aboutsummaryrefslogtreecommitdiff
path: root/hw/pci-host
diff options
context:
space:
mode:
Diffstat (limited to 'hw/pci-host')
-rw-r--r--hw/pci-host/Makefile.objs2
-rw-r--r--hw/pci-host/pnv_phb3.c1197
-rw-r--r--hw/pci-host/pnv_phb3_msi.c349
-rw-r--r--hw/pci-host/pnv_phb3_pbcq.c358
-rw-r--r--hw/pci-host/pnv_phb4.c1439
-rw-r--r--hw/pci-host/pnv_phb4_pec.c595
6 files changed, 3940 insertions, 0 deletions
diff --git a/hw/pci-host/Makefile.objs b/hw/pci-host/Makefile.objs
index 9c466fab01..8c87e8494d 100644
--- a/hw/pci-host/Makefile.objs
+++ b/hw/pci-host/Makefile.objs
@@ -20,3 +20,5 @@ common-obj-$(CONFIG_PCI_EXPRESS_GENERIC_BRIDGE) += gpex.o
common-obj-$(CONFIG_PCI_EXPRESS_XILINX) += xilinx-pcie.o
common-obj-$(CONFIG_PCI_EXPRESS_DESIGNWARE) += designware.o
+obj-$(CONFIG_POWERNV) += pnv_phb4.o pnv_phb4_pec.o
+obj-$(CONFIG_POWERNV) += pnv_phb3.o pnv_phb3_msi.o pnv_phb3_pbcq.o
diff --git a/hw/pci-host/pnv_phb3.c b/hw/pci-host/pnv_phb3.c
new file mode 100644
index 0000000000..74618fadf0
--- /dev/null
+++ b/hw/pci-host/pnv_phb3.c
@@ -0,0 +1,1197 @@
+/*
+ * QEMU PowerPC PowerNV (POWER8) PHB3 model
+ *
+ * Copyright (c) 2014-2020, IBM Corporation.
+ *
+ * This code is licensed under the GPL version 2 or later. See the
+ * COPYING file in the top-level directory.
+ */
+#include "qemu/osdep.h"
+#include "qemu/log.h"
+#include "qapi/visitor.h"
+#include "qapi/error.h"
+#include "qemu-common.h"
+#include "hw/pci-host/pnv_phb3_regs.h"
+#include "hw/pci-host/pnv_phb3.h"
+#include "hw/pci/pcie_host.h"
+#include "hw/pci/pcie_port.h"
+#include "hw/ppc/pnv.h"
+#include "hw/irq.h"
+#include "hw/qdev-properties.h"
+
+#define phb3_error(phb, fmt, ...) \
+ qemu_log_mask(LOG_GUEST_ERROR, "phb3[%d:%d]: " fmt "\n", \
+ (phb)->chip_id, (phb)->phb_id, ## __VA_ARGS__)
+
+static PCIDevice *pnv_phb3_find_cfg_dev(PnvPHB3 *phb)
+{
+ PCIHostState *pci = PCI_HOST_BRIDGE(phb);
+ uint64_t addr = phb->regs[PHB_CONFIG_ADDRESS >> 3];
+ uint8_t bus, devfn;
+
+ if (!(addr >> 63)) {
+ return NULL;
+ }
+ bus = (addr >> 52) & 0xff;
+ devfn = (addr >> 44) & 0xff;
+
+ return pci_find_device(pci->bus, bus, devfn);
+}
+
+/*
+ * The CONFIG_DATA register expects little endian accesses, but as the
+ * region is big endian, we have to swap the value.
+ */
+static void pnv_phb3_config_write(PnvPHB3 *phb, unsigned off,
+ unsigned size, uint64_t val)
+{
+ uint32_t cfg_addr, limit;
+ PCIDevice *pdev;
+
+ pdev = pnv_phb3_find_cfg_dev(phb);
+ if (!pdev) {
+ return;
+ }
+ cfg_addr = (phb->regs[PHB_CONFIG_ADDRESS >> 3] >> 32) & 0xffc;
+ cfg_addr |= off;
+ limit = pci_config_size(pdev);
+ if (limit <= cfg_addr) {
+ /*
+ * conventional pci device can be behind pcie-to-pci bridge.
+ * 256 <= addr < 4K has no effects.
+ */
+ return;
+ }
+ switch (size) {
+ case 1:
+ break;
+ case 2:
+ val = bswap16(val);
+ break;
+ case 4:
+ val = bswap32(val);
+ break;
+ default:
+ g_assert_not_reached();
+ }
+ pci_host_config_write_common(pdev, cfg_addr, limit, val, size);
+}
+
+static uint64_t pnv_phb3_config_read(PnvPHB3 *phb, unsigned off,
+ unsigned size)
+{
+ uint32_t cfg_addr, limit;
+ PCIDevice *pdev;
+ uint64_t val;
+
+ pdev = pnv_phb3_find_cfg_dev(phb);
+ if (!pdev) {
+ return ~0ull;
+ }
+ cfg_addr = (phb->regs[PHB_CONFIG_ADDRESS >> 3] >> 32) & 0xffc;
+ cfg_addr |= off;
+ limit = pci_config_size(pdev);
+ if (limit <= cfg_addr) {
+ /*
+ * conventional pci device can be behind pcie-to-pci bridge.
+ * 256 <= addr < 4K has no effects.
+ */
+ return ~0ull;
+ }
+ val = pci_host_config_read_common(pdev, cfg_addr, limit, size);
+ switch (size) {
+ case 1:
+ return val;
+ case 2:
+ return bswap16(val);
+ case 4:
+ return bswap32(val);
+ default:
+ g_assert_not_reached();
+ }
+}
+
+static void pnv_phb3_check_m32(PnvPHB3 *phb)
+{
+ uint64_t base, start, size;
+ MemoryRegion *parent;
+ PnvPBCQState *pbcq = &phb->pbcq;
+
+ if (memory_region_is_mapped(&phb->mr_m32)) {
+ memory_region_del_subregion(phb->mr_m32.container, &phb->mr_m32);
+ }
+
+ if (!(phb->regs[PHB_PHB3_CONFIG >> 3] & PHB_PHB3C_M32_EN)) {
+ return;
+ }
+
+ /* Grab geometry from registers */
+ base = phb->regs[PHB_M32_BASE_ADDR >> 3];
+ start = phb->regs[PHB_M32_START_ADDR >> 3];
+ size = ~(phb->regs[PHB_M32_BASE_MASK >> 3] | 0xfffc000000000000ull) + 1;
+
+ /* Check if it matches an enabled MMIO region in the PBCQ */
+ if (memory_region_is_mapped(&pbcq->mmbar0) &&
+ base >= pbcq->mmio0_base &&
+ (base + size) <= (pbcq->mmio0_base + pbcq->mmio0_size)) {
+ parent = &pbcq->mmbar0;
+ base -= pbcq->mmio0_base;
+ } else if (memory_region_is_mapped(&pbcq->mmbar1) &&
+ base >= pbcq->mmio1_base &&
+ (base + size) <= (pbcq->mmio1_base + pbcq->mmio1_size)) {
+ parent = &pbcq->mmbar1;
+ base -= pbcq->mmio1_base;
+ } else {
+ return;
+ }
+
+ /* Create alias */
+ memory_region_init_alias(&phb->mr_m32, OBJECT(phb), "phb3-m32",
+ &phb->pci_mmio, start, size);
+ memory_region_add_subregion(parent, base, &phb->mr_m32);
+}
+
+static void pnv_phb3_check_m64(PnvPHB3 *phb, uint32_t index)
+{
+ uint64_t base, start, size, m64;
+ MemoryRegion *parent;
+ PnvPBCQState *pbcq = &phb->pbcq;
+
+ if (memory_region_is_mapped(&phb->mr_m64[index])) {
+ /* Should we destroy it in RCU friendly way... ? */
+ memory_region_del_subregion(phb->mr_m64[index].container,
+ &phb->mr_m64[index]);
+ }
+
+ /* Get table entry */
+ m64 = phb->ioda_M64BT[index];
+
+ if (!(m64 & IODA2_M64BT_ENABLE)) {
+ return;
+ }
+
+ /* Grab geometry from registers */
+ base = GETFIELD(IODA2_M64BT_BASE, m64) << 20;
+ if (m64 & IODA2_M64BT_SINGLE_PE) {
+ base &= ~0x1ffffffull;
+ }
+ size = GETFIELD(IODA2_M64BT_MASK, m64) << 20;
+ size |= 0xfffc000000000000ull;
+ size = ~size + 1;
+ start = base | (phb->regs[PHB_M64_UPPER_BITS >> 3]);
+
+ /* Check if it matches an enabled MMIO region in the PBCQ */
+ if (memory_region_is_mapped(&pbcq->mmbar0) &&
+ base >= pbcq->mmio0_base &&
+ (base + size) <= (pbcq->mmio0_base + pbcq->mmio0_size)) {
+ parent = &pbcq->mmbar0;
+ base -= pbcq->mmio0_base;
+ } else if (memory_region_is_mapped(&pbcq->mmbar1) &&
+ base >= pbcq->mmio1_base &&
+ (base + size) <= (pbcq->mmio1_base + pbcq->mmio1_size)) {
+ parent = &pbcq->mmbar1;
+ base -= pbcq->mmio1_base;
+ } else {
+ return;
+ }
+
+ /* Create alias */
+ memory_region_init_alias(&phb->mr_m64[index], OBJECT(phb), "phb3-m64",
+ &phb->pci_mmio, start, size);
+ memory_region_add_subregion(parent, base, &phb->mr_m64[index]);
+}
+
+static void pnv_phb3_check_all_m64s(PnvPHB3 *phb)
+{
+ uint64_t i;
+
+ for (i = 0; i < PNV_PHB3_NUM_M64; i++) {
+ pnv_phb3_check_m64(phb, i);
+ }
+}
+
+static void pnv_phb3_lxivt_write(PnvPHB3 *phb, unsigned idx, uint64_t val)
+{
+ uint8_t server, prio;
+
+ phb->ioda_LXIVT[idx] = val & (IODA2_LXIVT_SERVER |
+ IODA2_LXIVT_PRIORITY |
+ IODA2_LXIVT_NODE_ID);
+ server = GETFIELD(IODA2_LXIVT_SERVER, val);
+ prio = GETFIELD(IODA2_LXIVT_PRIORITY, val);
+
+ /*
+ * The low order 2 bits are the link pointer (Type II interrupts).
+ * Shift back to get a valid IRQ server.
+ */
+ server >>= 2;
+
+ ics_write_xive(&phb->lsis, idx, server, prio, prio);
+}
+
+static uint64_t *pnv_phb3_ioda_access(PnvPHB3 *phb,
+ unsigned *out_table, unsigned *out_idx)
+{
+ uint64_t adreg = phb->regs[PHB_IODA_ADDR >> 3];
+ unsigned int index = GETFIELD(PHB_IODA_AD_TADR, adreg);
+ unsigned int table = GETFIELD(PHB_IODA_AD_TSEL, adreg);
+ unsigned int mask;
+ uint64_t *tptr = NULL;
+
+ switch (table) {
+ case IODA2_TBL_LIST:
+ tptr = phb->ioda_LIST;
+ mask = 7;
+ break;
+ case IODA2_TBL_LXIVT:
+ tptr = phb->ioda_LXIVT;
+ mask = 7;
+ break;
+ case IODA2_TBL_IVC_CAM:
+ case IODA2_TBL_RBA:
+ mask = 31;
+ break;
+ case IODA2_TBL_RCAM:
+ mask = 63;
+ break;
+ case IODA2_TBL_MRT:
+ mask = 7;
+ break;
+ case IODA2_TBL_PESTA:
+ case IODA2_TBL_PESTB:
+ mask = 255;
+ break;
+ case IODA2_TBL_TVT:
+ tptr = phb->ioda_TVT;
+ mask = 511;
+ break;
+ case IODA2_TBL_TCAM:
+ case IODA2_TBL_TDR:
+ mask = 63;
+ break;
+ case IODA2_TBL_M64BT:
+ tptr = phb->ioda_M64BT;
+ mask = 15;
+ break;
+ case IODA2_TBL_M32DT:
+ tptr = phb->ioda_MDT;
+ mask = 255;
+ break;
+ case IODA2_TBL_PEEV:
+ tptr = phb->ioda_PEEV;
+ mask = 3;
+ break;
+ default:
+ phb3_error(phb, "invalid IODA table %d", table);
+ return NULL;
+ }
+ index &= mask;
+ if (out_idx) {
+ *out_idx = index;
+ }
+ if (out_table) {
+ *out_table = table;
+ }
+ if (tptr) {
+ tptr += index;
+ }
+ if (adreg & PHB_IODA_AD_AUTOINC) {
+ index = (index + 1) & mask;
+ adreg = SETFIELD(PHB_IODA_AD_TADR, adreg, index);
+ }
+ phb->regs[PHB_IODA_ADDR >> 3] = adreg;
+ return tptr;
+}
+
+static uint64_t pnv_phb3_ioda_read(PnvPHB3 *phb)
+{
+ unsigned table;
+ uint64_t *tptr;
+
+ tptr = pnv_phb3_ioda_access(phb, &table, NULL);
+ if (!tptr) {
+ /* Return 0 on unsupported tables, not ff's */
+ return 0;
+ }
+ return *tptr;
+}
+
+static void pnv_phb3_ioda_write(PnvPHB3 *phb, uint64_t val)
+{
+ unsigned table, idx;
+ uint64_t *tptr;
+
+ tptr = pnv_phb3_ioda_access(phb, &table, &idx);
+ if (!tptr) {
+ return;
+ }
+
+ /* Handle side effects */
+ switch (table) {
+ case IODA2_TBL_LXIVT:
+ pnv_phb3_lxivt_write(phb, idx, val);
+ break;
+ case IODA2_TBL_M64BT:
+ *tptr = val;
+ pnv_phb3_check_m64(phb, idx);
+ break;
+ default:
+ *tptr = val;
+ }
+}
+
+/*
+ * This is called whenever the PHB LSI, MSI source ID register or
+ * the PBCQ irq filters are written.
+ */
+void pnv_phb3_remap_irqs(PnvPHB3 *phb)
+{
+ ICSState *ics = &phb->lsis;
+ uint32_t local, global, count, mask, comp;
+ uint64_t baren;
+ PnvPBCQState *pbcq = &phb->pbcq;
+
+ /*
+ * First check if we are enabled. Unlike real HW we don't separate
+ * TX and RX so we enable if both are set
+ */
+ baren = pbcq->nest_regs[PBCQ_NEST_BAR_EN];
+ if (!(baren & PBCQ_NEST_BAR_EN_IRSN_RX) ||
+ !(baren & PBCQ_NEST_BAR_EN_IRSN_TX)) {
+ ics->offset = 0;
+ return;
+ }
+
+ /* Grab local LSI source ID */
+ local = GETFIELD(PHB_LSI_SRC_ID, phb->regs[PHB_LSI_SOURCE_ID >> 3]) << 3;
+
+ /* Grab global one and compare */
+ global = GETFIELD(PBCQ_NEST_LSI_SRC,
+ pbcq->nest_regs[PBCQ_NEST_LSI_SRC_ID]) << 3;
+ if (global != local) {
+ /*
+ * This happens during initialization, let's come back when we
+ * are properly configured
+ */
+ ics->offset = 0;
+ return;
+ }
+
+ /* Get the base on the powerbus */
+ comp = GETFIELD(PBCQ_NEST_IRSN_COMP,
+ pbcq->nest_regs[PBCQ_NEST_IRSN_COMPARE]);
+ mask = GETFIELD(PBCQ_NEST_IRSN_COMP,
+ pbcq->nest_regs[PBCQ_NEST_IRSN_MASK]);
+ count = ((~mask) + 1) & 0x7ffff;
+ phb->total_irq = count;
+
+ /* Sanity checks */
+ if ((global + PNV_PHB3_NUM_LSI) > count) {
+ phb3_error(phb, "LSIs out of reach: LSI base=%d total irq=%d", global,
+ count);
+ }
+
+ if (count > 2048) {
+ phb3_error(phb, "More interrupts than supported: %d", count);
+ }
+
+ if ((comp & mask) != comp) {
+ phb3_error(phb, "IRQ compare bits not in mask: comp=0x%x mask=0x%x",
+ comp, mask);
+ comp &= mask;
+ }
+ /* Setup LSI offset */
+ ics->offset = comp + global;
+
+ /* Setup MSI offset */
+ pnv_phb3_msi_update_config(&phb->msis, comp, count - PNV_PHB3_NUM_LSI);
+}
+
+static void pnv_phb3_lsi_src_id_write(PnvPHB3 *phb, uint64_t val)
+{
+ /* Sanitize content */
+ val &= PHB_LSI_SRC_ID;
+ phb->regs[PHB_LSI_SOURCE_ID >> 3] = val;
+ pnv_phb3_remap_irqs(phb);
+}
+
+static void pnv_phb3_rtc_invalidate(PnvPHB3 *phb, uint64_t val)
+{
+ PnvPhb3DMASpace *ds;
+
+ /* Always invalidate all for now ... */
+ QLIST_FOREACH(ds, &phb->dma_spaces, list) {
+ ds->pe_num = PHB_INVALID_PE;
+ }
+}
+
+
+static void pnv_phb3_update_msi_regions(PnvPhb3DMASpace *ds)
+{
+ uint64_t cfg = ds->phb->regs[PHB_PHB3_CONFIG >> 3];
+
+ if (cfg & PHB_PHB3C_32BIT_MSI_EN) {
+ if (!memory_region_is_mapped(&ds->msi32_mr)) {
+ memory_region_add_subregion(MEMORY_REGION(&ds->dma_mr),
+ 0xffff0000, &ds->msi32_mr);
+ }
+ } else {
+ if (memory_region_is_mapped(&ds->msi32_mr)) {
+ memory_region_del_subregion(MEMORY_REGION(&ds->dma_mr),
+ &ds->msi32_mr);
+ }
+ }
+
+ if (cfg & PHB_PHB3C_64BIT_MSI_EN) {
+ if (!memory_region_is_mapped(&ds->msi64_mr)) {
+ memory_region_add_subregion(MEMORY_REGION(&ds->dma_mr),
+ (1ull << 60), &ds->msi64_mr);
+ }
+ } else {
+ if (memory_region_is_mapped(&ds->msi64_mr)) {
+ memory_region_del_subregion(MEMORY_REGION(&ds->dma_mr),
+ &ds->msi64_mr);
+ }
+ }
+}
+
+static void pnv_phb3_update_all_msi_regions(PnvPHB3 *phb)
+{
+ PnvPhb3DMASpace *ds;
+
+ QLIST_FOREACH(ds, &phb->dma_spaces, list) {
+ pnv_phb3_update_msi_regions(ds);
+ }
+}
+
+void pnv_phb3_reg_write(void *opaque, hwaddr off, uint64_t val, unsigned size)
+{
+ PnvPHB3 *phb = opaque;
+ bool changed;
+
+ /* Special case configuration data */
+ if ((off & 0xfffc) == PHB_CONFIG_DATA) {
+ pnv_phb3_config_write(phb, off & 0x3, size, val);
+ return;
+ }
+
+ /* Other registers are 64-bit only */
+ if (size != 8 || off & 0x7) {
+ phb3_error(phb, "Invalid register access, offset: 0x%"PRIx64" size: %d",
+ off, size);
+ return;
+ }
+
+ /* Handle masking & filtering */
+ switch (off) {
+ case PHB_M64_UPPER_BITS:
+ val &= 0xfffc000000000000ull;
+ break;
+ case PHB_Q_DMA_R:
+ /*
+ * This is enough logic to make SW happy but we aren't actually
+ * quiescing the DMAs
+ */
+ if (val & PHB_Q_DMA_R_AUTORESET) {
+ val = 0;
+ } else {
+ val &= PHB_Q_DMA_R_QUIESCE_DMA;
+ }
+ break;
+ /* LEM stuff */
+ case PHB_LEM_FIR_AND_MASK:
+ phb->regs[PHB_LEM_FIR_ACCUM >> 3] &= val;
+ return;
+ case PHB_LEM_FIR_OR_MASK:
+ phb->regs[PHB_LEM_FIR_ACCUM >> 3] |= val;
+ return;
+ case PHB_LEM_ERROR_AND_MASK:
+ phb->regs[PHB_LEM_ERROR_MASK >> 3] &= val;
+ return;
+ case PHB_LEM_ERROR_OR_MASK:
+ phb->regs[PHB_LEM_ERROR_MASK >> 3] |= val;
+ return;
+ case PHB_LEM_WOF:
+ val = 0;
+ break;
+ }
+
+ /* Record whether it changed */
+ changed = phb->regs[off >> 3] != val;
+
+ /* Store in register cache first */
+ phb->regs[off >> 3] = val;
+
+ /* Handle side effects */
+ switch (off) {
+ case PHB_PHB3_CONFIG:
+ if (changed) {
+ pnv_phb3_update_all_msi_regions(phb);
+ }
+ /* fall through */
+ case PHB_M32_BASE_ADDR:
+ case PHB_M32_BASE_MASK:
+ case PHB_M32_START_ADDR:
+ if (changed) {
+ pnv_phb3_check_m32(phb);
+ }
+ break;
+ case PHB_M64_UPPER_BITS:
+ if (changed) {
+ pnv_phb3_check_all_m64s(phb);
+ }
+ break;
+ case PHB_LSI_SOURCE_ID:
+ if (changed) {
+ pnv_phb3_lsi_src_id_write(phb, val);
+ }
+ break;
+
+ /* IODA table accesses */
+ case PHB_IODA_DATA0:
+ pnv_phb3_ioda_write(phb, val);
+ break;
+
+ /* RTC invalidation */
+ case PHB_RTC_INVALIDATE:
+ pnv_phb3_rtc_invalidate(phb, val);
+ break;
+
+ /* FFI request */
+ case PHB_FFI_REQUEST:
+ pnv_phb3_msi_ffi(&phb->msis, val);
+ break;
+
+ /* Silent simple writes */
+ case PHB_CONFIG_ADDRESS:
+ case PHB_IODA_ADDR:
+ case PHB_TCE_KILL:
+ case PHB_TCE_SPEC_CTL:
+ case PHB_PEST_BAR:
+ case PHB_PELTV_BAR:
+ case PHB_RTT_BAR:
+ case PHB_RBA_BAR:
+ case PHB_IVT_BAR:
+ case PHB_FFI_LOCK:
+ case PHB_LEM_FIR_ACCUM:
+ case PHB_LEM_ERROR_MASK:
+ case PHB_LEM_ACTION0:
+ case PHB_LEM_ACTION1:
+ break;
+
+ /* Noise on anything else */
+ default:
+ qemu_log_mask(LOG_UNIMP, "phb3: reg_write 0x%"PRIx64"=%"PRIx64"\n",
+ off, val);
+ }
+}
+
+uint64_t pnv_phb3_reg_read(void *opaque, hwaddr off, unsigned size)
+{
+ PnvPHB3 *phb = opaque;
+ PCIHostState *pci = PCI_HOST_BRIDGE(phb);
+ uint64_t val;
+
+ if ((off & 0xfffc) == PHB_CONFIG_DATA) {
+ return pnv_phb3_config_read(phb, off & 0x3, size);
+ }
+
+ /* Other registers are 64-bit only */
+ if (size != 8 || off & 0x7) {
+ phb3_error(phb, "Invalid register access, offset: 0x%"PRIx64" size: %d",
+ off, size);
+ return ~0ull;
+ }
+
+ /* Default read from cache */
+ val = phb->regs[off >> 3];
+
+ switch (off) {
+ /* Simulate venice DD2.0 */
+ case PHB_VERSION:
+ return 0x000000a300000005ull;
+ case PHB_PCIE_SYSTEM_CONFIG:
+ return 0x441100fc30000000;
+
+ /* IODA table accesses */
+ case PHB_IODA_DATA0:
+ return pnv_phb3_ioda_read(phb);
+
+ /* Link training always appears trained */
+ case PHB_PCIE_DLP_TRAIN_CTL:
+ if (!pci_find_device(pci->bus, 1, 0)) {
+ return 0;
+ }
+ return PHB_PCIE_DLP_INBAND_PRESENCE | PHB_PCIE_DLP_TC_DL_LINKACT;
+
+ /* FFI Lock */
+ case PHB_FFI_LOCK:
+ /* Set lock and return previous value */
+ phb->regs[off >> 3] |= PHB_FFI_LOCK_STATE;
+ return val;
+
+ /* DMA read sync: make it look like it's complete */
+ case PHB_DMARD_SYNC:
+ return PHB_DMARD_SYNC_COMPLETE;
+
+ /* Silent simple reads */
+ case PHB_PHB3_CONFIG:
+ case PHB_M32_BASE_ADDR:
+ case PHB_M32_BASE_MASK:
+ case PHB_M32_START_ADDR:
+ case PHB_CONFIG_ADDRESS:
+ case PHB_IODA_ADDR:
+ case PHB_RTC_INVALIDATE:
+ case PHB_TCE_KILL:
+ case PHB_TCE_SPEC_CTL:
+ case PHB_PEST_BAR:
+ case PHB_PELTV_BAR:
+ case PHB_RTT_BAR:
+ case PHB_RBA_BAR:
+ case PHB_IVT_BAR:
+ case PHB_M64_UPPER_BITS:
+ case PHB_LEM_FIR_ACCUM:
+ case PHB_LEM_ERROR_MASK:
+ case PHB_LEM_ACTION0:
+ case PHB_LEM_ACTION1:
+ break;
+
+ /* Noise on anything else */
+ default:
+ qemu_log_mask(LOG_UNIMP, "phb3: reg_read 0x%"PRIx64"=%"PRIx64"\n",
+ off, val);
+ }
+ return val;
+}
+
+static const MemoryRegionOps pnv_phb3_reg_ops = {
+ .read = pnv_phb3_reg_read,
+ .write = pnv_phb3_reg_write,
+ .valid.min_access_size = 1,
+ .valid.max_access_size = 8,
+ .impl.min_access_size = 1,
+ .impl.max_access_size = 8,
+ .endianness = DEVICE_BIG_ENDIAN,
+};
+
+static int pnv_phb3_map_irq(PCIDevice *pci_dev, int irq_num)
+{
+ /* Check that out properly ... */
+ return irq_num & 3;
+}
+
+static void pnv_phb3_set_irq(void *opaque, int irq_num, int level)
+{
+ PnvPHB3 *phb = opaque;
+
+ /* LSI only ... */
+ if (irq_num > 3) {
+ phb3_error(phb, "Unknown IRQ to set %d", irq_num);
+ }
+ qemu_set_irq(phb->qirqs[irq_num], level);
+}
+
+static bool pnv_phb3_resolve_pe(PnvPhb3DMASpace *ds)
+{
+ uint64_t rtt, addr;
+ uint16_t rte;
+ int bus_num;
+
+ /* Already resolved ? */
+ if (ds->pe_num != PHB_INVALID_PE) {
+ return true;
+ }
+
+ /* We need to lookup the RTT */
+ rtt = ds->phb->regs[PHB_RTT_BAR >> 3];
+ if (!(rtt & PHB_RTT_BAR_ENABLE)) {
+ phb3_error(ds->phb, "DMA with RTT BAR disabled !");
+ /* Set error bits ? fence ? ... */
+ return false;
+ }
+
+ /* Read RTE */
+ bus_num = pci_bus_num(ds->bus);
+ addr = rtt & PHB_RTT_BASE_ADDRESS_MASK;
+ addr += 2 * ((bus_num << 8) | ds->devfn);
+ if (dma_memory_read(&address_space_memory, addr, &rte, sizeof(rte))) {
+ phb3_error(ds->phb, "Failed to read RTT entry at 0x%"PRIx64, addr);
+ /* Set error bits ? fence ? ... */
+ return false;
+ }
+ rte = be16_to_cpu(rte);
+
+ /* Fail upon reading of invalid PE# */
+ if (rte >= PNV_PHB3_NUM_PE) {
+ phb3_error(ds->phb, "RTE for RID 0x%x invalid (%04x", ds->devfn, rte);
+ /* Set error bits ? fence ? ... */
+ return false;
+ }
+ ds->pe_num = rte;
+ return true;
+}
+
+static void pnv_phb3_translate_tve(PnvPhb3DMASpace *ds, hwaddr addr,
+ bool is_write, uint64_t tve,
+ IOMMUTLBEntry *tlb)
+{
+ uint64_t tta = GETFIELD(IODA2_TVT_TABLE_ADDR, tve);
+ int32_t lev = GETFIELD(IODA2_TVT_NUM_LEVELS, tve);
+ uint32_t tts = GETFIELD(IODA2_TVT_TCE_TABLE_SIZE, tve);
+ uint32_t tps = GETFIELD(IODA2_TVT_IO_PSIZE, tve);
+ PnvPHB3 *phb = ds->phb;
+
+ /* Invalid levels */
+ if (lev > 4) {
+ phb3_error(phb, "Invalid #levels in TVE %d", lev);
+ return;
+ }
+
+ /* IO Page Size of 0 means untranslated, else use TCEs */
+ if (tps == 0) {
+ /*
+ * We only support non-translate in top window.
+ *
+ * TODO: Venice/Murano support it on bottom window above 4G and
+ * Naples suports it on everything
+ */
+ if (!(tve & PPC_BIT(51))) {
+ phb3_error(phb, "xlate for invalid non-translate TVE");
+ return;
+ }
+ /* TODO: Handle boundaries */
+
+ /* Use 4k pages like q35 ... for now */
+ tlb->iova = addr & 0xfffffffffffff000ull;
+ tlb->translated_addr = addr & 0x0003fffffffff000ull;
+ tlb->addr_mask = 0xfffull;
+ tlb->perm = IOMMU_RW;
+ } else {
+ uint32_t tce_shift, tbl_shift, sh;
+ uint64_t base, taddr, tce, tce_mask;
+
+ /* TVE disabled ? */
+ if (tts == 0) {
+ phb3_error(phb, "xlate for invalid translated TVE");
+ return;
+ }
+
+ /* Address bits per bottom level TCE entry */
+ tce_shift = tps + 11;
+
+ /* Address bits per table level */
+ tbl_shift = tts + 8;
+
+ /* Top level table base address */
+ base = tta << 12;
+
+ /* Total shift to first level */
+ sh = tbl_shift * lev + tce_shift;
+
+ /* TODO: Multi-level untested */
+ while ((lev--) >= 0) {
+ /* Grab the TCE address */
+ taddr = base | (((addr >> sh) & ((1ul << tbl_shift) - 1)) << 3);
+ if (dma_memory_read(&address_space_memory, taddr, &tce,
+ sizeof(tce))) {
+ phb3_error(phb, "Failed to read TCE at 0x%"PRIx64, taddr);
+ return;
+ }
+ tce = be64_to_cpu(tce);
+
+ /* Check permission for indirect TCE */
+ if ((lev >= 0) && !(tce & 3)) {
+ phb3_error(phb, "Invalid indirect TCE at 0x%"PRIx64, taddr);
+ phb3_error(phb, " xlate %"PRIx64":%c TVE=%"PRIx64, addr,
+ is_write ? 'W' : 'R', tve);
+ phb3_error(phb, " tta=%"PRIx64" lev=%d tts=%d tps=%d",
+ tta, lev, tts, tps);
+ return;
+ }
+ sh -= tbl_shift;
+ base = tce & ~0xfffull;
+ }
+
+ /* We exit the loop with TCE being the final TCE */
+ tce_mask = ~((1ull << tce_shift) - 1);
+ tlb->iova = addr & tce_mask;
+ tlb->translated_addr = tce & tce_mask;
+ tlb->addr_mask = ~tce_mask;
+ tlb->perm = tce & 3;
+ if ((is_write & !(tce & 2)) || ((!is_write) && !(tce & 1))) {
+ phb3_error(phb, "TCE access fault at 0x%"PRIx64, taddr);
+ phb3_error(phb, " xlate %"PRIx64":%c TVE=%"PRIx64, addr,
+ is_write ? 'W' : 'R', tve);
+ phb3_error(phb, " tta=%"PRIx64" lev=%d tts=%d tps=%d",
+ tta, lev, tts, tps);
+ }
+ }
+}
+
+static IOMMUTLBEntry pnv_phb3_translate_iommu(IOMMUMemoryRegion *iommu,
+ hwaddr addr,
+ IOMMUAccessFlags flag,
+ int iommu_idx)
+{
+ PnvPhb3DMASpace *ds = container_of(iommu, PnvPhb3DMASpace, dma_mr);
+ int tve_sel;
+ uint64_t tve, cfg;
+ IOMMUTLBEntry ret = {
+ .target_as = &address_space_memory,
+ .iova = addr,
+ .translated_addr = 0,
+ .addr_mask = ~(hwaddr)0,
+ .perm = IOMMU_NONE,
+ };
+ PnvPHB3 *phb = ds->phb;
+
+ /* Resolve PE# */
+ if (!pnv_phb3_resolve_pe(ds)) {
+ phb3_error(phb, "Failed to resolve PE# for bus @%p (%d) devfn 0x%x",
+ ds->bus, pci_bus_num(ds->bus), ds->devfn);
+ return ret;
+ }
+
+ /* Check top bits */
+ switch (addr >> 60) {
+ case 00:
+ /* DMA or 32-bit MSI ? */
+ cfg = ds->phb->regs[PHB_PHB3_CONFIG >> 3];
+ if ((cfg & PHB_PHB3C_32BIT_MSI_EN) &&
+ ((addr & 0xffffffffffff0000ull) == 0xffff0000ull)) {
+ phb3_error(phb, "xlate on 32-bit MSI region");
+ return ret;
+ }
+ /* Choose TVE XXX Use PHB3 Control Register */
+ tve_sel = (addr >> 59) & 1;
+ tve = ds->phb->ioda_TVT[ds->pe_num * 2 + tve_sel];
+ pnv_phb3_translate_tve(ds, addr, flag & IOMMU_WO, tve, &ret);
+ break;
+ case 01:
+ phb3_error(phb, "xlate on 64-bit MSI region");
+ break;
+ default:
+ phb3_error(phb, "xlate on unsupported address 0x%"PRIx64, addr);
+ }
+ return ret;
+}
+
+#define TYPE_PNV_PHB3_IOMMU_MEMORY_REGION "pnv-phb3-iommu-memory-region"
+#define PNV_PHB3_IOMMU_MEMORY_REGION(obj) \
+ OBJECT_CHECK(IOMMUMemoryRegion, (obj), TYPE_PNV_PHB3_IOMMU_MEMORY_REGION)
+
+static void pnv_phb3_iommu_memory_region_class_init(ObjectClass *klass,
+ void *data)
+{
+ IOMMUMemoryRegionClass *imrc = IOMMU_MEMORY_REGION_CLASS(klass);
+
+ imrc->translate = pnv_phb3_translate_iommu;
+}
+
+static const TypeInfo pnv_phb3_iommu_memory_region_info = {
+ .parent = TYPE_IOMMU_MEMORY_REGION,
+ .name = TYPE_PNV_PHB3_IOMMU_MEMORY_REGION,
+ .class_init = pnv_phb3_iommu_memory_region_class_init,
+};
+
+/*
+ * MSI/MSIX memory region implementation.
+ * The handler handles both MSI and MSIX.
+ */
+static void pnv_phb3_msi_write(void *opaque, hwaddr addr,
+ uint64_t data, unsigned size)
+{
+ PnvPhb3DMASpace *ds = opaque;
+
+ /* Resolve PE# */
+ if (!pnv_phb3_resolve_pe(ds)) {
+ phb3_error(ds->phb, "Failed to resolve PE# for bus @%p (%d) devfn 0x%x",
+ ds->bus, pci_bus_num(ds->bus), ds->devfn);
+ return;
+ }
+
+ pnv_phb3_msi_send(&ds->phb->msis, addr, data, ds->pe_num);
+}
+
+/* There is no .read as the read result is undefined by PCI spec */
+static uint64_t pnv_phb3_msi_read(void *opaque, hwaddr addr, unsigned size)
+{
+ PnvPhb3DMASpace *ds = opaque;
+
+ phb3_error(ds->phb, "invalid read @ 0x%" HWADDR_PRIx, addr);
+ return -1;
+}
+
+static const MemoryRegionOps pnv_phb3_msi_ops = {
+ .read = pnv_phb3_msi_read,
+ .write = pnv_phb3_msi_write,
+ .endianness = DEVICE_LITTLE_ENDIAN
+};
+
+static AddressSpace *pnv_phb3_dma_iommu(PCIBus *bus, void *opaque, int devfn)
+{
+ PnvPHB3 *phb = opaque;
+ PnvPhb3DMASpace *ds;
+
+ QLIST_FOREACH(ds, &phb->dma_spaces, list) {
+ if (ds->bus == bus && ds->devfn == devfn) {
+ break;
+ }
+ }
+
+ if (ds == NULL) {
+ ds = g_malloc0(sizeof(PnvPhb3DMASpace));
+ ds->bus = bus;
+ ds->devfn = devfn;
+ ds->pe_num = PHB_INVALID_PE;
+ ds->phb = phb;
+ memory_region_init_iommu(&ds->dma_mr, sizeof(ds->dma_mr),
+ TYPE_PNV_PHB3_IOMMU_MEMORY_REGION,
+ OBJECT(phb), "phb3_iommu", UINT64_MAX);
+ address_space_init(&ds->dma_as, MEMORY_REGION(&ds->dma_mr),
+ "phb3_iommu");
+ memory_region_init_io(&ds->msi32_mr, OBJECT(phb), &pnv_phb3_msi_ops,
+ ds, "msi32", 0x10000);
+ memory_region_init_io(&ds->msi64_mr, OBJECT(phb), &pnv_phb3_msi_ops,
+ ds, "msi64", 0x100000);
+ pnv_phb3_update_msi_regions(ds);
+
+ QLIST_INSERT_HEAD(&phb->dma_spaces, ds, list);
+ }
+ return &ds->dma_as;
+}
+
+static void pnv_phb3_instance_init(Object *obj)
+{
+ PnvPHB3 *phb = PNV_PHB3(obj);
+
+ QLIST_INIT(&phb->dma_spaces);
+
+ /* LSI sources */
+ object_initialize_child(obj, "lsi", &phb->lsis, sizeof(phb->lsis),
+ TYPE_ICS, &error_abort, NULL);
+
+ /* Default init ... will be fixed by HW inits */
+ phb->lsis.offset = 0;
+
+ /* MSI sources */
+ object_initialize_child(obj, "msi", &phb->msis, sizeof(phb->msis),
+ TYPE_PHB3_MSI, &error_abort, NULL);
+
+ /* Power Bus Common Queue */
+ object_initialize_child(obj, "pbcq", &phb->pbcq, sizeof(phb->pbcq),
+ TYPE_PNV_PBCQ, &error_abort, NULL);
+
+ /* Root Port */
+ object_initialize_child(obj, "root", &phb->root, sizeof(phb->root),
+ TYPE_PNV_PHB3_ROOT_PORT, &error_abort, NULL);
+ qdev_prop_set_int32(DEVICE(&phb->root), "addr", PCI_DEVFN(0, 0));
+ qdev_prop_set_bit(DEVICE(&phb->root), "multifunction", false);
+}
+
+static void pnv_phb3_realize(DeviceState *dev, Error **errp)
+{
+ PnvPHB3 *phb = PNV_PHB3(dev);
+ PCIHostState *pci = PCI_HOST_BRIDGE(dev);
+ PnvMachineState *pnv = PNV_MACHINE(qdev_get_machine());
+ Error *local_err = NULL;
+ int i;
+
+ if (phb->phb_id >= PNV8_CHIP_PHB3_MAX) {
+ error_setg(errp, "invalid PHB index: %d", phb->phb_id);
+ return;
+ }
+
+ /* LSI sources */
+ object_property_set_link(OBJECT(&phb->lsis), OBJECT(pnv), "xics",
+ &error_abort);
+ object_property_set_int(OBJECT(&phb->lsis), PNV_PHB3_NUM_LSI, "nr-irqs",
+ &error_abort);
+ object_property_set_bool(OBJECT(&phb->lsis), true, "realized", &local_err);
+ if (local_err) {
+ error_propagate(errp, local_err);
+ return;
+ }
+
+ for (i = 0; i < phb->lsis.nr_irqs; i++) {
+ ics_set_irq_type(&phb->lsis, i, true);
+ }
+
+ phb->qirqs = qemu_allocate_irqs(ics_set_irq, &phb->lsis, phb->lsis.nr_irqs);
+
+ /* MSI sources */
+ object_property_set_link(OBJECT(&phb->msis), OBJECT(phb), "phb",
+ &error_abort);
+ object_property_set_link(OBJECT(&phb->msis), OBJECT(pnv), "xics",
+ &error_abort);
+ object_property_set_int(OBJECT(&phb->msis), PHB3_MAX_MSI, "nr-irqs",
+ &error_abort);
+ object_property_set_bool(OBJECT(&phb->msis), true, "realized", &local_err);
+ if (local_err) {
+ error_propagate(errp, local_err);
+ return;
+ }
+
+ /* Power Bus Common Queue */
+ object_property_set_link(OBJECT(&phb->pbcq), OBJECT(phb), "phb",
+ &error_abort);
+ object_property_set_bool(OBJECT(&phb->pbcq), true, "realized", &local_err);
+ if (local_err) {
+ error_propagate(errp, local_err);
+ return;
+ }
+
+ /* Controller Registers */
+ memory_region_init_io(&phb->mr_regs, OBJECT(phb), &pnv_phb3_reg_ops, phb,
+ "phb3-regs", 0x1000);
+
+ /*
+ * PHB3 doesn't support IO space. However, qemu gets very upset if
+ * we don't have an IO region to anchor IO BARs onto so we just
+ * initialize one which we never hook up to anything
+ */
+ memory_region_init(&phb->pci_io, OBJECT(phb), "pci-io", 0x10000);
+ memory_region_init(&phb->pci_mmio, OBJECT(phb), "pci-mmio",
+ PCI_MMIO_TOTAL_SIZE);
+
+ pci->bus = pci_register_root_bus(dev, "root-bus",
+ pnv_phb3_set_irq, pnv_phb3_map_irq, phb,
+ &phb->pci_mmio, &phb->pci_io,
+ 0, 4, TYPE_PNV_PHB3_ROOT_BUS);
+
+ pci_setup_iommu(pci->bus, pnv_phb3_dma_iommu, phb);
+
+ /* Add a single Root port */
+ qdev_prop_set_uint8(DEVICE(&phb->root), "chassis", phb->chip_id);
+ qdev_prop_set_uint16(DEVICE(&phb->root), "slot", phb->phb_id);
+ qdev_set_parent_bus(DEVICE(&phb->root), BUS(pci->bus));
+ qdev_init_nofail(DEVICE(&phb->root));
+}
+
+void pnv_phb3_update_regions(PnvPHB3 *phb)
+{
+ PnvPBCQState *pbcq = &phb->pbcq;
+
+ /* Unmap first always */
+ if (memory_region_is_mapped(&phb->mr_regs)) {
+ memory_region_del_subregion(&pbcq->phbbar, &phb->mr_regs);
+ }
+
+ /* Map registers if enabled */
+ if (memory_region_is_mapped(&pbcq->phbbar)) {
+ /* TODO: We should use the PHB BAR 2 register but we don't ... */
+ memory_region_add_subregion(&pbcq->phbbar, 0, &phb->mr_regs);
+ }
+
+ /* Check/update m32 */
+ if (memory_region_is_mapped(&phb->mr_m32)) {
+ pnv_phb3_check_m32(phb);
+ }
+ pnv_phb3_check_all_m64s(phb);
+}
+
+static const char *pnv_phb3_root_bus_path(PCIHostState *host_bridge,
+ PCIBus *rootbus)
+{
+ PnvPHB3 *phb = PNV_PHB3(host_bridge);
+
+ snprintf(phb->bus_path, sizeof(phb->bus_path), "00%02x:%02x",
+ phb->chip_id, phb->phb_id);
+ return phb->bus_path;
+}
+
+static Property pnv_phb3_properties[] = {
+ DEFINE_PROP_UINT32("index", PnvPHB3, phb_id, 0),
+ DEFINE_PROP_UINT32("chip-id", PnvPHB3, chip_id, 0),
+ DEFINE_PROP_END_OF_LIST(),
+};
+
+static void pnv_phb3_class_init(ObjectClass *klass, void *data)
+{
+ PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass);
+ DeviceClass *dc = DEVICE_CLASS(klass);
+
+ hc->root_bus_path = pnv_phb3_root_bus_path;
+ dc->realize = pnv_phb3_realize;
+ device_class_set_props(dc, pnv_phb3_properties);
+ set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+ dc->user_creatable = false;
+}
+
+static const TypeInfo pnv_phb3_type_info = {
+ .name = TYPE_PNV_PHB3,
+ .parent = TYPE_PCIE_HOST_BRIDGE,
+ .instance_size = sizeof(PnvPHB3),
+ .class_init = pnv_phb3_class_init,
+ .instance_init = pnv_phb3_instance_init,
+};
+
+static void pnv_phb3_root_bus_class_init(ObjectClass *klass, void *data)
+{
+ BusClass *k = BUS_CLASS(klass);
+
+ /*
+ * PHB3 has only a single root complex. Enforce the limit on the
+ * parent bus
+ */
+ k->max_dev = 1;
+}
+
+static const TypeInfo pnv_phb3_root_bus_info = {
+ .name = TYPE_PNV_PHB3_ROOT_BUS,
+ .parent = TYPE_PCIE_BUS,
+ .class_init = pnv_phb3_root_bus_class_init,
+ .interfaces = (InterfaceInfo[]) {
+ { INTERFACE_PCIE_DEVICE },
+ { }
+ },
+};
+
+static void pnv_phb3_root_port_realize(DeviceState *dev, Error **errp)
+{
+ PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
+ Error *local_err = NULL;
+
+ rpc->parent_realize(dev, &local_err);
+ if (local_err) {
+ error_propagate(errp, local_err);
+ return;
+ }
+}
+
+static void pnv_phb3_root_port_class_init(ObjectClass *klass, void *data)
+{
+ DeviceClass *dc = DEVICE_CLASS(klass);
+ PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
+ PCIERootPortClass *rpc = PCIE_ROOT_PORT_CLASS(klass);
+
+ dc->desc = "IBM PHB3 PCIE Root Port";
+
+ device_class_set_parent_realize(dc, pnv_phb3_root_port_realize,
+ &rpc->parent_realize);
+ dc->user_creatable = false;
+
+ k->vendor_id = PCI_VENDOR_ID_IBM;
+ k->device_id = 0x03dc;
+ k->revision = 0;
+
+ rpc->exp_offset = 0x48;
+ rpc->aer_offset = 0x100;
+}
+
+static const TypeInfo pnv_phb3_root_port_info = {
+ .name = TYPE_PNV_PHB3_ROOT_PORT,
+ .parent = TYPE_PCIE_ROOT_PORT,
+ .instance_size = sizeof(PnvPHB3RootPort),
+ .class_init = pnv_phb3_root_port_class_init,
+};
+
+static void pnv_phb3_register_types(void)
+{
+ type_register_static(&pnv_phb3_root_bus_info);
+ type_register_static(&pnv_phb3_root_port_info);
+ type_register_static(&pnv_phb3_type_info);
+ type_register_static(&pnv_phb3_iommu_memory_region_info);
+}
+
+type_init(pnv_phb3_register_types)
diff --git a/hw/pci-host/pnv_phb3_msi.c b/hw/pci-host/pnv_phb3_msi.c
new file mode 100644
index 0000000000..ecfc1b2c4e
--- /dev/null
+++ b/hw/pci-host/pnv_phb3_msi.c
@@ -0,0 +1,349 @@
+/*
+ * QEMU PowerPC PowerNV (POWER8) PHB3 model
+ *
+ * Copyright (c) 2014-2020, IBM Corporation.
+ *
+ * This code is licensed under the GPL version 2 or later. See the
+ * COPYING file in the top-level directory.
+ */
+#include "qemu/osdep.h"
+#include "qemu/log.h"
+#include "qapi/error.h"
+#include "qemu-common.h"
+#include "hw/pci-host/pnv_phb3_regs.h"
+#include "hw/pci-host/pnv_phb3.h"
+#include "hw/ppc/pnv.h"
+#include "hw/pci/msi.h"
+#include "monitor/monitor.h"
+#include "hw/irq.h"
+#include "hw/qdev-properties.h"
+#include "sysemu/reset.h"
+
+static uint64_t phb3_msi_ive_addr(PnvPHB3 *phb, int srcno)
+{
+ uint64_t ivtbar = phb->regs[PHB_IVT_BAR >> 3];
+ uint64_t phbctl = phb->regs[PHB_CONTROL >> 3];
+
+ if (!(ivtbar & PHB_IVT_BAR_ENABLE)) {
+ qemu_log_mask(LOG_GUEST_ERROR, "Failed access to disable IVT BAR !");
+ return 0;
+ }
+
+ if (srcno >= (ivtbar & PHB_IVT_LENGTH_MASK)) {
+ qemu_log_mask(LOG_GUEST_ERROR, "MSI out of bounds (%d vs 0x%"PRIx64")",
+ srcno, (uint64_t) (ivtbar & PHB_IVT_LENGTH_MASK));
+ return 0;
+ }
+
+ ivtbar &= PHB_IVT_BASE_ADDRESS_MASK;
+
+ if (phbctl & PHB_CTRL_IVE_128_BYTES) {
+ return ivtbar + 128 * srcno;
+ } else {
+ return ivtbar + 16 * srcno;
+ }
+}
+
+static bool phb3_msi_read_ive(PnvPHB3 *phb, int srcno, uint64_t *out_ive)
+{
+ uint64_t ive_addr, ive;
+
+ ive_addr = phb3_msi_ive_addr(phb, srcno);
+ if (!ive_addr) {
+ return false;
+ }
+
+ if (dma_memory_read(&address_space_memory, ive_addr, &ive, sizeof(ive))) {
+ qemu_log_mask(LOG_GUEST_ERROR, "Failed to read IVE at 0x%" PRIx64,
+ ive_addr);
+ return false;
+ }
+ *out_ive = be64_to_cpu(ive);
+
+ return true;
+}
+
+static void phb3_msi_set_p(Phb3MsiState *msi, int srcno, uint8_t gen)
+{
+ uint64_t ive_addr;
+ uint8_t p = 0x01 | (gen << 1);
+
+ ive_addr = phb3_msi_ive_addr(msi->phb, srcno);
+ if (!ive_addr) {
+ return;
+ }
+
+ if (dma_memory_write(&address_space_memory, ive_addr + 4, &p, 1)) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "Failed to write IVE (set P) at 0x%" PRIx64, ive_addr);
+ }
+}
+
+static void phb3_msi_set_q(Phb3MsiState *msi, int srcno)
+{
+ uint64_t ive_addr;
+ uint8_t q = 0x01;
+
+ ive_addr = phb3_msi_ive_addr(msi->phb, srcno);
+ if (!ive_addr) {
+ return;
+ }
+
+ if (dma_memory_write(&address_space_memory, ive_addr + 5, &q, 1)) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "Failed to write IVE (set Q) at 0x%" PRIx64, ive_addr);
+ }
+}
+
+static void phb3_msi_try_send(Phb3MsiState *msi, int srcno, bool force)
+{
+ ICSState *ics = ICS(msi);
+ uint64_t ive;
+ uint64_t server, prio, pq, gen;
+
+ if (!phb3_msi_read_ive(msi->phb, srcno, &ive)) {
+ return;
+ }
+
+ server = GETFIELD(IODA2_IVT_SERVER, ive);
+ prio = GETFIELD(IODA2_IVT_PRIORITY, ive);
+ if (!force) {
+ pq = GETFIELD(IODA2_IVT_Q, ive) | (GETFIELD(IODA2_IVT_P, ive) << 1);
+ } else {
+ pq = 0;
+ }
+ gen = GETFIELD(IODA2_IVT_GEN, ive);
+
+ /*
+ * The low order 2 bits are the link pointer (Type II interrupts).
+ * Shift back to get a valid IRQ server.
+ */
+ server >>= 2;
+
+ switch (pq) {
+ case 0: /* 00 */
+ if (prio == 0xff) {
+ /* Masked, set Q */
+ phb3_msi_set_q(msi, srcno);
+ } else {
+ /* Enabled, set P and send */
+ phb3_msi_set_p(msi, srcno, gen);
+ icp_irq(ics, server, srcno + ics->offset, prio);
+ }
+ break;
+ case 2: /* 10 */
+ /* Already pending, set Q */
+ phb3_msi_set_q(msi, srcno);
+ break;
+ case 1: /* 01 */
+ case 3: /* 11 */
+ default:
+ /* Just drop stuff if Q already set */
+ break;
+ }
+}
+
+static void phb3_msi_set_irq(void *opaque, int srcno, int val)
+{
+ Phb3MsiState *msi = PHB3_MSI(opaque);
+
+ if (val) {
+ phb3_msi_try_send(msi, srcno, false);
+ }
+}
+
+
+void pnv_phb3_msi_send(Phb3MsiState *msi, uint64_t addr, uint16_t data,
+ int32_t dev_pe)
+{
+ ICSState *ics = ICS(msi);
+ uint64_t ive;
+ uint16_t pe;
+ uint32_t src = ((addr >> 4) & 0xffff) | (data & 0x1f);
+
+ if (src >= ics->nr_irqs) {
+ qemu_log_mask(LOG_GUEST_ERROR, "MSI %d out of bounds", src);
+ return;
+ }
+ if (dev_pe >= 0) {
+ if (!phb3_msi_read_ive(msi->phb, src, &ive)) {
+ return;
+ }
+ pe = GETFIELD(IODA2_IVT_PE, ive);
+ if (pe != dev_pe) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "MSI %d send by PE#%d but assigned to PE#%d",
+ src, dev_pe, pe);
+ return;
+ }
+ }
+ qemu_irq_pulse(msi->qirqs[src]);
+}
+
+void pnv_phb3_msi_ffi(Phb3MsiState *msi, uint64_t val)
+{
+ /* Emit interrupt */
+ pnv_phb3_msi_send(msi, val, 0, -1);
+
+ /* Clear FFI lock */
+ msi->phb->regs[PHB_FFI_LOCK >> 3] = 0;
+}
+
+static void phb3_msi_reject(ICSState *ics, uint32_t nr)
+{
+ Phb3MsiState *msi = PHB3_MSI(ics);
+ unsigned int srcno = nr - ics->offset;
+ unsigned int idx = srcno >> 6;
+ unsigned int bit = 1ull << (srcno & 0x3f);
+
+ assert(srcno < PHB3_MAX_MSI);
+
+ msi->rba[idx] |= bit;
+ msi->rba_sum |= (1u << idx);
+}
+
+static void phb3_msi_resend(ICSState *ics)
+{
+ Phb3MsiState *msi = PHB3_MSI(ics);
+ unsigned int i, j;
+
+ if (msi->rba_sum == 0) {
+ return;
+ }
+
+ for (i = 0; i < 32; i++) {
+ if ((msi->rba_sum & (1u << i)) == 0) {
+ continue;
+ }
+ msi->rba_sum &= ~(1u << i);
+ for (j = 0; j < 64; j++) {
+ if ((msi->rba[i] & (1ull << j)) == 0) {
+ continue;
+ }
+ msi->rba[i] &= ~(1u << j);
+ phb3_msi_try_send(msi, i * 64 + j, true);
+ }
+ }
+}
+
+static void phb3_msi_reset(DeviceState *dev)
+{
+ Phb3MsiState *msi = PHB3_MSI(dev);
+ ICSStateClass *icsc = ICS_GET_CLASS(dev);
+
+ icsc->parent_reset(dev);
+
+ memset(msi->rba, 0, sizeof(msi->rba));
+ msi->rba_sum = 0;
+}
+
+static void phb3_msi_reset_handler(void *dev)
+{
+ phb3_msi_reset(dev);
+}
+
+void pnv_phb3_msi_update_config(Phb3MsiState *msi, uint32_t base,
+ uint32_t count)
+{
+ ICSState *ics = ICS(msi);
+
+ if (count > PHB3_MAX_MSI) {
+ count = PHB3_MAX_MSI;
+ }
+ ics->nr_irqs = count;
+ ics->offset = base;
+}
+
+static void phb3_msi_realize(DeviceState *dev, Error **errp)
+{
+ Phb3MsiState *msi = PHB3_MSI(dev);
+ ICSState *ics = ICS(msi);
+ ICSStateClass *icsc = ICS_GET_CLASS(ics);
+ Error *local_err = NULL;
+
+ assert(msi->phb);
+
+ icsc->parent_realize(dev, &local_err);
+ if (local_err) {
+ error_propagate(errp, local_err);
+ return;
+ }
+
+ msi->qirqs = qemu_allocate_irqs(phb3_msi_set_irq, msi, ics->nr_irqs);
+
+ qemu_register_reset(phb3_msi_reset_handler, dev);
+}
+
+static void phb3_msi_instance_init(Object *obj)
+{
+ Phb3MsiState *msi = PHB3_MSI(obj);
+ ICSState *ics = ICS(obj);
+
+ object_property_add_link(obj, "phb", TYPE_PNV_PHB3,
+ (Object **)&msi->phb,
+ object_property_allow_set_link,
+ OBJ_PROP_LINK_STRONG,
+ &error_abort);
+
+ /* Will be overriden later */
+ ics->offset = 0;
+}
+
+static void phb3_msi_class_init(ObjectClass *klass, void *data)
+{
+ DeviceClass *dc = DEVICE_CLASS(klass);
+ ICSStateClass *isc = ICS_CLASS(klass);
+
+ device_class_set_parent_realize(dc, phb3_msi_realize,
+ &isc->parent_realize);
+ device_class_set_parent_reset(dc, phb3_msi_reset,
+ &isc->parent_reset);
+
+ isc->reject = phb3_msi_reject;
+ isc->resend = phb3_msi_resend;
+}
+
+static const TypeInfo phb3_msi_info = {
+ .name = TYPE_PHB3_MSI,
+ .parent = TYPE_ICS,
+ .instance_size = sizeof(Phb3MsiState),
+ .class_init = phb3_msi_class_init,
+ .class_size = sizeof(ICSStateClass),
+ .instance_init = phb3_msi_instance_init,
+};
+
+static void pnv_phb3_msi_register_types(void)
+{
+ type_register_static(&phb3_msi_info);
+}
+
+type_init(pnv_phb3_msi_register_types);
+
+void pnv_phb3_msi_pic_print_info(Phb3MsiState *msi, Monitor *mon)
+{
+ ICSState *ics = ICS(msi);
+ int i;
+
+ monitor_printf(mon, "ICS %4x..%4x %p\n",
+ ics->offset, ics->offset + ics->nr_irqs - 1, ics);
+
+ for (i = 0; i < ics->nr_irqs; i++) {
+ uint64_t ive;
+
+ if (!phb3_msi_read_ive(msi->phb, i, &ive)) {
+ return;
+ }
+
+ if (GETFIELD(IODA2_IVT_PRIORITY, ive) == 0xff) {
+ continue;
+ }
+
+ monitor_printf(mon, " %4x %c%c server=%04x prio=%02x gen=%d\n",
+ ics->offset + i,
+ GETFIELD(IODA2_IVT_P, ive) ? 'P' : '-',
+ GETFIELD(IODA2_IVT_Q, ive) ? 'Q' : '-',
+ (uint32_t) GETFIELD(IODA2_IVT_SERVER, ive) >> 2,
+ (uint32_t) GETFIELD(IODA2_IVT_PRIORITY, ive),
+ (uint32_t) GETFIELD(IODA2_IVT_GEN, ive));
+ }
+}
diff --git a/hw/pci-host/pnv_phb3_pbcq.c b/hw/pci-host/pnv_phb3_pbcq.c
new file mode 100644
index 0000000000..f232228b0e
--- /dev/null
+++ b/hw/pci-host/pnv_phb3_pbcq.c
@@ -0,0 +1,358 @@
+/*
+ * QEMU PowerPC PowerNV (POWER8) PHB3 model
+ *
+ * Copyright (c) 2014-2020, IBM Corporation.
+ *
+ * This code is licensed under the GPL version 2 or later. See the
+ * COPYING file in the top-level directory.
+ */
+#include "qemu/osdep.h"
+#include "qapi/error.h"
+#include "qemu-common.h"
+#include "qemu/log.h"
+#include "target/ppc/cpu.h"
+#include "hw/ppc/fdt.h"
+#include "hw/pci-host/pnv_phb3_regs.h"
+#include "hw/pci-host/pnv_phb3.h"
+#include "hw/ppc/pnv.h"
+#include "hw/ppc/pnv_xscom.h"
+#include "hw/pci/pci_bridge.h"
+#include "hw/pci/pci_bus.h"
+
+#include <libfdt.h>
+
+#define phb3_pbcq_error(pbcq, fmt, ...) \
+ qemu_log_mask(LOG_GUEST_ERROR, "phb3_pbcq[%d:%d]: " fmt "\n", \
+ (pbcq)->phb->chip_id, (pbcq)->phb->phb_id, ## __VA_ARGS__)
+
+static uint64_t pnv_pbcq_nest_xscom_read(void *opaque, hwaddr addr,
+ unsigned size)
+{
+ PnvPBCQState *pbcq = PNV_PBCQ(opaque);
+ uint32_t offset = addr >> 3;
+
+ return pbcq->nest_regs[offset];
+}
+
+static uint64_t pnv_pbcq_pci_xscom_read(void *opaque, hwaddr addr,
+ unsigned size)
+{
+ PnvPBCQState *pbcq = PNV_PBCQ(opaque);
+ uint32_t offset = addr >> 3;
+
+ return pbcq->pci_regs[offset];
+}
+
+static uint64_t pnv_pbcq_spci_xscom_read(void *opaque, hwaddr addr,
+ unsigned size)
+{
+ PnvPBCQState *pbcq = PNV_PBCQ(opaque);
+ uint32_t offset = addr >> 3;
+
+ if (offset == PBCQ_SPCI_ASB_DATA) {
+ return pnv_phb3_reg_read(pbcq->phb,
+ pbcq->spci_regs[PBCQ_SPCI_ASB_ADDR], 8);
+ }
+ return pbcq->spci_regs[offset];
+}
+
+static void pnv_pbcq_update_map(PnvPBCQState *pbcq)
+{
+ uint64_t bar_en = pbcq->nest_regs[PBCQ_NEST_BAR_EN];
+ uint64_t bar, mask, size;
+
+ /*
+ * NOTE: This will really not work well if those are remapped
+ * after the PHB has created its sub regions. We could do better
+ * if we had a way to resize regions but we don't really care
+ * that much in practice as the stuff below really only happens
+ * once early during boot
+ */
+
+ /* Handle unmaps */
+ if (memory_region_is_mapped(&pbcq->mmbar0) &&
+ !(bar_en & PBCQ_NEST_BAR_EN_MMIO0)) {
+ memory_region_del_subregion(get_system_memory(), &pbcq->mmbar0);
+ }
+ if (memory_region_is_mapped(&pbcq->mmbar1) &&
+ !(bar_en & PBCQ_NEST_BAR_EN_MMIO1)) {
+ memory_region_del_subregion(get_system_memory(), &pbcq->mmbar1);
+ }
+ if (memory_region_is_mapped(&pbcq->phbbar) &&
+ !(bar_en & PBCQ_NEST_BAR_EN_PHB)) {
+ memory_region_del_subregion(get_system_memory(), &pbcq->phbbar);
+ }
+
+ /* Update PHB */
+ pnv_phb3_update_regions(pbcq->phb);
+
+ /* Handle maps */
+ if (!memory_region_is_mapped(&pbcq->mmbar0) &&
+ (bar_en & PBCQ_NEST_BAR_EN_MMIO0)) {
+ bar = pbcq->nest_regs[PBCQ_NEST_MMIO_BAR0] >> 14;
+ mask = pbcq->nest_regs[PBCQ_NEST_MMIO_MASK0];
+ size = ((~mask) >> 14) + 1;
+ memory_region_init(&pbcq->mmbar0, OBJECT(pbcq), "pbcq-mmio0", size);
+ memory_region_add_subregion(get_system_memory(), bar, &pbcq->mmbar0);
+ pbcq->mmio0_base = bar;
+ pbcq->mmio0_size = size;
+ }
+ if (!memory_region_is_mapped(&pbcq->mmbar1) &&
+ (bar_en & PBCQ_NEST_BAR_EN_MMIO1)) {
+ bar = pbcq->nest_regs[PBCQ_NEST_MMIO_BAR1] >> 14;
+ mask = pbcq->nest_regs[PBCQ_NEST_MMIO_MASK1];
+ size = ((~mask) >> 14) + 1;
+ memory_region_init(&pbcq->mmbar1, OBJECT(pbcq), "pbcq-mmio1", size);
+ memory_region_add_subregion(get_system_memory(), bar, &pbcq->mmbar1);
+ pbcq->mmio1_base = bar;
+ pbcq->mmio1_size = size;
+ }
+ if (!memory_region_is_mapped(&pbcq->phbbar)
+ && (bar_en & PBCQ_NEST_BAR_EN_PHB)) {
+ bar = pbcq->nest_regs[PBCQ_NEST_PHB_BAR] >> 14;
+ size = 0x1000;
+ memory_region_init(&pbcq->phbbar, OBJECT(pbcq), "pbcq-phb", size);
+ memory_region_add_subregion(get_system_memory(), bar, &pbcq->phbbar);
+ }
+
+ /* Update PHB */
+ pnv_phb3_update_regions(pbcq->phb);
+}
+
+static void pnv_pbcq_nest_xscom_write(void *opaque, hwaddr addr,
+ uint64_t val, unsigned size)
+{
+ PnvPBCQState *pbcq = PNV_PBCQ(opaque);
+ uint32_t reg = addr >> 3;
+
+ switch (reg) {
+ case PBCQ_NEST_MMIO_BAR0:
+ case PBCQ_NEST_MMIO_BAR1:
+ case PBCQ_NEST_MMIO_MASK0:
+ case PBCQ_NEST_MMIO_MASK1:
+ if (pbcq->nest_regs[PBCQ_NEST_BAR_EN] &
+ (PBCQ_NEST_BAR_EN_MMIO0 |
+ PBCQ_NEST_BAR_EN_MMIO1)) {
+ phb3_pbcq_error(pbcq, "Changing enabled BAR unsupported");
+ }
+ pbcq->nest_regs[reg] = val & 0xffffffffc0000000ull;
+ break;
+ case PBCQ_NEST_PHB_BAR:
+ if (pbcq->nest_regs[PBCQ_NEST_BAR_EN] & PBCQ_NEST_BAR_EN_PHB) {
+ phb3_pbcq_error(pbcq, "Changing enabled BAR unsupported");
+ }
+ pbcq->nest_regs[reg] = val & 0xfffffffffc000000ull;
+ break;
+ case PBCQ_NEST_BAR_EN:
+ pbcq->nest_regs[reg] = val & 0xf800000000000000ull;
+ pnv_pbcq_update_map(pbcq);
+ pnv_phb3_remap_irqs(pbcq->phb);
+ break;
+ case PBCQ_NEST_IRSN_COMPARE:
+ case PBCQ_NEST_IRSN_MASK:
+ pbcq->nest_regs[reg] = val & PBCQ_NEST_IRSN_COMP;
+ pnv_phb3_remap_irqs(pbcq->phb);
+ break;
+ case PBCQ_NEST_LSI_SRC_ID:
+ pbcq->nest_regs[reg] = val & PBCQ_NEST_LSI_SRC;
+ pnv_phb3_remap_irqs(pbcq->phb);
+ break;
+ default:
+ phb3_pbcq_error(pbcq, "%s @0x%"HWADDR_PRIx"=%"PRIx64, __func__,
+ addr, val);
+ }
+}
+
+static void pnv_pbcq_pci_xscom_write(void *opaque, hwaddr addr,
+ uint64_t val, unsigned size)
+{
+ PnvPBCQState *pbcq = PNV_PBCQ(opaque);
+ uint32_t reg = addr >> 3;
+
+ switch (reg) {
+ case PBCQ_PCI_BAR2:
+ pbcq->pci_regs[reg] = val & 0xfffffffffc000000ull;
+ pnv_pbcq_update_map(pbcq);
+ default:
+ phb3_pbcq_error(pbcq, "%s @0x%"HWADDR_PRIx"=%"PRIx64, __func__,
+ addr, val);
+ }
+}
+
+static void pnv_pbcq_spci_xscom_write(void *opaque, hwaddr addr,
+ uint64_t val, unsigned size)
+{
+ PnvPBCQState *pbcq = PNV_PBCQ(opaque);
+ uint32_t reg = addr >> 3;
+
+ switch (reg) {
+ case PBCQ_SPCI_ASB_ADDR:
+ pbcq->spci_regs[reg] = val & 0xfff;
+ break;
+ case PBCQ_SPCI_ASB_STATUS:
+ pbcq->spci_regs[reg] &= ~val;
+ break;
+ case PBCQ_SPCI_ASB_DATA:
+ pnv_phb3_reg_write(pbcq->phb, pbcq->spci_regs[PBCQ_SPCI_ASB_ADDR],
+ val, 8);
+ break;
+ case PBCQ_SPCI_AIB_CAPP_EN:
+ case PBCQ_SPCI_CAPP_SEC_TMR:
+ break;
+ default:
+ phb3_pbcq_error(pbcq, "%s @0x%"HWADDR_PRIx"=%"PRIx64, __func__,
+ addr, val);
+ }
+}
+
+static const MemoryRegionOps pnv_pbcq_nest_xscom_ops = {
+ .read = pnv_pbcq_nest_xscom_read,
+ .write = pnv_pbcq_nest_xscom_write,
+ .valid.min_access_size = 8,
+ .valid.max_access_size = 8,
+ .impl.min_access_size = 8,
+ .impl.max_access_size = 8,
+ .endianness = DEVICE_BIG_ENDIAN,
+};
+
+static const MemoryRegionOps pnv_pbcq_pci_xscom_ops = {
+ .read = pnv_pbcq_pci_xscom_read,
+ .write = pnv_pbcq_pci_xscom_write,
+ .valid.min_access_size = 8,
+ .valid.max_access_size = 8,
+ .impl.min_access_size = 8,
+ .impl.max_access_size = 8,
+ .endianness = DEVICE_BIG_ENDIAN,
+};
+
+static const MemoryRegionOps pnv_pbcq_spci_xscom_ops = {
+ .read = pnv_pbcq_spci_xscom_read,
+ .write = pnv_pbcq_spci_xscom_write,
+ .valid.min_access_size = 8,
+ .valid.max_access_size = 8,
+ .impl.min_access_size = 8,
+ .impl.max_access_size = 8,
+ .endianness = DEVICE_BIG_ENDIAN,
+};
+
+static void pnv_pbcq_default_bars(PnvPBCQState *pbcq)
+{
+ uint64_t mm0, mm1, reg;
+ PnvPHB3 *phb = pbcq->phb;
+
+ mm0 = 0x3d00000000000ull + 0x4000000000ull * phb->chip_id +
+ 0x1000000000ull * phb->phb_id;
+ mm1 = 0x3ff8000000000ull + 0x0200000000ull * phb->chip_id +
+ 0x0080000000ull * phb->phb_id;
+ reg = 0x3fffe40000000ull + 0x0000400000ull * phb->chip_id +
+ 0x0000100000ull * phb->phb_id;
+
+ pbcq->nest_regs[PBCQ_NEST_MMIO_BAR0] = mm0 << 14;
+ pbcq->nest_regs[PBCQ_NEST_MMIO_BAR1] = mm1 << 14;
+ pbcq->nest_regs[PBCQ_NEST_PHB_BAR] = reg << 14;
+ pbcq->nest_regs[PBCQ_NEST_MMIO_MASK0] = 0x3fff000000000ull << 14;
+ pbcq->nest_regs[PBCQ_NEST_MMIO_MASK1] = 0x3ffff80000000ull << 14;
+ pbcq->pci_regs[PBCQ_PCI_BAR2] = reg << 14;
+}
+
+static void pnv_pbcq_realize(DeviceState *dev, Error **errp)
+{
+ PnvPBCQState *pbcq = PNV_PBCQ(dev);
+ PnvPHB3 *phb;
+ char name[32];
+
+ assert(pbcq->phb);
+ phb = pbcq->phb;
+
+ /* TODO: Fix OPAL to do that: establish default BAR values */
+ pnv_pbcq_default_bars(pbcq);
+
+ /* Initialize the XSCOM region for the PBCQ registers */
+ snprintf(name, sizeof(name), "xscom-pbcq-nest-%d.%d",
+ phb->chip_id, phb->phb_id);
+ pnv_xscom_region_init(&pbcq->xscom_nest_regs, OBJECT(dev),
+ &pnv_pbcq_nest_xscom_ops, pbcq, name,
+ PNV_XSCOM_PBCQ_NEST_SIZE);
+ snprintf(name, sizeof(name), "xscom-pbcq-pci-%d.%d",
+ phb->chip_id, phb->phb_id);
+ pnv_xscom_region_init(&pbcq->xscom_pci_regs, OBJECT(dev),
+ &pnv_pbcq_pci_xscom_ops, pbcq, name,
+ PNV_XSCOM_PBCQ_PCI_SIZE);
+ snprintf(name, sizeof(name), "xscom-pbcq-spci-%d.%d",
+ phb->chip_id, phb->phb_id);
+ pnv_xscom_region_init(&pbcq->xscom_spci_regs, OBJECT(dev),
+ &pnv_pbcq_spci_xscom_ops, pbcq, name,
+ PNV_XSCOM_PBCQ_SPCI_SIZE);
+}
+
+static int pnv_pbcq_dt_xscom(PnvXScomInterface *dev, void *fdt,
+ int xscom_offset)
+{
+ const char compat[] = "ibm,power8-pbcq";
+ PnvPHB3 *phb = PNV_PBCQ(dev)->phb;
+ char *name;
+ int offset;
+ uint32_t lpc_pcba = PNV_XSCOM_PBCQ_NEST_BASE + 0x400 * phb->phb_id;
+ uint32_t reg[] = {
+ cpu_to_be32(lpc_pcba),
+ cpu_to_be32(PNV_XSCOM_PBCQ_NEST_SIZE),
+ cpu_to_be32(PNV_XSCOM_PBCQ_PCI_BASE + 0x400 * phb->phb_id),
+ cpu_to_be32(PNV_XSCOM_PBCQ_PCI_SIZE),
+ cpu_to_be32(PNV_XSCOM_PBCQ_SPCI_BASE + 0x040 * phb->phb_id),
+ cpu_to_be32(PNV_XSCOM_PBCQ_SPCI_SIZE)
+ };
+
+ name = g_strdup_printf("pbcq@%x", lpc_pcba);
+ offset = fdt_add_subnode(fdt, xscom_offset, name);
+ _FDT(offset);
+ g_free(name);
+
+ _FDT((fdt_setprop(fdt, offset, "reg", reg, sizeof(reg))));
+
+ _FDT((fdt_setprop_cell(fdt, offset, "ibm,phb-index", phb->phb_id)));
+ _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id", phb->chip_id)));
+ _FDT((fdt_setprop(fdt, offset, "compatible", compat,
+ sizeof(compat))));
+ return 0;
+}
+
+static void phb3_pbcq_instance_init(Object *obj)
+{
+ PnvPBCQState *pbcq = PNV_PBCQ(obj);
+
+ object_property_add_link(obj, "phb", TYPE_PNV_PHB3,
+ (Object **)&pbcq->phb,
+ object_property_allow_set_link,
+ OBJ_PROP_LINK_STRONG,
+ &error_abort);
+}
+
+static void pnv_pbcq_class_init(ObjectClass *klass, void *data)
+{
+ DeviceClass *dc = DEVICE_CLASS(klass);
+ PnvXScomInterfaceClass *xdc = PNV_XSCOM_INTERFACE_CLASS(klass);
+
+ xdc->dt_xscom = pnv_pbcq_dt_xscom;
+
+ dc->realize = pnv_pbcq_realize;
+ dc->user_creatable = false;
+}
+
+static const TypeInfo pnv_pbcq_type_info = {
+ .name = TYPE_PNV_PBCQ,
+ .parent = TYPE_DEVICE,
+ .instance_size = sizeof(PnvPBCQState),
+ .instance_init = phb3_pbcq_instance_init,
+ .class_init = pnv_pbcq_class_init,
+ .interfaces = (InterfaceInfo[]) {
+ { TYPE_PNV_XSCOM_INTERFACE },
+ { }
+ }
+};
+
+static void pnv_pbcq_register_types(void)
+{
+ type_register_static(&pnv_pbcq_type_info);
+}
+
+type_init(pnv_pbcq_register_types)
diff --git a/hw/pci-host/pnv_phb4.c b/hw/pci-host/pnv_phb4.c
new file mode 100644
index 0000000000..23cf093928
--- /dev/null
+++ b/hw/pci-host/pnv_phb4.c
@@ -0,0 +1,1439 @@
+/*
+ * QEMU PowerPC PowerNV (POWER9) PHB4 model
+ *
+ * Copyright (c) 2018-2020, IBM Corporation.
+ *
+ * This code is licensed under the GPL version 2 or later. See the
+ * COPYING file in the top-level directory.
+ */
+#include "qemu/osdep.h"
+#include "qemu/log.h"
+#include "qapi/visitor.h"
+#include "qapi/error.h"
+#include "qemu-common.h"
+#include "monitor/monitor.h"
+#include "target/ppc/cpu.h"
+#include "hw/pci-host/pnv_phb4_regs.h"
+#include "hw/pci-host/pnv_phb4.h"
+#include "hw/pci/pcie_host.h"
+#include "hw/pci/pcie_port.h"
+#include "hw/ppc/pnv.h"
+#include "hw/ppc/pnv_xscom.h"
+#include "hw/irq.h"
+#include "hw/qdev-properties.h"
+
+#define phb_error(phb, fmt, ...) \
+ qemu_log_mask(LOG_GUEST_ERROR, "phb4[%d:%d]: " fmt "\n", \
+ (phb)->chip_id, (phb)->phb_id, ## __VA_ARGS__)
+
+/*
+ * QEMU version of the GETFIELD/SETFIELD macros
+ *
+ * These are common with the PnvXive model.
+ */
+static inline uint64_t GETFIELD(uint64_t mask, uint64_t word)
+{
+ return (word & mask) >> ctz64(mask);
+}
+
+static inline uint64_t SETFIELD(uint64_t mask, uint64_t word,
+ uint64_t value)
+{
+ return (word & ~mask) | ((value << ctz64(mask)) & mask);
+}
+
+static PCIDevice *pnv_phb4_find_cfg_dev(PnvPHB4 *phb)
+{
+ PCIHostState *pci = PCI_HOST_BRIDGE(phb);
+ uint64_t addr = phb->regs[PHB_CONFIG_ADDRESS >> 3];
+ uint8_t bus, devfn;
+
+ if (!(addr >> 63)) {
+ return NULL;
+ }
+ bus = (addr >> 52) & 0xff;
+ devfn = (addr >> 44) & 0xff;
+
+ /* We don't access the root complex this way */
+ if (bus == 0 && devfn == 0) {
+ return NULL;
+ }
+ return pci_find_device(pci->bus, bus, devfn);
+}
+
+/*
+ * The CONFIG_DATA register expects little endian accesses, but as the
+ * region is big endian, we have to swap the value.
+ */
+static void pnv_phb4_config_write(PnvPHB4 *phb, unsigned off,
+ unsigned size, uint64_t val)
+{
+ uint32_t cfg_addr, limit;
+ PCIDevice *pdev;
+
+ pdev = pnv_phb4_find_cfg_dev(phb);
+ if (!pdev) {
+ return;
+ }
+ cfg_addr = (phb->regs[PHB_CONFIG_ADDRESS >> 3] >> 32) & 0xffc;
+ cfg_addr |= off;
+ limit = pci_config_size(pdev);
+ if (limit <= cfg_addr) {
+ /*
+ * conventional pci device can be behind pcie-to-pci bridge.
+ * 256 <= addr < 4K has no effects.
+ */
+ return;
+ }
+ switch (size) {
+ case 1:
+ break;
+ case 2:
+ val = bswap16(val);
+ break;
+ case 4:
+ val = bswap32(val);
+ break;
+ default:
+ g_assert_not_reached();
+ }
+ pci_host_config_write_common(pdev, cfg_addr, limit, val, size);
+}
+
+static uint64_t pnv_phb4_config_read(PnvPHB4 *phb, unsigned off,
+ unsigned size)
+{
+ uint32_t cfg_addr, limit;
+ PCIDevice *pdev;
+ uint64_t val;
+
+ pdev = pnv_phb4_find_cfg_dev(phb);
+ if (!pdev) {
+ return ~0ull;
+ }
+ cfg_addr = (phb->regs[PHB_CONFIG_ADDRESS >> 3] >> 32) & 0xffc;
+ cfg_addr |= off;
+ limit = pci_config_size(pdev);
+ if (limit <= cfg_addr) {
+ /*
+ * conventional pci device can be behind pcie-to-pci bridge.
+ * 256 <= addr < 4K has no effects.
+ */
+ return ~0ull;
+ }
+ val = pci_host_config_read_common(pdev, cfg_addr, limit, size);
+ switch (size) {
+ case 1:
+ return val;
+ case 2:
+ return bswap16(val);
+ case 4:
+ return bswap32(val);
+ default:
+ g_assert_not_reached();
+ }
+}
+
+/*
+ * Root complex register accesses are memory mapped.
+ */
+static void pnv_phb4_rc_config_write(PnvPHB4 *phb, unsigned off,
+ unsigned size, uint64_t val)
+{
+ PCIHostState *pci = PCI_HOST_BRIDGE(phb);
+ PCIDevice *pdev;
+
+ if (size != 4) {
+ phb_error(phb, "rc_config_write invalid size %d\n", size);
+ return;
+ }
+
+ pdev = pci_find_device(pci->bus, 0, 0);
+ assert(pdev);
+
+ pci_host_config_write_common(pdev, off, PHB_RC_CONFIG_SIZE,
+ bswap32(val), 4);
+}
+
+static uint64_t pnv_phb4_rc_config_read(PnvPHB4 *phb, unsigned off,
+ unsigned size)
+{
+ PCIHostState *pci = PCI_HOST_BRIDGE(phb);
+ PCIDevice *pdev;
+ uint64_t val;
+
+ if (size != 4) {
+ phb_error(phb, "rc_config_read invalid size %d\n", size);
+ return ~0ull;
+ }
+
+ pdev = pci_find_device(pci->bus, 0, 0);
+ assert(pdev);
+
+ val = pci_host_config_read_common(pdev, off, PHB_RC_CONFIG_SIZE, 4);
+ return bswap32(val);
+}
+
+static void pnv_phb4_check_mbt(PnvPHB4 *phb, uint32_t index)
+{
+ uint64_t base, start, size, mbe0, mbe1;
+ MemoryRegion *parent;
+ char name[64];
+
+ /* Unmap first */
+ if (memory_region_is_mapped(&phb->mr_mmio[index])) {
+ /* Should we destroy it in RCU friendly way... ? */
+ memory_region_del_subregion(phb->mr_mmio[index].container,
+ &phb->mr_mmio[index]);
+ }
+
+ /* Get table entry */
+ mbe0 = phb->ioda_MBT[(index << 1)];
+ mbe1 = phb->ioda_MBT[(index << 1) + 1];
+
+ if (!(mbe0 & IODA3_MBT0_ENABLE)) {
+ return;
+ }
+
+ /* Grab geometry from registers */
+ base = GETFIELD(IODA3_MBT0_BASE_ADDR, mbe0) << 12;
+ size = GETFIELD(IODA3_MBT1_MASK, mbe1) << 12;
+ size |= 0xff00000000000000ull;
+ size = ~size + 1;
+
+ /* Calculate PCI side start address based on M32/M64 window type */
+ if (mbe0 & IODA3_MBT0_TYPE_M32) {
+ start = phb->regs[PHB_M32_START_ADDR >> 3];
+ if ((start + size) > 0x100000000ull) {
+ phb_error(phb, "M32 set beyond 4GB boundary !");
+ size = 0x100000000 - start;
+ }
+ } else {
+ start = base | (phb->regs[PHB_M64_UPPER_BITS >> 3]);
+ }
+
+ /* TODO: Figure out how to implemet/decode AOMASK */
+
+ /* Check if it matches an enabled MMIO region in the PEC stack */
+ if (memory_region_is_mapped(&phb->stack->mmbar0) &&
+ base >= phb->stack->mmio0_base &&
+ (base + size) <= (phb->stack->mmio0_base + phb->stack->mmio0_size)) {
+ parent = &phb->stack->mmbar0;
+ base -= phb->stack->mmio0_base;
+ } else if (memory_region_is_mapped(&phb->stack->mmbar1) &&
+ base >= phb->stack->mmio1_base &&
+ (base + size) <= (phb->stack->mmio1_base + phb->stack->mmio1_size)) {
+ parent = &phb->stack->mmbar1;
+ base -= phb->stack->mmio1_base;
+ } else {
+ phb_error(phb, "PHB MBAR %d out of parent bounds", index);
+ return;
+ }
+
+ /* Create alias (better name ?) */
+ snprintf(name, sizeof(name), "phb4-mbar%d", index);
+ memory_region_init_alias(&phb->mr_mmio[index], OBJECT(phb), name,
+ &phb->pci_mmio, start, size);
+ memory_region_add_subregion(parent, base, &phb->mr_mmio[index]);
+}
+
+static void pnv_phb4_check_all_mbt(PnvPHB4 *phb)
+{
+ uint64_t i;
+ uint32_t num_windows = phb->big_phb ? PNV_PHB4_MAX_MMIO_WINDOWS :
+ PNV_PHB4_MIN_MMIO_WINDOWS;
+
+ for (i = 0; i < num_windows; i++) {
+ pnv_phb4_check_mbt(phb, i);
+ }
+}
+
+static uint64_t *pnv_phb4_ioda_access(PnvPHB4 *phb,
+ unsigned *out_table, unsigned *out_idx)
+{
+ uint64_t adreg = phb->regs[PHB_IODA_ADDR >> 3];
+ unsigned int index = GETFIELD(PHB_IODA_AD_TADR, adreg);
+ unsigned int table = GETFIELD(PHB_IODA_AD_TSEL, adreg);
+ unsigned int mask;
+ uint64_t *tptr = NULL;
+
+ switch (table) {
+ case IODA3_TBL_LIST:
+ tptr = phb->ioda_LIST;
+ mask = 7;
+ break;
+ case IODA3_TBL_MIST:
+ tptr = phb->ioda_MIST;
+ mask = phb->big_phb ? PNV_PHB4_MAX_MIST : (PNV_PHB4_MAX_MIST >> 1);
+ mask -= 1;
+ break;
+ case IODA3_TBL_RCAM:
+ mask = phb->big_phb ? 127 : 63;
+ break;
+ case IODA3_TBL_MRT:
+ mask = phb->big_phb ? 15 : 7;
+ break;
+ case IODA3_TBL_PESTA:
+ case IODA3_TBL_PESTB:
+ mask = phb->big_phb ? PNV_PHB4_MAX_PEs : (PNV_PHB4_MAX_PEs >> 1);
+ mask -= 1;
+ break;
+ case IODA3_TBL_TVT:
+ tptr = phb->ioda_TVT;
+ mask = phb->big_phb ? PNV_PHB4_MAX_TVEs : (PNV_PHB4_MAX_TVEs >> 1);
+ mask -= 1;
+ break;
+ case IODA3_TBL_TCR:
+ case IODA3_TBL_TDR:
+ mask = phb->big_phb ? 1023 : 511;
+ break;
+ case IODA3_TBL_MBT:
+ tptr = phb->ioda_MBT;
+ mask = phb->big_phb ? PNV_PHB4_MAX_MBEs : (PNV_PHB4_MAX_MBEs >> 1);
+ mask -= 1;
+ break;
+ case IODA3_TBL_MDT:
+ tptr = phb->ioda_MDT;
+ mask = phb->big_phb ? PNV_PHB4_MAX_PEs : (PNV_PHB4_MAX_PEs >> 1);
+ mask -= 1;
+ break;
+ case IODA3_TBL_PEEV:
+ tptr = phb->ioda_PEEV;
+ mask = phb->big_phb ? PNV_PHB4_MAX_PEEVs : (PNV_PHB4_MAX_PEEVs >> 1);
+ mask -= 1;
+ break;
+ default:
+ phb_error(phb, "invalid IODA table %d", table);
+ return NULL;
+ }
+ index &= mask;
+ if (out_idx) {
+ *out_idx = index;
+ }
+ if (out_table) {
+ *out_table = table;
+ }
+ if (tptr) {
+ tptr += index;
+ }
+ if (adreg & PHB_IODA_AD_AUTOINC) {
+ index = (index + 1) & mask;
+ adreg = SETFIELD(PHB_IODA_AD_TADR, adreg, index);
+ }
+
+ phb->regs[PHB_IODA_ADDR >> 3] = adreg;
+ return tptr;
+}
+
+static uint64_t pnv_phb4_ioda_read(PnvPHB4 *phb)
+{
+ unsigned table, idx;
+ uint64_t *tptr;
+
+ tptr = pnv_phb4_ioda_access(phb, &table, &idx);
+ if (!tptr) {
+ /* Special PESTA case */
+ if (table == IODA3_TBL_PESTA) {
+ return ((uint64_t)(phb->ioda_PEST_AB[idx] & 1)) << 63;
+ } else if (table == IODA3_TBL_PESTB) {
+ return ((uint64_t)(phb->ioda_PEST_AB[idx] & 2)) << 62;
+ }
+ /* Return 0 on unsupported tables, not ff's */
+ return 0;
+ }
+ return *tptr;
+}
+
+static void pnv_phb4_ioda_write(PnvPHB4 *phb, uint64_t val)
+{
+ unsigned table, idx;
+ uint64_t *tptr;
+
+ tptr = pnv_phb4_ioda_access(phb, &table, &idx);
+ if (!tptr) {
+ /* Special PESTA case */
+ if (table == IODA3_TBL_PESTA) {
+ phb->ioda_PEST_AB[idx] &= ~1;
+ phb->ioda_PEST_AB[idx] |= (val >> 63) & 1;
+ } else if (table == IODA3_TBL_PESTB) {
+ phb->ioda_PEST_AB[idx] &= ~2;
+ phb->ioda_PEST_AB[idx] |= (val >> 62) & 2;
+ }
+ return;
+ }
+
+ /* Handle side effects */
+ switch (table) {
+ case IODA3_TBL_LIST:
+ break;
+ case IODA3_TBL_MIST: {
+ /* Special mask for MIST partial write */
+ uint64_t adreg = phb->regs[PHB_IODA_ADDR >> 3];
+ uint32_t mmask = GETFIELD(PHB_IODA_AD_MIST_PWV, adreg);
+ uint64_t v = *tptr;
+ if (mmask == 0) {
+ mmask = 0xf;
+ }
+ if (mmask & 8) {
+ v &= 0x0000ffffffffffffull;
+ v |= 0xcfff000000000000ull & val;
+ }
+ if (mmask & 4) {
+ v &= 0xffff0000ffffffffull;
+ v |= 0x0000cfff00000000ull & val;
+ }
+ if (mmask & 2) {
+ v &= 0xffffffff0000ffffull;
+ v |= 0x00000000cfff0000ull & val;
+ }
+ if (mmask & 1) {
+ v &= 0xffffffffffff0000ull;
+ v |= 0x000000000000cfffull & val;
+ }
+ *tptr = val;
+ break;
+ }
+ case IODA3_TBL_MBT:
+ *tptr = val;
+
+ /* Copy accross the valid bit to the other half */
+ phb->ioda_MBT[idx ^ 1] &= 0x7fffffffffffffffull;
+ phb->ioda_MBT[idx ^ 1] |= 0x8000000000000000ull & val;
+
+ /* Update mappings */
+ pnv_phb4_check_mbt(phb, idx >> 1);
+ break;
+ default:
+ *tptr = val;
+ }
+}
+
+static void pnv_phb4_rtc_invalidate(PnvPHB4 *phb, uint64_t val)
+{
+ PnvPhb4DMASpace *ds;
+
+ /* Always invalidate all for now ... */
+ QLIST_FOREACH(ds, &phb->dma_spaces, list) {
+ ds->pe_num = PHB_INVALID_PE;
+ }
+}
+
+static void pnv_phb4_update_msi_regions(PnvPhb4DMASpace *ds)
+{
+ uint64_t cfg = ds->phb->regs[PHB_PHB4_CONFIG >> 3];
+
+ if (cfg & PHB_PHB4C_32BIT_MSI_EN) {
+ if (!memory_region_is_mapped(MEMORY_REGION(&ds->msi32_mr))) {
+ memory_region_add_subregion(MEMORY_REGION(&ds->dma_mr),
+ 0xffff0000, &ds->msi32_mr);
+ }
+ } else {
+ if (memory_region_is_mapped(MEMORY_REGION(&ds->msi32_mr))) {
+ memory_region_del_subregion(MEMORY_REGION(&ds->dma_mr),
+ &ds->msi32_mr);
+ }
+ }
+
+ if (cfg & PHB_PHB4C_64BIT_MSI_EN) {
+ if (!memory_region_is_mapped(MEMORY_REGION(&ds->msi64_mr))) {
+ memory_region_add_subregion(MEMORY_REGION(&ds->dma_mr),
+ (1ull << 60), &ds->msi64_mr);
+ }
+ } else {
+ if (memory_region_is_mapped(MEMORY_REGION(&ds->msi64_mr))) {
+ memory_region_del_subregion(MEMORY_REGION(&ds->dma_mr),
+ &ds->msi64_mr);
+ }
+ }
+}
+
+static void pnv_phb4_update_all_msi_regions(PnvPHB4 *phb)
+{
+ PnvPhb4DMASpace *ds;
+
+ QLIST_FOREACH(ds, &phb->dma_spaces, list) {
+ pnv_phb4_update_msi_regions(ds);
+ }
+}
+
+static void pnv_phb4_update_xsrc(PnvPHB4 *phb)
+{
+ int shift, flags, i, lsi_base;
+ XiveSource *xsrc = &phb->xsrc;
+
+ /* The XIVE source characteristics can be set at run time */
+ if (phb->regs[PHB_CTRLR >> 3] & PHB_CTRLR_IRQ_PGSZ_64K) {
+ shift = XIVE_ESB_64K;
+ } else {
+ shift = XIVE_ESB_4K;
+ }
+ if (phb->regs[PHB_CTRLR >> 3] & PHB_CTRLR_IRQ_STORE_EOI) {
+ flags = XIVE_SRC_STORE_EOI;
+ } else {
+ flags = 0;
+ }
+
+ phb->xsrc.esb_shift = shift;
+ phb->xsrc.esb_flags = flags;
+
+ lsi_base = GETFIELD(PHB_LSI_SRC_ID, phb->regs[PHB_LSI_SOURCE_ID >> 3]);
+ lsi_base <<= 3;
+
+ /* TODO: handle reset values of PHB_LSI_SRC_ID */
+ if (!lsi_base) {
+ return;
+ }
+
+ /* TODO: need a xive_source_irq_reset_lsi() */
+ bitmap_zero(xsrc->lsi_map, xsrc->nr_irqs);
+
+ for (i = 0; i < xsrc->nr_irqs; i++) {
+ bool msi = (i < lsi_base || i >= (lsi_base + 8));
+ if (!msi) {
+ xive_source_irq_set_lsi(xsrc, i);
+ }
+ }
+}
+
+static void pnv_phb4_reg_write(void *opaque, hwaddr off, uint64_t val,
+ unsigned size)
+{
+ PnvPHB4 *phb = PNV_PHB4(opaque);
+ bool changed;
+
+ /* Special case outbound configuration data */
+ if ((off & 0xfffc) == PHB_CONFIG_DATA) {
+ pnv_phb4_config_write(phb, off & 0x3, size, val);
+ return;
+ }
+
+ /* Special case RC configuration space */
+ if ((off & 0xf800) == PHB_RC_CONFIG_BASE) {
+ pnv_phb4_rc_config_write(phb, off & 0x7ff, size, val);
+ return;
+ }
+
+ /* Other registers are 64-bit only */
+ if (size != 8 || off & 0x7) {
+ phb_error(phb, "Invalid register access, offset: 0x%"PRIx64" size: %d",
+ off, size);
+ return;
+ }
+
+ /* Handle masking */
+ switch (off) {
+ case PHB_LSI_SOURCE_ID:
+ val &= PHB_LSI_SRC_ID;
+ break;
+ case PHB_M64_UPPER_BITS:
+ val &= 0xff00000000000000ull;
+ break;
+ /* TCE Kill */
+ case PHB_TCE_KILL:
+ /* Clear top 3 bits which HW does to indicate successful queuing */
+ val &= ~(PHB_TCE_KILL_ALL | PHB_TCE_KILL_PE | PHB_TCE_KILL_ONE);
+ break;
+ case PHB_Q_DMA_R:
+ /*
+ * This is enough logic to make SW happy but we aren't
+ * actually quiescing the DMAs
+ */
+ if (val & PHB_Q_DMA_R_AUTORESET) {
+ val = 0;
+ } else {
+ val &= PHB_Q_DMA_R_QUIESCE_DMA;
+ }
+ break;
+ /* LEM stuff */
+ case PHB_LEM_FIR_AND_MASK:
+ phb->regs[PHB_LEM_FIR_ACCUM >> 3] &= val;
+ return;
+ case PHB_LEM_FIR_OR_MASK:
+ phb->regs[PHB_LEM_FIR_ACCUM >> 3] |= val;
+ return;
+ case PHB_LEM_ERROR_AND_MASK:
+ phb->regs[PHB_LEM_ERROR_MASK >> 3] &= val;
+ return;
+ case PHB_LEM_ERROR_OR_MASK:
+ phb->regs[PHB_LEM_ERROR_MASK >> 3] |= val;
+ return;
+ case PHB_LEM_WOF:
+ val = 0;
+ break;
+ /* TODO: More regs ..., maybe create a table with masks... */
+
+ /* Read only registers */
+ case PHB_CPU_LOADSTORE_STATUS:
+ case PHB_ETU_ERR_SUMMARY:
+ case PHB_PHB4_GEN_CAP:
+ case PHB_PHB4_TCE_CAP:
+ case PHB_PHB4_IRQ_CAP:
+ case PHB_PHB4_EEH_CAP:
+ return;
+ }
+
+ /* Record whether it changed */
+ changed = phb->regs[off >> 3] != val;
+
+ /* Store in register cache first */
+ phb->regs[off >> 3] = val;
+
+ /* Handle side effects */
+ switch (off) {
+ case PHB_PHB4_CONFIG:
+ if (changed) {
+ pnv_phb4_update_all_msi_regions(phb);
+ }
+ break;
+ case PHB_M32_START_ADDR:
+ case PHB_M64_UPPER_BITS:
+ if (changed) {
+ pnv_phb4_check_all_mbt(phb);
+ }
+ break;
+
+ /* IODA table accesses */
+ case PHB_IODA_DATA0:
+ pnv_phb4_ioda_write(phb, val);
+ break;
+
+ /* RTC invalidation */
+ case PHB_RTC_INVALIDATE:
+ pnv_phb4_rtc_invalidate(phb, val);
+ break;
+
+ /* PHB Control (Affects XIVE source) */
+ case PHB_CTRLR:
+ case PHB_LSI_SOURCE_ID:
+ pnv_phb4_update_xsrc(phb);
+ break;
+
+ /* Silent simple writes */
+ case PHB_ASN_CMPM:
+ case PHB_CONFIG_ADDRESS:
+ case PHB_IODA_ADDR:
+ case PHB_TCE_KILL:
+ case PHB_TCE_SPEC_CTL:
+ case PHB_PEST_BAR:
+ case PHB_PELTV_BAR:
+ case PHB_RTT_BAR:
+ case PHB_LEM_FIR_ACCUM:
+ case PHB_LEM_ERROR_MASK:
+ case PHB_LEM_ACTION0:
+ case PHB_LEM_ACTION1:
+ case PHB_TCE_TAG_ENABLE:
+ case PHB_INT_NOTIFY_ADDR:
+ case PHB_INT_NOTIFY_INDEX:
+ case PHB_DMARD_SYNC:
+ break;
+
+ /* Noise on anything else */
+ default:
+ qemu_log_mask(LOG_UNIMP, "phb4: reg_write 0x%"PRIx64"=%"PRIx64"\n",
+ off, val);
+ }
+}
+
+static uint64_t pnv_phb4_reg_read(void *opaque, hwaddr off, unsigned size)
+{
+ PnvPHB4 *phb = PNV_PHB4(opaque);
+ uint64_t val;
+
+ if ((off & 0xfffc) == PHB_CONFIG_DATA) {
+ return pnv_phb4_config_read(phb, off & 0x3, size);
+ }
+
+ /* Special case RC configuration space */
+ if ((off & 0xf800) == PHB_RC_CONFIG_BASE) {
+ return pnv_phb4_rc_config_read(phb, off & 0x7ff, size);
+ }
+
+ /* Other registers are 64-bit only */
+ if (size != 8 || off & 0x7) {
+ phb_error(phb, "Invalid register access, offset: 0x%"PRIx64" size: %d",
+ off, size);
+ return ~0ull;
+ }
+
+ /* Default read from cache */
+ val = phb->regs[off >> 3];
+
+ switch (off) {
+ case PHB_VERSION:
+ return phb->version;
+
+ /* Read-only */
+ case PHB_PHB4_GEN_CAP:
+ return 0xe4b8000000000000ull;
+ case PHB_PHB4_TCE_CAP:
+ return phb->big_phb ? 0x4008440000000400ull : 0x2008440000000200ull;
+ case PHB_PHB4_IRQ_CAP:
+ return phb->big_phb ? 0x0800000000001000ull : 0x0800000000000800ull;
+ case PHB_PHB4_EEH_CAP:
+ return phb->big_phb ? 0x2000000000000000ull : 0x1000000000000000ull;
+
+ /* IODA table accesses */
+ case PHB_IODA_DATA0:
+ return pnv_phb4_ioda_read(phb);
+
+ /* Link training always appears trained */
+ case PHB_PCIE_DLP_TRAIN_CTL:
+ /* TODO: Do something sensible with speed ? */
+ return PHB_PCIE_DLP_INBAND_PRESENCE | PHB_PCIE_DLP_TL_LINKACT;
+
+ /* DMA read sync: make it look like it's complete */
+ case PHB_DMARD_SYNC:
+ return PHB_DMARD_SYNC_COMPLETE;
+
+ /* Silent simple reads */
+ case PHB_LSI_SOURCE_ID:
+ case PHB_CPU_LOADSTORE_STATUS:
+ case PHB_ASN_CMPM:
+ case PHB_PHB4_CONFIG:
+ case PHB_M32_START_ADDR:
+ case PHB_CONFIG_ADDRESS:
+ case PHB_IODA_ADDR:
+ case PHB_RTC_INVALIDATE:
+ case PHB_TCE_KILL:
+ case PHB_TCE_SPEC_CTL:
+ case PHB_PEST_BAR:
+ case PHB_PELTV_BAR:
+ case PHB_RTT_BAR:
+ case PHB_M64_UPPER_BITS:
+ case PHB_CTRLR:
+ case PHB_LEM_FIR_ACCUM:
+ case PHB_LEM_ERROR_MASK:
+ case PHB_LEM_ACTION0:
+ case PHB_LEM_ACTION1:
+ case PHB_TCE_TAG_ENABLE:
+ case PHB_INT_NOTIFY_ADDR:
+ case PHB_INT_NOTIFY_INDEX:
+ case PHB_Q_DMA_R:
+ case PHB_ETU_ERR_SUMMARY:
+ break;
+
+ /* Noise on anything else */
+ default:
+ qemu_log_mask(LOG_UNIMP, "phb4: reg_read 0x%"PRIx64"=%"PRIx64"\n",
+ off, val);
+ }
+ return val;
+}
+
+static const MemoryRegionOps pnv_phb4_reg_ops = {
+ .read = pnv_phb4_reg_read,
+ .write = pnv_phb4_reg_write,
+ .valid.min_access_size = 1,
+ .valid.max_access_size = 8,
+ .impl.min_access_size = 1,
+ .impl.max_access_size = 8,
+ .endianness = DEVICE_BIG_ENDIAN,
+};
+
+static uint64_t pnv_phb4_xscom_read(void *opaque, hwaddr addr, unsigned size)
+{
+ PnvPHB4 *phb = PNV_PHB4(opaque);
+ uint32_t reg = addr >> 3;
+ uint64_t val;
+ hwaddr offset;
+
+ switch (reg) {
+ case PHB_SCOM_HV_IND_ADDR:
+ return phb->scom_hv_ind_addr_reg;
+
+ case PHB_SCOM_HV_IND_DATA:
+ if (!(phb->scom_hv_ind_addr_reg & PHB_SCOM_HV_IND_ADDR_VALID)) {
+ phb_error(phb, "Invalid indirect address");
+ return ~0ull;
+ }
+ size = (phb->scom_hv_ind_addr_reg & PHB_SCOM_HV_IND_ADDR_4B) ? 4 : 8;
+ offset = GETFIELD(PHB_SCOM_HV_IND_ADDR_ADDR, phb->scom_hv_ind_addr_reg);
+ val = pnv_phb4_reg_read(phb, offset, size);
+ if (phb->scom_hv_ind_addr_reg & PHB_SCOM_HV_IND_ADDR_AUTOINC) {
+ offset += size;
+ offset &= 0x3fff;
+ phb->scom_hv_ind_addr_reg = SETFIELD(PHB_SCOM_HV_IND_ADDR_ADDR,
+ phb->scom_hv_ind_addr_reg,
+ offset);
+ }
+ return val;
+ case PHB_SCOM_ETU_LEM_FIR:
+ case PHB_SCOM_ETU_LEM_FIR_AND:
+ case PHB_SCOM_ETU_LEM_FIR_OR:
+ case PHB_SCOM_ETU_LEM_FIR_MSK:
+ case PHB_SCOM_ETU_LEM_ERR_MSK_AND:
+ case PHB_SCOM_ETU_LEM_ERR_MSK_OR:
+ case PHB_SCOM_ETU_LEM_ACT0:
+ case PHB_SCOM_ETU_LEM_ACT1:
+ case PHB_SCOM_ETU_LEM_WOF:
+ offset = ((reg - PHB_SCOM_ETU_LEM_FIR) << 3) + PHB_LEM_FIR_ACCUM;
+ return pnv_phb4_reg_read(phb, offset, size);
+ case PHB_SCOM_ETU_PMON_CONFIG:
+ case PHB_SCOM_ETU_PMON_CTR0:
+ case PHB_SCOM_ETU_PMON_CTR1:
+ case PHB_SCOM_ETU_PMON_CTR2:
+ case PHB_SCOM_ETU_PMON_CTR3:
+ offset = ((reg - PHB_SCOM_ETU_PMON_CONFIG) << 3) + PHB_PERFMON_CONFIG;
+ return pnv_phb4_reg_read(phb, offset, size);
+
+ default:
+ qemu_log_mask(LOG_UNIMP, "phb4: xscom_read 0x%"HWADDR_PRIx"\n", addr);
+ return ~0ull;
+ }
+}
+
+static void pnv_phb4_xscom_write(void *opaque, hwaddr addr,
+ uint64_t val, unsigned size)
+{
+ PnvPHB4 *phb = PNV_PHB4(opaque);
+ uint32_t reg = addr >> 3;
+ hwaddr offset;
+
+ switch (reg) {
+ case PHB_SCOM_HV_IND_ADDR:
+ phb->scom_hv_ind_addr_reg = val & 0xe000000000001fff;
+ break;
+ case PHB_SCOM_HV_IND_DATA:
+ if (!(phb->scom_hv_ind_addr_reg & PHB_SCOM_HV_IND_ADDR_VALID)) {
+ phb_error(phb, "Invalid indirect address");
+ break;
+ }
+ size = (phb->scom_hv_ind_addr_reg & PHB_SCOM_HV_IND_ADDR_4B) ? 4 : 8;
+ offset = GETFIELD(PHB_SCOM_HV_IND_ADDR_ADDR, phb->scom_hv_ind_addr_reg);
+ pnv_phb4_reg_write(phb, offset, val, size);
+ if (phb->scom_hv_ind_addr_reg & PHB_SCOM_HV_IND_ADDR_AUTOINC) {
+ offset += size;
+ offset &= 0x3fff;
+ phb->scom_hv_ind_addr_reg = SETFIELD(PHB_SCOM_HV_IND_ADDR_ADDR,
+ phb->scom_hv_ind_addr_reg,
+ offset);
+ }
+ break;
+ case PHB_SCOM_ETU_LEM_FIR:
+ case PHB_SCOM_ETU_LEM_FIR_AND:
+ case PHB_SCOM_ETU_LEM_FIR_OR:
+ case PHB_SCOM_ETU_LEM_FIR_MSK:
+ case PHB_SCOM_ETU_LEM_ERR_MSK_AND:
+ case PHB_SCOM_ETU_LEM_ERR_MSK_OR:
+ case PHB_SCOM_ETU_LEM_ACT0:
+ case PHB_SCOM_ETU_LEM_ACT1:
+ case PHB_SCOM_ETU_LEM_WOF:
+ offset = ((reg - PHB_SCOM_ETU_LEM_FIR) << 3) + PHB_LEM_FIR_ACCUM;
+ pnv_phb4_reg_write(phb, offset, val, size);
+ break;
+ case PHB_SCOM_ETU_PMON_CONFIG:
+ case PHB_SCOM_ETU_PMON_CTR0:
+ case PHB_SCOM_ETU_PMON_CTR1:
+ case PHB_SCOM_ETU_PMON_CTR2:
+ case PHB_SCOM_ETU_PMON_CTR3:
+ offset = ((reg - PHB_SCOM_ETU_PMON_CONFIG) << 3) + PHB_PERFMON_CONFIG;
+ pnv_phb4_reg_write(phb, offset, val, size);
+ break;
+ default:
+ qemu_log_mask(LOG_UNIMP, "phb4: xscom_write 0x%"HWADDR_PRIx
+ "=%"PRIx64"\n", addr, val);
+ }
+}
+
+const MemoryRegionOps pnv_phb4_xscom_ops = {
+ .read = pnv_phb4_xscom_read,
+ .write = pnv_phb4_xscom_write,
+ .valid.min_access_size = 8,
+ .valid.max_access_size = 8,
+ .impl.min_access_size = 8,
+ .impl.max_access_size = 8,
+ .endianness = DEVICE_BIG_ENDIAN,
+};
+
+static int pnv_phb4_map_irq(PCIDevice *pci_dev, int irq_num)
+{
+ /* Check that out properly ... */
+ return irq_num & 3;
+}
+
+static void pnv_phb4_set_irq(void *opaque, int irq_num, int level)
+{
+ PnvPHB4 *phb = PNV_PHB4(opaque);
+ uint32_t lsi_base;
+
+ /* LSI only ... */
+ if (irq_num > 3) {
+ phb_error(phb, "IRQ %x is not an LSI", irq_num);
+ }
+ lsi_base = GETFIELD(PHB_LSI_SRC_ID, phb->regs[PHB_LSI_SOURCE_ID >> 3]);
+ lsi_base <<= 3;
+ qemu_set_irq(phb->qirqs[lsi_base + irq_num], level);
+}
+
+static bool pnv_phb4_resolve_pe(PnvPhb4DMASpace *ds)
+{
+ uint64_t rtt, addr;
+ uint16_t rte;
+ int bus_num;
+ int num_PEs;
+
+ /* Already resolved ? */
+ if (ds->pe_num != PHB_INVALID_PE) {
+ return true;
+ }
+
+ /* We need to lookup the RTT */
+ rtt = ds->phb->regs[PHB_RTT_BAR >> 3];
+ if (!(rtt & PHB_RTT_BAR_ENABLE)) {
+ phb_error(ds->phb, "DMA with RTT BAR disabled !");
+ /* Set error bits ? fence ? ... */
+ return false;
+ }
+
+ /* Read RTE */
+ bus_num = pci_bus_num(ds->bus);
+ addr = rtt & PHB_RTT_BASE_ADDRESS_MASK;
+ addr += 2 * ((bus_num << 8) | ds->devfn);
+ if (dma_memory_read(&address_space_memory, addr, &rte, sizeof(rte))) {
+ phb_error(ds->phb, "Failed to read RTT entry at 0x%"PRIx64, addr);
+ /* Set error bits ? fence ? ... */
+ return false;
+ }
+ rte = be16_to_cpu(rte);
+
+ /* Fail upon reading of invalid PE# */
+ num_PEs = ds->phb->big_phb ? PNV_PHB4_MAX_PEs : (PNV_PHB4_MAX_PEs >> 1);
+ if (rte >= num_PEs) {
+ phb_error(ds->phb, "RTE for RID 0x%x invalid (%04x", ds->devfn, rte);
+ rte &= num_PEs - 1;
+ }
+ ds->pe_num = rte;
+ return true;
+}
+
+static void pnv_phb4_translate_tve(PnvPhb4DMASpace *ds, hwaddr addr,
+ bool is_write, uint64_t tve,
+ IOMMUTLBEntry *tlb)
+{
+ uint64_t tta = GETFIELD(IODA3_TVT_TABLE_ADDR, tve);
+ int32_t lev = GETFIELD(IODA3_TVT_NUM_LEVELS, tve);
+ uint32_t tts = GETFIELD(IODA3_TVT_TCE_TABLE_SIZE, tve);
+ uint32_t tps = GETFIELD(IODA3_TVT_IO_PSIZE, tve);
+
+ /* Invalid levels */
+ if (lev > 4) {
+ phb_error(ds->phb, "Invalid #levels in TVE %d", lev);
+ return;
+ }
+
+ /* Invalid entry */
+ if (tts == 0) {
+ phb_error(ds->phb, "Access to invalid TVE");
+ return;
+ }
+
+ /* IO Page Size of 0 means untranslated, else use TCEs */
+ if (tps == 0) {
+ /* TODO: Handle boundaries */
+
+ /* Use 4k pages like q35 ... for now */
+ tlb->iova = addr & 0xfffffffffffff000ull;
+ tlb->translated_addr = addr & 0x0003fffffffff000ull;
+ tlb->addr_mask = 0xfffull;
+ tlb->perm = IOMMU_RW;
+ } else {
+ uint32_t tce_shift, tbl_shift, sh;
+ uint64_t base, taddr, tce, tce_mask;
+
+ /* Address bits per bottom level TCE entry */
+ tce_shift = tps + 11;
+
+ /* Address bits per table level */
+ tbl_shift = tts + 8;
+
+ /* Top level table base address */
+ base = tta << 12;
+
+ /* Total shift to first level */
+ sh = tbl_shift * lev + tce_shift;
+
+ /* TODO: Limit to support IO page sizes */
+
+ /* TODO: Multi-level untested */
+ while ((lev--) >= 0) {
+ /* Grab the TCE address */
+ taddr = base | (((addr >> sh) & ((1ul << tbl_shift) - 1)) << 3);
+ if (dma_memory_read(&address_space_memory, taddr, &tce,
+ sizeof(tce))) {
+ phb_error(ds->phb, "Failed to read TCE at 0x%"PRIx64, taddr);
+ return;
+ }
+ tce = be64_to_cpu(tce);
+
+ /* Check permission for indirect TCE */
+ if ((lev >= 0) && !(tce & 3)) {
+ phb_error(ds->phb, "Invalid indirect TCE at 0x%"PRIx64, taddr);
+ phb_error(ds->phb, " xlate %"PRIx64":%c TVE=%"PRIx64, addr,
+ is_write ? 'W' : 'R', tve);
+ phb_error(ds->phb, " tta=%"PRIx64" lev=%d tts=%d tps=%d",
+ tta, lev, tts, tps);
+ return;
+ }
+ sh -= tbl_shift;
+ base = tce & ~0xfffull;
+ }
+
+ /* We exit the loop with TCE being the final TCE */
+ tce_mask = ~((1ull << tce_shift) - 1);
+ tlb->iova = addr & tce_mask;
+ tlb->translated_addr = tce & tce_mask;
+ tlb->addr_mask = ~tce_mask;
+ tlb->perm = tce & 3;
+ if ((is_write & !(tce & 2)) || ((!is_write) && !(tce & 1))) {
+ phb_error(ds->phb, "TCE access fault at 0x%"PRIx64, taddr);
+ phb_error(ds->phb, " xlate %"PRIx64":%c TVE=%"PRIx64, addr,
+ is_write ? 'W' : 'R', tve);
+ phb_error(ds->phb, " tta=%"PRIx64" lev=%d tts=%d tps=%d",
+ tta, lev, tts, tps);
+ }
+ }
+}
+
+static IOMMUTLBEntry pnv_phb4_translate_iommu(IOMMUMemoryRegion *iommu,
+ hwaddr addr,
+ IOMMUAccessFlags flag,
+ int iommu_idx)
+{
+ PnvPhb4DMASpace *ds = container_of(iommu, PnvPhb4DMASpace, dma_mr);
+ int tve_sel;
+ uint64_t tve, cfg;
+ IOMMUTLBEntry ret = {
+ .target_as = &address_space_memory,
+ .iova = addr,
+ .translated_addr = 0,
+ .addr_mask = ~(hwaddr)0,
+ .perm = IOMMU_NONE,
+ };
+
+ /* Resolve PE# */
+ if (!pnv_phb4_resolve_pe(ds)) {
+ phb_error(ds->phb, "Failed to resolve PE# for bus @%p (%d) devfn 0x%x",
+ ds->bus, pci_bus_num(ds->bus), ds->devfn);
+ return ret;
+ }
+
+ /* Check top bits */
+ switch (addr >> 60) {
+ case 00:
+ /* DMA or 32-bit MSI ? */
+ cfg = ds->phb->regs[PHB_PHB4_CONFIG >> 3];
+ if ((cfg & PHB_PHB4C_32BIT_MSI_EN) &&
+ ((addr & 0xffffffffffff0000ull) == 0xffff0000ull)) {
+ phb_error(ds->phb, "xlate on 32-bit MSI region");
+ return ret;
+ }
+ /* Choose TVE XXX Use PHB4 Control Register */
+ tve_sel = (addr >> 59) & 1;
+ tve = ds->phb->ioda_TVT[ds->pe_num * 2 + tve_sel];
+ pnv_phb4_translate_tve(ds, addr, flag & IOMMU_WO, tve, &ret);
+ break;
+ case 01:
+ phb_error(ds->phb, "xlate on 64-bit MSI region");
+ break;
+ default:
+ phb_error(ds->phb, "xlate on unsupported address 0x%"PRIx64, addr);
+ }
+ return ret;
+}
+
+#define TYPE_PNV_PHB4_IOMMU_MEMORY_REGION "pnv-phb4-iommu-memory-region"
+#define PNV_PHB4_IOMMU_MEMORY_REGION(obj) \
+ OBJECT_CHECK(IOMMUMemoryRegion, (obj), TYPE_PNV_PHB4_IOMMU_MEMORY_REGION)
+
+static void pnv_phb4_iommu_memory_region_class_init(ObjectClass *klass,
+ void *data)
+{
+ IOMMUMemoryRegionClass *imrc = IOMMU_MEMORY_REGION_CLASS(klass);
+
+ imrc->translate = pnv_phb4_translate_iommu;
+}
+
+static const TypeInfo pnv_phb4_iommu_memory_region_info = {
+ .parent = TYPE_IOMMU_MEMORY_REGION,
+ .name = TYPE_PNV_PHB4_IOMMU_MEMORY_REGION,
+ .class_init = pnv_phb4_iommu_memory_region_class_init,
+};
+
+/*
+ * MSI/MSIX memory region implementation.
+ * The handler handles both MSI and MSIX.
+ */
+static void pnv_phb4_msi_write(void *opaque, hwaddr addr,
+ uint64_t data, unsigned size)
+{
+ PnvPhb4DMASpace *ds = opaque;
+ PnvPHB4 *phb = ds->phb;
+
+ uint32_t src = ((addr >> 4) & 0xffff) | (data & 0x1f);
+
+ /* Resolve PE# */
+ if (!pnv_phb4_resolve_pe(ds)) {
+ phb_error(phb, "Failed to resolve PE# for bus @%p (%d) devfn 0x%x",
+ ds->bus, pci_bus_num(ds->bus), ds->devfn);
+ return;
+ }
+
+ /* TODO: Check it doesn't collide with LSIs */
+ if (src >= phb->xsrc.nr_irqs) {
+ phb_error(phb, "MSI %d out of bounds", src);
+ return;
+ }
+
+ /* TODO: check PE/MSI assignement */
+
+ qemu_irq_pulse(phb->qirqs[src]);
+}
+
+/* There is no .read as the read result is undefined by PCI spec */
+static uint64_t pnv_phb4_msi_read(void *opaque, hwaddr addr, unsigned size)
+{
+ PnvPhb4DMASpace *ds = opaque;
+
+ phb_error(ds->phb, "Invalid MSI read @ 0x%" HWADDR_PRIx, addr);
+ return -1;
+}
+
+static const MemoryRegionOps pnv_phb4_msi_ops = {
+ .read = pnv_phb4_msi_read,
+ .write = pnv_phb4_msi_write,
+ .endianness = DEVICE_LITTLE_ENDIAN
+};
+
+static PnvPhb4DMASpace *pnv_phb4_dma_find(PnvPHB4 *phb, PCIBus *bus, int devfn)
+{
+ PnvPhb4DMASpace *ds;
+
+ QLIST_FOREACH(ds, &phb->dma_spaces, list) {
+ if (ds->bus == bus && ds->devfn == devfn) {
+ break;
+ }
+ }
+ return ds;
+}
+
+static AddressSpace *pnv_phb4_dma_iommu(PCIBus *bus, void *opaque, int devfn)
+{
+ PnvPHB4 *phb = opaque;
+ PnvPhb4DMASpace *ds;
+ char name[32];
+
+ ds = pnv_phb4_dma_find(phb, bus, devfn);
+
+ if (ds == NULL) {
+ ds = g_malloc0(sizeof(PnvPhb4DMASpace));
+ ds->bus = bus;
+ ds->devfn = devfn;
+ ds->pe_num = PHB_INVALID_PE;
+ ds->phb = phb;
+ snprintf(name, sizeof(name), "phb4-%d.%d-iommu", phb->chip_id,
+ phb->phb_id);
+ memory_region_init_iommu(&ds->dma_mr, sizeof(ds->dma_mr),
+ TYPE_PNV_PHB4_IOMMU_MEMORY_REGION,
+ OBJECT(phb), name, UINT64_MAX);
+ address_space_init(&ds->dma_as, MEMORY_REGION(&ds->dma_mr),
+ name);
+ memory_region_init_io(&ds->msi32_mr, OBJECT(phb), &pnv_phb4_msi_ops,
+ ds, "msi32", 0x10000);
+ memory_region_init_io(&ds->msi64_mr, OBJECT(phb), &pnv_phb4_msi_ops,
+ ds, "msi64", 0x100000);
+ pnv_phb4_update_msi_regions(ds);
+
+ QLIST_INSERT_HEAD(&phb->dma_spaces, ds, list);
+ }
+ return &ds->dma_as;
+}
+
+static void pnv_phb4_instance_init(Object *obj)
+{
+ PnvPHB4 *phb = PNV_PHB4(obj);
+
+ QLIST_INIT(&phb->dma_spaces);
+
+ /* XIVE interrupt source object */
+ object_initialize_child(obj, "source", &phb->xsrc, sizeof(XiveSource),
+ TYPE_XIVE_SOURCE, &error_abort, NULL);
+
+ /* Root Port */
+ object_initialize_child(obj, "root", &phb->root, sizeof(phb->root),
+ TYPE_PNV_PHB4_ROOT_PORT, &error_abort, NULL);
+
+ qdev_prop_set_int32(DEVICE(&phb->root), "addr", PCI_DEVFN(0, 0));
+ qdev_prop_set_bit(DEVICE(&phb->root), "multifunction", false);
+}
+
+static void pnv_phb4_realize(DeviceState *dev, Error **errp)
+{
+ PnvPHB4 *phb = PNV_PHB4(dev);
+ PCIHostState *pci = PCI_HOST_BRIDGE(dev);
+ XiveSource *xsrc = &phb->xsrc;
+ Error *local_err = NULL;
+ int nr_irqs;
+ char name[32];
+
+ assert(phb->stack);
+
+ /* Set the "big_phb" flag */
+ phb->big_phb = phb->phb_id == 0 || phb->phb_id == 3;
+
+ /* Controller Registers */
+ snprintf(name, sizeof(name), "phb4-%d.%d-regs", phb->chip_id,
+ phb->phb_id);
+ memory_region_init_io(&phb->mr_regs, OBJECT(phb), &pnv_phb4_reg_ops, phb,
+ name, 0x2000);
+
+ /*
+ * PHB4 doesn't support IO space. However, qemu gets very upset if
+ * we don't have an IO region to anchor IO BARs onto so we just
+ * initialize one which we never hook up to anything
+ */
+
+ snprintf(name, sizeof(name), "phb4-%d.%d-pci-io", phb->chip_id,
+ phb->phb_id);
+ memory_region_init(&phb->pci_io, OBJECT(phb), name, 0x10000);
+
+ snprintf(name, sizeof(name), "phb4-%d.%d-pci-mmio", phb->chip_id,
+ phb->phb_id);
+ memory_region_init(&phb->pci_mmio, OBJECT(phb), name,
+ PCI_MMIO_TOTAL_SIZE);
+
+ pci->bus = pci_register_root_bus(dev, "root-bus",
+ pnv_phb4_set_irq, pnv_phb4_map_irq, phb,
+ &phb->pci_mmio, &phb->pci_io,
+ 0, 4, TYPE_PNV_PHB4_ROOT_BUS);
+ pci_setup_iommu(pci->bus, pnv_phb4_dma_iommu, phb);
+
+ /* Add a single Root port */
+ qdev_prop_set_uint8(DEVICE(&phb->root), "chassis", phb->chip_id);
+ qdev_prop_set_uint16(DEVICE(&phb->root), "slot", phb->phb_id);
+ qdev_set_parent_bus(DEVICE(&phb->root), BUS(pci->bus));
+ qdev_init_nofail(DEVICE(&phb->root));
+
+ /* Setup XIVE Source */
+ if (phb->big_phb) {
+ nr_irqs = PNV_PHB4_MAX_INTs;
+ } else {
+ nr_irqs = PNV_PHB4_MAX_INTs >> 1;
+ }
+ object_property_set_int(OBJECT(xsrc), nr_irqs, "nr-irqs", &error_fatal);
+ object_property_set_link(OBJECT(xsrc), OBJECT(phb), "xive", &error_fatal);
+ object_property_set_bool(OBJECT(xsrc), true, "realized", &local_err);
+ if (local_err) {
+ error_propagate(errp, local_err);
+ return;
+ }
+
+ pnv_phb4_update_xsrc(phb);
+
+ phb->qirqs = qemu_allocate_irqs(xive_source_set_irq, xsrc, xsrc->nr_irqs);
+}
+
+static void pnv_phb4_reset(DeviceState *dev)
+{
+ PnvPHB4 *phb = PNV_PHB4(dev);
+ PCIDevice *root_dev = PCI_DEVICE(&phb->root);
+
+ /*
+ * Configure PCI device id at reset using a property.
+ */
+ pci_config_set_vendor_id(root_dev->config, PCI_VENDOR_ID_IBM);
+ pci_config_set_device_id(root_dev->config, phb->device_id);
+}
+
+static const char *pnv_phb4_root_bus_path(PCIHostState *host_bridge,
+ PCIBus *rootbus)
+{
+ PnvPHB4 *phb = PNV_PHB4(host_bridge);
+
+ snprintf(phb->bus_path, sizeof(phb->bus_path), "00%02x:%02x",
+ phb->chip_id, phb->phb_id);
+ return phb->bus_path;
+}
+
+static void pnv_phb4_xive_notify(XiveNotifier *xf, uint32_t srcno)
+{
+ PnvPHB4 *phb = PNV_PHB4(xf);
+ uint64_t notif_port = phb->regs[PHB_INT_NOTIFY_ADDR >> 3];
+ uint32_t offset = phb->regs[PHB_INT_NOTIFY_INDEX >> 3];
+ uint64_t data = XIVE_TRIGGER_PQ | offset | srcno;
+ MemTxResult result;
+
+ address_space_stq_be(&address_space_memory, notif_port, data,
+ MEMTXATTRS_UNSPECIFIED, &result);
+ if (result != MEMTX_OK) {
+ phb_error(phb, "trigger failed @%"HWADDR_PRIx "\n", notif_port);
+ return;
+ }
+}
+
+static Property pnv_phb4_properties[] = {
+ DEFINE_PROP_UINT32("index", PnvPHB4, phb_id, 0),
+ DEFINE_PROP_UINT32("chip-id", PnvPHB4, chip_id, 0),
+ DEFINE_PROP_UINT64("version", PnvPHB4, version, 0),
+ DEFINE_PROP_UINT16("device-id", PnvPHB4, device_id, 0),
+ DEFINE_PROP_LINK("stack", PnvPHB4, stack, TYPE_PNV_PHB4_PEC_STACK,
+ PnvPhb4PecStack *),
+ DEFINE_PROP_END_OF_LIST(),
+};
+
+static void pnv_phb4_class_init(ObjectClass *klass, void *data)
+{
+ PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass);
+ DeviceClass *dc = DEVICE_CLASS(klass);
+ XiveNotifierClass *xfc = XIVE_NOTIFIER_CLASS(klass);
+
+ hc->root_bus_path = pnv_phb4_root_bus_path;
+ dc->realize = pnv_phb4_realize;
+ device_class_set_props(dc, pnv_phb4_properties);
+ set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+ dc->user_creatable = false;
+ dc->reset = pnv_phb4_reset;
+
+ xfc->notify = pnv_phb4_xive_notify;
+}
+
+static const TypeInfo pnv_phb4_type_info = {
+ .name = TYPE_PNV_PHB4,
+ .parent = TYPE_PCIE_HOST_BRIDGE,
+ .instance_init = pnv_phb4_instance_init,
+ .instance_size = sizeof(PnvPHB4),
+ .class_init = pnv_phb4_class_init,
+ .interfaces = (InterfaceInfo[]) {
+ { TYPE_XIVE_NOTIFIER },
+ { },
+ }
+};
+
+static void pnv_phb4_root_bus_class_init(ObjectClass *klass, void *data)
+{
+ BusClass *k = BUS_CLASS(klass);
+
+ /*
+ * PHB4 has only a single root complex. Enforce the limit on the
+ * parent bus
+ */
+ k->max_dev = 1;
+}
+
+static const TypeInfo pnv_phb4_root_bus_info = {
+ .name = TYPE_PNV_PHB4_ROOT_BUS,
+ .parent = TYPE_PCIE_BUS,
+ .class_init = pnv_phb4_root_bus_class_init,
+ .interfaces = (InterfaceInfo[]) {
+ { INTERFACE_PCIE_DEVICE },
+ { }
+ },
+};
+
+static void pnv_phb4_root_port_reset(DeviceState *dev)
+{
+ PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
+ PCIDevice *d = PCI_DEVICE(dev);
+ uint8_t *conf = d->config;
+
+ rpc->parent_reset(dev);
+
+ pci_byte_test_and_set_mask(conf + PCI_IO_BASE,
+ PCI_IO_RANGE_MASK & 0xff);
+ pci_byte_test_and_clear_mask(conf + PCI_IO_LIMIT,
+ PCI_IO_RANGE_MASK & 0xff);
+ pci_set_word(conf + PCI_MEMORY_BASE, 0);
+ pci_set_word(conf + PCI_MEMORY_LIMIT, 0xfff0);
+ pci_set_word(conf + PCI_PREF_MEMORY_BASE, 0x1);
+ pci_set_word(conf + PCI_PREF_MEMORY_LIMIT, 0xfff1);
+ pci_set_long(conf + PCI_PREF_BASE_UPPER32, 0x1); /* Hack */
+ pci_set_long(conf + PCI_PREF_LIMIT_UPPER32, 0xffffffff);
+}
+
+static void pnv_phb4_root_port_realize(DeviceState *dev, Error **errp)
+{
+ PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
+ Error *local_err = NULL;
+
+ rpc->parent_realize(dev, &local_err);
+ if (local_err) {
+ error_propagate(errp, local_err);
+ return;
+ }
+}
+
+static void pnv_phb4_root_port_class_init(ObjectClass *klass, void *data)
+{
+ DeviceClass *dc = DEVICE_CLASS(klass);
+ PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
+ PCIERootPortClass *rpc = PCIE_ROOT_PORT_CLASS(klass);
+
+ dc->desc = "IBM PHB4 PCIE Root Port";
+ dc->user_creatable = false;
+
+ device_class_set_parent_realize(dc, pnv_phb4_root_port_realize,
+ &rpc->parent_realize);
+ device_class_set_parent_reset(dc, pnv_phb4_root_port_reset,
+ &rpc->parent_reset);
+
+ k->vendor_id = PCI_VENDOR_ID_IBM;
+ k->device_id = PNV_PHB4_DEVICE_ID;
+ k->revision = 0;
+
+ rpc->exp_offset = 0x48;
+ rpc->aer_offset = 0x100;
+
+ dc->reset = &pnv_phb4_root_port_reset;
+}
+
+static const TypeInfo pnv_phb4_root_port_info = {
+ .name = TYPE_PNV_PHB4_ROOT_PORT,
+ .parent = TYPE_PCIE_ROOT_PORT,
+ .instance_size = sizeof(PnvPHB4RootPort),
+ .class_init = pnv_phb4_root_port_class_init,
+};
+
+static void pnv_phb4_register_types(void)
+{
+ type_register_static(&pnv_phb4_root_bus_info);
+ type_register_static(&pnv_phb4_root_port_info);
+ type_register_static(&pnv_phb4_type_info);
+ type_register_static(&pnv_phb4_iommu_memory_region_info);
+}
+
+type_init(pnv_phb4_register_types);
+
+void pnv_phb4_update_regions(PnvPhb4PecStack *stack)
+{
+ PnvPHB4 *phb = &stack->phb;
+
+ /* Unmap first always */
+ if (memory_region_is_mapped(&phb->mr_regs)) {
+ memory_region_del_subregion(&stack->phbbar, &phb->mr_regs);
+ }
+ if (memory_region_is_mapped(&phb->xsrc.esb_mmio)) {
+ memory_region_del_subregion(&stack->intbar, &phb->xsrc.esb_mmio);
+ }
+
+ /* Map registers if enabled */
+ if (memory_region_is_mapped(&stack->phbbar)) {
+ memory_region_add_subregion(&stack->phbbar, 0, &phb->mr_regs);
+ }
+
+ /* Map ESB if enabled */
+ if (memory_region_is_mapped(&stack->intbar)) {
+ memory_region_add_subregion(&stack->intbar, 0, &phb->xsrc.esb_mmio);
+ }
+
+ /* Check/update m32 */
+ pnv_phb4_check_all_mbt(phb);
+}
+
+void pnv_phb4_pic_print_info(PnvPHB4 *phb, Monitor *mon)
+{
+ uint32_t offset = phb->regs[PHB_INT_NOTIFY_INDEX >> 3];
+
+ monitor_printf(mon, "PHB4[%x:%x] Source %08x .. %08x\n",
+ phb->chip_id, phb->phb_id,
+ offset, offset + phb->xsrc.nr_irqs - 1);
+ xive_source_pic_print_info(&phb->xsrc, 0, mon);
+}
diff --git a/hw/pci-host/pnv_phb4_pec.c b/hw/pci-host/pnv_phb4_pec.c
new file mode 100644
index 0000000000..68e1db3eac
--- /dev/null
+++ b/hw/pci-host/pnv_phb4_pec.c
@@ -0,0 +1,595 @@
+/*
+ * QEMU PowerPC PowerNV (POWER9) PHB4 model
+ *
+ * Copyright (c) 2018-2020, IBM Corporation.
+ *
+ * This code is licensed under the GPL version 2 or later. See the
+ * COPYING file in the top-level directory.
+ */
+#include "qemu/osdep.h"
+#include "qapi/error.h"
+#include "qemu-common.h"
+#include "qemu/log.h"
+#include "target/ppc/cpu.h"
+#include "hw/ppc/fdt.h"
+#include "hw/pci-host/pnv_phb4_regs.h"
+#include "hw/pci-host/pnv_phb4.h"
+#include "hw/ppc/pnv_xscom.h"
+#include "hw/pci/pci_bridge.h"
+#include "hw/pci/pci_bus.h"
+#include "hw/ppc/pnv.h"
+#include "hw/qdev-properties.h"
+
+#include <libfdt.h>
+
+#define phb_pec_error(pec, fmt, ...) \
+ qemu_log_mask(LOG_GUEST_ERROR, "phb4_pec[%d:%d]: " fmt "\n", \
+ (pec)->chip_id, (pec)->index, ## __VA_ARGS__)
+
+
+static uint64_t pnv_pec_nest_xscom_read(void *opaque, hwaddr addr,
+ unsigned size)
+{
+ PnvPhb4PecState *pec = PNV_PHB4_PEC(opaque);
+ uint32_t reg = addr >> 3;
+
+ /* TODO: add list of allowed registers and error out if not */
+ return pec->nest_regs[reg];
+}
+
+static void pnv_pec_nest_xscom_write(void *opaque, hwaddr addr,
+ uint64_t val, unsigned size)
+{
+ PnvPhb4PecState *pec = PNV_PHB4_PEC(opaque);
+ uint32_t reg = addr >> 3;
+
+ switch (reg) {
+ case PEC_NEST_PBCQ_HW_CONFIG:
+ case PEC_NEST_DROP_PRIO_CTRL:
+ case PEC_NEST_PBCQ_ERR_INJECT:
+ case PEC_NEST_PCI_NEST_CLK_TRACE_CTL:
+ case PEC_NEST_PBCQ_PMON_CTRL:
+ case PEC_NEST_PBCQ_PBUS_ADDR_EXT:
+ case PEC_NEST_PBCQ_PRED_VEC_TIMEOUT:
+ case PEC_NEST_CAPP_CTRL:
+ case PEC_NEST_PBCQ_READ_STK_OVR:
+ case PEC_NEST_PBCQ_WRITE_STK_OVR:
+ case PEC_NEST_PBCQ_STORE_STK_OVR:
+ case PEC_NEST_PBCQ_RETRY_BKOFF_CTRL:
+ pec->nest_regs[reg] = val;
+ break;
+ default:
+ phb_pec_error(pec, "%s @0x%"HWADDR_PRIx"=%"PRIx64"\n", __func__,
+ addr, val);
+ }
+}
+
+static const MemoryRegionOps pnv_pec_nest_xscom_ops = {
+ .read = pnv_pec_nest_xscom_read,
+ .write = pnv_pec_nest_xscom_write,
+ .valid.min_access_size = 8,
+ .valid.max_access_size = 8,
+ .impl.min_access_size = 8,
+ .impl.max_access_size = 8,
+ .endianness = DEVICE_BIG_ENDIAN,
+};
+
+static uint64_t pnv_pec_pci_xscom_read(void *opaque, hwaddr addr,
+ unsigned size)
+{
+ PnvPhb4PecState *pec = PNV_PHB4_PEC(opaque);
+ uint32_t reg = addr >> 3;
+
+ /* TODO: add list of allowed registers and error out if not */
+ return pec->pci_regs[reg];
+}
+
+static void pnv_pec_pci_xscom_write(void *opaque, hwaddr addr,
+ uint64_t val, unsigned size)
+{
+ PnvPhb4PecState *pec = PNV_PHB4_PEC(opaque);
+ uint32_t reg = addr >> 3;
+
+ switch (reg) {
+ case PEC_PCI_PBAIB_HW_CONFIG:
+ case PEC_PCI_PBAIB_READ_STK_OVR:
+ pec->pci_regs[reg] = val;
+ break;
+ default:
+ phb_pec_error(pec, "%s @0x%"HWADDR_PRIx"=%"PRIx64"\n", __func__,
+ addr, val);
+ }
+}
+
+static const MemoryRegionOps pnv_pec_pci_xscom_ops = {
+ .read = pnv_pec_pci_xscom_read,
+ .write = pnv_pec_pci_xscom_write,
+ .valid.min_access_size = 8,
+ .valid.max_access_size = 8,
+ .impl.min_access_size = 8,
+ .impl.max_access_size = 8,
+ .endianness = DEVICE_BIG_ENDIAN,
+};
+
+static uint64_t pnv_pec_stk_nest_xscom_read(void *opaque, hwaddr addr,
+ unsigned size)
+{
+ PnvPhb4PecStack *stack = PNV_PHB4_PEC_STACK(opaque);
+ uint32_t reg = addr >> 3;
+
+ /* TODO: add list of allowed registers and error out if not */
+ return stack->nest_regs[reg];
+}
+
+static void pnv_pec_stk_update_map(PnvPhb4PecStack *stack)
+{
+ PnvPhb4PecState *pec = stack->pec;
+ MemoryRegion *sysmem = pec->system_memory;
+ uint64_t bar_en = stack->nest_regs[PEC_NEST_STK_BAR_EN];
+ uint64_t bar, mask, size;
+ char name[64];
+
+ /*
+ * NOTE: This will really not work well if those are remapped
+ * after the PHB has created its sub regions. We could do better
+ * if we had a way to resize regions but we don't really care
+ * that much in practice as the stuff below really only happens
+ * once early during boot
+ */
+
+ /* Handle unmaps */
+ if (memory_region_is_mapped(&stack->mmbar0) &&
+ !(bar_en & PEC_NEST_STK_BAR_EN_MMIO0)) {
+ memory_region_del_subregion(sysmem, &stack->mmbar0);
+ }
+ if (memory_region_is_mapped(&stack->mmbar1) &&
+ !(bar_en & PEC_NEST_STK_BAR_EN_MMIO1)) {
+ memory_region_del_subregion(sysmem, &stack->mmbar1);
+ }
+ if (memory_region_is_mapped(&stack->phbbar) &&
+ !(bar_en & PEC_NEST_STK_BAR_EN_PHB)) {
+ memory_region_del_subregion(sysmem, &stack->phbbar);
+ }
+ if (memory_region_is_mapped(&stack->intbar) &&
+ !(bar_en & PEC_NEST_STK_BAR_EN_INT)) {
+ memory_region_del_subregion(sysmem, &stack->intbar);
+ }
+
+ /* Update PHB */
+ pnv_phb4_update_regions(stack);
+
+ /* Handle maps */
+ if (!memory_region_is_mapped(&stack->mmbar0) &&
+ (bar_en & PEC_NEST_STK_BAR_EN_MMIO0)) {
+ bar = stack->nest_regs[PEC_NEST_STK_MMIO_BAR0] >> 8;
+ mask = stack->nest_regs[PEC_NEST_STK_MMIO_BAR0_MASK];
+ size = ((~mask) >> 8) + 1;
+ snprintf(name, sizeof(name), "pec-%d.%d-stack-%d-mmio0",
+ pec->chip_id, pec->index, stack->stack_no);
+ memory_region_init(&stack->mmbar0, OBJECT(stack), name, size);
+ memory_region_add_subregion(sysmem, bar, &stack->mmbar0);
+ stack->mmio0_base = bar;
+ stack->mmio0_size = size;
+ }
+ if (!memory_region_is_mapped(&stack->mmbar1) &&
+ (bar_en & PEC_NEST_STK_BAR_EN_MMIO1)) {
+ bar = stack->nest_regs[PEC_NEST_STK_MMIO_BAR1] >> 8;
+ mask = stack->nest_regs[PEC_NEST_STK_MMIO_BAR1_MASK];
+ size = ((~mask) >> 8) + 1;
+ snprintf(name, sizeof(name), "pec-%d.%d-stack-%d-mmio1",
+ pec->chip_id, pec->index, stack->stack_no);
+ memory_region_init(&stack->mmbar1, OBJECT(stack), name, size);
+ memory_region_add_subregion(sysmem, bar, &stack->mmbar1);
+ stack->mmio1_base = bar;
+ stack->mmio1_size = size;
+ }
+ if (!memory_region_is_mapped(&stack->phbbar) &&
+ (bar_en & PEC_NEST_STK_BAR_EN_PHB)) {
+ bar = stack->nest_regs[PEC_NEST_STK_PHB_REGS_BAR] >> 8;
+ size = PNV_PHB4_NUM_REGS << 3;
+ snprintf(name, sizeof(name), "pec-%d.%d-stack-%d-phb",
+ pec->chip_id, pec->index, stack->stack_no);
+ memory_region_init(&stack->phbbar, OBJECT(stack), name, size);
+ memory_region_add_subregion(sysmem, bar, &stack->phbbar);
+ }
+ if (!memory_region_is_mapped(&stack->intbar) &&
+ (bar_en & PEC_NEST_STK_BAR_EN_INT)) {
+ bar = stack->nest_regs[PEC_NEST_STK_INT_BAR] >> 8;
+ size = PNV_PHB4_MAX_INTs << 16;
+ snprintf(name, sizeof(name), "pec-%d.%d-stack-%d-int",
+ stack->pec->chip_id, stack->pec->index, stack->stack_no);
+ memory_region_init(&stack->intbar, OBJECT(stack), name, size);
+ memory_region_add_subregion(sysmem, bar, &stack->intbar);
+ }
+
+ /* Update PHB */
+ pnv_phb4_update_regions(stack);
+}
+
+static void pnv_pec_stk_nest_xscom_write(void *opaque, hwaddr addr,
+ uint64_t val, unsigned size)
+{
+ PnvPhb4PecStack *stack = PNV_PHB4_PEC_STACK(opaque);
+ PnvPhb4PecState *pec = stack->pec;
+ uint32_t reg = addr >> 3;
+
+ switch (reg) {
+ case PEC_NEST_STK_PCI_NEST_FIR:
+ stack->nest_regs[PEC_NEST_STK_PCI_NEST_FIR] = val;
+ break;
+ case PEC_NEST_STK_PCI_NEST_FIR_CLR:
+ stack->nest_regs[PEC_NEST_STK_PCI_NEST_FIR] &= val;
+ break;
+ case PEC_NEST_STK_PCI_NEST_FIR_SET:
+ stack->nest_regs[PEC_NEST_STK_PCI_NEST_FIR] |= val;
+ break;
+ case PEC_NEST_STK_PCI_NEST_FIR_MSK:
+ stack->nest_regs[PEC_NEST_STK_PCI_NEST_FIR_MSK] = val;
+ break;
+ case PEC_NEST_STK_PCI_NEST_FIR_MSKC:
+ stack->nest_regs[PEC_NEST_STK_PCI_NEST_FIR_MSK] &= val;
+ break;
+ case PEC_NEST_STK_PCI_NEST_FIR_MSKS:
+ stack->nest_regs[PEC_NEST_STK_PCI_NEST_FIR_MSK] |= val;
+ break;
+ case PEC_NEST_STK_PCI_NEST_FIR_ACT0:
+ case PEC_NEST_STK_PCI_NEST_FIR_ACT1:
+ stack->nest_regs[reg] = val;
+ break;
+ case PEC_NEST_STK_PCI_NEST_FIR_WOF:
+ stack->nest_regs[reg] = 0;
+ break;
+ case PEC_NEST_STK_ERR_REPORT_0:
+ case PEC_NEST_STK_ERR_REPORT_1:
+ case PEC_NEST_STK_PBCQ_GNRL_STATUS:
+ /* Flag error ? */
+ break;
+ case PEC_NEST_STK_PBCQ_MODE:
+ stack->nest_regs[reg] = val & 0xff00000000000000ull;
+ break;
+ case PEC_NEST_STK_MMIO_BAR0:
+ case PEC_NEST_STK_MMIO_BAR0_MASK:
+ case PEC_NEST_STK_MMIO_BAR1:
+ case PEC_NEST_STK_MMIO_BAR1_MASK:
+ if (stack->nest_regs[PEC_NEST_STK_BAR_EN] &
+ (PEC_NEST_STK_BAR_EN_MMIO0 |
+ PEC_NEST_STK_BAR_EN_MMIO1)) {
+ phb_pec_error(pec, "Changing enabled BAR unsupported\n");
+ }
+ stack->nest_regs[reg] = val & 0xffffffffff000000ull;
+ break;
+ case PEC_NEST_STK_PHB_REGS_BAR:
+ if (stack->nest_regs[PEC_NEST_STK_BAR_EN] & PEC_NEST_STK_BAR_EN_PHB) {
+ phb_pec_error(pec, "Changing enabled BAR unsupported\n");
+ }
+ stack->nest_regs[reg] = val & 0xffffffffffc00000ull;
+ break;
+ case PEC_NEST_STK_INT_BAR:
+ if (stack->nest_regs[PEC_NEST_STK_BAR_EN] & PEC_NEST_STK_BAR_EN_INT) {
+ phb_pec_error(pec, "Changing enabled BAR unsupported\n");
+ }
+ stack->nest_regs[reg] = val & 0xfffffff000000000ull;
+ break;
+ case PEC_NEST_STK_BAR_EN:
+ stack->nest_regs[reg] = val & 0xf000000000000000ull;
+ pnv_pec_stk_update_map(stack);
+ break;
+ case PEC_NEST_STK_DATA_FRZ_TYPE:
+ case PEC_NEST_STK_PBCQ_TUN_BAR:
+ /* Not used for now */
+ stack->nest_regs[reg] = val;
+ break;
+ default:
+ qemu_log_mask(LOG_UNIMP, "phb4_pec: nest_xscom_write 0x%"HWADDR_PRIx
+ "=%"PRIx64"\n", addr, val);
+ }
+}
+
+static const MemoryRegionOps pnv_pec_stk_nest_xscom_ops = {
+ .read = pnv_pec_stk_nest_xscom_read,
+ .write = pnv_pec_stk_nest_xscom_write,
+ .valid.min_access_size = 8,
+ .valid.max_access_size = 8,
+ .impl.min_access_size = 8,
+ .impl.max_access_size = 8,
+ .endianness = DEVICE_BIG_ENDIAN,
+};
+
+static uint64_t pnv_pec_stk_pci_xscom_read(void *opaque, hwaddr addr,
+ unsigned size)
+{
+ PnvPhb4PecStack *stack = PNV_PHB4_PEC_STACK(opaque);
+ uint32_t reg = addr >> 3;
+
+ /* TODO: add list of allowed registers and error out if not */
+ return stack->pci_regs[reg];
+}
+
+static void pnv_pec_stk_pci_xscom_write(void *opaque, hwaddr addr,
+ uint64_t val, unsigned size)
+{
+ PnvPhb4PecStack *stack = PNV_PHB4_PEC_STACK(opaque);
+ uint32_t reg = addr >> 3;
+
+ switch (reg) {
+ case PEC_PCI_STK_PCI_FIR:
+ stack->nest_regs[reg] = val;
+ break;
+ case PEC_PCI_STK_PCI_FIR_CLR:
+ stack->nest_regs[PEC_PCI_STK_PCI_FIR] &= val;
+ break;
+ case PEC_PCI_STK_PCI_FIR_SET:
+ stack->nest_regs[PEC_PCI_STK_PCI_FIR] |= val;
+ break;
+ case PEC_PCI_STK_PCI_FIR_MSK:
+ stack->nest_regs[reg] = val;
+ break;
+ case PEC_PCI_STK_PCI_FIR_MSKC:
+ stack->nest_regs[PEC_PCI_STK_PCI_FIR_MSK] &= val;
+ break;
+ case PEC_PCI_STK_PCI_FIR_MSKS:
+ stack->nest_regs[PEC_PCI_STK_PCI_FIR_MSK] |= val;
+ break;
+ case PEC_PCI_STK_PCI_FIR_ACT0:
+ case PEC_PCI_STK_PCI_FIR_ACT1:
+ stack->nest_regs[reg] = val;
+ break;
+ case PEC_PCI_STK_PCI_FIR_WOF:
+ stack->nest_regs[reg] = 0;
+ break;
+ case PEC_PCI_STK_ETU_RESET:
+ stack->nest_regs[reg] = val & 0x8000000000000000ull;
+ /* TODO: Implement reset */
+ break;
+ case PEC_PCI_STK_PBAIB_ERR_REPORT:
+ break;
+ case PEC_PCI_STK_PBAIB_TX_CMD_CRED:
+ case PEC_PCI_STK_PBAIB_TX_DAT_CRED:
+ stack->nest_regs[reg] = val;
+ break;
+ default:
+ qemu_log_mask(LOG_UNIMP, "phb4_pec_stk: pci_xscom_write 0x%"HWADDR_PRIx
+ "=%"PRIx64"\n", addr, val);
+ }
+}
+
+static const MemoryRegionOps pnv_pec_stk_pci_xscom_ops = {
+ .read = pnv_pec_stk_pci_xscom_read,
+ .write = pnv_pec_stk_pci_xscom_write,
+ .valid.min_access_size = 8,
+ .valid.max_access_size = 8,
+ .impl.min_access_size = 8,
+ .impl.max_access_size = 8,
+ .endianness = DEVICE_BIG_ENDIAN,
+};
+
+static void pnv_pec_instance_init(Object *obj)
+{
+ PnvPhb4PecState *pec = PNV_PHB4_PEC(obj);
+ int i;
+
+ for (i = 0; i < PHB4_PEC_MAX_STACKS; i++) {
+ object_initialize_child(obj, "stack[*]", &pec->stacks[i],
+ sizeof(pec->stacks[i]), TYPE_PNV_PHB4_PEC_STACK,
+ &error_abort, NULL);
+ }
+}
+
+static void pnv_pec_realize(DeviceState *dev, Error **errp)
+{
+ PnvPhb4PecState *pec = PNV_PHB4_PEC(dev);
+ Error *local_err = NULL;
+ char name[64];
+ int i;
+
+ assert(pec->system_memory);
+
+ /* Create stacks */
+ for (i = 0; i < pec->num_stacks; i++) {
+ PnvPhb4PecStack *stack = &pec->stacks[i];
+ Object *stk_obj = OBJECT(stack);
+
+ object_property_set_int(stk_obj, i, "stack-no", &error_abort);
+ object_property_set_link(stk_obj, OBJECT(pec), "pec", &error_abort);
+ object_property_set_bool(stk_obj, true, "realized", errp);
+ if (local_err) {
+ error_propagate(errp, local_err);
+ return;
+ }
+ }
+
+ /* Initialize the XSCOM regions for the PEC registers */
+ snprintf(name, sizeof(name), "xscom-pec-%d.%d-nest", pec->chip_id,
+ pec->index);
+ pnv_xscom_region_init(&pec->nest_regs_mr, OBJECT(dev),
+ &pnv_pec_nest_xscom_ops, pec, name,
+ PHB4_PEC_NEST_REGS_COUNT);
+
+ snprintf(name, sizeof(name), "xscom-pec-%d.%d-pci", pec->chip_id,
+ pec->index);
+ pnv_xscom_region_init(&pec->pci_regs_mr, OBJECT(dev),
+ &pnv_pec_pci_xscom_ops, pec, name,
+ PHB4_PEC_PCI_REGS_COUNT);
+}
+
+static int pnv_pec_dt_xscom(PnvXScomInterface *dev, void *fdt,
+ int xscom_offset)
+{
+ PnvPhb4PecState *pec = PNV_PHB4_PEC(dev);
+ PnvPhb4PecClass *pecc = PNV_PHB4_PEC_GET_CLASS(dev);
+ uint32_t nbase = pecc->xscom_nest_base(pec);
+ uint32_t pbase = pecc->xscom_pci_base(pec);
+ int offset, i;
+ char *name;
+ uint32_t reg[] = {
+ cpu_to_be32(nbase),
+ cpu_to_be32(pecc->xscom_nest_size),
+ cpu_to_be32(pbase),
+ cpu_to_be32(pecc->xscom_pci_size),
+ };
+
+ name = g_strdup_printf("pbcq@%x", nbase);
+ offset = fdt_add_subnode(fdt, xscom_offset, name);
+ _FDT(offset);
+ g_free(name);
+
+ _FDT((fdt_setprop(fdt, offset, "reg", reg, sizeof(reg))));
+
+ _FDT((fdt_setprop_cell(fdt, offset, "ibm,pec-index", pec->index)));
+ _FDT((fdt_setprop_cell(fdt, offset, "#address-cells", 1)));
+ _FDT((fdt_setprop_cell(fdt, offset, "#size-cells", 0)));
+ _FDT((fdt_setprop(fdt, offset, "compatible", pecc->compat,
+ pecc->compat_size)));
+
+ for (i = 0; i < pec->num_stacks; i++) {
+ PnvPhb4PecStack *stack = &pec->stacks[i];
+ PnvPHB4 *phb = &stack->phb;
+ int stk_offset;
+
+ name = g_strdup_printf("stack@%x", i);
+ stk_offset = fdt_add_subnode(fdt, offset, name);
+ _FDT(stk_offset);
+ g_free(name);
+ _FDT((fdt_setprop(fdt, stk_offset, "compatible", pecc->stk_compat,
+ pecc->stk_compat_size)));
+ _FDT((fdt_setprop_cell(fdt, stk_offset, "reg", i)));
+ _FDT((fdt_setprop_cell(fdt, stk_offset, "ibm,phb-index", phb->phb_id)));
+ }
+
+ return 0;
+}
+
+static Property pnv_pec_properties[] = {
+ DEFINE_PROP_UINT32("index", PnvPhb4PecState, index, 0),
+ DEFINE_PROP_UINT32("num-stacks", PnvPhb4PecState, num_stacks, 0),
+ DEFINE_PROP_UINT32("chip-id", PnvPhb4PecState, chip_id, 0),
+ DEFINE_PROP_LINK("system-memory", PnvPhb4PecState, system_memory,
+ TYPE_MEMORY_REGION, MemoryRegion *),
+ DEFINE_PROP_END_OF_LIST(),
+};
+
+static uint32_t pnv_pec_xscom_pci_base(PnvPhb4PecState *pec)
+{
+ return PNV9_XSCOM_PEC_PCI_BASE + 0x1000000 * pec->index;
+}
+
+static uint32_t pnv_pec_xscom_nest_base(PnvPhb4PecState *pec)
+{
+ return PNV9_XSCOM_PEC_NEST_BASE + 0x400 * pec->index;
+}
+
+static void pnv_pec_class_init(ObjectClass *klass, void *data)
+{
+ DeviceClass *dc = DEVICE_CLASS(klass);
+ PnvXScomInterfaceClass *xdc = PNV_XSCOM_INTERFACE_CLASS(klass);
+ PnvPhb4PecClass *pecc = PNV_PHB4_PEC_CLASS(klass);
+ static const char compat[] = "ibm,power9-pbcq";
+ static const char stk_compat[] = "ibm,power9-phb-stack";
+
+ xdc->dt_xscom = pnv_pec_dt_xscom;
+
+ dc->realize = pnv_pec_realize;
+ device_class_set_props(dc, pnv_pec_properties);
+ dc->user_creatable = false;
+
+ pecc->xscom_nest_base = pnv_pec_xscom_nest_base;
+ pecc->xscom_pci_base = pnv_pec_xscom_pci_base;
+ pecc->xscom_nest_size = PNV9_XSCOM_PEC_NEST_SIZE;
+ pecc->xscom_pci_size = PNV9_XSCOM_PEC_PCI_SIZE;
+ pecc->compat = compat;
+ pecc->compat_size = sizeof(compat);
+ pecc->stk_compat = stk_compat;
+ pecc->stk_compat_size = sizeof(stk_compat);
+}
+
+static const TypeInfo pnv_pec_type_info = {
+ .name = TYPE_PNV_PHB4_PEC,
+ .parent = TYPE_DEVICE,
+ .instance_size = sizeof(PnvPhb4PecState),
+ .instance_init = pnv_pec_instance_init,
+ .class_init = pnv_pec_class_init,
+ .class_size = sizeof(PnvPhb4PecClass),
+ .interfaces = (InterfaceInfo[]) {
+ { TYPE_PNV_XSCOM_INTERFACE },
+ { }
+ }
+};
+
+static void pnv_pec_stk_instance_init(Object *obj)
+{
+ PnvPhb4PecStack *stack = PNV_PHB4_PEC_STACK(obj);
+
+ object_initialize_child(obj, "phb", &stack->phb, sizeof(stack->phb),
+ TYPE_PNV_PHB4, &error_abort, NULL);
+}
+
+static void pnv_pec_stk_realize(DeviceState *dev, Error **errp)
+{
+ PnvPhb4PecStack *stack = PNV_PHB4_PEC_STACK(dev);
+ PnvPhb4PecState *pec = stack->pec;
+ char name[64];
+
+ assert(pec);
+
+ /* Initialize the XSCOM regions for the stack registers */
+ snprintf(name, sizeof(name), "xscom-pec-%d.%d-nest-stack-%d",
+ pec->chip_id, pec->index, stack->stack_no);
+ pnv_xscom_region_init(&stack->nest_regs_mr, OBJECT(stack),
+ &pnv_pec_stk_nest_xscom_ops, stack, name,
+ PHB4_PEC_NEST_STK_REGS_COUNT);
+
+ snprintf(name, sizeof(name), "xscom-pec-%d.%d-pci-stack-%d",
+ pec->chip_id, pec->index, stack->stack_no);
+ pnv_xscom_region_init(&stack->pci_regs_mr, OBJECT(stack),
+ &pnv_pec_stk_pci_xscom_ops, stack, name,
+ PHB4_PEC_PCI_STK_REGS_COUNT);
+
+ /* PHB pass-through */
+ snprintf(name, sizeof(name), "xscom-pec-%d.%d-pci-stack-%d-phb",
+ pec->chip_id, pec->index, stack->stack_no);
+ pnv_xscom_region_init(&stack->phb_regs_mr, OBJECT(&stack->phb),
+ &pnv_phb4_xscom_ops, &stack->phb, name, 0x40);
+
+ /*
+ * Let the machine/chip realize the PHB object to customize more
+ * easily some fields
+ */
+}
+
+static Property pnv_pec_stk_properties[] = {
+ DEFINE_PROP_UINT32("stack-no", PnvPhb4PecStack, stack_no, 0),
+ DEFINE_PROP_LINK("pec", PnvPhb4PecStack, pec, TYPE_PNV_PHB4_PEC,
+ PnvPhb4PecState *),
+ DEFINE_PROP_END_OF_LIST(),
+};
+
+static void pnv_pec_stk_class_init(ObjectClass *klass, void *data)
+{
+ DeviceClass *dc = DEVICE_CLASS(klass);
+
+ device_class_set_props(dc, pnv_pec_stk_properties);
+ dc->realize = pnv_pec_stk_realize;
+ dc->user_creatable = false;
+
+ /* TODO: reset regs ? */
+}
+
+static const TypeInfo pnv_pec_stk_type_info = {
+ .name = TYPE_PNV_PHB4_PEC_STACK,
+ .parent = TYPE_DEVICE,
+ .instance_size = sizeof(PnvPhb4PecStack),
+ .instance_init = pnv_pec_stk_instance_init,
+ .class_init = pnv_pec_stk_class_init,
+ .interfaces = (InterfaceInfo[]) {
+ { TYPE_PNV_XSCOM_INTERFACE },
+ { }
+ }
+};
+
+static void pnv_pec_register_types(void)
+{
+ type_register_static(&pnv_pec_type_info);
+ type_register_static(&pnv_pec_stk_type_info);
+}
+
+type_init(pnv_pec_register_types);