]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - arch/powerpc/kvm/book3s_hv_rmhandlers.S
powerpc/book3s: Add basic infrastructure to handle HMI in Linux.
[mirror_ubuntu-zesty-kernel.git] / arch / powerpc / kvm / book3s_hv_rmhandlers.S
CommitLineData
de56a948
PM
1/*
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.
5 *
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.
10 *
11 * Copyright 2011 Paul Mackerras, IBM Corp. <paulus@au1.ibm.com>
12 *
13 * Derived from book3s_rmhandlers.S and other files, which are:
14 *
15 * Copyright SUSE Linux Products GmbH 2009
16 *
17 * Authors: Alexander Graf <agraf@suse.de>
18 */
19
20#include <asm/ppc_asm.h>
21#include <asm/kvm_asm.h>
22#include <asm/reg.h>
177339d7 23#include <asm/mmu.h>
de56a948 24#include <asm/page.h>
177339d7
PM
25#include <asm/ptrace.h>
26#include <asm/hvcall.h>
de56a948
PM
27#include <asm/asm-offsets.h>
28#include <asm/exception-64s.h>
f0888f70 29#include <asm/kvm_book3s_asm.h>
b4072df4 30#include <asm/mmu-hash64.h>
e4e38121
MN
31#include <asm/tm.h>
32
33#define VCPU_GPRS_TM(reg) (((reg) * ULONG_SIZE) + VCPU_GPR_TM)
de56a948 34
7ffcf8ec
AB
35#ifdef __LITTLE_ENDIAN__
36#error Need to fix lppaca and SLB shadow accesses in little endian mode
37#endif
38
e0b7ec05
PM
39/* Values in HSTATE_NAPPING(r13) */
40#define NAPPING_CEDE 1
41#define NAPPING_NOVCPU 2
42
de56a948 43/*
19ccb76a 44 * Call kvmppc_hv_entry in real mode.
de56a948
PM
45 * Must be called with interrupts hard-disabled.
46 *
47 * Input Registers:
48 *
49 * LR = return address to continue at after eventually re-enabling MMU
50 */
51_GLOBAL(kvmppc_hv_entry_trampoline)
218309b7
PM
52 mflr r0
53 std r0, PPC_LR_STKOFF(r1)
54 stdu r1, -112(r1)
de56a948 55 mfmsr r10
218309b7 56 LOAD_REG_ADDR(r5, kvmppc_call_hv_entry)
de56a948
PM
57 li r0,MSR_RI
58 andc r0,r10,r0
59 li r6,MSR_IR | MSR_DR
60 andc r6,r10,r6
61 mtmsrd r0,1 /* clear RI in MSR */
62 mtsrr0 r5
63 mtsrr1 r6
64 RFI
65
218309b7 66kvmppc_call_hv_entry:
e0b7ec05 67 ld r4, HSTATE_KVM_VCPU(r13)
218309b7
PM
68 bl kvmppc_hv_entry
69
70 /* Back from guest - restore host state and return to caller */
71
eee7ff9d 72BEGIN_FTR_SECTION
218309b7
PM
73 /* Restore host DABR and DABRX */
74 ld r5,HSTATE_DABR(r13)
75 li r6,7
76 mtspr SPRN_DABR,r5
77 mtspr SPRN_DABRX,r6
eee7ff9d 78END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
218309b7
PM
79
80 /* Restore SPRG3 */
9d378dfa
SW
81 ld r3,PACA_SPRG_VDSO(r13)
82 mtspr SPRN_SPRG_VDSO_WRITE,r3
218309b7 83
218309b7
PM
84 /* Reload the host's PMU registers */
85 ld r3, PACALPPACAPTR(r13) /* is the host using the PMU? */
86 lbz r4, LPPACA_PMCINUSE(r3)
87 cmpwi r4, 0
88 beq 23f /* skip if not */
9bc01a9b
PM
89BEGIN_FTR_SECTION
90 ld r3, HSTATE_MMCR(r13)
91 andi. r4, r3, MMCR0_PMAO_SYNC | MMCR0_PMAO
92 cmpwi r4, MMCR0_PMAO
93 beql kvmppc_fix_pmao
94END_FTR_SECTION_IFSET(CPU_FTR_PMAO_BUG)
218309b7
PM
95 lwz r3, HSTATE_PMC(r13)
96 lwz r4, HSTATE_PMC + 4(r13)
97 lwz r5, HSTATE_PMC + 8(r13)
98 lwz r6, HSTATE_PMC + 12(r13)
99 lwz r8, HSTATE_PMC + 16(r13)
100 lwz r9, HSTATE_PMC + 20(r13)
101BEGIN_FTR_SECTION
102 lwz r10, HSTATE_PMC + 24(r13)
103 lwz r11, HSTATE_PMC + 28(r13)
104END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
105 mtspr SPRN_PMC1, r3
106 mtspr SPRN_PMC2, r4
107 mtspr SPRN_PMC3, r5
108 mtspr SPRN_PMC4, r6
109 mtspr SPRN_PMC5, r8
110 mtspr SPRN_PMC6, r9
111BEGIN_FTR_SECTION
112 mtspr SPRN_PMC7, r10
113 mtspr SPRN_PMC8, r11
114END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
115 ld r3, HSTATE_MMCR(r13)
116 ld r4, HSTATE_MMCR + 8(r13)
117 ld r5, HSTATE_MMCR + 16(r13)
72cde5a8
PM
118 ld r6, HSTATE_MMCR + 24(r13)
119 ld r7, HSTATE_MMCR + 32(r13)
218309b7
PM
120 mtspr SPRN_MMCR1, r4
121 mtspr SPRN_MMCRA, r5
72cde5a8
PM
122 mtspr SPRN_SIAR, r6
123 mtspr SPRN_SDAR, r7
124BEGIN_FTR_SECTION
125 ld r8, HSTATE_MMCR + 40(r13)
126 ld r9, HSTATE_MMCR + 48(r13)
127 mtspr SPRN_MMCR2, r8
128 mtspr SPRN_SIER, r9
129END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
218309b7
PM
130 mtspr SPRN_MMCR0, r3
131 isync
13223:
133
e0b7ec05
PM
134 /*
135 * Reload DEC. HDEC interrupts were disabled when
136 * we reloaded the host's LPCR value.
137 */
138 ld r3, HSTATE_DECEXP(r13)
139 mftb r4
140 subf r4, r4, r3
141 mtspr SPRN_DEC, r4
142
218309b7
PM
143 /*
144 * For external and machine check interrupts, we need
145 * to call the Linux handler to process the interrupt.
146 * We do that by jumping to absolute address 0x500 for
147 * external interrupts, or the machine_check_fwnmi label
148 * for machine checks (since firmware might have patched
149 * the vector area at 0x200). The [h]rfid at the end of the
150 * handler will return to the book3s_hv_interrupts.S code.
151 * For other interrupts we do the rfid to get back
152 * to the book3s_hv_interrupts.S code here.
153 */
154 ld r8, 112+PPC_LR_STKOFF(r1)
155 addi r1, r1, 112
156 ld r7, HSTATE_HOST_MSR(r13)
157
158 cmpwi cr1, r12, BOOK3S_INTERRUPT_MACHINE_CHECK
159 cmpwi r12, BOOK3S_INTERRUPT_EXTERNAL
160BEGIN_FTR_SECTION
161 beq 11f
0869b6fd
MS
162 cmpwi cr2, r12, BOOK3S_INTERRUPT_HMI
163 beq cr2, 14f /* HMI check */
218309b7
PM
164END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
165
166 /* RFI into the highmem handler, or branch to interrupt handler */
167 mfmsr r6
168 li r0, MSR_RI
169 andc r6, r6, r0
170 mtmsrd r6, 1 /* Clear RI in MSR */
171 mtsrr0 r8
172 mtsrr1 r7
173 beqa 0x500 /* external interrupt (PPC970) */
174 beq cr1, 13f /* machine check */
175 RFI
176
177 /* On POWER7, we have external interrupts set to use HSRR0/1 */
17811: mtspr SPRN_HSRR0, r8
179 mtspr SPRN_HSRR1, r7
180 ba 0x500
181
18213: b machine_check_fwnmi
183
0869b6fd
MS
18414: mtspr SPRN_HSRR0, r8
185 mtspr SPRN_HSRR1, r7
186 b hmi_exception_after_realmode
187
e0b7ec05
PM
188kvmppc_primary_no_guest:
189 /* We handle this much like a ceded vcpu */
190 /* set our bit in napping_threads */
191 ld r5, HSTATE_KVM_VCORE(r13)
192 lbz r7, HSTATE_PTID(r13)
193 li r0, 1
194 sld r0, r0, r7
195 addi r6, r5, VCORE_NAPPING_THREADS
1961: lwarx r3, 0, r6
197 or r3, r3, r0
198 stwcx. r3, 0, r6
199 bne 1b
200 /* order napping_threads update vs testing entry_exit_count */
201 isync
202 li r12, 0
203 lwz r7, VCORE_ENTRY_EXIT(r5)
204 cmpwi r7, 0x100
205 bge kvm_novcpu_exit /* another thread already exiting */
206 li r3, NAPPING_NOVCPU
207 stb r3, HSTATE_NAPPING(r13)
208 li r3, 1
209 stb r3, HSTATE_HWTHREAD_REQ(r13)
210
211 b kvm_do_nap
212
213kvm_novcpu_wakeup:
214 ld r1, HSTATE_HOST_R1(r13)
215 ld r5, HSTATE_KVM_VCORE(r13)
216 li r0, 0
217 stb r0, HSTATE_NAPPING(r13)
218 stb r0, HSTATE_HWTHREAD_REQ(r13)
219
e3bbbbfa
PM
220 /* check the wake reason */
221 bl kvmppc_check_wake_reason
222
e0b7ec05 223 /* see if any other thread is already exiting */
e0b7ec05
PM
224 lwz r0, VCORE_ENTRY_EXIT(r5)
225 cmpwi r0, 0x100
226 bge kvm_novcpu_exit
227
228 /* clear our bit in napping_threads */
229 lbz r7, HSTATE_PTID(r13)
230 li r0, 1
231 sld r0, r0, r7
232 addi r6, r5, VCORE_NAPPING_THREADS
e3bbbbfa
PM
2334: lwarx r7, 0, r6
234 andc r7, r7, r0
235 stwcx. r7, 0, r6
e0b7ec05
PM
236 bne 4b
237
e3bbbbfa 238 /* See if the wake reason means we need to exit */
e0b7ec05
PM
239 cmpdi r3, 0
240 bge kvm_novcpu_exit
e0b7ec05
PM
241
242 /* Got an IPI but other vcpus aren't yet exiting, must be a latecomer */
243 ld r4, HSTATE_KVM_VCPU(r13)
244 cmpdi r4, 0
245 bne kvmppc_got_guest
246
247kvm_novcpu_exit:
248 b hdec_soon
249
371fefd6 250/*
e0b7ec05 251 * We come in here when wakened from nap mode.
371fefd6
PM
252 * Relocation is off and most register values are lost.
253 * r13 points to the PACA.
254 */
255 .globl kvm_start_guest
256kvm_start_guest:
fd17dc7b
PM
257
258 /* Set runlatch bit the minute you wake up from nap */
259 mfspr r1, SPRN_CTRLF
260 ori r1, r1, 1
261 mtspr SPRN_CTRLT, r1
262
19ccb76a
PM
263 ld r2,PACATOC(r13)
264
f0888f70
PM
265 li r0,KVM_HWTHREAD_IN_KVM
266 stb r0,HSTATE_HWTHREAD_STATE(r13)
371fefd6 267
f0888f70
PM
268 /* NV GPR values from power7_idle() will no longer be valid */
269 li r0,1
270 stb r0,PACA_NAPSTATELOST(r13)
371fefd6 271
4619ac88
PM
272 /* were we napping due to cede? */
273 lbz r0,HSTATE_NAPPING(r13)
e0b7ec05
PM
274 cmpwi r0,NAPPING_CEDE
275 beq kvm_end_cede
276 cmpwi r0,NAPPING_NOVCPU
277 beq kvm_novcpu_wakeup
278
279 ld r1,PACAEMERGSP(r13)
280 subi r1,r1,STACK_FRAME_OVERHEAD
4619ac88
PM
281
282 /*
283 * We weren't napping due to cede, so this must be a secondary
284 * thread being woken up to run a guest, or being woken up due
285 * to a stray IPI. (Or due to some machine check or hypervisor
286 * maintenance interrupt while the core is in KVM.)
287 */
f0888f70
PM
288
289 /* Check the wake reason in SRR1 to see why we got here */
e3bbbbfa
PM
290 bl kvmppc_check_wake_reason
291 cmpdi r3, 0
292 bge kvm_no_guest
371fefd6 293
4619ac88 294 /* get vcpu pointer, NULL if we have no vcpu to run */
7b444c67
PM
295 ld r4,HSTATE_KVM_VCPU(r13)
296 cmpdi r4,0
f0888f70 297 /* if we have no vcpu to run, go back to sleep */
7b444c67 298 beq kvm_no_guest
f0888f70 299
e0b7ec05 300 /* Set HSTATE_DSCR(r13) to something sensible */
1739ea9e 301 ld r6, PACA_DSCR(r13)
e0b7ec05 302 std r6, HSTATE_DSCR(r13)
2fde6d20 303
e0b7ec05 304 bl kvmppc_hv_entry
218309b7
PM
305
306 /* Back from the guest, go back to nap */
307 /* Clear our vcpu pointer so we don't come back in early */
308 li r0, 0
309 std r0, HSTATE_KVM_VCPU(r13)
f019b7ad
PM
310 /*
311 * Make sure we clear HSTATE_KVM_VCPU(r13) before incrementing
312 * the nap_count, because once the increment to nap_count is
313 * visible we could be given another vcpu.
314 */
218309b7 315 lwsync
218309b7
PM
316
317 /* increment the nap count and then go to nap mode */
318 ld r4, HSTATE_KVM_VCORE(r13)
319 addi r4, r4, VCORE_NAP_COUNT
218309b7
PM
32051: lwarx r3, 0, r4
321 addi r3, r3, 1
322 stwcx. r3, 0, r4
323 bne 51b
324
325kvm_no_guest:
326 li r0, KVM_HWTHREAD_IN_NAP
327 stb r0, HSTATE_HWTHREAD_STATE(r13)
e0b7ec05 328kvm_do_nap:
582b910e
PM
329 /* Clear the runlatch bit before napping */
330 mfspr r2, SPRN_CTRLF
331 clrrdi r2, r2, 1
332 mtspr SPRN_CTRLT, r2
333
218309b7
PM
334 li r3, LPCR_PECE0
335 mfspr r4, SPRN_LPCR
336 rlwimi r4, r3, 0, LPCR_PECE0 | LPCR_PECE1
337 mtspr SPRN_LPCR, r4
338 isync
339 std r0, HSTATE_SCRATCH0(r13)
340 ptesync
341 ld r0, HSTATE_SCRATCH0(r13)
3421: cmpd r0, r0
343 bne 1b
344 nap
345 b .
346
347/******************************************************************************
348 * *
349 * Entry code *
350 * *
351 *****************************************************************************/
352
de56a948
PM
353.global kvmppc_hv_entry
354kvmppc_hv_entry:
355
356 /* Required state:
357 *
e0b7ec05 358 * R4 = vcpu pointer (or NULL)
de56a948
PM
359 * MSR = ~IR|DR
360 * R13 = PACA
361 * R1 = host R1
362 * all other volatile GPRS = free
363 */
364 mflr r0
218309b7
PM
365 std r0, PPC_LR_STKOFF(r1)
366 stdu r1, -112(r1)
de56a948 367
de56a948
PM
368 /* Save R1 in the PACA */
369 std r1, HSTATE_HOST_R1(r13)
370
44a3add8
PM
371 li r6, KVM_GUEST_MODE_HOST_HV
372 stb r6, HSTATE_IN_GUEST(r13)
373
de56a948
PM
374 /* Clear out SLB */
375 li r6,0
376 slbmte r6,r6
377 slbia
378 ptesync
379
9e368f29
PM
380BEGIN_FTR_SECTION
381 b 30f
382END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
383 /*
384 * POWER7 host -> guest partition switch code.
385 * We don't have to lock against concurrent tlbies,
386 * but we do have to coordinate across hardware threads.
387 */
371fefd6
PM
388 /* Increment entry count iff exit count is zero. */
389 ld r5,HSTATE_KVM_VCORE(r13)
390 addi r9,r5,VCORE_ENTRY_EXIT
39121: lwarx r3,0,r9
392 cmpwi r3,0x100 /* any threads starting to exit? */
393 bge secondary_too_late /* if so we're too late to the party */
394 addi r3,r3,1
395 stwcx. r3,0,r9
396 bne 21b
397
398 /* Primary thread switches to guest partition. */
e0b7ec05
PM
399 ld r9,VCORE_KVM(r5) /* pointer to struct kvm */
400 lbz r6,HSTATE_PTID(r13)
371fefd6
PM
401 cmpwi r6,0
402 bne 20f
de56a948
PM
403 ld r6,KVM_SDR1(r9)
404 lwz r7,KVM_LPID(r9)
405 li r0,LPID_RSVD /* switch to reserved LPID */
406 mtspr SPRN_LPID,r0
407 ptesync
408 mtspr SPRN_SDR1,r6 /* switch to partition page table */
409 mtspr SPRN_LPID,r7
410 isync
1b400ba0
PM
411
412 /* See if we need to flush the TLB */
413 lhz r6,PACAPACAINDEX(r13) /* test_bit(cpu, need_tlb_flush) */
414 clrldi r7,r6,64-6 /* extract bit number (6 bits) */
415 srdi r6,r6,6 /* doubleword number */
416 sldi r6,r6,3 /* address offset */
417 add r6,r6,r9
418 addi r6,r6,KVM_NEED_FLUSH /* dword in kvm->arch.need_tlb_flush */
371fefd6 419 li r0,1
1b400ba0
PM
420 sld r0,r0,r7
421 ld r7,0(r6)
422 and. r7,r7,r0
423 beq 22f
42423: ldarx r7,0,r6 /* if set, clear the bit */
425 andc r7,r7,r0
426 stdcx. r7,0,r6
427 bne 23b
ca252055
PM
428 /* Flush the TLB of any entries for this LPID */
429 /* use arch 2.07S as a proxy for POWER8 */
430BEGIN_FTR_SECTION
431 li r6,512 /* POWER8 has 512 sets */
432FTR_SECTION_ELSE
433 li r6,128 /* POWER7 has 128 sets */
434ALT_FTR_SECTION_END_IFSET(CPU_FTR_ARCH_207S)
1b400ba0
PM
435 mtctr r6
436 li r7,0x800 /* IS field = 0b10 */
437 ptesync
43828: tlbiel r7
439 addi r7,r7,0x1000
440 bdnz 28b
441 ptesync
442
93b0f4dc
PM
443 /* Add timebase offset onto timebase */
44422: ld r8,VCORE_TB_OFFSET(r5)
445 cmpdi r8,0
446 beq 37f
447 mftb r6 /* current host timebase */
448 add r8,r8,r6
449 mtspr SPRN_TBU40,r8 /* update upper 40 bits */
450 mftb r7 /* check if lower 24 bits overflowed */
451 clrldi r6,r6,40
452 clrldi r7,r7,40
453 cmpld r7,r6
454 bge 37f
455 addis r8,r8,0x100 /* if so, increment upper 40 bits */
456 mtspr SPRN_TBU40,r8
457
388cc6e1
PM
458 /* Load guest PCR value to select appropriate compat mode */
45937: ld r7, VCORE_PCR(r5)
460 cmpdi r7, 0
461 beq 38f
462 mtspr SPRN_PCR, r7
46338:
b005255e
MN
464
465BEGIN_FTR_SECTION
466 /* DPDES is shared between threads */
467 ld r8, VCORE_DPDES(r5)
468 mtspr SPRN_DPDES, r8
469END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
470
388cc6e1 471 li r0,1
371fefd6
PM
472 stb r0,VCORE_IN_GUEST(r5) /* signal secondaries to continue */
473 b 10f
474
475 /* Secondary threads wait for primary to have done partition switch */
47620: lbz r0,VCORE_IN_GUEST(r5)
477 cmpwi r0,0
478 beq 20b
aa04b4cc 479
19ccb76a 480 /* Set LPCR and RMOR. */
a0144e2a 48110: ld r8,VCORE_LPCR(r5)
19ccb76a 482 mtspr SPRN_LPCR,r8
aa04b4cc
PM
483 ld r8,KVM_RMOR(r9)
484 mtspr SPRN_RMOR,r8
de56a948
PM
485 isync
486
487 /* Check if HDEC expires soon */
488 mfspr r3,SPRN_HDEC
e0b7ec05 489 cmpwi r3,512 /* 1 microsecond */
de56a948 490 li r12,BOOK3S_INTERRUPT_HV_DECREMENTER
de56a948 491 blt hdec_soon
9e368f29
PM
492 b 31f
493
494 /*
495 * PPC970 host -> guest partition switch code.
496 * We have to lock against concurrent tlbies,
497 * using native_tlbie_lock to lock against host tlbies
498 * and kvm->arch.tlbie_lock to lock against guest tlbies.
499 * We also have to invalidate the TLB since its
500 * entries aren't tagged with the LPID.
501 */
e0b7ec05
PM
50230: ld r5,HSTATE_KVM_VCORE(r13)
503 ld r9,VCORE_KVM(r5) /* pointer to struct kvm */
9e368f29
PM
504
505 /* first take native_tlbie_lock */
506 .section ".toc","aw"
507toc_tlbie_lock:
508 .tc native_tlbie_lock[TC],native_tlbie_lock
509 .previous
510 ld r3,toc_tlbie_lock@toc(2)
54bb7f4b 511#ifdef __BIG_ENDIAN__
9e368f29 512 lwz r8,PACA_LOCK_TOKEN(r13)
54bb7f4b
AB
513#else
514 lwz r8,PACAPACAINDEX(r13)
515#endif
9e368f29
PM
51624: lwarx r0,0,r3
517 cmpwi r0,0
518 bne 24b
519 stwcx. r8,0,r3
520 bne 24b
521 isync
522
a0144e2a
PM
523 ld r5,HSTATE_KVM_VCORE(r13)
524 ld r7,VCORE_LPCR(r5) /* use vcore->lpcr to store HID4 */
9e368f29
PM
525 li r0,0x18f
526 rotldi r0,r0,HID4_LPID5_SH /* all lpid bits in HID4 = 1 */
527 or r0,r7,r0
528 ptesync
529 sync
530 mtspr SPRN_HID4,r0 /* switch to reserved LPID */
531 isync
532 li r0,0
533 stw r0,0(r3) /* drop native_tlbie_lock */
534
535 /* invalidate the whole TLB */
536 li r0,256
537 mtctr r0
538 li r6,0
53925: tlbiel r6
540 addi r6,r6,0x1000
541 bdnz 25b
542 ptesync
543
544 /* Take the guest's tlbie_lock */
545 addi r3,r9,KVM_TLBIE_LOCK
54624: lwarx r0,0,r3
547 cmpwi r0,0
548 bne 24b
549 stwcx. r8,0,r3
550 bne 24b
551 isync
552 ld r6,KVM_SDR1(r9)
553 mtspr SPRN_SDR1,r6 /* switch to partition page table */
554
555 /* Set up HID4 with the guest's LPID etc. */
556 sync
557 mtspr SPRN_HID4,r7
558 isync
559
560 /* drop the guest's tlbie_lock */
561 li r0,0
562 stw r0,0(r3)
563
564 /* Check if HDEC expires soon */
565 mfspr r3,SPRN_HDEC
566 cmpwi r3,10
567 li r12,BOOK3S_INTERRUPT_HV_DECREMENTER
9e368f29
PM
568 blt hdec_soon
569
570 /* Enable HDEC interrupts */
571 mfspr r0,SPRN_HID0
572 li r3,1
573 rldimi r0,r3, HID0_HDICE_SH, 64-HID0_HDICE_SH-1
574 sync
575 mtspr SPRN_HID0,r0
576 mfspr r0,SPRN_HID0
577 mfspr r0,SPRN_HID0
578 mfspr r0,SPRN_HID0
579 mfspr r0,SPRN_HID0
580 mfspr r0,SPRN_HID0
581 mfspr r0,SPRN_HID0
e0b7ec05
PM
58231:
583 /* Do we have a guest vcpu to run? */
584 cmpdi r4, 0
585 beq kvmppc_primary_no_guest
586kvmppc_got_guest:
de56a948
PM
587
588 /* Load up guest SLB entries */
e0b7ec05 589 lwz r5,VCPU_SLB_MAX(r4)
de56a948
PM
590 cmpwi r5,0
591 beq 9f
592 mtctr r5
593 addi r6,r4,VCPU_SLB
5941: ld r8,VCPU_SLB_E(r6)
595 ld r9,VCPU_SLB_V(r6)
596 slbmte r9,r8
597 addi r6,r6,VCPU_SLB_SIZE
598 bdnz 1b
5999:
e0b7ec05
PM
600 /* Increment yield count if they have a VPA */
601 ld r3, VCPU_VPA(r4)
602 cmpdi r3, 0
603 beq 25f
604 lwz r5, LPPACA_YIELDCOUNT(r3)
605 addi r5, r5, 1
606 stw r5, LPPACA_YIELDCOUNT(r3)
607 li r6, 1
608 stb r6, VCPU_VPA_DIRTY(r4)
60925:
610
611BEGIN_FTR_SECTION
612 /* Save purr/spurr */
613 mfspr r5,SPRN_PURR
614 mfspr r6,SPRN_SPURR
615 std r5,HSTATE_PURR(r13)
616 std r6,HSTATE_SPURR(r13)
617 ld r7,VCPU_PURR(r4)
618 ld r8,VCPU_SPURR(r4)
619 mtspr SPRN_PURR,r7
620 mtspr SPRN_SPURR,r8
621END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
622
623BEGIN_FTR_SECTION
624 /* Set partition DABR */
625 /* Do this before re-enabling PMU to avoid P7 DABR corruption bug */
8563bf52 626 lwz r5,VCPU_DABRX(r4)
e0b7ec05
PM
627 ld r6,VCPU_DABR(r4)
628 mtspr SPRN_DABRX,r5
629 mtspr SPRN_DABR,r6
630 BEGIN_FTR_SECTION_NESTED(89)
631 isync
632 END_FTR_SECTION_NESTED(CPU_FTR_ARCH_206, CPU_FTR_ARCH_206, 89)
633END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
634
e4e38121
MN
635#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
636BEGIN_FTR_SECTION
637 b skip_tm
638END_FTR_SECTION_IFCLR(CPU_FTR_TM)
639
640 /* Turn on TM/FP/VSX/VMX so we can restore them. */
641 mfmsr r5
642 li r6, MSR_TM >> 32
643 sldi r6, r6, 32
644 or r5, r5, r6
645 ori r5, r5, MSR_FP
646 oris r5, r5, (MSR_VEC | MSR_VSX)@h
647 mtmsrd r5
648
649 /*
650 * The user may change these outside of a transaction, so they must
651 * always be context switched.
652 */
653 ld r5, VCPU_TFHAR(r4)
654 ld r6, VCPU_TFIAR(r4)
655 ld r7, VCPU_TEXASR(r4)
656 mtspr SPRN_TFHAR, r5
657 mtspr SPRN_TFIAR, r6
658 mtspr SPRN_TEXASR, r7
659
660 ld r5, VCPU_MSR(r4)
661 rldicl. r5, r5, 64 - MSR_TS_S_LG, 62
662 beq skip_tm /* TM not active in guest */
663
664 /* Make sure the failure summary is set, otherwise we'll program check
665 * when we trechkpt. It's possible that this might have been not set
666 * on a kvmppc_set_one_reg() call but we shouldn't let this crash the
667 * host.
668 */
669 oris r7, r7, (TEXASR_FS)@h
670 mtspr SPRN_TEXASR, r7
671
672 /*
673 * We need to load up the checkpointed state for the guest.
674 * We need to do this early as it will blow away any GPRs, VSRs and
675 * some SPRs.
676 */
677
678 mr r31, r4
679 addi r3, r31, VCPU_FPRS_TM
680 bl .load_fp_state
681 addi r3, r31, VCPU_VRS_TM
682 bl .load_vr_state
683 mr r4, r31
684 lwz r7, VCPU_VRSAVE_TM(r4)
685 mtspr SPRN_VRSAVE, r7
686
687 ld r5, VCPU_LR_TM(r4)
688 lwz r6, VCPU_CR_TM(r4)
689 ld r7, VCPU_CTR_TM(r4)
690 ld r8, VCPU_AMR_TM(r4)
691 ld r9, VCPU_TAR_TM(r4)
692 mtlr r5
693 mtcr r6
694 mtctr r7
695 mtspr SPRN_AMR, r8
696 mtspr SPRN_TAR, r9
697
698 /*
699 * Load up PPR and DSCR values but don't put them in the actual SPRs
700 * till the last moment to avoid running with userspace PPR and DSCR for
701 * too long.
702 */
703 ld r29, VCPU_DSCR_TM(r4)
704 ld r30, VCPU_PPR_TM(r4)
705
706 std r2, PACATMSCRATCH(r13) /* Save TOC */
707
708 /* Clear the MSR RI since r1, r13 are all going to be foobar. */
709 li r5, 0
710 mtmsrd r5, 1
711
712 /* Load GPRs r0-r28 */
713 reg = 0
714 .rept 29
715 ld reg, VCPU_GPRS_TM(reg)(r31)
716 reg = reg + 1
717 .endr
718
719 mtspr SPRN_DSCR, r29
720 mtspr SPRN_PPR, r30
721
722 /* Load final GPRs */
723 ld 29, VCPU_GPRS_TM(29)(r31)
724 ld 30, VCPU_GPRS_TM(30)(r31)
725 ld 31, VCPU_GPRS_TM(31)(r31)
726
727 /* TM checkpointed state is now setup. All GPRs are now volatile. */
728 TRECHKPT
729
730 /* Now let's get back the state we need. */
731 HMT_MEDIUM
732 GET_PACA(r13)
733 ld r29, HSTATE_DSCR(r13)
734 mtspr SPRN_DSCR, r29
735 ld r4, HSTATE_KVM_VCPU(r13)
736 ld r1, HSTATE_HOST_R1(r13)
737 ld r2, PACATMSCRATCH(r13)
738
739 /* Set the MSR RI since we have our registers back. */
740 li r5, MSR_RI
741 mtmsrd r5, 1
742skip_tm:
743#endif
744
e0b7ec05
PM
745 /* Load guest PMU registers */
746 /* R4 is live here (vcpu pointer) */
747 li r3, 1
748 sldi r3, r3, 31 /* MMCR0_FC (freeze counters) bit */
749 mtspr SPRN_MMCR0, r3 /* freeze all counters, disable ints */
750 isync
9bc01a9b
PM
751BEGIN_FTR_SECTION
752 ld r3, VCPU_MMCR(r4)
753 andi. r5, r3, MMCR0_PMAO_SYNC | MMCR0_PMAO
754 cmpwi r5, MMCR0_PMAO
755 beql kvmppc_fix_pmao
756END_FTR_SECTION_IFSET(CPU_FTR_PMAO_BUG)
e0b7ec05
PM
757 lwz r3, VCPU_PMC(r4) /* always load up guest PMU registers */
758 lwz r5, VCPU_PMC + 4(r4) /* to prevent information leak */
759 lwz r6, VCPU_PMC + 8(r4)
760 lwz r7, VCPU_PMC + 12(r4)
761 lwz r8, VCPU_PMC + 16(r4)
762 lwz r9, VCPU_PMC + 20(r4)
763BEGIN_FTR_SECTION
764 lwz r10, VCPU_PMC + 24(r4)
765 lwz r11, VCPU_PMC + 28(r4)
766END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
767 mtspr SPRN_PMC1, r3
768 mtspr SPRN_PMC2, r5
769 mtspr SPRN_PMC3, r6
770 mtspr SPRN_PMC4, r7
771 mtspr SPRN_PMC5, r8
772 mtspr SPRN_PMC6, r9
773BEGIN_FTR_SECTION
774 mtspr SPRN_PMC7, r10
775 mtspr SPRN_PMC8, r11
776END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
777 ld r3, VCPU_MMCR(r4)
778 ld r5, VCPU_MMCR + 8(r4)
779 ld r6, VCPU_MMCR + 16(r4)
780 ld r7, VCPU_SIAR(r4)
781 ld r8, VCPU_SDAR(r4)
782 mtspr SPRN_MMCR1, r5
783 mtspr SPRN_MMCRA, r6
784 mtspr SPRN_SIAR, r7
785 mtspr SPRN_SDAR, r8
b005255e
MN
786BEGIN_FTR_SECTION
787 ld r5, VCPU_MMCR + 24(r4)
788 ld r6, VCPU_SIER(r4)
789 lwz r7, VCPU_PMC + 24(r4)
790 lwz r8, VCPU_PMC + 28(r4)
791 ld r9, VCPU_MMCR + 32(r4)
792 mtspr SPRN_MMCR2, r5
793 mtspr SPRN_SIER, r6
794 mtspr SPRN_SPMC1, r7
795 mtspr SPRN_SPMC2, r8
796 mtspr SPRN_MMCRS, r9
797END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
e0b7ec05
PM
798 mtspr SPRN_MMCR0, r3
799 isync
800
801 /* Load up FP, VMX and VSX registers */
802 bl kvmppc_load_fp
803
804 ld r14, VCPU_GPR(R14)(r4)
805 ld r15, VCPU_GPR(R15)(r4)
806 ld r16, VCPU_GPR(R16)(r4)
807 ld r17, VCPU_GPR(R17)(r4)
808 ld r18, VCPU_GPR(R18)(r4)
809 ld r19, VCPU_GPR(R19)(r4)
810 ld r20, VCPU_GPR(R20)(r4)
811 ld r21, VCPU_GPR(R21)(r4)
812 ld r22, VCPU_GPR(R22)(r4)
813 ld r23, VCPU_GPR(R23)(r4)
814 ld r24, VCPU_GPR(R24)(r4)
815 ld r25, VCPU_GPR(R25)(r4)
816 ld r26, VCPU_GPR(R26)(r4)
817 ld r27, VCPU_GPR(R27)(r4)
818 ld r28, VCPU_GPR(R28)(r4)
819 ld r29, VCPU_GPR(R29)(r4)
820 ld r30, VCPU_GPR(R30)(r4)
821 ld r31, VCPU_GPR(R31)(r4)
822
823BEGIN_FTR_SECTION
824 /* Switch DSCR to guest value */
825 ld r5, VCPU_DSCR(r4)
826 mtspr SPRN_DSCR, r5
827END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
828
b005255e
MN
829BEGIN_FTR_SECTION
830 /* Skip next section on POWER7 or PPC970 */
831 b 8f
832END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
833 /* Turn on TM so we can access TFHAR/TFIAR/TEXASR */
834 mfmsr r8
835 li r0, 1
836 rldimi r8, r0, MSR_TM_LG, 63-MSR_TM_LG
837 mtmsrd r8
838
839 /* Load up POWER8-specific registers */
840 ld r5, VCPU_IAMR(r4)
841 lwz r6, VCPU_PSPB(r4)
842 ld r7, VCPU_FSCR(r4)
843 mtspr SPRN_IAMR, r5
844 mtspr SPRN_PSPB, r6
845 mtspr SPRN_FSCR, r7
846 ld r5, VCPU_DAWR(r4)
847 ld r6, VCPU_DAWRX(r4)
848 ld r7, VCPU_CIABR(r4)
849 ld r8, VCPU_TAR(r4)
850 mtspr SPRN_DAWR, r5
851 mtspr SPRN_DAWRX, r6
852 mtspr SPRN_CIABR, r7
853 mtspr SPRN_TAR, r8
854 ld r5, VCPU_IC(r4)
855 ld r6, VCPU_VTB(r4)
856 mtspr SPRN_IC, r5
857 mtspr SPRN_VTB, r6
7b490411 858 ld r8, VCPU_EBBHR(r4)
b005255e
MN
859 mtspr SPRN_EBBHR, r8
860 ld r5, VCPU_EBBRR(r4)
861 ld r6, VCPU_BESCR(r4)
862 ld r7, VCPU_CSIGR(r4)
863 ld r8, VCPU_TACR(r4)
864 mtspr SPRN_EBBRR, r5
865 mtspr SPRN_BESCR, r6
866 mtspr SPRN_CSIGR, r7
867 mtspr SPRN_TACR, r8
868 ld r5, VCPU_TCSCR(r4)
869 ld r6, VCPU_ACOP(r4)
870 lwz r7, VCPU_GUEST_PID(r4)
871 ld r8, VCPU_WORT(r4)
872 mtspr SPRN_TCSCR, r5
873 mtspr SPRN_ACOP, r6
874 mtspr SPRN_PID, r7
875 mtspr SPRN_WORT, r8
8768:
877
e0b7ec05
PM
878 /*
879 * Set the decrementer to the guest decrementer.
880 */
881 ld r8,VCPU_DEC_EXPIRES(r4)
c5fb80d3
PM
882 /* r8 is a host timebase value here, convert to guest TB */
883 ld r5,HSTATE_KVM_VCORE(r13)
884 ld r6,VCORE_TB_OFFSET(r5)
885 add r8,r8,r6
e0b7ec05
PM
886 mftb r7
887 subf r3,r7,r8
888 mtspr SPRN_DEC,r3
889 stw r3,VCPU_DEC(r4)
890
891 ld r5, VCPU_SPRG0(r4)
892 ld r6, VCPU_SPRG1(r4)
893 ld r7, VCPU_SPRG2(r4)
894 ld r8, VCPU_SPRG3(r4)
895 mtspr SPRN_SPRG0, r5
896 mtspr SPRN_SPRG1, r6
897 mtspr SPRN_SPRG2, r7
898 mtspr SPRN_SPRG3, r8
899
900 /* Load up DAR and DSISR */
901 ld r5, VCPU_DAR(r4)
902 lwz r6, VCPU_DSISR(r4)
903 mtspr SPRN_DAR, r5
904 mtspr SPRN_DSISR, r6
905
906BEGIN_FTR_SECTION
907 /* Restore AMR and UAMOR, set AMOR to all 1s */
908 ld r5,VCPU_AMR(r4)
909 ld r6,VCPU_UAMOR(r4)
910 li r7,-1
911 mtspr SPRN_AMR,r5
912 mtspr SPRN_UAMOR,r6
913 mtspr SPRN_AMOR,r7
914END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
de56a948
PM
915
916 /* Restore state of CTRL run bit; assume 1 on entry */
917 lwz r5,VCPU_CTRL(r4)
918 andi. r5,r5,1
919 bne 4f
920 mfspr r6,SPRN_CTRLF
921 clrrdi r6,r6,1
922 mtspr SPRN_CTRLT,r6
9234:
924 ld r6, VCPU_CTR(r4)
925 lwz r7, VCPU_XER(r4)
926
927 mtctr r6
928 mtxer r7
929
e3bbbbfa 930kvmppc_cede_reentry: /* r4 = vcpu, r13 = paca */
4619ac88
PM
931 ld r10, VCPU_PC(r4)
932 ld r11, VCPU_MSR(r4)
de56a948
PM
933 ld r6, VCPU_SRR0(r4)
934 ld r7, VCPU_SRR1(r4)
e3bbbbfa
PM
935 mtspr SPRN_SRR0, r6
936 mtspr SPRN_SRR1, r7
de56a948 937
e3bbbbfa 938deliver_guest_interrupt:
4619ac88 939 /* r11 = vcpu->arch.msr & ~MSR_HV */
de56a948
PM
940 rldicl r11, r11, 63 - MSR_HV_LG, 1
941 rotldi r11, r11, 1 + MSR_HV_LG
942 ori r11, r11, MSR_ME
943
19ccb76a 944 /* Check if we can deliver an external or decrementer interrupt now */
e3bbbbfa
PM
945 ld r0, VCPU_PENDING_EXC(r4)
946 rldicl r0, r0, 64 - BOOK3S_IRQPRIO_EXTERNAL_LEVEL, 63
947 cmpdi cr1, r0, 0
948 andi. r8, r11, MSR_EE
19ccb76a 949BEGIN_FTR_SECTION
e3bbbbfa
PM
950 mfspr r8, SPRN_LPCR
951 /* Insert EXTERNAL_LEVEL bit into LPCR at the MER bit position */
952 rldimi r8, r0, LPCR_MER_SH, 63 - LPCR_MER_SH
953 mtspr SPRN_LPCR, r8
19ccb76a
PM
954 isync
955END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
956 beq 5f
e3bbbbfa
PM
957 li r0, BOOK3S_INTERRUPT_EXTERNAL
958 bne cr1, 12f
959 mfspr r0, SPRN_DEC
960 cmpwi r0, 0
961 li r0, BOOK3S_INTERRUPT_DECREMENTER
962 bge 5f
19ccb76a 963
e3bbbbfa 96412: mtspr SPRN_SRR0, r10
19ccb76a 965 mr r10,r0
e3bbbbfa 966 mtspr SPRN_SRR1, r11
e4e38121
MN
967 mr r9, r4
968 bl kvmppc_msr_interrupt
e3bbbbfa 9695:
19ccb76a 970
27025a60
LPF
971/*
972 * Required state:
973 * R4 = vcpu
974 * R10: value for HSRR0
975 * R11: value for HSRR1
976 * R13 = PACA
977 */
de56a948 978fast_guest_return:
4619ac88
PM
979 li r0,0
980 stb r0,VCPU_CEDED(r4) /* cancel cede */
de56a948
PM
981 mtspr SPRN_HSRR0,r10
982 mtspr SPRN_HSRR1,r11
983
984 /* Activate guest mode, so faults get handled by KVM */
44a3add8 985 li r9, KVM_GUEST_MODE_GUEST_HV
de56a948
PM
986 stb r9, HSTATE_IN_GUEST(r13)
987
988 /* Enter guest */
989
0acb9111
PM
990BEGIN_FTR_SECTION
991 ld r5, VCPU_CFAR(r4)
992 mtspr SPRN_CFAR, r5
993END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
4b8473c9
PM
994BEGIN_FTR_SECTION
995 ld r0, VCPU_PPR(r4)
996END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
0acb9111 997
de56a948
PM
998 ld r5, VCPU_LR(r4)
999 lwz r6, VCPU_CR(r4)
1000 mtlr r5
1001 mtcr r6
1002
c75df6f9
MN
1003 ld r1, VCPU_GPR(R1)(r4)
1004 ld r2, VCPU_GPR(R2)(r4)
1005 ld r3, VCPU_GPR(R3)(r4)
1006 ld r5, VCPU_GPR(R5)(r4)
1007 ld r6, VCPU_GPR(R6)(r4)
1008 ld r7, VCPU_GPR(R7)(r4)
1009 ld r8, VCPU_GPR(R8)(r4)
1010 ld r9, VCPU_GPR(R9)(r4)
1011 ld r10, VCPU_GPR(R10)(r4)
1012 ld r11, VCPU_GPR(R11)(r4)
1013 ld r12, VCPU_GPR(R12)(r4)
1014 ld r13, VCPU_GPR(R13)(r4)
1015
4b8473c9
PM
1016BEGIN_FTR_SECTION
1017 mtspr SPRN_PPR, r0
1018END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
1019 ld r0, VCPU_GPR(R0)(r4)
c75df6f9 1020 ld r4, VCPU_GPR(R4)(r4)
de56a948
PM
1021
1022 hrfid
1023 b .
1024
1025/******************************************************************************
1026 * *
1027 * Exit code *
1028 * *
1029 *****************************************************************************/
1030
1031/*
1032 * We come here from the first-level interrupt handlers.
1033 */
dd96b2c2
AK
1034 .globl kvmppc_interrupt_hv
1035kvmppc_interrupt_hv:
de56a948
PM
1036 /*
1037 * Register contents:
1038 * R12 = interrupt vector
1039 * R13 = PACA
1040 * guest CR, R12 saved in shadow VCPU SCRATCH1/0
1041 * guest R13 saved in SPRN_SCRATCH0
1042 */
36e7bb38 1043 std r9, HSTATE_SCRATCH2(r13)
44a3add8
PM
1044
1045 lbz r9, HSTATE_IN_GUEST(r13)
1046 cmpwi r9, KVM_GUEST_MODE_HOST_HV
1047 beq kvmppc_bad_host_intr
dd96b2c2
AK
1048#ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
1049 cmpwi r9, KVM_GUEST_MODE_GUEST
36e7bb38 1050 ld r9, HSTATE_SCRATCH2(r13)
dd96b2c2
AK
1051 beq kvmppc_interrupt_pr
1052#endif
44a3add8
PM
1053 /* We're now back in the host but in guest MMU context */
1054 li r9, KVM_GUEST_MODE_HOST_HV
1055 stb r9, HSTATE_IN_GUEST(r13)
1056
de56a948
PM
1057 ld r9, HSTATE_KVM_VCPU(r13)
1058
1059 /* Save registers */
1060
c75df6f9
MN
1061 std r0, VCPU_GPR(R0)(r9)
1062 std r1, VCPU_GPR(R1)(r9)
1063 std r2, VCPU_GPR(R2)(r9)
1064 std r3, VCPU_GPR(R3)(r9)
1065 std r4, VCPU_GPR(R4)(r9)
1066 std r5, VCPU_GPR(R5)(r9)
1067 std r6, VCPU_GPR(R6)(r9)
1068 std r7, VCPU_GPR(R7)(r9)
1069 std r8, VCPU_GPR(R8)(r9)
36e7bb38 1070 ld r0, HSTATE_SCRATCH2(r13)
c75df6f9
MN
1071 std r0, VCPU_GPR(R9)(r9)
1072 std r10, VCPU_GPR(R10)(r9)
1073 std r11, VCPU_GPR(R11)(r9)
de56a948
PM
1074 ld r3, HSTATE_SCRATCH0(r13)
1075 lwz r4, HSTATE_SCRATCH1(r13)
c75df6f9 1076 std r3, VCPU_GPR(R12)(r9)
de56a948 1077 stw r4, VCPU_CR(r9)
0acb9111
PM
1078BEGIN_FTR_SECTION
1079 ld r3, HSTATE_CFAR(r13)
1080 std r3, VCPU_CFAR(r9)
1081END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
4b8473c9
PM
1082BEGIN_FTR_SECTION
1083 ld r4, HSTATE_PPR(r13)
1084 std r4, VCPU_PPR(r9)
1085END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
de56a948
PM
1086
1087 /* Restore R1/R2 so we can handle faults */
1088 ld r1, HSTATE_HOST_R1(r13)
1089 ld r2, PACATOC(r13)
1090
1091 mfspr r10, SPRN_SRR0
1092 mfspr r11, SPRN_SRR1
1093 std r10, VCPU_SRR0(r9)
1094 std r11, VCPU_SRR1(r9)
1095 andi. r0, r12, 2 /* need to read HSRR0/1? */
1096 beq 1f
1097 mfspr r10, SPRN_HSRR0
1098 mfspr r11, SPRN_HSRR1
1099 clrrdi r12, r12, 2
11001: std r10, VCPU_PC(r9)
1101 std r11, VCPU_MSR(r9)
1102
1103 GET_SCRATCH0(r3)
1104 mflr r4
c75df6f9 1105 std r3, VCPU_GPR(R13)(r9)
de56a948
PM
1106 std r4, VCPU_LR(r9)
1107
de56a948
PM
1108 stw r12,VCPU_TRAP(r9)
1109
697d3899
PM
1110 /* Save HEIR (HV emulation assist reg) in last_inst
1111 if this is an HEI (HV emulation interrupt, e40) */
1112 li r3,KVM_INST_FETCH_FAILED
1113BEGIN_FTR_SECTION
1114 cmpwi r12,BOOK3S_INTERRUPT_H_EMUL_ASSIST
1115 bne 11f
1116 mfspr r3,SPRN_HEIR
1117END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
111811: stw r3,VCPU_LAST_INST(r9)
1119
1120 /* these are volatile across C function calls */
1121 mfctr r3
1122 mfxer r4
1123 std r3, VCPU_CTR(r9)
1124 stw r4, VCPU_XER(r9)
1125
1126BEGIN_FTR_SECTION
1127 /* If this is a page table miss then see if it's theirs or ours */
1128 cmpwi r12, BOOK3S_INTERRUPT_H_DATA_STORAGE
1129 beq kvmppc_hdsi
342d3db7
PM
1130 cmpwi r12, BOOK3S_INTERRUPT_H_INST_STORAGE
1131 beq kvmppc_hisi
697d3899
PM
1132END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1133
de56a948
PM
1134 /* See if this is a leftover HDEC interrupt */
1135 cmpwi r12,BOOK3S_INTERRUPT_HV_DECREMENTER
1136 bne 2f
1137 mfspr r3,SPRN_HDEC
1138 cmpwi r3,0
1139 bge ignore_hdec
11402:
697d3899 1141 /* See if this is an hcall we can handle in real mode */
a8606e20
PM
1142 cmpwi r12,BOOK3S_INTERRUPT_SYSCALL
1143 beq hcall_try_real_mode
de56a948 1144
54695c30 1145 /* Only handle external interrupts here on arch 206 and later */
9e368f29 1146BEGIN_FTR_SECTION
54695c30
BH
1147 b ext_interrupt_to_host
1148END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_206)
1149
1150 /* External interrupt ? */
1151 cmpwi r12, BOOK3S_INTERRUPT_EXTERNAL
1152 bne+ ext_interrupt_to_host
1153
1154 /* External interrupt, first check for host_ipi. If this is
1155 * set, we know the host wants us out so let's do it now
1156 */
c934243c
PM
1157 bl kvmppc_read_intr
1158 cmpdi r3, 0
1159 bgt ext_interrupt_to_host
54695c30 1160
4619ac88
PM
1161 /* Check if any CPU is heading out to the host, if so head out too */
1162 ld r5, HSTATE_KVM_VCORE(r13)
1163 lwz r0, VCORE_ENTRY_EXIT(r5)
1164 cmpwi r0, 0x100
1165 bge ext_interrupt_to_host
1166
e3bbbbfa
PM
1167 /* Return to guest after delivering any pending interrupt */
1168 mr r4, r9
1169 b deliver_guest_interrupt
54695c30 1170
54695c30 1171ext_interrupt_to_host:
de56a948 1172
b4072df4 1173guest_exit_cont: /* r9 = vcpu, r12 = trap, r13 = paca */
de56a948 1174 /* Save more register state */
de56a948
PM
1175 mfdar r6
1176 mfdsisr r7
de56a948
PM
1177 std r6, VCPU_DAR(r9)
1178 stw r7, VCPU_DSISR(r9)
9e368f29 1179BEGIN_FTR_SECTION
697d3899 1180 /* don't overwrite fault_dar/fault_dsisr if HDSI */
de56a948
PM
1181 cmpwi r12,BOOK3S_INTERRUPT_H_DATA_STORAGE
1182 beq 6f
9e368f29 1183END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
697d3899 1184 std r6, VCPU_FAULT_DAR(r9)
de56a948
PM
1185 stw r7, VCPU_FAULT_DSISR(r9)
1186
b4072df4
PM
1187 /* See if it is a machine check */
1188 cmpwi r12, BOOK3S_INTERRUPT_MACHINE_CHECK
1189 beq machine_check_realmode
1190mc_cont:
1191
de56a948 1192 /* Save guest CTRL register, set runlatch to 1 */
697d3899 11936: mfspr r6,SPRN_CTRLF
de56a948
PM
1194 stw r6,VCPU_CTRL(r9)
1195 andi. r0,r6,1
1196 bne 4f
1197 ori r6,r6,1
1198 mtspr SPRN_CTRLT,r6
11994:
1200 /* Read the guest SLB and save it away */
1201 lwz r0,VCPU_SLB_NR(r9) /* number of entries in SLB */
1202 mtctr r0
1203 li r6,0
1204 addi r7,r9,VCPU_SLB
1205 li r5,0
12061: slbmfee r8,r6
1207 andis. r0,r8,SLB_ESID_V@h
1208 beq 2f
1209 add r8,r8,r6 /* put index in */
1210 slbmfev r3,r6
1211 std r8,VCPU_SLB_E(r7)
1212 std r3,VCPU_SLB_V(r7)
1213 addi r7,r7,VCPU_SLB_SIZE
1214 addi r5,r5,1
12152: addi r6,r6,1
1216 bdnz 1b
1217 stw r5,VCPU_SLB_MAX(r9)
1218
1219 /*
1220 * Save the guest PURR/SPURR
1221 */
9e368f29 1222BEGIN_FTR_SECTION
de56a948
PM
1223 mfspr r5,SPRN_PURR
1224 mfspr r6,SPRN_SPURR
1225 ld r7,VCPU_PURR(r9)
1226 ld r8,VCPU_SPURR(r9)
1227 std r5,VCPU_PURR(r9)
1228 std r6,VCPU_SPURR(r9)
1229 subf r5,r7,r5
1230 subf r6,r8,r6
1231
1232 /*
1233 * Restore host PURR/SPURR and add guest times
1234 * so that the time in the guest gets accounted.
1235 */
1236 ld r3,HSTATE_PURR(r13)
1237 ld r4,HSTATE_SPURR(r13)
1238 add r3,r3,r5
1239 add r4,r4,r6
1240 mtspr SPRN_PURR,r3
1241 mtspr SPRN_SPURR,r4
9e368f29 1242END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_201)
de56a948 1243
e0b7ec05
PM
1244 /* Save DEC */
1245 mfspr r5,SPRN_DEC
1246 mftb r6
1247 extsw r5,r5
1248 add r5,r5,r6
c5fb80d3
PM
1249 /* r5 is a guest timebase value here, convert to host TB */
1250 ld r3,HSTATE_KVM_VCORE(r13)
1251 ld r4,VCORE_TB_OFFSET(r3)
1252 subf r5,r4,r5
e0b7ec05
PM
1253 std r5,VCPU_DEC_EXPIRES(r9)
1254
b005255e
MN
1255BEGIN_FTR_SECTION
1256 b 8f
1257END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_207S)
b005255e
MN
1258 /* Save POWER8-specific registers */
1259 mfspr r5, SPRN_IAMR
1260 mfspr r6, SPRN_PSPB
1261 mfspr r7, SPRN_FSCR
1262 std r5, VCPU_IAMR(r9)
1263 stw r6, VCPU_PSPB(r9)
1264 std r7, VCPU_FSCR(r9)
1265 mfspr r5, SPRN_IC
1266 mfspr r6, SPRN_VTB
1267 mfspr r7, SPRN_TAR
1268 std r5, VCPU_IC(r9)
1269 std r6, VCPU_VTB(r9)
1270 std r7, VCPU_TAR(r9)
7b490411 1271 mfspr r8, SPRN_EBBHR
b005255e
MN
1272 std r8, VCPU_EBBHR(r9)
1273 mfspr r5, SPRN_EBBRR
1274 mfspr r6, SPRN_BESCR
1275 mfspr r7, SPRN_CSIGR
1276 mfspr r8, SPRN_TACR
1277 std r5, VCPU_EBBRR(r9)
1278 std r6, VCPU_BESCR(r9)
1279 std r7, VCPU_CSIGR(r9)
1280 std r8, VCPU_TACR(r9)
1281 mfspr r5, SPRN_TCSCR
1282 mfspr r6, SPRN_ACOP
1283 mfspr r7, SPRN_PID
1284 mfspr r8, SPRN_WORT
1285 std r5, VCPU_TCSCR(r9)
1286 std r6, VCPU_ACOP(r9)
1287 stw r7, VCPU_GUEST_PID(r9)
1288 std r8, VCPU_WORT(r9)
12898:
1290
e0b7ec05
PM
1291 /* Save and reset AMR and UAMOR before turning on the MMU */
1292BEGIN_FTR_SECTION
1293 mfspr r5,SPRN_AMR
1294 mfspr r6,SPRN_UAMOR
1295 std r5,VCPU_AMR(r9)
1296 std r6,VCPU_UAMOR(r9)
1297 li r6,0
1298 mtspr SPRN_AMR,r6
1299END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1300
1301 /* Switch DSCR back to host value */
1302BEGIN_FTR_SECTION
1303 mfspr r8, SPRN_DSCR
1304 ld r7, HSTATE_DSCR(r13)
1305 std r8, VCPU_DSCR(r9)
1306 mtspr SPRN_DSCR, r7
1307END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1308
1309 /* Save non-volatile GPRs */
1310 std r14, VCPU_GPR(R14)(r9)
1311 std r15, VCPU_GPR(R15)(r9)
1312 std r16, VCPU_GPR(R16)(r9)
1313 std r17, VCPU_GPR(R17)(r9)
1314 std r18, VCPU_GPR(R18)(r9)
1315 std r19, VCPU_GPR(R19)(r9)
1316 std r20, VCPU_GPR(R20)(r9)
1317 std r21, VCPU_GPR(R21)(r9)
1318 std r22, VCPU_GPR(R22)(r9)
1319 std r23, VCPU_GPR(R23)(r9)
1320 std r24, VCPU_GPR(R24)(r9)
1321 std r25, VCPU_GPR(R25)(r9)
1322 std r26, VCPU_GPR(R26)(r9)
1323 std r27, VCPU_GPR(R27)(r9)
1324 std r28, VCPU_GPR(R28)(r9)
1325 std r29, VCPU_GPR(R29)(r9)
1326 std r30, VCPU_GPR(R30)(r9)
1327 std r31, VCPU_GPR(R31)(r9)
1328
1329 /* Save SPRGs */
1330 mfspr r3, SPRN_SPRG0
1331 mfspr r4, SPRN_SPRG1
1332 mfspr r5, SPRN_SPRG2
1333 mfspr r6, SPRN_SPRG3
1334 std r3, VCPU_SPRG0(r9)
1335 std r4, VCPU_SPRG1(r9)
1336 std r5, VCPU_SPRG2(r9)
1337 std r6, VCPU_SPRG3(r9)
1338
1339 /* save FP state */
1340 mr r3, r9
1341 bl kvmppc_save_fp
de56a948 1342
0a8eccef
PM
1343#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1344BEGIN_FTR_SECTION
1345 b 2f
1346END_FTR_SECTION_IFCLR(CPU_FTR_TM)
1347 /* Turn on TM. */
1348 mfmsr r8
1349 li r0, 1
1350 rldimi r8, r0, MSR_TM_LG, 63-MSR_TM_LG
1351 mtmsrd r8
1352
1353 ld r5, VCPU_MSR(r9)
1354 rldicl. r5, r5, 64 - MSR_TS_S_LG, 62
1355 beq 1f /* TM not active in guest. */
1356
1357 li r3, TM_CAUSE_KVM_RESCHED
1358
1359 /* Clear the MSR RI since r1, r13 are all going to be foobar. */
1360 li r5, 0
1361 mtmsrd r5, 1
1362
1363 /* All GPRs are volatile at this point. */
1364 TRECLAIM(R3)
1365
1366 /* Temporarily store r13 and r9 so we have some regs to play with */
1367 SET_SCRATCH0(r13)
1368 GET_PACA(r13)
1369 std r9, PACATMSCRATCH(r13)
1370 ld r9, HSTATE_KVM_VCPU(r13)
1371
1372 /* Get a few more GPRs free. */
1373 std r29, VCPU_GPRS_TM(29)(r9)
1374 std r30, VCPU_GPRS_TM(30)(r9)
1375 std r31, VCPU_GPRS_TM(31)(r9)
1376
1377 /* Save away PPR and DSCR soon so don't run with user values. */
1378 mfspr r31, SPRN_PPR
1379 HMT_MEDIUM
1380 mfspr r30, SPRN_DSCR
1381 ld r29, HSTATE_DSCR(r13)
1382 mtspr SPRN_DSCR, r29
1383
1384 /* Save all but r9, r13 & r29-r31 */
1385 reg = 0
1386 .rept 29
1387 .if (reg != 9) && (reg != 13)
1388 std reg, VCPU_GPRS_TM(reg)(r9)
1389 .endif
1390 reg = reg + 1
1391 .endr
1392 /* ... now save r13 */
1393 GET_SCRATCH0(r4)
1394 std r4, VCPU_GPRS_TM(13)(r9)
1395 /* ... and save r9 */
1396 ld r4, PACATMSCRATCH(r13)
1397 std r4, VCPU_GPRS_TM(9)(r9)
1398
1399 /* Reload stack pointer and TOC. */
1400 ld r1, HSTATE_HOST_R1(r13)
1401 ld r2, PACATOC(r13)
1402
1403 /* Set MSR RI now we have r1 and r13 back. */
1404 li r5, MSR_RI
1405 mtmsrd r5, 1
1406
1407 /* Save away checkpinted SPRs. */
1408 std r31, VCPU_PPR_TM(r9)
1409 std r30, VCPU_DSCR_TM(r9)
1410 mflr r5
1411 mfcr r6
1412 mfctr r7
1413 mfspr r8, SPRN_AMR
1414 mfspr r10, SPRN_TAR
1415 std r5, VCPU_LR_TM(r9)
1416 stw r6, VCPU_CR_TM(r9)
1417 std r7, VCPU_CTR_TM(r9)
1418 std r8, VCPU_AMR_TM(r9)
1419 std r10, VCPU_TAR_TM(r9)
1420
1421 /* Restore r12 as trap number. */
1422 lwz r12, VCPU_TRAP(r9)
1423
1424 /* Save FP/VSX. */
1425 addi r3, r9, VCPU_FPRS_TM
1426 bl .store_fp_state
1427 addi r3, r9, VCPU_VRS_TM
1428 bl .store_vr_state
1429 mfspr r6, SPRN_VRSAVE
1430 stw r6, VCPU_VRSAVE_TM(r9)
14311:
1432 /*
1433 * We need to save these SPRs after the treclaim so that the software
1434 * error code is recorded correctly in the TEXASR. Also the user may
1435 * change these outside of a transaction, so they must always be
1436 * context switched.
1437 */
1438 mfspr r5, SPRN_TFHAR
1439 mfspr r6, SPRN_TFIAR
1440 mfspr r7, SPRN_TEXASR
1441 std r5, VCPU_TFHAR(r9)
1442 std r6, VCPU_TFIAR(r9)
1443 std r7, VCPU_TEXASR(r9)
14442:
1445#endif
1446
e0b7ec05
PM
1447 /* Increment yield count if they have a VPA */
1448 ld r8, VCPU_VPA(r9) /* do they have a VPA? */
1449 cmpdi r8, 0
1450 beq 25f
1451 lwz r3, LPPACA_YIELDCOUNT(r8)
1452 addi r3, r3, 1
1453 stw r3, LPPACA_YIELDCOUNT(r8)
1454 li r3, 1
1455 stb r3, VCPU_VPA_DIRTY(r9)
145625:
1457 /* Save PMU registers if requested */
1458 /* r8 and cr0.eq are live here */
9bc01a9b
PM
1459BEGIN_FTR_SECTION
1460 /*
1461 * POWER8 seems to have a hardware bug where setting
1462 * MMCR0[PMAE] along with MMCR0[PMC1CE] and/or MMCR0[PMCjCE]
1463 * when some counters are already negative doesn't seem
1464 * to cause a performance monitor alert (and hence interrupt).
1465 * The effect of this is that when saving the PMU state,
1466 * if there is no PMU alert pending when we read MMCR0
1467 * before freezing the counters, but one becomes pending
1468 * before we read the counters, we lose it.
1469 * To work around this, we need a way to freeze the counters
1470 * before reading MMCR0. Normally, freezing the counters
1471 * is done by writing MMCR0 (to set MMCR0[FC]) which
1472 * unavoidably writes MMCR0[PMA0] as well. On POWER8,
1473 * we can also freeze the counters using MMCR2, by writing
1474 * 1s to all the counter freeze condition bits (there are
1475 * 9 bits each for 6 counters).
1476 */
1477 li r3, -1 /* set all freeze bits */
1478 clrrdi r3, r3, 10
1479 mfspr r10, SPRN_MMCR2
1480 mtspr SPRN_MMCR2, r3
1481 isync
1482END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
e0b7ec05
PM
1483 li r3, 1
1484 sldi r3, r3, 31 /* MMCR0_FC (freeze counters) bit */
1485 mfspr r4, SPRN_MMCR0 /* save MMCR0 */
1486 mtspr SPRN_MMCR0, r3 /* freeze all counters, disable ints */
1487 mfspr r6, SPRN_MMCRA
1488BEGIN_FTR_SECTION
1489 /* On P7, clear MMCRA in order to disable SDAR updates */
1490 li r7, 0
1491 mtspr SPRN_MMCRA, r7
1492END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
1493 isync
1494 beq 21f /* if no VPA, save PMU stuff anyway */
1495 lbz r7, LPPACA_PMCINUSE(r8)
1496 cmpwi r7, 0 /* did they ask for PMU stuff to be saved? */
1497 bne 21f
1498 std r3, VCPU_MMCR(r9) /* if not, set saved MMCR0 to FC */
1499 b 22f
150021: mfspr r5, SPRN_MMCR1
1501 mfspr r7, SPRN_SIAR
1502 mfspr r8, SPRN_SDAR
1503 std r4, VCPU_MMCR(r9)
1504 std r5, VCPU_MMCR + 8(r9)
1505 std r6, VCPU_MMCR + 16(r9)
9bc01a9b
PM
1506BEGIN_FTR_SECTION
1507 std r10, VCPU_MMCR + 24(r9)
1508END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
e0b7ec05
PM
1509 std r7, VCPU_SIAR(r9)
1510 std r8, VCPU_SDAR(r9)
1511 mfspr r3, SPRN_PMC1
1512 mfspr r4, SPRN_PMC2
1513 mfspr r5, SPRN_PMC3
1514 mfspr r6, SPRN_PMC4
1515 mfspr r7, SPRN_PMC5
1516 mfspr r8, SPRN_PMC6
1517BEGIN_FTR_SECTION
1518 mfspr r10, SPRN_PMC7
1519 mfspr r11, SPRN_PMC8
1520END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
1521 stw r3, VCPU_PMC(r9)
1522 stw r4, VCPU_PMC + 4(r9)
1523 stw r5, VCPU_PMC + 8(r9)
1524 stw r6, VCPU_PMC + 12(r9)
1525 stw r7, VCPU_PMC + 16(r9)
1526 stw r8, VCPU_PMC + 20(r9)
1527BEGIN_FTR_SECTION
1528 stw r10, VCPU_PMC + 24(r9)
1529 stw r11, VCPU_PMC + 28(r9)
1530END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
b005255e 1531BEGIN_FTR_SECTION
b005255e
MN
1532 mfspr r5, SPRN_SIER
1533 mfspr r6, SPRN_SPMC1
1534 mfspr r7, SPRN_SPMC2
1535 mfspr r8, SPRN_MMCRS
b005255e
MN
1536 std r5, VCPU_SIER(r9)
1537 stw r6, VCPU_PMC + 24(r9)
1538 stw r7, VCPU_PMC + 28(r9)
1539 std r8, VCPU_MMCR + 32(r9)
1540 lis r4, 0x8000
1541 mtspr SPRN_MMCRS, r4
1542END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
e0b7ec05 154322:
de56a948
PM
1544 /* Clear out SLB */
1545 li r5,0
1546 slbmte r5,r5
1547 slbia
1548 ptesync
1549
e0b7ec05 1550hdec_soon: /* r12 = trap, r13 = paca */
9e368f29
PM
1551BEGIN_FTR_SECTION
1552 b 32f
1553END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
1554 /*
1555 * POWER7 guest -> host partition switch code.
1556 * We don't have to lock against tlbies but we do
1557 * have to coordinate the hardware threads.
1558 */
371fefd6
PM
1559 /* Increment the threads-exiting-guest count in the 0xff00
1560 bits of vcore->entry_exit_count */
371fefd6
PM
1561 ld r5,HSTATE_KVM_VCORE(r13)
1562 addi r6,r5,VCORE_ENTRY_EXIT
156341: lwarx r3,0,r6
1564 addi r0,r3,0x100
1565 stwcx. r0,0,r6
1566 bne 41b
f019b7ad 1567 isync /* order stwcx. vs. reading napping_threads */
371fefd6
PM
1568
1569 /*
1570 * At this point we have an interrupt that we have to pass
1571 * up to the kernel or qemu; we can't handle it in real mode.
1572 * Thus we have to do a partition switch, so we have to
1573 * collect the other threads, if we are the first thread
1574 * to take an interrupt. To do this, we set the HDEC to 0,
1575 * which causes an HDEC interrupt in all threads within 2ns
1576 * because the HDEC register is shared between all 4 threads.
1577 * However, we don't need to bother if this is an HDEC
1578 * interrupt, since the other threads will already be on their
1579 * way here in that case.
1580 */
19ccb76a
PM
1581 cmpwi r3,0x100 /* Are we the first here? */
1582 bge 43f
371fefd6
PM
1583 cmpwi r12,BOOK3S_INTERRUPT_HV_DECREMENTER
1584 beq 40f
371fefd6
PM
1585 li r0,0
1586 mtspr SPRN_HDEC,r0
158740:
19ccb76a
PM
1588 /*
1589 * Send an IPI to any napping threads, since an HDEC interrupt
1590 * doesn't wake CPUs up from nap.
1591 */
1592 lwz r3,VCORE_NAPPING_THREADS(r5)
e0b7ec05 1593 lbz r4,HSTATE_PTID(r13)
19ccb76a 1594 li r0,1
2f584a14 1595 sld r0,r0,r4
19ccb76a
PM
1596 andc. r3,r3,r0 /* no sense IPI'ing ourselves */
1597 beq 43f
f019b7ad
PM
1598 /* Order entry/exit update vs. IPIs */
1599 sync
19ccb76a
PM
1600 mulli r4,r4,PACA_SIZE /* get paca for thread 0 */
1601 subf r6,r4,r13
160242: andi. r0,r3,1
1603 beq 44f
1604 ld r8,HSTATE_XICS_PHYS(r6) /* get thread's XICS reg addr */
1605 li r0,IPI_PRIORITY
54695c30 1606 li r7,XICS_MFRR
19ccb76a
PM
1607 stbcix r0,r7,r8 /* trigger the IPI */
160844: srdi. r3,r3,1
1609 addi r6,r6,PACA_SIZE
1610 bne 42b
371fefd6 1611
e0b7ec05 1612secondary_too_late:
371fefd6 1613 /* Secondary threads wait for primary to do partition switch */
e0b7ec05
PM
161443: ld r5,HSTATE_KVM_VCORE(r13)
1615 ld r4,VCORE_KVM(r5) /* pointer to struct kvm */
1616 lbz r3,HSTATE_PTID(r13)
371fefd6
PM
1617 cmpwi r3,0
1618 beq 15f
1619 HMT_LOW
162013: lbz r3,VCORE_IN_GUEST(r5)
1621 cmpwi r3,0
1622 bne 13b
1623 HMT_MEDIUM
1624 b 16f
1625
1626 /* Primary thread waits for all the secondaries to exit guest */
162715: lwz r3,VCORE_ENTRY_EXIT(r5)
1628 srwi r0,r3,8
1629 clrldi r3,r3,56
1630 cmpw r3,r0
1631 bne 15b
1632 isync
1633
1634 /* Primary thread switches back to host partition */
de56a948
PM
1635 ld r6,KVM_HOST_SDR1(r4)
1636 lwz r7,KVM_HOST_LPID(r4)
1637 li r8,LPID_RSVD /* switch to reserved LPID */
1638 mtspr SPRN_LPID,r8
1639 ptesync
1640 mtspr SPRN_SDR1,r6 /* switch to partition page table */
1641 mtspr SPRN_LPID,r7
1642 isync
93b0f4dc 1643
b005255e
MN
1644BEGIN_FTR_SECTION
1645 /* DPDES is shared between threads */
1646 mfspr r7, SPRN_DPDES
1647 std r7, VCORE_DPDES(r5)
1648 /* clear DPDES so we don't get guest doorbells in the host */
1649 li r8, 0
1650 mtspr SPRN_DPDES, r8
1651END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
1652
93b0f4dc
PM
1653 /* Subtract timebase offset from timebase */
1654 ld r8,VCORE_TB_OFFSET(r5)
1655 cmpdi r8,0
1656 beq 17f
c5fb80d3 1657 mftb r6 /* current guest timebase */
93b0f4dc
PM
1658 subf r8,r8,r6
1659 mtspr SPRN_TBU40,r8 /* update upper 40 bits */
1660 mftb r7 /* check if lower 24 bits overflowed */
1661 clrldi r6,r6,40
1662 clrldi r7,r7,40
1663 cmpld r7,r6
1664 bge 17f
1665 addis r8,r8,0x100 /* if so, increment upper 40 bits */
1666 mtspr SPRN_TBU40,r8
1667
388cc6e1
PM
1668 /* Reset PCR */
166917: ld r0, VCORE_PCR(r5)
1670 cmpdi r0, 0
1671 beq 18f
1672 li r0, 0
1673 mtspr SPRN_PCR, r0
167418:
93b0f4dc 1675 /* Signal secondary CPUs to continue */
371fefd6 1676 stb r0,VCORE_IN_GUEST(r5)
de56a948
PM
1677 lis r8,0x7fff /* MAX_INT@h */
1678 mtspr SPRN_HDEC,r8
1679
371fefd6 168016: ld r8,KVM_HOST_LPCR(r4)
de56a948
PM
1681 mtspr SPRN_LPCR,r8
1682 isync
9e368f29
PM
1683 b 33f
1684
1685 /*
1686 * PPC970 guest -> host partition switch code.
1687 * We have to lock against concurrent tlbies, and
1688 * we have to flush the whole TLB.
1689 */
e0b7ec05
PM
169032: ld r5,HSTATE_KVM_VCORE(r13)
1691 ld r4,VCORE_KVM(r5) /* pointer to struct kvm */
9e368f29
PM
1692
1693 /* Take the guest's tlbie_lock */
54bb7f4b 1694#ifdef __BIG_ENDIAN__
9e368f29 1695 lwz r8,PACA_LOCK_TOKEN(r13)
54bb7f4b
AB
1696#else
1697 lwz r8,PACAPACAINDEX(r13)
1698#endif
9e368f29
PM
1699 addi r3,r4,KVM_TLBIE_LOCK
170024: lwarx r0,0,r3
1701 cmpwi r0,0
1702 bne 24b
1703 stwcx. r8,0,r3
1704 bne 24b
1705 isync
1706
1707 ld r7,KVM_HOST_LPCR(r4) /* use kvm->arch.host_lpcr for HID4 */
1708 li r0,0x18f
1709 rotldi r0,r0,HID4_LPID5_SH /* all lpid bits in HID4 = 1 */
1710 or r0,r7,r0
1711 ptesync
1712 sync
1713 mtspr SPRN_HID4,r0 /* switch to reserved LPID */
1714 isync
1715 li r0,0
1716 stw r0,0(r3) /* drop guest tlbie_lock */
1717
1718 /* invalidate the whole TLB */
1719 li r0,256
1720 mtctr r0
1721 li r6,0
172225: tlbiel r6
1723 addi r6,r6,0x1000
1724 bdnz 25b
1725 ptesync
1726
1727 /* take native_tlbie_lock */
1728 ld r3,toc_tlbie_lock@toc(2)
172924: lwarx r0,0,r3
1730 cmpwi r0,0
1731 bne 24b
1732 stwcx. r8,0,r3
1733 bne 24b
1734 isync
1735
1736 ld r6,KVM_HOST_SDR1(r4)
1737 mtspr SPRN_SDR1,r6 /* switch to host page table */
1738
1739 /* Set up host HID4 value */
1740 sync
1741 mtspr SPRN_HID4,r7
1742 isync
1743 li r0,0
1744 stw r0,0(r3) /* drop native_tlbie_lock */
1745
1746 lis r8,0x7fff /* MAX_INT@h */
1747 mtspr SPRN_HDEC,r8
1748
1749 /* Disable HDEC interrupts */
1750 mfspr r0,SPRN_HID0
1751 li r3,0
1752 rldimi r0,r3, HID0_HDICE_SH, 64-HID0_HDICE_SH-1
1753 sync
1754 mtspr SPRN_HID0,r0
1755 mfspr r0,SPRN_HID0
1756 mfspr r0,SPRN_HID0
1757 mfspr r0,SPRN_HID0
1758 mfspr r0,SPRN_HID0
1759 mfspr r0,SPRN_HID0
1760 mfspr r0,SPRN_HID0
de56a948
PM
1761
1762 /* load host SLB entries */
9e368f29 176333: ld r8,PACA_SLBSHADOWPTR(r13)
de56a948
PM
1764
1765 .rept SLB_NUM_BOLTED
1766 ld r5,SLBSHADOW_SAVEAREA(r8)
1767 ld r6,SLBSHADOW_SAVEAREA+8(r8)
1768 andis. r7,r5,SLB_ESID_V@h
1769 beq 1f
1770 slbmte r6,r5
17711: addi r8,r8,16
1772 .endr
1773
44a3add8
PM
1774 /* Unset guest mode */
1775 li r0, KVM_GUEST_MODE_NONE
1776 stb r0, HSTATE_IN_GUEST(r13)
1777
218309b7
PM
1778 ld r0, 112+PPC_LR_STKOFF(r1)
1779 addi r1, r1, 112
1780 mtlr r0
1781 blr
b4072df4 1782
697d3899
PM
1783/*
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.
1789 */
1790kvmppc_hdsi:
1791 mfspr r4, SPRN_HDAR
1792 mfspr r6, SPRN_HDSISR
4cf302bc
PM
1793 /* HPTE not found fault or protection fault? */
1794 andis. r0, r6, (DSISR_NOHPTE | DSISR_PROTFAULT)@h
697d3899
PM
1795 beq 1f /* if not, send it to the guest */
1796 andi. r0, r11, MSR_DR /* data relocation enabled? */
1797 beq 3f
1798 clrrdi r0, r4, 28
c75df6f9 1799 PPC_SLBFEE_DOT(R5, R0) /* if so, look up SLB */
697d3899
PM
1800 bne 1f /* if no SLB entry found */
18014: std r4, VCPU_FAULT_DAR(r9)
1802 stw r6, VCPU_FAULT_DSISR(r9)
1803
1804 /* Search the hash table. */
1805 mr r3, r9 /* vcpu pointer */
342d3db7 1806 li r7, 1 /* data fault */
b1576fec 1807 bl kvmppc_hpte_hv_fault
697d3899
PM
1808 ld r9, HSTATE_KVM_VCPU(r13)
1809 ld r10, VCPU_PC(r9)
1810 ld r11, VCPU_MSR(r9)
1811 li r12, BOOK3S_INTERRUPT_H_DATA_STORAGE
1812 cmpdi r3, 0 /* retry the instruction */
1813 beq 6f
1814 cmpdi r3, -1 /* handle in kernel mode */
b4072df4 1815 beq guest_exit_cont
697d3899
PM
1816 cmpdi r3, -2 /* MMIO emulation; need instr word */
1817 beq 2f
1818
1819 /* Synthesize a DSI for the guest */
1820 ld r4, VCPU_FAULT_DAR(r9)
1821 mr r6, r3
18221: 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
e4e38121 1827 bl kvmppc_msr_interrupt
b4072df4 1828fast_interrupt_c_return:
697d3899
PM
18296: ld r7, VCPU_CTR(r9)
1830 lwz r8, VCPU_XER(r9)
1831 mtctr r7
1832 mtxer r8
1833 mr r4, r9
1834 b fast_guest_return
1835
18363: ld r5, VCPU_KVM(r9) /* not relocated, use VRMA */
1837 ld r5, KVM_VRMA_SLB_V(r5)
1838 b 4b
1839
1840 /* If this is for emulated MMIO, load the instruction word */
18412: li r8, KVM_INST_FETCH_FAILED /* In case lwz faults */
1842
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)
1847
1848 /* Do the access with MSR:DR enabled */
1849 mfmsr r3
1850 ori r4, r3, MSR_DR /* Enable paging for data */
1851 mtmsrd r4
1852 lwz r8, 0(r10)
1853 mtmsrd r3
1854
1855 /* Store the result */
1856 stw r8, VCPU_LAST_INST(r9)
1857
1858 /* Unset guest mode. */
44a3add8 1859 li r0, KVM_GUEST_MODE_HOST_HV
697d3899 1860 stb r0, HSTATE_IN_GUEST(r13)
b4072df4 1861 b guest_exit_cont
de56a948 1862
342d3db7
PM
1863/*
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.
1866 */
1867kvmppc_hisi:
1868 andis. r0, r11, SRR1_ISI_NOPT@h
1869 beq 1f
1870 andi. r0, r11, MSR_IR /* instruction relocation enabled? */
1871 beq 3f
1872 clrrdi r0, r10, 28
c75df6f9 1873 PPC_SLBFEE_DOT(R5, R0) /* if so, look up SLB */
342d3db7
PM
1874 bne 1f /* if no SLB entry found */
18754:
1876 /* Search the hash table. */
1877 mr r3, r9 /* vcpu pointer */
1878 mr r4, r10
1879 mr r6, r11
1880 li r7, 0 /* instruction fault */
b1576fec 1881 bl kvmppc_hpte_hv_fault
342d3db7
PM
1882 ld r9, HSTATE_KVM_VCPU(r13)
1883 ld r10, VCPU_PC(r9)
1884 ld r11, VCPU_MSR(r9)
1885 li r12, BOOK3S_INTERRUPT_H_INST_STORAGE
1886 cmpdi r3, 0 /* retry the instruction */
b4072df4 1887 beq fast_interrupt_c_return
342d3db7 1888 cmpdi r3, -1 /* handle in kernel mode */
b4072df4 1889 beq guest_exit_cont
342d3db7
PM
1890
1891 /* Synthesize an ISI for the guest */
1892 mr r11, r3
18931: mtspr SPRN_SRR0, r10
1894 mtspr SPRN_SRR1, r11
1895 li r10, BOOK3S_INTERRUPT_INST_STORAGE
e4e38121 1896 bl kvmppc_msr_interrupt
b4072df4 1897 b fast_interrupt_c_return
342d3db7
PM
1898
18993: ld r6, VCPU_KVM(r9) /* not relocated, use VRMA */
1900 ld r5, KVM_VRMA_SLB_V(r6)
1901 b 4b
1902
a8606e20
PM
1903/*
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).
1908 */
1909 .globl hcall_try_real_mode
1910hcall_try_real_mode:
c75df6f9 1911 ld r3,VCPU_GPR(R3)(r9)
a8606e20 1912 andi. r0,r11,MSR_PR
27025a60
LPF
1913 /* sc 1 from userspace - reflect to guest syscall */
1914 bne sc_1_fast_return
a8606e20
PM
1915 clrrdi r3,r3,2
1916 cmpldi r3,hcall_real_table_end - hcall_real_table
b4072df4 1917 bge guest_exit_cont
a8606e20 1918 LOAD_REG_ADDR(r4, hcall_real_table)
4baa1d87 1919 lwax r3,r3,r4
a8606e20 1920 cmpwi r3,0
b4072df4 1921 beq guest_exit_cont
a8606e20
PM
1922 add r3,r3,r4
1923 mtctr r3
1924 mr r3,r9 /* get vcpu pointer */
c75df6f9 1925 ld r4,VCPU_GPR(R4)(r9)
a8606e20
PM
1926 bctrl
1927 cmpdi r3,H_TOO_HARD
1928 beq hcall_real_fallback
1929 ld r4,HSTATE_KVM_VCPU(r13)
c75df6f9 1930 std r3,VCPU_GPR(R3)(r4)
a8606e20
PM
1931 ld r10,VCPU_PC(r4)
1932 ld r11,VCPU_MSR(r4)
1933 b fast_guest_return
1934
27025a60
LPF
1935sc_1_fast_return:
1936 mtspr SPRN_SRR0,r10
1937 mtspr SPRN_SRR1,r11
1938 li r10, BOOK3S_INTERRUPT_SYSCALL
e4e38121 1939 bl kvmppc_msr_interrupt
27025a60
LPF
1940 mr r4,r9
1941 b fast_guest_return
1942
a8606e20
PM
1943 /* We've attempted a real mode hcall, but it's punted it back
1944 * to userspace. We need to restore some clobbered volatiles
1945 * before resuming the pass-it-to-qemu path */
1946hcall_real_fallback:
1947 li r12,BOOK3S_INTERRUPT_SYSCALL
1948 ld r9, HSTATE_KVM_VCPU(r13)
a8606e20 1949
b4072df4 1950 b guest_exit_cont
a8606e20
PM
1951
1952 .globl hcall_real_table
1953hcall_real_table:
1954 .long 0 /* 0 - unused */
c1fb0194
AB
1955 .long DOTSYM(kvmppc_h_remove) - hcall_real_table
1956 .long DOTSYM(kvmppc_h_enter) - hcall_real_table
1957 .long DOTSYM(kvmppc_h_read) - hcall_real_table
a8606e20
PM
1958 .long 0 /* 0x10 - H_CLEAR_MOD */
1959 .long 0 /* 0x14 - H_CLEAR_REF */
c1fb0194
AB
1960 .long DOTSYM(kvmppc_h_protect) - hcall_real_table
1961 .long DOTSYM(kvmppc_h_get_tce) - hcall_real_table
1962 .long DOTSYM(kvmppc_h_put_tce) - hcall_real_table
a8606e20 1963 .long 0 /* 0x24 - H_SET_SPRG0 */
c1fb0194 1964 .long DOTSYM(kvmppc_h_set_dabr) - hcall_real_table
a8606e20
PM
1965 .long 0 /* 0x2c */
1966 .long 0 /* 0x30 */
1967 .long 0 /* 0x34 */
1968 .long 0 /* 0x38 */
1969 .long 0 /* 0x3c */
1970 .long 0 /* 0x40 */
1971 .long 0 /* 0x44 */
1972 .long 0 /* 0x48 */
1973 .long 0 /* 0x4c */
1974 .long 0 /* 0x50 */
1975 .long 0 /* 0x54 */
1976 .long 0 /* 0x58 */
1977 .long 0 /* 0x5c */
1978 .long 0 /* 0x60 */
e7d26f28 1979#ifdef CONFIG_KVM_XICS
c1fb0194
AB
1980 .long DOTSYM(kvmppc_rm_h_eoi) - hcall_real_table
1981 .long DOTSYM(kvmppc_rm_h_cppr) - hcall_real_table
1982 .long DOTSYM(kvmppc_rm_h_ipi) - hcall_real_table
e7d26f28 1983 .long 0 /* 0x70 - H_IPOLL */
c1fb0194 1984 .long DOTSYM(kvmppc_rm_h_xirr) - hcall_real_table
e7d26f28
BH
1985#else
1986 .long 0 /* 0x64 - H_EOI */
1987 .long 0 /* 0x68 - H_CPPR */
1988 .long 0 /* 0x6c - H_IPI */
1989 .long 0 /* 0x70 - H_IPOLL */
1990 .long 0 /* 0x74 - H_XIRR */
1991#endif
a8606e20
PM
1992 .long 0 /* 0x78 */
1993 .long 0 /* 0x7c */
1994 .long 0 /* 0x80 */
1995 .long 0 /* 0x84 */
1996 .long 0 /* 0x88 */
1997 .long 0 /* 0x8c */
1998 .long 0 /* 0x90 */
1999 .long 0 /* 0x94 */
2000 .long 0 /* 0x98 */
2001 .long 0 /* 0x9c */
2002 .long 0 /* 0xa0 */
2003 .long 0 /* 0xa4 */
2004 .long 0 /* 0xa8 */
2005 .long 0 /* 0xac */
2006 .long 0 /* 0xb0 */
2007 .long 0 /* 0xb4 */
2008 .long 0 /* 0xb8 */
2009 .long 0 /* 0xbc */
2010 .long 0 /* 0xc0 */
2011 .long 0 /* 0xc4 */
2012 .long 0 /* 0xc8 */
2013 .long 0 /* 0xcc */
2014 .long 0 /* 0xd0 */
2015 .long 0 /* 0xd4 */
2016 .long 0 /* 0xd8 */
2017 .long 0 /* 0xdc */
c1fb0194 2018 .long DOTSYM(kvmppc_h_cede) - hcall_real_table
a8606e20
PM
2019 .long 0 /* 0xe4 */
2020 .long 0 /* 0xe8 */
2021 .long 0 /* 0xec */
2022 .long 0 /* 0xf0 */
2023 .long 0 /* 0xf4 */
2024 .long 0 /* 0xf8 */
2025 .long 0 /* 0xfc */
2026 .long 0 /* 0x100 */
2027 .long 0 /* 0x104 */
2028 .long 0 /* 0x108 */
2029 .long 0 /* 0x10c */
2030 .long 0 /* 0x110 */
2031 .long 0 /* 0x114 */
2032 .long 0 /* 0x118 */
2033 .long 0 /* 0x11c */
2034 .long 0 /* 0x120 */
c1fb0194 2035 .long DOTSYM(kvmppc_h_bulk_remove) - hcall_real_table
8563bf52
PM
2036 .long 0 /* 0x128 */
2037 .long 0 /* 0x12c */
2038 .long 0 /* 0x130 */
c1fb0194 2039 .long DOTSYM(kvmppc_h_set_xdabr) - hcall_real_table
a8606e20
PM
2040hcall_real_table_end:
2041
de56a948
PM
2042ignore_hdec:
2043 mr r4,r9
2044 b fast_guest_return
2045
8563bf52
PM
2046_GLOBAL(kvmppc_h_set_xdabr)
2047 andi. r0, r5, DABRX_USER | DABRX_KERNEL
2048 beq 6f
2049 li r0, DABRX_USER | DABRX_KERNEL | DABRX_BTI
2050 andc. r0, r5, r0
2051 beq 3f
20526: li r3, H_PARAMETER
2053 blr
2054
a8606e20 2055_GLOBAL(kvmppc_h_set_dabr)
8563bf52
PM
2056 li r5, DABRX_USER | DABRX_KERNEL
20573:
eee7ff9d
MN
2058BEGIN_FTR_SECTION
2059 b 2f
2060END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
a8606e20 2061 std r4,VCPU_DABR(r3)
8563bf52
PM
2062 stw r5, VCPU_DABRX(r3)
2063 mtspr SPRN_DABRX, r5
8943633c
PM
2064 /* Work around P7 bug where DABR can get corrupted on mtspr */
20651: mtspr SPRN_DABR,r4
2066 mfspr r5, SPRN_DABR
2067 cmpd r4, r5
2068 bne 1b
2069 isync
a8606e20
PM
2070 li r3,0
2071 blr
2072
8563bf52
PM
2073 /* Emulate H_SET_DABR/X on P8 for the sake of compat mode guests */
20742: rlwimi r5, r4, 5, DAWRX_DR | DAWRX_DW
2075 rlwimi r5, r4, 1, DAWRX_WT
2076 clrrdi r4, r4, 3
2077 std r4, VCPU_DAWR(r3)
2078 std r5, VCPU_DAWRX(r3)
2079 mtspr SPRN_DAWR, r4
2080 mtspr SPRN_DAWRX, r5
2081 li r3, 0
a8606e20
PM
2082 blr
2083
19ccb76a
PM
2084_GLOBAL(kvmppc_h_cede)
2085 ori r11,r11,MSR_EE
2086 std r11,VCPU_MSR(r3)
2087 li r0,1
2088 stb r0,VCPU_CEDED(r3)
2089 sync /* order setting ceded vs. testing prodded */
2090 lbz r5,VCPU_PRODDED(r3)
2091 cmpwi r5,0
04f995a5 2092 bne kvm_cede_prodded
19ccb76a
PM
2093 li r0,0 /* set trap to 0 to say hcall is handled */
2094 stw r0,VCPU_TRAP(r3)
2095 li r0,H_SUCCESS
c75df6f9 2096 std r0,VCPU_GPR(R3)(r3)
19ccb76a 2097BEGIN_FTR_SECTION
04f995a5 2098 b kvm_cede_exit /* just send it up to host on 970 */
19ccb76a
PM
2099END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_206)
2100
2101 /*
2102 * Set our bit in the bitmask of napping threads unless all the
2103 * other threads are already napping, in which case we send this
2104 * up to the host.
2105 */
2106 ld r5,HSTATE_KVM_VCORE(r13)
e0b7ec05 2107 lbz r6,HSTATE_PTID(r13)
19ccb76a
PM
2108 lwz r8,VCORE_ENTRY_EXIT(r5)
2109 clrldi r8,r8,56
2110 li r0,1
2111 sld r0,r0,r6
2112 addi r6,r5,VCORE_NAPPING_THREADS
211331: lwarx r4,0,r6
2114 or r4,r4,r0
c75df6f9 2115 PPC_POPCNTW(R7,R4)
19ccb76a 2116 cmpw r7,r8
04f995a5 2117 bge kvm_cede_exit
19ccb76a
PM
2118 stwcx. r4,0,r6
2119 bne 31b
f019b7ad
PM
2120 /* order napping_threads update vs testing entry_exit_count */
2121 isync
e0b7ec05 2122 li r0,NAPPING_CEDE
19ccb76a 2123 stb r0,HSTATE_NAPPING(r13)
19ccb76a
PM
2124 lwz r7,VCORE_ENTRY_EXIT(r5)
2125 cmpwi r7,0x100
2126 bge 33f /* another thread already exiting */
2127
2128/*
2129 * Although not specifically required by the architecture, POWER7
2130 * preserves the following registers in nap mode, even if an SMT mode
2131 * switch occurs: SLB entries, PURR, SPURR, AMOR, UAMOR, AMR, SPRG0-3,
2132 * DAR, DSISR, DABR, DABRX, DSCR, PMCx, MMCRx, SIAR, SDAR.
2133 */
2134 /* Save non-volatile GPRs */
c75df6f9
MN
2135 std r14, VCPU_GPR(R14)(r3)
2136 std r15, VCPU_GPR(R15)(r3)
2137 std r16, VCPU_GPR(R16)(r3)
2138 std r17, VCPU_GPR(R17)(r3)
2139 std r18, VCPU_GPR(R18)(r3)
2140 std r19, VCPU_GPR(R19)(r3)
2141 std r20, VCPU_GPR(R20)(r3)
2142 std r21, VCPU_GPR(R21)(r3)
2143 std r22, VCPU_GPR(R22)(r3)
2144 std r23, VCPU_GPR(R23)(r3)
2145 std r24, VCPU_GPR(R24)(r3)
2146 std r25, VCPU_GPR(R25)(r3)
2147 std r26, VCPU_GPR(R26)(r3)
2148 std r27, VCPU_GPR(R27)(r3)
2149 std r28, VCPU_GPR(R28)(r3)
2150 std r29, VCPU_GPR(R29)(r3)
2151 std r30, VCPU_GPR(R30)(r3)
2152 std r31, VCPU_GPR(R31)(r3)
19ccb76a
PM
2153
2154 /* save FP state */
595e4f7e 2155 bl kvmppc_save_fp
19ccb76a
PM
2156
2157 /*
aa31e843 2158 * Take a nap until a decrementer or external or doobell interrupt
582b910e
PM
2159 * occurs, with PECE1, PECE0 and PECEDP set in LPCR. Also clear the
2160 * runlatch bit before napping.
19ccb76a 2161 */
582b910e
PM
2162 mfspr r2, SPRN_CTRLF
2163 clrrdi r2, r2, 1
2164 mtspr SPRN_CTRLT, r2
2165
f0888f70
PM
2166 li r0,1
2167 stb r0,HSTATE_HWTHREAD_REQ(r13)
19ccb76a
PM
2168 mfspr r5,SPRN_LPCR
2169 ori r5,r5,LPCR_PECE0 | LPCR_PECE1
aa31e843
PM
2170BEGIN_FTR_SECTION
2171 oris r5,r5,LPCR_PECEDP@h
2172END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
19ccb76a
PM
2173 mtspr SPRN_LPCR,r5
2174 isync
2175 li r0, 0
2176 std r0, HSTATE_SCRATCH0(r13)
2177 ptesync
2178 ld r0, HSTATE_SCRATCH0(r13)
21791: cmpd r0, r0
2180 bne 1b
2181 nap
2182 b .
2183
e3bbbbfa
PM
218433: mr r4, r3
2185 li r3, 0
2186 li r12, 0
2187 b 34f
2188
19ccb76a 2189kvm_end_cede:
4619ac88
PM
2190 /* get vcpu pointer */
2191 ld r4, HSTATE_KVM_VCPU(r13)
2192
19ccb76a
PM
2193 /* Woken by external or decrementer interrupt */
2194 ld r1, HSTATE_HOST_R1(r13)
19ccb76a 2195
19ccb76a
PM
2196 /* load up FP state */
2197 bl kvmppc_load_fp
2198
2199 /* Load NV GPRS */
c75df6f9
MN
2200 ld r14, VCPU_GPR(R14)(r4)
2201 ld r15, VCPU_GPR(R15)(r4)
2202 ld r16, VCPU_GPR(R16)(r4)
2203 ld r17, VCPU_GPR(R17)(r4)
2204 ld r18, VCPU_GPR(R18)(r4)
2205 ld r19, VCPU_GPR(R19)(r4)
2206 ld r20, VCPU_GPR(R20)(r4)
2207 ld r21, VCPU_GPR(R21)(r4)
2208 ld r22, VCPU_GPR(R22)(r4)
2209 ld r23, VCPU_GPR(R23)(r4)
2210 ld r24, VCPU_GPR(R24)(r4)
2211 ld r25, VCPU_GPR(R25)(r4)
2212 ld r26, VCPU_GPR(R26)(r4)
2213 ld r27, VCPU_GPR(R27)(r4)
2214 ld r28, VCPU_GPR(R28)(r4)
2215 ld r29, VCPU_GPR(R29)(r4)
2216 ld r30, VCPU_GPR(R30)(r4)
2217 ld r31, VCPU_GPR(R31)(r4)
e3bbbbfa
PM
2218
2219 /* Check the wake reason in SRR1 to see why we got here */
2220 bl kvmppc_check_wake_reason
19ccb76a
PM
2221
2222 /* clear our bit in vcore->napping_threads */
e3bbbbfa
PM
222334: ld r5,HSTATE_KVM_VCORE(r13)
2224 lbz r7,HSTATE_PTID(r13)
19ccb76a 2225 li r0,1
e3bbbbfa 2226 sld r0,r0,r7
19ccb76a
PM
2227 addi r6,r5,VCORE_NAPPING_THREADS
222832: lwarx r7,0,r6
2229 andc r7,r7,r0
2230 stwcx. r7,0,r6
2231 bne 32b
2232 li r0,0
2233 stb r0,HSTATE_NAPPING(r13)
2234
e3bbbbfa
PM
2235 /* See if the wake reason means we need to exit */
2236 stw r12, VCPU_TRAP(r4)
4619ac88 2237 mr r9, r4
e3bbbbfa
PM
2238 cmpdi r3, 0
2239 bgt guest_exit_cont
4619ac88 2240
19ccb76a
PM
2241 /* see if any other thread is already exiting */
2242 lwz r0,VCORE_ENTRY_EXIT(r5)
2243 cmpwi r0,0x100
e3bbbbfa 2244 bge guest_exit_cont
19ccb76a 2245
e3bbbbfa 2246 b kvmppc_cede_reentry /* if not go back to guest */
19ccb76a
PM
2247
2248 /* cede when already previously prodded case */
04f995a5
PM
2249kvm_cede_prodded:
2250 li r0,0
19ccb76a
PM
2251 stb r0,VCPU_PRODDED(r3)
2252 sync /* order testing prodded vs. clearing ceded */
2253 stb r0,VCPU_CEDED(r3)
2254 li r3,H_SUCCESS
2255 blr
2256
2257 /* we've ceded but we want to give control to the host */
04f995a5 2258kvm_cede_exit:
4619ac88 2259 b hcall_real_fallback
19ccb76a 2260
b4072df4
PM
2261 /* Try to handle a machine check in real mode */
2262machine_check_realmode:
2263 mr r3, r9 /* get vcpu pointer */
b1576fec 2264 bl kvmppc_realmode_machine_check
b4072df4 2265 nop
74845bc2 2266 cmpdi r3, 0 /* Did we handle MCE ? */
b4072df4
PM
2267 ld r9, HSTATE_KVM_VCPU(r13)
2268 li r12, BOOK3S_INTERRUPT_MACHINE_CHECK
74845bc2
MS
2269 /*
2270 * Deliver unhandled/fatal (e.g. UE) MCE errors to guest through
2271 * machine check interrupt (set HSRR0 to 0x200). And for handled
2272 * errors (no-fatal), just go back to guest execution with current
2273 * HSRR0 instead of exiting guest. This new approach will inject
2274 * machine check to guest for fatal error causing guest to crash.
2275 *
2276 * The old code used to return to host for unhandled errors which
2277 * was causing guest to hang with soft lockups inside guest and
2278 * makes it difficult to recover guest instance.
2279 */
2280 ld r10, VCPU_PC(r9)
2281 ld r11, VCPU_MSR(r9)
2282 bne 2f /* Continue guest execution. */
b4072df4
PM
2283 /* If not, deliver a machine check. SRR0/1 are already set */
2284 li r10, BOOK3S_INTERRUPT_MACHINE_CHECK
000a25dd 2285 ld r11, VCPU_MSR(r9)
e4e38121 2286 bl kvmppc_msr_interrupt
74845bc2 22872: b fast_interrupt_c_return
b4072df4 2288
e3bbbbfa
PM
2289/*
2290 * Check the reason we woke from nap, and take appropriate action.
2291 * Returns:
2292 * 0 if nothing needs to be done
2293 * 1 if something happened that needs to be handled by the host
2294 * -1 if there was a guest wakeup (IPI)
2295 *
2296 * Also sets r12 to the interrupt vector for any interrupt that needs
2297 * to be handled now by the host (0x500 for external interrupt), or zero.
2298 */
2299kvmppc_check_wake_reason:
2300 mfspr r6, SPRN_SRR1
aa31e843
PM
2301BEGIN_FTR_SECTION
2302 rlwinm r6, r6, 45-31, 0xf /* extract wake reason field (P8) */
2303FTR_SECTION_ELSE
2304 rlwinm r6, r6, 45-31, 0xe /* P7 wake reason field is 3 bits */
2305ALT_FTR_SECTION_END_IFSET(CPU_FTR_ARCH_207S)
2306 cmpwi r6, 8 /* was it an external interrupt? */
e3bbbbfa
PM
2307 li r12, BOOK3S_INTERRUPT_EXTERNAL
2308 beq kvmppc_read_intr /* if so, see what it was */
2309 li r3, 0
2310 li r12, 0
2311 cmpwi r6, 6 /* was it the decrementer? */
2312 beq 0f
aa31e843
PM
2313BEGIN_FTR_SECTION
2314 cmpwi r6, 5 /* privileged doorbell? */
2315 beq 0f
5d00f66b
PM
2316 cmpwi r6, 3 /* hypervisor doorbell? */
2317 beq 3f
aa31e843 2318END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
e3bbbbfa
PM
2319 li r3, 1 /* anything else, return 1 */
23200: blr
2321
5d00f66b
PM
2322 /* hypervisor doorbell */
23233: li r12, BOOK3S_INTERRUPT_H_DOORBELL
2324 li r3, 1
2325 blr
2326
c934243c
PM
2327/*
2328 * Determine what sort of external interrupt is pending (if any).
2329 * Returns:
2330 * 0 if no interrupt is pending
2331 * 1 if an interrupt is pending that needs to be handled by the host
2332 * -1 if there was a guest wakeup IPI (which has now been cleared)
2333 */
2334kvmppc_read_intr:
2335 /* see if a host IPI is pending */
2336 li r3, 1
2337 lbz r0, HSTATE_HOST_IPI(r13)
2338 cmpwi r0, 0
2339 bne 1f
371fefd6 2340
c934243c
PM
2341 /* Now read the interrupt from the ICP */
2342 ld r6, HSTATE_XICS_PHYS(r13)
19ccb76a 2343 li r7, XICS_XIRR
c934243c
PM
2344 cmpdi r6, 0
2345 beq- 1f
2346 lwzcix r0, r6, r7
2347 rlwinm. r3, r0, 0, 0xffffff
19ccb76a 2348 sync
c934243c 2349 beq 1f /* if nothing pending in the ICP */
371fefd6 2350
c934243c
PM
2351 /* We found something in the ICP...
2352 *
2353 * If it's not an IPI, stash it in the PACA and return to
2354 * the host, we don't (yet) handle directing real external
2355 * interrupts directly to the guest
2356 */
2357 cmpwi r3, XICS_IPI /* if there is, is it an IPI? */
c934243c 2358 bne 42f
371fefd6 2359
c934243c
PM
2360 /* It's an IPI, clear the MFRR and EOI it */
2361 li r3, 0xff
2362 li r8, XICS_MFRR
2363 stbcix r3, r6, r8 /* clear the IPI */
2364 stwcix r0, r6, r7 /* EOI it */
2365 sync
f0888f70 2366
c934243c
PM
2367 /* We need to re-check host IPI now in case it got set in the
2368 * meantime. If it's clear, we bounce the interrupt to the
2369 * guest
2370 */
2371 lbz r0, HSTATE_HOST_IPI(r13)
2372 cmpwi r0, 0
2373 bne- 43f
2374
2375 /* OK, it's an IPI for us */
2376 li r3, -1
23771: blr
2378
237942: /* It's not an IPI and it's for the host, stash it in the PACA
2380 * before exit, it will be picked up by the host ICP driver
2381 */
2382 stw r0, HSTATE_SAVED_XIRR(r13)
e3bbbbfa 2383 li r3, 1
c934243c
PM
2384 b 1b
2385
238643: /* We raced with the host, we need to resend that IPI, bummer */
2387 li r0, IPI_PRIORITY
2388 stbcix r0, r6, r8 /* set the IPI */
2389 sync
e3bbbbfa 2390 li r3, 1
c934243c 2391 b 1b
371fefd6 2392
de56a948
PM
2393/*
2394 * Save away FP, VMX and VSX registers.
2395 * r3 = vcpu pointer
595e4f7e
PM
2396 * N.B. r30 and r31 are volatile across this function,
2397 * thus it is not callable from C.
a8606e20 2398 */
595e4f7e
PM
2399kvmppc_save_fp:
2400 mflr r30
2401 mr r31,r3
8943633c
PM
2402 mfmsr r5
2403 ori r8,r5,MSR_FP
de56a948
PM
2404#ifdef CONFIG_ALTIVEC
2405BEGIN_FTR_SECTION
2406 oris r8,r8,MSR_VEC@h
2407END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2408#endif
2409#ifdef CONFIG_VSX
2410BEGIN_FTR_SECTION
2411 oris r8,r8,MSR_VSX@h
2412END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2413#endif
2414 mtmsrd r8
2415 isync
595e4f7e
PM
2416 addi r3,r3,VCPU_FPRS
2417 bl .store_fp_state
de56a948
PM
2418#ifdef CONFIG_ALTIVEC
2419BEGIN_FTR_SECTION
595e4f7e
PM
2420 addi r3,r31,VCPU_VRS
2421 bl .store_vr_state
de56a948
PM
2422END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2423#endif
2424 mfspr r6,SPRN_VRSAVE
e724f080 2425 stw r6,VCPU_VRSAVE(r31)
595e4f7e 2426 mtlr r30
de56a948
PM
2427 blr
2428
2429/*
2430 * Load up FP, VMX and VSX registers
2431 * r4 = vcpu pointer
595e4f7e
PM
2432 * N.B. r30 and r31 are volatile across this function,
2433 * thus it is not callable from C.
de56a948 2434 */
de56a948 2435kvmppc_load_fp:
595e4f7e
PM
2436 mflr r30
2437 mr r31,r4
de56a948
PM
2438 mfmsr r9
2439 ori r8,r9,MSR_FP
2440#ifdef CONFIG_ALTIVEC
2441BEGIN_FTR_SECTION
2442 oris r8,r8,MSR_VEC@h
2443END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2444#endif
2445#ifdef CONFIG_VSX
2446BEGIN_FTR_SECTION
2447 oris r8,r8,MSR_VSX@h
2448END_FTR_SECTION_IFSET(CPU_FTR_VSX)
2449#endif
2450 mtmsrd r8
2451 isync
595e4f7e
PM
2452 addi r3,r4,VCPU_FPRS
2453 bl .load_fp_state
de56a948
PM
2454#ifdef CONFIG_ALTIVEC
2455BEGIN_FTR_SECTION
595e4f7e
PM
2456 addi r3,r31,VCPU_VRS
2457 bl .load_vr_state
de56a948
PM
2458END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
2459#endif
e724f080 2460 lwz r7,VCPU_VRSAVE(r31)
de56a948 2461 mtspr SPRN_VRSAVE,r7
595e4f7e
PM
2462 mtlr r30
2463 mr r4,r31
de56a948 2464 blr
44a3add8
PM
2465
2466/*
2467 * We come here if we get any exception or interrupt while we are
2468 * executing host real mode code while in guest MMU context.
2469 * For now just spin, but we should do something better.
2470 */
2471kvmppc_bad_host_intr:
2472 b .
e4e38121
MN
2473
2474/*
2475 * This mimics the MSR transition on IRQ delivery. The new guest MSR is taken
2476 * from VCPU_INTR_MSR and is modified based on the required TM state changes.
2477 * r11 has the guest MSR value (in/out)
2478 * r9 has a vcpu pointer (in)
2479 * r0 is used as a scratch register
2480 */
2481kvmppc_msr_interrupt:
2482 rldicl r0, r11, 64 - MSR_TS_S_LG, 62
2483 cmpwi r0, 2 /* Check if we are in transactional state.. */
2484 ld r11, VCPU_INTR_MSR(r9)
2485 bne 1f
2486 /* ... if transactional, change to suspended */
2487 li r0, 1
24881: rldimi r11, r0, MSR_TS_S_LG, 63 - MSR_TS_T_LG
2489 blr
9bc01a9b
PM
2490
2491/*
2492 * This works around a hardware bug on POWER8E processors, where
2493 * writing a 1 to the MMCR0[PMAO] bit doesn't generate a
2494 * performance monitor interrupt. Instead, when we need to have
2495 * an interrupt pending, we have to arrange for a counter to overflow.
2496 */
2497kvmppc_fix_pmao:
2498 li r3, 0
2499 mtspr SPRN_MMCR2, r3
2500 lis r3, (MMCR0_PMXE | MMCR0_FCECE)@h
2501 ori r3, r3, MMCR0_PMCjCE | MMCR0_C56RUN
2502 mtspr SPRN_MMCR0, r3
2503 lis r3, 0x7fff
2504 ori r3, r3, 0xffff
2505 mtspr SPRN_PMC6, r3
2506 isync
2507 blr