cachepc-qemu

Fork of AMDESE/qemu with changes for cachepc side-channel attack
git clone https://git.sinitax.com/sinitax/cachepc-qemu
Log | Files | Refs | Submodules | LICENSE | sfeed.txt

spapr_pci.c (81716B)


      1/*
      2 * QEMU sPAPR PCI host originated from Uninorth PCI host
      3 *
      4 * Copyright (c) 2011 Alexey Kardashevskiy, IBM Corporation.
      5 * Copyright (C) 2011 David Gibson, IBM Corporation.
      6 *
      7 * Permission is hereby granted, free of charge, to any person obtaining a copy
      8 * of this software and associated documentation files (the "Software"), to deal
      9 * in the Software without restriction, including without limitation the rights
     10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
     11 * copies of the Software, and to permit persons to whom the Software is
     12 * furnished to do so, subject to the following conditions:
     13 *
     14 * The above copyright notice and this permission notice shall be included in
     15 * all copies or substantial portions of the Software.
     16 *
     17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
     18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
     19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
     20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
     21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
     22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
     23 * THE SOFTWARE.
     24 */
     25
     26#include "qemu/osdep.h"
     27#include "qapi/error.h"
     28#include "hw/irq.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"
     38#include <libfdt.h>
     39#include "trace.h"
     40#include "qemu/error-report.h"
     41#include "qemu/module.h"
     42#include "qapi/qmp/qerror.h"
     43#include "hw/ppc/fdt.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 "hw/qdev-properties.h"
     49#include "sysemu/device_tree.h"
     50#include "sysemu/kvm.h"
     51#include "sysemu/hostmem.h"
     52#include "sysemu/numa.h"
     53#include "hw/ppc/spapr_numa.h"
     54#include "qemu/log.h"
     55
     56/* Copied from the kernel arch/powerpc/platforms/pseries/msi.c */
     57#define RTAS_QUERY_FN           0
     58#define RTAS_CHANGE_FN          1
     59#define RTAS_RESET_FN           2
     60#define RTAS_CHANGE_MSI_FN      3
     61#define RTAS_CHANGE_MSIX_FN     4
     62
     63/* Interrupt types to return on RTAS_CHANGE_* */
     64#define RTAS_TYPE_MSI           1
     65#define RTAS_TYPE_MSIX          2
     66
     67SpaprPhbState *spapr_pci_find_phb(SpaprMachineState *spapr, uint64_t buid)
     68{
     69    SpaprPhbState *sphb;
     70
     71    QLIST_FOREACH(sphb, &spapr->phbs, list) {
     72        if (sphb->buid != buid) {
     73            continue;
     74        }
     75        return sphb;
     76    }
     77
     78    return NULL;
     79}
     80
     81PCIDevice *spapr_pci_find_dev(SpaprMachineState *spapr, uint64_t buid,
     82                              uint32_t config_addr)
     83{
     84    SpaprPhbState *sphb = spapr_pci_find_phb(spapr, buid);
     85    PCIHostState *phb = PCI_HOST_BRIDGE(sphb);
     86    int bus_num = (config_addr >> 16) & 0xFF;
     87    int devfn = (config_addr >> 8) & 0xFF;
     88
     89    if (!phb) {
     90        return NULL;
     91    }
     92
     93    return pci_find_device(phb->bus, bus_num, devfn);
     94}
     95
     96static uint32_t rtas_pci_cfgaddr(uint32_t arg)
     97{
     98    /* This handles the encoding of extended config space addresses */
     99    return ((arg >> 20) & 0xf00) | (arg & 0xff);
    100}
    101
    102static void finish_read_pci_config(SpaprMachineState *spapr, uint64_t buid,
    103                                   uint32_t addr, uint32_t size,
    104                                   target_ulong rets)
    105{
    106    PCIDevice *pci_dev;
    107    uint32_t val;
    108
    109    if ((size != 1) && (size != 2) && (size != 4)) {
    110        /* access must be 1, 2 or 4 bytes */
    111        rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
    112        return;
    113    }
    114
    115    pci_dev = spapr_pci_find_dev(spapr, buid, addr);
    116    addr = rtas_pci_cfgaddr(addr);
    117
    118    if (!pci_dev || (addr % size) || (addr >= pci_config_size(pci_dev))) {
    119        /* Access must be to a valid device, within bounds and
    120         * naturally aligned */
    121        rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
    122        return;
    123    }
    124
    125    val = pci_host_config_read_common(pci_dev, addr,
    126                                      pci_config_size(pci_dev), size);
    127
    128    rtas_st(rets, 0, RTAS_OUT_SUCCESS);
    129    rtas_st(rets, 1, val);
    130}
    131
    132static void rtas_ibm_read_pci_config(PowerPCCPU *cpu, SpaprMachineState *spapr,
    133                                     uint32_t token, uint32_t nargs,
    134                                     target_ulong args,
    135                                     uint32_t nret, target_ulong rets)
    136{
    137    uint64_t buid;
    138    uint32_t size, addr;
    139
    140    if ((nargs != 4) || (nret != 2)) {
    141        rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
    142        return;
    143    }
    144
    145    buid = rtas_ldq(args, 1);
    146    size = rtas_ld(args, 3);
    147    addr = rtas_ld(args, 0);
    148
    149    finish_read_pci_config(spapr, buid, addr, size, rets);
    150}
    151
    152static void rtas_read_pci_config(PowerPCCPU *cpu, SpaprMachineState *spapr,
    153                                 uint32_t token, uint32_t nargs,
    154                                 target_ulong args,
    155                                 uint32_t nret, target_ulong rets)
    156{
    157    uint32_t size, addr;
    158
    159    if ((nargs != 2) || (nret != 2)) {
    160        rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
    161        return;
    162    }
    163
    164    size = rtas_ld(args, 1);
    165    addr = rtas_ld(args, 0);
    166
    167    finish_read_pci_config(spapr, 0, addr, size, rets);
    168}
    169
    170static void finish_write_pci_config(SpaprMachineState *spapr, uint64_t buid,
    171                                    uint32_t addr, uint32_t size,
    172                                    uint32_t val, target_ulong rets)
    173{
    174    PCIDevice *pci_dev;
    175
    176    if ((size != 1) && (size != 2) && (size != 4)) {
    177        /* access must be 1, 2 or 4 bytes */
    178        rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
    179        return;
    180    }
    181
    182    pci_dev = spapr_pci_find_dev(spapr, buid, addr);
    183    addr = rtas_pci_cfgaddr(addr);
    184
    185    if (!pci_dev || (addr % size) || (addr >= pci_config_size(pci_dev))) {
    186        /* Access must be to a valid device, within bounds and
    187         * naturally aligned */
    188        rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
    189        return;
    190    }
    191
    192    pci_host_config_write_common(pci_dev, addr, pci_config_size(pci_dev),
    193                                 val, size);
    194
    195    rtas_st(rets, 0, RTAS_OUT_SUCCESS);
    196}
    197
    198static void rtas_ibm_write_pci_config(PowerPCCPU *cpu, SpaprMachineState *spapr,
    199                                      uint32_t token, uint32_t nargs,
    200                                      target_ulong args,
    201                                      uint32_t nret, target_ulong rets)
    202{
    203    uint64_t buid;
    204    uint32_t val, size, addr;
    205
    206    if ((nargs != 5) || (nret != 1)) {
    207        rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
    208        return;
    209    }
    210
    211    buid = rtas_ldq(args, 1);
    212    val = rtas_ld(args, 4);
    213    size = rtas_ld(args, 3);
    214    addr = rtas_ld(args, 0);
    215
    216    finish_write_pci_config(spapr, buid, addr, size, val, rets);
    217}
    218
    219static void rtas_write_pci_config(PowerPCCPU *cpu, SpaprMachineState *spapr,
    220                                  uint32_t token, uint32_t nargs,
    221                                  target_ulong args,
    222                                  uint32_t nret, target_ulong rets)
    223{
    224    uint32_t val, size, addr;
    225
    226    if ((nargs != 3) || (nret != 1)) {
    227        rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
    228        return;
    229    }
    230
    231
    232    val = rtas_ld(args, 2);
    233    size = rtas_ld(args, 1);
    234    addr = rtas_ld(args, 0);
    235
    236    finish_write_pci_config(spapr, 0, addr, size, val, rets);
    237}
    238
    239/*
    240 * Set MSI/MSIX message data.
    241 * This is required for msi_notify()/msix_notify() which
    242 * will write at the addresses via spapr_msi_write().
    243 *
    244 * If hwaddr == 0, all entries will have .data == first_irq i.e.
    245 * table will be reset.
    246 */
    247static void spapr_msi_setmsg(PCIDevice *pdev, hwaddr addr, bool msix,
    248                             unsigned first_irq, unsigned req_num)
    249{
    250    unsigned i;
    251    MSIMessage msg = { .address = addr, .data = first_irq };
    252
    253    if (!msix) {
    254        msi_set_message(pdev, msg);
    255        trace_spapr_pci_msi_setup(pdev->name, 0, msg.address);
    256        return;
    257    }
    258
    259    for (i = 0; i < req_num; ++i) {
    260        msix_set_message(pdev, i, msg);
    261        trace_spapr_pci_msi_setup(pdev->name, i, msg.address);
    262        if (addr) {
    263            ++msg.data;
    264        }
    265    }
    266}
    267
    268static void rtas_ibm_change_msi(PowerPCCPU *cpu, SpaprMachineState *spapr,
    269                                uint32_t token, uint32_t nargs,
    270                                target_ulong args, uint32_t nret,
    271                                target_ulong rets)
    272{
    273    SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
    274    uint32_t config_addr = rtas_ld(args, 0);
    275    uint64_t buid = rtas_ldq(args, 1);
    276    unsigned int func = rtas_ld(args, 3);
    277    unsigned int req_num = rtas_ld(args, 4); /* 0 == remove all */
    278    unsigned int seq_num = rtas_ld(args, 5);
    279    unsigned int ret_intr_type;
    280    unsigned int irq, max_irqs = 0;
    281    SpaprPhbState *phb = NULL;
    282    PCIDevice *pdev = NULL;
    283    SpaprPciMsi *msi;
    284    int *config_addr_key;
    285    Error *err = NULL;
    286    int i;
    287
    288    /* Fins SpaprPhbState */
    289    phb = spapr_pci_find_phb(spapr, buid);
    290    if (phb) {
    291        pdev = spapr_pci_find_dev(spapr, buid, config_addr);
    292    }
    293    if (!phb || !pdev) {
    294        rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
    295        return;
    296    }
    297
    298    switch (func) {
    299    case RTAS_CHANGE_FN:
    300        if (msi_present(pdev)) {
    301            ret_intr_type = RTAS_TYPE_MSI;
    302        } else if (msix_present(pdev)) {
    303            ret_intr_type = RTAS_TYPE_MSIX;
    304        } else {
    305            rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
    306            return;
    307        }
    308        break;
    309    case RTAS_CHANGE_MSI_FN:
    310        if (msi_present(pdev)) {
    311            ret_intr_type = RTAS_TYPE_MSI;
    312        } else {
    313            rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
    314            return;
    315        }
    316        break;
    317    case RTAS_CHANGE_MSIX_FN:
    318        if (msix_present(pdev)) {
    319            ret_intr_type = RTAS_TYPE_MSIX;
    320        } else {
    321            rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
    322            return;
    323        }
    324        break;
    325    default:
    326        error_report("rtas_ibm_change_msi(%u) is not implemented", func);
    327        rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
    328        return;
    329    }
    330
    331    msi = (SpaprPciMsi *) g_hash_table_lookup(phb->msi, &config_addr);
    332
    333    /* Releasing MSIs */
    334    if (!req_num) {
    335        if (!msi) {
    336            trace_spapr_pci_msi("Releasing wrong config", config_addr);
    337            rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
    338            return;
    339        }
    340
    341        if (msi_present(pdev)) {
    342            spapr_msi_setmsg(pdev, 0, false, 0, 0);
    343        }
    344        if (msix_present(pdev)) {
    345            spapr_msi_setmsg(pdev, 0, true, 0, 0);
    346        }
    347        g_hash_table_remove(phb->msi, &config_addr);
    348
    349        trace_spapr_pci_msi("Released MSIs", config_addr);
    350        rtas_st(rets, 0, RTAS_OUT_SUCCESS);
    351        rtas_st(rets, 1, 0);
    352        return;
    353    }
    354
    355    /* Enabling MSI */
    356
    357    /* Check if the device supports as many IRQs as requested */
    358    if (ret_intr_type == RTAS_TYPE_MSI) {
    359        max_irqs = msi_nr_vectors_allocated(pdev);
    360    } else if (ret_intr_type == RTAS_TYPE_MSIX) {
    361        max_irqs = pdev->msix_entries_nr;
    362    }
    363    if (!max_irqs) {
    364        error_report("Requested interrupt type %d is not enabled for device %x",
    365                     ret_intr_type, config_addr);
    366        rtas_st(rets, 0, -1); /* Hardware error */
    367        return;
    368    }
    369    /* Correct the number if the guest asked for too many */
    370    if (req_num > max_irqs) {
    371        trace_spapr_pci_msi_retry(config_addr, req_num, max_irqs);
    372        req_num = max_irqs;
    373        irq = 0; /* to avoid misleading trace */
    374        goto out;
    375    }
    376
    377    /* Allocate MSIs */
    378    if (smc->legacy_irq_allocation) {
    379        irq = spapr_irq_find(spapr, req_num, ret_intr_type == RTAS_TYPE_MSI,
    380                             &err);
    381    } else {
    382        irq = spapr_irq_msi_alloc(spapr, req_num,
    383                                  ret_intr_type == RTAS_TYPE_MSI, &err);
    384    }
    385    if (err) {
    386        error_reportf_err(err, "Can't allocate MSIs for device %x: ",
    387                          config_addr);
    388        rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
    389        return;
    390    }
    391
    392    for (i = 0; i < req_num; i++) {
    393        spapr_irq_claim(spapr, irq + i, false, &err);
    394        if (err) {
    395            if (i) {
    396                spapr_irq_free(spapr, irq, i);
    397            }
    398            if (!smc->legacy_irq_allocation) {
    399                spapr_irq_msi_free(spapr, irq, req_num);
    400            }
    401            error_reportf_err(err, "Can't allocate MSIs for device %x: ",
    402                              config_addr);
    403            rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
    404            return;
    405        }
    406    }
    407
    408    /* Release previous MSIs */
    409    if (msi) {
    410        g_hash_table_remove(phb->msi, &config_addr);
    411    }
    412
    413    /* Setup MSI/MSIX vectors in the device (via cfgspace or MSIX BAR) */
    414    spapr_msi_setmsg(pdev, SPAPR_PCI_MSI_WINDOW, ret_intr_type == RTAS_TYPE_MSIX,
    415                     irq, req_num);
    416
    417    /* Add MSI device to cache */
    418    msi = g_new(SpaprPciMsi, 1);
    419    msi->first_irq = irq;
    420    msi->num = req_num;
    421    config_addr_key = g_new(int, 1);
    422    *config_addr_key = config_addr;
    423    g_hash_table_insert(phb->msi, config_addr_key, msi);
    424
    425out:
    426    rtas_st(rets, 0, RTAS_OUT_SUCCESS);
    427    rtas_st(rets, 1, req_num);
    428    rtas_st(rets, 2, ++seq_num);
    429    if (nret > 3) {
    430        rtas_st(rets, 3, ret_intr_type);
    431    }
    432
    433    trace_spapr_pci_rtas_ibm_change_msi(config_addr, func, req_num, irq);
    434}
    435
    436static void rtas_ibm_query_interrupt_source_number(PowerPCCPU *cpu,
    437                                                   SpaprMachineState *spapr,
    438                                                   uint32_t token,
    439                                                   uint32_t nargs,
    440                                                   target_ulong args,
    441                                                   uint32_t nret,
    442                                                   target_ulong rets)
    443{
    444    uint32_t config_addr = rtas_ld(args, 0);
    445    uint64_t buid = rtas_ldq(args, 1);
    446    unsigned int intr_src_num = -1, ioa_intr_num = rtas_ld(args, 3);
    447    SpaprPhbState *phb = NULL;
    448    PCIDevice *pdev = NULL;
    449    SpaprPciMsi *msi;
    450
    451    /* Find SpaprPhbState */
    452    phb = spapr_pci_find_phb(spapr, buid);
    453    if (phb) {
    454        pdev = spapr_pci_find_dev(spapr, buid, config_addr);
    455    }
    456    if (!phb || !pdev) {
    457        rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
    458        return;
    459    }
    460
    461    /* Find device descriptor and start IRQ */
    462    msi = (SpaprPciMsi *) g_hash_table_lookup(phb->msi, &config_addr);
    463    if (!msi || !msi->first_irq || !msi->num || (ioa_intr_num >= msi->num)) {
    464        trace_spapr_pci_msi("Failed to return vector", config_addr);
    465        rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
    466        return;
    467    }
    468    intr_src_num = msi->first_irq + ioa_intr_num;
    469    trace_spapr_pci_rtas_ibm_query_interrupt_source_number(ioa_intr_num,
    470                                                           intr_src_num);
    471
    472    rtas_st(rets, 0, RTAS_OUT_SUCCESS);
    473    rtas_st(rets, 1, intr_src_num);
    474    rtas_st(rets, 2, 1);/* 0 == level; 1 == edge */
    475}
    476
    477static void rtas_ibm_set_eeh_option(PowerPCCPU *cpu,
    478                                    SpaprMachineState *spapr,
    479                                    uint32_t token, uint32_t nargs,
    480                                    target_ulong args, uint32_t nret,
    481                                    target_ulong rets)
    482{
    483    SpaprPhbState *sphb;
    484    uint32_t addr, option;
    485    uint64_t buid;
    486    int ret;
    487
    488    if ((nargs != 4) || (nret != 1)) {
    489        goto param_error_exit;
    490    }
    491
    492    buid = rtas_ldq(args, 1);
    493    addr = rtas_ld(args, 0);
    494    option = rtas_ld(args, 3);
    495
    496    sphb = spapr_pci_find_phb(spapr, buid);
    497    if (!sphb) {
    498        goto param_error_exit;
    499    }
    500
    501    if (!spapr_phb_eeh_available(sphb)) {
    502        goto param_error_exit;
    503    }
    504
    505    ret = spapr_phb_vfio_eeh_set_option(sphb, addr, option);
    506    rtas_st(rets, 0, ret);
    507    return;
    508
    509param_error_exit:
    510    rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
    511}
    512
    513static void rtas_ibm_get_config_addr_info2(PowerPCCPU *cpu,
    514                                           SpaprMachineState *spapr,
    515                                           uint32_t token, uint32_t nargs,
    516                                           target_ulong args, uint32_t nret,
    517                                           target_ulong rets)
    518{
    519    SpaprPhbState *sphb;
    520    PCIDevice *pdev;
    521    uint32_t addr, option;
    522    uint64_t buid;
    523
    524    if ((nargs != 4) || (nret != 2)) {
    525        goto param_error_exit;
    526    }
    527
    528    buid = rtas_ldq(args, 1);
    529    sphb = spapr_pci_find_phb(spapr, buid);
    530    if (!sphb) {
    531        goto param_error_exit;
    532    }
    533
    534    if (!spapr_phb_eeh_available(sphb)) {
    535        goto param_error_exit;
    536    }
    537
    538    /*
    539     * We always have PE address of form "00BB0001". "BB"
    540     * represents the bus number of PE's primary bus.
    541     */
    542    option = rtas_ld(args, 3);
    543    switch (option) {
    544    case RTAS_GET_PE_ADDR:
    545        addr = rtas_ld(args, 0);
    546        pdev = spapr_pci_find_dev(spapr, buid, addr);
    547        if (!pdev) {
    548            goto param_error_exit;
    549        }
    550
    551        rtas_st(rets, 1, (pci_bus_num(pci_get_bus(pdev)) << 16) + 1);
    552        break;
    553    case RTAS_GET_PE_MODE:
    554        rtas_st(rets, 1, RTAS_PE_MODE_SHARED);
    555        break;
    556    default:
    557        goto param_error_exit;
    558    }
    559
    560    rtas_st(rets, 0, RTAS_OUT_SUCCESS);
    561    return;
    562
    563param_error_exit:
    564    rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
    565}
    566
    567static void rtas_ibm_read_slot_reset_state2(PowerPCCPU *cpu,
    568                                            SpaprMachineState *spapr,
    569                                            uint32_t token, uint32_t nargs,
    570                                            target_ulong args, uint32_t nret,
    571                                            target_ulong rets)
    572{
    573    SpaprPhbState *sphb;
    574    uint64_t buid;
    575    int state, ret;
    576
    577    if ((nargs != 3) || (nret != 4 && nret != 5)) {
    578        goto param_error_exit;
    579    }
    580
    581    buid = rtas_ldq(args, 1);
    582    sphb = spapr_pci_find_phb(spapr, buid);
    583    if (!sphb) {
    584        goto param_error_exit;
    585    }
    586
    587    if (!spapr_phb_eeh_available(sphb)) {
    588        goto param_error_exit;
    589    }
    590
    591    ret = spapr_phb_vfio_eeh_get_state(sphb, &state);
    592    rtas_st(rets, 0, ret);
    593    if (ret != RTAS_OUT_SUCCESS) {
    594        return;
    595    }
    596
    597    rtas_st(rets, 1, state);
    598    rtas_st(rets, 2, RTAS_EEH_SUPPORT);
    599    rtas_st(rets, 3, RTAS_EEH_PE_UNAVAIL_INFO);
    600    if (nret >= 5) {
    601        rtas_st(rets, 4, RTAS_EEH_PE_RECOVER_INFO);
    602    }
    603    return;
    604
    605param_error_exit:
    606    rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
    607}
    608
    609static void rtas_ibm_set_slot_reset(PowerPCCPU *cpu,
    610                                    SpaprMachineState *spapr,
    611                                    uint32_t token, uint32_t nargs,
    612                                    target_ulong args, uint32_t nret,
    613                                    target_ulong rets)
    614{
    615    SpaprPhbState *sphb;
    616    uint32_t option;
    617    uint64_t buid;
    618    int ret;
    619
    620    if ((nargs != 4) || (nret != 1)) {
    621        goto param_error_exit;
    622    }
    623
    624    buid = rtas_ldq(args, 1);
    625    option = rtas_ld(args, 3);
    626    sphb = spapr_pci_find_phb(spapr, buid);
    627    if (!sphb) {
    628        goto param_error_exit;
    629    }
    630
    631    if (!spapr_phb_eeh_available(sphb)) {
    632        goto param_error_exit;
    633    }
    634
    635    ret = spapr_phb_vfio_eeh_reset(sphb, option);
    636    rtas_st(rets, 0, ret);
    637    return;
    638
    639param_error_exit:
    640    rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
    641}
    642
    643static void rtas_ibm_configure_pe(PowerPCCPU *cpu,
    644                                  SpaprMachineState *spapr,
    645                                  uint32_t token, uint32_t nargs,
    646                                  target_ulong args, uint32_t nret,
    647                                  target_ulong rets)
    648{
    649    SpaprPhbState *sphb;
    650    uint64_t buid;
    651    int ret;
    652
    653    if ((nargs != 3) || (nret != 1)) {
    654        goto param_error_exit;
    655    }
    656
    657    buid = rtas_ldq(args, 1);
    658    sphb = spapr_pci_find_phb(spapr, buid);
    659    if (!sphb) {
    660        goto param_error_exit;
    661    }
    662
    663    if (!spapr_phb_eeh_available(sphb)) {
    664        goto param_error_exit;
    665    }
    666
    667    ret = spapr_phb_vfio_eeh_configure(sphb);
    668    rtas_st(rets, 0, ret);
    669    return;
    670
    671param_error_exit:
    672    rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
    673}
    674
    675/* To support it later */
    676static void rtas_ibm_slot_error_detail(PowerPCCPU *cpu,
    677                                       SpaprMachineState *spapr,
    678                                       uint32_t token, uint32_t nargs,
    679                                       target_ulong args, uint32_t nret,
    680                                       target_ulong rets)
    681{
    682    SpaprPhbState *sphb;
    683    int option;
    684    uint64_t buid;
    685
    686    if ((nargs != 8) || (nret != 1)) {
    687        goto param_error_exit;
    688    }
    689
    690    buid = rtas_ldq(args, 1);
    691    sphb = spapr_pci_find_phb(spapr, buid);
    692    if (!sphb) {
    693        goto param_error_exit;
    694    }
    695
    696    if (!spapr_phb_eeh_available(sphb)) {
    697        goto param_error_exit;
    698    }
    699
    700    option = rtas_ld(args, 7);
    701    switch (option) {
    702    case RTAS_SLOT_TEMP_ERR_LOG:
    703    case RTAS_SLOT_PERM_ERR_LOG:
    704        break;
    705    default:
    706        goto param_error_exit;
    707    }
    708
    709    /* We don't have error log yet */
    710    rtas_st(rets, 0, RTAS_OUT_NO_ERRORS_FOUND);
    711    return;
    712
    713param_error_exit:
    714    rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
    715}
    716
    717static void pci_spapr_set_irq(void *opaque, int irq_num, int level)
    718{
    719    /*
    720     * Here we use the number returned by pci_swizzle_map_irq_fn to find a
    721     * corresponding qemu_irq.
    722     */
    723    SpaprPhbState *phb = opaque;
    724    SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
    725
    726    trace_spapr_pci_lsi_set(phb->dtbusname, irq_num, phb->lsi_table[irq_num].irq);
    727    qemu_set_irq(spapr_qirq(spapr, phb->lsi_table[irq_num].irq), level);
    728}
    729
    730static PCIINTxRoute spapr_route_intx_pin_to_irq(void *opaque, int pin)
    731{
    732    SpaprPhbState *sphb = SPAPR_PCI_HOST_BRIDGE(opaque);
    733    PCIINTxRoute route;
    734
    735    route.mode = PCI_INTX_ENABLED;
    736    route.irq = sphb->lsi_table[pin].irq;
    737
    738    return route;
    739}
    740
    741static uint64_t spapr_msi_read(void *opaque, hwaddr addr, unsigned size)
    742{
    743    qemu_log_mask(LOG_GUEST_ERROR, "%s: invalid access\n", __func__);
    744    return 0;
    745}
    746
    747/*
    748 * MSI/MSIX memory region implementation.
    749 * The handler handles both MSI and MSIX.
    750 * The vector number is encoded in least bits in data.
    751 */
    752static void spapr_msi_write(void *opaque, hwaddr addr,
    753                            uint64_t data, unsigned size)
    754{
    755    SpaprMachineState *spapr = opaque;
    756    uint32_t irq = data;
    757
    758    trace_spapr_pci_msi_write(addr, data, irq);
    759
    760    qemu_irq_pulse(spapr_qirq(spapr, irq));
    761}
    762
    763static const MemoryRegionOps spapr_msi_ops = {
    764    /*
    765     * .read result is undefined by PCI spec.
    766     * define .read method to avoid assert failure in memory_region_init_io
    767     */
    768    .read = spapr_msi_read,
    769    .write = spapr_msi_write,
    770    .endianness = DEVICE_LITTLE_ENDIAN
    771};
    772
    773/*
    774 * PHB PCI device
    775 */
    776static AddressSpace *spapr_pci_dma_iommu(PCIBus *bus, void *opaque, int devfn)
    777{
    778    SpaprPhbState *phb = opaque;
    779
    780    return &phb->iommu_as;
    781}
    782
    783static char *spapr_phb_vfio_get_loc_code(SpaprPhbState *sphb,  PCIDevice *pdev)
    784{
    785    g_autofree char *path = NULL;
    786    g_autofree char *host = NULL;
    787    g_autofree char *devspec = NULL;
    788    char *buf = NULL;
    789
    790    /* Get the PCI VFIO host id */
    791    host = object_property_get_str(OBJECT(pdev), "host", NULL);
    792    if (!host) {
    793        return NULL;
    794    }
    795
    796    /* Construct the path of the file that will give us the DT location */
    797    path = g_strdup_printf("/sys/bus/pci/devices/%s/devspec", host);
    798    if (!g_file_get_contents(path, &devspec, NULL, NULL)) {
    799        return NULL;
    800    }
    801
    802    /* Construct and read from host device tree the loc-code */
    803    path = g_strdup_printf("/proc/device-tree%s/ibm,loc-code", devspec);
    804    if (!g_file_get_contents(path, &buf, NULL, NULL)) {
    805        return NULL;
    806    }
    807    return buf;
    808}
    809
    810static char *spapr_phb_get_loc_code(SpaprPhbState *sphb, PCIDevice *pdev)
    811{
    812    char *buf;
    813    const char *devtype = "qemu";
    814    uint32_t busnr = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(pdev))));
    815
    816    if (object_dynamic_cast(OBJECT(pdev), "vfio-pci")) {
    817        buf = spapr_phb_vfio_get_loc_code(sphb, pdev);
    818        if (buf) {
    819            return buf;
    820        }
    821        devtype = "vfio";
    822    }
    823    /*
    824     * For emulated devices and VFIO-failure case, make up
    825     * the loc-code.
    826     */
    827    buf = g_strdup_printf("%s_%s:%04x:%02x:%02x.%x",
    828                          devtype, pdev->name, sphb->index, busnr,
    829                          PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
    830    return buf;
    831}
    832
    833/* Macros to operate with address in OF binding to PCI */
    834#define b_x(x, p, l)    (((x) & ((1<<(l))-1)) << (p))
    835#define b_n(x)          b_x((x), 31, 1) /* 0 if relocatable */
    836#define b_p(x)          b_x((x), 30, 1) /* 1 if prefetchable */
    837#define b_t(x)          b_x((x), 29, 1) /* 1 if the address is aliased */
    838#define b_ss(x)         b_x((x), 24, 2) /* the space code */
    839#define b_bbbbbbbb(x)   b_x((x), 16, 8) /* bus number */
    840#define b_ddddd(x)      b_x((x), 11, 5) /* device number */
    841#define b_fff(x)        b_x((x), 8, 3)  /* function number */
    842#define b_rrrrrrrr(x)   b_x((x), 0, 8)  /* register number */
    843
    844/* for 'reg' OF properties */
    845#define RESOURCE_CELLS_SIZE 2
    846#define RESOURCE_CELLS_ADDRESS 3
    847
    848typedef struct ResourceFields {
    849    uint32_t phys_hi;
    850    uint32_t phys_mid;
    851    uint32_t phys_lo;
    852    uint32_t size_hi;
    853    uint32_t size_lo;
    854} QEMU_PACKED ResourceFields;
    855
    856typedef struct ResourceProps {
    857    ResourceFields reg[8];
    858    uint32_t reg_len;
    859} ResourceProps;
    860
    861/* fill in the 'reg' OF properties for
    862 * a PCI device. 'reg' describes resource requirements for a
    863 * device's IO/MEM regions.
    864 *
    865 * the property is an array of ('phys-addr', 'size') pairs describing
    866 * the addressable regions of the PCI device, where 'phys-addr' is a
    867 * RESOURCE_CELLS_ADDRESS-tuple of 32-bit integers corresponding to
    868 * (phys.hi, phys.mid, phys.lo), and 'size' is a
    869 * RESOURCE_CELLS_SIZE-tuple corresponding to (size.hi, size.lo).
    870 *
    871 * phys.hi = 0xYYXXXXZZ, where:
    872 *   0xYY = npt000ss
    873 *          |||   |
    874 *          |||   +-- space code
    875 *          |||               |
    876 *          |||               +  00 if configuration space
    877 *          |||               +  01 if IO region,
    878 *          |||               +  10 if 32-bit MEM region
    879 *          |||               +  11 if 64-bit MEM region
    880 *          |||
    881 *          ||+------ for non-relocatable IO: 1 if aliased
    882 *          ||        for relocatable IO: 1 if below 64KB
    883 *          ||        for MEM: 1 if below 1MB
    884 *          |+------- 1 if region is prefetchable
    885 *          +-------- 1 if region is non-relocatable
    886 *   0xXXXX = bbbbbbbb dddddfff, encoding bus, slot, and function
    887 *            bits respectively
    888 *   0xZZ = rrrrrrrr, the register number of the BAR corresponding
    889 *          to the region
    890 *
    891 * phys.mid and phys.lo correspond respectively to the hi/lo portions
    892 * of the actual address of the region.
    893 *
    894 * note also that addresses defined in this property are, at least
    895 * for PAPR guests, relative to the PHBs IO/MEM windows, and
    896 * correspond directly to the addresses in the BARs.
    897 *
    898 * in accordance with PCI Bus Binding to Open Firmware,
    899 * IEEE Std 1275-1994, section 4.1.1, as implemented by PAPR+ v2.7,
    900 * Appendix C.
    901 */
    902static void populate_resource_props(PCIDevice *d, ResourceProps *rp)
    903{
    904    int bus_num = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(d))));
    905    uint32_t dev_id = (b_bbbbbbbb(bus_num) |
    906                       b_ddddd(PCI_SLOT(d->devfn)) |
    907                       b_fff(PCI_FUNC(d->devfn)));
    908    ResourceFields *reg;
    909    int i, reg_idx = 0;
    910
    911    /* config space region */
    912    reg = &rp->reg[reg_idx++];
    913    reg->phys_hi = cpu_to_be32(dev_id);
    914    reg->phys_mid = 0;
    915    reg->phys_lo = 0;
    916    reg->size_hi = 0;
    917    reg->size_lo = 0;
    918
    919    for (i = 0; i < PCI_NUM_REGIONS; i++) {
    920        if (!d->io_regions[i].size) {
    921            continue;
    922        }
    923
    924        reg = &rp->reg[reg_idx++];
    925
    926        reg->phys_hi = cpu_to_be32(dev_id | b_rrrrrrrr(pci_bar(d, i)));
    927        if (d->io_regions[i].type & PCI_BASE_ADDRESS_SPACE_IO) {
    928            reg->phys_hi |= cpu_to_be32(b_ss(1));
    929        } else if (d->io_regions[i].type & PCI_BASE_ADDRESS_MEM_TYPE_64) {
    930            reg->phys_hi |= cpu_to_be32(b_ss(3));
    931        } else {
    932            reg->phys_hi |= cpu_to_be32(b_ss(2));
    933        }
    934        reg->phys_mid = 0;
    935        reg->phys_lo = 0;
    936        reg->size_hi = cpu_to_be32(d->io_regions[i].size >> 32);
    937        reg->size_lo = cpu_to_be32(d->io_regions[i].size);
    938    }
    939
    940    rp->reg_len = reg_idx * sizeof(ResourceFields);
    941}
    942
    943typedef struct PCIClass PCIClass;
    944typedef struct PCISubClass PCISubClass;
    945typedef struct PCIIFace PCIIFace;
    946
    947struct PCIIFace {
    948    int iface;
    949    const char *name;
    950};
    951
    952struct PCISubClass {
    953    int subclass;
    954    const char *name;
    955    const PCIIFace *iface;
    956};
    957
    958struct PCIClass {
    959    const char *name;
    960    const PCISubClass *subc;
    961};
    962
    963static const PCISubClass undef_subclass[] = {
    964    { PCI_CLASS_NOT_DEFINED_VGA, "display", NULL },
    965    { 0xFF, NULL, NULL },
    966};
    967
    968static const PCISubClass mass_subclass[] = {
    969    { PCI_CLASS_STORAGE_SCSI, "scsi", NULL },
    970    { PCI_CLASS_STORAGE_IDE, "ide", NULL },
    971    { PCI_CLASS_STORAGE_FLOPPY, "fdc", NULL },
    972    { PCI_CLASS_STORAGE_IPI, "ipi", NULL },
    973    { PCI_CLASS_STORAGE_RAID, "raid", NULL },
    974    { PCI_CLASS_STORAGE_ATA, "ata", NULL },
    975    { PCI_CLASS_STORAGE_SATA, "sata", NULL },
    976    { PCI_CLASS_STORAGE_SAS, "sas", NULL },
    977    { 0xFF, NULL, NULL },
    978};
    979
    980static const PCISubClass net_subclass[] = {
    981    { PCI_CLASS_NETWORK_ETHERNET, "ethernet", NULL },
    982    { PCI_CLASS_NETWORK_TOKEN_RING, "token-ring", NULL },
    983    { PCI_CLASS_NETWORK_FDDI, "fddi", NULL },
    984    { PCI_CLASS_NETWORK_ATM, "atm", NULL },
    985    { PCI_CLASS_NETWORK_ISDN, "isdn", NULL },
    986    { PCI_CLASS_NETWORK_WORLDFIP, "worldfip", NULL },
    987    { PCI_CLASS_NETWORK_PICMG214, "picmg", NULL },
    988    { 0xFF, NULL, NULL },
    989};
    990
    991static const PCISubClass displ_subclass[] = {
    992    { PCI_CLASS_DISPLAY_VGA, "vga", NULL },
    993    { PCI_CLASS_DISPLAY_XGA, "xga", NULL },
    994    { PCI_CLASS_DISPLAY_3D, "3d-controller", NULL },
    995    { 0xFF, NULL, NULL },
    996};
    997
    998static const PCISubClass media_subclass[] = {
    999    { PCI_CLASS_MULTIMEDIA_VIDEO, "video", NULL },
   1000    { PCI_CLASS_MULTIMEDIA_AUDIO, "sound", NULL },
   1001    { PCI_CLASS_MULTIMEDIA_PHONE, "telephony", NULL },
   1002    { 0xFF, NULL, NULL },
   1003};
   1004
   1005static const PCISubClass mem_subclass[] = {
   1006    { PCI_CLASS_MEMORY_RAM, "memory", NULL },
   1007    { PCI_CLASS_MEMORY_FLASH, "flash", NULL },
   1008    { 0xFF, NULL, NULL },
   1009};
   1010
   1011static const PCISubClass bridg_subclass[] = {
   1012    { PCI_CLASS_BRIDGE_HOST, "host", NULL },
   1013    { PCI_CLASS_BRIDGE_ISA, "isa", NULL },
   1014    { PCI_CLASS_BRIDGE_EISA, "eisa", NULL },
   1015    { PCI_CLASS_BRIDGE_MC, "mca", NULL },
   1016    { PCI_CLASS_BRIDGE_PCI, "pci", NULL },
   1017    { PCI_CLASS_BRIDGE_PCMCIA, "pcmcia", NULL },
   1018    { PCI_CLASS_BRIDGE_NUBUS, "nubus", NULL },
   1019    { PCI_CLASS_BRIDGE_CARDBUS, "cardbus", NULL },
   1020    { PCI_CLASS_BRIDGE_RACEWAY, "raceway", NULL },
   1021    { PCI_CLASS_BRIDGE_PCI_SEMITP, "semi-transparent-pci", NULL },
   1022    { PCI_CLASS_BRIDGE_IB_PCI, "infiniband", NULL },
   1023    { 0xFF, NULL, NULL },
   1024};
   1025
   1026static const PCISubClass comm_subclass[] = {
   1027    { PCI_CLASS_COMMUNICATION_SERIAL, "serial", NULL },
   1028    { PCI_CLASS_COMMUNICATION_PARALLEL, "parallel", NULL },
   1029    { PCI_CLASS_COMMUNICATION_MULTISERIAL, "multiport-serial", NULL },
   1030    { PCI_CLASS_COMMUNICATION_MODEM, "modem", NULL },
   1031    { PCI_CLASS_COMMUNICATION_GPIB, "gpib", NULL },
   1032    { PCI_CLASS_COMMUNICATION_SC, "smart-card", NULL },
   1033    { 0xFF, NULL, NULL, },
   1034};
   1035
   1036static const PCIIFace pic_iface[] = {
   1037    { PCI_CLASS_SYSTEM_PIC_IOAPIC, "io-apic" },
   1038    { PCI_CLASS_SYSTEM_PIC_IOXAPIC, "io-xapic" },
   1039    { 0xFF, NULL },
   1040};
   1041
   1042static const PCISubClass sys_subclass[] = {
   1043    { PCI_CLASS_SYSTEM_PIC, "interrupt-controller", pic_iface },
   1044    { PCI_CLASS_SYSTEM_DMA, "dma-controller", NULL },
   1045    { PCI_CLASS_SYSTEM_TIMER, "timer", NULL },
   1046    { PCI_CLASS_SYSTEM_RTC, "rtc", NULL },
   1047    { PCI_CLASS_SYSTEM_PCI_HOTPLUG, "hot-plug-controller", NULL },
   1048    { PCI_CLASS_SYSTEM_SDHCI, "sd-host-controller", NULL },
   1049    { 0xFF, NULL, NULL },
   1050};
   1051
   1052static const PCISubClass inp_subclass[] = {
   1053    { PCI_CLASS_INPUT_KEYBOARD, "keyboard", NULL },
   1054    { PCI_CLASS_INPUT_PEN, "pen", NULL },
   1055    { PCI_CLASS_INPUT_MOUSE, "mouse", NULL },
   1056    { PCI_CLASS_INPUT_SCANNER, "scanner", NULL },
   1057    { PCI_CLASS_INPUT_GAMEPORT, "gameport", NULL },
   1058    { 0xFF, NULL, NULL },
   1059};
   1060
   1061static const PCISubClass dock_subclass[] = {
   1062    { PCI_CLASS_DOCKING_GENERIC, "dock", NULL },
   1063    { 0xFF, NULL, NULL },
   1064};
   1065
   1066static const PCISubClass cpu_subclass[] = {
   1067    { PCI_CLASS_PROCESSOR_PENTIUM, "pentium", NULL },
   1068    { PCI_CLASS_PROCESSOR_POWERPC, "powerpc", NULL },
   1069    { PCI_CLASS_PROCESSOR_MIPS, "mips", NULL },
   1070    { PCI_CLASS_PROCESSOR_CO, "co-processor", NULL },
   1071    { 0xFF, NULL, NULL },
   1072};
   1073
   1074static const PCIIFace usb_iface[] = {
   1075    { PCI_CLASS_SERIAL_USB_UHCI, "usb-uhci" },
   1076    { PCI_CLASS_SERIAL_USB_OHCI, "usb-ohci", },
   1077    { PCI_CLASS_SERIAL_USB_EHCI, "usb-ehci" },
   1078    { PCI_CLASS_SERIAL_USB_XHCI, "usb-xhci" },
   1079    { PCI_CLASS_SERIAL_USB_UNKNOWN, "usb-unknown" },
   1080    { PCI_CLASS_SERIAL_USB_DEVICE, "usb-device" },
   1081    { 0xFF, NULL },
   1082};
   1083
   1084static const PCISubClass ser_subclass[] = {
   1085    { PCI_CLASS_SERIAL_FIREWIRE, "firewire", NULL },
   1086    { PCI_CLASS_SERIAL_ACCESS, "access-bus", NULL },
   1087    { PCI_CLASS_SERIAL_SSA, "ssa", NULL },
   1088    { PCI_CLASS_SERIAL_USB, "usb", usb_iface },
   1089    { PCI_CLASS_SERIAL_FIBER, "fibre-channel", NULL },
   1090    { PCI_CLASS_SERIAL_SMBUS, "smb", NULL },
   1091    { PCI_CLASS_SERIAL_IB, "infiniband", NULL },
   1092    { PCI_CLASS_SERIAL_IPMI, "ipmi", NULL },
   1093    { PCI_CLASS_SERIAL_SERCOS, "sercos", NULL },
   1094    { PCI_CLASS_SERIAL_CANBUS, "canbus", NULL },
   1095    { 0xFF, NULL, NULL },
   1096};
   1097
   1098static const PCISubClass wrl_subclass[] = {
   1099    { PCI_CLASS_WIRELESS_IRDA, "irda", NULL },
   1100    { PCI_CLASS_WIRELESS_CIR, "consumer-ir", NULL },
   1101    { PCI_CLASS_WIRELESS_RF_CONTROLLER, "rf-controller", NULL },
   1102    { PCI_CLASS_WIRELESS_BLUETOOTH, "bluetooth", NULL },
   1103    { PCI_CLASS_WIRELESS_BROADBAND, "broadband", NULL },
   1104    { 0xFF, NULL, NULL },
   1105};
   1106
   1107static const PCISubClass sat_subclass[] = {
   1108    { PCI_CLASS_SATELLITE_TV, "satellite-tv", NULL },
   1109    { PCI_CLASS_SATELLITE_AUDIO, "satellite-audio", NULL },
   1110    { PCI_CLASS_SATELLITE_VOICE, "satellite-voice", NULL },
   1111    { PCI_CLASS_SATELLITE_DATA, "satellite-data", NULL },
   1112    { 0xFF, NULL, NULL },
   1113};
   1114
   1115static const PCISubClass crypt_subclass[] = {
   1116    { PCI_CLASS_CRYPT_NETWORK, "network-encryption", NULL },
   1117    { PCI_CLASS_CRYPT_ENTERTAINMENT,
   1118      "entertainment-encryption", NULL },
   1119    { 0xFF, NULL, NULL },
   1120};
   1121
   1122static const PCISubClass spc_subclass[] = {
   1123    { PCI_CLASS_SP_DPIO, "dpio", NULL },
   1124    { PCI_CLASS_SP_PERF, "counter", NULL },
   1125    { PCI_CLASS_SP_SYNCH, "measurement", NULL },
   1126    { PCI_CLASS_SP_MANAGEMENT, "management-card", NULL },
   1127    { 0xFF, NULL, NULL },
   1128};
   1129
   1130static const PCIClass pci_classes[] = {
   1131    { "legacy-device", undef_subclass },
   1132    { "mass-storage",  mass_subclass },
   1133    { "network", net_subclass },
   1134    { "display", displ_subclass, },
   1135    { "multimedia-device", media_subclass },
   1136    { "memory-controller", mem_subclass },
   1137    { "unknown-bridge", bridg_subclass },
   1138    { "communication-controller", comm_subclass},
   1139    { "system-peripheral", sys_subclass },
   1140    { "input-controller", inp_subclass },
   1141    { "docking-station", dock_subclass },
   1142    { "cpu", cpu_subclass },
   1143    { "serial-bus", ser_subclass },
   1144    { "wireless-controller", wrl_subclass },
   1145    { "intelligent-io", NULL },
   1146    { "satellite-device", sat_subclass },
   1147    { "encryption", crypt_subclass },
   1148    { "data-processing-controller", spc_subclass },
   1149};
   1150
   1151static const char *dt_name_from_class(uint8_t class, uint8_t subclass,
   1152                                      uint8_t iface)
   1153{
   1154    const PCIClass *pclass;
   1155    const PCISubClass *psubclass;
   1156    const PCIIFace *piface;
   1157    const char *name;
   1158
   1159    if (class >= ARRAY_SIZE(pci_classes)) {
   1160        return "pci";
   1161    }
   1162
   1163    pclass = pci_classes + class;
   1164    name = pclass->name;
   1165
   1166    if (pclass->subc == NULL) {
   1167        return name;
   1168    }
   1169
   1170    psubclass = pclass->subc;
   1171    while ((psubclass->subclass & 0xff) != 0xff) {
   1172        if ((psubclass->subclass & 0xff) == subclass) {
   1173            name = psubclass->name;
   1174            break;
   1175        }
   1176        psubclass++;
   1177    }
   1178
   1179    piface = psubclass->iface;
   1180    if (piface == NULL) {
   1181        return name;
   1182    }
   1183    while ((piface->iface & 0xff) != 0xff) {
   1184        if ((piface->iface & 0xff) == iface) {
   1185            name = piface->name;
   1186            break;
   1187        }
   1188        piface++;
   1189    }
   1190
   1191    return name;
   1192}
   1193
   1194/*
   1195 * DRC helper functions
   1196 */
   1197
   1198static uint32_t drc_id_from_devfn(SpaprPhbState *phb,
   1199                                  uint8_t chassis, int32_t devfn)
   1200{
   1201    return (phb->index << 16) | (chassis << 8) | devfn;
   1202}
   1203
   1204static SpaprDrc *drc_from_devfn(SpaprPhbState *phb,
   1205                                uint8_t chassis, int32_t devfn)
   1206{
   1207    return spapr_drc_by_id(TYPE_SPAPR_DRC_PCI,
   1208                           drc_id_from_devfn(phb, chassis, devfn));
   1209}
   1210
   1211static uint8_t chassis_from_bus(PCIBus *bus)
   1212{
   1213    if (pci_bus_is_root(bus)) {
   1214        return 0;
   1215    } else {
   1216        PCIDevice *bridge = pci_bridge_get_device(bus);
   1217
   1218        return object_property_get_uint(OBJECT(bridge), "chassis_nr",
   1219                                        &error_abort);
   1220    }
   1221}
   1222
   1223static SpaprDrc *drc_from_dev(SpaprPhbState *phb, PCIDevice *dev)
   1224{
   1225    uint8_t chassis = chassis_from_bus(pci_get_bus(dev));
   1226
   1227    return drc_from_devfn(phb, chassis, dev->devfn);
   1228}
   1229
   1230static void add_drcs(SpaprPhbState *phb, PCIBus *bus)
   1231{
   1232    Object *owner;
   1233    int i;
   1234    uint8_t chassis;
   1235
   1236    if (!phb->dr_enabled) {
   1237        return;
   1238    }
   1239
   1240    chassis = chassis_from_bus(bus);
   1241
   1242    if (pci_bus_is_root(bus)) {
   1243        owner = OBJECT(phb);
   1244    } else {
   1245        owner = OBJECT(pci_bridge_get_device(bus));
   1246    }
   1247
   1248    for (i = 0; i < PCI_SLOT_MAX * PCI_FUNC_MAX; i++) {
   1249        spapr_dr_connector_new(owner, TYPE_SPAPR_DRC_PCI,
   1250                               drc_id_from_devfn(phb, chassis, i));
   1251    }
   1252}
   1253
   1254static void remove_drcs(SpaprPhbState *phb, PCIBus *bus)
   1255{
   1256    int i;
   1257    uint8_t chassis;
   1258
   1259    if (!phb->dr_enabled) {
   1260        return;
   1261    }
   1262
   1263    chassis = chassis_from_bus(bus);
   1264
   1265    for (i = PCI_SLOT_MAX * PCI_FUNC_MAX - 1; i >= 0; i--) {
   1266        SpaprDrc *drc = drc_from_devfn(phb, chassis, i);
   1267
   1268        if (drc) {
   1269            object_unparent(OBJECT(drc));
   1270        }
   1271    }
   1272}
   1273
   1274typedef struct PciWalkFdt {
   1275    void *fdt;
   1276    int offset;
   1277    SpaprPhbState *sphb;
   1278    int err;
   1279} PciWalkFdt;
   1280
   1281static int spapr_dt_pci_device(SpaprPhbState *sphb, PCIDevice *dev,
   1282                               void *fdt, int parent_offset);
   1283
   1284static void spapr_dt_pci_device_cb(PCIBus *bus, PCIDevice *pdev,
   1285                                   void *opaque)
   1286{
   1287    PciWalkFdt *p = opaque;
   1288    int err;
   1289
   1290    if (p->err) {
   1291        /* Something's already broken, don't keep going */
   1292        return;
   1293    }
   1294
   1295    err = spapr_dt_pci_device(p->sphb, pdev, p->fdt, p->offset);
   1296    if (err < 0) {
   1297        p->err = err;
   1298    }
   1299}
   1300
   1301/* Augment PCI device node with bridge specific information */
   1302static int spapr_dt_pci_bus(SpaprPhbState *sphb, PCIBus *bus,
   1303                               void *fdt, int offset)
   1304{
   1305    Object *owner;
   1306    PciWalkFdt cbinfo = {
   1307        .fdt = fdt,
   1308        .offset = offset,
   1309        .sphb = sphb,
   1310        .err = 0,
   1311    };
   1312    int ret;
   1313
   1314    _FDT(fdt_setprop_cell(fdt, offset, "#address-cells",
   1315                          RESOURCE_CELLS_ADDRESS));
   1316    _FDT(fdt_setprop_cell(fdt, offset, "#size-cells",
   1317                          RESOURCE_CELLS_SIZE));
   1318
   1319    assert(bus);
   1320    pci_for_each_device_reverse(bus, pci_bus_num(bus),
   1321                                spapr_dt_pci_device_cb, &cbinfo);
   1322    if (cbinfo.err) {
   1323        return cbinfo.err;
   1324    }
   1325
   1326    if (pci_bus_is_root(bus)) {
   1327        owner = OBJECT(sphb);
   1328    } else {
   1329        owner = OBJECT(pci_bridge_get_device(bus));
   1330    }
   1331
   1332    ret = spapr_dt_drc(fdt, offset, owner,
   1333                       SPAPR_DR_CONNECTOR_TYPE_PCI);
   1334    if (ret) {
   1335        return ret;
   1336    }
   1337
   1338    return offset;
   1339}
   1340
   1341char *spapr_pci_fw_dev_name(PCIDevice *dev)
   1342{
   1343    const gchar *basename;
   1344    int slot = PCI_SLOT(dev->devfn);
   1345    int func = PCI_FUNC(dev->devfn);
   1346    uint32_t ccode = pci_default_read_config(dev, PCI_CLASS_PROG, 3);
   1347
   1348    basename = dt_name_from_class((ccode >> 16) & 0xff, (ccode >> 8) & 0xff,
   1349                                  ccode & 0xff);
   1350
   1351    if (func != 0) {
   1352        return g_strdup_printf("%s@%x,%x", basename, slot, func);
   1353    } else {
   1354        return g_strdup_printf("%s@%x", basename, slot);
   1355    }
   1356}
   1357
   1358/* create OF node for pci device and required OF DT properties */
   1359static int spapr_dt_pci_device(SpaprPhbState *sphb, PCIDevice *dev,
   1360                               void *fdt, int parent_offset)
   1361{
   1362    int offset;
   1363    g_autofree gchar *nodename = spapr_pci_fw_dev_name(dev);
   1364    PCIDeviceClass *pc = PCI_DEVICE_GET_CLASS(dev);
   1365    ResourceProps rp;
   1366    SpaprDrc *drc = drc_from_dev(sphb, dev);
   1367    uint32_t vendor_id = pci_default_read_config(dev, PCI_VENDOR_ID, 2);
   1368    uint32_t device_id = pci_default_read_config(dev, PCI_DEVICE_ID, 2);
   1369    uint32_t revision_id = pci_default_read_config(dev, PCI_REVISION_ID, 1);
   1370    uint32_t ccode = pci_default_read_config(dev, PCI_CLASS_PROG, 3);
   1371    uint32_t irq_pin = pci_default_read_config(dev, PCI_INTERRUPT_PIN, 1);
   1372    uint32_t subsystem_id = pci_default_read_config(dev, PCI_SUBSYSTEM_ID, 2);
   1373    uint32_t subsystem_vendor_id =
   1374        pci_default_read_config(dev, PCI_SUBSYSTEM_VENDOR_ID, 2);
   1375    uint32_t cache_line_size =
   1376        pci_default_read_config(dev, PCI_CACHE_LINE_SIZE, 1);
   1377    uint32_t pci_status = pci_default_read_config(dev, PCI_STATUS, 2);
   1378    gchar *loc_code;
   1379
   1380    _FDT(offset = fdt_add_subnode(fdt, parent_offset, nodename));
   1381
   1382    /* in accordance with PAPR+ v2.7 13.6.3, Table 181 */
   1383    _FDT(fdt_setprop_cell(fdt, offset, "vendor-id", vendor_id));
   1384    _FDT(fdt_setprop_cell(fdt, offset, "device-id", device_id));
   1385    _FDT(fdt_setprop_cell(fdt, offset, "revision-id", revision_id));
   1386
   1387    _FDT(fdt_setprop_cell(fdt, offset, "class-code", ccode));
   1388    if (irq_pin) {
   1389        _FDT(fdt_setprop_cell(fdt, offset, "interrupts", irq_pin));
   1390    }
   1391
   1392    if (subsystem_id) {
   1393        _FDT(fdt_setprop_cell(fdt, offset, "subsystem-id", subsystem_id));
   1394    }
   1395
   1396    if (subsystem_vendor_id) {
   1397        _FDT(fdt_setprop_cell(fdt, offset, "subsystem-vendor-id",
   1398                              subsystem_vendor_id));
   1399    }
   1400
   1401    _FDT(fdt_setprop_cell(fdt, offset, "cache-line-size", cache_line_size));
   1402
   1403
   1404    /* the following fdt cells are masked off the pci status register */
   1405    _FDT(fdt_setprop_cell(fdt, offset, "devsel-speed",
   1406                          PCI_STATUS_DEVSEL_MASK & pci_status));
   1407
   1408    if (pci_status & PCI_STATUS_FAST_BACK) {
   1409        _FDT(fdt_setprop(fdt, offset, "fast-back-to-back", NULL, 0));
   1410    }
   1411    if (pci_status & PCI_STATUS_66MHZ) {
   1412        _FDT(fdt_setprop(fdt, offset, "66mhz-capable", NULL, 0));
   1413    }
   1414    if (pci_status & PCI_STATUS_UDF) {
   1415        _FDT(fdt_setprop(fdt, offset, "udf-supported", NULL, 0));
   1416    }
   1417
   1418    loc_code = spapr_phb_get_loc_code(sphb, dev);
   1419    _FDT(fdt_setprop_string(fdt, offset, "ibm,loc-code", loc_code));
   1420    g_free(loc_code);
   1421
   1422    if (drc) {
   1423        _FDT(fdt_setprop_cell(fdt, offset, "ibm,my-drc-index",
   1424                              spapr_drc_index(drc)));
   1425    }
   1426
   1427    if (msi_present(dev)) {
   1428        uint32_t max_msi = msi_nr_vectors_allocated(dev);
   1429        if (max_msi) {
   1430            _FDT(fdt_setprop_cell(fdt, offset, "ibm,req#msi", max_msi));
   1431        }
   1432    }
   1433    if (msix_present(dev)) {
   1434        uint32_t max_msix = dev->msix_entries_nr;
   1435        if (max_msix) {
   1436            _FDT(fdt_setprop_cell(fdt, offset, "ibm,req#msi-x", max_msix));
   1437        }
   1438    }
   1439
   1440    populate_resource_props(dev, &rp);
   1441    _FDT(fdt_setprop(fdt, offset, "reg", (uint8_t *)rp.reg, rp.reg_len));
   1442
   1443    if (sphb->pcie_ecs && pci_is_express(dev)) {
   1444        _FDT(fdt_setprop_cell(fdt, offset, "ibm,pci-config-space-type", 0x1));
   1445    }
   1446
   1447    spapr_phb_nvgpu_populate_pcidev_dt(dev, fdt, offset, sphb);
   1448
   1449    if (!pc->is_bridge) {
   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));
   1455        return offset;
   1456    } else {
   1457        PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(dev));
   1458
   1459        return spapr_dt_pci_bus(sphb, sec_bus, fdt, offset);
   1460    }
   1461}
   1462
   1463/* Callback to be called during DRC release. */
   1464void spapr_phb_remove_pci_device_cb(DeviceState *dev)
   1465{
   1466    HotplugHandler *hotplug_ctrl = qdev_get_hotplug_handler(dev);
   1467
   1468    hotplug_handler_unplug(hotplug_ctrl, dev, &error_abort);
   1469    object_unparent(OBJECT(dev));
   1470}
   1471
   1472int spapr_pci_dt_populate(SpaprDrc *drc, SpaprMachineState *spapr,
   1473                          void *fdt, int *fdt_start_offset, Error **errp)
   1474{
   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);
   1478
   1479    *fdt_start_offset = spapr_dt_pci_device(sphb, pdev, fdt, 0);
   1480    return 0;
   1481}
   1482
   1483static void spapr_pci_bridge_plug(SpaprPhbState *phb,
   1484                                  PCIBridge *bridge)
   1485{
   1486    PCIBus *bus = pci_bridge_get_sec_bus(bridge);
   1487
   1488    add_drcs(phb, bus);
   1489}
   1490
   1491/* Returns non-zero if the value of "chassis_nr" is already in use */
   1492static int check_chassis_nr(Object *obj, void *opaque)
   1493{
   1494    int new_chassis_nr =
   1495        object_property_get_uint(opaque, "chassis_nr", &error_abort);
   1496    int chassis_nr =
   1497        object_property_get_uint(obj, "chassis_nr", NULL);
   1498
   1499    if (!object_dynamic_cast(obj, TYPE_PCI_BRIDGE)) {
   1500        return 0;
   1501    }
   1502
   1503    /* Skip unsupported bridge types */
   1504    if (!chassis_nr) {
   1505        return 0;
   1506    }
   1507
   1508    /* Skip self */
   1509    if (obj == opaque) {
   1510        return 0;
   1511    }
   1512
   1513    return chassis_nr == new_chassis_nr;
   1514}
   1515
   1516static bool bridge_has_valid_chassis_nr(Object *bridge, Error **errp)
   1517{
   1518    int chassis_nr =
   1519        object_property_get_uint(bridge, "chassis_nr", NULL);
   1520
   1521    /*
   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
   1524     * or not.
   1525     */
   1526    if (!chassis_nr) {
   1527        error_setg(errp, "PCI Bridge lacks a \"chassis_nr\" property");
   1528        error_append_hint(errp, "Try -device pci-bridge instead.\n");
   1529        return false;
   1530    }
   1531
   1532    /* We want unique values for "chassis_nr" */
   1533    if (object_child_foreach_recursive(object_get_root(), check_chassis_nr,
   1534                                       bridge)) {
   1535        error_setg(errp, "Bridge chassis %d already in use", chassis_nr);
   1536        return false;
   1537    }
   1538
   1539    return true;
   1540}
   1541
   1542static void spapr_pci_pre_plug(HotplugHandler *plug_handler,
   1543                               DeviceState *plugged_dev, Error **errp)
   1544{
   1545    SpaprPhbState *phb = SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler));
   1546    PCIDevice *pdev = PCI_DEVICE(plugged_dev);
   1547    PCIDeviceClass *pc = PCI_DEVICE_GET_CLASS(plugged_dev);
   1548    SpaprDrc *drc = drc_from_dev(phb, pdev);
   1549    PCIBus *bus = PCI_BUS(qdev_get_parent_bus(DEVICE(pdev)));
   1550    uint32_t slotnr = PCI_SLOT(pdev->devfn);
   1551
   1552    if (!phb->dr_enabled) {
   1553        /* if this is a hotplug operation initiated by the user
   1554         * we need to let them know it's not enabled
   1555         */
   1556        if (plugged_dev->hotplugged) {
   1557            error_setg(errp, QERR_BUS_NO_HOTPLUG,
   1558                       object_get_typename(OBJECT(phb)));
   1559            return;
   1560        }
   1561    }
   1562
   1563    if (pc->is_bridge) {
   1564        if (!bridge_has_valid_chassis_nr(OBJECT(plugged_dev), errp)) {
   1565            return;
   1566        }
   1567    }
   1568
   1569    /* Following the QEMU convention used for PCIe multifunction
   1570     * hotplug, we do not allow functions to be hotplugged to a
   1571     * slot that already has function 0 present
   1572     */
   1573    if (plugged_dev->hotplugged && bus->devices[PCI_DEVFN(slotnr, 0)] &&
   1574        PCI_FUNC(pdev->devfn) != 0) {
   1575        error_setg(errp, "PCI: slot %d function 0 already occupied by %s,"
   1576                   " additional functions can no longer be exposed to guest.",
   1577                   slotnr, bus->devices[PCI_DEVFN(slotnr, 0)]->name);
   1578    }
   1579
   1580    if (drc && drc->dev) {
   1581        error_setg(errp, "PCI: slot %d already occupied by %s", slotnr,
   1582                   pci_get_function_0(PCI_DEVICE(drc->dev))->name);
   1583        return;
   1584    }
   1585}
   1586
   1587static void spapr_pci_plug(HotplugHandler *plug_handler,
   1588                           DeviceState *plugged_dev, Error **errp)
   1589{
   1590    SpaprPhbState *phb = SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler));
   1591    PCIDevice *pdev = PCI_DEVICE(plugged_dev);
   1592    PCIDeviceClass *pc = PCI_DEVICE_GET_CLASS(plugged_dev);
   1593    SpaprDrc *drc = drc_from_dev(phb, pdev);
   1594    uint32_t slotnr = PCI_SLOT(pdev->devfn);
   1595
   1596    /*
   1597     * If DR is disabled we don't need to do anything in the case of
   1598     * hotplug or coldplug callbacks.
   1599     */
   1600    if (!phb->dr_enabled) {
   1601        return;
   1602    }
   1603
   1604    g_assert(drc);
   1605
   1606    if (pc->is_bridge) {
   1607        spapr_pci_bridge_plug(phb, PCI_BRIDGE(plugged_dev));
   1608    }
   1609
   1610    /* spapr_pci_pre_plug() already checked the DRC is attachable */
   1611    spapr_drc_attach(drc, DEVICE(pdev));
   1612
   1613    /* If this is function 0, signal hotplug for all the device functions.
   1614     * Otherwise defer sending the hotplug event.
   1615     */
   1616    if (!spapr_drc_hotplugged(plugged_dev)) {
   1617        spapr_drc_reset(drc);
   1618    } else if (PCI_FUNC(pdev->devfn) == 0) {
   1619        int i;
   1620        uint8_t chassis = chassis_from_bus(pci_get_bus(pdev));
   1621
   1622        for (i = 0; i < 8; i++) {
   1623            SpaprDrc *func_drc;
   1624            SpaprDrcClass *func_drck;
   1625            SpaprDREntitySense state;
   1626
   1627            func_drc = drc_from_devfn(phb, chassis, PCI_DEVFN(slotnr, i));
   1628            func_drck = SPAPR_DR_CONNECTOR_GET_CLASS(func_drc);
   1629            state = func_drck->dr_entity_sense(func_drc);
   1630
   1631            if (state == SPAPR_DR_ENTITY_SENSE_PRESENT) {
   1632                spapr_hotplug_req_add_by_index(func_drc);
   1633            }
   1634        }
   1635    }
   1636}
   1637
   1638static void spapr_pci_bridge_unplug(SpaprPhbState *phb,
   1639                                    PCIBridge *bridge)
   1640{
   1641    PCIBus *bus = pci_bridge_get_sec_bus(bridge);
   1642
   1643    remove_drcs(phb, bus);
   1644}
   1645
   1646static void spapr_pci_unplug(HotplugHandler *plug_handler,
   1647                             DeviceState *plugged_dev, Error **errp)
   1648{
   1649    PCIDeviceClass *pc = PCI_DEVICE_GET_CLASS(plugged_dev);
   1650    SpaprPhbState *phb = SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler));
   1651
   1652    /* some version guests do not wait for completion of a device
   1653     * cleanup (generally done asynchronously by the kernel) before
   1654     * signaling to QEMU that the device is safe, but instead sleep
   1655     * for some 'safe' period of time. unfortunately on a busy host
   1656     * this sleep isn't guaranteed to be long enough, resulting in
   1657     * bad things like IRQ lines being left asserted during final
   1658     * device removal. to deal with this we call reset just prior
   1659     * to finalizing the device, which will put the device back into
   1660     * an 'idle' state, as the device cleanup code expects.
   1661     */
   1662    pci_device_reset(PCI_DEVICE(plugged_dev));
   1663
   1664    if (pc->is_bridge) {
   1665        spapr_pci_bridge_unplug(phb, PCI_BRIDGE(plugged_dev));
   1666        return;
   1667    }
   1668
   1669    qdev_unrealize(plugged_dev);
   1670}
   1671
   1672static void spapr_pci_unplug_request(HotplugHandler *plug_handler,
   1673                                     DeviceState *plugged_dev, Error **errp)
   1674{
   1675    SpaprPhbState *phb = SPAPR_PCI_HOST_BRIDGE(DEVICE(plug_handler));
   1676    PCIDevice *pdev = PCI_DEVICE(plugged_dev);
   1677    SpaprDrc *drc = drc_from_dev(phb, pdev);
   1678
   1679    if (!phb->dr_enabled) {
   1680        error_setg(errp, QERR_BUS_NO_HOTPLUG,
   1681                   object_get_typename(OBJECT(phb)));
   1682        return;
   1683    }
   1684
   1685    g_assert(drc);
   1686    g_assert(drc->dev == plugged_dev);
   1687
   1688    if (!spapr_drc_unplug_requested(drc)) {
   1689        PCIDeviceClass *pc = PCI_DEVICE_GET_CLASS(plugged_dev);
   1690        uint32_t slotnr = PCI_SLOT(pdev->devfn);
   1691        SpaprDrc *func_drc;
   1692        SpaprDrcClass *func_drck;
   1693        SpaprDREntitySense state;
   1694        int i;
   1695        uint8_t chassis = chassis_from_bus(pci_get_bus(pdev));
   1696
   1697        if (pc->is_bridge) {
   1698            error_setg(errp, "PCI: Hot unplug of PCI bridges not supported");
   1699            return;
   1700        }
   1701        if (object_property_get_uint(OBJECT(pdev), "nvlink2-tgt", NULL)) {
   1702            error_setg(errp, "PCI: Cannot unplug NVLink2 devices");
   1703            return;
   1704        }
   1705
   1706        /* ensure any other present functions are pending unplug */
   1707        if (PCI_FUNC(pdev->devfn) == 0) {
   1708            for (i = 1; i < 8; i++) {
   1709                func_drc = drc_from_devfn(phb, chassis, PCI_DEVFN(slotnr, i));
   1710                func_drck = SPAPR_DR_CONNECTOR_GET_CLASS(func_drc);
   1711                state = func_drck->dr_entity_sense(func_drc);
   1712                if (state == SPAPR_DR_ENTITY_SENSE_PRESENT
   1713                    && !spapr_drc_unplug_requested(func_drc)) {
   1714                    /*
   1715                     * Attempting to remove function 0 of a multifunction
   1716                     * device will will cascade into removing all child
   1717                     * functions, even if their unplug weren't requested
   1718                     * beforehand.
   1719                     */
   1720                    spapr_drc_unplug_request(func_drc);
   1721                }
   1722            }
   1723        }
   1724
   1725        spapr_drc_unplug_request(drc);
   1726
   1727        /* if this isn't func 0, defer unplug event. otherwise signal removal
   1728         * for all present functions
   1729         */
   1730        if (PCI_FUNC(pdev->devfn) == 0) {
   1731            for (i = 7; i >= 0; i--) {
   1732                func_drc = drc_from_devfn(phb, chassis, PCI_DEVFN(slotnr, i));
   1733                func_drck = SPAPR_DR_CONNECTOR_GET_CLASS(func_drc);
   1734                state = func_drck->dr_entity_sense(func_drc);
   1735                if (state == SPAPR_DR_ENTITY_SENSE_PRESENT) {
   1736                    spapr_hotplug_req_remove_by_index(func_drc);
   1737                }
   1738            }
   1739        }
   1740    } else {
   1741        error_setg(errp,
   1742                   "PCI device unplug already in progress for device %s",
   1743                   drc->dev->id);
   1744    }
   1745}
   1746
   1747static void spapr_phb_finalizefn(Object *obj)
   1748{
   1749    SpaprPhbState *sphb = SPAPR_PCI_HOST_BRIDGE(obj);
   1750
   1751    g_free(sphb->dtbusname);
   1752    sphb->dtbusname = NULL;
   1753}
   1754
   1755static void spapr_phb_unrealize(DeviceState *dev)
   1756{
   1757    SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
   1758    SysBusDevice *s = SYS_BUS_DEVICE(dev);
   1759    PCIHostState *phb = PCI_HOST_BRIDGE(s);
   1760    SpaprPhbState *sphb = SPAPR_PCI_HOST_BRIDGE(phb);
   1761    SpaprTceTable *tcet;
   1762    int i;
   1763    const unsigned windows_supported = spapr_phb_windows_supported(sphb);
   1764
   1765    spapr_phb_nvgpu_free(sphb);
   1766
   1767    if (sphb->msi) {
   1768        g_hash_table_unref(sphb->msi);
   1769        sphb->msi = NULL;
   1770    }
   1771
   1772    /*
   1773     * Remove IO/MMIO subregions and aliases, rest should get cleaned
   1774     * via PHB's unrealize->object_finalize
   1775     */
   1776    for (i = windows_supported - 1; i >= 0; i--) {
   1777        tcet = spapr_tce_find_by_liobn(sphb->dma_liobn[i]);
   1778        if (tcet) {
   1779            memory_region_del_subregion(&sphb->iommu_root,
   1780                                        spapr_tce_get_iommu(tcet));
   1781        }
   1782    }
   1783
   1784    remove_drcs(sphb, phb->bus);
   1785
   1786    for (i = PCI_NUM_PINS - 1; i >= 0; i--) {
   1787        if (sphb->lsi_table[i].irq) {
   1788            spapr_irq_free(spapr, sphb->lsi_table[i].irq, 1);
   1789            sphb->lsi_table[i].irq = 0;
   1790        }
   1791    }
   1792
   1793    QLIST_REMOVE(sphb, list);
   1794
   1795    memory_region_del_subregion(&sphb->iommu_root, &sphb->msiwindow);
   1796
   1797    /*
   1798     * An attached PCI device may have memory listeners, eg. VFIO PCI. We have
   1799     * unmapped all sections. Remove the listeners now, before destroying the
   1800     * address space.
   1801     */
   1802    address_space_remove_listeners(&sphb->iommu_as);
   1803    address_space_destroy(&sphb->iommu_as);
   1804
   1805    qbus_set_hotplug_handler(BUS(phb->bus), NULL);
   1806    pci_unregister_root_bus(phb->bus);
   1807
   1808    memory_region_del_subregion(get_system_memory(), &sphb->iowindow);
   1809    if (sphb->mem64_win_pciaddr != (hwaddr)-1) {
   1810        memory_region_del_subregion(get_system_memory(), &sphb->mem64window);
   1811    }
   1812    memory_region_del_subregion(get_system_memory(), &sphb->mem32window);
   1813}
   1814
   1815static void spapr_phb_destroy_msi(gpointer opaque)
   1816{
   1817    SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
   1818    SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
   1819    SpaprPciMsi *msi = opaque;
   1820
   1821    if (!smc->legacy_irq_allocation) {
   1822        spapr_irq_msi_free(spapr, msi->first_irq, msi->num);
   1823    }
   1824    spapr_irq_free(spapr, msi->first_irq, msi->num);
   1825    g_free(msi);
   1826}
   1827
   1828static void spapr_phb_realize(DeviceState *dev, Error **errp)
   1829{
   1830    ERRP_GUARD();
   1831    /* We don't use SPAPR_MACHINE() in order to exit gracefully if the user
   1832     * tries to add a sPAPR PHB to a non-pseries machine.
   1833     */
   1834    SpaprMachineState *spapr =
   1835        (SpaprMachineState *) object_dynamic_cast(qdev_get_machine(),
   1836                                                  TYPE_SPAPR_MACHINE);
   1837    SpaprMachineClass *smc = spapr ? SPAPR_MACHINE_GET_CLASS(spapr) : NULL;
   1838    SysBusDevice *s = SYS_BUS_DEVICE(dev);
   1839    SpaprPhbState *sphb = SPAPR_PCI_HOST_BRIDGE(s);
   1840    PCIHostState *phb = PCI_HOST_BRIDGE(s);
   1841    MachineState *ms = MACHINE(spapr);
   1842    char *namebuf;
   1843    int i;
   1844    PCIBus *bus;
   1845    uint64_t msi_window_size = 4096;
   1846    SpaprTceTable *tcet;
   1847    const unsigned windows_supported = spapr_phb_windows_supported(sphb);
   1848
   1849    if (!spapr) {
   1850        error_setg(errp, TYPE_SPAPR_PCI_HOST_BRIDGE " needs a pseries machine");
   1851        return;
   1852    }
   1853
   1854    assert(sphb->index != (uint32_t)-1); /* checked in spapr_phb_pre_plug() */
   1855
   1856    if (sphb->mem64_win_size != 0) {
   1857        if (sphb->mem_win_size > SPAPR_PCI_MEM32_WIN_SIZE) {
   1858            error_setg(errp, "32-bit memory window of size 0x%"HWADDR_PRIx
   1859                       " (max 2 GiB)", sphb->mem_win_size);
   1860            return;
   1861        }
   1862
   1863        /* 64-bit window defaults to identity mapping */
   1864        sphb->mem64_win_pciaddr = sphb->mem64_win_addr;
   1865    } else if (sphb->mem_win_size > SPAPR_PCI_MEM32_WIN_SIZE) {
   1866        /*
   1867         * For compatibility with old configuration, if no 64-bit MMIO
   1868         * window is specified, but the ordinary (32-bit) memory
   1869         * window is specified as > 2GiB, we treat it as a 2GiB 32-bit
   1870         * window, with a 64-bit MMIO window following on immediately
   1871         * afterwards
   1872         */
   1873        sphb->mem64_win_size = sphb->mem_win_size - SPAPR_PCI_MEM32_WIN_SIZE;
   1874        sphb->mem64_win_addr = sphb->mem_win_addr + SPAPR_PCI_MEM32_WIN_SIZE;
   1875        sphb->mem64_win_pciaddr =
   1876            SPAPR_PCI_MEM_WIN_BUS_OFFSET + SPAPR_PCI_MEM32_WIN_SIZE;
   1877        sphb->mem_win_size = SPAPR_PCI_MEM32_WIN_SIZE;
   1878    }
   1879
   1880    if (spapr_pci_find_phb(spapr, sphb->buid)) {
   1881        SpaprPhbState *s;
   1882
   1883        error_setg(errp, "PCI host bridges must have unique indexes");
   1884        error_append_hint(errp, "The following indexes are already in use:");
   1885        QLIST_FOREACH(s, &spapr->phbs, list) {
   1886            error_append_hint(errp, " %d", s->index);
   1887        }
   1888        error_append_hint(errp, "\nTry another value for the index property\n");
   1889        return;
   1890    }
   1891
   1892    if (sphb->numa_node != -1 &&
   1893        (sphb->numa_node >= MAX_NODES ||
   1894         !ms->numa_state->nodes[sphb->numa_node].present)) {
   1895        error_setg(errp, "Invalid NUMA node ID for PCI host bridge");
   1896        return;
   1897    }
   1898
   1899    sphb->dtbusname = g_strdup_printf("pci@%" PRIx64, sphb->buid);
   1900
   1901    /* Initialize memory regions */
   1902    namebuf = g_strdup_printf("%s.mmio", sphb->dtbusname);
   1903    memory_region_init(&sphb->memspace, OBJECT(sphb), namebuf, UINT64_MAX);
   1904    g_free(namebuf);
   1905
   1906    namebuf = g_strdup_printf("%s.mmio32-alias", sphb->dtbusname);
   1907    memory_region_init_alias(&sphb->mem32window, OBJECT(sphb),
   1908                             namebuf, &sphb->memspace,
   1909                             SPAPR_PCI_MEM_WIN_BUS_OFFSET, sphb->mem_win_size);
   1910    g_free(namebuf);
   1911    memory_region_add_subregion(get_system_memory(), sphb->mem_win_addr,
   1912                                &sphb->mem32window);
   1913
   1914    if (sphb->mem64_win_size != 0) {
   1915        namebuf = g_strdup_printf("%s.mmio64-alias", sphb->dtbusname);
   1916        memory_region_init_alias(&sphb->mem64window, OBJECT(sphb),
   1917                                 namebuf, &sphb->memspace,
   1918                                 sphb->mem64_win_pciaddr, sphb->mem64_win_size);
   1919        g_free(namebuf);
   1920
   1921        memory_region_add_subregion(get_system_memory(),
   1922                                    sphb->mem64_win_addr,
   1923                                    &sphb->mem64window);
   1924    }
   1925
   1926    /* Initialize IO regions */
   1927    namebuf = g_strdup_printf("%s.io", sphb->dtbusname);
   1928    memory_region_init(&sphb->iospace, OBJECT(sphb),
   1929                       namebuf, SPAPR_PCI_IO_WIN_SIZE);
   1930    g_free(namebuf);
   1931
   1932    namebuf = g_strdup_printf("%s.io-alias", sphb->dtbusname);
   1933    memory_region_init_alias(&sphb->iowindow, OBJECT(sphb), namebuf,
   1934                             &sphb->iospace, 0, SPAPR_PCI_IO_WIN_SIZE);
   1935    g_free(namebuf);
   1936    memory_region_add_subregion(get_system_memory(), sphb->io_win_addr,
   1937                                &sphb->iowindow);
   1938
   1939    bus = pci_register_root_bus(dev, NULL,
   1940                                pci_spapr_set_irq, pci_swizzle_map_irq_fn, sphb,
   1941                                &sphb->memspace, &sphb->iospace,
   1942                                PCI_DEVFN(0, 0), PCI_NUM_PINS,
   1943                                TYPE_PCI_BUS);
   1944
   1945    /*
   1946     * Despite resembling a vanilla PCI bus in most ways, the PAPR
   1947     * para-virtualized PCI bus *does* permit PCI-E extended config
   1948     * space access
   1949     */
   1950    if (sphb->pcie_ecs) {
   1951        bus->flags |= PCI_BUS_EXTENDED_CONFIG_SPACE;
   1952    }
   1953    phb->bus = bus;
   1954    qbus_set_hotplug_handler(BUS(phb->bus), OBJECT(sphb));
   1955
   1956    /*
   1957     * Initialize PHB address space.
   1958     * By default there will be at least one subregion for default
   1959     * 32bit DMA window.
   1960     * Later the guest might want to create another DMA window
   1961     * which will become another memory subregion.
   1962     */
   1963    namebuf = g_strdup_printf("%s.iommu-root", sphb->dtbusname);
   1964    memory_region_init(&sphb->iommu_root, OBJECT(sphb),
   1965                       namebuf, UINT64_MAX);
   1966    g_free(namebuf);
   1967    address_space_init(&sphb->iommu_as, &sphb->iommu_root,
   1968                       sphb->dtbusname);
   1969
   1970    /*
   1971     * As MSI/MSIX interrupts trigger by writing at MSI/MSIX vectors,
   1972     * we need to allocate some memory to catch those writes coming
   1973     * from msi_notify()/msix_notify().
   1974     * As MSIMessage:addr is going to be the same and MSIMessage:data
   1975     * is going to be a VIRQ number, 4 bytes of the MSI MR will only
   1976     * be used.
   1977     *
   1978     * For KVM we want to ensure that this memory is a full page so that
   1979     * our memory slot is of page size granularity.
   1980     */
   1981    if (kvm_enabled()) {
   1982        msi_window_size = qemu_real_host_page_size;
   1983    }
   1984
   1985    memory_region_init_io(&sphb->msiwindow, OBJECT(sphb), &spapr_msi_ops, spapr,
   1986                          "msi", msi_window_size);
   1987    memory_region_add_subregion(&sphb->iommu_root, SPAPR_PCI_MSI_WINDOW,
   1988                                &sphb->msiwindow);
   1989
   1990    pci_setup_iommu(bus, spapr_pci_dma_iommu, sphb);
   1991
   1992    pci_bus_set_route_irq_fn(bus, spapr_route_intx_pin_to_irq);
   1993
   1994    QLIST_INSERT_HEAD(&spapr->phbs, sphb, list);
   1995
   1996    /* Initialize the LSI table */
   1997    for (i = 0; i < PCI_NUM_PINS; i++) {
   1998        int irq = SPAPR_IRQ_PCI_LSI + sphb->index * PCI_NUM_PINS + i;
   1999
   2000        if (smc->legacy_irq_allocation) {
   2001            irq = spapr_irq_findone(spapr, errp);
   2002            if (irq < 0) {
   2003                error_prepend(errp, "can't allocate LSIs: ");
   2004                /*
   2005                 * Older machines will never support PHB hotplug, ie, this is an
   2006                 * init only path and QEMU will terminate. No need to rollback.
   2007                 */
   2008                return;
   2009            }
   2010        }
   2011
   2012        if (spapr_irq_claim(spapr, irq, true, errp) < 0) {
   2013            error_prepend(errp, "can't allocate LSIs: ");
   2014            goto unrealize;
   2015        }
   2016
   2017        sphb->lsi_table[i].irq = irq;
   2018    }
   2019
   2020    /* allocate connectors for child PCI devices */
   2021    add_drcs(sphb, phb->bus);
   2022
   2023    /* DMA setup */
   2024    for (i = 0; i < windows_supported; ++i) {
   2025        tcet = spapr_tce_new_table(DEVICE(sphb), sphb->dma_liobn[i]);
   2026        if (!tcet) {
   2027            error_setg(errp, "Creating window#%d failed for %s",
   2028                       i, sphb->dtbusname);
   2029            goto unrealize;
   2030        }
   2031        memory_region_add_subregion(&sphb->iommu_root, 0,
   2032                                    spapr_tce_get_iommu(tcet));
   2033    }
   2034
   2035    sphb->msi = g_hash_table_new_full(g_int_hash, g_int_equal, g_free,
   2036                                      spapr_phb_destroy_msi);
   2037    return;
   2038
   2039unrealize:
   2040    spapr_phb_unrealize(dev);
   2041}
   2042
   2043static int spapr_phb_children_reset(Object *child, void *opaque)
   2044{
   2045    DeviceState *dev = (DeviceState *) object_dynamic_cast(child, TYPE_DEVICE);
   2046
   2047    if (dev) {
   2048        device_legacy_reset(dev);
   2049    }
   2050
   2051    return 0;
   2052}
   2053
   2054void spapr_phb_dma_reset(SpaprPhbState *sphb)
   2055{
   2056    int i;
   2057    SpaprTceTable *tcet;
   2058
   2059    for (i = 0; i < SPAPR_PCI_DMA_MAX_WINDOWS; ++i) {
   2060        tcet = spapr_tce_find_by_liobn(sphb->dma_liobn[i]);
   2061
   2062        if (tcet && tcet->nb_table) {
   2063            spapr_tce_table_disable(tcet);
   2064        }
   2065    }
   2066
   2067    /* Register default 32bit DMA window */
   2068    tcet = spapr_tce_find_by_liobn(sphb->dma_liobn[0]);
   2069    spapr_tce_table_enable(tcet, SPAPR_TCE_PAGE_SHIFT, sphb->dma_win_addr,
   2070                           sphb->dma_win_size >> SPAPR_TCE_PAGE_SHIFT);
   2071}
   2072
   2073static void spapr_phb_reset(DeviceState *qdev)
   2074{
   2075    SpaprPhbState *sphb = SPAPR_PCI_HOST_BRIDGE(qdev);
   2076    Error *err = NULL;
   2077
   2078    spapr_phb_dma_reset(sphb);
   2079    spapr_phb_nvgpu_free(sphb);
   2080    spapr_phb_nvgpu_setup(sphb, &err);
   2081    if (err) {
   2082        error_report_err(err);
   2083    }
   2084
   2085    /* Reset the IOMMU state */
   2086    object_child_foreach(OBJECT(qdev), spapr_phb_children_reset, NULL);
   2087
   2088    if (spapr_phb_eeh_available(SPAPR_PCI_HOST_BRIDGE(qdev))) {
   2089        spapr_phb_vfio_reset(qdev);
   2090    }
   2091
   2092    g_hash_table_remove_all(sphb->msi);
   2093}
   2094
   2095static Property spapr_phb_properties[] = {
   2096    DEFINE_PROP_UINT32("index", SpaprPhbState, index, -1),
   2097    DEFINE_PROP_UINT64("mem_win_size", SpaprPhbState, mem_win_size,
   2098                       SPAPR_PCI_MEM32_WIN_SIZE),
   2099    DEFINE_PROP_UINT64("mem64_win_size", SpaprPhbState, mem64_win_size,
   2100                       SPAPR_PCI_MEM64_WIN_SIZE),
   2101    DEFINE_PROP_UINT64("io_win_size", SpaprPhbState, io_win_size,
   2102                       SPAPR_PCI_IO_WIN_SIZE),
   2103    DEFINE_PROP_BOOL("dynamic-reconfiguration", SpaprPhbState, dr_enabled,
   2104                     true),
   2105    /* Default DMA window is 0..1GB */
   2106    DEFINE_PROP_UINT64("dma_win_addr", SpaprPhbState, dma_win_addr, 0),
   2107    DEFINE_PROP_UINT64("dma_win_size", SpaprPhbState, dma_win_size, 0x40000000),
   2108    DEFINE_PROP_UINT64("dma64_win_addr", SpaprPhbState, dma64_win_addr,
   2109                       0x800000000000000ULL),
   2110    DEFINE_PROP_BOOL("ddw", SpaprPhbState, ddw_enabled, true),
   2111    DEFINE_PROP_UINT64("pgsz", SpaprPhbState, page_size_mask,
   2112                       (1ULL << 12) | (1ULL << 16)
   2113                       | (1ULL << 21) | (1ULL << 24)),
   2114    DEFINE_PROP_UINT32("numa_node", SpaprPhbState, numa_node, -1),
   2115    DEFINE_PROP_BOOL("pre-2.8-migration", SpaprPhbState,
   2116                     pre_2_8_migration, false),
   2117    DEFINE_PROP_BOOL("pcie-extended-configuration-space", SpaprPhbState,
   2118                     pcie_ecs, true),
   2119    DEFINE_PROP_UINT64("gpa", SpaprPhbState, nv2_gpa_win_addr, 0),
   2120    DEFINE_PROP_UINT64("atsd", SpaprPhbState, nv2_atsd_win_addr, 0),
   2121    DEFINE_PROP_BOOL("pre-5.1-associativity", SpaprPhbState,
   2122                     pre_5_1_assoc, false),
   2123    DEFINE_PROP_END_OF_LIST(),
   2124};
   2125
   2126static const VMStateDescription vmstate_spapr_pci_lsi = {
   2127    .name = "spapr_pci/lsi",
   2128    .version_id = 1,
   2129    .minimum_version_id = 1,
   2130    .fields = (VMStateField[]) {
   2131        VMSTATE_UINT32_EQUAL(irq, SpaprPciLsi, NULL),
   2132
   2133        VMSTATE_END_OF_LIST()
   2134    },
   2135};
   2136
   2137static const VMStateDescription vmstate_spapr_pci_msi = {
   2138    .name = "spapr_pci/msi",
   2139    .version_id = 1,
   2140    .minimum_version_id = 1,
   2141    .fields = (VMStateField []) {
   2142        VMSTATE_UINT32(key, SpaprPciMsiMig),
   2143        VMSTATE_UINT32(value.first_irq, SpaprPciMsiMig),
   2144        VMSTATE_UINT32(value.num, SpaprPciMsiMig),
   2145        VMSTATE_END_OF_LIST()
   2146    },
   2147};
   2148
   2149static int spapr_pci_pre_save(void *opaque)
   2150{
   2151    SpaprPhbState *sphb = opaque;
   2152    GHashTableIter iter;
   2153    gpointer key, value;
   2154    int i;
   2155
   2156    if (sphb->pre_2_8_migration) {
   2157        sphb->mig_liobn = sphb->dma_liobn[0];
   2158        sphb->mig_mem_win_addr = sphb->mem_win_addr;
   2159        sphb->mig_mem_win_size = sphb->mem_win_size;
   2160        sphb->mig_io_win_addr = sphb->io_win_addr;
   2161        sphb->mig_io_win_size = sphb->io_win_size;
   2162
   2163        if ((sphb->mem64_win_size != 0)
   2164            && (sphb->mem64_win_addr
   2165                == (sphb->mem_win_addr + sphb->mem_win_size))) {
   2166            sphb->mig_mem_win_size += sphb->mem64_win_size;
   2167        }
   2168    }
   2169
   2170    g_free(sphb->msi_devs);
   2171    sphb->msi_devs = NULL;
   2172    sphb->msi_devs_num = g_hash_table_size(sphb->msi);
   2173    if (!sphb->msi_devs_num) {
   2174        return 0;
   2175    }
   2176    sphb->msi_devs = g_new(SpaprPciMsiMig, sphb->msi_devs_num);
   2177
   2178    g_hash_table_iter_init(&iter, sphb->msi);
   2179    for (i = 0; g_hash_table_iter_next(&iter, &key, &value); ++i) {
   2180        sphb->msi_devs[i].key = *(uint32_t *) key;
   2181        sphb->msi_devs[i].value = *(SpaprPciMsi *) value;
   2182    }
   2183
   2184    return 0;
   2185}
   2186
   2187static int spapr_pci_post_save(void *opaque)
   2188{
   2189    SpaprPhbState *sphb = opaque;
   2190
   2191    g_free(sphb->msi_devs);
   2192    sphb->msi_devs = NULL;
   2193    sphb->msi_devs_num = 0;
   2194    return 0;
   2195}
   2196
   2197static int spapr_pci_post_load(void *opaque, int version_id)
   2198{
   2199    SpaprPhbState *sphb = opaque;
   2200    gpointer key, value;
   2201    int i;
   2202
   2203    for (i = 0; i < sphb->msi_devs_num; ++i) {
   2204        key = g_memdup(&sphb->msi_devs[i].key,
   2205                       sizeof(sphb->msi_devs[i].key));
   2206        value = g_memdup(&sphb->msi_devs[i].value,
   2207                         sizeof(sphb->msi_devs[i].value));
   2208        g_hash_table_insert(sphb->msi, key, value);
   2209    }
   2210    g_free(sphb->msi_devs);
   2211    sphb->msi_devs = NULL;
   2212    sphb->msi_devs_num = 0;
   2213
   2214    return 0;
   2215}
   2216
   2217static bool pre_2_8_migration(void *opaque, int version_id)
   2218{
   2219    SpaprPhbState *sphb = opaque;
   2220
   2221    return sphb->pre_2_8_migration;
   2222}
   2223
   2224static const VMStateDescription vmstate_spapr_pci = {
   2225    .name = "spapr_pci",
   2226    .version_id = 2,
   2227    .minimum_version_id = 2,
   2228    .pre_save = spapr_pci_pre_save,
   2229    .post_save = spapr_pci_post_save,
   2230    .post_load = spapr_pci_post_load,
   2231    .fields = (VMStateField[]) {
   2232        VMSTATE_UINT64_EQUAL(buid, SpaprPhbState, NULL),
   2233        VMSTATE_UINT32_TEST(mig_liobn, SpaprPhbState, pre_2_8_migration),
   2234        VMSTATE_UINT64_TEST(mig_mem_win_addr, SpaprPhbState, pre_2_8_migration),
   2235        VMSTATE_UINT64_TEST(mig_mem_win_size, SpaprPhbState, pre_2_8_migration),
   2236        VMSTATE_UINT64_TEST(mig_io_win_addr, SpaprPhbState, pre_2_8_migration),
   2237        VMSTATE_UINT64_TEST(mig_io_win_size, SpaprPhbState, pre_2_8_migration),
   2238        VMSTATE_STRUCT_ARRAY(lsi_table, SpaprPhbState, PCI_NUM_PINS, 0,
   2239                             vmstate_spapr_pci_lsi, SpaprPciLsi),
   2240        VMSTATE_INT32(msi_devs_num, SpaprPhbState),
   2241        VMSTATE_STRUCT_VARRAY_ALLOC(msi_devs, SpaprPhbState, msi_devs_num, 0,
   2242                                    vmstate_spapr_pci_msi, SpaprPciMsiMig),
   2243        VMSTATE_END_OF_LIST()
   2244    },
   2245};
   2246
   2247static const char *spapr_phb_root_bus_path(PCIHostState *host_bridge,
   2248                                           PCIBus *rootbus)
   2249{
   2250    SpaprPhbState *sphb = SPAPR_PCI_HOST_BRIDGE(host_bridge);
   2251
   2252    return sphb->dtbusname;
   2253}
   2254
   2255static void spapr_phb_class_init(ObjectClass *klass, void *data)
   2256{
   2257    PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass);
   2258    DeviceClass *dc = DEVICE_CLASS(klass);
   2259    HotplugHandlerClass *hp = HOTPLUG_HANDLER_CLASS(klass);
   2260
   2261    hc->root_bus_path = spapr_phb_root_bus_path;
   2262    dc->realize = spapr_phb_realize;
   2263    dc->unrealize = spapr_phb_unrealize;
   2264    device_class_set_props(dc, spapr_phb_properties);
   2265    dc->reset = spapr_phb_reset;
   2266    dc->vmsd = &vmstate_spapr_pci;
   2267    /* Supported by TYPE_SPAPR_MACHINE */
   2268    dc->user_creatable = true;
   2269    set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
   2270    hp->pre_plug = spapr_pci_pre_plug;
   2271    hp->plug = spapr_pci_plug;
   2272    hp->unplug = spapr_pci_unplug;
   2273    hp->unplug_request = spapr_pci_unplug_request;
   2274}
   2275
   2276static const TypeInfo spapr_phb_info = {
   2277    .name          = TYPE_SPAPR_PCI_HOST_BRIDGE,
   2278    .parent        = TYPE_PCI_HOST_BRIDGE,
   2279    .instance_size = sizeof(SpaprPhbState),
   2280    .instance_finalize = spapr_phb_finalizefn,
   2281    .class_init    = spapr_phb_class_init,
   2282    .interfaces    = (InterfaceInfo[]) {
   2283        { TYPE_HOTPLUG_HANDLER },
   2284        { }
   2285    }
   2286};
   2287
   2288static void spapr_phb_pci_enumerate_bridge(PCIBus *bus, PCIDevice *pdev,
   2289                                           void *opaque)
   2290{
   2291    unsigned int *bus_no = opaque;
   2292    PCIBus *sec_bus = NULL;
   2293
   2294    if ((pci_default_read_config(pdev, PCI_HEADER_TYPE, 1) !=
   2295         PCI_HEADER_TYPE_BRIDGE)) {
   2296        return;
   2297    }
   2298
   2299    (*bus_no)++;
   2300    pci_default_write_config(pdev, PCI_PRIMARY_BUS, pci_dev_bus_num(pdev), 1);
   2301    pci_default_write_config(pdev, PCI_SECONDARY_BUS, *bus_no, 1);
   2302    pci_default_write_config(pdev, PCI_SUBORDINATE_BUS, *bus_no, 1);
   2303
   2304    sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
   2305    if (!sec_bus) {
   2306        return;
   2307    }
   2308
   2309    pci_for_each_device(sec_bus, pci_bus_num(sec_bus),
   2310                        spapr_phb_pci_enumerate_bridge, bus_no);
   2311    pci_default_write_config(pdev, PCI_SUBORDINATE_BUS, *bus_no, 1);
   2312}
   2313
   2314static void spapr_phb_pci_enumerate(SpaprPhbState *phb)
   2315{
   2316    PCIBus *bus = PCI_HOST_BRIDGE(phb)->bus;
   2317    unsigned int bus_no = 0;
   2318
   2319    pci_for_each_device(bus, pci_bus_num(bus),
   2320                        spapr_phb_pci_enumerate_bridge,
   2321                        &bus_no);
   2322
   2323}
   2324
   2325int spapr_dt_phb(SpaprMachineState *spapr, SpaprPhbState *phb,
   2326                 uint32_t intc_phandle, void *fdt, int *node_offset)
   2327{
   2328    int bus_off, i, j, ret;
   2329    uint32_t bus_range[] = { cpu_to_be32(0), cpu_to_be32(0xff) };
   2330    struct {
   2331        uint32_t hi;
   2332        uint64_t child;
   2333        uint64_t parent;
   2334        uint64_t size;
   2335    } QEMU_PACKED ranges[] = {
   2336        {
   2337            cpu_to_be32(b_ss(1)), cpu_to_be64(0),
   2338            cpu_to_be64(phb->io_win_addr),
   2339            cpu_to_be64(memory_region_size(&phb->iospace)),
   2340        },
   2341        {
   2342            cpu_to_be32(b_ss(2)), cpu_to_be64(SPAPR_PCI_MEM_WIN_BUS_OFFSET),
   2343            cpu_to_be64(phb->mem_win_addr),
   2344            cpu_to_be64(phb->mem_win_size),
   2345        },
   2346        {
   2347            cpu_to_be32(b_ss(3)), cpu_to_be64(phb->mem64_win_pciaddr),
   2348            cpu_to_be64(phb->mem64_win_addr),
   2349            cpu_to_be64(phb->mem64_win_size),
   2350        },
   2351    };
   2352    const unsigned sizeof_ranges =
   2353        (phb->mem64_win_size ? 3 : 2) * sizeof(ranges[0]);
   2354    uint64_t bus_reg[] = { cpu_to_be64(phb->buid), 0 };
   2355    uint32_t interrupt_map_mask[] = {
   2356        cpu_to_be32(b_ddddd(-1)|b_fff(0)), 0x0, 0x0, cpu_to_be32(-1)};
   2357    uint32_t interrupt_map[PCI_SLOT_MAX * PCI_NUM_PINS][7];
   2358    uint32_t ddw_applicable[] = {
   2359        cpu_to_be32(RTAS_IBM_QUERY_PE_DMA_WINDOW),
   2360        cpu_to_be32(RTAS_IBM_CREATE_PE_DMA_WINDOW),
   2361        cpu_to_be32(RTAS_IBM_REMOVE_PE_DMA_WINDOW)
   2362    };
   2363    uint32_t ddw_extensions[] = {
   2364        cpu_to_be32(1),
   2365        cpu_to_be32(RTAS_IBM_RESET_PE_DMA_WINDOW)
   2366    };
   2367    SpaprTceTable *tcet;
   2368    SpaprDrc *drc;
   2369    Error *err = NULL;
   2370
   2371    /* Start populating the FDT */
   2372    _FDT(bus_off = fdt_add_subnode(fdt, 0, phb->dtbusname));
   2373    if (node_offset) {
   2374        *node_offset = bus_off;
   2375    }
   2376
   2377    /* Write PHB properties */
   2378    _FDT(fdt_setprop_string(fdt, bus_off, "device_type", "pci"));
   2379    _FDT(fdt_setprop_string(fdt, bus_off, "compatible", "IBM,Logical_PHB"));
   2380    _FDT(fdt_setprop_cell(fdt, bus_off, "#interrupt-cells", 0x1));
   2381    _FDT(fdt_setprop(fdt, bus_off, "used-by-rtas", NULL, 0));
   2382    _FDT(fdt_setprop(fdt, bus_off, "bus-range", &bus_range, sizeof(bus_range)));
   2383    _FDT(fdt_setprop(fdt, bus_off, "ranges", &ranges, sizeof_ranges));
   2384    _FDT(fdt_setprop(fdt, bus_off, "reg", &bus_reg, sizeof(bus_reg)));
   2385    _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pci-config-space-type", 0x1));
   2386    _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pe-total-#msi",
   2387                          spapr_irq_nr_msis(spapr)));
   2388
   2389    /* Dynamic DMA window */
   2390    if (phb->ddw_enabled) {
   2391        _FDT(fdt_setprop(fdt, bus_off, "ibm,ddw-applicable", &ddw_applicable,
   2392                         sizeof(ddw_applicable)));
   2393        _FDT(fdt_setprop(fdt, bus_off, "ibm,ddw-extensions",
   2394                         &ddw_extensions, sizeof(ddw_extensions)));
   2395    }
   2396
   2397    /* Advertise NUMA via ibm,associativity */
   2398    if (phb->numa_node != -1) {
   2399        spapr_numa_write_associativity_dt(spapr, fdt, bus_off, phb->numa_node);
   2400    }
   2401
   2402    /* Build the interrupt-map, this must matches what is done
   2403     * in pci_swizzle_map_irq_fn
   2404     */
   2405    _FDT(fdt_setprop(fdt, bus_off, "interrupt-map-mask",
   2406                     &interrupt_map_mask, sizeof(interrupt_map_mask)));
   2407    for (i = 0; i < PCI_SLOT_MAX; i++) {
   2408        for (j = 0; j < PCI_NUM_PINS; j++) {
   2409            uint32_t *irqmap = interrupt_map[i*PCI_NUM_PINS + j];
   2410            int lsi_num = pci_swizzle(i, j);
   2411
   2412            irqmap[0] = cpu_to_be32(b_ddddd(i)|b_fff(0));
   2413            irqmap[1] = 0;
   2414            irqmap[2] = 0;
   2415            irqmap[3] = cpu_to_be32(j+1);
   2416            irqmap[4] = cpu_to_be32(intc_phandle);
   2417            spapr_dt_irq(&irqmap[5], phb->lsi_table[lsi_num].irq, true);
   2418        }
   2419    }
   2420    /* Write interrupt map */
   2421    _FDT(fdt_setprop(fdt, bus_off, "interrupt-map", &interrupt_map,
   2422                     sizeof(interrupt_map)));
   2423
   2424    tcet = spapr_tce_find_by_liobn(phb->dma_liobn[0]);
   2425    if (!tcet) {
   2426        return -1;
   2427    }
   2428    spapr_dma_dt(fdt, bus_off, "ibm,dma-window",
   2429                 tcet->liobn, tcet->bus_offset,
   2430                 tcet->nb_table << tcet->page_shift);
   2431
   2432    drc = spapr_drc_by_id(TYPE_SPAPR_DRC_PHB, phb->index);
   2433    if (drc) {
   2434        uint32_t drc_index = cpu_to_be32(spapr_drc_index(drc));
   2435
   2436        _FDT(fdt_setprop(fdt, bus_off, "ibm,my-drc-index", &drc_index,
   2437                         sizeof(drc_index)));
   2438    }
   2439
   2440    /* Walk the bridges and program the bus numbers*/
   2441    spapr_phb_pci_enumerate(phb);
   2442    _FDT(fdt_setprop_cell(fdt, bus_off, "qemu,phb-enumerated", 0x1));
   2443
   2444    /* Walk the bridge and subordinate buses */
   2445    ret = spapr_dt_pci_bus(phb, PCI_HOST_BRIDGE(phb)->bus, fdt, bus_off);
   2446    if (ret < 0) {
   2447        return ret;
   2448    }
   2449
   2450    spapr_phb_nvgpu_populate_dt(phb, fdt, bus_off, &err);
   2451    if (err) {
   2452        error_report_err(err);
   2453    }
   2454    spapr_phb_nvgpu_ram_populate_dt(phb, fdt);
   2455
   2456    return 0;
   2457}
   2458
   2459void spapr_pci_rtas_init(void)
   2460{
   2461    spapr_rtas_register(RTAS_READ_PCI_CONFIG, "read-pci-config",
   2462                        rtas_read_pci_config);
   2463    spapr_rtas_register(RTAS_WRITE_PCI_CONFIG, "write-pci-config",
   2464                        rtas_write_pci_config);
   2465    spapr_rtas_register(RTAS_IBM_READ_PCI_CONFIG, "ibm,read-pci-config",
   2466                        rtas_ibm_read_pci_config);
   2467    spapr_rtas_register(RTAS_IBM_WRITE_PCI_CONFIG, "ibm,write-pci-config",
   2468                        rtas_ibm_write_pci_config);
   2469    if (msi_nonbroken) {
   2470        spapr_rtas_register(RTAS_IBM_QUERY_INTERRUPT_SOURCE_NUMBER,
   2471                            "ibm,query-interrupt-source-number",
   2472                            rtas_ibm_query_interrupt_source_number);
   2473        spapr_rtas_register(RTAS_IBM_CHANGE_MSI, "ibm,change-msi",
   2474                            rtas_ibm_change_msi);
   2475    }
   2476
   2477    spapr_rtas_register(RTAS_IBM_SET_EEH_OPTION,
   2478                        "ibm,set-eeh-option",
   2479                        rtas_ibm_set_eeh_option);
   2480    spapr_rtas_register(RTAS_IBM_GET_CONFIG_ADDR_INFO2,
   2481                        "ibm,get-config-addr-info2",
   2482                        rtas_ibm_get_config_addr_info2);
   2483    spapr_rtas_register(RTAS_IBM_READ_SLOT_RESET_STATE2,
   2484                        "ibm,read-slot-reset-state2",
   2485                        rtas_ibm_read_slot_reset_state2);
   2486    spapr_rtas_register(RTAS_IBM_SET_SLOT_RESET,
   2487                        "ibm,set-slot-reset",
   2488                        rtas_ibm_set_slot_reset);
   2489    spapr_rtas_register(RTAS_IBM_CONFIGURE_PE,
   2490                        "ibm,configure-pe",
   2491                        rtas_ibm_configure_pe);
   2492    spapr_rtas_register(RTAS_IBM_SLOT_ERROR_DETAIL,
   2493                        "ibm,slot-error-detail",
   2494                        rtas_ibm_slot_error_detail);
   2495}
   2496
   2497static void spapr_pci_register_types(void)
   2498{
   2499    type_register_static(&spapr_phb_info);
   2500}
   2501
   2502type_init(spapr_pci_register_types)
   2503
   2504static int spapr_switch_one_vga(DeviceState *dev, void *opaque)
   2505{
   2506    bool be = *(bool *)opaque;
   2507
   2508    if (object_dynamic_cast(OBJECT(dev), "VGA")
   2509        || object_dynamic_cast(OBJECT(dev), "secondary-vga")
   2510        || object_dynamic_cast(OBJECT(dev), "bochs-display")
   2511        || object_dynamic_cast(OBJECT(dev), "virtio-vga")) {
   2512        object_property_set_bool(OBJECT(dev), "big-endian-framebuffer", be,
   2513                                 &error_abort);
   2514    }
   2515    return 0;
   2516}
   2517
   2518void spapr_pci_switch_vga(SpaprMachineState *spapr, bool big_endian)
   2519{
   2520    SpaprPhbState *sphb;
   2521
   2522    /*
   2523     * For backward compatibility with existing guests, we switch
   2524     * the endianness of the VGA controller when changing the guest
   2525     * interrupt mode
   2526     */
   2527    QLIST_FOREACH(sphb, &spapr->phbs, list) {
   2528        BusState *bus = &PCI_HOST_BRIDGE(sphb)->bus->qbus;
   2529        qbus_walk_children(bus, spapr_switch_one_vga, NULL, NULL, NULL,
   2530                           &big_endian);
   2531    }
   2532}