xref: /openbmc/qemu/hw/ppc/spapr.c (revision 1fd6bb44)
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/sysemu.h"
28 #include "hw/hw.h"
29 #include "elf.h"
30 #include "net/net.h"
31 #include "sysemu/blockdev.h"
32 #include "sysemu/cpus.h"
33 #include "sysemu/kvm.h"
34 #include "kvm_ppc.h"
35 
36 #include "hw/boards.h"
37 #include "hw/ppc/ppc.h"
38 #include "hw/loader.h"
39 
40 #include "hw/ppc/spapr.h"
41 #include "hw/ppc/spapr_vio.h"
42 #include "hw/pci-host/spapr.h"
43 #include "hw/ppc/xics.h"
44 #include "hw/pci/msi.h"
45 
46 #include "sysemu/kvm.h"
47 #include "kvm_ppc.h"
48 #include "hw/pci/pci.h"
49 
50 #include "exec/address-spaces.h"
51 #include "hw/usb.h"
52 #include "qemu/config-file.h"
53 
54 #include <libfdt.h>
55 
56 /* SLOF memory layout:
57  *
58  * SLOF raw image loaded at 0, copies its romfs right below the flat
59  * device-tree, then position SLOF itself 31M below that
60  *
61  * So we set FW_OVERHEAD to 40MB which should account for all of that
62  * and more
63  *
64  * We load our kernel at 4M, leaving space for SLOF initial image
65  */
66 #define FDT_MAX_SIZE            0x10000
67 #define RTAS_MAX_SIZE           0x10000
68 #define FW_MAX_SIZE             0x400000
69 #define FW_FILE_NAME            "slof.bin"
70 #define FW_OVERHEAD             0x2800000
71 #define KERNEL_LOAD_ADDR        FW_MAX_SIZE
72 
73 #define MIN_RMA_SLOF            128UL
74 
75 #define TIMEBASE_FREQ           512000000ULL
76 
77 #define MAX_CPUS                256
78 #define XICS_IRQS               1024
79 
80 #define PHANDLE_XICP            0x00001111
81 
82 #define HTAB_SIZE(spapr)        (1ULL << ((spapr)->htab_shift))
83 
84 sPAPREnvironment *spapr;
85 
86 int spapr_allocate_irq(int hint, bool lsi)
87 {
88     int irq;
89 
90     if (hint) {
91         irq = hint;
92         /* FIXME: we should probably check for collisions somehow */
93     } else {
94         irq = spapr->next_irq++;
95     }
96 
97     /* Configure irq type */
98     if (!xics_get_qirq(spapr->icp, irq)) {
99         return 0;
100     }
101 
102     xics_set_irq_type(spapr->icp, irq, lsi);
103 
104     return irq;
105 }
106 
107 /* Allocate block of consequtive IRQs, returns a number of the first */
108 int spapr_allocate_irq_block(int num, bool lsi)
109 {
110     int first = -1;
111     int i;
112 
113     for (i = 0; i < num; ++i) {
114         int irq;
115 
116         irq = spapr_allocate_irq(0, lsi);
117         if (!irq) {
118             return -1;
119         }
120 
121         if (0 == i) {
122             first = irq;
123         }
124 
125         /* If the above doesn't create a consecutive block then that's
126          * an internal bug */
127         assert(irq == (first + i));
128     }
129 
130     return first;
131 }
132 
133 static int spapr_fixup_cpu_dt(void *fdt, sPAPREnvironment *spapr)
134 {
135     int ret = 0, offset;
136     CPUPPCState *env;
137     CPUState *cpu;
138     char cpu_model[32];
139     int smt = kvmppc_smt_threads();
140     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
141 
142     assert(spapr->cpu_model);
143 
144     for (env = first_cpu; env != NULL; env = env->next_cpu) {
145         cpu = CPU(ppc_env_get_cpu(env));
146         uint32_t associativity[] = {cpu_to_be32(0x5),
147                                     cpu_to_be32(0x0),
148                                     cpu_to_be32(0x0),
149                                     cpu_to_be32(0x0),
150                                     cpu_to_be32(cpu->numa_node),
151                                     cpu_to_be32(cpu->cpu_index)};
152 
153         if ((cpu->cpu_index % smt) != 0) {
154             continue;
155         }
156 
157         snprintf(cpu_model, 32, "/cpus/%s@%x", spapr->cpu_model,
158                  cpu->cpu_index);
159 
160         offset = fdt_path_offset(fdt, cpu_model);
161         if (offset < 0) {
162             return offset;
163         }
164 
165         if (nb_numa_nodes > 1) {
166             ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
167                               sizeof(associativity));
168             if (ret < 0) {
169                 return ret;
170             }
171         }
172 
173         ret = fdt_setprop(fdt, offset, "ibm,pft-size",
174                           pft_size_prop, sizeof(pft_size_prop));
175         if (ret < 0) {
176             return ret;
177         }
178     }
179     return ret;
180 }
181 
182 
183 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
184                                      size_t maxsize)
185 {
186     size_t maxcells = maxsize / sizeof(uint32_t);
187     int i, j, count;
188     uint32_t *p = prop;
189 
190     for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
191         struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
192 
193         if (!sps->page_shift) {
194             break;
195         }
196         for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
197             if (sps->enc[count].page_shift == 0) {
198                 break;
199             }
200         }
201         if ((p - prop) >= (maxcells - 3 - count * 2)) {
202             break;
203         }
204         *(p++) = cpu_to_be32(sps->page_shift);
205         *(p++) = cpu_to_be32(sps->slb_enc);
206         *(p++) = cpu_to_be32(count);
207         for (j = 0; j < count; j++) {
208             *(p++) = cpu_to_be32(sps->enc[j].page_shift);
209             *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
210         }
211     }
212 
213     return (p - prop) * sizeof(uint32_t);
214 }
215 
216 #define _FDT(exp) \
217     do { \
218         int ret = (exp);                                           \
219         if (ret < 0) {                                             \
220             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
221                     #exp, fdt_strerror(ret));                      \
222             exit(1);                                               \
223         }                                                          \
224     } while (0)
225 
226 
227 static void *spapr_create_fdt_skel(const char *cpu_model,
228                                    hwaddr initrd_base,
229                                    hwaddr initrd_size,
230                                    hwaddr kernel_size,
231                                    const char *boot_device,
232                                    const char *kernel_cmdline,
233                                    uint32_t epow_irq)
234 {
235     void *fdt;
236     CPUPPCState *env;
237     uint32_t start_prop = cpu_to_be32(initrd_base);
238     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
239     char hypertas_prop[] = "hcall-pft\0hcall-term\0hcall-dabr\0hcall-interrupt"
240         "\0hcall-tce\0hcall-vio\0hcall-splpar\0hcall-bulk";
241     char qemu_hypertas_prop[] = "hcall-memop1";
242     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
243     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(smp_cpus)};
244     char *modelname;
245     int i, smt = kvmppc_smt_threads();
246     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
247 
248     fdt = g_malloc0(FDT_MAX_SIZE);
249     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
250 
251     if (kernel_size) {
252         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
253     }
254     if (initrd_size) {
255         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
256     }
257     _FDT((fdt_finish_reservemap(fdt)));
258 
259     /* Root node */
260     _FDT((fdt_begin_node(fdt, "")));
261     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
262     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
263     _FDT((fdt_property_string(fdt, "compatible", "qemu,pseries")));
264 
265     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
266     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
267 
268     /* /chosen */
269     _FDT((fdt_begin_node(fdt, "chosen")));
270 
271     /* Set Form1_affinity */
272     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
273 
274     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
275     _FDT((fdt_property(fdt, "linux,initrd-start",
276                        &start_prop, sizeof(start_prop))));
277     _FDT((fdt_property(fdt, "linux,initrd-end",
278                        &end_prop, sizeof(end_prop))));
279     if (kernel_size) {
280         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
281                               cpu_to_be64(kernel_size) };
282 
283         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
284     }
285     if (boot_device) {
286         _FDT((fdt_property_string(fdt, "qemu,boot-device", boot_device)));
287     }
288     _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
289     _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
290     _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
291 
292     _FDT((fdt_end_node(fdt)));
293 
294     /* cpus */
295     _FDT((fdt_begin_node(fdt, "cpus")));
296 
297     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
298     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
299 
300     modelname = g_strdup(cpu_model);
301 
302     for (i = 0; i < strlen(modelname); i++) {
303         modelname[i] = toupper(modelname[i]);
304     }
305 
306     /* This is needed during FDT finalization */
307     spapr->cpu_model = g_strdup(modelname);
308 
309     for (env = first_cpu; env != NULL; env = env->next_cpu) {
310         CPUState *cpu = CPU(ppc_env_get_cpu(env));
311         int index = cpu->cpu_index;
312         uint32_t servers_prop[smp_threads];
313         uint32_t gservers_prop[smp_threads * 2];
314         char *nodename;
315         uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
316                            0xffffffff, 0xffffffff};
317         uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
318         uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
319         uint32_t page_sizes_prop[64];
320         size_t page_sizes_prop_size;
321 
322         if ((index % smt) != 0) {
323             continue;
324         }
325 
326         nodename = g_strdup_printf("%s@%x", modelname, index);
327 
328         _FDT((fdt_begin_node(fdt, nodename)));
329 
330         g_free(nodename);
331 
332         _FDT((fdt_property_cell(fdt, "reg", index)));
333         _FDT((fdt_property_string(fdt, "device_type", "cpu")));
334 
335         _FDT((fdt_property_cell(fdt, "cpu-version", env->spr[SPR_PVR])));
336         _FDT((fdt_property_cell(fdt, "dcache-block-size",
337                                 env->dcache_line_size)));
338         _FDT((fdt_property_cell(fdt, "icache-block-size",
339                                 env->icache_line_size)));
340         _FDT((fdt_property_cell(fdt, "timebase-frequency", tbfreq)));
341         _FDT((fdt_property_cell(fdt, "clock-frequency", cpufreq)));
342         _FDT((fdt_property_cell(fdt, "ibm,slb-size", env->slb_nr)));
343         _FDT((fdt_property_string(fdt, "status", "okay")));
344         _FDT((fdt_property(fdt, "64-bit", NULL, 0)));
345 
346         /* Build interrupt servers and gservers properties */
347         for (i = 0; i < smp_threads; i++) {
348             servers_prop[i] = cpu_to_be32(index + i);
349             /* Hack, direct the group queues back to cpu 0 */
350             gservers_prop[i*2] = cpu_to_be32(index + i);
351             gservers_prop[i*2 + 1] = 0;
352         }
353         _FDT((fdt_property(fdt, "ibm,ppc-interrupt-server#s",
354                            servers_prop, sizeof(servers_prop))));
355         _FDT((fdt_property(fdt, "ibm,ppc-interrupt-gserver#s",
356                            gservers_prop, sizeof(gservers_prop))));
357 
358         if (env->mmu_model & POWERPC_MMU_1TSEG) {
359             _FDT((fdt_property(fdt, "ibm,processor-segment-sizes",
360                                segs, sizeof(segs))));
361         }
362 
363         /* Advertise VMX/VSX (vector extensions) if available
364          *   0 / no property == no vector extensions
365          *   1               == VMX / Altivec available
366          *   2               == VSX available */
367         if (env->insns_flags & PPC_ALTIVEC) {
368             uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
369 
370             _FDT((fdt_property_cell(fdt, "ibm,vmx", vmx)));
371         }
372 
373         /* Advertise DFP (Decimal Floating Point) if available
374          *   0 / no property == no DFP
375          *   1               == DFP available */
376         if (env->insns_flags2 & PPC2_DFP) {
377             _FDT((fdt_property_cell(fdt, "ibm,dfp", 1)));
378         }
379 
380         page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
381                                                       sizeof(page_sizes_prop));
382         if (page_sizes_prop_size) {
383             _FDT((fdt_property(fdt, "ibm,segment-page-sizes",
384                                page_sizes_prop, page_sizes_prop_size)));
385         }
386 
387         _FDT((fdt_end_node(fdt)));
388     }
389 
390     g_free(modelname);
391 
392     _FDT((fdt_end_node(fdt)));
393 
394     /* RTAS */
395     _FDT((fdt_begin_node(fdt, "rtas")));
396 
397     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas_prop,
398                        sizeof(hypertas_prop))));
399     _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas_prop,
400                        sizeof(qemu_hypertas_prop))));
401 
402     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
403         refpoints, sizeof(refpoints))));
404 
405     _FDT((fdt_property_cell(fdt, "rtas-error-log-max", RTAS_ERROR_LOG_MAX)));
406 
407     _FDT((fdt_end_node(fdt)));
408 
409     /* interrupt controller */
410     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
411 
412     _FDT((fdt_property_string(fdt, "device_type",
413                               "PowerPC-External-Interrupt-Presentation")));
414     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
415     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
416     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
417                        interrupt_server_ranges_prop,
418                        sizeof(interrupt_server_ranges_prop))));
419     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
420     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
421     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
422 
423     _FDT((fdt_end_node(fdt)));
424 
425     /* vdevice */
426     _FDT((fdt_begin_node(fdt, "vdevice")));
427 
428     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
429     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
430     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
431     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
432     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
433     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
434 
435     _FDT((fdt_end_node(fdt)));
436 
437     /* event-sources */
438     spapr_events_fdt_skel(fdt, epow_irq);
439 
440     _FDT((fdt_end_node(fdt))); /* close root node */
441     _FDT((fdt_finish(fdt)));
442 
443     return fdt;
444 }
445 
446 static int spapr_populate_memory(sPAPREnvironment *spapr, void *fdt)
447 {
448     uint32_t associativity[] = {cpu_to_be32(0x4), cpu_to_be32(0x0),
449                                 cpu_to_be32(0x0), cpu_to_be32(0x0),
450                                 cpu_to_be32(0x0)};
451     char mem_name[32];
452     hwaddr node0_size, mem_start;
453     uint64_t mem_reg_property[2];
454     int i, off;
455 
456     /* memory node(s) */
457     node0_size = (nb_numa_nodes > 1) ? node_mem[0] : ram_size;
458     if (spapr->rma_size > node0_size) {
459         spapr->rma_size = node0_size;
460     }
461 
462     /* RMA */
463     mem_reg_property[0] = 0;
464     mem_reg_property[1] = cpu_to_be64(spapr->rma_size);
465     off = fdt_add_subnode(fdt, 0, "memory@0");
466     _FDT(off);
467     _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
468     _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
469                       sizeof(mem_reg_property))));
470     _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
471                       sizeof(associativity))));
472 
473     /* RAM: Node 0 */
474     if (node0_size > spapr->rma_size) {
475         mem_reg_property[0] = cpu_to_be64(spapr->rma_size);
476         mem_reg_property[1] = cpu_to_be64(node0_size - spapr->rma_size);
477 
478         sprintf(mem_name, "memory@" TARGET_FMT_lx, spapr->rma_size);
479         off = fdt_add_subnode(fdt, 0, mem_name);
480         _FDT(off);
481         _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
482         _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
483                           sizeof(mem_reg_property))));
484         _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
485                           sizeof(associativity))));
486     }
487 
488     /* RAM: Node 1 and beyond */
489     mem_start = node0_size;
490     for (i = 1; i < nb_numa_nodes; i++) {
491         mem_reg_property[0] = cpu_to_be64(mem_start);
492         mem_reg_property[1] = cpu_to_be64(node_mem[i]);
493         associativity[3] = associativity[4] = cpu_to_be32(i);
494         sprintf(mem_name, "memory@" TARGET_FMT_lx, mem_start);
495         off = fdt_add_subnode(fdt, 0, mem_name);
496         _FDT(off);
497         _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
498         _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
499                           sizeof(mem_reg_property))));
500         _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
501                           sizeof(associativity))));
502         mem_start += node_mem[i];
503     }
504 
505     return 0;
506 }
507 
508 static void spapr_finalize_fdt(sPAPREnvironment *spapr,
509                                hwaddr fdt_addr,
510                                hwaddr rtas_addr,
511                                hwaddr rtas_size)
512 {
513     int ret;
514     void *fdt;
515     sPAPRPHBState *phb;
516 
517     fdt = g_malloc(FDT_MAX_SIZE);
518 
519     /* open out the base tree into a temp buffer for the final tweaks */
520     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
521 
522     ret = spapr_populate_memory(spapr, fdt);
523     if (ret < 0) {
524         fprintf(stderr, "couldn't setup memory nodes in fdt\n");
525         exit(1);
526     }
527 
528     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
529     if (ret < 0) {
530         fprintf(stderr, "couldn't setup vio devices in fdt\n");
531         exit(1);
532     }
533 
534     QLIST_FOREACH(phb, &spapr->phbs, list) {
535         ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
536     }
537 
538     if (ret < 0) {
539         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
540         exit(1);
541     }
542 
543     /* RTAS */
544     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
545     if (ret < 0) {
546         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
547     }
548 
549     /* Advertise NUMA via ibm,associativity */
550     ret = spapr_fixup_cpu_dt(fdt, spapr);
551     if (ret < 0) {
552         fprintf(stderr, "Couldn't finalize CPU device tree properties\n");
553     }
554 
555     if (!spapr->has_graphics) {
556         spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
557     }
558 
559     _FDT((fdt_pack(fdt)));
560 
561     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
562         hw_error("FDT too big ! 0x%x bytes (max is 0x%x)\n",
563                  fdt_totalsize(fdt), FDT_MAX_SIZE);
564         exit(1);
565     }
566 
567     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
568 
569     g_free(fdt);
570 }
571 
572 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
573 {
574     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
575 }
576 
577 static void emulate_spapr_hypercall(PowerPCCPU *cpu)
578 {
579     CPUPPCState *env = &cpu->env;
580 
581     if (msr_pr) {
582         hcall_dprintf("Hypercall made with MSR[PR]=1\n");
583         env->gpr[3] = H_PRIVILEGE;
584     } else {
585         env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]);
586     }
587 }
588 
589 static void spapr_reset_htab(sPAPREnvironment *spapr)
590 {
591     long shift;
592 
593     /* allocate hash page table.  For now we always make this 16mb,
594      * later we should probably make it scale to the size of guest
595      * RAM */
596 
597     shift = kvmppc_reset_htab(spapr->htab_shift);
598 
599     if (shift > 0) {
600         /* Kernel handles htab, we don't need to allocate one */
601         spapr->htab_shift = shift;
602     } else {
603         if (!spapr->htab) {
604             /* Allocate an htab if we don't yet have one */
605             spapr->htab = qemu_memalign(HTAB_SIZE(spapr), HTAB_SIZE(spapr));
606         }
607 
608         /* And clear it */
609         memset(spapr->htab, 0, HTAB_SIZE(spapr));
610     }
611 
612     /* Update the RMA size if necessary */
613     if (spapr->vrma_adjust) {
614         spapr->rma_size = kvmppc_rma_size(ram_size, spapr->htab_shift);
615     }
616 }
617 
618 static void ppc_spapr_reset(void)
619 {
620     CPUState *first_cpu_cpu;
621 
622     /* Reset the hash table & recalc the RMA */
623     spapr_reset_htab(spapr);
624 
625     qemu_devices_reset();
626 
627     /* Load the fdt */
628     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
629                        spapr->rtas_size);
630 
631     /* Set up the entry state */
632     first_cpu_cpu = ENV_GET_CPU(first_cpu);
633     first_cpu->gpr[3] = spapr->fdt_addr;
634     first_cpu->gpr[5] = 0;
635     first_cpu_cpu->halted = 0;
636     first_cpu->nip = spapr->entry_point;
637 
638 }
639 
640 static void spapr_cpu_reset(void *opaque)
641 {
642     PowerPCCPU *cpu = opaque;
643     CPUState *cs = CPU(cpu);
644     CPUPPCState *env = &cpu->env;
645 
646     cpu_reset(cs);
647 
648     /* All CPUs start halted.  CPU0 is unhalted from the machine level
649      * reset code and the rest are explicitly started up by the guest
650      * using an RTAS call */
651     cs->halted = 1;
652 
653     env->spr[SPR_HIOR] = 0;
654 
655     env->external_htab = spapr->htab;
656     env->htab_base = -1;
657     env->htab_mask = HTAB_SIZE(spapr) - 1;
658     env->spr[SPR_SDR1] = (unsigned long)spapr->htab |
659         (spapr->htab_shift - 18);
660 }
661 
662 static void spapr_create_nvram(sPAPREnvironment *spapr)
663 {
664     QemuOpts *machine_opts;
665     DeviceState *dev;
666 
667     dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram");
668 
669     machine_opts = qemu_opts_find(qemu_find_opts("machine"), 0);
670     if (machine_opts) {
671         const char *drivename;
672 
673         drivename = qemu_opt_get(machine_opts, "nvram");
674         if (drivename) {
675             BlockDriverState *bs;
676 
677             bs = bdrv_find(drivename);
678             if (!bs) {
679                 fprintf(stderr, "No such block device \"%s\" for nvram\n",
680                         drivename);
681                 exit(1);
682             }
683             qdev_prop_set_drive_nofail(dev, "drive", bs);
684         }
685     }
686 
687     qdev_init_nofail(dev);
688 
689     spapr->nvram = (struct sPAPRNVRAM *)dev;
690 }
691 
692 /* Returns whether we want to use VGA or not */
693 static int spapr_vga_init(PCIBus *pci_bus)
694 {
695     switch (vga_interface_type) {
696     case VGA_NONE:
697     case VGA_STD:
698         return pci_vga_init(pci_bus) != NULL;
699     default:
700         fprintf(stderr, "This vga model is not supported,"
701                 "currently it only supports -vga std\n");
702         exit(0);
703         break;
704     }
705 }
706 
707 /* pSeries LPAR / sPAPR hardware init */
708 static void ppc_spapr_init(QEMUMachineInitArgs *args)
709 {
710     ram_addr_t ram_size = args->ram_size;
711     const char *cpu_model = args->cpu_model;
712     const char *kernel_filename = args->kernel_filename;
713     const char *kernel_cmdline = args->kernel_cmdline;
714     const char *initrd_filename = args->initrd_filename;
715     const char *boot_device = args->boot_device;
716     PowerPCCPU *cpu;
717     CPUPPCState *env;
718     PCIHostState *phb;
719     int i;
720     MemoryRegion *sysmem = get_system_memory();
721     MemoryRegion *ram = g_new(MemoryRegion, 1);
722     hwaddr rma_alloc_size;
723     uint32_t initrd_base = 0;
724     long kernel_size = 0, initrd_size = 0;
725     long load_limit, rtas_limit, fw_size;
726     char *filename;
727 
728     msi_supported = true;
729 
730     spapr = g_malloc0(sizeof(*spapr));
731     QLIST_INIT(&spapr->phbs);
732 
733     cpu_ppc_hypercall = emulate_spapr_hypercall;
734 
735     /* Allocate RMA if necessary */
736     rma_alloc_size = kvmppc_alloc_rma("ppc_spapr.rma", sysmem);
737 
738     if (rma_alloc_size == -1) {
739         hw_error("qemu: Unable to create RMA\n");
740         exit(1);
741     }
742 
743     if (rma_alloc_size && (rma_alloc_size < ram_size)) {
744         spapr->rma_size = rma_alloc_size;
745     } else {
746         spapr->rma_size = ram_size;
747 
748         /* With KVM, we don't actually know whether KVM supports an
749          * unbounded RMA (PR KVM) or is limited by the hash table size
750          * (HV KVM using VRMA), so we always assume the latter
751          *
752          * In that case, we also limit the initial allocations for RTAS
753          * etc... to 256M since we have no way to know what the VRMA size
754          * is going to be as it depends on the size of the hash table
755          * isn't determined yet.
756          */
757         if (kvm_enabled()) {
758             spapr->vrma_adjust = 1;
759             spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
760         }
761     }
762 
763     /* We place the device tree and RTAS just below either the top of the RMA,
764      * or just below 2GB, whichever is lowere, so that it can be
765      * processed with 32-bit real mode code if necessary */
766     rtas_limit = MIN(spapr->rma_size, 0x80000000);
767     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
768     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
769     load_limit = spapr->fdt_addr - FW_OVERHEAD;
770 
771     /* We aim for a hash table of size 1/128 the size of RAM.  The
772      * normal rule of thumb is 1/64 the size of RAM, but that's much
773      * more than needed for the Linux guests we support. */
774     spapr->htab_shift = 18; /* Minimum architected size */
775     while (spapr->htab_shift <= 46) {
776         if ((1ULL << (spapr->htab_shift + 7)) >= ram_size) {
777             break;
778         }
779         spapr->htab_shift++;
780     }
781 
782     /* Set up Interrupt Controller before we create the VCPUs */
783     spapr->icp = xics_system_init(smp_cpus * kvmppc_smt_threads() / smp_threads,
784                                   XICS_IRQS);
785     spapr->next_irq = XICS_IRQ_BASE;
786 
787     /* init CPUs */
788     if (cpu_model == NULL) {
789         cpu_model = kvm_enabled() ? "host" : "POWER7";
790     }
791     for (i = 0; i < smp_cpus; i++) {
792         cpu = cpu_ppc_init(cpu_model);
793         if (cpu == NULL) {
794             fprintf(stderr, "Unable to find PowerPC CPU definition\n");
795             exit(1);
796         }
797         env = &cpu->env;
798 
799         xics_cpu_setup(spapr->icp, cpu);
800 
801         /* Set time-base frequency to 512 MHz */
802         cpu_ppc_tb_init(env, TIMEBASE_FREQ);
803 
804         /* PAPR always has exception vectors in RAM not ROM */
805         env->hreset_excp_prefix = 0;
806 
807         /* Tell KVM that we're in PAPR mode */
808         if (kvm_enabled()) {
809             kvmppc_set_papr(cpu);
810         }
811 
812         qemu_register_reset(spapr_cpu_reset, cpu);
813     }
814 
815     /* allocate RAM */
816     spapr->ram_limit = ram_size;
817     if (spapr->ram_limit > rma_alloc_size) {
818         ram_addr_t nonrma_base = rma_alloc_size;
819         ram_addr_t nonrma_size = spapr->ram_limit - rma_alloc_size;
820 
821         memory_region_init_ram(ram, "ppc_spapr.ram", nonrma_size);
822         vmstate_register_ram_global(ram);
823         memory_region_add_subregion(sysmem, nonrma_base, ram);
824     }
825 
826     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
827     spapr->rtas_size = load_image_targphys(filename, spapr->rtas_addr,
828                                            rtas_limit - spapr->rtas_addr);
829     if (spapr->rtas_size < 0) {
830         hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
831         exit(1);
832     }
833     if (spapr->rtas_size > RTAS_MAX_SIZE) {
834         hw_error("RTAS too big ! 0x%lx bytes (max is 0x%x)\n",
835                  spapr->rtas_size, RTAS_MAX_SIZE);
836         exit(1);
837     }
838     g_free(filename);
839 
840     /* Set up EPOW events infrastructure */
841     spapr_events_init(spapr);
842 
843     /* Set up IOMMU */
844     spapr_iommu_init();
845 
846     /* Set up VIO bus */
847     spapr->vio_bus = spapr_vio_bus_init();
848 
849     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
850         if (serial_hds[i]) {
851             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
852         }
853     }
854 
855     /* We always have at least the nvram device on VIO */
856     spapr_create_nvram(spapr);
857 
858     /* Set up PCI */
859     spapr_pci_rtas_init();
860 
861     phb = spapr_create_phb(spapr, 0);
862 
863     for (i = 0; i < nb_nics; i++) {
864         NICInfo *nd = &nd_table[i];
865 
866         if (!nd->model) {
867             nd->model = g_strdup("ibmveth");
868         }
869 
870         if (strcmp(nd->model, "ibmveth") == 0) {
871             spapr_vlan_create(spapr->vio_bus, nd);
872         } else {
873             pci_nic_init_nofail(&nd_table[i], nd->model, NULL);
874         }
875     }
876 
877     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
878         spapr_vscsi_create(spapr->vio_bus);
879     }
880 
881     /* Graphics */
882     if (spapr_vga_init(phb->bus)) {
883         spapr->has_graphics = true;
884     }
885 
886     if (usb_enabled(spapr->has_graphics)) {
887         pci_create_simple(phb->bus, -1, "pci-ohci");
888         if (spapr->has_graphics) {
889             usbdevice_create("keyboard");
890             usbdevice_create("mouse");
891         }
892     }
893 
894     if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
895         fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
896                 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
897         exit(1);
898     }
899 
900     if (kernel_filename) {
901         uint64_t lowaddr = 0;
902 
903         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
904                                NULL, &lowaddr, NULL, 1, ELF_MACHINE, 0);
905         if (kernel_size < 0) {
906             kernel_size = load_image_targphys(kernel_filename,
907                                               KERNEL_LOAD_ADDR,
908                                               load_limit - KERNEL_LOAD_ADDR);
909         }
910         if (kernel_size < 0) {
911             fprintf(stderr, "qemu: could not load kernel '%s'\n",
912                     kernel_filename);
913             exit(1);
914         }
915 
916         /* load initrd */
917         if (initrd_filename) {
918             /* Try to locate the initrd in the gap between the kernel
919              * and the firmware. Add a bit of space just in case
920              */
921             initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
922             initrd_size = load_image_targphys(initrd_filename, initrd_base,
923                                               load_limit - initrd_base);
924             if (initrd_size < 0) {
925                 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
926                         initrd_filename);
927                 exit(1);
928             }
929         } else {
930             initrd_base = 0;
931             initrd_size = 0;
932         }
933     }
934 
935     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, FW_FILE_NAME);
936     fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
937     if (fw_size < 0) {
938         hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
939         exit(1);
940     }
941     g_free(filename);
942 
943     spapr->entry_point = 0x100;
944 
945     /* Prepare the device tree */
946     spapr->fdt_skel = spapr_create_fdt_skel(cpu_model,
947                                             initrd_base, initrd_size,
948                                             kernel_size,
949                                             boot_device, kernel_cmdline,
950                                             spapr->epow_irq);
951     assert(spapr->fdt_skel != NULL);
952 }
953 
954 static QEMUMachine spapr_machine = {
955     .name = "pseries",
956     .desc = "pSeries Logical Partition (PAPR compliant)",
957     .init = ppc_spapr_init,
958     .reset = ppc_spapr_reset,
959     .block_default_type = IF_SCSI,
960     .max_cpus = MAX_CPUS,
961     .no_parallel = 1,
962     .boot_order = NULL,
963 };
964 
965 static void spapr_machine_init(void)
966 {
967     qemu_register_machine(&spapr_machine);
968 }
969 
970 machine_init(spapr_machine_init);
971