xref: /openbmc/qemu/hw/ppc/spapr_hcall.c (revision 88cd34ee)
1 #include "qemu/osdep.h"
2 #include "qemu/cutils.h"
3 #include "qapi/error.h"
4 #include "sysemu/hw_accel.h"
5 #include "sysemu/runstate.h"
6 #include "qemu/log.h"
7 #include "qemu/main-loop.h"
8 #include "qemu/module.h"
9 #include "qemu/error-report.h"
10 #include "cpu.h"
11 #include "exec/exec-all.h"
12 #include "helper_regs.h"
13 #include "hw/ppc/spapr.h"
14 #include "hw/ppc/spapr_cpu_core.h"
15 #include "mmu-hash64.h"
16 #include "cpu-models.h"
17 #include "trace.h"
18 #include "kvm_ppc.h"
19 #include "hw/ppc/fdt.h"
20 #include "hw/ppc/spapr_ovec.h"
21 #include "mmu-book3s-v3.h"
22 #include "hw/mem/memory-device.h"
23 
24 static bool has_spr(PowerPCCPU *cpu, int spr)
25 {
26     /* We can test whether the SPR is defined by checking for a valid name */
27     return cpu->env.spr_cb[spr].name != NULL;
28 }
29 
30 static inline bool valid_ptex(PowerPCCPU *cpu, target_ulong ptex)
31 {
32     /*
33      * hash value/pteg group index is normalized by HPT mask
34      */
35     if (((ptex & ~7ULL) / HPTES_PER_GROUP) & ~ppc_hash64_hpt_mask(cpu)) {
36         return false;
37     }
38     return true;
39 }
40 
41 static bool is_ram_address(SpaprMachineState *spapr, hwaddr addr)
42 {
43     MachineState *machine = MACHINE(spapr);
44     DeviceMemoryState *dms = machine->device_memory;
45 
46     if (addr < machine->ram_size) {
47         return true;
48     }
49     if ((addr >= dms->base)
50         && ((addr - dms->base) < memory_region_size(&dms->mr))) {
51         return true;
52     }
53 
54     return false;
55 }
56 
57 static target_ulong h_enter(PowerPCCPU *cpu, SpaprMachineState *spapr,
58                             target_ulong opcode, target_ulong *args)
59 {
60     target_ulong flags = args[0];
61     target_ulong ptex = args[1];
62     target_ulong pteh = args[2];
63     target_ulong ptel = args[3];
64     unsigned apshift;
65     target_ulong raddr;
66     target_ulong slot;
67     const ppc_hash_pte64_t *hptes;
68 
69     apshift = ppc_hash64_hpte_page_shift_noslb(cpu, pteh, ptel);
70     if (!apshift) {
71         /* Bad page size encoding */
72         return H_PARAMETER;
73     }
74 
75     raddr = (ptel & HPTE64_R_RPN) & ~((1ULL << apshift) - 1);
76 
77     if (is_ram_address(spapr, raddr)) {
78         /* Regular RAM - should have WIMG=0010 */
79         if ((ptel & HPTE64_R_WIMG) != HPTE64_R_M) {
80             return H_PARAMETER;
81         }
82     } else {
83         target_ulong wimg_flags;
84         /* Looks like an IO address */
85         /* FIXME: What WIMG combinations could be sensible for IO?
86          * For now we allow WIMG=010x, but are there others? */
87         /* FIXME: Should we check against registered IO addresses? */
88         wimg_flags = (ptel & (HPTE64_R_W | HPTE64_R_I | HPTE64_R_M));
89 
90         if (wimg_flags != HPTE64_R_I &&
91             wimg_flags != (HPTE64_R_I | HPTE64_R_M)) {
92             return H_PARAMETER;
93         }
94     }
95 
96     pteh &= ~0x60ULL;
97 
98     if (!valid_ptex(cpu, ptex)) {
99         return H_PARAMETER;
100     }
101 
102     slot = ptex & 7ULL;
103     ptex = ptex & ~7ULL;
104 
105     if (likely((flags & H_EXACT) == 0)) {
106         hptes = ppc_hash64_map_hptes(cpu, ptex, HPTES_PER_GROUP);
107         for (slot = 0; slot < 8; slot++) {
108             if (!(ppc_hash64_hpte0(cpu, hptes, slot) & HPTE64_V_VALID)) {
109                 break;
110             }
111         }
112         ppc_hash64_unmap_hptes(cpu, hptes, ptex, HPTES_PER_GROUP);
113         if (slot == 8) {
114             return H_PTEG_FULL;
115         }
116     } else {
117         hptes = ppc_hash64_map_hptes(cpu, ptex + slot, 1);
118         if (ppc_hash64_hpte0(cpu, hptes, 0) & HPTE64_V_VALID) {
119             ppc_hash64_unmap_hptes(cpu, hptes, ptex + slot, 1);
120             return H_PTEG_FULL;
121         }
122         ppc_hash64_unmap_hptes(cpu, hptes, ptex, 1);
123     }
124 
125     spapr_store_hpte(cpu, ptex + slot, pteh | HPTE64_V_HPTE_DIRTY, ptel);
126 
127     args[0] = ptex + slot;
128     return H_SUCCESS;
129 }
130 
131 typedef enum {
132     REMOVE_SUCCESS = 0,
133     REMOVE_NOT_FOUND = 1,
134     REMOVE_PARM = 2,
135     REMOVE_HW = 3,
136 } RemoveResult;
137 
138 static RemoveResult remove_hpte(PowerPCCPU *cpu
139                                 , target_ulong ptex,
140                                 target_ulong avpn,
141                                 target_ulong flags,
142                                 target_ulong *vp, target_ulong *rp)
143 {
144     const ppc_hash_pte64_t *hptes;
145     target_ulong v, r;
146 
147     if (!valid_ptex(cpu, ptex)) {
148         return REMOVE_PARM;
149     }
150 
151     hptes = ppc_hash64_map_hptes(cpu, ptex, 1);
152     v = ppc_hash64_hpte0(cpu, hptes, 0);
153     r = ppc_hash64_hpte1(cpu, hptes, 0);
154     ppc_hash64_unmap_hptes(cpu, hptes, ptex, 1);
155 
156     if ((v & HPTE64_V_VALID) == 0 ||
157         ((flags & H_AVPN) && (v & ~0x7fULL) != avpn) ||
158         ((flags & H_ANDCOND) && (v & avpn) != 0)) {
159         return REMOVE_NOT_FOUND;
160     }
161     *vp = v;
162     *rp = r;
163     spapr_store_hpte(cpu, ptex, HPTE64_V_HPTE_DIRTY, 0);
164     ppc_hash64_tlb_flush_hpte(cpu, ptex, v, r);
165     return REMOVE_SUCCESS;
166 }
167 
168 static target_ulong h_remove(PowerPCCPU *cpu, SpaprMachineState *spapr,
169                              target_ulong opcode, target_ulong *args)
170 {
171     CPUPPCState *env = &cpu->env;
172     target_ulong flags = args[0];
173     target_ulong ptex = args[1];
174     target_ulong avpn = args[2];
175     RemoveResult ret;
176 
177     ret = remove_hpte(cpu, ptex, avpn, flags,
178                       &args[0], &args[1]);
179 
180     switch (ret) {
181     case REMOVE_SUCCESS:
182         check_tlb_flush(env, true);
183         return H_SUCCESS;
184 
185     case REMOVE_NOT_FOUND:
186         return H_NOT_FOUND;
187 
188     case REMOVE_PARM:
189         return H_PARAMETER;
190 
191     case REMOVE_HW:
192         return H_HARDWARE;
193     }
194 
195     g_assert_not_reached();
196 }
197 
198 #define H_BULK_REMOVE_TYPE             0xc000000000000000ULL
199 #define   H_BULK_REMOVE_REQUEST        0x4000000000000000ULL
200 #define   H_BULK_REMOVE_RESPONSE       0x8000000000000000ULL
201 #define   H_BULK_REMOVE_END            0xc000000000000000ULL
202 #define H_BULK_REMOVE_CODE             0x3000000000000000ULL
203 #define   H_BULK_REMOVE_SUCCESS        0x0000000000000000ULL
204 #define   H_BULK_REMOVE_NOT_FOUND      0x1000000000000000ULL
205 #define   H_BULK_REMOVE_PARM           0x2000000000000000ULL
206 #define   H_BULK_REMOVE_HW             0x3000000000000000ULL
207 #define H_BULK_REMOVE_RC               0x0c00000000000000ULL
208 #define H_BULK_REMOVE_FLAGS            0x0300000000000000ULL
209 #define   H_BULK_REMOVE_ABSOLUTE       0x0000000000000000ULL
210 #define   H_BULK_REMOVE_ANDCOND        0x0100000000000000ULL
211 #define   H_BULK_REMOVE_AVPN           0x0200000000000000ULL
212 #define H_BULK_REMOVE_PTEX             0x00ffffffffffffffULL
213 
214 #define H_BULK_REMOVE_MAX_BATCH        4
215 
216 static target_ulong h_bulk_remove(PowerPCCPU *cpu, SpaprMachineState *spapr,
217                                   target_ulong opcode, target_ulong *args)
218 {
219     CPUPPCState *env = &cpu->env;
220     int i;
221     target_ulong rc = H_SUCCESS;
222 
223     for (i = 0; i < H_BULK_REMOVE_MAX_BATCH; i++) {
224         target_ulong *tsh = &args[i*2];
225         target_ulong tsl = args[i*2 + 1];
226         target_ulong v, r, ret;
227 
228         if ((*tsh & H_BULK_REMOVE_TYPE) == H_BULK_REMOVE_END) {
229             break;
230         } else if ((*tsh & H_BULK_REMOVE_TYPE) != H_BULK_REMOVE_REQUEST) {
231             return H_PARAMETER;
232         }
233 
234         *tsh &= H_BULK_REMOVE_PTEX | H_BULK_REMOVE_FLAGS;
235         *tsh |= H_BULK_REMOVE_RESPONSE;
236 
237         if ((*tsh & H_BULK_REMOVE_ANDCOND) && (*tsh & H_BULK_REMOVE_AVPN)) {
238             *tsh |= H_BULK_REMOVE_PARM;
239             return H_PARAMETER;
240         }
241 
242         ret = remove_hpte(cpu, *tsh & H_BULK_REMOVE_PTEX, tsl,
243                           (*tsh & H_BULK_REMOVE_FLAGS) >> 26,
244                           &v, &r);
245 
246         *tsh |= ret << 60;
247 
248         switch (ret) {
249         case REMOVE_SUCCESS:
250             *tsh |= (r & (HPTE64_R_C | HPTE64_R_R)) << 43;
251             break;
252 
253         case REMOVE_PARM:
254             rc = H_PARAMETER;
255             goto exit;
256 
257         case REMOVE_HW:
258             rc = H_HARDWARE;
259             goto exit;
260         }
261     }
262  exit:
263     check_tlb_flush(env, true);
264 
265     return rc;
266 }
267 
268 static target_ulong h_protect(PowerPCCPU *cpu, SpaprMachineState *spapr,
269                               target_ulong opcode, target_ulong *args)
270 {
271     CPUPPCState *env = &cpu->env;
272     target_ulong flags = args[0];
273     target_ulong ptex = args[1];
274     target_ulong avpn = args[2];
275     const ppc_hash_pte64_t *hptes;
276     target_ulong v, r;
277 
278     if (!valid_ptex(cpu, ptex)) {
279         return H_PARAMETER;
280     }
281 
282     hptes = ppc_hash64_map_hptes(cpu, ptex, 1);
283     v = ppc_hash64_hpte0(cpu, hptes, 0);
284     r = ppc_hash64_hpte1(cpu, hptes, 0);
285     ppc_hash64_unmap_hptes(cpu, hptes, ptex, 1);
286 
287     if ((v & HPTE64_V_VALID) == 0 ||
288         ((flags & H_AVPN) && (v & ~0x7fULL) != avpn)) {
289         return H_NOT_FOUND;
290     }
291 
292     r &= ~(HPTE64_R_PP0 | HPTE64_R_PP | HPTE64_R_N |
293            HPTE64_R_KEY_HI | HPTE64_R_KEY_LO);
294     r |= (flags << 55) & HPTE64_R_PP0;
295     r |= (flags << 48) & HPTE64_R_KEY_HI;
296     r |= flags & (HPTE64_R_PP | HPTE64_R_N | HPTE64_R_KEY_LO);
297     spapr_store_hpte(cpu, ptex,
298                      (v & ~HPTE64_V_VALID) | HPTE64_V_HPTE_DIRTY, 0);
299     ppc_hash64_tlb_flush_hpte(cpu, ptex, v, r);
300     /* Flush the tlb */
301     check_tlb_flush(env, true);
302     /* Don't need a memory barrier, due to qemu's global lock */
303     spapr_store_hpte(cpu, ptex, v | HPTE64_V_HPTE_DIRTY, r);
304     return H_SUCCESS;
305 }
306 
307 static target_ulong h_read(PowerPCCPU *cpu, SpaprMachineState *spapr,
308                            target_ulong opcode, target_ulong *args)
309 {
310     target_ulong flags = args[0];
311     target_ulong ptex = args[1];
312     int i, ridx, n_entries = 1;
313     const ppc_hash_pte64_t *hptes;
314 
315     if (!valid_ptex(cpu, ptex)) {
316         return H_PARAMETER;
317     }
318 
319     if (flags & H_READ_4) {
320         /* Clear the two low order bits */
321         ptex &= ~(3ULL);
322         n_entries = 4;
323     }
324 
325     hptes = ppc_hash64_map_hptes(cpu, ptex, n_entries);
326     for (i = 0, ridx = 0; i < n_entries; i++) {
327         args[ridx++] = ppc_hash64_hpte0(cpu, hptes, i);
328         args[ridx++] = ppc_hash64_hpte1(cpu, hptes, i);
329     }
330     ppc_hash64_unmap_hptes(cpu, hptes, ptex, n_entries);
331 
332     return H_SUCCESS;
333 }
334 
335 struct SpaprPendingHpt {
336     /* These fields are read-only after initialization */
337     int shift;
338     QemuThread thread;
339 
340     /* These fields are protected by the BQL */
341     bool complete;
342 
343     /* These fields are private to the preparation thread if
344      * !complete, otherwise protected by the BQL */
345     int ret;
346     void *hpt;
347 };
348 
349 static void free_pending_hpt(SpaprPendingHpt *pending)
350 {
351     if (pending->hpt) {
352         qemu_vfree(pending->hpt);
353     }
354 
355     g_free(pending);
356 }
357 
358 static void *hpt_prepare_thread(void *opaque)
359 {
360     SpaprPendingHpt *pending = opaque;
361     size_t size = 1ULL << pending->shift;
362 
363     pending->hpt = qemu_memalign(size, size);
364     if (pending->hpt) {
365         memset(pending->hpt, 0, size);
366         pending->ret = H_SUCCESS;
367     } else {
368         pending->ret = H_NO_MEM;
369     }
370 
371     qemu_mutex_lock_iothread();
372 
373     if (SPAPR_MACHINE(qdev_get_machine())->pending_hpt == pending) {
374         /* Ready to go */
375         pending->complete = true;
376     } else {
377         /* We've been cancelled, clean ourselves up */
378         free_pending_hpt(pending);
379     }
380 
381     qemu_mutex_unlock_iothread();
382     return NULL;
383 }
384 
385 /* Must be called with BQL held */
386 static void cancel_hpt_prepare(SpaprMachineState *spapr)
387 {
388     SpaprPendingHpt *pending = spapr->pending_hpt;
389 
390     /* Let the thread know it's cancelled */
391     spapr->pending_hpt = NULL;
392 
393     if (!pending) {
394         /* Nothing to do */
395         return;
396     }
397 
398     if (!pending->complete) {
399         /* thread will clean itself up */
400         return;
401     }
402 
403     free_pending_hpt(pending);
404 }
405 
406 /* Convert a return code from the KVM ioctl()s implementing resize HPT
407  * into a PAPR hypercall return code */
408 static target_ulong resize_hpt_convert_rc(int ret)
409 {
410     if (ret >= 100000) {
411         return H_LONG_BUSY_ORDER_100_SEC;
412     } else if (ret >= 10000) {
413         return H_LONG_BUSY_ORDER_10_SEC;
414     } else if (ret >= 1000) {
415         return H_LONG_BUSY_ORDER_1_SEC;
416     } else if (ret >= 100) {
417         return H_LONG_BUSY_ORDER_100_MSEC;
418     } else if (ret >= 10) {
419         return H_LONG_BUSY_ORDER_10_MSEC;
420     } else if (ret > 0) {
421         return H_LONG_BUSY_ORDER_1_MSEC;
422     }
423 
424     switch (ret) {
425     case 0:
426         return H_SUCCESS;
427     case -EPERM:
428         return H_AUTHORITY;
429     case -EINVAL:
430         return H_PARAMETER;
431     case -ENXIO:
432         return H_CLOSED;
433     case -ENOSPC:
434         return H_PTEG_FULL;
435     case -EBUSY:
436         return H_BUSY;
437     case -ENOMEM:
438         return H_NO_MEM;
439     default:
440         return H_HARDWARE;
441     }
442 }
443 
444 static target_ulong h_resize_hpt_prepare(PowerPCCPU *cpu,
445                                          SpaprMachineState *spapr,
446                                          target_ulong opcode,
447                                          target_ulong *args)
448 {
449     target_ulong flags = args[0];
450     int shift = args[1];
451     SpaprPendingHpt *pending = spapr->pending_hpt;
452     uint64_t current_ram_size;
453     int rc;
454 
455     if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
456         return H_AUTHORITY;
457     }
458 
459     if (!spapr->htab_shift) {
460         /* Radix guest, no HPT */
461         return H_NOT_AVAILABLE;
462     }
463 
464     trace_spapr_h_resize_hpt_prepare(flags, shift);
465 
466     if (flags != 0) {
467         return H_PARAMETER;
468     }
469 
470     if (shift && ((shift < 18) || (shift > 46))) {
471         return H_PARAMETER;
472     }
473 
474     current_ram_size = MACHINE(spapr)->ram_size + get_plugged_memory_size();
475 
476     /* We only allow the guest to allocate an HPT one order above what
477      * we'd normally give them (to stop a small guest claiming a huge
478      * chunk of resources in the HPT */
479     if (shift > (spapr_hpt_shift_for_ramsize(current_ram_size) + 1)) {
480         return H_RESOURCE;
481     }
482 
483     rc = kvmppc_resize_hpt_prepare(cpu, flags, shift);
484     if (rc != -ENOSYS) {
485         return resize_hpt_convert_rc(rc);
486     }
487 
488     if (pending) {
489         /* something already in progress */
490         if (pending->shift == shift) {
491             /* and it's suitable */
492             if (pending->complete) {
493                 return pending->ret;
494             } else {
495                 return H_LONG_BUSY_ORDER_100_MSEC;
496             }
497         }
498 
499         /* not suitable, cancel and replace */
500         cancel_hpt_prepare(spapr);
501     }
502 
503     if (!shift) {
504         /* nothing to do */
505         return H_SUCCESS;
506     }
507 
508     /* start new prepare */
509 
510     pending = g_new0(SpaprPendingHpt, 1);
511     pending->shift = shift;
512     pending->ret = H_HARDWARE;
513 
514     qemu_thread_create(&pending->thread, "sPAPR HPT prepare",
515                        hpt_prepare_thread, pending, QEMU_THREAD_DETACHED);
516 
517     spapr->pending_hpt = pending;
518 
519     /* In theory we could estimate the time more accurately based on
520      * the new size, but there's not much point */
521     return H_LONG_BUSY_ORDER_100_MSEC;
522 }
523 
524 static uint64_t new_hpte_load0(void *htab, uint64_t pteg, int slot)
525 {
526     uint8_t *addr = htab;
527 
528     addr += pteg * HASH_PTEG_SIZE_64;
529     addr += slot * HASH_PTE_SIZE_64;
530     return  ldq_p(addr);
531 }
532 
533 static void new_hpte_store(void *htab, uint64_t pteg, int slot,
534                            uint64_t pte0, uint64_t pte1)
535 {
536     uint8_t *addr = htab;
537 
538     addr += pteg * HASH_PTEG_SIZE_64;
539     addr += slot * HASH_PTE_SIZE_64;
540 
541     stq_p(addr, pte0);
542     stq_p(addr + HASH_PTE_SIZE_64 / 2, pte1);
543 }
544 
545 static int rehash_hpte(PowerPCCPU *cpu,
546                        const ppc_hash_pte64_t *hptes,
547                        void *old_hpt, uint64_t oldsize,
548                        void *new_hpt, uint64_t newsize,
549                        uint64_t pteg, int slot)
550 {
551     uint64_t old_hash_mask = (oldsize >> 7) - 1;
552     uint64_t new_hash_mask = (newsize >> 7) - 1;
553     target_ulong pte0 = ppc_hash64_hpte0(cpu, hptes, slot);
554     target_ulong pte1;
555     uint64_t avpn;
556     unsigned base_pg_shift;
557     uint64_t hash, new_pteg, replace_pte0;
558 
559     if (!(pte0 & HPTE64_V_VALID) || !(pte0 & HPTE64_V_BOLTED)) {
560         return H_SUCCESS;
561     }
562 
563     pte1 = ppc_hash64_hpte1(cpu, hptes, slot);
564 
565     base_pg_shift = ppc_hash64_hpte_page_shift_noslb(cpu, pte0, pte1);
566     assert(base_pg_shift); /* H_ENTER shouldn't allow a bad encoding */
567     avpn = HPTE64_V_AVPN_VAL(pte0) & ~(((1ULL << base_pg_shift) - 1) >> 23);
568 
569     if (pte0 & HPTE64_V_SECONDARY) {
570         pteg = ~pteg;
571     }
572 
573     if ((pte0 & HPTE64_V_SSIZE) == HPTE64_V_SSIZE_256M) {
574         uint64_t offset, vsid;
575 
576         /* We only have 28 - 23 bits of offset in avpn */
577         offset = (avpn & 0x1f) << 23;
578         vsid = avpn >> 5;
579         /* We can find more bits from the pteg value */
580         if (base_pg_shift < 23) {
581             offset |= ((vsid ^ pteg) & old_hash_mask) << base_pg_shift;
582         }
583 
584         hash = vsid ^ (offset >> base_pg_shift);
585     } else if ((pte0 & HPTE64_V_SSIZE) == HPTE64_V_SSIZE_1T) {
586         uint64_t offset, vsid;
587 
588         /* We only have 40 - 23 bits of seg_off in avpn */
589         offset = (avpn & 0x1ffff) << 23;
590         vsid = avpn >> 17;
591         if (base_pg_shift < 23) {
592             offset |= ((vsid ^ (vsid << 25) ^ pteg) & old_hash_mask)
593                 << base_pg_shift;
594         }
595 
596         hash = vsid ^ (vsid << 25) ^ (offset >> base_pg_shift);
597     } else {
598         error_report("rehash_pte: Bad segment size in HPTE");
599         return H_HARDWARE;
600     }
601 
602     new_pteg = hash & new_hash_mask;
603     if (pte0 & HPTE64_V_SECONDARY) {
604         assert(~pteg == (hash & old_hash_mask));
605         new_pteg = ~new_pteg;
606     } else {
607         assert(pteg == (hash & old_hash_mask));
608     }
609     assert((oldsize != newsize) || (pteg == new_pteg));
610     replace_pte0 = new_hpte_load0(new_hpt, new_pteg, slot);
611     /*
612      * Strictly speaking, we don't need all these tests, since we only
613      * ever rehash bolted HPTEs.  We might in future handle non-bolted
614      * HPTEs, though so make the logic correct for those cases as
615      * well.
616      */
617     if (replace_pte0 & HPTE64_V_VALID) {
618         assert(newsize < oldsize);
619         if (replace_pte0 & HPTE64_V_BOLTED) {
620             if (pte0 & HPTE64_V_BOLTED) {
621                 /* Bolted collision, nothing we can do */
622                 return H_PTEG_FULL;
623             } else {
624                 /* Discard this hpte */
625                 return H_SUCCESS;
626             }
627         }
628     }
629 
630     new_hpte_store(new_hpt, new_pteg, slot, pte0, pte1);
631     return H_SUCCESS;
632 }
633 
634 static int rehash_hpt(PowerPCCPU *cpu,
635                       void *old_hpt, uint64_t oldsize,
636                       void *new_hpt, uint64_t newsize)
637 {
638     uint64_t n_ptegs = oldsize >> 7;
639     uint64_t pteg;
640     int slot;
641     int rc;
642 
643     for (pteg = 0; pteg < n_ptegs; pteg++) {
644         hwaddr ptex = pteg * HPTES_PER_GROUP;
645         const ppc_hash_pte64_t *hptes
646             = ppc_hash64_map_hptes(cpu, ptex, HPTES_PER_GROUP);
647 
648         if (!hptes) {
649             return H_HARDWARE;
650         }
651 
652         for (slot = 0; slot < HPTES_PER_GROUP; slot++) {
653             rc = rehash_hpte(cpu, hptes, old_hpt, oldsize, new_hpt, newsize,
654                              pteg, slot);
655             if (rc != H_SUCCESS) {
656                 ppc_hash64_unmap_hptes(cpu, hptes, ptex, HPTES_PER_GROUP);
657                 return rc;
658             }
659         }
660         ppc_hash64_unmap_hptes(cpu, hptes, ptex, HPTES_PER_GROUP);
661     }
662 
663     return H_SUCCESS;
664 }
665 
666 static void do_push_sregs_to_kvm_pr(CPUState *cs, run_on_cpu_data data)
667 {
668     int ret;
669 
670     cpu_synchronize_state(cs);
671 
672     ret = kvmppc_put_books_sregs(POWERPC_CPU(cs));
673     if (ret < 0) {
674         error_report("failed to push sregs to KVM: %s", strerror(-ret));
675         exit(1);
676     }
677 }
678 
679 static void push_sregs_to_kvm_pr(SpaprMachineState *spapr)
680 {
681     CPUState *cs;
682 
683     /*
684      * This is a hack for the benefit of KVM PR - it abuses the SDR1
685      * slot in kvm_sregs to communicate the userspace address of the
686      * HPT
687      */
688     if (!kvm_enabled() || !spapr->htab) {
689         return;
690     }
691 
692     CPU_FOREACH(cs) {
693         run_on_cpu(cs, do_push_sregs_to_kvm_pr, RUN_ON_CPU_NULL);
694     }
695 }
696 
697 static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
698                                         SpaprMachineState *spapr,
699                                         target_ulong opcode,
700                                         target_ulong *args)
701 {
702     target_ulong flags = args[0];
703     target_ulong shift = args[1];
704     SpaprPendingHpt *pending = spapr->pending_hpt;
705     int rc;
706     size_t newsize;
707 
708     if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
709         return H_AUTHORITY;
710     }
711 
712     if (!spapr->htab_shift) {
713         /* Radix guest, no HPT */
714         return H_NOT_AVAILABLE;
715     }
716 
717     trace_spapr_h_resize_hpt_commit(flags, shift);
718 
719     rc = kvmppc_resize_hpt_commit(cpu, flags, shift);
720     if (rc != -ENOSYS) {
721         rc = resize_hpt_convert_rc(rc);
722         if (rc == H_SUCCESS) {
723             /* Need to set the new htab_shift in the machine state */
724             spapr->htab_shift = shift;
725         }
726         return rc;
727     }
728 
729     if (flags != 0) {
730         return H_PARAMETER;
731     }
732 
733     if (!pending || (pending->shift != shift)) {
734         /* no matching prepare */
735         return H_CLOSED;
736     }
737 
738     if (!pending->complete) {
739         /* prepare has not completed */
740         return H_BUSY;
741     }
742 
743     /* Shouldn't have got past PREPARE without an HPT */
744     g_assert(spapr->htab_shift);
745 
746     newsize = 1ULL << pending->shift;
747     rc = rehash_hpt(cpu, spapr->htab, HTAB_SIZE(spapr),
748                     pending->hpt, newsize);
749     if (rc == H_SUCCESS) {
750         qemu_vfree(spapr->htab);
751         spapr->htab = pending->hpt;
752         spapr->htab_shift = pending->shift;
753 
754         push_sregs_to_kvm_pr(spapr);
755 
756         pending->hpt = NULL; /* so it's not free()d */
757     }
758 
759     /* Clean up */
760     spapr->pending_hpt = NULL;
761     free_pending_hpt(pending);
762 
763     return rc;
764 }
765 
766 static target_ulong h_set_sprg0(PowerPCCPU *cpu, SpaprMachineState *spapr,
767                                 target_ulong opcode, target_ulong *args)
768 {
769     cpu_synchronize_state(CPU(cpu));
770     cpu->env.spr[SPR_SPRG0] = args[0];
771 
772     return H_SUCCESS;
773 }
774 
775 static target_ulong h_set_dabr(PowerPCCPU *cpu, SpaprMachineState *spapr,
776                                target_ulong opcode, target_ulong *args)
777 {
778     if (!has_spr(cpu, SPR_DABR)) {
779         return H_HARDWARE;              /* DABR register not available */
780     }
781     cpu_synchronize_state(CPU(cpu));
782 
783     if (has_spr(cpu, SPR_DABRX)) {
784         cpu->env.spr[SPR_DABRX] = 0x3;  /* Use Problem and Privileged state */
785     } else if (!(args[0] & 0x4)) {      /* Breakpoint Translation set? */
786         return H_RESERVED_DABR;
787     }
788 
789     cpu->env.spr[SPR_DABR] = args[0];
790     return H_SUCCESS;
791 }
792 
793 static target_ulong h_set_xdabr(PowerPCCPU *cpu, SpaprMachineState *spapr,
794                                 target_ulong opcode, target_ulong *args)
795 {
796     target_ulong dabrx = args[1];
797 
798     if (!has_spr(cpu, SPR_DABR) || !has_spr(cpu, SPR_DABRX)) {
799         return H_HARDWARE;
800     }
801 
802     if ((dabrx & ~0xfULL) != 0 || (dabrx & H_DABRX_HYPERVISOR) != 0
803         || (dabrx & (H_DABRX_KERNEL | H_DABRX_USER)) == 0) {
804         return H_PARAMETER;
805     }
806 
807     cpu_synchronize_state(CPU(cpu));
808     cpu->env.spr[SPR_DABRX] = dabrx;
809     cpu->env.spr[SPR_DABR] = args[0];
810 
811     return H_SUCCESS;
812 }
813 
814 static target_ulong h_page_init(PowerPCCPU *cpu, SpaprMachineState *spapr,
815                                 target_ulong opcode, target_ulong *args)
816 {
817     target_ulong flags = args[0];
818     hwaddr dst = args[1];
819     hwaddr src = args[2];
820     hwaddr len = TARGET_PAGE_SIZE;
821     uint8_t *pdst, *psrc;
822     target_long ret = H_SUCCESS;
823 
824     if (flags & ~(H_ICACHE_SYNCHRONIZE | H_ICACHE_INVALIDATE
825                   | H_COPY_PAGE | H_ZERO_PAGE)) {
826         qemu_log_mask(LOG_UNIMP, "h_page_init: Bad flags (" TARGET_FMT_lx "\n",
827                       flags);
828         return H_PARAMETER;
829     }
830 
831     /* Map-in destination */
832     if (!is_ram_address(spapr, dst) || (dst & ~TARGET_PAGE_MASK) != 0) {
833         return H_PARAMETER;
834     }
835     pdst = cpu_physical_memory_map(dst, &len, 1);
836     if (!pdst || len != TARGET_PAGE_SIZE) {
837         return H_PARAMETER;
838     }
839 
840     if (flags & H_COPY_PAGE) {
841         /* Map-in source, copy to destination, and unmap source again */
842         if (!is_ram_address(spapr, src) || (src & ~TARGET_PAGE_MASK) != 0) {
843             ret = H_PARAMETER;
844             goto unmap_out;
845         }
846         psrc = cpu_physical_memory_map(src, &len, 0);
847         if (!psrc || len != TARGET_PAGE_SIZE) {
848             ret = H_PARAMETER;
849             goto unmap_out;
850         }
851         memcpy(pdst, psrc, len);
852         cpu_physical_memory_unmap(psrc, len, 0, len);
853     } else if (flags & H_ZERO_PAGE) {
854         memset(pdst, 0, len);          /* Just clear the destination page */
855     }
856 
857     if (kvm_enabled() && (flags & H_ICACHE_SYNCHRONIZE) != 0) {
858         kvmppc_dcbst_range(cpu, pdst, len);
859     }
860     if (flags & (H_ICACHE_SYNCHRONIZE | H_ICACHE_INVALIDATE)) {
861         if (kvm_enabled()) {
862             kvmppc_icbi_range(cpu, pdst, len);
863         } else {
864             tb_flush(CPU(cpu));
865         }
866     }
867 
868 unmap_out:
869     cpu_physical_memory_unmap(pdst, TARGET_PAGE_SIZE, 1, len);
870     return ret;
871 }
872 
873 #define FLAGS_REGISTER_VPA         0x0000200000000000ULL
874 #define FLAGS_REGISTER_DTL         0x0000400000000000ULL
875 #define FLAGS_REGISTER_SLBSHADOW   0x0000600000000000ULL
876 #define FLAGS_DEREGISTER_VPA       0x0000a00000000000ULL
877 #define FLAGS_DEREGISTER_DTL       0x0000c00000000000ULL
878 #define FLAGS_DEREGISTER_SLBSHADOW 0x0000e00000000000ULL
879 
880 static target_ulong register_vpa(PowerPCCPU *cpu, target_ulong vpa)
881 {
882     CPUState *cs = CPU(cpu);
883     CPUPPCState *env = &cpu->env;
884     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
885     uint16_t size;
886     uint8_t tmp;
887 
888     if (vpa == 0) {
889         hcall_dprintf("Can't cope with registering a VPA at logical 0\n");
890         return H_HARDWARE;
891     }
892 
893     if (vpa % env->dcache_line_size) {
894         return H_PARAMETER;
895     }
896     /* FIXME: bounds check the address */
897 
898     size = lduw_be_phys(cs->as, vpa + 0x4);
899 
900     if (size < VPA_MIN_SIZE) {
901         return H_PARAMETER;
902     }
903 
904     /* VPA is not allowed to cross a page boundary */
905     if ((vpa / 4096) != ((vpa + size - 1) / 4096)) {
906         return H_PARAMETER;
907     }
908 
909     spapr_cpu->vpa_addr = vpa;
910 
911     tmp = ldub_phys(cs->as, spapr_cpu->vpa_addr + VPA_SHARED_PROC_OFFSET);
912     tmp |= VPA_SHARED_PROC_VAL;
913     stb_phys(cs->as, spapr_cpu->vpa_addr + VPA_SHARED_PROC_OFFSET, tmp);
914 
915     return H_SUCCESS;
916 }
917 
918 static target_ulong deregister_vpa(PowerPCCPU *cpu, target_ulong vpa)
919 {
920     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
921 
922     if (spapr_cpu->slb_shadow_addr) {
923         return H_RESOURCE;
924     }
925 
926     if (spapr_cpu->dtl_addr) {
927         return H_RESOURCE;
928     }
929 
930     spapr_cpu->vpa_addr = 0;
931     return H_SUCCESS;
932 }
933 
934 static target_ulong register_slb_shadow(PowerPCCPU *cpu, target_ulong addr)
935 {
936     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
937     uint32_t size;
938 
939     if (addr == 0) {
940         hcall_dprintf("Can't cope with SLB shadow at logical 0\n");
941         return H_HARDWARE;
942     }
943 
944     size = ldl_be_phys(CPU(cpu)->as, addr + 0x4);
945     if (size < 0x8) {
946         return H_PARAMETER;
947     }
948 
949     if ((addr / 4096) != ((addr + size - 1) / 4096)) {
950         return H_PARAMETER;
951     }
952 
953     if (!spapr_cpu->vpa_addr) {
954         return H_RESOURCE;
955     }
956 
957     spapr_cpu->slb_shadow_addr = addr;
958     spapr_cpu->slb_shadow_size = size;
959 
960     return H_SUCCESS;
961 }
962 
963 static target_ulong deregister_slb_shadow(PowerPCCPU *cpu, target_ulong addr)
964 {
965     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
966 
967     spapr_cpu->slb_shadow_addr = 0;
968     spapr_cpu->slb_shadow_size = 0;
969     return H_SUCCESS;
970 }
971 
972 static target_ulong register_dtl(PowerPCCPU *cpu, target_ulong addr)
973 {
974     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
975     uint32_t size;
976 
977     if (addr == 0) {
978         hcall_dprintf("Can't cope with DTL at logical 0\n");
979         return H_HARDWARE;
980     }
981 
982     size = ldl_be_phys(CPU(cpu)->as, addr + 0x4);
983 
984     if (size < 48) {
985         return H_PARAMETER;
986     }
987 
988     if (!spapr_cpu->vpa_addr) {
989         return H_RESOURCE;
990     }
991 
992     spapr_cpu->dtl_addr = addr;
993     spapr_cpu->dtl_size = size;
994 
995     return H_SUCCESS;
996 }
997 
998 static target_ulong deregister_dtl(PowerPCCPU *cpu, target_ulong addr)
999 {
1000     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
1001 
1002     spapr_cpu->dtl_addr = 0;
1003     spapr_cpu->dtl_size = 0;
1004 
1005     return H_SUCCESS;
1006 }
1007 
1008 static target_ulong h_register_vpa(PowerPCCPU *cpu, SpaprMachineState *spapr,
1009                                    target_ulong opcode, target_ulong *args)
1010 {
1011     target_ulong flags = args[0];
1012     target_ulong procno = args[1];
1013     target_ulong vpa = args[2];
1014     target_ulong ret = H_PARAMETER;
1015     PowerPCCPU *tcpu;
1016 
1017     tcpu = spapr_find_cpu(procno);
1018     if (!tcpu) {
1019         return H_PARAMETER;
1020     }
1021 
1022     switch (flags) {
1023     case FLAGS_REGISTER_VPA:
1024         ret = register_vpa(tcpu, vpa);
1025         break;
1026 
1027     case FLAGS_DEREGISTER_VPA:
1028         ret = deregister_vpa(tcpu, vpa);
1029         break;
1030 
1031     case FLAGS_REGISTER_SLBSHADOW:
1032         ret = register_slb_shadow(tcpu, vpa);
1033         break;
1034 
1035     case FLAGS_DEREGISTER_SLBSHADOW:
1036         ret = deregister_slb_shadow(tcpu, vpa);
1037         break;
1038 
1039     case FLAGS_REGISTER_DTL:
1040         ret = register_dtl(tcpu, vpa);
1041         break;
1042 
1043     case FLAGS_DEREGISTER_DTL:
1044         ret = deregister_dtl(tcpu, vpa);
1045         break;
1046     }
1047 
1048     return ret;
1049 }
1050 
1051 static target_ulong h_cede(PowerPCCPU *cpu, SpaprMachineState *spapr,
1052                            target_ulong opcode, target_ulong *args)
1053 {
1054     CPUPPCState *env = &cpu->env;
1055     CPUState *cs = CPU(cpu);
1056     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
1057 
1058     env->msr |= (1ULL << MSR_EE);
1059     hreg_compute_hflags(env);
1060 
1061     if (spapr_cpu->prod) {
1062         spapr_cpu->prod = false;
1063         return H_SUCCESS;
1064     }
1065 
1066     if (!cpu_has_work(cs)) {
1067         cs->halted = 1;
1068         cs->exception_index = EXCP_HLT;
1069         cs->exit_request = 1;
1070     }
1071 
1072     return H_SUCCESS;
1073 }
1074 
1075 /*
1076  * Confer to self, aka join. Cede could use the same pattern as well, if
1077  * EXCP_HLT can be changed to ECXP_HALTED.
1078  */
1079 static target_ulong h_confer_self(PowerPCCPU *cpu)
1080 {
1081     CPUState *cs = CPU(cpu);
1082     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
1083 
1084     if (spapr_cpu->prod) {
1085         spapr_cpu->prod = false;
1086         return H_SUCCESS;
1087     }
1088     cs->halted = 1;
1089     cs->exception_index = EXCP_HALTED;
1090     cs->exit_request = 1;
1091 
1092     return H_SUCCESS;
1093 }
1094 
1095 static target_ulong h_join(PowerPCCPU *cpu, SpaprMachineState *spapr,
1096                            target_ulong opcode, target_ulong *args)
1097 {
1098     CPUPPCState *env = &cpu->env;
1099     CPUState *cs;
1100     bool last_unjoined = true;
1101 
1102     if (env->msr & (1ULL << MSR_EE)) {
1103         return H_BAD_MODE;
1104     }
1105 
1106     /*
1107      * Must not join the last CPU running. Interestingly, no such restriction
1108      * for H_CONFER-to-self, but that is probably not intended to be used
1109      * when H_JOIN is available.
1110      */
1111     CPU_FOREACH(cs) {
1112         PowerPCCPU *c = POWERPC_CPU(cs);
1113         CPUPPCState *e = &c->env;
1114         if (c == cpu) {
1115             continue;
1116         }
1117 
1118         /* Don't have a way to indicate joined, so use halted && MSR[EE]=0 */
1119         if (!cs->halted || (e->msr & (1ULL << MSR_EE))) {
1120             last_unjoined = false;
1121             break;
1122         }
1123     }
1124     if (last_unjoined) {
1125         return H_CONTINUE;
1126     }
1127 
1128     return h_confer_self(cpu);
1129 }
1130 
1131 static target_ulong h_confer(PowerPCCPU *cpu, SpaprMachineState *spapr,
1132                            target_ulong opcode, target_ulong *args)
1133 {
1134     target_long target = args[0];
1135     uint32_t dispatch = args[1];
1136     CPUState *cs = CPU(cpu);
1137     SpaprCpuState *spapr_cpu;
1138 
1139     /*
1140      * -1 means confer to all other CPUs without dispatch counter check,
1141      *  otherwise it's a targeted confer.
1142      */
1143     if (target != -1) {
1144         PowerPCCPU *target_cpu = spapr_find_cpu(target);
1145         uint32_t target_dispatch;
1146 
1147         if (!target_cpu) {
1148             return H_PARAMETER;
1149         }
1150 
1151         /*
1152          * target == self is a special case, we wait until prodded, without
1153          * dispatch counter check.
1154          */
1155         if (cpu == target_cpu) {
1156             return h_confer_self(cpu);
1157         }
1158 
1159         spapr_cpu = spapr_cpu_state(target_cpu);
1160         if (!spapr_cpu->vpa_addr || ((dispatch & 1) == 0)) {
1161             return H_SUCCESS;
1162         }
1163 
1164         target_dispatch = ldl_be_phys(cs->as,
1165                                   spapr_cpu->vpa_addr + VPA_DISPATCH_COUNTER);
1166         if (target_dispatch != dispatch) {
1167             return H_SUCCESS;
1168         }
1169 
1170         /*
1171          * The targeted confer does not do anything special beyond yielding
1172          * the current vCPU, but even this should be better than nothing.
1173          * At least for single-threaded tcg, it gives the target a chance to
1174          * run before we run again. Multi-threaded tcg does not really do
1175          * anything with EXCP_YIELD yet.
1176          */
1177     }
1178 
1179     cs->exception_index = EXCP_YIELD;
1180     cs->exit_request = 1;
1181     cpu_loop_exit(cs);
1182 
1183     return H_SUCCESS;
1184 }
1185 
1186 static target_ulong h_prod(PowerPCCPU *cpu, SpaprMachineState *spapr,
1187                            target_ulong opcode, target_ulong *args)
1188 {
1189     target_long target = args[0];
1190     PowerPCCPU *tcpu;
1191     CPUState *cs;
1192     SpaprCpuState *spapr_cpu;
1193 
1194     tcpu = spapr_find_cpu(target);
1195     cs = CPU(tcpu);
1196     if (!cs) {
1197         return H_PARAMETER;
1198     }
1199 
1200     spapr_cpu = spapr_cpu_state(tcpu);
1201     spapr_cpu->prod = true;
1202     cs->halted = 0;
1203     qemu_cpu_kick(cs);
1204 
1205     return H_SUCCESS;
1206 }
1207 
1208 static target_ulong h_rtas(PowerPCCPU *cpu, SpaprMachineState *spapr,
1209                            target_ulong opcode, target_ulong *args)
1210 {
1211     target_ulong rtas_r3 = args[0];
1212     uint32_t token = rtas_ld(rtas_r3, 0);
1213     uint32_t nargs = rtas_ld(rtas_r3, 1);
1214     uint32_t nret = rtas_ld(rtas_r3, 2);
1215 
1216     return spapr_rtas_call(cpu, spapr, token, nargs, rtas_r3 + 12,
1217                            nret, rtas_r3 + 12 + 4*nargs);
1218 }
1219 
1220 static target_ulong h_logical_load(PowerPCCPU *cpu, SpaprMachineState *spapr,
1221                                    target_ulong opcode, target_ulong *args)
1222 {
1223     CPUState *cs = CPU(cpu);
1224     target_ulong size = args[0];
1225     target_ulong addr = args[1];
1226 
1227     switch (size) {
1228     case 1:
1229         args[0] = ldub_phys(cs->as, addr);
1230         return H_SUCCESS;
1231     case 2:
1232         args[0] = lduw_phys(cs->as, addr);
1233         return H_SUCCESS;
1234     case 4:
1235         args[0] = ldl_phys(cs->as, addr);
1236         return H_SUCCESS;
1237     case 8:
1238         args[0] = ldq_phys(cs->as, addr);
1239         return H_SUCCESS;
1240     }
1241     return H_PARAMETER;
1242 }
1243 
1244 static target_ulong h_logical_store(PowerPCCPU *cpu, SpaprMachineState *spapr,
1245                                     target_ulong opcode, target_ulong *args)
1246 {
1247     CPUState *cs = CPU(cpu);
1248 
1249     target_ulong size = args[0];
1250     target_ulong addr = args[1];
1251     target_ulong val  = args[2];
1252 
1253     switch (size) {
1254     case 1:
1255         stb_phys(cs->as, addr, val);
1256         return H_SUCCESS;
1257     case 2:
1258         stw_phys(cs->as, addr, val);
1259         return H_SUCCESS;
1260     case 4:
1261         stl_phys(cs->as, addr, val);
1262         return H_SUCCESS;
1263     case 8:
1264         stq_phys(cs->as, addr, val);
1265         return H_SUCCESS;
1266     }
1267     return H_PARAMETER;
1268 }
1269 
1270 static target_ulong h_logical_memop(PowerPCCPU *cpu, SpaprMachineState *spapr,
1271                                     target_ulong opcode, target_ulong *args)
1272 {
1273     CPUState *cs = CPU(cpu);
1274 
1275     target_ulong dst   = args[0]; /* Destination address */
1276     target_ulong src   = args[1]; /* Source address */
1277     target_ulong esize = args[2]; /* Element size (0=1,1=2,2=4,3=8) */
1278     target_ulong count = args[3]; /* Element count */
1279     target_ulong op    = args[4]; /* 0 = copy, 1 = invert */
1280     uint64_t tmp;
1281     unsigned int mask = (1 << esize) - 1;
1282     int step = 1 << esize;
1283 
1284     if (count > 0x80000000) {
1285         return H_PARAMETER;
1286     }
1287 
1288     if ((dst & mask) || (src & mask) || (op > 1)) {
1289         return H_PARAMETER;
1290     }
1291 
1292     if (dst >= src && dst < (src + (count << esize))) {
1293             dst = dst + ((count - 1) << esize);
1294             src = src + ((count - 1) << esize);
1295             step = -step;
1296     }
1297 
1298     while (count--) {
1299         switch (esize) {
1300         case 0:
1301             tmp = ldub_phys(cs->as, src);
1302             break;
1303         case 1:
1304             tmp = lduw_phys(cs->as, src);
1305             break;
1306         case 2:
1307             tmp = ldl_phys(cs->as, src);
1308             break;
1309         case 3:
1310             tmp = ldq_phys(cs->as, src);
1311             break;
1312         default:
1313             return H_PARAMETER;
1314         }
1315         if (op == 1) {
1316             tmp = ~tmp;
1317         }
1318         switch (esize) {
1319         case 0:
1320             stb_phys(cs->as, dst, tmp);
1321             break;
1322         case 1:
1323             stw_phys(cs->as, dst, tmp);
1324             break;
1325         case 2:
1326             stl_phys(cs->as, dst, tmp);
1327             break;
1328         case 3:
1329             stq_phys(cs->as, dst, tmp);
1330             break;
1331         }
1332         dst = dst + step;
1333         src = src + step;
1334     }
1335 
1336     return H_SUCCESS;
1337 }
1338 
1339 static target_ulong h_logical_icbi(PowerPCCPU *cpu, SpaprMachineState *spapr,
1340                                    target_ulong opcode, target_ulong *args)
1341 {
1342     /* Nothing to do on emulation, KVM will trap this in the kernel */
1343     return H_SUCCESS;
1344 }
1345 
1346 static target_ulong h_logical_dcbf(PowerPCCPU *cpu, SpaprMachineState *spapr,
1347                                    target_ulong opcode, target_ulong *args)
1348 {
1349     /* Nothing to do on emulation, KVM will trap this in the kernel */
1350     return H_SUCCESS;
1351 }
1352 
1353 static target_ulong h_set_mode_resource_le(PowerPCCPU *cpu,
1354                                            target_ulong mflags,
1355                                            target_ulong value1,
1356                                            target_ulong value2)
1357 {
1358     if (value1) {
1359         return H_P3;
1360     }
1361     if (value2) {
1362         return H_P4;
1363     }
1364 
1365     switch (mflags) {
1366     case H_SET_MODE_ENDIAN_BIG:
1367         spapr_set_all_lpcrs(0, LPCR_ILE);
1368         spapr_pci_switch_vga(true);
1369         return H_SUCCESS;
1370 
1371     case H_SET_MODE_ENDIAN_LITTLE:
1372         spapr_set_all_lpcrs(LPCR_ILE, LPCR_ILE);
1373         spapr_pci_switch_vga(false);
1374         return H_SUCCESS;
1375     }
1376 
1377     return H_UNSUPPORTED_FLAG;
1378 }
1379 
1380 static target_ulong h_set_mode_resource_addr_trans_mode(PowerPCCPU *cpu,
1381                                                         target_ulong mflags,
1382                                                         target_ulong value1,
1383                                                         target_ulong value2)
1384 {
1385     PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cpu);
1386 
1387     if (!(pcc->insns_flags2 & PPC2_ISA207S)) {
1388         return H_P2;
1389     }
1390     if (value1) {
1391         return H_P3;
1392     }
1393     if (value2) {
1394         return H_P4;
1395     }
1396 
1397     if (mflags == AIL_RESERVED) {
1398         return H_UNSUPPORTED_FLAG;
1399     }
1400 
1401     spapr_set_all_lpcrs(mflags << LPCR_AIL_SHIFT, LPCR_AIL);
1402 
1403     return H_SUCCESS;
1404 }
1405 
1406 static target_ulong h_set_mode(PowerPCCPU *cpu, SpaprMachineState *spapr,
1407                                target_ulong opcode, target_ulong *args)
1408 {
1409     target_ulong resource = args[1];
1410     target_ulong ret = H_P2;
1411 
1412     switch (resource) {
1413     case H_SET_MODE_RESOURCE_LE:
1414         ret = h_set_mode_resource_le(cpu, args[0], args[2], args[3]);
1415         break;
1416     case H_SET_MODE_RESOURCE_ADDR_TRANS_MODE:
1417         ret = h_set_mode_resource_addr_trans_mode(cpu, args[0],
1418                                                   args[2], args[3]);
1419         break;
1420     }
1421 
1422     return ret;
1423 }
1424 
1425 static target_ulong h_clean_slb(PowerPCCPU *cpu, SpaprMachineState *spapr,
1426                                 target_ulong opcode, target_ulong *args)
1427 {
1428     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x"TARGET_FMT_lx"%s\n",
1429                   opcode, " (H_CLEAN_SLB)");
1430     return H_FUNCTION;
1431 }
1432 
1433 static target_ulong h_invalidate_pid(PowerPCCPU *cpu, SpaprMachineState *spapr,
1434                                      target_ulong opcode, target_ulong *args)
1435 {
1436     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x"TARGET_FMT_lx"%s\n",
1437                   opcode, " (H_INVALIDATE_PID)");
1438     return H_FUNCTION;
1439 }
1440 
1441 static void spapr_check_setup_free_hpt(SpaprMachineState *spapr,
1442                                        uint64_t patbe_old, uint64_t patbe_new)
1443 {
1444     /*
1445      * We have 4 Options:
1446      * HASH->HASH || RADIX->RADIX || NOTHING->RADIX : Do Nothing
1447      * HASH->RADIX                                  : Free HPT
1448      * RADIX->HASH                                  : Allocate HPT
1449      * NOTHING->HASH                                : Allocate HPT
1450      * Note: NOTHING implies the case where we said the guest could choose
1451      *       later and so assumed radix and now it's called H_REG_PROC_TBL
1452      */
1453 
1454     if ((patbe_old & PATE1_GR) == (patbe_new & PATE1_GR)) {
1455         /* We assume RADIX, so this catches all the "Do Nothing" cases */
1456     } else if (!(patbe_old & PATE1_GR)) {
1457         /* HASH->RADIX : Free HPT */
1458         spapr_free_hpt(spapr);
1459     } else if (!(patbe_new & PATE1_GR)) {
1460         /* RADIX->HASH || NOTHING->HASH : Allocate HPT */
1461         spapr_setup_hpt_and_vrma(spapr);
1462     }
1463     return;
1464 }
1465 
1466 #define FLAGS_MASK              0x01FULL
1467 #define FLAG_MODIFY             0x10
1468 #define FLAG_REGISTER           0x08
1469 #define FLAG_RADIX              0x04
1470 #define FLAG_HASH_PROC_TBL      0x02
1471 #define FLAG_GTSE               0x01
1472 
1473 static target_ulong h_register_process_table(PowerPCCPU *cpu,
1474                                              SpaprMachineState *spapr,
1475                                              target_ulong opcode,
1476                                              target_ulong *args)
1477 {
1478     target_ulong flags = args[0];
1479     target_ulong proc_tbl = args[1];
1480     target_ulong page_size = args[2];
1481     target_ulong table_size = args[3];
1482     target_ulong update_lpcr = 0;
1483     uint64_t cproc;
1484 
1485     if (flags & ~FLAGS_MASK) { /* Check no reserved bits are set */
1486         return H_PARAMETER;
1487     }
1488     if (flags & FLAG_MODIFY) {
1489         if (flags & FLAG_REGISTER) {
1490             if (flags & FLAG_RADIX) { /* Register new RADIX process table */
1491                 if (proc_tbl & 0xfff || proc_tbl >> 60) {
1492                     return H_P2;
1493                 } else if (page_size) {
1494                     return H_P3;
1495                 } else if (table_size > 24) {
1496                     return H_P4;
1497                 }
1498                 cproc = PATE1_GR | proc_tbl | table_size;
1499             } else { /* Register new HPT process table */
1500                 if (flags & FLAG_HASH_PROC_TBL) { /* Hash with Segment Tables */
1501                     /* TODO - Not Supported */
1502                     /* Technically caused by flag bits => H_PARAMETER */
1503                     return H_PARAMETER;
1504                 } else { /* Hash with SLB */
1505                     if (proc_tbl >> 38) {
1506                         return H_P2;
1507                     } else if (page_size & ~0x7) {
1508                         return H_P3;
1509                     } else if (table_size > 24) {
1510                         return H_P4;
1511                     }
1512                 }
1513                 cproc = (proc_tbl << 25) | page_size << 5 | table_size;
1514             }
1515 
1516         } else { /* Deregister current process table */
1517             /*
1518              * Set to benign value: (current GR) | 0. This allows
1519              * deregistration in KVM to succeed even if the radix bit
1520              * in flags doesn't match the radix bit in the old PATE.
1521              */
1522             cproc = spapr->patb_entry & PATE1_GR;
1523         }
1524     } else { /* Maintain current registration */
1525         if (!(flags & FLAG_RADIX) != !(spapr->patb_entry & PATE1_GR)) {
1526             /* Technically caused by flag bits => H_PARAMETER */
1527             return H_PARAMETER; /* Existing Process Table Mismatch */
1528         }
1529         cproc = spapr->patb_entry;
1530     }
1531 
1532     /* Check if we need to setup OR free the hpt */
1533     spapr_check_setup_free_hpt(spapr, spapr->patb_entry, cproc);
1534 
1535     spapr->patb_entry = cproc; /* Save new process table */
1536 
1537     /* Update the UPRT, HR and GTSE bits in the LPCR for all cpus */
1538     if (flags & FLAG_RADIX)     /* Radix must use process tables, also set HR */
1539         update_lpcr |= (LPCR_UPRT | LPCR_HR);
1540     else if (flags & FLAG_HASH_PROC_TBL) /* Hash with process tables */
1541         update_lpcr |= LPCR_UPRT;
1542     if (flags & FLAG_GTSE)      /* Guest translation shootdown enable */
1543         update_lpcr |= LPCR_GTSE;
1544 
1545     spapr_set_all_lpcrs(update_lpcr, LPCR_UPRT | LPCR_HR | LPCR_GTSE);
1546 
1547     if (kvm_enabled()) {
1548         return kvmppc_configure_v3_mmu(cpu, flags & FLAG_RADIX,
1549                                        flags & FLAG_GTSE, cproc);
1550     }
1551     return H_SUCCESS;
1552 }
1553 
1554 #define H_SIGNAL_SYS_RESET_ALL         -1
1555 #define H_SIGNAL_SYS_RESET_ALLBUTSELF  -2
1556 
1557 static target_ulong h_signal_sys_reset(PowerPCCPU *cpu,
1558                                        SpaprMachineState *spapr,
1559                                        target_ulong opcode, target_ulong *args)
1560 {
1561     target_long target = args[0];
1562     CPUState *cs;
1563 
1564     if (target < 0) {
1565         /* Broadcast */
1566         if (target < H_SIGNAL_SYS_RESET_ALLBUTSELF) {
1567             return H_PARAMETER;
1568         }
1569 
1570         CPU_FOREACH(cs) {
1571             PowerPCCPU *c = POWERPC_CPU(cs);
1572 
1573             if (target == H_SIGNAL_SYS_RESET_ALLBUTSELF) {
1574                 if (c == cpu) {
1575                     continue;
1576                 }
1577             }
1578             run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL);
1579         }
1580         return H_SUCCESS;
1581 
1582     } else {
1583         /* Unicast */
1584         cs = CPU(spapr_find_cpu(target));
1585         if (cs) {
1586             run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL);
1587             return H_SUCCESS;
1588         }
1589         return H_PARAMETER;
1590     }
1591 }
1592 
1593 static uint32_t cas_check_pvr(SpaprMachineState *spapr, PowerPCCPU *cpu,
1594                               target_ulong *addr, bool *raw_mode_supported,
1595                               Error **errp)
1596 {
1597     bool explicit_match = false; /* Matched the CPU's real PVR */
1598     uint32_t max_compat = spapr->max_compat_pvr;
1599     uint32_t best_compat = 0;
1600     int i;
1601 
1602     /*
1603      * We scan the supplied table of PVRs looking for two things
1604      *   1. Is our real CPU PVR in the list?
1605      *   2. What's the "best" listed logical PVR
1606      */
1607     for (i = 0; i < 512; ++i) {
1608         uint32_t pvr, pvr_mask;
1609 
1610         pvr_mask = ldl_be_phys(&address_space_memory, *addr);
1611         pvr = ldl_be_phys(&address_space_memory, *addr + 4);
1612         *addr += 8;
1613 
1614         if (~pvr_mask & pvr) {
1615             break; /* Terminator record */
1616         }
1617 
1618         if ((cpu->env.spr[SPR_PVR] & pvr_mask) == (pvr & pvr_mask)) {
1619             explicit_match = true;
1620         } else {
1621             if (ppc_check_compat(cpu, pvr, best_compat, max_compat)) {
1622                 best_compat = pvr;
1623             }
1624         }
1625     }
1626 
1627     if ((best_compat == 0) && (!explicit_match || max_compat)) {
1628         /* We couldn't find a suitable compatibility mode, and either
1629          * the guest doesn't support "raw" mode for this CPU, or raw
1630          * mode is disabled because a maximum compat mode is set */
1631         error_setg(errp, "Couldn't negotiate a suitable PVR during CAS");
1632         return 0;
1633     }
1634 
1635     *raw_mode_supported = explicit_match;
1636 
1637     /* Parsing finished */
1638     trace_spapr_cas_pvr(cpu->compat_pvr, explicit_match, best_compat);
1639 
1640     return best_compat;
1641 }
1642 
1643 static bool spapr_transient_dev_before_cas(void)
1644 {
1645     Object *drc_container;
1646     ObjectProperty *prop;
1647     ObjectPropertyIterator iter;
1648 
1649     drc_container = container_get(object_get_root(), "/dr-connector");
1650     object_property_iter_init(&iter, drc_container);
1651     while ((prop = object_property_iter_next(&iter))) {
1652         SpaprDrc *drc;
1653 
1654         if (!strstart(prop->type, "link<", NULL)) {
1655             continue;
1656         }
1657         drc = SPAPR_DR_CONNECTOR(object_property_get_link(drc_container,
1658                                                           prop->name, NULL));
1659 
1660         if (spapr_drc_transient(drc)) {
1661             return true;
1662         }
1663     }
1664     return false;
1665 }
1666 
1667 static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
1668                                                   SpaprMachineState *spapr,
1669                                                   target_ulong opcode,
1670                                                   target_ulong *args)
1671 {
1672     /* Working address in data buffer */
1673     target_ulong addr = ppc64_phys_to_real(args[0]);
1674     target_ulong fdt_buf = args[1];
1675     target_ulong fdt_bufsize = args[2];
1676     target_ulong ov_table;
1677     uint32_t cas_pvr;
1678     SpaprOptionVector *ov1_guest, *ov5_guest, *ov5_cas_old;
1679     bool guest_radix;
1680     Error *local_err = NULL;
1681     bool raw_mode_supported = false;
1682     bool guest_xive;
1683     CPUState *cs;
1684 
1685     /* CAS is supposed to be called early when only the boot vCPU is active. */
1686     CPU_FOREACH(cs) {
1687         if (cs == CPU(cpu)) {
1688             continue;
1689         }
1690         if (!cs->halted) {
1691             warn_report("guest has multiple active vCPUs at CAS, which is not allowed");
1692             return H_MULTI_THREADS_ACTIVE;
1693         }
1694     }
1695 
1696     cas_pvr = cas_check_pvr(spapr, cpu, &addr, &raw_mode_supported, &local_err);
1697     if (local_err) {
1698         error_report_err(local_err);
1699         return H_HARDWARE;
1700     }
1701 
1702     /* Update CPUs */
1703     if (cpu->compat_pvr != cas_pvr) {
1704         ppc_set_compat_all(cas_pvr, &local_err);
1705         if (local_err) {
1706             /* We fail to set compat mode (likely because running with KVM PR),
1707              * but maybe we can fallback to raw mode if the guest supports it.
1708              */
1709             if (!raw_mode_supported) {
1710                 error_report_err(local_err);
1711                 return H_HARDWARE;
1712             }
1713             error_free(local_err);
1714             local_err = NULL;
1715         }
1716     }
1717 
1718     /* For the future use: here @ov_table points to the first option vector */
1719     ov_table = addr;
1720 
1721     ov1_guest = spapr_ovec_parse_vector(ov_table, 1);
1722     if (!ov1_guest) {
1723         warn_report("guest didn't provide option vector 1");
1724         return H_PARAMETER;
1725     }
1726     ov5_guest = spapr_ovec_parse_vector(ov_table, 5);
1727     if (!ov5_guest) {
1728         warn_report("guest didn't provide option vector 5");
1729         return H_PARAMETER;
1730     }
1731     if (spapr_ovec_test(ov5_guest, OV5_MMU_BOTH)) {
1732         error_report("guest requested hash and radix MMU, which is invalid.");
1733         exit(EXIT_FAILURE);
1734     }
1735     if (spapr_ovec_test(ov5_guest, OV5_XIVE_BOTH)) {
1736         error_report("guest requested an invalid interrupt mode");
1737         exit(EXIT_FAILURE);
1738     }
1739 
1740     /* The radix/hash bit in byte 24 requires special handling: */
1741     guest_radix = spapr_ovec_test(ov5_guest, OV5_MMU_RADIX_300);
1742     spapr_ovec_clear(ov5_guest, OV5_MMU_RADIX_300);
1743 
1744     guest_xive = spapr_ovec_test(ov5_guest, OV5_XIVE_EXPLOIT);
1745 
1746     /*
1747      * HPT resizing is a bit of a special case, because when enabled
1748      * we assume an HPT guest will support it until it says it
1749      * doesn't, instead of assuming it won't support it until it says
1750      * it does.  Strictly speaking that approach could break for
1751      * guests which don't make a CAS call, but those are so old we
1752      * don't care about them.  Without that assumption we'd have to
1753      * make at least a temporary allocation of an HPT sized for max
1754      * memory, which could be impossibly difficult under KVM HV if
1755      * maxram is large.
1756      */
1757     if (!guest_radix && !spapr_ovec_test(ov5_guest, OV5_HPT_RESIZE)) {
1758         int maxshift = spapr_hpt_shift_for_ramsize(MACHINE(spapr)->maxram_size);
1759 
1760         if (spapr->resize_hpt == SPAPR_RESIZE_HPT_REQUIRED) {
1761             error_report(
1762                 "h_client_architecture_support: Guest doesn't support HPT resizing, but resize-hpt=required");
1763             exit(1);
1764         }
1765 
1766         if (spapr->htab_shift < maxshift) {
1767             /* Guest doesn't know about HPT resizing, so we
1768              * pre-emptively resize for the maximum permitted RAM.  At
1769              * the point this is called, nothing should have been
1770              * entered into the existing HPT */
1771             spapr_reallocate_hpt(spapr, maxshift, &error_fatal);
1772             push_sregs_to_kvm_pr(spapr);
1773         }
1774     }
1775 
1776     /* NOTE: there are actually a number of ov5 bits where input from the
1777      * guest is always zero, and the platform/QEMU enables them independently
1778      * of guest input. To model these properly we'd want some sort of mask,
1779      * but since they only currently apply to memory migration as defined
1780      * by LoPAPR 1.1, 14.5.4.8, which QEMU doesn't implement, we don't need
1781      * to worry about this for now.
1782      */
1783     ov5_cas_old = spapr_ovec_clone(spapr->ov5_cas);
1784 
1785     /* also clear the radix/hash bit from the current ov5_cas bits to
1786      * be in sync with the newly ov5 bits. Else the radix bit will be
1787      * seen as being removed and this will generate a reset loop
1788      */
1789     spapr_ovec_clear(ov5_cas_old, OV5_MMU_RADIX_300);
1790 
1791     /* full range of negotiated ov5 capabilities */
1792     spapr_ovec_intersect(spapr->ov5_cas, spapr->ov5, ov5_guest);
1793     spapr_ovec_cleanup(ov5_guest);
1794     /* capabilities that have been added since CAS-generated guest reset.
1795      * if capabilities have since been removed, generate another reset
1796      */
1797     spapr->cas_reboot = !spapr_ovec_subset(ov5_cas_old, spapr->ov5_cas);
1798     spapr_ovec_cleanup(ov5_cas_old);
1799     /* Now that processing is finished, set the radix/hash bit for the
1800      * guest if it requested a valid mode; otherwise terminate the boot. */
1801     if (guest_radix) {
1802         if (kvm_enabled() && !kvmppc_has_cap_mmu_radix()) {
1803             error_report("Guest requested unavailable MMU mode (radix).");
1804             exit(EXIT_FAILURE);
1805         }
1806         spapr_ovec_set(spapr->ov5_cas, OV5_MMU_RADIX_300);
1807     } else {
1808         if (kvm_enabled() && kvmppc_has_cap_mmu_radix()
1809             && !kvmppc_has_cap_mmu_hash_v3()) {
1810             error_report("Guest requested unavailable MMU mode (hash).");
1811             exit(EXIT_FAILURE);
1812         }
1813     }
1814     spapr->cas_pre_isa3_guest = !spapr_ovec_test(ov1_guest, OV1_PPC_3_00);
1815     spapr_ovec_cleanup(ov1_guest);
1816 
1817     /*
1818      * Ensure the guest asks for an interrupt mode we support;
1819      * otherwise terminate the boot.
1820      */
1821     if (guest_xive) {
1822         if (!spapr->irq->xive) {
1823             error_report(
1824 "Guest requested unavailable interrupt mode (XIVE), try the ic-mode=xive or ic-mode=dual machine property");
1825             exit(EXIT_FAILURE);
1826         }
1827     } else {
1828         if (!spapr->irq->xics) {
1829             error_report(
1830 "Guest requested unavailable interrupt mode (XICS), either don't set the ic-mode machine property or try ic-mode=xics or ic-mode=dual");
1831             exit(EXIT_FAILURE);
1832         }
1833     }
1834 
1835     spapr_irq_update_active_intc(spapr);
1836 
1837     if (spapr_transient_dev_before_cas()) {
1838         spapr->cas_reboot = true;
1839     }
1840 
1841     if (!spapr->cas_reboot) {
1842         void *fdt;
1843         SpaprDeviceTreeUpdateHeader hdr = { .version_id = 1 };
1844 
1845         /* If spapr_machine_reset() did not set up a HPT but one is necessary
1846          * (because the guest isn't going to use radix) then set it up here. */
1847         if ((spapr->patb_entry & PATE1_GR) && !guest_radix) {
1848             /* legacy hash or new hash: */
1849             spapr_setup_hpt_and_vrma(spapr);
1850         }
1851 
1852         if (fdt_bufsize < sizeof(hdr)) {
1853             error_report("SLOF provided insufficient CAS buffer "
1854                          TARGET_FMT_lu " (min: %zu)", fdt_bufsize, sizeof(hdr));
1855             exit(EXIT_FAILURE);
1856         }
1857 
1858         fdt_bufsize -= sizeof(hdr);
1859 
1860         fdt = spapr_build_fdt(spapr, false, fdt_bufsize);
1861         _FDT((fdt_pack(fdt)));
1862 
1863         cpu_physical_memory_write(fdt_buf, &hdr, sizeof(hdr));
1864         cpu_physical_memory_write(fdt_buf + sizeof(hdr), fdt,
1865                                   fdt_totalsize(fdt));
1866         trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr));
1867 
1868         g_free(spapr->fdt_blob);
1869         spapr->fdt_size = fdt_totalsize(fdt);
1870         spapr->fdt_initial_size = spapr->fdt_size;
1871         spapr->fdt_blob = fdt;
1872     }
1873 
1874     if (spapr->cas_reboot) {
1875         qemu_system_reset_request(SHUTDOWN_CAUSE_SUBSYSTEM_RESET);
1876     }
1877 
1878     return H_SUCCESS;
1879 }
1880 
1881 static target_ulong h_home_node_associativity(PowerPCCPU *cpu,
1882                                               SpaprMachineState *spapr,
1883                                               target_ulong opcode,
1884                                               target_ulong *args)
1885 {
1886     target_ulong flags = args[0];
1887     target_ulong procno = args[1];
1888     PowerPCCPU *tcpu;
1889     int idx;
1890 
1891     /* only support procno from H_REGISTER_VPA */
1892     if (flags != 0x1) {
1893         return H_FUNCTION;
1894     }
1895 
1896     tcpu = spapr_find_cpu(procno);
1897     if (tcpu == NULL) {
1898         return H_P2;
1899     }
1900 
1901     /* sequence is the same as in the "ibm,associativity" property */
1902 
1903     idx = 0;
1904 #define ASSOCIATIVITY(a, b) (((uint64_t)(a) << 32) | \
1905                              ((uint64_t)(b) & 0xffffffff))
1906     args[idx++] = ASSOCIATIVITY(0, 0);
1907     args[idx++] = ASSOCIATIVITY(0, tcpu->node_id);
1908     args[idx++] = ASSOCIATIVITY(procno, -1);
1909     for ( ; idx < 6; idx++) {
1910         args[idx] = -1;
1911     }
1912 #undef ASSOCIATIVITY
1913 
1914     return H_SUCCESS;
1915 }
1916 
1917 static target_ulong h_get_cpu_characteristics(PowerPCCPU *cpu,
1918                                               SpaprMachineState *spapr,
1919                                               target_ulong opcode,
1920                                               target_ulong *args)
1921 {
1922     uint64_t characteristics = H_CPU_CHAR_HON_BRANCH_HINTS &
1923                                ~H_CPU_CHAR_THR_RECONF_TRIG;
1924     uint64_t behaviour = H_CPU_BEHAV_FAVOUR_SECURITY;
1925     uint8_t safe_cache = spapr_get_cap(spapr, SPAPR_CAP_CFPC);
1926     uint8_t safe_bounds_check = spapr_get_cap(spapr, SPAPR_CAP_SBBC);
1927     uint8_t safe_indirect_branch = spapr_get_cap(spapr, SPAPR_CAP_IBS);
1928     uint8_t count_cache_flush_assist = spapr_get_cap(spapr,
1929                                                      SPAPR_CAP_CCF_ASSIST);
1930 
1931     switch (safe_cache) {
1932     case SPAPR_CAP_WORKAROUND:
1933         characteristics |= H_CPU_CHAR_L1D_FLUSH_ORI30;
1934         characteristics |= H_CPU_CHAR_L1D_FLUSH_TRIG2;
1935         characteristics |= H_CPU_CHAR_L1D_THREAD_PRIV;
1936         behaviour |= H_CPU_BEHAV_L1D_FLUSH_PR;
1937         break;
1938     case SPAPR_CAP_FIXED:
1939         break;
1940     default: /* broken */
1941         assert(safe_cache == SPAPR_CAP_BROKEN);
1942         behaviour |= H_CPU_BEHAV_L1D_FLUSH_PR;
1943         break;
1944     }
1945 
1946     switch (safe_bounds_check) {
1947     case SPAPR_CAP_WORKAROUND:
1948         characteristics |= H_CPU_CHAR_SPEC_BAR_ORI31;
1949         behaviour |= H_CPU_BEHAV_BNDS_CHK_SPEC_BAR;
1950         break;
1951     case SPAPR_CAP_FIXED:
1952         break;
1953     default: /* broken */
1954         assert(safe_bounds_check == SPAPR_CAP_BROKEN);
1955         behaviour |= H_CPU_BEHAV_BNDS_CHK_SPEC_BAR;
1956         break;
1957     }
1958 
1959     switch (safe_indirect_branch) {
1960     case SPAPR_CAP_FIXED_NA:
1961         break;
1962     case SPAPR_CAP_FIXED_CCD:
1963         characteristics |= H_CPU_CHAR_CACHE_COUNT_DIS;
1964         break;
1965     case SPAPR_CAP_FIXED_IBS:
1966         characteristics |= H_CPU_CHAR_BCCTRL_SERIALISED;
1967         break;
1968     case SPAPR_CAP_WORKAROUND:
1969         behaviour |= H_CPU_BEHAV_FLUSH_COUNT_CACHE;
1970         if (count_cache_flush_assist) {
1971             characteristics |= H_CPU_CHAR_BCCTR_FLUSH_ASSIST;
1972         }
1973         break;
1974     default: /* broken */
1975         assert(safe_indirect_branch == SPAPR_CAP_BROKEN);
1976         break;
1977     }
1978 
1979     args[0] = characteristics;
1980     args[1] = behaviour;
1981     return H_SUCCESS;
1982 }
1983 
1984 static target_ulong h_update_dt(PowerPCCPU *cpu, SpaprMachineState *spapr,
1985                                 target_ulong opcode, target_ulong *args)
1986 {
1987     target_ulong dt = ppc64_phys_to_real(args[0]);
1988     struct fdt_header hdr = { 0 };
1989     unsigned cb;
1990     SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
1991     void *fdt;
1992 
1993     cpu_physical_memory_read(dt, &hdr, sizeof(hdr));
1994     cb = fdt32_to_cpu(hdr.totalsize);
1995 
1996     if (!smc->update_dt_enabled) {
1997         return H_SUCCESS;
1998     }
1999 
2000     /* Check that the fdt did not grow out of proportion */
2001     if (cb > spapr->fdt_initial_size * 2) {
2002         trace_spapr_update_dt_failed_size(spapr->fdt_initial_size, cb,
2003                                           fdt32_to_cpu(hdr.magic));
2004         return H_PARAMETER;
2005     }
2006 
2007     fdt = g_malloc0(cb);
2008     cpu_physical_memory_read(dt, fdt, cb);
2009 
2010     /* Check the fdt consistency */
2011     if (fdt_check_full(fdt, cb)) {
2012         trace_spapr_update_dt_failed_check(spapr->fdt_initial_size, cb,
2013                                            fdt32_to_cpu(hdr.magic));
2014         return H_PARAMETER;
2015     }
2016 
2017     g_free(spapr->fdt_blob);
2018     spapr->fdt_size = cb;
2019     spapr->fdt_blob = fdt;
2020     trace_spapr_update_dt(cb);
2021 
2022     return H_SUCCESS;
2023 }
2024 
2025 static spapr_hcall_fn papr_hypercall_table[(MAX_HCALL_OPCODE / 4) + 1];
2026 static spapr_hcall_fn kvmppc_hypercall_table[KVMPPC_HCALL_MAX - KVMPPC_HCALL_BASE + 1];
2027 static spapr_hcall_fn svm_hypercall_table[(SVM_HCALL_MAX - SVM_HCALL_BASE) / 4 + 1];
2028 
2029 void spapr_register_hypercall(target_ulong opcode, spapr_hcall_fn fn)
2030 {
2031     spapr_hcall_fn *slot;
2032 
2033     if (opcode <= MAX_HCALL_OPCODE) {
2034         assert((opcode & 0x3) == 0);
2035 
2036         slot = &papr_hypercall_table[opcode / 4];
2037     } else if (opcode >= SVM_HCALL_BASE && opcode <= SVM_HCALL_MAX) {
2038         /* we only have SVM-related hcall numbers assigned in multiples of 4 */
2039         assert((opcode & 0x3) == 0);
2040 
2041         slot = &svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
2042     } else {
2043         assert((opcode >= KVMPPC_HCALL_BASE) && (opcode <= KVMPPC_HCALL_MAX));
2044 
2045         slot = &kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
2046     }
2047 
2048     assert(!(*slot));
2049     *slot = fn;
2050 }
2051 
2052 target_ulong spapr_hypercall(PowerPCCPU *cpu, target_ulong opcode,
2053                              target_ulong *args)
2054 {
2055     SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
2056 
2057     if ((opcode <= MAX_HCALL_OPCODE)
2058         && ((opcode & 0x3) == 0)) {
2059         spapr_hcall_fn fn = papr_hypercall_table[opcode / 4];
2060 
2061         if (fn) {
2062             return fn(cpu, spapr, opcode, args);
2063         }
2064     } else if ((opcode >= SVM_HCALL_BASE) &&
2065                (opcode <= SVM_HCALL_MAX)) {
2066         spapr_hcall_fn fn = svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
2067 
2068         if (fn) {
2069             return fn(cpu, spapr, opcode, args);
2070         }
2071     } else if ((opcode >= KVMPPC_HCALL_BASE) &&
2072                (opcode <= KVMPPC_HCALL_MAX)) {
2073         spapr_hcall_fn fn = kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
2074 
2075         if (fn) {
2076             return fn(cpu, spapr, opcode, args);
2077         }
2078     }
2079 
2080     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x" TARGET_FMT_lx "\n",
2081                   opcode);
2082     return H_FUNCTION;
2083 }
2084 
2085 static void hypercall_register_types(void)
2086 {
2087     /* hcall-pft */
2088     spapr_register_hypercall(H_ENTER, h_enter);
2089     spapr_register_hypercall(H_REMOVE, h_remove);
2090     spapr_register_hypercall(H_PROTECT, h_protect);
2091     spapr_register_hypercall(H_READ, h_read);
2092 
2093     /* hcall-bulk */
2094     spapr_register_hypercall(H_BULK_REMOVE, h_bulk_remove);
2095 
2096     /* hcall-hpt-resize */
2097     spapr_register_hypercall(H_RESIZE_HPT_PREPARE, h_resize_hpt_prepare);
2098     spapr_register_hypercall(H_RESIZE_HPT_COMMIT, h_resize_hpt_commit);
2099 
2100     /* hcall-splpar */
2101     spapr_register_hypercall(H_REGISTER_VPA, h_register_vpa);
2102     spapr_register_hypercall(H_CEDE, h_cede);
2103     spapr_register_hypercall(H_CONFER, h_confer);
2104     spapr_register_hypercall(H_PROD, h_prod);
2105 
2106     /* hcall-join */
2107     spapr_register_hypercall(H_JOIN, h_join);
2108 
2109     spapr_register_hypercall(H_SIGNAL_SYS_RESET, h_signal_sys_reset);
2110 
2111     /* processor register resource access h-calls */
2112     spapr_register_hypercall(H_SET_SPRG0, h_set_sprg0);
2113     spapr_register_hypercall(H_SET_DABR, h_set_dabr);
2114     spapr_register_hypercall(H_SET_XDABR, h_set_xdabr);
2115     spapr_register_hypercall(H_PAGE_INIT, h_page_init);
2116     spapr_register_hypercall(H_SET_MODE, h_set_mode);
2117 
2118     /* In Memory Table MMU h-calls */
2119     spapr_register_hypercall(H_CLEAN_SLB, h_clean_slb);
2120     spapr_register_hypercall(H_INVALIDATE_PID, h_invalidate_pid);
2121     spapr_register_hypercall(H_REGISTER_PROC_TBL, h_register_process_table);
2122 
2123     /* hcall-get-cpu-characteristics */
2124     spapr_register_hypercall(H_GET_CPU_CHARACTERISTICS,
2125                              h_get_cpu_characteristics);
2126 
2127     /* "debugger" hcalls (also used by SLOF). Note: We do -not- differenciate
2128      * here between the "CI" and the "CACHE" variants, they will use whatever
2129      * mapping attributes qemu is using. When using KVM, the kernel will
2130      * enforce the attributes more strongly
2131      */
2132     spapr_register_hypercall(H_LOGICAL_CI_LOAD, h_logical_load);
2133     spapr_register_hypercall(H_LOGICAL_CI_STORE, h_logical_store);
2134     spapr_register_hypercall(H_LOGICAL_CACHE_LOAD, h_logical_load);
2135     spapr_register_hypercall(H_LOGICAL_CACHE_STORE, h_logical_store);
2136     spapr_register_hypercall(H_LOGICAL_ICBI, h_logical_icbi);
2137     spapr_register_hypercall(H_LOGICAL_DCBF, h_logical_dcbf);
2138     spapr_register_hypercall(KVMPPC_H_LOGICAL_MEMOP, h_logical_memop);
2139 
2140     /* qemu/KVM-PPC specific hcalls */
2141     spapr_register_hypercall(KVMPPC_H_RTAS, h_rtas);
2142 
2143     /* ibm,client-architecture-support support */
2144     spapr_register_hypercall(KVMPPC_H_CAS, h_client_architecture_support);
2145 
2146     spapr_register_hypercall(KVMPPC_H_UPDATE_DT, h_update_dt);
2147 
2148     /* Virtual Processor Home Node */
2149     spapr_register_hypercall(H_HOME_NODE_ASSOCIATIVITY,
2150                              h_home_node_associativity);
2151 }
2152 
2153 type_init(hypercall_register_types)
2154