xref: /qemu/hw/ppc/spapr_nested.c (revision 6ff5da16000f908140723e164d33a0b51a6c4162)
1 #include "qemu/osdep.h"
2 #include "qemu/cutils.h"
3 #include "exec/exec-all.h"
4 #include "exec/cputlb.h"
5 #include "helper_regs.h"
6 #include "hw/ppc/ppc.h"
7 #include "hw/ppc/spapr.h"
8 #include "hw/ppc/spapr_cpu_core.h"
9 #include "hw/ppc/spapr_nested.h"
10 #include "mmu-book3s-v3.h"
11 #include "cpu-models.h"
12 #include "qemu/log.h"
13 
14 void spapr_nested_reset(SpaprMachineState *spapr)
15 {
16     if (spapr_get_cap(spapr, SPAPR_CAP_NESTED_KVM_HV)) {
17         spapr_unregister_nested_hv();
18         spapr_register_nested_hv();
19     } else if (spapr_get_cap(spapr, SPAPR_CAP_NESTED_PAPR)) {
20         spapr->nested.capabilities_set = false;
21         spapr_unregister_nested_papr();
22         spapr_register_nested_papr();
23         spapr_nested_gsb_init();
24     } else {
25         spapr->nested.api = 0;
26     }
27 }
28 
29 uint8_t spapr_nested_api(SpaprMachineState *spapr)
30 {
31     return spapr->nested.api;
32 }
33 
34 #ifdef CONFIG_TCG
35 
36 bool spapr_get_pate_nested_hv(SpaprMachineState *spapr, PowerPCCPU *cpu,
37                               target_ulong lpid, ppc_v3_pate_t *entry)
38 {
39     uint64_t patb, pats;
40 
41     assert(lpid != 0);
42 
43     patb = spapr->nested.ptcr & PTCR_PATB;
44     pats = spapr->nested.ptcr & PTCR_PATS;
45 
46     /* Check if partition table is properly aligned */
47     if (patb & MAKE_64BIT_MASK(0, pats + 12)) {
48         return false;
49     }
50 
51     /* Calculate number of entries */
52     pats = 1ull << (pats + 12 - 4);
53     if (pats <= lpid) {
54         return false;
55     }
56 
57     /* Grab entry */
58     patb += 16 * lpid;
59     entry->dw0 = ldq_phys(CPU(cpu)->as, patb);
60     entry->dw1 = ldq_phys(CPU(cpu)->as, patb + 8);
61     return true;
62 }
63 
64 static
65 SpaprMachineStateNestedGuest *spapr_get_nested_guest(SpaprMachineState *spapr,
66                                                      target_ulong guestid)
67 {
68     SpaprMachineStateNestedGuest *guest;
69 
70     guest = g_hash_table_lookup(spapr->nested.guests, GINT_TO_POINTER(guestid));
71     return guest;
72 }
73 
74 bool spapr_get_pate_nested_papr(SpaprMachineState *spapr, PowerPCCPU *cpu,
75                                 target_ulong lpid, ppc_v3_pate_t *entry)
76 {
77     SpaprMachineStateNestedGuest *guest;
78     assert(lpid != 0);
79     guest = spapr_get_nested_guest(spapr, lpid);
80     if (!guest) {
81         return false;
82     }
83 
84     entry->dw0 = guest->parttbl[0];
85     entry->dw1 = guest->parttbl[1];
86     return true;
87 }
88 
89 #define PRTS_MASK      0x1f
90 
91 static target_ulong h_set_ptbl(PowerPCCPU *cpu,
92                                SpaprMachineState *spapr,
93                                target_ulong opcode,
94                                target_ulong *args)
95 {
96     target_ulong ptcr = args[0];
97 
98     if (!spapr_get_cap(spapr, SPAPR_CAP_NESTED_KVM_HV)) {
99         return H_FUNCTION;
100     }
101 
102     if ((ptcr & PRTS_MASK) + 12 - 4 > 12) {
103         return H_PARAMETER;
104     }
105 
106     spapr->nested.ptcr = ptcr; /* Save new partition table */
107 
108     return H_SUCCESS;
109 }
110 
111 static target_ulong h_tlb_invalidate(PowerPCCPU *cpu,
112                                      SpaprMachineState *spapr,
113                                      target_ulong opcode,
114                                      target_ulong *args)
115 {
116     /*
117      * The spapr virtual hypervisor nested HV implementation retains no L2
118      * translation state except for TLB. And the TLB is always invalidated
119      * across L1<->L2 transitions, so nothing is required here.
120      */
121 
122     return H_SUCCESS;
123 }
124 
125 static target_ulong h_copy_tofrom_guest(PowerPCCPU *cpu,
126                                         SpaprMachineState *spapr,
127                                         target_ulong opcode,
128                                         target_ulong *args)
129 {
130     /*
131      * This HCALL is not required, L1 KVM will take a slow path and walk the
132      * page tables manually to do the data copy.
133      */
134     return H_FUNCTION;
135 }
136 
137 static void nested_save_state(struct nested_ppc_state *save, PowerPCCPU *cpu)
138 {
139     CPUPPCState *env = &cpu->env;
140     SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
141 
142     memcpy(save->gpr, env->gpr, sizeof(save->gpr));
143 
144     save->lr = env->lr;
145     save->ctr = env->ctr;
146     save->cfar = env->cfar;
147     save->msr = env->msr;
148     save->nip = env->nip;
149 
150     save->cr = ppc_get_cr(env);
151     save->xer = cpu_read_xer(env);
152 
153     save->lpcr = env->spr[SPR_LPCR];
154     save->lpidr = env->spr[SPR_LPIDR];
155     save->pcr = env->spr[SPR_PCR];
156     save->dpdes = env->spr[SPR_DPDES];
157     save->hfscr = env->spr[SPR_HFSCR];
158     save->srr0 = env->spr[SPR_SRR0];
159     save->srr1 = env->spr[SPR_SRR1];
160     save->sprg0 = env->spr[SPR_SPRG0];
161     save->sprg1 = env->spr[SPR_SPRG1];
162     save->sprg2 = env->spr[SPR_SPRG2];
163     save->sprg3 = env->spr[SPR_SPRG3];
164     save->pidr = env->spr[SPR_BOOKS_PID];
165     save->ppr = env->spr[SPR_PPR];
166 
167     if (spapr_nested_api(spapr) == NESTED_API_PAPR) {
168         save->amor = env->spr[SPR_AMOR];
169         save->dawr0 = env->spr[SPR_DAWR0];
170         save->dawrx0 = env->spr[SPR_DAWRX0];
171         save->ciabr = env->spr[SPR_CIABR];
172         save->purr = env->spr[SPR_PURR];
173         save->spurr = env->spr[SPR_SPURR];
174         save->ic = env->spr[SPR_IC];
175         save->vtb = env->spr[SPR_VTB];
176         save->hdar = env->spr[SPR_HDAR];
177         save->hdsisr = env->spr[SPR_HDSISR];
178         save->heir = env->spr[SPR_HEIR];
179         save->asdr = env->spr[SPR_ASDR];
180         save->dawr1 = env->spr[SPR_DAWR1];
181         save->dawrx1 = env->spr[SPR_DAWRX1];
182         save->dexcr = env->spr[SPR_DEXCR];
183         save->hdexcr = env->spr[SPR_HDEXCR];
184         save->hashkeyr = env->spr[SPR_HASHKEYR];
185         save->hashpkeyr = env->spr[SPR_HASHPKEYR];
186         memcpy(save->vsr, env->vsr, sizeof(save->vsr));
187         save->ebbhr = env->spr[SPR_EBBHR];
188         save->tar = env->spr[SPR_TAR];
189         save->ebbrr = env->spr[SPR_EBBRR];
190         save->bescr = env->spr[SPR_BESCR];
191         save->iamr = env->spr[SPR_IAMR];
192         save->amr = env->spr[SPR_AMR];
193         save->uamor = env->spr[SPR_UAMOR];
194         save->dscr = env->spr[SPR_DSCR];
195         save->fscr = env->spr[SPR_FSCR];
196         save->pspb = env->spr[SPR_PSPB];
197         save->ctrl = env->spr[SPR_CTRL];
198         save->vrsave = env->spr[SPR_VRSAVE];
199         save->dar = env->spr[SPR_DAR];
200         save->dsisr = env->spr[SPR_DSISR];
201         save->pmc1 = env->spr[SPR_POWER_PMC1];
202         save->pmc2 = env->spr[SPR_POWER_PMC2];
203         save->pmc3 = env->spr[SPR_POWER_PMC3];
204         save->pmc4 = env->spr[SPR_POWER_PMC4];
205         save->pmc5 = env->spr[SPR_POWER_PMC5];
206         save->pmc6 = env->spr[SPR_POWER_PMC6];
207         save->mmcr0 = env->spr[SPR_POWER_MMCR0];
208         save->mmcr1 = env->spr[SPR_POWER_MMCR1];
209         save->mmcr2 = env->spr[SPR_POWER_MMCR2];
210         save->mmcra = env->spr[SPR_POWER_MMCRA];
211         save->sdar = env->spr[SPR_POWER_SDAR];
212         save->siar = env->spr[SPR_POWER_SIAR];
213         save->sier = env->spr[SPR_POWER_SIER];
214         save->vscr = ppc_get_vscr(env);
215         save->fpscr = env->fpscr;
216     } else if (spapr_nested_api(spapr) == NESTED_API_KVM_HV) {
217         save->tb_offset = env->tb_env->tb_offset;
218     }
219 }
220 
221 static void nested_post_load_state(CPUPPCState *env, CPUState *cs)
222 {
223     /*
224      * compute hflags and possible interrupts.
225      */
226     hreg_compute_hflags(env);
227     ppc_maybe_interrupt(env);
228     /*
229      * Nested HV does not tag TLB entries between L1 and L2, so must
230      * flush on transition.
231      */
232     tlb_flush(cs);
233     env->reserve_addr = -1; /* Reset the reservation */
234 }
235 
236 static void nested_load_state(PowerPCCPU *cpu, struct nested_ppc_state *load)
237 {
238     CPUPPCState *env = &cpu->env;
239     SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
240 
241     memcpy(env->gpr, load->gpr, sizeof(env->gpr));
242 
243     env->lr = load->lr;
244     env->ctr = load->ctr;
245     env->cfar = load->cfar;
246     env->msr = load->msr;
247     env->nip = load->nip;
248 
249     ppc_set_cr(env, load->cr);
250     cpu_write_xer(env, load->xer);
251 
252     env->spr[SPR_LPCR] = load->lpcr;
253     env->spr[SPR_LPIDR] = load->lpidr;
254     env->spr[SPR_PCR] = load->pcr;
255     env->spr[SPR_DPDES] = load->dpdes;
256     env->spr[SPR_HFSCR] = load->hfscr;
257     env->spr[SPR_SRR0] = load->srr0;
258     env->spr[SPR_SRR1] = load->srr1;
259     env->spr[SPR_SPRG0] = load->sprg0;
260     env->spr[SPR_SPRG1] = load->sprg1;
261     env->spr[SPR_SPRG2] = load->sprg2;
262     env->spr[SPR_SPRG3] = load->sprg3;
263     env->spr[SPR_BOOKS_PID] = load->pidr;
264     env->spr[SPR_PPR] = load->ppr;
265 
266     if (spapr_nested_api(spapr) == NESTED_API_PAPR) {
267         env->spr[SPR_AMOR] = load->amor;
268         env->spr[SPR_DAWR0] = load->dawr0;
269         env->spr[SPR_DAWRX0] = load->dawrx0;
270         env->spr[SPR_CIABR] = load->ciabr;
271         env->spr[SPR_PURR] = load->purr;
272         env->spr[SPR_SPURR] = load->purr;
273         env->spr[SPR_IC] = load->ic;
274         env->spr[SPR_VTB] = load->vtb;
275         env->spr[SPR_HDAR] = load->hdar;
276         env->spr[SPR_HDSISR] = load->hdsisr;
277         env->spr[SPR_HEIR] = load->heir;
278         env->spr[SPR_ASDR] = load->asdr;
279         env->spr[SPR_DAWR1] = load->dawr1;
280         env->spr[SPR_DAWRX1] = load->dawrx1;
281         env->spr[SPR_DEXCR] = load->dexcr;
282         env->spr[SPR_HDEXCR] = load->hdexcr;
283         env->spr[SPR_HASHKEYR] = load->hashkeyr;
284         env->spr[SPR_HASHPKEYR] = load->hashpkeyr;
285         memcpy(env->vsr, load->vsr, sizeof(env->vsr));
286         env->spr[SPR_EBBHR] = load->ebbhr;
287         env->spr[SPR_TAR] = load->tar;
288         env->spr[SPR_EBBRR] = load->ebbrr;
289         env->spr[SPR_BESCR] = load->bescr;
290         env->spr[SPR_IAMR] = load->iamr;
291         env->spr[SPR_AMR] = load->amr;
292         env->spr[SPR_UAMOR] = load->uamor;
293         env->spr[SPR_DSCR] = load->dscr;
294         env->spr[SPR_FSCR] = load->fscr;
295         env->spr[SPR_PSPB] = load->pspb;
296         env->spr[SPR_CTRL] = load->ctrl;
297         env->spr[SPR_VRSAVE] = load->vrsave;
298         env->spr[SPR_DAR] = load->dar;
299         env->spr[SPR_DSISR] = load->dsisr;
300         env->spr[SPR_POWER_PMC1] = load->pmc1;
301         env->spr[SPR_POWER_PMC2] = load->pmc2;
302         env->spr[SPR_POWER_PMC3] = load->pmc3;
303         env->spr[SPR_POWER_PMC4] = load->pmc4;
304         env->spr[SPR_POWER_PMC5] = load->pmc5;
305         env->spr[SPR_POWER_PMC6] = load->pmc6;
306         env->spr[SPR_POWER_MMCR0] = load->mmcr0;
307         env->spr[SPR_POWER_MMCR1] = load->mmcr1;
308         env->spr[SPR_POWER_MMCR2] = load->mmcr2;
309         env->spr[SPR_POWER_MMCRA] = load->mmcra;
310         env->spr[SPR_POWER_SDAR] = load->sdar;
311         env->spr[SPR_POWER_SIAR] = load->siar;
312         env->spr[SPR_POWER_SIER] = load->sier;
313         ppc_store_vscr(env, load->vscr);
314         ppc_store_fpscr(env, load->fpscr);
315     } else if (spapr_nested_api(spapr) == NESTED_API_KVM_HV) {
316         env->tb_env->tb_offset = load->tb_offset;
317     }
318 }
319 
320 /*
321  * When this handler returns, the environment is switched to the L2 guest
322  * and TCG begins running that. spapr_exit_nested() performs the switch from
323  * L2 back to L1 and returns from the H_ENTER_NESTED hcall.
324  */
325 static target_ulong h_enter_nested(PowerPCCPU *cpu,
326                                    SpaprMachineState *spapr,
327                                    target_ulong opcode,
328                                    target_ulong *args)
329 {
330     PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cpu);
331     CPUPPCState *env = &cpu->env;
332     CPUState *cs = CPU(cpu);
333     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
334     struct nested_ppc_state l2_state;
335     target_ulong hv_ptr = args[0];
336     target_ulong regs_ptr = args[1];
337     target_ulong hdec, now = cpu_ppc_load_tbl(env);
338     target_ulong lpcr, lpcr_mask;
339     struct kvmppc_hv_guest_state *hvstate;
340     struct kvmppc_hv_guest_state hv_state;
341     struct kvmppc_pt_regs *regs;
342     hwaddr len;
343 
344     if (spapr->nested.ptcr == 0) {
345         return H_NOT_AVAILABLE;
346     }
347 
348     len = sizeof(*hvstate);
349     hvstate = address_space_map(CPU(cpu)->as, hv_ptr, &len, false,
350                                 MEMTXATTRS_UNSPECIFIED);
351     if (len != sizeof(*hvstate)) {
352         address_space_unmap(CPU(cpu)->as, hvstate, len, 0, false);
353         return H_PARAMETER;
354     }
355 
356     memcpy(&hv_state, hvstate, len);
357 
358     address_space_unmap(CPU(cpu)->as, hvstate, len, len, false);
359 
360     /*
361      * We accept versions 1 and 2. Version 2 fields are unused because TCG
362      * does not implement DAWR*.
363      */
364     if (hv_state.version > HV_GUEST_STATE_VERSION) {
365         return H_PARAMETER;
366     }
367 
368     if (hv_state.lpid == 0) {
369         return H_PARAMETER;
370     }
371 
372     spapr_cpu->nested_host_state = g_try_new(struct nested_ppc_state, 1);
373     if (!spapr_cpu->nested_host_state) {
374         return H_NO_MEM;
375     }
376 
377     assert(env->spr[SPR_LPIDR] == 0);
378     assert(env->spr[SPR_DPDES] == 0);
379     nested_save_state(spapr_cpu->nested_host_state, cpu);
380 
381     len = sizeof(*regs);
382     regs = address_space_map(CPU(cpu)->as, regs_ptr, &len, false,
383                                 MEMTXATTRS_UNSPECIFIED);
384     if (!regs || len != sizeof(*regs)) {
385         address_space_unmap(CPU(cpu)->as, regs, len, 0, false);
386         g_free(spapr_cpu->nested_host_state);
387         return H_P2;
388     }
389 
390     len = sizeof(l2_state.gpr);
391     assert(len == sizeof(regs->gpr));
392     memcpy(l2_state.gpr, regs->gpr, len);
393 
394     l2_state.lr = regs->link;
395     l2_state.ctr = regs->ctr;
396     l2_state.xer = regs->xer;
397     l2_state.cr = regs->ccr;
398     l2_state.msr = regs->msr;
399     l2_state.nip = regs->nip;
400 
401     address_space_unmap(CPU(cpu)->as, regs, len, len, false);
402 
403     l2_state.cfar = hv_state.cfar;
404     l2_state.lpidr = hv_state.lpid;
405 
406     lpcr_mask = LPCR_DPFD | LPCR_ILE | LPCR_AIL | LPCR_LD | LPCR_MER;
407     lpcr = (env->spr[SPR_LPCR] & ~lpcr_mask) | (hv_state.lpcr & lpcr_mask);
408     lpcr |= LPCR_HR | LPCR_UPRT | LPCR_GTSE | LPCR_HVICE | LPCR_HDICE;
409     lpcr &= ~LPCR_LPES0;
410     l2_state.lpcr = lpcr & pcc->lpcr_mask;
411 
412     l2_state.pcr = hv_state.pcr;
413     /* hv_state.amor is not used */
414     l2_state.dpdes = hv_state.dpdes;
415     l2_state.hfscr = hv_state.hfscr;
416     /* TCG does not implement DAWR*, CIABR, PURR, SPURR, IC, VTB, HEIR SPRs*/
417     l2_state.srr0 = hv_state.srr0;
418     l2_state.srr1 = hv_state.srr1;
419     l2_state.sprg0 = hv_state.sprg[0];
420     l2_state.sprg1 = hv_state.sprg[1];
421     l2_state.sprg2 = hv_state.sprg[2];
422     l2_state.sprg3 = hv_state.sprg[3];
423     l2_state.pidr = hv_state.pidr;
424     l2_state.ppr = hv_state.ppr;
425     l2_state.tb_offset = env->tb_env->tb_offset + hv_state.tb_offset;
426 
427     /*
428      * Switch to the nested guest environment and start the "hdec" timer.
429      */
430     nested_load_state(cpu, &l2_state);
431     nested_post_load_state(env, cs);
432 
433     hdec = hv_state.hdec_expiry - now;
434     cpu_ppc_hdecr_init(env);
435     cpu_ppc_store_hdecr(env, hdec);
436 
437     /*
438      * The hv_state.vcpu_token is not needed. It is used by the KVM
439      * implementation to remember which L2 vCPU last ran on which physical
440      * CPU so as to invalidate process scope translations if it is moved
441      * between physical CPUs. For now TLBs are always flushed on L1<->L2
442      * transitions so this is not a problem.
443      *
444      * Could validate that the same vcpu_token does not attempt to run on
445      * different L1 vCPUs at the same time, but that would be a L1 KVM bug
446      * and it's not obviously worth a new data structure to do it.
447      */
448 
449     spapr_cpu->in_nested = true;
450 
451     /*
452      * The spapr hcall helper sets env->gpr[3] to the return value, but at
453      * this point the L1 is not returning from the hcall but rather we
454      * start running the L2, so r3 must not be clobbered, so return env->gpr[3]
455      * to leave it unchanged.
456      */
457     return env->gpr[3];
458 }
459 
460 static void spapr_exit_nested_hv(PowerPCCPU *cpu, int excp)
461 {
462     CPUPPCState *env = &cpu->env;
463     CPUState *cs = CPU(cpu);
464     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
465     struct nested_ppc_state l2_state;
466     target_ulong hv_ptr = spapr_cpu->nested_host_state->gpr[4];
467     target_ulong regs_ptr = spapr_cpu->nested_host_state->gpr[5];
468     target_ulong hsrr0, hsrr1, hdar, asdr, hdsisr;
469     struct kvmppc_hv_guest_state *hvstate;
470     struct kvmppc_pt_regs *regs;
471     hwaddr len;
472 
473     nested_save_state(&l2_state, cpu);
474     hsrr0 = env->spr[SPR_HSRR0];
475     hsrr1 = env->spr[SPR_HSRR1];
476     hdar = env->spr[SPR_HDAR];
477     hdsisr = env->spr[SPR_HDSISR];
478     asdr = env->spr[SPR_ASDR];
479 
480     /*
481      * Switch back to the host environment (including for any error).
482      */
483     assert(env->spr[SPR_LPIDR] != 0);
484     nested_load_state(cpu, spapr_cpu->nested_host_state);
485     nested_post_load_state(env, cs);
486     env->gpr[3] = env->excp_vectors[excp]; /* hcall return value */
487 
488     cpu_ppc_hdecr_exit(env);
489 
490     spapr_cpu->in_nested = false;
491 
492     g_free(spapr_cpu->nested_host_state);
493     spapr_cpu->nested_host_state = NULL;
494 
495     len = sizeof(*hvstate);
496     hvstate = address_space_map(CPU(cpu)->as, hv_ptr, &len, true,
497                                 MEMTXATTRS_UNSPECIFIED);
498     if (len != sizeof(*hvstate)) {
499         address_space_unmap(CPU(cpu)->as, hvstate, len, 0, true);
500         env->gpr[3] = H_PARAMETER;
501         return;
502     }
503 
504     hvstate->cfar = l2_state.cfar;
505     hvstate->lpcr = l2_state.lpcr;
506     hvstate->pcr = l2_state.pcr;
507     hvstate->dpdes = l2_state.dpdes;
508     hvstate->hfscr = l2_state.hfscr;
509 
510     if (excp == POWERPC_EXCP_HDSI) {
511         hvstate->hdar = hdar;
512         hvstate->hdsisr = hdsisr;
513         hvstate->asdr = asdr;
514     } else if (excp == POWERPC_EXCP_HISI) {
515         hvstate->asdr = asdr;
516     }
517 
518     /* HEIR should be implemented for HV mode and saved here. */
519     hvstate->srr0 = l2_state.srr0;
520     hvstate->srr1 = l2_state.srr1;
521     hvstate->sprg[0] = l2_state.sprg0;
522     hvstate->sprg[1] = l2_state.sprg1;
523     hvstate->sprg[2] = l2_state.sprg2;
524     hvstate->sprg[3] = l2_state.sprg3;
525     hvstate->pidr = l2_state.pidr;
526     hvstate->ppr = l2_state.ppr;
527 
528     /* Is it okay to specify write length larger than actual data written? */
529     address_space_unmap(CPU(cpu)->as, hvstate, len, len, true);
530 
531     len = sizeof(*regs);
532     regs = address_space_map(CPU(cpu)->as, regs_ptr, &len, true,
533                                 MEMTXATTRS_UNSPECIFIED);
534     if (!regs || len != sizeof(*regs)) {
535         address_space_unmap(CPU(cpu)->as, regs, len, 0, true);
536         env->gpr[3] = H_P2;
537         return;
538     }
539 
540     len = sizeof(env->gpr);
541     assert(len == sizeof(regs->gpr));
542     memcpy(regs->gpr, l2_state.gpr, len);
543 
544     regs->link = l2_state.lr;
545     regs->ctr = l2_state.ctr;
546     regs->xer = l2_state.xer;
547     regs->ccr = l2_state.cr;
548 
549     if (excp == POWERPC_EXCP_MCHECK ||
550         excp == POWERPC_EXCP_RESET ||
551         excp == POWERPC_EXCP_SYSCALL) {
552         regs->nip = l2_state.srr0;
553         regs->msr = l2_state.srr1 & env->msr_mask;
554     } else {
555         regs->nip = hsrr0;
556         regs->msr = hsrr1 & env->msr_mask;
557     }
558 
559     /* Is it okay to specify write length larger than actual data written? */
560     address_space_unmap(CPU(cpu)->as, regs, len, len, true);
561 }
562 
563 static bool spapr_nested_vcpu_check(SpaprMachineStateNestedGuest *guest,
564                                     target_ulong vcpuid, bool inoutbuf)
565 {
566     struct SpaprMachineStateNestedGuestVcpu *vcpu;
567     /*
568      * Perform sanity checks for the provided vcpuid of a guest.
569      * For now, ensure its valid, allocated and enabled for use.
570      */
571 
572     if (vcpuid >= PAPR_NESTED_GUEST_VCPU_MAX) {
573         return false;
574     }
575 
576     if (!(vcpuid < guest->nr_vcpus)) {
577         return false;
578     }
579 
580     vcpu = &guest->vcpus[vcpuid];
581     if (!vcpu->enabled) {
582         return false;
583     }
584 
585     if (!inoutbuf) {
586         return true;
587     }
588 
589     /* Check to see if the in/out buffers are registered */
590     if (vcpu->runbufin.addr && vcpu->runbufout.addr) {
591         return true;
592     }
593 
594     return false;
595 }
596 
597 static void *get_vcpu_state_ptr(SpaprMachineStateNestedGuest *guest,
598                               target_ulong vcpuid)
599 {
600     assert(spapr_nested_vcpu_check(guest, vcpuid, false));
601     return &guest->vcpus[vcpuid].state;
602 }
603 
604 static void *get_vcpu_ptr(SpaprMachineStateNestedGuest *guest,
605                                    target_ulong vcpuid)
606 {
607     assert(spapr_nested_vcpu_check(guest, vcpuid, false));
608     return &guest->vcpus[vcpuid];
609 }
610 
611 static void *get_guest_ptr(SpaprMachineStateNestedGuest *guest,
612                            target_ulong vcpuid)
613 {
614     return guest; /* for GSBE_NESTED */
615 }
616 
617 /*
618  * set=1 means the L1 is trying to set some state
619  * set=0 means the L1 is trying to get some state
620  */
621 static void copy_state_8to8(void *a, void *b, bool set)
622 {
623     /* set takes from the Big endian element_buf and sets internal buffer */
624 
625     if (set) {
626         *(uint64_t *)a = be64_to_cpu(*(uint64_t *)b);
627     } else {
628         *(uint64_t *)b = cpu_to_be64(*(uint64_t *)a);
629     }
630 }
631 
632 static void copy_state_4to4(void *a, void *b, bool set)
633 {
634     if (set) {
635         *(uint32_t *)a = be32_to_cpu(*(uint32_t *)b);
636     } else {
637         *(uint32_t *)b = cpu_to_be32(*((uint32_t *)a));
638     }
639 }
640 
641 static void copy_state_16to16(void *a, void *b, bool set)
642 {
643     uint64_t *src, *dst;
644 
645     if (set) {
646         src = b;
647         dst = a;
648 
649         dst[1] = be64_to_cpu(src[0]);
650         dst[0] = be64_to_cpu(src[1]);
651     } else {
652         src = a;
653         dst = b;
654 
655         dst[1] = cpu_to_be64(src[0]);
656         dst[0] = cpu_to_be64(src[1]);
657     }
658 }
659 
660 static void copy_state_4to8(void *a, void *b, bool set)
661 {
662     if (set) {
663         *(uint64_t *)a  = (uint64_t) be32_to_cpu(*(uint32_t *)b);
664     } else {
665         *(uint32_t *)b = cpu_to_be32((uint32_t) (*((uint64_t *)a)));
666     }
667 }
668 
669 static void copy_state_pagetbl(void *a, void *b, bool set)
670 {
671     uint64_t *pagetbl;
672     uint64_t *buf; /* 3 double words */
673     uint64_t rts;
674 
675     assert(set);
676 
677     pagetbl = a;
678     buf = b;
679 
680     *pagetbl = be64_to_cpu(buf[0]);
681     /* as per ISA section 6.7.6.1 */
682     *pagetbl |= PATE0_HR; /* Host Radix bit is 1 */
683 
684     /* RTS */
685     rts = be64_to_cpu(buf[1]);
686     assert(rts == 52);
687     rts = rts - 31; /* since radix tree size = 2^(RTS+31) */
688     *pagetbl |=  ((rts & 0x7) << 5); /* RTS2 is bit 56:58 */
689     *pagetbl |=  (((rts >> 3) & 0x3) << 61); /* RTS1 is bit 1:2 */
690 
691     /* RPDS {Size = 2^(RPDS+3) , RPDS >=5} */
692     *pagetbl |= 63 - clz64(be64_to_cpu(buf[2])) - 3;
693 }
694 
695 static void copy_state_proctbl(void *a, void *b, bool set)
696 {
697     uint64_t *proctbl;
698     uint64_t *buf; /* 2 double words */
699 
700     assert(set);
701 
702     proctbl = a;
703     buf = b;
704     /* PRTB: Process Table Base */
705     *proctbl = be64_to_cpu(buf[0]);
706     /* PRTS: Process Table Size = 2^(12+PRTS) */
707     if (be64_to_cpu(buf[1]) == (1ULL << 12)) {
708             *proctbl |= 0;
709     } else if (be64_to_cpu(buf[1]) == (1ULL << 24)) {
710             *proctbl |= 12;
711     } else {
712         g_assert_not_reached();
713     }
714 }
715 
716 static void copy_state_runbuf(void *a, void *b, bool set)
717 {
718     uint64_t *buf; /* 2 double words */
719     struct SpaprMachineStateNestedGuestVcpuRunBuf *runbuf;
720 
721     assert(set);
722 
723     runbuf = a;
724     buf = b;
725 
726     runbuf->addr = be64_to_cpu(buf[0]);
727     assert(runbuf->addr);
728 
729     /* per spec */
730     assert(be64_to_cpu(buf[1]) <= 16384);
731 
732     /*
733      * This will also hit in the input buffer but should be fine for
734      * now. If not we can split this function.
735      */
736     assert(be64_to_cpu(buf[1]) >= VCPU_OUT_BUF_MIN_SZ);
737 
738     runbuf->size = be64_to_cpu(buf[1]);
739 }
740 
741 /* tell the L1 how big we want the output vcpu run buffer */
742 static void out_buf_min_size(void *a, void *b, bool set)
743 {
744     uint64_t *buf; /* 1 double word */
745 
746     assert(!set);
747 
748     buf = b;
749 
750     buf[0] = cpu_to_be64(VCPU_OUT_BUF_MIN_SZ);
751 }
752 
753 static void copy_logical_pvr(void *a, void *b, bool set)
754 {
755     SpaprMachineStateNestedGuest *guest;
756     uint32_t *buf; /* 1 word */
757     uint32_t *pvr_logical_ptr;
758     uint32_t pvr_logical;
759     target_ulong pcr = 0;
760 
761     pvr_logical_ptr = a;
762     buf = b;
763 
764     if (!set) {
765         buf[0] = cpu_to_be32(*pvr_logical_ptr);
766         return;
767     }
768 
769     pvr_logical = be32_to_cpu(buf[0]);
770 
771     *pvr_logical_ptr = pvr_logical;
772 
773     if (*pvr_logical_ptr) {
774         switch (*pvr_logical_ptr) {
775         case CPU_POWERPC_LOGICAL_3_10_P11:
776         case CPU_POWERPC_LOGICAL_3_10:
777             pcr = PCR_COMPAT_3_10 | PCR_COMPAT_3_00;
778             break;
779         case CPU_POWERPC_LOGICAL_3_00:
780             pcr = PCR_COMPAT_3_00;
781             break;
782         default:
783             qemu_log_mask(LOG_GUEST_ERROR,
784                           "Could not set PCR for LPVR=0x%08x\n",
785                           *pvr_logical_ptr);
786             return;
787         }
788     }
789 
790     guest = container_of(pvr_logical_ptr,
791                          struct SpaprMachineStateNestedGuest,
792                          pvr_logical);
793     for (int i = 0; i < guest->nr_vcpus; i++) {
794         guest->vcpus[i].state.pcr = ~pcr | HVMASK_PCR;
795     }
796 }
797 
798 static void copy_tb_offset(void *a, void *b, bool set)
799 {
800     SpaprMachineStateNestedGuest *guest;
801     uint64_t *buf; /* 1 double word */
802     uint64_t *tb_offset_ptr;
803     uint64_t tb_offset;
804 
805     tb_offset_ptr = a;
806     buf = b;
807 
808     if (!set) {
809         buf[0] = cpu_to_be64(*tb_offset_ptr);
810         return;
811     }
812 
813     tb_offset = be64_to_cpu(buf[0]);
814     /* need to copy this to the individual tb_offset for each vcpu */
815     guest = container_of(tb_offset_ptr,
816                          struct SpaprMachineStateNestedGuest,
817                          tb_offset);
818     for (int i = 0; i < guest->nr_vcpus; i++) {
819         guest->vcpus[i].tb_offset = tb_offset;
820     }
821 }
822 
823 static void copy_state_hdecr(void *a, void *b, bool set)
824 {
825     uint64_t *buf; /* 1 double word */
826     uint64_t *hdecr_expiry_tb;
827 
828     hdecr_expiry_tb = a;
829     buf = b;
830 
831     if (!set) {
832         buf[0] = cpu_to_be64(*hdecr_expiry_tb);
833         return;
834     }
835 
836     *hdecr_expiry_tb = be64_to_cpu(buf[0]);
837 }
838 
839 struct guest_state_element_type guest_state_element_types[] = {
840     GUEST_STATE_ELEMENT_NOP(GSB_HV_VCPU_IGNORED_ID, 0),
841     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR0,  gpr[0]),
842     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR1,  gpr[1]),
843     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR2,  gpr[2]),
844     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR3,  gpr[3]),
845     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR4,  gpr[4]),
846     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR5,  gpr[5]),
847     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR6,  gpr[6]),
848     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR7,  gpr[7]),
849     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR8,  gpr[8]),
850     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR9,  gpr[9]),
851     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR10, gpr[10]),
852     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR11, gpr[11]),
853     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR12, gpr[12]),
854     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR13, gpr[13]),
855     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR14, gpr[14]),
856     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR15, gpr[15]),
857     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR16, gpr[16]),
858     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR17, gpr[17]),
859     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR18, gpr[18]),
860     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR19, gpr[19]),
861     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR20, gpr[20]),
862     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR21, gpr[21]),
863     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR22, gpr[22]),
864     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR23, gpr[23]),
865     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR24, gpr[24]),
866     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR25, gpr[25]),
867     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR26, gpr[26]),
868     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR27, gpr[27]),
869     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR28, gpr[28]),
870     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR29, gpr[29]),
871     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR30, gpr[30]),
872     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_GPR31, gpr[31]),
873     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_NIA, nip),
874     GSE_ENV_DWM(GSB_VCPU_SPR_MSR, msr, HVMASK_MSR),
875     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_CTR, ctr),
876     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_LR, lr),
877     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_XER, xer),
878     GUEST_STATE_ELEMENT_ENV_WW(GSB_VCPU_SPR_CR, cr),
879     GUEST_STATE_ELEMENT_NOP_DW(GSB_VCPU_SPR_MMCR3),
880     GUEST_STATE_ELEMENT_NOP_DW(GSB_VCPU_SPR_SIER2),
881     GUEST_STATE_ELEMENT_NOP_DW(GSB_VCPU_SPR_SIER3),
882     GUEST_STATE_ELEMENT_NOP_W(GSB_VCPU_SPR_WORT),
883     GSE_ENV_DWM(GSB_VCPU_SPR_LPCR, lpcr, HVMASK_LPCR),
884     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_AMOR, amor),
885     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_HFSCR, hfscr),
886     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_DAWR0, dawr0),
887     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_DAWRX0, dawrx0),
888     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_CIABR, ciabr),
889     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_PURR,  purr),
890     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_SPURR, spurr),
891     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_IC,    ic),
892     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_VTB,   vtb),
893     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_HDAR,  hdar),
894     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_HDSISR, hdsisr),
895     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_HEIR,   heir),
896     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_ASDR,  asdr),
897     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_SRR0,  srr0),
898     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_SRR1,  srr1),
899     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_SPRG0, sprg0),
900     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_SPRG1, sprg1),
901     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_SPRG2, sprg2),
902     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_SPRG3, sprg3),
903     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_PIDR,   pidr),
904     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_CFAR,  cfar),
905     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_PPR,   ppr),
906     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_DAWR1, dawr1),
907     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_DAWRX1, dawrx1),
908     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_DEXCR, dexcr),
909     GSE_ENV_DWM(GSB_VCPU_SPR_HDEXCR, hdexcr, HVMASK_HDEXCR),
910     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_HASHKEYR, hashkeyr),
911     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_HASHPKEYR, hashpkeyr),
912     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR0, vsr[0]),
913     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR1, vsr[1]),
914     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR2, vsr[2]),
915     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR3, vsr[3]),
916     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR4, vsr[4]),
917     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR5, vsr[5]),
918     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR6, vsr[6]),
919     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR7, vsr[7]),
920     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR8, vsr[8]),
921     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR9, vsr[9]),
922     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR10, vsr[10]),
923     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR11, vsr[11]),
924     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR12, vsr[12]),
925     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR13, vsr[13]),
926     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR14, vsr[14]),
927     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR15, vsr[15]),
928     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR16, vsr[16]),
929     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR17, vsr[17]),
930     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR18, vsr[18]),
931     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR19, vsr[19]),
932     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR20, vsr[20]),
933     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR21, vsr[21]),
934     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR22, vsr[22]),
935     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR23, vsr[23]),
936     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR24, vsr[24]),
937     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR25, vsr[25]),
938     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR26, vsr[26]),
939     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR27, vsr[27]),
940     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR28, vsr[28]),
941     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR29, vsr[29]),
942     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR30, vsr[30]),
943     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR31, vsr[31]),
944     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR32, vsr[32]),
945     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR33, vsr[33]),
946     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR34, vsr[34]),
947     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR35, vsr[35]),
948     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR36, vsr[36]),
949     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR37, vsr[37]),
950     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR38, vsr[38]),
951     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR39, vsr[39]),
952     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR40, vsr[40]),
953     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR41, vsr[41]),
954     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR42, vsr[42]),
955     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR43, vsr[43]),
956     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR44, vsr[44]),
957     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR45, vsr[45]),
958     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR46, vsr[46]),
959     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR47, vsr[47]),
960     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR48, vsr[48]),
961     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR49, vsr[49]),
962     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR50, vsr[50]),
963     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR51, vsr[51]),
964     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR52, vsr[52]),
965     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR53, vsr[53]),
966     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR54, vsr[54]),
967     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR55, vsr[55]),
968     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR56, vsr[56]),
969     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR57, vsr[57]),
970     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR58, vsr[58]),
971     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR59, vsr[59]),
972     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR60, vsr[60]),
973     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR61, vsr[61]),
974     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR62, vsr[62]),
975     GUEST_STATE_ELEMENT_ENV_QW(GSB_VCPU_SPR_VSR63, vsr[63]),
976     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_EBBHR, ebbhr),
977     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_TAR,   tar),
978     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_EBBRR, ebbrr),
979     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_BESCR, bescr),
980     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_IAMR,  iamr),
981     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_AMR,   amr),
982     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_UAMOR, uamor),
983     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_DSCR,  dscr),
984     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_FSCR,  fscr),
985     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_PSPB,   pspb),
986     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_CTRL,  ctrl),
987     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_DPDES, dpdes),
988     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_VRSAVE, vrsave),
989     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_DAR,   dar),
990     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_DSISR,  dsisr),
991     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_PMC1,   pmc1),
992     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_PMC2,   pmc2),
993     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_PMC3,   pmc3),
994     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_PMC4,   pmc4),
995     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_PMC5,   pmc5),
996     GUEST_STATE_ELEMENT_ENV_W(GSB_VCPU_SPR_PMC6,   pmc6),
997     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_MMCR0, mmcr0),
998     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_MMCR1, mmcr1),
999     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_MMCR2, mmcr2),
1000     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_MMCRA, mmcra),
1001     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_SDAR , sdar),
1002     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_SIAR , siar),
1003     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_SIER , sier),
1004     GUEST_STATE_ELEMENT_ENV_WW(GSB_VCPU_SPR_VSCR,  vscr),
1005     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_SPR_FPSCR, fpscr),
1006     GUEST_STATE_ELEMENT_ENV_DW(GSB_VCPU_DEC_EXPIRE_TB, dec_expiry_tb),
1007     GSBE_NESTED(GSB_PART_SCOPED_PAGETBL, 0x18, parttbl[0],  copy_state_pagetbl),
1008     GSBE_NESTED(GSB_PROCESS_TBL,         0x10, parttbl[1],  copy_state_proctbl),
1009     GSBE_NESTED(GSB_VCPU_LPVR,           0x4,  pvr_logical, copy_logical_pvr),
1010     GSBE_NESTED_MSK(GSB_TB_OFFSET, 0x8, tb_offset, copy_tb_offset,
1011                     HVMASK_TB_OFFSET),
1012     GSBE_NESTED_VCPU(GSB_VCPU_IN_BUFFER, 0x10, runbufin,    copy_state_runbuf),
1013     GSBE_NESTED_VCPU(GSB_VCPU_OUT_BUFFER, 0x10, runbufout,   copy_state_runbuf),
1014     GSBE_NESTED_VCPU(GSB_VCPU_OUT_BUF_MIN_SZ, 0x8, runbufout, out_buf_min_size),
1015     GSBE_NESTED_VCPU(GSB_VCPU_HDEC_EXPIRY_TB, 0x8, hdecr_expiry_tb,
1016                      copy_state_hdecr)
1017 };
1018 
1019 void spapr_nested_gsb_init(void)
1020 {
1021     struct guest_state_element_type *type;
1022 
1023     /* Init the guest state elements lookup table, flags for now */
1024     for (int i = 0; i < ARRAY_SIZE(guest_state_element_types); i++) {
1025         type = &guest_state_element_types[i];
1026 
1027         assert(type->id <= GSB_LAST);
1028         if (type->id >= GSB_VCPU_SPR_HDAR)
1029             /* 0xf000 - 0xf005 Thread + RO */
1030             type->flags = GUEST_STATE_ELEMENT_TYPE_FLAG_READ_ONLY;
1031         else if (type->id >= GSB_VCPU_IN_BUFFER)
1032             /* 0x0c00 - 0xf000 Thread + RW */
1033             type->flags = 0;
1034         else if (type->id >= GSB_VCPU_LPVR)
1035             /* 0x0003 - 0x0bff Guest + RW */
1036             type->flags = GUEST_STATE_ELEMENT_TYPE_FLAG_GUEST_WIDE;
1037         else if (type->id >= GSB_HV_VCPU_STATE_SIZE)
1038             /* 0x0001 - 0x0002 Guest + RO */
1039             type->flags = GUEST_STATE_ELEMENT_TYPE_FLAG_READ_ONLY |
1040                           GUEST_STATE_ELEMENT_TYPE_FLAG_GUEST_WIDE;
1041     }
1042 }
1043 
1044 static struct guest_state_element *guest_state_element_next(
1045     struct guest_state_element *element,
1046     int64_t *len,
1047     int64_t *num_elements)
1048 {
1049     uint16_t size;
1050 
1051     /* size is of element->value[] only. Not whole guest_state_element */
1052     size = be16_to_cpu(element->size);
1053 
1054     if (len) {
1055         *len -= size + offsetof(struct guest_state_element, value);
1056     }
1057 
1058     if (num_elements) {
1059         *num_elements -= 1;
1060     }
1061 
1062     return (struct guest_state_element *)(element->value + size);
1063 }
1064 
1065 static
1066 struct guest_state_element_type *guest_state_element_type_find(uint16_t id)
1067 {
1068     int i;
1069 
1070     for (i = 0; i < ARRAY_SIZE(guest_state_element_types); i++)
1071         if (id == guest_state_element_types[i].id) {
1072             return &guest_state_element_types[i];
1073         }
1074 
1075     return NULL;
1076 }
1077 
1078 static void log_element(struct guest_state_element *element,
1079                         struct guest_state_request *gsr)
1080 {
1081     qemu_log_mask(LOG_GUEST_ERROR, "h_guest_%s_state id:0x%04x size:0x%04x",
1082                   gsr->flags & GUEST_STATE_REQUEST_SET ? "set" : "get",
1083                   be16_to_cpu(element->id), be16_to_cpu(element->size));
1084     qemu_log_mask(LOG_GUEST_ERROR, "buf:0x%016"PRIx64" ...\n",
1085                   be64_to_cpu(*(uint64_t *)element->value));
1086 }
1087 
1088 static bool guest_state_request_check(struct guest_state_request *gsr)
1089 {
1090     int64_t num_elements, len = gsr->len;
1091     struct guest_state_buffer *gsb = gsr->gsb;
1092     struct guest_state_element *element;
1093     struct guest_state_element_type *type;
1094     uint16_t id, size;
1095 
1096     /* gsb->num_elements = 0 == 32 bits long */
1097     assert(len >= 4);
1098 
1099     num_elements = be32_to_cpu(gsb->num_elements);
1100     element = gsb->elements;
1101     len -= sizeof(gsb->num_elements);
1102 
1103     /* Walk the buffer to validate the length */
1104     while (num_elements) {
1105 
1106         id = be16_to_cpu(element->id);
1107         size = be16_to_cpu(element->size);
1108 
1109         if (false) {
1110             log_element(element, gsr);
1111         }
1112         /* buffer size too small */
1113         if (len < 0) {
1114             return false;
1115         }
1116 
1117         type = guest_state_element_type_find(id);
1118         if (!type) {
1119             qemu_log_mask(LOG_GUEST_ERROR, "Element ID %04x unknown\n", id);
1120             log_element(element, gsr);
1121             return false;
1122         }
1123 
1124         if (id == GSB_HV_VCPU_IGNORED_ID) {
1125             goto next_element;
1126         }
1127 
1128         if (size != type->size) {
1129             qemu_log_mask(LOG_GUEST_ERROR, "Size mismatch. Element ID:%04x."
1130                           "Size Exp:%i Got:%i\n", id, type->size, size);
1131             log_element(element, gsr);
1132             return false;
1133         }
1134 
1135         if ((type->flags & GUEST_STATE_ELEMENT_TYPE_FLAG_READ_ONLY) &&
1136             (gsr->flags & GUEST_STATE_REQUEST_SET)) {
1137             qemu_log_mask(LOG_GUEST_ERROR, "Trying to set a read-only Element "
1138                           "ID:%04x.\n", id);
1139             return false;
1140         }
1141 
1142         if (type->flags & GUEST_STATE_ELEMENT_TYPE_FLAG_GUEST_WIDE) {
1143             /* guest wide element type */
1144             if (!(gsr->flags & GUEST_STATE_REQUEST_GUEST_WIDE)) {
1145                 qemu_log_mask(LOG_GUEST_ERROR, "trying to set a guest wide "
1146                               "Element ID:%04x.\n", id);
1147                 return false;
1148             }
1149         } else {
1150             /* thread wide element type */
1151             if (gsr->flags & GUEST_STATE_REQUEST_GUEST_WIDE) {
1152                 qemu_log_mask(LOG_GUEST_ERROR, "trying to set a thread wide "
1153                               "Element ID:%04x.\n", id);
1154                 return false;
1155             }
1156         }
1157 next_element:
1158         element = guest_state_element_next(element, &len, &num_elements);
1159 
1160     }
1161     return true;
1162 }
1163 
1164 static bool is_gsr_invalid(struct guest_state_request *gsr,
1165                                    struct guest_state_element *element,
1166                                    struct guest_state_element_type *type)
1167 {
1168     if ((gsr->flags & GUEST_STATE_REQUEST_SET) &&
1169         (*(uint64_t *)(element->value) & ~(type->mask))) {
1170         log_element(element, gsr);
1171         qemu_log_mask(LOG_GUEST_ERROR, "L1 can't set reserved bits "
1172                       "(allowed mask: 0x%08"PRIx64")\n", type->mask);
1173         return true;
1174     }
1175     return false;
1176 }
1177 
1178 static target_ulong h_guest_get_capabilities(PowerPCCPU *cpu,
1179                                              SpaprMachineState *spapr,
1180                                              target_ulong opcode,
1181                                              target_ulong *args)
1182 {
1183     CPUPPCState *env = &cpu->env;
1184     target_ulong flags = args[0];
1185 
1186     if (flags) { /* don't handle any flags capabilities for now */
1187         return H_PARAMETER;
1188     }
1189 
1190     /* P11 capabilities */
1191     if (ppc_check_compat(cpu, CPU_POWERPC_LOGICAL_3_10_P11, 0,
1192         spapr->max_compat_pvr)) {
1193         env->gpr[4] |= H_GUEST_CAPABILITIES_P11_MODE;
1194     }
1195 
1196     /* P10 capabilities */
1197     if (ppc_check_compat(cpu, CPU_POWERPC_LOGICAL_3_10, 0,
1198         spapr->max_compat_pvr)) {
1199         env->gpr[4] |= H_GUEST_CAPABILITIES_P10_MODE;
1200     }
1201 
1202     /* P9 capabilities */
1203     if (ppc_check_compat(cpu, CPU_POWERPC_LOGICAL_3_00, 0,
1204         spapr->max_compat_pvr)) {
1205         env->gpr[4] |= H_GUEST_CAPABILITIES_P9_MODE;
1206     }
1207 
1208     return H_SUCCESS;
1209 }
1210 
1211 static target_ulong h_guest_set_capabilities(PowerPCCPU *cpu,
1212                                              SpaprMachineState *spapr,
1213                                              target_ulong opcode,
1214                                               target_ulong *args)
1215 {
1216     CPUPPCState *env = &cpu->env;
1217     target_ulong flags = args[0];
1218     target_ulong capabilities = args[1];
1219     env->gpr[4] = 0;
1220 
1221     if (flags) { /* don't handle any flags capabilities for now */
1222         return H_PARAMETER;
1223     }
1224 
1225     if (capabilities & H_GUEST_CAPABILITIES_COPY_MEM) {
1226         env->gpr[4] = 1;
1227         return H_P2; /* isn't supported */
1228     }
1229 
1230     /*
1231      * If there are no capabilities configured, set the R5 to the index of
1232      * the first supported Power Processor Mode
1233      */
1234     if (!capabilities) {
1235         env->gpr[4] = 1;
1236 
1237         /* set R5 to the first supported Power Processor Mode */
1238         if (ppc_check_compat(cpu, CPU_POWERPC_LOGICAL_3_10_P11, 0,
1239                              spapr->max_compat_pvr)) {
1240             env->gpr[5] = H_GUEST_CAP_P11_MODE_BMAP;
1241         } else if (ppc_check_compat(cpu, CPU_POWERPC_LOGICAL_3_10, 0,
1242                              spapr->max_compat_pvr)) {
1243             env->gpr[5] = H_GUEST_CAP_P10_MODE_BMAP;
1244         } else if (ppc_check_compat(cpu, CPU_POWERPC_LOGICAL_3_00, 0,
1245                                     spapr->max_compat_pvr)) {
1246             env->gpr[5] = H_GUEST_CAP_P9_MODE_BMAP;
1247         }
1248 
1249         return H_P2;
1250     }
1251 
1252     /*
1253      * If an invalid capability is set, R5 should contain the index of the
1254      * invalid capability bit
1255      */
1256     if (capabilities & ~H_GUEST_CAP_VALID_MASK) {
1257         env->gpr[4] = 1;
1258 
1259         /* Set R5 to the index of the invalid capability */
1260         env->gpr[5] = 63 - ctz64(capabilities);
1261 
1262         return H_P2;
1263     }
1264 
1265     if (!spapr->nested.capabilities_set) {
1266         spapr->nested.capabilities_set = true;
1267         spapr->nested.pvr_base = env->spr[SPR_PVR];
1268         return H_SUCCESS;
1269     } else {
1270         return H_STATE;
1271     }
1272 }
1273 
1274 static void
1275 destroy_guest_helper(gpointer value)
1276 {
1277     struct SpaprMachineStateNestedGuest *guest = value;
1278     g_free(guest->vcpus);
1279     g_free(guest);
1280 }
1281 
1282 static target_ulong h_guest_create(PowerPCCPU *cpu,
1283                                    SpaprMachineState *spapr,
1284                                    target_ulong opcode,
1285                                    target_ulong *args)
1286 {
1287     CPUPPCState *env = &cpu->env;
1288     target_ulong flags = args[0];
1289     target_ulong continue_token = args[1];
1290     uint64_t guestid;
1291     int nguests = 0;
1292     struct SpaprMachineStateNestedGuest *guest;
1293 
1294     if (flags) { /* don't handle any flags for now */
1295         return H_UNSUPPORTED_FLAG;
1296     }
1297 
1298     if (continue_token != -1) {
1299         return H_P2;
1300     }
1301 
1302     if (!spapr->nested.capabilities_set) {
1303         return H_STATE;
1304     }
1305 
1306     if (!spapr->nested.guests) {
1307         spapr->nested.guests = g_hash_table_new_full(NULL,
1308                                                      NULL,
1309                                                      NULL,
1310                                                      destroy_guest_helper);
1311     }
1312 
1313     nguests = g_hash_table_size(spapr->nested.guests);
1314 
1315     if (nguests == PAPR_NESTED_GUEST_MAX) {
1316         return H_NO_MEM;
1317     }
1318 
1319     /* Lookup for available guestid */
1320     for (guestid = 1; guestid < PAPR_NESTED_GUEST_MAX; guestid++) {
1321         if (!(g_hash_table_lookup(spapr->nested.guests,
1322                                   GINT_TO_POINTER(guestid)))) {
1323             break;
1324         }
1325     }
1326 
1327     if (guestid == PAPR_NESTED_GUEST_MAX) {
1328         return H_NO_MEM;
1329     }
1330 
1331     guest = g_try_new0(struct SpaprMachineStateNestedGuest, 1);
1332     if (!guest) {
1333         return H_NO_MEM;
1334     }
1335 
1336     guest->pvr_logical = spapr->nested.pvr_base;
1337     g_hash_table_insert(spapr->nested.guests, GINT_TO_POINTER(guestid), guest);
1338     env->gpr[4] = guestid;
1339 
1340     return H_SUCCESS;
1341 }
1342 
1343 static target_ulong h_guest_delete(PowerPCCPU *cpu,
1344                                    SpaprMachineState *spapr,
1345                                    target_ulong opcode,
1346                                    target_ulong *args)
1347 {
1348     target_ulong flags = args[0];
1349     target_ulong guestid = args[1];
1350     struct SpaprMachineStateNestedGuest *guest;
1351 
1352     /*
1353      * handle flag deleteAllGuests, if set:
1354      * guestid is ignored and all guests are deleted
1355      *
1356      */
1357     if (flags & ~H_GUEST_DELETE_ALL_FLAG) {
1358         return H_UNSUPPORTED_FLAG; /* other flag bits reserved */
1359     } else if (flags & H_GUEST_DELETE_ALL_FLAG) {
1360         g_hash_table_destroy(spapr->nested.guests);
1361         return H_SUCCESS;
1362     }
1363 
1364     guest = g_hash_table_lookup(spapr->nested.guests, GINT_TO_POINTER(guestid));
1365     if (!guest) {
1366         return H_P2;
1367     }
1368 
1369     g_hash_table_remove(spapr->nested.guests, GINT_TO_POINTER(guestid));
1370 
1371     return H_SUCCESS;
1372 }
1373 
1374 static target_ulong h_guest_create_vcpu(PowerPCCPU *cpu,
1375                                         SpaprMachineState *spapr,
1376                                         target_ulong opcode,
1377                                         target_ulong *args)
1378 {
1379     target_ulong flags = args[0];
1380     target_ulong guestid = args[1];
1381     target_ulong vcpuid = args[2];
1382     SpaprMachineStateNestedGuest *guest;
1383 
1384     if (flags) { /* don't handle any flags for now */
1385         return H_UNSUPPORTED_FLAG;
1386     }
1387 
1388     guest = spapr_get_nested_guest(spapr, guestid);
1389     if (!guest) {
1390         return H_P2;
1391     }
1392 
1393     if (vcpuid < guest->nr_vcpus) {
1394         qemu_log_mask(LOG_UNIMP, "vcpuid " TARGET_FMT_ld " already in use.",
1395                       vcpuid);
1396         return H_IN_USE;
1397     }
1398     /* linear vcpuid allocation only */
1399     assert(vcpuid == guest->nr_vcpus);
1400 
1401     if (guest->nr_vcpus >= PAPR_NESTED_GUEST_VCPU_MAX) {
1402         return H_P3;
1403     }
1404 
1405     SpaprMachineStateNestedGuestVcpu *vcpus, *curr_vcpu;
1406     vcpus = g_try_renew(struct SpaprMachineStateNestedGuestVcpu,
1407                         guest->vcpus,
1408                         guest->nr_vcpus + 1);
1409     if (!vcpus) {
1410         return H_NO_MEM;
1411     }
1412     guest->vcpus = vcpus;
1413     curr_vcpu = &vcpus[guest->nr_vcpus];
1414     memset(curr_vcpu, 0, sizeof(SpaprMachineStateNestedGuestVcpu));
1415 
1416     curr_vcpu->enabled = true;
1417     guest->nr_vcpus++;
1418 
1419     return H_SUCCESS;
1420 }
1421 
1422 static target_ulong getset_state(SpaprMachineStateNestedGuest *guest,
1423                                  uint64_t vcpuid,
1424                                  struct guest_state_request *gsr)
1425 {
1426     void *ptr;
1427     uint16_t id;
1428     struct guest_state_element *element;
1429     struct guest_state_element_type *type;
1430     int64_t lenleft, num_elements;
1431 
1432     lenleft = gsr->len;
1433 
1434     if (!guest_state_request_check(gsr)) {
1435         return H_P3;
1436     }
1437 
1438     num_elements = be32_to_cpu(gsr->gsb->num_elements);
1439     element = gsr->gsb->elements;
1440     /* Process the elements */
1441     while (num_elements) {
1442         type = NULL;
1443         /* log_element(element, gsr); */
1444 
1445         id = be16_to_cpu(element->id);
1446         if (id == GSB_HV_VCPU_IGNORED_ID) {
1447             goto next_element;
1448         }
1449 
1450         type = guest_state_element_type_find(id);
1451         assert(type);
1452 
1453         /* Get pointer to guest data to get/set */
1454         if (type->location && type->copy) {
1455             ptr = type->location(guest, vcpuid);
1456             assert(ptr);
1457             if (!~(type->mask) && is_gsr_invalid(gsr, element, type)) {
1458                 return H_INVALID_ELEMENT_VALUE;
1459             }
1460             type->copy(ptr + type->offset, element->value,
1461                        gsr->flags & GUEST_STATE_REQUEST_SET ? true : false);
1462         }
1463 
1464 next_element:
1465         element = guest_state_element_next(element, &lenleft, &num_elements);
1466     }
1467 
1468     return H_SUCCESS;
1469 }
1470 
1471 static target_ulong map_and_getset_state(PowerPCCPU *cpu,
1472                                          SpaprMachineStateNestedGuest *guest,
1473                                          uint64_t vcpuid,
1474                                          struct guest_state_request *gsr)
1475 {
1476     target_ulong rc;
1477     int64_t len;
1478     bool is_write;
1479 
1480     len = gsr->len;
1481     /* only get_state would require write access to the provided buffer */
1482     is_write = (gsr->flags & GUEST_STATE_REQUEST_SET) ? false : true;
1483     gsr->gsb = address_space_map(CPU(cpu)->as, gsr->buf, (uint64_t *)&len,
1484                                  is_write, MEMTXATTRS_UNSPECIFIED);
1485     if (!gsr->gsb) {
1486         rc = H_P3;
1487         goto out1;
1488     }
1489 
1490     if (len != gsr->len) {
1491         rc = H_P3;
1492         goto out1;
1493     }
1494 
1495     rc = getset_state(guest, vcpuid, gsr);
1496 
1497 out1:
1498     address_space_unmap(CPU(cpu)->as, gsr->gsb, len, is_write, len);
1499     return rc;
1500 }
1501 
1502 static target_ulong h_guest_getset_state(PowerPCCPU *cpu,
1503                                          SpaprMachineState *spapr,
1504                                          target_ulong *args,
1505                                          bool set)
1506 {
1507     target_ulong flags = args[0];
1508     target_ulong lpid = args[1];
1509     target_ulong vcpuid = args[2];
1510     target_ulong buf = args[3];
1511     target_ulong buflen = args[4];
1512     struct guest_state_request gsr;
1513     SpaprMachineStateNestedGuest *guest;
1514 
1515     guest = spapr_get_nested_guest(spapr, lpid);
1516     if (!guest) {
1517         return H_P2;
1518     }
1519     gsr.buf = buf;
1520     assert(buflen <= GSB_MAX_BUF_SIZE);
1521     gsr.len = buflen;
1522     gsr.flags = 0;
1523     if (flags & H_GUEST_GETSET_STATE_FLAG_GUEST_WIDE) {
1524         gsr.flags |= GUEST_STATE_REQUEST_GUEST_WIDE;
1525     }
1526     if (flags & ~H_GUEST_GETSET_STATE_FLAG_GUEST_WIDE) {
1527         return H_PARAMETER; /* flag not supported yet */
1528     }
1529 
1530     if (set) {
1531         gsr.flags |= GUEST_STATE_REQUEST_SET;
1532     }
1533     return map_and_getset_state(cpu, guest, vcpuid, &gsr);
1534 }
1535 
1536 static target_ulong h_guest_set_state(PowerPCCPU *cpu,
1537                                       SpaprMachineState *spapr,
1538                                       target_ulong opcode,
1539                                       target_ulong *args)
1540 {
1541     return h_guest_getset_state(cpu, spapr, args, true);
1542 }
1543 
1544 static target_ulong h_guest_get_state(PowerPCCPU *cpu,
1545                                       SpaprMachineState *spapr,
1546                                       target_ulong opcode,
1547                                       target_ulong *args)
1548 {
1549     return h_guest_getset_state(cpu, spapr, args, false);
1550 }
1551 
1552 static void exit_nested_store_l2(PowerPCCPU *cpu, int excp,
1553                                  SpaprMachineStateNestedGuestVcpu *vcpu)
1554 {
1555     CPUPPCState *env = &cpu->env;
1556     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
1557     target_ulong now, hdar, hdsisr, asdr;
1558 
1559     assert(sizeof(env->gpr) == sizeof(vcpu->state.gpr)); /* sanity check */
1560 
1561     now = cpu_ppc_load_tbl(env); /* L2 timebase */
1562     now -= vcpu->tb_offset; /* L1 timebase */
1563     vcpu->state.dec_expiry_tb = now - cpu_ppc_load_decr(env);
1564     cpu_ppc_store_decr(env, spapr_cpu->nested_host_state->dec_expiry_tb - now);
1565     /* backup hdar, hdsisr, asdr if reqd later below */
1566     hdar   = vcpu->state.hdar;
1567     hdsisr = vcpu->state.hdsisr;
1568     asdr   = vcpu->state.asdr;
1569 
1570     nested_save_state(&vcpu->state, cpu);
1571 
1572     if (excp == POWERPC_EXCP_MCHECK ||
1573         excp == POWERPC_EXCP_RESET ||
1574         excp == POWERPC_EXCP_SYSCALL) {
1575         vcpu->state.nip = env->spr[SPR_SRR0];
1576         vcpu->state.msr = env->spr[SPR_SRR1] & env->msr_mask;
1577     } else {
1578         vcpu->state.nip = env->spr[SPR_HSRR0];
1579         vcpu->state.msr = env->spr[SPR_HSRR1] & env->msr_mask;
1580     }
1581 
1582     /* hdar, hdsisr, asdr should be retained unless certain exceptions */
1583     if ((excp != POWERPC_EXCP_HDSI) && (excp != POWERPC_EXCP_HISI)) {
1584         vcpu->state.asdr = asdr;
1585     } else if (excp != POWERPC_EXCP_HDSI) {
1586         vcpu->state.hdar   = hdar;
1587         vcpu->state.hdsisr = hdsisr;
1588     }
1589 }
1590 
1591 static int get_exit_ids(uint64_t srr0, uint16_t ids[16])
1592 {
1593     int nr;
1594 
1595     switch (srr0) {
1596     case 0xc00:
1597         nr = 10;
1598         ids[0] = GSB_VCPU_GPR3;
1599         ids[1] = GSB_VCPU_GPR4;
1600         ids[2] = GSB_VCPU_GPR5;
1601         ids[3] = GSB_VCPU_GPR6;
1602         ids[4] = GSB_VCPU_GPR7;
1603         ids[5] = GSB_VCPU_GPR8;
1604         ids[6] = GSB_VCPU_GPR9;
1605         ids[7] = GSB_VCPU_GPR10;
1606         ids[8] = GSB_VCPU_GPR11;
1607         ids[9] = GSB_VCPU_GPR12;
1608         break;
1609     case 0xe00:
1610         nr = 5;
1611         ids[0] = GSB_VCPU_SPR_HDAR;
1612         ids[1] = GSB_VCPU_SPR_HDSISR;
1613         ids[2] = GSB_VCPU_SPR_ASDR;
1614         ids[3] = GSB_VCPU_SPR_NIA;
1615         ids[4] = GSB_VCPU_SPR_MSR;
1616         break;
1617     case 0xe20:
1618         nr = 4;
1619         ids[0] = GSB_VCPU_SPR_HDAR;
1620         ids[1] = GSB_VCPU_SPR_ASDR;
1621         ids[2] = GSB_VCPU_SPR_NIA;
1622         ids[3] = GSB_VCPU_SPR_MSR;
1623         break;
1624     case 0xe40:
1625         nr = 3;
1626         ids[0] = GSB_VCPU_SPR_HEIR;
1627         ids[1] = GSB_VCPU_SPR_NIA;
1628         ids[2] = GSB_VCPU_SPR_MSR;
1629         break;
1630     case 0xf80:
1631         nr = 3;
1632         ids[0] = GSB_VCPU_SPR_HFSCR;
1633         ids[1] = GSB_VCPU_SPR_NIA;
1634         ids[2] = GSB_VCPU_SPR_MSR;
1635         break;
1636     default:
1637         nr = 0;
1638         break;
1639     }
1640 
1641     return nr;
1642 }
1643 
1644 static void exit_process_output_buffer(PowerPCCPU *cpu,
1645                                        SpaprMachineStateNestedGuest *guest,
1646                                        target_ulong vcpuid,
1647                                        target_ulong *r3)
1648 {
1649     SpaprMachineStateNestedGuestVcpu *vcpu = &guest->vcpus[vcpuid];
1650     struct guest_state_request gsr;
1651     struct guest_state_buffer *gsb;
1652     struct guest_state_element *element;
1653     struct guest_state_element_type *type;
1654     int exit_id_count = 0;
1655     uint16_t exit_cause_ids[16];
1656     hwaddr len;
1657 
1658     len = vcpu->runbufout.size;
1659     gsb = address_space_map(CPU(cpu)->as, vcpu->runbufout.addr, &len, true,
1660                             MEMTXATTRS_UNSPECIFIED);
1661     if (!gsb || len != vcpu->runbufout.size) {
1662         address_space_unmap(CPU(cpu)->as, gsb, len, true, len);
1663         *r3 = H_P2;
1664         return;
1665     }
1666 
1667     exit_id_count = get_exit_ids(*r3, exit_cause_ids);
1668 
1669     /* Create a buffer of elements to send back */
1670     gsb->num_elements = cpu_to_be32(exit_id_count);
1671     element = gsb->elements;
1672     for (int i = 0; i < exit_id_count; i++) {
1673         type = guest_state_element_type_find(exit_cause_ids[i]);
1674         assert(type);
1675         element->id = cpu_to_be16(exit_cause_ids[i]);
1676         element->size = cpu_to_be16(type->size);
1677         element = guest_state_element_next(element, NULL, NULL);
1678     }
1679     gsr.gsb = gsb;
1680     gsr.len = VCPU_OUT_BUF_MIN_SZ;
1681     gsr.flags = 0; /* get + never guest wide */
1682     getset_state(guest, vcpuid, &gsr);
1683 
1684     address_space_unmap(CPU(cpu)->as, gsb, len, true, len);
1685     return;
1686 }
1687 
1688 static
1689 void spapr_exit_nested_papr(SpaprMachineState *spapr, PowerPCCPU *cpu, int excp)
1690 {
1691     CPUPPCState *env = &cpu->env;
1692     CPUState *cs = CPU(cpu);
1693     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
1694     target_ulong r3_return = env->excp_vectors[excp]; /* hcall return value */
1695     target_ulong lpid = 0, vcpuid = 0;
1696     struct SpaprMachineStateNestedGuestVcpu *vcpu = NULL;
1697     struct SpaprMachineStateNestedGuest *guest = NULL;
1698 
1699     lpid = spapr_cpu->nested_host_state->gpr[5];
1700     vcpuid = spapr_cpu->nested_host_state->gpr[6];
1701     guest = spapr_get_nested_guest(spapr, lpid);
1702     assert(guest);
1703     spapr_nested_vcpu_check(guest, vcpuid, false);
1704     vcpu = &guest->vcpus[vcpuid];
1705 
1706     exit_nested_store_l2(cpu, excp, vcpu);
1707     /* do the output buffer for run_vcpu*/
1708     exit_process_output_buffer(cpu, guest, vcpuid, &r3_return);
1709 
1710     assert(env->spr[SPR_LPIDR] != 0);
1711     nested_load_state(cpu, spapr_cpu->nested_host_state);
1712     cpu_ppc_decrease_tb_by_offset(env, vcpu->tb_offset);
1713     env->gpr[3] = H_SUCCESS;
1714     env->gpr[4] = r3_return;
1715     nested_post_load_state(env, cs);
1716     cpu_ppc_hdecr_exit(env);
1717 
1718     spapr_cpu->in_nested = false;
1719     g_free(spapr_cpu->nested_host_state);
1720     spapr_cpu->nested_host_state = NULL;
1721 }
1722 
1723 void spapr_exit_nested(PowerPCCPU *cpu, int excp)
1724 {
1725     SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1726     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
1727 
1728     assert(spapr_cpu->in_nested);
1729     if (spapr_nested_api(spapr) == NESTED_API_KVM_HV) {
1730         spapr_exit_nested_hv(cpu, excp);
1731     } else if (spapr_nested_api(spapr) == NESTED_API_PAPR) {
1732         spapr_exit_nested_papr(spapr, cpu, excp);
1733     } else {
1734         g_assert_not_reached();
1735     }
1736 }
1737 
1738 static void nested_papr_load_l2(PowerPCCPU *cpu,
1739                                 CPUPPCState *env,
1740                                 SpaprMachineStateNestedGuestVcpu *vcpu,
1741                                 target_ulong now)
1742 {
1743     PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cpu);
1744     target_ulong lpcr, lpcr_mask, hdec;
1745     lpcr_mask = LPCR_DPFD | LPCR_ILE | LPCR_AIL | LPCR_LD | LPCR_MER;
1746 
1747     assert(vcpu);
1748     assert(sizeof(env->gpr) == sizeof(vcpu->state.gpr));
1749     nested_load_state(cpu, &vcpu->state);
1750     lpcr = (env->spr[SPR_LPCR] & ~lpcr_mask) |
1751            (vcpu->state.lpcr & lpcr_mask);
1752     lpcr |= LPCR_HR | LPCR_UPRT | LPCR_GTSE | LPCR_HVICE | LPCR_HDICE;
1753     lpcr &= ~LPCR_LPES0;
1754     env->spr[SPR_LPCR] = lpcr & pcc->lpcr_mask;
1755 
1756     hdec = vcpu->hdecr_expiry_tb - now;
1757     cpu_ppc_store_decr(env, vcpu->state.dec_expiry_tb - now);
1758     cpu_ppc_hdecr_init(env);
1759     cpu_ppc_store_hdecr(env, hdec);
1760 
1761     cpu_ppc_increase_tb_by_offset(env, vcpu->tb_offset);
1762 }
1763 
1764 static void nested_papr_run_vcpu(PowerPCCPU *cpu,
1765                                  uint64_t lpid,
1766                                  SpaprMachineStateNestedGuestVcpu *vcpu)
1767 {
1768     SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1769     CPUPPCState *env = &cpu->env;
1770     CPUState *cs = CPU(cpu);
1771     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
1772     target_ulong now = cpu_ppc_load_tbl(env);
1773 
1774     assert(env->spr[SPR_LPIDR] == 0);
1775     assert(spapr->nested.api); /* ensure API version is initialized */
1776     spapr_cpu->nested_host_state = g_try_new(struct nested_ppc_state, 1);
1777     assert(spapr_cpu->nested_host_state);
1778     nested_save_state(spapr_cpu->nested_host_state, cpu);
1779     spapr_cpu->nested_host_state->dec_expiry_tb = now - cpu_ppc_load_decr(env);
1780     nested_papr_load_l2(cpu, env, vcpu, now);
1781     env->spr[SPR_LPIDR] = lpid; /* post load l2 */
1782 
1783     spapr_cpu->in_nested = true;
1784     nested_post_load_state(env, cs);
1785 }
1786 
1787 static target_ulong h_guest_run_vcpu(PowerPCCPU *cpu,
1788                                      SpaprMachineState *spapr,
1789                                      target_ulong opcode,
1790                                      target_ulong *args)
1791 {
1792     CPUPPCState *env = &cpu->env;
1793     target_ulong flags = args[0];
1794     target_ulong lpid = args[1];
1795     target_ulong vcpuid = args[2];
1796     struct SpaprMachineStateNestedGuestVcpu *vcpu;
1797     struct guest_state_request gsr;
1798     SpaprMachineStateNestedGuest *guest;
1799     target_ulong rc;
1800 
1801     if (flags) /* don't handle any flags for now */
1802         return H_PARAMETER;
1803 
1804     guest = spapr_get_nested_guest(spapr, lpid);
1805     if (!guest) {
1806         return H_P2;
1807     }
1808     if (!spapr_nested_vcpu_check(guest, vcpuid, true)) {
1809         return H_P3;
1810     }
1811 
1812     if (guest->parttbl[0] == 0) {
1813         /* At least need a partition scoped radix tree */
1814         return H_NOT_AVAILABLE;
1815     }
1816 
1817     vcpu = &guest->vcpus[vcpuid];
1818 
1819     /* Read run_vcpu input buffer to update state */
1820     gsr.buf = vcpu->runbufin.addr;
1821     gsr.len = vcpu->runbufin.size;
1822     gsr.flags = GUEST_STATE_REQUEST_SET; /* Thread wide + writing */
1823     rc = map_and_getset_state(cpu, guest, vcpuid, &gsr);
1824     if (rc == H_SUCCESS) {
1825         nested_papr_run_vcpu(cpu, lpid, vcpu);
1826     } else {
1827         env->gpr[3] = rc;
1828     }
1829     return env->gpr[3];
1830 }
1831 
1832 void spapr_register_nested_hv(void)
1833 {
1834     spapr_register_hypercall(KVMPPC_H_SET_PARTITION_TABLE, h_set_ptbl);
1835     spapr_register_hypercall(KVMPPC_H_ENTER_NESTED, h_enter_nested);
1836     spapr_register_hypercall(KVMPPC_H_TLB_INVALIDATE, h_tlb_invalidate);
1837     spapr_register_hypercall(KVMPPC_H_COPY_TOFROM_GUEST, h_copy_tofrom_guest);
1838 }
1839 
1840 void spapr_unregister_nested_hv(void)
1841 {
1842     spapr_unregister_hypercall(KVMPPC_H_SET_PARTITION_TABLE);
1843     spapr_unregister_hypercall(KVMPPC_H_ENTER_NESTED);
1844     spapr_unregister_hypercall(KVMPPC_H_TLB_INVALIDATE);
1845     spapr_unregister_hypercall(KVMPPC_H_COPY_TOFROM_GUEST);
1846 }
1847 
1848 void spapr_register_nested_papr(void)
1849 {
1850     spapr_register_hypercall(H_GUEST_GET_CAPABILITIES,
1851                              h_guest_get_capabilities);
1852     spapr_register_hypercall(H_GUEST_SET_CAPABILITIES,
1853                              h_guest_set_capabilities);
1854     spapr_register_hypercall(H_GUEST_CREATE, h_guest_create);
1855     spapr_register_hypercall(H_GUEST_DELETE, h_guest_delete);
1856     spapr_register_hypercall(H_GUEST_CREATE_VCPU, h_guest_create_vcpu);
1857     spapr_register_hypercall(H_GUEST_SET_STATE, h_guest_set_state);
1858     spapr_register_hypercall(H_GUEST_GET_STATE, h_guest_get_state);
1859     spapr_register_hypercall(H_GUEST_RUN_VCPU, h_guest_run_vcpu);
1860 }
1861 
1862 void spapr_unregister_nested_papr(void)
1863 {
1864     spapr_unregister_hypercall(H_GUEST_GET_CAPABILITIES);
1865     spapr_unregister_hypercall(H_GUEST_SET_CAPABILITIES);
1866     spapr_unregister_hypercall(H_GUEST_CREATE);
1867     spapr_unregister_hypercall(H_GUEST_DELETE);
1868     spapr_unregister_hypercall(H_GUEST_CREATE_VCPU);
1869     spapr_unregister_hypercall(H_GUEST_SET_STATE);
1870     spapr_unregister_hypercall(H_GUEST_GET_STATE);
1871     spapr_unregister_hypercall(H_GUEST_RUN_VCPU);
1872 }
1873 
1874 #else
1875 void spapr_exit_nested(PowerPCCPU *cpu, int excp)
1876 {
1877     g_assert_not_reached();
1878 }
1879 
1880 void spapr_register_nested_hv(void)
1881 {
1882     /* DO NOTHING */
1883 }
1884 
1885 void spapr_unregister_nested_hv(void)
1886 {
1887     /* DO NOTHING */
1888 }
1889 
1890 bool spapr_get_pate_nested_hv(SpaprMachineState *spapr, PowerPCCPU *cpu,
1891                               target_ulong lpid, ppc_v3_pate_t *entry)
1892 {
1893     return false;
1894 }
1895 
1896 bool spapr_get_pate_nested_papr(SpaprMachineState *spapr, PowerPCCPU *cpu,
1897                                 target_ulong lpid, ppc_v3_pate_t *entry)
1898 {
1899     return false;
1900 }
1901 
1902 void spapr_register_nested_papr(void)
1903 {
1904     /* DO NOTHING */
1905 }
1906 
1907 void spapr_unregister_nested_papr(void)
1908 {
1909     /* DO NOTHING */
1910 }
1911 
1912 void spapr_nested_gsb_init(void)
1913 {
1914     /* DO NOTHING */
1915 }
1916 
1917 #endif
1918