Annotation of qemu/hw/spapr.c, revision 1.1.1.2

1.1       root        1: /*
                      2:  * QEMU PowerPC pSeries Logical Partition (aka sPAPR) hardware System Emulator
                      3:  *
                      4:  * Copyright (c) 2004-2007 Fabrice Bellard
                      5:  * Copyright (c) 2007 Jocelyn Mayer
                      6:  * Copyright (c) 2010 David Gibson, IBM Corporation.
                      7:  *
                      8:  * Permission is hereby granted, free of charge, to any person obtaining a copy
                      9:  * of this software and associated documentation files (the "Software"), to deal
                     10:  * in the Software without restriction, including without limitation the rights
                     11:  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
                     12:  * copies of the Software, and to permit persons to whom the Software is
                     13:  * furnished to do so, subject to the following conditions:
                     14:  *
                     15:  * The above copyright notice and this permission notice shall be included in
                     16:  * all copies or substantial portions of the Software.
                     17:  *
                     18:  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
                     19:  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
                     20:  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
                     21:  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
                     22:  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
                     23:  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
                     24:  * THE SOFTWARE.
                     25:  *
                     26:  */
                     27: #include "sysemu.h"
                     28: #include "hw.h"
                     29: #include "elf.h"
                     30: #include "net.h"
                     31: #include "blockdev.h"
1.1.1.2 ! root       32: #include "cpus.h"
        !            33: #include "kvm.h"
        !            34: #include "kvm_ppc.h"
1.1       root       35: 
                     36: #include "hw/boards.h"
                     37: #include "hw/ppc.h"
                     38: #include "hw/loader.h"
                     39: 
                     40: #include "hw/spapr.h"
                     41: #include "hw/spapr_vio.h"
1.1.1.2 ! root       42: #include "hw/spapr_pci.h"
1.1       root       43: #include "hw/xics.h"
                     44: 
1.1.1.2 ! root       45: #include "kvm.h"
        !            46: #include "kvm_ppc.h"
        !            47: #include "pci.h"
        !            48: 
        !            49: #include "exec-memory.h"
        !            50: 
1.1       root       51: #include <libfdt.h>
                     52: 
                     53: #define KERNEL_LOAD_ADDR        0x00000000
                     54: #define INITRD_LOAD_ADDR        0x02800000
                     55: #define FDT_MAX_SIZE            0x10000
                     56: #define RTAS_MAX_SIZE           0x10000
                     57: #define FW_MAX_SIZE             0x400000
                     58: #define FW_FILE_NAME            "slof.bin"
                     59: 
1.1.1.2 ! root       60: #define MIN_RMA_SLOF           128UL
1.1       root       61: 
                     62: #define TIMEBASE_FREQ           512000000ULL
                     63: 
                     64: #define MAX_CPUS                256
                     65: #define XICS_IRQS              1024
                     66: 
1.1.1.2 ! root       67: #define SPAPR_PCI_BUID          0x800000020000001ULL
        !            68: #define SPAPR_PCI_MEM_WIN_ADDR  (0x10000000000ULL + 0xA0000000)
        !            69: #define SPAPR_PCI_MEM_WIN_SIZE  0x20000000
        !            70: #define SPAPR_PCI_IO_WIN_ADDR   (0x10000000000ULL + 0x80000000)
        !            71: 
        !            72: #define PHANDLE_XICP            0x00001111
        !            73: 
1.1       root       74: sPAPREnvironment *spapr;
                     75: 
1.1.1.2 ! root       76: qemu_irq spapr_allocate_irq(uint32_t hint, uint32_t *irq_num)
        !            77: {
        !            78:     uint32_t irq;
        !            79:     qemu_irq qirq;
        !            80: 
        !            81:     if (hint) {
        !            82:         irq = hint;
        !            83:         /* FIXME: we should probably check for collisions somehow */
        !            84:     } else {
        !            85:         irq = spapr->next_irq++;
        !            86:     }
        !            87: 
        !            88:     qirq = xics_find_qirq(spapr->icp, irq);
        !            89:     if (!qirq) {
        !            90:         return NULL;
        !            91:     }
        !            92: 
        !            93:     if (irq_num) {
        !            94:         *irq_num = irq;
        !            95:     }
        !            96: 
        !            97:     return qirq;
        !            98: }
        !            99: 
1.1       root      100: static void *spapr_create_fdt_skel(const char *cpu_model,
1.1.1.2 ! root      101:                                    target_phys_addr_t rma_size,
1.1       root      102:                                    target_phys_addr_t initrd_base,
                    103:                                    target_phys_addr_t initrd_size,
                    104:                                    const char *boot_device,
                    105:                                    const char *kernel_cmdline,
                    106:                                    long hash_shift)
                    107: {
                    108:     void *fdt;
                    109:     CPUState *env;
1.1.1.2 ! root      110:     uint64_t mem_reg_property_rma[] = { 0, cpu_to_be64(rma_size) };
        !           111:     uint64_t mem_reg_property_nonrma[] = { cpu_to_be64(rma_size),
        !           112:                                            cpu_to_be64(ram_size - rma_size) };
1.1       root      113:     uint32_t start_prop = cpu_to_be32(initrd_base);
                    114:     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
                    115:     uint32_t pft_size_prop[] = {0, cpu_to_be32(hash_shift)};
                    116:     char hypertas_prop[] = "hcall-pft\0hcall-term\0hcall-dabr\0hcall-interrupt"
1.1.1.2 ! root      117:         "\0hcall-tce\0hcall-vio\0hcall-splpar\0hcall-bulk";
1.1       root      118:     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(smp_cpus)};
                    119:     int i;
                    120:     char *modelname;
1.1.1.2 ! root      121:     int smt = kvmppc_smt_threads();
1.1       root      122: 
                    123: #define _FDT(exp) \
                    124:     do { \
                    125:         int ret = (exp);                                           \
                    126:         if (ret < 0) {                                             \
                    127:             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
                    128:                     #exp, fdt_strerror(ret));                      \
                    129:             exit(1);                                               \
                    130:         }                                                          \
                    131:     } while (0)
                    132: 
1.1.1.2 ! root      133:     fdt = g_malloc0(FDT_MAX_SIZE);
1.1       root      134:     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
                    135: 
                    136:     _FDT((fdt_finish_reservemap(fdt)));
                    137: 
                    138:     /* Root node */
                    139:     _FDT((fdt_begin_node(fdt, "")));
                    140:     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
                    141:     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
                    142: 
                    143:     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
                    144:     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
                    145: 
                    146:     /* /chosen */
                    147:     _FDT((fdt_begin_node(fdt, "chosen")));
                    148: 
                    149:     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
                    150:     _FDT((fdt_property(fdt, "linux,initrd-start",
                    151:                        &start_prop, sizeof(start_prop))));
                    152:     _FDT((fdt_property(fdt, "linux,initrd-end",
                    153:                        &end_prop, sizeof(end_prop))));
                    154:     _FDT((fdt_property_string(fdt, "qemu,boot-device", boot_device)));
                    155: 
1.1.1.2 ! root      156:     /*
        !           157:      * Because we don't always invoke any firmware, we can't rely on
        !           158:      * that to do BAR allocation.  Long term, we should probably do
        !           159:      * that ourselves, but for now, this setting (plus advertising the
        !           160:      * current BARs as 0) causes sufficiently recent kernels to to the
        !           161:      * BAR assignment themselves */
        !           162:     _FDT((fdt_property_cell(fdt, "linux,pci-probe-only", 0)));
        !           163: 
1.1       root      164:     _FDT((fdt_end_node(fdt)));
                    165: 
1.1.1.2 ! root      166:     /* memory node(s) */
1.1       root      167:     _FDT((fdt_begin_node(fdt, "memory@0")));
                    168: 
                    169:     _FDT((fdt_property_string(fdt, "device_type", "memory")));
1.1.1.2 ! root      170:     _FDT((fdt_property(fdt, "reg", mem_reg_property_rma,
        !           171:                        sizeof(mem_reg_property_rma))));
1.1       root      172:     _FDT((fdt_end_node(fdt)));
                    173: 
1.1.1.2 ! root      174:     if (ram_size > rma_size) {
        !           175:         char mem_name[32];
        !           176: 
        !           177:         sprintf(mem_name, "memory@%" PRIx64, (uint64_t)rma_size);
        !           178:         _FDT((fdt_begin_node(fdt, mem_name)));
        !           179:         _FDT((fdt_property_string(fdt, "device_type", "memory")));
        !           180:         _FDT((fdt_property(fdt, "reg", mem_reg_property_nonrma,
        !           181:                            sizeof(mem_reg_property_nonrma))));
        !           182:         _FDT((fdt_end_node(fdt)));
        !           183:     }
        !           184: 
1.1       root      185:     /* cpus */
                    186:     _FDT((fdt_begin_node(fdt, "cpus")));
                    187: 
                    188:     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
                    189:     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
                    190: 
1.1.1.2 ! root      191:     modelname = g_strdup(cpu_model);
1.1       root      192: 
                    193:     for (i = 0; i < strlen(modelname); i++) {
                    194:         modelname[i] = toupper(modelname[i]);
                    195:     }
                    196: 
                    197:     for (env = first_cpu; env != NULL; env = env->next_cpu) {
                    198:         int index = env->cpu_index;
1.1.1.2 ! root      199:         uint32_t servers_prop[smp_threads];
        !           200:         uint32_t gservers_prop[smp_threads * 2];
1.1       root      201:         char *nodename;
                    202:         uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
                    203:                            0xffffffff, 0xffffffff};
1.1.1.2 ! root      204:         uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
        !           205:         uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
        !           206: 
        !           207:         if ((index % smt) != 0) {
        !           208:             continue;
        !           209:         }
1.1       root      210: 
                    211:         if (asprintf(&nodename, "%s@%x", modelname, index) < 0) {
                    212:             fprintf(stderr, "Allocation failure\n");
                    213:             exit(1);
                    214:         }
                    215: 
                    216:         _FDT((fdt_begin_node(fdt, nodename)));
                    217: 
                    218:         free(nodename);
                    219: 
                    220:         _FDT((fdt_property_cell(fdt, "reg", index)));
                    221:         _FDT((fdt_property_string(fdt, "device_type", "cpu")));
                    222: 
                    223:         _FDT((fdt_property_cell(fdt, "cpu-version", env->spr[SPR_PVR])));
                    224:         _FDT((fdt_property_cell(fdt, "dcache-block-size",
                    225:                                 env->dcache_line_size)));
                    226:         _FDT((fdt_property_cell(fdt, "icache-block-size",
                    227:                                 env->icache_line_size)));
1.1.1.2 ! root      228:         _FDT((fdt_property_cell(fdt, "timebase-frequency", tbfreq)));
        !           229:         _FDT((fdt_property_cell(fdt, "clock-frequency", cpufreq)));
1.1       root      230:         _FDT((fdt_property_cell(fdt, "ibm,slb-size", env->slb_nr)));
                    231:         _FDT((fdt_property(fdt, "ibm,pft-size",
                    232:                            pft_size_prop, sizeof(pft_size_prop))));
                    233:         _FDT((fdt_property_string(fdt, "status", "okay")));
                    234:         _FDT((fdt_property(fdt, "64-bit", NULL, 0)));
1.1.1.2 ! root      235: 
        !           236:         /* Build interrupt servers and gservers properties */
        !           237:         for (i = 0; i < smp_threads; i++) {
        !           238:             servers_prop[i] = cpu_to_be32(index + i);
        !           239:             /* Hack, direct the group queues back to cpu 0 */
        !           240:             gservers_prop[i*2] = cpu_to_be32(index + i);
        !           241:             gservers_prop[i*2 + 1] = 0;
        !           242:         }
        !           243:         _FDT((fdt_property(fdt, "ibm,ppc-interrupt-server#s",
        !           244:                            servers_prop, sizeof(servers_prop))));
1.1       root      245:         _FDT((fdt_property(fdt, "ibm,ppc-interrupt-gserver#s",
1.1.1.2 ! root      246:                            gservers_prop, sizeof(gservers_prop))));
1.1       root      247: 
                    248:         if (env->mmu_model & POWERPC_MMU_1TSEG) {
                    249:             _FDT((fdt_property(fdt, "ibm,processor-segment-sizes",
                    250:                                segs, sizeof(segs))));
                    251:         }
                    252: 
1.1.1.2 ! root      253:         /* Advertise VMX/VSX (vector extensions) if available
        !           254:          *   0 / no property == no vector extensions
        !           255:          *   1               == VMX / Altivec available
        !           256:          *   2               == VSX available */
        !           257:         if (env->insns_flags & PPC_ALTIVEC) {
        !           258:             uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
        !           259: 
        !           260:             _FDT((fdt_property_cell(fdt, "ibm,vmx", vmx)));
        !           261:         }
        !           262: 
        !           263:         /* Advertise DFP (Decimal Floating Point) if available
        !           264:          *   0 / no property == no DFP
        !           265:          *   1               == DFP available */
        !           266:         if (env->insns_flags2 & PPC2_DFP) {
        !           267:             _FDT((fdt_property_cell(fdt, "ibm,dfp", 1)));
        !           268:         }
        !           269: 
1.1       root      270:         _FDT((fdt_end_node(fdt)));
                    271:     }
                    272: 
1.1.1.2 ! root      273:     g_free(modelname);
1.1       root      274: 
                    275:     _FDT((fdt_end_node(fdt)));
                    276: 
                    277:     /* RTAS */
                    278:     _FDT((fdt_begin_node(fdt, "rtas")));
                    279: 
                    280:     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas_prop,
                    281:                        sizeof(hypertas_prop))));
                    282: 
                    283:     _FDT((fdt_end_node(fdt)));
                    284: 
                    285:     /* interrupt controller */
1.1.1.2 ! root      286:     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
1.1       root      287: 
                    288:     _FDT((fdt_property_string(fdt, "device_type",
                    289:                               "PowerPC-External-Interrupt-Presentation")));
                    290:     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
                    291:     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
                    292:     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
                    293:                        interrupt_server_ranges_prop,
                    294:                        sizeof(interrupt_server_ranges_prop))));
1.1.1.2 ! root      295:     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
        !           296:     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
        !           297:     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
1.1       root      298: 
                    299:     _FDT((fdt_end_node(fdt)));
                    300: 
                    301:     /* vdevice */
                    302:     _FDT((fdt_begin_node(fdt, "vdevice")));
                    303: 
                    304:     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
                    305:     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
                    306:     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
                    307:     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
                    308:     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
                    309:     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
                    310: 
                    311:     _FDT((fdt_end_node(fdt)));
                    312: 
                    313:     _FDT((fdt_end_node(fdt))); /* close root node */
                    314:     _FDT((fdt_finish(fdt)));
                    315: 
                    316:     return fdt;
                    317: }
                    318: 
                    319: static void spapr_finalize_fdt(sPAPREnvironment *spapr,
                    320:                                target_phys_addr_t fdt_addr,
                    321:                                target_phys_addr_t rtas_addr,
                    322:                                target_phys_addr_t rtas_size)
                    323: {
                    324:     int ret;
                    325:     void *fdt;
1.1.1.2 ! root      326:     sPAPRPHBState *phb;
1.1       root      327: 
1.1.1.2 ! root      328:     fdt = g_malloc(FDT_MAX_SIZE);
1.1       root      329: 
                    330:     /* open out the base tree into a temp buffer for the final tweaks */
                    331:     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
                    332: 
                    333:     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
                    334:     if (ret < 0) {
                    335:         fprintf(stderr, "couldn't setup vio devices in fdt\n");
                    336:         exit(1);
                    337:     }
                    338: 
1.1.1.2 ! root      339:     QLIST_FOREACH(phb, &spapr->phbs, list) {
        !           340:         ret = spapr_populate_pci_devices(phb, PHANDLE_XICP, fdt);
        !           341:     }
        !           342: 
        !           343:     if (ret < 0) {
        !           344:         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
        !           345:         exit(1);
        !           346:     }
        !           347: 
1.1       root      348:     /* RTAS */
                    349:     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
                    350:     if (ret < 0) {
                    351:         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
                    352:     }
                    353: 
1.1.1.2 ! root      354:     spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
        !           355: 
1.1       root      356:     _FDT((fdt_pack(fdt)));
                    357: 
                    358:     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
                    359: 
1.1.1.2 ! root      360:     g_free(fdt);
1.1       root      361: }
                    362: 
                    363: static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
                    364: {
                    365:     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
                    366: }
                    367: 
                    368: static void emulate_spapr_hypercall(CPUState *env)
                    369: {
                    370:     env->gpr[3] = spapr_hypercall(env, env->gpr[3], &env->gpr[4]);
                    371: }
                    372: 
                    373: static void spapr_reset(void *opaque)
                    374: {
                    375:     sPAPREnvironment *spapr = (sPAPREnvironment *)opaque;
                    376: 
                    377:     fprintf(stderr, "sPAPR reset\n");
                    378: 
                    379:     /* flush out the hash table */
                    380:     memset(spapr->htab, 0, spapr->htab_size);
                    381: 
                    382:     /* Load the fdt */
                    383:     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
                    384:                        spapr->rtas_size);
                    385: 
                    386:     /* Set up the entry state */
                    387:     first_cpu->gpr[3] = spapr->fdt_addr;
                    388:     first_cpu->gpr[5] = 0;
                    389:     first_cpu->halted = 0;
                    390:     first_cpu->nip = spapr->entry_point;
                    391: 
                    392: }
                    393: 
                    394: /* pSeries LPAR / sPAPR hardware init */
                    395: static void ppc_spapr_init(ram_addr_t ram_size,
                    396:                            const char *boot_device,
                    397:                            const char *kernel_filename,
                    398:                            const char *kernel_cmdline,
                    399:                            const char *initrd_filename,
                    400:                            const char *cpu_model)
                    401: {
                    402:     CPUState *env;
                    403:     int i;
1.1.1.2 ! root      404:     MemoryRegion *sysmem = get_system_memory();
        !           405:     MemoryRegion *ram = g_new(MemoryRegion, 1);
        !           406:     target_phys_addr_t rma_alloc_size, rma_size;
1.1       root      407:     uint32_t initrd_base;
                    408:     long kernel_size, initrd_size, fw_size;
                    409:     long pteg_shift = 17;
                    410:     char *filename;
                    411: 
1.1.1.2 ! root      412:     spapr = g_malloc0(sizeof(*spapr));
        !           413:     QLIST_INIT(&spapr->phbs);
        !           414: 
1.1       root      415:     cpu_ppc_hypercall = emulate_spapr_hypercall;
                    416: 
1.1.1.2 ! root      417:     /* Allocate RMA if necessary */
        !           418:     rma_alloc_size = kvmppc_alloc_rma("ppc_spapr.rma", sysmem);
        !           419: 
        !           420:     if (rma_alloc_size == -1) {
        !           421:         hw_error("qemu: Unable to create RMA\n");
        !           422:         exit(1);
        !           423:     }
        !           424:     if (rma_alloc_size && (rma_alloc_size < ram_size)) {
        !           425:         rma_size = rma_alloc_size;
        !           426:     } else {
        !           427:         rma_size = ram_size;
        !           428:     }
        !           429: 
        !           430:     /* We place the device tree just below either the top of the RMA,
        !           431:      * or just below 2GB, whichever is lowere, so that it can be
        !           432:      * processed with 32-bit real mode code if necessary */
        !           433:     spapr->fdt_addr = MIN(rma_size, 0x80000000) - FDT_MAX_SIZE;
1.1       root      434:     spapr->rtas_addr = spapr->fdt_addr - RTAS_MAX_SIZE;
                    435: 
                    436:     /* init CPUs */
                    437:     if (cpu_model == NULL) {
1.1.1.2 ! root      438:         cpu_model = kvm_enabled() ? "host" : "POWER7";
1.1       root      439:     }
                    440:     for (i = 0; i < smp_cpus; i++) {
                    441:         env = cpu_init(cpu_model);
                    442: 
                    443:         if (!env) {
                    444:             fprintf(stderr, "Unable to find PowerPC CPU definition\n");
                    445:             exit(1);
                    446:         }
                    447:         /* Set time-base frequency to 512 MHz */
                    448:         cpu_ppc_tb_init(env, TIMEBASE_FREQ);
                    449:         qemu_register_reset((QEMUResetHandler *)&cpu_reset, env);
                    450: 
                    451:         env->hreset_vector = 0x60;
                    452:         env->hreset_excp_prefix = 0;
                    453:         env->gpr[3] = env->cpu_index;
                    454:     }
                    455: 
                    456:     /* allocate RAM */
1.1.1.2 ! root      457:     spapr->ram_limit = ram_size;
        !           458:     if (spapr->ram_limit > rma_alloc_size) {
        !           459:         ram_addr_t nonrma_base = rma_alloc_size;
        !           460:         ram_addr_t nonrma_size = spapr->ram_limit - rma_alloc_size;
        !           461: 
        !           462:         memory_region_init_ram(ram, NULL, "ppc_spapr.ram", nonrma_size);
        !           463:         memory_region_add_subregion(sysmem, nonrma_base, ram);
        !           464:     }
1.1       root      465: 
                    466:     /* allocate hash page table.  For now we always make this 16mb,
                    467:      * later we should probably make it scale to the size of guest
                    468:      * RAM */
                    469:     spapr->htab_size = 1ULL << (pteg_shift + 7);
1.1.1.2 ! root      470:     spapr->htab = qemu_memalign(spapr->htab_size, spapr->htab_size);
1.1       root      471: 
                    472:     for (env = first_cpu; env != NULL; env = env->next_cpu) {
                    473:         env->external_htab = spapr->htab;
                    474:         env->htab_base = -1;
                    475:         env->htab_mask = spapr->htab_size - 1;
1.1.1.2 ! root      476: 
        !           477:         /* Tell KVM that we're in PAPR mode */
        !           478:         env->spr[SPR_SDR1] = (unsigned long)spapr->htab |
        !           479:                              ((pteg_shift + 7) - 18);
        !           480:         env->spr[SPR_HIOR] = 0;
        !           481: 
        !           482:         if (kvm_enabled()) {
        !           483:             kvmppc_set_papr(env);
        !           484:         }
1.1       root      485:     }
                    486: 
                    487:     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
                    488:     spapr->rtas_size = load_image_targphys(filename, spapr->rtas_addr,
                    489:                                            ram_size - spapr->rtas_addr);
                    490:     if (spapr->rtas_size < 0) {
                    491:         hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
                    492:         exit(1);
                    493:     }
1.1.1.2 ! root      494:     g_free(filename);
1.1       root      495: 
                    496:     /* Set up Interrupt Controller */
                    497:     spapr->icp = xics_system_init(XICS_IRQS);
1.1.1.2 ! root      498:     spapr->next_irq = 16;
1.1       root      499: 
                    500:     /* Set up VIO bus */
                    501:     spapr->vio_bus = spapr_vio_bus_init();
                    502: 
1.1.1.2 ! root      503:     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
1.1       root      504:         if (serial_hds[i]) {
                    505:             spapr_vty_create(spapr->vio_bus, SPAPR_VTY_BASE_ADDRESS + i,
1.1.1.2 ! root      506:                              serial_hds[i]);
1.1       root      507:         }
                    508:     }
                    509: 
1.1.1.2 ! root      510:     /* Set up PCI */
        !           511:     spapr_create_phb(spapr, "pci", SPAPR_PCI_BUID,
        !           512:                      SPAPR_PCI_MEM_WIN_ADDR,
        !           513:                      SPAPR_PCI_MEM_WIN_SIZE,
        !           514:                      SPAPR_PCI_IO_WIN_ADDR);
        !           515: 
        !           516:     for (i = 0; i < nb_nics; i++) {
1.1       root      517:         NICInfo *nd = &nd_table[i];
                    518: 
                    519:         if (!nd->model) {
1.1.1.2 ! root      520:             nd->model = g_strdup("ibmveth");
1.1       root      521:         }
                    522: 
                    523:         if (strcmp(nd->model, "ibmveth") == 0) {
1.1.1.2 ! root      524:             spapr_vlan_create(spapr->vio_bus, 0x1000 + i, nd);
1.1       root      525:         } else {
1.1.1.2 ! root      526:             pci_nic_init_nofail(&nd_table[i], nd->model, NULL);
1.1       root      527:         }
                    528:     }
                    529: 
                    530:     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
1.1.1.2 ! root      531:         spapr_vscsi_create(spapr->vio_bus, 0x2000 + i);
1.1       root      532:     }
                    533: 
                    534:     if (kernel_filename) {
                    535:         uint64_t lowaddr = 0;
                    536: 
                    537:         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
                    538:                                NULL, &lowaddr, NULL, 1, ELF_MACHINE, 0);
                    539:         if (kernel_size < 0) {
                    540:             kernel_size = load_image_targphys(kernel_filename,
                    541:                                               KERNEL_LOAD_ADDR,
                    542:                                               ram_size - KERNEL_LOAD_ADDR);
                    543:         }
                    544:         if (kernel_size < 0) {
                    545:             fprintf(stderr, "qemu: could not load kernel '%s'\n",
                    546:                     kernel_filename);
                    547:             exit(1);
                    548:         }
                    549: 
                    550:         /* load initrd */
                    551:         if (initrd_filename) {
                    552:             initrd_base = INITRD_LOAD_ADDR;
                    553:             initrd_size = load_image_targphys(initrd_filename, initrd_base,
                    554:                                               ram_size - initrd_base);
                    555:             if (initrd_size < 0) {
                    556:                 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
                    557:                         initrd_filename);
                    558:                 exit(1);
                    559:             }
                    560:         } else {
                    561:             initrd_base = 0;
                    562:             initrd_size = 0;
                    563:         }
                    564: 
                    565:         spapr->entry_point = KERNEL_LOAD_ADDR;
                    566:     } else {
1.1.1.2 ! root      567:         if (rma_size < (MIN_RMA_SLOF << 20)) {
1.1       root      568:             fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
1.1.1.2 ! root      569:                     "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
1.1       root      570:             exit(1);
                    571:         }
1.1.1.2 ! root      572:         filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, FW_FILE_NAME);
1.1       root      573:         fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
                    574:         if (fw_size < 0) {
                    575:             hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
                    576:             exit(1);
                    577:         }
1.1.1.2 ! root      578:         g_free(filename);
1.1       root      579:         spapr->entry_point = 0x100;
                    580:         initrd_base = 0;
                    581:         initrd_size = 0;
                    582: 
                    583:         /* SLOF will startup the secondary CPUs using RTAS,
                    584:            rather than expecting a kexec() style entry */
                    585:         for (env = first_cpu; env != NULL; env = env->next_cpu) {
                    586:             env->halted = 1;
                    587:         }
                    588:     }
                    589: 
                    590:     /* Prepare the device tree */
1.1.1.2 ! root      591:     spapr->fdt_skel = spapr_create_fdt_skel(cpu_model, rma_size,
1.1       root      592:                                             initrd_base, initrd_size,
                    593:                                             boot_device, kernel_cmdline,
                    594:                                             pteg_shift + 7);
                    595:     assert(spapr->fdt_skel != NULL);
                    596: 
                    597:     qemu_register_reset(spapr_reset, spapr);
                    598: }
                    599: 
                    600: static QEMUMachine spapr_machine = {
                    601:     .name = "pseries",
                    602:     .desc = "pSeries Logical Partition (PAPR compliant)",
                    603:     .init = ppc_spapr_init,
                    604:     .max_cpus = MAX_CPUS,
                    605:     .no_vga = 1,
                    606:     .no_parallel = 1,
                    607:     .use_scsi = 1,
                    608: };
                    609: 
                    610: static void spapr_machine_init(void)
                    611: {
                    612:     qemu_register_machine(&spapr_machine);
                    613: }
                    614: 
                    615: machine_init(spapr_machine_init);

unix.superglobalmegacorp.com