2 * QEMU PowerPC pSeries Logical Partition (aka sPAPR) hardware System Emulator
4 * Copyright (c) 2004-2007 Fabrice Bellard
5 * Copyright (c) 2007 Jocelyn Mayer
6 * Copyright (c) 2010 David Gibson, IBM Corporation.
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:
15 * The above copyright notice and this permission notice shall be included in
16 * all copies or substantial portions of the Software.
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
27 #include "sysemu/sysemu.h"
28 #include "sysemu/numa.h"
30 #include "hw/fw-path-provider.h"
33 #include "sysemu/block-backend.h"
34 #include "sysemu/cpus.h"
35 #include "sysemu/kvm.h"
37 #include "mmu-hash64.h"
40 #include "hw/boards.h"
41 #include "hw/ppc/ppc.h"
42 #include "hw/loader.h"
44 #include "hw/ppc/spapr.h"
45 #include "hw/ppc/spapr_vio.h"
46 #include "hw/pci-host/spapr.h"
47 #include "hw/ppc/xics.h"
48 #include "hw/pci/msi.h"
50 #include "hw/pci/pci.h"
51 #include "hw/scsi/scsi.h"
52 #include "hw/virtio/virtio-scsi.h"
54 #include "exec/address-spaces.h"
56 #include "qemu/config-file.h"
57 #include "qemu/error-report.h"
61 #include "hw/compat.h"
65 /* SLOF memory layout:
67 * SLOF raw image loaded at 0, copies its romfs right below the flat
68 * device-tree, then position SLOF itself 31M below that
70 * So we set FW_OVERHEAD to 40MB which should account for all of that
73 * We load our kernel at 4M, leaving space for SLOF initial image
75 #define FDT_MAX_SIZE 0x40000
76 #define RTAS_MAX_SIZE 0x10000
77 #define RTAS_MAX_ADDR 0x80000000 /* RTAS must stay below that */
78 #define FW_MAX_SIZE 0x400000
79 #define FW_FILE_NAME "slof.bin"
80 #define FW_OVERHEAD 0x2800000
81 #define KERNEL_LOAD_ADDR FW_MAX_SIZE
83 #define MIN_RMA_SLOF 128UL
85 #define TIMEBASE_FREQ 512000000ULL
89 #define PHANDLE_XICP 0x00001111
91 #define HTAB_SIZE(spapr) (1ULL << ((spapr)->htab_shift))
93 typedef struct sPAPRMachineState sPAPRMachineState;
95 #define TYPE_SPAPR_MACHINE "spapr-machine"
96 #define SPAPR_MACHINE(obj) \
97 OBJECT_CHECK(sPAPRMachineState, (obj), TYPE_SPAPR_MACHINE)
102 struct sPAPRMachineState {
104 MachineState parent_obj;
110 sPAPREnvironment *spapr;
112 static XICSState *try_create_xics(const char *type, int nr_servers,
117 dev = qdev_create(NULL, type);
118 qdev_prop_set_uint32(dev, "nr_servers", nr_servers);
119 qdev_prop_set_uint32(dev, "nr_irqs", nr_irqs);
120 if (qdev_init(dev) < 0) {
124 return XICS_COMMON(dev);
127 static XICSState *xics_system_init(int nr_servers, int nr_irqs)
129 XICSState *icp = NULL;
132 QemuOpts *machine_opts = qemu_get_machine_opts();
133 bool irqchip_allowed = qemu_opt_get_bool(machine_opts,
134 "kernel_irqchip", true);
135 bool irqchip_required = qemu_opt_get_bool(machine_opts,
136 "kernel_irqchip", false);
137 if (irqchip_allowed) {
138 icp = try_create_xics(TYPE_KVM_XICS, nr_servers, nr_irqs);
141 if (irqchip_required && !icp) {
142 perror("Failed to create in-kernel XICS\n");
148 icp = try_create_xics(TYPE_XICS, nr_servers, nr_irqs);
152 perror("Failed to create XICS\n");
159 static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu,
163 uint32_t servers_prop[smt_threads];
164 uint32_t gservers_prop[smt_threads * 2];
165 int index = ppc_get_vcpu_dt_id(cpu);
167 if (cpu->cpu_version) {
168 ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->cpu_version);
174 /* Build interrupt servers and gservers properties */
175 for (i = 0; i < smt_threads; i++) {
176 servers_prop[i] = cpu_to_be32(index + i);
177 /* Hack, direct the group queues back to cpu 0 */
178 gservers_prop[i*2] = cpu_to_be32(index + i);
179 gservers_prop[i*2 + 1] = 0;
181 ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s",
182 servers_prop, sizeof(servers_prop));
186 ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s",
187 gservers_prop, sizeof(gservers_prop));
192 static int spapr_fixup_cpu_dt(void *fdt, sPAPREnvironment *spapr)
194 int ret = 0, offset, cpus_offset;
197 int smt = kvmppc_smt_threads();
198 uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
201 PowerPCCPU *cpu = POWERPC_CPU(cs);
202 DeviceClass *dc = DEVICE_GET_CLASS(cs);
203 int index = ppc_get_vcpu_dt_id(cpu);
204 uint32_t associativity[] = {cpu_to_be32(0x5),
208 cpu_to_be32(cs->numa_node),
211 if ((index % smt) != 0) {
215 snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index);
217 cpus_offset = fdt_path_offset(fdt, "/cpus");
218 if (cpus_offset < 0) {
219 cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"),
221 if (cpus_offset < 0) {
225 offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model);
227 offset = fdt_add_subnode(fdt, cpus_offset, cpu_model);
233 if (nb_numa_nodes > 1) {
234 ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
235 sizeof(associativity));
241 ret = fdt_setprop(fdt, offset, "ibm,pft-size",
242 pft_size_prop, sizeof(pft_size_prop));
247 ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
248 ppc_get_compat_smt_threads(cpu));
257 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
260 size_t maxcells = maxsize / sizeof(uint32_t);
264 for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
265 struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
267 if (!sps->page_shift) {
270 for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
271 if (sps->enc[count].page_shift == 0) {
275 if ((p - prop) >= (maxcells - 3 - count * 2)) {
278 *(p++) = cpu_to_be32(sps->page_shift);
279 *(p++) = cpu_to_be32(sps->slb_enc);
280 *(p++) = cpu_to_be32(count);
281 for (j = 0; j < count; j++) {
282 *(p++) = cpu_to_be32(sps->enc[j].page_shift);
283 *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
287 return (p - prop) * sizeof(uint32_t);
290 static hwaddr spapr_node0_size(void)
294 for (i = 0; i < nb_numa_nodes; ++i) {
295 if (numa_info[i].node_mem) {
296 return MIN(pow2floor(numa_info[i].node_mem), ram_size);
307 fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
308 #exp, fdt_strerror(ret)); \
313 static void add_str(GString *s, const gchar *s1)
315 g_string_append_len(s, s1, strlen(s1) + 1);
318 static void *spapr_create_fdt_skel(hwaddr initrd_base,
322 const char *boot_device,
323 const char *kernel_cmdline,
328 uint32_t start_prop = cpu_to_be32(initrd_base);
329 uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
330 GString *hypertas = g_string_sized_new(256);
331 GString *qemu_hypertas = g_string_sized_new(256);
332 uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
333 uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(smp_cpus)};
334 int smt = kvmppc_smt_threads();
335 unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
336 QemuOpts *opts = qemu_opts_find(qemu_find_opts("smp-opts"), NULL);
337 unsigned sockets = opts ? qemu_opt_get_number(opts, "sockets", 0) : 0;
338 uint32_t cpus_per_socket = sockets ? (smp_cpus / sockets) : 1;
341 add_str(hypertas, "hcall-pft");
342 add_str(hypertas, "hcall-term");
343 add_str(hypertas, "hcall-dabr");
344 add_str(hypertas, "hcall-interrupt");
345 add_str(hypertas, "hcall-tce");
346 add_str(hypertas, "hcall-vio");
347 add_str(hypertas, "hcall-splpar");
348 add_str(hypertas, "hcall-bulk");
349 add_str(hypertas, "hcall-set-mode");
350 add_str(qemu_hypertas, "hcall-memop1");
352 fdt = g_malloc0(FDT_MAX_SIZE);
353 _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
356 _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
359 _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
361 _FDT((fdt_finish_reservemap(fdt)));
364 _FDT((fdt_begin_node(fdt, "")));
365 _FDT((fdt_property_string(fdt, "device_type", "chrp")));
366 _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
367 _FDT((fdt_property_string(fdt, "compatible", "qemu,pseries")));
370 * Add info to guest to indentify which host is it being run on
371 * and what is the uuid of the guest
373 if (kvmppc_get_host_model(&buf)) {
374 _FDT((fdt_property_string(fdt, "host-model", buf)));
377 if (kvmppc_get_host_serial(&buf)) {
378 _FDT((fdt_property_string(fdt, "host-serial", buf)));
382 buf = g_strdup_printf(UUID_FMT, qemu_uuid[0], qemu_uuid[1],
383 qemu_uuid[2], qemu_uuid[3], qemu_uuid[4],
384 qemu_uuid[5], qemu_uuid[6], qemu_uuid[7],
385 qemu_uuid[8], qemu_uuid[9], qemu_uuid[10],
386 qemu_uuid[11], qemu_uuid[12], qemu_uuid[13],
387 qemu_uuid[14], qemu_uuid[15]);
389 _FDT((fdt_property_string(fdt, "vm,uuid", buf)));
392 _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
393 _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
396 _FDT((fdt_begin_node(fdt, "chosen")));
398 /* Set Form1_affinity */
399 _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
401 _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
402 _FDT((fdt_property(fdt, "linux,initrd-start",
403 &start_prop, sizeof(start_prop))));
404 _FDT((fdt_property(fdt, "linux,initrd-end",
405 &end_prop, sizeof(end_prop))));
407 uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
408 cpu_to_be64(kernel_size) };
410 _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
412 _FDT((fdt_property(fdt, "qemu,boot-kernel-le", NULL, 0)));
416 _FDT((fdt_property_string(fdt, "qemu,boot-device", boot_device)));
419 _FDT((fdt_property_cell(fdt, "qemu,boot-menu", boot_menu)));
421 _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
422 _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
423 _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
425 _FDT((fdt_end_node(fdt)));
428 _FDT((fdt_begin_node(fdt, "cpus")));
430 _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
431 _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
434 PowerPCCPU *cpu = POWERPC_CPU(cs);
435 CPUPPCState *env = &cpu->env;
436 DeviceClass *dc = DEVICE_GET_CLASS(cs);
437 PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs);
438 int index = ppc_get_vcpu_dt_id(cpu);
440 uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
441 0xffffffff, 0xffffffff};
442 uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
443 uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
444 uint32_t page_sizes_prop[64];
445 size_t page_sizes_prop_size;
447 if ((index % smt) != 0) {
451 nodename = g_strdup_printf("%s@%x", dc->fw_name, index);
453 _FDT((fdt_begin_node(fdt, nodename)));
457 _FDT((fdt_property_cell(fdt, "reg", index)));
458 _FDT((fdt_property_string(fdt, "device_type", "cpu")));
460 _FDT((fdt_property_cell(fdt, "cpu-version", env->spr[SPR_PVR])));
461 _FDT((fdt_property_cell(fdt, "d-cache-block-size",
462 env->dcache_line_size)));
463 _FDT((fdt_property_cell(fdt, "d-cache-line-size",
464 env->dcache_line_size)));
465 _FDT((fdt_property_cell(fdt, "i-cache-block-size",
466 env->icache_line_size)));
467 _FDT((fdt_property_cell(fdt, "i-cache-line-size",
468 env->icache_line_size)));
470 if (pcc->l1_dcache_size) {
471 _FDT((fdt_property_cell(fdt, "d-cache-size", pcc->l1_dcache_size)));
473 fprintf(stderr, "Warning: Unknown L1 dcache size for cpu\n");
475 if (pcc->l1_icache_size) {
476 _FDT((fdt_property_cell(fdt, "i-cache-size", pcc->l1_icache_size)));
478 fprintf(stderr, "Warning: Unknown L1 icache size for cpu\n");
481 _FDT((fdt_property_cell(fdt, "timebase-frequency", tbfreq)));
482 _FDT((fdt_property_cell(fdt, "clock-frequency", cpufreq)));
483 _FDT((fdt_property_cell(fdt, "ibm,slb-size", env->slb_nr)));
484 _FDT((fdt_property_string(fdt, "status", "okay")));
485 _FDT((fdt_property(fdt, "64-bit", NULL, 0)));
487 if (env->spr_cb[SPR_PURR].oea_read) {
488 _FDT((fdt_property(fdt, "ibm,purr", NULL, 0)));
491 if (env->mmu_model & POWERPC_MMU_1TSEG) {
492 _FDT((fdt_property(fdt, "ibm,processor-segment-sizes",
493 segs, sizeof(segs))));
496 /* Advertise VMX/VSX (vector extensions) if available
497 * 0 / no property == no vector extensions
498 * 1 == VMX / Altivec available
499 * 2 == VSX available */
500 if (env->insns_flags & PPC_ALTIVEC) {
501 uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
503 _FDT((fdt_property_cell(fdt, "ibm,vmx", vmx)));
506 /* Advertise DFP (Decimal Floating Point) if available
507 * 0 / no property == no DFP
508 * 1 == DFP available */
509 if (env->insns_flags2 & PPC2_DFP) {
510 _FDT((fdt_property_cell(fdt, "ibm,dfp", 1)));
513 page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
514 sizeof(page_sizes_prop));
515 if (page_sizes_prop_size) {
516 _FDT((fdt_property(fdt, "ibm,segment-page-sizes",
517 page_sizes_prop, page_sizes_prop_size)));
520 _FDT((fdt_property_cell(fdt, "ibm,chip-id",
521 cs->cpu_index / cpus_per_socket)));
523 _FDT((fdt_end_node(fdt)));
526 _FDT((fdt_end_node(fdt)));
529 _FDT((fdt_begin_node(fdt, "rtas")));
531 if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
532 add_str(hypertas, "hcall-multi-tce");
534 _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas->str,
536 g_string_free(hypertas, TRUE);
537 _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas->str,
538 qemu_hypertas->len)));
539 g_string_free(qemu_hypertas, TRUE);
541 _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
542 refpoints, sizeof(refpoints))));
544 _FDT((fdt_property_cell(fdt, "rtas-error-log-max", RTAS_ERROR_LOG_MAX)));
547 * According to PAPR, rtas ibm,os-term does not guarantee a return
548 * back to the guest cpu.
550 * While an additional ibm,extended-os-term property indicates that
551 * rtas call return will always occur. Set this property.
553 _FDT((fdt_property(fdt, "ibm,extended-os-term", NULL, 0)));
555 _FDT((fdt_end_node(fdt)));
557 /* interrupt controller */
558 _FDT((fdt_begin_node(fdt, "interrupt-controller")));
560 _FDT((fdt_property_string(fdt, "device_type",
561 "PowerPC-External-Interrupt-Presentation")));
562 _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
563 _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
564 _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
565 interrupt_server_ranges_prop,
566 sizeof(interrupt_server_ranges_prop))));
567 _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
568 _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
569 _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
571 _FDT((fdt_end_node(fdt)));
574 _FDT((fdt_begin_node(fdt, "vdevice")));
576 _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
577 _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
578 _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
579 _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
580 _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
581 _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
583 _FDT((fdt_end_node(fdt)));
586 spapr_events_fdt_skel(fdt, epow_irq);
588 /* /hypervisor node */
590 uint8_t hypercall[16];
592 /* indicate KVM hypercall interface */
593 _FDT((fdt_begin_node(fdt, "hypervisor")));
594 _FDT((fdt_property_string(fdt, "compatible", "linux,kvm")));
595 if (kvmppc_has_cap_fixup_hcalls()) {
597 * Older KVM versions with older guest kernels were broken with the
598 * magic page, don't allow the guest to map it.
600 kvmppc_get_hypercall(first_cpu->env_ptr, hypercall,
602 _FDT((fdt_property(fdt, "hcall-instructions", hypercall,
603 sizeof(hypercall))));
605 _FDT((fdt_end_node(fdt)));
608 _FDT((fdt_end_node(fdt))); /* close root node */
609 _FDT((fdt_finish(fdt)));
614 int spapr_h_cas_compose_response(target_ulong addr, target_ulong size)
616 void *fdt, *fdt_skel;
617 sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 };
621 /* Create sceleton */
622 fdt_skel = g_malloc0(size);
623 _FDT((fdt_create(fdt_skel, size)));
624 _FDT((fdt_begin_node(fdt_skel, "")));
625 _FDT((fdt_end_node(fdt_skel)));
626 _FDT((fdt_finish(fdt_skel)));
627 fdt = g_malloc0(size);
628 _FDT((fdt_open_into(fdt_skel, fdt, size)));
631 /* Fix skeleton up */
632 _FDT((spapr_fixup_cpu_dt(fdt, spapr)));
634 /* Pack resulting tree */
635 _FDT((fdt_pack(fdt)));
637 if (fdt_totalsize(fdt) + sizeof(hdr) > size) {
638 trace_spapr_cas_failed(size);
642 cpu_physical_memory_write(addr, &hdr, sizeof(hdr));
643 cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt));
644 trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr));
650 static void spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start,
653 uint32_t associativity[] = {
654 cpu_to_be32(0x4), /* length */
655 cpu_to_be32(0x0), cpu_to_be32(0x0),
656 cpu_to_be32(0x0), cpu_to_be32(nodeid)
659 uint64_t mem_reg_property[2];
662 mem_reg_property[0] = cpu_to_be64(start);
663 mem_reg_property[1] = cpu_to_be64(size);
665 sprintf(mem_name, "memory@" TARGET_FMT_lx, start);
666 off = fdt_add_subnode(fdt, 0, mem_name);
668 _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
669 _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
670 sizeof(mem_reg_property))));
671 _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
672 sizeof(associativity))));
675 static int spapr_populate_memory(sPAPREnvironment *spapr, void *fdt)
677 hwaddr mem_start, node_size;
678 int i, nb_nodes = nb_numa_nodes;
679 NodeInfo *nodes = numa_info;
682 /* No NUMA nodes, assume there is just one node with whole RAM */
683 if (!nb_numa_nodes) {
685 ramnode.node_mem = ram_size;
689 for (i = 0, mem_start = 0; i < nb_nodes; ++i) {
690 if (!nodes[i].node_mem) {
693 if (mem_start >= ram_size) {
696 node_size = nodes[i].node_mem;
697 if (node_size > ram_size - mem_start) {
698 node_size = ram_size - mem_start;
702 /* ppc_spapr_init() checks for rma_size <= node0_size already */
703 spapr_populate_memory_node(fdt, i, 0, spapr->rma_size);
704 mem_start += spapr->rma_size;
705 node_size -= spapr->rma_size;
707 for ( ; node_size; ) {
708 hwaddr sizetmp = pow2floor(node_size);
710 /* mem_start != 0 here */
711 if (ctzl(mem_start) < ctzl(sizetmp)) {
712 sizetmp = 1ULL << ctzl(mem_start);
715 spapr_populate_memory_node(fdt, i, mem_start, sizetmp);
716 node_size -= sizetmp;
717 mem_start += sizetmp;
724 static void spapr_finalize_fdt(sPAPREnvironment *spapr,
735 fdt = g_malloc(FDT_MAX_SIZE);
737 /* open out the base tree into a temp buffer for the final tweaks */
738 _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
740 ret = spapr_populate_memory(spapr, fdt);
742 fprintf(stderr, "couldn't setup memory nodes in fdt\n");
746 ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
748 fprintf(stderr, "couldn't setup vio devices in fdt\n");
752 QLIST_FOREACH(phb, &spapr->phbs, list) {
753 ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
757 fprintf(stderr, "couldn't setup PCI devices in fdt\n");
762 ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
764 fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
767 /* Advertise NUMA via ibm,associativity */
768 ret = spapr_fixup_cpu_dt(fdt, spapr);
770 fprintf(stderr, "Couldn't finalize CPU device tree properties\n");
773 bootlist = get_boot_devices_list(&cb, true);
774 if (cb && bootlist) {
775 int offset = fdt_path_offset(fdt, "/chosen");
779 for (i = 0; i < cb; i++) {
780 if (bootlist[i] == '\n') {
785 ret = fdt_setprop_string(fdt, offset, "qemu,boot-list", bootlist);
788 if (!spapr->has_graphics) {
789 spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
792 _FDT((fdt_pack(fdt)));
794 if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
795 hw_error("FDT too big ! 0x%x bytes (max is 0x%x)\n",
796 fdt_totalsize(fdt), FDT_MAX_SIZE);
800 cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
806 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
808 return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
811 static void emulate_spapr_hypercall(PowerPCCPU *cpu)
813 CPUPPCState *env = &cpu->env;
816 hcall_dprintf("Hypercall made with MSR[PR]=1\n");
817 env->gpr[3] = H_PRIVILEGE;
819 env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]);
823 #define HPTE(_table, _i) (void *)(((uint64_t *)(_table)) + ((_i) * 2))
824 #define HPTE_VALID(_hpte) (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID)
825 #define HPTE_DIRTY(_hpte) (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY)
826 #define CLEAN_HPTE(_hpte) ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY))
827 #define DIRTY_HPTE(_hpte) ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY))
829 static void spapr_reset_htab(sPAPREnvironment *spapr)
834 /* allocate hash page table. For now we always make this 16mb,
835 * later we should probably make it scale to the size of guest
838 shift = kvmppc_reset_htab(spapr->htab_shift);
841 /* Kernel handles htab, we don't need to allocate one */
842 spapr->htab_shift = shift;
843 kvmppc_kern_htab = true;
845 /* Tell readers to update their file descriptor */
846 if (spapr->htab_fd >= 0) {
847 spapr->htab_fd_stale = true;
851 /* Allocate an htab if we don't yet have one */
852 spapr->htab = qemu_memalign(HTAB_SIZE(spapr), HTAB_SIZE(spapr));
856 memset(spapr->htab, 0, HTAB_SIZE(spapr));
858 for (index = 0; index < HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; index++) {
859 DIRTY_HPTE(HPTE(spapr->htab, index));
863 /* Update the RMA size if necessary */
864 if (spapr->vrma_adjust) {
865 spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
870 static int find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
872 bool matched = false;
874 if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) {
879 error_report("Device %s is not supported by this machine yet.",
880 qdev_fw_name(DEVICE(sbdev)));
888 * A guest reset will cause spapr->htab_fd to become stale if being used.
889 * Reopen the file descriptor to make sure the whole HTAB is properly read.
891 static int spapr_check_htab_fd(sPAPREnvironment *spapr)
895 if (spapr->htab_fd_stale) {
896 close(spapr->htab_fd);
897 spapr->htab_fd = kvmppc_get_htab_fd(false);
898 if (spapr->htab_fd < 0) {
899 error_report("Unable to open fd for reading hash table from KVM: "
900 "%s", strerror(errno));
903 spapr->htab_fd_stale = false;
909 static void ppc_spapr_reset(void)
911 PowerPCCPU *first_ppc_cpu;
914 /* Check for unknown sysbus devices */
915 foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL);
917 /* Reset the hash table & recalc the RMA */
918 spapr_reset_htab(spapr);
920 qemu_devices_reset();
923 * We place the device tree and RTAS just below either the top of the RMA,
924 * or just below 2GB, whichever is lowere, so that it can be
925 * processed with 32-bit real mode code if necessary
927 rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR);
928 spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
929 spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
932 spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
936 cpu_physical_memory_write(spapr->rtas_addr, spapr->rtas_blob,
939 /* Set up the entry state */
940 first_ppc_cpu = POWERPC_CPU(first_cpu);
941 first_ppc_cpu->env.gpr[3] = spapr->fdt_addr;
942 first_ppc_cpu->env.gpr[5] = 0;
943 first_cpu->halted = 0;
944 first_ppc_cpu->env.nip = spapr->entry_point;
948 static void spapr_cpu_reset(void *opaque)
950 PowerPCCPU *cpu = opaque;
951 CPUState *cs = CPU(cpu);
952 CPUPPCState *env = &cpu->env;
956 /* All CPUs start halted. CPU0 is unhalted from the machine level
957 * reset code and the rest are explicitly started up by the guest
958 * using an RTAS call */
961 env->spr[SPR_HIOR] = 0;
963 env->external_htab = (uint8_t *)spapr->htab;
964 if (kvm_enabled() && !env->external_htab) {
966 * HV KVM, set external_htab to 1 so our ppc_hash64_load_hpte*
967 * functions do the right thing.
969 env->external_htab = (void *)1;
973 * htab_mask is the mask used to normalize hash value to PTEG index.
974 * htab_shift is log2 of hash table size.
975 * We have 8 hpte per group, and each hpte is 16 bytes.
976 * ie have 128 bytes per hpte entry.
978 env->htab_mask = (1ULL << ((spapr)->htab_shift - 7)) - 1;
979 env->spr[SPR_SDR1] = (target_ulong)(uintptr_t)spapr->htab |
980 (spapr->htab_shift - 18);
983 static void spapr_create_nvram(sPAPREnvironment *spapr)
985 DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram");
986 DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0);
989 qdev_prop_set_drive_nofail(dev, "drive", blk_by_legacy_dinfo(dinfo));
992 qdev_init_nofail(dev);
994 spapr->nvram = (struct sPAPRNVRAM *)dev;
997 /* Returns whether we want to use VGA or not */
998 static int spapr_vga_init(PCIBus *pci_bus)
1000 switch (vga_interface_type) {
1006 return pci_vga_init(pci_bus) != NULL;
1008 fprintf(stderr, "This vga model is not supported,"
1009 "currently it only supports -vga std\n");
1014 static const VMStateDescription vmstate_spapr = {
1017 .minimum_version_id = 1,
1018 .fields = (VMStateField[]) {
1019 VMSTATE_UNUSED(4), /* used to be @next_irq */
1022 VMSTATE_UINT64(rtc_offset, sPAPREnvironment),
1023 VMSTATE_PPC_TIMEBASE_V(tb, sPAPREnvironment, 2),
1024 VMSTATE_END_OF_LIST()
1028 static int htab_save_setup(QEMUFile *f, void *opaque)
1030 sPAPREnvironment *spapr = opaque;
1032 /* "Iteration" header */
1033 qemu_put_be32(f, spapr->htab_shift);
1036 spapr->htab_save_index = 0;
1037 spapr->htab_first_pass = true;
1039 assert(kvm_enabled());
1041 spapr->htab_fd = kvmppc_get_htab_fd(false);
1042 spapr->htab_fd_stale = false;
1043 if (spapr->htab_fd < 0) {
1044 fprintf(stderr, "Unable to open fd for reading hash table from KVM: %s\n",
1054 static void htab_save_first_pass(QEMUFile *f, sPAPREnvironment *spapr,
1057 int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1058 int index = spapr->htab_save_index;
1059 int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1061 assert(spapr->htab_first_pass);
1066 /* Consume invalid HPTEs */
1067 while ((index < htabslots)
1068 && !HPTE_VALID(HPTE(spapr->htab, index))) {
1070 CLEAN_HPTE(HPTE(spapr->htab, index));
1073 /* Consume valid HPTEs */
1075 while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1076 && HPTE_VALID(HPTE(spapr->htab, index))) {
1078 CLEAN_HPTE(HPTE(spapr->htab, index));
1081 if (index > chunkstart) {
1082 int n_valid = index - chunkstart;
1084 qemu_put_be32(f, chunkstart);
1085 qemu_put_be16(f, n_valid);
1086 qemu_put_be16(f, 0);
1087 qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1088 HASH_PTE_SIZE_64 * n_valid);
1090 if ((qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1094 } while ((index < htabslots) && !qemu_file_rate_limit(f));
1096 if (index >= htabslots) {
1097 assert(index == htabslots);
1099 spapr->htab_first_pass = false;
1101 spapr->htab_save_index = index;
1104 static int htab_save_later_pass(QEMUFile *f, sPAPREnvironment *spapr,
1107 bool final = max_ns < 0;
1108 int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1109 int examined = 0, sent = 0;
1110 int index = spapr->htab_save_index;
1111 int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1113 assert(!spapr->htab_first_pass);
1116 int chunkstart, invalidstart;
1118 /* Consume non-dirty HPTEs */
1119 while ((index < htabslots)
1120 && !HPTE_DIRTY(HPTE(spapr->htab, index))) {
1126 /* Consume valid dirty HPTEs */
1127 while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1128 && HPTE_DIRTY(HPTE(spapr->htab, index))
1129 && HPTE_VALID(HPTE(spapr->htab, index))) {
1130 CLEAN_HPTE(HPTE(spapr->htab, index));
1135 invalidstart = index;
1136 /* Consume invalid dirty HPTEs */
1137 while ((index < htabslots) && (index - invalidstart < USHRT_MAX)
1138 && HPTE_DIRTY(HPTE(spapr->htab, index))
1139 && !HPTE_VALID(HPTE(spapr->htab, index))) {
1140 CLEAN_HPTE(HPTE(spapr->htab, index));
1145 if (index > chunkstart) {
1146 int n_valid = invalidstart - chunkstart;
1147 int n_invalid = index - invalidstart;
1149 qemu_put_be32(f, chunkstart);
1150 qemu_put_be16(f, n_valid);
1151 qemu_put_be16(f, n_invalid);
1152 qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1153 HASH_PTE_SIZE_64 * n_valid);
1154 sent += index - chunkstart;
1156 if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1161 if (examined >= htabslots) {
1165 if (index >= htabslots) {
1166 assert(index == htabslots);
1169 } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final));
1171 if (index >= htabslots) {
1172 assert(index == htabslots);
1176 spapr->htab_save_index = index;
1178 return (examined >= htabslots) && (sent == 0) ? 1 : 0;
1181 #define MAX_ITERATION_NS 5000000 /* 5 ms */
1182 #define MAX_KVM_BUF_SIZE 2048
1184 static int htab_save_iterate(QEMUFile *f, void *opaque)
1186 sPAPREnvironment *spapr = opaque;
1189 /* Iteration header */
1190 qemu_put_be32(f, 0);
1193 assert(kvm_enabled());
1195 rc = spapr_check_htab_fd(spapr);
1200 rc = kvmppc_save_htab(f, spapr->htab_fd,
1201 MAX_KVM_BUF_SIZE, MAX_ITERATION_NS);
1205 } else if (spapr->htab_first_pass) {
1206 htab_save_first_pass(f, spapr, MAX_ITERATION_NS);
1208 rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS);
1212 qemu_put_be32(f, 0);
1213 qemu_put_be16(f, 0);
1214 qemu_put_be16(f, 0);
1219 static int htab_save_complete(QEMUFile *f, void *opaque)
1221 sPAPREnvironment *spapr = opaque;
1223 /* Iteration header */
1224 qemu_put_be32(f, 0);
1229 assert(kvm_enabled());
1231 rc = spapr_check_htab_fd(spapr);
1236 rc = kvmppc_save_htab(f, spapr->htab_fd, MAX_KVM_BUF_SIZE, -1);
1240 close(spapr->htab_fd);
1241 spapr->htab_fd = -1;
1243 htab_save_later_pass(f, spapr, -1);
1247 qemu_put_be32(f, 0);
1248 qemu_put_be16(f, 0);
1249 qemu_put_be16(f, 0);
1254 static int htab_load(QEMUFile *f, void *opaque, int version_id)
1256 sPAPREnvironment *spapr = opaque;
1257 uint32_t section_hdr;
1260 if (version_id < 1 || version_id > 1) {
1261 fprintf(stderr, "htab_load() bad version\n");
1265 section_hdr = qemu_get_be32(f);
1268 /* First section, just the hash shift */
1269 if (spapr->htab_shift != section_hdr) {
1276 assert(kvm_enabled());
1278 fd = kvmppc_get_htab_fd(true);
1280 fprintf(stderr, "Unable to open fd to restore KVM hash table: %s\n",
1287 uint16_t n_valid, n_invalid;
1289 index = qemu_get_be32(f);
1290 n_valid = qemu_get_be16(f);
1291 n_invalid = qemu_get_be16(f);
1293 if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) {
1298 if ((index + n_valid + n_invalid) >
1299 (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) {
1300 /* Bad index in stream */
1301 fprintf(stderr, "htab_load() bad index %d (%hd+%hd entries) "
1302 "in htab stream (htab_shift=%d)\n", index, n_valid, n_invalid,
1309 qemu_get_buffer(f, HPTE(spapr->htab, index),
1310 HASH_PTE_SIZE_64 * n_valid);
1313 memset(HPTE(spapr->htab, index + n_valid), 0,
1314 HASH_PTE_SIZE_64 * n_invalid);
1321 rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid);
1336 static SaveVMHandlers savevm_htab_handlers = {
1337 .save_live_setup = htab_save_setup,
1338 .save_live_iterate = htab_save_iterate,
1339 .save_live_complete = htab_save_complete,
1340 .load_state = htab_load,
1343 /* pSeries LPAR / sPAPR hardware init */
1344 static void ppc_spapr_init(MachineState *machine)
1346 ram_addr_t ram_size = machine->ram_size;
1347 const char *cpu_model = machine->cpu_model;
1348 const char *kernel_filename = machine->kernel_filename;
1349 const char *kernel_cmdline = machine->kernel_cmdline;
1350 const char *initrd_filename = machine->initrd_filename;
1351 const char *boot_device = machine->boot_order;
1356 MemoryRegion *sysmem = get_system_memory();
1357 MemoryRegion *ram = g_new(MemoryRegion, 1);
1358 MemoryRegion *rma_region;
1360 hwaddr rma_alloc_size;
1361 hwaddr node0_size = spapr_node0_size();
1362 uint32_t initrd_base = 0;
1363 long kernel_size = 0, initrd_size = 0;
1364 long load_limit, fw_size;
1365 bool kernel_le = false;
1368 msi_supported = true;
1370 spapr = g_malloc0(sizeof(*spapr));
1371 QLIST_INIT(&spapr->phbs);
1373 cpu_ppc_hypercall = emulate_spapr_hypercall;
1375 /* Allocate RMA if necessary */
1376 rma_alloc_size = kvmppc_alloc_rma(&rma);
1378 if (rma_alloc_size == -1) {
1379 hw_error("qemu: Unable to create RMA\n");
1383 if (rma_alloc_size && (rma_alloc_size < node0_size)) {
1384 spapr->rma_size = rma_alloc_size;
1386 spapr->rma_size = node0_size;
1388 /* With KVM, we don't actually know whether KVM supports an
1389 * unbounded RMA (PR KVM) or is limited by the hash table size
1390 * (HV KVM using VRMA), so we always assume the latter
1392 * In that case, we also limit the initial allocations for RTAS
1393 * etc... to 256M since we have no way to know what the VRMA size
1394 * is going to be as it depends on the size of the hash table
1395 * isn't determined yet.
1397 if (kvm_enabled()) {
1398 spapr->vrma_adjust = 1;
1399 spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
1403 if (spapr->rma_size > node0_size) {
1404 fprintf(stderr, "Error: Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")\n",
1409 /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */
1410 load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD;
1412 /* We aim for a hash table of size 1/128 the size of RAM. The
1413 * normal rule of thumb is 1/64 the size of RAM, but that's much
1414 * more than needed for the Linux guests we support. */
1415 spapr->htab_shift = 18; /* Minimum architected size */
1416 while (spapr->htab_shift <= 46) {
1417 if ((1ULL << (spapr->htab_shift + 7)) >= ram_size) {
1420 spapr->htab_shift++;
1423 /* Set up Interrupt Controller before we create the VCPUs */
1424 spapr->icp = xics_system_init(smp_cpus * kvmppc_smt_threads() / smp_threads,
1428 if (cpu_model == NULL) {
1429 cpu_model = kvm_enabled() ? "host" : "POWER7";
1431 for (i = 0; i < smp_cpus; i++) {
1432 cpu = cpu_ppc_init(cpu_model);
1434 fprintf(stderr, "Unable to find PowerPC CPU definition\n");
1439 /* Set time-base frequency to 512 MHz */
1440 cpu_ppc_tb_init(env, TIMEBASE_FREQ);
1442 /* PAPR always has exception vectors in RAM not ROM. To ensure this,
1443 * MSR[IP] should never be set.
1445 env->msr_mask &= ~(1 << 6);
1447 /* Tell KVM that we're in PAPR mode */
1448 if (kvm_enabled()) {
1449 kvmppc_set_papr(cpu);
1452 if (cpu->max_compat) {
1453 if (ppc_set_compat(cpu, cpu->max_compat) < 0) {
1458 xics_cpu_setup(spapr->icp, cpu);
1460 qemu_register_reset(spapr_cpu_reset, cpu);
1464 spapr->ram_limit = ram_size;
1465 memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram",
1467 memory_region_add_subregion(sysmem, 0, ram);
1469 if (rma_alloc_size && rma) {
1470 rma_region = g_new(MemoryRegion, 1);
1471 memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma",
1472 rma_alloc_size, rma);
1473 vmstate_register_ram_global(rma_region);
1474 memory_region_add_subregion(sysmem, 0, rma_region);
1477 filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
1478 spapr->rtas_size = get_image_size(filename);
1479 spapr->rtas_blob = g_malloc(spapr->rtas_size);
1480 if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) {
1481 hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
1484 if (spapr->rtas_size > RTAS_MAX_SIZE) {
1485 hw_error("RTAS too big ! 0x%zx bytes (max is 0x%x)\n",
1486 (size_t)spapr->rtas_size, RTAS_MAX_SIZE);
1491 /* Set up EPOW events infrastructure */
1492 spapr_events_init(spapr);
1494 /* Set up VIO bus */
1495 spapr->vio_bus = spapr_vio_bus_init();
1497 for (i = 0; i < MAX_SERIAL_PORTS; i++) {
1498 if (serial_hds[i]) {
1499 spapr_vty_create(spapr->vio_bus, serial_hds[i]);
1503 /* We always have at least the nvram device on VIO */
1504 spapr_create_nvram(spapr);
1507 spapr_pci_rtas_init();
1509 phb = spapr_create_phb(spapr, 0);
1511 for (i = 0; i < nb_nics; i++) {
1512 NICInfo *nd = &nd_table[i];
1515 nd->model = g_strdup("ibmveth");
1518 if (strcmp(nd->model, "ibmveth") == 0) {
1519 spapr_vlan_create(spapr->vio_bus, nd);
1521 pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL);
1525 for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
1526 spapr_vscsi_create(spapr->vio_bus);
1530 if (spapr_vga_init(phb->bus)) {
1531 spapr->has_graphics = true;
1532 machine->usb |= defaults_enabled();
1536 pci_create_simple(phb->bus, -1, "pci-ohci");
1538 if (spapr->has_graphics) {
1539 USBBus *usb_bus = usb_bus_find(-1);
1541 usb_create_simple(usb_bus, "usb-kbd");
1542 usb_create_simple(usb_bus, "usb-mouse");
1546 if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
1547 fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
1548 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
1552 if (kernel_filename) {
1553 uint64_t lowaddr = 0;
1555 kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
1556 NULL, &lowaddr, NULL, 1, ELF_MACHINE, 0);
1557 if (kernel_size == ELF_LOAD_WRONG_ENDIAN) {
1558 kernel_size = load_elf(kernel_filename,
1559 translate_kernel_address, NULL,
1560 NULL, &lowaddr, NULL, 0, ELF_MACHINE, 0);
1561 kernel_le = kernel_size > 0;
1563 if (kernel_size < 0) {
1564 fprintf(stderr, "qemu: error loading %s: %s\n",
1565 kernel_filename, load_elf_strerror(kernel_size));
1570 if (initrd_filename) {
1571 /* Try to locate the initrd in the gap between the kernel
1572 * and the firmware. Add a bit of space just in case
1574 initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
1575 initrd_size = load_image_targphys(initrd_filename, initrd_base,
1576 load_limit - initrd_base);
1577 if (initrd_size < 0) {
1578 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
1588 if (bios_name == NULL) {
1589 bios_name = FW_FILE_NAME;
1591 filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1592 fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
1594 hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
1599 spapr->entry_point = 0x100;
1601 vmstate_register(NULL, 0, &vmstate_spapr, spapr);
1602 register_savevm_live(NULL, "spapr/htab", -1, 1,
1603 &savevm_htab_handlers, spapr);
1605 /* Prepare the device tree */
1606 spapr->fdt_skel = spapr_create_fdt_skel(initrd_base, initrd_size,
1607 kernel_size, kernel_le,
1608 boot_device, kernel_cmdline,
1610 assert(spapr->fdt_skel != NULL);
1613 static int spapr_kvm_type(const char *vm_type)
1619 if (!strcmp(vm_type, "HV")) {
1623 if (!strcmp(vm_type, "PR")) {
1627 error_report("Unknown kvm-type specified '%s'", vm_type);
1632 * Implementation of an interface to adjust firmware path
1633 * for the bootindex property handling.
1635 static char *spapr_get_fw_dev_path(FWPathProvider *p, BusState *bus,
1638 #define CAST(type, obj, name) \
1639 ((type *)object_dynamic_cast(OBJECT(obj), (name)))
1640 SCSIDevice *d = CAST(SCSIDevice, dev, TYPE_SCSI_DEVICE);
1641 sPAPRPHBState *phb = CAST(sPAPRPHBState, dev, TYPE_SPAPR_PCI_HOST_BRIDGE);
1644 void *spapr = CAST(void, bus->parent, "spapr-vscsi");
1645 VirtIOSCSI *virtio = CAST(VirtIOSCSI, bus->parent, TYPE_VIRTIO_SCSI);
1646 USBDevice *usb = CAST(USBDevice, bus->parent, TYPE_USB_DEVICE);
1650 * Replace "channel@0/disk@0,0" with "disk@8000000000000000":
1651 * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun
1652 * in the top 16 bits of the 64-bit LUN
1654 unsigned id = 0x8000 | (d->id << 8) | d->lun;
1655 return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
1656 (uint64_t)id << 48);
1657 } else if (virtio) {
1659 * We use SRP luns of the form 01000000 | (target << 8) | lun
1660 * in the top 32 bits of the 64-bit LUN
1661 * Note: the quote above is from SLOF and it is wrong,
1662 * the actual binding is:
1663 * swap 0100 or 10 << or 20 << ( target lun-id -- srplun )
1665 unsigned id = 0x1000000 | (d->id << 16) | d->lun;
1666 return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
1667 (uint64_t)id << 32);
1670 * We use SRP luns of the form 01000000 | (usb-port << 16) | lun
1671 * in the top 32 bits of the 64-bit LUN
1673 unsigned usb_port = atoi(usb->port->path);
1674 unsigned id = 0x1000000 | (usb_port << 16) | d->lun;
1675 return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
1676 (uint64_t)id << 32);
1681 /* Replace "pci" with "pci@800000020000000" */
1682 return g_strdup_printf("pci@%"PRIX64, phb->buid);
1688 static char *spapr_get_kvm_type(Object *obj, Error **errp)
1690 sPAPRMachineState *sm = SPAPR_MACHINE(obj);
1692 return g_strdup(sm->kvm_type);
1695 static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
1697 sPAPRMachineState *sm = SPAPR_MACHINE(obj);
1699 g_free(sm->kvm_type);
1700 sm->kvm_type = g_strdup(value);
1703 static void spapr_machine_initfn(Object *obj)
1705 object_property_add_str(obj, "kvm-type",
1706 spapr_get_kvm_type, spapr_set_kvm_type, NULL);
1707 object_property_set_description(obj, "kvm-type",
1708 "Specifies the KVM virtualization mode (HV, PR)",
1712 static void ppc_cpu_do_nmi_on_cpu(void *arg)
1716 cpu_synchronize_state(cs);
1717 ppc_cpu_do_system_reset(cs);
1720 static void spapr_nmi(NMIState *n, int cpu_index, Error **errp)
1725 async_run_on_cpu(cs, ppc_cpu_do_nmi_on_cpu, cs);
1729 static void spapr_machine_class_init(ObjectClass *oc, void *data)
1731 MachineClass *mc = MACHINE_CLASS(oc);
1732 FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc);
1733 NMIClass *nc = NMI_CLASS(oc);
1735 mc->init = ppc_spapr_init;
1736 mc->reset = ppc_spapr_reset;
1737 mc->block_default_type = IF_SCSI;
1738 mc->max_cpus = MAX_CPUS;
1739 mc->no_parallel = 1;
1740 mc->default_boot_order = NULL;
1741 mc->kvm_type = spapr_kvm_type;
1742 mc->has_dynamic_sysbus = true;
1744 fwc->get_dev_path = spapr_get_fw_dev_path;
1745 nc->nmi_monitor_handler = spapr_nmi;
1748 static const TypeInfo spapr_machine_info = {
1749 .name = TYPE_SPAPR_MACHINE,
1750 .parent = TYPE_MACHINE,
1752 .instance_size = sizeof(sPAPRMachineState),
1753 .instance_init = spapr_machine_initfn,
1754 .class_init = spapr_machine_class_init,
1755 .interfaces = (InterfaceInfo[]) {
1756 { TYPE_FW_PATH_PROVIDER },
1762 static void spapr_machine_2_1_class_init(ObjectClass *oc, void *data)
1764 MachineClass *mc = MACHINE_CLASS(oc);
1765 static GlobalProperty compat_props[] = {
1767 { /* end of list */ }
1770 mc->name = "pseries-2.1";
1771 mc->desc = "pSeries Logical Partition (PAPR compliant) v2.1";
1772 mc->compat_props = compat_props;
1775 static const TypeInfo spapr_machine_2_1_info = {
1776 .name = TYPE_SPAPR_MACHINE "2.1",
1777 .parent = TYPE_SPAPR_MACHINE,
1778 .class_init = spapr_machine_2_1_class_init,
1781 static void spapr_machine_2_2_class_init(ObjectClass *oc, void *data)
1783 MachineClass *mc = MACHINE_CLASS(oc);
1785 mc->name = "pseries-2.2";
1786 mc->desc = "pSeries Logical Partition (PAPR compliant) v2.2";
1787 mc->alias = "pseries";
1791 static const TypeInfo spapr_machine_2_2_info = {
1792 .name = TYPE_SPAPR_MACHINE "2.2",
1793 .parent = TYPE_SPAPR_MACHINE,
1794 .class_init = spapr_machine_2_2_class_init,
1797 static void spapr_machine_register_types(void)
1799 type_register_static(&spapr_machine_info);
1800 type_register_static(&spapr_machine_2_1_info);
1801 type_register_static(&spapr_machine_2_2_info);
1804 type_init(spapr_machine_register_types)