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 static XICSState
*try_create_xics(const char *type
, int nr_servers
,
94 int nr_irqs
, Error
**errp
)
99 dev
= qdev_create(NULL
, type
);
100 qdev_prop_set_uint32(dev
, "nr_servers", nr_servers
);
101 qdev_prop_set_uint32(dev
, "nr_irqs", nr_irqs
);
102 object_property_set_bool(OBJECT(dev
), true, "realized", &err
);
104 error_propagate(errp
, err
);
105 object_unparent(OBJECT(dev
));
108 return XICS_COMMON(dev
);
111 static XICSState
*xics_system_init(MachineState
*machine
,
112 int nr_servers
, int nr_irqs
)
114 XICSState
*icp
= NULL
;
119 if (machine_kernel_irqchip_allowed(machine
)) {
120 icp
= try_create_xics(TYPE_KVM_XICS
, nr_servers
, nr_irqs
, &err
);
122 if (machine_kernel_irqchip_required(machine
) && !icp
) {
123 error_report("kernel_irqchip requested but unavailable: %s",
124 error_get_pretty(err
));
129 icp
= try_create_xics(TYPE_XICS
, nr_servers
, nr_irqs
, &error_abort
);
135 static int spapr_fixup_cpu_smt_dt(void *fdt
, int offset
, PowerPCCPU
*cpu
,
139 uint32_t servers_prop
[smt_threads
];
140 uint32_t gservers_prop
[smt_threads
* 2];
141 int index
= ppc_get_vcpu_dt_id(cpu
);
143 if (cpu
->cpu_version
) {
144 ret
= fdt_setprop_cell(fdt
, offset
, "cpu-version", cpu
->cpu_version
);
150 /* Build interrupt servers and gservers properties */
151 for (i
= 0; i
< smt_threads
; i
++) {
152 servers_prop
[i
] = cpu_to_be32(index
+ i
);
153 /* Hack, direct the group queues back to cpu 0 */
154 gservers_prop
[i
*2] = cpu_to_be32(index
+ i
);
155 gservers_prop
[i
*2 + 1] = 0;
157 ret
= fdt_setprop(fdt
, offset
, "ibm,ppc-interrupt-server#s",
158 servers_prop
, sizeof(servers_prop
));
162 ret
= fdt_setprop(fdt
, offset
, "ibm,ppc-interrupt-gserver#s",
163 gservers_prop
, sizeof(gservers_prop
));
168 static int spapr_fixup_cpu_numa_dt(void *fdt
, int offset
, CPUState
*cs
)
171 PowerPCCPU
*cpu
= POWERPC_CPU(cs
);
172 int index
= ppc_get_vcpu_dt_id(cpu
);
173 uint32_t associativity
[] = {cpu_to_be32(0x5),
177 cpu_to_be32(cs
->numa_node
),
180 /* Advertise NUMA via ibm,associativity */
181 if (nb_numa_nodes
> 1) {
182 ret
= fdt_setprop(fdt
, offset
, "ibm,associativity", associativity
,
183 sizeof(associativity
));
189 static int spapr_fixup_cpu_dt(void *fdt
, sPAPRMachineState
*spapr
)
191 int ret
= 0, offset
, cpus_offset
;
194 int smt
= kvmppc_smt_threads();
195 uint32_t pft_size_prop
[] = {0, cpu_to_be32(spapr
->htab_shift
)};
198 PowerPCCPU
*cpu
= POWERPC_CPU(cs
);
199 DeviceClass
*dc
= DEVICE_GET_CLASS(cs
);
200 int index
= ppc_get_vcpu_dt_id(cpu
);
202 if ((index
% smt
) != 0) {
206 snprintf(cpu_model
, 32, "%s@%x", dc
->fw_name
, index
);
208 cpus_offset
= fdt_path_offset(fdt
, "/cpus");
209 if (cpus_offset
< 0) {
210 cpus_offset
= fdt_add_subnode(fdt
, fdt_path_offset(fdt
, "/"),
212 if (cpus_offset
< 0) {
216 offset
= fdt_subnode_offset(fdt
, cpus_offset
, cpu_model
);
218 offset
= fdt_add_subnode(fdt
, cpus_offset
, cpu_model
);
224 ret
= fdt_setprop(fdt
, offset
, "ibm,pft-size",
225 pft_size_prop
, sizeof(pft_size_prop
));
230 ret
= spapr_fixup_cpu_numa_dt(fdt
, offset
, cs
);
235 ret
= spapr_fixup_cpu_smt_dt(fdt
, offset
, cpu
,
236 ppc_get_compat_smt_threads(cpu
));
245 static size_t create_page_sizes_prop(CPUPPCState
*env
, uint32_t *prop
,
248 size_t maxcells
= maxsize
/ sizeof(uint32_t);
252 for (i
= 0; i
< PPC_PAGE_SIZES_MAX_SZ
; i
++) {
253 struct ppc_one_seg_page_size
*sps
= &env
->sps
.sps
[i
];
255 if (!sps
->page_shift
) {
258 for (count
= 0; count
< PPC_PAGE_SIZES_MAX_SZ
; count
++) {
259 if (sps
->enc
[count
].page_shift
== 0) {
263 if ((p
- prop
) >= (maxcells
- 3 - count
* 2)) {
266 *(p
++) = cpu_to_be32(sps
->page_shift
);
267 *(p
++) = cpu_to_be32(sps
->slb_enc
);
268 *(p
++) = cpu_to_be32(count
);
269 for (j
= 0; j
< count
; j
++) {
270 *(p
++) = cpu_to_be32(sps
->enc
[j
].page_shift
);
271 *(p
++) = cpu_to_be32(sps
->enc
[j
].pte_enc
);
275 return (p
- prop
) * sizeof(uint32_t);
278 static hwaddr
spapr_node0_size(void)
280 MachineState
*machine
= MACHINE(qdev_get_machine());
284 for (i
= 0; i
< nb_numa_nodes
; ++i
) {
285 if (numa_info
[i
].node_mem
) {
286 return MIN(pow2floor(numa_info
[i
].node_mem
),
291 return machine
->ram_size
;
298 fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
299 #exp, fdt_strerror(ret)); \
304 static void add_str(GString
*s
, const gchar
*s1
)
306 g_string_append_len(s
, s1
, strlen(s1
) + 1);
309 static void *spapr_create_fdt_skel(hwaddr initrd_base
,
313 const char *kernel_cmdline
,
317 uint32_t start_prop
= cpu_to_be32(initrd_base
);
318 uint32_t end_prop
= cpu_to_be32(initrd_base
+ initrd_size
);
319 GString
*hypertas
= g_string_sized_new(256);
320 GString
*qemu_hypertas
= g_string_sized_new(256);
321 uint32_t refpoints
[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
322 uint32_t interrupt_server_ranges_prop
[] = {0, cpu_to_be32(max_cpus
)};
323 unsigned char vec5
[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
326 add_str(hypertas
, "hcall-pft");
327 add_str(hypertas
, "hcall-term");
328 add_str(hypertas
, "hcall-dabr");
329 add_str(hypertas
, "hcall-interrupt");
330 add_str(hypertas
, "hcall-tce");
331 add_str(hypertas
, "hcall-vio");
332 add_str(hypertas
, "hcall-splpar");
333 add_str(hypertas
, "hcall-bulk");
334 add_str(hypertas
, "hcall-set-mode");
335 add_str(qemu_hypertas
, "hcall-memop1");
337 fdt
= g_malloc0(FDT_MAX_SIZE
);
338 _FDT((fdt_create(fdt
, FDT_MAX_SIZE
)));
341 _FDT((fdt_add_reservemap_entry(fdt
, KERNEL_LOAD_ADDR
, kernel_size
)));
344 _FDT((fdt_add_reservemap_entry(fdt
, initrd_base
, initrd_size
)));
346 _FDT((fdt_finish_reservemap(fdt
)));
349 _FDT((fdt_begin_node(fdt
, "")));
350 _FDT((fdt_property_string(fdt
, "device_type", "chrp")));
351 _FDT((fdt_property_string(fdt
, "model", "IBM pSeries (emulated by qemu)")));
352 _FDT((fdt_property_string(fdt
, "compatible", "qemu,pseries")));
355 * Add info to guest to indentify which host is it being run on
356 * and what is the uuid of the guest
358 if (kvmppc_get_host_model(&buf
)) {
359 _FDT((fdt_property_string(fdt
, "host-model", buf
)));
362 if (kvmppc_get_host_serial(&buf
)) {
363 _FDT((fdt_property_string(fdt
, "host-serial", buf
)));
367 buf
= g_strdup_printf(UUID_FMT
, qemu_uuid
[0], qemu_uuid
[1],
368 qemu_uuid
[2], qemu_uuid
[3], qemu_uuid
[4],
369 qemu_uuid
[5], qemu_uuid
[6], qemu_uuid
[7],
370 qemu_uuid
[8], qemu_uuid
[9], qemu_uuid
[10],
371 qemu_uuid
[11], qemu_uuid
[12], qemu_uuid
[13],
372 qemu_uuid
[14], qemu_uuid
[15]);
374 _FDT((fdt_property_string(fdt
, "vm,uuid", buf
)));
377 _FDT((fdt_property_cell(fdt
, "#address-cells", 0x2)));
378 _FDT((fdt_property_cell(fdt
, "#size-cells", 0x2)));
381 _FDT((fdt_begin_node(fdt
, "chosen")));
383 /* Set Form1_affinity */
384 _FDT((fdt_property(fdt
, "ibm,architecture-vec-5", vec5
, sizeof(vec5
))));
386 _FDT((fdt_property_string(fdt
, "bootargs", kernel_cmdline
)));
387 _FDT((fdt_property(fdt
, "linux,initrd-start",
388 &start_prop
, sizeof(start_prop
))));
389 _FDT((fdt_property(fdt
, "linux,initrd-end",
390 &end_prop
, sizeof(end_prop
))));
392 uint64_t kprop
[2] = { cpu_to_be64(KERNEL_LOAD_ADDR
),
393 cpu_to_be64(kernel_size
) };
395 _FDT((fdt_property(fdt
, "qemu,boot-kernel", &kprop
, sizeof(kprop
))));
397 _FDT((fdt_property(fdt
, "qemu,boot-kernel-le", NULL
, 0)));
401 _FDT((fdt_property_cell(fdt
, "qemu,boot-menu", boot_menu
)));
403 _FDT((fdt_property_cell(fdt
, "qemu,graphic-width", graphic_width
)));
404 _FDT((fdt_property_cell(fdt
, "qemu,graphic-height", graphic_height
)));
405 _FDT((fdt_property_cell(fdt
, "qemu,graphic-depth", graphic_depth
)));
407 _FDT((fdt_end_node(fdt
)));
410 _FDT((fdt_begin_node(fdt
, "rtas")));
412 if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
413 add_str(hypertas
, "hcall-multi-tce");
415 _FDT((fdt_property(fdt
, "ibm,hypertas-functions", hypertas
->str
,
417 g_string_free(hypertas
, TRUE
);
418 _FDT((fdt_property(fdt
, "qemu,hypertas-functions", qemu_hypertas
->str
,
419 qemu_hypertas
->len
)));
420 g_string_free(qemu_hypertas
, TRUE
);
422 _FDT((fdt_property(fdt
, "ibm,associativity-reference-points",
423 refpoints
, sizeof(refpoints
))));
425 _FDT((fdt_property_cell(fdt
, "rtas-error-log-max", RTAS_ERROR_LOG_MAX
)));
426 _FDT((fdt_property_cell(fdt
, "rtas-event-scan-rate",
427 RTAS_EVENT_SCAN_RATE
)));
430 * According to PAPR, rtas ibm,os-term does not guarantee a return
431 * back to the guest cpu.
433 * While an additional ibm,extended-os-term property indicates that
434 * rtas call return will always occur. Set this property.
436 _FDT((fdt_property(fdt
, "ibm,extended-os-term", NULL
, 0)));
438 _FDT((fdt_end_node(fdt
)));
440 /* interrupt controller */
441 _FDT((fdt_begin_node(fdt
, "interrupt-controller")));
443 _FDT((fdt_property_string(fdt
, "device_type",
444 "PowerPC-External-Interrupt-Presentation")));
445 _FDT((fdt_property_string(fdt
, "compatible", "IBM,ppc-xicp")));
446 _FDT((fdt_property(fdt
, "interrupt-controller", NULL
, 0)));
447 _FDT((fdt_property(fdt
, "ibm,interrupt-server-ranges",
448 interrupt_server_ranges_prop
,
449 sizeof(interrupt_server_ranges_prop
))));
450 _FDT((fdt_property_cell(fdt
, "#interrupt-cells", 2)));
451 _FDT((fdt_property_cell(fdt
, "linux,phandle", PHANDLE_XICP
)));
452 _FDT((fdt_property_cell(fdt
, "phandle", PHANDLE_XICP
)));
454 _FDT((fdt_end_node(fdt
)));
457 _FDT((fdt_begin_node(fdt
, "vdevice")));
459 _FDT((fdt_property_string(fdt
, "device_type", "vdevice")));
460 _FDT((fdt_property_string(fdt
, "compatible", "IBM,vdevice")));
461 _FDT((fdt_property_cell(fdt
, "#address-cells", 0x1)));
462 _FDT((fdt_property_cell(fdt
, "#size-cells", 0x0)));
463 _FDT((fdt_property_cell(fdt
, "#interrupt-cells", 0x2)));
464 _FDT((fdt_property(fdt
, "interrupt-controller", NULL
, 0)));
466 _FDT((fdt_end_node(fdt
)));
469 spapr_events_fdt_skel(fdt
, epow_irq
);
471 /* /hypervisor node */
473 uint8_t hypercall
[16];
475 /* indicate KVM hypercall interface */
476 _FDT((fdt_begin_node(fdt
, "hypervisor")));
477 _FDT((fdt_property_string(fdt
, "compatible", "linux,kvm")));
478 if (kvmppc_has_cap_fixup_hcalls()) {
480 * Older KVM versions with older guest kernels were broken with the
481 * magic page, don't allow the guest to map it.
483 kvmppc_get_hypercall(first_cpu
->env_ptr
, hypercall
,
485 _FDT((fdt_property(fdt
, "hcall-instructions", hypercall
,
486 sizeof(hypercall
))));
488 _FDT((fdt_end_node(fdt
)));
491 _FDT((fdt_end_node(fdt
))); /* close root node */
492 _FDT((fdt_finish(fdt
)));
497 int spapr_h_cas_compose_response(sPAPRMachineState
*spapr
,
498 target_ulong addr
, target_ulong size
)
500 void *fdt
, *fdt_skel
;
501 sPAPRDeviceTreeUpdateHeader hdr
= { .version_id
= 1 };
505 /* Create sceleton */
506 fdt_skel
= g_malloc0(size
);
507 _FDT((fdt_create(fdt_skel
, size
)));
508 _FDT((fdt_begin_node(fdt_skel
, "")));
509 _FDT((fdt_end_node(fdt_skel
)));
510 _FDT((fdt_finish(fdt_skel
)));
511 fdt
= g_malloc0(size
);
512 _FDT((fdt_open_into(fdt_skel
, fdt
, size
)));
515 /* Fix skeleton up */
516 _FDT((spapr_fixup_cpu_dt(fdt
, spapr
)));
518 /* Pack resulting tree */
519 _FDT((fdt_pack(fdt
)));
521 if (fdt_totalsize(fdt
) + sizeof(hdr
) > size
) {
522 trace_spapr_cas_failed(size
);
526 cpu_physical_memory_write(addr
, &hdr
, sizeof(hdr
));
527 cpu_physical_memory_write(addr
+ sizeof(hdr
), fdt
, fdt_totalsize(fdt
));
528 trace_spapr_cas_continue(fdt_totalsize(fdt
) + sizeof(hdr
));
534 static void spapr_populate_memory_node(void *fdt
, int nodeid
, hwaddr start
,
537 uint32_t associativity
[] = {
538 cpu_to_be32(0x4), /* length */
539 cpu_to_be32(0x0), cpu_to_be32(0x0),
540 cpu_to_be32(0x0), cpu_to_be32(nodeid
)
543 uint64_t mem_reg_property
[2];
546 mem_reg_property
[0] = cpu_to_be64(start
);
547 mem_reg_property
[1] = cpu_to_be64(size
);
549 sprintf(mem_name
, "memory@" TARGET_FMT_lx
, start
);
550 off
= fdt_add_subnode(fdt
, 0, mem_name
);
552 _FDT((fdt_setprop_string(fdt
, off
, "device_type", "memory")));
553 _FDT((fdt_setprop(fdt
, off
, "reg", mem_reg_property
,
554 sizeof(mem_reg_property
))));
555 _FDT((fdt_setprop(fdt
, off
, "ibm,associativity", associativity
,
556 sizeof(associativity
))));
559 static int spapr_populate_memory(sPAPRMachineState
*spapr
, void *fdt
)
561 MachineState
*machine
= MACHINE(spapr
);
562 hwaddr mem_start
, node_size
;
563 int i
, nb_nodes
= nb_numa_nodes
;
564 NodeInfo
*nodes
= numa_info
;
567 /* No NUMA nodes, assume there is just one node with whole RAM */
568 if (!nb_numa_nodes
) {
570 ramnode
.node_mem
= machine
->ram_size
;
574 for (i
= 0, mem_start
= 0; i
< nb_nodes
; ++i
) {
575 if (!nodes
[i
].node_mem
) {
578 if (mem_start
>= machine
->ram_size
) {
581 node_size
= nodes
[i
].node_mem
;
582 if (node_size
> machine
->ram_size
- mem_start
) {
583 node_size
= machine
->ram_size
- mem_start
;
587 /* ppc_spapr_init() checks for rma_size <= node0_size already */
588 spapr_populate_memory_node(fdt
, i
, 0, spapr
->rma_size
);
589 mem_start
+= spapr
->rma_size
;
590 node_size
-= spapr
->rma_size
;
592 for ( ; node_size
; ) {
593 hwaddr sizetmp
= pow2floor(node_size
);
595 /* mem_start != 0 here */
596 if (ctzl(mem_start
) < ctzl(sizetmp
)) {
597 sizetmp
= 1ULL << ctzl(mem_start
);
600 spapr_populate_memory_node(fdt
, i
, mem_start
, sizetmp
);
601 node_size
-= sizetmp
;
602 mem_start
+= sizetmp
;
609 static void spapr_populate_cpu_dt(CPUState
*cs
, void *fdt
, int offset
,
610 sPAPRMachineState
*spapr
)
612 PowerPCCPU
*cpu
= POWERPC_CPU(cs
);
613 CPUPPCState
*env
= &cpu
->env
;
614 PowerPCCPUClass
*pcc
= POWERPC_CPU_GET_CLASS(cs
);
615 int index
= ppc_get_vcpu_dt_id(cpu
);
616 uint32_t segs
[] = {cpu_to_be32(28), cpu_to_be32(40),
617 0xffffffff, 0xffffffff};
618 uint32_t tbfreq
= kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ
;
619 uint32_t cpufreq
= kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
620 uint32_t page_sizes_prop
[64];
621 size_t page_sizes_prop_size
;
622 QemuOpts
*opts
= qemu_opts_find(qemu_find_opts("smp-opts"), NULL
);
623 unsigned sockets
= opts
? qemu_opt_get_number(opts
, "sockets", 0) : 0;
624 uint32_t cpus_per_socket
= sockets
? (smp_cpus
/ sockets
) : 1;
625 uint32_t pft_size_prop
[] = {0, cpu_to_be32(spapr
->htab_shift
)};
627 _FDT((fdt_setprop_cell(fdt
, offset
, "reg", index
)));
628 _FDT((fdt_setprop_string(fdt
, offset
, "device_type", "cpu")));
630 _FDT((fdt_setprop_cell(fdt
, offset
, "cpu-version", env
->spr
[SPR_PVR
])));
631 _FDT((fdt_setprop_cell(fdt
, offset
, "d-cache-block-size",
632 env
->dcache_line_size
)));
633 _FDT((fdt_setprop_cell(fdt
, offset
, "d-cache-line-size",
634 env
->dcache_line_size
)));
635 _FDT((fdt_setprop_cell(fdt
, offset
, "i-cache-block-size",
636 env
->icache_line_size
)));
637 _FDT((fdt_setprop_cell(fdt
, offset
, "i-cache-line-size",
638 env
->icache_line_size
)));
640 if (pcc
->l1_dcache_size
) {
641 _FDT((fdt_setprop_cell(fdt
, offset
, "d-cache-size",
642 pcc
->l1_dcache_size
)));
644 fprintf(stderr
, "Warning: Unknown L1 dcache size for cpu\n");
646 if (pcc
->l1_icache_size
) {
647 _FDT((fdt_setprop_cell(fdt
, offset
, "i-cache-size",
648 pcc
->l1_icache_size
)));
650 fprintf(stderr
, "Warning: Unknown L1 icache size for cpu\n");
653 _FDT((fdt_setprop_cell(fdt
, offset
, "timebase-frequency", tbfreq
)));
654 _FDT((fdt_setprop_cell(fdt
, offset
, "clock-frequency", cpufreq
)));
655 _FDT((fdt_setprop_cell(fdt
, offset
, "ibm,slb-size", env
->slb_nr
)));
656 _FDT((fdt_setprop_string(fdt
, offset
, "status", "okay")));
657 _FDT((fdt_setprop(fdt
, offset
, "64-bit", NULL
, 0)));
659 if (env
->spr_cb
[SPR_PURR
].oea_read
) {
660 _FDT((fdt_setprop(fdt
, offset
, "ibm,purr", NULL
, 0)));
663 if (env
->mmu_model
& POWERPC_MMU_1TSEG
) {
664 _FDT((fdt_setprop(fdt
, offset
, "ibm,processor-segment-sizes",
665 segs
, sizeof(segs
))));
668 /* Advertise VMX/VSX (vector extensions) if available
669 * 0 / no property == no vector extensions
670 * 1 == VMX / Altivec available
671 * 2 == VSX available */
672 if (env
->insns_flags
& PPC_ALTIVEC
) {
673 uint32_t vmx
= (env
->insns_flags2
& PPC2_VSX
) ? 2 : 1;
675 _FDT((fdt_setprop_cell(fdt
, offset
, "ibm,vmx", vmx
)));
678 /* Advertise DFP (Decimal Floating Point) if available
679 * 0 / no property == no DFP
680 * 1 == DFP available */
681 if (env
->insns_flags2
& PPC2_DFP
) {
682 _FDT((fdt_setprop_cell(fdt
, offset
, "ibm,dfp", 1)));
685 page_sizes_prop_size
= create_page_sizes_prop(env
, page_sizes_prop
,
686 sizeof(page_sizes_prop
));
687 if (page_sizes_prop_size
) {
688 _FDT((fdt_setprop(fdt
, offset
, "ibm,segment-page-sizes",
689 page_sizes_prop
, page_sizes_prop_size
)));
692 _FDT((fdt_setprop_cell(fdt
, offset
, "ibm,chip-id",
693 cs
->cpu_index
/ cpus_per_socket
)));
695 _FDT((fdt_setprop(fdt
, offset
, "ibm,pft-size",
696 pft_size_prop
, sizeof(pft_size_prop
))));
698 _FDT(spapr_fixup_cpu_numa_dt(fdt
, offset
, cs
));
700 _FDT(spapr_fixup_cpu_smt_dt(fdt
, offset
, cpu
,
701 ppc_get_compat_smt_threads(cpu
)));
704 static void spapr_populate_cpus_dt_node(void *fdt
, sPAPRMachineState
*spapr
)
709 int smt
= kvmppc_smt_threads();
711 cpus_offset
= fdt_add_subnode(fdt
, 0, "cpus");
713 _FDT((fdt_setprop_cell(fdt
, cpus_offset
, "#address-cells", 0x1)));
714 _FDT((fdt_setprop_cell(fdt
, cpus_offset
, "#size-cells", 0x0)));
717 * We walk the CPUs in reverse order to ensure that CPU DT nodes
718 * created by fdt_add_subnode() end up in the right order in FDT
719 * for the guest kernel the enumerate the CPUs correctly.
721 CPU_FOREACH_REVERSE(cs
) {
722 PowerPCCPU
*cpu
= POWERPC_CPU(cs
);
723 int index
= ppc_get_vcpu_dt_id(cpu
);
724 DeviceClass
*dc
= DEVICE_GET_CLASS(cs
);
727 if ((index
% smt
) != 0) {
731 nodename
= g_strdup_printf("%s@%x", dc
->fw_name
, index
);
732 offset
= fdt_add_subnode(fdt
, cpus_offset
, nodename
);
735 spapr_populate_cpu_dt(cs
, fdt
, offset
, spapr
);
740 static void spapr_finalize_fdt(sPAPRMachineState
*spapr
,
745 MachineState
*machine
= MACHINE(qdev_get_machine());
746 const char *boot_device
= machine
->boot_order
;
753 fdt
= g_malloc(FDT_MAX_SIZE
);
755 /* open out the base tree into a temp buffer for the final tweaks */
756 _FDT((fdt_open_into(spapr
->fdt_skel
, fdt
, FDT_MAX_SIZE
)));
758 ret
= spapr_populate_memory(spapr
, fdt
);
760 fprintf(stderr
, "couldn't setup memory nodes in fdt\n");
764 ret
= spapr_populate_vdevice(spapr
->vio_bus
, fdt
);
766 fprintf(stderr
, "couldn't setup vio devices in fdt\n");
770 QLIST_FOREACH(phb
, &spapr
->phbs
, list
) {
771 ret
= spapr_populate_pci_dt(phb
, PHANDLE_XICP
, fdt
);
775 fprintf(stderr
, "couldn't setup PCI devices in fdt\n");
780 ret
= spapr_rtas_device_tree_setup(fdt
, rtas_addr
, rtas_size
);
782 fprintf(stderr
, "Couldn't set up RTAS device tree properties\n");
786 spapr_populate_cpus_dt_node(fdt
, spapr
);
788 bootlist
= get_boot_devices_list(&cb
, true);
789 if (cb
&& bootlist
) {
790 int offset
= fdt_path_offset(fdt
, "/chosen");
794 for (i
= 0; i
< cb
; i
++) {
795 if (bootlist
[i
] == '\n') {
800 ret
= fdt_setprop_string(fdt
, offset
, "qemu,boot-list", bootlist
);
803 if (boot_device
&& strlen(boot_device
)) {
804 int offset
= fdt_path_offset(fdt
, "/chosen");
809 fdt_setprop_string(fdt
, offset
, "qemu,boot-device", boot_device
);
812 if (!spapr
->has_graphics
) {
813 spapr_populate_chosen_stdout(fdt
, spapr
->vio_bus
);
816 _FDT((fdt_pack(fdt
)));
818 if (fdt_totalsize(fdt
) > FDT_MAX_SIZE
) {
819 error_report("FDT too big ! 0x%x bytes (max is 0x%x)",
820 fdt_totalsize(fdt
), FDT_MAX_SIZE
);
824 cpu_physical_memory_write(fdt_addr
, fdt
, fdt_totalsize(fdt
));
830 static uint64_t translate_kernel_address(void *opaque
, uint64_t addr
)
832 return (addr
& 0x0fffffff) + KERNEL_LOAD_ADDR
;
835 static void emulate_spapr_hypercall(PowerPCCPU
*cpu
)
837 CPUPPCState
*env
= &cpu
->env
;
840 hcall_dprintf("Hypercall made with MSR[PR]=1\n");
841 env
->gpr
[3] = H_PRIVILEGE
;
843 env
->gpr
[3] = spapr_hypercall(cpu
, env
->gpr
[3], &env
->gpr
[4]);
847 #define HPTE(_table, _i) (void *)(((uint64_t *)(_table)) + ((_i) * 2))
848 #define HPTE_VALID(_hpte) (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID)
849 #define HPTE_DIRTY(_hpte) (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY)
850 #define CLEAN_HPTE(_hpte) ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY))
851 #define DIRTY_HPTE(_hpte) ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY))
853 static void spapr_reset_htab(sPAPRMachineState
*spapr
)
858 /* allocate hash page table. For now we always make this 16mb,
859 * later we should probably make it scale to the size of guest
862 shift
= kvmppc_reset_htab(spapr
->htab_shift
);
865 /* Kernel handles htab, we don't need to allocate one */
866 spapr
->htab_shift
= shift
;
867 kvmppc_kern_htab
= true;
869 /* Tell readers to update their file descriptor */
870 if (spapr
->htab_fd
>= 0) {
871 spapr
->htab_fd_stale
= true;
875 /* Allocate an htab if we don't yet have one */
876 spapr
->htab
= qemu_memalign(HTAB_SIZE(spapr
), HTAB_SIZE(spapr
));
880 memset(spapr
->htab
, 0, HTAB_SIZE(spapr
));
882 for (index
= 0; index
< HTAB_SIZE(spapr
) / HASH_PTE_SIZE_64
; index
++) {
883 DIRTY_HPTE(HPTE(spapr
->htab
, index
));
887 /* Update the RMA size if necessary */
888 if (spapr
->vrma_adjust
) {
889 spapr
->rma_size
= kvmppc_rma_size(spapr_node0_size(),
894 static int find_unknown_sysbus_device(SysBusDevice
*sbdev
, void *opaque
)
896 bool matched
= false;
898 if (object_dynamic_cast(OBJECT(sbdev
), TYPE_SPAPR_PCI_HOST_BRIDGE
)) {
903 error_report("Device %s is not supported by this machine yet.",
904 qdev_fw_name(DEVICE(sbdev
)));
912 * A guest reset will cause spapr->htab_fd to become stale if being used.
913 * Reopen the file descriptor to make sure the whole HTAB is properly read.
915 static int spapr_check_htab_fd(sPAPRMachineState
*spapr
)
919 if (spapr
->htab_fd_stale
) {
920 close(spapr
->htab_fd
);
921 spapr
->htab_fd
= kvmppc_get_htab_fd(false);
922 if (spapr
->htab_fd
< 0) {
923 error_report("Unable to open fd for reading hash table from KVM: "
924 "%s", strerror(errno
));
927 spapr
->htab_fd_stale
= false;
933 static void ppc_spapr_reset(void)
935 sPAPRMachineState
*spapr
= SPAPR_MACHINE(qdev_get_machine());
936 PowerPCCPU
*first_ppc_cpu
;
939 /* Check for unknown sysbus devices */
940 foreach_dynamic_sysbus_device(find_unknown_sysbus_device
, NULL
);
942 /* Reset the hash table & recalc the RMA */
943 spapr_reset_htab(spapr
);
945 qemu_devices_reset();
948 * We place the device tree and RTAS just below either the top of the RMA,
949 * or just below 2GB, whichever is lowere, so that it can be
950 * processed with 32-bit real mode code if necessary
952 rtas_limit
= MIN(spapr
->rma_size
, RTAS_MAX_ADDR
);
953 spapr
->rtas_addr
= rtas_limit
- RTAS_MAX_SIZE
;
954 spapr
->fdt_addr
= spapr
->rtas_addr
- FDT_MAX_SIZE
;
957 spapr_finalize_fdt(spapr
, spapr
->fdt_addr
, spapr
->rtas_addr
,
961 cpu_physical_memory_write(spapr
->rtas_addr
, spapr
->rtas_blob
,
964 /* Set up the entry state */
965 first_ppc_cpu
= POWERPC_CPU(first_cpu
);
966 first_ppc_cpu
->env
.gpr
[3] = spapr
->fdt_addr
;
967 first_ppc_cpu
->env
.gpr
[5] = 0;
968 first_cpu
->halted
= 0;
969 first_ppc_cpu
->env
.nip
= SPAPR_ENTRY_POINT
;
973 static void spapr_cpu_reset(void *opaque
)
975 sPAPRMachineState
*spapr
= SPAPR_MACHINE(qdev_get_machine());
976 PowerPCCPU
*cpu
= opaque
;
977 CPUState
*cs
= CPU(cpu
);
978 CPUPPCState
*env
= &cpu
->env
;
982 /* All CPUs start halted. CPU0 is unhalted from the machine level
983 * reset code and the rest are explicitly started up by the guest
984 * using an RTAS call */
987 env
->spr
[SPR_HIOR
] = 0;
989 env
->external_htab
= (uint8_t *)spapr
->htab
;
990 if (kvm_enabled() && !env
->external_htab
) {
992 * HV KVM, set external_htab to 1 so our ppc_hash64_load_hpte*
993 * functions do the right thing.
995 env
->external_htab
= (void *)1;
999 * htab_mask is the mask used to normalize hash value to PTEG index.
1000 * htab_shift is log2 of hash table size.
1001 * We have 8 hpte per group, and each hpte is 16 bytes.
1002 * ie have 128 bytes per hpte entry.
1004 env
->htab_mask
= (1ULL << (spapr
->htab_shift
- 7)) - 1;
1005 env
->spr
[SPR_SDR1
] = (target_ulong
)(uintptr_t)spapr
->htab
|
1006 (spapr
->htab_shift
- 18);
1009 static void spapr_create_nvram(sPAPRMachineState
*spapr
)
1011 DeviceState
*dev
= qdev_create(&spapr
->vio_bus
->bus
, "spapr-nvram");
1012 DriveInfo
*dinfo
= drive_get(IF_PFLASH
, 0, 0);
1015 qdev_prop_set_drive_nofail(dev
, "drive", blk_by_legacy_dinfo(dinfo
));
1018 qdev_init_nofail(dev
);
1020 spapr
->nvram
= (struct sPAPRNVRAM
*)dev
;
1023 static void spapr_rtc_create(sPAPRMachineState
*spapr
)
1025 DeviceState
*dev
= qdev_create(NULL
, TYPE_SPAPR_RTC
);
1027 qdev_init_nofail(dev
);
1030 object_property_add_alias(qdev_get_machine(), "rtc-time",
1031 OBJECT(spapr
->rtc
), "date", NULL
);
1034 /* Returns whether we want to use VGA or not */
1035 static int spapr_vga_init(PCIBus
*pci_bus
)
1037 switch (vga_interface_type
) {
1043 return pci_vga_init(pci_bus
) != NULL
;
1045 fprintf(stderr
, "This vga model is not supported,"
1046 "currently it only supports -vga std\n");
1051 static int spapr_post_load(void *opaque
, int version_id
)
1053 sPAPRMachineState
*spapr
= (sPAPRMachineState
*)opaque
;
1056 /* In earlier versions, there was no separate qdev for the PAPR
1057 * RTC, so the RTC offset was stored directly in sPAPREnvironment.
1058 * So when migrating from those versions, poke the incoming offset
1059 * value into the RTC device */
1060 if (version_id
< 3) {
1061 err
= spapr_rtc_import_offset(spapr
->rtc
, spapr
->rtc_offset
);
1067 static bool version_before_3(void *opaque
, int version_id
)
1069 return version_id
< 3;
1072 static const VMStateDescription vmstate_spapr
= {
1075 .minimum_version_id
= 1,
1076 .post_load
= spapr_post_load
,
1077 .fields
= (VMStateField
[]) {
1078 /* used to be @next_irq */
1079 VMSTATE_UNUSED_BUFFER(version_before_3
, 0, 4),
1082 VMSTATE_UINT64_TEST(rtc_offset
, sPAPRMachineState
, version_before_3
),
1084 VMSTATE_PPC_TIMEBASE_V(tb
, sPAPRMachineState
, 2),
1085 VMSTATE_END_OF_LIST()
1089 static int htab_save_setup(QEMUFile
*f
, void *opaque
)
1091 sPAPRMachineState
*spapr
= opaque
;
1093 /* "Iteration" header */
1094 qemu_put_be32(f
, spapr
->htab_shift
);
1097 spapr
->htab_save_index
= 0;
1098 spapr
->htab_first_pass
= true;
1100 assert(kvm_enabled());
1102 spapr
->htab_fd
= kvmppc_get_htab_fd(false);
1103 spapr
->htab_fd_stale
= false;
1104 if (spapr
->htab_fd
< 0) {
1105 fprintf(stderr
, "Unable to open fd for reading hash table from KVM: %s\n",
1115 static void htab_save_first_pass(QEMUFile
*f
, sPAPRMachineState
*spapr
,
1118 int htabslots
= HTAB_SIZE(spapr
) / HASH_PTE_SIZE_64
;
1119 int index
= spapr
->htab_save_index
;
1120 int64_t starttime
= qemu_clock_get_ns(QEMU_CLOCK_REALTIME
);
1122 assert(spapr
->htab_first_pass
);
1127 /* Consume invalid HPTEs */
1128 while ((index
< htabslots
)
1129 && !HPTE_VALID(HPTE(spapr
->htab
, index
))) {
1131 CLEAN_HPTE(HPTE(spapr
->htab
, index
));
1134 /* Consume valid HPTEs */
1136 while ((index
< htabslots
) && (index
- chunkstart
< USHRT_MAX
)
1137 && HPTE_VALID(HPTE(spapr
->htab
, index
))) {
1139 CLEAN_HPTE(HPTE(spapr
->htab
, index
));
1142 if (index
> chunkstart
) {
1143 int n_valid
= index
- chunkstart
;
1145 qemu_put_be32(f
, chunkstart
);
1146 qemu_put_be16(f
, n_valid
);
1147 qemu_put_be16(f
, 0);
1148 qemu_put_buffer(f
, HPTE(spapr
->htab
, chunkstart
),
1149 HASH_PTE_SIZE_64
* n_valid
);
1151 if ((qemu_clock_get_ns(QEMU_CLOCK_REALTIME
) - starttime
) > max_ns
) {
1155 } while ((index
< htabslots
) && !qemu_file_rate_limit(f
));
1157 if (index
>= htabslots
) {
1158 assert(index
== htabslots
);
1160 spapr
->htab_first_pass
= false;
1162 spapr
->htab_save_index
= index
;
1165 static int htab_save_later_pass(QEMUFile
*f
, sPAPRMachineState
*spapr
,
1168 bool final
= max_ns
< 0;
1169 int htabslots
= HTAB_SIZE(spapr
) / HASH_PTE_SIZE_64
;
1170 int examined
= 0, sent
= 0;
1171 int index
= spapr
->htab_save_index
;
1172 int64_t starttime
= qemu_clock_get_ns(QEMU_CLOCK_REALTIME
);
1174 assert(!spapr
->htab_first_pass
);
1177 int chunkstart
, invalidstart
;
1179 /* Consume non-dirty HPTEs */
1180 while ((index
< htabslots
)
1181 && !HPTE_DIRTY(HPTE(spapr
->htab
, index
))) {
1187 /* Consume valid dirty HPTEs */
1188 while ((index
< htabslots
) && (index
- chunkstart
< USHRT_MAX
)
1189 && HPTE_DIRTY(HPTE(spapr
->htab
, index
))
1190 && HPTE_VALID(HPTE(spapr
->htab
, index
))) {
1191 CLEAN_HPTE(HPTE(spapr
->htab
, index
));
1196 invalidstart
= index
;
1197 /* Consume invalid dirty HPTEs */
1198 while ((index
< htabslots
) && (index
- invalidstart
< USHRT_MAX
)
1199 && HPTE_DIRTY(HPTE(spapr
->htab
, index
))
1200 && !HPTE_VALID(HPTE(spapr
->htab
, index
))) {
1201 CLEAN_HPTE(HPTE(spapr
->htab
, index
));
1206 if (index
> chunkstart
) {
1207 int n_valid
= invalidstart
- chunkstart
;
1208 int n_invalid
= index
- invalidstart
;
1210 qemu_put_be32(f
, chunkstart
);
1211 qemu_put_be16(f
, n_valid
);
1212 qemu_put_be16(f
, n_invalid
);
1213 qemu_put_buffer(f
, HPTE(spapr
->htab
, chunkstart
),
1214 HASH_PTE_SIZE_64
* n_valid
);
1215 sent
+= index
- chunkstart
;
1217 if (!final
&& (qemu_clock_get_ns(QEMU_CLOCK_REALTIME
) - starttime
) > max_ns
) {
1222 if (examined
>= htabslots
) {
1226 if (index
>= htabslots
) {
1227 assert(index
== htabslots
);
1230 } while ((examined
< htabslots
) && (!qemu_file_rate_limit(f
) || final
));
1232 if (index
>= htabslots
) {
1233 assert(index
== htabslots
);
1237 spapr
->htab_save_index
= index
;
1239 return (examined
>= htabslots
) && (sent
== 0) ? 1 : 0;
1242 #define MAX_ITERATION_NS 5000000 /* 5 ms */
1243 #define MAX_KVM_BUF_SIZE 2048
1245 static int htab_save_iterate(QEMUFile
*f
, void *opaque
)
1247 sPAPRMachineState
*spapr
= opaque
;
1250 /* Iteration header */
1251 qemu_put_be32(f
, 0);
1254 assert(kvm_enabled());
1256 rc
= spapr_check_htab_fd(spapr
);
1261 rc
= kvmppc_save_htab(f
, spapr
->htab_fd
,
1262 MAX_KVM_BUF_SIZE
, MAX_ITERATION_NS
);
1266 } else if (spapr
->htab_first_pass
) {
1267 htab_save_first_pass(f
, spapr
, MAX_ITERATION_NS
);
1269 rc
= htab_save_later_pass(f
, spapr
, MAX_ITERATION_NS
);
1273 qemu_put_be32(f
, 0);
1274 qemu_put_be16(f
, 0);
1275 qemu_put_be16(f
, 0);
1280 static int htab_save_complete(QEMUFile
*f
, void *opaque
)
1282 sPAPRMachineState
*spapr
= opaque
;
1284 /* Iteration header */
1285 qemu_put_be32(f
, 0);
1290 assert(kvm_enabled());
1292 rc
= spapr_check_htab_fd(spapr
);
1297 rc
= kvmppc_save_htab(f
, spapr
->htab_fd
, MAX_KVM_BUF_SIZE
, -1);
1301 close(spapr
->htab_fd
);
1302 spapr
->htab_fd
= -1;
1304 htab_save_later_pass(f
, spapr
, -1);
1308 qemu_put_be32(f
, 0);
1309 qemu_put_be16(f
, 0);
1310 qemu_put_be16(f
, 0);
1315 static int htab_load(QEMUFile
*f
, void *opaque
, int version_id
)
1317 sPAPRMachineState
*spapr
= opaque
;
1318 uint32_t section_hdr
;
1321 if (version_id
< 1 || version_id
> 1) {
1322 fprintf(stderr
, "htab_load() bad version\n");
1326 section_hdr
= qemu_get_be32(f
);
1329 /* First section, just the hash shift */
1330 if (spapr
->htab_shift
!= section_hdr
) {
1337 assert(kvm_enabled());
1339 fd
= kvmppc_get_htab_fd(true);
1341 fprintf(stderr
, "Unable to open fd to restore KVM hash table: %s\n",
1348 uint16_t n_valid
, n_invalid
;
1350 index
= qemu_get_be32(f
);
1351 n_valid
= qemu_get_be16(f
);
1352 n_invalid
= qemu_get_be16(f
);
1354 if ((index
== 0) && (n_valid
== 0) && (n_invalid
== 0)) {
1359 if ((index
+ n_valid
+ n_invalid
) >
1360 (HTAB_SIZE(spapr
) / HASH_PTE_SIZE_64
)) {
1361 /* Bad index in stream */
1362 fprintf(stderr
, "htab_load() bad index %d (%hd+%hd entries) "
1363 "in htab stream (htab_shift=%d)\n", index
, n_valid
, n_invalid
,
1370 qemu_get_buffer(f
, HPTE(spapr
->htab
, index
),
1371 HASH_PTE_SIZE_64
* n_valid
);
1374 memset(HPTE(spapr
->htab
, index
+ n_valid
), 0,
1375 HASH_PTE_SIZE_64
* n_invalid
);
1382 rc
= kvmppc_load_htab_chunk(f
, fd
, index
, n_valid
, n_invalid
);
1397 static SaveVMHandlers savevm_htab_handlers
= {
1398 .save_live_setup
= htab_save_setup
,
1399 .save_live_iterate
= htab_save_iterate
,
1400 .save_live_complete
= htab_save_complete
,
1401 .load_state
= htab_load
,
1404 static void spapr_boot_set(void *opaque
, const char *boot_device
,
1407 MachineState
*machine
= MACHINE(qdev_get_machine());
1408 machine
->boot_order
= g_strdup(boot_device
);
1411 static void spapr_cpu_init(sPAPRMachineState
*spapr
, PowerPCCPU
*cpu
)
1413 CPUPPCState
*env
= &cpu
->env
;
1415 /* Set time-base frequency to 512 MHz */
1416 cpu_ppc_tb_init(env
, TIMEBASE_FREQ
);
1418 /* PAPR always has exception vectors in RAM not ROM. To ensure this,
1419 * MSR[IP] should never be set.
1421 env
->msr_mask
&= ~(1 << 6);
1423 /* Tell KVM that we're in PAPR mode */
1424 if (kvm_enabled()) {
1425 kvmppc_set_papr(cpu
);
1428 if (cpu
->max_compat
) {
1429 if (ppc_set_compat(cpu
, cpu
->max_compat
) < 0) {
1434 xics_cpu_setup(spapr
->icp
, cpu
);
1436 qemu_register_reset(spapr_cpu_reset
, cpu
);
1439 /* pSeries LPAR / sPAPR hardware init */
1440 static void ppc_spapr_init(MachineState
*machine
)
1442 sPAPRMachineState
*spapr
= SPAPR_MACHINE(machine
);
1443 const char *cpu_model
= machine
->cpu_model
;
1444 const char *kernel_filename
= machine
->kernel_filename
;
1445 const char *kernel_cmdline
= machine
->kernel_cmdline
;
1446 const char *initrd_filename
= machine
->initrd_filename
;
1450 MemoryRegion
*sysmem
= get_system_memory();
1451 MemoryRegion
*ram
= g_new(MemoryRegion
, 1);
1452 MemoryRegion
*rma_region
;
1454 hwaddr rma_alloc_size
;
1455 hwaddr node0_size
= spapr_node0_size();
1456 uint32_t initrd_base
= 0;
1457 long kernel_size
= 0, initrd_size
= 0;
1458 long load_limit
, fw_size
;
1459 bool kernel_le
= false;
1462 msi_supported
= true;
1464 QLIST_INIT(&spapr
->phbs
);
1466 cpu_ppc_hypercall
= emulate_spapr_hypercall
;
1468 /* Allocate RMA if necessary */
1469 rma_alloc_size
= kvmppc_alloc_rma(&rma
);
1471 if (rma_alloc_size
== -1) {
1472 error_report("Unable to create RMA");
1476 if (rma_alloc_size
&& (rma_alloc_size
< node0_size
)) {
1477 spapr
->rma_size
= rma_alloc_size
;
1479 spapr
->rma_size
= node0_size
;
1481 /* With KVM, we don't actually know whether KVM supports an
1482 * unbounded RMA (PR KVM) or is limited by the hash table size
1483 * (HV KVM using VRMA), so we always assume the latter
1485 * In that case, we also limit the initial allocations for RTAS
1486 * etc... to 256M since we have no way to know what the VRMA size
1487 * is going to be as it depends on the size of the hash table
1488 * isn't determined yet.
1490 if (kvm_enabled()) {
1491 spapr
->vrma_adjust
= 1;
1492 spapr
->rma_size
= MIN(spapr
->rma_size
, 0x10000000);
1496 if (spapr
->rma_size
> node0_size
) {
1497 fprintf(stderr
, "Error: Numa node 0 has to span the RMA (%#08"HWADDR_PRIx
")\n",
1502 /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */
1503 load_limit
= MIN(spapr
->rma_size
, RTAS_MAX_ADDR
) - FW_OVERHEAD
;
1505 /* We aim for a hash table of size 1/128 the size of RAM. The
1506 * normal rule of thumb is 1/64 the size of RAM, but that's much
1507 * more than needed for the Linux guests we support. */
1508 spapr
->htab_shift
= 18; /* Minimum architected size */
1509 while (spapr
->htab_shift
<= 46) {
1510 if ((1ULL << (spapr
->htab_shift
+ 7)) >= machine
->ram_size
) {
1513 spapr
->htab_shift
++;
1516 /* Set up Interrupt Controller before we create the VCPUs */
1517 spapr
->icp
= xics_system_init(machine
,
1518 DIV_ROUND_UP(max_cpus
* kvmppc_smt_threads(),
1523 if (cpu_model
== NULL
) {
1524 cpu_model
= kvm_enabled() ? "host" : "POWER7";
1526 for (i
= 0; i
< smp_cpus
; i
++) {
1527 cpu
= cpu_ppc_init(cpu_model
);
1529 fprintf(stderr
, "Unable to find PowerPC CPU definition\n");
1532 spapr_cpu_init(spapr
, cpu
);
1535 if (kvm_enabled()) {
1536 /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */
1537 kvmppc_enable_logical_ci_hcalls();
1541 memory_region_allocate_system_memory(ram
, NULL
, "ppc_spapr.ram",
1543 memory_region_add_subregion(sysmem
, 0, ram
);
1545 if (rma_alloc_size
&& rma
) {
1546 rma_region
= g_new(MemoryRegion
, 1);
1547 memory_region_init_ram_ptr(rma_region
, NULL
, "ppc_spapr.rma",
1548 rma_alloc_size
, rma
);
1549 vmstate_register_ram_global(rma_region
);
1550 memory_region_add_subregion(sysmem
, 0, rma_region
);
1553 filename
= qemu_find_file(QEMU_FILE_TYPE_BIOS
, "spapr-rtas.bin");
1555 error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin");
1558 spapr
->rtas_size
= get_image_size(filename
);
1559 spapr
->rtas_blob
= g_malloc(spapr
->rtas_size
);
1560 if (load_image_size(filename
, spapr
->rtas_blob
, spapr
->rtas_size
) < 0) {
1561 error_report("Could not load LPAR rtas '%s'", filename
);
1564 if (spapr
->rtas_size
> RTAS_MAX_SIZE
) {
1565 error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)",
1566 (size_t)spapr
->rtas_size
, RTAS_MAX_SIZE
);
1571 /* Set up EPOW events infrastructure */
1572 spapr_events_init(spapr
);
1574 /* Set up the RTC RTAS interfaces */
1575 spapr_rtc_create(spapr
);
1577 /* Set up VIO bus */
1578 spapr
->vio_bus
= spapr_vio_bus_init();
1580 for (i
= 0; i
< MAX_SERIAL_PORTS
; i
++) {
1581 if (serial_hds
[i
]) {
1582 spapr_vty_create(spapr
->vio_bus
, serial_hds
[i
]);
1586 /* We always have at least the nvram device on VIO */
1587 spapr_create_nvram(spapr
);
1590 spapr_pci_rtas_init();
1592 phb
= spapr_create_phb(spapr
, 0);
1594 for (i
= 0; i
< nb_nics
; i
++) {
1595 NICInfo
*nd
= &nd_table
[i
];
1598 nd
->model
= g_strdup("ibmveth");
1601 if (strcmp(nd
->model
, "ibmveth") == 0) {
1602 spapr_vlan_create(spapr
->vio_bus
, nd
);
1604 pci_nic_init_nofail(&nd_table
[i
], phb
->bus
, nd
->model
, NULL
);
1608 for (i
= 0; i
<= drive_get_max_bus(IF_SCSI
); i
++) {
1609 spapr_vscsi_create(spapr
->vio_bus
);
1613 if (spapr_vga_init(phb
->bus
)) {
1614 spapr
->has_graphics
= true;
1615 machine
->usb
|= defaults_enabled() && !machine
->usb_disabled
;
1619 pci_create_simple(phb
->bus
, -1, "pci-ohci");
1621 if (spapr
->has_graphics
) {
1622 USBBus
*usb_bus
= usb_bus_find(-1);
1624 usb_create_simple(usb_bus
, "usb-kbd");
1625 usb_create_simple(usb_bus
, "usb-mouse");
1629 if (spapr
->rma_size
< (MIN_RMA_SLOF
<< 20)) {
1630 fprintf(stderr
, "qemu: pSeries SLOF firmware requires >= "
1631 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF
);
1635 if (kernel_filename
) {
1636 uint64_t lowaddr
= 0;
1638 kernel_size
= load_elf(kernel_filename
, translate_kernel_address
, NULL
,
1639 NULL
, &lowaddr
, NULL
, 1, ELF_MACHINE
, 0);
1640 if (kernel_size
== ELF_LOAD_WRONG_ENDIAN
) {
1641 kernel_size
= load_elf(kernel_filename
,
1642 translate_kernel_address
, NULL
,
1643 NULL
, &lowaddr
, NULL
, 0, ELF_MACHINE
, 0);
1644 kernel_le
= kernel_size
> 0;
1646 if (kernel_size
< 0) {
1647 fprintf(stderr
, "qemu: error loading %s: %s\n",
1648 kernel_filename
, load_elf_strerror(kernel_size
));
1653 if (initrd_filename
) {
1654 /* Try to locate the initrd in the gap between the kernel
1655 * and the firmware. Add a bit of space just in case
1657 initrd_base
= (KERNEL_LOAD_ADDR
+ kernel_size
+ 0x1ffff) & ~0xffff;
1658 initrd_size
= load_image_targphys(initrd_filename
, initrd_base
,
1659 load_limit
- initrd_base
);
1660 if (initrd_size
< 0) {
1661 fprintf(stderr
, "qemu: could not load initial ram disk '%s'\n",
1671 if (bios_name
== NULL
) {
1672 bios_name
= FW_FILE_NAME
;
1674 filename
= qemu_find_file(QEMU_FILE_TYPE_BIOS
, bios_name
);
1676 error_report("Could not find LPAR firmware '%s'", bios_name
);
1679 fw_size
= load_image_targphys(filename
, 0, FW_MAX_SIZE
);
1681 error_report("Could not load LPAR firmware '%s'", filename
);
1686 /* FIXME: Should register things through the MachineState's qdev
1687 * interface, this is a legacy from the sPAPREnvironment structure
1688 * which predated MachineState but had a similar function */
1689 vmstate_register(NULL
, 0, &vmstate_spapr
, spapr
);
1690 register_savevm_live(NULL
, "spapr/htab", -1, 1,
1691 &savevm_htab_handlers
, spapr
);
1693 /* Prepare the device tree */
1694 spapr
->fdt_skel
= spapr_create_fdt_skel(initrd_base
, initrd_size
,
1695 kernel_size
, kernel_le
,
1697 spapr
->check_exception_irq
);
1698 assert(spapr
->fdt_skel
!= NULL
);
1701 QTAILQ_INIT(&spapr
->ccs_list
);
1702 qemu_register_reset(spapr_ccs_reset_hook
, spapr
);
1704 qemu_register_boot_set(spapr_boot_set
, spapr
);
1707 static int spapr_kvm_type(const char *vm_type
)
1713 if (!strcmp(vm_type
, "HV")) {
1717 if (!strcmp(vm_type
, "PR")) {
1721 error_report("Unknown kvm-type specified '%s'", vm_type
);
1726 * Implementation of an interface to adjust firmware path
1727 * for the bootindex property handling.
1729 static char *spapr_get_fw_dev_path(FWPathProvider
*p
, BusState
*bus
,
1732 #define CAST(type, obj, name) \
1733 ((type *)object_dynamic_cast(OBJECT(obj), (name)))
1734 SCSIDevice
*d
= CAST(SCSIDevice
, dev
, TYPE_SCSI_DEVICE
);
1735 sPAPRPHBState
*phb
= CAST(sPAPRPHBState
, dev
, TYPE_SPAPR_PCI_HOST_BRIDGE
);
1738 void *spapr
= CAST(void, bus
->parent
, "spapr-vscsi");
1739 VirtIOSCSI
*virtio
= CAST(VirtIOSCSI
, bus
->parent
, TYPE_VIRTIO_SCSI
);
1740 USBDevice
*usb
= CAST(USBDevice
, bus
->parent
, TYPE_USB_DEVICE
);
1744 * Replace "channel@0/disk@0,0" with "disk@8000000000000000":
1745 * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun
1746 * in the top 16 bits of the 64-bit LUN
1748 unsigned id
= 0x8000 | (d
->id
<< 8) | d
->lun
;
1749 return g_strdup_printf("%s@%"PRIX64
, qdev_fw_name(dev
),
1750 (uint64_t)id
<< 48);
1751 } else if (virtio
) {
1753 * We use SRP luns of the form 01000000 | (target << 8) | lun
1754 * in the top 32 bits of the 64-bit LUN
1755 * Note: the quote above is from SLOF and it is wrong,
1756 * the actual binding is:
1757 * swap 0100 or 10 << or 20 << ( target lun-id -- srplun )
1759 unsigned id
= 0x1000000 | (d
->id
<< 16) | d
->lun
;
1760 return g_strdup_printf("%s@%"PRIX64
, qdev_fw_name(dev
),
1761 (uint64_t)id
<< 32);
1764 * We use SRP luns of the form 01000000 | (usb-port << 16) | lun
1765 * in the top 32 bits of the 64-bit LUN
1767 unsigned usb_port
= atoi(usb
->port
->path
);
1768 unsigned id
= 0x1000000 | (usb_port
<< 16) | d
->lun
;
1769 return g_strdup_printf("%s@%"PRIX64
, qdev_fw_name(dev
),
1770 (uint64_t)id
<< 32);
1775 /* Replace "pci" with "pci@800000020000000" */
1776 return g_strdup_printf("pci@%"PRIX64
, phb
->buid
);
1782 static char *spapr_get_kvm_type(Object
*obj
, Error
**errp
)
1784 sPAPRMachineState
*spapr
= SPAPR_MACHINE(obj
);
1786 return g_strdup(spapr
->kvm_type
);
1789 static void spapr_set_kvm_type(Object
*obj
, const char *value
, Error
**errp
)
1791 sPAPRMachineState
*spapr
= SPAPR_MACHINE(obj
);
1793 g_free(spapr
->kvm_type
);
1794 spapr
->kvm_type
= g_strdup(value
);
1797 static void spapr_machine_initfn(Object
*obj
)
1799 object_property_add_str(obj
, "kvm-type",
1800 spapr_get_kvm_type
, spapr_set_kvm_type
, NULL
);
1801 object_property_set_description(obj
, "kvm-type",
1802 "Specifies the KVM virtualization mode (HV, PR)",
1806 static void ppc_cpu_do_nmi_on_cpu(void *arg
)
1810 cpu_synchronize_state(cs
);
1811 ppc_cpu_do_system_reset(cs
);
1814 static void spapr_nmi(NMIState
*n
, int cpu_index
, Error
**errp
)
1819 async_run_on_cpu(cs
, ppc_cpu_do_nmi_on_cpu
, cs
);
1823 static void spapr_machine_class_init(ObjectClass
*oc
, void *data
)
1825 MachineClass
*mc
= MACHINE_CLASS(oc
);
1826 FWPathProviderClass
*fwc
= FW_PATH_PROVIDER_CLASS(oc
);
1827 NMIClass
*nc
= NMI_CLASS(oc
);
1829 mc
->init
= ppc_spapr_init
;
1830 mc
->reset
= ppc_spapr_reset
;
1831 mc
->block_default_type
= IF_SCSI
;
1832 mc
->max_cpus
= MAX_CPUS
;
1833 mc
->no_parallel
= 1;
1834 mc
->default_boot_order
= "";
1835 mc
->default_ram_size
= 512 * M_BYTE
;
1836 mc
->kvm_type
= spapr_kvm_type
;
1837 mc
->has_dynamic_sysbus
= true;
1839 fwc
->get_dev_path
= spapr_get_fw_dev_path
;
1840 nc
->nmi_monitor_handler
= spapr_nmi
;
1843 static const TypeInfo spapr_machine_info
= {
1844 .name
= TYPE_SPAPR_MACHINE
,
1845 .parent
= TYPE_MACHINE
,
1847 .instance_size
= sizeof(sPAPRMachineState
),
1848 .instance_init
= spapr_machine_initfn
,
1849 .class_size
= sizeof(sPAPRMachineClass
),
1850 .class_init
= spapr_machine_class_init
,
1851 .interfaces
= (InterfaceInfo
[]) {
1852 { TYPE_FW_PATH_PROVIDER
},
1858 #define SPAPR_COMPAT_2_3 \
1861 .driver = "spapr-pci-host-bridge",\
1862 .property = "dynamic-reconfiguration",\
1866 #define SPAPR_COMPAT_2_2 \
1870 .driver = TYPE_SPAPR_PCI_HOST_BRIDGE,\
1871 .property = "mem_win_size",\
1872 .value = "0x20000000",\
1875 #define SPAPR_COMPAT_2_1 \
1879 static void spapr_compat_2_3(Object
*obj
)
1883 static void spapr_compat_2_2(Object
*obj
)
1885 spapr_compat_2_3(obj
);
1888 static void spapr_compat_2_1(Object
*obj
)
1890 spapr_compat_2_2(obj
);
1893 static void spapr_machine_2_3_instance_init(Object
*obj
)
1895 spapr_compat_2_3(obj
);
1896 spapr_machine_initfn(obj
);
1899 static void spapr_machine_2_2_instance_init(Object
*obj
)
1901 spapr_compat_2_2(obj
);
1902 spapr_machine_initfn(obj
);
1905 static void spapr_machine_2_1_instance_init(Object
*obj
)
1907 spapr_compat_2_1(obj
);
1908 spapr_machine_initfn(obj
);
1911 static void spapr_machine_2_1_class_init(ObjectClass
*oc
, void *data
)
1913 MachineClass
*mc
= MACHINE_CLASS(oc
);
1914 static GlobalProperty compat_props
[] = {
1916 { /* end of list */ }
1919 mc
->name
= "pseries-2.1";
1920 mc
->desc
= "pSeries Logical Partition (PAPR compliant) v2.1";
1921 mc
->compat_props
= compat_props
;
1924 static const TypeInfo spapr_machine_2_1_info
= {
1925 .name
= TYPE_SPAPR_MACHINE
"2.1",
1926 .parent
= TYPE_SPAPR_MACHINE
,
1927 .class_init
= spapr_machine_2_1_class_init
,
1928 .instance_init
= spapr_machine_2_1_instance_init
,
1931 static void spapr_machine_2_2_class_init(ObjectClass
*oc
, void *data
)
1933 static GlobalProperty compat_props
[] = {
1935 { /* end of list */ }
1937 MachineClass
*mc
= MACHINE_CLASS(oc
);
1939 mc
->name
= "pseries-2.2";
1940 mc
->desc
= "pSeries Logical Partition (PAPR compliant) v2.2";
1941 mc
->compat_props
= compat_props
;
1944 static const TypeInfo spapr_machine_2_2_info
= {
1945 .name
= TYPE_SPAPR_MACHINE
"2.2",
1946 .parent
= TYPE_SPAPR_MACHINE
,
1947 .class_init
= spapr_machine_2_2_class_init
,
1948 .instance_init
= spapr_machine_2_2_instance_init
,
1951 static void spapr_machine_2_3_class_init(ObjectClass
*oc
, void *data
)
1953 static GlobalProperty compat_props
[] = {
1955 { /* end of list */ }
1957 MachineClass
*mc
= MACHINE_CLASS(oc
);
1959 mc
->name
= "pseries-2.3";
1960 mc
->desc
= "pSeries Logical Partition (PAPR compliant) v2.3";
1961 mc
->compat_props
= compat_props
;
1964 static const TypeInfo spapr_machine_2_3_info
= {
1965 .name
= TYPE_SPAPR_MACHINE
"2.3",
1966 .parent
= TYPE_SPAPR_MACHINE
,
1967 .class_init
= spapr_machine_2_3_class_init
,
1968 .instance_init
= spapr_machine_2_3_instance_init
,
1971 static void spapr_machine_2_4_class_init(ObjectClass
*oc
, void *data
)
1973 MachineClass
*mc
= MACHINE_CLASS(oc
);
1975 mc
->name
= "pseries-2.4";
1976 mc
->desc
= "pSeries Logical Partition (PAPR compliant) v2.4";
1977 mc
->alias
= "pseries";
1981 static const TypeInfo spapr_machine_2_4_info
= {
1982 .name
= TYPE_SPAPR_MACHINE
"2.4",
1983 .parent
= TYPE_SPAPR_MACHINE
,
1984 .class_init
= spapr_machine_2_4_class_init
,
1987 static void spapr_machine_register_types(void)
1989 type_register_static(&spapr_machine_info
);
1990 type_register_static(&spapr_machine_2_1_info
);
1991 type_register_static(&spapr_machine_2_2_info
);
1992 type_register_static(&spapr_machine_2_3_info
);
1993 type_register_static(&spapr_machine_2_4_info
);
1996 type_init(spapr_machine_register_types
)