xref: /openbmc/qemu/hw/ppc/spapr_pci.c (revision c11b0583)
1 /*
2  * QEMU sPAPR PCI host originated from Uninorth PCI host
3  *
4  * Copyright (c) 2011 Alexey Kardashevskiy, IBM Corporation.
5  * Copyright (C) 2011 David Gibson, IBM Corporation.
6  *
7  * Permission is hereby granted, free of charge, to any person obtaining a copy
8  * of this software and associated documentation files (the "Software"), to deal
9  * in the Software without restriction, including without limitation the rights
10  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11  * copies of the Software, and to permit persons to whom the Software is
12  * furnished to do so, subject to the following conditions:
13  *
14  * The above copyright notice and this permission notice shall be included in
15  * all copies or substantial portions of the Software.
16  *
17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
23  * THE SOFTWARE.
24  */
25 #include "hw/hw.h"
26 #include "hw/sysbus.h"
27 #include "hw/pci/pci.h"
28 #include "hw/pci/msi.h"
29 #include "hw/pci/msix.h"
30 #include "hw/pci/pci_host.h"
31 #include "hw/ppc/spapr.h"
32 #include "hw/pci-host/spapr.h"
33 #include "exec/address-spaces.h"
34 #include <libfdt.h>
35 #include "trace.h"
36 #include "qemu/error-report.h"
37 #include "qapi/qmp/qerror.h"
38 
39 #include "hw/pci/pci_bridge.h"
40 #include "hw/pci/pci_bus.h"
41 #include "hw/ppc/spapr_drc.h"
42 #include "sysemu/device_tree.h"
43 
44 /* Copied from the kernel arch/powerpc/platforms/pseries/msi.c */
45 #define RTAS_QUERY_FN           0
46 #define RTAS_CHANGE_FN          1
47 #define RTAS_RESET_FN           2
48 #define RTAS_CHANGE_MSI_FN      3
49 #define RTAS_CHANGE_MSIX_FN     4
50 
51 /* Interrupt types to return on RTAS_CHANGE_* */
52 #define RTAS_TYPE_MSI           1
53 #define RTAS_TYPE_MSIX          2
54 
55 #define FDT_NAME_MAX          128
56 
57 #define _FDT(exp) \
58     do { \
59         int ret = (exp);                                           \
60         if (ret < 0) {                                             \
61             return ret;                                            \
62         }                                                          \
63     } while (0)
64 
65 sPAPRPHBState *spapr_pci_find_phb(sPAPRMachineState *spapr, uint64_t buid)
66 {
67     sPAPRPHBState *sphb;
68 
69     QLIST_FOREACH(sphb, &spapr->phbs, list) {
70         if (sphb->buid != buid) {
71             continue;
72         }
73         return sphb;
74     }
75 
76     return NULL;
77 }
78 
79 PCIDevice *spapr_pci_find_dev(sPAPRMachineState *spapr, uint64_t buid,
80                               uint32_t config_addr)
81 {
82     sPAPRPHBState *sphb = spapr_pci_find_phb(spapr, buid);
83     PCIHostState *phb = PCI_HOST_BRIDGE(sphb);
84     int bus_num = (config_addr >> 16) & 0xFF;
85     int devfn = (config_addr >> 8) & 0xFF;
86 
87     if (!phb) {
88         return NULL;
89     }
90 
91     return pci_find_device(phb->bus, bus_num, devfn);
92 }
93 
94 static uint32_t rtas_pci_cfgaddr(uint32_t arg)
95 {
96     /* This handles the encoding of extended config space addresses */
97     return ((arg >> 20) & 0xf00) | (arg & 0xff);
98 }
99 
100 static void finish_read_pci_config(sPAPRMachineState *spapr, uint64_t buid,
101                                    uint32_t addr, uint32_t size,
102                                    target_ulong rets)
103 {
104     PCIDevice *pci_dev;
105     uint32_t val;
106 
107     if ((size != 1) && (size != 2) && (size != 4)) {
108         /* access must be 1, 2 or 4 bytes */
109         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
110         return;
111     }
112 
113     pci_dev = spapr_pci_find_dev(spapr, buid, addr);
114     addr = rtas_pci_cfgaddr(addr);
115 
116     if (!pci_dev || (addr % size) || (addr >= pci_config_size(pci_dev))) {
117         /* Access must be to a valid device, within bounds and
118          * naturally aligned */
119         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
120         return;
121     }
122 
123     val = pci_host_config_read_common(pci_dev, addr,
124                                       pci_config_size(pci_dev), size);
125 
126     rtas_st(rets, 0, RTAS_OUT_SUCCESS);
127     rtas_st(rets, 1, val);
128 }
129 
130 static void rtas_ibm_read_pci_config(PowerPCCPU *cpu, sPAPRMachineState *spapr,
131                                      uint32_t token, uint32_t nargs,
132                                      target_ulong args,
133                                      uint32_t nret, target_ulong rets)
134 {
135     uint64_t buid;
136     uint32_t size, addr;
137 
138     if ((nargs != 4) || (nret != 2)) {
139         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
140         return;
141     }
142 
143     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
144     size = rtas_ld(args, 3);
145     addr = rtas_ld(args, 0);
146 
147     finish_read_pci_config(spapr, buid, addr, size, rets);
148 }
149 
150 static void rtas_read_pci_config(PowerPCCPU *cpu, sPAPRMachineState *spapr,
151                                  uint32_t token, uint32_t nargs,
152                                  target_ulong args,
153                                  uint32_t nret, target_ulong rets)
154 {
155     uint32_t size, addr;
156 
157     if ((nargs != 2) || (nret != 2)) {
158         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
159         return;
160     }
161 
162     size = rtas_ld(args, 1);
163     addr = rtas_ld(args, 0);
164 
165     finish_read_pci_config(spapr, 0, addr, size, rets);
166 }
167 
168 static void finish_write_pci_config(sPAPRMachineState *spapr, uint64_t buid,
169                                     uint32_t addr, uint32_t size,
170                                     uint32_t val, target_ulong rets)
171 {
172     PCIDevice *pci_dev;
173 
174     if ((size != 1) && (size != 2) && (size != 4)) {
175         /* access must be 1, 2 or 4 bytes */
176         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
177         return;
178     }
179 
180     pci_dev = spapr_pci_find_dev(spapr, buid, addr);
181     addr = rtas_pci_cfgaddr(addr);
182 
183     if (!pci_dev || (addr % size) || (addr >= pci_config_size(pci_dev))) {
184         /* Access must be to a valid device, within bounds and
185          * naturally aligned */
186         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
187         return;
188     }
189 
190     pci_host_config_write_common(pci_dev, addr, pci_config_size(pci_dev),
191                                  val, size);
192 
193     rtas_st(rets, 0, RTAS_OUT_SUCCESS);
194 }
195 
196 static void rtas_ibm_write_pci_config(PowerPCCPU *cpu, sPAPRMachineState *spapr,
197                                       uint32_t token, uint32_t nargs,
198                                       target_ulong args,
199                                       uint32_t nret, target_ulong rets)
200 {
201     uint64_t buid;
202     uint32_t val, size, addr;
203 
204     if ((nargs != 5) || (nret != 1)) {
205         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
206         return;
207     }
208 
209     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
210     val = rtas_ld(args, 4);
211     size = rtas_ld(args, 3);
212     addr = rtas_ld(args, 0);
213 
214     finish_write_pci_config(spapr, buid, addr, size, val, rets);
215 }
216 
217 static void rtas_write_pci_config(PowerPCCPU *cpu, sPAPRMachineState *spapr,
218                                   uint32_t token, uint32_t nargs,
219                                   target_ulong args,
220                                   uint32_t nret, target_ulong rets)
221 {
222     uint32_t val, size, addr;
223 
224     if ((nargs != 3) || (nret != 1)) {
225         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
226         return;
227     }
228 
229 
230     val = rtas_ld(args, 2);
231     size = rtas_ld(args, 1);
232     addr = rtas_ld(args, 0);
233 
234     finish_write_pci_config(spapr, 0, addr, size, val, rets);
235 }
236 
237 /*
238  * Set MSI/MSIX message data.
239  * This is required for msi_notify()/msix_notify() which
240  * will write at the addresses via spapr_msi_write().
241  *
242  * If hwaddr == 0, all entries will have .data == first_irq i.e.
243  * table will be reset.
244  */
245 static void spapr_msi_setmsg(PCIDevice *pdev, hwaddr addr, bool msix,
246                              unsigned first_irq, unsigned req_num)
247 {
248     unsigned i;
249     MSIMessage msg = { .address = addr, .data = first_irq };
250 
251     if (!msix) {
252         msi_set_message(pdev, msg);
253         trace_spapr_pci_msi_setup(pdev->name, 0, msg.address);
254         return;
255     }
256 
257     for (i = 0; i < req_num; ++i) {
258         msix_set_message(pdev, i, msg);
259         trace_spapr_pci_msi_setup(pdev->name, i, msg.address);
260         if (addr) {
261             ++msg.data;
262         }
263     }
264 }
265 
266 static void rtas_ibm_change_msi(PowerPCCPU *cpu, sPAPRMachineState *spapr,
267                                 uint32_t token, uint32_t nargs,
268                                 target_ulong args, uint32_t nret,
269                                 target_ulong rets)
270 {
271     uint32_t config_addr = rtas_ld(args, 0);
272     uint64_t buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
273     unsigned int func = rtas_ld(args, 3);
274     unsigned int req_num = rtas_ld(args, 4); /* 0 == remove all */
275     unsigned int seq_num = rtas_ld(args, 5);
276     unsigned int ret_intr_type;
277     unsigned int irq, max_irqs = 0, num = 0;
278     sPAPRPHBState *phb = NULL;
279     PCIDevice *pdev = NULL;
280     spapr_pci_msi *msi;
281     int *config_addr_key;
282 
283     switch (func) {
284     case RTAS_CHANGE_MSI_FN:
285     case RTAS_CHANGE_FN:
286         ret_intr_type = RTAS_TYPE_MSI;
287         break;
288     case RTAS_CHANGE_MSIX_FN:
289         ret_intr_type = RTAS_TYPE_MSIX;
290         break;
291     default:
292         error_report("rtas_ibm_change_msi(%u) is not implemented", func);
293         rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
294         return;
295     }
296 
297     /* Fins sPAPRPHBState */
298     phb = spapr_pci_find_phb(spapr, buid);
299     if (phb) {
300         pdev = spapr_pci_find_dev(spapr, buid, config_addr);
301     }
302     if (!phb || !pdev) {
303         rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
304         return;
305     }
306 
307     /* Releasing MSIs */
308     if (!req_num) {
309         msi = (spapr_pci_msi *) g_hash_table_lookup(phb->msi, &config_addr);
310         if (!msi) {
311             trace_spapr_pci_msi("Releasing wrong config", config_addr);
312             rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
313             return;
314         }
315 
316         xics_free(spapr->icp, msi->first_irq, msi->num);
317         if (msi_present(pdev)) {
318             spapr_msi_setmsg(pdev, 0, false, 0, num);
319         }
320         if (msix_present(pdev)) {
321             spapr_msi_setmsg(pdev, 0, true, 0, num);
322         }
323         g_hash_table_remove(phb->msi, &config_addr);
324 
325         trace_spapr_pci_msi("Released MSIs", config_addr);
326         rtas_st(rets, 0, RTAS_OUT_SUCCESS);
327         rtas_st(rets, 1, 0);
328         return;
329     }
330 
331     /* Enabling MSI */
332 
333     /* Check if the device supports as many IRQs as requested */
334     if (ret_intr_type == RTAS_TYPE_MSI) {
335         max_irqs = msi_nr_vectors_allocated(pdev);
336     } else if (ret_intr_type == RTAS_TYPE_MSIX) {
337         max_irqs = pdev->msix_entries_nr;
338     }
339     if (!max_irqs) {
340         error_report("Requested interrupt type %d is not enabled for device %x",
341                      ret_intr_type, config_addr);
342         rtas_st(rets, 0, -1); /* Hardware error */
343         return;
344     }
345     /* Correct the number if the guest asked for too many */
346     if (req_num > max_irqs) {
347         trace_spapr_pci_msi_retry(config_addr, req_num, max_irqs);
348         req_num = max_irqs;
349         irq = 0; /* to avoid misleading trace */
350         goto out;
351     }
352 
353     /* Allocate MSIs */
354     irq = xics_alloc_block(spapr->icp, 0, req_num, false,
355                            ret_intr_type == RTAS_TYPE_MSI);
356     if (!irq) {
357         error_report("Cannot allocate MSIs for device %x", config_addr);
358         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
359         return;
360     }
361 
362     /* Setup MSI/MSIX vectors in the device (via cfgspace or MSIX BAR) */
363     spapr_msi_setmsg(pdev, SPAPR_PCI_MSI_WINDOW, ret_intr_type == RTAS_TYPE_MSIX,
364                      irq, req_num);
365 
366     /* Add MSI device to cache */
367     msi = g_new(spapr_pci_msi, 1);
368     msi->first_irq = irq;
369     msi->num = req_num;
370     config_addr_key = g_new(int, 1);
371     *config_addr_key = config_addr;
372     g_hash_table_insert(phb->msi, config_addr_key, msi);
373 
374 out:
375     rtas_st(rets, 0, RTAS_OUT_SUCCESS);
376     rtas_st(rets, 1, req_num);
377     rtas_st(rets, 2, ++seq_num);
378     rtas_st(rets, 3, ret_intr_type);
379 
380     trace_spapr_pci_rtas_ibm_change_msi(config_addr, func, req_num, irq);
381 }
382 
383 static void rtas_ibm_query_interrupt_source_number(PowerPCCPU *cpu,
384                                                    sPAPRMachineState *spapr,
385                                                    uint32_t token,
386                                                    uint32_t nargs,
387                                                    target_ulong args,
388                                                    uint32_t nret,
389                                                    target_ulong rets)
390 {
391     uint32_t config_addr = rtas_ld(args, 0);
392     uint64_t buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
393     unsigned int intr_src_num = -1, ioa_intr_num = rtas_ld(args, 3);
394     sPAPRPHBState *phb = NULL;
395     PCIDevice *pdev = NULL;
396     spapr_pci_msi *msi;
397 
398     /* Find sPAPRPHBState */
399     phb = spapr_pci_find_phb(spapr, buid);
400     if (phb) {
401         pdev = spapr_pci_find_dev(spapr, buid, config_addr);
402     }
403     if (!phb || !pdev) {
404         rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
405         return;
406     }
407 
408     /* Find device descriptor and start IRQ */
409     msi = (spapr_pci_msi *) g_hash_table_lookup(phb->msi, &config_addr);
410     if (!msi || !msi->first_irq || !msi->num || (ioa_intr_num >= msi->num)) {
411         trace_spapr_pci_msi("Failed to return vector", config_addr);
412         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
413         return;
414     }
415     intr_src_num = msi->first_irq + ioa_intr_num;
416     trace_spapr_pci_rtas_ibm_query_interrupt_source_number(ioa_intr_num,
417                                                            intr_src_num);
418 
419     rtas_st(rets, 0, RTAS_OUT_SUCCESS);
420     rtas_st(rets, 1, intr_src_num);
421     rtas_st(rets, 2, 1);/* 0 == level; 1 == edge */
422 }
423 
424 static void rtas_ibm_set_eeh_option(PowerPCCPU *cpu,
425                                     sPAPRMachineState *spapr,
426                                     uint32_t token, uint32_t nargs,
427                                     target_ulong args, uint32_t nret,
428                                     target_ulong rets)
429 {
430     sPAPRPHBState *sphb;
431     sPAPRPHBClass *spc;
432     PCIDevice *pdev;
433     uint32_t addr, option;
434     uint64_t buid;
435     int ret;
436 
437     if ((nargs != 4) || (nret != 1)) {
438         goto param_error_exit;
439     }
440 
441     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
442     addr = rtas_ld(args, 0);
443     option = rtas_ld(args, 3);
444 
445     sphb = spapr_pci_find_phb(spapr, buid);
446     if (!sphb) {
447         goto param_error_exit;
448     }
449 
450     pdev = pci_find_device(PCI_HOST_BRIDGE(sphb)->bus,
451                            (addr >> 16) & 0xFF, (addr >> 8) & 0xFF);
452     if (!pdev || !object_dynamic_cast(OBJECT(pdev), "vfio-pci")) {
453         goto param_error_exit;
454     }
455 
456     spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
457     if (!spc->eeh_set_option) {
458         goto param_error_exit;
459     }
460 
461     ret = spc->eeh_set_option(sphb, addr, option);
462     rtas_st(rets, 0, ret);
463     return;
464 
465 param_error_exit:
466     rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
467 }
468 
469 static void rtas_ibm_get_config_addr_info2(PowerPCCPU *cpu,
470                                            sPAPRMachineState *spapr,
471                                            uint32_t token, uint32_t nargs,
472                                            target_ulong args, uint32_t nret,
473                                            target_ulong rets)
474 {
475     sPAPRPHBState *sphb;
476     sPAPRPHBClass *spc;
477     PCIDevice *pdev;
478     uint32_t addr, option;
479     uint64_t buid;
480 
481     if ((nargs != 4) || (nret != 2)) {
482         goto param_error_exit;
483     }
484 
485     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
486     sphb = spapr_pci_find_phb(spapr, buid);
487     if (!sphb) {
488         goto param_error_exit;
489     }
490 
491     spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
492     if (!spc->eeh_set_option) {
493         goto param_error_exit;
494     }
495 
496     /*
497      * We always have PE address of form "00BB0001". "BB"
498      * represents the bus number of PE's primary bus.
499      */
500     option = rtas_ld(args, 3);
501     switch (option) {
502     case RTAS_GET_PE_ADDR:
503         addr = rtas_ld(args, 0);
504         pdev = spapr_pci_find_dev(spapr, buid, addr);
505         if (!pdev) {
506             goto param_error_exit;
507         }
508 
509         rtas_st(rets, 1, (pci_bus_num(pdev->bus) << 16) + 1);
510         break;
511     case RTAS_GET_PE_MODE:
512         rtas_st(rets, 1, RTAS_PE_MODE_SHARED);
513         break;
514     default:
515         goto param_error_exit;
516     }
517 
518     rtas_st(rets, 0, RTAS_OUT_SUCCESS);
519     return;
520 
521 param_error_exit:
522     rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
523 }
524 
525 static void rtas_ibm_read_slot_reset_state2(PowerPCCPU *cpu,
526                                             sPAPRMachineState *spapr,
527                                             uint32_t token, uint32_t nargs,
528                                             target_ulong args, uint32_t nret,
529                                             target_ulong rets)
530 {
531     sPAPRPHBState *sphb;
532     sPAPRPHBClass *spc;
533     uint64_t buid;
534     int state, ret;
535 
536     if ((nargs != 3) || (nret != 4 && nret != 5)) {
537         goto param_error_exit;
538     }
539 
540     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
541     sphb = spapr_pci_find_phb(spapr, buid);
542     if (!sphb) {
543         goto param_error_exit;
544     }
545 
546     spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
547     if (!spc->eeh_get_state) {
548         goto param_error_exit;
549     }
550 
551     ret = spc->eeh_get_state(sphb, &state);
552     rtas_st(rets, 0, ret);
553     if (ret != RTAS_OUT_SUCCESS) {
554         return;
555     }
556 
557     rtas_st(rets, 1, state);
558     rtas_st(rets, 2, RTAS_EEH_SUPPORT);
559     rtas_st(rets, 3, RTAS_EEH_PE_UNAVAIL_INFO);
560     if (nret >= 5) {
561         rtas_st(rets, 4, RTAS_EEH_PE_RECOVER_INFO);
562     }
563     return;
564 
565 param_error_exit:
566     rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
567 }
568 
569 static void rtas_ibm_set_slot_reset(PowerPCCPU *cpu,
570                                     sPAPRMachineState *spapr,
571                                     uint32_t token, uint32_t nargs,
572                                     target_ulong args, uint32_t nret,
573                                     target_ulong rets)
574 {
575     sPAPRPHBState *sphb;
576     sPAPRPHBClass *spc;
577     uint32_t option;
578     uint64_t buid;
579     int ret;
580 
581     if ((nargs != 4) || (nret != 1)) {
582         goto param_error_exit;
583     }
584 
585     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
586     option = rtas_ld(args, 3);
587     sphb = spapr_pci_find_phb(spapr, buid);
588     if (!sphb) {
589         goto param_error_exit;
590     }
591 
592     spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
593     if (!spc->eeh_reset) {
594         goto param_error_exit;
595     }
596 
597     ret = spc->eeh_reset(sphb, option);
598     rtas_st(rets, 0, ret);
599     return;
600 
601 param_error_exit:
602     rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
603 }
604 
605 static void rtas_ibm_configure_pe(PowerPCCPU *cpu,
606                                   sPAPRMachineState *spapr,
607                                   uint32_t token, uint32_t nargs,
608                                   target_ulong args, uint32_t nret,
609                                   target_ulong rets)
610 {
611     sPAPRPHBState *sphb;
612     sPAPRPHBClass *spc;
613     uint64_t buid;
614     int ret;
615 
616     if ((nargs != 3) || (nret != 1)) {
617         goto param_error_exit;
618     }
619 
620     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
621     sphb = spapr_pci_find_phb(spapr, buid);
622     if (!sphb) {
623         goto param_error_exit;
624     }
625 
626     spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
627     if (!spc->eeh_configure) {
628         goto param_error_exit;
629     }
630 
631     ret = spc->eeh_configure(sphb);
632     rtas_st(rets, 0, ret);
633     return;
634 
635 param_error_exit:
636     rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
637 }
638 
639 /* To support it later */
640 static void rtas_ibm_slot_error_detail(PowerPCCPU *cpu,
641                                        sPAPRMachineState *spapr,
642                                        uint32_t token, uint32_t nargs,
643                                        target_ulong args, uint32_t nret,
644                                        target_ulong rets)
645 {
646     sPAPRPHBState *sphb;
647     sPAPRPHBClass *spc;
648     int option;
649     uint64_t buid;
650 
651     if ((nargs != 8) || (nret != 1)) {
652         goto param_error_exit;
653     }
654 
655     buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
656     sphb = spapr_pci_find_phb(spapr, buid);
657     if (!sphb) {
658         goto param_error_exit;
659     }
660 
661     spc = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(sphb);
662     if (!spc->eeh_set_option) {
663         goto param_error_exit;
664     }
665 
666     option = rtas_ld(args, 7);
667     switch (option) {
668     case RTAS_SLOT_TEMP_ERR_LOG:
669     case RTAS_SLOT_PERM_ERR_LOG:
670         break;
671     default:
672         goto param_error_exit;
673     }
674 
675     /* We don't have error log yet */
676     rtas_st(rets, 0, RTAS_OUT_NO_ERRORS_FOUND);
677     return;
678 
679 param_error_exit:
680     rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
681 }
682 
683 static int pci_spapr_swizzle(int slot, int pin)
684 {
685     return (slot + pin) % PCI_NUM_PINS;
686 }
687 
688 static int pci_spapr_map_irq(PCIDevice *pci_dev, int irq_num)
689 {
690     /*
691      * Here we need to convert pci_dev + irq_num to some unique value
692      * which is less than number of IRQs on the specific bus (4).  We
693      * use standard PCI swizzling, that is (slot number + pin number)
694      * % 4.
695      */
696     return pci_spapr_swizzle(PCI_SLOT(pci_dev->devfn), irq_num);
697 }
698 
699 static void pci_spapr_set_irq(void *opaque, int irq_num, int level)
700 {
701     /*
702      * Here we use the number returned by pci_spapr_map_irq to find a
703      * corresponding qemu_irq.
704      */
705     sPAPRPHBState *phb = opaque;
706 
707     trace_spapr_pci_lsi_set(phb->dtbusname, irq_num, phb->lsi_table[irq_num].irq);
708     qemu_set_irq(spapr_phb_lsi_qirq(phb, irq_num), level);
709 }
710 
711 static PCIINTxRoute spapr_route_intx_pin_to_irq(void *opaque, int pin)
712 {
713     sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(opaque);
714     PCIINTxRoute route;
715 
716     route.mode = PCI_INTX_ENABLED;
717     route.irq = sphb->lsi_table[pin].irq;
718 
719     return route;
720 }
721 
722 /*
723  * MSI/MSIX memory region implementation.
724  * The handler handles both MSI and MSIX.
725  * For MSI-X, the vector number is encoded as a part of the address,
726  * data is set to 0.
727  * For MSI, the vector number is encoded in least bits in data.
728  */
729 static void spapr_msi_write(void *opaque, hwaddr addr,
730                             uint64_t data, unsigned size)
731 {
732     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
733     uint32_t irq = data;
734 
735     trace_spapr_pci_msi_write(addr, data, irq);
736 
737     qemu_irq_pulse(xics_get_qirq(spapr->icp, irq));
738 }
739 
740 static const MemoryRegionOps spapr_msi_ops = {
741     /* There is no .read as the read result is undefined by PCI spec */
742     .read = NULL,
743     .write = spapr_msi_write,
744     .endianness = DEVICE_LITTLE_ENDIAN
745 };
746 
747 /*
748  * PHB PCI device
749  */
750 static AddressSpace *spapr_pci_dma_iommu(PCIBus *bus, void *opaque, int devfn)
751 {
752     sPAPRPHBState *phb = opaque;
753 
754     return &phb->iommu_as;
755 }
756 
757 static char *spapr_phb_vfio_get_loc_code(sPAPRPHBState *sphb,  PCIDevice *pdev)
758 {
759     char *path = NULL, *buf = NULL, *host = NULL;
760 
761     /* Get the PCI VFIO host id */
762     host = object_property_get_str(OBJECT(pdev), "host", NULL);
763     if (!host) {
764         goto err_out;
765     }
766 
767     /* Construct the path of the file that will give us the DT location */
768     path = g_strdup_printf("/sys/bus/pci/devices/%s/devspec", host);
769     g_free(host);
770     if (!path || !g_file_get_contents(path, &buf, NULL, NULL)) {
771         goto err_out;
772     }
773     g_free(path);
774 
775     /* Construct and read from host device tree the loc-code */
776     path = g_strdup_printf("/proc/device-tree%s/ibm,loc-code", buf);
777     g_free(buf);
778     if (!path || !g_file_get_contents(path, &buf, NULL, NULL)) {
779         goto err_out;
780     }
781     return buf;
782 
783 err_out:
784     g_free(path);
785     return NULL;
786 }
787 
788 static char *spapr_phb_get_loc_code(sPAPRPHBState *sphb, PCIDevice *pdev)
789 {
790     char *buf;
791     const char *devtype = "qemu";
792     uint32_t busnr = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(pdev))));
793 
794     if (object_dynamic_cast(OBJECT(pdev), "vfio-pci")) {
795         buf = spapr_phb_vfio_get_loc_code(sphb, pdev);
796         if (buf) {
797             return buf;
798         }
799         devtype = "vfio";
800     }
801     /*
802      * For emulated devices and VFIO-failure case, make up
803      * the loc-code.
804      */
805     buf = g_strdup_printf("%s_%s:%04x:%02x:%02x.%x",
806                           devtype, pdev->name, sphb->index, busnr,
807                           PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
808     return buf;
809 }
810 
811 /* Macros to operate with address in OF binding to PCI */
812 #define b_x(x, p, l)    (((x) & ((1<<(l))-1)) << (p))
813 #define b_n(x)          b_x((x), 31, 1) /* 0 if relocatable */
814 #define b_p(x)          b_x((x), 30, 1) /* 1 if prefetchable */
815 #define b_t(x)          b_x((x), 29, 1) /* 1 if the address is aliased */
816 #define b_ss(x)         b_x((x), 24, 2) /* the space code */
817 #define b_bbbbbbbb(x)   b_x((x), 16, 8) /* bus number */
818 #define b_ddddd(x)      b_x((x), 11, 5) /* device number */
819 #define b_fff(x)        b_x((x), 8, 3)  /* function number */
820 #define b_rrrrrrrr(x)   b_x((x), 0, 8)  /* register number */
821 
822 /* for 'reg'/'assigned-addresses' OF properties */
823 #define RESOURCE_CELLS_SIZE 2
824 #define RESOURCE_CELLS_ADDRESS 3
825 
826 typedef struct ResourceFields {
827     uint32_t phys_hi;
828     uint32_t phys_mid;
829     uint32_t phys_lo;
830     uint32_t size_hi;
831     uint32_t size_lo;
832 } QEMU_PACKED ResourceFields;
833 
834 typedef struct ResourceProps {
835     ResourceFields reg[8];
836     ResourceFields assigned[7];
837     uint32_t reg_len;
838     uint32_t assigned_len;
839 } ResourceProps;
840 
841 /* fill in the 'reg'/'assigned-resources' OF properties for
842  * a PCI device. 'reg' describes resource requirements for a
843  * device's IO/MEM regions, 'assigned-addresses' describes the
844  * actual resource assignments.
845  *
846  * the properties are arrays of ('phys-addr', 'size') pairs describing
847  * the addressable regions of the PCI device, where 'phys-addr' is a
848  * RESOURCE_CELLS_ADDRESS-tuple of 32-bit integers corresponding to
849  * (phys.hi, phys.mid, phys.lo), and 'size' is a
850  * RESOURCE_CELLS_SIZE-tuple corresponding to (size.hi, size.lo).
851  *
852  * phys.hi = 0xYYXXXXZZ, where:
853  *   0xYY = npt000ss
854  *          |||   |
855  *          |||   +-- space code
856  *          |||               |
857  *          |||               +  00 if configuration space
858  *          |||               +  01 if IO region,
859  *          |||               +  10 if 32-bit MEM region
860  *          |||               +  11 if 64-bit MEM region
861  *          |||
862  *          ||+------ for non-relocatable IO: 1 if aliased
863  *          ||        for relocatable IO: 1 if below 64KB
864  *          ||        for MEM: 1 if below 1MB
865  *          |+------- 1 if region is prefetchable
866  *          +-------- 1 if region is non-relocatable
867  *   0xXXXX = bbbbbbbb dddddfff, encoding bus, slot, and function
868  *            bits respectively
869  *   0xZZ = rrrrrrrr, the register number of the BAR corresponding
870  *          to the region
871  *
872  * phys.mid and phys.lo correspond respectively to the hi/lo portions
873  * of the actual address of the region.
874  *
875  * how the phys-addr/size values are used differ slightly between
876  * 'reg' and 'assigned-addresses' properties. namely, 'reg' has
877  * an additional description for the config space region of the
878  * device, and in the case of QEMU has n=0 and phys.mid=phys.lo=0
879  * to describe the region as relocatable, with an address-mapping
880  * that corresponds directly to the PHB's address space for the
881  * resource. 'assigned-addresses' always has n=1 set with an absolute
882  * address assigned for the resource. in general, 'assigned-addresses'
883  * won't be populated, since addresses for PCI devices are generally
884  * unmapped initially and left to the guest to assign.
885  *
886  * note also that addresses defined in these properties are, at least
887  * for PAPR guests, relative to the PHBs IO/MEM windows, and
888  * correspond directly to the addresses in the BARs.
889  *
890  * in accordance with PCI Bus Binding to Open Firmware,
891  * IEEE Std 1275-1994, section 4.1.1, as implemented by PAPR+ v2.7,
892  * Appendix C.
893  */
894 static void populate_resource_props(PCIDevice *d, ResourceProps *rp)
895 {
896     int bus_num = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(d))));
897     uint32_t dev_id = (b_bbbbbbbb(bus_num) |
898                        b_ddddd(PCI_SLOT(d->devfn)) |
899                        b_fff(PCI_FUNC(d->devfn)));
900     ResourceFields *reg, *assigned;
901     int i, reg_idx = 0, assigned_idx = 0;
902 
903     /* config space region */
904     reg = &rp->reg[reg_idx++];
905     reg->phys_hi = cpu_to_be32(dev_id);
906     reg->phys_mid = 0;
907     reg->phys_lo = 0;
908     reg->size_hi = 0;
909     reg->size_lo = 0;
910 
911     for (i = 0; i < PCI_NUM_REGIONS; i++) {
912         if (!d->io_regions[i].size) {
913             continue;
914         }
915 
916         reg = &rp->reg[reg_idx++];
917 
918         reg->phys_hi = cpu_to_be32(dev_id | b_rrrrrrrr(pci_bar(d, i)));
919         if (d->io_regions[i].type & PCI_BASE_ADDRESS_SPACE_IO) {
920             reg->phys_hi |= cpu_to_be32(b_ss(1));
921         } else if (d->io_regions[i].type & PCI_BASE_ADDRESS_MEM_TYPE_64) {
922             reg->phys_hi |= cpu_to_be32(b_ss(3));
923         } else {
924             reg->phys_hi |= cpu_to_be32(b_ss(2));
925         }
926         reg->phys_mid = 0;
927         reg->phys_lo = 0;
928         reg->size_hi = cpu_to_be32(d->io_regions[i].size >> 32);
929         reg->size_lo = cpu_to_be32(d->io_regions[i].size);
930 
931         if (d->io_regions[i].addr == PCI_BAR_UNMAPPED) {
932             continue;
933         }
934 
935         assigned = &rp->assigned[assigned_idx++];
936         assigned->phys_hi = cpu_to_be32(reg->phys_hi | b_n(1));
937         assigned->phys_mid = cpu_to_be32(d->io_regions[i].addr >> 32);
938         assigned->phys_lo = cpu_to_be32(d->io_regions[i].addr);
939         assigned->size_hi = reg->size_hi;
940         assigned->size_lo = reg->size_lo;
941     }
942 
943     rp->reg_len = reg_idx * sizeof(ResourceFields);
944     rp->assigned_len = assigned_idx * sizeof(ResourceFields);
945 }
946 
947 static uint32_t spapr_phb_get_pci_drc_index(sPAPRPHBState *phb,
948                                             PCIDevice *pdev);
949 
950 static int spapr_populate_pci_child_dt(PCIDevice *dev, void *fdt, int offset,
951                                        sPAPRPHBState *sphb)
952 {
953     ResourceProps rp;
954     bool is_bridge = false;
955     int pci_status, err;
956     char *buf = NULL;
957     uint32_t drc_index = spapr_phb_get_pci_drc_index(sphb, dev);
958 
959     if (pci_default_read_config(dev, PCI_HEADER_TYPE, 1) ==
960         PCI_HEADER_TYPE_BRIDGE) {
961         is_bridge = true;
962     }
963 
964     /* in accordance with PAPR+ v2.7 13.6.3, Table 181 */
965     _FDT(fdt_setprop_cell(fdt, offset, "vendor-id",
966                           pci_default_read_config(dev, PCI_VENDOR_ID, 2)));
967     _FDT(fdt_setprop_cell(fdt, offset, "device-id",
968                           pci_default_read_config(dev, PCI_DEVICE_ID, 2)));
969     _FDT(fdt_setprop_cell(fdt, offset, "revision-id",
970                           pci_default_read_config(dev, PCI_REVISION_ID, 1)));
971     _FDT(fdt_setprop_cell(fdt, offset, "class-code",
972                           pci_default_read_config(dev, PCI_CLASS_PROG, 3)));
973     if (pci_default_read_config(dev, PCI_INTERRUPT_PIN, 1)) {
974         _FDT(fdt_setprop_cell(fdt, offset, "interrupts",
975                  pci_default_read_config(dev, PCI_INTERRUPT_PIN, 1)));
976     }
977 
978     if (!is_bridge) {
979         _FDT(fdt_setprop_cell(fdt, offset, "min-grant",
980             pci_default_read_config(dev, PCI_MIN_GNT, 1)));
981         _FDT(fdt_setprop_cell(fdt, offset, "max-latency",
982             pci_default_read_config(dev, PCI_MAX_LAT, 1)));
983     }
984 
985     if (pci_default_read_config(dev, PCI_SUBSYSTEM_ID, 2)) {
986         _FDT(fdt_setprop_cell(fdt, offset, "subsystem-id",
987                  pci_default_read_config(dev, PCI_SUBSYSTEM_ID, 2)));
988     }
989 
990     if (pci_default_read_config(dev, PCI_SUBSYSTEM_VENDOR_ID, 2)) {
991         _FDT(fdt_setprop_cell(fdt, offset, "subsystem-vendor-id",
992                  pci_default_read_config(dev, PCI_SUBSYSTEM_VENDOR_ID, 2)));
993     }
994 
995     _FDT(fdt_setprop_cell(fdt, offset, "cache-line-size",
996         pci_default_read_config(dev, PCI_CACHE_LINE_SIZE, 1)));
997 
998     /* the following fdt cells are masked off the pci status register */
999     pci_status = pci_default_read_config(dev, PCI_STATUS, 2);
1000     _FDT(fdt_setprop_cell(fdt, offset, "devsel-speed",
1001                           PCI_STATUS_DEVSEL_MASK & pci_status));
1002 
1003     if (pci_status & PCI_STATUS_FAST_BACK) {
1004         _FDT(fdt_setprop(fdt, offset, "fast-back-to-back", NULL, 0));
1005     }
1006     if (pci_status & PCI_STATUS_66MHZ) {
1007         _FDT(fdt_setprop(fdt, offset, "66mhz-capable", NULL, 0));
1008     }
1009     if (pci_status & PCI_STATUS_UDF) {
1010         _FDT(fdt_setprop(fdt, offset, "udf-supported", NULL, 0));
1011     }
1012 
1013     /* NOTE: this is normally generated by firmware via path/unit name,
1014      * but in our case we must set it manually since it does not get
1015      * processed by OF beforehand
1016      */
1017     _FDT(fdt_setprop_string(fdt, offset, "name", "pci"));
1018     buf = spapr_phb_get_loc_code(sphb, dev);
1019     if (!buf) {
1020         error_report("Failed setting the ibm,loc-code");
1021         return -1;
1022     }
1023 
1024     err = fdt_setprop_string(fdt, offset, "ibm,loc-code", buf);
1025     g_free(buf);
1026     if (err < 0) {
1027         return err;
1028     }
1029 
1030     if (drc_index) {
1031         _FDT(fdt_setprop_cell(fdt, offset, "ibm,my-drc-index", drc_index));
1032     }
1033 
1034     _FDT(fdt_setprop_cell(fdt, offset, "#address-cells",
1035                           RESOURCE_CELLS_ADDRESS));
1036     _FDT(fdt_setprop_cell(fdt, offset, "#size-cells",
1037                           RESOURCE_CELLS_SIZE));
1038     _FDT(fdt_setprop_cell(fdt, offset, "ibm,req#msi-x",
1039                           RESOURCE_CELLS_SIZE));
1040 
1041     populate_resource_props(dev, &rp);
1042     _FDT(fdt_setprop(fdt, offset, "reg", (uint8_t *)rp.reg, rp.reg_len));
1043     _FDT(fdt_setprop(fdt, offset, "assigned-addresses",
1044                      (uint8_t *)rp.assigned, rp.assigned_len));
1045 
1046     return 0;
1047 }
1048 
1049 /* create OF node for pci device and required OF DT properties */
1050 static int spapr_create_pci_child_dt(sPAPRPHBState *phb, PCIDevice *dev,
1051                                      void *fdt, int node_offset)
1052 {
1053     int offset, ret;
1054     int slot = PCI_SLOT(dev->devfn);
1055     int func = PCI_FUNC(dev->devfn);
1056     char nodename[FDT_NAME_MAX];
1057 
1058     if (func != 0) {
1059         snprintf(nodename, FDT_NAME_MAX, "pci@%x,%x", slot, func);
1060     } else {
1061         snprintf(nodename, FDT_NAME_MAX, "pci@%x", slot);
1062     }
1063     offset = fdt_add_subnode(fdt, node_offset, nodename);
1064     ret = spapr_populate_pci_child_dt(dev, fdt, offset, phb);
1065 
1066     g_assert(!ret);
1067     if (ret) {
1068         return 0;
1069     }
1070     return offset;
1071 }
1072 
1073 static void spapr_phb_add_pci_device(sPAPRDRConnector *drc,
1074                                      sPAPRPHBState *phb,
1075                                      PCIDevice *pdev,
1076                                      Error **errp)
1077 {
1078     sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
1079     DeviceState *dev = DEVICE(pdev);
1080     void *fdt = NULL;
1081     int fdt_start_offset = 0, fdt_size;
1082 
1083     if (dev->hotplugged) {
1084         fdt = create_device_tree(&fdt_size);
1085         fdt_start_offset = spapr_create_pci_child_dt(phb, pdev, fdt, 0);
1086         if (!fdt_start_offset) {
1087             error_setg(errp, "Failed to create pci child device tree node");
1088             goto out;
1089         }
1090     }
1091 
1092     drck->attach(drc, DEVICE(pdev),
1093                  fdt, fdt_start_offset, !dev->hotplugged, errp);
1094 out:
1095     if (*errp) {
1096         g_free(fdt);
1097     }
1098 }
1099 
1100 static void spapr_phb_remove_pci_device_cb(DeviceState *dev, void *opaque)
1101 {
1102     /* some version guests do not wait for completion of a device
1103      * cleanup (generally done asynchronously by the kernel) before
1104      * signaling to QEMU that the device is safe, but instead sleep
1105      * for some 'safe' period of time. unfortunately on a busy host
1106      * this sleep isn't guaranteed to be long enough, resulting in
1107      * bad things like IRQ lines being left asserted during final
1108      * device removal. to deal with this we call reset just prior
1109      * to finalizing the device, which will put the device back into
1110      * an 'idle' state, as the device cleanup code expects.
1111      */
1112     pci_device_reset(PCI_DEVICE(dev));
1113     object_unparent(OBJECT(dev));
1114 }
1115 
1116 static void spapr_phb_remove_pci_device(sPAPRDRConnector *drc,
1117                                         sPAPRPHBState *phb,
1118                                         PCIDevice *pdev,
1119                                         Error **errp)
1120 {
1121     sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
1122 
1123     drck->detach(drc, DEVICE(pdev), spapr_phb_remove_pci_device_cb, phb, errp);
1124 }
1125 
1126 static sPAPRDRConnector *spapr_phb_get_pci_drc(sPAPRPHBState *phb,
1127                                                PCIDevice *pdev)
1128 {
1129     uint32_t busnr = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(pdev))));
1130     return spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_PCI,
1131                                     (phb->index << 16) |
1132                                     (busnr << 8) |
1133                                     pdev->devfn);
1134 }
1135 
1136 static uint32_t spapr_phb_get_pci_drc_index(sPAPRPHBState *phb,
1137                                             PCIDevice *pdev)
1138 {
1139     sPAPRDRConnector *drc = spapr_phb_get_pci_drc(phb, pdev);
1140     sPAPRDRConnectorClass *drck;
1141 
1142     if (!drc) {
1143         return 0;
1144     }
1145 
1146     drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
1147     return drck->get_index(drc);
1148 }
1149 
1150 static void spapr_phb_hot_plug_child(HotplugHandler *plug_handler,
1151                                      DeviceState *plugged_dev, Error **errp)
1152 {
1153     sPAPRPHBState *phb = SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler));
1154     PCIDevice *pdev = PCI_DEVICE(plugged_dev);
1155     sPAPRDRConnector *drc = spapr_phb_get_pci_drc(phb, pdev);
1156     Error *local_err = NULL;
1157 
1158     /* if DR is disabled we don't need to do anything in the case of
1159      * hotplug or coldplug callbacks
1160      */
1161     if (!phb->dr_enabled) {
1162         /* if this is a hotplug operation initiated by the user
1163          * we need to let them know it's not enabled
1164          */
1165         if (plugged_dev->hotplugged) {
1166             error_setg(errp, QERR_BUS_NO_HOTPLUG,
1167                        object_get_typename(OBJECT(phb)));
1168         }
1169         return;
1170     }
1171 
1172     g_assert(drc);
1173 
1174     spapr_phb_add_pci_device(drc, phb, pdev, &local_err);
1175     if (local_err) {
1176         error_propagate(errp, local_err);
1177         return;
1178     }
1179     if (plugged_dev->hotplugged) {
1180         spapr_hotplug_req_add_event(drc);
1181     }
1182 }
1183 
1184 static void spapr_phb_hot_unplug_child(HotplugHandler *plug_handler,
1185                                        DeviceState *plugged_dev, Error **errp)
1186 {
1187     sPAPRPHBState *phb = SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler));
1188     PCIDevice *pdev = PCI_DEVICE(plugged_dev);
1189     sPAPRDRConnectorClass *drck;
1190     sPAPRDRConnector *drc = spapr_phb_get_pci_drc(phb, pdev);
1191     Error *local_err = NULL;
1192 
1193     if (!phb->dr_enabled) {
1194         error_setg(errp, QERR_BUS_NO_HOTPLUG,
1195                    object_get_typename(OBJECT(phb)));
1196         return;
1197     }
1198 
1199     g_assert(drc);
1200 
1201     drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
1202     if (!drck->release_pending(drc)) {
1203         spapr_phb_remove_pci_device(drc, phb, pdev, &local_err);
1204         if (local_err) {
1205             error_propagate(errp, local_err);
1206             return;
1207         }
1208         spapr_hotplug_req_remove_event(drc);
1209     }
1210 }
1211 
1212 static void spapr_phb_realize(DeviceState *dev, Error **errp)
1213 {
1214     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1215     SysBusDevice *s = SYS_BUS_DEVICE(dev);
1216     sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(s);
1217     PCIHostState *phb = PCI_HOST_BRIDGE(s);
1218     sPAPRPHBClass *info = SPAPR_PCI_HOST_BRIDGE_GET_CLASS(s);
1219     char *namebuf;
1220     int i;
1221     PCIBus *bus;
1222     uint64_t msi_window_size = 4096;
1223 
1224     if (sphb->index != (uint32_t)-1) {
1225         hwaddr windows_base;
1226 
1227         if ((sphb->buid != (uint64_t)-1) || (sphb->dma_liobn != (uint32_t)-1)
1228             || (sphb->mem_win_addr != (hwaddr)-1)
1229             || (sphb->io_win_addr != (hwaddr)-1)) {
1230             error_setg(errp, "Either \"index\" or other parameters must"
1231                        " be specified for PAPR PHB, not both");
1232             return;
1233         }
1234 
1235         if (sphb->index > SPAPR_PCI_MAX_INDEX) {
1236             error_setg(errp, "\"index\" for PAPR PHB is too large (max %u)",
1237                        SPAPR_PCI_MAX_INDEX);
1238             return;
1239         }
1240 
1241         sphb->buid = SPAPR_PCI_BASE_BUID + sphb->index;
1242         sphb->dma_liobn = SPAPR_PCI_LIOBN(sphb->index, 0);
1243 
1244         windows_base = SPAPR_PCI_WINDOW_BASE
1245             + sphb->index * SPAPR_PCI_WINDOW_SPACING;
1246         sphb->mem_win_addr = windows_base + SPAPR_PCI_MMIO_WIN_OFF;
1247         sphb->io_win_addr = windows_base + SPAPR_PCI_IO_WIN_OFF;
1248     }
1249 
1250     if (sphb->buid == (uint64_t)-1) {
1251         error_setg(errp, "BUID not specified for PHB");
1252         return;
1253     }
1254 
1255     if (sphb->dma_liobn == (uint32_t)-1) {
1256         error_setg(errp, "LIOBN not specified for PHB");
1257         return;
1258     }
1259 
1260     if (sphb->mem_win_addr == (hwaddr)-1) {
1261         error_setg(errp, "Memory window address not specified for PHB");
1262         return;
1263     }
1264 
1265     if (sphb->io_win_addr == (hwaddr)-1) {
1266         error_setg(errp, "IO window address not specified for PHB");
1267         return;
1268     }
1269 
1270     if (spapr_pci_find_phb(spapr, sphb->buid)) {
1271         error_setg(errp, "PCI host bridges must have unique BUIDs");
1272         return;
1273     }
1274 
1275     sphb->dtbusname = g_strdup_printf("pci@%" PRIx64, sphb->buid);
1276 
1277     namebuf = alloca(strlen(sphb->dtbusname) + 32);
1278 
1279     /* Initialize memory regions */
1280     sprintf(namebuf, "%s.mmio", sphb->dtbusname);
1281     memory_region_init(&sphb->memspace, OBJECT(sphb), namebuf, UINT64_MAX);
1282 
1283     sprintf(namebuf, "%s.mmio-alias", sphb->dtbusname);
1284     memory_region_init_alias(&sphb->memwindow, OBJECT(sphb),
1285                              namebuf, &sphb->memspace,
1286                              SPAPR_PCI_MEM_WIN_BUS_OFFSET, sphb->mem_win_size);
1287     memory_region_add_subregion(get_system_memory(), sphb->mem_win_addr,
1288                                 &sphb->memwindow);
1289 
1290     /* Initialize IO regions */
1291     sprintf(namebuf, "%s.io", sphb->dtbusname);
1292     memory_region_init(&sphb->iospace, OBJECT(sphb),
1293                        namebuf, SPAPR_PCI_IO_WIN_SIZE);
1294 
1295     sprintf(namebuf, "%s.io-alias", sphb->dtbusname);
1296     memory_region_init_alias(&sphb->iowindow, OBJECT(sphb), namebuf,
1297                              &sphb->iospace, 0, SPAPR_PCI_IO_WIN_SIZE);
1298     memory_region_add_subregion(get_system_memory(), sphb->io_win_addr,
1299                                 &sphb->iowindow);
1300 
1301     bus = pci_register_bus(dev, NULL,
1302                            pci_spapr_set_irq, pci_spapr_map_irq, sphb,
1303                            &sphb->memspace, &sphb->iospace,
1304                            PCI_DEVFN(0, 0), PCI_NUM_PINS, TYPE_PCI_BUS);
1305     phb->bus = bus;
1306     qbus_set_hotplug_handler(BUS(phb->bus), DEVICE(sphb), NULL);
1307 
1308     /*
1309      * Initialize PHB address space.
1310      * By default there will be at least one subregion for default
1311      * 32bit DMA window.
1312      * Later the guest might want to create another DMA window
1313      * which will become another memory subregion.
1314      */
1315     sprintf(namebuf, "%s.iommu-root", sphb->dtbusname);
1316 
1317     memory_region_init(&sphb->iommu_root, OBJECT(sphb),
1318                        namebuf, UINT64_MAX);
1319     address_space_init(&sphb->iommu_as, &sphb->iommu_root,
1320                        sphb->dtbusname);
1321 
1322     /*
1323      * As MSI/MSIX interrupts trigger by writing at MSI/MSIX vectors,
1324      * we need to allocate some memory to catch those writes coming
1325      * from msi_notify()/msix_notify().
1326      * As MSIMessage:addr is going to be the same and MSIMessage:data
1327      * is going to be a VIRQ number, 4 bytes of the MSI MR will only
1328      * be used.
1329      *
1330      * For KVM we want to ensure that this memory is a full page so that
1331      * our memory slot is of page size granularity.
1332      */
1333 #ifdef CONFIG_KVM
1334     if (kvm_enabled()) {
1335         msi_window_size = getpagesize();
1336     }
1337 #endif
1338 
1339     memory_region_init_io(&sphb->msiwindow, NULL, &spapr_msi_ops, spapr,
1340                           "msi", msi_window_size);
1341     memory_region_add_subregion(&sphb->iommu_root, SPAPR_PCI_MSI_WINDOW,
1342                                 &sphb->msiwindow);
1343 
1344     pci_setup_iommu(bus, spapr_pci_dma_iommu, sphb);
1345 
1346     pci_bus_set_route_irq_fn(bus, spapr_route_intx_pin_to_irq);
1347 
1348     QLIST_INSERT_HEAD(&spapr->phbs, sphb, list);
1349 
1350     /* Initialize the LSI table */
1351     for (i = 0; i < PCI_NUM_PINS; i++) {
1352         uint32_t irq;
1353 
1354         irq = xics_alloc_block(spapr->icp, 0, 1, true, false);
1355         if (!irq) {
1356             error_setg(errp, "spapr_allocate_lsi failed");
1357             return;
1358         }
1359 
1360         sphb->lsi_table[i].irq = irq;
1361     }
1362 
1363     /* allocate connectors for child PCI devices */
1364     if (sphb->dr_enabled) {
1365         for (i = 0; i < PCI_SLOT_MAX * 8; i++) {
1366             spapr_dr_connector_new(OBJECT(phb),
1367                                    SPAPR_DR_CONNECTOR_TYPE_PCI,
1368                                    (sphb->index << 16) | i);
1369         }
1370     }
1371 
1372     if (!info->finish_realize) {
1373         error_setg(errp, "finish_realize not defined");
1374         return;
1375     }
1376 
1377     info->finish_realize(sphb, errp);
1378 
1379     sphb->msi = g_hash_table_new_full(g_int_hash, g_int_equal, g_free, g_free);
1380 }
1381 
1382 static void spapr_phb_finish_realize(sPAPRPHBState *sphb, Error **errp)
1383 {
1384     sPAPRTCETable *tcet;
1385     uint32_t nb_table;
1386 
1387     nb_table = SPAPR_PCI_DMA32_SIZE >> SPAPR_TCE_PAGE_SHIFT;
1388     tcet = spapr_tce_new_table(DEVICE(sphb), sphb->dma_liobn,
1389                                0, SPAPR_TCE_PAGE_SHIFT, nb_table, false);
1390     if (!tcet) {
1391         error_setg(errp, "Unable to create TCE table for %s",
1392                    sphb->dtbusname);
1393         return ;
1394     }
1395 
1396     /* Register default 32bit DMA window */
1397     memory_region_add_subregion(&sphb->iommu_root, 0,
1398                                 spapr_tce_get_iommu(tcet));
1399 }
1400 
1401 static int spapr_phb_children_reset(Object *child, void *opaque)
1402 {
1403     DeviceState *dev = (DeviceState *) object_dynamic_cast(child, TYPE_DEVICE);
1404 
1405     if (dev) {
1406         device_reset(dev);
1407     }
1408 
1409     return 0;
1410 }
1411 
1412 static void spapr_phb_reset(DeviceState *qdev)
1413 {
1414     /* Reset the IOMMU state */
1415     object_child_foreach(OBJECT(qdev), spapr_phb_children_reset, NULL);
1416 }
1417 
1418 static Property spapr_phb_properties[] = {
1419     DEFINE_PROP_UINT32("index", sPAPRPHBState, index, -1),
1420     DEFINE_PROP_UINT64("buid", sPAPRPHBState, buid, -1),
1421     DEFINE_PROP_UINT32("liobn", sPAPRPHBState, dma_liobn, -1),
1422     DEFINE_PROP_UINT64("mem_win_addr", sPAPRPHBState, mem_win_addr, -1),
1423     DEFINE_PROP_UINT64("mem_win_size", sPAPRPHBState, mem_win_size,
1424                        SPAPR_PCI_MMIO_WIN_SIZE),
1425     DEFINE_PROP_UINT64("io_win_addr", sPAPRPHBState, io_win_addr, -1),
1426     DEFINE_PROP_UINT64("io_win_size", sPAPRPHBState, io_win_size,
1427                        SPAPR_PCI_IO_WIN_SIZE),
1428     DEFINE_PROP_BOOL("dynamic-reconfiguration", sPAPRPHBState, dr_enabled,
1429                      true),
1430     DEFINE_PROP_END_OF_LIST(),
1431 };
1432 
1433 static const VMStateDescription vmstate_spapr_pci_lsi = {
1434     .name = "spapr_pci/lsi",
1435     .version_id = 1,
1436     .minimum_version_id = 1,
1437     .fields = (VMStateField[]) {
1438         VMSTATE_UINT32_EQUAL(irq, struct spapr_pci_lsi),
1439 
1440         VMSTATE_END_OF_LIST()
1441     },
1442 };
1443 
1444 static const VMStateDescription vmstate_spapr_pci_msi = {
1445     .name = "spapr_pci/msi",
1446     .version_id = 1,
1447     .minimum_version_id = 1,
1448     .fields = (VMStateField []) {
1449         VMSTATE_UINT32(key, spapr_pci_msi_mig),
1450         VMSTATE_UINT32(value.first_irq, spapr_pci_msi_mig),
1451         VMSTATE_UINT32(value.num, spapr_pci_msi_mig),
1452         VMSTATE_END_OF_LIST()
1453     },
1454 };
1455 
1456 static void spapr_pci_pre_save(void *opaque)
1457 {
1458     sPAPRPHBState *sphb = opaque;
1459     GHashTableIter iter;
1460     gpointer key, value;
1461     int i;
1462 
1463     g_free(sphb->msi_devs);
1464     sphb->msi_devs = NULL;
1465     sphb->msi_devs_num = g_hash_table_size(sphb->msi);
1466     if (!sphb->msi_devs_num) {
1467         return;
1468     }
1469     sphb->msi_devs = g_malloc(sphb->msi_devs_num * sizeof(spapr_pci_msi_mig));
1470 
1471     g_hash_table_iter_init(&iter, sphb->msi);
1472     for (i = 0; g_hash_table_iter_next(&iter, &key, &value); ++i) {
1473         sphb->msi_devs[i].key = *(uint32_t *) key;
1474         sphb->msi_devs[i].value = *(spapr_pci_msi *) value;
1475     }
1476 }
1477 
1478 static int spapr_pci_post_load(void *opaque, int version_id)
1479 {
1480     sPAPRPHBState *sphb = opaque;
1481     gpointer key, value;
1482     int i;
1483 
1484     for (i = 0; i < sphb->msi_devs_num; ++i) {
1485         key = g_memdup(&sphb->msi_devs[i].key,
1486                        sizeof(sphb->msi_devs[i].key));
1487         value = g_memdup(&sphb->msi_devs[i].value,
1488                          sizeof(sphb->msi_devs[i].value));
1489         g_hash_table_insert(sphb->msi, key, value);
1490     }
1491     g_free(sphb->msi_devs);
1492     sphb->msi_devs = NULL;
1493     sphb->msi_devs_num = 0;
1494 
1495     return 0;
1496 }
1497 
1498 static const VMStateDescription vmstate_spapr_pci = {
1499     .name = "spapr_pci",
1500     .version_id = 2,
1501     .minimum_version_id = 2,
1502     .pre_save = spapr_pci_pre_save,
1503     .post_load = spapr_pci_post_load,
1504     .fields = (VMStateField[]) {
1505         VMSTATE_UINT64_EQUAL(buid, sPAPRPHBState),
1506         VMSTATE_UINT32_EQUAL(dma_liobn, sPAPRPHBState),
1507         VMSTATE_UINT64_EQUAL(mem_win_addr, sPAPRPHBState),
1508         VMSTATE_UINT64_EQUAL(mem_win_size, sPAPRPHBState),
1509         VMSTATE_UINT64_EQUAL(io_win_addr, sPAPRPHBState),
1510         VMSTATE_UINT64_EQUAL(io_win_size, sPAPRPHBState),
1511         VMSTATE_STRUCT_ARRAY(lsi_table, sPAPRPHBState, PCI_NUM_PINS, 0,
1512                              vmstate_spapr_pci_lsi, struct spapr_pci_lsi),
1513         VMSTATE_INT32(msi_devs_num, sPAPRPHBState),
1514         VMSTATE_STRUCT_VARRAY_ALLOC(msi_devs, sPAPRPHBState, msi_devs_num, 0,
1515                                     vmstate_spapr_pci_msi, spapr_pci_msi_mig),
1516         VMSTATE_END_OF_LIST()
1517     },
1518 };
1519 
1520 static const char *spapr_phb_root_bus_path(PCIHostState *host_bridge,
1521                                            PCIBus *rootbus)
1522 {
1523     sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(host_bridge);
1524 
1525     return sphb->dtbusname;
1526 }
1527 
1528 static void spapr_phb_class_init(ObjectClass *klass, void *data)
1529 {
1530     PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass);
1531     DeviceClass *dc = DEVICE_CLASS(klass);
1532     sPAPRPHBClass *spc = SPAPR_PCI_HOST_BRIDGE_CLASS(klass);
1533     HotplugHandlerClass *hp = HOTPLUG_HANDLER_CLASS(klass);
1534 
1535     hc->root_bus_path = spapr_phb_root_bus_path;
1536     dc->realize = spapr_phb_realize;
1537     dc->props = spapr_phb_properties;
1538     dc->reset = spapr_phb_reset;
1539     dc->vmsd = &vmstate_spapr_pci;
1540     set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
1541     dc->cannot_instantiate_with_device_add_yet = false;
1542     spc->finish_realize = spapr_phb_finish_realize;
1543     hp->plug = spapr_phb_hot_plug_child;
1544     hp->unplug = spapr_phb_hot_unplug_child;
1545 }
1546 
1547 static const TypeInfo spapr_phb_info = {
1548     .name          = TYPE_SPAPR_PCI_HOST_BRIDGE,
1549     .parent        = TYPE_PCI_HOST_BRIDGE,
1550     .instance_size = sizeof(sPAPRPHBState),
1551     .class_init    = spapr_phb_class_init,
1552     .class_size    = sizeof(sPAPRPHBClass),
1553     .interfaces    = (InterfaceInfo[]) {
1554         { TYPE_HOTPLUG_HANDLER },
1555         { }
1556     }
1557 };
1558 
1559 PCIHostState *spapr_create_phb(sPAPRMachineState *spapr, int index)
1560 {
1561     DeviceState *dev;
1562 
1563     dev = qdev_create(NULL, TYPE_SPAPR_PCI_HOST_BRIDGE);
1564     qdev_prop_set_uint32(dev, "index", index);
1565     qdev_init_nofail(dev);
1566 
1567     return PCI_HOST_BRIDGE(dev);
1568 }
1569 
1570 typedef struct sPAPRFDT {
1571     void *fdt;
1572     int node_off;
1573     sPAPRPHBState *sphb;
1574 } sPAPRFDT;
1575 
1576 static void spapr_populate_pci_devices_dt(PCIBus *bus, PCIDevice *pdev,
1577                                           void *opaque)
1578 {
1579     PCIBus *sec_bus;
1580     sPAPRFDT *p = opaque;
1581     int offset;
1582     sPAPRFDT s_fdt;
1583 
1584     offset = spapr_create_pci_child_dt(p->sphb, pdev, p->fdt, p->node_off);
1585     if (!offset) {
1586         error_report("Failed to create pci child device tree node");
1587         return;
1588     }
1589 
1590     if ((pci_default_read_config(pdev, PCI_HEADER_TYPE, 1) !=
1591          PCI_HEADER_TYPE_BRIDGE)) {
1592         return;
1593     }
1594 
1595     sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
1596     if (!sec_bus) {
1597         return;
1598     }
1599 
1600     s_fdt.fdt = p->fdt;
1601     s_fdt.node_off = offset;
1602     s_fdt.sphb = p->sphb;
1603     pci_for_each_device(sec_bus, pci_bus_num(sec_bus),
1604                         spapr_populate_pci_devices_dt,
1605                         &s_fdt);
1606 }
1607 
1608 static void spapr_phb_pci_enumerate_bridge(PCIBus *bus, PCIDevice *pdev,
1609                                            void *opaque)
1610 {
1611     unsigned int *bus_no = opaque;
1612     unsigned int primary = *bus_no;
1613     unsigned int subordinate = 0xff;
1614     PCIBus *sec_bus = NULL;
1615 
1616     if ((pci_default_read_config(pdev, PCI_HEADER_TYPE, 1) !=
1617          PCI_HEADER_TYPE_BRIDGE)) {
1618         return;
1619     }
1620 
1621     (*bus_no)++;
1622     pci_default_write_config(pdev, PCI_PRIMARY_BUS, primary, 1);
1623     pci_default_write_config(pdev, PCI_SECONDARY_BUS, *bus_no, 1);
1624     pci_default_write_config(pdev, PCI_SUBORDINATE_BUS, *bus_no, 1);
1625 
1626     sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
1627     if (!sec_bus) {
1628         return;
1629     }
1630 
1631     pci_default_write_config(pdev, PCI_SUBORDINATE_BUS, subordinate, 1);
1632     pci_for_each_device(sec_bus, pci_bus_num(sec_bus),
1633                         spapr_phb_pci_enumerate_bridge, bus_no);
1634     pci_default_write_config(pdev, PCI_SUBORDINATE_BUS, *bus_no, 1);
1635 }
1636 
1637 static void spapr_phb_pci_enumerate(sPAPRPHBState *phb)
1638 {
1639     PCIBus *bus = PCI_HOST_BRIDGE(phb)->bus;
1640     unsigned int bus_no = 0;
1641 
1642     pci_for_each_device(bus, pci_bus_num(bus),
1643                         spapr_phb_pci_enumerate_bridge,
1644                         &bus_no);
1645 
1646 }
1647 
1648 int spapr_populate_pci_dt(sPAPRPHBState *phb,
1649                           uint32_t xics_phandle,
1650                           void *fdt)
1651 {
1652     int bus_off, i, j, ret;
1653     char nodename[FDT_NAME_MAX];
1654     uint32_t bus_range[] = { cpu_to_be32(0), cpu_to_be32(0xff) };
1655     const uint64_t mmiosize = memory_region_size(&phb->memwindow);
1656     const uint64_t w32max = (1ULL << 32) - SPAPR_PCI_MEM_WIN_BUS_OFFSET;
1657     const uint64_t w32size = MIN(w32max, mmiosize);
1658     const uint64_t w64size = (mmiosize > w32size) ? (mmiosize - w32size) : 0;
1659     struct {
1660         uint32_t hi;
1661         uint64_t child;
1662         uint64_t parent;
1663         uint64_t size;
1664     } QEMU_PACKED ranges[] = {
1665         {
1666             cpu_to_be32(b_ss(1)), cpu_to_be64(0),
1667             cpu_to_be64(phb->io_win_addr),
1668             cpu_to_be64(memory_region_size(&phb->iospace)),
1669         },
1670         {
1671             cpu_to_be32(b_ss(2)), cpu_to_be64(SPAPR_PCI_MEM_WIN_BUS_OFFSET),
1672             cpu_to_be64(phb->mem_win_addr),
1673             cpu_to_be64(w32size),
1674         },
1675         {
1676             cpu_to_be32(b_ss(3)), cpu_to_be64(1ULL << 32),
1677             cpu_to_be64(phb->mem_win_addr + w32size),
1678             cpu_to_be64(w64size)
1679         },
1680     };
1681     const unsigned sizeof_ranges = (w64size ? 3 : 2) * sizeof(ranges[0]);
1682     uint64_t bus_reg[] = { cpu_to_be64(phb->buid), 0 };
1683     uint32_t interrupt_map_mask[] = {
1684         cpu_to_be32(b_ddddd(-1)|b_fff(0)), 0x0, 0x0, cpu_to_be32(-1)};
1685     uint32_t interrupt_map[PCI_SLOT_MAX * PCI_NUM_PINS][7];
1686     sPAPRTCETable *tcet;
1687     PCIBus *bus = PCI_HOST_BRIDGE(phb)->bus;
1688     sPAPRFDT s_fdt;
1689 
1690     /* Start populating the FDT */
1691     snprintf(nodename, FDT_NAME_MAX, "pci@%" PRIx64, phb->buid);
1692     bus_off = fdt_add_subnode(fdt, 0, nodename);
1693     if (bus_off < 0) {
1694         return bus_off;
1695     }
1696 
1697     /* Write PHB properties */
1698     _FDT(fdt_setprop_string(fdt, bus_off, "device_type", "pci"));
1699     _FDT(fdt_setprop_string(fdt, bus_off, "compatible", "IBM,Logical_PHB"));
1700     _FDT(fdt_setprop_cell(fdt, bus_off, "#address-cells", 0x3));
1701     _FDT(fdt_setprop_cell(fdt, bus_off, "#size-cells", 0x2));
1702     _FDT(fdt_setprop_cell(fdt, bus_off, "#interrupt-cells", 0x1));
1703     _FDT(fdt_setprop(fdt, bus_off, "used-by-rtas", NULL, 0));
1704     _FDT(fdt_setprop(fdt, bus_off, "bus-range", &bus_range, sizeof(bus_range)));
1705     _FDT(fdt_setprop(fdt, bus_off, "ranges", &ranges, sizeof_ranges));
1706     _FDT(fdt_setprop(fdt, bus_off, "reg", &bus_reg, sizeof(bus_reg)));
1707     _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pci-config-space-type", 0x1));
1708     _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pe-total-#msi", XICS_IRQS));
1709 
1710     /* Build the interrupt-map, this must matches what is done
1711      * in pci_spapr_map_irq
1712      */
1713     _FDT(fdt_setprop(fdt, bus_off, "interrupt-map-mask",
1714                      &interrupt_map_mask, sizeof(interrupt_map_mask)));
1715     for (i = 0; i < PCI_SLOT_MAX; i++) {
1716         for (j = 0; j < PCI_NUM_PINS; j++) {
1717             uint32_t *irqmap = interrupt_map[i*PCI_NUM_PINS + j];
1718             int lsi_num = pci_spapr_swizzle(i, j);
1719 
1720             irqmap[0] = cpu_to_be32(b_ddddd(i)|b_fff(0));
1721             irqmap[1] = 0;
1722             irqmap[2] = 0;
1723             irqmap[3] = cpu_to_be32(j+1);
1724             irqmap[4] = cpu_to_be32(xics_phandle);
1725             irqmap[5] = cpu_to_be32(phb->lsi_table[lsi_num].irq);
1726             irqmap[6] = cpu_to_be32(0x8);
1727         }
1728     }
1729     /* Write interrupt map */
1730     _FDT(fdt_setprop(fdt, bus_off, "interrupt-map", &interrupt_map,
1731                      sizeof(interrupt_map)));
1732 
1733     tcet = spapr_tce_find_by_liobn(SPAPR_PCI_LIOBN(phb->index, 0));
1734     spapr_dma_dt(fdt, bus_off, "ibm,dma-window",
1735                  tcet->liobn, tcet->bus_offset,
1736                  tcet->nb_table << tcet->page_shift);
1737 
1738     /* Walk the bridges and program the bus numbers*/
1739     spapr_phb_pci_enumerate(phb);
1740     _FDT(fdt_setprop_cell(fdt, bus_off, "qemu,phb-enumerated", 0x1));
1741 
1742     /* Populate tree nodes with PCI devices attached */
1743     s_fdt.fdt = fdt;
1744     s_fdt.node_off = bus_off;
1745     s_fdt.sphb = phb;
1746     pci_for_each_device(bus, pci_bus_num(bus),
1747                         spapr_populate_pci_devices_dt,
1748                         &s_fdt);
1749 
1750     ret = spapr_drc_populate_dt(fdt, bus_off, OBJECT(phb),
1751                                 SPAPR_DR_CONNECTOR_TYPE_PCI);
1752     if (ret) {
1753         return ret;
1754     }
1755 
1756     return 0;
1757 }
1758 
1759 void spapr_pci_rtas_init(void)
1760 {
1761     spapr_rtas_register(RTAS_READ_PCI_CONFIG, "read-pci-config",
1762                         rtas_read_pci_config);
1763     spapr_rtas_register(RTAS_WRITE_PCI_CONFIG, "write-pci-config",
1764                         rtas_write_pci_config);
1765     spapr_rtas_register(RTAS_IBM_READ_PCI_CONFIG, "ibm,read-pci-config",
1766                         rtas_ibm_read_pci_config);
1767     spapr_rtas_register(RTAS_IBM_WRITE_PCI_CONFIG, "ibm,write-pci-config",
1768                         rtas_ibm_write_pci_config);
1769     if (msi_supported) {
1770         spapr_rtas_register(RTAS_IBM_QUERY_INTERRUPT_SOURCE_NUMBER,
1771                             "ibm,query-interrupt-source-number",
1772                             rtas_ibm_query_interrupt_source_number);
1773         spapr_rtas_register(RTAS_IBM_CHANGE_MSI, "ibm,change-msi",
1774                             rtas_ibm_change_msi);
1775     }
1776 
1777     spapr_rtas_register(RTAS_IBM_SET_EEH_OPTION,
1778                         "ibm,set-eeh-option",
1779                         rtas_ibm_set_eeh_option);
1780     spapr_rtas_register(RTAS_IBM_GET_CONFIG_ADDR_INFO2,
1781                         "ibm,get-config-addr-info2",
1782                         rtas_ibm_get_config_addr_info2);
1783     spapr_rtas_register(RTAS_IBM_READ_SLOT_RESET_STATE2,
1784                         "ibm,read-slot-reset-state2",
1785                         rtas_ibm_read_slot_reset_state2);
1786     spapr_rtas_register(RTAS_IBM_SET_SLOT_RESET,
1787                         "ibm,set-slot-reset",
1788                         rtas_ibm_set_slot_reset);
1789     spapr_rtas_register(RTAS_IBM_CONFIGURE_PE,
1790                         "ibm,configure-pe",
1791                         rtas_ibm_configure_pe);
1792     spapr_rtas_register(RTAS_IBM_SLOT_ERROR_DETAIL,
1793                         "ibm,slot-error-detail",
1794                         rtas_ibm_slot_error_detail);
1795 }
1796 
1797 static void spapr_pci_register_types(void)
1798 {
1799     type_register_static(&spapr_phb_info);
1800 }
1801 
1802 type_init(spapr_pci_register_types)
1803 
1804 static int spapr_switch_one_vga(DeviceState *dev, void *opaque)
1805 {
1806     bool be = *(bool *)opaque;
1807 
1808     if (object_dynamic_cast(OBJECT(dev), "VGA")
1809         || object_dynamic_cast(OBJECT(dev), "secondary-vga")) {
1810         object_property_set_bool(OBJECT(dev), be, "big-endian-framebuffer",
1811                                  &error_abort);
1812     }
1813     return 0;
1814 }
1815 
1816 void spapr_pci_switch_vga(bool big_endian)
1817 {
1818     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1819     sPAPRPHBState *sphb;
1820 
1821     /*
1822      * For backward compatibility with existing guests, we switch
1823      * the endianness of the VGA controller when changing the guest
1824      * interrupt mode
1825      */
1826     QLIST_FOREACH(sphb, &spapr->phbs, list) {
1827         BusState *bus = &PCI_HOST_BRIDGE(sphb)->bus->qbus;
1828         qbus_walk_children(bus, spapr_switch_one_vga, NULL, NULL, NULL,
1829                            &big_endian);
1830     }
1831 }
1832