xref: /qemu/hw/intc/spapr_xive.c (revision ab9056ff)
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 static void spapr_xive_end_reset(XiveEND *end)
209 {
210     memset(end, 0, sizeof(*end));
211 
212     /* switch off the escalation and notification ESBs */
213     end->w1 = cpu_to_be32(END_W1_ESe_Q | END_W1_ESn_Q);
214 }
215 
216 static void spapr_xive_reset(void *dev)
217 {
218     SpaprXive *xive = SPAPR_XIVE(dev);
219     int i;
220 
221     /*
222      * The XiveSource has its own reset handler, which mask off all
223      * IRQs (!P|Q)
224      */
225 
226     /* Mask all valid EASs in the IRQ number space. */
227     for (i = 0; i < xive->nr_irqs; i++) {
228         XiveEAS *eas = &xive->eat[i];
229         if (xive_eas_is_valid(eas)) {
230             eas->w = cpu_to_be64(EAS_VALID | EAS_MASKED);
231         } else {
232             eas->w = 0;
233         }
234     }
235 
236     /* Clear all ENDs */
237     for (i = 0; i < xive->nr_ends; i++) {
238         spapr_xive_end_reset(&xive->endt[i]);
239     }
240 }
241 
242 static void spapr_xive_instance_init(Object *obj)
243 {
244     SpaprXive *xive = SPAPR_XIVE(obj);
245 
246     object_initialize_child(obj, "source", &xive->source, sizeof(xive->source),
247                             TYPE_XIVE_SOURCE, &error_abort, NULL);
248 
249     object_initialize_child(obj, "end_source", &xive->end_source,
250                             sizeof(xive->end_source), TYPE_XIVE_END_SOURCE,
251                             &error_abort, NULL);
252 
253     /* Not connected to the KVM XIVE device */
254     xive->fd = -1;
255 }
256 
257 static void spapr_xive_realize(DeviceState *dev, Error **errp)
258 {
259     SpaprXive *xive = SPAPR_XIVE(dev);
260     XiveSource *xsrc = &xive->source;
261     XiveENDSource *end_xsrc = &xive->end_source;
262     Error *local_err = NULL;
263 
264     if (!xive->nr_irqs) {
265         error_setg(errp, "Number of interrupt needs to be greater 0");
266         return;
267     }
268 
269     if (!xive->nr_ends) {
270         error_setg(errp, "Number of interrupt needs to be greater 0");
271         return;
272     }
273 
274     /*
275      * Initialize the internal sources, for IPIs and virtual devices.
276      */
277     object_property_set_int(OBJECT(xsrc), xive->nr_irqs, "nr-irqs",
278                             &error_fatal);
279     object_property_add_const_link(OBJECT(xsrc), "xive", OBJECT(xive),
280                                    &error_fatal);
281     object_property_set_bool(OBJECT(xsrc), true, "realized", &local_err);
282     if (local_err) {
283         error_propagate(errp, local_err);
284         return;
285     }
286     sysbus_init_mmio(SYS_BUS_DEVICE(xive), &xsrc->esb_mmio);
287 
288     /*
289      * Initialize the END ESB source
290      */
291     object_property_set_int(OBJECT(end_xsrc), xive->nr_irqs, "nr-ends",
292                             &error_fatal);
293     object_property_add_const_link(OBJECT(end_xsrc), "xive", OBJECT(xive),
294                                    &error_fatal);
295     object_property_set_bool(OBJECT(end_xsrc), true, "realized", &local_err);
296     if (local_err) {
297         error_propagate(errp, local_err);
298         return;
299     }
300     sysbus_init_mmio(SYS_BUS_DEVICE(xive), &end_xsrc->esb_mmio);
301 
302     /* Set the mapping address of the END ESB pages after the source ESBs */
303     xive->end_base = xive->vc_base + (1ull << xsrc->esb_shift) * xsrc->nr_irqs;
304 
305     /*
306      * Allocate the routing tables
307      */
308     xive->eat = g_new0(XiveEAS, xive->nr_irqs);
309     xive->endt = g_new0(XiveEND, xive->nr_ends);
310 
311     xive->nodename = g_strdup_printf("interrupt-controller@%" PRIx64,
312                            xive->tm_base + XIVE_TM_USER_PAGE * (1 << TM_SHIFT));
313 
314     qemu_register_reset(spapr_xive_reset, dev);
315 
316     /* TIMA initialization */
317     memory_region_init_io(&xive->tm_mmio, OBJECT(xive), &xive_tm_ops, xive,
318                           "xive.tima", 4ull << TM_SHIFT);
319     sysbus_init_mmio(SYS_BUS_DEVICE(xive), &xive->tm_mmio);
320 
321     /*
322      * Map all regions. These will be enabled or disabled at reset and
323      * can also be overridden by KVM memory regions if active
324      */
325     sysbus_mmio_map(SYS_BUS_DEVICE(xive), 0, xive->vc_base);
326     sysbus_mmio_map(SYS_BUS_DEVICE(xive), 1, xive->end_base);
327     sysbus_mmio_map(SYS_BUS_DEVICE(xive), 2, xive->tm_base);
328 }
329 
330 static int spapr_xive_get_eas(XiveRouter *xrtr, uint8_t eas_blk,
331                               uint32_t eas_idx, XiveEAS *eas)
332 {
333     SpaprXive *xive = SPAPR_XIVE(xrtr);
334 
335     if (eas_idx >= xive->nr_irqs) {
336         return -1;
337     }
338 
339     *eas = xive->eat[eas_idx];
340     return 0;
341 }
342 
343 static int spapr_xive_get_end(XiveRouter *xrtr,
344                               uint8_t end_blk, uint32_t end_idx, XiveEND *end)
345 {
346     SpaprXive *xive = SPAPR_XIVE(xrtr);
347 
348     if (end_idx >= xive->nr_ends) {
349         return -1;
350     }
351 
352     memcpy(end, &xive->endt[end_idx], sizeof(XiveEND));
353     return 0;
354 }
355 
356 static int spapr_xive_write_end(XiveRouter *xrtr, uint8_t end_blk,
357                                 uint32_t end_idx, XiveEND *end,
358                                 uint8_t word_number)
359 {
360     SpaprXive *xive = SPAPR_XIVE(xrtr);
361 
362     if (end_idx >= xive->nr_ends) {
363         return -1;
364     }
365 
366     memcpy(&xive->endt[end_idx], end, sizeof(XiveEND));
367     return 0;
368 }
369 
370 static int spapr_xive_get_nvt(XiveRouter *xrtr,
371                               uint8_t nvt_blk, uint32_t nvt_idx, XiveNVT *nvt)
372 {
373     uint32_t vcpu_id = spapr_xive_nvt_to_target(nvt_blk, nvt_idx);
374     PowerPCCPU *cpu = spapr_find_cpu(vcpu_id);
375 
376     if (!cpu) {
377         /* TODO: should we assert() if we can find a NVT ? */
378         return -1;
379     }
380 
381     /*
382      * sPAPR does not maintain a NVT table. Return that the NVT is
383      * valid if we have found a matching CPU
384      */
385     nvt->w0 = cpu_to_be32(NVT_W0_VALID);
386     return 0;
387 }
388 
389 static int spapr_xive_write_nvt(XiveRouter *xrtr, uint8_t nvt_blk,
390                                 uint32_t nvt_idx, XiveNVT *nvt,
391                                 uint8_t word_number)
392 {
393     /*
394      * We don't need to write back to the NVTs because the sPAPR
395      * machine should never hit a non-scheduled NVT. It should never
396      * get called.
397      */
398     g_assert_not_reached();
399 }
400 
401 static XiveTCTX *spapr_xive_get_tctx(XiveRouter *xrtr, CPUState *cs)
402 {
403     PowerPCCPU *cpu = POWERPC_CPU(cs);
404 
405     return spapr_cpu_state(cpu)->tctx;
406 }
407 
408 static const VMStateDescription vmstate_spapr_xive_end = {
409     .name = TYPE_SPAPR_XIVE "/end",
410     .version_id = 1,
411     .minimum_version_id = 1,
412     .fields = (VMStateField []) {
413         VMSTATE_UINT32(w0, XiveEND),
414         VMSTATE_UINT32(w1, XiveEND),
415         VMSTATE_UINT32(w2, XiveEND),
416         VMSTATE_UINT32(w3, XiveEND),
417         VMSTATE_UINT32(w4, XiveEND),
418         VMSTATE_UINT32(w5, XiveEND),
419         VMSTATE_UINT32(w6, XiveEND),
420         VMSTATE_UINT32(w7, XiveEND),
421         VMSTATE_END_OF_LIST()
422     },
423 };
424 
425 static const VMStateDescription vmstate_spapr_xive_eas = {
426     .name = TYPE_SPAPR_XIVE "/eas",
427     .version_id = 1,
428     .minimum_version_id = 1,
429     .fields = (VMStateField []) {
430         VMSTATE_UINT64(w, XiveEAS),
431         VMSTATE_END_OF_LIST()
432     },
433 };
434 
435 static int vmstate_spapr_xive_pre_save(void *opaque)
436 {
437     if (kvm_irqchip_in_kernel()) {
438         return kvmppc_xive_pre_save(SPAPR_XIVE(opaque));
439     }
440 
441     return 0;
442 }
443 
444 /*
445  * Called by the sPAPR IRQ backend 'post_load' method at the machine
446  * level.
447  */
448 static int spapr_xive_post_load(SpaprInterruptController *intc, int version_id)
449 {
450     if (kvm_irqchip_in_kernel()) {
451         return kvmppc_xive_post_load(SPAPR_XIVE(intc), version_id);
452     }
453 
454     return 0;
455 }
456 
457 static const VMStateDescription vmstate_spapr_xive = {
458     .name = TYPE_SPAPR_XIVE,
459     .version_id = 1,
460     .minimum_version_id = 1,
461     .pre_save = vmstate_spapr_xive_pre_save,
462     .post_load = NULL, /* handled at the machine level */
463     .fields = (VMStateField[]) {
464         VMSTATE_UINT32_EQUAL(nr_irqs, SpaprXive, NULL),
465         VMSTATE_STRUCT_VARRAY_POINTER_UINT32(eat, SpaprXive, nr_irqs,
466                                      vmstate_spapr_xive_eas, XiveEAS),
467         VMSTATE_STRUCT_VARRAY_POINTER_UINT32(endt, SpaprXive, nr_ends,
468                                              vmstate_spapr_xive_end, XiveEND),
469         VMSTATE_END_OF_LIST()
470     },
471 };
472 
473 static int spapr_xive_claim_irq(SpaprInterruptController *intc, int lisn,
474                                 bool lsi, Error **errp)
475 {
476     SpaprXive *xive = SPAPR_XIVE(intc);
477     XiveSource *xsrc = &xive->source;
478 
479     assert(lisn < xive->nr_irqs);
480 
481     if (xive_eas_is_valid(&xive->eat[lisn])) {
482         error_setg(errp, "IRQ %d is not free", lisn);
483         return -EBUSY;
484     }
485 
486     /*
487      * Set default values when allocating an IRQ number
488      */
489     xive->eat[lisn].w |= cpu_to_be64(EAS_VALID | EAS_MASKED);
490     if (lsi) {
491         xive_source_irq_set_lsi(xsrc, lisn);
492     }
493 
494     if (kvm_irqchip_in_kernel()) {
495         return kvmppc_xive_source_reset_one(xsrc, lisn, errp);
496     }
497 
498     return 0;
499 }
500 
501 static void spapr_xive_free_irq(SpaprInterruptController *intc, int lisn)
502 {
503     SpaprXive *xive = SPAPR_XIVE(intc);
504     assert(lisn < xive->nr_irqs);
505 
506     xive->eat[lisn].w &= cpu_to_be64(~EAS_VALID);
507 }
508 
509 static Property spapr_xive_properties[] = {
510     DEFINE_PROP_UINT32("nr-irqs", SpaprXive, nr_irqs, 0),
511     DEFINE_PROP_UINT32("nr-ends", SpaprXive, nr_ends, 0),
512     DEFINE_PROP_UINT64("vc-base", SpaprXive, vc_base, SPAPR_XIVE_VC_BASE),
513     DEFINE_PROP_UINT64("tm-base", SpaprXive, tm_base, SPAPR_XIVE_TM_BASE),
514     DEFINE_PROP_END_OF_LIST(),
515 };
516 
517 static int spapr_xive_cpu_intc_create(SpaprInterruptController *intc,
518                                       PowerPCCPU *cpu, Error **errp)
519 {
520     SpaprXive *xive = SPAPR_XIVE(intc);
521     Object *obj;
522     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
523 
524     obj = xive_tctx_create(OBJECT(cpu), XIVE_ROUTER(xive), errp);
525     if (!obj) {
526         return -1;
527     }
528 
529     spapr_cpu->tctx = XIVE_TCTX(obj);
530     return 0;
531 }
532 
533 static void xive_tctx_set_os_cam(XiveTCTX *tctx, uint32_t os_cam)
534 {
535     uint32_t qw1w2 = cpu_to_be32(TM_QW1W2_VO | os_cam);
536     memcpy(&tctx->regs[TM_QW1_OS + TM_WORD2], &qw1w2, 4);
537 }
538 
539 static void spapr_xive_cpu_intc_reset(SpaprInterruptController *intc,
540                                      PowerPCCPU *cpu)
541 {
542     XiveTCTX *tctx = spapr_cpu_state(cpu)->tctx;
543     uint8_t  nvt_blk;
544     uint32_t nvt_idx;
545 
546     xive_tctx_reset(tctx);
547 
548     /*
549      * When a Virtual Processor is scheduled to run on a HW thread,
550      * the hypervisor pushes its identifier in the OS CAM line.
551      * Emulate the same behavior under QEMU.
552      */
553     spapr_xive_cpu_to_nvt(cpu, &nvt_blk, &nvt_idx);
554 
555     xive_tctx_set_os_cam(tctx, xive_nvt_cam_line(nvt_blk, nvt_idx));
556 }
557 
558 static void spapr_xive_set_irq(SpaprInterruptController *intc, int irq, int val)
559 {
560     SpaprXive *xive = SPAPR_XIVE(intc);
561 
562     if (kvm_irqchip_in_kernel()) {
563         kvmppc_xive_source_set_irq(&xive->source, irq, val);
564     } else {
565         xive_source_set_irq(&xive->source, irq, val);
566     }
567 }
568 
569 static void spapr_xive_print_info(SpaprInterruptController *intc, Monitor *mon)
570 {
571     SpaprXive *xive = SPAPR_XIVE(intc);
572     CPUState *cs;
573 
574     CPU_FOREACH(cs) {
575         PowerPCCPU *cpu = POWERPC_CPU(cs);
576 
577         xive_tctx_pic_print_info(spapr_cpu_state(cpu)->tctx, mon);
578     }
579 
580     spapr_xive_pic_print_info(xive, mon);
581 }
582 
583 static void spapr_xive_dt(SpaprInterruptController *intc, uint32_t nr_servers,
584                           void *fdt, uint32_t phandle)
585 {
586     SpaprXive *xive = SPAPR_XIVE(intc);
587     int node;
588     uint64_t timas[2 * 2];
589     /* Interrupt number ranges for the IPIs */
590     uint32_t lisn_ranges[] = {
591         cpu_to_be32(0),
592         cpu_to_be32(nr_servers),
593     };
594     /*
595      * EQ size - the sizes of pages supported by the system 4K, 64K,
596      * 2M, 16M. We only advertise 64K for the moment.
597      */
598     uint32_t eq_sizes[] = {
599         cpu_to_be32(16), /* 64K */
600     };
601     /*
602      * The following array is in sync with the reserved priorities
603      * defined by the 'spapr_xive_priority_is_reserved' routine.
604      */
605     uint32_t plat_res_int_priorities[] = {
606         cpu_to_be32(7),    /* start */
607         cpu_to_be32(0xf8), /* count */
608     };
609 
610     /* Thread Interrupt Management Area : User (ring 3) and OS (ring 2) */
611     timas[0] = cpu_to_be64(xive->tm_base +
612                            XIVE_TM_USER_PAGE * (1ull << TM_SHIFT));
613     timas[1] = cpu_to_be64(1ull << TM_SHIFT);
614     timas[2] = cpu_to_be64(xive->tm_base +
615                            XIVE_TM_OS_PAGE * (1ull << TM_SHIFT));
616     timas[3] = cpu_to_be64(1ull << TM_SHIFT);
617 
618     _FDT(node = fdt_add_subnode(fdt, 0, xive->nodename));
619 
620     _FDT(fdt_setprop_string(fdt, node, "device_type", "power-ivpe"));
621     _FDT(fdt_setprop(fdt, node, "reg", timas, sizeof(timas)));
622 
623     _FDT(fdt_setprop_string(fdt, node, "compatible", "ibm,power-ivpe"));
624     _FDT(fdt_setprop(fdt, node, "ibm,xive-eq-sizes", eq_sizes,
625                      sizeof(eq_sizes)));
626     _FDT(fdt_setprop(fdt, node, "ibm,xive-lisn-ranges", lisn_ranges,
627                      sizeof(lisn_ranges)));
628 
629     /* For Linux to link the LSIs to the interrupt controller. */
630     _FDT(fdt_setprop(fdt, node, "interrupt-controller", NULL, 0));
631     _FDT(fdt_setprop_cell(fdt, node, "#interrupt-cells", 2));
632 
633     /* For SLOF */
634     _FDT(fdt_setprop_cell(fdt, node, "linux,phandle", phandle));
635     _FDT(fdt_setprop_cell(fdt, node, "phandle", phandle));
636 
637     /*
638      * The "ibm,plat-res-int-priorities" property defines the priority
639      * ranges reserved by the hypervisor
640      */
641     _FDT(fdt_setprop(fdt, 0, "ibm,plat-res-int-priorities",
642                      plat_res_int_priorities, sizeof(plat_res_int_priorities)));
643 }
644 
645 static int spapr_xive_activate(SpaprInterruptController *intc, Error **errp)
646 {
647     SpaprXive *xive = SPAPR_XIVE(intc);
648 
649     if (kvm_enabled()) {
650         int rc = spapr_irq_init_kvm(kvmppc_xive_connect, intc, errp);
651         if (rc < 0) {
652             return rc;
653         }
654     }
655 
656     /* Activate the XIVE MMIOs */
657     spapr_xive_mmio_set_enabled(xive, true);
658 
659     return 0;
660 }
661 
662 static void spapr_xive_deactivate(SpaprInterruptController *intc)
663 {
664     SpaprXive *xive = SPAPR_XIVE(intc);
665 
666     spapr_xive_mmio_set_enabled(xive, false);
667 
668     if (kvm_irqchip_in_kernel()) {
669         kvmppc_xive_disconnect(intc);
670     }
671 }
672 
673 static void spapr_xive_class_init(ObjectClass *klass, void *data)
674 {
675     DeviceClass *dc = DEVICE_CLASS(klass);
676     XiveRouterClass *xrc = XIVE_ROUTER_CLASS(klass);
677     SpaprInterruptControllerClass *sicc = SPAPR_INTC_CLASS(klass);
678 
679     dc->desc    = "sPAPR XIVE Interrupt Controller";
680     dc->props   = spapr_xive_properties;
681     dc->realize = spapr_xive_realize;
682     dc->vmsd    = &vmstate_spapr_xive;
683 
684     xrc->get_eas = spapr_xive_get_eas;
685     xrc->get_end = spapr_xive_get_end;
686     xrc->write_end = spapr_xive_write_end;
687     xrc->get_nvt = spapr_xive_get_nvt;
688     xrc->write_nvt = spapr_xive_write_nvt;
689     xrc->get_tctx = spapr_xive_get_tctx;
690 
691     sicc->activate = spapr_xive_activate;
692     sicc->deactivate = spapr_xive_deactivate;
693     sicc->cpu_intc_create = spapr_xive_cpu_intc_create;
694     sicc->cpu_intc_reset = spapr_xive_cpu_intc_reset;
695     sicc->claim_irq = spapr_xive_claim_irq;
696     sicc->free_irq = spapr_xive_free_irq;
697     sicc->set_irq = spapr_xive_set_irq;
698     sicc->print_info = spapr_xive_print_info;
699     sicc->dt = spapr_xive_dt;
700     sicc->post_load = spapr_xive_post_load;
701 }
702 
703 static const TypeInfo spapr_xive_info = {
704     .name = TYPE_SPAPR_XIVE,
705     .parent = TYPE_XIVE_ROUTER,
706     .instance_init = spapr_xive_instance_init,
707     .instance_size = sizeof(SpaprXive),
708     .class_init = spapr_xive_class_init,
709     .interfaces = (InterfaceInfo[]) {
710         { TYPE_SPAPR_INTC },
711         { }
712     },
713 };
714 
715 static void spapr_xive_register_types(void)
716 {
717     type_register_static(&spapr_xive_info);
718 }
719 
720 type_init(spapr_xive_register_types)
721 
722 /*
723  * XIVE hcalls
724  *
725  * The terminology used by the XIVE hcalls is the following :
726  *
727  *   TARGET vCPU number
728  *   EQ     Event Queue assigned by OS to receive event data
729  *   ESB    page for source interrupt management
730  *   LISN   Logical Interrupt Source Number identifying a source in the
731  *          machine
732  *   EISN   Effective Interrupt Source Number used by guest OS to
733  *          identify source in the guest
734  *
735  * The EAS, END, NVT structures are not exposed.
736  */
737 
738 /*
739  * Linux hosts under OPAL reserve priority 7 for their own escalation
740  * interrupts (DD2.X POWER9). So we only allow the guest to use
741  * priorities [0..6].
742  */
743 static bool spapr_xive_priority_is_reserved(uint8_t priority)
744 {
745     switch (priority) {
746     case 0 ... 6:
747         return false;
748     case 7: /* OPAL escalation queue */
749     default:
750         return true;
751     }
752 }
753 
754 /*
755  * The H_INT_GET_SOURCE_INFO hcall() is used to obtain the logical
756  * real address of the MMIO page through which the Event State Buffer
757  * entry associated with the value of the "lisn" parameter is managed.
758  *
759  * Parameters:
760  * Input
761  * - R4: "flags"
762  *         Bits 0-63 reserved
763  * - R5: "lisn" is per "interrupts", "interrupt-map", or
764  *       "ibm,xive-lisn-ranges" properties, or as returned by the
765  *       ibm,query-interrupt-source-number RTAS call, or as returned
766  *       by the H_ALLOCATE_VAS_WINDOW hcall
767  *
768  * Output
769  * - R4: "flags"
770  *         Bits 0-59: Reserved
771  *         Bit 60: H_INT_ESB must be used for Event State Buffer
772  *                 management
773  *         Bit 61: 1 == LSI  0 == MSI
774  *         Bit 62: the full function page supports trigger
775  *         Bit 63: Store EOI Supported
776  * - R5: Logical Real address of full function Event State Buffer
777  *       management page, -1 if H_INT_ESB hcall flag is set to 1.
778  * - R6: Logical Real Address of trigger only Event State Buffer
779  *       management page or -1.
780  * - R7: Power of 2 page size for the ESB management pages returned in
781  *       R5 and R6.
782  */
783 
784 #define SPAPR_XIVE_SRC_H_INT_ESB     PPC_BIT(60) /* ESB manage with H_INT_ESB */
785 #define SPAPR_XIVE_SRC_LSI           PPC_BIT(61) /* Virtual LSI type */
786 #define SPAPR_XIVE_SRC_TRIGGER       PPC_BIT(62) /* Trigger and management
787                                                     on same page */
788 #define SPAPR_XIVE_SRC_STORE_EOI     PPC_BIT(63) /* Store EOI support */
789 
790 static target_ulong h_int_get_source_info(PowerPCCPU *cpu,
791                                           SpaprMachineState *spapr,
792                                           target_ulong opcode,
793                                           target_ulong *args)
794 {
795     SpaprXive *xive = spapr->xive;
796     XiveSource *xsrc = &xive->source;
797     target_ulong flags  = args[0];
798     target_ulong lisn   = args[1];
799 
800     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
801         return H_FUNCTION;
802     }
803 
804     if (flags) {
805         return H_PARAMETER;
806     }
807 
808     if (lisn >= xive->nr_irqs) {
809         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Unknown LISN " TARGET_FMT_lx "\n",
810                       lisn);
811         return H_P2;
812     }
813 
814     if (!xive_eas_is_valid(&xive->eat[lisn])) {
815         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Invalid LISN " TARGET_FMT_lx "\n",
816                       lisn);
817         return H_P2;
818     }
819 
820     /*
821      * All sources are emulated under the main XIVE object and share
822      * the same characteristics.
823      */
824     args[0] = 0;
825     if (!xive_source_esb_has_2page(xsrc)) {
826         args[0] |= SPAPR_XIVE_SRC_TRIGGER;
827     }
828     if (xsrc->esb_flags & XIVE_SRC_STORE_EOI) {
829         args[0] |= SPAPR_XIVE_SRC_STORE_EOI;
830     }
831 
832     /*
833      * Force the use of the H_INT_ESB hcall in case of an LSI
834      * interrupt. This is necessary under KVM to re-trigger the
835      * interrupt if the level is still asserted
836      */
837     if (xive_source_irq_is_lsi(xsrc, lisn)) {
838         args[0] |= SPAPR_XIVE_SRC_H_INT_ESB | SPAPR_XIVE_SRC_LSI;
839     }
840 
841     if (!(args[0] & SPAPR_XIVE_SRC_H_INT_ESB)) {
842         args[1] = xive->vc_base + xive_source_esb_mgmt(xsrc, lisn);
843     } else {
844         args[1] = -1;
845     }
846 
847     if (xive_source_esb_has_2page(xsrc) &&
848         !(args[0] & SPAPR_XIVE_SRC_H_INT_ESB)) {
849         args[2] = xive->vc_base + xive_source_esb_page(xsrc, lisn);
850     } else {
851         args[2] = -1;
852     }
853 
854     if (xive_source_esb_has_2page(xsrc)) {
855         args[3] = xsrc->esb_shift - 1;
856     } else {
857         args[3] = xsrc->esb_shift;
858     }
859 
860     return H_SUCCESS;
861 }
862 
863 /*
864  * The H_INT_SET_SOURCE_CONFIG hcall() is used to assign a Logical
865  * Interrupt Source to a target. The Logical Interrupt Source is
866  * designated with the "lisn" parameter and the target is designated
867  * with the "target" and "priority" parameters.  Upon return from the
868  * hcall(), no additional interrupts will be directed to the old EQ.
869  *
870  * Parameters:
871  * Input:
872  * - R4: "flags"
873  *         Bits 0-61: Reserved
874  *         Bit 62: set the "eisn" in the EAS
875  *         Bit 63: masks the interrupt source in the hardware interrupt
876  *       control structure. An interrupt masked by this mechanism will
877  *       be dropped, but it's source state bits will still be
878  *       set. There is no race-free way of unmasking and restoring the
879  *       source. Thus this should only be used in interrupts that are
880  *       also masked at the source, and only in cases where the
881  *       interrupt is not meant to be used for a large amount of time
882  *       because no valid target exists for it for example
883  * - R5: "lisn" is per "interrupts", "interrupt-map", or
884  *       "ibm,xive-lisn-ranges" properties, or as returned by the
885  *       ibm,query-interrupt-source-number RTAS call, or as returned by
886  *       the H_ALLOCATE_VAS_WINDOW hcall
887  * - R6: "target" is per "ibm,ppc-interrupt-server#s" or
888  *       "ibm,ppc-interrupt-gserver#s"
889  * - R7: "priority" is a valid priority not in
890  *       "ibm,plat-res-int-priorities"
891  * - R8: "eisn" is the guest EISN associated with the "lisn"
892  *
893  * Output:
894  * - None
895  */
896 
897 #define SPAPR_XIVE_SRC_SET_EISN PPC_BIT(62)
898 #define SPAPR_XIVE_SRC_MASK     PPC_BIT(63)
899 
900 static target_ulong h_int_set_source_config(PowerPCCPU *cpu,
901                                             SpaprMachineState *spapr,
902                                             target_ulong opcode,
903                                             target_ulong *args)
904 {
905     SpaprXive *xive = spapr->xive;
906     XiveEAS eas, new_eas;
907     target_ulong flags    = args[0];
908     target_ulong lisn     = args[1];
909     target_ulong target   = args[2];
910     target_ulong priority = args[3];
911     target_ulong eisn     = args[4];
912     uint8_t end_blk;
913     uint32_t end_idx;
914 
915     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
916         return H_FUNCTION;
917     }
918 
919     if (flags & ~(SPAPR_XIVE_SRC_SET_EISN | SPAPR_XIVE_SRC_MASK)) {
920         return H_PARAMETER;
921     }
922 
923     if (lisn >= xive->nr_irqs) {
924         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Unknown LISN " TARGET_FMT_lx "\n",
925                       lisn);
926         return H_P2;
927     }
928 
929     eas = xive->eat[lisn];
930     if (!xive_eas_is_valid(&eas)) {
931         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Invalid LISN " TARGET_FMT_lx "\n",
932                       lisn);
933         return H_P2;
934     }
935 
936     /* priority 0xff is used to reset the EAS */
937     if (priority == 0xff) {
938         new_eas.w = cpu_to_be64(EAS_VALID | EAS_MASKED);
939         goto out;
940     }
941 
942     if (flags & SPAPR_XIVE_SRC_MASK) {
943         new_eas.w = eas.w | cpu_to_be64(EAS_MASKED);
944     } else {
945         new_eas.w = eas.w & cpu_to_be64(~EAS_MASKED);
946     }
947 
948     if (spapr_xive_priority_is_reserved(priority)) {
949         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: priority " TARGET_FMT_ld
950                       " is reserved\n", priority);
951         return H_P4;
952     }
953 
954     /*
955      * Validate that "target" is part of the list of threads allocated
956      * to the partition. For that, find the END corresponding to the
957      * target.
958      */
959     if (spapr_xive_target_to_end(target, priority, &end_blk, &end_idx)) {
960         return H_P3;
961     }
962 
963     new_eas.w = xive_set_field64(EAS_END_BLOCK, new_eas.w, end_blk);
964     new_eas.w = xive_set_field64(EAS_END_INDEX, new_eas.w, end_idx);
965 
966     if (flags & SPAPR_XIVE_SRC_SET_EISN) {
967         new_eas.w = xive_set_field64(EAS_END_DATA, new_eas.w, eisn);
968     }
969 
970     if (kvm_irqchip_in_kernel()) {
971         Error *local_err = NULL;
972 
973         kvmppc_xive_set_source_config(xive, lisn, &new_eas, &local_err);
974         if (local_err) {
975             error_report_err(local_err);
976             return H_HARDWARE;
977         }
978     }
979 
980 out:
981     xive->eat[lisn] = new_eas;
982     return H_SUCCESS;
983 }
984 
985 /*
986  * The H_INT_GET_SOURCE_CONFIG hcall() is used to determine to which
987  * target/priority pair is assigned to the specified Logical Interrupt
988  * Source.
989  *
990  * Parameters:
991  * Input:
992  * - R4: "flags"
993  *         Bits 0-63 Reserved
994  * - R5: "lisn" is per "interrupts", "interrupt-map", or
995  *       "ibm,xive-lisn-ranges" properties, or as returned by the
996  *       ibm,query-interrupt-source-number RTAS call, or as
997  *       returned by the H_ALLOCATE_VAS_WINDOW hcall
998  *
999  * Output:
1000  * - R4: Target to which the specified Logical Interrupt Source is
1001  *       assigned
1002  * - R5: Priority to which the specified Logical Interrupt Source is
1003  *       assigned
1004  * - R6: EISN for the specified Logical Interrupt Source (this will be
1005  *       equivalent to the LISN if not changed by H_INT_SET_SOURCE_CONFIG)
1006  */
1007 static target_ulong h_int_get_source_config(PowerPCCPU *cpu,
1008                                             SpaprMachineState *spapr,
1009                                             target_ulong opcode,
1010                                             target_ulong *args)
1011 {
1012     SpaprXive *xive = spapr->xive;
1013     target_ulong flags = args[0];
1014     target_ulong lisn = args[1];
1015     XiveEAS eas;
1016     XiveEND *end;
1017     uint8_t nvt_blk;
1018     uint32_t end_idx, nvt_idx;
1019 
1020     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1021         return H_FUNCTION;
1022     }
1023 
1024     if (flags) {
1025         return H_PARAMETER;
1026     }
1027 
1028     if (lisn >= xive->nr_irqs) {
1029         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Unknown LISN " TARGET_FMT_lx "\n",
1030                       lisn);
1031         return H_P2;
1032     }
1033 
1034     eas = xive->eat[lisn];
1035     if (!xive_eas_is_valid(&eas)) {
1036         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Invalid LISN " TARGET_FMT_lx "\n",
1037                       lisn);
1038         return H_P2;
1039     }
1040 
1041     /* EAS_END_BLOCK is unused on sPAPR */
1042     end_idx = xive_get_field64(EAS_END_INDEX, eas.w);
1043 
1044     assert(end_idx < xive->nr_ends);
1045     end = &xive->endt[end_idx];
1046 
1047     nvt_blk = xive_get_field32(END_W6_NVT_BLOCK, end->w6);
1048     nvt_idx = xive_get_field32(END_W6_NVT_INDEX, end->w6);
1049     args[0] = spapr_xive_nvt_to_target(nvt_blk, nvt_idx);
1050 
1051     if (xive_eas_is_masked(&eas)) {
1052         args[1] = 0xff;
1053     } else {
1054         args[1] = xive_get_field32(END_W7_F0_PRIORITY, end->w7);
1055     }
1056 
1057     args[2] = xive_get_field64(EAS_END_DATA, eas.w);
1058 
1059     return H_SUCCESS;
1060 }
1061 
1062 /*
1063  * The H_INT_GET_QUEUE_INFO hcall() is used to get the logical real
1064  * address of the notification management page associated with the
1065  * specified target and priority.
1066  *
1067  * Parameters:
1068  * Input:
1069  * - R4: "flags"
1070  *         Bits 0-63 Reserved
1071  * - R5: "target" is per "ibm,ppc-interrupt-server#s" or
1072  *       "ibm,ppc-interrupt-gserver#s"
1073  * - R6: "priority" is a valid priority not in
1074  *       "ibm,plat-res-int-priorities"
1075  *
1076  * Output:
1077  * - R4: Logical real address of notification page
1078  * - R5: Power of 2 page size of the notification page
1079  */
1080 static target_ulong h_int_get_queue_info(PowerPCCPU *cpu,
1081                                          SpaprMachineState *spapr,
1082                                          target_ulong opcode,
1083                                          target_ulong *args)
1084 {
1085     SpaprXive *xive = spapr->xive;
1086     XiveENDSource *end_xsrc = &xive->end_source;
1087     target_ulong flags = args[0];
1088     target_ulong target = args[1];
1089     target_ulong priority = args[2];
1090     XiveEND *end;
1091     uint8_t end_blk;
1092     uint32_t end_idx;
1093 
1094     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1095         return H_FUNCTION;
1096     }
1097 
1098     if (flags) {
1099         return H_PARAMETER;
1100     }
1101 
1102     /*
1103      * H_STATE should be returned if a H_INT_RESET is in progress.
1104      * This is not needed when running the emulation under QEMU
1105      */
1106 
1107     if (spapr_xive_priority_is_reserved(priority)) {
1108         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: priority " TARGET_FMT_ld
1109                       " is reserved\n", priority);
1110         return H_P3;
1111     }
1112 
1113     /*
1114      * Validate that "target" is part of the list of threads allocated
1115      * to the partition. For that, find the END corresponding to the
1116      * target.
1117      */
1118     if (spapr_xive_target_to_end(target, priority, &end_blk, &end_idx)) {
1119         return H_P2;
1120     }
1121 
1122     assert(end_idx < xive->nr_ends);
1123     end = &xive->endt[end_idx];
1124 
1125     args[0] = xive->end_base + (1ull << (end_xsrc->esb_shift + 1)) * end_idx;
1126     if (xive_end_is_enqueue(end)) {
1127         args[1] = xive_get_field32(END_W0_QSIZE, end->w0) + 12;
1128     } else {
1129         args[1] = 0;
1130     }
1131 
1132     return H_SUCCESS;
1133 }
1134 
1135 /*
1136  * The H_INT_SET_QUEUE_CONFIG hcall() is used to set or reset a EQ for
1137  * a given "target" and "priority".  It is also used to set the
1138  * notification config associated with the EQ.  An EQ size of 0 is
1139  * used to reset the EQ config for a given target and priority. If
1140  * resetting the EQ config, the END associated with the given "target"
1141  * and "priority" will be changed to disable queueing.
1142  *
1143  * Upon return from the hcall(), no additional interrupts will be
1144  * directed to the old EQ (if one was set). The old EQ (if one was
1145  * set) should be investigated for interrupts that occurred prior to
1146  * or during the hcall().
1147  *
1148  * Parameters:
1149  * Input:
1150  * - R4: "flags"
1151  *         Bits 0-62: Reserved
1152  *         Bit 63: Unconditional Notify (n) per the XIVE spec
1153  * - R5: "target" is per "ibm,ppc-interrupt-server#s" or
1154  *       "ibm,ppc-interrupt-gserver#s"
1155  * - R6: "priority" is a valid priority not in
1156  *       "ibm,plat-res-int-priorities"
1157  * - R7: "eventQueue": The logical real address of the start of the EQ
1158  * - R8: "eventQueueSize": The power of 2 EQ size per "ibm,xive-eq-sizes"
1159  *
1160  * Output:
1161  * - None
1162  */
1163 
1164 #define SPAPR_XIVE_END_ALWAYS_NOTIFY PPC_BIT(63)
1165 
1166 static target_ulong h_int_set_queue_config(PowerPCCPU *cpu,
1167                                            SpaprMachineState *spapr,
1168                                            target_ulong opcode,
1169                                            target_ulong *args)
1170 {
1171     SpaprXive *xive = spapr->xive;
1172     target_ulong flags = args[0];
1173     target_ulong target = args[1];
1174     target_ulong priority = args[2];
1175     target_ulong qpage = args[3];
1176     target_ulong qsize = args[4];
1177     XiveEND end;
1178     uint8_t end_blk, nvt_blk;
1179     uint32_t end_idx, nvt_idx;
1180 
1181     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1182         return H_FUNCTION;
1183     }
1184 
1185     if (flags & ~SPAPR_XIVE_END_ALWAYS_NOTIFY) {
1186         return H_PARAMETER;
1187     }
1188 
1189     /*
1190      * H_STATE should be returned if a H_INT_RESET is in progress.
1191      * This is not needed when running the emulation under QEMU
1192      */
1193 
1194     if (spapr_xive_priority_is_reserved(priority)) {
1195         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: priority " TARGET_FMT_ld
1196                       " is reserved\n", priority);
1197         return H_P3;
1198     }
1199 
1200     /*
1201      * Validate that "target" is part of the list of threads allocated
1202      * to the partition. For that, find the END corresponding to the
1203      * target.
1204      */
1205 
1206     if (spapr_xive_target_to_end(target, priority, &end_blk, &end_idx)) {
1207         return H_P2;
1208     }
1209 
1210     assert(end_idx < xive->nr_ends);
1211     memcpy(&end, &xive->endt[end_idx], sizeof(XiveEND));
1212 
1213     switch (qsize) {
1214     case 12:
1215     case 16:
1216     case 21:
1217     case 24:
1218         if (!QEMU_IS_ALIGNED(qpage, 1ul << qsize)) {
1219             qemu_log_mask(LOG_GUEST_ERROR, "XIVE: EQ @0x%" HWADDR_PRIx
1220                           " is not naturally aligned with %" HWADDR_PRIx "\n",
1221                           qpage, (hwaddr)1 << qsize);
1222             return H_P4;
1223         }
1224         end.w2 = cpu_to_be32((qpage >> 32) & 0x0fffffff);
1225         end.w3 = cpu_to_be32(qpage & 0xffffffff);
1226         end.w0 |= cpu_to_be32(END_W0_ENQUEUE);
1227         end.w0 = xive_set_field32(END_W0_QSIZE, end.w0, qsize - 12);
1228         break;
1229     case 0:
1230         /* reset queue and disable queueing */
1231         spapr_xive_end_reset(&end);
1232         goto out;
1233 
1234     default:
1235         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: invalid EQ size %"PRIx64"\n",
1236                       qsize);
1237         return H_P5;
1238     }
1239 
1240     if (qsize) {
1241         hwaddr plen = 1 << qsize;
1242         void *eq;
1243 
1244         /*
1245          * Validate the guest EQ. We should also check that the queue
1246          * has been zeroed by the OS.
1247          */
1248         eq = address_space_map(CPU(cpu)->as, qpage, &plen, true,
1249                                MEMTXATTRS_UNSPECIFIED);
1250         if (plen != 1 << qsize) {
1251             qemu_log_mask(LOG_GUEST_ERROR, "XIVE: failed to map EQ @0x%"
1252                           HWADDR_PRIx "\n", qpage);
1253             return H_P4;
1254         }
1255         address_space_unmap(CPU(cpu)->as, eq, plen, true, plen);
1256     }
1257 
1258     /* "target" should have been validated above */
1259     if (spapr_xive_target_to_nvt(target, &nvt_blk, &nvt_idx)) {
1260         g_assert_not_reached();
1261     }
1262 
1263     /*
1264      * Ensure the priority and target are correctly set (they will not
1265      * be right after allocation)
1266      */
1267     end.w6 = xive_set_field32(END_W6_NVT_BLOCK, 0ul, nvt_blk) |
1268         xive_set_field32(END_W6_NVT_INDEX, 0ul, nvt_idx);
1269     end.w7 = xive_set_field32(END_W7_F0_PRIORITY, 0ul, priority);
1270 
1271     if (flags & SPAPR_XIVE_END_ALWAYS_NOTIFY) {
1272         end.w0 |= cpu_to_be32(END_W0_UCOND_NOTIFY);
1273     } else {
1274         end.w0 &= cpu_to_be32((uint32_t)~END_W0_UCOND_NOTIFY);
1275     }
1276 
1277     /*
1278      * The generation bit for the END starts at 1 and The END page
1279      * offset counter starts at 0.
1280      */
1281     end.w1 = cpu_to_be32(END_W1_GENERATION) |
1282         xive_set_field32(END_W1_PAGE_OFF, 0ul, 0ul);
1283     end.w0 |= cpu_to_be32(END_W0_VALID);
1284 
1285     /*
1286      * TODO: issue syncs required to ensure all in-flight interrupts
1287      * are complete on the old END
1288      */
1289 
1290 out:
1291     if (kvm_irqchip_in_kernel()) {
1292         Error *local_err = NULL;
1293 
1294         kvmppc_xive_set_queue_config(xive, end_blk, end_idx, &end, &local_err);
1295         if (local_err) {
1296             error_report_err(local_err);
1297             return H_HARDWARE;
1298         }
1299     }
1300 
1301     /* Update END */
1302     memcpy(&xive->endt[end_idx], &end, sizeof(XiveEND));
1303     return H_SUCCESS;
1304 }
1305 
1306 /*
1307  * The H_INT_GET_QUEUE_CONFIG hcall() is used to get a EQ for a given
1308  * target and priority.
1309  *
1310  * Parameters:
1311  * Input:
1312  * - R4: "flags"
1313  *         Bits 0-62: Reserved
1314  *         Bit 63: Debug: Return debug data
1315  * - R5: "target" is per "ibm,ppc-interrupt-server#s" or
1316  *       "ibm,ppc-interrupt-gserver#s"
1317  * - R6: "priority" is a valid priority not in
1318  *       "ibm,plat-res-int-priorities"
1319  *
1320  * Output:
1321  * - R4: "flags":
1322  *       Bits 0-61: Reserved
1323  *       Bit 62: The value of Event Queue Generation Number (g) per
1324  *              the XIVE spec if "Debug" = 1
1325  *       Bit 63: The value of Unconditional Notify (n) per the XIVE spec
1326  * - R5: The logical real address of the start of the EQ
1327  * - R6: The power of 2 EQ size per "ibm,xive-eq-sizes"
1328  * - R7: The value of Event Queue Offset Counter per XIVE spec
1329  *       if "Debug" = 1, else 0
1330  *
1331  */
1332 
1333 #define SPAPR_XIVE_END_DEBUG     PPC_BIT(63)
1334 
1335 static target_ulong h_int_get_queue_config(PowerPCCPU *cpu,
1336                                            SpaprMachineState *spapr,
1337                                            target_ulong opcode,
1338                                            target_ulong *args)
1339 {
1340     SpaprXive *xive = spapr->xive;
1341     target_ulong flags = args[0];
1342     target_ulong target = args[1];
1343     target_ulong priority = args[2];
1344     XiveEND *end;
1345     uint8_t end_blk;
1346     uint32_t end_idx;
1347 
1348     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1349         return H_FUNCTION;
1350     }
1351 
1352     if (flags & ~SPAPR_XIVE_END_DEBUG) {
1353         return H_PARAMETER;
1354     }
1355 
1356     /*
1357      * H_STATE should be returned if a H_INT_RESET is in progress.
1358      * This is not needed when running the emulation under QEMU
1359      */
1360 
1361     if (spapr_xive_priority_is_reserved(priority)) {
1362         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: priority " TARGET_FMT_ld
1363                       " is reserved\n", priority);
1364         return H_P3;
1365     }
1366 
1367     /*
1368      * Validate that "target" is part of the list of threads allocated
1369      * to the partition. For that, find the END corresponding to the
1370      * target.
1371      */
1372     if (spapr_xive_target_to_end(target, priority, &end_blk, &end_idx)) {
1373         return H_P2;
1374     }
1375 
1376     assert(end_idx < xive->nr_ends);
1377     end = &xive->endt[end_idx];
1378 
1379     args[0] = 0;
1380     if (xive_end_is_notify(end)) {
1381         args[0] |= SPAPR_XIVE_END_ALWAYS_NOTIFY;
1382     }
1383 
1384     if (xive_end_is_enqueue(end)) {
1385         args[1] = xive_end_qaddr(end);
1386         args[2] = xive_get_field32(END_W0_QSIZE, end->w0) + 12;
1387     } else {
1388         args[1] = 0;
1389         args[2] = 0;
1390     }
1391 
1392     if (kvm_irqchip_in_kernel()) {
1393         Error *local_err = NULL;
1394 
1395         kvmppc_xive_get_queue_config(xive, end_blk, end_idx, end, &local_err);
1396         if (local_err) {
1397             error_report_err(local_err);
1398             return H_HARDWARE;
1399         }
1400     }
1401 
1402     /* TODO: do we need any locking on the END ? */
1403     if (flags & SPAPR_XIVE_END_DEBUG) {
1404         /* Load the event queue generation number into the return flags */
1405         args[0] |= (uint64_t)xive_get_field32(END_W1_GENERATION, end->w1) << 62;
1406 
1407         /* Load R7 with the event queue offset counter */
1408         args[3] = xive_get_field32(END_W1_PAGE_OFF, end->w1);
1409     } else {
1410         args[3] = 0;
1411     }
1412 
1413     return H_SUCCESS;
1414 }
1415 
1416 /*
1417  * The H_INT_SET_OS_REPORTING_LINE hcall() is used to set the
1418  * reporting cache line pair for the calling thread.  The reporting
1419  * cache lines will contain the OS interrupt context when the OS
1420  * issues a CI store byte to @TIMA+0xC10 to acknowledge the OS
1421  * interrupt. The reporting cache lines can be reset by inputting -1
1422  * in "reportingLine".  Issuing the CI store byte without reporting
1423  * cache lines registered will result in the data not being accessible
1424  * to the OS.
1425  *
1426  * Parameters:
1427  * Input:
1428  * - R4: "flags"
1429  *         Bits 0-63: Reserved
1430  * - R5: "reportingLine": The logical real address of the reporting cache
1431  *       line pair
1432  *
1433  * Output:
1434  * - None
1435  */
1436 static target_ulong h_int_set_os_reporting_line(PowerPCCPU *cpu,
1437                                                 SpaprMachineState *spapr,
1438                                                 target_ulong opcode,
1439                                                 target_ulong *args)
1440 {
1441     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1442         return H_FUNCTION;
1443     }
1444 
1445     /*
1446      * H_STATE should be returned if a H_INT_RESET is in progress.
1447      * This is not needed when running the emulation under QEMU
1448      */
1449 
1450     /* TODO: H_INT_SET_OS_REPORTING_LINE */
1451     return H_FUNCTION;
1452 }
1453 
1454 /*
1455  * The H_INT_GET_OS_REPORTING_LINE hcall() is used to get the logical
1456  * real address of the reporting cache line pair set for the input
1457  * "target".  If no reporting cache line pair has been set, -1 is
1458  * returned.
1459  *
1460  * Parameters:
1461  * Input:
1462  * - R4: "flags"
1463  *         Bits 0-63: Reserved
1464  * - R5: "target" is per "ibm,ppc-interrupt-server#s" or
1465  *       "ibm,ppc-interrupt-gserver#s"
1466  * - R6: "reportingLine": The logical real address of the reporting
1467  *        cache line pair
1468  *
1469  * Output:
1470  * - R4: The logical real address of the reporting line if set, else -1
1471  */
1472 static target_ulong h_int_get_os_reporting_line(PowerPCCPU *cpu,
1473                                                 SpaprMachineState *spapr,
1474                                                 target_ulong opcode,
1475                                                 target_ulong *args)
1476 {
1477     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1478         return H_FUNCTION;
1479     }
1480 
1481     /*
1482      * H_STATE should be returned if a H_INT_RESET is in progress.
1483      * This is not needed when running the emulation under QEMU
1484      */
1485 
1486     /* TODO: H_INT_GET_OS_REPORTING_LINE */
1487     return H_FUNCTION;
1488 }
1489 
1490 /*
1491  * The H_INT_ESB hcall() is used to issue a load or store to the ESB
1492  * page for the input "lisn".  This hcall is only supported for LISNs
1493  * that have the ESB hcall flag set to 1 when returned from hcall()
1494  * H_INT_GET_SOURCE_INFO.
1495  *
1496  * Parameters:
1497  * Input:
1498  * - R4: "flags"
1499  *         Bits 0-62: Reserved
1500  *         bit 63: Store: Store=1, store operation, else load operation
1501  * - R5: "lisn" is per "interrupts", "interrupt-map", or
1502  *       "ibm,xive-lisn-ranges" properties, or as returned by the
1503  *       ibm,query-interrupt-source-number RTAS call, or as
1504  *       returned by the H_ALLOCATE_VAS_WINDOW hcall
1505  * - R6: "esbOffset" is the offset into the ESB page for the load or
1506  *       store operation
1507  * - R7: "storeData" is the data to write for a store operation
1508  *
1509  * Output:
1510  * - R4: The value of the load if load operation, else -1
1511  */
1512 
1513 #define SPAPR_XIVE_ESB_STORE PPC_BIT(63)
1514 
1515 static target_ulong h_int_esb(PowerPCCPU *cpu,
1516                               SpaprMachineState *spapr,
1517                               target_ulong opcode,
1518                               target_ulong *args)
1519 {
1520     SpaprXive *xive = spapr->xive;
1521     XiveEAS eas;
1522     target_ulong flags  = args[0];
1523     target_ulong lisn   = args[1];
1524     target_ulong offset = args[2];
1525     target_ulong data   = args[3];
1526     hwaddr mmio_addr;
1527     XiveSource *xsrc = &xive->source;
1528 
1529     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1530         return H_FUNCTION;
1531     }
1532 
1533     if (flags & ~SPAPR_XIVE_ESB_STORE) {
1534         return H_PARAMETER;
1535     }
1536 
1537     if (lisn >= xive->nr_irqs) {
1538         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Unknown LISN " TARGET_FMT_lx "\n",
1539                       lisn);
1540         return H_P2;
1541     }
1542 
1543     eas = xive->eat[lisn];
1544     if (!xive_eas_is_valid(&eas)) {
1545         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Invalid LISN " TARGET_FMT_lx "\n",
1546                       lisn);
1547         return H_P2;
1548     }
1549 
1550     if (offset > (1ull << xsrc->esb_shift)) {
1551         return H_P3;
1552     }
1553 
1554     if (kvm_irqchip_in_kernel()) {
1555         args[0] = kvmppc_xive_esb_rw(xsrc, lisn, offset, data,
1556                                      flags & SPAPR_XIVE_ESB_STORE);
1557     } else {
1558         mmio_addr = xive->vc_base + xive_source_esb_mgmt(xsrc, lisn) + offset;
1559 
1560         if (dma_memory_rw(&address_space_memory, mmio_addr, &data, 8,
1561                           (flags & SPAPR_XIVE_ESB_STORE))) {
1562             qemu_log_mask(LOG_GUEST_ERROR, "XIVE: failed to access ESB @0x%"
1563                           HWADDR_PRIx "\n", mmio_addr);
1564             return H_HARDWARE;
1565         }
1566         args[0] = (flags & SPAPR_XIVE_ESB_STORE) ? -1 : data;
1567     }
1568     return H_SUCCESS;
1569 }
1570 
1571 /*
1572  * The H_INT_SYNC hcall() is used to issue hardware syncs that will
1573  * ensure any in flight events for the input lisn are in the event
1574  * queue.
1575  *
1576  * Parameters:
1577  * Input:
1578  * - R4: "flags"
1579  *         Bits 0-63: Reserved
1580  * - R5: "lisn" is per "interrupts", "interrupt-map", or
1581  *       "ibm,xive-lisn-ranges" properties, or as returned by the
1582  *       ibm,query-interrupt-source-number RTAS call, or as
1583  *       returned by the H_ALLOCATE_VAS_WINDOW hcall
1584  *
1585  * Output:
1586  * - None
1587  */
1588 static target_ulong h_int_sync(PowerPCCPU *cpu,
1589                                SpaprMachineState *spapr,
1590                                target_ulong opcode,
1591                                target_ulong *args)
1592 {
1593     SpaprXive *xive = spapr->xive;
1594     XiveEAS eas;
1595     target_ulong flags = args[0];
1596     target_ulong lisn = args[1];
1597 
1598     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1599         return H_FUNCTION;
1600     }
1601 
1602     if (flags) {
1603         return H_PARAMETER;
1604     }
1605 
1606     if (lisn >= xive->nr_irqs) {
1607         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Unknown LISN " TARGET_FMT_lx "\n",
1608                       lisn);
1609         return H_P2;
1610     }
1611 
1612     eas = xive->eat[lisn];
1613     if (!xive_eas_is_valid(&eas)) {
1614         qemu_log_mask(LOG_GUEST_ERROR, "XIVE: Invalid LISN " TARGET_FMT_lx "\n",
1615                       lisn);
1616         return H_P2;
1617     }
1618 
1619     /*
1620      * H_STATE should be returned if a H_INT_RESET is in progress.
1621      * This is not needed when running the emulation under QEMU
1622      */
1623 
1624     /*
1625      * This is not real hardware. Nothing to be done unless when
1626      * under KVM
1627      */
1628 
1629     if (kvm_irqchip_in_kernel()) {
1630         Error *local_err = NULL;
1631 
1632         kvmppc_xive_sync_source(xive, lisn, &local_err);
1633         if (local_err) {
1634             error_report_err(local_err);
1635             return H_HARDWARE;
1636         }
1637     }
1638     return H_SUCCESS;
1639 }
1640 
1641 /*
1642  * The H_INT_RESET hcall() is used to reset all of the partition's
1643  * interrupt exploitation structures to their initial state.  This
1644  * means losing all previously set interrupt state set via
1645  * H_INT_SET_SOURCE_CONFIG and H_INT_SET_QUEUE_CONFIG.
1646  *
1647  * Parameters:
1648  * Input:
1649  * - R4: "flags"
1650  *         Bits 0-63: Reserved
1651  *
1652  * Output:
1653  * - None
1654  */
1655 static target_ulong h_int_reset(PowerPCCPU *cpu,
1656                                 SpaprMachineState *spapr,
1657                                 target_ulong opcode,
1658                                 target_ulong *args)
1659 {
1660     SpaprXive *xive = spapr->xive;
1661     target_ulong flags   = args[0];
1662 
1663     if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
1664         return H_FUNCTION;
1665     }
1666 
1667     if (flags) {
1668         return H_PARAMETER;
1669     }
1670 
1671     device_reset(DEVICE(xive));
1672 
1673     if (kvm_irqchip_in_kernel()) {
1674         Error *local_err = NULL;
1675 
1676         kvmppc_xive_reset(xive, &local_err);
1677         if (local_err) {
1678             error_report_err(local_err);
1679             return H_HARDWARE;
1680         }
1681     }
1682     return H_SUCCESS;
1683 }
1684 
1685 void spapr_xive_hcall_init(SpaprMachineState *spapr)
1686 {
1687     spapr_register_hypercall(H_INT_GET_SOURCE_INFO, h_int_get_source_info);
1688     spapr_register_hypercall(H_INT_SET_SOURCE_CONFIG, h_int_set_source_config);
1689     spapr_register_hypercall(H_INT_GET_SOURCE_CONFIG, h_int_get_source_config);
1690     spapr_register_hypercall(H_INT_GET_QUEUE_INFO, h_int_get_queue_info);
1691     spapr_register_hypercall(H_INT_SET_QUEUE_CONFIG, h_int_set_queue_config);
1692     spapr_register_hypercall(H_INT_GET_QUEUE_CONFIG, h_int_get_queue_config);
1693     spapr_register_hypercall(H_INT_SET_OS_REPORTING_LINE,
1694                              h_int_set_os_reporting_line);
1695     spapr_register_hypercall(H_INT_GET_OS_REPORTING_LINE,
1696                              h_int_get_os_reporting_line);
1697     spapr_register_hypercall(H_INT_ESB, h_int_esb);
1698     spapr_register_hypercall(H_INT_SYNC, h_int_sync);
1699     spapr_register_hypercall(H_INT_RESET, h_int_reset);
1700 }
1701