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 #ifdef __LITTLE_ENDIAN__
36 #error Need to fix lppaca and SLB shadow accesses in little endian mode
39 /* Values in HSTATE_NAPPING(r13) */
40 #define NAPPING_CEDE 1
41 #define NAPPING_NOVCPU 2
44 * Call kvmppc_hv_entry in real mode.
45 * Must be called with interrupts hard-disabled.
49 * LR = return address to continue at after eventually re-enabling MMU
51 _GLOBAL(kvmppc_hv_entry_trampoline)
53 std r0, PPC_LR_STKOFF(r1)
56 LOAD_REG_ADDR(r5, kvmppc_call_hv_entry)
61 mtmsrd r0,1 /* clear RI in MSR */
67 ld r4, HSTATE_KVM_VCPU(r13)
70 /* Back from guest - restore host state and return to caller */
73 /* Restore host DABR and DABRX */
74 ld r5,HSTATE_DABR(r13)
78 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
81 ld r3,PACA_SPRG_VDSO(r13)
82 mtspr SPRN_SPRG_VDSO_WRITE,r3
84 /* Reload the host's PMU registers */
85 ld r3, PACALPPACAPTR(r13) /* is the host using the PMU? */
86 lbz r4, LPPACA_PMCINUSE(r3)
88 beq 23f /* skip if not */
89 lwz r3, HSTATE_PMC(r13)
90 lwz r4, HSTATE_PMC + 4(r13)
91 lwz r5, HSTATE_PMC + 8(r13)
92 lwz r6, HSTATE_PMC + 12(r13)
93 lwz r8, HSTATE_PMC + 16(r13)
94 lwz r9, HSTATE_PMC + 20(r13)
96 lwz r10, HSTATE_PMC + 24(r13)
97 lwz r11, HSTATE_PMC + 28(r13)
98 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
108 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
109 ld r3, HSTATE_MMCR(r13)
110 ld r4, HSTATE_MMCR + 8(r13)
111 ld r5, HSTATE_MMCR + 16(r13)
112 ld r6, HSTATE_MMCR + 24(r13)
113 ld r7, HSTATE_MMCR + 32(r13)
119 ld r8, HSTATE_MMCR + 40(r13)
120 ld r9, HSTATE_MMCR + 48(r13)
123 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
129 * Reload DEC. HDEC interrupts were disabled when
130 * we reloaded the host's LPCR value.
132 ld r3, HSTATE_DECEXP(r13)
138 * For external and machine check interrupts, we need
139 * to call the Linux handler to process the interrupt.
140 * We do that by jumping to absolute address 0x500 for
141 * external interrupts, or the machine_check_fwnmi label
142 * for machine checks (since firmware might have patched
143 * the vector area at 0x200). The [h]rfid at the end of the
144 * handler will return to the book3s_hv_interrupts.S code.
145 * For other interrupts we do the rfid to get back
146 * to the book3s_hv_interrupts.S code here.
148 ld r8, 112+PPC_LR_STKOFF(r1)
150 ld r7, HSTATE_HOST_MSR(r13)
152 cmpwi cr1, r12, BOOK3S_INTERRUPT_MACHINE_CHECK
153 cmpwi r12, BOOK3S_INTERRUPT_EXTERNAL
156 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
158 /* RFI into the highmem handler, or branch to interrupt handler */
162 mtmsrd r6, 1 /* Clear RI in MSR */
165 beqa 0x500 /* external interrupt (PPC970) */
166 beq cr1, 13f /* machine check */
169 /* On POWER7, we have external interrupts set to use HSRR0/1 */
170 11: mtspr SPRN_HSRR0, r8
174 13: b machine_check_fwnmi
176 kvmppc_primary_no_guest:
177 /* We handle this much like a ceded vcpu */
178 /* set our bit in napping_threads */
179 ld r5, HSTATE_KVM_VCORE(r13)
180 lbz r7, HSTATE_PTID(r13)
183 addi r6, r5, VCORE_NAPPING_THREADS
188 /* order napping_threads update vs testing entry_exit_count */
191 lwz r7, VCORE_ENTRY_EXIT(r5)
193 bge kvm_novcpu_exit /* another thread already exiting */
194 li r3, NAPPING_NOVCPU
195 stb r3, HSTATE_NAPPING(r13)
197 stb r3, HSTATE_HWTHREAD_REQ(r13)
202 ld r1, HSTATE_HOST_R1(r13)
203 ld r5, HSTATE_KVM_VCORE(r13)
205 stb r0, HSTATE_NAPPING(r13)
206 stb r0, HSTATE_HWTHREAD_REQ(r13)
208 /* check the wake reason */
209 bl kvmppc_check_wake_reason
211 /* see if any other thread is already exiting */
212 lwz r0, VCORE_ENTRY_EXIT(r5)
216 /* clear our bit in napping_threads */
217 lbz r7, HSTATE_PTID(r13)
220 addi r6, r5, VCORE_NAPPING_THREADS
226 /* See if the wake reason means we need to exit */
230 /* Got an IPI but other vcpus aren't yet exiting, must be a latecomer */
231 ld r4, HSTATE_KVM_VCPU(r13)
239 * We come in here when wakened from nap mode.
240 * Relocation is off and most register values are lost.
241 * r13 points to the PACA.
243 .globl kvm_start_guest
246 /* Set runlatch bit the minute you wake up from nap */
253 li r0,KVM_HWTHREAD_IN_KVM
254 stb r0,HSTATE_HWTHREAD_STATE(r13)
256 /* NV GPR values from power7_idle() will no longer be valid */
258 stb r0,PACA_NAPSTATELOST(r13)
260 /* were we napping due to cede? */
261 lbz r0,HSTATE_NAPPING(r13)
262 cmpwi r0,NAPPING_CEDE
264 cmpwi r0,NAPPING_NOVCPU
265 beq kvm_novcpu_wakeup
267 ld r1,PACAEMERGSP(r13)
268 subi r1,r1,STACK_FRAME_OVERHEAD
271 * We weren't napping due to cede, so this must be a secondary
272 * thread being woken up to run a guest, or being woken up due
273 * to a stray IPI. (Or due to some machine check or hypervisor
274 * maintenance interrupt while the core is in KVM.)
277 /* Check the wake reason in SRR1 to see why we got here */
278 bl kvmppc_check_wake_reason
282 /* get vcpu pointer, NULL if we have no vcpu to run */
283 ld r4,HSTATE_KVM_VCPU(r13)
285 /* if we have no vcpu to run, go back to sleep */
288 /* Set HSTATE_DSCR(r13) to something sensible */
289 LOAD_REG_ADDR(r6, dscr_default)
291 std r6, HSTATE_DSCR(r13)
295 /* Back from the guest, go back to nap */
296 /* Clear our vcpu pointer so we don't come back in early */
298 std r0, HSTATE_KVM_VCPU(r13)
300 * Make sure we clear HSTATE_KVM_VCPU(r13) before incrementing
301 * the nap_count, because once the increment to nap_count is
302 * visible we could be given another vcpu.
306 /* increment the nap count and then go to nap mode */
307 ld r4, HSTATE_KVM_VCORE(r13)
308 addi r4, r4, VCORE_NAP_COUNT
315 li r0, KVM_HWTHREAD_IN_NAP
316 stb r0, HSTATE_HWTHREAD_STATE(r13)
318 /* Clear the runlatch bit before napping */
325 rlwimi r4, r3, 0, LPCR_PECE0 | LPCR_PECE1
328 std r0, HSTATE_SCRATCH0(r13)
330 ld r0, HSTATE_SCRATCH0(r13)
336 /******************************************************************************
340 *****************************************************************************/
342 .global kvmppc_hv_entry
347 * R4 = vcpu pointer (or NULL)
351 * all other volatile GPRS = free
354 std r0, PPC_LR_STKOFF(r1)
357 /* Save R1 in the PACA */
358 std r1, HSTATE_HOST_R1(r13)
360 li r6, KVM_GUEST_MODE_HOST_HV
361 stb r6, HSTATE_IN_GUEST(r13)
371 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
373 * POWER7 host -> guest partition switch code.
374 * We don't have to lock against concurrent tlbies,
375 * but we do have to coordinate across hardware threads.
377 /* Increment entry count iff exit count is zero. */
378 ld r5,HSTATE_KVM_VCORE(r13)
379 addi r9,r5,VCORE_ENTRY_EXIT
381 cmpwi r3,0x100 /* any threads starting to exit? */
382 bge secondary_too_late /* if so we're too late to the party */
387 /* Primary thread switches to guest partition. */
388 ld r9,VCORE_KVM(r5) /* pointer to struct kvm */
389 lbz r6,HSTATE_PTID(r13)
394 li r0,LPID_RSVD /* switch to reserved LPID */
397 mtspr SPRN_SDR1,r6 /* switch to partition page table */
401 /* See if we need to flush the TLB */
402 lhz r6,PACAPACAINDEX(r13) /* test_bit(cpu, need_tlb_flush) */
403 clrldi r7,r6,64-6 /* extract bit number (6 bits) */
404 srdi r6,r6,6 /* doubleword number */
405 sldi r6,r6,3 /* address offset */
407 addi r6,r6,KVM_NEED_FLUSH /* dword in kvm->arch.need_tlb_flush */
413 23: ldarx r7,0,r6 /* if set, clear the bit */
417 /* Flush the TLB of any entries for this LPID */
418 /* use arch 2.07S as a proxy for POWER8 */
420 li r6,512 /* POWER8 has 512 sets */
422 li r6,128 /* POWER7 has 128 sets */
423 ALT_FTR_SECTION_END_IFSET(CPU_FTR_ARCH_207S)
425 li r7,0x800 /* IS field = 0b10 */
432 /* Add timebase offset onto timebase */
433 22: ld r8,VCORE_TB_OFFSET(r5)
436 mftb r6 /* current host timebase */
438 mtspr SPRN_TBU40,r8 /* update upper 40 bits */
439 mftb r7 /* check if lower 24 bits overflowed */
444 addis r8,r8,0x100 /* if so, increment upper 40 bits */
447 /* Load guest PCR value to select appropriate compat mode */
448 37: ld r7, VCORE_PCR(r5)
455 /* DPDES is shared between threads */
456 ld r8, VCORE_DPDES(r5)
458 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
461 stb r0,VCORE_IN_GUEST(r5) /* signal secondaries to continue */
464 /* Secondary threads wait for primary to have done partition switch */
465 20: lbz r0,VCORE_IN_GUEST(r5)
469 /* Set LPCR and RMOR. */
470 10: ld r8,VCORE_LPCR(r5)
476 /* Check if HDEC expires soon */
478 cmpwi r3,512 /* 1 microsecond */
479 li r12,BOOK3S_INTERRUPT_HV_DECREMENTER
484 * PPC970 host -> guest partition switch code.
485 * We have to lock against concurrent tlbies,
486 * using native_tlbie_lock to lock against host tlbies
487 * and kvm->arch.tlbie_lock to lock against guest tlbies.
488 * We also have to invalidate the TLB since its
489 * entries aren't tagged with the LPID.
491 30: ld r5,HSTATE_KVM_VCORE(r13)
492 ld r9,VCORE_KVM(r5) /* pointer to struct kvm */
494 /* first take native_tlbie_lock */
497 .tc native_tlbie_lock[TC],native_tlbie_lock
499 ld r3,toc_tlbie_lock@toc(2)
500 #ifdef __BIG_ENDIAN__
501 lwz r8,PACA_LOCK_TOKEN(r13)
503 lwz r8,PACAPACAINDEX(r13)
512 ld r5,HSTATE_KVM_VCORE(r13)
513 ld r7,VCORE_LPCR(r5) /* use vcore->lpcr to store HID4 */
515 rotldi r0,r0,HID4_LPID5_SH /* all lpid bits in HID4 = 1 */
519 mtspr SPRN_HID4,r0 /* switch to reserved LPID */
522 stw r0,0(r3) /* drop native_tlbie_lock */
524 /* invalidate the whole TLB */
533 /* Take the guest's tlbie_lock */
534 addi r3,r9,KVM_TLBIE_LOCK
542 mtspr SPRN_SDR1,r6 /* switch to partition page table */
544 /* Set up HID4 with the guest's LPID etc. */
549 /* drop the guest's tlbie_lock */
553 /* Check if HDEC expires soon */
556 li r12,BOOK3S_INTERRUPT_HV_DECREMENTER
559 /* Enable HDEC interrupts */
562 rldimi r0,r3, HID0_HDICE_SH, 64-HID0_HDICE_SH-1
572 /* Do we have a guest vcpu to run? */
574 beq kvmppc_primary_no_guest
577 /* Load up guest SLB entries */
578 lwz r5,VCPU_SLB_MAX(r4)
583 1: ld r8,VCPU_SLB_E(r6)
586 addi r6,r6,VCPU_SLB_SIZE
589 /* Increment yield count if they have a VPA */
593 lwz r5, LPPACA_YIELDCOUNT(r3)
595 stw r5, LPPACA_YIELDCOUNT(r3)
597 stb r6, VCPU_VPA_DIRTY(r4)
601 /* Save purr/spurr */
604 std r5,HSTATE_PURR(r13)
605 std r6,HSTATE_SPURR(r13)
610 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
613 /* Set partition DABR */
614 /* Do this before re-enabling PMU to avoid P7 DABR corruption bug */
615 lwz r5,VCPU_DABRX(r4)
619 BEGIN_FTR_SECTION_NESTED(89)
621 END_FTR_SECTION_NESTED(CPU_FTR_ARCH_206, CPU_FTR_ARCH_206, 89)
622 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
624 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
627 END_FTR_SECTION_IFCLR(CPU_FTR_TM)
629 /* Turn on TM/FP/VSX/VMX so we can restore them. */
635 oris r5, r5, (MSR_VEC | MSR_VSX)@h
639 * The user may change these outside of a transaction, so they must
640 * always be context switched.
642 ld r5, VCPU_TFHAR(r4)
643 ld r6, VCPU_TFIAR(r4)
644 ld r7, VCPU_TEXASR(r4)
647 mtspr SPRN_TEXASR, r7
650 rldicl. r5, r5, 64 - MSR_TS_S_LG, 62
651 beq skip_tm /* TM not active in guest */
653 /* Make sure the failure summary is set, otherwise we'll program check
654 * when we trechkpt. It's possible that this might have been not set
655 * on a kvmppc_set_one_reg() call but we shouldn't let this crash the
658 oris r7, r7, (TEXASR_FS)@h
659 mtspr SPRN_TEXASR, r7
662 * We need to load up the checkpointed state for the guest.
663 * We need to do this early as it will blow away any GPRs, VSRs and
668 addi r3, r31, VCPU_FPRS_TM
670 addi r3, r31, VCPU_VRS_TM
673 lwz r7, VCPU_VRSAVE_TM(r4)
674 mtspr SPRN_VRSAVE, r7
676 ld r5, VCPU_LR_TM(r4)
677 lwz r6, VCPU_CR_TM(r4)
678 ld r7, VCPU_CTR_TM(r4)
679 ld r8, VCPU_AMR_TM(r4)
680 ld r9, VCPU_TAR_TM(r4)
688 * Load up PPR and DSCR values but don't put them in the actual SPRs
689 * till the last moment to avoid running with userspace PPR and DSCR for
692 ld r29, VCPU_DSCR_TM(r4)
693 ld r30, VCPU_PPR_TM(r4)
695 std r2, PACATMSCRATCH(r13) /* Save TOC */
697 /* Clear the MSR RI since r1, r13 are all going to be foobar. */
701 /* Load GPRs r0-r28 */
704 ld reg, VCPU_GPRS_TM(reg)(r31)
711 /* Load final GPRs */
712 ld 29, VCPU_GPRS_TM(29)(r31)
713 ld 30, VCPU_GPRS_TM(30)(r31)
714 ld 31, VCPU_GPRS_TM(31)(r31)
716 /* TM checkpointed state is now setup. All GPRs are now volatile. */
719 /* Now let's get back the state we need. */
722 ld r29, HSTATE_DSCR(r13)
724 ld r4, HSTATE_KVM_VCPU(r13)
725 ld r1, HSTATE_HOST_R1(r13)
726 ld r2, PACATMSCRATCH(r13)
728 /* Set the MSR RI since we have our registers back. */
734 /* Load guest PMU registers */
735 /* R4 is live here (vcpu pointer) */
737 sldi r3, r3, 31 /* MMCR0_FC (freeze counters) bit */
738 mtspr SPRN_MMCR0, r3 /* freeze all counters, disable ints */
740 lwz r3, VCPU_PMC(r4) /* always load up guest PMU registers */
741 lwz r5, VCPU_PMC + 4(r4) /* to prevent information leak */
742 lwz r6, VCPU_PMC + 8(r4)
743 lwz r7, VCPU_PMC + 12(r4)
744 lwz r8, VCPU_PMC + 16(r4)
745 lwz r9, VCPU_PMC + 20(r4)
747 lwz r10, VCPU_PMC + 24(r4)
748 lwz r11, VCPU_PMC + 28(r4)
749 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
759 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
761 ld r5, VCPU_MMCR + 8(r4)
762 ld r6, VCPU_MMCR + 16(r4)
770 ld r5, VCPU_MMCR + 24(r4)
772 lwz r7, VCPU_PMC + 24(r4)
773 lwz r8, VCPU_PMC + 28(r4)
774 ld r9, VCPU_MMCR + 32(r4)
780 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
784 /* Load up FP, VMX and VSX registers */
787 ld r14, VCPU_GPR(R14)(r4)
788 ld r15, VCPU_GPR(R15)(r4)
789 ld r16, VCPU_GPR(R16)(r4)
790 ld r17, VCPU_GPR(R17)(r4)
791 ld r18, VCPU_GPR(R18)(r4)
792 ld r19, VCPU_GPR(R19)(r4)
793 ld r20, VCPU_GPR(R20)(r4)
794 ld r21, VCPU_GPR(R21)(r4)
795 ld r22, VCPU_GPR(R22)(r4)
796 ld r23, VCPU_GPR(R23)(r4)
797 ld r24, VCPU_GPR(R24)(r4)
798 ld r25, VCPU_GPR(R25)(r4)
799 ld r26, VCPU_GPR(R26)(r4)
800 ld r27, VCPU_GPR(R27)(r4)
801 ld r28, VCPU_GPR(R28)(r4)
802 ld r29, VCPU_GPR(R29)(r4)
803 ld r30, VCPU_GPR(R30)(r4)
804 ld r31, VCPU_GPR(R31)(r4)
807 /* Switch DSCR to guest value */
810 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
813 /* Skip next section on POWER7 or PPC970 */
815 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
816 /* Turn on TM so we can access TFHAR/TFIAR/TEXASR */
819 rldimi r8, r0, MSR_TM_LG, 63-MSR_TM_LG
822 /* Load up POWER8-specific registers */
824 lwz r6, VCPU_PSPB(r4)
830 ld r6, VCPU_DAWRX(r4)
831 ld r7, VCPU_CIABR(r4)
841 ld r8, VCPU_EBBHR(r4)
843 ld r5, VCPU_EBBRR(r4)
844 ld r6, VCPU_BESCR(r4)
845 ld r7, VCPU_CSIGR(r4)
851 ld r5, VCPU_TCSCR(r4)
853 lwz r7, VCPU_GUEST_PID(r4)
862 * Set the decrementer to the guest decrementer.
864 ld r8,VCPU_DEC_EXPIRES(r4)
865 /* r8 is a host timebase value here, convert to guest TB */
866 ld r5,HSTATE_KVM_VCORE(r13)
867 ld r6,VCORE_TB_OFFSET(r5)
874 ld r5, VCPU_SPRG0(r4)
875 ld r6, VCPU_SPRG1(r4)
876 ld r7, VCPU_SPRG2(r4)
877 ld r8, VCPU_SPRG3(r4)
883 /* Load up DAR and DSISR */
885 lwz r6, VCPU_DSISR(r4)
890 /* Restore AMR and UAMOR, set AMOR to all 1s */
897 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
899 /* Restore state of CTRL run bit; assume 1 on entry */
913 kvmppc_cede_reentry: /* r4 = vcpu, r13 = paca */
921 deliver_guest_interrupt:
922 /* r11 = vcpu->arch.msr & ~MSR_HV */
923 rldicl r11, r11, 63 - MSR_HV_LG, 1
924 rotldi r11, r11, 1 + MSR_HV_LG
927 /* Check if we can deliver an external or decrementer interrupt now */
928 ld r0, VCPU_PENDING_EXC(r4)
929 rldicl r0, r0, 64 - BOOK3S_IRQPRIO_EXTERNAL_LEVEL, 63
931 andi. r8, r11, MSR_EE
934 /* Insert EXTERNAL_LEVEL bit into LPCR at the MER bit position */
935 rldimi r8, r0, LPCR_MER_SH, 63 - LPCR_MER_SH
938 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
940 li r0, BOOK3S_INTERRUPT_EXTERNAL
944 li r0, BOOK3S_INTERRUPT_DECREMENTER
947 12: mtspr SPRN_SRR0, r10
951 bl kvmppc_msr_interrupt
957 * R10: value for HSRR0
958 * R11: value for HSRR1
963 stb r0,VCPU_CEDED(r4) /* cancel cede */
967 /* Activate guest mode, so faults get handled by KVM */
968 li r9, KVM_GUEST_MODE_GUEST_HV
969 stb r9, HSTATE_IN_GUEST(r13)
976 END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
979 END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
986 ld r1, VCPU_GPR(R1)(r4)
987 ld r2, VCPU_GPR(R2)(r4)
988 ld r3, VCPU_GPR(R3)(r4)
989 ld r5, VCPU_GPR(R5)(r4)
990 ld r6, VCPU_GPR(R6)(r4)
991 ld r7, VCPU_GPR(R7)(r4)
992 ld r8, VCPU_GPR(R8)(r4)
993 ld r9, VCPU_GPR(R9)(r4)
994 ld r10, VCPU_GPR(R10)(r4)
995 ld r11, VCPU_GPR(R11)(r4)
996 ld r12, VCPU_GPR(R12)(r4)
997 ld r13, VCPU_GPR(R13)(r4)
1001 END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
1002 ld r0, VCPU_GPR(R0)(r4)
1003 ld r4, VCPU_GPR(R4)(r4)
1008 /******************************************************************************
1012 *****************************************************************************/
1015 * We come here from the first-level interrupt handlers.
1017 .globl kvmppc_interrupt_hv
1018 kvmppc_interrupt_hv:
1020 * Register contents:
1021 * R12 = interrupt vector
1023 * guest CR, R12 saved in shadow VCPU SCRATCH1/0
1024 * guest R13 saved in SPRN_SCRATCH0
1026 std r9, HSTATE_SCRATCH2(r13)
1028 lbz r9, HSTATE_IN_GUEST(r13)
1029 cmpwi r9, KVM_GUEST_MODE_HOST_HV
1030 beq kvmppc_bad_host_intr
1031 #ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1032 cmpwi r9, KVM_GUEST_MODE_GUEST
1033 ld r9, HSTATE_SCRATCH2(r13)
1034 beq kvmppc_interrupt_pr
1036 /* We're now back in the host but in guest MMU context */
1037 li r9, KVM_GUEST_MODE_HOST_HV
1038 stb r9, HSTATE_IN_GUEST(r13)
1040 ld r9, HSTATE_KVM_VCPU(r13)
1042 /* Save registers */
1044 std r0, VCPU_GPR(R0)(r9)
1045 std r1, VCPU_GPR(R1)(r9)
1046 std r2, VCPU_GPR(R2)(r9)
1047 std r3, VCPU_GPR(R3)(r9)
1048 std r4, VCPU_GPR(R4)(r9)
1049 std r5, VCPU_GPR(R5)(r9)
1050 std r6, VCPU_GPR(R6)(r9)
1051 std r7, VCPU_GPR(R7)(r9)
1052 std r8, VCPU_GPR(R8)(r9)
1053 ld r0, HSTATE_SCRATCH2(r13)
1054 std r0, VCPU_GPR(R9)(r9)
1055 std r10, VCPU_GPR(R10)(r9)
1056 std r11, VCPU_GPR(R11)(r9)
1057 ld r3, HSTATE_SCRATCH0(r13)
1058 lwz r4, HSTATE_SCRATCH1(r13)
1059 std r3, VCPU_GPR(R12)(r9)
1062 ld r3, HSTATE_CFAR(r13)
1063 std r3, VCPU_CFAR(r9)
1064 END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
1066 ld r4, HSTATE_PPR(r13)
1067 std r4, VCPU_PPR(r9)
1068 END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
1070 /* Restore R1/R2 so we can handle faults */
1071 ld r1, HSTATE_HOST_R1(r13)
1074 mfspr r10, SPRN_SRR0
1075 mfspr r11, SPRN_SRR1
1076 std r10, VCPU_SRR0(r9)
1077 std r11, VCPU_SRR1(r9)
1078 andi. r0, r12, 2 /* need to read HSRR0/1? */
1080 mfspr r10, SPRN_HSRR0
1081 mfspr r11, SPRN_HSRR1
1083 1: std r10, VCPU_PC(r9)
1084 std r11, VCPU_MSR(r9)
1088 std r3, VCPU_GPR(R13)(r9)
1091 stw r12,VCPU_TRAP(r9)
1093 /* Save HEIR (HV emulation assist reg) in last_inst
1094 if this is an HEI (HV emulation interrupt, e40) */
1095 li r3,KVM_INST_FETCH_FAILED
1097 cmpwi r12,BOOK3S_INTERRUPT_H_EMUL_ASSIST
1100 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1101 11: stw r3,VCPU_LAST_INST(r9)
1103 /* these are volatile across C function calls */
1106 std r3, VCPU_CTR(r9)
1107 stw r4, VCPU_XER(r9)
1110 /* If this is a page table miss then see if it's theirs or ours */
1111 cmpwi r12, BOOK3S_INTERRUPT_H_DATA_STORAGE
1113 cmpwi r12, BOOK3S_INTERRUPT_H_INST_STORAGE
1115 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1117 /* See if this is a leftover HDEC interrupt */
1118 cmpwi r12,BOOK3S_INTERRUPT_HV_DECREMENTER
1124 /* See if this is an hcall we can handle in real mode */
1125 cmpwi r12,BOOK3S_INTERRUPT_SYSCALL
1126 beq hcall_try_real_mode
1128 /* Only handle external interrupts here on arch 206 and later */
1130 b ext_interrupt_to_host
1131 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_206)
1133 /* External interrupt ? */
1134 cmpwi r12, BOOK3S_INTERRUPT_EXTERNAL
1135 bne+ ext_interrupt_to_host
1137 /* External interrupt, first check for host_ipi. If this is
1138 * set, we know the host wants us out so let's do it now
1142 bgt ext_interrupt_to_host
1144 /* Check if any CPU is heading out to the host, if so head out too */
1145 ld r5, HSTATE_KVM_VCORE(r13)
1146 lwz r0, VCORE_ENTRY_EXIT(r5)
1148 bge ext_interrupt_to_host
1150 /* Return to guest after delivering any pending interrupt */
1152 b deliver_guest_interrupt
1154 ext_interrupt_to_host:
1156 guest_exit_cont: /* r9 = vcpu, r12 = trap, r13 = paca */
1157 /* Save more register state */
1160 std r6, VCPU_DAR(r9)
1161 stw r7, VCPU_DSISR(r9)
1163 /* don't overwrite fault_dar/fault_dsisr if HDSI */
1164 cmpwi r12,BOOK3S_INTERRUPT_H_DATA_STORAGE
1166 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1167 std r6, VCPU_FAULT_DAR(r9)
1168 stw r7, VCPU_FAULT_DSISR(r9)
1170 /* See if it is a machine check */
1171 cmpwi r12, BOOK3S_INTERRUPT_MACHINE_CHECK
1172 beq machine_check_realmode
1175 /* Save guest CTRL register, set runlatch to 1 */
1176 6: mfspr r6,SPRN_CTRLF
1177 stw r6,VCPU_CTRL(r9)
1183 /* Read the guest SLB and save it away */
1184 lwz r0,VCPU_SLB_NR(r9) /* number of entries in SLB */
1190 andis. r0,r8,SLB_ESID_V@h
1192 add r8,r8,r6 /* put index in */
1194 std r8,VCPU_SLB_E(r7)
1195 std r3,VCPU_SLB_V(r7)
1196 addi r7,r7,VCPU_SLB_SIZE
1200 stw r5,VCPU_SLB_MAX(r9)
1203 * Save the guest PURR/SPURR
1209 ld r8,VCPU_SPURR(r9)
1210 std r5,VCPU_PURR(r9)
1211 std r6,VCPU_SPURR(r9)
1216 * Restore host PURR/SPURR and add guest times
1217 * so that the time in the guest gets accounted.
1219 ld r3,HSTATE_PURR(r13)
1220 ld r4,HSTATE_SPURR(r13)
1225 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_201)
1232 /* r5 is a guest timebase value here, convert to host TB */
1233 ld r3,HSTATE_KVM_VCORE(r13)
1234 ld r4,VCORE_TB_OFFSET(r3)
1236 std r5,VCPU_DEC_EXPIRES(r9)
1240 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
1241 /* Save POWER8-specific registers */
1245 std r5, VCPU_IAMR(r9)
1246 stw r6, VCPU_PSPB(r9)
1247 std r7, VCPU_FSCR(r9)
1252 std r6, VCPU_VTB(r9)
1253 std r7, VCPU_TAR(r9)
1254 mfspr r8, SPRN_EBBHR
1255 std r8, VCPU_EBBHR(r9)
1256 mfspr r5, SPRN_EBBRR
1257 mfspr r6, SPRN_BESCR
1258 mfspr r7, SPRN_CSIGR
1260 std r5, VCPU_EBBRR(r9)
1261 std r6, VCPU_BESCR(r9)
1262 std r7, VCPU_CSIGR(r9)
1263 std r8, VCPU_TACR(r9)
1264 mfspr r5, SPRN_TCSCR
1268 std r5, VCPU_TCSCR(r9)
1269 std r6, VCPU_ACOP(r9)
1270 stw r7, VCPU_GUEST_PID(r9)
1271 std r8, VCPU_WORT(r9)
1274 /* Save and reset AMR and UAMOR before turning on the MMU */
1279 std r6,VCPU_UAMOR(r9)
1282 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1284 /* Switch DSCR back to host value */
1287 ld r7, HSTATE_DSCR(r13)
1288 std r8, VCPU_DSCR(r9)
1290 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1292 /* Save non-volatile GPRs */
1293 std r14, VCPU_GPR(R14)(r9)
1294 std r15, VCPU_GPR(R15)(r9)
1295 std r16, VCPU_GPR(R16)(r9)
1296 std r17, VCPU_GPR(R17)(r9)
1297 std r18, VCPU_GPR(R18)(r9)
1298 std r19, VCPU_GPR(R19)(r9)
1299 std r20, VCPU_GPR(R20)(r9)
1300 std r21, VCPU_GPR(R21)(r9)
1301 std r22, VCPU_GPR(R22)(r9)
1302 std r23, VCPU_GPR(R23)(r9)
1303 std r24, VCPU_GPR(R24)(r9)
1304 std r25, VCPU_GPR(R25)(r9)
1305 std r26, VCPU_GPR(R26)(r9)
1306 std r27, VCPU_GPR(R27)(r9)
1307 std r28, VCPU_GPR(R28)(r9)
1308 std r29, VCPU_GPR(R29)(r9)
1309 std r30, VCPU_GPR(R30)(r9)
1310 std r31, VCPU_GPR(R31)(r9)
1313 mfspr r3, SPRN_SPRG0
1314 mfspr r4, SPRN_SPRG1
1315 mfspr r5, SPRN_SPRG2
1316 mfspr r6, SPRN_SPRG3
1317 std r3, VCPU_SPRG0(r9)
1318 std r4, VCPU_SPRG1(r9)
1319 std r5, VCPU_SPRG2(r9)
1320 std r6, VCPU_SPRG3(r9)
1326 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1329 END_FTR_SECTION_IFCLR(CPU_FTR_TM)
1333 rldimi r8, r0, MSR_TM_LG, 63-MSR_TM_LG
1337 rldicl. r5, r5, 64 - MSR_TS_S_LG, 62
1338 beq 1f /* TM not active in guest. */
1340 li r3, TM_CAUSE_KVM_RESCHED
1342 /* Clear the MSR RI since r1, r13 are all going to be foobar. */
1346 /* All GPRs are volatile at this point. */
1349 /* Temporarily store r13 and r9 so we have some regs to play with */
1352 std r9, PACATMSCRATCH(r13)
1353 ld r9, HSTATE_KVM_VCPU(r13)
1355 /* Get a few more GPRs free. */
1356 std r29, VCPU_GPRS_TM(29)(r9)
1357 std r30, VCPU_GPRS_TM(30)(r9)
1358 std r31, VCPU_GPRS_TM(31)(r9)
1360 /* Save away PPR and DSCR soon so don't run with user values. */
1363 mfspr r30, SPRN_DSCR
1364 ld r29, HSTATE_DSCR(r13)
1365 mtspr SPRN_DSCR, r29
1367 /* Save all but r9, r13 & r29-r31 */
1370 .if (reg != 9) && (reg != 13)
1371 std reg, VCPU_GPRS_TM(reg)(r9)
1375 /* ... now save r13 */
1377 std r4, VCPU_GPRS_TM(13)(r9)
1378 /* ... and save r9 */
1379 ld r4, PACATMSCRATCH(r13)
1380 std r4, VCPU_GPRS_TM(9)(r9)
1382 /* Reload stack pointer and TOC. */
1383 ld r1, HSTATE_HOST_R1(r13)
1386 /* Set MSR RI now we have r1 and r13 back. */
1390 /* Save away checkpinted SPRs. */
1391 std r31, VCPU_PPR_TM(r9)
1392 std r30, VCPU_DSCR_TM(r9)
1398 std r5, VCPU_LR_TM(r9)
1399 stw r6, VCPU_CR_TM(r9)
1400 std r7, VCPU_CTR_TM(r9)
1401 std r8, VCPU_AMR_TM(r9)
1402 std r10, VCPU_TAR_TM(r9)
1404 /* Restore r12 as trap number. */
1405 lwz r12, VCPU_TRAP(r9)
1408 addi r3, r9, VCPU_FPRS_TM
1410 addi r3, r9, VCPU_VRS_TM
1412 mfspr r6, SPRN_VRSAVE
1413 stw r6, VCPU_VRSAVE_TM(r9)
1416 * We need to save these SPRs after the treclaim so that the software
1417 * error code is recorded correctly in the TEXASR. Also the user may
1418 * change these outside of a transaction, so they must always be
1421 mfspr r5, SPRN_TFHAR
1422 mfspr r6, SPRN_TFIAR
1423 mfspr r7, SPRN_TEXASR
1424 std r5, VCPU_TFHAR(r9)
1425 std r6, VCPU_TFIAR(r9)
1426 std r7, VCPU_TEXASR(r9)
1430 /* Increment yield count if they have a VPA */
1431 ld r8, VCPU_VPA(r9) /* do they have a VPA? */
1434 lwz r3, LPPACA_YIELDCOUNT(r8)
1436 stw r3, LPPACA_YIELDCOUNT(r8)
1438 stb r3, VCPU_VPA_DIRTY(r9)
1440 /* Save PMU registers if requested */
1441 /* r8 and cr0.eq are live here */
1443 sldi r3, r3, 31 /* MMCR0_FC (freeze counters) bit */
1444 mfspr r4, SPRN_MMCR0 /* save MMCR0 */
1445 mtspr SPRN_MMCR0, r3 /* freeze all counters, disable ints */
1446 mfspr r6, SPRN_MMCRA
1448 /* On P7, clear MMCRA in order to disable SDAR updates */
1450 mtspr SPRN_MMCRA, r7
1451 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1453 beq 21f /* if no VPA, save PMU stuff anyway */
1454 lbz r7, LPPACA_PMCINUSE(r8)
1455 cmpwi r7, 0 /* did they ask for PMU stuff to be saved? */
1457 std r3, VCPU_MMCR(r9) /* if not, set saved MMCR0 to FC */
1459 21: mfspr r5, SPRN_MMCR1
1462 std r4, VCPU_MMCR(r9)
1463 std r5, VCPU_MMCR + 8(r9)
1464 std r6, VCPU_MMCR + 16(r9)
1465 std r7, VCPU_SIAR(r9)
1466 std r8, VCPU_SDAR(r9)
1474 mfspr r10, SPRN_PMC7
1475 mfspr r11, SPRN_PMC8
1476 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
1477 stw r3, VCPU_PMC(r9)
1478 stw r4, VCPU_PMC + 4(r9)
1479 stw r5, VCPU_PMC + 8(r9)
1480 stw r6, VCPU_PMC + 12(r9)
1481 stw r7, VCPU_PMC + 16(r9)
1482 stw r8, VCPU_PMC + 20(r9)
1484 stw r10, VCPU_PMC + 24(r9)
1485 stw r11, VCPU_PMC + 28(r9)
1486 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
1488 mfspr r4, SPRN_MMCR2
1490 mfspr r6, SPRN_SPMC1
1491 mfspr r7, SPRN_SPMC2
1492 mfspr r8, SPRN_MMCRS
1493 std r4, VCPU_MMCR + 24(r9)
1494 std r5, VCPU_SIER(r9)
1495 stw r6, VCPU_PMC + 24(r9)
1496 stw r7, VCPU_PMC + 28(r9)
1497 std r8, VCPU_MMCR + 32(r9)
1499 mtspr SPRN_MMCRS, r4
1500 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
1508 hdec_soon: /* r12 = trap, r13 = paca */
1511 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
1513 * POWER7 guest -> host partition switch code.
1514 * We don't have to lock against tlbies but we do
1515 * have to coordinate the hardware threads.
1517 /* Increment the threads-exiting-guest count in the 0xff00
1518 bits of vcore->entry_exit_count */
1519 ld r5,HSTATE_KVM_VCORE(r13)
1520 addi r6,r5,VCORE_ENTRY_EXIT
1525 isync /* order stwcx. vs. reading napping_threads */
1528 * At this point we have an interrupt that we have to pass
1529 * up to the kernel or qemu; we can't handle it in real mode.
1530 * Thus we have to do a partition switch, so we have to
1531 * collect the other threads, if we are the first thread
1532 * to take an interrupt. To do this, we set the HDEC to 0,
1533 * which causes an HDEC interrupt in all threads within 2ns
1534 * because the HDEC register is shared between all 4 threads.
1535 * However, we don't need to bother if this is an HDEC
1536 * interrupt, since the other threads will already be on their
1537 * way here in that case.
1539 cmpwi r3,0x100 /* Are we the first here? */
1541 cmpwi r12,BOOK3S_INTERRUPT_HV_DECREMENTER
1547 * Send an IPI to any napping threads, since an HDEC interrupt
1548 * doesn't wake CPUs up from nap.
1550 lwz r3,VCORE_NAPPING_THREADS(r5)
1551 lbz r4,HSTATE_PTID(r13)
1554 andc. r3,r3,r0 /* no sense IPI'ing ourselves */
1556 /* Order entry/exit update vs. IPIs */
1558 mulli r4,r4,PACA_SIZE /* get paca for thread 0 */
1562 ld r8,HSTATE_XICS_PHYS(r6) /* get thread's XICS reg addr */
1565 stbcix r0,r7,r8 /* trigger the IPI */
1567 addi r6,r6,PACA_SIZE
1571 /* Secondary threads wait for primary to do partition switch */
1572 43: ld r5,HSTATE_KVM_VCORE(r13)
1573 ld r4,VCORE_KVM(r5) /* pointer to struct kvm */
1574 lbz r3,HSTATE_PTID(r13)
1578 13: lbz r3,VCORE_IN_GUEST(r5)
1584 /* Primary thread waits for all the secondaries to exit guest */
1585 15: lwz r3,VCORE_ENTRY_EXIT(r5)
1592 /* Primary thread switches back to host partition */
1593 ld r6,KVM_HOST_SDR1(r4)
1594 lwz r7,KVM_HOST_LPID(r4)
1595 li r8,LPID_RSVD /* switch to reserved LPID */
1598 mtspr SPRN_SDR1,r6 /* switch to partition page table */
1603 /* DPDES is shared between threads */
1604 mfspr r7, SPRN_DPDES
1605 std r7, VCORE_DPDES(r5)
1606 /* clear DPDES so we don't get guest doorbells in the host */
1608 mtspr SPRN_DPDES, r8
1609 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
1611 /* Subtract timebase offset from timebase */
1612 ld r8,VCORE_TB_OFFSET(r5)
1615 mftb r6 /* current guest timebase */
1617 mtspr SPRN_TBU40,r8 /* update upper 40 bits */
1618 mftb r7 /* check if lower 24 bits overflowed */
1623 addis r8,r8,0x100 /* if so, increment upper 40 bits */
1627 17: ld r0, VCORE_PCR(r5)
1633 /* Signal secondary CPUs to continue */
1634 stb r0,VCORE_IN_GUEST(r5)
1635 lis r8,0x7fff /* MAX_INT@h */
1638 16: ld r8,KVM_HOST_LPCR(r4)
1644 * PPC970 guest -> host partition switch code.
1645 * We have to lock against concurrent tlbies, and
1646 * we have to flush the whole TLB.
1648 32: ld r5,HSTATE_KVM_VCORE(r13)
1649 ld r4,VCORE_KVM(r5) /* pointer to struct kvm */
1651 /* Take the guest's tlbie_lock */
1652 #ifdef __BIG_ENDIAN__
1653 lwz r8,PACA_LOCK_TOKEN(r13)
1655 lwz r8,PACAPACAINDEX(r13)
1657 addi r3,r4,KVM_TLBIE_LOCK
1665 ld r7,KVM_HOST_LPCR(r4) /* use kvm->arch.host_lpcr for HID4 */
1667 rotldi r0,r0,HID4_LPID5_SH /* all lpid bits in HID4 = 1 */
1671 mtspr SPRN_HID4,r0 /* switch to reserved LPID */
1674 stw r0,0(r3) /* drop guest tlbie_lock */
1676 /* invalidate the whole TLB */
1685 /* take native_tlbie_lock */
1686 ld r3,toc_tlbie_lock@toc(2)
1694 ld r6,KVM_HOST_SDR1(r4)
1695 mtspr SPRN_SDR1,r6 /* switch to host page table */
1697 /* Set up host HID4 value */
1702 stw r0,0(r3) /* drop native_tlbie_lock */
1704 lis r8,0x7fff /* MAX_INT@h */
1707 /* Disable HDEC interrupts */
1710 rldimi r0,r3, HID0_HDICE_SH, 64-HID0_HDICE_SH-1
1720 /* load host SLB entries */
1721 33: ld r8,PACA_SLBSHADOWPTR(r13)
1723 .rept SLB_NUM_BOLTED
1724 ld r5,SLBSHADOW_SAVEAREA(r8)
1725 ld r6,SLBSHADOW_SAVEAREA+8(r8)
1726 andis. r7,r5,SLB_ESID_V@h
1732 /* Unset guest mode */
1733 li r0, KVM_GUEST_MODE_NONE
1734 stb r0, HSTATE_IN_GUEST(r13)
1736 ld r0, 112+PPC_LR_STKOFF(r1)
1742 * Check whether an HDSI is an HPTE not found fault or something else.
1743 * If it is an HPTE not found fault that is due to the guest accessing
1744 * a page that they have mapped but which we have paged out, then
1745 * we continue on with the guest exit path. In all other cases,
1746 * reflect the HDSI to the guest as a DSI.
1750 mfspr r6, SPRN_HDSISR
1751 /* HPTE not found fault or protection fault? */
1752 andis. r0, r6, (DSISR_NOHPTE | DSISR_PROTFAULT)@h
1753 beq 1f /* if not, send it to the guest */
1754 andi. r0, r11, MSR_DR /* data relocation enabled? */
1757 PPC_SLBFEE_DOT(R5, R0) /* if so, look up SLB */
1758 bne 1f /* if no SLB entry found */
1759 4: std r4, VCPU_FAULT_DAR(r9)
1760 stw r6, VCPU_FAULT_DSISR(r9)
1762 /* Search the hash table. */
1763 mr r3, r9 /* vcpu pointer */
1764 li r7, 1 /* data fault */
1765 bl .kvmppc_hpte_hv_fault
1766 ld r9, HSTATE_KVM_VCPU(r13)
1768 ld r11, VCPU_MSR(r9)
1769 li r12, BOOK3S_INTERRUPT_H_DATA_STORAGE
1770 cmpdi r3, 0 /* retry the instruction */
1772 cmpdi r3, -1 /* handle in kernel mode */
1774 cmpdi r3, -2 /* MMIO emulation; need instr word */
1777 /* Synthesize a DSI for the guest */
1778 ld r4, VCPU_FAULT_DAR(r9)
1780 1: mtspr SPRN_DAR, r4
1781 mtspr SPRN_DSISR, r6
1782 mtspr SPRN_SRR0, r10
1783 mtspr SPRN_SRR1, r11
1784 li r10, BOOK3S_INTERRUPT_DATA_STORAGE
1785 bl kvmppc_msr_interrupt
1786 fast_interrupt_c_return:
1787 6: ld r7, VCPU_CTR(r9)
1788 lwz r8, VCPU_XER(r9)
1794 3: ld r5, VCPU_KVM(r9) /* not relocated, use VRMA */
1795 ld r5, KVM_VRMA_SLB_V(r5)
1798 /* If this is for emulated MMIO, load the instruction word */
1799 2: li r8, KVM_INST_FETCH_FAILED /* In case lwz faults */
1801 /* Set guest mode to 'jump over instruction' so if lwz faults
1802 * we'll just continue at the next IP. */
1803 li r0, KVM_GUEST_MODE_SKIP
1804 stb r0, HSTATE_IN_GUEST(r13)
1806 /* Do the access with MSR:DR enabled */
1808 ori r4, r3, MSR_DR /* Enable paging for data */
1813 /* Store the result */
1814 stw r8, VCPU_LAST_INST(r9)
1816 /* Unset guest mode. */
1817 li r0, KVM_GUEST_MODE_HOST_HV
1818 stb r0, HSTATE_IN_GUEST(r13)
1822 * Similarly for an HISI, reflect it to the guest as an ISI unless
1823 * it is an HPTE not found fault for a page that we have paged out.
1826 andis. r0, r11, SRR1_ISI_NOPT@h
1828 andi. r0, r11, MSR_IR /* instruction relocation enabled? */
1831 PPC_SLBFEE_DOT(R5, R0) /* if so, look up SLB */
1832 bne 1f /* if no SLB entry found */
1834 /* Search the hash table. */
1835 mr r3, r9 /* vcpu pointer */
1838 li r7, 0 /* instruction fault */
1839 bl .kvmppc_hpte_hv_fault
1840 ld r9, HSTATE_KVM_VCPU(r13)
1842 ld r11, VCPU_MSR(r9)
1843 li r12, BOOK3S_INTERRUPT_H_INST_STORAGE
1844 cmpdi r3, 0 /* retry the instruction */
1845 beq fast_interrupt_c_return
1846 cmpdi r3, -1 /* handle in kernel mode */
1849 /* Synthesize an ISI for the guest */
1851 1: mtspr SPRN_SRR0, r10
1852 mtspr SPRN_SRR1, r11
1853 li r10, BOOK3S_INTERRUPT_INST_STORAGE
1854 bl kvmppc_msr_interrupt
1855 b fast_interrupt_c_return
1857 3: ld r6, VCPU_KVM(r9) /* not relocated, use VRMA */
1858 ld r5, KVM_VRMA_SLB_V(r6)
1862 * Try to handle an hcall in real mode.
1863 * Returns to the guest if we handle it, or continues on up to
1864 * the kernel if we can't (i.e. if we don't have a handler for
1865 * it, or if the handler returns H_TOO_HARD).
1867 .globl hcall_try_real_mode
1868 hcall_try_real_mode:
1869 ld r3,VCPU_GPR(R3)(r9)
1871 /* sc 1 from userspace - reflect to guest syscall */
1872 bne sc_1_fast_return
1874 cmpldi r3,hcall_real_table_end - hcall_real_table
1876 LOAD_REG_ADDR(r4, hcall_real_table)
1882 mr r3,r9 /* get vcpu pointer */
1883 ld r4,VCPU_GPR(R4)(r9)
1886 beq hcall_real_fallback
1887 ld r4,HSTATE_KVM_VCPU(r13)
1888 std r3,VCPU_GPR(R3)(r4)
1896 li r10, BOOK3S_INTERRUPT_SYSCALL
1897 bl kvmppc_msr_interrupt
1901 /* We've attempted a real mode hcall, but it's punted it back
1902 * to userspace. We need to restore some clobbered volatiles
1903 * before resuming the pass-it-to-qemu path */
1904 hcall_real_fallback:
1905 li r12,BOOK3S_INTERRUPT_SYSCALL
1906 ld r9, HSTATE_KVM_VCPU(r13)
1910 .globl hcall_real_table
1912 .long 0 /* 0 - unused */
1913 .long .kvmppc_h_remove - hcall_real_table
1914 .long .kvmppc_h_enter - hcall_real_table
1915 .long .kvmppc_h_read - hcall_real_table
1916 .long 0 /* 0x10 - H_CLEAR_MOD */
1917 .long 0 /* 0x14 - H_CLEAR_REF */
1918 .long .kvmppc_h_protect - hcall_real_table
1919 .long .kvmppc_h_get_tce - hcall_real_table
1920 .long .kvmppc_h_put_tce - hcall_real_table
1921 .long 0 /* 0x24 - H_SET_SPRG0 */
1922 .long .kvmppc_h_set_dabr - hcall_real_table
1937 #ifdef CONFIG_KVM_XICS
1938 .long .kvmppc_rm_h_eoi - hcall_real_table
1939 .long .kvmppc_rm_h_cppr - hcall_real_table
1940 .long .kvmppc_rm_h_ipi - hcall_real_table
1941 .long 0 /* 0x70 - H_IPOLL */
1942 .long .kvmppc_rm_h_xirr - hcall_real_table
1944 .long 0 /* 0x64 - H_EOI */
1945 .long 0 /* 0x68 - H_CPPR */
1946 .long 0 /* 0x6c - H_IPI */
1947 .long 0 /* 0x70 - H_IPOLL */
1948 .long 0 /* 0x74 - H_XIRR */
1976 .long .kvmppc_h_cede - hcall_real_table
1993 .long .kvmppc_h_bulk_remove - hcall_real_table
1997 .long .kvmppc_h_set_xdabr - hcall_real_table
1998 hcall_real_table_end:
2004 _GLOBAL(kvmppc_h_set_xdabr)
2005 andi. r0, r5, DABRX_USER | DABRX_KERNEL
2007 li r0, DABRX_USER | DABRX_KERNEL | DABRX_BTI
2010 6: li r3, H_PARAMETER
2013 _GLOBAL(kvmppc_h_set_dabr)
2014 li r5, DABRX_USER | DABRX_KERNEL
2018 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
2019 std r4,VCPU_DABR(r3)
2020 stw r5, VCPU_DABRX(r3)
2021 mtspr SPRN_DABRX, r5
2022 /* Work around P7 bug where DABR can get corrupted on mtspr */
2023 1: mtspr SPRN_DABR,r4
2031 /* Emulate H_SET_DABR/X on P8 for the sake of compat mode guests */
2032 2: rlwimi r5, r4, 5, DAWRX_DR | DAWRX_DW
2033 rlwimi r5, r4, 1, DAWRX_WT
2035 std r4, VCPU_DAWR(r3)
2036 std r5, VCPU_DAWRX(r3)
2038 mtspr SPRN_DAWRX, r5
2042 _GLOBAL(kvmppc_h_cede)
2044 std r11,VCPU_MSR(r3)
2046 stb r0,VCPU_CEDED(r3)
2047 sync /* order setting ceded vs. testing prodded */
2048 lbz r5,VCPU_PRODDED(r3)
2050 bne kvm_cede_prodded
2051 li r0,0 /* set trap to 0 to say hcall is handled */
2052 stw r0,VCPU_TRAP(r3)
2054 std r0,VCPU_GPR(R3)(r3)
2056 b kvm_cede_exit /* just send it up to host on 970 */
2057 END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_206)
2060 * Set our bit in the bitmask of napping threads unless all the
2061 * other threads are already napping, in which case we send this
2064 ld r5,HSTATE_KVM_VCORE(r13)
2065 lbz r6,HSTATE_PTID(r13)
2066 lwz r8,VCORE_ENTRY_EXIT(r5)
2070 addi r6,r5,VCORE_NAPPING_THREADS
2078 /* order napping_threads update vs testing entry_exit_count */
2081 stb r0,HSTATE_NAPPING(r13)
2082 lwz r7,VCORE_ENTRY_EXIT(r5)
2084 bge 33f /* another thread already exiting */
2087 * Although not specifically required by the architecture, POWER7
2088 * preserves the following registers in nap mode, even if an SMT mode
2089 * switch occurs: SLB entries, PURR, SPURR, AMOR, UAMOR, AMR, SPRG0-3,
2090 * DAR, DSISR, DABR, DABRX, DSCR, PMCx, MMCRx, SIAR, SDAR.
2092 /* Save non-volatile GPRs */
2093 std r14, VCPU_GPR(R14)(r3)
2094 std r15, VCPU_GPR(R15)(r3)
2095 std r16, VCPU_GPR(R16)(r3)
2096 std r17, VCPU_GPR(R17)(r3)
2097 std r18, VCPU_GPR(R18)(r3)
2098 std r19, VCPU_GPR(R19)(r3)
2099 std r20, VCPU_GPR(R20)(r3)
2100 std r21, VCPU_GPR(R21)(r3)
2101 std r22, VCPU_GPR(R22)(r3)
2102 std r23, VCPU_GPR(R23)(r3)
2103 std r24, VCPU_GPR(R24)(r3)
2104 std r25, VCPU_GPR(R25)(r3)
2105 std r26, VCPU_GPR(R26)(r3)
2106 std r27, VCPU_GPR(R27)(r3)
2107 std r28, VCPU_GPR(R28)(r3)
2108 std r29, VCPU_GPR(R29)(r3)
2109 std r30, VCPU_GPR(R30)(r3)
2110 std r31, VCPU_GPR(R31)(r3)
2116 * Take a nap until a decrementer or external or doobell interrupt
2117 * occurs, with PECE1, PECE0 and PECEDP set in LPCR. Also clear the
2118 * runlatch bit before napping.
2120 mfspr r2, SPRN_CTRLF
2122 mtspr SPRN_CTRLT, r2
2125 stb r0,HSTATE_HWTHREAD_REQ(r13)
2127 ori r5,r5,LPCR_PECE0 | LPCR_PECE1
2129 oris r5,r5,LPCR_PECEDP@h
2130 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
2134 std r0, HSTATE_SCRATCH0(r13)
2136 ld r0, HSTATE_SCRATCH0(r13)
2148 /* get vcpu pointer */
2149 ld r4, HSTATE_KVM_VCPU(r13)
2151 /* Woken by external or decrementer interrupt */
2152 ld r1, HSTATE_HOST_R1(r13)
2154 /* load up FP state */
2158 ld r14, VCPU_GPR(R14)(r4)
2159 ld r15, VCPU_GPR(R15)(r4)
2160 ld r16, VCPU_GPR(R16)(r4)
2161 ld r17, VCPU_GPR(R17)(r4)
2162 ld r18, VCPU_GPR(R18)(r4)
2163 ld r19, VCPU_GPR(R19)(r4)
2164 ld r20, VCPU_GPR(R20)(r4)
2165 ld r21, VCPU_GPR(R21)(r4)
2166 ld r22, VCPU_GPR(R22)(r4)
2167 ld r23, VCPU_GPR(R23)(r4)
2168 ld r24, VCPU_GPR(R24)(r4)
2169 ld r25, VCPU_GPR(R25)(r4)
2170 ld r26, VCPU_GPR(R26)(r4)
2171 ld r27, VCPU_GPR(R27)(r4)
2172 ld r28, VCPU_GPR(R28)(r4)
2173 ld r29, VCPU_GPR(R29)(r4)
2174 ld r30, VCPU_GPR(R30)(r4)
2175 ld r31, VCPU_GPR(R31)(r4)
2177 /* Check the wake reason in SRR1 to see why we got here */
2178 bl kvmppc_check_wake_reason
2180 /* clear our bit in vcore->napping_threads */
2181 34: ld r5,HSTATE_KVM_VCORE(r13)
2182 lbz r7,HSTATE_PTID(r13)
2185 addi r6,r5,VCORE_NAPPING_THREADS
2191 stb r0,HSTATE_NAPPING(r13)
2193 /* See if the wake reason means we need to exit */
2194 stw r12, VCPU_TRAP(r4)
2199 /* see if any other thread is already exiting */
2200 lwz r0,VCORE_ENTRY_EXIT(r5)
2204 b kvmppc_cede_reentry /* if not go back to guest */
2206 /* cede when already previously prodded case */
2209 stb r0,VCPU_PRODDED(r3)
2210 sync /* order testing prodded vs. clearing ceded */
2211 stb r0,VCPU_CEDED(r3)
2215 /* we've ceded but we want to give control to the host */
2217 b hcall_real_fallback
2219 /* Try to handle a machine check in real mode */
2220 machine_check_realmode:
2221 mr r3, r9 /* get vcpu pointer */
2222 bl .kvmppc_realmode_machine_check
2224 cmpdi r3, 0 /* continue exiting from guest? */
2225 ld r9, HSTATE_KVM_VCPU(r13)
2226 li r12, BOOK3S_INTERRUPT_MACHINE_CHECK
2228 /* If not, deliver a machine check. SRR0/1 are already set */
2229 li r10, BOOK3S_INTERRUPT_MACHINE_CHECK
2230 bl kvmppc_msr_interrupt
2231 b fast_interrupt_c_return
2234 * Check the reason we woke from nap, and take appropriate action.
2236 * 0 if nothing needs to be done
2237 * 1 if something happened that needs to be handled by the host
2238 * -1 if there was a guest wakeup (IPI)
2240 * Also sets r12 to the interrupt vector for any interrupt that needs
2241 * to be handled now by the host (0x500 for external interrupt), or zero.
2243 kvmppc_check_wake_reason:
2246 rlwinm r6, r6, 45-31, 0xf /* extract wake reason field (P8) */
2248 rlwinm r6, r6, 45-31, 0xe /* P7 wake reason field is 3 bits */
2249 ALT_FTR_SECTION_END_IFSET(CPU_FTR_ARCH_207S)
2250 cmpwi r6, 8 /* was it an external interrupt? */
2251 li r12, BOOK3S_INTERRUPT_EXTERNAL
2252 beq kvmppc_read_intr /* if so, see what it was */
2255 cmpwi r6, 6 /* was it the decrementer? */
2258 cmpwi r6, 5 /* privileged doorbell? */
2260 cmpwi r6, 3 /* hypervisor doorbell? */
2262 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
2263 li r3, 1 /* anything else, return 1 */
2266 /* hypervisor doorbell */
2267 3: li r12, BOOK3S_INTERRUPT_H_DOORBELL
2272 * Determine what sort of external interrupt is pending (if any).
2274 * 0 if no interrupt is pending
2275 * 1 if an interrupt is pending that needs to be handled by the host
2276 * -1 if there was a guest wakeup IPI (which has now been cleared)
2279 /* see if a host IPI is pending */
2281 lbz r0, HSTATE_HOST_IPI(r13)
2285 /* Now read the interrupt from the ICP */
2286 ld r6, HSTATE_XICS_PHYS(r13)
2291 rlwinm. r3, r0, 0, 0xffffff
2293 beq 1f /* if nothing pending in the ICP */
2295 /* We found something in the ICP...
2297 * If it's not an IPI, stash it in the PACA and return to
2298 * the host, we don't (yet) handle directing real external
2299 * interrupts directly to the guest
2301 cmpwi r3, XICS_IPI /* if there is, is it an IPI? */
2304 /* It's an IPI, clear the MFRR and EOI it */
2307 stbcix r3, r6, r8 /* clear the IPI */
2308 stwcix r0, r6, r7 /* EOI it */
2311 /* We need to re-check host IPI now in case it got set in the
2312 * meantime. If it's clear, we bounce the interrupt to the
2315 lbz r0, HSTATE_HOST_IPI(r13)
2319 /* OK, it's an IPI for us */
2323 42: /* It's not an IPI and it's for the host, stash it in the PACA
2324 * before exit, it will be picked up by the host ICP driver
2326 stw r0, HSTATE_SAVED_XIRR(r13)
2330 43: /* We raced with the host, we need to resend that IPI, bummer */
2332 stbcix r0, r6, r8 /* set the IPI */
2338 * Save away FP, VMX and VSX registers.
2340 * N.B. r30 and r31 are volatile across this function,
2341 * thus it is not callable from C.
2348 #ifdef CONFIG_ALTIVEC
2350 oris r8,r8,MSR_VEC@h
2351 END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2355 oris r8,r8,MSR_VSX@h
2356 END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2360 addi r3,r3,VCPU_FPRS
2362 #ifdef CONFIG_ALTIVEC
2364 addi r3,r31,VCPU_VRS
2366 END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2368 mfspr r6,SPRN_VRSAVE
2369 stw r6,VCPU_VRSAVE(r31)
2374 * Load up FP, VMX and VSX registers
2376 * N.B. r30 and r31 are volatile across this function,
2377 * thus it is not callable from C.
2384 #ifdef CONFIG_ALTIVEC
2386 oris r8,r8,MSR_VEC@h
2387 END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2391 oris r8,r8,MSR_VSX@h
2392 END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2396 addi r3,r4,VCPU_FPRS
2398 #ifdef CONFIG_ALTIVEC
2400 addi r3,r31,VCPU_VRS
2402 END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2404 lwz r7,VCPU_VRSAVE(r31)
2405 mtspr SPRN_VRSAVE,r7
2411 * We come here if we get any exception or interrupt while we are
2412 * executing host real mode code while in guest MMU context.
2413 * For now just spin, but we should do something better.
2415 kvmppc_bad_host_intr:
2419 * This mimics the MSR transition on IRQ delivery. The new guest MSR is taken
2420 * from VCPU_INTR_MSR and is modified based on the required TM state changes.
2421 * r11 has the guest MSR value (in/out)
2422 * r9 has a vcpu pointer (in)
2423 * r0 is used as a scratch register
2425 kvmppc_msr_interrupt:
2426 rldicl r0, r11, 64 - MSR_TS_S_LG, 62
2427 cmpwi r0, 2 /* Check if we are in transactional state.. */
2428 ld r11, VCPU_INTR_MSR(r9)
2430 /* ... if transactional, change to suspended */
2432 1: rldimi r11, r0, MSR_TS_S_LG, 63 - MSR_TS_T_LG