xref: /openbmc/qemu/hw/ppc/spapr_hcall.c (revision 4b9fa0b4)
1 #include "qemu/osdep.h"
2 #include "qapi/error.h"
3 #include "sysemu/hw_accel.h"
4 #include "sysemu/runstate.h"
5 #include "qemu/log.h"
6 #include "qemu/main-loop.h"
7 #include "qemu/module.h"
8 #include "qemu/error-report.h"
9 #include "cpu.h"
10 #include "exec/exec-all.h"
11 #include "helper_regs.h"
12 #include "hw/ppc/spapr.h"
13 #include "hw/ppc/spapr_cpu_core.h"
14 #include "mmu-hash64.h"
15 #include "cpu-models.h"
16 #include "trace.h"
17 #include "kvm_ppc.h"
18 #include "hw/ppc/spapr_ovec.h"
19 #include "mmu-book3s-v3.h"
20 #include "hw/mem/memory-device.h"
21 
22 static bool has_spr(PowerPCCPU *cpu, int spr)
23 {
24     /* We can test whether the SPR is defined by checking for a valid name */
25     return cpu->env.spr_cb[spr].name != NULL;
26 }
27 
28 static inline bool valid_ptex(PowerPCCPU *cpu, target_ulong ptex)
29 {
30     /*
31      * hash value/pteg group index is normalized by HPT mask
32      */
33     if (((ptex & ~7ULL) / HPTES_PER_GROUP) & ~ppc_hash64_hpt_mask(cpu)) {
34         return false;
35     }
36     return true;
37 }
38 
39 static bool is_ram_address(SpaprMachineState *spapr, hwaddr addr)
40 {
41     MachineState *machine = MACHINE(spapr);
42     DeviceMemoryState *dms = machine->device_memory;
43 
44     if (addr < machine->ram_size) {
45         return true;
46     }
47     if ((addr >= dms->base)
48         && ((addr - dms->base) < memory_region_size(&dms->mr))) {
49         return true;
50     }
51 
52     return false;
53 }
54 
55 static target_ulong h_enter(PowerPCCPU *cpu, SpaprMachineState *spapr,
56                             target_ulong opcode, target_ulong *args)
57 {
58     target_ulong flags = args[0];
59     target_ulong ptex = args[1];
60     target_ulong pteh = args[2];
61     target_ulong ptel = args[3];
62     unsigned apshift;
63     target_ulong raddr;
64     target_ulong slot;
65     const ppc_hash_pte64_t *hptes;
66 
67     apshift = ppc_hash64_hpte_page_shift_noslb(cpu, pteh, ptel);
68     if (!apshift) {
69         /* Bad page size encoding */
70         return H_PARAMETER;
71     }
72 
73     raddr = (ptel & HPTE64_R_RPN) & ~((1ULL << apshift) - 1);
74 
75     if (is_ram_address(spapr, raddr)) {
76         /* Regular RAM - should have WIMG=0010 */
77         if ((ptel & HPTE64_R_WIMG) != HPTE64_R_M) {
78             return H_PARAMETER;
79         }
80     } else {
81         target_ulong wimg_flags;
82         /* Looks like an IO address */
83         /* FIXME: What WIMG combinations could be sensible for IO?
84          * For now we allow WIMG=010x, but are there others? */
85         /* FIXME: Should we check against registered IO addresses? */
86         wimg_flags = (ptel & (HPTE64_R_W | HPTE64_R_I | HPTE64_R_M));
87 
88         if (wimg_flags != HPTE64_R_I &&
89             wimg_flags != (HPTE64_R_I | HPTE64_R_M)) {
90             return H_PARAMETER;
91         }
92     }
93 
94     pteh &= ~0x60ULL;
95 
96     if (!valid_ptex(cpu, ptex)) {
97         return H_PARAMETER;
98     }
99 
100     slot = ptex & 7ULL;
101     ptex = ptex & ~7ULL;
102 
103     if (likely((flags & H_EXACT) == 0)) {
104         hptes = ppc_hash64_map_hptes(cpu, ptex, HPTES_PER_GROUP);
105         for (slot = 0; slot < 8; slot++) {
106             if (!(ppc_hash64_hpte0(cpu, hptes, slot) & HPTE64_V_VALID)) {
107                 break;
108             }
109         }
110         ppc_hash64_unmap_hptes(cpu, hptes, ptex, HPTES_PER_GROUP);
111         if (slot == 8) {
112             return H_PTEG_FULL;
113         }
114     } else {
115         hptes = ppc_hash64_map_hptes(cpu, ptex + slot, 1);
116         if (ppc_hash64_hpte0(cpu, hptes, 0) & HPTE64_V_VALID) {
117             ppc_hash64_unmap_hptes(cpu, hptes, ptex + slot, 1);
118             return H_PTEG_FULL;
119         }
120         ppc_hash64_unmap_hptes(cpu, hptes, ptex, 1);
121     }
122 
123     spapr_store_hpte(cpu, ptex + slot, pteh | HPTE64_V_HPTE_DIRTY, ptel);
124 
125     args[0] = ptex + slot;
126     return H_SUCCESS;
127 }
128 
129 typedef enum {
130     REMOVE_SUCCESS = 0,
131     REMOVE_NOT_FOUND = 1,
132     REMOVE_PARM = 2,
133     REMOVE_HW = 3,
134 } RemoveResult;
135 
136 static RemoveResult remove_hpte(PowerPCCPU *cpu
137                                 , target_ulong ptex,
138                                 target_ulong avpn,
139                                 target_ulong flags,
140                                 target_ulong *vp, target_ulong *rp)
141 {
142     const ppc_hash_pte64_t *hptes;
143     target_ulong v, r;
144 
145     if (!valid_ptex(cpu, ptex)) {
146         return REMOVE_PARM;
147     }
148 
149     hptes = ppc_hash64_map_hptes(cpu, ptex, 1);
150     v = ppc_hash64_hpte0(cpu, hptes, 0);
151     r = ppc_hash64_hpte1(cpu, hptes, 0);
152     ppc_hash64_unmap_hptes(cpu, hptes, ptex, 1);
153 
154     if ((v & HPTE64_V_VALID) == 0 ||
155         ((flags & H_AVPN) && (v & ~0x7fULL) != avpn) ||
156         ((flags & H_ANDCOND) && (v & avpn) != 0)) {
157         return REMOVE_NOT_FOUND;
158     }
159     *vp = v;
160     *rp = r;
161     spapr_store_hpte(cpu, ptex, HPTE64_V_HPTE_DIRTY, 0);
162     ppc_hash64_tlb_flush_hpte(cpu, ptex, v, r);
163     return REMOVE_SUCCESS;
164 }
165 
166 static target_ulong h_remove(PowerPCCPU *cpu, SpaprMachineState *spapr,
167                              target_ulong opcode, target_ulong *args)
168 {
169     CPUPPCState *env = &cpu->env;
170     target_ulong flags = args[0];
171     target_ulong ptex = args[1];
172     target_ulong avpn = args[2];
173     RemoveResult ret;
174 
175     ret = remove_hpte(cpu, ptex, avpn, flags,
176                       &args[0], &args[1]);
177 
178     switch (ret) {
179     case REMOVE_SUCCESS:
180         check_tlb_flush(env, true);
181         return H_SUCCESS;
182 
183     case REMOVE_NOT_FOUND:
184         return H_NOT_FOUND;
185 
186     case REMOVE_PARM:
187         return H_PARAMETER;
188 
189     case REMOVE_HW:
190         return H_HARDWARE;
191     }
192 
193     g_assert_not_reached();
194 }
195 
196 #define H_BULK_REMOVE_TYPE             0xc000000000000000ULL
197 #define   H_BULK_REMOVE_REQUEST        0x4000000000000000ULL
198 #define   H_BULK_REMOVE_RESPONSE       0x8000000000000000ULL
199 #define   H_BULK_REMOVE_END            0xc000000000000000ULL
200 #define H_BULK_REMOVE_CODE             0x3000000000000000ULL
201 #define   H_BULK_REMOVE_SUCCESS        0x0000000000000000ULL
202 #define   H_BULK_REMOVE_NOT_FOUND      0x1000000000000000ULL
203 #define   H_BULK_REMOVE_PARM           0x2000000000000000ULL
204 #define   H_BULK_REMOVE_HW             0x3000000000000000ULL
205 #define H_BULK_REMOVE_RC               0x0c00000000000000ULL
206 #define H_BULK_REMOVE_FLAGS            0x0300000000000000ULL
207 #define   H_BULK_REMOVE_ABSOLUTE       0x0000000000000000ULL
208 #define   H_BULK_REMOVE_ANDCOND        0x0100000000000000ULL
209 #define   H_BULK_REMOVE_AVPN           0x0200000000000000ULL
210 #define H_BULK_REMOVE_PTEX             0x00ffffffffffffffULL
211 
212 #define H_BULK_REMOVE_MAX_BATCH        4
213 
214 static target_ulong h_bulk_remove(PowerPCCPU *cpu, SpaprMachineState *spapr,
215                                   target_ulong opcode, target_ulong *args)
216 {
217     CPUPPCState *env = &cpu->env;
218     int i;
219     target_ulong rc = H_SUCCESS;
220 
221     for (i = 0; i < H_BULK_REMOVE_MAX_BATCH; i++) {
222         target_ulong *tsh = &args[i*2];
223         target_ulong tsl = args[i*2 + 1];
224         target_ulong v, r, ret;
225 
226         if ((*tsh & H_BULK_REMOVE_TYPE) == H_BULK_REMOVE_END) {
227             break;
228         } else if ((*tsh & H_BULK_REMOVE_TYPE) != H_BULK_REMOVE_REQUEST) {
229             return H_PARAMETER;
230         }
231 
232         *tsh &= H_BULK_REMOVE_PTEX | H_BULK_REMOVE_FLAGS;
233         *tsh |= H_BULK_REMOVE_RESPONSE;
234 
235         if ((*tsh & H_BULK_REMOVE_ANDCOND) && (*tsh & H_BULK_REMOVE_AVPN)) {
236             *tsh |= H_BULK_REMOVE_PARM;
237             return H_PARAMETER;
238         }
239 
240         ret = remove_hpte(cpu, *tsh & H_BULK_REMOVE_PTEX, tsl,
241                           (*tsh & H_BULK_REMOVE_FLAGS) >> 26,
242                           &v, &r);
243 
244         *tsh |= ret << 60;
245 
246         switch (ret) {
247         case REMOVE_SUCCESS:
248             *tsh |= (r & (HPTE64_R_C | HPTE64_R_R)) << 43;
249             break;
250 
251         case REMOVE_PARM:
252             rc = H_PARAMETER;
253             goto exit;
254 
255         case REMOVE_HW:
256             rc = H_HARDWARE;
257             goto exit;
258         }
259     }
260  exit:
261     check_tlb_flush(env, true);
262 
263     return rc;
264 }
265 
266 static target_ulong h_protect(PowerPCCPU *cpu, SpaprMachineState *spapr,
267                               target_ulong opcode, target_ulong *args)
268 {
269     CPUPPCState *env = &cpu->env;
270     target_ulong flags = args[0];
271     target_ulong ptex = args[1];
272     target_ulong avpn = args[2];
273     const ppc_hash_pte64_t *hptes;
274     target_ulong v, r;
275 
276     if (!valid_ptex(cpu, ptex)) {
277         return H_PARAMETER;
278     }
279 
280     hptes = ppc_hash64_map_hptes(cpu, ptex, 1);
281     v = ppc_hash64_hpte0(cpu, hptes, 0);
282     r = ppc_hash64_hpte1(cpu, hptes, 0);
283     ppc_hash64_unmap_hptes(cpu, hptes, ptex, 1);
284 
285     if ((v & HPTE64_V_VALID) == 0 ||
286         ((flags & H_AVPN) && (v & ~0x7fULL) != avpn)) {
287         return H_NOT_FOUND;
288     }
289 
290     r &= ~(HPTE64_R_PP0 | HPTE64_R_PP | HPTE64_R_N |
291            HPTE64_R_KEY_HI | HPTE64_R_KEY_LO);
292     r |= (flags << 55) & HPTE64_R_PP0;
293     r |= (flags << 48) & HPTE64_R_KEY_HI;
294     r |= flags & (HPTE64_R_PP | HPTE64_R_N | HPTE64_R_KEY_LO);
295     spapr_store_hpte(cpu, ptex,
296                      (v & ~HPTE64_V_VALID) | HPTE64_V_HPTE_DIRTY, 0);
297     ppc_hash64_tlb_flush_hpte(cpu, ptex, v, r);
298     /* Flush the tlb */
299     check_tlb_flush(env, true);
300     /* Don't need a memory barrier, due to qemu's global lock */
301     spapr_store_hpte(cpu, ptex, v | HPTE64_V_HPTE_DIRTY, r);
302     return H_SUCCESS;
303 }
304 
305 static target_ulong h_read(PowerPCCPU *cpu, SpaprMachineState *spapr,
306                            target_ulong opcode, target_ulong *args)
307 {
308     target_ulong flags = args[0];
309     target_ulong ptex = args[1];
310     int i, ridx, n_entries = 1;
311     const ppc_hash_pte64_t *hptes;
312 
313     if (!valid_ptex(cpu, ptex)) {
314         return H_PARAMETER;
315     }
316 
317     if (flags & H_READ_4) {
318         /* Clear the two low order bits */
319         ptex &= ~(3ULL);
320         n_entries = 4;
321     }
322 
323     hptes = ppc_hash64_map_hptes(cpu, ptex, n_entries);
324     for (i = 0, ridx = 0; i < n_entries; i++) {
325         args[ridx++] = ppc_hash64_hpte0(cpu, hptes, i);
326         args[ridx++] = ppc_hash64_hpte1(cpu, hptes, i);
327     }
328     ppc_hash64_unmap_hptes(cpu, hptes, ptex, n_entries);
329 
330     return H_SUCCESS;
331 }
332 
333 struct SpaprPendingHpt {
334     /* These fields are read-only after initialization */
335     int shift;
336     QemuThread thread;
337 
338     /* These fields are protected by the BQL */
339     bool complete;
340 
341     /* These fields are private to the preparation thread if
342      * !complete, otherwise protected by the BQL */
343     int ret;
344     void *hpt;
345 };
346 
347 static void free_pending_hpt(SpaprPendingHpt *pending)
348 {
349     if (pending->hpt) {
350         qemu_vfree(pending->hpt);
351     }
352 
353     g_free(pending);
354 }
355 
356 static void *hpt_prepare_thread(void *opaque)
357 {
358     SpaprPendingHpt *pending = opaque;
359     size_t size = 1ULL << pending->shift;
360 
361     pending->hpt = qemu_memalign(size, size);
362     if (pending->hpt) {
363         memset(pending->hpt, 0, size);
364         pending->ret = H_SUCCESS;
365     } else {
366         pending->ret = H_NO_MEM;
367     }
368 
369     qemu_mutex_lock_iothread();
370 
371     if (SPAPR_MACHINE(qdev_get_machine())->pending_hpt == pending) {
372         /* Ready to go */
373         pending->complete = true;
374     } else {
375         /* We've been cancelled, clean ourselves up */
376         free_pending_hpt(pending);
377     }
378 
379     qemu_mutex_unlock_iothread();
380     return NULL;
381 }
382 
383 /* Must be called with BQL held */
384 static void cancel_hpt_prepare(SpaprMachineState *spapr)
385 {
386     SpaprPendingHpt *pending = spapr->pending_hpt;
387 
388     /* Let the thread know it's cancelled */
389     spapr->pending_hpt = NULL;
390 
391     if (!pending) {
392         /* Nothing to do */
393         return;
394     }
395 
396     if (!pending->complete) {
397         /* thread will clean itself up */
398         return;
399     }
400 
401     free_pending_hpt(pending);
402 }
403 
404 /* Convert a return code from the KVM ioctl()s implementing resize HPT
405  * into a PAPR hypercall return code */
406 static target_ulong resize_hpt_convert_rc(int ret)
407 {
408     if (ret >= 100000) {
409         return H_LONG_BUSY_ORDER_100_SEC;
410     } else if (ret >= 10000) {
411         return H_LONG_BUSY_ORDER_10_SEC;
412     } else if (ret >= 1000) {
413         return H_LONG_BUSY_ORDER_1_SEC;
414     } else if (ret >= 100) {
415         return H_LONG_BUSY_ORDER_100_MSEC;
416     } else if (ret >= 10) {
417         return H_LONG_BUSY_ORDER_10_MSEC;
418     } else if (ret > 0) {
419         return H_LONG_BUSY_ORDER_1_MSEC;
420     }
421 
422     switch (ret) {
423     case 0:
424         return H_SUCCESS;
425     case -EPERM:
426         return H_AUTHORITY;
427     case -EINVAL:
428         return H_PARAMETER;
429     case -ENXIO:
430         return H_CLOSED;
431     case -ENOSPC:
432         return H_PTEG_FULL;
433     case -EBUSY:
434         return H_BUSY;
435     case -ENOMEM:
436         return H_NO_MEM;
437     default:
438         return H_HARDWARE;
439     }
440 }
441 
442 static target_ulong h_resize_hpt_prepare(PowerPCCPU *cpu,
443                                          SpaprMachineState *spapr,
444                                          target_ulong opcode,
445                                          target_ulong *args)
446 {
447     target_ulong flags = args[0];
448     int shift = args[1];
449     SpaprPendingHpt *pending = spapr->pending_hpt;
450     uint64_t current_ram_size;
451     int rc;
452 
453     if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
454         return H_AUTHORITY;
455     }
456 
457     if (!spapr->htab_shift) {
458         /* Radix guest, no HPT */
459         return H_NOT_AVAILABLE;
460     }
461 
462     trace_spapr_h_resize_hpt_prepare(flags, shift);
463 
464     if (flags != 0) {
465         return H_PARAMETER;
466     }
467 
468     if (shift && ((shift < 18) || (shift > 46))) {
469         return H_PARAMETER;
470     }
471 
472     current_ram_size = MACHINE(spapr)->ram_size + get_plugged_memory_size();
473 
474     /* We only allow the guest to allocate an HPT one order above what
475      * we'd normally give them (to stop a small guest claiming a huge
476      * chunk of resources in the HPT */
477     if (shift > (spapr_hpt_shift_for_ramsize(current_ram_size) + 1)) {
478         return H_RESOURCE;
479     }
480 
481     rc = kvmppc_resize_hpt_prepare(cpu, flags, shift);
482     if (rc != -ENOSYS) {
483         return resize_hpt_convert_rc(rc);
484     }
485 
486     if (pending) {
487         /* something already in progress */
488         if (pending->shift == shift) {
489             /* and it's suitable */
490             if (pending->complete) {
491                 return pending->ret;
492             } else {
493                 return H_LONG_BUSY_ORDER_100_MSEC;
494             }
495         }
496 
497         /* not suitable, cancel and replace */
498         cancel_hpt_prepare(spapr);
499     }
500 
501     if (!shift) {
502         /* nothing to do */
503         return H_SUCCESS;
504     }
505 
506     /* start new prepare */
507 
508     pending = g_new0(SpaprPendingHpt, 1);
509     pending->shift = shift;
510     pending->ret = H_HARDWARE;
511 
512     qemu_thread_create(&pending->thread, "sPAPR HPT prepare",
513                        hpt_prepare_thread, pending, QEMU_THREAD_DETACHED);
514 
515     spapr->pending_hpt = pending;
516 
517     /* In theory we could estimate the time more accurately based on
518      * the new size, but there's not much point */
519     return H_LONG_BUSY_ORDER_100_MSEC;
520 }
521 
522 static uint64_t new_hpte_load0(void *htab, uint64_t pteg, int slot)
523 {
524     uint8_t *addr = htab;
525 
526     addr += pteg * HASH_PTEG_SIZE_64;
527     addr += slot * HASH_PTE_SIZE_64;
528     return  ldq_p(addr);
529 }
530 
531 static void new_hpte_store(void *htab, uint64_t pteg, int slot,
532                            uint64_t pte0, uint64_t pte1)
533 {
534     uint8_t *addr = htab;
535 
536     addr += pteg * HASH_PTEG_SIZE_64;
537     addr += slot * HASH_PTE_SIZE_64;
538 
539     stq_p(addr, pte0);
540     stq_p(addr + HASH_PTE_SIZE_64 / 2, pte1);
541 }
542 
543 static int rehash_hpte(PowerPCCPU *cpu,
544                        const ppc_hash_pte64_t *hptes,
545                        void *old_hpt, uint64_t oldsize,
546                        void *new_hpt, uint64_t newsize,
547                        uint64_t pteg, int slot)
548 {
549     uint64_t old_hash_mask = (oldsize >> 7) - 1;
550     uint64_t new_hash_mask = (newsize >> 7) - 1;
551     target_ulong pte0 = ppc_hash64_hpte0(cpu, hptes, slot);
552     target_ulong pte1;
553     uint64_t avpn;
554     unsigned base_pg_shift;
555     uint64_t hash, new_pteg, replace_pte0;
556 
557     if (!(pte0 & HPTE64_V_VALID) || !(pte0 & HPTE64_V_BOLTED)) {
558         return H_SUCCESS;
559     }
560 
561     pte1 = ppc_hash64_hpte1(cpu, hptes, slot);
562 
563     base_pg_shift = ppc_hash64_hpte_page_shift_noslb(cpu, pte0, pte1);
564     assert(base_pg_shift); /* H_ENTER shouldn't allow a bad encoding */
565     avpn = HPTE64_V_AVPN_VAL(pte0) & ~(((1ULL << base_pg_shift) - 1) >> 23);
566 
567     if (pte0 & HPTE64_V_SECONDARY) {
568         pteg = ~pteg;
569     }
570 
571     if ((pte0 & HPTE64_V_SSIZE) == HPTE64_V_SSIZE_256M) {
572         uint64_t offset, vsid;
573 
574         /* We only have 28 - 23 bits of offset in avpn */
575         offset = (avpn & 0x1f) << 23;
576         vsid = avpn >> 5;
577         /* We can find more bits from the pteg value */
578         if (base_pg_shift < 23) {
579             offset |= ((vsid ^ pteg) & old_hash_mask) << base_pg_shift;
580         }
581 
582         hash = vsid ^ (offset >> base_pg_shift);
583     } else if ((pte0 & HPTE64_V_SSIZE) == HPTE64_V_SSIZE_1T) {
584         uint64_t offset, vsid;
585 
586         /* We only have 40 - 23 bits of seg_off in avpn */
587         offset = (avpn & 0x1ffff) << 23;
588         vsid = avpn >> 17;
589         if (base_pg_shift < 23) {
590             offset |= ((vsid ^ (vsid << 25) ^ pteg) & old_hash_mask)
591                 << base_pg_shift;
592         }
593 
594         hash = vsid ^ (vsid << 25) ^ (offset >> base_pg_shift);
595     } else {
596         error_report("rehash_pte: Bad segment size in HPTE");
597         return H_HARDWARE;
598     }
599 
600     new_pteg = hash & new_hash_mask;
601     if (pte0 & HPTE64_V_SECONDARY) {
602         assert(~pteg == (hash & old_hash_mask));
603         new_pteg = ~new_pteg;
604     } else {
605         assert(pteg == (hash & old_hash_mask));
606     }
607     assert((oldsize != newsize) || (pteg == new_pteg));
608     replace_pte0 = new_hpte_load0(new_hpt, new_pteg, slot);
609     /*
610      * Strictly speaking, we don't need all these tests, since we only
611      * ever rehash bolted HPTEs.  We might in future handle non-bolted
612      * HPTEs, though so make the logic correct for those cases as
613      * well.
614      */
615     if (replace_pte0 & HPTE64_V_VALID) {
616         assert(newsize < oldsize);
617         if (replace_pte0 & HPTE64_V_BOLTED) {
618             if (pte0 & HPTE64_V_BOLTED) {
619                 /* Bolted collision, nothing we can do */
620                 return H_PTEG_FULL;
621             } else {
622                 /* Discard this hpte */
623                 return H_SUCCESS;
624             }
625         }
626     }
627 
628     new_hpte_store(new_hpt, new_pteg, slot, pte0, pte1);
629     return H_SUCCESS;
630 }
631 
632 static int rehash_hpt(PowerPCCPU *cpu,
633                       void *old_hpt, uint64_t oldsize,
634                       void *new_hpt, uint64_t newsize)
635 {
636     uint64_t n_ptegs = oldsize >> 7;
637     uint64_t pteg;
638     int slot;
639     int rc;
640 
641     for (pteg = 0; pteg < n_ptegs; pteg++) {
642         hwaddr ptex = pteg * HPTES_PER_GROUP;
643         const ppc_hash_pte64_t *hptes
644             = ppc_hash64_map_hptes(cpu, ptex, HPTES_PER_GROUP);
645 
646         if (!hptes) {
647             return H_HARDWARE;
648         }
649 
650         for (slot = 0; slot < HPTES_PER_GROUP; slot++) {
651             rc = rehash_hpte(cpu, hptes, old_hpt, oldsize, new_hpt, newsize,
652                              pteg, slot);
653             if (rc != H_SUCCESS) {
654                 ppc_hash64_unmap_hptes(cpu, hptes, ptex, HPTES_PER_GROUP);
655                 return rc;
656             }
657         }
658         ppc_hash64_unmap_hptes(cpu, hptes, ptex, HPTES_PER_GROUP);
659     }
660 
661     return H_SUCCESS;
662 }
663 
664 static void do_push_sregs_to_kvm_pr(CPUState *cs, run_on_cpu_data data)
665 {
666     int ret;
667 
668     cpu_synchronize_state(cs);
669 
670     ret = kvmppc_put_books_sregs(POWERPC_CPU(cs));
671     if (ret < 0) {
672         error_report("failed to push sregs to KVM: %s", strerror(-ret));
673         exit(1);
674     }
675 }
676 
677 static void push_sregs_to_kvm_pr(SpaprMachineState *spapr)
678 {
679     CPUState *cs;
680 
681     /*
682      * This is a hack for the benefit of KVM PR - it abuses the SDR1
683      * slot in kvm_sregs to communicate the userspace address of the
684      * HPT
685      */
686     if (!kvm_enabled() || !spapr->htab) {
687         return;
688     }
689 
690     CPU_FOREACH(cs) {
691         run_on_cpu(cs, do_push_sregs_to_kvm_pr, RUN_ON_CPU_NULL);
692     }
693 }
694 
695 static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
696                                         SpaprMachineState *spapr,
697                                         target_ulong opcode,
698                                         target_ulong *args)
699 {
700     target_ulong flags = args[0];
701     target_ulong shift = args[1];
702     SpaprPendingHpt *pending = spapr->pending_hpt;
703     int rc;
704     size_t newsize;
705 
706     if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
707         return H_AUTHORITY;
708     }
709 
710     if (!spapr->htab_shift) {
711         /* Radix guest, no HPT */
712         return H_NOT_AVAILABLE;
713     }
714 
715     trace_spapr_h_resize_hpt_commit(flags, shift);
716 
717     rc = kvmppc_resize_hpt_commit(cpu, flags, shift);
718     if (rc != -ENOSYS) {
719         rc = resize_hpt_convert_rc(rc);
720         if (rc == H_SUCCESS) {
721             /* Need to set the new htab_shift in the machine state */
722             spapr->htab_shift = shift;
723         }
724         return rc;
725     }
726 
727     if (flags != 0) {
728         return H_PARAMETER;
729     }
730 
731     if (!pending || (pending->shift != shift)) {
732         /* no matching prepare */
733         return H_CLOSED;
734     }
735 
736     if (!pending->complete) {
737         /* prepare has not completed */
738         return H_BUSY;
739     }
740 
741     /* Shouldn't have got past PREPARE without an HPT */
742     g_assert(spapr->htab_shift);
743 
744     newsize = 1ULL << pending->shift;
745     rc = rehash_hpt(cpu, spapr->htab, HTAB_SIZE(spapr),
746                     pending->hpt, newsize);
747     if (rc == H_SUCCESS) {
748         qemu_vfree(spapr->htab);
749         spapr->htab = pending->hpt;
750         spapr->htab_shift = pending->shift;
751 
752         push_sregs_to_kvm_pr(spapr);
753 
754         pending->hpt = NULL; /* so it's not free()d */
755     }
756 
757     /* Clean up */
758     spapr->pending_hpt = NULL;
759     free_pending_hpt(pending);
760 
761     return rc;
762 }
763 
764 static target_ulong h_set_sprg0(PowerPCCPU *cpu, SpaprMachineState *spapr,
765                                 target_ulong opcode, target_ulong *args)
766 {
767     cpu_synchronize_state(CPU(cpu));
768     cpu->env.spr[SPR_SPRG0] = args[0];
769 
770     return H_SUCCESS;
771 }
772 
773 static target_ulong h_set_dabr(PowerPCCPU *cpu, SpaprMachineState *spapr,
774                                target_ulong opcode, target_ulong *args)
775 {
776     if (!has_spr(cpu, SPR_DABR)) {
777         return H_HARDWARE;              /* DABR register not available */
778     }
779     cpu_synchronize_state(CPU(cpu));
780 
781     if (has_spr(cpu, SPR_DABRX)) {
782         cpu->env.spr[SPR_DABRX] = 0x3;  /* Use Problem and Privileged state */
783     } else if (!(args[0] & 0x4)) {      /* Breakpoint Translation set? */
784         return H_RESERVED_DABR;
785     }
786 
787     cpu->env.spr[SPR_DABR] = args[0];
788     return H_SUCCESS;
789 }
790 
791 static target_ulong h_set_xdabr(PowerPCCPU *cpu, SpaprMachineState *spapr,
792                                 target_ulong opcode, target_ulong *args)
793 {
794     target_ulong dabrx = args[1];
795 
796     if (!has_spr(cpu, SPR_DABR) || !has_spr(cpu, SPR_DABRX)) {
797         return H_HARDWARE;
798     }
799 
800     if ((dabrx & ~0xfULL) != 0 || (dabrx & H_DABRX_HYPERVISOR) != 0
801         || (dabrx & (H_DABRX_KERNEL | H_DABRX_USER)) == 0) {
802         return H_PARAMETER;
803     }
804 
805     cpu_synchronize_state(CPU(cpu));
806     cpu->env.spr[SPR_DABRX] = dabrx;
807     cpu->env.spr[SPR_DABR] = args[0];
808 
809     return H_SUCCESS;
810 }
811 
812 static target_ulong h_page_init(PowerPCCPU *cpu, SpaprMachineState *spapr,
813                                 target_ulong opcode, target_ulong *args)
814 {
815     target_ulong flags = args[0];
816     hwaddr dst = args[1];
817     hwaddr src = args[2];
818     hwaddr len = TARGET_PAGE_SIZE;
819     uint8_t *pdst, *psrc;
820     target_long ret = H_SUCCESS;
821 
822     if (flags & ~(H_ICACHE_SYNCHRONIZE | H_ICACHE_INVALIDATE
823                   | H_COPY_PAGE | H_ZERO_PAGE)) {
824         qemu_log_mask(LOG_UNIMP, "h_page_init: Bad flags (" TARGET_FMT_lx "\n",
825                       flags);
826         return H_PARAMETER;
827     }
828 
829     /* Map-in destination */
830     if (!is_ram_address(spapr, dst) || (dst & ~TARGET_PAGE_MASK) != 0) {
831         return H_PARAMETER;
832     }
833     pdst = cpu_physical_memory_map(dst, &len, 1);
834     if (!pdst || len != TARGET_PAGE_SIZE) {
835         return H_PARAMETER;
836     }
837 
838     if (flags & H_COPY_PAGE) {
839         /* Map-in source, copy to destination, and unmap source again */
840         if (!is_ram_address(spapr, src) || (src & ~TARGET_PAGE_MASK) != 0) {
841             ret = H_PARAMETER;
842             goto unmap_out;
843         }
844         psrc = cpu_physical_memory_map(src, &len, 0);
845         if (!psrc || len != TARGET_PAGE_SIZE) {
846             ret = H_PARAMETER;
847             goto unmap_out;
848         }
849         memcpy(pdst, psrc, len);
850         cpu_physical_memory_unmap(psrc, len, 0, len);
851     } else if (flags & H_ZERO_PAGE) {
852         memset(pdst, 0, len);          /* Just clear the destination page */
853     }
854 
855     if (kvm_enabled() && (flags & H_ICACHE_SYNCHRONIZE) != 0) {
856         kvmppc_dcbst_range(cpu, pdst, len);
857     }
858     if (flags & (H_ICACHE_SYNCHRONIZE | H_ICACHE_INVALIDATE)) {
859         if (kvm_enabled()) {
860             kvmppc_icbi_range(cpu, pdst, len);
861         } else {
862             tb_flush(CPU(cpu));
863         }
864     }
865 
866 unmap_out:
867     cpu_physical_memory_unmap(pdst, TARGET_PAGE_SIZE, 1, len);
868     return ret;
869 }
870 
871 #define FLAGS_REGISTER_VPA         0x0000200000000000ULL
872 #define FLAGS_REGISTER_DTL         0x0000400000000000ULL
873 #define FLAGS_REGISTER_SLBSHADOW   0x0000600000000000ULL
874 #define FLAGS_DEREGISTER_VPA       0x0000a00000000000ULL
875 #define FLAGS_DEREGISTER_DTL       0x0000c00000000000ULL
876 #define FLAGS_DEREGISTER_SLBSHADOW 0x0000e00000000000ULL
877 
878 static target_ulong register_vpa(PowerPCCPU *cpu, target_ulong vpa)
879 {
880     CPUState *cs = CPU(cpu);
881     CPUPPCState *env = &cpu->env;
882     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
883     uint16_t size;
884     uint8_t tmp;
885 
886     if (vpa == 0) {
887         hcall_dprintf("Can't cope with registering a VPA at logical 0\n");
888         return H_HARDWARE;
889     }
890 
891     if (vpa % env->dcache_line_size) {
892         return H_PARAMETER;
893     }
894     /* FIXME: bounds check the address */
895 
896     size = lduw_be_phys(cs->as, vpa + 0x4);
897 
898     if (size < VPA_MIN_SIZE) {
899         return H_PARAMETER;
900     }
901 
902     /* VPA is not allowed to cross a page boundary */
903     if ((vpa / 4096) != ((vpa + size - 1) / 4096)) {
904         return H_PARAMETER;
905     }
906 
907     spapr_cpu->vpa_addr = vpa;
908 
909     tmp = ldub_phys(cs->as, spapr_cpu->vpa_addr + VPA_SHARED_PROC_OFFSET);
910     tmp |= VPA_SHARED_PROC_VAL;
911     stb_phys(cs->as, spapr_cpu->vpa_addr + VPA_SHARED_PROC_OFFSET, tmp);
912 
913     return H_SUCCESS;
914 }
915 
916 static target_ulong deregister_vpa(PowerPCCPU *cpu, target_ulong vpa)
917 {
918     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
919 
920     if (spapr_cpu->slb_shadow_addr) {
921         return H_RESOURCE;
922     }
923 
924     if (spapr_cpu->dtl_addr) {
925         return H_RESOURCE;
926     }
927 
928     spapr_cpu->vpa_addr = 0;
929     return H_SUCCESS;
930 }
931 
932 static target_ulong register_slb_shadow(PowerPCCPU *cpu, target_ulong addr)
933 {
934     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
935     uint32_t size;
936 
937     if (addr == 0) {
938         hcall_dprintf("Can't cope with SLB shadow at logical 0\n");
939         return H_HARDWARE;
940     }
941 
942     size = ldl_be_phys(CPU(cpu)->as, addr + 0x4);
943     if (size < 0x8) {
944         return H_PARAMETER;
945     }
946 
947     if ((addr / 4096) != ((addr + size - 1) / 4096)) {
948         return H_PARAMETER;
949     }
950 
951     if (!spapr_cpu->vpa_addr) {
952         return H_RESOURCE;
953     }
954 
955     spapr_cpu->slb_shadow_addr = addr;
956     spapr_cpu->slb_shadow_size = size;
957 
958     return H_SUCCESS;
959 }
960 
961 static target_ulong deregister_slb_shadow(PowerPCCPU *cpu, target_ulong addr)
962 {
963     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
964 
965     spapr_cpu->slb_shadow_addr = 0;
966     spapr_cpu->slb_shadow_size = 0;
967     return H_SUCCESS;
968 }
969 
970 static target_ulong register_dtl(PowerPCCPU *cpu, target_ulong addr)
971 {
972     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
973     uint32_t size;
974 
975     if (addr == 0) {
976         hcall_dprintf("Can't cope with DTL at logical 0\n");
977         return H_HARDWARE;
978     }
979 
980     size = ldl_be_phys(CPU(cpu)->as, addr + 0x4);
981 
982     if (size < 48) {
983         return H_PARAMETER;
984     }
985 
986     if (!spapr_cpu->vpa_addr) {
987         return H_RESOURCE;
988     }
989 
990     spapr_cpu->dtl_addr = addr;
991     spapr_cpu->dtl_size = size;
992 
993     return H_SUCCESS;
994 }
995 
996 static target_ulong deregister_dtl(PowerPCCPU *cpu, target_ulong addr)
997 {
998     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
999 
1000     spapr_cpu->dtl_addr = 0;
1001     spapr_cpu->dtl_size = 0;
1002 
1003     return H_SUCCESS;
1004 }
1005 
1006 static target_ulong h_register_vpa(PowerPCCPU *cpu, SpaprMachineState *spapr,
1007                                    target_ulong opcode, target_ulong *args)
1008 {
1009     target_ulong flags = args[0];
1010     target_ulong procno = args[1];
1011     target_ulong vpa = args[2];
1012     target_ulong ret = H_PARAMETER;
1013     PowerPCCPU *tcpu;
1014 
1015     tcpu = spapr_find_cpu(procno);
1016     if (!tcpu) {
1017         return H_PARAMETER;
1018     }
1019 
1020     switch (flags) {
1021     case FLAGS_REGISTER_VPA:
1022         ret = register_vpa(tcpu, vpa);
1023         break;
1024 
1025     case FLAGS_DEREGISTER_VPA:
1026         ret = deregister_vpa(tcpu, vpa);
1027         break;
1028 
1029     case FLAGS_REGISTER_SLBSHADOW:
1030         ret = register_slb_shadow(tcpu, vpa);
1031         break;
1032 
1033     case FLAGS_DEREGISTER_SLBSHADOW:
1034         ret = deregister_slb_shadow(tcpu, vpa);
1035         break;
1036 
1037     case FLAGS_REGISTER_DTL:
1038         ret = register_dtl(tcpu, vpa);
1039         break;
1040 
1041     case FLAGS_DEREGISTER_DTL:
1042         ret = deregister_dtl(tcpu, vpa);
1043         break;
1044     }
1045 
1046     return ret;
1047 }
1048 
1049 static target_ulong h_cede(PowerPCCPU *cpu, SpaprMachineState *spapr,
1050                            target_ulong opcode, target_ulong *args)
1051 {
1052     CPUPPCState *env = &cpu->env;
1053     CPUState *cs = CPU(cpu);
1054     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
1055 
1056     env->msr |= (1ULL << MSR_EE);
1057     hreg_compute_hflags(env);
1058 
1059     if (spapr_cpu->prod) {
1060         spapr_cpu->prod = false;
1061         return H_SUCCESS;
1062     }
1063 
1064     if (!cpu_has_work(cs)) {
1065         cs->halted = 1;
1066         cs->exception_index = EXCP_HLT;
1067         cs->exit_request = 1;
1068     }
1069 
1070     return H_SUCCESS;
1071 }
1072 
1073 /*
1074  * Confer to self, aka join. Cede could use the same pattern as well, if
1075  * EXCP_HLT can be changed to ECXP_HALTED.
1076  */
1077 static target_ulong h_confer_self(PowerPCCPU *cpu)
1078 {
1079     CPUState *cs = CPU(cpu);
1080     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
1081 
1082     if (spapr_cpu->prod) {
1083         spapr_cpu->prod = false;
1084         return H_SUCCESS;
1085     }
1086     cs->halted = 1;
1087     cs->exception_index = EXCP_HALTED;
1088     cs->exit_request = 1;
1089 
1090     return H_SUCCESS;
1091 }
1092 
1093 static target_ulong h_join(PowerPCCPU *cpu, SpaprMachineState *spapr,
1094                            target_ulong opcode, target_ulong *args)
1095 {
1096     CPUPPCState *env = &cpu->env;
1097     CPUState *cs;
1098     bool last_unjoined = true;
1099 
1100     if (env->msr & (1ULL << MSR_EE)) {
1101         return H_BAD_MODE;
1102     }
1103 
1104     /*
1105      * Must not join the last CPU running. Interestingly, no such restriction
1106      * for H_CONFER-to-self, but that is probably not intended to be used
1107      * when H_JOIN is available.
1108      */
1109     CPU_FOREACH(cs) {
1110         PowerPCCPU *c = POWERPC_CPU(cs);
1111         CPUPPCState *e = &c->env;
1112         if (c == cpu) {
1113             continue;
1114         }
1115 
1116         /* Don't have a way to indicate joined, so use halted && MSR[EE]=0 */
1117         if (!cs->halted || (e->msr & (1ULL << MSR_EE))) {
1118             last_unjoined = false;
1119             break;
1120         }
1121     }
1122     if (last_unjoined) {
1123         return H_CONTINUE;
1124     }
1125 
1126     return h_confer_self(cpu);
1127 }
1128 
1129 static target_ulong h_confer(PowerPCCPU *cpu, SpaprMachineState *spapr,
1130                            target_ulong opcode, target_ulong *args)
1131 {
1132     target_long target = args[0];
1133     uint32_t dispatch = args[1];
1134     CPUState *cs = CPU(cpu);
1135     SpaprCpuState *spapr_cpu;
1136 
1137     /*
1138      * -1 means confer to all other CPUs without dispatch counter check,
1139      *  otherwise it's a targeted confer.
1140      */
1141     if (target != -1) {
1142         PowerPCCPU *target_cpu = spapr_find_cpu(target);
1143         uint32_t target_dispatch;
1144 
1145         if (!target_cpu) {
1146             return H_PARAMETER;
1147         }
1148 
1149         /*
1150          * target == self is a special case, we wait until prodded, without
1151          * dispatch counter check.
1152          */
1153         if (cpu == target_cpu) {
1154             return h_confer_self(cpu);
1155         }
1156 
1157         spapr_cpu = spapr_cpu_state(target_cpu);
1158         if (!spapr_cpu->vpa_addr || ((dispatch & 1) == 0)) {
1159             return H_SUCCESS;
1160         }
1161 
1162         target_dispatch = ldl_be_phys(cs->as,
1163                                   spapr_cpu->vpa_addr + VPA_DISPATCH_COUNTER);
1164         if (target_dispatch != dispatch) {
1165             return H_SUCCESS;
1166         }
1167 
1168         /*
1169          * The targeted confer does not do anything special beyond yielding
1170          * the current vCPU, but even this should be better than nothing.
1171          * At least for single-threaded tcg, it gives the target a chance to
1172          * run before we run again. Multi-threaded tcg does not really do
1173          * anything with EXCP_YIELD yet.
1174          */
1175     }
1176 
1177     cs->exception_index = EXCP_YIELD;
1178     cs->exit_request = 1;
1179     cpu_loop_exit(cs);
1180 
1181     return H_SUCCESS;
1182 }
1183 
1184 static target_ulong h_prod(PowerPCCPU *cpu, SpaprMachineState *spapr,
1185                            target_ulong opcode, target_ulong *args)
1186 {
1187     target_long target = args[0];
1188     PowerPCCPU *tcpu;
1189     CPUState *cs;
1190     SpaprCpuState *spapr_cpu;
1191 
1192     tcpu = spapr_find_cpu(target);
1193     cs = CPU(tcpu);
1194     if (!cs) {
1195         return H_PARAMETER;
1196     }
1197 
1198     spapr_cpu = spapr_cpu_state(tcpu);
1199     spapr_cpu->prod = true;
1200     cs->halted = 0;
1201     qemu_cpu_kick(cs);
1202 
1203     return H_SUCCESS;
1204 }
1205 
1206 static target_ulong h_rtas(PowerPCCPU *cpu, SpaprMachineState *spapr,
1207                            target_ulong opcode, target_ulong *args)
1208 {
1209     target_ulong rtas_r3 = args[0];
1210     uint32_t token = rtas_ld(rtas_r3, 0);
1211     uint32_t nargs = rtas_ld(rtas_r3, 1);
1212     uint32_t nret = rtas_ld(rtas_r3, 2);
1213 
1214     return spapr_rtas_call(cpu, spapr, token, nargs, rtas_r3 + 12,
1215                            nret, rtas_r3 + 12 + 4*nargs);
1216 }
1217 
1218 static target_ulong h_logical_load(PowerPCCPU *cpu, SpaprMachineState *spapr,
1219                                    target_ulong opcode, target_ulong *args)
1220 {
1221     CPUState *cs = CPU(cpu);
1222     target_ulong size = args[0];
1223     target_ulong addr = args[1];
1224 
1225     switch (size) {
1226     case 1:
1227         args[0] = ldub_phys(cs->as, addr);
1228         return H_SUCCESS;
1229     case 2:
1230         args[0] = lduw_phys(cs->as, addr);
1231         return H_SUCCESS;
1232     case 4:
1233         args[0] = ldl_phys(cs->as, addr);
1234         return H_SUCCESS;
1235     case 8:
1236         args[0] = ldq_phys(cs->as, addr);
1237         return H_SUCCESS;
1238     }
1239     return H_PARAMETER;
1240 }
1241 
1242 static target_ulong h_logical_store(PowerPCCPU *cpu, SpaprMachineState *spapr,
1243                                     target_ulong opcode, target_ulong *args)
1244 {
1245     CPUState *cs = CPU(cpu);
1246 
1247     target_ulong size = args[0];
1248     target_ulong addr = args[1];
1249     target_ulong val  = args[2];
1250 
1251     switch (size) {
1252     case 1:
1253         stb_phys(cs->as, addr, val);
1254         return H_SUCCESS;
1255     case 2:
1256         stw_phys(cs->as, addr, val);
1257         return H_SUCCESS;
1258     case 4:
1259         stl_phys(cs->as, addr, val);
1260         return H_SUCCESS;
1261     case 8:
1262         stq_phys(cs->as, addr, val);
1263         return H_SUCCESS;
1264     }
1265     return H_PARAMETER;
1266 }
1267 
1268 static target_ulong h_logical_memop(PowerPCCPU *cpu, SpaprMachineState *spapr,
1269                                     target_ulong opcode, target_ulong *args)
1270 {
1271     CPUState *cs = CPU(cpu);
1272 
1273     target_ulong dst   = args[0]; /* Destination address */
1274     target_ulong src   = args[1]; /* Source address */
1275     target_ulong esize = args[2]; /* Element size (0=1,1=2,2=4,3=8) */
1276     target_ulong count = args[3]; /* Element count */
1277     target_ulong op    = args[4]; /* 0 = copy, 1 = invert */
1278     uint64_t tmp;
1279     unsigned int mask = (1 << esize) - 1;
1280     int step = 1 << esize;
1281 
1282     if (count > 0x80000000) {
1283         return H_PARAMETER;
1284     }
1285 
1286     if ((dst & mask) || (src & mask) || (op > 1)) {
1287         return H_PARAMETER;
1288     }
1289 
1290     if (dst >= src && dst < (src + (count << esize))) {
1291             dst = dst + ((count - 1) << esize);
1292             src = src + ((count - 1) << esize);
1293             step = -step;
1294     }
1295 
1296     while (count--) {
1297         switch (esize) {
1298         case 0:
1299             tmp = ldub_phys(cs->as, src);
1300             break;
1301         case 1:
1302             tmp = lduw_phys(cs->as, src);
1303             break;
1304         case 2:
1305             tmp = ldl_phys(cs->as, src);
1306             break;
1307         case 3:
1308             tmp = ldq_phys(cs->as, src);
1309             break;
1310         default:
1311             return H_PARAMETER;
1312         }
1313         if (op == 1) {
1314             tmp = ~tmp;
1315         }
1316         switch (esize) {
1317         case 0:
1318             stb_phys(cs->as, dst, tmp);
1319             break;
1320         case 1:
1321             stw_phys(cs->as, dst, tmp);
1322             break;
1323         case 2:
1324             stl_phys(cs->as, dst, tmp);
1325             break;
1326         case 3:
1327             stq_phys(cs->as, dst, tmp);
1328             break;
1329         }
1330         dst = dst + step;
1331         src = src + step;
1332     }
1333 
1334     return H_SUCCESS;
1335 }
1336 
1337 static target_ulong h_logical_icbi(PowerPCCPU *cpu, SpaprMachineState *spapr,
1338                                    target_ulong opcode, target_ulong *args)
1339 {
1340     /* Nothing to do on emulation, KVM will trap this in the kernel */
1341     return H_SUCCESS;
1342 }
1343 
1344 static target_ulong h_logical_dcbf(PowerPCCPU *cpu, SpaprMachineState *spapr,
1345                                    target_ulong opcode, target_ulong *args)
1346 {
1347     /* Nothing to do on emulation, KVM will trap this in the kernel */
1348     return H_SUCCESS;
1349 }
1350 
1351 static target_ulong h_set_mode_resource_le(PowerPCCPU *cpu,
1352                                            target_ulong mflags,
1353                                            target_ulong value1,
1354                                            target_ulong value2)
1355 {
1356     if (value1) {
1357         return H_P3;
1358     }
1359     if (value2) {
1360         return H_P4;
1361     }
1362 
1363     switch (mflags) {
1364     case H_SET_MODE_ENDIAN_BIG:
1365         spapr_set_all_lpcrs(0, LPCR_ILE);
1366         spapr_pci_switch_vga(true);
1367         return H_SUCCESS;
1368 
1369     case H_SET_MODE_ENDIAN_LITTLE:
1370         spapr_set_all_lpcrs(LPCR_ILE, LPCR_ILE);
1371         spapr_pci_switch_vga(false);
1372         return H_SUCCESS;
1373     }
1374 
1375     return H_UNSUPPORTED_FLAG;
1376 }
1377 
1378 static target_ulong h_set_mode_resource_addr_trans_mode(PowerPCCPU *cpu,
1379                                                         target_ulong mflags,
1380                                                         target_ulong value1,
1381                                                         target_ulong value2)
1382 {
1383     PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cpu);
1384 
1385     if (!(pcc->insns_flags2 & PPC2_ISA207S)) {
1386         return H_P2;
1387     }
1388     if (value1) {
1389         return H_P3;
1390     }
1391     if (value2) {
1392         return H_P4;
1393     }
1394 
1395     if (mflags == AIL_RESERVED) {
1396         return H_UNSUPPORTED_FLAG;
1397     }
1398 
1399     spapr_set_all_lpcrs(mflags << LPCR_AIL_SHIFT, LPCR_AIL);
1400 
1401     return H_SUCCESS;
1402 }
1403 
1404 static target_ulong h_set_mode(PowerPCCPU *cpu, SpaprMachineState *spapr,
1405                                target_ulong opcode, target_ulong *args)
1406 {
1407     target_ulong resource = args[1];
1408     target_ulong ret = H_P2;
1409 
1410     switch (resource) {
1411     case H_SET_MODE_RESOURCE_LE:
1412         ret = h_set_mode_resource_le(cpu, args[0], args[2], args[3]);
1413         break;
1414     case H_SET_MODE_RESOURCE_ADDR_TRANS_MODE:
1415         ret = h_set_mode_resource_addr_trans_mode(cpu, args[0],
1416                                                   args[2], args[3]);
1417         break;
1418     }
1419 
1420     return ret;
1421 }
1422 
1423 static target_ulong h_clean_slb(PowerPCCPU *cpu, SpaprMachineState *spapr,
1424                                 target_ulong opcode, target_ulong *args)
1425 {
1426     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x"TARGET_FMT_lx"%s\n",
1427                   opcode, " (H_CLEAN_SLB)");
1428     return H_FUNCTION;
1429 }
1430 
1431 static target_ulong h_invalidate_pid(PowerPCCPU *cpu, SpaprMachineState *spapr,
1432                                      target_ulong opcode, target_ulong *args)
1433 {
1434     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x"TARGET_FMT_lx"%s\n",
1435                   opcode, " (H_INVALIDATE_PID)");
1436     return H_FUNCTION;
1437 }
1438 
1439 static void spapr_check_setup_free_hpt(SpaprMachineState *spapr,
1440                                        uint64_t patbe_old, uint64_t patbe_new)
1441 {
1442     /*
1443      * We have 4 Options:
1444      * HASH->HASH || RADIX->RADIX || NOTHING->RADIX : Do Nothing
1445      * HASH->RADIX                                  : Free HPT
1446      * RADIX->HASH                                  : Allocate HPT
1447      * NOTHING->HASH                                : Allocate HPT
1448      * Note: NOTHING implies the case where we said the guest could choose
1449      *       later and so assumed radix and now it's called H_REG_PROC_TBL
1450      */
1451 
1452     if ((patbe_old & PATE1_GR) == (patbe_new & PATE1_GR)) {
1453         /* We assume RADIX, so this catches all the "Do Nothing" cases */
1454     } else if (!(patbe_old & PATE1_GR)) {
1455         /* HASH->RADIX : Free HPT */
1456         spapr_free_hpt(spapr);
1457     } else if (!(patbe_new & PATE1_GR)) {
1458         /* RADIX->HASH || NOTHING->HASH : Allocate HPT */
1459         spapr_setup_hpt_and_vrma(spapr);
1460     }
1461     return;
1462 }
1463 
1464 #define FLAGS_MASK              0x01FULL
1465 #define FLAG_MODIFY             0x10
1466 #define FLAG_REGISTER           0x08
1467 #define FLAG_RADIX              0x04
1468 #define FLAG_HASH_PROC_TBL      0x02
1469 #define FLAG_GTSE               0x01
1470 
1471 static target_ulong h_register_process_table(PowerPCCPU *cpu,
1472                                              SpaprMachineState *spapr,
1473                                              target_ulong opcode,
1474                                              target_ulong *args)
1475 {
1476     target_ulong flags = args[0];
1477     target_ulong proc_tbl = args[1];
1478     target_ulong page_size = args[2];
1479     target_ulong table_size = args[3];
1480     target_ulong update_lpcr = 0;
1481     uint64_t cproc;
1482 
1483     if (flags & ~FLAGS_MASK) { /* Check no reserved bits are set */
1484         return H_PARAMETER;
1485     }
1486     if (flags & FLAG_MODIFY) {
1487         if (flags & FLAG_REGISTER) {
1488             if (flags & FLAG_RADIX) { /* Register new RADIX process table */
1489                 if (proc_tbl & 0xfff || proc_tbl >> 60) {
1490                     return H_P2;
1491                 } else if (page_size) {
1492                     return H_P3;
1493                 } else if (table_size > 24) {
1494                     return H_P4;
1495                 }
1496                 cproc = PATE1_GR | proc_tbl | table_size;
1497             } else { /* Register new HPT process table */
1498                 if (flags & FLAG_HASH_PROC_TBL) { /* Hash with Segment Tables */
1499                     /* TODO - Not Supported */
1500                     /* Technically caused by flag bits => H_PARAMETER */
1501                     return H_PARAMETER;
1502                 } else { /* Hash with SLB */
1503                     if (proc_tbl >> 38) {
1504                         return H_P2;
1505                     } else if (page_size & ~0x7) {
1506                         return H_P3;
1507                     } else if (table_size > 24) {
1508                         return H_P4;
1509                     }
1510                 }
1511                 cproc = (proc_tbl << 25) | page_size << 5 | table_size;
1512             }
1513 
1514         } else { /* Deregister current process table */
1515             /*
1516              * Set to benign value: (current GR) | 0. This allows
1517              * deregistration in KVM to succeed even if the radix bit
1518              * in flags doesn't match the radix bit in the old PATE.
1519              */
1520             cproc = spapr->patb_entry & PATE1_GR;
1521         }
1522     } else { /* Maintain current registration */
1523         if (!(flags & FLAG_RADIX) != !(spapr->patb_entry & PATE1_GR)) {
1524             /* Technically caused by flag bits => H_PARAMETER */
1525             return H_PARAMETER; /* Existing Process Table Mismatch */
1526         }
1527         cproc = spapr->patb_entry;
1528     }
1529 
1530     /* Check if we need to setup OR free the hpt */
1531     spapr_check_setup_free_hpt(spapr, spapr->patb_entry, cproc);
1532 
1533     spapr->patb_entry = cproc; /* Save new process table */
1534 
1535     /* Update the UPRT, HR and GTSE bits in the LPCR for all cpus */
1536     if (flags & FLAG_RADIX)     /* Radix must use process tables, also set HR */
1537         update_lpcr |= (LPCR_UPRT | LPCR_HR);
1538     else if (flags & FLAG_HASH_PROC_TBL) /* Hash with process tables */
1539         update_lpcr |= LPCR_UPRT;
1540     if (flags & FLAG_GTSE)      /* Guest translation shootdown enable */
1541         update_lpcr |= LPCR_GTSE;
1542 
1543     spapr_set_all_lpcrs(update_lpcr, LPCR_UPRT | LPCR_HR | LPCR_GTSE);
1544 
1545     if (kvm_enabled()) {
1546         return kvmppc_configure_v3_mmu(cpu, flags & FLAG_RADIX,
1547                                        flags & FLAG_GTSE, cproc);
1548     }
1549     return H_SUCCESS;
1550 }
1551 
1552 #define H_SIGNAL_SYS_RESET_ALL         -1
1553 #define H_SIGNAL_SYS_RESET_ALLBUTSELF  -2
1554 
1555 static target_ulong h_signal_sys_reset(PowerPCCPU *cpu,
1556                                        SpaprMachineState *spapr,
1557                                        target_ulong opcode, target_ulong *args)
1558 {
1559     target_long target = args[0];
1560     CPUState *cs;
1561 
1562     if (target < 0) {
1563         /* Broadcast */
1564         if (target < H_SIGNAL_SYS_RESET_ALLBUTSELF) {
1565             return H_PARAMETER;
1566         }
1567 
1568         CPU_FOREACH(cs) {
1569             PowerPCCPU *c = POWERPC_CPU(cs);
1570 
1571             if (target == H_SIGNAL_SYS_RESET_ALLBUTSELF) {
1572                 if (c == cpu) {
1573                     continue;
1574                 }
1575             }
1576             run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL);
1577         }
1578         return H_SUCCESS;
1579 
1580     } else {
1581         /* Unicast */
1582         cs = CPU(spapr_find_cpu(target));
1583         if (cs) {
1584             run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL);
1585             return H_SUCCESS;
1586         }
1587         return H_PARAMETER;
1588     }
1589 }
1590 
1591 static uint32_t cas_check_pvr(SpaprMachineState *spapr, PowerPCCPU *cpu,
1592                               target_ulong *addr, bool *raw_mode_supported,
1593                               Error **errp)
1594 {
1595     bool explicit_match = false; /* Matched the CPU's real PVR */
1596     uint32_t max_compat = spapr->max_compat_pvr;
1597     uint32_t best_compat = 0;
1598     int i;
1599 
1600     /*
1601      * We scan the supplied table of PVRs looking for two things
1602      *   1. Is our real CPU PVR in the list?
1603      *   2. What's the "best" listed logical PVR
1604      */
1605     for (i = 0; i < 512; ++i) {
1606         uint32_t pvr, pvr_mask;
1607 
1608         pvr_mask = ldl_be_phys(&address_space_memory, *addr);
1609         pvr = ldl_be_phys(&address_space_memory, *addr + 4);
1610         *addr += 8;
1611 
1612         if (~pvr_mask & pvr) {
1613             break; /* Terminator record */
1614         }
1615 
1616         if ((cpu->env.spr[SPR_PVR] & pvr_mask) == (pvr & pvr_mask)) {
1617             explicit_match = true;
1618         } else {
1619             if (ppc_check_compat(cpu, pvr, best_compat, max_compat)) {
1620                 best_compat = pvr;
1621             }
1622         }
1623     }
1624 
1625     if ((best_compat == 0) && (!explicit_match || max_compat)) {
1626         /* We couldn't find a suitable compatibility mode, and either
1627          * the guest doesn't support "raw" mode for this CPU, or raw
1628          * mode is disabled because a maximum compat mode is set */
1629         error_setg(errp, "Couldn't negotiate a suitable PVR during CAS");
1630         return 0;
1631     }
1632 
1633     *raw_mode_supported = explicit_match;
1634 
1635     /* Parsing finished */
1636     trace_spapr_cas_pvr(cpu->compat_pvr, explicit_match, best_compat);
1637 
1638     return best_compat;
1639 }
1640 
1641 static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
1642                                                   SpaprMachineState *spapr,
1643                                                   target_ulong opcode,
1644                                                   target_ulong *args)
1645 {
1646     /* Working address in data buffer */
1647     target_ulong addr = ppc64_phys_to_real(args[0]);
1648     target_ulong ov_table;
1649     uint32_t cas_pvr;
1650     SpaprOptionVector *ov1_guest, *ov5_guest, *ov5_cas_old, *ov5_updates;
1651     bool guest_radix;
1652     Error *local_err = NULL;
1653     bool raw_mode_supported = false;
1654     bool guest_xive;
1655 
1656     cas_pvr = cas_check_pvr(spapr, cpu, &addr, &raw_mode_supported, &local_err);
1657     if (local_err) {
1658         error_report_err(local_err);
1659         return H_HARDWARE;
1660     }
1661 
1662     /* Update CPUs */
1663     if (cpu->compat_pvr != cas_pvr) {
1664         ppc_set_compat_all(cas_pvr, &local_err);
1665         if (local_err) {
1666             /* We fail to set compat mode (likely because running with KVM PR),
1667              * but maybe we can fallback to raw mode if the guest supports it.
1668              */
1669             if (!raw_mode_supported) {
1670                 error_report_err(local_err);
1671                 return H_HARDWARE;
1672             }
1673             error_free(local_err);
1674             local_err = NULL;
1675         }
1676     }
1677 
1678     /* For the future use: here @ov_table points to the first option vector */
1679     ov_table = addr;
1680 
1681     ov1_guest = spapr_ovec_parse_vector(ov_table, 1);
1682     ov5_guest = spapr_ovec_parse_vector(ov_table, 5);
1683     if (spapr_ovec_test(ov5_guest, OV5_MMU_BOTH)) {
1684         error_report("guest requested hash and radix MMU, which is invalid.");
1685         exit(EXIT_FAILURE);
1686     }
1687     if (spapr_ovec_test(ov5_guest, OV5_XIVE_BOTH)) {
1688         error_report("guest requested an invalid interrupt mode");
1689         exit(EXIT_FAILURE);
1690     }
1691 
1692     /* The radix/hash bit in byte 24 requires special handling: */
1693     guest_radix = spapr_ovec_test(ov5_guest, OV5_MMU_RADIX_300);
1694     spapr_ovec_clear(ov5_guest, OV5_MMU_RADIX_300);
1695 
1696     guest_xive = spapr_ovec_test(ov5_guest, OV5_XIVE_EXPLOIT);
1697 
1698     /*
1699      * HPT resizing is a bit of a special case, because when enabled
1700      * we assume an HPT guest will support it until it says it
1701      * doesn't, instead of assuming it won't support it until it says
1702      * it does.  Strictly speaking that approach could break for
1703      * guests which don't make a CAS call, but those are so old we
1704      * don't care about them.  Without that assumption we'd have to
1705      * make at least a temporary allocation of an HPT sized for max
1706      * memory, which could be impossibly difficult under KVM HV if
1707      * maxram is large.
1708      */
1709     if (!guest_radix && !spapr_ovec_test(ov5_guest, OV5_HPT_RESIZE)) {
1710         int maxshift = spapr_hpt_shift_for_ramsize(MACHINE(spapr)->maxram_size);
1711 
1712         if (spapr->resize_hpt == SPAPR_RESIZE_HPT_REQUIRED) {
1713             error_report(
1714                 "h_client_architecture_support: Guest doesn't support HPT resizing, but resize-hpt=required");
1715             exit(1);
1716         }
1717 
1718         if (spapr->htab_shift < maxshift) {
1719             /* Guest doesn't know about HPT resizing, so we
1720              * pre-emptively resize for the maximum permitted RAM.  At
1721              * the point this is called, nothing should have been
1722              * entered into the existing HPT */
1723             spapr_reallocate_hpt(spapr, maxshift, &error_fatal);
1724             push_sregs_to_kvm_pr(spapr);
1725         }
1726     }
1727 
1728     /* NOTE: there are actually a number of ov5 bits where input from the
1729      * guest is always zero, and the platform/QEMU enables them independently
1730      * of guest input. To model these properly we'd want some sort of mask,
1731      * but since they only currently apply to memory migration as defined
1732      * by LoPAPR 1.1, 14.5.4.8, which QEMU doesn't implement, we don't need
1733      * to worry about this for now.
1734      */
1735     ov5_cas_old = spapr_ovec_clone(spapr->ov5_cas);
1736 
1737     /* also clear the radix/hash bit from the current ov5_cas bits to
1738      * be in sync with the newly ov5 bits. Else the radix bit will be
1739      * seen as being removed and this will generate a reset loop
1740      */
1741     spapr_ovec_clear(ov5_cas_old, OV5_MMU_RADIX_300);
1742 
1743     /* full range of negotiated ov5 capabilities */
1744     spapr_ovec_intersect(spapr->ov5_cas, spapr->ov5, ov5_guest);
1745     spapr_ovec_cleanup(ov5_guest);
1746     /* capabilities that have been added since CAS-generated guest reset.
1747      * if capabilities have since been removed, generate another reset
1748      */
1749     ov5_updates = spapr_ovec_new();
1750     spapr->cas_reboot = spapr_ovec_diff(ov5_updates,
1751                                         ov5_cas_old, spapr->ov5_cas);
1752     spapr_ovec_cleanup(ov5_cas_old);
1753     /* Now that processing is finished, set the radix/hash bit for the
1754      * guest if it requested a valid mode; otherwise terminate the boot. */
1755     if (guest_radix) {
1756         if (kvm_enabled() && !kvmppc_has_cap_mmu_radix()) {
1757             error_report("Guest requested unavailable MMU mode (radix).");
1758             exit(EXIT_FAILURE);
1759         }
1760         spapr_ovec_set(spapr->ov5_cas, OV5_MMU_RADIX_300);
1761     } else {
1762         if (kvm_enabled() && kvmppc_has_cap_mmu_radix()
1763             && !kvmppc_has_cap_mmu_hash_v3()) {
1764             error_report("Guest requested unavailable MMU mode (hash).");
1765             exit(EXIT_FAILURE);
1766         }
1767     }
1768     spapr->cas_pre_isa3_guest = !spapr_ovec_test(ov1_guest, OV1_PPC_3_00);
1769     spapr_ovec_cleanup(ov1_guest);
1770     if (!spapr->cas_reboot) {
1771         /* If spapr_machine_reset() did not set up a HPT but one is necessary
1772          * (because the guest isn't going to use radix) then set it up here. */
1773         if ((spapr->patb_entry & PATE1_GR) && !guest_radix) {
1774             /* legacy hash or new hash: */
1775             spapr_setup_hpt_and_vrma(spapr);
1776         }
1777         spapr->cas_reboot =
1778             (spapr_h_cas_compose_response(spapr, args[1], args[2],
1779                                           ov5_updates) != 0);
1780     }
1781 
1782     /*
1783      * Ensure the guest asks for an interrupt mode we support; otherwise
1784      * terminate the boot.
1785      */
1786     if (guest_xive) {
1787         if (!spapr->irq->xive) {
1788             error_report(
1789 "Guest requested unavailable interrupt mode (XIVE), try the ic-mode=xive or ic-mode=dual machine property");
1790             exit(EXIT_FAILURE);
1791         }
1792     } else {
1793         if (!spapr->irq->xics) {
1794             error_report(
1795 "Guest requested unavailable interrupt mode (XICS), either don't set the ic-mode machine property or try ic-mode=xics or ic-mode=dual");
1796             exit(EXIT_FAILURE);
1797         }
1798     }
1799 
1800     /*
1801      * Generate a machine reset when we have an update of the
1802      * interrupt mode. Only required when the machine supports both
1803      * modes.
1804      */
1805     if (!spapr->cas_reboot) {
1806         spapr->cas_reboot = spapr_ovec_test(ov5_updates, OV5_XIVE_EXPLOIT)
1807             && spapr->irq->xics && spapr->irq->xive;
1808     }
1809 
1810     spapr_ovec_cleanup(ov5_updates);
1811 
1812     if (spapr->cas_reboot) {
1813         qemu_system_reset_request(SHUTDOWN_CAUSE_SUBSYSTEM_RESET);
1814     }
1815 
1816     return H_SUCCESS;
1817 }
1818 
1819 static target_ulong h_home_node_associativity(PowerPCCPU *cpu,
1820                                               SpaprMachineState *spapr,
1821                                               target_ulong opcode,
1822                                               target_ulong *args)
1823 {
1824     target_ulong flags = args[0];
1825     target_ulong procno = args[1];
1826     PowerPCCPU *tcpu;
1827     int idx;
1828 
1829     /* only support procno from H_REGISTER_VPA */
1830     if (flags != 0x1) {
1831         return H_FUNCTION;
1832     }
1833 
1834     tcpu = spapr_find_cpu(procno);
1835     if (tcpu == NULL) {
1836         return H_P2;
1837     }
1838 
1839     /* sequence is the same as in the "ibm,associativity" property */
1840 
1841     idx = 0;
1842 #define ASSOCIATIVITY(a, b) (((uint64_t)(a) << 32) | \
1843                              ((uint64_t)(b) & 0xffffffff))
1844     args[idx++] = ASSOCIATIVITY(0, 0);
1845     args[idx++] = ASSOCIATIVITY(0, tcpu->node_id);
1846     args[idx++] = ASSOCIATIVITY(procno, -1);
1847     for ( ; idx < 6; idx++) {
1848         args[idx] = -1;
1849     }
1850 #undef ASSOCIATIVITY
1851 
1852     return H_SUCCESS;
1853 }
1854 
1855 static target_ulong h_get_cpu_characteristics(PowerPCCPU *cpu,
1856                                               SpaprMachineState *spapr,
1857                                               target_ulong opcode,
1858                                               target_ulong *args)
1859 {
1860     uint64_t characteristics = H_CPU_CHAR_HON_BRANCH_HINTS &
1861                                ~H_CPU_CHAR_THR_RECONF_TRIG;
1862     uint64_t behaviour = H_CPU_BEHAV_FAVOUR_SECURITY;
1863     uint8_t safe_cache = spapr_get_cap(spapr, SPAPR_CAP_CFPC);
1864     uint8_t safe_bounds_check = spapr_get_cap(spapr, SPAPR_CAP_SBBC);
1865     uint8_t safe_indirect_branch = spapr_get_cap(spapr, SPAPR_CAP_IBS);
1866     uint8_t count_cache_flush_assist = spapr_get_cap(spapr,
1867                                                      SPAPR_CAP_CCF_ASSIST);
1868 
1869     switch (safe_cache) {
1870     case SPAPR_CAP_WORKAROUND:
1871         characteristics |= H_CPU_CHAR_L1D_FLUSH_ORI30;
1872         characteristics |= H_CPU_CHAR_L1D_FLUSH_TRIG2;
1873         characteristics |= H_CPU_CHAR_L1D_THREAD_PRIV;
1874         behaviour |= H_CPU_BEHAV_L1D_FLUSH_PR;
1875         break;
1876     case SPAPR_CAP_FIXED:
1877         break;
1878     default: /* broken */
1879         assert(safe_cache == SPAPR_CAP_BROKEN);
1880         behaviour |= H_CPU_BEHAV_L1D_FLUSH_PR;
1881         break;
1882     }
1883 
1884     switch (safe_bounds_check) {
1885     case SPAPR_CAP_WORKAROUND:
1886         characteristics |= H_CPU_CHAR_SPEC_BAR_ORI31;
1887         behaviour |= H_CPU_BEHAV_BNDS_CHK_SPEC_BAR;
1888         break;
1889     case SPAPR_CAP_FIXED:
1890         break;
1891     default: /* broken */
1892         assert(safe_bounds_check == SPAPR_CAP_BROKEN);
1893         behaviour |= H_CPU_BEHAV_BNDS_CHK_SPEC_BAR;
1894         break;
1895     }
1896 
1897     switch (safe_indirect_branch) {
1898     case SPAPR_CAP_FIXED_NA:
1899         break;
1900     case SPAPR_CAP_FIXED_CCD:
1901         characteristics |= H_CPU_CHAR_CACHE_COUNT_DIS;
1902         break;
1903     case SPAPR_CAP_FIXED_IBS:
1904         characteristics |= H_CPU_CHAR_BCCTRL_SERIALISED;
1905         break;
1906     case SPAPR_CAP_WORKAROUND:
1907         behaviour |= H_CPU_BEHAV_FLUSH_COUNT_CACHE;
1908         if (count_cache_flush_assist) {
1909             characteristics |= H_CPU_CHAR_BCCTR_FLUSH_ASSIST;
1910         }
1911         break;
1912     default: /* broken */
1913         assert(safe_indirect_branch == SPAPR_CAP_BROKEN);
1914         break;
1915     }
1916 
1917     args[0] = characteristics;
1918     args[1] = behaviour;
1919     return H_SUCCESS;
1920 }
1921 
1922 static target_ulong h_update_dt(PowerPCCPU *cpu, SpaprMachineState *spapr,
1923                                 target_ulong opcode, target_ulong *args)
1924 {
1925     target_ulong dt = ppc64_phys_to_real(args[0]);
1926     struct fdt_header hdr = { 0 };
1927     unsigned cb;
1928     SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
1929     void *fdt;
1930 
1931     cpu_physical_memory_read(dt, &hdr, sizeof(hdr));
1932     cb = fdt32_to_cpu(hdr.totalsize);
1933 
1934     if (!smc->update_dt_enabled) {
1935         return H_SUCCESS;
1936     }
1937 
1938     /* Check that the fdt did not grow out of proportion */
1939     if (cb > spapr->fdt_initial_size * 2) {
1940         trace_spapr_update_dt_failed_size(spapr->fdt_initial_size, cb,
1941                                           fdt32_to_cpu(hdr.magic));
1942         return H_PARAMETER;
1943     }
1944 
1945     fdt = g_malloc0(cb);
1946     cpu_physical_memory_read(dt, fdt, cb);
1947 
1948     /* Check the fdt consistency */
1949     if (fdt_check_full(fdt, cb)) {
1950         trace_spapr_update_dt_failed_check(spapr->fdt_initial_size, cb,
1951                                            fdt32_to_cpu(hdr.magic));
1952         return H_PARAMETER;
1953     }
1954 
1955     g_free(spapr->fdt_blob);
1956     spapr->fdt_size = cb;
1957     spapr->fdt_blob = fdt;
1958     trace_spapr_update_dt(cb);
1959 
1960     return H_SUCCESS;
1961 }
1962 
1963 static spapr_hcall_fn papr_hypercall_table[(MAX_HCALL_OPCODE / 4) + 1];
1964 static spapr_hcall_fn kvmppc_hypercall_table[KVMPPC_HCALL_MAX - KVMPPC_HCALL_BASE + 1];
1965 static spapr_hcall_fn svm_hypercall_table[(SVM_HCALL_MAX - SVM_HCALL_BASE) / 4 + 1];
1966 
1967 void spapr_register_hypercall(target_ulong opcode, spapr_hcall_fn fn)
1968 {
1969     spapr_hcall_fn *slot;
1970 
1971     if (opcode <= MAX_HCALL_OPCODE) {
1972         assert((opcode & 0x3) == 0);
1973 
1974         slot = &papr_hypercall_table[opcode / 4];
1975     } else if (opcode >= SVM_HCALL_BASE && opcode <= SVM_HCALL_MAX) {
1976         /* we only have SVM-related hcall numbers assigned in multiples of 4 */
1977         assert((opcode & 0x3) == 0);
1978 
1979         slot = &svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
1980     } else {
1981         assert((opcode >= KVMPPC_HCALL_BASE) && (opcode <= KVMPPC_HCALL_MAX));
1982 
1983         slot = &kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
1984     }
1985 
1986     assert(!(*slot));
1987     *slot = fn;
1988 }
1989 
1990 target_ulong spapr_hypercall(PowerPCCPU *cpu, target_ulong opcode,
1991                              target_ulong *args)
1992 {
1993     SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1994 
1995     if ((opcode <= MAX_HCALL_OPCODE)
1996         && ((opcode & 0x3) == 0)) {
1997         spapr_hcall_fn fn = papr_hypercall_table[opcode / 4];
1998 
1999         if (fn) {
2000             return fn(cpu, spapr, opcode, args);
2001         }
2002     } else if ((opcode >= SVM_HCALL_BASE) &&
2003                (opcode <= SVM_HCALL_MAX)) {
2004         spapr_hcall_fn fn = svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
2005 
2006         if (fn) {
2007             return fn(cpu, spapr, opcode, args);
2008         }
2009     } else if ((opcode >= KVMPPC_HCALL_BASE) &&
2010                (opcode <= KVMPPC_HCALL_MAX)) {
2011         spapr_hcall_fn fn = kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
2012 
2013         if (fn) {
2014             return fn(cpu, spapr, opcode, args);
2015         }
2016     }
2017 
2018     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x" TARGET_FMT_lx "\n",
2019                   opcode);
2020     return H_FUNCTION;
2021 }
2022 
2023 static void hypercall_register_types(void)
2024 {
2025     /* hcall-pft */
2026     spapr_register_hypercall(H_ENTER, h_enter);
2027     spapr_register_hypercall(H_REMOVE, h_remove);
2028     spapr_register_hypercall(H_PROTECT, h_protect);
2029     spapr_register_hypercall(H_READ, h_read);
2030 
2031     /* hcall-bulk */
2032     spapr_register_hypercall(H_BULK_REMOVE, h_bulk_remove);
2033 
2034     /* hcall-hpt-resize */
2035     spapr_register_hypercall(H_RESIZE_HPT_PREPARE, h_resize_hpt_prepare);
2036     spapr_register_hypercall(H_RESIZE_HPT_COMMIT, h_resize_hpt_commit);
2037 
2038     /* hcall-splpar */
2039     spapr_register_hypercall(H_REGISTER_VPA, h_register_vpa);
2040     spapr_register_hypercall(H_CEDE, h_cede);
2041     spapr_register_hypercall(H_CONFER, h_confer);
2042     spapr_register_hypercall(H_PROD, h_prod);
2043 
2044     /* hcall-join */
2045     spapr_register_hypercall(H_JOIN, h_join);
2046 
2047     spapr_register_hypercall(H_SIGNAL_SYS_RESET, h_signal_sys_reset);
2048 
2049     /* processor register resource access h-calls */
2050     spapr_register_hypercall(H_SET_SPRG0, h_set_sprg0);
2051     spapr_register_hypercall(H_SET_DABR, h_set_dabr);
2052     spapr_register_hypercall(H_SET_XDABR, h_set_xdabr);
2053     spapr_register_hypercall(H_PAGE_INIT, h_page_init);
2054     spapr_register_hypercall(H_SET_MODE, h_set_mode);
2055 
2056     /* In Memory Table MMU h-calls */
2057     spapr_register_hypercall(H_CLEAN_SLB, h_clean_slb);
2058     spapr_register_hypercall(H_INVALIDATE_PID, h_invalidate_pid);
2059     spapr_register_hypercall(H_REGISTER_PROC_TBL, h_register_process_table);
2060 
2061     /* hcall-get-cpu-characteristics */
2062     spapr_register_hypercall(H_GET_CPU_CHARACTERISTICS,
2063                              h_get_cpu_characteristics);
2064 
2065     /* "debugger" hcalls (also used by SLOF). Note: We do -not- differenciate
2066      * here between the "CI" and the "CACHE" variants, they will use whatever
2067      * mapping attributes qemu is using. When using KVM, the kernel will
2068      * enforce the attributes more strongly
2069      */
2070     spapr_register_hypercall(H_LOGICAL_CI_LOAD, h_logical_load);
2071     spapr_register_hypercall(H_LOGICAL_CI_STORE, h_logical_store);
2072     spapr_register_hypercall(H_LOGICAL_CACHE_LOAD, h_logical_load);
2073     spapr_register_hypercall(H_LOGICAL_CACHE_STORE, h_logical_store);
2074     spapr_register_hypercall(H_LOGICAL_ICBI, h_logical_icbi);
2075     spapr_register_hypercall(H_LOGICAL_DCBF, h_logical_dcbf);
2076     spapr_register_hypercall(KVMPPC_H_LOGICAL_MEMOP, h_logical_memop);
2077 
2078     /* qemu/KVM-PPC specific hcalls */
2079     spapr_register_hypercall(KVMPPC_H_RTAS, h_rtas);
2080 
2081     /* ibm,client-architecture-support support */
2082     spapr_register_hypercall(KVMPPC_H_CAS, h_client_architecture_support);
2083 
2084     spapr_register_hypercall(KVMPPC_H_UPDATE_DT, h_update_dt);
2085 
2086     /* Virtual Processor Home Node */
2087     spapr_register_hypercall(H_HOME_NODE_ASSOCIATIVITY,
2088                              h_home_node_associativity);
2089 }
2090 
2091 type_init(hypercall_register_types)
2092