153018216SPaolo Bonzini /* 253018216SPaolo Bonzini * QEMU PowerPC pSeries Logical Partition (aka sPAPR) hardware System Emulator 353018216SPaolo Bonzini * 453018216SPaolo Bonzini * Copyright (c) 2004-2007 Fabrice Bellard 553018216SPaolo Bonzini * Copyright (c) 2007 Jocelyn Mayer 653018216SPaolo Bonzini * Copyright (c) 2010 David Gibson, IBM Corporation. 753018216SPaolo Bonzini * 853018216SPaolo Bonzini * Permission is hereby granted, free of charge, to any person obtaining a copy 953018216SPaolo Bonzini * of this software and associated documentation files (the "Software"), to deal 1053018216SPaolo Bonzini * in the Software without restriction, including without limitation the rights 1153018216SPaolo Bonzini * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 1253018216SPaolo Bonzini * copies of the Software, and to permit persons to whom the Software is 1353018216SPaolo Bonzini * furnished to do so, subject to the following conditions: 1453018216SPaolo Bonzini * 1553018216SPaolo Bonzini * The above copyright notice and this permission notice shall be included in 1653018216SPaolo Bonzini * all copies or substantial portions of the Software. 1753018216SPaolo Bonzini * 1853018216SPaolo Bonzini * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 1953018216SPaolo Bonzini * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 2053018216SPaolo Bonzini * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 2153018216SPaolo Bonzini * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 2253018216SPaolo Bonzini * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 2353018216SPaolo Bonzini * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 2453018216SPaolo Bonzini * THE SOFTWARE. 2553018216SPaolo Bonzini * 2653018216SPaolo Bonzini */ 270d75590dSPeter Maydell #include "qemu/osdep.h" 28da34e65cSMarkus Armbruster #include "qapi/error.h" 2953018216SPaolo Bonzini #include "sysemu/sysemu.h" 30e35704baSEduardo Habkost #include "sysemu/numa.h" 3153018216SPaolo Bonzini #include "hw/hw.h" 3203dd024fSPaolo Bonzini #include "qemu/log.h" 3371461b0fSAlexey Kardashevskiy #include "hw/fw-path-provider.h" 3453018216SPaolo Bonzini #include "elf.h" 3553018216SPaolo Bonzini #include "net/net.h" 36ad440b4aSAndrew Jones #include "sysemu/device_tree.h" 37fa1d36dfSMarkus Armbruster #include "sysemu/block-backend.h" 3853018216SPaolo Bonzini #include "sysemu/cpus.h" 39b3946626SVincent Palatin #include "sysemu/hw_accel.h" 4053018216SPaolo Bonzini #include "kvm_ppc.h" 41ff14e817SDr. David Alan Gilbert #include "migration/migration.h" 424be21d56SDavid Gibson #include "mmu-hash64.h" 433794d548SAlexey Kardashevskiy #include "qom/cpu.h" 4453018216SPaolo Bonzini 4553018216SPaolo Bonzini #include "hw/boards.h" 460d09e41aSPaolo Bonzini #include "hw/ppc/ppc.h" 4753018216SPaolo Bonzini #include "hw/loader.h" 4853018216SPaolo Bonzini 497804c353SCédric Le Goater #include "hw/ppc/fdt.h" 500d09e41aSPaolo Bonzini #include "hw/ppc/spapr.h" 510d09e41aSPaolo Bonzini #include "hw/ppc/spapr_vio.h" 520d09e41aSPaolo Bonzini #include "hw/pci-host/spapr.h" 530d09e41aSPaolo Bonzini #include "hw/ppc/xics.h" 5453018216SPaolo Bonzini #include "hw/pci/msi.h" 5553018216SPaolo Bonzini 5653018216SPaolo Bonzini #include "hw/pci/pci.h" 5771461b0fSAlexey Kardashevskiy #include "hw/scsi/scsi.h" 5871461b0fSAlexey Kardashevskiy #include "hw/virtio/virtio-scsi.h" 5953018216SPaolo Bonzini 6053018216SPaolo Bonzini #include "exec/address-spaces.h" 6153018216SPaolo Bonzini #include "hw/usb.h" 6253018216SPaolo Bonzini #include "qemu/config-file.h" 63135a129aSAneesh Kumar K.V #include "qemu/error-report.h" 642a6593cbSAlexey Kardashevskiy #include "trace.h" 6534316482SAlexey Kardashevskiy #include "hw/nmi.h" 6653018216SPaolo Bonzini 6768a27b20SMichael S. Tsirkin #include "hw/compat.h" 68f348b6d1SVeronia Bahaa #include "qemu/cutils.h" 6994a94e4cSBharata B Rao #include "hw/ppc/spapr_cpu_core.h" 702474bfd4SIgor Mammedov #include "qmp-commands.h" 7168a27b20SMichael S. Tsirkin 7253018216SPaolo Bonzini #include <libfdt.h> 7353018216SPaolo Bonzini 7453018216SPaolo Bonzini /* SLOF memory layout: 7553018216SPaolo Bonzini * 7653018216SPaolo Bonzini * SLOF raw image loaded at 0, copies its romfs right below the flat 7753018216SPaolo Bonzini * device-tree, then position SLOF itself 31M below that 7853018216SPaolo Bonzini * 7953018216SPaolo Bonzini * So we set FW_OVERHEAD to 40MB which should account for all of that 8053018216SPaolo Bonzini * and more 8153018216SPaolo Bonzini * 8253018216SPaolo Bonzini * We load our kernel at 4M, leaving space for SLOF initial image 8353018216SPaolo Bonzini */ 8438b02bd8SAlexey Kardashevskiy #define FDT_MAX_SIZE 0x100000 8553018216SPaolo Bonzini #define RTAS_MAX_SIZE 0x10000 86b7d1f77aSBenjamin Herrenschmidt #define RTAS_MAX_ADDR 0x80000000 /* RTAS must stay below that */ 8753018216SPaolo Bonzini #define FW_MAX_SIZE 0x400000 8853018216SPaolo Bonzini #define FW_FILE_NAME "slof.bin" 8953018216SPaolo Bonzini #define FW_OVERHEAD 0x2800000 9053018216SPaolo Bonzini #define KERNEL_LOAD_ADDR FW_MAX_SIZE 9153018216SPaolo Bonzini 9253018216SPaolo Bonzini #define MIN_RMA_SLOF 128UL 9353018216SPaolo Bonzini 9453018216SPaolo Bonzini #define PHANDLE_XICP 0x00001111 9553018216SPaolo Bonzini 9653018216SPaolo Bonzini #define HTAB_SIZE(spapr) (1ULL << ((spapr)->htab_shift)) 9753018216SPaolo Bonzini 98*681bfadeSCédric Le Goater static XICSState *try_create_xics(sPAPRMachineState *spapr, 99*681bfadeSCédric Le Goater const char *type, const char *type_ics, 100817bb6a4SCédric Le Goater const char *type_icp, int nr_servers, 101817bb6a4SCédric Le Goater int nr_irqs, Error **errp) 102c04d6cfaSAnthony Liguori { 1034e4169f7SCédric Le Goater Error *err = NULL, *local_err = NULL; 1044e4169f7SCédric Le Goater XICSState *xics; 1054e4169f7SCédric Le Goater ICSState *ics = NULL; 106817bb6a4SCédric Le Goater int i; 107c04d6cfaSAnthony Liguori 1084e4169f7SCédric Le Goater xics = XICS_COMMON(object_new(type)); 1094e4169f7SCédric Le Goater qdev_set_parent_bus(DEVICE(xics), sysbus_get_default()); 110817bb6a4SCédric Le Goater object_property_set_bool(OBJECT(xics), true, "realized", &err); 11134f2af3dSMarkus Armbruster if (err) { 1124e4169f7SCédric Le Goater goto error; 113c04d6cfaSAnthony Liguori } 1144e4169f7SCédric Le Goater 1154e4169f7SCédric Le Goater ics = ICS_SIMPLE(object_new(type_ics)); 116*681bfadeSCédric Le Goater object_property_add_child(OBJECT(spapr), "ics", OBJECT(ics), NULL); 1174e4169f7SCédric Le Goater object_property_set_int(OBJECT(ics), nr_irqs, "nr-irqs", &err); 1184e4169f7SCédric Le Goater object_property_add_const_link(OBJECT(ics), "xics", OBJECT(xics), NULL); 1194e4169f7SCédric Le Goater object_property_set_bool(OBJECT(ics), true, "realized", &local_err); 1204e4169f7SCédric Le Goater error_propagate(&err, local_err); 1214e4169f7SCédric Le Goater if (err) { 1224e4169f7SCédric Le Goater goto error; 1234e4169f7SCédric Le Goater } 1244e4169f7SCédric Le Goater QLIST_INSERT_HEAD(&xics->ics, ics, list); 1254e4169f7SCédric Le Goater 126817bb6a4SCédric Le Goater xics->ss = g_malloc0(nr_servers * sizeof(ICPState)); 127817bb6a4SCédric Le Goater xics->nr_servers = nr_servers; 128817bb6a4SCédric Le Goater 129817bb6a4SCédric Le Goater for (i = 0; i < nr_servers; i++) { 130817bb6a4SCédric Le Goater ICPState *icp = &xics->ss[i]; 131817bb6a4SCédric Le Goater 132817bb6a4SCédric Le Goater object_initialize(icp, sizeof(*icp), type_icp); 133817bb6a4SCédric Le Goater object_property_add_child(OBJECT(xics), "icp[*]", OBJECT(icp), NULL); 134817bb6a4SCédric Le Goater object_property_add_const_link(OBJECT(icp), "xics", OBJECT(xics), NULL); 135817bb6a4SCédric Le Goater object_property_set_bool(OBJECT(icp), true, "realized", &err); 136817bb6a4SCédric Le Goater if (err) { 137817bb6a4SCédric Le Goater goto error; 138817bb6a4SCédric Le Goater } 139817bb6a4SCédric Le Goater object_unref(OBJECT(icp)); 140817bb6a4SCédric Le Goater } 141817bb6a4SCédric Le Goater 142*681bfadeSCédric Le Goater spapr->ics = ics; 1434e4169f7SCédric Le Goater return xics; 1444e4169f7SCédric Le Goater 1454e4169f7SCédric Le Goater error: 1464e4169f7SCédric Le Goater error_propagate(errp, err); 1474e4169f7SCédric Le Goater if (ics) { 1484e4169f7SCédric Le Goater object_unparent(OBJECT(ics)); 1494e4169f7SCédric Le Goater } 1504e4169f7SCédric Le Goater object_unparent(OBJECT(xics)); 1514e4169f7SCédric Le Goater return NULL; 152c04d6cfaSAnthony Liguori } 153c04d6cfaSAnthony Liguori 154446f16a6SMarcel Apfelbaum static XICSState *xics_system_init(MachineState *machine, 1551e49182dSDavid Gibson int nr_servers, int nr_irqs, Error **errp) 156c04d6cfaSAnthony Liguori { 15727f24582SBenjamin Herrenschmidt XICSState *xics = NULL; 158c04d6cfaSAnthony Liguori 15911ad93f6SDavid Gibson if (kvm_enabled()) { 16034f2af3dSMarkus Armbruster Error *err = NULL; 16134f2af3dSMarkus Armbruster 162446f16a6SMarcel Apfelbaum if (machine_kernel_irqchip_allowed(machine)) { 163*681bfadeSCédric Le Goater xics = try_create_xics(SPAPR_MACHINE(machine), 164*681bfadeSCédric Le Goater TYPE_XICS_SPAPR_KVM, TYPE_ICS_KVM, 165817bb6a4SCédric Le Goater TYPE_KVM_ICP, nr_servers, nr_irqs, &err); 16611ad93f6SDavid Gibson } 16727f24582SBenjamin Herrenschmidt if (machine_kernel_irqchip_required(machine) && !xics) { 168b83baa60SMarkus Armbruster error_reportf_err(err, 169b83baa60SMarkus Armbruster "kernel_irqchip requested but unavailable: "); 170b83baa60SMarkus Armbruster } else { 171903a41d3SStefano Dong (董兴水) error_free(err); 17211ad93f6SDavid Gibson } 173b83baa60SMarkus Armbruster } 17411ad93f6SDavid Gibson 17527f24582SBenjamin Herrenschmidt if (!xics) { 176*681bfadeSCédric Le Goater xics = try_create_xics(SPAPR_MACHINE(machine), 177*681bfadeSCédric Le Goater TYPE_XICS_SPAPR, TYPE_ICS_SIMPLE, 178*681bfadeSCédric Le Goater TYPE_ICP, nr_servers, nr_irqs, errp); 179c04d6cfaSAnthony Liguori } 180c04d6cfaSAnthony Liguori 18127f24582SBenjamin Herrenschmidt return xics; 182c04d6cfaSAnthony Liguori } 183c04d6cfaSAnthony Liguori 184833d4668SAlexey Kardashevskiy static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu, 185833d4668SAlexey Kardashevskiy int smt_threads) 186833d4668SAlexey Kardashevskiy { 187833d4668SAlexey Kardashevskiy int i, ret = 0; 188833d4668SAlexey Kardashevskiy uint32_t servers_prop[smt_threads]; 189833d4668SAlexey Kardashevskiy uint32_t gservers_prop[smt_threads * 2]; 190833d4668SAlexey Kardashevskiy int index = ppc_get_vcpu_dt_id(cpu); 191833d4668SAlexey Kardashevskiy 192d6e166c0SDavid Gibson if (cpu->compat_pvr) { 193d6e166c0SDavid Gibson ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->compat_pvr); 1946d9412eaSAlexey Kardashevskiy if (ret < 0) { 1956d9412eaSAlexey Kardashevskiy return ret; 1966d9412eaSAlexey Kardashevskiy } 1976d9412eaSAlexey Kardashevskiy } 1986d9412eaSAlexey Kardashevskiy 199833d4668SAlexey Kardashevskiy /* Build interrupt servers and gservers properties */ 200833d4668SAlexey Kardashevskiy for (i = 0; i < smt_threads; i++) { 201833d4668SAlexey Kardashevskiy servers_prop[i] = cpu_to_be32(index + i); 202833d4668SAlexey Kardashevskiy /* Hack, direct the group queues back to cpu 0 */ 203833d4668SAlexey Kardashevskiy gservers_prop[i*2] = cpu_to_be32(index + i); 204833d4668SAlexey Kardashevskiy gservers_prop[i*2 + 1] = 0; 205833d4668SAlexey Kardashevskiy } 206833d4668SAlexey Kardashevskiy ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s", 207833d4668SAlexey Kardashevskiy servers_prop, sizeof(servers_prop)); 208833d4668SAlexey Kardashevskiy if (ret < 0) { 209833d4668SAlexey Kardashevskiy return ret; 210833d4668SAlexey Kardashevskiy } 211833d4668SAlexey Kardashevskiy ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s", 212833d4668SAlexey Kardashevskiy gservers_prop, sizeof(gservers_prop)); 213833d4668SAlexey Kardashevskiy 214833d4668SAlexey Kardashevskiy return ret; 215833d4668SAlexey Kardashevskiy } 216833d4668SAlexey Kardashevskiy 2170da6f3feSBharata B Rao static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, CPUState *cs) 2180da6f3feSBharata B Rao { 2190da6f3feSBharata B Rao int ret = 0; 2200da6f3feSBharata B Rao PowerPCCPU *cpu = POWERPC_CPU(cs); 2210da6f3feSBharata B Rao int index = ppc_get_vcpu_dt_id(cpu); 2220da6f3feSBharata B Rao uint32_t associativity[] = {cpu_to_be32(0x5), 2230da6f3feSBharata B Rao cpu_to_be32(0x0), 2240da6f3feSBharata B Rao cpu_to_be32(0x0), 2250da6f3feSBharata B Rao cpu_to_be32(0x0), 2260da6f3feSBharata B Rao cpu_to_be32(cs->numa_node), 2270da6f3feSBharata B Rao cpu_to_be32(index)}; 2280da6f3feSBharata B Rao 2290da6f3feSBharata B Rao /* Advertise NUMA via ibm,associativity */ 2300da6f3feSBharata B Rao if (nb_numa_nodes > 1) { 2310da6f3feSBharata B Rao ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity, 2320da6f3feSBharata B Rao sizeof(associativity)); 2330da6f3feSBharata B Rao } 2340da6f3feSBharata B Rao 2350da6f3feSBharata B Rao return ret; 2360da6f3feSBharata B Rao } 2370da6f3feSBharata B Rao 23828e02042SDavid Gibson static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr) 23953018216SPaolo Bonzini { 24082677ed2SAlexey Kardashevskiy int ret = 0, offset, cpus_offset; 24182677ed2SAlexey Kardashevskiy CPUState *cs; 24253018216SPaolo Bonzini char cpu_model[32]; 24353018216SPaolo Bonzini int smt = kvmppc_smt_threads(); 24453018216SPaolo Bonzini uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)}; 24553018216SPaolo Bonzini 24682677ed2SAlexey Kardashevskiy CPU_FOREACH(cs) { 24782677ed2SAlexey Kardashevskiy PowerPCCPU *cpu = POWERPC_CPU(cs); 24882677ed2SAlexey Kardashevskiy DeviceClass *dc = DEVICE_GET_CLASS(cs); 24982677ed2SAlexey Kardashevskiy int index = ppc_get_vcpu_dt_id(cpu); 25012dbeb16SDavid Gibson int compat_smt = MIN(smp_threads, ppc_compat_max_threads(cpu)); 25153018216SPaolo Bonzini 2520f20ba62SAlexey Kardashevskiy if ((index % smt) != 0) { 25353018216SPaolo Bonzini continue; 25453018216SPaolo Bonzini } 25553018216SPaolo Bonzini 25682677ed2SAlexey Kardashevskiy snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index); 25753018216SPaolo Bonzini 25882677ed2SAlexey Kardashevskiy cpus_offset = fdt_path_offset(fdt, "/cpus"); 25982677ed2SAlexey Kardashevskiy if (cpus_offset < 0) { 26082677ed2SAlexey Kardashevskiy cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"), 26182677ed2SAlexey Kardashevskiy "cpus"); 26282677ed2SAlexey Kardashevskiy if (cpus_offset < 0) { 26382677ed2SAlexey Kardashevskiy return cpus_offset; 26482677ed2SAlexey Kardashevskiy } 26582677ed2SAlexey Kardashevskiy } 26682677ed2SAlexey Kardashevskiy offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model); 26782677ed2SAlexey Kardashevskiy if (offset < 0) { 26882677ed2SAlexey Kardashevskiy offset = fdt_add_subnode(fdt, cpus_offset, cpu_model); 26953018216SPaolo Bonzini if (offset < 0) { 27053018216SPaolo Bonzini return offset; 27153018216SPaolo Bonzini } 27282677ed2SAlexey Kardashevskiy } 27353018216SPaolo Bonzini 2740da6f3feSBharata B Rao ret = fdt_setprop(fdt, offset, "ibm,pft-size", 2750da6f3feSBharata B Rao pft_size_prop, sizeof(pft_size_prop)); 27653018216SPaolo Bonzini if (ret < 0) { 27753018216SPaolo Bonzini return ret; 27853018216SPaolo Bonzini } 27953018216SPaolo Bonzini 2800da6f3feSBharata B Rao ret = spapr_fixup_cpu_numa_dt(fdt, offset, cs); 28153018216SPaolo Bonzini if (ret < 0) { 28253018216SPaolo Bonzini return ret; 28353018216SPaolo Bonzini } 284833d4668SAlexey Kardashevskiy 28512dbeb16SDavid Gibson ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu, compat_smt); 286833d4668SAlexey Kardashevskiy if (ret < 0) { 287833d4668SAlexey Kardashevskiy return ret; 288833d4668SAlexey Kardashevskiy } 28953018216SPaolo Bonzini } 29053018216SPaolo Bonzini return ret; 29153018216SPaolo Bonzini } 29253018216SPaolo Bonzini 293b082d65aSAlexey Kardashevskiy static hwaddr spapr_node0_size(void) 294b082d65aSAlexey Kardashevskiy { 295fb164994SDavid Gibson MachineState *machine = MACHINE(qdev_get_machine()); 296fb164994SDavid Gibson 297b082d65aSAlexey Kardashevskiy if (nb_numa_nodes) { 298b082d65aSAlexey Kardashevskiy int i; 299b082d65aSAlexey Kardashevskiy for (i = 0; i < nb_numa_nodes; ++i) { 300b082d65aSAlexey Kardashevskiy if (numa_info[i].node_mem) { 301fb164994SDavid Gibson return MIN(pow2floor(numa_info[i].node_mem), 302fb164994SDavid Gibson machine->ram_size); 303b082d65aSAlexey Kardashevskiy } 304b082d65aSAlexey Kardashevskiy } 305b082d65aSAlexey Kardashevskiy } 306fb164994SDavid Gibson return machine->ram_size; 307b082d65aSAlexey Kardashevskiy } 308b082d65aSAlexey Kardashevskiy 309a1d59c0fSAlexey Kardashevskiy static void add_str(GString *s, const gchar *s1) 310a1d59c0fSAlexey Kardashevskiy { 311a1d59c0fSAlexey Kardashevskiy g_string_append_len(s, s1, strlen(s1) + 1); 312a1d59c0fSAlexey Kardashevskiy } 31353018216SPaolo Bonzini 31403d196b7SBharata B Rao static int spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start, 31526a8c353SAlexey Kardashevskiy hwaddr size) 31626a8c353SAlexey Kardashevskiy { 31726a8c353SAlexey Kardashevskiy uint32_t associativity[] = { 31826a8c353SAlexey Kardashevskiy cpu_to_be32(0x4), /* length */ 31926a8c353SAlexey Kardashevskiy cpu_to_be32(0x0), cpu_to_be32(0x0), 320c3b4f589SAlexey Kardashevskiy cpu_to_be32(0x0), cpu_to_be32(nodeid) 32126a8c353SAlexey Kardashevskiy }; 32226a8c353SAlexey Kardashevskiy char mem_name[32]; 32326a8c353SAlexey Kardashevskiy uint64_t mem_reg_property[2]; 32426a8c353SAlexey Kardashevskiy int off; 32526a8c353SAlexey Kardashevskiy 32626a8c353SAlexey Kardashevskiy mem_reg_property[0] = cpu_to_be64(start); 32726a8c353SAlexey Kardashevskiy mem_reg_property[1] = cpu_to_be64(size); 32826a8c353SAlexey Kardashevskiy 32926a8c353SAlexey Kardashevskiy sprintf(mem_name, "memory@" TARGET_FMT_lx, start); 33026a8c353SAlexey Kardashevskiy off = fdt_add_subnode(fdt, 0, mem_name); 33126a8c353SAlexey Kardashevskiy _FDT(off); 33226a8c353SAlexey Kardashevskiy _FDT((fdt_setprop_string(fdt, off, "device_type", "memory"))); 33326a8c353SAlexey Kardashevskiy _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property, 33426a8c353SAlexey Kardashevskiy sizeof(mem_reg_property)))); 33526a8c353SAlexey Kardashevskiy _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity, 33626a8c353SAlexey Kardashevskiy sizeof(associativity)))); 33703d196b7SBharata B Rao return off; 33826a8c353SAlexey Kardashevskiy } 33926a8c353SAlexey Kardashevskiy 34028e02042SDavid Gibson static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt) 34153018216SPaolo Bonzini { 342fb164994SDavid Gibson MachineState *machine = MACHINE(spapr); 3437db8a127SAlexey Kardashevskiy hwaddr mem_start, node_size; 3447db8a127SAlexey Kardashevskiy int i, nb_nodes = nb_numa_nodes; 3457db8a127SAlexey Kardashevskiy NodeInfo *nodes = numa_info; 3467db8a127SAlexey Kardashevskiy NodeInfo ramnode; 34753018216SPaolo Bonzini 3487db8a127SAlexey Kardashevskiy /* No NUMA nodes, assume there is just one node with whole RAM */ 3497db8a127SAlexey Kardashevskiy if (!nb_numa_nodes) { 3507db8a127SAlexey Kardashevskiy nb_nodes = 1; 351fb164994SDavid Gibson ramnode.node_mem = machine->ram_size; 3527db8a127SAlexey Kardashevskiy nodes = &ramnode; 3535fe269b1SPaul Mackerras } 35453018216SPaolo Bonzini 3557db8a127SAlexey Kardashevskiy for (i = 0, mem_start = 0; i < nb_nodes; ++i) { 3567db8a127SAlexey Kardashevskiy if (!nodes[i].node_mem) { 3577db8a127SAlexey Kardashevskiy continue; 35853018216SPaolo Bonzini } 359fb164994SDavid Gibson if (mem_start >= machine->ram_size) { 3605fe269b1SPaul Mackerras node_size = 0; 3615fe269b1SPaul Mackerras } else { 3627db8a127SAlexey Kardashevskiy node_size = nodes[i].node_mem; 363fb164994SDavid Gibson if (node_size > machine->ram_size - mem_start) { 364fb164994SDavid Gibson node_size = machine->ram_size - mem_start; 3655fe269b1SPaul Mackerras } 3665fe269b1SPaul Mackerras } 3677db8a127SAlexey Kardashevskiy if (!mem_start) { 3687db8a127SAlexey Kardashevskiy /* ppc_spapr_init() checks for rma_size <= node0_size already */ 369e8f986fcSBharata B Rao spapr_populate_memory_node(fdt, i, 0, spapr->rma_size); 3707db8a127SAlexey Kardashevskiy mem_start += spapr->rma_size; 3717db8a127SAlexey Kardashevskiy node_size -= spapr->rma_size; 3727db8a127SAlexey Kardashevskiy } 3736010818cSAlexey Kardashevskiy for ( ; node_size; ) { 3746010818cSAlexey Kardashevskiy hwaddr sizetmp = pow2floor(node_size); 3756010818cSAlexey Kardashevskiy 3766010818cSAlexey Kardashevskiy /* mem_start != 0 here */ 3776010818cSAlexey Kardashevskiy if (ctzl(mem_start) < ctzl(sizetmp)) { 3786010818cSAlexey Kardashevskiy sizetmp = 1ULL << ctzl(mem_start); 3796010818cSAlexey Kardashevskiy } 3806010818cSAlexey Kardashevskiy 3816010818cSAlexey Kardashevskiy spapr_populate_memory_node(fdt, i, mem_start, sizetmp); 3826010818cSAlexey Kardashevskiy node_size -= sizetmp; 3836010818cSAlexey Kardashevskiy mem_start += sizetmp; 3846010818cSAlexey Kardashevskiy } 38553018216SPaolo Bonzini } 38653018216SPaolo Bonzini 38753018216SPaolo Bonzini return 0; 38853018216SPaolo Bonzini } 38953018216SPaolo Bonzini 390230bf719SThomas Huth /* Populate the "ibm,pa-features" property */ 391230bf719SThomas Huth static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset) 392230bf719SThomas Huth { 393230bf719SThomas Huth uint8_t pa_features_206[] = { 6, 0, 394230bf719SThomas Huth 0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 }; 395230bf719SThomas Huth uint8_t pa_features_207[] = { 24, 0, 396230bf719SThomas Huth 0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0, 397230bf719SThomas Huth 0x80, 0x00, 0x00, 0x00, 0x00, 0x00, 398230bf719SThomas Huth 0x00, 0x00, 0x00, 0x00, 0x80, 0x00, 399bac3bf28SThomas Huth 0x80, 0x00, 0x80, 0x00, 0x00, 0x00 }; 400230bf719SThomas Huth uint8_t *pa_features; 401230bf719SThomas Huth size_t pa_size; 402230bf719SThomas Huth 4034cbec30dSThomas Huth switch (env->mmu_model) { 4044cbec30dSThomas Huth case POWERPC_MMU_2_06: 4054cbec30dSThomas Huth case POWERPC_MMU_2_06a: 406230bf719SThomas Huth pa_features = pa_features_206; 407230bf719SThomas Huth pa_size = sizeof(pa_features_206); 4084cbec30dSThomas Huth break; 4094cbec30dSThomas Huth case POWERPC_MMU_2_07: 4104cbec30dSThomas Huth case POWERPC_MMU_2_07a: 411230bf719SThomas Huth pa_features = pa_features_207; 412230bf719SThomas Huth pa_size = sizeof(pa_features_207); 4134cbec30dSThomas Huth break; 4144cbec30dSThomas Huth default: 4154cbec30dSThomas Huth return; 416230bf719SThomas Huth } 417230bf719SThomas Huth 418230bf719SThomas Huth if (env->ci_large_pages) { 419230bf719SThomas Huth /* 420230bf719SThomas Huth * Note: we keep CI large pages off by default because a 64K capable 421230bf719SThomas Huth * guest provisioned with large pages might otherwise try to map a qemu 422230bf719SThomas Huth * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages 423230bf719SThomas Huth * even if that qemu runs on a 4k host. 424230bf719SThomas Huth * We dd this bit back here if we are confident this is not an issue 425230bf719SThomas Huth */ 426230bf719SThomas Huth pa_features[3] |= 0x20; 427230bf719SThomas Huth } 428bac3bf28SThomas Huth if (kvmppc_has_cap_htm() && pa_size > 24) { 429bac3bf28SThomas Huth pa_features[24] |= 0x80; /* Transactional memory support */ 430bac3bf28SThomas Huth } 431230bf719SThomas Huth 432230bf719SThomas Huth _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size))); 433230bf719SThomas Huth } 434230bf719SThomas Huth 4350da6f3feSBharata B Rao static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset, 4360da6f3feSBharata B Rao sPAPRMachineState *spapr) 4370da6f3feSBharata B Rao { 4380da6f3feSBharata B Rao PowerPCCPU *cpu = POWERPC_CPU(cs); 4390da6f3feSBharata B Rao CPUPPCState *env = &cpu->env; 4400da6f3feSBharata B Rao PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs); 4410da6f3feSBharata B Rao int index = ppc_get_vcpu_dt_id(cpu); 4420da6f3feSBharata B Rao uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40), 4430da6f3feSBharata B Rao 0xffffffff, 0xffffffff}; 444afd10a0fSBharata B Rao uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() 445afd10a0fSBharata B Rao : SPAPR_TIMEBASE_FREQ; 4460da6f3feSBharata B Rao uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000; 4470da6f3feSBharata B Rao uint32_t page_sizes_prop[64]; 4480da6f3feSBharata B Rao size_t page_sizes_prop_size; 44922419c2aSDavid Gibson uint32_t vcpus_per_socket = smp_threads * smp_cores; 4500da6f3feSBharata B Rao uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)}; 45112dbeb16SDavid Gibson int compat_smt = MIN(smp_threads, ppc_compat_max_threads(cpu)); 452af81cf32SBharata B Rao sPAPRDRConnector *drc; 453af81cf32SBharata B Rao sPAPRDRConnectorClass *drck; 454af81cf32SBharata B Rao int drc_index; 455af81cf32SBharata B Rao 456af81cf32SBharata B Rao drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_CPU, index); 457af81cf32SBharata B Rao if (drc) { 458af81cf32SBharata B Rao drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); 459af81cf32SBharata B Rao drc_index = drck->get_index(drc); 460af81cf32SBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "ibm,my-drc-index", drc_index))); 461af81cf32SBharata B Rao } 4620da6f3feSBharata B Rao 4630da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "reg", index))); 4640da6f3feSBharata B Rao _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu"))); 4650da6f3feSBharata B Rao 4660da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "cpu-version", env->spr[SPR_PVR]))); 4670da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "d-cache-block-size", 4680da6f3feSBharata B Rao env->dcache_line_size))); 4690da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "d-cache-line-size", 4700da6f3feSBharata B Rao env->dcache_line_size))); 4710da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "i-cache-block-size", 4720da6f3feSBharata B Rao env->icache_line_size))); 4730da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "i-cache-line-size", 4740da6f3feSBharata B Rao env->icache_line_size))); 4750da6f3feSBharata B Rao 4760da6f3feSBharata B Rao if (pcc->l1_dcache_size) { 4770da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "d-cache-size", 4780da6f3feSBharata B Rao pcc->l1_dcache_size))); 4790da6f3feSBharata B Rao } else { 480ce9863b7SCédric Le Goater error_report("Warning: Unknown L1 dcache size for cpu"); 4810da6f3feSBharata B Rao } 4820da6f3feSBharata B Rao if (pcc->l1_icache_size) { 4830da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "i-cache-size", 4840da6f3feSBharata B Rao pcc->l1_icache_size))); 4850da6f3feSBharata B Rao } else { 486ce9863b7SCédric Le Goater error_report("Warning: Unknown L1 icache size for cpu"); 4870da6f3feSBharata B Rao } 4880da6f3feSBharata B Rao 4890da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "timebase-frequency", tbfreq))); 4900da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "clock-frequency", cpufreq))); 491fd5da5c4SThomas Huth _FDT((fdt_setprop_cell(fdt, offset, "slb-size", env->slb_nr))); 4920da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "ibm,slb-size", env->slb_nr))); 4930da6f3feSBharata B Rao _FDT((fdt_setprop_string(fdt, offset, "status", "okay"))); 4940da6f3feSBharata B Rao _FDT((fdt_setprop(fdt, offset, "64-bit", NULL, 0))); 4950da6f3feSBharata B Rao 4960da6f3feSBharata B Rao if (env->spr_cb[SPR_PURR].oea_read) { 4970da6f3feSBharata B Rao _FDT((fdt_setprop(fdt, offset, "ibm,purr", NULL, 0))); 4980da6f3feSBharata B Rao } 4990da6f3feSBharata B Rao 5000da6f3feSBharata B Rao if (env->mmu_model & POWERPC_MMU_1TSEG) { 5010da6f3feSBharata B Rao _FDT((fdt_setprop(fdt, offset, "ibm,processor-segment-sizes", 5020da6f3feSBharata B Rao segs, sizeof(segs)))); 5030da6f3feSBharata B Rao } 5040da6f3feSBharata B Rao 5050da6f3feSBharata B Rao /* Advertise VMX/VSX (vector extensions) if available 5060da6f3feSBharata B Rao * 0 / no property == no vector extensions 5070da6f3feSBharata B Rao * 1 == VMX / Altivec available 5080da6f3feSBharata B Rao * 2 == VSX available */ 5090da6f3feSBharata B Rao if (env->insns_flags & PPC_ALTIVEC) { 5100da6f3feSBharata B Rao uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1; 5110da6f3feSBharata B Rao 5120da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx))); 5130da6f3feSBharata B Rao } 5140da6f3feSBharata B Rao 5150da6f3feSBharata B Rao /* Advertise DFP (Decimal Floating Point) if available 5160da6f3feSBharata B Rao * 0 / no property == no DFP 5170da6f3feSBharata B Rao * 1 == DFP available */ 5180da6f3feSBharata B Rao if (env->insns_flags2 & PPC2_DFP) { 5190da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "ibm,dfp", 1))); 5200da6f3feSBharata B Rao } 5210da6f3feSBharata B Rao 5223654fa95SCédric Le Goater page_sizes_prop_size = ppc_create_page_sizes_prop(env, page_sizes_prop, 5230da6f3feSBharata B Rao sizeof(page_sizes_prop)); 5240da6f3feSBharata B Rao if (page_sizes_prop_size) { 5250da6f3feSBharata B Rao _FDT((fdt_setprop(fdt, offset, "ibm,segment-page-sizes", 5260da6f3feSBharata B Rao page_sizes_prop, page_sizes_prop_size))); 5270da6f3feSBharata B Rao } 5280da6f3feSBharata B Rao 529230bf719SThomas Huth spapr_populate_pa_features(env, fdt, offset); 53090da0d5aSBenjamin Herrenschmidt 5310da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id", 53222419c2aSDavid Gibson cs->cpu_index / vcpus_per_socket))); 5330da6f3feSBharata B Rao 5340da6f3feSBharata B Rao _FDT((fdt_setprop(fdt, offset, "ibm,pft-size", 5350da6f3feSBharata B Rao pft_size_prop, sizeof(pft_size_prop)))); 5360da6f3feSBharata B Rao 5370da6f3feSBharata B Rao _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cs)); 5380da6f3feSBharata B Rao 53912dbeb16SDavid Gibson _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu, compat_smt)); 5400da6f3feSBharata B Rao } 5410da6f3feSBharata B Rao 5420da6f3feSBharata B Rao static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr) 5430da6f3feSBharata B Rao { 5440da6f3feSBharata B Rao CPUState *cs; 5450da6f3feSBharata B Rao int cpus_offset; 5460da6f3feSBharata B Rao char *nodename; 5470da6f3feSBharata B Rao int smt = kvmppc_smt_threads(); 5480da6f3feSBharata B Rao 5490da6f3feSBharata B Rao cpus_offset = fdt_add_subnode(fdt, 0, "cpus"); 5500da6f3feSBharata B Rao _FDT(cpus_offset); 5510da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, cpus_offset, "#address-cells", 0x1))); 5520da6f3feSBharata B Rao _FDT((fdt_setprop_cell(fdt, cpus_offset, "#size-cells", 0x0))); 5530da6f3feSBharata B Rao 5540da6f3feSBharata B Rao /* 5550da6f3feSBharata B Rao * We walk the CPUs in reverse order to ensure that CPU DT nodes 5560da6f3feSBharata B Rao * created by fdt_add_subnode() end up in the right order in FDT 5570da6f3feSBharata B Rao * for the guest kernel the enumerate the CPUs correctly. 5580da6f3feSBharata B Rao */ 5590da6f3feSBharata B Rao CPU_FOREACH_REVERSE(cs) { 5600da6f3feSBharata B Rao PowerPCCPU *cpu = POWERPC_CPU(cs); 5610da6f3feSBharata B Rao int index = ppc_get_vcpu_dt_id(cpu); 5620da6f3feSBharata B Rao DeviceClass *dc = DEVICE_GET_CLASS(cs); 5630da6f3feSBharata B Rao int offset; 5640da6f3feSBharata B Rao 5650da6f3feSBharata B Rao if ((index % smt) != 0) { 5660da6f3feSBharata B Rao continue; 5670da6f3feSBharata B Rao } 5680da6f3feSBharata B Rao 5690da6f3feSBharata B Rao nodename = g_strdup_printf("%s@%x", dc->fw_name, index); 5700da6f3feSBharata B Rao offset = fdt_add_subnode(fdt, cpus_offset, nodename); 5710da6f3feSBharata B Rao g_free(nodename); 5720da6f3feSBharata B Rao _FDT(offset); 5730da6f3feSBharata B Rao spapr_populate_cpu_dt(cs, fdt, offset, spapr); 5740da6f3feSBharata B Rao } 5750da6f3feSBharata B Rao 5760da6f3feSBharata B Rao } 5770da6f3feSBharata B Rao 57803d196b7SBharata B Rao /* 57903d196b7SBharata B Rao * Adds ibm,dynamic-reconfiguration-memory node. 58003d196b7SBharata B Rao * Refer to docs/specs/ppc-spapr-hotplug.txt for the documentation 58103d196b7SBharata B Rao * of this device tree node. 58203d196b7SBharata B Rao */ 58303d196b7SBharata B Rao static int spapr_populate_drconf_memory(sPAPRMachineState *spapr, void *fdt) 58403d196b7SBharata B Rao { 58503d196b7SBharata B Rao MachineState *machine = MACHINE(spapr); 58603d196b7SBharata B Rao int ret, i, offset; 58703d196b7SBharata B Rao uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE; 58803d196b7SBharata B Rao uint32_t prop_lmb_size[] = {0, cpu_to_be32(lmb_size)}; 589d0e5a8f2SBharata B Rao uint32_t hotplug_lmb_start = spapr->hotplug_memory.base / lmb_size; 590d0e5a8f2SBharata B Rao uint32_t nr_lmbs = (spapr->hotplug_memory.base + 591d0e5a8f2SBharata B Rao memory_region_size(&spapr->hotplug_memory.mr)) / 592d0e5a8f2SBharata B Rao lmb_size; 59303d196b7SBharata B Rao uint32_t *int_buf, *cur_index, buf_len; 5946663864eSBharata B Rao int nr_nodes = nb_numa_nodes ? nb_numa_nodes : 1; 59503d196b7SBharata B Rao 596ef001f06SThomas Huth /* 597d0e5a8f2SBharata B Rao * Don't create the node if there is no hotpluggable memory 59816c25aefSBharata B Rao */ 599d0e5a8f2SBharata B Rao if (machine->ram_size == machine->maxram_size) { 60016c25aefSBharata B Rao return 0; 60116c25aefSBharata B Rao } 60216c25aefSBharata B Rao 60316c25aefSBharata B Rao /* 604ef001f06SThomas Huth * Allocate enough buffer size to fit in ibm,dynamic-memory 605ef001f06SThomas Huth * or ibm,associativity-lookup-arrays 606ef001f06SThomas Huth */ 607ef001f06SThomas Huth buf_len = MAX(nr_lmbs * SPAPR_DR_LMB_LIST_ENTRY_SIZE + 1, nr_nodes * 4 + 2) 608ef001f06SThomas Huth * sizeof(uint32_t); 60903d196b7SBharata B Rao cur_index = int_buf = g_malloc0(buf_len); 61003d196b7SBharata B Rao 61103d196b7SBharata B Rao offset = fdt_add_subnode(fdt, 0, "ibm,dynamic-reconfiguration-memory"); 61203d196b7SBharata B Rao 61303d196b7SBharata B Rao ret = fdt_setprop(fdt, offset, "ibm,lmb-size", prop_lmb_size, 61403d196b7SBharata B Rao sizeof(prop_lmb_size)); 61503d196b7SBharata B Rao if (ret < 0) { 61603d196b7SBharata B Rao goto out; 61703d196b7SBharata B Rao } 61803d196b7SBharata B Rao 61903d196b7SBharata B Rao ret = fdt_setprop_cell(fdt, offset, "ibm,memory-flags-mask", 0xff); 62003d196b7SBharata B Rao if (ret < 0) { 62103d196b7SBharata B Rao goto out; 62203d196b7SBharata B Rao } 62303d196b7SBharata B Rao 62403d196b7SBharata B Rao ret = fdt_setprop_cell(fdt, offset, "ibm,memory-preservation-time", 0x0); 62503d196b7SBharata B Rao if (ret < 0) { 62603d196b7SBharata B Rao goto out; 62703d196b7SBharata B Rao } 62803d196b7SBharata B Rao 62903d196b7SBharata B Rao /* ibm,dynamic-memory */ 63003d196b7SBharata B Rao int_buf[0] = cpu_to_be32(nr_lmbs); 63103d196b7SBharata B Rao cur_index++; 63203d196b7SBharata B Rao for (i = 0; i < nr_lmbs; i++) { 633d0e5a8f2SBharata B Rao uint64_t addr = i * lmb_size; 63403d196b7SBharata B Rao uint32_t *dynamic_memory = cur_index; 63503d196b7SBharata B Rao 636d0e5a8f2SBharata B Rao if (i >= hotplug_lmb_start) { 637d0e5a8f2SBharata B Rao sPAPRDRConnector *drc; 638d0e5a8f2SBharata B Rao sPAPRDRConnectorClass *drck; 639d0e5a8f2SBharata B Rao 640d0e5a8f2SBharata B Rao drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB, i); 64103d196b7SBharata B Rao g_assert(drc); 64203d196b7SBharata B Rao drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); 64303d196b7SBharata B Rao 64403d196b7SBharata B Rao dynamic_memory[0] = cpu_to_be32(addr >> 32); 64503d196b7SBharata B Rao dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff); 64603d196b7SBharata B Rao dynamic_memory[2] = cpu_to_be32(drck->get_index(drc)); 64703d196b7SBharata B Rao dynamic_memory[3] = cpu_to_be32(0); /* reserved */ 64803d196b7SBharata B Rao dynamic_memory[4] = cpu_to_be32(numa_get_node(addr, NULL)); 649d0e5a8f2SBharata B Rao if (memory_region_present(get_system_memory(), addr)) { 65003d196b7SBharata B Rao dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_ASSIGNED); 65103d196b7SBharata B Rao } else { 65203d196b7SBharata B Rao dynamic_memory[5] = cpu_to_be32(0); 65303d196b7SBharata B Rao } 654d0e5a8f2SBharata B Rao } else { 655d0e5a8f2SBharata B Rao /* 656d0e5a8f2SBharata B Rao * LMB information for RMA, boot time RAM and gap b/n RAM and 657d0e5a8f2SBharata B Rao * hotplug memory region -- all these are marked as reserved 658d0e5a8f2SBharata B Rao * and as having no valid DRC. 659d0e5a8f2SBharata B Rao */ 660d0e5a8f2SBharata B Rao dynamic_memory[0] = cpu_to_be32(addr >> 32); 661d0e5a8f2SBharata B Rao dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff); 662d0e5a8f2SBharata B Rao dynamic_memory[2] = cpu_to_be32(0); 663d0e5a8f2SBharata B Rao dynamic_memory[3] = cpu_to_be32(0); /* reserved */ 664d0e5a8f2SBharata B Rao dynamic_memory[4] = cpu_to_be32(-1); 665d0e5a8f2SBharata B Rao dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_RESERVED | 666d0e5a8f2SBharata B Rao SPAPR_LMB_FLAGS_DRC_INVALID); 667d0e5a8f2SBharata B Rao } 66803d196b7SBharata B Rao 66903d196b7SBharata B Rao cur_index += SPAPR_DR_LMB_LIST_ENTRY_SIZE; 67003d196b7SBharata B Rao } 67103d196b7SBharata B Rao ret = fdt_setprop(fdt, offset, "ibm,dynamic-memory", int_buf, buf_len); 67203d196b7SBharata B Rao if (ret < 0) { 67303d196b7SBharata B Rao goto out; 67403d196b7SBharata B Rao } 67503d196b7SBharata B Rao 67603d196b7SBharata B Rao /* ibm,associativity-lookup-arrays */ 67703d196b7SBharata B Rao cur_index = int_buf; 6786663864eSBharata B Rao int_buf[0] = cpu_to_be32(nr_nodes); 67903d196b7SBharata B Rao int_buf[1] = cpu_to_be32(4); /* Number of entries per associativity list */ 68003d196b7SBharata B Rao cur_index += 2; 6816663864eSBharata B Rao for (i = 0; i < nr_nodes; i++) { 68203d196b7SBharata B Rao uint32_t associativity[] = { 68303d196b7SBharata B Rao cpu_to_be32(0x0), 68403d196b7SBharata B Rao cpu_to_be32(0x0), 68503d196b7SBharata B Rao cpu_to_be32(0x0), 68603d196b7SBharata B Rao cpu_to_be32(i) 68703d196b7SBharata B Rao }; 68803d196b7SBharata B Rao memcpy(cur_index, associativity, sizeof(associativity)); 68903d196b7SBharata B Rao cur_index += 4; 69003d196b7SBharata B Rao } 69103d196b7SBharata B Rao ret = fdt_setprop(fdt, offset, "ibm,associativity-lookup-arrays", int_buf, 69203d196b7SBharata B Rao (cur_index - int_buf) * sizeof(uint32_t)); 69303d196b7SBharata B Rao out: 69403d196b7SBharata B Rao g_free(int_buf); 69503d196b7SBharata B Rao return ret; 69603d196b7SBharata B Rao } 69703d196b7SBharata B Rao 6986787d27bSMichael Roth static int spapr_dt_cas_updates(sPAPRMachineState *spapr, void *fdt, 6996787d27bSMichael Roth sPAPROptionVector *ov5_updates) 7006787d27bSMichael Roth { 7016787d27bSMichael Roth sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr); 702417ece33SMichael Roth int ret = 0, offset; 7036787d27bSMichael Roth 7046787d27bSMichael Roth /* Generate ibm,dynamic-reconfiguration-memory node if required */ 7056787d27bSMichael Roth if (spapr_ovec_test(ov5_updates, OV5_DRCONF_MEMORY)) { 7066787d27bSMichael Roth g_assert(smc->dr_lmb_enabled); 7076787d27bSMichael Roth ret = spapr_populate_drconf_memory(spapr, fdt); 708417ece33SMichael Roth if (ret) { 709417ece33SMichael Roth goto out; 710417ece33SMichael Roth } 7116787d27bSMichael Roth } 7126787d27bSMichael Roth 713417ece33SMichael Roth offset = fdt_path_offset(fdt, "/chosen"); 714417ece33SMichael Roth if (offset < 0) { 715417ece33SMichael Roth offset = fdt_add_subnode(fdt, 0, "chosen"); 716417ece33SMichael Roth if (offset < 0) { 717417ece33SMichael Roth return offset; 718417ece33SMichael Roth } 719417ece33SMichael Roth } 720417ece33SMichael Roth ret = spapr_ovec_populate_dt(fdt, offset, spapr->ov5_cas, 721417ece33SMichael Roth "ibm,architecture-vec-5"); 722417ece33SMichael Roth 723417ece33SMichael Roth out: 7246787d27bSMichael Roth return ret; 7256787d27bSMichael Roth } 7266787d27bSMichael Roth 72703d196b7SBharata B Rao int spapr_h_cas_compose_response(sPAPRMachineState *spapr, 72803d196b7SBharata B Rao target_ulong addr, target_ulong size, 7296787d27bSMichael Roth sPAPROptionVector *ov5_updates) 73003d196b7SBharata B Rao { 73103d196b7SBharata B Rao void *fdt, *fdt_skel; 73203d196b7SBharata B Rao sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 }; 73303d196b7SBharata B Rao 73403d196b7SBharata B Rao size -= sizeof(hdr); 73503d196b7SBharata B Rao 73603d196b7SBharata B Rao /* Create sceleton */ 73703d196b7SBharata B Rao fdt_skel = g_malloc0(size); 73803d196b7SBharata B Rao _FDT((fdt_create(fdt_skel, size))); 73903d196b7SBharata B Rao _FDT((fdt_begin_node(fdt_skel, ""))); 74003d196b7SBharata B Rao _FDT((fdt_end_node(fdt_skel))); 74103d196b7SBharata B Rao _FDT((fdt_finish(fdt_skel))); 74203d196b7SBharata B Rao fdt = g_malloc0(size); 74303d196b7SBharata B Rao _FDT((fdt_open_into(fdt_skel, fdt, size))); 74403d196b7SBharata B Rao g_free(fdt_skel); 74503d196b7SBharata B Rao 74603d196b7SBharata B Rao /* Fixup cpu nodes */ 74703d196b7SBharata B Rao _FDT((spapr_fixup_cpu_dt(fdt, spapr))); 74803d196b7SBharata B Rao 7496787d27bSMichael Roth if (spapr_dt_cas_updates(spapr, fdt, ov5_updates)) { 7506787d27bSMichael Roth return -1; 75103d196b7SBharata B Rao } 75203d196b7SBharata B Rao 75303d196b7SBharata B Rao /* Pack resulting tree */ 75403d196b7SBharata B Rao _FDT((fdt_pack(fdt))); 75503d196b7SBharata B Rao 75603d196b7SBharata B Rao if (fdt_totalsize(fdt) + sizeof(hdr) > size) { 75703d196b7SBharata B Rao trace_spapr_cas_failed(size); 75803d196b7SBharata B Rao return -1; 75903d196b7SBharata B Rao } 76003d196b7SBharata B Rao 76103d196b7SBharata B Rao cpu_physical_memory_write(addr, &hdr, sizeof(hdr)); 76203d196b7SBharata B Rao cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt)); 76303d196b7SBharata B Rao trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr)); 76403d196b7SBharata B Rao g_free(fdt); 76503d196b7SBharata B Rao 76603d196b7SBharata B Rao return 0; 76703d196b7SBharata B Rao } 76803d196b7SBharata B Rao 7693f5dabceSDavid Gibson static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt) 7703f5dabceSDavid Gibson { 7713f5dabceSDavid Gibson int rtas; 7723f5dabceSDavid Gibson GString *hypertas = g_string_sized_new(256); 7733f5dabceSDavid Gibson GString *qemu_hypertas = g_string_sized_new(256); 7743f5dabceSDavid Gibson uint32_t refpoints[] = { cpu_to_be32(0x4), cpu_to_be32(0x4) }; 7753f5dabceSDavid Gibson uint64_t max_hotplug_addr = spapr->hotplug_memory.base + 7763f5dabceSDavid Gibson memory_region_size(&spapr->hotplug_memory.mr); 7773f5dabceSDavid Gibson uint32_t lrdr_capacity[] = { 7783f5dabceSDavid Gibson cpu_to_be32(max_hotplug_addr >> 32), 7793f5dabceSDavid Gibson cpu_to_be32(max_hotplug_addr & 0xffffffff), 7803f5dabceSDavid Gibson 0, cpu_to_be32(SPAPR_MEMORY_BLOCK_SIZE), 7813f5dabceSDavid Gibson cpu_to_be32(max_cpus / smp_threads), 7823f5dabceSDavid Gibson }; 7833f5dabceSDavid Gibson 7843f5dabceSDavid Gibson _FDT(rtas = fdt_add_subnode(fdt, 0, "rtas")); 7853f5dabceSDavid Gibson 7863f5dabceSDavid Gibson /* hypertas */ 7873f5dabceSDavid Gibson add_str(hypertas, "hcall-pft"); 7883f5dabceSDavid Gibson add_str(hypertas, "hcall-term"); 7893f5dabceSDavid Gibson add_str(hypertas, "hcall-dabr"); 7903f5dabceSDavid Gibson add_str(hypertas, "hcall-interrupt"); 7913f5dabceSDavid Gibson add_str(hypertas, "hcall-tce"); 7923f5dabceSDavid Gibson add_str(hypertas, "hcall-vio"); 7933f5dabceSDavid Gibson add_str(hypertas, "hcall-splpar"); 7943f5dabceSDavid Gibson add_str(hypertas, "hcall-bulk"); 7953f5dabceSDavid Gibson add_str(hypertas, "hcall-set-mode"); 7963f5dabceSDavid Gibson add_str(hypertas, "hcall-sprg0"); 7973f5dabceSDavid Gibson add_str(hypertas, "hcall-copy"); 7983f5dabceSDavid Gibson add_str(hypertas, "hcall-debug"); 7993f5dabceSDavid Gibson add_str(qemu_hypertas, "hcall-memop1"); 8003f5dabceSDavid Gibson 8013f5dabceSDavid Gibson if (!kvm_enabled() || kvmppc_spapr_use_multitce()) { 8023f5dabceSDavid Gibson add_str(hypertas, "hcall-multi-tce"); 8033f5dabceSDavid Gibson } 8043f5dabceSDavid Gibson _FDT(fdt_setprop(fdt, rtas, "ibm,hypertas-functions", 8053f5dabceSDavid Gibson hypertas->str, hypertas->len)); 8063f5dabceSDavid Gibson g_string_free(hypertas, TRUE); 8073f5dabceSDavid Gibson _FDT(fdt_setprop(fdt, rtas, "qemu,hypertas-functions", 8083f5dabceSDavid Gibson qemu_hypertas->str, qemu_hypertas->len)); 8093f5dabceSDavid Gibson g_string_free(qemu_hypertas, TRUE); 8103f5dabceSDavid Gibson 8113f5dabceSDavid Gibson _FDT(fdt_setprop(fdt, rtas, "ibm,associativity-reference-points", 8123f5dabceSDavid Gibson refpoints, sizeof(refpoints))); 8133f5dabceSDavid Gibson 8143f5dabceSDavid Gibson _FDT(fdt_setprop_cell(fdt, rtas, "rtas-error-log-max", 8153f5dabceSDavid Gibson RTAS_ERROR_LOG_MAX)); 8163f5dabceSDavid Gibson _FDT(fdt_setprop_cell(fdt, rtas, "rtas-event-scan-rate", 8173f5dabceSDavid Gibson RTAS_EVENT_SCAN_RATE)); 8183f5dabceSDavid Gibson 8193f5dabceSDavid Gibson if (msi_nonbroken) { 8203f5dabceSDavid Gibson _FDT(fdt_setprop(fdt, rtas, "ibm,change-msix-capable", NULL, 0)); 8213f5dabceSDavid Gibson } 8223f5dabceSDavid Gibson 8233f5dabceSDavid Gibson /* 8243f5dabceSDavid Gibson * According to PAPR, rtas ibm,os-term does not guarantee a return 8253f5dabceSDavid Gibson * back to the guest cpu. 8263f5dabceSDavid Gibson * 8273f5dabceSDavid Gibson * While an additional ibm,extended-os-term property indicates 8283f5dabceSDavid Gibson * that rtas call return will always occur. Set this property. 8293f5dabceSDavid Gibson */ 8303f5dabceSDavid Gibson _FDT(fdt_setprop(fdt, rtas, "ibm,extended-os-term", NULL, 0)); 8313f5dabceSDavid Gibson 8323f5dabceSDavid Gibson _FDT(fdt_setprop(fdt, rtas, "ibm,lrdr-capacity", 8333f5dabceSDavid Gibson lrdr_capacity, sizeof(lrdr_capacity))); 8343f5dabceSDavid Gibson 8353f5dabceSDavid Gibson spapr_dt_rtas_tokens(fdt, rtas); 8363f5dabceSDavid Gibson } 8373f5dabceSDavid Gibson 8387c866c6aSDavid Gibson static void spapr_dt_chosen(sPAPRMachineState *spapr, void *fdt) 8397c866c6aSDavid Gibson { 8407c866c6aSDavid Gibson MachineState *machine = MACHINE(spapr); 8417c866c6aSDavid Gibson int chosen; 8427c866c6aSDavid Gibson const char *boot_device = machine->boot_order; 8437c866c6aSDavid Gibson char *stdout_path = spapr_vio_stdout_path(spapr->vio_bus); 8447c866c6aSDavid Gibson size_t cb = 0; 8457c866c6aSDavid Gibson char *bootlist = get_boot_devices_list(&cb, true); 8467c866c6aSDavid Gibson 8477c866c6aSDavid Gibson _FDT(chosen = fdt_add_subnode(fdt, 0, "chosen")); 8487c866c6aSDavid Gibson 8497c866c6aSDavid Gibson _FDT(fdt_setprop_string(fdt, chosen, "bootargs", machine->kernel_cmdline)); 8507c866c6aSDavid Gibson _FDT(fdt_setprop_cell(fdt, chosen, "linux,initrd-start", 8517c866c6aSDavid Gibson spapr->initrd_base)); 8527c866c6aSDavid Gibson _FDT(fdt_setprop_cell(fdt, chosen, "linux,initrd-end", 8537c866c6aSDavid Gibson spapr->initrd_base + spapr->initrd_size)); 8547c866c6aSDavid Gibson 8557c866c6aSDavid Gibson if (spapr->kernel_size) { 8567c866c6aSDavid Gibson uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR), 8577c866c6aSDavid Gibson cpu_to_be64(spapr->kernel_size) }; 8587c866c6aSDavid Gibson 8597c866c6aSDavid Gibson _FDT(fdt_setprop(fdt, chosen, "qemu,boot-kernel", 8607c866c6aSDavid Gibson &kprop, sizeof(kprop))); 8617c866c6aSDavid Gibson if (spapr->kernel_le) { 8627c866c6aSDavid Gibson _FDT(fdt_setprop(fdt, chosen, "qemu,boot-kernel-le", NULL, 0)); 8637c866c6aSDavid Gibson } 8647c866c6aSDavid Gibson } 8657c866c6aSDavid Gibson if (boot_menu) { 8667c866c6aSDavid Gibson _FDT((fdt_setprop_cell(fdt, chosen, "qemu,boot-menu", boot_menu))); 8677c866c6aSDavid Gibson } 8687c866c6aSDavid Gibson _FDT(fdt_setprop_cell(fdt, chosen, "qemu,graphic-width", graphic_width)); 8697c866c6aSDavid Gibson _FDT(fdt_setprop_cell(fdt, chosen, "qemu,graphic-height", graphic_height)); 8707c866c6aSDavid Gibson _FDT(fdt_setprop_cell(fdt, chosen, "qemu,graphic-depth", graphic_depth)); 8717c866c6aSDavid Gibson 8727c866c6aSDavid Gibson if (cb && bootlist) { 8737c866c6aSDavid Gibson int i; 8747c866c6aSDavid Gibson 8757c866c6aSDavid Gibson for (i = 0; i < cb; i++) { 8767c866c6aSDavid Gibson if (bootlist[i] == '\n') { 8777c866c6aSDavid Gibson bootlist[i] = ' '; 8787c866c6aSDavid Gibson } 8797c866c6aSDavid Gibson } 8807c866c6aSDavid Gibson _FDT(fdt_setprop_string(fdt, chosen, "qemu,boot-list", bootlist)); 8817c866c6aSDavid Gibson } 8827c866c6aSDavid Gibson 8837c866c6aSDavid Gibson if (boot_device && strlen(boot_device)) { 8847c866c6aSDavid Gibson _FDT(fdt_setprop_string(fdt, chosen, "qemu,boot-device", boot_device)); 8857c866c6aSDavid Gibson } 8867c866c6aSDavid Gibson 8877c866c6aSDavid Gibson if (!spapr->has_graphics && stdout_path) { 8887c866c6aSDavid Gibson _FDT(fdt_setprop_string(fdt, chosen, "linux,stdout-path", stdout_path)); 8897c866c6aSDavid Gibson } 8907c866c6aSDavid Gibson 8917c866c6aSDavid Gibson g_free(stdout_path); 8927c866c6aSDavid Gibson g_free(bootlist); 8937c866c6aSDavid Gibson } 8947c866c6aSDavid Gibson 895fca5f2dcSDavid Gibson static void spapr_dt_hypervisor(sPAPRMachineState *spapr, void *fdt) 896fca5f2dcSDavid Gibson { 897fca5f2dcSDavid Gibson /* The /hypervisor node isn't in PAPR - this is a hack to allow PR 898fca5f2dcSDavid Gibson * KVM to work under pHyp with some guest co-operation */ 899fca5f2dcSDavid Gibson int hypervisor; 900fca5f2dcSDavid Gibson uint8_t hypercall[16]; 901fca5f2dcSDavid Gibson 902fca5f2dcSDavid Gibson _FDT(hypervisor = fdt_add_subnode(fdt, 0, "hypervisor")); 903fca5f2dcSDavid Gibson /* indicate KVM hypercall interface */ 904fca5f2dcSDavid Gibson _FDT(fdt_setprop_string(fdt, hypervisor, "compatible", "linux,kvm")); 905fca5f2dcSDavid Gibson if (kvmppc_has_cap_fixup_hcalls()) { 906fca5f2dcSDavid Gibson /* 907fca5f2dcSDavid Gibson * Older KVM versions with older guest kernels were broken 908fca5f2dcSDavid Gibson * with the magic page, don't allow the guest to map it. 909fca5f2dcSDavid Gibson */ 910fca5f2dcSDavid Gibson if (!kvmppc_get_hypercall(first_cpu->env_ptr, hypercall, 911fca5f2dcSDavid Gibson sizeof(hypercall))) { 912fca5f2dcSDavid Gibson _FDT(fdt_setprop(fdt, hypervisor, "hcall-instructions", 913fca5f2dcSDavid Gibson hypercall, sizeof(hypercall))); 914fca5f2dcSDavid Gibson } 915fca5f2dcSDavid Gibson } 916fca5f2dcSDavid Gibson } 917fca5f2dcSDavid Gibson 918997b6cfcSDavid Gibson static void *spapr_build_fdt(sPAPRMachineState *spapr, 91953018216SPaolo Bonzini hwaddr rtas_addr, 92053018216SPaolo Bonzini hwaddr rtas_size) 92153018216SPaolo Bonzini { 9225b2128d2SAlexander Graf MachineState *machine = MACHINE(qdev_get_machine()); 9233c0c47e3SDavid Gibson MachineClass *mc = MACHINE_GET_CLASS(machine); 924c20d332aSBharata B Rao sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine); 9257c866c6aSDavid Gibson int ret; 92653018216SPaolo Bonzini void *fdt; 92753018216SPaolo Bonzini sPAPRPHBState *phb; 928398a0bd5SDavid Gibson char *buf; 92953018216SPaolo Bonzini 930398a0bd5SDavid Gibson fdt = g_malloc0(FDT_MAX_SIZE); 931398a0bd5SDavid Gibson _FDT((fdt_create_empty_tree(fdt, FDT_MAX_SIZE))); 93253018216SPaolo Bonzini 933398a0bd5SDavid Gibson /* Root node */ 934398a0bd5SDavid Gibson _FDT(fdt_setprop_string(fdt, 0, "device_type", "chrp")); 935398a0bd5SDavid Gibson _FDT(fdt_setprop_string(fdt, 0, "model", "IBM pSeries (emulated by qemu)")); 936398a0bd5SDavid Gibson _FDT(fdt_setprop_string(fdt, 0, "compatible", "qemu,pseries")); 937398a0bd5SDavid Gibson 938398a0bd5SDavid Gibson /* 939398a0bd5SDavid Gibson * Add info to guest to indentify which host is it being run on 940398a0bd5SDavid Gibson * and what is the uuid of the guest 941398a0bd5SDavid Gibson */ 942398a0bd5SDavid Gibson if (kvmppc_get_host_model(&buf)) { 943398a0bd5SDavid Gibson _FDT(fdt_setprop_string(fdt, 0, "host-model", buf)); 944398a0bd5SDavid Gibson g_free(buf); 945398a0bd5SDavid Gibson } 946398a0bd5SDavid Gibson if (kvmppc_get_host_serial(&buf)) { 947398a0bd5SDavid Gibson _FDT(fdt_setprop_string(fdt, 0, "host-serial", buf)); 948398a0bd5SDavid Gibson g_free(buf); 949398a0bd5SDavid Gibson } 950398a0bd5SDavid Gibson 951398a0bd5SDavid Gibson buf = qemu_uuid_unparse_strdup(&qemu_uuid); 952398a0bd5SDavid Gibson 953398a0bd5SDavid Gibson _FDT(fdt_setprop_string(fdt, 0, "vm,uuid", buf)); 954398a0bd5SDavid Gibson if (qemu_uuid_set) { 955398a0bd5SDavid Gibson _FDT(fdt_setprop_string(fdt, 0, "system-id", buf)); 956398a0bd5SDavid Gibson } 957398a0bd5SDavid Gibson g_free(buf); 958398a0bd5SDavid Gibson 959398a0bd5SDavid Gibson if (qemu_get_vm_name()) { 960398a0bd5SDavid Gibson _FDT(fdt_setprop_string(fdt, 0, "ibm,partition-name", 961398a0bd5SDavid Gibson qemu_get_vm_name())); 962398a0bd5SDavid Gibson } 963398a0bd5SDavid Gibson 964398a0bd5SDavid Gibson _FDT(fdt_setprop_cell(fdt, 0, "#address-cells", 2)); 965398a0bd5SDavid Gibson _FDT(fdt_setprop_cell(fdt, 0, "#size-cells", 2)); 96653018216SPaolo Bonzini 9679b9a1908SDavid Gibson /* /interrupt controller */ 9689b9a1908SDavid Gibson spapr_dt_xics(spapr->xics, fdt, PHANDLE_XICP); 9699b9a1908SDavid Gibson 970e8f986fcSBharata B Rao ret = spapr_populate_memory(spapr, fdt); 971e8f986fcSBharata B Rao if (ret < 0) { 972ce9863b7SCédric Le Goater error_report("couldn't setup memory nodes in fdt"); 973e8f986fcSBharata B Rao exit(1); 97453018216SPaolo Bonzini } 97553018216SPaolo Bonzini 976bf5a6696SDavid Gibson /* /vdevice */ 977bf5a6696SDavid Gibson spapr_dt_vdevice(spapr->vio_bus, fdt); 97853018216SPaolo Bonzini 9794d9392beSThomas Huth if (object_resolve_path_type("", TYPE_SPAPR_RNG, NULL)) { 9804d9392beSThomas Huth ret = spapr_rng_populate_dt(fdt); 9814d9392beSThomas Huth if (ret < 0) { 982ce9863b7SCédric Le Goater error_report("could not set up rng device in the fdt"); 9834d9392beSThomas Huth exit(1); 9844d9392beSThomas Huth } 9854d9392beSThomas Huth } 9864d9392beSThomas Huth 98753018216SPaolo Bonzini QLIST_FOREACH(phb, &spapr->phbs, list) { 98853018216SPaolo Bonzini ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt); 98953018216SPaolo Bonzini if (ret < 0) { 990da34fed7SThomas Huth error_report("couldn't setup PCI devices in fdt"); 99153018216SPaolo Bonzini exit(1); 99253018216SPaolo Bonzini } 993da34fed7SThomas Huth } 99453018216SPaolo Bonzini 9950da6f3feSBharata B Rao /* cpus */ 9960da6f3feSBharata B Rao spapr_populate_cpus_dt_node(fdt, spapr); 99753018216SPaolo Bonzini 998c20d332aSBharata B Rao if (smc->dr_lmb_enabled) { 999c20d332aSBharata B Rao _FDT(spapr_drc_populate_dt(fdt, 0, NULL, SPAPR_DR_CONNECTOR_TYPE_LMB)); 1000c20d332aSBharata B Rao } 1001c20d332aSBharata B Rao 1002c5514d0eSIgor Mammedov if (mc->has_hotpluggable_cpus) { 1003af81cf32SBharata B Rao int offset = fdt_path_offset(fdt, "/cpus"); 1004af81cf32SBharata B Rao ret = spapr_drc_populate_dt(fdt, offset, NULL, 1005af81cf32SBharata B Rao SPAPR_DR_CONNECTOR_TYPE_CPU); 1006af81cf32SBharata B Rao if (ret < 0) { 1007af81cf32SBharata B Rao error_report("Couldn't set up CPU DR device tree properties"); 1008af81cf32SBharata B Rao exit(1); 1009af81cf32SBharata B Rao } 1010af81cf32SBharata B Rao } 1011af81cf32SBharata B Rao 1012ffb1e275SDavid Gibson /* /event-sources */ 1013ffbb1705SMichael Roth spapr_dt_events(spapr, fdt); 1014ffb1e275SDavid Gibson 10153f5dabceSDavid Gibson /* /rtas */ 10163f5dabceSDavid Gibson spapr_dt_rtas(spapr, fdt); 10173f5dabceSDavid Gibson 10187c866c6aSDavid Gibson /* /chosen */ 10197c866c6aSDavid Gibson spapr_dt_chosen(spapr, fdt); 1020cf6e5223SDavid Gibson 1021fca5f2dcSDavid Gibson /* /hypervisor */ 1022fca5f2dcSDavid Gibson if (kvm_enabled()) { 1023fca5f2dcSDavid Gibson spapr_dt_hypervisor(spapr, fdt); 1024fca5f2dcSDavid Gibson } 1025fca5f2dcSDavid Gibson 1026cf6e5223SDavid Gibson /* Build memory reserve map */ 1027cf6e5223SDavid Gibson if (spapr->kernel_size) { 1028cf6e5223SDavid Gibson _FDT((fdt_add_mem_rsv(fdt, KERNEL_LOAD_ADDR, spapr->kernel_size))); 1029cf6e5223SDavid Gibson } 1030cf6e5223SDavid Gibson if (spapr->initrd_size) { 1031cf6e5223SDavid Gibson _FDT((fdt_add_mem_rsv(fdt, spapr->initrd_base, spapr->initrd_size))); 1032cf6e5223SDavid Gibson } 1033cf6e5223SDavid Gibson 10346787d27bSMichael Roth /* ibm,client-architecture-support updates */ 10356787d27bSMichael Roth ret = spapr_dt_cas_updates(spapr, fdt, spapr->ov5_cas); 10366787d27bSMichael Roth if (ret < 0) { 10376787d27bSMichael Roth error_report("couldn't setup CAS properties fdt"); 10386787d27bSMichael Roth exit(1); 10396787d27bSMichael Roth } 10406787d27bSMichael Roth 1041997b6cfcSDavid Gibson return fdt; 104253018216SPaolo Bonzini } 104353018216SPaolo Bonzini 104453018216SPaolo Bonzini static uint64_t translate_kernel_address(void *opaque, uint64_t addr) 104553018216SPaolo Bonzini { 104653018216SPaolo Bonzini return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR; 104753018216SPaolo Bonzini } 104853018216SPaolo Bonzini 10491d1be34dSDavid Gibson static void emulate_spapr_hypercall(PPCVirtualHypervisor *vhyp, 10501d1be34dSDavid Gibson PowerPCCPU *cpu) 105153018216SPaolo Bonzini { 105253018216SPaolo Bonzini CPUPPCState *env = &cpu->env; 105353018216SPaolo Bonzini 10548d04fb55SJan Kiszka /* The TCG path should also be holding the BQL at this point */ 10558d04fb55SJan Kiszka g_assert(qemu_mutex_iothread_locked()); 10568d04fb55SJan Kiszka 105753018216SPaolo Bonzini if (msr_pr) { 105853018216SPaolo Bonzini hcall_dprintf("Hypercall made with MSR[PR]=1\n"); 105953018216SPaolo Bonzini env->gpr[3] = H_PRIVILEGE; 106053018216SPaolo Bonzini } else { 106153018216SPaolo Bonzini env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]); 106253018216SPaolo Bonzini } 106353018216SPaolo Bonzini } 106453018216SPaolo Bonzini 1065e6b8fd24SSamuel Mendoza-Jonas #define HPTE(_table, _i) (void *)(((uint64_t *)(_table)) + ((_i) * 2)) 1066e6b8fd24SSamuel Mendoza-Jonas #define HPTE_VALID(_hpte) (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID) 1067e6b8fd24SSamuel Mendoza-Jonas #define HPTE_DIRTY(_hpte) (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY) 1068e6b8fd24SSamuel Mendoza-Jonas #define CLEAN_HPTE(_hpte) ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY)) 1069e6b8fd24SSamuel Mendoza-Jonas #define DIRTY_HPTE(_hpte) ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY)) 1070e6b8fd24SSamuel Mendoza-Jonas 1071715c5407SDavid Gibson /* 1072715c5407SDavid Gibson * Get the fd to access the kernel htab, re-opening it if necessary 1073715c5407SDavid Gibson */ 1074715c5407SDavid Gibson static int get_htab_fd(sPAPRMachineState *spapr) 1075715c5407SDavid Gibson { 1076715c5407SDavid Gibson if (spapr->htab_fd >= 0) { 1077715c5407SDavid Gibson return spapr->htab_fd; 1078715c5407SDavid Gibson } 1079715c5407SDavid Gibson 1080715c5407SDavid Gibson spapr->htab_fd = kvmppc_get_htab_fd(false); 1081715c5407SDavid Gibson if (spapr->htab_fd < 0) { 1082715c5407SDavid Gibson error_report("Unable to open fd for reading hash table from KVM: %s", 1083715c5407SDavid Gibson strerror(errno)); 1084715c5407SDavid Gibson } 1085715c5407SDavid Gibson 1086715c5407SDavid Gibson return spapr->htab_fd; 1087715c5407SDavid Gibson } 1088715c5407SDavid Gibson 1089715c5407SDavid Gibson static void close_htab_fd(sPAPRMachineState *spapr) 1090715c5407SDavid Gibson { 1091715c5407SDavid Gibson if (spapr->htab_fd >= 0) { 1092715c5407SDavid Gibson close(spapr->htab_fd); 1093715c5407SDavid Gibson } 1094715c5407SDavid Gibson spapr->htab_fd = -1; 1095715c5407SDavid Gibson } 1096715c5407SDavid Gibson 1097e57ca75cSDavid Gibson static hwaddr spapr_hpt_mask(PPCVirtualHypervisor *vhyp) 1098e57ca75cSDavid Gibson { 1099e57ca75cSDavid Gibson sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); 1100e57ca75cSDavid Gibson 1101e57ca75cSDavid Gibson return HTAB_SIZE(spapr) / HASH_PTEG_SIZE_64 - 1; 1102e57ca75cSDavid Gibson } 1103e57ca75cSDavid Gibson 1104e57ca75cSDavid Gibson static const ppc_hash_pte64_t *spapr_map_hptes(PPCVirtualHypervisor *vhyp, 1105e57ca75cSDavid Gibson hwaddr ptex, int n) 1106e57ca75cSDavid Gibson { 1107e57ca75cSDavid Gibson sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); 1108e57ca75cSDavid Gibson hwaddr pte_offset = ptex * HASH_PTE_SIZE_64; 1109e57ca75cSDavid Gibson 1110e57ca75cSDavid Gibson if (!spapr->htab) { 1111e57ca75cSDavid Gibson /* 1112e57ca75cSDavid Gibson * HTAB is controlled by KVM. Fetch into temporary buffer 1113e57ca75cSDavid Gibson */ 1114e57ca75cSDavid Gibson ppc_hash_pte64_t *hptes = g_malloc(n * HASH_PTE_SIZE_64); 1115e57ca75cSDavid Gibson kvmppc_read_hptes(hptes, ptex, n); 1116e57ca75cSDavid Gibson return hptes; 1117e57ca75cSDavid Gibson } 1118e57ca75cSDavid Gibson 1119e57ca75cSDavid Gibson /* 1120e57ca75cSDavid Gibson * HTAB is controlled by QEMU. Just point to the internally 1121e57ca75cSDavid Gibson * accessible PTEG. 1122e57ca75cSDavid Gibson */ 1123e57ca75cSDavid Gibson return (const ppc_hash_pte64_t *)(spapr->htab + pte_offset); 1124e57ca75cSDavid Gibson } 1125e57ca75cSDavid Gibson 1126e57ca75cSDavid Gibson static void spapr_unmap_hptes(PPCVirtualHypervisor *vhyp, 1127e57ca75cSDavid Gibson const ppc_hash_pte64_t *hptes, 1128e57ca75cSDavid Gibson hwaddr ptex, int n) 1129e57ca75cSDavid Gibson { 1130e57ca75cSDavid Gibson sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); 1131e57ca75cSDavid Gibson 1132e57ca75cSDavid Gibson if (!spapr->htab) { 1133e57ca75cSDavid Gibson g_free((void *)hptes); 1134e57ca75cSDavid Gibson } 1135e57ca75cSDavid Gibson 1136e57ca75cSDavid Gibson /* Nothing to do for qemu managed HPT */ 1137e57ca75cSDavid Gibson } 1138e57ca75cSDavid Gibson 1139e57ca75cSDavid Gibson static void spapr_store_hpte(PPCVirtualHypervisor *vhyp, hwaddr ptex, 1140e57ca75cSDavid Gibson uint64_t pte0, uint64_t pte1) 1141e57ca75cSDavid Gibson { 1142e57ca75cSDavid Gibson sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); 1143e57ca75cSDavid Gibson hwaddr offset = ptex * HASH_PTE_SIZE_64; 1144e57ca75cSDavid Gibson 1145e57ca75cSDavid Gibson if (!spapr->htab) { 1146e57ca75cSDavid Gibson kvmppc_write_hpte(ptex, pte0, pte1); 1147e57ca75cSDavid Gibson } else { 1148e57ca75cSDavid Gibson stq_p(spapr->htab + offset, pte0); 1149e57ca75cSDavid Gibson stq_p(spapr->htab + offset + HASH_PTE_SIZE_64 / 2, pte1); 1150e57ca75cSDavid Gibson } 1151e57ca75cSDavid Gibson } 1152e57ca75cSDavid Gibson 11538dfe8e7fSDavid Gibson static int spapr_hpt_shift_for_ramsize(uint64_t ramsize) 11548dfe8e7fSDavid Gibson { 11558dfe8e7fSDavid Gibson int shift; 11568dfe8e7fSDavid Gibson 11578dfe8e7fSDavid Gibson /* We aim for a hash table of size 1/128 the size of RAM (rounded 11588dfe8e7fSDavid Gibson * up). The PAPR recommendation is actually 1/64 of RAM size, but 11598dfe8e7fSDavid Gibson * that's much more than is needed for Linux guests */ 11608dfe8e7fSDavid Gibson shift = ctz64(pow2ceil(ramsize)) - 7; 11618dfe8e7fSDavid Gibson shift = MAX(shift, 18); /* Minimum architected size */ 11628dfe8e7fSDavid Gibson shift = MIN(shift, 46); /* Maximum architected size */ 11638dfe8e7fSDavid Gibson return shift; 11648dfe8e7fSDavid Gibson } 11658dfe8e7fSDavid Gibson 1166c5f54f3eSDavid Gibson static void spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift, 1167c5f54f3eSDavid Gibson Error **errp) 116853018216SPaolo Bonzini { 1169c5f54f3eSDavid Gibson long rc; 117053018216SPaolo Bonzini 1171c5f54f3eSDavid Gibson /* Clean up any HPT info from a previous boot */ 1172c5f54f3eSDavid Gibson g_free(spapr->htab); 1173c5f54f3eSDavid Gibson spapr->htab = NULL; 1174c5f54f3eSDavid Gibson spapr->htab_shift = 0; 1175c5f54f3eSDavid Gibson close_htab_fd(spapr); 117653018216SPaolo Bonzini 1177c5f54f3eSDavid Gibson rc = kvmppc_reset_htab(shift); 1178c5f54f3eSDavid Gibson if (rc < 0) { 1179c5f54f3eSDavid Gibson /* kernel-side HPT needed, but couldn't allocate one */ 1180c5f54f3eSDavid Gibson error_setg_errno(errp, errno, 1181c5f54f3eSDavid Gibson "Failed to allocate KVM HPT of order %d (try smaller maxmem?)", 1182c5f54f3eSDavid Gibson shift); 1183c5f54f3eSDavid Gibson /* This is almost certainly fatal, but if the caller really 1184c5f54f3eSDavid Gibson * wants to carry on with shift == 0, it's welcome to try */ 1185c5f54f3eSDavid Gibson } else if (rc > 0) { 1186c5f54f3eSDavid Gibson /* kernel-side HPT allocated */ 1187c5f54f3eSDavid Gibson if (rc != shift) { 1188c5f54f3eSDavid Gibson error_setg(errp, 1189c5f54f3eSDavid Gibson "Requested order %d HPT, but kernel allocated order %ld (try smaller maxmem?)", 1190c5f54f3eSDavid Gibson shift, rc); 11917735fedaSBharata B Rao } 11927735fedaSBharata B Rao 119353018216SPaolo Bonzini spapr->htab_shift = shift; 1194c18ad9a5SDavid Gibson spapr->htab = NULL; 1195b817772aSBharata B Rao } else { 1196c5f54f3eSDavid Gibson /* kernel-side HPT not needed, allocate in userspace instead */ 1197c5f54f3eSDavid Gibson size_t size = 1ULL << shift; 1198c5f54f3eSDavid Gibson int i; 119901a57972SSamuel Mendoza-Jonas 1200c5f54f3eSDavid Gibson spapr->htab = qemu_memalign(size, size); 1201c5f54f3eSDavid Gibson if (!spapr->htab) { 1202c5f54f3eSDavid Gibson error_setg_errno(errp, errno, 1203c5f54f3eSDavid Gibson "Could not allocate HPT of order %d", shift); 1204c5f54f3eSDavid Gibson return; 1205b817772aSBharata B Rao } 1206b817772aSBharata B Rao 1207c5f54f3eSDavid Gibson memset(spapr->htab, 0, size); 1208c5f54f3eSDavid Gibson spapr->htab_shift = shift; 1209b817772aSBharata B Rao 1210c5f54f3eSDavid Gibson for (i = 0; i < size / HASH_PTE_SIZE_64; i++) { 1211c5f54f3eSDavid Gibson DIRTY_HPTE(HPTE(spapr->htab, i)); 12127735fedaSBharata B Rao } 121353018216SPaolo Bonzini } 121453018216SPaolo Bonzini } 121553018216SPaolo Bonzini 12164f01a637SDavid Gibson static void find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque) 12179e3f9733SAlexander Graf { 12189e3f9733SAlexander Graf bool matched = false; 12199e3f9733SAlexander Graf 12209e3f9733SAlexander Graf if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) { 12219e3f9733SAlexander Graf matched = true; 12229e3f9733SAlexander Graf } 12239e3f9733SAlexander Graf 12249e3f9733SAlexander Graf if (!matched) { 12259e3f9733SAlexander Graf error_report("Device %s is not supported by this machine yet.", 12269e3f9733SAlexander Graf qdev_fw_name(DEVICE(sbdev))); 12279e3f9733SAlexander Graf exit(1); 12289e3f9733SAlexander Graf } 12299e3f9733SAlexander Graf } 12309e3f9733SAlexander Graf 123153018216SPaolo Bonzini static void ppc_spapr_reset(void) 123253018216SPaolo Bonzini { 1233c5f54f3eSDavid Gibson MachineState *machine = MACHINE(qdev_get_machine()); 1234c5f54f3eSDavid Gibson sPAPRMachineState *spapr = SPAPR_MACHINE(machine); 1235182735efSAndreas Färber PowerPCCPU *first_ppc_cpu; 1236b7d1f77aSBenjamin Herrenschmidt uint32_t rtas_limit; 1237cae172abSDavid Gibson hwaddr rtas_addr, fdt_addr; 1238997b6cfcSDavid Gibson void *fdt; 1239997b6cfcSDavid Gibson int rc; 1240259186a7SAndreas Färber 12419e3f9733SAlexander Graf /* Check for unknown sysbus devices */ 12429e3f9733SAlexander Graf foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL); 12439e3f9733SAlexander Graf 1244c5f54f3eSDavid Gibson /* Allocate and/or reset the hash page table */ 1245c5f54f3eSDavid Gibson spapr_reallocate_hpt(spapr, 1246c5f54f3eSDavid Gibson spapr_hpt_shift_for_ramsize(machine->maxram_size), 1247c5f54f3eSDavid Gibson &error_fatal); 1248c5f54f3eSDavid Gibson 1249c5f54f3eSDavid Gibson /* Update the RMA size if necessary */ 1250c5f54f3eSDavid Gibson if (spapr->vrma_adjust) { 1251c5f54f3eSDavid Gibson spapr->rma_size = kvmppc_rma_size(spapr_node0_size(), 1252c5f54f3eSDavid Gibson spapr->htab_shift); 1253c5f54f3eSDavid Gibson } 125453018216SPaolo Bonzini 125553018216SPaolo Bonzini qemu_devices_reset(); 125653018216SPaolo Bonzini 1257b7d1f77aSBenjamin Herrenschmidt /* 1258b7d1f77aSBenjamin Herrenschmidt * We place the device tree and RTAS just below either the top of the RMA, 1259b7d1f77aSBenjamin Herrenschmidt * or just below 2GB, whichever is lowere, so that it can be 1260b7d1f77aSBenjamin Herrenschmidt * processed with 32-bit real mode code if necessary 1261b7d1f77aSBenjamin Herrenschmidt */ 1262b7d1f77aSBenjamin Herrenschmidt rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR); 1263cae172abSDavid Gibson rtas_addr = rtas_limit - RTAS_MAX_SIZE; 1264cae172abSDavid Gibson fdt_addr = rtas_addr - FDT_MAX_SIZE; 1265b7d1f77aSBenjamin Herrenschmidt 12666787d27bSMichael Roth /* if this reset wasn't generated by CAS, we should reset our 12676787d27bSMichael Roth * negotiated options and start from scratch */ 12686787d27bSMichael Roth if (!spapr->cas_reboot) { 12696787d27bSMichael Roth spapr_ovec_cleanup(spapr->ov5_cas); 12706787d27bSMichael Roth spapr->ov5_cas = spapr_ovec_new(); 12716787d27bSMichael Roth } 12726787d27bSMichael Roth 1273cae172abSDavid Gibson fdt = spapr_build_fdt(spapr, rtas_addr, spapr->rtas_size); 127453018216SPaolo Bonzini 12752cac78c1SDavid Gibson spapr_load_rtas(spapr, fdt, rtas_addr); 1276b7d1f77aSBenjamin Herrenschmidt 1277997b6cfcSDavid Gibson rc = fdt_pack(fdt); 1278997b6cfcSDavid Gibson 1279997b6cfcSDavid Gibson /* Should only fail if we've built a corrupted tree */ 1280997b6cfcSDavid Gibson assert(rc == 0); 1281997b6cfcSDavid Gibson 1282997b6cfcSDavid Gibson if (fdt_totalsize(fdt) > FDT_MAX_SIZE) { 1283997b6cfcSDavid Gibson error_report("FDT too big ! 0x%x bytes (max is 0x%x)", 1284997b6cfcSDavid Gibson fdt_totalsize(fdt), FDT_MAX_SIZE); 1285997b6cfcSDavid Gibson exit(1); 1286997b6cfcSDavid Gibson } 1287997b6cfcSDavid Gibson 1288997b6cfcSDavid Gibson /* Load the fdt */ 1289997b6cfcSDavid Gibson qemu_fdt_dumpdtb(fdt, fdt_totalsize(fdt)); 1290cae172abSDavid Gibson cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt)); 1291997b6cfcSDavid Gibson g_free(fdt); 1292997b6cfcSDavid Gibson 129353018216SPaolo Bonzini /* Set up the entry state */ 1294182735efSAndreas Färber first_ppc_cpu = POWERPC_CPU(first_cpu); 1295cae172abSDavid Gibson first_ppc_cpu->env.gpr[3] = fdt_addr; 1296182735efSAndreas Färber first_ppc_cpu->env.gpr[5] = 0; 1297182735efSAndreas Färber first_cpu->halted = 0; 12981b718907SDavid Gibson first_ppc_cpu->env.nip = SPAPR_ENTRY_POINT; 129953018216SPaolo Bonzini 13006787d27bSMichael Roth spapr->cas_reboot = false; 130153018216SPaolo Bonzini } 130253018216SPaolo Bonzini 130328e02042SDavid Gibson static void spapr_create_nvram(sPAPRMachineState *spapr) 130453018216SPaolo Bonzini { 13052ff3de68SMarkus Armbruster DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram"); 13063978b863SPaolo Bonzini DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0); 130753018216SPaolo Bonzini 13083978b863SPaolo Bonzini if (dinfo) { 13096231a6daSMarkus Armbruster qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo), 13106231a6daSMarkus Armbruster &error_fatal); 131153018216SPaolo Bonzini } 131253018216SPaolo Bonzini 131353018216SPaolo Bonzini qdev_init_nofail(dev); 131453018216SPaolo Bonzini 131553018216SPaolo Bonzini spapr->nvram = (struct sPAPRNVRAM *)dev; 131653018216SPaolo Bonzini } 131753018216SPaolo Bonzini 131828e02042SDavid Gibson static void spapr_rtc_create(sPAPRMachineState *spapr) 131928df36a1SDavid Gibson { 132028df36a1SDavid Gibson DeviceState *dev = qdev_create(NULL, TYPE_SPAPR_RTC); 132128df36a1SDavid Gibson 132228df36a1SDavid Gibson qdev_init_nofail(dev); 132328df36a1SDavid Gibson spapr->rtc = dev; 132474e5ae28SDavid Gibson 132574e5ae28SDavid Gibson object_property_add_alias(qdev_get_machine(), "rtc-time", 132674e5ae28SDavid Gibson OBJECT(spapr->rtc), "date", NULL); 132728df36a1SDavid Gibson } 132828df36a1SDavid Gibson 132953018216SPaolo Bonzini /* Returns whether we want to use VGA or not */ 133014c6a894SDavid Gibson static bool spapr_vga_init(PCIBus *pci_bus, Error **errp) 133153018216SPaolo Bonzini { 133253018216SPaolo Bonzini switch (vga_interface_type) { 133353018216SPaolo Bonzini case VGA_NONE: 13347effdaa3SMark Wu return false; 13357effdaa3SMark Wu case VGA_DEVICE: 13367effdaa3SMark Wu return true; 133753018216SPaolo Bonzini case VGA_STD: 1338b798c190SBenjamin Herrenschmidt case VGA_VIRTIO: 133953018216SPaolo Bonzini return pci_vga_init(pci_bus) != NULL; 134053018216SPaolo Bonzini default: 134114c6a894SDavid Gibson error_setg(errp, 134214c6a894SDavid Gibson "Unsupported VGA mode, only -vga std or -vga virtio is supported"); 134314c6a894SDavid Gibson return false; 134453018216SPaolo Bonzini } 134553018216SPaolo Bonzini } 134653018216SPaolo Bonzini 1347880ae7deSDavid Gibson static int spapr_post_load(void *opaque, int version_id) 1348880ae7deSDavid Gibson { 134928e02042SDavid Gibson sPAPRMachineState *spapr = (sPAPRMachineState *)opaque; 1350880ae7deSDavid Gibson int err = 0; 1351880ae7deSDavid Gibson 1352631b22eaSStefan Weil /* In earlier versions, there was no separate qdev for the PAPR 1353880ae7deSDavid Gibson * RTC, so the RTC offset was stored directly in sPAPREnvironment. 1354880ae7deSDavid Gibson * So when migrating from those versions, poke the incoming offset 1355880ae7deSDavid Gibson * value into the RTC device */ 1356880ae7deSDavid Gibson if (version_id < 3) { 1357880ae7deSDavid Gibson err = spapr_rtc_import_offset(spapr->rtc, spapr->rtc_offset); 1358880ae7deSDavid Gibson } 1359880ae7deSDavid Gibson 1360880ae7deSDavid Gibson return err; 1361880ae7deSDavid Gibson } 1362880ae7deSDavid Gibson 1363880ae7deSDavid Gibson static bool version_before_3(void *opaque, int version_id) 1364880ae7deSDavid Gibson { 1365880ae7deSDavid Gibson return version_id < 3; 1366880ae7deSDavid Gibson } 1367880ae7deSDavid Gibson 136862ef3760SMichael Roth static bool spapr_ov5_cas_needed(void *opaque) 136962ef3760SMichael Roth { 137062ef3760SMichael Roth sPAPRMachineState *spapr = opaque; 137162ef3760SMichael Roth sPAPROptionVector *ov5_mask = spapr_ovec_new(); 137262ef3760SMichael Roth sPAPROptionVector *ov5_legacy = spapr_ovec_new(); 137362ef3760SMichael Roth sPAPROptionVector *ov5_removed = spapr_ovec_new(); 137462ef3760SMichael Roth bool cas_needed; 137562ef3760SMichael Roth 137662ef3760SMichael Roth /* Prior to the introduction of sPAPROptionVector, we had two option 137762ef3760SMichael Roth * vectors we dealt with: OV5_FORM1_AFFINITY, and OV5_DRCONF_MEMORY. 137862ef3760SMichael Roth * Both of these options encode machine topology into the device-tree 137962ef3760SMichael Roth * in such a way that the now-booted OS should still be able to interact 138062ef3760SMichael Roth * appropriately with QEMU regardless of what options were actually 138162ef3760SMichael Roth * negotiatied on the source side. 138262ef3760SMichael Roth * 138362ef3760SMichael Roth * As such, we can avoid migrating the CAS-negotiated options if these 138462ef3760SMichael Roth * are the only options available on the current machine/platform. 138562ef3760SMichael Roth * Since these are the only options available for pseries-2.7 and 138662ef3760SMichael Roth * earlier, this allows us to maintain old->new/new->old migration 138762ef3760SMichael Roth * compatibility. 138862ef3760SMichael Roth * 138962ef3760SMichael Roth * For QEMU 2.8+, there are additional CAS-negotiatable options available 139062ef3760SMichael Roth * via default pseries-2.8 machines and explicit command-line parameters. 139162ef3760SMichael Roth * Some of these options, like OV5_HP_EVT, *do* require QEMU to be aware 139262ef3760SMichael Roth * of the actual CAS-negotiated values to continue working properly. For 139362ef3760SMichael Roth * example, availability of memory unplug depends on knowing whether 139462ef3760SMichael Roth * OV5_HP_EVT was negotiated via CAS. 139562ef3760SMichael Roth * 139662ef3760SMichael Roth * Thus, for any cases where the set of available CAS-negotiatable 139762ef3760SMichael Roth * options extends beyond OV5_FORM1_AFFINITY and OV5_DRCONF_MEMORY, we 139862ef3760SMichael Roth * include the CAS-negotiated options in the migration stream. 139962ef3760SMichael Roth */ 140062ef3760SMichael Roth spapr_ovec_set(ov5_mask, OV5_FORM1_AFFINITY); 140162ef3760SMichael Roth spapr_ovec_set(ov5_mask, OV5_DRCONF_MEMORY); 140262ef3760SMichael Roth 140362ef3760SMichael Roth /* spapr_ovec_diff returns true if bits were removed. we avoid using 140462ef3760SMichael Roth * the mask itself since in the future it's possible "legacy" bits may be 140562ef3760SMichael Roth * removed via machine options, which could generate a false positive 140662ef3760SMichael Roth * that breaks migration. 140762ef3760SMichael Roth */ 140862ef3760SMichael Roth spapr_ovec_intersect(ov5_legacy, spapr->ov5, ov5_mask); 140962ef3760SMichael Roth cas_needed = spapr_ovec_diff(ov5_removed, spapr->ov5, ov5_legacy); 141062ef3760SMichael Roth 141162ef3760SMichael Roth spapr_ovec_cleanup(ov5_mask); 141262ef3760SMichael Roth spapr_ovec_cleanup(ov5_legacy); 141362ef3760SMichael Roth spapr_ovec_cleanup(ov5_removed); 141462ef3760SMichael Roth 141562ef3760SMichael Roth return cas_needed; 141662ef3760SMichael Roth } 141762ef3760SMichael Roth 141862ef3760SMichael Roth static const VMStateDescription vmstate_spapr_ov5_cas = { 141962ef3760SMichael Roth .name = "spapr_option_vector_ov5_cas", 142062ef3760SMichael Roth .version_id = 1, 142162ef3760SMichael Roth .minimum_version_id = 1, 142262ef3760SMichael Roth .needed = spapr_ov5_cas_needed, 142362ef3760SMichael Roth .fields = (VMStateField[]) { 142462ef3760SMichael Roth VMSTATE_STRUCT_POINTER_V(ov5_cas, sPAPRMachineState, 1, 142562ef3760SMichael Roth vmstate_spapr_ovec, sPAPROptionVector), 142662ef3760SMichael Roth VMSTATE_END_OF_LIST() 142762ef3760SMichael Roth }, 142862ef3760SMichael Roth }; 142962ef3760SMichael Roth 14304be21d56SDavid Gibson static const VMStateDescription vmstate_spapr = { 14314be21d56SDavid Gibson .name = "spapr", 1432880ae7deSDavid Gibson .version_id = 3, 14334be21d56SDavid Gibson .minimum_version_id = 1, 1434880ae7deSDavid Gibson .post_load = spapr_post_load, 14354be21d56SDavid Gibson .fields = (VMStateField[]) { 1436880ae7deSDavid Gibson /* used to be @next_irq */ 1437880ae7deSDavid Gibson VMSTATE_UNUSED_BUFFER(version_before_3, 0, 4), 14384be21d56SDavid Gibson 14394be21d56SDavid Gibson /* RTC offset */ 144028e02042SDavid Gibson VMSTATE_UINT64_TEST(rtc_offset, sPAPRMachineState, version_before_3), 1441880ae7deSDavid Gibson 144228e02042SDavid Gibson VMSTATE_PPC_TIMEBASE_V(tb, sPAPRMachineState, 2), 14434be21d56SDavid Gibson VMSTATE_END_OF_LIST() 14444be21d56SDavid Gibson }, 144562ef3760SMichael Roth .subsections = (const VMStateDescription*[]) { 144662ef3760SMichael Roth &vmstate_spapr_ov5_cas, 144762ef3760SMichael Roth NULL 144862ef3760SMichael Roth } 14494be21d56SDavid Gibson }; 14504be21d56SDavid Gibson 14514be21d56SDavid Gibson static int htab_save_setup(QEMUFile *f, void *opaque) 14524be21d56SDavid Gibson { 145328e02042SDavid Gibson sPAPRMachineState *spapr = opaque; 14544be21d56SDavid Gibson 14554be21d56SDavid Gibson /* "Iteration" header */ 14564be21d56SDavid Gibson qemu_put_be32(f, spapr->htab_shift); 14574be21d56SDavid Gibson 1458e68cb8b4SAlexey Kardashevskiy if (spapr->htab) { 1459e68cb8b4SAlexey Kardashevskiy spapr->htab_save_index = 0; 1460e68cb8b4SAlexey Kardashevskiy spapr->htab_first_pass = true; 1461e68cb8b4SAlexey Kardashevskiy } else { 1462e68cb8b4SAlexey Kardashevskiy assert(kvm_enabled()); 14634be21d56SDavid Gibson } 14644be21d56SDavid Gibson 1465e68cb8b4SAlexey Kardashevskiy 1466e68cb8b4SAlexey Kardashevskiy return 0; 1467e68cb8b4SAlexey Kardashevskiy } 14684be21d56SDavid Gibson 146928e02042SDavid Gibson static void htab_save_first_pass(QEMUFile *f, sPAPRMachineState *spapr, 14704be21d56SDavid Gibson int64_t max_ns) 14714be21d56SDavid Gibson { 1472378bc217SDavid Gibson bool has_timeout = max_ns != -1; 14734be21d56SDavid Gibson int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; 14744be21d56SDavid Gibson int index = spapr->htab_save_index; 1475bc72ad67SAlex Bligh int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); 14764be21d56SDavid Gibson 14774be21d56SDavid Gibson assert(spapr->htab_first_pass); 14784be21d56SDavid Gibson 14794be21d56SDavid Gibson do { 14804be21d56SDavid Gibson int chunkstart; 14814be21d56SDavid Gibson 14824be21d56SDavid Gibson /* Consume invalid HPTEs */ 14834be21d56SDavid Gibson while ((index < htabslots) 14844be21d56SDavid Gibson && !HPTE_VALID(HPTE(spapr->htab, index))) { 14854be21d56SDavid Gibson index++; 14864be21d56SDavid Gibson CLEAN_HPTE(HPTE(spapr->htab, index)); 14874be21d56SDavid Gibson } 14884be21d56SDavid Gibson 14894be21d56SDavid Gibson /* Consume valid HPTEs */ 14904be21d56SDavid Gibson chunkstart = index; 1491338c25b6SSamuel Mendoza-Jonas while ((index < htabslots) && (index - chunkstart < USHRT_MAX) 14924be21d56SDavid Gibson && HPTE_VALID(HPTE(spapr->htab, index))) { 14934be21d56SDavid Gibson index++; 14944be21d56SDavid Gibson CLEAN_HPTE(HPTE(spapr->htab, index)); 14954be21d56SDavid Gibson } 14964be21d56SDavid Gibson 14974be21d56SDavid Gibson if (index > chunkstart) { 14984be21d56SDavid Gibson int n_valid = index - chunkstart; 14994be21d56SDavid Gibson 15004be21d56SDavid Gibson qemu_put_be32(f, chunkstart); 15014be21d56SDavid Gibson qemu_put_be16(f, n_valid); 15024be21d56SDavid Gibson qemu_put_be16(f, 0); 15034be21d56SDavid Gibson qemu_put_buffer(f, HPTE(spapr->htab, chunkstart), 15044be21d56SDavid Gibson HASH_PTE_SIZE_64 * n_valid); 15054be21d56SDavid Gibson 1506378bc217SDavid Gibson if (has_timeout && 1507378bc217SDavid Gibson (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) { 15084be21d56SDavid Gibson break; 15094be21d56SDavid Gibson } 15104be21d56SDavid Gibson } 15114be21d56SDavid Gibson } while ((index < htabslots) && !qemu_file_rate_limit(f)); 15124be21d56SDavid Gibson 15134be21d56SDavid Gibson if (index >= htabslots) { 15144be21d56SDavid Gibson assert(index == htabslots); 15154be21d56SDavid Gibson index = 0; 15164be21d56SDavid Gibson spapr->htab_first_pass = false; 15174be21d56SDavid Gibson } 15184be21d56SDavid Gibson spapr->htab_save_index = index; 15194be21d56SDavid Gibson } 15204be21d56SDavid Gibson 152128e02042SDavid Gibson static int htab_save_later_pass(QEMUFile *f, sPAPRMachineState *spapr, 15224be21d56SDavid Gibson int64_t max_ns) 15234be21d56SDavid Gibson { 15244be21d56SDavid Gibson bool final = max_ns < 0; 15254be21d56SDavid Gibson int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; 15264be21d56SDavid Gibson int examined = 0, sent = 0; 15274be21d56SDavid Gibson int index = spapr->htab_save_index; 1528bc72ad67SAlex Bligh int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); 15294be21d56SDavid Gibson 15304be21d56SDavid Gibson assert(!spapr->htab_first_pass); 15314be21d56SDavid Gibson 15324be21d56SDavid Gibson do { 15334be21d56SDavid Gibson int chunkstart, invalidstart; 15344be21d56SDavid Gibson 15354be21d56SDavid Gibson /* Consume non-dirty HPTEs */ 15364be21d56SDavid Gibson while ((index < htabslots) 15374be21d56SDavid Gibson && !HPTE_DIRTY(HPTE(spapr->htab, index))) { 15384be21d56SDavid Gibson index++; 15394be21d56SDavid Gibson examined++; 15404be21d56SDavid Gibson } 15414be21d56SDavid Gibson 15424be21d56SDavid Gibson chunkstart = index; 15434be21d56SDavid Gibson /* Consume valid dirty HPTEs */ 1544338c25b6SSamuel Mendoza-Jonas while ((index < htabslots) && (index - chunkstart < USHRT_MAX) 15454be21d56SDavid Gibson && HPTE_DIRTY(HPTE(spapr->htab, index)) 15464be21d56SDavid Gibson && HPTE_VALID(HPTE(spapr->htab, index))) { 15474be21d56SDavid Gibson CLEAN_HPTE(HPTE(spapr->htab, index)); 15484be21d56SDavid Gibson index++; 15494be21d56SDavid Gibson examined++; 15504be21d56SDavid Gibson } 15514be21d56SDavid Gibson 15524be21d56SDavid Gibson invalidstart = index; 15534be21d56SDavid Gibson /* Consume invalid dirty HPTEs */ 1554338c25b6SSamuel Mendoza-Jonas while ((index < htabslots) && (index - invalidstart < USHRT_MAX) 15554be21d56SDavid Gibson && HPTE_DIRTY(HPTE(spapr->htab, index)) 15564be21d56SDavid Gibson && !HPTE_VALID(HPTE(spapr->htab, index))) { 15574be21d56SDavid Gibson CLEAN_HPTE(HPTE(spapr->htab, index)); 15584be21d56SDavid Gibson index++; 15594be21d56SDavid Gibson examined++; 15604be21d56SDavid Gibson } 15614be21d56SDavid Gibson 15624be21d56SDavid Gibson if (index > chunkstart) { 15634be21d56SDavid Gibson int n_valid = invalidstart - chunkstart; 15644be21d56SDavid Gibson int n_invalid = index - invalidstart; 15654be21d56SDavid Gibson 15664be21d56SDavid Gibson qemu_put_be32(f, chunkstart); 15674be21d56SDavid Gibson qemu_put_be16(f, n_valid); 15684be21d56SDavid Gibson qemu_put_be16(f, n_invalid); 15694be21d56SDavid Gibson qemu_put_buffer(f, HPTE(spapr->htab, chunkstart), 15704be21d56SDavid Gibson HASH_PTE_SIZE_64 * n_valid); 15714be21d56SDavid Gibson sent += index - chunkstart; 15724be21d56SDavid Gibson 1573bc72ad67SAlex Bligh if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) { 15744be21d56SDavid Gibson break; 15754be21d56SDavid Gibson } 15764be21d56SDavid Gibson } 15774be21d56SDavid Gibson 15784be21d56SDavid Gibson if (examined >= htabslots) { 15794be21d56SDavid Gibson break; 15804be21d56SDavid Gibson } 15814be21d56SDavid Gibson 15824be21d56SDavid Gibson if (index >= htabslots) { 15834be21d56SDavid Gibson assert(index == htabslots); 15844be21d56SDavid Gibson index = 0; 15854be21d56SDavid Gibson } 15864be21d56SDavid Gibson } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final)); 15874be21d56SDavid Gibson 15884be21d56SDavid Gibson if (index >= htabslots) { 15894be21d56SDavid Gibson assert(index == htabslots); 15904be21d56SDavid Gibson index = 0; 15914be21d56SDavid Gibson } 15924be21d56SDavid Gibson 15934be21d56SDavid Gibson spapr->htab_save_index = index; 15944be21d56SDavid Gibson 1595e68cb8b4SAlexey Kardashevskiy return (examined >= htabslots) && (sent == 0) ? 1 : 0; 15964be21d56SDavid Gibson } 15974be21d56SDavid Gibson 1598e68cb8b4SAlexey Kardashevskiy #define MAX_ITERATION_NS 5000000 /* 5 ms */ 1599e68cb8b4SAlexey Kardashevskiy #define MAX_KVM_BUF_SIZE 2048 1600e68cb8b4SAlexey Kardashevskiy 16014be21d56SDavid Gibson static int htab_save_iterate(QEMUFile *f, void *opaque) 16024be21d56SDavid Gibson { 160328e02042SDavid Gibson sPAPRMachineState *spapr = opaque; 1604715c5407SDavid Gibson int fd; 1605e68cb8b4SAlexey Kardashevskiy int rc = 0; 16064be21d56SDavid Gibson 16074be21d56SDavid Gibson /* Iteration header */ 16084be21d56SDavid Gibson qemu_put_be32(f, 0); 16094be21d56SDavid Gibson 1610e68cb8b4SAlexey Kardashevskiy if (!spapr->htab) { 1611e68cb8b4SAlexey Kardashevskiy assert(kvm_enabled()); 1612e68cb8b4SAlexey Kardashevskiy 1613715c5407SDavid Gibson fd = get_htab_fd(spapr); 1614715c5407SDavid Gibson if (fd < 0) { 1615715c5407SDavid Gibson return fd; 161601a57972SSamuel Mendoza-Jonas } 161701a57972SSamuel Mendoza-Jonas 1618715c5407SDavid Gibson rc = kvmppc_save_htab(f, fd, MAX_KVM_BUF_SIZE, MAX_ITERATION_NS); 1619e68cb8b4SAlexey Kardashevskiy if (rc < 0) { 1620e68cb8b4SAlexey Kardashevskiy return rc; 1621e68cb8b4SAlexey Kardashevskiy } 1622e68cb8b4SAlexey Kardashevskiy } else if (spapr->htab_first_pass) { 16234be21d56SDavid Gibson htab_save_first_pass(f, spapr, MAX_ITERATION_NS); 16244be21d56SDavid Gibson } else { 1625e68cb8b4SAlexey Kardashevskiy rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS); 16264be21d56SDavid Gibson } 16274be21d56SDavid Gibson 16284be21d56SDavid Gibson /* End marker */ 16294be21d56SDavid Gibson qemu_put_be32(f, 0); 16304be21d56SDavid Gibson qemu_put_be16(f, 0); 16314be21d56SDavid Gibson qemu_put_be16(f, 0); 16324be21d56SDavid Gibson 1633e68cb8b4SAlexey Kardashevskiy return rc; 16344be21d56SDavid Gibson } 16354be21d56SDavid Gibson 16364be21d56SDavid Gibson static int htab_save_complete(QEMUFile *f, void *opaque) 16374be21d56SDavid Gibson { 163828e02042SDavid Gibson sPAPRMachineState *spapr = opaque; 1639715c5407SDavid Gibson int fd; 16404be21d56SDavid Gibson 16414be21d56SDavid Gibson /* Iteration header */ 16424be21d56SDavid Gibson qemu_put_be32(f, 0); 16434be21d56SDavid Gibson 1644e68cb8b4SAlexey Kardashevskiy if (!spapr->htab) { 1645e68cb8b4SAlexey Kardashevskiy int rc; 1646e68cb8b4SAlexey Kardashevskiy 1647e68cb8b4SAlexey Kardashevskiy assert(kvm_enabled()); 1648e68cb8b4SAlexey Kardashevskiy 1649715c5407SDavid Gibson fd = get_htab_fd(spapr); 1650715c5407SDavid Gibson if (fd < 0) { 1651715c5407SDavid Gibson return fd; 165201a57972SSamuel Mendoza-Jonas } 165301a57972SSamuel Mendoza-Jonas 1654715c5407SDavid Gibson rc = kvmppc_save_htab(f, fd, MAX_KVM_BUF_SIZE, -1); 1655e68cb8b4SAlexey Kardashevskiy if (rc < 0) { 1656e68cb8b4SAlexey Kardashevskiy return rc; 1657e68cb8b4SAlexey Kardashevskiy } 1658e68cb8b4SAlexey Kardashevskiy } else { 1659378bc217SDavid Gibson if (spapr->htab_first_pass) { 1660378bc217SDavid Gibson htab_save_first_pass(f, spapr, -1); 1661378bc217SDavid Gibson } 16624be21d56SDavid Gibson htab_save_later_pass(f, spapr, -1); 1663e68cb8b4SAlexey Kardashevskiy } 16644be21d56SDavid Gibson 16654be21d56SDavid Gibson /* End marker */ 16664be21d56SDavid Gibson qemu_put_be32(f, 0); 16674be21d56SDavid Gibson qemu_put_be16(f, 0); 16684be21d56SDavid Gibson qemu_put_be16(f, 0); 16694be21d56SDavid Gibson 16704be21d56SDavid Gibson return 0; 16714be21d56SDavid Gibson } 16724be21d56SDavid Gibson 16734be21d56SDavid Gibson static int htab_load(QEMUFile *f, void *opaque, int version_id) 16744be21d56SDavid Gibson { 167528e02042SDavid Gibson sPAPRMachineState *spapr = opaque; 16764be21d56SDavid Gibson uint32_t section_hdr; 1677e68cb8b4SAlexey Kardashevskiy int fd = -1; 16784be21d56SDavid Gibson 16794be21d56SDavid Gibson if (version_id < 1 || version_id > 1) { 168098a5d100SDavid Gibson error_report("htab_load() bad version"); 16814be21d56SDavid Gibson return -EINVAL; 16824be21d56SDavid Gibson } 16834be21d56SDavid Gibson 16844be21d56SDavid Gibson section_hdr = qemu_get_be32(f); 16854be21d56SDavid Gibson 16864be21d56SDavid Gibson if (section_hdr) { 16879897e462SGreg Kurz Error *local_err = NULL; 1688c5f54f3eSDavid Gibson 1689c5f54f3eSDavid Gibson /* First section gives the htab size */ 1690c5f54f3eSDavid Gibson spapr_reallocate_hpt(spapr, section_hdr, &local_err); 1691c5f54f3eSDavid Gibson if (local_err) { 1692c5f54f3eSDavid Gibson error_report_err(local_err); 16934be21d56SDavid Gibson return -EINVAL; 16944be21d56SDavid Gibson } 16954be21d56SDavid Gibson return 0; 16964be21d56SDavid Gibson } 16974be21d56SDavid Gibson 1698e68cb8b4SAlexey Kardashevskiy if (!spapr->htab) { 1699e68cb8b4SAlexey Kardashevskiy assert(kvm_enabled()); 1700e68cb8b4SAlexey Kardashevskiy 1701e68cb8b4SAlexey Kardashevskiy fd = kvmppc_get_htab_fd(true); 1702e68cb8b4SAlexey Kardashevskiy if (fd < 0) { 170398a5d100SDavid Gibson error_report("Unable to open fd to restore KVM hash table: %s", 1704e68cb8b4SAlexey Kardashevskiy strerror(errno)); 1705e68cb8b4SAlexey Kardashevskiy } 1706e68cb8b4SAlexey Kardashevskiy } 1707e68cb8b4SAlexey Kardashevskiy 17084be21d56SDavid Gibson while (true) { 17094be21d56SDavid Gibson uint32_t index; 17104be21d56SDavid Gibson uint16_t n_valid, n_invalid; 17114be21d56SDavid Gibson 17124be21d56SDavid Gibson index = qemu_get_be32(f); 17134be21d56SDavid Gibson n_valid = qemu_get_be16(f); 17144be21d56SDavid Gibson n_invalid = qemu_get_be16(f); 17154be21d56SDavid Gibson 17164be21d56SDavid Gibson if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) { 17174be21d56SDavid Gibson /* End of Stream */ 17184be21d56SDavid Gibson break; 17194be21d56SDavid Gibson } 17204be21d56SDavid Gibson 1721e68cb8b4SAlexey Kardashevskiy if ((index + n_valid + n_invalid) > 17224be21d56SDavid Gibson (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) { 17234be21d56SDavid Gibson /* Bad index in stream */ 172498a5d100SDavid Gibson error_report( 172598a5d100SDavid Gibson "htab_load() bad index %d (%hd+%hd entries) in htab stream (htab_shift=%d)", 172698a5d100SDavid Gibson index, n_valid, n_invalid, spapr->htab_shift); 17274be21d56SDavid Gibson return -EINVAL; 17284be21d56SDavid Gibson } 17294be21d56SDavid Gibson 1730e68cb8b4SAlexey Kardashevskiy if (spapr->htab) { 17314be21d56SDavid Gibson if (n_valid) { 17324be21d56SDavid Gibson qemu_get_buffer(f, HPTE(spapr->htab, index), 17334be21d56SDavid Gibson HASH_PTE_SIZE_64 * n_valid); 17344be21d56SDavid Gibson } 17354be21d56SDavid Gibson if (n_invalid) { 17364be21d56SDavid Gibson memset(HPTE(spapr->htab, index + n_valid), 0, 17374be21d56SDavid Gibson HASH_PTE_SIZE_64 * n_invalid); 17384be21d56SDavid Gibson } 1739e68cb8b4SAlexey Kardashevskiy } else { 1740e68cb8b4SAlexey Kardashevskiy int rc; 1741e68cb8b4SAlexey Kardashevskiy 1742e68cb8b4SAlexey Kardashevskiy assert(fd >= 0); 1743e68cb8b4SAlexey Kardashevskiy 1744e68cb8b4SAlexey Kardashevskiy rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid); 1745e68cb8b4SAlexey Kardashevskiy if (rc < 0) { 1746e68cb8b4SAlexey Kardashevskiy return rc; 1747e68cb8b4SAlexey Kardashevskiy } 1748e68cb8b4SAlexey Kardashevskiy } 1749e68cb8b4SAlexey Kardashevskiy } 1750e68cb8b4SAlexey Kardashevskiy 1751e68cb8b4SAlexey Kardashevskiy if (!spapr->htab) { 1752e68cb8b4SAlexey Kardashevskiy assert(fd >= 0); 1753e68cb8b4SAlexey Kardashevskiy close(fd); 17544be21d56SDavid Gibson } 17554be21d56SDavid Gibson 17564be21d56SDavid Gibson return 0; 17574be21d56SDavid Gibson } 17584be21d56SDavid Gibson 1759c573fc03SThomas Huth static void htab_cleanup(void *opaque) 1760c573fc03SThomas Huth { 1761c573fc03SThomas Huth sPAPRMachineState *spapr = opaque; 1762c573fc03SThomas Huth 1763c573fc03SThomas Huth close_htab_fd(spapr); 1764c573fc03SThomas Huth } 1765c573fc03SThomas Huth 17664be21d56SDavid Gibson static SaveVMHandlers savevm_htab_handlers = { 17674be21d56SDavid Gibson .save_live_setup = htab_save_setup, 17684be21d56SDavid Gibson .save_live_iterate = htab_save_iterate, 1769a3e06c3dSDr. David Alan Gilbert .save_live_complete_precopy = htab_save_complete, 1770c573fc03SThomas Huth .cleanup = htab_cleanup, 17714be21d56SDavid Gibson .load_state = htab_load, 17724be21d56SDavid Gibson }; 17734be21d56SDavid Gibson 17745b2128d2SAlexander Graf static void spapr_boot_set(void *opaque, const char *boot_device, 17755b2128d2SAlexander Graf Error **errp) 17765b2128d2SAlexander Graf { 17775b2128d2SAlexander Graf MachineState *machine = MACHINE(qdev_get_machine()); 17785b2128d2SAlexander Graf machine->boot_order = g_strdup(boot_device); 17795b2128d2SAlexander Graf } 17805b2128d2SAlexander Graf 1781224245bfSDavid Gibson /* 1782224245bfSDavid Gibson * Reset routine for LMB DR devices. 1783224245bfSDavid Gibson * 1784224245bfSDavid Gibson * Unlike PCI DR devices, LMB DR devices explicitly register this reset 1785224245bfSDavid Gibson * routine. Reset for PCI DR devices will be handled by PHB reset routine 1786224245bfSDavid Gibson * when it walks all its children devices. LMB devices reset occurs 1787224245bfSDavid Gibson * as part of spapr_ppc_reset(). 1788224245bfSDavid Gibson */ 1789224245bfSDavid Gibson static void spapr_drc_reset(void *opaque) 1790224245bfSDavid Gibson { 1791224245bfSDavid Gibson sPAPRDRConnector *drc = opaque; 1792224245bfSDavid Gibson DeviceState *d = DEVICE(drc); 1793224245bfSDavid Gibson 1794224245bfSDavid Gibson if (d) { 1795224245bfSDavid Gibson device_reset(d); 1796224245bfSDavid Gibson } 1797224245bfSDavid Gibson } 1798224245bfSDavid Gibson 1799224245bfSDavid Gibson static void spapr_create_lmb_dr_connectors(sPAPRMachineState *spapr) 1800224245bfSDavid Gibson { 1801224245bfSDavid Gibson MachineState *machine = MACHINE(spapr); 1802224245bfSDavid Gibson uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE; 1803e8f986fcSBharata B Rao uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size; 1804224245bfSDavid Gibson int i; 1805224245bfSDavid Gibson 1806224245bfSDavid Gibson for (i = 0; i < nr_lmbs; i++) { 1807224245bfSDavid Gibson sPAPRDRConnector *drc; 1808224245bfSDavid Gibson uint64_t addr; 1809224245bfSDavid Gibson 1810e8f986fcSBharata B Rao addr = i * lmb_size + spapr->hotplug_memory.base; 1811224245bfSDavid Gibson drc = spapr_dr_connector_new(OBJECT(spapr), SPAPR_DR_CONNECTOR_TYPE_LMB, 1812224245bfSDavid Gibson addr/lmb_size); 1813224245bfSDavid Gibson qemu_register_reset(spapr_drc_reset, drc); 1814224245bfSDavid Gibson } 1815224245bfSDavid Gibson } 1816224245bfSDavid Gibson 1817224245bfSDavid Gibson /* 1818224245bfSDavid Gibson * If RAM size, maxmem size and individual node mem sizes aren't aligned 1819224245bfSDavid Gibson * to SPAPR_MEMORY_BLOCK_SIZE(256MB), then refuse to start the guest 1820224245bfSDavid Gibson * since we can't support such unaligned sizes with DRCONF_MEMORY. 1821224245bfSDavid Gibson */ 18227c150d6fSDavid Gibson static void spapr_validate_node_memory(MachineState *machine, Error **errp) 1823224245bfSDavid Gibson { 1824224245bfSDavid Gibson int i; 1825224245bfSDavid Gibson 18267c150d6fSDavid Gibson if (machine->ram_size % SPAPR_MEMORY_BLOCK_SIZE) { 18277c150d6fSDavid Gibson error_setg(errp, "Memory size 0x" RAM_ADDR_FMT 18287c150d6fSDavid Gibson " is not aligned to %llu MiB", 18297c150d6fSDavid Gibson machine->ram_size, 1830224245bfSDavid Gibson SPAPR_MEMORY_BLOCK_SIZE / M_BYTE); 18317c150d6fSDavid Gibson return; 18327c150d6fSDavid Gibson } 18337c150d6fSDavid Gibson 18347c150d6fSDavid Gibson if (machine->maxram_size % SPAPR_MEMORY_BLOCK_SIZE) { 18357c150d6fSDavid Gibson error_setg(errp, "Maximum memory size 0x" RAM_ADDR_FMT 18367c150d6fSDavid Gibson " is not aligned to %llu MiB", 18377c150d6fSDavid Gibson machine->ram_size, 18387c150d6fSDavid Gibson SPAPR_MEMORY_BLOCK_SIZE / M_BYTE); 18397c150d6fSDavid Gibson return; 1840224245bfSDavid Gibson } 1841224245bfSDavid Gibson 1842224245bfSDavid Gibson for (i = 0; i < nb_numa_nodes; i++) { 1843224245bfSDavid Gibson if (numa_info[i].node_mem % SPAPR_MEMORY_BLOCK_SIZE) { 18447c150d6fSDavid Gibson error_setg(errp, 18457c150d6fSDavid Gibson "Node %d memory size 0x%" PRIx64 18467c150d6fSDavid Gibson " is not aligned to %llu MiB", 18477c150d6fSDavid Gibson i, numa_info[i].node_mem, 1848224245bfSDavid Gibson SPAPR_MEMORY_BLOCK_SIZE / M_BYTE); 18497c150d6fSDavid Gibson return; 1850224245bfSDavid Gibson } 1851224245bfSDavid Gibson } 1852224245bfSDavid Gibson } 1853224245bfSDavid Gibson 1854535455fdSIgor Mammedov /* find cpu slot in machine->possible_cpus by core_id */ 1855535455fdSIgor Mammedov static CPUArchId *spapr_find_cpu_slot(MachineState *ms, uint32_t id, int *idx) 1856535455fdSIgor Mammedov { 1857535455fdSIgor Mammedov int index = id / smp_threads; 1858535455fdSIgor Mammedov 1859535455fdSIgor Mammedov if (index >= ms->possible_cpus->len) { 1860535455fdSIgor Mammedov return NULL; 1861535455fdSIgor Mammedov } 1862535455fdSIgor Mammedov if (idx) { 1863535455fdSIgor Mammedov *idx = index; 1864535455fdSIgor Mammedov } 1865535455fdSIgor Mammedov return &ms->possible_cpus->cpus[index]; 1866535455fdSIgor Mammedov } 1867535455fdSIgor Mammedov 18680c86d0fdSDavid Gibson static void spapr_init_cpus(sPAPRMachineState *spapr) 18690c86d0fdSDavid Gibson { 18700c86d0fdSDavid Gibson MachineState *machine = MACHINE(spapr); 18710c86d0fdSDavid Gibson MachineClass *mc = MACHINE_GET_CLASS(machine); 18720c86d0fdSDavid Gibson char *type = spapr_get_cpu_core_type(machine->cpu_model); 18730c86d0fdSDavid Gibson int smt = kvmppc_smt_threads(); 1874535455fdSIgor Mammedov const CPUArchIdList *possible_cpus; 1875535455fdSIgor Mammedov int boot_cores_nr = smp_cpus / smp_threads; 18760c86d0fdSDavid Gibson int i; 18770c86d0fdSDavid Gibson 18780c86d0fdSDavid Gibson if (!type) { 18790c86d0fdSDavid Gibson error_report("Unable to find sPAPR CPU Core definition"); 18800c86d0fdSDavid Gibson exit(1); 18810c86d0fdSDavid Gibson } 18820c86d0fdSDavid Gibson 1883535455fdSIgor Mammedov possible_cpus = mc->possible_cpu_arch_ids(machine); 1884c5514d0eSIgor Mammedov if (mc->has_hotpluggable_cpus) { 18850c86d0fdSDavid Gibson if (smp_cpus % smp_threads) { 18860c86d0fdSDavid Gibson error_report("smp_cpus (%u) must be multiple of threads (%u)", 18870c86d0fdSDavid Gibson smp_cpus, smp_threads); 18880c86d0fdSDavid Gibson exit(1); 18890c86d0fdSDavid Gibson } 18900c86d0fdSDavid Gibson if (max_cpus % smp_threads) { 18910c86d0fdSDavid Gibson error_report("max_cpus (%u) must be multiple of threads (%u)", 18920c86d0fdSDavid Gibson max_cpus, smp_threads); 18930c86d0fdSDavid Gibson exit(1); 18940c86d0fdSDavid Gibson } 18950c86d0fdSDavid Gibson } else { 18960c86d0fdSDavid Gibson if (max_cpus != smp_cpus) { 18970c86d0fdSDavid Gibson error_report("This machine version does not support CPU hotplug"); 18980c86d0fdSDavid Gibson exit(1); 18990c86d0fdSDavid Gibson } 1900535455fdSIgor Mammedov boot_cores_nr = possible_cpus->len; 19010c86d0fdSDavid Gibson } 19020c86d0fdSDavid Gibson 1903535455fdSIgor Mammedov for (i = 0; i < possible_cpus->len; i++) { 19040c86d0fdSDavid Gibson int core_id = i * smp_threads; 19050c86d0fdSDavid Gibson 1906c5514d0eSIgor Mammedov if (mc->has_hotpluggable_cpus) { 19070c86d0fdSDavid Gibson sPAPRDRConnector *drc = 19080c86d0fdSDavid Gibson spapr_dr_connector_new(OBJECT(spapr), 19090c86d0fdSDavid Gibson SPAPR_DR_CONNECTOR_TYPE_CPU, 19100c86d0fdSDavid Gibson (core_id / smp_threads) * smt); 19110c86d0fdSDavid Gibson 19120c86d0fdSDavid Gibson qemu_register_reset(spapr_drc_reset, drc); 19130c86d0fdSDavid Gibson } 19140c86d0fdSDavid Gibson 1915535455fdSIgor Mammedov if (i < boot_cores_nr) { 19160c86d0fdSDavid Gibson Object *core = object_new(type); 19170c86d0fdSDavid Gibson int nr_threads = smp_threads; 19180c86d0fdSDavid Gibson 19190c86d0fdSDavid Gibson /* Handle the partially filled core for older machine types */ 19200c86d0fdSDavid Gibson if ((i + 1) * smp_threads >= smp_cpus) { 19210c86d0fdSDavid Gibson nr_threads = smp_cpus - i * smp_threads; 19220c86d0fdSDavid Gibson } 19230c86d0fdSDavid Gibson 19240c86d0fdSDavid Gibson object_property_set_int(core, nr_threads, "nr-threads", 19250c86d0fdSDavid Gibson &error_fatal); 19260c86d0fdSDavid Gibson object_property_set_int(core, core_id, CPU_CORE_PROP_CORE_ID, 19270c86d0fdSDavid Gibson &error_fatal); 19280c86d0fdSDavid Gibson object_property_set_bool(core, true, "realized", &error_fatal); 19290c86d0fdSDavid Gibson } 19300c86d0fdSDavid Gibson } 19310c86d0fdSDavid Gibson g_free(type); 19320c86d0fdSDavid Gibson } 19330c86d0fdSDavid Gibson 193453018216SPaolo Bonzini /* pSeries LPAR / sPAPR hardware init */ 19353ef96221SMarcel Apfelbaum static void ppc_spapr_init(MachineState *machine) 193653018216SPaolo Bonzini { 193728e02042SDavid Gibson sPAPRMachineState *spapr = SPAPR_MACHINE(machine); 1938224245bfSDavid Gibson sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine); 19393ef96221SMarcel Apfelbaum const char *kernel_filename = machine->kernel_filename; 19403ef96221SMarcel Apfelbaum const char *initrd_filename = machine->initrd_filename; 194153018216SPaolo Bonzini PCIHostState *phb; 194253018216SPaolo Bonzini int i; 194353018216SPaolo Bonzini MemoryRegion *sysmem = get_system_memory(); 194453018216SPaolo Bonzini MemoryRegion *ram = g_new(MemoryRegion, 1); 1945658fa66bSAlexey Kardashevskiy MemoryRegion *rma_region; 1946658fa66bSAlexey Kardashevskiy void *rma = NULL; 194753018216SPaolo Bonzini hwaddr rma_alloc_size; 1948b082d65aSAlexey Kardashevskiy hwaddr node0_size = spapr_node0_size(); 1949b7d1f77aSBenjamin Herrenschmidt long load_limit, fw_size; 195053018216SPaolo Bonzini char *filename; 195194a94e4cSBharata B Rao int smt = kvmppc_smt_threads(); 195253018216SPaolo Bonzini 1953226419d6SMichael S. Tsirkin msi_nonbroken = true; 195453018216SPaolo Bonzini 195553018216SPaolo Bonzini QLIST_INIT(&spapr->phbs); 195653018216SPaolo Bonzini 195753018216SPaolo Bonzini /* Allocate RMA if necessary */ 1958658fa66bSAlexey Kardashevskiy rma_alloc_size = kvmppc_alloc_rma(&rma); 195953018216SPaolo Bonzini 196053018216SPaolo Bonzini if (rma_alloc_size == -1) { 1961730fce59SThomas Huth error_report("Unable to create RMA"); 196253018216SPaolo Bonzini exit(1); 196353018216SPaolo Bonzini } 196453018216SPaolo Bonzini 1965c4177479SAlexey Kardashevskiy if (rma_alloc_size && (rma_alloc_size < node0_size)) { 196653018216SPaolo Bonzini spapr->rma_size = rma_alloc_size; 196753018216SPaolo Bonzini } else { 1968c4177479SAlexey Kardashevskiy spapr->rma_size = node0_size; 196953018216SPaolo Bonzini 197053018216SPaolo Bonzini /* With KVM, we don't actually know whether KVM supports an 197153018216SPaolo Bonzini * unbounded RMA (PR KVM) or is limited by the hash table size 197253018216SPaolo Bonzini * (HV KVM using VRMA), so we always assume the latter 197353018216SPaolo Bonzini * 197453018216SPaolo Bonzini * In that case, we also limit the initial allocations for RTAS 197553018216SPaolo Bonzini * etc... to 256M since we have no way to know what the VRMA size 197653018216SPaolo Bonzini * is going to be as it depends on the size of the hash table 197753018216SPaolo Bonzini * isn't determined yet. 197853018216SPaolo Bonzini */ 197953018216SPaolo Bonzini if (kvm_enabled()) { 198053018216SPaolo Bonzini spapr->vrma_adjust = 1; 198153018216SPaolo Bonzini spapr->rma_size = MIN(spapr->rma_size, 0x10000000); 198253018216SPaolo Bonzini } 1983912acdf4SBenjamin Herrenschmidt 1984912acdf4SBenjamin Herrenschmidt /* Actually we don't support unbounded RMA anymore since we 1985912acdf4SBenjamin Herrenschmidt * added proper emulation of HV mode. The max we can get is 1986912acdf4SBenjamin Herrenschmidt * 16G which also happens to be what we configure for PAPR 1987912acdf4SBenjamin Herrenschmidt * mode so make sure we don't do anything bigger than that 1988912acdf4SBenjamin Herrenschmidt */ 1989912acdf4SBenjamin Herrenschmidt spapr->rma_size = MIN(spapr->rma_size, 0x400000000ull); 199053018216SPaolo Bonzini } 199153018216SPaolo Bonzini 1992c4177479SAlexey Kardashevskiy if (spapr->rma_size > node0_size) { 1993d54e4d76SDavid Gibson error_report("Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")", 1994c4177479SAlexey Kardashevskiy spapr->rma_size); 1995c4177479SAlexey Kardashevskiy exit(1); 1996c4177479SAlexey Kardashevskiy } 1997c4177479SAlexey Kardashevskiy 1998b7d1f77aSBenjamin Herrenschmidt /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */ 1999b7d1f77aSBenjamin Herrenschmidt load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD; 200053018216SPaolo Bonzini 20017b565160SDavid Gibson /* Set up Interrupt Controller before we create the VCPUs */ 200227f24582SBenjamin Herrenschmidt spapr->xics = xics_system_init(machine, 200394a94e4cSBharata B Rao DIV_ROUND_UP(max_cpus * smt, smp_threads), 2004161deaf2SBenjamin Herrenschmidt XICS_IRQS_SPAPR, &error_fatal); 20057b565160SDavid Gibson 2006facdb8b6SMichael Roth /* Set up containers for ibm,client-set-architecture negotiated options */ 2007facdb8b6SMichael Roth spapr->ov5 = spapr_ovec_new(); 2008facdb8b6SMichael Roth spapr->ov5_cas = spapr_ovec_new(); 2009facdb8b6SMichael Roth 2010224245bfSDavid Gibson if (smc->dr_lmb_enabled) { 2011facdb8b6SMichael Roth spapr_ovec_set(spapr->ov5, OV5_DRCONF_MEMORY); 20127c150d6fSDavid Gibson spapr_validate_node_memory(machine, &error_fatal); 2013224245bfSDavid Gibson } 2014224245bfSDavid Gibson 2015417ece33SMichael Roth spapr_ovec_set(spapr->ov5, OV5_FORM1_AFFINITY); 2016417ece33SMichael Roth 2017ffbb1705SMichael Roth /* advertise support for dedicated HP event source to guests */ 2018ffbb1705SMichael Roth if (spapr->use_hotplug_event_source) { 2019ffbb1705SMichael Roth spapr_ovec_set(spapr->ov5, OV5_HP_EVT); 2020ffbb1705SMichael Roth } 2021ffbb1705SMichael Roth 202253018216SPaolo Bonzini /* init CPUs */ 202319fb2c36SBharata B Rao if (machine->cpu_model == NULL) { 20243daa4a9fSThomas Huth machine->cpu_model = kvm_enabled() ? "host" : smc->tcg_default_cpu; 202553018216SPaolo Bonzini } 202694a94e4cSBharata B Rao 2027e703d2f7SGreg Kurz ppc_cpu_parse_features(machine->cpu_model); 2028e703d2f7SGreg Kurz 20290c86d0fdSDavid Gibson spapr_init_cpus(spapr); 203053018216SPaolo Bonzini 2031026bfd89SDavid Gibson if (kvm_enabled()) { 2032026bfd89SDavid Gibson /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */ 2033026bfd89SDavid Gibson kvmppc_enable_logical_ci_hcalls(); 2034ef9971ddSAlexey Kardashevskiy kvmppc_enable_set_mode_hcall(); 20355145ad4fSNathan Whitehorn 20365145ad4fSNathan Whitehorn /* H_CLEAR_MOD/_REF are mandatory in PAPR, but off by default */ 20375145ad4fSNathan Whitehorn kvmppc_enable_clear_ref_mod_hcalls(); 2038026bfd89SDavid Gibson } 2039026bfd89SDavid Gibson 204053018216SPaolo Bonzini /* allocate RAM */ 2041f92f5da1SAlexey Kardashevskiy memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram", 2042fb164994SDavid Gibson machine->ram_size); 2043f92f5da1SAlexey Kardashevskiy memory_region_add_subregion(sysmem, 0, ram); 204453018216SPaolo Bonzini 2045658fa66bSAlexey Kardashevskiy if (rma_alloc_size && rma) { 2046658fa66bSAlexey Kardashevskiy rma_region = g_new(MemoryRegion, 1); 2047658fa66bSAlexey Kardashevskiy memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma", 2048658fa66bSAlexey Kardashevskiy rma_alloc_size, rma); 2049658fa66bSAlexey Kardashevskiy vmstate_register_ram_global(rma_region); 2050658fa66bSAlexey Kardashevskiy memory_region_add_subregion(sysmem, 0, rma_region); 2051658fa66bSAlexey Kardashevskiy } 2052658fa66bSAlexey Kardashevskiy 20534a1c9cf0SBharata B Rao /* initialize hotplug memory address space */ 20544a1c9cf0SBharata B Rao if (machine->ram_size < machine->maxram_size) { 20554a1c9cf0SBharata B Rao ram_addr_t hotplug_mem_size = machine->maxram_size - machine->ram_size; 205671c9a3ddSBharata B Rao /* 205771c9a3ddSBharata B Rao * Limit the number of hotpluggable memory slots to half the number 205871c9a3ddSBharata B Rao * slots that KVM supports, leaving the other half for PCI and other 205971c9a3ddSBharata B Rao * devices. However ensure that number of slots doesn't drop below 32. 206071c9a3ddSBharata B Rao */ 206171c9a3ddSBharata B Rao int max_memslots = kvm_enabled() ? kvm_get_max_memslots() / 2 : 206271c9a3ddSBharata B Rao SPAPR_MAX_RAM_SLOTS; 20634a1c9cf0SBharata B Rao 206471c9a3ddSBharata B Rao if (max_memslots < SPAPR_MAX_RAM_SLOTS) { 206571c9a3ddSBharata B Rao max_memslots = SPAPR_MAX_RAM_SLOTS; 206671c9a3ddSBharata B Rao } 206771c9a3ddSBharata B Rao if (machine->ram_slots > max_memslots) { 2068d54e4d76SDavid Gibson error_report("Specified number of memory slots %" 2069d54e4d76SDavid Gibson PRIu64" exceeds max supported %d", 207071c9a3ddSBharata B Rao machine->ram_slots, max_memslots); 2071d54e4d76SDavid Gibson exit(1); 20724a1c9cf0SBharata B Rao } 20734a1c9cf0SBharata B Rao 20744a1c9cf0SBharata B Rao spapr->hotplug_memory.base = ROUND_UP(machine->ram_size, 20754a1c9cf0SBharata B Rao SPAPR_HOTPLUG_MEM_ALIGN); 20764a1c9cf0SBharata B Rao memory_region_init(&spapr->hotplug_memory.mr, OBJECT(spapr), 20774a1c9cf0SBharata B Rao "hotplug-memory", hotplug_mem_size); 20784a1c9cf0SBharata B Rao memory_region_add_subregion(sysmem, spapr->hotplug_memory.base, 20794a1c9cf0SBharata B Rao &spapr->hotplug_memory.mr); 20804a1c9cf0SBharata B Rao } 20814a1c9cf0SBharata B Rao 2082224245bfSDavid Gibson if (smc->dr_lmb_enabled) { 2083224245bfSDavid Gibson spapr_create_lmb_dr_connectors(spapr); 2084224245bfSDavid Gibson } 2085224245bfSDavid Gibson 208653018216SPaolo Bonzini filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin"); 20874c56440dSStefan Weil if (!filename) { 2088730fce59SThomas Huth error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin"); 20894c56440dSStefan Weil exit(1); 20904c56440dSStefan Weil } 2091b7d1f77aSBenjamin Herrenschmidt spapr->rtas_size = get_image_size(filename); 20928afc22a2SZhou Jie if (spapr->rtas_size < 0) { 20938afc22a2SZhou Jie error_report("Could not get size of LPAR rtas '%s'", filename); 20948afc22a2SZhou Jie exit(1); 20958afc22a2SZhou Jie } 2096b7d1f77aSBenjamin Herrenschmidt spapr->rtas_blob = g_malloc(spapr->rtas_size); 2097b7d1f77aSBenjamin Herrenschmidt if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) { 2098730fce59SThomas Huth error_report("Could not load LPAR rtas '%s'", filename); 209953018216SPaolo Bonzini exit(1); 210053018216SPaolo Bonzini } 210153018216SPaolo Bonzini if (spapr->rtas_size > RTAS_MAX_SIZE) { 2102730fce59SThomas Huth error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)", 21032f285bddSPeter Maydell (size_t)spapr->rtas_size, RTAS_MAX_SIZE); 210453018216SPaolo Bonzini exit(1); 210553018216SPaolo Bonzini } 210653018216SPaolo Bonzini g_free(filename); 210753018216SPaolo Bonzini 2108ffbb1705SMichael Roth /* Set up RTAS event infrastructure */ 210953018216SPaolo Bonzini spapr_events_init(spapr); 211053018216SPaolo Bonzini 211112f42174SDavid Gibson /* Set up the RTC RTAS interfaces */ 211228df36a1SDavid Gibson spapr_rtc_create(spapr); 211312f42174SDavid Gibson 211453018216SPaolo Bonzini /* Set up VIO bus */ 211553018216SPaolo Bonzini spapr->vio_bus = spapr_vio_bus_init(); 211653018216SPaolo Bonzini 211753018216SPaolo Bonzini for (i = 0; i < MAX_SERIAL_PORTS; i++) { 211853018216SPaolo Bonzini if (serial_hds[i]) { 211953018216SPaolo Bonzini spapr_vty_create(spapr->vio_bus, serial_hds[i]); 212053018216SPaolo Bonzini } 212153018216SPaolo Bonzini } 212253018216SPaolo Bonzini 212353018216SPaolo Bonzini /* We always have at least the nvram device on VIO */ 212453018216SPaolo Bonzini spapr_create_nvram(spapr); 212553018216SPaolo Bonzini 212653018216SPaolo Bonzini /* Set up PCI */ 212753018216SPaolo Bonzini spapr_pci_rtas_init(); 212853018216SPaolo Bonzini 212989dfd6e1SDavid Gibson phb = spapr_create_phb(spapr, 0); 213053018216SPaolo Bonzini 213153018216SPaolo Bonzini for (i = 0; i < nb_nics; i++) { 213253018216SPaolo Bonzini NICInfo *nd = &nd_table[i]; 213353018216SPaolo Bonzini 213453018216SPaolo Bonzini if (!nd->model) { 213553018216SPaolo Bonzini nd->model = g_strdup("ibmveth"); 213653018216SPaolo Bonzini } 213753018216SPaolo Bonzini 213853018216SPaolo Bonzini if (strcmp(nd->model, "ibmveth") == 0) { 213953018216SPaolo Bonzini spapr_vlan_create(spapr->vio_bus, nd); 214053018216SPaolo Bonzini } else { 214129b358f9SDavid Gibson pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL); 214253018216SPaolo Bonzini } 214353018216SPaolo Bonzini } 214453018216SPaolo Bonzini 214553018216SPaolo Bonzini for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) { 214653018216SPaolo Bonzini spapr_vscsi_create(spapr->vio_bus); 214753018216SPaolo Bonzini } 214853018216SPaolo Bonzini 214953018216SPaolo Bonzini /* Graphics */ 215014c6a894SDavid Gibson if (spapr_vga_init(phb->bus, &error_fatal)) { 215153018216SPaolo Bonzini spapr->has_graphics = true; 2152c6e76503SPaolo Bonzini machine->usb |= defaults_enabled() && !machine->usb_disabled; 215353018216SPaolo Bonzini } 215453018216SPaolo Bonzini 21554ee9ced9SMarcel Apfelbaum if (machine->usb) { 215657040d45SThomas Huth if (smc->use_ohci_by_default) { 215753018216SPaolo Bonzini pci_create_simple(phb->bus, -1, "pci-ohci"); 215857040d45SThomas Huth } else { 215957040d45SThomas Huth pci_create_simple(phb->bus, -1, "nec-usb-xhci"); 216057040d45SThomas Huth } 2161c86580b8SMarkus Armbruster 216253018216SPaolo Bonzini if (spapr->has_graphics) { 2163c86580b8SMarkus Armbruster USBBus *usb_bus = usb_bus_find(-1); 2164c86580b8SMarkus Armbruster 2165c86580b8SMarkus Armbruster usb_create_simple(usb_bus, "usb-kbd"); 2166c86580b8SMarkus Armbruster usb_create_simple(usb_bus, "usb-mouse"); 216753018216SPaolo Bonzini } 216853018216SPaolo Bonzini } 216953018216SPaolo Bonzini 217053018216SPaolo Bonzini if (spapr->rma_size < (MIN_RMA_SLOF << 20)) { 2171d54e4d76SDavid Gibson error_report( 2172d54e4d76SDavid Gibson "pSeries SLOF firmware requires >= %ldM guest RMA (Real Mode Area memory)", 2173d54e4d76SDavid Gibson MIN_RMA_SLOF); 217453018216SPaolo Bonzini exit(1); 217553018216SPaolo Bonzini } 217653018216SPaolo Bonzini 217753018216SPaolo Bonzini if (kernel_filename) { 217853018216SPaolo Bonzini uint64_t lowaddr = 0; 217953018216SPaolo Bonzini 2180a19f7fb0SDavid Gibson spapr->kernel_size = load_elf(kernel_filename, translate_kernel_address, 2181a19f7fb0SDavid Gibson NULL, NULL, &lowaddr, NULL, 1, 2182a19f7fb0SDavid Gibson PPC_ELF_MACHINE, 0, 0); 2183a19f7fb0SDavid Gibson if (spapr->kernel_size == ELF_LOAD_WRONG_ENDIAN) { 2184a19f7fb0SDavid Gibson spapr->kernel_size = load_elf(kernel_filename, 2185a19f7fb0SDavid Gibson translate_kernel_address, NULL, NULL, 2186a19f7fb0SDavid Gibson &lowaddr, NULL, 0, PPC_ELF_MACHINE, 21877ef295eaSPeter Crosthwaite 0, 0); 2188a19f7fb0SDavid Gibson spapr->kernel_le = spapr->kernel_size > 0; 218916457e7fSBenjamin Herrenschmidt } 2190a19f7fb0SDavid Gibson if (spapr->kernel_size < 0) { 2191a19f7fb0SDavid Gibson error_report("error loading %s: %s", kernel_filename, 2192a19f7fb0SDavid Gibson load_elf_strerror(spapr->kernel_size)); 219353018216SPaolo Bonzini exit(1); 219453018216SPaolo Bonzini } 219553018216SPaolo Bonzini 219653018216SPaolo Bonzini /* load initrd */ 219753018216SPaolo Bonzini if (initrd_filename) { 219853018216SPaolo Bonzini /* Try to locate the initrd in the gap between the kernel 219953018216SPaolo Bonzini * and the firmware. Add a bit of space just in case 220053018216SPaolo Bonzini */ 2201a19f7fb0SDavid Gibson spapr->initrd_base = (KERNEL_LOAD_ADDR + spapr->kernel_size 2202a19f7fb0SDavid Gibson + 0x1ffff) & ~0xffff; 2203a19f7fb0SDavid Gibson spapr->initrd_size = load_image_targphys(initrd_filename, 2204a19f7fb0SDavid Gibson spapr->initrd_base, 2205a19f7fb0SDavid Gibson load_limit 2206a19f7fb0SDavid Gibson - spapr->initrd_base); 2207a19f7fb0SDavid Gibson if (spapr->initrd_size < 0) { 2208d54e4d76SDavid Gibson error_report("could not load initial ram disk '%s'", 220953018216SPaolo Bonzini initrd_filename); 221053018216SPaolo Bonzini exit(1); 221153018216SPaolo Bonzini } 221253018216SPaolo Bonzini } 221353018216SPaolo Bonzini } 221453018216SPaolo Bonzini 22158e7ea787SAndreas Färber if (bios_name == NULL) { 22168e7ea787SAndreas Färber bios_name = FW_FILE_NAME; 22178e7ea787SAndreas Färber } 22188e7ea787SAndreas Färber filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name); 22194c56440dSStefan Weil if (!filename) { 222068fea5a0SThomas Huth error_report("Could not find LPAR firmware '%s'", bios_name); 22214c56440dSStefan Weil exit(1); 22224c56440dSStefan Weil } 222353018216SPaolo Bonzini fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE); 222468fea5a0SThomas Huth if (fw_size <= 0) { 222568fea5a0SThomas Huth error_report("Could not load LPAR firmware '%s'", filename); 222653018216SPaolo Bonzini exit(1); 222753018216SPaolo Bonzini } 222853018216SPaolo Bonzini g_free(filename); 222953018216SPaolo Bonzini 223028e02042SDavid Gibson /* FIXME: Should register things through the MachineState's qdev 223128e02042SDavid Gibson * interface, this is a legacy from the sPAPREnvironment structure 223228e02042SDavid Gibson * which predated MachineState but had a similar function */ 22334be21d56SDavid Gibson vmstate_register(NULL, 0, &vmstate_spapr, spapr); 22344be21d56SDavid Gibson register_savevm_live(NULL, "spapr/htab", -1, 1, 22354be21d56SDavid Gibson &savevm_htab_handlers, spapr); 22364be21d56SDavid Gibson 223746503c2bSMichael Roth /* used by RTAS */ 223846503c2bSMichael Roth QTAILQ_INIT(&spapr->ccs_list); 223946503c2bSMichael Roth qemu_register_reset(spapr_ccs_reset_hook, spapr); 224046503c2bSMichael Roth 22415b2128d2SAlexander Graf qemu_register_boot_set(spapr_boot_set, spapr); 224242043e4fSLaurent Vivier 224342043e4fSLaurent Vivier /* to stop and start vmclock */ 224442043e4fSLaurent Vivier if (kvm_enabled()) { 224542043e4fSLaurent Vivier qemu_add_vm_change_state_handler(cpu_ppc_clock_vm_state_change, 224642043e4fSLaurent Vivier &spapr->tb); 224742043e4fSLaurent Vivier } 224853018216SPaolo Bonzini } 224953018216SPaolo Bonzini 2250135a129aSAneesh Kumar K.V static int spapr_kvm_type(const char *vm_type) 2251135a129aSAneesh Kumar K.V { 2252135a129aSAneesh Kumar K.V if (!vm_type) { 2253135a129aSAneesh Kumar K.V return 0; 2254135a129aSAneesh Kumar K.V } 2255135a129aSAneesh Kumar K.V 2256135a129aSAneesh Kumar K.V if (!strcmp(vm_type, "HV")) { 2257135a129aSAneesh Kumar K.V return 1; 2258135a129aSAneesh Kumar K.V } 2259135a129aSAneesh Kumar K.V 2260135a129aSAneesh Kumar K.V if (!strcmp(vm_type, "PR")) { 2261135a129aSAneesh Kumar K.V return 2; 2262135a129aSAneesh Kumar K.V } 2263135a129aSAneesh Kumar K.V 2264135a129aSAneesh Kumar K.V error_report("Unknown kvm-type specified '%s'", vm_type); 2265135a129aSAneesh Kumar K.V exit(1); 2266135a129aSAneesh Kumar K.V } 2267135a129aSAneesh Kumar K.V 226871461b0fSAlexey Kardashevskiy /* 2269627b84f4SGonglei * Implementation of an interface to adjust firmware path 227071461b0fSAlexey Kardashevskiy * for the bootindex property handling. 227171461b0fSAlexey Kardashevskiy */ 227271461b0fSAlexey Kardashevskiy static char *spapr_get_fw_dev_path(FWPathProvider *p, BusState *bus, 227371461b0fSAlexey Kardashevskiy DeviceState *dev) 227471461b0fSAlexey Kardashevskiy { 227571461b0fSAlexey Kardashevskiy #define CAST(type, obj, name) \ 227671461b0fSAlexey Kardashevskiy ((type *)object_dynamic_cast(OBJECT(obj), (name))) 227771461b0fSAlexey Kardashevskiy SCSIDevice *d = CAST(SCSIDevice, dev, TYPE_SCSI_DEVICE); 227871461b0fSAlexey Kardashevskiy sPAPRPHBState *phb = CAST(sPAPRPHBState, dev, TYPE_SPAPR_PCI_HOST_BRIDGE); 227971461b0fSAlexey Kardashevskiy 228071461b0fSAlexey Kardashevskiy if (d) { 228171461b0fSAlexey Kardashevskiy void *spapr = CAST(void, bus->parent, "spapr-vscsi"); 228271461b0fSAlexey Kardashevskiy VirtIOSCSI *virtio = CAST(VirtIOSCSI, bus->parent, TYPE_VIRTIO_SCSI); 228371461b0fSAlexey Kardashevskiy USBDevice *usb = CAST(USBDevice, bus->parent, TYPE_USB_DEVICE); 228471461b0fSAlexey Kardashevskiy 228571461b0fSAlexey Kardashevskiy if (spapr) { 228671461b0fSAlexey Kardashevskiy /* 228771461b0fSAlexey Kardashevskiy * Replace "channel@0/disk@0,0" with "disk@8000000000000000": 228871461b0fSAlexey Kardashevskiy * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun 228971461b0fSAlexey Kardashevskiy * in the top 16 bits of the 64-bit LUN 229071461b0fSAlexey Kardashevskiy */ 229171461b0fSAlexey Kardashevskiy unsigned id = 0x8000 | (d->id << 8) | d->lun; 229271461b0fSAlexey Kardashevskiy return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev), 229371461b0fSAlexey Kardashevskiy (uint64_t)id << 48); 229471461b0fSAlexey Kardashevskiy } else if (virtio) { 229571461b0fSAlexey Kardashevskiy /* 229671461b0fSAlexey Kardashevskiy * We use SRP luns of the form 01000000 | (target << 8) | lun 229771461b0fSAlexey Kardashevskiy * in the top 32 bits of the 64-bit LUN 229871461b0fSAlexey Kardashevskiy * Note: the quote above is from SLOF and it is wrong, 229971461b0fSAlexey Kardashevskiy * the actual binding is: 230071461b0fSAlexey Kardashevskiy * swap 0100 or 10 << or 20 << ( target lun-id -- srplun ) 230171461b0fSAlexey Kardashevskiy */ 230271461b0fSAlexey Kardashevskiy unsigned id = 0x1000000 | (d->id << 16) | d->lun; 230371461b0fSAlexey Kardashevskiy return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev), 230471461b0fSAlexey Kardashevskiy (uint64_t)id << 32); 230571461b0fSAlexey Kardashevskiy } else if (usb) { 230671461b0fSAlexey Kardashevskiy /* 230771461b0fSAlexey Kardashevskiy * We use SRP luns of the form 01000000 | (usb-port << 16) | lun 230871461b0fSAlexey Kardashevskiy * in the top 32 bits of the 64-bit LUN 230971461b0fSAlexey Kardashevskiy */ 231071461b0fSAlexey Kardashevskiy unsigned usb_port = atoi(usb->port->path); 231171461b0fSAlexey Kardashevskiy unsigned id = 0x1000000 | (usb_port << 16) | d->lun; 231271461b0fSAlexey Kardashevskiy return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev), 231371461b0fSAlexey Kardashevskiy (uint64_t)id << 32); 231471461b0fSAlexey Kardashevskiy } 231571461b0fSAlexey Kardashevskiy } 231671461b0fSAlexey Kardashevskiy 2317b99260ebSThomas Huth /* 2318b99260ebSThomas Huth * SLOF probes the USB devices, and if it recognizes that the device is a 2319b99260ebSThomas Huth * storage device, it changes its name to "storage" instead of "usb-host", 2320b99260ebSThomas Huth * and additionally adds a child node for the SCSI LUN, so the correct 2321b99260ebSThomas Huth * boot path in SLOF is something like .../storage@1/disk@xxx" instead. 2322b99260ebSThomas Huth */ 2323b99260ebSThomas Huth if (strcmp("usb-host", qdev_fw_name(dev)) == 0) { 2324b99260ebSThomas Huth USBDevice *usbdev = CAST(USBDevice, dev, TYPE_USB_DEVICE); 2325b99260ebSThomas Huth if (usb_host_dev_is_scsi_storage(usbdev)) { 2326b99260ebSThomas Huth return g_strdup_printf("storage@%s/disk", usbdev->port->path); 2327b99260ebSThomas Huth } 2328b99260ebSThomas Huth } 2329b99260ebSThomas Huth 233071461b0fSAlexey Kardashevskiy if (phb) { 233171461b0fSAlexey Kardashevskiy /* Replace "pci" with "pci@800000020000000" */ 233271461b0fSAlexey Kardashevskiy return g_strdup_printf("pci@%"PRIX64, phb->buid); 233371461b0fSAlexey Kardashevskiy } 233471461b0fSAlexey Kardashevskiy 233571461b0fSAlexey Kardashevskiy return NULL; 233671461b0fSAlexey Kardashevskiy } 233771461b0fSAlexey Kardashevskiy 233823825581SEduardo Habkost static char *spapr_get_kvm_type(Object *obj, Error **errp) 233923825581SEduardo Habkost { 234028e02042SDavid Gibson sPAPRMachineState *spapr = SPAPR_MACHINE(obj); 234123825581SEduardo Habkost 234228e02042SDavid Gibson return g_strdup(spapr->kvm_type); 234323825581SEduardo Habkost } 234423825581SEduardo Habkost 234523825581SEduardo Habkost static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp) 234623825581SEduardo Habkost { 234728e02042SDavid Gibson sPAPRMachineState *spapr = SPAPR_MACHINE(obj); 234823825581SEduardo Habkost 234928e02042SDavid Gibson g_free(spapr->kvm_type); 235028e02042SDavid Gibson spapr->kvm_type = g_strdup(value); 235123825581SEduardo Habkost } 235223825581SEduardo Habkost 2353f6229214SMichael Roth static bool spapr_get_modern_hotplug_events(Object *obj, Error **errp) 2354f6229214SMichael Roth { 2355f6229214SMichael Roth sPAPRMachineState *spapr = SPAPR_MACHINE(obj); 2356f6229214SMichael Roth 2357f6229214SMichael Roth return spapr->use_hotplug_event_source; 2358f6229214SMichael Roth } 2359f6229214SMichael Roth 2360f6229214SMichael Roth static void spapr_set_modern_hotplug_events(Object *obj, bool value, 2361f6229214SMichael Roth Error **errp) 2362f6229214SMichael Roth { 2363f6229214SMichael Roth sPAPRMachineState *spapr = SPAPR_MACHINE(obj); 2364f6229214SMichael Roth 2365f6229214SMichael Roth spapr->use_hotplug_event_source = value; 2366f6229214SMichael Roth } 2367f6229214SMichael Roth 236823825581SEduardo Habkost static void spapr_machine_initfn(Object *obj) 236923825581SEduardo Habkost { 2370715c5407SDavid Gibson sPAPRMachineState *spapr = SPAPR_MACHINE(obj); 2371715c5407SDavid Gibson 2372715c5407SDavid Gibson spapr->htab_fd = -1; 2373f6229214SMichael Roth spapr->use_hotplug_event_source = true; 237423825581SEduardo Habkost object_property_add_str(obj, "kvm-type", 237523825581SEduardo Habkost spapr_get_kvm_type, spapr_set_kvm_type, NULL); 237649d2e648SMarcel Apfelbaum object_property_set_description(obj, "kvm-type", 237749d2e648SMarcel Apfelbaum "Specifies the KVM virtualization mode (HV, PR)", 237849d2e648SMarcel Apfelbaum NULL); 2379f6229214SMichael Roth object_property_add_bool(obj, "modern-hotplug-events", 2380f6229214SMichael Roth spapr_get_modern_hotplug_events, 2381f6229214SMichael Roth spapr_set_modern_hotplug_events, 2382f6229214SMichael Roth NULL); 2383f6229214SMichael Roth object_property_set_description(obj, "modern-hotplug-events", 2384f6229214SMichael Roth "Use dedicated hotplug event mechanism in" 2385f6229214SMichael Roth " place of standard EPOW events when possible" 2386f6229214SMichael Roth " (required for memory hot-unplug support)", 2387f6229214SMichael Roth NULL); 238823825581SEduardo Habkost } 238923825581SEduardo Habkost 239087bbdd9cSDavid Gibson static void spapr_machine_finalizefn(Object *obj) 239187bbdd9cSDavid Gibson { 239287bbdd9cSDavid Gibson sPAPRMachineState *spapr = SPAPR_MACHINE(obj); 239387bbdd9cSDavid Gibson 239487bbdd9cSDavid Gibson g_free(spapr->kvm_type); 239587bbdd9cSDavid Gibson } 239687bbdd9cSDavid Gibson 23971c7ad77eSNicholas Piggin void spapr_do_system_reset_on_cpu(CPUState *cs, run_on_cpu_data arg) 239834316482SAlexey Kardashevskiy { 239934316482SAlexey Kardashevskiy cpu_synchronize_state(cs); 240034316482SAlexey Kardashevskiy ppc_cpu_do_system_reset(cs); 240134316482SAlexey Kardashevskiy } 240234316482SAlexey Kardashevskiy 240334316482SAlexey Kardashevskiy static void spapr_nmi(NMIState *n, int cpu_index, Error **errp) 240434316482SAlexey Kardashevskiy { 240534316482SAlexey Kardashevskiy CPUState *cs; 240634316482SAlexey Kardashevskiy 240734316482SAlexey Kardashevskiy CPU_FOREACH(cs) { 24081c7ad77eSNicholas Piggin async_run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL); 240934316482SAlexey Kardashevskiy } 241034316482SAlexey Kardashevskiy } 241134316482SAlexey Kardashevskiy 241279b78a6bSMichael Roth static void spapr_add_lmbs(DeviceState *dev, uint64_t addr_start, uint64_t size, 241379b78a6bSMichael Roth uint32_t node, bool dedicated_hp_event_source, 241479b78a6bSMichael Roth Error **errp) 2415c20d332aSBharata B Rao { 2416c20d332aSBharata B Rao sPAPRDRConnector *drc; 2417c20d332aSBharata B Rao sPAPRDRConnectorClass *drck; 2418c20d332aSBharata B Rao uint32_t nr_lmbs = size/SPAPR_MEMORY_BLOCK_SIZE; 2419c20d332aSBharata B Rao int i, fdt_offset, fdt_size; 2420c20d332aSBharata B Rao void *fdt; 242179b78a6bSMichael Roth uint64_t addr = addr_start; 2422c20d332aSBharata B Rao 2423c20d332aSBharata B Rao for (i = 0; i < nr_lmbs; i++) { 2424c20d332aSBharata B Rao drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB, 2425c20d332aSBharata B Rao addr/SPAPR_MEMORY_BLOCK_SIZE); 2426c20d332aSBharata B Rao g_assert(drc); 2427c20d332aSBharata B Rao 2428c20d332aSBharata B Rao fdt = create_device_tree(&fdt_size); 2429c20d332aSBharata B Rao fdt_offset = spapr_populate_memory_node(fdt, node, addr, 2430c20d332aSBharata B Rao SPAPR_MEMORY_BLOCK_SIZE); 2431c20d332aSBharata B Rao 2432c20d332aSBharata B Rao drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); 2433c20d332aSBharata B Rao drck->attach(drc, dev, fdt, fdt_offset, !dev->hotplugged, errp); 2434c20d332aSBharata B Rao addr += SPAPR_MEMORY_BLOCK_SIZE; 24355c0139a8SMichael Roth if (!dev->hotplugged) { 24365c0139a8SMichael Roth /* guests expect coldplugged LMBs to be pre-allocated */ 24375c0139a8SMichael Roth drck->set_allocation_state(drc, SPAPR_DR_ALLOCATION_STATE_USABLE); 24385c0139a8SMichael Roth drck->set_isolation_state(drc, SPAPR_DR_ISOLATION_STATE_UNISOLATED); 24395c0139a8SMichael Roth } 2440c20d332aSBharata B Rao } 24415dd5238cSJianjun Duan /* send hotplug notification to the 24425dd5238cSJianjun Duan * guest only in case of hotplugged memory 24435dd5238cSJianjun Duan */ 24445dd5238cSJianjun Duan if (dev->hotplugged) { 244579b78a6bSMichael Roth if (dedicated_hp_event_source) { 244679b78a6bSMichael Roth drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB, 244779b78a6bSMichael Roth addr_start / SPAPR_MEMORY_BLOCK_SIZE); 244879b78a6bSMichael Roth drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); 244979b78a6bSMichael Roth spapr_hotplug_req_add_by_count_indexed(SPAPR_DR_CONNECTOR_TYPE_LMB, 245079b78a6bSMichael Roth nr_lmbs, 245179b78a6bSMichael Roth drck->get_index(drc)); 245279b78a6bSMichael Roth } else { 245379b78a6bSMichael Roth spapr_hotplug_req_add_by_count(SPAPR_DR_CONNECTOR_TYPE_LMB, 245479b78a6bSMichael Roth nr_lmbs); 245579b78a6bSMichael Roth } 2456c20d332aSBharata B Rao } 24575dd5238cSJianjun Duan } 2458c20d332aSBharata B Rao 2459c20d332aSBharata B Rao static void spapr_memory_plug(HotplugHandler *hotplug_dev, DeviceState *dev, 2460c20d332aSBharata B Rao uint32_t node, Error **errp) 2461c20d332aSBharata B Rao { 2462c20d332aSBharata B Rao Error *local_err = NULL; 2463c20d332aSBharata B Rao sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev); 2464c20d332aSBharata B Rao PCDIMMDevice *dimm = PC_DIMM(dev); 2465c20d332aSBharata B Rao PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); 2466c20d332aSBharata B Rao MemoryRegion *mr = ddc->get_memory_region(dimm); 2467c20d332aSBharata B Rao uint64_t align = memory_region_get_alignment(mr); 2468c20d332aSBharata B Rao uint64_t size = memory_region_size(mr); 2469c20d332aSBharata B Rao uint64_t addr; 2470df587133SThomas Huth char *mem_dev; 2471c20d332aSBharata B Rao 2472c20d332aSBharata B Rao if (size % SPAPR_MEMORY_BLOCK_SIZE) { 2473c20d332aSBharata B Rao error_setg(&local_err, "Hotplugged memory size must be a multiple of " 2474c20d332aSBharata B Rao "%lld MB", SPAPR_MEMORY_BLOCK_SIZE/M_BYTE); 2475c20d332aSBharata B Rao goto out; 2476c20d332aSBharata B Rao } 2477c20d332aSBharata B Rao 2478df587133SThomas Huth mem_dev = object_property_get_str(OBJECT(dimm), PC_DIMM_MEMDEV_PROP, NULL); 2479df587133SThomas Huth if (mem_dev && !kvmppc_is_mem_backend_page_size_ok(mem_dev)) { 2480df587133SThomas Huth error_setg(&local_err, "Memory backend has bad page size. " 2481df587133SThomas Huth "Use 'memory-backend-file' with correct mem-path."); 2482df587133SThomas Huth goto out; 2483df587133SThomas Huth } 2484df587133SThomas Huth 2485d6a9b0b8SMichael S. Tsirkin pc_dimm_memory_plug(dev, &ms->hotplug_memory, mr, align, &local_err); 2486c20d332aSBharata B Rao if (local_err) { 2487c20d332aSBharata B Rao goto out; 2488c20d332aSBharata B Rao } 2489c20d332aSBharata B Rao 2490c20d332aSBharata B Rao addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, &local_err); 2491c20d332aSBharata B Rao if (local_err) { 2492c20d332aSBharata B Rao pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr); 2493c20d332aSBharata B Rao goto out; 2494c20d332aSBharata B Rao } 2495c20d332aSBharata B Rao 249679b78a6bSMichael Roth spapr_add_lmbs(dev, addr, size, node, 249779b78a6bSMichael Roth spapr_ovec_test(ms->ov5_cas, OV5_HP_EVT), 249879b78a6bSMichael Roth &error_abort); 2499c20d332aSBharata B Rao 2500c20d332aSBharata B Rao out: 2501c20d332aSBharata B Rao error_propagate(errp, local_err); 2502c20d332aSBharata B Rao } 2503c20d332aSBharata B Rao 2504cf632463SBharata B Rao typedef struct sPAPRDIMMState { 2505cf632463SBharata B Rao uint32_t nr_lmbs; 2506cf632463SBharata B Rao } sPAPRDIMMState; 2507cf632463SBharata B Rao 2508cf632463SBharata B Rao static void spapr_lmb_release(DeviceState *dev, void *opaque) 2509cf632463SBharata B Rao { 2510cf632463SBharata B Rao sPAPRDIMMState *ds = (sPAPRDIMMState *)opaque; 2511cf632463SBharata B Rao HotplugHandler *hotplug_ctrl; 2512cf632463SBharata B Rao 2513cf632463SBharata B Rao if (--ds->nr_lmbs) { 2514cf632463SBharata B Rao return; 2515cf632463SBharata B Rao } 2516cf632463SBharata B Rao 2517cf632463SBharata B Rao g_free(ds); 2518cf632463SBharata B Rao 2519cf632463SBharata B Rao /* 2520cf632463SBharata B Rao * Now that all the LMBs have been removed by the guest, call the 2521cf632463SBharata B Rao * pc-dimm unplug handler to cleanup up the pc-dimm device. 2522cf632463SBharata B Rao */ 2523cf632463SBharata B Rao hotplug_ctrl = qdev_get_hotplug_handler(dev); 2524cf632463SBharata B Rao hotplug_handler_unplug(hotplug_ctrl, dev, &error_abort); 2525cf632463SBharata B Rao } 2526cf632463SBharata B Rao 2527cf632463SBharata B Rao static void spapr_del_lmbs(DeviceState *dev, uint64_t addr_start, uint64_t size, 2528cf632463SBharata B Rao Error **errp) 2529cf632463SBharata B Rao { 2530cf632463SBharata B Rao sPAPRDRConnector *drc; 2531cf632463SBharata B Rao sPAPRDRConnectorClass *drck; 2532cf632463SBharata B Rao uint32_t nr_lmbs = size / SPAPR_MEMORY_BLOCK_SIZE; 2533cf632463SBharata B Rao int i; 2534cf632463SBharata B Rao sPAPRDIMMState *ds = g_malloc0(sizeof(sPAPRDIMMState)); 2535cf632463SBharata B Rao uint64_t addr = addr_start; 2536cf632463SBharata B Rao 2537cf632463SBharata B Rao ds->nr_lmbs = nr_lmbs; 2538cf632463SBharata B Rao for (i = 0; i < nr_lmbs; i++) { 2539cf632463SBharata B Rao drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB, 2540cf632463SBharata B Rao addr / SPAPR_MEMORY_BLOCK_SIZE); 2541cf632463SBharata B Rao g_assert(drc); 2542cf632463SBharata B Rao 2543cf632463SBharata B Rao drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); 2544cf632463SBharata B Rao drck->detach(drc, dev, spapr_lmb_release, ds, errp); 2545cf632463SBharata B Rao addr += SPAPR_MEMORY_BLOCK_SIZE; 2546cf632463SBharata B Rao } 2547cf632463SBharata B Rao 2548cf632463SBharata B Rao drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB, 2549cf632463SBharata B Rao addr_start / SPAPR_MEMORY_BLOCK_SIZE); 2550cf632463SBharata B Rao drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); 2551cf632463SBharata B Rao spapr_hotplug_req_remove_by_count_indexed(SPAPR_DR_CONNECTOR_TYPE_LMB, 2552cf632463SBharata B Rao nr_lmbs, 2553cf632463SBharata B Rao drck->get_index(drc)); 2554cf632463SBharata B Rao } 2555cf632463SBharata B Rao 2556cf632463SBharata B Rao static void spapr_memory_unplug(HotplugHandler *hotplug_dev, DeviceState *dev, 2557cf632463SBharata B Rao Error **errp) 2558cf632463SBharata B Rao { 2559cf632463SBharata B Rao sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev); 2560cf632463SBharata B Rao PCDIMMDevice *dimm = PC_DIMM(dev); 2561cf632463SBharata B Rao PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); 2562cf632463SBharata B Rao MemoryRegion *mr = ddc->get_memory_region(dimm); 2563cf632463SBharata B Rao 2564cf632463SBharata B Rao pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr); 2565cf632463SBharata B Rao object_unparent(OBJECT(dev)); 2566cf632463SBharata B Rao } 2567cf632463SBharata B Rao 2568cf632463SBharata B Rao static void spapr_memory_unplug_request(HotplugHandler *hotplug_dev, 2569cf632463SBharata B Rao DeviceState *dev, Error **errp) 2570cf632463SBharata B Rao { 2571cf632463SBharata B Rao Error *local_err = NULL; 2572cf632463SBharata B Rao PCDIMMDevice *dimm = PC_DIMM(dev); 2573cf632463SBharata B Rao PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); 2574cf632463SBharata B Rao MemoryRegion *mr = ddc->get_memory_region(dimm); 2575cf632463SBharata B Rao uint64_t size = memory_region_size(mr); 2576cf632463SBharata B Rao uint64_t addr; 2577cf632463SBharata B Rao 2578cf632463SBharata B Rao addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, &local_err); 2579cf632463SBharata B Rao if (local_err) { 2580cf632463SBharata B Rao goto out; 2581cf632463SBharata B Rao } 2582cf632463SBharata B Rao 2583cf632463SBharata B Rao spapr_del_lmbs(dev, addr, size, &error_abort); 2584cf632463SBharata B Rao out: 2585cf632463SBharata B Rao error_propagate(errp, local_err); 2586cf632463SBharata B Rao } 2587cf632463SBharata B Rao 2588af81cf32SBharata B Rao void *spapr_populate_hotplug_cpu_dt(CPUState *cs, int *fdt_offset, 2589af81cf32SBharata B Rao sPAPRMachineState *spapr) 2590af81cf32SBharata B Rao { 2591af81cf32SBharata B Rao PowerPCCPU *cpu = POWERPC_CPU(cs); 2592af81cf32SBharata B Rao DeviceClass *dc = DEVICE_GET_CLASS(cs); 2593af81cf32SBharata B Rao int id = ppc_get_vcpu_dt_id(cpu); 2594af81cf32SBharata B Rao void *fdt; 2595af81cf32SBharata B Rao int offset, fdt_size; 2596af81cf32SBharata B Rao char *nodename; 2597af81cf32SBharata B Rao 2598af81cf32SBharata B Rao fdt = create_device_tree(&fdt_size); 2599af81cf32SBharata B Rao nodename = g_strdup_printf("%s@%x", dc->fw_name, id); 2600af81cf32SBharata B Rao offset = fdt_add_subnode(fdt, 0, nodename); 2601af81cf32SBharata B Rao 2602af81cf32SBharata B Rao spapr_populate_cpu_dt(cs, fdt, offset, spapr); 2603af81cf32SBharata B Rao g_free(nodename); 2604af81cf32SBharata B Rao 2605af81cf32SBharata B Rao *fdt_offset = offset; 2606af81cf32SBharata B Rao return fdt; 2607af81cf32SBharata B Rao } 2608af81cf32SBharata B Rao 2609115debf2SIgor Mammedov static void spapr_core_unplug(HotplugHandler *hotplug_dev, DeviceState *dev, 2610115debf2SIgor Mammedov Error **errp) 2611ff9006ddSIgor Mammedov { 2612535455fdSIgor Mammedov MachineState *ms = MACHINE(qdev_get_machine()); 2613ff9006ddSIgor Mammedov CPUCore *cc = CPU_CORE(dev); 2614535455fdSIgor Mammedov CPUArchId *core_slot = spapr_find_cpu_slot(ms, cc->core_id, NULL); 2615ff9006ddSIgor Mammedov 2616535455fdSIgor Mammedov core_slot->cpu = NULL; 2617ff9006ddSIgor Mammedov object_unparent(OBJECT(dev)); 2618ff9006ddSIgor Mammedov } 2619ff9006ddSIgor Mammedov 2620115debf2SIgor Mammedov static void spapr_core_release(DeviceState *dev, void *opaque) 2621115debf2SIgor Mammedov { 2622115debf2SIgor Mammedov HotplugHandler *hotplug_ctrl; 2623115debf2SIgor Mammedov 2624115debf2SIgor Mammedov hotplug_ctrl = qdev_get_hotplug_handler(dev); 2625115debf2SIgor Mammedov hotplug_handler_unplug(hotplug_ctrl, dev, &error_abort); 2626115debf2SIgor Mammedov } 2627115debf2SIgor Mammedov 2628115debf2SIgor Mammedov static 2629115debf2SIgor Mammedov void spapr_core_unplug_request(HotplugHandler *hotplug_dev, DeviceState *dev, 2630ff9006ddSIgor Mammedov Error **errp) 2631ff9006ddSIgor Mammedov { 2632535455fdSIgor Mammedov int index; 2633535455fdSIgor Mammedov sPAPRDRConnector *drc; 2634ff9006ddSIgor Mammedov sPAPRDRConnectorClass *drck; 2635ff9006ddSIgor Mammedov Error *local_err = NULL; 2636535455fdSIgor Mammedov CPUCore *cc = CPU_CORE(dev); 2637535455fdSIgor Mammedov int smt = kvmppc_smt_threads(); 2638ff9006ddSIgor Mammedov 2639535455fdSIgor Mammedov if (!spapr_find_cpu_slot(MACHINE(hotplug_dev), cc->core_id, &index)) { 2640535455fdSIgor Mammedov error_setg(errp, "Unable to find CPU core with core-id: %d", 2641535455fdSIgor Mammedov cc->core_id); 2642535455fdSIgor Mammedov return; 2643535455fdSIgor Mammedov } 2644ff9006ddSIgor Mammedov if (index == 0) { 2645ff9006ddSIgor Mammedov error_setg(errp, "Boot CPU core may not be unplugged"); 2646ff9006ddSIgor Mammedov return; 2647ff9006ddSIgor Mammedov } 2648ff9006ddSIgor Mammedov 2649535455fdSIgor Mammedov drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_CPU, index * smt); 2650ff9006ddSIgor Mammedov g_assert(drc); 2651ff9006ddSIgor Mammedov 2652ff9006ddSIgor Mammedov drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); 2653ff9006ddSIgor Mammedov drck->detach(drc, dev, spapr_core_release, NULL, &local_err); 2654ff9006ddSIgor Mammedov if (local_err) { 2655ff9006ddSIgor Mammedov error_propagate(errp, local_err); 2656ff9006ddSIgor Mammedov return; 2657ff9006ddSIgor Mammedov } 2658ff9006ddSIgor Mammedov 2659ff9006ddSIgor Mammedov spapr_hotplug_req_remove_by_index(drc); 2660ff9006ddSIgor Mammedov } 2661ff9006ddSIgor Mammedov 2662ff9006ddSIgor Mammedov static void spapr_core_plug(HotplugHandler *hotplug_dev, DeviceState *dev, 2663ff9006ddSIgor Mammedov Error **errp) 2664ff9006ddSIgor Mammedov { 2665ff9006ddSIgor Mammedov sPAPRMachineState *spapr = SPAPR_MACHINE(OBJECT(hotplug_dev)); 2666ff9006ddSIgor Mammedov MachineClass *mc = MACHINE_GET_CLASS(spapr); 2667ff9006ddSIgor Mammedov sPAPRCPUCore *core = SPAPR_CPU_CORE(OBJECT(dev)); 2668ff9006ddSIgor Mammedov CPUCore *cc = CPU_CORE(dev); 2669ff9006ddSIgor Mammedov CPUState *cs = CPU(core->threads); 2670ff9006ddSIgor Mammedov sPAPRDRConnector *drc; 2671ff9006ddSIgor Mammedov Error *local_err = NULL; 2672ff9006ddSIgor Mammedov void *fdt = NULL; 2673ff9006ddSIgor Mammedov int fdt_offset = 0; 2674ff9006ddSIgor Mammedov int smt = kvmppc_smt_threads(); 2675535455fdSIgor Mammedov CPUArchId *core_slot; 2676535455fdSIgor Mammedov int index; 2677ff9006ddSIgor Mammedov 2678535455fdSIgor Mammedov core_slot = spapr_find_cpu_slot(MACHINE(hotplug_dev), cc->core_id, &index); 2679535455fdSIgor Mammedov if (!core_slot) { 2680535455fdSIgor Mammedov error_setg(errp, "Unable to find CPU core with core-id: %d", 2681535455fdSIgor Mammedov cc->core_id); 2682535455fdSIgor Mammedov return; 2683535455fdSIgor Mammedov } 2684ff9006ddSIgor Mammedov drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_CPU, index * smt); 2685ff9006ddSIgor Mammedov 2686c5514d0eSIgor Mammedov g_assert(drc || !mc->has_hotpluggable_cpus); 2687ff9006ddSIgor Mammedov 2688ff9006ddSIgor Mammedov /* 2689ff9006ddSIgor Mammedov * Setup CPU DT entries only for hotplugged CPUs. For boot time or 2690ff9006ddSIgor Mammedov * coldplugged CPUs DT entries are setup in spapr_build_fdt(). 2691ff9006ddSIgor Mammedov */ 2692ff9006ddSIgor Mammedov if (dev->hotplugged) { 2693ff9006ddSIgor Mammedov fdt = spapr_populate_hotplug_cpu_dt(cs, &fdt_offset, spapr); 2694ff9006ddSIgor Mammedov } 2695ff9006ddSIgor Mammedov 2696ff9006ddSIgor Mammedov if (drc) { 2697ff9006ddSIgor Mammedov sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); 2698ff9006ddSIgor Mammedov drck->attach(drc, dev, fdt, fdt_offset, !dev->hotplugged, &local_err); 2699ff9006ddSIgor Mammedov if (local_err) { 2700ff9006ddSIgor Mammedov g_free(fdt); 2701ff9006ddSIgor Mammedov error_propagate(errp, local_err); 2702ff9006ddSIgor Mammedov return; 2703ff9006ddSIgor Mammedov } 2704ff9006ddSIgor Mammedov } 2705ff9006ddSIgor Mammedov 2706ff9006ddSIgor Mammedov if (dev->hotplugged) { 2707ff9006ddSIgor Mammedov /* 2708ff9006ddSIgor Mammedov * Send hotplug notification interrupt to the guest only in case 2709ff9006ddSIgor Mammedov * of hotplugged CPUs. 2710ff9006ddSIgor Mammedov */ 2711ff9006ddSIgor Mammedov spapr_hotplug_req_add_by_index(drc); 2712ff9006ddSIgor Mammedov } else { 2713ff9006ddSIgor Mammedov /* 2714ff9006ddSIgor Mammedov * Set the right DRC states for cold plugged CPU. 2715ff9006ddSIgor Mammedov */ 2716ff9006ddSIgor Mammedov if (drc) { 2717ff9006ddSIgor Mammedov sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); 2718ff9006ddSIgor Mammedov drck->set_allocation_state(drc, SPAPR_DR_ALLOCATION_STATE_USABLE); 2719ff9006ddSIgor Mammedov drck->set_isolation_state(drc, SPAPR_DR_ISOLATION_STATE_UNISOLATED); 2720ff9006ddSIgor Mammedov } 2721ff9006ddSIgor Mammedov } 2722535455fdSIgor Mammedov core_slot->cpu = OBJECT(dev); 2723ff9006ddSIgor Mammedov } 2724ff9006ddSIgor Mammedov 2725ff9006ddSIgor Mammedov static void spapr_core_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev, 2726ff9006ddSIgor Mammedov Error **errp) 2727ff9006ddSIgor Mammedov { 2728ff9006ddSIgor Mammedov MachineState *machine = MACHINE(OBJECT(hotplug_dev)); 2729ff9006ddSIgor Mammedov MachineClass *mc = MACHINE_GET_CLASS(hotplug_dev); 2730ff9006ddSIgor Mammedov Error *local_err = NULL; 2731ff9006ddSIgor Mammedov CPUCore *cc = CPU_CORE(dev); 2732ff9006ddSIgor Mammedov char *base_core_type = spapr_get_cpu_core_type(machine->cpu_model); 2733ff9006ddSIgor Mammedov const char *type = object_get_typename(OBJECT(dev)); 2734535455fdSIgor Mammedov CPUArchId *core_slot; 2735535455fdSIgor Mammedov int index; 2736ff9006ddSIgor Mammedov 2737c5514d0eSIgor Mammedov if (dev->hotplugged && !mc->has_hotpluggable_cpus) { 2738ff9006ddSIgor Mammedov error_setg(&local_err, "CPU hotplug not supported for this machine"); 2739ff9006ddSIgor Mammedov goto out; 2740ff9006ddSIgor Mammedov } 2741ff9006ddSIgor Mammedov 2742ff9006ddSIgor Mammedov if (strcmp(base_core_type, type)) { 2743ff9006ddSIgor Mammedov error_setg(&local_err, "CPU core type should be %s", base_core_type); 2744ff9006ddSIgor Mammedov goto out; 2745ff9006ddSIgor Mammedov } 2746ff9006ddSIgor Mammedov 2747ff9006ddSIgor Mammedov if (cc->core_id % smp_threads) { 2748ff9006ddSIgor Mammedov error_setg(&local_err, "invalid core id %d", cc->core_id); 2749ff9006ddSIgor Mammedov goto out; 2750ff9006ddSIgor Mammedov } 2751ff9006ddSIgor Mammedov 2752535455fdSIgor Mammedov core_slot = spapr_find_cpu_slot(MACHINE(hotplug_dev), cc->core_id, &index); 2753535455fdSIgor Mammedov if (!core_slot) { 2754ff9006ddSIgor Mammedov error_setg(&local_err, "core id %d out of range", cc->core_id); 2755ff9006ddSIgor Mammedov goto out; 2756ff9006ddSIgor Mammedov } 2757ff9006ddSIgor Mammedov 2758535455fdSIgor Mammedov if (core_slot->cpu) { 2759ff9006ddSIgor Mammedov error_setg(&local_err, "core %d already populated", cc->core_id); 2760ff9006ddSIgor Mammedov goto out; 2761ff9006ddSIgor Mammedov } 2762ff9006ddSIgor Mammedov 2763ff9006ddSIgor Mammedov out: 2764ff9006ddSIgor Mammedov g_free(base_core_type); 2765ff9006ddSIgor Mammedov error_propagate(errp, local_err); 2766ff9006ddSIgor Mammedov } 2767ff9006ddSIgor Mammedov 2768c20d332aSBharata B Rao static void spapr_machine_device_plug(HotplugHandler *hotplug_dev, 2769c20d332aSBharata B Rao DeviceState *dev, Error **errp) 2770c20d332aSBharata B Rao { 2771c20d332aSBharata B Rao sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine()); 2772c20d332aSBharata B Rao 2773c20d332aSBharata B Rao if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { 2774b556854bSBharata B Rao int node; 2775c20d332aSBharata B Rao 2776c20d332aSBharata B Rao if (!smc->dr_lmb_enabled) { 2777c20d332aSBharata B Rao error_setg(errp, "Memory hotplug not supported for this machine"); 2778c20d332aSBharata B Rao return; 2779c20d332aSBharata B Rao } 2780c20d332aSBharata B Rao node = object_property_get_int(OBJECT(dev), PC_DIMM_NODE_PROP, errp); 2781c20d332aSBharata B Rao if (*errp) { 2782c20d332aSBharata B Rao return; 2783c20d332aSBharata B Rao } 27841a5512bbSGonglei if (node < 0 || node >= MAX_NODES) { 27851a5512bbSGonglei error_setg(errp, "Invaild node %d", node); 27861a5512bbSGonglei return; 27871a5512bbSGonglei } 2788c20d332aSBharata B Rao 2789b556854bSBharata B Rao /* 2790b556854bSBharata B Rao * Currently PowerPC kernel doesn't allow hot-adding memory to 2791b556854bSBharata B Rao * memory-less node, but instead will silently add the memory 2792b556854bSBharata B Rao * to the first node that has some memory. This causes two 2793b556854bSBharata B Rao * unexpected behaviours for the user. 2794b556854bSBharata B Rao * 2795b556854bSBharata B Rao * - Memory gets hotplugged to a different node than what the user 2796b556854bSBharata B Rao * specified. 2797b556854bSBharata B Rao * - Since pc-dimm subsystem in QEMU still thinks that memory belongs 2798b556854bSBharata B Rao * to memory-less node, a reboot will set things accordingly 2799b556854bSBharata B Rao * and the previously hotplugged memory now ends in the right node. 2800b556854bSBharata B Rao * This appears as if some memory moved from one node to another. 2801b556854bSBharata B Rao * 2802b556854bSBharata B Rao * So until kernel starts supporting memory hotplug to memory-less 2803b556854bSBharata B Rao * nodes, just prevent such attempts upfront in QEMU. 2804b556854bSBharata B Rao */ 2805b556854bSBharata B Rao if (nb_numa_nodes && !numa_info[node].node_mem) { 2806b556854bSBharata B Rao error_setg(errp, "Can't hotplug memory to memory-less node %d", 2807b556854bSBharata B Rao node); 2808b556854bSBharata B Rao return; 2809b556854bSBharata B Rao } 2810b556854bSBharata B Rao 2811c20d332aSBharata B Rao spapr_memory_plug(hotplug_dev, dev, node, errp); 2812af81cf32SBharata B Rao } else if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { 2813af81cf32SBharata B Rao spapr_core_plug(hotplug_dev, dev, errp); 2814c20d332aSBharata B Rao } 2815c20d332aSBharata B Rao } 2816c20d332aSBharata B Rao 2817c20d332aSBharata B Rao static void spapr_machine_device_unplug(HotplugHandler *hotplug_dev, 2818c20d332aSBharata B Rao DeviceState *dev, Error **errp) 2819c20d332aSBharata B Rao { 2820cf632463SBharata B Rao sPAPRMachineState *sms = SPAPR_MACHINE(qdev_get_machine()); 28213c0c47e3SDavid Gibson MachineClass *mc = MACHINE_GET_CLASS(qdev_get_machine()); 28226f4b5c3eSBharata B Rao 2823c20d332aSBharata B Rao if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { 2824cf632463SBharata B Rao if (spapr_ovec_test(sms->ov5_cas, OV5_HP_EVT)) { 2825cf632463SBharata B Rao spapr_memory_unplug(hotplug_dev, dev, errp); 2826cf632463SBharata B Rao } else { 2827cf632463SBharata B Rao error_setg(errp, "Memory hot unplug not supported for this guest"); 2828cf632463SBharata B Rao } 2829cf632463SBharata B Rao } else if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { 2830c5514d0eSIgor Mammedov if (!mc->has_hotpluggable_cpus) { 2831cf632463SBharata B Rao error_setg(errp, "CPU hot unplug not supported on this machine"); 2832cf632463SBharata B Rao return; 2833cf632463SBharata B Rao } 2834cf632463SBharata B Rao spapr_core_unplug(hotplug_dev, dev, errp); 2835cf632463SBharata B Rao } 2836cf632463SBharata B Rao } 2837cf632463SBharata B Rao 2838cf632463SBharata B Rao static void spapr_machine_device_unplug_request(HotplugHandler *hotplug_dev, 2839cf632463SBharata B Rao DeviceState *dev, Error **errp) 2840cf632463SBharata B Rao { 2841cf632463SBharata B Rao sPAPRMachineState *sms = SPAPR_MACHINE(qdev_get_machine()); 2842cf632463SBharata B Rao MachineClass *mc = MACHINE_GET_CLASS(qdev_get_machine()); 2843cf632463SBharata B Rao 2844cf632463SBharata B Rao if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { 2845cf632463SBharata B Rao if (spapr_ovec_test(sms->ov5_cas, OV5_HP_EVT)) { 2846cf632463SBharata B Rao spapr_memory_unplug_request(hotplug_dev, dev, errp); 2847cf632463SBharata B Rao } else { 2848cf632463SBharata B Rao /* NOTE: this means there is a window after guest reset, prior to 2849cf632463SBharata B Rao * CAS negotiation, where unplug requests will fail due to the 2850cf632463SBharata B Rao * capability not being detected yet. This is a bit different than 2851cf632463SBharata B Rao * the case with PCI unplug, where the events will be queued and 2852cf632463SBharata B Rao * eventually handled by the guest after boot 2853cf632463SBharata B Rao */ 2854cf632463SBharata B Rao error_setg(errp, "Memory hot unplug not supported for this guest"); 2855cf632463SBharata B Rao } 28566f4b5c3eSBharata B Rao } else if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { 2857c5514d0eSIgor Mammedov if (!mc->has_hotpluggable_cpus) { 28586f4b5c3eSBharata B Rao error_setg(errp, "CPU hot unplug not supported on this machine"); 28596f4b5c3eSBharata B Rao return; 28606f4b5c3eSBharata B Rao } 2861115debf2SIgor Mammedov spapr_core_unplug_request(hotplug_dev, dev, errp); 2862c20d332aSBharata B Rao } 2863c20d332aSBharata B Rao } 2864c20d332aSBharata B Rao 286594a94e4cSBharata B Rao static void spapr_machine_device_pre_plug(HotplugHandler *hotplug_dev, 286694a94e4cSBharata B Rao DeviceState *dev, Error **errp) 286794a94e4cSBharata B Rao { 286894a94e4cSBharata B Rao if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { 286994a94e4cSBharata B Rao spapr_core_pre_plug(hotplug_dev, dev, errp); 287094a94e4cSBharata B Rao } 287194a94e4cSBharata B Rao } 287294a94e4cSBharata B Rao 28737ebaf795SBharata B Rao static HotplugHandler *spapr_get_hotplug_handler(MachineState *machine, 2874c20d332aSBharata B Rao DeviceState *dev) 2875c20d332aSBharata B Rao { 287694a94e4cSBharata B Rao if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) || 287794a94e4cSBharata B Rao object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { 2878c20d332aSBharata B Rao return HOTPLUG_HANDLER(machine); 2879c20d332aSBharata B Rao } 2880c20d332aSBharata B Rao return NULL; 2881c20d332aSBharata B Rao } 2882c20d332aSBharata B Rao 288320bb648dSDavid Gibson static unsigned spapr_cpu_index_to_socket_id(unsigned cpu_index) 288420bb648dSDavid Gibson { 288520bb648dSDavid Gibson /* Allocate to NUMA nodes on a "socket" basis (not that concept of 288620bb648dSDavid Gibson * socket means much for the paravirtualized PAPR platform) */ 288720bb648dSDavid Gibson return cpu_index / smp_threads / smp_cores; 288820bb648dSDavid Gibson } 288920bb648dSDavid Gibson 2890535455fdSIgor Mammedov static const CPUArchIdList *spapr_possible_cpu_arch_ids(MachineState *machine) 2891535455fdSIgor Mammedov { 2892535455fdSIgor Mammedov int i; 2893535455fdSIgor Mammedov int spapr_max_cores = max_cpus / smp_threads; 2894535455fdSIgor Mammedov MachineClass *mc = MACHINE_GET_CLASS(machine); 2895535455fdSIgor Mammedov 2896c5514d0eSIgor Mammedov if (!mc->has_hotpluggable_cpus) { 2897535455fdSIgor Mammedov spapr_max_cores = QEMU_ALIGN_UP(smp_cpus, smp_threads) / smp_threads; 2898535455fdSIgor Mammedov } 2899535455fdSIgor Mammedov if (machine->possible_cpus) { 2900535455fdSIgor Mammedov assert(machine->possible_cpus->len == spapr_max_cores); 2901535455fdSIgor Mammedov return machine->possible_cpus; 2902535455fdSIgor Mammedov } 2903535455fdSIgor Mammedov 2904535455fdSIgor Mammedov machine->possible_cpus = g_malloc0(sizeof(CPUArchIdList) + 2905535455fdSIgor Mammedov sizeof(CPUArchId) * spapr_max_cores); 2906535455fdSIgor Mammedov machine->possible_cpus->len = spapr_max_cores; 2907535455fdSIgor Mammedov for (i = 0; i < machine->possible_cpus->len; i++) { 2908535455fdSIgor Mammedov int core_id = i * smp_threads; 2909535455fdSIgor Mammedov 2910f2d672c2SIgor Mammedov machine->possible_cpus->cpus[i].vcpus_count = smp_threads; 2911535455fdSIgor Mammedov machine->possible_cpus->cpus[i].arch_id = core_id; 2912535455fdSIgor Mammedov machine->possible_cpus->cpus[i].props.has_core_id = true; 2913535455fdSIgor Mammedov machine->possible_cpus->cpus[i].props.core_id = core_id; 2914535455fdSIgor Mammedov /* TODO: add 'has_node/node' here to describe 2915535455fdSIgor Mammedov to which node core belongs */ 2916535455fdSIgor Mammedov } 2917535455fdSIgor Mammedov return machine->possible_cpus; 2918535455fdSIgor Mammedov } 2919535455fdSIgor Mammedov 29206737d9adSDavid Gibson static void spapr_phb_placement(sPAPRMachineState *spapr, uint32_t index, 2921daa23699SDavid Gibson uint64_t *buid, hwaddr *pio, 2922daa23699SDavid Gibson hwaddr *mmio32, hwaddr *mmio64, 29236737d9adSDavid Gibson unsigned n_dma, uint32_t *liobns, Error **errp) 29246737d9adSDavid Gibson { 2925357d1e3bSDavid Gibson /* 2926357d1e3bSDavid Gibson * New-style PHB window placement. 2927357d1e3bSDavid Gibson * 2928357d1e3bSDavid Gibson * Goals: Gives large (1TiB), naturally aligned 64-bit MMIO window 2929357d1e3bSDavid Gibson * for each PHB, in addition to 2GiB 32-bit MMIO and 64kiB PIO 2930357d1e3bSDavid Gibson * windows. 2931357d1e3bSDavid Gibson * 2932357d1e3bSDavid Gibson * Some guest kernels can't work with MMIO windows above 1<<46 2933357d1e3bSDavid Gibson * (64TiB), so we place up to 31 PHBs in the area 32TiB..64TiB 2934357d1e3bSDavid Gibson * 2935357d1e3bSDavid Gibson * 32TiB..(33TiB+1984kiB) contains the 64kiB PIO windows for each 2936357d1e3bSDavid Gibson * PHB stacked together. (32TiB+2GiB)..(32TiB+64GiB) contains the 2937357d1e3bSDavid Gibson * 2GiB 32-bit MMIO windows for each PHB. Then 33..64TiB has the 2938357d1e3bSDavid Gibson * 1TiB 64-bit MMIO windows for each PHB. 2939357d1e3bSDavid Gibson */ 29406737d9adSDavid Gibson const uint64_t base_buid = 0x800000020000000ULL; 294125e6a118SMichael S. Tsirkin #define SPAPR_MAX_PHBS ((SPAPR_PCI_LIMIT - SPAPR_PCI_BASE) / \ 294225e6a118SMichael S. Tsirkin SPAPR_PCI_MEM64_WIN_SIZE - 1) 29436737d9adSDavid Gibson int i; 29446737d9adSDavid Gibson 2945357d1e3bSDavid Gibson /* Sanity check natural alignments */ 2946357d1e3bSDavid Gibson QEMU_BUILD_BUG_ON((SPAPR_PCI_BASE % SPAPR_PCI_MEM64_WIN_SIZE) != 0); 2947357d1e3bSDavid Gibson QEMU_BUILD_BUG_ON((SPAPR_PCI_LIMIT % SPAPR_PCI_MEM64_WIN_SIZE) != 0); 2948357d1e3bSDavid Gibson QEMU_BUILD_BUG_ON((SPAPR_PCI_MEM64_WIN_SIZE % SPAPR_PCI_MEM32_WIN_SIZE) != 0); 2949357d1e3bSDavid Gibson QEMU_BUILD_BUG_ON((SPAPR_PCI_MEM32_WIN_SIZE % SPAPR_PCI_IO_WIN_SIZE) != 0); 2950357d1e3bSDavid Gibson /* Sanity check bounds */ 295125e6a118SMichael S. Tsirkin QEMU_BUILD_BUG_ON((SPAPR_MAX_PHBS * SPAPR_PCI_IO_WIN_SIZE) > 295225e6a118SMichael S. Tsirkin SPAPR_PCI_MEM32_WIN_SIZE); 295325e6a118SMichael S. Tsirkin QEMU_BUILD_BUG_ON((SPAPR_MAX_PHBS * SPAPR_PCI_MEM32_WIN_SIZE) > 295425e6a118SMichael S. Tsirkin SPAPR_PCI_MEM64_WIN_SIZE); 29552efff1c0SDavid Gibson 295625e6a118SMichael S. Tsirkin if (index >= SPAPR_MAX_PHBS) { 295725e6a118SMichael S. Tsirkin error_setg(errp, "\"index\" for PAPR PHB is too large (max %llu)", 295825e6a118SMichael S. Tsirkin SPAPR_MAX_PHBS - 1); 29596737d9adSDavid Gibson return; 29606737d9adSDavid Gibson } 29616737d9adSDavid Gibson 29626737d9adSDavid Gibson *buid = base_buid + index; 29636737d9adSDavid Gibson for (i = 0; i < n_dma; ++i) { 29646737d9adSDavid Gibson liobns[i] = SPAPR_PCI_LIOBN(index, i); 29656737d9adSDavid Gibson } 29666737d9adSDavid Gibson 2967357d1e3bSDavid Gibson *pio = SPAPR_PCI_BASE + index * SPAPR_PCI_IO_WIN_SIZE; 2968357d1e3bSDavid Gibson *mmio32 = SPAPR_PCI_BASE + (index + 1) * SPAPR_PCI_MEM32_WIN_SIZE; 2969357d1e3bSDavid Gibson *mmio64 = SPAPR_PCI_BASE + (index + 1) * SPAPR_PCI_MEM64_WIN_SIZE; 29706737d9adSDavid Gibson } 29716737d9adSDavid Gibson 297229ee3247SAlexey Kardashevskiy static void spapr_machine_class_init(ObjectClass *oc, void *data) 297353018216SPaolo Bonzini { 297429ee3247SAlexey Kardashevskiy MachineClass *mc = MACHINE_CLASS(oc); 2975224245bfSDavid Gibson sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(oc); 297671461b0fSAlexey Kardashevskiy FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc); 297734316482SAlexey Kardashevskiy NMIClass *nc = NMI_CLASS(oc); 2978c20d332aSBharata B Rao HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc); 29791d1be34dSDavid Gibson PPCVirtualHypervisorClass *vhc = PPC_VIRTUAL_HYPERVISOR_CLASS(oc); 298029ee3247SAlexey Kardashevskiy 29810eb9054cSDavid Gibson mc->desc = "pSeries Logical Partition (PAPR compliant)"; 2982fc9f38c3SDavid Gibson 2983fc9f38c3SDavid Gibson /* 2984fc9f38c3SDavid Gibson * We set up the default / latest behaviour here. The class_init 2985fc9f38c3SDavid Gibson * functions for the specific versioned machine types can override 2986fc9f38c3SDavid Gibson * these details for backwards compatibility 2987fc9f38c3SDavid Gibson */ 2988958db90cSMarcel Apfelbaum mc->init = ppc_spapr_init; 2989958db90cSMarcel Apfelbaum mc->reset = ppc_spapr_reset; 2990958db90cSMarcel Apfelbaum mc->block_default_type = IF_SCSI; 29916244bb7eSGreg Kurz mc->max_cpus = 1024; 2992958db90cSMarcel Apfelbaum mc->no_parallel = 1; 29935b2128d2SAlexander Graf mc->default_boot_order = ""; 2994a34944feSNikunj A Dadhania mc->default_ram_size = 512 * M_BYTE; 2995958db90cSMarcel Apfelbaum mc->kvm_type = spapr_kvm_type; 29969e3f9733SAlexander Graf mc->has_dynamic_sysbus = true; 2997e4024630SLaurent Vivier mc->pci_allow_0_address = true; 29987ebaf795SBharata B Rao mc->get_hotplug_handler = spapr_get_hotplug_handler; 299994a94e4cSBharata B Rao hc->pre_plug = spapr_machine_device_pre_plug; 3000c20d332aSBharata B Rao hc->plug = spapr_machine_device_plug; 3001c20d332aSBharata B Rao hc->unplug = spapr_machine_device_unplug; 300220bb648dSDavid Gibson mc->cpu_index_to_socket_id = spapr_cpu_index_to_socket_id; 3003535455fdSIgor Mammedov mc->possible_cpu_arch_ids = spapr_possible_cpu_arch_ids; 3004cf632463SBharata B Rao hc->unplug_request = spapr_machine_device_unplug_request; 300500b4fbe2SMarcel Apfelbaum 3006fc9f38c3SDavid Gibson smc->dr_lmb_enabled = true; 30073daa4a9fSThomas Huth smc->tcg_default_cpu = "POWER8"; 3008c5514d0eSIgor Mammedov mc->has_hotpluggable_cpus = true; 300971461b0fSAlexey Kardashevskiy fwc->get_dev_path = spapr_get_fw_dev_path; 301034316482SAlexey Kardashevskiy nc->nmi_monitor_handler = spapr_nmi; 30116737d9adSDavid Gibson smc->phb_placement = spapr_phb_placement; 30121d1be34dSDavid Gibson vhc->hypercall = emulate_spapr_hypercall; 3013e57ca75cSDavid Gibson vhc->hpt_mask = spapr_hpt_mask; 3014e57ca75cSDavid Gibson vhc->map_hptes = spapr_map_hptes; 3015e57ca75cSDavid Gibson vhc->unmap_hptes = spapr_unmap_hptes; 3016e57ca75cSDavid Gibson vhc->store_hpte = spapr_store_hpte; 301753018216SPaolo Bonzini } 301853018216SPaolo Bonzini 301929ee3247SAlexey Kardashevskiy static const TypeInfo spapr_machine_info = { 302029ee3247SAlexey Kardashevskiy .name = TYPE_SPAPR_MACHINE, 302129ee3247SAlexey Kardashevskiy .parent = TYPE_MACHINE, 30224aee7362SDavid Gibson .abstract = true, 30236ca1502eSAlexey Kardashevskiy .instance_size = sizeof(sPAPRMachineState), 302423825581SEduardo Habkost .instance_init = spapr_machine_initfn, 302587bbdd9cSDavid Gibson .instance_finalize = spapr_machine_finalizefn, 3026183930c0SDavid Gibson .class_size = sizeof(sPAPRMachineClass), 302729ee3247SAlexey Kardashevskiy .class_init = spapr_machine_class_init, 302871461b0fSAlexey Kardashevskiy .interfaces = (InterfaceInfo[]) { 302971461b0fSAlexey Kardashevskiy { TYPE_FW_PATH_PROVIDER }, 303034316482SAlexey Kardashevskiy { TYPE_NMI }, 3031c20d332aSBharata B Rao { TYPE_HOTPLUG_HANDLER }, 30321d1be34dSDavid Gibson { TYPE_PPC_VIRTUAL_HYPERVISOR }, 303371461b0fSAlexey Kardashevskiy { } 303471461b0fSAlexey Kardashevskiy }, 303529ee3247SAlexey Kardashevskiy }; 303629ee3247SAlexey Kardashevskiy 3037fccbc785SDavid Gibson #define DEFINE_SPAPR_MACHINE(suffix, verstr, latest) \ 30385013c547SDavid Gibson static void spapr_machine_##suffix##_class_init(ObjectClass *oc, \ 30395013c547SDavid Gibson void *data) \ 30405013c547SDavid Gibson { \ 30415013c547SDavid Gibson MachineClass *mc = MACHINE_CLASS(oc); \ 30425013c547SDavid Gibson spapr_machine_##suffix##_class_options(mc); \ 3043fccbc785SDavid Gibson if (latest) { \ 3044fccbc785SDavid Gibson mc->alias = "pseries"; \ 3045fccbc785SDavid Gibson mc->is_default = 1; \ 3046fccbc785SDavid Gibson } \ 30475013c547SDavid Gibson } \ 30485013c547SDavid Gibson static void spapr_machine_##suffix##_instance_init(Object *obj) \ 30495013c547SDavid Gibson { \ 30505013c547SDavid Gibson MachineState *machine = MACHINE(obj); \ 30515013c547SDavid Gibson spapr_machine_##suffix##_instance_options(machine); \ 30525013c547SDavid Gibson } \ 30535013c547SDavid Gibson static const TypeInfo spapr_machine_##suffix##_info = { \ 30545013c547SDavid Gibson .name = MACHINE_TYPE_NAME("pseries-" verstr), \ 30555013c547SDavid Gibson .parent = TYPE_SPAPR_MACHINE, \ 30565013c547SDavid Gibson .class_init = spapr_machine_##suffix##_class_init, \ 30575013c547SDavid Gibson .instance_init = spapr_machine_##suffix##_instance_init, \ 30585013c547SDavid Gibson }; \ 30595013c547SDavid Gibson static void spapr_machine_register_##suffix(void) \ 30605013c547SDavid Gibson { \ 30615013c547SDavid Gibson type_register(&spapr_machine_##suffix##_info); \ 30625013c547SDavid Gibson } \ 30630e6aac87SEduardo Habkost type_init(spapr_machine_register_##suffix) 30645013c547SDavid Gibson 30651c5f29bbSDavid Gibson /* 3066fa325e6cSDavid Gibson * pseries-2.9 3067db800b21SDavid Gibson */ 3068fa325e6cSDavid Gibson static void spapr_machine_2_9_instance_options(MachineState *machine) 3069db800b21SDavid Gibson { 3070db800b21SDavid Gibson } 3071db800b21SDavid Gibson 3072fa325e6cSDavid Gibson static void spapr_machine_2_9_class_options(MachineClass *mc) 3073db800b21SDavid Gibson { 3074db800b21SDavid Gibson /* Defaults for the latest behaviour inherited from the base class */ 3075db800b21SDavid Gibson } 3076db800b21SDavid Gibson 3077fa325e6cSDavid Gibson DEFINE_SPAPR_MACHINE(2_9, "2.9", true); 3078fa325e6cSDavid Gibson 3079fa325e6cSDavid Gibson /* 3080fa325e6cSDavid Gibson * pseries-2.8 3081fa325e6cSDavid Gibson */ 3082fa325e6cSDavid Gibson #define SPAPR_COMPAT_2_8 \ 3083fa325e6cSDavid Gibson HW_COMPAT_2_8 3084fa325e6cSDavid Gibson 3085fa325e6cSDavid Gibson static void spapr_machine_2_8_instance_options(MachineState *machine) 3086fa325e6cSDavid Gibson { 3087fa325e6cSDavid Gibson spapr_machine_2_9_instance_options(machine); 3088fa325e6cSDavid Gibson } 3089fa325e6cSDavid Gibson 3090fa325e6cSDavid Gibson static void spapr_machine_2_8_class_options(MachineClass *mc) 3091fa325e6cSDavid Gibson { 3092fa325e6cSDavid Gibson spapr_machine_2_9_class_options(mc); 3093fa325e6cSDavid Gibson SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_8); 3094fa325e6cSDavid Gibson } 3095fa325e6cSDavid Gibson 3096fa325e6cSDavid Gibson DEFINE_SPAPR_MACHINE(2_8, "2.8", false); 3097db800b21SDavid Gibson 3098db800b21SDavid Gibson /* 30991ea1eefcSBharata B Rao * pseries-2.7 31001ea1eefcSBharata B Rao */ 3101db800b21SDavid Gibson #define SPAPR_COMPAT_2_7 \ 3102db800b21SDavid Gibson HW_COMPAT_2_7 \ 3103357d1e3bSDavid Gibson { \ 3104357d1e3bSDavid Gibson .driver = TYPE_SPAPR_PCI_HOST_BRIDGE, \ 3105357d1e3bSDavid Gibson .property = "mem_win_size", \ 3106357d1e3bSDavid Gibson .value = stringify(SPAPR_PCI_2_7_MMIO_WIN_SIZE),\ 3107357d1e3bSDavid Gibson }, \ 3108357d1e3bSDavid Gibson { \ 3109357d1e3bSDavid Gibson .driver = TYPE_SPAPR_PCI_HOST_BRIDGE, \ 3110357d1e3bSDavid Gibson .property = "mem64_win_size", \ 3111357d1e3bSDavid Gibson .value = "0", \ 3112146c11f1SDavid Gibson }, \ 3113146c11f1SDavid Gibson { \ 3114146c11f1SDavid Gibson .driver = TYPE_POWERPC_CPU, \ 3115146c11f1SDavid Gibson .property = "pre-2.8-migration", \ 3116146c11f1SDavid Gibson .value = "on", \ 31175c4537bdSDavid Gibson }, \ 31185c4537bdSDavid Gibson { \ 31195c4537bdSDavid Gibson .driver = TYPE_SPAPR_PCI_HOST_BRIDGE, \ 31205c4537bdSDavid Gibson .property = "pre-2.8-migration", \ 31215c4537bdSDavid Gibson .value = "on", \ 3122357d1e3bSDavid Gibson }, 3123357d1e3bSDavid Gibson 3124357d1e3bSDavid Gibson static void phb_placement_2_7(sPAPRMachineState *spapr, uint32_t index, 3125357d1e3bSDavid Gibson uint64_t *buid, hwaddr *pio, 3126357d1e3bSDavid Gibson hwaddr *mmio32, hwaddr *mmio64, 3127357d1e3bSDavid Gibson unsigned n_dma, uint32_t *liobns, Error **errp) 3128357d1e3bSDavid Gibson { 3129357d1e3bSDavid Gibson /* Legacy PHB placement for pseries-2.7 and earlier machine types */ 3130357d1e3bSDavid Gibson const uint64_t base_buid = 0x800000020000000ULL; 3131357d1e3bSDavid Gibson const hwaddr phb_spacing = 0x1000000000ULL; /* 64 GiB */ 3132357d1e3bSDavid Gibson const hwaddr mmio_offset = 0xa0000000; /* 2 GiB + 512 MiB */ 3133357d1e3bSDavid Gibson const hwaddr pio_offset = 0x80000000; /* 2 GiB */ 3134357d1e3bSDavid Gibson const uint32_t max_index = 255; 3135357d1e3bSDavid Gibson const hwaddr phb0_alignment = 0x10000000000ULL; /* 1 TiB */ 3136357d1e3bSDavid Gibson 3137357d1e3bSDavid Gibson uint64_t ram_top = MACHINE(spapr)->ram_size; 3138357d1e3bSDavid Gibson hwaddr phb0_base, phb_base; 3139357d1e3bSDavid Gibson int i; 3140357d1e3bSDavid Gibson 3141357d1e3bSDavid Gibson /* Do we have hotpluggable memory? */ 3142357d1e3bSDavid Gibson if (MACHINE(spapr)->maxram_size > ram_top) { 3143357d1e3bSDavid Gibson /* Can't just use maxram_size, because there may be an 3144357d1e3bSDavid Gibson * alignment gap between normal and hotpluggable memory 3145357d1e3bSDavid Gibson * regions */ 3146357d1e3bSDavid Gibson ram_top = spapr->hotplug_memory.base + 3147357d1e3bSDavid Gibson memory_region_size(&spapr->hotplug_memory.mr); 3148357d1e3bSDavid Gibson } 3149357d1e3bSDavid Gibson 3150357d1e3bSDavid Gibson phb0_base = QEMU_ALIGN_UP(ram_top, phb0_alignment); 3151357d1e3bSDavid Gibson 3152357d1e3bSDavid Gibson if (index > max_index) { 3153357d1e3bSDavid Gibson error_setg(errp, "\"index\" for PAPR PHB is too large (max %u)", 3154357d1e3bSDavid Gibson max_index); 3155357d1e3bSDavid Gibson return; 3156357d1e3bSDavid Gibson } 3157357d1e3bSDavid Gibson 3158357d1e3bSDavid Gibson *buid = base_buid + index; 3159357d1e3bSDavid Gibson for (i = 0; i < n_dma; ++i) { 3160357d1e3bSDavid Gibson liobns[i] = SPAPR_PCI_LIOBN(index, i); 3161357d1e3bSDavid Gibson } 3162357d1e3bSDavid Gibson 3163357d1e3bSDavid Gibson phb_base = phb0_base + index * phb_spacing; 3164357d1e3bSDavid Gibson *pio = phb_base + pio_offset; 3165357d1e3bSDavid Gibson *mmio32 = phb_base + mmio_offset; 3166357d1e3bSDavid Gibson /* 3167357d1e3bSDavid Gibson * We don't set the 64-bit MMIO window, relying on the PHB's 3168357d1e3bSDavid Gibson * fallback behaviour of automatically splitting a large "32-bit" 3169357d1e3bSDavid Gibson * window into contiguous 32-bit and 64-bit windows 3170357d1e3bSDavid Gibson */ 3171357d1e3bSDavid Gibson } 3172db800b21SDavid Gibson 31731ea1eefcSBharata B Rao static void spapr_machine_2_7_instance_options(MachineState *machine) 31741ea1eefcSBharata B Rao { 3175f6229214SMichael Roth sPAPRMachineState *spapr = SPAPR_MACHINE(machine); 3176f6229214SMichael Roth 3177672de881SMichael Roth spapr_machine_2_8_instance_options(machine); 3178f6229214SMichael Roth spapr->use_hotplug_event_source = false; 31791ea1eefcSBharata B Rao } 31801ea1eefcSBharata B Rao 31811ea1eefcSBharata B Rao static void spapr_machine_2_7_class_options(MachineClass *mc) 31821ea1eefcSBharata B Rao { 31833daa4a9fSThomas Huth sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc); 31843daa4a9fSThomas Huth 3185db800b21SDavid Gibson spapr_machine_2_8_class_options(mc); 31863daa4a9fSThomas Huth smc->tcg_default_cpu = "POWER7"; 3187db800b21SDavid Gibson SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_7); 3188357d1e3bSDavid Gibson smc->phb_placement = phb_placement_2_7; 31891ea1eefcSBharata B Rao } 31901ea1eefcSBharata B Rao 3191db800b21SDavid Gibson DEFINE_SPAPR_MACHINE(2_7, "2.7", false); 31921ea1eefcSBharata B Rao 31931ea1eefcSBharata B Rao /* 31944b23699cSDavid Gibson * pseries-2.6 31954b23699cSDavid Gibson */ 31961ea1eefcSBharata B Rao #define SPAPR_COMPAT_2_6 \ 3197ae4de14cSAlexey Kardashevskiy HW_COMPAT_2_6 \ 3198ae4de14cSAlexey Kardashevskiy { \ 3199ae4de14cSAlexey Kardashevskiy .driver = TYPE_SPAPR_PCI_HOST_BRIDGE,\ 3200ae4de14cSAlexey Kardashevskiy .property = "ddw",\ 3201ae4de14cSAlexey Kardashevskiy .value = stringify(off),\ 3202ae4de14cSAlexey Kardashevskiy }, 32031ea1eefcSBharata B Rao 32044b23699cSDavid Gibson static void spapr_machine_2_6_instance_options(MachineState *machine) 32054b23699cSDavid Gibson { 3206672de881SMichael Roth spapr_machine_2_7_instance_options(machine); 32074b23699cSDavid Gibson } 32084b23699cSDavid Gibson 32094b23699cSDavid Gibson static void spapr_machine_2_6_class_options(MachineClass *mc) 32104b23699cSDavid Gibson { 32111ea1eefcSBharata B Rao spapr_machine_2_7_class_options(mc); 3212c5514d0eSIgor Mammedov mc->has_hotpluggable_cpus = false; 32131ea1eefcSBharata B Rao SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_6); 32144b23699cSDavid Gibson } 32154b23699cSDavid Gibson 32161ea1eefcSBharata B Rao DEFINE_SPAPR_MACHINE(2_6, "2.6", false); 32174b23699cSDavid Gibson 32184b23699cSDavid Gibson /* 32191c5f29bbSDavid Gibson * pseries-2.5 32201c5f29bbSDavid Gibson */ 32214b23699cSDavid Gibson #define SPAPR_COMPAT_2_5 \ 322257c522f4SThomas Huth HW_COMPAT_2_5 \ 322357c522f4SThomas Huth { \ 322457c522f4SThomas Huth .driver = "spapr-vlan", \ 322557c522f4SThomas Huth .property = "use-rx-buffer-pools", \ 322657c522f4SThomas Huth .value = "off", \ 322757c522f4SThomas Huth }, 32284b23699cSDavid Gibson 32295013c547SDavid Gibson static void spapr_machine_2_5_instance_options(MachineState *machine) 32301c5f29bbSDavid Gibson { 3231672de881SMichael Roth spapr_machine_2_6_instance_options(machine); 32325013c547SDavid Gibson } 32335013c547SDavid Gibson 32345013c547SDavid Gibson static void spapr_machine_2_5_class_options(MachineClass *mc) 32355013c547SDavid Gibson { 323657040d45SThomas Huth sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc); 323757040d45SThomas Huth 32384b23699cSDavid Gibson spapr_machine_2_6_class_options(mc); 323957040d45SThomas Huth smc->use_ohci_by_default = true; 32404b23699cSDavid Gibson SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_5); 32411c5f29bbSDavid Gibson } 32421c5f29bbSDavid Gibson 32434b23699cSDavid Gibson DEFINE_SPAPR_MACHINE(2_5, "2.5", false); 32441c5f29bbSDavid Gibson 32451c5f29bbSDavid Gibson /* 32461c5f29bbSDavid Gibson * pseries-2.4 32471c5f29bbSDavid Gibson */ 324880fd50f9SCornelia Huck #define SPAPR_COMPAT_2_4 \ 324980fd50f9SCornelia Huck HW_COMPAT_2_4 325080fd50f9SCornelia Huck 32515013c547SDavid Gibson static void spapr_machine_2_4_instance_options(MachineState *machine) 32521c5f29bbSDavid Gibson { 32535013c547SDavid Gibson spapr_machine_2_5_instance_options(machine); 32545013c547SDavid Gibson } 32551c5f29bbSDavid Gibson 32565013c547SDavid Gibson static void spapr_machine_2_4_class_options(MachineClass *mc) 32575013c547SDavid Gibson { 3258fc9f38c3SDavid Gibson sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc); 3259fc9f38c3SDavid Gibson 3260fc9f38c3SDavid Gibson spapr_machine_2_5_class_options(mc); 3261fc9f38c3SDavid Gibson smc->dr_lmb_enabled = false; 3262f949b4e5SDavid Gibson SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_4); 32631c5f29bbSDavid Gibson } 32641c5f29bbSDavid Gibson 3265fccbc785SDavid Gibson DEFINE_SPAPR_MACHINE(2_4, "2.4", false); 32661c5f29bbSDavid Gibson 32671c5f29bbSDavid Gibson /* 32681c5f29bbSDavid Gibson * pseries-2.3 32691c5f29bbSDavid Gibson */ 327038ff32c6SEduardo Habkost #define SPAPR_COMPAT_2_3 \ 32717619c7b0SMichael Roth HW_COMPAT_2_3 \ 32727619c7b0SMichael Roth {\ 32737619c7b0SMichael Roth .driver = "spapr-pci-host-bridge",\ 32747619c7b0SMichael Roth .property = "dynamic-reconfiguration",\ 32757619c7b0SMichael Roth .value = "off",\ 32767619c7b0SMichael Roth }, 327738ff32c6SEduardo Habkost 32785013c547SDavid Gibson static void spapr_machine_2_3_instance_options(MachineState *machine) 32791c5f29bbSDavid Gibson { 32805013c547SDavid Gibson spapr_machine_2_4_instance_options(machine); 32811c5f29bbSDavid Gibson savevm_skip_section_footers(); 32821c5f29bbSDavid Gibson global_state_set_optional(); 328309b5e30dSGreg Kurz savevm_skip_configuration(); 32841c5f29bbSDavid Gibson } 32851c5f29bbSDavid Gibson 32865013c547SDavid Gibson static void spapr_machine_2_3_class_options(MachineClass *mc) 32871c5f29bbSDavid Gibson { 3288fc9f38c3SDavid Gibson spapr_machine_2_4_class_options(mc); 3289f949b4e5SDavid Gibson SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_3); 32901c5f29bbSDavid Gibson } 3291fccbc785SDavid Gibson DEFINE_SPAPR_MACHINE(2_3, "2.3", false); 32921c5f29bbSDavid Gibson 32931c5f29bbSDavid Gibson /* 32941c5f29bbSDavid Gibson * pseries-2.2 32951c5f29bbSDavid Gibson */ 32961c5f29bbSDavid Gibson 3297b194df47SAlexey Kardashevskiy #define SPAPR_COMPAT_2_2 \ 32984dfd8eaaSEduardo Habkost HW_COMPAT_2_2 \ 3299b194df47SAlexey Kardashevskiy {\ 3300b194df47SAlexey Kardashevskiy .driver = TYPE_SPAPR_PCI_HOST_BRIDGE,\ 3301b194df47SAlexey Kardashevskiy .property = "mem_win_size",\ 3302b194df47SAlexey Kardashevskiy .value = "0x20000000",\ 3303dd754bafSEduardo Habkost }, 3304b194df47SAlexey Kardashevskiy 33055013c547SDavid Gibson static void spapr_machine_2_2_instance_options(MachineState *machine) 3306b0e966d0SJason Wang { 33075013c547SDavid Gibson spapr_machine_2_3_instance_options(machine); 3308cba0e779SGreg Kurz machine->suppress_vmdesc = true; 3309b0e966d0SJason Wang } 3310b0e966d0SJason Wang 33115013c547SDavid Gibson static void spapr_machine_2_2_class_options(MachineClass *mc) 3312b0e966d0SJason Wang { 3313fc9f38c3SDavid Gibson spapr_machine_2_3_class_options(mc); 3314f949b4e5SDavid Gibson SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_2); 33151c5f29bbSDavid Gibson } 3316fccbc785SDavid Gibson DEFINE_SPAPR_MACHINE(2_2, "2.2", false); 33171c5f29bbSDavid Gibson 33181c5f29bbSDavid Gibson /* 33191c5f29bbSDavid Gibson * pseries-2.1 33201c5f29bbSDavid Gibson */ 33211c5f29bbSDavid Gibson #define SPAPR_COMPAT_2_1 \ 33221c5f29bbSDavid Gibson HW_COMPAT_2_1 33231c5f29bbSDavid Gibson 33245013c547SDavid Gibson static void spapr_machine_2_1_instance_options(MachineState *machine) 33251c5f29bbSDavid Gibson { 33265013c547SDavid Gibson spapr_machine_2_2_instance_options(machine); 33271c5f29bbSDavid Gibson } 33281c5f29bbSDavid Gibson 33295013c547SDavid Gibson static void spapr_machine_2_1_class_options(MachineClass *mc) 3330b0e966d0SJason Wang { 3331fc9f38c3SDavid Gibson spapr_machine_2_2_class_options(mc); 3332f949b4e5SDavid Gibson SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_1); 33336026db45SAlexey Kardashevskiy } 3334fccbc785SDavid Gibson DEFINE_SPAPR_MACHINE(2_1, "2.1", false); 33356026db45SAlexey Kardashevskiy 333629ee3247SAlexey Kardashevskiy static void spapr_machine_register_types(void) 333729ee3247SAlexey Kardashevskiy { 333829ee3247SAlexey Kardashevskiy type_register_static(&spapr_machine_info); 333929ee3247SAlexey Kardashevskiy } 334029ee3247SAlexey Kardashevskiy 334129ee3247SAlexey Kardashevskiy type_init(spapr_machine_register_types) 3342