freezing the counters
1469 * is done by writing MMCR0 (to set MMCR0[FC]) which
1470 * unavoidably writes MMCR0[PMA0] as well. On POWER8,
1471 * we can also freeze the counters using MMCR2, by writing
1472 * 1s to all the counter freeze condition bits (there are
1473 * 9 bits each for 6 counters).
1475 li r3, -1 /* set all freeze bits */
1477 mfspr r10, SPRN_MMCR2
1478 mtspr SPRN_MMCR2, r3
1480 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
1482 sldi r3, r3, 31 /* MMCR0_FC (freeze counters) bit */
1483 mfspr r4, SPRN_MMCR0 /* save MMCR0 */
1484 mtspr SPRN_MMCR0, r3 /* freeze all counters, disable ints */
1485 mfspr r6, SPRN_MMCRA
1487 /* On P7, clear MMCRA in order to disable SDAR updates */
1489 mtspr SPRN_MMCRA, r7
1490 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1492 beq 21f /* if no VPA, save PMU stuff anyway */
1493 lbz r7, LPPACA_PMCINUSE(r8)
1494 cmpwi r7, 0 /* did they ask for PMU stuff to be saved? */
1496 std r3, VCPU_MMCR(r9) /* if not, set saved MMCR0 to FC */
1498 21: mfspr r5, SPRN_MMCR1
1501 std r4, VCPU_MMCR(r9)
1502 std r5, VCPU_MMCR + 8(r9)
1503 std r6, VCPU_MMCR + 16(r9)
1505 std r10, VCPU_MMCR + 24(r9)
1506 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
1507 std r7, VCPU_SIAR(r9)
1508 std r8, VCPU_SDAR(r9)
1516 mfspr r10, SPRN_PMC7
1517 mfspr r11, SPRN_PMC8
1518 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
1519 stw r3, VCPU_PMC(r9)
1520 stw r4, VCPU_PMC + 4(r9)
1521 stw r5, VCPU_PMC + 8(r9)
1522 stw r6, VCPU_PMC + 12(r9)
1523 stw r7, VCPU_PMC + 16(r9)
1524 stw r8, VCPU_PMC + 20(r9)
1526 stw r10, VCPU_PMC + 24(r9)
1527 stw r11, VCPU_PMC + 28(r9)
1528 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
1531 mfspr r6, SPRN_SPMC1
1532 mfspr r7, SPRN_SPMC2
1533 mfspr r8, SPRN_MMCRS
1534 std r5, VCPU_SIER(r9)
1535 stw r6, VCPU_PMC + 24(r9)
1536 stw r7, VCPU_PMC + 28(r9)
1537 std r8, VCPU_MMCR + 32(r9)
1539 mtspr SPRN_MMCRS, r4
1540 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
1548 hdec_soon: /* r12 = trap, r13 = paca */
1551 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
1553 * POWER7 guest -> host partition switch code.
1554 * We don't have to lock against tlbies but we do
1555 * have to coordinate the hardware threads.
1557 /* Increment the threads-exiting-guest count in the 0xff00
1558 bits of vcore->entry_exit_count */
1559 ld r5,HSTATE_KVM_VCORE(r13)
1560 addi r6,r5,VCORE_ENTRY_EXIT
1565 isync /* order stwcx. vs. reading napping_threads */
1568 * At this point we have an interrupt that we have to pass
1569 * up to the kernel or qemu; we can't handle it in real mode.
1570 * Thus we have to do a partition switch, so we have to
1571 * collect the other threads, if we are the first thread
1572 * to take an interrupt. To do this, we set the HDEC to 0,
1573 * which causes an HDEC interrupt in all threads within 2ns
1574 * because the HDEC register is shared between all 4 threads.
1575 * However, we don't need to bother if this is an HDEC
1576 * interrupt, since the other threads will already be on their
1577 * way here in that case.
1579 cmpwi r3,0x100 /* Are we the first here? */
1581 cmpwi r12,BOOK3S_INTERRUPT_HV_DECREMENTER
1587 * Send an IPI to any napping threads, since an HDEC interrupt
1588 * doesn't wake CPUs up from nap.
1590 lwz r3,VCORE_NAPPING_THREADS(r5)
1591 lbz r4,HSTATE_PTID(r13)
1594 andc. r3,r3,r0 /* no sense IPI'ing ourselves */
1596 /* Order entry/exit update vs. IPIs */
1598 mulli r4,r4,PACA_SIZE /* get paca for thread 0 */
1602 ld r8,HSTATE_XICS_PHYS(r6) /* get thread's XICS reg addr */
1605 stbcix r0,r7,r8 /* trigger the IPI */
1607 addi r6,r6,PACA_SIZE
1611 /* Secondary threads wait for primary to do partition switch */
1612 43: ld r5,HSTATE_KVM_VCORE(r13)
1613 ld r4,VCORE_KVM(r5) /* pointer to struct kvm */
1614 lbz r3,HSTATE_PTID(r13)
1618 13: lbz r3,VCORE_IN_GUEST(r5)
1624 /* Primary thread waits for all the secondaries to exit guest */
1625 15: lwz r3,VCORE_ENTRY_EXIT(r5)
1632 /* Primary thread switches back to host partition */
1633 ld r6,KVM_HOST_SDR1(r4)
1634 lwz r7,KVM_HOST_LPID(r4)
1635 li r8,LPID_RSVD /* switch to reserved LPID */
1638 mtspr SPRN_SDR1,r6 /* switch to partition page table */
1643 /* DPDES is shared between threads */
1644 mfspr r7, SPRN_DPDES
1645 std r7, VCORE_DPDES(r5)
1646 /* clear DPDES so we don't get guest doorbells in the host */
1648 mtspr SPRN_DPDES, r8
1649 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
1651 /* Subtract timebase offset from timebase */
1652 ld r8,VCORE_TB_OFFSET(r5)
1655 mftb r6 /* current guest timebase */
1657 mtspr SPRN_TBU40,r8 /* update upper 40 bits */
1658 mftb r7 /* check if lower 24 bits overflowed */
1663 addis r8,r8,0x100 /* if so, increment upper 40 bits */
1667 17: ld r0, VCORE_PCR(r5)
1673 /* Signal secondary CPUs to continue */
1674 stb r0,VCORE_IN_GUEST(r5)
1675 lis r8,0x7fff /* MAX_INT@h */
1678 16: ld r8,KVM_HOST_LPCR(r4)
1684 * PPC970 guest -> host partition switch code.
1685 * We have to lock against concurrent tlbies, and
1686 * we have to flush the whole TLB.
1688 32: ld r5,HSTATE_KVM_VCORE(r13)
1689 ld r4,VCORE_KVM(r5) /* pointer to struct kvm */
1691 /* Take the guest's tlbie_lock */
1692 #ifdef __BIG_ENDIAN__
1693 lwz r8,PACA_LOCK_TOKEN(r13)
1695 lwz r8,PACAPACAINDEX(r13)
1697 addi r3,r4,KVM_TLBIE_LOCK
1705 ld r7,KVM_HOST_LPCR(r4) /* use kvm->arch.host_lpcr for HID4 */
1707 rotldi r0,r0,HID4_LPID5_SH /* all lpid bits in HID4 = 1 */
1711 mtspr SPRN_HID4,r0 /* switch to reserved LPID */
1714 stw r0,0(r3) /* drop guest tlbie_lock */
1716 /* invalidate the whole TLB */
1725 /* take native_tlbie_lock */
1726 ld r3,toc_tlbie_lock@toc(2)
1734 ld r6,KVM_HOST_SDR1(r4)
1735 mtspr SPRN_SDR1,r6 /* switch to host page table */
1737 /* Set up host HID4 value */
1742 stw r0,0(r3) /* drop native_tlbie_lock */
1744 lis r8,0x7fff /* MAX_INT@h */
1747 /* Disable HDEC interrupts */
1750 rldimi r0,r3, HID0_HDICE_SH, 64-HID0_HDICE_SH-1
1760 /* load host SLB entries */
1761 33: ld r8,PACA_SLBSHADOWPTR(r13)
1763 .rept SLB_NUM_BOLTED
1764 li r3, SLBSHADOW_SAVEAREA
1768 andis. r7,r5,SLB_ESID_V@h
1774 /* Unset guest mode */
1775 li r0, KVM_GUEST_MODE_NONE
1776 stb r0, HSTATE_IN_GUEST(r13)
1778 ld r0, 112+PPC_LR_STKOFF(r1)
1784 * Check whether an HDSI is an HPTE not found fault or something else.
1785 * If it is an HPTE not found fault that is due to the guest accessing
1786 * a page that they have mapped but which we have paged out, then
1787 * we continue on with the guest exit path. In all other cases,
1788 * reflect the HDSI to the guest as a DSI.
1792 mfspr r6, SPRN_HDSISR
1793 /* HPTE not found fault or protection fault? */
1794 andis. r0, r6, (DSISR_NOHPTE | DSISR_PROTFAULT)@h
1795 beq 1f /* if not, send it to the guest */
1796 andi. r0, r11, MSR_DR /* data relocation enabled? */
1799 PPC_SLBFEE_DOT(R5, R0) /* if so, look up SLB */
1800 bne 1f /* if no SLB entry found */
1801 4: std r4, VCPU_FAULT_DAR(r9)
1802 stw r6, VCPU_FAULT_DSISR(r9)
1804 /* Search the hash table. */
1805 mr r3, r9 /* vcpu pointer */
1806 li r7, 1 /* data fault */
1807 bl kvmppc_hpte_hv_fault
1808 ld r9, HSTATE_KVM_VCPU(r13)
1810 ld r11, VCPU_MSR(r9)
1811 li r12, BOOK3S_INTERRUPT_H_DATA_STORAGE
1812 cmpdi r3, 0 /* retry the instruction */
1814 cmpdi r3, -1 /* handle in kernel mode */
1816 cmpdi r3, -2 /* MMIO emulation; need instr word */
1819 /* Synthesize a DSI for the guest */
1820 ld r4, VCPU_FAULT_DAR(r9)
1822 1: mtspr SPRN_DAR, r4
1823 mtspr SPRN_DSISR, r6
1824 mtspr SPRN_SRR0, r10
1825 mtspr SPRN_SRR1, r11
1826 li r10, BOOK3S_INTERRUPT_DATA_STORAGE
1827 bl kvmppc_msr_interrupt
1828 fast_interrupt_c_return:
1829 6: ld r7, VCPU_CTR(r9)
1830 lwz r8, VCPU_XER(r9)
1836 3: ld r5, VCPU_KVM(r9) /* not relocated, use VRMA */
1837 ld r5, KVM_VRMA_SLB_V(r5)
1840 /* If this is for emulated MMIO, load the instruction word */
1841 2: li r8, KVM_INST_FETCH_FAILED /* In case lwz faults */
1843 /* Set guest mode to 'jump over instruction' so if lwz faults
1844 * we'll just continue at the next IP. */
1845 li r0, KVM_GUEST_MODE_SKIP
1846 stb r0, HSTATE_IN_GUEST(r13)
1848 /* Do the access with MSR:DR enabled */
1850 ori r4, r3, MSR_DR /* Enable paging for data */
1855 /* Store the result */
1856 stw r8, VCPU_LAST_INST(r9)
1858 /* Unset guest mode. */
1859 li r0, KVM_GUEST_MODE_HOST_HV
1860 stb r0, HSTATE_IN_GUEST(r13)
1864 * Similarly for an HISI, reflect it to the guest as an ISI unless
1865 * it is an HPTE not found fault for a page that we have paged out.
1868 andis. r0, r11, SRR1_ISI_NOPT@h
1870 andi. r0, r11, MSR_IR /* instruction relocation enabled? */
1873 PPC_SLBFEE_DOT(R5, R0) /* if so, look up SLB */
1874 bne 1f /* if no SLB entry found */
1876 /* Search the hash table. */
1877 mr r3, r9 /* vcpu pointer */
1880 li r7, 0 /* instruction fault */
1881 bl kvmppc_hpte_hv_fault
1882 ld r9, HSTATE_KVM_VCPU(r13)
1884 ld r11, VCPU_MSR(r9)
1885 li r12, BOOK3S_INTERRUPT_H_INST_STORAGE
1886 cmpdi r3, 0 /* retry the instruction */
1887 beq fast_interrupt_c_return
1888 cmpdi r3, -1 /* handle in kernel mode */
1891 /* Synthesize an ISI for the guest */
1893 1: mtspr SPRN_SRR0, r10
1894 mtspr SPRN_SRR1, r11
1895 li r10, BOOK3S_INTERRUPT_INST_STORAGE
1896 bl kvmppc_msr_interrupt
1897 b fast_interrupt_c_return
1899 3: ld r6, VCPU_KVM(r9) /* not relocated, use VRMA */
1900 ld r5, KVM_VRMA_SLB_V(r6)
1904 * Try to handle an hcall in real mode.
1905 * Returns to the guest if we handle it, or continues on up to
1906 * the kernel if we can't (i.e. if we don't have a handler for
1907 * it, or if the handler returns H_TOO_HARD).
1909 .globl hcall_try_real_mode
1910 hcall_try_real_mode:
1911 ld r3,VCPU_GPR(R3)(r9)
1913 /* sc 1 from userspace - reflect to guest syscall */
1914 bne sc_1_fast_return
1916 cmpldi r3,hcall_real_table_end - hcall_real_table
1918 /* See if this hcall is enabled for in-kernel handling */
1920 srdi r0, r3, 8 /* r0 = (r3 / 4) >> 6 */
1921 sldi r0, r0, 3 /* index into kvm->arch.enabled_hcalls[] */
1923 ld r0, KVM_ENABLED_HCALLS(r4)
1924 rlwinm r4, r3, 32-2, 0x3f /* r4 = (r3 / 4) & 0x3f */
1928 /* Get pointer to handler, if any, and call it */
1929 LOAD_REG_ADDR(r4, hcall_real_table)
1935 mr r3,r9 /* get vcpu pointer */
1936 ld r4,VCPU_GPR(R4)(r9)
1939 beq hcall_real_fallback
1940 ld r4,HSTATE_KVM_VCPU(r13)
1941 std r3,VCPU_GPR(R3)(r4)
1949 li r10, BOOK3S_INTERRUPT_SYSCALL
1950 bl kvmppc_msr_interrupt
1954 /* We've attempted a real mode hcall, but it's punted it back
1955 * to userspace. We need to restore some clobbered volatiles
1956 * before resuming the pass-it-to-qemu path */
1957 hcall_real_fallback:
1958 li r12,BOOK3S_INTERRUPT_SYSCALL
1959 ld r9, HSTATE_KVM_VCPU(r13)
1963 .globl hcall_real_table
1965 .long 0 /* 0 - unused */
1966 .long DOTSYM(kvmppc_h_remove) - hcall_real_table
1967 .long DOTSYM(kvmppc_h_enter) - hcall_real_table
1968 .long DOTSYM(kvmppc_h_read) - hcall_real_table
1969 .long 0 /* 0x10 - H_CLEAR_MOD */
1970 .long 0 /* 0x14 - H_CLEAR_REF */
1971 .long DOTSYM(kvmppc_h_protect) - hcall_real_table
1972 .long DOTSYM(kvmppc_h_get_tce) - hcall_real_table
1973 .long DOTSYM(kvmppc_h_put_tce) - hcall_real_table
1974 .long 0 /* 0x24 - H_SET_SPRG0 */
1975 .long DOTSYM(kvmppc_h_set_dabr) - hcall_real_table
1990 #ifdef CONFIG_KVM_XICS
1991 .long DOTSYM(kvmppc_rm_h_eoi) - hcall_real_table
1992 .long DOTSYM(kvmppc_rm_h_cppr) - hcall_real_table
1993 .long DOTSYM(kvmppc_rm_h_ipi) - hcall_real_table
1994 .long 0 /* 0x70 - H_IPOLL */
1995 .long DOTSYM(kvmppc_rm_h_xirr) - hcall_real_table
1997 .long 0 /* 0x64 - H_EOI */
1998 .long 0 /* 0x68 - H_CPPR */
1999 .long 0 /* 0x6c - H_IPI */
2000 .long 0 /* 0x70 - H_IPOLL */
2001 .long 0 /* 0x74 - H_XIRR */
2029 .long DOTSYM(kvmppc_h_cede) - hcall_real_table
2046 .long DOTSYM(kvmppc_h_bulk_remove) - hcall_real_table
2050 .long DOTSYM(kvmppc_h_set_xdabr) - hcall_real_table
2051 .globl hcall_real_table_end
2052 hcall_real_table_end:
2058 _GLOBAL(kvmppc_h_set_xdabr)
2059 andi. r0, r5, DABRX_USER | DABRX_KERNEL
2061 li r0, DABRX_USER | DABRX_KERNEL | DABRX_BTI
2064 6: li r3, H_PARAMETER
2067 _GLOBAL(kvmppc_h_set_dabr)
2068 li r5, DABRX_USER | DABRX_KERNEL
2072 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
2073 std r4,VCPU_DABR(r3)
2074 stw r5, VCPU_DABRX(r3)
2075 mtspr SPRN_DABRX, r5
2076 /* Work around P7 bug where DABR can get corrupted on mtspr */
2077 1: mtspr SPRN_DABR,r4
2085 /* Emulate H_SET_DABR/X on P8 for the sake of compat mode guests */
2086 2: rlwimi r5, r4, 5, DAWRX_DR | DAWRX_DW
2087 rlwimi r5, r4, 1, DAWRX_WT
2089 std r4, VCPU_DAWR(r3)
2090 std r5, VCPU_DAWRX(r3)
2092 mtspr SPRN_DAWRX, r5
2096 _GLOBAL(kvmppc_h_cede)
2098 std r11,VCPU_MSR(r3)
2100 stb r0,VCPU_CEDED(r3)
2101 sync /* order setting ceded vs. testing prodded */
2102 lbz r5,VCPU_PRODDED(r3)
2104 bne kvm_cede_prodded
2105 li r0,0 /* set trap to 0 to say hcall is handled */
2106 stw r0,VCPU_TRAP(r3)
2108 std r0,VCPU_GPR(R3)(r3)
2110 b kvm_cede_exit /* just send it up to host on 970 */
2111 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_206)
2114 * Set our bit in the bitmask of napping threads unless all the
2115 * other threads are already napping, in which case we send this
2118 ld r5,HSTATE_KVM_VCORE(r13)
2119 lbz r6,HSTATE_PTID(r13)
2120 lwz r8,VCORE_ENTRY_EXIT(r5)
2124 addi r6,r5,VCORE_NAPPING_THREADS
2132 /* order napping_threads update vs testing entry_exit_count */
2135 stb r0,HSTATE_NAPPING(r13)
2136 lwz r7,VCORE_ENTRY_EXIT(r5)
2138 bge 33f /* another thread already exiting */
2141 * Although not specifically required by the architecture, POWER7
2142 * preserves the following registers in nap mode, even if an SMT mode
2143 * switch occurs: SLB entries, PURR, SPURR, AMOR, UAMOR, AMR, SPRG0-3,
2144 * DAR, DSISR, DABR, DABRX, DSCR, PMCx, MMCRx, SIAR, SDAR.
2146 /* Save non-volatile GPRs */
2147 std r14, VCPU_GPR(R14)(r3)
2148 std r15, VCPU_GPR(R15)(r3)
2149 std r16, VCPU_GPR(R16)(r3)
2150 std r17, VCPU_GPR(R17)(r3)
2151 std r18, VCPU_GPR(R18)(r3)
2152 std r19, VCPU_GPR(R19)(r3)
2153 std r20, VCPU_GPR(R20)(r3)
2154 std r21, VCPU_GPR(R21)(r3)
2155 std r22, VCPU_GPR(R22)(r3)
2156 std r23, VCPU_GPR(R23)(r3)
2157 std r24, VCPU_GPR(R24)(r3)
2158 std r25, VCPU_GPR(R25)(r3)
2159 std r26, VCPU_GPR(R26)(r3)
2160 std r27, VCPU_GPR(R27)(r3)
2161 std r28, VCPU_GPR(R28)(r3)
2162 std r29, VCPU_GPR(R29)(r3)
2163 std r30, VCPU_GPR(R30)(r3)
2164 std r31, VCPU_GPR(R31)(r3)
2170 * Take a nap until a decrementer or external or doobell interrupt
2171 * occurs, with PECE1, PECE0 and PECEDP set in LPCR. Also clear the
2172 * runlatch bit before napping.
2174 mfspr r2, SPRN_CTRLF
2176 mtspr SPRN_CTRLT, r2
2179 stb r0,HSTATE_HWTHREAD_REQ(r13)
2181 ori r5,r5,LPCR_PECE0 | LPCR_PECE1
2183 oris r5,r5,LPCR_PECEDP@h
2184 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
2188 std r0, HSTATE_SCRATCH0(r13)
2190 ld r0, HSTATE_SCRATCH0(r13)
2202 /* get vcpu pointer */
2203 ld r4, HSTATE_KVM_VCPU(r13)
2205 /* Woken by external or decrementer interrupt */
2206 ld r1, HSTATE_HOST_R1(r13)
2208 /* load up FP state */
2212 ld r14, VCPU_GPR(R14)(r4)
2213 ld r15, VCPU_GPR(R15)(r4)
2214 ld r16, VCPU_GPR(R16)(r4)
2215 ld r17, VCPU_GPR(R17)(r4)
2216 ld r18, VCPU_GPR(R18)(r4)
2217 ld r19, VCPU_GPR(R19)(r4)
2218 ld r20, VCPU_GPR(R20)(r4)
2219 ld r21, VCPU_GPR(R21)(r4)
2220 ld r22, VCPU_GPR(R22)(r4)
2221 ld r23, VCPU_GPR(R23)(r4)
2222 ld r24, VCPU_GPR(R24)(r4)
2223 ld r25, VCPU_GPR(R25)(r4)
2224 ld r26, VCPU_GPR(R26)(r4)
2225 ld r27, VCPU_GPR(R27)(r4)
2226 ld r28, VCPU_GPR(R28)(r4)
2227 ld r29, VCPU_GPR(R29)(r4)
2228 ld r30, VCPU_GPR(R30)(r4)
2229 ld r31, VCPU_GPR(R31)(r4)
2231 /* Check the wake reason in SRR1 to see why we got here */
2232 bl kvmppc_check_wake_reason
2234 /* clear our bit in vcore->napping_threads */
2235 34: ld r5,HSTATE_KVM_VCORE(r13)
2236 lbz r7,HSTATE_PTID(r13)
2239 addi r6,r5,VCORE_NAPPING_THREADS
2245 stb r0,HSTATE_NAPPING(r13)
2247 /* See if the wake reason means we need to exit */
2248 stw r12, VCPU_TRAP(r4)
2253 /* see if any other thread is already exiting */
2254 lwz r0,VCORE_ENTRY_EXIT(r5)
2258 b kvmppc_cede_reentry /* if not go back to guest */
2260 /* cede when already previously prodded case */
2263 stb r0,VCPU_PRODDED(r3)
2264 sync /* order testing prodded vs. clearing ceded */
2265 stb r0,VCPU_CEDED(r3)
2269 /* we've ceded but we want to give control to the host */
2271 b hcall_real_fallback
2273 /* Try to handle a machine check in real mode */
2274 machine_check_realmode:
2275 mr r3, r9 /* get vcpu pointer */
2276 bl kvmppc_realmode_machine_check
2278 cmpdi r3, 0 /* Did we handle MCE ? */
2279 ld r9, HSTATE_KVM_VCPU(r13)
2280 li r12, BOOK3S_INTERRUPT_MACHINE_CHECK
2282 * Deliver unhandled/fatal (e.g. UE) MCE errors to guest through
2283 * machine check interrupt (set HSRR0 to 0x200). And for handled
2284 * errors (no-fatal), just go back to guest execution with current
2285 * HSRR0 instead of exiting guest. This new approach will inject
2286 * machine check to guest for fatal error causing guest to crash.
2288 * The old code used to return to host for unhandled errors which
2289 * was causing guest to hang with soft lockups inside guest and
2290 * makes it difficult to recover guest instance.
2293 ld r11, VCPU_MSR(r9)
2294 bne 2f /* Continue guest execution. */
2295 /* If not, deliver a machine check. SRR0/1 are already set */
2296 li r10, BOOK3S_INTERRUPT_MACHINE_CHECK
2297 ld r11, VCPU_MSR(r9)
2298 bl kvmppc_msr_interrupt
2299 2: b fast_interrupt_c_return
2302 * Check the reason we woke from nap, and take appropriate action.
2304 * 0 if nothing needs to be done
2305 * 1 if something happened that needs to be handled by the host
2306 * -1 if there was a guest wakeup (IPI)
2308 * Also sets r12 to the interrupt vector for any interrupt that needs
2309 * to be handled now by the host (0x500 for external interrupt), or zero.
2311 kvmppc_check_wake_reason:
2314 rlwinm r6, r6, 45-31, 0xf /* extract wake reason field (P8) */
2316 rlwinm r6, r6, 45-31, 0xe /* P7 wake reason field is 3 bits */
2317 ALT_FTR_SECTION_END_IFSET(CPU_FTR_ARCH_207S)
2318 cmpwi r6, 8 /* was it an external interrupt? */
2319 li r12, BOOK3S_INTERRUPT_EXTERNAL
2320 beq kvmppc_read_intr /* if so, see what it was */
2323 cmpwi r6, 6 /* was it the decrementer? */
2326 cmpwi r6, 5 /* privileged doorbell? */
2328 cmpwi r6, 3 /* hypervisor doorbell? */
2330 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
2331 li r3, 1 /* anything else, return 1 */
2334 /* hypervisor doorbell */
2335 3: li r12, BOOK3S_INTERRUPT_H_DOORBELL
2340 * Determine what sort of external interrupt is pending (if any).
2342 * 0 if no interrupt is pending
2343 * 1 if an interrupt is pending that needs to be handled by the host
2344 * -1 if there was a guest wakeup IPI (which has now been cleared)
2347 /* see if a host IPI is pending */
2349 lbz r0, HSTATE_HOST_IPI(r13)
2353 /* Now read the interrupt from the ICP */
2354 ld r6, HSTATE_XICS_PHYS(r13)
2360 * Save XIRR for later. Since we get in in reverse endian on LE
2361 * systems, save it byte reversed and fetch it back in host endian.
2363 li r3, HSTATE_SAVED_XIRR
2365 #ifdef __LITTLE_ENDIAN__
2366 lwz r3, HSTATE_SAVED_XIRR(r13)
2370 rlwinm. r3, r3, 0, 0xffffff
2372 beq 1f /* if nothing pending in the ICP */
2374 /* We found something in the ICP...
2376 * If it's not an IPI, stash it in the PACA and return to
2377 * the host, we don't (yet) handle directing real external
2378 * interrupts directly to the guest
2380 cmpwi r3, XICS_IPI /* if there is, is it an IPI? */
2383 /* It's an IPI, clear the MFRR and EOI it */
2386 stbcix r3, r6, r8 /* clear the IPI */
2387 stwcix r0, r6, r7 /* EOI it */
2390 /* We need to re-check host IPI now in case it got set in the
2391 * meantime. If it's clear, we bounce the interrupt to the
2394 lbz r0, HSTATE_HOST_IPI(r13)
2398 /* OK, it's an IPI for us */
2402 42: /* It's not an IPI and it's for the host. We saved a copy of XIRR in
2403 * the PACA earlier, it will be picked up by the host ICP driver
2408 43: /* We raced with the host, we need to resend that IPI, bummer */
2410 stbcix r0, r6, r8 /* set the IPI */
2416 * Save away FP, VMX and VSX registers.
2418 * N.B. r30 and r31 are volatile across this function,
2419 * thus it is not callable from C.
2426 #ifdef CONFIG_ALTIVEC
2428 oris r8,r8,MSR_VEC@h
2429 END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2433 oris r8,r8,MSR_VSX@h
2434 END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2438 addi r3,r3,VCPU_FPRS
2440 #ifdef CONFIG_ALTIVEC
2442 addi r3,r31,VCPU_VRS
2444 END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2446 mfspr r6,SPRN_VRSAVE
2447 stw r6,VCPU_VRSAVE(r31)
2452 * Load up FP, VMX and VSX registers
2454 * N.B. r30 and r31 are volatile across this function,
2455 * thus it is not callable from C.
2462 #ifdef CONFIG_ALTIVEC
2464 oris r8,r8,MSR_VEC@h
2465 END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2469 oris r8,r8,MSR_VSX@h
2470 END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2474 addi r3,r4,VCPU_FPRS
2476 #ifdef CONFIG_ALTIVEC
2478 addi r3,r31,VCPU_VRS
2480 END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2482 lwz r7,VCPU_VRSAVE(r31)
2483 mtspr SPRN_VRSAVE,r7
2489 * We come here if we get any exception or interrupt while we are
2490 * executing host real mode code while in guest MMU context.
2491 * For now just spin, but we should do something better.
2493 kvmppc_bad_host_intr:
2497 * This mimics the MSR transition on IRQ delivery. The new guest MSR is taken
2498 * from VCPU_INTR_MSR and is modified based on the required TM state changes.
2499 * r11 has the guest MSR value (in/out)
2500 * r9 has a vcpu pointer (in)
2501 * r0 is used as a scratch register
2503 kvmppc_msr_interrupt:
2504 rldicl r0, r11, 64 - MSR_TS_S_LG, 62
2505 cmpwi r0, 2 /* Check if we are in transactional state.. */
2506 ld r11, VCPU_INTR_MSR(r9)
2508 /* ... if transactional, change to suspended */
2510 1: rldimi r11, r0, MSR_TS_S_LG, 63 - MSR_TS_T_LG
2514 * This works around a hardware bug on POWER8E processors, where
2515 * writing a 1 to the MMCR0[PMAO] bit doesn't generate a
2516 * performance monitor interrupt. Instead, when we need to have
2517 * an interrupt pending, we have to arrange for a counter to overflow.
2521 mtspr SPRN_MMCR2, r3
2522 lis r3, (MMCR0_PMXE | MMCR0_FCECE)@h
2523 ori r3, r3, MMCR0_PMCjCE | MMCR0_C56RUN
2524 mtspr SPRN_MMCR0, r3