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
25 #include "qemu/osdep.h"
26 #include "qapi/error.h"
27 #include "qemu-common.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"
41 #include "qemu/error-report.h"
42 #include "qapi/qmp/qerror.h"
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"
53 #include "hw/vfio/vfio.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 #define FDT_NAME_MAX 128
76 sPAPRPHBState
*spapr_pci_find_phb(sPAPRMachineState
*spapr
, uint64_t buid
)
80 QLIST_FOREACH(sphb
, &spapr
->phbs
, list
) {
81 if (sphb
->buid
!= buid
) {
90 PCIDevice
*spapr_pci_find_dev(sPAPRMachineState
*spapr
, uint64_t buid
,
93 sPAPRPHBState
*sphb
= spapr_pci_find_phb(spapr
, buid
);
94 PCIHostState
*phb
= PCI_HOST_BRIDGE(sphb
);
95 int bus_num
= (config_addr
>> 16) & 0xFF;
96 int devfn
= (config_addr
>> 8) & 0xFF;
102 return pci_find_device(phb
->bus
, bus_num
, devfn
);
105 static uint32_t rtas_pci_cfgaddr(uint32_t arg
)
107 /* This handles the encoding of extended config space addresses */
108 return ((arg
>> 20) & 0xf00) | (arg
& 0xff);
111 static void finish_read_pci_config(sPAPRMachineState
*spapr
, uint64_t buid
,
112 uint32_t addr
, uint32_t size
,
118 if ((size
!= 1) && (size
!= 2) && (size
!= 4)) {
119 /* access must be 1, 2 or 4 bytes */
120 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
124 pci_dev
= spapr_pci_find_dev(spapr
, buid
, addr
);
125 addr
= rtas_pci_cfgaddr(addr
);
127 if (!pci_dev
|| (addr
% size
) || (addr
>= pci_config_size(pci_dev
))) {
128 /* Access must be to a valid device, within bounds and
129 * naturally aligned */
130 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
134 val
= pci_host_config_read_common(pci_dev
, addr
,
135 pci_config_size(pci_dev
), size
);
137 rtas_st(rets
, 0, RTAS_OUT_SUCCESS
);
138 rtas_st(rets
, 1, val
);
141 static void rtas_ibm_read_pci_config(PowerPCCPU
*cpu
, sPAPRMachineState
*spapr
,
142 uint32_t token
, uint32_t nargs
,
144 uint32_t nret
, target_ulong rets
)
149 if ((nargs
!= 4) || (nret
!= 2)) {
150 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
154 buid
= rtas_ldq(args
, 1);
155 size
= rtas_ld(args
, 3);
156 addr
= rtas_ld(args
, 0);
158 finish_read_pci_config(spapr
, buid
, addr
, size
, rets
);
161 static void rtas_read_pci_config(PowerPCCPU
*cpu
, sPAPRMachineState
*spapr
,
162 uint32_t token
, uint32_t nargs
,
164 uint32_t nret
, target_ulong rets
)
168 if ((nargs
!= 2) || (nret
!= 2)) {
169 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
173 size
= rtas_ld(args
, 1);
174 addr
= rtas_ld(args
, 0);
176 finish_read_pci_config(spapr
, 0, addr
, size
, rets
);
179 static void finish_write_pci_config(sPAPRMachineState
*spapr
, uint64_t buid
,
180 uint32_t addr
, uint32_t size
,
181 uint32_t val
, target_ulong rets
)
185 if ((size
!= 1) && (size
!= 2) && (size
!= 4)) {
186 /* access must be 1, 2 or 4 bytes */
187 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
191 pci_dev
= spapr_pci_find_dev(spapr
, buid
, addr
);
192 addr
= rtas_pci_cfgaddr(addr
);
194 if (!pci_dev
|| (addr
% size
) || (addr
>= pci_config_size(pci_dev
))) {
195 /* Access must be to a valid device, within bounds and
196 * naturally aligned */
197 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
201 pci_host_config_write_common(pci_dev
, addr
, pci_config_size(pci_dev
),
204 rtas_st(rets
, 0, RTAS_OUT_SUCCESS
);
207 static void rtas_ibm_write_pci_config(PowerPCCPU
*cpu
, sPAPRMachineState
*spapr
,
208 uint32_t token
, uint32_t nargs
,
210 uint32_t nret
, target_ulong rets
)
213 uint32_t val
, size
, addr
;
215 if ((nargs
!= 5) || (nret
!= 1)) {
216 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
220 buid
= rtas_ldq(args
, 1);
221 val
= rtas_ld(args
, 4);
222 size
= rtas_ld(args
, 3);
223 addr
= rtas_ld(args
, 0);
225 finish_write_pci_config(spapr
, buid
, addr
, size
, val
, rets
);
228 static void rtas_write_pci_config(PowerPCCPU
*cpu
, sPAPRMachineState
*spapr
,
229 uint32_t token
, uint32_t nargs
,
231 uint32_t nret
, target_ulong rets
)
233 uint32_t val
, size
, addr
;
235 if ((nargs
!= 3) || (nret
!= 1)) {
236 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
241 val
= rtas_ld(args
, 2);
242 size
= rtas_ld(args
, 1);
243 addr
= rtas_ld(args
, 0);
245 finish_write_pci_config(spapr
, 0, addr
, size
, val
, rets
);
249 * Set MSI/MSIX message data.
250 * This is required for msi_notify()/msix_notify() which
251 * will write at the addresses via spapr_msi_write().
253 * If hwaddr == 0, all entries will have .data == first_irq i.e.
254 * table will be reset.
256 static void spapr_msi_setmsg(PCIDevice
*pdev
, hwaddr addr
, bool msix
,
257 unsigned first_irq
, unsigned req_num
)
260 MSIMessage msg
= { .address
= addr
, .data
= first_irq
};
263 msi_set_message(pdev
, msg
);
264 trace_spapr_pci_msi_setup(pdev
->name
, 0, msg
.address
);
268 for (i
= 0; i
< req_num
; ++i
) {
269 msix_set_message(pdev
, i
, msg
);
270 trace_spapr_pci_msi_setup(pdev
->name
, i
, msg
.address
);
277 static void rtas_ibm_change_msi(PowerPCCPU
*cpu
, sPAPRMachineState
*spapr
,
278 uint32_t token
, uint32_t nargs
,
279 target_ulong args
, uint32_t nret
,
282 uint32_t config_addr
= rtas_ld(args
, 0);
283 uint64_t buid
= rtas_ldq(args
, 1);
284 unsigned int func
= rtas_ld(args
, 3);
285 unsigned int req_num
= rtas_ld(args
, 4); /* 0 == remove all */
286 unsigned int seq_num
= rtas_ld(args
, 5);
287 unsigned int ret_intr_type
;
288 unsigned int irq
, max_irqs
= 0;
289 sPAPRPHBState
*phb
= NULL
;
290 PCIDevice
*pdev
= NULL
;
292 int *config_addr_key
;
296 case RTAS_CHANGE_MSI_FN
:
298 ret_intr_type
= RTAS_TYPE_MSI
;
300 case RTAS_CHANGE_MSIX_FN
:
301 ret_intr_type
= RTAS_TYPE_MSIX
;
304 error_report("rtas_ibm_change_msi(%u) is not implemented", func
);
305 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
309 /* Fins sPAPRPHBState */
310 phb
= spapr_pci_find_phb(spapr
, buid
);
312 pdev
= spapr_pci_find_dev(spapr
, buid
, config_addr
);
315 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
319 msi
= (spapr_pci_msi
*) g_hash_table_lookup(phb
->msi
, &config_addr
);
324 trace_spapr_pci_msi("Releasing wrong config", config_addr
);
325 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
329 spapr_ics_free(spapr
->ics
, msi
->first_irq
, msi
->num
);
330 if (msi_present(pdev
)) {
331 spapr_msi_setmsg(pdev
, 0, false, 0, 0);
333 if (msix_present(pdev
)) {
334 spapr_msi_setmsg(pdev
, 0, true, 0, 0);
336 g_hash_table_remove(phb
->msi
, &config_addr
);
338 trace_spapr_pci_msi("Released MSIs", config_addr
);
339 rtas_st(rets
, 0, RTAS_OUT_SUCCESS
);
346 /* Check if the device supports as many IRQs as requested */
347 if (ret_intr_type
== RTAS_TYPE_MSI
) {
348 max_irqs
= msi_nr_vectors_allocated(pdev
);
349 } else if (ret_intr_type
== RTAS_TYPE_MSIX
) {
350 max_irqs
= pdev
->msix_entries_nr
;
353 error_report("Requested interrupt type %d is not enabled for device %x",
354 ret_intr_type
, config_addr
);
355 rtas_st(rets
, 0, -1); /* Hardware error */
358 /* Correct the number if the guest asked for too many */
359 if (req_num
> max_irqs
) {
360 trace_spapr_pci_msi_retry(config_addr
, req_num
, max_irqs
);
362 irq
= 0; /* to avoid misleading trace */
367 irq
= spapr_ics_alloc_block(spapr
->ics
, req_num
, false,
368 ret_intr_type
== RTAS_TYPE_MSI
, &err
);
370 error_reportf_err(err
, "Can't allocate MSIs for device %x: ",
372 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
376 /* Release previous MSIs */
378 spapr_ics_free(spapr
->ics
, msi
->first_irq
, msi
->num
);
379 g_hash_table_remove(phb
->msi
, &config_addr
);
382 /* Setup MSI/MSIX vectors in the device (via cfgspace or MSIX BAR) */
383 spapr_msi_setmsg(pdev
, SPAPR_PCI_MSI_WINDOW
, ret_intr_type
== RTAS_TYPE_MSIX
,
386 /* Add MSI device to cache */
387 msi
= g_new(spapr_pci_msi
, 1);
388 msi
->first_irq
= irq
;
390 config_addr_key
= g_new(int, 1);
391 *config_addr_key
= config_addr
;
392 g_hash_table_insert(phb
->msi
, config_addr_key
, msi
);
395 rtas_st(rets
, 0, RTAS_OUT_SUCCESS
);
396 rtas_st(rets
, 1, req_num
);
397 rtas_st(rets
, 2, ++seq_num
);
399 rtas_st(rets
, 3, ret_intr_type
);
402 trace_spapr_pci_rtas_ibm_change_msi(config_addr
, func
, req_num
, irq
);
405 static void rtas_ibm_query_interrupt_source_number(PowerPCCPU
*cpu
,
406 sPAPRMachineState
*spapr
,
413 uint32_t config_addr
= rtas_ld(args
, 0);
414 uint64_t buid
= rtas_ldq(args
, 1);
415 unsigned int intr_src_num
= -1, ioa_intr_num
= rtas_ld(args
, 3);
416 sPAPRPHBState
*phb
= NULL
;
417 PCIDevice
*pdev
= NULL
;
420 /* Find sPAPRPHBState */
421 phb
= spapr_pci_find_phb(spapr
, buid
);
423 pdev
= spapr_pci_find_dev(spapr
, buid
, config_addr
);
426 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
430 /* Find device descriptor and start IRQ */
431 msi
= (spapr_pci_msi
*) g_hash_table_lookup(phb
->msi
, &config_addr
);
432 if (!msi
|| !msi
->first_irq
|| !msi
->num
|| (ioa_intr_num
>= msi
->num
)) {
433 trace_spapr_pci_msi("Failed to return vector", config_addr
);
434 rtas_st(rets
, 0, RTAS_OUT_HW_ERROR
);
437 intr_src_num
= msi
->first_irq
+ ioa_intr_num
;
438 trace_spapr_pci_rtas_ibm_query_interrupt_source_number(ioa_intr_num
,
441 rtas_st(rets
, 0, RTAS_OUT_SUCCESS
);
442 rtas_st(rets
, 1, intr_src_num
);
443 rtas_st(rets
, 2, 1);/* 0 == level; 1 == edge */
446 static void rtas_ibm_set_eeh_option(PowerPCCPU
*cpu
,
447 sPAPRMachineState
*spapr
,
448 uint32_t token
, uint32_t nargs
,
449 target_ulong args
, uint32_t nret
,
453 uint32_t addr
, option
;
457 if ((nargs
!= 4) || (nret
!= 1)) {
458 goto param_error_exit
;
461 buid
= rtas_ldq(args
, 1);
462 addr
= rtas_ld(args
, 0);
463 option
= rtas_ld(args
, 3);
465 sphb
= spapr_pci_find_phb(spapr
, buid
);
467 goto param_error_exit
;
470 if (!spapr_phb_eeh_available(sphb
)) {
471 goto param_error_exit
;
474 ret
= spapr_phb_vfio_eeh_set_option(sphb
, addr
, option
);
475 rtas_st(rets
, 0, ret
);
479 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
482 static void rtas_ibm_get_config_addr_info2(PowerPCCPU
*cpu
,
483 sPAPRMachineState
*spapr
,
484 uint32_t token
, uint32_t nargs
,
485 target_ulong args
, uint32_t nret
,
490 uint32_t addr
, option
;
493 if ((nargs
!= 4) || (nret
!= 2)) {
494 goto param_error_exit
;
497 buid
= rtas_ldq(args
, 1);
498 sphb
= spapr_pci_find_phb(spapr
, buid
);
500 goto param_error_exit
;
503 if (!spapr_phb_eeh_available(sphb
)) {
504 goto param_error_exit
;
508 * We always have PE address of form "00BB0001". "BB"
509 * represents the bus number of PE's primary bus.
511 option
= rtas_ld(args
, 3);
513 case RTAS_GET_PE_ADDR
:
514 addr
= rtas_ld(args
, 0);
515 pdev
= spapr_pci_find_dev(spapr
, buid
, addr
);
517 goto param_error_exit
;
520 rtas_st(rets
, 1, (pci_bus_num(pdev
->bus
) << 16) + 1);
522 case RTAS_GET_PE_MODE
:
523 rtas_st(rets
, 1, RTAS_PE_MODE_SHARED
);
526 goto param_error_exit
;
529 rtas_st(rets
, 0, RTAS_OUT_SUCCESS
);
533 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
536 static void rtas_ibm_read_slot_reset_state2(PowerPCCPU
*cpu
,
537 sPAPRMachineState
*spapr
,
538 uint32_t token
, uint32_t nargs
,
539 target_ulong args
, uint32_t nret
,
546 if ((nargs
!= 3) || (nret
!= 4 && nret
!= 5)) {
547 goto param_error_exit
;
550 buid
= rtas_ldq(args
, 1);
551 sphb
= spapr_pci_find_phb(spapr
, buid
);
553 goto param_error_exit
;
556 if (!spapr_phb_eeh_available(sphb
)) {
557 goto param_error_exit
;
560 ret
= spapr_phb_vfio_eeh_get_state(sphb
, &state
);
561 rtas_st(rets
, 0, ret
);
562 if (ret
!= RTAS_OUT_SUCCESS
) {
566 rtas_st(rets
, 1, state
);
567 rtas_st(rets
, 2, RTAS_EEH_SUPPORT
);
568 rtas_st(rets
, 3, RTAS_EEH_PE_UNAVAIL_INFO
);
570 rtas_st(rets
, 4, RTAS_EEH_PE_RECOVER_INFO
);
575 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
578 static void rtas_ibm_set_slot_reset(PowerPCCPU
*cpu
,
579 sPAPRMachineState
*spapr
,
580 uint32_t token
, uint32_t nargs
,
581 target_ulong args
, uint32_t nret
,
589 if ((nargs
!= 4) || (nret
!= 1)) {
590 goto param_error_exit
;
593 buid
= rtas_ldq(args
, 1);
594 option
= rtas_ld(args
, 3);
595 sphb
= spapr_pci_find_phb(spapr
, buid
);
597 goto param_error_exit
;
600 if (!spapr_phb_eeh_available(sphb
)) {
601 goto param_error_exit
;
604 ret
= spapr_phb_vfio_eeh_reset(sphb
, option
);
605 rtas_st(rets
, 0, ret
);
609 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
612 static void rtas_ibm_configure_pe(PowerPCCPU
*cpu
,
613 sPAPRMachineState
*spapr
,
614 uint32_t token
, uint32_t nargs
,
615 target_ulong args
, uint32_t nret
,
622 if ((nargs
!= 3) || (nret
!= 1)) {
623 goto param_error_exit
;
626 buid
= rtas_ldq(args
, 1);
627 sphb
= spapr_pci_find_phb(spapr
, buid
);
629 goto param_error_exit
;
632 if (!spapr_phb_eeh_available(sphb
)) {
633 goto param_error_exit
;
636 ret
= spapr_phb_vfio_eeh_configure(sphb
);
637 rtas_st(rets
, 0, ret
);
641 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
644 /* To support it later */
645 static void rtas_ibm_slot_error_detail(PowerPCCPU
*cpu
,
646 sPAPRMachineState
*spapr
,
647 uint32_t token
, uint32_t nargs
,
648 target_ulong args
, uint32_t nret
,
655 if ((nargs
!= 8) || (nret
!= 1)) {
656 goto param_error_exit
;
659 buid
= rtas_ldq(args
, 1);
660 sphb
= spapr_pci_find_phb(spapr
, buid
);
662 goto param_error_exit
;
665 if (!spapr_phb_eeh_available(sphb
)) {
666 goto param_error_exit
;
669 option
= rtas_ld(args
, 7);
671 case RTAS_SLOT_TEMP_ERR_LOG
:
672 case RTAS_SLOT_PERM_ERR_LOG
:
675 goto param_error_exit
;
678 /* We don't have error log yet */
679 rtas_st(rets
, 0, RTAS_OUT_NO_ERRORS_FOUND
);
683 rtas_st(rets
, 0, RTAS_OUT_PARAM_ERROR
);
686 static int pci_spapr_swizzle(int slot
, int pin
)
688 return (slot
+ pin
) % PCI_NUM_PINS
;
691 static int pci_spapr_map_irq(PCIDevice
*pci_dev
, int irq_num
)
694 * Here we need to convert pci_dev + irq_num to some unique value
695 * which is less than number of IRQs on the specific bus (4). We
696 * use standard PCI swizzling, that is (slot number + pin number)
699 return pci_spapr_swizzle(PCI_SLOT(pci_dev
->devfn
), irq_num
);
702 static void pci_spapr_set_irq(void *opaque
, int irq_num
, int level
)
705 * Here we use the number returned by pci_spapr_map_irq to find a
706 * corresponding qemu_irq.
708 sPAPRPHBState
*phb
= opaque
;
710 trace_spapr_pci_lsi_set(phb
->dtbusname
, irq_num
, phb
->lsi_table
[irq_num
].irq
);
711 qemu_set_irq(spapr_phb_lsi_qirq(phb
, irq_num
), level
);
714 static PCIINTxRoute
spapr_route_intx_pin_to_irq(void *opaque
, int pin
)
716 sPAPRPHBState
*sphb
= SPAPR_PCI_HOST_BRIDGE(opaque
);
719 route
.mode
= PCI_INTX_ENABLED
;
720 route
.irq
= sphb
->lsi_table
[pin
].irq
;
726 * MSI/MSIX memory region implementation.
727 * The handler handles both MSI and MSIX.
728 * For MSI-X, the vector number is encoded as a part of the address,
730 * For MSI, the vector number is encoded in least bits in data.
732 static void spapr_msi_write(void *opaque
, hwaddr addr
,
733 uint64_t data
, unsigned size
)
735 sPAPRMachineState
*spapr
= SPAPR_MACHINE(qdev_get_machine());
738 trace_spapr_pci_msi_write(addr
, data
, irq
);
740 qemu_irq_pulse(xics_get_qirq(XICS_FABRIC(spapr
), irq
));
743 static const MemoryRegionOps spapr_msi_ops
= {
744 /* There is no .read as the read result is undefined by PCI spec */
746 .write
= spapr_msi_write
,
747 .endianness
= DEVICE_LITTLE_ENDIAN
753 static AddressSpace
*spapr_pci_dma_iommu(PCIBus
*bus
, void *opaque
, int devfn
)
755 sPAPRPHBState
*phb
= opaque
;
757 return &phb
->iommu_as
;
760 static char *spapr_phb_vfio_get_loc_code(sPAPRPHBState
*sphb
, PCIDevice
*pdev
)
762 char *path
= NULL
, *buf
= NULL
, *host
= NULL
;
764 /* Get the PCI VFIO host id */
765 host
= object_property_get_str(OBJECT(pdev
), "host", NULL
);
770 /* Construct the path of the file that will give us the DT location */
771 path
= g_strdup_printf("/sys/bus/pci/devices/%s/devspec", host
);
773 if (!path
|| !g_file_get_contents(path
, &buf
, NULL
, NULL
)) {
778 /* Construct and read from host device tree the loc-code */
779 path
= g_strdup_printf("/proc/device-tree%s/ibm,loc-code", buf
);
781 if (!path
|| !g_file_get_contents(path
, &buf
, NULL
, NULL
)) {
791 static char *spapr_phb_get_loc_code(sPAPRPHBState
*sphb
, PCIDevice
*pdev
)
794 const char *devtype
= "qemu";
795 uint32_t busnr
= pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(pdev
))));
797 if (object_dynamic_cast(OBJECT(pdev
), "vfio-pci")) {
798 buf
= spapr_phb_vfio_get_loc_code(sphb
, pdev
);
805 * For emulated devices and VFIO-failure case, make up
808 buf
= g_strdup_printf("%s_%s:%04x:%02x:%02x.%x",
809 devtype
, pdev
->name
, sphb
->index
, busnr
,
810 PCI_SLOT(pdev
->devfn
), PCI_FUNC(pdev
->devfn
));
814 /* Macros to operate with address in OF binding to PCI */
815 #define b_x(x, p, l) (((x) & ((1<<(l))-1)) << (p))
816 #define b_n(x) b_x((x), 31, 1) /* 0 if relocatable */
817 #define b_p(x) b_x((x), 30, 1) /* 1 if prefetchable */
818 #define b_t(x) b_x((x), 29, 1) /* 1 if the address is aliased */
819 #define b_ss(x) b_x((x), 24, 2) /* the space code */
820 #define b_bbbbbbbb(x) b_x((x), 16, 8) /* bus number */
821 #define b_ddddd(x) b_x((x), 11, 5) /* device number */
822 #define b_fff(x) b_x((x), 8, 3) /* function number */
823 #define b_rrrrrrrr(x) b_x((x), 0, 8) /* register number */
825 /* for 'reg'/'assigned-addresses' OF properties */
826 #define RESOURCE_CELLS_SIZE 2
827 #define RESOURCE_CELLS_ADDRESS 3
829 typedef struct ResourceFields
{
835 } QEMU_PACKED ResourceFields
;
837 typedef struct ResourceProps
{
838 ResourceFields reg
[8];
839 ResourceFields assigned
[7];
841 uint32_t assigned_len
;
844 /* fill in the 'reg'/'assigned-resources' OF properties for
845 * a PCI device. 'reg' describes resource requirements for a
846 * device's IO/MEM regions, 'assigned-addresses' describes the
847 * actual resource assignments.
849 * the properties are arrays of ('phys-addr', 'size') pairs describing
850 * the addressable regions of the PCI device, where 'phys-addr' is a
851 * RESOURCE_CELLS_ADDRESS-tuple of 32-bit integers corresponding to
852 * (phys.hi, phys.mid, phys.lo), and 'size' is a
853 * RESOURCE_CELLS_SIZE-tuple corresponding to (size.hi, size.lo).
855 * phys.hi = 0xYYXXXXZZ, where:
860 * ||| + 00 if configuration space
861 * ||| + 01 if IO region,
862 * ||| + 10 if 32-bit MEM region
863 * ||| + 11 if 64-bit MEM region
865 * ||+------ for non-relocatable IO: 1 if aliased
866 * || for relocatable IO: 1 if below 64KB
867 * || for MEM: 1 if below 1MB
868 * |+------- 1 if region is prefetchable
869 * +-------- 1 if region is non-relocatable
870 * 0xXXXX = bbbbbbbb dddddfff, encoding bus, slot, and function
872 * 0xZZ = rrrrrrrr, the register number of the BAR corresponding
875 * phys.mid and phys.lo correspond respectively to the hi/lo portions
876 * of the actual address of the region.
878 * how the phys-addr/size values are used differ slightly between
879 * 'reg' and 'assigned-addresses' properties. namely, 'reg' has
880 * an additional description for the config space region of the
881 * device, and in the case of QEMU has n=0 and phys.mid=phys.lo=0
882 * to describe the region as relocatable, with an address-mapping
883 * that corresponds directly to the PHB's address space for the
884 * resource. 'assigned-addresses' always has n=1 set with an absolute
885 * address assigned for the resource. in general, 'assigned-addresses'
886 * won't be populated, since addresses for PCI devices are generally
887 * unmapped initially and left to the guest to assign.
889 * note also that addresses defined in these properties are, at least
890 * for PAPR guests, relative to the PHBs IO/MEM windows, and
891 * correspond directly to the addresses in the BARs.
893 * in accordance with PCI Bus Binding to Open Firmware,
894 * IEEE Std 1275-1994, section 4.1.1, as implemented by PAPR+ v2.7,
897 static void populate_resource_props(PCIDevice
*d
, ResourceProps
*rp
)
899 int bus_num
= pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(d
))));
900 uint32_t dev_id
= (b_bbbbbbbb(bus_num
) |
901 b_ddddd(PCI_SLOT(d
->devfn
)) |
902 b_fff(PCI_FUNC(d
->devfn
)));
903 ResourceFields
*reg
, *assigned
;
904 int i
, reg_idx
= 0, assigned_idx
= 0;
906 /* config space region */
907 reg
= &rp
->reg
[reg_idx
++];
908 reg
->phys_hi
= cpu_to_be32(dev_id
);
914 for (i
= 0; i
< PCI_NUM_REGIONS
; i
++) {
915 if (!d
->io_regions
[i
].size
) {
919 reg
= &rp
->reg
[reg_idx
++];
921 reg
->phys_hi
= cpu_to_be32(dev_id
| b_rrrrrrrr(pci_bar(d
, i
)));
922 if (d
->io_regions
[i
].type
& PCI_BASE_ADDRESS_SPACE_IO
) {
923 reg
->phys_hi
|= cpu_to_be32(b_ss(1));
924 } else if (d
->io_regions
[i
].type
& PCI_BASE_ADDRESS_MEM_TYPE_64
) {
925 reg
->phys_hi
|= cpu_to_be32(b_ss(3));
927 reg
->phys_hi
|= cpu_to_be32(b_ss(2));
931 reg
->size_hi
= cpu_to_be32(d
->io_regions
[i
].size
>> 32);
932 reg
->size_lo
= cpu_to_be32(d
->io_regions
[i
].size
);
934 if (d
->io_regions
[i
].addr
== PCI_BAR_UNMAPPED
) {
938 assigned
= &rp
->assigned
[assigned_idx
++];
939 assigned
->phys_hi
= cpu_to_be32(reg
->phys_hi
| b_n(1));
940 assigned
->phys_mid
= cpu_to_be32(d
->io_regions
[i
].addr
>> 32);
941 assigned
->phys_lo
= cpu_to_be32(d
->io_regions
[i
].addr
);
942 assigned
->size_hi
= reg
->size_hi
;
943 assigned
->size_lo
= reg
->size_lo
;
946 rp
->reg_len
= reg_idx
* sizeof(ResourceFields
);
947 rp
->assigned_len
= assigned_idx
* sizeof(ResourceFields
);
950 typedef struct PCIClass PCIClass
;
951 typedef struct PCISubClass PCISubClass
;
952 typedef struct PCIIFace PCIIFace
;
962 const PCIIFace
*iface
;
967 const PCISubClass
*subc
;
970 static const PCISubClass undef_subclass
[] = {
971 { PCI_CLASS_NOT_DEFINED_VGA
, "display", NULL
},
972 { 0xFF, NULL
, NULL
},
975 static const PCISubClass mass_subclass
[] = {
976 { PCI_CLASS_STORAGE_SCSI
, "scsi", NULL
},
977 { PCI_CLASS_STORAGE_IDE
, "ide", NULL
},
978 { PCI_CLASS_STORAGE_FLOPPY
, "fdc", NULL
},
979 { PCI_CLASS_STORAGE_IPI
, "ipi", NULL
},
980 { PCI_CLASS_STORAGE_RAID
, "raid", NULL
},
981 { PCI_CLASS_STORAGE_ATA
, "ata", NULL
},
982 { PCI_CLASS_STORAGE_SATA
, "sata", NULL
},
983 { PCI_CLASS_STORAGE_SAS
, "sas", NULL
},
984 { 0xFF, NULL
, NULL
},
987 static const PCISubClass net_subclass
[] = {
988 { PCI_CLASS_NETWORK_ETHERNET
, "ethernet", NULL
},
989 { PCI_CLASS_NETWORK_TOKEN_RING
, "token-ring", NULL
},
990 { PCI_CLASS_NETWORK_FDDI
, "fddi", NULL
},
991 { PCI_CLASS_NETWORK_ATM
, "atm", NULL
},
992 { PCI_CLASS_NETWORK_ISDN
, "isdn", NULL
},
993 { PCI_CLASS_NETWORK_WORLDFIP
, "worldfip", NULL
},
994 { PCI_CLASS_NETWORK_PICMG214
, "picmg", NULL
},
995 { 0xFF, NULL
, NULL
},
998 static const PCISubClass displ_subclass
[] = {
999 { PCI_CLASS_DISPLAY_VGA
, "vga", NULL
},
1000 { PCI_CLASS_DISPLAY_XGA
, "xga", NULL
},
1001 { PCI_CLASS_DISPLAY_3D
, "3d-controller", NULL
},
1002 { 0xFF, NULL
, NULL
},
1005 static const PCISubClass media_subclass
[] = {
1006 { PCI_CLASS_MULTIMEDIA_VIDEO
, "video", NULL
},
1007 { PCI_CLASS_MULTIMEDIA_AUDIO
, "sound", NULL
},
1008 { PCI_CLASS_MULTIMEDIA_PHONE
, "telephony", NULL
},
1009 { 0xFF, NULL
, NULL
},
1012 static const PCISubClass mem_subclass
[] = {
1013 { PCI_CLASS_MEMORY_RAM
, "memory", NULL
},
1014 { PCI_CLASS_MEMORY_FLASH
, "flash", NULL
},
1015 { 0xFF, NULL
, NULL
},
1018 static const PCISubClass bridg_subclass
[] = {
1019 { PCI_CLASS_BRIDGE_HOST
, "host", NULL
},
1020 { PCI_CLASS_BRIDGE_ISA
, "isa", NULL
},
1021 { PCI_CLASS_BRIDGE_EISA
, "eisa", NULL
},
1022 { PCI_CLASS_BRIDGE_MC
, "mca", NULL
},
1023 { PCI_CLASS_BRIDGE_PCI
, "pci", NULL
},
1024 { PCI_CLASS_BRIDGE_PCMCIA
, "pcmcia", NULL
},
1025 { PCI_CLASS_BRIDGE_NUBUS
, "nubus", NULL
},
1026 { PCI_CLASS_BRIDGE_CARDBUS
, "cardbus", NULL
},
1027 { PCI_CLASS_BRIDGE_RACEWAY
, "raceway", NULL
},
1028 { PCI_CLASS_BRIDGE_PCI_SEMITP
, "semi-transparent-pci", NULL
},
1029 { PCI_CLASS_BRIDGE_IB_PCI
, "infiniband", NULL
},
1030 { 0xFF, NULL
, NULL
},
1033 static const PCISubClass comm_subclass
[] = {
1034 { PCI_CLASS_COMMUNICATION_SERIAL
, "serial", NULL
},
1035 { PCI_CLASS_COMMUNICATION_PARALLEL
, "parallel", NULL
},
1036 { PCI_CLASS_COMMUNICATION_MULTISERIAL
, "multiport-serial", NULL
},
1037 { PCI_CLASS_COMMUNICATION_MODEM
, "modem", NULL
},
1038 { PCI_CLASS_COMMUNICATION_GPIB
, "gpib", NULL
},
1039 { PCI_CLASS_COMMUNICATION_SC
, "smart-card", NULL
},
1040 { 0xFF, NULL
, NULL
, },
1043 static const PCIIFace pic_iface
[] = {
1044 { PCI_CLASS_SYSTEM_PIC_IOAPIC
, "io-apic" },
1045 { PCI_CLASS_SYSTEM_PIC_IOXAPIC
, "io-xapic" },
1049 static const PCISubClass sys_subclass
[] = {
1050 { PCI_CLASS_SYSTEM_PIC
, "interrupt-controller", pic_iface
},
1051 { PCI_CLASS_SYSTEM_DMA
, "dma-controller", NULL
},
1052 { PCI_CLASS_SYSTEM_TIMER
, "timer", NULL
},
1053 { PCI_CLASS_SYSTEM_RTC
, "rtc", NULL
},
1054 { PCI_CLASS_SYSTEM_PCI_HOTPLUG
, "hot-plug-controller", NULL
},
1055 { PCI_CLASS_SYSTEM_SDHCI
, "sd-host-controller", NULL
},
1056 { 0xFF, NULL
, NULL
},
1059 static const PCISubClass inp_subclass
[] = {
1060 { PCI_CLASS_INPUT_KEYBOARD
, "keyboard", NULL
},
1061 { PCI_CLASS_INPUT_PEN
, "pen", NULL
},
1062 { PCI_CLASS_INPUT_MOUSE
, "mouse", NULL
},
1063 { PCI_CLASS_INPUT_SCANNER
, "scanner", NULL
},
1064 { PCI_CLASS_INPUT_GAMEPORT
, "gameport", NULL
},
1065 { 0xFF, NULL
, NULL
},
1068 static const PCISubClass dock_subclass
[] = {
1069 { PCI_CLASS_DOCKING_GENERIC
, "dock", NULL
},
1070 { 0xFF, NULL
, NULL
},
1073 static const PCISubClass cpu_subclass
[] = {
1074 { PCI_CLASS_PROCESSOR_PENTIUM
, "pentium", NULL
},
1075 { PCI_CLASS_PROCESSOR_POWERPC
, "powerpc", NULL
},
1076 { PCI_CLASS_PROCESSOR_MIPS
, "mips", NULL
},
1077 { PCI_CLASS_PROCESSOR_CO
, "co-processor", NULL
},
1078 { 0xFF, NULL
, NULL
},
1081 static const PCIIFace usb_iface
[] = {
1082 { PCI_CLASS_SERIAL_USB_UHCI
, "usb-uhci" },
1083 { PCI_CLASS_SERIAL_USB_OHCI
, "usb-ohci", },
1084 { PCI_CLASS_SERIAL_USB_EHCI
, "usb-ehci" },
1085 { PCI_CLASS_SERIAL_USB_XHCI
, "usb-xhci" },
1086 { PCI_CLASS_SERIAL_USB_UNKNOWN
, "usb-unknown" },
1087 { PCI_CLASS_SERIAL_USB_DEVICE
, "usb-device" },
1091 static const PCISubClass ser_subclass
[] = {
1092 { PCI_CLASS_SERIAL_FIREWIRE
, "firewire", NULL
},
1093 { PCI_CLASS_SERIAL_ACCESS
, "access-bus", NULL
},
1094 { PCI_CLASS_SERIAL_SSA
, "ssa", NULL
},
1095 { PCI_CLASS_SERIAL_USB
, "usb", usb_iface
},
1096 { PCI_CLASS_SERIAL_FIBER
, "fibre-channel", NULL
},
1097 { PCI_CLASS_SERIAL_SMBUS
, "smb", NULL
},
1098 { PCI_CLASS_SERIAL_IB
, "infiniband", NULL
},
1099 { PCI_CLASS_SERIAL_IPMI
, "ipmi", NULL
},
1100 { PCI_CLASS_SERIAL_SERCOS
, "sercos", NULL
},
1101 { PCI_CLASS_SERIAL_CANBUS
, "canbus", NULL
},
1102 { 0xFF, NULL
, NULL
},
1105 static const PCISubClass wrl_subclass
[] = {
1106 { PCI_CLASS_WIRELESS_IRDA
, "irda", NULL
},
1107 { PCI_CLASS_WIRELESS_CIR
, "consumer-ir", NULL
},
1108 { PCI_CLASS_WIRELESS_RF_CONTROLLER
, "rf-controller", NULL
},
1109 { PCI_CLASS_WIRELESS_BLUETOOTH
, "bluetooth", NULL
},
1110 { PCI_CLASS_WIRELESS_BROADBAND
, "broadband", NULL
},
1111 { 0xFF, NULL
, NULL
},
1114 static const PCISubClass sat_subclass
[] = {
1115 { PCI_CLASS_SATELLITE_TV
, "satellite-tv", NULL
},
1116 { PCI_CLASS_SATELLITE_AUDIO
, "satellite-audio", NULL
},
1117 { PCI_CLASS_SATELLITE_VOICE
, "satellite-voice", NULL
},
1118 { PCI_CLASS_SATELLITE_DATA
, "satellite-data", NULL
},
1119 { 0xFF, NULL
, NULL
},
1122 static const PCISubClass crypt_subclass
[] = {
1123 { PCI_CLASS_CRYPT_NETWORK
, "network-encryption", NULL
},
1124 { PCI_CLASS_CRYPT_ENTERTAINMENT
,
1125 "entertainment-encryption", NULL
},
1126 { 0xFF, NULL
, NULL
},
1129 static const PCISubClass spc_subclass
[] = {
1130 { PCI_CLASS_SP_DPIO
, "dpio", NULL
},
1131 { PCI_CLASS_SP_PERF
, "counter", NULL
},
1132 { PCI_CLASS_SP_SYNCH
, "measurement", NULL
},
1133 { PCI_CLASS_SP_MANAGEMENT
, "management-card", NULL
},
1134 { 0xFF, NULL
, NULL
},
1137 static const PCIClass pci_classes
[] = {
1138 { "legacy-device", undef_subclass
},
1139 { "mass-storage", mass_subclass
},
1140 { "network", net_subclass
},
1141 { "display", displ_subclass
, },
1142 { "multimedia-device", media_subclass
},
1143 { "memory-controller", mem_subclass
},
1144 { "unknown-bridge", bridg_subclass
},
1145 { "communication-controller", comm_subclass
},
1146 { "system-peripheral", sys_subclass
},
1147 { "input-controller", inp_subclass
},
1148 { "docking-station", dock_subclass
},
1149 { "cpu", cpu_subclass
},
1150 { "serial-bus", ser_subclass
},
1151 { "wireless-controller", wrl_subclass
},
1152 { "intelligent-io", NULL
},
1153 { "satellite-device", sat_subclass
},
1154 { "encryption", crypt_subclass
},
1155 { "data-processing-controller", spc_subclass
},
1158 static const char *pci_find_device_name(uint8_t class, uint8_t subclass
,
1161 const PCIClass
*pclass
;
1162 const PCISubClass
*psubclass
;
1163 const PCIIFace
*piface
;
1166 if (class >= ARRAY_SIZE(pci_classes
)) {
1170 pclass
= pci_classes
+ class;
1171 name
= pclass
->name
;
1173 if (pclass
->subc
== NULL
) {
1177 psubclass
= pclass
->subc
;
1178 while ((psubclass
->subclass
& 0xff) != 0xff) {
1179 if ((psubclass
->subclass
& 0xff) == subclass
) {
1180 name
= psubclass
->name
;
1186 piface
= psubclass
->iface
;
1187 if (piface
== NULL
) {
1190 while ((piface
->iface
& 0xff) != 0xff) {
1191 if ((piface
->iface
& 0xff) == iface
) {
1192 name
= piface
->name
;
1201 static void pci_get_node_name(char *nodename
, int len
, PCIDevice
*dev
)
1203 int slot
= PCI_SLOT(dev
->devfn
);
1204 int func
= PCI_FUNC(dev
->devfn
);
1205 uint32_t ccode
= pci_default_read_config(dev
, PCI_CLASS_PROG
, 3);
1208 name
= pci_find_device_name((ccode
>> 16) & 0xff, (ccode
>> 8) & 0xff,
1212 snprintf(nodename
, len
, "%s@%x,%x", name
, slot
, func
);
1214 snprintf(nodename
, len
, "%s@%x", name
, slot
);
1218 static uint32_t spapr_phb_get_pci_drc_index(sPAPRPHBState
*phb
,
1221 static int spapr_populate_pci_child_dt(PCIDevice
*dev
, void *fdt
, int offset
,
1222 sPAPRPHBState
*sphb
)
1225 bool is_bridge
= false;
1226 int pci_status
, err
;
1228 uint32_t drc_index
= spapr_phb_get_pci_drc_index(sphb
, dev
);
1229 uint32_t ccode
= pci_default_read_config(dev
, PCI_CLASS_PROG
, 3);
1230 uint32_t max_msi
, max_msix
;
1232 if (pci_default_read_config(dev
, PCI_HEADER_TYPE
, 1) ==
1233 PCI_HEADER_TYPE_BRIDGE
) {
1237 /* in accordance with PAPR+ v2.7 13.6.3, Table 181 */
1238 _FDT(fdt_setprop_cell(fdt
, offset
, "vendor-id",
1239 pci_default_read_config(dev
, PCI_VENDOR_ID
, 2)));
1240 _FDT(fdt_setprop_cell(fdt
, offset
, "device-id",
1241 pci_default_read_config(dev
, PCI_DEVICE_ID
, 2)));
1242 _FDT(fdt_setprop_cell(fdt
, offset
, "revision-id",
1243 pci_default_read_config(dev
, PCI_REVISION_ID
, 1)));
1244 _FDT(fdt_setprop_cell(fdt
, offset
, "class-code", ccode
));
1245 if (pci_default_read_config(dev
, PCI_INTERRUPT_PIN
, 1)) {
1246 _FDT(fdt_setprop_cell(fdt
, offset
, "interrupts",
1247 pci_default_read_config(dev
, PCI_INTERRUPT_PIN
, 1)));
1251 _FDT(fdt_setprop_cell(fdt
, offset
, "min-grant",
1252 pci_default_read_config(dev
, PCI_MIN_GNT
, 1)));
1253 _FDT(fdt_setprop_cell(fdt
, offset
, "max-latency",
1254 pci_default_read_config(dev
, PCI_MAX_LAT
, 1)));
1257 if (pci_default_read_config(dev
, PCI_SUBSYSTEM_ID
, 2)) {
1258 _FDT(fdt_setprop_cell(fdt
, offset
, "subsystem-id",
1259 pci_default_read_config(dev
, PCI_SUBSYSTEM_ID
, 2)));
1262 if (pci_default_read_config(dev
, PCI_SUBSYSTEM_VENDOR_ID
, 2)) {
1263 _FDT(fdt_setprop_cell(fdt
, offset
, "subsystem-vendor-id",
1264 pci_default_read_config(dev
, PCI_SUBSYSTEM_VENDOR_ID
, 2)));
1267 _FDT(fdt_setprop_cell(fdt
, offset
, "cache-line-size",
1268 pci_default_read_config(dev
, PCI_CACHE_LINE_SIZE
, 1)));
1270 /* the following fdt cells are masked off the pci status register */
1271 pci_status
= pci_default_read_config(dev
, PCI_STATUS
, 2);
1272 _FDT(fdt_setprop_cell(fdt
, offset
, "devsel-speed",
1273 PCI_STATUS_DEVSEL_MASK
& pci_status
));
1275 if (pci_status
& PCI_STATUS_FAST_BACK
) {
1276 _FDT(fdt_setprop(fdt
, offset
, "fast-back-to-back", NULL
, 0));
1278 if (pci_status
& PCI_STATUS_66MHZ
) {
1279 _FDT(fdt_setprop(fdt
, offset
, "66mhz-capable", NULL
, 0));
1281 if (pci_status
& PCI_STATUS_UDF
) {
1282 _FDT(fdt_setprop(fdt
, offset
, "udf-supported", NULL
, 0));
1285 _FDT(fdt_setprop_string(fdt
, offset
, "name",
1286 pci_find_device_name((ccode
>> 16) & 0xff,
1287 (ccode
>> 8) & 0xff,
1289 buf
= spapr_phb_get_loc_code(sphb
, dev
);
1291 error_report("Failed setting the ibm,loc-code");
1295 err
= fdt_setprop_string(fdt
, offset
, "ibm,loc-code", buf
);
1302 _FDT(fdt_setprop_cell(fdt
, offset
, "ibm,my-drc-index", drc_index
));
1305 _FDT(fdt_setprop_cell(fdt
, offset
, "#address-cells",
1306 RESOURCE_CELLS_ADDRESS
));
1307 _FDT(fdt_setprop_cell(fdt
, offset
, "#size-cells",
1308 RESOURCE_CELLS_SIZE
));
1310 max_msi
= msi_nr_vectors_allocated(dev
);
1312 _FDT(fdt_setprop_cell(fdt
, offset
, "ibm,req#msi", max_msi
));
1314 max_msix
= dev
->msix_entries_nr
;
1316 _FDT(fdt_setprop_cell(fdt
, offset
, "ibm,req#msi-x", max_msix
));
1319 populate_resource_props(dev
, &rp
);
1320 _FDT(fdt_setprop(fdt
, offset
, "reg", (uint8_t *)rp
.reg
, rp
.reg_len
));
1321 _FDT(fdt_setprop(fdt
, offset
, "assigned-addresses",
1322 (uint8_t *)rp
.assigned
, rp
.assigned_len
));
1324 if (sphb
->pcie_ecs
&& pci_is_express(dev
)) {
1325 _FDT(fdt_setprop_cell(fdt
, offset
, "ibm,pci-config-space-type", 0x1));
1331 /* create OF node for pci device and required OF DT properties */
1332 static int spapr_create_pci_child_dt(sPAPRPHBState
*phb
, PCIDevice
*dev
,
1333 void *fdt
, int node_offset
)
1336 char nodename
[FDT_NAME_MAX
];
1338 pci_get_node_name(nodename
, FDT_NAME_MAX
, dev
);
1339 offset
= fdt_add_subnode(fdt
, node_offset
, nodename
);
1340 ret
= spapr_populate_pci_child_dt(dev
, fdt
, offset
, phb
);
1349 static void spapr_phb_add_pci_device(sPAPRDRConnector
*drc
,
1354 sPAPRDRConnectorClass
*drck
= SPAPR_DR_CONNECTOR_GET_CLASS(drc
);
1355 DeviceState
*dev
= DEVICE(pdev
);
1357 int fdt_start_offset
= 0, fdt_size
;
1359 fdt
= create_device_tree(&fdt_size
);
1360 fdt_start_offset
= spapr_create_pci_child_dt(phb
, pdev
, fdt
, 0);
1361 if (!fdt_start_offset
) {
1362 error_setg(errp
, "Failed to create pci child device tree node");
1366 drck
->attach(drc
, DEVICE(pdev
),
1367 fdt
, fdt_start_offset
, !dev
->hotplugged
, errp
);
1374 static void spapr_phb_remove_pci_device_cb(DeviceState
*dev
, void *opaque
)
1376 /* some version guests do not wait for completion of a device
1377 * cleanup (generally done asynchronously by the kernel) before
1378 * signaling to QEMU that the device is safe, but instead sleep
1379 * for some 'safe' period of time. unfortunately on a busy host
1380 * this sleep isn't guaranteed to be long enough, resulting in
1381 * bad things like IRQ lines being left asserted during final
1382 * device removal. to deal with this we call reset just prior
1383 * to finalizing the device, which will put the device back into
1384 * an 'idle' state, as the device cleanup code expects.
1386 pci_device_reset(PCI_DEVICE(dev
));
1387 object_unparent(OBJECT(dev
));
1390 static void spapr_phb_remove_pci_device(sPAPRDRConnector
*drc
,
1395 sPAPRDRConnectorClass
*drck
= SPAPR_DR_CONNECTOR_GET_CLASS(drc
);
1397 drck
->detach(drc
, DEVICE(pdev
), spapr_phb_remove_pci_device_cb
, phb
, errp
);
1400 static sPAPRDRConnector
*spapr_phb_get_pci_func_drc(sPAPRPHBState
*phb
,
1404 return spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_PCI
,
1405 (phb
->index
<< 16) |
1410 static sPAPRDRConnector
*spapr_phb_get_pci_drc(sPAPRPHBState
*phb
,
1413 uint32_t busnr
= pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(pdev
))));
1414 return spapr_phb_get_pci_func_drc(phb
, busnr
, pdev
->devfn
);
1417 static uint32_t spapr_phb_get_pci_drc_index(sPAPRPHBState
*phb
,
1420 sPAPRDRConnector
*drc
= spapr_phb_get_pci_drc(phb
, pdev
);
1421 sPAPRDRConnectorClass
*drck
;
1427 drck
= SPAPR_DR_CONNECTOR_GET_CLASS(drc
);
1428 return drck
->get_index(drc
);
1431 static void spapr_phb_hot_plug_child(HotplugHandler
*plug_handler
,
1432 DeviceState
*plugged_dev
, Error
**errp
)
1434 sPAPRPHBState
*phb
= SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler
));
1435 PCIDevice
*pdev
= PCI_DEVICE(plugged_dev
);
1436 sPAPRDRConnector
*drc
= spapr_phb_get_pci_drc(phb
, pdev
);
1437 Error
*local_err
= NULL
;
1438 PCIBus
*bus
= PCI_BUS(qdev_get_parent_bus(DEVICE(pdev
)));
1439 uint32_t slotnr
= PCI_SLOT(pdev
->devfn
);
1441 /* if DR is disabled we don't need to do anything in the case of
1442 * hotplug or coldplug callbacks
1444 if (!phb
->dr_enabled
) {
1445 /* if this is a hotplug operation initiated by the user
1446 * we need to let them know it's not enabled
1448 if (plugged_dev
->hotplugged
) {
1449 error_setg(errp
, QERR_BUS_NO_HOTPLUG
,
1450 object_get_typename(OBJECT(phb
)));
1457 /* Following the QEMU convention used for PCIe multifunction
1458 * hotplug, we do not allow functions to be hotplugged to a
1459 * slot that already has function 0 present
1461 if (plugged_dev
->hotplugged
&& bus
->devices
[PCI_DEVFN(slotnr
, 0)] &&
1462 PCI_FUNC(pdev
->devfn
) != 0) {
1463 error_setg(errp
, "PCI: slot %d function 0 already ocuppied by %s,"
1464 " additional functions can no longer be exposed to guest.",
1465 slotnr
, bus
->devices
[PCI_DEVFN(slotnr
, 0)]->name
);
1469 spapr_phb_add_pci_device(drc
, phb
, pdev
, &local_err
);
1471 error_propagate(errp
, local_err
);
1475 /* If this is function 0, signal hotplug for all the device functions.
1476 * Otherwise defer sending the hotplug event.
1478 if (plugged_dev
->hotplugged
&& PCI_FUNC(pdev
->devfn
) == 0) {
1481 for (i
= 0; i
< 8; i
++) {
1482 sPAPRDRConnector
*func_drc
;
1483 sPAPRDRConnectorClass
*func_drck
;
1484 sPAPRDREntitySense state
;
1486 func_drc
= spapr_phb_get_pci_func_drc(phb
, pci_bus_num(bus
),
1487 PCI_DEVFN(slotnr
, i
));
1488 func_drck
= SPAPR_DR_CONNECTOR_GET_CLASS(func_drc
);
1489 func_drck
->entity_sense(func_drc
, &state
);
1491 if (state
== SPAPR_DR_ENTITY_SENSE_PRESENT
) {
1492 spapr_hotplug_req_add_by_index(func_drc
);
1498 static void spapr_phb_hot_unplug_child(HotplugHandler
*plug_handler
,
1499 DeviceState
*plugged_dev
, Error
**errp
)
1501 sPAPRPHBState
*phb
= SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler
));
1502 PCIDevice
*pdev
= PCI_DEVICE(plugged_dev
);
1503 sPAPRDRConnectorClass
*drck
;
1504 sPAPRDRConnector
*drc
= spapr_phb_get_pci_drc(phb
, pdev
);
1505 Error
*local_err
= NULL
;
1507 if (!phb
->dr_enabled
) {
1508 error_setg(errp
, QERR_BUS_NO_HOTPLUG
,
1509 object_get_typename(OBJECT(phb
)));
1515 drck
= SPAPR_DR_CONNECTOR_GET_CLASS(drc
);
1516 if (!drck
->release_pending(drc
)) {
1517 PCIBus
*bus
= PCI_BUS(qdev_get_parent_bus(DEVICE(pdev
)));
1518 uint32_t slotnr
= PCI_SLOT(pdev
->devfn
);
1519 sPAPRDRConnector
*func_drc
;
1520 sPAPRDRConnectorClass
*func_drck
;
1521 sPAPRDREntitySense state
;
1524 /* ensure any other present functions are pending unplug */
1525 if (PCI_FUNC(pdev
->devfn
) == 0) {
1526 for (i
= 1; i
< 8; i
++) {
1527 func_drc
= spapr_phb_get_pci_func_drc(phb
, pci_bus_num(bus
),
1528 PCI_DEVFN(slotnr
, i
));
1529 func_drck
= SPAPR_DR_CONNECTOR_GET_CLASS(func_drc
);
1530 func_drck
->entity_sense(func_drc
, &state
);
1531 if (state
== SPAPR_DR_ENTITY_SENSE_PRESENT
1532 && !func_drck
->release_pending(func_drc
)) {
1534 "PCI: slot %d, function %d still present. "
1535 "Must unplug all non-0 functions first.",
1542 spapr_phb_remove_pci_device(drc
, phb
, pdev
, &local_err
);
1544 error_propagate(errp
, local_err
);
1548 /* if this isn't func 0, defer unplug event. otherwise signal removal
1549 * for all present functions
1551 if (PCI_FUNC(pdev
->devfn
) == 0) {
1552 for (i
= 7; i
>= 0; i
--) {
1553 func_drc
= spapr_phb_get_pci_func_drc(phb
, pci_bus_num(bus
),
1554 PCI_DEVFN(slotnr
, i
));
1555 func_drck
= SPAPR_DR_CONNECTOR_GET_CLASS(func_drc
);
1556 func_drck
->entity_sense(func_drc
, &state
);
1557 if (state
== SPAPR_DR_ENTITY_SENSE_PRESENT
) {
1558 spapr_hotplug_req_remove_by_index(func_drc
);
1565 static void spapr_phb_realize(DeviceState
*dev
, Error
**errp
)
1567 sPAPRMachineState
*spapr
= SPAPR_MACHINE(qdev_get_machine());
1568 SysBusDevice
*s
= SYS_BUS_DEVICE(dev
);
1569 sPAPRPHBState
*sphb
= SPAPR_PCI_HOST_BRIDGE(s
);
1570 PCIHostState
*phb
= PCI_HOST_BRIDGE(s
);
1574 uint64_t msi_window_size
= 4096;
1575 sPAPRTCETable
*tcet
;
1576 const unsigned windows_supported
=
1577 sphb
->ddw_enabled
? SPAPR_PCI_DMA_MAX_WINDOWS
: 1;
1579 if (sphb
->index
!= (uint32_t)-1) {
1580 sPAPRMachineClass
*smc
= SPAPR_MACHINE_GET_CLASS(spapr
);
1581 Error
*local_err
= NULL
;
1583 if ((sphb
->buid
!= (uint64_t)-1) || (sphb
->dma_liobn
[0] != (uint32_t)-1)
1584 || (sphb
->dma_liobn
[1] != (uint32_t)-1 && windows_supported
== 2)
1585 || (sphb
->mem_win_addr
!= (hwaddr
)-1)
1586 || (sphb
->mem64_win_addr
!= (hwaddr
)-1)
1587 || (sphb
->io_win_addr
!= (hwaddr
)-1)) {
1588 error_setg(errp
, "Either \"index\" or other parameters must"
1589 " be specified for PAPR PHB, not both");
1593 smc
->phb_placement(spapr
, sphb
->index
,
1594 &sphb
->buid
, &sphb
->io_win_addr
,
1595 &sphb
->mem_win_addr
, &sphb
->mem64_win_addr
,
1596 windows_supported
, sphb
->dma_liobn
, &local_err
);
1598 error_propagate(errp
, local_err
);
1603 if (sphb
->buid
== (uint64_t)-1) {
1604 error_setg(errp
, "BUID not specified for PHB");
1608 if ((sphb
->dma_liobn
[0] == (uint32_t)-1) ||
1609 ((sphb
->dma_liobn
[1] == (uint32_t)-1) && (windows_supported
> 1))) {
1610 error_setg(errp
, "LIOBN(s) not specified for PHB");
1614 if (sphb
->mem_win_addr
== (hwaddr
)-1) {
1615 error_setg(errp
, "Memory window address not specified for PHB");
1619 if (sphb
->io_win_addr
== (hwaddr
)-1) {
1620 error_setg(errp
, "IO window address not specified for PHB");
1624 if (sphb
->mem64_win_size
!= 0) {
1625 if (sphb
->mem64_win_addr
== (hwaddr
)-1) {
1627 "64-bit memory window address not specified for PHB");
1631 if (sphb
->mem_win_size
> SPAPR_PCI_MEM32_WIN_SIZE
) {
1632 error_setg(errp
, "32-bit memory window of size 0x%"HWADDR_PRIx
1633 " (max 2 GiB)", sphb
->mem_win_size
);
1637 if (sphb
->mem64_win_pciaddr
== (hwaddr
)-1) {
1638 /* 64-bit window defaults to identity mapping */
1639 sphb
->mem64_win_pciaddr
= sphb
->mem64_win_addr
;
1641 } else if (sphb
->mem_win_size
> SPAPR_PCI_MEM32_WIN_SIZE
) {
1643 * For compatibility with old configuration, if no 64-bit MMIO
1644 * window is specified, but the ordinary (32-bit) memory
1645 * window is specified as > 2GiB, we treat it as a 2GiB 32-bit
1646 * window, with a 64-bit MMIO window following on immediately
1649 sphb
->mem64_win_size
= sphb
->mem_win_size
- SPAPR_PCI_MEM32_WIN_SIZE
;
1650 sphb
->mem64_win_addr
= sphb
->mem_win_addr
+ SPAPR_PCI_MEM32_WIN_SIZE
;
1651 sphb
->mem64_win_pciaddr
=
1652 SPAPR_PCI_MEM_WIN_BUS_OFFSET
+ SPAPR_PCI_MEM32_WIN_SIZE
;
1653 sphb
->mem_win_size
= SPAPR_PCI_MEM32_WIN_SIZE
;
1656 if (spapr_pci_find_phb(spapr
, sphb
->buid
)) {
1657 error_setg(errp
, "PCI host bridges must have unique BUIDs");
1661 if (sphb
->numa_node
!= -1 &&
1662 (sphb
->numa_node
>= MAX_NODES
|| !numa_info
[sphb
->numa_node
].present
)) {
1663 error_setg(errp
, "Invalid NUMA node ID for PCI host bridge");
1667 sphb
->dtbusname
= g_strdup_printf("pci@%" PRIx64
, sphb
->buid
);
1669 namebuf
= alloca(strlen(sphb
->dtbusname
) + 32);
1671 /* Initialize memory regions */
1672 sprintf(namebuf
, "%s.mmio", sphb
->dtbusname
);
1673 memory_region_init(&sphb
->memspace
, OBJECT(sphb
), namebuf
, UINT64_MAX
);
1675 sprintf(namebuf
, "%s.mmio32-alias", sphb
->dtbusname
);
1676 memory_region_init_alias(&sphb
->mem32window
, OBJECT(sphb
),
1677 namebuf
, &sphb
->memspace
,
1678 SPAPR_PCI_MEM_WIN_BUS_OFFSET
, sphb
->mem_win_size
);
1679 memory_region_add_subregion(get_system_memory(), sphb
->mem_win_addr
,
1680 &sphb
->mem32window
);
1682 sprintf(namebuf
, "%s.mmio64-alias", sphb
->dtbusname
);
1683 memory_region_init_alias(&sphb
->mem64window
, OBJECT(sphb
),
1684 namebuf
, &sphb
->memspace
,
1685 sphb
->mem64_win_pciaddr
, sphb
->mem64_win_size
);
1686 memory_region_add_subregion(get_system_memory(), sphb
->mem64_win_addr
,
1687 &sphb
->mem64window
);
1689 /* Initialize IO regions */
1690 sprintf(namebuf
, "%s.io", sphb
->dtbusname
);
1691 memory_region_init(&sphb
->iospace
, OBJECT(sphb
),
1692 namebuf
, SPAPR_PCI_IO_WIN_SIZE
);
1694 sprintf(namebuf
, "%s.io-alias", sphb
->dtbusname
);
1695 memory_region_init_alias(&sphb
->iowindow
, OBJECT(sphb
), namebuf
,
1696 &sphb
->iospace
, 0, SPAPR_PCI_IO_WIN_SIZE
);
1697 memory_region_add_subregion(get_system_memory(), sphb
->io_win_addr
,
1700 bus
= pci_register_bus(dev
, NULL
,
1701 pci_spapr_set_irq
, pci_spapr_map_irq
, sphb
,
1702 &sphb
->memspace
, &sphb
->iospace
,
1703 PCI_DEVFN(0, 0), PCI_NUM_PINS
, TYPE_PCI_BUS
);
1705 qbus_set_hotplug_handler(BUS(phb
->bus
), DEVICE(sphb
), NULL
);
1708 * Initialize PHB address space.
1709 * By default there will be at least one subregion for default
1711 * Later the guest might want to create another DMA window
1712 * which will become another memory subregion.
1714 sprintf(namebuf
, "%s.iommu-root", sphb
->dtbusname
);
1716 memory_region_init(&sphb
->iommu_root
, OBJECT(sphb
),
1717 namebuf
, UINT64_MAX
);
1718 address_space_init(&sphb
->iommu_as
, &sphb
->iommu_root
,
1722 * As MSI/MSIX interrupts trigger by writing at MSI/MSIX vectors,
1723 * we need to allocate some memory to catch those writes coming
1724 * from msi_notify()/msix_notify().
1725 * As MSIMessage:addr is going to be the same and MSIMessage:data
1726 * is going to be a VIRQ number, 4 bytes of the MSI MR will only
1729 * For KVM we want to ensure that this memory is a full page so that
1730 * our memory slot is of page size granularity.
1733 if (kvm_enabled()) {
1734 msi_window_size
= getpagesize();
1738 memory_region_init_io(&sphb
->msiwindow
, NULL
, &spapr_msi_ops
, spapr
,
1739 "msi", msi_window_size
);
1740 memory_region_add_subregion(&sphb
->iommu_root
, SPAPR_PCI_MSI_WINDOW
,
1743 pci_setup_iommu(bus
, spapr_pci_dma_iommu
, sphb
);
1745 pci_bus_set_route_irq_fn(bus
, spapr_route_intx_pin_to_irq
);
1747 QLIST_INSERT_HEAD(&spapr
->phbs
, sphb
, list
);
1749 /* Initialize the LSI table */
1750 for (i
= 0; i
< PCI_NUM_PINS
; i
++) {
1752 Error
*local_err
= NULL
;
1754 irq
= spapr_ics_alloc_block(spapr
->ics
, 1, true, false, &local_err
);
1756 error_propagate(errp
, local_err
);
1757 error_prepend(errp
, "can't allocate LSIs: ");
1761 sphb
->lsi_table
[i
].irq
= irq
;
1764 /* allocate connectors for child PCI devices */
1765 if (sphb
->dr_enabled
) {
1766 for (i
= 0; i
< PCI_SLOT_MAX
* 8; i
++) {
1767 spapr_dr_connector_new(OBJECT(phb
),
1768 SPAPR_DR_CONNECTOR_TYPE_PCI
,
1769 (sphb
->index
<< 16) | i
);
1774 if ((sphb
->page_size_mask
& qemu_getrampagesize()) == 0) {
1775 error_report("System page size 0x%lx is not enabled in page_size_mask "
1776 "(0x%"PRIx64
"). Performance may be slow",
1777 qemu_getrampagesize(), sphb
->page_size_mask
);
1780 for (i
= 0; i
< windows_supported
; ++i
) {
1781 tcet
= spapr_tce_new_table(DEVICE(sphb
), sphb
->dma_liobn
[i
]);
1783 error_setg(errp
, "Creating window#%d failed for %s",
1784 i
, sphb
->dtbusname
);
1787 memory_region_add_subregion_overlap(&sphb
->iommu_root
, 0,
1788 spapr_tce_get_iommu(tcet
), 0);
1791 sphb
->msi
= g_hash_table_new_full(g_int_hash
, g_int_equal
, g_free
, g_free
);
1794 static int spapr_phb_children_reset(Object
*child
, void *opaque
)
1796 DeviceState
*dev
= (DeviceState
*) object_dynamic_cast(child
, TYPE_DEVICE
);
1805 void spapr_phb_dma_reset(sPAPRPHBState
*sphb
)
1808 sPAPRTCETable
*tcet
;
1810 for (i
= 0; i
< SPAPR_PCI_DMA_MAX_WINDOWS
; ++i
) {
1811 tcet
= spapr_tce_find_by_liobn(sphb
->dma_liobn
[i
]);
1813 if (tcet
&& tcet
->nb_table
) {
1814 spapr_tce_table_disable(tcet
);
1818 /* Register default 32bit DMA window */
1819 tcet
= spapr_tce_find_by_liobn(sphb
->dma_liobn
[0]);
1820 spapr_tce_table_enable(tcet
, SPAPR_TCE_PAGE_SHIFT
, sphb
->dma_win_addr
,
1821 sphb
->dma_win_size
>> SPAPR_TCE_PAGE_SHIFT
);
1824 static void spapr_phb_reset(DeviceState
*qdev
)
1826 sPAPRPHBState
*sphb
= SPAPR_PCI_HOST_BRIDGE(qdev
);
1828 spapr_phb_dma_reset(sphb
);
1830 /* Reset the IOMMU state */
1831 object_child_foreach(OBJECT(qdev
), spapr_phb_children_reset
, NULL
);
1833 if (spapr_phb_eeh_available(SPAPR_PCI_HOST_BRIDGE(qdev
))) {
1834 spapr_phb_vfio_reset(qdev
);
1838 static Property spapr_phb_properties
[] = {
1839 DEFINE_PROP_UINT32("index", sPAPRPHBState
, index
, -1),
1840 DEFINE_PROP_UINT64("buid", sPAPRPHBState
, buid
, -1),
1841 DEFINE_PROP_UINT32("liobn", sPAPRPHBState
, dma_liobn
[0], -1),
1842 DEFINE_PROP_UINT32("liobn64", sPAPRPHBState
, dma_liobn
[1], -1),
1843 DEFINE_PROP_UINT64("mem_win_addr", sPAPRPHBState
, mem_win_addr
, -1),
1844 DEFINE_PROP_UINT64("mem_win_size", sPAPRPHBState
, mem_win_size
,
1845 SPAPR_PCI_MEM32_WIN_SIZE
),
1846 DEFINE_PROP_UINT64("mem64_win_addr", sPAPRPHBState
, mem64_win_addr
, -1),
1847 DEFINE_PROP_UINT64("mem64_win_size", sPAPRPHBState
, mem64_win_size
,
1848 SPAPR_PCI_MEM64_WIN_SIZE
),
1849 DEFINE_PROP_UINT64("mem64_win_pciaddr", sPAPRPHBState
, mem64_win_pciaddr
,
1851 DEFINE_PROP_UINT64("io_win_addr", sPAPRPHBState
, io_win_addr
, -1),
1852 DEFINE_PROP_UINT64("io_win_size", sPAPRPHBState
, io_win_size
,
1853 SPAPR_PCI_IO_WIN_SIZE
),
1854 DEFINE_PROP_BOOL("dynamic-reconfiguration", sPAPRPHBState
, dr_enabled
,
1856 /* Default DMA window is 0..1GB */
1857 DEFINE_PROP_UINT64("dma_win_addr", sPAPRPHBState
, dma_win_addr
, 0),
1858 DEFINE_PROP_UINT64("dma_win_size", sPAPRPHBState
, dma_win_size
, 0x40000000),
1859 DEFINE_PROP_UINT64("dma64_win_addr", sPAPRPHBState
, dma64_win_addr
,
1860 0x800000000000000ULL
),
1861 DEFINE_PROP_BOOL("ddw", sPAPRPHBState
, ddw_enabled
, true),
1862 DEFINE_PROP_UINT64("pgsz", sPAPRPHBState
, page_size_mask
,
1863 (1ULL << 12) | (1ULL << 16)),
1864 DEFINE_PROP_UINT32("numa_node", sPAPRPHBState
, numa_node
, -1),
1865 DEFINE_PROP_BOOL("pre-2.8-migration", sPAPRPHBState
,
1866 pre_2_8_migration
, false),
1867 DEFINE_PROP_BOOL("pcie-extended-configuration-space", sPAPRPHBState
,
1869 DEFINE_PROP_END_OF_LIST(),
1872 static const VMStateDescription vmstate_spapr_pci_lsi
= {
1873 .name
= "spapr_pci/lsi",
1875 .minimum_version_id
= 1,
1876 .fields
= (VMStateField
[]) {
1877 VMSTATE_UINT32_EQUAL(irq
, struct spapr_pci_lsi
),
1879 VMSTATE_END_OF_LIST()
1883 static const VMStateDescription vmstate_spapr_pci_msi
= {
1884 .name
= "spapr_pci/msi",
1886 .minimum_version_id
= 1,
1887 .fields
= (VMStateField
[]) {
1888 VMSTATE_UINT32(key
, spapr_pci_msi_mig
),
1889 VMSTATE_UINT32(value
.first_irq
, spapr_pci_msi_mig
),
1890 VMSTATE_UINT32(value
.num
, spapr_pci_msi_mig
),
1891 VMSTATE_END_OF_LIST()
1895 static void spapr_pci_pre_save(void *opaque
)
1897 sPAPRPHBState
*sphb
= opaque
;
1898 GHashTableIter iter
;
1899 gpointer key
, value
;
1902 g_free(sphb
->msi_devs
);
1903 sphb
->msi_devs
= NULL
;
1904 sphb
->msi_devs_num
= g_hash_table_size(sphb
->msi
);
1905 if (!sphb
->msi_devs_num
) {
1908 sphb
->msi_devs
= g_malloc(sphb
->msi_devs_num
* sizeof(spapr_pci_msi_mig
));
1910 g_hash_table_iter_init(&iter
, sphb
->msi
);
1911 for (i
= 0; g_hash_table_iter_next(&iter
, &key
, &value
); ++i
) {
1912 sphb
->msi_devs
[i
].key
= *(uint32_t *) key
;
1913 sphb
->msi_devs
[i
].value
= *(spapr_pci_msi
*) value
;
1916 if (sphb
->pre_2_8_migration
) {
1917 sphb
->mig_liobn
= sphb
->dma_liobn
[0];
1918 sphb
->mig_mem_win_addr
= sphb
->mem_win_addr
;
1919 sphb
->mig_mem_win_size
= sphb
->mem_win_size
;
1920 sphb
->mig_io_win_addr
= sphb
->io_win_addr
;
1921 sphb
->mig_io_win_size
= sphb
->io_win_size
;
1923 if ((sphb
->mem64_win_size
!= 0)
1924 && (sphb
->mem64_win_addr
1925 == (sphb
->mem_win_addr
+ sphb
->mem_win_size
))) {
1926 sphb
->mig_mem_win_size
+= sphb
->mem64_win_size
;
1931 static int spapr_pci_post_load(void *opaque
, int version_id
)
1933 sPAPRPHBState
*sphb
= opaque
;
1934 gpointer key
, value
;
1937 for (i
= 0; i
< sphb
->msi_devs_num
; ++i
) {
1938 key
= g_memdup(&sphb
->msi_devs
[i
].key
,
1939 sizeof(sphb
->msi_devs
[i
].key
));
1940 value
= g_memdup(&sphb
->msi_devs
[i
].value
,
1941 sizeof(sphb
->msi_devs
[i
].value
));
1942 g_hash_table_insert(sphb
->msi
, key
, value
);
1944 g_free(sphb
->msi_devs
);
1945 sphb
->msi_devs
= NULL
;
1946 sphb
->msi_devs_num
= 0;
1951 static bool pre_2_8_migration(void *opaque
, int version_id
)
1953 sPAPRPHBState
*sphb
= opaque
;
1955 return sphb
->pre_2_8_migration
;
1958 static const VMStateDescription vmstate_spapr_pci
= {
1959 .name
= "spapr_pci",
1961 .minimum_version_id
= 2,
1962 .pre_save
= spapr_pci_pre_save
,
1963 .post_load
= spapr_pci_post_load
,
1964 .fields
= (VMStateField
[]) {
1965 VMSTATE_UINT64_EQUAL(buid
, sPAPRPHBState
),
1966 VMSTATE_UINT32_TEST(mig_liobn
, sPAPRPHBState
, pre_2_8_migration
),
1967 VMSTATE_UINT64_TEST(mig_mem_win_addr
, sPAPRPHBState
, pre_2_8_migration
),
1968 VMSTATE_UINT64_TEST(mig_mem_win_size
, sPAPRPHBState
, pre_2_8_migration
),
1969 VMSTATE_UINT64_TEST(mig_io_win_addr
, sPAPRPHBState
, pre_2_8_migration
),
1970 VMSTATE_UINT64_TEST(mig_io_win_size
, sPAPRPHBState
, pre_2_8_migration
),
1971 VMSTATE_STRUCT_ARRAY(lsi_table
, sPAPRPHBState
, PCI_NUM_PINS
, 0,
1972 vmstate_spapr_pci_lsi
, struct spapr_pci_lsi
),
1973 VMSTATE_INT32(msi_devs_num
, sPAPRPHBState
),
1974 VMSTATE_STRUCT_VARRAY_ALLOC(msi_devs
, sPAPRPHBState
, msi_devs_num
, 0,
1975 vmstate_spapr_pci_msi
, spapr_pci_msi_mig
),
1976 VMSTATE_END_OF_LIST()
1980 static const char *spapr_phb_root_bus_path(PCIHostState
*host_bridge
,
1983 sPAPRPHBState
*sphb
= SPAPR_PCI_HOST_BRIDGE(host_bridge
);
1985 return sphb
->dtbusname
;
1988 static void spapr_phb_class_init(ObjectClass
*klass
, void *data
)
1990 PCIHostBridgeClass
*hc
= PCI_HOST_BRIDGE_CLASS(klass
);
1991 DeviceClass
*dc
= DEVICE_CLASS(klass
);
1992 HotplugHandlerClass
*hp
= HOTPLUG_HANDLER_CLASS(klass
);
1994 hc
->root_bus_path
= spapr_phb_root_bus_path
;
1995 dc
->realize
= spapr_phb_realize
;
1996 dc
->props
= spapr_phb_properties
;
1997 dc
->reset
= spapr_phb_reset
;
1998 dc
->vmsd
= &vmstate_spapr_pci
;
1999 set_bit(DEVICE_CATEGORY_BRIDGE
, dc
->categories
);
2000 hp
->plug
= spapr_phb_hot_plug_child
;
2001 hp
->unplug
= spapr_phb_hot_unplug_child
;
2004 static const TypeInfo spapr_phb_info
= {
2005 .name
= TYPE_SPAPR_PCI_HOST_BRIDGE
,
2006 .parent
= TYPE_PCI_HOST_BRIDGE
,
2007 .instance_size
= sizeof(sPAPRPHBState
),
2008 .class_init
= spapr_phb_class_init
,
2009 .interfaces
= (InterfaceInfo
[]) {
2010 { TYPE_HOTPLUG_HANDLER
},
2015 PCIHostState
*spapr_create_phb(sPAPRMachineState
*spapr
, int index
)
2019 dev
= qdev_create(NULL
, TYPE_SPAPR_PCI_HOST_BRIDGE
);
2020 qdev_prop_set_uint32(dev
, "index", index
);
2021 qdev_init_nofail(dev
);
2023 return PCI_HOST_BRIDGE(dev
);
2026 typedef struct sPAPRFDT
{
2029 sPAPRPHBState
*sphb
;
2032 static void spapr_populate_pci_devices_dt(PCIBus
*bus
, PCIDevice
*pdev
,
2036 sPAPRFDT
*p
= opaque
;
2040 offset
= spapr_create_pci_child_dt(p
->sphb
, pdev
, p
->fdt
, p
->node_off
);
2042 error_report("Failed to create pci child device tree node");
2046 if ((pci_default_read_config(pdev
, PCI_HEADER_TYPE
, 1) !=
2047 PCI_HEADER_TYPE_BRIDGE
)) {
2051 sec_bus
= pci_bridge_get_sec_bus(PCI_BRIDGE(pdev
));
2057 s_fdt
.node_off
= offset
;
2058 s_fdt
.sphb
= p
->sphb
;
2059 pci_for_each_device_reverse(sec_bus
, pci_bus_num(sec_bus
),
2060 spapr_populate_pci_devices_dt
,
2064 static void spapr_phb_pci_enumerate_bridge(PCIBus
*bus
, PCIDevice
*pdev
,
2067 unsigned int *bus_no
= opaque
;
2068 unsigned int primary
= *bus_no
;
2069 unsigned int subordinate
= 0xff;
2070 PCIBus
*sec_bus
= NULL
;
2072 if ((pci_default_read_config(pdev
, PCI_HEADER_TYPE
, 1) !=
2073 PCI_HEADER_TYPE_BRIDGE
)) {
2078 pci_default_write_config(pdev
, PCI_PRIMARY_BUS
, primary
, 1);
2079 pci_default_write_config(pdev
, PCI_SECONDARY_BUS
, *bus_no
, 1);
2080 pci_default_write_config(pdev
, PCI_SUBORDINATE_BUS
, *bus_no
, 1);
2082 sec_bus
= pci_bridge_get_sec_bus(PCI_BRIDGE(pdev
));
2087 pci_default_write_config(pdev
, PCI_SUBORDINATE_BUS
, subordinate
, 1);
2088 pci_for_each_device(sec_bus
, pci_bus_num(sec_bus
),
2089 spapr_phb_pci_enumerate_bridge
, bus_no
);
2090 pci_default_write_config(pdev
, PCI_SUBORDINATE_BUS
, *bus_no
, 1);
2093 static void spapr_phb_pci_enumerate(sPAPRPHBState
*phb
)
2095 PCIBus
*bus
= PCI_HOST_BRIDGE(phb
)->bus
;
2096 unsigned int bus_no
= 0;
2098 pci_for_each_device(bus
, pci_bus_num(bus
),
2099 spapr_phb_pci_enumerate_bridge
,
2104 int spapr_populate_pci_dt(sPAPRPHBState
*phb
,
2105 uint32_t xics_phandle
,
2108 int bus_off
, i
, j
, ret
;
2109 char nodename
[FDT_NAME_MAX
];
2110 uint32_t bus_range
[] = { cpu_to_be32(0), cpu_to_be32(0xff) };
2116 } QEMU_PACKED ranges
[] = {
2118 cpu_to_be32(b_ss(1)), cpu_to_be64(0),
2119 cpu_to_be64(phb
->io_win_addr
),
2120 cpu_to_be64(memory_region_size(&phb
->iospace
)),
2123 cpu_to_be32(b_ss(2)), cpu_to_be64(SPAPR_PCI_MEM_WIN_BUS_OFFSET
),
2124 cpu_to_be64(phb
->mem_win_addr
),
2125 cpu_to_be64(phb
->mem_win_size
),
2128 cpu_to_be32(b_ss(3)), cpu_to_be64(phb
->mem64_win_pciaddr
),
2129 cpu_to_be64(phb
->mem64_win_addr
),
2130 cpu_to_be64(phb
->mem64_win_size
),
2133 const unsigned sizeof_ranges
=
2134 (phb
->mem64_win_size
? 3 : 2) * sizeof(ranges
[0]);
2135 uint64_t bus_reg
[] = { cpu_to_be64(phb
->buid
), 0 };
2136 uint32_t interrupt_map_mask
[] = {
2137 cpu_to_be32(b_ddddd(-1)|b_fff(0)), 0x0, 0x0, cpu_to_be32(-1)};
2138 uint32_t interrupt_map
[PCI_SLOT_MAX
* PCI_NUM_PINS
][7];
2139 uint32_t ddw_applicable
[] = {
2140 cpu_to_be32(RTAS_IBM_QUERY_PE_DMA_WINDOW
),
2141 cpu_to_be32(RTAS_IBM_CREATE_PE_DMA_WINDOW
),
2142 cpu_to_be32(RTAS_IBM_REMOVE_PE_DMA_WINDOW
)
2144 uint32_t ddw_extensions
[] = {
2146 cpu_to_be32(RTAS_IBM_RESET_PE_DMA_WINDOW
)
2148 uint32_t associativity
[] = {cpu_to_be32(0x4),
2152 cpu_to_be32(phb
->numa_node
)};
2153 sPAPRTCETable
*tcet
;
2154 PCIBus
*bus
= PCI_HOST_BRIDGE(phb
)->bus
;
2157 /* Start populating the FDT */
2158 snprintf(nodename
, FDT_NAME_MAX
, "pci@%" PRIx64
, phb
->buid
);
2159 bus_off
= fdt_add_subnode(fdt
, 0, nodename
);
2164 /* Write PHB properties */
2165 _FDT(fdt_setprop_string(fdt
, bus_off
, "device_type", "pci"));
2166 _FDT(fdt_setprop_string(fdt
, bus_off
, "compatible", "IBM,Logical_PHB"));
2167 _FDT(fdt_setprop_cell(fdt
, bus_off
, "#address-cells", 0x3));
2168 _FDT(fdt_setprop_cell(fdt
, bus_off
, "#size-cells", 0x2));
2169 _FDT(fdt_setprop_cell(fdt
, bus_off
, "#interrupt-cells", 0x1));
2170 _FDT(fdt_setprop(fdt
, bus_off
, "used-by-rtas", NULL
, 0));
2171 _FDT(fdt_setprop(fdt
, bus_off
, "bus-range", &bus_range
, sizeof(bus_range
)));
2172 _FDT(fdt_setprop(fdt
, bus_off
, "ranges", &ranges
, sizeof_ranges
));
2173 _FDT(fdt_setprop(fdt
, bus_off
, "reg", &bus_reg
, sizeof(bus_reg
)));
2174 _FDT(fdt_setprop_cell(fdt
, bus_off
, "ibm,pci-config-space-type", 0x1));
2175 _FDT(fdt_setprop_cell(fdt
, bus_off
, "ibm,pe-total-#msi", XICS_IRQS_SPAPR
));
2177 /* Dynamic DMA window */
2178 if (phb
->ddw_enabled
) {
2179 _FDT(fdt_setprop(fdt
, bus_off
, "ibm,ddw-applicable", &ddw_applicable
,
2180 sizeof(ddw_applicable
)));
2181 _FDT(fdt_setprop(fdt
, bus_off
, "ibm,ddw-extensions",
2182 &ddw_extensions
, sizeof(ddw_extensions
)));
2185 /* Advertise NUMA via ibm,associativity */
2186 if (phb
->numa_node
!= -1) {
2187 _FDT(fdt_setprop(fdt
, bus_off
, "ibm,associativity", associativity
,
2188 sizeof(associativity
)));
2191 /* Build the interrupt-map, this must matches what is done
2192 * in pci_spapr_map_irq
2194 _FDT(fdt_setprop(fdt
, bus_off
, "interrupt-map-mask",
2195 &interrupt_map_mask
, sizeof(interrupt_map_mask
)));
2196 for (i
= 0; i
< PCI_SLOT_MAX
; i
++) {
2197 for (j
= 0; j
< PCI_NUM_PINS
; j
++) {
2198 uint32_t *irqmap
= interrupt_map
[i
*PCI_NUM_PINS
+ j
];
2199 int lsi_num
= pci_spapr_swizzle(i
, j
);
2201 irqmap
[0] = cpu_to_be32(b_ddddd(i
)|b_fff(0));
2204 irqmap
[3] = cpu_to_be32(j
+1);
2205 irqmap
[4] = cpu_to_be32(xics_phandle
);
2206 irqmap
[5] = cpu_to_be32(phb
->lsi_table
[lsi_num
].irq
);
2207 irqmap
[6] = cpu_to_be32(0x8);
2210 /* Write interrupt map */
2211 _FDT(fdt_setprop(fdt
, bus_off
, "interrupt-map", &interrupt_map
,
2212 sizeof(interrupt_map
)));
2214 tcet
= spapr_tce_find_by_liobn(phb
->dma_liobn
[0]);
2218 spapr_dma_dt(fdt
, bus_off
, "ibm,dma-window",
2219 tcet
->liobn
, tcet
->bus_offset
,
2220 tcet
->nb_table
<< tcet
->page_shift
);
2222 /* Walk the bridges and program the bus numbers*/
2223 spapr_phb_pci_enumerate(phb
);
2224 _FDT(fdt_setprop_cell(fdt
, bus_off
, "qemu,phb-enumerated", 0x1));
2226 /* Populate tree nodes with PCI devices attached */
2228 s_fdt
.node_off
= bus_off
;
2230 pci_for_each_device_reverse(bus
, pci_bus_num(bus
),
2231 spapr_populate_pci_devices_dt
,
2234 ret
= spapr_drc_populate_dt(fdt
, bus_off
, OBJECT(phb
),
2235 SPAPR_DR_CONNECTOR_TYPE_PCI
);
2243 void spapr_pci_rtas_init(void)
2245 spapr_rtas_register(RTAS_READ_PCI_CONFIG
, "read-pci-config",
2246 rtas_read_pci_config
);
2247 spapr_rtas_register(RTAS_WRITE_PCI_CONFIG
, "write-pci-config",
2248 rtas_write_pci_config
);
2249 spapr_rtas_register(RTAS_IBM_READ_PCI_CONFIG
, "ibm,read-pci-config",
2250 rtas_ibm_read_pci_config
);
2251 spapr_rtas_register(RTAS_IBM_WRITE_PCI_CONFIG
, "ibm,write-pci-config",
2252 rtas_ibm_write_pci_config
);
2253 if (msi_nonbroken
) {
2254 spapr_rtas_register(RTAS_IBM_QUERY_INTERRUPT_SOURCE_NUMBER
,
2255 "ibm,query-interrupt-source-number",
2256 rtas_ibm_query_interrupt_source_number
);
2257 spapr_rtas_register(RTAS_IBM_CHANGE_MSI
, "ibm,change-msi",
2258 rtas_ibm_change_msi
);
2261 spapr_rtas_register(RTAS_IBM_SET_EEH_OPTION
,
2262 "ibm,set-eeh-option",
2263 rtas_ibm_set_eeh_option
);
2264 spapr_rtas_register(RTAS_IBM_GET_CONFIG_ADDR_INFO2
,
2265 "ibm,get-config-addr-info2",
2266 rtas_ibm_get_config_addr_info2
);
2267 spapr_rtas_register(RTAS_IBM_READ_SLOT_RESET_STATE2
,
2268 "ibm,read-slot-reset-state2",
2269 rtas_ibm_read_slot_reset_state2
);
2270 spapr_rtas_register(RTAS_IBM_SET_SLOT_RESET
,
2271 "ibm,set-slot-reset",
2272 rtas_ibm_set_slot_reset
);
2273 spapr_rtas_register(RTAS_IBM_CONFIGURE_PE
,
2275 rtas_ibm_configure_pe
);
2276 spapr_rtas_register(RTAS_IBM_SLOT_ERROR_DETAIL
,
2277 "ibm,slot-error-detail",
2278 rtas_ibm_slot_error_detail
);
2281 static void spapr_pci_register_types(void)
2283 type_register_static(&spapr_phb_info
);
2286 type_init(spapr_pci_register_types
)
2288 static int spapr_switch_one_vga(DeviceState
*dev
, void *opaque
)
2290 bool be
= *(bool *)opaque
;
2292 if (object_dynamic_cast(OBJECT(dev
), "VGA")
2293 || object_dynamic_cast(OBJECT(dev
), "secondary-vga")) {
2294 object_property_set_bool(OBJECT(dev
), be
, "big-endian-framebuffer",
2300 void spapr_pci_switch_vga(bool big_endian
)
2302 sPAPRMachineState
*spapr
= SPAPR_MACHINE(qdev_get_machine());
2303 sPAPRPHBState
*sphb
;
2306 * For backward compatibility with existing guests, we switch
2307 * the endianness of the VGA controller when changing the guest
2310 QLIST_FOREACH(sphb
, &spapr
->phbs
, list
) {
2311 BusState
*bus
= &PCI_HOST_BRIDGE(sphb
)->bus
->qbus
;
2312 qbus_walk_children(bus
, spapr_switch_one_vga
, NULL
, NULL
, NULL
,