xref: /openbmc/qemu/hw/ppc/spapr_hcall.c (revision ac12b601)
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 "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/fdt.h"
19 #include "hw/ppc/spapr_ovec.h"
20 #include "mmu-book3s-v3.h"
21 #include "hw/mem/memory-device.h"
22 
23 static bool has_spr(PowerPCCPU *cpu, int spr)
24 {
25     /* We can test whether the SPR is defined by checking for a valid name */
26     return cpu->env.spr_cb[spr].name != NULL;
27 }
28 
29 static inline bool valid_ptex(PowerPCCPU *cpu, target_ulong ptex)
30 {
31     /*
32      * hash value/pteg group index is normalized by HPT mask
33      */
34     if (((ptex & ~7ULL) / HPTES_PER_GROUP) & ~ppc_hash64_hpt_mask(cpu)) {
35         return false;
36     }
37     return true;
38 }
39 
40 static bool is_ram_address(SpaprMachineState *spapr, hwaddr addr)
41 {
42     MachineState *machine = MACHINE(spapr);
43     DeviceMemoryState *dms = machine->device_memory;
44 
45     if (addr < machine->ram_size) {
46         return true;
47     }
48     if ((addr >= dms->base)
49         && ((addr - dms->base) < memory_region_size(&dms->mr))) {
50         return true;
51     }
52 
53     return false;
54 }
55 
56 static target_ulong h_enter(PowerPCCPU *cpu, SpaprMachineState *spapr,
57                             target_ulong opcode, target_ulong *args)
58 {
59     target_ulong flags = args[0];
60     target_ulong ptex = args[1];
61     target_ulong pteh = args[2];
62     target_ulong ptel = args[3];
63     unsigned apshift;
64     target_ulong raddr;
65     target_ulong slot;
66     const ppc_hash_pte64_t *hptes;
67 
68     apshift = ppc_hash64_hpte_page_shift_noslb(cpu, pteh, ptel);
69     if (!apshift) {
70         /* Bad page size encoding */
71         return H_PARAMETER;
72     }
73 
74     raddr = (ptel & HPTE64_R_RPN) & ~((1ULL << apshift) - 1);
75 
76     if (is_ram_address(spapr, raddr)) {
77         /* Regular RAM - should have WIMG=0010 */
78         if ((ptel & HPTE64_R_WIMG) != HPTE64_R_M) {
79             return H_PARAMETER;
80         }
81     } else {
82         target_ulong wimg_flags;
83         /* Looks like an IO address */
84         /* FIXME: What WIMG combinations could be sensible for IO?
85          * For now we allow WIMG=010x, but are there others? */
86         /* FIXME: Should we check against registered IO addresses? */
87         wimg_flags = (ptel & (HPTE64_R_W | HPTE64_R_I | HPTE64_R_M));
88 
89         if (wimg_flags != HPTE64_R_I &&
90             wimg_flags != (HPTE64_R_I | HPTE64_R_M)) {
91             return H_PARAMETER;
92         }
93     }
94 
95     pteh &= ~0x60ULL;
96 
97     if (!valid_ptex(cpu, ptex)) {
98         return H_PARAMETER;
99     }
100 
101     slot = ptex & 7ULL;
102     ptex = ptex & ~7ULL;
103 
104     if (likely((flags & H_EXACT) == 0)) {
105         hptes = ppc_hash64_map_hptes(cpu, ptex, HPTES_PER_GROUP);
106         for (slot = 0; slot < 8; slot++) {
107             if (!(ppc_hash64_hpte0(cpu, hptes, slot) & HPTE64_V_VALID)) {
108                 break;
109             }
110         }
111         ppc_hash64_unmap_hptes(cpu, hptes, ptex, HPTES_PER_GROUP);
112         if (slot == 8) {
113             return H_PTEG_FULL;
114         }
115     } else {
116         hptes = ppc_hash64_map_hptes(cpu, ptex + slot, 1);
117         if (ppc_hash64_hpte0(cpu, hptes, 0) & HPTE64_V_VALID) {
118             ppc_hash64_unmap_hptes(cpu, hptes, ptex + slot, 1);
119             return H_PTEG_FULL;
120         }
121         ppc_hash64_unmap_hptes(cpu, hptes, ptex, 1);
122     }
123 
124     spapr_store_hpte(cpu, ptex + slot, pteh | HPTE64_V_HPTE_DIRTY, ptel);
125 
126     args[0] = ptex + slot;
127     return H_SUCCESS;
128 }
129 
130 typedef enum {
131     REMOVE_SUCCESS = 0,
132     REMOVE_NOT_FOUND = 1,
133     REMOVE_PARM = 2,
134     REMOVE_HW = 3,
135 } RemoveResult;
136 
137 static RemoveResult remove_hpte(PowerPCCPU *cpu
138                                 , target_ulong ptex,
139                                 target_ulong avpn,
140                                 target_ulong flags,
141                                 target_ulong *vp, target_ulong *rp)
142 {
143     const ppc_hash_pte64_t *hptes;
144     target_ulong v, r;
145 
146     if (!valid_ptex(cpu, ptex)) {
147         return REMOVE_PARM;
148     }
149 
150     hptes = ppc_hash64_map_hptes(cpu, ptex, 1);
151     v = ppc_hash64_hpte0(cpu, hptes, 0);
152     r = ppc_hash64_hpte1(cpu, hptes, 0);
153     ppc_hash64_unmap_hptes(cpu, hptes, ptex, 1);
154 
155     if ((v & HPTE64_V_VALID) == 0 ||
156         ((flags & H_AVPN) && (v & ~0x7fULL) != avpn) ||
157         ((flags & H_ANDCOND) && (v & avpn) != 0)) {
158         return REMOVE_NOT_FOUND;
159     }
160     *vp = v;
161     *rp = r;
162     spapr_store_hpte(cpu, ptex, HPTE64_V_HPTE_DIRTY, 0);
163     ppc_hash64_tlb_flush_hpte(cpu, ptex, v, r);
164     return REMOVE_SUCCESS;
165 }
166 
167 static target_ulong h_remove(PowerPCCPU *cpu, SpaprMachineState *spapr,
168                              target_ulong opcode, target_ulong *args)
169 {
170     CPUPPCState *env = &cpu->env;
171     target_ulong flags = args[0];
172     target_ulong ptex = args[1];
173     target_ulong avpn = args[2];
174     RemoveResult ret;
175 
176     ret = remove_hpte(cpu, ptex, avpn, flags,
177                       &args[0], &args[1]);
178 
179     switch (ret) {
180     case REMOVE_SUCCESS:
181         check_tlb_flush(env, true);
182         return H_SUCCESS;
183 
184     case REMOVE_NOT_FOUND:
185         return H_NOT_FOUND;
186 
187     case REMOVE_PARM:
188         return H_PARAMETER;
189 
190     case REMOVE_HW:
191         return H_HARDWARE;
192     }
193 
194     g_assert_not_reached();
195 }
196 
197 #define H_BULK_REMOVE_TYPE             0xc000000000000000ULL
198 #define   H_BULK_REMOVE_REQUEST        0x4000000000000000ULL
199 #define   H_BULK_REMOVE_RESPONSE       0x8000000000000000ULL
200 #define   H_BULK_REMOVE_END            0xc000000000000000ULL
201 #define H_BULK_REMOVE_CODE             0x3000000000000000ULL
202 #define   H_BULK_REMOVE_SUCCESS        0x0000000000000000ULL
203 #define   H_BULK_REMOVE_NOT_FOUND      0x1000000000000000ULL
204 #define   H_BULK_REMOVE_PARM           0x2000000000000000ULL
205 #define   H_BULK_REMOVE_HW             0x3000000000000000ULL
206 #define H_BULK_REMOVE_RC               0x0c00000000000000ULL
207 #define H_BULK_REMOVE_FLAGS            0x0300000000000000ULL
208 #define   H_BULK_REMOVE_ABSOLUTE       0x0000000000000000ULL
209 #define   H_BULK_REMOVE_ANDCOND        0x0100000000000000ULL
210 #define   H_BULK_REMOVE_AVPN           0x0200000000000000ULL
211 #define H_BULK_REMOVE_PTEX             0x00ffffffffffffffULL
212 
213 #define H_BULK_REMOVE_MAX_BATCH        4
214 
215 static target_ulong h_bulk_remove(PowerPCCPU *cpu, SpaprMachineState *spapr,
216                                   target_ulong opcode, target_ulong *args)
217 {
218     CPUPPCState *env = &cpu->env;
219     int i;
220     target_ulong rc = H_SUCCESS;
221 
222     for (i = 0; i < H_BULK_REMOVE_MAX_BATCH; i++) {
223         target_ulong *tsh = &args[i*2];
224         target_ulong tsl = args[i*2 + 1];
225         target_ulong v, r, ret;
226 
227         if ((*tsh & H_BULK_REMOVE_TYPE) == H_BULK_REMOVE_END) {
228             break;
229         } else if ((*tsh & H_BULK_REMOVE_TYPE) != H_BULK_REMOVE_REQUEST) {
230             return H_PARAMETER;
231         }
232 
233         *tsh &= H_BULK_REMOVE_PTEX | H_BULK_REMOVE_FLAGS;
234         *tsh |= H_BULK_REMOVE_RESPONSE;
235 
236         if ((*tsh & H_BULK_REMOVE_ANDCOND) && (*tsh & H_BULK_REMOVE_AVPN)) {
237             *tsh |= H_BULK_REMOVE_PARM;
238             return H_PARAMETER;
239         }
240 
241         ret = remove_hpte(cpu, *tsh & H_BULK_REMOVE_PTEX, tsl,
242                           (*tsh & H_BULK_REMOVE_FLAGS) >> 26,
243                           &v, &r);
244 
245         *tsh |= ret << 60;
246 
247         switch (ret) {
248         case REMOVE_SUCCESS:
249             *tsh |= (r & (HPTE64_R_C | HPTE64_R_R)) << 43;
250             break;
251 
252         case REMOVE_PARM:
253             rc = H_PARAMETER;
254             goto exit;
255 
256         case REMOVE_HW:
257             rc = H_HARDWARE;
258             goto exit;
259         }
260     }
261  exit:
262     check_tlb_flush(env, true);
263 
264     return rc;
265 }
266 
267 static target_ulong h_protect(PowerPCCPU *cpu, SpaprMachineState *spapr,
268                               target_ulong opcode, target_ulong *args)
269 {
270     CPUPPCState *env = &cpu->env;
271     target_ulong flags = args[0];
272     target_ulong ptex = args[1];
273     target_ulong avpn = args[2];
274     const ppc_hash_pte64_t *hptes;
275     target_ulong v, r;
276 
277     if (!valid_ptex(cpu, ptex)) {
278         return H_PARAMETER;
279     }
280 
281     hptes = ppc_hash64_map_hptes(cpu, ptex, 1);
282     v = ppc_hash64_hpte0(cpu, hptes, 0);
283     r = ppc_hash64_hpte1(cpu, hptes, 0);
284     ppc_hash64_unmap_hptes(cpu, hptes, ptex, 1);
285 
286     if ((v & HPTE64_V_VALID) == 0 ||
287         ((flags & H_AVPN) && (v & ~0x7fULL) != avpn)) {
288         return H_NOT_FOUND;
289     }
290 
291     r &= ~(HPTE64_R_PP0 | HPTE64_R_PP | HPTE64_R_N |
292            HPTE64_R_KEY_HI | HPTE64_R_KEY_LO);
293     r |= (flags << 55) & HPTE64_R_PP0;
294     r |= (flags << 48) & HPTE64_R_KEY_HI;
295     r |= flags & (HPTE64_R_PP | HPTE64_R_N | HPTE64_R_KEY_LO);
296     spapr_store_hpte(cpu, ptex,
297                      (v & ~HPTE64_V_VALID) | HPTE64_V_HPTE_DIRTY, 0);
298     ppc_hash64_tlb_flush_hpte(cpu, ptex, v, r);
299     /* Flush the tlb */
300     check_tlb_flush(env, true);
301     /* Don't need a memory barrier, due to qemu's global lock */
302     spapr_store_hpte(cpu, ptex, v | HPTE64_V_HPTE_DIRTY, r);
303     return H_SUCCESS;
304 }
305 
306 static target_ulong h_read(PowerPCCPU *cpu, SpaprMachineState *spapr,
307                            target_ulong opcode, target_ulong *args)
308 {
309     target_ulong flags = args[0];
310     target_ulong ptex = args[1];
311     int i, ridx, n_entries = 1;
312     const ppc_hash_pte64_t *hptes;
313 
314     if (!valid_ptex(cpu, ptex)) {
315         return H_PARAMETER;
316     }
317 
318     if (flags & H_READ_4) {
319         /* Clear the two low order bits */
320         ptex &= ~(3ULL);
321         n_entries = 4;
322     }
323 
324     hptes = ppc_hash64_map_hptes(cpu, ptex, n_entries);
325     for (i = 0, ridx = 0; i < n_entries; i++) {
326         args[ridx++] = ppc_hash64_hpte0(cpu, hptes, i);
327         args[ridx++] = ppc_hash64_hpte1(cpu, hptes, i);
328     }
329     ppc_hash64_unmap_hptes(cpu, hptes, ptex, n_entries);
330 
331     return H_SUCCESS;
332 }
333 
334 struct SpaprPendingHpt {
335     /* These fields are read-only after initialization */
336     int shift;
337     QemuThread thread;
338 
339     /* These fields are protected by the BQL */
340     bool complete;
341 
342     /* These fields are private to the preparation thread if
343      * !complete, otherwise protected by the BQL */
344     int ret;
345     void *hpt;
346 };
347 
348 static void free_pending_hpt(SpaprPendingHpt *pending)
349 {
350     if (pending->hpt) {
351         qemu_vfree(pending->hpt);
352     }
353 
354     g_free(pending);
355 }
356 
357 static void *hpt_prepare_thread(void *opaque)
358 {
359     SpaprPendingHpt *pending = opaque;
360     size_t size = 1ULL << pending->shift;
361 
362     pending->hpt = qemu_try_memalign(size, size);
363     if (pending->hpt) {
364         memset(pending->hpt, 0, size);
365         pending->ret = H_SUCCESS;
366     } else {
367         pending->ret = H_NO_MEM;
368     }
369 
370     qemu_mutex_lock_iothread();
371 
372     if (SPAPR_MACHINE(qdev_get_machine())->pending_hpt == pending) {
373         /* Ready to go */
374         pending->complete = true;
375     } else {
376         /* We've been cancelled, clean ourselves up */
377         free_pending_hpt(pending);
378     }
379 
380     qemu_mutex_unlock_iothread();
381     return NULL;
382 }
383 
384 /* Must be called with BQL held */
385 static void cancel_hpt_prepare(SpaprMachineState *spapr)
386 {
387     SpaprPendingHpt *pending = spapr->pending_hpt;
388 
389     /* Let the thread know it's cancelled */
390     spapr->pending_hpt = NULL;
391 
392     if (!pending) {
393         /* Nothing to do */
394         return;
395     }
396 
397     if (!pending->complete) {
398         /* thread will clean itself up */
399         return;
400     }
401 
402     free_pending_hpt(pending);
403 }
404 
405 /* Convert a return code from the KVM ioctl()s implementing resize HPT
406  * into a PAPR hypercall return code */
407 static target_ulong resize_hpt_convert_rc(int ret)
408 {
409     if (ret >= 100000) {
410         return H_LONG_BUSY_ORDER_100_SEC;
411     } else if (ret >= 10000) {
412         return H_LONG_BUSY_ORDER_10_SEC;
413     } else if (ret >= 1000) {
414         return H_LONG_BUSY_ORDER_1_SEC;
415     } else if (ret >= 100) {
416         return H_LONG_BUSY_ORDER_100_MSEC;
417     } else if (ret >= 10) {
418         return H_LONG_BUSY_ORDER_10_MSEC;
419     } else if (ret > 0) {
420         return H_LONG_BUSY_ORDER_1_MSEC;
421     }
422 
423     switch (ret) {
424     case 0:
425         return H_SUCCESS;
426     case -EPERM:
427         return H_AUTHORITY;
428     case -EINVAL:
429         return H_PARAMETER;
430     case -ENXIO:
431         return H_CLOSED;
432     case -ENOSPC:
433         return H_PTEG_FULL;
434     case -EBUSY:
435         return H_BUSY;
436     case -ENOMEM:
437         return H_NO_MEM;
438     default:
439         return H_HARDWARE;
440     }
441 }
442 
443 static target_ulong h_resize_hpt_prepare(PowerPCCPU *cpu,
444                                          SpaprMachineState *spapr,
445                                          target_ulong opcode,
446                                          target_ulong *args)
447 {
448     target_ulong flags = args[0];
449     int shift = args[1];
450     SpaprPendingHpt *pending = spapr->pending_hpt;
451     uint64_t current_ram_size;
452     int rc;
453 
454     if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
455         return H_AUTHORITY;
456     }
457 
458     if (!spapr->htab_shift) {
459         /* Radix guest, no HPT */
460         return H_NOT_AVAILABLE;
461     }
462 
463     trace_spapr_h_resize_hpt_prepare(flags, shift);
464 
465     if (flags != 0) {
466         return H_PARAMETER;
467     }
468 
469     if (shift && ((shift < 18) || (shift > 46))) {
470         return H_PARAMETER;
471     }
472 
473     current_ram_size = MACHINE(spapr)->ram_size + get_plugged_memory_size();
474 
475     /* We only allow the guest to allocate an HPT one order above what
476      * we'd normally give them (to stop a small guest claiming a huge
477      * chunk of resources in the HPT */
478     if (shift > (spapr_hpt_shift_for_ramsize(current_ram_size) + 1)) {
479         return H_RESOURCE;
480     }
481 
482     rc = kvmppc_resize_hpt_prepare(cpu, flags, shift);
483     if (rc != -ENOSYS) {
484         return resize_hpt_convert_rc(rc);
485     }
486 
487     if (pending) {
488         /* something already in progress */
489         if (pending->shift == shift) {
490             /* and it's suitable */
491             if (pending->complete) {
492                 return pending->ret;
493             } else {
494                 return H_LONG_BUSY_ORDER_100_MSEC;
495             }
496         }
497 
498         /* not suitable, cancel and replace */
499         cancel_hpt_prepare(spapr);
500     }
501 
502     if (!shift) {
503         /* nothing to do */
504         return H_SUCCESS;
505     }
506 
507     /* start new prepare */
508 
509     pending = g_new0(SpaprPendingHpt, 1);
510     pending->shift = shift;
511     pending->ret = H_HARDWARE;
512 
513     qemu_thread_create(&pending->thread, "sPAPR HPT prepare",
514                        hpt_prepare_thread, pending, QEMU_THREAD_DETACHED);
515 
516     spapr->pending_hpt = pending;
517 
518     /* In theory we could estimate the time more accurately based on
519      * the new size, but there's not much point */
520     return H_LONG_BUSY_ORDER_100_MSEC;
521 }
522 
523 static uint64_t new_hpte_load0(void *htab, uint64_t pteg, int slot)
524 {
525     uint8_t *addr = htab;
526 
527     addr += pteg * HASH_PTEG_SIZE_64;
528     addr += slot * HASH_PTE_SIZE_64;
529     return  ldq_p(addr);
530 }
531 
532 static void new_hpte_store(void *htab, uint64_t pteg, int slot,
533                            uint64_t pte0, uint64_t pte1)
534 {
535     uint8_t *addr = htab;
536 
537     addr += pteg * HASH_PTEG_SIZE_64;
538     addr += slot * HASH_PTE_SIZE_64;
539 
540     stq_p(addr, pte0);
541     stq_p(addr + HASH_PTE_SIZE_64 / 2, pte1);
542 }
543 
544 static int rehash_hpte(PowerPCCPU *cpu,
545                        const ppc_hash_pte64_t *hptes,
546                        void *old_hpt, uint64_t oldsize,
547                        void *new_hpt, uint64_t newsize,
548                        uint64_t pteg, int slot)
549 {
550     uint64_t old_hash_mask = (oldsize >> 7) - 1;
551     uint64_t new_hash_mask = (newsize >> 7) - 1;
552     target_ulong pte0 = ppc_hash64_hpte0(cpu, hptes, slot);
553     target_ulong pte1;
554     uint64_t avpn;
555     unsigned base_pg_shift;
556     uint64_t hash, new_pteg, replace_pte0;
557 
558     if (!(pte0 & HPTE64_V_VALID) || !(pte0 & HPTE64_V_BOLTED)) {
559         return H_SUCCESS;
560     }
561 
562     pte1 = ppc_hash64_hpte1(cpu, hptes, slot);
563 
564     base_pg_shift = ppc_hash64_hpte_page_shift_noslb(cpu, pte0, pte1);
565     assert(base_pg_shift); /* H_ENTER shouldn't allow a bad encoding */
566     avpn = HPTE64_V_AVPN_VAL(pte0) & ~(((1ULL << base_pg_shift) - 1) >> 23);
567 
568     if (pte0 & HPTE64_V_SECONDARY) {
569         pteg = ~pteg;
570     }
571 
572     if ((pte0 & HPTE64_V_SSIZE) == HPTE64_V_SSIZE_256M) {
573         uint64_t offset, vsid;
574 
575         /* We only have 28 - 23 bits of offset in avpn */
576         offset = (avpn & 0x1f) << 23;
577         vsid = avpn >> 5;
578         /* We can find more bits from the pteg value */
579         if (base_pg_shift < 23) {
580             offset |= ((vsid ^ pteg) & old_hash_mask) << base_pg_shift;
581         }
582 
583         hash = vsid ^ (offset >> base_pg_shift);
584     } else if ((pte0 & HPTE64_V_SSIZE) == HPTE64_V_SSIZE_1T) {
585         uint64_t offset, vsid;
586 
587         /* We only have 40 - 23 bits of seg_off in avpn */
588         offset = (avpn & 0x1ffff) << 23;
589         vsid = avpn >> 17;
590         if (base_pg_shift < 23) {
591             offset |= ((vsid ^ (vsid << 25) ^ pteg) & old_hash_mask)
592                 << base_pg_shift;
593         }
594 
595         hash = vsid ^ (vsid << 25) ^ (offset >> base_pg_shift);
596     } else {
597         error_report("rehash_pte: Bad segment size in HPTE");
598         return H_HARDWARE;
599     }
600 
601     new_pteg = hash & new_hash_mask;
602     if (pte0 & HPTE64_V_SECONDARY) {
603         assert(~pteg == (hash & old_hash_mask));
604         new_pteg = ~new_pteg;
605     } else {
606         assert(pteg == (hash & old_hash_mask));
607     }
608     assert((oldsize != newsize) || (pteg == new_pteg));
609     replace_pte0 = new_hpte_load0(new_hpt, new_pteg, slot);
610     /*
611      * Strictly speaking, we don't need all these tests, since we only
612      * ever rehash bolted HPTEs.  We might in future handle non-bolted
613      * HPTEs, though so make the logic correct for those cases as
614      * well.
615      */
616     if (replace_pte0 & HPTE64_V_VALID) {
617         assert(newsize < oldsize);
618         if (replace_pte0 & HPTE64_V_BOLTED) {
619             if (pte0 & HPTE64_V_BOLTED) {
620                 /* Bolted collision, nothing we can do */
621                 return H_PTEG_FULL;
622             } else {
623                 /* Discard this hpte */
624                 return H_SUCCESS;
625             }
626         }
627     }
628 
629     new_hpte_store(new_hpt, new_pteg, slot, pte0, pte1);
630     return H_SUCCESS;
631 }
632 
633 static int rehash_hpt(PowerPCCPU *cpu,
634                       void *old_hpt, uint64_t oldsize,
635                       void *new_hpt, uint64_t newsize)
636 {
637     uint64_t n_ptegs = oldsize >> 7;
638     uint64_t pteg;
639     int slot;
640     int rc;
641 
642     for (pteg = 0; pteg < n_ptegs; pteg++) {
643         hwaddr ptex = pteg * HPTES_PER_GROUP;
644         const ppc_hash_pte64_t *hptes
645             = ppc_hash64_map_hptes(cpu, ptex, HPTES_PER_GROUP);
646 
647         if (!hptes) {
648             return H_HARDWARE;
649         }
650 
651         for (slot = 0; slot < HPTES_PER_GROUP; slot++) {
652             rc = rehash_hpte(cpu, hptes, old_hpt, oldsize, new_hpt, newsize,
653                              pteg, slot);
654             if (rc != H_SUCCESS) {
655                 ppc_hash64_unmap_hptes(cpu, hptes, ptex, HPTES_PER_GROUP);
656                 return rc;
657             }
658         }
659         ppc_hash64_unmap_hptes(cpu, hptes, ptex, HPTES_PER_GROUP);
660     }
661 
662     return H_SUCCESS;
663 }
664 
665 static void do_push_sregs_to_kvm_pr(CPUState *cs, run_on_cpu_data data)
666 {
667     int ret;
668 
669     cpu_synchronize_state(cs);
670 
671     ret = kvmppc_put_books_sregs(POWERPC_CPU(cs));
672     if (ret < 0) {
673         error_report("failed to push sregs to KVM: %s", strerror(-ret));
674         exit(1);
675     }
676 }
677 
678 static void push_sregs_to_kvm_pr(SpaprMachineState *spapr)
679 {
680     CPUState *cs;
681 
682     /*
683      * This is a hack for the benefit of KVM PR - it abuses the SDR1
684      * slot in kvm_sregs to communicate the userspace address of the
685      * HPT
686      */
687     if (!kvm_enabled() || !spapr->htab) {
688         return;
689     }
690 
691     CPU_FOREACH(cs) {
692         run_on_cpu(cs, do_push_sregs_to_kvm_pr, RUN_ON_CPU_NULL);
693     }
694 }
695 
696 static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
697                                         SpaprMachineState *spapr,
698                                         target_ulong opcode,
699                                         target_ulong *args)
700 {
701     target_ulong flags = args[0];
702     target_ulong shift = args[1];
703     SpaprPendingHpt *pending = spapr->pending_hpt;
704     int rc;
705     size_t newsize;
706 
707     if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
708         return H_AUTHORITY;
709     }
710 
711     if (!spapr->htab_shift) {
712         /* Radix guest, no HPT */
713         return H_NOT_AVAILABLE;
714     }
715 
716     trace_spapr_h_resize_hpt_commit(flags, shift);
717 
718     rc = kvmppc_resize_hpt_commit(cpu, flags, shift);
719     if (rc != -ENOSYS) {
720         rc = resize_hpt_convert_rc(rc);
721         if (rc == H_SUCCESS) {
722             /* Need to set the new htab_shift in the machine state */
723             spapr->htab_shift = shift;
724         }
725         return rc;
726     }
727 
728     if (flags != 0) {
729         return H_PARAMETER;
730     }
731 
732     if (!pending || (pending->shift != shift)) {
733         /* no matching prepare */
734         return H_CLOSED;
735     }
736 
737     if (!pending->complete) {
738         /* prepare has not completed */
739         return H_BUSY;
740     }
741 
742     /* Shouldn't have got past PREPARE without an HPT */
743     g_assert(spapr->htab_shift);
744 
745     newsize = 1ULL << pending->shift;
746     rc = rehash_hpt(cpu, spapr->htab, HTAB_SIZE(spapr),
747                     pending->hpt, newsize);
748     if (rc == H_SUCCESS) {
749         qemu_vfree(spapr->htab);
750         spapr->htab = pending->hpt;
751         spapr->htab_shift = pending->shift;
752 
753         push_sregs_to_kvm_pr(spapr);
754 
755         pending->hpt = NULL; /* so it's not free()d */
756     }
757 
758     /* Clean up */
759     spapr->pending_hpt = NULL;
760     free_pending_hpt(pending);
761 
762     return rc;
763 }
764 
765 static target_ulong h_set_sprg0(PowerPCCPU *cpu, SpaprMachineState *spapr,
766                                 target_ulong opcode, target_ulong *args)
767 {
768     cpu_synchronize_state(CPU(cpu));
769     cpu->env.spr[SPR_SPRG0] = args[0];
770 
771     return H_SUCCESS;
772 }
773 
774 static target_ulong h_set_dabr(PowerPCCPU *cpu, SpaprMachineState *spapr,
775                                target_ulong opcode, target_ulong *args)
776 {
777     if (!has_spr(cpu, SPR_DABR)) {
778         return H_HARDWARE;              /* DABR register not available */
779     }
780     cpu_synchronize_state(CPU(cpu));
781 
782     if (has_spr(cpu, SPR_DABRX)) {
783         cpu->env.spr[SPR_DABRX] = 0x3;  /* Use Problem and Privileged state */
784     } else if (!(args[0] & 0x4)) {      /* Breakpoint Translation set? */
785         return H_RESERVED_DABR;
786     }
787 
788     cpu->env.spr[SPR_DABR] = args[0];
789     return H_SUCCESS;
790 }
791 
792 static target_ulong h_set_xdabr(PowerPCCPU *cpu, SpaprMachineState *spapr,
793                                 target_ulong opcode, target_ulong *args)
794 {
795     target_ulong dabrx = args[1];
796 
797     if (!has_spr(cpu, SPR_DABR) || !has_spr(cpu, SPR_DABRX)) {
798         return H_HARDWARE;
799     }
800 
801     if ((dabrx & ~0xfULL) != 0 || (dabrx & H_DABRX_HYPERVISOR) != 0
802         || (dabrx & (H_DABRX_KERNEL | H_DABRX_USER)) == 0) {
803         return H_PARAMETER;
804     }
805 
806     cpu_synchronize_state(CPU(cpu));
807     cpu->env.spr[SPR_DABRX] = dabrx;
808     cpu->env.spr[SPR_DABR] = args[0];
809 
810     return H_SUCCESS;
811 }
812 
813 static target_ulong h_page_init(PowerPCCPU *cpu, SpaprMachineState *spapr,
814                                 target_ulong opcode, target_ulong *args)
815 {
816     target_ulong flags = args[0];
817     hwaddr dst = args[1];
818     hwaddr src = args[2];
819     hwaddr len = TARGET_PAGE_SIZE;
820     uint8_t *pdst, *psrc;
821     target_long ret = H_SUCCESS;
822 
823     if (flags & ~(H_ICACHE_SYNCHRONIZE | H_ICACHE_INVALIDATE
824                   | H_COPY_PAGE | H_ZERO_PAGE)) {
825         qemu_log_mask(LOG_UNIMP, "h_page_init: Bad flags (" TARGET_FMT_lx "\n",
826                       flags);
827         return H_PARAMETER;
828     }
829 
830     /* Map-in destination */
831     if (!is_ram_address(spapr, dst) || (dst & ~TARGET_PAGE_MASK) != 0) {
832         return H_PARAMETER;
833     }
834     pdst = cpu_physical_memory_map(dst, &len, true);
835     if (!pdst || len != TARGET_PAGE_SIZE) {
836         return H_PARAMETER;
837     }
838 
839     if (flags & H_COPY_PAGE) {
840         /* Map-in source, copy to destination, and unmap source again */
841         if (!is_ram_address(spapr, src) || (src & ~TARGET_PAGE_MASK) != 0) {
842             ret = H_PARAMETER;
843             goto unmap_out;
844         }
845         psrc = cpu_physical_memory_map(src, &len, false);
846         if (!psrc || len != TARGET_PAGE_SIZE) {
847             ret = H_PARAMETER;
848             goto unmap_out;
849         }
850         memcpy(pdst, psrc, len);
851         cpu_physical_memory_unmap(psrc, len, 0, len);
852     } else if (flags & H_ZERO_PAGE) {
853         memset(pdst, 0, len);          /* Just clear the destination page */
854     }
855 
856     if (kvm_enabled() && (flags & H_ICACHE_SYNCHRONIZE) != 0) {
857         kvmppc_dcbst_range(cpu, pdst, len);
858     }
859     if (flags & (H_ICACHE_SYNCHRONIZE | H_ICACHE_INVALIDATE)) {
860         if (kvm_enabled()) {
861             kvmppc_icbi_range(cpu, pdst, len);
862         } else {
863             tb_flush(CPU(cpu));
864         }
865     }
866 
867 unmap_out:
868     cpu_physical_memory_unmap(pdst, TARGET_PAGE_SIZE, 1, len);
869     return ret;
870 }
871 
872 #define FLAGS_REGISTER_VPA         0x0000200000000000ULL
873 #define FLAGS_REGISTER_DTL         0x0000400000000000ULL
874 #define FLAGS_REGISTER_SLBSHADOW   0x0000600000000000ULL
875 #define FLAGS_DEREGISTER_VPA       0x0000a00000000000ULL
876 #define FLAGS_DEREGISTER_DTL       0x0000c00000000000ULL
877 #define FLAGS_DEREGISTER_SLBSHADOW 0x0000e00000000000ULL
878 
879 static target_ulong register_vpa(PowerPCCPU *cpu, target_ulong vpa)
880 {
881     CPUState *cs = CPU(cpu);
882     CPUPPCState *env = &cpu->env;
883     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
884     uint16_t size;
885     uint8_t tmp;
886 
887     if (vpa == 0) {
888         hcall_dprintf("Can't cope with registering a VPA at logical 0\n");
889         return H_HARDWARE;
890     }
891 
892     if (vpa % env->dcache_line_size) {
893         return H_PARAMETER;
894     }
895     /* FIXME: bounds check the address */
896 
897     size = lduw_be_phys(cs->as, vpa + 0x4);
898 
899     if (size < VPA_MIN_SIZE) {
900         return H_PARAMETER;
901     }
902 
903     /* VPA is not allowed to cross a page boundary */
904     if ((vpa / 4096) != ((vpa + size - 1) / 4096)) {
905         return H_PARAMETER;
906     }
907 
908     spapr_cpu->vpa_addr = vpa;
909 
910     tmp = ldub_phys(cs->as, spapr_cpu->vpa_addr + VPA_SHARED_PROC_OFFSET);
911     tmp |= VPA_SHARED_PROC_VAL;
912     stb_phys(cs->as, spapr_cpu->vpa_addr + VPA_SHARED_PROC_OFFSET, tmp);
913 
914     return H_SUCCESS;
915 }
916 
917 static target_ulong deregister_vpa(PowerPCCPU *cpu, target_ulong vpa)
918 {
919     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
920 
921     if (spapr_cpu->slb_shadow_addr) {
922         return H_RESOURCE;
923     }
924 
925     if (spapr_cpu->dtl_addr) {
926         return H_RESOURCE;
927     }
928 
929     spapr_cpu->vpa_addr = 0;
930     return H_SUCCESS;
931 }
932 
933 static target_ulong register_slb_shadow(PowerPCCPU *cpu, target_ulong addr)
934 {
935     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
936     uint32_t size;
937 
938     if (addr == 0) {
939         hcall_dprintf("Can't cope with SLB shadow at logical 0\n");
940         return H_HARDWARE;
941     }
942 
943     size = ldl_be_phys(CPU(cpu)->as, addr + 0x4);
944     if (size < 0x8) {
945         return H_PARAMETER;
946     }
947 
948     if ((addr / 4096) != ((addr + size - 1) / 4096)) {
949         return H_PARAMETER;
950     }
951 
952     if (!spapr_cpu->vpa_addr) {
953         return H_RESOURCE;
954     }
955 
956     spapr_cpu->slb_shadow_addr = addr;
957     spapr_cpu->slb_shadow_size = size;
958 
959     return H_SUCCESS;
960 }
961 
962 static target_ulong deregister_slb_shadow(PowerPCCPU *cpu, target_ulong addr)
963 {
964     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
965 
966     spapr_cpu->slb_shadow_addr = 0;
967     spapr_cpu->slb_shadow_size = 0;
968     return H_SUCCESS;
969 }
970 
971 static target_ulong register_dtl(PowerPCCPU *cpu, target_ulong addr)
972 {
973     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
974     uint32_t size;
975 
976     if (addr == 0) {
977         hcall_dprintf("Can't cope with DTL at logical 0\n");
978         return H_HARDWARE;
979     }
980 
981     size = ldl_be_phys(CPU(cpu)->as, addr + 0x4);
982 
983     if (size < 48) {
984         return H_PARAMETER;
985     }
986 
987     if (!spapr_cpu->vpa_addr) {
988         return H_RESOURCE;
989     }
990 
991     spapr_cpu->dtl_addr = addr;
992     spapr_cpu->dtl_size = size;
993 
994     return H_SUCCESS;
995 }
996 
997 static target_ulong deregister_dtl(PowerPCCPU *cpu, target_ulong addr)
998 {
999     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
1000 
1001     spapr_cpu->dtl_addr = 0;
1002     spapr_cpu->dtl_size = 0;
1003 
1004     return H_SUCCESS;
1005 }
1006 
1007 static target_ulong h_register_vpa(PowerPCCPU *cpu, SpaprMachineState *spapr,
1008                                    target_ulong opcode, target_ulong *args)
1009 {
1010     target_ulong flags = args[0];
1011     target_ulong procno = args[1];
1012     target_ulong vpa = args[2];
1013     target_ulong ret = H_PARAMETER;
1014     PowerPCCPU *tcpu;
1015 
1016     tcpu = spapr_find_cpu(procno);
1017     if (!tcpu) {
1018         return H_PARAMETER;
1019     }
1020 
1021     switch (flags) {
1022     case FLAGS_REGISTER_VPA:
1023         ret = register_vpa(tcpu, vpa);
1024         break;
1025 
1026     case FLAGS_DEREGISTER_VPA:
1027         ret = deregister_vpa(tcpu, vpa);
1028         break;
1029 
1030     case FLAGS_REGISTER_SLBSHADOW:
1031         ret = register_slb_shadow(tcpu, vpa);
1032         break;
1033 
1034     case FLAGS_DEREGISTER_SLBSHADOW:
1035         ret = deregister_slb_shadow(tcpu, vpa);
1036         break;
1037 
1038     case FLAGS_REGISTER_DTL:
1039         ret = register_dtl(tcpu, vpa);
1040         break;
1041 
1042     case FLAGS_DEREGISTER_DTL:
1043         ret = deregister_dtl(tcpu, vpa);
1044         break;
1045     }
1046 
1047     return ret;
1048 }
1049 
1050 static target_ulong h_cede(PowerPCCPU *cpu, SpaprMachineState *spapr,
1051                            target_ulong opcode, target_ulong *args)
1052 {
1053     CPUPPCState *env = &cpu->env;
1054     CPUState *cs = CPU(cpu);
1055     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
1056 
1057     env->msr |= (1ULL << MSR_EE);
1058     hreg_compute_hflags(env);
1059 
1060     if (spapr_cpu->prod) {
1061         spapr_cpu->prod = false;
1062         return H_SUCCESS;
1063     }
1064 
1065     if (!cpu_has_work(cs)) {
1066         cs->halted = 1;
1067         cs->exception_index = EXCP_HLT;
1068         cs->exit_request = 1;
1069     }
1070 
1071     return H_SUCCESS;
1072 }
1073 
1074 /*
1075  * Confer to self, aka join. Cede could use the same pattern as well, if
1076  * EXCP_HLT can be changed to ECXP_HALTED.
1077  */
1078 static target_ulong h_confer_self(PowerPCCPU *cpu)
1079 {
1080     CPUState *cs = CPU(cpu);
1081     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
1082 
1083     if (spapr_cpu->prod) {
1084         spapr_cpu->prod = false;
1085         return H_SUCCESS;
1086     }
1087     cs->halted = 1;
1088     cs->exception_index = EXCP_HALTED;
1089     cs->exit_request = 1;
1090 
1091     return H_SUCCESS;
1092 }
1093 
1094 static target_ulong h_join(PowerPCCPU *cpu, SpaprMachineState *spapr,
1095                            target_ulong opcode, target_ulong *args)
1096 {
1097     CPUPPCState *env = &cpu->env;
1098     CPUState *cs;
1099     bool last_unjoined = true;
1100 
1101     if (env->msr & (1ULL << MSR_EE)) {
1102         return H_BAD_MODE;
1103     }
1104 
1105     /*
1106      * Must not join the last CPU running. Interestingly, no such restriction
1107      * for H_CONFER-to-self, but that is probably not intended to be used
1108      * when H_JOIN is available.
1109      */
1110     CPU_FOREACH(cs) {
1111         PowerPCCPU *c = POWERPC_CPU(cs);
1112         CPUPPCState *e = &c->env;
1113         if (c == cpu) {
1114             continue;
1115         }
1116 
1117         /* Don't have a way to indicate joined, so use halted && MSR[EE]=0 */
1118         if (!cs->halted || (e->msr & (1ULL << MSR_EE))) {
1119             last_unjoined = false;
1120             break;
1121         }
1122     }
1123     if (last_unjoined) {
1124         return H_CONTINUE;
1125     }
1126 
1127     return h_confer_self(cpu);
1128 }
1129 
1130 static target_ulong h_confer(PowerPCCPU *cpu, SpaprMachineState *spapr,
1131                            target_ulong opcode, target_ulong *args)
1132 {
1133     target_long target = args[0];
1134     uint32_t dispatch = args[1];
1135     CPUState *cs = CPU(cpu);
1136     SpaprCpuState *spapr_cpu;
1137 
1138     /*
1139      * -1 means confer to all other CPUs without dispatch counter check,
1140      *  otherwise it's a targeted confer.
1141      */
1142     if (target != -1) {
1143         PowerPCCPU *target_cpu = spapr_find_cpu(target);
1144         uint32_t target_dispatch;
1145 
1146         if (!target_cpu) {
1147             return H_PARAMETER;
1148         }
1149 
1150         /*
1151          * target == self is a special case, we wait until prodded, without
1152          * dispatch counter check.
1153          */
1154         if (cpu == target_cpu) {
1155             return h_confer_self(cpu);
1156         }
1157 
1158         spapr_cpu = spapr_cpu_state(target_cpu);
1159         if (!spapr_cpu->vpa_addr || ((dispatch & 1) == 0)) {
1160             return H_SUCCESS;
1161         }
1162 
1163         target_dispatch = ldl_be_phys(cs->as,
1164                                   spapr_cpu->vpa_addr + VPA_DISPATCH_COUNTER);
1165         if (target_dispatch != dispatch) {
1166             return H_SUCCESS;
1167         }
1168 
1169         /*
1170          * The targeted confer does not do anything special beyond yielding
1171          * the current vCPU, but even this should be better than nothing.
1172          * At least for single-threaded tcg, it gives the target a chance to
1173          * run before we run again. Multi-threaded tcg does not really do
1174          * anything with EXCP_YIELD yet.
1175          */
1176     }
1177 
1178     cs->exception_index = EXCP_YIELD;
1179     cs->exit_request = 1;
1180     cpu_loop_exit(cs);
1181 
1182     return H_SUCCESS;
1183 }
1184 
1185 static target_ulong h_prod(PowerPCCPU *cpu, SpaprMachineState *spapr,
1186                            target_ulong opcode, target_ulong *args)
1187 {
1188     target_long target = args[0];
1189     PowerPCCPU *tcpu;
1190     CPUState *cs;
1191     SpaprCpuState *spapr_cpu;
1192 
1193     tcpu = spapr_find_cpu(target);
1194     cs = CPU(tcpu);
1195     if (!cs) {
1196         return H_PARAMETER;
1197     }
1198 
1199     spapr_cpu = spapr_cpu_state(tcpu);
1200     spapr_cpu->prod = true;
1201     cs->halted = 0;
1202     qemu_cpu_kick(cs);
1203 
1204     return H_SUCCESS;
1205 }
1206 
1207 static target_ulong h_rtas(PowerPCCPU *cpu, SpaprMachineState *spapr,
1208                            target_ulong opcode, target_ulong *args)
1209 {
1210     target_ulong rtas_r3 = args[0];
1211     uint32_t token = rtas_ld(rtas_r3, 0);
1212     uint32_t nargs = rtas_ld(rtas_r3, 1);
1213     uint32_t nret = rtas_ld(rtas_r3, 2);
1214 
1215     return spapr_rtas_call(cpu, spapr, token, nargs, rtas_r3 + 12,
1216                            nret, rtas_r3 + 12 + 4*nargs);
1217 }
1218 
1219 static target_ulong h_logical_load(PowerPCCPU *cpu, SpaprMachineState *spapr,
1220                                    target_ulong opcode, target_ulong *args)
1221 {
1222     CPUState *cs = CPU(cpu);
1223     target_ulong size = args[0];
1224     target_ulong addr = args[1];
1225 
1226     switch (size) {
1227     case 1:
1228         args[0] = ldub_phys(cs->as, addr);
1229         return H_SUCCESS;
1230     case 2:
1231         args[0] = lduw_phys(cs->as, addr);
1232         return H_SUCCESS;
1233     case 4:
1234         args[0] = ldl_phys(cs->as, addr);
1235         return H_SUCCESS;
1236     case 8:
1237         args[0] = ldq_phys(cs->as, addr);
1238         return H_SUCCESS;
1239     }
1240     return H_PARAMETER;
1241 }
1242 
1243 static target_ulong h_logical_store(PowerPCCPU *cpu, SpaprMachineState *spapr,
1244                                     target_ulong opcode, target_ulong *args)
1245 {
1246     CPUState *cs = CPU(cpu);
1247 
1248     target_ulong size = args[0];
1249     target_ulong addr = args[1];
1250     target_ulong val  = args[2];
1251 
1252     switch (size) {
1253     case 1:
1254         stb_phys(cs->as, addr, val);
1255         return H_SUCCESS;
1256     case 2:
1257         stw_phys(cs->as, addr, val);
1258         return H_SUCCESS;
1259     case 4:
1260         stl_phys(cs->as, addr, val);
1261         return H_SUCCESS;
1262     case 8:
1263         stq_phys(cs->as, addr, val);
1264         return H_SUCCESS;
1265     }
1266     return H_PARAMETER;
1267 }
1268 
1269 static target_ulong h_logical_memop(PowerPCCPU *cpu, SpaprMachineState *spapr,
1270                                     target_ulong opcode, target_ulong *args)
1271 {
1272     CPUState *cs = CPU(cpu);
1273 
1274     target_ulong dst   = args[0]; /* Destination address */
1275     target_ulong src   = args[1]; /* Source address */
1276     target_ulong esize = args[2]; /* Element size (0=1,1=2,2=4,3=8) */
1277     target_ulong count = args[3]; /* Element count */
1278     target_ulong op    = args[4]; /* 0 = copy, 1 = invert */
1279     uint64_t tmp;
1280     unsigned int mask = (1 << esize) - 1;
1281     int step = 1 << esize;
1282 
1283     if (count > 0x80000000) {
1284         return H_PARAMETER;
1285     }
1286 
1287     if ((dst & mask) || (src & mask) || (op > 1)) {
1288         return H_PARAMETER;
1289     }
1290 
1291     if (dst >= src && dst < (src + (count << esize))) {
1292             dst = dst + ((count - 1) << esize);
1293             src = src + ((count - 1) << esize);
1294             step = -step;
1295     }
1296 
1297     while (count--) {
1298         switch (esize) {
1299         case 0:
1300             tmp = ldub_phys(cs->as, src);
1301             break;
1302         case 1:
1303             tmp = lduw_phys(cs->as, src);
1304             break;
1305         case 2:
1306             tmp = ldl_phys(cs->as, src);
1307             break;
1308         case 3:
1309             tmp = ldq_phys(cs->as, src);
1310             break;
1311         default:
1312             return H_PARAMETER;
1313         }
1314         if (op == 1) {
1315             tmp = ~tmp;
1316         }
1317         switch (esize) {
1318         case 0:
1319             stb_phys(cs->as, dst, tmp);
1320             break;
1321         case 1:
1322             stw_phys(cs->as, dst, tmp);
1323             break;
1324         case 2:
1325             stl_phys(cs->as, dst, tmp);
1326             break;
1327         case 3:
1328             stq_phys(cs->as, dst, tmp);
1329             break;
1330         }
1331         dst = dst + step;
1332         src = src + step;
1333     }
1334 
1335     return H_SUCCESS;
1336 }
1337 
1338 static target_ulong h_logical_icbi(PowerPCCPU *cpu, SpaprMachineState *spapr,
1339                                    target_ulong opcode, target_ulong *args)
1340 {
1341     /* Nothing to do on emulation, KVM will trap this in the kernel */
1342     return H_SUCCESS;
1343 }
1344 
1345 static target_ulong h_logical_dcbf(PowerPCCPU *cpu, SpaprMachineState *spapr,
1346                                    target_ulong opcode, target_ulong *args)
1347 {
1348     /* Nothing to do on emulation, KVM will trap this in the kernel */
1349     return H_SUCCESS;
1350 }
1351 
1352 static target_ulong h_set_mode_resource_le(PowerPCCPU *cpu,
1353                                            SpaprMachineState *spapr,
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(spapr, 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(spapr, 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 == 1) {
1398         /* AIL=1 is reserved in POWER8/POWER9/POWER10 */
1399         return H_UNSUPPORTED_FLAG;
1400     }
1401 
1402     if (mflags == 2 && (pcc->insns_flags2 & PPC2_ISA310)) {
1403         /* AIL=2 is reserved in POWER10 (ISA v3.1) */
1404         return H_UNSUPPORTED_FLAG;
1405     }
1406 
1407     spapr_set_all_lpcrs(mflags << LPCR_AIL_SHIFT, LPCR_AIL);
1408 
1409     return H_SUCCESS;
1410 }
1411 
1412 static target_ulong h_set_mode(PowerPCCPU *cpu, SpaprMachineState *spapr,
1413                                target_ulong opcode, target_ulong *args)
1414 {
1415     target_ulong resource = args[1];
1416     target_ulong ret = H_P2;
1417 
1418     switch (resource) {
1419     case H_SET_MODE_RESOURCE_LE:
1420         ret = h_set_mode_resource_le(cpu, spapr, args[0], args[2], args[3]);
1421         break;
1422     case H_SET_MODE_RESOURCE_ADDR_TRANS_MODE:
1423         ret = h_set_mode_resource_addr_trans_mode(cpu, args[0],
1424                                                   args[2], args[3]);
1425         break;
1426     }
1427 
1428     return ret;
1429 }
1430 
1431 static target_ulong h_clean_slb(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_CLEAN_SLB)");
1436     return H_FUNCTION;
1437 }
1438 
1439 static target_ulong h_invalidate_pid(PowerPCCPU *cpu, SpaprMachineState *spapr,
1440                                      target_ulong opcode, target_ulong *args)
1441 {
1442     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x"TARGET_FMT_lx"%s\n",
1443                   opcode, " (H_INVALIDATE_PID)");
1444     return H_FUNCTION;
1445 }
1446 
1447 static void spapr_check_setup_free_hpt(SpaprMachineState *spapr,
1448                                        uint64_t patbe_old, uint64_t patbe_new)
1449 {
1450     /*
1451      * We have 4 Options:
1452      * HASH->HASH || RADIX->RADIX || NOTHING->RADIX : Do Nothing
1453      * HASH->RADIX                                  : Free HPT
1454      * RADIX->HASH                                  : Allocate HPT
1455      * NOTHING->HASH                                : Allocate HPT
1456      * Note: NOTHING implies the case where we said the guest could choose
1457      *       later and so assumed radix and now it's called H_REG_PROC_TBL
1458      */
1459 
1460     if ((patbe_old & PATE1_GR) == (patbe_new & PATE1_GR)) {
1461         /* We assume RADIX, so this catches all the "Do Nothing" cases */
1462     } else if (!(patbe_old & PATE1_GR)) {
1463         /* HASH->RADIX : Free HPT */
1464         spapr_free_hpt(spapr);
1465     } else if (!(patbe_new & PATE1_GR)) {
1466         /* RADIX->HASH || NOTHING->HASH : Allocate HPT */
1467         spapr_setup_hpt(spapr);
1468     }
1469     return;
1470 }
1471 
1472 #define FLAGS_MASK              0x01FULL
1473 #define FLAG_MODIFY             0x10
1474 #define FLAG_REGISTER           0x08
1475 #define FLAG_RADIX              0x04
1476 #define FLAG_HASH_PROC_TBL      0x02
1477 #define FLAG_GTSE               0x01
1478 
1479 static target_ulong h_register_process_table(PowerPCCPU *cpu,
1480                                              SpaprMachineState *spapr,
1481                                              target_ulong opcode,
1482                                              target_ulong *args)
1483 {
1484     target_ulong flags = args[0];
1485     target_ulong proc_tbl = args[1];
1486     target_ulong page_size = args[2];
1487     target_ulong table_size = args[3];
1488     target_ulong update_lpcr = 0;
1489     uint64_t cproc;
1490 
1491     if (flags & ~FLAGS_MASK) { /* Check no reserved bits are set */
1492         return H_PARAMETER;
1493     }
1494     if (flags & FLAG_MODIFY) {
1495         if (flags & FLAG_REGISTER) {
1496             if (flags & FLAG_RADIX) { /* Register new RADIX process table */
1497                 if (proc_tbl & 0xfff || proc_tbl >> 60) {
1498                     return H_P2;
1499                 } else if (page_size) {
1500                     return H_P3;
1501                 } else if (table_size > 24) {
1502                     return H_P4;
1503                 }
1504                 cproc = PATE1_GR | proc_tbl | table_size;
1505             } else { /* Register new HPT process table */
1506                 if (flags & FLAG_HASH_PROC_TBL) { /* Hash with Segment Tables */
1507                     /* TODO - Not Supported */
1508                     /* Technically caused by flag bits => H_PARAMETER */
1509                     return H_PARAMETER;
1510                 } else { /* Hash with SLB */
1511                     if (proc_tbl >> 38) {
1512                         return H_P2;
1513                     } else if (page_size & ~0x7) {
1514                         return H_P3;
1515                     } else if (table_size > 24) {
1516                         return H_P4;
1517                     }
1518                 }
1519                 cproc = (proc_tbl << 25) | page_size << 5 | table_size;
1520             }
1521 
1522         } else { /* Deregister current process table */
1523             /*
1524              * Set to benign value: (current GR) | 0. This allows
1525              * deregistration in KVM to succeed even if the radix bit
1526              * in flags doesn't match the radix bit in the old PATE.
1527              */
1528             cproc = spapr->patb_entry & PATE1_GR;
1529         }
1530     } else { /* Maintain current registration */
1531         if (!(flags & FLAG_RADIX) != !(spapr->patb_entry & PATE1_GR)) {
1532             /* Technically caused by flag bits => H_PARAMETER */
1533             return H_PARAMETER; /* Existing Process Table Mismatch */
1534         }
1535         cproc = spapr->patb_entry;
1536     }
1537 
1538     /* Check if we need to setup OR free the hpt */
1539     spapr_check_setup_free_hpt(spapr, spapr->patb_entry, cproc);
1540 
1541     spapr->patb_entry = cproc; /* Save new process table */
1542 
1543     /* Update the UPRT, HR and GTSE bits in the LPCR for all cpus */
1544     if (flags & FLAG_RADIX)     /* Radix must use process tables, also set HR */
1545         update_lpcr |= (LPCR_UPRT | LPCR_HR);
1546     else if (flags & FLAG_HASH_PROC_TBL) /* Hash with process tables */
1547         update_lpcr |= LPCR_UPRT;
1548     if (flags & FLAG_GTSE)      /* Guest translation shootdown enable */
1549         update_lpcr |= LPCR_GTSE;
1550 
1551     spapr_set_all_lpcrs(update_lpcr, LPCR_UPRT | LPCR_HR | LPCR_GTSE);
1552 
1553     if (kvm_enabled()) {
1554         return kvmppc_configure_v3_mmu(cpu, flags & FLAG_RADIX,
1555                                        flags & FLAG_GTSE, cproc);
1556     }
1557     return H_SUCCESS;
1558 }
1559 
1560 #define H_SIGNAL_SYS_RESET_ALL         -1
1561 #define H_SIGNAL_SYS_RESET_ALLBUTSELF  -2
1562 
1563 static target_ulong h_signal_sys_reset(PowerPCCPU *cpu,
1564                                        SpaprMachineState *spapr,
1565                                        target_ulong opcode, target_ulong *args)
1566 {
1567     target_long target = args[0];
1568     CPUState *cs;
1569 
1570     if (target < 0) {
1571         /* Broadcast */
1572         if (target < H_SIGNAL_SYS_RESET_ALLBUTSELF) {
1573             return H_PARAMETER;
1574         }
1575 
1576         CPU_FOREACH(cs) {
1577             PowerPCCPU *c = POWERPC_CPU(cs);
1578 
1579             if (target == H_SIGNAL_SYS_RESET_ALLBUTSELF) {
1580                 if (c == cpu) {
1581                     continue;
1582                 }
1583             }
1584             run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL);
1585         }
1586         return H_SUCCESS;
1587 
1588     } else {
1589         /* Unicast */
1590         cs = CPU(spapr_find_cpu(target));
1591         if (cs) {
1592             run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL);
1593             return H_SUCCESS;
1594         }
1595         return H_PARAMETER;
1596     }
1597 }
1598 
1599 /* Returns either a logical PVR or zero if none was found */
1600 static uint32_t cas_check_pvr(PowerPCCPU *cpu, uint32_t max_compat,
1601                               target_ulong *addr, bool *raw_mode_supported)
1602 {
1603     bool explicit_match = false; /* Matched the CPU's real PVR */
1604     uint32_t best_compat = 0;
1605     int i;
1606 
1607     /*
1608      * We scan the supplied table of PVRs looking for two things
1609      *   1. Is our real CPU PVR in the list?
1610      *   2. What's the "best" listed logical PVR
1611      */
1612     for (i = 0; i < 512; ++i) {
1613         uint32_t pvr, pvr_mask;
1614 
1615         pvr_mask = ldl_be_phys(&address_space_memory, *addr);
1616         pvr = ldl_be_phys(&address_space_memory, *addr + 4);
1617         *addr += 8;
1618 
1619         if (~pvr_mask & pvr) {
1620             break; /* Terminator record */
1621         }
1622 
1623         if ((cpu->env.spr[SPR_PVR] & pvr_mask) == (pvr & pvr_mask)) {
1624             explicit_match = true;
1625         } else {
1626             if (ppc_check_compat(cpu, pvr, best_compat, max_compat)) {
1627                 best_compat = pvr;
1628             }
1629         }
1630     }
1631 
1632     *raw_mode_supported = explicit_match;
1633 
1634     /* Parsing finished */
1635     trace_spapr_cas_pvr(cpu->compat_pvr, explicit_match, best_compat);
1636 
1637     return best_compat;
1638 }
1639 
1640 static
1641 target_ulong do_client_architecture_support(PowerPCCPU *cpu,
1642                                             SpaprMachineState *spapr,
1643                                             target_ulong vec,
1644                                             target_ulong fdt_bufsize)
1645 {
1646     target_ulong ov_table; /* Working address in data buffer */
1647     uint32_t cas_pvr;
1648     SpaprOptionVector *ov1_guest, *ov5_guest;
1649     bool guest_radix;
1650     bool raw_mode_supported = false;
1651     bool guest_xive;
1652     CPUState *cs;
1653     void *fdt;
1654     uint32_t max_compat = spapr->max_compat_pvr;
1655 
1656     /* CAS is supposed to be called early when only the boot vCPU is active. */
1657     CPU_FOREACH(cs) {
1658         if (cs == CPU(cpu)) {
1659             continue;
1660         }
1661         if (!cs->halted) {
1662             warn_report("guest has multiple active vCPUs at CAS, which is not allowed");
1663             return H_MULTI_THREADS_ACTIVE;
1664         }
1665     }
1666 
1667     cas_pvr = cas_check_pvr(cpu, max_compat, &vec, &raw_mode_supported);
1668     if (!cas_pvr && (!raw_mode_supported || max_compat)) {
1669         /*
1670          * We couldn't find a suitable compatibility mode, and either
1671          * the guest doesn't support "raw" mode for this CPU, or "raw"
1672          * mode is disabled because a maximum compat mode is set.
1673          */
1674         error_report("Couldn't negotiate a suitable PVR during CAS");
1675         return H_HARDWARE;
1676     }
1677 
1678     /* Update CPUs */
1679     if (cpu->compat_pvr != cas_pvr) {
1680         Error *local_err = NULL;
1681 
1682         if (ppc_set_compat_all(cas_pvr, &local_err) < 0) {
1683             /* We fail to set compat mode (likely because running with KVM PR),
1684              * but maybe we can fallback to raw mode if the guest supports it.
1685              */
1686             if (!raw_mode_supported) {
1687                 error_report_err(local_err);
1688                 return H_HARDWARE;
1689             }
1690             error_free(local_err);
1691         }
1692     }
1693 
1694     /* For the future use: here @ov_table points to the first option vector */
1695     ov_table = vec;
1696 
1697     ov1_guest = spapr_ovec_parse_vector(ov_table, 1);
1698     if (!ov1_guest) {
1699         warn_report("guest didn't provide option vector 1");
1700         return H_PARAMETER;
1701     }
1702     ov5_guest = spapr_ovec_parse_vector(ov_table, 5);
1703     if (!ov5_guest) {
1704         spapr_ovec_cleanup(ov1_guest);
1705         warn_report("guest didn't provide option vector 5");
1706         return H_PARAMETER;
1707     }
1708     if (spapr_ovec_test(ov5_guest, OV5_MMU_BOTH)) {
1709         error_report("guest requested hash and radix MMU, which is invalid.");
1710         exit(EXIT_FAILURE);
1711     }
1712     if (spapr_ovec_test(ov5_guest, OV5_XIVE_BOTH)) {
1713         error_report("guest requested an invalid interrupt mode");
1714         exit(EXIT_FAILURE);
1715     }
1716 
1717     guest_radix = spapr_ovec_test(ov5_guest, OV5_MMU_RADIX_300);
1718 
1719     guest_xive = spapr_ovec_test(ov5_guest, OV5_XIVE_EXPLOIT);
1720 
1721     /*
1722      * HPT resizing is a bit of a special case, because when enabled
1723      * we assume an HPT guest will support it until it says it
1724      * doesn't, instead of assuming it won't support it until it says
1725      * it does.  Strictly speaking that approach could break for
1726      * guests which don't make a CAS call, but those are so old we
1727      * don't care about them.  Without that assumption we'd have to
1728      * make at least a temporary allocation of an HPT sized for max
1729      * memory, which could be impossibly difficult under KVM HV if
1730      * maxram is large.
1731      */
1732     if (!guest_radix && !spapr_ovec_test(ov5_guest, OV5_HPT_RESIZE)) {
1733         int maxshift = spapr_hpt_shift_for_ramsize(MACHINE(spapr)->maxram_size);
1734 
1735         if (spapr->resize_hpt == SPAPR_RESIZE_HPT_REQUIRED) {
1736             error_report(
1737                 "h_client_architecture_support: Guest doesn't support HPT resizing, but resize-hpt=required");
1738             exit(1);
1739         }
1740 
1741         if (spapr->htab_shift < maxshift) {
1742             /* Guest doesn't know about HPT resizing, so we
1743              * pre-emptively resize for the maximum permitted RAM.  At
1744              * the point this is called, nothing should have been
1745              * entered into the existing HPT */
1746             spapr_reallocate_hpt(spapr, maxshift, &error_fatal);
1747             push_sregs_to_kvm_pr(spapr);
1748         }
1749     }
1750 
1751     /* NOTE: there are actually a number of ov5 bits where input from the
1752      * guest is always zero, and the platform/QEMU enables them independently
1753      * of guest input. To model these properly we'd want some sort of mask,
1754      * but since they only currently apply to memory migration as defined
1755      * by LoPAPR 1.1, 14.5.4.8, which QEMU doesn't implement, we don't need
1756      * to worry about this for now.
1757      */
1758 
1759     /* full range of negotiated ov5 capabilities */
1760     spapr_ovec_intersect(spapr->ov5_cas, spapr->ov5, ov5_guest);
1761     spapr_ovec_cleanup(ov5_guest);
1762 
1763     if (guest_radix) {
1764         if (kvm_enabled() && !kvmppc_has_cap_mmu_radix()) {
1765             error_report("Guest requested unavailable MMU mode (radix).");
1766             exit(EXIT_FAILURE);
1767         }
1768     } else {
1769         if (kvm_enabled() && kvmppc_has_cap_mmu_radix()
1770             && !kvmppc_has_cap_mmu_hash_v3()) {
1771             error_report("Guest requested unavailable MMU mode (hash).");
1772             exit(EXIT_FAILURE);
1773         }
1774     }
1775     spapr->cas_pre_isa3_guest = !spapr_ovec_test(ov1_guest, OV1_PPC_3_00);
1776     spapr_ovec_cleanup(ov1_guest);
1777 
1778     /*
1779      * Ensure the guest asks for an interrupt mode we support;
1780      * otherwise terminate the boot.
1781      */
1782     if (guest_xive) {
1783         if (!spapr->irq->xive) {
1784             error_report(
1785 "Guest requested unavailable interrupt mode (XIVE), try the ic-mode=xive or ic-mode=dual machine property");
1786             exit(EXIT_FAILURE);
1787         }
1788     } else {
1789         if (!spapr->irq->xics) {
1790             error_report(
1791 "Guest requested unavailable interrupt mode (XICS), either don't set the ic-mode machine property or try ic-mode=xics or ic-mode=dual");
1792             exit(EXIT_FAILURE);
1793         }
1794     }
1795 
1796     spapr_irq_update_active_intc(spapr);
1797 
1798     /*
1799      * Process all pending hot-plug/unplug requests now. An updated full
1800      * rendered FDT will be returned to the guest.
1801      */
1802     spapr_drc_reset_all(spapr);
1803     spapr_clear_pending_hotplug_events(spapr);
1804 
1805     /*
1806      * If spapr_machine_reset() did not set up a HPT but one is necessary
1807      * (because the guest isn't going to use radix) then set it up here.
1808      */
1809     if ((spapr->patb_entry & PATE1_GR) && !guest_radix) {
1810         /* legacy hash or new hash: */
1811         spapr_setup_hpt(spapr);
1812     }
1813 
1814     fdt = spapr_build_fdt(spapr, false, fdt_bufsize);
1815 
1816     g_free(spapr->fdt_blob);
1817     spapr->fdt_size = fdt_totalsize(fdt);
1818     spapr->fdt_initial_size = spapr->fdt_size;
1819     spapr->fdt_blob = fdt;
1820 
1821     return H_SUCCESS;
1822 }
1823 
1824 static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
1825                                                   SpaprMachineState *spapr,
1826                                                   target_ulong opcode,
1827                                                   target_ulong *args)
1828 {
1829     target_ulong vec = ppc64_phys_to_real(args[0]);
1830     target_ulong fdt_buf = args[1];
1831     target_ulong fdt_bufsize = args[2];
1832     target_ulong ret;
1833     SpaprDeviceTreeUpdateHeader hdr = { .version_id = 1 };
1834 
1835     if (fdt_bufsize < sizeof(hdr)) {
1836         error_report("SLOF provided insufficient CAS buffer "
1837                      TARGET_FMT_lu " (min: %zu)", fdt_bufsize, sizeof(hdr));
1838         exit(EXIT_FAILURE);
1839     }
1840 
1841     fdt_bufsize -= sizeof(hdr);
1842 
1843     ret = do_client_architecture_support(cpu, spapr, vec, fdt_bufsize);
1844     if (ret == H_SUCCESS) {
1845         _FDT((fdt_pack(spapr->fdt_blob)));
1846         spapr->fdt_size = fdt_totalsize(spapr->fdt_blob);
1847         spapr->fdt_initial_size = spapr->fdt_size;
1848 
1849         cpu_physical_memory_write(fdt_buf, &hdr, sizeof(hdr));
1850         cpu_physical_memory_write(fdt_buf + sizeof(hdr), spapr->fdt_blob,
1851                                   spapr->fdt_size);
1852         trace_spapr_cas_continue(spapr->fdt_size + sizeof(hdr));
1853     }
1854 
1855     return ret;
1856 }
1857 
1858 static target_ulong h_get_cpu_characteristics(PowerPCCPU *cpu,
1859                                               SpaprMachineState *spapr,
1860                                               target_ulong opcode,
1861                                               target_ulong *args)
1862 {
1863     uint64_t characteristics = H_CPU_CHAR_HON_BRANCH_HINTS &
1864                                ~H_CPU_CHAR_THR_RECONF_TRIG;
1865     uint64_t behaviour = H_CPU_BEHAV_FAVOUR_SECURITY;
1866     uint8_t safe_cache = spapr_get_cap(spapr, SPAPR_CAP_CFPC);
1867     uint8_t safe_bounds_check = spapr_get_cap(spapr, SPAPR_CAP_SBBC);
1868     uint8_t safe_indirect_branch = spapr_get_cap(spapr, SPAPR_CAP_IBS);
1869     uint8_t count_cache_flush_assist = spapr_get_cap(spapr,
1870                                                      SPAPR_CAP_CCF_ASSIST);
1871 
1872     switch (safe_cache) {
1873     case SPAPR_CAP_WORKAROUND:
1874         characteristics |= H_CPU_CHAR_L1D_FLUSH_ORI30;
1875         characteristics |= H_CPU_CHAR_L1D_FLUSH_TRIG2;
1876         characteristics |= H_CPU_CHAR_L1D_THREAD_PRIV;
1877         behaviour |= H_CPU_BEHAV_L1D_FLUSH_PR;
1878         break;
1879     case SPAPR_CAP_FIXED:
1880         break;
1881     default: /* broken */
1882         assert(safe_cache == SPAPR_CAP_BROKEN);
1883         behaviour |= H_CPU_BEHAV_L1D_FLUSH_PR;
1884         break;
1885     }
1886 
1887     switch (safe_bounds_check) {
1888     case SPAPR_CAP_WORKAROUND:
1889         characteristics |= H_CPU_CHAR_SPEC_BAR_ORI31;
1890         behaviour |= H_CPU_BEHAV_BNDS_CHK_SPEC_BAR;
1891         break;
1892     case SPAPR_CAP_FIXED:
1893         break;
1894     default: /* broken */
1895         assert(safe_bounds_check == SPAPR_CAP_BROKEN);
1896         behaviour |= H_CPU_BEHAV_BNDS_CHK_SPEC_BAR;
1897         break;
1898     }
1899 
1900     switch (safe_indirect_branch) {
1901     case SPAPR_CAP_FIXED_NA:
1902         break;
1903     case SPAPR_CAP_FIXED_CCD:
1904         characteristics |= H_CPU_CHAR_CACHE_COUNT_DIS;
1905         break;
1906     case SPAPR_CAP_FIXED_IBS:
1907         characteristics |= H_CPU_CHAR_BCCTRL_SERIALISED;
1908         break;
1909     case SPAPR_CAP_WORKAROUND:
1910         behaviour |= H_CPU_BEHAV_FLUSH_COUNT_CACHE;
1911         if (count_cache_flush_assist) {
1912             characteristics |= H_CPU_CHAR_BCCTR_FLUSH_ASSIST;
1913         }
1914         break;
1915     default: /* broken */
1916         assert(safe_indirect_branch == SPAPR_CAP_BROKEN);
1917         break;
1918     }
1919 
1920     args[0] = characteristics;
1921     args[1] = behaviour;
1922     return H_SUCCESS;
1923 }
1924 
1925 static target_ulong h_update_dt(PowerPCCPU *cpu, SpaprMachineState *spapr,
1926                                 target_ulong opcode, target_ulong *args)
1927 {
1928     target_ulong dt = ppc64_phys_to_real(args[0]);
1929     struct fdt_header hdr = { 0 };
1930     unsigned cb;
1931     SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
1932     void *fdt;
1933 
1934     cpu_physical_memory_read(dt, &hdr, sizeof(hdr));
1935     cb = fdt32_to_cpu(hdr.totalsize);
1936 
1937     if (!smc->update_dt_enabled) {
1938         return H_SUCCESS;
1939     }
1940 
1941     /* Check that the fdt did not grow out of proportion */
1942     if (cb > spapr->fdt_initial_size * 2) {
1943         trace_spapr_update_dt_failed_size(spapr->fdt_initial_size, cb,
1944                                           fdt32_to_cpu(hdr.magic));
1945         return H_PARAMETER;
1946     }
1947 
1948     fdt = g_malloc0(cb);
1949     cpu_physical_memory_read(dt, fdt, cb);
1950 
1951     /* Check the fdt consistency */
1952     if (fdt_check_full(fdt, cb)) {
1953         trace_spapr_update_dt_failed_check(spapr->fdt_initial_size, cb,
1954                                            fdt32_to_cpu(hdr.magic));
1955         return H_PARAMETER;
1956     }
1957 
1958     g_free(spapr->fdt_blob);
1959     spapr->fdt_size = cb;
1960     spapr->fdt_blob = fdt;
1961     trace_spapr_update_dt(cb);
1962 
1963     return H_SUCCESS;
1964 }
1965 
1966 static spapr_hcall_fn papr_hypercall_table[(MAX_HCALL_OPCODE / 4) + 1];
1967 static spapr_hcall_fn kvmppc_hypercall_table[KVMPPC_HCALL_MAX - KVMPPC_HCALL_BASE + 1];
1968 static spapr_hcall_fn svm_hypercall_table[(SVM_HCALL_MAX - SVM_HCALL_BASE) / 4 + 1];
1969 
1970 void spapr_register_hypercall(target_ulong opcode, spapr_hcall_fn fn)
1971 {
1972     spapr_hcall_fn *slot;
1973 
1974     if (opcode <= MAX_HCALL_OPCODE) {
1975         assert((opcode & 0x3) == 0);
1976 
1977         slot = &papr_hypercall_table[opcode / 4];
1978     } else if (opcode >= SVM_HCALL_BASE && opcode <= SVM_HCALL_MAX) {
1979         /* we only have SVM-related hcall numbers assigned in multiples of 4 */
1980         assert((opcode & 0x3) == 0);
1981 
1982         slot = &svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
1983     } else {
1984         assert((opcode >= KVMPPC_HCALL_BASE) && (opcode <= KVMPPC_HCALL_MAX));
1985 
1986         slot = &kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
1987     }
1988 
1989     assert(!(*slot));
1990     *slot = fn;
1991 }
1992 
1993 target_ulong spapr_hypercall(PowerPCCPU *cpu, target_ulong opcode,
1994                              target_ulong *args)
1995 {
1996     SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1997 
1998     if ((opcode <= MAX_HCALL_OPCODE)
1999         && ((opcode & 0x3) == 0)) {
2000         spapr_hcall_fn fn = papr_hypercall_table[opcode / 4];
2001 
2002         if (fn) {
2003             return fn(cpu, spapr, opcode, args);
2004         }
2005     } else if ((opcode >= SVM_HCALL_BASE) &&
2006                (opcode <= SVM_HCALL_MAX)) {
2007         spapr_hcall_fn fn = svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
2008 
2009         if (fn) {
2010             return fn(cpu, spapr, opcode, args);
2011         }
2012     } else if ((opcode >= KVMPPC_HCALL_BASE) &&
2013                (opcode <= KVMPPC_HCALL_MAX)) {
2014         spapr_hcall_fn fn = kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
2015 
2016         if (fn) {
2017             return fn(cpu, spapr, opcode, args);
2018         }
2019     }
2020 
2021     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x" TARGET_FMT_lx "\n",
2022                   opcode);
2023     return H_FUNCTION;
2024 }
2025 
2026 static void hypercall_register_types(void)
2027 {
2028     /* hcall-pft */
2029     spapr_register_hypercall(H_ENTER, h_enter);
2030     spapr_register_hypercall(H_REMOVE, h_remove);
2031     spapr_register_hypercall(H_PROTECT, h_protect);
2032     spapr_register_hypercall(H_READ, h_read);
2033 
2034     /* hcall-bulk */
2035     spapr_register_hypercall(H_BULK_REMOVE, h_bulk_remove);
2036 
2037     /* hcall-hpt-resize */
2038     spapr_register_hypercall(H_RESIZE_HPT_PREPARE, h_resize_hpt_prepare);
2039     spapr_register_hypercall(H_RESIZE_HPT_COMMIT, h_resize_hpt_commit);
2040 
2041     /* hcall-splpar */
2042     spapr_register_hypercall(H_REGISTER_VPA, h_register_vpa);
2043     spapr_register_hypercall(H_CEDE, h_cede);
2044     spapr_register_hypercall(H_CONFER, h_confer);
2045     spapr_register_hypercall(H_PROD, h_prod);
2046 
2047     /* hcall-join */
2048     spapr_register_hypercall(H_JOIN, h_join);
2049 
2050     spapr_register_hypercall(H_SIGNAL_SYS_RESET, h_signal_sys_reset);
2051 
2052     /* processor register resource access h-calls */
2053     spapr_register_hypercall(H_SET_SPRG0, h_set_sprg0);
2054     spapr_register_hypercall(H_SET_DABR, h_set_dabr);
2055     spapr_register_hypercall(H_SET_XDABR, h_set_xdabr);
2056     spapr_register_hypercall(H_PAGE_INIT, h_page_init);
2057     spapr_register_hypercall(H_SET_MODE, h_set_mode);
2058 
2059     /* In Memory Table MMU h-calls */
2060     spapr_register_hypercall(H_CLEAN_SLB, h_clean_slb);
2061     spapr_register_hypercall(H_INVALIDATE_PID, h_invalidate_pid);
2062     spapr_register_hypercall(H_REGISTER_PROC_TBL, h_register_process_table);
2063 
2064     /* hcall-get-cpu-characteristics */
2065     spapr_register_hypercall(H_GET_CPU_CHARACTERISTICS,
2066                              h_get_cpu_characteristics);
2067 
2068     /* "debugger" hcalls (also used by SLOF). Note: We do -not- differenciate
2069      * here between the "CI" and the "CACHE" variants, they will use whatever
2070      * mapping attributes qemu is using. When using KVM, the kernel will
2071      * enforce the attributes more strongly
2072      */
2073     spapr_register_hypercall(H_LOGICAL_CI_LOAD, h_logical_load);
2074     spapr_register_hypercall(H_LOGICAL_CI_STORE, h_logical_store);
2075     spapr_register_hypercall(H_LOGICAL_CACHE_LOAD, h_logical_load);
2076     spapr_register_hypercall(H_LOGICAL_CACHE_STORE, h_logical_store);
2077     spapr_register_hypercall(H_LOGICAL_ICBI, h_logical_icbi);
2078     spapr_register_hypercall(H_LOGICAL_DCBF, h_logical_dcbf);
2079     spapr_register_hypercall(KVMPPC_H_LOGICAL_MEMOP, h_logical_memop);
2080 
2081     /* qemu/KVM-PPC specific hcalls */
2082     spapr_register_hypercall(KVMPPC_H_RTAS, h_rtas);
2083 
2084     /* ibm,client-architecture-support support */
2085     spapr_register_hypercall(KVMPPC_H_CAS, h_client_architecture_support);
2086 
2087     spapr_register_hypercall(KVMPPC_H_UPDATE_DT, h_update_dt);
2088 }
2089 
2090 type_init(hypercall_register_types)
2091