]> Git Repo - qemu.git/blob - hw/spapr.c
migration: Make all posible migration functions static
[qemu.git] / hw / spapr.c
1 /*
2  * QEMU PowerPC pSeries Logical Partition (aka sPAPR) hardware System Emulator
3  *
4  * Copyright (c) 2004-2007 Fabrice Bellard
5  * Copyright (c) 2007 Jocelyn Mayer
6  * Copyright (c) 2010 David Gibson, IBM Corporation.
7  *
8  * Permission is hereby granted, free of charge, to any person obtaining a copy
9  * of this software and associated documentation files (the "Software"), to deal
10  * in the Software without restriction, including without limitation the rights
11  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
12  * copies of the Software, and to permit persons to whom the Software is
13  * furnished to do so, subject to the following conditions:
14  *
15  * The above copyright notice and this permission notice shall be included in
16  * all copies or substantial portions of the Software.
17  *
18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
21  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
22  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
23  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24  * THE SOFTWARE.
25  *
26  */
27 #include "sysemu.h"
28 #include "hw.h"
29 #include "elf.h"
30 #include "net.h"
31 #include "blockdev.h"
32
33 #include "hw/boards.h"
34 #include "hw/ppc.h"
35 #include "hw/loader.h"
36
37 #include "hw/spapr.h"
38 #include "hw/spapr_vio.h"
39 #include "hw/xics.h"
40
41 #include "kvm.h"
42 #include "kvm_ppc.h"
43
44 #include <libfdt.h>
45
46 #define KERNEL_LOAD_ADDR        0x00000000
47 #define INITRD_LOAD_ADDR        0x02800000
48 #define FDT_MAX_SIZE            0x10000
49 #define RTAS_MAX_SIZE           0x10000
50 #define FW_MAX_SIZE             0x400000
51 #define FW_FILE_NAME            "slof.bin"
52
53 #define MIN_RAM_SLOF            512UL
54
55 #define TIMEBASE_FREQ           512000000ULL
56
57 #define MAX_CPUS                256
58 #define XICS_IRQS               1024
59
60 #define PHANDLE_XICP            0x00001111
61
62 sPAPREnvironment *spapr;
63
64 qemu_irq spapr_allocate_irq(uint32_t hint, uint32_t *irq_num)
65 {
66     uint32_t irq;
67     qemu_irq qirq;
68
69     if (hint) {
70         irq = hint;
71         /* FIXME: we should probably check for collisions somehow */
72     } else {
73         irq = spapr->next_irq++;
74     }
75
76     qirq = xics_find_qirq(spapr->icp, irq);
77     if (!qirq) {
78         return NULL;
79     }
80
81     if (irq_num) {
82         *irq_num = irq;
83     }
84
85     return qirq;
86 }
87
88 static void *spapr_create_fdt_skel(const char *cpu_model,
89                                    target_phys_addr_t initrd_base,
90                                    target_phys_addr_t initrd_size,
91                                    const char *boot_device,
92                                    const char *kernel_cmdline,
93                                    long hash_shift)
94 {
95     void *fdt;
96     CPUState *env;
97     uint64_t mem_reg_property[] = { 0, cpu_to_be64(ram_size) };
98     uint32_t start_prop = cpu_to_be32(initrd_base);
99     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
100     uint32_t pft_size_prop[] = {0, cpu_to_be32(hash_shift)};
101     char hypertas_prop[] = "hcall-pft\0hcall-term\0hcall-dabr\0hcall-interrupt"
102         "\0hcall-tce\0hcall-vio\0hcall-splpar\0hcall-bulk";
103     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(smp_cpus)};
104     int i;
105     char *modelname;
106
107 #define _FDT(exp) \
108     do { \
109         int ret = (exp);                                           \
110         if (ret < 0) {                                             \
111             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
112                     #exp, fdt_strerror(ret));                      \
113             exit(1);                                               \
114         }                                                          \
115     } while (0)
116
117     fdt = g_malloc0(FDT_MAX_SIZE);
118     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
119
120     _FDT((fdt_finish_reservemap(fdt)));
121
122     /* Root node */
123     _FDT((fdt_begin_node(fdt, "")));
124     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
125     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
126
127     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
128     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
129
130     /* /chosen */
131     _FDT((fdt_begin_node(fdt, "chosen")));
132
133     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
134     _FDT((fdt_property(fdt, "linux,initrd-start",
135                        &start_prop, sizeof(start_prop))));
136     _FDT((fdt_property(fdt, "linux,initrd-end",
137                        &end_prop, sizeof(end_prop))));
138     _FDT((fdt_property_string(fdt, "qemu,boot-device", boot_device)));
139
140     _FDT((fdt_end_node(fdt)));
141
142     /* memory node */
143     _FDT((fdt_begin_node(fdt, "memory@0")));
144
145     _FDT((fdt_property_string(fdt, "device_type", "memory")));
146     _FDT((fdt_property(fdt, "reg",
147                        mem_reg_property, sizeof(mem_reg_property))));
148
149     _FDT((fdt_end_node(fdt)));
150
151     /* cpus */
152     _FDT((fdt_begin_node(fdt, "cpus")));
153
154     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
155     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
156
157     modelname = g_strdup(cpu_model);
158
159     for (i = 0; i < strlen(modelname); i++) {
160         modelname[i] = toupper(modelname[i]);
161     }
162
163     for (env = first_cpu; env != NULL; env = env->next_cpu) {
164         int index = env->cpu_index;
165         uint32_t gserver_prop[] = {cpu_to_be32(index), 0}; /* HACK! */
166         char *nodename;
167         uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
168                            0xffffffff, 0xffffffff};
169         uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
170         uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
171
172         if (asprintf(&nodename, "%s@%x", modelname, index) < 0) {
173             fprintf(stderr, "Allocation failure\n");
174             exit(1);
175         }
176
177         _FDT((fdt_begin_node(fdt, nodename)));
178
179         free(nodename);
180
181         _FDT((fdt_property_cell(fdt, "reg", index)));
182         _FDT((fdt_property_string(fdt, "device_type", "cpu")));
183
184         _FDT((fdt_property_cell(fdt, "cpu-version", env->spr[SPR_PVR])));
185         _FDT((fdt_property_cell(fdt, "dcache-block-size",
186                                 env->dcache_line_size)));
187         _FDT((fdt_property_cell(fdt, "icache-block-size",
188                                 env->icache_line_size)));
189         _FDT((fdt_property_cell(fdt, "timebase-frequency", tbfreq)));
190         _FDT((fdt_property_cell(fdt, "clock-frequency", cpufreq)));
191         _FDT((fdt_property_cell(fdt, "ibm,slb-size", env->slb_nr)));
192         _FDT((fdt_property(fdt, "ibm,pft-size",
193                            pft_size_prop, sizeof(pft_size_prop))));
194         _FDT((fdt_property_string(fdt, "status", "okay")));
195         _FDT((fdt_property(fdt, "64-bit", NULL, 0)));
196         _FDT((fdt_property_cell(fdt, "ibm,ppc-interrupt-server#s", index)));
197         _FDT((fdt_property(fdt, "ibm,ppc-interrupt-gserver#s",
198                            gserver_prop, sizeof(gserver_prop))));
199
200         if (env->mmu_model & POWERPC_MMU_1TSEG) {
201             _FDT((fdt_property(fdt, "ibm,processor-segment-sizes",
202                                segs, sizeof(segs))));
203         }
204
205         _FDT((fdt_end_node(fdt)));
206     }
207
208     g_free(modelname);
209
210     _FDT((fdt_end_node(fdt)));
211
212     /* RTAS */
213     _FDT((fdt_begin_node(fdt, "rtas")));
214
215     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas_prop,
216                        sizeof(hypertas_prop))));
217
218     _FDT((fdt_end_node(fdt)));
219
220     /* interrupt controller */
221     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
222
223     _FDT((fdt_property_string(fdt, "device_type",
224                               "PowerPC-External-Interrupt-Presentation")));
225     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
226     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
227     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
228                        interrupt_server_ranges_prop,
229                        sizeof(interrupt_server_ranges_prop))));
230     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
231     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
232     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
233
234     _FDT((fdt_end_node(fdt)));
235
236     /* vdevice */
237     _FDT((fdt_begin_node(fdt, "vdevice")));
238
239     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
240     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
241     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
242     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
243     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
244     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
245
246     _FDT((fdt_end_node(fdt)));
247
248     _FDT((fdt_end_node(fdt))); /* close root node */
249     _FDT((fdt_finish(fdt)));
250
251     return fdt;
252 }
253
254 static void spapr_finalize_fdt(sPAPREnvironment *spapr,
255                                target_phys_addr_t fdt_addr,
256                                target_phys_addr_t rtas_addr,
257                                target_phys_addr_t rtas_size)
258 {
259     int ret;
260     void *fdt;
261
262     fdt = g_malloc(FDT_MAX_SIZE);
263
264     /* open out the base tree into a temp buffer for the final tweaks */
265     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
266
267     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
268     if (ret < 0) {
269         fprintf(stderr, "couldn't setup vio devices in fdt\n");
270         exit(1);
271     }
272
273     /* RTAS */
274     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
275     if (ret < 0) {
276         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
277     }
278
279     _FDT((fdt_pack(fdt)));
280
281     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
282
283     g_free(fdt);
284 }
285
286 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
287 {
288     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
289 }
290
291 static void emulate_spapr_hypercall(CPUState *env)
292 {
293     env->gpr[3] = spapr_hypercall(env, env->gpr[3], &env->gpr[4]);
294 }
295
296 static void spapr_reset(void *opaque)
297 {
298     sPAPREnvironment *spapr = (sPAPREnvironment *)opaque;
299
300     fprintf(stderr, "sPAPR reset\n");
301
302     /* flush out the hash table */
303     memset(spapr->htab, 0, spapr->htab_size);
304
305     /* Load the fdt */
306     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
307                        spapr->rtas_size);
308
309     /* Set up the entry state */
310     first_cpu->gpr[3] = spapr->fdt_addr;
311     first_cpu->gpr[5] = 0;
312     first_cpu->halted = 0;
313     first_cpu->nip = spapr->entry_point;
314
315 }
316
317 /* pSeries LPAR / sPAPR hardware init */
318 static void ppc_spapr_init(ram_addr_t ram_size,
319                            const char *boot_device,
320                            const char *kernel_filename,
321                            const char *kernel_cmdline,
322                            const char *initrd_filename,
323                            const char *cpu_model)
324 {
325     CPUState *env;
326     int i;
327     ram_addr_t ram_offset;
328     uint32_t initrd_base;
329     long kernel_size, initrd_size, fw_size;
330     long pteg_shift = 17;
331     char *filename;
332
333     spapr = g_malloc(sizeof(*spapr));
334     cpu_ppc_hypercall = emulate_spapr_hypercall;
335
336     /* We place the device tree just below either the top of RAM, or
337      * 2GB, so that it can be processed with 32-bit code if
338      * necessary */
339     spapr->fdt_addr = MIN(ram_size, 0x80000000) - FDT_MAX_SIZE;
340     spapr->rtas_addr = spapr->fdt_addr - RTAS_MAX_SIZE;
341
342     /* init CPUs */
343     if (cpu_model == NULL) {
344         cpu_model = "POWER7";
345     }
346     for (i = 0; i < smp_cpus; i++) {
347         env = cpu_init(cpu_model);
348
349         if (!env) {
350             fprintf(stderr, "Unable to find PowerPC CPU definition\n");
351             exit(1);
352         }
353         /* Set time-base frequency to 512 MHz */
354         cpu_ppc_tb_init(env, TIMEBASE_FREQ);
355         qemu_register_reset((QEMUResetHandler *)&cpu_reset, env);
356
357         env->hreset_vector = 0x60;
358         env->hreset_excp_prefix = 0;
359         env->gpr[3] = env->cpu_index;
360     }
361
362     /* allocate RAM */
363     spapr->ram_limit = ram_size;
364     ram_offset = qemu_ram_alloc(NULL, "ppc_spapr.ram", spapr->ram_limit);
365     cpu_register_physical_memory(0, ram_size, ram_offset);
366
367     /* allocate hash page table.  For now we always make this 16mb,
368      * later we should probably make it scale to the size of guest
369      * RAM */
370     spapr->htab_size = 1ULL << (pteg_shift + 7);
371     spapr->htab = qemu_memalign(spapr->htab_size, spapr->htab_size);
372
373     for (env = first_cpu; env != NULL; env = env->next_cpu) {
374         env->external_htab = spapr->htab;
375         env->htab_base = -1;
376         env->htab_mask = spapr->htab_size - 1;
377
378         /* Tell KVM that we're in PAPR mode */
379         env->spr[SPR_SDR1] = (unsigned long)spapr->htab |
380                              ((pteg_shift + 7) - 18);
381         env->spr[SPR_HIOR] = 0;
382
383         if (kvm_enabled()) {
384             kvmppc_set_papr(env);
385         }
386     }
387
388     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
389     spapr->rtas_size = load_image_targphys(filename, spapr->rtas_addr,
390                                            ram_size - spapr->rtas_addr);
391     if (spapr->rtas_size < 0) {
392         hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
393         exit(1);
394     }
395     g_free(filename);
396
397     /* Set up Interrupt Controller */
398     spapr->icp = xics_system_init(XICS_IRQS);
399     spapr->next_irq = 16;
400
401     /* Set up VIO bus */
402     spapr->vio_bus = spapr_vio_bus_init();
403
404     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
405         if (serial_hds[i]) {
406             spapr_vty_create(spapr->vio_bus, SPAPR_VTY_BASE_ADDRESS + i,
407                              serial_hds[i]);
408         }
409     }
410
411     for (i = 0; i < nb_nics; i++) {
412         NICInfo *nd = &nd_table[i];
413
414         if (!nd->model) {
415             nd->model = g_strdup("ibmveth");
416         }
417
418         if (strcmp(nd->model, "ibmveth") == 0) {
419             spapr_vlan_create(spapr->vio_bus, 0x1000 + i, nd);
420         } else {
421             fprintf(stderr, "pSeries (sPAPR) platform does not support "
422                     "NIC model '%s' (only ibmveth is supported)\n",
423                     nd->model);
424             exit(1);
425         }
426     }
427
428     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
429         spapr_vscsi_create(spapr->vio_bus, 0x2000 + i);
430     }
431
432     if (kernel_filename) {
433         uint64_t lowaddr = 0;
434
435         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
436                                NULL, &lowaddr, NULL, 1, ELF_MACHINE, 0);
437         if (kernel_size < 0) {
438             kernel_size = load_image_targphys(kernel_filename,
439                                               KERNEL_LOAD_ADDR,
440                                               ram_size - KERNEL_LOAD_ADDR);
441         }
442         if (kernel_size < 0) {
443             fprintf(stderr, "qemu: could not load kernel '%s'\n",
444                     kernel_filename);
445             exit(1);
446         }
447
448         /* load initrd */
449         if (initrd_filename) {
450             initrd_base = INITRD_LOAD_ADDR;
451             initrd_size = load_image_targphys(initrd_filename, initrd_base,
452                                               ram_size - initrd_base);
453             if (initrd_size < 0) {
454                 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
455                         initrd_filename);
456                 exit(1);
457             }
458         } else {
459             initrd_base = 0;
460             initrd_size = 0;
461         }
462
463         spapr->entry_point = KERNEL_LOAD_ADDR;
464     } else {
465         if (ram_size < (MIN_RAM_SLOF << 20)) {
466             fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
467                     "%ldM guest RAM\n", MIN_RAM_SLOF);
468             exit(1);
469         }
470         filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, FW_FILE_NAME);
471         fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
472         if (fw_size < 0) {
473             hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
474             exit(1);
475         }
476         g_free(filename);
477         spapr->entry_point = 0x100;
478         initrd_base = 0;
479         initrd_size = 0;
480
481         /* SLOF will startup the secondary CPUs using RTAS,
482            rather than expecting a kexec() style entry */
483         for (env = first_cpu; env != NULL; env = env->next_cpu) {
484             env->halted = 1;
485         }
486     }
487
488     /* Prepare the device tree */
489     spapr->fdt_skel = spapr_create_fdt_skel(cpu_model,
490                                             initrd_base, initrd_size,
491                                             boot_device, kernel_cmdline,
492                                             pteg_shift + 7);
493     assert(spapr->fdt_skel != NULL);
494
495     qemu_register_reset(spapr_reset, spapr);
496 }
497
498 static QEMUMachine spapr_machine = {
499     .name = "pseries",
500     .desc = "pSeries Logical Partition (PAPR compliant)",
501     .init = ppc_spapr_init,
502     .max_cpus = MAX_CPUS,
503     .no_vga = 1,
504     .no_parallel = 1,
505     .use_scsi = 1,
506 };
507
508 static void spapr_machine_init(void)
509 {
510     qemu_register_machine(&spapr_machine);
511 }
512
513 machine_init(spapr_machine_init);
This page took 0.072692 seconds and 4 git commands to generate.