xref: /openbmc/qemu/hw/intc/spapr_xive.c (revision 605994e5b7d17dcc275465b4f89816d29105b238)
1 /*
2  * QEMU PowerPC sPAPR XIVE interrupt controller model
3  *
4  * Copyright (c) 2017-2018, IBM Corporation.
5  *
6  * This code is licensed under the GPL version 2 or later. See the
7  * COPYING file in the top-level directory.
8  */
9 
10 #include "qemu/osdep.h"
11 #include "qemu/log.h"
12 #include "qemu/module.h"
13 #include "qapi/error.h"
14 #include "qemu/error-report.h"
15 #include "target/ppc/cpu.h"
16 #include "sysemu/cpus.h"
17 #include "sysemu/reset.h"
18 #include "migration/vmstate.h"
19 #include "monitor/monitor.h"
20 #include "hw/ppc/fdt.h"
21 #include "hw/ppc/spapr.h"
22 #include "hw/ppc/spapr_cpu_core.h"
23 #include "hw/ppc/spapr_xive.h"
24 #include "hw/ppc/xive.h"
25 #include "hw/ppc/xive_regs.h"
26 #include "hw/qdev-properties.h"
27 
28 /*
29  * XIVE Virtualization Controller BAR and Thread Managment BAR that we
30  * use for the ESB pages and the TIMA pages
31  */
32 #define SPAPR_XIVE_VC_BASE   0x0006010000000000ull
33 #define SPAPR_XIVE_TM_BASE   0x0006030203180000ull
34 
35 /*
36  * The allocation of VP blocks is a complex operation in OPAL and the
37  * VP identifiers have a relation with the number of HW chips, the
38  * size of the VP blocks, VP grouping, etc. The QEMU sPAPR XIVE
39  * controller model does not have the same constraints and can use a
40  * simple mapping scheme of the CPU vcpu_id
41  *
42  * These identifiers are never returned to the OS.
43  */
44 
45 #define SPAPR_XIVE_NVT_BASE 0x400
46 
47 /*
48  * sPAPR NVT and END indexing helpers
49  */
50 static uint32_t spapr_xive_nvt_to_target(uint8_t nvt_blk, uint32_t nvt_idx)
51 {
52     return nvt_idx - SPAPR_XIVE_NVT_BASE;
53 }
54 
55 static void spapr_xive_cpu_to_nvt(PowerPCCPU *cpu,
56                                   uint8_t *out_nvt_blk, uint32_t *out_nvt_idx)
57 {
58     assert(cpu);
59 
60     if (out_nvt_blk) {
61         *out_nvt_blk = SPAPR_XIVE_BLOCK_ID;
62     }
63 
64     if (out_nvt_blk) {
65         *out_nvt_idx = SPAPR_XIVE_NVT_BASE + cpu->vcpu_id;
66     }
67 }
68 
69 static int spapr_xive_target_to_nvt(uint32_t target,
70                                     uint8_t *out_nvt_blk, uint32_t *out_nvt_idx)
71 {
72     PowerPCCPU *cpu = spapr_find_cpu(target);
73 
74     if (!cpu) {
75         return -1;
76     }
77 
78     spapr_xive_cpu_to_nvt(cpu, out_nvt_blk, out_nvt_idx);
79     return 0;
80 }
81 
82 /*
83  * sPAPR END indexing uses a simple mapping of the CPU vcpu_id, 8
84  * priorities per CPU
85  */
86 int spapr_xive_end_to_target(uint8_t end_blk, uint32_t end_idx,
87                              uint32_t *out_server, uint8_t *out_prio)
88 {
89 
90     assert(end_blk == SPAPR_XIVE_BLOCK_ID);
91 
92     if (out_server) {
93         *out_server = end_idx >> 3;
94     }
95 
96     if (out_prio) {
97         *out_prio = end_idx & 0x7;
98     }
99     return 0;
100 }
101 
102 static void spapr_xive_cpu_to_end(PowerPCCPU *cpu, uint8_t prio,
103                                   uint8_t *out_end_blk, uint32_t *out_end_idx)
104 {
105     assert(cpu);
106 
107     if (out_end_blk) {
108         *out_end_blk = SPAPR_XIVE_BLOCK_ID;
109     }
110 
111     if (out_end_idx) {
112         *out_end_idx = (cpu->vcpu_id << 3) + prio;
113     }
114 }
115 
116 static int spapr_xive_target_to_end(uint32_t target, uint8_t prio,
117                                     uint8_t *out_end_blk, uint32_t *out_end_idx)
118 {
119     PowerPCCPU *cpu = spapr_find_cpu(target);
120 
121     if (!cpu) {
122         return -1;
123     }
124 
125     spapr_xive_cpu_to_end(cpu, prio, out_end_blk, out_end_idx);
126     return 0;
127 }
128 
129 /*
130  * On sPAPR machines, use a simplified output for the XIVE END
131  * structure dumping only the information related to the OS EQ.
132  */
133 static void spapr_xive_end_pic_print_info(SpaprXive *xive, XiveEND *end,
134                                           Monitor *mon)
135 {
136     uint64_t qaddr_base = xive_end_qaddr(end);
137     uint32_t qindex = xive_get_field32(END_W1_PAGE_OFF, end->w1);
138     uint32_t qgen = xive_get_field32(END_W1_GENERATION, end->w1);
139     uint32_t qsize = xive_get_field32(END_W0_QSIZE, end->w0);
140     uint32_t qentries = 1 << (qsize + 10);
141     uint32_t nvt = xive_get_field32(END_W6_NVT_INDEX, end->w6);
142     uint8_t priority = xive_get_field32(END_W7_F0_PRIORITY, end->w7);
143 
144     monitor_printf(mon, "%3d/%d % 6d/%5d @%"PRIx64" ^%d",
145                    spapr_xive_nvt_to_target(0, nvt),
146                    priority, qindex, qentries, qaddr_base, qgen);
147 
148     xive_end_queue_pic_print_info(end, 6, mon);
149 }
150 
151 void spapr_xive_pic_print_info(SpaprXive *xive, Monitor *mon)
152 {
153     XiveSource *xsrc = &xive->source;
154     int i;
155 
156     if (kvm_irqchip_in_kernel()) {
157         Error *local_err = NULL;
158 
159         kvmppc_xive_synchronize_state(xive, &local_err);
160         if (local_err) {
161             error_report_err(local_err);
162             return;
163         }
164     }
165 
166     monitor_printf(mon, "  LISN         PQ    EISN     CPU/PRIO EQ\n");
167 
168     for (i = 0; i < xive->nr_irqs; i++) {
169         uint8_t pq = xive_source_esb_get(xsrc, i);
170         XiveEAS *eas = &xive->eat[i];
171 
172         if (!xive_eas_is_valid(eas)) {
173             continue;
174         }
175 
176         monitor_printf(mon, "  %08x %s %c%c%c %s %08x ", i,
177                        xive_source_irq_is_lsi(xsrc, i) ? "LSI" : "MSI",
178                        pq & XIVE_ESB_VAL_P ? 'P' : '-',
179                        pq & XIVE_ESB_VAL_Q ? 'Q' : '-',
180                        xsrc->status[i] & XIVE_STATUS_ASSERTED ? 'A' : ' ',
181                        xive_eas_is_masked(eas) ? "M" : " ",
182                        (int) xive_get_field64(EAS_END_DATA, eas->w));
183 
184         if (!xive_eas_is_masked(eas)) {
185             uint32_t end_idx = xive_get_field64(EAS_END_INDEX, eas->w);
186             XiveEND *end;
187 
188             assert(end_idx < xive->nr_ends);
189             end = &xive->endt[end_idx];
190 
191             if (xive_end_is_valid(end)) {
192                 spapr_xive_end_pic_print_info(xive, end, mon);
193             }
194         }
195         monitor_printf(mon, "\n");
196     }
197 }
198 
199 void spapr_xive_mmio_set_enabled(SpaprXive *xive, bool enable)
200 {
201     memory_region_set_enabled(&xive->source.esb_mmio, enable);
202     memory_region_set_enabled(&xive->tm_mmio, enable);
203 
204     /* Disable the END ESBs until a guest OS makes use of them */
205     memory_region_set_enabled(&xive->end_source.esb_mmio, false);
206 }
207 
208 /*
209  * When a Virtual Processor is scheduled to run on a HW thread, the
210  * hypervisor pushes its identifier in the OS CAM line. Emulate the
211  * same behavior under QEMU.
212  */
213 void spapr_xive_set_tctx_os_cam(XiveTCTX *tctx)
214 {
215     uint8_t  nvt_blk;
216     uint32_t nvt_idx;
217     uint32_t nvt_cam;
218 
219     spapr_xive_cpu_to_nvt(POWERPC_CPU(tctx->cs), &nvt_blk, &nvt_idx);
220 
221     nvt_cam = cpu_to_be32(TM_QW1W2_VO | xive_nvt_cam_line(nvt_blk, nvt_idx));
222     memcpy(&tctx->regs[TM_QW1_OS + TM_WORD2], &nvt_cam, 4);
223 }
224 
225 static void spapr_xive_end_reset(XiveEND *end)
226 {
227     memset(end, 0, sizeof(*end));
228 
229     /* switch off the escalation and notification ESBs */
230     end->w1 = cpu_to_be32(END_W1_ESe_Q | END_W1_ESn_Q);
231 }
232 
233 static void spapr_xive_reset(void *dev)
234 {
235     SpaprXive *xive = SPAPR_XIVE(dev);
236     int i;
237 
238     /*
239      * The XiveSource has its own reset handler, which mask off all
240      * IRQs (!P|Q)
241      */
242 
243     /* Mask all valid EASs in the IRQ number space. */
244     for (i = 0; i < xive->nr_irqs; i++) {
245         XiveEAS *eas = &xive->eat[i];
246         if (xive_eas_is_valid(eas)) {
247             eas->w = cpu_to_be64(EAS_VALID | EAS_MASKED);
248         } else {
249             eas->w = 0;
250         }
251     }
252 
253     /* Clear all ENDs */
254     for (i = 0; i < xive->nr_ends; i++) {
255         spapr_xive_end_reset(&xive->endt[i]);
256     }
257 }
258 
259 static void spapr_xive_instance_init(Object *obj)
260 {
261     SpaprXive *xive = SPAPR_XIVE(obj);
262 
263     object_initialize_child(obj, "source", &xive->source, sizeof(xive->source),
264                             TYPE_XIVE_SOURCE, &error_abort, NULL);
265 
266     object_initialize_child(obj, "end_source", &xive->end_source,
267                             sizeof(xive->end_source), TYPE_XIVE_END_SOURCE,
268                             &error_abort, NULL);
269 
270     /* Not connected to the KVM XIVE device */
271     xive->fd = -1;
272 }
273 
274 static void spapr_xive_realize(DeviceState *dev, Error **errp)
275 {
276     SpaprXive *xive = SPAPR_XIVE(dev);
277     XiveSource *xsrc = &xive->source;
278     XiveENDSource *end_xsrc = &xive->end_source;
279     Error *local_err = NULL;
280 
281     if (!xive->nr_irqs) {
282         error_setg(errp, "Number of interrupt needs to be greater 0");
283         return;
284     }
285 
286     if (!xive->nr_ends) {
287         error_setg(errp, "Number of interrupt needs to be greater 0");
288         return;
289     }
290 
291     /*
292      * Initialize the internal sources, for IPIs and virtual devices.
293      */
294     object_property_set_int(OBJECT(xsrc), xive->nr_irqs, "nr-irqs",
295                             &error_fatal);
296     object_property_add_const_link(OBJECT(xsrc), "xive", OBJECT(xive),
297                                    &error_fatal);
298     object_property_set_bool(OBJECT(xsrc), true, "realized", &local_err);
299     if (local_err) {
300         error_propagate(errp, local_err);
301         return;
302     }
303     sysbus_init_mmio(SYS_BUS_DEVICE(xive), &xsrc->esb_mmio);
304 
305     /*
306      * Initialize the END ESB source
307      */
308     object_property_set_int(OBJECT(end_xsrc), xive->nr_irqs, "nr-ends",
309                             &error_fatal);
310     object_property_add_const_link(OBJECT(end_xsrc), "xive", OBJECT(xive),
311                                    &error_fatal);
312     object_property_set_bool(OBJECT(end_xsrc), true, "realized", &local_err);
313     if (local_err) {
314         error_propagate(errp, local_err);
315         return;
316     }
317     sysbus_init_mmio(SYS_BUS_DEVICE(xive), &end_xsrc->esb_mmio);
318 
319     /* Set the mapping address of the END ESB pages after the source ESBs */
320     xive->end_base = xive->vc_base + (1ull << xsrc->esb_shift) * xsrc->nr_irqs;
321 
322     /*
323      * Allocate the routing tables
324      */
325     xive->eat = g_new0(XiveEAS, xive->nr_irqs);
326     xive->endt = g_new0(XiveEND, xive->nr_ends);
327 
328     xive->nodename = g_strdup_printf("interrupt-controller@%" PRIx64,
329                            xive->tm_base + XIVE_TM_USER_PAGE * (1 << TM_SHIFT));
330 
331     qemu_register_reset(spapr_xive_reset, dev);
332 
333     /* TIMA initialization */
334     memory_region_init_io(&xive->tm_mmio, OBJECT(xive), &xive_tm_ops, xive,
335                           "xive.tima", 4ull << TM_SHIFT);
336     sysbus_init_mmio(SYS_BUS_DEVICE(xive), &xive->tm_mmio);
337 
338     /*
339      * Map all regions. These will be enabled or disabled at reset and
340      * can also be overridden by KVM memory regions if active
341      */
342     sysbus_mmio_map(SYS_BUS_DEVICE(xive), 0, xive->vc_base);
343     sysbus_mmio_map(SYS_BUS_DEVICE(xive), 1, xive->end_base);
344     sysbus_mmio_map(SYS_BUS_DEVICE(xive), 2, xive->tm_base);
345 }
346 
347 static int spapr_xive_get_eas(XiveRouter *xrtr, uint8_t eas_blk,
348                               uint32_t eas_idx, XiveEAS *eas)
349 {
350     SpaprXive *xive = SPAPR_XIVE(xrtr);
351 
352     if (eas_idx >= xive->nr_irqs) {
353         return -1;
354     }
355 
356     *eas = xive->eat[eas_idx];
357     return 0;
358 }
359 
360 static int spapr_xive_get_end(XiveRouter *xrtr,
361                               uint8_t end_blk, uint32_t end_idx, XiveEND *end)
362 {
363     SpaprXive *xive = SPAPR_XIVE(xrtr);
364 
365     if (end_idx >= xive->nr_ends) {
366         return -1;
367     }
368 
369     memcpy(end, &xive->endt[end_idx], sizeof(XiveEND));
370     return 0;
371 }
372 
373 static int spapr_xive_write_end(XiveRouter *xrtr, uint8_t end_blk,
374                                 uint32_t end_idx, XiveEND *end,
375                                 uint8_t word_number)
376 {
377     SpaprXive *xive = SPAPR_XIVE(xrtr);
378 
379     if (end_idx >= xive->nr_ends) {
380         return -1;
381     }
382 
383     memcpy(&xive->endt[end_idx], end, sizeof(XiveEND));
384     return 0;
385 }
386 
387 static int spapr_xive_get_nvt(XiveRouter *xrtr,
388                               uint8_t nvt_blk, uint32_t nvt_idx, XiveNVT *nvt)
389 {
390     uint32_t vcpu_id = spapr_xive_nvt_to_target(nvt_blk, nvt_idx);
391     PowerPCCPU *cpu = spapr_find_cpu(vcpu_id);
392 
393     if (!cpu) {
394         /* TODO: should we assert() if we can find a NVT ? */
395         return -1;
396     }
397 
398     /*
399      * sPAPR does not maintain a NVT table. Return that the NVT is
400      * valid if we have found a matching CPU
401      */
402     nvt->w0 = cpu_to_be32(NVT_W0_VALID);
403     return 0;
404 }
405 
406 static int spapr_xive_write_nvt(XiveRouter *xrtr, uint8_t nvt_blk,
407                                 uint32_t nvt_idx, XiveNVT *nvt,
408                                 uint8_t word_number)
409 {
410     /*
411      * We don't need to write back to the NVTs because the sPAPR
412      * machine should never hit a non-scheduled NVT. It should never
413      * get called.
414      */
415     g_assert_not_reached();
416 }
417 
418 static XiveTCTX *spapr_xive_get_tctx(XiveRouter *xrtr, CPUState *cs)
419 {
420     PowerPCCPU *cpu = POWERPC_CPU(cs);
421 
422     return spapr_cpu_state(cpu)->tctx;
423 }
424 
425 static const VMStateDescription vmstate_spapr_xive_end = {
426     .name = TYPE_SPAPR_XIVE "/end",
427     .version_id = 1,
428     .minimum_version_id = 1,
429     .fields = (VMStateField []) {
430         VMSTATE_UINT32(w0, XiveEND),
431         VMSTATE_UINT32(w1, XiveEND),
432         VMSTATE_UINT32(w2, XiveEND),
433         VMSTATE_UINT32(w3, XiveEND),
434         VMSTATE_UINT32(w4, XiveEND),
435         VMSTATE_UINT32(w5, XiveEND),
436         VMSTATE_UINT32(w6, XiveEND),
437         VMSTATE_UINT32(w7, XiveEND),
438         VMSTATE_END_OF_LIST()
439     },
440 };
441 
442 static const VMStateDescription vmstate_spapr_xive_eas = {
443     .name = TYPE_SPAPR_XIVE "/eas",
444     .version_id = 1,
445     .minimum_version_id = 1,
446     .fields = (VMStateField []) {
447         VMSTATE_UINT64(w, XiveEAS),
448         VMSTATE_END_OF_LIST()
449     },
450 };
451 
452 static int vmstate_spapr_xive_pre_save(void *opaque)
453 {
454     if (kvm_irqchip_in_kernel()) {
455         return kvmppc_xive_pre_save(SPAPR_XIVE(opaque));
456     }
457 
458     return 0;
459 }
460 
461 /*
462  * Called by the sPAPR IRQ backend 'post_load' method at the machine
463  * level.
464  */
465 static int spapr_xive_post_load(SpaprInterruptController *intc, int version_id)
466 {
467     if (kvm_irqchip_in_kernel()) {
468         return kvmppc_xive_post_load(SPAPR_XIVE(intc), version_id);
469     }
470 
471     return 0;
472 }
473 
474 static const VMStateDescription vmstate_spapr_xive = {
475     .name = TYPE_SPAPR_XIVE,
476     .version_id = 1,
477     .minimum_version_id = 1,
478     .pre_save = vmstate_spapr_xive_pre_save,
479     .post_load = NULL, /* handled at the machine level */
480     .fields = (VMStateField[]) {
481         VMSTATE_UINT32_EQUAL(nr_irqs, SpaprXive, NULL),
482         VMSTATE_STRUCT_VARRAY_POINTER_UINT32(eat, SpaprXive, nr_irqs,
483                                      vmstate_spapr_xive_eas, XiveEAS),
484         VMSTATE_STRUCT_VARRAY_POINTER_UINT32(endt, SpaprXive, nr_ends,
485                                              vmstate_spapr_xive_end, XiveEND),
486         VMSTATE_END_OF_LIST()
487     },
488 };
489 
490 static int spapr_xive_claim_irq(SpaprInterruptController *intc, int lisn,
491                                 bool lsi, Error **errp)
492 {
493     SpaprXive *xive = SPAPR_XIVE(intc);
494     XiveSource *xsrc = &xive->source;
495 
496     assert(lisn < xive->nr_irqs);
497 
498     if (xive_eas_is_valid(&xive->eat[lisn])) {
499         error_setg(errp, "IRQ %d is not free", lisn);
500         return -EBUSY;
501     }
502 
503     /*
504      * Set default values when allocating an IRQ number
505      */
506     xive->eat[lisn].w |= cpu_to_be64(EAS_VALID | EAS_MASKED);
507     if (lsi) {
508         xive_source_irq_set_lsi(xsrc, lisn);
509     }
510 
511     if (kvm_irqchip_in_kernel()) {
512         return kvmppc_xive_source_reset_one(xsrc, lisn, errp);
513     }
514 
515     return 0;
516 }
517 
518 static void spapr_xive_free_irq(SpaprInterruptController *intc, int lisn)
519 {
520     SpaprXive *xive = SPAPR_XIVE(intc);
521     assert(lisn < xive->nr_irqs);
522 
523     xive->eat[lisn].w &= cpu_to_be64(~EAS_VALID);
524 }
525 
526 static Property spapr_xive_properties[] = {
527     DEFINE_PROP_UINT32("nr-irqs", SpaprXive, nr_irqs, 0),
528     DEFINE_PROP_UINT32("nr-ends", SpaprXive, nr_ends, 0),
529     DEFINE_PROP_UINT64("vc-base", SpaprXive, vc_base, SPAPR_XIVE_VC_BASE),
530     DEFINE_PROP_UINT64("tm-base", SpaprXive, tm_base, SPAPR_XIVE_TM_BASE),
531     DEFINE_PROP_END_OF_LIST(),
532 };
533 
534 static int spapr_xive_cpu_intc_create(SpaprInterruptController *intc,
535                                       PowerPCCPU *cpu, Error **errp)
536 {
537     SpaprXive *xive = SPAPR_XIVE(intc);
538     Object *obj;
539     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
540 
541     obj = xive_tctx_create(OBJECT(cpu), XIVE_ROUTER(xive), errp);
542     if (!obj) {
543         return -1;
544     }
545 
546     spapr_cpu->tctx = XIVE_TCTX(obj);
547 
548     /*
549      * (TCG) Early setting the OS CAM line for hotplugged CPUs as they
550      * don't beneficiate from the reset of the XIVE IRQ backend
551      */
552     spapr_xive_set_tctx_os_cam(spapr_cpu->tctx);
553     return 0;
554 }
555 
556 static void spapr_xive_set_irq(SpaprInterruptController *intc, int irq, int val)
557 {
558     SpaprXive *xive = SPAPR_XIVE(intc);
559 
560     if (kvm_irqchip_in_kernel()) {
561         kvmppc_xive_source_set_irq(&xive->source, irq, val);
562     } else {
563         xive_source_set_irq(&xive->source, irq, val);
564     }
565 }
566 
567 static void spapr_xive_print_info(SpaprInterruptController *intc, Monitor *mon)
568 {
569     SpaprXive *xive = SPAPR_XIVE(intc);
570     CPUState *cs;
571 
572     CPU_FOREACH(cs) {
573         PowerPCCPU *cpu = POWERPC_CPU(cs);
574 
575         xive_tctx_pic_print_info(spapr_cpu_state(cpu)->tctx, mon);
576     }
577 
578     spapr_xive_pic_print_info(xive, mon);
579 }
580 
581 static void spapr_xive_dt(SpaprInterruptController *intc, uint32_t nr_servers,
582                           void *fdt, uint32_t phandle)
583 {
584     SpaprXive *xive = SPAPR_XIVE(intc);
585     int node;
586     uint64_t timas[2 * 2];
587     /* Interrupt number ranges for the IPIs */
588     uint32_t lisn_ranges[] = {
589         cpu_to_be32(0),
590         cpu_to_be32(nr_servers),
591     };
592     /*
593      * EQ size - the sizes of pages supported by the system 4K, 64K,
594      * 2M, 16M. We only advertise 64K for the moment.
595      */
596     uint32_t eq_sizes[] = {
597         cpu_to_be32(16), /* 64K */
598     };
599     /*
600      * The following array is in sync with the reserved priorities
601      * defined by the 'spapr_xive_priority_is_reserved' routine.
602      */
603     uint32_t plat_res_int_priorities[] = {
604         cpu_to_be32(7),    /* start */
605         cpu_to_be32(0xf8), /* count */
606     };
607 
608     /* Thread Interrupt Management Area : User (ring 3) and OS (ring 2) */
609     timas[0] = cpu_to_be64(xive->tm_base +
610                            XIVE_TM_USER_PAGE * (1ull << TM_SHIFT));
611     timas[1] = cpu_to_be64(1ull << TM_SHIFT);
612     timas[2] = cpu_to_be64(xive->tm_base +
613                            XIVE_TM_OS_PAGE * (1ull << TM_SHIFT));
614     timas[3] = cpu_to_be64(1ull << TM_SHIFT);
615 
616     _FDT(node = fdt_add_subnode(fdt, 0, xive->nodename));
617 
618     _FDT(fdt_setprop_string(fdt, node, "device_type", "power-ivpe"));
619     _FDT(fdt_setprop(fdt, node, "reg", timas, sizeof(timas)));
620 
621     _FDT(fdt_setprop_string(fdt, node, "compatible", "ibm,power-ivpe"));
622     _FDT(fdt_setprop(fdt, node, "ibm,xive-eq-sizes", eq_sizes,
623                      sizeof(eq_sizes)));
624     _FDT(fdt_setprop(fdt, node, "ibm,xive-lisn-ranges", lisn_ranges,
625                      sizeof(lisn_ranges)));
626 
627     /* For Linux to link the LSIs to the interrupt controller. */
628     _FDT(fdt_setprop(fdt, node, "interrupt-controller", NULL, 0));
629     _FDT(fdt_setprop_cell(fdt, node, "#interrupt-cells", 2));
630 
631     /* For SLOF */
632     _FDT(fdt_setprop_cell(fdt, node, "linux,phandle", phandle));
633     _FDT(fdt_setprop_cell(fdt, node, "phandle", phandle));
634 
635     /*
636      * The "ibm,plat-res-int-priorities" property defines the priority
637      * ranges reserved by the hypervisor
638      */
639     _FDT(fdt_setprop(fdt, 0, "ibm,plat-res-int-priorities",
640                      plat_res_int_priorities, sizeof(plat_res_int_priorities)));
641 }
642 
643 static int spapr_xive_activate(SpaprInterruptController *intc, Error **errp)
644 {
645     SpaprXive *xive = SPAPR_XIVE(intc);
646     CPUState *cs;
647 
648     CPU_FOREACH(cs) {
649         PowerPCCPU *cpu = POWERPC_CPU(cs);
650 
651         /* (TCG) Set the OS CAM line of the thread interrupt context. */
652         spapr_xive_set_tctx_os_cam(spapr_cpu_state(cpu)->tctx);
653     }
654 
655     if (kvm_enabled()) {
656         int rc = spapr_irq_init_kvm(kvmppc_xive_connect, intc, errp);
657         if (rc < 0) {
658             return rc;
659         }
660     }
661 
662     /* Activate the XIVE MMIOs */
663     spapr_xive_mmio_set_enabled(xive, true);
664 
665     return 0;
666 }
667 
668 static void spapr_xive_deactivate(SpaprInterruptController *intc)
669 {
670     SpaprXive *xive = SPAPR_XIVE(intc);
671 
672     spapr_xive_mmio_set_enabled(xive, false);
673 
674     if (kvm_irqchip_in_kernel()) {
675         kvmppc_xive_disconnect(intc);
676     }
677 }
678 
679 static void spapr_xive_class_init(ObjectClass *klass, void *data)
680 {
681     DeviceClass *dc = DEVICE_CLASS(klass);
682     XiveRouterClass *xrc = XIVE_ROUTER_CLASS(klass);
683     SpaprInterruptControllerClass *sicc = SPAPR_INTC_CLASS(klass);
684 
685     dc->desc    = "sPAPR XIVE Interrupt Controller";
686     dc->props   = spapr_xive_properties;
687     dc->realize = spapr_xive_realize;
688     dc->vmsd    = &vmstate_spapr_xive;
689 
690     xrc->get_eas = spapr_xive_get_eas;
691     xrc->get_end = spapr_xive_get_end;
692     xrc->write_end = spapr_xive_write_end;
693     xrc->get_nvt = spapr_xive_get_nvt;
694     xrc->write_nvt = spapr_xive_write_nvt;
695     xrc->get_tctx = spapr_xive_get_tctx;
696 
697     sicc->activate = spapr_xive_activate;
698     sicc->deactivate = spapr_xive_deactivate;
699     sicc->cpu_intc_create = spapr_xive_cpu_intc_create;
700     sicc->claim_irq = spapr_xive_claim_irq;
701     sicc->free_irq = spapr_xive_free_irq;
702     sicc->set_irq = spapr_xive_set_irq;
703     sicc->print_info = spapr_xive_print_info;
704     sicc->dt = spapr_xive_dt;
705     sicc->post_load = spapr_xive_post_load;
706 }
707 
708 static const TypeInfo spapr_xive_info = {
709     .name = TYPE_SPAPR_XIVE,
710     .parent = TYPE_XIVE_ROUTER,
711     .instance_init = spapr_xive_instance_init,
712     .instance_size = sizeof(SpaprXive),
713     .class_init = spapr_xive_class_init,
714     .interfaces = (InterfaceInfo[]) {
715         { TYPE_SPAPR_INTC },
716         { }
717     },
718 };
719 
720 static void spapr_xive_register_types(void)
721 {
722     type_register_static(&spapr_xive_info);
723 }
724 
725 type_init(spapr_xive_register_types)
726 
727 /*
728  * XIVE hcalls
729  *
730  * The terminology used by the XIVE hcalls is the following :
731  *
732  *   TARGET vCPU number
733  *   EQ     Event Queue assigned by OS to receive event data
734  *   ESB    page for source interrupt management
735  *   LISN   Logical Interrupt Source Number identifying a source in the
736  *          machine
737  *   EISN   Effective Interrupt Source Number used by guest OS to
738  *          identify source in the guest
739  *
740  * The EAS, END, NVT structures are not exposed.
741  */
742 
743 /*
744  * Linux hosts under OPAL reserve priority 7 for their own escalation
745  * interrupts (DD2.X POWER9). So we only allow the guest to use
746  * priorities [0..6].
747  */
748 static bool spapr_xive_priority_is_reserved(uint8_t priority)
749 {
750     switch (priority) {
751     case 0 ... 6:
752         return false;
753     case 7: /* OPAL escalation queue */
754     default:
755         return true;
756     }
757 }
758 
759 /*
760  * The H_INT_GET_SOURCE_INFO hcall() is used to obtain the logical
761  * real address of the MMIO page through which the Event State Buffer
762  * entry associated with the value of the "lisn" parameter is managed.
763  *
764  * Parameters:
765  * Input
766  * - R4: "flags"
767  *         Bits 0-63 reserved
768  * - R5: "lisn" is per "interrupts", "interrupt-map", or
769  *       "ibm,xive-lisn-ranges" properties, or as returned by the
770  *       ibm,query-interrupt-source-number RTAS call, or as returned
771  *       by the H_ALLOCATE_VAS_WINDOW hcall
772  *
773  * Output
774  * - R4: "flags"
775  *         Bits 0-59: Reserved
776  *         Bit 60: H_INT_ESB must be used for Event State Buffer
777  *                 management
778  *         Bit 61: 1 == LSI  0 == MSI
779  *         Bit 62: the full function page supports trigger
780  *         Bit 63: Store EOI Supported
781  * - R5: Logical Real address of full function Event State Buffer
782  *       management page, -1 if H_INT_ESB hcall flag is set to 1.
783  * - R6: Logical Real Address of trigger only Event State Buffer
784  *       management page or -1.
785  * - R7: Power of 2 page size for the ESB management pages returned in
786  *       R5 and R6.
787  */
788 
789 #define SPAPR_XIVE_SRC_H_INT_ESB     PPC_BIT(60) /* ESB manage with H_INT_ESB */
790 #define SPAPR_XIVE_SRC_LSI           PPC_BIT(61) /* Virtual LSI type */
791 #define SPAPR_XIVE_SRC_TRIGGER       PPC_BIT(62) /* Trigger and management
792                                                     on same page */
793 #define SPAPR_XIVE_SRC_STORE_EOI     PPC_BIT(63) /* Store EOI support */
794 
795 static target_ulong h_int_get_source_info(PowerPCCPU *cpu,
796                                           SpaprMachineState *spapr,
797                                           target_ulong opcode,
798                                           target_ulong *args)
799 {
800     SpaprXive *xive = spapr->xive;
801     XiveSource *xsrc = &xive->source;
802     target_ulong flags  = args[0];
803     target_ulong lisn   = args[1];
804 
805     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
806         return H_FUNCTION;
807     }
808 
809     if (flags) {
810         return H_PARAMETER;
811     }
812 
813     if (lisn >= xive->nr_irqs) {
814         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Unknown LISN " TARGET_FMT_lx "\n",
815                       lisn);
816         return H_P2;
817     }
818 
819     if (!xive_eas_is_valid(&xive->eat[lisn])) {
820         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Invalid LISN " TARGET_FMT_lx "\n",
821                       lisn);
822         return H_P2;
823     }
824 
825     /*
826      * All sources are emulated under the main XIVE object and share
827      * the same characteristics.
828      */
829     args[0] = 0;
830     if (!xive_source_esb_has_2page(xsrc)) {
831         args[0] |= SPAPR_XIVE_SRC_TRIGGER;
832     }
833     if (xsrc->esb_flags & XIVE_SRC_STORE_EOI) {
834         args[0] |= SPAPR_XIVE_SRC_STORE_EOI;
835     }
836 
837     /*
838      * Force the use of the H_INT_ESB hcall in case of an LSI
839      * interrupt. This is necessary under KVM to re-trigger the
840      * interrupt if the level is still asserted
841      */
842     if (xive_source_irq_is_lsi(xsrc, lisn)) {
843         args[0] |= SPAPR_XIVE_SRC_H_INT_ESB | SPAPR_XIVE_SRC_LSI;
844     }
845 
846     if (!(args[0] & SPAPR_XIVE_SRC_H_INT_ESB)) {
847         args[1] = xive->vc_base + xive_source_esb_mgmt(xsrc, lisn);
848     } else {
849         args[1] = -1;
850     }
851 
852     if (xive_source_esb_has_2page(xsrc) &&
853         !(args[0] & SPAPR_XIVE_SRC_H_INT_ESB)) {
854         args[2] = xive->vc_base + xive_source_esb_page(xsrc, lisn);
855     } else {
856         args[2] = -1;
857     }
858 
859     if (xive_source_esb_has_2page(xsrc)) {
860         args[3] = xsrc->esb_shift - 1;
861     } else {
862         args[3] = xsrc->esb_shift;
863     }
864 
865     return H_SUCCESS;
866 }
867 
868 /*
869  * The H_INT_SET_SOURCE_CONFIG hcall() is used to assign a Logical
870  * Interrupt Source to a target. The Logical Interrupt Source is
871  * designated with the "lisn" parameter and the target is designated
872  * with the "target" and "priority" parameters.  Upon return from the
873  * hcall(), no additional interrupts will be directed to the old EQ.
874  *
875  * Parameters:
876  * Input:
877  * - R4: "flags"
878  *         Bits 0-61: Reserved
879  *         Bit 62: set the "eisn" in the EAS
880  *         Bit 63: masks the interrupt source in the hardware interrupt
881  *       control structure. An interrupt masked by this mechanism will
882  *       be dropped, but it's source state bits will still be
883  *       set. There is no race-free way of unmasking and restoring the
884  *       source. Thus this should only be used in interrupts that are
885  *       also masked at the source, and only in cases where the
886  *       interrupt is not meant to be used for a large amount of time
887  *       because no valid target exists for it for example
888  * - R5: "lisn" is per "interrupts", "interrupt-map", or
889  *       "ibm,xive-lisn-ranges" properties, or as returned by the
890  *       ibm,query-interrupt-source-number RTAS call, or as returned by
891  *       the H_ALLOCATE_VAS_WINDOW hcall
892  * - R6: "target" is per "ibm,ppc-interrupt-server#s" or
893  *       "ibm,ppc-interrupt-gserver#s"
894  * - R7: "priority" is a valid priority not in
895  *       "ibm,plat-res-int-priorities"
896  * - R8: "eisn" is the guest EISN associated with the "lisn"
897  *
898  * Output:
899  * - None
900  */
901 
902 #define SPAPR_XIVE_SRC_SET_EISN PPC_BIT(62)
903 #define SPAPR_XIVE_SRC_MASK     PPC_BIT(63)
904 
905 static target_ulong h_int_set_source_config(PowerPCCPU *cpu,
906                                             SpaprMachineState *spapr,
907                                             target_ulong opcode,
908                                             target_ulong *args)
909 {
910     SpaprXive *xive = spapr->xive;
911     XiveEAS eas, new_eas;
912     target_ulong flags    = args[0];
913     target_ulong lisn     = args[1];
914     target_ulong target   = args[2];
915     target_ulong priority = args[3];
916     target_ulong eisn     = args[4];
917     uint8_t end_blk;
918     uint32_t end_idx;
919 
920     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
921         return H_FUNCTION;
922     }
923 
924     if (flags & ~(SPAPR_XIVE_SRC_SET_EISN | SPAPR_XIVE_SRC_MASK)) {
925         return H_PARAMETER;
926     }
927 
928     if (lisn >= xive->nr_irqs) {
929         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Unknown LISN " TARGET_FMT_lx "\n",
930                       lisn);
931         return H_P2;
932     }
933 
934     eas = xive->eat[lisn];
935     if (!xive_eas_is_valid(&eas)) {
936         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Invalid LISN " TARGET_FMT_lx "\n",
937                       lisn);
938         return H_P2;
939     }
940 
941     /* priority 0xff is used to reset the EAS */
942     if (priority == 0xff) {
943         new_eas.w = cpu_to_be64(EAS_VALID | EAS_MASKED);
944         goto out;
945     }
946 
947     if (flags & SPAPR_XIVE_SRC_MASK) {
948         new_eas.w = eas.w | cpu_to_be64(EAS_MASKED);
949     } else {
950         new_eas.w = eas.w & cpu_to_be64(~EAS_MASKED);
951     }
952 
953     if (spapr_xive_priority_is_reserved(priority)) {
954         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: priority " TARGET_FMT_ld
955                       " is reserved\n", priority);
956         return H_P4;
957     }
958 
959     /*
960      * Validate that "target" is part of the list of threads allocated
961      * to the partition. For that, find the END corresponding to the
962      * target.
963      */
964     if (spapr_xive_target_to_end(target, priority, &end_blk, &end_idx)) {
965         return H_P3;
966     }
967 
968     new_eas.w = xive_set_field64(EAS_END_BLOCK, new_eas.w, end_blk);
969     new_eas.w = xive_set_field64(EAS_END_INDEX, new_eas.w, end_idx);
970 
971     if (flags & SPAPR_XIVE_SRC_SET_EISN) {
972         new_eas.w = xive_set_field64(EAS_END_DATA, new_eas.w, eisn);
973     }
974 
975     if (kvm_irqchip_in_kernel()) {
976         Error *local_err = NULL;
977 
978         kvmppc_xive_set_source_config(xive, lisn, &new_eas, &local_err);
979         if (local_err) {
980             error_report_err(local_err);
981             return H_HARDWARE;
982         }
983     }
984 
985 out:
986     xive->eat[lisn] = new_eas;
987     return H_SUCCESS;
988 }
989 
990 /*
991  * The H_INT_GET_SOURCE_CONFIG hcall() is used to determine to which
992  * target/priority pair is assigned to the specified Logical Interrupt
993  * Source.
994  *
995  * Parameters:
996  * Input:
997  * - R4: "flags"
998  *         Bits 0-63 Reserved
999  * - R5: "lisn" is per "interrupts", "interrupt-map", or
1000  *       "ibm,xive-lisn-ranges" properties, or as returned by the
1001  *       ibm,query-interrupt-source-number RTAS call, or as
1002  *       returned by the H_ALLOCATE_VAS_WINDOW hcall
1003  *
1004  * Output:
1005  * - R4: Target to which the specified Logical Interrupt Source is
1006  *       assigned
1007  * - R5: Priority to which the specified Logical Interrupt Source is
1008  *       assigned
1009  * - R6: EISN for the specified Logical Interrupt Source (this will be
1010  *       equivalent to the LISN if not changed by H_INT_SET_SOURCE_CONFIG)
1011  */
1012 static target_ulong h_int_get_source_config(PowerPCCPU *cpu,
1013                                             SpaprMachineState *spapr,
1014                                             target_ulong opcode,
1015                                             target_ulong *args)
1016 {
1017     SpaprXive *xive = spapr->xive;
1018     target_ulong flags = args[0];
1019     target_ulong lisn = args[1];
1020     XiveEAS eas;
1021     XiveEND *end;
1022     uint8_t nvt_blk;
1023     uint32_t end_idx, nvt_idx;
1024 
1025     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1026         return H_FUNCTION;
1027     }
1028 
1029     if (flags) {
1030         return H_PARAMETER;
1031     }
1032 
1033     if (lisn >= xive->nr_irqs) {
1034         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Unknown LISN " TARGET_FMT_lx "\n",
1035                       lisn);
1036         return H_P2;
1037     }
1038 
1039     eas = xive->eat[lisn];
1040     if (!xive_eas_is_valid(&eas)) {
1041         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Invalid LISN " TARGET_FMT_lx "\n",
1042                       lisn);
1043         return H_P2;
1044     }
1045 
1046     /* EAS_END_BLOCK is unused on sPAPR */
1047     end_idx = xive_get_field64(EAS_END_INDEX, eas.w);
1048 
1049     assert(end_idx < xive->nr_ends);
1050     end = &xive->endt[end_idx];
1051 
1052     nvt_blk = xive_get_field32(END_W6_NVT_BLOCK, end->w6);
1053     nvt_idx = xive_get_field32(END_W6_NVT_INDEX, end->w6);
1054     args[0] = spapr_xive_nvt_to_target(nvt_blk, nvt_idx);
1055 
1056     if (xive_eas_is_masked(&eas)) {
1057         args[1] = 0xff;
1058     } else {
1059         args[1] = xive_get_field32(END_W7_F0_PRIORITY, end->w7);
1060     }
1061 
1062     args[2] = xive_get_field64(EAS_END_DATA, eas.w);
1063 
1064     return H_SUCCESS;
1065 }
1066 
1067 /*
1068  * The H_INT_GET_QUEUE_INFO hcall() is used to get the logical real
1069  * address of the notification management page associated with the
1070  * specified target and priority.
1071  *
1072  * Parameters:
1073  * Input:
1074  * - R4: "flags"
1075  *         Bits 0-63 Reserved
1076  * - R5: "target" is per "ibm,ppc-interrupt-server#s" or
1077  *       "ibm,ppc-interrupt-gserver#s"
1078  * - R6: "priority" is a valid priority not in
1079  *       "ibm,plat-res-int-priorities"
1080  *
1081  * Output:
1082  * - R4: Logical real address of notification page
1083  * - R5: Power of 2 page size of the notification page
1084  */
1085 static target_ulong h_int_get_queue_info(PowerPCCPU *cpu,
1086                                          SpaprMachineState *spapr,
1087                                          target_ulong opcode,
1088                                          target_ulong *args)
1089 {
1090     SpaprXive *xive = spapr->xive;
1091     XiveENDSource *end_xsrc = &xive->end_source;
1092     target_ulong flags = args[0];
1093     target_ulong target = args[1];
1094     target_ulong priority = args[2];
1095     XiveEND *end;
1096     uint8_t end_blk;
1097     uint32_t end_idx;
1098 
1099     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1100         return H_FUNCTION;
1101     }
1102 
1103     if (flags) {
1104         return H_PARAMETER;
1105     }
1106 
1107     /*
1108      * H_STATE should be returned if a H_INT_RESET is in progress.
1109      * This is not needed when running the emulation under QEMU
1110      */
1111 
1112     if (spapr_xive_priority_is_reserved(priority)) {
1113         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: priority " TARGET_FMT_ld
1114                       " is reserved\n", priority);
1115         return H_P3;
1116     }
1117 
1118     /*
1119      * Validate that "target" is part of the list of threads allocated
1120      * to the partition. For that, find the END corresponding to the
1121      * target.
1122      */
1123     if (spapr_xive_target_to_end(target, priority, &end_blk, &end_idx)) {
1124         return H_P2;
1125     }
1126 
1127     assert(end_idx < xive->nr_ends);
1128     end = &xive->endt[end_idx];
1129 
1130     args[0] = xive->end_base + (1ull << (end_xsrc->esb_shift + 1)) * end_idx;
1131     if (xive_end_is_enqueue(end)) {
1132         args[1] = xive_get_field32(END_W0_QSIZE, end->w0) + 12;
1133     } else {
1134         args[1] = 0;
1135     }
1136 
1137     return H_SUCCESS;
1138 }
1139 
1140 /*
1141  * The H_INT_SET_QUEUE_CONFIG hcall() is used to set or reset a EQ for
1142  * a given "target" and "priority".  It is also used to set the
1143  * notification config associated with the EQ.  An EQ size of 0 is
1144  * used to reset the EQ config for a given target and priority. If
1145  * resetting the EQ config, the END associated with the given "target"
1146  * and "priority" will be changed to disable queueing.
1147  *
1148  * Upon return from the hcall(), no additional interrupts will be
1149  * directed to the old EQ (if one was set). The old EQ (if one was
1150  * set) should be investigated for interrupts that occurred prior to
1151  * or during the hcall().
1152  *
1153  * Parameters:
1154  * Input:
1155  * - R4: "flags"
1156  *         Bits 0-62: Reserved
1157  *         Bit 63: Unconditional Notify (n) per the XIVE spec
1158  * - R5: "target" is per "ibm,ppc-interrupt-server#s" or
1159  *       "ibm,ppc-interrupt-gserver#s"
1160  * - R6: "priority" is a valid priority not in
1161  *       "ibm,plat-res-int-priorities"
1162  * - R7: "eventQueue": The logical real address of the start of the EQ
1163  * - R8: "eventQueueSize": The power of 2 EQ size per "ibm,xive-eq-sizes"
1164  *
1165  * Output:
1166  * - None
1167  */
1168 
1169 #define SPAPR_XIVE_END_ALWAYS_NOTIFY PPC_BIT(63)
1170 
1171 static target_ulong h_int_set_queue_config(PowerPCCPU *cpu,
1172                                            SpaprMachineState *spapr,
1173                                            target_ulong opcode,
1174                                            target_ulong *args)
1175 {
1176     SpaprXive *xive = spapr->xive;
1177     target_ulong flags = args[0];
1178     target_ulong target = args[1];
1179     target_ulong priority = args[2];
1180     target_ulong qpage = args[3];
1181     target_ulong qsize = args[4];
1182     XiveEND end;
1183     uint8_t end_blk, nvt_blk;
1184     uint32_t end_idx, nvt_idx;
1185 
1186     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1187         return H_FUNCTION;
1188     }
1189 
1190     if (flags & ~SPAPR_XIVE_END_ALWAYS_NOTIFY) {
1191         return H_PARAMETER;
1192     }
1193 
1194     /*
1195      * H_STATE should be returned if a H_INT_RESET is in progress.
1196      * This is not needed when running the emulation under QEMU
1197      */
1198 
1199     if (spapr_xive_priority_is_reserved(priority)) {
1200         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: priority " TARGET_FMT_ld
1201                       " is reserved\n", priority);
1202         return H_P3;
1203     }
1204 
1205     /*
1206      * Validate that "target" is part of the list of threads allocated
1207      * to the partition. For that, find the END corresponding to the
1208      * target.
1209      */
1210 
1211     if (spapr_xive_target_to_end(target, priority, &end_blk, &end_idx)) {
1212         return H_P2;
1213     }
1214 
1215     assert(end_idx < xive->nr_ends);
1216     memcpy(&end, &xive->endt[end_idx], sizeof(XiveEND));
1217 
1218     switch (qsize) {
1219     case 12:
1220     case 16:
1221     case 21:
1222     case 24:
1223         if (!QEMU_IS_ALIGNED(qpage, 1ul << qsize)) {
1224             qemu_log_mask(LOG_GUEST_ERROR, "XIVE: EQ @0x%" HWADDR_PRIx
1225                           " is not naturally aligned with %" HWADDR_PRIx "\n",
1226                           qpage, (hwaddr)1 << qsize);
1227             return H_P4;
1228         }
1229         end.w2 = cpu_to_be32((qpage >> 32) & 0x0fffffff);
1230         end.w3 = cpu_to_be32(qpage & 0xffffffff);
1231         end.w0 |= cpu_to_be32(END_W0_ENQUEUE);
1232         end.w0 = xive_set_field32(END_W0_QSIZE, end.w0, qsize - 12);
1233         break;
1234     case 0:
1235         /* reset queue and disable queueing */
1236         spapr_xive_end_reset(&end);
1237         goto out;
1238 
1239     default:
1240         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: invalid EQ size %"PRIx64"\n",
1241                       qsize);
1242         return H_P5;
1243     }
1244 
1245     if (qsize) {
1246         hwaddr plen = 1 << qsize;
1247         void *eq;
1248 
1249         /*
1250          * Validate the guest EQ. We should also check that the queue
1251          * has been zeroed by the OS.
1252          */
1253         eq = address_space_map(CPU(cpu)->as, qpage, &plen, true,
1254                                MEMTXATTRS_UNSPECIFIED);
1255         if (plen != 1 << qsize) {
1256             qemu_log_mask(LOG_GUEST_ERROR, "XIVE: failed to map EQ @0x%"
1257                           HWADDR_PRIx "\n", qpage);
1258             return H_P4;
1259         }
1260         address_space_unmap(CPU(cpu)->as, eq, plen, true, plen);
1261     }
1262 
1263     /* "target" should have been validated above */
1264     if (spapr_xive_target_to_nvt(target, &nvt_blk, &nvt_idx)) {
1265         g_assert_not_reached();
1266     }
1267 
1268     /*
1269      * Ensure the priority and target are correctly set (they will not
1270      * be right after allocation)
1271      */
1272     end.w6 = xive_set_field32(END_W6_NVT_BLOCK, 0ul, nvt_blk) |
1273         xive_set_field32(END_W6_NVT_INDEX, 0ul, nvt_idx);
1274     end.w7 = xive_set_field32(END_W7_F0_PRIORITY, 0ul, priority);
1275 
1276     if (flags & SPAPR_XIVE_END_ALWAYS_NOTIFY) {
1277         end.w0 |= cpu_to_be32(END_W0_UCOND_NOTIFY);
1278     } else {
1279         end.w0 &= cpu_to_be32((uint32_t)~END_W0_UCOND_NOTIFY);
1280     }
1281 
1282     /*
1283      * The generation bit for the END starts at 1 and The END page
1284      * offset counter starts at 0.
1285      */
1286     end.w1 = cpu_to_be32(END_W1_GENERATION) |
1287         xive_set_field32(END_W1_PAGE_OFF, 0ul, 0ul);
1288     end.w0 |= cpu_to_be32(END_W0_VALID);
1289 
1290     /*
1291      * TODO: issue syncs required to ensure all in-flight interrupts
1292      * are complete on the old END
1293      */
1294 
1295 out:
1296     if (kvm_irqchip_in_kernel()) {
1297         Error *local_err = NULL;
1298 
1299         kvmppc_xive_set_queue_config(xive, end_blk, end_idx, &end, &local_err);
1300         if (local_err) {
1301             error_report_err(local_err);
1302             return H_HARDWARE;
1303         }
1304     }
1305 
1306     /* Update END */
1307     memcpy(&xive->endt[end_idx], &end, sizeof(XiveEND));
1308     return H_SUCCESS;
1309 }
1310 
1311 /*
1312  * The H_INT_GET_QUEUE_CONFIG hcall() is used to get a EQ for a given
1313  * target and priority.
1314  *
1315  * Parameters:
1316  * Input:
1317  * - R4: "flags"
1318  *         Bits 0-62: Reserved
1319  *         Bit 63: Debug: Return debug data
1320  * - R5: "target" is per "ibm,ppc-interrupt-server#s" or
1321  *       "ibm,ppc-interrupt-gserver#s"
1322  * - R6: "priority" is a valid priority not in
1323  *       "ibm,plat-res-int-priorities"
1324  *
1325  * Output:
1326  * - R4: "flags":
1327  *       Bits 0-61: Reserved
1328  *       Bit 62: The value of Event Queue Generation Number (g) per
1329  *              the XIVE spec if "Debug" = 1
1330  *       Bit 63: The value of Unconditional Notify (n) per the XIVE spec
1331  * - R5: The logical real address of the start of the EQ
1332  * - R6: The power of 2 EQ size per "ibm,xive-eq-sizes"
1333  * - R7: The value of Event Queue Offset Counter per XIVE spec
1334  *       if "Debug" = 1, else 0
1335  *
1336  */
1337 
1338 #define SPAPR_XIVE_END_DEBUG     PPC_BIT(63)
1339 
1340 static target_ulong h_int_get_queue_config(PowerPCCPU *cpu,
1341                                            SpaprMachineState *spapr,
1342                                            target_ulong opcode,
1343                                            target_ulong *args)
1344 {
1345     SpaprXive *xive = spapr->xive;
1346     target_ulong flags = args[0];
1347     target_ulong target = args[1];
1348     target_ulong priority = args[2];
1349     XiveEND *end;
1350     uint8_t end_blk;
1351     uint32_t end_idx;
1352 
1353     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1354         return H_FUNCTION;
1355     }
1356 
1357     if (flags & ~SPAPR_XIVE_END_DEBUG) {
1358         return H_PARAMETER;
1359     }
1360 
1361     /*
1362      * H_STATE should be returned if a H_INT_RESET is in progress.
1363      * This is not needed when running the emulation under QEMU
1364      */
1365 
1366     if (spapr_xive_priority_is_reserved(priority)) {
1367         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: priority " TARGET_FMT_ld
1368                       " is reserved\n", priority);
1369         return H_P3;
1370     }
1371 
1372     /*
1373      * Validate that "target" is part of the list of threads allocated
1374      * to the partition. For that, find the END corresponding to the
1375      * target.
1376      */
1377     if (spapr_xive_target_to_end(target, priority, &end_blk, &end_idx)) {
1378         return H_P2;
1379     }
1380 
1381     assert(end_idx < xive->nr_ends);
1382     end = &xive->endt[end_idx];
1383 
1384     args[0] = 0;
1385     if (xive_end_is_notify(end)) {
1386         args[0] |= SPAPR_XIVE_END_ALWAYS_NOTIFY;
1387     }
1388 
1389     if (xive_end_is_enqueue(end)) {
1390         args[1] = xive_end_qaddr(end);
1391         args[2] = xive_get_field32(END_W0_QSIZE, end->w0) + 12;
1392     } else {
1393         args[1] = 0;
1394         args[2] = 0;
1395     }
1396 
1397     if (kvm_irqchip_in_kernel()) {
1398         Error *local_err = NULL;
1399 
1400         kvmppc_xive_get_queue_config(xive, end_blk, end_idx, end, &local_err);
1401         if (local_err) {
1402             error_report_err(local_err);
1403             return H_HARDWARE;
1404         }
1405     }
1406 
1407     /* TODO: do we need any locking on the END ? */
1408     if (flags & SPAPR_XIVE_END_DEBUG) {
1409         /* Load the event queue generation number into the return flags */
1410         args[0] |= (uint64_t)xive_get_field32(END_W1_GENERATION, end->w1) << 62;
1411 
1412         /* Load R7 with the event queue offset counter */
1413         args[3] = xive_get_field32(END_W1_PAGE_OFF, end->w1);
1414     } else {
1415         args[3] = 0;
1416     }
1417 
1418     return H_SUCCESS;
1419 }
1420 
1421 /*
1422  * The H_INT_SET_OS_REPORTING_LINE hcall() is used to set the
1423  * reporting cache line pair for the calling thread.  The reporting
1424  * cache lines will contain the OS interrupt context when the OS
1425  * issues a CI store byte to @TIMA+0xC10 to acknowledge the OS
1426  * interrupt. The reporting cache lines can be reset by inputting -1
1427  * in "reportingLine".  Issuing the CI store byte without reporting
1428  * cache lines registered will result in the data not being accessible
1429  * to the OS.
1430  *
1431  * Parameters:
1432  * Input:
1433  * - R4: "flags"
1434  *         Bits 0-63: Reserved
1435  * - R5: "reportingLine": The logical real address of the reporting cache
1436  *       line pair
1437  *
1438  * Output:
1439  * - None
1440  */
1441 static target_ulong h_int_set_os_reporting_line(PowerPCCPU *cpu,
1442                                                 SpaprMachineState *spapr,
1443                                                 target_ulong opcode,
1444                                                 target_ulong *args)
1445 {
1446     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1447         return H_FUNCTION;
1448     }
1449 
1450     /*
1451      * H_STATE should be returned if a H_INT_RESET is in progress.
1452      * This is not needed when running the emulation under QEMU
1453      */
1454 
1455     /* TODO: H_INT_SET_OS_REPORTING_LINE */
1456     return H_FUNCTION;
1457 }
1458 
1459 /*
1460  * The H_INT_GET_OS_REPORTING_LINE hcall() is used to get the logical
1461  * real address of the reporting cache line pair set for the input
1462  * "target".  If no reporting cache line pair has been set, -1 is
1463  * returned.
1464  *
1465  * Parameters:
1466  * Input:
1467  * - R4: "flags"
1468  *         Bits 0-63: Reserved
1469  * - R5: "target" is per "ibm,ppc-interrupt-server#s" or
1470  *       "ibm,ppc-interrupt-gserver#s"
1471  * - R6: "reportingLine": The logical real address of the reporting
1472  *        cache line pair
1473  *
1474  * Output:
1475  * - R4: The logical real address of the reporting line if set, else -1
1476  */
1477 static target_ulong h_int_get_os_reporting_line(PowerPCCPU *cpu,
1478                                                 SpaprMachineState *spapr,
1479                                                 target_ulong opcode,
1480                                                 target_ulong *args)
1481 {
1482     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1483         return H_FUNCTION;
1484     }
1485 
1486     /*
1487      * H_STATE should be returned if a H_INT_RESET is in progress.
1488      * This is not needed when running the emulation under QEMU
1489      */
1490 
1491     /* TODO: H_INT_GET_OS_REPORTING_LINE */
1492     return H_FUNCTION;
1493 }
1494 
1495 /*
1496  * The H_INT_ESB hcall() is used to issue a load or store to the ESB
1497  * page for the input "lisn".  This hcall is only supported for LISNs
1498  * that have the ESB hcall flag set to 1 when returned from hcall()
1499  * H_INT_GET_SOURCE_INFO.
1500  *
1501  * Parameters:
1502  * Input:
1503  * - R4: "flags"
1504  *         Bits 0-62: Reserved
1505  *         bit 63: Store: Store=1, store operation, else load operation
1506  * - R5: "lisn" is per "interrupts", "interrupt-map", or
1507  *       "ibm,xive-lisn-ranges" properties, or as returned by the
1508  *       ibm,query-interrupt-source-number RTAS call, or as
1509  *       returned by the H_ALLOCATE_VAS_WINDOW hcall
1510  * - R6: "esbOffset" is the offset into the ESB page for the load or
1511  *       store operation
1512  * - R7: "storeData" is the data to write for a store operation
1513  *
1514  * Output:
1515  * - R4: The value of the load if load operation, else -1
1516  */
1517 
1518 #define SPAPR_XIVE_ESB_STORE PPC_BIT(63)
1519 
1520 static target_ulong h_int_esb(PowerPCCPU *cpu,
1521                               SpaprMachineState *spapr,
1522                               target_ulong opcode,
1523                               target_ulong *args)
1524 {
1525     SpaprXive *xive = spapr->xive;
1526     XiveEAS eas;
1527     target_ulong flags  = args[0];
1528     target_ulong lisn   = args[1];
1529     target_ulong offset = args[2];
1530     target_ulong data   = args[3];
1531     hwaddr mmio_addr;
1532     XiveSource *xsrc = &xive->source;
1533 
1534     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1535         return H_FUNCTION;
1536     }
1537 
1538     if (flags & ~SPAPR_XIVE_ESB_STORE) {
1539         return H_PARAMETER;
1540     }
1541 
1542     if (lisn >= xive->nr_irqs) {
1543         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Unknown LISN " TARGET_FMT_lx "\n",
1544                       lisn);
1545         return H_P2;
1546     }
1547 
1548     eas = xive->eat[lisn];
1549     if (!xive_eas_is_valid(&eas)) {
1550         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Invalid LISN " TARGET_FMT_lx "\n",
1551                       lisn);
1552         return H_P2;
1553     }
1554 
1555     if (offset > (1ull << xsrc->esb_shift)) {
1556         return H_P3;
1557     }
1558 
1559     if (kvm_irqchip_in_kernel()) {
1560         args[0] = kvmppc_xive_esb_rw(xsrc, lisn, offset, data,
1561                                      flags & SPAPR_XIVE_ESB_STORE);
1562     } else {
1563         mmio_addr = xive->vc_base + xive_source_esb_mgmt(xsrc, lisn) + offset;
1564 
1565         if (dma_memory_rw(&address_space_memory, mmio_addr, &data, 8,
1566                           (flags & SPAPR_XIVE_ESB_STORE))) {
1567             qemu_log_mask(LOG_GUEST_ERROR, "XIVE: failed to access ESB @0x%"
1568                           HWADDR_PRIx "\n", mmio_addr);
1569             return H_HARDWARE;
1570         }
1571         args[0] = (flags & SPAPR_XIVE_ESB_STORE) ? -1 : data;
1572     }
1573     return H_SUCCESS;
1574 }
1575 
1576 /*
1577  * The H_INT_SYNC hcall() is used to issue hardware syncs that will
1578  * ensure any in flight events for the input lisn are in the event
1579  * queue.
1580  *
1581  * Parameters:
1582  * Input:
1583  * - R4: "flags"
1584  *         Bits 0-63: Reserved
1585  * - R5: "lisn" is per "interrupts", "interrupt-map", or
1586  *       "ibm,xive-lisn-ranges" properties, or as returned by the
1587  *       ibm,query-interrupt-source-number RTAS call, or as
1588  *       returned by the H_ALLOCATE_VAS_WINDOW hcall
1589  *
1590  * Output:
1591  * - None
1592  */
1593 static target_ulong h_int_sync(PowerPCCPU *cpu,
1594                                SpaprMachineState *spapr,
1595                                target_ulong opcode,
1596                                target_ulong *args)
1597 {
1598     SpaprXive *xive = spapr->xive;
1599     XiveEAS eas;
1600     target_ulong flags = args[0];
1601     target_ulong lisn = args[1];
1602 
1603     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1604         return H_FUNCTION;
1605     }
1606 
1607     if (flags) {
1608         return H_PARAMETER;
1609     }
1610 
1611     if (lisn >= xive->nr_irqs) {
1612         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Unknown LISN " TARGET_FMT_lx "\n",
1613                       lisn);
1614         return H_P2;
1615     }
1616 
1617     eas = xive->eat[lisn];
1618     if (!xive_eas_is_valid(&eas)) {
1619         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Invalid LISN " TARGET_FMT_lx "\n",
1620                       lisn);
1621         return H_P2;
1622     }
1623 
1624     /*
1625      * H_STATE should be returned if a H_INT_RESET is in progress.
1626      * This is not needed when running the emulation under QEMU
1627      */
1628 
1629     /*
1630      * This is not real hardware. Nothing to be done unless when
1631      * under KVM
1632      */
1633 
1634     if (kvm_irqchip_in_kernel()) {
1635         Error *local_err = NULL;
1636 
1637         kvmppc_xive_sync_source(xive, lisn, &local_err);
1638         if (local_err) {
1639             error_report_err(local_err);
1640             return H_HARDWARE;
1641         }
1642     }
1643     return H_SUCCESS;
1644 }
1645 
1646 /*
1647  * The H_INT_RESET hcall() is used to reset all of the partition's
1648  * interrupt exploitation structures to their initial state.  This
1649  * means losing all previously set interrupt state set via
1650  * H_INT_SET_SOURCE_CONFIG and H_INT_SET_QUEUE_CONFIG.
1651  *
1652  * Parameters:
1653  * Input:
1654  * - R4: "flags"
1655  *         Bits 0-63: Reserved
1656  *
1657  * Output:
1658  * - None
1659  */
1660 static target_ulong h_int_reset(PowerPCCPU *cpu,
1661                                 SpaprMachineState *spapr,
1662                                 target_ulong opcode,
1663                                 target_ulong *args)
1664 {
1665     SpaprXive *xive = spapr->xive;
1666     target_ulong flags   = args[0];
1667 
1668     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1669         return H_FUNCTION;
1670     }
1671 
1672     if (flags) {
1673         return H_PARAMETER;
1674     }
1675 
1676     device_reset(DEVICE(xive));
1677 
1678     if (kvm_irqchip_in_kernel()) {
1679         Error *local_err = NULL;
1680 
1681         kvmppc_xive_reset(xive, &local_err);
1682         if (local_err) {
1683             error_report_err(local_err);
1684             return H_HARDWARE;
1685         }
1686     }
1687     return H_SUCCESS;
1688 }
1689 
1690 void spapr_xive_hcall_init(SpaprMachineState *spapr)
1691 {
1692     spapr_register_hypercall(H_INT_GET_SOURCE_INFO, h_int_get_source_info);
1693     spapr_register_hypercall(H_INT_SET_SOURCE_CONFIG, h_int_set_source_config);
1694     spapr_register_hypercall(H_INT_GET_SOURCE_CONFIG, h_int_get_source_config);
1695     spapr_register_hypercall(H_INT_GET_QUEUE_INFO, h_int_get_queue_info);
1696     spapr_register_hypercall(H_INT_SET_QUEUE_CONFIG, h_int_set_queue_config);
1697     spapr_register_hypercall(H_INT_GET_QUEUE_CONFIG, h_int_get_queue_config);
1698     spapr_register_hypercall(H_INT_SET_OS_REPORTING_LINE,
1699                              h_int_set_os_reporting_line);
1700     spapr_register_hypercall(H_INT_GET_OS_REPORTING_LINE,
1701                              h_int_get_os_reporting_line);
1702     spapr_register_hypercall(H_INT_ESB, h_int_esb);
1703     spapr_register_hypercall(H_INT_SYNC, h_int_sync);
1704     spapr_register_hypercall(H_INT_RESET, h_int_reset);
1705 }
1706