]> git.proxmox.com Git - mirror_qemu.git/blame - hw/spapr.c
vga: remove CONFIG_BOCHS_VBE
[mirror_qemu.git] / hw / spapr.c
CommitLineData
9fdf0c29
DG
1/*
2 * QEMU PowerPC pSeries Logical Partition (aka sPAPR) hardware System Emulator
3 *
4 * Copyright (c) 2004-2007 Fabrice Bellard
5 * Copyright (c) 2007 Jocelyn Mayer
6 * Copyright (c) 2010 David Gibson, IBM Corporation.
7 *
8 * Permission is hereby granted, free of charge, to any person obtaining a copy
9 * of this software and associated documentation files (the "Software"), to deal
10 * in the Software without restriction, including without limitation the rights
11 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
12 * copies of the Software, and to permit persons to whom the Software is
13 * furnished to do so, subject to the following conditions:
14 *
15 * The above copyright notice and this permission notice shall be included in
16 * all copies or substantial portions of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
21 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
22 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
23 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24 * THE SOFTWARE.
25 *
26 */
27#include "sysemu.h"
9fdf0c29
DG
28#include "hw.h"
29#include "elf.h"
8d90ad90 30#include "net.h"
6e270446 31#include "blockdev.h"
e97c3636
DG
32#include "cpus.h"
33#include "kvm.h"
34#include "kvm_ppc.h"
9fdf0c29
DG
35
36#include "hw/boards.h"
37#include "hw/ppc.h"
38#include "hw/loader.h"
39
40#include "hw/spapr.h"
4040ab72 41#include "hw/spapr_vio.h"
3384f95c 42#include "hw/spapr_pci.h"
b5cec4c5 43#include "hw/xics.h"
0ee2c058 44#include "hw/msi.h"
9fdf0c29 45
f61b4bed
AG
46#include "kvm.h"
47#include "kvm_ppc.h"
3384f95c 48#include "pci.h"
f61b4bed 49
890c2b77 50#include "exec-memory.h"
35139a59 51#include "hw/usb.h"
890c2b77 52
9fdf0c29
DG
53#include <libfdt.h>
54
4d8d5467
BH
55/* SLOF memory layout:
56 *
57 * SLOF raw image loaded at 0, copies its romfs right below the flat
58 * device-tree, then position SLOF itself 31M below that
59 *
60 * So we set FW_OVERHEAD to 40MB which should account for all of that
61 * and more
62 *
63 * We load our kernel at 4M, leaving space for SLOF initial image
64 */
9fdf0c29 65#define FDT_MAX_SIZE 0x10000
39ac8455 66#define RTAS_MAX_SIZE 0x10000
a9f8ad8f
DG
67#define FW_MAX_SIZE 0x400000
68#define FW_FILE_NAME "slof.bin"
4d8d5467
BH
69#define FW_OVERHEAD 0x2800000
70#define KERNEL_LOAD_ADDR FW_MAX_SIZE
a9f8ad8f 71
4d8d5467 72#define MIN_RMA_SLOF 128UL
9fdf0c29
DG
73
74#define TIMEBASE_FREQ 512000000ULL
75
41019fec 76#define MAX_CPUS 256
4d8d5467 77#define XICS_IRQS 1024
9fdf0c29 78
3384f95c
DG
79#define SPAPR_PCI_BUID 0x800000020000001ULL
80#define SPAPR_PCI_MEM_WIN_ADDR (0x10000000000ULL + 0xA0000000)
81#define SPAPR_PCI_MEM_WIN_SIZE 0x20000000
82#define SPAPR_PCI_IO_WIN_ADDR (0x10000000000ULL + 0x80000000)
0ee2c058 83#define SPAPR_PCI_MSI_WIN_ADDR (0x10000000000ULL + 0x90000000)
3384f95c 84
0c103f8e
DG
85#define PHANDLE_XICP 0x00001111
86
7f763a5d
DG
87#define HTAB_SIZE(spapr) (1ULL << ((spapr)->htab_shift))
88
9fdf0c29
DG
89sPAPREnvironment *spapr;
90
ff9d2afa 91int spapr_allocate_irq(int hint, bool lsi)
e6c866d4 92{
a307d594 93 int irq;
e6c866d4
DG
94
95 if (hint) {
96 irq = hint;
97 /* FIXME: we should probably check for collisions somehow */
98 } else {
99 irq = spapr->next_irq++;
100 }
101
a307d594
AK
102 /* Configure irq type */
103 if (!xics_get_qirq(spapr->icp, irq)) {
104 return 0;
e6c866d4
DG
105 }
106
ff9d2afa 107 xics_set_irq_type(spapr->icp, irq, lsi);
e6c866d4 108
a307d594 109 return irq;
e6c866d4
DG
110}
111
f4b9523b 112/* Allocate block of consequtive IRQs, returns a number of the first */
ff9d2afa 113int spapr_allocate_irq_block(int num, bool lsi)
f4b9523b
AK
114{
115 int first = -1;
116 int i;
117
118 for (i = 0; i < num; ++i) {
119 int irq;
120
ff9d2afa 121 irq = spapr_allocate_irq(0, lsi);
f4b9523b
AK
122 if (!irq) {
123 return -1;
124 }
125
126 if (0 == i) {
127 first = irq;
128 }
129
130 /* If the above doesn't create a consecutive block then that's
131 * an internal bug */
132 assert(irq == (first + i));
133 }
134
135 return first;
136}
137
7f763a5d 138static int spapr_fixup_cpu_dt(void *fdt, sPAPREnvironment *spapr)
6e806cc3
BR
139{
140 int ret = 0, offset;
e2684c0b 141 CPUPPCState *env;
6e806cc3
BR
142 char cpu_model[32];
143 int smt = kvmppc_smt_threads();
7f763a5d 144 uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
6e806cc3
BR
145
146 assert(spapr->cpu_model);
147
148 for (env = first_cpu; env != NULL; env = env->next_cpu) {
149 uint32_t associativity[] = {cpu_to_be32(0x5),
150 cpu_to_be32(0x0),
151 cpu_to_be32(0x0),
152 cpu_to_be32(0x0),
153 cpu_to_be32(env->numa_node),
154 cpu_to_be32(env->cpu_index)};
155
156 if ((env->cpu_index % smt) != 0) {
157 continue;
158 }
159
160 snprintf(cpu_model, 32, "/cpus/%s@%x", spapr->cpu_model,
161 env->cpu_index);
162
163 offset = fdt_path_offset(fdt, cpu_model);
164 if (offset < 0) {
165 return offset;
166 }
167
7f763a5d
DG
168 if (nb_numa_nodes > 1) {
169 ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
170 sizeof(associativity));
171 if (ret < 0) {
172 return ret;
173 }
174 }
175
176 ret = fdt_setprop(fdt, offset, "ibm,pft-size",
177 pft_size_prop, sizeof(pft_size_prop));
6e806cc3
BR
178 if (ret < 0) {
179 return ret;
180 }
181 }
182 return ret;
183}
184
5af9873d
BH
185
186static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
187 size_t maxsize)
188{
189 size_t maxcells = maxsize / sizeof(uint32_t);
190 int i, j, count;
191 uint32_t *p = prop;
192
193 for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
194 struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
195
196 if (!sps->page_shift) {
197 break;
198 }
199 for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
200 if (sps->enc[count].page_shift == 0) {
201 break;
202 }
203 }
204 if ((p - prop) >= (maxcells - 3 - count * 2)) {
205 break;
206 }
207 *(p++) = cpu_to_be32(sps->page_shift);
208 *(p++) = cpu_to_be32(sps->slb_enc);
209 *(p++) = cpu_to_be32(count);
210 for (j = 0; j < count; j++) {
211 *(p++) = cpu_to_be32(sps->enc[j].page_shift);
212 *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
213 }
214 }
215
216 return (p - prop) * sizeof(uint32_t);
217}
218
7f763a5d
DG
219#define _FDT(exp) \
220 do { \
221 int ret = (exp); \
222 if (ret < 0) { \
223 fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
224 #exp, fdt_strerror(ret)); \
225 exit(1); \
226 } \
227 } while (0)
228
229
a3467baa
DG
230static void *spapr_create_fdt_skel(const char *cpu_model,
231 target_phys_addr_t initrd_base,
232 target_phys_addr_t initrd_size,
4d8d5467 233 target_phys_addr_t kernel_size,
a3467baa 234 const char *boot_device,
7f763a5d 235 const char *kernel_cmdline)
9fdf0c29
DG
236{
237 void *fdt;
e2684c0b 238 CPUPPCState *env;
9fdf0c29
DG
239 uint32_t start_prop = cpu_to_be32(initrd_base);
240 uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
ee86dfee 241 char hypertas_prop[] = "hcall-pft\0hcall-term\0hcall-dabr\0hcall-interrupt"
a3d0abae 242 "\0hcall-tce\0hcall-vio\0hcall-splpar\0hcall-bulk";
c73e3771 243 char qemu_hypertas_prop[] = "hcall-memop1";
7f763a5d 244 uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
b5cec4c5 245 uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(smp_cpus)};
9fdf0c29 246 char *modelname;
7f763a5d 247 int i, smt = kvmppc_smt_threads();
6e806cc3 248 unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
9fdf0c29 249
7267c094 250 fdt = g_malloc0(FDT_MAX_SIZE);
9fdf0c29
DG
251 _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
252
4d8d5467
BH
253 if (kernel_size) {
254 _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
255 }
256 if (initrd_size) {
257 _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
258 }
9fdf0c29
DG
259 _FDT((fdt_finish_reservemap(fdt)));
260
261 /* Root node */
262 _FDT((fdt_begin_node(fdt, "")));
263 _FDT((fdt_property_string(fdt, "device_type", "chrp")));
5d73dd66 264 _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
9fdf0c29
DG
265
266 _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
267 _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
268
269 /* /chosen */
270 _FDT((fdt_begin_node(fdt, "chosen")));
271
6e806cc3
BR
272 /* Set Form1_affinity */
273 _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
274
9fdf0c29
DG
275 _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
276 _FDT((fdt_property(fdt, "linux,initrd-start",
277 &start_prop, sizeof(start_prop))));
278 _FDT((fdt_property(fdt, "linux,initrd-end",
279 &end_prop, sizeof(end_prop))));
4d8d5467
BH
280 if (kernel_size) {
281 uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
282 cpu_to_be64(kernel_size) };
9fdf0c29 283
4d8d5467
BH
284 _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
285 }
286 _FDT((fdt_property_string(fdt, "qemu,boot-device", boot_device)));
f28359d8
LZ
287 _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
288 _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
289 _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
3384f95c 290
9fdf0c29
DG
291 _FDT((fdt_end_node(fdt)));
292
9fdf0c29
DG
293 /* cpus */
294 _FDT((fdt_begin_node(fdt, "cpus")));
295
296 _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
297 _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
298
7267c094 299 modelname = g_strdup(cpu_model);
9fdf0c29
DG
300
301 for (i = 0; i < strlen(modelname); i++) {
302 modelname[i] = toupper(modelname[i]);
303 }
304
6e806cc3
BR
305 /* This is needed during FDT finalization */
306 spapr->cpu_model = g_strdup(modelname);
307
c7a5c0c9
DG
308 for (env = first_cpu; env != NULL; env = env->next_cpu) {
309 int index = env->cpu_index;
e97c3636
DG
310 uint32_t servers_prop[smp_threads];
311 uint32_t gservers_prop[smp_threads * 2];
9fdf0c29
DG
312 char *nodename;
313 uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
314 0xffffffff, 0xffffffff};
0a8b2938
AG
315 uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
316 uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
5af9873d
BH
317 uint32_t page_sizes_prop[64];
318 size_t page_sizes_prop_size;
9fdf0c29 319
e97c3636
DG
320 if ((index % smt) != 0) {
321 continue;
322 }
323
c7a5c0c9 324 if (asprintf(&nodename, "%s@%x", modelname, index) < 0) {
9fdf0c29
DG
325 fprintf(stderr, "Allocation failure\n");
326 exit(1);
327 }
328
329 _FDT((fdt_begin_node(fdt, nodename)));
330
331 free(nodename);
332
c7a5c0c9 333 _FDT((fdt_property_cell(fdt, "reg", index)));
9fdf0c29
DG
334 _FDT((fdt_property_string(fdt, "device_type", "cpu")));
335
336 _FDT((fdt_property_cell(fdt, "cpu-version", env->spr[SPR_PVR])));
337 _FDT((fdt_property_cell(fdt, "dcache-block-size",
338 env->dcache_line_size)));
339 _FDT((fdt_property_cell(fdt, "icache-block-size",
340 env->icache_line_size)));
0a8b2938
AG
341 _FDT((fdt_property_cell(fdt, "timebase-frequency", tbfreq)));
342 _FDT((fdt_property_cell(fdt, "clock-frequency", cpufreq)));
9fdf0c29
DG
343 _FDT((fdt_property_cell(fdt, "ibm,slb-size", env->slb_nr)));
344 _FDT((fdt_property_string(fdt, "status", "okay")));
345 _FDT((fdt_property(fdt, "64-bit", NULL, 0)));
e97c3636
DG
346
347 /* Build interrupt servers and gservers properties */
348 for (i = 0; i < smp_threads; i++) {
349 servers_prop[i] = cpu_to_be32(index + i);
350 /* Hack, direct the group queues back to cpu 0 */
351 gservers_prop[i*2] = cpu_to_be32(index + i);
352 gservers_prop[i*2 + 1] = 0;
353 }
354 _FDT((fdt_property(fdt, "ibm,ppc-interrupt-server#s",
355 servers_prop, sizeof(servers_prop))));
b5cec4c5 356 _FDT((fdt_property(fdt, "ibm,ppc-interrupt-gserver#s",
e97c3636 357 gservers_prop, sizeof(gservers_prop))));
9fdf0c29 358
c7a5c0c9 359 if (env->mmu_model & POWERPC_MMU_1TSEG) {
9fdf0c29
DG
360 _FDT((fdt_property(fdt, "ibm,processor-segment-sizes",
361 segs, sizeof(segs))));
362 }
363
6659394f
DG
364 /* Advertise VMX/VSX (vector extensions) if available
365 * 0 / no property == no vector extensions
366 * 1 == VMX / Altivec available
367 * 2 == VSX available */
a7342588
DG
368 if (env->insns_flags & PPC_ALTIVEC) {
369 uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
370
6659394f
DG
371 _FDT((fdt_property_cell(fdt, "ibm,vmx", vmx)));
372 }
373
374 /* Advertise DFP (Decimal Floating Point) if available
375 * 0 / no property == no DFP
376 * 1 == DFP available */
a7342588
DG
377 if (env->insns_flags2 & PPC2_DFP) {
378 _FDT((fdt_property_cell(fdt, "ibm,dfp", 1)));
6659394f
DG
379 }
380
5af9873d
BH
381 page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
382 sizeof(page_sizes_prop));
383 if (page_sizes_prop_size) {
384 _FDT((fdt_property(fdt, "ibm,segment-page-sizes",
385 page_sizes_prop, page_sizes_prop_size)));
386 }
387
9fdf0c29
DG
388 _FDT((fdt_end_node(fdt)));
389 }
390
7267c094 391 g_free(modelname);
9fdf0c29
DG
392
393 _FDT((fdt_end_node(fdt)));
394
f43e3525
DG
395 /* RTAS */
396 _FDT((fdt_begin_node(fdt, "rtas")));
397
398 _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas_prop,
399 sizeof(hypertas_prop))));
c73e3771
BH
400 _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas_prop,
401 sizeof(qemu_hypertas_prop))));
f43e3525 402
6e806cc3
BR
403 _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
404 refpoints, sizeof(refpoints))));
405
f43e3525
DG
406 _FDT((fdt_end_node(fdt)));
407
b5cec4c5 408 /* interrupt controller */
9dfef5aa 409 _FDT((fdt_begin_node(fdt, "interrupt-controller")));
b5cec4c5
DG
410
411 _FDT((fdt_property_string(fdt, "device_type",
412 "PowerPC-External-Interrupt-Presentation")));
413 _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
b5cec4c5
DG
414 _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
415 _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
416 interrupt_server_ranges_prop,
417 sizeof(interrupt_server_ranges_prop))));
0c103f8e
DG
418 _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
419 _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
420 _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
b5cec4c5
DG
421
422 _FDT((fdt_end_node(fdt)));
423
4040ab72
DG
424 /* vdevice */
425 _FDT((fdt_begin_node(fdt, "vdevice")));
426
427 _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
428 _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
429 _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
430 _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
b5cec4c5
DG
431 _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
432 _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
4040ab72
DG
433
434 _FDT((fdt_end_node(fdt)));
435
9fdf0c29
DG
436 _FDT((fdt_end_node(fdt))); /* close root node */
437 _FDT((fdt_finish(fdt)));
438
a3467baa
DG
439 return fdt;
440}
441
7f763a5d
DG
442static int spapr_populate_memory(sPAPREnvironment *spapr, void *fdt)
443{
444 uint32_t associativity[] = {cpu_to_be32(0x4), cpu_to_be32(0x0),
445 cpu_to_be32(0x0), cpu_to_be32(0x0),
446 cpu_to_be32(0x0)};
447 char mem_name[32];
448 target_phys_addr_t node0_size, mem_start;
449 uint64_t mem_reg_property[2];
450 int i, off;
451
452 /* memory node(s) */
453 node0_size = (nb_numa_nodes > 1) ? node_mem[0] : ram_size;
454 if (spapr->rma_size > node0_size) {
455 spapr->rma_size = node0_size;
456 }
457
458 /* RMA */
459 mem_reg_property[0] = 0;
460 mem_reg_property[1] = cpu_to_be64(spapr->rma_size);
461 off = fdt_add_subnode(fdt, 0, "memory@0");
462 _FDT(off);
463 _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
464 _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
465 sizeof(mem_reg_property))));
466 _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
467 sizeof(associativity))));
468
469 /* RAM: Node 0 */
470 if (node0_size > spapr->rma_size) {
471 mem_reg_property[0] = cpu_to_be64(spapr->rma_size);
472 mem_reg_property[1] = cpu_to_be64(node0_size - spapr->rma_size);
473
474 sprintf(mem_name, "memory@" TARGET_FMT_lx, spapr->rma_size);
475 off = fdt_add_subnode(fdt, 0, mem_name);
476 _FDT(off);
477 _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
478 _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
479 sizeof(mem_reg_property))));
480 _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
481 sizeof(associativity))));
482 }
483
484 /* RAM: Node 1 and beyond */
485 mem_start = node0_size;
486 for (i = 1; i < nb_numa_nodes; i++) {
487 mem_reg_property[0] = cpu_to_be64(mem_start);
488 mem_reg_property[1] = cpu_to_be64(node_mem[i]);
489 associativity[3] = associativity[4] = cpu_to_be32(i);
490 sprintf(mem_name, "memory@" TARGET_FMT_lx, mem_start);
491 off = fdt_add_subnode(fdt, 0, mem_name);
492 _FDT(off);
493 _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
494 _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
495 sizeof(mem_reg_property))));
496 _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
497 sizeof(associativity))));
498 mem_start += node_mem[i];
499 }
500
501 return 0;
502}
503
a3467baa
DG
504static void spapr_finalize_fdt(sPAPREnvironment *spapr,
505 target_phys_addr_t fdt_addr,
506 target_phys_addr_t rtas_addr,
507 target_phys_addr_t rtas_size)
508{
509 int ret;
510 void *fdt;
3384f95c 511 sPAPRPHBState *phb;
a3467baa 512
7267c094 513 fdt = g_malloc(FDT_MAX_SIZE);
a3467baa
DG
514
515 /* open out the base tree into a temp buffer for the final tweaks */
516 _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
4040ab72 517
7f763a5d
DG
518 ret = spapr_populate_memory(spapr, fdt);
519 if (ret < 0) {
520 fprintf(stderr, "couldn't setup memory nodes in fdt\n");
521 exit(1);
522 }
523
4040ab72
DG
524 ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
525 if (ret < 0) {
526 fprintf(stderr, "couldn't setup vio devices in fdt\n");
527 exit(1);
528 }
529
3384f95c 530 QLIST_FOREACH(phb, &spapr->phbs, list) {
e0fdbd7c 531 ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
3384f95c
DG
532 }
533
534 if (ret < 0) {
535 fprintf(stderr, "couldn't setup PCI devices in fdt\n");
536 exit(1);
537 }
538
39ac8455
DG
539 /* RTAS */
540 ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
541 if (ret < 0) {
542 fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
543 }
544
6e806cc3 545 /* Advertise NUMA via ibm,associativity */
7f763a5d
DG
546 ret = spapr_fixup_cpu_dt(fdt, spapr);
547 if (ret < 0) {
548 fprintf(stderr, "Couldn't finalize CPU device tree properties\n");
6e806cc3
BR
549 }
550
3fc5acde 551 if (!spapr->has_graphics) {
f28359d8
LZ
552 spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
553 }
68f3a94c 554
4040ab72
DG
555 _FDT((fdt_pack(fdt)));
556
4d8d5467
BH
557 if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
558 hw_error("FDT too big ! 0x%x bytes (max is 0x%x)\n",
559 fdt_totalsize(fdt), FDT_MAX_SIZE);
560 exit(1);
561 }
562
a3467baa 563 cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
9fdf0c29 564
7267c094 565 g_free(fdt);
9fdf0c29
DG
566}
567
568static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
569{
570 return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
571}
572
e2684c0b 573static void emulate_spapr_hypercall(CPUPPCState *env)
9fdf0c29 574{
efcb9383
DG
575 if (msr_pr) {
576 hcall_dprintf("Hypercall made with MSR[PR]=1\n");
577 env->gpr[3] = H_PRIVILEGE;
578 } else {
579 env->gpr[3] = spapr_hypercall(env, env->gpr[3], &env->gpr[4]);
580 }
9fdf0c29
DG
581}
582
7f763a5d
DG
583static void spapr_reset_htab(sPAPREnvironment *spapr)
584{
585 long shift;
586
587 /* allocate hash page table. For now we always make this 16mb,
588 * later we should probably make it scale to the size of guest
589 * RAM */
590
591 shift = kvmppc_reset_htab(spapr->htab_shift);
592
593 if (shift > 0) {
594 /* Kernel handles htab, we don't need to allocate one */
595 spapr->htab_shift = shift;
596 } else {
597 if (!spapr->htab) {
598 /* Allocate an htab if we don't yet have one */
599 spapr->htab = qemu_memalign(HTAB_SIZE(spapr), HTAB_SIZE(spapr));
600 }
601
602 /* And clear it */
603 memset(spapr->htab, 0, HTAB_SIZE(spapr));
604 }
605
606 /* Update the RMA size if necessary */
607 if (spapr->vrma_adjust) {
608 spapr->rma_size = kvmppc_rma_size(ram_size, spapr->htab_shift);
609 }
9fdf0c29
DG
610}
611
c8787ad4 612static void ppc_spapr_reset(void)
a3467baa 613{
7f763a5d
DG
614 /* Reset the hash table & recalc the RMA */
615 spapr_reset_htab(spapr);
a3467baa 616
c8787ad4 617 qemu_devices_reset();
a3467baa
DG
618
619 /* Load the fdt */
620 spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
621 spapr->rtas_size);
622
623 /* Set up the entry state */
624 first_cpu->gpr[3] = spapr->fdt_addr;
625 first_cpu->gpr[5] = 0;
626 first_cpu->halted = 0;
627 first_cpu->nip = spapr->entry_point;
628
629}
630
1bba0dc9
AF
631static void spapr_cpu_reset(void *opaque)
632{
5b2038e0 633 PowerPCCPU *cpu = opaque;
048706d9 634 CPUPPCState *env = &cpu->env;
1bba0dc9 635
5b2038e0 636 cpu_reset(CPU(cpu));
048706d9
DG
637
638 /* All CPUs start halted. CPU0 is unhalted from the machine level
639 * reset code and the rest are explicitly started up by the guest
640 * using an RTAS call */
641 env->halted = 1;
642
643 env->spr[SPR_HIOR] = 0;
7f763a5d
DG
644
645 env->external_htab = spapr->htab;
646 env->htab_base = -1;
647 env->htab_mask = HTAB_SIZE(spapr) - 1;
648 env->spr[SPR_SDR1] = (unsigned long)spapr->htab |
649 (spapr->htab_shift - 18);
1bba0dc9
AF
650}
651
8c57b867 652/* Returns whether we want to use VGA or not */
f28359d8
LZ
653static int spapr_vga_init(PCIBus *pci_bus)
654{
8c57b867 655 switch (vga_interface_type) {
8c57b867 656 case VGA_NONE:
1ddcae82
AJ
657 case VGA_STD:
658 return pci_vga_init(pci_bus) != NULL;
8c57b867 659 default:
f28359d8
LZ
660 fprintf(stderr, "This vga model is not supported,"
661 "currently it only supports -vga std\n");
8c57b867
AG
662 exit(0);
663 break;
f28359d8 664 }
f28359d8
LZ
665}
666
9fdf0c29
DG
667/* pSeries LPAR / sPAPR hardware init */
668static void ppc_spapr_init(ram_addr_t ram_size,
669 const char *boot_device,
670 const char *kernel_filename,
671 const char *kernel_cmdline,
672 const char *initrd_filename,
673 const char *cpu_model)
674{
05769733 675 PowerPCCPU *cpu;
e2684c0b 676 CPUPPCState *env;
8c9f64df 677 PCIHostState *phb;
9fdf0c29 678 int i;
890c2b77
AK
679 MemoryRegion *sysmem = get_system_memory();
680 MemoryRegion *ram = g_new(MemoryRegion, 1);
7f763a5d 681 target_phys_addr_t rma_alloc_size;
4d8d5467
BH
682 uint32_t initrd_base = 0;
683 long kernel_size = 0, initrd_size = 0;
684 long load_limit, rtas_limit, fw_size;
39ac8455 685 char *filename;
9fdf0c29 686
0ee2c058
AK
687 msi_supported = true;
688
d43b45e2
DG
689 spapr = g_malloc0(sizeof(*spapr));
690 QLIST_INIT(&spapr->phbs);
691
9fdf0c29
DG
692 cpu_ppc_hypercall = emulate_spapr_hypercall;
693
354ac20a
DG
694 /* Allocate RMA if necessary */
695 rma_alloc_size = kvmppc_alloc_rma("ppc_spapr.rma", sysmem);
696
697 if (rma_alloc_size == -1) {
698 hw_error("qemu: Unable to create RMA\n");
699 exit(1);
700 }
7f763a5d 701
354ac20a 702 if (rma_alloc_size && (rma_alloc_size < ram_size)) {
7f763a5d 703 spapr->rma_size = rma_alloc_size;
354ac20a 704 } else {
7f763a5d
DG
705 spapr->rma_size = ram_size;
706
707 /* With KVM, we don't actually know whether KVM supports an
708 * unbounded RMA (PR KVM) or is limited by the hash table size
709 * (HV KVM using VRMA), so we always assume the latter
710 *
711 * In that case, we also limit the initial allocations for RTAS
712 * etc... to 256M since we have no way to know what the VRMA size
713 * is going to be as it depends on the size of the hash table
714 * isn't determined yet.
715 */
716 if (kvm_enabled()) {
717 spapr->vrma_adjust = 1;
718 spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
719 }
354ac20a
DG
720 }
721
4d8d5467 722 /* We place the device tree and RTAS just below either the top of the RMA,
354ac20a
DG
723 * or just below 2GB, whichever is lowere, so that it can be
724 * processed with 32-bit real mode code if necessary */
7f763a5d 725 rtas_limit = MIN(spapr->rma_size, 0x80000000);
4d8d5467
BH
726 spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
727 spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
728 load_limit = spapr->fdt_addr - FW_OVERHEAD;
9fdf0c29 729
382be75d
DG
730 /* We aim for a hash table of size 1/128 the size of RAM. The
731 * normal rule of thumb is 1/64 the size of RAM, but that's much
732 * more than needed for the Linux guests we support. */
733 spapr->htab_shift = 18; /* Minimum architected size */
734 while (spapr->htab_shift <= 46) {
735 if ((1ULL << (spapr->htab_shift + 7)) >= ram_size) {
736 break;
737 }
738 spapr->htab_shift++;
739 }
7f763a5d 740
9fdf0c29
DG
741 /* init CPUs */
742 if (cpu_model == NULL) {
6b7a2cf6 743 cpu_model = kvm_enabled() ? "host" : "POWER7";
9fdf0c29
DG
744 }
745 for (i = 0; i < smp_cpus; i++) {
05769733
AF
746 cpu = cpu_ppc_init(cpu_model);
747 if (cpu == NULL) {
9fdf0c29
DG
748 fprintf(stderr, "Unable to find PowerPC CPU definition\n");
749 exit(1);
750 }
05769733
AF
751 env = &cpu->env;
752
9fdf0c29
DG
753 /* Set time-base frequency to 512 MHz */
754 cpu_ppc_tb_init(env, TIMEBASE_FREQ);
9fdf0c29 755
048706d9 756 /* PAPR always has exception vectors in RAM not ROM */
9fdf0c29 757 env->hreset_excp_prefix = 0;
048706d9
DG
758
759 /* Tell KVM that we're in PAPR mode */
760 if (kvm_enabled()) {
761 kvmppc_set_papr(env);
762 }
763
764 qemu_register_reset(spapr_cpu_reset, cpu);
9fdf0c29
DG
765 }
766
767 /* allocate RAM */
f73a2575 768 spapr->ram_limit = ram_size;
354ac20a
DG
769 if (spapr->ram_limit > rma_alloc_size) {
770 ram_addr_t nonrma_base = rma_alloc_size;
771 ram_addr_t nonrma_size = spapr->ram_limit - rma_alloc_size;
772
c5705a77
AK
773 memory_region_init_ram(ram, "ppc_spapr.ram", nonrma_size);
774 vmstate_register_ram_global(ram);
354ac20a
DG
775 memory_region_add_subregion(sysmem, nonrma_base, ram);
776 }
9fdf0c29 777
39ac8455 778 filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
a3467baa 779 spapr->rtas_size = load_image_targphys(filename, spapr->rtas_addr,
4d8d5467 780 rtas_limit - spapr->rtas_addr);
a3467baa 781 if (spapr->rtas_size < 0) {
39ac8455
DG
782 hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
783 exit(1);
784 }
4d8d5467
BH
785 if (spapr->rtas_size > RTAS_MAX_SIZE) {
786 hw_error("RTAS too big ! 0x%lx bytes (max is 0x%x)\n",
787 spapr->rtas_size, RTAS_MAX_SIZE);
788 exit(1);
789 }
7267c094 790 g_free(filename);
39ac8455 791
4d8d5467 792
b5cec4c5 793 /* Set up Interrupt Controller */
c7a5c0c9 794 spapr->icp = xics_system_init(XICS_IRQS);
e6c866d4 795 spapr->next_irq = 16;
b5cec4c5 796
ad0ebb91
DG
797 /* Set up IOMMU */
798 spapr_iommu_init();
799
b5cec4c5 800 /* Set up VIO bus */
4040ab72
DG
801 spapr->vio_bus = spapr_vio_bus_init();
802
277f9acf 803 for (i = 0; i < MAX_SERIAL_PORTS; i++) {
4040ab72 804 if (serial_hds[i]) {
d601fac4 805 spapr_vty_create(spapr->vio_bus, serial_hds[i]);
4040ab72
DG
806 }
807 }
9fdf0c29 808
3384f95c 809 /* Set up PCI */
fa28f71b
AK
810 spapr_pci_rtas_init();
811
3384f95c
DG
812 spapr_create_phb(spapr, "pci", SPAPR_PCI_BUID,
813 SPAPR_PCI_MEM_WIN_ADDR,
814 SPAPR_PCI_MEM_WIN_SIZE,
0ee2c058
AK
815 SPAPR_PCI_IO_WIN_ADDR,
816 SPAPR_PCI_MSI_WIN_ADDR);
8558d942 817 phb = PCI_HOST_BRIDGE(QLIST_FIRST(&spapr->phbs));
3384f95c 818
277f9acf 819 for (i = 0; i < nb_nics; i++) {
8d90ad90
DG
820 NICInfo *nd = &nd_table[i];
821
822 if (!nd->model) {
7267c094 823 nd->model = g_strdup("ibmveth");
8d90ad90
DG
824 }
825
826 if (strcmp(nd->model, "ibmveth") == 0) {
d601fac4 827 spapr_vlan_create(spapr->vio_bus, nd);
8d90ad90 828 } else {
3384f95c 829 pci_nic_init_nofail(&nd_table[i], nd->model, NULL);
8d90ad90
DG
830 }
831 }
832
6e270446 833 for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
d601fac4 834 spapr_vscsi_create(spapr->vio_bus);
6e270446
BH
835 }
836
f28359d8 837 /* Graphics */
8c9f64df 838 if (spapr_vga_init(phb->bus)) {
3fc5acde 839 spapr->has_graphics = true;
f28359d8
LZ
840 }
841
35139a59 842 if (usb_enabled) {
8c9f64df 843 pci_create_simple(phb->bus, -1, "pci-ohci");
35139a59
DG
844 if (spapr->has_graphics) {
845 usbdevice_create("keyboard");
846 usbdevice_create("mouse");
847 }
848 }
849
7f763a5d 850 if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
4d8d5467
BH
851 fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
852 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
853 exit(1);
854 }
855
9fdf0c29
DG
856 if (kernel_filename) {
857 uint64_t lowaddr = 0;
858
9fdf0c29
DG
859 kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
860 NULL, &lowaddr, NULL, 1, ELF_MACHINE, 0);
861 if (kernel_size < 0) {
a3467baa
DG
862 kernel_size = load_image_targphys(kernel_filename,
863 KERNEL_LOAD_ADDR,
4d8d5467 864 load_limit - KERNEL_LOAD_ADDR);
9fdf0c29
DG
865 }
866 if (kernel_size < 0) {
867 fprintf(stderr, "qemu: could not load kernel '%s'\n",
868 kernel_filename);
869 exit(1);
870 }
871
872 /* load initrd */
873 if (initrd_filename) {
4d8d5467
BH
874 /* Try to locate the initrd in the gap between the kernel
875 * and the firmware. Add a bit of space just in case
876 */
877 initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
9fdf0c29 878 initrd_size = load_image_targphys(initrd_filename, initrd_base,
4d8d5467 879 load_limit - initrd_base);
9fdf0c29
DG
880 if (initrd_size < 0) {
881 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
882 initrd_filename);
883 exit(1);
884 }
885 } else {
886 initrd_base = 0;
887 initrd_size = 0;
888 }
4d8d5467 889 }
a3467baa 890
4d8d5467
BH
891 filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, FW_FILE_NAME);
892 fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
893 if (fw_size < 0) {
894 hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
895 exit(1);
896 }
897 g_free(filename);
4d8d5467
BH
898
899 spapr->entry_point = 0x100;
900
9fdf0c29 901 /* Prepare the device tree */
7f763a5d 902 spapr->fdt_skel = spapr_create_fdt_skel(cpu_model,
a3467baa 903 initrd_base, initrd_size,
4d8d5467 904 kernel_size,
7f763a5d 905 boot_device, kernel_cmdline);
a3467baa 906 assert(spapr->fdt_skel != NULL);
9fdf0c29
DG
907}
908
909static QEMUMachine spapr_machine = {
910 .name = "pseries",
911 .desc = "pSeries Logical Partition (PAPR compliant)",
912 .init = ppc_spapr_init,
c8787ad4 913 .reset = ppc_spapr_reset,
9fdf0c29 914 .max_cpus = MAX_CPUS,
9fdf0c29 915 .no_parallel = 1,
6e270446 916 .use_scsi = 1,
9fdf0c29
DG
917};
918
919static void spapr_machine_init(void)
920{
921 qemu_register_machine(&spapr_machine);
922}
923
924machine_init(spapr_machine_init);