xref: /openbmc/qemu/hw/ppc/spapr_pci.c (revision 795c40b8)
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 "qemu/osdep.h"
26 #include "qapi/error.h"
27 #include "qemu-common.h"
28 #include "cpu.h"
29 #include "hw/hw.h"
30 #include "hw/sysbus.h"
31 #include "hw/pci/pci.h"
32 #include "hw/pci/msi.h"
33 #include "hw/pci/msix.h"
34 #include "hw/pci/pci_host.h"
35 #include "hw/ppc/spapr.h"
36 #include "hw/pci-host/spapr.h"
37 #include "exec/address-spaces.h"
38 #include "exec/ram_addr.h"
39 #include <libfdt.h>
40 #include "trace.h"
41 #include "qemu/error-report.h"
42 #include "qapi/qmp/qerror.h"
43 
44 #include "hw/pci/pci_bridge.h"
45 #include "hw/pci/pci_bus.h"
46 #include "hw/pci/pci_ids.h"
47 #include "hw/ppc/spapr_drc.h"
48 #include "sysemu/device_tree.h"
49 #include "sysemu/kvm.h"
50 #include "sysemu/hostmem.h"
51 #include "sysemu/numa.h"
52 
53 /* Copied from the kernel arch/powerpc/platforms/pseries/msi.c */
54 #define RTAS_QUERY_FN           0
55 #define RTAS_CHANGE_FN          1
56 #define RTAS_RESET_FN           2
57 #define RTAS_CHANGE_MSI_FN      3
58 #define RTAS_CHANGE_MSIX_FN     4
59 
60 /* Interrupt types to return on RTAS_CHANGE_* */
61 #define RTAS_TYPE_MSI           1
62 #define RTAS_TYPE_MSIX          2
63 
64 #define FDT_NAME_MAX          128
65 
66 #define _FDT(exp) \
67     do { \
68         int ret = (exp);                                           \
69         if (ret < 0) {                                             \
70             return ret;                                            \
71         }                                                          \
72     } while (0)
73 
74 sPAPRPHBState *spapr_pci_find_phb(sPAPRMachineState *spapr, uint64_t buid)
75 {
76     sPAPRPHBState *sphb;
77 
78     QLIST_FOREACH(sphb, &spapr->phbs, list) {
79         if (sphb->buid != buid) {
80             continue;
81         }
82         return sphb;
83     }
84 
85     return NULL;
86 }
87 
88 PCIDevice *spapr_pci_find_dev(sPAPRMachineState *spapr, uint64_t buid,
89                               uint32_t config_addr)
90 {
91     sPAPRPHBState *sphb = spapr_pci_find_phb(spapr, buid);
92     PCIHostState *phb = PCI_HOST_BRIDGE(sphb);
93     int bus_num = (config_addr >> 16) & 0xFF;
94     int devfn = (config_addr >> 8) & 0xFF;
95 
96     if (!phb) {
97         return NULL;
98     }
99 
100     return pci_find_device(phb->bus, bus_num, devfn);
101 }
102 
103 static uint32_t rtas_pci_cfgaddr(uint32_t arg)
104 {
105     /* This handles the encoding of extended config space addresses */
106     return ((arg >> 20) & 0xf00) | (arg & 0xff);
107 }
108 
109 static void finish_read_pci_config(sPAPRMachineState *spapr, uint64_t buid,
110                                    uint32_t addr, uint32_t size,
111                                    target_ulong rets)
112 {
113     PCIDevice *pci_dev;
114     uint32_t val;
115 
116     if ((size != 1) && (size != 2) && (size != 4)) {
117         /* access must be 1, 2 or 4 bytes */
118         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
119         return;
120     }
121 
122     pci_dev = spapr_pci_find_dev(spapr, buid, addr);
123     addr = rtas_pci_cfgaddr(addr);
124 
125     if (!pci_dev || (addr % size) || (addr >= pci_config_size(pci_dev))) {
126         /* Access must be to a valid device, within bounds and
127          * naturally aligned */
128         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
129         return;
130     }
131 
132     val = pci_host_config_read_common(pci_dev, addr,
133                                       pci_config_size(pci_dev), size);
134 
135     rtas_st(rets, 0, RTAS_OUT_SUCCESS);
136     rtas_st(rets, 1, val);
137 }
138 
139 static void rtas_ibm_read_pci_config(PowerPCCPU *cpu, sPAPRMachineState *spapr,
140                                      uint32_t token, uint32_t nargs,
141                                      target_ulong args,
142                                      uint32_t nret, target_ulong rets)
143 {
144     uint64_t buid;
145     uint32_t size, addr;
146 
147     if ((nargs != 4) || (nret != 2)) {
148         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
149         return;
150     }
151 
152     buid = rtas_ldq(args, 1);
153     size = rtas_ld(args, 3);
154     addr = rtas_ld(args, 0);
155 
156     finish_read_pci_config(spapr, buid, addr, size, rets);
157 }
158 
159 static void rtas_read_pci_config(PowerPCCPU *cpu, sPAPRMachineState *spapr,
160                                  uint32_t token, uint32_t nargs,
161                                  target_ulong args,
162                                  uint32_t nret, target_ulong rets)
163 {
164     uint32_t size, addr;
165 
166     if ((nargs != 2) || (nret != 2)) {
167         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
168         return;
169     }
170 
171     size = rtas_ld(args, 1);
172     addr = rtas_ld(args, 0);
173 
174     finish_read_pci_config(spapr, 0, addr, size, rets);
175 }
176 
177 static void finish_write_pci_config(sPAPRMachineState *spapr, uint64_t buid,
178                                     uint32_t addr, uint32_t size,
179                                     uint32_t val, target_ulong rets)
180 {
181     PCIDevice *pci_dev;
182 
183     if ((size != 1) && (size != 2) && (size != 4)) {
184         /* access must be 1, 2 or 4 bytes */
185         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
186         return;
187     }
188 
189     pci_dev = spapr_pci_find_dev(spapr, buid, addr);
190     addr = rtas_pci_cfgaddr(addr);
191 
192     if (!pci_dev || (addr % size) || (addr >= pci_config_size(pci_dev))) {
193         /* Access must be to a valid device, within bounds and
194          * naturally aligned */
195         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
196         return;
197     }
198 
199     pci_host_config_write_common(pci_dev, addr, pci_config_size(pci_dev),
200                                  val, size);
201 
202     rtas_st(rets, 0, RTAS_OUT_SUCCESS);
203 }
204 
205 static void rtas_ibm_write_pci_config(PowerPCCPU *cpu, sPAPRMachineState *spapr,
206                                       uint32_t token, uint32_t nargs,
207                                       target_ulong args,
208                                       uint32_t nret, target_ulong rets)
209 {
210     uint64_t buid;
211     uint32_t val, size, addr;
212 
213     if ((nargs != 5) || (nret != 1)) {
214         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
215         return;
216     }
217 
218     buid = rtas_ldq(args, 1);
219     val = rtas_ld(args, 4);
220     size = rtas_ld(args, 3);
221     addr = rtas_ld(args, 0);
222 
223     finish_write_pci_config(spapr, buid, addr, size, val, rets);
224 }
225 
226 static void rtas_write_pci_config(PowerPCCPU *cpu, sPAPRMachineState *spapr,
227                                   uint32_t token, uint32_t nargs,
228                                   target_ulong args,
229                                   uint32_t nret, target_ulong rets)
230 {
231     uint32_t val, size, addr;
232 
233     if ((nargs != 3) || (nret != 1)) {
234         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
235         return;
236     }
237 
238 
239     val = rtas_ld(args, 2);
240     size = rtas_ld(args, 1);
241     addr = rtas_ld(args, 0);
242 
243     finish_write_pci_config(spapr, 0, addr, size, val, rets);
244 }
245 
246 /*
247  * Set MSI/MSIX message data.
248  * This is required for msi_notify()/msix_notify() which
249  * will write at the addresses via spapr_msi_write().
250  *
251  * If hwaddr == 0, all entries will have .data == first_irq i.e.
252  * table will be reset.
253  */
254 static void spapr_msi_setmsg(PCIDevice *pdev, hwaddr addr, bool msix,
255                              unsigned first_irq, unsigned req_num)
256 {
257     unsigned i;
258     MSIMessage msg = { .address = addr, .data = first_irq };
259 
260     if (!msix) {
261         msi_set_message(pdev, msg);
262         trace_spapr_pci_msi_setup(pdev->name, 0, msg.address);
263         return;
264     }
265 
266     for (i = 0; i < req_num; ++i) {
267         msix_set_message(pdev, i, msg);
268         trace_spapr_pci_msi_setup(pdev->name, i, msg.address);
269         if (addr) {
270             ++msg.data;
271         }
272     }
273 }
274 
275 static void rtas_ibm_change_msi(PowerPCCPU *cpu, sPAPRMachineState *spapr,
276                                 uint32_t token, uint32_t nargs,
277                                 target_ulong args, uint32_t nret,
278                                 target_ulong rets)
279 {
280     uint32_t config_addr = rtas_ld(args, 0);
281     uint64_t buid = rtas_ldq(args, 1);
282     unsigned int func = rtas_ld(args, 3);
283     unsigned int req_num = rtas_ld(args, 4); /* 0 == remove all */
284     unsigned int seq_num = rtas_ld(args, 5);
285     unsigned int ret_intr_type;
286     unsigned int irq, max_irqs = 0;
287     sPAPRPHBState *phb = NULL;
288     PCIDevice *pdev = NULL;
289     spapr_pci_msi *msi;
290     int *config_addr_key;
291     Error *err = NULL;
292 
293     switch (func) {
294     case RTAS_CHANGE_MSI_FN:
295     case RTAS_CHANGE_FN:
296         ret_intr_type = RTAS_TYPE_MSI;
297         break;
298     case RTAS_CHANGE_MSIX_FN:
299         ret_intr_type = RTAS_TYPE_MSIX;
300         break;
301     default:
302         error_report("rtas_ibm_change_msi(%u) is not implemented", func);
303         rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
304         return;
305     }
306 
307     /* Fins sPAPRPHBState */
308     phb = spapr_pci_find_phb(spapr, buid);
309     if (phb) {
310         pdev = spapr_pci_find_dev(spapr, buid, config_addr);
311     }
312     if (!phb || !pdev) {
313         rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
314         return;
315     }
316 
317     msi = (spapr_pci_msi *) g_hash_table_lookup(phb->msi, &config_addr);
318 
319     /* Releasing MSIs */
320     if (!req_num) {
321         if (!msi) {
322             trace_spapr_pci_msi("Releasing wrong config", config_addr);
323             rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
324             return;
325         }
326 
327         spapr_ics_free(spapr->ics, msi->first_irq, msi->num);
328         if (msi_present(pdev)) {
329             spapr_msi_setmsg(pdev, 0, false, 0, 0);
330         }
331         if (msix_present(pdev)) {
332             spapr_msi_setmsg(pdev, 0, true, 0, 0);
333         }
334         g_hash_table_remove(phb->msi, &config_addr);
335 
336         trace_spapr_pci_msi("Released MSIs", config_addr);
337         rtas_st(rets, 0, RTAS_OUT_SUCCESS);
338         rtas_st(rets, 1, 0);
339         return;
340     }
341 
342     /* Enabling MSI */
343 
344     /* Check if the device supports as many IRQs as requested */
345     if (ret_intr_type == RTAS_TYPE_MSI) {
346         max_irqs = msi_nr_vectors_allocated(pdev);
347     } else if (ret_intr_type == RTAS_TYPE_MSIX) {
348         max_irqs = pdev->msix_entries_nr;
349     }
350     if (!max_irqs) {
351         error_report("Requested interrupt type %d is not enabled for device %x",
352                      ret_intr_type, config_addr);
353         rtas_st(rets, 0, -1); /* Hardware error */
354         return;
355     }
356     /* Correct the number if the guest asked for too many */
357     if (req_num > max_irqs) {
358         trace_spapr_pci_msi_retry(config_addr, req_num, max_irqs);
359         req_num = max_irqs;
360         irq = 0; /* to avoid misleading trace */
361         goto out;
362     }
363 
364     /* Allocate MSIs */
365     irq = spapr_ics_alloc_block(spapr->ics, req_num, false,
366                            ret_intr_type == RTAS_TYPE_MSI, &err);
367     if (err) {
368         error_reportf_err(err, "Can't allocate MSIs for device %x: ",
369                           config_addr);
370         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
371         return;
372     }
373 
374     /* Release previous MSIs */
375     if (msi) {
376         spapr_ics_free(spapr->ics, msi->first_irq, msi->num);
377         g_hash_table_remove(phb->msi, &config_addr);
378     }
379 
380     /* Setup MSI/MSIX vectors in the device (via cfgspace or MSIX BAR) */
381     spapr_msi_setmsg(pdev, SPAPR_PCI_MSI_WINDOW, ret_intr_type == RTAS_TYPE_MSIX,
382                      irq, req_num);
383 
384     /* Add MSI device to cache */
385     msi = g_new(spapr_pci_msi, 1);
386     msi->first_irq = irq;
387     msi->num = req_num;
388     config_addr_key = g_new(int, 1);
389     *config_addr_key = config_addr;
390     g_hash_table_insert(phb->msi, config_addr_key, msi);
391 
392 out:
393     rtas_st(rets, 0, RTAS_OUT_SUCCESS);
394     rtas_st(rets, 1, req_num);
395     rtas_st(rets, 2, ++seq_num);
396     if (nret > 3) {
397         rtas_st(rets, 3, ret_intr_type);
398     }
399 
400     trace_spapr_pci_rtas_ibm_change_msi(config_addr, func, req_num, irq);
401 }
402 
403 static void rtas_ibm_query_interrupt_source_number(PowerPCCPU *cpu,
404                                                    sPAPRMachineState *spapr,
405                                                    uint32_t token,
406                                                    uint32_t nargs,
407                                                    target_ulong args,
408                                                    uint32_t nret,
409                                                    target_ulong rets)
410 {
411     uint32_t config_addr = rtas_ld(args, 0);
412     uint64_t buid = rtas_ldq(args, 1);
413     unsigned int intr_src_num = -1, ioa_intr_num = rtas_ld(args, 3);
414     sPAPRPHBState *phb = NULL;
415     PCIDevice *pdev = NULL;
416     spapr_pci_msi *msi;
417 
418     /* Find sPAPRPHBState */
419     phb = spapr_pci_find_phb(spapr, buid);
420     if (phb) {
421         pdev = spapr_pci_find_dev(spapr, buid, config_addr);
422     }
423     if (!phb || !pdev) {
424         rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
425         return;
426     }
427 
428     /* Find device descriptor and start IRQ */
429     msi = (spapr_pci_msi *) g_hash_table_lookup(phb->msi, &config_addr);
430     if (!msi || !msi->first_irq || !msi->num || (ioa_intr_num >= msi->num)) {
431         trace_spapr_pci_msi("Failed to return vector", config_addr);
432         rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
433         return;
434     }
435     intr_src_num = msi->first_irq + ioa_intr_num;
436     trace_spapr_pci_rtas_ibm_query_interrupt_source_number(ioa_intr_num,
437                                                            intr_src_num);
438 
439     rtas_st(rets, 0, RTAS_OUT_SUCCESS);
440     rtas_st(rets, 1, intr_src_num);
441     rtas_st(rets, 2, 1);/* 0 == level; 1 == edge */
442 }
443 
444 static void rtas_ibm_set_eeh_option(PowerPCCPU *cpu,
445                                     sPAPRMachineState *spapr,
446                                     uint32_t token, uint32_t nargs,
447                                     target_ulong args, uint32_t nret,
448                                     target_ulong rets)
449 {
450     sPAPRPHBState *sphb;
451     uint32_t addr, option;
452     uint64_t buid;
453     int ret;
454 
455     if ((nargs != 4) || (nret != 1)) {
456         goto param_error_exit;
457     }
458 
459     buid = rtas_ldq(args, 1);
460     addr = rtas_ld(args, 0);
461     option = rtas_ld(args, 3);
462 
463     sphb = spapr_pci_find_phb(spapr, buid);
464     if (!sphb) {
465         goto param_error_exit;
466     }
467 
468     if (!spapr_phb_eeh_available(sphb)) {
469         goto param_error_exit;
470     }
471 
472     ret = spapr_phb_vfio_eeh_set_option(sphb, addr, option);
473     rtas_st(rets, 0, ret);
474     return;
475 
476 param_error_exit:
477     rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
478 }
479 
480 static void rtas_ibm_get_config_addr_info2(PowerPCCPU *cpu,
481                                            sPAPRMachineState *spapr,
482                                            uint32_t token, uint32_t nargs,
483                                            target_ulong args, uint32_t nret,
484                                            target_ulong rets)
485 {
486     sPAPRPHBState *sphb;
487     PCIDevice *pdev;
488     uint32_t addr, option;
489     uint64_t buid;
490 
491     if ((nargs != 4) || (nret != 2)) {
492         goto param_error_exit;
493     }
494 
495     buid = rtas_ldq(args, 1);
496     sphb = spapr_pci_find_phb(spapr, buid);
497     if (!sphb) {
498         goto param_error_exit;
499     }
500 
501     if (!spapr_phb_eeh_available(sphb)) {
502         goto param_error_exit;
503     }
504 
505     /*
506      * We always have PE address of form "00BB0001". "BB"
507      * represents the bus number of PE's primary bus.
508      */
509     option = rtas_ld(args, 3);
510     switch (option) {
511     case RTAS_GET_PE_ADDR:
512         addr = rtas_ld(args, 0);
513         pdev = spapr_pci_find_dev(spapr, buid, addr);
514         if (!pdev) {
515             goto param_error_exit;
516         }
517 
518         rtas_st(rets, 1, (pci_bus_num(pdev->bus) << 16) + 1);
519         break;
520     case RTAS_GET_PE_MODE:
521         rtas_st(rets, 1, RTAS_PE_MODE_SHARED);
522         break;
523     default:
524         goto param_error_exit;
525     }
526 
527     rtas_st(rets, 0, RTAS_OUT_SUCCESS);
528     return;
529 
530 param_error_exit:
531     rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
532 }
533 
534 static void rtas_ibm_read_slot_reset_state2(PowerPCCPU *cpu,
535                                             sPAPRMachineState *spapr,
536                                             uint32_t token, uint32_t nargs,
537                                             target_ulong args, uint32_t nret,
538                                             target_ulong rets)
539 {
540     sPAPRPHBState *sphb;
541     uint64_t buid;
542     int state, ret;
543 
544     if ((nargs != 3) || (nret != 4 && nret != 5)) {
545         goto param_error_exit;
546     }
547 
548     buid = rtas_ldq(args, 1);
549     sphb = spapr_pci_find_phb(spapr, buid);
550     if (!sphb) {
551         goto param_error_exit;
552     }
553 
554     if (!spapr_phb_eeh_available(sphb)) {
555         goto param_error_exit;
556     }
557 
558     ret = spapr_phb_vfio_eeh_get_state(sphb, &state);
559     rtas_st(rets, 0, ret);
560     if (ret != RTAS_OUT_SUCCESS) {
561         return;
562     }
563 
564     rtas_st(rets, 1, state);
565     rtas_st(rets, 2, RTAS_EEH_SUPPORT);
566     rtas_st(rets, 3, RTAS_EEH_PE_UNAVAIL_INFO);
567     if (nret >= 5) {
568         rtas_st(rets, 4, RTAS_EEH_PE_RECOVER_INFO);
569     }
570     return;
571 
572 param_error_exit:
573     rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
574 }
575 
576 static void rtas_ibm_set_slot_reset(PowerPCCPU *cpu,
577                                     sPAPRMachineState *spapr,
578                                     uint32_t token, uint32_t nargs,
579                                     target_ulong args, uint32_t nret,
580                                     target_ulong rets)
581 {
582     sPAPRPHBState *sphb;
583     uint32_t option;
584     uint64_t buid;
585     int ret;
586 
587     if ((nargs != 4) || (nret != 1)) {
588         goto param_error_exit;
589     }
590 
591     buid = rtas_ldq(args, 1);
592     option = rtas_ld(args, 3);
593     sphb = spapr_pci_find_phb(spapr, buid);
594     if (!sphb) {
595         goto param_error_exit;
596     }
597 
598     if (!spapr_phb_eeh_available(sphb)) {
599         goto param_error_exit;
600     }
601 
602     ret = spapr_phb_vfio_eeh_reset(sphb, option);
603     rtas_st(rets, 0, ret);
604     return;
605 
606 param_error_exit:
607     rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
608 }
609 
610 static void rtas_ibm_configure_pe(PowerPCCPU *cpu,
611                                   sPAPRMachineState *spapr,
612                                   uint32_t token, uint32_t nargs,
613                                   target_ulong args, uint32_t nret,
614                                   target_ulong rets)
615 {
616     sPAPRPHBState *sphb;
617     uint64_t buid;
618     int ret;
619 
620     if ((nargs != 3) || (nret != 1)) {
621         goto param_error_exit;
622     }
623 
624     buid = rtas_ldq(args, 1);
625     sphb = spapr_pci_find_phb(spapr, buid);
626     if (!sphb) {
627         goto param_error_exit;
628     }
629 
630     if (!spapr_phb_eeh_available(sphb)) {
631         goto param_error_exit;
632     }
633 
634     ret = spapr_phb_vfio_eeh_configure(sphb);
635     rtas_st(rets, 0, ret);
636     return;
637 
638 param_error_exit:
639     rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
640 }
641 
642 /* To support it later */
643 static void rtas_ibm_slot_error_detail(PowerPCCPU *cpu,
644                                        sPAPRMachineState *spapr,
645                                        uint32_t token, uint32_t nargs,
646                                        target_ulong args, uint32_t nret,
647                                        target_ulong rets)
648 {
649     sPAPRPHBState *sphb;
650     int option;
651     uint64_t buid;
652 
653     if ((nargs != 8) || (nret != 1)) {
654         goto param_error_exit;
655     }
656 
657     buid = rtas_ldq(args, 1);
658     sphb = spapr_pci_find_phb(spapr, buid);
659     if (!sphb) {
660         goto param_error_exit;
661     }
662 
663     if (!spapr_phb_eeh_available(sphb)) {
664         goto param_error_exit;
665     }
666 
667     option = rtas_ld(args, 7);
668     switch (option) {
669     case RTAS_SLOT_TEMP_ERR_LOG:
670     case RTAS_SLOT_PERM_ERR_LOG:
671         break;
672     default:
673         goto param_error_exit;
674     }
675 
676     /* We don't have error log yet */
677     rtas_st(rets, 0, RTAS_OUT_NO_ERRORS_FOUND);
678     return;
679 
680 param_error_exit:
681     rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
682 }
683 
684 static int pci_spapr_swizzle(int slot, int pin)
685 {
686     return (slot + pin) % PCI_NUM_PINS;
687 }
688 
689 static int pci_spapr_map_irq(PCIDevice *pci_dev, int irq_num)
690 {
691     /*
692      * Here we need to convert pci_dev + irq_num to some unique value
693      * which is less than number of IRQs on the specific bus (4).  We
694      * use standard PCI swizzling, that is (slot number + pin number)
695      * % 4.
696      */
697     return pci_spapr_swizzle(PCI_SLOT(pci_dev->devfn), irq_num);
698 }
699 
700 static void pci_spapr_set_irq(void *opaque, int irq_num, int level)
701 {
702     /*
703      * Here we use the number returned by pci_spapr_map_irq to find a
704      * corresponding qemu_irq.
705      */
706     sPAPRPHBState *phb = opaque;
707 
708     trace_spapr_pci_lsi_set(phb->dtbusname, irq_num, phb->lsi_table[irq_num].irq);
709     qemu_set_irq(spapr_phb_lsi_qirq(phb, irq_num), level);
710 }
711 
712 static PCIINTxRoute spapr_route_intx_pin_to_irq(void *opaque, int pin)
713 {
714     sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(opaque);
715     PCIINTxRoute route;
716 
717     route.mode = PCI_INTX_ENABLED;
718     route.irq = sphb->lsi_table[pin].irq;
719 
720     return route;
721 }
722 
723 /*
724  * MSI/MSIX memory region implementation.
725  * The handler handles both MSI and MSIX.
726  * For MSI-X, the vector number is encoded as a part of the address,
727  * data is set to 0.
728  * For MSI, the vector number is encoded in least bits in data.
729  */
730 static void spapr_msi_write(void *opaque, hwaddr addr,
731                             uint64_t data, unsigned size)
732 {
733     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
734     uint32_t irq = data;
735 
736     trace_spapr_pci_msi_write(addr, data, irq);
737 
738     qemu_irq_pulse(xics_get_qirq(XICS_FABRIC(spapr), irq));
739 }
740 
741 static const MemoryRegionOps spapr_msi_ops = {
742     /* There is no .read as the read result is undefined by PCI spec */
743     .read = NULL,
744     .write = spapr_msi_write,
745     .endianness = DEVICE_LITTLE_ENDIAN
746 };
747 
748 /*
749  * PHB PCI device
750  */
751 static AddressSpace *spapr_pci_dma_iommu(PCIBus *bus, void *opaque, int devfn)
752 {
753     sPAPRPHBState *phb = opaque;
754 
755     return &phb->iommu_as;
756 }
757 
758 static char *spapr_phb_vfio_get_loc_code(sPAPRPHBState *sphb,  PCIDevice *pdev)
759 {
760     char *path = NULL, *buf = NULL, *host = NULL;
761 
762     /* Get the PCI VFIO host id */
763     host = object_property_get_str(OBJECT(pdev), "host", NULL);
764     if (!host) {
765         goto err_out;
766     }
767 
768     /* Construct the path of the file that will give us the DT location */
769     path = g_strdup_printf("/sys/bus/pci/devices/%s/devspec", host);
770     g_free(host);
771     if (!path || !g_file_get_contents(path, &buf, NULL, NULL)) {
772         goto err_out;
773     }
774     g_free(path);
775 
776     /* Construct and read from host device tree the loc-code */
777     path = g_strdup_printf("/proc/device-tree%s/ibm,loc-code", buf);
778     g_free(buf);
779     if (!path || !g_file_get_contents(path, &buf, NULL, NULL)) {
780         goto err_out;
781     }
782     return buf;
783 
784 err_out:
785     g_free(path);
786     return NULL;
787 }
788 
789 static char *spapr_phb_get_loc_code(sPAPRPHBState *sphb, PCIDevice *pdev)
790 {
791     char *buf;
792     const char *devtype = "qemu";
793     uint32_t busnr = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(pdev))));
794 
795     if (object_dynamic_cast(OBJECT(pdev), "vfio-pci")) {
796         buf = spapr_phb_vfio_get_loc_code(sphb, pdev);
797         if (buf) {
798             return buf;
799         }
800         devtype = "vfio";
801     }
802     /*
803      * For emulated devices and VFIO-failure case, make up
804      * the loc-code.
805      */
806     buf = g_strdup_printf("%s_%s:%04x:%02x:%02x.%x",
807                           devtype, pdev->name, sphb->index, busnr,
808                           PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
809     return buf;
810 }
811 
812 /* Macros to operate with address in OF binding to PCI */
813 #define b_x(x, p, l)    (((x) & ((1<<(l))-1)) << (p))
814 #define b_n(x)          b_x((x), 31, 1) /* 0 if relocatable */
815 #define b_p(x)          b_x((x), 30, 1) /* 1 if prefetchable */
816 #define b_t(x)          b_x((x), 29, 1) /* 1 if the address is aliased */
817 #define b_ss(x)         b_x((x), 24, 2) /* the space code */
818 #define b_bbbbbbbb(x)   b_x((x), 16, 8) /* bus number */
819 #define b_ddddd(x)      b_x((x), 11, 5) /* device number */
820 #define b_fff(x)        b_x((x), 8, 3)  /* function number */
821 #define b_rrrrrrrr(x)   b_x((x), 0, 8)  /* register number */
822 
823 /* for 'reg'/'assigned-addresses' OF properties */
824 #define RESOURCE_CELLS_SIZE 2
825 #define RESOURCE_CELLS_ADDRESS 3
826 
827 typedef struct ResourceFields {
828     uint32_t phys_hi;
829     uint32_t phys_mid;
830     uint32_t phys_lo;
831     uint32_t size_hi;
832     uint32_t size_lo;
833 } QEMU_PACKED ResourceFields;
834 
835 typedef struct ResourceProps {
836     ResourceFields reg[8];
837     ResourceFields assigned[7];
838     uint32_t reg_len;
839     uint32_t assigned_len;
840 } ResourceProps;
841 
842 /* fill in the 'reg'/'assigned-resources' OF properties for
843  * a PCI device. 'reg' describes resource requirements for a
844  * device's IO/MEM regions, 'assigned-addresses' describes the
845  * actual resource assignments.
846  *
847  * the properties are arrays of ('phys-addr', 'size') pairs describing
848  * the addressable regions of the PCI device, where 'phys-addr' is a
849  * RESOURCE_CELLS_ADDRESS-tuple of 32-bit integers corresponding to
850  * (phys.hi, phys.mid, phys.lo), and 'size' is a
851  * RESOURCE_CELLS_SIZE-tuple corresponding to (size.hi, size.lo).
852  *
853  * phys.hi = 0xYYXXXXZZ, where:
854  *   0xYY = npt000ss
855  *          |||   |
856  *          |||   +-- space code
857  *          |||               |
858  *          |||               +  00 if configuration space
859  *          |||               +  01 if IO region,
860  *          |||               +  10 if 32-bit MEM region
861  *          |||               +  11 if 64-bit MEM region
862  *          |||
863  *          ||+------ for non-relocatable IO: 1 if aliased
864  *          ||        for relocatable IO: 1 if below 64KB
865  *          ||        for MEM: 1 if below 1MB
866  *          |+------- 1 if region is prefetchable
867  *          +-------- 1 if region is non-relocatable
868  *   0xXXXX = bbbbbbbb dddddfff, encoding bus, slot, and function
869  *            bits respectively
870  *   0xZZ = rrrrrrrr, the register number of the BAR corresponding
871  *          to the region
872  *
873  * phys.mid and phys.lo correspond respectively to the hi/lo portions
874  * of the actual address of the region.
875  *
876  * how the phys-addr/size values are used differ slightly between
877  * 'reg' and 'assigned-addresses' properties. namely, 'reg' has
878  * an additional description for the config space region of the
879  * device, and in the case of QEMU has n=0 and phys.mid=phys.lo=0
880  * to describe the region as relocatable, with an address-mapping
881  * that corresponds directly to the PHB's address space for the
882  * resource. 'assigned-addresses' always has n=1 set with an absolute
883  * address assigned for the resource. in general, 'assigned-addresses'
884  * won't be populated, since addresses for PCI devices are generally
885  * unmapped initially and left to the guest to assign.
886  *
887  * note also that addresses defined in these properties are, at least
888  * for PAPR guests, relative to the PHBs IO/MEM windows, and
889  * correspond directly to the addresses in the BARs.
890  *
891  * in accordance with PCI Bus Binding to Open Firmware,
892  * IEEE Std 1275-1994, section 4.1.1, as implemented by PAPR+ v2.7,
893  * Appendix C.
894  */
895 static void populate_resource_props(PCIDevice *d, ResourceProps *rp)
896 {
897     int bus_num = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(d))));
898     uint32_t dev_id = (b_bbbbbbbb(bus_num) |
899                        b_ddddd(PCI_SLOT(d->devfn)) |
900                        b_fff(PCI_FUNC(d->devfn)));
901     ResourceFields *reg, *assigned;
902     int i, reg_idx = 0, assigned_idx = 0;
903 
904     /* config space region */
905     reg = &rp->reg[reg_idx++];
906     reg->phys_hi = cpu_to_be32(dev_id);
907     reg->phys_mid = 0;
908     reg->phys_lo = 0;
909     reg->size_hi = 0;
910     reg->size_lo = 0;
911 
912     for (i = 0; i < PCI_NUM_REGIONS; i++) {
913         if (!d->io_regions[i].size) {
914             continue;
915         }
916 
917         reg = &rp->reg[reg_idx++];
918 
919         reg->phys_hi = cpu_to_be32(dev_id | b_rrrrrrrr(pci_bar(d, i)));
920         if (d->io_regions[i].type & PCI_BASE_ADDRESS_SPACE_IO) {
921             reg->phys_hi |= cpu_to_be32(b_ss(1));
922         } else if (d->io_regions[i].type & PCI_BASE_ADDRESS_MEM_TYPE_64) {
923             reg->phys_hi |= cpu_to_be32(b_ss(3));
924         } else {
925             reg->phys_hi |= cpu_to_be32(b_ss(2));
926         }
927         reg->phys_mid = 0;
928         reg->phys_lo = 0;
929         reg->size_hi = cpu_to_be32(d->io_regions[i].size >> 32);
930         reg->size_lo = cpu_to_be32(d->io_regions[i].size);
931 
932         if (d->io_regions[i].addr == PCI_BAR_UNMAPPED) {
933             continue;
934         }
935 
936         assigned = &rp->assigned[assigned_idx++];
937         assigned->phys_hi = cpu_to_be32(reg->phys_hi | b_n(1));
938         assigned->phys_mid = cpu_to_be32(d->io_regions[i].addr >> 32);
939         assigned->phys_lo = cpu_to_be32(d->io_regions[i].addr);
940         assigned->size_hi = reg->size_hi;
941         assigned->size_lo = reg->size_lo;
942     }
943 
944     rp->reg_len = reg_idx * sizeof(ResourceFields);
945     rp->assigned_len = assigned_idx * sizeof(ResourceFields);
946 }
947 
948 typedef struct PCIClass PCIClass;
949 typedef struct PCISubClass PCISubClass;
950 typedef struct PCIIFace PCIIFace;
951 
952 struct PCIIFace {
953     int iface;
954     const char *name;
955 };
956 
957 struct PCISubClass {
958     int subclass;
959     const char *name;
960     const PCIIFace *iface;
961 };
962 
963 struct PCIClass {
964     const char *name;
965     const PCISubClass *subc;
966 };
967 
968 static const PCISubClass undef_subclass[] = {
969     { PCI_CLASS_NOT_DEFINED_VGA, "display", NULL },
970     { 0xFF, NULL, NULL },
971 };
972 
973 static const PCISubClass mass_subclass[] = {
974     { PCI_CLASS_STORAGE_SCSI, "scsi", NULL },
975     { PCI_CLASS_STORAGE_IDE, "ide", NULL },
976     { PCI_CLASS_STORAGE_FLOPPY, "fdc", NULL },
977     { PCI_CLASS_STORAGE_IPI, "ipi", NULL },
978     { PCI_CLASS_STORAGE_RAID, "raid", NULL },
979     { PCI_CLASS_STORAGE_ATA, "ata", NULL },
980     { PCI_CLASS_STORAGE_SATA, "sata", NULL },
981     { PCI_CLASS_STORAGE_SAS, "sas", NULL },
982     { 0xFF, NULL, NULL },
983 };
984 
985 static const PCISubClass net_subclass[] = {
986     { PCI_CLASS_NETWORK_ETHERNET, "ethernet", NULL },
987     { PCI_CLASS_NETWORK_TOKEN_RING, "token-ring", NULL },
988     { PCI_CLASS_NETWORK_FDDI, "fddi", NULL },
989     { PCI_CLASS_NETWORK_ATM, "atm", NULL },
990     { PCI_CLASS_NETWORK_ISDN, "isdn", NULL },
991     { PCI_CLASS_NETWORK_WORLDFIP, "worldfip", NULL },
992     { PCI_CLASS_NETWORK_PICMG214, "picmg", NULL },
993     { 0xFF, NULL, NULL },
994 };
995 
996 static const PCISubClass displ_subclass[] = {
997     { PCI_CLASS_DISPLAY_VGA, "vga", NULL },
998     { PCI_CLASS_DISPLAY_XGA, "xga", NULL },
999     { PCI_CLASS_DISPLAY_3D, "3d-controller", NULL },
1000     { 0xFF, NULL, NULL },
1001 };
1002 
1003 static const PCISubClass media_subclass[] = {
1004     { PCI_CLASS_MULTIMEDIA_VIDEO, "video", NULL },
1005     { PCI_CLASS_MULTIMEDIA_AUDIO, "sound", NULL },
1006     { PCI_CLASS_MULTIMEDIA_PHONE, "telephony", NULL },
1007     { 0xFF, NULL, NULL },
1008 };
1009 
1010 static const PCISubClass mem_subclass[] = {
1011     { PCI_CLASS_MEMORY_RAM, "memory", NULL },
1012     { PCI_CLASS_MEMORY_FLASH, "flash", NULL },
1013     { 0xFF, NULL, NULL },
1014 };
1015 
1016 static const PCISubClass bridg_subclass[] = {
1017     { PCI_CLASS_BRIDGE_HOST, "host", NULL },
1018     { PCI_CLASS_BRIDGE_ISA, "isa", NULL },
1019     { PCI_CLASS_BRIDGE_EISA, "eisa", NULL },
1020     { PCI_CLASS_BRIDGE_MC, "mca", NULL },
1021     { PCI_CLASS_BRIDGE_PCI, "pci", NULL },
1022     { PCI_CLASS_BRIDGE_PCMCIA, "pcmcia", NULL },
1023     { PCI_CLASS_BRIDGE_NUBUS, "nubus", NULL },
1024     { PCI_CLASS_BRIDGE_CARDBUS, "cardbus", NULL },
1025     { PCI_CLASS_BRIDGE_RACEWAY, "raceway", NULL },
1026     { PCI_CLASS_BRIDGE_PCI_SEMITP, "semi-transparent-pci", NULL },
1027     { PCI_CLASS_BRIDGE_IB_PCI, "infiniband", NULL },
1028     { 0xFF, NULL, NULL },
1029 };
1030 
1031 static const PCISubClass comm_subclass[] = {
1032     { PCI_CLASS_COMMUNICATION_SERIAL, "serial", NULL },
1033     { PCI_CLASS_COMMUNICATION_PARALLEL, "parallel", NULL },
1034     { PCI_CLASS_COMMUNICATION_MULTISERIAL, "multiport-serial", NULL },
1035     { PCI_CLASS_COMMUNICATION_MODEM, "modem", NULL },
1036     { PCI_CLASS_COMMUNICATION_GPIB, "gpib", NULL },
1037     { PCI_CLASS_COMMUNICATION_SC, "smart-card", NULL },
1038     { 0xFF, NULL, NULL, },
1039 };
1040 
1041 static const PCIIFace pic_iface[] = {
1042     { PCI_CLASS_SYSTEM_PIC_IOAPIC, "io-apic" },
1043     { PCI_CLASS_SYSTEM_PIC_IOXAPIC, "io-xapic" },
1044     { 0xFF, NULL },
1045 };
1046 
1047 static const PCISubClass sys_subclass[] = {
1048     { PCI_CLASS_SYSTEM_PIC, "interrupt-controller", pic_iface },
1049     { PCI_CLASS_SYSTEM_DMA, "dma-controller", NULL },
1050     { PCI_CLASS_SYSTEM_TIMER, "timer", NULL },
1051     { PCI_CLASS_SYSTEM_RTC, "rtc", NULL },
1052     { PCI_CLASS_SYSTEM_PCI_HOTPLUG, "hot-plug-controller", NULL },
1053     { PCI_CLASS_SYSTEM_SDHCI, "sd-host-controller", NULL },
1054     { 0xFF, NULL, NULL },
1055 };
1056 
1057 static const PCISubClass inp_subclass[] = {
1058     { PCI_CLASS_INPUT_KEYBOARD, "keyboard", NULL },
1059     { PCI_CLASS_INPUT_PEN, "pen", NULL },
1060     { PCI_CLASS_INPUT_MOUSE, "mouse", NULL },
1061     { PCI_CLASS_INPUT_SCANNER, "scanner", NULL },
1062     { PCI_CLASS_INPUT_GAMEPORT, "gameport", NULL },
1063     { 0xFF, NULL, NULL },
1064 };
1065 
1066 static const PCISubClass dock_subclass[] = {
1067     { PCI_CLASS_DOCKING_GENERIC, "dock", NULL },
1068     { 0xFF, NULL, NULL },
1069 };
1070 
1071 static const PCISubClass cpu_subclass[] = {
1072     { PCI_CLASS_PROCESSOR_PENTIUM, "pentium", NULL },
1073     { PCI_CLASS_PROCESSOR_POWERPC, "powerpc", NULL },
1074     { PCI_CLASS_PROCESSOR_MIPS, "mips", NULL },
1075     { PCI_CLASS_PROCESSOR_CO, "co-processor", NULL },
1076     { 0xFF, NULL, NULL },
1077 };
1078 
1079 static const PCIIFace usb_iface[] = {
1080     { PCI_CLASS_SERIAL_USB_UHCI, "usb-uhci" },
1081     { PCI_CLASS_SERIAL_USB_OHCI, "usb-ohci", },
1082     { PCI_CLASS_SERIAL_USB_EHCI, "usb-ehci" },
1083     { PCI_CLASS_SERIAL_USB_XHCI, "usb-xhci" },
1084     { PCI_CLASS_SERIAL_USB_UNKNOWN, "usb-unknown" },
1085     { PCI_CLASS_SERIAL_USB_DEVICE, "usb-device" },
1086     { 0xFF, NULL },
1087 };
1088 
1089 static const PCISubClass ser_subclass[] = {
1090     { PCI_CLASS_SERIAL_FIREWIRE, "firewire", NULL },
1091     { PCI_CLASS_SERIAL_ACCESS, "access-bus", NULL },
1092     { PCI_CLASS_SERIAL_SSA, "ssa", NULL },
1093     { PCI_CLASS_SERIAL_USB, "usb", usb_iface },
1094     { PCI_CLASS_SERIAL_FIBER, "fibre-channel", NULL },
1095     { PCI_CLASS_SERIAL_SMBUS, "smb", NULL },
1096     { PCI_CLASS_SERIAL_IB, "infiniband", NULL },
1097     { PCI_CLASS_SERIAL_IPMI, "ipmi", NULL },
1098     { PCI_CLASS_SERIAL_SERCOS, "sercos", NULL },
1099     { PCI_CLASS_SERIAL_CANBUS, "canbus", NULL },
1100     { 0xFF, NULL, NULL },
1101 };
1102 
1103 static const PCISubClass wrl_subclass[] = {
1104     { PCI_CLASS_WIRELESS_IRDA, "irda", NULL },
1105     { PCI_CLASS_WIRELESS_CIR, "consumer-ir", NULL },
1106     { PCI_CLASS_WIRELESS_RF_CONTROLLER, "rf-controller", NULL },
1107     { PCI_CLASS_WIRELESS_BLUETOOTH, "bluetooth", NULL },
1108     { PCI_CLASS_WIRELESS_BROADBAND, "broadband", NULL },
1109     { 0xFF, NULL, NULL },
1110 };
1111 
1112 static const PCISubClass sat_subclass[] = {
1113     { PCI_CLASS_SATELLITE_TV, "satellite-tv", NULL },
1114     { PCI_CLASS_SATELLITE_AUDIO, "satellite-audio", NULL },
1115     { PCI_CLASS_SATELLITE_VOICE, "satellite-voice", NULL },
1116     { PCI_CLASS_SATELLITE_DATA, "satellite-data", NULL },
1117     { 0xFF, NULL, NULL },
1118 };
1119 
1120 static const PCISubClass crypt_subclass[] = {
1121     { PCI_CLASS_CRYPT_NETWORK, "network-encryption", NULL },
1122     { PCI_CLASS_CRYPT_ENTERTAINMENT,
1123       "entertainment-encryption", NULL },
1124     { 0xFF, NULL, NULL },
1125 };
1126 
1127 static const PCISubClass spc_subclass[] = {
1128     { PCI_CLASS_SP_DPIO, "dpio", NULL },
1129     { PCI_CLASS_SP_PERF, "counter", NULL },
1130     { PCI_CLASS_SP_SYNCH, "measurement", NULL },
1131     { PCI_CLASS_SP_MANAGEMENT, "management-card", NULL },
1132     { 0xFF, NULL, NULL },
1133 };
1134 
1135 static const PCIClass pci_classes[] = {
1136     { "legacy-device", undef_subclass },
1137     { "mass-storage",  mass_subclass },
1138     { "network", net_subclass },
1139     { "display", displ_subclass, },
1140     { "multimedia-device", media_subclass },
1141     { "memory-controller", mem_subclass },
1142     { "unknown-bridge", bridg_subclass },
1143     { "communication-controller", comm_subclass},
1144     { "system-peripheral", sys_subclass },
1145     { "input-controller", inp_subclass },
1146     { "docking-station", dock_subclass },
1147     { "cpu", cpu_subclass },
1148     { "serial-bus", ser_subclass },
1149     { "wireless-controller", wrl_subclass },
1150     { "intelligent-io", NULL },
1151     { "satellite-device", sat_subclass },
1152     { "encryption", crypt_subclass },
1153     { "data-processing-controller", spc_subclass },
1154 };
1155 
1156 static const char *pci_find_device_name(uint8_t class, uint8_t subclass,
1157                                         uint8_t iface)
1158 {
1159     const PCIClass *pclass;
1160     const PCISubClass *psubclass;
1161     const PCIIFace *piface;
1162     const char *name;
1163 
1164     if (class >= ARRAY_SIZE(pci_classes)) {
1165         return "pci";
1166     }
1167 
1168     pclass = pci_classes + class;
1169     name = pclass->name;
1170 
1171     if (pclass->subc == NULL) {
1172         return name;
1173     }
1174 
1175     psubclass = pclass->subc;
1176     while ((psubclass->subclass & 0xff) != 0xff) {
1177         if ((psubclass->subclass & 0xff) == subclass) {
1178             name = psubclass->name;
1179             break;
1180         }
1181         psubclass++;
1182     }
1183 
1184     piface = psubclass->iface;
1185     if (piface == NULL) {
1186         return name;
1187     }
1188     while ((piface->iface & 0xff) != 0xff) {
1189         if ((piface->iface & 0xff) == iface) {
1190             name = piface->name;
1191             break;
1192         }
1193         piface++;
1194     }
1195 
1196     return name;
1197 }
1198 
1199 static void pci_get_node_name(char *nodename, int len, PCIDevice *dev)
1200 {
1201     int slot = PCI_SLOT(dev->devfn);
1202     int func = PCI_FUNC(dev->devfn);
1203     uint32_t ccode = pci_default_read_config(dev, PCI_CLASS_PROG, 3);
1204     const char *name;
1205 
1206     name = pci_find_device_name((ccode >> 16) & 0xff, (ccode >> 8) & 0xff,
1207                                 ccode & 0xff);
1208 
1209     if (func != 0) {
1210         snprintf(nodename, len, "%s@%x,%x", name, slot, func);
1211     } else {
1212         snprintf(nodename, len, "%s@%x", name, slot);
1213     }
1214 }
1215 
1216 static uint32_t spapr_phb_get_pci_drc_index(sPAPRPHBState *phb,
1217                                             PCIDevice *pdev);
1218 
1219 static int spapr_populate_pci_child_dt(PCIDevice *dev, void *fdt, int offset,
1220                                        sPAPRPHBState *sphb)
1221 {
1222     ResourceProps rp;
1223     bool is_bridge = false;
1224     int pci_status, err;
1225     char *buf = NULL;
1226     uint32_t drc_index = spapr_phb_get_pci_drc_index(sphb, dev);
1227     uint32_t ccode = pci_default_read_config(dev, PCI_CLASS_PROG, 3);
1228     uint32_t max_msi, max_msix;
1229 
1230     if (pci_default_read_config(dev, PCI_HEADER_TYPE, 1) ==
1231         PCI_HEADER_TYPE_BRIDGE) {
1232         is_bridge = true;
1233     }
1234 
1235     /* in accordance with PAPR+ v2.7 13.6.3, Table 181 */
1236     _FDT(fdt_setprop_cell(fdt, offset, "vendor-id",
1237                           pci_default_read_config(dev, PCI_VENDOR_ID, 2)));
1238     _FDT(fdt_setprop_cell(fdt, offset, "device-id",
1239                           pci_default_read_config(dev, PCI_DEVICE_ID, 2)));
1240     _FDT(fdt_setprop_cell(fdt, offset, "revision-id",
1241                           pci_default_read_config(dev, PCI_REVISION_ID, 1)));
1242     _FDT(fdt_setprop_cell(fdt, offset, "class-code", ccode));
1243     if (pci_default_read_config(dev, PCI_INTERRUPT_PIN, 1)) {
1244         _FDT(fdt_setprop_cell(fdt, offset, "interrupts",
1245                  pci_default_read_config(dev, PCI_INTERRUPT_PIN, 1)));
1246     }
1247 
1248     if (!is_bridge) {
1249         _FDT(fdt_setprop_cell(fdt, offset, "min-grant",
1250             pci_default_read_config(dev, PCI_MIN_GNT, 1)));
1251         _FDT(fdt_setprop_cell(fdt, offset, "max-latency",
1252             pci_default_read_config(dev, PCI_MAX_LAT, 1)));
1253     }
1254 
1255     if (pci_default_read_config(dev, PCI_SUBSYSTEM_ID, 2)) {
1256         _FDT(fdt_setprop_cell(fdt, offset, "subsystem-id",
1257                  pci_default_read_config(dev, PCI_SUBSYSTEM_ID, 2)));
1258     }
1259 
1260     if (pci_default_read_config(dev, PCI_SUBSYSTEM_VENDOR_ID, 2)) {
1261         _FDT(fdt_setprop_cell(fdt, offset, "subsystem-vendor-id",
1262                  pci_default_read_config(dev, PCI_SUBSYSTEM_VENDOR_ID, 2)));
1263     }
1264 
1265     _FDT(fdt_setprop_cell(fdt, offset, "cache-line-size",
1266         pci_default_read_config(dev, PCI_CACHE_LINE_SIZE, 1)));
1267 
1268     /* the following fdt cells are masked off the pci status register */
1269     pci_status = pci_default_read_config(dev, PCI_STATUS, 2);
1270     _FDT(fdt_setprop_cell(fdt, offset, "devsel-speed",
1271                           PCI_STATUS_DEVSEL_MASK & pci_status));
1272 
1273     if (pci_status & PCI_STATUS_FAST_BACK) {
1274         _FDT(fdt_setprop(fdt, offset, "fast-back-to-back", NULL, 0));
1275     }
1276     if (pci_status & PCI_STATUS_66MHZ) {
1277         _FDT(fdt_setprop(fdt, offset, "66mhz-capable", NULL, 0));
1278     }
1279     if (pci_status & PCI_STATUS_UDF) {
1280         _FDT(fdt_setprop(fdt, offset, "udf-supported", NULL, 0));
1281     }
1282 
1283     _FDT(fdt_setprop_string(fdt, offset, "name",
1284                             pci_find_device_name((ccode >> 16) & 0xff,
1285                                                  (ccode >> 8) & 0xff,
1286                                                  ccode & 0xff)));
1287     buf = spapr_phb_get_loc_code(sphb, dev);
1288     if (!buf) {
1289         error_report("Failed setting the ibm,loc-code");
1290         return -1;
1291     }
1292 
1293     err = fdt_setprop_string(fdt, offset, "ibm,loc-code", buf);
1294     g_free(buf);
1295     if (err < 0) {
1296         return err;
1297     }
1298 
1299     if (drc_index) {
1300         _FDT(fdt_setprop_cell(fdt, offset, "ibm,my-drc-index", drc_index));
1301     }
1302 
1303     _FDT(fdt_setprop_cell(fdt, offset, "#address-cells",
1304                           RESOURCE_CELLS_ADDRESS));
1305     _FDT(fdt_setprop_cell(fdt, offset, "#size-cells",
1306                           RESOURCE_CELLS_SIZE));
1307 
1308     max_msi = msi_nr_vectors_allocated(dev);
1309     if (max_msi) {
1310         _FDT(fdt_setprop_cell(fdt, offset, "ibm,req#msi", max_msi));
1311     }
1312     max_msix = dev->msix_entries_nr;
1313     if (max_msix) {
1314         _FDT(fdt_setprop_cell(fdt, offset, "ibm,req#msi-x", max_msix));
1315     }
1316 
1317     populate_resource_props(dev, &rp);
1318     _FDT(fdt_setprop(fdt, offset, "reg", (uint8_t *)rp.reg, rp.reg_len));
1319     _FDT(fdt_setprop(fdt, offset, "assigned-addresses",
1320                      (uint8_t *)rp.assigned, rp.assigned_len));
1321 
1322     if (sphb->pcie_ecs && pci_is_express(dev)) {
1323         _FDT(fdt_setprop_cell(fdt, offset, "ibm,pci-config-space-type", 0x1));
1324     }
1325 
1326     return 0;
1327 }
1328 
1329 /* create OF node for pci device and required OF DT properties */
1330 static int spapr_create_pci_child_dt(sPAPRPHBState *phb, PCIDevice *dev,
1331                                      void *fdt, int node_offset)
1332 {
1333     int offset, ret;
1334     char nodename[FDT_NAME_MAX];
1335 
1336     pci_get_node_name(nodename, FDT_NAME_MAX, dev);
1337     offset = fdt_add_subnode(fdt, node_offset, nodename);
1338     ret = spapr_populate_pci_child_dt(dev, fdt, offset, phb);
1339 
1340     g_assert(!ret);
1341     if (ret) {
1342         return 0;
1343     }
1344     return offset;
1345 }
1346 
1347 static void spapr_phb_add_pci_device(sPAPRDRConnector *drc,
1348                                      sPAPRPHBState *phb,
1349                                      PCIDevice *pdev,
1350                                      Error **errp)
1351 {
1352     sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
1353     DeviceState *dev = DEVICE(pdev);
1354     void *fdt = NULL;
1355     int fdt_start_offset = 0, fdt_size;
1356 
1357     fdt = create_device_tree(&fdt_size);
1358     fdt_start_offset = spapr_create_pci_child_dt(phb, pdev, fdt, 0);
1359     if (!fdt_start_offset) {
1360         error_setg(errp, "Failed to create pci child device tree node");
1361         goto out;
1362     }
1363 
1364     drck->attach(drc, DEVICE(pdev),
1365                  fdt, fdt_start_offset, !dev->hotplugged, errp);
1366 out:
1367     if (*errp) {
1368         g_free(fdt);
1369     }
1370 }
1371 
1372 static void spapr_phb_remove_pci_device_cb(DeviceState *dev, void *opaque)
1373 {
1374     /* some version guests do not wait for completion of a device
1375      * cleanup (generally done asynchronously by the kernel) before
1376      * signaling to QEMU that the device is safe, but instead sleep
1377      * for some 'safe' period of time. unfortunately on a busy host
1378      * this sleep isn't guaranteed to be long enough, resulting in
1379      * bad things like IRQ lines being left asserted during final
1380      * device removal. to deal with this we call reset just prior
1381      * to finalizing the device, which will put the device back into
1382      * an 'idle' state, as the device cleanup code expects.
1383      */
1384     pci_device_reset(PCI_DEVICE(dev));
1385     object_unparent(OBJECT(dev));
1386 }
1387 
1388 static void spapr_phb_remove_pci_device(sPAPRDRConnector *drc,
1389                                         sPAPRPHBState *phb,
1390                                         PCIDevice *pdev,
1391                                         Error **errp)
1392 {
1393     sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
1394 
1395     drck->detach(drc, DEVICE(pdev), spapr_phb_remove_pci_device_cb, phb, errp);
1396 }
1397 
1398 static sPAPRDRConnector *spapr_phb_get_pci_func_drc(sPAPRPHBState *phb,
1399                                                     uint32_t busnr,
1400                                                     int32_t devfn)
1401 {
1402     return spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_PCI,
1403                                     (phb->index << 16) |
1404                                     (busnr << 8) |
1405                                     devfn);
1406 }
1407 
1408 static sPAPRDRConnector *spapr_phb_get_pci_drc(sPAPRPHBState *phb,
1409                                                PCIDevice *pdev)
1410 {
1411     uint32_t busnr = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(pdev))));
1412     return spapr_phb_get_pci_func_drc(phb, busnr, pdev->devfn);
1413 }
1414 
1415 static uint32_t spapr_phb_get_pci_drc_index(sPAPRPHBState *phb,
1416                                             PCIDevice *pdev)
1417 {
1418     sPAPRDRConnector *drc = spapr_phb_get_pci_drc(phb, pdev);
1419     sPAPRDRConnectorClass *drck;
1420 
1421     if (!drc) {
1422         return 0;
1423     }
1424 
1425     drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
1426     return drck->get_index(drc);
1427 }
1428 
1429 static void spapr_phb_hot_plug_child(HotplugHandler *plug_handler,
1430                                      DeviceState *plugged_dev, Error **errp)
1431 {
1432     sPAPRPHBState *phb = SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler));
1433     PCIDevice *pdev = PCI_DEVICE(plugged_dev);
1434     sPAPRDRConnector *drc = spapr_phb_get_pci_drc(phb, pdev);
1435     Error *local_err = NULL;
1436     PCIBus *bus = PCI_BUS(qdev_get_parent_bus(DEVICE(pdev)));
1437     uint32_t slotnr = PCI_SLOT(pdev->devfn);
1438 
1439     /* if DR is disabled we don't need to do anything in the case of
1440      * hotplug or coldplug callbacks
1441      */
1442     if (!phb->dr_enabled) {
1443         /* if this is a hotplug operation initiated by the user
1444          * we need to let them know it's not enabled
1445          */
1446         if (plugged_dev->hotplugged) {
1447             error_setg(errp, QERR_BUS_NO_HOTPLUG,
1448                        object_get_typename(OBJECT(phb)));
1449         }
1450         return;
1451     }
1452 
1453     g_assert(drc);
1454 
1455     /* Following the QEMU convention used for PCIe multifunction
1456      * hotplug, we do not allow functions to be hotplugged to a
1457      * slot that already has function 0 present
1458      */
1459     if (plugged_dev->hotplugged && bus->devices[PCI_DEVFN(slotnr, 0)] &&
1460         PCI_FUNC(pdev->devfn) != 0) {
1461         error_setg(errp, "PCI: slot %d function 0 already ocuppied by %s,"
1462                    " additional functions can no longer be exposed to guest.",
1463                    slotnr, bus->devices[PCI_DEVFN(slotnr, 0)]->name);
1464         return;
1465     }
1466 
1467     spapr_phb_add_pci_device(drc, phb, pdev, &local_err);
1468     if (local_err) {
1469         error_propagate(errp, local_err);
1470         return;
1471     }
1472 
1473     /* If this is function 0, signal hotplug for all the device functions.
1474      * Otherwise defer sending the hotplug event.
1475      */
1476     if (plugged_dev->hotplugged && PCI_FUNC(pdev->devfn) == 0) {
1477         int i;
1478 
1479         for (i = 0; i < 8; i++) {
1480             sPAPRDRConnector *func_drc;
1481             sPAPRDRConnectorClass *func_drck;
1482             sPAPRDREntitySense state;
1483 
1484             func_drc = spapr_phb_get_pci_func_drc(phb, pci_bus_num(bus),
1485                                                   PCI_DEVFN(slotnr, i));
1486             func_drck = SPAPR_DR_CONNECTOR_GET_CLASS(func_drc);
1487             func_drck->entity_sense(func_drc, &state);
1488 
1489             if (state == SPAPR_DR_ENTITY_SENSE_PRESENT) {
1490                 spapr_hotplug_req_add_by_index(func_drc);
1491             }
1492         }
1493     }
1494 }
1495 
1496 static void spapr_phb_hot_unplug_child(HotplugHandler *plug_handler,
1497                                        DeviceState *plugged_dev, Error **errp)
1498 {
1499     sPAPRPHBState *phb = SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler));
1500     PCIDevice *pdev = PCI_DEVICE(plugged_dev);
1501     sPAPRDRConnectorClass *drck;
1502     sPAPRDRConnector *drc = spapr_phb_get_pci_drc(phb, pdev);
1503     Error *local_err = NULL;
1504 
1505     if (!phb->dr_enabled) {
1506         error_setg(errp, QERR_BUS_NO_HOTPLUG,
1507                    object_get_typename(OBJECT(phb)));
1508         return;
1509     }
1510 
1511     g_assert(drc);
1512 
1513     drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
1514     if (!drck->release_pending(drc)) {
1515         PCIBus *bus = PCI_BUS(qdev_get_parent_bus(DEVICE(pdev)));
1516         uint32_t slotnr = PCI_SLOT(pdev->devfn);
1517         sPAPRDRConnector *func_drc;
1518         sPAPRDRConnectorClass *func_drck;
1519         sPAPRDREntitySense state;
1520         int i;
1521 
1522         /* ensure any other present functions are pending unplug */
1523         if (PCI_FUNC(pdev->devfn) == 0) {
1524             for (i = 1; i < 8; i++) {
1525                 func_drc = spapr_phb_get_pci_func_drc(phb, pci_bus_num(bus),
1526                                                       PCI_DEVFN(slotnr, i));
1527                 func_drck = SPAPR_DR_CONNECTOR_GET_CLASS(func_drc);
1528                 func_drck->entity_sense(func_drc, &state);
1529                 if (state == SPAPR_DR_ENTITY_SENSE_PRESENT
1530                     && !func_drck->release_pending(func_drc)) {
1531                     error_setg(errp,
1532                                "PCI: slot %d, function %d still present. "
1533                                "Must unplug all non-0 functions first.",
1534                                slotnr, i);
1535                     return;
1536                 }
1537             }
1538         }
1539 
1540         spapr_phb_remove_pci_device(drc, phb, pdev, &local_err);
1541         if (local_err) {
1542             error_propagate(errp, local_err);
1543             return;
1544         }
1545 
1546         /* if this isn't func 0, defer unplug event. otherwise signal removal
1547          * for all present functions
1548          */
1549         if (PCI_FUNC(pdev->devfn) == 0) {
1550             for (i = 7; i >= 0; i--) {
1551                 func_drc = spapr_phb_get_pci_func_drc(phb, pci_bus_num(bus),
1552                                                       PCI_DEVFN(slotnr, i));
1553                 func_drck = SPAPR_DR_CONNECTOR_GET_CLASS(func_drc);
1554                 func_drck->entity_sense(func_drc, &state);
1555                 if (state == SPAPR_DR_ENTITY_SENSE_PRESENT) {
1556                     spapr_hotplug_req_remove_by_index(func_drc);
1557                 }
1558             }
1559         }
1560     }
1561 }
1562 
1563 static void spapr_phb_realize(DeviceState *dev, Error **errp)
1564 {
1565     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1566     SysBusDevice *s = SYS_BUS_DEVICE(dev);
1567     sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(s);
1568     PCIHostState *phb = PCI_HOST_BRIDGE(s);
1569     char *namebuf;
1570     int i;
1571     PCIBus *bus;
1572     uint64_t msi_window_size = 4096;
1573     sPAPRTCETable *tcet;
1574     const unsigned windows_supported =
1575         sphb->ddw_enabled ? SPAPR_PCI_DMA_MAX_WINDOWS : 1;
1576 
1577     if (sphb->index != (uint32_t)-1) {
1578         sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
1579         Error *local_err = NULL;
1580 
1581         if ((sphb->buid != (uint64_t)-1) || (sphb->dma_liobn[0] != (uint32_t)-1)
1582             || (sphb->dma_liobn[1] != (uint32_t)-1 && windows_supported == 2)
1583             || (sphb->mem_win_addr != (hwaddr)-1)
1584             || (sphb->mem64_win_addr != (hwaddr)-1)
1585             || (sphb->io_win_addr != (hwaddr)-1)) {
1586             error_setg(errp, "Either \"index\" or other parameters must"
1587                        " be specified for PAPR PHB, not both");
1588             return;
1589         }
1590 
1591         smc->phb_placement(spapr, sphb->index,
1592                            &sphb->buid, &sphb->io_win_addr,
1593                            &sphb->mem_win_addr, &sphb->mem64_win_addr,
1594                            windows_supported, sphb->dma_liobn, &local_err);
1595         if (local_err) {
1596             error_propagate(errp, local_err);
1597             return;
1598         }
1599     }
1600 
1601     if (sphb->buid == (uint64_t)-1) {
1602         error_setg(errp, "BUID not specified for PHB");
1603         return;
1604     }
1605 
1606     if ((sphb->dma_liobn[0] == (uint32_t)-1) ||
1607         ((sphb->dma_liobn[1] == (uint32_t)-1) && (windows_supported > 1))) {
1608         error_setg(errp, "LIOBN(s) not specified for PHB");
1609         return;
1610     }
1611 
1612     if (sphb->mem_win_addr == (hwaddr)-1) {
1613         error_setg(errp, "Memory window address not specified for PHB");
1614         return;
1615     }
1616 
1617     if (sphb->io_win_addr == (hwaddr)-1) {
1618         error_setg(errp, "IO window address not specified for PHB");
1619         return;
1620     }
1621 
1622     if (sphb->mem64_win_size != 0) {
1623         if (sphb->mem64_win_addr == (hwaddr)-1) {
1624             error_setg(errp,
1625                        "64-bit memory window address not specified for PHB");
1626             return;
1627         }
1628 
1629         if (sphb->mem_win_size > SPAPR_PCI_MEM32_WIN_SIZE) {
1630             error_setg(errp, "32-bit memory window of size 0x%"HWADDR_PRIx
1631                        " (max 2 GiB)", sphb->mem_win_size);
1632             return;
1633         }
1634 
1635         if (sphb->mem64_win_pciaddr == (hwaddr)-1) {
1636             /* 64-bit window defaults to identity mapping */
1637             sphb->mem64_win_pciaddr = sphb->mem64_win_addr;
1638         }
1639     } else if (sphb->mem_win_size > SPAPR_PCI_MEM32_WIN_SIZE) {
1640         /*
1641          * For compatibility with old configuration, if no 64-bit MMIO
1642          * window is specified, but the ordinary (32-bit) memory
1643          * window is specified as > 2GiB, we treat it as a 2GiB 32-bit
1644          * window, with a 64-bit MMIO window following on immediately
1645          * afterwards
1646          */
1647         sphb->mem64_win_size = sphb->mem_win_size - SPAPR_PCI_MEM32_WIN_SIZE;
1648         sphb->mem64_win_addr = sphb->mem_win_addr + SPAPR_PCI_MEM32_WIN_SIZE;
1649         sphb->mem64_win_pciaddr =
1650             SPAPR_PCI_MEM_WIN_BUS_OFFSET + SPAPR_PCI_MEM32_WIN_SIZE;
1651         sphb->mem_win_size = SPAPR_PCI_MEM32_WIN_SIZE;
1652     }
1653 
1654     if (spapr_pci_find_phb(spapr, sphb->buid)) {
1655         error_setg(errp, "PCI host bridges must have unique BUIDs");
1656         return;
1657     }
1658 
1659     if (sphb->numa_node != -1 &&
1660         (sphb->numa_node >= MAX_NODES || !numa_info[sphb->numa_node].present)) {
1661         error_setg(errp, "Invalid NUMA node ID for PCI host bridge");
1662         return;
1663     }
1664 
1665     sphb->dtbusname = g_strdup_printf("pci@%" PRIx64, sphb->buid);
1666 
1667     namebuf = alloca(strlen(sphb->dtbusname) + 32);
1668 
1669     /* Initialize memory regions */
1670     sprintf(namebuf, "%s.mmio", sphb->dtbusname);
1671     memory_region_init(&sphb->memspace, OBJECT(sphb), namebuf, UINT64_MAX);
1672 
1673     sprintf(namebuf, "%s.mmio32-alias", sphb->dtbusname);
1674     memory_region_init_alias(&sphb->mem32window, OBJECT(sphb),
1675                              namebuf, &sphb->memspace,
1676                              SPAPR_PCI_MEM_WIN_BUS_OFFSET, sphb->mem_win_size);
1677     memory_region_add_subregion(get_system_memory(), sphb->mem_win_addr,
1678                                 &sphb->mem32window);
1679 
1680     sprintf(namebuf, "%s.mmio64-alias", sphb->dtbusname);
1681     memory_region_init_alias(&sphb->mem64window, OBJECT(sphb),
1682                              namebuf, &sphb->memspace,
1683                              sphb->mem64_win_pciaddr, sphb->mem64_win_size);
1684     memory_region_add_subregion(get_system_memory(), sphb->mem64_win_addr,
1685                                 &sphb->mem64window);
1686 
1687     /* Initialize IO regions */
1688     sprintf(namebuf, "%s.io", sphb->dtbusname);
1689     memory_region_init(&sphb->iospace, OBJECT(sphb),
1690                        namebuf, SPAPR_PCI_IO_WIN_SIZE);
1691 
1692     sprintf(namebuf, "%s.io-alias", sphb->dtbusname);
1693     memory_region_init_alias(&sphb->iowindow, OBJECT(sphb), namebuf,
1694                              &sphb->iospace, 0, SPAPR_PCI_IO_WIN_SIZE);
1695     memory_region_add_subregion(get_system_memory(), sphb->io_win_addr,
1696                                 &sphb->iowindow);
1697 
1698     bus = pci_register_bus(dev, NULL,
1699                            pci_spapr_set_irq, pci_spapr_map_irq, sphb,
1700                            &sphb->memspace, &sphb->iospace,
1701                            PCI_DEVFN(0, 0), PCI_NUM_PINS, TYPE_PCI_BUS);
1702     phb->bus = bus;
1703     qbus_set_hotplug_handler(BUS(phb->bus), DEVICE(sphb), NULL);
1704 
1705     /*
1706      * Initialize PHB address space.
1707      * By default there will be at least one subregion for default
1708      * 32bit DMA window.
1709      * Later the guest might want to create another DMA window
1710      * which will become another memory subregion.
1711      */
1712     sprintf(namebuf, "%s.iommu-root", sphb->dtbusname);
1713 
1714     memory_region_init(&sphb->iommu_root, OBJECT(sphb),
1715                        namebuf, UINT64_MAX);
1716     address_space_init(&sphb->iommu_as, &sphb->iommu_root,
1717                        sphb->dtbusname);
1718 
1719     /*
1720      * As MSI/MSIX interrupts trigger by writing at MSI/MSIX vectors,
1721      * we need to allocate some memory to catch those writes coming
1722      * from msi_notify()/msix_notify().
1723      * As MSIMessage:addr is going to be the same and MSIMessage:data
1724      * is going to be a VIRQ number, 4 bytes of the MSI MR will only
1725      * be used.
1726      *
1727      * For KVM we want to ensure that this memory is a full page so that
1728      * our memory slot is of page size granularity.
1729      */
1730 #ifdef CONFIG_KVM
1731     if (kvm_enabled()) {
1732         msi_window_size = getpagesize();
1733     }
1734 #endif
1735 
1736     memory_region_init_io(&sphb->msiwindow, NULL, &spapr_msi_ops, spapr,
1737                           "msi", msi_window_size);
1738     memory_region_add_subregion(&sphb->iommu_root, SPAPR_PCI_MSI_WINDOW,
1739                                 &sphb->msiwindow);
1740 
1741     pci_setup_iommu(bus, spapr_pci_dma_iommu, sphb);
1742 
1743     pci_bus_set_route_irq_fn(bus, spapr_route_intx_pin_to_irq);
1744 
1745     QLIST_INSERT_HEAD(&spapr->phbs, sphb, list);
1746 
1747     /* Initialize the LSI table */
1748     for (i = 0; i < PCI_NUM_PINS; i++) {
1749         uint32_t irq;
1750         Error *local_err = NULL;
1751 
1752         irq = spapr_ics_alloc_block(spapr->ics, 1, true, false, &local_err);
1753         if (local_err) {
1754             error_propagate(errp, local_err);
1755             error_prepend(errp, "can't allocate LSIs: ");
1756             return;
1757         }
1758 
1759         sphb->lsi_table[i].irq = irq;
1760     }
1761 
1762     /* allocate connectors for child PCI devices */
1763     if (sphb->dr_enabled) {
1764         for (i = 0; i < PCI_SLOT_MAX * 8; i++) {
1765             spapr_dr_connector_new(OBJECT(phb),
1766                                    SPAPR_DR_CONNECTOR_TYPE_PCI,
1767                                    (sphb->index << 16) | i);
1768         }
1769     }
1770 
1771     /* DMA setup */
1772     if ((sphb->page_size_mask & qemu_getrampagesize()) == 0) {
1773         error_report("System page size 0x%lx is not enabled in page_size_mask "
1774                      "(0x%"PRIx64"). Performance may be slow",
1775                      qemu_getrampagesize(), sphb->page_size_mask);
1776     }
1777 
1778     for (i = 0; i < windows_supported; ++i) {
1779         tcet = spapr_tce_new_table(DEVICE(sphb), sphb->dma_liobn[i]);
1780         if (!tcet) {
1781             error_setg(errp, "Creating window#%d failed for %s",
1782                        i, sphb->dtbusname);
1783             return;
1784         }
1785         memory_region_add_subregion_overlap(&sphb->iommu_root, 0,
1786                                             spapr_tce_get_iommu(tcet), 0);
1787     }
1788 
1789     sphb->msi = g_hash_table_new_full(g_int_hash, g_int_equal, g_free, g_free);
1790 }
1791 
1792 static int spapr_phb_children_reset(Object *child, void *opaque)
1793 {
1794     DeviceState *dev = (DeviceState *) object_dynamic_cast(child, TYPE_DEVICE);
1795 
1796     if (dev) {
1797         device_reset(dev);
1798     }
1799 
1800     return 0;
1801 }
1802 
1803 void spapr_phb_dma_reset(sPAPRPHBState *sphb)
1804 {
1805     int i;
1806     sPAPRTCETable *tcet;
1807 
1808     for (i = 0; i < SPAPR_PCI_DMA_MAX_WINDOWS; ++i) {
1809         tcet = spapr_tce_find_by_liobn(sphb->dma_liobn[i]);
1810 
1811         if (tcet && tcet->nb_table) {
1812             spapr_tce_table_disable(tcet);
1813         }
1814     }
1815 
1816     /* Register default 32bit DMA window */
1817     tcet = spapr_tce_find_by_liobn(sphb->dma_liobn[0]);
1818     spapr_tce_table_enable(tcet, SPAPR_TCE_PAGE_SHIFT, sphb->dma_win_addr,
1819                            sphb->dma_win_size >> SPAPR_TCE_PAGE_SHIFT);
1820 }
1821 
1822 static void spapr_phb_reset(DeviceState *qdev)
1823 {
1824     sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(qdev);
1825 
1826     spapr_phb_dma_reset(sphb);
1827 
1828     /* Reset the IOMMU state */
1829     object_child_foreach(OBJECT(qdev), spapr_phb_children_reset, NULL);
1830 
1831     if (spapr_phb_eeh_available(SPAPR_PCI_HOST_BRIDGE(qdev))) {
1832         spapr_phb_vfio_reset(qdev);
1833     }
1834 }
1835 
1836 static Property spapr_phb_properties[] = {
1837     DEFINE_PROP_UINT32("index", sPAPRPHBState, index, -1),
1838     DEFINE_PROP_UINT64("buid", sPAPRPHBState, buid, -1),
1839     DEFINE_PROP_UINT32("liobn", sPAPRPHBState, dma_liobn[0], -1),
1840     DEFINE_PROP_UINT32("liobn64", sPAPRPHBState, dma_liobn[1], -1),
1841     DEFINE_PROP_UINT64("mem_win_addr", sPAPRPHBState, mem_win_addr, -1),
1842     DEFINE_PROP_UINT64("mem_win_size", sPAPRPHBState, mem_win_size,
1843                        SPAPR_PCI_MEM32_WIN_SIZE),
1844     DEFINE_PROP_UINT64("mem64_win_addr", sPAPRPHBState, mem64_win_addr, -1),
1845     DEFINE_PROP_UINT64("mem64_win_size", sPAPRPHBState, mem64_win_size,
1846                        SPAPR_PCI_MEM64_WIN_SIZE),
1847     DEFINE_PROP_UINT64("mem64_win_pciaddr", sPAPRPHBState, mem64_win_pciaddr,
1848                        -1),
1849     DEFINE_PROP_UINT64("io_win_addr", sPAPRPHBState, io_win_addr, -1),
1850     DEFINE_PROP_UINT64("io_win_size", sPAPRPHBState, io_win_size,
1851                        SPAPR_PCI_IO_WIN_SIZE),
1852     DEFINE_PROP_BOOL("dynamic-reconfiguration", sPAPRPHBState, dr_enabled,
1853                      true),
1854     /* Default DMA window is 0..1GB */
1855     DEFINE_PROP_UINT64("dma_win_addr", sPAPRPHBState, dma_win_addr, 0),
1856     DEFINE_PROP_UINT64("dma_win_size", sPAPRPHBState, dma_win_size, 0x40000000),
1857     DEFINE_PROP_UINT64("dma64_win_addr", sPAPRPHBState, dma64_win_addr,
1858                        0x800000000000000ULL),
1859     DEFINE_PROP_BOOL("ddw", sPAPRPHBState, ddw_enabled, true),
1860     DEFINE_PROP_UINT64("pgsz", sPAPRPHBState, page_size_mask,
1861                        (1ULL << 12) | (1ULL << 16)),
1862     DEFINE_PROP_UINT32("numa_node", sPAPRPHBState, numa_node, -1),
1863     DEFINE_PROP_BOOL("pre-2.8-migration", sPAPRPHBState,
1864                      pre_2_8_migration, false),
1865     DEFINE_PROP_BOOL("pcie-extended-configuration-space", sPAPRPHBState,
1866                      pcie_ecs, true),
1867     DEFINE_PROP_END_OF_LIST(),
1868 };
1869 
1870 static const VMStateDescription vmstate_spapr_pci_lsi = {
1871     .name = "spapr_pci/lsi",
1872     .version_id = 1,
1873     .minimum_version_id = 1,
1874     .fields = (VMStateField[]) {
1875         VMSTATE_UINT32_EQUAL(irq, struct spapr_pci_lsi),
1876 
1877         VMSTATE_END_OF_LIST()
1878     },
1879 };
1880 
1881 static const VMStateDescription vmstate_spapr_pci_msi = {
1882     .name = "spapr_pci/msi",
1883     .version_id = 1,
1884     .minimum_version_id = 1,
1885     .fields = (VMStateField []) {
1886         VMSTATE_UINT32(key, spapr_pci_msi_mig),
1887         VMSTATE_UINT32(value.first_irq, spapr_pci_msi_mig),
1888         VMSTATE_UINT32(value.num, spapr_pci_msi_mig),
1889         VMSTATE_END_OF_LIST()
1890     },
1891 };
1892 
1893 static void spapr_pci_pre_save(void *opaque)
1894 {
1895     sPAPRPHBState *sphb = opaque;
1896     GHashTableIter iter;
1897     gpointer key, value;
1898     int i;
1899 
1900     g_free(sphb->msi_devs);
1901     sphb->msi_devs = NULL;
1902     sphb->msi_devs_num = g_hash_table_size(sphb->msi);
1903     if (!sphb->msi_devs_num) {
1904         return;
1905     }
1906     sphb->msi_devs = g_malloc(sphb->msi_devs_num * sizeof(spapr_pci_msi_mig));
1907 
1908     g_hash_table_iter_init(&iter, sphb->msi);
1909     for (i = 0; g_hash_table_iter_next(&iter, &key, &value); ++i) {
1910         sphb->msi_devs[i].key = *(uint32_t *) key;
1911         sphb->msi_devs[i].value = *(spapr_pci_msi *) value;
1912     }
1913 
1914     if (sphb->pre_2_8_migration) {
1915         sphb->mig_liobn = sphb->dma_liobn[0];
1916         sphb->mig_mem_win_addr = sphb->mem_win_addr;
1917         sphb->mig_mem_win_size = sphb->mem_win_size;
1918         sphb->mig_io_win_addr = sphb->io_win_addr;
1919         sphb->mig_io_win_size = sphb->io_win_size;
1920 
1921         if ((sphb->mem64_win_size != 0)
1922             && (sphb->mem64_win_addr
1923                 == (sphb->mem_win_addr + sphb->mem_win_size))) {
1924             sphb->mig_mem_win_size += sphb->mem64_win_size;
1925         }
1926     }
1927 }
1928 
1929 static int spapr_pci_post_load(void *opaque, int version_id)
1930 {
1931     sPAPRPHBState *sphb = opaque;
1932     gpointer key, value;
1933     int i;
1934 
1935     for (i = 0; i < sphb->msi_devs_num; ++i) {
1936         key = g_memdup(&sphb->msi_devs[i].key,
1937                        sizeof(sphb->msi_devs[i].key));
1938         value = g_memdup(&sphb->msi_devs[i].value,
1939                          sizeof(sphb->msi_devs[i].value));
1940         g_hash_table_insert(sphb->msi, key, value);
1941     }
1942     g_free(sphb->msi_devs);
1943     sphb->msi_devs = NULL;
1944     sphb->msi_devs_num = 0;
1945 
1946     return 0;
1947 }
1948 
1949 static bool pre_2_8_migration(void *opaque, int version_id)
1950 {
1951     sPAPRPHBState *sphb = opaque;
1952 
1953     return sphb->pre_2_8_migration;
1954 }
1955 
1956 static const VMStateDescription vmstate_spapr_pci = {
1957     .name = "spapr_pci",
1958     .version_id = 2,
1959     .minimum_version_id = 2,
1960     .pre_save = spapr_pci_pre_save,
1961     .post_load = spapr_pci_post_load,
1962     .fields = (VMStateField[]) {
1963         VMSTATE_UINT64_EQUAL(buid, sPAPRPHBState),
1964         VMSTATE_UINT32_TEST(mig_liobn, sPAPRPHBState, pre_2_8_migration),
1965         VMSTATE_UINT64_TEST(mig_mem_win_addr, sPAPRPHBState, pre_2_8_migration),
1966         VMSTATE_UINT64_TEST(mig_mem_win_size, sPAPRPHBState, pre_2_8_migration),
1967         VMSTATE_UINT64_TEST(mig_io_win_addr, sPAPRPHBState, pre_2_8_migration),
1968         VMSTATE_UINT64_TEST(mig_io_win_size, sPAPRPHBState, pre_2_8_migration),
1969         VMSTATE_STRUCT_ARRAY(lsi_table, sPAPRPHBState, PCI_NUM_PINS, 0,
1970                              vmstate_spapr_pci_lsi, struct spapr_pci_lsi),
1971         VMSTATE_INT32(msi_devs_num, sPAPRPHBState),
1972         VMSTATE_STRUCT_VARRAY_ALLOC(msi_devs, sPAPRPHBState, msi_devs_num, 0,
1973                                     vmstate_spapr_pci_msi, spapr_pci_msi_mig),
1974         VMSTATE_END_OF_LIST()
1975     },
1976 };
1977 
1978 static const char *spapr_phb_root_bus_path(PCIHostState *host_bridge,
1979                                            PCIBus *rootbus)
1980 {
1981     sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(host_bridge);
1982 
1983     return sphb->dtbusname;
1984 }
1985 
1986 static void spapr_phb_class_init(ObjectClass *klass, void *data)
1987 {
1988     PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass);
1989     DeviceClass *dc = DEVICE_CLASS(klass);
1990     HotplugHandlerClass *hp = HOTPLUG_HANDLER_CLASS(klass);
1991 
1992     hc->root_bus_path = spapr_phb_root_bus_path;
1993     dc->realize = spapr_phb_realize;
1994     dc->props = spapr_phb_properties;
1995     dc->reset = spapr_phb_reset;
1996     dc->vmsd = &vmstate_spapr_pci;
1997     set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
1998     hp->plug = spapr_phb_hot_plug_child;
1999     hp->unplug = spapr_phb_hot_unplug_child;
2000 }
2001 
2002 static const TypeInfo spapr_phb_info = {
2003     .name          = TYPE_SPAPR_PCI_HOST_BRIDGE,
2004     .parent        = TYPE_PCI_HOST_BRIDGE,
2005     .instance_size = sizeof(sPAPRPHBState),
2006     .class_init    = spapr_phb_class_init,
2007     .interfaces    = (InterfaceInfo[]) {
2008         { TYPE_HOTPLUG_HANDLER },
2009         { }
2010     }
2011 };
2012 
2013 PCIHostState *spapr_create_phb(sPAPRMachineState *spapr, int index)
2014 {
2015     DeviceState *dev;
2016 
2017     dev = qdev_create(NULL, TYPE_SPAPR_PCI_HOST_BRIDGE);
2018     qdev_prop_set_uint32(dev, "index", index);
2019     qdev_init_nofail(dev);
2020 
2021     return PCI_HOST_BRIDGE(dev);
2022 }
2023 
2024 typedef struct sPAPRFDT {
2025     void *fdt;
2026     int node_off;
2027     sPAPRPHBState *sphb;
2028 } sPAPRFDT;
2029 
2030 static void spapr_populate_pci_devices_dt(PCIBus *bus, PCIDevice *pdev,
2031                                           void *opaque)
2032 {
2033     PCIBus *sec_bus;
2034     sPAPRFDT *p = opaque;
2035     int offset;
2036     sPAPRFDT s_fdt;
2037 
2038     offset = spapr_create_pci_child_dt(p->sphb, pdev, p->fdt, p->node_off);
2039     if (!offset) {
2040         error_report("Failed to create pci child device tree node");
2041         return;
2042     }
2043 
2044     if ((pci_default_read_config(pdev, PCI_HEADER_TYPE, 1) !=
2045          PCI_HEADER_TYPE_BRIDGE)) {
2046         return;
2047     }
2048 
2049     sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
2050     if (!sec_bus) {
2051         return;
2052     }
2053 
2054     s_fdt.fdt = p->fdt;
2055     s_fdt.node_off = offset;
2056     s_fdt.sphb = p->sphb;
2057     pci_for_each_device_reverse(sec_bus, pci_bus_num(sec_bus),
2058                                 spapr_populate_pci_devices_dt,
2059                                 &s_fdt);
2060 }
2061 
2062 static void spapr_phb_pci_enumerate_bridge(PCIBus *bus, PCIDevice *pdev,
2063                                            void *opaque)
2064 {
2065     unsigned int *bus_no = opaque;
2066     unsigned int primary = *bus_no;
2067     unsigned int subordinate = 0xff;
2068     PCIBus *sec_bus = NULL;
2069 
2070     if ((pci_default_read_config(pdev, PCI_HEADER_TYPE, 1) !=
2071          PCI_HEADER_TYPE_BRIDGE)) {
2072         return;
2073     }
2074 
2075     (*bus_no)++;
2076     pci_default_write_config(pdev, PCI_PRIMARY_BUS, primary, 1);
2077     pci_default_write_config(pdev, PCI_SECONDARY_BUS, *bus_no, 1);
2078     pci_default_write_config(pdev, PCI_SUBORDINATE_BUS, *bus_no, 1);
2079 
2080     sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
2081     if (!sec_bus) {
2082         return;
2083     }
2084 
2085     pci_default_write_config(pdev, PCI_SUBORDINATE_BUS, subordinate, 1);
2086     pci_for_each_device(sec_bus, pci_bus_num(sec_bus),
2087                         spapr_phb_pci_enumerate_bridge, bus_no);
2088     pci_default_write_config(pdev, PCI_SUBORDINATE_BUS, *bus_no, 1);
2089 }
2090 
2091 static void spapr_phb_pci_enumerate(sPAPRPHBState *phb)
2092 {
2093     PCIBus *bus = PCI_HOST_BRIDGE(phb)->bus;
2094     unsigned int bus_no = 0;
2095 
2096     pci_for_each_device(bus, pci_bus_num(bus),
2097                         spapr_phb_pci_enumerate_bridge,
2098                         &bus_no);
2099 
2100 }
2101 
2102 int spapr_populate_pci_dt(sPAPRPHBState *phb,
2103                           uint32_t xics_phandle,
2104                           void *fdt)
2105 {
2106     int bus_off, i, j, ret;
2107     char nodename[FDT_NAME_MAX];
2108     uint32_t bus_range[] = { cpu_to_be32(0), cpu_to_be32(0xff) };
2109     struct {
2110         uint32_t hi;
2111         uint64_t child;
2112         uint64_t parent;
2113         uint64_t size;
2114     } QEMU_PACKED ranges[] = {
2115         {
2116             cpu_to_be32(b_ss(1)), cpu_to_be64(0),
2117             cpu_to_be64(phb->io_win_addr),
2118             cpu_to_be64(memory_region_size(&phb->iospace)),
2119         },
2120         {
2121             cpu_to_be32(b_ss(2)), cpu_to_be64(SPAPR_PCI_MEM_WIN_BUS_OFFSET),
2122             cpu_to_be64(phb->mem_win_addr),
2123             cpu_to_be64(phb->mem_win_size),
2124         },
2125         {
2126             cpu_to_be32(b_ss(3)), cpu_to_be64(phb->mem64_win_pciaddr),
2127             cpu_to_be64(phb->mem64_win_addr),
2128             cpu_to_be64(phb->mem64_win_size),
2129         },
2130     };
2131     const unsigned sizeof_ranges =
2132         (phb->mem64_win_size ? 3 : 2) * sizeof(ranges[0]);
2133     uint64_t bus_reg[] = { cpu_to_be64(phb->buid), 0 };
2134     uint32_t interrupt_map_mask[] = {
2135         cpu_to_be32(b_ddddd(-1)|b_fff(0)), 0x0, 0x0, cpu_to_be32(-1)};
2136     uint32_t interrupt_map[PCI_SLOT_MAX * PCI_NUM_PINS][7];
2137     uint32_t ddw_applicable[] = {
2138         cpu_to_be32(RTAS_IBM_QUERY_PE_DMA_WINDOW),
2139         cpu_to_be32(RTAS_IBM_CREATE_PE_DMA_WINDOW),
2140         cpu_to_be32(RTAS_IBM_REMOVE_PE_DMA_WINDOW)
2141     };
2142     uint32_t ddw_extensions[] = {
2143         cpu_to_be32(1),
2144         cpu_to_be32(RTAS_IBM_RESET_PE_DMA_WINDOW)
2145     };
2146     uint32_t associativity[] = {cpu_to_be32(0x4),
2147                                 cpu_to_be32(0x0),
2148                                 cpu_to_be32(0x0),
2149                                 cpu_to_be32(0x0),
2150                                 cpu_to_be32(phb->numa_node)};
2151     sPAPRTCETable *tcet;
2152     PCIBus *bus = PCI_HOST_BRIDGE(phb)->bus;
2153     sPAPRFDT s_fdt;
2154 
2155     /* Start populating the FDT */
2156     snprintf(nodename, FDT_NAME_MAX, "pci@%" PRIx64, phb->buid);
2157     bus_off = fdt_add_subnode(fdt, 0, nodename);
2158     if (bus_off < 0) {
2159         return bus_off;
2160     }
2161 
2162     /* Write PHB properties */
2163     _FDT(fdt_setprop_string(fdt, bus_off, "device_type", "pci"));
2164     _FDT(fdt_setprop_string(fdt, bus_off, "compatible", "IBM,Logical_PHB"));
2165     _FDT(fdt_setprop_cell(fdt, bus_off, "#address-cells", 0x3));
2166     _FDT(fdt_setprop_cell(fdt, bus_off, "#size-cells", 0x2));
2167     _FDT(fdt_setprop_cell(fdt, bus_off, "#interrupt-cells", 0x1));
2168     _FDT(fdt_setprop(fdt, bus_off, "used-by-rtas", NULL, 0));
2169     _FDT(fdt_setprop(fdt, bus_off, "bus-range", &bus_range, sizeof(bus_range)));
2170     _FDT(fdt_setprop(fdt, bus_off, "ranges", &ranges, sizeof_ranges));
2171     _FDT(fdt_setprop(fdt, bus_off, "reg", &bus_reg, sizeof(bus_reg)));
2172     _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pci-config-space-type", 0x1));
2173     _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pe-total-#msi", XICS_IRQS_SPAPR));
2174 
2175     /* Dynamic DMA window */
2176     if (phb->ddw_enabled) {
2177         _FDT(fdt_setprop(fdt, bus_off, "ibm,ddw-applicable", &ddw_applicable,
2178                          sizeof(ddw_applicable)));
2179         _FDT(fdt_setprop(fdt, bus_off, "ibm,ddw-extensions",
2180                          &ddw_extensions, sizeof(ddw_extensions)));
2181     }
2182 
2183     /* Advertise NUMA via ibm,associativity */
2184     if (phb->numa_node != -1) {
2185         _FDT(fdt_setprop(fdt, bus_off, "ibm,associativity", associativity,
2186                          sizeof(associativity)));
2187     }
2188 
2189     /* Build the interrupt-map, this must matches what is done
2190      * in pci_spapr_map_irq
2191      */
2192     _FDT(fdt_setprop(fdt, bus_off, "interrupt-map-mask",
2193                      &interrupt_map_mask, sizeof(interrupt_map_mask)));
2194     for (i = 0; i < PCI_SLOT_MAX; i++) {
2195         for (j = 0; j < PCI_NUM_PINS; j++) {
2196             uint32_t *irqmap = interrupt_map[i*PCI_NUM_PINS + j];
2197             int lsi_num = pci_spapr_swizzle(i, j);
2198 
2199             irqmap[0] = cpu_to_be32(b_ddddd(i)|b_fff(0));
2200             irqmap[1] = 0;
2201             irqmap[2] = 0;
2202             irqmap[3] = cpu_to_be32(j+1);
2203             irqmap[4] = cpu_to_be32(xics_phandle);
2204             irqmap[5] = cpu_to_be32(phb->lsi_table[lsi_num].irq);
2205             irqmap[6] = cpu_to_be32(0x8);
2206         }
2207     }
2208     /* Write interrupt map */
2209     _FDT(fdt_setprop(fdt, bus_off, "interrupt-map", &interrupt_map,
2210                      sizeof(interrupt_map)));
2211 
2212     tcet = spapr_tce_find_by_liobn(phb->dma_liobn[0]);
2213     if (!tcet) {
2214         return -1;
2215     }
2216     spapr_dma_dt(fdt, bus_off, "ibm,dma-window",
2217                  tcet->liobn, tcet->bus_offset,
2218                  tcet->nb_table << tcet->page_shift);
2219 
2220     /* Walk the bridges and program the bus numbers*/
2221     spapr_phb_pci_enumerate(phb);
2222     _FDT(fdt_setprop_cell(fdt, bus_off, "qemu,phb-enumerated", 0x1));
2223 
2224     /* Populate tree nodes with PCI devices attached */
2225     s_fdt.fdt = fdt;
2226     s_fdt.node_off = bus_off;
2227     s_fdt.sphb = phb;
2228     pci_for_each_device_reverse(bus, pci_bus_num(bus),
2229                                 spapr_populate_pci_devices_dt,
2230                                 &s_fdt);
2231 
2232     ret = spapr_drc_populate_dt(fdt, bus_off, OBJECT(phb),
2233                                 SPAPR_DR_CONNECTOR_TYPE_PCI);
2234     if (ret) {
2235         return ret;
2236     }
2237 
2238     return 0;
2239 }
2240 
2241 void spapr_pci_rtas_init(void)
2242 {
2243     spapr_rtas_register(RTAS_READ_PCI_CONFIG, "read-pci-config",
2244                         rtas_read_pci_config);
2245     spapr_rtas_register(RTAS_WRITE_PCI_CONFIG, "write-pci-config",
2246                         rtas_write_pci_config);
2247     spapr_rtas_register(RTAS_IBM_READ_PCI_CONFIG, "ibm,read-pci-config",
2248                         rtas_ibm_read_pci_config);
2249     spapr_rtas_register(RTAS_IBM_WRITE_PCI_CONFIG, "ibm,write-pci-config",
2250                         rtas_ibm_write_pci_config);
2251     if (msi_nonbroken) {
2252         spapr_rtas_register(RTAS_IBM_QUERY_INTERRUPT_SOURCE_NUMBER,
2253                             "ibm,query-interrupt-source-number",
2254                             rtas_ibm_query_interrupt_source_number);
2255         spapr_rtas_register(RTAS_IBM_CHANGE_MSI, "ibm,change-msi",
2256                             rtas_ibm_change_msi);
2257     }
2258 
2259     spapr_rtas_register(RTAS_IBM_SET_EEH_OPTION,
2260                         "ibm,set-eeh-option",
2261                         rtas_ibm_set_eeh_option);
2262     spapr_rtas_register(RTAS_IBM_GET_CONFIG_ADDR_INFO2,
2263                         "ibm,get-config-addr-info2",
2264                         rtas_ibm_get_config_addr_info2);
2265     spapr_rtas_register(RTAS_IBM_READ_SLOT_RESET_STATE2,
2266                         "ibm,read-slot-reset-state2",
2267                         rtas_ibm_read_slot_reset_state2);
2268     spapr_rtas_register(RTAS_IBM_SET_SLOT_RESET,
2269                         "ibm,set-slot-reset",
2270                         rtas_ibm_set_slot_reset);
2271     spapr_rtas_register(RTAS_IBM_CONFIGURE_PE,
2272                         "ibm,configure-pe",
2273                         rtas_ibm_configure_pe);
2274     spapr_rtas_register(RTAS_IBM_SLOT_ERROR_DETAIL,
2275                         "ibm,slot-error-detail",
2276                         rtas_ibm_slot_error_detail);
2277 }
2278 
2279 static void spapr_pci_register_types(void)
2280 {
2281     type_register_static(&spapr_phb_info);
2282 }
2283 
2284 type_init(spapr_pci_register_types)
2285 
2286 static int spapr_switch_one_vga(DeviceState *dev, void *opaque)
2287 {
2288     bool be = *(bool *)opaque;
2289 
2290     if (object_dynamic_cast(OBJECT(dev), "VGA")
2291         || object_dynamic_cast(OBJECT(dev), "secondary-vga")) {
2292         object_property_set_bool(OBJECT(dev), be, "big-endian-framebuffer",
2293                                  &error_abort);
2294     }
2295     return 0;
2296 }
2297 
2298 void spapr_pci_switch_vga(bool big_endian)
2299 {
2300     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
2301     sPAPRPHBState *sphb;
2302 
2303     /*
2304      * For backward compatibility with existing guests, we switch
2305      * the endianness of the VGA controller when changing the guest
2306      * interrupt mode
2307      */
2308     QLIST_FOREACH(sphb, &spapr->phbs, list) {
2309         BusState *bus = &PCI_HOST_BRIDGE(sphb)->bus->qbus;
2310         qbus_walk_children(bus, spapr_switch_one_vga, NULL, NULL, NULL,
2311                            &big_endian);
2312     }
2313 }
2314