2 * QEMU sPAPR PCI host originated from Uninorth PCI host
4 * Copyright (c) 2011 Alexey Kardashevskiy, IBM Corporation.
5 * Copyright (C) 2011 David Gibson, IBM Corporation.
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:
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
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
26 #include "qemu/osdep.h"
27 #include "qapi/error.h"
29 #include "hw/sysbus.h"
30 #include "migration/vmstate.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/ram_addr.h"
40 #include "qemu/error-report.h"
41 #include "qemu/module.h"
42 #include "hw/ppc/fdt.h"
43 #include "hw/pci/pci_bridge.h"
44 #include "hw/pci/pci_bus.h"
45 #include "hw/pci/pci_ids.h"
46 #include "hw/ppc/spapr_drc.h"
47 #include "hw/qdev-properties.h"
48 #include "sysemu/device_tree.h"
49 #include "sysemu/kvm.h"
50 #include "sysemu/hostmem.h"
51 #include "sysemu/numa.h"
52 #include "hw/ppc/spapr_numa.h"
55 /* Copied from the kernel arch/powerpc/platforms/pseries/msi.c */
56 #define RTAS_QUERY_FN 0
57 #define RTAS_CHANGE_FN 1
58 #define RTAS_RESET_FN 2
59 #define RTAS_CHANGE_MSI_FN 3
60 #define RTAS_CHANGE_MSIX_FN 4
62 /* Interrupt types to return on RTAS_CHANGE_* */
63 #define RTAS_TYPE_MSI 1
64 #define RTAS_TYPE_MSIX 2
66 SpaprPhbState
*spapr_pci_find_phb(SpaprMachineState
*spapr
, uint64_t buid
)
70 QLIST_FOREACH(sphb
, &spapr
->phbs
, list
) {
71 if (sphb
->buid
!= buid
) {
80 PCIDevice
*spapr_pci_find_dev(SpaprMachineState
*spapr
, uint64_t buid
,
83 SpaprPhbState
*sphb
= spapr_pci_find_phb(spapr
, buid
);
84 PCIHostState
*phb
= PCI_HOST_BRIDGE(sphb
);
85 int bus_num
= (config_addr
>> 16) & 0xFF;
86 int devfn
= (config_addr
>> 8) & 0xFF;
92 return pci_find_device(phb
->bus
, bus_num
, devfn
);
95 static uint32_t rtas_pci_cfgaddr(uint32_t arg
)
97 /* This handles the encoding of extended config space addresses */
98 return ((arg
>> 20) & 0xf00) | (arg
& 0xff);
101 static void finish_read_pci_config(SpaprMachineState
*spapr
, uint64_t buid
,
102 uint32_t addr
, uint32_t size
,
108 if ((size
!= 1) && (size
!= 2) && (size
!= 4)) {
109 /* access must be 1, 2 or 4 bytes */
110 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
114 pci_dev
= spapr_pci_find_dev(spapr
, buid
, addr
);
115 addr
= rtas_pci_cfgaddr(addr
);
117 if (!pci_dev
|| (addr
% size
) || (addr
>= pci_config_size(pci_dev
))) {
118 /* Access must be to a valid device, within bounds and
119 * naturally aligned */
120 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
124 val
= pci_host_config_read_common(pci_dev
, addr
,
125 pci_config_size(pci_dev
), size
);
127 rtas_st(rets
, 0, RTAS_OUT_SUCCESS
);
128 rtas_st(rets
, 1, val
);
131 static void rtas_ibm_read_pci_config(PowerPCCPU
*cpu
, SpaprMachineState
*spapr
,
132 uint32_t token
, uint32_t nargs
,
134 uint32_t nret
, target_ulong rets
)
139 if ((nargs
!= 4) || (nret
!= 2)) {
140 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
144 buid
= rtas_ldq(args
, 1);
145 size
= rtas_ld(args
, 3);
146 addr
= rtas_ld(args
, 0);
148 finish_read_pci_config(spapr
, buid
, addr
, size
, rets
);
151 static void rtas_read_pci_config(PowerPCCPU
*cpu
, SpaprMachineState
*spapr
,
152 uint32_t token
, uint32_t nargs
,
154 uint32_t nret
, target_ulong rets
)
158 if ((nargs
!= 2) || (nret
!= 2)) {
159 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
163 size
= rtas_ld(args
, 1);
164 addr
= rtas_ld(args
, 0);
166 finish_read_pci_config(spapr
, 0, addr
, size
, rets
);
169 static void finish_write_pci_config(SpaprMachineState
*spapr
, uint64_t buid
,
170 uint32_t addr
, uint32_t size
,
171 uint32_t val
, target_ulong rets
)
175 if ((size
!= 1) && (size
!= 2) && (size
!= 4)) {
176 /* access must be 1, 2 or 4 bytes */
177 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
181 pci_dev
= spapr_pci_find_dev(spapr
, buid
, addr
);
182 addr
= rtas_pci_cfgaddr(addr
);
184 if (!pci_dev
|| (addr
% size
) || (addr
>= pci_config_size(pci_dev
))) {
185 /* Access must be to a valid device, within bounds and
186 * naturally aligned */
187 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
191 pci_host_config_write_common(pci_dev
, addr
, pci_config_size(pci_dev
),
194 rtas_st(rets
, 0, RTAS_OUT_SUCCESS
);
197 static void rtas_ibm_write_pci_config(PowerPCCPU
*cpu
, SpaprMachineState
*spapr
,
198 uint32_t token
, uint32_t nargs
,
200 uint32_t nret
, target_ulong rets
)
203 uint32_t val
, size
, addr
;
205 if ((nargs
!= 5) || (nret
!= 1)) {
206 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
210 buid
= rtas_ldq(args
, 1);
211 val
= rtas_ld(args
, 4);
212 size
= rtas_ld(args
, 3);
213 addr
= rtas_ld(args
, 0);
215 finish_write_pci_config(spapr
, buid
, addr
, size
, val
, rets
);
218 static void rtas_write_pci_config(PowerPCCPU
*cpu
, SpaprMachineState
*spapr
,
219 uint32_t token
, uint32_t nargs
,
221 uint32_t nret
, target_ulong rets
)
223 uint32_t val
, size
, addr
;
225 if ((nargs
!= 3) || (nret
!= 1)) {
226 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
231 val
= rtas_ld(args
, 2);
232 size
= rtas_ld(args
, 1);
233 addr
= rtas_ld(args
, 0);
235 finish_write_pci_config(spapr
, 0, addr
, size
, val
, rets
);
239 * Set MSI/MSIX message data.
240 * This is required for msi_notify()/msix_notify() which
241 * will write at the addresses via spapr_msi_write().
243 * If hwaddr == 0, all entries will have .data == first_irq i.e.
244 * table will be reset.
246 static void spapr_msi_setmsg(PCIDevice
*pdev
, hwaddr addr
, bool msix
,
247 unsigned first_irq
, unsigned req_num
)
250 MSIMessage msg
= { .address
= addr
, .data
= first_irq
};
253 msi_set_message(pdev
, msg
);
254 trace_spapr_pci_msi_setup(pdev
->name
, 0, msg
.address
);
258 for (i
= 0; i
< req_num
; ++i
) {
259 msix_set_message(pdev
, i
, msg
);
260 trace_spapr_pci_msi_setup(pdev
->name
, i
, msg
.address
);
267 static void rtas_ibm_change_msi(PowerPCCPU
*cpu
, SpaprMachineState
*spapr
,
268 uint32_t token
, uint32_t nargs
,
269 target_ulong args
, uint32_t nret
,
272 SpaprMachineClass
*smc
= SPAPR_MACHINE_GET_CLASS(spapr
);
273 uint32_t config_addr
= rtas_ld(args
, 0);
274 uint64_t buid
= rtas_ldq(args
, 1);
275 unsigned int func
= rtas_ld(args
, 3);
276 unsigned int req_num
= rtas_ld(args
, 4); /* 0 == remove all */
277 unsigned int seq_num
= rtas_ld(args
, 5);
278 unsigned int ret_intr_type
;
279 unsigned int irq
, max_irqs
= 0;
280 SpaprPhbState
*phb
= NULL
;
281 PCIDevice
*pdev
= NULL
;
283 int *config_addr_key
;
287 /* Fins SpaprPhbState */
288 phb
= spapr_pci_find_phb(spapr
, buid
);
290 pdev
= spapr_pci_find_dev(spapr
, buid
, config_addr
);
293 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
299 if (msi_present(pdev
)) {
300 ret_intr_type
= RTAS_TYPE_MSI
;
301 } else if (msix_present(pdev
)) {
302 ret_intr_type
= RTAS_TYPE_MSIX
;
304 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
308 case RTAS_CHANGE_MSI_FN
:
309 if (msi_present(pdev
)) {
310 ret_intr_type
= RTAS_TYPE_MSI
;
312 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
316 case RTAS_CHANGE_MSIX_FN
:
317 if (msix_present(pdev
)) {
318 ret_intr_type
= RTAS_TYPE_MSIX
;
320 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
325 error_report("rtas_ibm_change_msi(%u) is not implemented", func
);
326 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
330 msi
= (SpaprPciMsi
*) g_hash_table_lookup(phb
->msi
, &config_addr
);
335 trace_spapr_pci_msi("Releasing wrong config", config_addr
);
336 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
340 if (msi_present(pdev
)) {
341 spapr_msi_setmsg(pdev
, 0, false, 0, 0);
343 if (msix_present(pdev
)) {
344 spapr_msi_setmsg(pdev
, 0, true, 0, 0);
346 g_hash_table_remove(phb
->msi
, &config_addr
);
348 trace_spapr_pci_msi("Released MSIs", config_addr
);
349 rtas_st(rets
, 0, RTAS_OUT_SUCCESS
);
356 /* Check if the device supports as many IRQs as requested */
357 if (ret_intr_type
== RTAS_TYPE_MSI
) {
358 max_irqs
= msi_nr_vectors_allocated(pdev
);
359 } else if (ret_intr_type
== RTAS_TYPE_MSIX
) {
360 max_irqs
= pdev
->msix_entries_nr
;
363 error_report("Requested interrupt type %d is not enabled for device %x",
364 ret_intr_type
, config_addr
);
365 rtas_st(rets
, 0, -1); /* Hardware error */
368 /* Correct the number if the guest asked for too many */
369 if (req_num
> max_irqs
) {
370 trace_spapr_pci_msi_retry(config_addr
, req_num
, max_irqs
);
372 irq
= 0; /* to avoid misleading trace */
377 if (smc
->legacy_irq_allocation
) {
378 irq
= spapr_irq_find(spapr
, req_num
, ret_intr_type
== RTAS_TYPE_MSI
,
381 irq
= spapr_irq_msi_alloc(spapr
, req_num
,
382 ret_intr_type
== RTAS_TYPE_MSI
, &err
);
385 error_reportf_err(err
, "Can't allocate MSIs for device %x: ",
387 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
391 for (i
= 0; i
< req_num
; i
++) {
392 spapr_irq_claim(spapr
, irq
+ i
, false, &err
);
395 spapr_irq_free(spapr
, irq
, i
);
397 if (!smc
->legacy_irq_allocation
) {
398 spapr_irq_msi_free(spapr
, irq
, req_num
);
400 error_reportf_err(err
, "Can't allocate MSIs for device %x: ",
402 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
407 /* Release previous MSIs */
409 g_hash_table_remove(phb
->msi
, &config_addr
);
412 /* Setup MSI/MSIX vectors in the device (via cfgspace or MSIX BAR) */
413 spapr_msi_setmsg(pdev
, SPAPR_PCI_MSI_WINDOW
, ret_intr_type
== RTAS_TYPE_MSIX
,
416 /* Add MSI device to cache */
417 msi
= g_new(SpaprPciMsi
, 1);
418 msi
->first_irq
= irq
;
420 config_addr_key
= g_new(int, 1);
421 *config_addr_key
= config_addr
;
422 g_hash_table_insert(phb
->msi
, config_addr_key
, msi
);
425 rtas_st(rets
, 0, RTAS_OUT_SUCCESS
);
426 rtas_st(rets
, 1, req_num
);
427 rtas_st(rets
, 2, ++seq_num
);
429 rtas_st(rets
, 3, ret_intr_type
);
432 trace_spapr_pci_rtas_ibm_change_msi(config_addr
, func
, req_num
, irq
);
435 static void rtas_ibm_query_interrupt_source_number(PowerPCCPU
*cpu
,
436 SpaprMachineState
*spapr
,
443 uint32_t config_addr
= rtas_ld(args
, 0);
444 uint64_t buid
= rtas_ldq(args
, 1);
445 unsigned int intr_src_num
= -1, ioa_intr_num
= rtas_ld(args
, 3);
446 SpaprPhbState
*phb
= NULL
;
447 PCIDevice
*pdev
= NULL
;
450 /* Find SpaprPhbState */
451 phb
= spapr_pci_find_phb(spapr
, buid
);
453 pdev
= spapr_pci_find_dev(spapr
, buid
, config_addr
);
456 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
460 /* Find device descriptor and start IRQ */
461 msi
= (SpaprPciMsi
*) g_hash_table_lookup(phb
->msi
, &config_addr
);
462 if (!msi
|| !msi
->first_irq
|| !msi
->num
|| (ioa_intr_num
>= msi
->num
)) {
463 trace_spapr_pci_msi("Failed to return vector", config_addr
);
464 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
467 intr_src_num
= msi
->first_irq
+ ioa_intr_num
;
468 trace_spapr_pci_rtas_ibm_query_interrupt_source_number(ioa_intr_num
,
471 rtas_st(rets
, 0, RTAS_OUT_SUCCESS
);
472 rtas_st(rets
, 1, intr_src_num
);
473 rtas_st(rets
, 2, 1);/* 0 == level; 1 == edge */
476 static void rtas_ibm_set_eeh_option(PowerPCCPU
*cpu
,
477 SpaprMachineState
*spapr
,
478 uint32_t token
, uint32_t nargs
,
479 target_ulong args
, uint32_t nret
,
483 uint32_t addr
, option
;
487 if ((nargs
!= 4) || (nret
!= 1)) {
488 goto param_error_exit
;
491 buid
= rtas_ldq(args
, 1);
492 addr
= rtas_ld(args
, 0);
493 option
= rtas_ld(args
, 3);
495 sphb
= spapr_pci_find_phb(spapr
, buid
);
497 goto param_error_exit
;
500 if (!spapr_phb_eeh_available(sphb
)) {
501 goto param_error_exit
;
504 ret
= spapr_phb_vfio_eeh_set_option(sphb
, addr
, option
);
505 rtas_st(rets
, 0, ret
);
509 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
512 static void rtas_ibm_get_config_addr_info2(PowerPCCPU
*cpu
,
513 SpaprMachineState
*spapr
,
514 uint32_t token
, uint32_t nargs
,
515 target_ulong args
, uint32_t nret
,
520 uint32_t addr
, option
;
523 if ((nargs
!= 4) || (nret
!= 2)) {
524 goto param_error_exit
;
527 buid
= rtas_ldq(args
, 1);
528 sphb
= spapr_pci_find_phb(spapr
, buid
);
530 goto param_error_exit
;
533 if (!spapr_phb_eeh_available(sphb
)) {
534 goto param_error_exit
;
538 * We always have PE address of form "00BB0001". "BB"
539 * represents the bus number of PE's primary bus.
541 option
= rtas_ld(args
, 3);
543 case RTAS_GET_PE_ADDR
:
544 addr
= rtas_ld(args
, 0);
545 pdev
= spapr_pci_find_dev(spapr
, buid
, addr
);
547 goto param_error_exit
;
550 rtas_st(rets
, 1, (pci_bus_num(pci_get_bus(pdev
)) << 16) + 1);
552 case RTAS_GET_PE_MODE
:
553 rtas_st(rets
, 1, RTAS_PE_MODE_SHARED
);
556 goto param_error_exit
;
559 rtas_st(rets
, 0, RTAS_OUT_SUCCESS
);
563 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
566 static void rtas_ibm_read_slot_reset_state2(PowerPCCPU
*cpu
,
567 SpaprMachineState
*spapr
,
568 uint32_t token
, uint32_t nargs
,
569 target_ulong args
, uint32_t nret
,
576 if ((nargs
!= 3) || (nret
!= 4 && nret
!= 5)) {
577 goto param_error_exit
;
580 buid
= rtas_ldq(args
, 1);
581 sphb
= spapr_pci_find_phb(spapr
, buid
);
583 goto param_error_exit
;
586 if (!spapr_phb_eeh_available(sphb
)) {
587 goto param_error_exit
;
590 ret
= spapr_phb_vfio_eeh_get_state(sphb
, &state
);
591 rtas_st(rets
, 0, ret
);
592 if (ret
!= RTAS_OUT_SUCCESS
) {
596 rtas_st(rets
, 1, state
);
597 rtas_st(rets
, 2, RTAS_EEH_SUPPORT
);
598 rtas_st(rets
, 3, RTAS_EEH_PE_UNAVAIL_INFO
);
600 rtas_st(rets
, 4, RTAS_EEH_PE_RECOVER_INFO
);
605 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
608 static void rtas_ibm_set_slot_reset(PowerPCCPU
*cpu
,
609 SpaprMachineState
*spapr
,
610 uint32_t token
, uint32_t nargs
,
611 target_ulong args
, uint32_t nret
,
619 if ((nargs
!= 4) || (nret
!= 1)) {
620 goto param_error_exit
;
623 buid
= rtas_ldq(args
, 1);
624 option
= rtas_ld(args
, 3);
625 sphb
= spapr_pci_find_phb(spapr
, buid
);
627 goto param_error_exit
;
630 if (!spapr_phb_eeh_available(sphb
)) {
631 goto param_error_exit
;
634 ret
= spapr_phb_vfio_eeh_reset(sphb
, option
);
635 rtas_st(rets
, 0, ret
);
639 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
642 static void rtas_ibm_configure_pe(PowerPCCPU
*cpu
,
643 SpaprMachineState
*spapr
,
644 uint32_t token
, uint32_t nargs
,
645 target_ulong args
, uint32_t nret
,
652 if ((nargs
!= 3) || (nret
!= 1)) {
653 goto param_error_exit
;
656 buid
= rtas_ldq(args
, 1);
657 sphb
= spapr_pci_find_phb(spapr
, buid
);
659 goto param_error_exit
;
662 if (!spapr_phb_eeh_available(sphb
)) {
663 goto param_error_exit
;
666 ret
= spapr_phb_vfio_eeh_configure(sphb
);
667 rtas_st(rets
, 0, ret
);
671 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
674 /* To support it later */
675 static void rtas_ibm_slot_error_detail(PowerPCCPU
*cpu
,
676 SpaprMachineState
*spapr
,
677 uint32_t token
, uint32_t nargs
,
678 target_ulong args
, uint32_t nret
,
685 if ((nargs
!= 8) || (nret
!= 1)) {
686 goto param_error_exit
;
689 buid
= rtas_ldq(args
, 1);
690 sphb
= spapr_pci_find_phb(spapr
, buid
);
692 goto param_error_exit
;
695 if (!spapr_phb_eeh_available(sphb
)) {
696 goto param_error_exit
;
699 option
= rtas_ld(args
, 7);
701 case RTAS_SLOT_TEMP_ERR_LOG
:
702 case RTAS_SLOT_PERM_ERR_LOG
:
705 goto param_error_exit
;
708 /* We don't have error log yet */
709 rtas_st(rets
, 0, RTAS_OUT_NO_ERRORS_FOUND
);
713 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
716 static void pci_spapr_set_irq(void *opaque
, int irq_num
, int level
)
719 * Here we use the number returned by pci_swizzle_map_irq_fn to find a
720 * corresponding qemu_irq.
722 SpaprPhbState
*phb
= opaque
;
723 SpaprMachineState
*spapr
= SPAPR_MACHINE(qdev_get_machine());
725 trace_spapr_pci_lsi_set(phb
->dtbusname
, irq_num
, phb
->lsi_table
[irq_num
].irq
);
726 qemu_set_irq(spapr_qirq(spapr
, phb
->lsi_table
[irq_num
].irq
), level
);
729 static PCIINTxRoute
spapr_route_intx_pin_to_irq(void *opaque
, int pin
)
731 SpaprPhbState
*sphb
= SPAPR_PCI_HOST_BRIDGE(opaque
);
734 route
.mode
= PCI_INTX_ENABLED
;
735 route
.irq
= sphb
->lsi_table
[pin
].irq
;
740 static uint64_t spapr_msi_read(void *opaque
, hwaddr addr
, unsigned size
)
742 qemu_log_mask(LOG_GUEST_ERROR
, "%s: invalid access\n", __func__
);
747 * MSI/MSIX memory region implementation.
748 * The handler handles both MSI and MSIX.
749 * The vector number is encoded in least bits in data.
751 static void spapr_msi_write(void *opaque
, hwaddr addr
,
752 uint64_t data
, unsigned size
)
754 SpaprMachineState
*spapr
= opaque
;
757 trace_spapr_pci_msi_write(addr
, data
, irq
);
759 qemu_irq_pulse(spapr_qirq(spapr
, irq
));
762 static const MemoryRegionOps spapr_msi_ops
= {
764 * .read result is undefined by PCI spec.
765 * define .read method to avoid assert failure in memory_region_init_io
767 .read
= spapr_msi_read
,
768 .write
= spapr_msi_write
,
769 .endianness
= DEVICE_LITTLE_ENDIAN
775 static AddressSpace
*spapr_pci_dma_iommu(PCIBus
*bus
, void *opaque
, int devfn
)
777 SpaprPhbState
*phb
= opaque
;
779 return &phb
->iommu_as
;
782 static const PCIIOMMUOps spapr_iommu_ops
= {
783 .get_address_space
= spapr_pci_dma_iommu
,
786 static char *spapr_phb_vfio_get_loc_code(SpaprPhbState
*sphb
, PCIDevice
*pdev
)
788 g_autofree
char *path
= NULL
;
789 g_autofree
char *host
= NULL
;
790 g_autofree
char *devspec
= NULL
;
793 /* Get the PCI VFIO host id */
794 host
= object_property_get_str(OBJECT(pdev
), "host", NULL
);
799 /* Construct the path of the file that will give us the DT location */
800 path
= g_strdup_printf("/sys/bus/pci/devices/%s/devspec", host
);
801 if (!g_file_get_contents(path
, &devspec
, NULL
, NULL
)) {
805 /* Construct and read from host device tree the loc-code */
807 path
= g_strdup_printf("/proc/device-tree%s/ibm,loc-code", devspec
);
808 if (!g_file_get_contents(path
, &buf
, NULL
, NULL
)) {
814 static char *spapr_phb_get_loc_code(SpaprPhbState
*sphb
, PCIDevice
*pdev
)
817 const char *devtype
= "qemu";
818 uint32_t busnr
= pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(pdev
))));
820 if (object_dynamic_cast(OBJECT(pdev
), "vfio-pci")) {
821 buf
= spapr_phb_vfio_get_loc_code(sphb
, pdev
);
828 * For emulated devices and VFIO-failure case, make up
831 buf
= g_strdup_printf("%s_%s:%04x:%02x:%02x.%x",
832 devtype
, pdev
->name
, sphb
->index
, busnr
,
833 PCI_SLOT(pdev
->devfn
), PCI_FUNC(pdev
->devfn
));
837 /* Macros to operate with address in OF binding to PCI */
838 #define b_x(x, p, l) (((x) & ((1<<(l))-1)) << (p))
839 #define b_n(x) b_x((x), 31, 1) /* 0 if relocatable */
840 #define b_p(x) b_x((x), 30, 1) /* 1 if prefetchable */
841 #define b_t(x) b_x((x), 29, 1) /* 1 if the address is aliased */
842 #define b_ss(x) b_x((x), 24, 2) /* the space code */
843 #define b_bbbbbbbb(x) b_x((x), 16, 8) /* bus number */
844 #define b_ddddd(x) b_x((x), 11, 5) /* device number */
845 #define b_fff(x) b_x((x), 8, 3) /* function number */
846 #define b_rrrrrrrr(x) b_x((x), 0, 8) /* register number */
848 /* for 'reg' OF properties */
849 #define RESOURCE_CELLS_SIZE 2
850 #define RESOURCE_CELLS_ADDRESS 3
852 typedef struct ResourceFields
{
858 } QEMU_PACKED ResourceFields
;
860 typedef struct ResourceProps
{
861 ResourceFields reg
[8];
865 /* fill in the 'reg' OF properties for
866 * a PCI device. 'reg' describes resource requirements for a
867 * device's IO/MEM regions.
869 * the property is an array of ('phys-addr', 'size') pairs describing
870 * the addressable regions of the PCI device, where 'phys-addr' is a
871 * RESOURCE_CELLS_ADDRESS-tuple of 32-bit integers corresponding to
872 * (phys.hi, phys.mid, phys.lo), and 'size' is a
873 * RESOURCE_CELLS_SIZE-tuple corresponding to (size.hi, size.lo).
875 * phys.hi = 0xYYXXXXZZ, where:
880 * ||| + 00 if configuration space
881 * ||| + 01 if IO region,
882 * ||| + 10 if 32-bit MEM region
883 * ||| + 11 if 64-bit MEM region
885 * ||+------ for non-relocatable IO: 1 if aliased
886 * || for relocatable IO: 1 if below 64KB
887 * || for MEM: 1 if below 1MB
888 * |+------- 1 if region is prefetchable
889 * +-------- 1 if region is non-relocatable
890 * 0xXXXX = bbbbbbbb dddddfff, encoding bus, slot, and function
892 * 0xZZ = rrrrrrrr, the register number of the BAR corresponding
895 * phys.mid and phys.lo correspond respectively to the hi/lo portions
896 * of the actual address of the region.
898 * note also that addresses defined in this property are, at least
899 * for PAPR guests, relative to the PHBs IO/MEM windows, and
900 * correspond directly to the addresses in the BARs.
902 * in accordance with PCI Bus Binding to Open Firmware,
903 * IEEE Std 1275-1994, section 4.1.1, as implemented by PAPR+ v2.7,
906 static void populate_resource_props(PCIDevice
*d
, ResourceProps
*rp
)
908 int bus_num
= pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(d
))));
909 uint32_t dev_id
= (b_bbbbbbbb(bus_num
) |
910 b_ddddd(PCI_SLOT(d
->devfn
)) |
911 b_fff(PCI_FUNC(d
->devfn
)));
915 /* config space region */
916 reg
= &rp
->reg
[reg_idx
++];
917 reg
->phys_hi
= cpu_to_be32(dev_id
);
923 for (i
= 0; i
< PCI_NUM_REGIONS
; i
++) {
924 if (!d
->io_regions
[i
].size
) {
928 reg
= &rp
->reg
[reg_idx
++];
930 reg
->phys_hi
= cpu_to_be32(dev_id
| b_rrrrrrrr(pci_bar(d
, i
)));
931 if (d
->io_regions
[i
].type
& PCI_BASE_ADDRESS_SPACE_IO
) {
932 reg
->phys_hi
|= cpu_to_be32(b_ss(1));
933 } else if (d
->io_regions
[i
].type
& PCI_BASE_ADDRESS_MEM_TYPE_64
) {
934 reg
->phys_hi
|= cpu_to_be32(b_ss(3));
936 reg
->phys_hi
|= cpu_to_be32(b_ss(2));
940 reg
->size_hi
= cpu_to_be32(d
->io_regions
[i
].size
>> 32);
941 reg
->size_lo
= cpu_to_be32(d
->io_regions
[i
].size
);
944 rp
->reg_len
= reg_idx
* sizeof(ResourceFields
);
947 typedef struct PCIClass PCIClass
;
948 typedef struct PCISubClass PCISubClass
;
949 typedef struct PCIIFace PCIIFace
;
959 const PCIIFace
*iface
;
964 const PCISubClass
*subc
;
967 static const PCISubClass undef_subclass
[] = {
968 { PCI_CLASS_NOT_DEFINED_VGA
, "display", NULL
},
969 { 0xFF, NULL
, NULL
},
972 static const PCISubClass mass_subclass
[] = {
973 { PCI_CLASS_STORAGE_SCSI
, "scsi", NULL
},
974 { PCI_CLASS_STORAGE_IDE
, "ide", NULL
},
975 { PCI_CLASS_STORAGE_FLOPPY
, "fdc", NULL
},
976 { PCI_CLASS_STORAGE_IPI
, "ipi", NULL
},
977 { PCI_CLASS_STORAGE_RAID
, "raid", NULL
},
978 { PCI_CLASS_STORAGE_ATA
, "ata", NULL
},
979 { PCI_CLASS_STORAGE_SATA
, "sata", NULL
},
980 { PCI_CLASS_STORAGE_SAS
, "sas", NULL
},
981 { 0xFF, NULL
, NULL
},
984 static const PCISubClass net_subclass
[] = {
985 { PCI_CLASS_NETWORK_ETHERNET
, "ethernet", NULL
},
986 { PCI_CLASS_NETWORK_TOKEN_RING
, "token-ring", NULL
},
987 { PCI_CLASS_NETWORK_FDDI
, "fddi", NULL
},
988 { PCI_CLASS_NETWORK_ATM
, "atm", NULL
},
989 { PCI_CLASS_NETWORK_ISDN
, "isdn", NULL
},
990 { PCI_CLASS_NETWORK_WORLDFIP
, "worldfip", NULL
},
991 { PCI_CLASS_NETWORK_PICMG214
, "picmg", NULL
},
992 { 0xFF, NULL
, NULL
},
995 static const PCISubClass displ_subclass
[] = {
996 { PCI_CLASS_DISPLAY_VGA
, "vga", NULL
},
997 { PCI_CLASS_DISPLAY_XGA
, "xga", NULL
},
998 { PCI_CLASS_DISPLAY_3D
, "3d-controller", NULL
},
999 { 0xFF, NULL
, NULL
},
1002 static const PCISubClass media_subclass
[] = {
1003 { PCI_CLASS_MULTIMEDIA_VIDEO
, "video", NULL
},
1004 { PCI_CLASS_MULTIMEDIA_AUDIO
, "sound", NULL
},
1005 { PCI_CLASS_MULTIMEDIA_PHONE
, "telephony", NULL
},
1006 { 0xFF, NULL
, NULL
},
1009 static const PCISubClass mem_subclass
[] = {
1010 { PCI_CLASS_MEMORY_RAM
, "memory", NULL
},
1011 { PCI_CLASS_MEMORY_FLASH
, "flash", NULL
},
1012 { 0xFF, NULL
, NULL
},
1015 static const PCISubClass bridg_subclass
[] = {
1016 { PCI_CLASS_BRIDGE_HOST
, "host", NULL
},
1017 { PCI_CLASS_BRIDGE_ISA
, "isa", NULL
},
1018 { PCI_CLASS_BRIDGE_EISA
, "eisa", NULL
},
1019 { PCI_CLASS_BRIDGE_MC
, "mca", NULL
},
1020 { PCI_CLASS_BRIDGE_PCI
, "pci", NULL
},
1021 { PCI_CLASS_BRIDGE_PCMCIA
, "pcmcia", NULL
},
1022 { PCI_CLASS_BRIDGE_NUBUS
, "nubus", NULL
},
1023 { PCI_CLASS_BRIDGE_CARDBUS
, "cardbus", NULL
},
1024 { PCI_CLASS_BRIDGE_RACEWAY
, "raceway", NULL
},
1025 { PCI_CLASS_BRIDGE_PCI_SEMITP
, "semi-transparent-pci", NULL
},
1026 { PCI_CLASS_BRIDGE_IB_PCI
, "infiniband", NULL
},
1027 { 0xFF, NULL
, NULL
},
1030 static const PCISubClass comm_subclass
[] = {
1031 { PCI_CLASS_COMMUNICATION_SERIAL
, "serial", NULL
},
1032 { PCI_CLASS_COMMUNICATION_PARALLEL
, "parallel", NULL
},
1033 { PCI_CLASS_COMMUNICATION_MULTISERIAL
, "multiport-serial", NULL
},
1034 { PCI_CLASS_COMMUNICATION_MODEM
, "modem", NULL
},
1035 { PCI_CLASS_COMMUNICATION_GPIB
, "gpib", NULL
},
1036 { PCI_CLASS_COMMUNICATION_SC
, "smart-card", NULL
},
1037 { 0xFF, NULL
, NULL
, },
1040 static const PCIIFace pic_iface
[] = {
1041 { PCI_CLASS_SYSTEM_PIC_IOAPIC
, "io-apic" },
1042 { PCI_CLASS_SYSTEM_PIC_IOXAPIC
, "io-xapic" },
1046 static const PCISubClass sys_subclass
[] = {
1047 { PCI_CLASS_SYSTEM_PIC
, "interrupt-controller", pic_iface
},
1048 { PCI_CLASS_SYSTEM_DMA
, "dma-controller", NULL
},
1049 { PCI_CLASS_SYSTEM_TIMER
, "timer", NULL
},
1050 { PCI_CLASS_SYSTEM_RTC
, "rtc", NULL
},
1051 { PCI_CLASS_SYSTEM_PCI_HOTPLUG
, "hot-plug-controller", NULL
},
1052 { PCI_CLASS_SYSTEM_SDHCI
, "sd-host-controller", NULL
},
1053 { 0xFF, NULL
, NULL
},
1056 static const PCISubClass inp_subclass
[] = {
1057 { PCI_CLASS_INPUT_KEYBOARD
, "keyboard", NULL
},
1058 { PCI_CLASS_INPUT_PEN
, "pen", NULL
},
1059 { PCI_CLASS_INPUT_MOUSE
, "mouse", NULL
},
1060 { PCI_CLASS_INPUT_SCANNER
, "scanner", NULL
},
1061 { PCI_CLASS_INPUT_GAMEPORT
, "gameport", NULL
},
1062 { 0xFF, NULL
, NULL
},
1065 static const PCISubClass dock_subclass
[] = {
1066 { PCI_CLASS_DOCKING_GENERIC
, "dock", NULL
},
1067 { 0xFF, NULL
, NULL
},
1070 static const PCISubClass cpu_subclass
[] = {
1071 { PCI_CLASS_PROCESSOR_PENTIUM
, "pentium", NULL
},
1072 { PCI_CLASS_PROCESSOR_POWERPC
, "powerpc", NULL
},
1073 { PCI_CLASS_PROCESSOR_MIPS
, "mips", NULL
},
1074 { PCI_CLASS_PROCESSOR_CO
, "co-processor", NULL
},
1075 { 0xFF, NULL
, NULL
},
1078 static const PCIIFace usb_iface
[] = {
1079 { PCI_CLASS_SERIAL_USB_UHCI
, "usb-uhci" },
1080 { PCI_CLASS_SERIAL_USB_OHCI
, "usb-ohci", },
1081 { PCI_CLASS_SERIAL_USB_EHCI
, "usb-ehci" },
1082 { PCI_CLASS_SERIAL_USB_XHCI
, "usb-xhci" },
1083 { PCI_CLASS_SERIAL_USB_UNKNOWN
, "usb-unknown" },
1084 { PCI_CLASS_SERIAL_USB_DEVICE
, "usb-device" },
1088 static const PCISubClass ser_subclass
[] = {
1089 { PCI_CLASS_SERIAL_FIREWIRE
, "firewire", NULL
},
1090 { PCI_CLASS_SERIAL_ACCESS
, "access-bus", NULL
},
1091 { PCI_CLASS_SERIAL_SSA
, "ssa", NULL
},
1092 { PCI_CLASS_SERIAL_USB
, "usb", usb_iface
},
1093 { PCI_CLASS_SERIAL_FIBER
, "fibre-channel", NULL
},
1094 { PCI_CLASS_SERIAL_SMBUS
, "smb", NULL
},
1095 { PCI_CLASS_SERIAL_IB
, "infiniband", NULL
},
1096 { PCI_CLASS_SERIAL_IPMI
, "ipmi", NULL
},
1097 { PCI_CLASS_SERIAL_SERCOS
, "sercos", NULL
},
1098 { PCI_CLASS_SERIAL_CANBUS
, "canbus", NULL
},
1099 { 0xFF, NULL
, NULL
},
1102 static const PCISubClass wrl_subclass
[] = {
1103 { PCI_CLASS_WIRELESS_IRDA
, "irda", NULL
},
1104 { PCI_CLASS_WIRELESS_CIR
, "consumer-ir", NULL
},
1105 { PCI_CLASS_WIRELESS_RF_CONTROLLER
, "rf-controller", NULL
},
1106 { PCI_CLASS_WIRELESS_BLUETOOTH
, "bluetooth", NULL
},
1107 { PCI_CLASS_WIRELESS_BROADBAND
, "broadband", NULL
},
1108 { 0xFF, NULL
, NULL
},
1111 static const PCISubClass sat_subclass
[] = {
1112 { PCI_CLASS_SATELLITE_TV
, "satellite-tv", NULL
},
1113 { PCI_CLASS_SATELLITE_AUDIO
, "satellite-audio", NULL
},
1114 { PCI_CLASS_SATELLITE_VOICE
, "satellite-voice", NULL
},
1115 { PCI_CLASS_SATELLITE_DATA
, "satellite-data", NULL
},
1116 { 0xFF, NULL
, NULL
},
1119 static const PCISubClass crypt_subclass
[] = {
1120 { PCI_CLASS_CRYPT_NETWORK
, "network-encryption", NULL
},
1121 { PCI_CLASS_CRYPT_ENTERTAINMENT
,
1122 "entertainment-encryption", NULL
},
1123 { 0xFF, NULL
, NULL
},
1126 static const PCISubClass spc_subclass
[] = {
1127 { PCI_CLASS_SP_DPIO
, "dpio", NULL
},
1128 { PCI_CLASS_SP_PERF
, "counter", NULL
},
1129 { PCI_CLASS_SP_SYNCH
, "measurement", NULL
},
1130 { PCI_CLASS_SP_MANAGEMENT
, "management-card", NULL
},
1131 { 0xFF, NULL
, NULL
},
1134 static const PCIClass pci_classes
[] = {
1135 { "legacy-device", undef_subclass
},
1136 { "mass-storage", mass_subclass
},
1137 { "network", net_subclass
},
1138 { "display", displ_subclass
, },
1139 { "multimedia-device", media_subclass
},
1140 { "memory-controller", mem_subclass
},
1141 { "unknown-bridge", bridg_subclass
},
1142 { "communication-controller", comm_subclass
},
1143 { "system-peripheral", sys_subclass
},
1144 { "input-controller", inp_subclass
},
1145 { "docking-station", dock_subclass
},
1146 { "cpu", cpu_subclass
},
1147 { "serial-bus", ser_subclass
},
1148 { "wireless-controller", wrl_subclass
},
1149 { "intelligent-io", NULL
},
1150 { "satellite-device", sat_subclass
},
1151 { "encryption", crypt_subclass
},
1152 { "data-processing-controller", spc_subclass
},
1155 static const char *dt_name_from_class(uint8_t class, uint8_t subclass
,
1158 const PCIClass
*pclass
;
1159 const PCISubClass
*psubclass
;
1160 const PCIIFace
*piface
;
1163 if (class >= ARRAY_SIZE(pci_classes
)) {
1167 pclass
= pci_classes
+ class;
1168 name
= pclass
->name
;
1170 if (pclass
->subc
== NULL
) {
1174 psubclass
= pclass
->subc
;
1175 while ((psubclass
->subclass
& 0xff) != 0xff) {
1176 if ((psubclass
->subclass
& 0xff) == subclass
) {
1177 name
= psubclass
->name
;
1183 piface
= psubclass
->iface
;
1184 if (piface
== NULL
) {
1187 while ((piface
->iface
& 0xff) != 0xff) {
1188 if ((piface
->iface
& 0xff) == iface
) {
1189 name
= piface
->name
;
1199 * DRC helper functions
1202 static uint32_t drc_id_from_devfn(SpaprPhbState
*phb
,
1203 uint8_t chassis
, int32_t devfn
)
1205 return (phb
->index
<< 16) | (chassis
<< 8) | devfn
;
1208 static SpaprDrc
*drc_from_devfn(SpaprPhbState
*phb
,
1209 uint8_t chassis
, int32_t devfn
)
1211 return spapr_drc_by_id(TYPE_SPAPR_DRC_PCI
,
1212 drc_id_from_devfn(phb
, chassis
, devfn
));
1215 static uint8_t chassis_from_bus(PCIBus
*bus
)
1217 if (pci_bus_is_root(bus
)) {
1220 PCIDevice
*bridge
= pci_bridge_get_device(bus
);
1222 return object_property_get_uint(OBJECT(bridge
), "chassis_nr",
1227 static SpaprDrc
*drc_from_dev(SpaprPhbState
*phb
, PCIDevice
*dev
)
1229 uint8_t chassis
= chassis_from_bus(pci_get_bus(dev
));
1231 return drc_from_devfn(phb
, chassis
, dev
->devfn
);
1234 static void add_drcs(SpaprPhbState
*phb
, PCIBus
*bus
)
1240 if (!phb
->dr_enabled
) {
1244 chassis
= chassis_from_bus(bus
);
1246 if (pci_bus_is_root(bus
)) {
1247 owner
= OBJECT(phb
);
1249 owner
= OBJECT(pci_bridge_get_device(bus
));
1252 for (i
= 0; i
< PCI_SLOT_MAX
* PCI_FUNC_MAX
; i
++) {
1253 spapr_dr_connector_new(owner
, TYPE_SPAPR_DRC_PCI
,
1254 drc_id_from_devfn(phb
, chassis
, i
));
1258 static void remove_drcs(SpaprPhbState
*phb
, PCIBus
*bus
)
1263 if (!phb
->dr_enabled
) {
1267 chassis
= chassis_from_bus(bus
);
1269 for (i
= PCI_SLOT_MAX
* PCI_FUNC_MAX
- 1; i
>= 0; i
--) {
1270 SpaprDrc
*drc
= drc_from_devfn(phb
, chassis
, i
);
1273 object_unparent(OBJECT(drc
));
1278 typedef struct PciWalkFdt
{
1281 SpaprPhbState
*sphb
;
1285 static int spapr_dt_pci_device(SpaprPhbState
*sphb
, PCIDevice
*dev
,
1286 void *fdt
, int parent_offset
);
1288 static void spapr_dt_pci_device_cb(PCIBus
*bus
, PCIDevice
*pdev
,
1291 PciWalkFdt
*p
= opaque
;
1295 /* Something's already broken, don't keep going */
1299 err
= spapr_dt_pci_device(p
->sphb
, pdev
, p
->fdt
, p
->offset
);
1305 /* Augment PCI device node with bridge specific information */
1306 static int spapr_dt_pci_bus(SpaprPhbState
*sphb
, PCIBus
*bus
,
1307 void *fdt
, int offset
)
1310 PciWalkFdt cbinfo
= {
1318 _FDT(fdt_setprop_cell(fdt
, offset
, "#address-cells",
1319 RESOURCE_CELLS_ADDRESS
));
1320 _FDT(fdt_setprop_cell(fdt
, offset
, "#size-cells",
1321 RESOURCE_CELLS_SIZE
));
1324 pci_for_each_device_under_bus_reverse(bus
, spapr_dt_pci_device_cb
, &cbinfo
);
1329 if (pci_bus_is_root(bus
)) {
1330 owner
= OBJECT(sphb
);
1332 owner
= OBJECT(pci_bridge_get_device(bus
));
1335 ret
= spapr_dt_drc(fdt
, offset
, owner
,
1336 SPAPR_DR_CONNECTOR_TYPE_PCI
);
1344 char *spapr_pci_fw_dev_name(PCIDevice
*dev
)
1346 const gchar
*basename
;
1347 int slot
= PCI_SLOT(dev
->devfn
);
1348 int func
= PCI_FUNC(dev
->devfn
);
1349 uint32_t ccode
= pci_default_read_config(dev
, PCI_CLASS_PROG
, 3);
1351 basename
= dt_name_from_class((ccode
>> 16) & 0xff, (ccode
>> 8) & 0xff,
1355 return g_strdup_printf("%s@%x,%x", basename
, slot
, func
);
1357 return g_strdup_printf("%s@%x", basename
, slot
);
1361 /* create OF node for pci device and required OF DT properties */
1362 static int spapr_dt_pci_device(SpaprPhbState
*sphb
, PCIDevice
*dev
,
1363 void *fdt
, int parent_offset
)
1366 g_autofree gchar
*nodename
= spapr_pci_fw_dev_name(dev
);
1368 SpaprDrc
*drc
= drc_from_dev(sphb
, dev
);
1369 uint32_t vendor_id
= pci_default_read_config(dev
, PCI_VENDOR_ID
, 2);
1370 uint32_t device_id
= pci_default_read_config(dev
, PCI_DEVICE_ID
, 2);
1371 uint32_t revision_id
= pci_default_read_config(dev
, PCI_REVISION_ID
, 1);
1372 uint32_t ccode
= pci_default_read_config(dev
, PCI_CLASS_PROG
, 3);
1373 uint32_t irq_pin
= pci_default_read_config(dev
, PCI_INTERRUPT_PIN
, 1);
1374 uint32_t subsystem_id
= pci_default_read_config(dev
, PCI_SUBSYSTEM_ID
, 2);
1375 uint32_t subsystem_vendor_id
=
1376 pci_default_read_config(dev
, PCI_SUBSYSTEM_VENDOR_ID
, 2);
1377 uint32_t cache_line_size
=
1378 pci_default_read_config(dev
, PCI_CACHE_LINE_SIZE
, 1);
1379 uint32_t pci_status
= pci_default_read_config(dev
, PCI_STATUS
, 2);
1382 _FDT(offset
= fdt_add_subnode(fdt
, parent_offset
, nodename
));
1384 /* in accordance with PAPR+ v2.7 13.6.3, Table 181 */
1385 _FDT(fdt_setprop_cell(fdt
, offset
, "vendor-id", vendor_id
));
1386 _FDT(fdt_setprop_cell(fdt
, offset
, "device-id", device_id
));
1387 _FDT(fdt_setprop_cell(fdt
, offset
, "revision-id", revision_id
));
1389 _FDT(fdt_setprop_cell(fdt
, offset
, "class-code", ccode
));
1391 _FDT(fdt_setprop_cell(fdt
, offset
, "interrupts", irq_pin
));
1395 _FDT(fdt_setprop_cell(fdt
, offset
, "subsystem-id", subsystem_id
));
1398 if (subsystem_vendor_id
) {
1399 _FDT(fdt_setprop_cell(fdt
, offset
, "subsystem-vendor-id",
1400 subsystem_vendor_id
));
1403 _FDT(fdt_setprop_cell(fdt
, offset
, "cache-line-size", cache_line_size
));
1406 /* the following fdt cells are masked off the pci status register */
1407 _FDT(fdt_setprop_cell(fdt
, offset
, "devsel-speed",
1408 PCI_STATUS_DEVSEL_MASK
& pci_status
));
1410 if (pci_status
& PCI_STATUS_FAST_BACK
) {
1411 _FDT(fdt_setprop(fdt
, offset
, "fast-back-to-back", NULL
, 0));
1413 if (pci_status
& PCI_STATUS_66MHZ
) {
1414 _FDT(fdt_setprop(fdt
, offset
, "66mhz-capable", NULL
, 0));
1416 if (pci_status
& PCI_STATUS_UDF
) {
1417 _FDT(fdt_setprop(fdt
, offset
, "udf-supported", NULL
, 0));
1420 loc_code
= spapr_phb_get_loc_code(sphb
, dev
);
1421 _FDT(fdt_setprop_string(fdt
, offset
, "ibm,loc-code", loc_code
));
1425 _FDT(fdt_setprop_cell(fdt
, offset
, "ibm,my-drc-index",
1426 spapr_drc_index(drc
)));
1429 if (msi_present(dev
)) {
1430 uint32_t max_msi
= msi_nr_vectors_allocated(dev
);
1432 _FDT(fdt_setprop_cell(fdt
, offset
, "ibm,req#msi", max_msi
));
1435 if (msix_present(dev
)) {
1436 uint32_t max_msix
= dev
->msix_entries_nr
;
1438 _FDT(fdt_setprop_cell(fdt
, offset
, "ibm,req#msi-x", max_msix
));
1442 populate_resource_props(dev
, &rp
);
1443 _FDT(fdt_setprop(fdt
, offset
, "reg", (uint8_t *)rp
.reg
, rp
.reg_len
));
1445 if (sphb
->pcie_ecs
&& pci_is_express(dev
)) {
1446 _FDT(fdt_setprop_cell(fdt
, offset
, "ibm,pci-config-space-type", 0x1));
1449 if (!IS_PCI_BRIDGE(dev
)) {
1450 /* Properties only for non-bridges */
1451 uint32_t min_grant
= pci_default_read_config(dev
, PCI_MIN_GNT
, 1);
1452 uint32_t max_latency
= pci_default_read_config(dev
, PCI_MAX_LAT
, 1);
1453 _FDT(fdt_setprop_cell(fdt
, offset
, "min-grant", min_grant
));
1454 _FDT(fdt_setprop_cell(fdt
, offset
, "max-latency", max_latency
));
1457 PCIBus
*sec_bus
= pci_bridge_get_sec_bus(PCI_BRIDGE(dev
));
1459 return spapr_dt_pci_bus(sphb
, sec_bus
, fdt
, offset
);
1463 /* Callback to be called during DRC release. */
1464 void spapr_phb_remove_pci_device_cb(DeviceState
*dev
)
1466 HotplugHandler
*hotplug_ctrl
= qdev_get_hotplug_handler(dev
);
1468 hotplug_handler_unplug(hotplug_ctrl
, dev
, &error_abort
);
1469 object_unparent(OBJECT(dev
));
1472 int spapr_pci_dt_populate(SpaprDrc
*drc
, SpaprMachineState
*spapr
,
1473 void *fdt
, int *fdt_start_offset
, Error
**errp
)
1475 HotplugHandler
*plug_handler
= qdev_get_hotplug_handler(drc
->dev
);
1476 SpaprPhbState
*sphb
= SPAPR_PCI_HOST_BRIDGE(plug_handler
);
1477 PCIDevice
*pdev
= PCI_DEVICE(drc
->dev
);
1479 *fdt_start_offset
= spapr_dt_pci_device(sphb
, pdev
, fdt
, 0);
1483 static void spapr_pci_bridge_plug(SpaprPhbState
*phb
,
1486 PCIBus
*bus
= pci_bridge_get_sec_bus(bridge
);
1491 /* Returns non-zero if the value of "chassis_nr" is already in use */
1492 static int check_chassis_nr(Object
*obj
, void *opaque
)
1494 int new_chassis_nr
=
1495 object_property_get_uint(opaque
, "chassis_nr", &error_abort
);
1497 object_property_get_uint(obj
, "chassis_nr", NULL
);
1499 if (!object_dynamic_cast(obj
, TYPE_PCI_BRIDGE
)) {
1503 /* Skip unsupported bridge types */
1509 if (obj
== opaque
) {
1513 return chassis_nr
== new_chassis_nr
;
1516 static bool bridge_has_valid_chassis_nr(Object
*bridge
, Error
**errp
)
1519 object_property_get_uint(bridge
, "chassis_nr", NULL
);
1522 * slotid_cap_init() already ensures that "chassis_nr" isn't null for
1523 * standard PCI bridges, so this really tells if "chassis_nr" is present
1527 error_setg(errp
, "PCI Bridge lacks a \"chassis_nr\" property");
1528 error_append_hint(errp
, "Try -device pci-bridge instead.\n");
1532 /* We want unique values for "chassis_nr" */
1533 if (object_child_foreach_recursive(object_get_root(), check_chassis_nr
,
1535 error_setg(errp
, "Bridge chassis %d already in use", chassis_nr
);
1542 static void spapr_pci_pre_plug(HotplugHandler
*plug_handler
,
1543 DeviceState
*plugged_dev
, Error
**errp
)
1545 SpaprPhbState
*phb
= SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler
));
1546 PCIDevice
*pdev
= PCI_DEVICE(plugged_dev
);
1547 SpaprDrc
*drc
= drc_from_dev(phb
, pdev
);
1548 PCIBus
*bus
= PCI_BUS(qdev_get_parent_bus(DEVICE(pdev
)));
1549 uint32_t slotnr
= PCI_SLOT(pdev
->devfn
);
1551 if (!phb
->dr_enabled
) {
1552 /* if this is a hotplug operation initiated by the user
1553 * we need to let them know it's not enabled
1555 if (plugged_dev
->hotplugged
) {
1556 error_setg(errp
, "Bus '%s' does not support hotplugging",
1557 phb
->parent_obj
.bus
->qbus
.name
);
1562 if (IS_PCI_BRIDGE(plugged_dev
)) {
1563 if (!bridge_has_valid_chassis_nr(OBJECT(plugged_dev
), errp
)) {
1568 /* Following the QEMU convention used for PCIe multifunction
1569 * hotplug, we do not allow functions to be hotplugged to a
1570 * slot that already has function 0 present
1572 if (plugged_dev
->hotplugged
&& bus
->devices
[PCI_DEVFN(slotnr
, 0)] &&
1573 PCI_FUNC(pdev
->devfn
) != 0) {
1574 error_setg(errp
, "PCI: slot %d function 0 already occupied by %s,"
1575 " additional functions can no longer be exposed to guest.",
1576 slotnr
, bus
->devices
[PCI_DEVFN(slotnr
, 0)]->name
);
1579 if (drc
&& drc
->dev
) {
1580 error_setg(errp
, "PCI: slot %d already occupied by %s", slotnr
,
1581 pci_get_function_0(PCI_DEVICE(drc
->dev
))->name
);
1586 static void spapr_pci_plug(HotplugHandler
*plug_handler
,
1587 DeviceState
*plugged_dev
, Error
**errp
)
1589 SpaprPhbState
*phb
= SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler
));
1590 PCIDevice
*pdev
= PCI_DEVICE(plugged_dev
);
1591 SpaprDrc
*drc
= drc_from_dev(phb
, pdev
);
1592 uint32_t slotnr
= PCI_SLOT(pdev
->devfn
);
1595 * If DR is disabled we don't need to do anything in the case of
1596 * hotplug or coldplug callbacks.
1598 if (!phb
->dr_enabled
) {
1604 if (IS_PCI_BRIDGE(plugged_dev
)) {
1605 spapr_pci_bridge_plug(phb
, PCI_BRIDGE(plugged_dev
));
1608 /* spapr_pci_pre_plug() already checked the DRC is attachable */
1609 spapr_drc_attach(drc
, DEVICE(pdev
));
1611 /* If this is function 0, signal hotplug for all the device functions.
1612 * Otherwise defer sending the hotplug event.
1614 if (!spapr_drc_hotplugged(plugged_dev
)) {
1615 spapr_drc_reset(drc
);
1616 } else if (PCI_FUNC(pdev
->devfn
) == 0) {
1618 uint8_t chassis
= chassis_from_bus(pci_get_bus(pdev
));
1620 for (i
= 0; i
< 8; i
++) {
1622 SpaprDrcClass
*func_drck
;
1623 SpaprDREntitySense state
;
1625 func_drc
= drc_from_devfn(phb
, chassis
, PCI_DEVFN(slotnr
, i
));
1626 func_drck
= SPAPR_DR_CONNECTOR_GET_CLASS(func_drc
);
1627 state
= func_drck
->dr_entity_sense(func_drc
);
1629 if (state
== SPAPR_DR_ENTITY_SENSE_PRESENT
) {
1630 spapr_hotplug_req_add_by_index(func_drc
);
1636 static void spapr_pci_bridge_unplug(SpaprPhbState
*phb
,
1639 PCIBus
*bus
= pci_bridge_get_sec_bus(bridge
);
1641 remove_drcs(phb
, bus
);
1644 static void spapr_pci_unplug(HotplugHandler
*plug_handler
,
1645 DeviceState
*plugged_dev
, Error
**errp
)
1647 SpaprPhbState
*phb
= SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler
));
1649 /* some version guests do not wait for completion of a device
1650 * cleanup (generally done asynchronously by the kernel) before
1651 * signaling to QEMU that the device is safe, but instead sleep
1652 * for some 'safe' period of time. unfortunately on a busy host
1653 * this sleep isn't guaranteed to be long enough, resulting in
1654 * bad things like IRQ lines being left asserted during final
1655 * device removal. to deal with this we call reset just prior
1656 * to finalizing the device, which will put the device back into
1657 * an 'idle' state, as the device cleanup code expects.
1659 pci_device_reset(PCI_DEVICE(plugged_dev
));
1661 if (IS_PCI_BRIDGE(plugged_dev
)) {
1662 spapr_pci_bridge_unplug(phb
, PCI_BRIDGE(plugged_dev
));
1666 qdev_unrealize(plugged_dev
);
1669 static void spapr_pci_unplug_request(HotplugHandler
*plug_handler
,
1670 DeviceState
*plugged_dev
, Error
**errp
)
1672 SpaprPhbState
*phb
= SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler
));
1673 PCIDevice
*pdev
= PCI_DEVICE(plugged_dev
);
1674 SpaprDrc
*drc
= drc_from_dev(phb
, pdev
);
1676 if (!phb
->dr_enabled
) {
1677 error_setg(errp
, "Bus '%s' does not support hotplugging",
1678 phb
->parent_obj
.bus
->qbus
.name
);
1683 g_assert(drc
->dev
== plugged_dev
);
1685 if (!spapr_drc_unplug_requested(drc
)) {
1686 uint32_t slotnr
= PCI_SLOT(pdev
->devfn
);
1688 SpaprDrcClass
*func_drck
;
1689 SpaprDREntitySense state
;
1691 uint8_t chassis
= chassis_from_bus(pci_get_bus(pdev
));
1693 if (IS_PCI_BRIDGE(plugged_dev
)) {
1694 error_setg(errp
, "PCI: Hot unplug of PCI bridges not supported");
1697 if (object_property_get_uint(OBJECT(pdev
), "nvlink2-tgt", NULL
)) {
1698 error_setg(errp
, "PCI: Cannot unplug NVLink2 devices");
1702 /* ensure any other present functions are pending unplug */
1703 if (PCI_FUNC(pdev
->devfn
) == 0) {
1704 for (i
= 1; i
< 8; i
++) {
1705 func_drc
= drc_from_devfn(phb
, chassis
, PCI_DEVFN(slotnr
, i
));
1706 func_drck
= SPAPR_DR_CONNECTOR_GET_CLASS(func_drc
);
1707 state
= func_drck
->dr_entity_sense(func_drc
);
1708 if (state
== SPAPR_DR_ENTITY_SENSE_PRESENT
1709 && !spapr_drc_unplug_requested(func_drc
)) {
1711 * Attempting to remove function 0 of a multifunction
1712 * device will will cascade into removing all child
1713 * functions, even if their unplug weren't requested
1716 spapr_drc_unplug_request(func_drc
);
1721 spapr_drc_unplug_request(drc
);
1723 /* if this isn't func 0, defer unplug event. otherwise signal removal
1724 * for all present functions
1726 if (PCI_FUNC(pdev
->devfn
) == 0) {
1727 for (i
= 7; i
>= 0; i
--) {
1728 func_drc
= drc_from_devfn(phb
, chassis
, PCI_DEVFN(slotnr
, i
));
1729 func_drck
= SPAPR_DR_CONNECTOR_GET_CLASS(func_drc
);
1730 state
= func_drck
->dr_entity_sense(func_drc
);
1731 if (state
== SPAPR_DR_ENTITY_SENSE_PRESENT
) {
1732 spapr_hotplug_req_remove_by_index(func_drc
);
1738 "PCI device unplug already in progress for device %s",
1743 static void spapr_phb_finalizefn(Object
*obj
)
1745 SpaprPhbState
*sphb
= SPAPR_PCI_HOST_BRIDGE(obj
);
1747 g_free(sphb
->dtbusname
);
1748 sphb
->dtbusname
= NULL
;
1751 static void spapr_phb_unrealize(DeviceState
*dev
)
1753 SpaprMachineState
*spapr
= SPAPR_MACHINE(qdev_get_machine());
1754 SysBusDevice
*s
= SYS_BUS_DEVICE(dev
);
1755 PCIHostState
*phb
= PCI_HOST_BRIDGE(s
);
1756 SpaprPhbState
*sphb
= SPAPR_PCI_HOST_BRIDGE(phb
);
1757 SpaprTceTable
*tcet
;
1759 const unsigned windows_supported
= spapr_phb_windows_supported(sphb
);
1762 g_hash_table_unref(sphb
->msi
);
1767 * Remove IO/MMIO subregions and aliases, rest should get cleaned
1768 * via PHB's unrealize->object_finalize
1770 for (i
= windows_supported
- 1; i
>= 0; i
--) {
1771 tcet
= spapr_tce_find_by_liobn(sphb
->dma_liobn
[i
]);
1773 memory_region_del_subregion(&sphb
->iommu_root
,
1774 spapr_tce_get_iommu(tcet
));
1778 remove_drcs(sphb
, phb
->bus
);
1780 for (i
= PCI_NUM_PINS
- 1; i
>= 0; i
--) {
1781 if (sphb
->lsi_table
[i
].irq
) {
1782 spapr_irq_free(spapr
, sphb
->lsi_table
[i
].irq
, 1);
1783 sphb
->lsi_table
[i
].irq
= 0;
1787 QLIST_REMOVE(sphb
, list
);
1789 memory_region_del_subregion(&sphb
->iommu_root
, &sphb
->msiwindow
);
1792 * An attached PCI device may have memory listeners, eg. VFIO PCI. We have
1793 * unmapped all sections. Remove the listeners now, before destroying the
1796 address_space_remove_listeners(&sphb
->iommu_as
);
1797 address_space_destroy(&sphb
->iommu_as
);
1799 qbus_set_hotplug_handler(BUS(phb
->bus
), NULL
);
1800 pci_unregister_root_bus(phb
->bus
);
1802 memory_region_del_subregion(get_system_memory(), &sphb
->iowindow
);
1803 if (sphb
->mem64_win_pciaddr
!= (hwaddr
)-1) {
1804 memory_region_del_subregion(get_system_memory(), &sphb
->mem64window
);
1806 memory_region_del_subregion(get_system_memory(), &sphb
->mem32window
);
1809 static void spapr_phb_destroy_msi(gpointer opaque
)
1811 SpaprMachineState
*spapr
= SPAPR_MACHINE(qdev_get_machine());
1812 SpaprMachineClass
*smc
= SPAPR_MACHINE_GET_CLASS(spapr
);
1813 SpaprPciMsi
*msi
= opaque
;
1815 if (!smc
->legacy_irq_allocation
) {
1816 spapr_irq_msi_free(spapr
, msi
->first_irq
, msi
->num
);
1818 spapr_irq_free(spapr
, msi
->first_irq
, msi
->num
);
1822 static void spapr_phb_realize(DeviceState
*dev
, Error
**errp
)
1825 /* We don't use SPAPR_MACHINE() in order to exit gracefully if the user
1826 * tries to add a sPAPR PHB to a non-pseries machine.
1828 SpaprMachineState
*spapr
=
1829 (SpaprMachineState
*) object_dynamic_cast(qdev_get_machine(),
1830 TYPE_SPAPR_MACHINE
);
1831 SpaprMachineClass
*smc
= spapr
? SPAPR_MACHINE_GET_CLASS(spapr
) : NULL
;
1832 SysBusDevice
*sbd
= SYS_BUS_DEVICE(dev
);
1833 SpaprPhbState
*sphb
= SPAPR_PCI_HOST_BRIDGE(sbd
);
1834 PCIHostState
*phb
= PCI_HOST_BRIDGE(sbd
);
1835 MachineState
*ms
= MACHINE(spapr
);
1839 uint64_t msi_window_size
= 4096;
1840 SpaprTceTable
*tcet
;
1841 const unsigned windows_supported
= spapr_phb_windows_supported(sphb
);
1844 error_setg(errp
, TYPE_SPAPR_PCI_HOST_BRIDGE
" needs a pseries machine");
1848 assert(sphb
->index
!= (uint32_t)-1); /* checked in spapr_phb_pre_plug() */
1850 if (sphb
->mem64_win_size
!= 0) {
1851 if (sphb
->mem_win_size
> SPAPR_PCI_MEM32_WIN_SIZE
) {
1852 error_setg(errp
, "32-bit memory window of size 0x%"HWADDR_PRIx
1853 " (max 2 GiB)", sphb
->mem_win_size
);
1857 /* 64-bit window defaults to identity mapping */
1858 sphb
->mem64_win_pciaddr
= sphb
->mem64_win_addr
;
1859 } else if (sphb
->mem_win_size
> SPAPR_PCI_MEM32_WIN_SIZE
) {
1861 * For compatibility with old configuration, if no 64-bit MMIO
1862 * window is specified, but the ordinary (32-bit) memory
1863 * window is specified as > 2GiB, we treat it as a 2GiB 32-bit
1864 * window, with a 64-bit MMIO window following on immediately
1867 sphb
->mem64_win_size
= sphb
->mem_win_size
- SPAPR_PCI_MEM32_WIN_SIZE
;
1868 sphb
->mem64_win_addr
= sphb
->mem_win_addr
+ SPAPR_PCI_MEM32_WIN_SIZE
;
1869 sphb
->mem64_win_pciaddr
=
1870 SPAPR_PCI_MEM_WIN_BUS_OFFSET
+ SPAPR_PCI_MEM32_WIN_SIZE
;
1871 sphb
->mem_win_size
= SPAPR_PCI_MEM32_WIN_SIZE
;
1874 if (spapr_pci_find_phb(spapr
, sphb
->buid
)) {
1877 error_setg(errp
, "PCI host bridges must have unique indexes");
1878 error_append_hint(errp
, "The following indexes are already in use:");
1879 QLIST_FOREACH(s
, &spapr
->phbs
, list
) {
1880 error_append_hint(errp
, " %d", s
->index
);
1882 error_append_hint(errp
, "\nTry another value for the index property\n");
1886 if (sphb
->numa_node
!= -1 &&
1887 (sphb
->numa_node
>= MAX_NODES
||
1888 !ms
->numa_state
->nodes
[sphb
->numa_node
].present
)) {
1889 error_setg(errp
, "Invalid NUMA node ID for PCI host bridge");
1893 sphb
->dtbusname
= g_strdup_printf("pci@%" PRIx64
, sphb
->buid
);
1895 /* Initialize memory regions */
1896 namebuf
= g_strdup_printf("%s.mmio", sphb
->dtbusname
);
1897 memory_region_init(&sphb
->memspace
, OBJECT(sphb
), namebuf
, UINT64_MAX
);
1900 namebuf
= g_strdup_printf("%s.mmio32-alias", sphb
->dtbusname
);
1901 memory_region_init_alias(&sphb
->mem32window
, OBJECT(sphb
),
1902 namebuf
, &sphb
->memspace
,
1903 SPAPR_PCI_MEM_WIN_BUS_OFFSET
, sphb
->mem_win_size
);
1905 memory_region_add_subregion(get_system_memory(), sphb
->mem_win_addr
,
1906 &sphb
->mem32window
);
1908 if (sphb
->mem64_win_size
!= 0) {
1909 namebuf
= g_strdup_printf("%s.mmio64-alias", sphb
->dtbusname
);
1910 memory_region_init_alias(&sphb
->mem64window
, OBJECT(sphb
),
1911 namebuf
, &sphb
->memspace
,
1912 sphb
->mem64_win_pciaddr
, sphb
->mem64_win_size
);
1915 memory_region_add_subregion(get_system_memory(),
1916 sphb
->mem64_win_addr
,
1917 &sphb
->mem64window
);
1920 /* Initialize IO regions */
1921 namebuf
= g_strdup_printf("%s.io", sphb
->dtbusname
);
1922 memory_region_init(&sphb
->iospace
, OBJECT(sphb
),
1923 namebuf
, SPAPR_PCI_IO_WIN_SIZE
);
1926 namebuf
= g_strdup_printf("%s.io-alias", sphb
->dtbusname
);
1927 memory_region_init_alias(&sphb
->iowindow
, OBJECT(sphb
), namebuf
,
1928 &sphb
->iospace
, 0, SPAPR_PCI_IO_WIN_SIZE
);
1930 memory_region_add_subregion(get_system_memory(), sphb
->io_win_addr
,
1933 bus
= pci_register_root_bus(dev
, NULL
,
1934 pci_spapr_set_irq
, pci_swizzle_map_irq_fn
, sphb
,
1935 &sphb
->memspace
, &sphb
->iospace
,
1936 PCI_DEVFN(0, 0), PCI_NUM_PINS
,
1940 * Despite resembling a vanilla PCI bus in most ways, the PAPR
1941 * para-virtualized PCI bus *does* permit PCI-E extended config
1944 if (sphb
->pcie_ecs
) {
1945 bus
->flags
|= PCI_BUS_EXTENDED_CONFIG_SPACE
;
1948 qbus_set_hotplug_handler(BUS(phb
->bus
), OBJECT(sphb
));
1951 * Initialize PHB address space.
1952 * By default there will be at least one subregion for default
1954 * Later the guest might want to create another DMA window
1955 * which will become another memory subregion.
1957 namebuf
= g_strdup_printf("%s.iommu-root", sphb
->dtbusname
);
1958 memory_region_init(&sphb
->iommu_root
, OBJECT(sphb
),
1959 namebuf
, UINT64_MAX
);
1961 address_space_init(&sphb
->iommu_as
, &sphb
->iommu_root
,
1965 * As MSI/MSIX interrupts trigger by writing at MSI/MSIX vectors,
1966 * we need to allocate some memory to catch those writes coming
1967 * from msi_notify()/msix_notify().
1968 * As MSIMessage:addr is going to be the same and MSIMessage:data
1969 * is going to be a VIRQ number, 4 bytes of the MSI MR will only
1972 * For KVM we want to ensure that this memory is a full page so that
1973 * our memory slot is of page size granularity.
1975 if (kvm_enabled()) {
1976 msi_window_size
= qemu_real_host_page_size();
1979 memory_region_init_io(&sphb
->msiwindow
, OBJECT(sphb
), &spapr_msi_ops
, spapr
,
1980 "msi", msi_window_size
);
1981 memory_region_add_subregion(&sphb
->iommu_root
, SPAPR_PCI_MSI_WINDOW
,
1984 pci_setup_iommu(bus
, &spapr_iommu_ops
, sphb
);
1986 pci_bus_set_route_irq_fn(bus
, spapr_route_intx_pin_to_irq
);
1988 QLIST_INSERT_HEAD(&spapr
->phbs
, sphb
, list
);
1990 /* Initialize the LSI table */
1991 for (i
= 0; i
< PCI_NUM_PINS
; i
++) {
1992 int irq
= SPAPR_IRQ_PCI_LSI
+ sphb
->index
* PCI_NUM_PINS
+ i
;
1994 if (smc
->legacy_irq_allocation
) {
1995 irq
= spapr_irq_findone(spapr
, errp
);
1997 error_prepend(errp
, "can't allocate LSIs: ");
1999 * Older machines will never support PHB hotplug, ie, this is an
2000 * init only path and QEMU will terminate. No need to rollback.
2006 if (spapr_irq_claim(spapr
, irq
, true, errp
) < 0) {
2007 error_prepend(errp
, "can't allocate LSIs: ");
2011 sphb
->lsi_table
[i
].irq
= irq
;
2014 /* allocate connectors for child PCI devices */
2015 add_drcs(sphb
, phb
->bus
);
2018 for (i
= 0; i
< windows_supported
; ++i
) {
2019 tcet
= spapr_tce_new_table(DEVICE(sphb
), sphb
->dma_liobn
[i
]);
2021 error_setg(errp
, "Creating window#%d failed for %s",
2022 i
, sphb
->dtbusname
);
2025 memory_region_add_subregion(&sphb
->iommu_root
, 0,
2026 spapr_tce_get_iommu(tcet
));
2029 sphb
->msi
= g_hash_table_new_full(g_int_hash
, g_int_equal
, g_free
,
2030 spapr_phb_destroy_msi
);
2034 spapr_phb_unrealize(dev
);
2037 static int spapr_phb_children_reset(Object
*child
, void *opaque
)
2039 DeviceState
*dev
= (DeviceState
*) object_dynamic_cast(child
, TYPE_DEVICE
);
2042 device_cold_reset(dev
);
2048 void spapr_phb_dma_reset(SpaprPhbState
*sphb
)
2051 SpaprTceTable
*tcet
;
2053 for (i
= 0; i
< SPAPR_PCI_DMA_MAX_WINDOWS
; ++i
) {
2054 tcet
= spapr_tce_find_by_liobn(sphb
->dma_liobn
[i
]);
2056 if (tcet
&& tcet
->nb_table
) {
2057 spapr_tce_table_disable(tcet
);
2061 /* Register default 32bit DMA window */
2062 tcet
= spapr_tce_find_by_liobn(sphb
->dma_liobn
[0]);
2063 spapr_tce_table_enable(tcet
, SPAPR_TCE_PAGE_SHIFT
, sphb
->dma_win_addr
,
2064 sphb
->dma_win_size
>> SPAPR_TCE_PAGE_SHIFT
);
2065 tcet
->def_win
= true;
2068 static void spapr_phb_reset(DeviceState
*qdev
)
2070 SpaprPhbState
*sphb
= SPAPR_PCI_HOST_BRIDGE(qdev
);
2072 spapr_phb_dma_reset(sphb
);
2074 /* Reset the IOMMU state */
2075 object_child_foreach(OBJECT(qdev
), spapr_phb_children_reset
, NULL
);
2077 if (spapr_phb_eeh_available(SPAPR_PCI_HOST_BRIDGE(qdev
))) {
2078 spapr_phb_vfio_reset(qdev
);
2081 g_hash_table_remove_all(sphb
->msi
);
2084 static Property spapr_phb_properties
[] = {
2085 DEFINE_PROP_UINT32("index", SpaprPhbState
, index
, -1),
2086 DEFINE_PROP_UINT64("mem_win_size", SpaprPhbState
, mem_win_size
,
2087 SPAPR_PCI_MEM32_WIN_SIZE
),
2088 DEFINE_PROP_UINT64("mem64_win_size", SpaprPhbState
, mem64_win_size
,
2089 SPAPR_PCI_MEM64_WIN_SIZE
),
2090 DEFINE_PROP_UINT64("io_win_size", SpaprPhbState
, io_win_size
,
2091 SPAPR_PCI_IO_WIN_SIZE
),
2092 DEFINE_PROP_BOOL("dynamic-reconfiguration", SpaprPhbState
, dr_enabled
,
2094 /* Default DMA window is 0..1GB */
2095 DEFINE_PROP_UINT64("dma_win_addr", SpaprPhbState
, dma_win_addr
, 0),
2096 DEFINE_PROP_UINT64("dma_win_size", SpaprPhbState
, dma_win_size
, 0x40000000),
2097 DEFINE_PROP_UINT64("dma64_win_addr", SpaprPhbState
, dma64_win_addr
,
2098 0x800000000000000ULL
),
2099 DEFINE_PROP_BOOL("ddw", SpaprPhbState
, ddw_enabled
, true),
2100 DEFINE_PROP_UINT64("pgsz", SpaprPhbState
, page_size_mask
,
2101 (1ULL << 12) | (1ULL << 16)
2102 | (1ULL << 21) | (1ULL << 24)),
2103 DEFINE_PROP_UINT32("numa_node", SpaprPhbState
, numa_node
, -1),
2104 DEFINE_PROP_BOOL("pre-2.8-migration", SpaprPhbState
,
2105 pre_2_8_migration
, false),
2106 DEFINE_PROP_BOOL("pcie-extended-configuration-space", SpaprPhbState
,
2108 DEFINE_PROP_BOOL("pre-5.1-associativity", SpaprPhbState
,
2109 pre_5_1_assoc
, false),
2110 DEFINE_PROP_END_OF_LIST(),
2113 static const VMStateDescription vmstate_spapr_pci_lsi
= {
2114 .name
= "spapr_pci/lsi",
2116 .minimum_version_id
= 1,
2117 .fields
= (const VMStateField
[]) {
2118 VMSTATE_UINT32_EQUAL(irq
, SpaprPciLsi
, NULL
),
2120 VMSTATE_END_OF_LIST()
2124 static const VMStateDescription vmstate_spapr_pci_msi
= {
2125 .name
= "spapr_pci/msi",
2127 .minimum_version_id
= 1,
2128 .fields
= (const VMStateField
[]) {
2129 VMSTATE_UINT32(key
, SpaprPciMsiMig
),
2130 VMSTATE_UINT32(value
.first_irq
, SpaprPciMsiMig
),
2131 VMSTATE_UINT32(value
.num
, SpaprPciMsiMig
),
2132 VMSTATE_END_OF_LIST()
2136 static int spapr_pci_pre_save(void *opaque
)
2138 SpaprPhbState
*sphb
= opaque
;
2139 GHashTableIter iter
;
2140 gpointer key
, value
;
2143 if (sphb
->pre_2_8_migration
) {
2144 sphb
->mig_liobn
= sphb
->dma_liobn
[0];
2145 sphb
->mig_mem_win_addr
= sphb
->mem_win_addr
;
2146 sphb
->mig_mem_win_size
= sphb
->mem_win_size
;
2147 sphb
->mig_io_win_addr
= sphb
->io_win_addr
;
2148 sphb
->mig_io_win_size
= sphb
->io_win_size
;
2150 if ((sphb
->mem64_win_size
!= 0)
2151 && (sphb
->mem64_win_addr
2152 == (sphb
->mem_win_addr
+ sphb
->mem_win_size
))) {
2153 sphb
->mig_mem_win_size
+= sphb
->mem64_win_size
;
2157 g_free(sphb
->msi_devs
);
2158 sphb
->msi_devs
= NULL
;
2159 sphb
->msi_devs_num
= g_hash_table_size(sphb
->msi
);
2160 if (!sphb
->msi_devs_num
) {
2163 sphb
->msi_devs
= g_new(SpaprPciMsiMig
, sphb
->msi_devs_num
);
2165 g_hash_table_iter_init(&iter
, sphb
->msi
);
2166 for (i
= 0; g_hash_table_iter_next(&iter
, &key
, &value
); ++i
) {
2167 sphb
->msi_devs
[i
].key
= *(uint32_t *) key
;
2168 sphb
->msi_devs
[i
].value
= *(SpaprPciMsi
*) value
;
2174 static int spapr_pci_post_save(void *opaque
)
2176 SpaprPhbState
*sphb
= opaque
;
2178 g_free(sphb
->msi_devs
);
2179 sphb
->msi_devs
= NULL
;
2180 sphb
->msi_devs_num
= 0;
2184 static int spapr_pci_post_load(void *opaque
, int version_id
)
2186 SpaprPhbState
*sphb
= opaque
;
2187 gpointer key
, value
;
2190 for (i
= 0; i
< sphb
->msi_devs_num
; ++i
) {
2191 key
= g_memdup(&sphb
->msi_devs
[i
].key
,
2192 sizeof(sphb
->msi_devs
[i
].key
));
2193 value
= g_memdup(&sphb
->msi_devs
[i
].value
,
2194 sizeof(sphb
->msi_devs
[i
].value
));
2195 g_hash_table_insert(sphb
->msi
, key
, value
);
2197 g_free(sphb
->msi_devs
);
2198 sphb
->msi_devs
= NULL
;
2199 sphb
->msi_devs_num
= 0;
2204 static bool pre_2_8_migration(void *opaque
, int version_id
)
2206 SpaprPhbState
*sphb
= opaque
;
2208 return sphb
->pre_2_8_migration
;
2211 static const VMStateDescription vmstate_spapr_pci
= {
2212 .name
= "spapr_pci",
2214 .minimum_version_id
= 2,
2215 .pre_save
= spapr_pci_pre_save
,
2216 .post_save
= spapr_pci_post_save
,
2217 .post_load
= spapr_pci_post_load
,
2218 .fields
= (const VMStateField
[]) {
2219 VMSTATE_UINT64_EQUAL(buid
, SpaprPhbState
, NULL
),
2220 VMSTATE_UINT32_TEST(mig_liobn
, SpaprPhbState
, pre_2_8_migration
),
2221 VMSTATE_UINT64_TEST(mig_mem_win_addr
, SpaprPhbState
, pre_2_8_migration
),
2222 VMSTATE_UINT64_TEST(mig_mem_win_size
, SpaprPhbState
, pre_2_8_migration
),
2223 VMSTATE_UINT64_TEST(mig_io_win_addr
, SpaprPhbState
, pre_2_8_migration
),
2224 VMSTATE_UINT64_TEST(mig_io_win_size
, SpaprPhbState
, pre_2_8_migration
),
2225 VMSTATE_STRUCT_ARRAY(lsi_table
, SpaprPhbState
, PCI_NUM_PINS
, 0,
2226 vmstate_spapr_pci_lsi
, SpaprPciLsi
),
2227 VMSTATE_INT32(msi_devs_num
, SpaprPhbState
),
2228 VMSTATE_STRUCT_VARRAY_ALLOC(msi_devs
, SpaprPhbState
, msi_devs_num
, 0,
2229 vmstate_spapr_pci_msi
, SpaprPciMsiMig
),
2230 VMSTATE_END_OF_LIST()
2234 static const char *spapr_phb_root_bus_path(PCIHostState
*host_bridge
,
2237 SpaprPhbState
*sphb
= SPAPR_PCI_HOST_BRIDGE(host_bridge
);
2239 return sphb
->dtbusname
;
2242 static void spapr_phb_class_init(ObjectClass
*klass
, void *data
)
2244 PCIHostBridgeClass
*hc
= PCI_HOST_BRIDGE_CLASS(klass
);
2245 DeviceClass
*dc
= DEVICE_CLASS(klass
);
2246 HotplugHandlerClass
*hp
= HOTPLUG_HANDLER_CLASS(klass
);
2248 hc
->root_bus_path
= spapr_phb_root_bus_path
;
2249 dc
->realize
= spapr_phb_realize
;
2250 dc
->unrealize
= spapr_phb_unrealize
;
2251 device_class_set_props(dc
, spapr_phb_properties
);
2252 dc
->reset
= spapr_phb_reset
;
2253 dc
->vmsd
= &vmstate_spapr_pci
;
2254 /* Supported by TYPE_SPAPR_MACHINE */
2255 dc
->user_creatable
= true;
2256 set_bit(DEVICE_CATEGORY_BRIDGE
, dc
->categories
);
2257 hp
->pre_plug
= spapr_pci_pre_plug
;
2258 hp
->plug
= spapr_pci_plug
;
2259 hp
->unplug
= spapr_pci_unplug
;
2260 hp
->unplug_request
= spapr_pci_unplug_request
;
2263 static const TypeInfo spapr_phb_info
= {
2264 .name
= TYPE_SPAPR_PCI_HOST_BRIDGE
,
2265 .parent
= TYPE_PCI_HOST_BRIDGE
,
2266 .instance_size
= sizeof(SpaprPhbState
),
2267 .instance_finalize
= spapr_phb_finalizefn
,
2268 .class_init
= spapr_phb_class_init
,
2269 .interfaces
= (InterfaceInfo
[]) {
2270 { TYPE_HOTPLUG_HANDLER
},
2275 static void spapr_phb_pci_enumerate_bridge(PCIBus
*bus
, PCIDevice
*pdev
,
2278 unsigned int *bus_no
= opaque
;
2279 PCIBus
*sec_bus
= NULL
;
2281 if ((pci_default_read_config(pdev
, PCI_HEADER_TYPE
, 1) !=
2282 PCI_HEADER_TYPE_BRIDGE
)) {
2287 pci_default_write_config(pdev
, PCI_PRIMARY_BUS
, pci_dev_bus_num(pdev
), 1);
2288 pci_default_write_config(pdev
, PCI_SECONDARY_BUS
, *bus_no
, 1);
2289 pci_default_write_config(pdev
, PCI_SUBORDINATE_BUS
, *bus_no
, 1);
2291 sec_bus
= pci_bridge_get_sec_bus(PCI_BRIDGE(pdev
));
2296 pci_for_each_device_under_bus(sec_bus
, spapr_phb_pci_enumerate_bridge
,
2298 pci_default_write_config(pdev
, PCI_SUBORDINATE_BUS
, *bus_no
, 1);
2301 static void spapr_phb_pci_enumerate(SpaprPhbState
*phb
)
2303 PCIBus
*bus
= PCI_HOST_BRIDGE(phb
)->bus
;
2304 unsigned int bus_no
= 0;
2306 pci_for_each_device_under_bus(bus
, spapr_phb_pci_enumerate_bridge
,
2311 int spapr_dt_phb(SpaprMachineState
*spapr
, SpaprPhbState
*phb
,
2312 uint32_t intc_phandle
, void *fdt
, int *node_offset
)
2314 int bus_off
, i
, j
, ret
;
2315 uint32_t bus_range
[] = { cpu_to_be32(0), cpu_to_be32(0xff) };
2321 } QEMU_PACKED ranges
[] = {
2323 cpu_to_be32(b_ss(1)), cpu_to_be64(0),
2324 cpu_to_be64(phb
->io_win_addr
),
2325 cpu_to_be64(memory_region_size(&phb
->iospace
)),
2328 cpu_to_be32(b_ss(2)), cpu_to_be64(SPAPR_PCI_MEM_WIN_BUS_OFFSET
),
2329 cpu_to_be64(phb
->mem_win_addr
),
2330 cpu_to_be64(phb
->mem_win_size
),
2333 cpu_to_be32(b_ss(3)), cpu_to_be64(phb
->mem64_win_pciaddr
),
2334 cpu_to_be64(phb
->mem64_win_addr
),
2335 cpu_to_be64(phb
->mem64_win_size
),
2338 const unsigned sizeof_ranges
=
2339 (phb
->mem64_win_size
? 3 : 2) * sizeof(ranges
[0]);
2340 uint64_t bus_reg
[] = { cpu_to_be64(phb
->buid
), 0 };
2341 uint32_t interrupt_map_mask
[] = {
2342 cpu_to_be32(b_ddddd(-1)|b_fff(0)), 0x0, 0x0, cpu_to_be32(-1)};
2343 uint32_t interrupt_map
[PCI_SLOT_MAX
* PCI_NUM_PINS
][7];
2344 uint32_t ddw_applicable
[] = {
2345 cpu_to_be32(RTAS_IBM_QUERY_PE_DMA_WINDOW
),
2346 cpu_to_be32(RTAS_IBM_CREATE_PE_DMA_WINDOW
),
2347 cpu_to_be32(RTAS_IBM_REMOVE_PE_DMA_WINDOW
)
2349 uint32_t ddw_extensions
[] = {
2351 cpu_to_be32(RTAS_IBM_RESET_PE_DMA_WINDOW
),
2352 cpu_to_be32(1), /* 1: ibm,query-pe-dma-window 6 outputs, PAPR 2.8 */
2354 SpaprTceTable
*tcet
;
2357 /* Start populating the FDT */
2358 _FDT(bus_off
= fdt_add_subnode(fdt
, 0, phb
->dtbusname
));
2360 *node_offset
= bus_off
;
2363 /* Write PHB properties */
2364 _FDT(fdt_setprop_string(fdt
, bus_off
, "device_type", "pci"));
2365 _FDT(fdt_setprop_string(fdt
, bus_off
, "compatible", "IBM,Logical_PHB"));
2366 _FDT(fdt_setprop_cell(fdt
, bus_off
, "#interrupt-cells", 0x1));
2367 _FDT(fdt_setprop(fdt
, bus_off
, "used-by-rtas", NULL
, 0));
2368 _FDT(fdt_setprop(fdt
, bus_off
, "bus-range", &bus_range
, sizeof(bus_range
)));
2369 _FDT(fdt_setprop(fdt
, bus_off
, "ranges", &ranges
, sizeof_ranges
));
2370 _FDT(fdt_setprop(fdt
, bus_off
, "reg", &bus_reg
, sizeof(bus_reg
)));
2371 _FDT(fdt_setprop_cell(fdt
, bus_off
, "ibm,pci-config-space-type", 0x1));
2372 _FDT(fdt_setprop_cell(fdt
, bus_off
, "ibm,pe-total-#msi",
2373 spapr_irq_nr_msis(spapr
)));
2375 /* Dynamic DMA window */
2376 if (phb
->ddw_enabled
) {
2377 _FDT(fdt_setprop(fdt
, bus_off
, "ibm,ddw-applicable", &ddw_applicable
,
2378 sizeof(ddw_applicable
)));
2379 _FDT(fdt_setprop(fdt
, bus_off
, "ibm,ddw-extensions",
2380 &ddw_extensions
, sizeof(ddw_extensions
)));
2383 /* Advertise NUMA via ibm,associativity */
2384 if (phb
->numa_node
!= -1) {
2385 spapr_numa_write_associativity_dt(spapr
, fdt
, bus_off
, phb
->numa_node
);
2388 /* Build the interrupt-map, this must matches what is done
2389 * in pci_swizzle_map_irq_fn
2391 _FDT(fdt_setprop(fdt
, bus_off
, "interrupt-map-mask",
2392 &interrupt_map_mask
, sizeof(interrupt_map_mask
)));
2393 for (i
= 0; i
< PCI_SLOT_MAX
; i
++) {
2394 for (j
= 0; j
< PCI_NUM_PINS
; j
++) {
2395 uint32_t *irqmap
= interrupt_map
[i
*PCI_NUM_PINS
+ j
];
2396 int lsi_num
= pci_swizzle(i
, j
);
2398 irqmap
[0] = cpu_to_be32(b_ddddd(i
)|b_fff(0));
2401 irqmap
[3] = cpu_to_be32(j
+1);
2402 irqmap
[4] = cpu_to_be32(intc_phandle
);
2403 spapr_dt_irq(&irqmap
[5], phb
->lsi_table
[lsi_num
].irq
, true);
2406 /* Write interrupt map */
2407 _FDT(fdt_setprop(fdt
, bus_off
, "interrupt-map", &interrupt_map
,
2408 sizeof(interrupt_map
)));
2410 tcet
= spapr_tce_find_by_liobn(phb
->dma_liobn
[0]);
2414 spapr_dma_dt(fdt
, bus_off
, "ibm,dma-window",
2415 tcet
->liobn
, tcet
->bus_offset
,
2416 tcet
->nb_table
<< tcet
->page_shift
);
2418 drc
= spapr_drc_by_id(TYPE_SPAPR_DRC_PHB
, phb
->index
);
2420 uint32_t drc_index
= cpu_to_be32(spapr_drc_index(drc
));
2422 _FDT(fdt_setprop(fdt
, bus_off
, "ibm,my-drc-index", &drc_index
,
2423 sizeof(drc_index
)));
2426 /* Walk the bridges and program the bus numbers*/
2427 spapr_phb_pci_enumerate(phb
);
2428 _FDT(fdt_setprop_cell(fdt
, bus_off
, "qemu,phb-enumerated", 0x1));
2430 /* Walk the bridge and subordinate buses */
2431 ret
= spapr_dt_pci_bus(phb
, PCI_HOST_BRIDGE(phb
)->bus
, fdt
, bus_off
);
2439 void spapr_pci_rtas_init(void)
2441 spapr_rtas_register(RTAS_READ_PCI_CONFIG
, "read-pci-config",
2442 rtas_read_pci_config
);
2443 spapr_rtas_register(RTAS_WRITE_PCI_CONFIG
, "write-pci-config",
2444 rtas_write_pci_config
);
2445 spapr_rtas_register(RTAS_IBM_READ_PCI_CONFIG
, "ibm,read-pci-config",
2446 rtas_ibm_read_pci_config
);
2447 spapr_rtas_register(RTAS_IBM_WRITE_PCI_CONFIG
, "ibm,write-pci-config",
2448 rtas_ibm_write_pci_config
);
2449 if (msi_nonbroken
) {
2450 spapr_rtas_register(RTAS_IBM_QUERY_INTERRUPT_SOURCE_NUMBER
,
2451 "ibm,query-interrupt-source-number",
2452 rtas_ibm_query_interrupt_source_number
);
2453 spapr_rtas_register(RTAS_IBM_CHANGE_MSI
, "ibm,change-msi",
2454 rtas_ibm_change_msi
);
2457 spapr_rtas_register(RTAS_IBM_SET_EEH_OPTION
,
2458 "ibm,set-eeh-option",
2459 rtas_ibm_set_eeh_option
);
2460 spapr_rtas_register(RTAS_IBM_GET_CONFIG_ADDR_INFO2
,
2461 "ibm,get-config-addr-info2",
2462 rtas_ibm_get_config_addr_info2
);
2463 spapr_rtas_register(RTAS_IBM_READ_SLOT_RESET_STATE2
,
2464 "ibm,read-slot-reset-state2",
2465 rtas_ibm_read_slot_reset_state2
);
2466 spapr_rtas_register(RTAS_IBM_SET_SLOT_RESET
,
2467 "ibm,set-slot-reset",
2468 rtas_ibm_set_slot_reset
);
2469 spapr_rtas_register(RTAS_IBM_CONFIGURE_PE
,
2471 rtas_ibm_configure_pe
);
2472 spapr_rtas_register(RTAS_IBM_SLOT_ERROR_DETAIL
,
2473 "ibm,slot-error-detail",
2474 rtas_ibm_slot_error_detail
);
2477 static void spapr_pci_register_types(void)
2479 type_register_static(&spapr_phb_info
);
2482 type_init(spapr_pci_register_types
)
2484 static int spapr_switch_one_vga(DeviceState
*dev
, void *opaque
)
2486 bool be
= *(bool *)opaque
;
2488 if (object_dynamic_cast(OBJECT(dev
), "VGA")
2489 || object_dynamic_cast(OBJECT(dev
), "secondary-vga")
2490 || object_dynamic_cast(OBJECT(dev
), "bochs-display")
2491 || object_dynamic_cast(OBJECT(dev
), "virtio-vga")) {
2492 object_property_set_bool(OBJECT(dev
), "big-endian-framebuffer", be
,
2498 void spapr_pci_switch_vga(SpaprMachineState
*spapr
, bool big_endian
)
2500 SpaprPhbState
*sphb
;
2503 * For backward compatibility with existing guests, we switch
2504 * the endianness of the VGA controller when changing the guest
2507 QLIST_FOREACH(sphb
, &spapr
->phbs
, list
) {
2508 BusState
*bus
= &PCI_HOST_BRIDGE(sphb
)->bus
->qbus
;
2509 qbus_walk_children(bus
, spapr_switch_one_vga
, NULL
, NULL
, NULL
,